92 files changed, 73082 insertions, 18068 deletions

diff --git a/fs/btrfs/Kconfig b/fs/btrfs/Kconfig
index 7bb3c020e57..a66768ebc8d 100644
--- a/fs/btrfs/Kconfig
+++ b/fs/btrfs/Kconfig
@@ -1,16 +1,26 @@
 config BTRFS_FS
-	tristate "Btrfs filesystem (EXPERIMENTAL) Unstable disk format"
-	depends on EXPERIMENTAL
-	select LIBCRC32C
+	tristate "Btrfs filesystem support"
+	select CRYPTO
+	select CRYPTO_CRC32C
 	select ZLIB_INFLATE
 	select ZLIB_DEFLATE
+	select LZO_COMPRESS
+	select LZO_DECOMPRESS
+	select RAID6_PQ
+	select XOR_BLOCKS
+
 	help
-	  Btrfs is a new filesystem with extents, writable snapshotting,
-	  support for multiple devices and many more features.
+	  Btrfs is a general purpose copy-on-write filesystem with extents,
+	  writable snapshotting, support for multiple devices and many more
+	  features focused on fault tolerance, repair and easy administration.
+
+	  The filesystem disk format is no longer unstable, and it's not
+	  expected to change unless there are strong reasons to do so. If there
+	  is a format change, file systems with a unchanged format will
+	  continue to be mountable and usable by newer kernels.
 
-	  Btrfs is highly experimental, and THE DISK FORMAT IS NOT YET
-	  FINALIZED.  You should say N here unless you are interested in
-	  testing Btrfs with non-critical data.
+	  For more information, please see the web pages at
+	  http://btrfs.wiki.kernel.org.
 
 	  To compile this file system support as a module, choose M here. The
 	  module will be called btrfs.
@@ -29,3 +39,52 @@ config BTRFS_FS_POSIX_ACL
 	  Linux website <http://acl.bestbits.at/>.
 
 	  If you don't know what Access Control Lists are, say N
+
+config BTRFS_FS_CHECK_INTEGRITY
+	bool "Btrfs with integrity check tool compiled in (DANGEROUS)"
+	depends on BTRFS_FS
+	help
+	  Adds code that examines all block write requests (including
+	  writes of the super block). The goal is to verify that the
+	  state of the filesystem on disk is always consistent, i.e.,
+	  after a power-loss or kernel panic event the filesystem is
+	  in a consistent state.
+
+	  If the integrity check tool is included and activated in
+	  the mount options, plenty of kernel memory is used, and
+	  plenty of additional CPU cycles are spent. Enabling this
+	  functionality is not intended for normal use.
+
+	  In most cases, unless you are a btrfs developer who needs
+	  to verify the integrity of (super)-block write requests
+	  during the run of a regression test, say N
+
+config BTRFS_FS_RUN_SANITY_TESTS
+	bool "Btrfs will run sanity tests upon loading"
+	depends on BTRFS_FS
+	help
+	  This will run some basic sanity tests on the free space cache
+	  code to make sure it is acting as it should.  These are mostly
+	  regression tests and are only really interesting to btrfs
+	  developers.
+
+	  If unsure, say N.
+
+config BTRFS_DEBUG
+	bool "Btrfs debugging support"
+	depends on BTRFS_FS
+	help
+	  Enable run-time debugging support for the btrfs filesystem. This may
+	  enable additional and expensive checks with negative impact on
+	  performance, or export extra information via sysfs.
+
+	  If unsure, say N.
+
+config BTRFS_ASSERT
+	bool "Btrfs assert support"
+	depends on BTRFS_FS
+	help
+	  Enable run-time assertion checking.  This will result in panics if
+	  any of the assertions trip.  This is meant for btrfs developers only.
+
+	  If unsure, say N.
diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile
index a35eb36b32f..6d1d0b93b1a 100644
--- a/fs/btrfs/Makefile
+++ b/fs/btrfs/Makefile
@@ -6,5 +6,14 @@ btrfs-y += super.o ctree.o extent-tree.o print-tree.o root-tree.o dir-item.o \
 	   transaction.o inode.o file.o tree-defrag.o \
 	   extent_map.o sysfs.o struct-funcs.o xattr.o ordered-data.o \
 	   extent_io.o volumes.o async-thread.o ioctl.o locking.o orphan.o \
-	   export.o tree-log.o acl.o free-space-cache.o zlib.o \
-	   compression.o delayed-ref.o relocation.o
+	   export.o tree-log.o free-space-cache.o zlib.o lzo.o \
+	   compression.o delayed-ref.o relocation.o delayed-inode.o scrub.o \
+	   reada.o backref.o ulist.o qgroup.o send.o dev-replace.o raid56.o \
+	   uuid-tree.o props.o hash.o
+
+btrfs-$(CONFIG_BTRFS_FS_POSIX_ACL) += acl.o
+btrfs-$(CONFIG_BTRFS_FS_CHECK_INTEGRITY) += check-integrity.o
+
+btrfs-$(CONFIG_BTRFS_FS_RUN_SANITY_TESTS) += tests/free-space-tests.o \
+	tests/extent-buffer-tests.o tests/btrfs-tests.o \
+	tests/extent-io-tests.o tests/inode-tests.o tests/qgroup-tests.o
diff --git a/fs/btrfs/acl.c b/fs/btrfs/acl.c
index 6ef7b26724e..9a0124a9585 100644
--- a/fs/btrfs/acl.c
+++ b/fs/btrfs/acl.c
@@ -28,19 +28,13 @@
 #include "btrfs_inode.h"
 #include "xattr.h"
 
-#ifdef CONFIG_BTRFS_FS_POSIX_ACL
-
-static struct posix_acl *btrfs_get_acl(struct inode *inode, int type)
+struct posix_acl *btrfs_get_acl(struct inode *inode, int type)
 {
 	int size;
 	const char *name;
 	char *value = NULL;
 	struct posix_acl *acl;
 
-	acl = get_cached_acl(inode, type);
-	if (acl != ACL_NOT_CACHED)
-		return acl;
-
 	switch (type) {
 	case ACL_TYPE_ACCESS:
 		name = POSIX_ACL_XATTR_ACCESS;
@@ -58,67 +52,42 @@ static struct posix_acl *btrfs_get_acl(struct inode *inode, int type)
 		if (!value)
 			return ERR_PTR(-ENOMEM);
 		size = __btrfs_getxattr(inode, name, value, size);
-		if (size > 0) {
-			acl = posix_acl_from_xattr(value, size);
-			set_cached_acl(inode, type, acl);
-		}
-		kfree(value);
+	}
+	if (size > 0) {
+		acl = posix_acl_from_xattr(&init_user_ns, value, size);
 	} else if (size == -ENOENT || size == -ENODATA || size == 0) {
 		/* FIXME, who returns -ENOENT?  I think nobody */
 		acl = NULL;
-		set_cached_acl(inode, type, acl);
 	} else {
 		acl = ERR_PTR(-EIO);
 	}
+	kfree(value);
 
-	return acl;
-}
-
-static int btrfs_xattr_acl_get(struct dentry *dentry, const char *name,
-		void *value, size_t size, int type)
-{
-	struct posix_acl *acl;
-	int ret = 0;
-
-	acl = btrfs_get_acl(dentry->d_inode, type);
-
-	if (IS_ERR(acl))
-		return PTR_ERR(acl);
-	if (acl == NULL)
-		return -ENODATA;
-	ret = posix_acl_to_xattr(acl, value, size);
-	posix_acl_release(acl);
+	if (!IS_ERR(acl))
+		set_cached_acl(inode, type, acl);
 
-	return ret;
+	return acl;
 }
 
 /*
  * Needs to be called with fs_mutex held
  */
-static int btrfs_set_acl(struct btrfs_trans_handle *trans,
+static int __btrfs_set_acl(struct btrfs_trans_handle *trans,
 			 struct inode *inode, struct posix_acl *acl, int type)
 {
 	int ret, size = 0;
 	const char *name;
 	char *value = NULL;
-	mode_t mode;
-
-	if (acl) {
-		ret = posix_acl_valid(acl);
-		if (ret < 0)
-			return ret;
-		ret = 0;
-	}
 
 	switch (type) {
 	case ACL_TYPE_ACCESS:
-		mode = inode->i_mode;
 		name = POSIX_ACL_XATTR_ACCESS;
 		if (acl) {
-			ret = posix_acl_equiv_mode(acl, &mode);
+			ret = posix_acl_equiv_mode(acl, &inode->i_mode);
 			if (ret < 0)
 				return ret;
-			inode->i_mode = mode;
+			if (ret == 0)
+				acl = NULL;
 		}
 		ret = 0;
 		break;
@@ -139,7 +108,7 @@ static int btrfs_set_acl(struct btrfs_trans_handle *trans,
 			goto out;
 		}
 
-		ret = posix_acl_to_xattr(acl, value, size);
+		ret = posix_acl_to_xattr(&init_user_ns, acl, value, size);
 		if (ret < 0)
 			goto out;
 	}
@@ -154,44 +123,9 @@ out:
 	return ret;
 }
 
-static int btrfs_xattr_acl_set(struct dentry *dentry, const char *name,
-		const void *value, size_t size, int flags, int type)
+int btrfs_set_acl(struct inode *inode, struct posix_acl *acl, int type)
 {
-	int ret;
-	struct posix_acl *acl = NULL;
-
-	if (value) {
-		acl = posix_acl_from_xattr(value, size);
-		if (acl == NULL) {
-			value = NULL;
-			size = 0;
-		} else if (IS_ERR(acl)) {
-			return PTR_ERR(acl);
-		}
-	}
-
-	ret = btrfs_set_acl(NULL, dentry->d_inode, acl, type);
-
-	posix_acl_release(acl);
-
-	return ret;
-}
-
-int btrfs_check_acl(struct inode *inode, int mask)
-{
-	struct posix_acl *acl;
-	int error = -EAGAIN;
-
-	acl = btrfs_get_acl(inode, ACL_TYPE_ACCESS);
-
-	if (IS_ERR(acl))
-		return PTR_ERR(acl);
-	if (acl) {
-		error = posix_acl_permission(inode, acl, mask);
-		posix_acl_release(acl);
-	}
-
-	return error;
+	return __btrfs_set_acl(NULL, inode, acl, type);
 }
 
 /*
@@ -202,111 +136,31 @@ int btrfs_check_acl(struct inode *inode, int mask)
 int btrfs_init_acl(struct btrfs_trans_handle *trans,
 		   struct inode *inode, struct inode *dir)
 {
-	struct posix_acl *acl = NULL;
+	struct posix_acl *default_acl, *acl;
 	int ret = 0;
 
 	/* this happens with subvols */
 	if (!dir)
 		return 0;
 
-	if (!S_ISLNK(inode->i_mode)) {
-		if (IS_POSIXACL(dir)) {
-			acl = btrfs_get_acl(dir, ACL_TYPE_DEFAULT);
-			if (IS_ERR(acl))
-				return PTR_ERR(acl);
-		}
+	ret = posix_acl_create(dir, &inode->i_mode, &default_acl, &acl);
+	if (ret)
+		return ret;
 
-		if (!acl)
-			inode->i_mode &= ~current_umask();
+	if (default_acl) {
+		ret = __btrfs_set_acl(trans, inode, default_acl,
+				      ACL_TYPE_DEFAULT);
+		posix_acl_release(default_acl);
 	}
 
-	if (IS_POSIXACL(dir) && acl) {
-		struct posix_acl *clone;
-		mode_t mode;
-
-		if (S_ISDIR(inode->i_mode)) {
-			ret = btrfs_set_acl(trans, inode, acl,
-					    ACL_TYPE_DEFAULT);
-			if (ret)
-				goto failed;
-		}
-		clone = posix_acl_clone(acl, GFP_NOFS);
-		ret = -ENOMEM;
-		if (!clone)
-			goto failed;
-
-		mode = inode->i_mode;
-		ret = posix_acl_create_masq(clone, &mode);
-		if (ret >= 0) {
-			inode->i_mode = mode;
-			if (ret > 0) {
-				/* we need an acl */
-				ret = btrfs_set_acl(trans, inode, clone,
-						    ACL_TYPE_ACCESS);
-			}
-		}
-		posix_acl_release(clone);
+	if (acl) {
+		if (!ret)
+			ret = __btrfs_set_acl(trans, inode, acl,
+					      ACL_TYPE_ACCESS);
+		posix_acl_release(acl);
 	}
-failed:
-	posix_acl_release(acl);
-
-	return ret;
-}
-
-int btrfs_acl_chmod(struct inode *inode)
-{
-	struct posix_acl *acl, *clone;
-	int ret = 0;
-
-	if (S_ISLNK(inode->i_mode))
-		return -EOPNOTSUPP;
-
-	if (!IS_POSIXACL(inode))
-		return 0;
-
-	acl = btrfs_get_acl(inode, ACL_TYPE_ACCESS);
-	if (IS_ERR(acl) || !acl)
-		return PTR_ERR(acl);
-
-	clone = posix_acl_clone(acl, GFP_KERNEL);
-	posix_acl_release(acl);
-	if (!clone)
-		return -ENOMEM;
-
-	ret = posix_acl_chmod_masq(clone, inode->i_mode);
-	if (!ret)
-		ret = btrfs_set_acl(NULL, inode, clone, ACL_TYPE_ACCESS);
-
-	posix_acl_release(clone);
 
+	if (!default_acl && !acl)
+		cache_no_acl(inode);
 	return ret;
 }
-
-struct xattr_handler btrfs_xattr_acl_default_handler = {
-	.prefix = POSIX_ACL_XATTR_DEFAULT,
-	.flags	= ACL_TYPE_DEFAULT,
-	.get	= btrfs_xattr_acl_get,
-	.set	= btrfs_xattr_acl_set,
-};
-
-struct xattr_handler btrfs_xattr_acl_access_handler = {
-	.prefix = POSIX_ACL_XATTR_ACCESS,
-	.flags	= ACL_TYPE_ACCESS,
-	.get	= btrfs_xattr_acl_get,
-	.set	= btrfs_xattr_acl_set,
-};
-
-#else /* CONFIG_BTRFS_FS_POSIX_ACL */
-
-int btrfs_acl_chmod(struct inode *inode)
-{
-	return 0;
-}
-
-int btrfs_init_acl(struct btrfs_trans_handle *trans,
-		   struct inode *inode, struct inode *dir)
-{
-	return 0;
-}
-
-#endif /* CONFIG_BTRFS_FS_POSIX_ACL */
diff --git a/fs/btrfs/async-thread.c b/fs/btrfs/async-thread.c
index 462859a3014..5a201d81049 100644
--- a/fs/btrfs/async-thread.c
+++ b/fs/btrfs/async-thread.c
@@ -1,5 +1,6 @@
 /*
  * Copyright (C) 2007 Oracle.  All rights reserved.
+ * Copyright (C) 2014 Fujitsu.  All rights reserved.
  *
  * This program is free software; you can redistribute it and/or
  * modify it under the terms of the GNU General Public
@@ -21,697 +22,315 @@
 #include <linux/list.h>
 #include <linux/spinlock.h>
 #include <linux/freezer.h>
+#include <linux/workqueue.h>
 #include "async-thread.h"
+#include "ctree.h"
+
+#define WORK_DONE_BIT 0
+#define WORK_ORDER_DONE_BIT 1
+#define WORK_HIGH_PRIO_BIT 2
+
+#define NO_THRESHOLD (-1)
+#define DFT_THRESHOLD (32)
+
+struct __btrfs_workqueue {
+	struct workqueue_struct *normal_wq;
+	/* List head pointing to ordered work list */
+	struct list_head ordered_list;
+
+	/* Spinlock for ordered_list */
+	spinlock_t list_lock;
+
+	/* Thresholding related variants */
+	atomic_t pending;
+	int max_active;
+	int current_max;
+	int thresh;
+	unsigned int count;
+	spinlock_t thres_lock;
+};
 
-#define WORK_QUEUED_BIT 0
-#define WORK_DONE_BIT 1
-#define WORK_ORDER_DONE_BIT 2
-#define WORK_HIGH_PRIO_BIT 3
-
-/*
- * container for the kthread task pointer and the list of pending work
- * One of these is allocated per thread.
- */
-struct btrfs_worker_thread {
-	/* pool we belong to */
-	struct btrfs_workers *workers;
-
-	/* list of struct btrfs_work that are waiting for service */
-	struct list_head pending;
-	struct list_head prio_pending;
-
-	/* list of worker threads from struct btrfs_workers */
-	struct list_head worker_list;
+struct btrfs_workqueue {
+	struct __btrfs_workqueue *normal;
+	struct __btrfs_workqueue *high;
+};
 
-	/* kthread */
-	struct task_struct *task;
+static inline struct __btrfs_workqueue
+*__btrfs_alloc_workqueue(const char *name, int flags, int max_active,
+			 int thresh)
+{
+	struct __btrfs_workqueue *ret = kzalloc(sizeof(*ret), GFP_NOFS);
 
-	/* number of things on the pending list */
-	atomic_t num_pending;
+	if (unlikely(!ret))
+		return NULL;
 
-	/* reference counter for this struct */
-	atomic_t refs;
+	ret->max_active = max_active;
+	atomic_set(&ret->pending, 0);
+	if (thresh == 0)
+		thresh = DFT_THRESHOLD;
+	/* For low threshold, disabling threshold is a better choice */
+	if (thresh < DFT_THRESHOLD) {
+		ret->current_max = max_active;
+		ret->thresh = NO_THRESHOLD;
+	} else {
+		ret->current_max = 1;
+		ret->thresh = thresh;
+	}
 
-	unsigned long sequence;
+	if (flags & WQ_HIGHPRI)
+		ret->normal_wq = alloc_workqueue("%s-%s-high", flags,
+						 ret->max_active,
+						 "btrfs", name);
+	else
+		ret->normal_wq = alloc_workqueue("%s-%s", flags,
+						 ret->max_active, "btrfs",
+						 name);
+	if (unlikely(!ret->normal_wq)) {
+		kfree(ret);
+		return NULL;
+	}
 
-	/* protects the pending list. */
-	spinlock_t lock;
+	INIT_LIST_HEAD(&ret->ordered_list);
+	spin_lock_init(&ret->list_lock);
+	spin_lock_init(&ret->thres_lock);
+	trace_btrfs_workqueue_alloc(ret, name, flags & WQ_HIGHPRI);
+	return ret;
+}
 
-	/* set to non-zero when this thread is already awake and kicking */
-	int working;
+static inline void
+__btrfs_destroy_workqueue(struct __btrfs_workqueue *wq);
 
-	/* are we currently idle */
-	int idle;
-};
+struct btrfs_workqueue *btrfs_alloc_workqueue(const char *name,
+					      int flags,
+					      int max_active,
+					      int thresh)
+{
+	struct btrfs_workqueue *ret = kzalloc(sizeof(*ret), GFP_NOFS);
 
-/*
- * btrfs_start_workers uses kthread_run, which can block waiting for memory
- * for a very long time.  It will actually throttle on page writeback,
- * and so it may not make progress until after our btrfs worker threads
- * process all of the pending work structs in their queue
- *
- * This means we can't use btrfs_start_workers from inside a btrfs worker
- * thread that is used as part of cleaning dirty memory, which pretty much
- * involves all of the worker threads.
- *
- * Instead we have a helper queue who never has more than one thread
- * where we scheduler thread start operations.  This worker_start struct
- * is used to contain the work and hold a pointer to the queue that needs
- * another worker.
- */
-struct worker_start {
-	struct btrfs_work work;
-	struct btrfs_workers *queue;
-};
+	if (unlikely(!ret))
+		return NULL;
 
-static void start_new_worker_func(struct btrfs_work *work)
-{
-	struct worker_start *start;
-	start = container_of(work, struct worker_start, work);
-	btrfs_start_workers(start->queue, 1);
-	kfree(start);
-}
+	ret->normal = __btrfs_alloc_workqueue(name, flags & ~WQ_HIGHPRI,
+					      max_active, thresh);
+	if (unlikely(!ret->normal)) {
+		kfree(ret);
+		return NULL;
+	}
 
-static int start_new_worker(struct btrfs_workers *queue)
-{
-	struct worker_start *start;
-	int ret;
-
-	start = kzalloc(sizeof(*start), GFP_NOFS);
-	if (!start)
-		return -ENOMEM;
-
-	start->work.func = start_new_worker_func;
-	start->queue = queue;
-	ret = btrfs_queue_worker(queue->atomic_worker_start, &start->work);
-	if (ret)
-		kfree(start);
+	if (flags & WQ_HIGHPRI) {
+		ret->high = __btrfs_alloc_workqueue(name, flags, max_active,
+						    thresh);
+		if (unlikely(!ret->high)) {
+			__btrfs_destroy_workqueue(ret->normal);
+			kfree(ret);
+			return NULL;
+		}
+	}
 	return ret;
 }
 
 /*
- * helper function to move a thread onto the idle list after it
- * has finished some requests.
+ * Hook for threshold which will be called in btrfs_queue_work.
+ * This hook WILL be called in IRQ handler context,
+ * so workqueue_set_max_active MUST NOT be called in this hook
  */
-static void check_idle_worker(struct btrfs_worker_thread *worker)
+static inline void thresh_queue_hook(struct __btrfs_workqueue *wq)
 {
-	if (!worker->idle && atomic_read(&worker->num_pending) <
-	    worker->workers->idle_thresh / 2) {
-		unsigned long flags;
-		spin_lock_irqsave(&worker->workers->lock, flags);
-		worker->idle = 1;
-
-		/* the list may be empty if the worker is just starting */
-		if (!list_empty(&worker->worker_list)) {
-			list_move(&worker->worker_list,
-				 &worker->workers->idle_list);
-		}
-		spin_unlock_irqrestore(&worker->workers->lock, flags);
-	}
+	if (wq->thresh == NO_THRESHOLD)
+		return;
+	atomic_inc(&wq->pending);
 }
 
 /*
- * helper function to move a thread off the idle list after new
- * pending work is added.
+ * Hook for threshold which will be called before executing the work,
+ * This hook is called in kthread content.
+ * So workqueue_set_max_active is called here.
  */
-static void check_busy_worker(struct btrfs_worker_thread *worker)
-{
-	if (worker->idle && atomic_read(&worker->num_pending) >=
-	    worker->workers->idle_thresh) {
-		unsigned long flags;
-		spin_lock_irqsave(&worker->workers->lock, flags);
-		worker->idle = 0;
-
-		if (!list_empty(&worker->worker_list)) {
-			list_move_tail(&worker->worker_list,
-				      &worker->workers->worker_list);
-		}
-		spin_unlock_irqrestore(&worker->workers->lock, flags);
-	}
-}
-
-static void check_pending_worker_creates(struct btrfs_worker_thread *worker)
+static inline void thresh_exec_hook(struct __btrfs_workqueue *wq)
 {
-	struct btrfs_workers *workers = worker->workers;
-	unsigned long flags;
+	int new_max_active;
+	long pending;
+	int need_change = 0;
 
-	rmb();
-	if (!workers->atomic_start_pending)
+	if (wq->thresh == NO_THRESHOLD)
 		return;
 
-	spin_lock_irqsave(&workers->lock, flags);
-	if (!workers->atomic_start_pending)
-		goto out;
-
-	workers->atomic_start_pending = 0;
-	if (workers->num_workers + workers->num_workers_starting >=
-	    workers->max_workers)
-		goto out;
-
-	workers->num_workers_starting += 1;
-	spin_unlock_irqrestore(&workers->lock, flags);
-	start_new_worker(workers);
-	return;
+	atomic_dec(&wq->pending);
+	spin_lock(&wq->thres_lock);
+	/*
+	 * Use wq->count to limit the calling frequency of
+	 * workqueue_set_max_active.
+	 */
+	wq->count++;
+	wq->count %= (wq->thresh / 4);
+	if (!wq->count)
+		goto  out;
+	new_max_active = wq->current_max;
 
+	/*
+	 * pending may be changed later, but it's OK since we really
+	 * don't need it so accurate to calculate new_max_active.
+	 */
+	pending = atomic_read(&wq->pending);
+	if (pending > wq->thresh)
+		new_max_active++;
+	if (pending < wq->thresh / 2)
+		new_max_active--;
+	new_max_active = clamp_val(new_max_active, 1, wq->max_active);
+	if (new_max_active != wq->current_max)  {
+		need_change = 1;
+		wq->current_max = new_max_active;
+	}
 out:
-	spin_unlock_irqrestore(&workers->lock, flags);
+	spin_unlock(&wq->thres_lock);
+
+	if (need_change) {
+		workqueue_set_max_active(wq->normal_wq, wq->current_max);
+	}
 }
 
-static noinline int run_ordered_completions(struct btrfs_workers *workers,
-					    struct btrfs_work *work)
+static void run_ordered_work(struct __btrfs_workqueue *wq)
 {
-	if (!workers->ordered)
-		return 0;
-
-	set_bit(WORK_DONE_BIT, &work->flags);
-
-	spin_lock(&workers->order_lock);
+	struct list_head *list = &wq->ordered_list;
+	struct btrfs_work *work;
+	spinlock_t *lock = &wq->list_lock;
+	unsigned long flags;
 
 	while (1) {
-		if (!list_empty(&workers->prio_order_list)) {
-			work = list_entry(workers->prio_order_list.next,
-					  struct btrfs_work, order_list);
-		} else if (!list_empty(&workers->order_list)) {
-			work = list_entry(workers->order_list.next,
-					  struct btrfs_work, order_list);
-		} else {
+		spin_lock_irqsave(lock, flags);
+		if (list_empty(list))
 			break;
-		}
+		work = list_entry(list->next, struct btrfs_work,
+				  ordered_list);
 		if (!test_bit(WORK_DONE_BIT, &work->flags))
 			break;
 
-		/* we are going to call the ordered done function, but
+		/*
+		 * we are going to call the ordered done function, but
 		 * we leave the work item on the list as a barrier so
 		 * that later work items that are done don't have their
 		 * functions called before this one returns
 		 */
 		if (test_and_set_bit(WORK_ORDER_DONE_BIT, &work->flags))
 			break;
-
-		spin_unlock(&workers->order_lock);
-
+		trace_btrfs_ordered_sched(work);
+		spin_unlock_irqrestore(lock, flags);
 		work->ordered_func(work);
 
-		/* now take the lock again and call the freeing code */
-		spin_lock(&workers->order_lock);
-		list_del(&work->order_list);
-		work->ordered_free(work);
-	}
+		/* now take the lock again and drop our item from the list */
+		spin_lock_irqsave(lock, flags);
+		list_del(&work->ordered_list);
+		spin_unlock_irqrestore(lock, flags);
 
-	spin_unlock(&workers->order_lock);
-	return 0;
-}
-
-static void put_worker(struct btrfs_worker_thread *worker)
-{
-	if (atomic_dec_and_test(&worker->refs))
-		kfree(worker);
-}
-
-static int try_worker_shutdown(struct btrfs_worker_thread *worker)
-{
-	int freeit = 0;
-
-	spin_lock_irq(&worker->lock);
-	spin_lock(&worker->workers->lock);
-	if (worker->workers->num_workers > 1 &&
-	    worker->idle &&
-	    !worker->working &&
-	    !list_empty(&worker->worker_list) &&
-	    list_empty(&worker->prio_pending) &&
-	    list_empty(&worker->pending) &&
-	    atomic_read(&worker->num_pending) == 0) {
-		freeit = 1;
-		list_del_init(&worker->worker_list);
-		worker->workers->num_workers--;
+		/*
+		 * we don't want to call the ordered free functions
+		 * with the lock held though
+		 */
+		work->ordered_free(work);
+		trace_btrfs_all_work_done(work);
 	}
-	spin_unlock(&worker->workers->lock);
-	spin_unlock_irq(&worker->lock);
-
-	if (freeit)
-		put_worker(worker);
-	return freeit;
+	spin_unlock_irqrestore(lock, flags);
 }
 
-static struct btrfs_work *get_next_work(struct btrfs_worker_thread *worker,
-					struct list_head *prio_head,
-					struct list_head *head)
+static void normal_work_helper(struct work_struct *arg)
 {
-	struct btrfs_work *work = NULL;
-	struct list_head *cur = NULL;
-
-	if(!list_empty(prio_head))
-		cur = prio_head->next;
-
-	smp_mb();
-	if (!list_empty(&worker->prio_pending))
-		goto refill;
-
-	if (!list_empty(head))
-		cur = head->next;
-
-	if (cur)
-		goto out;
-
-refill:
-	spin_lock_irq(&worker->lock);
-	list_splice_tail_init(&worker->prio_pending, prio_head);
-	list_splice_tail_init(&worker->pending, head);
-
-	if (!list_empty(prio_head))
-		cur = prio_head->next;
-	else if (!list_empty(head))
-		cur = head->next;
-	spin_unlock_irq(&worker->lock);
-
-	if (!cur)
-		goto out_fail;
-
-out:
-	work = list_entry(cur, struct btrfs_work, list);
-
-out_fail:
-	return work;
-}
-
-/*
- * main loop for servicing work items
- */
-static int worker_loop(void *arg)
-{
-	struct btrfs_worker_thread *worker = arg;
-	struct list_head head;
-	struct list_head prio_head;
 	struct btrfs_work *work;
+	struct __btrfs_workqueue *wq;
+	int need_order = 0;
 
-	INIT_LIST_HEAD(&head);
-	INIT_LIST_HEAD(&prio_head);
-
-	do {
-again:
-		while (1) {
-
-
-			work = get_next_work(worker, &prio_head, &head);
-			if (!work)
-				break;
-
-			list_del(&work->list);
-			clear_bit(WORK_QUEUED_BIT, &work->flags);
-
-			work->worker = worker;
-
-			work->func(work);
-
-			atomic_dec(&worker->num_pending);
-			/*
-			 * unless this is an ordered work queue,
-			 * 'work' was probably freed by func above.
-			 */
-			run_ordered_completions(worker->workers, work);
-
-			check_pending_worker_creates(worker);
-
-		}
-
-		spin_lock_irq(&worker->lock);
-		check_idle_worker(worker);
-
-		if (freezing(current)) {
-			worker->working = 0;
-			spin_unlock_irq(&worker->lock);
-			refrigerator();
-		} else {
-			spin_unlock_irq(&worker->lock);
-			if (!kthread_should_stop()) {
-				cpu_relax();
-				/*
-				 * we've dropped the lock, did someone else
-				 * jump_in?
-				 */
-				smp_mb();
-				if (!list_empty(&worker->pending) ||
-				    !list_empty(&worker->prio_pending))
-					continue;
-
-				/*
-				 * this short schedule allows more work to
-				 * come in without the queue functions
-				 * needing to go through wake_up_process()
-				 *
-				 * worker->working is still 1, so nobody
-				 * is going to try and wake us up
-				 */
-				schedule_timeout(1);
-				smp_mb();
-				if (!list_empty(&worker->pending) ||
-				    !list_empty(&worker->prio_pending))
-					continue;
-
-				if (kthread_should_stop())
-					break;
-
-				/* still no more work?, sleep for real */
-				spin_lock_irq(&worker->lock);
-				set_current_state(TASK_INTERRUPTIBLE);
-				if (!list_empty(&worker->pending) ||
-				    !list_empty(&worker->prio_pending)) {
-					spin_unlock_irq(&worker->lock);
-					goto again;
-				}
-
-				/*
-				 * this makes sure we get a wakeup when someone
-				 * adds something new to the queue
-				 */
-				worker->working = 0;
-				spin_unlock_irq(&worker->lock);
-
-				if (!kthread_should_stop()) {
-					schedule_timeout(HZ * 120);
-					if (!worker->working &&
-					    try_worker_shutdown(worker)) {
-						return 0;
-					}
-				}
-			}
-			__set_current_state(TASK_RUNNING);
-		}
-	} while (!kthread_should_stop());
-	return 0;
-}
-
-/*
- * this will wait for all the worker threads to shutdown
- */
-int btrfs_stop_workers(struct btrfs_workers *workers)
-{
-	struct list_head *cur;
-	struct btrfs_worker_thread *worker;
-	int can_stop;
-
-	spin_lock_irq(&workers->lock);
-	list_splice_init(&workers->idle_list, &workers->worker_list);
-	while (!list_empty(&workers->worker_list)) {
-		cur = workers->worker_list.next;
-		worker = list_entry(cur, struct btrfs_worker_thread,
-				    worker_list);
-
-		atomic_inc(&worker->refs);
-		workers->num_workers -= 1;
-		if (!list_empty(&worker->worker_list)) {
-			list_del_init(&worker->worker_list);
-			put_worker(worker);
-			can_stop = 1;
-		} else
-			can_stop = 0;
-		spin_unlock_irq(&workers->lock);
-		if (can_stop)
-			kthread_stop(worker->task);
-		spin_lock_irq(&workers->lock);
-		put_worker(worker);
+	work = container_of(arg, struct btrfs_work, normal_work);
+	/*
+	 * We should not touch things inside work in the following cases:
+	 * 1) after work->func() if it has no ordered_free
+	 *    Since the struct is freed in work->func().
+	 * 2) after setting WORK_DONE_BIT
+	 *    The work may be freed in other threads almost instantly.
+	 * So we save the needed things here.
+	 */
+	if (work->ordered_func)
+		need_order = 1;
+	wq = work->wq;
+
+	trace_btrfs_work_sched(work);
+	thresh_exec_hook(wq);
+	work->func(work);
+	if (need_order) {
+		set_bit(WORK_DONE_BIT, &work->flags);
+		run_ordered_work(wq);
 	}
-	spin_unlock_irq(&workers->lock);
-	return 0;
+	if (!need_order)
+		trace_btrfs_all_work_done(work);
 }
 
-/*
- * simple init on struct btrfs_workers
- */
-void btrfs_init_workers(struct btrfs_workers *workers, char *name, int max,
-			struct btrfs_workers *async_helper)
+void btrfs_init_work(struct btrfs_work *work,
+		     btrfs_func_t func,
+		     btrfs_func_t ordered_func,
+		     btrfs_func_t ordered_free)
 {
-	workers->num_workers = 0;
-	workers->num_workers_starting = 0;
-	INIT_LIST_HEAD(&workers->worker_list);
-	INIT_LIST_HEAD(&workers->idle_list);
-	INIT_LIST_HEAD(&workers->order_list);
-	INIT_LIST_HEAD(&workers->prio_order_list);
-	spin_lock_init(&workers->lock);
-	spin_lock_init(&workers->order_lock);
-	workers->max_workers = max;
-	workers->idle_thresh = 32;
-	workers->name = name;
-	workers->ordered = 0;
-	workers->atomic_start_pending = 0;
-	workers->atomic_worker_start = async_helper;
+	work->func = func;
+	work->ordered_func = ordered_func;
+	work->ordered_free = ordered_free;
+	INIT_WORK(&work->normal_work, normal_work_helper);
+	INIT_LIST_HEAD(&work->ordered_list);
+	work->flags = 0;
 }
 
-/*
- * starts new worker threads.  This does not enforce the max worker
- * count in case you need to temporarily go past it.
- */
-static int __btrfs_start_workers(struct btrfs_workers *workers,
-				 int num_workers)
+static inline void __btrfs_queue_work(struct __btrfs_workqueue *wq,
+				      struct btrfs_work *work)
 {
-	struct btrfs_worker_thread *worker;
-	int ret = 0;
-	int i;
-
-	for (i = 0; i < num_workers; i++) {
-		worker = kzalloc(sizeof(*worker), GFP_NOFS);
-		if (!worker) {
-			ret = -ENOMEM;
-			goto fail;
-		}
+	unsigned long flags;
 
-		INIT_LIST_HEAD(&worker->pending);
-		INIT_LIST_HEAD(&worker->prio_pending);
-		INIT_LIST_HEAD(&worker->worker_list);
-		spin_lock_init(&worker->lock);
-
-		atomic_set(&worker->num_pending, 0);
-		atomic_set(&worker->refs, 1);
-		worker->workers = workers;
-		worker->task = kthread_run(worker_loop, worker,
-					   "btrfs-%s-%d", workers->name,
-					   workers->num_workers + i);
-		if (IS_ERR(worker->task)) {
-			ret = PTR_ERR(worker->task);
-			kfree(worker);
-			goto fail;
-		}
-		spin_lock_irq(&workers->lock);
-		list_add_tail(&worker->worker_list, &workers->idle_list);
-		worker->idle = 1;
-		workers->num_workers++;
-		workers->num_workers_starting--;
-		WARN_ON(workers->num_workers_starting < 0);
-		spin_unlock_irq(&workers->lock);
+	work->wq = wq;
+	thresh_queue_hook(wq);
+	if (work->ordered_func) {
+		spin_lock_irqsave(&wq->list_lock, flags);
+		list_add_tail(&work->ordered_list, &wq->ordered_list);
+		spin_unlock_irqrestore(&wq->list_lock, flags);
 	}
-	return 0;
-fail:
-	btrfs_stop_workers(workers);
-	return ret;
+	queue_work(wq->normal_wq, &work->normal_work);
+	trace_btrfs_work_queued(work);
 }
 
-int btrfs_start_workers(struct btrfs_workers *workers, int num_workers)
+void btrfs_queue_work(struct btrfs_workqueue *wq,
+		      struct btrfs_work *work)
 {
-	spin_lock_irq(&workers->lock);
-	workers->num_workers_starting += num_workers;
-	spin_unlock_irq(&workers->lock);
-	return __btrfs_start_workers(workers, num_workers);
-}
+	struct __btrfs_workqueue *dest_wq;
 
-/*
- * run through the list and find a worker thread that doesn't have a lot
- * to do right now.  This can return null if we aren't yet at the thread
- * count limit and all of the threads are busy.
- */
-static struct btrfs_worker_thread *next_worker(struct btrfs_workers *workers)
-{
-	struct btrfs_worker_thread *worker;
-	struct list_head *next;
-	int enforce_min;
-
-	enforce_min = (workers->num_workers + workers->num_workers_starting) <
-		workers->max_workers;
-
-	/*
-	 * if we find an idle thread, don't move it to the end of the
-	 * idle list.  This improves the chance that the next submission
-	 * will reuse the same thread, and maybe catch it while it is still
-	 * working
-	 */
-	if (!list_empty(&workers->idle_list)) {
-		next = workers->idle_list.next;
-		worker = list_entry(next, struct btrfs_worker_thread,
-				    worker_list);
-		return worker;
-	}
-	if (enforce_min || list_empty(&workers->worker_list))
-		return NULL;
-
-	/*
-	 * if we pick a busy task, move the task to the end of the list.
-	 * hopefully this will keep things somewhat evenly balanced.
-	 * Do the move in batches based on the sequence number.  This groups
-	 * requests submitted at roughly the same time onto the same worker.
-	 */
-	next = workers->worker_list.next;
-	worker = list_entry(next, struct btrfs_worker_thread, worker_list);
-	worker->sequence++;
-
-	if (worker->sequence % workers->idle_thresh == 0)
-		list_move_tail(next, &workers->worker_list);
-	return worker;
+	if (test_bit(WORK_HIGH_PRIO_BIT, &work->flags) && wq->high)
+		dest_wq = wq->high;
+	else
+		dest_wq = wq->normal;
+	__btrfs_queue_work(dest_wq, work);
 }
 
-/*
- * selects a worker thread to take the next job.  This will either find
- * an idle worker, start a new worker up to the max count, or just return
- * one of the existing busy workers.
- */
-static struct btrfs_worker_thread *find_worker(struct btrfs_workers *workers)
+static inline void
+__btrfs_destroy_workqueue(struct __btrfs_workqueue *wq)
 {
-	struct btrfs_worker_thread *worker;
-	unsigned long flags;
-	struct list_head *fallback;
-
-again:
-	spin_lock_irqsave(&workers->lock, flags);
-	worker = next_worker(workers);
-
-	if (!worker) {
-		if (workers->num_workers + workers->num_workers_starting >=
-		    workers->max_workers) {
-			goto fallback;
-		} else if (workers->atomic_worker_start) {
-			workers->atomic_start_pending = 1;
-			goto fallback;
-		} else {
-			workers->num_workers_starting++;
-			spin_unlock_irqrestore(&workers->lock, flags);
-			/* we're below the limit, start another worker */
-			__btrfs_start_workers(workers, 1);
-			goto again;
-		}
-	}
-	goto found;
-
-fallback:
-	fallback = NULL;
-	/*
-	 * we have failed to find any workers, just
-	 * return the first one we can find.
-	 */
-	if (!list_empty(&workers->worker_list))
-		fallback = workers->worker_list.next;
-	if (!list_empty(&workers->idle_list))
-		fallback = workers->idle_list.next;
-	BUG_ON(!fallback);
-	worker = list_entry(fallback,
-		  struct btrfs_worker_thread, worker_list);
-found:
-	/*
-	 * this makes sure the worker doesn't exit before it is placed
-	 * onto a busy/idle list
-	 */
-	atomic_inc(&worker->num_pending);
-	spin_unlock_irqrestore(&workers->lock, flags);
-	return worker;
+	destroy_workqueue(wq->normal_wq);
+	trace_btrfs_workqueue_destroy(wq);
+	kfree(wq);
 }
 
-/*
- * btrfs_requeue_work just puts the work item back on the tail of the list
- * it was taken from.  It is intended for use with long running work functions
- * that make some progress and want to give the cpu up for others.
- */
-int btrfs_requeue_work(struct btrfs_work *work)
+void btrfs_destroy_workqueue(struct btrfs_workqueue *wq)
 {
-	struct btrfs_worker_thread *worker = work->worker;
-	unsigned long flags;
-	int wake = 0;
-
-	if (test_and_set_bit(WORK_QUEUED_BIT, &work->flags))
-		goto out;
-
-	spin_lock_irqsave(&worker->lock, flags);
-	if (test_bit(WORK_HIGH_PRIO_BIT, &work->flags))
-		list_add_tail(&work->list, &worker->prio_pending);
-	else
-		list_add_tail(&work->list, &worker->pending);
-	atomic_inc(&worker->num_pending);
-
-	/* by definition we're busy, take ourselves off the idle
-	 * list
-	 */
-	if (worker->idle) {
-		spin_lock(&worker->workers->lock);
-		worker->idle = 0;
-		list_move_tail(&worker->worker_list,
-			      &worker->workers->worker_list);
-		spin_unlock(&worker->workers->lock);
-	}
-	if (!worker->working) {
-		wake = 1;
-		worker->working = 1;
-	}
-
-	if (wake)
-		wake_up_process(worker->task);
-	spin_unlock_irqrestore(&worker->lock, flags);
-out:
-
-	return 0;
+	if (!wq)
+		return;
+	if (wq->high)
+		__btrfs_destroy_workqueue(wq->high);
+	__btrfs_destroy_workqueue(wq->normal);
+	kfree(wq);
 }
 
-void btrfs_set_work_high_prio(struct btrfs_work *work)
+void btrfs_workqueue_set_max(struct btrfs_workqueue *wq, int max)
 {
-	set_bit(WORK_HIGH_PRIO_BIT, &work->flags);
+	if (!wq)
+		return;
+	wq->normal->max_active = max;
+	if (wq->high)
+		wq->high->max_active = max;
 }
 
-/*
- * places a struct btrfs_work into the pending queue of one of the kthreads
- */
-int btrfs_queue_worker(struct btrfs_workers *workers, struct btrfs_work *work)
+void btrfs_set_work_high_priority(struct btrfs_work *work)
 {
-	struct btrfs_worker_thread *worker;
-	unsigned long flags;
-	int wake = 0;
-
-	/* don't requeue something already on a list */
-	if (test_and_set_bit(WORK_QUEUED_BIT, &work->flags))
-		goto out;
-
-	worker = find_worker(workers);
-	if (workers->ordered) {
-		/*
-		 * you're not allowed to do ordered queues from an
-		 * interrupt handler
-		 */
-		spin_lock(&workers->order_lock);
-		if (test_bit(WORK_HIGH_PRIO_BIT, &work->flags)) {
-			list_add_tail(&work->order_list,
-				      &workers->prio_order_list);
-		} else {
-			list_add_tail(&work->order_list, &workers->order_list);
-		}
-		spin_unlock(&workers->order_lock);
-	} else {
-		INIT_LIST_HEAD(&work->order_list);
-	}
-
-	spin_lock_irqsave(&worker->lock, flags);
-
-	if (test_bit(WORK_HIGH_PRIO_BIT, &work->flags))
-		list_add_tail(&work->list, &worker->prio_pending);
-	else
-		list_add_tail(&work->list, &worker->pending);
-	check_busy_worker(worker);
-
-	/*
-	 * avoid calling into wake_up_process if this thread has already
-	 * been kicked
-	 */
-	if (!worker->working)
-		wake = 1;
-	worker->working = 1;
-
-	if (wake)
-		wake_up_process(worker->task);
-	spin_unlock_irqrestore(&worker->lock, flags);
-
-out:
-	return 0;
+	set_bit(WORK_HIGH_PRIO_BIT, &work->flags);
 }
diff --git a/fs/btrfs/async-thread.h b/fs/btrfs/async-thread.h
index 5077746cf85..9c6b66d15fb 100644
--- a/fs/btrfs/async-thread.h
+++ b/fs/btrfs/async-thread.h
@@ -1,5 +1,6 @@
 /*
  * Copyright (C) 2007 Oracle.  All rights reserved.
+ * Copyright (C) 2014 Fujitsu.  All rights reserved.
  *
  * This program is free software; you can redistribute it and/or
  * modify it under the terms of the GNU General Public
@@ -19,101 +20,35 @@
 #ifndef __BTRFS_ASYNC_THREAD_
 #define __BTRFS_ASYNC_THREAD_
 
-struct btrfs_worker_thread;
+struct btrfs_workqueue;
+/* Internal use only */
+struct __btrfs_workqueue;
+struct btrfs_work;
+typedef void (*btrfs_func_t)(struct btrfs_work *arg);
 
-/*
- * This is similar to a workqueue, but it is meant to spread the operations
- * across all available cpus instead of just the CPU that was used to
- * queue the work.  There is also some batching introduced to try and
- * cut down on context switches.
- *
- * By default threads are added on demand up to 2 * the number of cpus.
- * Changing struct btrfs_workers->max_workers is one way to prevent
- * demand creation of kthreads.
- *
- * the basic model of these worker threads is to embed a btrfs_work
- * structure in your own data struct, and use container_of in a
- * work function to get back to your data struct.
- */
 struct btrfs_work {
-	/*
-	 * func should be set to the function you want called
-	 * your work struct is passed as the only arg
-	 *
-	 * ordered_func must be set for work sent to an ordered work queue,
-	 * and it is called to complete a given work item in the same
-	 * order they were sent to the queue.
-	 */
-	void (*func)(struct btrfs_work *work);
-	void (*ordered_func)(struct btrfs_work *work);
-	void (*ordered_free)(struct btrfs_work *work);
-
-	/*
-	 * flags should be set to zero.  It is used to make sure the
-	 * struct is only inserted once into the list.
-	 */
+	btrfs_func_t func;
+	btrfs_func_t ordered_func;
+	btrfs_func_t ordered_free;
+
+	/* Don't touch things below */
+	struct work_struct normal_work;
+	struct list_head ordered_list;
+	struct __btrfs_workqueue *wq;
 	unsigned long flags;
-
-	/* don't touch these */
-	struct btrfs_worker_thread *worker;
-	struct list_head list;
-	struct list_head order_list;
-};
-
-struct btrfs_workers {
-	/* current number of running workers */
-	int num_workers;
-
-	int num_workers_starting;
-
-	/* max number of workers allowed.  changed by btrfs_start_workers */
-	int max_workers;
-
-	/* once a worker has this many requests or fewer, it is idle */
-	int idle_thresh;
-
-	/* force completions in the order they were queued */
-	int ordered;
-
-	/* more workers required, but in an interrupt handler */
-	int atomic_start_pending;
-
-	/*
-	 * are we allowed to sleep while starting workers or are we required
-	 * to start them at a later time?  If we can't sleep, this indicates
-	 * which queue we need to use to schedule thread creation.
-	 */
-	struct btrfs_workers *atomic_worker_start;
-
-	/* list with all the work threads.  The workers on the idle thread
-	 * may be actively servicing jobs, but they haven't yet hit the
-	 * idle thresh limit above.
-	 */
-	struct list_head worker_list;
-	struct list_head idle_list;
-
-	/*
-	 * when operating in ordered mode, this maintains the list
-	 * of work items waiting for completion
-	 */
-	struct list_head order_list;
-	struct list_head prio_order_list;
-
-	/* lock for finding the next worker thread to queue on */
-	spinlock_t lock;
-
-	/* lock for the ordered lists */
-	spinlock_t order_lock;
-
-	/* extra name for this worker, used for current->name */
-	char *name;
 };
 
-int btrfs_queue_worker(struct btrfs_workers *workers, struct btrfs_work *work);
-int btrfs_start_workers(struct btrfs_workers *workers, int num_workers);
-int btrfs_stop_workers(struct btrfs_workers *workers);
-void btrfs_init_workers(struct btrfs_workers *workers, char *name, int max,
-			struct btrfs_workers *async_starter);
-int btrfs_requeue_work(struct btrfs_work *work);
-void btrfs_set_work_high_prio(struct btrfs_work *work);
+struct btrfs_workqueue *btrfs_alloc_workqueue(const char *name,
+					      int flags,
+					      int max_active,
+					      int thresh);
+void btrfs_init_work(struct btrfs_work *work,
+		     btrfs_func_t func,
+		     btrfs_func_t ordered_func,
+		     btrfs_func_t ordered_free);
+void btrfs_queue_work(struct btrfs_workqueue *wq,
+		      struct btrfs_work *work);
+void btrfs_destroy_workqueue(struct btrfs_workqueue *wq);
+void btrfs_workqueue_set_max(struct btrfs_workqueue *wq, int max);
+void btrfs_set_work_high_priority(struct btrfs_work *work);
 #endif
diff --git a/fs/btrfs/backref.c b/fs/btrfs/backref.c
new file mode 100644
index 00000000000..e25564bfcb4
--- /dev/null
+++ b/fs/btrfs/backref.c
@@ -0,0 +1,1883 @@
+/*
+ * Copyright (C) 2011 STRATO.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/vmalloc.h>
+#include "ctree.h"
+#include "disk-io.h"
+#include "backref.h"
+#include "ulist.h"
+#include "transaction.h"
+#include "delayed-ref.h"
+#include "locking.h"
+
+struct extent_inode_elem {
+	u64 inum;
+	u64 offset;
+	struct extent_inode_elem *next;
+};
+
+static int check_extent_in_eb(struct btrfs_key *key, struct extent_buffer *eb,
+				struct btrfs_file_extent_item *fi,
+				u64 extent_item_pos,
+				struct extent_inode_elem **eie)
+{
+	u64 offset = 0;
+	struct extent_inode_elem *e;
+
+	if (!btrfs_file_extent_compression(eb, fi) &&
+	    !btrfs_file_extent_encryption(eb, fi) &&
+	    !btrfs_file_extent_other_encoding(eb, fi)) {
+		u64 data_offset;
+		u64 data_len;
+
+		data_offset = btrfs_file_extent_offset(eb, fi);
+		data_len = btrfs_file_extent_num_bytes(eb, fi);
+
+		if (extent_item_pos < data_offset ||
+		    extent_item_pos >= data_offset + data_len)
+			return 1;
+		offset = extent_item_pos - data_offset;
+	}
+
+	e = kmalloc(sizeof(*e), GFP_NOFS);
+	if (!e)
+		return -ENOMEM;
+
+	e->next = *eie;
+	e->inum = key->objectid;
+	e->offset = key->offset + offset;
+	*eie = e;
+
+	return 0;
+}
+
+static void free_inode_elem_list(struct extent_inode_elem *eie)
+{
+	struct extent_inode_elem *eie_next;
+
+	for (; eie; eie = eie_next) {
+		eie_next = eie->next;
+		kfree(eie);
+	}
+}
+
+static int find_extent_in_eb(struct extent_buffer *eb, u64 wanted_disk_byte,
+				u64 extent_item_pos,
+				struct extent_inode_elem **eie)
+{
+	u64 disk_byte;
+	struct btrfs_key key;
+	struct btrfs_file_extent_item *fi;
+	int slot;
+	int nritems;
+	int extent_type;
+	int ret;
+
+	/*
+	 * from the shared data ref, we only have the leaf but we need
+	 * the key. thus, we must look into all items and see that we
+	 * find one (some) with a reference to our extent item.
+	 */
+	nritems = btrfs_header_nritems(eb);
+	for (slot = 0; slot < nritems; ++slot) {
+		btrfs_item_key_to_cpu(eb, &key, slot);
+		if (key.type != BTRFS_EXTENT_DATA_KEY)
+			continue;
+		fi = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item);
+		extent_type = btrfs_file_extent_type(eb, fi);
+		if (extent_type == BTRFS_FILE_EXTENT_INLINE)
+			continue;
+		/* don't skip BTRFS_FILE_EXTENT_PREALLOC, we can handle that */
+		disk_byte = btrfs_file_extent_disk_bytenr(eb, fi);
+		if (disk_byte != wanted_disk_byte)
+			continue;
+
+		ret = check_extent_in_eb(&key, eb, fi, extent_item_pos, eie);
+		if (ret < 0)
+			return ret;
+	}
+
+	return 0;
+}
+
+/*
+ * this structure records all encountered refs on the way up to the root
+ */
+struct __prelim_ref {
+	struct list_head list;
+	u64 root_id;
+	struct btrfs_key key_for_search;
+	int level;
+	int count;
+	struct extent_inode_elem *inode_list;
+	u64 parent;
+	u64 wanted_disk_byte;
+};
+
+static struct kmem_cache *btrfs_prelim_ref_cache;
+
+int __init btrfs_prelim_ref_init(void)
+{
+	btrfs_prelim_ref_cache = kmem_cache_create("btrfs_prelim_ref",
+					sizeof(struct __prelim_ref),
+					0,
+					SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD,
+					NULL);
+	if (!btrfs_prelim_ref_cache)
+		return -ENOMEM;
+	return 0;
+}
+
+void btrfs_prelim_ref_exit(void)
+{
+	if (btrfs_prelim_ref_cache)
+		kmem_cache_destroy(btrfs_prelim_ref_cache);
+}
+
+/*
+ * the rules for all callers of this function are:
+ * - obtaining the parent is the goal
+ * - if you add a key, you must know that it is a correct key
+ * - if you cannot add the parent or a correct key, then we will look into the
+ *   block later to set a correct key
+ *
+ * delayed refs
+ * ============
+ *        backref type | shared | indirect | shared | indirect
+ * information         |   tree |     tree |   data |     data
+ * --------------------+--------+----------+--------+----------
+ *      parent logical |    y   |     -    |    -   |     -
+ *      key to resolve |    -   |     y    |    y   |     y
+ *  tree block logical |    -   |     -    |    -   |     -
+ *  root for resolving |    y   |     y    |    y   |     y
+ *
+ * - column 1:       we've the parent -> done
+ * - column 2, 3, 4: we use the key to find the parent
+ *
+ * on disk refs (inline or keyed)
+ * ==============================
+ *        backref type | shared | indirect | shared | indirect
+ * information         |   tree |     tree |   data |     data
+ * --------------------+--------+----------+--------+----------
+ *      parent logical |    y   |     -    |    y   |     -
+ *      key to resolve |    -   |     -    |    -   |     y
+ *  tree block logical |    y   |     y    |    y   |     y
+ *  root for resolving |    -   |     y    |    y   |     y
+ *
+ * - column 1, 3: we've the parent -> done
+ * - column 2:    we take the first key from the block to find the parent
+ *                (see __add_missing_keys)
+ * - column 4:    we use the key to find the parent
+ *
+ * additional information that's available but not required to find the parent
+ * block might help in merging entries to gain some speed.
+ */
+
+static int __add_prelim_ref(struct list_head *head, u64 root_id,
+			    struct btrfs_key *key, int level,
+			    u64 parent, u64 wanted_disk_byte, int count,
+			    gfp_t gfp_mask)
+{
+	struct __prelim_ref *ref;
+
+	if (root_id == BTRFS_DATA_RELOC_TREE_OBJECTID)
+		return 0;
+
+	ref = kmem_cache_alloc(btrfs_prelim_ref_cache, gfp_mask);
+	if (!ref)
+		return -ENOMEM;
+
+	ref->root_id = root_id;
+	if (key)
+		ref->key_for_search = *key;
+	else
+		memset(&ref->key_for_search, 0, sizeof(ref->key_for_search));
+
+	ref->inode_list = NULL;
+	ref->level = level;
+	ref->count = count;
+	ref->parent = parent;
+	ref->wanted_disk_byte = wanted_disk_byte;
+	list_add_tail(&ref->list, head);
+
+	return 0;
+}
+
+static int add_all_parents(struct btrfs_root *root, struct btrfs_path *path,
+			   struct ulist *parents, struct __prelim_ref *ref,
+			   int level, u64 time_seq, const u64 *extent_item_pos,
+			   u64 total_refs)
+{
+	int ret = 0;
+	int slot;
+	struct extent_buffer *eb;
+	struct btrfs_key key;
+	struct btrfs_key *key_for_search = &ref->key_for_search;
+	struct btrfs_file_extent_item *fi;
+	struct extent_inode_elem *eie = NULL, *old = NULL;
+	u64 disk_byte;
+	u64 wanted_disk_byte = ref->wanted_disk_byte;
+	u64 count = 0;
+
+	if (level != 0) {
+		eb = path->nodes[level];
+		ret = ulist_add(parents, eb->start, 0, GFP_NOFS);
+		if (ret < 0)
+			return ret;
+		return 0;
+	}
+
+	/*
+	 * We normally enter this function with the path already pointing to
+	 * the first item to check. But sometimes, we may enter it with
+	 * slot==nritems. In that case, go to the next leaf before we continue.
+	 */
+	if (path->slots[0] >= btrfs_header_nritems(path->nodes[0]))
+		ret = btrfs_next_old_leaf(root, path, time_seq);
+
+	while (!ret && count < total_refs) {
+		eb = path->nodes[0];
+		slot = path->slots[0];
+
+		btrfs_item_key_to_cpu(eb, &key, slot);
+
+		if (key.objectid != key_for_search->objectid ||
+		    key.type != BTRFS_EXTENT_DATA_KEY)
+			break;
+
+		fi = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item);
+		disk_byte = btrfs_file_extent_disk_bytenr(eb, fi);
+
+		if (disk_byte == wanted_disk_byte) {
+			eie = NULL;
+			old = NULL;
+			count++;
+			if (extent_item_pos) {
+				ret = check_extent_in_eb(&key, eb, fi,
+						*extent_item_pos,
+						&eie);
+				if (ret < 0)
+					break;
+			}
+			if (ret > 0)
+				goto next;
+			ret = ulist_add_merge(parents, eb->start,
+					      (uintptr_t)eie,
+					      (u64 *)&old, GFP_NOFS);
+			if (ret < 0)
+				break;
+			if (!ret && extent_item_pos) {
+				while (old->next)
+					old = old->next;
+				old->next = eie;
+			}
+			eie = NULL;
+		}
+next:
+		ret = btrfs_next_old_item(root, path, time_seq);
+	}
+
+	if (ret > 0)
+		ret = 0;
+	else if (ret < 0)
+		free_inode_elem_list(eie);
+	return ret;
+}
+
+/*
+ * resolve an indirect backref in the form (root_id, key, level)
+ * to a logical address
+ */
+static int __resolve_indirect_ref(struct btrfs_fs_info *fs_info,
+				  struct btrfs_path *path, u64 time_seq,
+				  struct __prelim_ref *ref,
+				  struct ulist *parents,
+				  const u64 *extent_item_pos, u64 total_refs)
+{
+	struct btrfs_root *root;
+	struct btrfs_key root_key;
+	struct extent_buffer *eb;
+	int ret = 0;
+	int root_level;
+	int level = ref->level;
+	int index;
+
+	root_key.objectid = ref->root_id;
+	root_key.type = BTRFS_ROOT_ITEM_KEY;
+	root_key.offset = (u64)-1;
+
+	index = srcu_read_lock(&fs_info->subvol_srcu);
+
+	root = btrfs_read_fs_root_no_name(fs_info, &root_key);
+	if (IS_ERR(root)) {
+		srcu_read_unlock(&fs_info->subvol_srcu, index);
+		ret = PTR_ERR(root);
+		goto out;
+	}
+
+	if (path->search_commit_root)
+		root_level = btrfs_header_level(root->commit_root);
+	else
+		root_level = btrfs_old_root_level(root, time_seq);
+
+	if (root_level + 1 == level) {
+		srcu_read_unlock(&fs_info->subvol_srcu, index);
+		goto out;
+	}
+
+	path->lowest_level = level;
+	ret = btrfs_search_old_slot(root, &ref->key_for_search, path, time_seq);
+
+	/* root node has been locked, we can release @subvol_srcu safely here */
+	srcu_read_unlock(&fs_info->subvol_srcu, index);
+
+	pr_debug("search slot in root %llu (level %d, ref count %d) returned "
+		 "%d for key (%llu %u %llu)\n",
+		 ref->root_id, level, ref->count, ret,
+		 ref->key_for_search.objectid, ref->key_for_search.type,
+		 ref->key_for_search.offset);
+	if (ret < 0)
+		goto out;
+
+	eb = path->nodes[level];
+	while (!eb) {
+		if (WARN_ON(!level)) {
+			ret = 1;
+			goto out;
+		}
+		level--;
+		eb = path->nodes[level];
+	}
+
+	ret = add_all_parents(root, path, parents, ref, level, time_seq,
+			      extent_item_pos, total_refs);
+out:
+	path->lowest_level = 0;
+	btrfs_release_path(path);
+	return ret;
+}
+
+/*
+ * resolve all indirect backrefs from the list
+ */
+static int __resolve_indirect_refs(struct btrfs_fs_info *fs_info,
+				   struct btrfs_path *path, u64 time_seq,
+				   struct list_head *head,
+				   const u64 *extent_item_pos, u64 total_refs)
+{
+	int err;
+	int ret = 0;
+	struct __prelim_ref *ref;
+	struct __prelim_ref *ref_safe;
+	struct __prelim_ref *new_ref;
+	struct ulist *parents;
+	struct ulist_node *node;
+	struct ulist_iterator uiter;
+
+	parents = ulist_alloc(GFP_NOFS);
+	if (!parents)
+		return -ENOMEM;
+
+	/*
+	 * _safe allows us to insert directly after the current item without
+	 * iterating over the newly inserted items.
+	 * we're also allowed to re-assign ref during iteration.
+	 */
+	list_for_each_entry_safe(ref, ref_safe, head, list) {
+		if (ref->parent)	/* already direct */
+			continue;
+		if (ref->count == 0)
+			continue;
+		err = __resolve_indirect_ref(fs_info, path, time_seq, ref,
+					     parents, extent_item_pos,
+					     total_refs);
+		/*
+		 * we can only tolerate ENOENT,otherwise,we should catch error
+		 * and return directly.
+		 */
+		if (err == -ENOENT) {
+			continue;
+		} else if (err) {
+			ret = err;
+			goto out;
+		}
+
+		/* we put the first parent into the ref at hand */
+		ULIST_ITER_INIT(&uiter);
+		node = ulist_next(parents, &uiter);
+		ref->parent = node ? node->val : 0;
+		ref->inode_list = node ?
+			(struct extent_inode_elem *)(uintptr_t)node->aux : NULL;
+
+		/* additional parents require new refs being added here */
+		while ((node = ulist_next(parents, &uiter))) {
+			new_ref = kmem_cache_alloc(btrfs_prelim_ref_cache,
+						   GFP_NOFS);
+			if (!new_ref) {
+				ret = -ENOMEM;
+				goto out;
+			}
+			memcpy(new_ref, ref, sizeof(*ref));
+			new_ref->parent = node->val;
+			new_ref->inode_list = (struct extent_inode_elem *)
+							(uintptr_t)node->aux;
+			list_add(&new_ref->list, &ref->list);
+		}
+		ulist_reinit(parents);
+	}
+out:
+	ulist_free(parents);
+	return ret;
+}
+
+static inline int ref_for_same_block(struct __prelim_ref *ref1,
+				     struct __prelim_ref *ref2)
+{
+	if (ref1->level != ref2->level)
+		return 0;
+	if (ref1->root_id != ref2->root_id)
+		return 0;
+	if (ref1->key_for_search.type != ref2->key_for_search.type)
+		return 0;
+	if (ref1->key_for_search.objectid != ref2->key_for_search.objectid)
+		return 0;
+	if (ref1->key_for_search.offset != ref2->key_for_search.offset)
+		return 0;
+	if (ref1->parent != ref2->parent)
+		return 0;
+
+	return 1;
+}
+
+/*
+ * read tree blocks and add keys where required.
+ */
+static int __add_missing_keys(struct btrfs_fs_info *fs_info,
+			      struct list_head *head)
+{
+	struct list_head *pos;
+	struct extent_buffer *eb;
+
+	list_for_each(pos, head) {
+		struct __prelim_ref *ref;
+		ref = list_entry(pos, struct __prelim_ref, list);
+
+		if (ref->parent)
+			continue;
+		if (ref->key_for_search.type)
+			continue;
+		BUG_ON(!ref->wanted_disk_byte);
+		eb = read_tree_block(fs_info->tree_root, ref->wanted_disk_byte,
+				     fs_info->tree_root->leafsize, 0);
+		if (!eb || !extent_buffer_uptodate(eb)) {
+			free_extent_buffer(eb);
+			return -EIO;
+		}
+		btrfs_tree_read_lock(eb);
+		if (btrfs_header_level(eb) == 0)
+			btrfs_item_key_to_cpu(eb, &ref->key_for_search, 0);
+		else
+			btrfs_node_key_to_cpu(eb, &ref->key_for_search, 0);
+		btrfs_tree_read_unlock(eb);
+		free_extent_buffer(eb);
+	}
+	return 0;
+}
+
+/*
+ * merge two lists of backrefs and adjust counts accordingly
+ *
+ * mode = 1: merge identical keys, if key is set
+ *    FIXME: if we add more keys in __add_prelim_ref, we can merge more here.
+ *           additionally, we could even add a key range for the blocks we
+ *           looked into to merge even more (-> replace unresolved refs by those
+ *           having a parent).
+ * mode = 2: merge identical parents
+ */
+static void __merge_refs(struct list_head *head, int mode)
+{
+	struct list_head *pos1;
+
+	list_for_each(pos1, head) {
+		struct list_head *n2;
+		struct list_head *pos2;
+		struct __prelim_ref *ref1;
+
+		ref1 = list_entry(pos1, struct __prelim_ref, list);
+
+		for (pos2 = pos1->next, n2 = pos2->next; pos2 != head;
+		     pos2 = n2, n2 = pos2->next) {
+			struct __prelim_ref *ref2;
+			struct __prelim_ref *xchg;
+			struct extent_inode_elem *eie;
+
+			ref2 = list_entry(pos2, struct __prelim_ref, list);
+
+			if (mode == 1) {
+				if (!ref_for_same_block(ref1, ref2))
+					continue;
+				if (!ref1->parent && ref2->parent) {
+					xchg = ref1;
+					ref1 = ref2;
+					ref2 = xchg;
+				}
+			} else {
+				if (ref1->parent != ref2->parent)
+					continue;
+			}
+
+			eie = ref1->inode_list;
+			while (eie && eie->next)
+				eie = eie->next;
+			if (eie)
+				eie->next = ref2->inode_list;
+			else
+				ref1->inode_list = ref2->inode_list;
+			ref1->count += ref2->count;
+
+			list_del(&ref2->list);
+			kmem_cache_free(btrfs_prelim_ref_cache, ref2);
+		}
+
+	}
+}
+
+/*
+ * add all currently queued delayed refs from this head whose seq nr is
+ * smaller or equal that seq to the list
+ */
+static int __add_delayed_refs(struct btrfs_delayed_ref_head *head, u64 seq,
+			      struct list_head *prefs, u64 *total_refs)
+{
+	struct btrfs_delayed_extent_op *extent_op = head->extent_op;
+	struct rb_node *n = &head->node.rb_node;
+	struct btrfs_key key;
+	struct btrfs_key op_key = {0};
+	int sgn;
+	int ret = 0;
+
+	if (extent_op && extent_op->update_key)
+		btrfs_disk_key_to_cpu(&op_key, &extent_op->key);
+
+	spin_lock(&head->lock);
+	n = rb_first(&head->ref_root);
+	while (n) {
+		struct btrfs_delayed_ref_node *node;
+		node = rb_entry(n, struct btrfs_delayed_ref_node,
+				rb_node);
+		n = rb_next(n);
+		if (node->seq > seq)
+			continue;
+
+		switch (node->action) {
+		case BTRFS_ADD_DELAYED_EXTENT:
+		case BTRFS_UPDATE_DELAYED_HEAD:
+			WARN_ON(1);
+			continue;
+		case BTRFS_ADD_DELAYED_REF:
+			sgn = 1;
+			break;
+		case BTRFS_DROP_DELAYED_REF:
+			sgn = -1;
+			break;
+		default:
+			BUG_ON(1);
+		}
+		*total_refs += (node->ref_mod * sgn);
+		switch (node->type) {
+		case BTRFS_TREE_BLOCK_REF_KEY: {
+			struct btrfs_delayed_tree_ref *ref;
+
+			ref = btrfs_delayed_node_to_tree_ref(node);
+			ret = __add_prelim_ref(prefs, ref->root, &op_key,
+					       ref->level + 1, 0, node->bytenr,
+					       node->ref_mod * sgn, GFP_ATOMIC);
+			break;
+		}
+		case BTRFS_SHARED_BLOCK_REF_KEY: {
+			struct btrfs_delayed_tree_ref *ref;
+
+			ref = btrfs_delayed_node_to_tree_ref(node);
+			ret = __add_prelim_ref(prefs, ref->root, NULL,
+					       ref->level + 1, ref->parent,
+					       node->bytenr,
+					       node->ref_mod * sgn, GFP_ATOMIC);
+			break;
+		}
+		case BTRFS_EXTENT_DATA_REF_KEY: {
+			struct btrfs_delayed_data_ref *ref;
+			ref = btrfs_delayed_node_to_data_ref(node);
+
+			key.objectid = ref->objectid;
+			key.type = BTRFS_EXTENT_DATA_KEY;
+			key.offset = ref->offset;
+			ret = __add_prelim_ref(prefs, ref->root, &key, 0, 0,
+					       node->bytenr,
+					       node->ref_mod * sgn, GFP_ATOMIC);
+			break;
+		}
+		case BTRFS_SHARED_DATA_REF_KEY: {
+			struct btrfs_delayed_data_ref *ref;
+
+			ref = btrfs_delayed_node_to_data_ref(node);
+
+			key.objectid = ref->objectid;
+			key.type = BTRFS_EXTENT_DATA_KEY;
+			key.offset = ref->offset;
+			ret = __add_prelim_ref(prefs, ref->root, &key, 0,
+					       ref->parent, node->bytenr,
+					       node->ref_mod * sgn, GFP_ATOMIC);
+			break;
+		}
+		default:
+			WARN_ON(1);
+		}
+		if (ret)
+			break;
+	}
+	spin_unlock(&head->lock);
+	return ret;
+}
+
+/*
+ * add all inline backrefs for bytenr to the list
+ */
+static int __add_inline_refs(struct btrfs_fs_info *fs_info,
+			     struct btrfs_path *path, u64 bytenr,
+			     int *info_level, struct list_head *prefs,
+			     u64 *total_refs)
+{
+	int ret = 0;
+	int slot;
+	struct extent_buffer *leaf;
+	struct btrfs_key key;
+	struct btrfs_key found_key;
+	unsigned long ptr;
+	unsigned long end;
+	struct btrfs_extent_item *ei;
+	u64 flags;
+	u64 item_size;
+
+	/*
+	 * enumerate all inline refs
+	 */
+	leaf = path->nodes[0];
+	slot = path->slots[0];
+
+	item_size = btrfs_item_size_nr(leaf, slot);
+	BUG_ON(item_size < sizeof(*ei));
+
+	ei = btrfs_item_ptr(leaf, slot, struct btrfs_extent_item);
+	flags = btrfs_extent_flags(leaf, ei);
+	*total_refs += btrfs_extent_refs(leaf, ei);
+	btrfs_item_key_to_cpu(leaf, &found_key, slot);
+
+	ptr = (unsigned long)(ei + 1);
+	end = (unsigned long)ei + item_size;
+
+	if (found_key.type == BTRFS_EXTENT_ITEM_KEY &&
+	    flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
+		struct btrfs_tree_block_info *info;
+
+		info = (struct btrfs_tree_block_info *)ptr;
+		*info_level = btrfs_tree_block_level(leaf, info);
+		ptr += sizeof(struct btrfs_tree_block_info);
+		BUG_ON(ptr > end);
+	} else if (found_key.type == BTRFS_METADATA_ITEM_KEY) {
+		*info_level = found_key.offset;
+	} else {
+		BUG_ON(!(flags & BTRFS_EXTENT_FLAG_DATA));
+	}
+
+	while (ptr < end) {
+		struct btrfs_extent_inline_ref *iref;
+		u64 offset;
+		int type;
+
+		iref = (struct btrfs_extent_inline_ref *)ptr;
+		type = btrfs_extent_inline_ref_type(leaf, iref);
+		offset = btrfs_extent_inline_ref_offset(leaf, iref);
+
+		switch (type) {
+		case BTRFS_SHARED_BLOCK_REF_KEY:
+			ret = __add_prelim_ref(prefs, 0, NULL,
+						*info_level + 1, offset,
+						bytenr, 1, GFP_NOFS);
+			break;
+		case BTRFS_SHARED_DATA_REF_KEY: {
+			struct btrfs_shared_data_ref *sdref;
+			int count;
+
+			sdref = (struct btrfs_shared_data_ref *)(iref + 1);
+			count = btrfs_shared_data_ref_count(leaf, sdref);
+			ret = __add_prelim_ref(prefs, 0, NULL, 0, offset,
+					       bytenr, count, GFP_NOFS);
+			break;
+		}
+		case BTRFS_TREE_BLOCK_REF_KEY:
+			ret = __add_prelim_ref(prefs, offset, NULL,
+					       *info_level + 1, 0,
+					       bytenr, 1, GFP_NOFS);
+			break;
+		case BTRFS_EXTENT_DATA_REF_KEY: {
+			struct btrfs_extent_data_ref *dref;
+			int count;
+			u64 root;
+
+			dref = (struct btrfs_extent_data_ref *)(&iref->offset);
+			count = btrfs_extent_data_ref_count(leaf, dref);
+			key.objectid = btrfs_extent_data_ref_objectid(leaf,
+								      dref);
+			key.type = BTRFS_EXTENT_DATA_KEY;
+			key.offset = btrfs_extent_data_ref_offset(leaf, dref);
+			root = btrfs_extent_data_ref_root(leaf, dref);
+			ret = __add_prelim_ref(prefs, root, &key, 0, 0,
+					       bytenr, count, GFP_NOFS);
+			break;
+		}
+		default:
+			WARN_ON(1);
+		}
+		if (ret)
+			return ret;
+		ptr += btrfs_extent_inline_ref_size(type);
+	}
+
+	return 0;
+}
+
+/*
+ * add all non-inline backrefs for bytenr to the list
+ */
+static int __add_keyed_refs(struct btrfs_fs_info *fs_info,
+			    struct btrfs_path *path, u64 bytenr,
+			    int info_level, struct list_head *prefs)
+{
+	struct btrfs_root *extent_root = fs_info->extent_root;
+	int ret;
+	int slot;
+	struct extent_buffer *leaf;
+	struct btrfs_key key;
+
+	while (1) {
+		ret = btrfs_next_item(extent_root, path);
+		if (ret < 0)
+			break;
+		if (ret) {
+			ret = 0;
+			break;
+		}
+
+		slot = path->slots[0];
+		leaf = path->nodes[0];
+		btrfs_item_key_to_cpu(leaf, &key, slot);
+
+		if (key.objectid != bytenr)
+			break;
+		if (key.type < BTRFS_TREE_BLOCK_REF_KEY)
+			continue;
+		if (key.type > BTRFS_SHARED_DATA_REF_KEY)
+			break;
+
+		switch (key.type) {
+		case BTRFS_SHARED_BLOCK_REF_KEY:
+			ret = __add_prelim_ref(prefs, 0, NULL,
+						info_level + 1, key.offset,
+						bytenr, 1, GFP_NOFS);
+			break;
+		case BTRFS_SHARED_DATA_REF_KEY: {
+			struct btrfs_shared_data_ref *sdref;
+			int count;
+
+			sdref = btrfs_item_ptr(leaf, slot,
+					      struct btrfs_shared_data_ref);
+			count = btrfs_shared_data_ref_count(leaf, sdref);
+			ret = __add_prelim_ref(prefs, 0, NULL, 0, key.offset,
+						bytenr, count, GFP_NOFS);
+			break;
+		}
+		case BTRFS_TREE_BLOCK_REF_KEY:
+			ret = __add_prelim_ref(prefs, key.offset, NULL,
+					       info_level + 1, 0,
+					       bytenr, 1, GFP_NOFS);
+			break;
+		case BTRFS_EXTENT_DATA_REF_KEY: {
+			struct btrfs_extent_data_ref *dref;
+			int count;
+			u64 root;
+
+			dref = btrfs_item_ptr(leaf, slot,
+					      struct btrfs_extent_data_ref);
+			count = btrfs_extent_data_ref_count(leaf, dref);
+			key.objectid = btrfs_extent_data_ref_objectid(leaf,
+								      dref);
+			key.type = BTRFS_EXTENT_DATA_KEY;
+			key.offset = btrfs_extent_data_ref_offset(leaf, dref);
+			root = btrfs_extent_data_ref_root(leaf, dref);
+			ret = __add_prelim_ref(prefs, root, &key, 0, 0,
+					       bytenr, count, GFP_NOFS);
+			break;
+		}
+		default:
+			WARN_ON(1);
+		}
+		if (ret)
+			return ret;
+
+	}
+
+	return ret;
+}
+
+/*
+ * this adds all existing backrefs (inline backrefs, backrefs and delayed
+ * refs) for the given bytenr to the refs list, merges duplicates and resolves
+ * indirect refs to their parent bytenr.
+ * When roots are found, they're added to the roots list
+ *
+ * FIXME some caching might speed things up
+ */
+static int find_parent_nodes(struct btrfs_trans_handle *trans,
+			     struct btrfs_fs_info *fs_info, u64 bytenr,
+			     u64 time_seq, struct ulist *refs,
+			     struct ulist *roots, const u64 *extent_item_pos)
+{
+	struct btrfs_key key;
+	struct btrfs_path *path;
+	struct btrfs_delayed_ref_root *delayed_refs = NULL;
+	struct btrfs_delayed_ref_head *head;
+	int info_level = 0;
+	int ret;
+	struct list_head prefs_delayed;
+	struct list_head prefs;
+	struct __prelim_ref *ref;
+	struct extent_inode_elem *eie = NULL;
+	u64 total_refs = 0;
+
+	INIT_LIST_HEAD(&prefs);
+	INIT_LIST_HEAD(&prefs_delayed);
+
+	key.objectid = bytenr;
+	key.offset = (u64)-1;
+	if (btrfs_fs_incompat(fs_info, SKINNY_METADATA))
+		key.type = BTRFS_METADATA_ITEM_KEY;
+	else
+		key.type = BTRFS_EXTENT_ITEM_KEY;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+	if (!trans) {
+		path->search_commit_root = 1;
+		path->skip_locking = 1;
+	}
+
+	/*
+	 * grab both a lock on the path and a lock on the delayed ref head.
+	 * We need both to get a consistent picture of how the refs look
+	 * at a specified point in time
+	 */
+again:
+	head = NULL;
+
+	ret = btrfs_search_slot(trans, fs_info->extent_root, &key, path, 0, 0);
+	if (ret < 0)
+		goto out;
+	BUG_ON(ret == 0);
+
+#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
+	if (trans && likely(trans->type != __TRANS_DUMMY)) {
+#else
+	if (trans) {
+#endif
+		/*
+		 * look if there are updates for this ref queued and lock the
+		 * head
+		 */
+		delayed_refs = &trans->transaction->delayed_refs;
+		spin_lock(&delayed_refs->lock);
+		head = btrfs_find_delayed_ref_head(trans, bytenr);
+		if (head) {
+			if (!mutex_trylock(&head->mutex)) {
+				atomic_inc(&head->node.refs);
+				spin_unlock(&delayed_refs->lock);
+
+				btrfs_release_path(path);
+
+				/*
+				 * Mutex was contended, block until it's
+				 * released and try again
+				 */
+				mutex_lock(&head->mutex);
+				mutex_unlock(&head->mutex);
+				btrfs_put_delayed_ref(&head->node);
+				goto again;
+			}
+			spin_unlock(&delayed_refs->lock);
+			ret = __add_delayed_refs(head, time_seq,
+						 &prefs_delayed, &total_refs);
+			mutex_unlock(&head->mutex);
+			if (ret)
+				goto out;
+		} else {
+			spin_unlock(&delayed_refs->lock);
+		}
+	}
+
+	if (path->slots[0]) {
+		struct extent_buffer *leaf;
+		int slot;
+
+		path->slots[0]--;
+		leaf = path->nodes[0];
+		slot = path->slots[0];
+		btrfs_item_key_to_cpu(leaf, &key, slot);
+		if (key.objectid == bytenr &&
+		    (key.type == BTRFS_EXTENT_ITEM_KEY ||
+		     key.type == BTRFS_METADATA_ITEM_KEY)) {
+			ret = __add_inline_refs(fs_info, path, bytenr,
+						&info_level, &prefs,
+						&total_refs);
+			if (ret)
+				goto out;
+			ret = __add_keyed_refs(fs_info, path, bytenr,
+					       info_level, &prefs);
+			if (ret)
+				goto out;
+		}
+	}
+	btrfs_release_path(path);
+
+	list_splice_init(&prefs_delayed, &prefs);
+
+	ret = __add_missing_keys(fs_info, &prefs);
+	if (ret)
+		goto out;
+
+	__merge_refs(&prefs, 1);
+
+	ret = __resolve_indirect_refs(fs_info, path, time_seq, &prefs,
+				      extent_item_pos, total_refs);
+	if (ret)
+		goto out;
+
+	__merge_refs(&prefs, 2);
+
+	while (!list_empty(&prefs)) {
+		ref = list_first_entry(&prefs, struct __prelim_ref, list);
+		WARN_ON(ref->count < 0);
+		if (roots && ref->count && ref->root_id && ref->parent == 0) {
+			/* no parent == root of tree */
+			ret = ulist_add(roots, ref->root_id, 0, GFP_NOFS);
+			if (ret < 0)
+				goto out;
+		}
+		if (ref->count && ref->parent) {
+			if (extent_item_pos && !ref->inode_list &&
+			    ref->level == 0) {
+				u32 bsz;
+				struct extent_buffer *eb;
+				bsz = btrfs_level_size(fs_info->extent_root,
+							ref->level);
+				eb = read_tree_block(fs_info->extent_root,
+							   ref->parent, bsz, 0);
+				if (!eb || !extent_buffer_uptodate(eb)) {
+					free_extent_buffer(eb);
+					ret = -EIO;
+					goto out;
+				}
+				ret = find_extent_in_eb(eb, bytenr,
+							*extent_item_pos, &eie);
+				free_extent_buffer(eb);
+				if (ret < 0)
+					goto out;
+				ref->inode_list = eie;
+			}
+			ret = ulist_add_merge(refs, ref->parent,
+					      (uintptr_t)ref->inode_list,
+					      (u64 *)&eie, GFP_NOFS);
+			if (ret < 0)
+				goto out;
+			if (!ret && extent_item_pos) {
+				/*
+				 * we've recorded that parent, so we must extend
+				 * its inode list here
+				 */
+				BUG_ON(!eie);
+				while (eie->next)
+					eie = eie->next;
+				eie->next = ref->inode_list;
+			}
+			eie = NULL;
+		}
+		list_del(&ref->list);
+		kmem_cache_free(btrfs_prelim_ref_cache, ref);
+	}
+
+out:
+	btrfs_free_path(path);
+	while (!list_empty(&prefs)) {
+		ref = list_first_entry(&prefs, struct __prelim_ref, list);
+		list_del(&ref->list);
+		kmem_cache_free(btrfs_prelim_ref_cache, ref);
+	}
+	while (!list_empty(&prefs_delayed)) {
+		ref = list_first_entry(&prefs_delayed, struct __prelim_ref,
+				       list);
+		list_del(&ref->list);
+		kmem_cache_free(btrfs_prelim_ref_cache, ref);
+	}
+	if (ret < 0)
+		free_inode_elem_list(eie);
+	return ret;
+}
+
+static void free_leaf_list(struct ulist *blocks)
+{
+	struct ulist_node *node = NULL;
+	struct extent_inode_elem *eie;
+	struct ulist_iterator uiter;
+
+	ULIST_ITER_INIT(&uiter);
+	while ((node = ulist_next(blocks, &uiter))) {
+		if (!node->aux)
+			continue;
+		eie = (struct extent_inode_elem *)(uintptr_t)node->aux;
+		free_inode_elem_list(eie);
+		node->aux = 0;
+	}
+
+	ulist_free(blocks);
+}
+
+/*
+ * Finds all leafs with a reference to the specified combination of bytenr and
+ * offset. key_list_head will point to a list of corresponding keys (caller must
+ * free each list element). The leafs will be stored in the leafs ulist, which
+ * must be freed with ulist_free.
+ *
+ * returns 0 on success, <0 on error
+ */
+static int btrfs_find_all_leafs(struct btrfs_trans_handle *trans,
+				struct btrfs_fs_info *fs_info, u64 bytenr,
+				u64 time_seq, struct ulist **leafs,
+				const u64 *extent_item_pos)
+{
+	int ret;
+
+	*leafs = ulist_alloc(GFP_NOFS);
+	if (!*leafs)
+		return -ENOMEM;
+
+	ret = find_parent_nodes(trans, fs_info, bytenr,
+				time_seq, *leafs, NULL, extent_item_pos);
+	if (ret < 0 && ret != -ENOENT) {
+		free_leaf_list(*leafs);
+		return ret;
+	}
+
+	return 0;
+}
+
+/*
+ * walk all backrefs for a given extent to find all roots that reference this
+ * extent. Walking a backref means finding all extents that reference this
+ * extent and in turn walk the backrefs of those, too. Naturally this is a
+ * recursive process, but here it is implemented in an iterative fashion: We
+ * find all referencing extents for the extent in question and put them on a
+ * list. In turn, we find all referencing extents for those, further appending
+ * to the list. The way we iterate the list allows adding more elements after
+ * the current while iterating. The process stops when we reach the end of the
+ * list. Found roots are added to the roots list.
+ *
+ * returns 0 on success, < 0 on error.
+ */
+static int __btrfs_find_all_roots(struct btrfs_trans_handle *trans,
+				  struct btrfs_fs_info *fs_info, u64 bytenr,
+				  u64 time_seq, struct ulist **roots)
+{
+	struct ulist *tmp;
+	struct ulist_node *node = NULL;
+	struct ulist_iterator uiter;
+	int ret;
+
+	tmp = ulist_alloc(GFP_NOFS);
+	if (!tmp)
+		return -ENOMEM;
+	*roots = ulist_alloc(GFP_NOFS);
+	if (!*roots) {
+		ulist_free(tmp);
+		return -ENOMEM;
+	}
+
+	ULIST_ITER_INIT(&uiter);
+	while (1) {
+		ret = find_parent_nodes(trans, fs_info, bytenr,
+					time_seq, tmp, *roots, NULL);
+		if (ret < 0 && ret != -ENOENT) {
+			ulist_free(tmp);
+			ulist_free(*roots);
+			return ret;
+		}
+		node = ulist_next(tmp, &uiter);
+		if (!node)
+			break;
+		bytenr = node->val;
+		cond_resched();
+	}
+
+	ulist_free(tmp);
+	return 0;
+}
+
+int btrfs_find_all_roots(struct btrfs_trans_handle *trans,
+			 struct btrfs_fs_info *fs_info, u64 bytenr,
+			 u64 time_seq, struct ulist **roots)
+{
+	int ret;
+
+	if (!trans)
+		down_read(&fs_info->commit_root_sem);
+	ret = __btrfs_find_all_roots(trans, fs_info, bytenr, time_seq, roots);
+	if (!trans)
+		up_read(&fs_info->commit_root_sem);
+	return ret;
+}
+
+/*
+ * this makes the path point to (inum INODE_ITEM ioff)
+ */
+int inode_item_info(u64 inum, u64 ioff, struct btrfs_root *fs_root,
+			struct btrfs_path *path)
+{
+	struct btrfs_key key;
+	return btrfs_find_item(fs_root, path, inum, ioff,
+			BTRFS_INODE_ITEM_KEY, &key);
+}
+
+static int inode_ref_info(u64 inum, u64 ioff, struct btrfs_root *fs_root,
+				struct btrfs_path *path,
+				struct btrfs_key *found_key)
+{
+	return btrfs_find_item(fs_root, path, inum, ioff,
+			BTRFS_INODE_REF_KEY, found_key);
+}
+
+int btrfs_find_one_extref(struct btrfs_root *root, u64 inode_objectid,
+			  u64 start_off, struct btrfs_path *path,
+			  struct btrfs_inode_extref **ret_extref,
+			  u64 *found_off)
+{
+	int ret, slot;
+	struct btrfs_key key;
+	struct btrfs_key found_key;
+	struct btrfs_inode_extref *extref;
+	struct extent_buffer *leaf;
+	unsigned long ptr;
+
+	key.objectid = inode_objectid;
+	btrfs_set_key_type(&key, BTRFS_INODE_EXTREF_KEY);
+	key.offset = start_off;
+
+	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+	if (ret < 0)
+		return ret;
+
+	while (1) {
+		leaf = path->nodes[0];
+		slot = path->slots[0];
+		if (slot >= btrfs_header_nritems(leaf)) {
+			/*
+			 * If the item at offset is not found,
+			 * btrfs_search_slot will point us to the slot
+			 * where it should be inserted. In our case
+			 * that will be the slot directly before the
+			 * next INODE_REF_KEY_V2 item. In the case
+			 * that we're pointing to the last slot in a
+			 * leaf, we must move one leaf over.
+			 */
+			ret = btrfs_next_leaf(root, path);
+			if (ret) {
+				if (ret >= 1)
+					ret = -ENOENT;
+				break;
+			}
+			continue;
+		}
+
+		btrfs_item_key_to_cpu(leaf, &found_key, slot);
+
+		/*
+		 * Check that we're still looking at an extended ref key for
+		 * this particular objectid. If we have different
+		 * objectid or type then there are no more to be found
+		 * in the tree and we can exit.
+		 */
+		ret = -ENOENT;
+		if (found_key.objectid != inode_objectid)
+			break;
+		if (btrfs_key_type(&found_key) != BTRFS_INODE_EXTREF_KEY)
+			break;
+
+		ret = 0;
+		ptr = btrfs_item_ptr_offset(leaf, path->slots[0]);
+		extref = (struct btrfs_inode_extref *)ptr;
+		*ret_extref = extref;
+		if (found_off)
+			*found_off = found_key.offset;
+		break;
+	}
+
+	return ret;
+}
+
+/*
+ * this iterates to turn a name (from iref/extref) into a full filesystem path.
+ * Elements of the path are separated by '/' and the path is guaranteed to be
+ * 0-terminated. the path is only given within the current file system.
+ * Therefore, it never starts with a '/'. the caller is responsible to provide
+ * "size" bytes in "dest". the dest buffer will be filled backwards. finally,
+ * the start point of the resulting string is returned. this pointer is within
+ * dest, normally.
+ * in case the path buffer would overflow, the pointer is decremented further
+ * as if output was written to the buffer, though no more output is actually
+ * generated. that way, the caller can determine how much space would be
+ * required for the path to fit into the buffer. in that case, the returned
+ * value will be smaller than dest. callers must check this!
+ */
+char *btrfs_ref_to_path(struct btrfs_root *fs_root, struct btrfs_path *path,
+			u32 name_len, unsigned long name_off,
+			struct extent_buffer *eb_in, u64 parent,
+			char *dest, u32 size)
+{
+	int slot;
+	u64 next_inum;
+	int ret;
+	s64 bytes_left = ((s64)size) - 1;
+	struct extent_buffer *eb = eb_in;
+	struct btrfs_key found_key;
+	int leave_spinning = path->leave_spinning;
+	struct btrfs_inode_ref *iref;
+
+	if (bytes_left >= 0)
+		dest[bytes_left] = '\0';
+
+	path->leave_spinning = 1;
+	while (1) {
+		bytes_left -= name_len;
+		if (bytes_left >= 0)
+			read_extent_buffer(eb, dest + bytes_left,
+					   name_off, name_len);
+		if (eb != eb_in) {
+			btrfs_tree_read_unlock_blocking(eb);
+			free_extent_buffer(eb);
+		}
+		ret = inode_ref_info(parent, 0, fs_root, path, &found_key);
+		if (ret > 0)
+			ret = -ENOENT;
+		if (ret)
+			break;
+
+		next_inum = found_key.offset;
+
+		/* regular exit ahead */
+		if (parent == next_inum)
+			break;
+
+		slot = path->slots[0];
+		eb = path->nodes[0];
+		/* make sure we can use eb after releasing the path */
+		if (eb != eb_in) {
+			atomic_inc(&eb->refs);
+			btrfs_tree_read_lock(eb);
+			btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
+		}
+		btrfs_release_path(path);
+		iref = btrfs_item_ptr(eb, slot, struct btrfs_inode_ref);
+
+		name_len = btrfs_inode_ref_name_len(eb, iref);
+		name_off = (unsigned long)(iref + 1);
+
+		parent = next_inum;
+		--bytes_left;
+		if (bytes_left >= 0)
+			dest[bytes_left] = '/';
+	}
+
+	btrfs_release_path(path);
+	path->leave_spinning = leave_spinning;
+
+	if (ret)
+		return ERR_PTR(ret);
+
+	return dest + bytes_left;
+}
+
+/*
+ * this makes the path point to (logical EXTENT_ITEM *)
+ * returns BTRFS_EXTENT_FLAG_DATA for data, BTRFS_EXTENT_FLAG_TREE_BLOCK for
+ * tree blocks and <0 on error.
+ */
+int extent_from_logical(struct btrfs_fs_info *fs_info, u64 logical,
+			struct btrfs_path *path, struct btrfs_key *found_key,
+			u64 *flags_ret)
+{
+	int ret;
+	u64 flags;
+	u64 size = 0;
+	u32 item_size;
+	struct extent_buffer *eb;
+	struct btrfs_extent_item *ei;
+	struct btrfs_key key;
+
+	if (btrfs_fs_incompat(fs_info, SKINNY_METADATA))
+		key.type = BTRFS_METADATA_ITEM_KEY;
+	else
+		key.type = BTRFS_EXTENT_ITEM_KEY;
+	key.objectid = logical;
+	key.offset = (u64)-1;
+
+	ret = btrfs_search_slot(NULL, fs_info->extent_root, &key, path, 0, 0);
+	if (ret < 0)
+		return ret;
+
+	ret = btrfs_previous_extent_item(fs_info->extent_root, path, 0);
+	if (ret) {
+		if (ret > 0)
+			ret = -ENOENT;
+		return ret;
+	}
+	btrfs_item_key_to_cpu(path->nodes[0], found_key, path->slots[0]);
+	if (found_key->type == BTRFS_METADATA_ITEM_KEY)
+		size = fs_info->extent_root->leafsize;
+	else if (found_key->type == BTRFS_EXTENT_ITEM_KEY)
+		size = found_key->offset;
+
+	if (found_key->objectid > logical ||
+	    found_key->objectid + size <= logical) {
+		pr_debug("logical %llu is not within any extent\n", logical);
+		return -ENOENT;
+	}
+
+	eb = path->nodes[0];
+	item_size = btrfs_item_size_nr(eb, path->slots[0]);
+	BUG_ON(item_size < sizeof(*ei));
+
+	ei = btrfs_item_ptr(eb, path->slots[0], struct btrfs_extent_item);
+	flags = btrfs_extent_flags(eb, ei);
+
+	pr_debug("logical %llu is at position %llu within the extent (%llu "
+		 "EXTENT_ITEM %llu) flags %#llx size %u\n",
+		 logical, logical - found_key->objectid, found_key->objectid,
+		 found_key->offset, flags, item_size);
+
+	WARN_ON(!flags_ret);
+	if (flags_ret) {
+		if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK)
+			*flags_ret = BTRFS_EXTENT_FLAG_TREE_BLOCK;
+		else if (flags & BTRFS_EXTENT_FLAG_DATA)
+			*flags_ret = BTRFS_EXTENT_FLAG_DATA;
+		else
+			BUG_ON(1);
+		return 0;
+	}
+
+	return -EIO;
+}
+
+/*
+ * helper function to iterate extent inline refs. ptr must point to a 0 value
+ * for the first call and may be modified. it is used to track state.
+ * if more refs exist, 0 is returned and the next call to
+ * __get_extent_inline_ref must pass the modified ptr parameter to get the
+ * next ref. after the last ref was processed, 1 is returned.
+ * returns <0 on error
+ */
+static int __get_extent_inline_ref(unsigned long *ptr, struct extent_buffer *eb,
+				   struct btrfs_key *key,
+				   struct btrfs_extent_item *ei, u32 item_size,
+				   struct btrfs_extent_inline_ref **out_eiref,
+				   int *out_type)
+{
+	unsigned long end;
+	u64 flags;
+	struct btrfs_tree_block_info *info;
+
+	if (!*ptr) {
+		/* first call */
+		flags = btrfs_extent_flags(eb, ei);
+		if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
+			if (key->type == BTRFS_METADATA_ITEM_KEY) {
+				/* a skinny metadata extent */
+				*out_eiref =
+				     (struct btrfs_extent_inline_ref *)(ei + 1);
+			} else {
+				WARN_ON(key->type != BTRFS_EXTENT_ITEM_KEY);
+				info = (struct btrfs_tree_block_info *)(ei + 1);
+				*out_eiref =
+				   (struct btrfs_extent_inline_ref *)(info + 1);
+			}
+		} else {
+			*out_eiref = (struct btrfs_extent_inline_ref *)(ei + 1);
+		}
+		*ptr = (unsigned long)*out_eiref;
+		if ((unsigned long)(*ptr) >= (unsigned long)ei + item_size)
+			return -ENOENT;
+	}
+
+	end = (unsigned long)ei + item_size;
+	*out_eiref = (struct btrfs_extent_inline_ref *)(*ptr);
+	*out_type = btrfs_extent_inline_ref_type(eb, *out_eiref);
+
+	*ptr += btrfs_extent_inline_ref_size(*out_type);
+	WARN_ON(*ptr > end);
+	if (*ptr == end)
+		return 1; /* last */
+
+	return 0;
+}
+
+/*
+ * reads the tree block backref for an extent. tree level and root are returned
+ * through out_level and out_root. ptr must point to a 0 value for the first
+ * call and may be modified (see __get_extent_inline_ref comment).
+ * returns 0 if data was provided, 1 if there was no more data to provide or
+ * <0 on error.
+ */
+int tree_backref_for_extent(unsigned long *ptr, struct extent_buffer *eb,
+			    struct btrfs_key *key, struct btrfs_extent_item *ei,
+			    u32 item_size, u64 *out_root, u8 *out_level)
+{
+	int ret;
+	int type;
+	struct btrfs_tree_block_info *info;
+	struct btrfs_extent_inline_ref *eiref;
+
+	if (*ptr == (unsigned long)-1)
+		return 1;
+
+	while (1) {
+		ret = __get_extent_inline_ref(ptr, eb, key, ei, item_size,
+					      &eiref, &type);
+		if (ret < 0)
+			return ret;
+
+		if (type == BTRFS_TREE_BLOCK_REF_KEY ||
+		    type == BTRFS_SHARED_BLOCK_REF_KEY)
+			break;
+
+		if (ret == 1)
+			return 1;
+	}
+
+	/* we can treat both ref types equally here */
+	info = (struct btrfs_tree_block_info *)(ei + 1);
+	*out_root = btrfs_extent_inline_ref_offset(eb, eiref);
+	*out_level = btrfs_tree_block_level(eb, info);
+
+	if (ret == 1)
+		*ptr = (unsigned long)-1;
+
+	return 0;
+}
+
+static int iterate_leaf_refs(struct extent_inode_elem *inode_list,
+				u64 root, u64 extent_item_objectid,
+				iterate_extent_inodes_t *iterate, void *ctx)
+{
+	struct extent_inode_elem *eie;
+	int ret = 0;
+
+	for (eie = inode_list; eie; eie = eie->next) {
+		pr_debug("ref for %llu resolved, key (%llu EXTEND_DATA %llu), "
+			 "root %llu\n", extent_item_objectid,
+			 eie->inum, eie->offset, root);
+		ret = iterate(eie->inum, eie->offset, root, ctx);
+		if (ret) {
+			pr_debug("stopping iteration for %llu due to ret=%d\n",
+				 extent_item_objectid, ret);
+			break;
+		}
+	}
+
+	return ret;
+}
+
+/*
+ * calls iterate() for every inode that references the extent identified by
+ * the given parameters.
+ * when the iterator function returns a non-zero value, iteration stops.
+ */
+int iterate_extent_inodes(struct btrfs_fs_info *fs_info,
+				u64 extent_item_objectid, u64 extent_item_pos,
+				int search_commit_root,
+				iterate_extent_inodes_t *iterate, void *ctx)
+{
+	int ret;
+	struct btrfs_trans_handle *trans = NULL;
+	struct ulist *refs = NULL;
+	struct ulist *roots = NULL;
+	struct ulist_node *ref_node = NULL;
+	struct ulist_node *root_node = NULL;
+	struct seq_list tree_mod_seq_elem = {};
+	struct ulist_iterator ref_uiter;
+	struct ulist_iterator root_uiter;
+
+	pr_debug("resolving all inodes for extent %llu\n",
+			extent_item_objectid);
+
+	if (!search_commit_root) {
+		trans = btrfs_join_transaction(fs_info->extent_root);
+		if (IS_ERR(trans))
+			return PTR_ERR(trans);
+		btrfs_get_tree_mod_seq(fs_info, &tree_mod_seq_elem);
+	} else {
+		down_read(&fs_info->commit_root_sem);
+	}
+
+	ret = btrfs_find_all_leafs(trans, fs_info, extent_item_objectid,
+				   tree_mod_seq_elem.seq, &refs,
+				   &extent_item_pos);
+	if (ret)
+		goto out;
+
+	ULIST_ITER_INIT(&ref_uiter);
+	while (!ret && (ref_node = ulist_next(refs, &ref_uiter))) {
+		ret = __btrfs_find_all_roots(trans, fs_info, ref_node->val,
+					     tree_mod_seq_elem.seq, &roots);
+		if (ret)
+			break;
+		ULIST_ITER_INIT(&root_uiter);
+		while (!ret && (root_node = ulist_next(roots, &root_uiter))) {
+			pr_debug("root %llu references leaf %llu, data list "
+				 "%#llx\n", root_node->val, ref_node->val,
+				 ref_node->aux);
+			ret = iterate_leaf_refs((struct extent_inode_elem *)
+						(uintptr_t)ref_node->aux,
+						root_node->val,
+						extent_item_objectid,
+						iterate, ctx);
+		}
+		ulist_free(roots);
+	}
+
+	free_leaf_list(refs);
+out:
+	if (!search_commit_root) {
+		btrfs_put_tree_mod_seq(fs_info, &tree_mod_seq_elem);
+		btrfs_end_transaction(trans, fs_info->extent_root);
+	} else {
+		up_read(&fs_info->commit_root_sem);
+	}
+
+	return ret;
+}
+
+int iterate_inodes_from_logical(u64 logical, struct btrfs_fs_info *fs_info,
+				struct btrfs_path *path,
+				iterate_extent_inodes_t *iterate, void *ctx)
+{
+	int ret;
+	u64 extent_item_pos;
+	u64 flags = 0;
+	struct btrfs_key found_key;
+	int search_commit_root = path->search_commit_root;
+
+	ret = extent_from_logical(fs_info, logical, path, &found_key, &flags);
+	btrfs_release_path(path);
+	if (ret < 0)
+		return ret;
+	if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK)
+		return -EINVAL;
+
+	extent_item_pos = logical - found_key.objectid;
+	ret = iterate_extent_inodes(fs_info, found_key.objectid,
+					extent_item_pos, search_commit_root,
+					iterate, ctx);
+
+	return ret;
+}
+
+typedef int (iterate_irefs_t)(u64 parent, u32 name_len, unsigned long name_off,
+			      struct extent_buffer *eb, void *ctx);
+
+static int iterate_inode_refs(u64 inum, struct btrfs_root *fs_root,
+			      struct btrfs_path *path,
+			      iterate_irefs_t *iterate, void *ctx)
+{
+	int ret = 0;
+	int slot;
+	u32 cur;
+	u32 len;
+	u32 name_len;
+	u64 parent = 0;
+	int found = 0;
+	struct extent_buffer *eb;
+	struct btrfs_item *item;
+	struct btrfs_inode_ref *iref;
+	struct btrfs_key found_key;
+
+	while (!ret) {
+		ret = inode_ref_info(inum, parent ? parent+1 : 0, fs_root, path,
+				     &found_key);
+		if (ret < 0)
+			break;
+		if (ret) {
+			ret = found ? 0 : -ENOENT;
+			break;
+		}
+		++found;
+
+		parent = found_key.offset;
+		slot = path->slots[0];
+		eb = btrfs_clone_extent_buffer(path->nodes[0]);
+		if (!eb) {
+			ret = -ENOMEM;
+			break;
+		}
+		extent_buffer_get(eb);
+		btrfs_tree_read_lock(eb);
+		btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
+		btrfs_release_path(path);
+
+		item = btrfs_item_nr(slot);
+		iref = btrfs_item_ptr(eb, slot, struct btrfs_inode_ref);
+
+		for (cur = 0; cur < btrfs_item_size(eb, item); cur += len) {
+			name_len = btrfs_inode_ref_name_len(eb, iref);
+			/* path must be released before calling iterate()! */
+			pr_debug("following ref at offset %u for inode %llu in "
+				 "tree %llu\n", cur, found_key.objectid,
+				 fs_root->objectid);
+			ret = iterate(parent, name_len,
+				      (unsigned long)(iref + 1), eb, ctx);
+			if (ret)
+				break;
+			len = sizeof(*iref) + name_len;
+			iref = (struct btrfs_inode_ref *)((char *)iref + len);
+		}
+		btrfs_tree_read_unlock_blocking(eb);
+		free_extent_buffer(eb);
+	}
+
+	btrfs_release_path(path);
+
+	return ret;
+}
+
+static int iterate_inode_extrefs(u64 inum, struct btrfs_root *fs_root,
+				 struct btrfs_path *path,
+				 iterate_irefs_t *iterate, void *ctx)
+{
+	int ret;
+	int slot;
+	u64 offset = 0;
+	u64 parent;
+	int found = 0;
+	struct extent_buffer *eb;
+	struct btrfs_inode_extref *extref;
+	struct extent_buffer *leaf;
+	u32 item_size;
+	u32 cur_offset;
+	unsigned long ptr;
+
+	while (1) {
+		ret = btrfs_find_one_extref(fs_root, inum, offset, path, &extref,
+					    &offset);
+		if (ret < 0)
+			break;
+		if (ret) {
+			ret = found ? 0 : -ENOENT;
+			break;
+		}
+		++found;
+
+		slot = path->slots[0];
+		eb = btrfs_clone_extent_buffer(path->nodes[0]);
+		if (!eb) {
+			ret = -ENOMEM;
+			break;
+		}
+		extent_buffer_get(eb);
+
+		btrfs_tree_read_lock(eb);
+		btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
+		btrfs_release_path(path);
+
+		leaf = path->nodes[0];
+		item_size = btrfs_item_size_nr(leaf, slot);
+		ptr = btrfs_item_ptr_offset(leaf, slot);
+		cur_offset = 0;
+
+		while (cur_offset < item_size) {
+			u32 name_len;
+
+			extref = (struct btrfs_inode_extref *)(ptr + cur_offset);
+			parent = btrfs_inode_extref_parent(eb, extref);
+			name_len = btrfs_inode_extref_name_len(eb, extref);
+			ret = iterate(parent, name_len,
+				      (unsigned long)&extref->name, eb, ctx);
+			if (ret)
+				break;
+
+			cur_offset += btrfs_inode_extref_name_len(leaf, extref);
+			cur_offset += sizeof(*extref);
+		}
+		btrfs_tree_read_unlock_blocking(eb);
+		free_extent_buffer(eb);
+
+		offset++;
+	}
+
+	btrfs_release_path(path);
+
+	return ret;
+}
+
+static int iterate_irefs(u64 inum, struct btrfs_root *fs_root,
+			 struct btrfs_path *path, iterate_irefs_t *iterate,
+			 void *ctx)
+{
+	int ret;
+	int found_refs = 0;
+
+	ret = iterate_inode_refs(inum, fs_root, path, iterate, ctx);
+	if (!ret)
+		++found_refs;
+	else if (ret != -ENOENT)
+		return ret;
+
+	ret = iterate_inode_extrefs(inum, fs_root, path, iterate, ctx);
+	if (ret == -ENOENT && found_refs)
+		return 0;
+
+	return ret;
+}
+
+/*
+ * returns 0 if the path could be dumped (probably truncated)
+ * returns <0 in case of an error
+ */
+static int inode_to_path(u64 inum, u32 name_len, unsigned long name_off,
+			 struct extent_buffer *eb, void *ctx)
+{
+	struct inode_fs_paths *ipath = ctx;
+	char *fspath;
+	char *fspath_min;
+	int i = ipath->fspath->elem_cnt;
+	const int s_ptr = sizeof(char *);
+	u32 bytes_left;
+
+	bytes_left = ipath->fspath->bytes_left > s_ptr ?
+					ipath->fspath->bytes_left - s_ptr : 0;
+
+	fspath_min = (char *)ipath->fspath->val + (i + 1) * s_ptr;
+	fspath = btrfs_ref_to_path(ipath->fs_root, ipath->btrfs_path, name_len,
+				   name_off, eb, inum, fspath_min, bytes_left);
+	if (IS_ERR(fspath))
+		return PTR_ERR(fspath);
+
+	if (fspath > fspath_min) {
+		ipath->fspath->val[i] = (u64)(unsigned long)fspath;
+		++ipath->fspath->elem_cnt;
+		ipath->fspath->bytes_left = fspath - fspath_min;
+	} else {
+		++ipath->fspath->elem_missed;
+		ipath->fspath->bytes_missing += fspath_min - fspath;
+		ipath->fspath->bytes_left = 0;
+	}
+
+	return 0;
+}
+
+/*
+ * this dumps all file system paths to the inode into the ipath struct, provided
+ * is has been created large enough. each path is zero-terminated and accessed
+ * from ipath->fspath->val[i].
+ * when it returns, there are ipath->fspath->elem_cnt number of paths available
+ * in ipath->fspath->val[]. when the allocated space wasn't sufficient, the
+ * number of missed paths in recored in ipath->fspath->elem_missed, otherwise,
+ * it's zero. ipath->fspath->bytes_missing holds the number of bytes that would
+ * have been needed to return all paths.
+ */
+int paths_from_inode(u64 inum, struct inode_fs_paths *ipath)
+{
+	return iterate_irefs(inum, ipath->fs_root, ipath->btrfs_path,
+			     inode_to_path, ipath);
+}
+
+struct btrfs_data_container *init_data_container(u32 total_bytes)
+{
+	struct btrfs_data_container *data;
+	size_t alloc_bytes;
+
+	alloc_bytes = max_t(size_t, total_bytes, sizeof(*data));
+	data = vmalloc(alloc_bytes);
+	if (!data)
+		return ERR_PTR(-ENOMEM);
+
+	if (total_bytes >= sizeof(*data)) {
+		data->bytes_left = total_bytes - sizeof(*data);
+		data->bytes_missing = 0;
+	} else {
+		data->bytes_missing = sizeof(*data) - total_bytes;
+		data->bytes_left = 0;
+	}
+
+	data->elem_cnt = 0;
+	data->elem_missed = 0;
+
+	return data;
+}
+
+/*
+ * allocates space to return multiple file system paths for an inode.
+ * total_bytes to allocate are passed, note that space usable for actual path
+ * information will be total_bytes - sizeof(struct inode_fs_paths).
+ * the returned pointer must be freed with free_ipath() in the end.
+ */
+struct inode_fs_paths *init_ipath(s32 total_bytes, struct btrfs_root *fs_root,
+					struct btrfs_path *path)
+{
+	struct inode_fs_paths *ifp;
+	struct btrfs_data_container *fspath;
+
+	fspath = init_data_container(total_bytes);
+	if (IS_ERR(fspath))
+		return (void *)fspath;
+
+	ifp = kmalloc(sizeof(*ifp), GFP_NOFS);
+	if (!ifp) {
+		kfree(fspath);
+		return ERR_PTR(-ENOMEM);
+	}
+
+	ifp->btrfs_path = path;
+	ifp->fspath = fspath;
+	ifp->fs_root = fs_root;
+
+	return ifp;
+}
+
+void free_ipath(struct inode_fs_paths *ipath)
+{
+	if (!ipath)
+		return;
+	vfree(ipath->fspath);
+	kfree(ipath);
+}
diff --git a/fs/btrfs/backref.h b/fs/btrfs/backref.h
new file mode 100644
index 00000000000..86fc20fec28
--- /dev/null
+++ b/fs/btrfs/backref.h
@@ -0,0 +1,77 @@
+/*
+ * Copyright (C) 2011 STRATO.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef __BTRFS_BACKREF__
+#define __BTRFS_BACKREF__
+
+#include <linux/btrfs.h>
+#include "ulist.h"
+#include "extent_io.h"
+
+struct inode_fs_paths {
+	struct btrfs_path		*btrfs_path;
+	struct btrfs_root		*fs_root;
+	struct btrfs_data_container	*fspath;
+};
+
+typedef int (iterate_extent_inodes_t)(u64 inum, u64 offset, u64 root,
+		void *ctx);
+
+int inode_item_info(u64 inum, u64 ioff, struct btrfs_root *fs_root,
+			struct btrfs_path *path);
+
+int extent_from_logical(struct btrfs_fs_info *fs_info, u64 logical,
+			struct btrfs_path *path, struct btrfs_key *found_key,
+			u64 *flags);
+
+int tree_backref_for_extent(unsigned long *ptr, struct extent_buffer *eb,
+			    struct btrfs_key *key, struct btrfs_extent_item *ei,
+			    u32 item_size, u64 *out_root, u8 *out_level);
+
+int iterate_extent_inodes(struct btrfs_fs_info *fs_info,
+				u64 extent_item_objectid,
+				u64 extent_offset, int search_commit_root,
+				iterate_extent_inodes_t *iterate, void *ctx);
+
+int iterate_inodes_from_logical(u64 logical, struct btrfs_fs_info *fs_info,
+				struct btrfs_path *path,
+				iterate_extent_inodes_t *iterate, void *ctx);
+
+int paths_from_inode(u64 inum, struct inode_fs_paths *ipath);
+
+int btrfs_find_all_roots(struct btrfs_trans_handle *trans,
+			 struct btrfs_fs_info *fs_info, u64 bytenr,
+			 u64 time_seq, struct ulist **roots);
+char *btrfs_ref_to_path(struct btrfs_root *fs_root, struct btrfs_path *path,
+			u32 name_len, unsigned long name_off,
+			struct extent_buffer *eb_in, u64 parent,
+			char *dest, u32 size);
+
+struct btrfs_data_container *init_data_container(u32 total_bytes);
+struct inode_fs_paths *init_ipath(s32 total_bytes, struct btrfs_root *fs_root,
+					struct btrfs_path *path);
+void free_ipath(struct inode_fs_paths *ipath);
+
+int btrfs_find_one_extref(struct btrfs_root *root, u64 inode_objectid,
+			  u64 start_off, struct btrfs_path *path,
+			  struct btrfs_inode_extref **ret_extref,
+			  u64 *found_off);
+
+int __init btrfs_prelim_ref_init(void);
+void btrfs_prelim_ref_exit(void);
+#endif
diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h
index 7a4dee19983..4794923c410 100644
--- a/fs/btrfs/btrfs_inode.h
+++ b/fs/btrfs/btrfs_inode.h
@@ -19,9 +19,31 @@
 #ifndef __BTRFS_I__
 #define __BTRFS_I__
 
+#include <linux/hash.h>
 #include "extent_map.h"
 #include "extent_io.h"
 #include "ordered-data.h"
+#include "delayed-inode.h"
+
+/*
+ * ordered_data_close is set by truncate when a file that used
+ * to have good data has been truncated to zero.  When it is set
+ * the btrfs file release call will add this inode to the
+ * ordered operations list so that we make sure to flush out any
+ * new data the application may have written before commit.
+ */
+#define BTRFS_INODE_ORDERED_DATA_CLOSE		0
+#define BTRFS_INODE_ORPHAN_META_RESERVED	1
+#define BTRFS_INODE_DUMMY			2
+#define BTRFS_INODE_IN_DEFRAG			3
+#define BTRFS_INODE_DELALLOC_META_RESERVED	4
+#define BTRFS_INODE_HAS_ORPHAN_ITEM		5
+#define BTRFS_INODE_HAS_ASYNC_EXTENT		6
+#define BTRFS_INODE_NEEDS_FULL_SYNC		7
+#define BTRFS_INODE_COPY_EVERYTHING		8
+#define BTRFS_INODE_IN_DELALLOC_LIST		9
+#define BTRFS_INODE_READDIO_NEED_LOCK		10
+#define BTRFS_INODE_HAS_PROPS		        11
 
 /* in memory btrfs inode */
 struct btrfs_inode {
@@ -33,6 +55,9 @@ struct btrfs_inode {
 	 */
 	struct btrfs_key location;
 
+	/* Lock for counters */
+	spinlock_t lock;
+
 	/* the extent_tree has caches of all the extent mappings to disk */
 	struct extent_map_tree extent_tree;
 
@@ -47,12 +72,12 @@ struct btrfs_inode {
 	/* held while logging the inode in tree-log.c */
 	struct mutex log_mutex;
 
+	/* held while doing delalloc reservations */
+	struct mutex delalloc_mutex;
+
 	/* used to order data wrt metadata */
 	struct btrfs_ordered_inode_tree ordered_tree;
 
-	/* for keeping track of orphaned inodes */
-	struct list_head i_orphan;
-
 	/* list of all the delalloc inodes in the FS.  There are times we need
 	 * to write all the delalloc pages to disk, and this list is used
 	 * to walk them all.
@@ -68,42 +93,39 @@ struct btrfs_inode {
 	/* node for the red-black tree that links inodes in subvolume root */
 	struct rb_node rb_node;
 
-	/* the space_info for where this inode's data allocations are done */
-	struct btrfs_space_info *space_info;
+	unsigned long runtime_flags;
+
+	/* Keep track of who's O_SYNC/fsyncing currently */
+	atomic_t sync_writers;
 
 	/* full 64 bit generation number, struct vfs_inode doesn't have a big
 	 * enough field for this.
 	 */
 	u64 generation;
 
-	/* sequence number for NFS changes */
-	u64 sequence;
-
 	/*
 	 * transid of the trans_handle that last modified this inode
 	 */
 	u64 last_trans;
 
 	/*
-	 * log transid when this inode was last modified
+	 * transid that last logged this inode
 	 */
-	u64 last_sub_trans;
+	u64 logged_trans;
 
 	/*
-	 * transid that last logged this inode
+	 * log transid when this inode was last modified
 	 */
-	u64 logged_trans;
+	int last_sub_trans;
+
+	/* a local copy of root's last_log_commit */
+	int last_log_commit;
 
 	/* total number of bytes pending delalloc, used by stat to calc the
 	 * real block usage of the file
 	 */
 	u64 delalloc_bytes;
 
-	/* total number of bytes that may be used for this inode for
-	 * delalloc
-	 */
-	u64 reserved_bytes;
-
 	/*
 	 * the size of the file stored in the metadata on disk.  data=ordered
 	 * means the in-memory i_size might be larger than the size on disk
@@ -111,17 +133,14 @@ struct btrfs_inode {
 	 */
 	u64 disk_i_size;
 
-	/* flags field from the on disk inode */
-	u32 flags;
-
 	/*
 	 * if this is a directory then index_cnt is the counter for the index
 	 * number for new files that are created
 	 */
 	u64 index_cnt;
 
-	/* the start of block group preferred for allocations. */
-	u64 block_group;
+	/* Cache the directory index number to speed the dir/file remove */
+	u64 dir_index;
 
 	/* the fsync log has some corner cases that mean we have to check
 	 * directories to see if any unlinks have been done before
@@ -131,45 +150,140 @@ struct btrfs_inode {
 	u64 last_unlink_trans;
 
 	/*
+	 * Number of bytes outstanding that are going to need csums.  This is
+	 * used in ENOSPC accounting.
+	 */
+	u64 csum_bytes;
+
+	/* flags field from the on disk inode */
+	u32 flags;
+
+	/*
 	 * Counters to keep track of the number of extent item's we may use due
 	 * to delalloc and such.  outstanding_extents is the number of extent
 	 * items we think we'll end up using, and reserved_extents is the number
 	 * of extent items we've reserved metadata for.
 	 */
-	spinlock_t accounting_lock;
-	int reserved_extents;
-	int outstanding_extents;
-
-	/*
-	 * ordered_data_close is set by truncate when a file that used
-	 * to have good data has been truncated to zero.  When it is set
-	 * the btrfs file release call will add this inode to the
-	 * ordered operations list so that we make sure to flush out any
-	 * new data the application may have written before commit.
-	 *
-	 * yes, its silly to have a single bitflag, but we might grow more
-	 * of these.
-	 */
-	unsigned ordered_data_close:1;
-	unsigned dummy_inode:1;
+	unsigned outstanding_extents;
+	unsigned reserved_extents;
 
 	/*
 	 * always compress this one file
 	 */
-	unsigned force_compress:1;
+	unsigned force_compress;
+
+	struct btrfs_delayed_node *delayed_node;
 
 	struct inode vfs_inode;
 };
 
+extern unsigned char btrfs_filetype_table[];
+
 static inline struct btrfs_inode *BTRFS_I(struct inode *inode)
 {
 	return container_of(inode, struct btrfs_inode, vfs_inode);
 }
 
+static inline unsigned long btrfs_inode_hash(u64 objectid,
+					     const struct btrfs_root *root)
+{
+	u64 h = objectid ^ (root->objectid * GOLDEN_RATIO_PRIME);
+
+#if BITS_PER_LONG == 32
+	h = (h >> 32) ^ (h & 0xffffffff);
+#endif
+
+	return (unsigned long)h;
+}
+
+static inline void btrfs_insert_inode_hash(struct inode *inode)
+{
+	unsigned long h = btrfs_inode_hash(inode->i_ino, BTRFS_I(inode)->root);
+
+	__insert_inode_hash(inode, h);
+}
+
+static inline u64 btrfs_ino(struct inode *inode)
+{
+	u64 ino = BTRFS_I(inode)->location.objectid;
+
+	/*
+	 * !ino: btree_inode
+	 * type == BTRFS_ROOT_ITEM_KEY: subvol dir
+	 */
+	if (!ino || BTRFS_I(inode)->location.type == BTRFS_ROOT_ITEM_KEY)
+		ino = inode->i_ino;
+	return ino;
+}
+
 static inline void btrfs_i_size_write(struct inode *inode, u64 size)
 {
 	i_size_write(inode, size);
 	BTRFS_I(inode)->disk_i_size = size;
 }
 
+static inline bool btrfs_is_free_space_inode(struct inode *inode)
+{
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+
+	if (root == root->fs_info->tree_root &&
+	    btrfs_ino(inode) != BTRFS_BTREE_INODE_OBJECTID)
+		return true;
+	if (BTRFS_I(inode)->location.objectid == BTRFS_FREE_INO_OBJECTID)
+		return true;
+	return false;
+}
+
+static inline int btrfs_inode_in_log(struct inode *inode, u64 generation)
+{
+	if (BTRFS_I(inode)->logged_trans == generation &&
+	    BTRFS_I(inode)->last_sub_trans <=
+	    BTRFS_I(inode)->last_log_commit &&
+	    BTRFS_I(inode)->last_sub_trans <=
+	    BTRFS_I(inode)->root->last_log_commit)
+		return 1;
+	return 0;
+}
+
+struct btrfs_dio_private {
+	struct inode *inode;
+	u64 logical_offset;
+	u64 disk_bytenr;
+	u64 bytes;
+	void *private;
+
+	/* number of bios pending for this dio */
+	atomic_t pending_bios;
+
+	/* IO errors */
+	int errors;
+
+	/* orig_bio is our btrfs_io_bio */
+	struct bio *orig_bio;
+
+	/* dio_bio came from fs/direct-io.c */
+	struct bio *dio_bio;
+	u8 csum[0];
+};
+
+/*
+ * Disable DIO read nolock optimization, so new dio readers will be forced
+ * to grab i_mutex. It is used to avoid the endless truncate due to
+ * nonlocked dio read.
+ */
+static inline void btrfs_inode_block_unlocked_dio(struct inode *inode)
+{
+	set_bit(BTRFS_INODE_READDIO_NEED_LOCK, &BTRFS_I(inode)->runtime_flags);
+	smp_mb();
+}
+
+static inline void btrfs_inode_resume_unlocked_dio(struct inode *inode)
+{
+	smp_mb__before_atomic();
+	clear_bit(BTRFS_INODE_READDIO_NEED_LOCK,
+		  &BTRFS_I(inode)->runtime_flags);
+}
+
+bool btrfs_page_exists_in_range(struct inode *inode, loff_t start, loff_t end);
+
 #endif
diff --git a/fs/btrfs/check-integrity.c b/fs/btrfs/check-integrity.c
new file mode 100644
index 00000000000..ce92ae30250
--- /dev/null
+++ b/fs/btrfs/check-integrity.c
@@ -0,0 +1,3295 @@
+/*
+ * Copyright (C) STRATO AG 2011.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+/*
+ * This module can be used to catch cases when the btrfs kernel
+ * code executes write requests to the disk that bring the file
+ * system in an inconsistent state. In such a state, a power-loss
+ * or kernel panic event would cause that the data on disk is
+ * lost or at least damaged.
+ *
+ * Code is added that examines all block write requests during
+ * runtime (including writes of the super block). Three rules
+ * are verified and an error is printed on violation of the
+ * rules:
+ * 1. It is not allowed to write a disk block which is
+ *    currently referenced by the super block (either directly
+ *    or indirectly).
+ * 2. When a super block is written, it is verified that all
+ *    referenced (directly or indirectly) blocks fulfill the
+ *    following requirements:
+ *    2a. All referenced blocks have either been present when
+ *        the file system was mounted, (i.e., they have been
+ *        referenced by the super block) or they have been
+ *        written since then and the write completion callback
+ *        was called and no write error was indicated and a
+ *        FLUSH request to the device where these blocks are
+ *        located was received and completed.
+ *    2b. All referenced blocks need to have a generation
+ *        number which is equal to the parent's number.
+ *
+ * One issue that was found using this module was that the log
+ * tree on disk became temporarily corrupted because disk blocks
+ * that had been in use for the log tree had been freed and
+ * reused too early, while being referenced by the written super
+ * block.
+ *
+ * The search term in the kernel log that can be used to filter
+ * on the existence of detected integrity issues is
+ * "btrfs: attempt".
+ *
+ * The integrity check is enabled via mount options. These
+ * mount options are only supported if the integrity check
+ * tool is compiled by defining BTRFS_FS_CHECK_INTEGRITY.
+ *
+ * Example #1, apply integrity checks to all metadata:
+ * mount /dev/sdb1 /mnt -o check_int
+ *
+ * Example #2, apply integrity checks to all metadata and
+ * to data extents:
+ * mount /dev/sdb1 /mnt -o check_int_data
+ *
+ * Example #3, apply integrity checks to all metadata and dump
+ * the tree that the super block references to kernel messages
+ * each time after a super block was written:
+ * mount /dev/sdb1 /mnt -o check_int,check_int_print_mask=263
+ *
+ * If the integrity check tool is included and activated in
+ * the mount options, plenty of kernel memory is used, and
+ * plenty of additional CPU cycles are spent. Enabling this
+ * functionality is not intended for normal use. In most
+ * cases, unless you are a btrfs developer who needs to verify
+ * the integrity of (super)-block write requests, do not
+ * enable the config option BTRFS_FS_CHECK_INTEGRITY to
+ * include and compile the integrity check tool.
+ *
+ * Expect millions of lines of information in the kernel log with an
+ * enabled check_int_print_mask. Therefore set LOG_BUF_SHIFT in the
+ * kernel config to at least 26 (which is 64MB). Usually the value is
+ * limited to 21 (which is 2MB) in init/Kconfig. The file needs to be
+ * changed like this before LOG_BUF_SHIFT can be set to a high value:
+ * config LOG_BUF_SHIFT
+ *       int "Kernel log buffer size (16 => 64KB, 17 => 128KB)"
+ *       range 12 30
+ */
+
+#include <linux/sched.h>
+#include <linux/slab.h>
+#include <linux/buffer_head.h>
+#include <linux/mutex.h>
+#include <linux/genhd.h>
+#include <linux/blkdev.h>
+#include "ctree.h"
+#include "disk-io.h"
+#include "hash.h"
+#include "transaction.h"
+#include "extent_io.h"
+#include "volumes.h"
+#include "print-tree.h"
+#include "locking.h"
+#include "check-integrity.h"
+#include "rcu-string.h"
+
+#define BTRFSIC_BLOCK_HASHTABLE_SIZE 0x10000
+#define BTRFSIC_BLOCK_LINK_HASHTABLE_SIZE 0x10000
+#define BTRFSIC_DEV2STATE_HASHTABLE_SIZE 0x100
+#define BTRFSIC_BLOCK_MAGIC_NUMBER 0x14491051
+#define BTRFSIC_BLOCK_LINK_MAGIC_NUMBER 0x11070807
+#define BTRFSIC_DEV2STATE_MAGIC_NUMBER 0x20111530
+#define BTRFSIC_BLOCK_STACK_FRAME_MAGIC_NUMBER 20111300
+#define BTRFSIC_TREE_DUMP_MAX_INDENT_LEVEL (200 - 6)	/* in characters,
+							 * excluding " [...]" */
+#define BTRFSIC_GENERATION_UNKNOWN ((u64)-1)
+
+/*
+ * The definition of the bitmask fields for the print_mask.
+ * They are specified with the mount option check_integrity_print_mask.
+ */
+#define BTRFSIC_PRINT_MASK_SUPERBLOCK_WRITE			0x00000001
+#define BTRFSIC_PRINT_MASK_ROOT_CHUNK_LOG_TREE_LOCATION		0x00000002
+#define BTRFSIC_PRINT_MASK_TREE_AFTER_SB_WRITE			0x00000004
+#define BTRFSIC_PRINT_MASK_TREE_BEFORE_SB_WRITE			0x00000008
+#define BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH			0x00000010
+#define BTRFSIC_PRINT_MASK_END_IO_BIO_BH			0x00000020
+#define BTRFSIC_PRINT_MASK_VERBOSE				0x00000040
+#define BTRFSIC_PRINT_MASK_VERY_VERBOSE				0x00000080
+#define BTRFSIC_PRINT_MASK_INITIAL_TREE				0x00000100
+#define BTRFSIC_PRINT_MASK_INITIAL_ALL_TREES			0x00000200
+#define BTRFSIC_PRINT_MASK_INITIAL_DATABASE			0x00000400
+#define BTRFSIC_PRINT_MASK_NUM_COPIES				0x00000800
+#define BTRFSIC_PRINT_MASK_TREE_WITH_ALL_MIRRORS		0x00001000
+#define BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH_VERBOSE		0x00002000
+
+struct btrfsic_dev_state;
+struct btrfsic_state;
+
+struct btrfsic_block {
+	u32 magic_num;		/* only used for debug purposes */
+	unsigned int is_metadata:1;	/* if it is meta-data, not data-data */
+	unsigned int is_superblock:1;	/* if it is one of the superblocks */
+	unsigned int is_iodone:1;	/* if is done by lower subsystem */
+	unsigned int iodone_w_error:1;	/* error was indicated to endio */
+	unsigned int never_written:1;	/* block was added because it was
+					 * referenced, not because it was
+					 * written */
+	unsigned int mirror_num;	/* large enough to hold
+					 * BTRFS_SUPER_MIRROR_MAX */
+	struct btrfsic_dev_state *dev_state;
+	u64 dev_bytenr;		/* key, physical byte num on disk */
+	u64 logical_bytenr;	/* logical byte num on disk */
+	u64 generation;
+	struct btrfs_disk_key disk_key;	/* extra info to print in case of
+					 * issues, will not always be correct */
+	struct list_head collision_resolving_node;	/* list node */
+	struct list_head all_blocks_node;	/* list node */
+
+	/* the following two lists contain block_link items */
+	struct list_head ref_to_list;	/* list */
+	struct list_head ref_from_list;	/* list */
+	struct btrfsic_block *next_in_same_bio;
+	void *orig_bio_bh_private;
+	union {
+		bio_end_io_t *bio;
+		bh_end_io_t *bh;
+	} orig_bio_bh_end_io;
+	int submit_bio_bh_rw;
+	u64 flush_gen; /* only valid if !never_written */
+};
+
+/*
+ * Elements of this type are allocated dynamically and required because
+ * each block object can refer to and can be ref from multiple blocks.
+ * The key to lookup them in the hashtable is the dev_bytenr of
+ * the block ref to plus the one from the block refered from.
+ * The fact that they are searchable via a hashtable and that a
+ * ref_cnt is maintained is not required for the btrfs integrity
+ * check algorithm itself, it is only used to make the output more
+ * beautiful in case that an error is detected (an error is defined
+ * as a write operation to a block while that block is still referenced).
+ */
+struct btrfsic_block_link {
+	u32 magic_num;		/* only used for debug purposes */
+	u32 ref_cnt;
+	struct list_head node_ref_to;	/* list node */
+	struct list_head node_ref_from;	/* list node */
+	struct list_head collision_resolving_node;	/* list node */
+	struct btrfsic_block *block_ref_to;
+	struct btrfsic_block *block_ref_from;
+	u64 parent_generation;
+};
+
+struct btrfsic_dev_state {
+	u32 magic_num;		/* only used for debug purposes */
+	struct block_device *bdev;
+	struct btrfsic_state *state;
+	struct list_head collision_resolving_node;	/* list node */
+	struct btrfsic_block dummy_block_for_bio_bh_flush;
+	u64 last_flush_gen;
+	char name[BDEVNAME_SIZE];
+};
+
+struct btrfsic_block_hashtable {
+	struct list_head table[BTRFSIC_BLOCK_HASHTABLE_SIZE];
+};
+
+struct btrfsic_block_link_hashtable {
+	struct list_head table[BTRFSIC_BLOCK_LINK_HASHTABLE_SIZE];
+};
+
+struct btrfsic_dev_state_hashtable {
+	struct list_head table[BTRFSIC_DEV2STATE_HASHTABLE_SIZE];
+};
+
+struct btrfsic_block_data_ctx {
+	u64 start;		/* virtual bytenr */
+	u64 dev_bytenr;		/* physical bytenr on device */
+	u32 len;
+	struct btrfsic_dev_state *dev;
+	char **datav;
+	struct page **pagev;
+	void *mem_to_free;
+};
+
+/* This structure is used to implement recursion without occupying
+ * any stack space, refer to btrfsic_process_metablock() */
+struct btrfsic_stack_frame {
+	u32 magic;
+	u32 nr;
+	int error;
+	int i;
+	int limit_nesting;
+	int num_copies;
+	int mirror_num;
+	struct btrfsic_block *block;
+	struct btrfsic_block_data_ctx *block_ctx;
+	struct btrfsic_block *next_block;
+	struct btrfsic_block_data_ctx next_block_ctx;
+	struct btrfs_header *hdr;
+	struct btrfsic_stack_frame *prev;
+};
+
+/* Some state per mounted filesystem */
+struct btrfsic_state {
+	u32 print_mask;
+	int include_extent_data;
+	int csum_size;
+	struct list_head all_blocks_list;
+	struct btrfsic_block_hashtable block_hashtable;
+	struct btrfsic_block_link_hashtable block_link_hashtable;
+	struct btrfs_root *root;
+	u64 max_superblock_generation;
+	struct btrfsic_block *latest_superblock;
+	u32 metablock_size;
+	u32 datablock_size;
+};
+
+static void btrfsic_block_init(struct btrfsic_block *b);
+static struct btrfsic_block *btrfsic_block_alloc(void);
+static void btrfsic_block_free(struct btrfsic_block *b);
+static void btrfsic_block_link_init(struct btrfsic_block_link *n);
+static struct btrfsic_block_link *btrfsic_block_link_alloc(void);
+static void btrfsic_block_link_free(struct btrfsic_block_link *n);
+static void btrfsic_dev_state_init(struct btrfsic_dev_state *ds);
+static struct btrfsic_dev_state *btrfsic_dev_state_alloc(void);
+static void btrfsic_dev_state_free(struct btrfsic_dev_state *ds);
+static void btrfsic_block_hashtable_init(struct btrfsic_block_hashtable *h);
+static void btrfsic_block_hashtable_add(struct btrfsic_block *b,
+					struct btrfsic_block_hashtable *h);
+static void btrfsic_block_hashtable_remove(struct btrfsic_block *b);
+static struct btrfsic_block *btrfsic_block_hashtable_lookup(
+		struct block_device *bdev,
+		u64 dev_bytenr,
+		struct btrfsic_block_hashtable *h);
+static void btrfsic_block_link_hashtable_init(
+		struct btrfsic_block_link_hashtable *h);
+static void btrfsic_block_link_hashtable_add(
+		struct btrfsic_block_link *l,
+		struct btrfsic_block_link_hashtable *h);
+static void btrfsic_block_link_hashtable_remove(struct btrfsic_block_link *l);
+static struct btrfsic_block_link *btrfsic_block_link_hashtable_lookup(
+		struct block_device *bdev_ref_to,
+		u64 dev_bytenr_ref_to,
+		struct block_device *bdev_ref_from,
+		u64 dev_bytenr_ref_from,
+		struct btrfsic_block_link_hashtable *h);
+static void btrfsic_dev_state_hashtable_init(
+		struct btrfsic_dev_state_hashtable *h);
+static void btrfsic_dev_state_hashtable_add(
+		struct btrfsic_dev_state *ds,
+		struct btrfsic_dev_state_hashtable *h);
+static void btrfsic_dev_state_hashtable_remove(struct btrfsic_dev_state *ds);
+static struct btrfsic_dev_state *btrfsic_dev_state_hashtable_lookup(
+		struct block_device *bdev,
+		struct btrfsic_dev_state_hashtable *h);
+static struct btrfsic_stack_frame *btrfsic_stack_frame_alloc(void);
+static void btrfsic_stack_frame_free(struct btrfsic_stack_frame *sf);
+static int btrfsic_process_superblock(struct btrfsic_state *state,
+				      struct btrfs_fs_devices *fs_devices);
+static int btrfsic_process_metablock(struct btrfsic_state *state,
+				     struct btrfsic_block *block,
+				     struct btrfsic_block_data_ctx *block_ctx,
+				     int limit_nesting, int force_iodone_flag);
+static void btrfsic_read_from_block_data(
+	struct btrfsic_block_data_ctx *block_ctx,
+	void *dst, u32 offset, size_t len);
+static int btrfsic_create_link_to_next_block(
+		struct btrfsic_state *state,
+		struct btrfsic_block *block,
+		struct btrfsic_block_data_ctx
+		*block_ctx, u64 next_bytenr,
+		int limit_nesting,
+		struct btrfsic_block_data_ctx *next_block_ctx,
+		struct btrfsic_block **next_blockp,
+		int force_iodone_flag,
+		int *num_copiesp, int *mirror_nump,
+		struct btrfs_disk_key *disk_key,
+		u64 parent_generation);
+static int btrfsic_handle_extent_data(struct btrfsic_state *state,
+				      struct btrfsic_block *block,
+				      struct btrfsic_block_data_ctx *block_ctx,
+				      u32 item_offset, int force_iodone_flag);
+static int btrfsic_map_block(struct btrfsic_state *state, u64 bytenr, u32 len,
+			     struct btrfsic_block_data_ctx *block_ctx_out,
+			     int mirror_num);
+static int btrfsic_map_superblock(struct btrfsic_state *state, u64 bytenr,
+				  u32 len, struct block_device *bdev,
+				  struct btrfsic_block_data_ctx *block_ctx_out);
+static void btrfsic_release_block_ctx(struct btrfsic_block_data_ctx *block_ctx);
+static int btrfsic_read_block(struct btrfsic_state *state,
+			      struct btrfsic_block_data_ctx *block_ctx);
+static void btrfsic_dump_database(struct btrfsic_state *state);
+static int btrfsic_test_for_metadata(struct btrfsic_state *state,
+				     char **datav, unsigned int num_pages);
+static void btrfsic_process_written_block(struct btrfsic_dev_state *dev_state,
+					  u64 dev_bytenr, char **mapped_datav,
+					  unsigned int num_pages,
+					  struct bio *bio, int *bio_is_patched,
+					  struct buffer_head *bh,
+					  int submit_bio_bh_rw);
+static int btrfsic_process_written_superblock(
+		struct btrfsic_state *state,
+		struct btrfsic_block *const block,
+		struct btrfs_super_block *const super_hdr);
+static void btrfsic_bio_end_io(struct bio *bp, int bio_error_status);
+static void btrfsic_bh_end_io(struct buffer_head *bh, int uptodate);
+static int btrfsic_is_block_ref_by_superblock(const struct btrfsic_state *state,
+					      const struct btrfsic_block *block,
+					      int recursion_level);
+static int btrfsic_check_all_ref_blocks(struct btrfsic_state *state,
+					struct btrfsic_block *const block,
+					int recursion_level);
+static void btrfsic_print_add_link(const struct btrfsic_state *state,
+				   const struct btrfsic_block_link *l);
+static void btrfsic_print_rem_link(const struct btrfsic_state *state,
+				   const struct btrfsic_block_link *l);
+static char btrfsic_get_block_type(const struct btrfsic_state *state,
+				   const struct btrfsic_block *block);
+static void btrfsic_dump_tree(const struct btrfsic_state *state);
+static void btrfsic_dump_tree_sub(const struct btrfsic_state *state,
+				  const struct btrfsic_block *block,
+				  int indent_level);
+static struct btrfsic_block_link *btrfsic_block_link_lookup_or_add(
+		struct btrfsic_state *state,
+		struct btrfsic_block_data_ctx *next_block_ctx,
+		struct btrfsic_block *next_block,
+		struct btrfsic_block *from_block,
+		u64 parent_generation);
+static struct btrfsic_block *btrfsic_block_lookup_or_add(
+		struct btrfsic_state *state,
+		struct btrfsic_block_data_ctx *block_ctx,
+		const char *additional_string,
+		int is_metadata,
+		int is_iodone,
+		int never_written,
+		int mirror_num,
+		int *was_created);
+static int btrfsic_process_superblock_dev_mirror(
+		struct btrfsic_state *state,
+		struct btrfsic_dev_state *dev_state,
+		struct btrfs_device *device,
+		int superblock_mirror_num,
+		struct btrfsic_dev_state **selected_dev_state,
+		struct btrfs_super_block *selected_super);
+static struct btrfsic_dev_state *btrfsic_dev_state_lookup(
+		struct block_device *bdev);
+static void btrfsic_cmp_log_and_dev_bytenr(struct btrfsic_state *state,
+					   u64 bytenr,
+					   struct btrfsic_dev_state *dev_state,
+					   u64 dev_bytenr);
+
+static struct mutex btrfsic_mutex;
+static int btrfsic_is_initialized;
+static struct btrfsic_dev_state_hashtable btrfsic_dev_state_hashtable;
+
+
+static void btrfsic_block_init(struct btrfsic_block *b)
+{
+	b->magic_num = BTRFSIC_BLOCK_MAGIC_NUMBER;
+	b->dev_state = NULL;
+	b->dev_bytenr = 0;
+	b->logical_bytenr = 0;
+	b->generation = BTRFSIC_GENERATION_UNKNOWN;
+	b->disk_key.objectid = 0;
+	b->disk_key.type = 0;
+	b->disk_key.offset = 0;
+	b->is_metadata = 0;
+	b->is_superblock = 0;
+	b->is_iodone = 0;
+	b->iodone_w_error = 0;
+	b->never_written = 0;
+	b->mirror_num = 0;
+	b->next_in_same_bio = NULL;
+	b->orig_bio_bh_private = NULL;
+	b->orig_bio_bh_end_io.bio = NULL;
+	INIT_LIST_HEAD(&b->collision_resolving_node);
+	INIT_LIST_HEAD(&b->all_blocks_node);
+	INIT_LIST_HEAD(&b->ref_to_list);
+	INIT_LIST_HEAD(&b->ref_from_list);
+	b->submit_bio_bh_rw = 0;
+	b->flush_gen = 0;
+}
+
+static struct btrfsic_block *btrfsic_block_alloc(void)
+{
+	struct btrfsic_block *b;
+
+	b = kzalloc(sizeof(*b), GFP_NOFS);
+	if (NULL != b)
+		btrfsic_block_init(b);
+
+	return b;
+}
+
+static void btrfsic_block_free(struct btrfsic_block *b)
+{
+	BUG_ON(!(NULL == b || BTRFSIC_BLOCK_MAGIC_NUMBER == b->magic_num));
+	kfree(b);
+}
+
+static void btrfsic_block_link_init(struct btrfsic_block_link *l)
+{
+	l->magic_num = BTRFSIC_BLOCK_LINK_MAGIC_NUMBER;
+	l->ref_cnt = 1;
+	INIT_LIST_HEAD(&l->node_ref_to);
+	INIT_LIST_HEAD(&l->node_ref_from);
+	INIT_LIST_HEAD(&l->collision_resolving_node);
+	l->block_ref_to = NULL;
+	l->block_ref_from = NULL;
+}
+
+static struct btrfsic_block_link *btrfsic_block_link_alloc(void)
+{
+	struct btrfsic_block_link *l;
+
+	l = kzalloc(sizeof(*l), GFP_NOFS);
+	if (NULL != l)
+		btrfsic_block_link_init(l);
+
+	return l;
+}
+
+static void btrfsic_block_link_free(struct btrfsic_block_link *l)
+{
+	BUG_ON(!(NULL == l || BTRFSIC_BLOCK_LINK_MAGIC_NUMBER == l->magic_num));
+	kfree(l);
+}
+
+static void btrfsic_dev_state_init(struct btrfsic_dev_state *ds)
+{
+	ds->magic_num = BTRFSIC_DEV2STATE_MAGIC_NUMBER;
+	ds->bdev = NULL;
+	ds->state = NULL;
+	ds->name[0] = '\0';
+	INIT_LIST_HEAD(&ds->collision_resolving_node);
+	ds->last_flush_gen = 0;
+	btrfsic_block_init(&ds->dummy_block_for_bio_bh_flush);
+	ds->dummy_block_for_bio_bh_flush.is_iodone = 1;
+	ds->dummy_block_for_bio_bh_flush.dev_state = ds;
+}
+
+static struct btrfsic_dev_state *btrfsic_dev_state_alloc(void)
+{
+	struct btrfsic_dev_state *ds;
+
+	ds = kzalloc(sizeof(*ds), GFP_NOFS);
+	if (NULL != ds)
+		btrfsic_dev_state_init(ds);
+
+	return ds;
+}
+
+static void btrfsic_dev_state_free(struct btrfsic_dev_state *ds)
+{
+	BUG_ON(!(NULL == ds ||
+		 BTRFSIC_DEV2STATE_MAGIC_NUMBER == ds->magic_num));
+	kfree(ds);
+}
+
+static void btrfsic_block_hashtable_init(struct btrfsic_block_hashtable *h)
+{
+	int i;
+
+	for (i = 0; i < BTRFSIC_BLOCK_HASHTABLE_SIZE; i++)
+		INIT_LIST_HEAD(h->table + i);
+}
+
+static void btrfsic_block_hashtable_add(struct btrfsic_block *b,
+					struct btrfsic_block_hashtable *h)
+{
+	const unsigned int hashval =
+	    (((unsigned int)(b->dev_bytenr >> 16)) ^
+	     ((unsigned int)((uintptr_t)b->dev_state->bdev))) &
+	     (BTRFSIC_BLOCK_HASHTABLE_SIZE - 1);
+
+	list_add(&b->collision_resolving_node, h->table + hashval);
+}
+
+static void btrfsic_block_hashtable_remove(struct btrfsic_block *b)
+{
+	list_del(&b->collision_resolving_node);
+}
+
+static struct btrfsic_block *btrfsic_block_hashtable_lookup(
+		struct block_device *bdev,
+		u64 dev_bytenr,
+		struct btrfsic_block_hashtable *h)
+{
+	const unsigned int hashval =
+	    (((unsigned int)(dev_bytenr >> 16)) ^
+	     ((unsigned int)((uintptr_t)bdev))) &
+	     (BTRFSIC_BLOCK_HASHTABLE_SIZE - 1);
+	struct list_head *elem;
+
+	list_for_each(elem, h->table + hashval) {
+		struct btrfsic_block *const b =
+		    list_entry(elem, struct btrfsic_block,
+			       collision_resolving_node);
+
+		if (b->dev_state->bdev == bdev && b->dev_bytenr == dev_bytenr)
+			return b;
+	}
+
+	return NULL;
+}
+
+static void btrfsic_block_link_hashtable_init(
+		struct btrfsic_block_link_hashtable *h)
+{
+	int i;
+
+	for (i = 0; i < BTRFSIC_BLOCK_LINK_HASHTABLE_SIZE; i++)
+		INIT_LIST_HEAD(h->table + i);
+}
+
+static void btrfsic_block_link_hashtable_add(
+		struct btrfsic_block_link *l,
+		struct btrfsic_block_link_hashtable *h)
+{
+	const unsigned int hashval =
+	    (((unsigned int)(l->block_ref_to->dev_bytenr >> 16)) ^
+	     ((unsigned int)(l->block_ref_from->dev_bytenr >> 16)) ^
+	     ((unsigned int)((uintptr_t)l->block_ref_to->dev_state->bdev)) ^
+	     ((unsigned int)((uintptr_t)l->block_ref_from->dev_state->bdev)))
+	     & (BTRFSIC_BLOCK_LINK_HASHTABLE_SIZE - 1);
+
+	BUG_ON(NULL == l->block_ref_to);
+	BUG_ON(NULL == l->block_ref_from);
+	list_add(&l->collision_resolving_node, h->table + hashval);
+}
+
+static void btrfsic_block_link_hashtable_remove(struct btrfsic_block_link *l)
+{
+	list_del(&l->collision_resolving_node);
+}
+
+static struct btrfsic_block_link *btrfsic_block_link_hashtable_lookup(
+		struct block_device *bdev_ref_to,
+		u64 dev_bytenr_ref_to,
+		struct block_device *bdev_ref_from,
+		u64 dev_bytenr_ref_from,
+		struct btrfsic_block_link_hashtable *h)
+{
+	const unsigned int hashval =
+	    (((unsigned int)(dev_bytenr_ref_to >> 16)) ^
+	     ((unsigned int)(dev_bytenr_ref_from >> 16)) ^
+	     ((unsigned int)((uintptr_t)bdev_ref_to)) ^
+	     ((unsigned int)((uintptr_t)bdev_ref_from))) &
+	     (BTRFSIC_BLOCK_LINK_HASHTABLE_SIZE - 1);
+	struct list_head *elem;
+
+	list_for_each(elem, h->table + hashval) {
+		struct btrfsic_block_link *const l =
+		    list_entry(elem, struct btrfsic_block_link,
+			       collision_resolving_node);
+
+		BUG_ON(NULL == l->block_ref_to);
+		BUG_ON(NULL == l->block_ref_from);
+		if (l->block_ref_to->dev_state->bdev == bdev_ref_to &&
+		    l->block_ref_to->dev_bytenr == dev_bytenr_ref_to &&
+		    l->block_ref_from->dev_state->bdev == bdev_ref_from &&
+		    l->block_ref_from->dev_bytenr == dev_bytenr_ref_from)
+			return l;
+	}
+
+	return NULL;
+}
+
+static void btrfsic_dev_state_hashtable_init(
+		struct btrfsic_dev_state_hashtable *h)
+{
+	int i;
+
+	for (i = 0; i < BTRFSIC_DEV2STATE_HASHTABLE_SIZE; i++)
+		INIT_LIST_HEAD(h->table + i);
+}
+
+static void btrfsic_dev_state_hashtable_add(
+		struct btrfsic_dev_state *ds,
+		struct btrfsic_dev_state_hashtable *h)
+{
+	const unsigned int hashval =
+	    (((unsigned int)((uintptr_t)ds->bdev)) &
+	     (BTRFSIC_DEV2STATE_HASHTABLE_SIZE - 1));
+
+	list_add(&ds->collision_resolving_node, h->table + hashval);
+}
+
+static void btrfsic_dev_state_hashtable_remove(struct btrfsic_dev_state *ds)
+{
+	list_del(&ds->collision_resolving_node);
+}
+
+static struct btrfsic_dev_state *btrfsic_dev_state_hashtable_lookup(
+		struct block_device *bdev,
+		struct btrfsic_dev_state_hashtable *h)
+{
+	const unsigned int hashval =
+	    (((unsigned int)((uintptr_t)bdev)) &
+	     (BTRFSIC_DEV2STATE_HASHTABLE_SIZE - 1));
+	struct list_head *elem;
+
+	list_for_each(elem, h->table + hashval) {
+		struct btrfsic_dev_state *const ds =
+		    list_entry(elem, struct btrfsic_dev_state,
+			       collision_resolving_node);
+
+		if (ds->bdev == bdev)
+			return ds;
+	}
+
+	return NULL;
+}
+
+static int btrfsic_process_superblock(struct btrfsic_state *state,
+				      struct btrfs_fs_devices *fs_devices)
+{
+	int ret = 0;
+	struct btrfs_super_block *selected_super;
+	struct list_head *dev_head = &fs_devices->devices;
+	struct btrfs_device *device;
+	struct btrfsic_dev_state *selected_dev_state = NULL;
+	int pass;
+
+	BUG_ON(NULL == state);
+	selected_super = kzalloc(sizeof(*selected_super), GFP_NOFS);
+	if (NULL == selected_super) {
+		printk(KERN_INFO "btrfsic: error, kmalloc failed!\n");
+		return -1;
+	}
+
+	list_for_each_entry(device, dev_head, dev_list) {
+		int i;
+		struct btrfsic_dev_state *dev_state;
+
+		if (!device->bdev || !device->name)
+			continue;
+
+		dev_state = btrfsic_dev_state_lookup(device->bdev);
+		BUG_ON(NULL == dev_state);
+		for (i = 0; i < BTRFS_SUPER_MIRROR_MAX; i++) {
+			ret = btrfsic_process_superblock_dev_mirror(
+					state, dev_state, device, i,
+					&selected_dev_state, selected_super);
+			if (0 != ret && 0 == i) {
+				kfree(selected_super);
+				return ret;
+			}
+		}
+	}
+
+	if (NULL == state->latest_superblock) {
+		printk(KERN_INFO "btrfsic: no superblock found!\n");
+		kfree(selected_super);
+		return -1;
+	}
+
+	state->csum_size = btrfs_super_csum_size(selected_super);
+
+	for (pass = 0; pass < 3; pass++) {
+		int num_copies;
+		int mirror_num;
+		u64 next_bytenr;
+
+		switch (pass) {
+		case 0:
+			next_bytenr = btrfs_super_root(selected_super);
+			if (state->print_mask &
+			    BTRFSIC_PRINT_MASK_ROOT_CHUNK_LOG_TREE_LOCATION)
+				printk(KERN_INFO "root@%llu\n", next_bytenr);
+			break;
+		case 1:
+			next_bytenr = btrfs_super_chunk_root(selected_super);
+			if (state->print_mask &
+			    BTRFSIC_PRINT_MASK_ROOT_CHUNK_LOG_TREE_LOCATION)
+				printk(KERN_INFO "chunk@%llu\n", next_bytenr);
+			break;
+		case 2:
+			next_bytenr = btrfs_super_log_root(selected_super);
+			if (0 == next_bytenr)
+				continue;
+			if (state->print_mask &
+			    BTRFSIC_PRINT_MASK_ROOT_CHUNK_LOG_TREE_LOCATION)
+				printk(KERN_INFO "log@%llu\n", next_bytenr);
+			break;
+		}
+
+		num_copies =
+		    btrfs_num_copies(state->root->fs_info,
+				     next_bytenr, state->metablock_size);
+		if (state->print_mask & BTRFSIC_PRINT_MASK_NUM_COPIES)
+			printk(KERN_INFO "num_copies(log_bytenr=%llu) = %d\n",
+			       next_bytenr, num_copies);
+
+		for (mirror_num = 1; mirror_num <= num_copies; mirror_num++) {
+			struct btrfsic_block *next_block;
+			struct btrfsic_block_data_ctx tmp_next_block_ctx;
+			struct btrfsic_block_link *l;
+
+			ret = btrfsic_map_block(state, next_bytenr,
+						state->metablock_size,
+						&tmp_next_block_ctx,
+						mirror_num);
+			if (ret) {
+				printk(KERN_INFO "btrfsic:"
+				       " btrfsic_map_block(root @%llu,"
+				       " mirror %d) failed!\n",
+				       next_bytenr, mirror_num);
+				kfree(selected_super);
+				return -1;
+			}
+
+			next_block = btrfsic_block_hashtable_lookup(
+					tmp_next_block_ctx.dev->bdev,
+					tmp_next_block_ctx.dev_bytenr,
+					&state->block_hashtable);
+			BUG_ON(NULL == next_block);
+
+			l = btrfsic_block_link_hashtable_lookup(
+					tmp_next_block_ctx.dev->bdev,
+					tmp_next_block_ctx.dev_bytenr,
+					state->latest_superblock->dev_state->
+					bdev,
+					state->latest_superblock->dev_bytenr,
+					&state->block_link_hashtable);
+			BUG_ON(NULL == l);
+
+			ret = btrfsic_read_block(state, &tmp_next_block_ctx);
+			if (ret < (int)PAGE_CACHE_SIZE) {
+				printk(KERN_INFO
+				       "btrfsic: read @logical %llu failed!\n",
+				       tmp_next_block_ctx.start);
+				btrfsic_release_block_ctx(&tmp_next_block_ctx);
+				kfree(selected_super);
+				return -1;
+			}
+
+			ret = btrfsic_process_metablock(state,
+							next_block,
+							&tmp_next_block_ctx,
+							BTRFS_MAX_LEVEL + 3, 1);
+			btrfsic_release_block_ctx(&tmp_next_block_ctx);
+		}
+	}
+
+	kfree(selected_super);
+	return ret;
+}
+
+static int btrfsic_process_superblock_dev_mirror(
+		struct btrfsic_state *state,
+		struct btrfsic_dev_state *dev_state,
+		struct btrfs_device *device,
+		int superblock_mirror_num,
+		struct btrfsic_dev_state **selected_dev_state,
+		struct btrfs_super_block *selected_super)
+{
+	struct btrfs_super_block *super_tmp;
+	u64 dev_bytenr;
+	struct buffer_head *bh;
+	struct btrfsic_block *superblock_tmp;
+	int pass;
+	struct block_device *const superblock_bdev = device->bdev;
+
+	/* super block bytenr is always the unmapped device bytenr */
+	dev_bytenr = btrfs_sb_offset(superblock_mirror_num);
+	if (dev_bytenr + BTRFS_SUPER_INFO_SIZE > device->total_bytes)
+		return -1;
+	bh = __bread(superblock_bdev, dev_bytenr / 4096,
+		     BTRFS_SUPER_INFO_SIZE);
+	if (NULL == bh)
+		return -1;
+	super_tmp = (struct btrfs_super_block *)
+	    (bh->b_data + (dev_bytenr & 4095));
+
+	if (btrfs_super_bytenr(super_tmp) != dev_bytenr ||
+	    btrfs_super_magic(super_tmp) != BTRFS_MAGIC ||
+	    memcmp(device->uuid, super_tmp->dev_item.uuid, BTRFS_UUID_SIZE) ||
+	    btrfs_super_nodesize(super_tmp) != state->metablock_size ||
+	    btrfs_super_leafsize(super_tmp) != state->metablock_size ||
+	    btrfs_super_sectorsize(super_tmp) != state->datablock_size) {
+		brelse(bh);
+		return 0;
+	}
+
+	superblock_tmp =
+	    btrfsic_block_hashtable_lookup(superblock_bdev,
+					   dev_bytenr,
+					   &state->block_hashtable);
+	if (NULL == superblock_tmp) {
+		superblock_tmp = btrfsic_block_alloc();
+		if (NULL == superblock_tmp) {
+			printk(KERN_INFO "btrfsic: error, kmalloc failed!\n");
+			brelse(bh);
+			return -1;
+		}
+		/* for superblock, only the dev_bytenr makes sense */
+		superblock_tmp->dev_bytenr = dev_bytenr;
+		superblock_tmp->dev_state = dev_state;
+		superblock_tmp->logical_bytenr = dev_bytenr;
+		superblock_tmp->generation = btrfs_super_generation(super_tmp);
+		superblock_tmp->is_metadata = 1;
+		superblock_tmp->is_superblock = 1;
+		superblock_tmp->is_iodone = 1;
+		superblock_tmp->never_written = 0;
+		superblock_tmp->mirror_num = 1 + superblock_mirror_num;
+		if (state->print_mask & BTRFSIC_PRINT_MASK_SUPERBLOCK_WRITE)
+			printk_in_rcu(KERN_INFO "New initial S-block (bdev %p, %s)"
+				     " @%llu (%s/%llu/%d)\n",
+				     superblock_bdev,
+				     rcu_str_deref(device->name), dev_bytenr,
+				     dev_state->name, dev_bytenr,
+				     superblock_mirror_num);
+		list_add(&superblock_tmp->all_blocks_node,
+			 &state->all_blocks_list);
+		btrfsic_block_hashtable_add(superblock_tmp,
+					    &state->block_hashtable);
+	}
+
+	/* select the one with the highest generation field */
+	if (btrfs_super_generation(super_tmp) >
+	    state->max_superblock_generation ||
+	    0 == state->max_superblock_generation) {
+		memcpy(selected_super, super_tmp, sizeof(*selected_super));
+		*selected_dev_state = dev_state;
+		state->max_superblock_generation =
+		    btrfs_super_generation(super_tmp);
+		state->latest_superblock = superblock_tmp;
+	}
+
+	for (pass = 0; pass < 3; pass++) {
+		u64 next_bytenr;
+		int num_copies;
+		int mirror_num;
+		const char *additional_string = NULL;
+		struct btrfs_disk_key tmp_disk_key;
+
+		tmp_disk_key.type = BTRFS_ROOT_ITEM_KEY;
+		tmp_disk_key.offset = 0;
+		switch (pass) {
+		case 0:
+			btrfs_set_disk_key_objectid(&tmp_disk_key,
+						    BTRFS_ROOT_TREE_OBJECTID);
+			additional_string = "initial root ";
+			next_bytenr = btrfs_super_root(super_tmp);
+			break;
+		case 1:
+			btrfs_set_disk_key_objectid(&tmp_disk_key,
+						    BTRFS_CHUNK_TREE_OBJECTID);
+			additional_string = "initial chunk ";
+			next_bytenr = btrfs_super_chunk_root(super_tmp);
+			break;
+		case 2:
+			btrfs_set_disk_key_objectid(&tmp_disk_key,
+						    BTRFS_TREE_LOG_OBJECTID);
+			additional_string = "initial log ";
+			next_bytenr = btrfs_super_log_root(super_tmp);
+			if (0 == next_bytenr)
+				continue;
+			break;
+		}
+
+		num_copies =
+		    btrfs_num_copies(state->root->fs_info,
+				     next_bytenr, state->metablock_size);
+		if (state->print_mask & BTRFSIC_PRINT_MASK_NUM_COPIES)
+			printk(KERN_INFO "num_copies(log_bytenr=%llu) = %d\n",
+			       next_bytenr, num_copies);
+		for (mirror_num = 1; mirror_num <= num_copies; mirror_num++) {
+			struct btrfsic_block *next_block;
+			struct btrfsic_block_data_ctx tmp_next_block_ctx;
+			struct btrfsic_block_link *l;
+
+			if (btrfsic_map_block(state, next_bytenr,
+					      state->metablock_size,
+					      &tmp_next_block_ctx,
+					      mirror_num)) {
+				printk(KERN_INFO "btrfsic: btrfsic_map_block("
+				       "bytenr @%llu, mirror %d) failed!\n",
+				       next_bytenr, mirror_num);
+				brelse(bh);
+				return -1;
+			}
+
+			next_block = btrfsic_block_lookup_or_add(
+					state, &tmp_next_block_ctx,
+					additional_string, 1, 1, 0,
+					mirror_num, NULL);
+			if (NULL == next_block) {
+				btrfsic_release_block_ctx(&tmp_next_block_ctx);
+				brelse(bh);
+				return -1;
+			}
+
+			next_block->disk_key = tmp_disk_key;
+			next_block->generation = BTRFSIC_GENERATION_UNKNOWN;
+			l = btrfsic_block_link_lookup_or_add(
+					state, &tmp_next_block_ctx,
+					next_block, superblock_tmp,
+					BTRFSIC_GENERATION_UNKNOWN);
+			btrfsic_release_block_ctx(&tmp_next_block_ctx);
+			if (NULL == l) {
+				brelse(bh);
+				return -1;
+			}
+		}
+	}
+	if (state->print_mask & BTRFSIC_PRINT_MASK_INITIAL_ALL_TREES)
+		btrfsic_dump_tree_sub(state, superblock_tmp, 0);
+
+	brelse(bh);
+	return 0;
+}
+
+static struct btrfsic_stack_frame *btrfsic_stack_frame_alloc(void)
+{
+	struct btrfsic_stack_frame *sf;
+
+	sf = kzalloc(sizeof(*sf), GFP_NOFS);
+	if (NULL == sf)
+		printk(KERN_INFO "btrfsic: alloc memory failed!\n");
+	else
+		sf->magic = BTRFSIC_BLOCK_STACK_FRAME_MAGIC_NUMBER;
+	return sf;
+}
+
+static void btrfsic_stack_frame_free(struct btrfsic_stack_frame *sf)
+{
+	BUG_ON(!(NULL == sf ||
+		 BTRFSIC_BLOCK_STACK_FRAME_MAGIC_NUMBER == sf->magic));
+	kfree(sf);
+}
+
+static int btrfsic_process_metablock(
+		struct btrfsic_state *state,
+		struct btrfsic_block *const first_block,
+		struct btrfsic_block_data_ctx *const first_block_ctx,
+		int first_limit_nesting, int force_iodone_flag)
+{
+	struct btrfsic_stack_frame initial_stack_frame = { 0 };
+	struct btrfsic_stack_frame *sf;
+	struct btrfsic_stack_frame *next_stack;
+	struct btrfs_header *const first_hdr =
+		(struct btrfs_header *)first_block_ctx->datav[0];
+
+	BUG_ON(!first_hdr);
+	sf = &initial_stack_frame;
+	sf->error = 0;
+	sf->i = -1;
+	sf->limit_nesting = first_limit_nesting;
+	sf->block = first_block;
+	sf->block_ctx = first_block_ctx;
+	sf->next_block = NULL;
+	sf->hdr = first_hdr;
+	sf->prev = NULL;
+
+continue_with_new_stack_frame:
+	sf->block->generation = le64_to_cpu(sf->hdr->generation);
+	if (0 == sf->hdr->level) {
+		struct btrfs_leaf *const leafhdr =
+		    (struct btrfs_leaf *)sf->hdr;
+
+		if (-1 == sf->i) {
+			sf->nr = btrfs_stack_header_nritems(&leafhdr->header);
+
+			if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
+				printk(KERN_INFO
+				       "leaf %llu items %d generation %llu"
+				       " owner %llu\n",
+				       sf->block_ctx->start, sf->nr,
+				       btrfs_stack_header_generation(
+					       &leafhdr->header),
+				       btrfs_stack_header_owner(
+					       &leafhdr->header));
+		}
+
+continue_with_current_leaf_stack_frame:
+		if (0 == sf->num_copies || sf->mirror_num > sf->num_copies) {
+			sf->i++;
+			sf->num_copies = 0;
+		}
+
+		if (sf->i < sf->nr) {
+			struct btrfs_item disk_item;
+			u32 disk_item_offset =
+				(uintptr_t)(leafhdr->items + sf->i) -
+				(uintptr_t)leafhdr;
+			struct btrfs_disk_key *disk_key;
+			u8 type;
+			u32 item_offset;
+			u32 item_size;
+
+			if (disk_item_offset + sizeof(struct btrfs_item) >
+			    sf->block_ctx->len) {
+leaf_item_out_of_bounce_error:
+				printk(KERN_INFO
+				       "btrfsic: leaf item out of bounce at logical %llu, dev %s\n",
+				       sf->block_ctx->start,
+				       sf->block_ctx->dev->name);
+				goto one_stack_frame_backwards;
+			}
+			btrfsic_read_from_block_data(sf->block_ctx,
+						     &disk_item,
+						     disk_item_offset,
+						     sizeof(struct btrfs_item));
+			item_offset = btrfs_stack_item_offset(&disk_item);
+			item_size = btrfs_stack_item_size(&disk_item);
+			disk_key = &disk_item.key;
+			type = btrfs_disk_key_type(disk_key);
+
+			if (BTRFS_ROOT_ITEM_KEY == type) {
+				struct btrfs_root_item root_item;
+				u32 root_item_offset;
+				u64 next_bytenr;
+
+				root_item_offset = item_offset +
+					offsetof(struct btrfs_leaf, items);
+				if (root_item_offset + item_size >
+				    sf->block_ctx->len)
+					goto leaf_item_out_of_bounce_error;
+				btrfsic_read_from_block_data(
+					sf->block_ctx, &root_item,
+					root_item_offset,
+					item_size);
+				next_bytenr = btrfs_root_bytenr(&root_item);
+
+				sf->error =
+				    btrfsic_create_link_to_next_block(
+						state,
+						sf->block,
+						sf->block_ctx,
+						next_bytenr,
+						sf->limit_nesting,
+						&sf->next_block_ctx,
+						&sf->next_block,
+						force_iodone_flag,
+						&sf->num_copies,
+						&sf->mirror_num,
+						disk_key,
+						btrfs_root_generation(
+						&root_item));
+				if (sf->error)
+					goto one_stack_frame_backwards;
+
+				if (NULL != sf->next_block) {
+					struct btrfs_header *const next_hdr =
+					    (struct btrfs_header *)
+					    sf->next_block_ctx.datav[0];
+
+					next_stack =
+					    btrfsic_stack_frame_alloc();
+					if (NULL == next_stack) {
+						sf->error = -1;
+						btrfsic_release_block_ctx(
+								&sf->
+								next_block_ctx);
+						goto one_stack_frame_backwards;
+					}
+
+					next_stack->i = -1;
+					next_stack->block = sf->next_block;
+					next_stack->block_ctx =
+					    &sf->next_block_ctx;
+					next_stack->next_block = NULL;
+					next_stack->hdr = next_hdr;
+					next_stack->limit_nesting =
+					    sf->limit_nesting - 1;
+					next_stack->prev = sf;
+					sf = next_stack;
+					goto continue_with_new_stack_frame;
+				}
+			} else if (BTRFS_EXTENT_DATA_KEY == type &&
+				   state->include_extent_data) {
+				sf->error = btrfsic_handle_extent_data(
+						state,
+						sf->block,
+						sf->block_ctx,
+						item_offset,
+						force_iodone_flag);
+				if (sf->error)
+					goto one_stack_frame_backwards;
+			}
+
+			goto continue_with_current_leaf_stack_frame;
+		}
+	} else {
+		struct btrfs_node *const nodehdr = (struct btrfs_node *)sf->hdr;
+
+		if (-1 == sf->i) {
+			sf->nr = btrfs_stack_header_nritems(&nodehdr->header);
+
+			if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
+				printk(KERN_INFO "node %llu level %d items %d"
+				       " generation %llu owner %llu\n",
+				       sf->block_ctx->start,
+				       nodehdr->header.level, sf->nr,
+				       btrfs_stack_header_generation(
+				       &nodehdr->header),
+				       btrfs_stack_header_owner(
+				       &nodehdr->header));
+		}
+
+continue_with_current_node_stack_frame:
+		if (0 == sf->num_copies || sf->mirror_num > sf->num_copies) {
+			sf->i++;
+			sf->num_copies = 0;
+		}
+
+		if (sf->i < sf->nr) {
+			struct btrfs_key_ptr key_ptr;
+			u32 key_ptr_offset;
+			u64 next_bytenr;
+
+			key_ptr_offset = (uintptr_t)(nodehdr->ptrs + sf->i) -
+					  (uintptr_t)nodehdr;
+			if (key_ptr_offset + sizeof(struct btrfs_key_ptr) >
+			    sf->block_ctx->len) {
+				printk(KERN_INFO
+				       "btrfsic: node item out of bounce at logical %llu, dev %s\n",
+				       sf->block_ctx->start,
+				       sf->block_ctx->dev->name);
+				goto one_stack_frame_backwards;
+			}
+			btrfsic_read_from_block_data(
+				sf->block_ctx, &key_ptr, key_ptr_offset,
+				sizeof(struct btrfs_key_ptr));
+			next_bytenr = btrfs_stack_key_blockptr(&key_ptr);
+
+			sf->error = btrfsic_create_link_to_next_block(
+					state,
+					sf->block,
+					sf->block_ctx,
+					next_bytenr,
+					sf->limit_nesting,
+					&sf->next_block_ctx,
+					&sf->next_block,
+					force_iodone_flag,
+					&sf->num_copies,
+					&sf->mirror_num,
+					&key_ptr.key,
+					btrfs_stack_key_generation(&key_ptr));
+			if (sf->error)
+				goto one_stack_frame_backwards;
+
+			if (NULL != sf->next_block) {
+				struct btrfs_header *const next_hdr =
+				    (struct btrfs_header *)
+				    sf->next_block_ctx.datav[0];
+
+				next_stack = btrfsic_stack_frame_alloc();
+				if (NULL == next_stack) {
+					sf->error = -1;
+					goto one_stack_frame_backwards;
+				}
+
+				next_stack->i = -1;
+				next_stack->block = sf->next_block;
+				next_stack->block_ctx = &sf->next_block_ctx;
+				next_stack->next_block = NULL;
+				next_stack->hdr = next_hdr;
+				next_stack->limit_nesting =
+				    sf->limit_nesting - 1;
+				next_stack->prev = sf;
+				sf = next_stack;
+				goto continue_with_new_stack_frame;
+			}
+
+			goto continue_with_current_node_stack_frame;
+		}
+	}
+
+one_stack_frame_backwards:
+	if (NULL != sf->prev) {
+		struct btrfsic_stack_frame *const prev = sf->prev;
+
+		/* the one for the initial block is freed in the caller */
+		btrfsic_release_block_ctx(sf->block_ctx);
+
+		if (sf->error) {
+			prev->error = sf->error;
+			btrfsic_stack_frame_free(sf);
+			sf = prev;
+			goto one_stack_frame_backwards;
+		}
+
+		btrfsic_stack_frame_free(sf);
+		sf = prev;
+		goto continue_with_new_stack_frame;
+	} else {
+		BUG_ON(&initial_stack_frame != sf);
+	}
+
+	return sf->error;
+}
+
+static void btrfsic_read_from_block_data(
+	struct btrfsic_block_data_ctx *block_ctx,
+	void *dstv, u32 offset, size_t len)
+{
+	size_t cur;
+	size_t offset_in_page;
+	char *kaddr;
+	char *dst = (char *)dstv;
+	size_t start_offset = block_ctx->start & ((u64)PAGE_CACHE_SIZE - 1);
+	unsigned long i = (start_offset + offset) >> PAGE_CACHE_SHIFT;
+
+	WARN_ON(offset + len > block_ctx->len);
+	offset_in_page = (start_offset + offset) & (PAGE_CACHE_SIZE - 1);
+
+	while (len > 0) {
+		cur = min(len, ((size_t)PAGE_CACHE_SIZE - offset_in_page));
+		BUG_ON(i >= (block_ctx->len + PAGE_CACHE_SIZE - 1) >>
+			    PAGE_CACHE_SHIFT);
+		kaddr = block_ctx->datav[i];
+		memcpy(dst, kaddr + offset_in_page, cur);
+
+		dst += cur;
+		len -= cur;
+		offset_in_page = 0;
+		i++;
+	}
+}
+
+static int btrfsic_create_link_to_next_block(
+		struct btrfsic_state *state,
+		struct btrfsic_block *block,
+		struct btrfsic_block_data_ctx *block_ctx,
+		u64 next_bytenr,
+		int limit_nesting,
+		struct btrfsic_block_data_ctx *next_block_ctx,
+		struct btrfsic_block **next_blockp,
+		int force_iodone_flag,
+		int *num_copiesp, int *mirror_nump,
+		struct btrfs_disk_key *disk_key,
+		u64 parent_generation)
+{
+	struct btrfsic_block *next_block = NULL;
+	int ret;
+	struct btrfsic_block_link *l;
+	int did_alloc_block_link;
+	int block_was_created;
+
+	*next_blockp = NULL;
+	if (0 == *num_copiesp) {
+		*num_copiesp =
+		    btrfs_num_copies(state->root->fs_info,
+				     next_bytenr, state->metablock_size);
+		if (state->print_mask & BTRFSIC_PRINT_MASK_NUM_COPIES)
+			printk(KERN_INFO "num_copies(log_bytenr=%llu) = %d\n",
+			       next_bytenr, *num_copiesp);
+		*mirror_nump = 1;
+	}
+
+	if (*mirror_nump > *num_copiesp)
+		return 0;
+
+	if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
+		printk(KERN_INFO
+		       "btrfsic_create_link_to_next_block(mirror_num=%d)\n",
+		       *mirror_nump);
+	ret = btrfsic_map_block(state, next_bytenr,
+				state->metablock_size,
+				next_block_ctx, *mirror_nump);
+	if (ret) {
+		printk(KERN_INFO
+		       "btrfsic: btrfsic_map_block(@%llu, mirror=%d) failed!\n",
+		       next_bytenr, *mirror_nump);
+		btrfsic_release_block_ctx(next_block_ctx);
+		*next_blockp = NULL;
+		return -1;
+	}
+
+	next_block = btrfsic_block_lookup_or_add(state,
+						 next_block_ctx, "referenced ",
+						 1, force_iodone_flag,
+						 !force_iodone_flag,
+						 *mirror_nump,
+						 &block_was_created);
+	if (NULL == next_block) {
+		btrfsic_release_block_ctx(next_block_ctx);
+		*next_blockp = NULL;
+		return -1;
+	}
+	if (block_was_created) {
+		l = NULL;
+		next_block->generation = BTRFSIC_GENERATION_UNKNOWN;
+	} else {
+		if (next_block->logical_bytenr != next_bytenr &&
+		    !(!next_block->is_metadata &&
+		      0 == next_block->logical_bytenr)) {
+			printk(KERN_INFO
+			       "Referenced block @%llu (%s/%llu/%d)"
+			       " found in hash table, %c,"
+			       " bytenr mismatch (!= stored %llu).\n",
+			       next_bytenr, next_block_ctx->dev->name,
+			       next_block_ctx->dev_bytenr, *mirror_nump,
+			       btrfsic_get_block_type(state, next_block),
+			       next_block->logical_bytenr);
+		} else if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
+			printk(KERN_INFO
+			       "Referenced block @%llu (%s/%llu/%d)"
+			       " found in hash table, %c.\n",
+			       next_bytenr, next_block_ctx->dev->name,
+			       next_block_ctx->dev_bytenr, *mirror_nump,
+			       btrfsic_get_block_type(state, next_block));
+		next_block->logical_bytenr = next_bytenr;
+
+		next_block->mirror_num = *mirror_nump;
+		l = btrfsic_block_link_hashtable_lookup(
+				next_block_ctx->dev->bdev,
+				next_block_ctx->dev_bytenr,
+				block_ctx->dev->bdev,
+				block_ctx->dev_bytenr,
+				&state->block_link_hashtable);
+	}
+
+	next_block->disk_key = *disk_key;
+	if (NULL == l) {
+		l = btrfsic_block_link_alloc();
+		if (NULL == l) {
+			printk(KERN_INFO "btrfsic: error, kmalloc failed!\n");
+			btrfsic_release_block_ctx(next_block_ctx);
+			*next_blockp = NULL;
+			return -1;
+		}
+
+		did_alloc_block_link = 1;
+		l->block_ref_to = next_block;
+		l->block_ref_from = block;
+		l->ref_cnt = 1;
+		l->parent_generation = parent_generation;
+
+		if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
+			btrfsic_print_add_link(state, l);
+
+		list_add(&l->node_ref_to, &block->ref_to_list);
+		list_add(&l->node_ref_from, &next_block->ref_from_list);
+
+		btrfsic_block_link_hashtable_add(l,
+						 &state->block_link_hashtable);
+	} else {
+		did_alloc_block_link = 0;
+		if (0 == limit_nesting) {
+			l->ref_cnt++;
+			l->parent_generation = parent_generation;
+			if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
+				btrfsic_print_add_link(state, l);
+		}
+	}
+
+	if (limit_nesting > 0 && did_alloc_block_link) {
+		ret = btrfsic_read_block(state, next_block_ctx);
+		if (ret < (int)next_block_ctx->len) {
+			printk(KERN_INFO
+			       "btrfsic: read block @logical %llu failed!\n",
+			       next_bytenr);
+			btrfsic_release_block_ctx(next_block_ctx);
+			*next_blockp = NULL;
+			return -1;
+		}
+
+		*next_blockp = next_block;
+	} else {
+		*next_blockp = NULL;
+	}
+	(*mirror_nump)++;
+
+	return 0;
+}
+
+static int btrfsic_handle_extent_data(
+		struct btrfsic_state *state,
+		struct btrfsic_block *block,
+		struct btrfsic_block_data_ctx *block_ctx,
+		u32 item_offset, int force_iodone_flag)
+{
+	int ret;
+	struct btrfs_file_extent_item file_extent_item;
+	u64 file_extent_item_offset;
+	u64 next_bytenr;
+	u64 num_bytes;
+	u64 generation;
+	struct btrfsic_block_link *l;
+
+	file_extent_item_offset = offsetof(struct btrfs_leaf, items) +
+				  item_offset;
+	if (file_extent_item_offset +
+	    offsetof(struct btrfs_file_extent_item, disk_num_bytes) >
+	    block_ctx->len) {
+		printk(KERN_INFO
+		       "btrfsic: file item out of bounce at logical %llu, dev %s\n",
+		       block_ctx->start, block_ctx->dev->name);
+		return -1;
+	}
+
+	btrfsic_read_from_block_data(block_ctx, &file_extent_item,
+		file_extent_item_offset,
+		offsetof(struct btrfs_file_extent_item, disk_num_bytes));
+	if (BTRFS_FILE_EXTENT_REG != file_extent_item.type ||
+	    btrfs_stack_file_extent_disk_bytenr(&file_extent_item) == 0) {
+		if (state->print_mask & BTRFSIC_PRINT_MASK_VERY_VERBOSE)
+			printk(KERN_INFO "extent_data: type %u, disk_bytenr = %llu\n",
+			       file_extent_item.type,
+			       btrfs_stack_file_extent_disk_bytenr(
+			       &file_extent_item));
+		return 0;
+	}
+
+	if (file_extent_item_offset + sizeof(struct btrfs_file_extent_item) >
+	    block_ctx->len) {
+		printk(KERN_INFO
+		       "btrfsic: file item out of bounce at logical %llu, dev %s\n",
+		       block_ctx->start, block_ctx->dev->name);
+		return -1;
+	}
+	btrfsic_read_from_block_data(block_ctx, &file_extent_item,
+				     file_extent_item_offset,
+				     sizeof(struct btrfs_file_extent_item));
+	next_bytenr = btrfs_stack_file_extent_disk_bytenr(&file_extent_item);
+	if (btrfs_stack_file_extent_compression(&file_extent_item) ==
+	    BTRFS_COMPRESS_NONE) {
+		next_bytenr += btrfs_stack_file_extent_offset(&file_extent_item);
+		num_bytes = btrfs_stack_file_extent_num_bytes(&file_extent_item);
+	} else {
+		num_bytes = btrfs_stack_file_extent_disk_num_bytes(&file_extent_item);
+	}
+	generation = btrfs_stack_file_extent_generation(&file_extent_item);
+
+	if (state->print_mask & BTRFSIC_PRINT_MASK_VERY_VERBOSE)
+		printk(KERN_INFO "extent_data: type %u, disk_bytenr = %llu,"
+		       " offset = %llu, num_bytes = %llu\n",
+		       file_extent_item.type,
+		       btrfs_stack_file_extent_disk_bytenr(&file_extent_item),
+		       btrfs_stack_file_extent_offset(&file_extent_item),
+		       num_bytes);
+	while (num_bytes > 0) {
+		u32 chunk_len;
+		int num_copies;
+		int mirror_num;
+
+		if (num_bytes > state->datablock_size)
+			chunk_len = state->datablock_size;
+		else
+			chunk_len = num_bytes;
+
+		num_copies =
+		    btrfs_num_copies(state->root->fs_info,
+				     next_bytenr, state->datablock_size);
+		if (state->print_mask & BTRFSIC_PRINT_MASK_NUM_COPIES)
+			printk(KERN_INFO "num_copies(log_bytenr=%llu) = %d\n",
+			       next_bytenr, num_copies);
+		for (mirror_num = 1; mirror_num <= num_copies; mirror_num++) {
+			struct btrfsic_block_data_ctx next_block_ctx;
+			struct btrfsic_block *next_block;
+			int block_was_created;
+
+			if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
+				printk(KERN_INFO "btrfsic_handle_extent_data("
+				       "mirror_num=%d)\n", mirror_num);
+			if (state->print_mask & BTRFSIC_PRINT_MASK_VERY_VERBOSE)
+				printk(KERN_INFO
+				       "\tdisk_bytenr = %llu, num_bytes %u\n",
+				       next_bytenr, chunk_len);
+			ret = btrfsic_map_block(state, next_bytenr,
+						chunk_len, &next_block_ctx,
+						mirror_num);
+			if (ret) {
+				printk(KERN_INFO
+				       "btrfsic: btrfsic_map_block(@%llu,"
+				       " mirror=%d) failed!\n",
+				       next_bytenr, mirror_num);
+				return -1;
+			}
+
+			next_block = btrfsic_block_lookup_or_add(
+					state,
+					&next_block_ctx,
+					"referenced ",
+					0,
+					force_iodone_flag,
+					!force_iodone_flag,
+					mirror_num,
+					&block_was_created);
+			if (NULL == next_block) {
+				printk(KERN_INFO
+				       "btrfsic: error, kmalloc failed!\n");
+				btrfsic_release_block_ctx(&next_block_ctx);
+				return -1;
+			}
+			if (!block_was_created) {
+				if (next_block->logical_bytenr != next_bytenr &&
+				    !(!next_block->is_metadata &&
+				      0 == next_block->logical_bytenr)) {
+					printk(KERN_INFO
+					       "Referenced block"
+					       " @%llu (%s/%llu/%d)"
+					       " found in hash table, D,"
+					       " bytenr mismatch"
+					       " (!= stored %llu).\n",
+					       next_bytenr,
+					       next_block_ctx.dev->name,
+					       next_block_ctx.dev_bytenr,
+					       mirror_num,
+					       next_block->logical_bytenr);
+				}
+				next_block->logical_bytenr = next_bytenr;
+				next_block->mirror_num = mirror_num;
+			}
+
+			l = btrfsic_block_link_lookup_or_add(state,
+							     &next_block_ctx,
+							     next_block, block,
+							     generation);
+			btrfsic_release_block_ctx(&next_block_ctx);
+			if (NULL == l)
+				return -1;
+		}
+
+		next_bytenr += chunk_len;
+		num_bytes -= chunk_len;
+	}
+
+	return 0;
+}
+
+static int btrfsic_map_block(struct btrfsic_state *state, u64 bytenr, u32 len,
+			     struct btrfsic_block_data_ctx *block_ctx_out,
+			     int mirror_num)
+{
+	int ret;
+	u64 length;
+	struct btrfs_bio *multi = NULL;
+	struct btrfs_device *device;
+
+	length = len;
+	ret = btrfs_map_block(state->root->fs_info, READ,
+			      bytenr, &length, &multi, mirror_num);
+
+	if (ret) {
+		block_ctx_out->start = 0;
+		block_ctx_out->dev_bytenr = 0;
+		block_ctx_out->len = 0;
+		block_ctx_out->dev = NULL;
+		block_ctx_out->datav = NULL;
+		block_ctx_out->pagev = NULL;
+		block_ctx_out->mem_to_free = NULL;
+
+		return ret;
+	}
+
+	device = multi->stripes[0].dev;
+	block_ctx_out->dev = btrfsic_dev_state_lookup(device->bdev);
+	block_ctx_out->dev_bytenr = multi->stripes[0].physical;
+	block_ctx_out->start = bytenr;
+	block_ctx_out->len = len;
+	block_ctx_out->datav = NULL;
+	block_ctx_out->pagev = NULL;
+	block_ctx_out->mem_to_free = NULL;
+
+	kfree(multi);
+	if (NULL == block_ctx_out->dev) {
+		ret = -ENXIO;
+		printk(KERN_INFO "btrfsic: error, cannot lookup dev (#1)!\n");
+	}
+
+	return ret;
+}
+
+static int btrfsic_map_superblock(struct btrfsic_state *state, u64 bytenr,
+				  u32 len, struct block_device *bdev,
+				  struct btrfsic_block_data_ctx *block_ctx_out)
+{
+	block_ctx_out->dev = btrfsic_dev_state_lookup(bdev);
+	block_ctx_out->dev_bytenr = bytenr;
+	block_ctx_out->start = bytenr;
+	block_ctx_out->len = len;
+	block_ctx_out->datav = NULL;
+	block_ctx_out->pagev = NULL;
+	block_ctx_out->mem_to_free = NULL;
+	if (NULL != block_ctx_out->dev) {
+		return 0;
+	} else {
+		printk(KERN_INFO "btrfsic: error, cannot lookup dev (#2)!\n");
+		return -ENXIO;
+	}
+}
+
+static void btrfsic_release_block_ctx(struct btrfsic_block_data_ctx *block_ctx)
+{
+	if (block_ctx->mem_to_free) {
+		unsigned int num_pages;
+
+		BUG_ON(!block_ctx->datav);
+		BUG_ON(!block_ctx->pagev);
+		num_pages = (block_ctx->len + (u64)PAGE_CACHE_SIZE - 1) >>
+			    PAGE_CACHE_SHIFT;
+		while (num_pages > 0) {
+			num_pages--;
+			if (block_ctx->datav[num_pages]) {
+				kunmap(block_ctx->pagev[num_pages]);
+				block_ctx->datav[num_pages] = NULL;
+			}
+			if (block_ctx->pagev[num_pages]) {
+				__free_page(block_ctx->pagev[num_pages]);
+				block_ctx->pagev[num_pages] = NULL;
+			}
+		}
+
+		kfree(block_ctx->mem_to_free);
+		block_ctx->mem_to_free = NULL;
+		block_ctx->pagev = NULL;
+		block_ctx->datav = NULL;
+	}
+}
+
+static int btrfsic_read_block(struct btrfsic_state *state,
+			      struct btrfsic_block_data_ctx *block_ctx)
+{
+	unsigned int num_pages;
+	unsigned int i;
+	u64 dev_bytenr;
+	int ret;
+
+	BUG_ON(block_ctx->datav);
+	BUG_ON(block_ctx->pagev);
+	BUG_ON(block_ctx->mem_to_free);
+	if (block_ctx->dev_bytenr & ((u64)PAGE_CACHE_SIZE - 1)) {
+		printk(KERN_INFO
+		       "btrfsic: read_block() with unaligned bytenr %llu\n",
+		       block_ctx->dev_bytenr);
+		return -1;
+	}
+
+	num_pages = (block_ctx->len + (u64)PAGE_CACHE_SIZE - 1) >>
+		    PAGE_CACHE_SHIFT;
+	block_ctx->mem_to_free = kzalloc((sizeof(*block_ctx->datav) +
+					  sizeof(*block_ctx->pagev)) *
+					 num_pages, GFP_NOFS);
+	if (!block_ctx->mem_to_free)
+		return -1;
+	block_ctx->datav = block_ctx->mem_to_free;
+	block_ctx->pagev = (struct page **)(block_ctx->datav + num_pages);
+	for (i = 0; i < num_pages; i++) {
+		block_ctx->pagev[i] = alloc_page(GFP_NOFS);
+		if (!block_ctx->pagev[i])
+			return -1;
+	}
+
+	dev_bytenr = block_ctx->dev_bytenr;
+	for (i = 0; i < num_pages;) {
+		struct bio *bio;
+		unsigned int j;
+
+		bio = btrfs_io_bio_alloc(GFP_NOFS, num_pages - i);
+		if (!bio) {
+			printk(KERN_INFO
+			       "btrfsic: bio_alloc() for %u pages failed!\n",
+			       num_pages - i);
+			return -1;
+		}
+		bio->bi_bdev = block_ctx->dev->bdev;
+		bio->bi_iter.bi_sector = dev_bytenr >> 9;
+
+		for (j = i; j < num_pages; j++) {
+			ret = bio_add_page(bio, block_ctx->pagev[j],
+					   PAGE_CACHE_SIZE, 0);
+			if (PAGE_CACHE_SIZE != ret)
+				break;
+		}
+		if (j == i) {
+			printk(KERN_INFO
+			       "btrfsic: error, failed to add a single page!\n");
+			return -1;
+		}
+		if (submit_bio_wait(READ, bio)) {
+			printk(KERN_INFO
+			       "btrfsic: read error at logical %llu dev %s!\n",
+			       block_ctx->start, block_ctx->dev->name);
+			bio_put(bio);
+			return -1;
+		}
+		bio_put(bio);
+		dev_bytenr += (j - i) * PAGE_CACHE_SIZE;
+		i = j;
+	}
+	for (i = 0; i < num_pages; i++) {
+		block_ctx->datav[i] = kmap(block_ctx->pagev[i]);
+		if (!block_ctx->datav[i]) {
+			printk(KERN_INFO "btrfsic: kmap() failed (dev %s)!\n",
+			       block_ctx->dev->name);
+			return -1;
+		}
+	}
+
+	return block_ctx->len;
+}
+
+static void btrfsic_dump_database(struct btrfsic_state *state)
+{
+	struct list_head *elem_all;
+
+	BUG_ON(NULL == state);
+
+	printk(KERN_INFO "all_blocks_list:\n");
+	list_for_each(elem_all, &state->all_blocks_list) {
+		const struct btrfsic_block *const b_all =
+		    list_entry(elem_all, struct btrfsic_block,
+			       all_blocks_node);
+		struct list_head *elem_ref_to;
+		struct list_head *elem_ref_from;
+
+		printk(KERN_INFO "%c-block @%llu (%s/%llu/%d)\n",
+		       btrfsic_get_block_type(state, b_all),
+		       b_all->logical_bytenr, b_all->dev_state->name,
+		       b_all->dev_bytenr, b_all->mirror_num);
+
+		list_for_each(elem_ref_to, &b_all->ref_to_list) {
+			const struct btrfsic_block_link *const l =
+			    list_entry(elem_ref_to,
+				       struct btrfsic_block_link,
+				       node_ref_to);
+
+			printk(KERN_INFO " %c @%llu (%s/%llu/%d)"
+			       " refers %u* to"
+			       " %c @%llu (%s/%llu/%d)\n",
+			       btrfsic_get_block_type(state, b_all),
+			       b_all->logical_bytenr, b_all->dev_state->name,
+			       b_all->dev_bytenr, b_all->mirror_num,
+			       l->ref_cnt,
+			       btrfsic_get_block_type(state, l->block_ref_to),
+			       l->block_ref_to->logical_bytenr,
+			       l->block_ref_to->dev_state->name,
+			       l->block_ref_to->dev_bytenr,
+			       l->block_ref_to->mirror_num);
+		}
+
+		list_for_each(elem_ref_from, &b_all->ref_from_list) {
+			const struct btrfsic_block_link *const l =
+			    list_entry(elem_ref_from,
+				       struct btrfsic_block_link,
+				       node_ref_from);
+
+			printk(KERN_INFO " %c @%llu (%s/%llu/%d)"
+			       " is ref %u* from"
+			       " %c @%llu (%s/%llu/%d)\n",
+			       btrfsic_get_block_type(state, b_all),
+			       b_all->logical_bytenr, b_all->dev_state->name,
+			       b_all->dev_bytenr, b_all->mirror_num,
+			       l->ref_cnt,
+			       btrfsic_get_block_type(state, l->block_ref_from),
+			       l->block_ref_from->logical_bytenr,
+			       l->block_ref_from->dev_state->name,
+			       l->block_ref_from->dev_bytenr,
+			       l->block_ref_from->mirror_num);
+		}
+
+		printk(KERN_INFO "\n");
+	}
+}
+
+/*
+ * Test whether the disk block contains a tree block (leaf or node)
+ * (note that this test fails for the super block)
+ */
+static int btrfsic_test_for_metadata(struct btrfsic_state *state,
+				     char **datav, unsigned int num_pages)
+{
+	struct btrfs_header *h;
+	u8 csum[BTRFS_CSUM_SIZE];
+	u32 crc = ~(u32)0;
+	unsigned int i;
+
+	if (num_pages * PAGE_CACHE_SIZE < state->metablock_size)
+		return 1; /* not metadata */
+	num_pages = state->metablock_size >> PAGE_CACHE_SHIFT;
+	h = (struct btrfs_header *)datav[0];
+
+	if (memcmp(h->fsid, state->root->fs_info->fsid, BTRFS_UUID_SIZE))
+		return 1;
+
+	for (i = 0; i < num_pages; i++) {
+		u8 *data = i ? datav[i] : (datav[i] + BTRFS_CSUM_SIZE);
+		size_t sublen = i ? PAGE_CACHE_SIZE :
+				    (PAGE_CACHE_SIZE - BTRFS_CSUM_SIZE);
+
+		crc = btrfs_crc32c(crc, data, sublen);
+	}
+	btrfs_csum_final(crc, csum);
+	if (memcmp(csum, h->csum, state->csum_size))
+		return 1;
+
+	return 0; /* is metadata */
+}
+
+static void btrfsic_process_written_block(struct btrfsic_dev_state *dev_state,
+					  u64 dev_bytenr, char **mapped_datav,
+					  unsigned int num_pages,
+					  struct bio *bio, int *bio_is_patched,
+					  struct buffer_head *bh,
+					  int submit_bio_bh_rw)
+{
+	int is_metadata;
+	struct btrfsic_block *block;
+	struct btrfsic_block_data_ctx block_ctx;
+	int ret;
+	struct btrfsic_state *state = dev_state->state;
+	struct block_device *bdev = dev_state->bdev;
+	unsigned int processed_len;
+
+	if (NULL != bio_is_patched)
+		*bio_is_patched = 0;
+
+again:
+	if (num_pages == 0)
+		return;
+
+	processed_len = 0;
+	is_metadata = (0 == btrfsic_test_for_metadata(state, mapped_datav,
+						      num_pages));
+
+	block = btrfsic_block_hashtable_lookup(bdev, dev_bytenr,
+					       &state->block_hashtable);
+	if (NULL != block) {
+		u64 bytenr = 0;
+		struct list_head *elem_ref_to;
+		struct list_head *tmp_ref_to;
+
+		if (block->is_superblock) {
+			bytenr = btrfs_super_bytenr((struct btrfs_super_block *)
+						    mapped_datav[0]);
+			if (num_pages * PAGE_CACHE_SIZE <
+			    BTRFS_SUPER_INFO_SIZE) {
+				printk(KERN_INFO
+				       "btrfsic: cannot work with too short bios!\n");
+				return;
+			}
+			is_metadata = 1;
+			BUG_ON(BTRFS_SUPER_INFO_SIZE & (PAGE_CACHE_SIZE - 1));
+			processed_len = BTRFS_SUPER_INFO_SIZE;
+			if (state->print_mask &
+			    BTRFSIC_PRINT_MASK_TREE_BEFORE_SB_WRITE) {
+				printk(KERN_INFO
+				       "[before new superblock is written]:\n");
+				btrfsic_dump_tree_sub(state, block, 0);
+			}
+		}
+		if (is_metadata) {
+			if (!block->is_superblock) {
+				if (num_pages * PAGE_CACHE_SIZE <
+				    state->metablock_size) {
+					printk(KERN_INFO
+					       "btrfsic: cannot work with too short bios!\n");
+					return;
+				}
+				processed_len = state->metablock_size;
+				bytenr = btrfs_stack_header_bytenr(
+						(struct btrfs_header *)
+						mapped_datav[0]);
+				btrfsic_cmp_log_and_dev_bytenr(state, bytenr,
+							       dev_state,
+							       dev_bytenr);
+			}
+			if (block->logical_bytenr != bytenr &&
+			    !(!block->is_metadata &&
+			      block->logical_bytenr == 0))
+				printk(KERN_INFO
+				       "Written block @%llu (%s/%llu/%d)"
+				       " found in hash table, %c,"
+				       " bytenr mismatch"
+				       " (!= stored %llu).\n",
+				       bytenr, dev_state->name, dev_bytenr,
+				       block->mirror_num,
+				       btrfsic_get_block_type(state, block),
+				       block->logical_bytenr);
+			else if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
+				printk(KERN_INFO
+				       "Written block @%llu (%s/%llu/%d)"
+				       " found in hash table, %c.\n",
+				       bytenr, dev_state->name, dev_bytenr,
+				       block->mirror_num,
+				       btrfsic_get_block_type(state, block));
+			block->logical_bytenr = bytenr;
+		} else {
+			if (num_pages * PAGE_CACHE_SIZE <
+			    state->datablock_size) {
+				printk(KERN_INFO
+				       "btrfsic: cannot work with too short bios!\n");
+				return;
+			}
+			processed_len = state->datablock_size;
+			bytenr = block->logical_bytenr;
+			if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
+				printk(KERN_INFO
+				       "Written block @%llu (%s/%llu/%d)"
+				       " found in hash table, %c.\n",
+				       bytenr, dev_state->name, dev_bytenr,
+				       block->mirror_num,
+				       btrfsic_get_block_type(state, block));
+		}
+
+		if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
+			printk(KERN_INFO
+			       "ref_to_list: %cE, ref_from_list: %cE\n",
+			       list_empty(&block->ref_to_list) ? ' ' : '!',
+			       list_empty(&block->ref_from_list) ? ' ' : '!');
+		if (btrfsic_is_block_ref_by_superblock(state, block, 0)) {
+			printk(KERN_INFO "btrfs: attempt to overwrite %c-block"
+			       " @%llu (%s/%llu/%d), old(gen=%llu,"
+			       " objectid=%llu, type=%d, offset=%llu),"
+			       " new(gen=%llu),"
+			       " which is referenced by most recent superblock"
+			       " (superblockgen=%llu)!\n",
+			       btrfsic_get_block_type(state, block), bytenr,
+			       dev_state->name, dev_bytenr, block->mirror_num,
+			       block->generation,
+			       btrfs_disk_key_objectid(&block->disk_key),
+			       block->disk_key.type,
+			       btrfs_disk_key_offset(&block->disk_key),
+			       btrfs_stack_header_generation(
+				       (struct btrfs_header *) mapped_datav[0]),
+			       state->max_superblock_generation);
+			btrfsic_dump_tree(state);
+		}
+
+		if (!block->is_iodone && !block->never_written) {
+			printk(KERN_INFO "btrfs: attempt to overwrite %c-block"
+			       " @%llu (%s/%llu/%d), oldgen=%llu, newgen=%llu,"
+			       " which is not yet iodone!\n",
+			       btrfsic_get_block_type(state, block), bytenr,
+			       dev_state->name, dev_bytenr, block->mirror_num,
+			       block->generation,
+			       btrfs_stack_header_generation(
+				       (struct btrfs_header *)
+				       mapped_datav[0]));
+			/* it would not be safe to go on */
+			btrfsic_dump_tree(state);
+			goto continue_loop;
+		}
+
+		/*
+		 * Clear all references of this block. Do not free
+		 * the block itself even if is not referenced anymore
+		 * because it still carries valueable information
+		 * like whether it was ever written and IO completed.
+		 */
+		list_for_each_safe(elem_ref_to, tmp_ref_to,
+				   &block->ref_to_list) {
+			struct btrfsic_block_link *const l =
+			    list_entry(elem_ref_to,
+				       struct btrfsic_block_link,
+				       node_ref_to);
+
+			if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
+				btrfsic_print_rem_link(state, l);
+			l->ref_cnt--;
+			if (0 == l->ref_cnt) {
+				list_del(&l->node_ref_to);
+				list_del(&l->node_ref_from);
+				btrfsic_block_link_hashtable_remove(l);
+				btrfsic_block_link_free(l);
+			}
+		}
+
+		if (block->is_superblock)
+			ret = btrfsic_map_superblock(state, bytenr,
+						     processed_len,
+						     bdev, &block_ctx);
+		else
+			ret = btrfsic_map_block(state, bytenr, processed_len,
+						&block_ctx, 0);
+		if (ret) {
+			printk(KERN_INFO
+			       "btrfsic: btrfsic_map_block(root @%llu)"
+			       " failed!\n", bytenr);
+			goto continue_loop;
+		}
+		block_ctx.datav = mapped_datav;
+		/* the following is required in case of writes to mirrors,
+		 * use the same that was used for the lookup */
+		block_ctx.dev = dev_state;
+		block_ctx.dev_bytenr = dev_bytenr;
+
+		if (is_metadata || state->include_extent_data) {
+			block->never_written = 0;
+			block->iodone_w_error = 0;
+			if (NULL != bio) {
+				block->is_iodone = 0;
+				BUG_ON(NULL == bio_is_patched);
+				if (!*bio_is_patched) {
+					block->orig_bio_bh_private =
+					    bio->bi_private;
+					block->orig_bio_bh_end_io.bio =
+					    bio->bi_end_io;
+					block->next_in_same_bio = NULL;
+					bio->bi_private = block;
+					bio->bi_end_io = btrfsic_bio_end_io;
+					*bio_is_patched = 1;
+				} else {
+					struct btrfsic_block *chained_block =
+					    (struct btrfsic_block *)
+					    bio->bi_private;
+
+					BUG_ON(NULL == chained_block);
+					block->orig_bio_bh_private =
+					    chained_block->orig_bio_bh_private;
+					block->orig_bio_bh_end_io.bio =
+					    chained_block->orig_bio_bh_end_io.
+					    bio;
+					block->next_in_same_bio = chained_block;
+					bio->bi_private = block;
+				}
+			} else if (NULL != bh) {
+				block->is_iodone = 0;
+				block->orig_bio_bh_private = bh->b_private;
+				block->orig_bio_bh_end_io.bh = bh->b_end_io;
+				block->next_in_same_bio = NULL;
+				bh->b_private = block;
+				bh->b_end_io = btrfsic_bh_end_io;
+			} else {
+				block->is_iodone = 1;
+				block->orig_bio_bh_private = NULL;
+				block->orig_bio_bh_end_io.bio = NULL;
+				block->next_in_same_bio = NULL;
+			}
+		}
+
+		block->flush_gen = dev_state->last_flush_gen + 1;
+		block->submit_bio_bh_rw = submit_bio_bh_rw;
+		if (is_metadata) {
+			block->logical_bytenr = bytenr;
+			block->is_metadata = 1;
+			if (block->is_superblock) {
+				BUG_ON(PAGE_CACHE_SIZE !=
+				       BTRFS_SUPER_INFO_SIZE);
+				ret = btrfsic_process_written_superblock(
+						state,
+						block,
+						(struct btrfs_super_block *)
+						mapped_datav[0]);
+				if (state->print_mask &
+				    BTRFSIC_PRINT_MASK_TREE_AFTER_SB_WRITE) {
+					printk(KERN_INFO
+					"[after new superblock is written]:\n");
+					btrfsic_dump_tree_sub(state, block, 0);
+				}
+			} else {
+				block->mirror_num = 0;	/* unknown */
+				ret = btrfsic_process_metablock(
+						state,
+						block,
+						&block_ctx,
+						0, 0);
+			}
+			if (ret)
+				printk(KERN_INFO
+				       "btrfsic: btrfsic_process_metablock"
+				       "(root @%llu) failed!\n",
+				       dev_bytenr);
+		} else {
+			block->is_metadata = 0;
+			block->mirror_num = 0;	/* unknown */
+			block->generation = BTRFSIC_GENERATION_UNKNOWN;
+			if (!state->include_extent_data
+			    && list_empty(&block->ref_from_list)) {
+				/*
+				 * disk block is overwritten with extent
+				 * data (not meta data) and we are configured
+				 * to not include extent data: take the
+				 * chance and free the block's memory
+				 */
+				btrfsic_block_hashtable_remove(block);
+				list_del(&block->all_blocks_node);
+				btrfsic_block_free(block);
+			}
+		}
+		btrfsic_release_block_ctx(&block_ctx);
+	} else {
+		/* block has not been found in hash table */
+		u64 bytenr;
+
+		if (!is_metadata) {
+			processed_len = state->datablock_size;
+			if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
+				printk(KERN_INFO "Written block (%s/%llu/?)"
+				       " !found in hash table, D.\n",
+				       dev_state->name, dev_bytenr);
+			if (!state->include_extent_data) {
+				/* ignore that written D block */
+				goto continue_loop;
+			}
+
+			/* this is getting ugly for the
+			 * include_extent_data case... */
+			bytenr = 0;	/* unknown */
+			block_ctx.start = bytenr;
+			block_ctx.len = processed_len;
+			block_ctx.mem_to_free = NULL;
+			block_ctx.pagev = NULL;
+		} else {
+			processed_len = state->metablock_size;
+			bytenr = btrfs_stack_header_bytenr(
+					(struct btrfs_header *)
+					mapped_datav[0]);
+			btrfsic_cmp_log_and_dev_bytenr(state, bytenr, dev_state,
+						       dev_bytenr);
+			if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
+				printk(KERN_INFO
+				       "Written block @%llu (%s/%llu/?)"
+				       " !found in hash table, M.\n",
+				       bytenr, dev_state->name, dev_bytenr);
+
+			ret = btrfsic_map_block(state, bytenr, processed_len,
+						&block_ctx, 0);
+			if (ret) {
+				printk(KERN_INFO
+				       "btrfsic: btrfsic_map_block(root @%llu)"
+				       " failed!\n",
+				       dev_bytenr);
+				goto continue_loop;
+			}
+		}
+		block_ctx.datav = mapped_datav;
+		/* the following is required in case of writes to mirrors,
+		 * use the same that was used for the lookup */
+		block_ctx.dev = dev_state;
+		block_ctx.dev_bytenr = dev_bytenr;
+
+		block = btrfsic_block_alloc();
+		if (NULL == block) {
+			printk(KERN_INFO "btrfsic: error, kmalloc failed!\n");
+			btrfsic_release_block_ctx(&block_ctx);
+			goto continue_loop;
+		}
+		block->dev_state = dev_state;
+		block->dev_bytenr = dev_bytenr;
+		block->logical_bytenr = bytenr;
+		block->is_metadata = is_metadata;
+		block->never_written = 0;
+		block->iodone_w_error = 0;
+		block->mirror_num = 0;	/* unknown */
+		block->flush_gen = dev_state->last_flush_gen + 1;
+		block->submit_bio_bh_rw = submit_bio_bh_rw;
+		if (NULL != bio) {
+			block->is_iodone = 0;
+			BUG_ON(NULL == bio_is_patched);
+			if (!*bio_is_patched) {
+				block->orig_bio_bh_private = bio->bi_private;
+				block->orig_bio_bh_end_io.bio = bio->bi_end_io;
+				block->next_in_same_bio = NULL;
+				bio->bi_private = block;
+				bio->bi_end_io = btrfsic_bio_end_io;
+				*bio_is_patched = 1;
+			} else {
+				struct btrfsic_block *chained_block =
+				    (struct btrfsic_block *)
+				    bio->bi_private;
+
+				BUG_ON(NULL == chained_block);
+				block->orig_bio_bh_private =
+				    chained_block->orig_bio_bh_private;
+				block->orig_bio_bh_end_io.bio =
+				    chained_block->orig_bio_bh_end_io.bio;
+				block->next_in_same_bio = chained_block;
+				bio->bi_private = block;
+			}
+		} else if (NULL != bh) {
+			block->is_iodone = 0;
+			block->orig_bio_bh_private = bh->b_private;
+			block->orig_bio_bh_end_io.bh = bh->b_end_io;
+			block->next_in_same_bio = NULL;
+			bh->b_private = block;
+			bh->b_end_io = btrfsic_bh_end_io;
+		} else {
+			block->is_iodone = 1;
+			block->orig_bio_bh_private = NULL;
+			block->orig_bio_bh_end_io.bio = NULL;
+			block->next_in_same_bio = NULL;
+		}
+		if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
+			printk(KERN_INFO
+			       "New written %c-block @%llu (%s/%llu/%d)\n",
+			       is_metadata ? 'M' : 'D',
+			       block->logical_bytenr, block->dev_state->name,
+			       block->dev_bytenr, block->mirror_num);
+		list_add(&block->all_blocks_node, &state->all_blocks_list);
+		btrfsic_block_hashtable_add(block, &state->block_hashtable);
+
+		if (is_metadata) {
+			ret = btrfsic_process_metablock(state, block,
+							&block_ctx, 0, 0);
+			if (ret)
+				printk(KERN_INFO
+				       "btrfsic: process_metablock(root @%llu)"
+				       " failed!\n",
+				       dev_bytenr);
+		}
+		btrfsic_release_block_ctx(&block_ctx);
+	}
+
+continue_loop:
+	BUG_ON(!processed_len);
+	dev_bytenr += processed_len;
+	mapped_datav += processed_len >> PAGE_CACHE_SHIFT;
+	num_pages -= processed_len >> PAGE_CACHE_SHIFT;
+	goto again;
+}
+
+static void btrfsic_bio_end_io(struct bio *bp, int bio_error_status)
+{
+	struct btrfsic_block *block = (struct btrfsic_block *)bp->bi_private;
+	int iodone_w_error;
+
+	/* mutex is not held! This is not save if IO is not yet completed
+	 * on umount */
+	iodone_w_error = 0;
+	if (bio_error_status)
+		iodone_w_error = 1;
+
+	BUG_ON(NULL == block);
+	bp->bi_private = block->orig_bio_bh_private;
+	bp->bi_end_io = block->orig_bio_bh_end_io.bio;
+
+	do {
+		struct btrfsic_block *next_block;
+		struct btrfsic_dev_state *const dev_state = block->dev_state;
+
+		if ((dev_state->state->print_mask &
+		     BTRFSIC_PRINT_MASK_END_IO_BIO_BH))
+			printk(KERN_INFO
+			       "bio_end_io(err=%d) for %c @%llu (%s/%llu/%d)\n",
+			       bio_error_status,
+			       btrfsic_get_block_type(dev_state->state, block),
+			       block->logical_bytenr, dev_state->name,
+			       block->dev_bytenr, block->mirror_num);
+		next_block = block->next_in_same_bio;
+		block->iodone_w_error = iodone_w_error;
+		if (block->submit_bio_bh_rw & REQ_FLUSH) {
+			dev_state->last_flush_gen++;
+			if ((dev_state->state->print_mask &
+			     BTRFSIC_PRINT_MASK_END_IO_BIO_BH))
+				printk(KERN_INFO
+				       "bio_end_io() new %s flush_gen=%llu\n",
+				       dev_state->name,
+				       dev_state->last_flush_gen);
+		}
+		if (block->submit_bio_bh_rw & REQ_FUA)
+			block->flush_gen = 0; /* FUA completed means block is
+					       * on disk */
+		block->is_iodone = 1; /* for FLUSH, this releases the block */
+		block = next_block;
+	} while (NULL != block);
+
+	bp->bi_end_io(bp, bio_error_status);
+}
+
+static void btrfsic_bh_end_io(struct buffer_head *bh, int uptodate)
+{
+	struct btrfsic_block *block = (struct btrfsic_block *)bh->b_private;
+	int iodone_w_error = !uptodate;
+	struct btrfsic_dev_state *dev_state;
+
+	BUG_ON(NULL == block);
+	dev_state = block->dev_state;
+	if ((dev_state->state->print_mask & BTRFSIC_PRINT_MASK_END_IO_BIO_BH))
+		printk(KERN_INFO
+		       "bh_end_io(error=%d) for %c @%llu (%s/%llu/%d)\n",
+		       iodone_w_error,
+		       btrfsic_get_block_type(dev_state->state, block),
+		       block->logical_bytenr, block->dev_state->name,
+		       block->dev_bytenr, block->mirror_num);
+
+	block->iodone_w_error = iodone_w_error;
+	if (block->submit_bio_bh_rw & REQ_FLUSH) {
+		dev_state->last_flush_gen++;
+		if ((dev_state->state->print_mask &
+		     BTRFSIC_PRINT_MASK_END_IO_BIO_BH))
+			printk(KERN_INFO
+			       "bh_end_io() new %s flush_gen=%llu\n",
+			       dev_state->name, dev_state->last_flush_gen);
+	}
+	if (block->submit_bio_bh_rw & REQ_FUA)
+		block->flush_gen = 0; /* FUA completed means block is on disk */
+
+	bh->b_private = block->orig_bio_bh_private;
+	bh->b_end_io = block->orig_bio_bh_end_io.bh;
+	block->is_iodone = 1; /* for FLUSH, this releases the block */
+	bh->b_end_io(bh, uptodate);
+}
+
+static int btrfsic_process_written_superblock(
+		struct btrfsic_state *state,
+		struct btrfsic_block *const superblock,
+		struct btrfs_super_block *const super_hdr)
+{
+	int pass;
+
+	superblock->generation = btrfs_super_generation(super_hdr);
+	if (!(superblock->generation > state->max_superblock_generation ||
+	      0 == state->max_superblock_generation)) {
+		if (state->print_mask & BTRFSIC_PRINT_MASK_SUPERBLOCK_WRITE)
+			printk(KERN_INFO
+			       "btrfsic: superblock @%llu (%s/%llu/%d)"
+			       " with old gen %llu <= %llu\n",
+			       superblock->logical_bytenr,
+			       superblock->dev_state->name,
+			       superblock->dev_bytenr, superblock->mirror_num,
+			       btrfs_super_generation(super_hdr),
+			       state->max_superblock_generation);
+	} else {
+		if (state->print_mask & BTRFSIC_PRINT_MASK_SUPERBLOCK_WRITE)
+			printk(KERN_INFO
+			       "btrfsic: got new superblock @%llu (%s/%llu/%d)"
+			       " with new gen %llu > %llu\n",
+			       superblock->logical_bytenr,
+			       superblock->dev_state->name,
+			       superblock->dev_bytenr, superblock->mirror_num,
+			       btrfs_super_generation(super_hdr),
+			       state->max_superblock_generation);
+
+		state->max_superblock_generation =
+		    btrfs_super_generation(super_hdr);
+		state->latest_superblock = superblock;
+	}
+
+	for (pass = 0; pass < 3; pass++) {
+		int ret;
+		u64 next_bytenr;
+		struct btrfsic_block *next_block;
+		struct btrfsic_block_data_ctx tmp_next_block_ctx;
+		struct btrfsic_block_link *l;
+		int num_copies;
+		int mirror_num;
+		const char *additional_string = NULL;
+		struct btrfs_disk_key tmp_disk_key = {0};
+
+		btrfs_set_disk_key_objectid(&tmp_disk_key,
+					    BTRFS_ROOT_ITEM_KEY);
+		btrfs_set_disk_key_objectid(&tmp_disk_key, 0);
+
+		switch (pass) {
+		case 0:
+			btrfs_set_disk_key_objectid(&tmp_disk_key,
+						    BTRFS_ROOT_TREE_OBJECTID);
+			additional_string = "root ";
+			next_bytenr = btrfs_super_root(super_hdr);
+			if (state->print_mask &
+			    BTRFSIC_PRINT_MASK_ROOT_CHUNK_LOG_TREE_LOCATION)
+				printk(KERN_INFO "root@%llu\n", next_bytenr);
+			break;
+		case 1:
+			btrfs_set_disk_key_objectid(&tmp_disk_key,
+						    BTRFS_CHUNK_TREE_OBJECTID);
+			additional_string = "chunk ";
+			next_bytenr = btrfs_super_chunk_root(super_hdr);
+			if (state->print_mask &
+			    BTRFSIC_PRINT_MASK_ROOT_CHUNK_LOG_TREE_LOCATION)
+				printk(KERN_INFO "chunk@%llu\n", next_bytenr);
+			break;
+		case 2:
+			btrfs_set_disk_key_objectid(&tmp_disk_key,
+						    BTRFS_TREE_LOG_OBJECTID);
+			additional_string = "log ";
+			next_bytenr = btrfs_super_log_root(super_hdr);
+			if (0 == next_bytenr)
+				continue;
+			if (state->print_mask &
+			    BTRFSIC_PRINT_MASK_ROOT_CHUNK_LOG_TREE_LOCATION)
+				printk(KERN_INFO "log@%llu\n", next_bytenr);
+			break;
+		}
+
+		num_copies =
+		    btrfs_num_copies(state->root->fs_info,
+				     next_bytenr, BTRFS_SUPER_INFO_SIZE);
+		if (state->print_mask & BTRFSIC_PRINT_MASK_NUM_COPIES)
+			printk(KERN_INFO "num_copies(log_bytenr=%llu) = %d\n",
+			       next_bytenr, num_copies);
+		for (mirror_num = 1; mirror_num <= num_copies; mirror_num++) {
+			int was_created;
+
+			if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
+				printk(KERN_INFO
+				       "btrfsic_process_written_superblock("
+				       "mirror_num=%d)\n", mirror_num);
+			ret = btrfsic_map_block(state, next_bytenr,
+						BTRFS_SUPER_INFO_SIZE,
+						&tmp_next_block_ctx,
+						mirror_num);
+			if (ret) {
+				printk(KERN_INFO
+				       "btrfsic: btrfsic_map_block(@%llu,"
+				       " mirror=%d) failed!\n",
+				       next_bytenr, mirror_num);
+				return -1;
+			}
+
+			next_block = btrfsic_block_lookup_or_add(
+					state,
+					&tmp_next_block_ctx,
+					additional_string,
+					1, 0, 1,
+					mirror_num,
+					&was_created);
+			if (NULL == next_block) {
+				printk(KERN_INFO
+				       "btrfsic: error, kmalloc failed!\n");
+				btrfsic_release_block_ctx(&tmp_next_block_ctx);
+				return -1;
+			}
+
+			next_block->disk_key = tmp_disk_key;
+			if (was_created)
+				next_block->generation =
+				    BTRFSIC_GENERATION_UNKNOWN;
+			l = btrfsic_block_link_lookup_or_add(
+					state,
+					&tmp_next_block_ctx,
+					next_block,
+					superblock,
+					BTRFSIC_GENERATION_UNKNOWN);
+			btrfsic_release_block_ctx(&tmp_next_block_ctx);
+			if (NULL == l)
+				return -1;
+		}
+	}
+
+	if (WARN_ON(-1 == btrfsic_check_all_ref_blocks(state, superblock, 0)))
+		btrfsic_dump_tree(state);
+
+	return 0;
+}
+
+static int btrfsic_check_all_ref_blocks(struct btrfsic_state *state,
+					struct btrfsic_block *const block,
+					int recursion_level)
+{
+	struct list_head *elem_ref_to;
+	int ret = 0;
+
+	if (recursion_level >= 3 + BTRFS_MAX_LEVEL) {
+		/*
+		 * Note that this situation can happen and does not
+		 * indicate an error in regular cases. It happens
+		 * when disk blocks are freed and later reused.
+		 * The check-integrity module is not aware of any
+		 * block free operations, it just recognizes block
+		 * write operations. Therefore it keeps the linkage
+		 * information for a block until a block is
+		 * rewritten. This can temporarily cause incorrect
+		 * and even circular linkage informations. This
+		 * causes no harm unless such blocks are referenced
+		 * by the most recent super block.
+		 */
+		if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
+			printk(KERN_INFO
+			       "btrfsic: abort cyclic linkage (case 1).\n");
+
+		return ret;
+	}
+
+	/*
+	 * This algorithm is recursive because the amount of used stack
+	 * space is very small and the max recursion depth is limited.
+	 */
+	list_for_each(elem_ref_to, &block->ref_to_list) {
+		const struct btrfsic_block_link *const l =
+		    list_entry(elem_ref_to, struct btrfsic_block_link,
+			       node_ref_to);
+
+		if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
+			printk(KERN_INFO
+			       "rl=%d, %c @%llu (%s/%llu/%d)"
+			       " %u* refers to %c @%llu (%s/%llu/%d)\n",
+			       recursion_level,
+			       btrfsic_get_block_type(state, block),
+			       block->logical_bytenr, block->dev_state->name,
+			       block->dev_bytenr, block->mirror_num,
+			       l->ref_cnt,
+			       btrfsic_get_block_type(state, l->block_ref_to),
+			       l->block_ref_to->logical_bytenr,
+			       l->block_ref_to->dev_state->name,
+			       l->block_ref_to->dev_bytenr,
+			       l->block_ref_to->mirror_num);
+		if (l->block_ref_to->never_written) {
+			printk(KERN_INFO "btrfs: attempt to write superblock"
+			       " which references block %c @%llu (%s/%llu/%d)"
+			       " which is never written!\n",
+			       btrfsic_get_block_type(state, l->block_ref_to),
+			       l->block_ref_to->logical_bytenr,
+			       l->block_ref_to->dev_state->name,
+			       l->block_ref_to->dev_bytenr,
+			       l->block_ref_to->mirror_num);
+			ret = -1;
+		} else if (!l->block_ref_to->is_iodone) {
+			printk(KERN_INFO "btrfs: attempt to write superblock"
+			       " which references block %c @%llu (%s/%llu/%d)"
+			       " which is not yet iodone!\n",
+			       btrfsic_get_block_type(state, l->block_ref_to),
+			       l->block_ref_to->logical_bytenr,
+			       l->block_ref_to->dev_state->name,
+			       l->block_ref_to->dev_bytenr,
+			       l->block_ref_to->mirror_num);
+			ret = -1;
+		} else if (l->block_ref_to->iodone_w_error) {
+			printk(KERN_INFO "btrfs: attempt to write superblock"
+			       " which references block %c @%llu (%s/%llu/%d)"
+			       " which has write error!\n",
+			       btrfsic_get_block_type(state, l->block_ref_to),
+			       l->block_ref_to->logical_bytenr,
+			       l->block_ref_to->dev_state->name,
+			       l->block_ref_to->dev_bytenr,
+			       l->block_ref_to->mirror_num);
+			ret = -1;
+		} else if (l->parent_generation !=
+			   l->block_ref_to->generation &&
+			   BTRFSIC_GENERATION_UNKNOWN !=
+			   l->parent_generation &&
+			   BTRFSIC_GENERATION_UNKNOWN !=
+			   l->block_ref_to->generation) {
+			printk(KERN_INFO "btrfs: attempt to write superblock"
+			       " which references block %c @%llu (%s/%llu/%d)"
+			       " with generation %llu !="
+			       " parent generation %llu!\n",
+			       btrfsic_get_block_type(state, l->block_ref_to),
+			       l->block_ref_to->logical_bytenr,
+			       l->block_ref_to->dev_state->name,
+			       l->block_ref_to->dev_bytenr,
+			       l->block_ref_to->mirror_num,
+			       l->block_ref_to->generation,
+			       l->parent_generation);
+			ret = -1;
+		} else if (l->block_ref_to->flush_gen >
+			   l->block_ref_to->dev_state->last_flush_gen) {
+			printk(KERN_INFO "btrfs: attempt to write superblock"
+			       " which references block %c @%llu (%s/%llu/%d)"
+			       " which is not flushed out of disk's write cache"
+			       " (block flush_gen=%llu,"
+			       " dev->flush_gen=%llu)!\n",
+			       btrfsic_get_block_type(state, l->block_ref_to),
+			       l->block_ref_to->logical_bytenr,
+			       l->block_ref_to->dev_state->name,
+			       l->block_ref_to->dev_bytenr,
+			       l->block_ref_to->mirror_num, block->flush_gen,
+			       l->block_ref_to->dev_state->last_flush_gen);
+			ret = -1;
+		} else if (-1 == btrfsic_check_all_ref_blocks(state,
+							      l->block_ref_to,
+							      recursion_level +
+							      1)) {
+			ret = -1;
+		}
+	}
+
+	return ret;
+}
+
+static int btrfsic_is_block_ref_by_superblock(
+		const struct btrfsic_state *state,
+		const struct btrfsic_block *block,
+		int recursion_level)
+{
+	struct list_head *elem_ref_from;
+
+	if (recursion_level >= 3 + BTRFS_MAX_LEVEL) {
+		/* refer to comment at "abort cyclic linkage (case 1)" */
+		if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
+			printk(KERN_INFO
+			       "btrfsic: abort cyclic linkage (case 2).\n");
+
+		return 0;
+	}
+
+	/*
+	 * This algorithm is recursive because the amount of used stack space
+	 * is very small and the max recursion depth is limited.
+	 */
+	list_for_each(elem_ref_from, &block->ref_from_list) {
+		const struct btrfsic_block_link *const l =
+		    list_entry(elem_ref_from, struct btrfsic_block_link,
+			       node_ref_from);
+
+		if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
+			printk(KERN_INFO
+			       "rl=%d, %c @%llu (%s/%llu/%d)"
+			       " is ref %u* from %c @%llu (%s/%llu/%d)\n",
+			       recursion_level,
+			       btrfsic_get_block_type(state, block),
+			       block->logical_bytenr, block->dev_state->name,
+			       block->dev_bytenr, block->mirror_num,
+			       l->ref_cnt,
+			       btrfsic_get_block_type(state, l->block_ref_from),
+			       l->block_ref_from->logical_bytenr,
+			       l->block_ref_from->dev_state->name,
+			       l->block_ref_from->dev_bytenr,
+			       l->block_ref_from->mirror_num);
+		if (l->block_ref_from->is_superblock &&
+		    state->latest_superblock->dev_bytenr ==
+		    l->block_ref_from->dev_bytenr &&
+		    state->latest_superblock->dev_state->bdev ==
+		    l->block_ref_from->dev_state->bdev)
+			return 1;
+		else if (btrfsic_is_block_ref_by_superblock(state,
+							    l->block_ref_from,
+							    recursion_level +
+							    1))
+			return 1;
+	}
+
+	return 0;
+}
+
+static void btrfsic_print_add_link(const struct btrfsic_state *state,
+				   const struct btrfsic_block_link *l)
+{
+	printk(KERN_INFO
+	       "Add %u* link from %c @%llu (%s/%llu/%d)"
+	       " to %c @%llu (%s/%llu/%d).\n",
+	       l->ref_cnt,
+	       btrfsic_get_block_type(state, l->block_ref_from),
+	       l->block_ref_from->logical_bytenr,
+	       l->block_ref_from->dev_state->name,
+	       l->block_ref_from->dev_bytenr, l->block_ref_from->mirror_num,
+	       btrfsic_get_block_type(state, l->block_ref_to),
+	       l->block_ref_to->logical_bytenr,
+	       l->block_ref_to->dev_state->name, l->block_ref_to->dev_bytenr,
+	       l->block_ref_to->mirror_num);
+}
+
+static void btrfsic_print_rem_link(const struct btrfsic_state *state,
+				   const struct btrfsic_block_link *l)
+{
+	printk(KERN_INFO
+	       "Rem %u* link from %c @%llu (%s/%llu/%d)"
+	       " to %c @%llu (%s/%llu/%d).\n",
+	       l->ref_cnt,
+	       btrfsic_get_block_type(state, l->block_ref_from),
+	       l->block_ref_from->logical_bytenr,
+	       l->block_ref_from->dev_state->name,
+	       l->block_ref_from->dev_bytenr, l->block_ref_from->mirror_num,
+	       btrfsic_get_block_type(state, l->block_ref_to),
+	       l->block_ref_to->logical_bytenr,
+	       l->block_ref_to->dev_state->name, l->block_ref_to->dev_bytenr,
+	       l->block_ref_to->mirror_num);
+}
+
+static char btrfsic_get_block_type(const struct btrfsic_state *state,
+				   const struct btrfsic_block *block)
+{
+	if (block->is_superblock &&
+	    state->latest_superblock->dev_bytenr == block->dev_bytenr &&
+	    state->latest_superblock->dev_state->bdev == block->dev_state->bdev)
+		return 'S';
+	else if (block->is_superblock)
+		return 's';
+	else if (block->is_metadata)
+		return 'M';
+	else
+		return 'D';
+}
+
+static void btrfsic_dump_tree(const struct btrfsic_state *state)
+{
+	btrfsic_dump_tree_sub(state, state->latest_superblock, 0);
+}
+
+static void btrfsic_dump_tree_sub(const struct btrfsic_state *state,
+				  const struct btrfsic_block *block,
+				  int indent_level)
+{
+	struct list_head *elem_ref_to;
+	int indent_add;
+	static char buf[80];
+	int cursor_position;
+
+	/*
+	 * Should better fill an on-stack buffer with a complete line and
+	 * dump it at once when it is time to print a newline character.
+	 */
+
+	/*
+	 * This algorithm is recursive because the amount of used stack space
+	 * is very small and the max recursion depth is limited.
+	 */
+	indent_add = sprintf(buf, "%c-%llu(%s/%llu/%d)",
+			     btrfsic_get_block_type(state, block),
+			     block->logical_bytenr, block->dev_state->name,
+			     block->dev_bytenr, block->mirror_num);
+	if (indent_level + indent_add > BTRFSIC_TREE_DUMP_MAX_INDENT_LEVEL) {
+		printk("[...]\n");
+		return;
+	}
+	printk(buf);
+	indent_level += indent_add;
+	if (list_empty(&block->ref_to_list)) {
+		printk("\n");
+		return;
+	}
+	if (block->mirror_num > 1 &&
+	    !(state->print_mask & BTRFSIC_PRINT_MASK_TREE_WITH_ALL_MIRRORS)) {
+		printk(" [...]\n");
+		return;
+	}
+
+	cursor_position = indent_level;
+	list_for_each(elem_ref_to, &block->ref_to_list) {
+		const struct btrfsic_block_link *const l =
+		    list_entry(elem_ref_to, struct btrfsic_block_link,
+			       node_ref_to);
+
+		while (cursor_position < indent_level) {
+			printk(" ");
+			cursor_position++;
+		}
+		if (l->ref_cnt > 1)
+			indent_add = sprintf(buf, " %d*--> ", l->ref_cnt);
+		else
+			indent_add = sprintf(buf, " --> ");
+		if (indent_level + indent_add >
+		    BTRFSIC_TREE_DUMP_MAX_INDENT_LEVEL) {
+			printk("[...]\n");
+			cursor_position = 0;
+			continue;
+		}
+
+		printk(buf);
+
+		btrfsic_dump_tree_sub(state, l->block_ref_to,
+				      indent_level + indent_add);
+		cursor_position = 0;
+	}
+}
+
+static struct btrfsic_block_link *btrfsic_block_link_lookup_or_add(
+		struct btrfsic_state *state,
+		struct btrfsic_block_data_ctx *next_block_ctx,
+		struct btrfsic_block *next_block,
+		struct btrfsic_block *from_block,
+		u64 parent_generation)
+{
+	struct btrfsic_block_link *l;
+
+	l = btrfsic_block_link_hashtable_lookup(next_block_ctx->dev->bdev,
+						next_block_ctx->dev_bytenr,
+						from_block->dev_state->bdev,
+						from_block->dev_bytenr,
+						&state->block_link_hashtable);
+	if (NULL == l) {
+		l = btrfsic_block_link_alloc();
+		if (NULL == l) {
+			printk(KERN_INFO
+			       "btrfsic: error, kmalloc" " failed!\n");
+			return NULL;
+		}
+
+		l->block_ref_to = next_block;
+		l->block_ref_from = from_block;
+		l->ref_cnt = 1;
+		l->parent_generation = parent_generation;
+
+		if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
+			btrfsic_print_add_link(state, l);
+
+		list_add(&l->node_ref_to, &from_block->ref_to_list);
+		list_add(&l->node_ref_from, &next_block->ref_from_list);
+
+		btrfsic_block_link_hashtable_add(l,
+						 &state->block_link_hashtable);
+	} else {
+		l->ref_cnt++;
+		l->parent_generation = parent_generation;
+		if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
+			btrfsic_print_add_link(state, l);
+	}
+
+	return l;
+}
+
+static struct btrfsic_block *btrfsic_block_lookup_or_add(
+		struct btrfsic_state *state,
+		struct btrfsic_block_data_ctx *block_ctx,
+		const char *additional_string,
+		int is_metadata,
+		int is_iodone,
+		int never_written,
+		int mirror_num,
+		int *was_created)
+{
+	struct btrfsic_block *block;
+
+	block = btrfsic_block_hashtable_lookup(block_ctx->dev->bdev,
+					       block_ctx->dev_bytenr,
+					       &state->block_hashtable);
+	if (NULL == block) {
+		struct btrfsic_dev_state *dev_state;
+
+		block = btrfsic_block_alloc();
+		if (NULL == block) {
+			printk(KERN_INFO "btrfsic: error, kmalloc failed!\n");
+			return NULL;
+		}
+		dev_state = btrfsic_dev_state_lookup(block_ctx->dev->bdev);
+		if (NULL == dev_state) {
+			printk(KERN_INFO
+			       "btrfsic: error, lookup dev_state failed!\n");
+			btrfsic_block_free(block);
+			return NULL;
+		}
+		block->dev_state = dev_state;
+		block->dev_bytenr = block_ctx->dev_bytenr;
+		block->logical_bytenr = block_ctx->start;
+		block->is_metadata = is_metadata;
+		block->is_iodone = is_iodone;
+		block->never_written = never_written;
+		block->mirror_num = mirror_num;
+		if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
+			printk(KERN_INFO
+			       "New %s%c-block @%llu (%s/%llu/%d)\n",
+			       additional_string,
+			       btrfsic_get_block_type(state, block),
+			       block->logical_bytenr, dev_state->name,
+			       block->dev_bytenr, mirror_num);
+		list_add(&block->all_blocks_node, &state->all_blocks_list);
+		btrfsic_block_hashtable_add(block, &state->block_hashtable);
+		if (NULL != was_created)
+			*was_created = 1;
+	} else {
+		if (NULL != was_created)
+			*was_created = 0;
+	}
+
+	return block;
+}
+
+static void btrfsic_cmp_log_and_dev_bytenr(struct btrfsic_state *state,
+					   u64 bytenr,
+					   struct btrfsic_dev_state *dev_state,
+					   u64 dev_bytenr)
+{
+	int num_copies;
+	int mirror_num;
+	int ret;
+	struct btrfsic_block_data_ctx block_ctx;
+	int match = 0;
+
+	num_copies = btrfs_num_copies(state->root->fs_info,
+				      bytenr, state->metablock_size);
+
+	for (mirror_num = 1; mirror_num <= num_copies; mirror_num++) {
+		ret = btrfsic_map_block(state, bytenr, state->metablock_size,
+					&block_ctx, mirror_num);
+		if (ret) {
+			printk(KERN_INFO "btrfsic:"
+			       " btrfsic_map_block(logical @%llu,"
+			       " mirror %d) failed!\n",
+			       bytenr, mirror_num);
+			continue;
+		}
+
+		if (dev_state->bdev == block_ctx.dev->bdev &&
+		    dev_bytenr == block_ctx.dev_bytenr) {
+			match++;
+			btrfsic_release_block_ctx(&block_ctx);
+			break;
+		}
+		btrfsic_release_block_ctx(&block_ctx);
+	}
+
+	if (WARN_ON(!match)) {
+		printk(KERN_INFO "btrfs: attempt to write M-block which contains logical bytenr that doesn't map to dev+physical bytenr of submit_bio,"
+		       " buffer->log_bytenr=%llu, submit_bio(bdev=%s,"
+		       " phys_bytenr=%llu)!\n",
+		       bytenr, dev_state->name, dev_bytenr);
+		for (mirror_num = 1; mirror_num <= num_copies; mirror_num++) {
+			ret = btrfsic_map_block(state, bytenr,
+						state->metablock_size,
+						&block_ctx, mirror_num);
+			if (ret)
+				continue;
+
+			printk(KERN_INFO "Read logical bytenr @%llu maps to"
+			       " (%s/%llu/%d)\n",
+			       bytenr, block_ctx.dev->name,
+			       block_ctx.dev_bytenr, mirror_num);
+		}
+	}
+}
+
+static struct btrfsic_dev_state *btrfsic_dev_state_lookup(
+		struct block_device *bdev)
+{
+	struct btrfsic_dev_state *ds;
+
+	ds = btrfsic_dev_state_hashtable_lookup(bdev,
+						&btrfsic_dev_state_hashtable);
+	return ds;
+}
+
+int btrfsic_submit_bh(int rw, struct buffer_head *bh)
+{
+	struct btrfsic_dev_state *dev_state;
+
+	if (!btrfsic_is_initialized)
+		return submit_bh(rw, bh);
+
+	mutex_lock(&btrfsic_mutex);
+	/* since btrfsic_submit_bh() might also be called before
+	 * btrfsic_mount(), this might return NULL */
+	dev_state = btrfsic_dev_state_lookup(bh->b_bdev);
+
+	/* Only called to write the superblock (incl. FLUSH/FUA) */
+	if (NULL != dev_state &&
+	    (rw & WRITE) && bh->b_size > 0) {
+		u64 dev_bytenr;
+
+		dev_bytenr = 4096 * bh->b_blocknr;
+		if (dev_state->state->print_mask &
+		    BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH)
+			printk(KERN_INFO
+			       "submit_bh(rw=0x%x, blocknr=%llu (bytenr %llu),"
+			       " size=%zu, data=%p, bdev=%p)\n",
+			       rw, (unsigned long long)bh->b_blocknr,
+			       dev_bytenr, bh->b_size, bh->b_data, bh->b_bdev);
+		btrfsic_process_written_block(dev_state, dev_bytenr,
+					      &bh->b_data, 1, NULL,
+					      NULL, bh, rw);
+	} else if (NULL != dev_state && (rw & REQ_FLUSH)) {
+		if (dev_state->state->print_mask &
+		    BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH)
+			printk(KERN_INFO
+			       "submit_bh(rw=0x%x FLUSH, bdev=%p)\n",
+			       rw, bh->b_bdev);
+		if (!dev_state->dummy_block_for_bio_bh_flush.is_iodone) {
+			if ((dev_state->state->print_mask &
+			     (BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH |
+			      BTRFSIC_PRINT_MASK_VERBOSE)))
+				printk(KERN_INFO
+				       "btrfsic_submit_bh(%s) with FLUSH"
+				       " but dummy block already in use"
+				       " (ignored)!\n",
+				       dev_state->name);
+		} else {
+			struct btrfsic_block *const block =
+				&dev_state->dummy_block_for_bio_bh_flush;
+
+			block->is_iodone = 0;
+			block->never_written = 0;
+			block->iodone_w_error = 0;
+			block->flush_gen = dev_state->last_flush_gen + 1;
+			block->submit_bio_bh_rw = rw;
+			block->orig_bio_bh_private = bh->b_private;
+			block->orig_bio_bh_end_io.bh = bh->b_end_io;
+			block->next_in_same_bio = NULL;
+			bh->b_private = block;
+			bh->b_end_io = btrfsic_bh_end_io;
+		}
+	}
+	mutex_unlock(&btrfsic_mutex);
+	return submit_bh(rw, bh);
+}
+
+static void __btrfsic_submit_bio(int rw, struct bio *bio)
+{
+	struct btrfsic_dev_state *dev_state;
+
+	if (!btrfsic_is_initialized)
+		return;
+
+	mutex_lock(&btrfsic_mutex);
+	/* since btrfsic_submit_bio() is also called before
+	 * btrfsic_mount(), this might return NULL */
+	dev_state = btrfsic_dev_state_lookup(bio->bi_bdev);
+	if (NULL != dev_state &&
+	    (rw & WRITE) && NULL != bio->bi_io_vec) {
+		unsigned int i;
+		u64 dev_bytenr;
+		u64 cur_bytenr;
+		int bio_is_patched;
+		char **mapped_datav;
+
+		dev_bytenr = 512 * bio->bi_iter.bi_sector;
+		bio_is_patched = 0;
+		if (dev_state->state->print_mask &
+		    BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH)
+			printk(KERN_INFO
+			       "submit_bio(rw=0x%x, bi_vcnt=%u,"
+			       " bi_sector=%llu (bytenr %llu), bi_bdev=%p)\n",
+			       rw, bio->bi_vcnt,
+			       (unsigned long long)bio->bi_iter.bi_sector,
+			       dev_bytenr, bio->bi_bdev);
+
+		mapped_datav = kmalloc(sizeof(*mapped_datav) * bio->bi_vcnt,
+				       GFP_NOFS);
+		if (!mapped_datav)
+			goto leave;
+		cur_bytenr = dev_bytenr;
+		for (i = 0; i < bio->bi_vcnt; i++) {
+			BUG_ON(bio->bi_io_vec[i].bv_len != PAGE_CACHE_SIZE);
+			mapped_datav[i] = kmap(bio->bi_io_vec[i].bv_page);
+			if (!mapped_datav[i]) {
+				while (i > 0) {
+					i--;
+					kunmap(bio->bi_io_vec[i].bv_page);
+				}
+				kfree(mapped_datav);
+				goto leave;
+			}
+			if (dev_state->state->print_mask &
+			    BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH_VERBOSE)
+				printk(KERN_INFO
+				       "#%u: bytenr=%llu, len=%u, offset=%u\n",
+				       i, cur_bytenr, bio->bi_io_vec[i].bv_len,
+				       bio->bi_io_vec[i].bv_offset);
+			cur_bytenr += bio->bi_io_vec[i].bv_len;
+		}
+		btrfsic_process_written_block(dev_state, dev_bytenr,
+					      mapped_datav, bio->bi_vcnt,
+					      bio, &bio_is_patched,
+					      NULL, rw);
+		while (i > 0) {
+			i--;
+			kunmap(bio->bi_io_vec[i].bv_page);
+		}
+		kfree(mapped_datav);
+	} else if (NULL != dev_state && (rw & REQ_FLUSH)) {
+		if (dev_state->state->print_mask &
+		    BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH)
+			printk(KERN_INFO
+			       "submit_bio(rw=0x%x FLUSH, bdev=%p)\n",
+			       rw, bio->bi_bdev);
+		if (!dev_state->dummy_block_for_bio_bh_flush.is_iodone) {
+			if ((dev_state->state->print_mask &
+			     (BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH |
+			      BTRFSIC_PRINT_MASK_VERBOSE)))
+				printk(KERN_INFO
+				       "btrfsic_submit_bio(%s) with FLUSH"
+				       " but dummy block already in use"
+				       " (ignored)!\n",
+				       dev_state->name);
+		} else {
+			struct btrfsic_block *const block =
+				&dev_state->dummy_block_for_bio_bh_flush;
+
+			block->is_iodone = 0;
+			block->never_written = 0;
+			block->iodone_w_error = 0;
+			block->flush_gen = dev_state->last_flush_gen + 1;
+			block->submit_bio_bh_rw = rw;
+			block->orig_bio_bh_private = bio->bi_private;
+			block->orig_bio_bh_end_io.bio = bio->bi_end_io;
+			block->next_in_same_bio = NULL;
+			bio->bi_private = block;
+			bio->bi_end_io = btrfsic_bio_end_io;
+		}
+	}
+leave:
+	mutex_unlock(&btrfsic_mutex);
+}
+
+void btrfsic_submit_bio(int rw, struct bio *bio)
+{
+	__btrfsic_submit_bio(rw, bio);
+	submit_bio(rw, bio);
+}
+
+int btrfsic_submit_bio_wait(int rw, struct bio *bio)
+{
+	__btrfsic_submit_bio(rw, bio);
+	return submit_bio_wait(rw, bio);
+}
+
+int btrfsic_mount(struct btrfs_root *root,
+		  struct btrfs_fs_devices *fs_devices,
+		  int including_extent_data, u32 print_mask)
+{
+	int ret;
+	struct btrfsic_state *state;
+	struct list_head *dev_head = &fs_devices->devices;
+	struct btrfs_device *device;
+
+	if (root->nodesize != root->leafsize) {
+		printk(KERN_INFO
+		       "btrfsic: cannot handle nodesize %d != leafsize %d!\n",
+		       root->nodesize, root->leafsize);
+		return -1;
+	}
+	if (root->nodesize & ((u64)PAGE_CACHE_SIZE - 1)) {
+		printk(KERN_INFO
+		       "btrfsic: cannot handle nodesize %d not being a multiple of PAGE_CACHE_SIZE %ld!\n",
+		       root->nodesize, PAGE_CACHE_SIZE);
+		return -1;
+	}
+	if (root->leafsize & ((u64)PAGE_CACHE_SIZE - 1)) {
+		printk(KERN_INFO
+		       "btrfsic: cannot handle leafsize %d not being a multiple of PAGE_CACHE_SIZE %ld!\n",
+		       root->leafsize, PAGE_CACHE_SIZE);
+		return -1;
+	}
+	if (root->sectorsize & ((u64)PAGE_CACHE_SIZE - 1)) {
+		printk(KERN_INFO
+		       "btrfsic: cannot handle sectorsize %d not being a multiple of PAGE_CACHE_SIZE %ld!\n",
+		       root->sectorsize, PAGE_CACHE_SIZE);
+		return -1;
+	}
+	state = kzalloc(sizeof(*state), GFP_NOFS);
+	if (NULL == state) {
+		printk(KERN_INFO "btrfs check-integrity: kmalloc() failed!\n");
+		return -1;
+	}
+
+	if (!btrfsic_is_initialized) {
+		mutex_init(&btrfsic_mutex);
+		btrfsic_dev_state_hashtable_init(&btrfsic_dev_state_hashtable);
+		btrfsic_is_initialized = 1;
+	}
+	mutex_lock(&btrfsic_mutex);
+	state->root = root;
+	state->print_mask = print_mask;
+	state->include_extent_data = including_extent_data;
+	state->csum_size = 0;
+	state->metablock_size = root->nodesize;
+	state->datablock_size = root->sectorsize;
+	INIT_LIST_HEAD(&state->all_blocks_list);
+	btrfsic_block_hashtable_init(&state->block_hashtable);
+	btrfsic_block_link_hashtable_init(&state->block_link_hashtable);
+	state->max_superblock_generation = 0;
+	state->latest_superblock = NULL;
+
+	list_for_each_entry(device, dev_head, dev_list) {
+		struct btrfsic_dev_state *ds;
+		char *p;
+
+		if (!device->bdev || !device->name)
+			continue;
+
+		ds = btrfsic_dev_state_alloc();
+		if (NULL == ds) {
+			printk(KERN_INFO
+			       "btrfs check-integrity: kmalloc() failed!\n");
+			mutex_unlock(&btrfsic_mutex);
+			return -1;
+		}
+		ds->bdev = device->bdev;
+		ds->state = state;
+		bdevname(ds->bdev, ds->name);
+		ds->name[BDEVNAME_SIZE - 1] = '\0';
+		for (p = ds->name; *p != '\0'; p++);
+		while (p > ds->name && *p != '/')
+			p--;
+		if (*p == '/')
+			p++;
+		strlcpy(ds->name, p, sizeof(ds->name));
+		btrfsic_dev_state_hashtable_add(ds,
+						&btrfsic_dev_state_hashtable);
+	}
+
+	ret = btrfsic_process_superblock(state, fs_devices);
+	if (0 != ret) {
+		mutex_unlock(&btrfsic_mutex);
+		btrfsic_unmount(root, fs_devices);
+		return ret;
+	}
+
+	if (state->print_mask & BTRFSIC_PRINT_MASK_INITIAL_DATABASE)
+		btrfsic_dump_database(state);
+	if (state->print_mask & BTRFSIC_PRINT_MASK_INITIAL_TREE)
+		btrfsic_dump_tree(state);
+
+	mutex_unlock(&btrfsic_mutex);
+	return 0;
+}
+
+void btrfsic_unmount(struct btrfs_root *root,
+		     struct btrfs_fs_devices *fs_devices)
+{
+	struct list_head *elem_all;
+	struct list_head *tmp_all;
+	struct btrfsic_state *state;
+	struct list_head *dev_head = &fs_devices->devices;
+	struct btrfs_device *device;
+
+	if (!btrfsic_is_initialized)
+		return;
+
+	mutex_lock(&btrfsic_mutex);
+
+	state = NULL;
+	list_for_each_entry(device, dev_head, dev_list) {
+		struct btrfsic_dev_state *ds;
+
+		if (!device->bdev || !device->name)
+			continue;
+
+		ds = btrfsic_dev_state_hashtable_lookup(
+				device->bdev,
+				&btrfsic_dev_state_hashtable);
+		if (NULL != ds) {
+			state = ds->state;
+			btrfsic_dev_state_hashtable_remove(ds);
+			btrfsic_dev_state_free(ds);
+		}
+	}
+
+	if (NULL == state) {
+		printk(KERN_INFO
+		       "btrfsic: error, cannot find state information"
+		       " on umount!\n");
+		mutex_unlock(&btrfsic_mutex);
+		return;
+	}
+
+	/*
+	 * Don't care about keeping the lists' state up to date,
+	 * just free all memory that was allocated dynamically.
+	 * Free the blocks and the block_links.
+	 */
+	list_for_each_safe(elem_all, tmp_all, &state->all_blocks_list) {
+		struct btrfsic_block *const b_all =
+		    list_entry(elem_all, struct btrfsic_block,
+			       all_blocks_node);
+		struct list_head *elem_ref_to;
+		struct list_head *tmp_ref_to;
+
+		list_for_each_safe(elem_ref_to, tmp_ref_to,
+				   &b_all->ref_to_list) {
+			struct btrfsic_block_link *const l =
+			    list_entry(elem_ref_to,
+				       struct btrfsic_block_link,
+				       node_ref_to);
+
+			if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
+				btrfsic_print_rem_link(state, l);
+
+			l->ref_cnt--;
+			if (0 == l->ref_cnt)
+				btrfsic_block_link_free(l);
+		}
+
+		if (b_all->is_iodone || b_all->never_written)
+			btrfsic_block_free(b_all);
+		else
+			printk(KERN_INFO "btrfs: attempt to free %c-block"
+			       " @%llu (%s/%llu/%d) on umount which is"
+			       " not yet iodone!\n",
+			       btrfsic_get_block_type(state, b_all),
+			       b_all->logical_bytenr, b_all->dev_state->name,
+			       b_all->dev_bytenr, b_all->mirror_num);
+	}
+
+	mutex_unlock(&btrfsic_mutex);
+
+	kfree(state);
+}
diff --git a/fs/btrfs/check-integrity.h b/fs/btrfs/check-integrity.h
new file mode 100644
index 00000000000..13b8566c97a
--- /dev/null
+++ b/fs/btrfs/check-integrity.h
@@ -0,0 +1,38 @@
+/*
+ * Copyright (C) STRATO AG 2011.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#if !defined(__BTRFS_CHECK_INTEGRITY__)
+#define __BTRFS_CHECK_INTEGRITY__
+
+#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
+int btrfsic_submit_bh(int rw, struct buffer_head *bh);
+void btrfsic_submit_bio(int rw, struct bio *bio);
+int btrfsic_submit_bio_wait(int rw, struct bio *bio);
+#else
+#define btrfsic_submit_bh submit_bh
+#define btrfsic_submit_bio submit_bio
+#define btrfsic_submit_bio_wait submit_bio_wait
+#endif
+
+int btrfsic_mount(struct btrfs_root *root,
+		  struct btrfs_fs_devices *fs_devices,
+		  int including_extent_data, u32 print_mask);
+void btrfsic_unmount(struct btrfs_root *root,
+		     struct btrfs_fs_devices *fs_devices);
+
+#endif
diff --git a/fs/btrfs/compat.h b/fs/btrfs/compat.h
deleted file mode 100644
index 7c4503ef6ef..00000000000
--- a/fs/btrfs/compat.h
+++ /dev/null
@@ -1,7 +0,0 @@
-#ifndef _COMPAT_H_
-#define _COMPAT_H_
-
-#define btrfs_drop_nlink(inode) drop_nlink(inode)
-#define btrfs_inc_nlink(inode)	inc_nlink(inode)
-
-#endif /* _COMPAT_H_ */
diff --git a/fs/btrfs/compression.c b/fs/btrfs/compression.c
index 396039b3a8a..1daea0b4718 100644
--- a/fs/btrfs/compression.c
+++ b/fs/btrfs/compression.c
@@ -32,7 +32,6 @@
 #include <linux/writeback.h>
 #include <linux/bit_spinlock.h>
 #include <linux/slab.h>
-#include "compat.h"
 #include "ctree.h"
 #include "disk-io.h"
 #include "transaction.h"
@@ -62,6 +61,9 @@ struct compressed_bio {
 	/* number of bytes on disk */
 	unsigned long compressed_len;
 
+	/* the compression algorithm for this bio */
+	int compress_type;
+
 	/* number of compressed pages in the array */
 	unsigned long nr_pages;
 
@@ -79,10 +81,15 @@ struct compressed_bio {
 	u32 sums;
 };
 
+static int btrfs_decompress_biovec(int type, struct page **pages_in,
+				   u64 disk_start, struct bio_vec *bvec,
+				   int vcnt, size_t srclen);
+
 static inline int compressed_bio_size(struct btrfs_root *root,
 				      unsigned long disk_size)
 {
-	u16 csum_size = btrfs_super_csum_size(&root->fs_info->super_copy);
+	u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
+
 	return sizeof(struct compressed_bio) +
 		((disk_size + root->sectorsize - 1) / root->sectorsize) *
 		csum_size;
@@ -91,23 +98,10 @@ static inline int compressed_bio_size(struct btrfs_root *root,
 static struct bio *compressed_bio_alloc(struct block_device *bdev,
 					u64 first_byte, gfp_t gfp_flags)
 {
-	struct bio *bio;
 	int nr_vecs;
 
 	nr_vecs = bio_get_nr_vecs(bdev);
-	bio = bio_alloc(gfp_flags, nr_vecs);
-
-	if (bio == NULL && (current->flags & PF_MEMALLOC)) {
-		while (!bio && (nr_vecs /= 2))
-			bio = bio_alloc(gfp_flags, nr_vecs);
-	}
-
-	if (bio) {
-		bio->bi_size = 0;
-		bio->bi_bdev = bdev;
-		bio->bi_sector = first_byte >> 9;
-	}
-	return bio;
+	return btrfs_bio_alloc(bdev, first_byte >> 9, nr_vecs, gfp_flags);
 }
 
 static int check_compressed_csum(struct inode *inode,
@@ -115,7 +109,6 @@ static int check_compressed_csum(struct inode *inode,
 				 u64 disk_start)
 {
 	int ret;
-	struct btrfs_root *root = BTRFS_I(inode)->root;
 	struct page *page;
 	unsigned long i;
 	char *kaddr;
@@ -129,17 +122,16 @@ static int check_compressed_csum(struct inode *inode,
 		page = cb->compressed_pages[i];
 		csum = ~(u32)0;
 
-		kaddr = kmap_atomic(page, KM_USER0);
-		csum = btrfs_csum_data(root, kaddr, csum, PAGE_CACHE_SIZE);
+		kaddr = kmap_atomic(page);
+		csum = btrfs_csum_data(kaddr, csum, PAGE_CACHE_SIZE);
 		btrfs_csum_final(csum, (char *)&csum);
-		kunmap_atomic(kaddr, KM_USER0);
+		kunmap_atomic(kaddr);
 
 		if (csum != *cb_sum) {
-			printk(KERN_INFO "btrfs csum failed ino %lu "
-			       "extent %llu csum %u "
-			       "wanted %u mirror %d\n", inode->i_ino,
-			       (unsigned long long)disk_start,
-			       csum, *cb_sum, cb->mirror_num);
+			btrfs_info(BTRFS_I(inode)->root->fs_info,
+			   "csum failed ino %llu extent %llu csum %u wanted %u mirror %d",
+			   btrfs_ino(inode), disk_start, csum, *cb_sum,
+			   cb->mirror_num);
 			ret = -EIO;
 			goto fail;
 		}
@@ -163,7 +155,6 @@ fail:
  */
 static void end_compressed_bio_read(struct bio *bio, int err)
 {
-	struct extent_io_tree *tree;
 	struct compressed_bio *cb = bio->bi_private;
 	struct inode *inode;
 	struct page *page;
@@ -180,19 +171,20 @@ static void end_compressed_bio_read(struct bio *bio, int err)
 		goto out;
 
 	inode = cb->inode;
-	ret = check_compressed_csum(inode, cb, (u64)bio->bi_sector << 9);
+	ret = check_compressed_csum(inode, cb,
+				    (u64)bio->bi_iter.bi_sector << 9);
 	if (ret)
 		goto csum_failed;
 
 	/* ok, we're the last bio for this extent, lets start
 	 * the decompression.
 	 */
-	tree = &BTRFS_I(inode)->io_tree;
-	ret = btrfs_zlib_decompress_biovec(cb->compressed_pages,
-					cb->start,
-					cb->orig_bio->bi_io_vec,
-					cb->orig_bio->bi_vcnt,
-					cb->compressed_len);
+	ret = btrfs_decompress_biovec(cb->compress_type,
+				      cb->compressed_pages,
+				      cb->start,
+				      cb->orig_bio->bi_io_vec,
+				      cb->orig_bio->bi_vcnt,
+				      cb->compressed_len);
 csum_failed:
 	if (ret)
 		cb->errors = 1;
@@ -209,18 +201,16 @@ csum_failed:
 	if (cb->errors) {
 		bio_io_error(cb->orig_bio);
 	} else {
-		int bio_index = 0;
-		struct bio_vec *bvec = cb->orig_bio->bi_io_vec;
+		int i;
+		struct bio_vec *bvec;
 
 		/*
 		 * we have verified the checksum already, set page
 		 * checked so the end_io handlers know about it
 		 */
-		while (bio_index < cb->orig_bio->bi_vcnt) {
+		bio_for_each_segment_all(bvec, cb->orig_bio, i)
 			SetPageChecked(bvec->bv_page);
-			bvec++;
-			bio_index++;
-		}
+
 		bio_endio(cb->orig_bio, 0);
 	}
 
@@ -235,8 +225,8 @@ out:
  * Clear the writeback bits on all of the file
  * pages for a compressed write
  */
-static noinline int end_compressed_writeback(struct inode *inode, u64 start,
-					     unsigned long ram_size)
+static noinline void end_compressed_writeback(struct inode *inode, u64 start,
+					      unsigned long ram_size)
 {
 	unsigned long index = start >> PAGE_CACHE_SHIFT;
 	unsigned long end_index = (start + ram_size - 1) >> PAGE_CACHE_SHIFT;
@@ -262,7 +252,6 @@ static noinline int end_compressed_writeback(struct inode *inode, u64 start,
 		index += ret;
 	}
 	/* the inode may be gone now */
-	return 0;
 }
 
 /*
@@ -343,14 +332,17 @@ int btrfs_submit_compressed_write(struct inode *inode, u64 start,
 	struct compressed_bio *cb;
 	unsigned long bytes_left;
 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
-	int page_index = 0;
+	int pg_index = 0;
 	struct page *page;
 	u64 first_byte = disk_start;
 	struct block_device *bdev;
 	int ret;
+	int skip_sum = BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM;
 
 	WARN_ON(start & ((u64)PAGE_CACHE_SIZE - 1));
 	cb = kmalloc(compressed_bio_size(root, compressed_len), GFP_NOFS);
+	if (!cb)
+		return -ENOMEM;
 	atomic_set(&cb->pending_bios, 0);
 	cb->errors = 0;
 	cb->inode = inode;
@@ -365,17 +357,21 @@ int btrfs_submit_compressed_write(struct inode *inode, u64 start,
 	bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
 
 	bio = compressed_bio_alloc(bdev, first_byte, GFP_NOFS);
+	if (!bio) {
+		kfree(cb);
+		return -ENOMEM;
+	}
 	bio->bi_private = cb;
 	bio->bi_end_io = end_compressed_bio_write;
 	atomic_inc(&cb->pending_bios);
 
 	/* create and submit bios for the compressed pages */
 	bytes_left = compressed_len;
-	for (page_index = 0; page_index < cb->nr_pages; page_index++) {
-		page = compressed_pages[page_index];
+	for (pg_index = 0; pg_index < cb->nr_pages; pg_index++) {
+		page = compressed_pages[pg_index];
 		page->mapping = inode->i_mapping;
-		if (bio->bi_size)
-			ret = io_tree->ops->merge_bio_hook(page, 0,
+		if (bio->bi_iter.bi_size)
+			ret = io_tree->ops->merge_bio_hook(WRITE, page, 0,
 							   PAGE_CACHE_SIZE,
 							   bio, 0);
 		else
@@ -394,23 +390,28 @@ int btrfs_submit_compressed_write(struct inode *inode, u64 start,
 			 */
 			atomic_inc(&cb->pending_bios);
 			ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0);
-			BUG_ON(ret);
+			BUG_ON(ret); /* -ENOMEM */
 
-			ret = btrfs_csum_one_bio(root, inode, bio, start, 1);
-			BUG_ON(ret);
+			if (!skip_sum) {
+				ret = btrfs_csum_one_bio(root, inode, bio,
+							 start, 1);
+				BUG_ON(ret); /* -ENOMEM */
+			}
 
 			ret = btrfs_map_bio(root, WRITE, bio, 0, 1);
-			BUG_ON(ret);
+			BUG_ON(ret); /* -ENOMEM */
 
 			bio_put(bio);
 
 			bio = compressed_bio_alloc(bdev, first_byte, GFP_NOFS);
+			BUG_ON(!bio);
 			bio->bi_private = cb;
 			bio->bi_end_io = end_compressed_bio_write;
 			bio_add_page(bio, page, PAGE_CACHE_SIZE, 0);
 		}
 		if (bytes_left < PAGE_CACHE_SIZE) {
-			printk("bytes left %lu compress len %lu nr %lu\n",
+			btrfs_info(BTRFS_I(inode)->root->fs_info,
+					"bytes left %lu compress len %lu nr %lu",
 			       bytes_left, cb->compressed_len, cb->nr_pages);
 		}
 		bytes_left -= PAGE_CACHE_SIZE;
@@ -420,13 +421,15 @@ int btrfs_submit_compressed_write(struct inode *inode, u64 start,
 	bio_get(bio);
 
 	ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0);
-	BUG_ON(ret);
+	BUG_ON(ret); /* -ENOMEM */
 
-	ret = btrfs_csum_one_bio(root, inode, bio, start, 1);
-	BUG_ON(ret);
+	if (!skip_sum) {
+		ret = btrfs_csum_one_bio(root, inode, bio, start, 1);
+		BUG_ON(ret); /* -ENOMEM */
+	}
 
 	ret = btrfs_map_bio(root, WRITE, bio, 0, 1);
-	BUG_ON(ret);
+	BUG_ON(ret); /* -ENOMEM */
 
 	bio_put(bio);
 	return 0;
@@ -437,7 +440,7 @@ static noinline int add_ra_bio_pages(struct inode *inode,
 				     struct compressed_bio *cb)
 {
 	unsigned long end_index;
-	unsigned long page_index;
+	unsigned long pg_index;
 	u64 last_offset;
 	u64 isize = i_size_read(inode);
 	int ret;
@@ -461,15 +464,15 @@ static noinline int add_ra_bio_pages(struct inode *inode,
 	end_index = (i_size_read(inode) - 1) >> PAGE_CACHE_SHIFT;
 
 	while (last_offset < compressed_end) {
-		page_index = last_offset >> PAGE_CACHE_SHIFT;
+		pg_index = last_offset >> PAGE_CACHE_SHIFT;
 
-		if (page_index > end_index)
+		if (pg_index > end_index)
 			break;
 
 		rcu_read_lock();
-		page = radix_tree_lookup(&mapping->page_tree, page_index);
+		page = radix_tree_lookup(&mapping->page_tree, pg_index);
 		rcu_read_unlock();
-		if (page) {
+		if (page && !radix_tree_exceptional_entry(page)) {
 			misses++;
 			if (misses > 4)
 				break;
@@ -481,7 +484,7 @@ static noinline int add_ra_bio_pages(struct inode *inode,
 		if (!page)
 			break;
 
-		if (add_to_page_cache_lru(page, mapping, page_index,
+		if (add_to_page_cache_lru(page, mapping, pg_index,
 								GFP_NOFS)) {
 			page_cache_release(page);
 			goto next;
@@ -494,7 +497,7 @@ static noinline int add_ra_bio_pages(struct inode *inode,
 		 * sure they map to this compressed extent on disk.
 		 */
 		set_page_extent_mapped(page);
-		lock_extent(tree, last_offset, end, GFP_NOFS);
+		lock_extent(tree, last_offset, end);
 		read_lock(&em_tree->lock);
 		em = lookup_extent_mapping(em_tree, last_offset,
 					   PAGE_CACHE_SIZE);
@@ -502,9 +505,9 @@ static noinline int add_ra_bio_pages(struct inode *inode,
 
 		if (!em || last_offset < em->start ||
 		    (last_offset + PAGE_CACHE_SIZE > extent_map_end(em)) ||
-		    (em->block_start >> 9) != cb->orig_bio->bi_sector) {
+		    (em->block_start >> 9) != cb->orig_bio->bi_iter.bi_sector) {
 			free_extent_map(em);
-			unlock_extent(tree, last_offset, end, GFP_NOFS);
+			unlock_extent(tree, last_offset, end);
 			unlock_page(page);
 			page_cache_release(page);
 			break;
@@ -518,10 +521,10 @@ static noinline int add_ra_bio_pages(struct inode *inode,
 			if (zero_offset) {
 				int zeros;
 				zeros = PAGE_CACHE_SIZE - zero_offset;
-				userpage = kmap_atomic(page, KM_USER0);
+				userpage = kmap_atomic(page);
 				memset(userpage + zero_offset, 0, zeros);
 				flush_dcache_page(page);
-				kunmap_atomic(userpage, KM_USER0);
+				kunmap_atomic(userpage);
 			}
 		}
 
@@ -532,7 +535,7 @@ static noinline int add_ra_bio_pages(struct inode *inode,
 			nr_pages++;
 			page_cache_release(page);
 		} else {
-			unlock_extent(tree, last_offset, end, GFP_NOFS);
+			unlock_extent(tree, last_offset, end);
 			unlock_page(page);
 			page_cache_release(page);
 			break;
@@ -548,7 +551,7 @@ next:
  * in it.  We don't actually do IO on those pages but allocate new ones
  * to hold the compressed pages on disk.
  *
- * bio->bi_sector points to the compressed extent on disk
+ * bio->bi_iter.bi_sector points to the compressed extent on disk
  * bio->bi_io_vec points to all of the inode pages
  * bio->bi_vcnt is a count of pages
  *
@@ -565,15 +568,16 @@ int btrfs_submit_compressed_read(struct inode *inode, struct bio *bio,
 	unsigned long uncompressed_len = bio->bi_vcnt * PAGE_CACHE_SIZE;
 	unsigned long compressed_len;
 	unsigned long nr_pages;
-	unsigned long page_index;
+	unsigned long pg_index;
 	struct page *page;
 	struct block_device *bdev;
 	struct bio *comp_bio;
-	u64 cur_disk_byte = (u64)bio->bi_sector << 9;
+	u64 cur_disk_byte = (u64)bio->bi_iter.bi_sector << 9;
 	u64 em_len;
 	u64 em_start;
 	struct extent_map *em;
-	int ret;
+	int ret = -ENOMEM;
+	int faili = 0;
 	u32 *sums;
 
 	tree = &BTRFS_I(inode)->io_tree;
@@ -585,9 +589,14 @@ int btrfs_submit_compressed_read(struct inode *inode, struct bio *bio,
 				   page_offset(bio->bi_io_vec->bv_page),
 				   PAGE_CACHE_SIZE);
 	read_unlock(&em_tree->lock);
+	if (!em)
+		return -EIO;
 
 	compressed_len = em->block_len;
 	cb = kmalloc(compressed_bio_size(root, compressed_len), GFP_NOFS);
+	if (!cb)
+		goto out;
+
 	atomic_set(&cb->pending_bios, 0);
 	cb->errors = 0;
 	cb->inode = inode;
@@ -603,38 +612,54 @@ int btrfs_submit_compressed_read(struct inode *inode, struct bio *bio,
 
 	cb->len = uncompressed_len;
 	cb->compressed_len = compressed_len;
+	cb->compress_type = extent_compress_type(bio_flags);
 	cb->orig_bio = bio;
 
 	nr_pages = (compressed_len + PAGE_CACHE_SIZE - 1) /
 				 PAGE_CACHE_SIZE;
-	cb->compressed_pages = kmalloc(sizeof(struct page *) * nr_pages,
+	cb->compressed_pages = kzalloc(sizeof(struct page *) * nr_pages,
 				       GFP_NOFS);
+	if (!cb->compressed_pages)
+		goto fail1;
+
 	bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
 
-	for (page_index = 0; page_index < nr_pages; page_index++) {
-		cb->compressed_pages[page_index] = alloc_page(GFP_NOFS |
+	for (pg_index = 0; pg_index < nr_pages; pg_index++) {
+		cb->compressed_pages[pg_index] = alloc_page(GFP_NOFS |
 							      __GFP_HIGHMEM);
+		if (!cb->compressed_pages[pg_index]) {
+			faili = pg_index - 1;
+			ret = -ENOMEM;
+			goto fail2;
+		}
 	}
+	faili = nr_pages - 1;
 	cb->nr_pages = nr_pages;
 
-	add_ra_bio_pages(inode, em_start + em_len, cb);
+	/* In the parent-locked case, we only locked the range we are
+	 * interested in.  In all other cases, we can opportunistically
+	 * cache decompressed data that goes beyond the requested range. */
+	if (!(bio_flags & EXTENT_BIO_PARENT_LOCKED))
+		add_ra_bio_pages(inode, em_start + em_len, cb);
 
 	/* include any pages we added in add_ra-bio_pages */
 	uncompressed_len = bio->bi_vcnt * PAGE_CACHE_SIZE;
 	cb->len = uncompressed_len;
 
 	comp_bio = compressed_bio_alloc(bdev, cur_disk_byte, GFP_NOFS);
+	if (!comp_bio)
+		goto fail2;
 	comp_bio->bi_private = cb;
 	comp_bio->bi_end_io = end_compressed_bio_read;
 	atomic_inc(&cb->pending_bios);
 
-	for (page_index = 0; page_index < nr_pages; page_index++) {
-		page = cb->compressed_pages[page_index];
+	for (pg_index = 0; pg_index < nr_pages; pg_index++) {
+		page = cb->compressed_pages[pg_index];
 		page->mapping = inode->i_mapping;
 		page->index = em_start >> PAGE_CACHE_SHIFT;
 
-		if (comp_bio->bi_size)
-			ret = tree->ops->merge_bio_hook(page, 0,
+		if (comp_bio->bi_iter.bi_size)
+			ret = tree->ops->merge_bio_hook(READ, page, 0,
 							PAGE_CACHE_SIZE,
 							comp_bio, 0);
 		else
@@ -646,7 +671,7 @@ int btrfs_submit_compressed_read(struct inode *inode, struct bio *bio,
 			bio_get(comp_bio);
 
 			ret = btrfs_bio_wq_end_io(root->fs_info, comp_bio, 0);
-			BUG_ON(ret);
+			BUG_ON(ret); /* -ENOMEM */
 
 			/*
 			 * inc the count before we submit the bio so
@@ -657,20 +682,23 @@ int btrfs_submit_compressed_read(struct inode *inode, struct bio *bio,
 			atomic_inc(&cb->pending_bios);
 
 			if (!(BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)) {
-				btrfs_lookup_bio_sums(root, inode, comp_bio,
-						      sums);
+				ret = btrfs_lookup_bio_sums(root, inode,
+							comp_bio, sums);
+				BUG_ON(ret); /* -ENOMEM */
 			}
-			sums += (comp_bio->bi_size + root->sectorsize - 1) /
-				root->sectorsize;
+			sums += (comp_bio->bi_iter.bi_size +
+				 root->sectorsize - 1) / root->sectorsize;
 
 			ret = btrfs_map_bio(root, READ, comp_bio,
 					    mirror_num, 0);
-			BUG_ON(ret);
+			if (ret)
+				bio_endio(comp_bio, ret);
 
 			bio_put(comp_bio);
 
 			comp_bio = compressed_bio_alloc(bdev, cur_disk_byte,
 							GFP_NOFS);
+			BUG_ON(!comp_bio);
 			comp_bio->bi_private = cb;
 			comp_bio->bi_end_io = end_compressed_bio_read;
 
@@ -681,14 +709,347 @@ int btrfs_submit_compressed_read(struct inode *inode, struct bio *bio,
 	bio_get(comp_bio);
 
 	ret = btrfs_bio_wq_end_io(root->fs_info, comp_bio, 0);
-	BUG_ON(ret);
+	BUG_ON(ret); /* -ENOMEM */
 
-	if (!(BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
-		btrfs_lookup_bio_sums(root, inode, comp_bio, sums);
+	if (!(BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)) {
+		ret = btrfs_lookup_bio_sums(root, inode, comp_bio, sums);
+		BUG_ON(ret); /* -ENOMEM */
+	}
 
 	ret = btrfs_map_bio(root, READ, comp_bio, mirror_num, 0);
-	BUG_ON(ret);
+	if (ret)
+		bio_endio(comp_bio, ret);
 
 	bio_put(comp_bio);
 	return 0;
+
+fail2:
+	while (faili >= 0) {
+		__free_page(cb->compressed_pages[faili]);
+		faili--;
+	}
+
+	kfree(cb->compressed_pages);
+fail1:
+	kfree(cb);
+out:
+	free_extent_map(em);
+	return ret;
+}
+
+static struct list_head comp_idle_workspace[BTRFS_COMPRESS_TYPES];
+static spinlock_t comp_workspace_lock[BTRFS_COMPRESS_TYPES];
+static int comp_num_workspace[BTRFS_COMPRESS_TYPES];
+static atomic_t comp_alloc_workspace[BTRFS_COMPRESS_TYPES];
+static wait_queue_head_t comp_workspace_wait[BTRFS_COMPRESS_TYPES];
+
+static struct btrfs_compress_op *btrfs_compress_op[] = {
+	&btrfs_zlib_compress,
+	&btrfs_lzo_compress,
+};
+
+void __init btrfs_init_compress(void)
+{
+	int i;
+
+	for (i = 0; i < BTRFS_COMPRESS_TYPES; i++) {
+		INIT_LIST_HEAD(&comp_idle_workspace[i]);
+		spin_lock_init(&comp_workspace_lock[i]);
+		atomic_set(&comp_alloc_workspace[i], 0);
+		init_waitqueue_head(&comp_workspace_wait[i]);
+	}
+}
+
+/*
+ * this finds an available workspace or allocates a new one
+ * ERR_PTR is returned if things go bad.
+ */
+static struct list_head *find_workspace(int type)
+{
+	struct list_head *workspace;
+	int cpus = num_online_cpus();
+	int idx = type - 1;
+
+	struct list_head *idle_workspace	= &comp_idle_workspace[idx];
+	spinlock_t *workspace_lock		= &comp_workspace_lock[idx];
+	atomic_t *alloc_workspace		= &comp_alloc_workspace[idx];
+	wait_queue_head_t *workspace_wait	= &comp_workspace_wait[idx];
+	int *num_workspace			= &comp_num_workspace[idx];
+again:
+	spin_lock(workspace_lock);
+	if (!list_empty(idle_workspace)) {
+		workspace = idle_workspace->next;
+		list_del(workspace);
+		(*num_workspace)--;
+		spin_unlock(workspace_lock);
+		return workspace;
+
+	}
+	if (atomic_read(alloc_workspace) > cpus) {
+		DEFINE_WAIT(wait);
+
+		spin_unlock(workspace_lock);
+		prepare_to_wait(workspace_wait, &wait, TASK_UNINTERRUPTIBLE);
+		if (atomic_read(alloc_workspace) > cpus && !*num_workspace)
+			schedule();
+		finish_wait(workspace_wait, &wait);
+		goto again;
+	}
+	atomic_inc(alloc_workspace);
+	spin_unlock(workspace_lock);
+
+	workspace = btrfs_compress_op[idx]->alloc_workspace();
+	if (IS_ERR(workspace)) {
+		atomic_dec(alloc_workspace);
+		wake_up(workspace_wait);
+	}
+	return workspace;
+}
+
+/*
+ * put a workspace struct back on the list or free it if we have enough
+ * idle ones sitting around
+ */
+static void free_workspace(int type, struct list_head *workspace)
+{
+	int idx = type - 1;
+	struct list_head *idle_workspace	= &comp_idle_workspace[idx];
+	spinlock_t *workspace_lock		= &comp_workspace_lock[idx];
+	atomic_t *alloc_workspace		= &comp_alloc_workspace[idx];
+	wait_queue_head_t *workspace_wait	= &comp_workspace_wait[idx];
+	int *num_workspace			= &comp_num_workspace[idx];
+
+	spin_lock(workspace_lock);
+	if (*num_workspace < num_online_cpus()) {
+		list_add(workspace, idle_workspace);
+		(*num_workspace)++;
+		spin_unlock(workspace_lock);
+		goto wake;
+	}
+	spin_unlock(workspace_lock);
+
+	btrfs_compress_op[idx]->free_workspace(workspace);
+	atomic_dec(alloc_workspace);
+wake:
+	smp_mb();
+	if (waitqueue_active(workspace_wait))
+		wake_up(workspace_wait);
+}
+
+/*
+ * cleanup function for module exit
+ */
+static void free_workspaces(void)
+{
+	struct list_head *workspace;
+	int i;
+
+	for (i = 0; i < BTRFS_COMPRESS_TYPES; i++) {
+		while (!list_empty(&comp_idle_workspace[i])) {
+			workspace = comp_idle_workspace[i].next;
+			list_del(workspace);
+			btrfs_compress_op[i]->free_workspace(workspace);
+			atomic_dec(&comp_alloc_workspace[i]);
+		}
+	}
+}
+
+/*
+ * given an address space and start/len, compress the bytes.
+ *
+ * pages are allocated to hold the compressed result and stored
+ * in 'pages'
+ *
+ * out_pages is used to return the number of pages allocated.  There
+ * may be pages allocated even if we return an error
+ *
+ * total_in is used to return the number of bytes actually read.  It
+ * may be smaller then len if we had to exit early because we
+ * ran out of room in the pages array or because we cross the
+ * max_out threshold.
+ *
+ * total_out is used to return the total number of compressed bytes
+ *
+ * max_out tells us the max number of bytes that we're allowed to
+ * stuff into pages
+ */
+int btrfs_compress_pages(int type, struct address_space *mapping,
+			 u64 start, unsigned long len,
+			 struct page **pages,
+			 unsigned long nr_dest_pages,
+			 unsigned long *out_pages,
+			 unsigned long *total_in,
+			 unsigned long *total_out,
+			 unsigned long max_out)
+{
+	struct list_head *workspace;
+	int ret;
+
+	workspace = find_workspace(type);
+	if (IS_ERR(workspace))
+		return PTR_ERR(workspace);
+
+	ret = btrfs_compress_op[type-1]->compress_pages(workspace, mapping,
+						      start, len, pages,
+						      nr_dest_pages, out_pages,
+						      total_in, total_out,
+						      max_out);
+	free_workspace(type, workspace);
+	return ret;
+}
+
+/*
+ * pages_in is an array of pages with compressed data.
+ *
+ * disk_start is the starting logical offset of this array in the file
+ *
+ * bvec is a bio_vec of pages from the file that we want to decompress into
+ *
+ * vcnt is the count of pages in the biovec
+ *
+ * srclen is the number of bytes in pages_in
+ *
+ * The basic idea is that we have a bio that was created by readpages.
+ * The pages in the bio are for the uncompressed data, and they may not
+ * be contiguous.  They all correspond to the range of bytes covered by
+ * the compressed extent.
+ */
+static int btrfs_decompress_biovec(int type, struct page **pages_in,
+				   u64 disk_start, struct bio_vec *bvec,
+				   int vcnt, size_t srclen)
+{
+	struct list_head *workspace;
+	int ret;
+
+	workspace = find_workspace(type);
+	if (IS_ERR(workspace))
+		return PTR_ERR(workspace);
+
+	ret = btrfs_compress_op[type-1]->decompress_biovec(workspace, pages_in,
+							 disk_start,
+							 bvec, vcnt, srclen);
+	free_workspace(type, workspace);
+	return ret;
+}
+
+/*
+ * a less complex decompression routine.  Our compressed data fits in a
+ * single page, and we want to read a single page out of it.
+ * start_byte tells us the offset into the compressed data we're interested in
+ */
+int btrfs_decompress(int type, unsigned char *data_in, struct page *dest_page,
+		     unsigned long start_byte, size_t srclen, size_t destlen)
+{
+	struct list_head *workspace;
+	int ret;
+
+	workspace = find_workspace(type);
+	if (IS_ERR(workspace))
+		return PTR_ERR(workspace);
+
+	ret = btrfs_compress_op[type-1]->decompress(workspace, data_in,
+						  dest_page, start_byte,
+						  srclen, destlen);
+
+	free_workspace(type, workspace);
+	return ret;
+}
+
+void btrfs_exit_compress(void)
+{
+	free_workspaces();
+}
+
+/*
+ * Copy uncompressed data from working buffer to pages.
+ *
+ * buf_start is the byte offset we're of the start of our workspace buffer.
+ *
+ * total_out is the last byte of the buffer
+ */
+int btrfs_decompress_buf2page(char *buf, unsigned long buf_start,
+			      unsigned long total_out, u64 disk_start,
+			      struct bio_vec *bvec, int vcnt,
+			      unsigned long *pg_index,
+			      unsigned long *pg_offset)
+{
+	unsigned long buf_offset;
+	unsigned long current_buf_start;
+	unsigned long start_byte;
+	unsigned long working_bytes = total_out - buf_start;
+	unsigned long bytes;
+	char *kaddr;
+	struct page *page_out = bvec[*pg_index].bv_page;
+
+	/*
+	 * start byte is the first byte of the page we're currently
+	 * copying into relative to the start of the compressed data.
+	 */
+	start_byte = page_offset(page_out) - disk_start;
+
+	/* we haven't yet hit data corresponding to this page */
+	if (total_out <= start_byte)
+		return 1;
+
+	/*
+	 * the start of the data we care about is offset into
+	 * the middle of our working buffer
+	 */
+	if (total_out > start_byte && buf_start < start_byte) {
+		buf_offset = start_byte - buf_start;
+		working_bytes -= buf_offset;
+	} else {
+		buf_offset = 0;
+	}
+	current_buf_start = buf_start;
+
+	/* copy bytes from the working buffer into the pages */
+	while (working_bytes > 0) {
+		bytes = min(PAGE_CACHE_SIZE - *pg_offset,
+			    PAGE_CACHE_SIZE - buf_offset);
+		bytes = min(bytes, working_bytes);
+		kaddr = kmap_atomic(page_out);
+		memcpy(kaddr + *pg_offset, buf + buf_offset, bytes);
+		if (*pg_index == (vcnt - 1) && *pg_offset == 0)
+			memset(kaddr + bytes, 0, PAGE_CACHE_SIZE - bytes);
+		kunmap_atomic(kaddr);
+		flush_dcache_page(page_out);
+
+		*pg_offset += bytes;
+		buf_offset += bytes;
+		working_bytes -= bytes;
+		current_buf_start += bytes;
+
+		/* check if we need to pick another page */
+		if (*pg_offset == PAGE_CACHE_SIZE) {
+			(*pg_index)++;
+			if (*pg_index >= vcnt)
+				return 0;
+
+			page_out = bvec[*pg_index].bv_page;
+			*pg_offset = 0;
+			start_byte = page_offset(page_out) - disk_start;
+
+			/*
+			 * make sure our new page is covered by this
+			 * working buffer
+			 */
+			if (total_out <= start_byte)
+				return 1;
+
+			/*
+			 * the next page in the biovec might not be adjacent
+			 * to the last page, but it might still be found
+			 * inside this working buffer. bump our offset pointer
+			 */
+			if (total_out > start_byte &&
+			    current_buf_start < start_byte) {
+				buf_offset = start_byte - buf_start;
+				working_bytes = total_out - start_byte;
+				current_buf_start = buf_start + buf_offset;
+			}
+		}
+	}
+
+	return 1;
 }
diff --git a/fs/btrfs/compression.h b/fs/btrfs/compression.h
index 421f5b4aa71..0c803b4fbf9 100644
--- a/fs/btrfs/compression.h
+++ b/fs/btrfs/compression.h
@@ -19,24 +19,25 @@
 #ifndef __BTRFS_COMPRESSION_
 #define __BTRFS_COMPRESSION_
 
-int btrfs_zlib_decompress(unsigned char *data_in,
-			  struct page *dest_page,
-			  unsigned long start_byte,
-			  size_t srclen, size_t destlen);
-int btrfs_zlib_compress_pages(struct address_space *mapping,
-			      u64 start, unsigned long len,
-			      struct page **pages,
-			      unsigned long nr_dest_pages,
-			      unsigned long *out_pages,
-			      unsigned long *total_in,
-			      unsigned long *total_out,
-			      unsigned long max_out);
-int btrfs_zlib_decompress_biovec(struct page **pages_in,
-			      u64 disk_start,
-			      struct bio_vec *bvec,
-			      int vcnt,
-			      size_t srclen);
-void btrfs_zlib_exit(void);
+void btrfs_init_compress(void);
+void btrfs_exit_compress(void);
+
+int btrfs_compress_pages(int type, struct address_space *mapping,
+			 u64 start, unsigned long len,
+			 struct page **pages,
+			 unsigned long nr_dest_pages,
+			 unsigned long *out_pages,
+			 unsigned long *total_in,
+			 unsigned long *total_out,
+			 unsigned long max_out);
+int btrfs_decompress(int type, unsigned char *data_in, struct page *dest_page,
+		     unsigned long start_byte, size_t srclen, size_t destlen);
+int btrfs_decompress_buf2page(char *buf, unsigned long buf_start,
+			      unsigned long total_out, u64 disk_start,
+			      struct bio_vec *bvec, int vcnt,
+			      unsigned long *pg_index,
+			      unsigned long *pg_offset);
+
 int btrfs_submit_compressed_write(struct inode *inode, u64 start,
 				  unsigned long len, u64 disk_start,
 				  unsigned long compressed_len,
@@ -44,4 +45,37 @@ int btrfs_submit_compressed_write(struct inode *inode, u64 start,
 				  unsigned long nr_pages);
 int btrfs_submit_compressed_read(struct inode *inode, struct bio *bio,
 				 int mirror_num, unsigned long bio_flags);
+
+struct btrfs_compress_op {
+	struct list_head *(*alloc_workspace)(void);
+
+	void (*free_workspace)(struct list_head *workspace);
+
+	int (*compress_pages)(struct list_head *workspace,
+			      struct address_space *mapping,
+			      u64 start, unsigned long len,
+			      struct page **pages,
+			      unsigned long nr_dest_pages,
+			      unsigned long *out_pages,
+			      unsigned long *total_in,
+			      unsigned long *total_out,
+			      unsigned long max_out);
+
+	int (*decompress_biovec)(struct list_head *workspace,
+				 struct page **pages_in,
+				 u64 disk_start,
+				 struct bio_vec *bvec,
+				 int vcnt,
+				 size_t srclen);
+
+	int (*decompress)(struct list_head *workspace,
+			  unsigned char *data_in,
+			  struct page *dest_page,
+			  unsigned long start_byte,
+			  size_t srclen, size_t destlen);
+};
+
+extern struct btrfs_compress_op btrfs_zlib_compress;
+extern struct btrfs_compress_op btrfs_lzo_compress;
+
 #endif
diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index 6795a713b20..aeab453b8e2 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -18,6 +18,7 @@
 
 #include <linux/sched.h>
 #include <linux/slab.h>
+#include <linux/rbtree.h>
 #include "ctree.h"
 #include "disk-io.h"
 #include "transaction.h"
@@ -36,20 +37,15 @@ static int balance_node_right(struct btrfs_trans_handle *trans,
 			      struct btrfs_root *root,
 			      struct extent_buffer *dst_buf,
 			      struct extent_buffer *src_buf);
-static int del_ptr(struct btrfs_trans_handle *trans, struct btrfs_root *root,
-		   struct btrfs_path *path, int level, int slot);
-static int setup_items_for_insert(struct btrfs_trans_handle *trans,
-			struct btrfs_root *root, struct btrfs_path *path,
-			struct btrfs_key *cpu_key, u32 *data_size,
-			u32 total_data, u32 total_size, int nr);
-
+static void del_ptr(struct btrfs_root *root, struct btrfs_path *path,
+		    int level, int slot);
+static int tree_mod_log_free_eb(struct btrfs_fs_info *fs_info,
+				 struct extent_buffer *eb);
 
 struct btrfs_path *btrfs_alloc_path(void)
 {
 	struct btrfs_path *path;
 	path = kmem_cache_zalloc(btrfs_path_cachep, GFP_NOFS);
-	if (path)
-		path->reada = 1;
 	return path;
 }
 
@@ -61,8 +57,13 @@ noinline void btrfs_set_path_blocking(struct btrfs_path *p)
 {
 	int i;
 	for (i = 0; i < BTRFS_MAX_LEVEL; i++) {
-		if (p->nodes[i] && p->locks[i])
-			btrfs_set_lock_blocking(p->nodes[i]);
+		if (!p->nodes[i] || !p->locks[i])
+			continue;
+		btrfs_set_lock_blocking_rw(p->nodes[i], p->locks[i]);
+		if (p->locks[i] == BTRFS_READ_LOCK)
+			p->locks[i] = BTRFS_READ_LOCK_BLOCKING;
+		else if (p->locks[i] == BTRFS_WRITE_LOCK)
+			p->locks[i] = BTRFS_WRITE_LOCK_BLOCKING;
 	}
 }
 
@@ -75,7 +76,7 @@ noinline void btrfs_set_path_blocking(struct btrfs_path *p)
  * for held
  */
 noinline void btrfs_clear_path_blocking(struct btrfs_path *p,
-					struct extent_buffer *held)
+					struct extent_buffer *held, int held_rw)
 {
 	int i;
 
@@ -86,26 +87,38 @@ noinline void btrfs_clear_path_blocking(struct btrfs_path *p,
 	 * really sure by forcing the path to blocking before we clear
 	 * the path blocking.
 	 */
-	if (held)
-		btrfs_set_lock_blocking(held);
+	if (held) {
+		btrfs_set_lock_blocking_rw(held, held_rw);
+		if (held_rw == BTRFS_WRITE_LOCK)
+			held_rw = BTRFS_WRITE_LOCK_BLOCKING;
+		else if (held_rw == BTRFS_READ_LOCK)
+			held_rw = BTRFS_READ_LOCK_BLOCKING;
+	}
 	btrfs_set_path_blocking(p);
 #endif
 
 	for (i = BTRFS_MAX_LEVEL - 1; i >= 0; i--) {
-		if (p->nodes[i] && p->locks[i])
-			btrfs_clear_lock_blocking(p->nodes[i]);
+		if (p->nodes[i] && p->locks[i]) {
+			btrfs_clear_lock_blocking_rw(p->nodes[i], p->locks[i]);
+			if (p->locks[i] == BTRFS_WRITE_LOCK_BLOCKING)
+				p->locks[i] = BTRFS_WRITE_LOCK;
+			else if (p->locks[i] == BTRFS_READ_LOCK_BLOCKING)
+				p->locks[i] = BTRFS_READ_LOCK;
+		}
 	}
 
 #ifdef CONFIG_DEBUG_LOCK_ALLOC
 	if (held)
-		btrfs_clear_lock_blocking(held);
+		btrfs_clear_lock_blocking_rw(held, held_rw);
 #endif
 }
 
 /* this also releases the path */
 void btrfs_free_path(struct btrfs_path *p)
 {
-	btrfs_release_path(NULL, p);
+	if (!p)
+		return;
+	btrfs_release_path(p);
 	kmem_cache_free(btrfs_path_cachep, p);
 }
 
@@ -115,7 +128,7 @@ void btrfs_free_path(struct btrfs_path *p)
  *
  * It is safe to call this on paths that no locks or extent buffers held.
  */
-noinline void btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p)
+noinline void btrfs_release_path(struct btrfs_path *p)
 {
 	int i;
 
@@ -124,7 +137,7 @@ noinline void btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p)
 		if (!p->nodes[i])
 			continue;
 		if (p->locks[i]) {
-			btrfs_tree_unlock(p->nodes[i]);
+			btrfs_tree_unlock_rw(p->nodes[i], p->locks[i]);
 			p->locks[i] = 0;
 		}
 		free_extent_buffer(p->nodes[i]);
@@ -145,10 +158,24 @@ noinline void btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p)
 struct extent_buffer *btrfs_root_node(struct btrfs_root *root)
 {
 	struct extent_buffer *eb;
-	spin_lock(&root->node_lock);
-	eb = root->node;
-	extent_buffer_get(eb);
-	spin_unlock(&root->node_lock);
+
+	while (1) {
+		rcu_read_lock();
+		eb = rcu_dereference(root->node);
+
+		/*
+		 * RCU really hurts here, we could free up the root node because
+		 * it was cow'ed but we may not get the new root node yet so do
+		 * the inc_not_zero dance and if it doesn't work then
+		 * synchronize_rcu and try again.
+		 */
+		if (atomic_inc_not_zero(&eb->refs)) {
+			rcu_read_unlock();
+			break;
+		}
+		rcu_read_unlock();
+		synchronize_rcu();
+	}
 	return eb;
 }
 
@@ -163,30 +190,46 @@ struct extent_buffer *btrfs_lock_root_node(struct btrfs_root *root)
 	while (1) {
 		eb = btrfs_root_node(root);
 		btrfs_tree_lock(eb);
-
-		spin_lock(&root->node_lock);
-		if (eb == root->node) {
-			spin_unlock(&root->node_lock);
+		if (eb == root->node)
 			break;
-		}
-		spin_unlock(&root->node_lock);
-
 		btrfs_tree_unlock(eb);
 		free_extent_buffer(eb);
 	}
 	return eb;
 }
 
+/* loop around taking references on and locking the root node of the
+ * tree until you end up with a lock on the root.  A locked buffer
+ * is returned, with a reference held.
+ */
+static struct extent_buffer *btrfs_read_lock_root_node(struct btrfs_root *root)
+{
+	struct extent_buffer *eb;
+
+	while (1) {
+		eb = btrfs_root_node(root);
+		btrfs_tree_read_lock(eb);
+		if (eb == root->node)
+			break;
+		btrfs_tree_read_unlock(eb);
+		free_extent_buffer(eb);
+	}
+	return eb;
+}
+
 /* cowonly root (everything not a reference counted cow subvolume), just get
  * put onto a simple dirty list.  transaction.c walks this to make sure they
  * get properly updated on disk.
  */
 static void add_root_to_dirty_list(struct btrfs_root *root)
 {
-	if (root->track_dirty && list_empty(&root->dirty_list)) {
+	spin_lock(&root->fs_info->trans_lock);
+	if (test_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state) &&
+	    list_empty(&root->dirty_list)) {
 		list_add(&root->dirty_list,
 			 &root->fs_info->dirty_cowonly_roots);
 	}
+	spin_unlock(&root->fs_info->trans_lock);
 }
 
 /*
@@ -200,17 +243,16 @@ int btrfs_copy_root(struct btrfs_trans_handle *trans,
 		      struct extent_buffer **cow_ret, u64 new_root_objectid)
 {
 	struct extent_buffer *cow;
-	u32 nritems;
 	int ret = 0;
 	int level;
 	struct btrfs_disk_key disk_key;
 
-	WARN_ON(root->ref_cows && trans->transid !=
-		root->fs_info->running_transaction->transid);
-	WARN_ON(root->ref_cows && trans->transid != root->last_trans);
+	WARN_ON(test_bit(BTRFS_ROOT_REF_COWS, &root->state) &&
+		trans->transid != root->fs_info->running_transaction->transid);
+	WARN_ON(test_bit(BTRFS_ROOT_REF_COWS, &root->state) &&
+		trans->transid != root->last_trans);
 
 	level = btrfs_header_level(buf);
-	nritems = btrfs_header_nritems(buf);
 	if (level == 0)
 		btrfs_item_key(buf, &disk_key, 0);
 	else
@@ -233,15 +275,14 @@ int btrfs_copy_root(struct btrfs_trans_handle *trans,
 	else
 		btrfs_set_header_owner(cow, new_root_objectid);
 
-	write_extent_buffer(cow, root->fs_info->fsid,
-			    (unsigned long)btrfs_header_fsid(cow),
+	write_extent_buffer(cow, root->fs_info->fsid, btrfs_header_fsid(),
 			    BTRFS_FSID_SIZE);
 
 	WARN_ON(btrfs_header_generation(buf) > trans->transid);
 	if (new_root_objectid == BTRFS_TREE_RELOC_OBJECTID)
-		ret = btrfs_inc_ref(trans, root, cow, 1);
+		ret = btrfs_inc_ref(trans, root, cow, 1, 1);
 	else
-		ret = btrfs_inc_ref(trans, root, cow, 0);
+		ret = btrfs_inc_ref(trans, root, cow, 0, 1);
 
 	if (ret)
 		return ret;
@@ -251,6 +292,666 @@ int btrfs_copy_root(struct btrfs_trans_handle *trans,
 	return 0;
 }
 
+enum mod_log_op {
+	MOD_LOG_KEY_REPLACE,
+	MOD_LOG_KEY_ADD,
+	MOD_LOG_KEY_REMOVE,
+	MOD_LOG_KEY_REMOVE_WHILE_FREEING,
+	MOD_LOG_KEY_REMOVE_WHILE_MOVING,
+	MOD_LOG_MOVE_KEYS,
+	MOD_LOG_ROOT_REPLACE,
+};
+
+struct tree_mod_move {
+	int dst_slot;
+	int nr_items;
+};
+
+struct tree_mod_root {
+	u64 logical;
+	u8 level;
+};
+
+struct tree_mod_elem {
+	struct rb_node node;
+	u64 index;		/* shifted logical */
+	u64 seq;
+	enum mod_log_op op;
+
+	/* this is used for MOD_LOG_KEY_* and MOD_LOG_MOVE_KEYS operations */
+	int slot;
+
+	/* this is used for MOD_LOG_KEY* and MOD_LOG_ROOT_REPLACE */
+	u64 generation;
+
+	/* those are used for op == MOD_LOG_KEY_{REPLACE,REMOVE} */
+	struct btrfs_disk_key key;
+	u64 blockptr;
+
+	/* this is used for op == MOD_LOG_MOVE_KEYS */
+	struct tree_mod_move move;
+
+	/* this is used for op == MOD_LOG_ROOT_REPLACE */
+	struct tree_mod_root old_root;
+};
+
+static inline void tree_mod_log_read_lock(struct btrfs_fs_info *fs_info)
+{
+	read_lock(&fs_info->tree_mod_log_lock);
+}
+
+static inline void tree_mod_log_read_unlock(struct btrfs_fs_info *fs_info)
+{
+	read_unlock(&fs_info->tree_mod_log_lock);
+}
+
+static inline void tree_mod_log_write_lock(struct btrfs_fs_info *fs_info)
+{
+	write_lock(&fs_info->tree_mod_log_lock);
+}
+
+static inline void tree_mod_log_write_unlock(struct btrfs_fs_info *fs_info)
+{
+	write_unlock(&fs_info->tree_mod_log_lock);
+}
+
+/*
+ * Pull a new tree mod seq number for our operation.
+ */
+static inline u64 btrfs_inc_tree_mod_seq(struct btrfs_fs_info *fs_info)
+{
+	return atomic64_inc_return(&fs_info->tree_mod_seq);
+}
+
+/*
+ * This adds a new blocker to the tree mod log's blocker list if the @elem
+ * passed does not already have a sequence number set. So when a caller expects
+ * to record tree modifications, it should ensure to set elem->seq to zero
+ * before calling btrfs_get_tree_mod_seq.
+ * Returns a fresh, unused tree log modification sequence number, even if no new
+ * blocker was added.
+ */
+u64 btrfs_get_tree_mod_seq(struct btrfs_fs_info *fs_info,
+			   struct seq_list *elem)
+{
+	tree_mod_log_write_lock(fs_info);
+	spin_lock(&fs_info->tree_mod_seq_lock);
+	if (!elem->seq) {
+		elem->seq = btrfs_inc_tree_mod_seq(fs_info);
+		list_add_tail(&elem->list, &fs_info->tree_mod_seq_list);
+	}
+	spin_unlock(&fs_info->tree_mod_seq_lock);
+	tree_mod_log_write_unlock(fs_info);
+
+	return elem->seq;
+}
+
+void btrfs_put_tree_mod_seq(struct btrfs_fs_info *fs_info,
+			    struct seq_list *elem)
+{
+	struct rb_root *tm_root;
+	struct rb_node *node;
+	struct rb_node *next;
+	struct seq_list *cur_elem;
+	struct tree_mod_elem *tm;
+	u64 min_seq = (u64)-1;
+	u64 seq_putting = elem->seq;
+
+	if (!seq_putting)
+		return;
+
+	spin_lock(&fs_info->tree_mod_seq_lock);
+	list_del(&elem->list);
+	elem->seq = 0;
+
+	list_for_each_entry(cur_elem, &fs_info->tree_mod_seq_list, list) {
+		if (cur_elem->seq < min_seq) {
+			if (seq_putting > cur_elem->seq) {
+				/*
+				 * blocker with lower sequence number exists, we
+				 * cannot remove anything from the log
+				 */
+				spin_unlock(&fs_info->tree_mod_seq_lock);
+				return;
+			}
+			min_seq = cur_elem->seq;
+		}
+	}
+	spin_unlock(&fs_info->tree_mod_seq_lock);
+
+	/*
+	 * anything that's lower than the lowest existing (read: blocked)
+	 * sequence number can be removed from the tree.
+	 */
+	tree_mod_log_write_lock(fs_info);
+	tm_root = &fs_info->tree_mod_log;
+	for (node = rb_first(tm_root); node; node = next) {
+		next = rb_next(node);
+		tm = container_of(node, struct tree_mod_elem, node);
+		if (tm->seq > min_seq)
+			continue;
+		rb_erase(node, tm_root);
+		kfree(tm);
+	}
+	tree_mod_log_write_unlock(fs_info);
+}
+
+/*
+ * key order of the log:
+ *       index -> sequence
+ *
+ * the index is the shifted logical of the *new* root node for root replace
+ * operations, or the shifted logical of the affected block for all other
+ * operations.
+ *
+ * Note: must be called with write lock (tree_mod_log_write_lock).
+ */
+static noinline int
+__tree_mod_log_insert(struct btrfs_fs_info *fs_info, struct tree_mod_elem *tm)
+{
+	struct rb_root *tm_root;
+	struct rb_node **new;
+	struct rb_node *parent = NULL;
+	struct tree_mod_elem *cur;
+
+	BUG_ON(!tm);
+
+	tm->seq = btrfs_inc_tree_mod_seq(fs_info);
+
+	tm_root = &fs_info->tree_mod_log;
+	new = &tm_root->rb_node;
+	while (*new) {
+		cur = container_of(*new, struct tree_mod_elem, node);
+		parent = *new;
+		if (cur->index < tm->index)
+			new = &((*new)->rb_left);
+		else if (cur->index > tm->index)
+			new = &((*new)->rb_right);
+		else if (cur->seq < tm->seq)
+			new = &((*new)->rb_left);
+		else if (cur->seq > tm->seq)
+			new = &((*new)->rb_right);
+		else
+			return -EEXIST;
+	}
+
+	rb_link_node(&tm->node, parent, new);
+	rb_insert_color(&tm->node, tm_root);
+	return 0;
+}
+
+/*
+ * Determines if logging can be omitted. Returns 1 if it can. Otherwise, it
+ * returns zero with the tree_mod_log_lock acquired. The caller must hold
+ * this until all tree mod log insertions are recorded in the rb tree and then
+ * call tree_mod_log_write_unlock() to release.
+ */
+static inline int tree_mod_dont_log(struct btrfs_fs_info *fs_info,
+				    struct extent_buffer *eb) {
+	smp_mb();
+	if (list_empty(&(fs_info)->tree_mod_seq_list))
+		return 1;
+	if (eb && btrfs_header_level(eb) == 0)
+		return 1;
+
+	tree_mod_log_write_lock(fs_info);
+	if (list_empty(&(fs_info)->tree_mod_seq_list)) {
+		tree_mod_log_write_unlock(fs_info);
+		return 1;
+	}
+
+	return 0;
+}
+
+/* Similar to tree_mod_dont_log, but doesn't acquire any locks. */
+static inline int tree_mod_need_log(const struct btrfs_fs_info *fs_info,
+				    struct extent_buffer *eb)
+{
+	smp_mb();
+	if (list_empty(&(fs_info)->tree_mod_seq_list))
+		return 0;
+	if (eb && btrfs_header_level(eb) == 0)
+		return 0;
+
+	return 1;
+}
+
+static struct tree_mod_elem *
+alloc_tree_mod_elem(struct extent_buffer *eb, int slot,
+		    enum mod_log_op op, gfp_t flags)
+{
+	struct tree_mod_elem *tm;
+
+	tm = kzalloc(sizeof(*tm), flags);
+	if (!tm)
+		return NULL;
+
+	tm->index = eb->start >> PAGE_CACHE_SHIFT;
+	if (op != MOD_LOG_KEY_ADD) {
+		btrfs_node_key(eb, &tm->key, slot);
+		tm->blockptr = btrfs_node_blockptr(eb, slot);
+	}
+	tm->op = op;
+	tm->slot = slot;
+	tm->generation = btrfs_node_ptr_generation(eb, slot);
+	RB_CLEAR_NODE(&tm->node);
+
+	return tm;
+}
+
+static noinline int
+tree_mod_log_insert_key(struct btrfs_fs_info *fs_info,
+			struct extent_buffer *eb, int slot,
+			enum mod_log_op op, gfp_t flags)
+{
+	struct tree_mod_elem *tm;
+	int ret;
+
+	if (!tree_mod_need_log(fs_info, eb))
+		return 0;
+
+	tm = alloc_tree_mod_elem(eb, slot, op, flags);
+	if (!tm)
+		return -ENOMEM;
+
+	if (tree_mod_dont_log(fs_info, eb)) {
+		kfree(tm);
+		return 0;
+	}
+
+	ret = __tree_mod_log_insert(fs_info, tm);
+	tree_mod_log_write_unlock(fs_info);
+	if (ret)
+		kfree(tm);
+
+	return ret;
+}
+
+static noinline int
+tree_mod_log_insert_move(struct btrfs_fs_info *fs_info,
+			 struct extent_buffer *eb, int dst_slot, int src_slot,
+			 int nr_items, gfp_t flags)
+{
+	struct tree_mod_elem *tm = NULL;
+	struct tree_mod_elem **tm_list = NULL;
+	int ret = 0;
+	int i;
+	int locked = 0;
+
+	if (!tree_mod_need_log(fs_info, eb))
+		return 0;
+
+	tm_list = kzalloc(nr_items * sizeof(struct tree_mod_elem *), flags);
+	if (!tm_list)
+		return -ENOMEM;
+
+	tm = kzalloc(sizeof(*tm), flags);
+	if (!tm) {
+		ret = -ENOMEM;
+		goto free_tms;
+	}
+
+	tm->index = eb->start >> PAGE_CACHE_SHIFT;
+	tm->slot = src_slot;
+	tm->move.dst_slot = dst_slot;
+	tm->move.nr_items = nr_items;
+	tm->op = MOD_LOG_MOVE_KEYS;
+
+	for (i = 0; i + dst_slot < src_slot && i < nr_items; i++) {
+		tm_list[i] = alloc_tree_mod_elem(eb, i + dst_slot,
+		    MOD_LOG_KEY_REMOVE_WHILE_MOVING, flags);
+		if (!tm_list[i]) {
+			ret = -ENOMEM;
+			goto free_tms;
+		}
+	}
+
+	if (tree_mod_dont_log(fs_info, eb))
+		goto free_tms;
+	locked = 1;
+
+	/*
+	 * When we override something during the move, we log these removals.
+	 * This can only happen when we move towards the beginning of the
+	 * buffer, i.e. dst_slot < src_slot.
+	 */
+	for (i = 0; i + dst_slot < src_slot && i < nr_items; i++) {
+		ret = __tree_mod_log_insert(fs_info, tm_list[i]);
+		if (ret)
+			goto free_tms;
+	}
+
+	ret = __tree_mod_log_insert(fs_info, tm);
+	if (ret)
+		goto free_tms;
+	tree_mod_log_write_unlock(fs_info);
+	kfree(tm_list);
+
+	return 0;
+free_tms:
+	for (i = 0; i < nr_items; i++) {
+		if (tm_list[i] && !RB_EMPTY_NODE(&tm_list[i]->node))
+			rb_erase(&tm_list[i]->node, &fs_info->tree_mod_log);
+		kfree(tm_list[i]);
+	}
+	if (locked)
+		tree_mod_log_write_unlock(fs_info);
+	kfree(tm_list);
+	kfree(tm);
+
+	return ret;
+}
+
+static inline int
+__tree_mod_log_free_eb(struct btrfs_fs_info *fs_info,
+		       struct tree_mod_elem **tm_list,
+		       int nritems)
+{
+	int i, j;
+	int ret;
+
+	for (i = nritems - 1; i >= 0; i--) {
+		ret = __tree_mod_log_insert(fs_info, tm_list[i]);
+		if (ret) {
+			for (j = nritems - 1; j > i; j--)
+				rb_erase(&tm_list[j]->node,
+					 &fs_info->tree_mod_log);
+			return ret;
+		}
+	}
+
+	return 0;
+}
+
+static noinline int
+tree_mod_log_insert_root(struct btrfs_fs_info *fs_info,
+			 struct extent_buffer *old_root,
+			 struct extent_buffer *new_root, gfp_t flags,
+			 int log_removal)
+{
+	struct tree_mod_elem *tm = NULL;
+	struct tree_mod_elem **tm_list = NULL;
+	int nritems = 0;
+	int ret = 0;
+	int i;
+
+	if (!tree_mod_need_log(fs_info, NULL))
+		return 0;
+
+	if (log_removal && btrfs_header_level(old_root) > 0) {
+		nritems = btrfs_header_nritems(old_root);
+		tm_list = kzalloc(nritems * sizeof(struct tree_mod_elem *),
+				  flags);
+		if (!tm_list) {
+			ret = -ENOMEM;
+			goto free_tms;
+		}
+		for (i = 0; i < nritems; i++) {
+			tm_list[i] = alloc_tree_mod_elem(old_root, i,
+			    MOD_LOG_KEY_REMOVE_WHILE_FREEING, flags);
+			if (!tm_list[i]) {
+				ret = -ENOMEM;
+				goto free_tms;
+			}
+		}
+	}
+
+	tm = kzalloc(sizeof(*tm), flags);
+	if (!tm) {
+		ret = -ENOMEM;
+		goto free_tms;
+	}
+
+	tm->index = new_root->start >> PAGE_CACHE_SHIFT;
+	tm->old_root.logical = old_root->start;
+	tm->old_root.level = btrfs_header_level(old_root);
+	tm->generation = btrfs_header_generation(old_root);
+	tm->op = MOD_LOG_ROOT_REPLACE;
+
+	if (tree_mod_dont_log(fs_info, NULL))
+		goto free_tms;
+
+	if (tm_list)
+		ret = __tree_mod_log_free_eb(fs_info, tm_list, nritems);
+	if (!ret)
+		ret = __tree_mod_log_insert(fs_info, tm);
+
+	tree_mod_log_write_unlock(fs_info);
+	if (ret)
+		goto free_tms;
+	kfree(tm_list);
+
+	return ret;
+
+free_tms:
+	if (tm_list) {
+		for (i = 0; i < nritems; i++)
+			kfree(tm_list[i]);
+		kfree(tm_list);
+	}
+	kfree(tm);
+
+	return ret;
+}
+
+static struct tree_mod_elem *
+__tree_mod_log_search(struct btrfs_fs_info *fs_info, u64 start, u64 min_seq,
+		      int smallest)
+{
+	struct rb_root *tm_root;
+	struct rb_node *node;
+	struct tree_mod_elem *cur = NULL;
+	struct tree_mod_elem *found = NULL;
+	u64 index = start >> PAGE_CACHE_SHIFT;
+
+	tree_mod_log_read_lock(fs_info);
+	tm_root = &fs_info->tree_mod_log;
+	node = tm_root->rb_node;
+	while (node) {
+		cur = container_of(node, struct tree_mod_elem, node);
+		if (cur->index < index) {
+			node = node->rb_left;
+		} else if (cur->index > index) {
+			node = node->rb_right;
+		} else if (cur->seq < min_seq) {
+			node = node->rb_left;
+		} else if (!smallest) {
+			/* we want the node with the highest seq */
+			if (found)
+				BUG_ON(found->seq > cur->seq);
+			found = cur;
+			node = node->rb_left;
+		} else if (cur->seq > min_seq) {
+			/* we want the node with the smallest seq */
+			if (found)
+				BUG_ON(found->seq < cur->seq);
+			found = cur;
+			node = node->rb_right;
+		} else {
+			found = cur;
+			break;
+		}
+	}
+	tree_mod_log_read_unlock(fs_info);
+
+	return found;
+}
+
+/*
+ * this returns the element from the log with the smallest time sequence
+ * value that's in the log (the oldest log item). any element with a time
+ * sequence lower than min_seq will be ignored.
+ */
+static struct tree_mod_elem *
+tree_mod_log_search_oldest(struct btrfs_fs_info *fs_info, u64 start,
+			   u64 min_seq)
+{
+	return __tree_mod_log_search(fs_info, start, min_seq, 1);
+}
+
+/*
+ * this returns the element from the log with the largest time sequence
+ * value that's in the log (the most recent log item). any element with
+ * a time sequence lower than min_seq will be ignored.
+ */
+static struct tree_mod_elem *
+tree_mod_log_search(struct btrfs_fs_info *fs_info, u64 start, u64 min_seq)
+{
+	return __tree_mod_log_search(fs_info, start, min_seq, 0);
+}
+
+static noinline int
+tree_mod_log_eb_copy(struct btrfs_fs_info *fs_info, struct extent_buffer *dst,
+		     struct extent_buffer *src, unsigned long dst_offset,
+		     unsigned long src_offset, int nr_items)
+{
+	int ret = 0;
+	struct tree_mod_elem **tm_list = NULL;
+	struct tree_mod_elem **tm_list_add, **tm_list_rem;
+	int i;
+	int locked = 0;
+
+	if (!tree_mod_need_log(fs_info, NULL))
+		return 0;
+
+	if (btrfs_header_level(dst) == 0 && btrfs_header_level(src) == 0)
+		return 0;
+
+	tm_list = kzalloc(nr_items * 2 * sizeof(struct tree_mod_elem *),
+			  GFP_NOFS);
+	if (!tm_list)
+		return -ENOMEM;
+
+	tm_list_add = tm_list;
+	tm_list_rem = tm_list + nr_items;
+	for (i = 0; i < nr_items; i++) {
+		tm_list_rem[i] = alloc_tree_mod_elem(src, i + src_offset,
+		    MOD_LOG_KEY_REMOVE, GFP_NOFS);
+		if (!tm_list_rem[i]) {
+			ret = -ENOMEM;
+			goto free_tms;
+		}
+
+		tm_list_add[i] = alloc_tree_mod_elem(dst, i + dst_offset,
+		    MOD_LOG_KEY_ADD, GFP_NOFS);
+		if (!tm_list_add[i]) {
+			ret = -ENOMEM;
+			goto free_tms;
+		}
+	}
+
+	if (tree_mod_dont_log(fs_info, NULL))
+		goto free_tms;
+	locked = 1;
+
+	for (i = 0; i < nr_items; i++) {
+		ret = __tree_mod_log_insert(fs_info, tm_list_rem[i]);
+		if (ret)
+			goto free_tms;
+		ret = __tree_mod_log_insert(fs_info, tm_list_add[i]);
+		if (ret)
+			goto free_tms;
+	}
+
+	tree_mod_log_write_unlock(fs_info);
+	kfree(tm_list);
+
+	return 0;
+
+free_tms:
+	for (i = 0; i < nr_items * 2; i++) {
+		if (tm_list[i] && !RB_EMPTY_NODE(&tm_list[i]->node))
+			rb_erase(&tm_list[i]->node, &fs_info->tree_mod_log);
+		kfree(tm_list[i]);
+	}
+	if (locked)
+		tree_mod_log_write_unlock(fs_info);
+	kfree(tm_list);
+
+	return ret;
+}
+
+static inline void
+tree_mod_log_eb_move(struct btrfs_fs_info *fs_info, struct extent_buffer *dst,
+		     int dst_offset, int src_offset, int nr_items)
+{
+	int ret;
+	ret = tree_mod_log_insert_move(fs_info, dst, dst_offset, src_offset,
+				       nr_items, GFP_NOFS);
+	BUG_ON(ret < 0);
+}
+
+static noinline void
+tree_mod_log_set_node_key(struct btrfs_fs_info *fs_info,
+			  struct extent_buffer *eb, int slot, int atomic)
+{
+	int ret;
+
+	ret = tree_mod_log_insert_key(fs_info, eb, slot,
+					MOD_LOG_KEY_REPLACE,
+					atomic ? GFP_ATOMIC : GFP_NOFS);
+	BUG_ON(ret < 0);
+}
+
+static noinline int
+tree_mod_log_free_eb(struct btrfs_fs_info *fs_info, struct extent_buffer *eb)
+{
+	struct tree_mod_elem **tm_list = NULL;
+	int nritems = 0;
+	int i;
+	int ret = 0;
+
+	if (btrfs_header_level(eb) == 0)
+		return 0;
+
+	if (!tree_mod_need_log(fs_info, NULL))
+		return 0;
+
+	nritems = btrfs_header_nritems(eb);
+	tm_list = kzalloc(nritems * sizeof(struct tree_mod_elem *),
+			  GFP_NOFS);
+	if (!tm_list)
+		return -ENOMEM;
+
+	for (i = 0; i < nritems; i++) {
+		tm_list[i] = alloc_tree_mod_elem(eb, i,
+		    MOD_LOG_KEY_REMOVE_WHILE_FREEING, GFP_NOFS);
+		if (!tm_list[i]) {
+			ret = -ENOMEM;
+			goto free_tms;
+		}
+	}
+
+	if (tree_mod_dont_log(fs_info, eb))
+		goto free_tms;
+
+	ret = __tree_mod_log_free_eb(fs_info, tm_list, nritems);
+	tree_mod_log_write_unlock(fs_info);
+	if (ret)
+		goto free_tms;
+	kfree(tm_list);
+
+	return 0;
+
+free_tms:
+	for (i = 0; i < nritems; i++)
+		kfree(tm_list[i]);
+	kfree(tm_list);
+
+	return ret;
+}
+
+static noinline void
+tree_mod_log_set_root_pointer(struct btrfs_root *root,
+			      struct extent_buffer *new_root_node,
+			      int log_removal)
+{
+	int ret;
+	ret = tree_mod_log_insert_root(root->fs_info, root->node,
+				       new_root_node, GFP_NOFS, log_removal);
+	BUG_ON(ret < 0);
+}
+
 /*
  * check if the tree block can be shared by multiple trees
  */
@@ -263,14 +964,14 @@ int btrfs_block_can_be_shared(struct btrfs_root *root,
 	 * snapshot and the block was not allocated by tree relocation,
 	 * we know the block is not shared.
 	 */
-	if (root->ref_cows &&
+	if (test_bit(BTRFS_ROOT_REF_COWS, &root->state) &&
 	    buf != root->node && buf != root->commit_root &&
 	    (btrfs_header_generation(buf) <=
 	     btrfs_root_last_snapshot(&root->root_item) ||
 	     btrfs_header_flag(buf, BTRFS_HEADER_FLAG_RELOC)))
 		return 1;
 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
-	if (root->ref_cows &&
+	if (test_bit(BTRFS_ROOT_REF_COWS, &root->state) &&
 	    btrfs_header_backref_rev(buf) < BTRFS_MIXED_BACKREF_REV)
 		return 1;
 #endif
@@ -280,7 +981,8 @@ int btrfs_block_can_be_shared(struct btrfs_root *root,
 static noinline int update_ref_for_cow(struct btrfs_trans_handle *trans,
 				       struct btrfs_root *root,
 				       struct extent_buffer *buf,
-				       struct extent_buffer *cow)
+				       struct extent_buffer *cow,
+				       int *last_ref)
 {
 	u64 refs;
 	u64 owner;
@@ -307,9 +1009,15 @@ static noinline int update_ref_for_cow(struct btrfs_trans_handle *trans,
 
 	if (btrfs_block_can_be_shared(root, buf)) {
 		ret = btrfs_lookup_extent_info(trans, root, buf->start,
-					       buf->len, &refs, &flags);
-		BUG_ON(ret);
-		BUG_ON(refs == 0);
+					       btrfs_header_level(buf), 1,
+					       &refs, &flags);
+		if (ret)
+			return ret;
+		if (refs == 0) {
+			ret = -EROFS;
+			btrfs_std_error(root->fs_info, ret);
+			return ret;
+		}
 	} else {
 		refs = 1;
 		if (root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID ||
@@ -327,45 +1035,49 @@ static noinline int update_ref_for_cow(struct btrfs_trans_handle *trans,
 		if ((owner == root->root_key.objectid ||
 		     root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID) &&
 		    !(flags & BTRFS_BLOCK_FLAG_FULL_BACKREF)) {
-			ret = btrfs_inc_ref(trans, root, buf, 1);
-			BUG_ON(ret);
+			ret = btrfs_inc_ref(trans, root, buf, 1, 1);
+			BUG_ON(ret); /* -ENOMEM */
 
 			if (root->root_key.objectid ==
 			    BTRFS_TREE_RELOC_OBJECTID) {
-				ret = btrfs_dec_ref(trans, root, buf, 0);
-				BUG_ON(ret);
-				ret = btrfs_inc_ref(trans, root, cow, 1);
-				BUG_ON(ret);
+				ret = btrfs_dec_ref(trans, root, buf, 0, 1);
+				BUG_ON(ret); /* -ENOMEM */
+				ret = btrfs_inc_ref(trans, root, cow, 1, 1);
+				BUG_ON(ret); /* -ENOMEM */
 			}
 			new_flags |= BTRFS_BLOCK_FLAG_FULL_BACKREF;
 		} else {
 
 			if (root->root_key.objectid ==
 			    BTRFS_TREE_RELOC_OBJECTID)
-				ret = btrfs_inc_ref(trans, root, cow, 1);
+				ret = btrfs_inc_ref(trans, root, cow, 1, 1);
 			else
-				ret = btrfs_inc_ref(trans, root, cow, 0);
-			BUG_ON(ret);
+				ret = btrfs_inc_ref(trans, root, cow, 0, 1);
+			BUG_ON(ret); /* -ENOMEM */
 		}
 		if (new_flags != 0) {
+			int level = btrfs_header_level(buf);
+
 			ret = btrfs_set_disk_extent_flags(trans, root,
 							  buf->start,
 							  buf->len,
-							  new_flags, 0);
-			BUG_ON(ret);
+							  new_flags, level, 0);
+			if (ret)
+				return ret;
 		}
 	} else {
 		if (flags & BTRFS_BLOCK_FLAG_FULL_BACKREF) {
 			if (root->root_key.objectid ==
 			    BTRFS_TREE_RELOC_OBJECTID)
-				ret = btrfs_inc_ref(trans, root, cow, 1);
+				ret = btrfs_inc_ref(trans, root, cow, 1, 1);
 			else
-				ret = btrfs_inc_ref(trans, root, cow, 0);
-			BUG_ON(ret);
-			ret = btrfs_dec_ref(trans, root, buf, 1);
-			BUG_ON(ret);
+				ret = btrfs_inc_ref(trans, root, cow, 0, 1);
+			BUG_ON(ret); /* -ENOMEM */
+			ret = btrfs_dec_ref(trans, root, buf, 1, 1);
+			BUG_ON(ret); /* -ENOMEM */
 		}
 		clean_tree_block(trans, root, buf);
+		*last_ref = 1;
 	}
 	return 0;
 }
@@ -391,7 +1103,8 @@ static noinline int __btrfs_cow_block(struct btrfs_trans_handle *trans,
 {
 	struct btrfs_disk_key disk_key;
 	struct extent_buffer *cow;
-	int level;
+	int level, ret;
+	int last_ref = 0;
 	int unlock_orig = 0;
 	u64 parent_start;
 
@@ -400,9 +1113,10 @@ static noinline int __btrfs_cow_block(struct btrfs_trans_handle *trans,
 
 	btrfs_assert_tree_locked(buf);
 
-	WARN_ON(root->ref_cows && trans->transid !=
-		root->fs_info->running_transaction->transid);
-	WARN_ON(root->ref_cows && trans->transid != root->last_trans);
+	WARN_ON(test_bit(BTRFS_ROOT_REF_COWS, &root->state) &&
+		trans->transid != root->fs_info->running_transaction->transid);
+	WARN_ON(test_bit(BTRFS_ROOT_REF_COWS, &root->state) &&
+		trans->transid != root->last_trans);
 
 	level = btrfs_header_level(buf);
 
@@ -438,11 +1152,20 @@ static noinline int __btrfs_cow_block(struct btrfs_trans_handle *trans,
 	else
 		btrfs_set_header_owner(cow, root->root_key.objectid);
 
-	write_extent_buffer(cow, root->fs_info->fsid,
-			    (unsigned long)btrfs_header_fsid(cow),
+	write_extent_buffer(cow, root->fs_info->fsid, btrfs_header_fsid(),
 			    BTRFS_FSID_SIZE);
 
-	update_ref_for_cow(trans, root, buf, cow);
+	ret = update_ref_for_cow(trans, root, buf, cow, &last_ref);
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
+		return ret;
+	}
+
+	if (test_bit(BTRFS_ROOT_REF_COWS, &root->state)) {
+		ret = btrfs_reloc_cow_block(trans, root, buf, cow);
+		if (ret)
+			return ret;
+	}
 
 	if (buf == root->node) {
 		WARN_ON(parent && parent != buf);
@@ -452,13 +1175,12 @@ static noinline int __btrfs_cow_block(struct btrfs_trans_handle *trans,
 		else
 			parent_start = 0;
 
-		spin_lock(&root->node_lock);
-		root->node = cow;
 		extent_buffer_get(cow);
-		spin_unlock(&root->node_lock);
+		tree_mod_log_set_root_pointer(root, cow, 1);
+		rcu_assign_pointer(root->node, cow);
 
-		btrfs_free_tree_block(trans, root, buf->start, buf->len,
-				parent_start, root->root_key.objectid, level);
+		btrfs_free_tree_block(trans, root, buf, parent_start,
+				      last_ref);
 		free_extent_buffer(buf);
 		add_root_to_dirty_list(root);
 	} else {
@@ -468,30 +1190,345 @@ static noinline int __btrfs_cow_block(struct btrfs_trans_handle *trans,
 			parent_start = 0;
 
 		WARN_ON(trans->transid != btrfs_header_generation(parent));
+		tree_mod_log_insert_key(root->fs_info, parent, parent_slot,
+					MOD_LOG_KEY_REPLACE, GFP_NOFS);
 		btrfs_set_node_blockptr(parent, parent_slot,
 					cow->start);
 		btrfs_set_node_ptr_generation(parent, parent_slot,
 					      trans->transid);
 		btrfs_mark_buffer_dirty(parent);
-		btrfs_free_tree_block(trans, root, buf->start, buf->len,
-				parent_start, root->root_key.objectid, level);
+		if (last_ref) {
+			ret = tree_mod_log_free_eb(root->fs_info, buf);
+			if (ret) {
+				btrfs_abort_transaction(trans, root, ret);
+				return ret;
+			}
+		}
+		btrfs_free_tree_block(trans, root, buf, parent_start,
+				      last_ref);
 	}
 	if (unlock_orig)
 		btrfs_tree_unlock(buf);
-	free_extent_buffer(buf);
+	free_extent_buffer_stale(buf);
 	btrfs_mark_buffer_dirty(cow);
 	*cow_ret = cow;
 	return 0;
 }
 
+/*
+ * returns the logical address of the oldest predecessor of the given root.
+ * entries older than time_seq are ignored.
+ */
+static struct tree_mod_elem *
+__tree_mod_log_oldest_root(struct btrfs_fs_info *fs_info,
+			   struct extent_buffer *eb_root, u64 time_seq)
+{
+	struct tree_mod_elem *tm;
+	struct tree_mod_elem *found = NULL;
+	u64 root_logical = eb_root->start;
+	int looped = 0;
+
+	if (!time_seq)
+		return NULL;
+
+	/*
+	 * the very last operation that's logged for a root is the replacement
+	 * operation (if it is replaced at all). this has the index of the *new*
+	 * root, making it the very first operation that's logged for this root.
+	 */
+	while (1) {
+		tm = tree_mod_log_search_oldest(fs_info, root_logical,
+						time_seq);
+		if (!looped && !tm)
+			return NULL;
+		/*
+		 * if there are no tree operation for the oldest root, we simply
+		 * return it. this should only happen if that (old) root is at
+		 * level 0.
+		 */
+		if (!tm)
+			break;
+
+		/*
+		 * if there's an operation that's not a root replacement, we
+		 * found the oldest version of our root. normally, we'll find a
+		 * MOD_LOG_KEY_REMOVE_WHILE_FREEING operation here.
+		 */
+		if (tm->op != MOD_LOG_ROOT_REPLACE)
+			break;
+
+		found = tm;
+		root_logical = tm->old_root.logical;
+		looped = 1;
+	}
+
+	/* if there's no old root to return, return what we found instead */
+	if (!found)
+		found = tm;
+
+	return found;
+}
+
+/*
+ * tm is a pointer to the first operation to rewind within eb. then, all
+ * previous operations will be rewinded (until we reach something older than
+ * time_seq).
+ */
+static void
+__tree_mod_log_rewind(struct btrfs_fs_info *fs_info, struct extent_buffer *eb,
+		      u64 time_seq, struct tree_mod_elem *first_tm)
+{
+	u32 n;
+	struct rb_node *next;
+	struct tree_mod_elem *tm = first_tm;
+	unsigned long o_dst;
+	unsigned long o_src;
+	unsigned long p_size = sizeof(struct btrfs_key_ptr);
+
+	n = btrfs_header_nritems(eb);
+	tree_mod_log_read_lock(fs_info);
+	while (tm && tm->seq >= time_seq) {
+		/*
+		 * all the operations are recorded with the operator used for
+		 * the modification. as we're going backwards, we do the
+		 * opposite of each operation here.
+		 */
+		switch (tm->op) {
+		case MOD_LOG_KEY_REMOVE_WHILE_FREEING:
+			BUG_ON(tm->slot < n);
+			/* Fallthrough */
+		case MOD_LOG_KEY_REMOVE_WHILE_MOVING:
+		case MOD_LOG_KEY_REMOVE:
+			btrfs_set_node_key(eb, &tm->key, tm->slot);
+			btrfs_set_node_blockptr(eb, tm->slot, tm->blockptr);
+			btrfs_set_node_ptr_generation(eb, tm->slot,
+						      tm->generation);
+			n++;
+			break;
+		case MOD_LOG_KEY_REPLACE:
+			BUG_ON(tm->slot >= n);
+			btrfs_set_node_key(eb, &tm->key, tm->slot);
+			btrfs_set_node_blockptr(eb, tm->slot, tm->blockptr);
+			btrfs_set_node_ptr_generation(eb, tm->slot,
+						      tm->generation);
+			break;
+		case MOD_LOG_KEY_ADD:
+			/* if a move operation is needed it's in the log */
+			n--;
+			break;
+		case MOD_LOG_MOVE_KEYS:
+			o_dst = btrfs_node_key_ptr_offset(tm->slot);
+			o_src = btrfs_node_key_ptr_offset(tm->move.dst_slot);
+			memmove_extent_buffer(eb, o_dst, o_src,
+					      tm->move.nr_items * p_size);
+			break;
+		case MOD_LOG_ROOT_REPLACE:
+			/*
+			 * this operation is special. for roots, this must be
+			 * handled explicitly before rewinding.
+			 * for non-roots, this operation may exist if the node
+			 * was a root: root A -> child B; then A gets empty and
+			 * B is promoted to the new root. in the mod log, we'll
+			 * have a root-replace operation for B, a tree block
+			 * that is no root. we simply ignore that operation.
+			 */
+			break;
+		}
+		next = rb_next(&tm->node);
+		if (!next)
+			break;
+		tm = container_of(next, struct tree_mod_elem, node);
+		if (tm->index != first_tm->index)
+			break;
+	}
+	tree_mod_log_read_unlock(fs_info);
+	btrfs_set_header_nritems(eb, n);
+}
+
+/*
+ * Called with eb read locked. If the buffer cannot be rewinded, the same buffer
+ * is returned. If rewind operations happen, a fresh buffer is returned. The
+ * returned buffer is always read-locked. If the returned buffer is not the
+ * input buffer, the lock on the input buffer is released and the input buffer
+ * is freed (its refcount is decremented).
+ */
+static struct extent_buffer *
+tree_mod_log_rewind(struct btrfs_fs_info *fs_info, struct btrfs_path *path,
+		    struct extent_buffer *eb, u64 time_seq)
+{
+	struct extent_buffer *eb_rewin;
+	struct tree_mod_elem *tm;
+
+	if (!time_seq)
+		return eb;
+
+	if (btrfs_header_level(eb) == 0)
+		return eb;
+
+	tm = tree_mod_log_search(fs_info, eb->start, time_seq);
+	if (!tm)
+		return eb;
+
+	btrfs_set_path_blocking(path);
+	btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
+
+	if (tm->op == MOD_LOG_KEY_REMOVE_WHILE_FREEING) {
+		BUG_ON(tm->slot != 0);
+		eb_rewin = alloc_dummy_extent_buffer(eb->start,
+						fs_info->tree_root->nodesize);
+		if (!eb_rewin) {
+			btrfs_tree_read_unlock_blocking(eb);
+			free_extent_buffer(eb);
+			return NULL;
+		}
+		btrfs_set_header_bytenr(eb_rewin, eb->start);
+		btrfs_set_header_backref_rev(eb_rewin,
+					     btrfs_header_backref_rev(eb));
+		btrfs_set_header_owner(eb_rewin, btrfs_header_owner(eb));
+		btrfs_set_header_level(eb_rewin, btrfs_header_level(eb));
+	} else {
+		eb_rewin = btrfs_clone_extent_buffer(eb);
+		if (!eb_rewin) {
+			btrfs_tree_read_unlock_blocking(eb);
+			free_extent_buffer(eb);
+			return NULL;
+		}
+	}
+
+	btrfs_clear_path_blocking(path, NULL, BTRFS_READ_LOCK);
+	btrfs_tree_read_unlock_blocking(eb);
+	free_extent_buffer(eb);
+
+	extent_buffer_get(eb_rewin);
+	btrfs_tree_read_lock(eb_rewin);
+	__tree_mod_log_rewind(fs_info, eb_rewin, time_seq, tm);
+	WARN_ON(btrfs_header_nritems(eb_rewin) >
+		BTRFS_NODEPTRS_PER_BLOCK(fs_info->tree_root));
+
+	return eb_rewin;
+}
+
+/*
+ * get_old_root() rewinds the state of @root's root node to the given @time_seq
+ * value. If there are no changes, the current root->root_node is returned. If
+ * anything changed in between, there's a fresh buffer allocated on which the
+ * rewind operations are done. In any case, the returned buffer is read locked.
+ * Returns NULL on error (with no locks held).
+ */
+static inline struct extent_buffer *
+get_old_root(struct btrfs_root *root, u64 time_seq)
+{
+	struct tree_mod_elem *tm;
+	struct extent_buffer *eb = NULL;
+	struct extent_buffer *eb_root;
+	struct extent_buffer *old;
+	struct tree_mod_root *old_root = NULL;
+	u64 old_generation = 0;
+	u64 logical;
+	u32 blocksize;
+
+	eb_root = btrfs_read_lock_root_node(root);
+	tm = __tree_mod_log_oldest_root(root->fs_info, eb_root, time_seq);
+	if (!tm)
+		return eb_root;
+
+	if (tm->op == MOD_LOG_ROOT_REPLACE) {
+		old_root = &tm->old_root;
+		old_generation = tm->generation;
+		logical = old_root->logical;
+	} else {
+		logical = eb_root->start;
+	}
+
+	tm = tree_mod_log_search(root->fs_info, logical, time_seq);
+	if (old_root && tm && tm->op != MOD_LOG_KEY_REMOVE_WHILE_FREEING) {
+		btrfs_tree_read_unlock(eb_root);
+		free_extent_buffer(eb_root);
+		blocksize = btrfs_level_size(root, old_root->level);
+		old = read_tree_block(root, logical, blocksize, 0);
+		if (WARN_ON(!old || !extent_buffer_uptodate(old))) {
+			free_extent_buffer(old);
+			btrfs_warn(root->fs_info,
+				"failed to read tree block %llu from get_old_root", logical);
+		} else {
+			eb = btrfs_clone_extent_buffer(old);
+			free_extent_buffer(old);
+		}
+	} else if (old_root) {
+		btrfs_tree_read_unlock(eb_root);
+		free_extent_buffer(eb_root);
+		eb = alloc_dummy_extent_buffer(logical, root->nodesize);
+	} else {
+		btrfs_set_lock_blocking_rw(eb_root, BTRFS_READ_LOCK);
+		eb = btrfs_clone_extent_buffer(eb_root);
+		btrfs_tree_read_unlock_blocking(eb_root);
+		free_extent_buffer(eb_root);
+	}
+
+	if (!eb)
+		return NULL;
+	extent_buffer_get(eb);
+	btrfs_tree_read_lock(eb);
+	if (old_root) {
+		btrfs_set_header_bytenr(eb, eb->start);
+		btrfs_set_header_backref_rev(eb, BTRFS_MIXED_BACKREF_REV);
+		btrfs_set_header_owner(eb, btrfs_header_owner(eb_root));
+		btrfs_set_header_level(eb, old_root->level);
+		btrfs_set_header_generation(eb, old_generation);
+	}
+	if (tm)
+		__tree_mod_log_rewind(root->fs_info, eb, time_seq, tm);
+	else
+		WARN_ON(btrfs_header_level(eb) != 0);
+	WARN_ON(btrfs_header_nritems(eb) > BTRFS_NODEPTRS_PER_BLOCK(root));
+
+	return eb;
+}
+
+int btrfs_old_root_level(struct btrfs_root *root, u64 time_seq)
+{
+	struct tree_mod_elem *tm;
+	int level;
+	struct extent_buffer *eb_root = btrfs_root_node(root);
+
+	tm = __tree_mod_log_oldest_root(root->fs_info, eb_root, time_seq);
+	if (tm && tm->op == MOD_LOG_ROOT_REPLACE) {
+		level = tm->old_root.level;
+	} else {
+		level = btrfs_header_level(eb_root);
+	}
+	free_extent_buffer(eb_root);
+
+	return level;
+}
+
 static inline int should_cow_block(struct btrfs_trans_handle *trans,
 				   struct btrfs_root *root,
 				   struct extent_buffer *buf)
 {
+#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
+	if (unlikely(test_bit(BTRFS_ROOT_DUMMY_ROOT, &root->state)))
+		return 0;
+#endif
+	/* ensure we can see the force_cow */
+	smp_rmb();
+
+	/*
+	 * We do not need to cow a block if
+	 * 1) this block is not created or changed in this transaction;
+	 * 2) this block does not belong to TREE_RELOC tree;
+	 * 3) the root is not forced COW.
+	 *
+	 * What is forced COW:
+	 *    when we create snapshot during commiting the transaction,
+	 *    after we've finished coping src root, we must COW the shared
+	 *    block to ensure the metadata consistency.
+	 */
 	if (btrfs_header_generation(buf) == trans->transid &&
 	    !btrfs_header_flag(buf, BTRFS_HEADER_FLAG_WRITTEN) &&
 	    !(root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID &&
-	      btrfs_header_flag(buf, BTRFS_HEADER_FLAG_RELOC)))
+	      btrfs_header_flag(buf, BTRFS_HEADER_FLAG_RELOC)) &&
+	    !test_bit(BTRFS_ROOT_FORCE_COW, &root->state))
 		return 0;
 	return 1;
 }
@@ -509,19 +1546,14 @@ noinline int btrfs_cow_block(struct btrfs_trans_handle *trans,
 	u64 search_start;
 	int ret;
 
-	if (trans->transaction != root->fs_info->running_transaction) {
-		printk(KERN_CRIT "trans %llu running %llu\n",
-		       (unsigned long long)trans->transid,
-		       (unsigned long long)
+	if (trans->transaction != root->fs_info->running_transaction)
+		WARN(1, KERN_CRIT "trans %llu running %llu\n",
+		       trans->transid,
 		       root->fs_info->running_transaction->transid);
-		WARN_ON(1);
-	}
-	if (trans->transid != root->fs_info->generation) {
-		printk(KERN_CRIT "trans %llu running %llu\n",
-		       (unsigned long long)trans->transid,
-		       (unsigned long long)root->fs_info->generation);
-		WARN_ON(1);
-	}
+
+	if (trans->transid != root->fs_info->generation)
+		WARN(1, KERN_CRIT "trans %llu running %llu\n",
+		       trans->transid, root->fs_info->generation);
 
 	if (!should_cow_block(trans, root, buf)) {
 		*cow_ret = buf;
@@ -536,6 +1568,9 @@ noinline int btrfs_cow_block(struct btrfs_trans_handle *trans,
 
 	ret = __btrfs_cow_block(trans, root, buf, parent,
 				 parent_slot, cow_ret, search_start, 0);
+
+	trace_btrfs_cow_block(root, buf, *cow_ret);
+
 	return ret;
 }
 
@@ -591,7 +1626,7 @@ int btrfs_comp_cpu_keys(struct btrfs_key *k1, struct btrfs_key *k2)
  */
 int btrfs_realloc_node(struct btrfs_trans_handle *trans,
 		       struct btrfs_root *root, struct extent_buffer *parent,
-		       int start_slot, int cache_only, u64 *last_ret,
+		       int start_slot, u64 *last_ret,
 		       struct btrfs_key *progress)
 {
 	struct extent_buffer *cur;
@@ -611,13 +1646,9 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans,
 	struct btrfs_disk_key disk_key;
 
 	parent_level = btrfs_header_level(parent);
-	if (cache_only && parent_level != 1)
-		return 0;
 
-	if (trans->transaction != root->fs_info->running_transaction)
-		WARN_ON(1);
-	if (trans->transid != root->fs_info->generation)
-		WARN_ON(1);
+	WARN_ON(trans->transaction != root->fs_info->running_transaction);
+	WARN_ON(trans->transid != root->fs_info->generation);
 
 	parent_nritems = btrfs_header_nritems(parent);
 	blocksize = btrfs_level_size(root, parent_level - 1);
@@ -631,14 +1662,6 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans,
 	for (i = start_slot; i < end_slot; i++) {
 		int close = 1;
 
-		if (!parent->map_token) {
-			map_extent_buffer(parent,
-					btrfs_node_key_ptr_offset(i),
-					sizeof(struct btrfs_key_ptr),
-					&parent->map_token, &parent->kaddr,
-					&parent->map_start, &parent->map_len,
-					KM_USER1);
-		}
 		btrfs_node_key(parent, &disk_key, i);
 		if (!progress_passed && comp_keys(&disk_key, progress) < 0)
 			continue;
@@ -661,27 +1684,26 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans,
 			last_block = blocknr;
 			continue;
 		}
-		if (parent->map_token) {
-			unmap_extent_buffer(parent, parent->map_token,
-					    KM_USER1);
-			parent->map_token = NULL;
-		}
 
 		cur = btrfs_find_tree_block(root, blocknr, blocksize);
 		if (cur)
-			uptodate = btrfs_buffer_uptodate(cur, gen);
+			uptodate = btrfs_buffer_uptodate(cur, gen, 0);
 		else
 			uptodate = 0;
 		if (!cur || !uptodate) {
-			if (cache_only) {
-				free_extent_buffer(cur);
-				continue;
-			}
 			if (!cur) {
 				cur = read_tree_block(root, blocknr,
 							 blocksize, gen);
+				if (!cur || !extent_buffer_uptodate(cur)) {
+					free_extent_buffer(cur);
+					return -EIO;
+				}
 			} else if (!uptodate) {
-				btrfs_read_buffer(cur, gen);
+				err = btrfs_read_buffer(cur, gen);
+				if (err) {
+					free_extent_buffer(cur);
+					return err;
+				}
 			}
 		}
 		if (search_start == 0)
@@ -704,11 +1726,6 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans,
 		btrfs_tree_unlock(cur);
 		free_extent_buffer(cur);
 	}
-	if (parent->map_token) {
-		unmap_extent_buffer(parent, parent->map_token,
-				    KM_USER1);
-		parent->map_token = NULL;
-	}
 	return err;
 }
 
@@ -726,122 +1743,6 @@ static inline unsigned int leaf_data_end(struct btrfs_root *root,
 	return btrfs_item_offset_nr(leaf, nr - 1);
 }
 
-/*
- * extra debugging checks to make sure all the items in a key are
- * well formed and in the proper order
- */
-static int check_node(struct btrfs_root *root, struct btrfs_path *path,
-		      int level)
-{
-	struct extent_buffer *parent = NULL;
-	struct extent_buffer *node = path->nodes[level];
-	struct btrfs_disk_key parent_key;
-	struct btrfs_disk_key node_key;
-	int parent_slot;
-	int slot;
-	struct btrfs_key cpukey;
-	u32 nritems = btrfs_header_nritems(node);
-
-	if (path->nodes[level + 1])
-		parent = path->nodes[level + 1];
-
-	slot = path->slots[level];
-	BUG_ON(nritems == 0);
-	if (parent) {
-		parent_slot = path->slots[level + 1];
-		btrfs_node_key(parent, &parent_key, parent_slot);
-		btrfs_node_key(node, &node_key, 0);
-		BUG_ON(memcmp(&parent_key, &node_key,
-			      sizeof(struct btrfs_disk_key)));
-		BUG_ON(btrfs_node_blockptr(parent, parent_slot) !=
-		       btrfs_header_bytenr(node));
-	}
-	BUG_ON(nritems > BTRFS_NODEPTRS_PER_BLOCK(root));
-	if (slot != 0) {
-		btrfs_node_key_to_cpu(node, &cpukey, slot - 1);
-		btrfs_node_key(node, &node_key, slot);
-		BUG_ON(comp_keys(&node_key, &cpukey) <= 0);
-	}
-	if (slot < nritems - 1) {
-		btrfs_node_key_to_cpu(node, &cpukey, slot + 1);
-		btrfs_node_key(node, &node_key, slot);
-		BUG_ON(comp_keys(&node_key, &cpukey) >= 0);
-	}
-	return 0;
-}
-
-/*
- * extra checking to make sure all the items in a leaf are
- * well formed and in the proper order
- */
-static int check_leaf(struct btrfs_root *root, struct btrfs_path *path,
-		      int level)
-{
-	struct extent_buffer *leaf = path->nodes[level];
-	struct extent_buffer *parent = NULL;
-	int parent_slot;
-	struct btrfs_key cpukey;
-	struct btrfs_disk_key parent_key;
-	struct btrfs_disk_key leaf_key;
-	int slot = path->slots[0];
-
-	u32 nritems = btrfs_header_nritems(leaf);
-
-	if (path->nodes[level + 1])
-		parent = path->nodes[level + 1];
-
-	if (nritems == 0)
-		return 0;
-
-	if (parent) {
-		parent_slot = path->slots[level + 1];
-		btrfs_node_key(parent, &parent_key, parent_slot);
-		btrfs_item_key(leaf, &leaf_key, 0);
-
-		BUG_ON(memcmp(&parent_key, &leaf_key,
-		       sizeof(struct btrfs_disk_key)));
-		BUG_ON(btrfs_node_blockptr(parent, parent_slot) !=
-		       btrfs_header_bytenr(leaf));
-	}
-	if (slot != 0 && slot < nritems - 1) {
-		btrfs_item_key(leaf, &leaf_key, slot);
-		btrfs_item_key_to_cpu(leaf, &cpukey, slot - 1);
-		if (comp_keys(&leaf_key, &cpukey) <= 0) {
-			btrfs_print_leaf(root, leaf);
-			printk(KERN_CRIT "slot %d offset bad key\n", slot);
-			BUG_ON(1);
-		}
-		if (btrfs_item_offset_nr(leaf, slot - 1) !=
-		       btrfs_item_end_nr(leaf, slot)) {
-			btrfs_print_leaf(root, leaf);
-			printk(KERN_CRIT "slot %d offset bad\n", slot);
-			BUG_ON(1);
-		}
-	}
-	if (slot < nritems - 1) {
-		btrfs_item_key(leaf, &leaf_key, slot);
-		btrfs_item_key_to_cpu(leaf, &cpukey, slot + 1);
-		BUG_ON(comp_keys(&leaf_key, &cpukey) >= 0);
-		if (btrfs_item_offset_nr(leaf, slot) !=
-			btrfs_item_end_nr(leaf, slot + 1)) {
-			btrfs_print_leaf(root, leaf);
-			printk(KERN_CRIT "slot %d offset bad\n", slot);
-			BUG_ON(1);
-		}
-	}
-	BUG_ON(btrfs_item_offset_nr(leaf, 0) +
-	       btrfs_item_size_nr(leaf, 0) != BTRFS_LEAF_DATA_SIZE(root));
-	return 0;
-}
-
-static noinline int check_block(struct btrfs_root *root,
-				struct btrfs_path *path, int level)
-{
-	return 0;
-	if (level == 0)
-		return check_leaf(root, path, level);
-	return check_node(root, path, level);
-}
 
 /*
  * search for key in the extent_buffer.  The items start at offset p,
@@ -865,7 +1766,6 @@ static noinline int generic_bin_search(struct extent_buffer *eb,
 	struct btrfs_disk_key *tmp = NULL;
 	struct btrfs_disk_key unaligned;
 	unsigned long offset;
-	char *map_token = NULL;
 	char *kaddr = NULL;
 	unsigned long map_start = 0;
 	unsigned long map_len = 0;
@@ -875,18 +1775,13 @@ static noinline int generic_bin_search(struct extent_buffer *eb,
 		mid = (low + high) / 2;
 		offset = p + mid * item_size;
 
-		if (!map_token || offset < map_start ||
+		if (!kaddr || offset < map_start ||
 		    (offset + sizeof(struct btrfs_disk_key)) >
 		    map_start + map_len) {
-			if (map_token) {
-				unmap_extent_buffer(eb, map_token, KM_USER0);
-				map_token = NULL;
-			}
 
 			err = map_private_extent_buffer(eb, offset,
 						sizeof(struct btrfs_disk_key),
-						&map_token, &kaddr,
-						&map_start, &map_len, KM_USER0);
+						&kaddr, &map_start, &map_len);
 
 			if (!err) {
 				tmp = (struct btrfs_disk_key *)(kaddr + offset -
@@ -909,14 +1804,10 @@ static noinline int generic_bin_search(struct extent_buffer *eb,
 			high = mid;
 		else {
 			*slot = mid;
-			if (map_token)
-				unmap_extent_buffer(eb, map_token, KM_USER0);
 			return 0;
 		}
 	}
 	*slot = low;
-	if (map_token)
-		unmap_extent_buffer(eb, map_token, KM_USER0);
 	return 1;
 }
 
@@ -927,20 +1818,18 @@ static noinline int generic_bin_search(struct extent_buffer *eb,
 static int bin_search(struct extent_buffer *eb, struct btrfs_key *key,
 		      int level, int *slot)
 {
-	if (level == 0) {
+	if (level == 0)
 		return generic_bin_search(eb,
 					  offsetof(struct btrfs_leaf, items),
 					  sizeof(struct btrfs_item),
 					  key, btrfs_header_nritems(eb),
 					  slot);
-	} else {
+	else
 		return generic_bin_search(eb,
 					  offsetof(struct btrfs_node, ptrs),
 					  sizeof(struct btrfs_key_ptr),
 					  key, btrfs_header_nritems(eb),
 					  slot);
-	}
-	return -1;
 }
 
 int btrfs_bin_search(struct extent_buffer *eb, struct btrfs_key *key,
@@ -949,6 +1838,22 @@ int btrfs_bin_search(struct extent_buffer *eb, struct btrfs_key *key,
 	return bin_search(eb, key, level, slot);
 }
 
+static void root_add_used(struct btrfs_root *root, u32 size)
+{
+	spin_lock(&root->accounting_lock);
+	btrfs_set_root_used(&root->root_item,
+			    btrfs_root_used(&root->root_item) + size);
+	spin_unlock(&root->accounting_lock);
+}
+
+static void root_sub_used(struct btrfs_root *root, u32 size)
+{
+	spin_lock(&root->accounting_lock);
+	btrfs_set_root_used(&root->root_item,
+			    btrfs_root_used(&root->root_item) - size);
+	spin_unlock(&root->accounting_lock);
+}
+
 /* given a node and slot number, this reads the blocks it points to.  The
  * extent buffer is returned with a reference taken (but unlocked).
  * NULL is returned on error.
@@ -957,6 +1862,8 @@ static noinline struct extent_buffer *read_node_slot(struct btrfs_root *root,
 				   struct extent_buffer *parent, int slot)
 {
 	int level = btrfs_header_level(parent);
+	struct extent_buffer *eb;
+
 	if (slot < 0)
 		return NULL;
 	if (slot >= btrfs_header_nritems(parent))
@@ -964,9 +1871,15 @@ static noinline struct extent_buffer *read_node_slot(struct btrfs_root *root,
 
 	BUG_ON(level == 0);
 
-	return read_tree_block(root, btrfs_node_blockptr(parent, slot),
-		       btrfs_level_size(root, level - 1),
-		       btrfs_node_ptr_generation(parent, slot));
+	eb = read_tree_block(root, btrfs_node_blockptr(parent, slot),
+			     btrfs_level_size(root, level - 1),
+			     btrfs_node_ptr_generation(parent, slot));
+	if (eb && !extent_buffer_uptodate(eb)) {
+		free_extent_buffer(eb);
+		eb = NULL;
+	}
+
+	return eb;
 }
 
 /*
@@ -986,7 +1899,6 @@ static noinline int balance_level(struct btrfs_trans_handle *trans,
 	int wret;
 	int pslot;
 	int orig_slot = path->slots[level];
-	int err_on_enospc = 0;
 	u64 orig_ptr;
 
 	if (level == 0)
@@ -994,14 +1906,16 @@ static noinline int balance_level(struct btrfs_trans_handle *trans,
 
 	mid = path->nodes[level];
 
-	WARN_ON(!path->locks[level]);
+	WARN_ON(path->locks[level] != BTRFS_WRITE_LOCK &&
+		path->locks[level] != BTRFS_WRITE_LOCK_BLOCKING);
 	WARN_ON(btrfs_header_generation(mid) != trans->transid);
 
 	orig_ptr = btrfs_node_blockptr(mid, orig_slot);
 
-	if (level < BTRFS_MAX_LEVEL - 1)
+	if (level < BTRFS_MAX_LEVEL - 1) {
 		parent = path->nodes[level + 1];
-	pslot = path->slots[level + 1];
+		pslot = path->slots[level + 1];
+	}
 
 	/*
 	 * deal with the case where there is only one pointer in the root
@@ -1015,15 +1929,23 @@ static noinline int balance_level(struct btrfs_trans_handle *trans,
 
 		/* promote the child to a root */
 		child = read_node_slot(root, mid, 0);
-		BUG_ON(!child);
+		if (!child) {
+			ret = -EROFS;
+			btrfs_std_error(root->fs_info, ret);
+			goto enospc;
+		}
+
 		btrfs_tree_lock(child);
 		btrfs_set_lock_blocking(child);
 		ret = btrfs_cow_block(trans, root, child, mid, 0, &child);
-		BUG_ON(ret);
+		if (ret) {
+			btrfs_tree_unlock(child);
+			free_extent_buffer(child);
+			goto enospc;
+		}
 
-		spin_lock(&root->node_lock);
-		root->node = child;
-		spin_unlock(&root->node_lock);
+		tree_mod_log_set_root_pointer(root, child, 1);
+		rcu_assign_pointer(root->node, child);
 
 		add_root_to_dirty_list(root);
 		btrfs_tree_unlock(child);
@@ -1034,19 +1956,17 @@ static noinline int balance_level(struct btrfs_trans_handle *trans,
 		btrfs_tree_unlock(mid);
 		/* once for the path */
 		free_extent_buffer(mid);
-		ret = btrfs_free_tree_block(trans, root, mid->start, mid->len,
-					    0, root->root_key.objectid, level);
+
+		root_sub_used(root, mid->len);
+		btrfs_free_tree_block(trans, root, mid, 0, 1);
 		/* once for the root ptr */
-		free_extent_buffer(mid);
-		return ret;
+		free_extent_buffer_stale(mid);
+		return 0;
 	}
 	if (btrfs_header_nritems(mid) >
 	    BTRFS_NODEPTRS_PER_BLOCK(root) / 4)
 		return 0;
 
-	if (btrfs_header_nritems(mid) < 2)
-		err_on_enospc = 1;
-
 	left = read_node_slot(root, parent, pslot - 1);
 	if (left) {
 		btrfs_tree_lock(left);
@@ -1076,8 +1996,6 @@ static noinline int balance_level(struct btrfs_trans_handle *trans,
 		wret = push_node_left(trans, root, left, mid, 1);
 		if (wret < 0)
 			ret = wret;
-		if (btrfs_header_nritems(mid) < 2)
-			err_on_enospc = 1;
 	}
 
 	/*
@@ -1088,26 +2006,18 @@ static noinline int balance_level(struct btrfs_trans_handle *trans,
 		if (wret < 0 && wret != -ENOSPC)
 			ret = wret;
 		if (btrfs_header_nritems(right) == 0) {
-			u64 bytenr = right->start;
-			u32 blocksize = right->len;
-
 			clean_tree_block(trans, root, right);
 			btrfs_tree_unlock(right);
-			free_extent_buffer(right);
+			del_ptr(root, path, level + 1, pslot + 1);
+			root_sub_used(root, right->len);
+			btrfs_free_tree_block(trans, root, right, 0, 1);
+			free_extent_buffer_stale(right);
 			right = NULL;
-			wret = del_ptr(trans, root, path, level + 1, pslot +
-				       1);
-			if (wret)
-				ret = wret;
-			wret = btrfs_free_tree_block(trans, root,
-						     bytenr, blocksize, 0,
-						     root->root_key.objectid,
-						     level);
-			if (wret)
-				ret = wret;
 		} else {
 			struct btrfs_disk_key right_key;
 			btrfs_node_key(right, &right_key, 0);
+			tree_mod_log_set_node_key(root->fs_info, parent,
+						  pslot + 1, 0);
 			btrfs_set_node_key(parent, &right_key, pslot + 1);
 			btrfs_mark_buffer_dirty(parent);
 		}
@@ -1122,7 +2032,11 @@ static noinline int balance_level(struct btrfs_trans_handle *trans,
 		 * otherwise we would have pulled some pointers from the
 		 * right
 		 */
-		BUG_ON(!left);
+		if (!left) {
+			ret = -EROFS;
+			btrfs_std_error(root->fs_info, ret);
+			goto enospc;
+		}
 		wret = balance_node_right(trans, root, mid, left);
 		if (wret < 0) {
 			ret = wret;
@@ -1136,25 +2050,19 @@ static noinline int balance_level(struct btrfs_trans_handle *trans,
 		BUG_ON(wret == 1);
 	}
 	if (btrfs_header_nritems(mid) == 0) {
-		/* we've managed to empty the middle node, drop it */
-		u64 bytenr = mid->start;
-		u32 blocksize = mid->len;
-
 		clean_tree_block(trans, root, mid);
 		btrfs_tree_unlock(mid);
-		free_extent_buffer(mid);
+		del_ptr(root, path, level + 1, pslot);
+		root_sub_used(root, mid->len);
+		btrfs_free_tree_block(trans, root, mid, 0, 1);
+		free_extent_buffer_stale(mid);
 		mid = NULL;
-		wret = del_ptr(trans, root, path, level + 1, pslot);
-		if (wret)
-			ret = wret;
-		wret = btrfs_free_tree_block(trans, root, bytenr, blocksize,
-					 0, root->root_key.objectid, level);
-		if (wret)
-			ret = wret;
 	} else {
 		/* update the parent key to reflect our changes */
 		struct btrfs_disk_key mid_key;
 		btrfs_node_key(mid, &mid_key, 0);
+		tree_mod_log_set_node_key(root->fs_info, parent,
+					  pslot, 0);
 		btrfs_set_node_key(parent, &mid_key, pslot);
 		btrfs_mark_buffer_dirty(parent);
 	}
@@ -1177,7 +2085,6 @@ static noinline int balance_level(struct btrfs_trans_handle *trans,
 		}
 	}
 	/* double check we haven't messed things up */
-	check_block(root, path, level);
 	if (orig_ptr !=
 	    btrfs_node_blockptr(path->nodes[level], path->slots[level]))
 		BUG();
@@ -1210,18 +2117,17 @@ static noinline int push_nodes_for_insert(struct btrfs_trans_handle *trans,
 	int wret;
 	int pslot;
 	int orig_slot = path->slots[level];
-	u64 orig_ptr;
 
 	if (level == 0)
 		return 1;
 
 	mid = path->nodes[level];
 	WARN_ON(btrfs_header_generation(mid) != trans->transid);
-	orig_ptr = btrfs_node_blockptr(mid, orig_slot);
 
-	if (level < BTRFS_MAX_LEVEL - 1)
+	if (level < BTRFS_MAX_LEVEL - 1) {
 		parent = path->nodes[level + 1];
-	pslot = path->slots[level + 1];
+		pslot = path->slots[level + 1];
+	}
 
 	if (!parent)
 		return 1;
@@ -1254,6 +2160,8 @@ static noinline int push_nodes_for_insert(struct btrfs_trans_handle *trans,
 			struct btrfs_disk_key disk_key;
 			orig_slot += left_nr;
 			btrfs_node_key(mid, &disk_key, 0);
+			tree_mod_log_set_node_key(root->fs_info, parent,
+						  pslot, 0);
 			btrfs_set_node_key(parent, &disk_key, pslot);
 			btrfs_mark_buffer_dirty(parent);
 			if (btrfs_header_nritems(left) > orig_slot) {
@@ -1305,6 +2213,8 @@ static noinline int push_nodes_for_insert(struct btrfs_trans_handle *trans,
 			struct btrfs_disk_key disk_key;
 
 			btrfs_node_key(right, &disk_key, 0);
+			tree_mod_log_set_node_key(root->fs_info, parent,
+						  pslot + 1, 0);
 			btrfs_set_node_key(parent, &disk_key, pslot + 1);
 			btrfs_mark_buffer_dirty(parent);
 
@@ -1341,6 +2251,7 @@ static void reada_for_search(struct btrfs_root *root,
 	u64 search;
 	u64 target;
 	u64 nread = 0;
+	u64 gen;
 	int direction = path->reada;
 	struct extent_buffer *eb;
 	u32 nr;
@@ -1367,6 +2278,7 @@ static void reada_for_search(struct btrfs_root *root,
 
 	nritems = btrfs_header_nritems(node);
 	nr = slot;
+
 	while (1) {
 		if (direction < 0) {
 			if (nr == 0)
@@ -1385,8 +2297,8 @@ static void reada_for_search(struct btrfs_root *root,
 		search = btrfs_node_blockptr(node, nr);
 		if ((search <= target && target - search <= 65536) ||
 		    (search > target && search - target <= 65536)) {
-			readahead_tree_block(root, search, blocksize,
-				     btrfs_node_ptr_generation(node, nr));
+			gen = btrfs_node_ptr_generation(node, nr);
+			readahead_tree_block(root, search, blocksize, gen);
 			nread += blocksize;
 		}
 		nscan++;
@@ -1395,12 +2307,8 @@ static void reada_for_search(struct btrfs_root *root,
 	}
 }
 
-/*
- * returns -EAGAIN if it had to drop the path, or zero if everything was in
- * cache
- */
-static noinline int reada_for_balance(struct btrfs_root *root,
-				      struct btrfs_path *path, int level)
+static noinline void reada_for_balance(struct btrfs_root *root,
+				       struct btrfs_path *path, int level)
 {
 	int slot;
 	int nritems;
@@ -1409,12 +2317,11 @@ static noinline int reada_for_balance(struct btrfs_root *root,
 	u64 gen;
 	u64 block1 = 0;
 	u64 block2 = 0;
-	int ret = 0;
 	int blocksize;
 
 	parent = path->nodes[level + 1];
 	if (!parent)
-		return 0;
+		return;
 
 	nritems = btrfs_header_nritems(parent);
 	slot = path->slots[level + 1];
@@ -1424,7 +2331,12 @@ static noinline int reada_for_balance(struct btrfs_root *root,
 		block1 = btrfs_node_blockptr(parent, slot - 1);
 		gen = btrfs_node_ptr_generation(parent, slot - 1);
 		eb = btrfs_find_tree_block(root, block1, blocksize);
-		if (eb && btrfs_buffer_uptodate(eb, gen))
+		/*
+		 * if we get -eagain from btrfs_buffer_uptodate, we
+		 * don't want to return eagain here.  That will loop
+		 * forever
+		 */
+		if (eb && btrfs_buffer_uptodate(eb, gen, 1) != 0)
 			block1 = 0;
 		free_extent_buffer(eb);
 	}
@@ -1432,32 +2344,15 @@ static noinline int reada_for_balance(struct btrfs_root *root,
 		block2 = btrfs_node_blockptr(parent, slot + 1);
 		gen = btrfs_node_ptr_generation(parent, slot + 1);
 		eb = btrfs_find_tree_block(root, block2, blocksize);
-		if (eb && btrfs_buffer_uptodate(eb, gen))
+		if (eb && btrfs_buffer_uptodate(eb, gen, 1) != 0)
 			block2 = 0;
 		free_extent_buffer(eb);
 	}
-	if (block1 || block2) {
-		ret = -EAGAIN;
-
-		/* release the whole path */
-		btrfs_release_path(root, path);
 
-		/* read the blocks */
-		if (block1)
-			readahead_tree_block(root, block1, blocksize, 0);
-		if (block2)
-			readahead_tree_block(root, block2, blocksize, 0);
-
-		if (block1) {
-			eb = read_tree_block(root, block1, blocksize, 0);
-			free_extent_buffer(eb);
-		}
-		if (block2) {
-			eb = read_tree_block(root, block2, blocksize, 0);
-			free_extent_buffer(eb);
-		}
-	}
-	return ret;
+	if (block1)
+		readahead_tree_block(root, block1, blocksize, 0);
+	if (block2)
+		readahead_tree_block(root, block2, blocksize, 0);
 }
 
 
@@ -1475,7 +2370,8 @@ static noinline int reada_for_balance(struct btrfs_root *root,
  * if lowest_unlock is 1, level 0 won't be unlocked
  */
 static noinline void unlock_up(struct btrfs_path *path, int level,
-			       int lowest_unlock)
+			       int lowest_unlock, int min_write_lock_level,
+			       int *write_lock_level)
 {
 	int i;
 	int skip_level = level;
@@ -1505,8 +2401,13 @@ static noinline void unlock_up(struct btrfs_path *path, int level,
 
 		t = path->nodes[i];
 		if (i >= lowest_unlock && i > skip_level && path->locks[i]) {
-			btrfs_tree_unlock(t);
+			btrfs_tree_unlock_rw(t, path->locks[i]);
 			path->locks[i] = 0;
+			if (write_lock_level &&
+			    i > min_write_lock_level &&
+			    i <= *write_lock_level) {
+				*write_lock_level = i - 1;
+			}
 		}
 	}
 }
@@ -1532,7 +2433,7 @@ noinline void btrfs_unlock_up_safe(struct btrfs_path *path, int level)
 			continue;
 		if (!path->locks[i])
 			continue;
-		btrfs_tree_unlock(path->nodes[i]);
+		btrfs_tree_unlock_rw(path->nodes[i], path->locks[i]);
 		path->locks[i] = 0;
 	}
 }
@@ -1549,7 +2450,7 @@ static int
 read_block_for_search(struct btrfs_trans_handle *trans,
 		       struct btrfs_root *root, struct btrfs_path *p,
 		       struct extent_buffer **eb_ret, int level, int slot,
-		       struct btrfs_key *key)
+		       struct btrfs_key *key, u64 time_seq)
 {
 	u64 blocknr;
 	u64 gen;
@@ -1563,13 +2464,30 @@ read_block_for_search(struct btrfs_trans_handle *trans,
 	blocksize = btrfs_level_size(root, level - 1);
 
 	tmp = btrfs_find_tree_block(root, blocknr, blocksize);
-	if (tmp && btrfs_buffer_uptodate(tmp, gen)) {
-		/*
-		 * we found an up to date block without sleeping, return
-		 * right away
+	if (tmp) {
+		/* first we do an atomic uptodate check */
+		if (btrfs_buffer_uptodate(tmp, gen, 1) > 0) {
+			*eb_ret = tmp;
+			return 0;
+		}
+
+		/* the pages were up to date, but we failed
+		 * the generation number check.  Do a full
+		 * read for the generation number that is correct.
+		 * We must do this without dropping locks so
+		 * we can trust our generation number
 		 */
-		*eb_ret = tmp;
-		return 0;
+		btrfs_set_path_blocking(p);
+
+		/* now we're allowed to do a blocking uptodate check */
+		ret = btrfs_read_buffer(tmp, gen);
+		if (!ret) {
+			*eb_ret = tmp;
+			return 0;
+		}
+		free_extent_buffer(tmp);
+		btrfs_release_path(p);
+		return -EIO;
 	}
 
 	/*
@@ -1582,15 +2500,14 @@ read_block_for_search(struct btrfs_trans_handle *trans,
 	btrfs_unlock_up_safe(p, level + 1);
 	btrfs_set_path_blocking(p);
 
-	if (tmp)
-		free_extent_buffer(tmp);
+	free_extent_buffer(tmp);
 	if (p->reada)
 		reada_for_search(root, p, level, slot, key->objectid);
 
-	btrfs_release_path(NULL, p);
+	btrfs_release_path(p);
 
 	ret = -EAGAIN;
-	tmp = read_tree_block(root, blocknr, blocksize, gen);
+	tmp = read_tree_block(root, blocknr, blocksize, 0);
 	if (tmp) {
 		/*
 		 * If the read above didn't mark this buffer up to date,
@@ -1598,7 +2515,7 @@ read_block_for_search(struct btrfs_trans_handle *trans,
 		 * and give up so that our caller doesn't loop forever
 		 * on our EAGAINs.
 		 */
-		if (!btrfs_buffer_uptodate(tmp, 0))
+		if (!btrfs_buffer_uptodate(tmp, 0, 0))
 			ret = -EIO;
 		free_extent_buffer(tmp);
 	}
@@ -1617,20 +2534,24 @@ read_block_for_search(struct btrfs_trans_handle *trans,
 static int
 setup_nodes_for_search(struct btrfs_trans_handle *trans,
 		       struct btrfs_root *root, struct btrfs_path *p,
-		       struct extent_buffer *b, int level, int ins_len)
+		       struct extent_buffer *b, int level, int ins_len,
+		       int *write_lock_level)
 {
 	int ret;
 	if ((p->search_for_split || ins_len > 0) && btrfs_header_nritems(b) >=
 	    BTRFS_NODEPTRS_PER_BLOCK(root) - 3) {
 		int sret;
 
-		sret = reada_for_balance(root, p, level);
-		if (sret)
+		if (*write_lock_level < level + 1) {
+			*write_lock_level = level + 1;
+			btrfs_release_path(p);
 			goto again;
+		}
 
 		btrfs_set_path_blocking(p);
+		reada_for_balance(root, p, level);
 		sret = split_node(trans, root, p, level);
-		btrfs_clear_path_blocking(p, NULL);
+		btrfs_clear_path_blocking(p, NULL, 0);
 
 		BUG_ON(sret > 0);
 		if (sret) {
@@ -1642,13 +2563,16 @@ setup_nodes_for_search(struct btrfs_trans_handle *trans,
 		   BTRFS_NODEPTRS_PER_BLOCK(root) / 2) {
 		int sret;
 
-		sret = reada_for_balance(root, p, level);
-		if (sret)
+		if (*write_lock_level < level + 1) {
+			*write_lock_level = level + 1;
+			btrfs_release_path(p);
 			goto again;
+		}
 
 		btrfs_set_path_blocking(p);
+		reada_for_balance(root, p, level);
 		sret = balance_level(trans, root, p, level);
-		btrfs_clear_path_blocking(p, NULL);
+		btrfs_clear_path_blocking(p, NULL, 0);
 
 		if (sret) {
 			ret = sret;
@@ -1656,7 +2580,7 @@ setup_nodes_for_search(struct btrfs_trans_handle *trans,
 		}
 		b = p->nodes[level];
 		if (!b) {
-			btrfs_release_path(NULL, p);
+			btrfs_release_path(p);
 			goto again;
 		}
 		BUG_ON(btrfs_header_nritems(b) == 1);
@@ -1669,6 +2593,83 @@ done:
 	return ret;
 }
 
+static void key_search_validate(struct extent_buffer *b,
+				struct btrfs_key *key,
+				int level)
+{
+#ifdef CONFIG_BTRFS_ASSERT
+	struct btrfs_disk_key disk_key;
+
+	btrfs_cpu_key_to_disk(&disk_key, key);
+
+	if (level == 0)
+		ASSERT(!memcmp_extent_buffer(b, &disk_key,
+		    offsetof(struct btrfs_leaf, items[0].key),
+		    sizeof(disk_key)));
+	else
+		ASSERT(!memcmp_extent_buffer(b, &disk_key,
+		    offsetof(struct btrfs_node, ptrs[0].key),
+		    sizeof(disk_key)));
+#endif
+}
+
+static int key_search(struct extent_buffer *b, struct btrfs_key *key,
+		      int level, int *prev_cmp, int *slot)
+{
+	if (*prev_cmp != 0) {
+		*prev_cmp = bin_search(b, key, level, slot);
+		return *prev_cmp;
+	}
+
+	key_search_validate(b, key, level);
+	*slot = 0;
+
+	return 0;
+}
+
+int btrfs_find_item(struct btrfs_root *fs_root, struct btrfs_path *found_path,
+		u64 iobjectid, u64 ioff, u8 key_type,
+		struct btrfs_key *found_key)
+{
+	int ret;
+	struct btrfs_key key;
+	struct extent_buffer *eb;
+	struct btrfs_path *path;
+
+	key.type = key_type;
+	key.objectid = iobjectid;
+	key.offset = ioff;
+
+	if (found_path == NULL) {
+		path = btrfs_alloc_path();
+		if (!path)
+			return -ENOMEM;
+	} else
+		path = found_path;
+
+	ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
+	if ((ret < 0) || (found_key == NULL)) {
+		if (path != found_path)
+			btrfs_free_path(path);
+		return ret;
+	}
+
+	eb = path->nodes[0];
+	if (ret && path->slots[0] >= btrfs_header_nritems(eb)) {
+		ret = btrfs_next_leaf(fs_root, path);
+		if (ret)
+			return ret;
+		eb = path->nodes[0];
+	}
+
+	btrfs_item_key_to_cpu(eb, found_key, path->slots[0]);
+	if (found_key->type != key.type ||
+			found_key->objectid != key.objectid)
+		return 1;
+
+	return 0;
+}
+
 /*
  * look for key in the tree.  path is filled in with nodes along the way
  * if key is found, we return zero and you can find the item in the leaf
@@ -1692,27 +2693,88 @@ int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root
 	int err;
 	int level;
 	int lowest_unlock = 1;
+	int root_lock;
+	/* everything at write_lock_level or lower must be write locked */
+	int write_lock_level = 0;
 	u8 lowest_level = 0;
+	int min_write_lock_level;
+	int prev_cmp;
 
 	lowest_level = p->lowest_level;
 	WARN_ON(lowest_level && ins_len > 0);
 	WARN_ON(p->nodes[0] != NULL);
+	BUG_ON(!cow && ins_len);
 
-	if (ins_len < 0)
+	if (ins_len < 0) {
 		lowest_unlock = 2;
 
+		/* when we are removing items, we might have to go up to level
+		 * two as we update tree pointers  Make sure we keep write
+		 * for those levels as well
+		 */
+		write_lock_level = 2;
+	} else if (ins_len > 0) {
+		/*
+		 * for inserting items, make sure we have a write lock on
+		 * level 1 so we can update keys
+		 */
+		write_lock_level = 1;
+	}
+
+	if (!cow)
+		write_lock_level = -1;
+
+	if (cow && (p->keep_locks || p->lowest_level))
+		write_lock_level = BTRFS_MAX_LEVEL;
+
+	min_write_lock_level = write_lock_level;
+
 again:
+	prev_cmp = -1;
+	/*
+	 * we try very hard to do read locks on the root
+	 */
+	root_lock = BTRFS_READ_LOCK;
+	level = 0;
 	if (p->search_commit_root) {
+		/*
+		 * the commit roots are read only
+		 * so we always do read locks
+		 */
+		if (p->need_commit_sem)
+			down_read(&root->fs_info->commit_root_sem);
 		b = root->commit_root;
 		extent_buffer_get(b);
+		level = btrfs_header_level(b);
+		if (p->need_commit_sem)
+			up_read(&root->fs_info->commit_root_sem);
 		if (!p->skip_locking)
-			btrfs_tree_lock(b);
+			btrfs_tree_read_lock(b);
 	} else {
-		if (p->skip_locking)
+		if (p->skip_locking) {
 			b = btrfs_root_node(root);
-		else
-			b = btrfs_lock_root_node(root);
+			level = btrfs_header_level(b);
+		} else {
+			/* we don't know the level of the root node
+			 * until we actually have it read locked
+			 */
+			b = btrfs_read_lock_root_node(root);
+			level = btrfs_header_level(b);
+			if (level <= write_lock_level) {
+				/* whoops, must trade for write lock */
+				btrfs_tree_read_unlock(b);
+				free_extent_buffer(b);
+				b = btrfs_lock_root_node(root);
+				root_lock = BTRFS_WRITE_LOCK;
+
+				/* the level might have changed, check again */
+				level = btrfs_header_level(b);
+			}
+		}
 	}
+	p->nodes[level] = b;
+	if (!p->skip_locking)
+		p->locks[level] = root_lock;
 
 	while (b) {
 		level = btrfs_header_level(b);
@@ -1721,10 +2783,6 @@ again:
 		 * setup the path here so we can release it under lock
 		 * contention with the cow code
 		 */
-		p->nodes[level] = b;
-		if (!p->skip_locking)
-			p->locks[level] = 1;
-
 		if (cow) {
 			/*
 			 * if we don't really need to cow this block
@@ -1736,26 +2794,30 @@ again:
 
 			btrfs_set_path_blocking(p);
 
+			/*
+			 * must have write locks on this node and the
+			 * parent
+			 */
+			if (level > write_lock_level ||
+			    (level + 1 > write_lock_level &&
+			    level + 1 < BTRFS_MAX_LEVEL &&
+			    p->nodes[level + 1])) {
+				write_lock_level = level + 1;
+				btrfs_release_path(p);
+				goto again;
+			}
+
 			err = btrfs_cow_block(trans, root, b,
 					      p->nodes[level + 1],
 					      p->slots[level + 1], &b);
 			if (err) {
-				free_extent_buffer(b);
 				ret = err;
 				goto done;
 			}
 		}
 cow_done:
-		BUG_ON(!cow && ins_len);
-		if (level != btrfs_header_level(b))
-			WARN_ON(1);
-		level = btrfs_header_level(b);
-
 		p->nodes[level] = b;
-		if (!p->skip_locking)
-			p->locks[level] = 1;
-
-		btrfs_clear_path_blocking(p, NULL);
+		btrfs_clear_path_blocking(p, NULL, 0);
 
 		/*
 		 * we have a lock on b and as long as we aren't changing
@@ -1763,21 +2825,21 @@ cow_done:
 		 * It is safe to drop the lock on our parent before we
 		 * go through the expensive btree search on b.
 		 *
-		 * If cow is true, then we might be changing slot zero,
-		 * which may require changing the parent.  So, we can't
-		 * drop the lock until after we know which slot we're
-		 * operating on.
+		 * If we're inserting or deleting (ins_len != 0), then we might
+		 * be changing slot zero, which may require changing the parent.
+		 * So, we can't drop the lock until after we know which slot
+		 * we're operating on.
 		 */
-		if (!cow)
-			btrfs_unlock_up_safe(p, level + 1);
+		if (!ins_len && !p->keep_locks) {
+			int u = level + 1;
 
-		ret = check_block(root, p, level);
-		if (ret) {
-			ret = -1;
-			goto done;
+			if (u < BTRFS_MAX_LEVEL && p->locks[u]) {
+				btrfs_tree_unlock_rw(p->nodes[u], p->locks[u]);
+				p->locks[u] = 0;
+			}
 		}
 
-		ret = bin_search(b, key, level, &slot);
+		ret = key_search(b, key, level, &prev_cmp, &slot);
 
 		if (level != 0) {
 			int dec = 0;
@@ -1787,7 +2849,7 @@ cow_done:
 			}
 			p->slots[level] = slot;
 			err = setup_nodes_for_search(trans, root, p, b, level,
-						     ins_len);
+					     ins_len, &write_lock_level);
 			if (err == -EAGAIN)
 				goto again;
 			if (err) {
@@ -1797,7 +2859,21 @@ cow_done:
 			b = p->nodes[level];
 			slot = p->slots[level];
 
-			unlock_up(p, level, lowest_unlock);
+			/*
+			 * slot 0 is special, if we change the key
+			 * we have to update the parent pointer
+			 * which means we must have a write lock
+			 * on the parent
+			 */
+			if (slot == 0 && ins_len &&
+			    write_lock_level < level + 1) {
+				write_lock_level = level + 1;
+				btrfs_release_path(p);
+				goto again;
+			}
+
+			unlock_up(p, level, lowest_unlock,
+				  min_write_lock_level, &write_lock_level);
 
 			if (level == lowest_level) {
 				if (dec)
@@ -1806,7 +2882,7 @@ cow_done:
 			}
 
 			err = read_block_for_search(trans, root, p,
-						    &b, level, slot, key);
+						    &b, level, slot, key, 0);
 			if (err == -EAGAIN)
 				goto again;
 			if (err) {
@@ -1815,23 +2891,42 @@ cow_done:
 			}
 
 			if (!p->skip_locking) {
-				btrfs_clear_path_blocking(p, NULL);
-				err = btrfs_try_spin_lock(b);
-
-				if (!err) {
-					btrfs_set_path_blocking(p);
-					btrfs_tree_lock(b);
-					btrfs_clear_path_blocking(p, b);
+				level = btrfs_header_level(b);
+				if (level <= write_lock_level) {
+					err = btrfs_try_tree_write_lock(b);
+					if (!err) {
+						btrfs_set_path_blocking(p);
+						btrfs_tree_lock(b);
+						btrfs_clear_path_blocking(p, b,
+								  BTRFS_WRITE_LOCK);
+					}
+					p->locks[level] = BTRFS_WRITE_LOCK;
+				} else {
+					err = btrfs_try_tree_read_lock(b);
+					if (!err) {
+						btrfs_set_path_blocking(p);
+						btrfs_tree_read_lock(b);
+						btrfs_clear_path_blocking(p, b,
+								  BTRFS_READ_LOCK);
+					}
+					p->locks[level] = BTRFS_READ_LOCK;
 				}
+				p->nodes[level] = b;
 			}
 		} else {
 			p->slots[level] = slot;
 			if (ins_len > 0 &&
 			    btrfs_leaf_free_space(root, b) < ins_len) {
+				if (write_lock_level < 1) {
+					write_lock_level = 1;
+					btrfs_release_path(p);
+					goto again;
+				}
+
 				btrfs_set_path_blocking(p);
 				err = split_leaf(trans, root, key,
 						 p, ins_len, ret == 0);
-				btrfs_clear_path_blocking(p, NULL);
+				btrfs_clear_path_blocking(p, NULL, 0);
 
 				BUG_ON(err > 0);
 				if (err) {
@@ -1840,7 +2935,8 @@ cow_done:
 				}
 			}
 			if (!p->search_for_split)
-				unlock_up(p, level, lowest_unlock);
+				unlock_up(p, level, lowest_unlock,
+					  min_write_lock_level, &write_lock_level);
 			goto done;
 		}
 	}
@@ -1853,26 +2949,209 @@ done:
 	if (!p->leave_spinning)
 		btrfs_set_path_blocking(p);
 	if (ret < 0)
-		btrfs_release_path(root, p);
+		btrfs_release_path(p);
 	return ret;
 }
 
 /*
+ * Like btrfs_search_slot, this looks for a key in the given tree. It uses the
+ * current state of the tree together with the operations recorded in the tree
+ * modification log to search for the key in a previous version of this tree, as
+ * denoted by the time_seq parameter.
+ *
+ * Naturally, there is no support for insert, delete or cow operations.
+ *
+ * The resulting path and return value will be set up as if we called
+ * btrfs_search_slot at that point in time with ins_len and cow both set to 0.
+ */
+int btrfs_search_old_slot(struct btrfs_root *root, struct btrfs_key *key,
+			  struct btrfs_path *p, u64 time_seq)
+{
+	struct extent_buffer *b;
+	int slot;
+	int ret;
+	int err;
+	int level;
+	int lowest_unlock = 1;
+	u8 lowest_level = 0;
+	int prev_cmp = -1;
+
+	lowest_level = p->lowest_level;
+	WARN_ON(p->nodes[0] != NULL);
+
+	if (p->search_commit_root) {
+		BUG_ON(time_seq);
+		return btrfs_search_slot(NULL, root, key, p, 0, 0);
+	}
+
+again:
+	b = get_old_root(root, time_seq);
+	level = btrfs_header_level(b);
+	p->locks[level] = BTRFS_READ_LOCK;
+
+	while (b) {
+		level = btrfs_header_level(b);
+		p->nodes[level] = b;
+		btrfs_clear_path_blocking(p, NULL, 0);
+
+		/*
+		 * we have a lock on b and as long as we aren't changing
+		 * the tree, there is no way to for the items in b to change.
+		 * It is safe to drop the lock on our parent before we
+		 * go through the expensive btree search on b.
+		 */
+		btrfs_unlock_up_safe(p, level + 1);
+
+		/*
+		 * Since we can unwind eb's we want to do a real search every
+		 * time.
+		 */
+		prev_cmp = -1;
+		ret = key_search(b, key, level, &prev_cmp, &slot);
+
+		if (level != 0) {
+			int dec = 0;
+			if (ret && slot > 0) {
+				dec = 1;
+				slot -= 1;
+			}
+			p->slots[level] = slot;
+			unlock_up(p, level, lowest_unlock, 0, NULL);
+
+			if (level == lowest_level) {
+				if (dec)
+					p->slots[level]++;
+				goto done;
+			}
+
+			err = read_block_for_search(NULL, root, p, &b, level,
+						    slot, key, time_seq);
+			if (err == -EAGAIN)
+				goto again;
+			if (err) {
+				ret = err;
+				goto done;
+			}
+
+			level = btrfs_header_level(b);
+			err = btrfs_try_tree_read_lock(b);
+			if (!err) {
+				btrfs_set_path_blocking(p);
+				btrfs_tree_read_lock(b);
+				btrfs_clear_path_blocking(p, b,
+							  BTRFS_READ_LOCK);
+			}
+			b = tree_mod_log_rewind(root->fs_info, p, b, time_seq);
+			if (!b) {
+				ret = -ENOMEM;
+				goto done;
+			}
+			p->locks[level] = BTRFS_READ_LOCK;
+			p->nodes[level] = b;
+		} else {
+			p->slots[level] = slot;
+			unlock_up(p, level, lowest_unlock, 0, NULL);
+			goto done;
+		}
+	}
+	ret = 1;
+done:
+	if (!p->leave_spinning)
+		btrfs_set_path_blocking(p);
+	if (ret < 0)
+		btrfs_release_path(p);
+
+	return ret;
+}
+
+/*
+ * helper to use instead of search slot if no exact match is needed but
+ * instead the next or previous item should be returned.
+ * When find_higher is true, the next higher item is returned, the next lower
+ * otherwise.
+ * When return_any and find_higher are both true, and no higher item is found,
+ * return the next lower instead.
+ * When return_any is true and find_higher is false, and no lower item is found,
+ * return the next higher instead.
+ * It returns 0 if any item is found, 1 if none is found (tree empty), and
+ * < 0 on error
+ */
+int btrfs_search_slot_for_read(struct btrfs_root *root,
+			       struct btrfs_key *key, struct btrfs_path *p,
+			       int find_higher, int return_any)
+{
+	int ret;
+	struct extent_buffer *leaf;
+
+again:
+	ret = btrfs_search_slot(NULL, root, key, p, 0, 0);
+	if (ret <= 0)
+		return ret;
+	/*
+	 * a return value of 1 means the path is at the position where the
+	 * item should be inserted. Normally this is the next bigger item,
+	 * but in case the previous item is the last in a leaf, path points
+	 * to the first free slot in the previous leaf, i.e. at an invalid
+	 * item.
+	 */
+	leaf = p->nodes[0];
+
+	if (find_higher) {
+		if (p->slots[0] >= btrfs_header_nritems(leaf)) {
+			ret = btrfs_next_leaf(root, p);
+			if (ret <= 0)
+				return ret;
+			if (!return_any)
+				return 1;
+			/*
+			 * no higher item found, return the next
+			 * lower instead
+			 */
+			return_any = 0;
+			find_higher = 0;
+			btrfs_release_path(p);
+			goto again;
+		}
+	} else {
+		if (p->slots[0] == 0) {
+			ret = btrfs_prev_leaf(root, p);
+			if (ret < 0)
+				return ret;
+			if (!ret) {
+				leaf = p->nodes[0];
+				if (p->slots[0] == btrfs_header_nritems(leaf))
+					p->slots[0]--;
+				return 0;
+			}
+			if (!return_any)
+				return 1;
+			/*
+			 * no lower item found, return the next
+			 * higher instead
+			 */
+			return_any = 0;
+			find_higher = 1;
+			btrfs_release_path(p);
+			goto again;
+		} else {
+			--p->slots[0];
+		}
+	}
+	return 0;
+}
+
+/*
  * adjust the pointers going up the tree, starting at level
  * making sure the right key of each node is points to 'key'.
  * This is used after shifting pointers to the left, so it stops
  * fixing up pointers when a given leaf/node is not in slot 0 of the
  * higher levels
  *
- * If this fails to write a tree block, it returns -1, but continues
- * fixing up the blocks in ram so the tree is consistent.
  */
-static int fixup_low_keys(struct btrfs_trans_handle *trans,
-			  struct btrfs_root *root, struct btrfs_path *path,
-			  struct btrfs_disk_key *key, int level)
+static void fixup_low_keys(struct btrfs_root *root, struct btrfs_path *path,
+			   struct btrfs_disk_key *key, int level)
 {
 	int i;
-	int ret = 0;
 	struct extent_buffer *t;
 
 	for (i = level; i < BTRFS_MAX_LEVEL; i++) {
@@ -1880,12 +3159,12 @@ static int fixup_low_keys(struct btrfs_trans_handle *trans,
 		if (!path->nodes[i])
 			break;
 		t = path->nodes[i];
+		tree_mod_log_set_node_key(root->fs_info, t, tslot, 1);
 		btrfs_set_node_key(t, key, tslot);
 		btrfs_mark_buffer_dirty(path->nodes[i]);
 		if (tslot != 0)
 			break;
 	}
-	return ret;
 }
 
 /*
@@ -1894,9 +3173,8 @@ static int fixup_low_keys(struct btrfs_trans_handle *trans,
  * This function isn't completely safe. It's the caller's responsibility
  * that the new key won't break the order
  */
-int btrfs_set_item_key_safe(struct btrfs_trans_handle *trans,
-			    struct btrfs_root *root, struct btrfs_path *path,
-			    struct btrfs_key *new_key)
+void btrfs_set_item_key_safe(struct btrfs_root *root, struct btrfs_path *path,
+			     struct btrfs_key *new_key)
 {
 	struct btrfs_disk_key disk_key;
 	struct extent_buffer *eb;
@@ -1906,21 +3184,18 @@ int btrfs_set_item_key_safe(struct btrfs_trans_handle *trans,
 	slot = path->slots[0];
 	if (slot > 0) {
 		btrfs_item_key(eb, &disk_key, slot - 1);
-		if (comp_keys(&disk_key, new_key) >= 0)
-			return -1;
+		BUG_ON(comp_keys(&disk_key, new_key) >= 0);
 	}
 	if (slot < btrfs_header_nritems(eb) - 1) {
 		btrfs_item_key(eb, &disk_key, slot + 1);
-		if (comp_keys(&disk_key, new_key) <= 0)
-			return -1;
+		BUG_ON(comp_keys(&disk_key, new_key) <= 0);
 	}
 
 	btrfs_cpu_key_to_disk(&disk_key, new_key);
 	btrfs_set_item_key(eb, &disk_key, slot);
 	btrfs_mark_buffer_dirty(eb);
 	if (slot == 0)
-		fixup_low_keys(trans, root, path, &disk_key, 1);
-	return 0;
+		fixup_low_keys(root, path, &disk_key, 1);
 }
 
 /*
@@ -1966,12 +3241,22 @@ static int push_node_left(struct btrfs_trans_handle *trans,
 	} else
 		push_items = min(src_nritems - 8, push_items);
 
+	ret = tree_mod_log_eb_copy(root->fs_info, dst, src, dst_nritems, 0,
+				   push_items);
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
+		return ret;
+	}
 	copy_extent_buffer(dst, src,
 			   btrfs_node_key_ptr_offset(dst_nritems),
 			   btrfs_node_key_ptr_offset(0),
 			   push_items * sizeof(struct btrfs_key_ptr));
 
 	if (push_items < src_nritems) {
+		/*
+		 * don't call tree_mod_log_eb_move here, key removal was already
+		 * fully logged by tree_mod_log_eb_copy above.
+		 */
 		memmove_extent_buffer(src, btrfs_node_key_ptr_offset(0),
 				      btrfs_node_key_ptr_offset(push_items),
 				      (src_nritems - push_items) *
@@ -2025,11 +3310,18 @@ static int balance_node_right(struct btrfs_trans_handle *trans,
 	if (max_push < push_items)
 		push_items = max_push;
 
+	tree_mod_log_eb_move(root->fs_info, dst, push_items, 0, dst_nritems);
 	memmove_extent_buffer(dst, btrfs_node_key_ptr_offset(push_items),
 				      btrfs_node_key_ptr_offset(0),
 				      (dst_nritems) *
 				      sizeof(struct btrfs_key_ptr));
 
+	ret = tree_mod_log_eb_copy(root->fs_info, dst, src, 0,
+				   src_nritems - push_items, push_items);
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
+		return ret;
+	}
 	copy_extent_buffer(dst, src,
 			   btrfs_node_key_ptr_offset(0),
 			   btrfs_node_key_ptr_offset(src_nritems - push_items),
@@ -2076,6 +3368,8 @@ static noinline int insert_new_root(struct btrfs_trans_handle *trans,
 	if (IS_ERR(c))
 		return PTR_ERR(c);
 
+	root_add_used(root, root->nodesize);
+
 	memset_extent_buffer(c, 0, 0, sizeof(struct btrfs_header));
 	btrfs_set_header_nritems(c, 1);
 	btrfs_set_header_level(c, level);
@@ -2084,13 +3378,11 @@ static noinline int insert_new_root(struct btrfs_trans_handle *trans,
 	btrfs_set_header_backref_rev(c, BTRFS_MIXED_BACKREF_REV);
 	btrfs_set_header_owner(c, root->root_key.objectid);
 
-	write_extent_buffer(c, root->fs_info->fsid,
-			    (unsigned long)btrfs_header_fsid(c),
+	write_extent_buffer(c, root->fs_info->fsid, btrfs_header_fsid(),
 			    BTRFS_FSID_SIZE);
 
 	write_extent_buffer(c, root->fs_info->chunk_tree_uuid,
-			    (unsigned long)btrfs_header_chunk_tree_uuid(c),
-			    BTRFS_UUID_SIZE);
+			    btrfs_header_chunk_tree_uuid(c), BTRFS_UUID_SIZE);
 
 	btrfs_set_node_key(c, &lower_key, 0);
 	btrfs_set_node_blockptr(c, 0, lower->start);
@@ -2101,10 +3393,9 @@ static noinline int insert_new_root(struct btrfs_trans_handle *trans,
 
 	btrfs_mark_buffer_dirty(c);
 
-	spin_lock(&root->node_lock);
 	old = root->node;
-	root->node = c;
-	spin_unlock(&root->node_lock);
+	tree_mod_log_set_root_pointer(root, c, 0);
+	rcu_assign_pointer(root->node, c);
 
 	/* the super has an extra ref to root->node */
 	free_extent_buffer(old);
@@ -2112,7 +3403,7 @@ static noinline int insert_new_root(struct btrfs_trans_handle *trans,
 	add_root_to_dirty_list(root);
 	extent_buffer_get(c);
 	path->nodes[level] = c;
-	path->locks[level] = 1;
+	path->locks[level] = BTRFS_WRITE_LOCK;
 	path->slots[level] = 0;
 	return 0;
 }
@@ -2123,35 +3414,42 @@ static noinline int insert_new_root(struct btrfs_trans_handle *trans,
  *
  * slot and level indicate where you want the key to go, and
  * blocknr is the block the key points to.
- *
- * returns zero on success and < 0 on any error
  */
-static int insert_ptr(struct btrfs_trans_handle *trans, struct btrfs_root
-		      *root, struct btrfs_path *path, struct btrfs_disk_key
-		      *key, u64 bytenr, int slot, int level)
+static void insert_ptr(struct btrfs_trans_handle *trans,
+		       struct btrfs_root *root, struct btrfs_path *path,
+		       struct btrfs_disk_key *key, u64 bytenr,
+		       int slot, int level)
 {
 	struct extent_buffer *lower;
 	int nritems;
+	int ret;
 
 	BUG_ON(!path->nodes[level]);
+	btrfs_assert_tree_locked(path->nodes[level]);
 	lower = path->nodes[level];
 	nritems = btrfs_header_nritems(lower);
 	BUG_ON(slot > nritems);
-	if (nritems == BTRFS_NODEPTRS_PER_BLOCK(root))
-		BUG();
+	BUG_ON(nritems == BTRFS_NODEPTRS_PER_BLOCK(root));
 	if (slot != nritems) {
+		if (level)
+			tree_mod_log_eb_move(root->fs_info, lower, slot + 1,
+					     slot, nritems - slot);
 		memmove_extent_buffer(lower,
 			      btrfs_node_key_ptr_offset(slot + 1),
 			      btrfs_node_key_ptr_offset(slot),
 			      (nritems - slot) * sizeof(struct btrfs_key_ptr));
 	}
+	if (level) {
+		ret = tree_mod_log_insert_key(root->fs_info, lower, slot,
+					      MOD_LOG_KEY_ADD, GFP_NOFS);
+		BUG_ON(ret < 0);
+	}
 	btrfs_set_node_key(lower, key, slot);
 	btrfs_set_node_blockptr(lower, slot, bytenr);
 	WARN_ON(trans->transid == 0);
 	btrfs_set_node_ptr_generation(lower, slot, trans->transid);
 	btrfs_set_header_nritems(lower, nritems + 1);
 	btrfs_mark_buffer_dirty(lower);
-	return 0;
 }
 
 /*
@@ -2172,13 +3470,21 @@ static noinline int split_node(struct btrfs_trans_handle *trans,
 	struct btrfs_disk_key disk_key;
 	int mid;
 	int ret;
-	int wret;
 	u32 c_nritems;
 
 	c = path->nodes[level];
 	WARN_ON(btrfs_header_generation(c) != trans->transid);
 	if (c == root->node) {
-		/* trying to split the root, lets make a new one */
+		/*
+		 * trying to split the root, lets make a new one
+		 *
+		 * tree mod log: We don't log_removal old root in
+		 * insert_new_root, because that root buffer will be kept as a
+		 * normal node. We are going to log removal of half of the
+		 * elements below with tree_mod_log_eb_copy. We're holding a
+		 * tree lock on the buffer, which is why we cannot race with
+		 * other tree_mod_log users.
+		 */
 		ret = insert_new_root(trans, root, path, level + 1);
 		if (ret)
 			return ret;
@@ -2202,6 +3508,8 @@ static noinline int split_node(struct btrfs_trans_handle *trans,
 	if (IS_ERR(split))
 		return PTR_ERR(split);
 
+	root_add_used(root, root->nodesize);
+
 	memset_extent_buffer(split, 0, 0, sizeof(struct btrfs_header));
 	btrfs_set_header_level(split, btrfs_header_level(c));
 	btrfs_set_header_bytenr(split, split->start);
@@ -2209,13 +3517,17 @@ static noinline int split_node(struct btrfs_trans_handle *trans,
 	btrfs_set_header_backref_rev(split, BTRFS_MIXED_BACKREF_REV);
 	btrfs_set_header_owner(split, root->root_key.objectid);
 	write_extent_buffer(split, root->fs_info->fsid,
-			    (unsigned long)btrfs_header_fsid(split),
-			    BTRFS_FSID_SIZE);
+			    btrfs_header_fsid(), BTRFS_FSID_SIZE);
 	write_extent_buffer(split, root->fs_info->chunk_tree_uuid,
-			    (unsigned long)btrfs_header_chunk_tree_uuid(split),
+			    btrfs_header_chunk_tree_uuid(split),
 			    BTRFS_UUID_SIZE);
 
-
+	ret = tree_mod_log_eb_copy(root->fs_info, split, c, 0,
+				   mid, c_nritems - mid);
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
+		return ret;
+	}
 	copy_extent_buffer(split, c,
 			   btrfs_node_key_ptr_offset(0),
 			   btrfs_node_key_ptr_offset(mid),
@@ -2227,11 +3539,8 @@ static noinline int split_node(struct btrfs_trans_handle *trans,
 	btrfs_mark_buffer_dirty(c);
 	btrfs_mark_buffer_dirty(split);
 
-	wret = insert_ptr(trans, root, path, &disk_key, split->start,
-			  path->slots[level + 1] + 1,
-			  level + 1);
-	if (wret)
-		ret = wret;
+	insert_ptr(trans, root, path, &disk_key, split->start,
+		   path->slots[level + 1] + 1, level + 1);
 
 	if (path->slots[level] >= mid) {
 		path->slots[level] -= mid;
@@ -2253,14 +3562,21 @@ static noinline int split_node(struct btrfs_trans_handle *trans,
  */
 static int leaf_space_used(struct extent_buffer *l, int start, int nr)
 {
+	struct btrfs_item *start_item;
+	struct btrfs_item *end_item;
+	struct btrfs_map_token token;
 	int data_len;
 	int nritems = btrfs_header_nritems(l);
 	int end = min(nritems, start + nr) - 1;
 
 	if (!nr)
 		return 0;
-	data_len = btrfs_item_end_nr(l, start);
-	data_len = data_len - btrfs_item_offset_nr(l, end);
+	btrfs_init_map_token(&token);
+	start_item = btrfs_item_nr(start);
+	end_item = btrfs_item_nr(end);
+	data_len = btrfs_token_item_offset(l, start_item, &token) +
+		btrfs_token_item_size(l, start_item, &token);
+	data_len = data_len - btrfs_token_item_offset(l, end_item, &token);
 	data_len += sizeof(struct btrfs_item) * nr;
 	WARN_ON(data_len < 0);
 	return data_len;
@@ -2278,23 +3594,29 @@ noinline int btrfs_leaf_free_space(struct btrfs_root *root,
 	int ret;
 	ret = BTRFS_LEAF_DATA_SIZE(root) - leaf_space_used(leaf, 0, nritems);
 	if (ret < 0) {
-		printk(KERN_CRIT "leaf free space ret %d, leaf data size %lu, "
-		       "used %d nritems %d\n",
+		btrfs_crit(root->fs_info,
+			"leaf free space ret %d, leaf data size %lu, used %d nritems %d",
 		       ret, (unsigned long) BTRFS_LEAF_DATA_SIZE(root),
 		       leaf_space_used(leaf, 0, nritems), nritems);
 	}
 	return ret;
 }
 
+/*
+ * min slot controls the lowest index we're willing to push to the
+ * right.  We'll push up to and including min_slot, but no lower
+ */
 static noinline int __push_leaf_right(struct btrfs_trans_handle *trans,
 				      struct btrfs_root *root,
 				      struct btrfs_path *path,
 				      int data_size, int empty,
 				      struct extent_buffer *right,
-				      int free_space, u32 left_nritems)
+				      int free_space, u32 left_nritems,
+				      u32 min_slot)
 {
 	struct extent_buffer *left = path->nodes[0];
 	struct extent_buffer *upper = path->nodes[1];
+	struct btrfs_map_token token;
 	struct btrfs_disk_key disk_key;
 	int slot;
 	u32 i;
@@ -2306,10 +3628,12 @@ static noinline int __push_leaf_right(struct btrfs_trans_handle *trans,
 	u32 data_end;
 	u32 this_item_size;
 
+	btrfs_init_map_token(&token);
+
 	if (empty)
 		nr = 0;
 	else
-		nr = 1;
+		nr = max_t(u32, 1, min_slot);
 
 	if (path->slots[0] >= left_nritems)
 		push_space += data_size;
@@ -2317,7 +3641,7 @@ static noinline int __push_leaf_right(struct btrfs_trans_handle *trans,
 	slot = path->slots[1];
 	i = left_nritems - 1;
 	while (i >= nr) {
-		item = btrfs_item_nr(left, i);
+		item = btrfs_item_nr(i);
 
 		if (!empty && push_items > 0) {
 			if (path->slots[0] > i)
@@ -2332,14 +3656,6 @@ static noinline int __push_leaf_right(struct btrfs_trans_handle *trans,
 		if (path->slots[0] == i)
 			push_space += data_size;
 
-		if (!left->map_token) {
-			map_extent_buffer(left, (unsigned long)item,
-					sizeof(struct btrfs_item),
-					&left->map_token, &left->kaddr,
-					&left->map_start, &left->map_len,
-					KM_USER1);
-		}
-
 		this_item_size = btrfs_item_size(left, item);
 		if (this_item_size + sizeof(*item) + push_space > free_space)
 			break;
@@ -2350,16 +3666,11 @@ static noinline int __push_leaf_right(struct btrfs_trans_handle *trans,
 			break;
 		i--;
 	}
-	if (left->map_token) {
-		unmap_extent_buffer(left, left->map_token, KM_USER1);
-		left->map_token = NULL;
-	}
 
 	if (push_items == 0)
 		goto out_unlock;
 
-	if (!empty && push_items == left_nritems)
-		WARN_ON(1);
+	WARN_ON(!empty && push_items == left_nritems);
 
 	/* push left to right */
 	right_nritems = btrfs_header_nritems(right);
@@ -2394,27 +3705,19 @@ static noinline int __push_leaf_right(struct btrfs_trans_handle *trans,
 	btrfs_set_header_nritems(right, right_nritems);
 	push_space = BTRFS_LEAF_DATA_SIZE(root);
 	for (i = 0; i < right_nritems; i++) {
-		item = btrfs_item_nr(right, i);
-		if (!right->map_token) {
-			map_extent_buffer(right, (unsigned long)item,
-					sizeof(struct btrfs_item),
-					&right->map_token, &right->kaddr,
-					&right->map_start, &right->map_len,
-					KM_USER1);
-		}
-		push_space -= btrfs_item_size(right, item);
-		btrfs_set_item_offset(right, item, push_space);
+		item = btrfs_item_nr(i);
+		push_space -= btrfs_token_item_size(right, item, &token);
+		btrfs_set_token_item_offset(right, item, push_space, &token);
 	}
 
-	if (right->map_token) {
-		unmap_extent_buffer(right, right->map_token, KM_USER1);
-		right->map_token = NULL;
-	}
 	left_nritems -= push_items;
 	btrfs_set_header_nritems(left, left_nritems);
 
 	if (left_nritems)
 		btrfs_mark_buffer_dirty(left);
+	else
+		clean_tree_block(trans, root, left);
+
 	btrfs_mark_buffer_dirty(right);
 
 	btrfs_item_key(right, &disk_key, 0);
@@ -2448,10 +3751,14 @@ out_unlock:
  *
  * returns 1 if the push failed because the other node didn't have enough
  * room, 0 if everything worked out and < 0 if there were major errors.
+ *
+ * this will push starting from min_slot to the end of the leaf.  It won't
+ * push any slot lower than min_slot
  */
 static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root
-			   *root, struct btrfs_path *path, int data_size,
-			   int empty)
+			   *root, struct btrfs_path *path,
+			   int min_data_size, int data_size,
+			   int empty, u32 min_slot)
 {
 	struct extent_buffer *left = path->nodes[0];
 	struct extent_buffer *right;
@@ -2472,6 +3779,9 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root
 	btrfs_assert_tree_locked(path->nodes[1]);
 
 	right = read_node_slot(root, upper, slot + 1);
+	if (right == NULL)
+		return 1;
+
 	btrfs_tree_lock(right);
 	btrfs_set_lock_blocking(right);
 
@@ -2493,8 +3803,21 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root
 	if (left_nritems == 0)
 		goto out_unlock;
 
-	return __push_leaf_right(trans, root, path, data_size, empty,
-				right, free_space, left_nritems);
+	if (path->slots[0] == left_nritems && !empty) {
+		/* Key greater than all keys in the leaf, right neighbor has
+		 * enough room for it and we're not emptying our leaf to delete
+		 * it, therefore use right neighbor to insert the new item and
+		 * no need to touch/dirty our left leaft. */
+		btrfs_tree_unlock(left);
+		free_extent_buffer(left);
+		path->nodes[0] = right;
+		path->slots[0] = 0;
+		path->slots[1]++;
+		return 0;
+	}
+
+	return __push_leaf_right(trans, root, path, min_data_size, empty,
+				right, free_space, left_nritems, min_slot);
 out_unlock:
 	btrfs_tree_unlock(right);
 	free_extent_buffer(right);
@@ -2504,16 +3827,20 @@ out_unlock:
 /*
  * push some data in the path leaf to the left, trying to free up at
  * least data_size bytes.  returns zero if the push worked, nonzero otherwise
+ *
+ * max_slot can put a limit on how far into the leaf we'll push items.  The
+ * item at 'max_slot' won't be touched.  Use (u32)-1 to make us do all the
+ * items
  */
 static noinline int __push_leaf_left(struct btrfs_trans_handle *trans,
 				     struct btrfs_root *root,
 				     struct btrfs_path *path, int data_size,
 				     int empty, struct extent_buffer *left,
-				     int free_space, int right_nritems)
+				     int free_space, u32 right_nritems,
+				     u32 max_slot)
 {
 	struct btrfs_disk_key disk_key;
 	struct extent_buffer *right = path->nodes[0];
-	int slot;
 	int i;
 	int push_space = 0;
 	int push_items = 0;
@@ -2521,26 +3848,19 @@ static noinline int __push_leaf_left(struct btrfs_trans_handle *trans,
 	u32 old_left_nritems;
 	u32 nr;
 	int ret = 0;
-	int wret;
 	u32 this_item_size;
 	u32 old_left_item_size;
+	struct btrfs_map_token token;
 
-	slot = path->slots[1];
+	btrfs_init_map_token(&token);
 
 	if (empty)
-		nr = right_nritems;
+		nr = min(right_nritems, max_slot);
 	else
-		nr = right_nritems - 1;
+		nr = min(right_nritems - 1, max_slot);
 
 	for (i = 0; i < nr; i++) {
-		item = btrfs_item_nr(right, i);
-		if (!right->map_token) {
-			map_extent_buffer(right, (unsigned long)item,
-					sizeof(struct btrfs_item),
-					&right->map_token, &right->kaddr,
-					&right->map_start, &right->map_len,
-					KM_USER1);
-		}
+		item = btrfs_item_nr(i);
 
 		if (!empty && push_items > 0) {
 			if (path->slots[0] < i)
@@ -2563,17 +3883,11 @@ static noinline int __push_leaf_left(struct btrfs_trans_handle *trans,
 		push_space += this_item_size + sizeof(*item);
 	}
 
-	if (right->map_token) {
-		unmap_extent_buffer(right, right->map_token, KM_USER1);
-		right->map_token = NULL;
-	}
-
 	if (push_items == 0) {
 		ret = 1;
 		goto out;
 	}
-	if (!empty && push_items == btrfs_header_nritems(right))
-		WARN_ON(1);
+	WARN_ON(!empty && push_items == btrfs_header_nritems(right));
 
 	/* push data from right to left */
 	copy_extent_buffer(left, right,
@@ -2596,31 +3910,19 @@ static noinline int __push_leaf_left(struct btrfs_trans_handle *trans,
 	for (i = old_left_nritems; i < old_left_nritems + push_items; i++) {
 		u32 ioff;
 
-		item = btrfs_item_nr(left, i);
-		if (!left->map_token) {
-			map_extent_buffer(left, (unsigned long)item,
-					sizeof(struct btrfs_item),
-					&left->map_token, &left->kaddr,
-					&left->map_start, &left->map_len,
-					KM_USER1);
-		}
+		item = btrfs_item_nr(i);
 
-		ioff = btrfs_item_offset(left, item);
-		btrfs_set_item_offset(left, item,
-		      ioff - (BTRFS_LEAF_DATA_SIZE(root) - old_left_item_size));
+		ioff = btrfs_token_item_offset(left, item, &token);
+		btrfs_set_token_item_offset(left, item,
+		      ioff - (BTRFS_LEAF_DATA_SIZE(root) - old_left_item_size),
+		      &token);
 	}
 	btrfs_set_header_nritems(left, old_left_nritems + push_items);
-	if (left->map_token) {
-		unmap_extent_buffer(left, left->map_token, KM_USER1);
-		left->map_token = NULL;
-	}
 
 	/* fixup right node */
-	if (push_items > right_nritems) {
-		printk(KERN_CRIT "push items %d nr %u\n", push_items,
+	if (push_items > right_nritems)
+		WARN(1, KERN_CRIT "push items %d nr %u\n", push_items,
 		       right_nritems);
-		WARN_ON(1);
-	}
 
 	if (push_items < right_nritems) {
 		push_space = btrfs_item_offset_nr(right, push_items - 1) -
@@ -2639,38 +3941,25 @@ static noinline int __push_leaf_left(struct btrfs_trans_handle *trans,
 	btrfs_set_header_nritems(right, right_nritems);
 	push_space = BTRFS_LEAF_DATA_SIZE(root);
 	for (i = 0; i < right_nritems; i++) {
-		item = btrfs_item_nr(right, i);
-
-		if (!right->map_token) {
-			map_extent_buffer(right, (unsigned long)item,
-					sizeof(struct btrfs_item),
-					&right->map_token, &right->kaddr,
-					&right->map_start, &right->map_len,
-					KM_USER1);
-		}
+		item = btrfs_item_nr(i);
 
-		push_space = push_space - btrfs_item_size(right, item);
-		btrfs_set_item_offset(right, item, push_space);
-	}
-	if (right->map_token) {
-		unmap_extent_buffer(right, right->map_token, KM_USER1);
-		right->map_token = NULL;
+		push_space = push_space - btrfs_token_item_size(right,
+								item, &token);
+		btrfs_set_token_item_offset(right, item, push_space, &token);
 	}
 
 	btrfs_mark_buffer_dirty(left);
 	if (right_nritems)
 		btrfs_mark_buffer_dirty(right);
+	else
+		clean_tree_block(trans, root, right);
 
 	btrfs_item_key(right, &disk_key, 0);
-	wret = fixup_low_keys(trans, root, path, &disk_key, 1);
-	if (wret)
-		ret = wret;
+	fixup_low_keys(root, path, &disk_key, 1);
 
 	/* then fixup the leaf pointer in the path */
 	if (path->slots[0] < push_items) {
 		path->slots[0] += old_left_nritems;
-		if (btrfs_header_nritems(path->nodes[0]) == 0)
-			clean_tree_block(trans, root, path->nodes[0]);
 		btrfs_tree_unlock(path->nodes[0]);
 		free_extent_buffer(path->nodes[0]);
 		path->nodes[0] = left;
@@ -2691,10 +3980,14 @@ out:
 /*
  * push some data in the path leaf to the left, trying to free up at
  * least data_size bytes.  returns zero if the push worked, nonzero otherwise
+ *
+ * max_slot can put a limit on how far into the leaf we'll push items.  The
+ * item at 'max_slot' won't be touched.  Use (u32)-1 to make us push all the
+ * items
  */
 static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root
-			  *root, struct btrfs_path *path, int data_size,
-			  int empty)
+			  *root, struct btrfs_path *path, int min_data_size,
+			  int data_size, int empty, u32 max_slot)
 {
 	struct extent_buffer *right = path->nodes[0];
 	struct extent_buffer *left;
@@ -2716,6 +4009,9 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root
 	btrfs_assert_tree_locked(path->nodes[1]);
 
 	left = read_node_slot(root, path->nodes[1], slot - 1);
+	if (left == NULL)
+		return 1;
+
 	btrfs_tree_lock(left);
 	btrfs_set_lock_blocking(left);
 
@@ -2730,7 +4026,8 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root
 			      path->nodes[1], slot - 1, &left);
 	if (ret) {
 		/* we hit -ENOSPC, but it isn't fatal here */
-		ret = 1;
+		if (ret == -ENOSPC)
+			ret = 1;
 		goto out;
 	}
 
@@ -2740,8 +4037,9 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root
 		goto out;
 	}
 
-	return __push_leaf_left(trans, root, path, data_size,
-			       empty, left, free_space, right_nritems);
+	return __push_leaf_left(trans, root, path, min_data_size,
+			       empty, left, free_space, right_nritems,
+			       max_slot);
 out:
 	btrfs_tree_unlock(left);
 	free_extent_buffer(left);
@@ -2751,22 +4049,21 @@ out:
 /*
  * split the path's leaf in two, making sure there is at least data_size
  * available for the resulting leaf level of the path.
- *
- * returns 0 if all went well and < 0 on failure.
  */
-static noinline int copy_for_split(struct btrfs_trans_handle *trans,
-			       struct btrfs_root *root,
-			       struct btrfs_path *path,
-			       struct extent_buffer *l,
-			       struct extent_buffer *right,
-			       int slot, int mid, int nritems)
+static noinline void copy_for_split(struct btrfs_trans_handle *trans,
+				    struct btrfs_root *root,
+				    struct btrfs_path *path,
+				    struct extent_buffer *l,
+				    struct extent_buffer *right,
+				    int slot, int mid, int nritems)
 {
 	int data_copy_size;
 	int rt_data_off;
 	int i;
-	int ret = 0;
-	int wret;
 	struct btrfs_disk_key disk_key;
+	struct btrfs_map_token token;
+
+	btrfs_init_map_token(&token);
 
 	nritems = nritems - mid;
 	btrfs_set_header_nritems(right, nritems);
@@ -2785,33 +4082,18 @@ static noinline int copy_for_split(struct btrfs_trans_handle *trans,
 		      btrfs_item_end_nr(l, mid);
 
 	for (i = 0; i < nritems; i++) {
-		struct btrfs_item *item = btrfs_item_nr(right, i);
+		struct btrfs_item *item = btrfs_item_nr(i);
 		u32 ioff;
 
-		if (!right->map_token) {
-			map_extent_buffer(right, (unsigned long)item,
-					sizeof(struct btrfs_item),
-					&right->map_token, &right->kaddr,
-					&right->map_start, &right->map_len,
-					KM_USER1);
-		}
-
-		ioff = btrfs_item_offset(right, item);
-		btrfs_set_item_offset(right, item, ioff + rt_data_off);
-	}
-
-	if (right->map_token) {
-		unmap_extent_buffer(right, right->map_token, KM_USER1);
-		right->map_token = NULL;
+		ioff = btrfs_token_item_offset(right, item, &token);
+		btrfs_set_token_item_offset(right, item,
+					    ioff + rt_data_off, &token);
 	}
 
 	btrfs_set_header_nritems(l, mid);
-	ret = 0;
 	btrfs_item_key(right, &disk_key, 0);
-	wret = insert_ptr(trans, root, path, &disk_key, right->start,
-			  path->slots[1] + 1, 1);
-	if (wret)
-		ret = wret;
+	insert_ptr(trans, root, path, &disk_key, right->start,
+		   path->slots[1] + 1, 1);
 
 	btrfs_mark_buffer_dirty(right);
 	btrfs_mark_buffer_dirty(l);
@@ -2829,8 +4111,67 @@ static noinline int copy_for_split(struct btrfs_trans_handle *trans,
 	}
 
 	BUG_ON(path->slots[0] < 0);
+}
 
-	return ret;
+/*
+ * double splits happen when we need to insert a big item in the middle
+ * of a leaf.  A double split can leave us with 3 mostly empty leaves:
+ * leaf: [ slots 0 - N] [ our target ] [ N + 1 - total in leaf ]
+ *          A                 B                 C
+ *
+ * We avoid this by trying to push the items on either side of our target
+ * into the adjacent leaves.  If all goes well we can avoid the double split
+ * completely.
+ */
+static noinline int push_for_double_split(struct btrfs_trans_handle *trans,
+					  struct btrfs_root *root,
+					  struct btrfs_path *path,
+					  int data_size)
+{
+	int ret;
+	int progress = 0;
+	int slot;
+	u32 nritems;
+	int space_needed = data_size;
+
+	slot = path->slots[0];
+	if (slot < btrfs_header_nritems(path->nodes[0]))
+		space_needed -= btrfs_leaf_free_space(root, path->nodes[0]);
+
+	/*
+	 * try to push all the items after our slot into the
+	 * right leaf
+	 */
+	ret = push_leaf_right(trans, root, path, 1, space_needed, 0, slot);
+	if (ret < 0)
+		return ret;
+
+	if (ret == 0)
+		progress++;
+
+	nritems = btrfs_header_nritems(path->nodes[0]);
+	/*
+	 * our goal is to get our slot at the start or end of a leaf.  If
+	 * we've done so we're done
+	 */
+	if (path->slots[0] == 0 || path->slots[0] == nritems)
+		return 0;
+
+	if (btrfs_leaf_free_space(root, path->nodes[0]) >= data_size)
+		return 0;
+
+	/* try to push all the items before our slot into the next leaf */
+	slot = path->slots[0];
+	ret = push_leaf_left(trans, root, path, 1, space_needed, 0, slot);
+	if (ret < 0)
+		return ret;
+
+	if (ret == 0)
+		progress++;
+
+	if (progress)
+		return 0;
+	return 1;
 }
 
 /*
@@ -2855,6 +4196,7 @@ static noinline int split_leaf(struct btrfs_trans_handle *trans,
 	int wret;
 	int split;
 	int num_doubles = 0;
+	int tried_avoid_double = 0;
 
 	l = path->nodes[0];
 	slot = path->slots[0];
@@ -2863,12 +4205,19 @@ static noinline int split_leaf(struct btrfs_trans_handle *trans,
 		return -EOVERFLOW;
 
 	/* first try to make some room by pushing left and right */
-	if (data_size && ins_key->type != BTRFS_DIR_ITEM_KEY) {
-		wret = push_leaf_right(trans, root, path, data_size, 0);
+	if (data_size && path->nodes[1]) {
+		int space_needed = data_size;
+
+		if (slot < btrfs_header_nritems(l))
+			space_needed -= btrfs_leaf_free_space(root, l);
+
+		wret = push_leaf_right(trans, root, path, space_needed,
+				       space_needed, 0, 0);
 		if (wret < 0)
 			return wret;
 		if (wret) {
-			wret = push_leaf_left(trans, root, path, data_size, 0);
+			wret = push_leaf_left(trans, root, path, space_needed,
+					      space_needed, 0, (u32)-1);
 			if (wret < 0)
 				return wret;
 		}
@@ -2902,6 +4251,8 @@ again:
 				if (mid != nritems &&
 				    leaf_space_used(l, mid, nritems - mid) +
 				    data_size > BTRFS_LEAF_DATA_SIZE(root)) {
+					if (data_size && !tried_avoid_double)
+						goto push_for_double;
 					split = 2;
 				}
 			}
@@ -2918,7 +4269,9 @@ again:
 				if (mid != nritems &&
 				    leaf_space_used(l, mid, nritems - mid) +
 				    data_size > BTRFS_LEAF_DATA_SIZE(root)) {
-					split = 2 ;
+					if (data_size && !tried_avoid_double)
+						goto push_for_double;
+					split = 2;
 				}
 			}
 		}
@@ -2932,10 +4285,10 @@ again:
 	right = btrfs_alloc_free_block(trans, root, root->leafsize, 0,
 					root->root_key.objectid,
 					&disk_key, 0, l->start, 0);
-	if (IS_ERR(right)) {
-		BUG_ON(1);
+	if (IS_ERR(right))
 		return PTR_ERR(right);
-	}
+
+	root_add_used(root, root->leafsize);
 
 	memset_extent_buffer(right, 0, 0, sizeof(struct btrfs_header));
 	btrfs_set_header_bytenr(right, right->start);
@@ -2944,22 +4297,17 @@ again:
 	btrfs_set_header_owner(right, root->root_key.objectid);
 	btrfs_set_header_level(right, 0);
 	write_extent_buffer(right, root->fs_info->fsid,
-			    (unsigned long)btrfs_header_fsid(right),
-			    BTRFS_FSID_SIZE);
+			    btrfs_header_fsid(), BTRFS_FSID_SIZE);
 
 	write_extent_buffer(right, root->fs_info->chunk_tree_uuid,
-			    (unsigned long)btrfs_header_chunk_tree_uuid(right),
+			    btrfs_header_chunk_tree_uuid(right),
 			    BTRFS_UUID_SIZE);
 
 	if (split == 0) {
 		if (mid <= slot) {
 			btrfs_set_header_nritems(right, 0);
-			wret = insert_ptr(trans, root, path,
-					  &disk_key, right->start,
-					  path->slots[1] + 1, 1);
-			if (wret)
-				ret = wret;
-
+			insert_ptr(trans, root, path, &disk_key, right->start,
+				   path->slots[1] + 1, 1);
 			btrfs_tree_unlock(path->nodes[0]);
 			free_extent_buffer(path->nodes[0]);
 			path->nodes[0] = right;
@@ -2967,29 +4315,20 @@ again:
 			path->slots[1] += 1;
 		} else {
 			btrfs_set_header_nritems(right, 0);
-			wret = insert_ptr(trans, root, path,
-					  &disk_key,
-					  right->start,
+			insert_ptr(trans, root, path, &disk_key, right->start,
 					  path->slots[1], 1);
-			if (wret)
-				ret = wret;
 			btrfs_tree_unlock(path->nodes[0]);
 			free_extent_buffer(path->nodes[0]);
 			path->nodes[0] = right;
 			path->slots[0] = 0;
-			if (path->slots[1] == 0) {
-				wret = fixup_low_keys(trans, root,
-						path, &disk_key, 1);
-				if (wret)
-					ret = wret;
-			}
+			if (path->slots[1] == 0)
+				fixup_low_keys(root, path, &disk_key, 1);
 		}
 		btrfs_mark_buffer_dirty(right);
 		return ret;
 	}
 
-	ret = copy_for_split(trans, root, path, l, right, slot, mid, nritems);
-	BUG_ON(ret);
+	copy_for_split(trans, root, path, l, right, slot, mid, nritems);
 
 	if (split == 2) {
 		BUG_ON(num_doubles != 0);
@@ -2997,7 +4336,14 @@ again:
 		goto again;
 	}
 
-	return ret;
+	return 0;
+
+push_for_double:
+	push_for_double_split(trans, root, path, data_size);
+	tried_avoid_double = 1;
+	if (btrfs_leaf_free_space(root, path->nodes[0]) >= data_size)
+		return 0;
+	goto again;
 }
 
 static noinline int setup_leaf_for_split(struct btrfs_trans_handle *trans,
@@ -3026,7 +4372,7 @@ static noinline int setup_leaf_for_split(struct btrfs_trans_handle *trans,
 				    struct btrfs_file_extent_item);
 		extent_len = btrfs_file_extent_num_bytes(leaf, fi);
 	}
-	btrfs_release_path(root, path);
+	btrfs_release_path(path);
 
 	path->keep_locks = 1;
 	path->search_for_split = 1;
@@ -3054,7 +4400,8 @@ static noinline int setup_leaf_for_split(struct btrfs_trans_handle *trans,
 
 	btrfs_set_path_blocking(path);
 	ret = split_leaf(trans, root, &key, path, ins_len, 1);
-	BUG_ON(ret);
+	if (ret)
+		goto err;
 
 	path->keep_locks = 0;
 	btrfs_unlock_up_safe(path, 1);
@@ -3085,7 +4432,7 @@ static noinline int split_item(struct btrfs_trans_handle *trans,
 
 	btrfs_set_path_blocking(path);
 
-	item = btrfs_item_nr(leaf, path->slots[0]);
+	item = btrfs_item_nr(path->slots[0]);
 	orig_offset = btrfs_item_offset(leaf, item);
 	item_size = btrfs_item_size(leaf, item);
 
@@ -3108,7 +4455,7 @@ static noinline int split_item(struct btrfs_trans_handle *trans,
 	btrfs_cpu_key_to_disk(&disk_key, new_key);
 	btrfs_set_item_key(leaf, &disk_key, slot);
 
-	new_item = btrfs_item_nr(leaf, slot);
+	new_item = btrfs_item_nr(slot);
 
 	btrfs_set_item_offset(leaf, new_item, orig_offset);
 	btrfs_set_item_size(leaf, new_item, item_size - split_offset);
@@ -3191,11 +4538,9 @@ int btrfs_duplicate_item(struct btrfs_trans_handle *trans,
 		return ret;
 
 	path->slots[0]++;
-	ret = setup_items_for_insert(trans, root, path, new_key, &item_size,
-				     item_size, item_size +
-				     sizeof(struct btrfs_item), 1);
-	BUG_ON(ret);
-
+	setup_items_for_insert(root, path, new_key, &item_size,
+			       item_size, item_size +
+			       sizeof(struct btrfs_item), 1);
 	leaf = path->nodes[0];
 	memcpy_extent_buffer(leaf,
 			     btrfs_item_ptr_offset(leaf, path->slots[0]),
@@ -3210,14 +4555,10 @@ int btrfs_duplicate_item(struct btrfs_trans_handle *trans,
  * off the end of the item or if we shift the item to chop bytes off
  * the front.
  */
-int btrfs_truncate_item(struct btrfs_trans_handle *trans,
-			struct btrfs_root *root,
-			struct btrfs_path *path,
-			u32 new_size, int from_end)
+void btrfs_truncate_item(struct btrfs_root *root, struct btrfs_path *path,
+			 u32 new_size, int from_end)
 {
-	int ret = 0;
 	int slot;
-	int slot_orig;
 	struct extent_buffer *leaf;
 	struct btrfs_item *item;
 	u32 nritems;
@@ -3226,14 +4567,16 @@ int btrfs_truncate_item(struct btrfs_trans_handle *trans,
 	unsigned int old_size;
 	unsigned int size_diff;
 	int i;
+	struct btrfs_map_token token;
+
+	btrfs_init_map_token(&token);
 
-	slot_orig = path->slots[0];
 	leaf = path->nodes[0];
 	slot = path->slots[0];
 
 	old_size = btrfs_item_size_nr(leaf, slot);
 	if (old_size == new_size)
-		return 0;
+		return;
 
 	nritems = btrfs_header_nritems(leaf);
 	data_end = leaf_data_end(root, leaf);
@@ -3251,23 +4594,11 @@ int btrfs_truncate_item(struct btrfs_trans_handle *trans,
 	/* first correct the data pointers */
 	for (i = slot; i < nritems; i++) {
 		u32 ioff;
-		item = btrfs_item_nr(leaf, i);
-
-		if (!leaf->map_token) {
-			map_extent_buffer(leaf, (unsigned long)item,
-					sizeof(struct btrfs_item),
-					&leaf->map_token, &leaf->kaddr,
-					&leaf->map_start, &leaf->map_len,
-					KM_USER1);
-		}
-
-		ioff = btrfs_item_offset(leaf, item);
-		btrfs_set_item_offset(leaf, item, ioff + size_diff);
-	}
+		item = btrfs_item_nr(i);
 
-	if (leaf->map_token) {
-		unmap_extent_buffer(leaf, leaf->map_token, KM_USER1);
-		leaf->map_token = NULL;
+		ioff = btrfs_token_item_offset(leaf, item, &token);
+		btrfs_set_token_item_offset(leaf, item,
+					    ioff + size_diff, &token);
 	}
 
 	/* shift the data */
@@ -3308,31 +4639,26 @@ int btrfs_truncate_item(struct btrfs_trans_handle *trans,
 		btrfs_set_disk_key_offset(&disk_key, offset + size_diff);
 		btrfs_set_item_key(leaf, &disk_key, slot);
 		if (slot == 0)
-			fixup_low_keys(trans, root, path, &disk_key, 1);
+			fixup_low_keys(root, path, &disk_key, 1);
 	}
 
-	item = btrfs_item_nr(leaf, slot);
+	item = btrfs_item_nr(slot);
 	btrfs_set_item_size(leaf, item, new_size);
 	btrfs_mark_buffer_dirty(leaf);
 
-	ret = 0;
 	if (btrfs_leaf_free_space(root, leaf) < 0) {
 		btrfs_print_leaf(root, leaf);
 		BUG();
 	}
-	return ret;
 }
 
 /*
- * make the item pointed to by the path bigger, data_size is the new size.
+ * make the item pointed to by the path bigger, data_size is the added size.
  */
-int btrfs_extend_item(struct btrfs_trans_handle *trans,
-		      struct btrfs_root *root, struct btrfs_path *path,
-		      u32 data_size)
+void btrfs_extend_item(struct btrfs_root *root, struct btrfs_path *path,
+		       u32 data_size)
 {
-	int ret = 0;
 	int slot;
-	int slot_orig;
 	struct extent_buffer *leaf;
 	struct btrfs_item *item;
 	u32 nritems;
@@ -3340,8 +4666,10 @@ int btrfs_extend_item(struct btrfs_trans_handle *trans,
 	unsigned int old_data;
 	unsigned int old_size;
 	int i;
+	struct btrfs_map_token token;
+
+	btrfs_init_map_token(&token);
 
-	slot_orig = path->slots[0];
 	leaf = path->nodes[0];
 
 	nritems = btrfs_header_nritems(leaf);
@@ -3357,7 +4685,7 @@ int btrfs_extend_item(struct btrfs_trans_handle *trans,
 	BUG_ON(slot < 0);
 	if (slot >= nritems) {
 		btrfs_print_leaf(root, leaf);
-		printk(KERN_CRIT "slot %d too large, nritems %d\n",
+		btrfs_crit(root->fs_info, "slot %d too large, nritems %d",
 		       slot, nritems);
 		BUG_ON(1);
 	}
@@ -3368,22 +4696,11 @@ int btrfs_extend_item(struct btrfs_trans_handle *trans,
 	/* first correct the data pointers */
 	for (i = slot; i < nritems; i++) {
 		u32 ioff;
-		item = btrfs_item_nr(leaf, i);
-
-		if (!leaf->map_token) {
-			map_extent_buffer(leaf, (unsigned long)item,
-					sizeof(struct btrfs_item),
-					&leaf->map_token, &leaf->kaddr,
-					&leaf->map_start, &leaf->map_len,
-					KM_USER1);
-		}
-		ioff = btrfs_item_offset(leaf, item);
-		btrfs_set_item_offset(leaf, item, ioff - data_size);
-	}
+		item = btrfs_item_nr(i);
 
-	if (leaf->map_token) {
-		unmap_extent_buffer(leaf, leaf->map_token, KM_USER1);
-		leaf->map_token = NULL;
+		ioff = btrfs_token_item_offset(leaf, item, &token);
+		btrfs_set_token_item_offset(leaf, item,
+					    ioff - data_size, &token);
 	}
 
 	/* shift the data */
@@ -3393,168 +4710,14 @@ int btrfs_extend_item(struct btrfs_trans_handle *trans,
 
 	data_end = old_data;
 	old_size = btrfs_item_size_nr(leaf, slot);
-	item = btrfs_item_nr(leaf, slot);
+	item = btrfs_item_nr(slot);
 	btrfs_set_item_size(leaf, item, old_size + data_size);
 	btrfs_mark_buffer_dirty(leaf);
 
-	ret = 0;
 	if (btrfs_leaf_free_space(root, leaf) < 0) {
 		btrfs_print_leaf(root, leaf);
 		BUG();
 	}
-	return ret;
-}
-
-/*
- * Given a key and some data, insert items into the tree.
- * This does all the path init required, making room in the tree if needed.
- * Returns the number of keys that were inserted.
- */
-int btrfs_insert_some_items(struct btrfs_trans_handle *trans,
-			    struct btrfs_root *root,
-			    struct btrfs_path *path,
-			    struct btrfs_key *cpu_key, u32 *data_size,
-			    int nr)
-{
-	struct extent_buffer *leaf;
-	struct btrfs_item *item;
-	int ret = 0;
-	int slot;
-	int i;
-	u32 nritems;
-	u32 total_data = 0;
-	u32 total_size = 0;
-	unsigned int data_end;
-	struct btrfs_disk_key disk_key;
-	struct btrfs_key found_key;
-
-	for (i = 0; i < nr; i++) {
-		if (total_size + data_size[i] + sizeof(struct btrfs_item) >
-		    BTRFS_LEAF_DATA_SIZE(root)) {
-			break;
-			nr = i;
-		}
-		total_data += data_size[i];
-		total_size += data_size[i] + sizeof(struct btrfs_item);
-	}
-	BUG_ON(nr == 0);
-
-	ret = btrfs_search_slot(trans, root, cpu_key, path, total_size, 1);
-	if (ret == 0)
-		return -EEXIST;
-	if (ret < 0)
-		goto out;
-
-	leaf = path->nodes[0];
-
-	nritems = btrfs_header_nritems(leaf);
-	data_end = leaf_data_end(root, leaf);
-
-	if (btrfs_leaf_free_space(root, leaf) < total_size) {
-		for (i = nr; i >= 0; i--) {
-			total_data -= data_size[i];
-			total_size -= data_size[i] + sizeof(struct btrfs_item);
-			if (total_size < btrfs_leaf_free_space(root, leaf))
-				break;
-		}
-		nr = i;
-	}
-
-	slot = path->slots[0];
-	BUG_ON(slot < 0);
-
-	if (slot != nritems) {
-		unsigned int old_data = btrfs_item_end_nr(leaf, slot);
-
-		item = btrfs_item_nr(leaf, slot);
-		btrfs_item_key_to_cpu(leaf, &found_key, slot);
-
-		/* figure out how many keys we can insert in here */
-		total_data = data_size[0];
-		for (i = 1; i < nr; i++) {
-			if (btrfs_comp_cpu_keys(&found_key, cpu_key + i) <= 0)
-				break;
-			total_data += data_size[i];
-		}
-		nr = i;
-
-		if (old_data < data_end) {
-			btrfs_print_leaf(root, leaf);
-			printk(KERN_CRIT "slot %d old_data %d data_end %d\n",
-			       slot, old_data, data_end);
-			BUG_ON(1);
-		}
-		/*
-		 * item0..itemN ... dataN.offset..dataN.size .. data0.size
-		 */
-		/* first correct the data pointers */
-		WARN_ON(leaf->map_token);
-		for (i = slot; i < nritems; i++) {
-			u32 ioff;
-
-			item = btrfs_item_nr(leaf, i);
-			if (!leaf->map_token) {
-				map_extent_buffer(leaf, (unsigned long)item,
-					sizeof(struct btrfs_item),
-					&leaf->map_token, &leaf->kaddr,
-					&leaf->map_start, &leaf->map_len,
-					KM_USER1);
-			}
-
-			ioff = btrfs_item_offset(leaf, item);
-			btrfs_set_item_offset(leaf, item, ioff - total_data);
-		}
-		if (leaf->map_token) {
-			unmap_extent_buffer(leaf, leaf->map_token, KM_USER1);
-			leaf->map_token = NULL;
-		}
-
-		/* shift the items */
-		memmove_extent_buffer(leaf, btrfs_item_nr_offset(slot + nr),
-			      btrfs_item_nr_offset(slot),
-			      (nritems - slot) * sizeof(struct btrfs_item));
-
-		/* shift the data */
-		memmove_extent_buffer(leaf, btrfs_leaf_data(leaf) +
-			      data_end - total_data, btrfs_leaf_data(leaf) +
-			      data_end, old_data - data_end);
-		data_end = old_data;
-	} else {
-		/*
-		 * this sucks but it has to be done, if we are inserting at
-		 * the end of the leaf only insert 1 of the items, since we
-		 * have no way of knowing whats on the next leaf and we'd have
-		 * to drop our current locks to figure it out
-		 */
-		nr = 1;
-	}
-
-	/* setup the item for the new data */
-	for (i = 0; i < nr; i++) {
-		btrfs_cpu_key_to_disk(&disk_key, cpu_key + i);
-		btrfs_set_item_key(leaf, &disk_key, slot + i);
-		item = btrfs_item_nr(leaf, slot + i);
-		btrfs_set_item_offset(leaf, item, data_end - data_size[i]);
-		data_end -= data_size[i];
-		btrfs_set_item_size(leaf, item, data_size[i]);
-	}
-	btrfs_set_header_nritems(leaf, nritems + nr);
-	btrfs_mark_buffer_dirty(leaf);
-
-	ret = 0;
-	if (slot == 0) {
-		btrfs_cpu_key_to_disk(&disk_key, cpu_key);
-		ret = fixup_low_keys(trans, root, path, &disk_key, 1);
-	}
-
-	if (btrfs_leaf_free_space(root, leaf) < 0) {
-		btrfs_print_leaf(root, leaf);
-		BUG();
-	}
-out:
-	if (!ret)
-		ret = nr;
-	return ret;
 }
 
 /*
@@ -3562,20 +4725,20 @@ out:
  * to save stack depth by doing the bulk of the work in a function
  * that doesn't call btrfs_search_slot
  */
-static noinline_for_stack int
-setup_items_for_insert(struct btrfs_trans_handle *trans,
-		      struct btrfs_root *root, struct btrfs_path *path,
-		      struct btrfs_key *cpu_key, u32 *data_size,
-		      u32 total_data, u32 total_size, int nr)
+void setup_items_for_insert(struct btrfs_root *root, struct btrfs_path *path,
+			    struct btrfs_key *cpu_key, u32 *data_size,
+			    u32 total_data, u32 total_size, int nr)
 {
 	struct btrfs_item *item;
 	int i;
 	u32 nritems;
 	unsigned int data_end;
 	struct btrfs_disk_key disk_key;
-	int ret;
 	struct extent_buffer *leaf;
 	int slot;
+	struct btrfs_map_token token;
+
+	btrfs_init_map_token(&token);
 
 	leaf = path->nodes[0];
 	slot = path->slots[0];
@@ -3585,7 +4748,7 @@ setup_items_for_insert(struct btrfs_trans_handle *trans,
 
 	if (btrfs_leaf_free_space(root, leaf) < total_size) {
 		btrfs_print_leaf(root, leaf);
-		printk(KERN_CRIT "not enough freespace need %u have %d\n",
+		btrfs_crit(root->fs_info, "not enough freespace need %u have %d",
 		       total_size, btrfs_leaf_free_space(root, leaf));
 		BUG();
 	}
@@ -3595,7 +4758,7 @@ setup_items_for_insert(struct btrfs_trans_handle *trans,
 
 		if (old_data < data_end) {
 			btrfs_print_leaf(root, leaf);
-			printk(KERN_CRIT "slot %d old_data %d data_end %d\n",
+			btrfs_crit(root->fs_info, "slot %d old_data %d data_end %d",
 			       slot, old_data, data_end);
 			BUG_ON(1);
 		}
@@ -3603,27 +4766,14 @@ setup_items_for_insert(struct btrfs_trans_handle *trans,
 		 * item0..itemN ... dataN.offset..dataN.size .. data0.size
 		 */
 		/* first correct the data pointers */
-		WARN_ON(leaf->map_token);
 		for (i = slot; i < nritems; i++) {
 			u32 ioff;
 
-			item = btrfs_item_nr(leaf, i);
-			if (!leaf->map_token) {
-				map_extent_buffer(leaf, (unsigned long)item,
-					sizeof(struct btrfs_item),
-					&leaf->map_token, &leaf->kaddr,
-					&leaf->map_start, &leaf->map_len,
-					KM_USER1);
-			}
-
-			ioff = btrfs_item_offset(leaf, item);
-			btrfs_set_item_offset(leaf, item, ioff - total_data);
+			item = btrfs_item_nr( i);
+			ioff = btrfs_token_item_offset(leaf, item, &token);
+			btrfs_set_token_item_offset(leaf, item,
+						    ioff - total_data, &token);
 		}
-		if (leaf->map_token) {
-			unmap_extent_buffer(leaf, leaf->map_token, KM_USER1);
-			leaf->map_token = NULL;
-		}
-
 		/* shift the items */
 		memmove_extent_buffer(leaf, btrfs_item_nr_offset(slot + nr),
 			      btrfs_item_nr_offset(slot),
@@ -3640,19 +4790,18 @@ setup_items_for_insert(struct btrfs_trans_handle *trans,
 	for (i = 0; i < nr; i++) {
 		btrfs_cpu_key_to_disk(&disk_key, cpu_key + i);
 		btrfs_set_item_key(leaf, &disk_key, slot + i);
-		item = btrfs_item_nr(leaf, slot + i);
-		btrfs_set_item_offset(leaf, item, data_end - data_size[i]);
+		item = btrfs_item_nr(slot + i);
+		btrfs_set_token_item_offset(leaf, item,
+					    data_end - data_size[i], &token);
 		data_end -= data_size[i];
-		btrfs_set_item_size(leaf, item, data_size[i]);
+		btrfs_set_token_item_size(leaf, item, data_size[i], &token);
 	}
 
 	btrfs_set_header_nritems(leaf, nritems + nr);
 
-	ret = 0;
 	if (slot == 0) {
-		struct btrfs_disk_key disk_key;
 		btrfs_cpu_key_to_disk(&disk_key, cpu_key);
-		ret = fixup_low_keys(trans, root, path, &disk_key, 1);
+		fixup_low_keys(root, path, &disk_key, 1);
 	}
 	btrfs_unlock_up_safe(path, 1);
 	btrfs_mark_buffer_dirty(leaf);
@@ -3661,7 +4810,6 @@ setup_items_for_insert(struct btrfs_trans_handle *trans,
 		btrfs_print_leaf(root, leaf);
 		BUG();
 	}
-	return ret;
 }
 
 /*
@@ -3674,7 +4822,6 @@ int btrfs_insert_empty_items(struct btrfs_trans_handle *trans,
 			    struct btrfs_key *cpu_key, u32 *data_size,
 			    int nr)
 {
-	struct extent_buffer *leaf;
 	int ret = 0;
 	int slot;
 	int i;
@@ -3689,17 +4836,14 @@ int btrfs_insert_empty_items(struct btrfs_trans_handle *trans,
 	if (ret == 0)
 		return -EEXIST;
 	if (ret < 0)
-		goto out;
+		return ret;
 
-	leaf = path->nodes[0];
 	slot = path->slots[0];
 	BUG_ON(slot < 0);
 
-	ret = setup_items_for_insert(trans, root, path, cpu_key, data_size,
+	setup_items_for_insert(root, path, cpu_key, data_size,
 			       total_data, total_size, nr);
-
-out:
-	return ret;
+	return 0;
 }
 
 /*
@@ -3716,7 +4860,8 @@ int btrfs_insert_item(struct btrfs_trans_handle *trans, struct btrfs_root
 	unsigned long ptr;
 
 	path = btrfs_alloc_path();
-	BUG_ON(!path);
+	if (!path)
+		return -ENOMEM;
 	ret = btrfs_insert_empty_item(trans, root, path, cpu_key, data_size);
 	if (!ret) {
 		leaf = path->nodes[0];
@@ -3734,22 +4879,29 @@ int btrfs_insert_item(struct btrfs_trans_handle *trans, struct btrfs_root
  * the tree should have been previously balanced so the deletion does not
  * empty a node.
  */
-static int del_ptr(struct btrfs_trans_handle *trans, struct btrfs_root *root,
-		   struct btrfs_path *path, int level, int slot)
+static void del_ptr(struct btrfs_root *root, struct btrfs_path *path,
+		    int level, int slot)
 {
 	struct extent_buffer *parent = path->nodes[level];
 	u32 nritems;
-	int ret = 0;
-	int wret;
+	int ret;
 
 	nritems = btrfs_header_nritems(parent);
 	if (slot != nritems - 1) {
+		if (level)
+			tree_mod_log_eb_move(root->fs_info, parent, slot,
+					     slot + 1, nritems - slot - 1);
 		memmove_extent_buffer(parent,
 			      btrfs_node_key_ptr_offset(slot),
 			      btrfs_node_key_ptr_offset(slot + 1),
 			      sizeof(struct btrfs_key_ptr) *
 			      (nritems - slot - 1));
+	} else if (level) {
+		ret = tree_mod_log_insert_key(root->fs_info, parent, slot,
+					      MOD_LOG_KEY_REMOVE, GFP_NOFS);
+		BUG_ON(ret < 0);
 	}
+
 	nritems--;
 	btrfs_set_header_nritems(parent, nritems);
 	if (nritems == 0 && parent == root->node) {
@@ -3760,12 +4912,9 @@ static int del_ptr(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 		struct btrfs_disk_key disk_key;
 
 		btrfs_node_key(parent, &disk_key, 0);
-		wret = fixup_low_keys(trans, root, path, &disk_key, level + 1);
-		if (wret)
-			ret = wret;
+		fixup_low_keys(root, path, &disk_key, level + 1);
 	}
 	btrfs_mark_buffer_dirty(parent);
-	return ret;
 }
 
 /*
@@ -3778,17 +4927,13 @@ static int del_ptr(struct btrfs_trans_handle *trans, struct btrfs_root *root,
  * The path must have already been setup for deleting the leaf, including
  * all the proper balancing.  path->nodes[1] must be locked.
  */
-static noinline int btrfs_del_leaf(struct btrfs_trans_handle *trans,
-				   struct btrfs_root *root,
-				   struct btrfs_path *path,
-				   struct extent_buffer *leaf)
+static noinline void btrfs_del_leaf(struct btrfs_trans_handle *trans,
+				    struct btrfs_root *root,
+				    struct btrfs_path *path,
+				    struct extent_buffer *leaf)
 {
-	int ret;
-
 	WARN_ON(btrfs_header_generation(leaf) != trans->transid);
-	ret = del_ptr(trans, root, path, 1, path->slots[1]);
-	if (ret)
-		return ret;
+	del_ptr(root, path, 1, path->slots[1]);
 
 	/*
 	 * btrfs_free_extent is expensive, we want to make sure we
@@ -3796,9 +4941,11 @@ static noinline int btrfs_del_leaf(struct btrfs_trans_handle *trans,
 	 */
 	btrfs_unlock_up_safe(path, 0);
 
-	ret = btrfs_free_tree_block(trans, root, leaf->start, leaf->len,
-				    0, root->root_key.objectid, 0);
-	return ret;
+	root_sub_used(root, leaf->len);
+
+	extent_buffer_get(leaf);
+	btrfs_free_tree_block(trans, root, leaf, 0, 1);
+	free_extent_buffer_stale(leaf);
 }
 /*
  * delete the item at the leaf level in path.  If that empties
@@ -3815,6 +4962,9 @@ int btrfs_del_items(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 	int wret;
 	int i;
 	u32 nritems;
+	struct btrfs_map_token token;
+
+	btrfs_init_map_token(&token);
 
 	leaf = path->nodes[0];
 	last_off = btrfs_item_offset_nr(leaf, slot + nr - 1);
@@ -3835,21 +4985,10 @@ int btrfs_del_items(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 		for (i = slot + nr; i < nritems; i++) {
 			u32 ioff;
 
-			item = btrfs_item_nr(leaf, i);
-			if (!leaf->map_token) {
-				map_extent_buffer(leaf, (unsigned long)item,
-					sizeof(struct btrfs_item),
-					&leaf->map_token, &leaf->kaddr,
-					&leaf->map_start, &leaf->map_len,
-					KM_USER1);
-			}
-			ioff = btrfs_item_offset(leaf, item);
-			btrfs_set_item_offset(leaf, item, ioff + dsize);
-		}
-
-		if (leaf->map_token) {
-			unmap_extent_buffer(leaf, leaf->map_token, KM_USER1);
-			leaf->map_token = NULL;
+			item = btrfs_item_nr(i);
+			ioff = btrfs_token_item_offset(leaf, item, &token);
+			btrfs_set_token_item_offset(leaf, item,
+						    ioff + dsize, &token);
 		}
 
 		memmove_extent_buffer(leaf, btrfs_item_nr_offset(slot),
@@ -3865,8 +5004,9 @@ int btrfs_del_items(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 		if (leaf == root->node) {
 			btrfs_set_header_level(leaf, 0);
 		} else {
-			ret = btrfs_del_leaf(trans, root, path, leaf);
-			BUG_ON(ret);
+			btrfs_set_path_blocking(path);
+			clean_tree_block(trans, root, leaf);
+			btrfs_del_leaf(trans, root, path, leaf);
 		}
 	} else {
 		int used = leaf_space_used(leaf, 0, nritems);
@@ -3874,10 +5014,7 @@ int btrfs_del_items(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 			struct btrfs_disk_key disk_key;
 
 			btrfs_item_key(leaf, &disk_key, 0);
-			wret = fixup_low_keys(trans, root, path,
-					      &disk_key, 1);
-			if (wret)
-				ret = wret;
+			fixup_low_keys(root, path, &disk_key, 1);
 		}
 
 		/* delete the leaf if it is mostly empty */
@@ -3890,22 +5027,24 @@ int btrfs_del_items(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 			extent_buffer_get(leaf);
 
 			btrfs_set_path_blocking(path);
-			wret = push_leaf_left(trans, root, path, 1, 1);
+			wret = push_leaf_left(trans, root, path, 1, 1,
+					      1, (u32)-1);
 			if (wret < 0 && wret != -ENOSPC)
 				ret = wret;
 
 			if (path->nodes[0] == leaf &&
 			    btrfs_header_nritems(leaf)) {
-				wret = push_leaf_right(trans, root, path, 1, 1);
+				wret = push_leaf_right(trans, root, path, 1,
+						       1, 1, 0);
 				if (wret < 0 && wret != -ENOSPC)
 					ret = wret;
 			}
 
 			if (btrfs_header_nritems(leaf) == 0) {
 				path->slots[1] = slot;
-				ret = btrfs_del_leaf(trans, root, path, leaf);
-				BUG_ON(ret);
+				btrfs_del_leaf(trans, root, path, leaf);
 				free_extent_buffer(leaf);
+				ret = 0;
 			} else {
 				/* if we're still in the path, make sure
 				 * we're dirty.  Otherwise, one of the
@@ -3939,30 +5078,44 @@ int btrfs_prev_leaf(struct btrfs_root *root, struct btrfs_path *path)
 
 	btrfs_item_key_to_cpu(path->nodes[0], &key, 0);
 
-	if (key.offset > 0)
+	if (key.offset > 0) {
 		key.offset--;
-	else if (key.type > 0)
+	} else if (key.type > 0) {
 		key.type--;
-	else if (key.objectid > 0)
+		key.offset = (u64)-1;
+	} else if (key.objectid > 0) {
 		key.objectid--;
-	else
+		key.type = (u8)-1;
+		key.offset = (u64)-1;
+	} else {
 		return 1;
+	}
 
-	btrfs_release_path(root, path);
+	btrfs_release_path(path);
 	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
 	if (ret < 0)
 		return ret;
 	btrfs_item_key(path->nodes[0], &found_key, 0);
 	ret = comp_keys(&found_key, &key);
-	if (ret < 0)
+	/*
+	 * We might have had an item with the previous key in the tree right
+	 * before we released our path. And after we released our path, that
+	 * item might have been pushed to the first slot (0) of the leaf we
+	 * were holding due to a tree balance. Alternatively, an item with the
+	 * previous key can exist as the only element of a leaf (big fat item).
+	 * Therefore account for these 2 cases, so that our callers (like
+	 * btrfs_previous_item) don't miss an existing item with a key matching
+	 * the previous key we computed above.
+	 */
+	if (ret <= 0)
 		return 0;
 	return 1;
 }
 
 /*
  * A helper function to walk down the tree starting at min_key, and looking
- * for nodes or leaves that are either in cache or have a minimum
- * transaction id.  This is used by the btree defrag code, and tree logging
+ * for nodes or leaves that are have a minimum transaction id.
+ * This is used by the btree defrag code, and tree logging
  *
  * This does not cow, but it does stuff the starting key it finds back
  * into min_key, so you can call btrfs_search_slot with cow=1 on the
@@ -3982,8 +5135,7 @@ int btrfs_prev_leaf(struct btrfs_root *root, struct btrfs_path *path)
  * was nothing in the tree that matched the search criteria.
  */
 int btrfs_search_forward(struct btrfs_root *root, struct btrfs_key *min_key,
-			 struct btrfs_key *max_key,
-			 struct btrfs_path *path, int cache_only,
+			 struct btrfs_path *path,
 			 u64 min_trans)
 {
 	struct extent_buffer *cur;
@@ -3996,11 +5148,11 @@ int btrfs_search_forward(struct btrfs_root *root, struct btrfs_key *min_key,
 
 	WARN_ON(!path->keep_locks);
 again:
-	cur = btrfs_lock_root_node(root);
+	cur = btrfs_read_lock_root_node(root);
 	level = btrfs_header_level(cur);
 	WARN_ON(path->nodes[level]);
 	path->nodes[level] = cur;
-	path->locks[level] = 1;
+	path->locks[level] = BTRFS_READ_LOCK;
 
 	if (btrfs_header_generation(cur) < min_trans) {
 		ret = 1;
@@ -4023,43 +5175,18 @@ again:
 		if (sret && slot > 0)
 			slot--;
 		/*
-		 * check this node pointer against the cache_only and
-		 * min_trans parameters.  If it isn't in cache or is too
-		 * old, skip to the next one.
+		 * check this node pointer against the min_trans parameters.
+		 * If it is too old, old, skip to the next one.
 		 */
 		while (slot < nritems) {
-			u64 blockptr;
 			u64 gen;
-			struct extent_buffer *tmp;
-			struct btrfs_disk_key disk_key;
 
-			blockptr = btrfs_node_blockptr(cur, slot);
 			gen = btrfs_node_ptr_generation(cur, slot);
 			if (gen < min_trans) {
 				slot++;
 				continue;
 			}
-			if (!cache_only)
-				break;
-
-			if (max_key) {
-				btrfs_node_key(cur, &disk_key, slot);
-				if (comp_keys(&disk_key, max_key) >= 0) {
-					ret = 1;
-					goto out;
-				}
-			}
-
-			tmp = btrfs_find_tree_block(root, blockptr,
-					    btrfs_level_size(root, level - 1));
-
-			if (tmp && btrfs_buffer_uptodate(tmp, gen)) {
-				free_extent_buffer(tmp);
-				break;
-			}
-			if (tmp)
-				free_extent_buffer(tmp);
-			slot++;
+			break;
 		}
 find_next_key:
 		/*
@@ -4070,9 +5197,9 @@ find_next_key:
 			path->slots[level] = slot;
 			btrfs_set_path_blocking(path);
 			sret = btrfs_find_next_key(root, path, min_key, level,
-						  cache_only, min_trans);
+						  min_trans);
 			if (sret == 0) {
-				btrfs_release_path(root, path);
+				btrfs_release_path(path);
 				goto again;
 			} else {
 				goto out;
@@ -4083,18 +5210,19 @@ find_next_key:
 		path->slots[level] = slot;
 		if (level == path->lowest_level) {
 			ret = 0;
-			unlock_up(path, level, 1);
+			unlock_up(path, level, 1, 0, NULL);
 			goto out;
 		}
 		btrfs_set_path_blocking(path);
 		cur = read_node_slot(root, cur, slot);
+		BUG_ON(!cur); /* -ENOMEM */
 
-		btrfs_tree_lock(cur);
+		btrfs_tree_read_lock(cur);
 
-		path->locks[level - 1] = 1;
+		path->locks[level - 1] = BTRFS_READ_LOCK;
 		path->nodes[level - 1] = cur;
-		unlock_up(path, level, 1);
-		btrfs_clear_path_blocking(path, NULL);
+		unlock_up(path, level, 1, 0, NULL);
+		btrfs_clear_path_blocking(path, NULL, 0);
 	}
 out:
 	if (ret == 0)
@@ -4103,11 +5231,362 @@ out:
 	return ret;
 }
 
+static void tree_move_down(struct btrfs_root *root,
+			   struct btrfs_path *path,
+			   int *level, int root_level)
+{
+	BUG_ON(*level == 0);
+	path->nodes[*level - 1] = read_node_slot(root, path->nodes[*level],
+					path->slots[*level]);
+	path->slots[*level - 1] = 0;
+	(*level)--;
+}
+
+static int tree_move_next_or_upnext(struct btrfs_root *root,
+				    struct btrfs_path *path,
+				    int *level, int root_level)
+{
+	int ret = 0;
+	int nritems;
+	nritems = btrfs_header_nritems(path->nodes[*level]);
+
+	path->slots[*level]++;
+
+	while (path->slots[*level] >= nritems) {
+		if (*level == root_level)
+			return -1;
+
+		/* move upnext */
+		path->slots[*level] = 0;
+		free_extent_buffer(path->nodes[*level]);
+		path->nodes[*level] = NULL;
+		(*level)++;
+		path->slots[*level]++;
+
+		nritems = btrfs_header_nritems(path->nodes[*level]);
+		ret = 1;
+	}
+	return ret;
+}
+
+/*
+ * Returns 1 if it had to move up and next. 0 is returned if it moved only next
+ * or down.
+ */
+static int tree_advance(struct btrfs_root *root,
+			struct btrfs_path *path,
+			int *level, int root_level,
+			int allow_down,
+			struct btrfs_key *key)
+{
+	int ret;
+
+	if (*level == 0 || !allow_down) {
+		ret = tree_move_next_or_upnext(root, path, level, root_level);
+	} else {
+		tree_move_down(root, path, level, root_level);
+		ret = 0;
+	}
+	if (ret >= 0) {
+		if (*level == 0)
+			btrfs_item_key_to_cpu(path->nodes[*level], key,
+					path->slots[*level]);
+		else
+			btrfs_node_key_to_cpu(path->nodes[*level], key,
+					path->slots[*level]);
+	}
+	return ret;
+}
+
+static int tree_compare_item(struct btrfs_root *left_root,
+			     struct btrfs_path *left_path,
+			     struct btrfs_path *right_path,
+			     char *tmp_buf)
+{
+	int cmp;
+	int len1, len2;
+	unsigned long off1, off2;
+
+	len1 = btrfs_item_size_nr(left_path->nodes[0], left_path->slots[0]);
+	len2 = btrfs_item_size_nr(right_path->nodes[0], right_path->slots[0]);
+	if (len1 != len2)
+		return 1;
+
+	off1 = btrfs_item_ptr_offset(left_path->nodes[0], left_path->slots[0]);
+	off2 = btrfs_item_ptr_offset(right_path->nodes[0],
+				right_path->slots[0]);
+
+	read_extent_buffer(left_path->nodes[0], tmp_buf, off1, len1);
+
+	cmp = memcmp_extent_buffer(right_path->nodes[0], tmp_buf, off2, len1);
+	if (cmp)
+		return 1;
+	return 0;
+}
+
+#define ADVANCE 1
+#define ADVANCE_ONLY_NEXT -1
+
+/*
+ * This function compares two trees and calls the provided callback for
+ * every changed/new/deleted item it finds.
+ * If shared tree blocks are encountered, whole subtrees are skipped, making
+ * the compare pretty fast on snapshotted subvolumes.
+ *
+ * This currently works on commit roots only. As commit roots are read only,
+ * we don't do any locking. The commit roots are protected with transactions.
+ * Transactions are ended and rejoined when a commit is tried in between.
+ *
+ * This function checks for modifications done to the trees while comparing.
+ * If it detects a change, it aborts immediately.
+ */
+int btrfs_compare_trees(struct btrfs_root *left_root,
+			struct btrfs_root *right_root,
+			btrfs_changed_cb_t changed_cb, void *ctx)
+{
+	int ret;
+	int cmp;
+	struct btrfs_path *left_path = NULL;
+	struct btrfs_path *right_path = NULL;
+	struct btrfs_key left_key;
+	struct btrfs_key right_key;
+	char *tmp_buf = NULL;
+	int left_root_level;
+	int right_root_level;
+	int left_level;
+	int right_level;
+	int left_end_reached;
+	int right_end_reached;
+	int advance_left;
+	int advance_right;
+	u64 left_blockptr;
+	u64 right_blockptr;
+	u64 left_gen;
+	u64 right_gen;
+
+	left_path = btrfs_alloc_path();
+	if (!left_path) {
+		ret = -ENOMEM;
+		goto out;
+	}
+	right_path = btrfs_alloc_path();
+	if (!right_path) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	tmp_buf = kmalloc(left_root->leafsize, GFP_NOFS);
+	if (!tmp_buf) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	left_path->search_commit_root = 1;
+	left_path->skip_locking = 1;
+	right_path->search_commit_root = 1;
+	right_path->skip_locking = 1;
+
+	/*
+	 * Strategy: Go to the first items of both trees. Then do
+	 *
+	 * If both trees are at level 0
+	 *   Compare keys of current items
+	 *     If left < right treat left item as new, advance left tree
+	 *       and repeat
+	 *     If left > right treat right item as deleted, advance right tree
+	 *       and repeat
+	 *     If left == right do deep compare of items, treat as changed if
+	 *       needed, advance both trees and repeat
+	 * If both trees are at the same level but not at level 0
+	 *   Compare keys of current nodes/leafs
+	 *     If left < right advance left tree and repeat
+	 *     If left > right advance right tree and repeat
+	 *     If left == right compare blockptrs of the next nodes/leafs
+	 *       If they match advance both trees but stay at the same level
+	 *         and repeat
+	 *       If they don't match advance both trees while allowing to go
+	 *         deeper and repeat
+	 * If tree levels are different
+	 *   Advance the tree that needs it and repeat
+	 *
+	 * Advancing a tree means:
+	 *   If we are at level 0, try to go to the next slot. If that's not
+	 *   possible, go one level up and repeat. Stop when we found a level
+	 *   where we could go to the next slot. We may at this point be on a
+	 *   node or a leaf.
+	 *
+	 *   If we are not at level 0 and not on shared tree blocks, go one
+	 *   level deeper.
+	 *
+	 *   If we are not at level 0 and on shared tree blocks, go one slot to
+	 *   the right if possible or go up and right.
+	 */
+
+	down_read(&left_root->fs_info->commit_root_sem);
+	left_level = btrfs_header_level(left_root->commit_root);
+	left_root_level = left_level;
+	left_path->nodes[left_level] = left_root->commit_root;
+	extent_buffer_get(left_path->nodes[left_level]);
+
+	right_level = btrfs_header_level(right_root->commit_root);
+	right_root_level = right_level;
+	right_path->nodes[right_level] = right_root->commit_root;
+	extent_buffer_get(right_path->nodes[right_level]);
+	up_read(&left_root->fs_info->commit_root_sem);
+
+	if (left_level == 0)
+		btrfs_item_key_to_cpu(left_path->nodes[left_level],
+				&left_key, left_path->slots[left_level]);
+	else
+		btrfs_node_key_to_cpu(left_path->nodes[left_level],
+				&left_key, left_path->slots[left_level]);
+	if (right_level == 0)
+		btrfs_item_key_to_cpu(right_path->nodes[right_level],
+				&right_key, right_path->slots[right_level]);
+	else
+		btrfs_node_key_to_cpu(right_path->nodes[right_level],
+				&right_key, right_path->slots[right_level]);
+
+	left_end_reached = right_end_reached = 0;
+	advance_left = advance_right = 0;
+
+	while (1) {
+		if (advance_left && !left_end_reached) {
+			ret = tree_advance(left_root, left_path, &left_level,
+					left_root_level,
+					advance_left != ADVANCE_ONLY_NEXT,
+					&left_key);
+			if (ret < 0)
+				left_end_reached = ADVANCE;
+			advance_left = 0;
+		}
+		if (advance_right && !right_end_reached) {
+			ret = tree_advance(right_root, right_path, &right_level,
+					right_root_level,
+					advance_right != ADVANCE_ONLY_NEXT,
+					&right_key);
+			if (ret < 0)
+				right_end_reached = ADVANCE;
+			advance_right = 0;
+		}
+
+		if (left_end_reached && right_end_reached) {
+			ret = 0;
+			goto out;
+		} else if (left_end_reached) {
+			if (right_level == 0) {
+				ret = changed_cb(left_root, right_root,
+						left_path, right_path,
+						&right_key,
+						BTRFS_COMPARE_TREE_DELETED,
+						ctx);
+				if (ret < 0)
+					goto out;
+			}
+			advance_right = ADVANCE;
+			continue;
+		} else if (right_end_reached) {
+			if (left_level == 0) {
+				ret = changed_cb(left_root, right_root,
+						left_path, right_path,
+						&left_key,
+						BTRFS_COMPARE_TREE_NEW,
+						ctx);
+				if (ret < 0)
+					goto out;
+			}
+			advance_left = ADVANCE;
+			continue;
+		}
+
+		if (left_level == 0 && right_level == 0) {
+			cmp = btrfs_comp_cpu_keys(&left_key, &right_key);
+			if (cmp < 0) {
+				ret = changed_cb(left_root, right_root,
+						left_path, right_path,
+						&left_key,
+						BTRFS_COMPARE_TREE_NEW,
+						ctx);
+				if (ret < 0)
+					goto out;
+				advance_left = ADVANCE;
+			} else if (cmp > 0) {
+				ret = changed_cb(left_root, right_root,
+						left_path, right_path,
+						&right_key,
+						BTRFS_COMPARE_TREE_DELETED,
+						ctx);
+				if (ret < 0)
+					goto out;
+				advance_right = ADVANCE;
+			} else {
+				enum btrfs_compare_tree_result cmp;
+
+				WARN_ON(!extent_buffer_uptodate(left_path->nodes[0]));
+				ret = tree_compare_item(left_root, left_path,
+						right_path, tmp_buf);
+				if (ret)
+					cmp = BTRFS_COMPARE_TREE_CHANGED;
+				else
+					cmp = BTRFS_COMPARE_TREE_SAME;
+				ret = changed_cb(left_root, right_root,
+						 left_path, right_path,
+						 &left_key, cmp, ctx);
+				if (ret < 0)
+					goto out;
+				advance_left = ADVANCE;
+				advance_right = ADVANCE;
+			}
+		} else if (left_level == right_level) {
+			cmp = btrfs_comp_cpu_keys(&left_key, &right_key);
+			if (cmp < 0) {
+				advance_left = ADVANCE;
+			} else if (cmp > 0) {
+				advance_right = ADVANCE;
+			} else {
+				left_blockptr = btrfs_node_blockptr(
+						left_path->nodes[left_level],
+						left_path->slots[left_level]);
+				right_blockptr = btrfs_node_blockptr(
+						right_path->nodes[right_level],
+						right_path->slots[right_level]);
+				left_gen = btrfs_node_ptr_generation(
+						left_path->nodes[left_level],
+						left_path->slots[left_level]);
+				right_gen = btrfs_node_ptr_generation(
+						right_path->nodes[right_level],
+						right_path->slots[right_level]);
+				if (left_blockptr == right_blockptr &&
+				    left_gen == right_gen) {
+					/*
+					 * As we're on a shared block, don't
+					 * allow to go deeper.
+					 */
+					advance_left = ADVANCE_ONLY_NEXT;
+					advance_right = ADVANCE_ONLY_NEXT;
+				} else {
+					advance_left = ADVANCE;
+					advance_right = ADVANCE;
+				}
+			}
+		} else if (left_level < right_level) {
+			advance_right = ADVANCE;
+		} else {
+			advance_left = ADVANCE;
+		}
+	}
+
+out:
+	btrfs_free_path(left_path);
+	btrfs_free_path(right_path);
+	kfree(tmp_buf);
+	return ret;
+}
+
 /*
  * this is similar to btrfs_next_leaf, but does not try to preserve
  * and fixup the path.  It looks for and returns the next key in the
- * tree based on the current path and the cache_only and min_trans
- * parameters.
+ * tree based on the current path and the min_trans parameters.
  *
  * 0 is returned if another key is found, < 0 if there are any errors
  * and 1 is returned if there are no higher keys in the tree
@@ -4116,8 +5595,7 @@ out:
  * calling this function.
  */
 int btrfs_find_next_key(struct btrfs_root *root, struct btrfs_path *path,
-			struct btrfs_key *key, int level,
-			int cache_only, u64 min_trans)
+			struct btrfs_key *key, int level, u64 min_trans)
 {
 	int slot;
 	struct extent_buffer *c;
@@ -4150,7 +5628,7 @@ next:
 				btrfs_node_key_to_cpu(c, &cur_key, slot);
 
 			orig_lowest = path->lowest_level;
-			btrfs_release_path(root, path);
+			btrfs_release_path(path);
 			path->lowest_level = level;
 			ret = btrfs_search_slot(NULL, root, &cur_key, path,
 						0, 0);
@@ -4168,21 +5646,8 @@ next:
 		if (level == 0)
 			btrfs_item_key_to_cpu(c, key, slot);
 		else {
-			u64 blockptr = btrfs_node_blockptr(c, slot);
 			u64 gen = btrfs_node_ptr_generation(c, slot);
 
-			if (cache_only) {
-				struct extent_buffer *cur;
-				cur = btrfs_find_tree_block(root, blockptr,
-					    btrfs_level_size(root, level - 1));
-				if (!cur || !btrfs_buffer_uptodate(cur, gen)) {
-					slot++;
-					if (cur)
-						free_extent_buffer(cur);
-					goto next;
-				}
-				free_extent_buffer(cur);
-			}
 			if (gen < min_trans) {
 				slot++;
 				goto next;
@@ -4201,6 +5666,12 @@ next:
  */
 int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path)
 {
+	return btrfs_next_old_leaf(root, path, 0);
+}
+
+int btrfs_next_old_leaf(struct btrfs_root *root, struct btrfs_path *path,
+			u64 time_seq)
+{
 	int slot;
 	int level;
 	struct extent_buffer *c;
@@ -4209,32 +5680,26 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path)
 	u32 nritems;
 	int ret;
 	int old_spinning = path->leave_spinning;
-	int force_blocking = 0;
+	int next_rw_lock = 0;
 
 	nritems = btrfs_header_nritems(path->nodes[0]);
 	if (nritems == 0)
 		return 1;
 
-	/*
-	 * we take the blocks in an order that upsets lockdep.  Using
-	 * blocking mode is the only way around it.
-	 */
-#ifdef CONFIG_DEBUG_LOCK_ALLOC
-	force_blocking = 1;
-#endif
-
 	btrfs_item_key_to_cpu(path->nodes[0], &key, nritems - 1);
 again:
 	level = 1;
 	next = NULL;
-	btrfs_release_path(root, path);
+	next_rw_lock = 0;
+	btrfs_release_path(path);
 
 	path->keep_locks = 1;
+	path->leave_spinning = 1;
 
-	if (!force_blocking)
-		path->leave_spinning = 1;
-
-	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+	if (time_seq)
+		ret = btrfs_search_old_slot(root, &key, path, time_seq);
+	else
+		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
 	path->keep_locks = 0;
 
 	if (ret < 0)
@@ -4253,6 +5718,24 @@ again:
 		ret = 0;
 		goto done;
 	}
+	/*
+	 * So the above check misses one case:
+	 * - after releasing the path above, someone has removed the item that
+	 *   used to be at the very end of the block, and balance between leafs
+	 *   gets another one with bigger key.offset to replace it.
+	 *
+	 * This one should be returned as well, or we can get leaf corruption
+	 * later(esp. in __btrfs_drop_extents()).
+	 *
+	 * And a bit more explanation about this check,
+	 * with ret > 0, the key isn't found, the path points to the slot
+	 * where it should be inserted, so the path->slots[0] item must be the
+	 * bigger one.
+	 */
+	if (nritems > 0 && ret > 0 && path->slots[0] == nritems - 1) {
+		ret = 0;
+		goto done;
+	}
 
 	while (level < BTRFS_MAX_LEVEL) {
 		if (!path->nodes[level]) {
@@ -4272,31 +5755,44 @@ again:
 		}
 
 		if (next) {
-			btrfs_tree_unlock(next);
+			btrfs_tree_unlock_rw(next, next_rw_lock);
 			free_extent_buffer(next);
 		}
 
 		next = c;
+		next_rw_lock = path->locks[level];
 		ret = read_block_for_search(NULL, root, path, &next, level,
-					    slot, &key);
+					    slot, &key, 0);
 		if (ret == -EAGAIN)
 			goto again;
 
 		if (ret < 0) {
-			btrfs_release_path(root, path);
+			btrfs_release_path(path);
 			goto done;
 		}
 
 		if (!path->skip_locking) {
-			ret = btrfs_try_spin_lock(next);
+			ret = btrfs_try_tree_read_lock(next);
+			if (!ret && time_seq) {
+				/*
+				 * If we don't get the lock, we may be racing
+				 * with push_leaf_left, holding that lock while
+				 * itself waiting for the leaf we've currently
+				 * locked. To solve this situation, we give up
+				 * on our lock and cycle.
+				 */
+				free_extent_buffer(next);
+				btrfs_release_path(path);
+				cond_resched();
+				goto again;
+			}
 			if (!ret) {
 				btrfs_set_path_blocking(path);
-				btrfs_tree_lock(next);
-				if (!force_blocking)
-					btrfs_clear_path_blocking(path, next);
+				btrfs_tree_read_lock(next);
+				btrfs_clear_path_blocking(path, next,
+							  BTRFS_READ_LOCK);
 			}
-			if (force_blocking)
-				btrfs_set_lock_blocking(next);
+			next_rw_lock = BTRFS_READ_LOCK;
 		}
 		break;
 	}
@@ -4305,43 +5801,40 @@ again:
 		level--;
 		c = path->nodes[level];
 		if (path->locks[level])
-			btrfs_tree_unlock(c);
+			btrfs_tree_unlock_rw(c, path->locks[level]);
 
 		free_extent_buffer(c);
 		path->nodes[level] = next;
 		path->slots[level] = 0;
 		if (!path->skip_locking)
-			path->locks[level] = 1;
-
+			path->locks[level] = next_rw_lock;
 		if (!level)
 			break;
 
 		ret = read_block_for_search(NULL, root, path, &next, level,
-					    0, &key);
+					    0, &key, 0);
 		if (ret == -EAGAIN)
 			goto again;
 
 		if (ret < 0) {
-			btrfs_release_path(root, path);
+			btrfs_release_path(path);
 			goto done;
 		}
 
 		if (!path->skip_locking) {
-			btrfs_assert_tree_locked(path->nodes[level]);
-			ret = btrfs_try_spin_lock(next);
+			ret = btrfs_try_tree_read_lock(next);
 			if (!ret) {
 				btrfs_set_path_blocking(path);
-				btrfs_tree_lock(next);
-				if (!force_blocking)
-					btrfs_clear_path_blocking(path, next);
+				btrfs_tree_read_lock(next);
+				btrfs_clear_path_blocking(path, next,
+							  BTRFS_READ_LOCK);
 			}
-			if (force_blocking)
-				btrfs_set_lock_blocking(next);
+			next_rw_lock = BTRFS_READ_LOCK;
 		}
 	}
 	ret = 0;
 done:
-	unlock_up(path, 0, 1);
+	unlock_up(path, 0, 1, 0, NULL);
 	path->leave_spinning = old_spinning;
 	if (!old_spinning)
 		btrfs_set_path_blocking(path);
@@ -4391,3 +5884,46 @@ int btrfs_previous_item(struct btrfs_root *root,
 	}
 	return 1;
 }
+
+/*
+ * search in extent tree to find a previous Metadata/Data extent item with
+ * min objecitd.
+ *
+ * returns 0 if something is found, 1 if nothing was found and < 0 on error
+ */
+int btrfs_previous_extent_item(struct btrfs_root *root,
+			struct btrfs_path *path, u64 min_objectid)
+{
+	struct btrfs_key found_key;
+	struct extent_buffer *leaf;
+	u32 nritems;
+	int ret;
+
+	while (1) {
+		if (path->slots[0] == 0) {
+			btrfs_set_path_blocking(path);
+			ret = btrfs_prev_leaf(root, path);
+			if (ret != 0)
+				return ret;
+		} else {
+			path->slots[0]--;
+		}
+		leaf = path->nodes[0];
+		nritems = btrfs_header_nritems(leaf);
+		if (nritems == 0)
+			return 1;
+		if (path->slots[0] == nritems)
+			path->slots[0]--;
+
+		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
+		if (found_key.objectid < min_objectid)
+			break;
+		if (found_key.type == BTRFS_EXTENT_ITEM_KEY ||
+		    found_key.type == BTRFS_METADATA_ITEM_KEY)
+			return 0;
+		if (found_key.objectid == min_objectid &&
+		    found_key.type < BTRFS_EXTENT_ITEM_KEY)
+			break;
+	}
+	return 1;
+}
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 746a7248678..be91397f4e9 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -19,28 +19,44 @@
 #ifndef __BTRFS_CTREE__
 #define __BTRFS_CTREE__
 
-#include <linux/version.h>
 #include <linux/mm.h>
 #include <linux/highmem.h>
 #include <linux/fs.h>
+#include <linux/rwsem.h>
+#include <linux/semaphore.h>
 #include <linux/completion.h>
 #include <linux/backing-dev.h>
 #include <linux/wait.h>
 #include <linux/slab.h>
+#include <linux/kobject.h>
+#include <trace/events/btrfs.h>
 #include <asm/kmap_types.h>
+#include <linux/pagemap.h>
+#include <linux/btrfs.h>
+#include <linux/workqueue.h>
 #include "extent_io.h"
 #include "extent_map.h"
 #include "async-thread.h"
 
 struct btrfs_trans_handle;
 struct btrfs_transaction;
+struct btrfs_pending_snapshot;
 extern struct kmem_cache *btrfs_trans_handle_cachep;
 extern struct kmem_cache *btrfs_transaction_cachep;
 extern struct kmem_cache *btrfs_bit_radix_cachep;
 extern struct kmem_cache *btrfs_path_cachep;
+extern struct kmem_cache *btrfs_free_space_cachep;
 struct btrfs_ordered_sum;
 
-#define BTRFS_MAGIC "_BHRfS_M"
+#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
+#define STATIC noinline
+#else
+#define STATIC static noinline
+#endif
+
+#define BTRFS_MAGIC 0x4D5F53665248425FULL /* ascii _BHRfS_M, no null */
+
+#define BTRFS_MAX_MIRRORS 3
 
 #define BTRFS_MAX_LEVEL 8
 
@@ -80,6 +96,15 @@ struct btrfs_ordered_sum;
 /* holds checksums of all the data extents */
 #define BTRFS_CSUM_TREE_OBJECTID 7ULL
 
+/* holds quota configuration and tracking */
+#define BTRFS_QUOTA_TREE_OBJECTID 8ULL
+
+/* for storing items that use the BTRFS_UUID_KEY* types */
+#define BTRFS_UUID_TREE_OBJECTID 9ULL
+
+/* for storing balance parameters in the root tree */
+#define BTRFS_BALANCE_OBJECTID -4ULL
+
 /* orhpan objectid for tracking unlinked/truncated files */
 #define BTRFS_ORPHAN_OBJECTID -5ULL
 
@@ -98,6 +123,15 @@ struct btrfs_ordered_sum;
  */
 #define BTRFS_EXTENT_CSUM_OBJECTID -10ULL
 
+/* For storing free space cache */
+#define BTRFS_FREE_SPACE_OBJECTID -11ULL
+
+/*
+ * The inode number assigned to the special inode for storing
+ * free ino cache
+ */
+#define BTRFS_FREE_INO_OBJECTID -12ULL
+
 /* dummy objectid represents multiple objectids */
 #define BTRFS_MULTIPLE_OBJECTIDS -255ULL
 
@@ -119,12 +153,27 @@ struct btrfs_ordered_sum;
 
 #define BTRFS_EMPTY_SUBVOL_DIR_OBJECTID 2
 
+#define BTRFS_DEV_REPLACE_DEVID 0ULL
+
+/*
+ * the max metadata block size.  This limit is somewhat artificial,
+ * but the memmove costs go through the roof for larger blocks.
+ */
+#define BTRFS_MAX_METADATA_BLOCKSIZE 65536
+
 /*
  * we can actually store much bigger names, but lets not confuse the rest
  * of linux
  */
 #define BTRFS_NAME_LEN 255
 
+/*
+ * Theoretical limit is larger, but we keep this down to a sane
+ * value. That should limit greatly the possibility of collisions on
+ * inode ref items.
+ */
+#define BTRFS_LINK_MAX 65535U
+
 /* 32 bytes in various csum fields */
 #define BTRFS_CSUM_SIZE 32
 
@@ -136,6 +185,9 @@ static int btrfs_csum_sizes[] = { 4, 0 };
 /* four bytes for CRC32 */
 #define BTRFS_EMPTY_DIR_SIZE 0
 
+/* spefic to btrfs_map_block(), therefore not in include/linux/blk_types.h */
+#define REQ_GET_READ_MIRRORS	(1 << 30)
+
 #define BTRFS_FT_UNKNOWN	0
 #define BTRFS_FT_REG_FILE	1
 #define BTRFS_FT_DIR		2
@@ -147,6 +199,11 @@ static int btrfs_csum_sizes[] = { 4, 0 };
 #define BTRFS_FT_XATTR		8
 #define BTRFS_FT_MAX		9
 
+/* ioprio of readahead is set to idle */
+#define BTRFS_IOPRIO_READA (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_IDLE, 0))
+
+#define BTRFS_DIRTY_METADATA_THRESH	(32 * 1024 * 1024)
+
 /*
  * The key defines the order in the tree, and so it also defines (optimal)
  * block layout.
@@ -180,7 +237,6 @@ struct btrfs_mapping_tree {
 	struct extent_map_tree map_tree;
 };
 
-#define BTRFS_UUID_SIZE 16
 struct btrfs_dev_item {
 	/* the internal btrfs device id */
 	__le64 devid;
@@ -264,6 +320,22 @@ struct btrfs_chunk {
 	/* additional stripes go here */
 } __attribute__ ((__packed__));
 
+#define BTRFS_FREE_SPACE_EXTENT	1
+#define BTRFS_FREE_SPACE_BITMAP	2
+
+struct btrfs_free_space_entry {
+	__le64 offset;
+	__le64 bytes;
+	u8 type;
+} __attribute__ ((__packed__));
+
+struct btrfs_free_space_header {
+	struct btrfs_disk_key location;
+	__le64 generation;
+	__le64 num_entries;
+	__le64 num_bitmaps;
+} __attribute__ ((__packed__));
+
 static inline unsigned long btrfs_chunk_item_size(int num_stripes)
 {
 	BUG_ON(num_stripes == 0);
@@ -271,9 +343,21 @@ static inline unsigned long btrfs_chunk_item_size(int num_stripes)
 		sizeof(struct btrfs_stripe) * (num_stripes - 1);
 }
 
-#define BTRFS_FSID_SIZE 16
 #define BTRFS_HEADER_FLAG_WRITTEN	(1ULL << 0)
 #define BTRFS_HEADER_FLAG_RELOC		(1ULL << 1)
+
+/*
+ * File system states
+ */
+#define BTRFS_FS_STATE_ERROR		0
+#define BTRFS_FS_STATE_REMOUNTING	1
+#define BTRFS_FS_STATE_TRANS_ABORTED	2
+#define BTRFS_FS_STATE_DEV_REPLACING	3
+
+/* Super block flags */
+/* Errors detected */
+#define BTRFS_SUPER_FLAG_ERROR		(1ULL << 2)
+
 #define BTRFS_SUPER_FLAG_SEEDING	(1ULL << 32)
 #define BTRFS_SUPER_FLAG_METADUMP	(1ULL << 33)
 
@@ -324,6 +408,47 @@ struct btrfs_header {
 #define BTRFS_LABEL_SIZE 256
 
 /*
+ * just in case we somehow lose the roots and are not able to mount,
+ * we store an array of the roots from previous transactions
+ * in the super.
+ */
+#define BTRFS_NUM_BACKUP_ROOTS 4
+struct btrfs_root_backup {
+	__le64 tree_root;
+	__le64 tree_root_gen;
+
+	__le64 chunk_root;
+	__le64 chunk_root_gen;
+
+	__le64 extent_root;
+	__le64 extent_root_gen;
+
+	__le64 fs_root;
+	__le64 fs_root_gen;
+
+	__le64 dev_root;
+	__le64 dev_root_gen;
+
+	__le64 csum_root;
+	__le64 csum_root_gen;
+
+	__le64 total_bytes;
+	__le64 bytes_used;
+	__le64 num_devices;
+	/* future */
+	__le64 unused_64[4];
+
+	u8 tree_root_level;
+	u8 chunk_root_level;
+	u8 extent_root_level;
+	u8 fs_root_level;
+	u8 dev_root_level;
+	u8 csum_root_level;
+	/* future and to align */
+	u8 unused_8[10];
+} __attribute__ ((__packed__));
+
+/*
  * the super block basically lists the main trees of the FS
  * it currently lacks any block count etc etc
  */
@@ -364,9 +489,13 @@ struct btrfs_super_block {
 
 	char label[BTRFS_LABEL_SIZE];
 
+	__le64 cache_generation;
+	__le64 uuid_tree_generation;
+
 	/* future expansion */
-	__le64 reserved[32];
+	__le64 reserved[30];
 	u8 sys_chunk_array[BTRFS_SYSTEM_CHUNK_ARRAY_SIZE];
+	struct btrfs_root_backup super_roots[BTRFS_NUM_BACKUP_ROOTS];
 } __attribute__ ((__packed__));
 
 /*
@@ -374,13 +503,49 @@ struct btrfs_super_block {
  * ones specified below then we will fail to mount
  */
 #define BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF	(1ULL << 0)
-#define BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL	(2ULL << 0)
+#define BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL	(1ULL << 1)
+#define BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS	(1ULL << 2)
+#define BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO	(1ULL << 3)
+/*
+ * some patches floated around with a second compression method
+ * lets save that incompat here for when they do get in
+ * Note we don't actually support it, we're just reserving the
+ * number
+ */
+#define BTRFS_FEATURE_INCOMPAT_COMPRESS_LZOv2	(1ULL << 4)
+
+/*
+ * older kernels tried to do bigger metadata blocks, but the
+ * code was pretty buggy.  Lets not let them try anymore.
+ */
+#define BTRFS_FEATURE_INCOMPAT_BIG_METADATA	(1ULL << 5)
+
+#define BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF	(1ULL << 6)
+#define BTRFS_FEATURE_INCOMPAT_RAID56		(1ULL << 7)
+#define BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA	(1ULL << 8)
+#define BTRFS_FEATURE_INCOMPAT_NO_HOLES		(1ULL << 9)
 
 #define BTRFS_FEATURE_COMPAT_SUPP		0ULL
+#define BTRFS_FEATURE_COMPAT_SAFE_SET		0ULL
+#define BTRFS_FEATURE_COMPAT_SAFE_CLEAR		0ULL
 #define BTRFS_FEATURE_COMPAT_RO_SUPP		0ULL
-#define BTRFS_FEATURE_INCOMPAT_SUPP		\
-	(BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF |	\
-	 BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)
+#define BTRFS_FEATURE_COMPAT_RO_SAFE_SET	0ULL
+#define BTRFS_FEATURE_COMPAT_RO_SAFE_CLEAR	0ULL
+
+#define BTRFS_FEATURE_INCOMPAT_SUPP			\
+	(BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF |		\
+	 BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL |	\
+	 BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS |		\
+	 BTRFS_FEATURE_INCOMPAT_BIG_METADATA |		\
+	 BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO |		\
+	 BTRFS_FEATURE_INCOMPAT_RAID56 |		\
+	 BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF |		\
+	 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA |	\
+	 BTRFS_FEATURE_INCOMPAT_NO_HOLES)
+
+#define BTRFS_FEATURE_INCOMPAT_SAFE_SET			\
+	(BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF)
+#define BTRFS_FEATURE_INCOMPAT_SAFE_CLEAR		0ULL
 
 /*
  * A leaf is full of items. offset and size tell us where to find
@@ -445,6 +610,7 @@ struct btrfs_path {
 	unsigned int skip_locking:1;
 	unsigned int leave_spinning:1;
 	unsigned int search_commit_root:1;
+	unsigned int need_commit_sem:1;
 };
 
 /*
@@ -473,6 +639,12 @@ struct btrfs_extent_item_v0 {
 /* use full backrefs for extent pointers in the block */
 #define BTRFS_BLOCK_FLAG_FULL_BACKREF	(1ULL << 8)
 
+/*
+ * this flag is only used internally by scrub and may be changed at any time
+ * it is only declared here to avoid collisions
+ */
+#define BTRFS_EXTENT_FLAG_SUPER		(1ULL << 48)
+
 struct btrfs_tree_block_info {
 	struct btrfs_disk_key key;
 	u8 level;
@@ -521,15 +693,25 @@ struct btrfs_inode_ref {
 	/* name goes here */
 } __attribute__ ((__packed__));
 
+struct btrfs_inode_extref {
+	__le64 parent_objectid;
+	__le64 index;
+	__le16 name_len;
+	__u8   name[0];
+	/* name goes here */
+} __attribute__ ((__packed__));
+
 struct btrfs_timespec {
 	__le64 sec;
 	__le32 nsec;
 } __attribute__ ((__packed__));
 
 enum btrfs_compression_type {
-	BTRFS_COMPRESS_NONE = 0,
-	BTRFS_COMPRESS_ZLIB = 1,
-	BTRFS_COMPRESS_LAST = 2,
+	BTRFS_COMPRESS_NONE  = 0,
+	BTRFS_COMPRESS_ZLIB  = 1,
+	BTRFS_COMPRESS_LZO   = 2,
+	BTRFS_COMPRESS_TYPES = 2,
+	BTRFS_COMPRESS_LAST  = 3,
 };
 
 struct btrfs_inode_item {
@@ -573,6 +755,14 @@ struct btrfs_dir_item {
 	u8 type;
 } __attribute__ ((__packed__));
 
+#define BTRFS_ROOT_SUBVOL_RDONLY	(1ULL << 0)
+
+/*
+ * Internal in-memory flag that a subvolume has been marked for deletion but
+ * still visible as a directory
+ */
+#define BTRFS_ROOT_SUBVOL_DEAD		(1ULL << 48)
+
 struct btrfs_root_item {
 	struct btrfs_inode_item inode;
 	__le64 generation;
@@ -586,6 +776,36 @@ struct btrfs_root_item {
 	struct btrfs_disk_key drop_progress;
 	u8 drop_level;
 	u8 level;
+
+	/*
+	 * The following fields appear after subvol_uuids+subvol_times
+	 * were introduced.
+	 */
+
+	/*
+	 * This generation number is used to test if the new fields are valid
+	 * and up to date while reading the root item. Everytime the root item
+	 * is written out, the "generation" field is copied into this field. If
+	 * anyone ever mounted the fs with an older kernel, we will have
+	 * mismatching generation values here and thus must invalidate the
+	 * new fields. See btrfs_update_root and btrfs_find_last_root for
+	 * details.
+	 * the offset of generation_v2 is also used as the start for the memset
+	 * when invalidating the fields.
+	 */
+	__le64 generation_v2;
+	u8 uuid[BTRFS_UUID_SIZE];
+	u8 parent_uuid[BTRFS_UUID_SIZE];
+	u8 received_uuid[BTRFS_UUID_SIZE];
+	__le64 ctransid; /* updated when an inode changes */
+	__le64 otransid; /* trans when created */
+	__le64 stransid; /* trans when sent. non-zero for received subvol */
+	__le64 rtransid; /* trans when received. non-zero for received subvol */
+	struct btrfs_timespec ctime;
+	struct btrfs_timespec otime;
+	struct btrfs_timespec stime;
+	struct btrfs_timespec rtime;
+	__le64 reserved[8]; /* for future */
 } __attribute__ ((__packed__));
 
 /*
@@ -597,6 +817,57 @@ struct btrfs_root_ref {
 	__le16 name_len;
 } __attribute__ ((__packed__));
 
+struct btrfs_disk_balance_args {
+	/*
+	 * profiles to operate on, single is denoted by
+	 * BTRFS_AVAIL_ALLOC_BIT_SINGLE
+	 */
+	__le64 profiles;
+
+	/* usage filter */
+	__le64 usage;
+
+	/* devid filter */
+	__le64 devid;
+
+	/* devid subset filter [pstart..pend) */
+	__le64 pstart;
+	__le64 pend;
+
+	/* btrfs virtual address space subset filter [vstart..vend) */
+	__le64 vstart;
+	__le64 vend;
+
+	/*
+	 * profile to convert to, single is denoted by
+	 * BTRFS_AVAIL_ALLOC_BIT_SINGLE
+	 */
+	__le64 target;
+
+	/* BTRFS_BALANCE_ARGS_* */
+	__le64 flags;
+
+	/* BTRFS_BALANCE_ARGS_LIMIT value */
+	__le64 limit;
+
+	__le64 unused[7];
+} __attribute__ ((__packed__));
+
+/*
+ * store balance parameters to disk so that balance can be properly
+ * resumed after crash or unmount
+ */
+struct btrfs_balance_item {
+	/* BTRFS_BALANCE_* */
+	__le64 flags;
+
+	struct btrfs_disk_balance_args data;
+	struct btrfs_disk_balance_args meta;
+	struct btrfs_disk_balance_args sys;
+
+	__le64 unused[4];
+} __attribute__ ((__packed__));
+
 #define BTRFS_FILE_EXTENT_INLINE 0
 #define BTRFS_FILE_EXTENT_REG 1
 #define BTRFS_FILE_EXTENT_PREALLOC 2
@@ -655,14 +926,130 @@ struct btrfs_csum_item {
 	u8 csum;
 } __attribute__ ((__packed__));
 
+struct btrfs_dev_stats_item {
+	/*
+	 * grow this item struct at the end for future enhancements and keep
+	 * the existing values unchanged
+	 */
+	__le64 values[BTRFS_DEV_STAT_VALUES_MAX];
+} __attribute__ ((__packed__));
+
+#define BTRFS_DEV_REPLACE_ITEM_CONT_READING_FROM_SRCDEV_MODE_ALWAYS	0
+#define BTRFS_DEV_REPLACE_ITEM_CONT_READING_FROM_SRCDEV_MODE_AVOID	1
+#define BTRFS_DEV_REPLACE_ITEM_STATE_NEVER_STARTED	0
+#define BTRFS_DEV_REPLACE_ITEM_STATE_STARTED		1
+#define BTRFS_DEV_REPLACE_ITEM_STATE_SUSPENDED		2
+#define BTRFS_DEV_REPLACE_ITEM_STATE_FINISHED		3
+#define BTRFS_DEV_REPLACE_ITEM_STATE_CANCELED		4
+
+struct btrfs_dev_replace {
+	u64 replace_state;	/* see #define above */
+	u64 time_started;	/* seconds since 1-Jan-1970 */
+	u64 time_stopped;	/* seconds since 1-Jan-1970 */
+	atomic64_t num_write_errors;
+	atomic64_t num_uncorrectable_read_errors;
+
+	u64 cursor_left;
+	u64 committed_cursor_left;
+	u64 cursor_left_last_write_of_item;
+	u64 cursor_right;
+
+	u64 cont_reading_from_srcdev_mode;	/* see #define above */
+
+	int is_valid;
+	int item_needs_writeback;
+	struct btrfs_device *srcdev;
+	struct btrfs_device *tgtdev;
+
+	pid_t lock_owner;
+	atomic_t nesting_level;
+	struct mutex lock_finishing_cancel_unmount;
+	struct mutex lock_management_lock;
+	struct mutex lock;
+
+	struct btrfs_scrub_progress scrub_progress;
+};
+
+struct btrfs_dev_replace_item {
+	/*
+	 * grow this item struct at the end for future enhancements and keep
+	 * the existing values unchanged
+	 */
+	__le64 src_devid;
+	__le64 cursor_left;
+	__le64 cursor_right;
+	__le64 cont_reading_from_srcdev_mode;
+
+	__le64 replace_state;
+	__le64 time_started;
+	__le64 time_stopped;
+	__le64 num_write_errors;
+	__le64 num_uncorrectable_read_errors;
+} __attribute__ ((__packed__));
+
 /* different types of block groups (and chunks) */
-#define BTRFS_BLOCK_GROUP_DATA     (1 << 0)
-#define BTRFS_BLOCK_GROUP_SYSTEM   (1 << 1)
-#define BTRFS_BLOCK_GROUP_METADATA (1 << 2)
-#define BTRFS_BLOCK_GROUP_RAID0    (1 << 3)
-#define BTRFS_BLOCK_GROUP_RAID1    (1 << 4)
-#define BTRFS_BLOCK_GROUP_DUP	   (1 << 5)
-#define BTRFS_BLOCK_GROUP_RAID10   (1 << 6)
+#define BTRFS_BLOCK_GROUP_DATA		(1ULL << 0)
+#define BTRFS_BLOCK_GROUP_SYSTEM	(1ULL << 1)
+#define BTRFS_BLOCK_GROUP_METADATA	(1ULL << 2)
+#define BTRFS_BLOCK_GROUP_RAID0		(1ULL << 3)
+#define BTRFS_BLOCK_GROUP_RAID1		(1ULL << 4)
+#define BTRFS_BLOCK_GROUP_DUP		(1ULL << 5)
+#define BTRFS_BLOCK_GROUP_RAID10	(1ULL << 6)
+#define BTRFS_BLOCK_GROUP_RAID5         (1ULL << 7)
+#define BTRFS_BLOCK_GROUP_RAID6         (1ULL << 8)
+#define BTRFS_BLOCK_GROUP_RESERVED	(BTRFS_AVAIL_ALLOC_BIT_SINGLE | \
+					 BTRFS_SPACE_INFO_GLOBAL_RSV)
+
+enum btrfs_raid_types {
+	BTRFS_RAID_RAID10,
+	BTRFS_RAID_RAID1,
+	BTRFS_RAID_DUP,
+	BTRFS_RAID_RAID0,
+	BTRFS_RAID_SINGLE,
+	BTRFS_RAID_RAID5,
+	BTRFS_RAID_RAID6,
+	BTRFS_NR_RAID_TYPES
+};
+
+#define BTRFS_BLOCK_GROUP_TYPE_MASK	(BTRFS_BLOCK_GROUP_DATA |    \
+					 BTRFS_BLOCK_GROUP_SYSTEM |  \
+					 BTRFS_BLOCK_GROUP_METADATA)
+
+#define BTRFS_BLOCK_GROUP_PROFILE_MASK	(BTRFS_BLOCK_GROUP_RAID0 |   \
+					 BTRFS_BLOCK_GROUP_RAID1 |   \
+					 BTRFS_BLOCK_GROUP_RAID5 |   \
+					 BTRFS_BLOCK_GROUP_RAID6 |   \
+					 BTRFS_BLOCK_GROUP_DUP |     \
+					 BTRFS_BLOCK_GROUP_RAID10)
+/*
+ * We need a bit for restriper to be able to tell when chunks of type
+ * SINGLE are available.  This "extended" profile format is used in
+ * fs_info->avail_*_alloc_bits (in-memory) and balance item fields
+ * (on-disk).  The corresponding on-disk bit in chunk.type is reserved
+ * to avoid remappings between two formats in future.
+ */
+#define BTRFS_AVAIL_ALLOC_BIT_SINGLE	(1ULL << 48)
+
+/*
+ * A fake block group type that is used to communicate global block reserve
+ * size to userspace via the SPACE_INFO ioctl.
+ */
+#define BTRFS_SPACE_INFO_GLOBAL_RSV	(1ULL << 49)
+
+#define BTRFS_EXTENDED_PROFILE_MASK	(BTRFS_BLOCK_GROUP_PROFILE_MASK | \
+					 BTRFS_AVAIL_ALLOC_BIT_SINGLE)
+
+static inline u64 chunk_to_extended(u64 flags)
+{
+	if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0)
+		flags |= BTRFS_AVAIL_ALLOC_BIT_SINGLE;
+
+	return flags;
+}
+static inline u64 extended_to_chunk(u64 flags)
+{
+	return flags & ~BTRFS_AVAIL_ALLOC_BIT_SINGLE;
+}
 
 struct btrfs_block_group_item {
 	__le64 used;
@@ -670,44 +1057,147 @@ struct btrfs_block_group_item {
 	__le64 flags;
 } __attribute__ ((__packed__));
 
+/*
+ * is subvolume quota turned on?
+ */
+#define BTRFS_QGROUP_STATUS_FLAG_ON		(1ULL << 0)
+/*
+ * RESCAN is set during the initialization phase
+ */
+#define BTRFS_QGROUP_STATUS_FLAG_RESCAN		(1ULL << 1)
+/*
+ * Some qgroup entries are known to be out of date,
+ * either because the configuration has changed in a way that
+ * makes a rescan necessary, or because the fs has been mounted
+ * with a non-qgroup-aware version.
+ * Turning qouta off and on again makes it inconsistent, too.
+ */
+#define BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT	(1ULL << 2)
+
+#define BTRFS_QGROUP_STATUS_VERSION        1
+
+struct btrfs_qgroup_status_item {
+	__le64 version;
+	/*
+	 * the generation is updated during every commit. As older
+	 * versions of btrfs are not aware of qgroups, it will be
+	 * possible to detect inconsistencies by checking the
+	 * generation on mount time
+	 */
+	__le64 generation;
+
+	/* flag definitions see above */
+	__le64 flags;
+
+	/*
+	 * only used during scanning to record the progress
+	 * of the scan. It contains a logical address
+	 */
+	__le64 rescan;
+} __attribute__ ((__packed__));
+
+struct btrfs_qgroup_info_item {
+	__le64 generation;
+	__le64 rfer;
+	__le64 rfer_cmpr;
+	__le64 excl;
+	__le64 excl_cmpr;
+} __attribute__ ((__packed__));
+
+/* flags definition for qgroup limits */
+#define BTRFS_QGROUP_LIMIT_MAX_RFER	(1ULL << 0)
+#define BTRFS_QGROUP_LIMIT_MAX_EXCL	(1ULL << 1)
+#define BTRFS_QGROUP_LIMIT_RSV_RFER	(1ULL << 2)
+#define BTRFS_QGROUP_LIMIT_RSV_EXCL	(1ULL << 3)
+#define BTRFS_QGROUP_LIMIT_RFER_CMPR	(1ULL << 4)
+#define BTRFS_QGROUP_LIMIT_EXCL_CMPR	(1ULL << 5)
+
+struct btrfs_qgroup_limit_item {
+	/*
+	 * only updated when any of the other values change
+	 */
+	__le64 flags;
+	__le64 max_rfer;
+	__le64 max_excl;
+	__le64 rsv_rfer;
+	__le64 rsv_excl;
+} __attribute__ ((__packed__));
+
+/* For raid type sysfs entries */
+struct raid_kobject {
+	int raid_type;
+	struct kobject kobj;
+};
+
 struct btrfs_space_info {
-	u64 flags;
+	spinlock_t lock;
 
-	u64 total_bytes;	/* total bytes in the space */
-	u64 bytes_used;		/* total bytes used on disk */
+	u64 total_bytes;	/* total bytes in the space,
+				   this doesn't take mirrors into account */
+	u64 bytes_used;		/* total bytes used,
+				   this doesn't take mirrors into account */
 	u64 bytes_pinned;	/* total bytes pinned, will be freed when the
 				   transaction finishes */
 	u64 bytes_reserved;	/* total bytes the allocator has reserved for
 				   current allocations */
-	u64 bytes_readonly;	/* total bytes that are read only */
-	u64 bytes_super;	/* total bytes reserved for the super blocks */
-	u64 bytes_root;		/* the number of bytes needed to commit a
-				   transaction */
 	u64 bytes_may_use;	/* number of bytes that may be used for
 				   delalloc/allocations */
-	u64 bytes_delalloc;	/* number of bytes currently reserved for
-				   delayed allocation */
+	u64 bytes_readonly;	/* total bytes that are read only */
 
-	int full;		/* indicates that we cannot allocate any more
+	unsigned int full:1;	/* indicates that we cannot allocate any more
 				   chunks for this space */
-	int force_alloc;	/* set if we need to force a chunk alloc for
-				   this space */
-	int force_delalloc;	/* make people start doing filemap_flush until
-				   we're under a threshold */
+	unsigned int chunk_alloc:1;	/* set if we are allocating a chunk */
 
-	struct list_head list;
+	unsigned int flush:1;		/* set if we are trying to make space */
+
+	unsigned int force_alloc;	/* set if we need to force a chunk
+					   alloc for this space */
 
-	/* for controlling how we free up space for allocations */
-	wait_queue_head_t allocate_wait;
-	wait_queue_head_t flush_wait;
-	int allocating_chunk;
-	int flushing;
+	u64 disk_used;		/* total bytes used on disk */
+	u64 disk_total;		/* total bytes on disk, takes mirrors into
+				   account */
 
+	u64 flags;
+
+	/*
+	 * bytes_pinned is kept in line with what is actually pinned, as in
+	 * we've called update_block_group and dropped the bytes_used counter
+	 * and increased the bytes_pinned counter.  However this means that
+	 * bytes_pinned does not reflect the bytes that will be pinned once the
+	 * delayed refs are flushed, so this counter is inc'ed everytime we call
+	 * btrfs_free_extent so it is a realtime count of what will be freed
+	 * once the transaction is committed.  It will be zero'ed everytime the
+	 * transaction commits.
+	 */
+	struct percpu_counter total_bytes_pinned;
+
+	struct list_head list;
+
+	struct rw_semaphore groups_sem;
 	/* for block groups in our same type */
-	struct list_head block_groups;
+	struct list_head block_groups[BTRFS_NR_RAID_TYPES];
+	wait_queue_head_t wait;
+
+	struct kobject kobj;
+	struct kobject *block_group_kobjs[BTRFS_NR_RAID_TYPES];
+};
+
+#define	BTRFS_BLOCK_RSV_GLOBAL		1
+#define	BTRFS_BLOCK_RSV_DELALLOC	2
+#define	BTRFS_BLOCK_RSV_TRANS		3
+#define	BTRFS_BLOCK_RSV_CHUNK		4
+#define	BTRFS_BLOCK_RSV_DELOPS		5
+#define	BTRFS_BLOCK_RSV_EMPTY		6
+#define	BTRFS_BLOCK_RSV_TEMP		7
+
+struct btrfs_block_rsv {
+	u64 size;
+	u64 reserved;
+	struct btrfs_space_info *space_info;
 	spinlock_t lock;
-	struct rw_semaphore groups_sem;
-	atomic_t caching_threads;
+	unsigned short full;
+	unsigned short type;
+	unsigned short failfast;
 };
 
 /*
@@ -726,9 +1216,6 @@ struct btrfs_free_cluster {
 	/* first extent starting offset */
 	u64 window_start;
 
-	/* if this cluster simply points at a bitmap in the block group */
-	bool points_to_bitmap;
-
 	struct btrfs_block_group_cache *block_group;
 	/*
 	 * when a cluster is allocated from a block group, we put the
@@ -741,13 +1228,24 @@ struct btrfs_free_cluster {
 enum btrfs_caching_type {
 	BTRFS_CACHE_NO		= 0,
 	BTRFS_CACHE_STARTED	= 1,
-	BTRFS_CACHE_FINISHED	= 2,
+	BTRFS_CACHE_FAST	= 2,
+	BTRFS_CACHE_FINISHED	= 3,
+	BTRFS_CACHE_ERROR	= 4,
+};
+
+enum btrfs_disk_cache_state {
+	BTRFS_DC_WRITTEN	= 0,
+	BTRFS_DC_ERROR		= 1,
+	BTRFS_DC_CLEAR		= 2,
+	BTRFS_DC_SETUP		= 3,
+	BTRFS_DC_NEED_WRITE	= 4,
 };
 
 struct btrfs_caching_control {
 	struct list_head list;
 	struct mutex mutex;
 	wait_queue_head_t wait;
+	struct btrfs_work work;
 	struct btrfs_block_group_cache *block_group;
 	u64 progress;
 	atomic_t count;
@@ -757,17 +1255,31 @@ struct btrfs_block_group_cache {
 	struct btrfs_key key;
 	struct btrfs_block_group_item item;
 	struct btrfs_fs_info *fs_info;
+	struct inode *inode;
 	spinlock_t lock;
 	u64 pinned;
 	u64 reserved;
+	u64 delalloc_bytes;
 	u64 bytes_super;
 	u64 flags;
 	u64 sectorsize;
-	int extents_thresh;
-	int free_extents;
-	int total_bitmaps;
-	int ro;
-	int dirty;
+	u64 cache_generation;
+
+	/*
+	 * It is just used for the delayed data space allocation because
+	 * only the data space allocation and the relative metadata update
+	 * can be done cross the transaction.
+	 */
+	struct rw_semaphore data_rwsem;
+
+	/* for raid56, this is a full stripe, without parity */
+	unsigned long full_stripe_len;
+
+	unsigned int ro:1;
+	unsigned int dirty:1;
+	unsigned int iref:1;
+
+	int disk_cache_state;
 
 	/* cache tracking stuff */
 	int cached;
@@ -777,9 +1289,7 @@ struct btrfs_block_group_cache {
 	struct btrfs_space_info *space_info;
 
 	/* free space cache stuff */
-	spinlock_t tree_lock;
-	struct rb_root free_space_offset;
-	u64 free_space;
+	struct btrfs_free_space_ctl *free_space_ctl;
 
 	/* block group cache stuff */
 	struct rb_node cache_node;
@@ -794,11 +1304,47 @@ struct btrfs_block_group_cache {
 	 * Today it will only have one thing on it, but that may change
 	 */
 	struct list_head cluster_list;
+
+	/* For delayed block group creation */
+	struct list_head new_bg_list;
+};
+
+/* delayed seq elem */
+struct seq_list {
+	struct list_head list;
+	u64 seq;
+};
+
+enum btrfs_orphan_cleanup_state {
+	ORPHAN_CLEANUP_STARTED	= 1,
+	ORPHAN_CLEANUP_DONE	= 2,
+};
+
+/* used by the raid56 code to lock stripes for read/modify/write */
+struct btrfs_stripe_hash {
+	struct list_head hash_list;
+	wait_queue_head_t wait;
+	spinlock_t lock;
+};
+
+/* used by the raid56 code to lock stripes for read/modify/write */
+struct btrfs_stripe_hash_table {
+	struct list_head stripe_cache;
+	spinlock_t cache_lock;
+	int cache_size;
+	struct btrfs_stripe_hash table[];
 };
 
+#define BTRFS_STRIPE_HASH_TABLE_BITS 11
+
+void btrfs_init_async_reclaim_work(struct work_struct *work);
+
+/* fs_info */
 struct reloc_control;
 struct btrfs_device;
 struct btrfs_fs_devices;
+struct btrfs_balance_control;
+struct btrfs_delayed_root;
 struct btrfs_fs_info {
 	u8 fsid[BTRFS_FSID_SIZE];
 	u8 chunk_tree_uuid[BTRFS_UUID_SIZE];
@@ -808,6 +1354,8 @@ struct btrfs_fs_info {
 	struct btrfs_root *dev_root;
 	struct btrfs_root *fs_root;
 	struct btrfs_root *csum_root;
+	struct btrfs_root *quota_root;
+	struct btrfs_root *uuid_root;
 
 	/* the log root tree is a directory of all the other log roots */
 	struct btrfs_root *log_root_tree;
@@ -817,43 +1365,100 @@ struct btrfs_fs_info {
 
 	/* block group cache stuff */
 	spinlock_t block_group_cache_lock;
+	u64 first_logical_byte;
 	struct rb_root block_group_cache_tree;
 
+	/* keep track of unallocated space */
+	spinlock_t free_chunk_lock;
+	u64 free_chunk_space;
+
 	struct extent_io_tree freed_extents[2];
 	struct extent_io_tree *pinned_extents;
 
 	/* logical->physical extent mapping */
 	struct btrfs_mapping_tree mapping_tree;
 
+	/*
+	 * block reservation for extent, checksum, root tree and
+	 * delayed dir index item
+	 */
+	struct btrfs_block_rsv global_block_rsv;
+	/* block reservation for delay allocation */
+	struct btrfs_block_rsv delalloc_block_rsv;
+	/* block reservation for metadata operations */
+	struct btrfs_block_rsv trans_block_rsv;
+	/* block reservation for chunk tree */
+	struct btrfs_block_rsv chunk_block_rsv;
+	/* block reservation for delayed operations */
+	struct btrfs_block_rsv delayed_block_rsv;
+
+	struct btrfs_block_rsv empty_block_rsv;
+
 	u64 generation;
 	u64 last_trans_committed;
+	u64 avg_delayed_ref_runtime;
 
 	/*
 	 * this is updated to the current trans every time a full commit
 	 * is required instead of the faster short fsync log commits
 	 */
 	u64 last_trans_log_full_commit;
-	u64 open_ioctl_trans;
 	unsigned long mount_opt;
+	unsigned long compress_type:4;
+	int commit_interval;
+	/*
+	 * It is a suggestive number, the read side is safe even it gets a
+	 * wrong number because we will write out the data into a regular
+	 * extent. The write side(mount/remount) is under ->s_umount lock,
+	 * so it is also safe.
+	 */
 	u64 max_inline;
+	/*
+	 * Protected by ->chunk_mutex and sb->s_umount.
+	 *
+	 * The reason that we use two lock to protect it is because only
+	 * remount and mount operations can change it and these two operations
+	 * are under sb->s_umount, but the read side (chunk allocation) can not
+	 * acquire sb->s_umount or the deadlock would happen. So we use two
+	 * locks to protect it. On the write side, we must acquire two locks,
+	 * and on the read side, we just need acquire one of them.
+	 */
 	u64 alloc_start;
 	struct btrfs_transaction *running_transaction;
 	wait_queue_head_t transaction_throttle;
 	wait_queue_head_t transaction_wait;
+	wait_queue_head_t transaction_blocked_wait;
 	wait_queue_head_t async_submit_wait;
 
-	struct btrfs_super_block super_copy;
-	struct btrfs_super_block super_for_commit;
+	/*
+	 * Used to protect the incompat_flags, compat_flags, compat_ro_flags
+	 * when they are updated.
+	 *
+	 * Because we do not clear the flags for ever, so we needn't use
+	 * the lock on the read side.
+	 *
+	 * We also needn't use the lock when we mount the fs, because
+	 * there is no other task which will update the flag.
+	 */
+	spinlock_t super_lock;
+	struct btrfs_super_block *super_copy;
+	struct btrfs_super_block *super_for_commit;
 	struct block_device *__bdev;
 	struct super_block *sb;
 	struct inode *btree_inode;
 	struct backing_dev_info bdi;
-	struct mutex trans_mutex;
 	struct mutex tree_log_mutex;
 	struct mutex transaction_kthread_mutex;
 	struct mutex cleaner_mutex;
 	struct mutex chunk_mutex;
 	struct mutex volume_mutex;
+
+	/* this is used during read/modify/write to make sure
+	 * no two ios are trying to mod the same stripe at the same
+	 * time
+	 */
+	struct btrfs_stripe_hash_table *stripe_hash_table;
+
 	/*
 	 * this protects the ordered operations list only while we are
 	 * processing all of the entries on it.  This way we make
@@ -862,52 +1467,67 @@ struct btrfs_fs_info {
 	 * before jumping into the main commit.
 	 */
 	struct mutex ordered_operations_mutex;
-	struct rw_semaphore extent_commit_sem;
+
+	/*
+	 * Same as ordered_operations_mutex except this is for ordered extents
+	 * and not the operations.
+	 */
+	struct mutex ordered_extent_flush_mutex;
+
+	struct rw_semaphore commit_root_sem;
 
 	struct rw_semaphore cleanup_work_sem;
 
 	struct rw_semaphore subvol_sem;
 	struct srcu_struct subvol_srcu;
 
+	spinlock_t trans_lock;
+	/*
+	 * the reloc mutex goes with the trans lock, it is taken
+	 * during commit to protect us from the relocation code
+	 */
+	struct mutex reloc_mutex;
+
 	struct list_head trans_list;
-	struct list_head hashers;
 	struct list_head dead_roots;
 	struct list_head caching_block_groups;
 
 	spinlock_t delayed_iput_lock;
 	struct list_head delayed_iputs;
 
+	/* this protects tree_mod_seq_list */
+	spinlock_t tree_mod_seq_lock;
+	atomic64_t tree_mod_seq;
+	struct list_head tree_mod_seq_list;
+
+	/* this protects tree_mod_log */
+	rwlock_t tree_mod_log_lock;
+	struct rb_root tree_mod_log;
+
 	atomic_t nr_async_submits;
 	atomic_t async_submit_draining;
 	atomic_t nr_async_bios;
 	atomic_t async_delalloc_pages;
+	atomic_t open_ioctl_trans;
 
 	/*
-	 * this is used by the balancing code to wait for all the pending
-	 * ordered extents
+	 * this is used to protect the following list -- ordered_roots.
 	 */
-	spinlock_t ordered_extent_lock;
+	spinlock_t ordered_root_lock;
 
 	/*
-	 * all of the data=ordered extents pending writeback
+	 * all fs/file tree roots in which there are data=ordered extents
+	 * pending writeback are added into this list.
+	 *
 	 * these can span multiple transactions and basically include
 	 * every dirty data page that isn't from nodatacow
 	 */
-	struct list_head ordered_extents;
+	struct list_head ordered_roots;
 
-	/*
-	 * all of the inodes that have delalloc bytes.  It is possible for
-	 * this list to be empty even when there is still dirty data=ordered
-	 * extents waiting to finish IO.
-	 */
-	struct list_head delalloc_inodes;
-
-	/*
-	 * special rename and truncate targets that must be on disk before
-	 * we're allowed to commit.  This is basically the ext3 style
-	 * data=ordered list.
-	 */
-	struct list_head ordered_operations;
+	struct mutex delalloc_root_mutex;
+	spinlock_t delalloc_root_lock;
+	/* all fs/file tree roots that have delalloc inodes. */
+	struct list_head delalloc_roots;
 
 	/*
 	 * there is a pool of worker threads for checksumming during writes
@@ -919,26 +1539,37 @@ struct btrfs_fs_info {
 	 * A third pool does submit_bio to avoid deadlocking with the other
 	 * two
 	 */
-	struct btrfs_workers generic_worker;
-	struct btrfs_workers workers;
-	struct btrfs_workers delalloc_workers;
-	struct btrfs_workers endio_workers;
-	struct btrfs_workers endio_meta_workers;
-	struct btrfs_workers endio_meta_write_workers;
-	struct btrfs_workers endio_write_workers;
-	struct btrfs_workers submit_workers;
-	struct btrfs_workers enospc_workers;
+	struct btrfs_workqueue *workers;
+	struct btrfs_workqueue *delalloc_workers;
+	struct btrfs_workqueue *flush_workers;
+	struct btrfs_workqueue *endio_workers;
+	struct btrfs_workqueue *endio_meta_workers;
+	struct btrfs_workqueue *endio_raid56_workers;
+	struct btrfs_workqueue *rmw_workers;
+	struct btrfs_workqueue *endio_meta_write_workers;
+	struct btrfs_workqueue *endio_write_workers;
+	struct btrfs_workqueue *endio_freespace_worker;
+	struct btrfs_workqueue *submit_workers;
+	struct btrfs_workqueue *caching_workers;
+	struct btrfs_workqueue *readahead_workers;
+
 	/*
 	 * fixup workers take dirty pages that didn't properly go through
 	 * the cow mechanism and make them safe to write.  It happens
 	 * for the sys_munmap function call path
 	 */
-	struct btrfs_workers fixup_workers;
+	struct btrfs_workqueue *fixup_workers;
+	struct btrfs_workqueue *delayed_workers;
+
+	/* the extent workers do delayed refs on the extent allocation tree */
+	struct btrfs_workqueue *extent_workers;
 	struct task_struct *transaction_kthread;
 	struct task_struct *cleaner_kthread;
 	int thread_pool_size;
 
 	struct kobject super_kobj;
+	struct kobject *space_info_kobj;
+	struct kobject *device_dir_kobj;
 	struct completion kobj_unregister;
 	int do_barriers;
 	int closing;
@@ -946,10 +1577,12 @@ struct btrfs_fs_info {
 
 	u64 total_pinned;
 
-	/* protected by the delalloc lock, used to keep from writing
-	 * metadata until there is a nice batch
-	 */
-	u64 dirty_metadata_bytes;
+	/* used to keep from writing metadata until there is a nice batch */
+	struct percpu_counter dirty_metadata_bytes;
+	struct percpu_counter delalloc_bytes;
+	s32 dirty_metadata_batch;
+	s32 delalloc_batch;
+
 	struct list_head dirty_cowonly_roots;
 
 	struct btrfs_fs_devices *fs_devices;
@@ -961,11 +1594,9 @@ struct btrfs_fs_info {
 	 */
 	struct list_head space_info;
 
-	struct reloc_control *reloc_ctl;
+	struct btrfs_space_info *data_sinfo;
 
-	spinlock_t delalloc_lock;
-	spinlock_t new_trans_lock;
-	u64 delalloc_bytes;
+	struct reloc_control *reloc_ctl;
 
 	/* data_alloc_cluster is only used in ssd mode */
 	struct btrfs_free_cluster data_alloc_cluster;
@@ -973,22 +1604,152 @@ struct btrfs_fs_info {
 	/* all metadata allocations go through this cluster */
 	struct btrfs_free_cluster meta_alloc_cluster;
 
-	spinlock_t ref_cache_lock;
-	u64 total_ref_cache_size;
+	/* auto defrag inodes go here */
+	spinlock_t defrag_inodes_lock;
+	struct rb_root defrag_inodes;
+	atomic_t defrag_running;
 
+	/* Used to protect avail_{data, metadata, system}_alloc_bits */
+	seqlock_t profiles_lock;
+	/*
+	 * these three are in extended format (availability of single
+	 * chunks is denoted by BTRFS_AVAIL_ALLOC_BIT_SINGLE bit, other
+	 * types are denoted by corresponding BTRFS_BLOCK_GROUP_* bits)
+	 */
 	u64 avail_data_alloc_bits;
 	u64 avail_metadata_alloc_bits;
 	u64 avail_system_alloc_bits;
-	u64 data_alloc_profile;
-	u64 metadata_alloc_profile;
-	u64 system_alloc_profile;
+
+	/* restriper state */
+	spinlock_t balance_lock;
+	struct mutex balance_mutex;
+	atomic_t balance_running;
+	atomic_t balance_pause_req;
+	atomic_t balance_cancel_req;
+	struct btrfs_balance_control *balance_ctl;
+	wait_queue_head_t balance_wait_q;
 
 	unsigned data_chunk_allocations;
 	unsigned metadata_ratio;
 
 	void *bdev_holder;
+
+	/* private scrub information */
+	struct mutex scrub_lock;
+	atomic_t scrubs_running;
+	atomic_t scrub_pause_req;
+	atomic_t scrubs_paused;
+	atomic_t scrub_cancel_req;
+	wait_queue_head_t scrub_pause_wait;
+	int scrub_workers_refcnt;
+	struct btrfs_workqueue *scrub_workers;
+	struct btrfs_workqueue *scrub_wr_completion_workers;
+	struct btrfs_workqueue *scrub_nocow_workers;
+
+#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
+	u32 check_integrity_print_mask;
+#endif
+	/*
+	 * quota information
+	 */
+	unsigned int quota_enabled:1;
+
+	/*
+	 * quota_enabled only changes state after a commit. This holds the
+	 * next state.
+	 */
+	unsigned int pending_quota_state:1;
+
+	/* is qgroup tracking in a consistent state? */
+	u64 qgroup_flags;
+
+	/* holds configuration and tracking. Protected by qgroup_lock */
+	struct rb_root qgroup_tree;
+	struct rb_root qgroup_op_tree;
+	spinlock_t qgroup_lock;
+	spinlock_t qgroup_op_lock;
+	atomic_t qgroup_op_seq;
+
+	/*
+	 * used to avoid frequently calling ulist_alloc()/ulist_free()
+	 * when doing qgroup accounting, it must be protected by qgroup_lock.
+	 */
+	struct ulist *qgroup_ulist;
+
+	/* protect user change for quota operations */
+	struct mutex qgroup_ioctl_lock;
+
+	/* list of dirty qgroups to be written at next commit */
+	struct list_head dirty_qgroups;
+
+	/* used by btrfs_qgroup_record_ref for an efficient tree traversal */
+	u64 qgroup_seq;
+
+	/* qgroup rescan items */
+	struct mutex qgroup_rescan_lock; /* protects the progress item */
+	struct btrfs_key qgroup_rescan_progress;
+	struct btrfs_workqueue *qgroup_rescan_workers;
+	struct completion qgroup_rescan_completion;
+	struct btrfs_work qgroup_rescan_work;
+
+	/* filesystem state */
+	unsigned long fs_state;
+
+	struct btrfs_delayed_root *delayed_root;
+
+	/* readahead tree */
+	spinlock_t reada_lock;
+	struct radix_tree_root reada_tree;
+
+	/* Extent buffer radix tree */
+	spinlock_t buffer_lock;
+	struct radix_tree_root buffer_radix;
+
+	/* next backup root to be overwritten */
+	int backup_root_index;
+
+	int num_tolerated_disk_barrier_failures;
+
+	/* device replace state */
+	struct btrfs_dev_replace dev_replace;
+
+	atomic_t mutually_exclusive_operation_running;
+
+	struct percpu_counter bio_counter;
+	wait_queue_head_t replace_wait;
+
+	struct semaphore uuid_tree_rescan_sem;
+	unsigned int update_uuid_tree_gen:1;
+
+	/* Used to reclaim the metadata space in the background. */
+	struct work_struct async_reclaim_work;
 };
 
+struct btrfs_subvolume_writers {
+	struct percpu_counter	counter;
+	wait_queue_head_t	wait;
+};
+
+/*
+ * The state of btrfs root
+ */
+/*
+ * btrfs_record_root_in_trans is a multi-step process,
+ * and it can race with the balancing code.   But the
+ * race is very small, and only the first time the root
+ * is added to each transaction.  So IN_TRANS_SETUP
+ * is used to tell us when more checks are required
+ */
+#define BTRFS_ROOT_IN_TRANS_SETUP	0
+#define BTRFS_ROOT_REF_COWS		1
+#define BTRFS_ROOT_TRACK_DIRTY		2
+#define BTRFS_ROOT_IN_RADIX		3
+#define BTRFS_ROOT_DUMMY_ROOT		4
+#define BTRFS_ROOT_ORPHAN_ITEM_INSERTED	5
+#define BTRFS_ROOT_DEFRAG_RUNNING	6
+#define BTRFS_ROOT_FORCE_COW		7
+#define BTRFS_ROOT_MULTI_LOG_TASKS	8
+
 /*
  * in ram representation of the tree.  extent_root is used for all allocations
  * and for the extent tree extent_root root.
@@ -996,13 +1757,11 @@ struct btrfs_fs_info {
 struct btrfs_root {
 	struct extent_buffer *node;
 
-	/* the node lock is held while changing the node pointer */
-	spinlock_t node_lock;
-
 	struct extent_buffer *commit_root;
 	struct btrfs_root *log_root;
 	struct btrfs_root *reloc_root;
 
+	unsigned long state;
 	struct btrfs_root_item root_item;
 	struct btrfs_key root_key;
 	struct btrfs_fs_info *fs_info;
@@ -1012,16 +1771,31 @@ struct btrfs_root {
 	struct completion kobj_unregister;
 	struct mutex objectid_mutex;
 
+	spinlock_t accounting_lock;
+	struct btrfs_block_rsv *block_rsv;
+
+	/* free ino cache stuff */
+	struct btrfs_free_space_ctl *free_ino_ctl;
+	enum btrfs_caching_type cached;
+	spinlock_t cache_lock;
+	wait_queue_head_t cache_wait;
+	struct btrfs_free_space_ctl *free_ino_pinned;
+	u64 cache_progress;
+	struct inode *cache_inode;
+
 	struct mutex log_mutex;
 	wait_queue_head_t log_writer_wait;
 	wait_queue_head_t log_commit_wait[2];
+	struct list_head log_ctxs[2];
 	atomic_t log_writers;
 	atomic_t log_commit[2];
-	unsigned long log_transid;
-	unsigned long last_log_commit;
-	unsigned long log_batch;
+	atomic_t log_batch;
+	int log_transid;
+	/* No matter the commit succeeds or not*/
+	int log_transid_committed;
+	/* Just be updated when the commit succeeds. */
+	int last_log_commit;
 	pid_t log_start_pid;
-	bool log_multiple_pids;
 
 	u64 objectid;
 	u64 last_trans;
@@ -1040,37 +1814,115 @@ struct btrfs_root {
 	u32 type;
 
 	u64 highest_objectid;
-	int ref_cows;
-	int track_dirty;
-	int in_radix;
-	int clean_orphans;
+
+#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
+	u64 alloc_bytenr;
+#endif
 
 	u64 defrag_trans_start;
 	struct btrfs_key defrag_progress;
 	struct btrfs_key defrag_max;
-	int defrag_running;
 	char *name;
-	int in_sysfs;
 
 	/* the dirty list is only used by non-reference counted roots */
 	struct list_head dirty_list;
 
 	struct list_head root_list;
 
-	spinlock_t list_lock;
-	struct list_head orphan_list;
+	spinlock_t log_extents_lock[2];
+	struct list_head logged_list[2];
+
+	spinlock_t orphan_lock;
+	atomic_t orphan_inodes;
+	struct btrfs_block_rsv *orphan_block_rsv;
+	int orphan_cleanup_state;
 
 	spinlock_t inode_lock;
 	/* red-black tree that keeps track of in-memory inodes */
 	struct rb_root inode_tree;
 
 	/*
+	 * radix tree that keeps track of delayed nodes of every inode,
+	 * protected by inode_lock
+	 */
+	struct radix_tree_root delayed_nodes_tree;
+	/*
 	 * right now this just gets used so that a root has its own devid
 	 * for stat.  It may be used for more later
 	 */
-	struct super_block anon_super;
+	dev_t anon_dev;
+
+	spinlock_t root_item_lock;
+	atomic_t refs;
+
+	struct mutex delalloc_mutex;
+	spinlock_t delalloc_lock;
+	/*
+	 * all of the inodes that have delalloc bytes.  It is possible for
+	 * this list to be empty even when there is still dirty data=ordered
+	 * extents waiting to finish IO.
+	 */
+	struct list_head delalloc_inodes;
+	struct list_head delalloc_root;
+	u64 nr_delalloc_inodes;
+
+	struct mutex ordered_extent_mutex;
+	/*
+	 * this is used by the balancing code to wait for all the pending
+	 * ordered extents
+	 */
+	spinlock_t ordered_extent_lock;
+
+	/*
+	 * all of the data=ordered extents pending writeback
+	 * these can span multiple transactions and basically include
+	 * every dirty data page that isn't from nodatacow
+	 */
+	struct list_head ordered_extents;
+	struct list_head ordered_root;
+	u64 nr_ordered_extents;
+
+	/*
+	 * Number of currently running SEND ioctls to prevent
+	 * manipulation with the read-only status via SUBVOL_SETFLAGS
+	 */
+	int send_in_progress;
+	struct btrfs_subvolume_writers *subv_writers;
+	atomic_t will_be_snapshoted;
+};
+
+struct btrfs_ioctl_defrag_range_args {
+	/* start of the defrag operation */
+	__u64 start;
+
+	/* number of bytes to defrag, use (u64)-1 to say all */
+	__u64 len;
+
+	/*
+	 * flags for the operation, which can include turning
+	 * on compression for this one defrag
+	 */
+	__u64 flags;
+
+	/*
+	 * any extent bigger than this will be considered
+	 * already defragged.  Use 0 to take the kernel default
+	 * Use 1 to say every single extent must be rewritten
+	 */
+	__u32 extent_thresh;
+
+	/*
+	 * which compression method to use if turning on compression
+	 * for this defrag operation.  If unspecified, zlib will
+	 * be used
+	 */
+	__u32 compress_type;
+
+	/* spare for later */
+	__u32 unused[4];
 };
 
+
 /*
  * inode items have the data typically returned from stat and store other
  * info about object characteristics.  There is one for every file and dir in
@@ -1078,6 +1930,7 @@ struct btrfs_root {
  */
 #define BTRFS_INODE_ITEM_KEY		1
 #define BTRFS_INODE_REF_KEY		12
+#define BTRFS_INODE_EXTREF_KEY		13
 #define BTRFS_XATTR_ITEM_KEY		24
 #define BTRFS_ORPHAN_ITEM_KEY		48
 /* reserve 2-15 close to the inode for later flexibility */
@@ -1126,6 +1979,12 @@ struct btrfs_root {
  */
 #define BTRFS_EXTENT_ITEM_KEY	168
 
+/*
+ * The same as the BTRFS_EXTENT_ITEM_KEY, except it's metadata we already know
+ * the length, so we save the level in key->offset instead of the length.
+ */
+#define BTRFS_METADATA_ITEM_KEY	169
+
 #define BTRFS_TREE_BLOCK_REF_KEY	176
 
 #define BTRFS_EXTENT_DATA_REF_KEY	178
@@ -1147,11 +2006,67 @@ struct btrfs_root {
 #define BTRFS_CHUNK_ITEM_KEY	228
 
 /*
+ * Records the overall state of the qgroups.
+ * There's only one instance of this key present,
+ * (0, BTRFS_QGROUP_STATUS_KEY, 0)
+ */
+#define BTRFS_QGROUP_STATUS_KEY         240
+/*
+ * Records the currently used space of the qgroup.
+ * One key per qgroup, (0, BTRFS_QGROUP_INFO_KEY, qgroupid).
+ */
+#define BTRFS_QGROUP_INFO_KEY           242
+/*
+ * Contains the user configured limits for the qgroup.
+ * One key per qgroup, (0, BTRFS_QGROUP_LIMIT_KEY, qgroupid).
+ */
+#define BTRFS_QGROUP_LIMIT_KEY          244
+/*
+ * Records the child-parent relationship of qgroups. For
+ * each relation, 2 keys are present:
+ * (childid, BTRFS_QGROUP_RELATION_KEY, parentid)
+ * (parentid, BTRFS_QGROUP_RELATION_KEY, childid)
+ */
+#define BTRFS_QGROUP_RELATION_KEY       246
+
+#define BTRFS_BALANCE_ITEM_KEY	248
+
+/*
+ * Persistantly stores the io stats in the device tree.
+ * One key for all stats, (0, BTRFS_DEV_STATS_KEY, devid).
+ */
+#define BTRFS_DEV_STATS_KEY	249
+
+/*
+ * Persistantly stores the device replace state in the device tree.
+ * The key is built like this: (0, BTRFS_DEV_REPLACE_KEY, 0).
+ */
+#define BTRFS_DEV_REPLACE_KEY	250
+
+/*
+ * Stores items that allow to quickly map UUIDs to something else.
+ * These items are part of the filesystem UUID tree.
+ * The key is built like this:
+ * (UUID_upper_64_bits, BTRFS_UUID_KEY*, UUID_lower_64_bits).
+ */
+#if BTRFS_UUID_SIZE != 16
+#error "UUID items require BTRFS_UUID_SIZE == 16!"
+#endif
+#define BTRFS_UUID_KEY_SUBVOL	251	/* for UUIDs assigned to subvols */
+#define BTRFS_UUID_KEY_RECEIVED_SUBVOL	252	/* for UUIDs assigned to
+						 * received subvols */
+
+/*
  * string items are for debugging.  They just store a short string of
  * data in the FS
  */
 #define BTRFS_STRING_ITEM_KEY	253
 
+/*
+ * Flags for mount options.
+ *
+ * Note: don't forget to add new options to btrfs_show_options()
+ */
 #define BTRFS_MOUNT_NODATASUM		(1 << 0)
 #define BTRFS_MOUNT_NODATACOW		(1 << 1)
 #define BTRFS_MOUNT_NOBARRIER		(1 << 2)
@@ -1164,11 +2079,41 @@ struct btrfs_root {
 #define BTRFS_MOUNT_NOSSD		(1 << 9)
 #define BTRFS_MOUNT_DISCARD		(1 << 10)
 #define BTRFS_MOUNT_FORCE_COMPRESS      (1 << 11)
+#define BTRFS_MOUNT_SPACE_CACHE		(1 << 12)
+#define BTRFS_MOUNT_CLEAR_CACHE		(1 << 13)
+#define BTRFS_MOUNT_USER_SUBVOL_RM_ALLOWED (1 << 14)
+#define BTRFS_MOUNT_ENOSPC_DEBUG	 (1 << 15)
+#define BTRFS_MOUNT_AUTO_DEFRAG		(1 << 16)
+#define BTRFS_MOUNT_INODE_MAP_CACHE	(1 << 17)
+#define BTRFS_MOUNT_RECOVERY		(1 << 18)
+#define BTRFS_MOUNT_SKIP_BALANCE	(1 << 19)
+#define BTRFS_MOUNT_CHECK_INTEGRITY	(1 << 20)
+#define BTRFS_MOUNT_CHECK_INTEGRITY_INCLUDING_EXTENT_DATA (1 << 21)
+#define BTRFS_MOUNT_PANIC_ON_FATAL_ERROR	(1 << 22)
+#define BTRFS_MOUNT_RESCAN_UUID_TREE	(1 << 23)
+#define	BTRFS_MOUNT_CHANGE_INODE_CACHE	(1 << 24)
+
+#define BTRFS_DEFAULT_COMMIT_INTERVAL	(30)
 
 #define btrfs_clear_opt(o, opt)		((o) &= ~BTRFS_MOUNT_##opt)
 #define btrfs_set_opt(o, opt)		((o) |= BTRFS_MOUNT_##opt)
+#define btrfs_raw_test_opt(o, opt)	((o) & BTRFS_MOUNT_##opt)
 #define btrfs_test_opt(root, opt)	((root)->fs_info->mount_opt & \
 					 BTRFS_MOUNT_##opt)
+#define btrfs_set_and_info(root, opt, fmt, args...)			\
+{									\
+	if (!btrfs_test_opt(root, opt))					\
+		btrfs_info(root->fs_info, fmt, ##args);			\
+	btrfs_set_opt(root->fs_info->mount_opt, opt);			\
+}
+
+#define btrfs_clear_and_info(root, opt, fmt, args...)			\
+{									\
+	if (btrfs_test_opt(root, opt))					\
+		btrfs_info(root->fs_info, fmt, ##args);			\
+	btrfs_clear_opt(root->fs_info->mount_opt, opt);			\
+}
+
 /*
  * Inode flags
  */
@@ -1183,6 +2128,20 @@ struct btrfs_root {
 #define BTRFS_INODE_NODUMP		(1 << 8)
 #define BTRFS_INODE_NOATIME		(1 << 9)
 #define BTRFS_INODE_DIRSYNC		(1 << 10)
+#define BTRFS_INODE_COMPRESS		(1 << 11)
+
+#define BTRFS_INODE_ROOT_ITEM_INIT	(1 << 31)
+
+struct btrfs_map_token {
+	struct extent_buffer *eb;
+	char *kaddr;
+	unsigned long offset;
+};
+
+static inline void btrfs_init_map_token (struct btrfs_map_token *token)
+{
+	token->kaddr = NULL;
+}
 
 /* some macros to generate set/get funcs for the struct fields.  This
  * assumes there is a lefoo_to_cpu for every type, so lets make a simple
@@ -1204,26 +2163,67 @@ struct btrfs_root {
 			    offsetof(type, member),			\
 			   sizeof(((type *)0)->member)))
 
-#ifndef BTRFS_SETGET_FUNCS
+#define DECLARE_BTRFS_SETGET_BITS(bits)					\
+u##bits btrfs_get_token_##bits(struct extent_buffer *eb, void *ptr,	\
+			       unsigned long off,			\
+                              struct btrfs_map_token *token);		\
+void btrfs_set_token_##bits(struct extent_buffer *eb, void *ptr,	\
+			    unsigned long off, u##bits val,		\
+			    struct btrfs_map_token *token);		\
+static inline u##bits btrfs_get_##bits(struct extent_buffer *eb, void *ptr, \
+				       unsigned long off)		\
+{									\
+	return btrfs_get_token_##bits(eb, ptr, off, NULL);		\
+}									\
+static inline void btrfs_set_##bits(struct extent_buffer *eb, void *ptr, \
+				    unsigned long off, u##bits val)	\
+{									\
+       btrfs_set_token_##bits(eb, ptr, off, val, NULL);			\
+}
+
+DECLARE_BTRFS_SETGET_BITS(8)
+DECLARE_BTRFS_SETGET_BITS(16)
+DECLARE_BTRFS_SETGET_BITS(32)
+DECLARE_BTRFS_SETGET_BITS(64)
+
 #define BTRFS_SETGET_FUNCS(name, type, member, bits)			\
-u##bits btrfs_##name(struct extent_buffer *eb, type *s);		\
-void btrfs_set_##name(struct extent_buffer *eb, type *s, u##bits val);
-#endif
+static inline u##bits btrfs_##name(struct extent_buffer *eb, type *s)	\
+{									\
+	BUILD_BUG_ON(sizeof(u##bits) != sizeof(((type *)0))->member);	\
+	return btrfs_get_##bits(eb, s, offsetof(type, member));		\
+}									\
+static inline void btrfs_set_##name(struct extent_buffer *eb, type *s,	\
+				    u##bits val)			\
+{									\
+	BUILD_BUG_ON(sizeof(u##bits) != sizeof(((type *)0))->member);	\
+	btrfs_set_##bits(eb, s, offsetof(type, member), val);		\
+}									\
+static inline u##bits btrfs_token_##name(struct extent_buffer *eb, type *s, \
+					 struct btrfs_map_token *token)	\
+{									\
+	BUILD_BUG_ON(sizeof(u##bits) != sizeof(((type *)0))->member);	\
+	return btrfs_get_token_##bits(eb, s, offsetof(type, member), token); \
+}									\
+static inline void btrfs_set_token_##name(struct extent_buffer *eb,	\
+					  type *s, u##bits val,		\
+                                         struct btrfs_map_token *token)	\
+{									\
+	BUILD_BUG_ON(sizeof(u##bits) != sizeof(((type *)0))->member);	\
+	btrfs_set_token_##bits(eb, s, offsetof(type, member), val, token); \
+}
 
 #define BTRFS_SETGET_HEADER_FUNCS(name, type, member, bits)		\
 static inline u##bits btrfs_##name(struct extent_buffer *eb)		\
 {									\
-	type *p = kmap_atomic(eb->first_page, KM_USER0);		\
+	type *p = page_address(eb->pages[0]);				\
 	u##bits res = le##bits##_to_cpu(p->member);			\
-	kunmap_atomic(p, KM_USER0);					\
 	return res;							\
 }									\
 static inline void btrfs_set_##name(struct extent_buffer *eb,		\
 				    u##bits val)			\
 {									\
-	type *p = kmap_atomic(eb->first_page, KM_USER0);		\
+	type *p = page_address(eb->pages[0]);				\
 	p->member = cpu_to_le##bits(val);				\
-	kunmap_atomic(p, KM_USER0);					\
 }
 
 #define BTRFS_SETGET_STACK_FUNCS(name, type, member, bits)		\
@@ -1271,14 +2271,14 @@ BTRFS_SETGET_STACK_FUNCS(stack_device_bandwidth, struct btrfs_dev_item,
 BTRFS_SETGET_STACK_FUNCS(stack_device_generation, struct btrfs_dev_item,
 			 generation, 64);
 
-static inline char *btrfs_device_uuid(struct btrfs_dev_item *d)
+static inline unsigned long btrfs_device_uuid(struct btrfs_dev_item *d)
 {
-	return (char *)d + offsetof(struct btrfs_dev_item, uuid);
+	return (unsigned long)d + offsetof(struct btrfs_dev_item, uuid);
 }
 
-static inline char *btrfs_device_fsid(struct btrfs_dev_item *d)
+static inline unsigned long btrfs_device_fsid(struct btrfs_dev_item *d)
 {
-	return (char *)d + offsetof(struct btrfs_dev_item, fsid);
+	return (unsigned long)d + offsetof(struct btrfs_dev_item, fsid);
 }
 
 BTRFS_SETGET_FUNCS(chunk_length, struct btrfs_chunk, length, 64);
@@ -1336,26 +2336,12 @@ static inline u64 btrfs_stripe_offset_nr(struct extent_buffer *eb,
 	return btrfs_stripe_offset(eb, btrfs_stripe_nr(c, nr));
 }
 
-static inline void btrfs_set_stripe_offset_nr(struct extent_buffer *eb,
-					     struct btrfs_chunk *c, int nr,
-					     u64 val)
-{
-	btrfs_set_stripe_offset(eb, btrfs_stripe_nr(c, nr), val);
-}
-
 static inline u64 btrfs_stripe_devid_nr(struct extent_buffer *eb,
 					 struct btrfs_chunk *c, int nr)
 {
 	return btrfs_stripe_devid(eb, btrfs_stripe_nr(c, nr));
 }
 
-static inline void btrfs_set_stripe_devid_nr(struct extent_buffer *eb,
-					     struct btrfs_chunk *c, int nr,
-					     u64 val)
-{
-	btrfs_set_stripe_devid(eb, btrfs_stripe_nr(c, nr), val);
-}
-
 /* struct btrfs_block_group_item */
 BTRFS_SETGET_STACK_FUNCS(block_group_used, struct btrfs_block_group_item,
 			 used, 64);
@@ -1375,6 +2361,13 @@ BTRFS_SETGET_STACK_FUNCS(block_group_flags,
 BTRFS_SETGET_FUNCS(inode_ref_name_len, struct btrfs_inode_ref, name_len, 16);
 BTRFS_SETGET_FUNCS(inode_ref_index, struct btrfs_inode_ref, index, 64);
 
+/* struct btrfs_inode_extref */
+BTRFS_SETGET_FUNCS(inode_extref_parent, struct btrfs_inode_extref,
+		   parent_objectid, 64);
+BTRFS_SETGET_FUNCS(inode_extref_name_len, struct btrfs_inode_extref,
+		   name_len, 16);
+BTRFS_SETGET_FUNCS(inode_extref_index, struct btrfs_inode_extref, index, 64);
+
 /* struct btrfs_inode_item */
 BTRFS_SETGET_FUNCS(inode_generation, struct btrfs_inode_item, generation, 64);
 BTRFS_SETGET_FUNCS(inode_sequence, struct btrfs_inode_item, sequence, 64);
@@ -1388,6 +2381,23 @@ BTRFS_SETGET_FUNCS(inode_gid, struct btrfs_inode_item, gid, 32);
 BTRFS_SETGET_FUNCS(inode_mode, struct btrfs_inode_item, mode, 32);
 BTRFS_SETGET_FUNCS(inode_rdev, struct btrfs_inode_item, rdev, 64);
 BTRFS_SETGET_FUNCS(inode_flags, struct btrfs_inode_item, flags, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_inode_generation, struct btrfs_inode_item,
+			 generation, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_inode_sequence, struct btrfs_inode_item,
+			 sequence, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_inode_transid, struct btrfs_inode_item,
+			 transid, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_inode_size, struct btrfs_inode_item, size, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_inode_nbytes, struct btrfs_inode_item,
+			 nbytes, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_inode_block_group, struct btrfs_inode_item,
+			 block_group, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_inode_nlink, struct btrfs_inode_item, nlink, 32);
+BTRFS_SETGET_STACK_FUNCS(stack_inode_uid, struct btrfs_inode_item, uid, 32);
+BTRFS_SETGET_STACK_FUNCS(stack_inode_gid, struct btrfs_inode_item, gid, 32);
+BTRFS_SETGET_STACK_FUNCS(stack_inode_mode, struct btrfs_inode_item, mode, 32);
+BTRFS_SETGET_STACK_FUNCS(stack_inode_rdev, struct btrfs_inode_item, rdev, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_inode_flags, struct btrfs_inode_item, flags, 64);
 
 static inline struct btrfs_timespec *
 btrfs_inode_atime(struct btrfs_inode_item *inode_item)
@@ -1413,16 +2423,10 @@ btrfs_inode_ctime(struct btrfs_inode_item *inode_item)
 	return (struct btrfs_timespec *)ptr;
 }
 
-static inline struct btrfs_timespec *
-btrfs_inode_otime(struct btrfs_inode_item *inode_item)
-{
-	unsigned long ptr = (unsigned long)inode_item;
-	ptr += offsetof(struct btrfs_inode_item, otime);
-	return (struct btrfs_timespec *)ptr;
-}
-
 BTRFS_SETGET_FUNCS(timespec_sec, struct btrfs_timespec, sec, 64);
 BTRFS_SETGET_FUNCS(timespec_nsec, struct btrfs_timespec, nsec, 32);
+BTRFS_SETGET_STACK_FUNCS(stack_timespec_sec, struct btrfs_timespec, sec, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_timespec_nsec, struct btrfs_timespec, nsec, 32);
 
 /* struct btrfs_dev_extent */
 BTRFS_SETGET_FUNCS(dev_extent_chunk_tree, struct btrfs_dev_extent,
@@ -1433,10 +2437,10 @@ BTRFS_SETGET_FUNCS(dev_extent_chunk_offset, struct btrfs_dev_extent,
 		   chunk_offset, 64);
 BTRFS_SETGET_FUNCS(dev_extent_length, struct btrfs_dev_extent, length, 64);
 
-static inline u8 *btrfs_dev_extent_chunk_tree_uuid(struct btrfs_dev_extent *dev)
+static inline unsigned long btrfs_dev_extent_chunk_tree_uuid(struct btrfs_dev_extent *dev)
 {
 	unsigned long ptr = offsetof(struct btrfs_dev_extent, chunk_tree_uuid);
-	return (u8 *)((unsigned long)dev + ptr);
+	return (unsigned long)dev + ptr;
 }
 
 BTRFS_SETGET_FUNCS(extent_refs, struct btrfs_extent_item, refs, 64);
@@ -1504,6 +2508,10 @@ BTRFS_SETGET_FUNCS(ref_count_v0, struct btrfs_extent_ref_v0, count, 32);
 /* struct btrfs_node */
 BTRFS_SETGET_FUNCS(key_blockptr, struct btrfs_key_ptr, blockptr, 64);
 BTRFS_SETGET_FUNCS(key_generation, struct btrfs_key_ptr, generation, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_key_blockptr, struct btrfs_key_ptr,
+			 blockptr, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_key_generation, struct btrfs_key_ptr,
+			 generation, 64);
 
 static inline u64 btrfs_node_blockptr(struct extent_buffer *eb, int nr)
 {
@@ -1560,6 +2568,8 @@ static inline void btrfs_set_node_key(struct extent_buffer *eb,
 /* struct btrfs_item */
 BTRFS_SETGET_FUNCS(item_offset, struct btrfs_item, offset, 32);
 BTRFS_SETGET_FUNCS(item_size, struct btrfs_item, size, 32);
+BTRFS_SETGET_STACK_FUNCS(stack_item_offset, struct btrfs_item, offset, 32);
+BTRFS_SETGET_STACK_FUNCS(stack_item_size, struct btrfs_item, size, 32);
 
 static inline unsigned long btrfs_item_nr_offset(int nr)
 {
@@ -1567,8 +2577,7 @@ static inline unsigned long btrfs_item_nr_offset(int nr)
 		sizeof(struct btrfs_item) * nr;
 }
 
-static inline struct btrfs_item *btrfs_item_nr(struct extent_buffer *eb,
-					       int nr)
+static inline struct btrfs_item *btrfs_item_nr(int nr)
 {
 	return (struct btrfs_item *)btrfs_item_nr_offset(nr);
 }
@@ -1581,30 +2590,30 @@ static inline u32 btrfs_item_end(struct extent_buffer *eb,
 
 static inline u32 btrfs_item_end_nr(struct extent_buffer *eb, int nr)
 {
-	return btrfs_item_end(eb, btrfs_item_nr(eb, nr));
+	return btrfs_item_end(eb, btrfs_item_nr(nr));
 }
 
 static inline u32 btrfs_item_offset_nr(struct extent_buffer *eb, int nr)
 {
-	return btrfs_item_offset(eb, btrfs_item_nr(eb, nr));
+	return btrfs_item_offset(eb, btrfs_item_nr(nr));
 }
 
 static inline u32 btrfs_item_size_nr(struct extent_buffer *eb, int nr)
 {
-	return btrfs_item_size(eb, btrfs_item_nr(eb, nr));
+	return btrfs_item_size(eb, btrfs_item_nr(nr));
 }
 
 static inline void btrfs_item_key(struct extent_buffer *eb,
 			   struct btrfs_disk_key *disk_key, int nr)
 {
-	struct btrfs_item *item = btrfs_item_nr(eb, nr);
+	struct btrfs_item *item = btrfs_item_nr(nr);
 	read_eb_member(eb, item, struct btrfs_item, key, disk_key);
 }
 
 static inline void btrfs_set_item_key(struct extent_buffer *eb,
 			       struct btrfs_disk_key *disk_key, int nr)
 {
-	struct btrfs_item *item = btrfs_item_nr(eb, nr);
+	struct btrfs_item *item = btrfs_item_nr(nr);
 	write_eb_member(eb, item, struct btrfs_item, key, disk_key);
 }
 
@@ -1622,6 +2631,13 @@ BTRFS_SETGET_FUNCS(dir_data_len, struct btrfs_dir_item, data_len, 16);
 BTRFS_SETGET_FUNCS(dir_type, struct btrfs_dir_item, type, 8);
 BTRFS_SETGET_FUNCS(dir_name_len, struct btrfs_dir_item, name_len, 16);
 BTRFS_SETGET_FUNCS(dir_transid, struct btrfs_dir_item, transid, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_dir_type, struct btrfs_dir_item, type, 8);
+BTRFS_SETGET_STACK_FUNCS(stack_dir_data_len, struct btrfs_dir_item,
+			 data_len, 16);
+BTRFS_SETGET_STACK_FUNCS(stack_dir_name_len, struct btrfs_dir_item,
+			 name_len, 16);
+BTRFS_SETGET_STACK_FUNCS(stack_dir_transid, struct btrfs_dir_item,
+			 transid, 64);
 
 static inline void btrfs_dir_item_key(struct extent_buffer *eb,
 				      struct btrfs_dir_item *item,
@@ -1637,6 +2653,27 @@ static inline void btrfs_set_dir_item_key(struct extent_buffer *eb,
 	write_eb_member(eb, item, struct btrfs_dir_item, location, key);
 }
 
+BTRFS_SETGET_FUNCS(free_space_entries, struct btrfs_free_space_header,
+		   num_entries, 64);
+BTRFS_SETGET_FUNCS(free_space_bitmaps, struct btrfs_free_space_header,
+		   num_bitmaps, 64);
+BTRFS_SETGET_FUNCS(free_space_generation, struct btrfs_free_space_header,
+		   generation, 64);
+
+static inline void btrfs_free_space_key(struct extent_buffer *eb,
+					struct btrfs_free_space_header *h,
+					struct btrfs_disk_key *key)
+{
+	read_eb_member(eb, h, struct btrfs_free_space_header, location, key);
+}
+
+static inline void btrfs_set_free_space_key(struct extent_buffer *eb,
+					    struct btrfs_free_space_header *h,
+					    struct btrfs_disk_key *key)
+{
+	write_eb_member(eb, h, struct btrfs_free_space_header, location, key);
+}
+
 /* struct btrfs_disk_key */
 BTRFS_SETGET_STACK_FUNCS(disk_key_objectid, struct btrfs_disk_key,
 			 objectid, 64);
@@ -1703,6 +2740,12 @@ BTRFS_SETGET_HEADER_FUNCS(header_owner, struct btrfs_header, owner, 64);
 BTRFS_SETGET_HEADER_FUNCS(header_nritems, struct btrfs_header, nritems, 32);
 BTRFS_SETGET_HEADER_FUNCS(header_flags, struct btrfs_header, flags, 64);
 BTRFS_SETGET_HEADER_FUNCS(header_level, struct btrfs_header, level, 8);
+BTRFS_SETGET_STACK_FUNCS(stack_header_generation, struct btrfs_header,
+			 generation, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_header_owner, struct btrfs_header, owner, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_header_nritems, struct btrfs_header,
+			 nritems, 32);
+BTRFS_SETGET_STACK_FUNCS(stack_header_bytenr, struct btrfs_header, bytenr, 64);
 
 static inline int btrfs_header_flag(struct extent_buffer *eb, u64 flag)
 {
@@ -1738,43 +2781,14 @@ static inline void btrfs_set_header_backref_rev(struct extent_buffer *eb,
 	btrfs_set_header_flags(eb, flags);
 }
 
-static inline u8 *btrfs_header_fsid(struct extent_buffer *eb)
+static inline unsigned long btrfs_header_fsid(void)
 {
-	unsigned long ptr = offsetof(struct btrfs_header, fsid);
-	return (u8 *)ptr;
+	return offsetof(struct btrfs_header, fsid);
 }
 
-static inline u8 *btrfs_header_chunk_tree_uuid(struct extent_buffer *eb)
+static inline unsigned long btrfs_header_chunk_tree_uuid(struct extent_buffer *eb)
 {
-	unsigned long ptr = offsetof(struct btrfs_header, chunk_tree_uuid);
-	return (u8 *)ptr;
-}
-
-static inline u8 *btrfs_super_fsid(struct extent_buffer *eb)
-{
-	unsigned long ptr = offsetof(struct btrfs_super_block, fsid);
-	return (u8 *)ptr;
-}
-
-static inline u8 *btrfs_header_csum(struct extent_buffer *eb)
-{
-	unsigned long ptr = offsetof(struct btrfs_header, csum);
-	return (u8 *)ptr;
-}
-
-static inline struct btrfs_node *btrfs_buffer_node(struct extent_buffer *eb)
-{
-	return NULL;
-}
-
-static inline struct btrfs_leaf *btrfs_buffer_leaf(struct extent_buffer *eb)
-{
-	return NULL;
-}
-
-static inline struct btrfs_header *btrfs_buffer_header(struct extent_buffer *eb)
-{
-	return NULL;
+	return offsetof(struct btrfs_header, chunk_tree_uuid);
 }
 
 static inline int btrfs_is_leaf(struct extent_buffer *eb)
@@ -1800,9 +2814,158 @@ BTRFS_SETGET_STACK_FUNCS(root_used, struct btrfs_root_item, bytes_used, 64);
 BTRFS_SETGET_STACK_FUNCS(root_limit, struct btrfs_root_item, byte_limit, 64);
 BTRFS_SETGET_STACK_FUNCS(root_last_snapshot, struct btrfs_root_item,
 			 last_snapshot, 64);
+BTRFS_SETGET_STACK_FUNCS(root_generation_v2, struct btrfs_root_item,
+			 generation_v2, 64);
+BTRFS_SETGET_STACK_FUNCS(root_ctransid, struct btrfs_root_item,
+			 ctransid, 64);
+BTRFS_SETGET_STACK_FUNCS(root_otransid, struct btrfs_root_item,
+			 otransid, 64);
+BTRFS_SETGET_STACK_FUNCS(root_stransid, struct btrfs_root_item,
+			 stransid, 64);
+BTRFS_SETGET_STACK_FUNCS(root_rtransid, struct btrfs_root_item,
+			 rtransid, 64);
+
+static inline bool btrfs_root_readonly(struct btrfs_root *root)
+{
+	return (root->root_item.flags & cpu_to_le64(BTRFS_ROOT_SUBVOL_RDONLY)) != 0;
+}
 
-/* struct btrfs_super_block */
+static inline bool btrfs_root_dead(struct btrfs_root *root)
+{
+	return (root->root_item.flags & cpu_to_le64(BTRFS_ROOT_SUBVOL_DEAD)) != 0;
+}
+
+/* struct btrfs_root_backup */
+BTRFS_SETGET_STACK_FUNCS(backup_tree_root, struct btrfs_root_backup,
+		   tree_root, 64);
+BTRFS_SETGET_STACK_FUNCS(backup_tree_root_gen, struct btrfs_root_backup,
+		   tree_root_gen, 64);
+BTRFS_SETGET_STACK_FUNCS(backup_tree_root_level, struct btrfs_root_backup,
+		   tree_root_level, 8);
+
+BTRFS_SETGET_STACK_FUNCS(backup_chunk_root, struct btrfs_root_backup,
+		   chunk_root, 64);
+BTRFS_SETGET_STACK_FUNCS(backup_chunk_root_gen, struct btrfs_root_backup,
+		   chunk_root_gen, 64);
+BTRFS_SETGET_STACK_FUNCS(backup_chunk_root_level, struct btrfs_root_backup,
+		   chunk_root_level, 8);
+
+BTRFS_SETGET_STACK_FUNCS(backup_extent_root, struct btrfs_root_backup,
+		   extent_root, 64);
+BTRFS_SETGET_STACK_FUNCS(backup_extent_root_gen, struct btrfs_root_backup,
+		   extent_root_gen, 64);
+BTRFS_SETGET_STACK_FUNCS(backup_extent_root_level, struct btrfs_root_backup,
+		   extent_root_level, 8);
+
+BTRFS_SETGET_STACK_FUNCS(backup_fs_root, struct btrfs_root_backup,
+		   fs_root, 64);
+BTRFS_SETGET_STACK_FUNCS(backup_fs_root_gen, struct btrfs_root_backup,
+		   fs_root_gen, 64);
+BTRFS_SETGET_STACK_FUNCS(backup_fs_root_level, struct btrfs_root_backup,
+		   fs_root_level, 8);
+
+BTRFS_SETGET_STACK_FUNCS(backup_dev_root, struct btrfs_root_backup,
+		   dev_root, 64);
+BTRFS_SETGET_STACK_FUNCS(backup_dev_root_gen, struct btrfs_root_backup,
+		   dev_root_gen, 64);
+BTRFS_SETGET_STACK_FUNCS(backup_dev_root_level, struct btrfs_root_backup,
+		   dev_root_level, 8);
+
+BTRFS_SETGET_STACK_FUNCS(backup_csum_root, struct btrfs_root_backup,
+		   csum_root, 64);
+BTRFS_SETGET_STACK_FUNCS(backup_csum_root_gen, struct btrfs_root_backup,
+		   csum_root_gen, 64);
+BTRFS_SETGET_STACK_FUNCS(backup_csum_root_level, struct btrfs_root_backup,
+		   csum_root_level, 8);
+BTRFS_SETGET_STACK_FUNCS(backup_total_bytes, struct btrfs_root_backup,
+		   total_bytes, 64);
+BTRFS_SETGET_STACK_FUNCS(backup_bytes_used, struct btrfs_root_backup,
+		   bytes_used, 64);
+BTRFS_SETGET_STACK_FUNCS(backup_num_devices, struct btrfs_root_backup,
+		   num_devices, 64);
+
+/* struct btrfs_balance_item */
+BTRFS_SETGET_FUNCS(balance_flags, struct btrfs_balance_item, flags, 64);
+
+static inline void btrfs_balance_data(struct extent_buffer *eb,
+				      struct btrfs_balance_item *bi,
+				      struct btrfs_disk_balance_args *ba)
+{
+	read_eb_member(eb, bi, struct btrfs_balance_item, data, ba);
+}
+
+static inline void btrfs_set_balance_data(struct extent_buffer *eb,
+					  struct btrfs_balance_item *bi,
+					  struct btrfs_disk_balance_args *ba)
+{
+	write_eb_member(eb, bi, struct btrfs_balance_item, data, ba);
+}
+
+static inline void btrfs_balance_meta(struct extent_buffer *eb,
+				      struct btrfs_balance_item *bi,
+				      struct btrfs_disk_balance_args *ba)
+{
+	read_eb_member(eb, bi, struct btrfs_balance_item, meta, ba);
+}
+
+static inline void btrfs_set_balance_meta(struct extent_buffer *eb,
+					  struct btrfs_balance_item *bi,
+					  struct btrfs_disk_balance_args *ba)
+{
+	write_eb_member(eb, bi, struct btrfs_balance_item, meta, ba);
+}
+
+static inline void btrfs_balance_sys(struct extent_buffer *eb,
+				     struct btrfs_balance_item *bi,
+				     struct btrfs_disk_balance_args *ba)
+{
+	read_eb_member(eb, bi, struct btrfs_balance_item, sys, ba);
+}
+
+static inline void btrfs_set_balance_sys(struct extent_buffer *eb,
+					 struct btrfs_balance_item *bi,
+					 struct btrfs_disk_balance_args *ba)
+{
+	write_eb_member(eb, bi, struct btrfs_balance_item, sys, ba);
+}
+
+static inline void
+btrfs_disk_balance_args_to_cpu(struct btrfs_balance_args *cpu,
+			       struct btrfs_disk_balance_args *disk)
+{
+	memset(cpu, 0, sizeof(*cpu));
+
+	cpu->profiles = le64_to_cpu(disk->profiles);
+	cpu->usage = le64_to_cpu(disk->usage);
+	cpu->devid = le64_to_cpu(disk->devid);
+	cpu->pstart = le64_to_cpu(disk->pstart);
+	cpu->pend = le64_to_cpu(disk->pend);
+	cpu->vstart = le64_to_cpu(disk->vstart);
+	cpu->vend = le64_to_cpu(disk->vend);
+	cpu->target = le64_to_cpu(disk->target);
+	cpu->flags = le64_to_cpu(disk->flags);
+	cpu->limit = le64_to_cpu(disk->limit);
+}
+
+static inline void
+btrfs_cpu_balance_args_to_disk(struct btrfs_disk_balance_args *disk,
+			       struct btrfs_balance_args *cpu)
+{
+	memset(disk, 0, sizeof(*disk));
+
+	disk->profiles = cpu_to_le64(cpu->profiles);
+	disk->usage = cpu_to_le64(cpu->usage);
+	disk->devid = cpu_to_le64(cpu->devid);
+	disk->pstart = cpu_to_le64(cpu->pstart);
+	disk->pend = cpu_to_le64(cpu->pend);
+	disk->vstart = cpu_to_le64(cpu->vstart);
+	disk->vend = cpu_to_le64(cpu->vend);
+	disk->target = cpu_to_le64(cpu->target);
+	disk->flags = cpu_to_le64(cpu->flags);
+	disk->limit = cpu_to_le64(cpu->limit);
+}
 
+/* struct btrfs_super_block */
 BTRFS_SETGET_STACK_FUNCS(super_bytenr, struct btrfs_super_block, bytenr, 64);
 BTRFS_SETGET_STACK_FUNCS(super_flags, struct btrfs_super_block, flags, 64);
 BTRFS_SETGET_STACK_FUNCS(super_generation, struct btrfs_super_block,
@@ -1848,11 +3011,18 @@ BTRFS_SETGET_STACK_FUNCS(super_incompat_flags, struct btrfs_super_block,
 			 incompat_flags, 64);
 BTRFS_SETGET_STACK_FUNCS(super_csum_type, struct btrfs_super_block,
 			 csum_type, 16);
+BTRFS_SETGET_STACK_FUNCS(super_cache_generation, struct btrfs_super_block,
+			 cache_generation, 64);
+BTRFS_SETGET_STACK_FUNCS(super_magic, struct btrfs_super_block, magic, 64);
+BTRFS_SETGET_STACK_FUNCS(super_uuid_tree_generation, struct btrfs_super_block,
+			 uuid_tree_generation, 64);
 
 static inline int btrfs_super_csum_size(struct btrfs_super_block *s)
 {
-	int t = btrfs_super_csum_type(s);
-	BUG_ON(t >= ARRAY_SIZE(btrfs_csum_sizes));
+	u16 t = btrfs_super_csum_type(s);
+	/*
+	 * csum type is validated at mount time
+	 */
 	return btrfs_csum_sizes[t];
 }
 
@@ -1863,6 +3033,18 @@ static inline unsigned long btrfs_leaf_data(struct extent_buffer *l)
 
 /* struct btrfs_file_extent_item */
 BTRFS_SETGET_FUNCS(file_extent_type, struct btrfs_file_extent_item, type, 8);
+BTRFS_SETGET_STACK_FUNCS(stack_file_extent_disk_bytenr,
+			 struct btrfs_file_extent_item, disk_bytenr, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_file_extent_offset,
+			 struct btrfs_file_extent_item, offset, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_file_extent_generation,
+			 struct btrfs_file_extent_item, generation, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_file_extent_num_bytes,
+			 struct btrfs_file_extent_item, num_bytes, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_file_extent_disk_num_bytes,
+			 struct btrfs_file_extent_item, disk_num_bytes, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_file_extent_compression,
+			 struct btrfs_file_extent_item, compression, 8);
 
 static inline unsigned long
 btrfs_file_extent_inline_start(struct btrfs_file_extent_item *e)
@@ -1896,15 +3078,6 @@ BTRFS_SETGET_FUNCS(file_extent_encryption, struct btrfs_file_extent_item,
 BTRFS_SETGET_FUNCS(file_extent_other_encoding, struct btrfs_file_extent_item,
 		   other_encoding, 16);
 
-/* this returns the number of file bytes represented by the inline item.
- * If an item is compressed, this is the uncompressed size
- */
-static inline u32 btrfs_file_extent_inline_len(struct extent_buffer *eb,
-					       struct btrfs_file_extent_item *e)
-{
-	return btrfs_file_extent_ram_bytes(eb, e);
-}
-
 /*
  * this returns the number of bytes used by the item on disk, minus the
  * size of any extent headers.  If a file is compressed on disk, this is
@@ -1918,25 +3091,145 @@ static inline u32 btrfs_file_extent_inline_item_len(struct extent_buffer *eb,
 	return btrfs_item_size(eb, e) - offset;
 }
 
-static inline struct btrfs_root *btrfs_sb(struct super_block *sb)
+/* this returns the number of file bytes represented by the inline item.
+ * If an item is compressed, this is the uncompressed size
+ */
+static inline u32 btrfs_file_extent_inline_len(struct extent_buffer *eb,
+					       int slot,
+					       struct btrfs_file_extent_item *fi)
 {
-	return sb->s_fs_info;
+	struct btrfs_map_token token;
+
+	btrfs_init_map_token(&token);
+	/*
+	 * return the space used on disk if this item isn't
+	 * compressed or encoded
+	 */
+	if (btrfs_token_file_extent_compression(eb, fi, &token) == 0 &&
+	    btrfs_token_file_extent_encryption(eb, fi, &token) == 0 &&
+	    btrfs_token_file_extent_other_encoding(eb, fi, &token) == 0) {
+		return btrfs_file_extent_inline_item_len(eb,
+							 btrfs_item_nr(slot));
+	}
+
+	/* otherwise use the ram bytes field */
+	return btrfs_token_file_extent_ram_bytes(eb, fi, &token);
 }
 
-static inline int btrfs_set_root_name(struct btrfs_root *root,
-				      const char *name, int len)
+
+/* btrfs_dev_stats_item */
+static inline u64 btrfs_dev_stats_value(struct extent_buffer *eb,
+					struct btrfs_dev_stats_item *ptr,
+					int index)
 {
-	/* if we already have a name just free it */
-	kfree(root->name);
+	u64 val;
 
-	root->name = kmalloc(len+1, GFP_KERNEL);
-	if (!root->name)
-		return -ENOMEM;
+	read_extent_buffer(eb, &val,
+			   offsetof(struct btrfs_dev_stats_item, values) +
+			    ((unsigned long)ptr) + (index * sizeof(u64)),
+			   sizeof(val));
+	return val;
+}
 
-	memcpy(root->name, name, len);
-	root->name[len] = '\0';
+static inline void btrfs_set_dev_stats_value(struct extent_buffer *eb,
+					     struct btrfs_dev_stats_item *ptr,
+					     int index, u64 val)
+{
+	write_extent_buffer(eb, &val,
+			    offsetof(struct btrfs_dev_stats_item, values) +
+			     ((unsigned long)ptr) + (index * sizeof(u64)),
+			    sizeof(val));
+}
 
-	return 0;
+/* btrfs_qgroup_status_item */
+BTRFS_SETGET_FUNCS(qgroup_status_generation, struct btrfs_qgroup_status_item,
+		   generation, 64);
+BTRFS_SETGET_FUNCS(qgroup_status_version, struct btrfs_qgroup_status_item,
+		   version, 64);
+BTRFS_SETGET_FUNCS(qgroup_status_flags, struct btrfs_qgroup_status_item,
+		   flags, 64);
+BTRFS_SETGET_FUNCS(qgroup_status_rescan, struct btrfs_qgroup_status_item,
+		   rescan, 64);
+
+/* btrfs_qgroup_info_item */
+BTRFS_SETGET_FUNCS(qgroup_info_generation, struct btrfs_qgroup_info_item,
+		   generation, 64);
+BTRFS_SETGET_FUNCS(qgroup_info_rfer, struct btrfs_qgroup_info_item, rfer, 64);
+BTRFS_SETGET_FUNCS(qgroup_info_rfer_cmpr, struct btrfs_qgroup_info_item,
+		   rfer_cmpr, 64);
+BTRFS_SETGET_FUNCS(qgroup_info_excl, struct btrfs_qgroup_info_item, excl, 64);
+BTRFS_SETGET_FUNCS(qgroup_info_excl_cmpr, struct btrfs_qgroup_info_item,
+		   excl_cmpr, 64);
+
+BTRFS_SETGET_STACK_FUNCS(stack_qgroup_info_generation,
+			 struct btrfs_qgroup_info_item, generation, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_qgroup_info_rfer, struct btrfs_qgroup_info_item,
+			 rfer, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_qgroup_info_rfer_cmpr,
+			 struct btrfs_qgroup_info_item, rfer_cmpr, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_qgroup_info_excl, struct btrfs_qgroup_info_item,
+			 excl, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_qgroup_info_excl_cmpr,
+			 struct btrfs_qgroup_info_item, excl_cmpr, 64);
+
+/* btrfs_qgroup_limit_item */
+BTRFS_SETGET_FUNCS(qgroup_limit_flags, struct btrfs_qgroup_limit_item,
+		   flags, 64);
+BTRFS_SETGET_FUNCS(qgroup_limit_max_rfer, struct btrfs_qgroup_limit_item,
+		   max_rfer, 64);
+BTRFS_SETGET_FUNCS(qgroup_limit_max_excl, struct btrfs_qgroup_limit_item,
+		   max_excl, 64);
+BTRFS_SETGET_FUNCS(qgroup_limit_rsv_rfer, struct btrfs_qgroup_limit_item,
+		   rsv_rfer, 64);
+BTRFS_SETGET_FUNCS(qgroup_limit_rsv_excl, struct btrfs_qgroup_limit_item,
+		   rsv_excl, 64);
+
+/* btrfs_dev_replace_item */
+BTRFS_SETGET_FUNCS(dev_replace_src_devid,
+		   struct btrfs_dev_replace_item, src_devid, 64);
+BTRFS_SETGET_FUNCS(dev_replace_cont_reading_from_srcdev_mode,
+		   struct btrfs_dev_replace_item, cont_reading_from_srcdev_mode,
+		   64);
+BTRFS_SETGET_FUNCS(dev_replace_replace_state, struct btrfs_dev_replace_item,
+		   replace_state, 64);
+BTRFS_SETGET_FUNCS(dev_replace_time_started, struct btrfs_dev_replace_item,
+		   time_started, 64);
+BTRFS_SETGET_FUNCS(dev_replace_time_stopped, struct btrfs_dev_replace_item,
+		   time_stopped, 64);
+BTRFS_SETGET_FUNCS(dev_replace_num_write_errors, struct btrfs_dev_replace_item,
+		   num_write_errors, 64);
+BTRFS_SETGET_FUNCS(dev_replace_num_uncorrectable_read_errors,
+		   struct btrfs_dev_replace_item, num_uncorrectable_read_errors,
+		   64);
+BTRFS_SETGET_FUNCS(dev_replace_cursor_left, struct btrfs_dev_replace_item,
+		   cursor_left, 64);
+BTRFS_SETGET_FUNCS(dev_replace_cursor_right, struct btrfs_dev_replace_item,
+		   cursor_right, 64);
+
+BTRFS_SETGET_STACK_FUNCS(stack_dev_replace_src_devid,
+			 struct btrfs_dev_replace_item, src_devid, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_dev_replace_cont_reading_from_srcdev_mode,
+			 struct btrfs_dev_replace_item,
+			 cont_reading_from_srcdev_mode, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_dev_replace_replace_state,
+			 struct btrfs_dev_replace_item, replace_state, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_dev_replace_time_started,
+			 struct btrfs_dev_replace_item, time_started, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_dev_replace_time_stopped,
+			 struct btrfs_dev_replace_item, time_stopped, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_dev_replace_num_write_errors,
+			 struct btrfs_dev_replace_item, num_write_errors, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_dev_replace_num_uncorrectable_read_errors,
+			 struct btrfs_dev_replace_item,
+			 num_uncorrectable_read_errors, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_dev_replace_cursor_left,
+			 struct btrfs_dev_replace_item, cursor_left, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_dev_replace_cursor_right,
+			 struct btrfs_dev_replace_item, cursor_right, 64);
+
+static inline struct btrfs_fs_info *btrfs_sb(struct super_block *sb)
+{
+	return sb->s_fs_info;
 }
 
 static inline u32 btrfs_level_size(struct btrfs_root *root, int level)
@@ -1955,43 +3248,72 @@ static inline u32 btrfs_level_size(struct btrfs_root *root, int level)
 	((unsigned long)(btrfs_leaf_data(leaf) + \
 	btrfs_item_offset_nr(leaf, slot)))
 
-static inline struct dentry *fdentry(struct file *file)
+static inline bool btrfs_mixed_space_info(struct btrfs_space_info *space_info)
 {
-	return file->f_path.dentry;
+	return ((space_info->flags & BTRFS_BLOCK_GROUP_METADATA) &&
+		(space_info->flags & BTRFS_BLOCK_GROUP_DATA));
+}
+
+static inline gfp_t btrfs_alloc_write_mask(struct address_space *mapping)
+{
+	return mapping_gfp_mask(mapping) & ~__GFP_FS;
 }
 
 /* extent-tree.c */
+static inline u64 btrfs_calc_trans_metadata_size(struct btrfs_root *root,
+						 unsigned num_items)
+{
+	return (root->leafsize + root->nodesize * (BTRFS_MAX_LEVEL - 1)) *
+		2 * num_items;
+}
+
+/*
+ * Doing a truncate won't result in new nodes or leaves, just what we need for
+ * COW.
+ */
+static inline u64 btrfs_calc_trunc_metadata_size(struct btrfs_root *root,
+						 unsigned num_items)
+{
+	return (root->leafsize + root->nodesize * (BTRFS_MAX_LEVEL - 1)) *
+		num_items;
+}
+
+int btrfs_should_throttle_delayed_refs(struct btrfs_trans_handle *trans,
+				       struct btrfs_root *root);
+int btrfs_check_space_for_delayed_refs(struct btrfs_trans_handle *trans,
+				       struct btrfs_root *root);
 void btrfs_put_block_group(struct btrfs_block_group_cache *cache);
 int btrfs_run_delayed_refs(struct btrfs_trans_handle *trans,
 			   struct btrfs_root *root, unsigned long count);
+int btrfs_async_run_delayed_refs(struct btrfs_root *root,
+				 unsigned long count, int wait);
 int btrfs_lookup_extent(struct btrfs_root *root, u64 start, u64 len);
+int btrfs_lookup_extent_info(struct btrfs_trans_handle *trans,
+			     struct btrfs_root *root, u64 bytenr,
+			     u64 offset, int metadata, u64 *refs, u64 *flags);
 int btrfs_pin_extent(struct btrfs_root *root,
 		     u64 bytenr, u64 num, int reserved);
-int btrfs_drop_leaf_ref(struct btrfs_trans_handle *trans,
-			struct btrfs_root *root, struct extent_buffer *leaf);
+int btrfs_pin_extent_for_log_replay(struct btrfs_root *root,
+				    u64 bytenr, u64 num_bytes);
+int btrfs_exclude_logged_extents(struct btrfs_root *root,
+				 struct extent_buffer *eb);
 int btrfs_cross_ref_exist(struct btrfs_trans_handle *trans,
 			  struct btrfs_root *root,
 			  u64 objectid, u64 offset, u64 bytenr);
-int btrfs_copy_pinned(struct btrfs_root *root, struct extent_io_tree *copy);
 struct btrfs_block_group_cache *btrfs_lookup_block_group(
 						 struct btrfs_fs_info *info,
 						 u64 bytenr);
 void btrfs_put_block_group(struct btrfs_block_group_cache *cache);
-u64 btrfs_find_block_group(struct btrfs_root *root,
-			   u64 search_start, u64 search_hint, int owner);
+int get_block_group_index(struct btrfs_block_group_cache *cache);
 struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
 					struct btrfs_root *root, u32 blocksize,
 					u64 parent, u64 root_objectid,
 					struct btrfs_disk_key *key, int level,
 					u64 hint, u64 empty_size);
-int btrfs_free_tree_block(struct btrfs_trans_handle *trans,
-			  struct btrfs_root *root,
-			  u64 bytenr, u32 blocksize,
-			  u64 parent, u64 root_objectid, int level);
-struct extent_buffer *btrfs_init_new_buffer(struct btrfs_trans_handle *trans,
-					    struct btrfs_root *root,
-					    u64 bytenr, u32 blocksize,
-					    int level);
+void btrfs_free_tree_block(struct btrfs_trans_handle *trans,
+			   struct btrfs_root *root,
+			   struct extent_buffer *buf,
+			   u64 parent, int last_ref);
 int btrfs_alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
 				     struct btrfs_root *root,
 				     u64 root_objectid, u64 owner,
@@ -2000,34 +3322,34 @@ int btrfs_alloc_logged_file_extent(struct btrfs_trans_handle *trans,
 				   struct btrfs_root *root,
 				   u64 root_objectid, u64 owner, u64 offset,
 				   struct btrfs_key *ins);
-int btrfs_reserve_extent(struct btrfs_trans_handle *trans,
-				  struct btrfs_root *root,
-				  u64 num_bytes, u64 min_alloc_size,
-				  u64 empty_size, u64 hint_byte,
-				  u64 search_end, struct btrfs_key *ins,
-				  u64 data);
+int btrfs_reserve_extent(struct btrfs_root *root, u64 num_bytes,
+			 u64 min_alloc_size, u64 empty_size, u64 hint_byte,
+			 struct btrfs_key *ins, int is_data, int delalloc);
 int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
-		  struct extent_buffer *buf, int full_backref);
+		  struct extent_buffer *buf, int full_backref, int no_quota);
 int btrfs_dec_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
-		  struct extent_buffer *buf, int full_backref);
+		  struct extent_buffer *buf, int full_backref, int no_quota);
 int btrfs_set_disk_extent_flags(struct btrfs_trans_handle *trans,
 				struct btrfs_root *root,
 				u64 bytenr, u64 num_bytes, u64 flags,
-				int is_data);
+				int level, int is_data);
 int btrfs_free_extent(struct btrfs_trans_handle *trans,
 		      struct btrfs_root *root,
-		      u64 bytenr, u64 num_bytes, u64 parent,
-		      u64 root_objectid, u64 owner, u64 offset);
-
-int btrfs_free_reserved_extent(struct btrfs_root *root, u64 start, u64 len);
-int btrfs_prepare_extent_commit(struct btrfs_trans_handle *trans,
-				struct btrfs_root *root);
+		      u64 bytenr, u64 num_bytes, u64 parent, u64 root_objectid,
+		      u64 owner, u64 offset, int no_quota);
+
+int btrfs_free_reserved_extent(struct btrfs_root *root, u64 start, u64 len,
+			       int delalloc);
+int btrfs_free_and_pin_reserved_extent(struct btrfs_root *root,
+				       u64 start, u64 len);
+void btrfs_prepare_extent_commit(struct btrfs_trans_handle *trans,
+				 struct btrfs_root *root);
 int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans,
 			       struct btrfs_root *root);
 int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
 			 struct btrfs_root *root,
 			 u64 bytenr, u64 num_bytes, u64 parent,
-			 u64 root_objectid, u64 owner, u64 offset);
+			 u64 root_objectid, u64 owner, u64 offset, int no_quota);
 
 int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans,
 				    struct btrfs_root *root);
@@ -2041,27 +3363,82 @@ int btrfs_make_block_group(struct btrfs_trans_handle *trans,
 			   u64 size);
 int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
 			     struct btrfs_root *root, u64 group_start);
-int btrfs_prepare_block_group_relocation(struct btrfs_root *root,
-				struct btrfs_block_group_cache *group);
-
-u64 btrfs_reduce_alloc_profile(struct btrfs_root *root, u64 flags);
-void btrfs_set_inode_space_info(struct btrfs_root *root, struct inode *ionde);
+void btrfs_create_pending_block_groups(struct btrfs_trans_handle *trans,
+				       struct btrfs_root *root);
+u64 btrfs_get_alloc_profile(struct btrfs_root *root, int data);
 void btrfs_clear_space_info_full(struct btrfs_fs_info *info);
 
-int btrfs_reserve_metadata_space(struct btrfs_root *root, int num_items);
-int btrfs_unreserve_metadata_space(struct btrfs_root *root, int num_items);
-int btrfs_unreserve_metadata_for_delalloc(struct btrfs_root *root,
-					  struct inode *inode, int num_items);
-int btrfs_reserve_metadata_for_delalloc(struct btrfs_root *root,
-					struct inode *inode, int num_items);
-int btrfs_check_data_free_space(struct btrfs_root *root, struct inode *inode,
-				u64 bytes);
-void btrfs_free_reserved_data_space(struct btrfs_root *root,
-				    struct inode *inode, u64 bytes);
-void btrfs_delalloc_reserve_space(struct btrfs_root *root, struct inode *inode,
-				 u64 bytes);
-void btrfs_delalloc_free_space(struct btrfs_root *root, struct inode *inode,
-			      u64 bytes);
+enum btrfs_reserve_flush_enum {
+	/* If we are in the transaction, we can't flush anything.*/
+	BTRFS_RESERVE_NO_FLUSH,
+	/*
+	 * Flushing delalloc may cause deadlock somewhere, in this
+	 * case, use FLUSH LIMIT
+	 */
+	BTRFS_RESERVE_FLUSH_LIMIT,
+	BTRFS_RESERVE_FLUSH_ALL,
+};
+
+int btrfs_check_data_free_space(struct inode *inode, u64 bytes);
+void btrfs_free_reserved_data_space(struct inode *inode, u64 bytes);
+void btrfs_trans_release_metadata(struct btrfs_trans_handle *trans,
+				struct btrfs_root *root);
+int btrfs_orphan_reserve_metadata(struct btrfs_trans_handle *trans,
+				  struct inode *inode);
+void btrfs_orphan_release_metadata(struct inode *inode);
+int btrfs_subvolume_reserve_metadata(struct btrfs_root *root,
+				     struct btrfs_block_rsv *rsv,
+				     int nitems,
+				     u64 *qgroup_reserved, bool use_global_rsv);
+void btrfs_subvolume_release_metadata(struct btrfs_root *root,
+				      struct btrfs_block_rsv *rsv,
+				      u64 qgroup_reserved);
+int btrfs_delalloc_reserve_metadata(struct inode *inode, u64 num_bytes);
+void btrfs_delalloc_release_metadata(struct inode *inode, u64 num_bytes);
+int btrfs_delalloc_reserve_space(struct inode *inode, u64 num_bytes);
+void btrfs_delalloc_release_space(struct inode *inode, u64 num_bytes);
+void btrfs_init_block_rsv(struct btrfs_block_rsv *rsv, unsigned short type);
+struct btrfs_block_rsv *btrfs_alloc_block_rsv(struct btrfs_root *root,
+					      unsigned short type);
+void btrfs_free_block_rsv(struct btrfs_root *root,
+			  struct btrfs_block_rsv *rsv);
+int btrfs_block_rsv_add(struct btrfs_root *root,
+			struct btrfs_block_rsv *block_rsv, u64 num_bytes,
+			enum btrfs_reserve_flush_enum flush);
+int btrfs_block_rsv_check(struct btrfs_root *root,
+			  struct btrfs_block_rsv *block_rsv, int min_factor);
+int btrfs_block_rsv_refill(struct btrfs_root *root,
+			   struct btrfs_block_rsv *block_rsv, u64 min_reserved,
+			   enum btrfs_reserve_flush_enum flush);
+int btrfs_block_rsv_migrate(struct btrfs_block_rsv *src_rsv,
+			    struct btrfs_block_rsv *dst_rsv,
+			    u64 num_bytes);
+int btrfs_cond_migrate_bytes(struct btrfs_fs_info *fs_info,
+			     struct btrfs_block_rsv *dest, u64 num_bytes,
+			     int min_factor);
+void btrfs_block_rsv_release(struct btrfs_root *root,
+			     struct btrfs_block_rsv *block_rsv,
+			     u64 num_bytes);
+int btrfs_set_block_group_ro(struct btrfs_root *root,
+			     struct btrfs_block_group_cache *cache);
+void btrfs_set_block_group_rw(struct btrfs_root *root,
+			      struct btrfs_block_group_cache *cache);
+void btrfs_put_block_group_cache(struct btrfs_fs_info *info);
+u64 btrfs_account_ro_block_groups_free_space(struct btrfs_space_info *sinfo);
+int btrfs_error_unpin_extent_range(struct btrfs_root *root,
+				   u64 start, u64 end);
+int btrfs_error_discard_extent(struct btrfs_root *root, u64 bytenr,
+			       u64 num_bytes, u64 *actual_bytes);
+int btrfs_force_chunk_alloc(struct btrfs_trans_handle *trans,
+			    struct btrfs_root *root, u64 type);
+int btrfs_trim_fs(struct btrfs_root *root, struct fstrim_range *range);
+
+int btrfs_init_space_info(struct btrfs_fs_info *fs_info);
+int btrfs_delayed_refs_qgroup_accounting(struct btrfs_trans_handle *trans,
+					 struct btrfs_fs_info *fs_info);
+int __get_raid_index(u64 flags);
+int btrfs_start_nocow_write(struct btrfs_root *root);
+void btrfs_end_nocow_write(struct btrfs_root *root);
 /* ctree.c */
 int btrfs_bin_search(struct extent_buffer *eb, struct btrfs_key *key,
 		     int level, int *slot);
@@ -2069,18 +3446,34 @@ int btrfs_comp_cpu_keys(struct btrfs_key *k1, struct btrfs_key *k2);
 int btrfs_previous_item(struct btrfs_root *root,
 			struct btrfs_path *path, u64 min_objectid,
 			int type);
-int btrfs_set_item_key_safe(struct btrfs_trans_handle *trans,
-			    struct btrfs_root *root, struct btrfs_path *path,
-			    struct btrfs_key *new_key);
+int btrfs_previous_extent_item(struct btrfs_root *root,
+			struct btrfs_path *path, u64 min_objectid);
+void btrfs_set_item_key_safe(struct btrfs_root *root, struct btrfs_path *path,
+			     struct btrfs_key *new_key);
 struct extent_buffer *btrfs_root_node(struct btrfs_root *root);
 struct extent_buffer *btrfs_lock_root_node(struct btrfs_root *root);
 int btrfs_find_next_key(struct btrfs_root *root, struct btrfs_path *path,
 			struct btrfs_key *key, int lowest_level,
-			int cache_only, u64 min_trans);
+			u64 min_trans);
 int btrfs_search_forward(struct btrfs_root *root, struct btrfs_key *min_key,
-			 struct btrfs_key *max_key,
-			 struct btrfs_path *path, int cache_only,
+			 struct btrfs_path *path,
 			 u64 min_trans);
+enum btrfs_compare_tree_result {
+	BTRFS_COMPARE_TREE_NEW,
+	BTRFS_COMPARE_TREE_DELETED,
+	BTRFS_COMPARE_TREE_CHANGED,
+	BTRFS_COMPARE_TREE_SAME,
+};
+typedef int (*btrfs_changed_cb_t)(struct btrfs_root *left_root,
+				  struct btrfs_root *right_root,
+				  struct btrfs_path *left_path,
+				  struct btrfs_path *right_path,
+				  struct btrfs_key *key,
+				  enum btrfs_compare_tree_result result,
+				  void *ctx);
+int btrfs_compare_trees(struct btrfs_root *left_root,
+			struct btrfs_root *right_root,
+			btrfs_changed_cb_t cb, void *ctx);
 int btrfs_cow_block(struct btrfs_trans_handle *trans,
 		    struct btrfs_root *root, struct extent_buffer *buf,
 		    struct extent_buffer *parent, int parent_slot,
@@ -2091,12 +3484,10 @@ int btrfs_copy_root(struct btrfs_trans_handle *trans,
 		      struct extent_buffer **cow_ret, u64 new_root_objectid);
 int btrfs_block_can_be_shared(struct btrfs_root *root,
 			      struct extent_buffer *buf);
-int btrfs_extend_item(struct btrfs_trans_handle *trans, struct btrfs_root
-		      *root, struct btrfs_path *path, u32 data_size);
-int btrfs_truncate_item(struct btrfs_trans_handle *trans,
-			struct btrfs_root *root,
-			struct btrfs_path *path,
-			u32 new_size, int from_end);
+void btrfs_extend_item(struct btrfs_root *root, struct btrfs_path *path,
+		       u32 data_size);
+void btrfs_truncate_item(struct btrfs_root *root, struct btrfs_path *path,
+			 u32 new_size, int from_end);
 int btrfs_split_item(struct btrfs_trans_handle *trans,
 		     struct btrfs_root *root,
 		     struct btrfs_path *path,
@@ -2106,17 +3497,26 @@ int btrfs_duplicate_item(struct btrfs_trans_handle *trans,
 			 struct btrfs_root *root,
 			 struct btrfs_path *path,
 			 struct btrfs_key *new_key);
+int btrfs_find_item(struct btrfs_root *fs_root, struct btrfs_path *path,
+		u64 inum, u64 ioff, u8 key_type, struct btrfs_key *found_key);
 int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root
 		      *root, struct btrfs_key *key, struct btrfs_path *p, int
 		      ins_len, int cow);
+int btrfs_search_old_slot(struct btrfs_root *root, struct btrfs_key *key,
+			  struct btrfs_path *p, u64 time_seq);
+int btrfs_search_slot_for_read(struct btrfs_root *root,
+			       struct btrfs_key *key, struct btrfs_path *p,
+			       int find_higher, int return_any);
 int btrfs_realloc_node(struct btrfs_trans_handle *trans,
 		       struct btrfs_root *root, struct extent_buffer *parent,
-		       int start_slot, int cache_only, u64 *last_ret,
+		       int start_slot, u64 *last_ret,
 		       struct btrfs_key *progress);
-void btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p);
+void btrfs_release_path(struct btrfs_path *p);
 struct btrfs_path *btrfs_alloc_path(void);
 void btrfs_free_path(struct btrfs_path *p);
 void btrfs_set_path_blocking(struct btrfs_path *p);
+void btrfs_clear_path_blocking(struct btrfs_path *p,
+			       struct extent_buffer *held, int held_rw);
 void btrfs_unlock_up_safe(struct btrfs_path *p, int level);
 
 int btrfs_del_items(struct btrfs_trans_handle *trans, struct btrfs_root *root,
@@ -2128,13 +3528,11 @@ static inline int btrfs_del_item(struct btrfs_trans_handle *trans,
 	return btrfs_del_items(trans, root, path, path->slots[0], 1);
 }
 
+void setup_items_for_insert(struct btrfs_root *root, struct btrfs_path *path,
+			    struct btrfs_key *cpu_key, u32 *data_size,
+			    u32 total_data, u32 total_size, int nr);
 int btrfs_insert_item(struct btrfs_trans_handle *trans, struct btrfs_root
 		      *root, struct btrfs_key *key, void *data, u32 data_size);
-int btrfs_insert_some_items(struct btrfs_trans_handle *trans,
-			    struct btrfs_root *root,
-			    struct btrfs_path *path,
-			    struct btrfs_key *cpu_key, u32 *data_size,
-			    int nr);
 int btrfs_insert_empty_items(struct btrfs_trans_handle *trans,
 			     struct btrfs_root *root,
 			     struct btrfs_path *path,
@@ -2151,12 +3549,71 @@ static inline int btrfs_insert_empty_item(struct btrfs_trans_handle *trans,
 
 int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path);
 int btrfs_prev_leaf(struct btrfs_root *root, struct btrfs_path *path);
+int btrfs_next_old_leaf(struct btrfs_root *root, struct btrfs_path *path,
+			u64 time_seq);
+static inline int btrfs_next_old_item(struct btrfs_root *root,
+				      struct btrfs_path *p, u64 time_seq)
+{
+	++p->slots[0];
+	if (p->slots[0] >= btrfs_header_nritems(p->nodes[0]))
+		return btrfs_next_old_leaf(root, p, time_seq);
+	return 0;
+}
+static inline int btrfs_next_item(struct btrfs_root *root, struct btrfs_path *p)
+{
+	return btrfs_next_old_item(root, p, 0);
+}
 int btrfs_leaf_free_space(struct btrfs_root *root, struct extent_buffer *leaf);
-int btrfs_drop_snapshot(struct btrfs_root *root, int update_ref);
+int __must_check btrfs_drop_snapshot(struct btrfs_root *root,
+				     struct btrfs_block_rsv *block_rsv,
+				     int update_ref, int for_reloc);
 int btrfs_drop_subtree(struct btrfs_trans_handle *trans,
 			struct btrfs_root *root,
 			struct extent_buffer *node,
 			struct extent_buffer *parent);
+static inline int btrfs_fs_closing(struct btrfs_fs_info *fs_info)
+{
+	/*
+	 * Get synced with close_ctree()
+	 */
+	smp_mb();
+	return fs_info->closing;
+}
+
+/*
+ * If we remount the fs to be R/O or umount the fs, the cleaner needn't do
+ * anything except sleeping. This function is used to check the status of
+ * the fs.
+ */
+static inline int btrfs_need_cleaner_sleep(struct btrfs_root *root)
+{
+	return (root->fs_info->sb->s_flags & MS_RDONLY ||
+		btrfs_fs_closing(root->fs_info));
+}
+
+static inline void free_fs_info(struct btrfs_fs_info *fs_info)
+{
+	kfree(fs_info->balance_ctl);
+	kfree(fs_info->delayed_root);
+	kfree(fs_info->extent_root);
+	kfree(fs_info->tree_root);
+	kfree(fs_info->chunk_root);
+	kfree(fs_info->dev_root);
+	kfree(fs_info->csum_root);
+	kfree(fs_info->quota_root);
+	kfree(fs_info->uuid_root);
+	kfree(fs_info->super_copy);
+	kfree(fs_info->super_for_commit);
+	kfree(fs_info);
+}
+
+/* tree mod log functions from ctree.c */
+u64 btrfs_get_tree_mod_seq(struct btrfs_fs_info *fs_info,
+			   struct seq_list *elem);
+void btrfs_put_tree_mod_seq(struct btrfs_fs_info *fs_info,
+			    struct seq_list *elem);
+int btrfs_old_root_level(struct btrfs_root *root, u64 time_seq);
+
 /* root-item.c */
 int btrfs_find_root_ref(struct btrfs_root *tree_root,
 			struct btrfs_path *path,
@@ -2174,21 +3631,37 @@ int btrfs_del_root(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 int btrfs_insert_root(struct btrfs_trans_handle *trans, struct btrfs_root
 		      *root, struct btrfs_key *key, struct btrfs_root_item
 		      *item);
-int btrfs_update_root(struct btrfs_trans_handle *trans, struct btrfs_root
-		      *root, struct btrfs_key *key, struct btrfs_root_item
-		      *item);
-int btrfs_find_last_root(struct btrfs_root *root, u64 objectid, struct
-			 btrfs_root_item *item, struct btrfs_key *key);
-int btrfs_search_root(struct btrfs_root *root, u64 search_start,
-		      u64 *found_objectid);
-int btrfs_find_dead_roots(struct btrfs_root *root, u64 objectid);
+int __must_check btrfs_update_root(struct btrfs_trans_handle *trans,
+				   struct btrfs_root *root,
+				   struct btrfs_key *key,
+				   struct btrfs_root_item *item);
+int btrfs_find_root(struct btrfs_root *root, struct btrfs_key *search_key,
+		    struct btrfs_path *path, struct btrfs_root_item *root_item,
+		    struct btrfs_key *root_key);
 int btrfs_find_orphan_roots(struct btrfs_root *tree_root);
-int btrfs_set_root_node(struct btrfs_root_item *item,
-			struct extent_buffer *node);
+void btrfs_set_root_node(struct btrfs_root_item *item,
+			 struct extent_buffer *node);
+void btrfs_check_and_init_root_item(struct btrfs_root_item *item);
+void btrfs_update_root_times(struct btrfs_trans_handle *trans,
+			     struct btrfs_root *root);
+
+/* uuid-tree.c */
+int btrfs_uuid_tree_add(struct btrfs_trans_handle *trans,
+			struct btrfs_root *uuid_root, u8 *uuid, u8 type,
+			u64 subid);
+int btrfs_uuid_tree_rem(struct btrfs_trans_handle *trans,
+			struct btrfs_root *uuid_root, u8 *uuid, u8 type,
+			u64 subid);
+int btrfs_uuid_tree_iterate(struct btrfs_fs_info *fs_info,
+			    int (*check_func)(struct btrfs_fs_info *, u8 *, u8,
+					      u64));
+
 /* dir-item.c */
+int btrfs_check_dir_item_collision(struct btrfs_root *root, u64 dir,
+			  const char *name, int name_len);
 int btrfs_insert_dir_item(struct btrfs_trans_handle *trans,
 			  struct btrfs_root *root, const char *name,
-			  int name_len, u64 dir,
+			  int name_len, struct inode *dir,
 			  struct btrfs_key *location, u8 type, u64 index);
 struct btrfs_dir_item *btrfs_lookup_dir_item(struct btrfs_trans_handle *trans,
 					     struct btrfs_root *root,
@@ -2205,9 +3678,6 @@ struct btrfs_dir_item *
 btrfs_search_dir_index_item(struct btrfs_root *root,
 			    struct btrfs_path *path, u64 dirid,
 			    const char *name, int name_len);
-struct btrfs_dir_item *btrfs_match_dir_item_name(struct btrfs_root *root,
-			      struct btrfs_path *path,
-			      const char *name, int name_len);
 int btrfs_delete_one_dir_name(struct btrfs_trans_handle *trans,
 			      struct btrfs_root *root,
 			      struct btrfs_path *path,
@@ -2222,6 +3692,9 @@ struct btrfs_dir_item *btrfs_lookup_xattr(struct btrfs_trans_handle *trans,
 					  struct btrfs_path *path, u64 dir,
 					  const char *name, u16 name_len,
 					  int mod);
+int verify_dir_item(struct btrfs_root *root,
+		    struct extent_buffer *leaf,
+		    struct btrfs_dir_item *dir_item);
 
 /* orphan.c */
 int btrfs_insert_orphan_item(struct btrfs_trans_handle *trans,
@@ -2230,12 +3703,6 @@ int btrfs_del_orphan_item(struct btrfs_trans_handle *trans,
 			  struct btrfs_root *root, u64 offset);
 int btrfs_find_orphan_item(struct btrfs_root *root, u64 offset);
 
-/* inode-map.c */
-int btrfs_find_free_objectid(struct btrfs_trans_handle *trans,
-			     struct btrfs_root *fs_root,
-			     u64 dirid, u64 *objectid);
-int btrfs_find_highest_inode(struct btrfs_root *fs_root, u64 *objectid);
-
 /* inode-item.c */
 int btrfs_insert_inode_ref(struct btrfs_trans_handle *trans,
 			   struct btrfs_root *root,
@@ -2252,11 +3719,28 @@ int btrfs_lookup_inode(struct btrfs_trans_handle *trans, struct btrfs_root
 		       *root, struct btrfs_path *path,
 		       struct btrfs_key *location, int mod);
 
+struct btrfs_inode_extref *
+btrfs_lookup_inode_extref(struct btrfs_trans_handle *trans,
+			  struct btrfs_root *root,
+			  struct btrfs_path *path,
+			  const char *name, int name_len,
+			  u64 inode_objectid, u64 ref_objectid, int ins_len,
+			  int cow);
+
+int btrfs_find_name_in_ext_backref(struct btrfs_path *path,
+				   u64 ref_objectid, const char *name,
+				   int name_len,
+				   struct btrfs_inode_extref **extref_ret);
+
 /* file-item.c */
+struct btrfs_dio_private;
 int btrfs_del_csums(struct btrfs_trans_handle *trans,
 		    struct btrfs_root *root, u64 bytenr, u64 len);
 int btrfs_lookup_bio_sums(struct btrfs_root *root, struct inode *inode,
 			  struct bio *bio, u32 *dst);
+int btrfs_lookup_bio_sums_dio(struct btrfs_root *root, struct inode *inode,
+			      struct btrfs_dio_private *dip, struct bio *bio,
+			      u64 logical_offset);
 int btrfs_insert_file_extent(struct btrfs_trans_handle *trans,
 			     struct btrfs_root *root,
 			     u64 objectid, u64 pos,
@@ -2272,18 +3756,34 @@ int btrfs_csum_file_blocks(struct btrfs_trans_handle *trans,
 			   struct btrfs_ordered_sum *sums);
 int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode,
 		       struct bio *bio, u64 file_start, int contig);
-int btrfs_csum_file_bytes(struct btrfs_root *root, struct inode *inode,
-			  u64 start, unsigned long len);
-struct btrfs_csum_item *btrfs_lookup_csum(struct btrfs_trans_handle *trans,
-					  struct btrfs_root *root,
-					  struct btrfs_path *path,
-					  u64 bytenr, int cow);
-int btrfs_csum_truncate(struct btrfs_trans_handle *trans,
-			struct btrfs_root *root, struct btrfs_path *path,
-			u64 isize);
-int btrfs_lookup_csums_range(struct btrfs_root *root, u64 start,
-			     u64 end, struct list_head *list);
+int btrfs_lookup_csums_range(struct btrfs_root *root, u64 start, u64 end,
+			     struct list_head *list, int search_commit);
+void btrfs_extent_item_to_extent_map(struct inode *inode,
+				     const struct btrfs_path *path,
+				     struct btrfs_file_extent_item *fi,
+				     const bool new_inline,
+				     struct extent_map *em);
+
 /* inode.c */
+struct btrfs_delalloc_work {
+	struct inode *inode;
+	int wait;
+	int delay_iput;
+	struct completion completion;
+	struct list_head list;
+	struct btrfs_work work;
+};
+
+struct btrfs_delalloc_work *btrfs_alloc_delalloc_work(struct inode *inode,
+						    int wait, int delay_iput);
+void btrfs_wait_and_free_delalloc_work(struct btrfs_delalloc_work *work);
+
+struct extent_map *btrfs_get_extent_fiemap(struct inode *inode, struct page *page,
+					   size_t pg_offset, u64 start, u64 len,
+					   int create);
+noinline int can_nocow_extent(struct inode *inode, u64 offset, u64 *len,
+			      u64 *orig_start, u64 *orig_block_len,
+			      u64 *ram_bytes);
 
 /* RHEL and EL kernels have a patch that renames PG_checked to FsMisc */
 #if defined(ClearPageFsMisc) && !defined(ClearPageChecked)
@@ -2292,6 +3792,14 @@ int btrfs_lookup_csums_range(struct btrfs_root *root, u64 start,
 #define PageChecked PageFsMisc
 #endif
 
+/* This forces readahead on a given range of bytes in an inode */
+static inline void btrfs_force_ra(struct address_space *mapping,
+				  struct file_ra_state *ra, struct file *file,
+				  pgoff_t offset, unsigned long req_size)
+{
+	page_cache_sync_readahead(mapping, ra, file, offset, req_size);
+}
+
 struct inode *btrfs_lookup_dentry(struct inode *dir, struct dentry *dentry);
 int btrfs_set_inode_index(struct inode *dir, u64 *index);
 int btrfs_unlink_inode(struct btrfs_trans_handle *trans,
@@ -2305,84 +3813,114 @@ int btrfs_unlink_subvol(struct btrfs_trans_handle *trans,
 			struct btrfs_root *root,
 			struct inode *dir, u64 objectid,
 			const char *name, int name_len);
+int btrfs_truncate_page(struct inode *inode, loff_t from, loff_t len,
+			int front);
 int btrfs_truncate_inode_items(struct btrfs_trans_handle *trans,
 			       struct btrfs_root *root,
 			       struct inode *inode, u64 new_size,
 			       u32 min_type);
 
 int btrfs_start_delalloc_inodes(struct btrfs_root *root, int delay_iput);
+int btrfs_start_delalloc_roots(struct btrfs_fs_info *fs_info, int delay_iput,
+			       int nr);
 int btrfs_set_extent_delalloc(struct inode *inode, u64 start, u64 end,
 			      struct extent_state **cached_state);
-int btrfs_writepages(struct address_space *mapping,
-		     struct writeback_control *wbc);
 int btrfs_create_subvol_root(struct btrfs_trans_handle *trans,
 			     struct btrfs_root *new_root,
-			     u64 new_dirid, u64 alloc_hint);
-int btrfs_merge_bio_hook(struct page *page, unsigned long offset,
-			 size_t size, struct bio *bio, unsigned long bio_flags);
-
-unsigned long btrfs_force_ra(struct address_space *mapping,
-			      struct file_ra_state *ra, struct file *file,
-			      pgoff_t offset, pgoff_t last_index);
+			     struct btrfs_root *parent_root,
+			     u64 new_dirid);
+int btrfs_merge_bio_hook(int rw, struct page *page, unsigned long offset,
+			 size_t size, struct bio *bio,
+			 unsigned long bio_flags);
 int btrfs_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
 int btrfs_readpage(struct file *file, struct page *page);
-void btrfs_delete_inode(struct inode *inode);
-void btrfs_put_inode(struct inode *inode);
+void btrfs_evict_inode(struct inode *inode);
 int btrfs_write_inode(struct inode *inode, struct writeback_control *wbc);
-void btrfs_dirty_inode(struct inode *inode);
 struct inode *btrfs_alloc_inode(struct super_block *sb);
 void btrfs_destroy_inode(struct inode *inode);
-void btrfs_drop_inode(struct inode *inode);
+int btrfs_drop_inode(struct inode *inode);
 int btrfs_init_cachep(void);
 void btrfs_destroy_cachep(void);
 long btrfs_ioctl_trans_end(struct file *file);
 struct inode *btrfs_iget(struct super_block *s, struct btrfs_key *location,
 			 struct btrfs_root *root, int *was_new);
-int btrfs_commit_write(struct file *file, struct page *page,
-		       unsigned from, unsigned to);
 struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
-				    size_t page_offset, u64 start, u64 end,
+				    size_t pg_offset, u64 start, u64 end,
 				    int create);
 int btrfs_update_inode(struct btrfs_trans_handle *trans,
 			      struct btrfs_root *root,
 			      struct inode *inode);
+int btrfs_update_inode_fallback(struct btrfs_trans_handle *trans,
+				struct btrfs_root *root, struct inode *inode);
 int btrfs_orphan_add(struct btrfs_trans_handle *trans, struct inode *inode);
-int btrfs_orphan_del(struct btrfs_trans_handle *trans, struct inode *inode);
-void btrfs_orphan_cleanup(struct btrfs_root *root);
-int btrfs_cont_expand(struct inode *inode, loff_t size);
-int btrfs_invalidate_inodes(struct btrfs_root *root);
+int btrfs_orphan_cleanup(struct btrfs_root *root);
+void btrfs_orphan_commit_root(struct btrfs_trans_handle *trans,
+			      struct btrfs_root *root);
+int btrfs_cont_expand(struct inode *inode, loff_t oldsize, loff_t size);
+void btrfs_invalidate_inodes(struct btrfs_root *root);
 void btrfs_add_delayed_iput(struct inode *inode);
 void btrfs_run_delayed_iputs(struct btrfs_root *root);
+int btrfs_prealloc_file_range(struct inode *inode, int mode,
+			      u64 start, u64 num_bytes, u64 min_size,
+			      loff_t actual_len, u64 *alloc_hint);
+int btrfs_prealloc_file_range_trans(struct inode *inode,
+				    struct btrfs_trans_handle *trans, int mode,
+				    u64 start, u64 num_bytes, u64 min_size,
+				    loff_t actual_len, u64 *alloc_hint);
 extern const struct dentry_operations btrfs_dentry_operations;
 
 /* ioctl.c */
 long btrfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
 void btrfs_update_iflags(struct inode *inode);
 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir);
+int btrfs_is_empty_uuid(u8 *uuid);
+int btrfs_defrag_file(struct inode *inode, struct file *file,
+		      struct btrfs_ioctl_defrag_range_args *range,
+		      u64 newer_than, unsigned long max_pages);
+void btrfs_get_block_group_info(struct list_head *groups_list,
+				struct btrfs_ioctl_space_info *space);
+void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
+			       struct btrfs_ioctl_balance_args *bargs);
+
 
 /* file.c */
-int btrfs_sync_file(struct file *file, struct dentry *dentry, int datasync);
-int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end,
-			    int skip_pinned);
-int btrfs_check_file(struct btrfs_root *root, struct inode *inode);
+int btrfs_auto_defrag_init(void);
+void btrfs_auto_defrag_exit(void);
+int btrfs_add_inode_defrag(struct btrfs_trans_handle *trans,
+			   struct inode *inode);
+int btrfs_run_defrag_inodes(struct btrfs_fs_info *fs_info);
+void btrfs_cleanup_defrag_inodes(struct btrfs_fs_info *fs_info);
+int btrfs_sync_file(struct file *file, loff_t start, loff_t end, int datasync);
+void btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end,
+			     int skip_pinned);
 extern const struct file_operations btrfs_file_operations;
-int btrfs_drop_extents(struct btrfs_trans_handle *trans, struct inode *inode,
-		       u64 start, u64 end, u64 *hint_byte, int drop_cache);
+int __btrfs_drop_extents(struct btrfs_trans_handle *trans,
+			 struct btrfs_root *root, struct inode *inode,
+			 struct btrfs_path *path, u64 start, u64 end,
+			 u64 *drop_end, int drop_cache,
+			 int replace_extent,
+			 u32 extent_item_size,
+			 int *key_inserted);
+int btrfs_drop_extents(struct btrfs_trans_handle *trans,
+		       struct btrfs_root *root, struct inode *inode, u64 start,
+		       u64 end, int drop_cache);
 int btrfs_mark_extent_written(struct btrfs_trans_handle *trans,
 			      struct inode *inode, u64 start, u64 end);
 int btrfs_release_file(struct inode *inode, struct file *file);
+int btrfs_dirty_pages(struct btrfs_root *root, struct inode *inode,
+		      struct page **pages, size_t num_pages,
+		      loff_t pos, size_t write_bytes,
+		      struct extent_state **cached);
 
 /* tree-defrag.c */
 int btrfs_defrag_leaves(struct btrfs_trans_handle *trans,
-			struct btrfs_root *root, int cache_only);
+			struct btrfs_root *root);
 
 /* sysfs.c */
 int btrfs_init_sysfs(void);
 void btrfs_exit_sysfs(void);
-int btrfs_sysfs_add_super(struct btrfs_fs_info *fs);
-int btrfs_sysfs_add_root(struct btrfs_root *root);
-void btrfs_sysfs_del_root(struct btrfs_root *root);
-void btrfs_sysfs_del_super(struct btrfs_fs_info *root);
+int btrfs_sysfs_add_one(struct btrfs_fs_info *fs_info);
+void btrfs_sysfs_remove_one(struct btrfs_fs_info *fs_info);
 
 /* xattr.c */
 ssize_t btrfs_listxattr(struct dentry *dentry, char *buffer, size_t size);
@@ -2391,15 +3929,151 @@ ssize_t btrfs_listxattr(struct dentry *dentry, char *buffer, size_t size);
 int btrfs_parse_options(struct btrfs_root *root, char *options);
 int btrfs_sync_fs(struct super_block *sb, int wait);
 
-/* acl.c */
-#ifdef CONFIG_BTRFS_FS_POSIX_ACL
-int btrfs_check_acl(struct inode *inode, int mask);
+#ifdef CONFIG_PRINTK
+__printf(2, 3)
+void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...);
 #else
-#define btrfs_check_acl NULL
+static inline __printf(2, 3)
+void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
+{
+}
 #endif
+
+#define btrfs_emerg(fs_info, fmt, args...) \
+	btrfs_printk(fs_info, KERN_EMERG fmt, ##args)
+#define btrfs_alert(fs_info, fmt, args...) \
+	btrfs_printk(fs_info, KERN_ALERT fmt, ##args)
+#define btrfs_crit(fs_info, fmt, args...) \
+	btrfs_printk(fs_info, KERN_CRIT fmt, ##args)
+#define btrfs_err(fs_info, fmt, args...) \
+	btrfs_printk(fs_info, KERN_ERR fmt, ##args)
+#define btrfs_warn(fs_info, fmt, args...) \
+	btrfs_printk(fs_info, KERN_WARNING fmt, ##args)
+#define btrfs_notice(fs_info, fmt, args...) \
+	btrfs_printk(fs_info, KERN_NOTICE fmt, ##args)
+#define btrfs_info(fs_info, fmt, args...) \
+	btrfs_printk(fs_info, KERN_INFO fmt, ##args)
+
+#ifdef DEBUG
+#define btrfs_debug(fs_info, fmt, args...) \
+	btrfs_printk(fs_info, KERN_DEBUG fmt, ##args)
+#else
+#define btrfs_debug(fs_info, fmt, args...) \
+    no_printk(KERN_DEBUG fmt, ##args)
+#endif
+
+#ifdef CONFIG_BTRFS_ASSERT
+
+static inline void assfail(char *expr, char *file, int line)
+{
+	pr_err("BTRFS: assertion failed: %s, file: %s, line: %d",
+	       expr, file, line);
+	BUG();
+}
+
+#define ASSERT(expr)	\
+	(likely(expr) ? (void)0 : assfail(#expr, __FILE__, __LINE__))
+#else
+#define ASSERT(expr)	((void)0)
+#endif
+
+#define btrfs_assert()
+__printf(5, 6)
+void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
+		     unsigned int line, int errno, const char *fmt, ...);
+
+
+void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
+			       struct btrfs_root *root, const char *function,
+			       unsigned int line, int errno);
+
+#define btrfs_set_fs_incompat(__fs_info, opt) \
+	__btrfs_set_fs_incompat((__fs_info), BTRFS_FEATURE_INCOMPAT_##opt)
+
+static inline void __btrfs_set_fs_incompat(struct btrfs_fs_info *fs_info,
+					   u64 flag)
+{
+	struct btrfs_super_block *disk_super;
+	u64 features;
+
+	disk_super = fs_info->super_copy;
+	features = btrfs_super_incompat_flags(disk_super);
+	if (!(features & flag)) {
+		spin_lock(&fs_info->super_lock);
+		features = btrfs_super_incompat_flags(disk_super);
+		if (!(features & flag)) {
+			features |= flag;
+			btrfs_set_super_incompat_flags(disk_super, features);
+			btrfs_info(fs_info, "setting %llu feature flag",
+					 flag);
+		}
+		spin_unlock(&fs_info->super_lock);
+	}
+}
+
+#define btrfs_fs_incompat(fs_info, opt) \
+	__btrfs_fs_incompat((fs_info), BTRFS_FEATURE_INCOMPAT_##opt)
+
+static inline int __btrfs_fs_incompat(struct btrfs_fs_info *fs_info, u64 flag)
+{
+	struct btrfs_super_block *disk_super;
+	disk_super = fs_info->super_copy;
+	return !!(btrfs_super_incompat_flags(disk_super) & flag);
+}
+
+/*
+ * Call btrfs_abort_transaction as early as possible when an error condition is
+ * detected, that way the exact line number is reported.
+ */
+
+#define btrfs_abort_transaction(trans, root, errno)		\
+do {								\
+	__btrfs_abort_transaction(trans, root, __func__,	\
+				  __LINE__, errno);		\
+} while (0)
+
+#define btrfs_std_error(fs_info, errno)				\
+do {								\
+	if ((errno))						\
+		__btrfs_std_error((fs_info), __func__,		\
+				   __LINE__, (errno), NULL);	\
+} while (0)
+
+#define btrfs_error(fs_info, errno, fmt, args...)		\
+do {								\
+	__btrfs_std_error((fs_info), __func__, __LINE__,	\
+			  (errno), fmt, ##args);		\
+} while (0)
+
+__printf(5, 6)
+void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
+		   unsigned int line, int errno, const char *fmt, ...);
+
+/*
+ * If BTRFS_MOUNT_PANIC_ON_FATAL_ERROR is in mount_opt, __btrfs_panic
+ * will panic().  Otherwise we BUG() here.
+ */
+#define btrfs_panic(fs_info, errno, fmt, args...)			\
+do {									\
+	__btrfs_panic(fs_info, __func__, __LINE__, errno, fmt, ##args);	\
+	BUG();								\
+} while (0)
+
+/* acl.c */
+#ifdef CONFIG_BTRFS_FS_POSIX_ACL
+struct posix_acl *btrfs_get_acl(struct inode *inode, int type);
+int btrfs_set_acl(struct inode *inode, struct posix_acl *acl, int type);
 int btrfs_init_acl(struct btrfs_trans_handle *trans,
 		   struct inode *inode, struct inode *dir);
-int btrfs_acl_chmod(struct inode *inode);
+#else
+#define btrfs_get_acl NULL
+#define btrfs_set_acl NULL
+static inline int btrfs_init_acl(struct btrfs_trans_handle *trans,
+				 struct inode *inode, struct inode *dir)
+{
+	return 0;
+}
+#endif
 
 /* relocation.c */
 int btrfs_relocate_block_group(struct btrfs_root *root, u64 group_start);
@@ -2409,4 +4083,66 @@ int btrfs_update_reloc_root(struct btrfs_trans_handle *trans,
 			    struct btrfs_root *root);
 int btrfs_recover_relocation(struct btrfs_root *root);
 int btrfs_reloc_clone_csums(struct inode *inode, u64 file_pos, u64 len);
+int btrfs_reloc_cow_block(struct btrfs_trans_handle *trans,
+			  struct btrfs_root *root, struct extent_buffer *buf,
+			  struct extent_buffer *cow);
+void btrfs_reloc_pre_snapshot(struct btrfs_trans_handle *trans,
+			      struct btrfs_pending_snapshot *pending,
+			      u64 *bytes_to_reserve);
+int btrfs_reloc_post_snapshot(struct btrfs_trans_handle *trans,
+			      struct btrfs_pending_snapshot *pending);
+
+/* scrub.c */
+int btrfs_scrub_dev(struct btrfs_fs_info *fs_info, u64 devid, u64 start,
+		    u64 end, struct btrfs_scrub_progress *progress,
+		    int readonly, int is_dev_replace);
+void btrfs_scrub_pause(struct btrfs_root *root);
+void btrfs_scrub_continue(struct btrfs_root *root);
+int btrfs_scrub_cancel(struct btrfs_fs_info *info);
+int btrfs_scrub_cancel_dev(struct btrfs_fs_info *info,
+			   struct btrfs_device *dev);
+int btrfs_scrub_progress(struct btrfs_root *root, u64 devid,
+			 struct btrfs_scrub_progress *progress);
+
+/* dev-replace.c */
+void btrfs_bio_counter_inc_blocked(struct btrfs_fs_info *fs_info);
+void btrfs_bio_counter_inc_noblocked(struct btrfs_fs_info *fs_info);
+void btrfs_bio_counter_dec(struct btrfs_fs_info *fs_info);
+
+/* reada.c */
+struct reada_control {
+	struct btrfs_root	*root;		/* tree to prefetch */
+	struct btrfs_key	key_start;
+	struct btrfs_key	key_end;	/* exclusive */
+	atomic_t		elems;
+	struct kref		refcnt;
+	wait_queue_head_t	wait;
+};
+struct reada_control *btrfs_reada_add(struct btrfs_root *root,
+			      struct btrfs_key *start, struct btrfs_key *end);
+int btrfs_reada_wait(void *handle);
+void btrfs_reada_detach(void *handle);
+int btree_readahead_hook(struct btrfs_root *root, struct extent_buffer *eb,
+			 u64 start, int err);
+
+static inline int is_fstree(u64 rootid)
+{
+	if (rootid == BTRFS_FS_TREE_OBJECTID ||
+	    (s64)rootid >= (s64)BTRFS_FIRST_FREE_OBJECTID)
+		return 1;
+	return 0;
+}
+
+static inline int btrfs_defrag_cancelled(struct btrfs_fs_info *fs_info)
+{
+	return signal_pending(current);
+}
+
+/* Sanity test specific functions */
+#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
+void btrfs_test_destroy_inode(struct inode *inode);
+int btrfs_verify_qgroup_counts(struct btrfs_fs_info *fs_info, u64 qgroupid,
+			       u64 rfer, u64 excl);
+#endif
+
 #endif
diff --git a/fs/btrfs/delayed-inode.c b/fs/btrfs/delayed-inode.c
new file mode 100644
index 00000000000..da775bfdebc
--- /dev/null
+++ b/fs/btrfs/delayed-inode.c
@@ -0,0 +1,1980 @@
+/*
+ * Copyright (C) 2011 Fujitsu.  All rights reserved.
+ * Written by Miao Xie <miaox@cn.fujitsu.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/slab.h>
+#include "delayed-inode.h"
+#include "disk-io.h"
+#include "transaction.h"
+#include "ctree.h"
+
+#define BTRFS_DELAYED_WRITEBACK		512
+#define BTRFS_DELAYED_BACKGROUND	128
+#define BTRFS_DELAYED_BATCH		16
+
+static struct kmem_cache *delayed_node_cache;
+
+int __init btrfs_delayed_inode_init(void)
+{
+	delayed_node_cache = kmem_cache_create("btrfs_delayed_node",
+					sizeof(struct btrfs_delayed_node),
+					0,
+					SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD,
+					NULL);
+	if (!delayed_node_cache)
+		return -ENOMEM;
+	return 0;
+}
+
+void btrfs_delayed_inode_exit(void)
+{
+	if (delayed_node_cache)
+		kmem_cache_destroy(delayed_node_cache);
+}
+
+static inline void btrfs_init_delayed_node(
+				struct btrfs_delayed_node *delayed_node,
+				struct btrfs_root *root, u64 inode_id)
+{
+	delayed_node->root = root;
+	delayed_node->inode_id = inode_id;
+	atomic_set(&delayed_node->refs, 0);
+	delayed_node->count = 0;
+	delayed_node->flags = 0;
+	delayed_node->ins_root = RB_ROOT;
+	delayed_node->del_root = RB_ROOT;
+	mutex_init(&delayed_node->mutex);
+	delayed_node->index_cnt = 0;
+	INIT_LIST_HEAD(&delayed_node->n_list);
+	INIT_LIST_HEAD(&delayed_node->p_list);
+	delayed_node->bytes_reserved = 0;
+	memset(&delayed_node->inode_item, 0, sizeof(delayed_node->inode_item));
+}
+
+static inline int btrfs_is_continuous_delayed_item(
+					struct btrfs_delayed_item *item1,
+					struct btrfs_delayed_item *item2)
+{
+	if (item1->key.type == BTRFS_DIR_INDEX_KEY &&
+	    item1->key.objectid == item2->key.objectid &&
+	    item1->key.type == item2->key.type &&
+	    item1->key.offset + 1 == item2->key.offset)
+		return 1;
+	return 0;
+}
+
+static inline struct btrfs_delayed_root *btrfs_get_delayed_root(
+							struct btrfs_root *root)
+{
+	return root->fs_info->delayed_root;
+}
+
+static struct btrfs_delayed_node *btrfs_get_delayed_node(struct inode *inode)
+{
+	struct btrfs_inode *btrfs_inode = BTRFS_I(inode);
+	struct btrfs_root *root = btrfs_inode->root;
+	u64 ino = btrfs_ino(inode);
+	struct btrfs_delayed_node *node;
+
+	node = ACCESS_ONCE(btrfs_inode->delayed_node);
+	if (node) {
+		atomic_inc(&node->refs);
+		return node;
+	}
+
+	spin_lock(&root->inode_lock);
+	node = radix_tree_lookup(&root->delayed_nodes_tree, ino);
+	if (node) {
+		if (btrfs_inode->delayed_node) {
+			atomic_inc(&node->refs);	/* can be accessed */
+			BUG_ON(btrfs_inode->delayed_node != node);
+			spin_unlock(&root->inode_lock);
+			return node;
+		}
+		btrfs_inode->delayed_node = node;
+		/* can be accessed and cached in the inode */
+		atomic_add(2, &node->refs);
+		spin_unlock(&root->inode_lock);
+		return node;
+	}
+	spin_unlock(&root->inode_lock);
+
+	return NULL;
+}
+
+/* Will return either the node or PTR_ERR(-ENOMEM) */
+static struct btrfs_delayed_node *btrfs_get_or_create_delayed_node(
+							struct inode *inode)
+{
+	struct btrfs_delayed_node *node;
+	struct btrfs_inode *btrfs_inode = BTRFS_I(inode);
+	struct btrfs_root *root = btrfs_inode->root;
+	u64 ino = btrfs_ino(inode);
+	int ret;
+
+again:
+	node = btrfs_get_delayed_node(inode);
+	if (node)
+		return node;
+
+	node = kmem_cache_alloc(delayed_node_cache, GFP_NOFS);
+	if (!node)
+		return ERR_PTR(-ENOMEM);
+	btrfs_init_delayed_node(node, root, ino);
+
+	/* cached in the btrfs inode and can be accessed */
+	atomic_add(2, &node->refs);
+
+	ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
+	if (ret) {
+		kmem_cache_free(delayed_node_cache, node);
+		return ERR_PTR(ret);
+	}
+
+	spin_lock(&root->inode_lock);
+	ret = radix_tree_insert(&root->delayed_nodes_tree, ino, node);
+	if (ret == -EEXIST) {
+		spin_unlock(&root->inode_lock);
+		kmem_cache_free(delayed_node_cache, node);
+		radix_tree_preload_end();
+		goto again;
+	}
+	btrfs_inode->delayed_node = node;
+	spin_unlock(&root->inode_lock);
+	radix_tree_preload_end();
+
+	return node;
+}
+
+/*
+ * Call it when holding delayed_node->mutex
+ *
+ * If mod = 1, add this node into the prepared list.
+ */
+static void btrfs_queue_delayed_node(struct btrfs_delayed_root *root,
+				     struct btrfs_delayed_node *node,
+				     int mod)
+{
+	spin_lock(&root->lock);
+	if (test_bit(BTRFS_DELAYED_NODE_IN_LIST, &node->flags)) {
+		if (!list_empty(&node->p_list))
+			list_move_tail(&node->p_list, &root->prepare_list);
+		else if (mod)
+			list_add_tail(&node->p_list, &root->prepare_list);
+	} else {
+		list_add_tail(&node->n_list, &root->node_list);
+		list_add_tail(&node->p_list, &root->prepare_list);
+		atomic_inc(&node->refs);	/* inserted into list */
+		root->nodes++;
+		set_bit(BTRFS_DELAYED_NODE_IN_LIST, &node->flags);
+	}
+	spin_unlock(&root->lock);
+}
+
+/* Call it when holding delayed_node->mutex */
+static void btrfs_dequeue_delayed_node(struct btrfs_delayed_root *root,
+				       struct btrfs_delayed_node *node)
+{
+	spin_lock(&root->lock);
+	if (test_bit(BTRFS_DELAYED_NODE_IN_LIST, &node->flags)) {
+		root->nodes--;
+		atomic_dec(&node->refs);	/* not in the list */
+		list_del_init(&node->n_list);
+		if (!list_empty(&node->p_list))
+			list_del_init(&node->p_list);
+		clear_bit(BTRFS_DELAYED_NODE_IN_LIST, &node->flags);
+	}
+	spin_unlock(&root->lock);
+}
+
+static struct btrfs_delayed_node *btrfs_first_delayed_node(
+			struct btrfs_delayed_root *delayed_root)
+{
+	struct list_head *p;
+	struct btrfs_delayed_node *node = NULL;
+
+	spin_lock(&delayed_root->lock);
+	if (list_empty(&delayed_root->node_list))
+		goto out;
+
+	p = delayed_root->node_list.next;
+	node = list_entry(p, struct btrfs_delayed_node, n_list);
+	atomic_inc(&node->refs);
+out:
+	spin_unlock(&delayed_root->lock);
+
+	return node;
+}
+
+static struct btrfs_delayed_node *btrfs_next_delayed_node(
+						struct btrfs_delayed_node *node)
+{
+	struct btrfs_delayed_root *delayed_root;
+	struct list_head *p;
+	struct btrfs_delayed_node *next = NULL;
+
+	delayed_root = node->root->fs_info->delayed_root;
+	spin_lock(&delayed_root->lock);
+	if (!test_bit(BTRFS_DELAYED_NODE_IN_LIST, &node->flags)) {
+		/* not in the list */
+		if (list_empty(&delayed_root->node_list))
+			goto out;
+		p = delayed_root->node_list.next;
+	} else if (list_is_last(&node->n_list, &delayed_root->node_list))
+		goto out;
+	else
+		p = node->n_list.next;
+
+	next = list_entry(p, struct btrfs_delayed_node, n_list);
+	atomic_inc(&next->refs);
+out:
+	spin_unlock(&delayed_root->lock);
+
+	return next;
+}
+
+static void __btrfs_release_delayed_node(
+				struct btrfs_delayed_node *delayed_node,
+				int mod)
+{
+	struct btrfs_delayed_root *delayed_root;
+
+	if (!delayed_node)
+		return;
+
+	delayed_root = delayed_node->root->fs_info->delayed_root;
+
+	mutex_lock(&delayed_node->mutex);
+	if (delayed_node->count)
+		btrfs_queue_delayed_node(delayed_root, delayed_node, mod);
+	else
+		btrfs_dequeue_delayed_node(delayed_root, delayed_node);
+	mutex_unlock(&delayed_node->mutex);
+
+	if (atomic_dec_and_test(&delayed_node->refs)) {
+		bool free = false;
+		struct btrfs_root *root = delayed_node->root;
+		spin_lock(&root->inode_lock);
+		if (atomic_read(&delayed_node->refs) == 0) {
+			radix_tree_delete(&root->delayed_nodes_tree,
+					  delayed_node->inode_id);
+			free = true;
+		}
+		spin_unlock(&root->inode_lock);
+		if (free)
+			kmem_cache_free(delayed_node_cache, delayed_node);
+	}
+}
+
+static inline void btrfs_release_delayed_node(struct btrfs_delayed_node *node)
+{
+	__btrfs_release_delayed_node(node, 0);
+}
+
+static struct btrfs_delayed_node *btrfs_first_prepared_delayed_node(
+					struct btrfs_delayed_root *delayed_root)
+{
+	struct list_head *p;
+	struct btrfs_delayed_node *node = NULL;
+
+	spin_lock(&delayed_root->lock);
+	if (list_empty(&delayed_root->prepare_list))
+		goto out;
+
+	p = delayed_root->prepare_list.next;
+	list_del_init(p);
+	node = list_entry(p, struct btrfs_delayed_node, p_list);
+	atomic_inc(&node->refs);
+out:
+	spin_unlock(&delayed_root->lock);
+
+	return node;
+}
+
+static inline void btrfs_release_prepared_delayed_node(
+					struct btrfs_delayed_node *node)
+{
+	__btrfs_release_delayed_node(node, 1);
+}
+
+static struct btrfs_delayed_item *btrfs_alloc_delayed_item(u32 data_len)
+{
+	struct btrfs_delayed_item *item;
+	item = kmalloc(sizeof(*item) + data_len, GFP_NOFS);
+	if (item) {
+		item->data_len = data_len;
+		item->ins_or_del = 0;
+		item->bytes_reserved = 0;
+		item->delayed_node = NULL;
+		atomic_set(&item->refs, 1);
+	}
+	return item;
+}
+
+/*
+ * __btrfs_lookup_delayed_item - look up the delayed item by key
+ * @delayed_node: pointer to the delayed node
+ * @key:	  the key to look up
+ * @prev:	  used to store the prev item if the right item isn't found
+ * @next:	  used to store the next item if the right item isn't found
+ *
+ * Note: if we don't find the right item, we will return the prev item and
+ * the next item.
+ */
+static struct btrfs_delayed_item *__btrfs_lookup_delayed_item(
+				struct rb_root *root,
+				struct btrfs_key *key,
+				struct btrfs_delayed_item **prev,
+				struct btrfs_delayed_item **next)
+{
+	struct rb_node *node, *prev_node = NULL;
+	struct btrfs_delayed_item *delayed_item = NULL;
+	int ret = 0;
+
+	node = root->rb_node;
+
+	while (node) {
+		delayed_item = rb_entry(node, struct btrfs_delayed_item,
+					rb_node);
+		prev_node = node;
+		ret = btrfs_comp_cpu_keys(&delayed_item->key, key);
+		if (ret < 0)
+			node = node->rb_right;
+		else if (ret > 0)
+			node = node->rb_left;
+		else
+			return delayed_item;
+	}
+
+	if (prev) {
+		if (!prev_node)
+			*prev = NULL;
+		else if (ret < 0)
+			*prev = delayed_item;
+		else if ((node = rb_prev(prev_node)) != NULL) {
+			*prev = rb_entry(node, struct btrfs_delayed_item,
+					 rb_node);
+		} else
+			*prev = NULL;
+	}
+
+	if (next) {
+		if (!prev_node)
+			*next = NULL;
+		else if (ret > 0)
+			*next = delayed_item;
+		else if ((node = rb_next(prev_node)) != NULL) {
+			*next = rb_entry(node, struct btrfs_delayed_item,
+					 rb_node);
+		} else
+			*next = NULL;
+	}
+	return NULL;
+}
+
+static struct btrfs_delayed_item *__btrfs_lookup_delayed_insertion_item(
+					struct btrfs_delayed_node *delayed_node,
+					struct btrfs_key *key)
+{
+	struct btrfs_delayed_item *item;
+
+	item = __btrfs_lookup_delayed_item(&delayed_node->ins_root, key,
+					   NULL, NULL);
+	return item;
+}
+
+static int __btrfs_add_delayed_item(struct btrfs_delayed_node *delayed_node,
+				    struct btrfs_delayed_item *ins,
+				    int action)
+{
+	struct rb_node **p, *node;
+	struct rb_node *parent_node = NULL;
+	struct rb_root *root;
+	struct btrfs_delayed_item *item;
+	int cmp;
+
+	if (action == BTRFS_DELAYED_INSERTION_ITEM)
+		root = &delayed_node->ins_root;
+	else if (action == BTRFS_DELAYED_DELETION_ITEM)
+		root = &delayed_node->del_root;
+	else
+		BUG();
+	p = &root->rb_node;
+	node = &ins->rb_node;
+
+	while (*p) {
+		parent_node = *p;
+		item = rb_entry(parent_node, struct btrfs_delayed_item,
+				 rb_node);
+
+		cmp = btrfs_comp_cpu_keys(&item->key, &ins->key);
+		if (cmp < 0)
+			p = &(*p)->rb_right;
+		else if (cmp > 0)
+			p = &(*p)->rb_left;
+		else
+			return -EEXIST;
+	}
+
+	rb_link_node(node, parent_node, p);
+	rb_insert_color(node, root);
+	ins->delayed_node = delayed_node;
+	ins->ins_or_del = action;
+
+	if (ins->key.type == BTRFS_DIR_INDEX_KEY &&
+	    action == BTRFS_DELAYED_INSERTION_ITEM &&
+	    ins->key.offset >= delayed_node->index_cnt)
+			delayed_node->index_cnt = ins->key.offset + 1;
+
+	delayed_node->count++;
+	atomic_inc(&delayed_node->root->fs_info->delayed_root->items);
+	return 0;
+}
+
+static int __btrfs_add_delayed_insertion_item(struct btrfs_delayed_node *node,
+					      struct btrfs_delayed_item *item)
+{
+	return __btrfs_add_delayed_item(node, item,
+					BTRFS_DELAYED_INSERTION_ITEM);
+}
+
+static int __btrfs_add_delayed_deletion_item(struct btrfs_delayed_node *node,
+					     struct btrfs_delayed_item *item)
+{
+	return __btrfs_add_delayed_item(node, item,
+					BTRFS_DELAYED_DELETION_ITEM);
+}
+
+static void finish_one_item(struct btrfs_delayed_root *delayed_root)
+{
+	int seq = atomic_inc_return(&delayed_root->items_seq);
+	if ((atomic_dec_return(&delayed_root->items) <
+	    BTRFS_DELAYED_BACKGROUND || seq % BTRFS_DELAYED_BATCH == 0) &&
+	    waitqueue_active(&delayed_root->wait))
+		wake_up(&delayed_root->wait);
+}
+
+static void __btrfs_remove_delayed_item(struct btrfs_delayed_item *delayed_item)
+{
+	struct rb_root *root;
+	struct btrfs_delayed_root *delayed_root;
+
+	delayed_root = delayed_item->delayed_node->root->fs_info->delayed_root;
+
+	BUG_ON(!delayed_root);
+	BUG_ON(delayed_item->ins_or_del != BTRFS_DELAYED_DELETION_ITEM &&
+	       delayed_item->ins_or_del != BTRFS_DELAYED_INSERTION_ITEM);
+
+	if (delayed_item->ins_or_del == BTRFS_DELAYED_INSERTION_ITEM)
+		root = &delayed_item->delayed_node->ins_root;
+	else
+		root = &delayed_item->delayed_node->del_root;
+
+	rb_erase(&delayed_item->rb_node, root);
+	delayed_item->delayed_node->count--;
+
+	finish_one_item(delayed_root);
+}
+
+static void btrfs_release_delayed_item(struct btrfs_delayed_item *item)
+{
+	if (item) {
+		__btrfs_remove_delayed_item(item);
+		if (atomic_dec_and_test(&item->refs))
+			kfree(item);
+	}
+}
+
+static struct btrfs_delayed_item *__btrfs_first_delayed_insertion_item(
+					struct btrfs_delayed_node *delayed_node)
+{
+	struct rb_node *p;
+	struct btrfs_delayed_item *item = NULL;
+
+	p = rb_first(&delayed_node->ins_root);
+	if (p)
+		item = rb_entry(p, struct btrfs_delayed_item, rb_node);
+
+	return item;
+}
+
+static struct btrfs_delayed_item *__btrfs_first_delayed_deletion_item(
+					struct btrfs_delayed_node *delayed_node)
+{
+	struct rb_node *p;
+	struct btrfs_delayed_item *item = NULL;
+
+	p = rb_first(&delayed_node->del_root);
+	if (p)
+		item = rb_entry(p, struct btrfs_delayed_item, rb_node);
+
+	return item;
+}
+
+static struct btrfs_delayed_item *__btrfs_next_delayed_item(
+						struct btrfs_delayed_item *item)
+{
+	struct rb_node *p;
+	struct btrfs_delayed_item *next = NULL;
+
+	p = rb_next(&item->rb_node);
+	if (p)
+		next = rb_entry(p, struct btrfs_delayed_item, rb_node);
+
+	return next;
+}
+
+static int btrfs_delayed_item_reserve_metadata(struct btrfs_trans_handle *trans,
+					       struct btrfs_root *root,
+					       struct btrfs_delayed_item *item)
+{
+	struct btrfs_block_rsv *src_rsv;
+	struct btrfs_block_rsv *dst_rsv;
+	u64 num_bytes;
+	int ret;
+
+	if (!trans->bytes_reserved)
+		return 0;
+
+	src_rsv = trans->block_rsv;
+	dst_rsv = &root->fs_info->delayed_block_rsv;
+
+	num_bytes = btrfs_calc_trans_metadata_size(root, 1);
+	ret = btrfs_block_rsv_migrate(src_rsv, dst_rsv, num_bytes);
+	if (!ret) {
+		trace_btrfs_space_reservation(root->fs_info, "delayed_item",
+					      item->key.objectid,
+					      num_bytes, 1);
+		item->bytes_reserved = num_bytes;
+	}
+
+	return ret;
+}
+
+static void btrfs_delayed_item_release_metadata(struct btrfs_root *root,
+						struct btrfs_delayed_item *item)
+{
+	struct btrfs_block_rsv *rsv;
+
+	if (!item->bytes_reserved)
+		return;
+
+	rsv = &root->fs_info->delayed_block_rsv;
+	trace_btrfs_space_reservation(root->fs_info, "delayed_item",
+				      item->key.objectid, item->bytes_reserved,
+				      0);
+	btrfs_block_rsv_release(root, rsv,
+				item->bytes_reserved);
+}
+
+static int btrfs_delayed_inode_reserve_metadata(
+					struct btrfs_trans_handle *trans,
+					struct btrfs_root *root,
+					struct inode *inode,
+					struct btrfs_delayed_node *node)
+{
+	struct btrfs_block_rsv *src_rsv;
+	struct btrfs_block_rsv *dst_rsv;
+	u64 num_bytes;
+	int ret;
+	bool release = false;
+
+	src_rsv = trans->block_rsv;
+	dst_rsv = &root->fs_info->delayed_block_rsv;
+
+	num_bytes = btrfs_calc_trans_metadata_size(root, 1);
+
+	/*
+	 * btrfs_dirty_inode will update the inode under btrfs_join_transaction
+	 * which doesn't reserve space for speed.  This is a problem since we
+	 * still need to reserve space for this update, so try to reserve the
+	 * space.
+	 *
+	 * Now if src_rsv == delalloc_block_rsv we'll let it just steal since
+	 * we're accounted for.
+	 */
+	if (!src_rsv || (!trans->bytes_reserved &&
+			 src_rsv->type != BTRFS_BLOCK_RSV_DELALLOC)) {
+		ret = btrfs_block_rsv_add(root, dst_rsv, num_bytes,
+					  BTRFS_RESERVE_NO_FLUSH);
+		/*
+		 * Since we're under a transaction reserve_metadata_bytes could
+		 * try to commit the transaction which will make it return
+		 * EAGAIN to make us stop the transaction we have, so return
+		 * ENOSPC instead so that btrfs_dirty_inode knows what to do.
+		 */
+		if (ret == -EAGAIN)
+			ret = -ENOSPC;
+		if (!ret) {
+			node->bytes_reserved = num_bytes;
+			trace_btrfs_space_reservation(root->fs_info,
+						      "delayed_inode",
+						      btrfs_ino(inode),
+						      num_bytes, 1);
+		}
+		return ret;
+	} else if (src_rsv->type == BTRFS_BLOCK_RSV_DELALLOC) {
+		spin_lock(&BTRFS_I(inode)->lock);
+		if (test_and_clear_bit(BTRFS_INODE_DELALLOC_META_RESERVED,
+				       &BTRFS_I(inode)->runtime_flags)) {
+			spin_unlock(&BTRFS_I(inode)->lock);
+			release = true;
+			goto migrate;
+		}
+		spin_unlock(&BTRFS_I(inode)->lock);
+
+		/* Ok we didn't have space pre-reserved.  This shouldn't happen
+		 * too often but it can happen if we do delalloc to an existing
+		 * inode which gets dirtied because of the time update, and then
+		 * isn't touched again until after the transaction commits and
+		 * then we try to write out the data.  First try to be nice and
+		 * reserve something strictly for us.  If not be a pain and try
+		 * to steal from the delalloc block rsv.
+		 */
+		ret = btrfs_block_rsv_add(root, dst_rsv, num_bytes,
+					  BTRFS_RESERVE_NO_FLUSH);
+		if (!ret)
+			goto out;
+
+		ret = btrfs_block_rsv_migrate(src_rsv, dst_rsv, num_bytes);
+		if (!WARN_ON(ret))
+			goto out;
+
+		/*
+		 * Ok this is a problem, let's just steal from the global rsv
+		 * since this really shouldn't happen that often.
+		 */
+		ret = btrfs_block_rsv_migrate(&root->fs_info->global_block_rsv,
+					      dst_rsv, num_bytes);
+		goto out;
+	}
+
+migrate:
+	ret = btrfs_block_rsv_migrate(src_rsv, dst_rsv, num_bytes);
+
+out:
+	/*
+	 * Migrate only takes a reservation, it doesn't touch the size of the
+	 * block_rsv.  This is to simplify people who don't normally have things
+	 * migrated from their block rsv.  If they go to release their
+	 * reservation, that will decrease the size as well, so if migrate
+	 * reduced size we'd end up with a negative size.  But for the
+	 * delalloc_meta_reserved stuff we will only know to drop 1 reservation,
+	 * but we could in fact do this reserve/migrate dance several times
+	 * between the time we did the original reservation and we'd clean it
+	 * up.  So to take care of this, release the space for the meta
+	 * reservation here.  I think it may be time for a documentation page on
+	 * how block rsvs. work.
+	 */
+	if (!ret) {
+		trace_btrfs_space_reservation(root->fs_info, "delayed_inode",
+					      btrfs_ino(inode), num_bytes, 1);
+		node->bytes_reserved = num_bytes;
+	}
+
+	if (release) {
+		trace_btrfs_space_reservation(root->fs_info, "delalloc",
+					      btrfs_ino(inode), num_bytes, 0);
+		btrfs_block_rsv_release(root, src_rsv, num_bytes);
+	}
+
+	return ret;
+}
+
+static void btrfs_delayed_inode_release_metadata(struct btrfs_root *root,
+						struct btrfs_delayed_node *node)
+{
+	struct btrfs_block_rsv *rsv;
+
+	if (!node->bytes_reserved)
+		return;
+
+	rsv = &root->fs_info->delayed_block_rsv;
+	trace_btrfs_space_reservation(root->fs_info, "delayed_inode",
+				      node->inode_id, node->bytes_reserved, 0);
+	btrfs_block_rsv_release(root, rsv,
+				node->bytes_reserved);
+	node->bytes_reserved = 0;
+}
+
+/*
+ * This helper will insert some continuous items into the same leaf according
+ * to the free space of the leaf.
+ */
+static int btrfs_batch_insert_items(struct btrfs_root *root,
+				    struct btrfs_path *path,
+				    struct btrfs_delayed_item *item)
+{
+	struct btrfs_delayed_item *curr, *next;
+	int free_space;
+	int total_data_size = 0, total_size = 0;
+	struct extent_buffer *leaf;
+	char *data_ptr;
+	struct btrfs_key *keys;
+	u32 *data_size;
+	struct list_head head;
+	int slot;
+	int nitems;
+	int i;
+	int ret = 0;
+
+	BUG_ON(!path->nodes[0]);
+
+	leaf = path->nodes[0];
+	free_space = btrfs_leaf_free_space(root, leaf);
+	INIT_LIST_HEAD(&head);
+
+	next = item;
+	nitems = 0;
+
+	/*
+	 * count the number of the continuous items that we can insert in batch
+	 */
+	while (total_size + next->data_len + sizeof(struct btrfs_item) <=
+	       free_space) {
+		total_data_size += next->data_len;
+		total_size += next->data_len + sizeof(struct btrfs_item);
+		list_add_tail(&next->tree_list, &head);
+		nitems++;
+
+		curr = next;
+		next = __btrfs_next_delayed_item(curr);
+		if (!next)
+			break;
+
+		if (!btrfs_is_continuous_delayed_item(curr, next))
+			break;
+	}
+
+	if (!nitems) {
+		ret = 0;
+		goto out;
+	}
+
+	/*
+	 * we need allocate some memory space, but it might cause the task
+	 * to sleep, so we set all locked nodes in the path to blocking locks
+	 * first.
+	 */
+	btrfs_set_path_blocking(path);
+
+	keys = kmalloc_array(nitems, sizeof(struct btrfs_key), GFP_NOFS);
+	if (!keys) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	data_size = kmalloc_array(nitems, sizeof(u32), GFP_NOFS);
+	if (!data_size) {
+		ret = -ENOMEM;
+		goto error;
+	}
+
+	/* get keys of all the delayed items */
+	i = 0;
+	list_for_each_entry(next, &head, tree_list) {
+		keys[i] = next->key;
+		data_size[i] = next->data_len;
+		i++;
+	}
+
+	/* reset all the locked nodes in the patch to spinning locks. */
+	btrfs_clear_path_blocking(path, NULL, 0);
+
+	/* insert the keys of the items */
+	setup_items_for_insert(root, path, keys, data_size,
+			       total_data_size, total_size, nitems);
+
+	/* insert the dir index items */
+	slot = path->slots[0];
+	list_for_each_entry_safe(curr, next, &head, tree_list) {
+		data_ptr = btrfs_item_ptr(leaf, slot, char);
+		write_extent_buffer(leaf, &curr->data,
+				    (unsigned long)data_ptr,
+				    curr->data_len);
+		slot++;
+
+		btrfs_delayed_item_release_metadata(root, curr);
+
+		list_del(&curr->tree_list);
+		btrfs_release_delayed_item(curr);
+	}
+
+error:
+	kfree(data_size);
+	kfree(keys);
+out:
+	return ret;
+}
+
+/*
+ * This helper can just do simple insertion that needn't extend item for new
+ * data, such as directory name index insertion, inode insertion.
+ */
+static int btrfs_insert_delayed_item(struct btrfs_trans_handle *trans,
+				     struct btrfs_root *root,
+				     struct btrfs_path *path,
+				     struct btrfs_delayed_item *delayed_item)
+{
+	struct extent_buffer *leaf;
+	char *ptr;
+	int ret;
+
+	ret = btrfs_insert_empty_item(trans, root, path, &delayed_item->key,
+				      delayed_item->data_len);
+	if (ret < 0 && ret != -EEXIST)
+		return ret;
+
+	leaf = path->nodes[0];
+
+	ptr = btrfs_item_ptr(leaf, path->slots[0], char);
+
+	write_extent_buffer(leaf, delayed_item->data, (unsigned long)ptr,
+			    delayed_item->data_len);
+	btrfs_mark_buffer_dirty(leaf);
+
+	btrfs_delayed_item_release_metadata(root, delayed_item);
+	return 0;
+}
+
+/*
+ * we insert an item first, then if there are some continuous items, we try
+ * to insert those items into the same leaf.
+ */
+static int btrfs_insert_delayed_items(struct btrfs_trans_handle *trans,
+				      struct btrfs_path *path,
+				      struct btrfs_root *root,
+				      struct btrfs_delayed_node *node)
+{
+	struct btrfs_delayed_item *curr, *prev;
+	int ret = 0;
+
+do_again:
+	mutex_lock(&node->mutex);
+	curr = __btrfs_first_delayed_insertion_item(node);
+	if (!curr)
+		goto insert_end;
+
+	ret = btrfs_insert_delayed_item(trans, root, path, curr);
+	if (ret < 0) {
+		btrfs_release_path(path);
+		goto insert_end;
+	}
+
+	prev = curr;
+	curr = __btrfs_next_delayed_item(prev);
+	if (curr && btrfs_is_continuous_delayed_item(prev, curr)) {
+		/* insert the continuous items into the same leaf */
+		path->slots[0]++;
+		btrfs_batch_insert_items(root, path, curr);
+	}
+	btrfs_release_delayed_item(prev);
+	btrfs_mark_buffer_dirty(path->nodes[0]);
+
+	btrfs_release_path(path);
+	mutex_unlock(&node->mutex);
+	goto do_again;
+
+insert_end:
+	mutex_unlock(&node->mutex);
+	return ret;
+}
+
+static int btrfs_batch_delete_items(struct btrfs_trans_handle *trans,
+				    struct btrfs_root *root,
+				    struct btrfs_path *path,
+				    struct btrfs_delayed_item *item)
+{
+	struct btrfs_delayed_item *curr, *next;
+	struct extent_buffer *leaf;
+	struct btrfs_key key;
+	struct list_head head;
+	int nitems, i, last_item;
+	int ret = 0;
+
+	BUG_ON(!path->nodes[0]);
+
+	leaf = path->nodes[0];
+
+	i = path->slots[0];
+	last_item = btrfs_header_nritems(leaf) - 1;
+	if (i > last_item)
+		return -ENOENT;	/* FIXME: Is errno suitable? */
+
+	next = item;
+	INIT_LIST_HEAD(&head);
+	btrfs_item_key_to_cpu(leaf, &key, i);
+	nitems = 0;
+	/*
+	 * count the number of the dir index items that we can delete in batch
+	 */
+	while (btrfs_comp_cpu_keys(&next->key, &key) == 0) {
+		list_add_tail(&next->tree_list, &head);
+		nitems++;
+
+		curr = next;
+		next = __btrfs_next_delayed_item(curr);
+		if (!next)
+			break;
+
+		if (!btrfs_is_continuous_delayed_item(curr, next))
+			break;
+
+		i++;
+		if (i > last_item)
+			break;
+		btrfs_item_key_to_cpu(leaf, &key, i);
+	}
+
+	if (!nitems)
+		return 0;
+
+	ret = btrfs_del_items(trans, root, path, path->slots[0], nitems);
+	if (ret)
+		goto out;
+
+	list_for_each_entry_safe(curr, next, &head, tree_list) {
+		btrfs_delayed_item_release_metadata(root, curr);
+		list_del(&curr->tree_list);
+		btrfs_release_delayed_item(curr);
+	}
+
+out:
+	return ret;
+}
+
+static int btrfs_delete_delayed_items(struct btrfs_trans_handle *trans,
+				      struct btrfs_path *path,
+				      struct btrfs_root *root,
+				      struct btrfs_delayed_node *node)
+{
+	struct btrfs_delayed_item *curr, *prev;
+	int ret = 0;
+
+do_again:
+	mutex_lock(&node->mutex);
+	curr = __btrfs_first_delayed_deletion_item(node);
+	if (!curr)
+		goto delete_fail;
+
+	ret = btrfs_search_slot(trans, root, &curr->key, path, -1, 1);
+	if (ret < 0)
+		goto delete_fail;
+	else if (ret > 0) {
+		/*
+		 * can't find the item which the node points to, so this node
+		 * is invalid, just drop it.
+		 */
+		prev = curr;
+		curr = __btrfs_next_delayed_item(prev);
+		btrfs_release_delayed_item(prev);
+		ret = 0;
+		btrfs_release_path(path);
+		if (curr) {
+			mutex_unlock(&node->mutex);
+			goto do_again;
+		} else
+			goto delete_fail;
+	}
+
+	btrfs_batch_delete_items(trans, root, path, curr);
+	btrfs_release_path(path);
+	mutex_unlock(&node->mutex);
+	goto do_again;
+
+delete_fail:
+	btrfs_release_path(path);
+	mutex_unlock(&node->mutex);
+	return ret;
+}
+
+static void btrfs_release_delayed_inode(struct btrfs_delayed_node *delayed_node)
+{
+	struct btrfs_delayed_root *delayed_root;
+
+	if (delayed_node &&
+	    test_bit(BTRFS_DELAYED_NODE_INODE_DIRTY, &delayed_node->flags)) {
+		BUG_ON(!delayed_node->root);
+		clear_bit(BTRFS_DELAYED_NODE_INODE_DIRTY, &delayed_node->flags);
+		delayed_node->count--;
+
+		delayed_root = delayed_node->root->fs_info->delayed_root;
+		finish_one_item(delayed_root);
+	}
+}
+
+static void btrfs_release_delayed_iref(struct btrfs_delayed_node *delayed_node)
+{
+	struct btrfs_delayed_root *delayed_root;
+
+	ASSERT(delayed_node->root);
+	clear_bit(BTRFS_DELAYED_NODE_DEL_IREF, &delayed_node->flags);
+	delayed_node->count--;
+
+	delayed_root = delayed_node->root->fs_info->delayed_root;
+	finish_one_item(delayed_root);
+}
+
+static int __btrfs_update_delayed_inode(struct btrfs_trans_handle *trans,
+					struct btrfs_root *root,
+					struct btrfs_path *path,
+					struct btrfs_delayed_node *node)
+{
+	struct btrfs_key key;
+	struct btrfs_inode_item *inode_item;
+	struct extent_buffer *leaf;
+	int mod;
+	int ret;
+
+	key.objectid = node->inode_id;
+	btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
+	key.offset = 0;
+
+	if (test_bit(BTRFS_DELAYED_NODE_DEL_IREF, &node->flags))
+		mod = -1;
+	else
+		mod = 1;
+
+	ret = btrfs_lookup_inode(trans, root, path, &key, mod);
+	if (ret > 0) {
+		btrfs_release_path(path);
+		return -ENOENT;
+	} else if (ret < 0) {
+		return ret;
+	}
+
+	leaf = path->nodes[0];
+	inode_item = btrfs_item_ptr(leaf, path->slots[0],
+				    struct btrfs_inode_item);
+	write_extent_buffer(leaf, &node->inode_item, (unsigned long)inode_item,
+			    sizeof(struct btrfs_inode_item));
+	btrfs_mark_buffer_dirty(leaf);
+
+	if (!test_bit(BTRFS_DELAYED_NODE_DEL_IREF, &node->flags))
+		goto no_iref;
+
+	path->slots[0]++;
+	if (path->slots[0] >= btrfs_header_nritems(leaf))
+		goto search;
+again:
+	btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
+	if (key.objectid != node->inode_id)
+		goto out;
+
+	if (key.type != BTRFS_INODE_REF_KEY &&
+	    key.type != BTRFS_INODE_EXTREF_KEY)
+		goto out;
+
+	/*
+	 * Delayed iref deletion is for the inode who has only one link,
+	 * so there is only one iref. The case that several irefs are
+	 * in the same item doesn't exist.
+	 */
+	btrfs_del_item(trans, root, path);
+out:
+	btrfs_release_delayed_iref(node);
+no_iref:
+	btrfs_release_path(path);
+err_out:
+	btrfs_delayed_inode_release_metadata(root, node);
+	btrfs_release_delayed_inode(node);
+
+	return ret;
+
+search:
+	btrfs_release_path(path);
+
+	btrfs_set_key_type(&key, BTRFS_INODE_EXTREF_KEY);
+	key.offset = -1;
+	ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
+	if (ret < 0)
+		goto err_out;
+	ASSERT(ret);
+
+	ret = 0;
+	leaf = path->nodes[0];
+	path->slots[0]--;
+	goto again;
+}
+
+static inline int btrfs_update_delayed_inode(struct btrfs_trans_handle *trans,
+					     struct btrfs_root *root,
+					     struct btrfs_path *path,
+					     struct btrfs_delayed_node *node)
+{
+	int ret;
+
+	mutex_lock(&node->mutex);
+	if (!test_bit(BTRFS_DELAYED_NODE_INODE_DIRTY, &node->flags)) {
+		mutex_unlock(&node->mutex);
+		return 0;
+	}
+
+	ret = __btrfs_update_delayed_inode(trans, root, path, node);
+	mutex_unlock(&node->mutex);
+	return ret;
+}
+
+static inline int
+__btrfs_commit_inode_delayed_items(struct btrfs_trans_handle *trans,
+				   struct btrfs_path *path,
+				   struct btrfs_delayed_node *node)
+{
+	int ret;
+
+	ret = btrfs_insert_delayed_items(trans, path, node->root, node);
+	if (ret)
+		return ret;
+
+	ret = btrfs_delete_delayed_items(trans, path, node->root, node);
+	if (ret)
+		return ret;
+
+	ret = btrfs_update_delayed_inode(trans, node->root, path, node);
+	return ret;
+}
+
+/*
+ * Called when committing the transaction.
+ * Returns 0 on success.
+ * Returns < 0 on error and returns with an aborted transaction with any
+ * outstanding delayed items cleaned up.
+ */
+static int __btrfs_run_delayed_items(struct btrfs_trans_handle *trans,
+				     struct btrfs_root *root, int nr)
+{
+	struct btrfs_delayed_root *delayed_root;
+	struct btrfs_delayed_node *curr_node, *prev_node;
+	struct btrfs_path *path;
+	struct btrfs_block_rsv *block_rsv;
+	int ret = 0;
+	bool count = (nr > 0);
+
+	if (trans->aborted)
+		return -EIO;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+	path->leave_spinning = 1;
+
+	block_rsv = trans->block_rsv;
+	trans->block_rsv = &root->fs_info->delayed_block_rsv;
+
+	delayed_root = btrfs_get_delayed_root(root);
+
+	curr_node = btrfs_first_delayed_node(delayed_root);
+	while (curr_node && (!count || (count && nr--))) {
+		ret = __btrfs_commit_inode_delayed_items(trans, path,
+							 curr_node);
+		if (ret) {
+			btrfs_release_delayed_node(curr_node);
+			curr_node = NULL;
+			btrfs_abort_transaction(trans, root, ret);
+			break;
+		}
+
+		prev_node = curr_node;
+		curr_node = btrfs_next_delayed_node(curr_node);
+		btrfs_release_delayed_node(prev_node);
+	}
+
+	if (curr_node)
+		btrfs_release_delayed_node(curr_node);
+	btrfs_free_path(path);
+	trans->block_rsv = block_rsv;
+
+	return ret;
+}
+
+int btrfs_run_delayed_items(struct btrfs_trans_handle *trans,
+			    struct btrfs_root *root)
+{
+	return __btrfs_run_delayed_items(trans, root, -1);
+}
+
+int btrfs_run_delayed_items_nr(struct btrfs_trans_handle *trans,
+			       struct btrfs_root *root, int nr)
+{
+	return __btrfs_run_delayed_items(trans, root, nr);
+}
+
+int btrfs_commit_inode_delayed_items(struct btrfs_trans_handle *trans,
+				     struct inode *inode)
+{
+	struct btrfs_delayed_node *delayed_node = btrfs_get_delayed_node(inode);
+	struct btrfs_path *path;
+	struct btrfs_block_rsv *block_rsv;
+	int ret;
+
+	if (!delayed_node)
+		return 0;
+
+	mutex_lock(&delayed_node->mutex);
+	if (!delayed_node->count) {
+		mutex_unlock(&delayed_node->mutex);
+		btrfs_release_delayed_node(delayed_node);
+		return 0;
+	}
+	mutex_unlock(&delayed_node->mutex);
+
+	path = btrfs_alloc_path();
+	if (!path) {
+		btrfs_release_delayed_node(delayed_node);
+		return -ENOMEM;
+	}
+	path->leave_spinning = 1;
+
+	block_rsv = trans->block_rsv;
+	trans->block_rsv = &delayed_node->root->fs_info->delayed_block_rsv;
+
+	ret = __btrfs_commit_inode_delayed_items(trans, path, delayed_node);
+
+	btrfs_release_delayed_node(delayed_node);
+	btrfs_free_path(path);
+	trans->block_rsv = block_rsv;
+
+	return ret;
+}
+
+int btrfs_commit_inode_delayed_inode(struct inode *inode)
+{
+	struct btrfs_trans_handle *trans;
+	struct btrfs_delayed_node *delayed_node = btrfs_get_delayed_node(inode);
+	struct btrfs_path *path;
+	struct btrfs_block_rsv *block_rsv;
+	int ret;
+
+	if (!delayed_node)
+		return 0;
+
+	mutex_lock(&delayed_node->mutex);
+	if (!test_bit(BTRFS_DELAYED_NODE_INODE_DIRTY, &delayed_node->flags)) {
+		mutex_unlock(&delayed_node->mutex);
+		btrfs_release_delayed_node(delayed_node);
+		return 0;
+	}
+	mutex_unlock(&delayed_node->mutex);
+
+	trans = btrfs_join_transaction(delayed_node->root);
+	if (IS_ERR(trans)) {
+		ret = PTR_ERR(trans);
+		goto out;
+	}
+
+	path = btrfs_alloc_path();
+	if (!path) {
+		ret = -ENOMEM;
+		goto trans_out;
+	}
+	path->leave_spinning = 1;
+
+	block_rsv = trans->block_rsv;
+	trans->block_rsv = &delayed_node->root->fs_info->delayed_block_rsv;
+
+	mutex_lock(&delayed_node->mutex);
+	if (test_bit(BTRFS_DELAYED_NODE_INODE_DIRTY, &delayed_node->flags))
+		ret = __btrfs_update_delayed_inode(trans, delayed_node->root,
+						   path, delayed_node);
+	else
+		ret = 0;
+	mutex_unlock(&delayed_node->mutex);
+
+	btrfs_free_path(path);
+	trans->block_rsv = block_rsv;
+trans_out:
+	btrfs_end_transaction(trans, delayed_node->root);
+	btrfs_btree_balance_dirty(delayed_node->root);
+out:
+	btrfs_release_delayed_node(delayed_node);
+
+	return ret;
+}
+
+void btrfs_remove_delayed_node(struct inode *inode)
+{
+	struct btrfs_delayed_node *delayed_node;
+
+	delayed_node = ACCESS_ONCE(BTRFS_I(inode)->delayed_node);
+	if (!delayed_node)
+		return;
+
+	BTRFS_I(inode)->delayed_node = NULL;
+	btrfs_release_delayed_node(delayed_node);
+}
+
+struct btrfs_async_delayed_work {
+	struct btrfs_delayed_root *delayed_root;
+	int nr;
+	struct btrfs_work work;
+};
+
+static void btrfs_async_run_delayed_root(struct btrfs_work *work)
+{
+	struct btrfs_async_delayed_work *async_work;
+	struct btrfs_delayed_root *delayed_root;
+	struct btrfs_trans_handle *trans;
+	struct btrfs_path *path;
+	struct btrfs_delayed_node *delayed_node = NULL;
+	struct btrfs_root *root;
+	struct btrfs_block_rsv *block_rsv;
+	int total_done = 0;
+
+	async_work = container_of(work, struct btrfs_async_delayed_work, work);
+	delayed_root = async_work->delayed_root;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		goto out;
+
+again:
+	if (atomic_read(&delayed_root->items) < BTRFS_DELAYED_BACKGROUND / 2)
+		goto free_path;
+
+	delayed_node = btrfs_first_prepared_delayed_node(delayed_root);
+	if (!delayed_node)
+		goto free_path;
+
+	path->leave_spinning = 1;
+	root = delayed_node->root;
+
+	trans = btrfs_join_transaction(root);
+	if (IS_ERR(trans))
+		goto release_path;
+
+	block_rsv = trans->block_rsv;
+	trans->block_rsv = &root->fs_info->delayed_block_rsv;
+
+	__btrfs_commit_inode_delayed_items(trans, path, delayed_node);
+
+	trans->block_rsv = block_rsv;
+	btrfs_end_transaction(trans, root);
+	btrfs_btree_balance_dirty_nodelay(root);
+
+release_path:
+	btrfs_release_path(path);
+	total_done++;
+
+	btrfs_release_prepared_delayed_node(delayed_node);
+	if (async_work->nr == 0 || total_done < async_work->nr)
+		goto again;
+
+free_path:
+	btrfs_free_path(path);
+out:
+	wake_up(&delayed_root->wait);
+	kfree(async_work);
+}
+
+
+static int btrfs_wq_run_delayed_node(struct btrfs_delayed_root *delayed_root,
+				     struct btrfs_root *root, int nr)
+{
+	struct btrfs_async_delayed_work *async_work;
+
+	if (atomic_read(&delayed_root->items) < BTRFS_DELAYED_BACKGROUND)
+		return 0;
+
+	async_work = kmalloc(sizeof(*async_work), GFP_NOFS);
+	if (!async_work)
+		return -ENOMEM;
+
+	async_work->delayed_root = delayed_root;
+	btrfs_init_work(&async_work->work, btrfs_async_run_delayed_root,
+			NULL, NULL);
+	async_work->nr = nr;
+
+	btrfs_queue_work(root->fs_info->delayed_workers, &async_work->work);
+	return 0;
+}
+
+void btrfs_assert_delayed_root_empty(struct btrfs_root *root)
+{
+	struct btrfs_delayed_root *delayed_root;
+	delayed_root = btrfs_get_delayed_root(root);
+	WARN_ON(btrfs_first_delayed_node(delayed_root));
+}
+
+static int could_end_wait(struct btrfs_delayed_root *delayed_root, int seq)
+{
+	int val = atomic_read(&delayed_root->items_seq);
+
+	if (val < seq || val >= seq + BTRFS_DELAYED_BATCH)
+		return 1;
+
+	if (atomic_read(&delayed_root->items) < BTRFS_DELAYED_BACKGROUND)
+		return 1;
+
+	return 0;
+}
+
+void btrfs_balance_delayed_items(struct btrfs_root *root)
+{
+	struct btrfs_delayed_root *delayed_root;
+
+	delayed_root = btrfs_get_delayed_root(root);
+
+	if (atomic_read(&delayed_root->items) < BTRFS_DELAYED_BACKGROUND)
+		return;
+
+	if (atomic_read(&delayed_root->items) >= BTRFS_DELAYED_WRITEBACK) {
+		int seq;
+		int ret;
+
+		seq = atomic_read(&delayed_root->items_seq);
+
+		ret = btrfs_wq_run_delayed_node(delayed_root, root, 0);
+		if (ret)
+			return;
+
+		wait_event_interruptible(delayed_root->wait,
+					 could_end_wait(delayed_root, seq));
+		return;
+	}
+
+	btrfs_wq_run_delayed_node(delayed_root, root, BTRFS_DELAYED_BATCH);
+}
+
+/* Will return 0 or -ENOMEM */
+int btrfs_insert_delayed_dir_index(struct btrfs_trans_handle *trans,
+				   struct btrfs_root *root, const char *name,
+				   int name_len, struct inode *dir,
+				   struct btrfs_disk_key *disk_key, u8 type,
+				   u64 index)
+{
+	struct btrfs_delayed_node *delayed_node;
+	struct btrfs_delayed_item *delayed_item;
+	struct btrfs_dir_item *dir_item;
+	int ret;
+
+	delayed_node = btrfs_get_or_create_delayed_node(dir);
+	if (IS_ERR(delayed_node))
+		return PTR_ERR(delayed_node);
+
+	delayed_item = btrfs_alloc_delayed_item(sizeof(*dir_item) + name_len);
+	if (!delayed_item) {
+		ret = -ENOMEM;
+		goto release_node;
+	}
+
+	delayed_item->key.objectid = btrfs_ino(dir);
+	btrfs_set_key_type(&delayed_item->key, BTRFS_DIR_INDEX_KEY);
+	delayed_item->key.offset = index;
+
+	dir_item = (struct btrfs_dir_item *)delayed_item->data;
+	dir_item->location = *disk_key;
+	btrfs_set_stack_dir_transid(dir_item, trans->transid);
+	btrfs_set_stack_dir_data_len(dir_item, 0);
+	btrfs_set_stack_dir_name_len(dir_item, name_len);
+	btrfs_set_stack_dir_type(dir_item, type);
+	memcpy((char *)(dir_item + 1), name, name_len);
+
+	ret = btrfs_delayed_item_reserve_metadata(trans, root, delayed_item);
+	/*
+	 * we have reserved enough space when we start a new transaction,
+	 * so reserving metadata failure is impossible
+	 */
+	BUG_ON(ret);
+
+
+	mutex_lock(&delayed_node->mutex);
+	ret = __btrfs_add_delayed_insertion_item(delayed_node, delayed_item);
+	if (unlikely(ret)) {
+		btrfs_err(root->fs_info, "err add delayed dir index item(name: %.*s) "
+				"into the insertion tree of the delayed node"
+				"(root id: %llu, inode id: %llu, errno: %d)",
+				name_len, name, delayed_node->root->objectid,
+				delayed_node->inode_id, ret);
+		BUG();
+	}
+	mutex_unlock(&delayed_node->mutex);
+
+release_node:
+	btrfs_release_delayed_node(delayed_node);
+	return ret;
+}
+
+static int btrfs_delete_delayed_insertion_item(struct btrfs_root *root,
+					       struct btrfs_delayed_node *node,
+					       struct btrfs_key *key)
+{
+	struct btrfs_delayed_item *item;
+
+	mutex_lock(&node->mutex);
+	item = __btrfs_lookup_delayed_insertion_item(node, key);
+	if (!item) {
+		mutex_unlock(&node->mutex);
+		return 1;
+	}
+
+	btrfs_delayed_item_release_metadata(root, item);
+	btrfs_release_delayed_item(item);
+	mutex_unlock(&node->mutex);
+	return 0;
+}
+
+int btrfs_delete_delayed_dir_index(struct btrfs_trans_handle *trans,
+				   struct btrfs_root *root, struct inode *dir,
+				   u64 index)
+{
+	struct btrfs_delayed_node *node;
+	struct btrfs_delayed_item *item;
+	struct btrfs_key item_key;
+	int ret;
+
+	node = btrfs_get_or_create_delayed_node(dir);
+	if (IS_ERR(node))
+		return PTR_ERR(node);
+
+	item_key.objectid = btrfs_ino(dir);
+	btrfs_set_key_type(&item_key, BTRFS_DIR_INDEX_KEY);
+	item_key.offset = index;
+
+	ret = btrfs_delete_delayed_insertion_item(root, node, &item_key);
+	if (!ret)
+		goto end;
+
+	item = btrfs_alloc_delayed_item(0);
+	if (!item) {
+		ret = -ENOMEM;
+		goto end;
+	}
+
+	item->key = item_key;
+
+	ret = btrfs_delayed_item_reserve_metadata(trans, root, item);
+	/*
+	 * we have reserved enough space when we start a new transaction,
+	 * so reserving metadata failure is impossible.
+	 */
+	BUG_ON(ret);
+
+	mutex_lock(&node->mutex);
+	ret = __btrfs_add_delayed_deletion_item(node, item);
+	if (unlikely(ret)) {
+		btrfs_err(root->fs_info, "err add delayed dir index item(index: %llu) "
+				"into the deletion tree of the delayed node"
+				"(root id: %llu, inode id: %llu, errno: %d)",
+				index, node->root->objectid, node->inode_id,
+				ret);
+		BUG();
+	}
+	mutex_unlock(&node->mutex);
+end:
+	btrfs_release_delayed_node(node);
+	return ret;
+}
+
+int btrfs_inode_delayed_dir_index_count(struct inode *inode)
+{
+	struct btrfs_delayed_node *delayed_node = btrfs_get_delayed_node(inode);
+
+	if (!delayed_node)
+		return -ENOENT;
+
+	/*
+	 * Since we have held i_mutex of this directory, it is impossible that
+	 * a new directory index is added into the delayed node and index_cnt
+	 * is updated now. So we needn't lock the delayed node.
+	 */
+	if (!delayed_node->index_cnt) {
+		btrfs_release_delayed_node(delayed_node);
+		return -EINVAL;
+	}
+
+	BTRFS_I(inode)->index_cnt = delayed_node->index_cnt;
+	btrfs_release_delayed_node(delayed_node);
+	return 0;
+}
+
+void btrfs_get_delayed_items(struct inode *inode, struct list_head *ins_list,
+			     struct list_head *del_list)
+{
+	struct btrfs_delayed_node *delayed_node;
+	struct btrfs_delayed_item *item;
+
+	delayed_node = btrfs_get_delayed_node(inode);
+	if (!delayed_node)
+		return;
+
+	mutex_lock(&delayed_node->mutex);
+	item = __btrfs_first_delayed_insertion_item(delayed_node);
+	while (item) {
+		atomic_inc(&item->refs);
+		list_add_tail(&item->readdir_list, ins_list);
+		item = __btrfs_next_delayed_item(item);
+	}
+
+	item = __btrfs_first_delayed_deletion_item(delayed_node);
+	while (item) {
+		atomic_inc(&item->refs);
+		list_add_tail(&item->readdir_list, del_list);
+		item = __btrfs_next_delayed_item(item);
+	}
+	mutex_unlock(&delayed_node->mutex);
+	/*
+	 * This delayed node is still cached in the btrfs inode, so refs
+	 * must be > 1 now, and we needn't check it is going to be freed
+	 * or not.
+	 *
+	 * Besides that, this function is used to read dir, we do not
+	 * insert/delete delayed items in this period. So we also needn't
+	 * requeue or dequeue this delayed node.
+	 */
+	atomic_dec(&delayed_node->refs);
+}
+
+void btrfs_put_delayed_items(struct list_head *ins_list,
+			     struct list_head *del_list)
+{
+	struct btrfs_delayed_item *curr, *next;
+
+	list_for_each_entry_safe(curr, next, ins_list, readdir_list) {
+		list_del(&curr->readdir_list);
+		if (atomic_dec_and_test(&curr->refs))
+			kfree(curr);
+	}
+
+	list_for_each_entry_safe(curr, next, del_list, readdir_list) {
+		list_del(&curr->readdir_list);
+		if (atomic_dec_and_test(&curr->refs))
+			kfree(curr);
+	}
+}
+
+int btrfs_should_delete_dir_index(struct list_head *del_list,
+				  u64 index)
+{
+	struct btrfs_delayed_item *curr, *next;
+	int ret;
+
+	if (list_empty(del_list))
+		return 0;
+
+	list_for_each_entry_safe(curr, next, del_list, readdir_list) {
+		if (curr->key.offset > index)
+			break;
+
+		list_del(&curr->readdir_list);
+		ret = (curr->key.offset == index);
+
+		if (atomic_dec_and_test(&curr->refs))
+			kfree(curr);
+
+		if (ret)
+			return 1;
+		else
+			continue;
+	}
+	return 0;
+}
+
+/*
+ * btrfs_readdir_delayed_dir_index - read dir info stored in the delayed tree
+ *
+ */
+int btrfs_readdir_delayed_dir_index(struct dir_context *ctx,
+				    struct list_head *ins_list)
+{
+	struct btrfs_dir_item *di;
+	struct btrfs_delayed_item *curr, *next;
+	struct btrfs_key location;
+	char *name;
+	int name_len;
+	int over = 0;
+	unsigned char d_type;
+
+	if (list_empty(ins_list))
+		return 0;
+
+	/*
+	 * Changing the data of the delayed item is impossible. So
+	 * we needn't lock them. And we have held i_mutex of the
+	 * directory, nobody can delete any directory indexes now.
+	 */
+	list_for_each_entry_safe(curr, next, ins_list, readdir_list) {
+		list_del(&curr->readdir_list);
+
+		if (curr->key.offset < ctx->pos) {
+			if (atomic_dec_and_test(&curr->refs))
+				kfree(curr);
+			continue;
+		}
+
+		ctx->pos = curr->key.offset;
+
+		di = (struct btrfs_dir_item *)curr->data;
+		name = (char *)(di + 1);
+		name_len = btrfs_stack_dir_name_len(di);
+
+		d_type = btrfs_filetype_table[di->type];
+		btrfs_disk_key_to_cpu(&location, &di->location);
+
+		over = !dir_emit(ctx, name, name_len,
+			       location.objectid, d_type);
+
+		if (atomic_dec_and_test(&curr->refs))
+			kfree(curr);
+
+		if (over)
+			return 1;
+	}
+	return 0;
+}
+
+static void fill_stack_inode_item(struct btrfs_trans_handle *trans,
+				  struct btrfs_inode_item *inode_item,
+				  struct inode *inode)
+{
+	btrfs_set_stack_inode_uid(inode_item, i_uid_read(inode));
+	btrfs_set_stack_inode_gid(inode_item, i_gid_read(inode));
+	btrfs_set_stack_inode_size(inode_item, BTRFS_I(inode)->disk_i_size);
+	btrfs_set_stack_inode_mode(inode_item, inode->i_mode);
+	btrfs_set_stack_inode_nlink(inode_item, inode->i_nlink);
+	btrfs_set_stack_inode_nbytes(inode_item, inode_get_bytes(inode));
+	btrfs_set_stack_inode_generation(inode_item,
+					 BTRFS_I(inode)->generation);
+	btrfs_set_stack_inode_sequence(inode_item, inode->i_version);
+	btrfs_set_stack_inode_transid(inode_item, trans->transid);
+	btrfs_set_stack_inode_rdev(inode_item, inode->i_rdev);
+	btrfs_set_stack_inode_flags(inode_item, BTRFS_I(inode)->flags);
+	btrfs_set_stack_inode_block_group(inode_item, 0);
+
+	btrfs_set_stack_timespec_sec(btrfs_inode_atime(inode_item),
+				     inode->i_atime.tv_sec);
+	btrfs_set_stack_timespec_nsec(btrfs_inode_atime(inode_item),
+				      inode->i_atime.tv_nsec);
+
+	btrfs_set_stack_timespec_sec(btrfs_inode_mtime(inode_item),
+				     inode->i_mtime.tv_sec);
+	btrfs_set_stack_timespec_nsec(btrfs_inode_mtime(inode_item),
+				      inode->i_mtime.tv_nsec);
+
+	btrfs_set_stack_timespec_sec(btrfs_inode_ctime(inode_item),
+				     inode->i_ctime.tv_sec);
+	btrfs_set_stack_timespec_nsec(btrfs_inode_ctime(inode_item),
+				      inode->i_ctime.tv_nsec);
+}
+
+int btrfs_fill_inode(struct inode *inode, u32 *rdev)
+{
+	struct btrfs_delayed_node *delayed_node;
+	struct btrfs_inode_item *inode_item;
+	struct btrfs_timespec *tspec;
+
+	delayed_node = btrfs_get_delayed_node(inode);
+	if (!delayed_node)
+		return -ENOENT;
+
+	mutex_lock(&delayed_node->mutex);
+	if (!test_bit(BTRFS_DELAYED_NODE_INODE_DIRTY, &delayed_node->flags)) {
+		mutex_unlock(&delayed_node->mutex);
+		btrfs_release_delayed_node(delayed_node);
+		return -ENOENT;
+	}
+
+	inode_item = &delayed_node->inode_item;
+
+	i_uid_write(inode, btrfs_stack_inode_uid(inode_item));
+	i_gid_write(inode, btrfs_stack_inode_gid(inode_item));
+	btrfs_i_size_write(inode, btrfs_stack_inode_size(inode_item));
+	inode->i_mode = btrfs_stack_inode_mode(inode_item);
+	set_nlink(inode, btrfs_stack_inode_nlink(inode_item));
+	inode_set_bytes(inode, btrfs_stack_inode_nbytes(inode_item));
+	BTRFS_I(inode)->generation = btrfs_stack_inode_generation(inode_item);
+	inode->i_version = btrfs_stack_inode_sequence(inode_item);
+	inode->i_rdev = 0;
+	*rdev = btrfs_stack_inode_rdev(inode_item);
+	BTRFS_I(inode)->flags = btrfs_stack_inode_flags(inode_item);
+
+	tspec = btrfs_inode_atime(inode_item);
+	inode->i_atime.tv_sec = btrfs_stack_timespec_sec(tspec);
+	inode->i_atime.tv_nsec = btrfs_stack_timespec_nsec(tspec);
+
+	tspec = btrfs_inode_mtime(inode_item);
+	inode->i_mtime.tv_sec = btrfs_stack_timespec_sec(tspec);
+	inode->i_mtime.tv_nsec = btrfs_stack_timespec_nsec(tspec);
+
+	tspec = btrfs_inode_ctime(inode_item);
+	inode->i_ctime.tv_sec = btrfs_stack_timespec_sec(tspec);
+	inode->i_ctime.tv_nsec = btrfs_stack_timespec_nsec(tspec);
+
+	inode->i_generation = BTRFS_I(inode)->generation;
+	BTRFS_I(inode)->index_cnt = (u64)-1;
+
+	mutex_unlock(&delayed_node->mutex);
+	btrfs_release_delayed_node(delayed_node);
+	return 0;
+}
+
+int btrfs_delayed_update_inode(struct btrfs_trans_handle *trans,
+			       struct btrfs_root *root, struct inode *inode)
+{
+	struct btrfs_delayed_node *delayed_node;
+	int ret = 0;
+
+	delayed_node = btrfs_get_or_create_delayed_node(inode);
+	if (IS_ERR(delayed_node))
+		return PTR_ERR(delayed_node);
+
+	mutex_lock(&delayed_node->mutex);
+	if (test_bit(BTRFS_DELAYED_NODE_INODE_DIRTY, &delayed_node->flags)) {
+		fill_stack_inode_item(trans, &delayed_node->inode_item, inode);
+		goto release_node;
+	}
+
+	ret = btrfs_delayed_inode_reserve_metadata(trans, root, inode,
+						   delayed_node);
+	if (ret)
+		goto release_node;
+
+	fill_stack_inode_item(trans, &delayed_node->inode_item, inode);
+	set_bit(BTRFS_DELAYED_NODE_INODE_DIRTY, &delayed_node->flags);
+	delayed_node->count++;
+	atomic_inc(&root->fs_info->delayed_root->items);
+release_node:
+	mutex_unlock(&delayed_node->mutex);
+	btrfs_release_delayed_node(delayed_node);
+	return ret;
+}
+
+int btrfs_delayed_delete_inode_ref(struct inode *inode)
+{
+	struct btrfs_delayed_node *delayed_node;
+
+	delayed_node = btrfs_get_or_create_delayed_node(inode);
+	if (IS_ERR(delayed_node))
+		return PTR_ERR(delayed_node);
+
+	/*
+	 * We don't reserve space for inode ref deletion is because:
+	 * - We ONLY do async inode ref deletion for the inode who has only
+	 *   one link(i_nlink == 1), it means there is only one inode ref.
+	 *   And in most case, the inode ref and the inode item are in the
+	 *   same leaf, and we will deal with them at the same time.
+	 *   Since we are sure we will reserve the space for the inode item,
+	 *   it is unnecessary to reserve space for inode ref deletion.
+	 * - If the inode ref and the inode item are not in the same leaf,
+	 *   We also needn't worry about enospc problem, because we reserve
+	 *   much more space for the inode update than it needs.
+	 * - At the worst, we can steal some space from the global reservation.
+	 *   It is very rare.
+	 */
+	mutex_lock(&delayed_node->mutex);
+	if (test_bit(BTRFS_DELAYED_NODE_DEL_IREF, &delayed_node->flags))
+		goto release_node;
+
+	set_bit(BTRFS_DELAYED_NODE_DEL_IREF, &delayed_node->flags);
+	delayed_node->count++;
+	atomic_inc(&BTRFS_I(inode)->root->fs_info->delayed_root->items);
+release_node:
+	mutex_unlock(&delayed_node->mutex);
+	btrfs_release_delayed_node(delayed_node);
+	return 0;
+}
+
+static void __btrfs_kill_delayed_node(struct btrfs_delayed_node *delayed_node)
+{
+	struct btrfs_root *root = delayed_node->root;
+	struct btrfs_delayed_item *curr_item, *prev_item;
+
+	mutex_lock(&delayed_node->mutex);
+	curr_item = __btrfs_first_delayed_insertion_item(delayed_node);
+	while (curr_item) {
+		btrfs_delayed_item_release_metadata(root, curr_item);
+		prev_item = curr_item;
+		curr_item = __btrfs_next_delayed_item(prev_item);
+		btrfs_release_delayed_item(prev_item);
+	}
+
+	curr_item = __btrfs_first_delayed_deletion_item(delayed_node);
+	while (curr_item) {
+		btrfs_delayed_item_release_metadata(root, curr_item);
+		prev_item = curr_item;
+		curr_item = __btrfs_next_delayed_item(prev_item);
+		btrfs_release_delayed_item(prev_item);
+	}
+
+	if (test_bit(BTRFS_DELAYED_NODE_DEL_IREF, &delayed_node->flags))
+		btrfs_release_delayed_iref(delayed_node);
+
+	if (test_bit(BTRFS_DELAYED_NODE_INODE_DIRTY, &delayed_node->flags)) {
+		btrfs_delayed_inode_release_metadata(root, delayed_node);
+		btrfs_release_delayed_inode(delayed_node);
+	}
+	mutex_unlock(&delayed_node->mutex);
+}
+
+void btrfs_kill_delayed_inode_items(struct inode *inode)
+{
+	struct btrfs_delayed_node *delayed_node;
+
+	delayed_node = btrfs_get_delayed_node(inode);
+	if (!delayed_node)
+		return;
+
+	__btrfs_kill_delayed_node(delayed_node);
+	btrfs_release_delayed_node(delayed_node);
+}
+
+void btrfs_kill_all_delayed_nodes(struct btrfs_root *root)
+{
+	u64 inode_id = 0;
+	struct btrfs_delayed_node *delayed_nodes[8];
+	int i, n;
+
+	while (1) {
+		spin_lock(&root->inode_lock);
+		n = radix_tree_gang_lookup(&root->delayed_nodes_tree,
+					   (void **)delayed_nodes, inode_id,
+					   ARRAY_SIZE(delayed_nodes));
+		if (!n) {
+			spin_unlock(&root->inode_lock);
+			break;
+		}
+
+		inode_id = delayed_nodes[n - 1]->inode_id + 1;
+
+		for (i = 0; i < n; i++)
+			atomic_inc(&delayed_nodes[i]->refs);
+		spin_unlock(&root->inode_lock);
+
+		for (i = 0; i < n; i++) {
+			__btrfs_kill_delayed_node(delayed_nodes[i]);
+			btrfs_release_delayed_node(delayed_nodes[i]);
+		}
+	}
+}
+
+void btrfs_destroy_delayed_inodes(struct btrfs_root *root)
+{
+	struct btrfs_delayed_root *delayed_root;
+	struct btrfs_delayed_node *curr_node, *prev_node;
+
+	delayed_root = btrfs_get_delayed_root(root);
+
+	curr_node = btrfs_first_delayed_node(delayed_root);
+	while (curr_node) {
+		__btrfs_kill_delayed_node(curr_node);
+
+		prev_node = curr_node;
+		curr_node = btrfs_next_delayed_node(curr_node);
+		btrfs_release_delayed_node(prev_node);
+	}
+}
+
diff --git a/fs/btrfs/delayed-inode.h b/fs/btrfs/delayed-inode.h
new file mode 100644
index 00000000000..f70119f2542
--- /dev/null
+++ b/fs/btrfs/delayed-inode.h
@@ -0,0 +1,156 @@
+/*
+ * Copyright (C) 2011 Fujitsu.  All rights reserved.
+ * Written by Miao Xie <miaox@cn.fujitsu.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef __DELAYED_TREE_OPERATION_H
+#define __DELAYED_TREE_OPERATION_H
+
+#include <linux/rbtree.h>
+#include <linux/spinlock.h>
+#include <linux/mutex.h>
+#include <linux/list.h>
+#include <linux/wait.h>
+#include <linux/atomic.h>
+
+#include "ctree.h"
+
+/* types of the delayed item */
+#define BTRFS_DELAYED_INSERTION_ITEM	1
+#define BTRFS_DELAYED_DELETION_ITEM	2
+
+struct btrfs_delayed_root {
+	spinlock_t lock;
+	struct list_head node_list;
+	/*
+	 * Used for delayed nodes which is waiting to be dealt with by the
+	 * worker. If the delayed node is inserted into the work queue, we
+	 * drop it from this list.
+	 */
+	struct list_head prepare_list;
+	atomic_t items;		/* for delayed items */
+	atomic_t items_seq;	/* for delayed items */
+	int nodes;		/* for delayed nodes */
+	wait_queue_head_t wait;
+};
+
+#define BTRFS_DELAYED_NODE_IN_LIST	0
+#define BTRFS_DELAYED_NODE_INODE_DIRTY	1
+#define BTRFS_DELAYED_NODE_DEL_IREF	2
+
+struct btrfs_delayed_node {
+	u64 inode_id;
+	u64 bytes_reserved;
+	struct btrfs_root *root;
+	/* Used to add the node into the delayed root's node list. */
+	struct list_head n_list;
+	/*
+	 * Used to add the node into the prepare list, the nodes in this list
+	 * is waiting to be dealt with by the async worker.
+	 */
+	struct list_head p_list;
+	struct rb_root ins_root;
+	struct rb_root del_root;
+	struct mutex mutex;
+	struct btrfs_inode_item inode_item;
+	atomic_t refs;
+	u64 index_cnt;
+	unsigned long flags;
+	int count;
+};
+
+struct btrfs_delayed_item {
+	struct rb_node rb_node;
+	struct btrfs_key key;
+	struct list_head tree_list;	/* used for batch insert/delete items */
+	struct list_head readdir_list;	/* used for readdir items */
+	u64 bytes_reserved;
+	struct btrfs_delayed_node *delayed_node;
+	atomic_t refs;
+	int ins_or_del;
+	u32 data_len;
+	char data[0];
+};
+
+static inline void btrfs_init_delayed_root(
+				struct btrfs_delayed_root *delayed_root)
+{
+	atomic_set(&delayed_root->items, 0);
+	atomic_set(&delayed_root->items_seq, 0);
+	delayed_root->nodes = 0;
+	spin_lock_init(&delayed_root->lock);
+	init_waitqueue_head(&delayed_root->wait);
+	INIT_LIST_HEAD(&delayed_root->node_list);
+	INIT_LIST_HEAD(&delayed_root->prepare_list);
+}
+
+int btrfs_insert_delayed_dir_index(struct btrfs_trans_handle *trans,
+				   struct btrfs_root *root, const char *name,
+				   int name_len, struct inode *dir,
+				   struct btrfs_disk_key *disk_key, u8 type,
+				   u64 index);
+
+int btrfs_delete_delayed_dir_index(struct btrfs_trans_handle *trans,
+				   struct btrfs_root *root, struct inode *dir,
+				   u64 index);
+
+int btrfs_inode_delayed_dir_index_count(struct inode *inode);
+
+int btrfs_run_delayed_items(struct btrfs_trans_handle *trans,
+			    struct btrfs_root *root);
+int btrfs_run_delayed_items_nr(struct btrfs_trans_handle *trans,
+			       struct btrfs_root *root, int nr);
+
+void btrfs_balance_delayed_items(struct btrfs_root *root);
+
+int btrfs_commit_inode_delayed_items(struct btrfs_trans_handle *trans,
+				     struct inode *inode);
+/* Used for evicting the inode. */
+void btrfs_remove_delayed_node(struct inode *inode);
+void btrfs_kill_delayed_inode_items(struct inode *inode);
+int btrfs_commit_inode_delayed_inode(struct inode *inode);
+
+
+int btrfs_delayed_update_inode(struct btrfs_trans_handle *trans,
+			       struct btrfs_root *root, struct inode *inode);
+int btrfs_fill_inode(struct inode *inode, u32 *rdev);
+int btrfs_delayed_delete_inode_ref(struct inode *inode);
+
+/* Used for drop dead root */
+void btrfs_kill_all_delayed_nodes(struct btrfs_root *root);
+
+/* Used for clean the transaction */
+void btrfs_destroy_delayed_inodes(struct btrfs_root *root);
+
+/* Used for readdir() */
+void btrfs_get_delayed_items(struct inode *inode, struct list_head *ins_list,
+			     struct list_head *del_list);
+void btrfs_put_delayed_items(struct list_head *ins_list,
+			     struct list_head *del_list);
+int btrfs_should_delete_dir_index(struct list_head *del_list,
+				  u64 index);
+int btrfs_readdir_delayed_dir_index(struct dir_context *ctx,
+				    struct list_head *ins_list);
+
+/* for init */
+int __init btrfs_delayed_inode_init(void);
+void btrfs_delayed_inode_exit(void);
+
+/* for debugging */
+void btrfs_assert_delayed_root_empty(struct btrfs_root *root);
+
+#endif
diff --git a/fs/btrfs/delayed-ref.c b/fs/btrfs/delayed-ref.c
index 902ce507c4e..6d16bea94e1 100644
--- a/fs/btrfs/delayed-ref.c
+++ b/fs/btrfs/delayed-ref.c
@@ -23,6 +23,10 @@
 #include "delayed-ref.h"
 #include "transaction.h"
 
+struct kmem_cache *btrfs_delayed_ref_head_cachep;
+struct kmem_cache *btrfs_delayed_tree_ref_cachep;
+struct kmem_cache *btrfs_delayed_data_ref_cachep;
+struct kmem_cache *btrfs_delayed_extent_op_cachep;
 /*
  * delayed back reference update tracking.  For subvolume trees
  * we queue up extent allocations and backref maintenance for
@@ -36,9 +40,9 @@
  * compare two delayed tree backrefs with same bytenr and type
  */
 static int comp_tree_refs(struct btrfs_delayed_tree_ref *ref2,
-			  struct btrfs_delayed_tree_ref *ref1)
+			  struct btrfs_delayed_tree_ref *ref1, int type)
 {
-	if (ref1->node.type == BTRFS_TREE_BLOCK_REF_KEY) {
+	if (type == BTRFS_TREE_BLOCK_REF_KEY) {
 		if (ref1->root < ref2->root)
 			return -1;
 		if (ref1->root > ref2->root)
@@ -85,7 +89,8 @@ static int comp_data_refs(struct btrfs_delayed_data_ref *ref2,
  * type of the delayed backrefs and content of delayed backrefs.
  */
 static int comp_entry(struct btrfs_delayed_ref_node *ref2,
-		      struct btrfs_delayed_ref_node *ref1)
+		      struct btrfs_delayed_ref_node *ref1,
+		      bool compare_seq)
 {
 	if (ref1->bytenr < ref2->bytenr)
 		return -1;
@@ -101,10 +106,22 @@ static int comp_entry(struct btrfs_delayed_ref_node *ref2,
 		return -1;
 	if (ref1->type > ref2->type)
 		return 1;
+	if (ref1->no_quota > ref2->no_quota)
+		return 1;
+	if (ref1->no_quota < ref2->no_quota)
+		return -1;
+	/* merging of sequenced refs is not allowed */
+	if (compare_seq) {
+		if (ref1->seq < ref2->seq)
+			return -1;
+		if (ref1->seq > ref2->seq)
+			return 1;
+	}
 	if (ref1->type == BTRFS_TREE_BLOCK_REF_KEY ||
 	    ref1->type == BTRFS_SHARED_BLOCK_REF_KEY) {
 		return comp_tree_refs(btrfs_delayed_node_to_tree_ref(ref2),
-				      btrfs_delayed_node_to_tree_ref(ref1));
+				      btrfs_delayed_node_to_tree_ref(ref1),
+				      ref1->type);
 	} else if (ref1->type == BTRFS_EXTENT_DATA_REF_KEY ||
 		   ref1->type == BTRFS_SHARED_DATA_REF_KEY) {
 		return comp_data_refs(btrfs_delayed_node_to_data_ref(ref2),
@@ -134,7 +151,7 @@ static struct btrfs_delayed_ref_node *tree_insert(struct rb_root *root,
 		entry = rb_entry(parent_node, struct btrfs_delayed_ref_node,
 				 rb_node);
 
-		cmp = comp_entry(entry, ins);
+		cmp = comp_entry(entry, ins, 1);
 		if (cmp < 0)
 			p = &(*p)->rb_left;
 		else if (cmp > 0)
@@ -148,40 +165,72 @@ static struct btrfs_delayed_ref_node *tree_insert(struct rb_root *root,
 	return NULL;
 }
 
+/* insert a new ref to head ref rbtree */
+static struct btrfs_delayed_ref_head *htree_insert(struct rb_root *root,
+						   struct rb_node *node)
+{
+	struct rb_node **p = &root->rb_node;
+	struct rb_node *parent_node = NULL;
+	struct btrfs_delayed_ref_head *entry;
+	struct btrfs_delayed_ref_head *ins;
+	u64 bytenr;
+
+	ins = rb_entry(node, struct btrfs_delayed_ref_head, href_node);
+	bytenr = ins->node.bytenr;
+	while (*p) {
+		parent_node = *p;
+		entry = rb_entry(parent_node, struct btrfs_delayed_ref_head,
+				 href_node);
+
+		if (bytenr < entry->node.bytenr)
+			p = &(*p)->rb_left;
+		else if (bytenr > entry->node.bytenr)
+			p = &(*p)->rb_right;
+		else
+			return entry;
+	}
+
+	rb_link_node(node, parent_node, p);
+	rb_insert_color(node, root);
+	return NULL;
+}
+
 /*
  * find an head entry based on bytenr. This returns the delayed ref
- * head if it was able to find one, or NULL if nothing was in that spot
+ * head if it was able to find one, or NULL if nothing was in that spot.
+ * If return_bigger is given, the next bigger entry is returned if no exact
+ * match is found.
  */
-static struct btrfs_delayed_ref_node *find_ref_head(struct rb_root *root,
-				  u64 bytenr,
-				  struct btrfs_delayed_ref_node **last)
+static struct btrfs_delayed_ref_head *
+find_ref_head(struct rb_root *root, u64 bytenr,
+	      int return_bigger)
 {
-	struct rb_node *n = root->rb_node;
-	struct btrfs_delayed_ref_node *entry;
-	int cmp;
+	struct rb_node *n;
+	struct btrfs_delayed_ref_head *entry;
 
+	n = root->rb_node;
+	entry = NULL;
 	while (n) {
-		entry = rb_entry(n, struct btrfs_delayed_ref_node, rb_node);
-		WARN_ON(!entry->in_tree);
-		if (last)
-			*last = entry;
-
-		if (bytenr < entry->bytenr)
-			cmp = -1;
-		else if (bytenr > entry->bytenr)
-			cmp = 1;
-		else if (!btrfs_delayed_ref_is_head(entry))
-			cmp = 1;
-		else
-			cmp = 0;
+		entry = rb_entry(n, struct btrfs_delayed_ref_head, href_node);
 
-		if (cmp < 0)
+		if (bytenr < entry->node.bytenr)
 			n = n->rb_left;
-		else if (cmp > 0)
+		else if (bytenr > entry->node.bytenr)
 			n = n->rb_right;
 		else
 			return entry;
 	}
+	if (entry && return_bigger) {
+		if (bytenr > entry->node.bytenr) {
+			n = rb_next(&entry->href_node);
+			if (!n)
+				n = rb_first(root);
+			entry = rb_entry(n, struct btrfs_delayed_ref_head,
+					 href_node);
+			return entry;
+		}
+		return entry;
+	}
 	return NULL;
 }
 
@@ -209,214 +258,188 @@ int btrfs_delayed_ref_lock(struct btrfs_trans_handle *trans,
 	return 0;
 }
 
-int btrfs_find_ref_cluster(struct btrfs_trans_handle *trans,
-			   struct list_head *cluster, u64 start)
+static inline void drop_delayed_ref(struct btrfs_trans_handle *trans,
+				    struct btrfs_delayed_ref_root *delayed_refs,
+				    struct btrfs_delayed_ref_head *head,
+				    struct btrfs_delayed_ref_node *ref)
 {
-	int count = 0;
-	struct btrfs_delayed_ref_root *delayed_refs;
-	struct rb_node *node;
-	struct btrfs_delayed_ref_node *ref;
-	struct btrfs_delayed_ref_head *head;
-
-	delayed_refs = &trans->transaction->delayed_refs;
-	if (start == 0) {
-		node = rb_first(&delayed_refs->root);
+	if (btrfs_delayed_ref_is_head(ref)) {
+		head = btrfs_delayed_node_to_head(ref);
+		rb_erase(&head->href_node, &delayed_refs->href_root);
 	} else {
-		ref = NULL;
-		find_ref_head(&delayed_refs->root, start, &ref);
-		if (ref) {
-			struct btrfs_delayed_ref_node *tmp;
-
-			node = rb_prev(&ref->rb_node);
-			while (node) {
-				tmp = rb_entry(node,
-					       struct btrfs_delayed_ref_node,
-					       rb_node);
-				if (tmp->bytenr < start)
-					break;
-				ref = tmp;
-				node = rb_prev(&ref->rb_node);
-			}
-			node = &ref->rb_node;
-		} else
-			node = rb_first(&delayed_refs->root);
+		assert_spin_locked(&head->lock);
+		rb_erase(&ref->rb_node, &head->ref_root);
 	}
-again:
-	while (node && count < 32) {
-		ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);
-		if (btrfs_delayed_ref_is_head(ref)) {
-			head = btrfs_delayed_node_to_head(ref);
-			if (list_empty(&head->cluster)) {
-				list_add_tail(&head->cluster, cluster);
-				delayed_refs->run_delayed_start =
-					head->node.bytenr;
-				count++;
-
-				WARN_ON(delayed_refs->num_heads_ready == 0);
-				delayed_refs->num_heads_ready--;
-			} else if (count) {
-				/* the goal of the clustering is to find extents
-				 * that are likely to end up in the same extent
-				 * leaf on disk.  So, we don't want them spread
-				 * all over the tree.  Stop now if we've hit
-				 * a head that was already in use
-				 */
-				break;
+	ref->in_tree = 0;
+	btrfs_put_delayed_ref(ref);
+	atomic_dec(&delayed_refs->num_entries);
+	if (trans->delayed_ref_updates)
+		trans->delayed_ref_updates--;
+}
+
+static int merge_ref(struct btrfs_trans_handle *trans,
+		     struct btrfs_delayed_ref_root *delayed_refs,
+		     struct btrfs_delayed_ref_head *head,
+		     struct btrfs_delayed_ref_node *ref, u64 seq)
+{
+	struct rb_node *node;
+	int mod = 0;
+	int done = 0;
+
+	node = rb_next(&ref->rb_node);
+	while (!done && node) {
+		struct btrfs_delayed_ref_node *next;
+
+		next = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);
+		node = rb_next(node);
+		if (seq && next->seq >= seq)
+			break;
+		if (comp_entry(ref, next, 0))
+			continue;
+
+		if (ref->action == next->action) {
+			mod = next->ref_mod;
+		} else {
+			if (ref->ref_mod < next->ref_mod) {
+				struct btrfs_delayed_ref_node *tmp;
+
+				tmp = ref;
+				ref = next;
+				next = tmp;
+				done = 1;
 			}
+			mod = -next->ref_mod;
+		}
+
+		drop_delayed_ref(trans, delayed_refs, head, next);
+		ref->ref_mod += mod;
+		if (ref->ref_mod == 0) {
+			drop_delayed_ref(trans, delayed_refs, head, ref);
+			done = 1;
+		} else {
+			/*
+			 * You can't have multiples of the same ref on a tree
+			 * block.
+			 */
+			WARN_ON(ref->type == BTRFS_TREE_BLOCK_REF_KEY ||
+				ref->type == BTRFS_SHARED_BLOCK_REF_KEY);
 		}
-		node = rb_next(node);
-	}
-	if (count) {
-		return 0;
-	} else if (start) {
-		/*
-		 * we've gone to the end of the rbtree without finding any
-		 * clusters.  start from the beginning and try again
-		 */
-		start = 0;
-		node = rb_first(&delayed_refs->root);
-		goto again;
 	}
-	return 1;
+	return done;
 }
 
-/*
- * This checks to see if there are any delayed refs in the
- * btree for a given bytenr.  It returns one if it finds any
- * and zero otherwise.
- *
- * If it only finds a head node, it returns 0.
- *
- * The idea is to use this when deciding if you can safely delete an
- * extent from the extent allocation tree.  There may be a pending
- * ref in the rbtree that adds or removes references, so as long as this
- * returns one you need to leave the BTRFS_EXTENT_ITEM in the extent
- * allocation tree.
- */
-int btrfs_delayed_ref_pending(struct btrfs_trans_handle *trans, u64 bytenr)
+void btrfs_merge_delayed_refs(struct btrfs_trans_handle *trans,
+			      struct btrfs_fs_info *fs_info,
+			      struct btrfs_delayed_ref_root *delayed_refs,
+			      struct btrfs_delayed_ref_head *head)
 {
-	struct btrfs_delayed_ref_node *ref;
-	struct btrfs_delayed_ref_root *delayed_refs;
-	struct rb_node *prev_node;
-	int ret = 0;
+	struct rb_node *node;
+	u64 seq = 0;
 
-	delayed_refs = &trans->transaction->delayed_refs;
-	spin_lock(&delayed_refs->lock);
+	assert_spin_locked(&head->lock);
+	/*
+	 * We don't have too much refs to merge in the case of delayed data
+	 * refs.
+	 */
+	if (head->is_data)
+		return;
 
-	ref = find_ref_head(&delayed_refs->root, bytenr, NULL);
-	if (ref) {
-		prev_node = rb_prev(&ref->rb_node);
-		if (!prev_node)
-			goto out;
-		ref = rb_entry(prev_node, struct btrfs_delayed_ref_node,
+	spin_lock(&fs_info->tree_mod_seq_lock);
+	if (!list_empty(&fs_info->tree_mod_seq_list)) {
+		struct seq_list *elem;
+
+		elem = list_first_entry(&fs_info->tree_mod_seq_list,
+					struct seq_list, list);
+		seq = elem->seq;
+	}
+	spin_unlock(&fs_info->tree_mod_seq_lock);
+
+	node = rb_first(&head->ref_root);
+	while (node) {
+		struct btrfs_delayed_ref_node *ref;
+
+		ref = rb_entry(node, struct btrfs_delayed_ref_node,
 			       rb_node);
-		if (ref->bytenr == bytenr)
+		/* We can't merge refs that are outside of our seq count */
+		if (seq && ref->seq >= seq)
+			break;
+		if (merge_ref(trans, delayed_refs, head, ref, seq))
+			node = rb_first(&head->ref_root);
+		else
+			node = rb_next(&ref->rb_node);
+	}
+}
+
+int btrfs_check_delayed_seq(struct btrfs_fs_info *fs_info,
+			    struct btrfs_delayed_ref_root *delayed_refs,
+			    u64 seq)
+{
+	struct seq_list *elem;
+	int ret = 0;
+
+	spin_lock(&fs_info->tree_mod_seq_lock);
+	if (!list_empty(&fs_info->tree_mod_seq_list)) {
+		elem = list_first_entry(&fs_info->tree_mod_seq_list,
+					struct seq_list, list);
+		if (seq >= elem->seq) {
+			pr_debug("holding back delayed_ref %#x.%x, lowest is %#x.%x (%p)\n",
+				 (u32)(seq >> 32), (u32)seq,
+				 (u32)(elem->seq >> 32), (u32)elem->seq,
+				 delayed_refs);
 			ret = 1;
+		}
 	}
-out:
-	spin_unlock(&delayed_refs->lock);
+
+	spin_unlock(&fs_info->tree_mod_seq_lock);
 	return ret;
 }
 
-/*
- * helper function to lookup reference count and flags of extent.
- *
- * the head node for delayed ref is used to store the sum of all the
- * reference count modifications queued up in the rbtree. the head
- * node may also store the extent flags to set. This way you can check
- * to see what the reference count and extent flags would be if all of
- * the delayed refs are not processed.
- */
-int btrfs_lookup_extent_info(struct btrfs_trans_handle *trans,
-			     struct btrfs_root *root, u64 bytenr,
-			     u64 num_bytes, u64 *refs, u64 *flags)
+struct btrfs_delayed_ref_head *
+btrfs_select_ref_head(struct btrfs_trans_handle *trans)
 {
-	struct btrfs_delayed_ref_node *ref;
-	struct btrfs_delayed_ref_head *head;
 	struct btrfs_delayed_ref_root *delayed_refs;
-	struct btrfs_path *path;
-	struct btrfs_extent_item *ei;
-	struct extent_buffer *leaf;
-	struct btrfs_key key;
-	u32 item_size;
-	u64 num_refs;
-	u64 extent_flags;
-	int ret;
-
-	path = btrfs_alloc_path();
-	if (!path)
-		return -ENOMEM;
+	struct btrfs_delayed_ref_head *head;
+	u64 start;
+	bool loop = false;
 
-	key.objectid = bytenr;
-	key.type = BTRFS_EXTENT_ITEM_KEY;
-	key.offset = num_bytes;
 	delayed_refs = &trans->transaction->delayed_refs;
+
 again:
-	ret = btrfs_search_slot(trans, root->fs_info->extent_root,
-				&key, path, 0, 0);
-	if (ret < 0)
-		goto out;
-
-	if (ret == 0) {
-		leaf = path->nodes[0];
-		item_size = btrfs_item_size_nr(leaf, path->slots[0]);
-		if (item_size >= sizeof(*ei)) {
-			ei = btrfs_item_ptr(leaf, path->slots[0],
-					    struct btrfs_extent_item);
-			num_refs = btrfs_extent_refs(leaf, ei);
-			extent_flags = btrfs_extent_flags(leaf, ei);
-		} else {
-#ifdef BTRFS_COMPAT_EXTENT_TREE_V0
-			struct btrfs_extent_item_v0 *ei0;
-			BUG_ON(item_size != sizeof(*ei0));
-			ei0 = btrfs_item_ptr(leaf, path->slots[0],
-					     struct btrfs_extent_item_v0);
-			num_refs = btrfs_extent_refs_v0(leaf, ei0);
-			/* FIXME: this isn't correct for data */
-			extent_flags = BTRFS_BLOCK_FLAG_FULL_BACKREF;
-#else
-			BUG();
-#endif
-		}
-		BUG_ON(num_refs == 0);
-	} else {
-		num_refs = 0;
-		extent_flags = 0;
-		ret = 0;
+	start = delayed_refs->run_delayed_start;
+	head = find_ref_head(&delayed_refs->href_root, start, 1);
+	if (!head && !loop) {
+		delayed_refs->run_delayed_start = 0;
+		start = 0;
+		loop = true;
+		head = find_ref_head(&delayed_refs->href_root, start, 1);
+		if (!head)
+			return NULL;
+	} else if (!head && loop) {
+		return NULL;
 	}
 
-	spin_lock(&delayed_refs->lock);
-	ref = find_ref_head(&delayed_refs->root, bytenr, NULL);
-	if (ref) {
-		head = btrfs_delayed_node_to_head(ref);
-		if (!mutex_trylock(&head->mutex)) {
-			atomic_inc(&ref->refs);
-			spin_unlock(&delayed_refs->lock);
-
-			btrfs_release_path(root->fs_info->extent_root, path);
+	while (head->processing) {
+		struct rb_node *node;
 
-			mutex_lock(&head->mutex);
-			mutex_unlock(&head->mutex);
-			btrfs_put_delayed_ref(ref);
+		node = rb_next(&head->href_node);
+		if (!node) {
+			if (loop)
+				return NULL;
+			delayed_refs->run_delayed_start = 0;
+			start = 0;
+			loop = true;
 			goto again;
 		}
-		if (head->extent_op && head->extent_op->update_flags)
-			extent_flags |= head->extent_op->flags_to_set;
-		else
-			BUG_ON(num_refs == 0);
-
-		num_refs += ref->ref_mod;
-		mutex_unlock(&head->mutex);
+		head = rb_entry(node, struct btrfs_delayed_ref_head,
+				href_node);
 	}
-	WARN_ON(num_refs == 0);
-	if (refs)
-		*refs = num_refs;
-	if (flags)
-		*flags = extent_flags;
-out:
-	spin_unlock(&delayed_refs->lock);
-	btrfs_free_path(path);
-	return ret;
+
+	head->processing = 1;
+	WARN_ON(delayed_refs->num_heads_ready == 0);
+	delayed_refs->num_heads_ready--;
+	delayed_refs->run_delayed_start = head->node.bytenr +
+		head->node.num_bytes;
+	return head;
 }
 
 /*
@@ -430,6 +453,7 @@ out:
 static noinline void
 update_existing_ref(struct btrfs_trans_handle *trans,
 		    struct btrfs_delayed_ref_root *delayed_refs,
+		    struct btrfs_delayed_ref_head *head,
 		    struct btrfs_delayed_ref_node *existing,
 		    struct btrfs_delayed_ref_node *update)
 {
@@ -441,18 +465,11 @@ update_existing_ref(struct btrfs_trans_handle *trans,
 		 * every changing the extent allocation tree.
 		 */
 		existing->ref_mod--;
-		if (existing->ref_mod == 0) {
-			rb_erase(&existing->rb_node,
-				 &delayed_refs->root);
-			existing->in_tree = 0;
-			btrfs_put_delayed_ref(existing);
-			delayed_refs->num_entries--;
-			if (trans->delayed_ref_updates)
-				trans->delayed_ref_updates--;
-		} else {
+		if (existing->ref_mod == 0)
+			drop_delayed_ref(trans, delayed_refs, head, existing);
+		else
 			WARN_ON(existing->type == BTRFS_TREE_BLOCK_REF_KEY ||
 				existing->type == BTRFS_SHARED_BLOCK_REF_KEY);
-		}
 	} else {
 		WARN_ON(existing->type == BTRFS_TREE_BLOCK_REF_KEY ||
 			existing->type == BTRFS_SHARED_BLOCK_REF_KEY);
@@ -482,6 +499,7 @@ update_existing_head_ref(struct btrfs_delayed_ref_node *existing,
 	ref = btrfs_delayed_node_to_head(update);
 	BUG_ON(existing_ref->is_data != ref->is_data);
 
+	spin_lock(&existing_ref->lock);
 	if (ref->must_insert_reserved) {
 		/* if the extent was freed and then
 		 * reallocated before the delayed ref
@@ -515,13 +533,16 @@ update_existing_head_ref(struct btrfs_delayed_ref_node *existing,
 					ref->extent_op->flags_to_set;
 				existing_ref->extent_op->update_flags = 1;
 			}
-			kfree(ref->extent_op);
+			btrfs_free_delayed_extent_op(ref->extent_op);
 		}
 	}
 	/*
-	 * update the reference mod on the head to reflect this new operation
+	 * update the reference mod on the head to reflect this new operation,
+	 * only need the lock for this case cause we could be processing it
+	 * currently, for refs we just added we know we're a-ok.
 	 */
 	existing->ref_mod += update->ref_mod;
+	spin_unlock(&existing_ref->lock);
 }
 
 /*
@@ -529,12 +550,13 @@ update_existing_head_ref(struct btrfs_delayed_ref_node *existing,
  * this does all the dirty work in terms of maintaining the correct
  * overall modification count.
  */
-static noinline int add_delayed_ref_head(struct btrfs_trans_handle *trans,
-					struct btrfs_delayed_ref_node *ref,
-					u64 bytenr, u64 num_bytes,
-					int action, int is_data)
+static noinline struct btrfs_delayed_ref_head *
+add_delayed_ref_head(struct btrfs_fs_info *fs_info,
+		     struct btrfs_trans_handle *trans,
+		     struct btrfs_delayed_ref_node *ref, u64 bytenr,
+		     u64 num_bytes, int action, int is_data)
 {
-	struct btrfs_delayed_ref_node *existing;
+	struct btrfs_delayed_ref_head *existing;
 	struct btrfs_delayed_ref_head *head_ref = NULL;
 	struct btrfs_delayed_ref_root *delayed_refs;
 	int count_mod = 1;
@@ -576,47 +598,59 @@ static noinline int add_delayed_ref_head(struct btrfs_trans_handle *trans,
 	ref->action  = 0;
 	ref->is_head = 1;
 	ref->in_tree = 1;
+	ref->seq = 0;
 
 	head_ref = btrfs_delayed_node_to_head(ref);
 	head_ref->must_insert_reserved = must_insert_reserved;
 	head_ref->is_data = is_data;
+	head_ref->ref_root = RB_ROOT;
+	head_ref->processing = 0;
 
-	INIT_LIST_HEAD(&head_ref->cluster);
+	spin_lock_init(&head_ref->lock);
 	mutex_init(&head_ref->mutex);
 
-	existing = tree_insert(&delayed_refs->root, &ref->rb_node);
+	trace_add_delayed_ref_head(ref, head_ref, action);
 
+	existing = htree_insert(&delayed_refs->href_root,
+				&head_ref->href_node);
 	if (existing) {
-		update_existing_head_ref(existing, ref);
+		update_existing_head_ref(&existing->node, ref);
 		/*
 		 * we've updated the existing ref, free the newly
 		 * allocated ref
 		 */
-		kfree(ref);
+		kmem_cache_free(btrfs_delayed_ref_head_cachep, head_ref);
+		head_ref = existing;
 	} else {
 		delayed_refs->num_heads++;
 		delayed_refs->num_heads_ready++;
-		delayed_refs->num_entries++;
+		atomic_inc(&delayed_refs->num_entries);
 		trans->delayed_ref_updates++;
 	}
-	return 0;
+	return head_ref;
 }
 
 /*
  * helper to insert a delayed tree ref into the rbtree.
  */
-static noinline int add_delayed_tree_ref(struct btrfs_trans_handle *trans,
-					 struct btrfs_delayed_ref_node *ref,
-					 u64 bytenr, u64 num_bytes, u64 parent,
-					 u64 ref_root, int level, int action)
+static noinline void
+add_delayed_tree_ref(struct btrfs_fs_info *fs_info,
+		     struct btrfs_trans_handle *trans,
+		     struct btrfs_delayed_ref_head *head_ref,
+		     struct btrfs_delayed_ref_node *ref, u64 bytenr,
+		     u64 num_bytes, u64 parent, u64 ref_root, int level,
+		     int action, int no_quota)
 {
 	struct btrfs_delayed_ref_node *existing;
 	struct btrfs_delayed_tree_ref *full_ref;
 	struct btrfs_delayed_ref_root *delayed_refs;
+	u64 seq = 0;
 
 	if (action == BTRFS_ADD_DELAYED_EXTENT)
 		action = BTRFS_ADD_DELAYED_REF;
 
+	if (is_fstree(ref_root))
+		seq = atomic64_read(&fs_info->tree_mod_seq);
 	delayed_refs = &trans->transaction->delayed_refs;
 
 	/* first set the basic ref node struct up */
@@ -627,51 +661,61 @@ static noinline int add_delayed_tree_ref(struct btrfs_trans_handle *trans,
 	ref->action = action;
 	ref->is_head = 0;
 	ref->in_tree = 1;
+	ref->no_quota = no_quota;
+	ref->seq = seq;
 
 	full_ref = btrfs_delayed_node_to_tree_ref(ref);
-	if (parent) {
-		full_ref->parent = parent;
+	full_ref->parent = parent;
+	full_ref->root = ref_root;
+	if (parent)
 		ref->type = BTRFS_SHARED_BLOCK_REF_KEY;
-	} else {
-		full_ref->root = ref_root;
+	else
 		ref->type = BTRFS_TREE_BLOCK_REF_KEY;
-	}
 	full_ref->level = level;
 
-	existing = tree_insert(&delayed_refs->root, &ref->rb_node);
+	trace_add_delayed_tree_ref(ref, full_ref, action);
 
+	spin_lock(&head_ref->lock);
+	existing = tree_insert(&head_ref->ref_root, &ref->rb_node);
 	if (existing) {
-		update_existing_ref(trans, delayed_refs, existing, ref);
+		update_existing_ref(trans, delayed_refs, head_ref, existing,
+				    ref);
 		/*
 		 * we've updated the existing ref, free the newly
 		 * allocated ref
 		 */
-		kfree(ref);
+		kmem_cache_free(btrfs_delayed_tree_ref_cachep, full_ref);
 	} else {
-		delayed_refs->num_entries++;
+		atomic_inc(&delayed_refs->num_entries);
 		trans->delayed_ref_updates++;
 	}
-	return 0;
+	spin_unlock(&head_ref->lock);
 }
 
 /*
  * helper to insert a delayed data ref into the rbtree.
  */
-static noinline int add_delayed_data_ref(struct btrfs_trans_handle *trans,
-					 struct btrfs_delayed_ref_node *ref,
-					 u64 bytenr, u64 num_bytes, u64 parent,
-					 u64 ref_root, u64 owner, u64 offset,
-					 int action)
+static noinline void
+add_delayed_data_ref(struct btrfs_fs_info *fs_info,
+		     struct btrfs_trans_handle *trans,
+		     struct btrfs_delayed_ref_head *head_ref,
+		     struct btrfs_delayed_ref_node *ref, u64 bytenr,
+		     u64 num_bytes, u64 parent, u64 ref_root, u64 owner,
+		     u64 offset, int action, int no_quota)
 {
 	struct btrfs_delayed_ref_node *existing;
 	struct btrfs_delayed_data_ref *full_ref;
 	struct btrfs_delayed_ref_root *delayed_refs;
+	u64 seq = 0;
 
 	if (action == BTRFS_ADD_DELAYED_EXTENT)
 		action = BTRFS_ADD_DELAYED_REF;
 
 	delayed_refs = &trans->transaction->delayed_refs;
 
+	if (is_fstree(ref_root))
+		seq = atomic64_read(&fs_info->tree_mod_seq);
+
 	/* first set the basic ref node struct up */
 	atomic_set(&ref->refs, 1);
 	ref->bytenr = bytenr;
@@ -680,32 +724,37 @@ static noinline int add_delayed_data_ref(struct btrfs_trans_handle *trans,
 	ref->action = action;
 	ref->is_head = 0;
 	ref->in_tree = 1;
+	ref->no_quota = no_quota;
+	ref->seq = seq;
 
 	full_ref = btrfs_delayed_node_to_data_ref(ref);
-	if (parent) {
-		full_ref->parent = parent;
+	full_ref->parent = parent;
+	full_ref->root = ref_root;
+	if (parent)
 		ref->type = BTRFS_SHARED_DATA_REF_KEY;
-	} else {
-		full_ref->root = ref_root;
+	else
 		ref->type = BTRFS_EXTENT_DATA_REF_KEY;
-	}
+
 	full_ref->objectid = owner;
 	full_ref->offset = offset;
 
-	existing = tree_insert(&delayed_refs->root, &ref->rb_node);
+	trace_add_delayed_data_ref(ref, full_ref, action);
 
+	spin_lock(&head_ref->lock);
+	existing = tree_insert(&head_ref->ref_root, &ref->rb_node);
 	if (existing) {
-		update_existing_ref(trans, delayed_refs, existing, ref);
+		update_existing_ref(trans, delayed_refs, head_ref, existing,
+				    ref);
 		/*
 		 * we've updated the existing ref, free the newly
 		 * allocated ref
 		 */
-		kfree(ref);
+		kmem_cache_free(btrfs_delayed_data_ref_cachep, full_ref);
 	} else {
-		delayed_refs->num_entries++;
+		atomic_inc(&delayed_refs->num_entries);
 		trans->delayed_ref_updates++;
 	}
-	return 0;
+	spin_unlock(&head_ref->lock);
 }
 
 /*
@@ -713,24 +762,28 @@ static noinline int add_delayed_data_ref(struct btrfs_trans_handle *trans,
  * to make sure the delayed ref is eventually processed before this
  * transaction commits.
  */
-int btrfs_add_delayed_tree_ref(struct btrfs_trans_handle *trans,
+int btrfs_add_delayed_tree_ref(struct btrfs_fs_info *fs_info,
+			       struct btrfs_trans_handle *trans,
 			       u64 bytenr, u64 num_bytes, u64 parent,
 			       u64 ref_root,  int level, int action,
-			       struct btrfs_delayed_extent_op *extent_op)
+			       struct btrfs_delayed_extent_op *extent_op,
+			       int no_quota)
 {
 	struct btrfs_delayed_tree_ref *ref;
 	struct btrfs_delayed_ref_head *head_ref;
 	struct btrfs_delayed_ref_root *delayed_refs;
-	int ret;
+
+	if (!is_fstree(ref_root) || !fs_info->quota_enabled)
+		no_quota = 0;
 
 	BUG_ON(extent_op && extent_op->is_data);
-	ref = kmalloc(sizeof(*ref), GFP_NOFS);
+	ref = kmem_cache_alloc(btrfs_delayed_tree_ref_cachep, GFP_NOFS);
 	if (!ref)
 		return -ENOMEM;
 
-	head_ref = kmalloc(sizeof(*head_ref), GFP_NOFS);
+	head_ref = kmem_cache_alloc(btrfs_delayed_ref_head_cachep, GFP_NOFS);
 	if (!head_ref) {
-		kfree(ref);
+		kmem_cache_free(btrfs_delayed_tree_ref_cachep, ref);
 		return -ENOMEM;
 	}
 
@@ -743,39 +796,43 @@ int btrfs_add_delayed_tree_ref(struct btrfs_trans_handle *trans,
 	 * insert both the head node and the new ref without dropping
 	 * the spin lock
 	 */
-	ret = add_delayed_ref_head(trans, &head_ref->node, bytenr, num_bytes,
-				   action, 0);
-	BUG_ON(ret);
+	head_ref = add_delayed_ref_head(fs_info, trans, &head_ref->node,
+					bytenr, num_bytes, action, 0);
 
-	ret = add_delayed_tree_ref(trans, &ref->node, bytenr, num_bytes,
-				   parent, ref_root, level, action);
-	BUG_ON(ret);
+	add_delayed_tree_ref(fs_info, trans, head_ref, &ref->node, bytenr,
+				   num_bytes, parent, ref_root, level, action,
+				   no_quota);
 	spin_unlock(&delayed_refs->lock);
+
 	return 0;
 }
 
 /*
  * add a delayed data ref. it's similar to btrfs_add_delayed_tree_ref.
  */
-int btrfs_add_delayed_data_ref(struct btrfs_trans_handle *trans,
+int btrfs_add_delayed_data_ref(struct btrfs_fs_info *fs_info,
+			       struct btrfs_trans_handle *trans,
 			       u64 bytenr, u64 num_bytes,
 			       u64 parent, u64 ref_root,
 			       u64 owner, u64 offset, int action,
-			       struct btrfs_delayed_extent_op *extent_op)
+			       struct btrfs_delayed_extent_op *extent_op,
+			       int no_quota)
 {
 	struct btrfs_delayed_data_ref *ref;
 	struct btrfs_delayed_ref_head *head_ref;
 	struct btrfs_delayed_ref_root *delayed_refs;
-	int ret;
+
+	if (!is_fstree(ref_root) || !fs_info->quota_enabled)
+		no_quota = 0;
 
 	BUG_ON(extent_op && !extent_op->is_data);
-	ref = kmalloc(sizeof(*ref), GFP_NOFS);
+	ref = kmem_cache_alloc(btrfs_delayed_data_ref_cachep, GFP_NOFS);
 	if (!ref)
 		return -ENOMEM;
 
-	head_ref = kmalloc(sizeof(*head_ref), GFP_NOFS);
+	head_ref = kmem_cache_alloc(btrfs_delayed_ref_head_cachep, GFP_NOFS);
 	if (!head_ref) {
-		kfree(ref);
+		kmem_cache_free(btrfs_delayed_data_ref_cachep, ref);
 		return -ENOMEM;
 	}
 
@@ -788,26 +845,26 @@ int btrfs_add_delayed_data_ref(struct btrfs_trans_handle *trans,
 	 * insert both the head node and the new ref without dropping
 	 * the spin lock
 	 */
-	ret = add_delayed_ref_head(trans, &head_ref->node, bytenr, num_bytes,
-				   action, 1);
-	BUG_ON(ret);
+	head_ref = add_delayed_ref_head(fs_info, trans, &head_ref->node,
+					bytenr, num_bytes, action, 1);
 
-	ret = add_delayed_data_ref(trans, &ref->node, bytenr, num_bytes,
-				   parent, ref_root, owner, offset, action);
-	BUG_ON(ret);
+	add_delayed_data_ref(fs_info, trans, head_ref, &ref->node, bytenr,
+				   num_bytes, parent, ref_root, owner, offset,
+				   action, no_quota);
 	spin_unlock(&delayed_refs->lock);
+
 	return 0;
 }
 
-int btrfs_add_delayed_extent_op(struct btrfs_trans_handle *trans,
+int btrfs_add_delayed_extent_op(struct btrfs_fs_info *fs_info,
+				struct btrfs_trans_handle *trans,
 				u64 bytenr, u64 num_bytes,
 				struct btrfs_delayed_extent_op *extent_op)
 {
 	struct btrfs_delayed_ref_head *head_ref;
 	struct btrfs_delayed_ref_root *delayed_refs;
-	int ret;
 
-	head_ref = kmalloc(sizeof(*head_ref), GFP_NOFS);
+	head_ref = kmem_cache_alloc(btrfs_delayed_ref_head_cachep, GFP_NOFS);
 	if (!head_ref)
 		return -ENOMEM;
 
@@ -816,10 +873,9 @@ int btrfs_add_delayed_extent_op(struct btrfs_trans_handle *trans,
 	delayed_refs = &trans->transaction->delayed_refs;
 	spin_lock(&delayed_refs->lock);
 
-	ret = add_delayed_ref_head(trans, &head_ref->node, bytenr,
+	add_delayed_ref_head(fs_info, trans, &head_ref->node, bytenr,
 				   num_bytes, BTRFS_UPDATE_DELAYED_HEAD,
 				   extent_op->is_data);
-	BUG_ON(ret);
 
 	spin_unlock(&delayed_refs->lock);
 	return 0;
@@ -833,88 +889,56 @@ int btrfs_add_delayed_extent_op(struct btrfs_trans_handle *trans,
 struct btrfs_delayed_ref_head *
 btrfs_find_delayed_ref_head(struct btrfs_trans_handle *trans, u64 bytenr)
 {
-	struct btrfs_delayed_ref_node *ref;
 	struct btrfs_delayed_ref_root *delayed_refs;
 
 	delayed_refs = &trans->transaction->delayed_refs;
-	ref = find_ref_head(&delayed_refs->root, bytenr, NULL);
-	if (ref)
-		return btrfs_delayed_node_to_head(ref);
-	return NULL;
+	return find_ref_head(&delayed_refs->href_root, bytenr, 0);
 }
 
-/*
- * add a delayed ref to the tree.  This does all of the accounting required
- * to make sure the delayed ref is eventually processed before this
- * transaction commits.
- *
- * The main point of this call is to add and remove a backreference in a single
- * shot, taking the lock only once, and only searching for the head node once.
- *
- * It is the same as doing a ref add and delete in two separate calls.
- */
-#if 0
-int btrfs_update_delayed_ref(struct btrfs_trans_handle *trans,
-			  u64 bytenr, u64 num_bytes, u64 orig_parent,
-			  u64 parent, u64 orig_ref_root, u64 ref_root,
-			  u64 orig_ref_generation, u64 ref_generation,
-			  u64 owner_objectid, int pin)
+void btrfs_delayed_ref_exit(void)
 {
-	struct btrfs_delayed_ref *ref;
-	struct btrfs_delayed_ref *old_ref;
-	struct btrfs_delayed_ref_head *head_ref;
-	struct btrfs_delayed_ref_root *delayed_refs;
-	int ret;
-
-	ref = kmalloc(sizeof(*ref), GFP_NOFS);
-	if (!ref)
-		return -ENOMEM;
-
-	old_ref = kmalloc(sizeof(*old_ref), GFP_NOFS);
-	if (!old_ref) {
-		kfree(ref);
-		return -ENOMEM;
-	}
-
-	/*
-	 * the parent = 0 case comes from cases where we don't actually
-	 * know the parent yet.  It will get updated later via a add/drop
-	 * pair.
-	 */
-	if (parent == 0)
-		parent = bytenr;
-	if (orig_parent == 0)
-		orig_parent = bytenr;
+	if (btrfs_delayed_ref_head_cachep)
+		kmem_cache_destroy(btrfs_delayed_ref_head_cachep);
+	if (btrfs_delayed_tree_ref_cachep)
+		kmem_cache_destroy(btrfs_delayed_tree_ref_cachep);
+	if (btrfs_delayed_data_ref_cachep)
+		kmem_cache_destroy(btrfs_delayed_data_ref_cachep);
+	if (btrfs_delayed_extent_op_cachep)
+		kmem_cache_destroy(btrfs_delayed_extent_op_cachep);
+}
 
-	head_ref = kmalloc(sizeof(*head_ref), GFP_NOFS);
-	if (!head_ref) {
-		kfree(ref);
-		kfree(old_ref);
-		return -ENOMEM;
-	}
-	delayed_refs = &trans->transaction->delayed_refs;
-	spin_lock(&delayed_refs->lock);
+int btrfs_delayed_ref_init(void)
+{
+	btrfs_delayed_ref_head_cachep = kmem_cache_create(
+				"btrfs_delayed_ref_head",
+				sizeof(struct btrfs_delayed_ref_head), 0,
+				SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
+	if (!btrfs_delayed_ref_head_cachep)
+		goto fail;
+
+	btrfs_delayed_tree_ref_cachep = kmem_cache_create(
+				"btrfs_delayed_tree_ref",
+				sizeof(struct btrfs_delayed_tree_ref), 0,
+				SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
+	if (!btrfs_delayed_tree_ref_cachep)
+		goto fail;
+
+	btrfs_delayed_data_ref_cachep = kmem_cache_create(
+				"btrfs_delayed_data_ref",
+				sizeof(struct btrfs_delayed_data_ref), 0,
+				SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
+	if (!btrfs_delayed_data_ref_cachep)
+		goto fail;
+
+	btrfs_delayed_extent_op_cachep = kmem_cache_create(
+				"btrfs_delayed_extent_op",
+				sizeof(struct btrfs_delayed_extent_op), 0,
+				SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
+	if (!btrfs_delayed_extent_op_cachep)
+		goto fail;
 
-	/*
-	 * insert both the head node and the new ref without dropping
-	 * the spin lock
-	 */
-	ret = __btrfs_add_delayed_ref(trans, &head_ref->node, bytenr, num_bytes,
-				      (u64)-1, 0, 0, 0,
-				      BTRFS_UPDATE_DELAYED_HEAD, 0);
-	BUG_ON(ret);
-
-	ret = __btrfs_add_delayed_ref(trans, &ref->node, bytenr, num_bytes,
-				      parent, ref_root, ref_generation,
-				      owner_objectid, BTRFS_ADD_DELAYED_REF, 0);
-	BUG_ON(ret);
-
-	ret = __btrfs_add_delayed_ref(trans, &old_ref->node, bytenr, num_bytes,
-				      orig_parent, orig_ref_root,
-				      orig_ref_generation, owner_objectid,
-				      BTRFS_DROP_DELAYED_REF, pin);
-	BUG_ON(ret);
-	spin_unlock(&delayed_refs->lock);
 	return 0;
+fail:
+	btrfs_delayed_ref_exit();
+	return -ENOMEM;
 }
-#endif
diff --git a/fs/btrfs/delayed-ref.h b/fs/btrfs/delayed-ref.h
index f6fc67ddad3..a764e2340d4 100644
--- a/fs/btrfs/delayed-ref.h
+++ b/fs/btrfs/delayed-ref.h
@@ -18,7 +18,7 @@
 #ifndef __DELAYED_REF__
 #define __DELAYED_REF__
 
-/* these are the possible values of struct btrfs_delayed_ref->action */
+/* these are the possible values of struct btrfs_delayed_ref_node->action */
 #define BTRFS_ADD_DELAYED_REF    1 /* add one backref to the tree */
 #define BTRFS_DROP_DELAYED_REF   2 /* delete one backref from the tree */
 #define BTRFS_ADD_DELAYED_EXTENT 3 /* record a full extent allocation */
@@ -33,6 +33,9 @@ struct btrfs_delayed_ref_node {
 	/* the size of the extent */
 	u64 num_bytes;
 
+	/* seq number to keep track of insertion order */
+	u64 seq;
+
 	/* ref count on this data structure */
 	atomic_t refs;
 
@@ -49,6 +52,7 @@ struct btrfs_delayed_ref_node {
 
 	unsigned int action:8;
 	unsigned int type:8;
+	unsigned int no_quota:1;
 	/* is this node still in the rbtree? */
 	unsigned int is_head:1;
 	unsigned int in_tree:1;
@@ -57,6 +61,7 @@ struct btrfs_delayed_ref_node {
 struct btrfs_delayed_extent_op {
 	struct btrfs_disk_key key;
 	u64 flags_to_set;
+	int level;
 	unsigned int update_key:1;
 	unsigned int update_flags:1;
 	unsigned int is_data:1;
@@ -77,7 +82,10 @@ struct btrfs_delayed_ref_head {
 	 */
 	struct mutex mutex;
 
-	struct list_head cluster;
+	spinlock_t lock;
+	struct rb_root ref_root;
+
+	struct rb_node href_node;
 
 	struct btrfs_delayed_extent_op *extent_op;
 	/*
@@ -94,29 +102,27 @@ struct btrfs_delayed_ref_head {
 	 */
 	unsigned int must_insert_reserved:1;
 	unsigned int is_data:1;
+	unsigned int processing:1;
 };
 
 struct btrfs_delayed_tree_ref {
 	struct btrfs_delayed_ref_node node;
-	union {
-		u64 root;
-		u64 parent;
-	};
+	u64 root;
+	u64 parent;
 	int level;
 };
 
 struct btrfs_delayed_data_ref {
 	struct btrfs_delayed_ref_node node;
-	union {
-		u64 root;
-		u64 parent;
-	};
+	u64 root;
+	u64 parent;
 	u64 objectid;
 	u64 offset;
 };
 
 struct btrfs_delayed_ref_root {
-	struct rb_root root;
+	/* head ref rbtree */
+	struct rb_root href_root;
 
 	/* this spin lock protects the rbtree and the entries inside */
 	spinlock_t lock;
@@ -124,7 +130,7 @@ struct btrfs_delayed_ref_root {
 	/* how many delayed ref updates we've queued, used by the
 	 * throttling code
 	 */
-	unsigned long num_entries;
+	atomic_t num_entries;
 
 	/* total number of head nodes in tree */
 	unsigned long num_heads;
@@ -142,43 +148,89 @@ struct btrfs_delayed_ref_root {
 	u64 run_delayed_start;
 };
 
+extern struct kmem_cache *btrfs_delayed_ref_head_cachep;
+extern struct kmem_cache *btrfs_delayed_tree_ref_cachep;
+extern struct kmem_cache *btrfs_delayed_data_ref_cachep;
+extern struct kmem_cache *btrfs_delayed_extent_op_cachep;
+
+int btrfs_delayed_ref_init(void);
+void btrfs_delayed_ref_exit(void);
+
+static inline struct btrfs_delayed_extent_op *
+btrfs_alloc_delayed_extent_op(void)
+{
+	return kmem_cache_alloc(btrfs_delayed_extent_op_cachep, GFP_NOFS);
+}
+
+static inline void
+btrfs_free_delayed_extent_op(struct btrfs_delayed_extent_op *op)
+{
+	if (op)
+		kmem_cache_free(btrfs_delayed_extent_op_cachep, op);
+}
+
 static inline void btrfs_put_delayed_ref(struct btrfs_delayed_ref_node *ref)
 {
 	WARN_ON(atomic_read(&ref->refs) == 0);
 	if (atomic_dec_and_test(&ref->refs)) {
 		WARN_ON(ref->in_tree);
-		kfree(ref);
+		switch (ref->type) {
+		case BTRFS_TREE_BLOCK_REF_KEY:
+		case BTRFS_SHARED_BLOCK_REF_KEY:
+			kmem_cache_free(btrfs_delayed_tree_ref_cachep, ref);
+			break;
+		case BTRFS_EXTENT_DATA_REF_KEY:
+		case BTRFS_SHARED_DATA_REF_KEY:
+			kmem_cache_free(btrfs_delayed_data_ref_cachep, ref);
+			break;
+		case 0:
+			kmem_cache_free(btrfs_delayed_ref_head_cachep, ref);
+			break;
+		default:
+			BUG();
+		}
 	}
 }
 
-int btrfs_add_delayed_tree_ref(struct btrfs_trans_handle *trans,
+int btrfs_add_delayed_tree_ref(struct btrfs_fs_info *fs_info,
+			       struct btrfs_trans_handle *trans,
 			       u64 bytenr, u64 num_bytes, u64 parent,
 			       u64 ref_root, int level, int action,
-			       struct btrfs_delayed_extent_op *extent_op);
-int btrfs_add_delayed_data_ref(struct btrfs_trans_handle *trans,
+			       struct btrfs_delayed_extent_op *extent_op,
+			       int no_quota);
+int btrfs_add_delayed_data_ref(struct btrfs_fs_info *fs_info,
+			       struct btrfs_trans_handle *trans,
 			       u64 bytenr, u64 num_bytes,
 			       u64 parent, u64 ref_root,
 			       u64 owner, u64 offset, int action,
-			       struct btrfs_delayed_extent_op *extent_op);
-int btrfs_add_delayed_extent_op(struct btrfs_trans_handle *trans,
+			       struct btrfs_delayed_extent_op *extent_op,
+			       int no_quota);
+int btrfs_add_delayed_extent_op(struct btrfs_fs_info *fs_info,
+				struct btrfs_trans_handle *trans,
 				u64 bytenr, u64 num_bytes,
 				struct btrfs_delayed_extent_op *extent_op);
+void btrfs_merge_delayed_refs(struct btrfs_trans_handle *trans,
+			      struct btrfs_fs_info *fs_info,
+			      struct btrfs_delayed_ref_root *delayed_refs,
+			      struct btrfs_delayed_ref_head *head);
 
 struct btrfs_delayed_ref_head *
 btrfs_find_delayed_ref_head(struct btrfs_trans_handle *trans, u64 bytenr);
-int btrfs_delayed_ref_pending(struct btrfs_trans_handle *trans, u64 bytenr);
-int btrfs_lookup_extent_info(struct btrfs_trans_handle *trans,
-			     struct btrfs_root *root, u64 bytenr,
-			     u64 num_bytes, u64 *refs, u64 *flags);
-int btrfs_update_delayed_ref(struct btrfs_trans_handle *trans,
-			  u64 bytenr, u64 num_bytes, u64 orig_parent,
-			  u64 parent, u64 orig_ref_root, u64 ref_root,
-			  u64 orig_ref_generation, u64 ref_generation,
-			  u64 owner_objectid, int pin);
 int btrfs_delayed_ref_lock(struct btrfs_trans_handle *trans,
 			   struct btrfs_delayed_ref_head *head);
-int btrfs_find_ref_cluster(struct btrfs_trans_handle *trans,
-			   struct list_head *cluster, u64 search_start);
+static inline void btrfs_delayed_ref_unlock(struct btrfs_delayed_ref_head *head)
+{
+	mutex_unlock(&head->mutex);
+}
+
+
+struct btrfs_delayed_ref_head *
+btrfs_select_ref_head(struct btrfs_trans_handle *trans);
+
+int btrfs_check_delayed_seq(struct btrfs_fs_info *fs_info,
+			    struct btrfs_delayed_ref_root *delayed_refs,
+			    u64 seq);
+
 /*
  * a node might live in a head or a regular ref, this lets you
  * test for the proper type to use.
diff --git a/fs/btrfs/dev-replace.c b/fs/btrfs/dev-replace.c
new file mode 100644
index 00000000000..eea26e1b2fd
--- /dev/null
+++ b/fs/btrfs/dev-replace.c
@@ -0,0 +1,932 @@
+/*
+ * Copyright (C) STRATO AG 2012.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+#include <linux/sched.h>
+#include <linux/bio.h>
+#include <linux/slab.h>
+#include <linux/buffer_head.h>
+#include <linux/blkdev.h>
+#include <linux/random.h>
+#include <linux/iocontext.h>
+#include <linux/capability.h>
+#include <linux/kthread.h>
+#include <linux/math64.h>
+#include <asm/div64.h>
+#include "ctree.h"
+#include "extent_map.h"
+#include "disk-io.h"
+#include "transaction.h"
+#include "print-tree.h"
+#include "volumes.h"
+#include "async-thread.h"
+#include "check-integrity.h"
+#include "rcu-string.h"
+#include "dev-replace.h"
+#include "sysfs.h"
+
+static int btrfs_dev_replace_finishing(struct btrfs_fs_info *fs_info,
+				       int scrub_ret);
+static void btrfs_dev_replace_update_device_in_mapping_tree(
+						struct btrfs_fs_info *fs_info,
+						struct btrfs_device *srcdev,
+						struct btrfs_device *tgtdev);
+static int btrfs_dev_replace_find_srcdev(struct btrfs_root *root, u64 srcdevid,
+					 char *srcdev_name,
+					 struct btrfs_device **device);
+static u64 __btrfs_dev_replace_cancel(struct btrfs_fs_info *fs_info);
+static int btrfs_dev_replace_kthread(void *data);
+static int btrfs_dev_replace_continue_on_mount(struct btrfs_fs_info *fs_info);
+
+
+int btrfs_init_dev_replace(struct btrfs_fs_info *fs_info)
+{
+	struct btrfs_key key;
+	struct btrfs_root *dev_root = fs_info->dev_root;
+	struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
+	struct extent_buffer *eb;
+	int slot;
+	int ret = 0;
+	struct btrfs_path *path = NULL;
+	int item_size;
+	struct btrfs_dev_replace_item *ptr;
+	u64 src_devid;
+
+	path = btrfs_alloc_path();
+	if (!path) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	key.objectid = 0;
+	key.type = BTRFS_DEV_REPLACE_KEY;
+	key.offset = 0;
+	ret = btrfs_search_slot(NULL, dev_root, &key, path, 0, 0);
+	if (ret) {
+no_valid_dev_replace_entry_found:
+		ret = 0;
+		dev_replace->replace_state =
+			BTRFS_DEV_REPLACE_ITEM_STATE_NEVER_STARTED;
+		dev_replace->cont_reading_from_srcdev_mode =
+		    BTRFS_DEV_REPLACE_ITEM_CONT_READING_FROM_SRCDEV_MODE_ALWAYS;
+		dev_replace->replace_state = 0;
+		dev_replace->time_started = 0;
+		dev_replace->time_stopped = 0;
+		atomic64_set(&dev_replace->num_write_errors, 0);
+		atomic64_set(&dev_replace->num_uncorrectable_read_errors, 0);
+		dev_replace->cursor_left = 0;
+		dev_replace->committed_cursor_left = 0;
+		dev_replace->cursor_left_last_write_of_item = 0;
+		dev_replace->cursor_right = 0;
+		dev_replace->srcdev = NULL;
+		dev_replace->tgtdev = NULL;
+		dev_replace->is_valid = 0;
+		dev_replace->item_needs_writeback = 0;
+		goto out;
+	}
+	slot = path->slots[0];
+	eb = path->nodes[0];
+	item_size = btrfs_item_size_nr(eb, slot);
+	ptr = btrfs_item_ptr(eb, slot, struct btrfs_dev_replace_item);
+
+	if (item_size != sizeof(struct btrfs_dev_replace_item)) {
+		btrfs_warn(fs_info,
+			"dev_replace entry found has unexpected size, ignore entry");
+		goto no_valid_dev_replace_entry_found;
+	}
+
+	src_devid = btrfs_dev_replace_src_devid(eb, ptr);
+	dev_replace->cont_reading_from_srcdev_mode =
+		btrfs_dev_replace_cont_reading_from_srcdev_mode(eb, ptr);
+	dev_replace->replace_state = btrfs_dev_replace_replace_state(eb, ptr);
+	dev_replace->time_started = btrfs_dev_replace_time_started(eb, ptr);
+	dev_replace->time_stopped =
+		btrfs_dev_replace_time_stopped(eb, ptr);
+	atomic64_set(&dev_replace->num_write_errors,
+		     btrfs_dev_replace_num_write_errors(eb, ptr));
+	atomic64_set(&dev_replace->num_uncorrectable_read_errors,
+		     btrfs_dev_replace_num_uncorrectable_read_errors(eb, ptr));
+	dev_replace->cursor_left = btrfs_dev_replace_cursor_left(eb, ptr);
+	dev_replace->committed_cursor_left = dev_replace->cursor_left;
+	dev_replace->cursor_left_last_write_of_item = dev_replace->cursor_left;
+	dev_replace->cursor_right = btrfs_dev_replace_cursor_right(eb, ptr);
+	dev_replace->is_valid = 1;
+
+	dev_replace->item_needs_writeback = 0;
+	switch (dev_replace->replace_state) {
+	case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED:
+	case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED:
+	case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED:
+		dev_replace->srcdev = NULL;
+		dev_replace->tgtdev = NULL;
+		break;
+	case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED:
+	case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED:
+		dev_replace->srcdev = btrfs_find_device(fs_info, src_devid,
+							NULL, NULL);
+		dev_replace->tgtdev = btrfs_find_device(fs_info,
+							BTRFS_DEV_REPLACE_DEVID,
+							NULL, NULL);
+		/*
+		 * allow 'btrfs dev replace_cancel' if src/tgt device is
+		 * missing
+		 */
+		if (!dev_replace->srcdev &&
+		    !btrfs_test_opt(dev_root, DEGRADED)) {
+			ret = -EIO;
+			btrfs_warn(fs_info,
+			   "cannot mount because device replace operation is ongoing and");
+			btrfs_warn(fs_info,
+			   "srcdev (devid %llu) is missing, need to run 'btrfs dev scan'?",
+			   src_devid);
+		}
+		if (!dev_replace->tgtdev &&
+		    !btrfs_test_opt(dev_root, DEGRADED)) {
+			ret = -EIO;
+			btrfs_warn(fs_info,
+			   "cannot mount because device replace operation is ongoing and");
+			btrfs_warn(fs_info,
+			   "tgtdev (devid %llu) is missing, need to run 'btrfs dev scan'?",
+				BTRFS_DEV_REPLACE_DEVID);
+		}
+		if (dev_replace->tgtdev) {
+			if (dev_replace->srcdev) {
+				dev_replace->tgtdev->total_bytes =
+					dev_replace->srcdev->total_bytes;
+				dev_replace->tgtdev->disk_total_bytes =
+					dev_replace->srcdev->disk_total_bytes;
+				dev_replace->tgtdev->bytes_used =
+					dev_replace->srcdev->bytes_used;
+			}
+			dev_replace->tgtdev->is_tgtdev_for_dev_replace = 1;
+			btrfs_init_dev_replace_tgtdev_for_resume(fs_info,
+				dev_replace->tgtdev);
+		}
+		break;
+	}
+
+out:
+	if (path)
+		btrfs_free_path(path);
+	return ret;
+}
+
+/*
+ * called from commit_transaction. Writes changed device replace state to
+ * disk.
+ */
+int btrfs_run_dev_replace(struct btrfs_trans_handle *trans,
+			  struct btrfs_fs_info *fs_info)
+{
+	int ret;
+	struct btrfs_root *dev_root = fs_info->dev_root;
+	struct btrfs_path *path;
+	struct btrfs_key key;
+	struct extent_buffer *eb;
+	struct btrfs_dev_replace_item *ptr;
+	struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
+
+	btrfs_dev_replace_lock(dev_replace);
+	if (!dev_replace->is_valid ||
+	    !dev_replace->item_needs_writeback) {
+		btrfs_dev_replace_unlock(dev_replace);
+		return 0;
+	}
+	btrfs_dev_replace_unlock(dev_replace);
+
+	key.objectid = 0;
+	key.type = BTRFS_DEV_REPLACE_KEY;
+	key.offset = 0;
+
+	path = btrfs_alloc_path();
+	if (!path) {
+		ret = -ENOMEM;
+		goto out;
+	}
+	ret = btrfs_search_slot(trans, dev_root, &key, path, -1, 1);
+	if (ret < 0) {
+		btrfs_warn(fs_info, "error %d while searching for dev_replace item!",
+			ret);
+		goto out;
+	}
+
+	if (ret == 0 &&
+	    btrfs_item_size_nr(path->nodes[0], path->slots[0]) < sizeof(*ptr)) {
+		/*
+		 * need to delete old one and insert a new one.
+		 * Since no attempt is made to recover any old state, if the
+		 * dev_replace state is 'running', the data on the target
+		 * drive is lost.
+		 * It would be possible to recover the state: just make sure
+		 * that the beginning of the item is never changed and always
+		 * contains all the essential information. Then read this
+		 * minimal set of information and use it as a base for the
+		 * new state.
+		 */
+		ret = btrfs_del_item(trans, dev_root, path);
+		if (ret != 0) {
+			btrfs_warn(fs_info, "delete too small dev_replace item failed %d!",
+				ret);
+			goto out;
+		}
+		ret = 1;
+	}
+
+	if (ret == 1) {
+		/* need to insert a new item */
+		btrfs_release_path(path);
+		ret = btrfs_insert_empty_item(trans, dev_root, path,
+					      &key, sizeof(*ptr));
+		if (ret < 0) {
+			btrfs_warn(fs_info, "insert dev_replace item failed %d!",
+				ret);
+			goto out;
+		}
+	}
+
+	eb = path->nodes[0];
+	ptr = btrfs_item_ptr(eb, path->slots[0],
+			     struct btrfs_dev_replace_item);
+
+	btrfs_dev_replace_lock(dev_replace);
+	if (dev_replace->srcdev)
+		btrfs_set_dev_replace_src_devid(eb, ptr,
+			dev_replace->srcdev->devid);
+	else
+		btrfs_set_dev_replace_src_devid(eb, ptr, (u64)-1);
+	btrfs_set_dev_replace_cont_reading_from_srcdev_mode(eb, ptr,
+		dev_replace->cont_reading_from_srcdev_mode);
+	btrfs_set_dev_replace_replace_state(eb, ptr,
+		dev_replace->replace_state);
+	btrfs_set_dev_replace_time_started(eb, ptr, dev_replace->time_started);
+	btrfs_set_dev_replace_time_stopped(eb, ptr, dev_replace->time_stopped);
+	btrfs_set_dev_replace_num_write_errors(eb, ptr,
+		atomic64_read(&dev_replace->num_write_errors));
+	btrfs_set_dev_replace_num_uncorrectable_read_errors(eb, ptr,
+		atomic64_read(&dev_replace->num_uncorrectable_read_errors));
+	dev_replace->cursor_left_last_write_of_item =
+		dev_replace->cursor_left;
+	btrfs_set_dev_replace_cursor_left(eb, ptr,
+		dev_replace->cursor_left_last_write_of_item);
+	btrfs_set_dev_replace_cursor_right(eb, ptr,
+		dev_replace->cursor_right);
+	dev_replace->item_needs_writeback = 0;
+	btrfs_dev_replace_unlock(dev_replace);
+
+	btrfs_mark_buffer_dirty(eb);
+
+out:
+	btrfs_free_path(path);
+
+	return ret;
+}
+
+void btrfs_after_dev_replace_commit(struct btrfs_fs_info *fs_info)
+{
+	struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
+
+	dev_replace->committed_cursor_left =
+		dev_replace->cursor_left_last_write_of_item;
+}
+
+int btrfs_dev_replace_start(struct btrfs_root *root,
+			    struct btrfs_ioctl_dev_replace_args *args)
+{
+	struct btrfs_trans_handle *trans;
+	struct btrfs_fs_info *fs_info = root->fs_info;
+	struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
+	int ret;
+	struct btrfs_device *tgt_device = NULL;
+	struct btrfs_device *src_device = NULL;
+
+	if (btrfs_fs_incompat(fs_info, RAID56)) {
+		btrfs_warn(fs_info, "dev_replace cannot yet handle RAID5/RAID6");
+		return -EOPNOTSUPP;
+	}
+
+	switch (args->start.cont_reading_from_srcdev_mode) {
+	case BTRFS_IOCTL_DEV_REPLACE_CONT_READING_FROM_SRCDEV_MODE_ALWAYS:
+	case BTRFS_IOCTL_DEV_REPLACE_CONT_READING_FROM_SRCDEV_MODE_AVOID:
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	if ((args->start.srcdevid == 0 && args->start.srcdev_name[0] == '\0') ||
+	    args->start.tgtdev_name[0] == '\0')
+		return -EINVAL;
+
+	mutex_lock(&fs_info->volume_mutex);
+	ret = btrfs_init_dev_replace_tgtdev(root, args->start.tgtdev_name,
+					    &tgt_device);
+	if (ret) {
+		btrfs_err(fs_info, "target device %s is invalid!",
+		       args->start.tgtdev_name);
+		mutex_unlock(&fs_info->volume_mutex);
+		return -EINVAL;
+	}
+
+	ret = btrfs_dev_replace_find_srcdev(root, args->start.srcdevid,
+					    args->start.srcdev_name,
+					    &src_device);
+	mutex_unlock(&fs_info->volume_mutex);
+	if (ret) {
+		ret = -EINVAL;
+		goto leave_no_lock;
+	}
+
+	if (tgt_device->total_bytes < src_device->total_bytes) {
+		btrfs_err(fs_info, "target device is smaller than source device!");
+		ret = -EINVAL;
+		goto leave_no_lock;
+	}
+
+	btrfs_dev_replace_lock(dev_replace);
+	switch (dev_replace->replace_state) {
+	case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED:
+	case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED:
+	case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED:
+		break;
+	case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED:
+	case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED:
+		args->result = BTRFS_IOCTL_DEV_REPLACE_RESULT_ALREADY_STARTED;
+		goto leave;
+	}
+
+	dev_replace->cont_reading_from_srcdev_mode =
+		args->start.cont_reading_from_srcdev_mode;
+	WARN_ON(!src_device);
+	dev_replace->srcdev = src_device;
+	WARN_ON(!tgt_device);
+	dev_replace->tgtdev = tgt_device;
+
+	printk_in_rcu(KERN_INFO
+		      "BTRFS: dev_replace from %s (devid %llu) to %s started\n",
+		      src_device->missing ? "<missing disk>" :
+		        rcu_str_deref(src_device->name),
+		      src_device->devid,
+		      rcu_str_deref(tgt_device->name));
+
+	tgt_device->total_bytes = src_device->total_bytes;
+	tgt_device->disk_total_bytes = src_device->disk_total_bytes;
+	tgt_device->bytes_used = src_device->bytes_used;
+
+	/*
+	 * from now on, the writes to the srcdev are all duplicated to
+	 * go to the tgtdev as well (refer to btrfs_map_block()).
+	 */
+	dev_replace->replace_state = BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED;
+	dev_replace->time_started = get_seconds();
+	dev_replace->cursor_left = 0;
+	dev_replace->committed_cursor_left = 0;
+	dev_replace->cursor_left_last_write_of_item = 0;
+	dev_replace->cursor_right = 0;
+	dev_replace->is_valid = 1;
+	dev_replace->item_needs_writeback = 1;
+	args->result = BTRFS_IOCTL_DEV_REPLACE_RESULT_NO_ERROR;
+	btrfs_dev_replace_unlock(dev_replace);
+
+	btrfs_wait_ordered_roots(root->fs_info, -1);
+
+	/* force writing the updated state information to disk */
+	trans = btrfs_start_transaction(root, 0);
+	if (IS_ERR(trans)) {
+		ret = PTR_ERR(trans);
+		btrfs_dev_replace_lock(dev_replace);
+		goto leave;
+	}
+
+	ret = btrfs_commit_transaction(trans, root);
+	WARN_ON(ret);
+
+	/* the disk copy procedure reuses the scrub code */
+	ret = btrfs_scrub_dev(fs_info, src_device->devid, 0,
+			      src_device->total_bytes,
+			      &dev_replace->scrub_progress, 0, 1);
+
+	ret = btrfs_dev_replace_finishing(root->fs_info, ret);
+	WARN_ON(ret);
+
+	return 0;
+
+leave:
+	dev_replace->srcdev = NULL;
+	dev_replace->tgtdev = NULL;
+	btrfs_dev_replace_unlock(dev_replace);
+leave_no_lock:
+	if (tgt_device)
+		btrfs_destroy_dev_replace_tgtdev(fs_info, tgt_device);
+	return ret;
+}
+
+/*
+ * blocked until all flighting bios are finished.
+ */
+static void btrfs_rm_dev_replace_blocked(struct btrfs_fs_info *fs_info)
+{
+	s64 writers;
+	DEFINE_WAIT(wait);
+
+	set_bit(BTRFS_FS_STATE_DEV_REPLACING, &fs_info->fs_state);
+	do {
+		prepare_to_wait(&fs_info->replace_wait, &wait,
+				TASK_UNINTERRUPTIBLE);
+		writers = percpu_counter_sum(&fs_info->bio_counter);
+		if (writers)
+			schedule();
+		finish_wait(&fs_info->replace_wait, &wait);
+	} while (writers);
+}
+
+/*
+ * we have removed target device, it is safe to allow new bios request.
+ */
+static void btrfs_rm_dev_replace_unblocked(struct btrfs_fs_info *fs_info)
+{
+	clear_bit(BTRFS_FS_STATE_DEV_REPLACING, &fs_info->fs_state);
+	if (waitqueue_active(&fs_info->replace_wait))
+		wake_up(&fs_info->replace_wait);
+}
+
+static int btrfs_dev_replace_finishing(struct btrfs_fs_info *fs_info,
+				       int scrub_ret)
+{
+	struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
+	struct btrfs_device *tgt_device;
+	struct btrfs_device *src_device;
+	struct btrfs_root *root = fs_info->tree_root;
+	u8 uuid_tmp[BTRFS_UUID_SIZE];
+	struct btrfs_trans_handle *trans;
+	int ret = 0;
+
+	/* don't allow cancel or unmount to disturb the finishing procedure */
+	mutex_lock(&dev_replace->lock_finishing_cancel_unmount);
+
+	btrfs_dev_replace_lock(dev_replace);
+	/* was the operation canceled, or is it finished? */
+	if (dev_replace->replace_state !=
+	    BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED) {
+		btrfs_dev_replace_unlock(dev_replace);
+		mutex_unlock(&dev_replace->lock_finishing_cancel_unmount);
+		return 0;
+	}
+
+	tgt_device = dev_replace->tgtdev;
+	src_device = dev_replace->srcdev;
+	btrfs_dev_replace_unlock(dev_replace);
+
+	/*
+	 * flush all outstanding I/O and inode extent mappings before the
+	 * copy operation is declared as being finished
+	 */
+	ret = btrfs_start_delalloc_roots(root->fs_info, 0, -1);
+	if (ret) {
+		mutex_unlock(&dev_replace->lock_finishing_cancel_unmount);
+		return ret;
+	}
+	btrfs_wait_ordered_roots(root->fs_info, -1);
+
+	trans = btrfs_start_transaction(root, 0);
+	if (IS_ERR(trans)) {
+		mutex_unlock(&dev_replace->lock_finishing_cancel_unmount);
+		return PTR_ERR(trans);
+	}
+	ret = btrfs_commit_transaction(trans, root);
+	WARN_ON(ret);
+
+	/* keep away write_all_supers() during the finishing procedure */
+	mutex_lock(&root->fs_info->chunk_mutex);
+	mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
+	btrfs_dev_replace_lock(dev_replace);
+	dev_replace->replace_state =
+		scrub_ret ? BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED
+			  : BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED;
+	dev_replace->tgtdev = NULL;
+	dev_replace->srcdev = NULL;
+	dev_replace->time_stopped = get_seconds();
+	dev_replace->item_needs_writeback = 1;
+
+	/* replace old device with new one in mapping tree */
+	if (!scrub_ret) {
+		btrfs_dev_replace_update_device_in_mapping_tree(fs_info,
+								src_device,
+								tgt_device);
+	} else {
+		printk_in_rcu(KERN_ERR
+			      "BTRFS: btrfs_scrub_dev(%s, %llu, %s) failed %d\n",
+			      src_device->missing ? "<missing disk>" :
+			        rcu_str_deref(src_device->name),
+			      src_device->devid,
+			      rcu_str_deref(tgt_device->name), scrub_ret);
+		btrfs_dev_replace_unlock(dev_replace);
+		mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
+		mutex_unlock(&root->fs_info->chunk_mutex);
+		if (tgt_device)
+			btrfs_destroy_dev_replace_tgtdev(fs_info, tgt_device);
+		mutex_unlock(&dev_replace->lock_finishing_cancel_unmount);
+
+		return 0;
+	}
+
+	printk_in_rcu(KERN_INFO
+		      "BTRFS: dev_replace from %s (devid %llu) to %s) finished\n",
+		      src_device->missing ? "<missing disk>" :
+		        rcu_str_deref(src_device->name),
+		      src_device->devid,
+		      rcu_str_deref(tgt_device->name));
+	tgt_device->is_tgtdev_for_dev_replace = 0;
+	tgt_device->devid = src_device->devid;
+	src_device->devid = BTRFS_DEV_REPLACE_DEVID;
+	tgt_device->bytes_used = src_device->bytes_used;
+	memcpy(uuid_tmp, tgt_device->uuid, sizeof(uuid_tmp));
+	memcpy(tgt_device->uuid, src_device->uuid, sizeof(tgt_device->uuid));
+	memcpy(src_device->uuid, uuid_tmp, sizeof(src_device->uuid));
+	tgt_device->total_bytes = src_device->total_bytes;
+	tgt_device->disk_total_bytes = src_device->disk_total_bytes;
+	tgt_device->bytes_used = src_device->bytes_used;
+	if (fs_info->sb->s_bdev == src_device->bdev)
+		fs_info->sb->s_bdev = tgt_device->bdev;
+	if (fs_info->fs_devices->latest_bdev == src_device->bdev)
+		fs_info->fs_devices->latest_bdev = tgt_device->bdev;
+	list_add(&tgt_device->dev_alloc_list, &fs_info->fs_devices->alloc_list);
+
+	/* replace the sysfs entry */
+	btrfs_kobj_rm_device(fs_info, src_device);
+	btrfs_kobj_add_device(fs_info, tgt_device);
+
+	btrfs_rm_dev_replace_blocked(fs_info);
+
+	btrfs_rm_dev_replace_srcdev(fs_info, src_device);
+
+	btrfs_rm_dev_replace_unblocked(fs_info);
+
+	/*
+	 * this is again a consistent state where no dev_replace procedure
+	 * is running, the target device is part of the filesystem, the
+	 * source device is not part of the filesystem anymore and its 1st
+	 * superblock is scratched out so that it is no longer marked to
+	 * belong to this filesystem.
+	 */
+	btrfs_dev_replace_unlock(dev_replace);
+	mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
+	mutex_unlock(&root->fs_info->chunk_mutex);
+
+	/* write back the superblocks */
+	trans = btrfs_start_transaction(root, 0);
+	if (!IS_ERR(trans))
+		btrfs_commit_transaction(trans, root);
+
+	mutex_unlock(&dev_replace->lock_finishing_cancel_unmount);
+
+	return 0;
+}
+
+static void btrfs_dev_replace_update_device_in_mapping_tree(
+						struct btrfs_fs_info *fs_info,
+						struct btrfs_device *srcdev,
+						struct btrfs_device *tgtdev)
+{
+	struct extent_map_tree *em_tree = &fs_info->mapping_tree.map_tree;
+	struct extent_map *em;
+	struct map_lookup *map;
+	u64 start = 0;
+	int i;
+
+	write_lock(&em_tree->lock);
+	do {
+		em = lookup_extent_mapping(em_tree, start, (u64)-1);
+		if (!em)
+			break;
+		map = (struct map_lookup *)em->bdev;
+		for (i = 0; i < map->num_stripes; i++)
+			if (srcdev == map->stripes[i].dev)
+				map->stripes[i].dev = tgtdev;
+		start = em->start + em->len;
+		free_extent_map(em);
+	} while (start);
+	write_unlock(&em_tree->lock);
+}
+
+static int btrfs_dev_replace_find_srcdev(struct btrfs_root *root, u64 srcdevid,
+					 char *srcdev_name,
+					 struct btrfs_device **device)
+{
+	int ret;
+
+	if (srcdevid) {
+		ret = 0;
+		*device = btrfs_find_device(root->fs_info, srcdevid, NULL,
+					    NULL);
+		if (!*device)
+			ret = -ENOENT;
+	} else {
+		ret = btrfs_find_device_missing_or_by_path(root, srcdev_name,
+							   device);
+	}
+	return ret;
+}
+
+void btrfs_dev_replace_status(struct btrfs_fs_info *fs_info,
+			      struct btrfs_ioctl_dev_replace_args *args)
+{
+	struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
+
+	btrfs_dev_replace_lock(dev_replace);
+	/* even if !dev_replace_is_valid, the values are good enough for
+	 * the replace_status ioctl */
+	args->result = BTRFS_IOCTL_DEV_REPLACE_RESULT_NO_ERROR;
+	args->status.replace_state = dev_replace->replace_state;
+	args->status.time_started = dev_replace->time_started;
+	args->status.time_stopped = dev_replace->time_stopped;
+	args->status.num_write_errors =
+		atomic64_read(&dev_replace->num_write_errors);
+	args->status.num_uncorrectable_read_errors =
+		atomic64_read(&dev_replace->num_uncorrectable_read_errors);
+	switch (dev_replace->replace_state) {
+	case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED:
+	case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED:
+		args->status.progress_1000 = 0;
+		break;
+	case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED:
+		args->status.progress_1000 = 1000;
+		break;
+	case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED:
+	case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED:
+		args->status.progress_1000 = div64_u64(dev_replace->cursor_left,
+			div64_u64(dev_replace->srcdev->total_bytes, 1000));
+		break;
+	}
+	btrfs_dev_replace_unlock(dev_replace);
+}
+
+int btrfs_dev_replace_cancel(struct btrfs_fs_info *fs_info,
+			     struct btrfs_ioctl_dev_replace_args *args)
+{
+	args->result = __btrfs_dev_replace_cancel(fs_info);
+	return 0;
+}
+
+static u64 __btrfs_dev_replace_cancel(struct btrfs_fs_info *fs_info)
+{
+	struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
+	struct btrfs_device *tgt_device = NULL;
+	struct btrfs_trans_handle *trans;
+	struct btrfs_root *root = fs_info->tree_root;
+	u64 result;
+	int ret;
+
+	if (fs_info->sb->s_flags & MS_RDONLY)
+		return -EROFS;
+
+	mutex_lock(&dev_replace->lock_finishing_cancel_unmount);
+	btrfs_dev_replace_lock(dev_replace);
+	switch (dev_replace->replace_state) {
+	case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED:
+	case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED:
+	case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED:
+		result = BTRFS_IOCTL_DEV_REPLACE_RESULT_NOT_STARTED;
+		btrfs_dev_replace_unlock(dev_replace);
+		goto leave;
+	case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED:
+	case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED:
+		result = BTRFS_IOCTL_DEV_REPLACE_RESULT_NO_ERROR;
+		tgt_device = dev_replace->tgtdev;
+		dev_replace->tgtdev = NULL;
+		dev_replace->srcdev = NULL;
+		break;
+	}
+	dev_replace->replace_state = BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED;
+	dev_replace->time_stopped = get_seconds();
+	dev_replace->item_needs_writeback = 1;
+	btrfs_dev_replace_unlock(dev_replace);
+	btrfs_scrub_cancel(fs_info);
+
+	trans = btrfs_start_transaction(root, 0);
+	if (IS_ERR(trans)) {
+		mutex_unlock(&dev_replace->lock_finishing_cancel_unmount);
+		return PTR_ERR(trans);
+	}
+	ret = btrfs_commit_transaction(trans, root);
+	WARN_ON(ret);
+	if (tgt_device)
+		btrfs_destroy_dev_replace_tgtdev(fs_info, tgt_device);
+
+leave:
+	mutex_unlock(&dev_replace->lock_finishing_cancel_unmount);
+	return result;
+}
+
+void btrfs_dev_replace_suspend_for_unmount(struct btrfs_fs_info *fs_info)
+{
+	struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
+
+	mutex_lock(&dev_replace->lock_finishing_cancel_unmount);
+	btrfs_dev_replace_lock(dev_replace);
+	switch (dev_replace->replace_state) {
+	case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED:
+	case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED:
+	case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED:
+	case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED:
+		break;
+	case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED:
+		dev_replace->replace_state =
+			BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED;
+		dev_replace->time_stopped = get_seconds();
+		dev_replace->item_needs_writeback = 1;
+		btrfs_info(fs_info, "suspending dev_replace for unmount");
+		break;
+	}
+
+	btrfs_dev_replace_unlock(dev_replace);
+	mutex_unlock(&dev_replace->lock_finishing_cancel_unmount);
+}
+
+/* resume dev_replace procedure that was interrupted by unmount */
+int btrfs_resume_dev_replace_async(struct btrfs_fs_info *fs_info)
+{
+	struct task_struct *task;
+	struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
+
+	btrfs_dev_replace_lock(dev_replace);
+	switch (dev_replace->replace_state) {
+	case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED:
+	case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED:
+	case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED:
+		btrfs_dev_replace_unlock(dev_replace);
+		return 0;
+	case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED:
+		break;
+	case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED:
+		dev_replace->replace_state =
+			BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED;
+		break;
+	}
+	if (!dev_replace->tgtdev || !dev_replace->tgtdev->bdev) {
+		btrfs_info(fs_info, "cannot continue dev_replace, tgtdev is missing");
+		btrfs_info(fs_info,
+			"you may cancel the operation after 'mount -o degraded'");
+		btrfs_dev_replace_unlock(dev_replace);
+		return 0;
+	}
+	btrfs_dev_replace_unlock(dev_replace);
+
+	WARN_ON(atomic_xchg(
+		&fs_info->mutually_exclusive_operation_running, 1));
+	task = kthread_run(btrfs_dev_replace_kthread, fs_info, "btrfs-devrepl");
+	return PTR_ERR_OR_ZERO(task);
+}
+
+static int btrfs_dev_replace_kthread(void *data)
+{
+	struct btrfs_fs_info *fs_info = data;
+	struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
+	struct btrfs_ioctl_dev_replace_args *status_args;
+	u64 progress;
+
+	status_args = kzalloc(sizeof(*status_args), GFP_NOFS);
+	if (status_args) {
+		btrfs_dev_replace_status(fs_info, status_args);
+		progress = status_args->status.progress_1000;
+		kfree(status_args);
+		do_div(progress, 10);
+		printk_in_rcu(KERN_INFO
+			"BTRFS: continuing dev_replace from %s (devid %llu) to %s @%u%%\n",
+			dev_replace->srcdev->missing ? "<missing disk>" :
+			rcu_str_deref(dev_replace->srcdev->name),
+			dev_replace->srcdev->devid,
+			dev_replace->tgtdev ?
+			rcu_str_deref(dev_replace->tgtdev->name) :
+			"<missing target disk>",
+			(unsigned int)progress);
+	}
+	btrfs_dev_replace_continue_on_mount(fs_info);
+	atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
+
+	return 0;
+}
+
+static int btrfs_dev_replace_continue_on_mount(struct btrfs_fs_info *fs_info)
+{
+	struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
+	int ret;
+
+	ret = btrfs_scrub_dev(fs_info, dev_replace->srcdev->devid,
+			      dev_replace->committed_cursor_left,
+			      dev_replace->srcdev->total_bytes,
+			      &dev_replace->scrub_progress, 0, 1);
+	ret = btrfs_dev_replace_finishing(fs_info, ret);
+	WARN_ON(ret);
+	return 0;
+}
+
+int btrfs_dev_replace_is_ongoing(struct btrfs_dev_replace *dev_replace)
+{
+	if (!dev_replace->is_valid)
+		return 0;
+
+	switch (dev_replace->replace_state) {
+	case BTRFS_IOCTL_DEV_REPLACE_STATE_NEVER_STARTED:
+	case BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED:
+	case BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED:
+		return 0;
+	case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED:
+	case BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED:
+		/*
+		 * return true even if tgtdev is missing (this is
+		 * something that can happen if the dev_replace
+		 * procedure is suspended by an umount and then
+		 * the tgtdev is missing (or "btrfs dev scan") was
+		 * not called and the the filesystem is remounted
+		 * in degraded state. This does not stop the
+		 * dev_replace procedure. It needs to be canceled
+		 * manually if the cancelation is wanted.
+		 */
+		break;
+	}
+	return 1;
+}
+
+void btrfs_dev_replace_lock(struct btrfs_dev_replace *dev_replace)
+{
+	/* the beginning is just an optimization for the typical case */
+	if (atomic_read(&dev_replace->nesting_level) == 0) {
+acquire_lock:
+		/* this is not a nested case where the same thread
+		 * is trying to acqurire the same lock twice */
+		mutex_lock(&dev_replace->lock);
+		mutex_lock(&dev_replace->lock_management_lock);
+		dev_replace->lock_owner = current->pid;
+		atomic_inc(&dev_replace->nesting_level);
+		mutex_unlock(&dev_replace->lock_management_lock);
+		return;
+	}
+
+	mutex_lock(&dev_replace->lock_management_lock);
+	if (atomic_read(&dev_replace->nesting_level) > 0 &&
+	    dev_replace->lock_owner == current->pid) {
+		WARN_ON(!mutex_is_locked(&dev_replace->lock));
+		atomic_inc(&dev_replace->nesting_level);
+		mutex_unlock(&dev_replace->lock_management_lock);
+		return;
+	}
+
+	mutex_unlock(&dev_replace->lock_management_lock);
+	goto acquire_lock;
+}
+
+void btrfs_dev_replace_unlock(struct btrfs_dev_replace *dev_replace)
+{
+	WARN_ON(!mutex_is_locked(&dev_replace->lock));
+	mutex_lock(&dev_replace->lock_management_lock);
+	WARN_ON(atomic_read(&dev_replace->nesting_level) < 1);
+	WARN_ON(dev_replace->lock_owner != current->pid);
+	atomic_dec(&dev_replace->nesting_level);
+	if (atomic_read(&dev_replace->nesting_level) == 0) {
+		dev_replace->lock_owner = 0;
+		mutex_unlock(&dev_replace->lock_management_lock);
+		mutex_unlock(&dev_replace->lock);
+	} else {
+		mutex_unlock(&dev_replace->lock_management_lock);
+	}
+}
+
+void btrfs_bio_counter_inc_noblocked(struct btrfs_fs_info *fs_info)
+{
+	percpu_counter_inc(&fs_info->bio_counter);
+}
+
+void btrfs_bio_counter_dec(struct btrfs_fs_info *fs_info)
+{
+	percpu_counter_dec(&fs_info->bio_counter);
+
+	if (waitqueue_active(&fs_info->replace_wait))
+		wake_up(&fs_info->replace_wait);
+}
+
+void btrfs_bio_counter_inc_blocked(struct btrfs_fs_info *fs_info)
+{
+	DEFINE_WAIT(wait);
+again:
+	percpu_counter_inc(&fs_info->bio_counter);
+	if (test_bit(BTRFS_FS_STATE_DEV_REPLACING, &fs_info->fs_state)) {
+		btrfs_bio_counter_dec(fs_info);
+		wait_event(fs_info->replace_wait,
+			   !test_bit(BTRFS_FS_STATE_DEV_REPLACING,
+				     &fs_info->fs_state));
+		goto again;
+	}
+
+}
diff --git a/fs/btrfs/dev-replace.h b/fs/btrfs/dev-replace.h
new file mode 100644
index 00000000000..20035cbbf02
--- /dev/null
+++ b/fs/btrfs/dev-replace.h
@@ -0,0 +1,44 @@
+/*
+ * Copyright (C) STRATO AG 2012.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#if !defined(__BTRFS_DEV_REPLACE__)
+#define __BTRFS_DEV_REPLACE__
+
+struct btrfs_ioctl_dev_replace_args;
+
+int btrfs_init_dev_replace(struct btrfs_fs_info *fs_info);
+int btrfs_run_dev_replace(struct btrfs_trans_handle *trans,
+			  struct btrfs_fs_info *fs_info);
+void btrfs_after_dev_replace_commit(struct btrfs_fs_info *fs_info);
+int btrfs_dev_replace_start(struct btrfs_root *root,
+			    struct btrfs_ioctl_dev_replace_args *args);
+void btrfs_dev_replace_status(struct btrfs_fs_info *fs_info,
+			      struct btrfs_ioctl_dev_replace_args *args);
+int btrfs_dev_replace_cancel(struct btrfs_fs_info *fs_info,
+			     struct btrfs_ioctl_dev_replace_args *args);
+void btrfs_dev_replace_suspend_for_unmount(struct btrfs_fs_info *fs_info);
+int btrfs_resume_dev_replace_async(struct btrfs_fs_info *fs_info);
+int btrfs_dev_replace_is_ongoing(struct btrfs_dev_replace *dev_replace);
+void btrfs_dev_replace_lock(struct btrfs_dev_replace *dev_replace);
+void btrfs_dev_replace_unlock(struct btrfs_dev_replace *dev_replace);
+
+static inline void btrfs_dev_replace_stats_inc(atomic64_t *stat_value)
+{
+	atomic64_inc(stat_value);
+}
+#endif
diff --git a/fs/btrfs/dir-item.c b/fs/btrfs/dir-item.c
index e9103b3baa4..a0691df5dce 100644
--- a/fs/btrfs/dir-item.c
+++ b/fs/btrfs/dir-item.c
@@ -21,6 +21,10 @@
 #include "hash.h"
 #include "transaction.h"
 
+static struct btrfs_dir_item *btrfs_match_dir_item_name(struct btrfs_root *root,
+			      struct btrfs_path *path,
+			      const char *name, int name_len);
+
 /*
  * insert a name into a directory, doing overflow properly if there is a hash
  * collision.  data_size indicates how big the item inserted should be.  On
@@ -49,14 +53,12 @@ static struct btrfs_dir_item *insert_with_overflow(struct btrfs_trans_handle
 		di = btrfs_match_dir_item_name(root, path, name, name_len);
 		if (di)
 			return ERR_PTR(-EEXIST);
-		ret = btrfs_extend_item(trans, root, path, data_size);
-		WARN_ON(ret > 0);
-	}
-	if (ret < 0)
+		btrfs_extend_item(root, path, data_size);
+	} else if (ret < 0)
 		return ERR_PTR(ret);
 	WARN_ON(ret > 0);
 	leaf = path->nodes[0];
-	item = btrfs_item_nr(leaf, path->slots[0]);
+	item = btrfs_item_nr(path->slots[0]);
 	ptr = btrfs_item_ptr(leaf, path->slots[0], char);
 	BUG_ON(data_size > btrfs_item_size(leaf, item));
 	ptr += btrfs_item_size(leaf, item) - data_size;
@@ -90,13 +92,8 @@ int btrfs_insert_xattr_item(struct btrfs_trans_handle *trans,
 	data_size = sizeof(*dir_item) + name_len + data_len;
 	dir_item = insert_with_overflow(trans, root, path, &key, data_size,
 					name, name_len);
-	/*
-	 * FIXME: at some point we should handle xattr's that are larger than
-	 * what we can fit in our leaf.  We set location to NULL b/c we arent
-	 * pointing at anything else, that will change if we store the xattr
-	 * data in a separate inode.
-	 */
-	BUG_ON(IS_ERR(dir_item));
+	if (IS_ERR(dir_item))
+		return PTR_ERR(dir_item);
 	memset(&location, 0, sizeof(location));
 
 	leaf = path->nodes[0];
@@ -122,10 +119,12 @@ int btrfs_insert_xattr_item(struct btrfs_trans_handle *trans,
  * 'location' is the key to stuff into the directory item, 'type' is the
  * type of the inode we're pointing to, and 'index' is the sequence number
  * to use for the second index (if one is created).
+ * Will return 0 or -ENOMEM
  */
 int btrfs_insert_dir_item(struct btrfs_trans_handle *trans, struct btrfs_root
-			  *root, const char *name, int name_len, u64 dir,
-			  struct btrfs_key *location, u8 type, u64 index)
+			  *root, const char *name, int name_len,
+			  struct inode *dir, struct btrfs_key *location,
+			  u8 type, u64 index)
 {
 	int ret = 0;
 	int ret2 = 0;
@@ -137,13 +136,17 @@ int btrfs_insert_dir_item(struct btrfs_trans_handle *trans, struct btrfs_root
 	struct btrfs_disk_key disk_key;
 	u32 data_size;
 
-	key.objectid = dir;
+	key.objectid = btrfs_ino(dir);
 	btrfs_set_key_type(&key, BTRFS_DIR_ITEM_KEY);
 	key.offset = btrfs_name_hash(name, name_len);
 
 	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
 	path->leave_spinning = 1;
 
+	btrfs_cpu_key_to_disk(&disk_key, location);
+
 	data_size = sizeof(*dir_item) + name_len;
 	dir_item = insert_with_overflow(trans, root, path, &key, data_size,
 					name, name_len);
@@ -151,11 +154,10 @@ int btrfs_insert_dir_item(struct btrfs_trans_handle *trans, struct btrfs_root
 		ret = PTR_ERR(dir_item);
 		if (ret == -EEXIST)
 			goto second_insert;
-		goto out;
+		goto out_free;
 	}
 
 	leaf = path->nodes[0];
-	btrfs_cpu_key_to_disk(&disk_key, location);
 	btrfs_set_dir_item_key(leaf, dir_item, &disk_key);
 	btrfs_set_dir_type(leaf, dir_item, type);
 	btrfs_set_dir_data_len(leaf, dir_item, 0);
@@ -170,29 +172,13 @@ second_insert:
 	/* FIXME, use some real flag for selecting the extra index */
 	if (root == root->fs_info->tree_root) {
 		ret = 0;
-		goto out;
+		goto out_free;
 	}
-	btrfs_release_path(root, path);
+	btrfs_release_path(path);
 
-	btrfs_set_key_type(&key, BTRFS_DIR_INDEX_KEY);
-	key.offset = index;
-	dir_item = insert_with_overflow(trans, root, path, &key, data_size,
-					name, name_len);
-	if (IS_ERR(dir_item)) {
-		ret2 = PTR_ERR(dir_item);
-		goto out;
-	}
-	leaf = path->nodes[0];
-	btrfs_cpu_key_to_disk(&disk_key, location);
-	btrfs_set_dir_item_key(leaf, dir_item, &disk_key);
-	btrfs_set_dir_type(leaf, dir_item, type);
-	btrfs_set_dir_data_len(leaf, dir_item, 0);
-	btrfs_set_dir_name_len(leaf, dir_item, name_len);
-	btrfs_set_dir_transid(leaf, dir_item, trans->transid);
-	name_ptr = (unsigned long)(dir_item + 1);
-	write_extent_buffer(leaf, name, name_ptr, name_len);
-	btrfs_mark_buffer_dirty(leaf);
-out:
+	ret2 = btrfs_insert_delayed_dir_index(trans, root, name, name_len, dir,
+					      &disk_key, type, index);
+out_free:
 	btrfs_free_path(path);
 	if (ret)
 		return ret;
@@ -216,8 +202,6 @@ struct btrfs_dir_item *btrfs_lookup_dir_item(struct btrfs_trans_handle *trans,
 	struct btrfs_key key;
 	int ins_len = mod < 0 ? -1 : 0;
 	int cow = mod != 0;
-	struct btrfs_key found_key;
-	struct extent_buffer *leaf;
 
 	key.objectid = dir;
 	btrfs_set_key_type(&key, BTRFS_DIR_ITEM_KEY);
@@ -227,21 +211,69 @@ struct btrfs_dir_item *btrfs_lookup_dir_item(struct btrfs_trans_handle *trans,
 	ret = btrfs_search_slot(trans, root, &key, path, ins_len, cow);
 	if (ret < 0)
 		return ERR_PTR(ret);
+	if (ret > 0)
+		return NULL;
+
+	return btrfs_match_dir_item_name(root, path, name, name_len);
+}
+
+int btrfs_check_dir_item_collision(struct btrfs_root *root, u64 dir,
+				   const char *name, int name_len)
+{
+	int ret;
+	struct btrfs_key key;
+	struct btrfs_dir_item *di;
+	int data_size;
+	struct extent_buffer *leaf;
+	int slot;
+	struct btrfs_path *path;
+
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	key.objectid = dir;
+	btrfs_set_key_type(&key, BTRFS_DIR_ITEM_KEY);
+	key.offset = btrfs_name_hash(name, name_len);
+
+	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+
+	/* return back any errors */
+	if (ret < 0)
+		goto out;
+
+	/* nothing found, we're safe */
 	if (ret > 0) {
-		if (path->slots[0] == 0)
-			return NULL;
-		path->slots[0]--;
+		ret = 0;
+		goto out;
 	}
 
-	leaf = path->nodes[0];
-	btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
-
-	if (found_key.objectid != dir ||
-	    btrfs_key_type(&found_key) != BTRFS_DIR_ITEM_KEY ||
-	    found_key.offset != key.offset)
-		return NULL;
+	/* we found an item, look for our name in the item */
+	di = btrfs_match_dir_item_name(root, path, name, name_len);
+	if (di) {
+		/* our exact name was found */
+		ret = -EEXIST;
+		goto out;
+	}
 
-	return btrfs_match_dir_item_name(root, path, name, name_len);
+	/*
+	 * see if there is room in the item to insert this
+	 * name
+	 */
+	data_size = sizeof(*di) + name_len;
+	leaf = path->nodes[0];
+	slot = path->slots[0];
+	if (data_size + btrfs_item_size_nr(leaf, slot) +
+	    sizeof(struct btrfs_item) > BTRFS_LEAF_DATA_SIZE(root)) {
+		ret = -EOVERFLOW;
+	} else {
+		/* plenty of insertion room */
+		ret = 0;
+	}
+out:
+	btrfs_free_path(path);
+	return ret;
 }
 
 /*
@@ -333,8 +365,6 @@ struct btrfs_dir_item *btrfs_lookup_xattr(struct btrfs_trans_handle *trans,
 	struct btrfs_key key;
 	int ins_len = mod < 0 ? -1 : 0;
 	int cow = mod != 0;
-	struct btrfs_key found_key;
-	struct extent_buffer *leaf;
 
 	key.objectid = dir;
 	btrfs_set_key_type(&key, BTRFS_XATTR_ITEM_KEY);
@@ -342,18 +372,7 @@ struct btrfs_dir_item *btrfs_lookup_xattr(struct btrfs_trans_handle *trans,
 	ret = btrfs_search_slot(trans, root, &key, path, ins_len, cow);
 	if (ret < 0)
 		return ERR_PTR(ret);
-	if (ret > 0) {
-		if (path->slots[0] == 0)
-			return NULL;
-		path->slots[0]--;
-	}
-
-	leaf = path->nodes[0];
-	btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
-
-	if (found_key.objectid != dir ||
-	    btrfs_key_type(&found_key) != BTRFS_XATTR_ITEM_KEY ||
-	    found_key.offset != key.offset)
+	if (ret > 0)
 		return NULL;
 
 	return btrfs_match_dir_item_name(root, path, name, name_len);
@@ -364,7 +383,7 @@ struct btrfs_dir_item *btrfs_lookup_xattr(struct btrfs_trans_handle *trans,
  * this walks through all the entries in a dir item and finds one
  * for a specific name.
  */
-struct btrfs_dir_item *btrfs_match_dir_item_name(struct btrfs_root *root,
+static struct btrfs_dir_item *btrfs_match_dir_item_name(struct btrfs_root *root,
 			      struct btrfs_path *path,
 			      const char *name, int name_len)
 {
@@ -377,6 +396,9 @@ struct btrfs_dir_item *btrfs_match_dir_item_name(struct btrfs_root *root,
 
 	leaf = path->nodes[0];
 	dir_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_dir_item);
+	if (verify_dir_item(root, leaf, dir_item))
+		return NULL;
+
 	total_len = btrfs_item_size_nr(leaf, path->slots[0]);
 	while (cur < total_len) {
 		this_len = sizeof(*dir_item) +
@@ -424,8 +446,41 @@ int btrfs_delete_one_dir_name(struct btrfs_trans_handle *trans,
 		start = btrfs_item_ptr_offset(leaf, path->slots[0]);
 		memmove_extent_buffer(leaf, ptr, ptr + sub_item_len,
 			item_len - (ptr + sub_item_len - start));
-		ret = btrfs_truncate_item(trans, root, path,
-					  item_len - sub_item_len, 1);
+		btrfs_truncate_item(root, path, item_len - sub_item_len, 1);
 	}
+	return ret;
+}
+
+int verify_dir_item(struct btrfs_root *root,
+		    struct extent_buffer *leaf,
+		    struct btrfs_dir_item *dir_item)
+{
+	u16 namelen = BTRFS_NAME_LEN;
+	u8 type = btrfs_dir_type(leaf, dir_item);
+
+	if (type >= BTRFS_FT_MAX) {
+		btrfs_crit(root->fs_info, "invalid dir item type: %d",
+		       (int)type);
+		return 1;
+	}
+
+	if (type == BTRFS_FT_XATTR)
+		namelen = XATTR_NAME_MAX;
+
+	if (btrfs_dir_name_len(leaf, dir_item) > namelen) {
+		btrfs_crit(root->fs_info, "invalid dir item name len: %u",
+		       (unsigned)btrfs_dir_data_len(leaf, dir_item));
+		return 1;
+	}
+
+	/* BTRFS_MAX_XATTR_SIZE is the same for all dir items */
+	if ((btrfs_dir_data_len(leaf, dir_item) +
+	     btrfs_dir_name_len(leaf, dir_item)) > BTRFS_MAX_XATTR_SIZE(root)) {
+		btrfs_crit(root->fs_info, "invalid dir item name + data len: %u + %u",
+		       (unsigned)btrfs_dir_name_len(leaf, dir_item),
+		       (unsigned)btrfs_dir_data_len(leaf, dir_item));
+		return 1;
+	}
+
 	return 0;
 }
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index e7b8f2c89cc..08e65e9cf2a 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -26,11 +26,15 @@
 #include <linux/workqueue.h>
 #include <linux/kthread.h>
 #include <linux/freezer.h>
-#include <linux/crc32c.h>
 #include <linux/slab.h>
-#include "compat.h"
+#include <linux/migrate.h>
+#include <linux/ratelimit.h>
+#include <linux/uuid.h>
+#include <linux/semaphore.h>
+#include <asm/unaligned.h>
 #include "ctree.h"
 #include "disk-io.h"
+#include "hash.h"
 #include "transaction.h"
 #include "btrfs_inode.h"
 #include "volumes.h"
@@ -39,12 +43,36 @@
 #include "locking.h"
 #include "tree-log.h"
 #include "free-space-cache.h"
+#include "inode-map.h"
+#include "check-integrity.h"
+#include "rcu-string.h"
+#include "dev-replace.h"
+#include "raid56.h"
+#include "sysfs.h"
+#include "qgroup.h"
+
+#ifdef CONFIG_X86
+#include <asm/cpufeature.h>
+#endif
 
 static struct extent_io_ops btree_extent_io_ops;
 static void end_workqueue_fn(struct btrfs_work *work);
 static void free_fs_root(struct btrfs_root *root);
-
-static atomic_t btrfs_bdi_num = ATOMIC_INIT(0);
+static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
+				    int read_only);
+static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t,
+					     struct btrfs_root *root);
+static void btrfs_destroy_ordered_extents(struct btrfs_root *root);
+static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
+				      struct btrfs_root *root);
+static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root);
+static int btrfs_destroy_marked_extents(struct btrfs_root *root,
+					struct extent_io_tree *dirty_pages,
+					int mark);
+static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
+				       struct extent_io_tree *pinned_extents);
+static int btrfs_cleanup_transaction(struct btrfs_root *root);
+static void btrfs_error_commit_super(struct btrfs_root *root);
 
 /*
  * end_io_wq structs are used to do processing in task context when an IO is
@@ -76,41 +104,93 @@ struct async_submit_bio {
 	int rw;
 	int mirror_num;
 	unsigned long bio_flags;
+	/*
+	 * bio_offset is optional, can be used if the pages in the bio
+	 * can't tell us where in the file the bio should go
+	 */
+	u64 bio_offset;
 	struct btrfs_work work;
+	int error;
 };
 
-/* These are used to set the lockdep class on the extent buffer locks.
- * The class is set by the readpage_end_io_hook after the buffer has
- * passed csum validation but before the pages are unlocked.
+/*
+ * Lockdep class keys for extent_buffer->lock's in this root.  For a given
+ * eb, the lockdep key is determined by the btrfs_root it belongs to and
+ * the level the eb occupies in the tree.
+ *
+ * Different roots are used for different purposes and may nest inside each
+ * other and they require separate keysets.  As lockdep keys should be
+ * static, assign keysets according to the purpose of the root as indicated
+ * by btrfs_root->objectid.  This ensures that all special purpose roots
+ * have separate keysets.
  *
- * The lockdep class is also set by btrfs_init_new_buffer on freshly
- * allocated blocks.
+ * Lock-nesting across peer nodes is always done with the immediate parent
+ * node locked thus preventing deadlock.  As lockdep doesn't know this, use
+ * subclass to avoid triggering lockdep warning in such cases.
  *
- * The class is based on the level in the tree block, which allows lockdep
- * to know that lower nodes nest inside the locks of higher nodes.
+ * The key is set by the readpage_end_io_hook after the buffer has passed
+ * csum validation but before the pages are unlocked.  It is also set by
+ * btrfs_init_new_buffer on freshly allocated blocks.
  *
- * We also add a check to make sure the highest level of the tree is
- * the same as our lockdep setup here.  If BTRFS_MAX_LEVEL changes, this
- * code needs update as well.
+ * We also add a check to make sure the highest level of the tree is the
+ * same as our lockdep setup here.  If BTRFS_MAX_LEVEL changes, this code
+ * needs update as well.
  */
 #ifdef CONFIG_DEBUG_LOCK_ALLOC
 # if BTRFS_MAX_LEVEL != 8
 #  error
 # endif
-static struct lock_class_key btrfs_eb_class[BTRFS_MAX_LEVEL + 1];
-static const char *btrfs_eb_name[BTRFS_MAX_LEVEL + 1] = {
-	/* leaf */
-	"btrfs-extent-00",
-	"btrfs-extent-01",
-	"btrfs-extent-02",
-	"btrfs-extent-03",
-	"btrfs-extent-04",
-	"btrfs-extent-05",
-	"btrfs-extent-06",
-	"btrfs-extent-07",
-	/* highest possible level */
-	"btrfs-extent-08",
+
+static struct btrfs_lockdep_keyset {
+	u64			id;		/* root objectid */
+	const char		*name_stem;	/* lock name stem */
+	char			names[BTRFS_MAX_LEVEL + 1][20];
+	struct lock_class_key	keys[BTRFS_MAX_LEVEL + 1];
+} btrfs_lockdep_keysets[] = {
+	{ .id = BTRFS_ROOT_TREE_OBJECTID,	.name_stem = "root"	},
+	{ .id = BTRFS_EXTENT_TREE_OBJECTID,	.name_stem = "extent"	},
+	{ .id = BTRFS_CHUNK_TREE_OBJECTID,	.name_stem = "chunk"	},
+	{ .id = BTRFS_DEV_TREE_OBJECTID,	.name_stem = "dev"	},
+	{ .id = BTRFS_FS_TREE_OBJECTID,		.name_stem = "fs"	},
+	{ .id = BTRFS_CSUM_TREE_OBJECTID,	.name_stem = "csum"	},
+	{ .id = BTRFS_QUOTA_TREE_OBJECTID,	.name_stem = "quota"	},
+	{ .id = BTRFS_TREE_LOG_OBJECTID,	.name_stem = "log"	},
+	{ .id = BTRFS_TREE_RELOC_OBJECTID,	.name_stem = "treloc"	},
+	{ .id = BTRFS_DATA_RELOC_TREE_OBJECTID,	.name_stem = "dreloc"	},
+	{ .id = BTRFS_UUID_TREE_OBJECTID,	.name_stem = "uuid"	},
+	{ .id = 0,				.name_stem = "tree"	},
 };
+
+void __init btrfs_init_lockdep(void)
+{
+	int i, j;
+
+	/* initialize lockdep class names */
+	for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) {
+		struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i];
+
+		for (j = 0; j < ARRAY_SIZE(ks->names); j++)
+			snprintf(ks->names[j], sizeof(ks->names[j]),
+				 "btrfs-%s-%02d", ks->name_stem, j);
+	}
+}
+
+void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb,
+				    int level)
+{
+	struct btrfs_lockdep_keyset *ks;
+
+	BUG_ON(level >= ARRAY_SIZE(ks->keys));
+
+	/* find the matching keyset, id 0 is the default entry */
+	for (ks = btrfs_lockdep_keysets; ks->id; ks++)
+		if (ks->id == objectid)
+			break;
+
+	lockdep_set_class_and_name(&eb->lock,
+				   &ks->keys[level], ks->names[level]);
+}
+
 #endif
 
 /*
@@ -118,7 +198,7 @@ static const char *btrfs_eb_name[BTRFS_MAX_LEVEL + 1] = {
  * that covers the entire device
  */
 static struct extent_map *btree_get_extent(struct inode *inode,
-		struct page *page, size_t page_offset, u64 start, u64 len,
+		struct page *page, size_t pg_offset, u64 start, u64 len,
 		int create)
 {
 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
@@ -135,7 +215,7 @@ static struct extent_map *btree_get_extent(struct inode *inode,
 	}
 	read_unlock(&em_tree->lock);
 
-	em = alloc_extent_map(GFP_NOFS);
+	em = alloc_extent_map();
 	if (!em) {
 		em = ERR_PTR(-ENOMEM);
 		goto out;
@@ -147,40 +227,30 @@ static struct extent_map *btree_get_extent(struct inode *inode,
 	em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
 
 	write_lock(&em_tree->lock);
-	ret = add_extent_mapping(em_tree, em);
+	ret = add_extent_mapping(em_tree, em, 0);
 	if (ret == -EEXIST) {
-		u64 failed_start = em->start;
-		u64 failed_len = em->len;
-
 		free_extent_map(em);
 		em = lookup_extent_mapping(em_tree, start, len);
-		if (em) {
-			ret = 0;
-		} else {
-			em = lookup_extent_mapping(em_tree, failed_start,
-						   failed_len);
-			ret = -EIO;
-		}
+		if (!em)
+			em = ERR_PTR(-EIO);
 	} else if (ret) {
 		free_extent_map(em);
-		em = NULL;
+		em = ERR_PTR(ret);
 	}
 	write_unlock(&em_tree->lock);
 
-	if (ret)
-		em = ERR_PTR(ret);
 out:
 	return em;
 }
 
-u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len)
+u32 btrfs_csum_data(char *data, u32 seed, size_t len)
 {
-	return crc32c(seed, data, len);
+	return btrfs_crc32c(seed, data, len);
 }
 
 void btrfs_csum_final(u32 crc, char *result)
 {
-	*(__le32 *)result = ~cpu_to_le32(crc);
+	put_unaligned_le32(~crc, result);
 }
 
 /*
@@ -190,13 +260,11 @@ void btrfs_csum_final(u32 crc, char *result)
 static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
 			   int verify)
 {
-	u16 csum_size =
-		btrfs_super_csum_size(&root->fs_info->super_copy);
+	u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
 	char *result = NULL;
 	unsigned long len;
 	unsigned long cur_len;
 	unsigned long offset = BTRFS_CSUM_SIZE;
-	char *map_token = NULL;
 	char *kaddr;
 	unsigned long map_start;
 	unsigned long map_len;
@@ -207,16 +275,14 @@ static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
 	len = buf->len - offset;
 	while (len > 0) {
 		err = map_private_extent_buffer(buf, offset, 32,
-					&map_token, &kaddr,
-					&map_start, &map_len, KM_USER0);
+					&kaddr, &map_start, &map_len);
 		if (err)
 			return 1;
 		cur_len = min(len, map_len - (offset - map_start));
-		crc = btrfs_csum_data(root, kaddr + offset - map_start,
+		crc = btrfs_csum_data(kaddr + offset - map_start,
 				      crc, cur_len);
 		len -= cur_len;
 		offset += cur_len;
-		unmap_extent_buffer(buf, map_token, KM_USER0);
 	}
 	if (csum_size > sizeof(inline_result)) {
 		result = kzalloc(csum_size * sizeof(char), GFP_NOFS);
@@ -235,14 +301,11 @@ static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
 			memcpy(&found, result, csum_size);
 
 			read_extent_buffer(buf, &val, 0, csum_size);
-			if (printk_ratelimit()) {
-				printk(KERN_INFO "btrfs: %s checksum verify "
-				       "failed on %llu wanted %X found %X "
-				       "level %d\n",
-				       root->fs_info->sb->s_id,
-				       (unsigned long long)buf->start, val, found,
-				       btrfs_header_level(buf));
-			}
+			printk_ratelimited(KERN_INFO
+				"BTRFS: %s checksum verify failed on %llu wanted %X found %X "
+				"level %d\n",
+				root->fs_info->sb->s_id, buf->start,
+				val, found, btrfs_header_level(buf));
 			if (result != (char *)&inline_result)
 				kfree(result);
 			return 1;
@@ -262,33 +325,96 @@ static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
  * in the wrong place.
  */
 static int verify_parent_transid(struct extent_io_tree *io_tree,
-				 struct extent_buffer *eb, u64 parent_transid)
+				 struct extent_buffer *eb, u64 parent_transid,
+				 int atomic)
 {
 	struct extent_state *cached_state = NULL;
 	int ret;
+	bool need_lock = (current->journal_info ==
+			  (void *)BTRFS_SEND_TRANS_STUB);
 
 	if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
 		return 0;
 
+	if (atomic)
+		return -EAGAIN;
+
+	if (need_lock) {
+		btrfs_tree_read_lock(eb);
+		btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
+	}
+
 	lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1,
-			 0, &cached_state, GFP_NOFS);
-	if (extent_buffer_uptodate(io_tree, eb, cached_state) &&
+			 0, &cached_state);
+	if (extent_buffer_uptodate(eb) &&
 	    btrfs_header_generation(eb) == parent_transid) {
 		ret = 0;
 		goto out;
 	}
-	if (printk_ratelimit()) {
-		printk("parent transid verify failed on %llu wanted %llu "
+	printk_ratelimited("parent transid verify failed on %llu wanted %llu "
 		       "found %llu\n",
-		       (unsigned long long)eb->start,
-		       (unsigned long long)parent_transid,
-		       (unsigned long long)btrfs_header_generation(eb));
-	}
+		       eb->start, parent_transid, btrfs_header_generation(eb));
 	ret = 1;
-	clear_extent_buffer_uptodate(io_tree, eb, &cached_state);
+
+	/*
+	 * Things reading via commit roots that don't have normal protection,
+	 * like send, can have a really old block in cache that may point at a
+	 * block that has been free'd and re-allocated.  So don't clear uptodate
+	 * if we find an eb that is under IO (dirty/writeback) because we could
+	 * end up reading in the stale data and then writing it back out and
+	 * making everybody very sad.
+	 */
+	if (!extent_buffer_under_io(eb))
+		clear_extent_buffer_uptodate(eb);
 out:
 	unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1,
 			     &cached_state, GFP_NOFS);
+	if (need_lock)
+		btrfs_tree_read_unlock_blocking(eb);
+	return ret;
+}
+
+/*
+ * Return 0 if the superblock checksum type matches the checksum value of that
+ * algorithm. Pass the raw disk superblock data.
+ */
+static int btrfs_check_super_csum(char *raw_disk_sb)
+{
+	struct btrfs_super_block *disk_sb =
+		(struct btrfs_super_block *)raw_disk_sb;
+	u16 csum_type = btrfs_super_csum_type(disk_sb);
+	int ret = 0;
+
+	if (csum_type == BTRFS_CSUM_TYPE_CRC32) {
+		u32 crc = ~(u32)0;
+		const int csum_size = sizeof(crc);
+		char result[csum_size];
+
+		/*
+		 * The super_block structure does not span the whole
+		 * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space
+		 * is filled with zeros and is included in the checkum.
+		 */
+		crc = btrfs_csum_data(raw_disk_sb + BTRFS_CSUM_SIZE,
+				crc, BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
+		btrfs_csum_final(crc, result);
+
+		if (memcmp(raw_disk_sb, result, csum_size))
+			ret = 1;
+
+		if (ret && btrfs_super_generation(disk_sb) < 10) {
+			printk(KERN_WARNING
+				"BTRFS: super block crcs don't match, older mkfs detected\n");
+			ret = 0;
+		}
+	}
+
+	if (csum_type >= ARRAY_SIZE(btrfs_csum_sizes)) {
+		printk(KERN_ERR "BTRFS: unsupported checksum algorithm %u\n",
+				csum_type);
+		ret = 1;
+	}
+
 	return ret;
 }
 
@@ -301,28 +427,56 @@ static int btree_read_extent_buffer_pages(struct btrfs_root *root,
 					  u64 start, u64 parent_transid)
 {
 	struct extent_io_tree *io_tree;
+	int failed = 0;
 	int ret;
 	int num_copies = 0;
 	int mirror_num = 0;
+	int failed_mirror = 0;
 
+	clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
 	io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
 	while (1) {
-		ret = read_extent_buffer_pages(io_tree, eb, start, 1,
+		ret = read_extent_buffer_pages(io_tree, eb, start,
+					       WAIT_COMPLETE,
 					       btree_get_extent, mirror_num);
-		if (!ret &&
-		    !verify_parent_transid(io_tree, eb, parent_transid))
-			return ret;
+		if (!ret) {
+			if (!verify_parent_transid(io_tree, eb,
+						   parent_transid, 0))
+				break;
+			else
+				ret = -EIO;
+		}
 
-		num_copies = btrfs_num_copies(&root->fs_info->mapping_tree,
+		/*
+		 * This buffer's crc is fine, but its contents are corrupted, so
+		 * there is no reason to read the other copies, they won't be
+		 * any less wrong.
+		 */
+		if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags))
+			break;
+
+		num_copies = btrfs_num_copies(root->fs_info,
 					      eb->start, eb->len);
 		if (num_copies == 1)
-			return ret;
+			break;
+
+		if (!failed_mirror) {
+			failed = 1;
+			failed_mirror = eb->read_mirror;
+		}
 
 		mirror_num++;
+		if (mirror_num == failed_mirror)
+			mirror_num++;
+
 		if (mirror_num > num_copies)
-			return ret;
+			break;
 	}
-	return -EIO;
+
+	if (failed && !ret && failed_mirror)
+		repair_eb_io_failure(root, eb, failed_mirror);
+
+	return ret;
 }
 
 /*
@@ -332,46 +486,17 @@ static int btree_read_extent_buffer_pages(struct btrfs_root *root,
 
 static int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
 {
-	struct extent_io_tree *tree;
-	u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
+	u64 start = page_offset(page);
 	u64 found_start;
-	int found_level;
-	unsigned long len;
 	struct extent_buffer *eb;
-	int ret;
-
-	tree = &BTRFS_I(page->mapping->host)->io_tree;
 
-	if (page->private == EXTENT_PAGE_PRIVATE)
-		goto out;
-	if (!page->private)
-		goto out;
-	len = page->private >> 2;
-	WARN_ON(len == 0);
-
-	eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS);
-	ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE,
-					     btrfs_header_generation(eb));
-	BUG_ON(ret);
+	eb = (struct extent_buffer *)page->private;
+	if (page != eb->pages[0])
+		return 0;
 	found_start = btrfs_header_bytenr(eb);
-	if (found_start != start) {
-		WARN_ON(1);
-		goto err;
-	}
-	if (eb->first_page != page) {
-		WARN_ON(1);
-		goto err;
-	}
-	if (!PageUptodate(page)) {
-		WARN_ON(1);
-		goto err;
-	}
-	found_level = btrfs_header_level(eb);
-
+	if (WARN_ON(found_start != start || !PageUptodate(page)))
+		return 0;
 	csum_tree_block(root, eb, 0);
-err:
-	free_extent_buffer(eb);
-out:
 	return 0;
 }
 
@@ -382,8 +507,7 @@ static int check_tree_block_fsid(struct btrfs_root *root,
 	u8 fsid[BTRFS_UUID_SIZE];
 	int ret = 1;
 
-	read_extent_buffer(eb, fsid, (unsigned long)btrfs_header_fsid(eb),
-			   BTRFS_FSID_SIZE);
+	read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE);
 	while (fs_devices) {
 		if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) {
 			ret = 0;
@@ -394,79 +518,179 @@ static int check_tree_block_fsid(struct btrfs_root *root,
 	return ret;
 }
 
-#ifdef CONFIG_DEBUG_LOCK_ALLOC
-void btrfs_set_buffer_lockdep_class(struct extent_buffer *eb, int level)
+#define CORRUPT(reason, eb, root, slot)				\
+	btrfs_crit(root->fs_info, "corrupt leaf, %s: block=%llu,"	\
+		   "root=%llu, slot=%d", reason,			\
+	       btrfs_header_bytenr(eb),	root->objectid, slot)
+
+static noinline int check_leaf(struct btrfs_root *root,
+			       struct extent_buffer *leaf)
 {
-	lockdep_set_class_and_name(&eb->lock,
-			   &btrfs_eb_class[level],
-			   btrfs_eb_name[level]);
+	struct btrfs_key key;
+	struct btrfs_key leaf_key;
+	u32 nritems = btrfs_header_nritems(leaf);
+	int slot;
+
+	if (nritems == 0)
+		return 0;
+
+	/* Check the 0 item */
+	if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) !=
+	    BTRFS_LEAF_DATA_SIZE(root)) {
+		CORRUPT("invalid item offset size pair", leaf, root, 0);
+		return -EIO;
+	}
+
+	/*
+	 * Check to make sure each items keys are in the correct order and their
+	 * offsets make sense.  We only have to loop through nritems-1 because
+	 * we check the current slot against the next slot, which verifies the
+	 * next slot's offset+size makes sense and that the current's slot
+	 * offset is correct.
+	 */
+	for (slot = 0; slot < nritems - 1; slot++) {
+		btrfs_item_key_to_cpu(leaf, &leaf_key, slot);
+		btrfs_item_key_to_cpu(leaf, &key, slot + 1);
+
+		/* Make sure the keys are in the right order */
+		if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) {
+			CORRUPT("bad key order", leaf, root, slot);
+			return -EIO;
+		}
+
+		/*
+		 * Make sure the offset and ends are right, remember that the
+		 * item data starts at the end of the leaf and grows towards the
+		 * front.
+		 */
+		if (btrfs_item_offset_nr(leaf, slot) !=
+			btrfs_item_end_nr(leaf, slot + 1)) {
+			CORRUPT("slot offset bad", leaf, root, slot);
+			return -EIO;
+		}
+
+		/*
+		 * Check to make sure that we don't point outside of the leaf,
+		 * just incase all the items are consistent to eachother, but
+		 * all point outside of the leaf.
+		 */
+		if (btrfs_item_end_nr(leaf, slot) >
+		    BTRFS_LEAF_DATA_SIZE(root)) {
+			CORRUPT("slot end outside of leaf", leaf, root, slot);
+			return -EIO;
+		}
+	}
+
+	return 0;
 }
-#endif
 
-static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end,
-			       struct extent_state *state)
+static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio,
+				      u64 phy_offset, struct page *page,
+				      u64 start, u64 end, int mirror)
 {
-	struct extent_io_tree *tree;
 	u64 found_start;
 	int found_level;
-	unsigned long len;
 	struct extent_buffer *eb;
 	struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
 	int ret = 0;
+	int reads_done;
 
-	tree = &BTRFS_I(page->mapping->host)->io_tree;
-	if (page->private == EXTENT_PAGE_PRIVATE)
-		goto out;
 	if (!page->private)
 		goto out;
 
-	len = page->private >> 2;
-	WARN_ON(len == 0);
+	eb = (struct extent_buffer *)page->private;
 
-	eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS);
+	/* the pending IO might have been the only thing that kept this buffer
+	 * in memory.  Make sure we have a ref for all this other checks
+	 */
+	extent_buffer_get(eb);
 
-	found_start = btrfs_header_bytenr(eb);
-	if (found_start != start) {
-		if (printk_ratelimit()) {
-			printk(KERN_INFO "btrfs bad tree block start "
-			       "%llu %llu\n",
-			       (unsigned long long)found_start,
-			       (unsigned long long)eb->start);
-		}
+	reads_done = atomic_dec_and_test(&eb->io_pages);
+	if (!reads_done)
+		goto err;
+
+	eb->read_mirror = mirror;
+	if (test_bit(EXTENT_BUFFER_IOERR, &eb->bflags)) {
 		ret = -EIO;
 		goto err;
 	}
-	if (eb->first_page != page) {
-		printk(KERN_INFO "btrfs bad first page %lu %lu\n",
-		       eb->first_page->index, page->index);
-		WARN_ON(1);
+
+	found_start = btrfs_header_bytenr(eb);
+	if (found_start != eb->start) {
+		printk_ratelimited(KERN_INFO "BTRFS: bad tree block start "
+			       "%llu %llu\n",
+			       found_start, eb->start);
 		ret = -EIO;
 		goto err;
 	}
 	if (check_tree_block_fsid(root, eb)) {
-		if (printk_ratelimit()) {
-			printk(KERN_INFO "btrfs bad fsid on block %llu\n",
-			       (unsigned long long)eb->start);
-		}
+		printk_ratelimited(KERN_INFO "BTRFS: bad fsid on block %llu\n",
+			       eb->start);
 		ret = -EIO;
 		goto err;
 	}
 	found_level = btrfs_header_level(eb);
+	if (found_level >= BTRFS_MAX_LEVEL) {
+		btrfs_info(root->fs_info, "bad tree block level %d",
+			   (int)btrfs_header_level(eb));
+		ret = -EIO;
+		goto err;
+	}
 
-	btrfs_set_buffer_lockdep_class(eb, found_level);
+	btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb),
+				       eb, found_level);
 
 	ret = csum_tree_block(root, eb, 1);
-	if (ret)
+	if (ret) {
 		ret = -EIO;
+		goto err;
+	}
 
-	end = min_t(u64, eb->len, PAGE_CACHE_SIZE);
-	end = eb->start + end - 1;
+	/*
+	 * If this is a leaf block and it is corrupt, set the corrupt bit so
+	 * that we don't try and read the other copies of this block, just
+	 * return -EIO.
+	 */
+	if (found_level == 0 && check_leaf(root, eb)) {
+		set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
+		ret = -EIO;
+	}
+
+	if (!ret)
+		set_extent_buffer_uptodate(eb);
 err:
+	if (reads_done &&
+	    test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
+		btree_readahead_hook(root, eb, eb->start, ret);
+
+	if (ret) {
+		/*
+		 * our io error hook is going to dec the io pages
+		 * again, we have to make sure it has something
+		 * to decrement
+		 */
+		atomic_inc(&eb->io_pages);
+		clear_extent_buffer_uptodate(eb);
+	}
 	free_extent_buffer(eb);
 out:
 	return ret;
 }
 
+static int btree_io_failed_hook(struct page *page, int failed_mirror)
+{
+	struct extent_buffer *eb;
+	struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
+
+	eb = (struct extent_buffer *)page->private;
+	set_bit(EXTENT_BUFFER_IOERR, &eb->bflags);
+	eb->read_mirror = failed_mirror;
+	atomic_dec(&eb->io_pages);
+	if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
+		btree_readahead_hook(root, eb, eb->start, -EIO);
+	return -EIO;	/* we fixed nothing */
+}
+
 static void end_workqueue_bio(struct bio *bio, int err)
 {
 	struct end_io_wq *end_io_wq = bio->bi_private;
@@ -474,26 +698,42 @@ static void end_workqueue_bio(struct bio *bio, int err)
 
 	fs_info = end_io_wq->info;
 	end_io_wq->error = err;
-	end_io_wq->work.func = end_workqueue_fn;
-	end_io_wq->work.flags = 0;
-
-	if (bio->bi_rw & (1 << BIO_RW)) {
-		if (end_io_wq->metadata)
-			btrfs_queue_worker(&fs_info->endio_meta_write_workers,
-					   &end_io_wq->work);
+	btrfs_init_work(&end_io_wq->work, end_workqueue_fn, NULL, NULL);
+
+	if (bio->bi_rw & REQ_WRITE) {
+		if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA)
+			btrfs_queue_work(fs_info->endio_meta_write_workers,
+					 &end_io_wq->work);
+		else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE)
+			btrfs_queue_work(fs_info->endio_freespace_worker,
+					 &end_io_wq->work);
+		else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56)
+			btrfs_queue_work(fs_info->endio_raid56_workers,
+					 &end_io_wq->work);
 		else
-			btrfs_queue_worker(&fs_info->endio_write_workers,
-					   &end_io_wq->work);
+			btrfs_queue_work(fs_info->endio_write_workers,
+					 &end_io_wq->work);
 	} else {
-		if (end_io_wq->metadata)
-			btrfs_queue_worker(&fs_info->endio_meta_workers,
-					   &end_io_wq->work);
+		if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56)
+			btrfs_queue_work(fs_info->endio_raid56_workers,
+					 &end_io_wq->work);
+		else if (end_io_wq->metadata)
+			btrfs_queue_work(fs_info->endio_meta_workers,
+					 &end_io_wq->work);
 		else
-			btrfs_queue_worker(&fs_info->endio_workers,
-					   &end_io_wq->work);
+			btrfs_queue_work(fs_info->endio_workers,
+					 &end_io_wq->work);
 	}
 }
 
+/*
+ * For the metadata arg you want
+ *
+ * 0 - if data
+ * 1 - if normal metadta
+ * 2 - if writing to the free space cache area
+ * 3 - raid parity work
+ */
 int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
 			int metadata)
 {
@@ -517,26 +757,22 @@ int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
 unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
 {
 	unsigned long limit = min_t(unsigned long,
-				    info->workers.max_workers,
+				    info->thread_pool_size,
 				    info->fs_devices->open_devices);
 	return 256 * limit;
 }
 
-int btrfs_congested_async(struct btrfs_fs_info *info, int iodone)
-{
-	return atomic_read(&info->nr_async_bios) >
-		btrfs_async_submit_limit(info);
-}
-
 static void run_one_async_start(struct btrfs_work *work)
 {
-	struct btrfs_fs_info *fs_info;
 	struct async_submit_bio *async;
+	int ret;
 
 	async = container_of(work, struct  async_submit_bio, work);
-	fs_info = BTRFS_I(async->inode)->root->fs_info;
-	async->submit_bio_start(async->inode, async->rw, async->bio,
-			       async->mirror_num, async->bio_flags);
+	ret = async->submit_bio_start(async->inode, async->rw, async->bio,
+				      async->mirror_num, async->bio_flags,
+				      async->bio_offset);
+	if (ret)
+		async->error = ret;
 }
 
 static void run_one_async_done(struct btrfs_work *work)
@@ -551,14 +787,19 @@ static void run_one_async_done(struct btrfs_work *work)
 	limit = btrfs_async_submit_limit(fs_info);
 	limit = limit * 2 / 3;
 
-	atomic_dec(&fs_info->nr_async_submits);
-
-	if (atomic_read(&fs_info->nr_async_submits) < limit &&
+	if (atomic_dec_return(&fs_info->nr_async_submits) < limit &&
 	    waitqueue_active(&fs_info->async_submit_wait))
 		wake_up(&fs_info->async_submit_wait);
 
+	/* If an error occured we just want to clean up the bio and move on */
+	if (async->error) {
+		bio_endio(async->bio, async->error);
+		return;
+	}
+
 	async->submit_bio_done(async->inode, async->rw, async->bio,
-			       async->mirror_num, async->bio_flags);
+			       async->mirror_num, async->bio_flags,
+			       async->bio_offset);
 }
 
 static void run_one_async_free(struct btrfs_work *work)
@@ -572,6 +813,7 @@ static void run_one_async_free(struct btrfs_work *work)
 int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
 			int rw, struct bio *bio, int mirror_num,
 			unsigned long bio_flags,
+			u64 bio_offset,
 			extent_submit_bio_hook_t *submit_bio_start,
 			extent_submit_bio_hook_t *submit_bio_done)
 {
@@ -588,19 +830,20 @@ int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
 	async->submit_bio_start = submit_bio_start;
 	async->submit_bio_done = submit_bio_done;
 
-	async->work.func = run_one_async_start;
-	async->work.ordered_func = run_one_async_done;
-	async->work.ordered_free = run_one_async_free;
+	btrfs_init_work(&async->work, run_one_async_start,
+			run_one_async_done, run_one_async_free);
 
-	async->work.flags = 0;
 	async->bio_flags = bio_flags;
+	async->bio_offset = bio_offset;
+
+	async->error = 0;
 
 	atomic_inc(&fs_info->nr_async_submits);
 
-	if (rw & (1 << BIO_RW_SYNCIO))
-		btrfs_set_work_high_prio(&async->work);
+	if (rw & REQ_SYNC)
+		btrfs_set_work_high_priority(&async->work);
 
-	btrfs_queue_worker(&fs_info->workers, &async->work);
+	btrfs_queue_work(fs_info->workers, &async->work);
 
 	while (atomic_read(&fs_info->async_submit_draining) &&
 	      atomic_read(&fs_info->nr_async_submits)) {
@@ -613,176 +856,202 @@ int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
 
 static int btree_csum_one_bio(struct bio *bio)
 {
-	struct bio_vec *bvec = bio->bi_io_vec;
-	int bio_index = 0;
+	struct bio_vec *bvec;
 	struct btrfs_root *root;
+	int i, ret = 0;
 
-	WARN_ON(bio->bi_vcnt <= 0);
-	while (bio_index < bio->bi_vcnt) {
+	bio_for_each_segment_all(bvec, bio, i) {
 		root = BTRFS_I(bvec->bv_page->mapping->host)->root;
-		csum_dirty_buffer(root, bvec->bv_page);
-		bio_index++;
-		bvec++;
+		ret = csum_dirty_buffer(root, bvec->bv_page);
+		if (ret)
+			break;
 	}
-	return 0;
+
+	return ret;
 }
 
 static int __btree_submit_bio_start(struct inode *inode, int rw,
 				    struct bio *bio, int mirror_num,
-				    unsigned long bio_flags)
+				    unsigned long bio_flags,
+				    u64 bio_offset)
 {
 	/*
 	 * when we're called for a write, we're already in the async
 	 * submission context.  Just jump into btrfs_map_bio
 	 */
-	btree_csum_one_bio(bio);
-	return 0;
+	return btree_csum_one_bio(bio);
 }
 
 static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
-				 int mirror_num, unsigned long bio_flags)
+				 int mirror_num, unsigned long bio_flags,
+				 u64 bio_offset)
 {
+	int ret;
+
 	/*
 	 * when we're called for a write, we're already in the async
 	 * submission context.  Just jump into btrfs_map_bio
 	 */
-	return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
+	ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
+	if (ret)
+		bio_endio(bio, ret);
+	return ret;
+}
+
+static int check_async_write(struct inode *inode, unsigned long bio_flags)
+{
+	if (bio_flags & EXTENT_BIO_TREE_LOG)
+		return 0;
+#ifdef CONFIG_X86
+	if (cpu_has_xmm4_2)
+		return 0;
+#endif
+	return 1;
 }
 
 static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
-				 int mirror_num, unsigned long bio_flags)
+				 int mirror_num, unsigned long bio_flags,
+				 u64 bio_offset)
 {
+	int async = check_async_write(inode, bio_flags);
 	int ret;
 
-	ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
-					  bio, 1);
-	BUG_ON(ret);
-
-	if (!(rw & (1 << BIO_RW))) {
+	if (!(rw & REQ_WRITE)) {
 		/*
 		 * called for a read, do the setup so that checksum validation
 		 * can happen in the async kernel threads
 		 */
-		return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
-				     mirror_num, 0);
+		ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
+					  bio, 1);
+		if (ret)
+			goto out_w_error;
+		ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
+				    mirror_num, 0);
+	} else if (!async) {
+		ret = btree_csum_one_bio(bio);
+		if (ret)
+			goto out_w_error;
+		ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
+				    mirror_num, 0);
+	} else {
+		/*
+		 * kthread helpers are used to submit writes so that
+		 * checksumming can happen in parallel across all CPUs
+		 */
+		ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
+					  inode, rw, bio, mirror_num, 0,
+					  bio_offset,
+					  __btree_submit_bio_start,
+					  __btree_submit_bio_done);
 	}
 
-	/*
-	 * kthread helpers are used to submit writes so that checksumming
-	 * can happen in parallel across all CPUs
-	 */
-	return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
-				   inode, rw, bio, mirror_num, 0,
-				   __btree_submit_bio_start,
-				   __btree_submit_bio_done);
+	if (ret) {
+out_w_error:
+		bio_endio(bio, ret);
+	}
+	return ret;
 }
 
-static int btree_writepage(struct page *page, struct writeback_control *wbc)
+#ifdef CONFIG_MIGRATION
+static int btree_migratepage(struct address_space *mapping,
+			struct page *newpage, struct page *page,
+			enum migrate_mode mode)
 {
-	struct extent_io_tree *tree;
-	struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
-	struct extent_buffer *eb;
-	int was_dirty;
-
-	tree = &BTRFS_I(page->mapping->host)->io_tree;
-	if (!(current->flags & PF_MEMALLOC)) {
-		return extent_write_full_page(tree, page,
-					      btree_get_extent, wbc);
-	}
-
-	redirty_page_for_writepage(wbc, page);
-	eb = btrfs_find_tree_block(root, page_offset(page),
-				      PAGE_CACHE_SIZE);
-	WARN_ON(!eb);
-
-	was_dirty = test_and_set_bit(EXTENT_BUFFER_DIRTY, &eb->bflags);
-	if (!was_dirty) {
-		spin_lock(&root->fs_info->delalloc_lock);
-		root->fs_info->dirty_metadata_bytes += PAGE_CACHE_SIZE;
-		spin_unlock(&root->fs_info->delalloc_lock);
-	}
-	free_extent_buffer(eb);
-
-	unlock_page(page);
-	return 0;
+	/*
+	 * we can't safely write a btree page from here,
+	 * we haven't done the locking hook
+	 */
+	if (PageDirty(page))
+		return -EAGAIN;
+	/*
+	 * Buffers may be managed in a filesystem specific way.
+	 * We must have no buffers or drop them.
+	 */
+	if (page_has_private(page) &&
+	    !try_to_release_page(page, GFP_KERNEL))
+		return -EAGAIN;
+	return migrate_page(mapping, newpage, page, mode);
 }
+#endif
+
 
 static int btree_writepages(struct address_space *mapping,
 			    struct writeback_control *wbc)
 {
-	struct extent_io_tree *tree;
-	tree = &BTRFS_I(mapping->host)->io_tree;
+	struct btrfs_fs_info *fs_info;
+	int ret;
+
 	if (wbc->sync_mode == WB_SYNC_NONE) {
-		struct btrfs_root *root = BTRFS_I(mapping->host)->root;
-		u64 num_dirty;
-		unsigned long thresh = 32 * 1024 * 1024;
 
 		if (wbc->for_kupdate)
 			return 0;
 
+		fs_info = BTRFS_I(mapping->host)->root->fs_info;
 		/* this is a bit racy, but that's ok */
-		num_dirty = root->fs_info->dirty_metadata_bytes;
-		if (num_dirty < thresh)
+		ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes,
+					     BTRFS_DIRTY_METADATA_THRESH);
+		if (ret < 0)
 			return 0;
 	}
-	return extent_writepages(tree, mapping, btree_get_extent, wbc);
+	return btree_write_cache_pages(mapping, wbc);
 }
 
 static int btree_readpage(struct file *file, struct page *page)
 {
 	struct extent_io_tree *tree;
 	tree = &BTRFS_I(page->mapping->host)->io_tree;
-	return extent_read_full_page(tree, page, btree_get_extent);
+	return extent_read_full_page(tree, page, btree_get_extent, 0);
 }
 
 static int btree_releasepage(struct page *page, gfp_t gfp_flags)
 {
-	struct extent_io_tree *tree;
-	struct extent_map_tree *map;
-	int ret;
-
 	if (PageWriteback(page) || PageDirty(page))
 		return 0;
 
-	tree = &BTRFS_I(page->mapping->host)->io_tree;
-	map = &BTRFS_I(page->mapping->host)->extent_tree;
-
-	ret = try_release_extent_state(map, tree, page, gfp_flags);
-	if (!ret)
-		return 0;
-
-	ret = try_release_extent_buffer(tree, page);
-	if (ret == 1) {
-		ClearPagePrivate(page);
-		set_page_private(page, 0);
-		page_cache_release(page);
-	}
-
-	return ret;
+	return try_release_extent_buffer(page);
 }
 
-static void btree_invalidatepage(struct page *page, unsigned long offset)
+static void btree_invalidatepage(struct page *page, unsigned int offset,
+				 unsigned int length)
 {
 	struct extent_io_tree *tree;
 	tree = &BTRFS_I(page->mapping->host)->io_tree;
 	extent_invalidatepage(tree, page, offset);
 	btree_releasepage(page, GFP_NOFS);
 	if (PagePrivate(page)) {
-		printk(KERN_WARNING "btrfs warning page private not zero "
-		       "on page %llu\n", (unsigned long long)page_offset(page));
+		btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info,
+			   "page private not zero on page %llu",
+			   (unsigned long long)page_offset(page));
 		ClearPagePrivate(page);
 		set_page_private(page, 0);
 		page_cache_release(page);
 	}
 }
 
+static int btree_set_page_dirty(struct page *page)
+{
+#ifdef DEBUG
+	struct extent_buffer *eb;
+
+	BUG_ON(!PagePrivate(page));
+	eb = (struct extent_buffer *)page->private;
+	BUG_ON(!eb);
+	BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags));
+	BUG_ON(!atomic_read(&eb->refs));
+	btrfs_assert_tree_locked(eb);
+#endif
+	return __set_page_dirty_nobuffers(page);
+}
+
 static const struct address_space_operations btree_aops = {
 	.readpage	= btree_readpage,
-	.writepage	= btree_writepage,
 	.writepages	= btree_writepages,
 	.releasepage	= btree_releasepage,
 	.invalidatepage = btree_invalidatepage,
-	.sync_page	= block_sync_page,
+#ifdef CONFIG_MIGRATION
+	.migratepage	= btree_migratepage,
+#endif
+	.set_page_dirty = btree_set_page_dirty,
 };
 
 int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
@@ -796,42 +1065,70 @@ int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
 	if (!buf)
 		return 0;
 	read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
-				 buf, 0, 0, btree_get_extent, 0);
+				 buf, 0, WAIT_NONE, btree_get_extent, 0);
 	free_extent_buffer(buf);
 	return ret;
 }
 
+int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, u32 blocksize,
+			 int mirror_num, struct extent_buffer **eb)
+{
+	struct extent_buffer *buf = NULL;
+	struct inode *btree_inode = root->fs_info->btree_inode;
+	struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree;
+	int ret;
+
+	buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
+	if (!buf)
+		return 0;
+
+	set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags);
+
+	ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK,
+				       btree_get_extent, mirror_num);
+	if (ret) {
+		free_extent_buffer(buf);
+		return ret;
+	}
+
+	if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) {
+		free_extent_buffer(buf);
+		return -EIO;
+	} else if (extent_buffer_uptodate(buf)) {
+		*eb = buf;
+	} else {
+		free_extent_buffer(buf);
+	}
+	return 0;
+}
+
 struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
 					    u64 bytenr, u32 blocksize)
 {
-	struct inode *btree_inode = root->fs_info->btree_inode;
-	struct extent_buffer *eb;
-	eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
-				bytenr, blocksize, GFP_NOFS);
-	return eb;
+	return find_extent_buffer(root->fs_info, bytenr);
 }
 
 struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
 						 u64 bytenr, u32 blocksize)
 {
-	struct inode *btree_inode = root->fs_info->btree_inode;
-	struct extent_buffer *eb;
-
-	eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree,
-				 bytenr, blocksize, NULL, GFP_NOFS);
-	return eb;
+#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
+	if (unlikely(test_bit(BTRFS_ROOT_DUMMY_ROOT, &root->state)))
+		return alloc_test_extent_buffer(root->fs_info, bytenr,
+						blocksize);
+#endif
+	return alloc_extent_buffer(root->fs_info, bytenr, blocksize);
 }
 
 
 int btrfs_write_tree_block(struct extent_buffer *buf)
 {
-	return filemap_fdatawrite_range(buf->first_page->mapping, buf->start,
+	return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start,
 					buf->start + buf->len - 1);
 }
 
 int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
 {
-	return filemap_fdatawait_range(buf->first_page->mapping,
+	return filemap_fdatawait_range(buf->pages[0]->mapping,
 				       buf->start, buf->start + buf->len - 1);
 }
 
@@ -839,53 +1136,71 @@ struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
 				      u32 blocksize, u64 parent_transid)
 {
 	struct extent_buffer *buf = NULL;
-	struct inode *btree_inode = root->fs_info->btree_inode;
-	struct extent_io_tree *io_tree;
 	int ret;
 
-	io_tree = &BTRFS_I(btree_inode)->io_tree;
-
 	buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
 	if (!buf)
 		return NULL;
 
 	ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
-
-	if (ret == 0)
-		set_bit(EXTENT_BUFFER_UPTODATE, &buf->bflags);
+	if (ret) {
+		free_extent_buffer(buf);
+		return NULL;
+	}
 	return buf;
 
 }
 
-int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
-		     struct extent_buffer *buf)
+void clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+		      struct extent_buffer *buf)
 {
-	struct inode *btree_inode = root->fs_info->btree_inode;
+	struct btrfs_fs_info *fs_info = root->fs_info;
+
 	if (btrfs_header_generation(buf) ==
-	    root->fs_info->running_transaction->transid) {
+	    fs_info->running_transaction->transid) {
 		btrfs_assert_tree_locked(buf);
 
 		if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) {
-			spin_lock(&root->fs_info->delalloc_lock);
-			if (root->fs_info->dirty_metadata_bytes >= buf->len)
-				root->fs_info->dirty_metadata_bytes -= buf->len;
-			else
-				WARN_ON(1);
-			spin_unlock(&root->fs_info->delalloc_lock);
+			__percpu_counter_add(&fs_info->dirty_metadata_bytes,
+					     -buf->len,
+					     fs_info->dirty_metadata_batch);
+			/* ugh, clear_extent_buffer_dirty needs to lock the page */
+			btrfs_set_lock_blocking(buf);
+			clear_extent_buffer_dirty(buf);
 		}
+	}
+}
+
+static struct btrfs_subvolume_writers *btrfs_alloc_subvolume_writers(void)
+{
+	struct btrfs_subvolume_writers *writers;
+	int ret;
 
-		/* ugh, clear_extent_buffer_dirty needs to lock the page */
-		btrfs_set_lock_blocking(buf);
-		clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree,
-					  buf);
+	writers = kmalloc(sizeof(*writers), GFP_NOFS);
+	if (!writers)
+		return ERR_PTR(-ENOMEM);
+
+	ret = percpu_counter_init(&writers->counter, 0);
+	if (ret < 0) {
+		kfree(writers);
+		return ERR_PTR(ret);
 	}
-	return 0;
+
+	init_waitqueue_head(&writers->wait);
+	return writers;
+}
+
+static void
+btrfs_free_subvolume_writers(struct btrfs_subvolume_writers *writers)
+{
+	percpu_counter_destroy(&writers->counter);
+	kfree(writers);
 }
 
-static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
-			u32 stripesize, struct btrfs_root *root,
-			struct btrfs_fs_info *fs_info,
-			u64 objectid)
+static void __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
+			 u32 stripesize, struct btrfs_root *root,
+			 struct btrfs_fs_info *fs_info,
+			 u64 objectid)
 {
 	root->node = NULL;
 	root->commit_root = NULL;
@@ -893,117 +1208,178 @@ static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
 	root->nodesize = nodesize;
 	root->leafsize = leafsize;
 	root->stripesize = stripesize;
-	root->ref_cows = 0;
-	root->track_dirty = 0;
-	root->in_radix = 0;
-	root->clean_orphans = 0;
+	root->state = 0;
+	root->orphan_cleanup_state = 0;
 
-	root->fs_info = fs_info;
 	root->objectid = objectid;
 	root->last_trans = 0;
 	root->highest_objectid = 0;
+	root->nr_delalloc_inodes = 0;
+	root->nr_ordered_extents = 0;
 	root->name = NULL;
-	root->in_sysfs = 0;
 	root->inode_tree = RB_ROOT;
+	INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC);
+	root->block_rsv = NULL;
+	root->orphan_block_rsv = NULL;
 
 	INIT_LIST_HEAD(&root->dirty_list);
-	INIT_LIST_HEAD(&root->orphan_list);
 	INIT_LIST_HEAD(&root->root_list);
-	spin_lock_init(&root->node_lock);
-	spin_lock_init(&root->list_lock);
+	INIT_LIST_HEAD(&root->delalloc_inodes);
+	INIT_LIST_HEAD(&root->delalloc_root);
+	INIT_LIST_HEAD(&root->ordered_extents);
+	INIT_LIST_HEAD(&root->ordered_root);
+	INIT_LIST_HEAD(&root->logged_list[0]);
+	INIT_LIST_HEAD(&root->logged_list[1]);
+	spin_lock_init(&root->orphan_lock);
 	spin_lock_init(&root->inode_lock);
+	spin_lock_init(&root->delalloc_lock);
+	spin_lock_init(&root->ordered_extent_lock);
+	spin_lock_init(&root->accounting_lock);
+	spin_lock_init(&root->log_extents_lock[0]);
+	spin_lock_init(&root->log_extents_lock[1]);
 	mutex_init(&root->objectid_mutex);
 	mutex_init(&root->log_mutex);
+	mutex_init(&root->ordered_extent_mutex);
+	mutex_init(&root->delalloc_mutex);
 	init_waitqueue_head(&root->log_writer_wait);
 	init_waitqueue_head(&root->log_commit_wait[0]);
 	init_waitqueue_head(&root->log_commit_wait[1]);
+	INIT_LIST_HEAD(&root->log_ctxs[0]);
+	INIT_LIST_HEAD(&root->log_ctxs[1]);
 	atomic_set(&root->log_commit[0], 0);
 	atomic_set(&root->log_commit[1], 0);
 	atomic_set(&root->log_writers, 0);
-	root->log_batch = 0;
+	atomic_set(&root->log_batch, 0);
+	atomic_set(&root->orphan_inodes, 0);
+	atomic_set(&root->refs, 1);
+	atomic_set(&root->will_be_snapshoted, 0);
 	root->log_transid = 0;
+	root->log_transid_committed = -1;
 	root->last_log_commit = 0;
-	extent_io_tree_init(&root->dirty_log_pages,
-			     fs_info->btree_inode->i_mapping, GFP_NOFS);
+	if (fs_info)
+		extent_io_tree_init(&root->dirty_log_pages,
+				     fs_info->btree_inode->i_mapping);
 
 	memset(&root->root_key, 0, sizeof(root->root_key));
 	memset(&root->root_item, 0, sizeof(root->root_item));
 	memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
 	memset(&root->root_kobj, 0, sizeof(root->root_kobj));
-	root->defrag_trans_start = fs_info->generation;
+	if (fs_info)
+		root->defrag_trans_start = fs_info->generation;
+	else
+		root->defrag_trans_start = 0;
 	init_completion(&root->kobj_unregister);
-	root->defrag_running = 0;
 	root->root_key.objectid = objectid;
-	root->anon_super.s_root = NULL;
-	root->anon_super.s_dev = 0;
-	INIT_LIST_HEAD(&root->anon_super.s_list);
-	INIT_LIST_HEAD(&root->anon_super.s_instances);
-	init_rwsem(&root->anon_super.s_umount);
+	root->anon_dev = 0;
 
-	return 0;
+	spin_lock_init(&root->root_item_lock);
 }
 
-static int find_and_setup_root(struct btrfs_root *tree_root,
-			       struct btrfs_fs_info *fs_info,
-			       u64 objectid,
-			       struct btrfs_root *root)
+static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info)
 {
-	int ret;
-	u32 blocksize;
-	u64 generation;
+	struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS);
+	if (root)
+		root->fs_info = fs_info;
+	return root;
+}
 
-	__setup_root(tree_root->nodesize, tree_root->leafsize,
-		     tree_root->sectorsize, tree_root->stripesize,
-		     root, fs_info, objectid);
-	ret = btrfs_find_last_root(tree_root, objectid,
-				   &root->root_item, &root->root_key);
-	if (ret > 0)
-		return -ENOENT;
-	BUG_ON(ret);
+#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
+/* Should only be used by the testing infrastructure */
+struct btrfs_root *btrfs_alloc_dummy_root(void)
+{
+	struct btrfs_root *root;
 
-	generation = btrfs_root_generation(&root->root_item);
-	blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
-	root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
-				     blocksize, generation);
-	BUG_ON(!root->node);
-	root->commit_root = btrfs_root_node(root);
-	return 0;
+	root = btrfs_alloc_root(NULL);
+	if (!root)
+		return ERR_PTR(-ENOMEM);
+	__setup_root(4096, 4096, 4096, 4096, root, NULL, 1);
+	set_bit(BTRFS_ROOT_DUMMY_ROOT, &root->state);
+	root->alloc_bytenr = 0;
+
+	return root;
 }
+#endif
 
-int btrfs_free_log_root_tree(struct btrfs_trans_handle *trans,
-			     struct btrfs_fs_info *fs_info)
+struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans,
+				     struct btrfs_fs_info *fs_info,
+				     u64 objectid)
 {
-	struct extent_buffer *eb;
-	struct btrfs_root *log_root_tree = fs_info->log_root_tree;
-	u64 start = 0;
-	u64 end = 0;
-	int ret;
+	struct extent_buffer *leaf;
+	struct btrfs_root *tree_root = fs_info->tree_root;
+	struct btrfs_root *root;
+	struct btrfs_key key;
+	int ret = 0;
+	uuid_le uuid;
 
-	if (!log_root_tree)
-		return 0;
+	root = btrfs_alloc_root(fs_info);
+	if (!root)
+		return ERR_PTR(-ENOMEM);
 
-	while (1) {
-		ret = find_first_extent_bit(&log_root_tree->dirty_log_pages,
-				0, &start, &end, EXTENT_DIRTY | EXTENT_NEW);
-		if (ret)
-			break;
+	__setup_root(tree_root->nodesize, tree_root->leafsize,
+		     tree_root->sectorsize, tree_root->stripesize,
+		     root, fs_info, objectid);
+	root->root_key.objectid = objectid;
+	root->root_key.type = BTRFS_ROOT_ITEM_KEY;
+	root->root_key.offset = 0;
 
-		clear_extent_bits(&log_root_tree->dirty_log_pages, start, end,
-				  EXTENT_DIRTY | EXTENT_NEW, GFP_NOFS);
+	leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
+				      0, objectid, NULL, 0, 0, 0);
+	if (IS_ERR(leaf)) {
+		ret = PTR_ERR(leaf);
+		leaf = NULL;
+		goto fail;
 	}
-	eb = fs_info->log_root_tree->node;
 
-	WARN_ON(btrfs_header_level(eb) != 0);
-	WARN_ON(btrfs_header_nritems(eb) != 0);
+	memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
+	btrfs_set_header_bytenr(leaf, leaf->start);
+	btrfs_set_header_generation(leaf, trans->transid);
+	btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
+	btrfs_set_header_owner(leaf, objectid);
+	root->node = leaf;
 
-	ret = btrfs_free_reserved_extent(fs_info->tree_root,
-				eb->start, eb->len);
-	BUG_ON(ret);
+	write_extent_buffer(leaf, fs_info->fsid, btrfs_header_fsid(),
+			    BTRFS_FSID_SIZE);
+	write_extent_buffer(leaf, fs_info->chunk_tree_uuid,
+			    btrfs_header_chunk_tree_uuid(leaf),
+			    BTRFS_UUID_SIZE);
+	btrfs_mark_buffer_dirty(leaf);
 
-	free_extent_buffer(eb);
-	kfree(fs_info->log_root_tree);
-	fs_info->log_root_tree = NULL;
-	return 0;
+	root->commit_root = btrfs_root_node(root);
+	set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
+
+	root->root_item.flags = 0;
+	root->root_item.byte_limit = 0;
+	btrfs_set_root_bytenr(&root->root_item, leaf->start);
+	btrfs_set_root_generation(&root->root_item, trans->transid);
+	btrfs_set_root_level(&root->root_item, 0);
+	btrfs_set_root_refs(&root->root_item, 1);
+	btrfs_set_root_used(&root->root_item, leaf->len);
+	btrfs_set_root_last_snapshot(&root->root_item, 0);
+	btrfs_set_root_dirid(&root->root_item, 0);
+	uuid_le_gen(&uuid);
+	memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE);
+	root->root_item.drop_level = 0;
+
+	key.objectid = objectid;
+	key.type = BTRFS_ROOT_ITEM_KEY;
+	key.offset = 0;
+	ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item);
+	if (ret)
+		goto fail;
+
+	btrfs_tree_unlock(leaf);
+
+	return root;
+
+fail:
+	if (leaf) {
+		btrfs_tree_unlock(leaf);
+		free_extent_buffer(root->commit_root);
+		free_extent_buffer(leaf);
+	}
+	kfree(root);
+
+	return ERR_PTR(ret);
 }
 
 static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
@@ -1013,7 +1389,7 @@ static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
 	struct btrfs_root *tree_root = fs_info->tree_root;
 	struct extent_buffer *leaf;
 
-	root = kzalloc(sizeof(*root), GFP_NOFS);
+	root = btrfs_alloc_root(fs_info);
 	if (!root)
 		return ERR_PTR(-ENOMEM);
 
@@ -1024,16 +1400,19 @@ static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
 	root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
 	root->root_key.type = BTRFS_ROOT_ITEM_KEY;
 	root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
+
 	/*
+	 * DON'T set REF_COWS for log trees
+	 *
 	 * log trees do not get reference counted because they go away
 	 * before a real commit is actually done.  They do store pointers
 	 * to file data extents, and those reference counts still get
 	 * updated (along with back refs to the log tree).
 	 */
-	root->ref_cows = 0;
 
 	leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0,
-				      BTRFS_TREE_LOG_OBJECTID, NULL, 0, 0, 0);
+				      BTRFS_TREE_LOG_OBJECTID, NULL,
+				      0, 0, 0);
 	if (IS_ERR(leaf)) {
 		kfree(root);
 		return ERR_CAST(leaf);
@@ -1047,8 +1426,7 @@ static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
 	root->node = leaf;
 
 	write_extent_buffer(root->node, root->fs_info->fsid,
-			    (unsigned long)btrfs_header_fsid(root->node),
-			    BTRFS_FSID_SIZE);
+			    btrfs_header_fsid(), BTRFS_FSID_SIZE);
 	btrfs_mark_buffer_dirty(root->node);
 	btrfs_tree_unlock(root->node);
 	return root;
@@ -1081,96 +1459,169 @@ int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
 	log_root->root_key.offset = root->root_key.objectid;
 
 	inode_item = &log_root->root_item.inode;
-	inode_item->generation = cpu_to_le64(1);
-	inode_item->size = cpu_to_le64(3);
-	inode_item->nlink = cpu_to_le32(1);
-	inode_item->nbytes = cpu_to_le64(root->leafsize);
-	inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
+	btrfs_set_stack_inode_generation(inode_item, 1);
+	btrfs_set_stack_inode_size(inode_item, 3);
+	btrfs_set_stack_inode_nlink(inode_item, 1);
+	btrfs_set_stack_inode_nbytes(inode_item, root->leafsize);
+	btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
 
 	btrfs_set_root_node(&log_root->root_item, log_root->node);
 
 	WARN_ON(root->log_root);
 	root->log_root = log_root;
 	root->log_transid = 0;
+	root->log_transid_committed = -1;
 	root->last_log_commit = 0;
 	return 0;
 }
 
-struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root,
-					       struct btrfs_key *location)
+static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root,
+					       struct btrfs_key *key)
 {
 	struct btrfs_root *root;
 	struct btrfs_fs_info *fs_info = tree_root->fs_info;
 	struct btrfs_path *path;
-	struct extent_buffer *l;
 	u64 generation;
 	u32 blocksize;
-	int ret = 0;
+	int ret;
 
-	root = kzalloc(sizeof(*root), GFP_NOFS);
-	if (!root)
+	path = btrfs_alloc_path();
+	if (!path)
 		return ERR_PTR(-ENOMEM);
-	if (location->offset == (u64)-1) {
-		ret = find_and_setup_root(tree_root, fs_info,
-					  location->objectid, root);
-		if (ret) {
-			kfree(root);
-			return ERR_PTR(ret);
-		}
-		goto out;
+
+	root = btrfs_alloc_root(fs_info);
+	if (!root) {
+		ret = -ENOMEM;
+		goto alloc_fail;
 	}
 
 	__setup_root(tree_root->nodesize, tree_root->leafsize,
 		     tree_root->sectorsize, tree_root->stripesize,
-		     root, fs_info, location->objectid);
+		     root, fs_info, key->objectid);
 
-	path = btrfs_alloc_path();
-	BUG_ON(!path);
-	ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
-	if (ret == 0) {
-		l = path->nodes[0];
-		read_extent_buffer(l, &root->root_item,
-				btrfs_item_ptr_offset(l, path->slots[0]),
-				sizeof(root->root_item));
-		memcpy(&root->root_key, location, sizeof(*location));
-	}
-	btrfs_free_path(path);
+	ret = btrfs_find_root(tree_root, key, path,
+			      &root->root_item, &root->root_key);
 	if (ret) {
 		if (ret > 0)
 			ret = -ENOENT;
-		return ERR_PTR(ret);
+		goto find_fail;
 	}
 
 	generation = btrfs_root_generation(&root->root_item);
 	blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
 	root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
 				     blocksize, generation);
+	if (!root->node) {
+		ret = -ENOMEM;
+		goto find_fail;
+	} else if (!btrfs_buffer_uptodate(root->node, generation, 0)) {
+		ret = -EIO;
+		goto read_fail;
+	}
 	root->commit_root = btrfs_root_node(root);
-	BUG_ON(!root->node);
 out:
-	if (location->objectid != BTRFS_TREE_LOG_OBJECTID)
-		root->ref_cows = 1;
-
+	btrfs_free_path(path);
 	return root;
+
+read_fail:
+	free_extent_buffer(root->node);
+find_fail:
+	kfree(root);
+alloc_fail:
+	root = ERR_PTR(ret);
+	goto out;
 }
 
-struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
-					u64 root_objectid)
+struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root,
+				      struct btrfs_key *location)
 {
 	struct btrfs_root *root;
 
-	if (root_objectid == BTRFS_ROOT_TREE_OBJECTID)
-		return fs_info->tree_root;
-	if (root_objectid == BTRFS_EXTENT_TREE_OBJECTID)
-		return fs_info->extent_root;
+	root = btrfs_read_tree_root(tree_root, location);
+	if (IS_ERR(root))
+		return root;
+
+	if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
+		set_bit(BTRFS_ROOT_REF_COWS, &root->state);
+		btrfs_check_and_init_root_item(&root->root_item);
+	}
+
+	return root;
+}
 
+int btrfs_init_fs_root(struct btrfs_root *root)
+{
+	int ret;
+	struct btrfs_subvolume_writers *writers;
+
+	root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS);
+	root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned),
+					GFP_NOFS);
+	if (!root->free_ino_pinned || !root->free_ino_ctl) {
+		ret = -ENOMEM;
+		goto fail;
+	}
+
+	writers = btrfs_alloc_subvolume_writers();
+	if (IS_ERR(writers)) {
+		ret = PTR_ERR(writers);
+		goto fail;
+	}
+	root->subv_writers = writers;
+
+	btrfs_init_free_ino_ctl(root);
+	spin_lock_init(&root->cache_lock);
+	init_waitqueue_head(&root->cache_wait);
+
+	ret = get_anon_bdev(&root->anon_dev);
+	if (ret)
+		goto free_writers;
+	return 0;
+
+free_writers:
+	btrfs_free_subvolume_writers(root->subv_writers);
+fail:
+	kfree(root->free_ino_ctl);
+	kfree(root->free_ino_pinned);
+	return ret;
+}
+
+static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
+					       u64 root_id)
+{
+	struct btrfs_root *root;
+
+	spin_lock(&fs_info->fs_roots_radix_lock);
 	root = radix_tree_lookup(&fs_info->fs_roots_radix,
-				 (unsigned long)root_objectid);
+				 (unsigned long)root_id);
+	spin_unlock(&fs_info->fs_roots_radix_lock);
 	return root;
 }
 
-struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
-					      struct btrfs_key *location)
+int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info,
+			 struct btrfs_root *root)
+{
+	int ret;
+
+	ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
+	if (ret)
+		return ret;
+
+	spin_lock(&fs_info->fs_roots_radix_lock);
+	ret = radix_tree_insert(&fs_info->fs_roots_radix,
+				(unsigned long)root->root_key.objectid,
+				root);
+	if (ret == 0)
+		set_bit(BTRFS_ROOT_IN_RADIX, &root->state);
+	spin_unlock(&fs_info->fs_roots_radix_lock);
+	radix_tree_preload_end();
+
+	return ret;
+}
+
+struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info,
+				     struct btrfs_key *location,
+				     bool check_ref)
 {
 	struct btrfs_root *root;
 	int ret;
@@ -1185,41 +1636,41 @@ struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
 		return fs_info->dev_root;
 	if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
 		return fs_info->csum_root;
+	if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID)
+		return fs_info->quota_root ? fs_info->quota_root :
+					     ERR_PTR(-ENOENT);
+	if (location->objectid == BTRFS_UUID_TREE_OBJECTID)
+		return fs_info->uuid_root ? fs_info->uuid_root :
+					    ERR_PTR(-ENOENT);
 again:
-	spin_lock(&fs_info->fs_roots_radix_lock);
-	root = radix_tree_lookup(&fs_info->fs_roots_radix,
-				 (unsigned long)location->objectid);
-	spin_unlock(&fs_info->fs_roots_radix_lock);
-	if (root)
+	root = btrfs_lookup_fs_root(fs_info, location->objectid);
+	if (root) {
+		if (check_ref && btrfs_root_refs(&root->root_item) == 0)
+			return ERR_PTR(-ENOENT);
 		return root;
+	}
 
-	ret = btrfs_find_orphan_item(fs_info->tree_root, location->objectid);
-	if (ret == 0)
-		ret = -ENOENT;
-	if (ret < 0)
-		return ERR_PTR(ret);
-
-	root = btrfs_read_fs_root_no_radix(fs_info->tree_root, location);
+	root = btrfs_read_fs_root(fs_info->tree_root, location);
 	if (IS_ERR(root))
 		return root;
 
-	WARN_ON(btrfs_root_refs(&root->root_item) == 0);
-	set_anon_super(&root->anon_super, NULL);
+	if (check_ref && btrfs_root_refs(&root->root_item) == 0) {
+		ret = -ENOENT;
+		goto fail;
+	}
 
-	ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
+	ret = btrfs_init_fs_root(root);
 	if (ret)
 		goto fail;
 
-	spin_lock(&fs_info->fs_roots_radix_lock);
-	ret = radix_tree_insert(&fs_info->fs_roots_radix,
-				(unsigned long)root->root_key.objectid,
-				root);
-	if (ret == 0) {
-		root->in_radix = 1;
-		root->clean_orphans = 1;
-	}
-	spin_unlock(&fs_info->fs_roots_radix_lock);
-	radix_tree_preload_end();
+	ret = btrfs_find_item(fs_info->tree_root, NULL, BTRFS_ORPHAN_OBJECTID,
+			location->objectid, BTRFS_ORPHAN_ITEM_KEY, NULL);
+	if (ret < 0)
+		goto fail;
+	if (ret == 0)
+		set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state);
+
+	ret = btrfs_insert_fs_root(fs_info, root);
 	if (ret) {
 		if (ret == -EEXIST) {
 			free_fs_root(root);
@@ -1227,51 +1678,12 @@ again:
 		}
 		goto fail;
 	}
-
-	ret = btrfs_find_dead_roots(fs_info->tree_root,
-				    root->root_key.objectid);
-	WARN_ON(ret);
 	return root;
 fail:
 	free_fs_root(root);
 	return ERR_PTR(ret);
 }
 
-struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
-				      struct btrfs_key *location,
-				      const char *name, int namelen)
-{
-	return btrfs_read_fs_root_no_name(fs_info, location);
-#if 0
-	struct btrfs_root *root;
-	int ret;
-
-	root = btrfs_read_fs_root_no_name(fs_info, location);
-	if (!root)
-		return NULL;
-
-	if (root->in_sysfs)
-		return root;
-
-	ret = btrfs_set_root_name(root, name, namelen);
-	if (ret) {
-		free_extent_buffer(root->node);
-		kfree(root);
-		return ERR_PTR(ret);
-	}
-
-	ret = btrfs_sysfs_add_root(root);
-	if (ret) {
-		free_extent_buffer(root->node);
-		kfree(root->name);
-		kfree(root);
-		return ERR_PTR(ret);
-	}
-	root->in_sysfs = 1;
-	return root;
-#endif
-}
-
 static int btrfs_congested_fn(void *congested_data, int bdi_bits)
 {
 	struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
@@ -1279,7 +1691,8 @@ static int btrfs_congested_fn(void *congested_data, int bdi_bits)
 	struct btrfs_device *device;
 	struct backing_dev_info *bdi;
 
-	list_for_each_entry(device, &info->fs_devices->devices, dev_list) {
+	rcu_read_lock();
+	list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) {
 		if (!device->bdev)
 			continue;
 		bdi = blk_get_backing_dev_info(device->bdev);
@@ -1288,86 +1701,11 @@ static int btrfs_congested_fn(void *congested_data, int bdi_bits)
 			break;
 		}
 	}
+	rcu_read_unlock();
 	return ret;
 }
 
 /*
- * this unplugs every device on the box, and it is only used when page
- * is null
- */
-static void __unplug_io_fn(struct backing_dev_info *bdi, struct page *page)
-{
-	struct btrfs_device *device;
-	struct btrfs_fs_info *info;
-
-	info = (struct btrfs_fs_info *)bdi->unplug_io_data;
-	list_for_each_entry(device, &info->fs_devices->devices, dev_list) {
-		if (!device->bdev)
-			continue;
-
-		bdi = blk_get_backing_dev_info(device->bdev);
-		if (bdi->unplug_io_fn)
-			bdi->unplug_io_fn(bdi, page);
-	}
-}
-
-static void btrfs_unplug_io_fn(struct backing_dev_info *bdi, struct page *page)
-{
-	struct inode *inode;
-	struct extent_map_tree *em_tree;
-	struct extent_map *em;
-	struct address_space *mapping;
-	u64 offset;
-
-	/* the generic O_DIRECT read code does this */
-	if (1 || !page) {
-		__unplug_io_fn(bdi, page);
-		return;
-	}
-
-	/*
-	 * page->mapping may change at any time.  Get a consistent copy
-	 * and use that for everything below
-	 */
-	smp_mb();
-	mapping = page->mapping;
-	if (!mapping)
-		return;
-
-	inode = mapping->host;
-
-	/*
-	 * don't do the expensive searching for a small number of
-	 * devices
-	 */
-	if (BTRFS_I(inode)->root->fs_info->fs_devices->open_devices <= 2) {
-		__unplug_io_fn(bdi, page);
-		return;
-	}
-
-	offset = page_offset(page);
-
-	em_tree = &BTRFS_I(inode)->extent_tree;
-	read_lock(&em_tree->lock);
-	em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
-	read_unlock(&em_tree->lock);
-	if (!em) {
-		__unplug_io_fn(bdi, page);
-		return;
-	}
-
-	if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
-		free_extent_map(em);
-		__unplug_io_fn(bdi, page);
-		return;
-	}
-	offset = offset - em->start;
-	btrfs_unplug_page(&BTRFS_I(inode)->root->fs_info->mapping_tree,
-			  em->block_start + offset, page);
-	free_extent_map(em);
-}
-
-/*
  * If this fails, caller must call bdi_destroy() to get rid of the
  * bdi again.
  */
@@ -1375,64 +1713,17 @@ static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
 {
 	int err;
 
-	bdi->name = "btrfs";
 	bdi->capabilities = BDI_CAP_MAP_COPY;
-	err = bdi_init(bdi);
+	err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY);
 	if (err)
 		return err;
 
-	err = bdi_register(bdi, NULL, "btrfs-%d",
-				atomic_inc_return(&btrfs_bdi_num));
-	if (err) {
-		bdi_destroy(bdi);
-		return err;
-	}
-
 	bdi->ra_pages	= default_backing_dev_info.ra_pages;
-	bdi->unplug_io_fn	= btrfs_unplug_io_fn;
-	bdi->unplug_io_data	= info;
 	bdi->congested_fn	= btrfs_congested_fn;
 	bdi->congested_data	= info;
 	return 0;
 }
 
-static int bio_ready_for_csum(struct bio *bio)
-{
-	u64 length = 0;
-	u64 buf_len = 0;
-	u64 start = 0;
-	struct page *page;
-	struct extent_io_tree *io_tree = NULL;
-	struct btrfs_fs_info *info = NULL;
-	struct bio_vec *bvec;
-	int i;
-	int ret;
-
-	bio_for_each_segment(bvec, bio, i) {
-		page = bvec->bv_page;
-		if (page->private == EXTENT_PAGE_PRIVATE) {
-			length += bvec->bv_len;
-			continue;
-		}
-		if (!page->private) {
-			length += bvec->bv_len;
-			continue;
-		}
-		length = bvec->bv_len;
-		buf_len = page->private >> 2;
-		start = page_offset(page) + bvec->bv_offset;
-		io_tree = &BTRFS_I(page->mapping->host)->io_tree;
-		info = BTRFS_I(page->mapping->host)->root->fs_info;
-	}
-	/* are we fully contained in this bio? */
-	if (buf_len <= length)
-		return 1;
-
-	ret = extent_range_uptodate(io_tree, start + length,
-				    start + buf_len - 1);
-	return ret;
-}
-
 /*
  * called by the kthread helper functions to finally call the bio end_io
  * functions.  This is where read checksum verification actually happens
@@ -1441,57 +1732,56 @@ static void end_workqueue_fn(struct btrfs_work *work)
 {
 	struct bio *bio;
 	struct end_io_wq *end_io_wq;
-	struct btrfs_fs_info *fs_info;
 	int error;
 
 	end_io_wq = container_of(work, struct end_io_wq, work);
 	bio = end_io_wq->bio;
-	fs_info = end_io_wq->info;
 
-	/* metadata bio reads are special because the whole tree block must
-	 * be checksummed at once.  This makes sure the entire block is in
-	 * ram and up to date before trying to verify things.  For
-	 * blocksize <= pagesize, it is basically a noop
-	 */
-	if (!(bio->bi_rw & (1 << BIO_RW)) && end_io_wq->metadata &&
-	    !bio_ready_for_csum(bio)) {
-		btrfs_queue_worker(&fs_info->endio_meta_workers,
-				   &end_io_wq->work);
-		return;
-	}
 	error = end_io_wq->error;
 	bio->bi_private = end_io_wq->private;
 	bio->bi_end_io = end_io_wq->end_io;
 	kfree(end_io_wq);
-	bio_endio(bio, error);
+	bio_endio_nodec(bio, error);
 }
 
 static int cleaner_kthread(void *arg)
 {
 	struct btrfs_root *root = arg;
+	int again;
 
 	do {
-		smp_mb();
-		if (root->fs_info->closing)
-			break;
+		again = 0;
 
-		vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE);
+		/* Make the cleaner go to sleep early. */
+		if (btrfs_need_cleaner_sleep(root))
+			goto sleep;
+
+		if (!mutex_trylock(&root->fs_info->cleaner_mutex))
+			goto sleep;
 
-		if (!(root->fs_info->sb->s_flags & MS_RDONLY) &&
-		    mutex_trylock(&root->fs_info->cleaner_mutex)) {
-			btrfs_run_delayed_iputs(root);
-			btrfs_clean_old_snapshots(root);
+		/*
+		 * Avoid the problem that we change the status of the fs
+		 * during the above check and trylock.
+		 */
+		if (btrfs_need_cleaner_sleep(root)) {
 			mutex_unlock(&root->fs_info->cleaner_mutex);
+			goto sleep;
 		}
 
-		if (freezing(current)) {
-			refrigerator();
-		} else {
-			smp_mb();
-			if (root->fs_info->closing)
-				break;
+		btrfs_run_delayed_iputs(root);
+		again = btrfs_clean_one_deleted_snapshot(root);
+		mutex_unlock(&root->fs_info->cleaner_mutex);
+
+		/*
+		 * The defragger has dealt with the R/O remount and umount,
+		 * needn't do anything special here.
+		 */
+		btrfs_run_defrag_inodes(root->fs_info);
+sleep:
+		if (!try_to_freeze() && !again) {
 			set_current_state(TASK_INTERRUPTIBLE);
-			schedule();
+			if (!kthread_should_stop())
+				schedule();
 			__set_current_state(TASK_RUNNING);
 		}
 	} while (!kthread_should_stop());
@@ -1503,56 +1793,350 @@ static int transaction_kthread(void *arg)
 	struct btrfs_root *root = arg;
 	struct btrfs_trans_handle *trans;
 	struct btrfs_transaction *cur;
+	u64 transid;
 	unsigned long now;
 	unsigned long delay;
-	int ret;
+	bool cannot_commit;
 
 	do {
-		smp_mb();
-		if (root->fs_info->closing)
-			break;
-
-		delay = HZ * 30;
-		vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE);
+		cannot_commit = false;
+		delay = HZ * root->fs_info->commit_interval;
 		mutex_lock(&root->fs_info->transaction_kthread_mutex);
 
-		mutex_lock(&root->fs_info->trans_mutex);
+		spin_lock(&root->fs_info->trans_lock);
 		cur = root->fs_info->running_transaction;
 		if (!cur) {
-			mutex_unlock(&root->fs_info->trans_mutex);
+			spin_unlock(&root->fs_info->trans_lock);
 			goto sleep;
 		}
 
 		now = get_seconds();
-		if (now < cur->start_time || now - cur->start_time < 30) {
-			mutex_unlock(&root->fs_info->trans_mutex);
+		if (cur->state < TRANS_STATE_BLOCKED &&
+		    (now < cur->start_time ||
+		     now - cur->start_time < root->fs_info->commit_interval)) {
+			spin_unlock(&root->fs_info->trans_lock);
 			delay = HZ * 5;
 			goto sleep;
 		}
-		mutex_unlock(&root->fs_info->trans_mutex);
-		trans = btrfs_start_transaction(root, 1);
-		ret = btrfs_commit_transaction(trans, root);
-
+		transid = cur->transid;
+		spin_unlock(&root->fs_info->trans_lock);
+
+		/* If the file system is aborted, this will always fail. */
+		trans = btrfs_attach_transaction(root);
+		if (IS_ERR(trans)) {
+			if (PTR_ERR(trans) != -ENOENT)
+				cannot_commit = true;
+			goto sleep;
+		}
+		if (transid == trans->transid) {
+			btrfs_commit_transaction(trans, root);
+		} else {
+			btrfs_end_transaction(trans, root);
+		}
 sleep:
 		wake_up_process(root->fs_info->cleaner_kthread);
 		mutex_unlock(&root->fs_info->transaction_kthread_mutex);
 
-		if (freezing(current)) {
-			refrigerator();
-		} else {
-			if (root->fs_info->closing)
-				break;
+		if (unlikely(test_bit(BTRFS_FS_STATE_ERROR,
+				      &root->fs_info->fs_state)))
+			btrfs_cleanup_transaction(root);
+		if (!try_to_freeze()) {
 			set_current_state(TASK_INTERRUPTIBLE);
-			schedule_timeout(delay);
+			if (!kthread_should_stop() &&
+			    (!btrfs_transaction_blocked(root->fs_info) ||
+			     cannot_commit))
+				schedule_timeout(delay);
 			__set_current_state(TASK_RUNNING);
 		}
 	} while (!kthread_should_stop());
 	return 0;
 }
 
-struct btrfs_root *open_ctree(struct super_block *sb,
-			      struct btrfs_fs_devices *fs_devices,
-			      char *options)
+/*
+ * this will find the highest generation in the array of
+ * root backups.  The index of the highest array is returned,
+ * or -1 if we can't find anything.
+ *
+ * We check to make sure the array is valid by comparing the
+ * generation of the latest  root in the array with the generation
+ * in the super block.  If they don't match we pitch it.
+ */
+static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen)
+{
+	u64 cur;
+	int newest_index = -1;
+	struct btrfs_root_backup *root_backup;
+	int i;
+
+	for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) {
+		root_backup = info->super_copy->super_roots + i;
+		cur = btrfs_backup_tree_root_gen(root_backup);
+		if (cur == newest_gen)
+			newest_index = i;
+	}
+
+	/* check to see if we actually wrapped around */
+	if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) {
+		root_backup = info->super_copy->super_roots;
+		cur = btrfs_backup_tree_root_gen(root_backup);
+		if (cur == newest_gen)
+			newest_index = 0;
+	}
+	return newest_index;
+}
+
+
+/*
+ * find the oldest backup so we know where to store new entries
+ * in the backup array.  This will set the backup_root_index
+ * field in the fs_info struct
+ */
+static void find_oldest_super_backup(struct btrfs_fs_info *info,
+				     u64 newest_gen)
+{
+	int newest_index = -1;
+
+	newest_index = find_newest_super_backup(info, newest_gen);
+	/* if there was garbage in there, just move along */
+	if (newest_index == -1) {
+		info->backup_root_index = 0;
+	} else {
+		info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS;
+	}
+}
+
+/*
+ * copy all the root pointers into the super backup array.
+ * this will bump the backup pointer by one when it is
+ * done
+ */
+static void backup_super_roots(struct btrfs_fs_info *info)
+{
+	int next_backup;
+	struct btrfs_root_backup *root_backup;
+	int last_backup;
+
+	next_backup = info->backup_root_index;
+	last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) %
+		BTRFS_NUM_BACKUP_ROOTS;
+
+	/*
+	 * just overwrite the last backup if we're at the same generation
+	 * this happens only at umount
+	 */
+	root_backup = info->super_for_commit->super_roots + last_backup;
+	if (btrfs_backup_tree_root_gen(root_backup) ==
+	    btrfs_header_generation(info->tree_root->node))
+		next_backup = last_backup;
+
+	root_backup = info->super_for_commit->super_roots + next_backup;
+
+	/*
+	 * make sure all of our padding and empty slots get zero filled
+	 * regardless of which ones we use today
+	 */
+	memset(root_backup, 0, sizeof(*root_backup));
+
+	info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS;
+
+	btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start);
+	btrfs_set_backup_tree_root_gen(root_backup,
+			       btrfs_header_generation(info->tree_root->node));
+
+	btrfs_set_backup_tree_root_level(root_backup,
+			       btrfs_header_level(info->tree_root->node));
+
+	btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start);
+	btrfs_set_backup_chunk_root_gen(root_backup,
+			       btrfs_header_generation(info->chunk_root->node));
+	btrfs_set_backup_chunk_root_level(root_backup,
+			       btrfs_header_level(info->chunk_root->node));
+
+	btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start);
+	btrfs_set_backup_extent_root_gen(root_backup,
+			       btrfs_header_generation(info->extent_root->node));
+	btrfs_set_backup_extent_root_level(root_backup,
+			       btrfs_header_level(info->extent_root->node));
+
+	/*
+	 * we might commit during log recovery, which happens before we set
+	 * the fs_root.  Make sure it is valid before we fill it in.
+	 */
+	if (info->fs_root && info->fs_root->node) {
+		btrfs_set_backup_fs_root(root_backup,
+					 info->fs_root->node->start);
+		btrfs_set_backup_fs_root_gen(root_backup,
+			       btrfs_header_generation(info->fs_root->node));
+		btrfs_set_backup_fs_root_level(root_backup,
+			       btrfs_header_level(info->fs_root->node));
+	}
+
+	btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start);
+	btrfs_set_backup_dev_root_gen(root_backup,
+			       btrfs_header_generation(info->dev_root->node));
+	btrfs_set_backup_dev_root_level(root_backup,
+				       btrfs_header_level(info->dev_root->node));
+
+	btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start);
+	btrfs_set_backup_csum_root_gen(root_backup,
+			       btrfs_header_generation(info->csum_root->node));
+	btrfs_set_backup_csum_root_level(root_backup,
+			       btrfs_header_level(info->csum_root->node));
+
+	btrfs_set_backup_total_bytes(root_backup,
+			     btrfs_super_total_bytes(info->super_copy));
+	btrfs_set_backup_bytes_used(root_backup,
+			     btrfs_super_bytes_used(info->super_copy));
+	btrfs_set_backup_num_devices(root_backup,
+			     btrfs_super_num_devices(info->super_copy));
+
+	/*
+	 * if we don't copy this out to the super_copy, it won't get remembered
+	 * for the next commit
+	 */
+	memcpy(&info->super_copy->super_roots,
+	       &info->super_for_commit->super_roots,
+	       sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS);
+}
+
+/*
+ * this copies info out of the root backup array and back into
+ * the in-memory super block.  It is meant to help iterate through
+ * the array, so you send it the number of backups you've already
+ * tried and the last backup index you used.
+ *
+ * this returns -1 when it has tried all the backups
+ */
+static noinline int next_root_backup(struct btrfs_fs_info *info,
+				     struct btrfs_super_block *super,
+				     int *num_backups_tried, int *backup_index)
+{
+	struct btrfs_root_backup *root_backup;
+	int newest = *backup_index;
+
+	if (*num_backups_tried == 0) {
+		u64 gen = btrfs_super_generation(super);
+
+		newest = find_newest_super_backup(info, gen);
+		if (newest == -1)
+			return -1;
+
+		*backup_index = newest;
+		*num_backups_tried = 1;
+	} else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) {
+		/* we've tried all the backups, all done */
+		return -1;
+	} else {
+		/* jump to the next oldest backup */
+		newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) %
+			BTRFS_NUM_BACKUP_ROOTS;
+		*backup_index = newest;
+		*num_backups_tried += 1;
+	}
+	root_backup = super->super_roots + newest;
+
+	btrfs_set_super_generation(super,
+				   btrfs_backup_tree_root_gen(root_backup));
+	btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup));
+	btrfs_set_super_root_level(super,
+				   btrfs_backup_tree_root_level(root_backup));
+	btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup));
+
+	/*
+	 * fixme: the total bytes and num_devices need to match or we should
+	 * need a fsck
+	 */
+	btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup));
+	btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup));
+	return 0;
+}
+
+/* helper to cleanup workers */
+static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info)
+{
+	btrfs_destroy_workqueue(fs_info->fixup_workers);
+	btrfs_destroy_workqueue(fs_info->delalloc_workers);
+	btrfs_destroy_workqueue(fs_info->workers);
+	btrfs_destroy_workqueue(fs_info->endio_workers);
+	btrfs_destroy_workqueue(fs_info->endio_meta_workers);
+	btrfs_destroy_workqueue(fs_info->endio_raid56_workers);
+	btrfs_destroy_workqueue(fs_info->rmw_workers);
+	btrfs_destroy_workqueue(fs_info->endio_meta_write_workers);
+	btrfs_destroy_workqueue(fs_info->endio_write_workers);
+	btrfs_destroy_workqueue(fs_info->endio_freespace_worker);
+	btrfs_destroy_workqueue(fs_info->submit_workers);
+	btrfs_destroy_workqueue(fs_info->delayed_workers);
+	btrfs_destroy_workqueue(fs_info->caching_workers);
+	btrfs_destroy_workqueue(fs_info->readahead_workers);
+	btrfs_destroy_workqueue(fs_info->flush_workers);
+	btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers);
+	btrfs_destroy_workqueue(fs_info->extent_workers);
+}
+
+static void free_root_extent_buffers(struct btrfs_root *root)
+{
+	if (root) {
+		free_extent_buffer(root->node);
+		free_extent_buffer(root->commit_root);
+		root->node = NULL;
+		root->commit_root = NULL;
+	}
+}
+
+/* helper to cleanup tree roots */
+static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root)
+{
+	free_root_extent_buffers(info->tree_root);
+
+	free_root_extent_buffers(info->dev_root);
+	free_root_extent_buffers(info->extent_root);
+	free_root_extent_buffers(info->csum_root);
+	free_root_extent_buffers(info->quota_root);
+	free_root_extent_buffers(info->uuid_root);
+	if (chunk_root)
+		free_root_extent_buffers(info->chunk_root);
+}
+
+void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info)
+{
+	int ret;
+	struct btrfs_root *gang[8];
+	int i;
+
+	while (!list_empty(&fs_info->dead_roots)) {
+		gang[0] = list_entry(fs_info->dead_roots.next,
+				     struct btrfs_root, root_list);
+		list_del(&gang[0]->root_list);
+
+		if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) {
+			btrfs_drop_and_free_fs_root(fs_info, gang[0]);
+		} else {
+			free_extent_buffer(gang[0]->node);
+			free_extent_buffer(gang[0]->commit_root);
+			btrfs_put_fs_root(gang[0]);
+		}
+	}
+
+	while (1) {
+		ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
+					     (void **)gang, 0,
+					     ARRAY_SIZE(gang));
+		if (!ret)
+			break;
+		for (i = 0; i < ret; i++)
+			btrfs_drop_and_free_fs_root(fs_info, gang[i]);
+	}
+
+	if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
+		btrfs_free_log_root_tree(NULL, fs_info);
+		btrfs_destroy_pinned_extent(fs_info->tree_root,
+					    fs_info->pinned_extents);
+	}
+}
+
+int open_ctree(struct super_block *sb,
+	       struct btrfs_fs_devices *fs_devices,
+	       char *options)
 {
 	u32 sectorsize;
 	u32 nodesize;
@@ -1563,27 +2147,28 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	u64 features;
 	struct btrfs_key location;
 	struct buffer_head *bh;
-	struct btrfs_root *extent_root = kzalloc(sizeof(struct btrfs_root),
-						 GFP_NOFS);
-	struct btrfs_root *csum_root = kzalloc(sizeof(struct btrfs_root),
-						 GFP_NOFS);
-	struct btrfs_root *tree_root = kzalloc(sizeof(struct btrfs_root),
-					       GFP_NOFS);
-	struct btrfs_fs_info *fs_info = kzalloc(sizeof(*fs_info),
-						GFP_NOFS);
-	struct btrfs_root *chunk_root = kzalloc(sizeof(struct btrfs_root),
-						GFP_NOFS);
-	struct btrfs_root *dev_root = kzalloc(sizeof(struct btrfs_root),
-					      GFP_NOFS);
+	struct btrfs_super_block *disk_super;
+	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
+	struct btrfs_root *tree_root;
+	struct btrfs_root *extent_root;
+	struct btrfs_root *csum_root;
+	struct btrfs_root *chunk_root;
+	struct btrfs_root *dev_root;
+	struct btrfs_root *quota_root;
+	struct btrfs_root *uuid_root;
 	struct btrfs_root *log_tree_root;
-
 	int ret;
 	int err = -EINVAL;
-
-	struct btrfs_super_block *disk_super;
-
-	if (!extent_root || !tree_root || !fs_info ||
-	    !chunk_root || !dev_root || !csum_root) {
+	int num_backups_tried = 0;
+	int backup_index = 0;
+	int max_active;
+	int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND;
+	bool create_uuid_tree;
+	bool check_uuid_tree;
+
+	tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info);
+	chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info);
+	if (!tree_root || !chunk_root) {
 		err = -ENOMEM;
 		goto fail;
 	}
@@ -1600,56 +2185,129 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 		goto fail_srcu;
 	}
 
+	ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0);
+	if (ret) {
+		err = ret;
+		goto fail_bdi;
+	}
+	fs_info->dirty_metadata_batch = PAGE_CACHE_SIZE *
+					(1 + ilog2(nr_cpu_ids));
+
+	ret = percpu_counter_init(&fs_info->delalloc_bytes, 0);
+	if (ret) {
+		err = ret;
+		goto fail_dirty_metadata_bytes;
+	}
+
+	ret = percpu_counter_init(&fs_info->bio_counter, 0);
+	if (ret) {
+		err = ret;
+		goto fail_delalloc_bytes;
+	}
+
 	fs_info->btree_inode = new_inode(sb);
 	if (!fs_info->btree_inode) {
 		err = -ENOMEM;
-		goto fail_bdi;
+		goto fail_bio_counter;
 	}
 
+	mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
+
 	INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
+	INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC);
 	INIT_LIST_HEAD(&fs_info->trans_list);
 	INIT_LIST_HEAD(&fs_info->dead_roots);
 	INIT_LIST_HEAD(&fs_info->delayed_iputs);
-	INIT_LIST_HEAD(&fs_info->hashers);
-	INIT_LIST_HEAD(&fs_info->delalloc_inodes);
-	INIT_LIST_HEAD(&fs_info->ordered_operations);
+	INIT_LIST_HEAD(&fs_info->delalloc_roots);
 	INIT_LIST_HEAD(&fs_info->caching_block_groups);
-	spin_lock_init(&fs_info->delalloc_lock);
-	spin_lock_init(&fs_info->new_trans_lock);
-	spin_lock_init(&fs_info->ref_cache_lock);
+	spin_lock_init(&fs_info->delalloc_root_lock);
+	spin_lock_init(&fs_info->trans_lock);
 	spin_lock_init(&fs_info->fs_roots_radix_lock);
 	spin_lock_init(&fs_info->delayed_iput_lock);
+	spin_lock_init(&fs_info->defrag_inodes_lock);
+	spin_lock_init(&fs_info->free_chunk_lock);
+	spin_lock_init(&fs_info->tree_mod_seq_lock);
+	spin_lock_init(&fs_info->super_lock);
+	spin_lock_init(&fs_info->qgroup_op_lock);
+	spin_lock_init(&fs_info->buffer_lock);
+	rwlock_init(&fs_info->tree_mod_log_lock);
+	mutex_init(&fs_info->reloc_mutex);
+	mutex_init(&fs_info->delalloc_root_mutex);
+	seqlock_init(&fs_info->profiles_lock);
 
 	init_completion(&fs_info->kobj_unregister);
-	fs_info->tree_root = tree_root;
-	fs_info->extent_root = extent_root;
-	fs_info->csum_root = csum_root;
-	fs_info->chunk_root = chunk_root;
-	fs_info->dev_root = dev_root;
-	fs_info->fs_devices = fs_devices;
 	INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
 	INIT_LIST_HEAD(&fs_info->space_info);
+	INIT_LIST_HEAD(&fs_info->tree_mod_seq_list);
 	btrfs_mapping_init(&fs_info->mapping_tree);
+	btrfs_init_block_rsv(&fs_info->global_block_rsv,
+			     BTRFS_BLOCK_RSV_GLOBAL);
+	btrfs_init_block_rsv(&fs_info->delalloc_block_rsv,
+			     BTRFS_BLOCK_RSV_DELALLOC);
+	btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS);
+	btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK);
+	btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY);
+	btrfs_init_block_rsv(&fs_info->delayed_block_rsv,
+			     BTRFS_BLOCK_RSV_DELOPS);
 	atomic_set(&fs_info->nr_async_submits, 0);
 	atomic_set(&fs_info->async_delalloc_pages, 0);
 	atomic_set(&fs_info->async_submit_draining, 0);
 	atomic_set(&fs_info->nr_async_bios, 0);
+	atomic_set(&fs_info->defrag_running, 0);
+	atomic_set(&fs_info->qgroup_op_seq, 0);
+	atomic64_set(&fs_info->tree_mod_seq, 0);
 	fs_info->sb = sb;
 	fs_info->max_inline = 8192 * 1024;
 	fs_info->metadata_ratio = 0;
+	fs_info->defrag_inodes = RB_ROOT;
+	fs_info->free_chunk_space = 0;
+	fs_info->tree_mod_log = RB_ROOT;
+	fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
+	fs_info->avg_delayed_ref_runtime = div64_u64(NSEC_PER_SEC, 64);
+	/* readahead state */
+	INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT);
+	spin_lock_init(&fs_info->reada_lock);
 
 	fs_info->thread_pool_size = min_t(unsigned long,
 					  num_online_cpus() + 2, 8);
 
-	INIT_LIST_HEAD(&fs_info->ordered_extents);
-	spin_lock_init(&fs_info->ordered_extent_lock);
+	INIT_LIST_HEAD(&fs_info->ordered_roots);
+	spin_lock_init(&fs_info->ordered_root_lock);
+	fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root),
+					GFP_NOFS);
+	if (!fs_info->delayed_root) {
+		err = -ENOMEM;
+		goto fail_iput;
+	}
+	btrfs_init_delayed_root(fs_info->delayed_root);
+
+	mutex_init(&fs_info->scrub_lock);
+	atomic_set(&fs_info->scrubs_running, 0);
+	atomic_set(&fs_info->scrub_pause_req, 0);
+	atomic_set(&fs_info->scrubs_paused, 0);
+	atomic_set(&fs_info->scrub_cancel_req, 0);
+	init_waitqueue_head(&fs_info->replace_wait);
+	init_waitqueue_head(&fs_info->scrub_pause_wait);
+	fs_info->scrub_workers_refcnt = 0;
+#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
+	fs_info->check_integrity_print_mask = 0;
+#endif
+
+	spin_lock_init(&fs_info->balance_lock);
+	mutex_init(&fs_info->balance_mutex);
+	atomic_set(&fs_info->balance_running, 0);
+	atomic_set(&fs_info->balance_pause_req, 0);
+	atomic_set(&fs_info->balance_cancel_req, 0);
+	fs_info->balance_ctl = NULL;
+	init_waitqueue_head(&fs_info->balance_wait_q);
+	btrfs_init_async_reclaim_work(&fs_info->async_reclaim_work);
 
 	sb->s_blocksize = 4096;
 	sb->s_blocksize_bits = blksize_bits(4096);
 	sb->s_bdi = &fs_info->bdi;
 
 	fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID;
-	fs_info->btree_inode->i_nlink = 1;
+	set_nlink(fs_info->btree_inode, 1);
 	/*
 	 * we set the i_size on the btree inode to the max possible int.
 	 * the real end of the address space is determined by all of
@@ -1661,71 +2319,141 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 
 	RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node);
 	extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
-			     fs_info->btree_inode->i_mapping,
-			     GFP_NOFS);
-	extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree,
-			     GFP_NOFS);
+			     fs_info->btree_inode->i_mapping);
+	BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0;
+	extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree);
 
 	BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
 
 	BTRFS_I(fs_info->btree_inode)->root = tree_root;
 	memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
 	       sizeof(struct btrfs_key));
-	BTRFS_I(fs_info->btree_inode)->dummy_inode = 1;
-	insert_inode_hash(fs_info->btree_inode);
+	set_bit(BTRFS_INODE_DUMMY,
+		&BTRFS_I(fs_info->btree_inode)->runtime_flags);
+	btrfs_insert_inode_hash(fs_info->btree_inode);
 
 	spin_lock_init(&fs_info->block_group_cache_lock);
 	fs_info->block_group_cache_tree = RB_ROOT;
+	fs_info->first_logical_byte = (u64)-1;
 
 	extent_io_tree_init(&fs_info->freed_extents[0],
-			     fs_info->btree_inode->i_mapping, GFP_NOFS);
+			     fs_info->btree_inode->i_mapping);
 	extent_io_tree_init(&fs_info->freed_extents[1],
-			     fs_info->btree_inode->i_mapping, GFP_NOFS);
+			     fs_info->btree_inode->i_mapping);
 	fs_info->pinned_extents = &fs_info->freed_extents[0];
 	fs_info->do_barriers = 1;
 
 
-	mutex_init(&fs_info->trans_mutex);
 	mutex_init(&fs_info->ordered_operations_mutex);
+	mutex_init(&fs_info->ordered_extent_flush_mutex);
 	mutex_init(&fs_info->tree_log_mutex);
 	mutex_init(&fs_info->chunk_mutex);
 	mutex_init(&fs_info->transaction_kthread_mutex);
 	mutex_init(&fs_info->cleaner_mutex);
 	mutex_init(&fs_info->volume_mutex);
-	init_rwsem(&fs_info->extent_commit_sem);
+	init_rwsem(&fs_info->commit_root_sem);
 	init_rwsem(&fs_info->cleanup_work_sem);
 	init_rwsem(&fs_info->subvol_sem);
+	sema_init(&fs_info->uuid_tree_rescan_sem, 1);
+	fs_info->dev_replace.lock_owner = 0;
+	atomic_set(&fs_info->dev_replace.nesting_level, 0);
+	mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount);
+	mutex_init(&fs_info->dev_replace.lock_management_lock);
+	mutex_init(&fs_info->dev_replace.lock);
+
+	spin_lock_init(&fs_info->qgroup_lock);
+	mutex_init(&fs_info->qgroup_ioctl_lock);
+	fs_info->qgroup_tree = RB_ROOT;
+	fs_info->qgroup_op_tree = RB_ROOT;
+	INIT_LIST_HEAD(&fs_info->dirty_qgroups);
+	fs_info->qgroup_seq = 1;
+	fs_info->quota_enabled = 0;
+	fs_info->pending_quota_state = 0;
+	fs_info->qgroup_ulist = NULL;
+	mutex_init(&fs_info->qgroup_rescan_lock);
 
 	btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
 	btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
 
 	init_waitqueue_head(&fs_info->transaction_throttle);
 	init_waitqueue_head(&fs_info->transaction_wait);
+	init_waitqueue_head(&fs_info->transaction_blocked_wait);
 	init_waitqueue_head(&fs_info->async_submit_wait);
 
+	ret = btrfs_alloc_stripe_hash_table(fs_info);
+	if (ret) {
+		err = ret;
+		goto fail_alloc;
+	}
+
 	__setup_root(4096, 4096, 4096, 4096, tree_root,
 		     fs_info, BTRFS_ROOT_TREE_OBJECTID);
 
+	invalidate_bdev(fs_devices->latest_bdev);
 
+	/*
+	 * Read super block and check the signature bytes only
+	 */
 	bh = btrfs_read_dev_super(fs_devices->latest_bdev);
-	if (!bh)
-		goto fail_iput;
+	if (!bh) {
+		err = -EINVAL;
+		goto fail_alloc;
+	}
 
-	memcpy(&fs_info->super_copy, bh->b_data, sizeof(fs_info->super_copy));
-	memcpy(&fs_info->super_for_commit, &fs_info->super_copy,
-	       sizeof(fs_info->super_for_commit));
+	/*
+	 * We want to check superblock checksum, the type is stored inside.
+	 * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k).
+	 */
+	if (btrfs_check_super_csum(bh->b_data)) {
+		printk(KERN_ERR "BTRFS: superblock checksum mismatch\n");
+		err = -EINVAL;
+		goto fail_alloc;
+	}
+
+	/*
+	 * super_copy is zeroed at allocation time and we never touch the
+	 * following bytes up to INFO_SIZE, the checksum is calculated from
+	 * the whole block of INFO_SIZE
+	 */
+	memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy));
+	memcpy(fs_info->super_for_commit, fs_info->super_copy,
+	       sizeof(*fs_info->super_for_commit));
 	brelse(bh);
 
-	memcpy(fs_info->fsid, fs_info->super_copy.fsid, BTRFS_FSID_SIZE);
+	memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE);
+
+	ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY);
+	if (ret) {
+		printk(KERN_ERR "BTRFS: superblock contains fatal errors\n");
+		err = -EINVAL;
+		goto fail_alloc;
+	}
 
-	disk_super = &fs_info->super_copy;
+	disk_super = fs_info->super_copy;
 	if (!btrfs_super_root(disk_super))
-		goto fail_iput;
+		goto fail_alloc;
+
+	/* check FS state, whether FS is broken. */
+	if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR)
+		set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
+
+	/*
+	 * run through our array of backup supers and setup
+	 * our ring pointer to the oldest one
+	 */
+	generation = btrfs_super_generation(disk_super);
+	find_oldest_super_backup(fs_info, generation);
+
+	/*
+	 * In the long term, we'll store the compression type in the super
+	 * block, and it'll be used for per file compression control.
+	 */
+	fs_info->compress_type = BTRFS_COMPRESS_ZLIB;
 
 	ret = btrfs_parse_options(tree_root, options);
 	if (ret) {
 		err = ret;
-		goto fail_iput;
+		goto fail_alloc;
 	}
 
 	features = btrfs_super_incompat_flags(disk_super) &
@@ -1733,102 +2461,156 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	if (features) {
 		printk(KERN_ERR "BTRFS: couldn't mount because of "
 		       "unsupported optional features (%Lx).\n",
-		       (unsigned long long)features);
+		       features);
 		err = -EINVAL;
-		goto fail_iput;
+		goto fail_alloc;
+	}
+
+	if (btrfs_super_leafsize(disk_super) !=
+	    btrfs_super_nodesize(disk_super)) {
+		printk(KERN_ERR "BTRFS: couldn't mount because metadata "
+		       "blocksizes don't match.  node %d leaf %d\n",
+		       btrfs_super_nodesize(disk_super),
+		       btrfs_super_leafsize(disk_super));
+		err = -EINVAL;
+		goto fail_alloc;
+	}
+	if (btrfs_super_leafsize(disk_super) > BTRFS_MAX_METADATA_BLOCKSIZE) {
+		printk(KERN_ERR "BTRFS: couldn't mount because metadata "
+		       "blocksize (%d) was too large\n",
+		       btrfs_super_leafsize(disk_super));
+		err = -EINVAL;
+		goto fail_alloc;
 	}
 
 	features = btrfs_super_incompat_flags(disk_super);
-	if (!(features & BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF)) {
-		features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
-		btrfs_set_super_incompat_flags(disk_super, features);
+	features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
+	if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
+		features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
+
+	if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA)
+		printk(KERN_ERR "BTRFS: has skinny extents\n");
+
+	/*
+	 * flag our filesystem as having big metadata blocks if
+	 * they are bigger than the page size
+	 */
+	if (btrfs_super_leafsize(disk_super) > PAGE_CACHE_SIZE) {
+		if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA))
+			printk(KERN_INFO "BTRFS: flagging fs with big metadata feature\n");
+		features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA;
+	}
+
+	nodesize = btrfs_super_nodesize(disk_super);
+	leafsize = btrfs_super_leafsize(disk_super);
+	sectorsize = btrfs_super_sectorsize(disk_super);
+	stripesize = btrfs_super_stripesize(disk_super);
+	fs_info->dirty_metadata_batch = leafsize * (1 + ilog2(nr_cpu_ids));
+	fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids));
+
+	/*
+	 * mixed block groups end up with duplicate but slightly offset
+	 * extent buffers for the same range.  It leads to corruptions
+	 */
+	if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) &&
+	    (sectorsize != leafsize)) {
+		printk(KERN_WARNING "BTRFS: unequal leaf/node/sector sizes "
+				"are not allowed for mixed block groups on %s\n",
+				sb->s_id);
+		goto fail_alloc;
 	}
 
+	/*
+	 * Needn't use the lock because there is no other task which will
+	 * update the flag.
+	 */
+	btrfs_set_super_incompat_flags(disk_super, features);
+
 	features = btrfs_super_compat_ro_flags(disk_super) &
 		~BTRFS_FEATURE_COMPAT_RO_SUPP;
 	if (!(sb->s_flags & MS_RDONLY) && features) {
 		printk(KERN_ERR "BTRFS: couldn't mount RDWR because of "
 		       "unsupported option features (%Lx).\n",
-		       (unsigned long long)features);
+		       features);
 		err = -EINVAL;
-		goto fail_iput;
+		goto fail_alloc;
 	}
 
-	btrfs_init_workers(&fs_info->generic_worker,
-			   "genwork", 1, NULL);
+	max_active = fs_info->thread_pool_size;
+
+	fs_info->workers =
+		btrfs_alloc_workqueue("worker", flags | WQ_HIGHPRI,
+				      max_active, 16);
 
-	btrfs_init_workers(&fs_info->workers, "worker",
-			   fs_info->thread_pool_size,
-			   &fs_info->generic_worker);
+	fs_info->delalloc_workers =
+		btrfs_alloc_workqueue("delalloc", flags, max_active, 2);
 
-	btrfs_init_workers(&fs_info->delalloc_workers, "delalloc",
-			   fs_info->thread_pool_size,
-			   &fs_info->generic_worker);
+	fs_info->flush_workers =
+		btrfs_alloc_workqueue("flush_delalloc", flags, max_active, 0);
 
-	btrfs_init_workers(&fs_info->submit_workers, "submit",
-			   min_t(u64, fs_devices->num_devices,
-			   fs_info->thread_pool_size),
-			   &fs_info->generic_worker);
-	btrfs_init_workers(&fs_info->enospc_workers, "enospc",
-			   fs_info->thread_pool_size,
-			   &fs_info->generic_worker);
+	fs_info->caching_workers =
+		btrfs_alloc_workqueue("cache", flags, max_active, 0);
 
-	/* a higher idle thresh on the submit workers makes it much more
+	/*
+	 * a higher idle thresh on the submit workers makes it much more
 	 * likely that bios will be send down in a sane order to the
 	 * devices
 	 */
-	fs_info->submit_workers.idle_thresh = 64;
-
-	fs_info->workers.idle_thresh = 16;
-	fs_info->workers.ordered = 1;
-
-	fs_info->delalloc_workers.idle_thresh = 2;
-	fs_info->delalloc_workers.ordered = 1;
-
-	btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1,
-			   &fs_info->generic_worker);
-	btrfs_init_workers(&fs_info->endio_workers, "endio",
-			   fs_info->thread_pool_size,
-			   &fs_info->generic_worker);
-	btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta",
-			   fs_info->thread_pool_size,
-			   &fs_info->generic_worker);
-	btrfs_init_workers(&fs_info->endio_meta_write_workers,
-			   "endio-meta-write", fs_info->thread_pool_size,
-			   &fs_info->generic_worker);
-	btrfs_init_workers(&fs_info->endio_write_workers, "endio-write",
-			   fs_info->thread_pool_size,
-			   &fs_info->generic_worker);
+	fs_info->submit_workers =
+		btrfs_alloc_workqueue("submit", flags,
+				      min_t(u64, fs_devices->num_devices,
+					    max_active), 64);
+
+	fs_info->fixup_workers =
+		btrfs_alloc_workqueue("fixup", flags, 1, 0);
 
 	/*
 	 * endios are largely parallel and should have a very
 	 * low idle thresh
 	 */
-	fs_info->endio_workers.idle_thresh = 4;
-	fs_info->endio_meta_workers.idle_thresh = 4;
-
-	fs_info->endio_write_workers.idle_thresh = 2;
-	fs_info->endio_meta_write_workers.idle_thresh = 2;
-
-	btrfs_start_workers(&fs_info->workers, 1);
-	btrfs_start_workers(&fs_info->generic_worker, 1);
-	btrfs_start_workers(&fs_info->submit_workers, 1);
-	btrfs_start_workers(&fs_info->delalloc_workers, 1);
-	btrfs_start_workers(&fs_info->fixup_workers, 1);
-	btrfs_start_workers(&fs_info->endio_workers, 1);
-	btrfs_start_workers(&fs_info->endio_meta_workers, 1);
-	btrfs_start_workers(&fs_info->endio_meta_write_workers, 1);
-	btrfs_start_workers(&fs_info->endio_write_workers, 1);
-	btrfs_start_workers(&fs_info->enospc_workers, 1);
+	fs_info->endio_workers =
+		btrfs_alloc_workqueue("endio", flags, max_active, 4);
+	fs_info->endio_meta_workers =
+		btrfs_alloc_workqueue("endio-meta", flags, max_active, 4);
+	fs_info->endio_meta_write_workers =
+		btrfs_alloc_workqueue("endio-meta-write", flags, max_active, 2);
+	fs_info->endio_raid56_workers =
+		btrfs_alloc_workqueue("endio-raid56", flags, max_active, 4);
+	fs_info->rmw_workers =
+		btrfs_alloc_workqueue("rmw", flags, max_active, 2);
+	fs_info->endio_write_workers =
+		btrfs_alloc_workqueue("endio-write", flags, max_active, 2);
+	fs_info->endio_freespace_worker =
+		btrfs_alloc_workqueue("freespace-write", flags, max_active, 0);
+	fs_info->delayed_workers =
+		btrfs_alloc_workqueue("delayed-meta", flags, max_active, 0);
+	fs_info->readahead_workers =
+		btrfs_alloc_workqueue("readahead", flags, max_active, 2);
+	fs_info->qgroup_rescan_workers =
+		btrfs_alloc_workqueue("qgroup-rescan", flags, 1, 0);
+	fs_info->extent_workers =
+		btrfs_alloc_workqueue("extent-refs", flags,
+				      min_t(u64, fs_devices->num_devices,
+					    max_active), 8);
+
+	if (!(fs_info->workers && fs_info->delalloc_workers &&
+	      fs_info->submit_workers && fs_info->flush_workers &&
+	      fs_info->endio_workers && fs_info->endio_meta_workers &&
+	      fs_info->endio_meta_write_workers &&
+	      fs_info->endio_write_workers && fs_info->endio_raid56_workers &&
+	      fs_info->endio_freespace_worker && fs_info->rmw_workers &&
+	      fs_info->caching_workers && fs_info->readahead_workers &&
+	      fs_info->fixup_workers && fs_info->delayed_workers &&
+	      fs_info->fixup_workers && fs_info->extent_workers &&
+	      fs_info->qgroup_rescan_workers)) {
+		err = -ENOMEM;
+		goto fail_sb_buffer;
+	}
 
 	fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
 	fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
 				    4 * 1024 * 1024 / PAGE_CACHE_SIZE);
 
-	nodesize = btrfs_super_nodesize(disk_super);
-	leafsize = btrfs_super_leafsize(disk_super);
-	sectorsize = btrfs_super_sectorsize(disk_super);
-	stripesize = btrfs_super_stripesize(disk_super);
 	tree_root->nodesize = nodesize;
 	tree_root->leafsize = leafsize;
 	tree_root->sectorsize = sectorsize;
@@ -1837,9 +2619,14 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	sb->s_blocksize = sectorsize;
 	sb->s_blocksize_bits = blksize_bits(sectorsize);
 
-	if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
-		    sizeof(disk_super->magic))) {
-		printk(KERN_INFO "btrfs: valid FS not found on %s\n", sb->s_id);
+	if (btrfs_super_magic(disk_super) != BTRFS_MAGIC) {
+		printk(KERN_INFO "BTRFS: valid FS not found on %s\n", sb->s_id);
+		goto fail_sb_buffer;
+	}
+
+	if (sectorsize != PAGE_SIZE) {
+		printk(KERN_WARNING "BTRFS: Incompatible sector size(%lu) "
+		       "found on %s\n", (unsigned long)sectorsize, sb->s_id);
 		goto fail_sb_buffer;
 	}
 
@@ -1847,7 +2634,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	ret = btrfs_read_sys_array(tree_root);
 	mutex_unlock(&fs_info->chunk_mutex);
 	if (ret) {
-		printk(KERN_WARNING "btrfs: failed to read the system "
+		printk(KERN_WARNING "BTRFS: failed to read the system "
 		       "array on %s\n", sb->s_id);
 		goto fail_sb_buffer;
 	}
@@ -1862,30 +2649,38 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	chunk_root->node = read_tree_block(chunk_root,
 					   btrfs_super_chunk_root(disk_super),
 					   blocksize, generation);
-	BUG_ON(!chunk_root->node);
-	if (!test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) {
-		printk(KERN_WARNING "btrfs: failed to read chunk root on %s\n",
+	if (!chunk_root->node ||
+	    !test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) {
+		printk(KERN_WARNING "BTRFS: failed to read chunk root on %s\n",
 		       sb->s_id);
-		goto fail_chunk_root;
+		goto fail_tree_roots;
 	}
 	btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
 	chunk_root->commit_root = btrfs_root_node(chunk_root);
 
 	read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
-	   (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node),
-	   BTRFS_UUID_SIZE);
+	   btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE);
 
-	mutex_lock(&fs_info->chunk_mutex);
 	ret = btrfs_read_chunk_tree(chunk_root);
-	mutex_unlock(&fs_info->chunk_mutex);
 	if (ret) {
-		printk(KERN_WARNING "btrfs: failed to read chunk tree on %s\n",
+		printk(KERN_WARNING "BTRFS: failed to read chunk tree on %s\n",
 		       sb->s_id);
-		goto fail_chunk_root;
+		goto fail_tree_roots;
 	}
 
-	btrfs_close_extra_devices(fs_devices);
+	/*
+	 * keep the device that is marked to be the target device for the
+	 * dev_replace procedure
+	 */
+	btrfs_close_extra_devices(fs_info, fs_devices, 0);
+
+	if (!fs_devices->latest_bdev) {
+		printk(KERN_CRIT "BTRFS: failed to read devices on %s\n",
+		       sb->s_id);
+		goto fail_tree_roots;
+	}
 
+retry_root_backup:
 	blocksize = btrfs_level_size(tree_root,
 				     btrfs_super_root_level(disk_super));
 	generation = btrfs_super_generation(disk_super);
@@ -1893,50 +2688,129 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	tree_root->node = read_tree_block(tree_root,
 					  btrfs_super_root(disk_super),
 					  blocksize, generation);
-	if (!tree_root->node)
-		goto fail_chunk_root;
-	if (!test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) {
-		printk(KERN_WARNING "btrfs: failed to read tree root on %s\n",
+	if (!tree_root->node ||
+	    !test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) {
+		printk(KERN_WARNING "BTRFS: failed to read tree root on %s\n",
 		       sb->s_id);
-		goto fail_tree_root;
+
+		goto recovery_tree_root;
 	}
+
 	btrfs_set_root_node(&tree_root->root_item, tree_root->node);
 	tree_root->commit_root = btrfs_root_node(tree_root);
+	btrfs_set_root_refs(&tree_root->root_item, 1);
 
-	ret = find_and_setup_root(tree_root, fs_info,
-				  BTRFS_EXTENT_TREE_OBJECTID, extent_root);
-	if (ret)
-		goto fail_tree_root;
-	extent_root->track_dirty = 1;
+	location.objectid = BTRFS_EXTENT_TREE_OBJECTID;
+	location.type = BTRFS_ROOT_ITEM_KEY;
+	location.offset = 0;
 
-	ret = find_and_setup_root(tree_root, fs_info,
-				  BTRFS_DEV_TREE_OBJECTID, dev_root);
-	if (ret)
-		goto fail_extent_root;
-	dev_root->track_dirty = 1;
+	extent_root = btrfs_read_tree_root(tree_root, &location);
+	if (IS_ERR(extent_root)) {
+		ret = PTR_ERR(extent_root);
+		goto recovery_tree_root;
+	}
+	set_bit(BTRFS_ROOT_TRACK_DIRTY, &extent_root->state);
+	fs_info->extent_root = extent_root;
 
-	ret = find_and_setup_root(tree_root, fs_info,
-				  BTRFS_CSUM_TREE_OBJECTID, csum_root);
-	if (ret)
-		goto fail_dev_root;
+	location.objectid = BTRFS_DEV_TREE_OBJECTID;
+	dev_root = btrfs_read_tree_root(tree_root, &location);
+	if (IS_ERR(dev_root)) {
+		ret = PTR_ERR(dev_root);
+		goto recovery_tree_root;
+	}
+	set_bit(BTRFS_ROOT_TRACK_DIRTY, &dev_root->state);
+	fs_info->dev_root = dev_root;
+	btrfs_init_devices_late(fs_info);
 
-	csum_root->track_dirty = 1;
+	location.objectid = BTRFS_CSUM_TREE_OBJECTID;
+	csum_root = btrfs_read_tree_root(tree_root, &location);
+	if (IS_ERR(csum_root)) {
+		ret = PTR_ERR(csum_root);
+		goto recovery_tree_root;
+	}
+	set_bit(BTRFS_ROOT_TRACK_DIRTY, &csum_root->state);
+	fs_info->csum_root = csum_root;
 
-	ret = btrfs_read_block_groups(extent_root);
-	if (ret) {
-		printk(KERN_ERR "Failed to read block groups: %d\n", ret);
-		goto fail_block_groups;
+	location.objectid = BTRFS_QUOTA_TREE_OBJECTID;
+	quota_root = btrfs_read_tree_root(tree_root, &location);
+	if (!IS_ERR(quota_root)) {
+		set_bit(BTRFS_ROOT_TRACK_DIRTY, &quota_root->state);
+		fs_info->quota_enabled = 1;
+		fs_info->pending_quota_state = 1;
+		fs_info->quota_root = quota_root;
+	}
+
+	location.objectid = BTRFS_UUID_TREE_OBJECTID;
+	uuid_root = btrfs_read_tree_root(tree_root, &location);
+	if (IS_ERR(uuid_root)) {
+		ret = PTR_ERR(uuid_root);
+		if (ret != -ENOENT)
+			goto recovery_tree_root;
+		create_uuid_tree = true;
+		check_uuid_tree = false;
+	} else {
+		set_bit(BTRFS_ROOT_TRACK_DIRTY, &uuid_root->state);
+		fs_info->uuid_root = uuid_root;
+		create_uuid_tree = false;
+		check_uuid_tree =
+		    generation != btrfs_super_uuid_tree_generation(disk_super);
 	}
 
 	fs_info->generation = generation;
 	fs_info->last_trans_committed = generation;
-	fs_info->data_alloc_profile = (u64)-1;
-	fs_info->metadata_alloc_profile = (u64)-1;
-	fs_info->system_alloc_profile = fs_info->metadata_alloc_profile;
+
+	ret = btrfs_recover_balance(fs_info);
+	if (ret) {
+		printk(KERN_WARNING "BTRFS: failed to recover balance\n");
+		goto fail_block_groups;
+	}
+
+	ret = btrfs_init_dev_stats(fs_info);
+	if (ret) {
+		printk(KERN_ERR "BTRFS: failed to init dev_stats: %d\n",
+		       ret);
+		goto fail_block_groups;
+	}
+
+	ret = btrfs_init_dev_replace(fs_info);
+	if (ret) {
+		pr_err("BTRFS: failed to init dev_replace: %d\n", ret);
+		goto fail_block_groups;
+	}
+
+	btrfs_close_extra_devices(fs_info, fs_devices, 1);
+
+	ret = btrfs_sysfs_add_one(fs_info);
+	if (ret) {
+		pr_err("BTRFS: failed to init sysfs interface: %d\n", ret);
+		goto fail_block_groups;
+	}
+
+	ret = btrfs_init_space_info(fs_info);
+	if (ret) {
+		printk(KERN_ERR "BTRFS: Failed to initial space info: %d\n", ret);
+		goto fail_sysfs;
+	}
+
+	ret = btrfs_read_block_groups(extent_root);
+	if (ret) {
+		printk(KERN_ERR "BTRFS: Failed to read block groups: %d\n", ret);
+		goto fail_sysfs;
+	}
+	fs_info->num_tolerated_disk_barrier_failures =
+		btrfs_calc_num_tolerated_disk_barrier_failures(fs_info);
+	if (fs_info->fs_devices->missing_devices >
+	     fs_info->num_tolerated_disk_barrier_failures &&
+	    !(sb->s_flags & MS_RDONLY)) {
+		printk(KERN_WARNING "BTRFS: "
+			"too many missing devices, writeable mount is not allowed\n");
+		goto fail_sysfs;
+	}
+
 	fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
 					       "btrfs-cleaner");
 	if (IS_ERR(fs_info->cleaner_kthread))
-		goto fail_block_groups;
+		goto fail_sysfs;
 
 	fs_info->transaction_kthread = kthread_run(transaction_kthread,
 						   tree_root,
@@ -1947,26 +2821,50 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	if (!btrfs_test_opt(tree_root, SSD) &&
 	    !btrfs_test_opt(tree_root, NOSSD) &&
 	    !fs_info->fs_devices->rotating) {
-		printk(KERN_INFO "Btrfs detected SSD devices, enabling SSD "
+		printk(KERN_INFO "BTRFS: detected SSD devices, enabling SSD "
 		       "mode\n");
 		btrfs_set_opt(fs_info->mount_opt, SSD);
 	}
 
+	/* Set the real inode map cache flag */
+	if (btrfs_test_opt(tree_root, CHANGE_INODE_CACHE))
+		btrfs_set_opt(tree_root->fs_info->mount_opt, INODE_MAP_CACHE);
+
+#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
+	if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) {
+		ret = btrfsic_mount(tree_root, fs_devices,
+				    btrfs_test_opt(tree_root,
+					CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ?
+				    1 : 0,
+				    fs_info->check_integrity_print_mask);
+		if (ret)
+			printk(KERN_WARNING "BTRFS: failed to initialize"
+			       " integrity check module %s\n", sb->s_id);
+	}
+#endif
+	ret = btrfs_read_qgroup_config(fs_info);
+	if (ret)
+		goto fail_trans_kthread;
+
+	/* do not make disk changes in broken FS */
 	if (btrfs_super_log_root(disk_super) != 0) {
 		u64 bytenr = btrfs_super_log_root(disk_super);
 
 		if (fs_devices->rw_devices == 0) {
-			printk(KERN_WARNING "Btrfs log replay required "
+			printk(KERN_WARNING "BTRFS: log replay required "
 			       "on RO media\n");
 			err = -EIO;
-			goto fail_trans_kthread;
+			goto fail_qgroup;
 		}
 		blocksize =
 		     btrfs_level_size(tree_root,
 				      btrfs_super_log_root_level(disk_super));
 
-		log_tree_root = kzalloc(sizeof(struct btrfs_root),
-						      GFP_NOFS);
+		log_tree_root = btrfs_alloc_root(fs_info);
+		if (!log_tree_root) {
+			err = -ENOMEM;
+			goto fail_qgroup;
+		}
 
 		__setup_root(nodesize, leafsize, sectorsize, stripesize,
 			     log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
@@ -1974,46 +2872,119 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 		log_tree_root->node = read_tree_block(tree_root, bytenr,
 						      blocksize,
 						      generation + 1);
+		if (!log_tree_root->node ||
+		    !extent_buffer_uptodate(log_tree_root->node)) {
+			printk(KERN_ERR "BTRFS: failed to read log tree\n");
+			free_extent_buffer(log_tree_root->node);
+			kfree(log_tree_root);
+			goto fail_qgroup;
+		}
+		/* returns with log_tree_root freed on success */
 		ret = btrfs_recover_log_trees(log_tree_root);
-		BUG_ON(ret);
+		if (ret) {
+			btrfs_error(tree_root->fs_info, ret,
+				    "Failed to recover log tree");
+			free_extent_buffer(log_tree_root->node);
+			kfree(log_tree_root);
+			goto fail_qgroup;
+		}
 
 		if (sb->s_flags & MS_RDONLY) {
-			ret =  btrfs_commit_super(tree_root);
-			BUG_ON(ret);
+			ret = btrfs_commit_super(tree_root);
+			if (ret)
+				goto fail_qgroup;
 		}
 	}
 
 	ret = btrfs_find_orphan_roots(tree_root);
-	BUG_ON(ret);
+	if (ret)
+		goto fail_qgroup;
 
 	if (!(sb->s_flags & MS_RDONLY)) {
+		ret = btrfs_cleanup_fs_roots(fs_info);
+		if (ret)
+			goto fail_qgroup;
+
+		mutex_lock(&fs_info->cleaner_mutex);
 		ret = btrfs_recover_relocation(tree_root);
+		mutex_unlock(&fs_info->cleaner_mutex);
 		if (ret < 0) {
 			printk(KERN_WARNING
-			       "btrfs: failed to recover relocation\n");
+			       "BTRFS: failed to recover relocation\n");
 			err = -EINVAL;
-			goto fail_trans_kthread;
+			goto fail_qgroup;
 		}
 	}
 
 	location.objectid = BTRFS_FS_TREE_OBJECTID;
 	location.type = BTRFS_ROOT_ITEM_KEY;
-	location.offset = (u64)-1;
+	location.offset = 0;
 
 	fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
-	if (!fs_info->fs_root)
-		goto fail_trans_kthread;
+	if (IS_ERR(fs_info->fs_root)) {
+		err = PTR_ERR(fs_info->fs_root);
+		goto fail_qgroup;
+	}
 
-	if (!(sb->s_flags & MS_RDONLY)) {
-		down_read(&fs_info->cleanup_work_sem);
-		btrfs_orphan_cleanup(fs_info->fs_root);
+	if (sb->s_flags & MS_RDONLY)
+		return 0;
+
+	down_read(&fs_info->cleanup_work_sem);
+	if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) ||
+	    (ret = btrfs_orphan_cleanup(fs_info->tree_root))) {
 		up_read(&fs_info->cleanup_work_sem);
+		close_ctree(tree_root);
+		return ret;
+	}
+	up_read(&fs_info->cleanup_work_sem);
+
+	ret = btrfs_resume_balance_async(fs_info);
+	if (ret) {
+		printk(KERN_WARNING "BTRFS: failed to resume balance\n");
+		close_ctree(tree_root);
+		return ret;
+	}
+
+	ret = btrfs_resume_dev_replace_async(fs_info);
+	if (ret) {
+		pr_warn("BTRFS: failed to resume dev_replace\n");
+		close_ctree(tree_root);
+		return ret;
 	}
 
-	return tree_root;
+	btrfs_qgroup_rescan_resume(fs_info);
 
+	if (create_uuid_tree) {
+		pr_info("BTRFS: creating UUID tree\n");
+		ret = btrfs_create_uuid_tree(fs_info);
+		if (ret) {
+			pr_warn("BTRFS: failed to create the UUID tree %d\n",
+				ret);
+			close_ctree(tree_root);
+			return ret;
+		}
+	} else if (check_uuid_tree ||
+		   btrfs_test_opt(tree_root, RESCAN_UUID_TREE)) {
+		pr_info("BTRFS: checking UUID tree\n");
+		ret = btrfs_check_uuid_tree(fs_info);
+		if (ret) {
+			pr_warn("BTRFS: failed to check the UUID tree %d\n",
+				ret);
+			close_ctree(tree_root);
+			return ret;
+		}
+	} else {
+		fs_info->update_uuid_tree_gen = 1;
+	}
+
+	return 0;
+
+fail_qgroup:
+	btrfs_free_qgroup_config(fs_info);
 fail_trans_kthread:
 	kthread_stop(fs_info->transaction_kthread);
+	btrfs_cleanup_transaction(fs_info->tree_root);
+	btrfs_free_fs_roots(fs_info);
 fail_cleaner:
 	kthread_stop(fs_info->cleaner_kthread);
 
@@ -2022,71 +2993,75 @@ fail_cleaner:
 	 * kthreads
 	 */
 	filemap_write_and_wait(fs_info->btree_inode->i_mapping);
-	invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
+
+fail_sysfs:
+	btrfs_sysfs_remove_one(fs_info);
 
 fail_block_groups:
+	btrfs_put_block_group_cache(fs_info);
 	btrfs_free_block_groups(fs_info);
-	free_extent_buffer(csum_root->node);
-	free_extent_buffer(csum_root->commit_root);
-fail_dev_root:
-	free_extent_buffer(dev_root->node);
-	free_extent_buffer(dev_root->commit_root);
-fail_extent_root:
-	free_extent_buffer(extent_root->node);
-	free_extent_buffer(extent_root->commit_root);
-fail_tree_root:
-	free_extent_buffer(tree_root->node);
-	free_extent_buffer(tree_root->commit_root);
-fail_chunk_root:
-	free_extent_buffer(chunk_root->node);
-	free_extent_buffer(chunk_root->commit_root);
-fail_sb_buffer:
-	btrfs_stop_workers(&fs_info->generic_worker);
-	btrfs_stop_workers(&fs_info->fixup_workers);
-	btrfs_stop_workers(&fs_info->delalloc_workers);
-	btrfs_stop_workers(&fs_info->workers);
-	btrfs_stop_workers(&fs_info->endio_workers);
-	btrfs_stop_workers(&fs_info->endio_meta_workers);
-	btrfs_stop_workers(&fs_info->endio_meta_write_workers);
-	btrfs_stop_workers(&fs_info->endio_write_workers);
-	btrfs_stop_workers(&fs_info->submit_workers);
-	btrfs_stop_workers(&fs_info->enospc_workers);
-fail_iput:
+
+fail_tree_roots:
+	free_root_pointers(fs_info, 1);
 	invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
-	iput(fs_info->btree_inode);
 
-	btrfs_close_devices(fs_info->fs_devices);
+fail_sb_buffer:
+	btrfs_stop_all_workers(fs_info);
+fail_alloc:
+fail_iput:
 	btrfs_mapping_tree_free(&fs_info->mapping_tree);
+
+	iput(fs_info->btree_inode);
+fail_bio_counter:
+	percpu_counter_destroy(&fs_info->bio_counter);
+fail_delalloc_bytes:
+	percpu_counter_destroy(&fs_info->delalloc_bytes);
+fail_dirty_metadata_bytes:
+	percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
 fail_bdi:
 	bdi_destroy(&fs_info->bdi);
 fail_srcu:
 	cleanup_srcu_struct(&fs_info->subvol_srcu);
 fail:
-	kfree(extent_root);
-	kfree(tree_root);
-	kfree(fs_info);
-	kfree(chunk_root);
-	kfree(dev_root);
-	kfree(csum_root);
-	return ERR_PTR(err);
+	btrfs_free_stripe_hash_table(fs_info);
+	btrfs_close_devices(fs_info->fs_devices);
+	return err;
+
+recovery_tree_root:
+	if (!btrfs_test_opt(tree_root, RECOVERY))
+		goto fail_tree_roots;
+
+	free_root_pointers(fs_info, 0);
+
+	/* don't use the log in recovery mode, it won't be valid */
+	btrfs_set_super_log_root(disk_super, 0);
+
+	/* we can't trust the free space cache either */
+	btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE);
+
+	ret = next_root_backup(fs_info, fs_info->super_copy,
+			       &num_backups_tried, &backup_index);
+	if (ret == -1)
+		goto fail_block_groups;
+	goto retry_root_backup;
 }
 
 static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
 {
-	char b[BDEVNAME_SIZE];
-
 	if (uptodate) {
 		set_buffer_uptodate(bh);
 	} else {
-		if (!buffer_eopnotsupp(bh) && printk_ratelimit()) {
-			printk(KERN_WARNING "lost page write due to "
-					"I/O error on %s\n",
-				       bdevname(bh->b_bdev, b));
-		}
+		struct btrfs_device *device = (struct btrfs_device *)
+			bh->b_private;
+
+		printk_ratelimited_in_rcu(KERN_WARNING "BTRFS: lost page write due to "
+					  "I/O error on %s\n",
+					  rcu_str_deref(device->name));
 		/* note, we dont' set_buffer_write_io_error because we have
 		 * our own ways of dealing with the IO errors
 		 */
 		clear_buffer_uptodate(bh);
+		btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS);
 	}
 	unlock_buffer(bh);
 	put_bh(bh);
@@ -2108,16 +3083,17 @@ struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
 	 */
 	for (i = 0; i < 1; i++) {
 		bytenr = btrfs_sb_offset(i);
-		if (bytenr + 4096 >= i_size_read(bdev->bd_inode))
+		if (bytenr + BTRFS_SUPER_INFO_SIZE >=
+					i_size_read(bdev->bd_inode))
 			break;
-		bh = __bread(bdev, bytenr / 4096, 4096);
+		bh = __bread(bdev, bytenr / 4096,
+					BTRFS_SUPER_INFO_SIZE);
 		if (!bh)
 			continue;
 
 		super = (struct btrfs_super_block *)bh->b_data;
 		if (btrfs_super_bytenr(super) != bytenr ||
-		    strncmp((char *)(&super->magic), BTRFS_MAGIC,
-			    sizeof(super->magic))) {
+		    btrfs_super_magic(super) != BTRFS_MAGIC) {
 			brelse(bh);
 			continue;
 		}
@@ -2154,22 +3130,10 @@ static int write_dev_supers(struct btrfs_device *device,
 	int errors = 0;
 	u32 crc;
 	u64 bytenr;
-	int last_barrier = 0;
 
 	if (max_mirrors == 0)
 		max_mirrors = BTRFS_SUPER_MIRROR_MAX;
 
-	/* make sure only the last submit_bh does a barrier */
-	if (do_barriers) {
-		for (i = 0; i < max_mirrors; i++) {
-			bytenr = btrfs_sb_offset(i);
-			if (bytenr + BTRFS_SUPER_INFO_SIZE >=
-			    device->total_bytes)
-				break;
-			last_barrier = i;
-		}
-	}
-
 	for (i = 0; i < max_mirrors; i++) {
 		bytenr = btrfs_sb_offset(i);
 		if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes)
@@ -2178,7 +3142,10 @@ static int write_dev_supers(struct btrfs_device *device,
 		if (wait) {
 			bh = __find_get_block(device->bdev, bytenr / 4096,
 					      BTRFS_SUPER_INFO_SIZE);
-			BUG_ON(!bh);
+			if (!bh) {
+				errors++;
+				continue;
+			}
 			wait_on_buffer(bh);
 			if (!buffer_uptodate(bh))
 				errors++;
@@ -2193,7 +3160,7 @@ static int write_dev_supers(struct btrfs_device *device,
 			btrfs_set_super_bytenr(sb, bytenr);
 
 			crc = ~(u32)0;
-			crc = btrfs_csum_data(NULL, (char *)sb +
+			crc = btrfs_csum_data((char *)sb +
 					      BTRFS_CSUM_SIZE, crc,
 					      BTRFS_SUPER_INFO_SIZE -
 					      BTRFS_CSUM_SIZE);
@@ -2205,6 +3172,13 @@ static int write_dev_supers(struct btrfs_device *device,
 			 */
 			bh = __getblk(device->bdev, bytenr / 4096,
 				      BTRFS_SUPER_INFO_SIZE);
+			if (!bh) {
+				printk(KERN_ERR "BTRFS: couldn't get super "
+				       "buffer head for bytenr %Lu\n", bytenr);
+				errors++;
+				continue;
+			}
+
 			memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
 
 			/* one reference for submit_bh */
@@ -2213,31 +3187,228 @@ static int write_dev_supers(struct btrfs_device *device,
 			set_buffer_uptodate(bh);
 			lock_buffer(bh);
 			bh->b_end_io = btrfs_end_buffer_write_sync;
+			bh->b_private = device;
 		}
 
-		if (i == last_barrier && do_barriers && device->barriers) {
-			ret = submit_bh(WRITE_BARRIER, bh);
-			if (ret == -EOPNOTSUPP) {
-				printk("btrfs: disabling barriers on dev %s\n",
-				       device->name);
-				set_buffer_uptodate(bh);
-				device->barriers = 0;
-				/* one reference for submit_bh */
-				get_bh(bh);
-				lock_buffer(bh);
-				ret = submit_bh(WRITE_SYNC, bh);
-			}
-		} else {
-			ret = submit_bh(WRITE_SYNC, bh);
-		}
-
+		/*
+		 * we fua the first super.  The others we allow
+		 * to go down lazy.
+		 */
+		if (i == 0)
+			ret = btrfsic_submit_bh(WRITE_FUA, bh);
+		else
+			ret = btrfsic_submit_bh(WRITE_SYNC, bh);
 		if (ret)
 			errors++;
 	}
 	return errors < i ? 0 : -1;
 }
 
-int write_all_supers(struct btrfs_root *root, int max_mirrors)
+/*
+ * endio for the write_dev_flush, this will wake anyone waiting
+ * for the barrier when it is done
+ */
+static void btrfs_end_empty_barrier(struct bio *bio, int err)
+{
+	if (err) {
+		if (err == -EOPNOTSUPP)
+			set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
+		clear_bit(BIO_UPTODATE, &bio->bi_flags);
+	}
+	if (bio->bi_private)
+		complete(bio->bi_private);
+	bio_put(bio);
+}
+
+/*
+ * trigger flushes for one the devices.  If you pass wait == 0, the flushes are
+ * sent down.  With wait == 1, it waits for the previous flush.
+ *
+ * any device where the flush fails with eopnotsupp are flagged as not-barrier
+ * capable
+ */
+static int write_dev_flush(struct btrfs_device *device, int wait)
+{
+	struct bio *bio;
+	int ret = 0;
+
+	if (device->nobarriers)
+		return 0;
+
+	if (wait) {
+		bio = device->flush_bio;
+		if (!bio)
+			return 0;
+
+		wait_for_completion(&device->flush_wait);
+
+		if (bio_flagged(bio, BIO_EOPNOTSUPP)) {
+			printk_in_rcu("BTRFS: disabling barriers on dev %s\n",
+				      rcu_str_deref(device->name));
+			device->nobarriers = 1;
+		} else if (!bio_flagged(bio, BIO_UPTODATE)) {
+			ret = -EIO;
+			btrfs_dev_stat_inc_and_print(device,
+				BTRFS_DEV_STAT_FLUSH_ERRS);
+		}
+
+		/* drop the reference from the wait == 0 run */
+		bio_put(bio);
+		device->flush_bio = NULL;
+
+		return ret;
+	}
+
+	/*
+	 * one reference for us, and we leave it for the
+	 * caller
+	 */
+	device->flush_bio = NULL;
+	bio = btrfs_io_bio_alloc(GFP_NOFS, 0);
+	if (!bio)
+		return -ENOMEM;
+
+	bio->bi_end_io = btrfs_end_empty_barrier;
+	bio->bi_bdev = device->bdev;
+	init_completion(&device->flush_wait);
+	bio->bi_private = &device->flush_wait;
+	device->flush_bio = bio;
+
+	bio_get(bio);
+	btrfsic_submit_bio(WRITE_FLUSH, bio);
+
+	return 0;
+}
+
+/*
+ * send an empty flush down to each device in parallel,
+ * then wait for them
+ */
+static int barrier_all_devices(struct btrfs_fs_info *info)
+{
+	struct list_head *head;
+	struct btrfs_device *dev;
+	int errors_send = 0;
+	int errors_wait = 0;
+	int ret;
+
+	/* send down all the barriers */
+	head = &info->fs_devices->devices;
+	list_for_each_entry_rcu(dev, head, dev_list) {
+		if (dev->missing)
+			continue;
+		if (!dev->bdev) {
+			errors_send++;
+			continue;
+		}
+		if (!dev->in_fs_metadata || !dev->writeable)
+			continue;
+
+		ret = write_dev_flush(dev, 0);
+		if (ret)
+			errors_send++;
+	}
+
+	/* wait for all the barriers */
+	list_for_each_entry_rcu(dev, head, dev_list) {
+		if (dev->missing)
+			continue;
+		if (!dev->bdev) {
+			errors_wait++;
+			continue;
+		}
+		if (!dev->in_fs_metadata || !dev->writeable)
+			continue;
+
+		ret = write_dev_flush(dev, 1);
+		if (ret)
+			errors_wait++;
+	}
+	if (errors_send > info->num_tolerated_disk_barrier_failures ||
+	    errors_wait > info->num_tolerated_disk_barrier_failures)
+		return -EIO;
+	return 0;
+}
+
+int btrfs_calc_num_tolerated_disk_barrier_failures(
+	struct btrfs_fs_info *fs_info)
+{
+	struct btrfs_ioctl_space_info space;
+	struct btrfs_space_info *sinfo;
+	u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
+		       BTRFS_BLOCK_GROUP_SYSTEM,
+		       BTRFS_BLOCK_GROUP_METADATA,
+		       BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
+	int num_types = 4;
+	int i;
+	int c;
+	int num_tolerated_disk_barrier_failures =
+		(int)fs_info->fs_devices->num_devices;
+
+	for (i = 0; i < num_types; i++) {
+		struct btrfs_space_info *tmp;
+
+		sinfo = NULL;
+		rcu_read_lock();
+		list_for_each_entry_rcu(tmp, &fs_info->space_info, list) {
+			if (tmp->flags == types[i]) {
+				sinfo = tmp;
+				break;
+			}
+		}
+		rcu_read_unlock();
+
+		if (!sinfo)
+			continue;
+
+		down_read(&sinfo->groups_sem);
+		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
+			if (!list_empty(&sinfo->block_groups[c])) {
+				u64 flags;
+
+				btrfs_get_block_group_info(
+					&sinfo->block_groups[c], &space);
+				if (space.total_bytes == 0 ||
+				    space.used_bytes == 0)
+					continue;
+				flags = space.flags;
+				/*
+				 * return
+				 * 0: if dup, single or RAID0 is configured for
+				 *    any of metadata, system or data, else
+				 * 1: if RAID5 is configured, or if RAID1 or
+				 *    RAID10 is configured and only two mirrors
+				 *    are used, else
+				 * 2: if RAID6 is configured, else
+				 * num_mirrors - 1: if RAID1 or RAID10 is
+				 *                  configured and more than
+				 *                  2 mirrors are used.
+				 */
+				if (num_tolerated_disk_barrier_failures > 0 &&
+				    ((flags & (BTRFS_BLOCK_GROUP_DUP |
+					       BTRFS_BLOCK_GROUP_RAID0)) ||
+				     ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK)
+				      == 0)))
+					num_tolerated_disk_barrier_failures = 0;
+				else if (num_tolerated_disk_barrier_failures > 1) {
+					if (flags & (BTRFS_BLOCK_GROUP_RAID1 |
+					    BTRFS_BLOCK_GROUP_RAID5 |
+					    BTRFS_BLOCK_GROUP_RAID10)) {
+						num_tolerated_disk_barrier_failures = 1;
+					} else if (flags &
+						   BTRFS_BLOCK_GROUP_RAID6) {
+						num_tolerated_disk_barrier_failures = 2;
+					}
+				}
+			}
+		}
+		up_read(&sinfo->groups_sem);
+	}
+
+	return num_tolerated_disk_barrier_failures;
+}
+
+static int write_all_supers(struct btrfs_root *root, int max_mirrors)
 {
 	struct list_head *head;
 	struct btrfs_device *dev;
@@ -2249,15 +3420,28 @@ int write_all_supers(struct btrfs_root *root, int max_mirrors)
 	int total_errors = 0;
 	u64 flags;
 
-	max_errors = btrfs_super_num_devices(&root->fs_info->super_copy) - 1;
 	do_barriers = !btrfs_test_opt(root, NOBARRIER);
+	backup_super_roots(root->fs_info);
 
-	sb = &root->fs_info->super_for_commit;
+	sb = root->fs_info->super_for_commit;
 	dev_item = &sb->dev_item;
 
 	mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
 	head = &root->fs_info->fs_devices->devices;
-	list_for_each_entry(dev, head, dev_list) {
+	max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1;
+
+	if (do_barriers) {
+		ret = barrier_all_devices(root->fs_info);
+		if (ret) {
+			mutex_unlock(
+				&root->fs_info->fs_devices->device_list_mutex);
+			btrfs_error(root->fs_info, ret,
+				    "errors while submitting device barriers.");
+			return ret;
+		}
+	}
+
+	list_for_each_entry_rcu(dev, head, dev_list) {
 		if (!dev->bdev) {
 			total_errors++;
 			continue;
@@ -2284,13 +3468,18 @@ int write_all_supers(struct btrfs_root *root, int max_mirrors)
 			total_errors++;
 	}
 	if (total_errors > max_errors) {
-		printk(KERN_ERR "btrfs: %d errors while writing supers\n",
+		btrfs_err(root->fs_info, "%d errors while writing supers",
 		       total_errors);
-		BUG();
+		mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
+
+		/* FUA is masked off if unsupported and can't be the reason */
+		btrfs_error(root->fs_info, -EIO,
+			    "%d errors while writing supers", total_errors);
+		return -EIO;
 	}
 
 	total_errors = 0;
-	list_for_each_entry(dev, head, dev_list) {
+	list_for_each_entry_rcu(dev, head, dev_list) {
 		if (!dev->bdev)
 			continue;
 		if (!dev->in_fs_metadata || !dev->writeable)
@@ -2302,9 +3491,9 @@ int write_all_supers(struct btrfs_root *root, int max_mirrors)
 	}
 	mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
 	if (total_errors > max_errors) {
-		printk(KERN_ERR "btrfs: %d errors while writing supers\n",
-		       total_errors);
-		BUG();
+		btrfs_error(root->fs_info, -EIO,
+			    "%d errors while writing supers", total_errors);
+		return -EIO;
 	}
 	return 0;
 }
@@ -2312,13 +3501,12 @@ int write_all_supers(struct btrfs_root *root, int max_mirrors)
 int write_ctree_super(struct btrfs_trans_handle *trans,
 		      struct btrfs_root *root, int max_mirrors)
 {
-	int ret;
-
-	ret = write_all_supers(root, max_mirrors);
-	return ret;
+	return write_all_supers(root, max_mirrors);
 }
 
-int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
+/* Drop a fs root from the radix tree and free it. */
+void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info,
+				  struct btrfs_root *root)
 {
 	spin_lock(&fs_info->fs_roots_radix_lock);
 	radix_tree_delete(&fs_info->fs_roots_radix,
@@ -2328,104 +3516,107 @@ int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
 	if (btrfs_root_refs(&root->root_item) == 0)
 		synchronize_srcu(&fs_info->subvol_srcu);
 
+	if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
+		btrfs_free_log(NULL, root);
+
+	if (root->free_ino_pinned)
+		__btrfs_remove_free_space_cache(root->free_ino_pinned);
+	if (root->free_ino_ctl)
+		__btrfs_remove_free_space_cache(root->free_ino_ctl);
 	free_fs_root(root);
-	return 0;
 }
 
 static void free_fs_root(struct btrfs_root *root)
 {
+	iput(root->cache_inode);
 	WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
-	if (root->anon_super.s_dev) {
-		down_write(&root->anon_super.s_umount);
-		kill_anon_super(&root->anon_super);
-	}
+	btrfs_free_block_rsv(root, root->orphan_block_rsv);
+	root->orphan_block_rsv = NULL;
+	if (root->anon_dev)
+		free_anon_bdev(root->anon_dev);
+	if (root->subv_writers)
+		btrfs_free_subvolume_writers(root->subv_writers);
 	free_extent_buffer(root->node);
 	free_extent_buffer(root->commit_root);
+	kfree(root->free_ino_ctl);
+	kfree(root->free_ino_pinned);
 	kfree(root->name);
-	kfree(root);
+	btrfs_put_fs_root(root);
 }
 
-static int del_fs_roots(struct btrfs_fs_info *fs_info)
+void btrfs_free_fs_root(struct btrfs_root *root)
 {
-	int ret;
-	struct btrfs_root *gang[8];
-	int i;
-
-	while (!list_empty(&fs_info->dead_roots)) {
-		gang[0] = list_entry(fs_info->dead_roots.next,
-				     struct btrfs_root, root_list);
-		list_del(&gang[0]->root_list);
-
-		if (gang[0]->in_radix) {
-			btrfs_free_fs_root(fs_info, gang[0]);
-		} else {
-			free_extent_buffer(gang[0]->node);
-			free_extent_buffer(gang[0]->commit_root);
-			kfree(gang[0]);
-		}
-	}
-
-	while (1) {
-		ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
-					     (void **)gang, 0,
-					     ARRAY_SIZE(gang));
-		if (!ret)
-			break;
-		for (i = 0; i < ret; i++)
-			btrfs_free_fs_root(fs_info, gang[i]);
-	}
-	return 0;
+	free_fs_root(root);
 }
 
 int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
 {
 	u64 root_objectid = 0;
 	struct btrfs_root *gang[8];
-	int i;
-	int ret;
+	int i = 0;
+	int err = 0;
+	unsigned int ret = 0;
+	int index;
 
 	while (1) {
+		index = srcu_read_lock(&fs_info->subvol_srcu);
 		ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
 					     (void **)gang, root_objectid,
 					     ARRAY_SIZE(gang));
-		if (!ret)
+		if (!ret) {
+			srcu_read_unlock(&fs_info->subvol_srcu, index);
 			break;
-
+		}
 		root_objectid = gang[ret - 1]->root_key.objectid + 1;
+
+		for (i = 0; i < ret; i++) {
+			/* Avoid to grab roots in dead_roots */
+			if (btrfs_root_refs(&gang[i]->root_item) == 0) {
+				gang[i] = NULL;
+				continue;
+			}
+			/* grab all the search result for later use */
+			gang[i] = btrfs_grab_fs_root(gang[i]);
+		}
+		srcu_read_unlock(&fs_info->subvol_srcu, index);
+
 		for (i = 0; i < ret; i++) {
+			if (!gang[i])
+				continue;
 			root_objectid = gang[i]->root_key.objectid;
-			btrfs_orphan_cleanup(gang[i]);
+			err = btrfs_orphan_cleanup(gang[i]);
+			if (err)
+				break;
+			btrfs_put_fs_root(gang[i]);
 		}
 		root_objectid++;
 	}
-	return 0;
+
+	/* release the uncleaned roots due to error */
+	for (; i < ret; i++) {
+		if (gang[i])
+			btrfs_put_fs_root(gang[i]);
+	}
+	return err;
 }
 
 int btrfs_commit_super(struct btrfs_root *root)
 {
 	struct btrfs_trans_handle *trans;
-	int ret;
 
 	mutex_lock(&root->fs_info->cleaner_mutex);
 	btrfs_run_delayed_iputs(root);
-	btrfs_clean_old_snapshots(root);
 	mutex_unlock(&root->fs_info->cleaner_mutex);
+	wake_up_process(root->fs_info->cleaner_kthread);
 
 	/* wait until ongoing cleanup work done */
 	down_write(&root->fs_info->cleanup_work_sem);
 	up_write(&root->fs_info->cleanup_work_sem);
 
-	trans = btrfs_start_transaction(root, 1);
-	ret = btrfs_commit_transaction(trans, root);
-	BUG_ON(ret);
-	/* run commit again to drop the original snapshot */
-	trans = btrfs_start_transaction(root, 1);
-	btrfs_commit_transaction(trans, root);
-	ret = btrfs_write_and_wait_transaction(NULL, root);
-	BUG_ON(ret);
-
-	ret = write_ctree_super(NULL, root, 0);
-	return ret;
+	trans = btrfs_join_transaction(root);
+	if (IS_ERR(trans))
+		return PTR_ERR(trans);
+	return btrfs_commit_transaction(trans, root);
 }
 
 int close_ctree(struct btrfs_root *root)
@@ -2436,186 +3627,555 @@ int close_ctree(struct btrfs_root *root)
 	fs_info->closing = 1;
 	smp_mb();
 
-	kthread_stop(root->fs_info->transaction_kthread);
-	kthread_stop(root->fs_info->cleaner_kthread);
+	/* wait for the uuid_scan task to finish */
+	down(&fs_info->uuid_tree_rescan_sem);
+	/* avoid complains from lockdep et al., set sem back to initial state */
+	up(&fs_info->uuid_tree_rescan_sem);
+
+	/* pause restriper - we want to resume on mount */
+	btrfs_pause_balance(fs_info);
+
+	btrfs_dev_replace_suspend_for_unmount(fs_info);
+
+	btrfs_scrub_cancel(fs_info);
+
+	/* wait for any defraggers to finish */
+	wait_event(fs_info->transaction_wait,
+		   (atomic_read(&fs_info->defrag_running) == 0));
+
+	/* clear out the rbtree of defraggable inodes */
+	btrfs_cleanup_defrag_inodes(fs_info);
+
+	cancel_work_sync(&fs_info->async_reclaim_work);
 
 	if (!(fs_info->sb->s_flags & MS_RDONLY)) {
-		ret =  btrfs_commit_super(root);
+		ret = btrfs_commit_super(root);
 		if (ret)
-			printk(KERN_ERR "btrfs: commit super ret %d\n", ret);
+			btrfs_err(root->fs_info, "commit super ret %d", ret);
 	}
 
+	if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
+		btrfs_error_commit_super(root);
+
+	kthread_stop(fs_info->transaction_kthread);
+	kthread_stop(fs_info->cleaner_kthread);
+
 	fs_info->closing = 2;
 	smp_mb();
 
-	if (fs_info->delalloc_bytes) {
-		printk(KERN_INFO "btrfs: at unmount delalloc count %llu\n",
-		       (unsigned long long)fs_info->delalloc_bytes);
-	}
-	if (fs_info->total_ref_cache_size) {
-		printk(KERN_INFO "btrfs: at umount reference cache size %llu\n",
-		       (unsigned long long)fs_info->total_ref_cache_size);
+	btrfs_free_qgroup_config(root->fs_info);
+
+	if (percpu_counter_sum(&fs_info->delalloc_bytes)) {
+		btrfs_info(root->fs_info, "at unmount delalloc count %lld",
+		       percpu_counter_sum(&fs_info->delalloc_bytes));
 	}
 
-	free_extent_buffer(fs_info->extent_root->node);
-	free_extent_buffer(fs_info->extent_root->commit_root);
-	free_extent_buffer(fs_info->tree_root->node);
-	free_extent_buffer(fs_info->tree_root->commit_root);
-	free_extent_buffer(root->fs_info->chunk_root->node);
-	free_extent_buffer(root->fs_info->chunk_root->commit_root);
-	free_extent_buffer(root->fs_info->dev_root->node);
-	free_extent_buffer(root->fs_info->dev_root->commit_root);
-	free_extent_buffer(root->fs_info->csum_root->node);
-	free_extent_buffer(root->fs_info->csum_root->commit_root);
+	btrfs_sysfs_remove_one(fs_info);
+
+	btrfs_free_fs_roots(fs_info);
 
-	btrfs_free_block_groups(root->fs_info);
+	btrfs_put_block_group_cache(fs_info);
 
-	del_fs_roots(fs_info);
+	btrfs_free_block_groups(fs_info);
+
+	/*
+	 * we must make sure there is not any read request to
+	 * submit after we stopping all workers.
+	 */
+	invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
+	btrfs_stop_all_workers(fs_info);
+
+	free_root_pointers(fs_info, 1);
 
 	iput(fs_info->btree_inode);
 
-	btrfs_stop_workers(&fs_info->generic_worker);
-	btrfs_stop_workers(&fs_info->fixup_workers);
-	btrfs_stop_workers(&fs_info->delalloc_workers);
-	btrfs_stop_workers(&fs_info->workers);
-	btrfs_stop_workers(&fs_info->endio_workers);
-	btrfs_stop_workers(&fs_info->endio_meta_workers);
-	btrfs_stop_workers(&fs_info->endio_meta_write_workers);
-	btrfs_stop_workers(&fs_info->endio_write_workers);
-	btrfs_stop_workers(&fs_info->submit_workers);
-	btrfs_stop_workers(&fs_info->enospc_workers);
+#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
+	if (btrfs_test_opt(root, CHECK_INTEGRITY))
+		btrfsic_unmount(root, fs_info->fs_devices);
+#endif
 
 	btrfs_close_devices(fs_info->fs_devices);
 	btrfs_mapping_tree_free(&fs_info->mapping_tree);
 
+	percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
+	percpu_counter_destroy(&fs_info->delalloc_bytes);
+	percpu_counter_destroy(&fs_info->bio_counter);
 	bdi_destroy(&fs_info->bdi);
 	cleanup_srcu_struct(&fs_info->subvol_srcu);
 
-	kfree(fs_info->extent_root);
-	kfree(fs_info->tree_root);
-	kfree(fs_info->chunk_root);
-	kfree(fs_info->dev_root);
-	kfree(fs_info->csum_root);
+	btrfs_free_stripe_hash_table(fs_info);
+
+	btrfs_free_block_rsv(root, root->orphan_block_rsv);
+	root->orphan_block_rsv = NULL;
+
 	return 0;
 }
 
-int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid)
+int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid,
+			  int atomic)
 {
 	int ret;
-	struct inode *btree_inode = buf->first_page->mapping->host;
+	struct inode *btree_inode = buf->pages[0]->mapping->host;
 
-	ret = extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf,
-				     NULL);
+	ret = extent_buffer_uptodate(buf);
 	if (!ret)
 		return ret;
 
 	ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
-				    parent_transid);
+				    parent_transid, atomic);
+	if (ret == -EAGAIN)
+		return ret;
 	return !ret;
 }
 
 int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
 {
-	struct inode *btree_inode = buf->first_page->mapping->host;
-	return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree,
-					  buf);
+	return set_extent_buffer_uptodate(buf);
 }
 
 void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
 {
-	struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
+	struct btrfs_root *root;
 	u64 transid = btrfs_header_generation(buf);
-	struct inode *btree_inode = root->fs_info->btree_inode;
 	int was_dirty;
 
+#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
+	/*
+	 * This is a fast path so only do this check if we have sanity tests
+	 * enabled.  Normal people shouldn't be marking dummy buffers as dirty
+	 * outside of the sanity tests.
+	 */
+	if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags)))
+		return;
+#endif
+	root = BTRFS_I(buf->pages[0]->mapping->host)->root;
 	btrfs_assert_tree_locked(buf);
-	if (transid != root->fs_info->generation) {
-		printk(KERN_CRIT "btrfs transid mismatch buffer %llu, "
+	if (transid != root->fs_info->generation)
+		WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, "
 		       "found %llu running %llu\n",
-			(unsigned long long)buf->start,
-			(unsigned long long)transid,
-			(unsigned long long)root->fs_info->generation);
-		WARN_ON(1);
-	}
-	was_dirty = set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree,
-					    buf);
-	if (!was_dirty) {
-		spin_lock(&root->fs_info->delalloc_lock);
-		root->fs_info->dirty_metadata_bytes += buf->len;
-		spin_unlock(&root->fs_info->delalloc_lock);
+			buf->start, transid, root->fs_info->generation);
+	was_dirty = set_extent_buffer_dirty(buf);
+	if (!was_dirty)
+		__percpu_counter_add(&root->fs_info->dirty_metadata_bytes,
+				     buf->len,
+				     root->fs_info->dirty_metadata_batch);
+#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
+	if (btrfs_header_level(buf) == 0 && check_leaf(root, buf)) {
+		btrfs_print_leaf(root, buf);
+		ASSERT(0);
 	}
+#endif
 }
 
-void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
+static void __btrfs_btree_balance_dirty(struct btrfs_root *root,
+					int flush_delayed)
 {
 	/*
 	 * looks as though older kernels can get into trouble with
 	 * this code, they end up stuck in balance_dirty_pages forever
 	 */
-	u64 num_dirty;
-	unsigned long thresh = 32 * 1024 * 1024;
+	int ret;
 
 	if (current->flags & PF_MEMALLOC)
 		return;
 
-	num_dirty = root->fs_info->dirty_metadata_bytes;
+	if (flush_delayed)
+		btrfs_balance_delayed_items(root);
 
-	if (num_dirty > thresh) {
-		balance_dirty_pages_ratelimited_nr(
-				   root->fs_info->btree_inode->i_mapping, 1);
+	ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes,
+				     BTRFS_DIRTY_METADATA_THRESH);
+	if (ret > 0) {
+		balance_dirty_pages_ratelimited(
+				   root->fs_info->btree_inode->i_mapping);
 	}
 	return;
 }
 
+void btrfs_btree_balance_dirty(struct btrfs_root *root)
+{
+	__btrfs_btree_balance_dirty(root, 1);
+}
+
+void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root)
+{
+	__btrfs_btree_balance_dirty(root, 0);
+}
+
 int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
 {
-	struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
-	int ret;
-	ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
-	if (ret == 0)
-		set_bit(EXTENT_BUFFER_UPTODATE, &buf->bflags);
+	struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
+	return btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
+}
+
+static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
+			      int read_only)
+{
+	/*
+	 * Placeholder for checks
+	 */
+	return 0;
+}
+
+static void btrfs_error_commit_super(struct btrfs_root *root)
+{
+	mutex_lock(&root->fs_info->cleaner_mutex);
+	btrfs_run_delayed_iputs(root);
+	mutex_unlock(&root->fs_info->cleaner_mutex);
+
+	down_write(&root->fs_info->cleanup_work_sem);
+	up_write(&root->fs_info->cleanup_work_sem);
+
+	/* cleanup FS via transaction */
+	btrfs_cleanup_transaction(root);
+}
+
+static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t,
+					     struct btrfs_root *root)
+{
+	struct btrfs_inode *btrfs_inode;
+	struct list_head splice;
+
+	INIT_LIST_HEAD(&splice);
+
+	mutex_lock(&root->fs_info->ordered_operations_mutex);
+	spin_lock(&root->fs_info->ordered_root_lock);
+
+	list_splice_init(&t->ordered_operations, &splice);
+	while (!list_empty(&splice)) {
+		btrfs_inode = list_entry(splice.next, struct btrfs_inode,
+					 ordered_operations);
+
+		list_del_init(&btrfs_inode->ordered_operations);
+		spin_unlock(&root->fs_info->ordered_root_lock);
+
+		btrfs_invalidate_inodes(btrfs_inode->root);
+
+		spin_lock(&root->fs_info->ordered_root_lock);
+	}
+
+	spin_unlock(&root->fs_info->ordered_root_lock);
+	mutex_unlock(&root->fs_info->ordered_operations_mutex);
+}
+
+static void btrfs_destroy_ordered_extents(struct btrfs_root *root)
+{
+	struct btrfs_ordered_extent *ordered;
+
+	spin_lock(&root->ordered_extent_lock);
+	/*
+	 * This will just short circuit the ordered completion stuff which will
+	 * make sure the ordered extent gets properly cleaned up.
+	 */
+	list_for_each_entry(ordered, &root->ordered_extents,
+			    root_extent_list)
+		set_bit(BTRFS_ORDERED_IOERR, &ordered->flags);
+	spin_unlock(&root->ordered_extent_lock);
+}
+
+static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info)
+{
+	struct btrfs_root *root;
+	struct list_head splice;
+
+	INIT_LIST_HEAD(&splice);
+
+	spin_lock(&fs_info->ordered_root_lock);
+	list_splice_init(&fs_info->ordered_roots, &splice);
+	while (!list_empty(&splice)) {
+		root = list_first_entry(&splice, struct btrfs_root,
+					ordered_root);
+		list_move_tail(&root->ordered_root,
+			       &fs_info->ordered_roots);
+
+		spin_unlock(&fs_info->ordered_root_lock);
+		btrfs_destroy_ordered_extents(root);
+
+		cond_resched();
+		spin_lock(&fs_info->ordered_root_lock);
+	}
+	spin_unlock(&fs_info->ordered_root_lock);
+}
+
+static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
+				      struct btrfs_root *root)
+{
+	struct rb_node *node;
+	struct btrfs_delayed_ref_root *delayed_refs;
+	struct btrfs_delayed_ref_node *ref;
+	int ret = 0;
+
+	delayed_refs = &trans->delayed_refs;
+
+	spin_lock(&delayed_refs->lock);
+	if (atomic_read(&delayed_refs->num_entries) == 0) {
+		spin_unlock(&delayed_refs->lock);
+		btrfs_info(root->fs_info, "delayed_refs has NO entry");
+		return ret;
+	}
+
+	while ((node = rb_first(&delayed_refs->href_root)) != NULL) {
+		struct btrfs_delayed_ref_head *head;
+		bool pin_bytes = false;
+
+		head = rb_entry(node, struct btrfs_delayed_ref_head,
+				href_node);
+		if (!mutex_trylock(&head->mutex)) {
+			atomic_inc(&head->node.refs);
+			spin_unlock(&delayed_refs->lock);
+
+			mutex_lock(&head->mutex);
+			mutex_unlock(&head->mutex);
+			btrfs_put_delayed_ref(&head->node);
+			spin_lock(&delayed_refs->lock);
+			continue;
+		}
+		spin_lock(&head->lock);
+		while ((node = rb_first(&head->ref_root)) != NULL) {
+			ref = rb_entry(node, struct btrfs_delayed_ref_node,
+				       rb_node);
+			ref->in_tree = 0;
+			rb_erase(&ref->rb_node, &head->ref_root);
+			atomic_dec(&delayed_refs->num_entries);
+			btrfs_put_delayed_ref(ref);
+		}
+		if (head->must_insert_reserved)
+			pin_bytes = true;
+		btrfs_free_delayed_extent_op(head->extent_op);
+		delayed_refs->num_heads--;
+		if (head->processing == 0)
+			delayed_refs->num_heads_ready--;
+		atomic_dec(&delayed_refs->num_entries);
+		head->node.in_tree = 0;
+		rb_erase(&head->href_node, &delayed_refs->href_root);
+		spin_unlock(&head->lock);
+		spin_unlock(&delayed_refs->lock);
+		mutex_unlock(&head->mutex);
+
+		if (pin_bytes)
+			btrfs_pin_extent(root, head->node.bytenr,
+					 head->node.num_bytes, 1);
+		btrfs_put_delayed_ref(&head->node);
+		cond_resched();
+		spin_lock(&delayed_refs->lock);
+	}
+
+	spin_unlock(&delayed_refs->lock);
+
 	return ret;
 }
 
-int btree_lock_page_hook(struct page *page)
+static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root)
+{
+	struct btrfs_inode *btrfs_inode;
+	struct list_head splice;
+
+	INIT_LIST_HEAD(&splice);
+
+	spin_lock(&root->delalloc_lock);
+	list_splice_init(&root->delalloc_inodes, &splice);
+
+	while (!list_empty(&splice)) {
+		btrfs_inode = list_first_entry(&splice, struct btrfs_inode,
+					       delalloc_inodes);
+
+		list_del_init(&btrfs_inode->delalloc_inodes);
+		clear_bit(BTRFS_INODE_IN_DELALLOC_LIST,
+			  &btrfs_inode->runtime_flags);
+		spin_unlock(&root->delalloc_lock);
+
+		btrfs_invalidate_inodes(btrfs_inode->root);
+
+		spin_lock(&root->delalloc_lock);
+	}
+
+	spin_unlock(&root->delalloc_lock);
+}
+
+static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info)
+{
+	struct btrfs_root *root;
+	struct list_head splice;
+
+	INIT_LIST_HEAD(&splice);
+
+	spin_lock(&fs_info->delalloc_root_lock);
+	list_splice_init(&fs_info->delalloc_roots, &splice);
+	while (!list_empty(&splice)) {
+		root = list_first_entry(&splice, struct btrfs_root,
+					 delalloc_root);
+		list_del_init(&root->delalloc_root);
+		root = btrfs_grab_fs_root(root);
+		BUG_ON(!root);
+		spin_unlock(&fs_info->delalloc_root_lock);
+
+		btrfs_destroy_delalloc_inodes(root);
+		btrfs_put_fs_root(root);
+
+		spin_lock(&fs_info->delalloc_root_lock);
+	}
+	spin_unlock(&fs_info->delalloc_root_lock);
+}
+
+static int btrfs_destroy_marked_extents(struct btrfs_root *root,
+					struct extent_io_tree *dirty_pages,
+					int mark)
 {
-	struct inode *inode = page->mapping->host;
-	struct btrfs_root *root = BTRFS_I(inode)->root;
-	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
+	int ret;
 	struct extent_buffer *eb;
-	unsigned long len;
-	u64 bytenr = page_offset(page);
+	u64 start = 0;
+	u64 end;
 
-	if (page->private == EXTENT_PAGE_PRIVATE)
-		goto out;
+	while (1) {
+		ret = find_first_extent_bit(dirty_pages, start, &start, &end,
+					    mark, NULL);
+		if (ret)
+			break;
 
-	len = page->private >> 2;
-	eb = find_extent_buffer(io_tree, bytenr, len, GFP_NOFS);
-	if (!eb)
-		goto out;
+		clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS);
+		while (start <= end) {
+			eb = btrfs_find_tree_block(root, start,
+						   root->leafsize);
+			start += root->leafsize;
+			if (!eb)
+				continue;
+			wait_on_extent_buffer_writeback(eb);
+
+			if (test_and_clear_bit(EXTENT_BUFFER_DIRTY,
+					       &eb->bflags))
+				clear_extent_buffer_dirty(eb);
+			free_extent_buffer_stale(eb);
+		}
+	}
 
-	btrfs_tree_lock(eb);
-	btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
+	return ret;
+}
 
-	if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) {
-		spin_lock(&root->fs_info->delalloc_lock);
-		if (root->fs_info->dirty_metadata_bytes >= eb->len)
-			root->fs_info->dirty_metadata_bytes -= eb->len;
+static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
+				       struct extent_io_tree *pinned_extents)
+{
+	struct extent_io_tree *unpin;
+	u64 start;
+	u64 end;
+	int ret;
+	bool loop = true;
+
+	unpin = pinned_extents;
+again:
+	while (1) {
+		ret = find_first_extent_bit(unpin, 0, &start, &end,
+					    EXTENT_DIRTY, NULL);
+		if (ret)
+			break;
+
+		/* opt_discard */
+		if (btrfs_test_opt(root, DISCARD))
+			ret = btrfs_error_discard_extent(root, start,
+							 end + 1 - start,
+							 NULL);
+
+		clear_extent_dirty(unpin, start, end, GFP_NOFS);
+		btrfs_error_unpin_extent_range(root, start, end);
+		cond_resched();
+	}
+
+	if (loop) {
+		if (unpin == &root->fs_info->freed_extents[0])
+			unpin = &root->fs_info->freed_extents[1];
 		else
-			WARN_ON(1);
-		spin_unlock(&root->fs_info->delalloc_lock);
+			unpin = &root->fs_info->freed_extents[0];
+		loop = false;
+		goto again;
 	}
 
-	btrfs_tree_unlock(eb);
-	free_extent_buffer(eb);
-out:
-	lock_page(page);
+	return 0;
+}
+
+void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans,
+				   struct btrfs_root *root)
+{
+	btrfs_destroy_ordered_operations(cur_trans, root);
+
+	btrfs_destroy_delayed_refs(cur_trans, root);
+
+	cur_trans->state = TRANS_STATE_COMMIT_START;
+	wake_up(&root->fs_info->transaction_blocked_wait);
+
+	cur_trans->state = TRANS_STATE_UNBLOCKED;
+	wake_up(&root->fs_info->transaction_wait);
+
+	btrfs_destroy_delayed_inodes(root);
+	btrfs_assert_delayed_root_empty(root);
+
+	btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages,
+				     EXTENT_DIRTY);
+	btrfs_destroy_pinned_extent(root,
+				    root->fs_info->pinned_extents);
+
+	cur_trans->state =TRANS_STATE_COMPLETED;
+	wake_up(&cur_trans->commit_wait);
+
+	/*
+	memset(cur_trans, 0, sizeof(*cur_trans));
+	kmem_cache_free(btrfs_transaction_cachep, cur_trans);
+	*/
+}
+
+static int btrfs_cleanup_transaction(struct btrfs_root *root)
+{
+	struct btrfs_transaction *t;
+
+	mutex_lock(&root->fs_info->transaction_kthread_mutex);
+
+	spin_lock(&root->fs_info->trans_lock);
+	while (!list_empty(&root->fs_info->trans_list)) {
+		t = list_first_entry(&root->fs_info->trans_list,
+				     struct btrfs_transaction, list);
+		if (t->state >= TRANS_STATE_COMMIT_START) {
+			atomic_inc(&t->use_count);
+			spin_unlock(&root->fs_info->trans_lock);
+			btrfs_wait_for_commit(root, t->transid);
+			btrfs_put_transaction(t);
+			spin_lock(&root->fs_info->trans_lock);
+			continue;
+		}
+		if (t == root->fs_info->running_transaction) {
+			t->state = TRANS_STATE_COMMIT_DOING;
+			spin_unlock(&root->fs_info->trans_lock);
+			/*
+			 * We wait for 0 num_writers since we don't hold a trans
+			 * handle open currently for this transaction.
+			 */
+			wait_event(t->writer_wait,
+				   atomic_read(&t->num_writers) == 0);
+		} else {
+			spin_unlock(&root->fs_info->trans_lock);
+		}
+		btrfs_cleanup_one_transaction(t, root);
+
+		spin_lock(&root->fs_info->trans_lock);
+		if (t == root->fs_info->running_transaction)
+			root->fs_info->running_transaction = NULL;
+		list_del_init(&t->list);
+		spin_unlock(&root->fs_info->trans_lock);
+
+		btrfs_put_transaction(t);
+		trace_btrfs_transaction_commit(root);
+		spin_lock(&root->fs_info->trans_lock);
+	}
+	spin_unlock(&root->fs_info->trans_lock);
+	btrfs_destroy_all_ordered_extents(root->fs_info);
+	btrfs_destroy_delayed_inodes(root);
+	btrfs_assert_delayed_root_empty(root);
+	btrfs_destroy_pinned_extent(root, root->fs_info->pinned_extents);
+	btrfs_destroy_all_delalloc_inodes(root->fs_info);
+	mutex_unlock(&root->fs_info->transaction_kthread_mutex);
+
 	return 0;
 }
 
 static struct extent_io_ops btree_extent_io_ops = {
-	.write_cache_pages_lock_hook = btree_lock_page_hook,
 	.readpage_end_io_hook = btree_readpage_end_io_hook,
+	.readpage_io_failed_hook = btree_io_failed_hook,
 	.submit_bio_hook = btree_submit_bio_hook,
 	/* note we're sharing with inode.c for the merge bio hook */
 	.merge_bio_hook = btrfs_merge_bio_hook,
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index c958ecbc191..23ce3ceba0a 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -25,6 +25,13 @@
 #define BTRFS_SUPER_MIRROR_MAX	 3
 #define BTRFS_SUPER_MIRROR_SHIFT 12
 
+enum {
+	BTRFS_WQ_ENDIO_DATA = 0,
+	BTRFS_WQ_ENDIO_METADATA = 1,
+	BTRFS_WQ_ENDIO_FREE_SPACE = 2,
+	BTRFS_WQ_ENDIO_RAID56 = 3,
+};
+
 static inline u64 btrfs_sb_offset(int mirror)
 {
 	u64 start = 16 * 1024;
@@ -40,13 +47,15 @@ struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
 				      u32 blocksize, u64 parent_transid);
 int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
 			 u64 parent_transid);
+int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, u32 blocksize,
+			 int mirror_num, struct extent_buffer **eb);
 struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
 						   u64 bytenr, u32 blocksize);
-int clean_tree_block(struct btrfs_trans_handle *trans,
-		     struct btrfs_root *root, struct extent_buffer *buf);
-struct btrfs_root *open_ctree(struct super_block *sb,
-			      struct btrfs_fs_devices *fs_devices,
-			      char *options);
+void clean_tree_block(struct btrfs_trans_handle *trans,
+		      struct btrfs_root *root, struct extent_buffer *buf);
+int open_ctree(struct super_block *sb,
+	       struct btrfs_fs_devices *fs_devices,
+	       char *options);
 int close_ctree(struct btrfs_root *root);
 int write_ctree_super(struct btrfs_trans_handle *trans,
 		      struct btrfs_root *root, int max_mirrors);
@@ -54,61 +63,94 @@ struct buffer_head *btrfs_read_dev_super(struct block_device *bdev);
 int btrfs_commit_super(struct btrfs_root *root);
 struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
 					    u64 bytenr, u32 blocksize);
-struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
-					u64 root_objectid);
-struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
-				      struct btrfs_key *location,
-				      const char *name, int namelen);
-struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root,
-					       struct btrfs_key *location);
-struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
-					      struct btrfs_key *location);
+struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root,
+				      struct btrfs_key *location);
+int btrfs_init_fs_root(struct btrfs_root *root);
+int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info,
+			 struct btrfs_root *root);
+void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info);
+
+struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info,
+				     struct btrfs_key *key,
+				     bool check_ref);
+static inline struct btrfs_root *
+btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
+			   struct btrfs_key *location)
+{
+	return btrfs_get_fs_root(fs_info, location, true);
+}
+
 int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info);
-int btrfs_insert_dev_radix(struct btrfs_root *root,
-			   struct block_device *bdev,
-			   u64 device_id,
-			   u64 block_start,
-			   u64 num_blocks);
-void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr);
-int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root);
+void btrfs_btree_balance_dirty(struct btrfs_root *root);
+void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root);
+void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info,
+				 struct btrfs_root *root);
+void btrfs_free_fs_root(struct btrfs_root *root);
+
+#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
+struct btrfs_root *btrfs_alloc_dummy_root(void);
+#endif
+
+/*
+ * This function is used to grab the root, and avoid it is freed when we
+ * access it. But it doesn't ensure that the tree is not dropped.
+ *
+ * If you want to ensure the whole tree is safe, you should use
+ * 	fs_info->subvol_srcu
+ */
+static inline struct btrfs_root *btrfs_grab_fs_root(struct btrfs_root *root)
+{
+	if (atomic_inc_not_zero(&root->refs))
+		return root;
+	return NULL;
+}
+
+static inline void btrfs_put_fs_root(struct btrfs_root *root)
+{
+	if (atomic_dec_and_test(&root->refs))
+		kfree(root);
+}
+
 void btrfs_mark_buffer_dirty(struct extent_buffer *buf);
-void btrfs_mark_buffer_dirty_nonblocking(struct extent_buffer *buf);
-int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid);
+int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid,
+			  int atomic);
 int btrfs_set_buffer_uptodate(struct extent_buffer *buf);
-int wait_on_tree_block_writeback(struct btrfs_root *root,
-				 struct extent_buffer *buf);
 int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid);
-u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len);
+u32 btrfs_csum_data(char *data, u32 seed, size_t len);
 void btrfs_csum_final(u32 crc, char *result);
-int btrfs_open_device(struct btrfs_device *dev);
-int btrfs_verify_block_csum(struct btrfs_root *root,
-			    struct extent_buffer *buf);
 int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
 			int metadata);
 int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
 			int rw, struct bio *bio, int mirror_num,
-			unsigned long bio_flags,
+			unsigned long bio_flags, u64 bio_offset,
 			extent_submit_bio_hook_t *submit_bio_start,
 			extent_submit_bio_hook_t *submit_bio_done);
-
-int btrfs_congested_async(struct btrfs_fs_info *info, int iodone);
 unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info);
 int btrfs_write_tree_block(struct extent_buffer *buf);
 int btrfs_wait_tree_block_writeback(struct extent_buffer *buf);
-int btrfs_free_log_root_tree(struct btrfs_trans_handle *trans,
-			     struct btrfs_fs_info *fs_info);
 int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
 			     struct btrfs_fs_info *fs_info);
 int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
 		       struct btrfs_root *root);
-int btree_lock_page_hook(struct page *page);
-
+void btrfs_cleanup_one_transaction(struct btrfs_transaction *trans,
+				  struct btrfs_root *root);
+struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans,
+				     struct btrfs_fs_info *fs_info,
+				     u64 objectid);
+int btree_lock_page_hook(struct page *page, void *data,
+				void (*flush_fn)(void *));
+int btrfs_calc_num_tolerated_disk_barrier_failures(
+	struct btrfs_fs_info *fs_info);
 
 #ifdef CONFIG_DEBUG_LOCK_ALLOC
-void btrfs_set_buffer_lockdep_class(struct extent_buffer *eb, int level);
+void btrfs_init_lockdep(void);
+void btrfs_set_buffer_lockdep_class(u64 objectid,
+			            struct extent_buffer *eb, int level);
 #else
-static inline void btrfs_set_buffer_lockdep_class(struct extent_buffer *eb,
-						 int level)
+static inline void btrfs_init_lockdep(void)
+{ }
+static inline void btrfs_set_buffer_lockdep_class(u64 objectid,
+					struct extent_buffer *eb, int level)
 {
 }
 #endif
diff --git a/fs/btrfs/export.c b/fs/btrfs/export.c
index 951ef09b82f..41422a3de8e 100644
--- a/fs/btrfs/export.c
+++ b/fs/btrfs/export.c
@@ -5,7 +5,6 @@
 #include "btrfs_inode.h"
 #include "print-tree.h"
 #include "export.h"
-#include "compat.h"
 
 #define BTRFS_FID_SIZE_NON_CONNECTABLE (offsetof(struct btrfs_fid, \
 						 parent_objectid) / 4)
@@ -13,38 +12,35 @@
 					     parent_root_objectid) / 4)
 #define BTRFS_FID_SIZE_CONNECTABLE_ROOT (sizeof(struct btrfs_fid) / 4)
 
-static int btrfs_encode_fh(struct dentry *dentry, u32 *fh, int *max_len,
-			   int connectable)
+static int btrfs_encode_fh(struct inode *inode, u32 *fh, int *max_len,
+			   struct inode *parent)
 {
 	struct btrfs_fid *fid = (struct btrfs_fid *)fh;
-	struct inode *inode = dentry->d_inode;
 	int len = *max_len;
 	int type;
 
-	if ((len < BTRFS_FID_SIZE_NON_CONNECTABLE) ||
-	    (connectable && len < BTRFS_FID_SIZE_CONNECTABLE))
-		return 255;
+	if (parent && (len < BTRFS_FID_SIZE_CONNECTABLE)) {
+		*max_len = BTRFS_FID_SIZE_CONNECTABLE;
+		return FILEID_INVALID;
+	} else if (len < BTRFS_FID_SIZE_NON_CONNECTABLE) {
+		*max_len = BTRFS_FID_SIZE_NON_CONNECTABLE;
+		return FILEID_INVALID;
+	}
 
 	len  = BTRFS_FID_SIZE_NON_CONNECTABLE;
 	type = FILEID_BTRFS_WITHOUT_PARENT;
 
-	fid->objectid = inode->i_ino;
+	fid->objectid = btrfs_ino(inode);
 	fid->root_objectid = BTRFS_I(inode)->root->objectid;
 	fid->gen = inode->i_generation;
 
-	if (connectable && !S_ISDIR(inode->i_mode)) {
-		struct inode *parent;
+	if (parent) {
 		u64 parent_root_id;
 
-		spin_lock(&dentry->d_lock);
-
-		parent = dentry->d_parent->d_inode;
 		fid->parent_objectid = BTRFS_I(parent)->location.objectid;
 		fid->parent_gen = parent->i_generation;
 		parent_root_id = BTRFS_I(parent)->root->objectid;
 
-		spin_unlock(&dentry->d_lock);
-
 		if (parent_root_id != fid->root_objectid) {
 			fid->parent_root_objectid = parent_root_id;
 			len = BTRFS_FID_SIZE_CONNECTABLE_ROOT;
@@ -63,9 +59,8 @@ static struct dentry *btrfs_get_dentry(struct super_block *sb, u64 objectid,
 				       u64 root_objectid, u32 generation,
 				       int check_generation)
 {
-	struct btrfs_fs_info *fs_info = btrfs_sb(sb)->fs_info;
+	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 	struct btrfs_root *root;
-	struct dentry *dentry;
 	struct inode *inode;
 	struct btrfs_key key;
 	int index;
@@ -86,11 +81,6 @@ static struct dentry *btrfs_get_dentry(struct super_block *sb, u64 objectid,
 		goto fail;
 	}
 
-	if (btrfs_root_refs(&root->root_item) == 0) {
-		err = -ENOENT;
-		goto fail;
-	}
-
 	key.objectid = objectid;
 	btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
 	key.offset = 0;
@@ -108,10 +98,7 @@ static struct dentry *btrfs_get_dentry(struct super_block *sb, u64 objectid,
 		return ERR_PTR(-ESTALE);
 	}
 
-	dentry = d_obtain_alias(inode);
-	if (!IS_ERR(dentry))
-		dentry->d_op = &btrfs_dentry_operations;
-	return dentry;
+	return d_obtain_alias(inode);
 fail:
 	srcu_read_unlock(&fs_info->subvol_srcu, index);
 	return ERR_PTR(err);
@@ -166,7 +153,6 @@ static struct dentry *btrfs_fh_to_dentry(struct super_block *sb, struct fid *fh,
 static struct dentry *btrfs_get_parent(struct dentry *child)
 {
 	struct inode *dir = child->d_inode;
-	static struct dentry *dentry;
 	struct btrfs_root *root = BTRFS_I(dir)->root;
 	struct btrfs_path *path;
 	struct extent_buffer *leaf;
@@ -176,14 +162,16 @@ static struct dentry *btrfs_get_parent(struct dentry *child)
 	int ret;
 
 	path = btrfs_alloc_path();
+	if (!path)
+		return ERR_PTR(-ENOMEM);
 
-	if (dir->i_ino == BTRFS_FIRST_FREE_OBJECTID) {
+	if (btrfs_ino(dir) == BTRFS_FIRST_FREE_OBJECTID) {
 		key.objectid = root->root_key.objectid;
 		key.type = BTRFS_ROOT_BACKREF_KEY;
 		key.offset = (u64)-1;
 		root = root->fs_info->tree_root;
 	} else {
-		key.objectid = dir->i_ino;
+		key.objectid = btrfs_ino(dir);
 		key.type = BTRFS_INODE_REF_KEY;
 		key.offset = (u64)-1;
 	}
@@ -192,7 +180,7 @@ static struct dentry *btrfs_get_parent(struct dentry *child)
 	if (ret < 0)
 		goto fail;
 
-	BUG_ON(ret == 0);
+	BUG_ON(ret == 0); /* Key with offset of -1 found */
 	if (path->slots[0] == 0) {
 		ret = -ENOENT;
 		goto fail;
@@ -223,18 +211,94 @@ static struct dentry *btrfs_get_parent(struct dentry *child)
 
 	key.type = BTRFS_INODE_ITEM_KEY;
 	key.offset = 0;
-	dentry = d_obtain_alias(btrfs_iget(root->fs_info->sb, &key, root, NULL));
-	if (!IS_ERR(dentry))
-		dentry->d_op = &btrfs_dentry_operations;
-	return dentry;
+	return d_obtain_alias(btrfs_iget(root->fs_info->sb, &key, root, NULL));
 fail:
 	btrfs_free_path(path);
 	return ERR_PTR(ret);
 }
 
+static int btrfs_get_name(struct dentry *parent, char *name,
+			  struct dentry *child)
+{
+	struct inode *inode = child->d_inode;
+	struct inode *dir = parent->d_inode;
+	struct btrfs_path *path;
+	struct btrfs_root *root = BTRFS_I(dir)->root;
+	struct btrfs_inode_ref *iref;
+	struct btrfs_root_ref *rref;
+	struct extent_buffer *leaf;
+	unsigned long name_ptr;
+	struct btrfs_key key;
+	int name_len;
+	int ret;
+	u64 ino;
+
+	if (!dir || !inode)
+		return -EINVAL;
+
+	if (!S_ISDIR(dir->i_mode))
+		return -EINVAL;
+
+	ino = btrfs_ino(inode);
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+	path->leave_spinning = 1;
+
+	if (ino == BTRFS_FIRST_FREE_OBJECTID) {
+		key.objectid = BTRFS_I(inode)->root->root_key.objectid;
+		key.type = BTRFS_ROOT_BACKREF_KEY;
+		key.offset = (u64)-1;
+		root = root->fs_info->tree_root;
+	} else {
+		key.objectid = ino;
+		key.offset = btrfs_ino(dir);
+		key.type = BTRFS_INODE_REF_KEY;
+	}
+
+	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+	if (ret < 0) {
+		btrfs_free_path(path);
+		return ret;
+	} else if (ret > 0) {
+		if (ino == BTRFS_FIRST_FREE_OBJECTID) {
+			path->slots[0]--;
+		} else {
+			btrfs_free_path(path);
+			return -ENOENT;
+		}
+	}
+	leaf = path->nodes[0];
+
+	if (ino == BTRFS_FIRST_FREE_OBJECTID) {
+		rref = btrfs_item_ptr(leaf, path->slots[0],
+				     struct btrfs_root_ref);
+		name_ptr = (unsigned long)(rref + 1);
+		name_len = btrfs_root_ref_name_len(leaf, rref);
+	} else {
+		iref = btrfs_item_ptr(leaf, path->slots[0],
+				      struct btrfs_inode_ref);
+		name_ptr = (unsigned long)(iref + 1);
+		name_len = btrfs_inode_ref_name_len(leaf, iref);
+	}
+
+	read_extent_buffer(leaf, name, name_ptr, name_len);
+	btrfs_free_path(path);
+
+	/*
+	 * have to add the null termination to make sure that reconnect_path
+	 * gets the right len for strlen
+	 */
+	name[name_len] = '\0';
+
+	return 0;
+}
+
 const struct export_operations btrfs_export_ops = {
 	.encode_fh	= btrfs_encode_fh,
 	.fh_to_dentry	= btrfs_fh_to_dentry,
 	.fh_to_parent	= btrfs_fh_to_parent,
 	.get_parent	= btrfs_get_parent,
+	.get_name	= btrfs_get_name,
 };
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 9e23ffea7f5..813537f362f 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -23,28 +23,66 @@
 #include <linux/rcupdate.h>
 #include <linux/kthread.h>
 #include <linux/slab.h>
-#include "compat.h"
+#include <linux/ratelimit.h>
+#include <linux/percpu_counter.h>
 #include "hash.h"
-#include "ctree.h"
+#include "tree-log.h"
 #include "disk-io.h"
 #include "print-tree.h"
-#include "transaction.h"
 #include "volumes.h"
+#include "raid56.h"
 #include "locking.h"
 #include "free-space-cache.h"
+#include "math.h"
+#include "sysfs.h"
+#include "qgroup.h"
 
-static int update_block_group(struct btrfs_trans_handle *trans,
-			      struct btrfs_root *root,
-			      u64 bytenr, u64 num_bytes, int alloc,
-			      int mark_free);
-static int update_reserved_extents(struct btrfs_block_group_cache *cache,
-				   u64 num_bytes, int reserve);
+#undef SCRAMBLE_DELAYED_REFS
+
+/*
+ * control flags for do_chunk_alloc's force field
+ * CHUNK_ALLOC_NO_FORCE means to only allocate a chunk
+ * if we really need one.
+ *
+ * CHUNK_ALLOC_LIMITED means to only try and allocate one
+ * if we have very few chunks already allocated.  This is
+ * used as part of the clustering code to help make sure
+ * we have a good pool of storage to cluster in, without
+ * filling the FS with empty chunks
+ *
+ * CHUNK_ALLOC_FORCE means it must try to allocate one
+ *
+ */
+enum {
+	CHUNK_ALLOC_NO_FORCE = 0,
+	CHUNK_ALLOC_LIMITED = 1,
+	CHUNK_ALLOC_FORCE = 2,
+};
+
+/*
+ * Control how reservations are dealt with.
+ *
+ * RESERVE_FREE - freeing a reservation.
+ * RESERVE_ALLOC - allocating space and we need to update bytes_may_use for
+ *   ENOSPC accounting
+ * RESERVE_ALLOC_NO_ACCOUNT - allocating space and we should not update
+ *   bytes_may_use as the ENOSPC accounting is done elsewhere
+ */
+enum {
+	RESERVE_FREE = 0,
+	RESERVE_ALLOC = 1,
+	RESERVE_ALLOC_NO_ACCOUNT = 2,
+};
+
+static int update_block_group(struct btrfs_root *root,
+			      u64 bytenr, u64 num_bytes, int alloc);
 static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
 				struct btrfs_root *root,
 				u64 bytenr, u64 num_bytes, u64 parent,
 				u64 root_objectid, u64 owner_objectid,
 				u64 owner_offset, int refs_to_drop,
-				struct btrfs_delayed_extent_op *extra_op);
+				struct btrfs_delayed_extent_op *extra_op,
+				int no_quota);
 static void __run_delayed_extent_op(struct btrfs_delayed_extent_op *extent_op,
 				    struct extent_buffer *leaf,
 				    struct btrfs_extent_item *ei);
@@ -57,26 +95,29 @@ static int alloc_reserved_tree_block(struct btrfs_trans_handle *trans,
 				     struct btrfs_root *root,
 				     u64 parent, u64 root_objectid,
 				     u64 flags, struct btrfs_disk_key *key,
-				     int level, struct btrfs_key *ins);
+				     int level, struct btrfs_key *ins,
+				     int no_quota);
 static int do_chunk_alloc(struct btrfs_trans_handle *trans,
-			  struct btrfs_root *extent_root, u64 alloc_bytes,
-			  u64 flags, int force);
-static int pin_down_bytes(struct btrfs_trans_handle *trans,
-			  struct btrfs_root *root,
-			  struct btrfs_path *path,
-			  u64 bytenr, u64 num_bytes,
-			  int is_data, int reserved,
-			  struct extent_buffer **must_clean);
+			  struct btrfs_root *extent_root, u64 flags,
+			  int force);
 static int find_next_key(struct btrfs_path *path, int level,
 			 struct btrfs_key *key);
 static void dump_space_info(struct btrfs_space_info *info, u64 bytes,
 			    int dump_block_groups);
+static int btrfs_update_reserved_bytes(struct btrfs_block_group_cache *cache,
+				       u64 num_bytes, int reserve,
+				       int delalloc);
+static int block_rsv_use_bytes(struct btrfs_block_rsv *block_rsv,
+			       u64 num_bytes);
+int btrfs_pin_extent(struct btrfs_root *root,
+		     u64 bytenr, u64 num_bytes, int reserved);
 
 static noinline int
 block_group_cache_done(struct btrfs_block_group_cache *cache)
 {
 	smp_mb();
-	return cache->cached == BTRFS_CACHE_FINISHED;
+	return cache->cached == BTRFS_CACHE_FINISHED ||
+		cache->cached == BTRFS_CACHE_ERROR;
 }
 
 static int block_group_bits(struct btrfs_block_group_cache *cache, u64 bits)
@@ -84,15 +125,19 @@ static int block_group_bits(struct btrfs_block_group_cache *cache, u64 bits)
 	return (cache->flags & bits) == bits;
 }
 
-void btrfs_get_block_group(struct btrfs_block_group_cache *cache)
+static void btrfs_get_block_group(struct btrfs_block_group_cache *cache)
 {
 	atomic_inc(&cache->count);
 }
 
 void btrfs_put_block_group(struct btrfs_block_group_cache *cache)
 {
-	if (atomic_dec_and_test(&cache->count))
+	if (atomic_dec_and_test(&cache->count)) {
+		WARN_ON(cache->pinned > 0);
+		WARN_ON(cache->reserved > 0);
+		kfree(cache->free_space_ctl);
 		kfree(cache);
+	}
 }
 
 /*
@@ -126,6 +171,10 @@ static int btrfs_add_block_group_cache(struct btrfs_fs_info *info,
 	rb_link_node(&block_group->cache_node, parent, p);
 	rb_insert_color(&block_group->cache_node,
 			&info->block_group_cache_tree);
+
+	if (info->first_logical_byte > block_group->key.objectid)
+		info->first_logical_byte = block_group->key.objectid;
+
 	spin_unlock(&info->block_group_cache_lock);
 
 	return 0;
@@ -167,8 +216,11 @@ block_group_cache_tree_search(struct btrfs_fs_info *info, u64 bytenr,
 			break;
 		}
 	}
-	if (ret)
+	if (ret) {
 		btrfs_get_block_group(ret);
+		if (bytenr == 0 && info->first_logical_byte > ret->key.objectid)
+			info->first_logical_byte = ret->key.objectid;
+	}
 	spin_unlock(&info->block_group_cache_lock);
 
 	return ret;
@@ -212,7 +264,8 @@ static int exclude_super_stripes(struct btrfs_root *root,
 		cache->bytes_super += stripe_len;
 		ret = add_excluded_extent(root, cache->key.objectid,
 					  stripe_len);
-		BUG_ON(ret);
+		if (ret)
+			return ret;
 	}
 
 	for (i = 0; i < BTRFS_SUPER_MIRROR_MAX; i++) {
@@ -220,13 +273,35 @@ static int exclude_super_stripes(struct btrfs_root *root,
 		ret = btrfs_rmap_block(&root->fs_info->mapping_tree,
 				       cache->key.objectid, bytenr,
 				       0, &logical, &nr, &stripe_len);
-		BUG_ON(ret);
+		if (ret)
+			return ret;
 
 		while (nr--) {
-			cache->bytes_super += stripe_len;
-			ret = add_excluded_extent(root, logical[nr],
-						  stripe_len);
-			BUG_ON(ret);
+			u64 start, len;
+
+			if (logical[nr] > cache->key.objectid +
+			    cache->key.offset)
+				continue;
+
+			if (logical[nr] + stripe_len <= cache->key.objectid)
+				continue;
+
+			start = logical[nr];
+			if (start < cache->key.objectid) {
+				start = cache->key.objectid;
+				len = (logical[nr] + stripe_len) - start;
+			} else {
+				len = min_t(u64, stripe_len,
+					    cache->key.objectid +
+					    cache->key.offset - start);
+			}
+
+			cache->bytes_super += len;
+			ret = add_excluded_extent(root, start, len);
+			if (ret) {
+				kfree(logical);
+				return ret;
+			}
 		}
 
 		kfree(logical);
@@ -245,6 +320,12 @@ get_caching_control(struct btrfs_block_group_cache *cache)
 		return NULL;
 	}
 
+	/* We're loading it the fast way, so we don't have a caching_ctl. */
+	if (!cache->caching_ctl) {
+		spin_unlock(&cache->lock);
+		return NULL;
+	}
+
 	ctl = cache->caching_ctl;
 	atomic_inc(&ctl->count);
 	spin_unlock(&cache->lock);
@@ -271,7 +352,8 @@ static u64 add_new_free_space(struct btrfs_block_group_cache *block_group,
 	while (start < end) {
 		ret = find_first_extent_bit(info->pinned_extents, start,
 					    &extent_start, &extent_end,
-					    EXTENT_DIRTY | EXTENT_UPTODATE);
+					    EXTENT_DIRTY | EXTENT_UPTODATE,
+					    NULL);
 		if (ret)
 			break;
 
@@ -282,7 +364,7 @@ static u64 add_new_free_space(struct btrfs_block_group_cache *block_group,
 			total_added += size;
 			ret = btrfs_add_free_space(block_group, start,
 						   size);
-			BUG_ON(ret);
+			BUG_ON(ret); /* -ENOMEM or logic error */
 			start = extent_end + 1;
 		} else {
 			break;
@@ -293,34 +375,34 @@ static u64 add_new_free_space(struct btrfs_block_group_cache *block_group,
 		size = end - start;
 		total_added += size;
 		ret = btrfs_add_free_space(block_group, start, size);
-		BUG_ON(ret);
+		BUG_ON(ret); /* -ENOMEM or logic error */
 	}
 
 	return total_added;
 }
 
-static int caching_kthread(void *data)
+static noinline void caching_thread(struct btrfs_work *work)
 {
-	struct btrfs_block_group_cache *block_group = data;
-	struct btrfs_fs_info *fs_info = block_group->fs_info;
-	struct btrfs_caching_control *caching_ctl = block_group->caching_ctl;
-	struct btrfs_root *extent_root = fs_info->extent_root;
+	struct btrfs_block_group_cache *block_group;
+	struct btrfs_fs_info *fs_info;
+	struct btrfs_caching_control *caching_ctl;
+	struct btrfs_root *extent_root;
 	struct btrfs_path *path;
 	struct extent_buffer *leaf;
 	struct btrfs_key key;
 	u64 total_found = 0;
 	u64 last = 0;
 	u32 nritems;
-	int ret = 0;
+	int ret = -ENOMEM;
+
+	caching_ctl = container_of(work, struct btrfs_caching_control, work);
+	block_group = caching_ctl->block_group;
+	fs_info = block_group->fs_info;
+	extent_root = fs_info->extent_root;
 
 	path = btrfs_alloc_path();
 	if (!path)
-		return -ENOMEM;
-
-	exclude_super_stripes(extent_root, block_group);
-	spin_lock(&block_group->space_info->lock);
-	block_group->space_info->bytes_super += block_group->bytes_super;
-	spin_unlock(&block_group->space_info->lock);
+		goto out;
 
 	last = max_t(u64, block_group->key.objectid, BTRFS_SUPER_INFO_OFFSET);
 
@@ -332,7 +414,7 @@ static int caching_kthread(void *data)
 	 */
 	path->skip_locking = 1;
 	path->search_commit_root = 1;
-	path->reada = 2;
+	path->reada = 1;
 
 	key.objectid = last;
 	key.offset = 0;
@@ -340,8 +422,9 @@ static int caching_kthread(void *data)
 again:
 	mutex_lock(&caching_ctl->mutex);
 	/* need to make sure the commit_root doesn't disappear */
-	down_read(&fs_info->extent_commit_sem);
+	down_read(&fs_info->commit_root_sem);
 
+next:
 	ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0);
 	if (ret < 0)
 		goto err;
@@ -350,8 +433,7 @@ again:
 	nritems = btrfs_header_nritems(leaf);
 
 	while (1) {
-		smp_mb();
-		if (fs_info->closing > 1) {
+		if (btrfs_fs_closing(fs_info) > 1) {
 			last = (u64)-1;
 			break;
 		}
@@ -363,15 +445,34 @@ again:
 			if (ret)
 				break;
 
-			caching_ctl->progress = last;
-			btrfs_release_path(extent_root, path);
-			up_read(&fs_info->extent_commit_sem);
-			mutex_unlock(&caching_ctl->mutex);
-			if (btrfs_transaction_in_commit(fs_info))
-				schedule_timeout(1);
-			else
+			if (need_resched() ||
+			    rwsem_is_contended(&fs_info->commit_root_sem)) {
+				caching_ctl->progress = last;
+				btrfs_release_path(path);
+				up_read(&fs_info->commit_root_sem);
+				mutex_unlock(&caching_ctl->mutex);
 				cond_resched();
-			goto again;
+				goto again;
+			}
+
+			ret = btrfs_next_leaf(extent_root, path);
+			if (ret < 0)
+				goto err;
+			if (ret)
+				break;
+			leaf = path->nodes[0];
+			nritems = btrfs_header_nritems(leaf);
+			continue;
+		}
+
+		if (key.objectid < last) {
+			key.objectid = last;
+			key.offset = 0;
+			key.type = BTRFS_EXTENT_ITEM_KEY;
+
+			caching_ctl->progress = last;
+			btrfs_release_path(path);
+			goto next;
 		}
 
 		if (key.objectid < block_group->key.objectid) {
@@ -383,11 +484,16 @@ again:
 		    block_group->key.offset)
 			break;
 
-		if (key.type == BTRFS_EXTENT_ITEM_KEY) {
+		if (key.type == BTRFS_EXTENT_ITEM_KEY ||
+		    key.type == BTRFS_METADATA_ITEM_KEY) {
 			total_found += add_new_free_space(block_group,
 							  fs_info, last,
 							  key.objectid);
-			last = key.objectid + key.offset;
+			if (key.type == BTRFS_METADATA_ITEM_KEY)
+				last = key.objectid +
+					fs_info->tree_root->leafsize;
+			else
+				last = key.objectid + key.offset;
 
 			if (total_found > (1024 * 1024 * 2)) {
 				total_found = 0;
@@ -410,66 +516,134 @@ again:
 
 err:
 	btrfs_free_path(path);
-	up_read(&fs_info->extent_commit_sem);
+	up_read(&fs_info->commit_root_sem);
 
 	free_excluded_extents(extent_root, block_group);
 
 	mutex_unlock(&caching_ctl->mutex);
+out:
+	if (ret) {
+		spin_lock(&block_group->lock);
+		block_group->caching_ctl = NULL;
+		block_group->cached = BTRFS_CACHE_ERROR;
+		spin_unlock(&block_group->lock);
+	}
 	wake_up(&caching_ctl->wait);
 
 	put_caching_control(caching_ctl);
-	atomic_dec(&block_group->space_info->caching_threads);
 	btrfs_put_block_group(block_group);
-
-	return 0;
 }
 
-static int cache_block_group(struct btrfs_block_group_cache *cache)
+static int cache_block_group(struct btrfs_block_group_cache *cache,
+			     int load_cache_only)
 {
+	DEFINE_WAIT(wait);
 	struct btrfs_fs_info *fs_info = cache->fs_info;
 	struct btrfs_caching_control *caching_ctl;
-	struct task_struct *tsk;
 	int ret = 0;
 
-	smp_mb();
-	if (cache->cached != BTRFS_CACHE_NO)
-		return 0;
-
-	caching_ctl = kzalloc(sizeof(*caching_ctl), GFP_KERNEL);
-	BUG_ON(!caching_ctl);
+	caching_ctl = kzalloc(sizeof(*caching_ctl), GFP_NOFS);
+	if (!caching_ctl)
+		return -ENOMEM;
 
 	INIT_LIST_HEAD(&caching_ctl->list);
 	mutex_init(&caching_ctl->mutex);
 	init_waitqueue_head(&caching_ctl->wait);
 	caching_ctl->block_group = cache;
 	caching_ctl->progress = cache->key.objectid;
-	/* one for caching kthread, one for caching block group list */
-	atomic_set(&caching_ctl->count, 2);
+	atomic_set(&caching_ctl->count, 1);
+	btrfs_init_work(&caching_ctl->work, caching_thread, NULL, NULL);
 
 	spin_lock(&cache->lock);
+	/*
+	 * This should be a rare occasion, but this could happen I think in the
+	 * case where one thread starts to load the space cache info, and then
+	 * some other thread starts a transaction commit which tries to do an
+	 * allocation while the other thread is still loading the space cache
+	 * info.  The previous loop should have kept us from choosing this block
+	 * group, but if we've moved to the state where we will wait on caching
+	 * block groups we need to first check if we're doing a fast load here,
+	 * so we can wait for it to finish, otherwise we could end up allocating
+	 * from a block group who's cache gets evicted for one reason or
+	 * another.
+	 */
+	while (cache->cached == BTRFS_CACHE_FAST) {
+		struct btrfs_caching_control *ctl;
+
+		ctl = cache->caching_ctl;
+		atomic_inc(&ctl->count);
+		prepare_to_wait(&ctl->wait, &wait, TASK_UNINTERRUPTIBLE);
+		spin_unlock(&cache->lock);
+
+		schedule();
+
+		finish_wait(&ctl->wait, &wait);
+		put_caching_control(ctl);
+		spin_lock(&cache->lock);
+	}
+
 	if (cache->cached != BTRFS_CACHE_NO) {
 		spin_unlock(&cache->lock);
 		kfree(caching_ctl);
 		return 0;
 	}
+	WARN_ON(cache->caching_ctl);
 	cache->caching_ctl = caching_ctl;
-	cache->cached = BTRFS_CACHE_STARTED;
+	cache->cached = BTRFS_CACHE_FAST;
 	spin_unlock(&cache->lock);
 
-	down_write(&fs_info->extent_commit_sem);
+	if (fs_info->mount_opt & BTRFS_MOUNT_SPACE_CACHE) {
+		ret = load_free_space_cache(fs_info, cache);
+
+		spin_lock(&cache->lock);
+		if (ret == 1) {
+			cache->caching_ctl = NULL;
+			cache->cached = BTRFS_CACHE_FINISHED;
+			cache->last_byte_to_unpin = (u64)-1;
+		} else {
+			if (load_cache_only) {
+				cache->caching_ctl = NULL;
+				cache->cached = BTRFS_CACHE_NO;
+			} else {
+				cache->cached = BTRFS_CACHE_STARTED;
+			}
+		}
+		spin_unlock(&cache->lock);
+		wake_up(&caching_ctl->wait);
+		if (ret == 1) {
+			put_caching_control(caching_ctl);
+			free_excluded_extents(fs_info->extent_root, cache);
+			return 0;
+		}
+	} else {
+		/*
+		 * We are not going to do the fast caching, set cached to the
+		 * appropriate value and wakeup any waiters.
+		 */
+		spin_lock(&cache->lock);
+		if (load_cache_only) {
+			cache->caching_ctl = NULL;
+			cache->cached = BTRFS_CACHE_NO;
+		} else {
+			cache->cached = BTRFS_CACHE_STARTED;
+		}
+		spin_unlock(&cache->lock);
+		wake_up(&caching_ctl->wait);
+	}
+
+	if (load_cache_only) {
+		put_caching_control(caching_ctl);
+		return 0;
+	}
+
+	down_write(&fs_info->commit_root_sem);
+	atomic_inc(&caching_ctl->count);
 	list_add_tail(&caching_ctl->list, &fs_info->caching_block_groups);
-	up_write(&fs_info->extent_commit_sem);
+	up_write(&fs_info->commit_root_sem);
 
-	atomic_inc(&cache->space_info->caching_threads);
 	btrfs_get_block_group(cache);
 
-	tsk = kthread_run(caching_kthread, cache, "btrfs-cache-%llu\n",
-			  cache->key.objectid);
-	if (IS_ERR(tsk)) {
-		ret = PTR_ERR(tsk);
-		printk(KERN_ERR "error running thread %d\n", ret);
-		BUG();
-	}
+	btrfs_queue_work(fs_info->caching_workers, &caching_ctl->work);
 
 	return ret;
 }
@@ -507,9 +681,11 @@ static struct btrfs_space_info *__find_space_info(struct btrfs_fs_info *info,
 	struct list_head *head = &info->space_info;
 	struct btrfs_space_info *found;
 
+	flags &= BTRFS_BLOCK_GROUP_TYPE_MASK;
+
 	rcu_read_lock();
 	list_for_each_entry_rcu(found, head, list) {
-		if (found->flags == flags) {
+		if (found->flags & flags) {
 			rcu_read_unlock();
 			return found;
 		}
@@ -533,64 +709,6 @@ void btrfs_clear_space_info_full(struct btrfs_fs_info *info)
 	rcu_read_unlock();
 }
 
-static u64 div_factor(u64 num, int factor)
-{
-	if (factor == 10)
-		return num;
-	num *= factor;
-	do_div(num, 10);
-	return num;
-}
-
-u64 btrfs_find_block_group(struct btrfs_root *root,
-			   u64 search_start, u64 search_hint, int owner)
-{
-	struct btrfs_block_group_cache *cache;
-	u64 used;
-	u64 last = max(search_hint, search_start);
-	u64 group_start = 0;
-	int full_search = 0;
-	int factor = 9;
-	int wrapped = 0;
-again:
-	while (1) {
-		cache = btrfs_lookup_first_block_group(root->fs_info, last);
-		if (!cache)
-			break;
-
-		spin_lock(&cache->lock);
-		last = cache->key.objectid + cache->key.offset;
-		used = btrfs_block_group_used(&cache->item);
-
-		if ((full_search || !cache->ro) &&
-		    block_group_bits(cache, BTRFS_BLOCK_GROUP_METADATA)) {
-			if (used + cache->pinned + cache->reserved <
-			    div_factor(cache->key.offset, factor)) {
-				group_start = cache->key.objectid;
-				spin_unlock(&cache->lock);
-				btrfs_put_block_group(cache);
-				goto found;
-			}
-		}
-		spin_unlock(&cache->lock);
-		btrfs_put_block_group(cache);
-		cond_resched();
-	}
-	if (!wrapped) {
-		last = search_start;
-		wrapped = 1;
-		goto again;
-	}
-	if (!full_search && factor < 10) {
-		last = search_start;
-		full_search = 1;
-		factor = 10;
-		goto again;
-	}
-found:
-	return group_start;
-}
-
 /* simple helper to search for an existing extent at a given offset */
 int btrfs_lookup_extent(struct btrfs_root *root, u64 start, u64 len)
 {
@@ -599,12 +717,167 @@ int btrfs_lookup_extent(struct btrfs_root *root, u64 start, u64 len)
 	struct btrfs_path *path;
 
 	path = btrfs_alloc_path();
-	BUG_ON(!path);
+	if (!path)
+		return -ENOMEM;
+
 	key.objectid = start;
 	key.offset = len;
-	btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
+	key.type = BTRFS_EXTENT_ITEM_KEY;
 	ret = btrfs_search_slot(NULL, root->fs_info->extent_root, &key, path,
 				0, 0);
+	if (ret > 0) {
+		btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
+		if (key.objectid == start &&
+		    key.type == BTRFS_METADATA_ITEM_KEY)
+			ret = 0;
+	}
+	btrfs_free_path(path);
+	return ret;
+}
+
+/*
+ * helper function to lookup reference count and flags of a tree block.
+ *
+ * the head node for delayed ref is used to store the sum of all the
+ * reference count modifications queued up in the rbtree. the head
+ * node may also store the extent flags to set. This way you can check
+ * to see what the reference count and extent flags would be if all of
+ * the delayed refs are not processed.
+ */
+int btrfs_lookup_extent_info(struct btrfs_trans_handle *trans,
+			     struct btrfs_root *root, u64 bytenr,
+			     u64 offset, int metadata, u64 *refs, u64 *flags)
+{
+	struct btrfs_delayed_ref_head *head;
+	struct btrfs_delayed_ref_root *delayed_refs;
+	struct btrfs_path *path;
+	struct btrfs_extent_item *ei;
+	struct extent_buffer *leaf;
+	struct btrfs_key key;
+	u32 item_size;
+	u64 num_refs;
+	u64 extent_flags;
+	int ret;
+
+	/*
+	 * If we don't have skinny metadata, don't bother doing anything
+	 * different
+	 */
+	if (metadata && !btrfs_fs_incompat(root->fs_info, SKINNY_METADATA)) {
+		offset = root->leafsize;
+		metadata = 0;
+	}
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	if (!trans) {
+		path->skip_locking = 1;
+		path->search_commit_root = 1;
+	}
+
+search_again:
+	key.objectid = bytenr;
+	key.offset = offset;
+	if (metadata)
+		key.type = BTRFS_METADATA_ITEM_KEY;
+	else
+		key.type = BTRFS_EXTENT_ITEM_KEY;
+
+again:
+	ret = btrfs_search_slot(trans, root->fs_info->extent_root,
+				&key, path, 0, 0);
+	if (ret < 0)
+		goto out_free;
+
+	if (ret > 0 && metadata && key.type == BTRFS_METADATA_ITEM_KEY) {
+		if (path->slots[0]) {
+			path->slots[0]--;
+			btrfs_item_key_to_cpu(path->nodes[0], &key,
+					      path->slots[0]);
+			if (key.objectid == bytenr &&
+			    key.type == BTRFS_EXTENT_ITEM_KEY &&
+			    key.offset == root->leafsize)
+				ret = 0;
+		}
+		if (ret) {
+			key.objectid = bytenr;
+			key.type = BTRFS_EXTENT_ITEM_KEY;
+			key.offset = root->leafsize;
+			btrfs_release_path(path);
+			goto again;
+		}
+	}
+
+	if (ret == 0) {
+		leaf = path->nodes[0];
+		item_size = btrfs_item_size_nr(leaf, path->slots[0]);
+		if (item_size >= sizeof(*ei)) {
+			ei = btrfs_item_ptr(leaf, path->slots[0],
+					    struct btrfs_extent_item);
+			num_refs = btrfs_extent_refs(leaf, ei);
+			extent_flags = btrfs_extent_flags(leaf, ei);
+		} else {
+#ifdef BTRFS_COMPAT_EXTENT_TREE_V0
+			struct btrfs_extent_item_v0 *ei0;
+			BUG_ON(item_size != sizeof(*ei0));
+			ei0 = btrfs_item_ptr(leaf, path->slots[0],
+					     struct btrfs_extent_item_v0);
+			num_refs = btrfs_extent_refs_v0(leaf, ei0);
+			/* FIXME: this isn't correct for data */
+			extent_flags = BTRFS_BLOCK_FLAG_FULL_BACKREF;
+#else
+			BUG();
+#endif
+		}
+		BUG_ON(num_refs == 0);
+	} else {
+		num_refs = 0;
+		extent_flags = 0;
+		ret = 0;
+	}
+
+	if (!trans)
+		goto out;
+
+	delayed_refs = &trans->transaction->delayed_refs;
+	spin_lock(&delayed_refs->lock);
+	head = btrfs_find_delayed_ref_head(trans, bytenr);
+	if (head) {
+		if (!mutex_trylock(&head->mutex)) {
+			atomic_inc(&head->node.refs);
+			spin_unlock(&delayed_refs->lock);
+
+			btrfs_release_path(path);
+
+			/*
+			 * Mutex was contended, block until it's released and try
+			 * again
+			 */
+			mutex_lock(&head->mutex);
+			mutex_unlock(&head->mutex);
+			btrfs_put_delayed_ref(&head->node);
+			goto search_again;
+		}
+		spin_lock(&head->lock);
+		if (head->extent_op && head->extent_op->update_flags)
+			extent_flags |= head->extent_op->flags_to_set;
+		else
+			BUG_ON(num_refs == 0);
+
+		num_refs += head->node.ref_mod;
+		spin_unlock(&head->lock);
+		mutex_unlock(&head->mutex);
+	}
+	spin_unlock(&delayed_refs->lock);
+out:
+	WARN_ON(num_refs == 0);
+	if (refs)
+		*refs = num_refs;
+	if (flags)
+		*flags = extent_flags;
+out_free:
 	btrfs_free_path(path);
 	return ret;
 }
@@ -746,7 +1019,7 @@ static int convert_extent_item_v0(struct btrfs_trans_handle *trans,
 				ret = btrfs_next_leaf(root, path);
 				if (ret < 0)
 					return ret;
-				BUG_ON(ret > 0);
+				BUG_ON(ret > 0); /* Corruption */
 				leaf = path->nodes[0];
 			}
 			btrfs_item_key_to_cpu(leaf, &found_key,
@@ -762,7 +1035,7 @@ static int convert_extent_item_v0(struct btrfs_trans_handle *trans,
 			break;
 		}
 	}
-	btrfs_release_path(root, path);
+	btrfs_release_path(path);
 
 	if (owner < BTRFS_FIRST_FREE_OBJECTID)
 		new_size += sizeof(*bi);
@@ -772,10 +1045,9 @@ static int convert_extent_item_v0(struct btrfs_trans_handle *trans,
 				new_size + extra_size, 1);
 	if (ret < 0)
 		return ret;
-	BUG_ON(ret);
+	BUG_ON(ret); /* Corruption */
 
-	ret = btrfs_extend_item(trans, root, path, new_size);
-	BUG_ON(ret);
+	btrfs_extend_item(root, path, new_size);
 
 	leaf = path->nodes[0];
 	item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
@@ -805,11 +1077,11 @@ static u64 hash_extent_data_ref(u64 root_objectid, u64 owner, u64 offset)
 	__le64 lenum;
 
 	lenum = cpu_to_le64(root_objectid);
-	high_crc = crc32c(high_crc, &lenum, sizeof(lenum));
+	high_crc = btrfs_crc32c(high_crc, &lenum, sizeof(lenum));
 	lenum = cpu_to_le64(owner);
-	low_crc = crc32c(low_crc, &lenum, sizeof(lenum));
+	low_crc = btrfs_crc32c(low_crc, &lenum, sizeof(lenum));
 	lenum = cpu_to_le64(offset);
-	low_crc = crc32c(low_crc, &lenum, sizeof(lenum));
+	low_crc = btrfs_crc32c(low_crc, &lenum, sizeof(lenum));
 
 	return ((u64)high_crc << 31) ^ (u64)low_crc;
 }
@@ -870,7 +1142,7 @@ again:
 			return 0;
 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
 		key.type = BTRFS_EXTENT_REF_V0_KEY;
-		btrfs_release_path(root, path);
+		btrfs_release_path(path);
 		ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
 		if (ret < 0) {
 			err = ret;
@@ -908,7 +1180,7 @@ again:
 		if (match_extent_data_ref(leaf, ref, root_objectid,
 					  owner, offset)) {
 			if (recow) {
-				btrfs_release_path(root, path);
+				btrfs_release_path(path);
 				goto again;
 			}
 			err = 0;
@@ -969,7 +1241,7 @@ static noinline int insert_extent_data_ref(struct btrfs_trans_handle *trans,
 			if (match_extent_data_ref(leaf, ref, root_objectid,
 						  owner, offset))
 				break;
-			btrfs_release_path(root, path);
+			btrfs_release_path(path);
 			key.offset++;
 			ret = btrfs_insert_empty_item(trans, root, path, &key,
 						      size);
@@ -995,14 +1267,14 @@ static noinline int insert_extent_data_ref(struct btrfs_trans_handle *trans,
 	btrfs_mark_buffer_dirty(leaf);
 	ret = 0;
 fail:
-	btrfs_release_path(root, path);
+	btrfs_release_path(path);
 	return ret;
 }
 
 static noinline int remove_extent_data_ref(struct btrfs_trans_handle *trans,
 					   struct btrfs_root *root,
 					   struct btrfs_path *path,
-					   int refs_to_drop)
+					   int refs_to_drop, int *last_ref)
 {
 	struct btrfs_key key;
 	struct btrfs_extent_data_ref *ref1 = NULL;
@@ -1038,6 +1310,7 @@ static noinline int remove_extent_data_ref(struct btrfs_trans_handle *trans,
 
 	if (num_refs == 0) {
 		ret = btrfs_del_item(trans, root, path);
+		*last_ref = 1;
 	} else {
 		if (key.type == BTRFS_EXTENT_DATA_REF_KEY)
 			btrfs_set_extent_data_ref_count(leaf, ref1, num_refs);
@@ -1121,7 +1394,7 @@ static noinline int lookup_tree_block_ref(struct btrfs_trans_handle *trans,
 		ret = -ENOENT;
 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
 	if (ret == -ENOENT && parent) {
-		btrfs_release_path(root, path);
+		btrfs_release_path(path);
 		key.type = BTRFS_EXTENT_REF_V0_KEY;
 		ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
 		if (ret > 0)
@@ -1150,7 +1423,7 @@ static noinline int insert_tree_block_ref(struct btrfs_trans_handle *trans,
 	}
 
 	ret = btrfs_insert_empty_item(trans, root, path, &key, 0);
-	btrfs_release_path(root, path);
+	btrfs_release_path(path);
 	return ret;
 }
 
@@ -1227,6 +1500,8 @@ int lookup_inline_extent_backref(struct btrfs_trans_handle *trans,
 	int want;
 	int ret;
 	int err = 0;
+	bool skinny_metadata = btrfs_fs_incompat(root->fs_info,
+						 SKINNY_METADATA);
 
 	key.objectid = bytenr;
 	key.type = BTRFS_EXTENT_ITEM_KEY;
@@ -1238,12 +1513,54 @@ int lookup_inline_extent_backref(struct btrfs_trans_handle *trans,
 		path->keep_locks = 1;
 	} else
 		extra_size = -1;
+
+	/*
+	 * Owner is our parent level, so we can just add one to get the level
+	 * for the block we are interested in.
+	 */
+	if (skinny_metadata && owner < BTRFS_FIRST_FREE_OBJECTID) {
+		key.type = BTRFS_METADATA_ITEM_KEY;
+		key.offset = owner;
+	}
+
+again:
 	ret = btrfs_search_slot(trans, root, &key, path, extra_size, 1);
 	if (ret < 0) {
 		err = ret;
 		goto out;
 	}
-	BUG_ON(ret);
+
+	/*
+	 * We may be a newly converted file system which still has the old fat
+	 * extent entries for metadata, so try and see if we have one of those.
+	 */
+	if (ret > 0 && skinny_metadata) {
+		skinny_metadata = false;
+		if (path->slots[0]) {
+			path->slots[0]--;
+			btrfs_item_key_to_cpu(path->nodes[0], &key,
+					      path->slots[0]);
+			if (key.objectid == bytenr &&
+			    key.type == BTRFS_EXTENT_ITEM_KEY &&
+			    key.offset == num_bytes)
+				ret = 0;
+		}
+		if (ret) {
+			key.objectid = bytenr;
+			key.type = BTRFS_EXTENT_ITEM_KEY;
+			key.offset = num_bytes;
+			btrfs_release_path(path);
+			goto again;
+		}
+	}
+
+	if (ret && !insert) {
+		err = -ENOENT;
+		goto out;
+	} else if (WARN_ON(ret)) {
+		err = -EIO;
+		goto out;
+	}
 
 	leaf = path->nodes[0];
 	item_size = btrfs_item_size_nr(leaf, path->slots[0]);
@@ -1271,11 +1588,9 @@ int lookup_inline_extent_backref(struct btrfs_trans_handle *trans,
 	ptr = (unsigned long)(ei + 1);
 	end = (unsigned long)ei + item_size;
 
-	if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
+	if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK && !skinny_metadata) {
 		ptr += sizeof(struct btrfs_tree_block_info);
 		BUG_ON(ptr > end);
-	} else {
-		BUG_ON(!(flags & BTRFS_EXTENT_FLAG_DATA));
 	}
 
 	err = -ENOENT;
@@ -1357,13 +1672,12 @@ out:
  * helper to add new inline back ref
  */
 static noinline_for_stack
-int setup_inline_extent_backref(struct btrfs_trans_handle *trans,
-				struct btrfs_root *root,
-				struct btrfs_path *path,
-				struct btrfs_extent_inline_ref *iref,
-				u64 parent, u64 root_objectid,
-				u64 owner, u64 offset, int refs_to_add,
-				struct btrfs_delayed_extent_op *extent_op)
+void setup_inline_extent_backref(struct btrfs_root *root,
+				 struct btrfs_path *path,
+				 struct btrfs_extent_inline_ref *iref,
+				 u64 parent, u64 root_objectid,
+				 u64 owner, u64 offset, int refs_to_add,
+				 struct btrfs_delayed_extent_op *extent_op)
 {
 	struct extent_buffer *leaf;
 	struct btrfs_extent_item *ei;
@@ -1373,7 +1687,6 @@ int setup_inline_extent_backref(struct btrfs_trans_handle *trans,
 	u64 refs;
 	int size;
 	int type;
-	int ret;
 
 	leaf = path->nodes[0];
 	ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
@@ -1382,8 +1695,7 @@ int setup_inline_extent_backref(struct btrfs_trans_handle *trans,
 	type = extent_ref_type(parent, owner);
 	size = btrfs_extent_inline_ref_size(type);
 
-	ret = btrfs_extend_item(trans, root, path, size);
-	BUG_ON(ret);
+	btrfs_extend_item(root, path, size);
 
 	ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
 	refs = btrfs_extent_refs(leaf, ei);
@@ -1418,7 +1730,6 @@ int setup_inline_extent_backref(struct btrfs_trans_handle *trans,
 		btrfs_set_extent_inline_ref_offset(leaf, iref, root_objectid);
 	}
 	btrfs_mark_buffer_dirty(leaf);
-	return 0;
 }
 
 static int lookup_extent_backref(struct btrfs_trans_handle *trans,
@@ -1436,7 +1747,7 @@ static int lookup_extent_backref(struct btrfs_trans_handle *trans,
 	if (ret != -ENOENT)
 		return ret;
 
-	btrfs_release_path(root, path);
+	btrfs_release_path(path);
 	*ref_ret = NULL;
 
 	if (owner < BTRFS_FIRST_FREE_OBJECTID) {
@@ -1453,12 +1764,12 @@ static int lookup_extent_backref(struct btrfs_trans_handle *trans,
  * helper to update/remove inline back ref
  */
 static noinline_for_stack
-int update_inline_extent_backref(struct btrfs_trans_handle *trans,
-				 struct btrfs_root *root,
-				 struct btrfs_path *path,
-				 struct btrfs_extent_inline_ref *iref,
-				 int refs_to_mod,
-				 struct btrfs_delayed_extent_op *extent_op)
+void update_inline_extent_backref(struct btrfs_root *root,
+				  struct btrfs_path *path,
+				  struct btrfs_extent_inline_ref *iref,
+				  int refs_to_mod,
+				  struct btrfs_delayed_extent_op *extent_op,
+				  int *last_ref)
 {
 	struct extent_buffer *leaf;
 	struct btrfs_extent_item *ei;
@@ -1469,7 +1780,6 @@ int update_inline_extent_backref(struct btrfs_trans_handle *trans,
 	u32 item_size;
 	int size;
 	int type;
-	int ret;
 	u64 refs;
 
 	leaf = path->nodes[0];
@@ -1503,6 +1813,7 @@ int update_inline_extent_backref(struct btrfs_trans_handle *trans,
 		else
 			btrfs_set_shared_data_ref_count(leaf, sref, refs);
 	} else {
+		*last_ref = 1;
 		size =  btrfs_extent_inline_ref_size(type);
 		item_size = btrfs_item_size_nr(leaf, path->slots[0]);
 		ptr = (unsigned long)iref;
@@ -1511,11 +1822,9 @@ int update_inline_extent_backref(struct btrfs_trans_handle *trans,
 			memmove_extent_buffer(leaf, ptr, ptr + size,
 					      end - ptr - size);
 		item_size -= size;
-		ret = btrfs_truncate_item(trans, root, path, item_size, 1);
-		BUG_ON(ret);
+		btrfs_truncate_item(root, path, item_size, 1);
 	}
 	btrfs_mark_buffer_dirty(leaf);
-	return 0;
 }
 
 static noinline_for_stack
@@ -1535,13 +1844,13 @@ int insert_inline_extent_backref(struct btrfs_trans_handle *trans,
 					   root_objectid, owner, offset, 1);
 	if (ret == 0) {
 		BUG_ON(owner < BTRFS_FIRST_FREE_OBJECTID);
-		ret = update_inline_extent_backref(trans, root, path, iref,
-						   refs_to_add, extent_op);
+		update_inline_extent_backref(root, path, iref,
+					     refs_to_add, extent_op, NULL);
 	} else if (ret == -ENOENT) {
-		ret = setup_inline_extent_backref(trans, root, path, iref,
-						  parent, root_objectid,
-						  owner, offset, refs_to_add,
-						  extent_op);
+		setup_inline_extent_backref(root, path, iref, parent,
+					    root_objectid, owner, offset,
+					    refs_to_add, extent_op);
+		ret = 0;
 	}
 	return ret;
 }
@@ -1569,77 +1878,101 @@ static int remove_extent_backref(struct btrfs_trans_handle *trans,
 				 struct btrfs_root *root,
 				 struct btrfs_path *path,
 				 struct btrfs_extent_inline_ref *iref,
-				 int refs_to_drop, int is_data)
+				 int refs_to_drop, int is_data, int *last_ref)
 {
-	int ret;
+	int ret = 0;
 
 	BUG_ON(!is_data && refs_to_drop != 1);
 	if (iref) {
-		ret = update_inline_extent_backref(trans, root, path, iref,
-						   -refs_to_drop, NULL);
+		update_inline_extent_backref(root, path, iref,
+					     -refs_to_drop, NULL, last_ref);
 	} else if (is_data) {
-		ret = remove_extent_data_ref(trans, root, path, refs_to_drop);
+		ret = remove_extent_data_ref(trans, root, path, refs_to_drop,
+					     last_ref);
 	} else {
+		*last_ref = 1;
 		ret = btrfs_del_item(trans, root, path);
 	}
 	return ret;
 }
 
-static void btrfs_issue_discard(struct block_device *bdev,
+static int btrfs_issue_discard(struct block_device *bdev,
 				u64 start, u64 len)
 {
-	blkdev_issue_discard(bdev, start >> 9, len >> 9, GFP_KERNEL,
-			     DISCARD_FL_BARRIER);
+	return blkdev_issue_discard(bdev, start >> 9, len >> 9, GFP_NOFS, 0);
 }
 
 static int btrfs_discard_extent(struct btrfs_root *root, u64 bytenr,
-				u64 num_bytes)
+				u64 num_bytes, u64 *actual_bytes)
 {
 	int ret;
-	u64 map_length = num_bytes;
-	struct btrfs_multi_bio *multi = NULL;
+	u64 discarded_bytes = 0;
+	struct btrfs_bio *bbio = NULL;
 
-	if (!btrfs_test_opt(root, DISCARD))
-		return 0;
 
 	/* Tell the block device(s) that the sectors can be discarded */
-	ret = btrfs_map_block(&root->fs_info->mapping_tree, READ,
-			      bytenr, &map_length, &multi, 0);
+	ret = btrfs_map_block(root->fs_info, REQ_DISCARD,
+			      bytenr, &num_bytes, &bbio, 0);
+	/* Error condition is -ENOMEM */
 	if (!ret) {
-		struct btrfs_bio_stripe *stripe = multi->stripes;
+		struct btrfs_bio_stripe *stripe = bbio->stripes;
 		int i;
 
-		if (map_length > num_bytes)
-			map_length = num_bytes;
 
-		for (i = 0; i < multi->num_stripes; i++, stripe++) {
-			btrfs_issue_discard(stripe->dev->bdev,
-					    stripe->physical,
-					    map_length);
+		for (i = 0; i < bbio->num_stripes; i++, stripe++) {
+			if (!stripe->dev->can_discard)
+				continue;
+
+			ret = btrfs_issue_discard(stripe->dev->bdev,
+						  stripe->physical,
+						  stripe->length);
+			if (!ret)
+				discarded_bytes += stripe->length;
+			else if (ret != -EOPNOTSUPP)
+				break; /* Logic errors or -ENOMEM, or -EIO but I don't know how that could happen JDM */
+
+			/*
+			 * Just in case we get back EOPNOTSUPP for some reason,
+			 * just ignore the return value so we don't screw up
+			 * people calling discard_extent.
+			 */
+			ret = 0;
 		}
-		kfree(multi);
+		kfree(bbio);
 	}
 
+	if (actual_bytes)
+		*actual_bytes = discarded_bytes;
+
+
+	if (ret == -EOPNOTSUPP)
+		ret = 0;
 	return ret;
 }
 
+/* Can return -ENOMEM */
 int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
 			 struct btrfs_root *root,
 			 u64 bytenr, u64 num_bytes, u64 parent,
-			 u64 root_objectid, u64 owner, u64 offset)
+			 u64 root_objectid, u64 owner, u64 offset,
+			 int no_quota)
 {
 	int ret;
+	struct btrfs_fs_info *fs_info = root->fs_info;
+
 	BUG_ON(owner < BTRFS_FIRST_FREE_OBJECTID &&
 	       root_objectid == BTRFS_TREE_LOG_OBJECTID);
 
 	if (owner < BTRFS_FIRST_FREE_OBJECTID) {
-		ret = btrfs_add_delayed_tree_ref(trans, bytenr, num_bytes,
+		ret = btrfs_add_delayed_tree_ref(fs_info, trans, bytenr,
+					num_bytes,
 					parent, root_objectid, (int)owner,
-					BTRFS_ADD_DELAYED_REF, NULL);
+					BTRFS_ADD_DELAYED_REF, NULL, no_quota);
 	} else {
-		ret = btrfs_add_delayed_data_ref(trans, bytenr, num_bytes,
+		ret = btrfs_add_delayed_data_ref(fs_info, trans, bytenr,
+					num_bytes,
 					parent, root_objectid, owner, offset,
-					BTRFS_ADD_DELAYED_REF, NULL);
+					BTRFS_ADD_DELAYED_REF, NULL, no_quota);
 	}
 	return ret;
 }
@@ -1649,55 +1982,89 @@ static int __btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
 				  u64 bytenr, u64 num_bytes,
 				  u64 parent, u64 root_objectid,
 				  u64 owner, u64 offset, int refs_to_add,
+				  int no_quota,
 				  struct btrfs_delayed_extent_op *extent_op)
 {
+	struct btrfs_fs_info *fs_info = root->fs_info;
 	struct btrfs_path *path;
 	struct extent_buffer *leaf;
 	struct btrfs_extent_item *item;
+	struct btrfs_key key;
 	u64 refs;
 	int ret;
-	int err = 0;
+	enum btrfs_qgroup_operation_type type = BTRFS_QGROUP_OPER_ADD_EXCL;
 
 	path = btrfs_alloc_path();
 	if (!path)
 		return -ENOMEM;
 
+	if (!is_fstree(root_objectid) || !root->fs_info->quota_enabled)
+		no_quota = 1;
+
 	path->reada = 1;
 	path->leave_spinning = 1;
 	/* this will setup the path even if it fails to insert the back ref */
-	ret = insert_inline_extent_backref(trans, root->fs_info->extent_root,
-					   path, bytenr, num_bytes, parent,
+	ret = insert_inline_extent_backref(trans, fs_info->extent_root, path,
+					   bytenr, num_bytes, parent,
 					   root_objectid, owner, offset,
 					   refs_to_add, extent_op);
-	if (ret == 0)
+	if ((ret < 0 && ret != -EAGAIN) || (!ret && no_quota))
 		goto out;
-
-	if (ret != -EAGAIN) {
-		err = ret;
+	/*
+	 * Ok we were able to insert an inline extent and it appears to be a new
+	 * reference, deal with the qgroup accounting.
+	 */
+	if (!ret && !no_quota) {
+		ASSERT(root->fs_info->quota_enabled);
+		leaf = path->nodes[0];
+		btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
+		item = btrfs_item_ptr(leaf, path->slots[0],
+				      struct btrfs_extent_item);
+		if (btrfs_extent_refs(leaf, item) > (u64)refs_to_add)
+			type = BTRFS_QGROUP_OPER_ADD_SHARED;
+		btrfs_release_path(path);
+
+		ret = btrfs_qgroup_record_ref(trans, fs_info, root_objectid,
+					      bytenr, num_bytes, type, 0);
 		goto out;
 	}
 
+	/*
+	 * Ok we had -EAGAIN which means we didn't have space to insert and
+	 * inline extent ref, so just update the reference count and add a
+	 * normal backref.
+	 */
 	leaf = path->nodes[0];
+	btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
 	item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
 	refs = btrfs_extent_refs(leaf, item);
+	if (refs)
+		type = BTRFS_QGROUP_OPER_ADD_SHARED;
 	btrfs_set_extent_refs(leaf, item, refs + refs_to_add);
 	if (extent_op)
 		__run_delayed_extent_op(extent_op, leaf, item);
 
 	btrfs_mark_buffer_dirty(leaf);
-	btrfs_release_path(root->fs_info->extent_root, path);
+	btrfs_release_path(path);
+
+	if (!no_quota) {
+		ret = btrfs_qgroup_record_ref(trans, fs_info, root_objectid,
+					      bytenr, num_bytes, type, 0);
+		if (ret)
+			goto out;
+	}
 
 	path->reada = 1;
 	path->leave_spinning = 1;
-
 	/* now insert the actual backref */
 	ret = insert_extent_backref(trans, root->fs_info->extent_root,
 				    path, bytenr, parent, root_objectid,
 				    owner, offset, refs_to_add);
-	BUG_ON(ret);
+	if (ret)
+		btrfs_abort_transaction(trans, root, ret);
 out:
 	btrfs_free_path(path);
-	return err;
+	return ret;
 }
 
 static int run_delayed_data_ref(struct btrfs_trans_handle *trans,
@@ -1718,16 +2085,15 @@ static int run_delayed_data_ref(struct btrfs_trans_handle *trans,
 	ins.type = BTRFS_EXTENT_ITEM_KEY;
 
 	ref = btrfs_delayed_node_to_data_ref(node);
+	trace_run_delayed_data_ref(node, ref, node->action);
+
 	if (node->type == BTRFS_SHARED_DATA_REF_KEY)
 		parent = ref->parent;
-	else
-		ref_root = ref->root;
+	ref_root = ref->root;
 
 	if (node->action == BTRFS_ADD_DELAYED_REF && insert_reserved) {
-		if (extent_op) {
-			BUG_ON(extent_op->update_key);
+		if (extent_op)
 			flags |= extent_op->flags_to_set;
-		}
 		ret = alloc_reserved_file_extent(trans, root,
 						 parent, ref_root, flags,
 						 ref->objectid, ref->offset,
@@ -1737,13 +2103,13 @@ static int run_delayed_data_ref(struct btrfs_trans_handle *trans,
 					     node->num_bytes, parent,
 					     ref_root, ref->objectid,
 					     ref->offset, node->ref_mod,
-					     extent_op);
+					     node->no_quota, extent_op);
 	} else if (node->action == BTRFS_DROP_DELAYED_REF) {
 		ret = __btrfs_free_extent(trans, root, node->bytenr,
 					  node->num_bytes, parent,
 					  ref_root, ref->objectid,
 					  ref->offset, node->ref_mod,
-					  extent_op);
+					  extent_op, node->no_quota);
 	} else {
 		BUG();
 	}
@@ -1780,15 +2146,29 @@ static int run_delayed_extent_op(struct btrfs_trans_handle *trans,
 	u32 item_size;
 	int ret;
 	int err = 0;
+	int metadata = !extent_op->is_data;
+
+	if (trans->aborted)
+		return 0;
+
+	if (metadata && !btrfs_fs_incompat(root->fs_info, SKINNY_METADATA))
+		metadata = 0;
 
 	path = btrfs_alloc_path();
 	if (!path)
 		return -ENOMEM;
 
 	key.objectid = node->bytenr;
-	key.type = BTRFS_EXTENT_ITEM_KEY;
-	key.offset = node->num_bytes;
 
+	if (metadata) {
+		key.type = BTRFS_METADATA_ITEM_KEY;
+		key.offset = extent_op->level;
+	} else {
+		key.type = BTRFS_EXTENT_ITEM_KEY;
+		key.offset = node->num_bytes;
+	}
+
+again:
 	path->reada = 1;
 	path->leave_spinning = 1;
 	ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key,
@@ -1798,8 +2178,29 @@ static int run_delayed_extent_op(struct btrfs_trans_handle *trans,
 		goto out;
 	}
 	if (ret > 0) {
-		err = -EIO;
-		goto out;
+		if (metadata) {
+			if (path->slots[0] > 0) {
+				path->slots[0]--;
+				btrfs_item_key_to_cpu(path->nodes[0], &key,
+						      path->slots[0]);
+				if (key.objectid == node->bytenr &&
+				    key.type == BTRFS_EXTENT_ITEM_KEY &&
+				    key.offset == node->num_bytes)
+					ret = 0;
+			}
+			if (ret > 0) {
+				btrfs_release_path(path);
+				metadata = 0;
+
+				key.objectid = node->bytenr;
+				key.offset = node->num_bytes;
+				key.type = BTRFS_EXTENT_ITEM_KEY;
+				goto again;
+			}
+		} else {
+			err = -EIO;
+			goto out;
+		}
 	}
 
 	leaf = path->nodes[0];
@@ -1837,41 +2238,50 @@ static int run_delayed_tree_ref(struct btrfs_trans_handle *trans,
 	struct btrfs_key ins;
 	u64 parent = 0;
 	u64 ref_root = 0;
-
-	ins.objectid = node->bytenr;
-	ins.offset = node->num_bytes;
-	ins.type = BTRFS_EXTENT_ITEM_KEY;
+	bool skinny_metadata = btrfs_fs_incompat(root->fs_info,
+						 SKINNY_METADATA);
 
 	ref = btrfs_delayed_node_to_tree_ref(node);
+	trace_run_delayed_tree_ref(node, ref, node->action);
+
 	if (node->type == BTRFS_SHARED_BLOCK_REF_KEY)
 		parent = ref->parent;
-	else
-		ref_root = ref->root;
+	ref_root = ref->root;
+
+	ins.objectid = node->bytenr;
+	if (skinny_metadata) {
+		ins.offset = ref->level;
+		ins.type = BTRFS_METADATA_ITEM_KEY;
+	} else {
+		ins.offset = node->num_bytes;
+		ins.type = BTRFS_EXTENT_ITEM_KEY;
+	}
 
 	BUG_ON(node->ref_mod != 1);
 	if (node->action == BTRFS_ADD_DELAYED_REF && insert_reserved) {
-		BUG_ON(!extent_op || !extent_op->update_flags ||
-		       !extent_op->update_key);
+		BUG_ON(!extent_op || !extent_op->update_flags);
 		ret = alloc_reserved_tree_block(trans, root,
 						parent, ref_root,
 						extent_op->flags_to_set,
 						&extent_op->key,
-						ref->level, &ins);
+						ref->level, &ins,
+						node->no_quota);
 	} else if (node->action == BTRFS_ADD_DELAYED_REF) {
 		ret = __btrfs_inc_extent_ref(trans, root, node->bytenr,
 					     node->num_bytes, parent, ref_root,
-					     ref->level, 0, 1, extent_op);
+					     ref->level, 0, 1, node->no_quota,
+					     extent_op);
 	} else if (node->action == BTRFS_DROP_DELAYED_REF) {
 		ret = __btrfs_free_extent(trans, root, node->bytenr,
 					  node->num_bytes, parent, ref_root,
-					  ref->level, 0, 1, extent_op);
+					  ref->level, 0, 1, extent_op,
+					  node->no_quota);
 	} else {
 		BUG();
 	}
 	return ret;
 }
 
-
 /* helper function to actually process a single delayed ref entry */
 static int run_one_delayed_ref(struct btrfs_trans_handle *trans,
 			       struct btrfs_root *root,
@@ -1879,7 +2289,15 @@ static int run_one_delayed_ref(struct btrfs_trans_handle *trans,
 			       struct btrfs_delayed_extent_op *extent_op,
 			       int insert_reserved)
 {
-	int ret;
+	int ret = 0;
+
+	if (trans->aborted) {
+		if (insert_reserved)
+			btrfs_pin_extent(root, node->bytenr,
+					 node->num_bytes, 1);
+		return 0;
+	}
+
 	if (btrfs_delayed_ref_is_head(node)) {
 		struct btrfs_delayed_ref_head *head;
 		/*
@@ -1890,36 +2308,18 @@ static int run_one_delayed_ref(struct btrfs_trans_handle *trans,
 		 */
 		BUG_ON(extent_op);
 		head = btrfs_delayed_node_to_head(node);
-		if (insert_reserved) {
-			int mark_free = 0;
-			struct extent_buffer *must_clean = NULL;
-
-			ret = pin_down_bytes(trans, root, NULL,
-					     node->bytenr, node->num_bytes,
-					     head->is_data, 1, &must_clean);
-			if (ret > 0)
-				mark_free = 1;
+		trace_run_delayed_ref_head(node, head, node->action);
 
-			if (must_clean) {
-				clean_tree_block(NULL, root, must_clean);
-				btrfs_tree_unlock(must_clean);
-				free_extent_buffer(must_clean);
-			}
+		if (insert_reserved) {
+			btrfs_pin_extent(root, node->bytenr,
+					 node->num_bytes, 1);
 			if (head->is_data) {
 				ret = btrfs_del_csums(trans, root,
 						      node->bytenr,
 						      node->num_bytes);
-				BUG_ON(ret);
-			}
-			if (mark_free) {
-				ret = btrfs_free_reserved_extent(root,
-							node->bytenr,
-							node->num_bytes);
-				BUG_ON(ret);
 			}
 		}
-		mutex_unlock(&head->mutex);
-		return 0;
+		return ret;
 	}
 
 	if (node->type == BTRFS_TREE_BLOCK_REF_KEY ||
@@ -1939,59 +2339,62 @@ static noinline struct btrfs_delayed_ref_node *
 select_delayed_ref(struct btrfs_delayed_ref_head *head)
 {
 	struct rb_node *node;
-	struct btrfs_delayed_ref_node *ref;
-	int action = BTRFS_ADD_DELAYED_REF;
-again:
+	struct btrfs_delayed_ref_node *ref, *last = NULL;;
+
 	/*
 	 * select delayed ref of type BTRFS_ADD_DELAYED_REF first.
 	 * this prevents ref count from going down to zero when
 	 * there still are pending delayed ref.
 	 */
-	node = rb_prev(&head->node.rb_node);
-	while (1) {
-		if (!node)
-			break;
+	node = rb_first(&head->ref_root);
+	while (node) {
 		ref = rb_entry(node, struct btrfs_delayed_ref_node,
 				rb_node);
-		if (ref->bytenr != head->node.bytenr)
-			break;
-		if (ref->action == action)
+		if (ref->action == BTRFS_ADD_DELAYED_REF)
 			return ref;
-		node = rb_prev(node);
-	}
-	if (action == BTRFS_ADD_DELAYED_REF) {
-		action = BTRFS_DROP_DELAYED_REF;
-		goto again;
+		else if (last == NULL)
+			last = ref;
+		node = rb_next(node);
 	}
-	return NULL;
+	return last;
 }
 
-static noinline int run_clustered_refs(struct btrfs_trans_handle *trans,
-				       struct btrfs_root *root,
-				       struct list_head *cluster)
+/*
+ * Returns 0 on success or if called with an already aborted transaction.
+ * Returns -ENOMEM or -EIO on failure and will abort the transaction.
+ */
+static noinline int __btrfs_run_delayed_refs(struct btrfs_trans_handle *trans,
+					     struct btrfs_root *root,
+					     unsigned long nr)
 {
 	struct btrfs_delayed_ref_root *delayed_refs;
 	struct btrfs_delayed_ref_node *ref;
 	struct btrfs_delayed_ref_head *locked_ref = NULL;
 	struct btrfs_delayed_extent_op *extent_op;
+	struct btrfs_fs_info *fs_info = root->fs_info;
+	ktime_t start = ktime_get();
 	int ret;
-	int count = 0;
+	unsigned long count = 0;
+	unsigned long actual_count = 0;
 	int must_insert_reserved = 0;
 
 	delayed_refs = &trans->transaction->delayed_refs;
 	while (1) {
 		if (!locked_ref) {
-			/* pick a new head ref from the cluster list */
-			if (list_empty(cluster))
+			if (count >= nr)
 				break;
 
-			locked_ref = list_entry(cluster->next,
-				     struct btrfs_delayed_ref_head, cluster);
+			spin_lock(&delayed_refs->lock);
+			locked_ref = btrfs_select_ref_head(trans);
+			if (!locked_ref) {
+				spin_unlock(&delayed_refs->lock);
+				break;
+			}
 
 			/* grab the lock that says we are going to process
 			 * all the refs for this head */
 			ret = btrfs_delayed_ref_lock(trans, locked_ref);
-
+			spin_unlock(&delayed_refs->lock);
 			/*
 			 * we may have dropped the spin lock to get the head
 			 * mutex lock, and that might have given someone else
@@ -2006,6 +2409,37 @@ static noinline int run_clustered_refs(struct btrfs_trans_handle *trans,
 		}
 
 		/*
+		 * We need to try and merge add/drops of the same ref since we
+		 * can run into issues with relocate dropping the implicit ref
+		 * and then it being added back again before the drop can
+		 * finish.  If we merged anything we need to re-loop so we can
+		 * get a good ref.
+		 */
+		spin_lock(&locked_ref->lock);
+		btrfs_merge_delayed_refs(trans, fs_info, delayed_refs,
+					 locked_ref);
+
+		/*
+		 * locked_ref is the head node, so we have to go one
+		 * node back for any delayed ref updates
+		 */
+		ref = select_delayed_ref(locked_ref);
+
+		if (ref && ref->seq &&
+		    btrfs_check_delayed_seq(fs_info, delayed_refs, ref->seq)) {
+			spin_unlock(&locked_ref->lock);
+			btrfs_delayed_ref_unlock(locked_ref);
+			spin_lock(&delayed_refs->lock);
+			locked_ref->processing = 0;
+			delayed_refs->num_heads_ready++;
+			spin_unlock(&delayed_refs->lock);
+			locked_ref = NULL;
+			cond_resched();
+			count++;
+			continue;
+		}
+
+		/*
 		 * record the must insert reserved flag before we
 		 * drop the spin lock.
 		 */
@@ -2015,12 +2449,9 @@ static noinline int run_clustered_refs(struct btrfs_trans_handle *trans,
 		extent_op = locked_ref->extent_op;
 		locked_ref->extent_op = NULL;
 
-		/*
-		 * locked_ref is the head node, so we have to go one
-		 * node back for any delayed ref updates
-		 */
-		ref = select_delayed_ref(locked_ref);
 		if (!ref) {
+
+
 			/* All delayed refs have been processed, Go ahead
 			 * and send the head node to run_one_delayed_ref,
 			 * so that any accounting fixes can happen
@@ -2028,45 +2459,308 @@ static noinline int run_clustered_refs(struct btrfs_trans_handle *trans,
 			ref = &locked_ref->node;
 
 			if (extent_op && must_insert_reserved) {
-				kfree(extent_op);
+				btrfs_free_delayed_extent_op(extent_op);
 				extent_op = NULL;
 			}
 
 			if (extent_op) {
-				spin_unlock(&delayed_refs->lock);
-
+				spin_unlock(&locked_ref->lock);
 				ret = run_delayed_extent_op(trans, root,
 							    ref, extent_op);
-				BUG_ON(ret);
-				kfree(extent_op);
-
-				cond_resched();
-				spin_lock(&delayed_refs->lock);
+				btrfs_free_delayed_extent_op(extent_op);
+
+				if (ret) {
+					/*
+					 * Need to reset must_insert_reserved if
+					 * there was an error so the abort stuff
+					 * can cleanup the reserved space
+					 * properly.
+					 */
+					if (must_insert_reserved)
+						locked_ref->must_insert_reserved = 1;
+					locked_ref->processing = 0;
+					btrfs_debug(fs_info, "run_delayed_extent_op returned %d", ret);
+					btrfs_delayed_ref_unlock(locked_ref);
+					return ret;
+				}
 				continue;
 			}
 
-			list_del_init(&locked_ref->cluster);
-			locked_ref = NULL;
+			/*
+			 * Need to drop our head ref lock and re-aqcuire the
+			 * delayed ref lock and then re-check to make sure
+			 * nobody got added.
+			 */
+			spin_unlock(&locked_ref->lock);
+			spin_lock(&delayed_refs->lock);
+			spin_lock(&locked_ref->lock);
+			if (rb_first(&locked_ref->ref_root) ||
+			    locked_ref->extent_op) {
+				spin_unlock(&locked_ref->lock);
+				spin_unlock(&delayed_refs->lock);
+				continue;
+			}
+			ref->in_tree = 0;
+			delayed_refs->num_heads--;
+			rb_erase(&locked_ref->href_node,
+				 &delayed_refs->href_root);
+			spin_unlock(&delayed_refs->lock);
+		} else {
+			actual_count++;
+			ref->in_tree = 0;
+			rb_erase(&ref->rb_node, &locked_ref->ref_root);
 		}
+		atomic_dec(&delayed_refs->num_entries);
 
-		ref->in_tree = 0;
-		rb_erase(&ref->rb_node, &delayed_refs->root);
-		delayed_refs->num_entries--;
-
-		spin_unlock(&delayed_refs->lock);
+		if (!btrfs_delayed_ref_is_head(ref)) {
+			/*
+			 * when we play the delayed ref, also correct the
+			 * ref_mod on head
+			 */
+			switch (ref->action) {
+			case BTRFS_ADD_DELAYED_REF:
+			case BTRFS_ADD_DELAYED_EXTENT:
+				locked_ref->node.ref_mod -= ref->ref_mod;
+				break;
+			case BTRFS_DROP_DELAYED_REF:
+				locked_ref->node.ref_mod += ref->ref_mod;
+				break;
+			default:
+				WARN_ON(1);
+			}
+		}
+		spin_unlock(&locked_ref->lock);
 
 		ret = run_one_delayed_ref(trans, root, ref, extent_op,
 					  must_insert_reserved);
-		BUG_ON(ret);
 
+		btrfs_free_delayed_extent_op(extent_op);
+		if (ret) {
+			locked_ref->processing = 0;
+			btrfs_delayed_ref_unlock(locked_ref);
+			btrfs_put_delayed_ref(ref);
+			btrfs_debug(fs_info, "run_one_delayed_ref returned %d", ret);
+			return ret;
+		}
+
+		/*
+		 * If this node is a head, that means all the refs in this head
+		 * have been dealt with, and we will pick the next head to deal
+		 * with, so we must unlock the head and drop it from the cluster
+		 * list before we release it.
+		 */
+		if (btrfs_delayed_ref_is_head(ref)) {
+			btrfs_delayed_ref_unlock(locked_ref);
+			locked_ref = NULL;
+		}
 		btrfs_put_delayed_ref(ref);
-		kfree(extent_op);
 		count++;
-
 		cond_resched();
+	}
+
+	/*
+	 * We don't want to include ref heads since we can have empty ref heads
+	 * and those will drastically skew our runtime down since we just do
+	 * accounting, no actual extent tree updates.
+	 */
+	if (actual_count > 0) {
+		u64 runtime = ktime_to_ns(ktime_sub(ktime_get(), start));
+		u64 avg;
+
+		/*
+		 * We weigh the current average higher than our current runtime
+		 * to avoid large swings in the average.
+		 */
 		spin_lock(&delayed_refs->lock);
+		avg = fs_info->avg_delayed_ref_runtime * 3 + runtime;
+		avg = div64_u64(avg, 4);
+		fs_info->avg_delayed_ref_runtime = avg;
+		spin_unlock(&delayed_refs->lock);
 	}
-	return count;
+	return 0;
+}
+
+#ifdef SCRAMBLE_DELAYED_REFS
+/*
+ * Normally delayed refs get processed in ascending bytenr order. This
+ * correlates in most cases to the order added. To expose dependencies on this
+ * order, we start to process the tree in the middle instead of the beginning
+ */
+static u64 find_middle(struct rb_root *root)
+{
+	struct rb_node *n = root->rb_node;
+	struct btrfs_delayed_ref_node *entry;
+	int alt = 1;
+	u64 middle;
+	u64 first = 0, last = 0;
+
+	n = rb_first(root);
+	if (n) {
+		entry = rb_entry(n, struct btrfs_delayed_ref_node, rb_node);
+		first = entry->bytenr;
+	}
+	n = rb_last(root);
+	if (n) {
+		entry = rb_entry(n, struct btrfs_delayed_ref_node, rb_node);
+		last = entry->bytenr;
+	}
+	n = root->rb_node;
+
+	while (n) {
+		entry = rb_entry(n, struct btrfs_delayed_ref_node, rb_node);
+		WARN_ON(!entry->in_tree);
+
+		middle = entry->bytenr;
+
+		if (alt)
+			n = n->rb_left;
+		else
+			n = n->rb_right;
+
+		alt = 1 - alt;
+	}
+	return middle;
+}
+#endif
+
+static inline u64 heads_to_leaves(struct btrfs_root *root, u64 heads)
+{
+	u64 num_bytes;
+
+	num_bytes = heads * (sizeof(struct btrfs_extent_item) +
+			     sizeof(struct btrfs_extent_inline_ref));
+	if (!btrfs_fs_incompat(root->fs_info, SKINNY_METADATA))
+		num_bytes += heads * sizeof(struct btrfs_tree_block_info);
+
+	/*
+	 * We don't ever fill up leaves all the way so multiply by 2 just to be
+	 * closer to what we're really going to want to ouse.
+	 */
+	return div64_u64(num_bytes, BTRFS_LEAF_DATA_SIZE(root));
+}
+
+int btrfs_check_space_for_delayed_refs(struct btrfs_trans_handle *trans,
+				       struct btrfs_root *root)
+{
+	struct btrfs_block_rsv *global_rsv;
+	u64 num_heads = trans->transaction->delayed_refs.num_heads_ready;
+	u64 num_bytes;
+	int ret = 0;
+
+	num_bytes = btrfs_calc_trans_metadata_size(root, 1);
+	num_heads = heads_to_leaves(root, num_heads);
+	if (num_heads > 1)
+		num_bytes += (num_heads - 1) * root->leafsize;
+	num_bytes <<= 1;
+	global_rsv = &root->fs_info->global_block_rsv;
+
+	/*
+	 * If we can't allocate any more chunks lets make sure we have _lots_ of
+	 * wiggle room since running delayed refs can create more delayed refs.
+	 */
+	if (global_rsv->space_info->full)
+		num_bytes <<= 1;
+
+	spin_lock(&global_rsv->lock);
+	if (global_rsv->reserved <= num_bytes)
+		ret = 1;
+	spin_unlock(&global_rsv->lock);
+	return ret;
+}
+
+int btrfs_should_throttle_delayed_refs(struct btrfs_trans_handle *trans,
+				       struct btrfs_root *root)
+{
+	struct btrfs_fs_info *fs_info = root->fs_info;
+	u64 num_entries =
+		atomic_read(&trans->transaction->delayed_refs.num_entries);
+	u64 avg_runtime;
+	u64 val;
+
+	smp_mb();
+	avg_runtime = fs_info->avg_delayed_ref_runtime;
+	val = num_entries * avg_runtime;
+	if (num_entries * avg_runtime >= NSEC_PER_SEC)
+		return 1;
+	if (val >= NSEC_PER_SEC / 2)
+		return 2;
+
+	return btrfs_check_space_for_delayed_refs(trans, root);
+}
+
+struct async_delayed_refs {
+	struct btrfs_root *root;
+	int count;
+	int error;
+	int sync;
+	struct completion wait;
+	struct btrfs_work work;
+};
+
+static void delayed_ref_async_start(struct btrfs_work *work)
+{
+	struct async_delayed_refs *async;
+	struct btrfs_trans_handle *trans;
+	int ret;
+
+	async = container_of(work, struct async_delayed_refs, work);
+
+	trans = btrfs_join_transaction(async->root);
+	if (IS_ERR(trans)) {
+		async->error = PTR_ERR(trans);
+		goto done;
+	}
+
+	/*
+	 * trans->sync means that when we call end_transaciton, we won't
+	 * wait on delayed refs
+	 */
+	trans->sync = true;
+	ret = btrfs_run_delayed_refs(trans, async->root, async->count);
+	if (ret)
+		async->error = ret;
+
+	ret = btrfs_end_transaction(trans, async->root);
+	if (ret && !async->error)
+		async->error = ret;
+done:
+	if (async->sync)
+		complete(&async->wait);
+	else
+		kfree(async);
+}
+
+int btrfs_async_run_delayed_refs(struct btrfs_root *root,
+				 unsigned long count, int wait)
+{
+	struct async_delayed_refs *async;
+	int ret;
+
+	async = kmalloc(sizeof(*async), GFP_NOFS);
+	if (!async)
+		return -ENOMEM;
+
+	async->root = root->fs_info->tree_root;
+	async->count = count;
+	async->error = 0;
+	if (wait)
+		async->sync = 1;
+	else
+		async->sync = 0;
+	init_completion(&async->wait);
+
+	btrfs_init_work(&async->work, delayed_ref_async_start,
+			NULL, NULL);
+
+	btrfs_queue_work(root->fs_info->extent_workers, &async->work);
+
+	if (wait) {
+		wait_for_completion(&async->wait);
+		ret = async->error;
+		kfree(async);
+		return ret;
+	}
+	return 0;
 }
 
 /*
@@ -2075,97 +2769,101 @@ static noinline int run_clustered_refs(struct btrfs_trans_handle *trans,
  * 0, which means to process everything in the tree at the start
  * of the run (but not newly added entries), or it can be some target
  * number you'd like to process.
+ *
+ * Returns 0 on success or if called with an aborted transaction
+ * Returns <0 on error and aborts the transaction
  */
 int btrfs_run_delayed_refs(struct btrfs_trans_handle *trans,
 			   struct btrfs_root *root, unsigned long count)
 {
 	struct rb_node *node;
 	struct btrfs_delayed_ref_root *delayed_refs;
-	struct btrfs_delayed_ref_node *ref;
-	struct list_head cluster;
+	struct btrfs_delayed_ref_head *head;
 	int ret;
 	int run_all = count == (unsigned long)-1;
 	int run_most = 0;
 
+	/* We'll clean this up in btrfs_cleanup_transaction */
+	if (trans->aborted)
+		return 0;
+
 	if (root == root->fs_info->extent_root)
 		root = root->fs_info->tree_root;
 
 	delayed_refs = &trans->transaction->delayed_refs;
-	INIT_LIST_HEAD(&cluster);
-again:
-	spin_lock(&delayed_refs->lock);
 	if (count == 0) {
-		count = delayed_refs->num_entries * 2;
+		count = atomic_read(&delayed_refs->num_entries) * 2;
 		run_most = 1;
 	}
-	while (1) {
-		if (!(run_all || run_most) &&
-		    delayed_refs->num_heads_ready < 64)
-			break;
-
-		/*
-		 * go find something we can process in the rbtree.  We start at
-		 * the beginning of the tree, and then build a cluster
-		 * of refs to process starting at the first one we are able to
-		 * lock
-		 */
-		ret = btrfs_find_ref_cluster(trans, &cluster,
-					     delayed_refs->run_delayed_start);
-		if (ret)
-			break;
-
-		ret = run_clustered_refs(trans, root, &cluster);
-		BUG_ON(ret < 0);
-
-		count -= min_t(unsigned long, ret, count);
 
-		if (count == 0)
-			break;
+again:
+#ifdef SCRAMBLE_DELAYED_REFS
+	delayed_refs->run_delayed_start = find_middle(&delayed_refs->root);
+#endif
+	ret = __btrfs_run_delayed_refs(trans, root, count);
+	if (ret < 0) {
+		btrfs_abort_transaction(trans, root, ret);
+		return ret;
 	}
 
 	if (run_all) {
-		node = rb_first(&delayed_refs->root);
-		if (!node)
+		if (!list_empty(&trans->new_bgs))
+			btrfs_create_pending_block_groups(trans, root);
+
+		spin_lock(&delayed_refs->lock);
+		node = rb_first(&delayed_refs->href_root);
+		if (!node) {
+			spin_unlock(&delayed_refs->lock);
 			goto out;
+		}
 		count = (unsigned long)-1;
 
 		while (node) {
-			ref = rb_entry(node, struct btrfs_delayed_ref_node,
-				       rb_node);
-			if (btrfs_delayed_ref_is_head(ref)) {
-				struct btrfs_delayed_ref_head *head;
+			head = rb_entry(node, struct btrfs_delayed_ref_head,
+					href_node);
+			if (btrfs_delayed_ref_is_head(&head->node)) {
+				struct btrfs_delayed_ref_node *ref;
 
-				head = btrfs_delayed_node_to_head(ref);
+				ref = &head->node;
 				atomic_inc(&ref->refs);
 
 				spin_unlock(&delayed_refs->lock);
+				/*
+				 * Mutex was contended, block until it's
+				 * released and try again
+				 */
 				mutex_lock(&head->mutex);
 				mutex_unlock(&head->mutex);
 
 				btrfs_put_delayed_ref(ref);
 				cond_resched();
 				goto again;
+			} else {
+				WARN_ON(1);
 			}
 			node = rb_next(node);
 		}
 		spin_unlock(&delayed_refs->lock);
-		schedule_timeout(1);
+		cond_resched();
 		goto again;
 	}
 out:
-	spin_unlock(&delayed_refs->lock);
+	ret = btrfs_delayed_qgroup_accounting(trans, root->fs_info);
+	if (ret)
+		return ret;
+	assert_qgroups_uptodate(trans);
 	return 0;
 }
 
 int btrfs_set_disk_extent_flags(struct btrfs_trans_handle *trans,
 				struct btrfs_root *root,
 				u64 bytenr, u64 num_bytes, u64 flags,
-				int is_data)
+				int level, int is_data)
 {
 	struct btrfs_delayed_extent_op *extent_op;
 	int ret;
 
-	extent_op = kmalloc(sizeof(*extent_op), GFP_NOFS);
+	extent_op = btrfs_alloc_delayed_extent_op();
 	if (!extent_op)
 		return -ENOMEM;
 
@@ -2173,10 +2871,12 @@ int btrfs_set_disk_extent_flags(struct btrfs_trans_handle *trans,
 	extent_op->update_flags = 1;
 	extent_op->update_key = 0;
 	extent_op->is_data = is_data ? 1 : 0;
+	extent_op->level = level;
 
-	ret = btrfs_add_delayed_extent_op(trans, bytenr, num_bytes, extent_op);
+	ret = btrfs_add_delayed_extent_op(root->fs_info, trans, bytenr,
+					  num_bytes, extent_op);
 	if (ret)
-		kfree(extent_op);
+		btrfs_free_delayed_extent_op(extent_op);
 	return ret;
 }
 
@@ -2192,56 +2892,58 @@ static noinline int check_delayed_ref(struct btrfs_trans_handle *trans,
 	struct rb_node *node;
 	int ret = 0;
 
-	ret = -ENOENT;
 	delayed_refs = &trans->transaction->delayed_refs;
 	spin_lock(&delayed_refs->lock);
 	head = btrfs_find_delayed_ref_head(trans, bytenr);
-	if (!head)
-		goto out;
+	if (!head) {
+		spin_unlock(&delayed_refs->lock);
+		return 0;
+	}
 
 	if (!mutex_trylock(&head->mutex)) {
 		atomic_inc(&head->node.refs);
 		spin_unlock(&delayed_refs->lock);
 
-		btrfs_release_path(root->fs_info->extent_root, path);
+		btrfs_release_path(path);
 
+		/*
+		 * Mutex was contended, block until it's released and let
+		 * caller try again
+		 */
 		mutex_lock(&head->mutex);
 		mutex_unlock(&head->mutex);
 		btrfs_put_delayed_ref(&head->node);
 		return -EAGAIN;
 	}
+	spin_unlock(&delayed_refs->lock);
 
-	node = rb_prev(&head->node.rb_node);
-	if (!node)
-		goto out_unlock;
-
-	ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);
-
-	if (ref->bytenr != bytenr)
-		goto out_unlock;
+	spin_lock(&head->lock);
+	node = rb_first(&head->ref_root);
+	while (node) {
+		ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);
+		node = rb_next(node);
 
-	ret = 1;
-	if (ref->type != BTRFS_EXTENT_DATA_REF_KEY)
-		goto out_unlock;
+		/* If it's a shared ref we know a cross reference exists */
+		if (ref->type != BTRFS_EXTENT_DATA_REF_KEY) {
+			ret = 1;
+			break;
+		}
 
-	data_ref = btrfs_delayed_node_to_data_ref(ref);
+		data_ref = btrfs_delayed_node_to_data_ref(ref);
 
-	node = rb_prev(node);
-	if (node) {
-		ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);
-		if (ref->bytenr == bytenr)
-			goto out_unlock;
+		/*
+		 * If our ref doesn't match the one we're currently looking at
+		 * then we have a cross reference.
+		 */
+		if (data_ref->root != root->root_key.objectid ||
+		    data_ref->objectid != objectid ||
+		    data_ref->offset != offset) {
+			ret = 1;
+			break;
+		}
 	}
-
-	if (data_ref->root != root->root_key.objectid ||
-	    data_ref->objectid != objectid || data_ref->offset != offset)
-		goto out_unlock;
-
-	ret = 0;
-out_unlock:
+	spin_unlock(&head->lock);
 	mutex_unlock(&head->mutex);
-out:
-	spin_unlock(&delayed_refs->lock);
 	return ret;
 }
 
@@ -2266,7 +2968,7 @@ static noinline int check_committed_ref(struct btrfs_trans_handle *trans,
 	ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0);
 	if (ret < 0)
 		goto out;
-	BUG_ON(ret == 0);
+	BUG_ON(ret == 0); /* Corruption */
 
 	ret = -ENOENT;
 	if (path->slots[0] == 0)
@@ -2347,133 +3049,15 @@ int btrfs_cross_ref_exist(struct btrfs_trans_handle *trans,
 		ret = 0;
 out:
 	btrfs_free_path(path);
+	if (root->root_key.objectid == BTRFS_DATA_RELOC_TREE_OBJECTID)
+		WARN_ON(ret > 0);
 	return ret;
 }
 
-#if 0
-int btrfs_cache_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
-		    struct extent_buffer *buf, u32 nr_extents)
-{
-	struct btrfs_key key;
-	struct btrfs_file_extent_item *fi;
-	u64 root_gen;
-	u32 nritems;
-	int i;
-	int level;
-	int ret = 0;
-	int shared = 0;
-
-	if (!root->ref_cows)
-		return 0;
-
-	if (root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID) {
-		shared = 0;
-		root_gen = root->root_key.offset;
-	} else {
-		shared = 1;
-		root_gen = trans->transid - 1;
-	}
-
-	level = btrfs_header_level(buf);
-	nritems = btrfs_header_nritems(buf);
-
-	if (level == 0) {
-		struct btrfs_leaf_ref *ref;
-		struct btrfs_extent_info *info;
-
-		ref = btrfs_alloc_leaf_ref(root, nr_extents);
-		if (!ref) {
-			ret = -ENOMEM;
-			goto out;
-		}
-
-		ref->root_gen = root_gen;
-		ref->bytenr = buf->start;
-		ref->owner = btrfs_header_owner(buf);
-		ref->generation = btrfs_header_generation(buf);
-		ref->nritems = nr_extents;
-		info = ref->extents;
-
-		for (i = 0; nr_extents > 0 && i < nritems; i++) {
-			u64 disk_bytenr;
-			btrfs_item_key_to_cpu(buf, &key, i);
-			if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
-				continue;
-			fi = btrfs_item_ptr(buf, i,
-					    struct btrfs_file_extent_item);
-			if (btrfs_file_extent_type(buf, fi) ==
-			    BTRFS_FILE_EXTENT_INLINE)
-				continue;
-			disk_bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
-			if (disk_bytenr == 0)
-				continue;
-
-			info->bytenr = disk_bytenr;
-			info->num_bytes =
-				btrfs_file_extent_disk_num_bytes(buf, fi);
-			info->objectid = key.objectid;
-			info->offset = key.offset;
-			info++;
-		}
-
-		ret = btrfs_add_leaf_ref(root, ref, shared);
-		if (ret == -EEXIST && shared) {
-			struct btrfs_leaf_ref *old;
-			old = btrfs_lookup_leaf_ref(root, ref->bytenr);
-			BUG_ON(!old);
-			btrfs_remove_leaf_ref(root, old);
-			btrfs_free_leaf_ref(root, old);
-			ret = btrfs_add_leaf_ref(root, ref, shared);
-		}
-		WARN_ON(ret);
-		btrfs_free_leaf_ref(root, ref);
-	}
-out:
-	return ret;
-}
-
-/* when a block goes through cow, we update the reference counts of
- * everything that block points to.  The internal pointers of the block
- * can be in just about any order, and it is likely to have clusters of
- * things that are close together and clusters of things that are not.
- *
- * To help reduce the seeks that come with updating all of these reference
- * counts, sort them by byte number before actual updates are done.
- *
- * struct refsort is used to match byte number to slot in the btree block.
- * we sort based on the byte number and then use the slot to actually
- * find the item.
- *
- * struct refsort is smaller than strcut btrfs_item and smaller than
- * struct btrfs_key_ptr.  Since we're currently limited to the page size
- * for a btree block, there's no way for a kmalloc of refsorts for a
- * single node to be bigger than a page.
- */
-struct refsort {
-	u64 bytenr;
-	u32 slot;
-};
-
-/*
- * for passing into sort()
- */
-static int refsort_cmp(const void *a_void, const void *b_void)
-{
-	const struct refsort *a = a_void;
-	const struct refsort *b = b_void;
-
-	if (a->bytenr < b->bytenr)
-		return -1;
-	if (a->bytenr > b->bytenr)
-		return 1;
-	return 0;
-}
-#endif
-
 static int __btrfs_mod_ref(struct btrfs_trans_handle *trans,
 			   struct btrfs_root *root,
 			   struct extent_buffer *buf,
-			   int full_backref, int inc)
+			   int full_backref, int inc, int no_quota)
 {
 	u64 bytenr;
 	u64 num_bytes;
@@ -2486,13 +3070,17 @@ static int __btrfs_mod_ref(struct btrfs_trans_handle *trans,
 	int level;
 	int ret = 0;
 	int (*process_func)(struct btrfs_trans_handle *, struct btrfs_root *,
-			    u64, u64, u64, u64, u64, u64);
+			    u64, u64, u64, u64, u64, u64, int);
 
+#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
+	if (unlikely(test_bit(BTRFS_ROOT_DUMMY_ROOT, &root->state)))
+		return 0;
+#endif
 	ref_root = btrfs_header_owner(buf);
 	nritems = btrfs_header_nritems(buf);
 	level = btrfs_header_level(buf);
 
-	if (!root->ref_cows && level == 0)
+	if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state) && level == 0)
 		return 0;
 
 	if (inc)
@@ -2523,34 +3111,34 @@ static int __btrfs_mod_ref(struct btrfs_trans_handle *trans,
 			key.offset -= btrfs_file_extent_offset(buf, fi);
 			ret = process_func(trans, root, bytenr, num_bytes,
 					   parent, ref_root, key.objectid,
-					   key.offset);
+					   key.offset, no_quota);
 			if (ret)
 				goto fail;
 		} else {
 			bytenr = btrfs_node_blockptr(buf, i);
 			num_bytes = btrfs_level_size(root, level - 1);
 			ret = process_func(trans, root, bytenr, num_bytes,
-					   parent, ref_root, level - 1, 0);
+					   parent, ref_root, level - 1, 0,
+					   no_quota);
 			if (ret)
 				goto fail;
 		}
 	}
 	return 0;
 fail:
-	BUG();
 	return ret;
 }
 
 int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
-		  struct extent_buffer *buf, int full_backref)
+		  struct extent_buffer *buf, int full_backref, int no_quota)
 {
-	return __btrfs_mod_ref(trans, root, buf, full_backref, 1);
+	return __btrfs_mod_ref(trans, root, buf, full_backref, 1, no_quota);
 }
 
 int btrfs_dec_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
-		  struct extent_buffer *buf, int full_backref)
+		  struct extent_buffer *buf, int full_backref, int no_quota)
 {
-	return __btrfs_mod_ref(trans, root, buf, full_backref, 0);
+	return __btrfs_mod_ref(trans, root, buf, full_backref, 0, no_quota);
 }
 
 static int write_one_cache_group(struct btrfs_trans_handle *trans,
@@ -2566,16 +3154,18 @@ static int write_one_cache_group(struct btrfs_trans_handle *trans,
 	ret = btrfs_search_slot(trans, extent_root, &cache->key, path, 0, 1);
 	if (ret < 0)
 		goto fail;
-	BUG_ON(ret);
+	BUG_ON(ret); /* Corruption */
 
 	leaf = path->nodes[0];
 	bi = btrfs_item_ptr_offset(leaf, path->slots[0]);
 	write_extent_buffer(leaf, &cache->item, bi, sizeof(cache->item));
 	btrfs_mark_buffer_dirty(leaf);
-	btrfs_release_path(extent_root, path);
+	btrfs_release_path(path);
 fail:
-	if (ret)
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
 		return ret;
+	}
 	return 0;
 
 }
@@ -2598,6 +3188,130 @@ next_block_group(struct btrfs_root *root,
 	return cache;
 }
 
+static int cache_save_setup(struct btrfs_block_group_cache *block_group,
+			    struct btrfs_trans_handle *trans,
+			    struct btrfs_path *path)
+{
+	struct btrfs_root *root = block_group->fs_info->tree_root;
+	struct inode *inode = NULL;
+	u64 alloc_hint = 0;
+	int dcs = BTRFS_DC_ERROR;
+	int num_pages = 0;
+	int retries = 0;
+	int ret = 0;
+
+	/*
+	 * If this block group is smaller than 100 megs don't bother caching the
+	 * block group.
+	 */
+	if (block_group->key.offset < (100 * 1024 * 1024)) {
+		spin_lock(&block_group->lock);
+		block_group->disk_cache_state = BTRFS_DC_WRITTEN;
+		spin_unlock(&block_group->lock);
+		return 0;
+	}
+
+again:
+	inode = lookup_free_space_inode(root, block_group, path);
+	if (IS_ERR(inode) && PTR_ERR(inode) != -ENOENT) {
+		ret = PTR_ERR(inode);
+		btrfs_release_path(path);
+		goto out;
+	}
+
+	if (IS_ERR(inode)) {
+		BUG_ON(retries);
+		retries++;
+
+		if (block_group->ro)
+			goto out_free;
+
+		ret = create_free_space_inode(root, trans, block_group, path);
+		if (ret)
+			goto out_free;
+		goto again;
+	}
+
+	/* We've already setup this transaction, go ahead and exit */
+	if (block_group->cache_generation == trans->transid &&
+	    i_size_read(inode)) {
+		dcs = BTRFS_DC_SETUP;
+		goto out_put;
+	}
+
+	/*
+	 * We want to set the generation to 0, that way if anything goes wrong
+	 * from here on out we know not to trust this cache when we load up next
+	 * time.
+	 */
+	BTRFS_I(inode)->generation = 0;
+	ret = btrfs_update_inode(trans, root, inode);
+	WARN_ON(ret);
+
+	if (i_size_read(inode) > 0) {
+		ret = btrfs_check_trunc_cache_free_space(root,
+					&root->fs_info->global_block_rsv);
+		if (ret)
+			goto out_put;
+
+		ret = btrfs_truncate_free_space_cache(root, trans, inode);
+		if (ret)
+			goto out_put;
+	}
+
+	spin_lock(&block_group->lock);
+	if (block_group->cached != BTRFS_CACHE_FINISHED ||
+	    !btrfs_test_opt(root, SPACE_CACHE) ||
+	    block_group->delalloc_bytes) {
+		/*
+		 * don't bother trying to write stuff out _if_
+		 * a) we're not cached,
+		 * b) we're with nospace_cache mount option.
+		 */
+		dcs = BTRFS_DC_WRITTEN;
+		spin_unlock(&block_group->lock);
+		goto out_put;
+	}
+	spin_unlock(&block_group->lock);
+
+	/*
+	 * Try to preallocate enough space based on how big the block group is.
+	 * Keep in mind this has to include any pinned space which could end up
+	 * taking up quite a bit since it's not folded into the other space
+	 * cache.
+	 */
+	num_pages = (int)div64_u64(block_group->key.offset, 256 * 1024 * 1024);
+	if (!num_pages)
+		num_pages = 1;
+
+	num_pages *= 16;
+	num_pages *= PAGE_CACHE_SIZE;
+
+	ret = btrfs_check_data_free_space(inode, num_pages);
+	if (ret)
+		goto out_put;
+
+	ret = btrfs_prealloc_file_range_trans(inode, trans, 0, 0, num_pages,
+					      num_pages, num_pages,
+					      &alloc_hint);
+	if (!ret)
+		dcs = BTRFS_DC_SETUP;
+	btrfs_free_reserved_data_space(inode, num_pages);
+
+out_put:
+	iput(inode);
+out_free:
+	btrfs_release_path(path);
+out:
+	spin_lock(&block_group->lock);
+	if (!ret && dcs == BTRFS_DC_SETUP)
+		block_group->cache_generation = trans->transid;
+	block_group->disk_cache_state = dcs;
+	spin_unlock(&block_group->lock);
+
+	return ret;
+}
+
 int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans,
 				   struct btrfs_root *root)
 {
@@ -2610,15 +3324,40 @@ int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans,
 	if (!path)
 		return -ENOMEM;
 
+again:
+	while (1) {
+		cache = btrfs_lookup_first_block_group(root->fs_info, last);
+		while (cache) {
+			if (cache->disk_cache_state == BTRFS_DC_CLEAR)
+				break;
+			cache = next_block_group(root, cache);
+		}
+		if (!cache) {
+			if (last == 0)
+				break;
+			last = 0;
+			continue;
+		}
+		err = cache_save_setup(cache, trans, path);
+		last = cache->key.objectid + cache->key.offset;
+		btrfs_put_block_group(cache);
+	}
+
 	while (1) {
 		if (last == 0) {
 			err = btrfs_run_delayed_refs(trans, root,
 						     (unsigned long)-1);
-			BUG_ON(err);
+			if (err) /* File system offline */
+				goto out;
 		}
 
 		cache = btrfs_lookup_first_block_group(root->fs_info, last);
 		while (cache) {
+			if (cache->disk_cache_state == BTRFS_DC_CLEAR) {
+				btrfs_put_block_group(cache);
+				goto again;
+			}
+
 			if (cache->dirty)
 				break;
 			cache = next_block_group(root, cache);
@@ -2630,16 +3369,67 @@ int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans,
 			continue;
 		}
 
+		if (cache->disk_cache_state == BTRFS_DC_SETUP)
+			cache->disk_cache_state = BTRFS_DC_NEED_WRITE;
 		cache->dirty = 0;
 		last = cache->key.objectid + cache->key.offset;
 
 		err = write_one_cache_group(trans, root, path, cache);
-		BUG_ON(err);
+		btrfs_put_block_group(cache);
+		if (err) /* File system offline */
+			goto out;
+	}
+
+	while (1) {
+		/*
+		 * I don't think this is needed since we're just marking our
+		 * preallocated extent as written, but just in case it can't
+		 * hurt.
+		 */
+		if (last == 0) {
+			err = btrfs_run_delayed_refs(trans, root,
+						     (unsigned long)-1);
+			if (err) /* File system offline */
+				goto out;
+		}
+
+		cache = btrfs_lookup_first_block_group(root->fs_info, last);
+		while (cache) {
+			/*
+			 * Really this shouldn't happen, but it could if we
+			 * couldn't write the entire preallocated extent and
+			 * splitting the extent resulted in a new block.
+			 */
+			if (cache->dirty) {
+				btrfs_put_block_group(cache);
+				goto again;
+			}
+			if (cache->disk_cache_state == BTRFS_DC_NEED_WRITE)
+				break;
+			cache = next_block_group(root, cache);
+		}
+		if (!cache) {
+			if (last == 0)
+				break;
+			last = 0;
+			continue;
+		}
+
+		err = btrfs_write_out_cache(root, trans, cache, path);
+
+		/*
+		 * If we didn't have an error then the cache state is still
+		 * NEED_WRITE, so we can set it to WRITTEN.
+		 */
+		if (!err && cache->disk_cache_state == BTRFS_DC_NEED_WRITE)
+			cache->disk_cache_state = BTRFS_DC_WRITTEN;
+		last = cache->key.objectid + cache->key.offset;
 		btrfs_put_block_group(cache);
 	}
+out:
 
 	btrfs_free_path(path);
-	return 0;
+	return err;
 }
 
 int btrfs_extent_readonly(struct btrfs_root *root, u64 bytenr)
@@ -2655,17 +3445,45 @@ int btrfs_extent_readonly(struct btrfs_root *root, u64 bytenr)
 	return readonly;
 }
 
+static const char *alloc_name(u64 flags)
+{
+	switch (flags) {
+	case BTRFS_BLOCK_GROUP_METADATA|BTRFS_BLOCK_GROUP_DATA:
+		return "mixed";
+	case BTRFS_BLOCK_GROUP_METADATA:
+		return "metadata";
+	case BTRFS_BLOCK_GROUP_DATA:
+		return "data";
+	case BTRFS_BLOCK_GROUP_SYSTEM:
+		return "system";
+	default:
+		WARN_ON(1);
+		return "invalid-combination";
+	};
+}
+
 static int update_space_info(struct btrfs_fs_info *info, u64 flags,
 			     u64 total_bytes, u64 bytes_used,
 			     struct btrfs_space_info **space_info)
 {
 	struct btrfs_space_info *found;
+	int i;
+	int factor;
+	int ret;
+
+	if (flags & (BTRFS_BLOCK_GROUP_DUP | BTRFS_BLOCK_GROUP_RAID1 |
+		     BTRFS_BLOCK_GROUP_RAID10))
+		factor = 2;
+	else
+		factor = 1;
 
 	found = __find_space_info(info, flags);
 	if (found) {
 		spin_lock(&found->lock);
 		found->total_bytes += total_bytes;
+		found->disk_total += total_bytes * factor;
 		found->bytes_used += bytes_used;
+		found->disk_used += bytes_used * factor;
 		found->full = 0;
 		spin_unlock(&found->lock);
 		*space_info = found;
@@ -2675,812 +3493,1926 @@ static int update_space_info(struct btrfs_fs_info *info, u64 flags,
 	if (!found)
 		return -ENOMEM;
 
-	INIT_LIST_HEAD(&found->block_groups);
+	ret = percpu_counter_init(&found->total_bytes_pinned, 0);
+	if (ret) {
+		kfree(found);
+		return ret;
+	}
+
+	for (i = 0; i < BTRFS_NR_RAID_TYPES; i++)
+		INIT_LIST_HEAD(&found->block_groups[i]);
 	init_rwsem(&found->groups_sem);
-	init_waitqueue_head(&found->flush_wait);
-	init_waitqueue_head(&found->allocate_wait);
 	spin_lock_init(&found->lock);
-	found->flags = flags;
+	found->flags = flags & BTRFS_BLOCK_GROUP_TYPE_MASK;
 	found->total_bytes = total_bytes;
+	found->disk_total = total_bytes * factor;
 	found->bytes_used = bytes_used;
+	found->disk_used = bytes_used * factor;
 	found->bytes_pinned = 0;
 	found->bytes_reserved = 0;
 	found->bytes_readonly = 0;
-	found->bytes_delalloc = 0;
+	found->bytes_may_use = 0;
 	found->full = 0;
-	found->force_alloc = 0;
+	found->force_alloc = CHUNK_ALLOC_NO_FORCE;
+	found->chunk_alloc = 0;
+	found->flush = 0;
+	init_waitqueue_head(&found->wait);
+
+	ret = kobject_init_and_add(&found->kobj, &space_info_ktype,
+				    info->space_info_kobj, "%s",
+				    alloc_name(found->flags));
+	if (ret) {
+		kfree(found);
+		return ret;
+	}
+
 	*space_info = found;
 	list_add_rcu(&found->list, &info->space_info);
-	atomic_set(&found->caching_threads, 0);
-	return 0;
+	if (flags & BTRFS_BLOCK_GROUP_DATA)
+		info->data_sinfo = found;
+
+	return ret;
 }
 
 static void set_avail_alloc_bits(struct btrfs_fs_info *fs_info, u64 flags)
 {
-	u64 extra_flags = flags & (BTRFS_BLOCK_GROUP_RAID0 |
-				   BTRFS_BLOCK_GROUP_RAID1 |
-				   BTRFS_BLOCK_GROUP_RAID10 |
-				   BTRFS_BLOCK_GROUP_DUP);
-	if (extra_flags) {
-		if (flags & BTRFS_BLOCK_GROUP_DATA)
-			fs_info->avail_data_alloc_bits |= extra_flags;
-		if (flags & BTRFS_BLOCK_GROUP_METADATA)
-			fs_info->avail_metadata_alloc_bits |= extra_flags;
-		if (flags & BTRFS_BLOCK_GROUP_SYSTEM)
-			fs_info->avail_system_alloc_bits |= extra_flags;
-	}
+	u64 extra_flags = chunk_to_extended(flags) &
+				BTRFS_EXTENDED_PROFILE_MASK;
+
+	write_seqlock(&fs_info->profiles_lock);
+	if (flags & BTRFS_BLOCK_GROUP_DATA)
+		fs_info->avail_data_alloc_bits |= extra_flags;
+	if (flags & BTRFS_BLOCK_GROUP_METADATA)
+		fs_info->avail_metadata_alloc_bits |= extra_flags;
+	if (flags & BTRFS_BLOCK_GROUP_SYSTEM)
+		fs_info->avail_system_alloc_bits |= extra_flags;
+	write_sequnlock(&fs_info->profiles_lock);
 }
 
-static void set_block_group_readonly(struct btrfs_block_group_cache *cache)
+/*
+ * returns target flags in extended format or 0 if restripe for this
+ * chunk_type is not in progress
+ *
+ * should be called with either volume_mutex or balance_lock held
+ */
+static u64 get_restripe_target(struct btrfs_fs_info *fs_info, u64 flags)
 {
-	spin_lock(&cache->space_info->lock);
-	spin_lock(&cache->lock);
-	if (!cache->ro) {
-		cache->space_info->bytes_readonly += cache->key.offset -
-					btrfs_block_group_used(&cache->item);
-		cache->ro = 1;
+	struct btrfs_balance_control *bctl = fs_info->balance_ctl;
+	u64 target = 0;
+
+	if (!bctl)
+		return 0;
+
+	if (flags & BTRFS_BLOCK_GROUP_DATA &&
+	    bctl->data.flags & BTRFS_BALANCE_ARGS_CONVERT) {
+		target = BTRFS_BLOCK_GROUP_DATA | bctl->data.target;
+	} else if (flags & BTRFS_BLOCK_GROUP_SYSTEM &&
+		   bctl->sys.flags & BTRFS_BALANCE_ARGS_CONVERT) {
+		target = BTRFS_BLOCK_GROUP_SYSTEM | bctl->sys.target;
+	} else if (flags & BTRFS_BLOCK_GROUP_METADATA &&
+		   bctl->meta.flags & BTRFS_BALANCE_ARGS_CONVERT) {
+		target = BTRFS_BLOCK_GROUP_METADATA | bctl->meta.target;
 	}
-	spin_unlock(&cache->lock);
-	spin_unlock(&cache->space_info->lock);
+
+	return target;
 }
 
-u64 btrfs_reduce_alloc_profile(struct btrfs_root *root, u64 flags)
+/*
+ * @flags: available profiles in extended format (see ctree.h)
+ *
+ * Returns reduced profile in chunk format.  If profile changing is in
+ * progress (either running or paused) picks the target profile (if it's
+ * already available), otherwise falls back to plain reducing.
+ */
+static u64 btrfs_reduce_alloc_profile(struct btrfs_root *root, u64 flags)
 {
-	u64 num_devices = root->fs_info->fs_devices->rw_devices;
+	/*
+	 * we add in the count of missing devices because we want
+	 * to make sure that any RAID levels on a degraded FS
+	 * continue to be honored.
+	 */
+	u64 num_devices = root->fs_info->fs_devices->rw_devices +
+		root->fs_info->fs_devices->missing_devices;
+	u64 target;
+	u64 tmp;
 
+	/*
+	 * see if restripe for this chunk_type is in progress, if so
+	 * try to reduce to the target profile
+	 */
+	spin_lock(&root->fs_info->balance_lock);
+	target = get_restripe_target(root->fs_info, flags);
+	if (target) {
+		/* pick target profile only if it's already available */
+		if ((flags & target) & BTRFS_EXTENDED_PROFILE_MASK) {
+			spin_unlock(&root->fs_info->balance_lock);
+			return extended_to_chunk(target);
+		}
+	}
+	spin_unlock(&root->fs_info->balance_lock);
+
+	/* First, mask out the RAID levels which aren't possible */
 	if (num_devices == 1)
-		flags &= ~(BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID0);
+		flags &= ~(BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID0 |
+			   BTRFS_BLOCK_GROUP_RAID5);
+	if (num_devices < 3)
+		flags &= ~BTRFS_BLOCK_GROUP_RAID6;
 	if (num_devices < 4)
 		flags &= ~BTRFS_BLOCK_GROUP_RAID10;
 
-	if ((flags & BTRFS_BLOCK_GROUP_DUP) &&
-	    (flags & (BTRFS_BLOCK_GROUP_RAID1 |
-		      BTRFS_BLOCK_GROUP_RAID10))) {
-		flags &= ~BTRFS_BLOCK_GROUP_DUP;
+	tmp = flags & (BTRFS_BLOCK_GROUP_DUP | BTRFS_BLOCK_GROUP_RAID0 |
+		       BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID5 |
+		       BTRFS_BLOCK_GROUP_RAID6 | BTRFS_BLOCK_GROUP_RAID10);
+	flags &= ~tmp;
+
+	if (tmp & BTRFS_BLOCK_GROUP_RAID6)
+		tmp = BTRFS_BLOCK_GROUP_RAID6;
+	else if (tmp & BTRFS_BLOCK_GROUP_RAID5)
+		tmp = BTRFS_BLOCK_GROUP_RAID5;
+	else if (tmp & BTRFS_BLOCK_GROUP_RAID10)
+		tmp = BTRFS_BLOCK_GROUP_RAID10;
+	else if (tmp & BTRFS_BLOCK_GROUP_RAID1)
+		tmp = BTRFS_BLOCK_GROUP_RAID1;
+	else if (tmp & BTRFS_BLOCK_GROUP_RAID0)
+		tmp = BTRFS_BLOCK_GROUP_RAID0;
+
+	return extended_to_chunk(flags | tmp);
+}
+
+static u64 get_alloc_profile(struct btrfs_root *root, u64 orig_flags)
+{
+	unsigned seq;
+	u64 flags;
+
+	do {
+		flags = orig_flags;
+		seq = read_seqbegin(&root->fs_info->profiles_lock);
+
+		if (flags & BTRFS_BLOCK_GROUP_DATA)
+			flags |= root->fs_info->avail_data_alloc_bits;
+		else if (flags & BTRFS_BLOCK_GROUP_SYSTEM)
+			flags |= root->fs_info->avail_system_alloc_bits;
+		else if (flags & BTRFS_BLOCK_GROUP_METADATA)
+			flags |= root->fs_info->avail_metadata_alloc_bits;
+	} while (read_seqretry(&root->fs_info->profiles_lock, seq));
+
+	return btrfs_reduce_alloc_profile(root, flags);
+}
+
+u64 btrfs_get_alloc_profile(struct btrfs_root *root, int data)
+{
+	u64 flags;
+	u64 ret;
+
+	if (data)
+		flags = BTRFS_BLOCK_GROUP_DATA;
+	else if (root == root->fs_info->chunk_root)
+		flags = BTRFS_BLOCK_GROUP_SYSTEM;
+	else
+		flags = BTRFS_BLOCK_GROUP_METADATA;
+
+	ret = get_alloc_profile(root, flags);
+	return ret;
+}
+
+/*
+ * This will check the space that the inode allocates from to make sure we have
+ * enough space for bytes.
+ */
+int btrfs_check_data_free_space(struct inode *inode, u64 bytes)
+{
+	struct btrfs_space_info *data_sinfo;
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct btrfs_fs_info *fs_info = root->fs_info;
+	u64 used;
+	int ret = 0, committed = 0, alloc_chunk = 1;
+
+	/* make sure bytes are sectorsize aligned */
+	bytes = ALIGN(bytes, root->sectorsize);
+
+	if (btrfs_is_free_space_inode(inode)) {
+		committed = 1;
+		ASSERT(current->journal_info);
 	}
 
-	if ((flags & BTRFS_BLOCK_GROUP_RAID1) &&
-	    (flags & BTRFS_BLOCK_GROUP_RAID10)) {
-		flags &= ~BTRFS_BLOCK_GROUP_RAID1;
+	data_sinfo = fs_info->data_sinfo;
+	if (!data_sinfo)
+		goto alloc;
+
+again:
+	/* make sure we have enough space to handle the data first */
+	spin_lock(&data_sinfo->lock);
+	used = data_sinfo->bytes_used + data_sinfo->bytes_reserved +
+		data_sinfo->bytes_pinned + data_sinfo->bytes_readonly +
+		data_sinfo->bytes_may_use;
+
+	if (used + bytes > data_sinfo->total_bytes) {
+		struct btrfs_trans_handle *trans;
+
+		/*
+		 * if we don't have enough free bytes in this space then we need
+		 * to alloc a new chunk.
+		 */
+		if (!data_sinfo->full && alloc_chunk) {
+			u64 alloc_target;
+
+			data_sinfo->force_alloc = CHUNK_ALLOC_FORCE;
+			spin_unlock(&data_sinfo->lock);
+alloc:
+			alloc_target = btrfs_get_alloc_profile(root, 1);
+			/*
+			 * It is ugly that we don't call nolock join
+			 * transaction for the free space inode case here.
+			 * But it is safe because we only do the data space
+			 * reservation for the free space cache in the
+			 * transaction context, the common join transaction
+			 * just increase the counter of the current transaction
+			 * handler, doesn't try to acquire the trans_lock of
+			 * the fs.
+			 */
+			trans = btrfs_join_transaction(root);
+			if (IS_ERR(trans))
+				return PTR_ERR(trans);
+
+			ret = do_chunk_alloc(trans, root->fs_info->extent_root,
+					     alloc_target,
+					     CHUNK_ALLOC_NO_FORCE);
+			btrfs_end_transaction(trans, root);
+			if (ret < 0) {
+				if (ret != -ENOSPC)
+					return ret;
+				else
+					goto commit_trans;
+			}
+
+			if (!data_sinfo)
+				data_sinfo = fs_info->data_sinfo;
+
+			goto again;
+		}
+
+		/*
+		 * If we don't have enough pinned space to deal with this
+		 * allocation don't bother committing the transaction.
+		 */
+		if (percpu_counter_compare(&data_sinfo->total_bytes_pinned,
+					   bytes) < 0)
+			committed = 1;
+		spin_unlock(&data_sinfo->lock);
+
+		/* commit the current transaction and try again */
+commit_trans:
+		if (!committed &&
+		    !atomic_read(&root->fs_info->open_ioctl_trans)) {
+			committed = 1;
+
+			trans = btrfs_join_transaction(root);
+			if (IS_ERR(trans))
+				return PTR_ERR(trans);
+			ret = btrfs_commit_transaction(trans, root);
+			if (ret)
+				return ret;
+			goto again;
+		}
+
+		trace_btrfs_space_reservation(root->fs_info,
+					      "space_info:enospc",
+					      data_sinfo->flags, bytes, 1);
+		return -ENOSPC;
 	}
+	data_sinfo->bytes_may_use += bytes;
+	trace_btrfs_space_reservation(root->fs_info, "space_info",
+				      data_sinfo->flags, bytes, 1);
+	spin_unlock(&data_sinfo->lock);
 
-	if ((flags & BTRFS_BLOCK_GROUP_RAID0) &&
-	    ((flags & BTRFS_BLOCK_GROUP_RAID1) |
-	     (flags & BTRFS_BLOCK_GROUP_RAID10) |
-	     (flags & BTRFS_BLOCK_GROUP_DUP)))
-		flags &= ~BTRFS_BLOCK_GROUP_RAID0;
-	return flags;
+	return 0;
 }
 
-static u64 btrfs_get_alloc_profile(struct btrfs_root *root, u64 data)
+/*
+ * Called if we need to clear a data reservation for this inode.
+ */
+void btrfs_free_reserved_data_space(struct inode *inode, u64 bytes)
 {
-	struct btrfs_fs_info *info = root->fs_info;
-	u64 alloc_profile;
-
-	if (data) {
-		alloc_profile = info->avail_data_alloc_bits &
-			info->data_alloc_profile;
-		data = BTRFS_BLOCK_GROUP_DATA | alloc_profile;
-	} else if (root == root->fs_info->chunk_root) {
-		alloc_profile = info->avail_system_alloc_bits &
-			info->system_alloc_profile;
-		data = BTRFS_BLOCK_GROUP_SYSTEM | alloc_profile;
-	} else {
-		alloc_profile = info->avail_metadata_alloc_bits &
-			info->metadata_alloc_profile;
-		data = BTRFS_BLOCK_GROUP_METADATA | alloc_profile;
-	}
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct btrfs_space_info *data_sinfo;
 
-	return btrfs_reduce_alloc_profile(root, data);
+	/* make sure bytes are sectorsize aligned */
+	bytes = ALIGN(bytes, root->sectorsize);
+
+	data_sinfo = root->fs_info->data_sinfo;
+	spin_lock(&data_sinfo->lock);
+	WARN_ON(data_sinfo->bytes_may_use < bytes);
+	data_sinfo->bytes_may_use -= bytes;
+	trace_btrfs_space_reservation(root->fs_info, "space_info",
+				      data_sinfo->flags, bytes, 0);
+	spin_unlock(&data_sinfo->lock);
 }
 
-void btrfs_set_inode_space_info(struct btrfs_root *root, struct inode *inode)
+static void force_metadata_allocation(struct btrfs_fs_info *info)
 {
-	u64 alloc_target;
+	struct list_head *head = &info->space_info;
+	struct btrfs_space_info *found;
 
-	alloc_target = btrfs_get_alloc_profile(root, 1);
-	BTRFS_I(inode)->space_info = __find_space_info(root->fs_info,
-						       alloc_target);
+	rcu_read_lock();
+	list_for_each_entry_rcu(found, head, list) {
+		if (found->flags & BTRFS_BLOCK_GROUP_METADATA)
+			found->force_alloc = CHUNK_ALLOC_FORCE;
+	}
+	rcu_read_unlock();
 }
 
-static u64 calculate_bytes_needed(struct btrfs_root *root, int num_items)
+static inline u64 calc_global_rsv_need_space(struct btrfs_block_rsv *global)
 {
-	u64 num_bytes;
-	int level;
+	return (global->size << 1);
+}
+
+static int should_alloc_chunk(struct btrfs_root *root,
+			      struct btrfs_space_info *sinfo, int force)
+{
+	struct btrfs_block_rsv *global_rsv = &root->fs_info->global_block_rsv;
+	u64 num_bytes = sinfo->total_bytes - sinfo->bytes_readonly;
+	u64 num_allocated = sinfo->bytes_used + sinfo->bytes_reserved;
+	u64 thresh;
+
+	if (force == CHUNK_ALLOC_FORCE)
+		return 1;
 
-	level = BTRFS_MAX_LEVEL - 2;
 	/*
-	 * NOTE: these calculations are absolutely the worst possible case.
-	 * This assumes that _every_ item we insert will require a new leaf, and
-	 * that the tree has grown to its maximum level size.
+	 * We need to take into account the global rsv because for all intents
+	 * and purposes it's used space.  Don't worry about locking the
+	 * global_rsv, it doesn't change except when the transaction commits.
 	 */
+	if (sinfo->flags & BTRFS_BLOCK_GROUP_METADATA)
+		num_allocated += calc_global_rsv_need_space(global_rsv);
 
 	/*
-	 * for every item we insert we could insert both an extent item and a
-	 * extent ref item.  Then for ever item we insert, we will need to cow
-	 * both the original leaf, plus the leaf to the left and right of it.
-	 *
-	 * Unless we are talking about the extent root, then we just want the
-	 * number of items * 2, since we just need the extent item plus its ref.
+	 * in limited mode, we want to have some free space up to
+	 * about 1% of the FS size.
 	 */
-	if (root == root->fs_info->extent_root)
-		num_bytes = num_items * 2;
+	if (force == CHUNK_ALLOC_LIMITED) {
+		thresh = btrfs_super_total_bytes(root->fs_info->super_copy);
+		thresh = max_t(u64, 64 * 1024 * 1024,
+			       div_factor_fine(thresh, 1));
+
+		if (num_bytes - num_allocated < thresh)
+			return 1;
+	}
+
+	if (num_allocated + 2 * 1024 * 1024 < div_factor(num_bytes, 8))
+		return 0;
+	return 1;
+}
+
+static u64 get_system_chunk_thresh(struct btrfs_root *root, u64 type)
+{
+	u64 num_dev;
+
+	if (type & (BTRFS_BLOCK_GROUP_RAID10 |
+		    BTRFS_BLOCK_GROUP_RAID0 |
+		    BTRFS_BLOCK_GROUP_RAID5 |
+		    BTRFS_BLOCK_GROUP_RAID6))
+		num_dev = root->fs_info->fs_devices->rw_devices;
+	else if (type & BTRFS_BLOCK_GROUP_RAID1)
+		num_dev = 2;
 	else
-		num_bytes = (num_items + (2 * num_items)) * 3;
+		num_dev = 1;	/* DUP or single */
 
-	/*
-	 * num_bytes is total number of leaves we could need times the leaf
-	 * size, and then for every leaf we could end up cow'ing 2 nodes per
-	 * level, down to the leaf level.
-	 */
-	num_bytes = (num_bytes * root->leafsize) +
-		(num_bytes * (level * 2)) * root->nodesize;
+	/* metadata for updaing devices and chunk tree */
+	return btrfs_calc_trans_metadata_size(root, num_dev + 1);
+}
 
-	return num_bytes;
+static void check_system_chunk(struct btrfs_trans_handle *trans,
+			       struct btrfs_root *root, u64 type)
+{
+	struct btrfs_space_info *info;
+	u64 left;
+	u64 thresh;
+
+	info = __find_space_info(root->fs_info, BTRFS_BLOCK_GROUP_SYSTEM);
+	spin_lock(&info->lock);
+	left = info->total_bytes - info->bytes_used - info->bytes_pinned -
+		info->bytes_reserved - info->bytes_readonly;
+	spin_unlock(&info->lock);
+
+	thresh = get_system_chunk_thresh(root, type);
+	if (left < thresh && btrfs_test_opt(root, ENOSPC_DEBUG)) {
+		btrfs_info(root->fs_info, "left=%llu, need=%llu, flags=%llu",
+			left, thresh, type);
+		dump_space_info(info, 0, 0);
+	}
+
+	if (left < thresh) {
+		u64 flags;
+
+		flags = btrfs_get_alloc_profile(root->fs_info->chunk_root, 0);
+		btrfs_alloc_chunk(trans, root, flags);
+	}
 }
 
-/*
- * Unreserve metadata space for delalloc.  If we have less reserved credits than
- * we have extents, this function does nothing.
- */
-int btrfs_unreserve_metadata_for_delalloc(struct btrfs_root *root,
-					  struct inode *inode, int num_items)
+static int do_chunk_alloc(struct btrfs_trans_handle *trans,
+			  struct btrfs_root *extent_root, u64 flags, int force)
 {
-	struct btrfs_fs_info *info = root->fs_info;
-	struct btrfs_space_info *meta_sinfo;
-	u64 num_bytes;
-	u64 alloc_target;
-	bool bug = false;
-
-	/* get the space info for where the metadata will live */
-	alloc_target = btrfs_get_alloc_profile(root, 0);
-	meta_sinfo = __find_space_info(info, alloc_target);
-
-	num_bytes = calculate_bytes_needed(root->fs_info->extent_root,
-					   num_items);
-
-	spin_lock(&meta_sinfo->lock);
-	spin_lock(&BTRFS_I(inode)->accounting_lock);
-	if (BTRFS_I(inode)->reserved_extents <=
-	    BTRFS_I(inode)->outstanding_extents) {
-		spin_unlock(&BTRFS_I(inode)->accounting_lock);
-		spin_unlock(&meta_sinfo->lock);
-		return 0;
+	struct btrfs_space_info *space_info;
+	struct btrfs_fs_info *fs_info = extent_root->fs_info;
+	int wait_for_alloc = 0;
+	int ret = 0;
+
+	/* Don't re-enter if we're already allocating a chunk */
+	if (trans->allocating_chunk)
+		return -ENOSPC;
+
+	space_info = __find_space_info(extent_root->fs_info, flags);
+	if (!space_info) {
+		ret = update_space_info(extent_root->fs_info, flags,
+					0, 0, &space_info);
+		BUG_ON(ret); /* -ENOMEM */
 	}
-	spin_unlock(&BTRFS_I(inode)->accounting_lock);
+	BUG_ON(!space_info); /* Logic error */
 
-	BTRFS_I(inode)->reserved_extents -= num_items;
-	BUG_ON(BTRFS_I(inode)->reserved_extents < 0);
+again:
+	spin_lock(&space_info->lock);
+	if (force < space_info->force_alloc)
+		force = space_info->force_alloc;
+	if (space_info->full) {
+		if (should_alloc_chunk(extent_root, space_info, force))
+			ret = -ENOSPC;
+		else
+			ret = 0;
+		spin_unlock(&space_info->lock);
+		return ret;
+	}
 
-	if (meta_sinfo->bytes_delalloc < num_bytes) {
-		bug = true;
-		meta_sinfo->bytes_delalloc = 0;
+	if (!should_alloc_chunk(extent_root, space_info, force)) {
+		spin_unlock(&space_info->lock);
+		return 0;
+	} else if (space_info->chunk_alloc) {
+		wait_for_alloc = 1;
 	} else {
-		meta_sinfo->bytes_delalloc -= num_bytes;
+		space_info->chunk_alloc = 1;
 	}
-	spin_unlock(&meta_sinfo->lock);
 
-	BUG_ON(bug);
+	spin_unlock(&space_info->lock);
 
-	return 0;
+	mutex_lock(&fs_info->chunk_mutex);
+
+	/*
+	 * The chunk_mutex is held throughout the entirety of a chunk
+	 * allocation, so once we've acquired the chunk_mutex we know that the
+	 * other guy is done and we need to recheck and see if we should
+	 * allocate.
+	 */
+	if (wait_for_alloc) {
+		mutex_unlock(&fs_info->chunk_mutex);
+		wait_for_alloc = 0;
+		goto again;
+	}
+
+	trans->allocating_chunk = true;
+
+	/*
+	 * If we have mixed data/metadata chunks we want to make sure we keep
+	 * allocating mixed chunks instead of individual chunks.
+	 */
+	if (btrfs_mixed_space_info(space_info))
+		flags |= (BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA);
+
+	/*
+	 * if we're doing a data chunk, go ahead and make sure that
+	 * we keep a reasonable number of metadata chunks allocated in the
+	 * FS as well.
+	 */
+	if (flags & BTRFS_BLOCK_GROUP_DATA && fs_info->metadata_ratio) {
+		fs_info->data_chunk_allocations++;
+		if (!(fs_info->data_chunk_allocations %
+		      fs_info->metadata_ratio))
+			force_metadata_allocation(fs_info);
+	}
+
+	/*
+	 * Check if we have enough space in SYSTEM chunk because we may need
+	 * to update devices.
+	 */
+	check_system_chunk(trans, extent_root, flags);
+
+	ret = btrfs_alloc_chunk(trans, extent_root, flags);
+	trans->allocating_chunk = false;
+
+	spin_lock(&space_info->lock);
+	if (ret < 0 && ret != -ENOSPC)
+		goto out;
+	if (ret)
+		space_info->full = 1;
+	else
+		ret = 1;
+
+	space_info->force_alloc = CHUNK_ALLOC_NO_FORCE;
+out:
+	space_info->chunk_alloc = 0;
+	spin_unlock(&space_info->lock);
+	mutex_unlock(&fs_info->chunk_mutex);
+	return ret;
 }
 
-static void check_force_delalloc(struct btrfs_space_info *meta_sinfo)
+static int can_overcommit(struct btrfs_root *root,
+			  struct btrfs_space_info *space_info, u64 bytes,
+			  enum btrfs_reserve_flush_enum flush)
 {
-	u64 thresh;
+	struct btrfs_block_rsv *global_rsv = &root->fs_info->global_block_rsv;
+	u64 profile = btrfs_get_alloc_profile(root, 0);
+	u64 space_size;
+	u64 avail;
+	u64 used;
+
+	used = space_info->bytes_used + space_info->bytes_reserved +
+		space_info->bytes_pinned + space_info->bytes_readonly;
+
+	/*
+	 * We only want to allow over committing if we have lots of actual space
+	 * free, but if we don't have enough space to handle the global reserve
+	 * space then we could end up having a real enospc problem when trying
+	 * to allocate a chunk or some other such important allocation.
+	 */
+	spin_lock(&global_rsv->lock);
+	space_size = calc_global_rsv_need_space(global_rsv);
+	spin_unlock(&global_rsv->lock);
+	if (used + space_size >= space_info->total_bytes)
+		return 0;
+
+	used += space_info->bytes_may_use;
 
-	thresh = meta_sinfo->bytes_used + meta_sinfo->bytes_reserved +
-		meta_sinfo->bytes_pinned + meta_sinfo->bytes_readonly +
-		meta_sinfo->bytes_super + meta_sinfo->bytes_root +
-		meta_sinfo->bytes_may_use;
+	spin_lock(&root->fs_info->free_chunk_lock);
+	avail = root->fs_info->free_chunk_space;
+	spin_unlock(&root->fs_info->free_chunk_lock);
 
-	thresh = meta_sinfo->total_bytes - thresh;
-	thresh *= 80;
-	do_div(thresh, 100);
-	if (thresh <= meta_sinfo->bytes_delalloc)
-		meta_sinfo->force_delalloc = 1;
+	/*
+	 * If we have dup, raid1 or raid10 then only half of the free
+	 * space is actually useable.  For raid56, the space info used
+	 * doesn't include the parity drive, so we don't have to
+	 * change the math
+	 */
+	if (profile & (BTRFS_BLOCK_GROUP_DUP |
+		       BTRFS_BLOCK_GROUP_RAID1 |
+		       BTRFS_BLOCK_GROUP_RAID10))
+		avail >>= 1;
+
+	/*
+	 * If we aren't flushing all things, let us overcommit up to
+	 * 1/2th of the space. If we can flush, don't let us overcommit
+	 * too much, let it overcommit up to 1/8 of the space.
+	 */
+	if (flush == BTRFS_RESERVE_FLUSH_ALL)
+		avail >>= 3;
 	else
-		meta_sinfo->force_delalloc = 0;
+		avail >>= 1;
+
+	if (used + bytes < space_info->total_bytes + avail)
+		return 1;
+	return 0;
 }
 
-struct async_flush {
-	struct btrfs_root *root;
-	struct btrfs_space_info *info;
-	struct btrfs_work work;
-};
+static void btrfs_writeback_inodes_sb_nr(struct btrfs_root *root,
+					 unsigned long nr_pages, int nr_items)
+{
+	struct super_block *sb = root->fs_info->sb;
+
+	if (down_read_trylock(&sb->s_umount)) {
+		writeback_inodes_sb_nr(sb, nr_pages, WB_REASON_FS_FREE_SPACE);
+		up_read(&sb->s_umount);
+	} else {
+		/*
+		 * We needn't worry the filesystem going from r/w to r/o though
+		 * we don't acquire ->s_umount mutex, because the filesystem
+		 * should guarantee the delalloc inodes list be empty after
+		 * the filesystem is readonly(all dirty pages are written to
+		 * the disk).
+		 */
+		btrfs_start_delalloc_roots(root->fs_info, 0, nr_items);
+		if (!current->journal_info)
+			btrfs_wait_ordered_roots(root->fs_info, nr_items);
+	}
+}
 
-static noinline void flush_delalloc_async(struct btrfs_work *work)
+static inline int calc_reclaim_items_nr(struct btrfs_root *root, u64 to_reclaim)
 {
-	struct async_flush *async;
-	struct btrfs_root *root;
-	struct btrfs_space_info *info;
+	u64 bytes;
+	int nr;
 
-	async = container_of(work, struct async_flush, work);
-	root = async->root;
-	info = async->info;
+	bytes = btrfs_calc_trans_metadata_size(root, 1);
+	nr = (int)div64_u64(to_reclaim, bytes);
+	if (!nr)
+		nr = 1;
+	return nr;
+}
 
-	btrfs_start_delalloc_inodes(root, 0);
-	wake_up(&info->flush_wait);
-	btrfs_wait_ordered_extents(root, 0, 0);
+#define EXTENT_SIZE_PER_ITEM	(256 * 1024)
 
-	spin_lock(&info->lock);
-	info->flushing = 0;
-	spin_unlock(&info->lock);
-	wake_up(&info->flush_wait);
+/*
+ * shrink metadata reservation for delalloc
+ */
+static void shrink_delalloc(struct btrfs_root *root, u64 to_reclaim, u64 orig,
+			    bool wait_ordered)
+{
+	struct btrfs_block_rsv *block_rsv;
+	struct btrfs_space_info *space_info;
+	struct btrfs_trans_handle *trans;
+	u64 delalloc_bytes;
+	u64 max_reclaim;
+	long time_left;
+	unsigned long nr_pages;
+	int loops;
+	int items;
+	enum btrfs_reserve_flush_enum flush;
+
+	/* Calc the number of the pages we need flush for space reservation */
+	items = calc_reclaim_items_nr(root, to_reclaim);
+	to_reclaim = items * EXTENT_SIZE_PER_ITEM;
+
+	trans = (struct btrfs_trans_handle *)current->journal_info;
+	block_rsv = &root->fs_info->delalloc_block_rsv;
+	space_info = block_rsv->space_info;
+
+	delalloc_bytes = percpu_counter_sum_positive(
+						&root->fs_info->delalloc_bytes);
+	if (delalloc_bytes == 0) {
+		if (trans)
+			return;
+		if (wait_ordered)
+			btrfs_wait_ordered_roots(root->fs_info, items);
+		return;
+	}
+
+	loops = 0;
+	while (delalloc_bytes && loops < 3) {
+		max_reclaim = min(delalloc_bytes, to_reclaim);
+		nr_pages = max_reclaim >> PAGE_CACHE_SHIFT;
+		btrfs_writeback_inodes_sb_nr(root, nr_pages, items);
+		/*
+		 * We need to wait for the async pages to actually start before
+		 * we do anything.
+		 */
+		max_reclaim = atomic_read(&root->fs_info->async_delalloc_pages);
+		if (!max_reclaim)
+			goto skip_async;
 
-	kfree(async);
+		if (max_reclaim <= nr_pages)
+			max_reclaim = 0;
+		else
+			max_reclaim -= nr_pages;
+
+		wait_event(root->fs_info->async_submit_wait,
+			   atomic_read(&root->fs_info->async_delalloc_pages) <=
+			   (int)max_reclaim);
+skip_async:
+		if (!trans)
+			flush = BTRFS_RESERVE_FLUSH_ALL;
+		else
+			flush = BTRFS_RESERVE_NO_FLUSH;
+		spin_lock(&space_info->lock);
+		if (can_overcommit(root, space_info, orig, flush)) {
+			spin_unlock(&space_info->lock);
+			break;
+		}
+		spin_unlock(&space_info->lock);
+
+		loops++;
+		if (wait_ordered && !trans) {
+			btrfs_wait_ordered_roots(root->fs_info, items);
+		} else {
+			time_left = schedule_timeout_killable(1);
+			if (time_left)
+				break;
+		}
+		delalloc_bytes = percpu_counter_sum_positive(
+						&root->fs_info->delalloc_bytes);
+	}
 }
 
-static void wait_on_flush(struct btrfs_space_info *info)
+/**
+ * maybe_commit_transaction - possibly commit the transaction if its ok to
+ * @root - the root we're allocating for
+ * @bytes - the number of bytes we want to reserve
+ * @force - force the commit
+ *
+ * This will check to make sure that committing the transaction will actually
+ * get us somewhere and then commit the transaction if it does.  Otherwise it
+ * will return -ENOSPC.
+ */
+static int may_commit_transaction(struct btrfs_root *root,
+				  struct btrfs_space_info *space_info,
+				  u64 bytes, int force)
 {
-	DEFINE_WAIT(wait);
-	u64 used;
+	struct btrfs_block_rsv *delayed_rsv = &root->fs_info->delayed_block_rsv;
+	struct btrfs_trans_handle *trans;
 
-	while (1) {
-		prepare_to_wait(&info->flush_wait, &wait,
-				TASK_UNINTERRUPTIBLE);
-		spin_lock(&info->lock);
-		if (!info->flushing) {
-			spin_unlock(&info->lock);
+	trans = (struct btrfs_trans_handle *)current->journal_info;
+	if (trans)
+		return -EAGAIN;
+
+	if (force)
+		goto commit;
+
+	/* See if there is enough pinned space to make this reservation */
+	if (percpu_counter_compare(&space_info->total_bytes_pinned,
+				   bytes) >= 0)
+		goto commit;
+
+	/*
+	 * See if there is some space in the delayed insertion reservation for
+	 * this reservation.
+	 */
+	if (space_info != delayed_rsv->space_info)
+		return -ENOSPC;
+
+	spin_lock(&delayed_rsv->lock);
+	if (percpu_counter_compare(&space_info->total_bytes_pinned,
+				   bytes - delayed_rsv->size) >= 0) {
+		spin_unlock(&delayed_rsv->lock);
+		return -ENOSPC;
+	}
+	spin_unlock(&delayed_rsv->lock);
+
+commit:
+	trans = btrfs_join_transaction(root);
+	if (IS_ERR(trans))
+		return -ENOSPC;
+
+	return btrfs_commit_transaction(trans, root);
+}
+
+enum flush_state {
+	FLUSH_DELAYED_ITEMS_NR	=	1,
+	FLUSH_DELAYED_ITEMS	=	2,
+	FLUSH_DELALLOC		=	3,
+	FLUSH_DELALLOC_WAIT	=	4,
+	ALLOC_CHUNK		=	5,
+	COMMIT_TRANS		=	6,
+};
+
+static int flush_space(struct btrfs_root *root,
+		       struct btrfs_space_info *space_info, u64 num_bytes,
+		       u64 orig_bytes, int state)
+{
+	struct btrfs_trans_handle *trans;
+	int nr;
+	int ret = 0;
+
+	switch (state) {
+	case FLUSH_DELAYED_ITEMS_NR:
+	case FLUSH_DELAYED_ITEMS:
+		if (state == FLUSH_DELAYED_ITEMS_NR)
+			nr = calc_reclaim_items_nr(root, num_bytes) * 2;
+		else
+			nr = -1;
+
+		trans = btrfs_join_transaction(root);
+		if (IS_ERR(trans)) {
+			ret = PTR_ERR(trans);
 			break;
 		}
-
-		used = info->bytes_used + info->bytes_reserved +
-			info->bytes_pinned + info->bytes_readonly +
-			info->bytes_super + info->bytes_root +
-			info->bytes_may_use + info->bytes_delalloc;
-		if (used < info->total_bytes) {
-			spin_unlock(&info->lock);
+		ret = btrfs_run_delayed_items_nr(trans, root, nr);
+		btrfs_end_transaction(trans, root);
+		break;
+	case FLUSH_DELALLOC:
+	case FLUSH_DELALLOC_WAIT:
+		shrink_delalloc(root, num_bytes * 2, orig_bytes,
+				state == FLUSH_DELALLOC_WAIT);
+		break;
+	case ALLOC_CHUNK:
+		trans = btrfs_join_transaction(root);
+		if (IS_ERR(trans)) {
+			ret = PTR_ERR(trans);
 			break;
 		}
-		spin_unlock(&info->lock);
-		schedule();
+		ret = do_chunk_alloc(trans, root->fs_info->extent_root,
+				     btrfs_get_alloc_profile(root, 0),
+				     CHUNK_ALLOC_NO_FORCE);
+		btrfs_end_transaction(trans, root);
+		if (ret == -ENOSPC)
+			ret = 0;
+		break;
+	case COMMIT_TRANS:
+		ret = may_commit_transaction(root, space_info, orig_bytes, 0);
+		break;
+	default:
+		ret = -ENOSPC;
+		break;
 	}
-	finish_wait(&info->flush_wait, &wait);
+
+	return ret;
 }
 
-static void flush_delalloc(struct btrfs_root *root,
-				 struct btrfs_space_info *info)
+static inline u64
+btrfs_calc_reclaim_metadata_size(struct btrfs_root *root,
+				 struct btrfs_space_info *space_info)
 {
-	struct async_flush *async;
-	bool wait = false;
+	u64 used;
+	u64 expected;
+	u64 to_reclaim;
 
-	spin_lock(&info->lock);
+	to_reclaim = min_t(u64, num_online_cpus() * 1024 * 1024,
+				16 * 1024 * 1024);
+	spin_lock(&space_info->lock);
+	if (can_overcommit(root, space_info, to_reclaim,
+			   BTRFS_RESERVE_FLUSH_ALL)) {
+		to_reclaim = 0;
+		goto out;
+	}
 
-	if (!info->flushing)
-		info->flushing = 1;
+	used = space_info->bytes_used + space_info->bytes_reserved +
+	       space_info->bytes_pinned + space_info->bytes_readonly +
+	       space_info->bytes_may_use;
+	if (can_overcommit(root, space_info, 1024 * 1024,
+			   BTRFS_RESERVE_FLUSH_ALL))
+		expected = div_factor_fine(space_info->total_bytes, 95);
 	else
-		wait = true;
+		expected = div_factor_fine(space_info->total_bytes, 90);
 
-	spin_unlock(&info->lock);
+	if (used > expected)
+		to_reclaim = used - expected;
+	else
+		to_reclaim = 0;
+	to_reclaim = min(to_reclaim, space_info->bytes_may_use +
+				     space_info->bytes_reserved);
+out:
+	spin_unlock(&space_info->lock);
 
-	if (wait) {
-		wait_on_flush(info);
-		return;
+	return to_reclaim;
+}
+
+static inline int need_do_async_reclaim(struct btrfs_space_info *space_info,
+					struct btrfs_fs_info *fs_info, u64 used)
+{
+	return (used >= div_factor_fine(space_info->total_bytes, 98) &&
+		!btrfs_fs_closing(fs_info) &&
+		!test_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state));
+}
+
+static int btrfs_need_do_async_reclaim(struct btrfs_space_info *space_info,
+				       struct btrfs_fs_info *fs_info)
+{
+	u64 used;
+
+	spin_lock(&space_info->lock);
+	used = space_info->bytes_used + space_info->bytes_reserved +
+	       space_info->bytes_pinned + space_info->bytes_readonly +
+	       space_info->bytes_may_use;
+	if (need_do_async_reclaim(space_info, fs_info, used)) {
+		spin_unlock(&space_info->lock);
+		return 1;
 	}
+	spin_unlock(&space_info->lock);
 
-	async = kzalloc(sizeof(*async), GFP_NOFS);
-	if (!async)
-		goto flush;
+	return 0;
+}
 
-	async->root = root;
-	async->info = info;
-	async->work.func = flush_delalloc_async;
+static void btrfs_async_reclaim_metadata_space(struct work_struct *work)
+{
+	struct btrfs_fs_info *fs_info;
+	struct btrfs_space_info *space_info;
+	u64 to_reclaim;
+	int flush_state;
 
-	btrfs_queue_worker(&root->fs_info->enospc_workers,
-			   &async->work);
-	wait_on_flush(info);
-	return;
+	fs_info = container_of(work, struct btrfs_fs_info, async_reclaim_work);
+	space_info = __find_space_info(fs_info, BTRFS_BLOCK_GROUP_METADATA);
 
-flush:
-	btrfs_start_delalloc_inodes(root, 0);
-	btrfs_wait_ordered_extents(root, 0, 0);
+	to_reclaim = btrfs_calc_reclaim_metadata_size(fs_info->fs_root,
+						      space_info);
+	if (!to_reclaim)
+		return;
 
-	spin_lock(&info->lock);
-	info->flushing = 0;
-	spin_unlock(&info->lock);
-	wake_up(&info->flush_wait);
+	flush_state = FLUSH_DELAYED_ITEMS_NR;
+	do {
+		flush_space(fs_info->fs_root, space_info, to_reclaim,
+			    to_reclaim, flush_state);
+		flush_state++;
+		if (!btrfs_need_do_async_reclaim(space_info, fs_info))
+			return;
+	} while (flush_state <= COMMIT_TRANS);
+
+	if (btrfs_need_do_async_reclaim(space_info, fs_info))
+		queue_work(system_unbound_wq, work);
 }
 
-static int maybe_allocate_chunk(struct btrfs_root *root,
-				 struct btrfs_space_info *info)
+void btrfs_init_async_reclaim_work(struct work_struct *work)
 {
-	struct btrfs_super_block *disk_super = &root->fs_info->super_copy;
-	struct btrfs_trans_handle *trans;
-	bool wait = false;
+	INIT_WORK(work, btrfs_async_reclaim_metadata_space);
+}
+
+/**
+ * reserve_metadata_bytes - try to reserve bytes from the block_rsv's space
+ * @root - the root we're allocating for
+ * @block_rsv - the block_rsv we're allocating for
+ * @orig_bytes - the number of bytes we want
+ * @flush - whether or not we can flush to make our reservation
+ *
+ * This will reserve orgi_bytes number of bytes from the space info associated
+ * with the block_rsv.  If there is not enough space it will make an attempt to
+ * flush out space to make room.  It will do this by flushing delalloc if
+ * possible or committing the transaction.  If flush is 0 then no attempts to
+ * regain reservations will be made and this will fail if there is not enough
+ * space already.
+ */
+static int reserve_metadata_bytes(struct btrfs_root *root,
+				  struct btrfs_block_rsv *block_rsv,
+				  u64 orig_bytes,
+				  enum btrfs_reserve_flush_enum flush)
+{
+	struct btrfs_space_info *space_info = block_rsv->space_info;
+	u64 used;
+	u64 num_bytes = orig_bytes;
+	int flush_state = FLUSH_DELAYED_ITEMS_NR;
 	int ret = 0;
-	u64 min_metadata;
-	u64 free_space;
+	bool flushing = false;
 
-	free_space = btrfs_super_total_bytes(disk_super);
+again:
+	ret = 0;
+	spin_lock(&space_info->lock);
 	/*
-	 * we allow the metadata to grow to a max of either 10gb or 5% of the
-	 * space in the volume.
+	 * We only want to wait if somebody other than us is flushing and we
+	 * are actually allowed to flush all things.
 	 */
-	min_metadata = min((u64)10 * 1024 * 1024 * 1024,
-			     div64_u64(free_space * 5, 100));
-	if (info->total_bytes >= min_metadata) {
-		spin_unlock(&info->lock);
-		return 0;
-	}
+	while (flush == BTRFS_RESERVE_FLUSH_ALL && !flushing &&
+	       space_info->flush) {
+		spin_unlock(&space_info->lock);
+		/*
+		 * If we have a trans handle we can't wait because the flusher
+		 * may have to commit the transaction, which would mean we would
+		 * deadlock since we are waiting for the flusher to finish, but
+		 * hold the current transaction open.
+		 */
+		if (current->journal_info)
+			return -EAGAIN;
+		ret = wait_event_killable(space_info->wait, !space_info->flush);
+		/* Must have been killed, return */
+		if (ret)
+			return -EINTR;
 
-	if (info->full) {
-		spin_unlock(&info->lock);
-		return 0;
+		spin_lock(&space_info->lock);
 	}
 
-	if (!info->allocating_chunk) {
-		info->force_alloc = 1;
-		info->allocating_chunk = 1;
+	ret = -ENOSPC;
+	used = space_info->bytes_used + space_info->bytes_reserved +
+		space_info->bytes_pinned + space_info->bytes_readonly +
+		space_info->bytes_may_use;
+
+	/*
+	 * The idea here is that we've not already over-reserved the block group
+	 * then we can go ahead and save our reservation first and then start
+	 * flushing if we need to.  Otherwise if we've already overcommitted
+	 * lets start flushing stuff first and then come back and try to make
+	 * our reservation.
+	 */
+	if (used <= space_info->total_bytes) {
+		if (used + orig_bytes <= space_info->total_bytes) {
+			space_info->bytes_may_use += orig_bytes;
+			trace_btrfs_space_reservation(root->fs_info,
+				"space_info", space_info->flags, orig_bytes, 1);
+			ret = 0;
+		} else {
+			/*
+			 * Ok set num_bytes to orig_bytes since we aren't
+			 * overocmmitted, this way we only try and reclaim what
+			 * we need.
+			 */
+			num_bytes = orig_bytes;
+		}
 	} else {
-		wait = true;
+		/*
+		 * Ok we're over committed, set num_bytes to the overcommitted
+		 * amount plus the amount of bytes that we need for this
+		 * reservation.
+		 */
+		num_bytes = used - space_info->total_bytes +
+			(orig_bytes * 2);
 	}
 
-	spin_unlock(&info->lock);
-
-	if (wait) {
-		wait_event(info->allocate_wait,
-			   !info->allocating_chunk);
-		return 1;
+	if (ret && can_overcommit(root, space_info, orig_bytes, flush)) {
+		space_info->bytes_may_use += orig_bytes;
+		trace_btrfs_space_reservation(root->fs_info, "space_info",
+					      space_info->flags, orig_bytes,
+					      1);
+		ret = 0;
 	}
 
-	trans = btrfs_start_transaction(root, 1);
-	if (!trans) {
-		ret = -ENOMEM;
-		goto out;
+	/*
+	 * Couldn't make our reservation, save our place so while we're trying
+	 * to reclaim space we can actually use it instead of somebody else
+	 * stealing it from us.
+	 *
+	 * We make the other tasks wait for the flush only when we can flush
+	 * all things.
+	 */
+	if (ret && flush != BTRFS_RESERVE_NO_FLUSH) {
+		flushing = true;
+		space_info->flush = 1;
+	} else if (!ret && space_info->flags & BTRFS_BLOCK_GROUP_METADATA) {
+		used += orig_bytes;
+		if (need_do_async_reclaim(space_info, root->fs_info, used) &&
+		    !work_busy(&root->fs_info->async_reclaim_work))
+			queue_work(system_unbound_wq,
+				   &root->fs_info->async_reclaim_work);
 	}
+	spin_unlock(&space_info->lock);
 
-	ret = do_chunk_alloc(trans, root->fs_info->extent_root,
-			     4096 + 2 * 1024 * 1024,
-			     info->flags, 0);
-	btrfs_end_transaction(trans, root);
-	if (ret)
+	if (!ret || flush == BTRFS_RESERVE_NO_FLUSH)
 		goto out;
+
+	ret = flush_space(root, space_info, num_bytes, orig_bytes,
+			  flush_state);
+	flush_state++;
+
+	/*
+	 * If we are FLUSH_LIMIT, we can not flush delalloc, or the deadlock
+	 * would happen. So skip delalloc flush.
+	 */
+	if (flush == BTRFS_RESERVE_FLUSH_LIMIT &&
+	    (flush_state == FLUSH_DELALLOC ||
+	     flush_state == FLUSH_DELALLOC_WAIT))
+		flush_state = ALLOC_CHUNK;
+
+	if (!ret)
+		goto again;
+	else if (flush == BTRFS_RESERVE_FLUSH_LIMIT &&
+		 flush_state < COMMIT_TRANS)
+		goto again;
+	else if (flush == BTRFS_RESERVE_FLUSH_ALL &&
+		 flush_state <= COMMIT_TRANS)
+		goto again;
+
 out:
-	spin_lock(&info->lock);
-	info->allocating_chunk = 0;
-	spin_unlock(&info->lock);
-	wake_up(&info->allocate_wait);
+	if (ret == -ENOSPC &&
+	    unlikely(root->orphan_cleanup_state == ORPHAN_CLEANUP_STARTED)) {
+		struct btrfs_block_rsv *global_rsv =
+			&root->fs_info->global_block_rsv;
 
-	if (ret)
-		return 0;
-	return 1;
+		if (block_rsv != global_rsv &&
+		    !block_rsv_use_bytes(global_rsv, orig_bytes))
+			ret = 0;
+	}
+	if (ret == -ENOSPC)
+		trace_btrfs_space_reservation(root->fs_info,
+					      "space_info:enospc",
+					      space_info->flags, orig_bytes, 1);
+	if (flushing) {
+		spin_lock(&space_info->lock);
+		space_info->flush = 0;
+		wake_up_all(&space_info->wait);
+		spin_unlock(&space_info->lock);
+	}
+	return ret;
 }
 
-/*
- * Reserve metadata space for delalloc.
- */
-int btrfs_reserve_metadata_for_delalloc(struct btrfs_root *root,
-					struct inode *inode, int num_items)
+static struct btrfs_block_rsv *get_block_rsv(
+					const struct btrfs_trans_handle *trans,
+					const struct btrfs_root *root)
 {
-	struct btrfs_fs_info *info = root->fs_info;
-	struct btrfs_space_info *meta_sinfo;
-	u64 num_bytes;
-	u64 used;
-	u64 alloc_target;
-	int flushed = 0;
-	int force_delalloc;
+	struct btrfs_block_rsv *block_rsv = NULL;
 
-	/* get the space info for where the metadata will live */
-	alloc_target = btrfs_get_alloc_profile(root, 0);
-	meta_sinfo = __find_space_info(info, alloc_target);
+	if (test_bit(BTRFS_ROOT_REF_COWS, &root->state))
+		block_rsv = trans->block_rsv;
 
-	num_bytes = calculate_bytes_needed(root->fs_info->extent_root,
-					   num_items);
-again:
-	spin_lock(&meta_sinfo->lock);
+	if (root == root->fs_info->csum_root && trans->adding_csums)
+		block_rsv = trans->block_rsv;
 
-	force_delalloc = meta_sinfo->force_delalloc;
+	if (root == root->fs_info->uuid_root)
+		block_rsv = trans->block_rsv;
 
-	if (unlikely(!meta_sinfo->bytes_root))
-		meta_sinfo->bytes_root = calculate_bytes_needed(root, 6);
+	if (!block_rsv)
+		block_rsv = root->block_rsv;
 
-	if (!flushed)
-		meta_sinfo->bytes_delalloc += num_bytes;
+	if (!block_rsv)
+		block_rsv = &root->fs_info->empty_block_rsv;
 
-	used = meta_sinfo->bytes_used + meta_sinfo->bytes_reserved +
-		meta_sinfo->bytes_pinned + meta_sinfo->bytes_readonly +
-		meta_sinfo->bytes_super + meta_sinfo->bytes_root +
-		meta_sinfo->bytes_may_use + meta_sinfo->bytes_delalloc;
+	return block_rsv;
+}
 
-	if (used > meta_sinfo->total_bytes) {
-		flushed++;
+static int block_rsv_use_bytes(struct btrfs_block_rsv *block_rsv,
+			       u64 num_bytes)
+{
+	int ret = -ENOSPC;
+	spin_lock(&block_rsv->lock);
+	if (block_rsv->reserved >= num_bytes) {
+		block_rsv->reserved -= num_bytes;
+		if (block_rsv->reserved < block_rsv->size)
+			block_rsv->full = 0;
+		ret = 0;
+	}
+	spin_unlock(&block_rsv->lock);
+	return ret;
+}
 
-		if (flushed == 1) {
-			if (maybe_allocate_chunk(root, meta_sinfo))
-				goto again;
-			flushed++;
-		} else {
-			spin_unlock(&meta_sinfo->lock);
-		}
+static void block_rsv_add_bytes(struct btrfs_block_rsv *block_rsv,
+				u64 num_bytes, int update_size)
+{
+	spin_lock(&block_rsv->lock);
+	block_rsv->reserved += num_bytes;
+	if (update_size)
+		block_rsv->size += num_bytes;
+	else if (block_rsv->reserved >= block_rsv->size)
+		block_rsv->full = 1;
+	spin_unlock(&block_rsv->lock);
+}
 
-		if (flushed == 2) {
-			filemap_flush(inode->i_mapping);
-			goto again;
-		} else if (flushed == 3) {
-			flush_delalloc(root, meta_sinfo);
-			goto again;
-		}
-		spin_lock(&meta_sinfo->lock);
-		meta_sinfo->bytes_delalloc -= num_bytes;
-		spin_unlock(&meta_sinfo->lock);
-		printk(KERN_ERR "enospc, has %d, reserved %d\n",
-		       BTRFS_I(inode)->outstanding_extents,
-		       BTRFS_I(inode)->reserved_extents);
-		dump_space_info(meta_sinfo, 0, 0);
+int btrfs_cond_migrate_bytes(struct btrfs_fs_info *fs_info,
+			     struct btrfs_block_rsv *dest, u64 num_bytes,
+			     int min_factor)
+{
+	struct btrfs_block_rsv *global_rsv = &fs_info->global_block_rsv;
+	u64 min_bytes;
+
+	if (global_rsv->space_info != dest->space_info)
+		return -ENOSPC;
+
+	spin_lock(&global_rsv->lock);
+	min_bytes = div_factor(global_rsv->size, min_factor);
+	if (global_rsv->reserved < min_bytes + num_bytes) {
+		spin_unlock(&global_rsv->lock);
 		return -ENOSPC;
 	}
+	global_rsv->reserved -= num_bytes;
+	if (global_rsv->reserved < global_rsv->size)
+		global_rsv->full = 0;
+	spin_unlock(&global_rsv->lock);
+
+	block_rsv_add_bytes(dest, num_bytes, 1);
+	return 0;
+}
 
-	BTRFS_I(inode)->reserved_extents += num_items;
-	check_force_delalloc(meta_sinfo);
-	spin_unlock(&meta_sinfo->lock);
+static void block_rsv_release_bytes(struct btrfs_fs_info *fs_info,
+				    struct btrfs_block_rsv *block_rsv,
+				    struct btrfs_block_rsv *dest, u64 num_bytes)
+{
+	struct btrfs_space_info *space_info = block_rsv->space_info;
 
-	if (!flushed && force_delalloc)
-		filemap_flush(inode->i_mapping);
+	spin_lock(&block_rsv->lock);
+	if (num_bytes == (u64)-1)
+		num_bytes = block_rsv->size;
+	block_rsv->size -= num_bytes;
+	if (block_rsv->reserved >= block_rsv->size) {
+		num_bytes = block_rsv->reserved - block_rsv->size;
+		block_rsv->reserved = block_rsv->size;
+		block_rsv->full = 1;
+	} else {
+		num_bytes = 0;
+	}
+	spin_unlock(&block_rsv->lock);
+
+	if (num_bytes > 0) {
+		if (dest) {
+			spin_lock(&dest->lock);
+			if (!dest->full) {
+				u64 bytes_to_add;
+
+				bytes_to_add = dest->size - dest->reserved;
+				bytes_to_add = min(num_bytes, bytes_to_add);
+				dest->reserved += bytes_to_add;
+				if (dest->reserved >= dest->size)
+					dest->full = 1;
+				num_bytes -= bytes_to_add;
+			}
+			spin_unlock(&dest->lock);
+		}
+		if (num_bytes) {
+			spin_lock(&space_info->lock);
+			space_info->bytes_may_use -= num_bytes;
+			trace_btrfs_space_reservation(fs_info, "space_info",
+					space_info->flags, num_bytes, 0);
+			spin_unlock(&space_info->lock);
+		}
+	}
+}
+
+static int block_rsv_migrate_bytes(struct btrfs_block_rsv *src,
+				   struct btrfs_block_rsv *dst, u64 num_bytes)
+{
+	int ret;
+
+	ret = block_rsv_use_bytes(src, num_bytes);
+	if (ret)
+		return ret;
 
+	block_rsv_add_bytes(dst, num_bytes, 1);
 	return 0;
 }
 
-/*
- * unreserve num_items number of items worth of metadata space.  This needs to
- * be paired with btrfs_reserve_metadata_space.
- *
- * NOTE: if you have the option, run this _AFTER_ you do a
- * btrfs_end_transaction, since btrfs_end_transaction will run delayed ref
- * oprations which will result in more used metadata, so we want to make sure we
- * can do that without issue.
- */
-int btrfs_unreserve_metadata_space(struct btrfs_root *root, int num_items)
+void btrfs_init_block_rsv(struct btrfs_block_rsv *rsv, unsigned short type)
 {
-	struct btrfs_fs_info *info = root->fs_info;
-	struct btrfs_space_info *meta_sinfo;
-	u64 num_bytes;
-	u64 alloc_target;
-	bool bug = false;
+	memset(rsv, 0, sizeof(*rsv));
+	spin_lock_init(&rsv->lock);
+	rsv->type = type;
+}
 
-	/* get the space info for where the metadata will live */
-	alloc_target = btrfs_get_alloc_profile(root, 0);
-	meta_sinfo = __find_space_info(info, alloc_target);
+struct btrfs_block_rsv *btrfs_alloc_block_rsv(struct btrfs_root *root,
+					      unsigned short type)
+{
+	struct btrfs_block_rsv *block_rsv;
+	struct btrfs_fs_info *fs_info = root->fs_info;
 
-	num_bytes = calculate_bytes_needed(root, num_items);
+	block_rsv = kmalloc(sizeof(*block_rsv), GFP_NOFS);
+	if (!block_rsv)
+		return NULL;
 
-	spin_lock(&meta_sinfo->lock);
-	if (meta_sinfo->bytes_may_use < num_bytes) {
-		bug = true;
-		meta_sinfo->bytes_may_use = 0;
-	} else {
-		meta_sinfo->bytes_may_use -= num_bytes;
-	}
-	spin_unlock(&meta_sinfo->lock);
+	btrfs_init_block_rsv(block_rsv, type);
+	block_rsv->space_info = __find_space_info(fs_info,
+						  BTRFS_BLOCK_GROUP_METADATA);
+	return block_rsv;
+}
 
-	BUG_ON(bug);
+void btrfs_free_block_rsv(struct btrfs_root *root,
+			  struct btrfs_block_rsv *rsv)
+{
+	if (!rsv)
+		return;
+	btrfs_block_rsv_release(root, rsv, (u64)-1);
+	kfree(rsv);
+}
 
-	return 0;
+int btrfs_block_rsv_add(struct btrfs_root *root,
+			struct btrfs_block_rsv *block_rsv, u64 num_bytes,
+			enum btrfs_reserve_flush_enum flush)
+{
+	int ret;
+
+	if (num_bytes == 0)
+		return 0;
+
+	ret = reserve_metadata_bytes(root, block_rsv, num_bytes, flush);
+	if (!ret) {
+		block_rsv_add_bytes(block_rsv, num_bytes, 1);
+		return 0;
+	}
+
+	return ret;
 }
 
-/*
- * Reserve some metadata space for use.  We'll calculate the worste case number
- * of bytes that would be needed to modify num_items number of items.  If we
- * have space, fantastic, if not, you get -ENOSPC.  Please call
- * btrfs_unreserve_metadata_space when you are done for the _SAME_ number of
- * items you reserved, since whatever metadata you needed should have already
- * been allocated.
- *
- * This will commit the transaction to make more space if we don't have enough
- * metadata space.  THe only time we don't do this is if we're reserving space
- * inside of a transaction, then we will just return -ENOSPC and it is the
- * callers responsibility to handle it properly.
- */
-int btrfs_reserve_metadata_space(struct btrfs_root *root, int num_items)
+int btrfs_block_rsv_check(struct btrfs_root *root,
+			  struct btrfs_block_rsv *block_rsv, int min_factor)
 {
-	struct btrfs_fs_info *info = root->fs_info;
-	struct btrfs_space_info *meta_sinfo;
-	u64 num_bytes;
-	u64 used;
-	u64 alloc_target;
-	int retries = 0;
+	u64 num_bytes = 0;
+	int ret = -ENOSPC;
 
-	/* get the space info for where the metadata will live */
-	alloc_target = btrfs_get_alloc_profile(root, 0);
-	meta_sinfo = __find_space_info(info, alloc_target);
+	if (!block_rsv)
+		return 0;
 
-	num_bytes = calculate_bytes_needed(root, num_items);
-again:
-	spin_lock(&meta_sinfo->lock);
+	spin_lock(&block_rsv->lock);
+	num_bytes = div_factor(block_rsv->size, min_factor);
+	if (block_rsv->reserved >= num_bytes)
+		ret = 0;
+	spin_unlock(&block_rsv->lock);
 
-	if (unlikely(!meta_sinfo->bytes_root))
-		meta_sinfo->bytes_root = calculate_bytes_needed(root, 6);
+	return ret;
+}
 
-	if (!retries)
-		meta_sinfo->bytes_may_use += num_bytes;
+int btrfs_block_rsv_refill(struct btrfs_root *root,
+			   struct btrfs_block_rsv *block_rsv, u64 min_reserved,
+			   enum btrfs_reserve_flush_enum flush)
+{
+	u64 num_bytes = 0;
+	int ret = -ENOSPC;
 
-	used = meta_sinfo->bytes_used + meta_sinfo->bytes_reserved +
-		meta_sinfo->bytes_pinned + meta_sinfo->bytes_readonly +
-		meta_sinfo->bytes_super + meta_sinfo->bytes_root +
-		meta_sinfo->bytes_may_use + meta_sinfo->bytes_delalloc;
+	if (!block_rsv)
+		return 0;
 
-	if (used > meta_sinfo->total_bytes) {
-		retries++;
-		if (retries == 1) {
-			if (maybe_allocate_chunk(root, meta_sinfo))
-				goto again;
-			retries++;
-		} else {
-			spin_unlock(&meta_sinfo->lock);
-		}
+	spin_lock(&block_rsv->lock);
+	num_bytes = min_reserved;
+	if (block_rsv->reserved >= num_bytes)
+		ret = 0;
+	else
+		num_bytes -= block_rsv->reserved;
+	spin_unlock(&block_rsv->lock);
 
-		if (retries == 2) {
-			flush_delalloc(root, meta_sinfo);
-			goto again;
-		}
-		spin_lock(&meta_sinfo->lock);
-		meta_sinfo->bytes_may_use -= num_bytes;
-		spin_unlock(&meta_sinfo->lock);
+	if (!ret)
+		return 0;
 
-		dump_space_info(meta_sinfo, 0, 0);
-		return -ENOSPC;
+	ret = reserve_metadata_bytes(root, block_rsv, num_bytes, flush);
+	if (!ret) {
+		block_rsv_add_bytes(block_rsv, num_bytes, 0);
+		return 0;
 	}
 
-	check_force_delalloc(meta_sinfo);
-	spin_unlock(&meta_sinfo->lock);
+	return ret;
+}
+
+int btrfs_block_rsv_migrate(struct btrfs_block_rsv *src_rsv,
+			    struct btrfs_block_rsv *dst_rsv,
+			    u64 num_bytes)
+{
+	return block_rsv_migrate_bytes(src_rsv, dst_rsv, num_bytes);
+}
 
-	return 0;
+void btrfs_block_rsv_release(struct btrfs_root *root,
+			     struct btrfs_block_rsv *block_rsv,
+			     u64 num_bytes)
+{
+	struct btrfs_block_rsv *global_rsv = &root->fs_info->global_block_rsv;
+	if (global_rsv == block_rsv ||
+	    block_rsv->space_info != global_rsv->space_info)
+		global_rsv = NULL;
+	block_rsv_release_bytes(root->fs_info, block_rsv, global_rsv,
+				num_bytes);
 }
 
 /*
- * This will check the space that the inode allocates from to make sure we have
- * enough space for bytes.
+ * helper to calculate size of global block reservation.
+ * the desired value is sum of space used by extent tree,
+ * checksum tree and root tree
  */
-int btrfs_check_data_free_space(struct btrfs_root *root, struct inode *inode,
-				u64 bytes)
+static u64 calc_global_metadata_size(struct btrfs_fs_info *fs_info)
 {
-	struct btrfs_space_info *data_sinfo;
-	int ret = 0, committed = 0;
+	struct btrfs_space_info *sinfo;
+	u64 num_bytes;
+	u64 meta_used;
+	u64 data_used;
+	int csum_size = btrfs_super_csum_size(fs_info->super_copy);
 
-	/* make sure bytes are sectorsize aligned */
-	bytes = (bytes + root->sectorsize - 1) & ~((u64)root->sectorsize - 1);
+	sinfo = __find_space_info(fs_info, BTRFS_BLOCK_GROUP_DATA);
+	spin_lock(&sinfo->lock);
+	data_used = sinfo->bytes_used;
+	spin_unlock(&sinfo->lock);
 
-	data_sinfo = BTRFS_I(inode)->space_info;
-	if (!data_sinfo)
-		goto alloc;
+	sinfo = __find_space_info(fs_info, BTRFS_BLOCK_GROUP_METADATA);
+	spin_lock(&sinfo->lock);
+	if (sinfo->flags & BTRFS_BLOCK_GROUP_DATA)
+		data_used = 0;
+	meta_used = sinfo->bytes_used;
+	spin_unlock(&sinfo->lock);
 
-again:
-	/* make sure we have enough space to handle the data first */
-	spin_lock(&data_sinfo->lock);
-	if (data_sinfo->total_bytes - data_sinfo->bytes_used -
-	    data_sinfo->bytes_delalloc - data_sinfo->bytes_reserved -
-	    data_sinfo->bytes_pinned - data_sinfo->bytes_readonly -
-	    data_sinfo->bytes_may_use - data_sinfo->bytes_super < bytes) {
-		struct btrfs_trans_handle *trans;
+	num_bytes = (data_used >> fs_info->sb->s_blocksize_bits) *
+		    csum_size * 2;
+	num_bytes += div64_u64(data_used + meta_used, 50);
 
-		/*
-		 * if we don't have enough free bytes in this space then we need
-		 * to alloc a new chunk.
-		 */
-		if (!data_sinfo->full) {
-			u64 alloc_target;
+	if (num_bytes * 3 > meta_used)
+		num_bytes = div64_u64(meta_used, 3);
 
-			data_sinfo->force_alloc = 1;
-			spin_unlock(&data_sinfo->lock);
-alloc:
-			alloc_target = btrfs_get_alloc_profile(root, 1);
-			trans = btrfs_start_transaction(root, 1);
-			if (!trans)
-				return -ENOMEM;
+	return ALIGN(num_bytes, fs_info->extent_root->leafsize << 10);
+}
 
-			ret = do_chunk_alloc(trans, root->fs_info->extent_root,
-					     bytes + 2 * 1024 * 1024,
-					     alloc_target, 0);
-			btrfs_end_transaction(trans, root);
-			if (ret)
-				return ret;
+static void update_global_block_rsv(struct btrfs_fs_info *fs_info)
+{
+	struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
+	struct btrfs_space_info *sinfo = block_rsv->space_info;
+	u64 num_bytes;
 
-			if (!data_sinfo) {
-				btrfs_set_inode_space_info(root, inode);
-				data_sinfo = BTRFS_I(inode)->space_info;
-			}
-			goto again;
-		}
-		spin_unlock(&data_sinfo->lock);
+	num_bytes = calc_global_metadata_size(fs_info);
 
-		/* commit the current transaction and try again */
-		if (!committed && !root->fs_info->open_ioctl_trans) {
-			committed = 1;
-			trans = btrfs_join_transaction(root, 1);
-			if (!trans)
-				return -ENOMEM;
-			ret = btrfs_commit_transaction(trans, root);
-			if (ret)
-				return ret;
-			goto again;
-		}
+	spin_lock(&sinfo->lock);
+	spin_lock(&block_rsv->lock);
 
-		printk(KERN_ERR "no space left, need %llu, %llu delalloc bytes"
-		       ", %llu bytes_used, %llu bytes_reserved, "
-		       "%llu bytes_pinned, %llu bytes_readonly, %llu may use "
-		       "%llu total\n", (unsigned long long)bytes,
-		       (unsigned long long)data_sinfo->bytes_delalloc,
-		       (unsigned long long)data_sinfo->bytes_used,
-		       (unsigned long long)data_sinfo->bytes_reserved,
-		       (unsigned long long)data_sinfo->bytes_pinned,
-		       (unsigned long long)data_sinfo->bytes_readonly,
-		       (unsigned long long)data_sinfo->bytes_may_use,
-		       (unsigned long long)data_sinfo->total_bytes);
-		return -ENOSPC;
+	block_rsv->size = min_t(u64, num_bytes, 512 * 1024 * 1024);
+
+	num_bytes = sinfo->bytes_used + sinfo->bytes_pinned +
+		    sinfo->bytes_reserved + sinfo->bytes_readonly +
+		    sinfo->bytes_may_use;
+
+	if (sinfo->total_bytes > num_bytes) {
+		num_bytes = sinfo->total_bytes - num_bytes;
+		block_rsv->reserved += num_bytes;
+		sinfo->bytes_may_use += num_bytes;
+		trace_btrfs_space_reservation(fs_info, "space_info",
+				      sinfo->flags, num_bytes, 1);
 	}
-	data_sinfo->bytes_may_use += bytes;
-	BTRFS_I(inode)->reserved_bytes += bytes;
-	spin_unlock(&data_sinfo->lock);
 
-	return 0;
+	if (block_rsv->reserved >= block_rsv->size) {
+		num_bytes = block_rsv->reserved - block_rsv->size;
+		sinfo->bytes_may_use -= num_bytes;
+		trace_btrfs_space_reservation(fs_info, "space_info",
+				      sinfo->flags, num_bytes, 0);
+		block_rsv->reserved = block_rsv->size;
+		block_rsv->full = 1;
+	}
+
+	spin_unlock(&block_rsv->lock);
+	spin_unlock(&sinfo->lock);
 }
 
-/*
- * if there was an error for whatever reason after calling
- * btrfs_check_data_free_space, call this so we can cleanup the counters.
- */
-void btrfs_free_reserved_data_space(struct btrfs_root *root,
-				    struct inode *inode, u64 bytes)
+static void init_global_block_rsv(struct btrfs_fs_info *fs_info)
 {
-	struct btrfs_space_info *data_sinfo;
+	struct btrfs_space_info *space_info;
 
-	/* make sure bytes are sectorsize aligned */
-	bytes = (bytes + root->sectorsize - 1) & ~((u64)root->sectorsize - 1);
+	space_info = __find_space_info(fs_info, BTRFS_BLOCK_GROUP_SYSTEM);
+	fs_info->chunk_block_rsv.space_info = space_info;
 
-	data_sinfo = BTRFS_I(inode)->space_info;
-	spin_lock(&data_sinfo->lock);
-	data_sinfo->bytes_may_use -= bytes;
-	BTRFS_I(inode)->reserved_bytes -= bytes;
-	spin_unlock(&data_sinfo->lock);
+	space_info = __find_space_info(fs_info, BTRFS_BLOCK_GROUP_METADATA);
+	fs_info->global_block_rsv.space_info = space_info;
+	fs_info->delalloc_block_rsv.space_info = space_info;
+	fs_info->trans_block_rsv.space_info = space_info;
+	fs_info->empty_block_rsv.space_info = space_info;
+	fs_info->delayed_block_rsv.space_info = space_info;
+
+	fs_info->extent_root->block_rsv = &fs_info->global_block_rsv;
+	fs_info->csum_root->block_rsv = &fs_info->global_block_rsv;
+	fs_info->dev_root->block_rsv = &fs_info->global_block_rsv;
+	fs_info->tree_root->block_rsv = &fs_info->global_block_rsv;
+	if (fs_info->quota_root)
+		fs_info->quota_root->block_rsv = &fs_info->global_block_rsv;
+	fs_info->chunk_root->block_rsv = &fs_info->chunk_block_rsv;
+
+	update_global_block_rsv(fs_info);
 }
 
-/* called when we are adding a delalloc extent to the inode's io_tree */
-void btrfs_delalloc_reserve_space(struct btrfs_root *root, struct inode *inode,
-				  u64 bytes)
+static void release_global_block_rsv(struct btrfs_fs_info *fs_info)
 {
-	struct btrfs_space_info *data_sinfo;
+	block_rsv_release_bytes(fs_info, &fs_info->global_block_rsv, NULL,
+				(u64)-1);
+	WARN_ON(fs_info->delalloc_block_rsv.size > 0);
+	WARN_ON(fs_info->delalloc_block_rsv.reserved > 0);
+	WARN_ON(fs_info->trans_block_rsv.size > 0);
+	WARN_ON(fs_info->trans_block_rsv.reserved > 0);
+	WARN_ON(fs_info->chunk_block_rsv.size > 0);
+	WARN_ON(fs_info->chunk_block_rsv.reserved > 0);
+	WARN_ON(fs_info->delayed_block_rsv.size > 0);
+	WARN_ON(fs_info->delayed_block_rsv.reserved > 0);
+}
 
-	/* get the space info for where this inode will be storing its data */
-	data_sinfo = BTRFS_I(inode)->space_info;
+void btrfs_trans_release_metadata(struct btrfs_trans_handle *trans,
+				  struct btrfs_root *root)
+{
+	if (!trans->block_rsv)
+		return;
 
-	/* make sure we have enough space to handle the data first */
-	spin_lock(&data_sinfo->lock);
-	data_sinfo->bytes_delalloc += bytes;
+	if (!trans->bytes_reserved)
+		return;
+
+	trace_btrfs_space_reservation(root->fs_info, "transaction",
+				      trans->transid, trans->bytes_reserved, 0);
+	btrfs_block_rsv_release(root, trans->block_rsv, trans->bytes_reserved);
+	trans->bytes_reserved = 0;
+}
+
+/* Can only return 0 or -ENOSPC */
+int btrfs_orphan_reserve_metadata(struct btrfs_trans_handle *trans,
+				  struct inode *inode)
+{
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct btrfs_block_rsv *src_rsv = get_block_rsv(trans, root);
+	struct btrfs_block_rsv *dst_rsv = root->orphan_block_rsv;
 
 	/*
-	 * we are adding a delalloc extent without calling
-	 * btrfs_check_data_free_space first.  This happens on a weird
-	 * writepage condition, but shouldn't hurt our accounting
+	 * We need to hold space in order to delete our orphan item once we've
+	 * added it, so this takes the reservation so we can release it later
+	 * when we are truly done with the orphan item.
 	 */
-	if (unlikely(bytes > BTRFS_I(inode)->reserved_bytes)) {
-		data_sinfo->bytes_may_use -= BTRFS_I(inode)->reserved_bytes;
-		BTRFS_I(inode)->reserved_bytes = 0;
+	u64 num_bytes = btrfs_calc_trans_metadata_size(root, 1);
+	trace_btrfs_space_reservation(root->fs_info, "orphan",
+				      btrfs_ino(inode), num_bytes, 1);
+	return block_rsv_migrate_bytes(src_rsv, dst_rsv, num_bytes);
+}
+
+void btrfs_orphan_release_metadata(struct inode *inode)
+{
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	u64 num_bytes = btrfs_calc_trans_metadata_size(root, 1);
+	trace_btrfs_space_reservation(root->fs_info, "orphan",
+				      btrfs_ino(inode), num_bytes, 0);
+	btrfs_block_rsv_release(root, root->orphan_block_rsv, num_bytes);
+}
+
+/*
+ * btrfs_subvolume_reserve_metadata() - reserve space for subvolume operation
+ * root: the root of the parent directory
+ * rsv: block reservation
+ * items: the number of items that we need do reservation
+ * qgroup_reserved: used to return the reserved size in qgroup
+ *
+ * This function is used to reserve the space for snapshot/subvolume
+ * creation and deletion. Those operations are different with the
+ * common file/directory operations, they change two fs/file trees
+ * and root tree, the number of items that the qgroup reserves is
+ * different with the free space reservation. So we can not use
+ * the space reseravtion mechanism in start_transaction().
+ */
+int btrfs_subvolume_reserve_metadata(struct btrfs_root *root,
+				     struct btrfs_block_rsv *rsv,
+				     int items,
+				     u64 *qgroup_reserved,
+				     bool use_global_rsv)
+{
+	u64 num_bytes;
+	int ret;
+	struct btrfs_block_rsv *global_rsv = &root->fs_info->global_block_rsv;
+
+	if (root->fs_info->quota_enabled) {
+		/* One for parent inode, two for dir entries */
+		num_bytes = 3 * root->leafsize;
+		ret = btrfs_qgroup_reserve(root, num_bytes);
+		if (ret)
+			return ret;
 	} else {
-		data_sinfo->bytes_may_use -= bytes;
-		BTRFS_I(inode)->reserved_bytes -= bytes;
+		num_bytes = 0;
 	}
 
-	spin_unlock(&data_sinfo->lock);
+	*qgroup_reserved = num_bytes;
+
+	num_bytes = btrfs_calc_trans_metadata_size(root, items);
+	rsv->space_info = __find_space_info(root->fs_info,
+					    BTRFS_BLOCK_GROUP_METADATA);
+	ret = btrfs_block_rsv_add(root, rsv, num_bytes,
+				  BTRFS_RESERVE_FLUSH_ALL);
+
+	if (ret == -ENOSPC && use_global_rsv)
+		ret = btrfs_block_rsv_migrate(global_rsv, rsv, num_bytes);
+
+	if (ret) {
+		if (*qgroup_reserved)
+			btrfs_qgroup_free(root, *qgroup_reserved);
+	}
+
+	return ret;
 }
 
-/* called when we are clearing an delalloc extent from the inode's io_tree */
-void btrfs_delalloc_free_space(struct btrfs_root *root, struct inode *inode,
-			      u64 bytes)
+void btrfs_subvolume_release_metadata(struct btrfs_root *root,
+				      struct btrfs_block_rsv *rsv,
+				      u64 qgroup_reserved)
 {
-	struct btrfs_space_info *info;
+	btrfs_block_rsv_release(root, rsv, (u64)-1);
+	if (qgroup_reserved)
+		btrfs_qgroup_free(root, qgroup_reserved);
+}
+
+/**
+ * drop_outstanding_extent - drop an outstanding extent
+ * @inode: the inode we're dropping the extent for
+ *
+ * This is called when we are freeing up an outstanding extent, either called
+ * after an error or after an extent is written.  This will return the number of
+ * reserved extents that need to be freed.  This must be called with
+ * BTRFS_I(inode)->lock held.
+ */
+static unsigned drop_outstanding_extent(struct inode *inode)
+{
+	unsigned drop_inode_space = 0;
+	unsigned dropped_extents = 0;
 
-	info = BTRFS_I(inode)->space_info;
+	BUG_ON(!BTRFS_I(inode)->outstanding_extents);
+	BTRFS_I(inode)->outstanding_extents--;
 
-	spin_lock(&info->lock);
-	info->bytes_delalloc -= bytes;
-	spin_unlock(&info->lock);
-}
+	if (BTRFS_I(inode)->outstanding_extents == 0 &&
+	    test_and_clear_bit(BTRFS_INODE_DELALLOC_META_RESERVED,
+			       &BTRFS_I(inode)->runtime_flags))
+		drop_inode_space = 1;
 
-static void force_metadata_allocation(struct btrfs_fs_info *info)
+	/*
+	 * If we have more or the same amount of outsanding extents than we have
+	 * reserved then we need to leave the reserved extents count alone.
+	 */
+	if (BTRFS_I(inode)->outstanding_extents >=
+	    BTRFS_I(inode)->reserved_extents)
+		return drop_inode_space;
+
+	dropped_extents = BTRFS_I(inode)->reserved_extents -
+		BTRFS_I(inode)->outstanding_extents;
+	BTRFS_I(inode)->reserved_extents -= dropped_extents;
+	return dropped_extents + drop_inode_space;
+}
+
+/**
+ * calc_csum_metadata_size - return the amount of metada space that must be
+ *	reserved/free'd for the given bytes.
+ * @inode: the inode we're manipulating
+ * @num_bytes: the number of bytes in question
+ * @reserve: 1 if we are reserving space, 0 if we are freeing space
+ *
+ * This adjusts the number of csum_bytes in the inode and then returns the
+ * correct amount of metadata that must either be reserved or freed.  We
+ * calculate how many checksums we can fit into one leaf and then divide the
+ * number of bytes that will need to be checksumed by this value to figure out
+ * how many checksums will be required.  If we are adding bytes then the number
+ * may go up and we will return the number of additional bytes that must be
+ * reserved.  If it is going down we will return the number of bytes that must
+ * be freed.
+ *
+ * This must be called with BTRFS_I(inode)->lock held.
+ */
+static u64 calc_csum_metadata_size(struct inode *inode, u64 num_bytes,
+				   int reserve)
 {
-	struct list_head *head = &info->space_info;
-	struct btrfs_space_info *found;
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	u64 csum_size;
+	int num_csums_per_leaf;
+	int num_csums;
+	int old_csums;
 
-	rcu_read_lock();
-	list_for_each_entry_rcu(found, head, list) {
-		if (found->flags & BTRFS_BLOCK_GROUP_METADATA)
-			found->force_alloc = 1;
-	}
-	rcu_read_unlock();
+	if (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM &&
+	    BTRFS_I(inode)->csum_bytes == 0)
+		return 0;
+
+	old_csums = (int)div64_u64(BTRFS_I(inode)->csum_bytes, root->sectorsize);
+	if (reserve)
+		BTRFS_I(inode)->csum_bytes += num_bytes;
+	else
+		BTRFS_I(inode)->csum_bytes -= num_bytes;
+	csum_size = BTRFS_LEAF_DATA_SIZE(root) - sizeof(struct btrfs_item);
+	num_csums_per_leaf = (int)div64_u64(csum_size,
+					    sizeof(struct btrfs_csum_item) +
+					    sizeof(struct btrfs_disk_key));
+	num_csums = (int)div64_u64(BTRFS_I(inode)->csum_bytes, root->sectorsize);
+	num_csums = num_csums + num_csums_per_leaf - 1;
+	num_csums = num_csums / num_csums_per_leaf;
+
+	old_csums = old_csums + num_csums_per_leaf - 1;
+	old_csums = old_csums / num_csums_per_leaf;
+
+	/* No change, no need to reserve more */
+	if (old_csums == num_csums)
+		return 0;
+
+	if (reserve)
+		return btrfs_calc_trans_metadata_size(root,
+						      num_csums - old_csums);
+
+	return btrfs_calc_trans_metadata_size(root, old_csums - num_csums);
 }
 
-static int do_chunk_alloc(struct btrfs_trans_handle *trans,
-			  struct btrfs_root *extent_root, u64 alloc_bytes,
-			  u64 flags, int force)
+int btrfs_delalloc_reserve_metadata(struct inode *inode, u64 num_bytes)
 {
-	struct btrfs_space_info *space_info;
-	struct btrfs_fs_info *fs_info = extent_root->fs_info;
-	u64 thresh;
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct btrfs_block_rsv *block_rsv = &root->fs_info->delalloc_block_rsv;
+	u64 to_reserve = 0;
+	u64 csum_bytes;
+	unsigned nr_extents = 0;
+	int extra_reserve = 0;
+	enum btrfs_reserve_flush_enum flush = BTRFS_RESERVE_FLUSH_ALL;
 	int ret = 0;
+	bool delalloc_lock = true;
+	u64 to_free = 0;
+	unsigned dropped;
+
+	/* If we are a free space inode we need to not flush since we will be in
+	 * the middle of a transaction commit.  We also don't need the delalloc
+	 * mutex since we won't race with anybody.  We need this mostly to make
+	 * lockdep shut its filthy mouth.
+	 */
+	if (btrfs_is_free_space_inode(inode)) {
+		flush = BTRFS_RESERVE_NO_FLUSH;
+		delalloc_lock = false;
+	}
 
-	mutex_lock(&fs_info->chunk_mutex);
+	if (flush != BTRFS_RESERVE_NO_FLUSH &&
+	    btrfs_transaction_in_commit(root->fs_info))
+		schedule_timeout(1);
 
-	flags = btrfs_reduce_alloc_profile(extent_root, flags);
+	if (delalloc_lock)
+		mutex_lock(&BTRFS_I(inode)->delalloc_mutex);
 
-	space_info = __find_space_info(extent_root->fs_info, flags);
-	if (!space_info) {
-		ret = update_space_info(extent_root->fs_info, flags,
-					0, 0, &space_info);
-		BUG_ON(ret);
+	num_bytes = ALIGN(num_bytes, root->sectorsize);
+
+	spin_lock(&BTRFS_I(inode)->lock);
+	BTRFS_I(inode)->outstanding_extents++;
+
+	if (BTRFS_I(inode)->outstanding_extents >
+	    BTRFS_I(inode)->reserved_extents)
+		nr_extents = BTRFS_I(inode)->outstanding_extents -
+			BTRFS_I(inode)->reserved_extents;
+
+	/*
+	 * Add an item to reserve for updating the inode when we complete the
+	 * delalloc io.
+	 */
+	if (!test_bit(BTRFS_INODE_DELALLOC_META_RESERVED,
+		      &BTRFS_I(inode)->runtime_flags)) {
+		nr_extents++;
+		extra_reserve = 1;
 	}
-	BUG_ON(!space_info);
 
-	spin_lock(&space_info->lock);
-	if (space_info->force_alloc)
-		force = 1;
-	if (space_info->full) {
-		spin_unlock(&space_info->lock);
-		goto out;
+	to_reserve = btrfs_calc_trans_metadata_size(root, nr_extents);
+	to_reserve += calc_csum_metadata_size(inode, num_bytes, 1);
+	csum_bytes = BTRFS_I(inode)->csum_bytes;
+	spin_unlock(&BTRFS_I(inode)->lock);
+
+	if (root->fs_info->quota_enabled) {
+		ret = btrfs_qgroup_reserve(root, num_bytes +
+					   nr_extents * root->leafsize);
+		if (ret)
+			goto out_fail;
 	}
 
-	thresh = space_info->total_bytes - space_info->bytes_readonly;
-	thresh = div_factor(thresh, 8);
-	if (!force &&
-	   (space_info->bytes_used + space_info->bytes_pinned +
-	    space_info->bytes_reserved + alloc_bytes) < thresh) {
-		spin_unlock(&space_info->lock);
-		goto out;
+	ret = reserve_metadata_bytes(root, block_rsv, to_reserve, flush);
+	if (unlikely(ret)) {
+		if (root->fs_info->quota_enabled)
+			btrfs_qgroup_free(root, num_bytes +
+						nr_extents * root->leafsize);
+		goto out_fail;
 	}
-	spin_unlock(&space_info->lock);
 
+	spin_lock(&BTRFS_I(inode)->lock);
+	if (extra_reserve) {
+		set_bit(BTRFS_INODE_DELALLOC_META_RESERVED,
+			&BTRFS_I(inode)->runtime_flags);
+		nr_extents--;
+	}
+	BTRFS_I(inode)->reserved_extents += nr_extents;
+	spin_unlock(&BTRFS_I(inode)->lock);
+
+	if (delalloc_lock)
+		mutex_unlock(&BTRFS_I(inode)->delalloc_mutex);
+
+	if (to_reserve)
+		trace_btrfs_space_reservation(root->fs_info, "delalloc",
+					      btrfs_ino(inode), to_reserve, 1);
+	block_rsv_add_bytes(block_rsv, to_reserve, 1);
+
+	return 0;
+
+out_fail:
+	spin_lock(&BTRFS_I(inode)->lock);
+	dropped = drop_outstanding_extent(inode);
 	/*
-	 * if we're doing a data chunk, go ahead and make sure that
-	 * we keep a reasonable number of metadata chunks allocated in the
-	 * FS as well.
+	 * If the inodes csum_bytes is the same as the original
+	 * csum_bytes then we know we haven't raced with any free()ers
+	 * so we can just reduce our inodes csum bytes and carry on.
 	 */
-	if (flags & BTRFS_BLOCK_GROUP_DATA && fs_info->metadata_ratio) {
-		fs_info->data_chunk_allocations++;
-		if (!(fs_info->data_chunk_allocations %
-		      fs_info->metadata_ratio))
-			force_metadata_allocation(fs_info);
+	if (BTRFS_I(inode)->csum_bytes == csum_bytes) {
+		calc_csum_metadata_size(inode, num_bytes, 0);
+	} else {
+		u64 orig_csum_bytes = BTRFS_I(inode)->csum_bytes;
+		u64 bytes;
+
+		/*
+		 * This is tricky, but first we need to figure out how much we
+		 * free'd from any free-ers that occured during this
+		 * reservation, so we reset ->csum_bytes to the csum_bytes
+		 * before we dropped our lock, and then call the free for the
+		 * number of bytes that were freed while we were trying our
+		 * reservation.
+		 */
+		bytes = csum_bytes - BTRFS_I(inode)->csum_bytes;
+		BTRFS_I(inode)->csum_bytes = csum_bytes;
+		to_free = calc_csum_metadata_size(inode, bytes, 0);
+
+
+		/*
+		 * Now we need to see how much we would have freed had we not
+		 * been making this reservation and our ->csum_bytes were not
+		 * artificially inflated.
+		 */
+		BTRFS_I(inode)->csum_bytes = csum_bytes - num_bytes;
+		bytes = csum_bytes - orig_csum_bytes;
+		bytes = calc_csum_metadata_size(inode, bytes, 0);
+
+		/*
+		 * Now reset ->csum_bytes to what it should be.  If bytes is
+		 * more than to_free then we would have free'd more space had we
+		 * not had an artificially high ->csum_bytes, so we need to free
+		 * the remainder.  If bytes is the same or less then we don't
+		 * need to do anything, the other free-ers did the correct
+		 * thing.
+		 */
+		BTRFS_I(inode)->csum_bytes = orig_csum_bytes - num_bytes;
+		if (bytes > to_free)
+			to_free = bytes - to_free;
+		else
+			to_free = 0;
 	}
+	spin_unlock(&BTRFS_I(inode)->lock);
+	if (dropped)
+		to_free += btrfs_calc_trans_metadata_size(root, dropped);
 
-	ret = btrfs_alloc_chunk(trans, extent_root, flags);
-	spin_lock(&space_info->lock);
-	if (ret)
-		space_info->full = 1;
-	space_info->force_alloc = 0;
-	spin_unlock(&space_info->lock);
-out:
-	mutex_unlock(&extent_root->fs_info->chunk_mutex);
+	if (to_free) {
+		btrfs_block_rsv_release(root, block_rsv, to_free);
+		trace_btrfs_space_reservation(root->fs_info, "delalloc",
+					      btrfs_ino(inode), to_free, 0);
+	}
+	if (delalloc_lock)
+		mutex_unlock(&BTRFS_I(inode)->delalloc_mutex);
 	return ret;
 }
 
-static int update_block_group(struct btrfs_trans_handle *trans,
-			      struct btrfs_root *root,
-			      u64 bytenr, u64 num_bytes, int alloc,
-			      int mark_free)
+/**
+ * btrfs_delalloc_release_metadata - release a metadata reservation for an inode
+ * @inode: the inode to release the reservation for
+ * @num_bytes: the number of bytes we're releasing
+ *
+ * This will release the metadata reservation for an inode.  This can be called
+ * once we complete IO for a given set of bytes to release their metadata
+ * reservations.
+ */
+void btrfs_delalloc_release_metadata(struct inode *inode, u64 num_bytes)
 {
-	struct btrfs_block_group_cache *cache;
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	u64 to_free = 0;
+	unsigned dropped;
+
+	num_bytes = ALIGN(num_bytes, root->sectorsize);
+	spin_lock(&BTRFS_I(inode)->lock);
+	dropped = drop_outstanding_extent(inode);
+
+	if (num_bytes)
+		to_free = calc_csum_metadata_size(inode, num_bytes, 0);
+	spin_unlock(&BTRFS_I(inode)->lock);
+	if (dropped > 0)
+		to_free += btrfs_calc_trans_metadata_size(root, dropped);
+
+	trace_btrfs_space_reservation(root->fs_info, "delalloc",
+				      btrfs_ino(inode), to_free, 0);
+	if (root->fs_info->quota_enabled) {
+		btrfs_qgroup_free(root, num_bytes +
+					dropped * root->leafsize);
+	}
+
+	btrfs_block_rsv_release(root, &root->fs_info->delalloc_block_rsv,
+				to_free);
+}
+
+/**
+ * btrfs_delalloc_reserve_space - reserve data and metadata space for delalloc
+ * @inode: inode we're writing to
+ * @num_bytes: the number of bytes we want to allocate
+ *
+ * This will do the following things
+ *
+ * o reserve space in the data space info for num_bytes
+ * o reserve space in the metadata space info based on number of outstanding
+ *   extents and how much csums will be needed
+ * o add to the inodes ->delalloc_bytes
+ * o add it to the fs_info's delalloc inodes list.
+ *
+ * This will return 0 for success and -ENOSPC if there is no space left.
+ */
+int btrfs_delalloc_reserve_space(struct inode *inode, u64 num_bytes)
+{
+	int ret;
+
+	ret = btrfs_check_data_free_space(inode, num_bytes);
+	if (ret)
+		return ret;
+
+	ret = btrfs_delalloc_reserve_metadata(inode, num_bytes);
+	if (ret) {
+		btrfs_free_reserved_data_space(inode, num_bytes);
+		return ret;
+	}
+
+	return 0;
+}
+
+/**
+ * btrfs_delalloc_release_space - release data and metadata space for delalloc
+ * @inode: inode we're releasing space for
+ * @num_bytes: the number of bytes we want to free up
+ *
+ * This must be matched with a call to btrfs_delalloc_reserve_space.  This is
+ * called in the case that we don't need the metadata AND data reservations
+ * anymore.  So if there is an error or we insert an inline extent.
+ *
+ * This function will release the metadata space that was not used and will
+ * decrement ->delalloc_bytes and remove it from the fs_info delalloc_inodes
+ * list if there are no delalloc bytes left.
+ */
+void btrfs_delalloc_release_space(struct inode *inode, u64 num_bytes)
+{
+	btrfs_delalloc_release_metadata(inode, num_bytes);
+	btrfs_free_reserved_data_space(inode, num_bytes);
+}
+
+static int update_block_group(struct btrfs_root *root,
+			      u64 bytenr, u64 num_bytes, int alloc)
+{
+	struct btrfs_block_group_cache *cache = NULL;
 	struct btrfs_fs_info *info = root->fs_info;
 	u64 total = num_bytes;
 	u64 old_val;
 	u64 byte_in_group;
+	int factor;
 
 	/* block accounting for super block */
-	spin_lock(&info->delalloc_lock);
-	old_val = btrfs_super_bytes_used(&info->super_copy);
+	spin_lock(&info->delalloc_root_lock);
+	old_val = btrfs_super_bytes_used(info->super_copy);
 	if (alloc)
 		old_val += num_bytes;
 	else
 		old_val -= num_bytes;
-	btrfs_set_super_bytes_used(&info->super_copy, old_val);
-	spin_unlock(&info->delalloc_lock);
+	btrfs_set_super_bytes_used(info->super_copy, old_val);
+	spin_unlock(&info->delalloc_root_lock);
 
 	while (total) {
 		cache = btrfs_lookup_block_group(info, bytenr);
 		if (!cache)
-			return -1;
+			return -ENOENT;
+		if (cache->flags & (BTRFS_BLOCK_GROUP_DUP |
+				    BTRFS_BLOCK_GROUP_RAID1 |
+				    BTRFS_BLOCK_GROUP_RAID10))
+			factor = 2;
+		else
+			factor = 1;
+		/*
+		 * If this block group has free space cache written out, we
+		 * need to make sure to load it if we are removing space.  This
+		 * is because we need the unpinning stage to actually add the
+		 * space back to the block group, otherwise we will leak space.
+		 */
+		if (!alloc && cache->cached == BTRFS_CACHE_NO)
+			cache_block_group(cache, 1);
+
 		byte_in_group = bytenr - cache->key.objectid;
 		WARN_ON(byte_in_group > cache->key.offset);
 
 		spin_lock(&cache->space_info->lock);
 		spin_lock(&cache->lock);
+
+		if (btrfs_test_opt(root, SPACE_CACHE) &&
+		    cache->disk_cache_state < BTRFS_DC_CLEAR)
+			cache->disk_cache_state = BTRFS_DC_CLEAR;
+
 		cache->dirty = 1;
 		old_val = btrfs_block_group_used(&cache->item);
 		num_bytes = min(total, cache->key.offset - byte_in_group);
@@ -3488,31 +5420,24 @@ static int update_block_group(struct btrfs_trans_handle *trans,
 			old_val += num_bytes;
 			btrfs_set_block_group_used(&cache->item, old_val);
 			cache->reserved -= num_bytes;
-			cache->space_info->bytes_used += num_bytes;
 			cache->space_info->bytes_reserved -= num_bytes;
-			if (cache->ro)
-				cache->space_info->bytes_readonly -= num_bytes;
+			cache->space_info->bytes_used += num_bytes;
+			cache->space_info->disk_used += num_bytes * factor;
 			spin_unlock(&cache->lock);
 			spin_unlock(&cache->space_info->lock);
 		} else {
 			old_val -= num_bytes;
-			cache->space_info->bytes_used -= num_bytes;
-			if (cache->ro)
-				cache->space_info->bytes_readonly += num_bytes;
 			btrfs_set_block_group_used(&cache->item, old_val);
+			cache->pinned += num_bytes;
+			cache->space_info->bytes_pinned += num_bytes;
+			cache->space_info->bytes_used -= num_bytes;
+			cache->space_info->disk_used -= num_bytes * factor;
 			spin_unlock(&cache->lock);
 			spin_unlock(&cache->space_info->lock);
-			if (mark_free) {
-				int ret;
 
-				ret = btrfs_discard_extent(root, bytenr,
-							   num_bytes);
-				WARN_ON(ret);
-
-				ret = btrfs_add_free_space(cache, bytenr,
-							   num_bytes);
-				WARN_ON(ret);
-			}
+			set_extent_dirty(info->pinned_extents,
+					 bytenr, bytenr + num_bytes - 1,
+					 GFP_NOFS | __GFP_NOFAIL);
 		}
 		btrfs_put_block_group(cache);
 		total -= num_bytes;
@@ -3526,6 +5451,13 @@ static u64 first_logical_byte(struct btrfs_root *root, u64 search_start)
 	struct btrfs_block_group_cache *cache;
 	u64 bytenr;
 
+	spin_lock(&root->fs_info->block_group_cache_lock);
+	bytenr = root->fs_info->first_logical_byte;
+	spin_unlock(&root->fs_info->block_group_cache_lock);
+
+	if (bytenr < (u64)-1)
+		return bytenr;
+
 	cache = btrfs_lookup_first_block_group(root->fs_info, search_start);
 	if (!cache)
 		return 0;
@@ -3536,18 +5468,10 @@ static u64 first_logical_byte(struct btrfs_root *root, u64 search_start)
 	return bytenr;
 }
 
-/*
- * this function must be called within transaction
- */
-int btrfs_pin_extent(struct btrfs_root *root,
-		     u64 bytenr, u64 num_bytes, int reserved)
+static int pin_down_extent(struct btrfs_root *root,
+			   struct btrfs_block_group_cache *cache,
+			   u64 bytenr, u64 num_bytes, int reserved)
 {
-	struct btrfs_fs_info *fs_info = root->fs_info;
-	struct btrfs_block_group_cache *cache;
-
-	cache = btrfs_lookup_block_group(fs_info, bytenr);
-	BUG_ON(!cache);
-
 	spin_lock(&cache->space_info->lock);
 	spin_lock(&cache->lock);
 	cache->pinned += num_bytes;
@@ -3559,31 +5483,195 @@ int btrfs_pin_extent(struct btrfs_root *root,
 	spin_unlock(&cache->lock);
 	spin_unlock(&cache->space_info->lock);
 
+	set_extent_dirty(root->fs_info->pinned_extents, bytenr,
+			 bytenr + num_bytes - 1, GFP_NOFS | __GFP_NOFAIL);
+	if (reserved)
+		trace_btrfs_reserved_extent_free(root, bytenr, num_bytes);
+	return 0;
+}
+
+/*
+ * this function must be called within transaction
+ */
+int btrfs_pin_extent(struct btrfs_root *root,
+		     u64 bytenr, u64 num_bytes, int reserved)
+{
+	struct btrfs_block_group_cache *cache;
+
+	cache = btrfs_lookup_block_group(root->fs_info, bytenr);
+	BUG_ON(!cache); /* Logic error */
+
+	pin_down_extent(root, cache, bytenr, num_bytes, reserved);
+
 	btrfs_put_block_group(cache);
+	return 0;
+}
+
+/*
+ * this function must be called within transaction
+ */
+int btrfs_pin_extent_for_log_replay(struct btrfs_root *root,
+				    u64 bytenr, u64 num_bytes)
+{
+	struct btrfs_block_group_cache *cache;
+	int ret;
+
+	cache = btrfs_lookup_block_group(root->fs_info, bytenr);
+	if (!cache)
+		return -EINVAL;
+
+	/*
+	 * pull in the free space cache (if any) so that our pin
+	 * removes the free space from the cache.  We have load_only set
+	 * to one because the slow code to read in the free extents does check
+	 * the pinned extents.
+	 */
+	cache_block_group(cache, 1);
+
+	pin_down_extent(root, cache, bytenr, num_bytes, 0);
+
+	/* remove us from the free space cache (if we're there at all) */
+	ret = btrfs_remove_free_space(cache, bytenr, num_bytes);
+	btrfs_put_block_group(cache);
+	return ret;
+}
+
+static int __exclude_logged_extent(struct btrfs_root *root, u64 start, u64 num_bytes)
+{
+	int ret;
+	struct btrfs_block_group_cache *block_group;
+	struct btrfs_caching_control *caching_ctl;
+
+	block_group = btrfs_lookup_block_group(root->fs_info, start);
+	if (!block_group)
+		return -EINVAL;
+
+	cache_block_group(block_group, 0);
+	caching_ctl = get_caching_control(block_group);
+
+	if (!caching_ctl) {
+		/* Logic error */
+		BUG_ON(!block_group_cache_done(block_group));
+		ret = btrfs_remove_free_space(block_group, start, num_bytes);
+	} else {
+		mutex_lock(&caching_ctl->mutex);
+
+		if (start >= caching_ctl->progress) {
+			ret = add_excluded_extent(root, start, num_bytes);
+		} else if (start + num_bytes <= caching_ctl->progress) {
+			ret = btrfs_remove_free_space(block_group,
+						      start, num_bytes);
+		} else {
+			num_bytes = caching_ctl->progress - start;
+			ret = btrfs_remove_free_space(block_group,
+						      start, num_bytes);
+			if (ret)
+				goto out_lock;
+
+			num_bytes = (start + num_bytes) -
+				caching_ctl->progress;
+			start = caching_ctl->progress;
+			ret = add_excluded_extent(root, start, num_bytes);
+		}
+out_lock:
+		mutex_unlock(&caching_ctl->mutex);
+		put_caching_control(caching_ctl);
+	}
+	btrfs_put_block_group(block_group);
+	return ret;
+}
+
+int btrfs_exclude_logged_extents(struct btrfs_root *log,
+				 struct extent_buffer *eb)
+{
+	struct btrfs_file_extent_item *item;
+	struct btrfs_key key;
+	int found_type;
+	int i;
+
+	if (!btrfs_fs_incompat(log->fs_info, MIXED_GROUPS))
+		return 0;
+
+	for (i = 0; i < btrfs_header_nritems(eb); i++) {
+		btrfs_item_key_to_cpu(eb, &key, i);
+		if (key.type != BTRFS_EXTENT_DATA_KEY)
+			continue;
+		item = btrfs_item_ptr(eb, i, struct btrfs_file_extent_item);
+		found_type = btrfs_file_extent_type(eb, item);
+		if (found_type == BTRFS_FILE_EXTENT_INLINE)
+			continue;
+		if (btrfs_file_extent_disk_bytenr(eb, item) == 0)
+			continue;
+		key.objectid = btrfs_file_extent_disk_bytenr(eb, item);
+		key.offset = btrfs_file_extent_disk_num_bytes(eb, item);
+		__exclude_logged_extent(log, key.objectid, key.offset);
+	}
 
-	set_extent_dirty(fs_info->pinned_extents,
-			 bytenr, bytenr + num_bytes - 1, GFP_NOFS);
 	return 0;
 }
 
-static int update_reserved_extents(struct btrfs_block_group_cache *cache,
-				   u64 num_bytes, int reserve)
+/**
+ * btrfs_update_reserved_bytes - update the block_group and space info counters
+ * @cache:	The cache we are manipulating
+ * @num_bytes:	The number of bytes in question
+ * @reserve:	One of the reservation enums
+ * @delalloc:   The blocks are allocated for the delalloc write
+ *
+ * This is called by the allocator when it reserves space, or by somebody who is
+ * freeing space that was never actually used on disk.  For example if you
+ * reserve some space for a new leaf in transaction A and before transaction A
+ * commits you free that leaf, you call this with reserve set to 0 in order to
+ * clear the reservation.
+ *
+ * Metadata reservations should be called with RESERVE_ALLOC so we do the proper
+ * ENOSPC accounting.  For data we handle the reservation through clearing the
+ * delalloc bits in the io_tree.  We have to do this since we could end up
+ * allocating less disk space for the amount of data we have reserved in the
+ * case of compression.
+ *
+ * If this is a reservation and the block group has become read only we cannot
+ * make the reservation and return -EAGAIN, otherwise this function always
+ * succeeds.
+ */
+static int btrfs_update_reserved_bytes(struct btrfs_block_group_cache *cache,
+				       u64 num_bytes, int reserve, int delalloc)
 {
-	spin_lock(&cache->space_info->lock);
+	struct btrfs_space_info *space_info = cache->space_info;
+	int ret = 0;
+
+	spin_lock(&space_info->lock);
 	spin_lock(&cache->lock);
-	if (reserve) {
-		cache->reserved += num_bytes;
-		cache->space_info->bytes_reserved += num_bytes;
+	if (reserve != RESERVE_FREE) {
+		if (cache->ro) {
+			ret = -EAGAIN;
+		} else {
+			cache->reserved += num_bytes;
+			space_info->bytes_reserved += num_bytes;
+			if (reserve == RESERVE_ALLOC) {
+				trace_btrfs_space_reservation(cache->fs_info,
+						"space_info", space_info->flags,
+						num_bytes, 0);
+				space_info->bytes_may_use -= num_bytes;
+			}
+
+			if (delalloc)
+				cache->delalloc_bytes += num_bytes;
+		}
 	} else {
+		if (cache->ro)
+			space_info->bytes_readonly += num_bytes;
 		cache->reserved -= num_bytes;
-		cache->space_info->bytes_reserved -= num_bytes;
+		space_info->bytes_reserved -= num_bytes;
+
+		if (delalloc)
+			cache->delalloc_bytes -= num_bytes;
 	}
 	spin_unlock(&cache->lock);
-	spin_unlock(&cache->space_info->lock);
-	return 0;
+	spin_unlock(&space_info->lock);
+	return ret;
 }
 
-int btrfs_prepare_extent_commit(struct btrfs_trans_handle *trans,
+void btrfs_prepare_extent_commit(struct btrfs_trans_handle *trans,
 				struct btrfs_root *root)
 {
 	struct btrfs_fs_info *fs_info = root->fs_info;
@@ -3591,7 +5679,7 @@ int btrfs_prepare_extent_commit(struct btrfs_trans_handle *trans,
 	struct btrfs_caching_control *caching_ctl;
 	struct btrfs_block_group_cache *cache;
 
-	down_write(&fs_info->extent_commit_sem);
+	down_write(&fs_info->commit_root_sem);
 
 	list_for_each_entry_safe(caching_ctl, next,
 				 &fs_info->caching_block_groups, list) {
@@ -3610,23 +5698,28 @@ int btrfs_prepare_extent_commit(struct btrfs_trans_handle *trans,
 	else
 		fs_info->pinned_extents = &fs_info->freed_extents[0];
 
-	up_write(&fs_info->extent_commit_sem);
-	return 0;
+	up_write(&fs_info->commit_root_sem);
+
+	update_global_block_rsv(fs_info);
 }
 
 static int unpin_extent_range(struct btrfs_root *root, u64 start, u64 end)
 {
 	struct btrfs_fs_info *fs_info = root->fs_info;
 	struct btrfs_block_group_cache *cache = NULL;
+	struct btrfs_space_info *space_info;
+	struct btrfs_block_rsv *global_rsv = &fs_info->global_block_rsv;
 	u64 len;
+	bool readonly;
 
 	while (start <= end) {
+		readonly = false;
 		if (!cache ||
 		    start >= cache->key.objectid + cache->key.offset) {
 			if (cache)
 				btrfs_put_block_group(cache);
 			cache = btrfs_lookup_block_group(fs_info, start);
-			BUG_ON(!cache);
+			BUG_ON(!cache); /* Logic error */
 		}
 
 		len = cache->key.objectid + cache->key.offset - start;
@@ -3637,14 +5730,32 @@ static int unpin_extent_range(struct btrfs_root *root, u64 start, u64 end)
 			btrfs_add_free_space(cache, start, len);
 		}
 
-		spin_lock(&cache->space_info->lock);
+		start += len;
+		space_info = cache->space_info;
+
+		spin_lock(&space_info->lock);
 		spin_lock(&cache->lock);
 		cache->pinned -= len;
-		cache->space_info->bytes_pinned -= len;
+		space_info->bytes_pinned -= len;
+		percpu_counter_add(&space_info->total_bytes_pinned, -len);
+		if (cache->ro) {
+			space_info->bytes_readonly += len;
+			readonly = true;
+		}
 		spin_unlock(&cache->lock);
-		spin_unlock(&cache->space_info->lock);
-
-		start += len;
+		if (!readonly && global_rsv->space_info == space_info) {
+			spin_lock(&global_rsv->lock);
+			if (!global_rsv->full) {
+				len = min(len, global_rsv->size -
+					  global_rsv->reserved);
+				global_rsv->reserved += len;
+				space_info->bytes_may_use += len;
+				if (global_rsv->reserved >= global_rsv->size)
+					global_rsv->full = 1;
+			}
+			spin_unlock(&global_rsv->lock);
+		}
+		spin_unlock(&space_info->lock);
 	}
 
 	if (cache)
@@ -3661,6 +5772,9 @@ int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans,
 	u64 end;
 	int ret;
 
+	if (trans->aborted)
+		return 0;
+
 	if (fs_info->pinned_extents == &fs_info->freed_extents[0])
 		unpin = &fs_info->freed_extents[1];
 	else
@@ -3668,79 +5782,50 @@ int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans,
 
 	while (1) {
 		ret = find_first_extent_bit(unpin, 0, &start, &end,
-					    EXTENT_DIRTY);
+					    EXTENT_DIRTY, NULL);
 		if (ret)
 			break;
 
-		ret = btrfs_discard_extent(root, start, end + 1 - start);
+		if (btrfs_test_opt(root, DISCARD))
+			ret = btrfs_discard_extent(root, start,
+						   end + 1 - start, NULL);
 
 		clear_extent_dirty(unpin, start, end, GFP_NOFS);
 		unpin_extent_range(root, start, end);
 		cond_resched();
 	}
 
-	return ret;
+	return 0;
 }
 
-static int pin_down_bytes(struct btrfs_trans_handle *trans,
-			  struct btrfs_root *root,
-			  struct btrfs_path *path,
-			  u64 bytenr, u64 num_bytes,
-			  int is_data, int reserved,
-			  struct extent_buffer **must_clean)
+static void add_pinned_bytes(struct btrfs_fs_info *fs_info, u64 num_bytes,
+			     u64 owner, u64 root_objectid)
 {
-	int err = 0;
-	struct extent_buffer *buf;
-
-	if (is_data)
-		goto pinit;
-
-	/*
-	 * discard is sloooow, and so triggering discards on
-	 * individual btree blocks isn't a good plan.  Just
-	 * pin everything in discard mode.
-	 */
-	if (btrfs_test_opt(root, DISCARD))
-		goto pinit;
-
-	buf = btrfs_find_tree_block(root, bytenr, num_bytes);
-	if (!buf)
-		goto pinit;
+	struct btrfs_space_info *space_info;
+	u64 flags;
 
-	/* we can reuse a block if it hasn't been written
-	 * and it is from this transaction.  We can't
-	 * reuse anything from the tree log root because
-	 * it has tiny sub-transactions.
-	 */
-	if (btrfs_buffer_uptodate(buf, 0) &&
-	    btrfs_try_tree_lock(buf)) {
-		u64 header_owner = btrfs_header_owner(buf);
-		u64 header_transid = btrfs_header_generation(buf);
-		if (header_owner != BTRFS_TREE_LOG_OBJECTID &&
-		    header_transid == trans->transid &&
-		    !btrfs_header_flag(buf, BTRFS_HEADER_FLAG_WRITTEN)) {
-			*must_clean = buf;
-			return 1;
-		}
-		btrfs_tree_unlock(buf);
+	if (owner < BTRFS_FIRST_FREE_OBJECTID) {
+		if (root_objectid == BTRFS_CHUNK_TREE_OBJECTID)
+			flags = BTRFS_BLOCK_GROUP_SYSTEM;
+		else
+			flags = BTRFS_BLOCK_GROUP_METADATA;
+	} else {
+		flags = BTRFS_BLOCK_GROUP_DATA;
 	}
-	free_extent_buffer(buf);
-pinit:
-	if (path)
-		btrfs_set_path_blocking(path);
-	/* unlocks the pinned mutex */
-	btrfs_pin_extent(root, bytenr, num_bytes, reserved);
 
-	BUG_ON(err < 0);
-	return 0;
+	space_info = __find_space_info(fs_info, flags);
+	BUG_ON(!space_info); /* Logic bug */
+	percpu_counter_add(&space_info->total_bytes_pinned, num_bytes);
 }
 
+
 static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
 				struct btrfs_root *root,
 				u64 bytenr, u64 num_bytes, u64 parent,
 				u64 root_objectid, u64 owner_objectid,
 				u64 owner_offset, int refs_to_drop,
-				struct btrfs_delayed_extent_op *extent_op)
+				struct btrfs_delayed_extent_op *extent_op,
+				int no_quota)
 {
 	struct btrfs_key key;
 	struct btrfs_path *path;
@@ -3756,6 +5841,13 @@ static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
 	int num_to_del = 1;
 	u32 item_size;
 	u64 refs;
+	int last_ref = 0;
+	enum btrfs_qgroup_operation_type type = BTRFS_QGROUP_OPER_SUB_EXCL;
+	bool skinny_metadata = btrfs_fs_incompat(root->fs_info,
+						 SKINNY_METADATA);
+
+	if (!info->quota_enabled || !is_fstree(root_objectid))
+		no_quota = 1;
 
 	path = btrfs_alloc_path();
 	if (!path)
@@ -3767,6 +5859,9 @@ static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
 	is_data = owner_objectid >= BTRFS_FIRST_FREE_OBJECTID;
 	BUG_ON(!is_data && refs_to_drop != 1);
 
+	if (is_data)
+		skinny_metadata = 0;
+
 	ret = lookup_extent_backref(trans, extent_root, path, &iref,
 				    bytenr, num_bytes, parent,
 				    root_objectid, owner_objectid,
@@ -3783,6 +5878,11 @@ static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
 				found_extent = 1;
 				break;
 			}
+			if (key.type == BTRFS_METADATA_ITEM_KEY &&
+			    key.offset == owner_objectid) {
+				found_extent = 1;
+				break;
+			}
 			if (path->slots[0] - extent_slot > 5)
 				break;
 			extent_slot--;
@@ -3796,36 +5896,73 @@ static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
 			BUG_ON(iref);
 			ret = remove_extent_backref(trans, extent_root, path,
 						    NULL, refs_to_drop,
-						    is_data);
-			BUG_ON(ret);
-			btrfs_release_path(extent_root, path);
+						    is_data, &last_ref);
+			if (ret) {
+				btrfs_abort_transaction(trans, extent_root, ret);
+				goto out;
+			}
+			btrfs_release_path(path);
 			path->leave_spinning = 1;
 
 			key.objectid = bytenr;
 			key.type = BTRFS_EXTENT_ITEM_KEY;
 			key.offset = num_bytes;
 
+			if (!is_data && skinny_metadata) {
+				key.type = BTRFS_METADATA_ITEM_KEY;
+				key.offset = owner_objectid;
+			}
+
 			ret = btrfs_search_slot(trans, extent_root,
 						&key, path, -1, 1);
+			if (ret > 0 && skinny_metadata && path->slots[0]) {
+				/*
+				 * Couldn't find our skinny metadata item,
+				 * see if we have ye olde extent item.
+				 */
+				path->slots[0]--;
+				btrfs_item_key_to_cpu(path->nodes[0], &key,
+						      path->slots[0]);
+				if (key.objectid == bytenr &&
+				    key.type == BTRFS_EXTENT_ITEM_KEY &&
+				    key.offset == num_bytes)
+					ret = 0;
+			}
+
+			if (ret > 0 && skinny_metadata) {
+				skinny_metadata = false;
+				key.objectid = bytenr;
+				key.type = BTRFS_EXTENT_ITEM_KEY;
+				key.offset = num_bytes;
+				btrfs_release_path(path);
+				ret = btrfs_search_slot(trans, extent_root,
+							&key, path, -1, 1);
+			}
+
 			if (ret) {
-				printk(KERN_ERR "umm, got %d back from search"
-				       ", was looking for %llu\n", ret,
-				       (unsigned long long)bytenr);
-				btrfs_print_leaf(extent_root, path->nodes[0]);
+				btrfs_err(info, "umm, got %d back from search, was looking for %llu",
+					ret, bytenr);
+				if (ret > 0)
+					btrfs_print_leaf(extent_root,
+							 path->nodes[0]);
+			}
+			if (ret < 0) {
+				btrfs_abort_transaction(trans, extent_root, ret);
+				goto out;
 			}
-			BUG_ON(ret);
 			extent_slot = path->slots[0];
 		}
-	} else {
+	} else if (WARN_ON(ret == -ENOENT)) {
 		btrfs_print_leaf(extent_root, path->nodes[0]);
-		WARN_ON(1);
-		printk(KERN_ERR "btrfs unable to find ref byte nr %llu "
-		       "parent %llu root %llu  owner %llu offset %llu\n",
-		       (unsigned long long)bytenr,
-		       (unsigned long long)parent,
-		       (unsigned long long)root_objectid,
-		       (unsigned long long)owner_objectid,
-		       (unsigned long long)owner_offset);
+		btrfs_err(info,
+			"unable to find ref byte nr %llu parent %llu root %llu  owner %llu offset %llu",
+			bytenr, parent, root_objectid, owner_objectid,
+			owner_offset);
+		btrfs_abort_transaction(trans, extent_root, ret);
+		goto out;
+	} else {
+		btrfs_abort_transaction(trans, extent_root, ret);
+		goto out;
 	}
 
 	leaf = path->nodes[0];
@@ -3835,9 +5972,12 @@ static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
 		BUG_ON(found_extent || extent_slot != path->slots[0]);
 		ret = convert_extent_item_v0(trans, extent_root, path,
 					     owner_objectid, 0);
-		BUG_ON(ret < 0);
+		if (ret < 0) {
+			btrfs_abort_transaction(trans, extent_root, ret);
+			goto out;
+		}
 
-		btrfs_release_path(extent_root, path);
+		btrfs_release_path(path);
 		path->leave_spinning = 1;
 
 		key.objectid = bytenr;
@@ -3847,12 +5987,15 @@ static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
 		ret = btrfs_search_slot(trans, extent_root, &key, path,
 					-1, 1);
 		if (ret) {
-			printk(KERN_ERR "umm, got %d back from search"
-			       ", was looking for %llu\n", ret,
-			       (unsigned long long)bytenr);
+			btrfs_err(info, "umm, got %d back from search, was looking for %llu",
+				ret, bytenr);
 			btrfs_print_leaf(extent_root, path->nodes[0]);
 		}
-		BUG_ON(ret);
+		if (ret < 0) {
+			btrfs_abort_transaction(trans, extent_root, ret);
+			goto out;
+		}
+
 		extent_slot = path->slots[0];
 		leaf = path->nodes[0];
 		item_size = btrfs_item_size_nr(leaf, extent_slot);
@@ -3861,7 +6004,8 @@ static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
 	BUG_ON(item_size < sizeof(*ei));
 	ei = btrfs_item_ptr(leaf, extent_slot,
 			    struct btrfs_extent_item);
-	if (owner_objectid < BTRFS_FIRST_FREE_OBJECTID) {
+	if (owner_objectid < BTRFS_FIRST_FREE_OBJECTID &&
+	    key.type == BTRFS_EXTENT_ITEM_KEY) {
 		struct btrfs_tree_block_info *bi;
 		BUG_ON(item_size < sizeof(*ei) + sizeof(*bi));
 		bi = (struct btrfs_tree_block_info *)(ei + 1);
@@ -3869,10 +6013,17 @@ static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
 	}
 
 	refs = btrfs_extent_refs(leaf, ei);
-	BUG_ON(refs < refs_to_drop);
+	if (refs < refs_to_drop) {
+		btrfs_err(info, "trying to drop %d refs but we only have %Lu "
+			  "for bytenr %Lu", refs_to_drop, refs, bytenr);
+		ret = -EINVAL;
+		btrfs_abort_transaction(trans, extent_root, ret);
+		goto out;
+	}
 	refs -= refs_to_drop;
 
 	if (refs > 0) {
+		type = BTRFS_QGROUP_OPER_SUB_SHARED;
 		if (extent_op)
 			__run_delayed_extent_op(extent_op, leaf, ei);
 		/*
@@ -3888,13 +6039,15 @@ static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
 		if (found_extent) {
 			ret = remove_extent_backref(trans, extent_root, path,
 						    iref, refs_to_drop,
-						    is_data);
-			BUG_ON(ret);
+						    is_data, &last_ref);
+			if (ret) {
+				btrfs_abort_transaction(trans, extent_root, ret);
+				goto out;
+			}
 		}
+		add_pinned_bytes(root->fs_info, -num_bytes, owner_objectid,
+				 root_objectid);
 	} else {
-		int mark_free = 0;
-		struct extent_buffer *must_clean = NULL;
-
 		if (found_extent) {
 			BUG_ON(is_data && refs_to_drop !=
 			       extent_data_ref_count(root, path, iref));
@@ -3907,50 +6060,50 @@ static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
 			}
 		}
 
-		ret = pin_down_bytes(trans, root, path, bytenr,
-				     num_bytes, is_data, 0, &must_clean);
-		if (ret > 0)
-			mark_free = 1;
-		BUG_ON(ret < 0);
-		/*
-		 * it is going to be very rare for someone to be waiting
-		 * on the block we're freeing.  del_items might need to
-		 * schedule, so rather than get fancy, just force it
-		 * to blocking here
-		 */
-		if (must_clean)
-			btrfs_set_lock_blocking(must_clean);
-
+		last_ref = 1;
 		ret = btrfs_del_items(trans, extent_root, path, path->slots[0],
 				      num_to_del);
-		BUG_ON(ret);
-		btrfs_release_path(extent_root, path);
-
-		if (must_clean) {
-			clean_tree_block(NULL, root, must_clean);
-			btrfs_tree_unlock(must_clean);
-			free_extent_buffer(must_clean);
+		if (ret) {
+			btrfs_abort_transaction(trans, extent_root, ret);
+			goto out;
 		}
+		btrfs_release_path(path);
 
 		if (is_data) {
 			ret = btrfs_del_csums(trans, root, bytenr, num_bytes);
-			BUG_ON(ret);
-		} else {
-			invalidate_mapping_pages(info->btree_inode->i_mapping,
-			     bytenr >> PAGE_CACHE_SHIFT,
-			     (bytenr + num_bytes - 1) >> PAGE_CACHE_SHIFT);
+			if (ret) {
+				btrfs_abort_transaction(trans, extent_root, ret);
+				goto out;
+			}
+		}
+
+		ret = update_block_group(root, bytenr, num_bytes, 0);
+		if (ret) {
+			btrfs_abort_transaction(trans, extent_root, ret);
+			goto out;
 		}
+	}
+	btrfs_release_path(path);
+
+	/* Deal with the quota accounting */
+	if (!ret && last_ref && !no_quota) {
+		int mod_seq = 0;
+
+		if (owner_objectid >= BTRFS_FIRST_FREE_OBJECTID &&
+		    type == BTRFS_QGROUP_OPER_SUB_SHARED)
+			mod_seq = 1;
 
-		ret = update_block_group(trans, root, bytenr, num_bytes, 0,
-					 mark_free);
-		BUG_ON(ret);
+		ret = btrfs_qgroup_record_ref(trans, info, root_objectid,
+					      bytenr, num_bytes, type,
+					      mod_seq);
 	}
+out:
 	btrfs_free_path(path);
 	return ret;
 }
 
 /*
- * when we free an extent, it is possible (and likely) that we free the last
+ * when we free an block, it is possible (and likely) that we free the last
  * delayed ref for that extent as well.  This searches the delayed ref tree for
  * a given extent, and if there are no other delayed refs to be processed, it
  * removes it from the tree.
@@ -3960,30 +6113,22 @@ static noinline int check_ref_cleanup(struct btrfs_trans_handle *trans,
 {
 	struct btrfs_delayed_ref_head *head;
 	struct btrfs_delayed_ref_root *delayed_refs;
-	struct btrfs_delayed_ref_node *ref;
-	struct rb_node *node;
-	int ret;
+	int ret = 0;
 
 	delayed_refs = &trans->transaction->delayed_refs;
 	spin_lock(&delayed_refs->lock);
 	head = btrfs_find_delayed_ref_head(trans, bytenr);
 	if (!head)
-		goto out;
+		goto out_delayed_unlock;
 
-	node = rb_prev(&head->node.rb_node);
-	if (!node)
-		goto out;
-
-	ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);
-
-	/* there are still entries for this ref, we can't drop it */
-	if (ref->bytenr == bytenr)
+	spin_lock(&head->lock);
+	if (rb_first(&head->ref_root))
 		goto out;
 
 	if (head->extent_op) {
 		if (!head->must_insert_reserved)
 			goto out;
-		kfree(head->extent_op);
+		btrfs_free_delayed_extent_op(head->extent_op);
 		head->extent_op = NULL;
 	}
 
@@ -3999,38 +6144,105 @@ static noinline int check_ref_cleanup(struct btrfs_trans_handle *trans,
 	 * ahead and process it.
 	 */
 	head->node.in_tree = 0;
-	rb_erase(&head->node.rb_node, &delayed_refs->root);
+	rb_erase(&head->href_node, &delayed_refs->href_root);
 
-	delayed_refs->num_entries--;
+	atomic_dec(&delayed_refs->num_entries);
 
 	/*
 	 * we don't take a ref on the node because we're removing it from the
 	 * tree, so we just steal the ref the tree was holding.
 	 */
 	delayed_refs->num_heads--;
-	if (list_empty(&head->cluster))
+	if (head->processing == 0)
 		delayed_refs->num_heads_ready--;
-
-	list_del_init(&head->cluster);
+	head->processing = 0;
+	spin_unlock(&head->lock);
 	spin_unlock(&delayed_refs->lock);
 
-	ret = run_one_delayed_ref(trans, root->fs_info->tree_root,
-				  &head->node, head->extent_op,
-				  head->must_insert_reserved);
-	BUG_ON(ret);
+	BUG_ON(head->extent_op);
+	if (head->must_insert_reserved)
+		ret = 1;
+
+	mutex_unlock(&head->mutex);
 	btrfs_put_delayed_ref(&head->node);
-	return 0;
+	return ret;
 out:
+	spin_unlock(&head->lock);
+
+out_delayed_unlock:
 	spin_unlock(&delayed_refs->lock);
 	return 0;
 }
 
-int btrfs_free_extent(struct btrfs_trans_handle *trans,
-		      struct btrfs_root *root,
-		      u64 bytenr, u64 num_bytes, u64 parent,
-		      u64 root_objectid, u64 owner, u64 offset)
+void btrfs_free_tree_block(struct btrfs_trans_handle *trans,
+			   struct btrfs_root *root,
+			   struct extent_buffer *buf,
+			   u64 parent, int last_ref)
+{
+	struct btrfs_block_group_cache *cache = NULL;
+	int pin = 1;
+	int ret;
+
+	if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
+		ret = btrfs_add_delayed_tree_ref(root->fs_info, trans,
+					buf->start, buf->len,
+					parent, root->root_key.objectid,
+					btrfs_header_level(buf),
+					BTRFS_DROP_DELAYED_REF, NULL, 0);
+		BUG_ON(ret); /* -ENOMEM */
+	}
+
+	if (!last_ref)
+		return;
+
+	cache = btrfs_lookup_block_group(root->fs_info, buf->start);
+
+	if (btrfs_header_generation(buf) == trans->transid) {
+		if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
+			ret = check_ref_cleanup(trans, root, buf->start);
+			if (!ret)
+				goto out;
+		}
+
+		if (btrfs_header_flag(buf, BTRFS_HEADER_FLAG_WRITTEN)) {
+			pin_down_extent(root, cache, buf->start, buf->len, 1);
+			goto out;
+		}
+
+		WARN_ON(test_bit(EXTENT_BUFFER_DIRTY, &buf->bflags));
+
+		btrfs_add_free_space(cache, buf->start, buf->len);
+		btrfs_update_reserved_bytes(cache, buf->len, RESERVE_FREE, 0);
+		trace_btrfs_reserved_extent_free(root, buf->start, buf->len);
+		pin = 0;
+	}
+out:
+	if (pin)
+		add_pinned_bytes(root->fs_info, buf->len,
+				 btrfs_header_level(buf),
+				 root->root_key.objectid);
+
+	/*
+	 * Deleting the buffer, clear the corrupt flag since it doesn't matter
+	 * anymore.
+	 */
+	clear_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags);
+	btrfs_put_block_group(cache);
+}
+
+/* Can return -ENOMEM */
+int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+		      u64 bytenr, u64 num_bytes, u64 parent, u64 root_objectid,
+		      u64 owner, u64 offset, int no_quota)
 {
 	int ret;
+	struct btrfs_fs_info *fs_info = root->fs_info;
+
+#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
+	if (unlikely(test_bit(BTRFS_ROOT_DUMMY_ROOT, &root->state)))
+		return 0;
+#endif
+	add_pinned_bytes(root->fs_info, num_bytes, owner, root_objectid);
 
 	/*
 	 * tree log blocks never actually go into the extent allocation
@@ -4042,40 +6254,25 @@ int btrfs_free_extent(struct btrfs_trans_handle *trans,
 		btrfs_pin_extent(root, bytenr, num_bytes, 1);
 		ret = 0;
 	} else if (owner < BTRFS_FIRST_FREE_OBJECTID) {
-		ret = btrfs_add_delayed_tree_ref(trans, bytenr, num_bytes,
+		ret = btrfs_add_delayed_tree_ref(fs_info, trans, bytenr,
+					num_bytes,
 					parent, root_objectid, (int)owner,
-					BTRFS_DROP_DELAYED_REF, NULL);
-		BUG_ON(ret);
-		ret = check_ref_cleanup(trans, root, bytenr);
-		BUG_ON(ret);
+					BTRFS_DROP_DELAYED_REF, NULL, no_quota);
 	} else {
-		ret = btrfs_add_delayed_data_ref(trans, bytenr, num_bytes,
-					parent, root_objectid, owner,
-					offset, BTRFS_DROP_DELAYED_REF, NULL);
-		BUG_ON(ret);
+		ret = btrfs_add_delayed_data_ref(fs_info, trans, bytenr,
+						num_bytes,
+						parent, root_objectid, owner,
+						offset, BTRFS_DROP_DELAYED_REF,
+						NULL, no_quota);
 	}
 	return ret;
 }
 
-int btrfs_free_tree_block(struct btrfs_trans_handle *trans,
-			  struct btrfs_root *root,
-			  u64 bytenr, u32 blocksize,
-			  u64 parent, u64 root_objectid, int level)
-{
-	u64 used;
-	spin_lock(&root->node_lock);
-	used = btrfs_root_used(&root->root_item) - blocksize;
-	btrfs_set_root_used(&root->root_item, used);
-	spin_unlock(&root->node_lock);
-
-	return btrfs_free_extent(trans, root, bytenr, blocksize,
-				 parent, root_objectid, level, 0);
-}
-
-static u64 stripe_align(struct btrfs_root *root, u64 val)
+static u64 stripe_align(struct btrfs_root *root,
+			struct btrfs_block_group_cache *cache,
+			u64 val, u64 num_bytes)
 {
-	u64 mask = ((u64)root->stripesize - 1);
-	u64 ret = (val + mask) & ~mask;
+	u64 ret = ALIGN(val, root->stripesize);
 	return ret;
 }
 
@@ -4089,102 +6286,216 @@ static u64 stripe_align(struct btrfs_root *root, u64 val)
  * for our min num_bytes.  Another option is to have it go ahead
  * and look in the rbtree for a free extent of a given size, but this
  * is a good start.
+ *
+ * Callers of this must check if cache->cached == BTRFS_CACHE_ERROR before using
+ * any of the information in this block group.
  */
-static noinline int
+static noinline void
 wait_block_group_cache_progress(struct btrfs_block_group_cache *cache,
 				u64 num_bytes)
 {
 	struct btrfs_caching_control *caching_ctl;
-	DEFINE_WAIT(wait);
 
 	caching_ctl = get_caching_control(cache);
 	if (!caching_ctl)
-		return 0;
+		return;
 
 	wait_event(caching_ctl->wait, block_group_cache_done(cache) ||
-		   (cache->free_space >= num_bytes));
+		   (cache->free_space_ctl->free_space >= num_bytes));
 
 	put_caching_control(caching_ctl);
-	return 0;
 }
 
 static noinline int
 wait_block_group_cache_done(struct btrfs_block_group_cache *cache)
 {
 	struct btrfs_caching_control *caching_ctl;
-	DEFINE_WAIT(wait);
+	int ret = 0;
 
 	caching_ctl = get_caching_control(cache);
 	if (!caching_ctl)
-		return 0;
+		return (cache->cached == BTRFS_CACHE_ERROR) ? -EIO : 0;
 
 	wait_event(caching_ctl->wait, block_group_cache_done(cache));
-
+	if (cache->cached == BTRFS_CACHE_ERROR)
+		ret = -EIO;
 	put_caching_control(caching_ctl);
-	return 0;
+	return ret;
+}
+
+int __get_raid_index(u64 flags)
+{
+	if (flags & BTRFS_BLOCK_GROUP_RAID10)
+		return BTRFS_RAID_RAID10;
+	else if (flags & BTRFS_BLOCK_GROUP_RAID1)
+		return BTRFS_RAID_RAID1;
+	else if (flags & BTRFS_BLOCK_GROUP_DUP)
+		return BTRFS_RAID_DUP;
+	else if (flags & BTRFS_BLOCK_GROUP_RAID0)
+		return BTRFS_RAID_RAID0;
+	else if (flags & BTRFS_BLOCK_GROUP_RAID5)
+		return BTRFS_RAID_RAID5;
+	else if (flags & BTRFS_BLOCK_GROUP_RAID6)
+		return BTRFS_RAID_RAID6;
+
+	return BTRFS_RAID_SINGLE; /* BTRFS_BLOCK_GROUP_SINGLE */
+}
+
+int get_block_group_index(struct btrfs_block_group_cache *cache)
+{
+	return __get_raid_index(cache->flags);
+}
+
+static const char *btrfs_raid_type_names[BTRFS_NR_RAID_TYPES] = {
+	[BTRFS_RAID_RAID10]	= "raid10",
+	[BTRFS_RAID_RAID1]	= "raid1",
+	[BTRFS_RAID_DUP]	= "dup",
+	[BTRFS_RAID_RAID0]	= "raid0",
+	[BTRFS_RAID_SINGLE]	= "single",
+	[BTRFS_RAID_RAID5]	= "raid5",
+	[BTRFS_RAID_RAID6]	= "raid6",
+};
+
+static const char *get_raid_name(enum btrfs_raid_types type)
+{
+	if (type >= BTRFS_NR_RAID_TYPES)
+		return NULL;
+
+	return btrfs_raid_type_names[type];
 }
 
 enum btrfs_loop_type {
-	LOOP_FIND_IDEAL = 0,
-	LOOP_CACHING_NOWAIT = 1,
-	LOOP_CACHING_WAIT = 2,
-	LOOP_ALLOC_CHUNK = 3,
-	LOOP_NO_EMPTY_SIZE = 4,
+	LOOP_CACHING_NOWAIT = 0,
+	LOOP_CACHING_WAIT = 1,
+	LOOP_ALLOC_CHUNK = 2,
+	LOOP_NO_EMPTY_SIZE = 3,
 };
 
+static inline void
+btrfs_lock_block_group(struct btrfs_block_group_cache *cache,
+		       int delalloc)
+{
+	if (delalloc)
+		down_read(&cache->data_rwsem);
+}
+
+static inline void
+btrfs_grab_block_group(struct btrfs_block_group_cache *cache,
+		       int delalloc)
+{
+	btrfs_get_block_group(cache);
+	if (delalloc)
+		down_read(&cache->data_rwsem);
+}
+
+static struct btrfs_block_group_cache *
+btrfs_lock_cluster(struct btrfs_block_group_cache *block_group,
+		   struct btrfs_free_cluster *cluster,
+		   int delalloc)
+{
+	struct btrfs_block_group_cache *used_bg;
+	bool locked = false;
+again:
+	spin_lock(&cluster->refill_lock);
+	if (locked) {
+		if (used_bg == cluster->block_group)
+			return used_bg;
+
+		up_read(&used_bg->data_rwsem);
+		btrfs_put_block_group(used_bg);
+	}
+
+	used_bg = cluster->block_group;
+	if (!used_bg)
+		return NULL;
+
+	if (used_bg == block_group)
+		return used_bg;
+
+	btrfs_get_block_group(used_bg);
+
+	if (!delalloc)
+		return used_bg;
+
+	if (down_read_trylock(&used_bg->data_rwsem))
+		return used_bg;
+
+	spin_unlock(&cluster->refill_lock);
+	down_read(&used_bg->data_rwsem);
+	locked = true;
+	goto again;
+}
+
+static inline void
+btrfs_release_block_group(struct btrfs_block_group_cache *cache,
+			 int delalloc)
+{
+	if (delalloc)
+		up_read(&cache->data_rwsem);
+	btrfs_put_block_group(cache);
+}
+
 /*
  * walks the btree of allocated extents and find a hole of a given size.
  * The key ins is changed to record the hole:
- * ins->objectid == block start
+ * ins->objectid == start position
  * ins->flags = BTRFS_EXTENT_ITEM_KEY
- * ins->offset == number of blocks
+ * ins->offset == the size of the hole.
  * Any available blocks before search_start are skipped.
+ *
+ * If there is no suitable free space, we will record the max size of
+ * the free space extent currently.
  */
-static noinline int find_free_extent(struct btrfs_trans_handle *trans,
-				     struct btrfs_root *orig_root,
+static noinline int find_free_extent(struct btrfs_root *orig_root,
 				     u64 num_bytes, u64 empty_size,
-				     u64 search_start, u64 search_end,
 				     u64 hint_byte, struct btrfs_key *ins,
-				     u64 exclude_start, u64 exclude_nr,
-				     int data)
+				     u64 flags, int delalloc)
 {
 	int ret = 0;
 	struct btrfs_root *root = orig_root->fs_info->extent_root;
 	struct btrfs_free_cluster *last_ptr = NULL;
 	struct btrfs_block_group_cache *block_group = NULL;
+	u64 search_start = 0;
+	u64 max_extent_size = 0;
 	int empty_cluster = 2 * 1024 * 1024;
-	int allowed_chunk_alloc = 0;
-	int done_chunk_alloc = 0;
 	struct btrfs_space_info *space_info;
-	int last_ptr_loop = 0;
 	int loop = 0;
-	bool found_uncached_bg = false;
+	int index = __get_raid_index(flags);
+	int alloc_type = (flags & BTRFS_BLOCK_GROUP_DATA) ?
+		RESERVE_ALLOC_NO_ACCOUNT : RESERVE_ALLOC;
 	bool failed_cluster_refill = false;
 	bool failed_alloc = false;
-	u64 ideal_cache_percent = 0;
-	u64 ideal_cache_offset = 0;
+	bool use_cluster = true;
+	bool have_caching_bg = false;
 
 	WARN_ON(num_bytes < root->sectorsize);
 	btrfs_set_key_type(ins, BTRFS_EXTENT_ITEM_KEY);
 	ins->objectid = 0;
 	ins->offset = 0;
 
-	space_info = __find_space_info(root->fs_info, data);
+	trace_find_free_extent(orig_root, num_bytes, empty_size, flags);
+
+	space_info = __find_space_info(root->fs_info, flags);
 	if (!space_info) {
-		printk(KERN_ERR "No space info for %d\n", data);
+		btrfs_err(root->fs_info, "No space info for %llu", flags);
 		return -ENOSPC;
 	}
 
-	if (orig_root->ref_cows || empty_size)
-		allowed_chunk_alloc = 1;
+	/*
+	 * If the space info is for both data and metadata it means we have a
+	 * small filesystem and we can't use the clustering stuff.
+	 */
+	if (btrfs_mixed_space_info(space_info))
+		use_cluster = false;
 
-	if (data & BTRFS_BLOCK_GROUP_METADATA) {
+	if (flags & BTRFS_BLOCK_GROUP_METADATA && use_cluster) {
 		last_ptr = &root->fs_info->meta_alloc_cluster;
 		if (!btrfs_test_opt(root, SSD))
 			empty_cluster = 64 * 1024;
 	}
 
-	if ((data & BTRFS_BLOCK_GROUP_DATA) && btrfs_test_opt(root, SSD)) {
+	if ((flags & BTRFS_BLOCK_GROUP_DATA) && use_cluster &&
+	    btrfs_test_opt(root, SSD)) {
 		last_ptr = &root->fs_info->data_alloc_cluster;
 	}
 
@@ -4202,7 +6513,6 @@ static noinline int find_free_extent(struct btrfs_trans_handle *trans,
 		empty_cluster = 0;
 
 	if (search_start == hint_byte) {
-ideal_cache:
 		block_group = btrfs_lookup_block_group(root->fs_info,
 						       search_start);
 		/*
@@ -4212,9 +6522,8 @@ ideal_cache:
 		 * However if we are re-searching with an ideal block group
 		 * picked out then we don't care that the block group is cached.
 		 */
-		if (block_group && block_group_bits(block_group, data) &&
-		    (block_group->cached != BTRFS_CACHE_NO ||
-		     search_start == ideal_cache_offset)) {
+		if (block_group && block_group_bits(block_group, flags) &&
+		    block_group->cached != BTRFS_CACHE_NO) {
 			down_read(&space_info->groups_sem);
 			if (list_empty(&block_group->list) ||
 			    block_group->ro) {
@@ -4227,6 +6536,8 @@ ideal_cache:
 				btrfs_put_block_group(block_group);
 				up_read(&space_info->groups_sem);
 			} else {
+				index = get_block_group_index(block_group);
+				btrfs_lock_block_group(block_group, delalloc);
 				goto have_block_group;
 			}
 		} else if (block_group) {
@@ -4234,140 +6545,156 @@ ideal_cache:
 		}
 	}
 search:
+	have_caching_bg = false;
 	down_read(&space_info->groups_sem);
-	list_for_each_entry(block_group, &space_info->block_groups, list) {
+	list_for_each_entry(block_group, &space_info->block_groups[index],
+			    list) {
 		u64 offset;
 		int cached;
 
-		btrfs_get_block_group(block_group);
+		btrfs_grab_block_group(block_group, delalloc);
 		search_start = block_group->key.objectid;
 
-have_block_group:
-		if (unlikely(block_group->cached == BTRFS_CACHE_NO)) {
-			u64 free_percent;
-
-			free_percent = btrfs_block_group_used(&block_group->item);
-			free_percent *= 100;
-			free_percent = div64_u64(free_percent,
-						 block_group->key.offset);
-			free_percent = 100 - free_percent;
-			if (free_percent > ideal_cache_percent &&
-			    likely(!block_group->ro)) {
-				ideal_cache_offset = block_group->key.objectid;
-				ideal_cache_percent = free_percent;
-			}
-
-			/*
-			 * We only want to start kthread caching if we are at
-			 * the point where we will wait for caching to make
-			 * progress, or if our ideal search is over and we've
-			 * found somebody to start caching.
-			 */
-			if (loop > LOOP_CACHING_NOWAIT ||
-			    (loop > LOOP_FIND_IDEAL &&
-			     atomic_read(&space_info->caching_threads) < 2)) {
-				ret = cache_block_group(block_group);
-				BUG_ON(ret);
-			}
-			found_uncached_bg = true;
+		/*
+		 * this can happen if we end up cycling through all the
+		 * raid types, but we want to make sure we only allocate
+		 * for the proper type.
+		 */
+		if (!block_group_bits(block_group, flags)) {
+		    u64 extra = BTRFS_BLOCK_GROUP_DUP |
+				BTRFS_BLOCK_GROUP_RAID1 |
+				BTRFS_BLOCK_GROUP_RAID5 |
+				BTRFS_BLOCK_GROUP_RAID6 |
+				BTRFS_BLOCK_GROUP_RAID10;
 
 			/*
-			 * If loop is set for cached only, try the next block
-			 * group.
+			 * if they asked for extra copies and this block group
+			 * doesn't provide them, bail.  This does allow us to
+			 * fill raid0 from raid1.
 			 */
-			if (loop == LOOP_FIND_IDEAL)
+			if ((flags & extra) && !(block_group->flags & extra))
 				goto loop;
 		}
 
+have_block_group:
 		cached = block_group_cache_done(block_group);
-		if (unlikely(!cached))
-			found_uncached_bg = true;
+		if (unlikely(!cached)) {
+			ret = cache_block_group(block_group, 0);
+			BUG_ON(ret < 0);
+			ret = 0;
+		}
 
+		if (unlikely(block_group->cached == BTRFS_CACHE_ERROR))
+			goto loop;
 		if (unlikely(block_group->ro))
 			goto loop;
 
 		/*
-		 * Ok we want to try and use the cluster allocator, so lets look
-		 * there, unless we are on LOOP_NO_EMPTY_SIZE, since we will
-		 * have tried the cluster allocator plenty of times at this
-		 * point and not have found anything, so we are likely way too
-		 * fragmented for the clustering stuff to find anything, so lets
-		 * just skip it and let the allocator find whatever block it can
-		 * find
+		 * Ok we want to try and use the cluster allocator, so
+		 * lets look there
 		 */
-		if (last_ptr && loop < LOOP_NO_EMPTY_SIZE) {
+		if (last_ptr) {
+			struct btrfs_block_group_cache *used_block_group;
+			unsigned long aligned_cluster;
 			/*
 			 * the refill lock keeps out other
 			 * people trying to start a new cluster
 			 */
-			spin_lock(&last_ptr->refill_lock);
-			if (last_ptr->block_group &&
-			    (last_ptr->block_group->ro ||
-			    !block_group_bits(last_ptr->block_group, data))) {
-				offset = 0;
+			used_block_group = btrfs_lock_cluster(block_group,
+							      last_ptr,
+							      delalloc);
+			if (!used_block_group)
 				goto refill_cluster;
-			}
 
-			offset = btrfs_alloc_from_cluster(block_group, last_ptr,
-						 num_bytes, search_start);
+			if (used_block_group != block_group &&
+			    (used_block_group->ro ||
+			     !block_group_bits(used_block_group, flags)))
+				goto release_cluster;
+
+			offset = btrfs_alloc_from_cluster(used_block_group,
+						last_ptr,
+						num_bytes,
+						used_block_group->key.objectid,
+						&max_extent_size);
 			if (offset) {
 				/* we have a block, we're done */
 				spin_unlock(&last_ptr->refill_lock);
+				trace_btrfs_reserve_extent_cluster(root,
+						used_block_group,
+						search_start, num_bytes);
+				if (used_block_group != block_group) {
+					btrfs_release_block_group(block_group,
+								  delalloc);
+					block_group = used_block_group;
+				}
 				goto checks;
 			}
 
-			spin_lock(&last_ptr->lock);
-			/*
-			 * whoops, this cluster doesn't actually point to
-			 * this block group.  Get a ref on the block
-			 * group is does point to and try again
-			 */
-			if (!last_ptr_loop && last_ptr->block_group &&
-			    last_ptr->block_group != block_group) {
-
-				btrfs_put_block_group(block_group);
-				block_group = last_ptr->block_group;
-				btrfs_get_block_group(block_group);
-				spin_unlock(&last_ptr->lock);
+			WARN_ON(last_ptr->block_group != used_block_group);
+release_cluster:
+			/* If we are on LOOP_NO_EMPTY_SIZE, we can't
+			 * set up a new clusters, so lets just skip it
+			 * and let the allocator find whatever block
+			 * it can find.  If we reach this point, we
+			 * will have tried the cluster allocator
+			 * plenty of times and not have found
+			 * anything, so we are likely way too
+			 * fragmented for the clustering stuff to find
+			 * anything.
+			 *
+			 * However, if the cluster is taken from the
+			 * current block group, release the cluster
+			 * first, so that we stand a better chance of
+			 * succeeding in the unclustered
+			 * allocation.  */
+			if (loop >= LOOP_NO_EMPTY_SIZE &&
+			    used_block_group != block_group) {
 				spin_unlock(&last_ptr->refill_lock);
-
-				last_ptr_loop = 1;
-				search_start = block_group->key.objectid;
-				/*
-				 * we know this block group is properly
-				 * in the list because
-				 * btrfs_remove_block_group, drops the
-				 * cluster before it removes the block
-				 * group from the list
-				 */
-				goto have_block_group;
+				btrfs_release_block_group(used_block_group,
+							  delalloc);
+				goto unclustered_alloc;
 			}
-			spin_unlock(&last_ptr->lock);
-refill_cluster:
+
 			/*
 			 * this cluster didn't work out, free it and
 			 * start over
 			 */
 			btrfs_return_cluster_to_free_space(NULL, last_ptr);
 
-			last_ptr_loop = 0;
+			if (used_block_group != block_group)
+				btrfs_release_block_group(used_block_group,
+							  delalloc);
+refill_cluster:
+			if (loop >= LOOP_NO_EMPTY_SIZE) {
+				spin_unlock(&last_ptr->refill_lock);
+				goto unclustered_alloc;
+			}
+
+			aligned_cluster = max_t(unsigned long,
+						empty_cluster + empty_size,
+					      block_group->full_stripe_len);
 
 			/* allocate a cluster in this block group */
-			ret = btrfs_find_space_cluster(trans, root,
-					       block_group, last_ptr,
-					       offset, num_bytes,
-					       empty_cluster + empty_size);
+			ret = btrfs_find_space_cluster(root, block_group,
+						       last_ptr, search_start,
+						       num_bytes,
+						       aligned_cluster);
 			if (ret == 0) {
 				/*
 				 * now pull our allocation out of this
 				 * cluster
 				 */
 				offset = btrfs_alloc_from_cluster(block_group,
-						  last_ptr, num_bytes,
-						  search_start);
+							last_ptr,
+							num_bytes,
+							search_start,
+							&max_extent_size);
 				if (offset) {
 					/* we found one, proceed */
 					spin_unlock(&last_ptr->refill_lock);
+					trace_btrfs_reserve_extent_cluster(root,
+						block_group, search_start,
+						num_bytes);
 					goto checks;
 				}
 			} else if (!cached && loop > LOOP_CACHING_NOWAIT
@@ -4391,8 +6718,23 @@ refill_cluster:
 			goto loop;
 		}
 
+unclustered_alloc:
+		spin_lock(&block_group->free_space_ctl->tree_lock);
+		if (cached &&
+		    block_group->free_space_ctl->free_space <
+		    num_bytes + empty_cluster + empty_size) {
+			if (block_group->free_space_ctl->free_space >
+			    max_extent_size)
+				max_extent_size =
+					block_group->free_space_ctl->free_space;
+			spin_unlock(&block_group->free_space_ctl->tree_lock);
+			goto loop;
+		}
+		spin_unlock(&block_group->free_space_ctl->tree_lock);
+
 		offset = btrfs_find_space_for_alloc(block_group, search_start,
-						    num_bytes, empty_size);
+						    num_bytes, empty_size,
+						    &max_extent_size);
 		/*
 		 * If we didn't find a chunk, and we haven't failed on this
 		 * block group before, and this block group is in the middle of
@@ -4409,15 +6751,13 @@ refill_cluster:
 			failed_alloc = true;
 			goto have_block_group;
 		} else if (!offset) {
+			if (!cached)
+				have_caching_bg = true;
 			goto loop;
 		}
 checks:
-		search_start = stripe_align(root, offset);
-		/* move on to the next group */
-		if (search_start + num_bytes >= search_end) {
-			btrfs_add_free_space(block_group, offset, num_bytes);
-			goto loop;
-		}
+		search_start = stripe_align(root, block_group,
+					    offset, num_bytes);
 
 		/* move on to the next group */
 		if (search_start + num_bytes >
@@ -4426,45 +6766,41 @@ checks:
 			goto loop;
 		}
 
-		if (exclude_nr > 0 &&
-		    (search_start + num_bytes > exclude_start &&
-		     search_start < exclude_start + exclude_nr)) {
-			search_start = exclude_start + exclude_nr;
+		if (offset < search_start)
+			btrfs_add_free_space(block_group, offset,
+					     search_start - offset);
+		BUG_ON(offset > search_start);
 
+		ret = btrfs_update_reserved_bytes(block_group, num_bytes,
+						  alloc_type, delalloc);
+		if (ret == -EAGAIN) {
 			btrfs_add_free_space(block_group, offset, num_bytes);
-			/*
-			 * if search_start is still in this block group
-			 * then we just re-search this block group
-			 */
-			if (search_start >= block_group->key.objectid &&
-			    search_start < (block_group->key.objectid +
-					    block_group->key.offset))
-				goto have_block_group;
 			goto loop;
 		}
 
+		/* we are all good, lets return */
 		ins->objectid = search_start;
 		ins->offset = num_bytes;
 
-		if (offset < search_start)
-			btrfs_add_free_space(block_group, offset,
-					     search_start - offset);
-		BUG_ON(offset > search_start);
-
-		update_reserved_extents(block_group, num_bytes, 1);
-
-		/* we are all good, lets return */
+		trace_btrfs_reserve_extent(orig_root, block_group,
+					   search_start, num_bytes);
+		btrfs_release_block_group(block_group, delalloc);
 		break;
 loop:
 		failed_cluster_refill = false;
 		failed_alloc = false;
-		btrfs_put_block_group(block_group);
+		BUG_ON(index != get_block_group_index(block_group));
+		btrfs_release_block_group(block_group, delalloc);
 	}
 	up_read(&space_info->groups_sem);
 
-	/* LOOP_FIND_IDEAL, only search caching/cached bg's, and don't wait for
-	 *			for them to make caching progress.  Also
-	 *			determine the best possible bg to cache
+	if (!ins->objectid && loop >= LOOP_CACHING_WAIT && have_caching_bg)
+		goto search;
+
+	if (!ins->objectid && ++index < BTRFS_NR_RAID_TYPES)
+		goto search;
+
+	/*
 	 * LOOP_CACHING_NOWAIT, search partially cached block groups, kicking
 	 *			caching kthreads as we move along
 	 * LOOP_CACHING_WAIT, search everything, and wait if our bg is caching
@@ -4472,84 +6808,55 @@ loop:
 	 * LOOP_NO_EMPTY_SIZE, set empty_size and empty_cluster to 0 and try
 	 *			again
 	 */
-	if (!ins->objectid && loop < LOOP_NO_EMPTY_SIZE &&
-	    (found_uncached_bg || empty_size || empty_cluster ||
-	     allowed_chunk_alloc)) {
-		if (loop == LOOP_FIND_IDEAL && found_uncached_bg) {
-			found_uncached_bg = false;
-			loop++;
-			if (!ideal_cache_percent &&
-			    atomic_read(&space_info->caching_threads))
-				goto search;
+	if (!ins->objectid && loop < LOOP_NO_EMPTY_SIZE) {
+		index = 0;
+		loop++;
+		if (loop == LOOP_ALLOC_CHUNK) {
+			struct btrfs_trans_handle *trans;
+			int exist = 0;
 
+			trans = current->journal_info;
+			if (trans)
+				exist = 1;
+			else
+				trans = btrfs_join_transaction(root);
+
+			if (IS_ERR(trans)) {
+				ret = PTR_ERR(trans);
+				goto out;
+			}
+
+			ret = do_chunk_alloc(trans, root, flags,
+					     CHUNK_ALLOC_FORCE);
 			/*
-			 * 1 of the following 2 things have happened so far
-			 *
-			 * 1) We found an ideal block group for caching that
-			 * is mostly full and will cache quickly, so we might
-			 * as well wait for it.
-			 *
-			 * 2) We searched for cached only and we didn't find
-			 * anything, and we didn't start any caching kthreads
-			 * either, so chances are we will loop through and
-			 * start a couple caching kthreads, and then come back
-			 * around and just wait for them.  This will be slower
-			 * because we will have 2 caching kthreads reading at
-			 * the same time when we could have just started one
-			 * and waited for it to get far enough to give us an
-			 * allocation, so go ahead and go to the wait caching
-			 * loop.
-			 */
-			loop = LOOP_CACHING_WAIT;
-			search_start = ideal_cache_offset;
-			ideal_cache_percent = 0;
-			goto ideal_cache;
-		} else if (loop == LOOP_FIND_IDEAL) {
-			/*
-			 * Didn't find a uncached bg, wait on anything we find
-			 * next.
+			 * Do not bail out on ENOSPC since we
+			 * can do more things.
 			 */
-			loop = LOOP_CACHING_WAIT;
-			goto search;
-		}
-
-		if (loop < LOOP_CACHING_WAIT) {
-			loop++;
-			goto search;
+			if (ret < 0 && ret != -ENOSPC)
+				btrfs_abort_transaction(trans,
+							root, ret);
+			else
+				ret = 0;
+			if (!exist)
+				btrfs_end_transaction(trans, root);
+			if (ret)
+				goto out;
 		}
 
-		if (loop == LOOP_ALLOC_CHUNK) {
+		if (loop == LOOP_NO_EMPTY_SIZE) {
 			empty_size = 0;
 			empty_cluster = 0;
 		}
 
-		if (allowed_chunk_alloc) {
-			ret = do_chunk_alloc(trans, root, num_bytes +
-					     2 * 1024 * 1024, data, 1);
-			allowed_chunk_alloc = 0;
-			done_chunk_alloc = 1;
-		} else if (!done_chunk_alloc) {
-			space_info->force_alloc = 1;
-		}
-
-		if (loop < LOOP_NO_EMPTY_SIZE) {
-			loop++;
-			goto search;
-		}
-		ret = -ENOSPC;
+		goto search;
 	} else if (!ins->objectid) {
 		ret = -ENOSPC;
-	}
-
-	/* we found what we needed */
-	if (ins->objectid) {
-		if (!(data & BTRFS_BLOCK_GROUP_DATA))
-			trans->block_group = block_group->key.objectid;
-
-		btrfs_put_block_group(block_group);
+	} else if (ins->objectid) {
 		ret = 0;
 	}
-
+out:
+	if (ret == -ENOSPC)
+		ins->offset = max_extent_size;
 	return ret;
 }
 
@@ -4557,113 +6864,121 @@ static void dump_space_info(struct btrfs_space_info *info, u64 bytes,
 			    int dump_block_groups)
 {
 	struct btrfs_block_group_cache *cache;
+	int index = 0;
 
 	spin_lock(&info->lock);
-	printk(KERN_INFO "space_info has %llu free, is %sfull\n",
-	       (unsigned long long)(info->total_bytes - info->bytes_used -
-				    info->bytes_pinned - info->bytes_reserved -
-				    info->bytes_super),
+	printk(KERN_INFO "BTRFS: space_info %llu has %llu free, is %sfull\n",
+	       info->flags,
+	       info->total_bytes - info->bytes_used - info->bytes_pinned -
+	       info->bytes_reserved - info->bytes_readonly,
 	       (info->full) ? "" : "not ");
-	printk(KERN_INFO "space_info total=%llu, pinned=%llu, delalloc=%llu,"
-	       " may_use=%llu, used=%llu, root=%llu, super=%llu, reserved=%llu"
-	       "\n",
-	       (unsigned long long)info->total_bytes,
-	       (unsigned long long)info->bytes_pinned,
-	       (unsigned long long)info->bytes_delalloc,
-	       (unsigned long long)info->bytes_may_use,
-	       (unsigned long long)info->bytes_used,
-	       (unsigned long long)info->bytes_root,
-	       (unsigned long long)info->bytes_super,
-	       (unsigned long long)info->bytes_reserved);
+	printk(KERN_INFO "BTRFS: space_info total=%llu, used=%llu, pinned=%llu, "
+	       "reserved=%llu, may_use=%llu, readonly=%llu\n",
+	       info->total_bytes, info->bytes_used, info->bytes_pinned,
+	       info->bytes_reserved, info->bytes_may_use,
+	       info->bytes_readonly);
 	spin_unlock(&info->lock);
 
 	if (!dump_block_groups)
 		return;
 
 	down_read(&info->groups_sem);
-	list_for_each_entry(cache, &info->block_groups, list) {
+again:
+	list_for_each_entry(cache, &info->block_groups[index], list) {
 		spin_lock(&cache->lock);
-		printk(KERN_INFO "block group %llu has %llu bytes, %llu used "
-		       "%llu pinned %llu reserved\n",
-		       (unsigned long long)cache->key.objectid,
-		       (unsigned long long)cache->key.offset,
-		       (unsigned long long)btrfs_block_group_used(&cache->item),
-		       (unsigned long long)cache->pinned,
-		       (unsigned long long)cache->reserved);
+		printk(KERN_INFO "BTRFS: "
+			   "block group %llu has %llu bytes, "
+			   "%llu used %llu pinned %llu reserved %s\n",
+		       cache->key.objectid, cache->key.offset,
+		       btrfs_block_group_used(&cache->item), cache->pinned,
+		       cache->reserved, cache->ro ? "[readonly]" : "");
 		btrfs_dump_free_space(cache, bytes);
 		spin_unlock(&cache->lock);
 	}
+	if (++index < BTRFS_NR_RAID_TYPES)
+		goto again;
 	up_read(&info->groups_sem);
 }
 
-int btrfs_reserve_extent(struct btrfs_trans_handle *trans,
-			 struct btrfs_root *root,
+int btrfs_reserve_extent(struct btrfs_root *root,
 			 u64 num_bytes, u64 min_alloc_size,
 			 u64 empty_size, u64 hint_byte,
-			 u64 search_end, struct btrfs_key *ins,
-			 u64 data)
+			 struct btrfs_key *ins, int is_data, int delalloc)
 {
+	bool final_tried = false;
+	u64 flags;
 	int ret;
-	u64 search_start = 0;
 
-	data = btrfs_get_alloc_profile(root, data);
+	flags = btrfs_get_alloc_profile(root, is_data);
 again:
-	/*
-	 * the only place that sets empty_size is btrfs_realloc_node, which
-	 * is not called recursively on allocations
-	 */
-	if (empty_size || root->ref_cows)
-		ret = do_chunk_alloc(trans, root->fs_info->extent_root,
-				     num_bytes + 2 * 1024 * 1024, data, 0);
-
 	WARN_ON(num_bytes < root->sectorsize);
-	ret = find_free_extent(trans, root, num_bytes, empty_size,
-			       search_start, search_end, hint_byte, ins,
-			       trans->alloc_exclude_start,
-			       trans->alloc_exclude_nr, data);
-
-	if (ret == -ENOSPC && num_bytes > min_alloc_size) {
-		num_bytes = num_bytes >> 1;
-		num_bytes = num_bytes & ~(root->sectorsize - 1);
-		num_bytes = max(num_bytes, min_alloc_size);
-		do_chunk_alloc(trans, root->fs_info->extent_root,
-			       num_bytes, data, 1);
-		goto again;
-	}
+	ret = find_free_extent(root, num_bytes, empty_size, hint_byte, ins,
+			       flags, delalloc);
+
 	if (ret == -ENOSPC) {
-		struct btrfs_space_info *sinfo;
+		if (!final_tried && ins->offset) {
+			num_bytes = min(num_bytes >> 1, ins->offset);
+			num_bytes = round_down(num_bytes, root->sectorsize);
+			num_bytes = max(num_bytes, min_alloc_size);
+			if (num_bytes == min_alloc_size)
+				final_tried = true;
+			goto again;
+		} else if (btrfs_test_opt(root, ENOSPC_DEBUG)) {
+			struct btrfs_space_info *sinfo;
 
-		sinfo = __find_space_info(root->fs_info, data);
-		printk(KERN_ERR "btrfs allocation failed flags %llu, "
-		       "wanted %llu\n", (unsigned long long)data,
-		       (unsigned long long)num_bytes);
-		dump_space_info(sinfo, num_bytes, 1);
+			sinfo = __find_space_info(root->fs_info, flags);
+			btrfs_err(root->fs_info, "allocation failed flags %llu, wanted %llu",
+				flags, num_bytes);
+			if (sinfo)
+				dump_space_info(sinfo, num_bytes, 1);
+		}
 	}
 
 	return ret;
 }
 
-int btrfs_free_reserved_extent(struct btrfs_root *root, u64 start, u64 len)
+static int __btrfs_free_reserved_extent(struct btrfs_root *root,
+					u64 start, u64 len,
+					int pin, int delalloc)
 {
 	struct btrfs_block_group_cache *cache;
 	int ret = 0;
 
 	cache = btrfs_lookup_block_group(root->fs_info, start);
 	if (!cache) {
-		printk(KERN_ERR "Unable to find block group for %llu\n",
-		       (unsigned long long)start);
+		btrfs_err(root->fs_info, "Unable to find block group for %llu",
+			start);
 		return -ENOSPC;
 	}
 
-	ret = btrfs_discard_extent(root, start, len);
+	if (btrfs_test_opt(root, DISCARD))
+		ret = btrfs_discard_extent(root, start, len, NULL);
 
-	btrfs_add_free_space(cache, start, len);
-	update_reserved_extents(cache, len, 0);
+	if (pin)
+		pin_down_extent(root, cache, start, len, 1);
+	else {
+		btrfs_add_free_space(cache, start, len);
+		btrfs_update_reserved_bytes(cache, len, RESERVE_FREE, delalloc);
+	}
 	btrfs_put_block_group(cache);
 
+	trace_btrfs_reserved_extent_free(root, start, len);
+
 	return ret;
 }
 
+int btrfs_free_reserved_extent(struct btrfs_root *root,
+			       u64 start, u64 len, int delalloc)
+{
+	return __btrfs_free_reserved_extent(root, start, len, 0, delalloc);
+}
+
+int btrfs_free_and_pin_reserved_extent(struct btrfs_root *root,
+				       u64 start, u64 len)
+{
+	return __btrfs_free_reserved_extent(root, start, len, 1, 0);
+}
+
 static int alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
 				      struct btrfs_root *root,
 				      u64 parent, u64 root_objectid,
@@ -4687,12 +7002,16 @@ static int alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
 	size = sizeof(*extent_item) + btrfs_extent_inline_ref_size(type);
 
 	path = btrfs_alloc_path();
-	BUG_ON(!path);
+	if (!path)
+		return -ENOMEM;
 
 	path->leave_spinning = 1;
 	ret = btrfs_insert_empty_item(trans, fs_info->extent_root, path,
 				      ins, size);
-	BUG_ON(ret);
+	if (ret) {
+		btrfs_free_path(path);
+		return ret;
+	}
 
 	leaf = path->nodes[0];
 	extent_item = btrfs_item_ptr(leaf, path->slots[0],
@@ -4721,14 +7040,20 @@ static int alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
 	btrfs_mark_buffer_dirty(path->nodes[0]);
 	btrfs_free_path(path);
 
-	ret = update_block_group(trans, root, ins->objectid, ins->offset,
-				 1, 0);
-	if (ret) {
-		printk(KERN_ERR "btrfs update block group failed for %llu "
-		       "%llu\n", (unsigned long long)ins->objectid,
-		       (unsigned long long)ins->offset);
+	/* Always set parent to 0 here since its exclusive anyway. */
+	ret = btrfs_qgroup_record_ref(trans, fs_info, root_objectid,
+				      ins->objectid, ins->offset,
+				      BTRFS_QGROUP_OPER_ADD_EXCL, 0);
+	if (ret)
+		return ret;
+
+	ret = update_block_group(root, ins->objectid, ins->offset, 1);
+	if (ret) { /* -ENOENT, logic error */
+		btrfs_err(fs_info, "update block group failed for %llu %llu",
+			ins->objectid, ins->offset);
 		BUG();
 	}
+	trace_btrfs_reserved_extent_alloc(root, ins->objectid, ins->offset);
 	return ret;
 }
 
@@ -4736,7 +7061,8 @@ static int alloc_reserved_tree_block(struct btrfs_trans_handle *trans,
 				     struct btrfs_root *root,
 				     u64 parent, u64 root_objectid,
 				     u64 flags, struct btrfs_disk_key *key,
-				     int level, struct btrfs_key *ins)
+				     int level, struct btrfs_key *ins,
+				     int no_quota)
 {
 	int ret;
 	struct btrfs_fs_info *fs_info = root->fs_info;
@@ -4745,15 +7071,30 @@ static int alloc_reserved_tree_block(struct btrfs_trans_handle *trans,
 	struct btrfs_extent_inline_ref *iref;
 	struct btrfs_path *path;
 	struct extent_buffer *leaf;
-	u32 size = sizeof(*extent_item) + sizeof(*block_info) + sizeof(*iref);
+	u32 size = sizeof(*extent_item) + sizeof(*iref);
+	u64 num_bytes = ins->offset;
+	bool skinny_metadata = btrfs_fs_incompat(root->fs_info,
+						 SKINNY_METADATA);
+
+	if (!skinny_metadata)
+		size += sizeof(*block_info);
 
 	path = btrfs_alloc_path();
-	BUG_ON(!path);
+	if (!path) {
+		btrfs_free_and_pin_reserved_extent(root, ins->objectid,
+						   root->leafsize);
+		return -ENOMEM;
+	}
 
 	path->leave_spinning = 1;
 	ret = btrfs_insert_empty_item(trans, fs_info->extent_root, path,
 				      ins, size);
-	BUG_ON(ret);
+	if (ret) {
+		btrfs_free_and_pin_reserved_extent(root, ins->objectid,
+						   root->leafsize);
+		btrfs_free_path(path);
+		return ret;
+	}
 
 	leaf = path->nodes[0];
 	extent_item = btrfs_item_ptr(leaf, path->slots[0],
@@ -4762,12 +7103,17 @@ static int alloc_reserved_tree_block(struct btrfs_trans_handle *trans,
 	btrfs_set_extent_generation(leaf, extent_item, trans->transid);
 	btrfs_set_extent_flags(leaf, extent_item,
 			       flags | BTRFS_EXTENT_FLAG_TREE_BLOCK);
-	block_info = (struct btrfs_tree_block_info *)(extent_item + 1);
 
-	btrfs_set_tree_block_key(leaf, block_info, key);
-	btrfs_set_tree_block_level(leaf, block_info, level);
+	if (skinny_metadata) {
+		iref = (struct btrfs_extent_inline_ref *)(extent_item + 1);
+		num_bytes = root->leafsize;
+	} else {
+		block_info = (struct btrfs_tree_block_info *)(extent_item + 1);
+		btrfs_set_tree_block_key(leaf, block_info, key);
+		btrfs_set_tree_block_level(leaf, block_info, level);
+		iref = (struct btrfs_extent_inline_ref *)(block_info + 1);
+	}
 
-	iref = (struct btrfs_extent_inline_ref *)(block_info + 1);
 	if (parent > 0) {
 		BUG_ON(!(flags & BTRFS_BLOCK_FLAG_FULL_BACKREF));
 		btrfs_set_extent_inline_ref_type(leaf, iref,
@@ -4782,14 +7128,22 @@ static int alloc_reserved_tree_block(struct btrfs_trans_handle *trans,
 	btrfs_mark_buffer_dirty(leaf);
 	btrfs_free_path(path);
 
-	ret = update_block_group(trans, root, ins->objectid, ins->offset,
-				 1, 0);
-	if (ret) {
-		printk(KERN_ERR "btrfs update block group failed for %llu "
-		       "%llu\n", (unsigned long long)ins->objectid,
-		       (unsigned long long)ins->offset);
+	if (!no_quota) {
+		ret = btrfs_qgroup_record_ref(trans, fs_info, root_objectid,
+					      ins->objectid, num_bytes,
+					      BTRFS_QGROUP_OPER_ADD_EXCL, 0);
+		if (ret)
+			return ret;
+	}
+
+	ret = update_block_group(root, ins->objectid, root->leafsize, 1);
+	if (ret) { /* -ENOENT, logic error */
+		btrfs_err(fs_info, "update block group failed for %llu %llu",
+			ins->objectid, ins->offset);
 		BUG();
 	}
+
+	trace_btrfs_reserved_extent_alloc(root, ins->objectid, root->leafsize);
 	return ret;
 }
 
@@ -4802,9 +7156,10 @@ int btrfs_alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
 
 	BUG_ON(root_objectid == BTRFS_TREE_LOG_OBJECTID);
 
-	ret = btrfs_add_delayed_data_ref(trans, ins->objectid, ins->offset,
-					 0, root_objectid, owner, offset,
-					 BTRFS_ADD_DELAYED_EXTENT, NULL);
+	ret = btrfs_add_delayed_data_ref(root->fs_info, trans, ins->objectid,
+					 ins->offset, 0,
+					 root_objectid, owner, offset,
+					 BTRFS_ADD_DELAYED_EXTENT, NULL, 0);
 	return ret;
 }
 
@@ -4820,116 +7175,33 @@ int btrfs_alloc_logged_file_extent(struct btrfs_trans_handle *trans,
 {
 	int ret;
 	struct btrfs_block_group_cache *block_group;
-	struct btrfs_caching_control *caching_ctl;
-	u64 start = ins->objectid;
-	u64 num_bytes = ins->offset;
-
-	block_group = btrfs_lookup_block_group(root->fs_info, ins->objectid);
-	cache_block_group(block_group);
-	caching_ctl = get_caching_control(block_group);
-
-	if (!caching_ctl) {
-		BUG_ON(!block_group_cache_done(block_group));
-		ret = btrfs_remove_free_space(block_group, start, num_bytes);
-		BUG_ON(ret);
-	} else {
-		mutex_lock(&caching_ctl->mutex);
 
-		if (start >= caching_ctl->progress) {
-			ret = add_excluded_extent(root, start, num_bytes);
-			BUG_ON(ret);
-		} else if (start + num_bytes <= caching_ctl->progress) {
-			ret = btrfs_remove_free_space(block_group,
-						      start, num_bytes);
-			BUG_ON(ret);
-		} else {
-			num_bytes = caching_ctl->progress - start;
-			ret = btrfs_remove_free_space(block_group,
-						      start, num_bytes);
-			BUG_ON(ret);
-
-			start = caching_ctl->progress;
-			num_bytes = ins->objectid + ins->offset -
-				    caching_ctl->progress;
-			ret = add_excluded_extent(root, start, num_bytes);
-			BUG_ON(ret);
-		}
-
-		mutex_unlock(&caching_ctl->mutex);
-		put_caching_control(caching_ctl);
+	/*
+	 * Mixed block groups will exclude before processing the log so we only
+	 * need to do the exlude dance if this fs isn't mixed.
+	 */
+	if (!btrfs_fs_incompat(root->fs_info, MIXED_GROUPS)) {
+		ret = __exclude_logged_extent(root, ins->objectid, ins->offset);
+		if (ret)
+			return ret;
 	}
 
-	update_reserved_extents(block_group, ins->offset, 1);
-	btrfs_put_block_group(block_group);
+	block_group = btrfs_lookup_block_group(root->fs_info, ins->objectid);
+	if (!block_group)
+		return -EINVAL;
+
+	ret = btrfs_update_reserved_bytes(block_group, ins->offset,
+					  RESERVE_ALLOC_NO_ACCOUNT, 0);
+	BUG_ON(ret); /* logic error */
 	ret = alloc_reserved_file_extent(trans, root, 0, root_objectid,
 					 0, owner, offset, ins, 1);
+	btrfs_put_block_group(block_group);
 	return ret;
 }
 
-/*
- * finds a free extent and does all the dirty work required for allocation
- * returns the key for the extent through ins, and a tree buffer for
- * the first block of the extent through buf.
- *
- * returns 0 if everything worked, non-zero otherwise.
- */
-static int alloc_tree_block(struct btrfs_trans_handle *trans,
-			    struct btrfs_root *root,
-			    u64 num_bytes, u64 parent, u64 root_objectid,
-			    struct btrfs_disk_key *key, int level,
-			    u64 empty_size, u64 hint_byte, u64 search_end,
-			    struct btrfs_key *ins)
-{
-	int ret;
-	u64 flags = 0;
-
-	ret = btrfs_reserve_extent(trans, root, num_bytes, num_bytes,
-				   empty_size, hint_byte, search_end,
-				   ins, 0);
-	if (ret)
-		return ret;
-
-	if (root_objectid == BTRFS_TREE_RELOC_OBJECTID) {
-		if (parent == 0)
-			parent = ins->objectid;
-		flags |= BTRFS_BLOCK_FLAG_FULL_BACKREF;
-	} else
-		BUG_ON(parent > 0);
-
-	if (root_objectid != BTRFS_TREE_LOG_OBJECTID) {
-		struct btrfs_delayed_extent_op *extent_op;
-		extent_op = kmalloc(sizeof(*extent_op), GFP_NOFS);
-		BUG_ON(!extent_op);
-		if (key)
-			memcpy(&extent_op->key, key, sizeof(extent_op->key));
-		else
-			memset(&extent_op->key, 0, sizeof(extent_op->key));
-		extent_op->flags_to_set = flags;
-		extent_op->update_key = 1;
-		extent_op->update_flags = 1;
-		extent_op->is_data = 0;
-
-		ret = btrfs_add_delayed_tree_ref(trans, ins->objectid,
-					ins->offset, parent, root_objectid,
-					level, BTRFS_ADD_DELAYED_EXTENT,
-					extent_op);
-		BUG_ON(ret);
-	}
-
-	if (root_objectid == root->root_key.objectid) {
-		u64 used;
-		spin_lock(&root->node_lock);
-		used = btrfs_root_used(&root->root_item) + num_bytes;
-		btrfs_set_root_used(&root->root_item, used);
-		spin_unlock(&root->node_lock);
-	}
-	return ret;
-}
-
-struct extent_buffer *btrfs_init_new_buffer(struct btrfs_trans_handle *trans,
-					    struct btrfs_root *root,
-					    u64 bytenr, u32 blocksize,
-					    int level)
+static struct extent_buffer *
+btrfs_init_new_buffer(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+		      u64 bytenr, u32 blocksize, int level)
 {
 	struct extent_buffer *buf;
 
@@ -4937,9 +7209,10 @@ struct extent_buffer *btrfs_init_new_buffer(struct btrfs_trans_handle *trans,
 	if (!buf)
 		return ERR_PTR(-ENOMEM);
 	btrfs_set_header_generation(buf, trans->transid);
-	btrfs_set_buffer_lockdep_class(buf, level);
+	btrfs_set_buffer_lockdep_class(root->root_key.objectid, buf, level);
 	btrfs_tree_lock(buf);
 	clean_tree_block(trans, root, buf);
+	clear_bit(EXTENT_BUFFER_STALE, &buf->bflags);
 
 	btrfs_set_lock_blocking(buf);
 	btrfs_set_buffer_uptodate(buf);
@@ -4964,8 +7237,72 @@ struct extent_buffer *btrfs_init_new_buffer(struct btrfs_trans_handle *trans,
 	return buf;
 }
 
+static struct btrfs_block_rsv *
+use_block_rsv(struct btrfs_trans_handle *trans,
+	      struct btrfs_root *root, u32 blocksize)
+{
+	struct btrfs_block_rsv *block_rsv;
+	struct btrfs_block_rsv *global_rsv = &root->fs_info->global_block_rsv;
+	int ret;
+	bool global_updated = false;
+
+	block_rsv = get_block_rsv(trans, root);
+
+	if (unlikely(block_rsv->size == 0))
+		goto try_reserve;
+again:
+	ret = block_rsv_use_bytes(block_rsv, blocksize);
+	if (!ret)
+		return block_rsv;
+
+	if (block_rsv->failfast)
+		return ERR_PTR(ret);
+
+	if (block_rsv->type == BTRFS_BLOCK_RSV_GLOBAL && !global_updated) {
+		global_updated = true;
+		update_global_block_rsv(root->fs_info);
+		goto again;
+	}
+
+	if (btrfs_test_opt(root, ENOSPC_DEBUG)) {
+		static DEFINE_RATELIMIT_STATE(_rs,
+				DEFAULT_RATELIMIT_INTERVAL * 10,
+				/*DEFAULT_RATELIMIT_BURST*/ 1);
+		if (__ratelimit(&_rs))
+			WARN(1, KERN_DEBUG
+				"BTRFS: block rsv returned %d\n", ret);
+	}
+try_reserve:
+	ret = reserve_metadata_bytes(root, block_rsv, blocksize,
+				     BTRFS_RESERVE_NO_FLUSH);
+	if (!ret)
+		return block_rsv;
+	/*
+	 * If we couldn't reserve metadata bytes try and use some from
+	 * the global reserve if its space type is the same as the global
+	 * reservation.
+	 */
+	if (block_rsv->type != BTRFS_BLOCK_RSV_GLOBAL &&
+	    block_rsv->space_info == global_rsv->space_info) {
+		ret = block_rsv_use_bytes(global_rsv, blocksize);
+		if (!ret)
+			return global_rsv;
+	}
+	return ERR_PTR(ret);
+}
+
+static void unuse_block_rsv(struct btrfs_fs_info *fs_info,
+			    struct btrfs_block_rsv *block_rsv, u32 blocksize)
+{
+	block_rsv_add_bytes(block_rsv, blocksize, 0);
+	block_rsv_release_bytes(fs_info, block_rsv, NULL, 0);
+}
+
 /*
- * helper function to allocate a block for a given tree
+ * finds a free extent and does all the dirty work required for allocation
+ * returns the key for the extent through ins, and a tree buffer for
+ * the first block of the extent through buf.
+ *
  * returns the tree buffer or NULL.
  */
 struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
@@ -4975,18 +7312,68 @@ struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
 					u64 hint, u64 empty_size)
 {
 	struct btrfs_key ins;
-	int ret;
+	struct btrfs_block_rsv *block_rsv;
 	struct extent_buffer *buf;
+	u64 flags = 0;
+	int ret;
+	bool skinny_metadata = btrfs_fs_incompat(root->fs_info,
+						 SKINNY_METADATA);
+
+#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
+	if (unlikely(test_bit(BTRFS_ROOT_DUMMY_ROOT, &root->state))) {
+		buf = btrfs_init_new_buffer(trans, root, root->alloc_bytenr,
+					    blocksize, level);
+		if (!IS_ERR(buf))
+			root->alloc_bytenr += blocksize;
+		return buf;
+	}
+#endif
+	block_rsv = use_block_rsv(trans, root, blocksize);
+	if (IS_ERR(block_rsv))
+		return ERR_CAST(block_rsv);
 
-	ret = alloc_tree_block(trans, root, blocksize, parent, root_objectid,
-			       key, level, empty_size, hint, (u64)-1, &ins);
+	ret = btrfs_reserve_extent(root, blocksize, blocksize,
+				   empty_size, hint, &ins, 0, 0);
 	if (ret) {
-		BUG_ON(ret > 0);
+		unuse_block_rsv(root->fs_info, block_rsv, blocksize);
 		return ERR_PTR(ret);
 	}
 
 	buf = btrfs_init_new_buffer(trans, root, ins.objectid,
 				    blocksize, level);
+	BUG_ON(IS_ERR(buf)); /* -ENOMEM */
+
+	if (root_objectid == BTRFS_TREE_RELOC_OBJECTID) {
+		if (parent == 0)
+			parent = ins.objectid;
+		flags |= BTRFS_BLOCK_FLAG_FULL_BACKREF;
+	} else
+		BUG_ON(parent > 0);
+
+	if (root_objectid != BTRFS_TREE_LOG_OBJECTID) {
+		struct btrfs_delayed_extent_op *extent_op;
+		extent_op = btrfs_alloc_delayed_extent_op();
+		BUG_ON(!extent_op); /* -ENOMEM */
+		if (key)
+			memcpy(&extent_op->key, key, sizeof(extent_op->key));
+		else
+			memset(&extent_op->key, 0, sizeof(extent_op->key));
+		extent_op->flags_to_set = flags;
+		if (skinny_metadata)
+			extent_op->update_key = 0;
+		else
+			extent_op->update_key = 1;
+		extent_op->update_flags = 1;
+		extent_op->is_data = 0;
+		extent_op->level = level;
+
+		ret = btrfs_add_delayed_tree_ref(root->fs_info, trans,
+					ins.objectid,
+					ins.offset, parent, root_objectid,
+					level, BTRFS_ADD_DELAYED_EXTENT,
+					extent_op, 0);
+		BUG_ON(ret); /* -ENOMEM */
+	}
 	return buf;
 }
 
@@ -5001,6 +7388,7 @@ struct walk_control {
 	int keep_locks;
 	int reada_slot;
 	int reada_count;
+	int for_reloc;
 };
 
 #define DROP_REFERENCE	1
@@ -5015,7 +7403,6 @@ static noinline void reada_walk_down(struct btrfs_trans_handle *trans,
 	u64 generation;
 	u64 refs;
 	u64 flags;
-	u64 last = 0;
 	u32 nritems;
 	u32 blocksize;
 	struct btrfs_key key;
@@ -5053,9 +7440,12 @@ static noinline void reada_walk_down(struct btrfs_trans_handle *trans,
 			continue;
 
 		/* We don't lock the tree block, it's OK to be racy here */
-		ret = btrfs_lookup_extent_info(trans, root, bytenr, blocksize,
-					       &refs, &flags);
-		BUG_ON(ret);
+		ret = btrfs_lookup_extent_info(trans, root, bytenr,
+					       wc->level - 1, 1, &refs,
+					       &flags);
+		/* We don't care about errors in readahead. */
+		if (ret < 0)
+			continue;
 		BUG_ON(refs == 0);
 
 		if (wc->stage == DROP_REFERENCE) {
@@ -5083,14 +7473,13 @@ reada:
 					   generation);
 		if (ret)
 			break;
-		last = bytenr + blocksize;
 		nread++;
 	}
 	wc->reada_slot = slot;
 }
 
 /*
- * hepler to process tree block while walking down the tree.
+ * helper to process tree block while walking down the tree.
  *
  * when wc->stage == UPDATE_BACKREF, this function updates
  * back refs for pointers in the block.
@@ -5120,10 +7509,12 @@ static noinline int walk_down_proc(struct btrfs_trans_handle *trans,
 	     (wc->stage == UPDATE_BACKREF && !(wc->flags[level] & flag)))) {
 		BUG_ON(!path->locks[level]);
 		ret = btrfs_lookup_extent_info(trans, root,
-					       eb->start, eb->len,
+					       eb->start, level, 1,
 					       &wc->refs[level],
 					       &wc->flags[level]);
-		BUG_ON(ret);
+		BUG_ON(ret == -ENOMEM);
+		if (ret)
+			return ret;
 		BUG_ON(wc->refs[level] == 0);
 	}
 
@@ -5132,7 +7523,7 @@ static noinline int walk_down_proc(struct btrfs_trans_handle *trans,
 			return 1;
 
 		if (path->locks[level] && !wc->keep_locks) {
-			btrfs_tree_unlock(eb);
+			btrfs_tree_unlock_rw(eb, path->locks[level]);
 			path->locks[level] = 0;
 		}
 		return 0;
@@ -5141,13 +7532,14 @@ static noinline int walk_down_proc(struct btrfs_trans_handle *trans,
 	/* wc->stage == UPDATE_BACKREF */
 	if (!(wc->flags[level] & flag)) {
 		BUG_ON(!path->locks[level]);
-		ret = btrfs_inc_ref(trans, root, eb, 1);
-		BUG_ON(ret);
-		ret = btrfs_dec_ref(trans, root, eb, 0);
-		BUG_ON(ret);
+		ret = btrfs_inc_ref(trans, root, eb, 1, wc->for_reloc);
+		BUG_ON(ret); /* -ENOMEM */
+		ret = btrfs_dec_ref(trans, root, eb, 0, wc->for_reloc);
+		BUG_ON(ret); /* -ENOMEM */
 		ret = btrfs_set_disk_extent_flags(trans, root, eb->start,
-						  eb->len, flag, 0);
-		BUG_ON(ret);
+						  eb->len, flag,
+						  btrfs_header_level(eb), 0);
+		BUG_ON(ret); /* -ENOMEM */
 		wc->flags[level] |= flag;
 	}
 
@@ -5156,14 +7548,14 @@ static noinline int walk_down_proc(struct btrfs_trans_handle *trans,
 	 * keep the tree lock
 	 */
 	if (path->locks[level] && level > 0) {
-		btrfs_tree_unlock(eb);
+		btrfs_tree_unlock_rw(eb, path->locks[level]);
 		path->locks[level] = 0;
 	}
 	return 0;
 }
 
 /*
- * hepler to process tree block pointer.
+ * helper to process tree block pointer.
  *
  * when wc->stage == DROP_REFERENCE, this function checks
  * reference count of the block pointed to. if the block
@@ -5211,16 +7603,25 @@ static noinline int do_walk_down(struct btrfs_trans_handle *trans,
 		next = btrfs_find_create_tree_block(root, bytenr, blocksize);
 		if (!next)
 			return -ENOMEM;
+		btrfs_set_buffer_lockdep_class(root->root_key.objectid, next,
+					       level - 1);
 		reada = 1;
 	}
 	btrfs_tree_lock(next);
 	btrfs_set_lock_blocking(next);
 
-	ret = btrfs_lookup_extent_info(trans, root, bytenr, blocksize,
+	ret = btrfs_lookup_extent_info(trans, root, bytenr, level - 1, 1,
 				       &wc->refs[level - 1],
 				       &wc->flags[level - 1]);
-	BUG_ON(ret);
-	BUG_ON(wc->refs[level - 1] == 0);
+	if (ret < 0) {
+		btrfs_tree_unlock(next);
+		return ret;
+	}
+
+	if (unlikely(wc->refs[level - 1] == 0)) {
+		btrfs_err(root->fs_info, "Missing references.");
+		BUG();
+	}
 	*lookup_info = 0;
 
 	if (wc->stage == DROP_REFERENCE) {
@@ -5248,7 +7649,7 @@ static noinline int do_walk_down(struct btrfs_trans_handle *trans,
 			goto skip;
 	}
 
-	if (!btrfs_buffer_uptodate(next, generation)) {
+	if (!btrfs_buffer_uptodate(next, generation, 0)) {
 		btrfs_tree_unlock(next);
 		free_extent_buffer(next);
 		next = NULL;
@@ -5259,6 +7660,10 @@ static noinline int do_walk_down(struct btrfs_trans_handle *trans,
 		if (reada && level == 1)
 			reada_walk_down(trans, root, wc, path);
 		next = read_tree_block(root, bytenr, blocksize, generation);
+		if (!next || !extent_buffer_uptodate(next)) {
+			free_extent_buffer(next);
+			return -EIO;
+		}
 		btrfs_tree_lock(next);
 		btrfs_set_lock_blocking(next);
 	}
@@ -5267,7 +7672,7 @@ static noinline int do_walk_down(struct btrfs_trans_handle *trans,
 	BUG_ON(level != btrfs_header_level(next));
 	path->nodes[level] = next;
 	path->slots[level] = 0;
-	path->locks[level] = 1;
+	path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
 	wc->level = level;
 	if (wc->level == 1)
 		wc->reada_slot = 0;
@@ -5285,8 +7690,8 @@ skip:
 		}
 
 		ret = btrfs_free_extent(trans, root, bytenr, blocksize, parent,
-					root->root_key.objectid, level - 1, 0);
-		BUG_ON(ret);
+				root->root_key.objectid, level - 1, 0, 0);
+		BUG_ON(ret); /* -ENOMEM */
 	}
 	btrfs_tree_unlock(next);
 	free_extent_buffer(next);
@@ -5295,7 +7700,7 @@ skip:
 }
 
 /*
- * hepler to process tree block while walking up the tree.
+ * helper to process tree block while walking up the tree.
  *
  * when wc->stage == DROP_REFERENCE, this function drops
  * reference count on the block.
@@ -5311,7 +7716,7 @@ static noinline int walk_up_proc(struct btrfs_trans_handle *trans,
 				 struct btrfs_path *path,
 				 struct walk_control *wc)
 {
-	int ret = 0;
+	int ret;
 	int level = wc->level;
 	struct extent_buffer *eb = path->nodes[level];
 	u64 parent = 0;
@@ -5338,16 +7743,20 @@ static noinline int walk_up_proc(struct btrfs_trans_handle *trans,
 			BUG_ON(level == 0);
 			btrfs_tree_lock(eb);
 			btrfs_set_lock_blocking(eb);
-			path->locks[level] = 1;
+			path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
 
 			ret = btrfs_lookup_extent_info(trans, root,
-						       eb->start, eb->len,
+						       eb->start, level, 1,
 						       &wc->refs[level],
 						       &wc->flags[level]);
-			BUG_ON(ret);
+			if (ret < 0) {
+				btrfs_tree_unlock_rw(eb, path->locks[level]);
+				path->locks[level] = 0;
+				return ret;
+			}
 			BUG_ON(wc->refs[level] == 0);
 			if (wc->refs[level] == 1) {
-				btrfs_tree_unlock(eb);
+				btrfs_tree_unlock_rw(eb, path->locks[level]);
 				path->locks[level] = 0;
 				return 1;
 			}
@@ -5360,17 +7769,19 @@ static noinline int walk_up_proc(struct btrfs_trans_handle *trans,
 	if (wc->refs[level] == 1) {
 		if (level == 0) {
 			if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
-				ret = btrfs_dec_ref(trans, root, eb, 1);
+				ret = btrfs_dec_ref(trans, root, eb, 1,
+						    wc->for_reloc);
 			else
-				ret = btrfs_dec_ref(trans, root, eb, 0);
-			BUG_ON(ret);
+				ret = btrfs_dec_ref(trans, root, eb, 0,
+						    wc->for_reloc);
+			BUG_ON(ret); /* -ENOMEM */
 		}
 		/* make block locked assertion in clean_tree_block happy */
 		if (!path->locks[level] &&
 		    btrfs_header_generation(eb) == trans->transid) {
 			btrfs_tree_lock(eb);
 			btrfs_set_lock_blocking(eb);
-			path->locks[level] = 1;
+			path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
 		}
 		clean_tree_block(trans, root, eb);
 	}
@@ -5389,13 +7800,11 @@ static noinline int walk_up_proc(struct btrfs_trans_handle *trans,
 			       btrfs_header_owner(path->nodes[level + 1]));
 	}
 
-	ret = btrfs_free_extent(trans, root, eb->start, eb->len, parent,
-				root->root_key.objectid, level, 0);
-	BUG_ON(ret);
+	btrfs_free_tree_block(trans, root, eb, parent, wc->refs[level] == 1);
 out:
 	wc->refs[level] = 0;
 	wc->flags[level] = 0;
-	return ret;
+	return 0;
 }
 
 static noinline int walk_down_tree(struct btrfs_trans_handle *trans,
@@ -5451,7 +7860,8 @@ static noinline int walk_up_tree(struct btrfs_trans_handle *trans,
 				return 0;
 
 			if (path->locks[level]) {
-				btrfs_tree_unlock(path->nodes[level]);
+				btrfs_tree_unlock_rw(path->nodes[level],
+						     path->locks[level]);
 				path->locks[level] = 0;
 			}
 			free_extent_buffer(path->nodes[level]);
@@ -5472,8 +7882,12 @@ static noinline int walk_up_tree(struct btrfs_trans_handle *trans,
  * reference count by one. if update_ref is true, this function
  * also make sure backrefs for the shared block and all lower level
  * blocks are properly updated.
+ *
+ * If called with for_reloc == 0, may exit early with -EAGAIN
  */
-int btrfs_drop_snapshot(struct btrfs_root *root, int update_ref)
+int btrfs_drop_snapshot(struct btrfs_root *root,
+			 struct btrfs_block_rsv *block_rsv, int update_ref,
+			 int for_reloc)
 {
 	struct btrfs_path *path;
 	struct btrfs_trans_handle *trans;
@@ -5484,21 +7898,36 @@ int btrfs_drop_snapshot(struct btrfs_root *root, int update_ref)
 	int err = 0;
 	int ret;
 	int level;
+	bool root_dropped = false;
 
 	path = btrfs_alloc_path();
-	BUG_ON(!path);
+	if (!path) {
+		err = -ENOMEM;
+		goto out;
+	}
 
 	wc = kzalloc(sizeof(*wc), GFP_NOFS);
-	BUG_ON(!wc);
+	if (!wc) {
+		btrfs_free_path(path);
+		err = -ENOMEM;
+		goto out;
+	}
+
+	trans = btrfs_start_transaction(tree_root, 0);
+	if (IS_ERR(trans)) {
+		err = PTR_ERR(trans);
+		goto out_free;
+	}
 
-	trans = btrfs_start_transaction(tree_root, 1);
+	if (block_rsv)
+		trans->block_rsv = block_rsv;
 
 	if (btrfs_disk_key_objectid(&root_item->drop_progress) == 0) {
 		level = btrfs_header_level(root->node);
 		path->nodes[level] = btrfs_lock_root_node(root);
 		btrfs_set_lock_blocking(path->nodes[level]);
 		path->slots[level] = 0;
-		path->locks[level] = 1;
+		path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
 		memset(&wc->update_progress, 0,
 		       sizeof(wc->update_progress));
 	} else {
@@ -5513,7 +7942,7 @@ int btrfs_drop_snapshot(struct btrfs_root *root, int update_ref)
 		path->lowest_level = 0;
 		if (ret < 0) {
 			err = ret;
-			goto out;
+			goto out_end_trans;
 		}
 		WARN_ON(ret > 0);
 
@@ -5527,19 +7956,23 @@ int btrfs_drop_snapshot(struct btrfs_root *root, int update_ref)
 		while (1) {
 			btrfs_tree_lock(path->nodes[level]);
 			btrfs_set_lock_blocking(path->nodes[level]);
+			path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
 
 			ret = btrfs_lookup_extent_info(trans, root,
 						path->nodes[level]->start,
-						path->nodes[level]->len,
-						&wc->refs[level],
+						level, 1, &wc->refs[level],
 						&wc->flags[level]);
-			BUG_ON(ret);
+			if (ret < 0) {
+				err = ret;
+				goto out_end_trans;
+			}
 			BUG_ON(wc->refs[level] == 0);
 
 			if (level == root_item->drop_level)
 				break;
 
 			btrfs_tree_unlock(path->nodes[level]);
+			path->locks[level] = 0;
 			WARN_ON(wc->refs[level] != 1);
 			level--;
 		}
@@ -5550,9 +7983,11 @@ int btrfs_drop_snapshot(struct btrfs_root *root, int update_ref)
 	wc->stage = DROP_REFERENCE;
 	wc->update_ref = update_ref;
 	wc->keep_locks = 0;
+	wc->for_reloc = for_reloc;
 	wc->reada_count = BTRFS_NODEPTRS_PER_BLOCK(root);
 
 	while (1) {
+
 		ret = walk_down_tree(trans, root, path, wc);
 		if (ret < 0) {
 			err = ret;
@@ -5579,52 +8014,86 @@ int btrfs_drop_snapshot(struct btrfs_root *root, int update_ref)
 		}
 
 		BUG_ON(wc->level == 0);
-		if (trans->transaction->in_commit ||
-		    trans->transaction->delayed_refs.flushing) {
+		if (btrfs_should_end_transaction(trans, tree_root) ||
+		    (!for_reloc && btrfs_need_cleaner_sleep(root))) {
 			ret = btrfs_update_root(trans, tree_root,
 						&root->root_key,
 						root_item);
-			BUG_ON(ret);
+			if (ret) {
+				btrfs_abort_transaction(trans, tree_root, ret);
+				err = ret;
+				goto out_end_trans;
+			}
 
-			btrfs_end_transaction(trans, tree_root);
-			trans = btrfs_start_transaction(tree_root, 1);
-		} else {
-			unsigned long update;
-			update = trans->delayed_ref_updates;
-			trans->delayed_ref_updates = 0;
-			if (update)
-				btrfs_run_delayed_refs(trans, tree_root,
-						       update);
+			btrfs_end_transaction_throttle(trans, tree_root);
+			if (!for_reloc && btrfs_need_cleaner_sleep(root)) {
+				pr_debug("BTRFS: drop snapshot early exit\n");
+				err = -EAGAIN;
+				goto out_free;
+			}
+
+			trans = btrfs_start_transaction(tree_root, 0);
+			if (IS_ERR(trans)) {
+				err = PTR_ERR(trans);
+				goto out_free;
+			}
+			if (block_rsv)
+				trans->block_rsv = block_rsv;
 		}
 	}
-	btrfs_release_path(root, path);
-	BUG_ON(err);
+	btrfs_release_path(path);
+	if (err)
+		goto out_end_trans;
 
 	ret = btrfs_del_root(trans, tree_root, &root->root_key);
-	BUG_ON(ret);
+	if (ret) {
+		btrfs_abort_transaction(trans, tree_root, ret);
+		goto out_end_trans;
+	}
 
 	if (root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID) {
-		ret = btrfs_find_last_root(tree_root, root->root_key.objectid,
-					   NULL, NULL);
-		BUG_ON(ret < 0);
-		if (ret > 0) {
-			ret = btrfs_del_orphan_item(trans, tree_root,
-						    root->root_key.objectid);
-			BUG_ON(ret);
+		ret = btrfs_find_root(tree_root, &root->root_key, path,
+				      NULL, NULL);
+		if (ret < 0) {
+			btrfs_abort_transaction(trans, tree_root, ret);
+			err = ret;
+			goto out_end_trans;
+		} else if (ret > 0) {
+			/* if we fail to delete the orphan item this time
+			 * around, it'll get picked up the next time.
+			 *
+			 * The most common failure here is just -ENOENT.
+			 */
+			btrfs_del_orphan_item(trans, tree_root,
+					      root->root_key.objectid);
 		}
 	}
 
-	if (root->in_radix) {
-		btrfs_free_fs_root(tree_root->fs_info, root);
+	if (test_bit(BTRFS_ROOT_IN_RADIX, &root->state)) {
+		btrfs_drop_and_free_fs_root(tree_root->fs_info, root);
 	} else {
 		free_extent_buffer(root->node);
 		free_extent_buffer(root->commit_root);
-		kfree(root);
+		btrfs_put_fs_root(root);
 	}
-out:
-	btrfs_end_transaction(trans, tree_root);
+	root_dropped = true;
+out_end_trans:
+	btrfs_end_transaction_throttle(trans, tree_root);
+out_free:
 	kfree(wc);
 	btrfs_free_path(path);
+out:
+	/*
+	 * So if we need to stop dropping the snapshot for whatever reason we
+	 * need to make sure to add it back to the dead root list so that we
+	 * keep trying to do the work later.  This also cleans up roots if we
+	 * don't have it in the radix (like when we recover after a power fail
+	 * or unmount) so we don't leak memory.
+	 */
+	if (!for_reloc && root_dropped == false)
+		btrfs_add_dead_root(root);
+	if (err && err != -EAGAIN)
+		btrfs_std_error(root->fs_info, err);
 	return err;
 }
 
@@ -5632,6 +8101,7 @@ out:
  * drop subtree rooted at tree block 'node'.
  *
  * NOTE: this function will unlock and release tree block 'node'
+ * only used by relocation code
  */
 int btrfs_drop_subtree(struct btrfs_trans_handle *trans,
 			struct btrfs_root *root,
@@ -5648,10 +8118,14 @@ int btrfs_drop_subtree(struct btrfs_trans_handle *trans,
 	BUG_ON(root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID);
 
 	path = btrfs_alloc_path();
-	BUG_ON(!path);
+	if (!path)
+		return -ENOMEM;
 
 	wc = kzalloc(sizeof(*wc), GFP_NOFS);
-	BUG_ON(!wc);
+	if (!wc) {
+		btrfs_free_path(path);
+		return -ENOMEM;
+	}
 
 	btrfs_assert_tree_locked(parent);
 	parent_level = btrfs_header_level(parent);
@@ -5663,7 +8137,7 @@ int btrfs_drop_subtree(struct btrfs_trans_handle *trans,
 	level = btrfs_header_level(node);
 	path->nodes[level] = node;
 	path->slots[level] = 0;
-	path->locks[level] = 1;
+	path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
 
 	wc->refs[parent_level] = 1;
 	wc->flags[parent_level] = BTRFS_BLOCK_FLAG_FULL_BACKREF;
@@ -5672,6 +8146,7 @@ int btrfs_drop_subtree(struct btrfs_trans_handle *trans,
 	wc->stage = DROP_REFERENCE;
 	wc->update_ref = 0;
 	wc->keep_locks = 1;
+	wc->for_reloc = 1;
 	wc->reada_count = BTRFS_NODEPTRS_PER_BLOCK(root);
 
 	while (1) {
@@ -5693,1574 +8168,221 @@ int btrfs_drop_subtree(struct btrfs_trans_handle *trans,
 	return ret;
 }
 
-#if 0
-static unsigned long calc_ra(unsigned long start, unsigned long last,
-			     unsigned long nr)
-{
-	return min(last, start + nr - 1);
-}
-
-static noinline int relocate_inode_pages(struct inode *inode, u64 start,
-					 u64 len)
+static u64 update_block_group_flags(struct btrfs_root *root, u64 flags)
 {
-	u64 page_start;
-	u64 page_end;
-	unsigned long first_index;
-	unsigned long last_index;
-	unsigned long i;
-	struct page *page;
-	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
-	struct file_ra_state *ra;
-	struct btrfs_ordered_extent *ordered;
-	unsigned int total_read = 0;
-	unsigned int total_dirty = 0;
-	int ret = 0;
-
-	ra = kzalloc(sizeof(*ra), GFP_NOFS);
+	u64 num_devices;
+	u64 stripped;
 
-	mutex_lock(&inode->i_mutex);
-	first_index = start >> PAGE_CACHE_SHIFT;
-	last_index = (start + len - 1) >> PAGE_CACHE_SHIFT;
+	/*
+	 * if restripe for this chunk_type is on pick target profile and
+	 * return, otherwise do the usual balance
+	 */
+	stripped = get_restripe_target(root->fs_info, flags);
+	if (stripped)
+		return extended_to_chunk(stripped);
 
-	/* make sure the dirty trick played by the caller work */
-	ret = invalidate_inode_pages2_range(inode->i_mapping,
-					    first_index, last_index);
-	if (ret)
-		goto out_unlock;
+	/*
+	 * we add in the count of missing devices because we want
+	 * to make sure that any RAID levels on a degraded FS
+	 * continue to be honored.
+	 */
+	num_devices = root->fs_info->fs_devices->rw_devices +
+		root->fs_info->fs_devices->missing_devices;
 
-	file_ra_state_init(ra, inode->i_mapping);
+	stripped = BTRFS_BLOCK_GROUP_RAID0 |
+		BTRFS_BLOCK_GROUP_RAID5 | BTRFS_BLOCK_GROUP_RAID6 |
+		BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID10;
 
-	for (i = first_index ; i <= last_index; i++) {
-		if (total_read % ra->ra_pages == 0) {
-			btrfs_force_ra(inode->i_mapping, ra, NULL, i,
-				       calc_ra(i, last_index, ra->ra_pages));
-		}
-		total_read++;
-again:
-		if (((u64)i << PAGE_CACHE_SHIFT) > i_size_read(inode))
-			BUG_ON(1);
-		page = grab_cache_page(inode->i_mapping, i);
-		if (!page) {
-			ret = -ENOMEM;
-			goto out_unlock;
-		}
-		if (!PageUptodate(page)) {
-			btrfs_readpage(NULL, page);
-			lock_page(page);
-			if (!PageUptodate(page)) {
-				unlock_page(page);
-				page_cache_release(page);
-				ret = -EIO;
-				goto out_unlock;
-			}
-		}
-		wait_on_page_writeback(page);
+	if (num_devices == 1) {
+		stripped |= BTRFS_BLOCK_GROUP_DUP;
+		stripped = flags & ~stripped;
 
-		page_start = (u64)page->index << PAGE_CACHE_SHIFT;
-		page_end = page_start + PAGE_CACHE_SIZE - 1;
-		lock_extent(io_tree, page_start, page_end, GFP_NOFS);
+		/* turn raid0 into single device chunks */
+		if (flags & BTRFS_BLOCK_GROUP_RAID0)
+			return stripped;
 
-		ordered = btrfs_lookup_ordered_extent(inode, page_start);
-		if (ordered) {
-			unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
-			unlock_page(page);
-			page_cache_release(page);
-			btrfs_start_ordered_extent(inode, ordered, 1);
-			btrfs_put_ordered_extent(ordered);
-			goto again;
-		}
-		set_page_extent_mapped(page);
+		/* turn mirroring into duplication */
+		if (flags & (BTRFS_BLOCK_GROUP_RAID1 |
+			     BTRFS_BLOCK_GROUP_RAID10))
+			return stripped | BTRFS_BLOCK_GROUP_DUP;
+	} else {
+		/* they already had raid on here, just return */
+		if (flags & stripped)
+			return flags;
 
-		if (i == first_index)
-			set_extent_bits(io_tree, page_start, page_end,
-					EXTENT_BOUNDARY, GFP_NOFS);
-		btrfs_set_extent_delalloc(inode, page_start, page_end);
+		stripped |= BTRFS_BLOCK_GROUP_DUP;
+		stripped = flags & ~stripped;
 
-		set_page_dirty(page);
-		total_dirty++;
+		/* switch duplicated blocks with raid1 */
+		if (flags & BTRFS_BLOCK_GROUP_DUP)
+			return stripped | BTRFS_BLOCK_GROUP_RAID1;
 
-		unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
-		unlock_page(page);
-		page_cache_release(page);
+		/* this is drive concat, leave it alone */
 	}
 
-out_unlock:
-	kfree(ra);
-	mutex_unlock(&inode->i_mutex);
-	balance_dirty_pages_ratelimited_nr(inode->i_mapping, total_dirty);
-	return ret;
+	return flags;
 }
 
-static noinline int relocate_data_extent(struct inode *reloc_inode,
-					 struct btrfs_key *extent_key,
-					 u64 offset)
+static int set_block_group_ro(struct btrfs_block_group_cache *cache, int force)
 {
-	struct btrfs_root *root = BTRFS_I(reloc_inode)->root;
-	struct extent_map_tree *em_tree = &BTRFS_I(reloc_inode)->extent_tree;
-	struct extent_map *em;
-	u64 start = extent_key->objectid - offset;
-	u64 end = start + extent_key->offset - 1;
-
-	em = alloc_extent_map(GFP_NOFS);
-	BUG_ON(!em || IS_ERR(em));
-
-	em->start = start;
-	em->len = extent_key->offset;
-	em->block_len = extent_key->offset;
-	em->block_start = extent_key->objectid;
-	em->bdev = root->fs_info->fs_devices->latest_bdev;
-	set_bit(EXTENT_FLAG_PINNED, &em->flags);
-
-	/* setup extent map to cheat btrfs_readpage */
-	lock_extent(&BTRFS_I(reloc_inode)->io_tree, start, end, GFP_NOFS);
-	while (1) {
-		int ret;
-		write_lock(&em_tree->lock);
-		ret = add_extent_mapping(em_tree, em);
-		write_unlock(&em_tree->lock);
-		if (ret != -EEXIST) {
-			free_extent_map(em);
-			break;
-		}
-		btrfs_drop_extent_cache(reloc_inode, start, end, 0);
-	}
-	unlock_extent(&BTRFS_I(reloc_inode)->io_tree, start, end, GFP_NOFS);
-
-	return relocate_inode_pages(reloc_inode, start, extent_key->offset);
-}
-
-struct btrfs_ref_path {
-	u64 extent_start;
-	u64 nodes[BTRFS_MAX_LEVEL];
-	u64 root_objectid;
-	u64 root_generation;
-	u64 owner_objectid;
-	u32 num_refs;
-	int lowest_level;
-	int current_level;
-	int shared_level;
-
-	struct btrfs_key node_keys[BTRFS_MAX_LEVEL];
-	u64 new_nodes[BTRFS_MAX_LEVEL];
-};
-
-struct disk_extent {
-	u64 ram_bytes;
-	u64 disk_bytenr;
-	u64 disk_num_bytes;
-	u64 offset;
+	struct btrfs_space_info *sinfo = cache->space_info;
 	u64 num_bytes;
-	u8 compression;
-	u8 encryption;
-	u16 other_encoding;
-};
-
-static int is_cowonly_root(u64 root_objectid)
-{
-	if (root_objectid == BTRFS_ROOT_TREE_OBJECTID ||
-	    root_objectid == BTRFS_EXTENT_TREE_OBJECTID ||
-	    root_objectid == BTRFS_CHUNK_TREE_OBJECTID ||
-	    root_objectid == BTRFS_DEV_TREE_OBJECTID ||
-	    root_objectid == BTRFS_TREE_LOG_OBJECTID ||
-	    root_objectid == BTRFS_CSUM_TREE_OBJECTID)
-		return 1;
-	return 0;
-}
-
-static noinline int __next_ref_path(struct btrfs_trans_handle *trans,
-				    struct btrfs_root *extent_root,
-				    struct btrfs_ref_path *ref_path,
-				    int first_time)
-{
-	struct extent_buffer *leaf;
-	struct btrfs_path *path;
-	struct btrfs_extent_ref *ref;
-	struct btrfs_key key;
-	struct btrfs_key found_key;
-	u64 bytenr;
-	u32 nritems;
-	int level;
-	int ret = 1;
-
-	path = btrfs_alloc_path();
-	if (!path)
-		return -ENOMEM;
-
-	if (first_time) {
-		ref_path->lowest_level = -1;
-		ref_path->current_level = -1;
-		ref_path->shared_level = -1;
-		goto walk_up;
-	}
-walk_down:
-	level = ref_path->current_level - 1;
-	while (level >= -1) {
-		u64 parent;
-		if (level < ref_path->lowest_level)
-			break;
-
-		if (level >= 0)
-			bytenr = ref_path->nodes[level];
-		else
-			bytenr = ref_path->extent_start;
-		BUG_ON(bytenr == 0);
+	u64 min_allocable_bytes;
+	int ret = -ENOSPC;
 
-		parent = ref_path->nodes[level + 1];
-		ref_path->nodes[level + 1] = 0;
-		ref_path->current_level = level;
-		BUG_ON(parent == 0);
 
-		key.objectid = bytenr;
-		key.offset = parent + 1;
-		key.type = BTRFS_EXTENT_REF_KEY;
-
-		ret = btrfs_search_slot(trans, extent_root, &key, path, 0, 0);
-		if (ret < 0)
-			goto out;
-		BUG_ON(ret == 0);
+	/*
+	 * We need some metadata space and system metadata space for
+	 * allocating chunks in some corner cases until we force to set
+	 * it to be readonly.
+	 */
+	if ((sinfo->flags &
+	     (BTRFS_BLOCK_GROUP_SYSTEM | BTRFS_BLOCK_GROUP_METADATA)) &&
+	    !force)
+		min_allocable_bytes = 1 * 1024 * 1024;
+	else
+		min_allocable_bytes = 0;
 
-		leaf = path->nodes[0];
-		nritems = btrfs_header_nritems(leaf);
-		if (path->slots[0] >= nritems) {
-			ret = btrfs_next_leaf(extent_root, path);
-			if (ret < 0)
-				goto out;
-			if (ret > 0)
-				goto next;
-			leaf = path->nodes[0];
-		}
+	spin_lock(&sinfo->lock);
+	spin_lock(&cache->lock);
 
-		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
-		if (found_key.objectid == bytenr &&
-		    found_key.type == BTRFS_EXTENT_REF_KEY) {
-			if (level < ref_path->shared_level)
-				ref_path->shared_level = level;
-			goto found;
-		}
-next:
-		level--;
-		btrfs_release_path(extent_root, path);
-		cond_resched();
+	if (cache->ro) {
+		ret = 0;
+		goto out;
 	}
-	/* reached lowest level */
-	ret = 1;
-	goto out;
-walk_up:
-	level = ref_path->current_level;
-	while (level < BTRFS_MAX_LEVEL - 1) {
-		u64 ref_objectid;
-
-		if (level >= 0)
-			bytenr = ref_path->nodes[level];
-		else
-			bytenr = ref_path->extent_start;
-
-		BUG_ON(bytenr == 0);
-
-		key.objectid = bytenr;
-		key.offset = 0;
-		key.type = BTRFS_EXTENT_REF_KEY;
-
-		ret = btrfs_search_slot(trans, extent_root, &key, path, 0, 0);
-		if (ret < 0)
-			goto out;
-
-		leaf = path->nodes[0];
-		nritems = btrfs_header_nritems(leaf);
-		if (path->slots[0] >= nritems) {
-			ret = btrfs_next_leaf(extent_root, path);
-			if (ret < 0)
-				goto out;
-			if (ret > 0) {
-				/* the extent was freed by someone */
-				if (ref_path->lowest_level == level)
-					goto out;
-				btrfs_release_path(extent_root, path);
-				goto walk_down;
-			}
-			leaf = path->nodes[0];
-		}
-
-		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
-		if (found_key.objectid != bytenr ||
-				found_key.type != BTRFS_EXTENT_REF_KEY) {
-			/* the extent was freed by someone */
-			if (ref_path->lowest_level == level) {
-				ret = 1;
-				goto out;
-			}
-			btrfs_release_path(extent_root, path);
-			goto walk_down;
-		}
-found:
-		ref = btrfs_item_ptr(leaf, path->slots[0],
-				struct btrfs_extent_ref);
-		ref_objectid = btrfs_ref_objectid(leaf, ref);
-		if (ref_objectid < BTRFS_FIRST_FREE_OBJECTID) {
-			if (first_time) {
-				level = (int)ref_objectid;
-				BUG_ON(level >= BTRFS_MAX_LEVEL);
-				ref_path->lowest_level = level;
-				ref_path->current_level = level;
-				ref_path->nodes[level] = bytenr;
-			} else {
-				WARN_ON(ref_objectid != level);
-			}
-		} else {
-			WARN_ON(level != -1);
-		}
-		first_time = 0;
-
-		if (ref_path->lowest_level == level) {
-			ref_path->owner_objectid = ref_objectid;
-			ref_path->num_refs = btrfs_ref_num_refs(leaf, ref);
-		}
-
-		/*
-		 * the block is tree root or the block isn't in reference
-		 * counted tree.
-		 */
-		if (found_key.objectid == found_key.offset ||
-		    is_cowonly_root(btrfs_ref_root(leaf, ref))) {
-			ref_path->root_objectid = btrfs_ref_root(leaf, ref);
-			ref_path->root_generation =
-				btrfs_ref_generation(leaf, ref);
-			if (level < 0) {
-				/* special reference from the tree log */
-				ref_path->nodes[0] = found_key.offset;
-				ref_path->current_level = 0;
-			}
-			ret = 0;
-			goto out;
-		}
-
-		level++;
-		BUG_ON(ref_path->nodes[level] != 0);
-		ref_path->nodes[level] = found_key.offset;
-		ref_path->current_level = level;
 
-		/*
-		 * the reference was created in the running transaction,
-		 * no need to continue walking up.
-		 */
-		if (btrfs_ref_generation(leaf, ref) == trans->transid) {
-			ref_path->root_objectid = btrfs_ref_root(leaf, ref);
-			ref_path->root_generation =
-				btrfs_ref_generation(leaf, ref);
-			ret = 0;
-			goto out;
-		}
+	num_bytes = cache->key.offset - cache->reserved - cache->pinned -
+		    cache->bytes_super - btrfs_block_group_used(&cache->item);
 
-		btrfs_release_path(extent_root, path);
-		cond_resched();
+	if (sinfo->bytes_used + sinfo->bytes_reserved + sinfo->bytes_pinned +
+	    sinfo->bytes_may_use + sinfo->bytes_readonly + num_bytes +
+	    min_allocable_bytes <= sinfo->total_bytes) {
+		sinfo->bytes_readonly += num_bytes;
+		cache->ro = 1;
+		ret = 0;
 	}
-	/* reached max tree level, but no tree root found. */
-	BUG();
 out:
-	btrfs_free_path(path);
+	spin_unlock(&cache->lock);
+	spin_unlock(&sinfo->lock);
 	return ret;
 }
 
-static int btrfs_first_ref_path(struct btrfs_trans_handle *trans,
-				struct btrfs_root *extent_root,
-				struct btrfs_ref_path *ref_path,
-				u64 extent_start)
-{
-	memset(ref_path, 0, sizeof(*ref_path));
-	ref_path->extent_start = extent_start;
-
-	return __next_ref_path(trans, extent_root, ref_path, 1);
-}
-
-static int btrfs_next_ref_path(struct btrfs_trans_handle *trans,
-			       struct btrfs_root *extent_root,
-			       struct btrfs_ref_path *ref_path)
-{
-	return __next_ref_path(trans, extent_root, ref_path, 0);
-}
+int btrfs_set_block_group_ro(struct btrfs_root *root,
+			     struct btrfs_block_group_cache *cache)
 
-static noinline int get_new_locations(struct inode *reloc_inode,
-				      struct btrfs_key *extent_key,
-				      u64 offset, int no_fragment,
-				      struct disk_extent **extents,
-				      int *nr_extents)
 {
-	struct btrfs_root *root = BTRFS_I(reloc_inode)->root;
-	struct btrfs_path *path;
-	struct btrfs_file_extent_item *fi;
-	struct extent_buffer *leaf;
-	struct disk_extent *exts = *extents;
-	struct btrfs_key found_key;
-	u64 cur_pos;
-	u64 last_byte;
-	u32 nritems;
-	int nr = 0;
-	int max = *nr_extents;
+	struct btrfs_trans_handle *trans;
+	u64 alloc_flags;
 	int ret;
 
-	WARN_ON(!no_fragment && *extents);
-	if (!exts) {
-		max = 1;
-		exts = kmalloc(sizeof(*exts) * max, GFP_NOFS);
-		if (!exts)
-			return -ENOMEM;
-	}
-
-	path = btrfs_alloc_path();
-	BUG_ON(!path);
-
-	cur_pos = extent_key->objectid - offset;
-	last_byte = extent_key->objectid + extent_key->offset;
-	ret = btrfs_lookup_file_extent(NULL, root, path, reloc_inode->i_ino,
-				       cur_pos, 0);
-	if (ret < 0)
-		goto out;
-	if (ret > 0) {
-		ret = -ENOENT;
-		goto out;
-	}
-
-	while (1) {
-		leaf = path->nodes[0];
-		nritems = btrfs_header_nritems(leaf);
-		if (path->slots[0] >= nritems) {
-			ret = btrfs_next_leaf(root, path);
-			if (ret < 0)
-				goto out;
-			if (ret > 0)
-				break;
-			leaf = path->nodes[0];
-		}
-
-		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
-		if (found_key.offset != cur_pos ||
-		    found_key.type != BTRFS_EXTENT_DATA_KEY ||
-		    found_key.objectid != reloc_inode->i_ino)
-			break;
-
-		fi = btrfs_item_ptr(leaf, path->slots[0],
-				    struct btrfs_file_extent_item);
-		if (btrfs_file_extent_type(leaf, fi) !=
-		    BTRFS_FILE_EXTENT_REG ||
-		    btrfs_file_extent_disk_bytenr(leaf, fi) == 0)
-			break;
+	BUG_ON(cache->ro);
 
-		if (nr == max) {
-			struct disk_extent *old = exts;
-			max *= 2;
-			exts = kzalloc(sizeof(*exts) * max, GFP_NOFS);
-			memcpy(exts, old, sizeof(*exts) * nr);
-			if (old != *extents)
-				kfree(old);
-		}
-
-		exts[nr].disk_bytenr =
-			btrfs_file_extent_disk_bytenr(leaf, fi);
-		exts[nr].disk_num_bytes =
-			btrfs_file_extent_disk_num_bytes(leaf, fi);
-		exts[nr].offset = btrfs_file_extent_offset(leaf, fi);
-		exts[nr].num_bytes = btrfs_file_extent_num_bytes(leaf, fi);
-		exts[nr].ram_bytes = btrfs_file_extent_ram_bytes(leaf, fi);
-		exts[nr].compression = btrfs_file_extent_compression(leaf, fi);
-		exts[nr].encryption = btrfs_file_extent_encryption(leaf, fi);
-		exts[nr].other_encoding = btrfs_file_extent_other_encoding(leaf,
-									   fi);
-		BUG_ON(exts[nr].offset > 0);
-		BUG_ON(exts[nr].compression || exts[nr].encryption);
-		BUG_ON(exts[nr].num_bytes != exts[nr].disk_num_bytes);
-
-		cur_pos += exts[nr].num_bytes;
-		nr++;
-
-		if (cur_pos + offset >= last_byte)
-			break;
+	trans = btrfs_join_transaction(root);
+	if (IS_ERR(trans))
+		return PTR_ERR(trans);
 
-		if (no_fragment) {
-			ret = 1;
+	alloc_flags = update_block_group_flags(root, cache->flags);
+	if (alloc_flags != cache->flags) {
+		ret = do_chunk_alloc(trans, root, alloc_flags,
+				     CHUNK_ALLOC_FORCE);
+		if (ret < 0)
 			goto out;
-		}
-		path->slots[0]++;
 	}
 
-	BUG_ON(cur_pos + offset > last_byte);
-	if (cur_pos + offset < last_byte) {
-		ret = -ENOENT;
+	ret = set_block_group_ro(cache, 0);
+	if (!ret)
 		goto out;
-	}
-	ret = 0;
+	alloc_flags = get_alloc_profile(root, cache->space_info->flags);
+	ret = do_chunk_alloc(trans, root, alloc_flags,
+			     CHUNK_ALLOC_FORCE);
+	if (ret < 0)
+		goto out;
+	ret = set_block_group_ro(cache, 0);
 out:
-	btrfs_free_path(path);
-	if (ret) {
-		if (exts != *extents)
-			kfree(exts);
-	} else {
-		*extents = exts;
-		*nr_extents = nr;
-	}
+	btrfs_end_transaction(trans, root);
 	return ret;
 }
 
-static noinline int replace_one_extent(struct btrfs_trans_handle *trans,
-					struct btrfs_root *root,
-					struct btrfs_path *path,
-					struct btrfs_key *extent_key,
-					struct btrfs_key *leaf_key,
-					struct btrfs_ref_path *ref_path,
-					struct disk_extent *new_extents,
-					int nr_extents)
+int btrfs_force_chunk_alloc(struct btrfs_trans_handle *trans,
+			    struct btrfs_root *root, u64 type)
 {
-	struct extent_buffer *leaf;
-	struct btrfs_file_extent_item *fi;
-	struct inode *inode = NULL;
-	struct btrfs_key key;
-	u64 lock_start = 0;
-	u64 lock_end = 0;
-	u64 num_bytes;
-	u64 ext_offset;
-	u64 search_end = (u64)-1;
-	u32 nritems;
-	int nr_scaned = 0;
-	int extent_locked = 0;
-	int extent_type;
-	int ret;
-
-	memcpy(&key, leaf_key, sizeof(key));
-	if (ref_path->owner_objectid != BTRFS_MULTIPLE_OBJECTIDS) {
-		if (key.objectid < ref_path->owner_objectid ||
-		    (key.objectid == ref_path->owner_objectid &&
-		     key.type < BTRFS_EXTENT_DATA_KEY)) {
-			key.objectid = ref_path->owner_objectid;
-			key.type = BTRFS_EXTENT_DATA_KEY;
-			key.offset = 0;
-		}
-	}
-
-	while (1) {
-		ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
-		if (ret < 0)
-			goto out;
-
-		leaf = path->nodes[0];
-		nritems = btrfs_header_nritems(leaf);
-next:
-		if (extent_locked && ret > 0) {
-			/*
-			 * the file extent item was modified by someone
-			 * before the extent got locked.
-			 */
-			unlock_extent(&BTRFS_I(inode)->io_tree, lock_start,
-				      lock_end, GFP_NOFS);
-			extent_locked = 0;
-		}
-
-		if (path->slots[0] >= nritems) {
-			if (++nr_scaned > 2)
-				break;
-
-			BUG_ON(extent_locked);
-			ret = btrfs_next_leaf(root, path);
-			if (ret < 0)
-				goto out;
-			if (ret > 0)
-				break;
-			leaf = path->nodes[0];
-			nritems = btrfs_header_nritems(leaf);
-		}
-
-		btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
-
-		if (ref_path->owner_objectid != BTRFS_MULTIPLE_OBJECTIDS) {
-			if ((key.objectid > ref_path->owner_objectid) ||
-			    (key.objectid == ref_path->owner_objectid &&
-			     key.type > BTRFS_EXTENT_DATA_KEY) ||
-			    key.offset >= search_end)
-				break;
-		}
-
-		if (inode && key.objectid != inode->i_ino) {
-			BUG_ON(extent_locked);
-			btrfs_release_path(root, path);
-			mutex_unlock(&inode->i_mutex);
-			iput(inode);
-			inode = NULL;
-			continue;
-		}
-
-		if (key.type != BTRFS_EXTENT_DATA_KEY) {
-			path->slots[0]++;
-			ret = 1;
-			goto next;
-		}
-		fi = btrfs_item_ptr(leaf, path->slots[0],
-				    struct btrfs_file_extent_item);
-		extent_type = btrfs_file_extent_type(leaf, fi);
-		if ((extent_type != BTRFS_FILE_EXTENT_REG &&
-		     extent_type != BTRFS_FILE_EXTENT_PREALLOC) ||
-		    (btrfs_file_extent_disk_bytenr(leaf, fi) !=
-		     extent_key->objectid)) {
-			path->slots[0]++;
-			ret = 1;
-			goto next;
-		}
-
-		num_bytes = btrfs_file_extent_num_bytes(leaf, fi);
-		ext_offset = btrfs_file_extent_offset(leaf, fi);
-
-		if (search_end == (u64)-1) {
-			search_end = key.offset - ext_offset +
-				btrfs_file_extent_ram_bytes(leaf, fi);
-		}
-
-		if (!extent_locked) {
-			lock_start = key.offset;
-			lock_end = lock_start + num_bytes - 1;
-		} else {
-			if (lock_start > key.offset ||
-			    lock_end + 1 < key.offset + num_bytes) {
-				unlock_extent(&BTRFS_I(inode)->io_tree,
-					      lock_start, lock_end, GFP_NOFS);
-				extent_locked = 0;
-			}
-		}
-
-		if (!inode) {
-			btrfs_release_path(root, path);
-
-			inode = btrfs_iget_locked(root->fs_info->sb,
-						  key.objectid, root);
-			if (inode->i_state & I_NEW) {
-				BTRFS_I(inode)->root = root;
-				BTRFS_I(inode)->location.objectid =
-					key.objectid;
-				BTRFS_I(inode)->location.type =
-					BTRFS_INODE_ITEM_KEY;
-				BTRFS_I(inode)->location.offset = 0;
-				btrfs_read_locked_inode(inode);
-				unlock_new_inode(inode);
-			}
-			/*
-			 * some code call btrfs_commit_transaction while
-			 * holding the i_mutex, so we can't use mutex_lock
-			 * here.
-			 */
-			if (is_bad_inode(inode) ||
-			    !mutex_trylock(&inode->i_mutex)) {
-				iput(inode);
-				inode = NULL;
-				key.offset = (u64)-1;
-				goto skip;
-			}
-		}
-
-		if (!extent_locked) {
-			struct btrfs_ordered_extent *ordered;
-
-			btrfs_release_path(root, path);
-
-			lock_extent(&BTRFS_I(inode)->io_tree, lock_start,
-				    lock_end, GFP_NOFS);
-			ordered = btrfs_lookup_first_ordered_extent(inode,
-								    lock_end);
-			if (ordered &&
-			    ordered->file_offset <= lock_end &&
-			    ordered->file_offset + ordered->len > lock_start) {
-				unlock_extent(&BTRFS_I(inode)->io_tree,
-					      lock_start, lock_end, GFP_NOFS);
-				btrfs_start_ordered_extent(inode, ordered, 1);
-				btrfs_put_ordered_extent(ordered);
-				key.offset += num_bytes;
-				goto skip;
-			}
-			if (ordered)
-				btrfs_put_ordered_extent(ordered);
-
-			extent_locked = 1;
-			continue;
-		}
-
-		if (nr_extents == 1) {
-			/* update extent pointer in place */
-			btrfs_set_file_extent_disk_bytenr(leaf, fi,
-						new_extents[0].disk_bytenr);
-			btrfs_set_file_extent_disk_num_bytes(leaf, fi,
-						new_extents[0].disk_num_bytes);
-			btrfs_mark_buffer_dirty(leaf);
-
-			btrfs_drop_extent_cache(inode, key.offset,
-						key.offset + num_bytes - 1, 0);
-
-			ret = btrfs_inc_extent_ref(trans, root,
-						new_extents[0].disk_bytenr,
-						new_extents[0].disk_num_bytes,
-						leaf->start,
-						root->root_key.objectid,
-						trans->transid,
-						key.objectid);
-			BUG_ON(ret);
-
-			ret = btrfs_free_extent(trans, root,
-						extent_key->objectid,
-						extent_key->offset,
-						leaf->start,
-						btrfs_header_owner(leaf),
-						btrfs_header_generation(leaf),
-						key.objectid, 0);
-			BUG_ON(ret);
-
-			btrfs_release_path(root, path);
-			key.offset += num_bytes;
-		} else {
-			BUG_ON(1);
-#if 0
-			u64 alloc_hint;
-			u64 extent_len;
-			int i;
-			/*
-			 * drop old extent pointer at first, then insert the
-			 * new pointers one bye one
-			 */
-			btrfs_release_path(root, path);
-			ret = btrfs_drop_extents(trans, root, inode, key.offset,
-						 key.offset + num_bytes,
-						 key.offset, &alloc_hint);
-			BUG_ON(ret);
-
-			for (i = 0; i < nr_extents; i++) {
-				if (ext_offset >= new_extents[i].num_bytes) {
-					ext_offset -= new_extents[i].num_bytes;
-					continue;
-				}
-				extent_len = min(new_extents[i].num_bytes -
-						 ext_offset, num_bytes);
-
-				ret = btrfs_insert_empty_item(trans, root,
-							      path, &key,
-							      sizeof(*fi));
-				BUG_ON(ret);
-
-				leaf = path->nodes[0];
-				fi = btrfs_item_ptr(leaf, path->slots[0],
-						struct btrfs_file_extent_item);
-				btrfs_set_file_extent_generation(leaf, fi,
-							trans->transid);
-				btrfs_set_file_extent_type(leaf, fi,
-							BTRFS_FILE_EXTENT_REG);
-				btrfs_set_file_extent_disk_bytenr(leaf, fi,
-						new_extents[i].disk_bytenr);
-				btrfs_set_file_extent_disk_num_bytes(leaf, fi,
-						new_extents[i].disk_num_bytes);
-				btrfs_set_file_extent_ram_bytes(leaf, fi,
-						new_extents[i].ram_bytes);
-
-				btrfs_set_file_extent_compression(leaf, fi,
-						new_extents[i].compression);
-				btrfs_set_file_extent_encryption(leaf, fi,
-						new_extents[i].encryption);
-				btrfs_set_file_extent_other_encoding(leaf, fi,
-						new_extents[i].other_encoding);
-
-				btrfs_set_file_extent_num_bytes(leaf, fi,
-							extent_len);
-				ext_offset += new_extents[i].offset;
-				btrfs_set_file_extent_offset(leaf, fi,
-							ext_offset);
-				btrfs_mark_buffer_dirty(leaf);
-
-				btrfs_drop_extent_cache(inode, key.offset,
-						key.offset + extent_len - 1, 0);
-
-				ret = btrfs_inc_extent_ref(trans, root,
-						new_extents[i].disk_bytenr,
-						new_extents[i].disk_num_bytes,
-						leaf->start,
-						root->root_key.objectid,
-						trans->transid, key.objectid);
-				BUG_ON(ret);
-				btrfs_release_path(root, path);
-
-				inode_add_bytes(inode, extent_len);
-
-				ext_offset = 0;
-				num_bytes -= extent_len;
-				key.offset += extent_len;
-
-				if (num_bytes == 0)
-					break;
-			}
-			BUG_ON(i >= nr_extents);
-#endif
-		}
-
-		if (extent_locked) {
-			unlock_extent(&BTRFS_I(inode)->io_tree, lock_start,
-				      lock_end, GFP_NOFS);
-			extent_locked = 0;
-		}
-skip:
-		if (ref_path->owner_objectid != BTRFS_MULTIPLE_OBJECTIDS &&
-		    key.offset >= search_end)
-			break;
-
-		cond_resched();
-	}
-	ret = 0;
-out:
-	btrfs_release_path(root, path);
-	if (inode) {
-		mutex_unlock(&inode->i_mutex);
-		if (extent_locked) {
-			unlock_extent(&BTRFS_I(inode)->io_tree, lock_start,
-				      lock_end, GFP_NOFS);
-		}
-		iput(inode);
-	}
-	return ret;
+	u64 alloc_flags = get_alloc_profile(root, type);
+	return do_chunk_alloc(trans, root, alloc_flags,
+			      CHUNK_ALLOC_FORCE);
 }
 
-int btrfs_reloc_tree_cache_ref(struct btrfs_trans_handle *trans,
-			       struct btrfs_root *root,
-			       struct extent_buffer *buf, u64 orig_start)
-{
-	int level;
-	int ret;
-
-	BUG_ON(btrfs_header_generation(buf) != trans->transid);
-	BUG_ON(root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID);
-
-	level = btrfs_header_level(buf);
-	if (level == 0) {
-		struct btrfs_leaf_ref *ref;
-		struct btrfs_leaf_ref *orig_ref;
-
-		orig_ref = btrfs_lookup_leaf_ref(root, orig_start);
-		if (!orig_ref)
-			return -ENOENT;
-
-		ref = btrfs_alloc_leaf_ref(root, orig_ref->nritems);
-		if (!ref) {
-			btrfs_free_leaf_ref(root, orig_ref);
-			return -ENOMEM;
-		}
-
-		ref->nritems = orig_ref->nritems;
-		memcpy(ref->extents, orig_ref->extents,
-			sizeof(ref->extents[0]) * ref->nritems);
-
-		btrfs_free_leaf_ref(root, orig_ref);
-
-		ref->root_gen = trans->transid;
-		ref->bytenr = buf->start;
-		ref->owner = btrfs_header_owner(buf);
-		ref->generation = btrfs_header_generation(buf);
-
-		ret = btrfs_add_leaf_ref(root, ref, 0);
-		WARN_ON(ret);
-		btrfs_free_leaf_ref(root, ref);
-	}
-	return 0;
-}
-
-static noinline int invalidate_extent_cache(struct btrfs_root *root,
-					struct extent_buffer *leaf,
-					struct btrfs_block_group_cache *group,
-					struct btrfs_root *target_root)
-{
-	struct btrfs_key key;
-	struct inode *inode = NULL;
-	struct btrfs_file_extent_item *fi;
-	struct extent_state *cached_state = NULL;
-	u64 num_bytes;
-	u64 skip_objectid = 0;
-	u32 nritems;
-	u32 i;
-
-	nritems = btrfs_header_nritems(leaf);
-	for (i = 0; i < nritems; i++) {
-		btrfs_item_key_to_cpu(leaf, &key, i);
-		if (key.objectid == skip_objectid ||
-		    key.type != BTRFS_EXTENT_DATA_KEY)
-			continue;
-		fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
-		if (btrfs_file_extent_type(leaf, fi) ==
-		    BTRFS_FILE_EXTENT_INLINE)
-			continue;
-		if (btrfs_file_extent_disk_bytenr(leaf, fi) == 0)
-			continue;
-		if (!inode || inode->i_ino != key.objectid) {
-			iput(inode);
-			inode = btrfs_ilookup(target_root->fs_info->sb,
-					      key.objectid, target_root, 1);
-		}
-		if (!inode) {
-			skip_objectid = key.objectid;
-			continue;
-		}
-		num_bytes = btrfs_file_extent_num_bytes(leaf, fi);
-
-		lock_extent_bits(&BTRFS_I(inode)->io_tree, key.offset,
-				 key.offset + num_bytes - 1, 0, &cached_state,
-				 GFP_NOFS);
-		btrfs_drop_extent_cache(inode, key.offset,
-					key.offset + num_bytes - 1, 1);
-		unlock_extent_cached(&BTRFS_I(inode)->io_tree, key.offset,
-				     key.offset + num_bytes - 1, &cached_state,
-				     GFP_NOFS);
-		cond_resched();
-	}
-	iput(inode);
-	return 0;
-}
-
-static noinline int replace_extents_in_leaf(struct btrfs_trans_handle *trans,
-					struct btrfs_root *root,
-					struct extent_buffer *leaf,
-					struct btrfs_block_group_cache *group,
-					struct inode *reloc_inode)
+/*
+ * helper to account the unused space of all the readonly block group in the
+ * list. takes mirrors into account.
+ */
+static u64 __btrfs_get_ro_block_group_free_space(struct list_head *groups_list)
 {
-	struct btrfs_key key;
-	struct btrfs_key extent_key;
-	struct btrfs_file_extent_item *fi;
-	struct btrfs_leaf_ref *ref;
-	struct disk_extent *new_extent;
-	u64 bytenr;
-	u64 num_bytes;
-	u32 nritems;
-	u32 i;
-	int ext_index;
-	int nr_extent;
-	int ret;
-
-	new_extent = kmalloc(sizeof(*new_extent), GFP_NOFS);
-	BUG_ON(!new_extent);
-
-	ref = btrfs_lookup_leaf_ref(root, leaf->start);
-	BUG_ON(!ref);
-
-	ext_index = -1;
-	nritems = btrfs_header_nritems(leaf);
-	for (i = 0; i < nritems; i++) {
-		btrfs_item_key_to_cpu(leaf, &key, i);
-		if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
-			continue;
-		fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
-		if (btrfs_file_extent_type(leaf, fi) ==
-		    BTRFS_FILE_EXTENT_INLINE)
-			continue;
-		bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
-		num_bytes = btrfs_file_extent_disk_num_bytes(leaf, fi);
-		if (bytenr == 0)
-			continue;
+	struct btrfs_block_group_cache *block_group;
+	u64 free_bytes = 0;
+	int factor;
 
-		ext_index++;
-		if (bytenr >= group->key.objectid + group->key.offset ||
-		    bytenr + num_bytes <= group->key.objectid)
-			continue;
+	list_for_each_entry(block_group, groups_list, list) {
+		spin_lock(&block_group->lock);
 
-		extent_key.objectid = bytenr;
-		extent_key.offset = num_bytes;
-		extent_key.type = BTRFS_EXTENT_ITEM_KEY;
-		nr_extent = 1;
-		ret = get_new_locations(reloc_inode, &extent_key,
-					group->key.objectid, 1,
-					&new_extent, &nr_extent);
-		if (ret > 0)
+		if (!block_group->ro) {
+			spin_unlock(&block_group->lock);
 			continue;
-		BUG_ON(ret < 0);
-
-		BUG_ON(ref->extents[ext_index].bytenr != bytenr);
-		BUG_ON(ref->extents[ext_index].num_bytes != num_bytes);
-		ref->extents[ext_index].bytenr = new_extent->disk_bytenr;
-		ref->extents[ext_index].num_bytes = new_extent->disk_num_bytes;
-
-		btrfs_set_file_extent_disk_bytenr(leaf, fi,
-						new_extent->disk_bytenr);
-		btrfs_set_file_extent_disk_num_bytes(leaf, fi,
-						new_extent->disk_num_bytes);
-		btrfs_mark_buffer_dirty(leaf);
-
-		ret = btrfs_inc_extent_ref(trans, root,
-					new_extent->disk_bytenr,
-					new_extent->disk_num_bytes,
-					leaf->start,
-					root->root_key.objectid,
-					trans->transid, key.objectid);
-		BUG_ON(ret);
-
-		ret = btrfs_free_extent(trans, root,
-					bytenr, num_bytes, leaf->start,
-					btrfs_header_owner(leaf),
-					btrfs_header_generation(leaf),
-					key.objectid, 0);
-		BUG_ON(ret);
-		cond_resched();
-	}
-	kfree(new_extent);
-	BUG_ON(ext_index + 1 != ref->nritems);
-	btrfs_free_leaf_ref(root, ref);
-	return 0;
-}
-
-int btrfs_free_reloc_root(struct btrfs_trans_handle *trans,
-			  struct btrfs_root *root)
-{
-	struct btrfs_root *reloc_root;
-	int ret;
-
-	if (root->reloc_root) {
-		reloc_root = root->reloc_root;
-		root->reloc_root = NULL;
-		list_add(&reloc_root->dead_list,
-			 &root->fs_info->dead_reloc_roots);
-
-		btrfs_set_root_bytenr(&reloc_root->root_item,
-				      reloc_root->node->start);
-		btrfs_set_root_level(&root->root_item,
-				     btrfs_header_level(reloc_root->node));
-		memset(&reloc_root->root_item.drop_progress, 0,
-			sizeof(struct btrfs_disk_key));
-		reloc_root->root_item.drop_level = 0;
-
-		ret = btrfs_update_root(trans, root->fs_info->tree_root,
-					&reloc_root->root_key,
-					&reloc_root->root_item);
-		BUG_ON(ret);
-	}
-	return 0;
-}
-
-int btrfs_drop_dead_reloc_roots(struct btrfs_root *root)
-{
-	struct btrfs_trans_handle *trans;
-	struct btrfs_root *reloc_root;
-	struct btrfs_root *prev_root = NULL;
-	struct list_head dead_roots;
-	int ret;
-	unsigned long nr;
-
-	INIT_LIST_HEAD(&dead_roots);
-	list_splice_init(&root->fs_info->dead_reloc_roots, &dead_roots);
-
-	while (!list_empty(&dead_roots)) {
-		reloc_root = list_entry(dead_roots.prev,
-					struct btrfs_root, dead_list);
-		list_del_init(&reloc_root->dead_list);
-
-		BUG_ON(reloc_root->commit_root != NULL);
-		while (1) {
-			trans = btrfs_join_transaction(root, 1);
-			BUG_ON(!trans);
-
-			mutex_lock(&root->fs_info->drop_mutex);
-			ret = btrfs_drop_snapshot(trans, reloc_root);
-			if (ret != -EAGAIN)
-				break;
-			mutex_unlock(&root->fs_info->drop_mutex);
-
-			nr = trans->blocks_used;
-			ret = btrfs_end_transaction(trans, root);
-			BUG_ON(ret);
-			btrfs_btree_balance_dirty(root, nr);
 		}
 
-		free_extent_buffer(reloc_root->node);
-
-		ret = btrfs_del_root(trans, root->fs_info->tree_root,
-				     &reloc_root->root_key);
-		BUG_ON(ret);
-		mutex_unlock(&root->fs_info->drop_mutex);
-
-		nr = trans->blocks_used;
-		ret = btrfs_end_transaction(trans, root);
-		BUG_ON(ret);
-		btrfs_btree_balance_dirty(root, nr);
-
-		kfree(prev_root);
-		prev_root = reloc_root;
-	}
-	if (prev_root) {
-		btrfs_remove_leaf_refs(prev_root, (u64)-1, 0);
-		kfree(prev_root);
-	}
-	return 0;
-}
-
-int btrfs_add_dead_reloc_root(struct btrfs_root *root)
-{
-	list_add(&root->dead_list, &root->fs_info->dead_reloc_roots);
-	return 0;
-}
-
-int btrfs_cleanup_reloc_trees(struct btrfs_root *root)
-{
-	struct btrfs_root *reloc_root;
-	struct btrfs_trans_handle *trans;
-	struct btrfs_key location;
-	int found;
-	int ret;
+		if (block_group->flags & (BTRFS_BLOCK_GROUP_RAID1 |
+					  BTRFS_BLOCK_GROUP_RAID10 |
+					  BTRFS_BLOCK_GROUP_DUP))
+			factor = 2;
+		else
+			factor = 1;
 
-	mutex_lock(&root->fs_info->tree_reloc_mutex);
-	ret = btrfs_find_dead_roots(root, BTRFS_TREE_RELOC_OBJECTID, NULL);
-	BUG_ON(ret);
-	found = !list_empty(&root->fs_info->dead_reloc_roots);
-	mutex_unlock(&root->fs_info->tree_reloc_mutex);
+		free_bytes += (block_group->key.offset -
+			       btrfs_block_group_used(&block_group->item)) *
+			       factor;
 
-	if (found) {
-		trans = btrfs_start_transaction(root, 1);
-		BUG_ON(!trans);
-		ret = btrfs_commit_transaction(trans, root);
-		BUG_ON(ret);
+		spin_unlock(&block_group->lock);
 	}
 
-	location.objectid = BTRFS_DATA_RELOC_TREE_OBJECTID;
-	location.offset = (u64)-1;
-	location.type = BTRFS_ROOT_ITEM_KEY;
-
-	reloc_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
-	BUG_ON(!reloc_root);
-	btrfs_orphan_cleanup(reloc_root);
-	return 0;
-}
-
-static noinline int init_reloc_tree(struct btrfs_trans_handle *trans,
-				    struct btrfs_root *root)
-{
-	struct btrfs_root *reloc_root;
-	struct extent_buffer *eb;
-	struct btrfs_root_item *root_item;
-	struct btrfs_key root_key;
-	int ret;
-
-	BUG_ON(!root->ref_cows);
-	if (root->reloc_root)
-		return 0;
-
-	root_item = kmalloc(sizeof(*root_item), GFP_NOFS);
-	BUG_ON(!root_item);
-
-	ret = btrfs_copy_root(trans, root, root->commit_root,
-			      &eb, BTRFS_TREE_RELOC_OBJECTID);
-	BUG_ON(ret);
-
-	root_key.objectid = BTRFS_TREE_RELOC_OBJECTID;
-	root_key.offset = root->root_key.objectid;
-	root_key.type = BTRFS_ROOT_ITEM_KEY;
-
-	memcpy(root_item, &root->root_item, sizeof(root_item));
-	btrfs_set_root_refs(root_item, 0);
-	btrfs_set_root_bytenr(root_item, eb->start);
-	btrfs_set_root_level(root_item, btrfs_header_level(eb));
-	btrfs_set_root_generation(root_item, trans->transid);
-
-	btrfs_tree_unlock(eb);
-	free_extent_buffer(eb);
-
-	ret = btrfs_insert_root(trans, root->fs_info->tree_root,
-				&root_key, root_item);
-	BUG_ON(ret);
-	kfree(root_item);
-
-	reloc_root = btrfs_read_fs_root_no_radix(root->fs_info->tree_root,
-						 &root_key);
-	BUG_ON(!reloc_root);
-	reloc_root->last_trans = trans->transid;
-	reloc_root->commit_root = NULL;
-	reloc_root->ref_tree = &root->fs_info->reloc_ref_tree;
-
-	root->reloc_root = reloc_root;
-	return 0;
+	return free_bytes;
 }
 
 /*
- * Core function of space balance.
- *
- * The idea is using reloc trees to relocate tree blocks in reference
- * counted roots. There is one reloc tree for each subvol, and all
- * reloc trees share same root key objectid. Reloc trees are snapshots
- * of the latest committed roots of subvols (root->commit_root).
- *
- * To relocate a tree block referenced by a subvol, there are two steps.
- * COW the block through subvol's reloc tree, then update block pointer
- * in the subvol to point to the new block. Since all reloc trees share
- * same root key objectid, doing special handing for tree blocks owned
- * by them is easy. Once a tree block has been COWed in one reloc tree,
- * we can use the resulting new block directly when the same block is
- * required to COW again through other reloc trees. By this way, relocated
- * tree blocks are shared between reloc trees, so they are also shared
- * between subvols.
+ * helper to account the unused space of all the readonly block group in the
+ * space_info. takes mirrors into account.
  */
-static noinline int relocate_one_path(struct btrfs_trans_handle *trans,
-				      struct btrfs_root *root,
-				      struct btrfs_path *path,
-				      struct btrfs_key *first_key,
-				      struct btrfs_ref_path *ref_path,
-				      struct btrfs_block_group_cache *group,
-				      struct inode *reloc_inode)
-{
-	struct btrfs_root *reloc_root;
-	struct extent_buffer *eb = NULL;
-	struct btrfs_key *keys;
-	u64 *nodes;
-	int level;
-	int shared_level;
-	int lowest_level = 0;
-	int ret;
-
-	if (ref_path->owner_objectid < BTRFS_FIRST_FREE_OBJECTID)
-		lowest_level = ref_path->owner_objectid;
-
-	if (!root->ref_cows) {
-		path->lowest_level = lowest_level;
-		ret = btrfs_search_slot(trans, root, first_key, path, 0, 1);
-		BUG_ON(ret < 0);
-		path->lowest_level = 0;
-		btrfs_release_path(root, path);
-		return 0;
-	}
-
-	mutex_lock(&root->fs_info->tree_reloc_mutex);
-	ret = init_reloc_tree(trans, root);
-	BUG_ON(ret);
-	reloc_root = root->reloc_root;
-
-	shared_level = ref_path->shared_level;
-	ref_path->shared_level = BTRFS_MAX_LEVEL - 1;
-
-	keys = ref_path->node_keys;
-	nodes = ref_path->new_nodes;
-	memset(&keys[shared_level + 1], 0,
-	       sizeof(*keys) * (BTRFS_MAX_LEVEL - shared_level - 1));
-	memset(&nodes[shared_level + 1], 0,
-	       sizeof(*nodes) * (BTRFS_MAX_LEVEL - shared_level - 1));
-
-	if (nodes[lowest_level] == 0) {
-		path->lowest_level = lowest_level;
-		ret = btrfs_search_slot(trans, reloc_root, first_key, path,
-					0, 1);
-		BUG_ON(ret);
-		for (level = lowest_level; level < BTRFS_MAX_LEVEL; level++) {
-			eb = path->nodes[level];
-			if (!eb || eb == reloc_root->node)
-				break;
-			nodes[level] = eb->start;
-			if (level == 0)
-				btrfs_item_key_to_cpu(eb, &keys[level], 0);
-			else
-				btrfs_node_key_to_cpu(eb, &keys[level], 0);
-		}
-		if (nodes[0] &&
-		    ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
-			eb = path->nodes[0];
-			ret = replace_extents_in_leaf(trans, reloc_root, eb,
-						      group, reloc_inode);
-			BUG_ON(ret);
-		}
-		btrfs_release_path(reloc_root, path);
-	} else {
-		ret = btrfs_merge_path(trans, reloc_root, keys, nodes,
-				       lowest_level);
-		BUG_ON(ret);
-	}
-
-	/*
-	 * replace tree blocks in the fs tree with tree blocks in
-	 * the reloc tree.
-	 */
-	ret = btrfs_merge_path(trans, root, keys, nodes, lowest_level);
-	BUG_ON(ret < 0);
-
-	if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
-		ret = btrfs_search_slot(trans, reloc_root, first_key, path,
-					0, 0);
-		BUG_ON(ret);
-		extent_buffer_get(path->nodes[0]);
-		eb = path->nodes[0];
-		btrfs_release_path(reloc_root, path);
-		ret = invalidate_extent_cache(reloc_root, eb, group, root);
-		BUG_ON(ret);
-		free_extent_buffer(eb);
-	}
-
-	mutex_unlock(&root->fs_info->tree_reloc_mutex);
-	path->lowest_level = 0;
-	return 0;
-}
-
-static noinline int relocate_tree_block(struct btrfs_trans_handle *trans,
-					struct btrfs_root *root,
-					struct btrfs_path *path,
-					struct btrfs_key *first_key,
-					struct btrfs_ref_path *ref_path)
-{
-	int ret;
-
-	ret = relocate_one_path(trans, root, path, first_key,
-				ref_path, NULL, NULL);
-	BUG_ON(ret);
-
-	return 0;
-}
-
-static noinline int del_extent_zero(struct btrfs_trans_handle *trans,
-				    struct btrfs_root *extent_root,
-				    struct btrfs_path *path,
-				    struct btrfs_key *extent_key)
-{
-	int ret;
-
-	ret = btrfs_search_slot(trans, extent_root, extent_key, path, -1, 1);
-	if (ret)
-		goto out;
-	ret = btrfs_del_item(trans, extent_root, path);
-out:
-	btrfs_release_path(extent_root, path);
-	return ret;
-}
-
-static noinline struct btrfs_root *read_ref_root(struct btrfs_fs_info *fs_info,
-						struct btrfs_ref_path *ref_path)
-{
-	struct btrfs_key root_key;
-
-	root_key.objectid = ref_path->root_objectid;
-	root_key.type = BTRFS_ROOT_ITEM_KEY;
-	if (is_cowonly_root(ref_path->root_objectid))
-		root_key.offset = 0;
-	else
-		root_key.offset = (u64)-1;
-
-	return btrfs_read_fs_root_no_name(fs_info, &root_key);
-}
-
-static noinline int relocate_one_extent(struct btrfs_root *extent_root,
-					struct btrfs_path *path,
-					struct btrfs_key *extent_key,
-					struct btrfs_block_group_cache *group,
-					struct inode *reloc_inode, int pass)
+u64 btrfs_account_ro_block_groups_free_space(struct btrfs_space_info *sinfo)
 {
-	struct btrfs_trans_handle *trans;
-	struct btrfs_root *found_root;
-	struct btrfs_ref_path *ref_path = NULL;
-	struct disk_extent *new_extents = NULL;
-	int nr_extents = 0;
-	int loops;
-	int ret;
-	int level;
-	struct btrfs_key first_key;
-	u64 prev_block = 0;
-
-
-	trans = btrfs_start_transaction(extent_root, 1);
-	BUG_ON(!trans);
-
-	if (extent_key->objectid == 0) {
-		ret = del_extent_zero(trans, extent_root, path, extent_key);
-		goto out;
-	}
-
-	ref_path = kmalloc(sizeof(*ref_path), GFP_NOFS);
-	if (!ref_path) {
-		ret = -ENOMEM;
-		goto out;
-	}
-
-	for (loops = 0; ; loops++) {
-		if (loops == 0) {
-			ret = btrfs_first_ref_path(trans, extent_root, ref_path,
-						   extent_key->objectid);
-		} else {
-			ret = btrfs_next_ref_path(trans, extent_root, ref_path);
-		}
-		if (ret < 0)
-			goto out;
-		if (ret > 0)
-			break;
-
-		if (ref_path->root_objectid == BTRFS_TREE_LOG_OBJECTID ||
-		    ref_path->root_objectid == BTRFS_TREE_RELOC_OBJECTID)
-			continue;
-
-		found_root = read_ref_root(extent_root->fs_info, ref_path);
-		BUG_ON(!found_root);
-		/*
-		 * for reference counted tree, only process reference paths
-		 * rooted at the latest committed root.
-		 */
-		if (found_root->ref_cows &&
-		    ref_path->root_generation != found_root->root_key.offset)
-			continue;
-
-		if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
-			if (pass == 0) {
-				/*
-				 * copy data extents to new locations
-				 */
-				u64 group_start = group->key.objectid;
-				ret = relocate_data_extent(reloc_inode,
-							   extent_key,
-							   group_start);
-				if (ret < 0)
-					goto out;
-				break;
-			}
-			level = 0;
-		} else {
-			level = ref_path->owner_objectid;
-		}
-
-		if (prev_block != ref_path->nodes[level]) {
-			struct extent_buffer *eb;
-			u64 block_start = ref_path->nodes[level];
-			u64 block_size = btrfs_level_size(found_root, level);
-
-			eb = read_tree_block(found_root, block_start,
-					     block_size, 0);
-			btrfs_tree_lock(eb);
-			BUG_ON(level != btrfs_header_level(eb));
-
-			if (level == 0)
-				btrfs_item_key_to_cpu(eb, &first_key, 0);
-			else
-				btrfs_node_key_to_cpu(eb, &first_key, 0);
-
-			btrfs_tree_unlock(eb);
-			free_extent_buffer(eb);
-			prev_block = block_start;
-		}
-
-		mutex_lock(&extent_root->fs_info->trans_mutex);
-		btrfs_record_root_in_trans(found_root);
-		mutex_unlock(&extent_root->fs_info->trans_mutex);
-		if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
-			/*
-			 * try to update data extent references while
-			 * keeping metadata shared between snapshots.
-			 */
-			if (pass == 1) {
-				ret = relocate_one_path(trans, found_root,
-						path, &first_key, ref_path,
-						group, reloc_inode);
-				if (ret < 0)
-					goto out;
-				continue;
-			}
-			/*
-			 * use fallback method to process the remaining
-			 * references.
-			 */
-			if (!new_extents) {
-				u64 group_start = group->key.objectid;
-				new_extents = kmalloc(sizeof(*new_extents),
-						      GFP_NOFS);
-				nr_extents = 1;
-				ret = get_new_locations(reloc_inode,
-							extent_key,
-							group_start, 1,
-							&new_extents,
-							&nr_extents);
-				if (ret)
-					goto out;
-			}
-			ret = replace_one_extent(trans, found_root,
-						path, extent_key,
-						&first_key, ref_path,
-						new_extents, nr_extents);
-		} else {
-			ret = relocate_tree_block(trans, found_root, path,
-						  &first_key, ref_path);
-		}
-		if (ret < 0)
-			goto out;
-	}
-	ret = 0;
-out:
-	btrfs_end_transaction(trans, extent_root);
-	kfree(new_extents);
-	kfree(ref_path);
-	return ret;
-}
-#endif
-
-static u64 update_block_group_flags(struct btrfs_root *root, u64 flags)
-{
-	u64 num_devices;
-	u64 stripped = BTRFS_BLOCK_GROUP_RAID0 |
-		BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID10;
-
-	num_devices = root->fs_info->fs_devices->rw_devices;
-	if (num_devices == 1) {
-		stripped |= BTRFS_BLOCK_GROUP_DUP;
-		stripped = flags & ~stripped;
-
-		/* turn raid0 into single device chunks */
-		if (flags & BTRFS_BLOCK_GROUP_RAID0)
-			return stripped;
+	int i;
+	u64 free_bytes = 0;
 
-		/* turn mirroring into duplication */
-		if (flags & (BTRFS_BLOCK_GROUP_RAID1 |
-			     BTRFS_BLOCK_GROUP_RAID10))
-			return stripped | BTRFS_BLOCK_GROUP_DUP;
-		return flags;
-	} else {
-		/* they already had raid on here, just return */
-		if (flags & stripped)
-			return flags;
+	spin_lock(&sinfo->lock);
 
-		stripped |= BTRFS_BLOCK_GROUP_DUP;
-		stripped = flags & ~stripped;
+	for (i = 0; i < BTRFS_NR_RAID_TYPES; i++)
+		if (!list_empty(&sinfo->block_groups[i]))
+			free_bytes += __btrfs_get_ro_block_group_free_space(
+						&sinfo->block_groups[i]);
 
-		/* switch duplicated blocks with raid1 */
-		if (flags & BTRFS_BLOCK_GROUP_DUP)
-			return stripped | BTRFS_BLOCK_GROUP_RAID1;
+	spin_unlock(&sinfo->lock);
 
-		/* turn single device chunks into raid0 */
-		return stripped | BTRFS_BLOCK_GROUP_RAID0;
-	}
-	return flags;
+	return free_bytes;
 }
 
-static int __alloc_chunk_for_shrink(struct btrfs_root *root,
-		     struct btrfs_block_group_cache *shrink_block_group,
-		     int force)
+void btrfs_set_block_group_rw(struct btrfs_root *root,
+			      struct btrfs_block_group_cache *cache)
 {
-	struct btrfs_trans_handle *trans;
-	u64 new_alloc_flags;
-	u64 calc;
-
-	spin_lock(&shrink_block_group->lock);
-	if (btrfs_block_group_used(&shrink_block_group->item) +
-	    shrink_block_group->reserved > 0) {
-		spin_unlock(&shrink_block_group->lock);
-
-		trans = btrfs_start_transaction(root, 1);
-		spin_lock(&shrink_block_group->lock);
-
-		new_alloc_flags = update_block_group_flags(root,
-						   shrink_block_group->flags);
-		if (new_alloc_flags != shrink_block_group->flags) {
-			calc =
-			     btrfs_block_group_used(&shrink_block_group->item);
-		} else {
-			calc = shrink_block_group->key.offset;
-		}
-		spin_unlock(&shrink_block_group->lock);
-
-		do_chunk_alloc(trans, root->fs_info->extent_root,
-			       calc + 2 * 1024 * 1024, new_alloc_flags, force);
-
-		btrfs_end_transaction(trans, root);
-	} else
-		spin_unlock(&shrink_block_group->lock);
-	return 0;
-}
-
+	struct btrfs_space_info *sinfo = cache->space_info;
+	u64 num_bytes;
 
-int btrfs_prepare_block_group_relocation(struct btrfs_root *root,
-					 struct btrfs_block_group_cache *group)
+	BUG_ON(!cache->ro);
 
-{
-	__alloc_chunk_for_shrink(root, group, 1);
-	set_block_group_readonly(group);
-	return 0;
+	spin_lock(&sinfo->lock);
+	spin_lock(&cache->lock);
+	num_bytes = cache->key.offset - cache->reserved - cache->pinned -
+		    cache->bytes_super - btrfs_block_group_used(&cache->item);
+	sinfo->bytes_readonly -= num_bytes;
+	cache->ro = 0;
+	spin_unlock(&cache->lock);
+	spin_unlock(&sinfo->lock);
 }
 
 /*
@@ -7275,6 +8397,12 @@ int btrfs_can_relocate(struct btrfs_root *root, u64 bytenr)
 	struct btrfs_space_info *space_info;
 	struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
 	struct btrfs_device *device;
+	struct btrfs_trans_handle *trans;
+	u64 min_free;
+	u64 dev_min = 1;
+	u64 dev_nr = 0;
+	u64 target;
+	int index;
 	int full = 0;
 	int ret = 0;
 
@@ -7284,8 +8412,10 @@ int btrfs_can_relocate(struct btrfs_root *root, u64 bytenr)
 	if (!block_group)
 		return -1;
 
+	min_free = btrfs_block_group_used(&block_group->item);
+
 	/* no bytes used, we're good */
-	if (!btrfs_block_group_used(&block_group->item))
+	if (!min_free)
 		goto out;
 
 	space_info = block_group->space_info;
@@ -7301,10 +8431,9 @@ int btrfs_can_relocate(struct btrfs_root *root, u64 bytenr)
 	 * all of the extents from this block group.  If we can, we're good
 	 */
 	if ((space_info->total_bytes != block_group->key.offset) &&
-	   (space_info->bytes_used + space_info->bytes_reserved +
-	    space_info->bytes_pinned + space_info->bytes_readonly +
-	    btrfs_block_group_used(&block_group->item) <
-	    space_info->total_bytes)) {
+	    (space_info->bytes_used + space_info->bytes_reserved +
+	     space_info->bytes_pinned + space_info->bytes_readonly +
+	     min_free < space_info->total_bytes)) {
 		spin_unlock(&space_info->lock);
 		goto out;
 	}
@@ -7313,32 +8442,78 @@ int btrfs_can_relocate(struct btrfs_root *root, u64 bytenr)
 	/*
 	 * ok we don't have enough space, but maybe we have free space on our
 	 * devices to allocate new chunks for relocation, so loop through our
-	 * alloc devices and guess if we have enough space.  However, if we
-	 * were marked as full, then we know there aren't enough chunks, and we
-	 * can just return.
+	 * alloc devices and guess if we have enough space.  if this block
+	 * group is going to be restriped, run checks against the target
+	 * profile instead of the current one.
 	 */
 	ret = -1;
-	if (full)
+
+	/*
+	 * index:
+	 *      0: raid10
+	 *      1: raid1
+	 *      2: dup
+	 *      3: raid0
+	 *      4: single
+	 */
+	target = get_restripe_target(root->fs_info, block_group->flags);
+	if (target) {
+		index = __get_raid_index(extended_to_chunk(target));
+	} else {
+		/*
+		 * this is just a balance, so if we were marked as full
+		 * we know there is no space for a new chunk
+		 */
+		if (full)
+			goto out;
+
+		index = get_block_group_index(block_group);
+	}
+
+	if (index == BTRFS_RAID_RAID10) {
+		dev_min = 4;
+		/* Divide by 2 */
+		min_free >>= 1;
+	} else if (index == BTRFS_RAID_RAID1) {
+		dev_min = 2;
+	} else if (index == BTRFS_RAID_DUP) {
+		/* Multiply by 2 */
+		min_free <<= 1;
+	} else if (index == BTRFS_RAID_RAID0) {
+		dev_min = fs_devices->rw_devices;
+		do_div(min_free, dev_min);
+	}
+
+	/* We need to do this so that we can look at pending chunks */
+	trans = btrfs_join_transaction(root);
+	if (IS_ERR(trans)) {
+		ret = PTR_ERR(trans);
 		goto out;
+	}
 
 	mutex_lock(&root->fs_info->chunk_mutex);
 	list_for_each_entry(device, &fs_devices->alloc_list, dev_alloc_list) {
-		u64 min_free = btrfs_block_group_used(&block_group->item);
-		u64 dev_offset, max_avail;
+		u64 dev_offset;
 
 		/*
 		 * check to make sure we can actually find a chunk with enough
 		 * space to fit our block group in.
 		 */
-		if (device->total_bytes > device->bytes_used + min_free) {
-			ret = find_free_dev_extent(NULL, device, min_free,
-						   &dev_offset, &max_avail);
+		if (device->total_bytes > device->bytes_used + min_free &&
+		    !device->is_tgtdev_for_dev_replace) {
+			ret = find_free_dev_extent(trans, device, min_free,
+						   &dev_offset, NULL);
 			if (!ret)
+				dev_nr++;
+
+			if (dev_nr >= dev_min)
 				break;
+
 			ret = -1;
 		}
 	}
 	mutex_unlock(&root->fs_info->chunk_mutex);
+	btrfs_end_transaction(trans, root);
 out:
 	btrfs_put_block_group(block_group);
 	return ret;
@@ -7380,6 +8555,40 @@ out:
 	return ret;
 }
 
+void btrfs_put_block_group_cache(struct btrfs_fs_info *info)
+{
+	struct btrfs_block_group_cache *block_group;
+	u64 last = 0;
+
+	while (1) {
+		struct inode *inode;
+
+		block_group = btrfs_lookup_first_block_group(info, last);
+		while (block_group) {
+			spin_lock(&block_group->lock);
+			if (block_group->iref)
+				break;
+			spin_unlock(&block_group->lock);
+			block_group = next_block_group(info->tree_root,
+						       block_group);
+		}
+		if (!block_group) {
+			if (last == 0)
+				break;
+			last = 0;
+			continue;
+		}
+
+		inode = block_group->inode;
+		block_group->iref = 0;
+		block_group->inode = NULL;
+		spin_unlock(&block_group->lock);
+		iput(inode);
+		last = block_group->key.objectid + block_group->key.offset;
+		btrfs_put_block_group(block_group);
+	}
+}
+
 int btrfs_free_block_groups(struct btrfs_fs_info *info)
 {
 	struct btrfs_block_group_cache *block_group;
@@ -7387,14 +8596,14 @@ int btrfs_free_block_groups(struct btrfs_fs_info *info)
 	struct btrfs_caching_control *caching_ctl;
 	struct rb_node *n;
 
-	down_write(&info->extent_commit_sem);
+	down_write(&info->commit_root_sem);
 	while (!list_empty(&info->caching_block_groups)) {
 		caching_ctl = list_entry(info->caching_block_groups.next,
 					 struct btrfs_caching_control, list);
 		list_del(&caching_ctl->list);
 		put_caching_control(caching_ctl);
 	}
-	up_write(&info->extent_commit_sem);
+	up_write(&info->commit_root_sem);
 
 	spin_lock(&info->block_group_cache_lock);
 	while ((n = rb_last(&info->block_group_cache_tree)) != NULL) {
@@ -7411,6 +8620,14 @@ int btrfs_free_block_groups(struct btrfs_fs_info *info)
 		if (block_group->cached == BTRFS_CACHE_STARTED)
 			wait_block_group_cache_done(block_group);
 
+		/*
+		 * We haven't cached this block group, which means we could
+		 * possibly have excluded extents on this block group.
+		 */
+		if (block_group->cached == BTRFS_CACHE_NO ||
+		    block_group->cached == BTRFS_CACHE_ERROR)
+			free_excluded_extents(info->extent_root, block_group);
+
 		btrfs_remove_free_space_cache(block_group);
 		btrfs_put_block_group(block_group);
 
@@ -7426,17 +8643,108 @@ int btrfs_free_block_groups(struct btrfs_fs_info *info)
 	 */
 	synchronize_rcu();
 
-	while(!list_empty(&info->space_info)) {
+	release_global_block_rsv(info);
+
+	while (!list_empty(&info->space_info)) {
+		int i;
+
 		space_info = list_entry(info->space_info.next,
 					struct btrfs_space_info,
 					list);
-
+		if (btrfs_test_opt(info->tree_root, ENOSPC_DEBUG)) {
+			if (WARN_ON(space_info->bytes_pinned > 0 ||
+			    space_info->bytes_reserved > 0 ||
+			    space_info->bytes_may_use > 0)) {
+				dump_space_info(space_info, 0, 0);
+			}
+		}
 		list_del(&space_info->list);
-		kfree(space_info);
+		for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
+			struct kobject *kobj;
+			kobj = space_info->block_group_kobjs[i];
+			space_info->block_group_kobjs[i] = NULL;
+			if (kobj) {
+				kobject_del(kobj);
+				kobject_put(kobj);
+			}
+		}
+		kobject_del(&space_info->kobj);
+		kobject_put(&space_info->kobj);
 	}
 	return 0;
 }
 
+static void __link_block_group(struct btrfs_space_info *space_info,
+			       struct btrfs_block_group_cache *cache)
+{
+	int index = get_block_group_index(cache);
+	bool first = false;
+
+	down_write(&space_info->groups_sem);
+	if (list_empty(&space_info->block_groups[index]))
+		first = true;
+	list_add_tail(&cache->list, &space_info->block_groups[index]);
+	up_write(&space_info->groups_sem);
+
+	if (first) {
+		struct raid_kobject *rkobj;
+		int ret;
+
+		rkobj = kzalloc(sizeof(*rkobj), GFP_NOFS);
+		if (!rkobj)
+			goto out_err;
+		rkobj->raid_type = index;
+		kobject_init(&rkobj->kobj, &btrfs_raid_ktype);
+		ret = kobject_add(&rkobj->kobj, &space_info->kobj,
+				  "%s", get_raid_name(index));
+		if (ret) {
+			kobject_put(&rkobj->kobj);
+			goto out_err;
+		}
+		space_info->block_group_kobjs[index] = &rkobj->kobj;
+	}
+
+	return;
+out_err:
+	pr_warn("BTRFS: failed to add kobject for block cache. ignoring.\n");
+}
+
+static struct btrfs_block_group_cache *
+btrfs_create_block_group_cache(struct btrfs_root *root, u64 start, u64 size)
+{
+	struct btrfs_block_group_cache *cache;
+
+	cache = kzalloc(sizeof(*cache), GFP_NOFS);
+	if (!cache)
+		return NULL;
+
+	cache->free_space_ctl = kzalloc(sizeof(*cache->free_space_ctl),
+					GFP_NOFS);
+	if (!cache->free_space_ctl) {
+		kfree(cache);
+		return NULL;
+	}
+
+	cache->key.objectid = start;
+	cache->key.offset = size;
+	cache->key.type = BTRFS_BLOCK_GROUP_ITEM_KEY;
+
+	cache->sectorsize = root->sectorsize;
+	cache->fs_info = root->fs_info;
+	cache->full_stripe_len = btrfs_full_stripe_len(root,
+					       &root->fs_info->mapping_tree,
+					       start);
+	atomic_set(&cache->count, 1);
+	spin_lock_init(&cache->lock);
+	init_rwsem(&cache->data_rwsem);
+	INIT_LIST_HEAD(&cache->list);
+	INIT_LIST_HEAD(&cache->cluster_list);
+	INIT_LIST_HEAD(&cache->new_bg_list);
+	btrfs_init_free_space_ctl(cache);
+
+	return cache;
+}
+
 int btrfs_read_block_groups(struct btrfs_root *root)
 {
 	struct btrfs_path *path;
@@ -7447,6 +8755,8 @@ int btrfs_read_block_groups(struct btrfs_root *root)
 	struct btrfs_key key;
 	struct btrfs_key found_key;
 	struct extent_buffer *leaf;
+	int need_clear = 0;
+	u64 cache_gen;
 
 	root = info->extent_root;
 	key.objectid = 0;
@@ -7455,48 +8765,71 @@ int btrfs_read_block_groups(struct btrfs_root *root)
 	path = btrfs_alloc_path();
 	if (!path)
 		return -ENOMEM;
+	path->reada = 1;
+
+	cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
+	if (btrfs_test_opt(root, SPACE_CACHE) &&
+	    btrfs_super_generation(root->fs_info->super_copy) != cache_gen)
+		need_clear = 1;
+	if (btrfs_test_opt(root, CLEAR_CACHE))
+		need_clear = 1;
 
 	while (1) {
 		ret = find_first_block_group(root, path, &key);
-		if (ret > 0) {
-			ret = 0;
-			goto error;
-		}
+		if (ret > 0)
+			break;
 		if (ret != 0)
 			goto error;
 
 		leaf = path->nodes[0];
 		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
-		cache = kzalloc(sizeof(*cache), GFP_NOFS);
+
+		cache = btrfs_create_block_group_cache(root, found_key.objectid,
+						       found_key.offset);
 		if (!cache) {
 			ret = -ENOMEM;
-			break;
+			goto error;
 		}
 
-		atomic_set(&cache->count, 1);
-		spin_lock_init(&cache->lock);
-		spin_lock_init(&cache->tree_lock);
-		cache->fs_info = info;
-		INIT_LIST_HEAD(&cache->list);
-		INIT_LIST_HEAD(&cache->cluster_list);
-
-		/*
-		 * we only want to have 32k of ram per block group for keeping
-		 * track of free space, and if we pass 1/2 of that we want to
-		 * start converting things over to using bitmaps
-		 */
-		cache->extents_thresh = ((1024 * 32) / 2) /
-			sizeof(struct btrfs_free_space);
+		if (need_clear) {
+			/*
+			 * When we mount with old space cache, we need to
+			 * set BTRFS_DC_CLEAR and set dirty flag.
+			 *
+			 * a) Setting 'BTRFS_DC_CLEAR' makes sure that we
+			 *    truncate the old free space cache inode and
+			 *    setup a new one.
+			 * b) Setting 'dirty flag' makes sure that we flush
+			 *    the new space cache info onto disk.
+			 */
+			cache->disk_cache_state = BTRFS_DC_CLEAR;
+			if (btrfs_test_opt(root, SPACE_CACHE))
+				cache->dirty = 1;
+		}
 
 		read_extent_buffer(leaf, &cache->item,
 				   btrfs_item_ptr_offset(leaf, path->slots[0]),
 				   sizeof(cache->item));
-		memcpy(&cache->key, &found_key, sizeof(found_key));
+		cache->flags = btrfs_block_group_flags(&cache->item);
 
 		key.objectid = found_key.objectid + found_key.offset;
-		btrfs_release_path(root, path);
-		cache->flags = btrfs_block_group_flags(&cache->item);
-		cache->sectorsize = root->sectorsize;
+		btrfs_release_path(path);
+
+		/*
+		 * We need to exclude the super stripes now so that the space
+		 * info has super bytes accounted for, otherwise we'll think
+		 * we have more space than we actually do.
+		 */
+		ret = exclude_super_stripes(root, cache);
+		if (ret) {
+			/*
+			 * We may have excluded something, so call this just in
+			 * case.
+			 */
+			free_excluded_extents(root, cache);
+			btrfs_put_block_group(cache);
+			goto error;
+		}
 
 		/*
 		 * check for two cases, either we are full, and therefore
@@ -7506,12 +8839,10 @@ int btrfs_read_block_groups(struct btrfs_root *root)
 		 * time, particularly in the full case.
 		 */
 		if (found_key.offset == btrfs_block_group_used(&cache->item)) {
-			exclude_super_stripes(root, cache);
 			cache->last_byte_to_unpin = (u64)-1;
 			cache->cached = BTRFS_CACHE_FINISHED;
 			free_excluded_extents(root, cache);
 		} else if (btrfs_block_group_used(&cache->item) == 0) {
-			exclude_super_stripes(root, cache);
 			cache->last_byte_to_unpin = (u64)-1;
 			cache->cached = BTRFS_CACHE_FINISHED;
 			add_new_free_space(cache, root->fs_info,
@@ -7521,32 +8852,99 @@ int btrfs_read_block_groups(struct btrfs_root *root)
 			free_excluded_extents(root, cache);
 		}
 
+		ret = btrfs_add_block_group_cache(root->fs_info, cache);
+		if (ret) {
+			btrfs_remove_free_space_cache(cache);
+			btrfs_put_block_group(cache);
+			goto error;
+		}
+
 		ret = update_space_info(info, cache->flags, found_key.offset,
 					btrfs_block_group_used(&cache->item),
 					&space_info);
-		BUG_ON(ret);
+		if (ret) {
+			btrfs_remove_free_space_cache(cache);
+			spin_lock(&info->block_group_cache_lock);
+			rb_erase(&cache->cache_node,
+				 &info->block_group_cache_tree);
+			spin_unlock(&info->block_group_cache_lock);
+			btrfs_put_block_group(cache);
+			goto error;
+		}
+
 		cache->space_info = space_info;
 		spin_lock(&cache->space_info->lock);
-		cache->space_info->bytes_super += cache->bytes_super;
+		cache->space_info->bytes_readonly += cache->bytes_super;
 		spin_unlock(&cache->space_info->lock);
 
-		down_write(&space_info->groups_sem);
-		list_add_tail(&cache->list, &space_info->block_groups);
-		up_write(&space_info->groups_sem);
-
-		ret = btrfs_add_block_group_cache(root->fs_info, cache);
-		BUG_ON(ret);
+		__link_block_group(space_info, cache);
 
 		set_avail_alloc_bits(root->fs_info, cache->flags);
 		if (btrfs_chunk_readonly(root, cache->key.objectid))
-			set_block_group_readonly(cache);
+			set_block_group_ro(cache, 1);
 	}
+
+	list_for_each_entry_rcu(space_info, &root->fs_info->space_info, list) {
+		if (!(get_alloc_profile(root, space_info->flags) &
+		      (BTRFS_BLOCK_GROUP_RAID10 |
+		       BTRFS_BLOCK_GROUP_RAID1 |
+		       BTRFS_BLOCK_GROUP_RAID5 |
+		       BTRFS_BLOCK_GROUP_RAID6 |
+		       BTRFS_BLOCK_GROUP_DUP)))
+			continue;
+		/*
+		 * avoid allocating from un-mirrored block group if there are
+		 * mirrored block groups.
+		 */
+		list_for_each_entry(cache,
+				&space_info->block_groups[BTRFS_RAID_RAID0],
+				list)
+			set_block_group_ro(cache, 1);
+		list_for_each_entry(cache,
+				&space_info->block_groups[BTRFS_RAID_SINGLE],
+				list)
+			set_block_group_ro(cache, 1);
+	}
+
+	init_global_block_rsv(info);
 	ret = 0;
 error:
 	btrfs_free_path(path);
 	return ret;
 }
 
+void btrfs_create_pending_block_groups(struct btrfs_trans_handle *trans,
+				       struct btrfs_root *root)
+{
+	struct btrfs_block_group_cache *block_group, *tmp;
+	struct btrfs_root *extent_root = root->fs_info->extent_root;
+	struct btrfs_block_group_item item;
+	struct btrfs_key key;
+	int ret = 0;
+
+	list_for_each_entry_safe(block_group, tmp, &trans->new_bgs,
+				 new_bg_list) {
+		list_del_init(&block_group->new_bg_list);
+
+		if (ret)
+			continue;
+
+		spin_lock(&block_group->lock);
+		memcpy(&item, &block_group->item, sizeof(item));
+		memcpy(&key, &block_group->key, sizeof(key));
+		spin_unlock(&block_group->lock);
+
+		ret = btrfs_insert_item(trans, extent_root, &key, &item,
+					sizeof(item));
+		if (ret)
+			btrfs_abort_transaction(trans, extent_root, ret);
+		ret = btrfs_finish_chunk_alloc(trans, extent_root,
+					       key.objectid, key.offset);
+		if (ret)
+			btrfs_abort_transaction(trans, extent_root, ret);
+	}
+}
+
 int btrfs_make_block_group(struct btrfs_trans_handle *trans,
 			   struct btrfs_root *root, u64 bytes_used,
 			   u64 type, u64 chunk_objectid, u64 chunk_offset,
@@ -7558,76 +8956,96 @@ int btrfs_make_block_group(struct btrfs_trans_handle *trans,
 
 	extent_root = root->fs_info->extent_root;
 
-	root->fs_info->last_trans_log_full_commit = trans->transid;
+	btrfs_set_log_full_commit(root->fs_info, trans);
 
-	cache = kzalloc(sizeof(*cache), GFP_NOFS);
+	cache = btrfs_create_block_group_cache(root, chunk_offset, size);
 	if (!cache)
 		return -ENOMEM;
 
-	cache->key.objectid = chunk_offset;
-	cache->key.offset = size;
-	cache->key.type = BTRFS_BLOCK_GROUP_ITEM_KEY;
-	cache->sectorsize = root->sectorsize;
-
-	/*
-	 * we only want to have 32k of ram per block group for keeping track
-	 * of free space, and if we pass 1/2 of that we want to start
-	 * converting things over to using bitmaps
-	 */
-	cache->extents_thresh = ((1024 * 32) / 2) /
-		sizeof(struct btrfs_free_space);
-	atomic_set(&cache->count, 1);
-	spin_lock_init(&cache->lock);
-	spin_lock_init(&cache->tree_lock);
-	INIT_LIST_HEAD(&cache->list);
-	INIT_LIST_HEAD(&cache->cluster_list);
-
 	btrfs_set_block_group_used(&cache->item, bytes_used);
 	btrfs_set_block_group_chunk_objectid(&cache->item, chunk_objectid);
-	cache->flags = type;
 	btrfs_set_block_group_flags(&cache->item, type);
 
+	cache->flags = type;
 	cache->last_byte_to_unpin = (u64)-1;
 	cache->cached = BTRFS_CACHE_FINISHED;
-	exclude_super_stripes(root, cache);
+	ret = exclude_super_stripes(root, cache);
+	if (ret) {
+		/*
+		 * We may have excluded something, so call this just in
+		 * case.
+		 */
+		free_excluded_extents(root, cache);
+		btrfs_put_block_group(cache);
+		return ret;
+	}
 
 	add_new_free_space(cache, root->fs_info, chunk_offset,
 			   chunk_offset + size);
 
 	free_excluded_extents(root, cache);
 
+	ret = btrfs_add_block_group_cache(root->fs_info, cache);
+	if (ret) {
+		btrfs_remove_free_space_cache(cache);
+		btrfs_put_block_group(cache);
+		return ret;
+	}
+
 	ret = update_space_info(root->fs_info, cache->flags, size, bytes_used,
 				&cache->space_info);
-	BUG_ON(ret);
+	if (ret) {
+		btrfs_remove_free_space_cache(cache);
+		spin_lock(&root->fs_info->block_group_cache_lock);
+		rb_erase(&cache->cache_node,
+			 &root->fs_info->block_group_cache_tree);
+		spin_unlock(&root->fs_info->block_group_cache_lock);
+		btrfs_put_block_group(cache);
+		return ret;
+	}
+	update_global_block_rsv(root->fs_info);
 
 	spin_lock(&cache->space_info->lock);
-	cache->space_info->bytes_super += cache->bytes_super;
+	cache->space_info->bytes_readonly += cache->bytes_super;
 	spin_unlock(&cache->space_info->lock);
 
-	down_write(&cache->space_info->groups_sem);
-	list_add_tail(&cache->list, &cache->space_info->block_groups);
-	up_write(&cache->space_info->groups_sem);
+	__link_block_group(cache->space_info, cache);
 
-	ret = btrfs_add_block_group_cache(root->fs_info, cache);
-	BUG_ON(ret);
-
-	ret = btrfs_insert_item(trans, extent_root, &cache->key, &cache->item,
-				sizeof(cache->item));
-	BUG_ON(ret);
+	list_add_tail(&cache->new_bg_list, &trans->new_bgs);
 
 	set_avail_alloc_bits(extent_root->fs_info, type);
 
 	return 0;
 }
 
+static void clear_avail_alloc_bits(struct btrfs_fs_info *fs_info, u64 flags)
+{
+	u64 extra_flags = chunk_to_extended(flags) &
+				BTRFS_EXTENDED_PROFILE_MASK;
+
+	write_seqlock(&fs_info->profiles_lock);
+	if (flags & BTRFS_BLOCK_GROUP_DATA)
+		fs_info->avail_data_alloc_bits &= ~extra_flags;
+	if (flags & BTRFS_BLOCK_GROUP_METADATA)
+		fs_info->avail_metadata_alloc_bits &= ~extra_flags;
+	if (flags & BTRFS_BLOCK_GROUP_SYSTEM)
+		fs_info->avail_system_alloc_bits &= ~extra_flags;
+	write_sequnlock(&fs_info->profiles_lock);
+}
+
 int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
 			     struct btrfs_root *root, u64 group_start)
 {
 	struct btrfs_path *path;
 	struct btrfs_block_group_cache *block_group;
 	struct btrfs_free_cluster *cluster;
+	struct btrfs_root *tree_root = root->fs_info->tree_root;
 	struct btrfs_key key;
+	struct inode *inode;
+	struct kobject *kobj = NULL;
 	int ret;
+	int index;
+	int factor;
 
 	root = root->fs_info->extent_root;
 
@@ -7635,7 +9053,20 @@ int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
 	BUG_ON(!block_group);
 	BUG_ON(!block_group->ro);
 
+	/*
+	 * Free the reserved super bytes from this block group before
+	 * remove it.
+	 */
+	free_excluded_extents(root, block_group);
+
 	memcpy(&key, &block_group->key, sizeof(key));
+	index = get_block_group_index(block_group);
+	if (block_group->flags & (BTRFS_BLOCK_GROUP_DUP |
+				  BTRFS_BLOCK_GROUP_RAID1 |
+				  BTRFS_BLOCK_GROUP_RAID10))
+		factor = 2;
+	else
+		factor = 1;
 
 	/* make sure this block group isn't part of an allocation cluster */
 	cluster = &root->fs_info->data_alloc_cluster;
@@ -7653,11 +9084,55 @@ int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
 	spin_unlock(&cluster->refill_lock);
 
 	path = btrfs_alloc_path();
-	BUG_ON(!path);
+	if (!path) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	inode = lookup_free_space_inode(tree_root, block_group, path);
+	if (!IS_ERR(inode)) {
+		ret = btrfs_orphan_add(trans, inode);
+		if (ret) {
+			btrfs_add_delayed_iput(inode);
+			goto out;
+		}
+		clear_nlink(inode);
+		/* One for the block groups ref */
+		spin_lock(&block_group->lock);
+		if (block_group->iref) {
+			block_group->iref = 0;
+			block_group->inode = NULL;
+			spin_unlock(&block_group->lock);
+			iput(inode);
+		} else {
+			spin_unlock(&block_group->lock);
+		}
+		/* One for our lookup ref */
+		btrfs_add_delayed_iput(inode);
+	}
+
+	key.objectid = BTRFS_FREE_SPACE_OBJECTID;
+	key.offset = block_group->key.objectid;
+	key.type = 0;
+
+	ret = btrfs_search_slot(trans, tree_root, &key, path, -1, 1);
+	if (ret < 0)
+		goto out;
+	if (ret > 0)
+		btrfs_release_path(path);
+	if (ret == 0) {
+		ret = btrfs_del_item(trans, tree_root, path);
+		if (ret)
+			goto out;
+		btrfs_release_path(path);
+	}
 
 	spin_lock(&root->fs_info->block_group_cache_lock);
 	rb_erase(&block_group->cache_node,
 		 &root->fs_info->block_group_cache_tree);
+
+	if (root->fs_info->first_logical_byte == block_group->key.objectid)
+		root->fs_info->first_logical_byte = (u64)-1;
 	spin_unlock(&root->fs_info->block_group_cache_lock);
 
 	down_write(&block_group->space_info->groups_sem);
@@ -7666,7 +9141,16 @@ int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
 	 * are still on the list after taking the semaphore
 	 */
 	list_del_init(&block_group->list);
+	if (list_empty(&block_group->space_info->block_groups[index])) {
+		kobj = block_group->space_info->block_group_kobjs[index];
+		block_group->space_info->block_group_kobjs[index] = NULL;
+		clear_avail_alloc_bits(root->fs_info, block_group->flags);
+	}
 	up_write(&block_group->space_info->groups_sem);
+	if (kobj) {
+		kobject_del(kobj);
+		kobject_put(kobj);
+	}
 
 	if (block_group->cached == BTRFS_CACHE_STARTED)
 		wait_block_group_cache_done(block_group);
@@ -7676,8 +9160,11 @@ int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
 	spin_lock(&block_group->space_info->lock);
 	block_group->space_info->total_bytes -= block_group->key.offset;
 	block_group->space_info->bytes_readonly -= block_group->key.offset;
+	block_group->space_info->disk_total -= block_group->key.offset * factor;
 	spin_unlock(&block_group->space_info->lock);
 
+	memcpy(&key, &block_group->key, sizeof(key));
+
 	btrfs_clear_space_info_full(root->fs_info);
 
 	btrfs_put_block_group(block_group);
@@ -7694,3 +9181,149 @@ out:
 	btrfs_free_path(path);
 	return ret;
 }
+
+int btrfs_init_space_info(struct btrfs_fs_info *fs_info)
+{
+	struct btrfs_space_info *space_info;
+	struct btrfs_super_block *disk_super;
+	u64 features;
+	u64 flags;
+	int mixed = 0;
+	int ret;
+
+	disk_super = fs_info->super_copy;
+	if (!btrfs_super_root(disk_super))
+		return 1;
+
+	features = btrfs_super_incompat_flags(disk_super);
+	if (features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS)
+		mixed = 1;
+
+	flags = BTRFS_BLOCK_GROUP_SYSTEM;
+	ret = update_space_info(fs_info, flags, 0, 0, &space_info);
+	if (ret)
+		goto out;
+
+	if (mixed) {
+		flags = BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DATA;
+		ret = update_space_info(fs_info, flags, 0, 0, &space_info);
+	} else {
+		flags = BTRFS_BLOCK_GROUP_METADATA;
+		ret = update_space_info(fs_info, flags, 0, 0, &space_info);
+		if (ret)
+			goto out;
+
+		flags = BTRFS_BLOCK_GROUP_DATA;
+		ret = update_space_info(fs_info, flags, 0, 0, &space_info);
+	}
+out:
+	return ret;
+}
+
+int btrfs_error_unpin_extent_range(struct btrfs_root *root, u64 start, u64 end)
+{
+	return unpin_extent_range(root, start, end);
+}
+
+int btrfs_error_discard_extent(struct btrfs_root *root, u64 bytenr,
+			       u64 num_bytes, u64 *actual_bytes)
+{
+	return btrfs_discard_extent(root, bytenr, num_bytes, actual_bytes);
+}
+
+int btrfs_trim_fs(struct btrfs_root *root, struct fstrim_range *range)
+{
+	struct btrfs_fs_info *fs_info = root->fs_info;
+	struct btrfs_block_group_cache *cache = NULL;
+	u64 group_trimmed;
+	u64 start;
+	u64 end;
+	u64 trimmed = 0;
+	u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
+	int ret = 0;
+
+	/*
+	 * try to trim all FS space, our block group may start from non-zero.
+	 */
+	if (range->len == total_bytes)
+		cache = btrfs_lookup_first_block_group(fs_info, range->start);
+	else
+		cache = btrfs_lookup_block_group(fs_info, range->start);
+
+	while (cache) {
+		if (cache->key.objectid >= (range->start + range->len)) {
+			btrfs_put_block_group(cache);
+			break;
+		}
+
+		start = max(range->start, cache->key.objectid);
+		end = min(range->start + range->len,
+				cache->key.objectid + cache->key.offset);
+
+		if (end - start >= range->minlen) {
+			if (!block_group_cache_done(cache)) {
+				ret = cache_block_group(cache, 0);
+				if (ret) {
+					btrfs_put_block_group(cache);
+					break;
+				}
+				ret = wait_block_group_cache_done(cache);
+				if (ret) {
+					btrfs_put_block_group(cache);
+					break;
+				}
+			}
+			ret = btrfs_trim_block_group(cache,
+						     &group_trimmed,
+						     start,
+						     end,
+						     range->minlen);
+
+			trimmed += group_trimmed;
+			if (ret) {
+				btrfs_put_block_group(cache);
+				break;
+			}
+		}
+
+		cache = next_block_group(fs_info->tree_root, cache);
+	}
+
+	range->len = trimmed;
+	return ret;
+}
+
+/*
+ * btrfs_{start,end}_write() is similar to mnt_{want, drop}_write(),
+ * they are used to prevent the some tasks writing data into the page cache
+ * by nocow before the subvolume is snapshoted, but flush the data into
+ * the disk after the snapshot creation.
+ */
+void btrfs_end_nocow_write(struct btrfs_root *root)
+{
+	percpu_counter_dec(&root->subv_writers->counter);
+	/*
+	 * Make sure counter is updated before we wake up
+	 * waiters.
+	 */
+	smp_mb();
+	if (waitqueue_active(&root->subv_writers->wait))
+		wake_up(&root->subv_writers->wait);
+}
+
+int btrfs_start_nocow_write(struct btrfs_root *root)
+{
+	if (unlikely(atomic_read(&root->will_be_snapshoted)))
+		return 0;
+
+	percpu_counter_inc(&root->subv_writers->counter);
+	/*
+	 * Make sure counter is updated before we check for snapshot creation.
+	 */
+	smp_mb();
+	if (unlikely(atomic_read(&root->will_be_snapshoted))) {
+		btrfs_end_nocow_write(root);
+		return 0;
+	}
+	return 1;
+}
diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
index d2d03684fab..a389820d158 100644
--- a/fs/btrfs/extent_io.c
+++ b/fs/btrfs/extent_io.c
@@ -4,27 +4,103 @@
 #include <linux/mm.h>
 #include <linux/pagemap.h>
 #include <linux/page-flags.h>
-#include <linux/module.h>
 #include <linux/spinlock.h>
 #include <linux/blkdev.h>
 #include <linux/swap.h>
 #include <linux/writeback.h>
 #include <linux/pagevec.h>
+#include <linux/prefetch.h>
+#include <linux/cleancache.h>
 #include "extent_io.h"
 #include "extent_map.h"
-#include "compat.h"
 #include "ctree.h"
 #include "btrfs_inode.h"
+#include "volumes.h"
+#include "check-integrity.h"
+#include "locking.h"
+#include "rcu-string.h"
+#include "backref.h"
 
 static struct kmem_cache *extent_state_cache;
 static struct kmem_cache *extent_buffer_cache;
+static struct bio_set *btrfs_bioset;
 
+#ifdef CONFIG_BTRFS_DEBUG
 static LIST_HEAD(buffers);
 static LIST_HEAD(states);
 
-#define LEAK_DEBUG 0
-#if LEAK_DEBUG
 static DEFINE_SPINLOCK(leak_lock);
+
+static inline
+void btrfs_leak_debug_add(struct list_head *new, struct list_head *head)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&leak_lock, flags);
+	list_add(new, head);
+	spin_unlock_irqrestore(&leak_lock, flags);
+}
+
+static inline
+void btrfs_leak_debug_del(struct list_head *entry)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&leak_lock, flags);
+	list_del(entry);
+	spin_unlock_irqrestore(&leak_lock, flags);
+}
+
+static inline
+void btrfs_leak_debug_check(void)
+{
+	struct extent_state *state;
+	struct extent_buffer *eb;
+
+	while (!list_empty(&states)) {
+		state = list_entry(states.next, struct extent_state, leak_list);
+		printk(KERN_ERR "BTRFS: state leak: start %llu end %llu "
+		       "state %lu in tree %p refs %d\n",
+		       state->start, state->end, state->state, state->tree,
+		       atomic_read(&state->refs));
+		list_del(&state->leak_list);
+		kmem_cache_free(extent_state_cache, state);
+	}
+
+	while (!list_empty(&buffers)) {
+		eb = list_entry(buffers.next, struct extent_buffer, leak_list);
+		printk(KERN_ERR "BTRFS: buffer leak start %llu len %lu "
+		       "refs %d\n",
+		       eb->start, eb->len, atomic_read(&eb->refs));
+		list_del(&eb->leak_list);
+		kmem_cache_free(extent_buffer_cache, eb);
+	}
+}
+
+#define btrfs_debug_check_extent_io_range(tree, start, end)		\
+	__btrfs_debug_check_extent_io_range(__func__, (tree), (start), (end))
+static inline void __btrfs_debug_check_extent_io_range(const char *caller,
+		struct extent_io_tree *tree, u64 start, u64 end)
+{
+	struct inode *inode;
+	u64 isize;
+
+	if (!tree->mapping)
+		return;
+
+	inode = tree->mapping->host;
+	isize = i_size_read(inode);
+	if (end >= PAGE_SIZE && (end % 2) == 0 && end != isize - 1) {
+		printk_ratelimited(KERN_DEBUG
+		    "BTRFS: %s: ino %llu isize %llu odd range [%llu,%llu]\n",
+				caller, btrfs_ino(inode), isize, start, end);
+	}
+}
+#else
+#define btrfs_leak_debug_add(new, head)	do {} while (0)
+#define btrfs_leak_debug_del(entry)	do {} while (0)
+#define btrfs_leak_debug_check()	do {} while (0)
+#define btrfs_debug_check_extent_io_range(c, s, e)	do {} while (0)
 #endif
 
 #define BUFFER_LRU_MAX 64
@@ -39,6 +115,7 @@ struct extent_page_data {
 	struct bio *bio;
 	struct extent_io_tree *tree;
 	get_extent_t *get_extent;
+	unsigned long bio_flags;
 
 	/* tells writepage not to lock the state bits for this range
 	 * it still does the unlocking
@@ -49,75 +126,83 @@ struct extent_page_data {
 	unsigned int sync_io:1;
 };
 
+static noinline void flush_write_bio(void *data);
+static inline struct btrfs_fs_info *
+tree_fs_info(struct extent_io_tree *tree)
+{
+	if (!tree->mapping)
+		return NULL;
+	return btrfs_sb(tree->mapping->host->i_sb);
+}
+
 int __init extent_io_init(void)
 {
-	extent_state_cache = kmem_cache_create("extent_state",
+	extent_state_cache = kmem_cache_create("btrfs_extent_state",
 			sizeof(struct extent_state), 0,
 			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
 	if (!extent_state_cache)
 		return -ENOMEM;
 
-	extent_buffer_cache = kmem_cache_create("extent_buffers",
+	extent_buffer_cache = kmem_cache_create("btrfs_extent_buffer",
 			sizeof(struct extent_buffer), 0,
 			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
 	if (!extent_buffer_cache)
 		goto free_state_cache;
+
+	btrfs_bioset = bioset_create(BIO_POOL_SIZE,
+				     offsetof(struct btrfs_io_bio, bio));
+	if (!btrfs_bioset)
+		goto free_buffer_cache;
+
+	if (bioset_integrity_create(btrfs_bioset, BIO_POOL_SIZE))
+		goto free_bioset;
+
 	return 0;
 
+free_bioset:
+	bioset_free(btrfs_bioset);
+	btrfs_bioset = NULL;
+
+free_buffer_cache:
+	kmem_cache_destroy(extent_buffer_cache);
+	extent_buffer_cache = NULL;
+
 free_state_cache:
 	kmem_cache_destroy(extent_state_cache);
+	extent_state_cache = NULL;
 	return -ENOMEM;
 }
 
 void extent_io_exit(void)
 {
-	struct extent_state *state;
-	struct extent_buffer *eb;
+	btrfs_leak_debug_check();
 
-	while (!list_empty(&states)) {
-		state = list_entry(states.next, struct extent_state, leak_list);
-		printk(KERN_ERR "btrfs state leak: start %llu end %llu "
-		       "state %lu in tree %p refs %d\n",
-		       (unsigned long long)state->start,
-		       (unsigned long long)state->end,
-		       state->state, state->tree, atomic_read(&state->refs));
-		list_del(&state->leak_list);
-		kmem_cache_free(extent_state_cache, state);
-
-	}
-
-	while (!list_empty(&buffers)) {
-		eb = list_entry(buffers.next, struct extent_buffer, leak_list);
-		printk(KERN_ERR "btrfs buffer leak start %llu len %lu "
-		       "refs %d\n", (unsigned long long)eb->start,
-		       eb->len, atomic_read(&eb->refs));
-		list_del(&eb->leak_list);
-		kmem_cache_free(extent_buffer_cache, eb);
-	}
+	/*
+	 * Make sure all delayed rcu free are flushed before we
+	 * destroy caches.
+	 */
+	rcu_barrier();
 	if (extent_state_cache)
 		kmem_cache_destroy(extent_state_cache);
 	if (extent_buffer_cache)
 		kmem_cache_destroy(extent_buffer_cache);
+	if (btrfs_bioset)
+		bioset_free(btrfs_bioset);
 }
 
 void extent_io_tree_init(struct extent_io_tree *tree,
-			  struct address_space *mapping, gfp_t mask)
+			 struct address_space *mapping)
 {
 	tree->state = RB_ROOT;
-	tree->buffer = RB_ROOT;
 	tree->ops = NULL;
 	tree->dirty_bytes = 0;
 	spin_lock_init(&tree->lock);
-	spin_lock_init(&tree->buffer_lock);
 	tree->mapping = mapping;
 }
 
 static struct extent_state *alloc_extent_state(gfp_t mask)
 {
 	struct extent_state *state;
-#if LEAK_DEBUG
-	unsigned long flags;
-#endif
 
 	state = kmem_cache_alloc(extent_state_cache, mask);
 	if (!state)
@@ -125,41 +210,43 @@ static struct extent_state *alloc_extent_state(gfp_t mask)
 	state->state = 0;
 	state->private = 0;
 	state->tree = NULL;
-#if LEAK_DEBUG
-	spin_lock_irqsave(&leak_lock, flags);
-	list_add(&state->leak_list, &states);
-	spin_unlock_irqrestore(&leak_lock, flags);
-#endif
+	btrfs_leak_debug_add(&state->leak_list, &states);
 	atomic_set(&state->refs, 1);
 	init_waitqueue_head(&state->wq);
+	trace_alloc_extent_state(state, mask, _RET_IP_);
 	return state;
 }
 
-static void free_extent_state(struct extent_state *state)
+void free_extent_state(struct extent_state *state)
 {
 	if (!state)
 		return;
 	if (atomic_dec_and_test(&state->refs)) {
-#if LEAK_DEBUG
-		unsigned long flags;
-#endif
 		WARN_ON(state->tree);
-#if LEAK_DEBUG
-		spin_lock_irqsave(&leak_lock, flags);
-		list_del(&state->leak_list);
-		spin_unlock_irqrestore(&leak_lock, flags);
-#endif
+		btrfs_leak_debug_del(&state->leak_list);
+		trace_free_extent_state(state, _RET_IP_);
 		kmem_cache_free(extent_state_cache, state);
 	}
 }
 
-static struct rb_node *tree_insert(struct rb_root *root, u64 offset,
-				   struct rb_node *node)
+static struct rb_node *tree_insert(struct rb_root *root,
+				   struct rb_node *search_start,
+				   u64 offset,
+				   struct rb_node *node,
+				   struct rb_node ***p_in,
+				   struct rb_node **parent_in)
 {
-	struct rb_node **p = &root->rb_node;
+	struct rb_node **p;
 	struct rb_node *parent = NULL;
 	struct tree_entry *entry;
 
+	if (p_in && parent_in) {
+		p = *p_in;
+		parent = *parent_in;
+		goto do_insert;
+	}
+
+	p = search_start ? &search_start : &root->rb_node;
 	while (*p) {
 		parent = *p;
 		entry = rb_entry(parent, struct tree_entry, rb_node);
@@ -172,36 +259,43 @@ static struct rb_node *tree_insert(struct rb_root *root, u64 offset,
 			return parent;
 	}
 
-	entry = rb_entry(node, struct tree_entry, rb_node);
+do_insert:
 	rb_link_node(node, parent, p);
 	rb_insert_color(node, root);
 	return NULL;
 }
 
 static struct rb_node *__etree_search(struct extent_io_tree *tree, u64 offset,
-				     struct rb_node **prev_ret,
-				     struct rb_node **next_ret)
+				      struct rb_node **prev_ret,
+				      struct rb_node **next_ret,
+				      struct rb_node ***p_ret,
+				      struct rb_node **parent_ret)
 {
 	struct rb_root *root = &tree->state;
-	struct rb_node *n = root->rb_node;
+	struct rb_node **n = &root->rb_node;
 	struct rb_node *prev = NULL;
 	struct rb_node *orig_prev = NULL;
 	struct tree_entry *entry;
 	struct tree_entry *prev_entry = NULL;
 
-	while (n) {
-		entry = rb_entry(n, struct tree_entry, rb_node);
-		prev = n;
+	while (*n) {
+		prev = *n;
+		entry = rb_entry(prev, struct tree_entry, rb_node);
 		prev_entry = entry;
 
 		if (offset < entry->start)
-			n = n->rb_left;
+			n = &(*n)->rb_left;
 		else if (offset > entry->end)
-			n = n->rb_right;
+			n = &(*n)->rb_right;
 		else
-			return n;
+			return *n;
 	}
 
+	if (p_ret)
+		*p_ret = n;
+	if (parent_ret)
+		*parent_ret = prev;
+
 	if (prev_ret) {
 		orig_prev = prev;
 		while (prev && offset > prev_entry->end) {
@@ -223,60 +317,25 @@ static struct rb_node *__etree_search(struct extent_io_tree *tree, u64 offset,
 	return NULL;
 }
 
-static inline struct rb_node *tree_search(struct extent_io_tree *tree,
-					  u64 offset)
+static inline struct rb_node *
+tree_search_for_insert(struct extent_io_tree *tree,
+		       u64 offset,
+		       struct rb_node ***p_ret,
+		       struct rb_node **parent_ret)
 {
 	struct rb_node *prev = NULL;
 	struct rb_node *ret;
 
-	ret = __etree_search(tree, offset, &prev, NULL);
+	ret = __etree_search(tree, offset, &prev, NULL, p_ret, parent_ret);
 	if (!ret)
 		return prev;
 	return ret;
 }
 
-static struct extent_buffer *buffer_tree_insert(struct extent_io_tree *tree,
-					  u64 offset, struct rb_node *node)
+static inline struct rb_node *tree_search(struct extent_io_tree *tree,
+					  u64 offset)
 {
-	struct rb_root *root = &tree->buffer;
-	struct rb_node **p = &root->rb_node;
-	struct rb_node *parent = NULL;
-	struct extent_buffer *eb;
-
-	while (*p) {
-		parent = *p;
-		eb = rb_entry(parent, struct extent_buffer, rb_node);
-
-		if (offset < eb->start)
-			p = &(*p)->rb_left;
-		else if (offset > eb->start)
-			p = &(*p)->rb_right;
-		else
-			return eb;
-	}
-
-	rb_link_node(node, parent, p);
-	rb_insert_color(node, root);
-	return NULL;
-}
-
-static struct extent_buffer *buffer_search(struct extent_io_tree *tree,
-					   u64 offset)
-{
-	struct rb_root *root = &tree->buffer;
-	struct rb_node *n = root->rb_node;
-	struct extent_buffer *eb;
-
-	while (n) {
-		eb = rb_entry(n, struct extent_buffer, rb_node);
-		if (offset < eb->start)
-			n = n->rb_left;
-		else if (offset > eb->start)
-			n = n->rb_right;
-		else
-			return eb;
-	}
-	return NULL;
+	return tree_search_for_insert(tree, offset, NULL, NULL);
 }
 
 static void merge_cb(struct extent_io_tree *tree, struct extent_state *new,
@@ -296,14 +355,14 @@ static void merge_cb(struct extent_io_tree *tree, struct extent_state *new,
  *
  * This should be called with the tree lock held.
  */
-static int merge_state(struct extent_io_tree *tree,
-		       struct extent_state *state)
+static void merge_state(struct extent_io_tree *tree,
+		        struct extent_state *state)
 {
 	struct extent_state *other;
 	struct rb_node *other_node;
 
 	if (state->state & (EXTENT_IOBITS | EXTENT_BOUNDARY))
-		return 0;
+		return;
 
 	other_node = rb_prev(&state->rb_node);
 	if (other_node) {
@@ -323,38 +382,31 @@ static int merge_state(struct extent_io_tree *tree,
 		if (other->start == state->end + 1 &&
 		    other->state == state->state) {
 			merge_cb(tree, state, other);
-			other->start = state->start;
-			state->tree = NULL;
-			rb_erase(&state->rb_node, &tree->state);
-			free_extent_state(state);
-			state = NULL;
+			state->end = other->end;
+			other->tree = NULL;
+			rb_erase(&other->rb_node, &tree->state);
+			free_extent_state(other);
 		}
 	}
-
-	return 0;
 }
 
-static int set_state_cb(struct extent_io_tree *tree,
-			 struct extent_state *state,
-			 unsigned long bits)
+static void set_state_cb(struct extent_io_tree *tree,
+			 struct extent_state *state, unsigned long *bits)
 {
-	if (tree->ops && tree->ops->set_bit_hook) {
-		return tree->ops->set_bit_hook(tree->mapping->host,
-					       state->start, state->end,
-					       state->state, bits);
-	}
-
-	return 0;
+	if (tree->ops && tree->ops->set_bit_hook)
+		tree->ops->set_bit_hook(tree->mapping->host, state, bits);
 }
 
 static void clear_state_cb(struct extent_io_tree *tree,
-			   struct extent_state *state,
-			   unsigned long bits)
+			   struct extent_state *state, unsigned long *bits)
 {
 	if (tree->ops && tree->ops->clear_bit_hook)
 		tree->ops->clear_bit_hook(tree->mapping->host, state, bits);
 }
 
+static void set_state_bits(struct extent_io_tree *tree,
+			   struct extent_state *state, unsigned long *bits);
+
 /*
  * insert an extent_state struct into the tree.  'bits' are set on the
  * struct before it is inserted.
@@ -367,35 +419,27 @@ static void clear_state_cb(struct extent_io_tree *tree,
  */
 static int insert_state(struct extent_io_tree *tree,
 			struct extent_state *state, u64 start, u64 end,
-			int bits)
+			struct rb_node ***p,
+			struct rb_node **parent,
+			unsigned long *bits)
 {
 	struct rb_node *node;
-	int ret;
 
-	if (end < start) {
-		printk(KERN_ERR "btrfs end < start %llu %llu\n",
-		       (unsigned long long)end,
-		       (unsigned long long)start);
-		WARN_ON(1);
-	}
+	if (end < start)
+		WARN(1, KERN_ERR "BTRFS: end < start %llu %llu\n",
+		       end, start);
 	state->start = start;
 	state->end = end;
-	ret = set_state_cb(tree, state, bits);
-	if (ret)
-		return ret;
 
-	if (bits & EXTENT_DIRTY)
-		tree->dirty_bytes += end - start + 1;
-	state->state |= bits;
-	node = tree_insert(&tree->state, end, &state->rb_node);
+	set_state_bits(tree, state, bits);
+
+	node = tree_insert(&tree->state, NULL, end, &state->rb_node, p, parent);
 	if (node) {
 		struct extent_state *found;
 		found = rb_entry(node, struct extent_state, rb_node);
-		printk(KERN_ERR "btrfs found node %llu %llu on insert of "
-		       "%llu %llu\n", (unsigned long long)found->start,
-		       (unsigned long long)found->end,
-		       (unsigned long long)start, (unsigned long long)end);
-		free_extent_state(state);
+		printk(KERN_ERR "BTRFS: found node %llu %llu on insert of "
+		       "%llu %llu\n",
+		       found->start, found->end, start, end);
 		return -EEXIST;
 	}
 	state->tree = tree;
@@ -403,13 +447,11 @@ static int insert_state(struct extent_io_tree *tree,
 	return 0;
 }
 
-static int split_cb(struct extent_io_tree *tree, struct extent_state *orig,
+static void split_cb(struct extent_io_tree *tree, struct extent_state *orig,
 		     u64 split)
 {
 	if (tree->ops && tree->ops->split_extent_hook)
-		return tree->ops->split_extent_hook(tree->mapping->host,
-						    orig, split);
-	return 0;
+		tree->ops->split_extent_hook(tree->mapping->host, orig, split);
 }
 
 /*
@@ -438,7 +480,8 @@ static int split_state(struct extent_io_tree *tree, struct extent_state *orig,
 	prealloc->state = orig->state;
 	orig->start = split;
 
-	node = tree_insert(&tree->state, prealloc->end, &prealloc->rb_node);
+	node = tree_insert(&tree->state, &orig->rb_node, prealloc->end,
+			   &prealloc->rb_node, NULL, NULL);
 	if (node) {
 		free_extent_state(prealloc);
 		return -EEXIST;
@@ -447,22 +490,30 @@ static int split_state(struct extent_io_tree *tree, struct extent_state *orig,
 	return 0;
 }
 
+static struct extent_state *next_state(struct extent_state *state)
+{
+	struct rb_node *next = rb_next(&state->rb_node);
+	if (next)
+		return rb_entry(next, struct extent_state, rb_node);
+	else
+		return NULL;
+}
+
 /*
  * utility function to clear some bits in an extent state struct.
- * it will optionally wake up any one waiting on this state (wake == 1), or
- * forcibly remove the state from the tree (delete == 1).
+ * it will optionally wake up any one waiting on this state (wake == 1).
  *
  * If no bits are set on the state struct after clearing things, the
  * struct is freed and removed from the tree
  */
-static int clear_state_bit(struct extent_io_tree *tree,
-			    struct extent_state *state, int bits, int wake,
-			    int delete)
+static struct extent_state *clear_state_bit(struct extent_io_tree *tree,
+					    struct extent_state *state,
+					    unsigned long *bits, int wake)
 {
-	int bits_to_clear = bits & ~EXTENT_DO_ACCOUNTING;
-	int ret = state->state & bits_to_clear;
+	struct extent_state *next;
+	unsigned long bits_to_clear = *bits & ~EXTENT_CTLBITS;
 
-	if ((bits & EXTENT_DIRTY) && (state->state & EXTENT_DIRTY)) {
+	if ((bits_to_clear & EXTENT_DIRTY) && (state->state & EXTENT_DIRTY)) {
 		u64 range = state->end - state->start + 1;
 		WARN_ON(range > tree->dirty_bytes);
 		tree->dirty_bytes -= range;
@@ -471,9 +522,9 @@ static int clear_state_bit(struct extent_io_tree *tree,
 	state->state &= ~bits_to_clear;
 	if (wake)
 		wake_up(&state->wq);
-	if (delete || state->state == 0) {
+	if (state->state == 0) {
+		next = next_state(state);
 		if (state->tree) {
-			clear_state_cb(tree, state, state->state);
 			rb_erase(&state->rb_node, &tree->state);
 			state->tree = NULL;
 			free_extent_state(state);
@@ -482,8 +533,25 @@ static int clear_state_bit(struct extent_io_tree *tree,
 		}
 	} else {
 		merge_state(tree, state);
+		next = next_state(state);
 	}
-	return ret;
+	return next;
+}
+
+static struct extent_state *
+alloc_extent_state_atomic(struct extent_state *prealloc)
+{
+	if (!prealloc)
+		prealloc = alloc_extent_state(GFP_ATOMIC);
+
+	return prealloc;
+}
+
+static void extent_io_tree_panic(struct extent_io_tree *tree, int err)
+{
+	btrfs_panic(tree_fs_info(tree), err, "Locking error: "
+		    "Extent tree was modified by another "
+		    "thread while locked.");
 }
 
 /*
@@ -496,24 +564,30 @@ static int clear_state_bit(struct extent_io_tree *tree,
  *
  * the range [start, end] is inclusive.
  *
- * This takes the tree lock, and returns < 0 on error, > 0 if any of the
- * bits were already set, or zero if none of the bits were already set.
+ * This takes the tree lock, and returns 0 on success and < 0 on error.
  */
 int clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
-		     int bits, int wake, int delete,
+		     unsigned long bits, int wake, int delete,
 		     struct extent_state **cached_state,
 		     gfp_t mask)
 {
 	struct extent_state *state;
 	struct extent_state *cached;
 	struct extent_state *prealloc = NULL;
-	struct rb_node *next_node;
 	struct rb_node *node;
 	u64 last_end;
 	int err;
-	int set = 0;
 	int clear = 0;
 
+	btrfs_debug_check_extent_io_range(tree, start, end);
+
+	if (bits & EXTENT_DELALLOC)
+		bits |= EXTENT_NORESERVE;
+
+	if (delete)
+		bits |= ~EXTENT_CTLBITS;
+	bits |= EXTENT_FIRST_DELALLOC;
+
 	if (bits & (EXTENT_IOBITS | EXTENT_BOUNDARY))
 		clear = 1;
 again:
@@ -532,7 +606,8 @@ again:
 			cached_state = NULL;
 		}
 
-		if (cached && cached->tree && cached->start == start) {
+		if (cached && cached->tree && cached->start <= start &&
+		    cached->end > start) {
 			if (clear)
 				atomic_dec(&cached->refs);
 			state = cached;
@@ -555,6 +630,12 @@ hit_next:
 	WARN_ON(state->end < start);
 	last_end = state->end;
 
+	/* the state doesn't have the wanted bits, go ahead */
+	if (!(state->state & bits)) {
+		state = next_state(state);
+		goto next;
+	}
+
 	/*
 	 *     | ---- desired range ---- |
 	 *  | state | or
@@ -572,19 +653,18 @@ hit_next:
 	 */
 
 	if (state->start < start) {
-		if (!prealloc)
-			prealloc = alloc_extent_state(GFP_ATOMIC);
+		prealloc = alloc_extent_state_atomic(prealloc);
+		BUG_ON(!prealloc);
 		err = split_state(tree, state, prealloc, start);
-		BUG_ON(err == -EEXIST);
+		if (err)
+			extent_io_tree_panic(tree, err);
+
 		prealloc = NULL;
 		if (err)
 			goto out;
 		if (state->end <= end) {
-			set |= clear_state_bit(tree, state, bits, wake,
-					       delete);
-			if (last_end == (u64)-1)
-				goto out;
-			start = last_end + 1;
+			state = clear_state_bit(tree, state, &bits, wake);
+			goto next;
 		}
 		goto search_again;
 	}
@@ -595,34 +675,28 @@ hit_next:
 	 * on the first half
 	 */
 	if (state->start <= end && state->end > end) {
-		if (!prealloc)
-			prealloc = alloc_extent_state(GFP_ATOMIC);
+		prealloc = alloc_extent_state_atomic(prealloc);
+		BUG_ON(!prealloc);
 		err = split_state(tree, state, prealloc, end + 1);
-		BUG_ON(err == -EEXIST);
+		if (err)
+			extent_io_tree_panic(tree, err);
+
 		if (wake)
 			wake_up(&state->wq);
 
-		set |= clear_state_bit(tree, prealloc, bits, wake, delete);
+		clear_state_bit(tree, prealloc, &bits, wake);
 
 		prealloc = NULL;
 		goto out;
 	}
 
-	if (state->end < end && prealloc && !need_resched())
-		next_node = rb_next(&state->rb_node);
-	else
-		next_node = NULL;
-
-	set |= clear_state_bit(tree, state, bits, wake, delete);
+	state = clear_state_bit(tree, state, &bits, wake);
+next:
 	if (last_end == (u64)-1)
 		goto out;
 	start = last_end + 1;
-	if (start <= end && next_node) {
-		state = rb_entry(next_node, struct extent_state,
-				 rb_node);
-		if (state->start == start)
-			goto hit_next;
-	}
+	if (start <= end && state && !need_resched())
+		goto hit_next;
 	goto search_again;
 
 out:
@@ -630,7 +704,7 @@ out:
 	if (prealloc)
 		free_extent_state(prealloc);
 
-	return set;
+	return 0;
 
 search_again:
 	if (start > end)
@@ -641,8 +715,8 @@ search_again:
 	goto again;
 }
 
-static int wait_on_state(struct extent_io_tree *tree,
-			 struct extent_state *state)
+static void wait_on_state(struct extent_io_tree *tree,
+			  struct extent_state *state)
 		__releases(tree->lock)
 		__acquires(tree->lock)
 {
@@ -652,7 +726,6 @@ static int wait_on_state(struct extent_io_tree *tree,
 	schedule();
 	spin_lock(&tree->lock);
 	finish_wait(&state->wq, &wait);
-	return 0;
 }
 
 /*
@@ -660,11 +733,14 @@ static int wait_on_state(struct extent_io_tree *tree,
  * The range [start, end] is inclusive.
  * The tree lock is taken by this function
  */
-int wait_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, int bits)
+static void wait_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
+			    unsigned long bits)
 {
 	struct extent_state *state;
 	struct rb_node *node;
 
+	btrfs_debug_check_extent_io_range(tree, start, end);
+
 	spin_lock(&tree->lock);
 again:
 	while (1) {
@@ -673,6 +749,7 @@ again:
 		 * our range starts
 		 */
 		node = tree_search(tree, start);
+process_node:
 		if (!node)
 			break;
 
@@ -693,34 +770,27 @@ again:
 		if (start > end)
 			break;
 
-		if (need_resched()) {
-			spin_unlock(&tree->lock);
-			cond_resched();
-			spin_lock(&tree->lock);
+		if (!cond_resched_lock(&tree->lock)) {
+			node = rb_next(node);
+			goto process_node;
 		}
 	}
 out:
 	spin_unlock(&tree->lock);
-	return 0;
 }
 
-static int set_state_bits(struct extent_io_tree *tree,
+static void set_state_bits(struct extent_io_tree *tree,
 			   struct extent_state *state,
-			   int bits)
+			   unsigned long *bits)
 {
-	int ret;
-
-	ret = set_state_cb(tree, state, bits);
-	if (ret)
-		return ret;
+	unsigned long bits_to_set = *bits & ~EXTENT_CTLBITS;
 
-	if ((bits & EXTENT_DIRTY) && !(state->state & EXTENT_DIRTY)) {
+	set_state_cb(tree, state, bits);
+	if ((bits_to_set & EXTENT_DIRTY) && !(state->state & EXTENT_DIRTY)) {
 		u64 range = state->end - state->start + 1;
 		tree->dirty_bytes += range;
 	}
-	state->state |= bits;
-
-	return 0;
+	state->state |= bits_to_set;
 }
 
 static void cache_state(struct extent_state *state,
@@ -745,29 +815,35 @@ static void cache_state(struct extent_state *state,
  * [start, end] is inclusive This takes the tree lock.
  */
 
-static int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
-			  int bits, int exclusive_bits, u64 *failed_start,
-			  struct extent_state **cached_state,
-			  gfp_t mask)
+static int __must_check
+__set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
+		 unsigned long bits, unsigned long exclusive_bits,
+		 u64 *failed_start, struct extent_state **cached_state,
+		 gfp_t mask)
 {
 	struct extent_state *state;
 	struct extent_state *prealloc = NULL;
 	struct rb_node *node;
+	struct rb_node **p;
+	struct rb_node *parent;
 	int err = 0;
 	u64 last_start;
 	u64 last_end;
 
+	btrfs_debug_check_extent_io_range(tree, start, end);
+
+	bits |= EXTENT_FIRST_DELALLOC;
 again:
 	if (!prealloc && (mask & __GFP_WAIT)) {
 		prealloc = alloc_extent_state(mask);
-		if (!prealloc)
-			return -ENOMEM;
+		BUG_ON(!prealloc);
 	}
 
 	spin_lock(&tree->lock);
 	if (cached_state && *cached_state) {
 		state = *cached_state;
-		if (state->start == start && state->tree) {
+		if (state->start <= start && state->end > start &&
+		    state->tree) {
 			node = &state->rb_node;
 			goto hit_next;
 		}
@@ -776,11 +852,17 @@ again:
 	 * this search will find all the extents that end after
 	 * our range starts.
 	 */
-	node = tree_search(tree, start);
+	node = tree_search_for_insert(tree, start, &p, &parent);
 	if (!node) {
-		err = insert_state(tree, prealloc, start, end, bits);
+		prealloc = alloc_extent_state_atomic(prealloc);
+		BUG_ON(!prealloc);
+		err = insert_state(tree, prealloc, start, end,
+				   &p, &parent, &bits);
+		if (err)
+			extent_io_tree_panic(tree, err);
+
+		cache_state(prealloc, cached_state);
 		prealloc = NULL;
-		BUG_ON(err == -EEXIST);
 		goto out;
 	}
 	state = rb_entry(node, struct extent_state, rb_node);
@@ -795,32 +877,22 @@ hit_next:
 	 * Just lock what we found and keep going
 	 */
 	if (state->start == start && state->end <= end) {
-		struct rb_node *next_node;
 		if (state->state & exclusive_bits) {
 			*failed_start = state->start;
 			err = -EEXIST;
 			goto out;
 		}
 
-		err = set_state_bits(tree, state, bits);
-		if (err)
-			goto out;
-
+		set_state_bits(tree, state, &bits);
 		cache_state(state, cached_state);
 		merge_state(tree, state);
 		if (last_end == (u64)-1)
 			goto out;
-
 		start = last_end + 1;
-		if (start < end && prealloc && !need_resched()) {
-			next_node = rb_next(node);
-			if (next_node) {
-				state = rb_entry(next_node, struct extent_state,
-						 rb_node);
-				if (state->start == start)
-					goto hit_next;
-			}
-		}
+		state = next_state(state);
+		if (start < end && state && state->start == start &&
+		    !need_resched())
+			goto hit_next;
 		goto search_again;
 	}
 
@@ -846,20 +918,27 @@ hit_next:
 			err = -EEXIST;
 			goto out;
 		}
+
+		prealloc = alloc_extent_state_atomic(prealloc);
+		BUG_ON(!prealloc);
 		err = split_state(tree, state, prealloc, start);
-		BUG_ON(err == -EEXIST);
+		if (err)
+			extent_io_tree_panic(tree, err);
+
 		prealloc = NULL;
 		if (err)
 			goto out;
 		if (state->end <= end) {
-			err = set_state_bits(tree, state, bits);
-			if (err)
-				goto out;
+			set_state_bits(tree, state, &bits);
 			cache_state(state, cached_state);
 			merge_state(tree, state);
 			if (last_end == (u64)-1)
 				goto out;
 			start = last_end + 1;
+			state = next_state(state);
+			if (start < end && state && state->start == start &&
+			    !need_resched())
+				goto hit_next;
 		}
 		goto search_again;
 	}
@@ -876,13 +955,19 @@ hit_next:
 			this_end = end;
 		else
 			this_end = last_start - 1;
+
+		prealloc = alloc_extent_state_atomic(prealloc);
+		BUG_ON(!prealloc);
+
+		/*
+		 * Avoid to free 'prealloc' if it can be merged with
+		 * the later extent.
+		 */
 		err = insert_state(tree, prealloc, start, this_end,
-				   bits);
-		BUG_ON(err == -EEXIST);
-		if (err) {
-			prealloc = NULL;
-			goto out;
-		}
+				   NULL, NULL, &bits);
+		if (err)
+			extent_io_tree_panic(tree, err);
+
 		cache_state(prealloc, cached_state);
 		prealloc = NULL;
 		start = this_end + 1;
@@ -900,16 +985,233 @@ hit_next:
 			err = -EEXIST;
 			goto out;
 		}
+
+		prealloc = alloc_extent_state_atomic(prealloc);
+		BUG_ON(!prealloc);
 		err = split_state(tree, state, prealloc, end + 1);
-		BUG_ON(err == -EEXIST);
+		if (err)
+			extent_io_tree_panic(tree, err);
 
-		err = set_state_bits(tree, prealloc, bits);
-		if (err) {
-			prealloc = NULL;
+		set_state_bits(tree, prealloc, &bits);
+		cache_state(prealloc, cached_state);
+		merge_state(tree, prealloc);
+		prealloc = NULL;
+		goto out;
+	}
+
+	goto search_again;
+
+out:
+	spin_unlock(&tree->lock);
+	if (prealloc)
+		free_extent_state(prealloc);
+
+	return err;
+
+search_again:
+	if (start > end)
+		goto out;
+	spin_unlock(&tree->lock);
+	if (mask & __GFP_WAIT)
+		cond_resched();
+	goto again;
+}
+
+int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
+		   unsigned long bits, u64 * failed_start,
+		   struct extent_state **cached_state, gfp_t mask)
+{
+	return __set_extent_bit(tree, start, end, bits, 0, failed_start,
+				cached_state, mask);
+}
+
+
+/**
+ * convert_extent_bit - convert all bits in a given range from one bit to
+ * 			another
+ * @tree:	the io tree to search
+ * @start:	the start offset in bytes
+ * @end:	the end offset in bytes (inclusive)
+ * @bits:	the bits to set in this range
+ * @clear_bits:	the bits to clear in this range
+ * @cached_state:	state that we're going to cache
+ * @mask:	the allocation mask
+ *
+ * This will go through and set bits for the given range.  If any states exist
+ * already in this range they are set with the given bit and cleared of the
+ * clear_bits.  This is only meant to be used by things that are mergeable, ie
+ * converting from say DELALLOC to DIRTY.  This is not meant to be used with
+ * boundary bits like LOCK.
+ */
+int convert_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
+		       unsigned long bits, unsigned long clear_bits,
+		       struct extent_state **cached_state, gfp_t mask)
+{
+	struct extent_state *state;
+	struct extent_state *prealloc = NULL;
+	struct rb_node *node;
+	struct rb_node **p;
+	struct rb_node *parent;
+	int err = 0;
+	u64 last_start;
+	u64 last_end;
+
+	btrfs_debug_check_extent_io_range(tree, start, end);
+
+again:
+	if (!prealloc && (mask & __GFP_WAIT)) {
+		prealloc = alloc_extent_state(mask);
+		if (!prealloc)
+			return -ENOMEM;
+	}
+
+	spin_lock(&tree->lock);
+	if (cached_state && *cached_state) {
+		state = *cached_state;
+		if (state->start <= start && state->end > start &&
+		    state->tree) {
+			node = &state->rb_node;
+			goto hit_next;
+		}
+	}
+
+	/*
+	 * this search will find all the extents that end after
+	 * our range starts.
+	 */
+	node = tree_search_for_insert(tree, start, &p, &parent);
+	if (!node) {
+		prealloc = alloc_extent_state_atomic(prealloc);
+		if (!prealloc) {
+			err = -ENOMEM;
 			goto out;
 		}
+		err = insert_state(tree, prealloc, start, end,
+				   &p, &parent, &bits);
+		if (err)
+			extent_io_tree_panic(tree, err);
+		cache_state(prealloc, cached_state);
+		prealloc = NULL;
+		goto out;
+	}
+	state = rb_entry(node, struct extent_state, rb_node);
+hit_next:
+	last_start = state->start;
+	last_end = state->end;
+
+	/*
+	 * | ---- desired range ---- |
+	 * | state |
+	 *
+	 * Just lock what we found and keep going
+	 */
+	if (state->start == start && state->end <= end) {
+		set_state_bits(tree, state, &bits);
+		cache_state(state, cached_state);
+		state = clear_state_bit(tree, state, &clear_bits, 0);
+		if (last_end == (u64)-1)
+			goto out;
+		start = last_end + 1;
+		if (start < end && state && state->start == start &&
+		    !need_resched())
+			goto hit_next;
+		goto search_again;
+	}
+
+	/*
+	 *     | ---- desired range ---- |
+	 * | state |
+	 *   or
+	 * | ------------- state -------------- |
+	 *
+	 * We need to split the extent we found, and may flip bits on
+	 * second half.
+	 *
+	 * If the extent we found extends past our
+	 * range, we just split and search again.  It'll get split
+	 * again the next time though.
+	 *
+	 * If the extent we found is inside our range, we set the
+	 * desired bit on it.
+	 */
+	if (state->start < start) {
+		prealloc = alloc_extent_state_atomic(prealloc);
+		if (!prealloc) {
+			err = -ENOMEM;
+			goto out;
+		}
+		err = split_state(tree, state, prealloc, start);
+		if (err)
+			extent_io_tree_panic(tree, err);
+		prealloc = NULL;
+		if (err)
+			goto out;
+		if (state->end <= end) {
+			set_state_bits(tree, state, &bits);
+			cache_state(state, cached_state);
+			state = clear_state_bit(tree, state, &clear_bits, 0);
+			if (last_end == (u64)-1)
+				goto out;
+			start = last_end + 1;
+			if (start < end && state && state->start == start &&
+			    !need_resched())
+				goto hit_next;
+		}
+		goto search_again;
+	}
+	/*
+	 * | ---- desired range ---- |
+	 *     | state | or               | state |
+	 *
+	 * There's a hole, we need to insert something in it and
+	 * ignore the extent we found.
+	 */
+	if (state->start > start) {
+		u64 this_end;
+		if (end < last_start)
+			this_end = end;
+		else
+			this_end = last_start - 1;
+
+		prealloc = alloc_extent_state_atomic(prealloc);
+		if (!prealloc) {
+			err = -ENOMEM;
+			goto out;
+		}
+
+		/*
+		 * Avoid to free 'prealloc' if it can be merged with
+		 * the later extent.
+		 */
+		err = insert_state(tree, prealloc, start, this_end,
+				   NULL, NULL, &bits);
+		if (err)
+			extent_io_tree_panic(tree, err);
 		cache_state(prealloc, cached_state);
-		merge_state(tree, prealloc);
+		prealloc = NULL;
+		start = this_end + 1;
+		goto search_again;
+	}
+	/*
+	 * | ---- desired range ---- |
+	 *                        | state |
+	 * We need to split the extent, and set the bit
+	 * on the first half
+	 */
+	if (state->start <= end && state->end > end) {
+		prealloc = alloc_extent_state_atomic(prealloc);
+		if (!prealloc) {
+			err = -ENOMEM;
+			goto out;
+		}
+
+		err = split_state(tree, state, prealloc, end + 1);
+		if (err)
+			extent_io_tree_panic(tree, err);
+
+		set_state_bits(tree, prealloc, &bits);
+		cache_state(prealloc, cached_state);
+		clear_state_bit(tree, prealloc, &clear_bits, 0);
 		prealloc = NULL;
 		goto out;
 	}
@@ -936,19 +1238,19 @@ search_again:
 int set_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end,
 		     gfp_t mask)
 {
-	return set_extent_bit(tree, start, end, EXTENT_DIRTY, 0, NULL,
+	return set_extent_bit(tree, start, end, EXTENT_DIRTY, NULL,
 			      NULL, mask);
 }
 
 int set_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
-		    int bits, gfp_t mask)
+		    unsigned long bits, gfp_t mask)
 {
-	return set_extent_bit(tree, start, end, bits, 0, NULL,
+	return set_extent_bit(tree, start, end, bits, NULL,
 			      NULL, mask);
 }
 
 int clear_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
-		      int bits, gfp_t mask)
+		      unsigned long bits, gfp_t mask)
 {
 	return clear_extent_bit(tree, start, end, bits, 0, 0, NULL, mask);
 }
@@ -957,8 +1259,16 @@ int set_extent_delalloc(struct extent_io_tree *tree, u64 start, u64 end,
 			struct extent_state **cached_state, gfp_t mask)
 {
 	return set_extent_bit(tree, start, end,
-			      EXTENT_DELALLOC | EXTENT_DIRTY | EXTENT_UPTODATE,
-			      0, NULL, cached_state, mask);
+			      EXTENT_DELALLOC | EXTENT_UPTODATE,
+			      NULL, cached_state, mask);
+}
+
+int set_extent_defrag(struct extent_io_tree *tree, u64 start, u64 end,
+		      struct extent_state **cached_state, gfp_t mask)
+{
+	return set_extent_bit(tree, start, end,
+			      EXTENT_DELALLOC | EXTENT_UPTODATE | EXTENT_DEFRAG,
+			      NULL, cached_state, mask);
 }
 
 int clear_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end,
@@ -966,85 +1276,69 @@ int clear_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end,
 {
 	return clear_extent_bit(tree, start, end,
 				EXTENT_DIRTY | EXTENT_DELALLOC |
-				EXTENT_DO_ACCOUNTING, 0, 0,
-				NULL, mask);
+				EXTENT_DO_ACCOUNTING, 0, 0, NULL, mask);
 }
 
 int set_extent_new(struct extent_io_tree *tree, u64 start, u64 end,
 		     gfp_t mask)
 {
-	return set_extent_bit(tree, start, end, EXTENT_NEW, 0, NULL,
+	return set_extent_bit(tree, start, end, EXTENT_NEW, NULL,
 			      NULL, mask);
 }
 
-static int clear_extent_new(struct extent_io_tree *tree, u64 start, u64 end,
-		       gfp_t mask)
-{
-	return clear_extent_bit(tree, start, end, EXTENT_NEW, 0, 0,
-				NULL, mask);
-}
-
 int set_extent_uptodate(struct extent_io_tree *tree, u64 start, u64 end,
-			gfp_t mask)
+			struct extent_state **cached_state, gfp_t mask)
 {
-	return set_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, NULL,
-			      NULL, mask);
+	return set_extent_bit(tree, start, end, EXTENT_UPTODATE, NULL,
+			      cached_state, mask);
 }
 
-static int clear_extent_uptodate(struct extent_io_tree *tree, u64 start,
-				 u64 end, struct extent_state **cached_state,
-				 gfp_t mask)
+int clear_extent_uptodate(struct extent_io_tree *tree, u64 start, u64 end,
+			  struct extent_state **cached_state, gfp_t mask)
 {
 	return clear_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, 0,
 				cached_state, mask);
 }
 
-int wait_on_extent_writeback(struct extent_io_tree *tree, u64 start, u64 end)
-{
-	return wait_extent_bit(tree, start, end, EXTENT_WRITEBACK);
-}
-
 /*
  * either insert or lock state struct between start and end use mask to tell
  * us if waiting is desired.
  */
 int lock_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
-		     int bits, struct extent_state **cached_state, gfp_t mask)
+		     unsigned long bits, struct extent_state **cached_state)
 {
 	int err;
 	u64 failed_start;
 	while (1) {
-		err = set_extent_bit(tree, start, end, EXTENT_LOCKED | bits,
-				     EXTENT_LOCKED, &failed_start,
-				     cached_state, mask);
-		if (err == -EEXIST && (mask & __GFP_WAIT)) {
+		err = __set_extent_bit(tree, start, end, EXTENT_LOCKED | bits,
+				       EXTENT_LOCKED, &failed_start,
+				       cached_state, GFP_NOFS);
+		if (err == -EEXIST) {
 			wait_extent_bit(tree, failed_start, end, EXTENT_LOCKED);
 			start = failed_start;
-		} else {
+		} else
 			break;
-		}
 		WARN_ON(start > end);
 	}
 	return err;
 }
 
-int lock_extent(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask)
+int lock_extent(struct extent_io_tree *tree, u64 start, u64 end)
 {
-	return lock_extent_bits(tree, start, end, 0, NULL, mask);
+	return lock_extent_bits(tree, start, end, 0, NULL);
 }
 
-int try_lock_extent(struct extent_io_tree *tree, u64 start, u64 end,
-		    gfp_t mask)
+int try_lock_extent(struct extent_io_tree *tree, u64 start, u64 end)
 {
 	int err;
 	u64 failed_start;
 
-	err = set_extent_bit(tree, start, end, EXTENT_LOCKED, EXTENT_LOCKED,
-			     &failed_start, NULL, mask);
+	err = __set_extent_bit(tree, start, end, EXTENT_LOCKED, EXTENT_LOCKED,
+			       &failed_start, NULL, GFP_NOFS);
 	if (err == -EEXIST) {
 		if (failed_start > start)
 			clear_extent_bit(tree, start, failed_start - 1,
-					 EXTENT_LOCKED, 1, 0, NULL, mask);
+					 EXTENT_LOCKED, 1, 0, NULL, GFP_NOFS);
 		return 0;
 	}
 	return 1;
@@ -1057,25 +1351,38 @@ int unlock_extent_cached(struct extent_io_tree *tree, u64 start, u64 end,
 				mask);
 }
 
-int unlock_extent(struct extent_io_tree *tree, u64 start, u64 end,
-		  gfp_t mask)
+int unlock_extent(struct extent_io_tree *tree, u64 start, u64 end)
 {
 	return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, NULL,
-				mask);
+				GFP_NOFS);
 }
 
-/*
- * helper function to set pages and extents in the tree dirty
- */
-int set_range_dirty(struct extent_io_tree *tree, u64 start, u64 end)
+int extent_range_clear_dirty_for_io(struct inode *inode, u64 start, u64 end)
 {
 	unsigned long index = start >> PAGE_CACHE_SHIFT;
 	unsigned long end_index = end >> PAGE_CACHE_SHIFT;
 	struct page *page;
 
 	while (index <= end_index) {
-		page = find_get_page(tree->mapping, index);
-		BUG_ON(!page);
+		page = find_get_page(inode->i_mapping, index);
+		BUG_ON(!page); /* Pages should be in the extent_io_tree */
+		clear_page_dirty_for_io(page);
+		page_cache_release(page);
+		index++;
+	}
+	return 0;
+}
+
+int extent_range_redirty_for_io(struct inode *inode, u64 start, u64 end)
+{
+	unsigned long index = start >> PAGE_CACHE_SHIFT;
+	unsigned long end_index = end >> PAGE_CACHE_SHIFT;
+	struct page *page;
+
+	while (index <= end_index) {
+		page = find_get_page(inode->i_mapping, index);
+		BUG_ON(!page); /* Pages should be in the extent_io_tree */
+		account_page_redirty(page);
 		__set_page_dirty_nobuffers(page);
 		page_cache_release(page);
 		index++;
@@ -1094,7 +1401,7 @@ static int set_range_writeback(struct extent_io_tree *tree, u64 start, u64 end)
 
 	while (index <= end_index) {
 		page = find_get_page(tree->mapping, index);
-		BUG_ON(!page);
+		BUG_ON(!page); /* Pages should be in the extent_io_tree */
 		set_page_writeback(page);
 		page_cache_release(page);
 		index++;
@@ -1102,21 +1409,17 @@ static int set_range_writeback(struct extent_io_tree *tree, u64 start, u64 end)
 	return 0;
 }
 
-/*
- * find the first offset in the io tree with 'bits' set. zero is
- * returned if we find something, and *start_ret and *end_ret are
- * set to reflect the state struct that was found.
- *
- * If nothing was found, 1 is returned, < 0 on error
+/* find the first state struct with 'bits' set after 'start', and
+ * return it.  tree->lock must be held.  NULL will returned if
+ * nothing was found after 'start'
  */
-int find_first_extent_bit(struct extent_io_tree *tree, u64 start,
-			  u64 *start_ret, u64 *end_ret, int bits)
+static struct extent_state *
+find_first_extent_bit_state(struct extent_io_tree *tree,
+			    u64 start, unsigned long bits)
 {
 	struct rb_node *node;
 	struct extent_state *state;
-	int ret = 1;
 
-	spin_lock(&tree->lock);
 	/*
 	 * this search will find all the extents that end after
 	 * our range starts.
@@ -1127,50 +1430,63 @@ int find_first_extent_bit(struct extent_io_tree *tree, u64 start,
 
 	while (1) {
 		state = rb_entry(node, struct extent_state, rb_node);
-		if (state->end >= start && (state->state & bits)) {
-			*start_ret = state->start;
-			*end_ret = state->end;
-			ret = 0;
-			break;
-		}
+		if (state->end >= start && (state->state & bits))
+			return state;
+
 		node = rb_next(node);
 		if (!node)
 			break;
 	}
 out:
-	spin_unlock(&tree->lock);
-	return ret;
+	return NULL;
 }
 
-/* find the first state struct with 'bits' set after 'start', and
- * return it.  tree->lock must be held.  NULL will returned if
- * nothing was found after 'start'
+/*
+ * find the first offset in the io tree with 'bits' set. zero is
+ * returned if we find something, and *start_ret and *end_ret are
+ * set to reflect the state struct that was found.
+ *
+ * If nothing was found, 1 is returned. If found something, return 0.
  */
-struct extent_state *find_first_extent_bit_state(struct extent_io_tree *tree,
-						 u64 start, int bits)
+int find_first_extent_bit(struct extent_io_tree *tree, u64 start,
+			  u64 *start_ret, u64 *end_ret, unsigned long bits,
+			  struct extent_state **cached_state)
 {
-	struct rb_node *node;
 	struct extent_state *state;
+	struct rb_node *n;
+	int ret = 1;
 
-	/*
-	 * this search will find all the extents that end after
-	 * our range starts.
-	 */
-	node = tree_search(tree, start);
-	if (!node)
-		goto out;
-
-	while (1) {
-		state = rb_entry(node, struct extent_state, rb_node);
-		if (state->end >= start && (state->state & bits))
-			return state;
+	spin_lock(&tree->lock);
+	if (cached_state && *cached_state) {
+		state = *cached_state;
+		if (state->end == start - 1 && state->tree) {
+			n = rb_next(&state->rb_node);
+			while (n) {
+				state = rb_entry(n, struct extent_state,
+						 rb_node);
+				if (state->state & bits)
+					goto got_it;
+				n = rb_next(n);
+			}
+			free_extent_state(*cached_state);
+			*cached_state = NULL;
+			goto out;
+		}
+		free_extent_state(*cached_state);
+		*cached_state = NULL;
+	}
 
-		node = rb_next(node);
-		if (!node)
-			break;
+	state = find_first_extent_bit_state(tree, start, bits);
+got_it:
+	if (state) {
+		cache_state(state, cached_state);
+		*start_ret = state->start;
+		*end_ret = state->end;
+		ret = 0;
 	}
 out:
-	return NULL;
+	spin_unlock(&tree->lock);
+	return ret;
 }
 
 /*
@@ -1222,20 +1538,20 @@ static noinline u64 find_delalloc_range(struct extent_io_tree *tree,
 		*end = state->end;
 		cur_start = state->end + 1;
 		node = rb_next(node);
-		if (!node)
-			break;
 		total_bytes += state->end - state->start + 1;
 		if (total_bytes >= max_bytes)
 			break;
+		if (!node)
+			break;
 	}
 out:
 	spin_unlock(&tree->lock);
 	return found;
 }
 
-static noinline int __unlock_for_delalloc(struct inode *inode,
-					  struct page *locked_page,
-					  u64 start, u64 end)
+static noinline void __unlock_for_delalloc(struct inode *inode,
+					   struct page *locked_page,
+					   u64 start, u64 end)
 {
 	int ret;
 	struct page *pages[16];
@@ -1245,7 +1561,7 @@ static noinline int __unlock_for_delalloc(struct inode *inode,
 	int i;
 
 	if (index == locked_page->index && end_index == index)
-		return 0;
+		return;
 
 	while (nr_pages > 0) {
 		ret = find_get_pages_contig(inode->i_mapping, index,
@@ -1260,7 +1576,6 @@ static noinline int __unlock_for_delalloc(struct inode *inode,
 		index += ret;
 		cond_resched();
 	}
-	return 0;
 }
 
 static noinline int lock_delalloc_pages(struct inode *inode,
@@ -1331,11 +1646,10 @@ done:
  *
  * 1 is returned if we find something, 0 if nothing was in the tree
  */
-static noinline u64 find_lock_delalloc_range(struct inode *inode,
-					     struct extent_io_tree *tree,
-					     struct page *locked_page,
-					     u64 *start, u64 *end,
-					     u64 max_bytes)
+STATIC u64 find_lock_delalloc_range(struct inode *inode,
+				    struct extent_io_tree *tree,
+				    struct page *locked_page, u64 *start,
+				    u64 *end, u64 max_bytes)
 {
 	u64 delalloc_start;
 	u64 delalloc_end;
@@ -1354,7 +1668,7 @@ again:
 		*start = delalloc_start;
 		*end = delalloc_end;
 		free_extent_state(cached_state);
-		return found;
+		return 0;
 	}
 
 	/*
@@ -1367,10 +1681,9 @@ again:
 
 	/*
 	 * make sure to limit the number of pages we try to lock down
-	 * if we're looping.
 	 */
-	if (delalloc_end + 1 - delalloc_start > max_bytes && loops)
-		delalloc_end = delalloc_start + PAGE_CACHE_SIZE - 1;
+	if (delalloc_end + 1 - delalloc_start > max_bytes)
+		delalloc_end = delalloc_start + max_bytes - 1;
 
 	/* step two, lock all the pages after the page that has start */
 	ret = lock_delalloc_pages(inode, locked_page,
@@ -1380,9 +1693,9 @@ again:
 		 * shortening the size of the delalloc range we're searching
 		 */
 		free_extent_state(cached_state);
+		cached_state = NULL;
 		if (!loops) {
-			unsigned long offset = (*start) & (PAGE_CACHE_SIZE - 1);
-			max_bytes = PAGE_CACHE_SIZE - offset;
+			max_bytes = PAGE_CACHE_SIZE;
 			loops = 1;
 			goto again;
 		} else {
@@ -1390,11 +1703,10 @@ again:
 			goto out_failed;
 		}
 	}
-	BUG_ON(ret);
+	BUG_ON(ret); /* Only valid values are 0 and -EAGAIN */
 
 	/* step three, lock the state bits for the whole range */
-	lock_extent_bits(tree, delalloc_start, delalloc_end,
-			 0, &cached_state, GFP_NOFS);
+	lock_extent_bits(tree, delalloc_start, delalloc_end, 0, &cached_state);
 
 	/* then test to make sure it is all still delalloc */
 	ret = test_range_bit(tree, delalloc_start, delalloc_end,
@@ -1414,34 +1726,21 @@ out_failed:
 	return found;
 }
 
-int extent_clear_unlock_delalloc(struct inode *inode,
-				struct extent_io_tree *tree,
-				u64 start, u64 end, struct page *locked_page,
-				unsigned long op)
+int extent_clear_unlock_delalloc(struct inode *inode, u64 start, u64 end,
+				 struct page *locked_page,
+				 unsigned long clear_bits,
+				 unsigned long page_ops)
 {
+	struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
 	int ret;
 	struct page *pages[16];
 	unsigned long index = start >> PAGE_CACHE_SHIFT;
 	unsigned long end_index = end >> PAGE_CACHE_SHIFT;
 	unsigned long nr_pages = end_index - index + 1;
 	int i;
-	int clear_bits = 0;
-
-	if (op & EXTENT_CLEAR_UNLOCK)
-		clear_bits |= EXTENT_LOCKED;
-	if (op & EXTENT_CLEAR_DIRTY)
-		clear_bits |= EXTENT_DIRTY;
-
-	if (op & EXTENT_CLEAR_DELALLOC)
-		clear_bits |= EXTENT_DELALLOC;
-
-	if (op & EXTENT_CLEAR_ACCOUNTING)
-		clear_bits |= EXTENT_DO_ACCOUNTING;
 
 	clear_extent_bit(tree, start, end, clear_bits, 1, 0, NULL, GFP_NOFS);
-	if (!(op & (EXTENT_CLEAR_UNLOCK_PAGE | EXTENT_CLEAR_DIRTY |
-		    EXTENT_SET_WRITEBACK | EXTENT_END_WRITEBACK |
-		    EXTENT_SET_PRIVATE2)))
+	if (page_ops == 0)
 		return 0;
 
 	while (nr_pages > 0) {
@@ -1450,20 +1749,20 @@ int extent_clear_unlock_delalloc(struct inode *inode,
 				     nr_pages, ARRAY_SIZE(pages)), pages);
 		for (i = 0; i < ret; i++) {
 
-			if (op & EXTENT_SET_PRIVATE2)
+			if (page_ops & PAGE_SET_PRIVATE2)
 				SetPagePrivate2(pages[i]);
 
 			if (pages[i] == locked_page) {
 				page_cache_release(pages[i]);
 				continue;
 			}
-			if (op & EXTENT_CLEAR_DIRTY)
+			if (page_ops & PAGE_CLEAR_DIRTY)
 				clear_page_dirty_for_io(pages[i]);
-			if (op & EXTENT_SET_WRITEBACK)
+			if (page_ops & PAGE_SET_WRITEBACK)
 				set_page_writeback(pages[i]);
-			if (op & EXTENT_END_WRITEBACK)
+			if (page_ops & PAGE_END_WRITEBACK)
 				end_page_writeback(pages[i]);
-			if (op & EXTENT_CLEAR_UNLOCK_PAGE)
+			if (page_ops & PAGE_UNLOCK)
 				unlock_page(pages[i]);
 			page_cache_release(pages[i]);
 		}
@@ -1481,18 +1780,17 @@ int extent_clear_unlock_delalloc(struct inode *inode,
  */
 u64 count_range_bits(struct extent_io_tree *tree,
 		     u64 *start, u64 search_end, u64 max_bytes,
-		     unsigned long bits)
+		     unsigned long bits, int contig)
 {
 	struct rb_node *node;
 	struct extent_state *state;
 	u64 cur_start = *start;
 	u64 total_bytes = 0;
+	u64 last = 0;
 	int found = 0;
 
-	if (search_end <= cur_start) {
-		WARN_ON(1);
+	if (WARN_ON(search_end <= cur_start))
 		return 0;
-	}
 
 	spin_lock(&tree->lock);
 	if (cur_start == 0 && bits == EXTENT_DIRTY) {
@@ -1511,15 +1809,20 @@ u64 count_range_bits(struct extent_io_tree *tree,
 		state = rb_entry(node, struct extent_state, rb_node);
 		if (state->start > search_end)
 			break;
-		if (state->end >= cur_start && (state->state & bits)) {
+		if (contig && found && state->start > last + 1)
+			break;
+		if (state->end >= cur_start && (state->state & bits) == bits) {
 			total_bytes += min(search_end, state->end) + 1 -
 				       max(cur_start, state->start);
 			if (total_bytes >= max_bytes)
 				break;
 			if (!found) {
-				*start = state->start;
+				*start = max(cur_start, state->start);
 				found = 1;
 			}
+			last = state->end;
+		} else if (contig && found) {
+			break;
 		}
 		node = rb_next(node);
 		if (!node)
@@ -1534,7 +1837,7 @@ out:
  * set the private field for a given byte offset in the tree.  If there isn't
  * an extent_state there already, this does nothing.
  */
-int set_state_private(struct extent_io_tree *tree, u64 start, u64 private)
+static int set_state_private(struct extent_io_tree *tree, u64 start, u64 private)
 {
 	struct rb_node *node;
 	struct extent_state *state;
@@ -1595,14 +1898,15 @@ out:
  * range is found set.
  */
 int test_range_bit(struct extent_io_tree *tree, u64 start, u64 end,
-		   int bits, int filled, struct extent_state *cached)
+		   unsigned long bits, int filled, struct extent_state *cached)
 {
 	struct extent_state *state = NULL;
 	struct rb_node *node;
 	int bitset = 0;
 
 	spin_lock(&tree->lock);
-	if (cached && cached->tree && cached->start == start)
+	if (cached && cached->tree && cached->start <= start &&
+	    cached->end > start)
 		node = &cached->rb_node;
 	else
 		node = tree_search(tree, start);
@@ -1647,43 +1951,429 @@ int test_range_bit(struct extent_io_tree *tree, u64 start, u64 end,
  * helper function to set a given page up to date if all the
  * extents in the tree for that page are up to date
  */
-static int check_page_uptodate(struct extent_io_tree *tree,
-			       struct page *page)
+static void check_page_uptodate(struct extent_io_tree *tree, struct page *page)
 {
-	u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
+	u64 start = page_offset(page);
 	u64 end = start + PAGE_CACHE_SIZE - 1;
 	if (test_range_bit(tree, start, end, EXTENT_UPTODATE, 1, NULL))
 		SetPageUptodate(page);
-	return 0;
 }
 
 /*
- * helper function to unlock a page if all the extents in the tree
- * for that page are unlocked
+ * When IO fails, either with EIO or csum verification fails, we
+ * try other mirrors that might have a good copy of the data.  This
+ * io_failure_record is used to record state as we go through all the
+ * mirrors.  If another mirror has good data, the page is set up to date
+ * and things continue.  If a good mirror can't be found, the original
+ * bio end_io callback is called to indicate things have failed.
  */
-static int check_page_locked(struct extent_io_tree *tree,
-			     struct page *page)
+struct io_failure_record {
+	struct page *page;
+	u64 start;
+	u64 len;
+	u64 logical;
+	unsigned long bio_flags;
+	int this_mirror;
+	int failed_mirror;
+	int in_validation;
+};
+
+static int free_io_failure(struct inode *inode, struct io_failure_record *rec,
+				int did_repair)
 {
-	u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
-	u64 end = start + PAGE_CACHE_SIZE - 1;
-	if (!test_range_bit(tree, start, end, EXTENT_LOCKED, 0, NULL))
-		unlock_page(page);
-	return 0;
+	int ret;
+	int err = 0;
+	struct extent_io_tree *failure_tree = &BTRFS_I(inode)->io_failure_tree;
+
+	set_state_private(failure_tree, rec->start, 0);
+	ret = clear_extent_bits(failure_tree, rec->start,
+				rec->start + rec->len - 1,
+				EXTENT_LOCKED | EXTENT_DIRTY, GFP_NOFS);
+	if (ret)
+		err = ret;
+
+	ret = clear_extent_bits(&BTRFS_I(inode)->io_tree, rec->start,
+				rec->start + rec->len - 1,
+				EXTENT_DAMAGED, GFP_NOFS);
+	if (ret && !err)
+		err = ret;
+
+	kfree(rec);
+	return err;
 }
 
 /*
- * helper function to end page writeback if all the extents
- * in the tree for that page are done with writeback
+ * this bypasses the standard btrfs submit functions deliberately, as
+ * the standard behavior is to write all copies in a raid setup. here we only
+ * want to write the one bad copy. so we do the mapping for ourselves and issue
+ * submit_bio directly.
+ * to avoid any synchronization issues, wait for the data after writing, which
+ * actually prevents the read that triggered the error from finishing.
+ * currently, there can be no more than two copies of every data bit. thus,
+ * exactly one rewrite is required.
  */
-static int check_page_writeback(struct extent_io_tree *tree,
-			     struct page *page)
+int repair_io_failure(struct btrfs_fs_info *fs_info, u64 start,
+			u64 length, u64 logical, struct page *page,
+			int mirror_num)
 {
-	end_page_writeback(page);
+	struct bio *bio;
+	struct btrfs_device *dev;
+	u64 map_length = 0;
+	u64 sector;
+	struct btrfs_bio *bbio = NULL;
+	struct btrfs_mapping_tree *map_tree = &fs_info->mapping_tree;
+	int ret;
+
+	ASSERT(!(fs_info->sb->s_flags & MS_RDONLY));
+	BUG_ON(!mirror_num);
+
+	/* we can't repair anything in raid56 yet */
+	if (btrfs_is_parity_mirror(map_tree, logical, length, mirror_num))
+		return 0;
+
+	bio = btrfs_io_bio_alloc(GFP_NOFS, 1);
+	if (!bio)
+		return -EIO;
+	bio->bi_iter.bi_size = 0;
+	map_length = length;
+
+	ret = btrfs_map_block(fs_info, WRITE, logical,
+			      &map_length, &bbio, mirror_num);
+	if (ret) {
+		bio_put(bio);
+		return -EIO;
+	}
+	BUG_ON(mirror_num != bbio->mirror_num);
+	sector = bbio->stripes[mirror_num-1].physical >> 9;
+	bio->bi_iter.bi_sector = sector;
+	dev = bbio->stripes[mirror_num-1].dev;
+	kfree(bbio);
+	if (!dev || !dev->bdev || !dev->writeable) {
+		bio_put(bio);
+		return -EIO;
+	}
+	bio->bi_bdev = dev->bdev;
+	bio_add_page(bio, page, length, start - page_offset(page));
+
+	if (btrfsic_submit_bio_wait(WRITE_SYNC, bio)) {
+		/* try to remap that extent elsewhere? */
+		bio_put(bio);
+		btrfs_dev_stat_inc_and_print(dev, BTRFS_DEV_STAT_WRITE_ERRS);
+		return -EIO;
+	}
+
+	printk_ratelimited_in_rcu(KERN_INFO
+			"BTRFS: read error corrected: ino %lu off %llu "
+		    "(dev %s sector %llu)\n", page->mapping->host->i_ino,
+		    start, rcu_str_deref(dev->name), sector);
+
+	bio_put(bio);
 	return 0;
 }
 
+int repair_eb_io_failure(struct btrfs_root *root, struct extent_buffer *eb,
+			 int mirror_num)
+{
+	u64 start = eb->start;
+	unsigned long i, num_pages = num_extent_pages(eb->start, eb->len);
+	int ret = 0;
+
+	if (root->fs_info->sb->s_flags & MS_RDONLY)
+		return -EROFS;
+
+	for (i = 0; i < num_pages; i++) {
+		struct page *p = extent_buffer_page(eb, i);
+		ret = repair_io_failure(root->fs_info, start, PAGE_CACHE_SIZE,
+					start, p, mirror_num);
+		if (ret)
+			break;
+		start += PAGE_CACHE_SIZE;
+	}
+
+	return ret;
+}
+
+/*
+ * each time an IO finishes, we do a fast check in the IO failure tree
+ * to see if we need to process or clean up an io_failure_record
+ */
+static int clean_io_failure(u64 start, struct page *page)
+{
+	u64 private;
+	u64 private_failure;
+	struct io_failure_record *failrec;
+	struct inode *inode = page->mapping->host;
+	struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
+	struct extent_state *state;
+	int num_copies;
+	int did_repair = 0;
+	int ret;
+
+	private = 0;
+	ret = count_range_bits(&BTRFS_I(inode)->io_failure_tree, &private,
+				(u64)-1, 1, EXTENT_DIRTY, 0);
+	if (!ret)
+		return 0;
+
+	ret = get_state_private(&BTRFS_I(inode)->io_failure_tree, start,
+				&private_failure);
+	if (ret)
+		return 0;
+
+	failrec = (struct io_failure_record *)(unsigned long) private_failure;
+	BUG_ON(!failrec->this_mirror);
+
+	if (failrec->in_validation) {
+		/* there was no real error, just free the record */
+		pr_debug("clean_io_failure: freeing dummy error at %llu\n",
+			 failrec->start);
+		did_repair = 1;
+		goto out;
+	}
+	if (fs_info->sb->s_flags & MS_RDONLY)
+		goto out;
+
+	spin_lock(&BTRFS_I(inode)->io_tree.lock);
+	state = find_first_extent_bit_state(&BTRFS_I(inode)->io_tree,
+					    failrec->start,
+					    EXTENT_LOCKED);
+	spin_unlock(&BTRFS_I(inode)->io_tree.lock);
+
+	if (state && state->start <= failrec->start &&
+	    state->end >= failrec->start + failrec->len - 1) {
+		num_copies = btrfs_num_copies(fs_info, failrec->logical,
+					      failrec->len);
+		if (num_copies > 1)  {
+			ret = repair_io_failure(fs_info, start, failrec->len,
+						failrec->logical, page,
+						failrec->failed_mirror);
+			did_repair = !ret;
+		}
+		ret = 0;
+	}
+
+out:
+	if (!ret)
+		ret = free_io_failure(inode, failrec, did_repair);
+
+	return ret;
+}
+
+/*
+ * this is a generic handler for readpage errors (default
+ * readpage_io_failed_hook). if other copies exist, read those and write back
+ * good data to the failed position. does not investigate in remapping the
+ * failed extent elsewhere, hoping the device will be smart enough to do this as
+ * needed
+ */
+
+static int bio_readpage_error(struct bio *failed_bio, u64 phy_offset,
+			      struct page *page, u64 start, u64 end,
+			      int failed_mirror)
+{
+	struct io_failure_record *failrec = NULL;
+	u64 private;
+	struct extent_map *em;
+	struct inode *inode = page->mapping->host;
+	struct extent_io_tree *failure_tree = &BTRFS_I(inode)->io_failure_tree;
+	struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
+	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
+	struct bio *bio;
+	struct btrfs_io_bio *btrfs_failed_bio;
+	struct btrfs_io_bio *btrfs_bio;
+	int num_copies;
+	int ret;
+	int read_mode;
+	u64 logical;
+
+	BUG_ON(failed_bio->bi_rw & REQ_WRITE);
+
+	ret = get_state_private(failure_tree, start, &private);
+	if (ret) {
+		failrec = kzalloc(sizeof(*failrec), GFP_NOFS);
+		if (!failrec)
+			return -ENOMEM;
+		failrec->start = start;
+		failrec->len = end - start + 1;
+		failrec->this_mirror = 0;
+		failrec->bio_flags = 0;
+		failrec->in_validation = 0;
+
+		read_lock(&em_tree->lock);
+		em = lookup_extent_mapping(em_tree, start, failrec->len);
+		if (!em) {
+			read_unlock(&em_tree->lock);
+			kfree(failrec);
+			return -EIO;
+		}
+
+		if (em->start > start || em->start + em->len <= start) {
+			free_extent_map(em);
+			em = NULL;
+		}
+		read_unlock(&em_tree->lock);
+
+		if (!em) {
+			kfree(failrec);
+			return -EIO;
+		}
+		logical = start - em->start;
+		logical = em->block_start + logical;
+		if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) {
+			logical = em->block_start;
+			failrec->bio_flags = EXTENT_BIO_COMPRESSED;
+			extent_set_compress_type(&failrec->bio_flags,
+						 em->compress_type);
+		}
+		pr_debug("bio_readpage_error: (new) logical=%llu, start=%llu, "
+			 "len=%llu\n", logical, start, failrec->len);
+		failrec->logical = logical;
+		free_extent_map(em);
+
+		/* set the bits in the private failure tree */
+		ret = set_extent_bits(failure_tree, start, end,
+					EXTENT_LOCKED | EXTENT_DIRTY, GFP_NOFS);
+		if (ret >= 0)
+			ret = set_state_private(failure_tree, start,
+						(u64)(unsigned long)failrec);
+		/* set the bits in the inode's tree */
+		if (ret >= 0)
+			ret = set_extent_bits(tree, start, end, EXTENT_DAMAGED,
+						GFP_NOFS);
+		if (ret < 0) {
+			kfree(failrec);
+			return ret;
+		}
+	} else {
+		failrec = (struct io_failure_record *)(unsigned long)private;
+		pr_debug("bio_readpage_error: (found) logical=%llu, "
+			 "start=%llu, len=%llu, validation=%d\n",
+			 failrec->logical, failrec->start, failrec->len,
+			 failrec->in_validation);
+		/*
+		 * when data can be on disk more than twice, add to failrec here
+		 * (e.g. with a list for failed_mirror) to make
+		 * clean_io_failure() clean all those errors at once.
+		 */
+	}
+	num_copies = btrfs_num_copies(BTRFS_I(inode)->root->fs_info,
+				      failrec->logical, failrec->len);
+	if (num_copies == 1) {
+		/*
+		 * we only have a single copy of the data, so don't bother with
+		 * all the retry and error correction code that follows. no
+		 * matter what the error is, it is very likely to persist.
+		 */
+		pr_debug("bio_readpage_error: cannot repair, num_copies=%d, next_mirror %d, failed_mirror %d\n",
+			 num_copies, failrec->this_mirror, failed_mirror);
+		free_io_failure(inode, failrec, 0);
+		return -EIO;
+	}
+
+	/*
+	 * there are two premises:
+	 *	a) deliver good data to the caller
+	 *	b) correct the bad sectors on disk
+	 */
+	if (failed_bio->bi_vcnt > 1) {
+		/*
+		 * to fulfill b), we need to know the exact failing sectors, as
+		 * we don't want to rewrite any more than the failed ones. thus,
+		 * we need separate read requests for the failed bio
+		 *
+		 * if the following BUG_ON triggers, our validation request got
+		 * merged. we need separate requests for our algorithm to work.
+		 */
+		BUG_ON(failrec->in_validation);
+		failrec->in_validation = 1;
+		failrec->this_mirror = failed_mirror;
+		read_mode = READ_SYNC | REQ_FAILFAST_DEV;
+	} else {
+		/*
+		 * we're ready to fulfill a) and b) alongside. get a good copy
+		 * of the failed sector and if we succeed, we have setup
+		 * everything for repair_io_failure to do the rest for us.
+		 */
+		if (failrec->in_validation) {
+			BUG_ON(failrec->this_mirror != failed_mirror);
+			failrec->in_validation = 0;
+			failrec->this_mirror = 0;
+		}
+		failrec->failed_mirror = failed_mirror;
+		failrec->this_mirror++;
+		if (failrec->this_mirror == failed_mirror)
+			failrec->this_mirror++;
+		read_mode = READ_SYNC;
+	}
+
+	if (failrec->this_mirror > num_copies) {
+		pr_debug("bio_readpage_error: (fail) num_copies=%d, next_mirror %d, failed_mirror %d\n",
+			 num_copies, failrec->this_mirror, failed_mirror);
+		free_io_failure(inode, failrec, 0);
+		return -EIO;
+	}
+
+	bio = btrfs_io_bio_alloc(GFP_NOFS, 1);
+	if (!bio) {
+		free_io_failure(inode, failrec, 0);
+		return -EIO;
+	}
+	bio->bi_end_io = failed_bio->bi_end_io;
+	bio->bi_iter.bi_sector = failrec->logical >> 9;
+	bio->bi_bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
+	bio->bi_iter.bi_size = 0;
+
+	btrfs_failed_bio = btrfs_io_bio(failed_bio);
+	if (btrfs_failed_bio->csum) {
+		struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
+		u16 csum_size = btrfs_super_csum_size(fs_info->super_copy);
+
+		btrfs_bio = btrfs_io_bio(bio);
+		btrfs_bio->csum = btrfs_bio->csum_inline;
+		phy_offset >>= inode->i_sb->s_blocksize_bits;
+		phy_offset *= csum_size;
+		memcpy(btrfs_bio->csum, btrfs_failed_bio->csum + phy_offset,
+		       csum_size);
+	}
+
+	bio_add_page(bio, page, failrec->len, start - page_offset(page));
+
+	pr_debug("bio_readpage_error: submitting new read[%#x] to "
+		 "this_mirror=%d, num_copies=%d, in_validation=%d\n", read_mode,
+		 failrec->this_mirror, num_copies, failrec->in_validation);
+
+	ret = tree->ops->submit_bio_hook(inode, read_mode, bio,
+					 failrec->this_mirror,
+					 failrec->bio_flags, 0);
+	return ret;
+}
+
 /* lots and lots of room for performance fixes in the end_bio funcs */
 
+int end_extent_writepage(struct page *page, int err, u64 start, u64 end)
+{
+	int uptodate = (err == 0);
+	struct extent_io_tree *tree;
+	int ret = 0;
+
+	tree = &BTRFS_I(page->mapping->host)->io_tree;
+
+	if (tree->ops && tree->ops->writepage_end_io_hook) {
+		ret = tree->ops->writepage_end_io_hook(page, start,
+					       end, NULL, uptodate);
+		if (ret)
+			uptodate = 0;
+	}
+
+	if (!uptodate) {
+		ClearPageUptodate(page);
+		SetPageError(page);
+		ret = ret < 0 ? ret : -EIO;
+		mapping_set_error(page->mapping, ret);
+	}
+	return 0;
+}
+
 /*
  * after a writepage IO is done, we need to:
  * clear the uptodate bits on error
@@ -1695,61 +2385,55 @@ static int check_page_writeback(struct extent_io_tree *tree,
  */
 static void end_bio_extent_writepage(struct bio *bio, int err)
 {
-	int uptodate = err == 0;
-	struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1;
-	struct extent_io_tree *tree;
+	struct bio_vec *bvec;
 	u64 start;
 	u64 end;
-	int whole_page;
-	int ret;
+	int i;
 
-	do {
+	bio_for_each_segment_all(bvec, bio, i) {
 		struct page *page = bvec->bv_page;
-		tree = &BTRFS_I(page->mapping->host)->io_tree;
-
-		start = ((u64)page->index << PAGE_CACHE_SHIFT) +
-			 bvec->bv_offset;
-		end = start + bvec->bv_len - 1;
 
-		if (bvec->bv_offset == 0 && bvec->bv_len == PAGE_CACHE_SIZE)
-			whole_page = 1;
-		else
-			whole_page = 0;
-
-		if (--bvec >= bio->bi_io_vec)
-			prefetchw(&bvec->bv_page->flags);
-		if (tree->ops && tree->ops->writepage_end_io_hook) {
-			ret = tree->ops->writepage_end_io_hook(page, start,
-						       end, NULL, uptodate);
-			if (ret)
-				uptodate = 0;
+		/* We always issue full-page reads, but if some block
+		 * in a page fails to read, blk_update_request() will
+		 * advance bv_offset and adjust bv_len to compensate.
+		 * Print a warning for nonzero offsets, and an error
+		 * if they don't add up to a full page.  */
+		if (bvec->bv_offset || bvec->bv_len != PAGE_CACHE_SIZE) {
+			if (bvec->bv_offset + bvec->bv_len != PAGE_CACHE_SIZE)
+				btrfs_err(BTRFS_I(page->mapping->host)->root->fs_info,
+				   "partial page write in btrfs with offset %u and length %u",
+					bvec->bv_offset, bvec->bv_len);
+			else
+				btrfs_info(BTRFS_I(page->mapping->host)->root->fs_info,
+				   "incomplete page write in btrfs with offset %u and "
+				   "length %u",
+					bvec->bv_offset, bvec->bv_len);
 		}
 
-		if (!uptodate && tree->ops &&
-		    tree->ops->writepage_io_failed_hook) {
-			ret = tree->ops->writepage_io_failed_hook(bio, page,
-							 start, end, NULL);
-			if (ret == 0) {
-				uptodate = (err == 0);
-				continue;
-			}
-		}
+		start = page_offset(page);
+		end = start + bvec->bv_offset + bvec->bv_len - 1;
 
-		if (!uptodate) {
-			clear_extent_uptodate(tree, start, end, NULL, GFP_NOFS);
-			ClearPageUptodate(page);
-			SetPageError(page);
-		}
+		if (end_extent_writepage(page, err, start, end))
+			continue;
 
-		if (whole_page)
-			end_page_writeback(page);
-		else
-			check_page_writeback(tree, page);
-	} while (bvec >= bio->bi_io_vec);
+		end_page_writeback(page);
+	}
 
 	bio_put(bio);
 }
 
+static void
+endio_readpage_release_extent(struct extent_io_tree *tree, u64 start, u64 len,
+			      int uptodate)
+{
+	struct extent_state *cached = NULL;
+	u64 end = start + len - 1;
+
+	if (uptodate && tree->track_uptodate)
+		set_extent_uptodate(tree, start, end, &cached, GFP_ATOMIC);
+	unlock_extent_cached(tree, start, end, &cached, GFP_ATOMIC);
+}
+
 /*
  * after a readpage IO is done, we need to:
  * clear the uptodate bits on error
@@ -1763,44 +2447,86 @@ static void end_bio_extent_writepage(struct bio *bio, int err)
  */
 static void end_bio_extent_readpage(struct bio *bio, int err)
 {
+	struct bio_vec *bvec;
 	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
-	struct bio_vec *bvec_end = bio->bi_io_vec + bio->bi_vcnt - 1;
-	struct bio_vec *bvec = bio->bi_io_vec;
+	struct btrfs_io_bio *io_bio = btrfs_io_bio(bio);
 	struct extent_io_tree *tree;
+	u64 offset = 0;
 	u64 start;
 	u64 end;
-	int whole_page;
+	u64 len;
+	u64 extent_start = 0;
+	u64 extent_len = 0;
+	int mirror;
 	int ret;
+	int i;
 
 	if (err)
 		uptodate = 0;
 
-	do {
+	bio_for_each_segment_all(bvec, bio, i) {
 		struct page *page = bvec->bv_page;
-		tree = &BTRFS_I(page->mapping->host)->io_tree;
-
-		start = ((u64)page->index << PAGE_CACHE_SHIFT) +
-			bvec->bv_offset;
-		end = start + bvec->bv_len - 1;
-
-		if (bvec->bv_offset == 0 && bvec->bv_len == PAGE_CACHE_SIZE)
-			whole_page = 1;
-		else
-			whole_page = 0;
+		struct inode *inode = page->mapping->host;
+
+		pr_debug("end_bio_extent_readpage: bi_sector=%llu, err=%d, "
+			 "mirror=%lu\n", (u64)bio->bi_iter.bi_sector, err,
+			 io_bio->mirror_num);
+		tree = &BTRFS_I(inode)->io_tree;
+
+		/* We always issue full-page reads, but if some block
+		 * in a page fails to read, blk_update_request() will
+		 * advance bv_offset and adjust bv_len to compensate.
+		 * Print a warning for nonzero offsets, and an error
+		 * if they don't add up to a full page.  */
+		if (bvec->bv_offset || bvec->bv_len != PAGE_CACHE_SIZE) {
+			if (bvec->bv_offset + bvec->bv_len != PAGE_CACHE_SIZE)
+				btrfs_err(BTRFS_I(page->mapping->host)->root->fs_info,
+				   "partial page read in btrfs with offset %u and length %u",
+					bvec->bv_offset, bvec->bv_len);
+			else
+				btrfs_info(BTRFS_I(page->mapping->host)->root->fs_info,
+				   "incomplete page read in btrfs with offset %u and "
+				   "length %u",
+					bvec->bv_offset, bvec->bv_len);
+		}
 
-		if (++bvec <= bvec_end)
-			prefetchw(&bvec->bv_page->flags);
+		start = page_offset(page);
+		end = start + bvec->bv_offset + bvec->bv_len - 1;
+		len = bvec->bv_len;
 
-		if (uptodate && tree->ops && tree->ops->readpage_end_io_hook) {
-			ret = tree->ops->readpage_end_io_hook(page, start, end,
-							      NULL);
+		mirror = io_bio->mirror_num;
+		if (likely(uptodate && tree->ops &&
+			   tree->ops->readpage_end_io_hook)) {
+			ret = tree->ops->readpage_end_io_hook(io_bio, offset,
+							      page, start, end,
+							      mirror);
 			if (ret)
 				uptodate = 0;
+			else
+				clean_io_failure(start, page);
 		}
-		if (!uptodate && tree->ops &&
-		    tree->ops->readpage_io_failed_hook) {
-			ret = tree->ops->readpage_io_failed_hook(bio, page,
-							 start, end, NULL);
+
+		if (likely(uptodate))
+			goto readpage_ok;
+
+		if (tree->ops && tree->ops->readpage_io_failed_hook) {
+			ret = tree->ops->readpage_io_failed_hook(page, mirror);
+			if (!ret && !err &&
+			    test_bit(BIO_UPTODATE, &bio->bi_flags))
+				uptodate = 1;
+		} else {
+			/*
+			 * The generic bio_readpage_error handles errors the
+			 * following way: If possible, new read requests are
+			 * created and submitted and will end up in
+			 * end_bio_extent_readpage as well (if we're lucky, not
+			 * in the !uptodate case). In that case it returns 0 and
+			 * we just go on with the next page in our bio. If it
+			 * can't handle the error it will return -EIO and we
+			 * remain responsible for that page.
+			 */
+			ret = bio_readpage_error(bio, offset, page, start, end,
+						 mirror);
 			if (ret == 0) {
 				uptodate =
 					test_bit(BIO_UPTODATE, &bio->bi_flags);
@@ -1809,122 +2535,149 @@ static void end_bio_extent_readpage(struct bio *bio, int err)
 				continue;
 			}
 		}
-
-		if (uptodate) {
-			set_extent_uptodate(tree, start, end,
-					    GFP_ATOMIC);
+readpage_ok:
+		if (likely(uptodate)) {
+			loff_t i_size = i_size_read(inode);
+			pgoff_t end_index = i_size >> PAGE_CACHE_SHIFT;
+			unsigned offset;
+
+			/* Zero out the end if this page straddles i_size */
+			offset = i_size & (PAGE_CACHE_SIZE-1);
+			if (page->index == end_index && offset)
+				zero_user_segment(page, offset, PAGE_CACHE_SIZE);
+			SetPageUptodate(page);
+		} else {
+			ClearPageUptodate(page);
+			SetPageError(page);
 		}
-		unlock_extent(tree, start, end, GFP_ATOMIC);
-
-		if (whole_page) {
-			if (uptodate) {
-				SetPageUptodate(page);
-			} else {
-				ClearPageUptodate(page);
-				SetPageError(page);
+		unlock_page(page);
+		offset += len;
+
+		if (unlikely(!uptodate)) {
+			if (extent_len) {
+				endio_readpage_release_extent(tree,
+							      extent_start,
+							      extent_len, 1);
+				extent_start = 0;
+				extent_len = 0;
 			}
-			unlock_page(page);
+			endio_readpage_release_extent(tree, start,
+						      end - start + 1, 0);
+		} else if (!extent_len) {
+			extent_start = start;
+			extent_len = end + 1 - start;
+		} else if (extent_start + extent_len == start) {
+			extent_len += end + 1 - start;
 		} else {
-			if (uptodate) {
-				check_page_uptodate(tree, page);
-			} else {
-				ClearPageUptodate(page);
-				SetPageError(page);
-			}
-			check_page_locked(tree, page);
+			endio_readpage_release_extent(tree, extent_start,
+						      extent_len, uptodate);
+			extent_start = start;
+			extent_len = end + 1 - start;
 		}
-	} while (bvec <= bvec_end);
+	}
 
+	if (extent_len)
+		endio_readpage_release_extent(tree, extent_start, extent_len,
+					      uptodate);
+	if (io_bio->end_io)
+		io_bio->end_io(io_bio, err);
 	bio_put(bio);
 }
 
 /*
- * IO done from prepare_write is pretty simple, we just unlock
- * the structs in the extent tree when done, and set the uptodate bits
- * as appropriate.
+ * this allocates from the btrfs_bioset.  We're returning a bio right now
+ * but you can call btrfs_io_bio for the appropriate container_of magic
  */
-static void end_bio_extent_preparewrite(struct bio *bio, int err)
+struct bio *
+btrfs_bio_alloc(struct block_device *bdev, u64 first_sector, int nr_vecs,
+		gfp_t gfp_flags)
 {
-	const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
-	struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1;
-	struct extent_io_tree *tree;
-	u64 start;
-	u64 end;
-
-	do {
-		struct page *page = bvec->bv_page;
-		tree = &BTRFS_I(page->mapping->host)->io_tree;
-
-		start = ((u64)page->index << PAGE_CACHE_SHIFT) +
-			bvec->bv_offset;
-		end = start + bvec->bv_len - 1;
+	struct btrfs_io_bio *btrfs_bio;
+	struct bio *bio;
 
-		if (--bvec >= bio->bi_io_vec)
-			prefetchw(&bvec->bv_page->flags);
+	bio = bio_alloc_bioset(gfp_flags, nr_vecs, btrfs_bioset);
 
-		if (uptodate) {
-			set_extent_uptodate(tree, start, end, GFP_ATOMIC);
-		} else {
-			ClearPageUptodate(page);
-			SetPageError(page);
+	if (bio == NULL && (current->flags & PF_MEMALLOC)) {
+		while (!bio && (nr_vecs /= 2)) {
+			bio = bio_alloc_bioset(gfp_flags,
+					       nr_vecs, btrfs_bioset);
 		}
+	}
 
-		unlock_extent(tree, start, end, GFP_ATOMIC);
-
-	} while (bvec >= bio->bi_io_vec);
-
-	bio_put(bio);
+	if (bio) {
+		bio->bi_bdev = bdev;
+		bio->bi_iter.bi_sector = first_sector;
+		btrfs_bio = btrfs_io_bio(bio);
+		btrfs_bio->csum = NULL;
+		btrfs_bio->csum_allocated = NULL;
+		btrfs_bio->end_io = NULL;
+	}
+	return bio;
 }
 
-static struct bio *
-extent_bio_alloc(struct block_device *bdev, u64 first_sector, int nr_vecs,
-		 gfp_t gfp_flags)
+struct bio *btrfs_bio_clone(struct bio *bio, gfp_t gfp_mask)
 {
-	struct bio *bio;
+	return bio_clone_bioset(bio, gfp_mask, btrfs_bioset);
+}
 
-	bio = bio_alloc(gfp_flags, nr_vecs);
 
-	if (bio == NULL && (current->flags & PF_MEMALLOC)) {
-		while (!bio && (nr_vecs /= 2))
-			bio = bio_alloc(gfp_flags, nr_vecs);
-	}
+/* this also allocates from the btrfs_bioset */
+struct bio *btrfs_io_bio_alloc(gfp_t gfp_mask, unsigned int nr_iovecs)
+{
+	struct btrfs_io_bio *btrfs_bio;
+	struct bio *bio;
 
+	bio = bio_alloc_bioset(gfp_mask, nr_iovecs, btrfs_bioset);
 	if (bio) {
-		bio->bi_size = 0;
-		bio->bi_bdev = bdev;
-		bio->bi_sector = first_sector;
+		btrfs_bio = btrfs_io_bio(bio);
+		btrfs_bio->csum = NULL;
+		btrfs_bio->csum_allocated = NULL;
+		btrfs_bio->end_io = NULL;
 	}
 	return bio;
 }
 
-static int submit_one_bio(int rw, struct bio *bio, int mirror_num,
-			  unsigned long bio_flags)
+
+static int __must_check submit_one_bio(int rw, struct bio *bio,
+				       int mirror_num, unsigned long bio_flags)
 {
 	int ret = 0;
 	struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1;
 	struct page *page = bvec->bv_page;
 	struct extent_io_tree *tree = bio->bi_private;
 	u64 start;
-	u64 end;
 
-	start = ((u64)page->index << PAGE_CACHE_SHIFT) + bvec->bv_offset;
-	end = start + bvec->bv_len - 1;
+	start = page_offset(page) + bvec->bv_offset;
 
 	bio->bi_private = NULL;
 
 	bio_get(bio);
 
 	if (tree->ops && tree->ops->submit_bio_hook)
-		tree->ops->submit_bio_hook(page->mapping->host, rw, bio,
-					   mirror_num, bio_flags);
+		ret = tree->ops->submit_bio_hook(page->mapping->host, rw, bio,
+					   mirror_num, bio_flags, start);
 	else
-		submit_bio(rw, bio);
+		btrfsic_submit_bio(rw, bio);
+
 	if (bio_flagged(bio, BIO_EOPNOTSUPP))
 		ret = -EOPNOTSUPP;
 	bio_put(bio);
 	return ret;
 }
 
+static int merge_bio(int rw, struct extent_io_tree *tree, struct page *page,
+		     unsigned long offset, size_t size, struct bio *bio,
+		     unsigned long bio_flags)
+{
+	int ret = 0;
+	if (tree->ops && tree->ops->merge_bio_hook)
+		ret = tree->ops->merge_bio_hook(rw, page, offset, size, bio,
+						bio_flags);
+	BUG_ON(ret < 0);
+	return ret;
+
+}
+
 static int submit_extent_page(int rw, struct extent_io_tree *tree,
 			      struct page *page, sector_t sector,
 			      size_t size, unsigned long offset,
@@ -1947,18 +2700,17 @@ static int submit_extent_page(int rw, struct extent_io_tree *tree,
 	if (bio_ret && *bio_ret) {
 		bio = *bio_ret;
 		if (old_compressed)
-			contig = bio->bi_sector == sector;
+			contig = bio->bi_iter.bi_sector == sector;
 		else
-			contig = bio->bi_sector + (bio->bi_size >> 9) ==
-				sector;
+			contig = bio_end_sector(bio) == sector;
 
 		if (prev_bio_flags != bio_flags || !contig ||
-		    (tree->ops && tree->ops->merge_bio_hook &&
-		     tree->ops->merge_bio_hook(page, offset, page_size, bio,
-					       bio_flags)) ||
+		    merge_bio(rw, tree, page, offset, page_size, bio, bio_flags) ||
 		    bio_add_page(bio, page, page_size, offset) < page_size) {
 			ret = submit_one_bio(rw, bio, mirror_num,
 					     prev_bio_flags);
+			if (ret < 0)
+				return ret;
 			bio = NULL;
 		} else {
 			return 0;
@@ -1969,7 +2721,9 @@ static int submit_extent_page(int rw, struct extent_io_tree *tree,
 	else
 		nr = bio_get_nr_vecs(bdev);
 
-	bio = extent_bio_alloc(bdev, sector, nr, GFP_NOFS | __GFP_HIGH);
+	bio = btrfs_bio_alloc(bdev, sector, nr, GFP_NOFS | __GFP_HIGH);
+	if (!bio)
+		return -ENOMEM;
 
 	bio_add_page(bio, page, page_size, offset);
 	bio->bi_end_io = end_io_func;
@@ -1983,6 +2737,18 @@ static int submit_extent_page(int rw, struct extent_io_tree *tree,
 	return ret;
 }
 
+static void attach_extent_buffer_page(struct extent_buffer *eb,
+				      struct page *page)
+{
+	if (!PagePrivate(page)) {
+		SetPagePrivate(page);
+		page_cache_get(page);
+		set_page_private(page, (unsigned long)eb);
+	} else {
+		WARN_ON(page->private != (unsigned long)eb);
+	}
+}
+
 void set_page_extent_mapped(struct page *page)
 {
 	if (!PagePrivate(page)) {
@@ -1992,24 +2758,48 @@ void set_page_extent_mapped(struct page *page)
 	}
 }
 
-static void set_page_extent_head(struct page *page, unsigned long len)
+static struct extent_map *
+__get_extent_map(struct inode *inode, struct page *page, size_t pg_offset,
+		 u64 start, u64 len, get_extent_t *get_extent,
+		 struct extent_map **em_cached)
 {
-	set_page_private(page, EXTENT_PAGE_PRIVATE_FIRST_PAGE | len << 2);
-}
+	struct extent_map *em;
 
+	if (em_cached && *em_cached) {
+		em = *em_cached;
+		if (extent_map_in_tree(em) && start >= em->start &&
+		    start < extent_map_end(em)) {
+			atomic_inc(&em->refs);
+			return em;
+		}
+
+		free_extent_map(em);
+		*em_cached = NULL;
+	}
+
+	em = get_extent(inode, page, pg_offset, start, len, 0);
+	if (em_cached && !IS_ERR_OR_NULL(em)) {
+		BUG_ON(*em_cached);
+		atomic_inc(&em->refs);
+		*em_cached = em;
+	}
+	return em;
+}
 /*
  * basic readpage implementation.  Locked extent state structs are inserted
  * into the tree that are removed when the IO is done (by the end_io
  * handlers)
+ * XXX JDM: This needs looking at to ensure proper page locking
  */
-static int __extent_read_full_page(struct extent_io_tree *tree,
-				   struct page *page,
-				   get_extent_t *get_extent,
-				   struct bio **bio, int mirror_num,
-				   unsigned long *bio_flags)
+static int __do_readpage(struct extent_io_tree *tree,
+			 struct page *page,
+			 get_extent_t *get_extent,
+			 struct extent_map **em_cached,
+			 struct bio **bio, int mirror_num,
+			 unsigned long *bio_flags, int rw)
 {
 	struct inode *inode = page->mapping->host;
-	u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
+	u64 start = page_offset(page);
 	u64 page_end = start + PAGE_CACHE_SIZE - 1;
 	u64 end;
 	u64 cur = start;
@@ -2022,16 +2812,23 @@ static int __extent_read_full_page(struct extent_io_tree *tree,
 	struct block_device *bdev;
 	int ret;
 	int nr = 0;
-	size_t page_offset = 0;
+	int parent_locked = *bio_flags & EXTENT_BIO_PARENT_LOCKED;
+	size_t pg_offset = 0;
 	size_t iosize;
 	size_t disk_io_size;
 	size_t blocksize = inode->i_sb->s_blocksize;
-	unsigned long this_bio_flag = 0;
+	unsigned long this_bio_flag = *bio_flags & EXTENT_BIO_PARENT_LOCKED;
 
 	set_page_extent_mapped(page);
 
 	end = page_end;
-	lock_extent(tree, start, end, GFP_NOFS);
+	if (!PageUptodate(page)) {
+		if (cleancache_get_page(page) == 0) {
+			BUG_ON(blocksize != PAGE_SIZE);
+			unlock_extent(tree, start, end);
+			goto out;
+		}
+	}
 
 	if (page->index == last_byte >> PAGE_CACHE_SHIFT) {
 		char *userpage;
@@ -2039,42 +2836,53 @@ static int __extent_read_full_page(struct extent_io_tree *tree,
 
 		if (zero_offset) {
 			iosize = PAGE_CACHE_SIZE - zero_offset;
-			userpage = kmap_atomic(page, KM_USER0);
+			userpage = kmap_atomic(page);
 			memset(userpage + zero_offset, 0, iosize);
 			flush_dcache_page(page);
-			kunmap_atomic(userpage, KM_USER0);
+			kunmap_atomic(userpage);
 		}
 	}
 	while (cur <= end) {
+		unsigned long pnr = (last_byte >> PAGE_CACHE_SHIFT) + 1;
+
 		if (cur >= last_byte) {
 			char *userpage;
-			iosize = PAGE_CACHE_SIZE - page_offset;
-			userpage = kmap_atomic(page, KM_USER0);
-			memset(userpage + page_offset, 0, iosize);
+			struct extent_state *cached = NULL;
+
+			iosize = PAGE_CACHE_SIZE - pg_offset;
+			userpage = kmap_atomic(page);
+			memset(userpage + pg_offset, 0, iosize);
 			flush_dcache_page(page);
-			kunmap_atomic(userpage, KM_USER0);
+			kunmap_atomic(userpage);
 			set_extent_uptodate(tree, cur, cur + iosize - 1,
-					    GFP_NOFS);
-			unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS);
+					    &cached, GFP_NOFS);
+			if (!parent_locked)
+				unlock_extent_cached(tree, cur,
+						     cur + iosize - 1,
+						     &cached, GFP_NOFS);
 			break;
 		}
-		em = get_extent(inode, page, page_offset, cur,
-				end - cur + 1, 0);
-		if (IS_ERR(em) || !em) {
+		em = __get_extent_map(inode, page, pg_offset, cur,
+				      end - cur + 1, get_extent, em_cached);
+		if (IS_ERR_OR_NULL(em)) {
 			SetPageError(page);
-			unlock_extent(tree, cur, end, GFP_NOFS);
+			if (!parent_locked)
+				unlock_extent(tree, cur, end);
 			break;
 		}
 		extent_offset = cur - em->start;
 		BUG_ON(extent_map_end(em) <= cur);
 		BUG_ON(end < cur);
 
-		if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags))
-			this_bio_flag = EXTENT_BIO_COMPRESSED;
+		if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) {
+			this_bio_flag |= EXTENT_BIO_COMPRESSED;
+			extent_set_compress_type(&this_bio_flag,
+						 em->compress_type);
+		}
 
 		iosize = min(extent_map_end(em) - cur, end - cur + 1);
 		cur_end = min(extent_map_end(em) - 1, end);
-		iosize = (iosize + blocksize - 1) & ~((u64)blocksize - 1);
+		iosize = ALIGN(iosize, blocksize);
 		if (this_bio_flag & EXTENT_BIO_COMPRESSED) {
 			disk_io_size = em->block_len;
 			sector = em->block_start >> 9;
@@ -2092,25 +2900,29 @@ static int __extent_read_full_page(struct extent_io_tree *tree,
 		/* we've found a hole, just zero and go on */
 		if (block_start == EXTENT_MAP_HOLE) {
 			char *userpage;
-			userpage = kmap_atomic(page, KM_USER0);
-			memset(userpage + page_offset, 0, iosize);
+			struct extent_state *cached = NULL;
+
+			userpage = kmap_atomic(page);
+			memset(userpage + pg_offset, 0, iosize);
 			flush_dcache_page(page);
-			kunmap_atomic(userpage, KM_USER0);
+			kunmap_atomic(userpage);
 
 			set_extent_uptodate(tree, cur, cur + iosize - 1,
-					    GFP_NOFS);
-			unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS);
+					    &cached, GFP_NOFS);
+			unlock_extent_cached(tree, cur, cur + iosize - 1,
+			                     &cached, GFP_NOFS);
 			cur = cur + iosize;
-			page_offset += iosize;
+			pg_offset += iosize;
 			continue;
 		}
 		/* the get_extent function already copied into the page */
 		if (test_range_bit(tree, cur, cur_end,
 				   EXTENT_UPTODATE, 1, NULL)) {
 			check_page_uptodate(tree, page);
-			unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS);
+			if (!parent_locked)
+				unlock_extent(tree, cur, cur + iosize - 1);
 			cur = cur + iosize;
-			page_offset += iosize;
+			pg_offset += iosize;
 			continue;
 		}
 		/* we have an inline extent but it didn't get marked up
@@ -2118,34 +2930,32 @@ static int __extent_read_full_page(struct extent_io_tree *tree,
 		 */
 		if (block_start == EXTENT_MAP_INLINE) {
 			SetPageError(page);
-			unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS);
+			if (!parent_locked)
+				unlock_extent(tree, cur, cur + iosize - 1);
 			cur = cur + iosize;
-			page_offset += iosize;
+			pg_offset += iosize;
 			continue;
 		}
 
-		ret = 0;
-		if (tree->ops && tree->ops->readpage_io_hook) {
-			ret = tree->ops->readpage_io_hook(page, cur,
-							  cur + iosize - 1);
-		}
-		if (!ret) {
-			unsigned long pnr = (last_byte >> PAGE_CACHE_SHIFT) + 1;
-			pnr -= page->index;
-			ret = submit_extent_page(READ, tree, page,
-					 sector, disk_io_size, page_offset,
+		pnr -= page->index;
+		ret = submit_extent_page(rw, tree, page,
+					 sector, disk_io_size, pg_offset,
 					 bdev, bio, pnr,
 					 end_bio_extent_readpage, mirror_num,
 					 *bio_flags,
 					 this_bio_flag);
+		if (!ret) {
 			nr++;
 			*bio_flags = this_bio_flag;
-		}
-		if (ret)
+		} else {
 			SetPageError(page);
+			if (!parent_locked)
+				unlock_extent(tree, cur, cur + iosize - 1);
+		}
 		cur = cur + iosize;
-		page_offset += iosize;
+		pg_offset += iosize;
 	}
+out:
 	if (!nr) {
 		if (!PageError(page))
 			SetPageUptodate(page);
@@ -2154,17 +2964,129 @@ static int __extent_read_full_page(struct extent_io_tree *tree,
 	return 0;
 }
 
+static inline void __do_contiguous_readpages(struct extent_io_tree *tree,
+					     struct page *pages[], int nr_pages,
+					     u64 start, u64 end,
+					     get_extent_t *get_extent,
+					     struct extent_map **em_cached,
+					     struct bio **bio, int mirror_num,
+					     unsigned long *bio_flags, int rw)
+{
+	struct inode *inode;
+	struct btrfs_ordered_extent *ordered;
+	int index;
+
+	inode = pages[0]->mapping->host;
+	while (1) {
+		lock_extent(tree, start, end);
+		ordered = btrfs_lookup_ordered_range(inode, start,
+						     end - start + 1);
+		if (!ordered)
+			break;
+		unlock_extent(tree, start, end);
+		btrfs_start_ordered_extent(inode, ordered, 1);
+		btrfs_put_ordered_extent(ordered);
+	}
+
+	for (index = 0; index < nr_pages; index++) {
+		__do_readpage(tree, pages[index], get_extent, em_cached, bio,
+			      mirror_num, bio_flags, rw);
+		page_cache_release(pages[index]);
+	}
+}
+
+static void __extent_readpages(struct extent_io_tree *tree,
+			       struct page *pages[],
+			       int nr_pages, get_extent_t *get_extent,
+			       struct extent_map **em_cached,
+			       struct bio **bio, int mirror_num,
+			       unsigned long *bio_flags, int rw)
+{
+	u64 start = 0;
+	u64 end = 0;
+	u64 page_start;
+	int index;
+	int first_index = 0;
+
+	for (index = 0; index < nr_pages; index++) {
+		page_start = page_offset(pages[index]);
+		if (!end) {
+			start = page_start;
+			end = start + PAGE_CACHE_SIZE - 1;
+			first_index = index;
+		} else if (end + 1 == page_start) {
+			end += PAGE_CACHE_SIZE;
+		} else {
+			__do_contiguous_readpages(tree, &pages[first_index],
+						  index - first_index, start,
+						  end, get_extent, em_cached,
+						  bio, mirror_num, bio_flags,
+						  rw);
+			start = page_start;
+			end = start + PAGE_CACHE_SIZE - 1;
+			first_index = index;
+		}
+	}
+
+	if (end)
+		__do_contiguous_readpages(tree, &pages[first_index],
+					  index - first_index, start,
+					  end, get_extent, em_cached, bio,
+					  mirror_num, bio_flags, rw);
+}
+
+static int __extent_read_full_page(struct extent_io_tree *tree,
+				   struct page *page,
+				   get_extent_t *get_extent,
+				   struct bio **bio, int mirror_num,
+				   unsigned long *bio_flags, int rw)
+{
+	struct inode *inode = page->mapping->host;
+	struct btrfs_ordered_extent *ordered;
+	u64 start = page_offset(page);
+	u64 end = start + PAGE_CACHE_SIZE - 1;
+	int ret;
+
+	while (1) {
+		lock_extent(tree, start, end);
+		ordered = btrfs_lookup_ordered_extent(inode, start);
+		if (!ordered)
+			break;
+		unlock_extent(tree, start, end);
+		btrfs_start_ordered_extent(inode, ordered, 1);
+		btrfs_put_ordered_extent(ordered);
+	}
+
+	ret = __do_readpage(tree, page, get_extent, NULL, bio, mirror_num,
+			    bio_flags, rw);
+	return ret;
+}
+
 int extent_read_full_page(struct extent_io_tree *tree, struct page *page,
-			    get_extent_t *get_extent)
+			    get_extent_t *get_extent, int mirror_num)
 {
 	struct bio *bio = NULL;
 	unsigned long bio_flags = 0;
 	int ret;
 
-	ret = __extent_read_full_page(tree, page, get_extent, &bio, 0,
-				      &bio_flags);
+	ret = __extent_read_full_page(tree, page, get_extent, &bio, mirror_num,
+				      &bio_flags, READ);
 	if (bio)
-		submit_one_bio(READ, bio, 0, bio_flags);
+		ret = submit_one_bio(READ, bio, mirror_num, bio_flags);
+	return ret;
+}
+
+int extent_read_full_page_nolock(struct extent_io_tree *tree, struct page *page,
+				 get_extent_t *get_extent, int mirror_num)
+{
+	struct bio *bio = NULL;
+	unsigned long bio_flags = EXTENT_BIO_PARENT_LOCKED;
+	int ret;
+
+	ret = __do_readpage(tree, page, get_extent, NULL, &bio, mirror_num,
+				      &bio_flags, READ);
+	if (bio)
+		ret = submit_one_bio(READ, bio, mirror_num, bio_flags);
 	return ret;
 }
 
@@ -2179,136 +3101,143 @@ static noinline void update_nr_written(struct page *page,
 }
 
 /*
- * the writepage semantics are similar to regular writepage.  extent
- * records are inserted to lock ranges in the tree, and as dirty areas
- * are found, they are marked writeback.  Then the lock bits are removed
- * and the end_io handler clears the writeback ranges
+ * helper for __extent_writepage, doing all of the delayed allocation setup.
+ *
+ * This returns 1 if our fill_delalloc function did all the work required
+ * to write the page (copy into inline extent).  In this case the IO has
+ * been started and the page is already unlocked.
+ *
+ * This returns 0 if all went well (page still locked)
+ * This returns < 0 if there were errors (page still locked)
  */
-static int __extent_writepage(struct page *page, struct writeback_control *wbc,
-			      void *data)
+static noinline_for_stack int writepage_delalloc(struct inode *inode,
+			      struct page *page, struct writeback_control *wbc,
+			      struct extent_page_data *epd,
+			      u64 delalloc_start,
+			      unsigned long *nr_written)
 {
-	struct inode *inode = page->mapping->host;
-	struct extent_page_data *epd = data;
 	struct extent_io_tree *tree = epd->tree;
-	u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
-	u64 delalloc_start;
+	u64 page_end = delalloc_start + PAGE_CACHE_SIZE - 1;
+	u64 nr_delalloc;
+	u64 delalloc_to_write = 0;
+	u64 delalloc_end = 0;
+	int ret;
+	int page_started = 0;
+
+	if (epd->extent_locked || !tree->ops || !tree->ops->fill_delalloc)
+		return 0;
+
+	while (delalloc_end < page_end) {
+		nr_delalloc = find_lock_delalloc_range(inode, tree,
+					       page,
+					       &delalloc_start,
+					       &delalloc_end,
+					       128 * 1024 * 1024);
+		if (nr_delalloc == 0) {
+			delalloc_start = delalloc_end + 1;
+			continue;
+		}
+		ret = tree->ops->fill_delalloc(inode, page,
+					       delalloc_start,
+					       delalloc_end,
+					       &page_started,
+					       nr_written);
+		/* File system has been set read-only */
+		if (ret) {
+			SetPageError(page);
+			/* fill_delalloc should be return < 0 for error
+			 * but just in case, we use > 0 here meaning the
+			 * IO is started, so we don't want to return > 0
+			 * unless things are going well.
+			 */
+			ret = ret < 0 ? ret : -EIO;
+			goto done;
+		}
+		/*
+		 * delalloc_end is already one less than the total
+		 * length, so we don't subtract one from
+		 * PAGE_CACHE_SIZE
+		 */
+		delalloc_to_write += (delalloc_end - delalloc_start +
+				      PAGE_CACHE_SIZE) >>
+				      PAGE_CACHE_SHIFT;
+		delalloc_start = delalloc_end + 1;
+	}
+	if (wbc->nr_to_write < delalloc_to_write) {
+		int thresh = 8192;
+
+		if (delalloc_to_write < thresh * 2)
+			thresh = delalloc_to_write;
+		wbc->nr_to_write = min_t(u64, delalloc_to_write,
+					 thresh);
+	}
+
+	/* did the fill delalloc function already unlock and start
+	 * the IO?
+	 */
+	if (page_started) {
+		/*
+		 * we've unlocked the page, so we can't update
+		 * the mapping's writeback index, just update
+		 * nr_to_write.
+		 */
+		wbc->nr_to_write -= *nr_written;
+		return 1;
+	}
+
+	ret = 0;
+
+done:
+	return ret;
+}
+
+/*
+ * helper for __extent_writepage.  This calls the writepage start hooks,
+ * and does the loop to map the page into extents and bios.
+ *
+ * We return 1 if the IO is started and the page is unlocked,
+ * 0 if all went well (page still locked)
+ * < 0 if there were errors (page still locked)
+ */
+static noinline_for_stack int __extent_writepage_io(struct inode *inode,
+				 struct page *page,
+				 struct writeback_control *wbc,
+				 struct extent_page_data *epd,
+				 loff_t i_size,
+				 unsigned long nr_written,
+				 int write_flags, int *nr_ret)
+{
+	struct extent_io_tree *tree = epd->tree;
+	u64 start = page_offset(page);
 	u64 page_end = start + PAGE_CACHE_SIZE - 1;
 	u64 end;
 	u64 cur = start;
 	u64 extent_offset;
-	u64 last_byte = i_size_read(inode);
 	u64 block_start;
 	u64 iosize;
-	u64 unlock_start;
 	sector_t sector;
 	struct extent_state *cached_state = NULL;
 	struct extent_map *em;
 	struct block_device *bdev;
-	int ret;
-	int nr = 0;
 	size_t pg_offset = 0;
 	size_t blocksize;
-	loff_t i_size = i_size_read(inode);
-	unsigned long end_index = i_size >> PAGE_CACHE_SHIFT;
-	u64 nr_delalloc;
-	u64 delalloc_end;
-	int page_started;
-	int compressed;
-	int write_flags;
-	unsigned long nr_written = 0;
-
-	if (wbc->sync_mode == WB_SYNC_ALL)
-		write_flags = WRITE_SYNC_PLUG;
-	else
-		write_flags = WRITE;
-
-	WARN_ON(!PageLocked(page));
-	pg_offset = i_size & (PAGE_CACHE_SIZE - 1);
-	if (page->index > end_index ||
-	   (page->index == end_index && !pg_offset)) {
-		page->mapping->a_ops->invalidatepage(page, 0);
-		unlock_page(page);
-		return 0;
-	}
-
-	if (page->index == end_index) {
-		char *userpage;
-
-		userpage = kmap_atomic(page, KM_USER0);
-		memset(userpage + pg_offset, 0,
-		       PAGE_CACHE_SIZE - pg_offset);
-		kunmap_atomic(userpage, KM_USER0);
-		flush_dcache_page(page);
-	}
-	pg_offset = 0;
-
-	set_page_extent_mapped(page);
-
-	delalloc_start = start;
-	delalloc_end = 0;
-	page_started = 0;
-	if (!epd->extent_locked) {
-		u64 delalloc_to_write = 0;
-		/*
-		 * make sure the wbc mapping index is at least updated
-		 * to this page.
-		 */
-		update_nr_written(page, wbc, 0);
-
-		while (delalloc_end < page_end) {
-			nr_delalloc = find_lock_delalloc_range(inode, tree,
-						       page,
-						       &delalloc_start,
-						       &delalloc_end,
-						       128 * 1024 * 1024);
-			if (nr_delalloc == 0) {
-				delalloc_start = delalloc_end + 1;
-				continue;
-			}
-			tree->ops->fill_delalloc(inode, page, delalloc_start,
-						 delalloc_end, &page_started,
-						 &nr_written);
-			/*
-			 * delalloc_end is already one less than the total
-			 * length, so we don't subtract one from
-			 * PAGE_CACHE_SIZE
-			 */
-			delalloc_to_write += (delalloc_end - delalloc_start +
-					      PAGE_CACHE_SIZE) >>
-					      PAGE_CACHE_SHIFT;
-			delalloc_start = delalloc_end + 1;
-		}
-		if (wbc->nr_to_write < delalloc_to_write) {
-			int thresh = 8192;
-
-			if (delalloc_to_write < thresh * 2)
-				thresh = delalloc_to_write;
-			wbc->nr_to_write = min_t(u64, delalloc_to_write,
-						 thresh);
-		}
+	int ret = 0;
+	int nr = 0;
+	bool compressed;
 
-		/* did the fill delalloc function already unlock and start
-		 * the IO?
-		 */
-		if (page_started) {
-			ret = 0;
-			/*
-			 * we've unlocked the page, so we can't update
-			 * the mapping's writeback index, just update
-			 * nr_to_write.
-			 */
-			wbc->nr_to_write -= nr_written;
-			goto done_unlocked;
-		}
-	}
 	if (tree->ops && tree->ops->writepage_start_hook) {
 		ret = tree->ops->writepage_start_hook(page, start,
 						      page_end);
-		if (ret == -EAGAIN) {
-			redirty_page_for_writepage(wbc, page);
+		if (ret) {
+			/* Fixup worker will requeue */
+			if (ret == -EBUSY)
+				wbc->pages_skipped++;
+			else
+				redirty_page_for_writepage(wbc, page);
+
 			update_nr_written(page, wbc, nr_written);
 			unlock_page(page);
-			ret = 0;
+			ret = 1;
 			goto done_unlocked;
 		}
 	}
@@ -2320,36 +3249,37 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc,
 	update_nr_written(page, wbc, nr_written + 1);
 
 	end = page_end;
-	if (last_byte <= start) {
+	if (i_size <= start) {
 		if (tree->ops && tree->ops->writepage_end_io_hook)
 			tree->ops->writepage_end_io_hook(page, start,
 							 page_end, NULL, 1);
-		unlock_start = page_end + 1;
 		goto done;
 	}
 
 	blocksize = inode->i_sb->s_blocksize;
 
 	while (cur <= end) {
-		if (cur >= last_byte) {
+		u64 em_end;
+		if (cur >= i_size) {
 			if (tree->ops && tree->ops->writepage_end_io_hook)
 				tree->ops->writepage_end_io_hook(page, cur,
 							 page_end, NULL, 1);
-			unlock_start = page_end + 1;
 			break;
 		}
 		em = epd->get_extent(inode, page, pg_offset, cur,
 				     end - cur + 1, 1);
-		if (IS_ERR(em) || !em) {
+		if (IS_ERR_OR_NULL(em)) {
 			SetPageError(page);
+			ret = PTR_ERR_OR_ZERO(em);
 			break;
 		}
 
 		extent_offset = cur - em->start;
-		BUG_ON(extent_map_end(em) <= cur);
+		em_end = extent_map_end(em);
+		BUG_ON(em_end <= cur);
 		BUG_ON(end < cur);
-		iosize = min(extent_map_end(em) - cur, end - cur + 1);
-		iosize = (iosize + blocksize - 1) & ~((u64)blocksize - 1);
+		iosize = min(em_end - cur, end - cur + 1);
+		iosize = ALIGN(iosize, blocksize);
 		sector = (em->block_start + extent_offset) >> 9;
 		bdev = em->bdev;
 		block_start = em->block_start;
@@ -2382,14 +3312,6 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc,
 
 			cur += iosize;
 			pg_offset += iosize;
-			unlock_start = cur;
-			continue;
-		}
-		/* leave this out until we have a page_mkwrite call */
-		if (0 && !test_range_bit(tree, cur, cur + iosize - 1,
-				   EXTENT_DIRTY, 0, NULL)) {
-			cur = cur + iosize;
-			pg_offset += iosize;
 			continue;
 		}
 
@@ -2402,14 +3324,13 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc,
 		if (ret) {
 			SetPageError(page);
 		} else {
-			unsigned long max_nr = end_index + 1;
+			unsigned long max_nr = (i_size >> PAGE_CACHE_SHIFT) + 1;
 
 			set_range_writeback(tree, cur, cur + iosize - 1);
 			if (!PageWriteback(page)) {
-				printk(KERN_ERR "btrfs warning page %lu not "
-				       "writeback, cur %llu end %llu\n",
-				       page->index, (unsigned long long)cur,
-				       (unsigned long long)end);
+				btrfs_err(BTRFS_I(inode)->root->fs_info,
+					   "page %lu not writeback, cur %llu end %llu",
+				       page->index, cur, end);
 			}
 
 			ret = submit_extent_page(write_flags, tree, page,
@@ -2425,20 +3346,391 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc,
 		nr++;
 	}
 done:
+	*nr_ret = nr;
+
+done_unlocked:
+
+	/* drop our reference on any cached states */
+	free_extent_state(cached_state);
+	return ret;
+}
+
+/*
+ * the writepage semantics are similar to regular writepage.  extent
+ * records are inserted to lock ranges in the tree, and as dirty areas
+ * are found, they are marked writeback.  Then the lock bits are removed
+ * and the end_io handler clears the writeback ranges
+ */
+static int __extent_writepage(struct page *page, struct writeback_control *wbc,
+			      void *data)
+{
+	struct inode *inode = page->mapping->host;
+	struct extent_page_data *epd = data;
+	u64 start = page_offset(page);
+	u64 page_end = start + PAGE_CACHE_SIZE - 1;
+	int ret;
+	int nr = 0;
+	size_t pg_offset = 0;
+	loff_t i_size = i_size_read(inode);
+	unsigned long end_index = i_size >> PAGE_CACHE_SHIFT;
+	int write_flags;
+	unsigned long nr_written = 0;
+
+	if (wbc->sync_mode == WB_SYNC_ALL)
+		write_flags = WRITE_SYNC;
+	else
+		write_flags = WRITE;
+
+	trace___extent_writepage(page, inode, wbc);
+
+	WARN_ON(!PageLocked(page));
+
+	ClearPageError(page);
+
+	pg_offset = i_size & (PAGE_CACHE_SIZE - 1);
+	if (page->index > end_index ||
+	   (page->index == end_index && !pg_offset)) {
+		page->mapping->a_ops->invalidatepage(page, 0, PAGE_CACHE_SIZE);
+		unlock_page(page);
+		return 0;
+	}
+
+	if (page->index == end_index) {
+		char *userpage;
+
+		userpage = kmap_atomic(page);
+		memset(userpage + pg_offset, 0,
+		       PAGE_CACHE_SIZE - pg_offset);
+		kunmap_atomic(userpage);
+		flush_dcache_page(page);
+	}
+
+	pg_offset = 0;
+
+	set_page_extent_mapped(page);
+
+	ret = writepage_delalloc(inode, page, wbc, epd, start, &nr_written);
+	if (ret == 1)
+		goto done_unlocked;
+	if (ret)
+		goto done;
+
+	ret = __extent_writepage_io(inode, page, wbc, epd,
+				    i_size, nr_written, write_flags, &nr);
+	if (ret == 1)
+		goto done_unlocked;
+
+done:
 	if (nr == 0) {
 		/* make sure the mapping tag for page dirty gets cleared */
 		set_page_writeback(page);
 		end_page_writeback(page);
 	}
+	if (PageError(page)) {
+		ret = ret < 0 ? ret : -EIO;
+		end_extent_writepage(page, ret, start, page_end);
+	}
 	unlock_page(page);
+	return ret;
 
 done_unlocked:
+	return 0;
+}
 
-	/* drop our reference on any cached states */
-	free_extent_state(cached_state);
+static int eb_wait(void *word)
+{
+	io_schedule();
 	return 0;
 }
 
+void wait_on_extent_buffer_writeback(struct extent_buffer *eb)
+{
+	wait_on_bit(&eb->bflags, EXTENT_BUFFER_WRITEBACK, eb_wait,
+		    TASK_UNINTERRUPTIBLE);
+}
+
+static noinline_for_stack int
+lock_extent_buffer_for_io(struct extent_buffer *eb,
+			  struct btrfs_fs_info *fs_info,
+			  struct extent_page_data *epd)
+{
+	unsigned long i, num_pages;
+	int flush = 0;
+	int ret = 0;
+
+	if (!btrfs_try_tree_write_lock(eb)) {
+		flush = 1;
+		flush_write_bio(epd);
+		btrfs_tree_lock(eb);
+	}
+
+	if (test_bit(EXTENT_BUFFER_WRITEBACK, &eb->bflags)) {
+		btrfs_tree_unlock(eb);
+		if (!epd->sync_io)
+			return 0;
+		if (!flush) {
+			flush_write_bio(epd);
+			flush = 1;
+		}
+		while (1) {
+			wait_on_extent_buffer_writeback(eb);
+			btrfs_tree_lock(eb);
+			if (!test_bit(EXTENT_BUFFER_WRITEBACK, &eb->bflags))
+				break;
+			btrfs_tree_unlock(eb);
+		}
+	}
+
+	/*
+	 * We need to do this to prevent races in people who check if the eb is
+	 * under IO since we can end up having no IO bits set for a short period
+	 * of time.
+	 */
+	spin_lock(&eb->refs_lock);
+	if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) {
+		set_bit(EXTENT_BUFFER_WRITEBACK, &eb->bflags);
+		spin_unlock(&eb->refs_lock);
+		btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
+		__percpu_counter_add(&fs_info->dirty_metadata_bytes,
+				     -eb->len,
+				     fs_info->dirty_metadata_batch);
+		ret = 1;
+	} else {
+		spin_unlock(&eb->refs_lock);
+	}
+
+	btrfs_tree_unlock(eb);
+
+	if (!ret)
+		return ret;
+
+	num_pages = num_extent_pages(eb->start, eb->len);
+	for (i = 0; i < num_pages; i++) {
+		struct page *p = extent_buffer_page(eb, i);
+
+		if (!trylock_page(p)) {
+			if (!flush) {
+				flush_write_bio(epd);
+				flush = 1;
+			}
+			lock_page(p);
+		}
+	}
+
+	return ret;
+}
+
+static void end_extent_buffer_writeback(struct extent_buffer *eb)
+{
+	clear_bit(EXTENT_BUFFER_WRITEBACK, &eb->bflags);
+	smp_mb__after_atomic();
+	wake_up_bit(&eb->bflags, EXTENT_BUFFER_WRITEBACK);
+}
+
+static void end_bio_extent_buffer_writepage(struct bio *bio, int err)
+{
+	struct bio_vec *bvec;
+	struct extent_buffer *eb;
+	int i, done;
+
+	bio_for_each_segment_all(bvec, bio, i) {
+		struct page *page = bvec->bv_page;
+
+		eb = (struct extent_buffer *)page->private;
+		BUG_ON(!eb);
+		done = atomic_dec_and_test(&eb->io_pages);
+
+		if (err || test_bit(EXTENT_BUFFER_IOERR, &eb->bflags)) {
+			set_bit(EXTENT_BUFFER_IOERR, &eb->bflags);
+			ClearPageUptodate(page);
+			SetPageError(page);
+		}
+
+		end_page_writeback(page);
+
+		if (!done)
+			continue;
+
+		end_extent_buffer_writeback(eb);
+	}
+
+	bio_put(bio);
+}
+
+static noinline_for_stack int write_one_eb(struct extent_buffer *eb,
+			struct btrfs_fs_info *fs_info,
+			struct writeback_control *wbc,
+			struct extent_page_data *epd)
+{
+	struct block_device *bdev = fs_info->fs_devices->latest_bdev;
+	struct extent_io_tree *tree = &BTRFS_I(fs_info->btree_inode)->io_tree;
+	u64 offset = eb->start;
+	unsigned long i, num_pages;
+	unsigned long bio_flags = 0;
+	int rw = (epd->sync_io ? WRITE_SYNC : WRITE) | REQ_META;
+	int ret = 0;
+
+	clear_bit(EXTENT_BUFFER_IOERR, &eb->bflags);
+	num_pages = num_extent_pages(eb->start, eb->len);
+	atomic_set(&eb->io_pages, num_pages);
+	if (btrfs_header_owner(eb) == BTRFS_TREE_LOG_OBJECTID)
+		bio_flags = EXTENT_BIO_TREE_LOG;
+
+	for (i = 0; i < num_pages; i++) {
+		struct page *p = extent_buffer_page(eb, i);
+
+		clear_page_dirty_for_io(p);
+		set_page_writeback(p);
+		ret = submit_extent_page(rw, tree, p, offset >> 9,
+					 PAGE_CACHE_SIZE, 0, bdev, &epd->bio,
+					 -1, end_bio_extent_buffer_writepage,
+					 0, epd->bio_flags, bio_flags);
+		epd->bio_flags = bio_flags;
+		if (ret) {
+			set_bit(EXTENT_BUFFER_IOERR, &eb->bflags);
+			SetPageError(p);
+			if (atomic_sub_and_test(num_pages - i, &eb->io_pages))
+				end_extent_buffer_writeback(eb);
+			ret = -EIO;
+			break;
+		}
+		offset += PAGE_CACHE_SIZE;
+		update_nr_written(p, wbc, 1);
+		unlock_page(p);
+	}
+
+	if (unlikely(ret)) {
+		for (; i < num_pages; i++) {
+			struct page *p = extent_buffer_page(eb, i);
+			unlock_page(p);
+		}
+	}
+
+	return ret;
+}
+
+int btree_write_cache_pages(struct address_space *mapping,
+				   struct writeback_control *wbc)
+{
+	struct extent_io_tree *tree = &BTRFS_I(mapping->host)->io_tree;
+	struct btrfs_fs_info *fs_info = BTRFS_I(mapping->host)->root->fs_info;
+	struct extent_buffer *eb, *prev_eb = NULL;
+	struct extent_page_data epd = {
+		.bio = NULL,
+		.tree = tree,
+		.extent_locked = 0,
+		.sync_io = wbc->sync_mode == WB_SYNC_ALL,
+		.bio_flags = 0,
+	};
+	int ret = 0;
+	int done = 0;
+	int nr_to_write_done = 0;
+	struct pagevec pvec;
+	int nr_pages;
+	pgoff_t index;
+	pgoff_t end;		/* Inclusive */
+	int scanned = 0;
+	int tag;
+
+	pagevec_init(&pvec, 0);
+	if (wbc->range_cyclic) {
+		index = mapping->writeback_index; /* Start from prev offset */
+		end = -1;
+	} else {
+		index = wbc->range_start >> PAGE_CACHE_SHIFT;
+		end = wbc->range_end >> PAGE_CACHE_SHIFT;
+		scanned = 1;
+	}
+	if (wbc->sync_mode == WB_SYNC_ALL)
+		tag = PAGECACHE_TAG_TOWRITE;
+	else
+		tag = PAGECACHE_TAG_DIRTY;
+retry:
+	if (wbc->sync_mode == WB_SYNC_ALL)
+		tag_pages_for_writeback(mapping, index, end);
+	while (!done && !nr_to_write_done && (index <= end) &&
+	       (nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, tag,
+			min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1))) {
+		unsigned i;
+
+		scanned = 1;
+		for (i = 0; i < nr_pages; i++) {
+			struct page *page = pvec.pages[i];
+
+			if (!PagePrivate(page))
+				continue;
+
+			if (!wbc->range_cyclic && page->index > end) {
+				done = 1;
+				break;
+			}
+
+			spin_lock(&mapping->private_lock);
+			if (!PagePrivate(page)) {
+				spin_unlock(&mapping->private_lock);
+				continue;
+			}
+
+			eb = (struct extent_buffer *)page->private;
+
+			/*
+			 * Shouldn't happen and normally this would be a BUG_ON
+			 * but no sense in crashing the users box for something
+			 * we can survive anyway.
+			 */
+			if (WARN_ON(!eb)) {
+				spin_unlock(&mapping->private_lock);
+				continue;
+			}
+
+			if (eb == prev_eb) {
+				spin_unlock(&mapping->private_lock);
+				continue;
+			}
+
+			ret = atomic_inc_not_zero(&eb->refs);
+			spin_unlock(&mapping->private_lock);
+			if (!ret)
+				continue;
+
+			prev_eb = eb;
+			ret = lock_extent_buffer_for_io(eb, fs_info, &epd);
+			if (!ret) {
+				free_extent_buffer(eb);
+				continue;
+			}
+
+			ret = write_one_eb(eb, fs_info, wbc, &epd);
+			if (ret) {
+				done = 1;
+				free_extent_buffer(eb);
+				break;
+			}
+			free_extent_buffer(eb);
+
+			/*
+			 * the filesystem may choose to bump up nr_to_write.
+			 * We have to make sure to honor the new nr_to_write
+			 * at any time
+			 */
+			nr_to_write_done = wbc->nr_to_write <= 0;
+		}
+		pagevec_release(&pvec);
+		cond_resched();
+	}
+	if (!scanned && !done) {
+		/*
+		 * We hit the last page and there is more work to be done: wrap
+		 * back to the start of the file
+		 */
+		scanned = 1;
+		index = 0;
+		goto retry;
+	}
+	flush_write_bio(&epd);
+	return ret;
+}
+
 /**
  * write_cache_pages - walk the list of dirty pages of the given address space and write all of them.
  * @mapping: address space structure to write
@@ -2460,15 +3752,29 @@ static int extent_write_cache_pages(struct extent_io_tree *tree,
 			     writepage_t writepage, void *data,
 			     void (*flush_fn)(void *))
 {
+	struct inode *inode = mapping->host;
 	int ret = 0;
 	int done = 0;
+	int err = 0;
 	int nr_to_write_done = 0;
 	struct pagevec pvec;
 	int nr_pages;
 	pgoff_t index;
 	pgoff_t end;		/* Inclusive */
 	int scanned = 0;
-	int range_whole = 0;
+	int tag;
+
+	/*
+	 * We have to hold onto the inode so that ordered extents can do their
+	 * work when the IO finishes.  The alternative to this is failing to add
+	 * an ordered extent if the igrab() fails there and that is a huge pain
+	 * to deal with, so instead just hold onto the inode throughout the
+	 * writepages operation.  If it fails here we are freeing up the inode
+	 * anyway and we'd rather not waste our time writing out stuff that is
+	 * going to be truncated anyway.
+	 */
+	if (!igrab(inode))
+		return 0;
 
 	pagevec_init(&pvec, 0);
 	if (wbc->range_cyclic) {
@@ -2477,15 +3783,18 @@ static int extent_write_cache_pages(struct extent_io_tree *tree,
 	} else {
 		index = wbc->range_start >> PAGE_CACHE_SHIFT;
 		end = wbc->range_end >> PAGE_CACHE_SHIFT;
-		if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX)
-			range_whole = 1;
 		scanned = 1;
 	}
+	if (wbc->sync_mode == WB_SYNC_ALL)
+		tag = PAGECACHE_TAG_TOWRITE;
+	else
+		tag = PAGECACHE_TAG_DIRTY;
 retry:
+	if (wbc->sync_mode == WB_SYNC_ALL)
+		tag_pages_for_writeback(mapping, index, end);
 	while (!done && !nr_to_write_done && (index <= end) &&
-	       (nr_pages = pagevec_lookup_tag(&pvec, mapping, &index,
-			      PAGECACHE_TAG_DIRTY, min(end - index,
-				  (pgoff_t)PAGEVEC_SIZE-1) + 1))) {
+	       (nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, tag,
+			min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1))) {
 		unsigned i;
 
 		scanned = 1;
@@ -2499,10 +3808,10 @@ retry:
 			 * swizzled back from swapper_space to tmpfs file
 			 * mapping
 			 */
-			if (tree->ops && tree->ops->write_cache_pages_lock_hook)
-				tree->ops->write_cache_pages_lock_hook(page);
-			else
+			if (!trylock_page(page)) {
+				flush_fn(data);
 				lock_page(page);
+			}
 
 			if (unlikely(page->mapping != mapping)) {
 				unlock_page(page);
@@ -2533,8 +3842,8 @@ retry:
 				unlock_page(page);
 				ret = 0;
 			}
-			if (ret)
-				done = 1;
+			if (!err && ret < 0)
+				err = ret;
 
 			/*
 			 * the filesystem may choose to bump up nr_to_write.
@@ -2546,7 +3855,7 @@ retry:
 		pagevec_release(&pvec);
 		cond_resched();
 	}
-	if (!scanned && !done) {
+	if (!scanned && !done && !err) {
 		/*
 		 * We hit the last page and there is more work to be done: wrap
 		 * back to the start of the file
@@ -2555,16 +3864,21 @@ retry:
 		index = 0;
 		goto retry;
 	}
-	return ret;
+	btrfs_add_delayed_iput(inode);
+	return err;
 }
 
 static void flush_epd_write_bio(struct extent_page_data *epd)
 {
 	if (epd->bio) {
+		int rw = WRITE;
+		int ret;
+
 		if (epd->sync_io)
-			submit_one_bio(WRITE_SYNC, epd->bio, 0, 0);
-		else
-			submit_one_bio(WRITE, epd->bio, 0, 0);
+			rw = WRITE_SYNC;
+
+		ret = submit_one_bio(rw, epd->bio, 0, epd->bio_flags);
+		BUG_ON(ret < 0); /* -ENOMEM */
 		epd->bio = NULL;
 	}
 }
@@ -2580,27 +3894,17 @@ int extent_write_full_page(struct extent_io_tree *tree, struct page *page,
 			  struct writeback_control *wbc)
 {
 	int ret;
-	struct address_space *mapping = page->mapping;
 	struct extent_page_data epd = {
 		.bio = NULL,
 		.tree = tree,
 		.get_extent = get_extent,
 		.extent_locked = 0,
 		.sync_io = wbc->sync_mode == WB_SYNC_ALL,
-	};
-	struct writeback_control wbc_writepages = {
-		.bdi		= wbc->bdi,
-		.sync_mode	= wbc->sync_mode,
-		.older_than_this = NULL,
-		.nr_to_write	= 64,
-		.range_start	= page_offset(page) + PAGE_CACHE_SIZE,
-		.range_end	= (loff_t)-1,
+		.bio_flags = 0,
 	};
 
 	ret = __extent_writepage(page, wbc, &epd);
 
-	extent_write_cache_pages(tree, mapping, &wbc_writepages,
-				 __extent_writepage, &epd, flush_write_bio);
 	flush_epd_write_bio(&epd);
 	return ret;
 }
@@ -2621,11 +3925,10 @@ int extent_write_locked_range(struct extent_io_tree *tree, struct inode *inode,
 		.get_extent = get_extent,
 		.extent_locked = 1,
 		.sync_io = mode == WB_SYNC_ALL,
+		.bio_flags = 0,
 	};
 	struct writeback_control wbc_writepages = {
-		.bdi		= inode->i_mapping->backing_dev_info,
 		.sync_mode	= mode,
-		.older_than_this = NULL,
 		.nr_to_write	= nr_pages * 2,
 		.range_start	= start,
 		.range_end	= end + 1,
@@ -2662,6 +3965,7 @@ int extent_writepages(struct extent_io_tree *tree,
 		.get_extent = get_extent,
 		.extent_locked = 0,
 		.sync_io = wbc->sync_mode == WB_SYNC_ALL,
+		.bio_flags = 0,
 	};
 
 	ret = extent_write_cache_pages(tree, mapping, wbc,
@@ -2679,22 +3983,39 @@ int extent_readpages(struct extent_io_tree *tree,
 	struct bio *bio = NULL;
 	unsigned page_idx;
 	unsigned long bio_flags = 0;
+	struct page *pagepool[16];
+	struct page *page;
+	struct extent_map *em_cached = NULL;
+	int nr = 0;
 
 	for (page_idx = 0; page_idx < nr_pages; page_idx++) {
-		struct page *page = list_entry(pages->prev, struct page, lru);
+		page = list_entry(pages->prev, struct page, lru);
 
 		prefetchw(&page->flags);
 		list_del(&page->lru);
-		if (!add_to_page_cache_lru(page, mapping,
-					page->index, GFP_KERNEL)) {
-			__extent_read_full_page(tree, page, get_extent,
-						&bio, 0, &bio_flags);
+		if (add_to_page_cache_lru(page, mapping,
+					page->index, GFP_NOFS)) {
+			page_cache_release(page);
+			continue;
 		}
-		page_cache_release(page);
+
+		pagepool[nr++] = page;
+		if (nr < ARRAY_SIZE(pagepool))
+			continue;
+		__extent_readpages(tree, pagepool, nr, get_extent, &em_cached,
+				   &bio, 0, &bio_flags, READ);
+		nr = 0;
 	}
+	if (nr)
+		__extent_readpages(tree, pagepool, nr, get_extent, &em_cached,
+				   &bio, 0, &bio_flags, READ);
+
+	if (em_cached)
+		free_extent_map(em_cached);
+
 	BUG_ON(!list_empty(pages));
 	if (bio)
-		submit_one_bio(READ, bio, 0, bio_flags);
+		return submit_one_bio(READ, bio, 0, bio_flags);
 	return 0;
 }
 
@@ -2707,15 +4028,15 @@ int extent_invalidatepage(struct extent_io_tree *tree,
 			  struct page *page, unsigned long offset)
 {
 	struct extent_state *cached_state = NULL;
-	u64 start = ((u64)page->index << PAGE_CACHE_SHIFT);
+	u64 start = page_offset(page);
 	u64 end = start + PAGE_CACHE_SIZE - 1;
 	size_t blocksize = page->mapping->host->i_sb->s_blocksize;
 
-	start += (offset + blocksize - 1) & ~(blocksize - 1);
+	start += ALIGN(offset, blocksize);
 	if (start > end)
 		return 0;
 
-	lock_extent_bits(tree, start, end, 0, &cached_state, GFP_NOFS);
+	lock_extent_bits(tree, start, end, 0, &cached_state);
 	wait_on_page_writeback(page);
 	clear_extent_bit(tree, start, end,
 			 EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC |
@@ -2725,132 +4046,15 @@ int extent_invalidatepage(struct extent_io_tree *tree,
 }
 
 /*
- * simple commit_write call, set_range_dirty is used to mark both
- * the pages and the extent records as dirty
- */
-int extent_commit_write(struct extent_io_tree *tree,
-			struct inode *inode, struct page *page,
-			unsigned from, unsigned to)
-{
-	loff_t pos = ((loff_t)page->index << PAGE_CACHE_SHIFT) + to;
-
-	set_page_extent_mapped(page);
-	set_page_dirty(page);
-
-	if (pos > inode->i_size) {
-		i_size_write(inode, pos);
-		mark_inode_dirty(inode);
-	}
-	return 0;
-}
-
-int extent_prepare_write(struct extent_io_tree *tree,
-			 struct inode *inode, struct page *page,
-			 unsigned from, unsigned to, get_extent_t *get_extent)
-{
-	u64 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
-	u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
-	u64 block_start;
-	u64 orig_block_start;
-	u64 block_end;
-	u64 cur_end;
-	struct extent_map *em;
-	unsigned blocksize = 1 << inode->i_blkbits;
-	size_t page_offset = 0;
-	size_t block_off_start;
-	size_t block_off_end;
-	int err = 0;
-	int iocount = 0;
-	int ret = 0;
-	int isnew;
-
-	set_page_extent_mapped(page);
-
-	block_start = (page_start + from) & ~((u64)blocksize - 1);
-	block_end = (page_start + to - 1) | (blocksize - 1);
-	orig_block_start = block_start;
-
-	lock_extent(tree, page_start, page_end, GFP_NOFS);
-	while (block_start <= block_end) {
-		em = get_extent(inode, page, page_offset, block_start,
-				block_end - block_start + 1, 1);
-		if (IS_ERR(em) || !em)
-			goto err;
-
-		cur_end = min(block_end, extent_map_end(em) - 1);
-		block_off_start = block_start & (PAGE_CACHE_SIZE - 1);
-		block_off_end = block_off_start + blocksize;
-		isnew = clear_extent_new(tree, block_start, cur_end, GFP_NOFS);
-
-		if (!PageUptodate(page) && isnew &&
-		    (block_off_end > to || block_off_start < from)) {
-			void *kaddr;
-
-			kaddr = kmap_atomic(page, KM_USER0);
-			if (block_off_end > to)
-				memset(kaddr + to, 0, block_off_end - to);
-			if (block_off_start < from)
-				memset(kaddr + block_off_start, 0,
-				       from - block_off_start);
-			flush_dcache_page(page);
-			kunmap_atomic(kaddr, KM_USER0);
-		}
-		if ((em->block_start != EXTENT_MAP_HOLE &&
-		     em->block_start != EXTENT_MAP_INLINE) &&
-		    !isnew && !PageUptodate(page) &&
-		    (block_off_end > to || block_off_start < from) &&
-		    !test_range_bit(tree, block_start, cur_end,
-				    EXTENT_UPTODATE, 1, NULL)) {
-			u64 sector;
-			u64 extent_offset = block_start - em->start;
-			size_t iosize;
-			sector = (em->block_start + extent_offset) >> 9;
-			iosize = (cur_end - block_start + blocksize) &
-				~((u64)blocksize - 1);
-			/*
-			 * we've already got the extent locked, but we
-			 * need to split the state such that our end_bio
-			 * handler can clear the lock.
-			 */
-			set_extent_bit(tree, block_start,
-				       block_start + iosize - 1,
-				       EXTENT_LOCKED, 0, NULL, NULL, GFP_NOFS);
-			ret = submit_extent_page(READ, tree, page,
-					 sector, iosize, page_offset, em->bdev,
-					 NULL, 1,
-					 end_bio_extent_preparewrite, 0,
-					 0, 0);
-			iocount++;
-			block_start = block_start + iosize;
-		} else {
-			set_extent_uptodate(tree, block_start, cur_end,
-					    GFP_NOFS);
-			unlock_extent(tree, block_start, cur_end, GFP_NOFS);
-			block_start = cur_end + 1;
-		}
-		page_offset = block_start & (PAGE_CACHE_SIZE - 1);
-		free_extent_map(em);
-	}
-	if (iocount) {
-		wait_extent_bit(tree, orig_block_start,
-				block_end, EXTENT_LOCKED);
-	}
-	check_page_uptodate(tree, page);
-err:
-	/* FIXME, zero out newly allocated blocks on error */
-	return err;
-}
-
-/*
  * a helper for releasepage, this tests for areas of the page that
  * are locked or under IO and drops the related state bits if it is safe
  * to drop the page.
  */
-int try_release_extent_state(struct extent_map_tree *map,
-			     struct extent_io_tree *tree, struct page *page,
-			     gfp_t mask)
+static int try_release_extent_state(struct extent_map_tree *map,
+				    struct extent_io_tree *tree,
+				    struct page *page, gfp_t mask)
 {
-	u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
+	u64 start = page_offset(page);
 	u64 end = start + PAGE_CACHE_SIZE - 1;
 	int ret = 1;
 
@@ -2864,9 +4068,17 @@ int try_release_extent_state(struct extent_map_tree *map,
 		 * at this point we can safely clear everything except the
 		 * locked bit and the nodatasum bit
 		 */
-		clear_extent_bit(tree, start, end,
+		ret = clear_extent_bit(tree, start, end,
 				 ~(EXTENT_LOCKED | EXTENT_NODATASUM),
 				 0, 0, NULL, mask);
+
+		/* if clear_extent_bit failed for enomem reasons,
+		 * we can't allow the release to continue.
+		 */
+		if (ret < 0)
+			ret = 0;
+		else
+			ret = 1;
 	}
 	return ret;
 }
@@ -2881,7 +4093,7 @@ int try_release_extent_mapping(struct extent_map_tree *map,
 			       gfp_t mask)
 {
 	struct extent_map *em;
-	u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
+	u64 start = page_offset(page);
 	u64 end = start + PAGE_CACHE_SIZE - 1;
 
 	if ((mask & __GFP_WAIT) &&
@@ -2891,7 +4103,7 @@ int try_release_extent_mapping(struct extent_map_tree *map,
 			len = end - start + 1;
 			write_lock(&map->lock);
 			em = lookup_extent_mapping(map, start, len);
-			if (!em || IS_ERR(em)) {
+			if (!em) {
 				write_unlock(&map->lock);
 				break;
 			}
@@ -2919,76 +4131,183 @@ int try_release_extent_mapping(struct extent_map_tree *map,
 	return try_release_extent_state(map, tree, page, mask);
 }
 
-sector_t extent_bmap(struct address_space *mapping, sector_t iblock,
-		get_extent_t *get_extent)
+/*
+ * helper function for fiemap, which doesn't want to see any holes.
+ * This maps until we find something past 'last'
+ */
+static struct extent_map *get_extent_skip_holes(struct inode *inode,
+						u64 offset,
+						u64 last,
+						get_extent_t *get_extent)
 {
-	struct inode *inode = mapping->host;
-	struct extent_state *cached_state = NULL;
-	u64 start = iblock << inode->i_blkbits;
-	sector_t sector = 0;
-	size_t blksize = (1 << inode->i_blkbits);
+	u64 sectorsize = BTRFS_I(inode)->root->sectorsize;
 	struct extent_map *em;
+	u64 len;
 
-	lock_extent_bits(&BTRFS_I(inode)->io_tree, start, start + blksize - 1,
-			 0, &cached_state, GFP_NOFS);
-	em = get_extent(inode, NULL, 0, start, blksize, 0);
-	unlock_extent_cached(&BTRFS_I(inode)->io_tree, start,
-			     start + blksize - 1, &cached_state, GFP_NOFS);
-	if (!em || IS_ERR(em))
-		return 0;
+	if (offset >= last)
+		return NULL;
 
-	if (em->block_start > EXTENT_MAP_LAST_BYTE)
-		goto out;
+	while (1) {
+		len = last - offset;
+		if (len == 0)
+			break;
+		len = ALIGN(len, sectorsize);
+		em = get_extent(inode, NULL, 0, offset, len, 0);
+		if (IS_ERR_OR_NULL(em))
+			return em;
+
+		/* if this isn't a hole return it */
+		if (!test_bit(EXTENT_FLAG_VACANCY, &em->flags) &&
+		    em->block_start != EXTENT_MAP_HOLE) {
+			return em;
+		}
 
-	sector = (em->block_start + start - em->start) >> inode->i_blkbits;
-out:
-	free_extent_map(em);
-	return sector;
+		/* this is a hole, advance to the next extent */
+		offset = extent_map_end(em);
+		free_extent_map(em);
+		if (offset >= last)
+			break;
+	}
+	return NULL;
+}
+
+static noinline int count_ext_ref(u64 inum, u64 offset, u64 root_id, void *ctx)
+{
+	unsigned long cnt = *((unsigned long *)ctx);
+
+	cnt++;
+	*((unsigned long *)ctx) = cnt;
+
+	/* Now we're sure that the extent is shared. */
+	if (cnt > 1)
+		return 1;
+	return 0;
 }
 
 int extent_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
 		__u64 start, __u64 len, get_extent_t *get_extent)
 {
-	int ret;
+	int ret = 0;
 	u64 off = start;
 	u64 max = start + len;
 	u32 flags = 0;
+	u32 found_type;
+	u64 last;
+	u64 last_for_get_extent = 0;
 	u64 disko = 0;
+	u64 isize = i_size_read(inode);
+	struct btrfs_key found_key;
 	struct extent_map *em = NULL;
 	struct extent_state *cached_state = NULL;
+	struct btrfs_path *path;
 	int end = 0;
-	u64 em_start = 0, em_len = 0;
-	unsigned long emflags;
-	ret = 0;
+	u64 em_start = 0;
+	u64 em_len = 0;
+	u64 em_end = 0;
 
 	if (len == 0)
 		return -EINVAL;
 
-	lock_extent_bits(&BTRFS_I(inode)->io_tree, start, start + len, 0,
-			 &cached_state, GFP_NOFS);
-	em = get_extent(inode, NULL, 0, off, max - off, 0);
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+	path->leave_spinning = 1;
+
+	start = ALIGN(start, BTRFS_I(inode)->root->sectorsize);
+	len = ALIGN(len, BTRFS_I(inode)->root->sectorsize);
+
+	/*
+	 * lookup the last file extent.  We're not using i_size here
+	 * because there might be preallocation past i_size
+	 */
+	ret = btrfs_lookup_file_extent(NULL, BTRFS_I(inode)->root,
+				       path, btrfs_ino(inode), -1, 0);
+	if (ret < 0) {
+		btrfs_free_path(path);
+		return ret;
+	}
+	WARN_ON(!ret);
+	path->slots[0]--;
+	btrfs_item_key_to_cpu(path->nodes[0], &found_key, path->slots[0]);
+	found_type = btrfs_key_type(&found_key);
+
+	/* No extents, but there might be delalloc bits */
+	if (found_key.objectid != btrfs_ino(inode) ||
+	    found_type != BTRFS_EXTENT_DATA_KEY) {
+		/* have to trust i_size as the end */
+		last = (u64)-1;
+		last_for_get_extent = isize;
+	} else {
+		/*
+		 * remember the start of the last extent.  There are a
+		 * bunch of different factors that go into the length of the
+		 * extent, so its much less complex to remember where it started
+		 */
+		last = found_key.offset;
+		last_for_get_extent = last + 1;
+	}
+	btrfs_release_path(path);
+
+	/*
+	 * we might have some extents allocated but more delalloc past those
+	 * extents.  so, we trust isize unless the start of the last extent is
+	 * beyond isize
+	 */
+	if (last < isize) {
+		last = (u64)-1;
+		last_for_get_extent = isize;
+	}
+
+	lock_extent_bits(&BTRFS_I(inode)->io_tree, start, start + len - 1, 0,
+			 &cached_state);
+
+	em = get_extent_skip_holes(inode, start, last_for_get_extent,
+				   get_extent);
 	if (!em)
 		goto out;
 	if (IS_ERR(em)) {
 		ret = PTR_ERR(em);
 		goto out;
 	}
+
 	while (!end) {
-		off = em->start + em->len;
-		if (off >= max)
-			end = 1;
+		u64 offset_in_extent = 0;
+
+		/* break if the extent we found is outside the range */
+		if (em->start >= max || extent_map_end(em) < off)
+			break;
 
-		em_start = em->start;
-		em_len = em->len;
+		/*
+		 * get_extent may return an extent that starts before our
+		 * requested range.  We have to make sure the ranges
+		 * we return to fiemap always move forward and don't
+		 * overlap, so adjust the offsets here
+		 */
+		em_start = max(em->start, off);
 
+		/*
+		 * record the offset from the start of the extent
+		 * for adjusting the disk offset below.  Only do this if the
+		 * extent isn't compressed since our in ram offset may be past
+		 * what we have actually allocated on disk.
+		 */
+		if (!test_bit(EXTENT_FLAG_COMPRESSED, &em->flags))
+			offset_in_extent = em_start - em->start;
+		em_end = extent_map_end(em);
+		em_len = em_end - em_start;
 		disko = 0;
 		flags = 0;
 
+		/*
+		 * bump off for our next call to get_extent
+		 */
+		off = extent_map_end(em);
+		if (off >= max)
+			end = 1;
+
 		if (em->block_start == EXTENT_MAP_LAST_BYTE) {
 			end = 1;
 			flags |= FIEMAP_EXTENT_LAST;
-		} else if (em->block_start == EXTENT_MAP_HOLE) {
-			flags |= FIEMAP_EXTENT_UNWRITTEN;
 		} else if (em->block_start == EXTENT_MAP_INLINE) {
 			flags |= (FIEMAP_EXTENT_DATA_INLINE |
 				  FIEMAP_EXTENT_NOT_ALIGNED);
@@ -2996,115 +4315,349 @@ int extent_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
 			flags |= (FIEMAP_EXTENT_DELALLOC |
 				  FIEMAP_EXTENT_UNKNOWN);
 		} else {
-			disko = em->block_start;
+			unsigned long ref_cnt = 0;
+
+			disko = em->block_start + offset_in_extent;
+
+			/*
+			 * As btrfs supports shared space, this information
+			 * can be exported to userspace tools via
+			 * flag FIEMAP_EXTENT_SHARED.
+			 */
+			ret = iterate_inodes_from_logical(
+					em->block_start,
+					BTRFS_I(inode)->root->fs_info,
+					path, count_ext_ref, &ref_cnt);
+			if (ret < 0 && ret != -ENOENT)
+				goto out_free;
+
+			if (ref_cnt > 1)
+				flags |= FIEMAP_EXTENT_SHARED;
 		}
 		if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags))
 			flags |= FIEMAP_EXTENT_ENCODED;
 
-		emflags = em->flags;
 		free_extent_map(em);
 		em = NULL;
+		if ((em_start >= last) || em_len == (u64)-1 ||
+		   (last == (u64)-1 && isize <= em_end)) {
+			flags |= FIEMAP_EXTENT_LAST;
+			end = 1;
+		}
 
-		if (!end) {
-			em = get_extent(inode, NULL, 0, off, max - off, 0);
-			if (!em)
-				goto out;
-			if (IS_ERR(em)) {
-				ret = PTR_ERR(em);
-				goto out;
-			}
-			emflags = em->flags;
+		/* now scan forward to see if this is really the last extent. */
+		em = get_extent_skip_holes(inode, off, last_for_get_extent,
+					   get_extent);
+		if (IS_ERR(em)) {
+			ret = PTR_ERR(em);
+			goto out;
 		}
-		if (test_bit(EXTENT_FLAG_VACANCY, &emflags)) {
+		if (!em) {
 			flags |= FIEMAP_EXTENT_LAST;
 			end = 1;
 		}
-
 		ret = fiemap_fill_next_extent(fieinfo, em_start, disko,
-					em_len, flags);
+					      em_len, flags);
 		if (ret)
 			goto out_free;
 	}
 out_free:
 	free_extent_map(em);
 out:
-	unlock_extent_cached(&BTRFS_I(inode)->io_tree, start, start + len,
+	btrfs_free_path(path);
+	unlock_extent_cached(&BTRFS_I(inode)->io_tree, start, start + len - 1,
 			     &cached_state, GFP_NOFS);
 	return ret;
 }
 
-static inline struct page *extent_buffer_page(struct extent_buffer *eb,
-					      unsigned long i)
+static void __free_extent_buffer(struct extent_buffer *eb)
 {
-	struct page *p;
-	struct address_space *mapping;
+	btrfs_leak_debug_del(&eb->leak_list);
+	kmem_cache_free(extent_buffer_cache, eb);
+}
 
-	if (i == 0)
-		return eb->first_page;
-	i += eb->start >> PAGE_CACHE_SHIFT;
-	mapping = eb->first_page->mapping;
-	if (!mapping)
-		return NULL;
+int extent_buffer_under_io(struct extent_buffer *eb)
+{
+	return (atomic_read(&eb->io_pages) ||
+		test_bit(EXTENT_BUFFER_WRITEBACK, &eb->bflags) ||
+		test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags));
+}
 
-	/*
-	 * extent_buffer_page is only called after pinning the page
-	 * by increasing the reference count.  So we know the page must
-	 * be in the radix tree.
-	 */
-	rcu_read_lock();
-	p = radix_tree_lookup(&mapping->page_tree, i);
-	rcu_read_unlock();
+/*
+ * Helper for releasing extent buffer page.
+ */
+static void btrfs_release_extent_buffer_page(struct extent_buffer *eb,
+						unsigned long start_idx)
+{
+	unsigned long index;
+	unsigned long num_pages;
+	struct page *page;
+	int mapped = !test_bit(EXTENT_BUFFER_DUMMY, &eb->bflags);
+
+	BUG_ON(extent_buffer_under_io(eb));
 
-	return p;
+	num_pages = num_extent_pages(eb->start, eb->len);
+	index = start_idx + num_pages;
+	if (start_idx >= index)
+		return;
+
+	do {
+		index--;
+		page = extent_buffer_page(eb, index);
+		if (page && mapped) {
+			spin_lock(&page->mapping->private_lock);
+			/*
+			 * We do this since we'll remove the pages after we've
+			 * removed the eb from the radix tree, so we could race
+			 * and have this page now attached to the new eb.  So
+			 * only clear page_private if it's still connected to
+			 * this eb.
+			 */
+			if (PagePrivate(page) &&
+			    page->private == (unsigned long)eb) {
+				BUG_ON(test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags));
+				BUG_ON(PageDirty(page));
+				BUG_ON(PageWriteback(page));
+				/*
+				 * We need to make sure we haven't be attached
+				 * to a new eb.
+				 */
+				ClearPagePrivate(page);
+				set_page_private(page, 0);
+				/* One for the page private */
+				page_cache_release(page);
+			}
+			spin_unlock(&page->mapping->private_lock);
+
+		}
+		if (page) {
+			/* One for when we alloced the page */
+			page_cache_release(page);
+		}
+	} while (index != start_idx);
 }
 
-static inline unsigned long num_extent_pages(u64 start, u64 len)
+/*
+ * Helper for releasing the extent buffer.
+ */
+static inline void btrfs_release_extent_buffer(struct extent_buffer *eb)
 {
-	return ((start + len + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT) -
-		(start >> PAGE_CACHE_SHIFT);
+	btrfs_release_extent_buffer_page(eb, 0);
+	__free_extent_buffer(eb);
 }
 
-static struct extent_buffer *__alloc_extent_buffer(struct extent_io_tree *tree,
-						   u64 start,
-						   unsigned long len,
-						   gfp_t mask)
+static struct extent_buffer *
+__alloc_extent_buffer(struct btrfs_fs_info *fs_info, u64 start,
+		      unsigned long len, gfp_t mask)
 {
 	struct extent_buffer *eb = NULL;
-#if LEAK_DEBUG
-	unsigned long flags;
-#endif
 
 	eb = kmem_cache_zalloc(extent_buffer_cache, mask);
+	if (eb == NULL)
+		return NULL;
 	eb->start = start;
 	eb->len = len;
-	spin_lock_init(&eb->lock);
-	init_waitqueue_head(&eb->lock_wq);
-
-#if LEAK_DEBUG
-	spin_lock_irqsave(&leak_lock, flags);
-	list_add(&eb->leak_list, &buffers);
-	spin_unlock_irqrestore(&leak_lock, flags);
-#endif
+	eb->fs_info = fs_info;
+	eb->bflags = 0;
+	rwlock_init(&eb->lock);
+	atomic_set(&eb->write_locks, 0);
+	atomic_set(&eb->read_locks, 0);
+	atomic_set(&eb->blocking_readers, 0);
+	atomic_set(&eb->blocking_writers, 0);
+	atomic_set(&eb->spinning_readers, 0);
+	atomic_set(&eb->spinning_writers, 0);
+	eb->lock_nested = 0;
+	init_waitqueue_head(&eb->write_lock_wq);
+	init_waitqueue_head(&eb->read_lock_wq);
+
+	btrfs_leak_debug_add(&eb->leak_list, &buffers);
+
+	spin_lock_init(&eb->refs_lock);
 	atomic_set(&eb->refs, 1);
+	atomic_set(&eb->io_pages, 0);
+
+	/*
+	 * Sanity checks, currently the maximum is 64k covered by 16x 4k pages
+	 */
+	BUILD_BUG_ON(BTRFS_MAX_METADATA_BLOCKSIZE
+		> MAX_INLINE_EXTENT_BUFFER_SIZE);
+	BUG_ON(len > MAX_INLINE_EXTENT_BUFFER_SIZE);
 
 	return eb;
 }
 
-static void __free_extent_buffer(struct extent_buffer *eb)
+struct extent_buffer *btrfs_clone_extent_buffer(struct extent_buffer *src)
 {
-#if LEAK_DEBUG
-	unsigned long flags;
-	spin_lock_irqsave(&leak_lock, flags);
-	list_del(&eb->leak_list);
-	spin_unlock_irqrestore(&leak_lock, flags);
-#endif
-	kmem_cache_free(extent_buffer_cache, eb);
+	unsigned long i;
+	struct page *p;
+	struct extent_buffer *new;
+	unsigned long num_pages = num_extent_pages(src->start, src->len);
+
+	new = __alloc_extent_buffer(NULL, src->start, src->len, GFP_NOFS);
+	if (new == NULL)
+		return NULL;
+
+	for (i = 0; i < num_pages; i++) {
+		p = alloc_page(GFP_NOFS);
+		if (!p) {
+			btrfs_release_extent_buffer(new);
+			return NULL;
+		}
+		attach_extent_buffer_page(new, p);
+		WARN_ON(PageDirty(p));
+		SetPageUptodate(p);
+		new->pages[i] = p;
+	}
+
+	copy_extent_buffer(new, src, 0, 0, src->len);
+	set_bit(EXTENT_BUFFER_UPTODATE, &new->bflags);
+	set_bit(EXTENT_BUFFER_DUMMY, &new->bflags);
+
+	return new;
+}
+
+struct extent_buffer *alloc_dummy_extent_buffer(u64 start, unsigned long len)
+{
+	struct extent_buffer *eb;
+	unsigned long num_pages = num_extent_pages(0, len);
+	unsigned long i;
+
+	eb = __alloc_extent_buffer(NULL, start, len, GFP_NOFS);
+	if (!eb)
+		return NULL;
+
+	for (i = 0; i < num_pages; i++) {
+		eb->pages[i] = alloc_page(GFP_NOFS);
+		if (!eb->pages[i])
+			goto err;
+	}
+	set_extent_buffer_uptodate(eb);
+	btrfs_set_header_nritems(eb, 0);
+	set_bit(EXTENT_BUFFER_DUMMY, &eb->bflags);
+
+	return eb;
+err:
+	for (; i > 0; i--)
+		__free_page(eb->pages[i - 1]);
+	__free_extent_buffer(eb);
+	return NULL;
 }
 
-struct extent_buffer *alloc_extent_buffer(struct extent_io_tree *tree,
-					  u64 start, unsigned long len,
-					  struct page *page0,
-					  gfp_t mask)
+static void check_buffer_tree_ref(struct extent_buffer *eb)
+{
+	int refs;
+	/* the ref bit is tricky.  We have to make sure it is set
+	 * if we have the buffer dirty.   Otherwise the
+	 * code to free a buffer can end up dropping a dirty
+	 * page
+	 *
+	 * Once the ref bit is set, it won't go away while the
+	 * buffer is dirty or in writeback, and it also won't
+	 * go away while we have the reference count on the
+	 * eb bumped.
+	 *
+	 * We can't just set the ref bit without bumping the
+	 * ref on the eb because free_extent_buffer might
+	 * see the ref bit and try to clear it.  If this happens
+	 * free_extent_buffer might end up dropping our original
+	 * ref by mistake and freeing the page before we are able
+	 * to add one more ref.
+	 *
+	 * So bump the ref count first, then set the bit.  If someone
+	 * beat us to it, drop the ref we added.
+	 */
+	refs = atomic_read(&eb->refs);
+	if (refs >= 2 && test_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags))
+		return;
+
+	spin_lock(&eb->refs_lock);
+	if (!test_and_set_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags))
+		atomic_inc(&eb->refs);
+	spin_unlock(&eb->refs_lock);
+}
+
+static void mark_extent_buffer_accessed(struct extent_buffer *eb,
+		struct page *accessed)
+{
+	unsigned long num_pages, i;
+
+	check_buffer_tree_ref(eb);
+
+	num_pages = num_extent_pages(eb->start, eb->len);
+	for (i = 0; i < num_pages; i++) {
+		struct page *p = extent_buffer_page(eb, i);
+		if (p != accessed)
+			mark_page_accessed(p);
+	}
+}
+
+struct extent_buffer *find_extent_buffer(struct btrfs_fs_info *fs_info,
+					 u64 start)
+{
+	struct extent_buffer *eb;
+
+	rcu_read_lock();
+	eb = radix_tree_lookup(&fs_info->buffer_radix,
+			       start >> PAGE_CACHE_SHIFT);
+	if (eb && atomic_inc_not_zero(&eb->refs)) {
+		rcu_read_unlock();
+		mark_extent_buffer_accessed(eb, NULL);
+		return eb;
+	}
+	rcu_read_unlock();
+
+	return NULL;
+}
+
+#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
+struct extent_buffer *alloc_test_extent_buffer(struct btrfs_fs_info *fs_info,
+					       u64 start, unsigned long len)
+{
+	struct extent_buffer *eb, *exists = NULL;
+	int ret;
+
+	eb = find_extent_buffer(fs_info, start);
+	if (eb)
+		return eb;
+	eb = alloc_dummy_extent_buffer(start, len);
+	if (!eb)
+		return NULL;
+	eb->fs_info = fs_info;
+again:
+	ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
+	if (ret)
+		goto free_eb;
+	spin_lock(&fs_info->buffer_lock);
+	ret = radix_tree_insert(&fs_info->buffer_radix,
+				start >> PAGE_CACHE_SHIFT, eb);
+	spin_unlock(&fs_info->buffer_lock);
+	radix_tree_preload_end();
+	if (ret == -EEXIST) {
+		exists = find_extent_buffer(fs_info, start);
+		if (exists)
+			goto free_eb;
+		else
+			goto again;
+	}
+	check_buffer_tree_ref(eb);
+	set_bit(EXTENT_BUFFER_IN_TREE, &eb->bflags);
+
+	/*
+	 * We will free dummy extent buffer's if they come into
+	 * free_extent_buffer with a ref count of 2, but if we are using this we
+	 * want the buffers to stay in memory until we're done with them, so
+	 * bump the ref count again.
+	 */
+	atomic_inc(&eb->refs);
+	return eb;
+free_eb:
+	btrfs_release_extent_buffer(eb);
+	return exists;
+}
+#endif
+
+struct extent_buffer *alloc_extent_buffer(struct btrfs_fs_info *fs_info,
+					  u64 start, unsigned long len)
 {
 	unsigned long num_pages = num_extent_pages(start, len);
 	unsigned long i;
@@ -3112,110 +4665,198 @@ struct extent_buffer *alloc_extent_buffer(struct extent_io_tree *tree,
 	struct extent_buffer *eb;
 	struct extent_buffer *exists = NULL;
 	struct page *p;
-	struct address_space *mapping = tree->mapping;
+	struct address_space *mapping = fs_info->btree_inode->i_mapping;
 	int uptodate = 1;
+	int ret;
 
-	spin_lock(&tree->buffer_lock);
-	eb = buffer_search(tree, start);
-	if (eb) {
-		atomic_inc(&eb->refs);
-		spin_unlock(&tree->buffer_lock);
-		mark_page_accessed(eb->first_page);
+	eb = find_extent_buffer(fs_info, start);
+	if (eb)
 		return eb;
-	}
-	spin_unlock(&tree->buffer_lock);
 
-	eb = __alloc_extent_buffer(tree, start, len, mask);
+	eb = __alloc_extent_buffer(fs_info, start, len, GFP_NOFS);
 	if (!eb)
 		return NULL;
 
-	if (page0) {
-		eb->first_page = page0;
-		i = 1;
-		index++;
-		page_cache_get(page0);
-		mark_page_accessed(page0);
-		set_page_extent_mapped(page0);
-		set_page_extent_head(page0, len);
-		uptodate = PageUptodate(page0);
-	} else {
-		i = 0;
-	}
-	for (; i < num_pages; i++, index++) {
-		p = find_or_create_page(mapping, index, mask | __GFP_HIGHMEM);
-		if (!p) {
-			WARN_ON(1);
+	for (i = 0; i < num_pages; i++, index++) {
+		p = find_or_create_page(mapping, index, GFP_NOFS);
+		if (!p)
 			goto free_eb;
+
+		spin_lock(&mapping->private_lock);
+		if (PagePrivate(p)) {
+			/*
+			 * We could have already allocated an eb for this page
+			 * and attached one so lets see if we can get a ref on
+			 * the existing eb, and if we can we know it's good and
+			 * we can just return that one, else we know we can just
+			 * overwrite page->private.
+			 */
+			exists = (struct extent_buffer *)p->private;
+			if (atomic_inc_not_zero(&exists->refs)) {
+				spin_unlock(&mapping->private_lock);
+				unlock_page(p);
+				page_cache_release(p);
+				mark_extent_buffer_accessed(exists, p);
+				goto free_eb;
+			}
+
+			/*
+			 * Do this so attach doesn't complain and we need to
+			 * drop the ref the old guy had.
+			 */
+			ClearPagePrivate(p);
+			WARN_ON(PageDirty(p));
+			page_cache_release(p);
 		}
-		set_page_extent_mapped(p);
-		mark_page_accessed(p);
-		if (i == 0) {
-			eb->first_page = p;
-			set_page_extent_head(p, len);
-		} else {
-			set_page_private(p, EXTENT_PAGE_PRIVATE);
-		}
+		attach_extent_buffer_page(eb, p);
+		spin_unlock(&mapping->private_lock);
+		WARN_ON(PageDirty(p));
+		eb->pages[i] = p;
 		if (!PageUptodate(p))
 			uptodate = 0;
-		unlock_page(p);
+
+		/*
+		 * see below about how we avoid a nasty race with release page
+		 * and why we unlock later
+		 */
 	}
 	if (uptodate)
 		set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
-
-	spin_lock(&tree->buffer_lock);
-	exists = buffer_tree_insert(tree, start, &eb->rb_node);
-	if (exists) {
-		/* add one reference for the caller */
-		atomic_inc(&exists->refs);
-		spin_unlock(&tree->buffer_lock);
+again:
+	ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
+	if (ret)
 		goto free_eb;
+
+	spin_lock(&fs_info->buffer_lock);
+	ret = radix_tree_insert(&fs_info->buffer_radix,
+				start >> PAGE_CACHE_SHIFT, eb);
+	spin_unlock(&fs_info->buffer_lock);
+	radix_tree_preload_end();
+	if (ret == -EEXIST) {
+		exists = find_extent_buffer(fs_info, start);
+		if (exists)
+			goto free_eb;
+		else
+			goto again;
 	}
 	/* add one reference for the tree */
-	atomic_inc(&eb->refs);
-	spin_unlock(&tree->buffer_lock);
+	check_buffer_tree_ref(eb);
+	set_bit(EXTENT_BUFFER_IN_TREE, &eb->bflags);
+
+	/*
+	 * there is a race where release page may have
+	 * tried to find this extent buffer in the radix
+	 * but failed.  It will tell the VM it is safe to
+	 * reclaim the, and it will clear the page private bit.
+	 * We must make sure to set the page private bit properly
+	 * after the extent buffer is in the radix tree so
+	 * it doesn't get lost
+	 */
+	SetPageChecked(eb->pages[0]);
+	for (i = 1; i < num_pages; i++) {
+		p = extent_buffer_page(eb, i);
+		ClearPageChecked(p);
+		unlock_page(p);
+	}
+	unlock_page(eb->pages[0]);
 	return eb;
 
 free_eb:
-	if (!atomic_dec_and_test(&eb->refs))
-		return exists;
-	for (index = 1; index < i; index++)
-		page_cache_release(extent_buffer_page(eb, index));
-	page_cache_release(extent_buffer_page(eb, 0));
-	__free_extent_buffer(eb);
+	for (i = 0; i < num_pages; i++) {
+		if (eb->pages[i])
+			unlock_page(eb->pages[i]);
+	}
+
+	WARN_ON(!atomic_dec_and_test(&eb->refs));
+	btrfs_release_extent_buffer(eb);
 	return exists;
 }
 
-struct extent_buffer *find_extent_buffer(struct extent_io_tree *tree,
-					 u64 start, unsigned long len,
-					  gfp_t mask)
+static inline void btrfs_release_extent_buffer_rcu(struct rcu_head *head)
 {
-	struct extent_buffer *eb;
+	struct extent_buffer *eb =
+			container_of(head, struct extent_buffer, rcu_head);
 
-	spin_lock(&tree->buffer_lock);
-	eb = buffer_search(tree, start);
-	if (eb)
-		atomic_inc(&eb->refs);
-	spin_unlock(&tree->buffer_lock);
+	__free_extent_buffer(eb);
+}
 
-	if (eb)
-		mark_page_accessed(eb->first_page);
+/* Expects to have eb->eb_lock already held */
+static int release_extent_buffer(struct extent_buffer *eb)
+{
+	WARN_ON(atomic_read(&eb->refs) == 0);
+	if (atomic_dec_and_test(&eb->refs)) {
+		if (test_and_clear_bit(EXTENT_BUFFER_IN_TREE, &eb->bflags)) {
+			struct btrfs_fs_info *fs_info = eb->fs_info;
 
-	return eb;
+			spin_unlock(&eb->refs_lock);
+
+			spin_lock(&fs_info->buffer_lock);
+			radix_tree_delete(&fs_info->buffer_radix,
+					  eb->start >> PAGE_CACHE_SHIFT);
+			spin_unlock(&fs_info->buffer_lock);
+		} else {
+			spin_unlock(&eb->refs_lock);
+		}
+
+		/* Should be safe to release our pages at this point */
+		btrfs_release_extent_buffer_page(eb, 0);
+		call_rcu(&eb->rcu_head, btrfs_release_extent_buffer_rcu);
+		return 1;
+	}
+	spin_unlock(&eb->refs_lock);
+
+	return 0;
 }
 
 void free_extent_buffer(struct extent_buffer *eb)
 {
+	int refs;
+	int old;
 	if (!eb)
 		return;
 
-	if (!atomic_dec_and_test(&eb->refs))
+	while (1) {
+		refs = atomic_read(&eb->refs);
+		if (refs <= 3)
+			break;
+		old = atomic_cmpxchg(&eb->refs, refs, refs - 1);
+		if (old == refs)
+			return;
+	}
+
+	spin_lock(&eb->refs_lock);
+	if (atomic_read(&eb->refs) == 2 &&
+	    test_bit(EXTENT_BUFFER_DUMMY, &eb->bflags))
+		atomic_dec(&eb->refs);
+
+	if (atomic_read(&eb->refs) == 2 &&
+	    test_bit(EXTENT_BUFFER_STALE, &eb->bflags) &&
+	    !extent_buffer_under_io(eb) &&
+	    test_and_clear_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags))
+		atomic_dec(&eb->refs);
+
+	/*
+	 * I know this is terrible, but it's temporary until we stop tracking
+	 * the uptodate bits and such for the extent buffers.
+	 */
+	release_extent_buffer(eb);
+}
+
+void free_extent_buffer_stale(struct extent_buffer *eb)
+{
+	if (!eb)
 		return;
 
-	WARN_ON(1);
+	spin_lock(&eb->refs_lock);
+	set_bit(EXTENT_BUFFER_STALE, &eb->bflags);
+
+	if (atomic_read(&eb->refs) == 2 && !extent_buffer_under_io(eb) &&
+	    test_and_clear_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags))
+		atomic_dec(&eb->refs);
+	release_extent_buffer(eb);
 }
 
-int clear_extent_buffer_dirty(struct extent_io_tree *tree,
-			      struct extent_buffer *eb)
+void clear_extent_buffer_dirty(struct extent_buffer *eb)
 {
 	unsigned long i;
 	unsigned long num_pages;
@@ -3229,10 +4870,7 @@ int clear_extent_buffer_dirty(struct extent_io_tree *tree,
 			continue;
 
 		lock_page(page);
-		if (i == 0)
-			set_page_extent_head(page, eb->len);
-		else
-			set_page_private(page, EXTENT_PAGE_PRIVATE);
+		WARN_ON(!PagePrivate(page));
 
 		clear_page_dirty_for_io(page);
 		spin_lock_irq(&page->mapping->tree_lock);
@@ -3242,45 +4880,39 @@ int clear_extent_buffer_dirty(struct extent_io_tree *tree,
 						PAGECACHE_TAG_DIRTY);
 		}
 		spin_unlock_irq(&page->mapping->tree_lock);
+		ClearPageError(page);
 		unlock_page(page);
 	}
-	return 0;
-}
-
-int wait_on_extent_buffer_writeback(struct extent_io_tree *tree,
-				    struct extent_buffer *eb)
-{
-	return wait_on_extent_writeback(tree, eb->start,
-					eb->start + eb->len - 1);
+	WARN_ON(atomic_read(&eb->refs) == 0);
 }
 
-int set_extent_buffer_dirty(struct extent_io_tree *tree,
-			     struct extent_buffer *eb)
+int set_extent_buffer_dirty(struct extent_buffer *eb)
 {
 	unsigned long i;
 	unsigned long num_pages;
 	int was_dirty = 0;
 
+	check_buffer_tree_ref(eb);
+
 	was_dirty = test_and_set_bit(EXTENT_BUFFER_DIRTY, &eb->bflags);
+
 	num_pages = num_extent_pages(eb->start, eb->len);
+	WARN_ON(atomic_read(&eb->refs) == 0);
+	WARN_ON(!test_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags));
+
 	for (i = 0; i < num_pages; i++)
-		__set_page_dirty_nobuffers(extent_buffer_page(eb, i));
+		set_page_dirty(extent_buffer_page(eb, i));
 	return was_dirty;
 }
 
-int clear_extent_buffer_uptodate(struct extent_io_tree *tree,
-				struct extent_buffer *eb,
-				struct extent_state **cached_state)
+int clear_extent_buffer_uptodate(struct extent_buffer *eb)
 {
 	unsigned long i;
 	struct page *page;
 	unsigned long num_pages;
 
-	num_pages = num_extent_pages(eb->start, eb->len);
 	clear_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
-
-	clear_extent_uptodate(tree, eb->start, eb->start + eb->len - 1,
-			      cached_state, GFP_NOFS);
+	num_pages = num_extent_pages(eb->start, eb->len);
 	for (i = 0; i < num_pages; i++) {
 		page = extent_buffer_page(eb, i);
 		if (page)
@@ -3289,88 +4921,28 @@ int clear_extent_buffer_uptodate(struct extent_io_tree *tree,
 	return 0;
 }
 
-int set_extent_buffer_uptodate(struct extent_io_tree *tree,
-				struct extent_buffer *eb)
+int set_extent_buffer_uptodate(struct extent_buffer *eb)
 {
 	unsigned long i;
 	struct page *page;
 	unsigned long num_pages;
 
+	set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
 	num_pages = num_extent_pages(eb->start, eb->len);
-
-	set_extent_uptodate(tree, eb->start, eb->start + eb->len - 1,
-			    GFP_NOFS);
 	for (i = 0; i < num_pages; i++) {
 		page = extent_buffer_page(eb, i);
-		if ((i == 0 && (eb->start & (PAGE_CACHE_SIZE - 1))) ||
-		    ((i == num_pages - 1) &&
-		     ((eb->start + eb->len) & (PAGE_CACHE_SIZE - 1)))) {
-			check_page_uptodate(tree, page);
-			continue;
-		}
 		SetPageUptodate(page);
 	}
 	return 0;
 }
 
-int extent_range_uptodate(struct extent_io_tree *tree,
-			  u64 start, u64 end)
+int extent_buffer_uptodate(struct extent_buffer *eb)
 {
-	struct page *page;
-	int ret;
-	int pg_uptodate = 1;
-	int uptodate;
-	unsigned long index;
-
-	ret = test_range_bit(tree, start, end, EXTENT_UPTODATE, 1, NULL);
-	if (ret)
-		return 1;
-	while (start <= end) {
-		index = start >> PAGE_CACHE_SHIFT;
-		page = find_get_page(tree->mapping, index);
-		uptodate = PageUptodate(page);
-		page_cache_release(page);
-		if (!uptodate) {
-			pg_uptodate = 0;
-			break;
-		}
-		start += PAGE_CACHE_SIZE;
-	}
-	return pg_uptodate;
-}
-
-int extent_buffer_uptodate(struct extent_io_tree *tree,
-			   struct extent_buffer *eb,
-			   struct extent_state *cached_state)
-{
-	int ret = 0;
-	unsigned long num_pages;
-	unsigned long i;
-	struct page *page;
-	int pg_uptodate = 1;
-
-	if (test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags))
-		return 1;
-
-	ret = test_range_bit(tree, eb->start, eb->start + eb->len - 1,
-			   EXTENT_UPTODATE, 1, cached_state);
-	if (ret)
-		return ret;
-
-	num_pages = num_extent_pages(eb->start, eb->len);
-	for (i = 0; i < num_pages; i++) {
-		page = extent_buffer_page(eb, i);
-		if (!PageUptodate(page)) {
-			pg_uptodate = 0;
-			break;
-		}
-	}
-	return pg_uptodate;
+	return test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
 }
 
 int read_extent_buffer_pages(struct extent_io_tree *tree,
-			     struct extent_buffer *eb,
-			     u64 start, int wait,
+			     struct extent_buffer *eb, u64 start, int wait,
 			     get_extent_t *get_extent, int mirror_num)
 {
 	unsigned long i;
@@ -3380,19 +4952,14 @@ int read_extent_buffer_pages(struct extent_io_tree *tree,
 	int ret = 0;
 	int locked_pages = 0;
 	int all_uptodate = 1;
-	int inc_all_pages = 0;
 	unsigned long num_pages;
+	unsigned long num_reads = 0;
 	struct bio *bio = NULL;
 	unsigned long bio_flags = 0;
 
 	if (test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags))
 		return 0;
 
-	if (test_range_bit(tree, eb->start, eb->start + eb->len - 1,
-			   EXTENT_UPTODATE, 1, NULL)) {
-		return 0;
-	}
-
 	if (start) {
 		WARN_ON(start < eb->start);
 		start_i = (start >> PAGE_CACHE_SHIFT) -
@@ -3404,15 +4971,17 @@ int read_extent_buffer_pages(struct extent_io_tree *tree,
 	num_pages = num_extent_pages(eb->start, eb->len);
 	for (i = start_i; i < num_pages; i++) {
 		page = extent_buffer_page(eb, i);
-		if (!wait) {
+		if (wait == WAIT_NONE) {
 			if (!trylock_page(page))
 				goto unlock_exit;
 		} else {
 			lock_page(page);
 		}
 		locked_pages++;
-		if (!PageUptodate(page))
+		if (!PageUptodate(page)) {
+			num_reads++;
 			all_uptodate = 0;
+		}
 	}
 	if (all_uptodate) {
 		if (start_i == 0)
@@ -3420,17 +4989,17 @@ int read_extent_buffer_pages(struct extent_io_tree *tree,
 		goto unlock_exit;
 	}
 
+	clear_bit(EXTENT_BUFFER_IOERR, &eb->bflags);
+	eb->read_mirror = 0;
+	atomic_set(&eb->io_pages, num_reads);
 	for (i = start_i; i < num_pages; i++) {
 		page = extent_buffer_page(eb, i);
-		if (inc_all_pages)
-			page_cache_get(page);
 		if (!PageUptodate(page)) {
-			if (start_i == 0)
-				inc_all_pages = 1;
 			ClearPageError(page);
 			err = __extent_read_full_page(tree, page,
 						      get_extent, &bio,
-						      mirror_num, &bio_flags);
+						      mirror_num, &bio_flags,
+						      READ | REQ_META);
 			if (err)
 				ret = err;
 		} else {
@@ -3438,10 +5007,14 @@ int read_extent_buffer_pages(struct extent_io_tree *tree,
 		}
 	}
 
-	if (bio)
-		submit_one_bio(READ, bio, mirror_num, bio_flags);
+	if (bio) {
+		err = submit_one_bio(READ | REQ_META, bio, mirror_num,
+				     bio_flags);
+		if (err)
+			return err;
+	}
 
-	if (ret || !wait)
+	if (ret || wait != WAIT_COMPLETE)
 		return ret;
 
 	for (i = start_i; i < num_pages; i++) {
@@ -3451,8 +5024,6 @@ int read_extent_buffer_pages(struct extent_io_tree *tree,
 			ret = -EIO;
 	}
 
-	if (!ret)
-		set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
 	return ret;
 
 unlock_exit:
@@ -3481,15 +5052,14 @@ void read_extent_buffer(struct extent_buffer *eb, void *dstv,
 	WARN_ON(start > eb->len);
 	WARN_ON(start + len > eb->start + eb->len);
 
-	offset = (start_offset + start) & ((unsigned long)PAGE_CACHE_SIZE - 1);
+	offset = (start_offset + start) & (PAGE_CACHE_SIZE - 1);
 
 	while (len > 0) {
 		page = extent_buffer_page(eb, i);
 
 		cur = min(len, (PAGE_CACHE_SIZE - offset));
-		kaddr = kmap_atomic(page, KM_USER1);
+		kaddr = page_address(page);
 		memcpy(dst, kaddr + offset, cur);
-		kunmap_atomic(kaddr, KM_USER1);
 
 		dst += cur;
 		len -= cur;
@@ -3498,10 +5068,47 @@ void read_extent_buffer(struct extent_buffer *eb, void *dstv,
 	}
 }
 
+int read_extent_buffer_to_user(struct extent_buffer *eb, void __user *dstv,
+			unsigned long start,
+			unsigned long len)
+{
+	size_t cur;
+	size_t offset;
+	struct page *page;
+	char *kaddr;
+	char __user *dst = (char __user *)dstv;
+	size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
+	unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;
+	int ret = 0;
+
+	WARN_ON(start > eb->len);
+	WARN_ON(start + len > eb->start + eb->len);
+
+	offset = (start_offset + start) & (PAGE_CACHE_SIZE - 1);
+
+	while (len > 0) {
+		page = extent_buffer_page(eb, i);
+
+		cur = min(len, (PAGE_CACHE_SIZE - offset));
+		kaddr = page_address(page);
+		if (copy_to_user(dst, kaddr + offset, cur)) {
+			ret = -EFAULT;
+			break;
+		}
+
+		dst += cur;
+		len -= cur;
+		offset = 0;
+		i++;
+	}
+
+	return ret;
+}
+
 int map_private_extent_buffer(struct extent_buffer *eb, unsigned long start,
-			       unsigned long min_len, char **token, char **map,
+			       unsigned long min_len, char **map,
 			       unsigned long *map_start,
-			       unsigned long *map_len, int km)
+			       unsigned long *map_len)
 {
 	size_t offset = start & (PAGE_CACHE_SIZE - 1);
 	char *kaddr;
@@ -3523,49 +5130,19 @@ int map_private_extent_buffer(struct extent_buffer *eb, unsigned long start,
 	}
 
 	if (start + min_len > eb->len) {
-		printk(KERN_ERR "btrfs bad mapping eb start %llu len %lu, "
-		       "wanted %lu %lu\n", (unsigned long long)eb->start,
-		       eb->len, start, min_len);
-		WARN_ON(1);
+		WARN(1, KERN_ERR "btrfs bad mapping eb start %llu len %lu, "
+		       "wanted %lu %lu\n",
+		       eb->start, eb->len, start, min_len);
+		return -EINVAL;
 	}
 
 	p = extent_buffer_page(eb, i);
-	kaddr = kmap_atomic(p, km);
-	*token = kaddr;
+	kaddr = page_address(p);
 	*map = kaddr + offset;
 	*map_len = PAGE_CACHE_SIZE - offset;
 	return 0;
 }
 
-int map_extent_buffer(struct extent_buffer *eb, unsigned long start,
-		      unsigned long min_len,
-		      char **token, char **map,
-		      unsigned long *map_start,
-		      unsigned long *map_len, int km)
-{
-	int err;
-	int save = 0;
-	if (eb->map_token) {
-		unmap_extent_buffer(eb, eb->map_token, km);
-		eb->map_token = NULL;
-		save = 1;
-	}
-	err = map_private_extent_buffer(eb, start, min_len, token, map,
-				       map_start, map_len, km);
-	if (!err && save) {
-		eb->map_token = *token;
-		eb->kaddr = *map;
-		eb->map_start = *map_start;
-		eb->map_len = *map_len;
-	}
-	return err;
-}
-
-void unmap_extent_buffer(struct extent_buffer *eb, char *token, int km)
-{
-	kunmap_atomic(token, km);
-}
-
 int memcmp_extent_buffer(struct extent_buffer *eb, const void *ptrv,
 			  unsigned long start,
 			  unsigned long len)
@@ -3582,16 +5159,15 @@ int memcmp_extent_buffer(struct extent_buffer *eb, const void *ptrv,
 	WARN_ON(start > eb->len);
 	WARN_ON(start + len > eb->start + eb->len);
 
-	offset = (start_offset + start) & ((unsigned long)PAGE_CACHE_SIZE - 1);
+	offset = (start_offset + start) & (PAGE_CACHE_SIZE - 1);
 
 	while (len > 0) {
 		page = extent_buffer_page(eb, i);
 
 		cur = min(len, (PAGE_CACHE_SIZE - offset));
 
-		kaddr = kmap_atomic(page, KM_USER0);
+		kaddr = page_address(page);
 		ret = memcmp(ptr, kaddr + offset, cur);
-		kunmap_atomic(kaddr, KM_USER0);
 		if (ret)
 			break;
 
@@ -3617,16 +5193,15 @@ void write_extent_buffer(struct extent_buffer *eb, const void *srcv,
 	WARN_ON(start > eb->len);
 	WARN_ON(start + len > eb->start + eb->len);
 
-	offset = (start_offset + start) & ((unsigned long)PAGE_CACHE_SIZE - 1);
+	offset = (start_offset + start) & (PAGE_CACHE_SIZE - 1);
 
 	while (len > 0) {
 		page = extent_buffer_page(eb, i);
 		WARN_ON(!PageUptodate(page));
 
 		cur = min(len, PAGE_CACHE_SIZE - offset);
-		kaddr = kmap_atomic(page, KM_USER1);
+		kaddr = page_address(page);
 		memcpy(kaddr + offset, src, cur);
-		kunmap_atomic(kaddr, KM_USER1);
 
 		src += cur;
 		len -= cur;
@@ -3648,16 +5223,15 @@ void memset_extent_buffer(struct extent_buffer *eb, char c,
 	WARN_ON(start > eb->len);
 	WARN_ON(start + len > eb->start + eb->len);
 
-	offset = (start_offset + start) & ((unsigned long)PAGE_CACHE_SIZE - 1);
+	offset = (start_offset + start) & (PAGE_CACHE_SIZE - 1);
 
 	while (len > 0) {
 		page = extent_buffer_page(eb, i);
 		WARN_ON(!PageUptodate(page));
 
 		cur = min(len, PAGE_CACHE_SIZE - offset);
-		kaddr = kmap_atomic(page, KM_USER0);
+		kaddr = page_address(page);
 		memset(kaddr + offset, c, cur);
-		kunmap_atomic(kaddr, KM_USER0);
 
 		len -= cur;
 		offset = 0;
@@ -3680,7 +5254,7 @@ void copy_extent_buffer(struct extent_buffer *dst, struct extent_buffer *src,
 	WARN_ON(src->len != dst_len);
 
 	offset = (start_offset + dst_offset) &
-		((unsigned long)PAGE_CACHE_SIZE - 1);
+		(PAGE_CACHE_SIZE - 1);
 
 	while (len > 0) {
 		page = extent_buffer_page(dst, i);
@@ -3688,9 +5262,8 @@ void copy_extent_buffer(struct extent_buffer *dst, struct extent_buffer *src,
 
 		cur = min(len, (unsigned long)(PAGE_CACHE_SIZE - offset));
 
-		kaddr = kmap_atomic(page, KM_USER0);
+		kaddr = page_address(page);
 		read_extent_buffer(src, kaddr + offset, src_offset, cur);
-		kunmap_atomic(kaddr, KM_USER0);
 
 		src_offset += cur;
 		len -= cur;
@@ -3699,42 +5272,32 @@ void copy_extent_buffer(struct extent_buffer *dst, struct extent_buffer *src,
 	}
 }
 
-static void move_pages(struct page *dst_page, struct page *src_page,
-		       unsigned long dst_off, unsigned long src_off,
-		       unsigned long len)
+static inline bool areas_overlap(unsigned long src, unsigned long dst, unsigned long len)
 {
-	char *dst_kaddr = kmap_atomic(dst_page, KM_USER0);
-	if (dst_page == src_page) {
-		memmove(dst_kaddr + dst_off, dst_kaddr + src_off, len);
-	} else {
-		char *src_kaddr = kmap_atomic(src_page, KM_USER1);
-		char *p = dst_kaddr + dst_off + len;
-		char *s = src_kaddr + src_off + len;
-
-		while (len--)
-			*--p = *--s;
-
-		kunmap_atomic(src_kaddr, KM_USER1);
-	}
-	kunmap_atomic(dst_kaddr, KM_USER0);
+	unsigned long distance = (src > dst) ? src - dst : dst - src;
+	return distance < len;
 }
 
 static void copy_pages(struct page *dst_page, struct page *src_page,
 		       unsigned long dst_off, unsigned long src_off,
 		       unsigned long len)
 {
-	char *dst_kaddr = kmap_atomic(dst_page, KM_USER0);
+	char *dst_kaddr = page_address(dst_page);
 	char *src_kaddr;
+	int must_memmove = 0;
 
-	if (dst_page != src_page)
-		src_kaddr = kmap_atomic(src_page, KM_USER1);
-	else
+	if (dst_page != src_page) {
+		src_kaddr = page_address(src_page);
+	} else {
 		src_kaddr = dst_kaddr;
+		if (areas_overlap(src_off, dst_off, len))
+			must_memmove = 1;
+	}
 
-	memcpy(dst_kaddr + dst_off, src_kaddr + src_off, len);
-	kunmap_atomic(dst_kaddr, KM_USER0);
-	if (dst_page != src_page)
-		kunmap_atomic(src_kaddr, KM_USER1);
+	if (must_memmove)
+		memmove(dst_kaddr + dst_off, src_kaddr + src_off, len);
+	else
+		memcpy(dst_kaddr + dst_off, src_kaddr + src_off, len);
 }
 
 void memcpy_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
@@ -3748,21 +5311,21 @@ void memcpy_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
 	unsigned long src_i;
 
 	if (src_offset + len > dst->len) {
-		printk(KERN_ERR "btrfs memmove bogus src_offset %lu move "
+		printk(KERN_ERR "BTRFS: memmove bogus src_offset %lu move "
 		       "len %lu dst len %lu\n", src_offset, len, dst->len);
 		BUG_ON(1);
 	}
 	if (dst_offset + len > dst->len) {
-		printk(KERN_ERR "btrfs memmove bogus dst_offset %lu move "
+		printk(KERN_ERR "BTRFS: memmove bogus dst_offset %lu move "
 		       "len %lu dst len %lu\n", dst_offset, len, dst->len);
 		BUG_ON(1);
 	}
 
 	while (len > 0) {
 		dst_off_in_page = (start_offset + dst_offset) &
-			((unsigned long)PAGE_CACHE_SIZE - 1);
+			(PAGE_CACHE_SIZE - 1);
 		src_off_in_page = (start_offset + src_offset) &
-			((unsigned long)PAGE_CACHE_SIZE - 1);
+			(PAGE_CACHE_SIZE - 1);
 
 		dst_i = (start_offset + dst_offset) >> PAGE_CACHE_SHIFT;
 		src_i = (start_offset + src_offset) >> PAGE_CACHE_SHIFT;
@@ -3795,12 +5358,12 @@ void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
 	unsigned long src_i;
 
 	if (src_offset + len > dst->len) {
-		printk(KERN_ERR "btrfs memmove bogus src_offset %lu move "
+		printk(KERN_ERR "BTRFS: memmove bogus src_offset %lu move "
 		       "len %lu len %lu\n", src_offset, len, dst->len);
 		BUG_ON(1);
 	}
 	if (dst_offset + len > dst->len) {
-		printk(KERN_ERR "btrfs memmove bogus dst_offset %lu move "
+		printk(KERN_ERR "BTRFS: memmove bogus dst_offset %lu move "
 		       "len %lu len %lu\n", dst_offset, len, dst->len);
 		BUG_ON(1);
 	}
@@ -3813,13 +5376,13 @@ void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
 		src_i = (start_offset + src_end) >> PAGE_CACHE_SHIFT;
 
 		dst_off_in_page = (start_offset + dst_end) &
-			((unsigned long)PAGE_CACHE_SIZE - 1);
+			(PAGE_CACHE_SIZE - 1);
 		src_off_in_page = (start_offset + src_end) &
-			((unsigned long)PAGE_CACHE_SIZE - 1);
+			(PAGE_CACHE_SIZE - 1);
 
 		cur = min_t(unsigned long, len, src_off_in_page + 1);
 		cur = min(cur, dst_off_in_page + 1);
-		move_pages(extent_buffer_page(dst, dst_i),
+		copy_pages(extent_buffer_page(dst, dst_i),
 			   extent_buffer_page(dst, src_i),
 			   dst_off_in_page - cur + 1,
 			   src_off_in_page - cur + 1, cur);
@@ -3830,34 +5393,44 @@ void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
 	}
 }
 
-int try_release_extent_buffer(struct extent_io_tree *tree, struct page *page)
+int try_release_extent_buffer(struct page *page)
 {
-	u64 start = page_offset(page);
 	struct extent_buffer *eb;
-	int ret = 1;
-	unsigned long i;
-	unsigned long num_pages;
 
-	spin_lock(&tree->buffer_lock);
-	eb = buffer_search(tree, start);
-	if (!eb)
-		goto out;
+	/*
+	 * We need to make sure noboody is attaching this page to an eb right
+	 * now.
+	 */
+	spin_lock(&page->mapping->private_lock);
+	if (!PagePrivate(page)) {
+		spin_unlock(&page->mapping->private_lock);
+		return 1;
+	}
 
-	if (atomic_read(&eb->refs) > 1) {
-		ret = 0;
-		goto out;
+	eb = (struct extent_buffer *)page->private;
+	BUG_ON(!eb);
+
+	/*
+	 * This is a little awful but should be ok, we need to make sure that
+	 * the eb doesn't disappear out from under us while we're looking at
+	 * this page.
+	 */
+	spin_lock(&eb->refs_lock);
+	if (atomic_read(&eb->refs) != 1 || extent_buffer_under_io(eb)) {
+		spin_unlock(&eb->refs_lock);
+		spin_unlock(&page->mapping->private_lock);
+		return 0;
 	}
-	if (test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) {
-		ret = 0;
-		goto out;
+	spin_unlock(&page->mapping->private_lock);
+
+	/*
+	 * If tree ref isn't set then we know the ref on this eb is a real ref,
+	 * so just return, this page will likely be freed soon anyway.
+	 */
+	if (!test_and_clear_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags)) {
+		spin_unlock(&eb->refs_lock);
+		return 0;
 	}
-	/* at this point we can safely release the extent buffer */
-	num_pages = num_extent_pages(eb->start, eb->len);
-	for (i = 0; i < num_pages; i++)
-		page_cache_release(extent_buffer_page(eb, i));
-	rb_erase(&eb->rb_node, &tree->buffer);
-	__free_extent_buffer(eb);
-out:
-	spin_unlock(&tree->buffer_lock);
-	return ret;
+
+	return release_extent_buffer(eb);
 }
diff --git a/fs/btrfs/extent_io.h b/fs/btrfs/extent_io.h
index bbab4813646..ccc264e7bde 100644
--- a/fs/btrfs/extent_io.h
+++ b/fs/btrfs/extent_io.h
@@ -16,25 +16,41 @@
 #define EXTENT_BOUNDARY (1 << 9)
 #define EXTENT_NODATASUM (1 << 10)
 #define EXTENT_DO_ACCOUNTING (1 << 11)
+#define EXTENT_FIRST_DELALLOC (1 << 12)
+#define EXTENT_NEED_WAIT (1 << 13)
+#define EXTENT_DAMAGED (1 << 14)
+#define EXTENT_NORESERVE (1 << 15)
 #define EXTENT_IOBITS (EXTENT_LOCKED | EXTENT_WRITEBACK)
+#define EXTENT_CTLBITS (EXTENT_DO_ACCOUNTING | EXTENT_FIRST_DELALLOC)
 
-/* flags for bio submission */
+/*
+ * flags for bio submission. The high bits indicate the compression
+ * type for this bio
+ */
 #define EXTENT_BIO_COMPRESSED 1
+#define EXTENT_BIO_TREE_LOG 2
+#define EXTENT_BIO_PARENT_LOCKED 4
+#define EXTENT_BIO_FLAG_SHIFT 16
 
 /* these are bit numbers for test/set bit */
 #define EXTENT_BUFFER_UPTODATE 0
 #define EXTENT_BUFFER_BLOCKING 1
 #define EXTENT_BUFFER_DIRTY 2
+#define EXTENT_BUFFER_CORRUPT 3
+#define EXTENT_BUFFER_READAHEAD 4	/* this got triggered by readahead */
+#define EXTENT_BUFFER_TREE_REF 5
+#define EXTENT_BUFFER_STALE 6
+#define EXTENT_BUFFER_WRITEBACK 7
+#define EXTENT_BUFFER_IOERR 8
+#define EXTENT_BUFFER_DUMMY 9
+#define EXTENT_BUFFER_IN_TREE 10
 
 /* these are flags for extent_clear_unlock_delalloc */
-#define EXTENT_CLEAR_UNLOCK_PAGE 0x1
-#define EXTENT_CLEAR_UNLOCK	 0x2
-#define EXTENT_CLEAR_DELALLOC	 0x4
-#define EXTENT_CLEAR_DIRTY	 0x8
-#define EXTENT_SET_WRITEBACK	 0x10
-#define EXTENT_END_WRITEBACK	 0x20
-#define EXTENT_SET_PRIVATE2	 0x40
-#define EXTENT_CLEAR_ACCOUNTING  0x80
+#define PAGE_UNLOCK		(1 << 0)
+#define PAGE_CLEAR_DIRTY	(1 << 1)
+#define PAGE_SET_WRITEBACK	(1 << 2)
+#define PAGE_END_WRITEBACK	(1 << 3)
+#define PAGE_SET_PRIVATE2	(1 << 4)
 
 /*
  * page->private values.  Every page that is controlled by the extent
@@ -44,10 +60,12 @@
 #define EXTENT_PAGE_PRIVATE_FIRST_PAGE 3
 
 struct extent_state;
+struct btrfs_root;
+struct btrfs_io_bio;
 
 typedef	int (extent_submit_bio_hook_t)(struct inode *inode, int rw,
 				       struct bio *bio, int mirror_num,
-				       unsigned long bio_flags);
+				       unsigned long bio_flags, u64 bio_offset);
 struct extent_io_ops {
 	int (*fill_delalloc)(struct inode *inode, struct page *locked_page,
 			     u64 start, u64 end, int *page_started,
@@ -55,39 +73,32 @@ struct extent_io_ops {
 	int (*writepage_start_hook)(struct page *page, u64 start, u64 end);
 	int (*writepage_io_hook)(struct page *page, u64 start, u64 end);
 	extent_submit_bio_hook_t *submit_bio_hook;
-	int (*merge_bio_hook)(struct page *page, unsigned long offset,
+	int (*merge_bio_hook)(int rw, struct page *page, unsigned long offset,
 			      size_t size, struct bio *bio,
 			      unsigned long bio_flags);
-	int (*readpage_io_hook)(struct page *page, u64 start, u64 end);
-	int (*readpage_io_failed_hook)(struct bio *bio, struct page *page,
-				       u64 start, u64 end,
-				       struct extent_state *state);
-	int (*writepage_io_failed_hook)(struct bio *bio, struct page *page,
-					u64 start, u64 end,
-				       struct extent_state *state);
-	int (*readpage_end_io_hook)(struct page *page, u64 start, u64 end,
-				    struct extent_state *state);
+	int (*readpage_io_failed_hook)(struct page *page, int failed_mirror);
+	int (*readpage_end_io_hook)(struct btrfs_io_bio *io_bio, u64 phy_offset,
+				    struct page *page, u64 start, u64 end,
+				    int mirror);
 	int (*writepage_end_io_hook)(struct page *page, u64 start, u64 end,
 				      struct extent_state *state, int uptodate);
-	int (*set_bit_hook)(struct inode *inode, u64 start, u64 end,
-			    unsigned long old, unsigned long bits);
-	int (*clear_bit_hook)(struct inode *inode, struct extent_state *state,
-			      unsigned long bits);
-	int (*merge_extent_hook)(struct inode *inode,
-				 struct extent_state *new,
-				 struct extent_state *other);
-	int (*split_extent_hook)(struct inode *inode,
-				 struct extent_state *orig, u64 split);
-	int (*write_cache_pages_lock_hook)(struct page *page);
+	void (*set_bit_hook)(struct inode *inode, struct extent_state *state,
+			     unsigned long *bits);
+	void (*clear_bit_hook)(struct inode *inode, struct extent_state *state,
+			       unsigned long *bits);
+	void (*merge_extent_hook)(struct inode *inode,
+				  struct extent_state *new,
+				  struct extent_state *other);
+	void (*split_extent_hook)(struct inode *inode,
+				  struct extent_state *orig, u64 split);
 };
 
 struct extent_io_tree {
 	struct rb_root state;
-	struct rb_root buffer;
 	struct address_space *mapping;
 	u64 dirty_bytes;
+	int track_uptodate;
 	spinlock_t lock;
-	spinlock_t buffer_lock;
 	struct extent_io_ops *ops;
 };
 
@@ -101,110 +112,135 @@ struct extent_state {
 	wait_queue_head_t wq;
 	atomic_t refs;
 	unsigned long state;
-	u64 split_start;
-	u64 split_end;
 
 	/* for use by the FS */
 	u64 private;
 
+#ifdef CONFIG_BTRFS_DEBUG
 	struct list_head leak_list;
+#endif
 };
 
+#define INLINE_EXTENT_BUFFER_PAGES 16
+#define MAX_INLINE_EXTENT_BUFFER_SIZE (INLINE_EXTENT_BUFFER_PAGES * PAGE_CACHE_SIZE)
 struct extent_buffer {
 	u64 start;
 	unsigned long len;
-	char *map_token;
-	char *kaddr;
 	unsigned long map_start;
 	unsigned long map_len;
-	struct page *first_page;
 	unsigned long bflags;
+	struct btrfs_fs_info *fs_info;
+	spinlock_t refs_lock;
 	atomic_t refs;
-	struct list_head leak_list;
-	struct rb_node rb_node;
+	atomic_t io_pages;
+	int read_mirror;
+	struct rcu_head rcu_head;
+	pid_t lock_owner;
 
-	/* the spinlock is used to protect most operations */
-	spinlock_t lock;
+	/* count of read lock holders on the extent buffer */
+	atomic_t write_locks;
+	atomic_t read_locks;
+	atomic_t blocking_writers;
+	atomic_t blocking_readers;
+	atomic_t spinning_readers;
+	atomic_t spinning_writers;
+	int lock_nested;
 
-	/*
-	 * when we keep the lock held while blocking, waiters go onto
-	 * the wq
+	/* protects write locks */
+	rwlock_t lock;
+
+	/* readers use lock_wq while they wait for the write
+	 * lock holders to unlock
 	 */
-	wait_queue_head_t lock_wq;
+	wait_queue_head_t write_lock_wq;
+
+	/* writers use read_lock_wq while they wait for readers
+	 * to unlock
+	 */
+	wait_queue_head_t read_lock_wq;
+	struct page *pages[INLINE_EXTENT_BUFFER_PAGES];
+#ifdef CONFIG_BTRFS_DEBUG
+	struct list_head leak_list;
+#endif
 };
 
-struct extent_map_tree;
+static inline void extent_set_compress_type(unsigned long *bio_flags,
+					    int compress_type)
+{
+	*bio_flags |= compress_type << EXTENT_BIO_FLAG_SHIFT;
+}
 
-static inline struct extent_state *extent_state_next(struct extent_state *state)
+static inline int extent_compress_type(unsigned long bio_flags)
 {
-	struct rb_node *node;
-	node = rb_next(&state->rb_node);
-	if (!node)
-		return NULL;
-	return rb_entry(node, struct extent_state, rb_node);
+	return bio_flags >> EXTENT_BIO_FLAG_SHIFT;
 }
 
+struct extent_map_tree;
+
 typedef struct extent_map *(get_extent_t)(struct inode *inode,
 					  struct page *page,
-					  size_t page_offset,
+					  size_t pg_offset,
 					  u64 start, u64 len,
 					  int create);
 
 void extent_io_tree_init(struct extent_io_tree *tree,
-			  struct address_space *mapping, gfp_t mask);
+			 struct address_space *mapping);
 int try_release_extent_mapping(struct extent_map_tree *map,
 			       struct extent_io_tree *tree, struct page *page,
 			       gfp_t mask);
-int try_release_extent_buffer(struct extent_io_tree *tree, struct page *page);
-int try_release_extent_state(struct extent_map_tree *map,
-			     struct extent_io_tree *tree, struct page *page,
-			     gfp_t mask);
-int lock_extent(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask);
+int try_release_extent_buffer(struct page *page);
+int lock_extent(struct extent_io_tree *tree, u64 start, u64 end);
 int lock_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
-		     int bits, struct extent_state **cached, gfp_t mask);
-int unlock_extent(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask);
+		     unsigned long bits, struct extent_state **cached);
+int unlock_extent(struct extent_io_tree *tree, u64 start, u64 end);
 int unlock_extent_cached(struct extent_io_tree *tree, u64 start, u64 end,
 			 struct extent_state **cached, gfp_t mask);
-int try_lock_extent(struct extent_io_tree *tree, u64 start, u64 end,
-		    gfp_t mask);
+int try_lock_extent(struct extent_io_tree *tree, u64 start, u64 end);
 int extent_read_full_page(struct extent_io_tree *tree, struct page *page,
-			  get_extent_t *get_extent);
+			  get_extent_t *get_extent, int mirror_num);
+int extent_read_full_page_nolock(struct extent_io_tree *tree, struct page *page,
+				 get_extent_t *get_extent, int mirror_num);
 int __init extent_io_init(void);
 void extent_io_exit(void);
 
 u64 count_range_bits(struct extent_io_tree *tree,
 		     u64 *start, u64 search_end,
-		     u64 max_bytes, unsigned long bits);
+		     u64 max_bytes, unsigned long bits, int contig);
 
+void free_extent_state(struct extent_state *state);
 int test_range_bit(struct extent_io_tree *tree, u64 start, u64 end,
-		   int bits, int filled, struct extent_state *cached_state);
+		   unsigned long bits, int filled,
+		   struct extent_state *cached_state);
 int clear_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
-		      int bits, gfp_t mask);
+		      unsigned long bits, gfp_t mask);
 int clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
-		     int bits, int wake, int delete, struct extent_state **cached,
-		     gfp_t mask);
+		     unsigned long bits, int wake, int delete,
+		     struct extent_state **cached, gfp_t mask);
 int set_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
-		    int bits, gfp_t mask);
+		    unsigned long bits, gfp_t mask);
+int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
+		   unsigned long bits, u64 *failed_start,
+		   struct extent_state **cached_state, gfp_t mask);
 int set_extent_uptodate(struct extent_io_tree *tree, u64 start, u64 end,
-			gfp_t mask);
+			struct extent_state **cached_state, gfp_t mask);
+int clear_extent_uptodate(struct extent_io_tree *tree, u64 start, u64 end,
+			  struct extent_state **cached_state, gfp_t mask);
 int set_extent_new(struct extent_io_tree *tree, u64 start, u64 end,
 		   gfp_t mask);
 int set_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end,
 		     gfp_t mask);
 int clear_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end,
 		       gfp_t mask);
-int clear_extent_ordered(struct extent_io_tree *tree, u64 start, u64 end,
-		       gfp_t mask);
-int clear_extent_ordered_metadata(struct extent_io_tree *tree, u64 start,
-				  u64 end, gfp_t mask);
+int convert_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
+		       unsigned long bits, unsigned long clear_bits,
+		       struct extent_state **cached_state, gfp_t mask);
 int set_extent_delalloc(struct extent_io_tree *tree, u64 start, u64 end,
 			struct extent_state **cached_state, gfp_t mask);
-int set_extent_ordered(struct extent_io_tree *tree, u64 start, u64 end,
-		     gfp_t mask);
+int set_extent_defrag(struct extent_io_tree *tree, u64 start, u64 end,
+		      struct extent_state **cached_state, gfp_t mask);
 int find_first_extent_bit(struct extent_io_tree *tree, u64 start,
-			  u64 *start_ret, u64 *end_ret, int bits);
-struct extent_state *find_first_extent_bit_state(struct extent_io_tree *tree,
-						 u64 start, int bits);
+			  u64 *start_ret, u64 *end_ret, unsigned long bits,
+			  struct extent_state **cached_state);
 int extent_invalidatepage(struct extent_io_tree *tree,
 			  struct page *page, unsigned long offset);
 int extent_write_full_page(struct extent_io_tree *tree, struct page *page,
@@ -217,36 +253,44 @@ int extent_writepages(struct extent_io_tree *tree,
 		      struct address_space *mapping,
 		      get_extent_t *get_extent,
 		      struct writeback_control *wbc);
+int btree_write_cache_pages(struct address_space *mapping,
+			    struct writeback_control *wbc);
 int extent_readpages(struct extent_io_tree *tree,
 		     struct address_space *mapping,
 		     struct list_head *pages, unsigned nr_pages,
 		     get_extent_t get_extent);
-int extent_prepare_write(struct extent_io_tree *tree,
-			 struct inode *inode, struct page *page,
-			 unsigned from, unsigned to, get_extent_t *get_extent);
-int extent_commit_write(struct extent_io_tree *tree,
-			struct inode *inode, struct page *page,
-			unsigned from, unsigned to);
-sector_t extent_bmap(struct address_space *mapping, sector_t iblock,
-		get_extent_t *get_extent);
 int extent_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
 		__u64 start, __u64 len, get_extent_t *get_extent);
-int set_range_dirty(struct extent_io_tree *tree, u64 start, u64 end);
-int set_state_private(struct extent_io_tree *tree, u64 start, u64 private);
 int get_state_private(struct extent_io_tree *tree, u64 start, u64 *private);
 void set_page_extent_mapped(struct page *page);
 
-struct extent_buffer *alloc_extent_buffer(struct extent_io_tree *tree,
-					  u64 start, unsigned long len,
-					  struct page *page0,
-					  gfp_t mask);
-struct extent_buffer *find_extent_buffer(struct extent_io_tree *tree,
-					 u64 start, unsigned long len,
-					  gfp_t mask);
+struct extent_buffer *alloc_extent_buffer(struct btrfs_fs_info *fs_info,
+					  u64 start, unsigned long len);
+struct extent_buffer *alloc_dummy_extent_buffer(u64 start, unsigned long len);
+struct extent_buffer *btrfs_clone_extent_buffer(struct extent_buffer *src);
+struct extent_buffer *find_extent_buffer(struct btrfs_fs_info *fs_info,
+					 u64 start);
 void free_extent_buffer(struct extent_buffer *eb);
+void free_extent_buffer_stale(struct extent_buffer *eb);
+#define WAIT_NONE	0
+#define WAIT_COMPLETE	1
+#define WAIT_PAGE_LOCK	2
 int read_extent_buffer_pages(struct extent_io_tree *tree,
 			     struct extent_buffer *eb, u64 start, int wait,
 			     get_extent_t *get_extent, int mirror_num);
+void wait_on_extent_buffer_writeback(struct extent_buffer *eb);
+
+static inline unsigned long num_extent_pages(u64 start, u64 len)
+{
+	return ((start + len + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT) -
+		(start >> PAGE_CACHE_SHIFT);
+}
+
+static inline struct page *extent_buffer_page(struct extent_buffer *eb,
+					      unsigned long i)
+{
+	return eb->pages[i];
+}
 
 static inline void extent_buffer_get(struct extent_buffer *eb)
 {
@@ -259,6 +303,9 @@ int memcmp_extent_buffer(struct extent_buffer *eb, const void *ptrv,
 void read_extent_buffer(struct extent_buffer *eb, void *dst,
 			unsigned long start,
 			unsigned long len);
+int read_extent_buffer_to_user(struct extent_buffer *eb, void __user *dst,
+			       unsigned long start,
+			       unsigned long len);
 void write_extent_buffer(struct extent_buffer *eb, const void *src,
 			 unsigned long start, unsigned long len);
 void copy_extent_buffer(struct extent_buffer *dst, struct extent_buffer *src,
@@ -270,38 +317,42 @@ void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
 			   unsigned long src_offset, unsigned long len);
 void memset_extent_buffer(struct extent_buffer *eb, char c,
 			  unsigned long start, unsigned long len);
-int wait_on_extent_buffer_writeback(struct extent_io_tree *tree,
-				    struct extent_buffer *eb);
-int wait_on_extent_writeback(struct extent_io_tree *tree, u64 start, u64 end);
-int wait_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, int bits);
-int clear_extent_buffer_dirty(struct extent_io_tree *tree,
-			      struct extent_buffer *eb);
-int set_extent_buffer_dirty(struct extent_io_tree *tree,
-			     struct extent_buffer *eb);
-int test_extent_buffer_dirty(struct extent_io_tree *tree,
-			     struct extent_buffer *eb);
-int set_extent_buffer_uptodate(struct extent_io_tree *tree,
-			       struct extent_buffer *eb);
-int clear_extent_buffer_uptodate(struct extent_io_tree *tree,
-				struct extent_buffer *eb,
-				struct extent_state **cached_state);
-int extent_buffer_uptodate(struct extent_io_tree *tree,
-			   struct extent_buffer *eb,
-			   struct extent_state *cached_state);
-int map_extent_buffer(struct extent_buffer *eb, unsigned long offset,
-		      unsigned long min_len, char **token, char **map,
-		      unsigned long *map_start,
-		      unsigned long *map_len, int km);
+void clear_extent_buffer_dirty(struct extent_buffer *eb);
+int set_extent_buffer_dirty(struct extent_buffer *eb);
+int set_extent_buffer_uptodate(struct extent_buffer *eb);
+int clear_extent_buffer_uptodate(struct extent_buffer *eb);
+int extent_buffer_uptodate(struct extent_buffer *eb);
+int extent_buffer_under_io(struct extent_buffer *eb);
 int map_private_extent_buffer(struct extent_buffer *eb, unsigned long offset,
-		      unsigned long min_len, char **token, char **map,
+		      unsigned long min_len, char **map,
 		      unsigned long *map_start,
-		      unsigned long *map_len, int km);
-void unmap_extent_buffer(struct extent_buffer *eb, char *token, int km);
-int release_extent_buffer_tail_pages(struct extent_buffer *eb);
-int extent_range_uptodate(struct extent_io_tree *tree,
-			  u64 start, u64 end);
-int extent_clear_unlock_delalloc(struct inode *inode,
-				struct extent_io_tree *tree,
-				u64 start, u64 end, struct page *locked_page,
-				unsigned long op);
+		      unsigned long *map_len);
+int extent_range_clear_dirty_for_io(struct inode *inode, u64 start, u64 end);
+int extent_range_redirty_for_io(struct inode *inode, u64 start, u64 end);
+int extent_clear_unlock_delalloc(struct inode *inode, u64 start, u64 end,
+				 struct page *locked_page,
+				 unsigned long bits_to_clear,
+				 unsigned long page_ops);
+struct bio *
+btrfs_bio_alloc(struct block_device *bdev, u64 first_sector, int nr_vecs,
+		gfp_t gfp_flags);
+struct bio *btrfs_io_bio_alloc(gfp_t gfp_mask, unsigned int nr_iovecs);
+struct bio *btrfs_bio_clone(struct bio *bio, gfp_t gfp_mask);
+
+struct btrfs_fs_info;
+
+int repair_io_failure(struct btrfs_fs_info *fs_info, u64 start,
+			u64 length, u64 logical, struct page *page,
+			int mirror_num);
+int end_extent_writepage(struct page *page, int err, u64 start, u64 end);
+int repair_eb_io_failure(struct btrfs_root *root, struct extent_buffer *eb,
+			 int mirror_num);
+#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
+noinline u64 find_lock_delalloc_range(struct inode *inode,
+				      struct extent_io_tree *tree,
+				      struct page *locked_page, u64 *start,
+				      u64 *end, u64 max_bytes);
+struct extent_buffer *alloc_test_extent_buffer(struct btrfs_fs_info *fs_info,
+					       u64 start, unsigned long len);
+#endif
 #endif
diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c
index 454ca52d645..225302b39af 100644
--- a/fs/btrfs/extent_map.c
+++ b/fs/btrfs/extent_map.c
@@ -1,8 +1,8 @@
 #include <linux/err.h>
 #include <linux/slab.h>
-#include <linux/module.h>
 #include <linux/spinlock.h>
 #include <linux/hardirq.h>
+#include "ctree.h"
 #include "extent_map.h"
 
 
@@ -10,7 +10,7 @@ static struct kmem_cache *extent_map_cache;
 
 int __init extent_map_init(void)
 {
-	extent_map_cache = kmem_cache_create("extent_map",
+	extent_map_cache = kmem_cache_create("btrfs_extent_map",
 			sizeof(struct extent_map), 0,
 			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
 	if (!extent_map_cache)
@@ -27,34 +27,36 @@ void extent_map_exit(void)
 /**
  * extent_map_tree_init - initialize extent map tree
  * @tree:		tree to initialize
- * @mask:		flags for memory allocations during tree operations
  *
  * Initialize the extent tree @tree.  Should be called for each new inode
  * or other user of the extent_map interface.
  */
-void extent_map_tree_init(struct extent_map_tree *tree, gfp_t mask)
+void extent_map_tree_init(struct extent_map_tree *tree)
 {
 	tree->map = RB_ROOT;
+	INIT_LIST_HEAD(&tree->modified_extents);
 	rwlock_init(&tree->lock);
 }
 
 /**
  * alloc_extent_map - allocate new extent map structure
- * @mask:	memory allocation flags
  *
  * Allocate a new extent_map structure.  The new structure is
  * returned with a reference count of one and needs to be
  * freed using free_extent_map()
  */
-struct extent_map *alloc_extent_map(gfp_t mask)
+struct extent_map *alloc_extent_map(void)
 {
 	struct extent_map *em;
-	em = kmem_cache_alloc(extent_map_cache, mask);
-	if (!em || IS_ERR(em))
-		return em;
-	em->in_tree = 0;
+	em = kmem_cache_zalloc(extent_map_cache, GFP_NOFS);
+	if (!em)
+		return NULL;
+	RB_CLEAR_NODE(&em->rb_node);
 	em->flags = 0;
+	em->compress_type = BTRFS_COMPRESS_NONE;
+	em->generation = 0;
 	atomic_set(&em->refs, 1);
+	INIT_LIST_HEAD(&em->list);
 	return em;
 }
 
@@ -71,37 +73,64 @@ void free_extent_map(struct extent_map *em)
 		return;
 	WARN_ON(atomic_read(&em->refs) == 0);
 	if (atomic_dec_and_test(&em->refs)) {
-		WARN_ON(em->in_tree);
+		WARN_ON(extent_map_in_tree(em));
+		WARN_ON(!list_empty(&em->list));
+		if (test_bit(EXTENT_FLAG_FS_MAPPING, &em->flags))
+			kfree(em->bdev);
 		kmem_cache_free(extent_map_cache, em);
 	}
 }
 
-static struct rb_node *tree_insert(struct rb_root *root, u64 offset,
-				   struct rb_node *node)
+/* simple helper to do math around the end of an extent, handling wrap */
+static u64 range_end(u64 start, u64 len)
+{
+	if (start + len < start)
+		return (u64)-1;
+	return start + len;
+}
+
+static int tree_insert(struct rb_root *root, struct extent_map *em)
 {
 	struct rb_node **p = &root->rb_node;
 	struct rb_node *parent = NULL;
-	struct extent_map *entry;
+	struct extent_map *entry = NULL;
+	struct rb_node *orig_parent = NULL;
+	u64 end = range_end(em->start, em->len);
 
 	while (*p) {
 		parent = *p;
 		entry = rb_entry(parent, struct extent_map, rb_node);
 
-		WARN_ON(!entry->in_tree);
-
-		if (offset < entry->start)
+		if (em->start < entry->start)
 			p = &(*p)->rb_left;
-		else if (offset >= extent_map_end(entry))
+		else if (em->start >= extent_map_end(entry))
 			p = &(*p)->rb_right;
 		else
-			return parent;
+			return -EEXIST;
 	}
 
-	entry = rb_entry(node, struct extent_map, rb_node);
-	entry->in_tree = 1;
-	rb_link_node(node, parent, p);
-	rb_insert_color(node, root);
-	return NULL;
+	orig_parent = parent;
+	while (parent && em->start >= extent_map_end(entry)) {
+		parent = rb_next(parent);
+		entry = rb_entry(parent, struct extent_map, rb_node);
+	}
+	if (parent)
+		if (end > entry->start && em->start < extent_map_end(entry))
+			return -EEXIST;
+
+	parent = orig_parent;
+	entry = rb_entry(parent, struct extent_map, rb_node);
+	while (parent && em->start < entry->start) {
+		parent = rb_prev(parent);
+		entry = rb_entry(parent, struct extent_map, rb_node);
+	}
+	if (parent)
+		if (end > entry->start && em->start < extent_map_end(entry))
+			return -EEXIST;
+
+	rb_link_node(&em->rb_node, orig_parent, p);
+	rb_insert_color(&em->rb_node, root);
+	return 0;
 }
 
 /*
@@ -123,8 +152,6 @@ static struct rb_node *__tree_search(struct rb_root *root, u64 offset,
 		prev = n;
 		prev_entry = entry;
 
-		WARN_ON(!entry->in_tree);
-
 		if (offset < entry->start)
 			n = n->rb_left;
 		else if (offset >= extent_map_end(entry))
@@ -167,6 +194,18 @@ static int mergable_maps(struct extent_map *prev, struct extent_map *next)
 	if (test_bit(EXTENT_FLAG_COMPRESSED, &prev->flags))
 		return 0;
 
+	if (test_bit(EXTENT_FLAG_LOGGING, &prev->flags) ||
+	    test_bit(EXTENT_FLAG_LOGGING, &next->flags))
+		return 0;
+
+	/*
+	 * We don't want to merge stuff that hasn't been written to the log yet
+	 * since it may not reflect exactly what is on disk, and that would be
+	 * bad.
+	 */
+	if (!list_empty(&prev->list) || !list_empty(&next->list))
+		return 0;
+
 	if (extent_map_end(prev) == next->start &&
 	    prev->flags == next->flags &&
 	    prev->bdev == next->bdev &&
@@ -183,22 +222,10 @@ static int mergable_maps(struct extent_map *prev, struct extent_map *next)
 	return 0;
 }
 
-int unpin_extent_cache(struct extent_map_tree *tree, u64 start, u64 len)
+static void try_merge_map(struct extent_map_tree *tree, struct extent_map *em)
 {
-	int ret = 0;
 	struct extent_map *merge = NULL;
 	struct rb_node *rb;
-	struct extent_map *em;
-
-	write_lock(&tree->lock);
-	em = lookup_extent_mapping(tree, start, len);
-
-	WARN_ON(!em || em->start != start);
-
-	if (!em)
-		goto out;
-
-	clear_bit(EXTENT_FLAG_PINNED, &em->flags);
 
 	if (em->start != 0) {
 		rb = rb_prev(&em->rb_node);
@@ -206,11 +233,16 @@ int unpin_extent_cache(struct extent_map_tree *tree, u64 start, u64 len)
 			merge = rb_entry(rb, struct extent_map, rb_node);
 		if (rb && mergable_maps(merge, em)) {
 			em->start = merge->start;
+			em->orig_start = merge->orig_start;
 			em->len += merge->len;
 			em->block_len += merge->block_len;
 			em->block_start = merge->block_start;
-			merge->in_tree = 0;
+			em->mod_len = (em->mod_len + em->mod_start) - merge->mod_start;
+			em->mod_start = merge->mod_start;
+			em->generation = max(em->generation, merge->generation);
+
 			rb_erase(&merge->rb_node, &tree->map);
+			RB_CLEAR_NODE(&merge->rb_node);
 			free_extent_map(merge);
 		}
 	}
@@ -220,11 +252,59 @@ int unpin_extent_cache(struct extent_map_tree *tree, u64 start, u64 len)
 		merge = rb_entry(rb, struct extent_map, rb_node);
 	if (rb && mergable_maps(em, merge)) {
 		em->len += merge->len;
-		em->block_len += merge->len;
+		em->block_len += merge->block_len;
 		rb_erase(&merge->rb_node, &tree->map);
-		merge->in_tree = 0;
+		RB_CLEAR_NODE(&merge->rb_node);
+		em->mod_len = (merge->mod_start + merge->mod_len) - em->mod_start;
+		em->generation = max(em->generation, merge->generation);
 		free_extent_map(merge);
 	}
+}
+
+/**
+ * unpin_extent_cache - unpin an extent from the cache
+ * @tree:	tree to unpin the extent in
+ * @start:	logical offset in the file
+ * @len:	length of the extent
+ * @gen:	generation that this extent has been modified in
+ *
+ * Called after an extent has been written to disk properly.  Set the generation
+ * to the generation that actually added the file item to the inode so we know
+ * we need to sync this extent when we call fsync().
+ */
+int unpin_extent_cache(struct extent_map_tree *tree, u64 start, u64 len,
+		       u64 gen)
+{
+	int ret = 0;
+	struct extent_map *em;
+	bool prealloc = false;
+
+	write_lock(&tree->lock);
+	em = lookup_extent_mapping(tree, start, len);
+
+	WARN_ON(!em || em->start != start);
+
+	if (!em)
+		goto out;
+
+	if (!test_bit(EXTENT_FLAG_LOGGING, &em->flags))
+		list_move(&em->list, &tree->modified_extents);
+	em->generation = gen;
+	clear_bit(EXTENT_FLAG_PINNED, &em->flags);
+	em->mod_start = em->start;
+	em->mod_len = em->len;
+
+	if (test_bit(EXTENT_FLAG_FILLING, &em->flags)) {
+		prealloc = true;
+		clear_bit(EXTENT_FLAG_FILLING, &em->flags);
+	}
+
+	try_merge_map(tree, em);
+
+	if (prealloc) {
+		em->mod_start = em->start;
+		em->mod_len = em->len;
+	}
 
 	free_extent_map(em);
 out:
@@ -233,6 +313,27 @@ out:
 
 }
 
+void clear_em_logging(struct extent_map_tree *tree, struct extent_map *em)
+{
+	clear_bit(EXTENT_FLAG_LOGGING, &em->flags);
+	if (extent_map_in_tree(em))
+		try_merge_map(tree, em);
+}
+
+static inline void setup_extent_mapping(struct extent_map_tree *tree,
+					struct extent_map *em,
+					int modified)
+{
+	atomic_inc(&em->refs);
+	em->mod_start = em->start;
+	em->mod_len = em->len;
+
+	if (modified)
+		list_move(&em->list, &tree->modified_extents);
+	else
+		try_merge_map(tree, em);
+}
+
 /**
  * add_extent_mapping - add new extent map to the extent tree
  * @tree:	tree to insert new map in
@@ -241,62 +342,49 @@ out:
  * Insert @em into @tree or perform a simple forward/backward merge with
  * existing mappings.  The extent_map struct passed in will be inserted
  * into the tree directly, with an additional reference taken, or a
- * reference dropped if the merge attempt was successfull.
+ * reference dropped if the merge attempt was successful.
  */
 int add_extent_mapping(struct extent_map_tree *tree,
-		       struct extent_map *em)
+		       struct extent_map *em, int modified)
 {
 	int ret = 0;
-	struct extent_map *merge = NULL;
-	struct rb_node *rb;
-	struct extent_map *exist;
 
-	exist = lookup_extent_mapping(tree, em->start, em->len);
-	if (exist) {
-		free_extent_map(exist);
-		ret = -EEXIST;
+	ret = tree_insert(&tree->map, em);
+	if (ret)
 		goto out;
-	}
-	rb = tree_insert(&tree->map, em->start, &em->rb_node);
-	if (rb) {
-		ret = -EEXIST;
-		goto out;
-	}
-	atomic_inc(&em->refs);
-	if (em->start != 0) {
-		rb = rb_prev(&em->rb_node);
-		if (rb)
-			merge = rb_entry(rb, struct extent_map, rb_node);
-		if (rb && mergable_maps(merge, em)) {
-			em->start = merge->start;
-			em->len += merge->len;
-			em->block_len += merge->block_len;
-			em->block_start = merge->block_start;
-			merge->in_tree = 0;
-			rb_erase(&merge->rb_node, &tree->map);
-			free_extent_map(merge);
-		}
-	 }
-	rb = rb_next(&em->rb_node);
-	if (rb)
-		merge = rb_entry(rb, struct extent_map, rb_node);
-	if (rb && mergable_maps(em, merge)) {
-		em->len += merge->len;
-		em->block_len += merge->len;
-		rb_erase(&merge->rb_node, &tree->map);
-		merge->in_tree = 0;
-		free_extent_map(merge);
-	}
+
+	setup_extent_mapping(tree, em, modified);
 out:
 	return ret;
 }
 
-/* simple helper to do math around the end of an extent, handling wrap */
-static u64 range_end(u64 start, u64 len)
+static struct extent_map *
+__lookup_extent_mapping(struct extent_map_tree *tree,
+			u64 start, u64 len, int strict)
 {
-	if (start + len < start)
-		return (u64)-1;
-	return start + len;
+	struct extent_map *em;
+	struct rb_node *rb_node;
+	struct rb_node *prev = NULL;
+	struct rb_node *next = NULL;
+	u64 end = range_end(start, len);
+
+	rb_node = __tree_search(&tree->map, start, &prev, &next);
+	if (!rb_node) {
+		if (prev)
+			rb_node = prev;
+		else if (next)
+			rb_node = next;
+		else
+			return NULL;
+	}
+
+	em = rb_entry(rb_node, struct extent_map, rb_node);
+
+	if (strict && !(end > em->start && start < extent_map_end(em)))
+		return NULL;
+
+	atomic_inc(&em->refs);
+	return em;
 }
 
 /**
@@ -313,42 +401,7 @@ static u64 range_end(u64 start, u64 len)
 struct extent_map *lookup_extent_mapping(struct extent_map_tree *tree,
 					 u64 start, u64 len)
 {
-	struct extent_map *em;
-	struct rb_node *rb_node;
-	struct rb_node *prev = NULL;
-	struct rb_node *next = NULL;
-	u64 end = range_end(start, len);
-
-	rb_node = __tree_search(&tree->map, start, &prev, &next);
-	if (!rb_node && prev) {
-		em = rb_entry(prev, struct extent_map, rb_node);
-		if (end > em->start && start < extent_map_end(em))
-			goto found;
-	}
-	if (!rb_node && next) {
-		em = rb_entry(next, struct extent_map, rb_node);
-		if (end > em->start && start < extent_map_end(em))
-			goto found;
-	}
-	if (!rb_node) {
-		em = NULL;
-		goto out;
-	}
-	if (IS_ERR(rb_node)) {
-		em = ERR_PTR(PTR_ERR(rb_node));
-		goto out;
-	}
-	em = rb_entry(rb_node, struct extent_map, rb_node);
-	if (end > em->start && start < extent_map_end(em))
-		goto found;
-
-	em = NULL;
-	goto out;
-
-found:
-	atomic_inc(&em->refs);
-out:
-	return em;
+	return __lookup_extent_mapping(tree, start, len, 1);
 }
 
 /**
@@ -365,38 +418,7 @@ out:
 struct extent_map *search_extent_mapping(struct extent_map_tree *tree,
 					 u64 start, u64 len)
 {
-	struct extent_map *em;
-	struct rb_node *rb_node;
-	struct rb_node *prev = NULL;
-	struct rb_node *next = NULL;
-
-	rb_node = __tree_search(&tree->map, start, &prev, &next);
-	if (!rb_node && prev) {
-		em = rb_entry(prev, struct extent_map, rb_node);
-		goto found;
-	}
-	if (!rb_node && next) {
-		em = rb_entry(next, struct extent_map, rb_node);
-		goto found;
-	}
-	if (!rb_node) {
-		em = NULL;
-		goto out;
-	}
-	if (IS_ERR(rb_node)) {
-		em = ERR_PTR(PTR_ERR(rb_node));
-		goto out;
-	}
-	em = rb_entry(rb_node, struct extent_map, rb_node);
-	goto found;
-
-	em = NULL;
-	goto out;
-
-found:
-	atomic_inc(&em->refs);
-out:
-	return em;
+	return __lookup_extent_mapping(tree, start, len, 0);
 }
 
 /**
@@ -413,6 +435,23 @@ int remove_extent_mapping(struct extent_map_tree *tree, struct extent_map *em)
 
 	WARN_ON(test_bit(EXTENT_FLAG_PINNED, &em->flags));
 	rb_erase(&em->rb_node, &tree->map);
-	em->in_tree = 0;
+	if (!test_bit(EXTENT_FLAG_LOGGING, &em->flags))
+		list_del_init(&em->list);
+	RB_CLEAR_NODE(&em->rb_node);
 	return ret;
 }
+
+void replace_extent_mapping(struct extent_map_tree *tree,
+			    struct extent_map *cur,
+			    struct extent_map *new,
+			    int modified)
+{
+	WARN_ON(test_bit(EXTENT_FLAG_PINNED, &cur->flags));
+	ASSERT(extent_map_in_tree(cur));
+	if (!test_bit(EXTENT_FLAG_LOGGING, &cur->flags))
+		list_del_init(&cur->list);
+	rb_replace_node(&cur->rb_node, &new->rb_node, &tree->map);
+	RB_CLEAR_NODE(&cur->rb_node);
+
+	setup_extent_mapping(tree, new, modified);
+}
diff --git a/fs/btrfs/extent_map.h b/fs/btrfs/extent_map.h
index ab6d74b6e64..b2991fd8583 100644
--- a/fs/btrfs/extent_map.h
+++ b/fs/btrfs/extent_map.h
@@ -3,16 +3,19 @@
 
 #include <linux/rbtree.h>
 
-#define EXTENT_MAP_LAST_BYTE (u64)-4
-#define EXTENT_MAP_HOLE (u64)-3
-#define EXTENT_MAP_INLINE (u64)-2
-#define EXTENT_MAP_DELALLOC (u64)-1
+#define EXTENT_MAP_LAST_BYTE ((u64)-4)
+#define EXTENT_MAP_HOLE ((u64)-3)
+#define EXTENT_MAP_INLINE ((u64)-2)
+#define EXTENT_MAP_DELALLOC ((u64)-1)
 
 /* bits for the flags field */
 #define EXTENT_FLAG_PINNED 0 /* this entry not yet on disk, don't free it */
 #define EXTENT_FLAG_COMPRESSED 1
 #define EXTENT_FLAG_VACANCY 2 /* no file extent item found */
 #define EXTENT_FLAG_PREALLOC 3 /* pre-allocated extent */
+#define EXTENT_FLAG_LOGGING 4 /* Logging this extent */
+#define EXTENT_FLAG_FILLING 5 /* Filling in a preallocated extent */
+#define EXTENT_FLAG_FS_MAPPING 6 /* filesystem extent mapping type */
 
 struct extent_map {
 	struct rb_node rb_node;
@@ -20,20 +23,32 @@ struct extent_map {
 	/* all of these are in bytes */
 	u64 start;
 	u64 len;
+	u64 mod_start;
+	u64 mod_len;
 	u64 orig_start;
+	u64 orig_block_len;
+	u64 ram_bytes;
 	u64 block_start;
 	u64 block_len;
+	u64 generation;
 	unsigned long flags;
 	struct block_device *bdev;
 	atomic_t refs;
-	int in_tree;
+	unsigned int compress_type;
+	struct list_head list;
 };
 
 struct extent_map_tree {
 	struct rb_root map;
+	struct list_head modified_extents;
 	rwlock_t lock;
 };
 
+static inline int extent_map_in_tree(const struct extent_map *em)
+{
+	return !RB_EMPTY_NODE(&em->rb_node);
+}
+
 static inline u64 extent_map_end(struct extent_map *em)
 {
 	if (em->start + em->len < em->start)
@@ -48,18 +63,23 @@ static inline u64 extent_map_block_end(struct extent_map *em)
 	return em->block_start + em->block_len;
 }
 
-void extent_map_tree_init(struct extent_map_tree *tree, gfp_t mask);
+void extent_map_tree_init(struct extent_map_tree *tree);
 struct extent_map *lookup_extent_mapping(struct extent_map_tree *tree,
 					 u64 start, u64 len);
 int add_extent_mapping(struct extent_map_tree *tree,
-		       struct extent_map *em);
+		       struct extent_map *em, int modified);
 int remove_extent_mapping(struct extent_map_tree *tree, struct extent_map *em);
+void replace_extent_mapping(struct extent_map_tree *tree,
+			    struct extent_map *cur,
+			    struct extent_map *new,
+			    int modified);
 
-struct extent_map *alloc_extent_map(gfp_t mask);
+struct extent_map *alloc_extent_map(void);
 void free_extent_map(struct extent_map *em);
 int __init extent_map_init(void);
 void extent_map_exit(void);
-int unpin_extent_cache(struct extent_map_tree *tree, u64 start, u64 len);
+int unpin_extent_cache(struct extent_map_tree *tree, u64 start, u64 len, u64 gen);
+void clear_em_logging(struct extent_map_tree *tree, struct extent_map *em);
 struct extent_map *search_extent_mapping(struct extent_map_tree *tree,
 					 u64 start, u64 len);
 #endif
diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c
index 54a255065aa..f46cfe45d68 100644
--- a/fs/btrfs/file-item.c
+++ b/fs/btrfs/file-item.c
@@ -23,16 +23,19 @@
 #include "ctree.h"
 #include "disk-io.h"
 #include "transaction.h"
+#include "volumes.h"
 #include "print-tree.h"
 
-#define MAX_CSUM_ITEMS(r, size) ((((BTRFS_LEAF_DATA_SIZE(r) - \
+#define __MAX_CSUM_ITEMS(r, size) ((unsigned long)(((BTRFS_LEAF_DATA_SIZE(r) - \
 				   sizeof(struct btrfs_item) * 2) / \
 				  size) - 1))
 
+#define MAX_CSUM_ITEMS(r, size) (min_t(u32, __MAX_CSUM_ITEMS(r, size), \
+				       PAGE_CACHE_SIZE))
+
 #define MAX_ORDERED_SUM_BYTES(r) ((PAGE_SIZE - \
 				   sizeof(struct btrfs_ordered_sum)) / \
-				   sizeof(struct btrfs_sector_sum) * \
-				   (r)->sectorsize - (r)->sectorsize)
+				   sizeof(u32) * (r)->sectorsize)
 
 int btrfs_insert_file_extent(struct btrfs_trans_handle *trans,
 			     struct btrfs_root *root,
@@ -48,7 +51,8 @@ int btrfs_insert_file_extent(struct btrfs_trans_handle *trans,
 	struct extent_buffer *leaf;
 
 	path = btrfs_alloc_path();
-	BUG_ON(!path);
+	if (!path)
+		return -ENOMEM;
 	file_key.objectid = objectid;
 	file_key.offset = pos;
 	btrfs_set_key_type(&file_key, BTRFS_EXTENT_DATA_KEY);
@@ -58,7 +62,7 @@ int btrfs_insert_file_extent(struct btrfs_trans_handle *trans,
 				      sizeof(*item));
 	if (ret < 0)
 		goto out;
-	BUG_ON(ret);
+	BUG_ON(ret); /* Can't happen */
 	leaf = path->nodes[0];
 	item = btrfs_item_ptr(leaf, path->slots[0],
 			      struct btrfs_file_extent_item);
@@ -79,10 +83,11 @@ out:
 	return ret;
 }
 
-struct btrfs_csum_item *btrfs_lookup_csum(struct btrfs_trans_handle *trans,
-					  struct btrfs_root *root,
-					  struct btrfs_path *path,
-					  u64 bytenr, int cow)
+static struct btrfs_csum_item *
+btrfs_lookup_csum(struct btrfs_trans_handle *trans,
+		  struct btrfs_root *root,
+		  struct btrfs_path *path,
+		  u64 bytenr, int cow)
 {
 	int ret;
 	struct btrfs_key file_key;
@@ -90,8 +95,7 @@ struct btrfs_csum_item *btrfs_lookup_csum(struct btrfs_trans_handle *trans,
 	struct btrfs_csum_item *item;
 	struct extent_buffer *leaf;
 	u64 csum_offset = 0;
-	u16 csum_size =
-		btrfs_super_csum_size(&root->fs_info->super_copy);
+	u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
 	int csums_in_item;
 
 	file_key.objectid = BTRFS_EXTENT_CSUM_OBJECTID;
@@ -115,9 +119,11 @@ struct btrfs_csum_item *btrfs_lookup_csum(struct btrfs_trans_handle *trans,
 		csums_in_item = btrfs_item_size_nr(leaf, path->slots[0]);
 		csums_in_item /= csum_size;
 
-		if (csum_offset >= csums_in_item) {
+		if (csum_offset == csums_in_item) {
 			ret = -EFBIG;
 			goto fail;
+		} else if (csum_offset > csums_in_item) {
+			goto fail;
 		}
 	}
 	item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_csum_item);
@@ -130,7 +136,6 @@ fail:
 	return ERR_PTR(ret);
 }
 
-
 int btrfs_lookup_file_extent(struct btrfs_trans_handle *trans,
 			     struct btrfs_root *root,
 			     struct btrfs_path *path, u64 objectid,
@@ -148,36 +153,79 @@ int btrfs_lookup_file_extent(struct btrfs_trans_handle *trans,
 	return ret;
 }
 
+static void btrfs_io_bio_endio_readpage(struct btrfs_io_bio *bio, int err)
+{
+	kfree(bio->csum_allocated);
+}
 
-int btrfs_lookup_bio_sums(struct btrfs_root *root, struct inode *inode,
-			  struct bio *bio, u32 *dst)
+static int __btrfs_lookup_bio_sums(struct btrfs_root *root,
+				   struct inode *inode, struct bio *bio,
+				   u64 logical_offset, u32 *dst, int dio)
 {
-	u32 sum;
 	struct bio_vec *bvec = bio->bi_io_vec;
-	int bio_index = 0;
-	u64 offset;
+	struct btrfs_io_bio *btrfs_bio = btrfs_io_bio(bio);
+	struct btrfs_csum_item *item = NULL;
+	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
+	struct btrfs_path *path;
+	u8 *csum;
+	u64 offset = 0;
 	u64 item_start_offset = 0;
 	u64 item_last_offset = 0;
 	u64 disk_bytenr;
 	u32 diff;
-	u16 csum_size =
-		btrfs_super_csum_size(&root->fs_info->super_copy);
-	int ret;
-	struct btrfs_path *path;
-	struct btrfs_csum_item *item = NULL;
-	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
+	int nblocks;
+	int bio_index = 0;
+	int count;
+	u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
 
 	path = btrfs_alloc_path();
-	if (bio->bi_size > PAGE_CACHE_SIZE * 8)
+	if (!path)
+		return -ENOMEM;
+
+	nblocks = bio->bi_iter.bi_size >> inode->i_sb->s_blocksize_bits;
+	if (!dst) {
+		if (nblocks * csum_size > BTRFS_BIO_INLINE_CSUM_SIZE) {
+			btrfs_bio->csum_allocated = kmalloc(nblocks * csum_size,
+							    GFP_NOFS);
+			if (!btrfs_bio->csum_allocated) {
+				btrfs_free_path(path);
+				return -ENOMEM;
+			}
+			btrfs_bio->csum = btrfs_bio->csum_allocated;
+			btrfs_bio->end_io = btrfs_io_bio_endio_readpage;
+		} else {
+			btrfs_bio->csum = btrfs_bio->csum_inline;
+		}
+		csum = btrfs_bio->csum;
+	} else {
+		csum = (u8 *)dst;
+	}
+
+	if (bio->bi_iter.bi_size > PAGE_CACHE_SIZE * 8)
 		path->reada = 2;
 
 	WARN_ON(bio->bi_vcnt <= 0);
 
-	disk_bytenr = (u64)bio->bi_sector << 9;
+	/*
+	 * the free space stuff is only read when it hasn't been
+	 * updated in the current transaction.  So, we can safely
+	 * read from the commit root and sidestep a nasty deadlock
+	 * between reading the free space cache and updating the csum tree.
+	 */
+	if (btrfs_is_free_space_inode(inode)) {
+		path->search_commit_root = 1;
+		path->skip_locking = 1;
+	}
+
+	disk_bytenr = (u64)bio->bi_iter.bi_sector << 9;
+	if (dio)
+		offset = logical_offset;
 	while (bio_index < bio->bi_vcnt) {
-		offset = page_offset(bvec->bv_page) + bvec->bv_offset;
-		ret = btrfs_find_ordered_sum(inode, offset, disk_bytenr, &sum);
-		if (ret == 0)
+		if (!dio)
+			offset = page_offset(bvec->bv_page) + bvec->bv_offset;
+		count = btrfs_find_ordered_sum(inode, offset, disk_bytenr,
+					       (u32 *)csum, nblocks);
+		if (count)
 			goto found;
 
 		if (!item || disk_bytenr < item_start_offset ||
@@ -186,27 +234,24 @@ int btrfs_lookup_bio_sums(struct btrfs_root *root, struct inode *inode,
 			u32 item_size;
 
 			if (item)
-				btrfs_release_path(root, path);
+				btrfs_release_path(path);
 			item = btrfs_lookup_csum(NULL, root->fs_info->csum_root,
 						 path, disk_bytenr, 0);
 			if (IS_ERR(item)) {
-				ret = PTR_ERR(item);
-				if (ret == -ENOENT || ret == -EFBIG)
-					ret = 0;
-				sum = 0;
+				count = 1;
+				memset(csum, 0, csum_size);
 				if (BTRFS_I(inode)->root->root_key.objectid ==
 				    BTRFS_DATA_RELOC_TREE_OBJECTID) {
 					set_extent_bits(io_tree, offset,
 						offset + bvec->bv_len - 1,
 						EXTENT_NODATASUM, GFP_NOFS);
 				} else {
-					printk(KERN_INFO "btrfs no csum found "
-					       "for inode %lu start %llu\n",
-					       inode->i_ino,
-					       (unsigned long long)offset);
+					btrfs_info(BTRFS_I(inode)->root->fs_info,
+						   "no csum found for inode %llu start %llu",
+					       btrfs_ino(inode), offset);
 				}
 				item = NULL;
-				btrfs_release_path(root, path);
+				btrfs_release_path(path);
 				goto found;
 			}
 			btrfs_item_key_to_cpu(path->nodes[0], &found_key,
@@ -228,40 +273,74 @@ int btrfs_lookup_bio_sums(struct btrfs_root *root, struct inode *inode,
 		diff = disk_bytenr - item_start_offset;
 		diff = diff / root->sectorsize;
 		diff = diff * csum_size;
-
-		read_extent_buffer(path->nodes[0], &sum,
+		count = min_t(int, nblocks, (item_last_offset - disk_bytenr) >>
+					    inode->i_sb->s_blocksize_bits);
+		read_extent_buffer(path->nodes[0], csum,
 				   ((unsigned long)item) + diff,
-				   csum_size);
+				   csum_size * count);
 found:
-		if (dst)
-			*dst++ = sum;
-		else
-			set_state_private(io_tree, offset, sum);
-		disk_bytenr += bvec->bv_len;
-		bio_index++;
-		bvec++;
+		csum += count * csum_size;
+		nblocks -= count;
+		bio_index += count;
+		while (count--) {
+			disk_bytenr += bvec->bv_len;
+			offset += bvec->bv_len;
+			bvec++;
+		}
 	}
 	btrfs_free_path(path);
 	return 0;
 }
 
+int btrfs_lookup_bio_sums(struct btrfs_root *root, struct inode *inode,
+			  struct bio *bio, u32 *dst)
+{
+	return __btrfs_lookup_bio_sums(root, inode, bio, 0, dst, 0);
+}
+
+int btrfs_lookup_bio_sums_dio(struct btrfs_root *root, struct inode *inode,
+			      struct btrfs_dio_private *dip, struct bio *bio,
+			      u64 offset)
+{
+	int len = (bio->bi_iter.bi_sector << 9) - dip->disk_bytenr;
+	u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
+	int ret;
+
+	len >>= inode->i_sb->s_blocksize_bits;
+	len *= csum_size;
+
+	ret = __btrfs_lookup_bio_sums(root, inode, bio, offset,
+				      (u32 *)(dip->csum + len), 1);
+	return ret;
+}
+
 int btrfs_lookup_csums_range(struct btrfs_root *root, u64 start, u64 end,
-			     struct list_head *list)
+			     struct list_head *list, int search_commit)
 {
 	struct btrfs_key key;
 	struct btrfs_path *path;
 	struct extent_buffer *leaf;
 	struct btrfs_ordered_sum *sums;
-	struct btrfs_sector_sum *sector_sum;
 	struct btrfs_csum_item *item;
+	LIST_HEAD(tmplist);
 	unsigned long offset;
 	int ret;
 	size_t size;
 	u64 csum_end;
-	u16 csum_size = btrfs_super_csum_size(&root->fs_info->super_copy);
+	u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
+
+	ASSERT(start == ALIGN(start, root->sectorsize) &&
+	       (end + 1) == ALIGN(end + 1, root->sectorsize));
 
 	path = btrfs_alloc_path();
-	BUG_ON(!path);
+	if (!path)
+		return -ENOMEM;
+
+	if (search_commit) {
+		path->skip_locking = 1;
+		path->reada = 2;
+		path->search_commit_root = 1;
+	}
 
 	key.objectid = BTRFS_EXTENT_CSUM_OBJECTID;
 	key.offset = start;
@@ -296,11 +375,8 @@ int btrfs_lookup_csums_range(struct btrfs_root *root, u64 start, u64 end,
 
 		btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
 		if (key.objectid != BTRFS_EXTENT_CSUM_OBJECTID ||
-		    key.type != BTRFS_EXTENT_CSUM_KEY)
-			break;
-
-		btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
-		if (key.offset > end)
+		    key.type != BTRFS_EXTENT_CSUM_KEY ||
+		    key.offset > end)
 			break;
 
 		if (key.offset > start)
@@ -318,37 +394,41 @@ int btrfs_lookup_csums_range(struct btrfs_root *root, u64 start, u64 end,
 				      struct btrfs_csum_item);
 		while (start < csum_end) {
 			size = min_t(size_t, csum_end - start,
-					MAX_ORDERED_SUM_BYTES(root));
+				     MAX_ORDERED_SUM_BYTES(root));
 			sums = kzalloc(btrfs_ordered_sum_size(root, size),
-					GFP_NOFS);
-			BUG_ON(!sums);
+				       GFP_NOFS);
+			if (!sums) {
+				ret = -ENOMEM;
+				goto fail;
+			}
 
-			sector_sum = sums->sums;
 			sums->bytenr = start;
-			sums->len = size;
+			sums->len = (int)size;
 
 			offset = (start - key.offset) >>
 				root->fs_info->sb->s_blocksize_bits;
 			offset *= csum_size;
+			size >>= root->fs_info->sb->s_blocksize_bits;
 
-			while (size > 0) {
-				read_extent_buffer(path->nodes[0],
-						&sector_sum->sum,
-						((unsigned long)item) +
-						offset, csum_size);
-				sector_sum->bytenr = start;
-
-				size -= root->sectorsize;
-				start += root->sectorsize;
-				offset += csum_size;
-				sector_sum++;
-			}
-			list_add_tail(&sums->list, list);
+			read_extent_buffer(path->nodes[0],
+					   sums->sums,
+					   ((unsigned long)item) + offset,
+					   csum_size * size);
+
+			start += root->sectorsize * size;
+			list_add_tail(&sums->list, &tmplist);
 		}
 		path->slots[0]++;
 	}
 	ret = 0;
 fail:
+	while (ret < 0 && !list_empty(&tmplist)) {
+		sums = list_entry(&tmplist, struct btrfs_ordered_sum, list);
+		list_del(&sums->list);
+		kfree(sums);
+	}
+	list_splice_tail(&tmplist, list);
+
 	btrfs_free_path(path);
 	return ret;
 }
@@ -357,24 +437,22 @@ int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode,
 		       struct bio *bio, u64 file_start, int contig)
 {
 	struct btrfs_ordered_sum *sums;
-	struct btrfs_sector_sum *sector_sum;
 	struct btrfs_ordered_extent *ordered;
 	char *data;
 	struct bio_vec *bvec = bio->bi_io_vec;
 	int bio_index = 0;
+	int index;
 	unsigned long total_bytes = 0;
 	unsigned long this_sum_bytes = 0;
 	u64 offset;
-	u64 disk_bytenr;
 
 	WARN_ON(bio->bi_vcnt <= 0);
-	sums = kzalloc(btrfs_ordered_sum_size(root, bio->bi_size), GFP_NOFS);
+	sums = kzalloc(btrfs_ordered_sum_size(root, bio->bi_iter.bi_size),
+		       GFP_NOFS);
 	if (!sums)
 		return -ENOMEM;
 
-	sector_sum = sums->sums;
-	disk_bytenr = (u64)bio->bi_sector << 9;
-	sums->len = bio->bi_size;
+	sums->len = bio->bi_iter.bi_size;
 	INIT_LIST_HEAD(&sums->list);
 
 	if (contig)
@@ -383,49 +461,48 @@ int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode,
 		offset = page_offset(bvec->bv_page) + bvec->bv_offset;
 
 	ordered = btrfs_lookup_ordered_extent(inode, offset);
-	BUG_ON(!ordered);
-	sums->bytenr = ordered->start;
+	BUG_ON(!ordered); /* Logic error */
+	sums->bytenr = (u64)bio->bi_iter.bi_sector << 9;
+	index = 0;
 
 	while (bio_index < bio->bi_vcnt) {
 		if (!contig)
 			offset = page_offset(bvec->bv_page) + bvec->bv_offset;
 
-		if (!contig && (offset >= ordered->file_offset + ordered->len ||
-		    offset < ordered->file_offset)) {
+		if (offset >= ordered->file_offset + ordered->len ||
+		    offset < ordered->file_offset) {
 			unsigned long bytes_left;
 			sums->len = this_sum_bytes;
 			this_sum_bytes = 0;
 			btrfs_add_ordered_sum(inode, ordered, sums);
 			btrfs_put_ordered_extent(ordered);
 
-			bytes_left = bio->bi_size - total_bytes;
+			bytes_left = bio->bi_iter.bi_size - total_bytes;
 
 			sums = kzalloc(btrfs_ordered_sum_size(root, bytes_left),
 				       GFP_NOFS);
-			BUG_ON(!sums);
-			sector_sum = sums->sums;
+			BUG_ON(!sums); /* -ENOMEM */
 			sums->len = bytes_left;
 			ordered = btrfs_lookup_ordered_extent(inode, offset);
-			BUG_ON(!ordered);
-			sums->bytenr = ordered->start;
+			BUG_ON(!ordered); /* Logic error */
+			sums->bytenr = ((u64)bio->bi_iter.bi_sector << 9) +
+				       total_bytes;
+			index = 0;
 		}
 
-		data = kmap_atomic(bvec->bv_page, KM_USER0);
-		sector_sum->sum = ~(u32)0;
-		sector_sum->sum = btrfs_csum_data(root,
-						  data + bvec->bv_offset,
-						  sector_sum->sum,
-						  bvec->bv_len);
-		kunmap_atomic(data, KM_USER0);
-		btrfs_csum_final(sector_sum->sum,
-				 (char *)&sector_sum->sum);
-		sector_sum->bytenr = disk_bytenr;
-
-		sector_sum++;
+		data = kmap_atomic(bvec->bv_page);
+		sums->sums[index] = ~(u32)0;
+		sums->sums[index] = btrfs_csum_data(data + bvec->bv_offset,
+						    sums->sums[index],
+						    bvec->bv_len);
+		kunmap_atomic(data);
+		btrfs_csum_final(sums->sums[index],
+				 (char *)(sums->sums + index));
+
 		bio_index++;
+		index++;
 		total_bytes += bvec->bv_len;
 		this_sum_bytes += bvec->bv_len;
-		disk_bytenr += bvec->bv_len;
 		offset += bvec->bv_len;
 		bvec++;
 	}
@@ -446,19 +523,16 @@ int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode,
  * This calls btrfs_truncate_item with the correct args based on the
  * overlap, and fixes up the key as required.
  */
-static noinline int truncate_one_csum(struct btrfs_trans_handle *trans,
-				      struct btrfs_root *root,
-				      struct btrfs_path *path,
-				      struct btrfs_key *key,
-				      u64 bytenr, u64 len)
+static noinline void truncate_one_csum(struct btrfs_root *root,
+				       struct btrfs_path *path,
+				       struct btrfs_key *key,
+				       u64 bytenr, u64 len)
 {
 	struct extent_buffer *leaf;
-	u16 csum_size =
-		btrfs_super_csum_size(&root->fs_info->super_copy);
+	u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
 	u64 csum_end;
 	u64 end_byte = bytenr + len;
 	u32 blocksize_bits = root->fs_info->sb->s_blocksize_bits;
-	int ret;
 
 	leaf = path->nodes[0];
 	csum_end = btrfs_item_size_nr(leaf, path->slots[0]) / csum_size;
@@ -474,8 +548,7 @@ static noinline int truncate_one_csum(struct btrfs_trans_handle *trans,
 		 */
 		u32 new_size = (bytenr - key->offset) >> blocksize_bits;
 		new_size *= csum_size;
-		ret = btrfs_truncate_item(trans, root, path, new_size, 1);
-		BUG_ON(ret);
+		btrfs_truncate_item(root, path, new_size, 1);
 	} else if (key->offset >= bytenr && csum_end > end_byte &&
 		   end_byte > key->offset) {
 		/*
@@ -487,16 +560,13 @@ static noinline int truncate_one_csum(struct btrfs_trans_handle *trans,
 		u32 new_size = (csum_end - end_byte) >> blocksize_bits;
 		new_size *= csum_size;
 
-		ret = btrfs_truncate_item(trans, root, path, new_size, 0);
-		BUG_ON(ret);
+		btrfs_truncate_item(root, path, new_size, 0);
 
 		key->offset = end_byte;
-		ret = btrfs_set_item_key_safe(trans, root, path, key);
-		BUG_ON(ret);
+		btrfs_set_item_key_safe(root, path, key);
 	} else {
 		BUG();
 	}
-	return 0;
 }
 
 /*
@@ -512,13 +582,14 @@ int btrfs_del_csums(struct btrfs_trans_handle *trans,
 	u64 csum_end;
 	struct extent_buffer *leaf;
 	int ret;
-	u16 csum_size =
-		btrfs_super_csum_size(&root->fs_info->super_copy);
+	u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
 	int blocksize_bits = root->fs_info->sb->s_blocksize_bits;
 
 	root = root->fs_info->csum_root;
 
 	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
 
 	while (1) {
 		key.objectid = BTRFS_EXTENT_CSUM_OBJECTID;
@@ -529,9 +600,12 @@ int btrfs_del_csums(struct btrfs_trans_handle *trans,
 		ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
 		if (ret > 0) {
 			if (path->slots[0] == 0)
-				goto out;
+				break;
 			path->slots[0]--;
+		} else if (ret < 0) {
+			break;
 		}
+
 		leaf = path->nodes[0];
 		btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
 
@@ -554,7 +628,8 @@ int btrfs_del_csums(struct btrfs_trans_handle *trans,
 		/* delete the entire item, it is inside our range */
 		if (key.offset >= bytenr && csum_end <= end_byte) {
 			ret = btrfs_del_item(trans, root, path);
-			BUG_ON(ret);
+			if (ret)
+				goto out;
 			if (key.offset == bytenr)
 				break;
 		} else if (key.offset < bytenr && csum_end > end_byte) {
@@ -596,67 +671,71 @@ int btrfs_del_csums(struct btrfs_trans_handle *trans,
 			 * item changed size or key
 			 */
 			ret = btrfs_split_item(trans, root, path, &key, offset);
-			BUG_ON(ret && ret != -EAGAIN);
+			if (ret && ret != -EAGAIN) {
+				btrfs_abort_transaction(trans, root, ret);
+				goto out;
+			}
 
 			key.offset = end_byte - 1;
 		} else {
-			ret = truncate_one_csum(trans, root, path,
-						&key, bytenr, len);
-			BUG_ON(ret);
+			truncate_one_csum(root, path, &key, bytenr, len);
 			if (key.offset < bytenr)
 				break;
 		}
-		btrfs_release_path(root, path);
+		btrfs_release_path(path);
 	}
+	ret = 0;
 out:
 	btrfs_free_path(path);
-	return 0;
+	return ret;
 }
 
 int btrfs_csum_file_blocks(struct btrfs_trans_handle *trans,
 			   struct btrfs_root *root,
 			   struct btrfs_ordered_sum *sums)
 {
-	u64 bytenr;
-	int ret;
 	struct btrfs_key file_key;
 	struct btrfs_key found_key;
-	u64 next_offset;
-	u64 total_bytes = 0;
-	int found_next;
 	struct btrfs_path *path;
 	struct btrfs_csum_item *item;
 	struct btrfs_csum_item *item_end;
 	struct extent_buffer *leaf = NULL;
+	u64 next_offset;
+	u64 total_bytes = 0;
 	u64 csum_offset;
-	struct btrfs_sector_sum *sector_sum;
+	u64 bytenr;
 	u32 nritems;
 	u32 ins_size;
-	char *eb_map;
-	char *eb_token;
-	unsigned long map_len;
-	unsigned long map_start;
-	u16 csum_size =
-		btrfs_super_csum_size(&root->fs_info->super_copy);
+	int index = 0;
+	int found_next;
+	int ret;
+	u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
 
 	path = btrfs_alloc_path();
-	BUG_ON(!path);
-	sector_sum = sums->sums;
+	if (!path)
+		return -ENOMEM;
 again:
 	next_offset = (u64)-1;
 	found_next = 0;
+	bytenr = sums->bytenr + total_bytes;
 	file_key.objectid = BTRFS_EXTENT_CSUM_OBJECTID;
-	file_key.offset = sector_sum->bytenr;
-	bytenr = sector_sum->bytenr;
+	file_key.offset = bytenr;
 	btrfs_set_key_type(&file_key, BTRFS_EXTENT_CSUM_KEY);
 
-	item = btrfs_lookup_csum(trans, root, path, sector_sum->bytenr, 1);
+	item = btrfs_lookup_csum(trans, root, path, bytenr, 1);
 	if (!IS_ERR(item)) {
-		leaf = path->nodes[0];
 		ret = 0;
+		leaf = path->nodes[0];
+		item_end = btrfs_item_ptr(leaf, path->slots[0],
+					  struct btrfs_csum_item);
+		item_end = (struct btrfs_csum_item *)((char *)item_end +
+			   btrfs_item_size_nr(leaf, path->slots[0]));
 		goto found;
 	}
 	ret = PTR_ERR(item);
+	if (ret != -EFBIG && ret != -ENOENT)
+		goto fail_unlock;
+
 	if (ret == -EFBIG) {
 		u32 item_size;
 		/* we found one, but it isn't big enough yet */
@@ -671,7 +750,7 @@ again:
 		int slot = path->slots[0] + 1;
 		/* we didn't find a csum item, insert one */
 		nritems = btrfs_header_nritems(path->nodes[0]);
-		if (path->slots[0] >= nritems - 1) {
+		if (!nritems || (path->slots[0] >= nritems - 1)) {
 			ret = btrfs_next_leaf(root, path);
 			if (ret == 1)
 				found_next = 1;
@@ -694,7 +773,7 @@ again:
 	 * at this point, we know the tree has an item, but it isn't big
 	 * enough yet to put our csum in.  Grow it
 	 */
-	btrfs_release_path(root, path);
+	btrfs_release_path(path);
 	ret = btrfs_search_slot(trans, root, &file_key, path,
 				csum_size, 1);
 	if (ret < 0)
@@ -717,43 +796,48 @@ again:
 		goto insert;
 	}
 
-	if (csum_offset >= btrfs_item_size_nr(leaf, path->slots[0]) /
+	if (csum_offset == btrfs_item_size_nr(leaf, path->slots[0]) /
 	    csum_size) {
-		u32 diff = (csum_offset + 1) * csum_size;
+		int extend_nr;
+		u64 tmp;
+		u32 diff;
+		u32 free_space;
 
-		/*
-		 * is the item big enough already?  we dropped our lock
-		 * before and need to recheck
-		 */
-		if (diff < btrfs_item_size_nr(leaf, path->slots[0]))
-			goto csum;
+		if (btrfs_leaf_free_space(root, leaf) <
+				 sizeof(struct btrfs_item) + csum_size * 2)
+			goto insert;
+
+		free_space = btrfs_leaf_free_space(root, leaf) -
+					 sizeof(struct btrfs_item) - csum_size;
+		tmp = sums->len - total_bytes;
+		tmp >>= root->fs_info->sb->s_blocksize_bits;
+		WARN_ON(tmp < 1);
+
+		extend_nr = max_t(int, 1, (int)tmp);
+		diff = (csum_offset + extend_nr) * csum_size;
+		diff = min(diff, MAX_CSUM_ITEMS(root, csum_size) * csum_size);
 
 		diff = diff - btrfs_item_size_nr(leaf, path->slots[0]);
-		if (diff != csum_size)
-			goto insert;
+		diff = min(free_space, diff);
+		diff /= csum_size;
+		diff *= csum_size;
 
-		ret = btrfs_extend_item(trans, root, path, diff);
-		BUG_ON(ret);
+		btrfs_extend_item(root, path, diff);
+		ret = 0;
 		goto csum;
 	}
 
 insert:
-	btrfs_release_path(root, path);
+	btrfs_release_path(path);
 	csum_offset = 0;
 	if (found_next) {
-		u64 tmp = total_bytes + root->sectorsize;
-		u64 next_sector = sector_sum->bytenr;
-		struct btrfs_sector_sum *next = sector_sum + 1;
+		u64 tmp;
 
-		while (tmp < sums->len) {
-			if (next_sector + root->sectorsize != next->bytenr)
-				break;
-			tmp += root->sectorsize;
-			next_sector = next->bytenr;
-			next++;
-		}
-		tmp = min(tmp, next_offset - file_key.offset);
+		tmp = sums->len - total_bytes;
 		tmp >>= root->fs_info->sb->s_blocksize_bits;
+		tmp = min(tmp, (next_offset - file_key.offset) >>
+					 root->fs_info->sb->s_blocksize_bits);
+
 		tmp = max((u64)1, tmp);
 		tmp = min(tmp, (u64)MAX_CSUM_ITEMS(root, csum_size));
 		ins_size = csum_size * tmp;
@@ -766,63 +850,31 @@ insert:
 	path->leave_spinning = 0;
 	if (ret < 0)
 		goto fail_unlock;
-	if (ret != 0) {
-		WARN_ON(1);
+	if (WARN_ON(ret != 0))
 		goto fail_unlock;
-	}
-csum:
 	leaf = path->nodes[0];
+csum:
 	item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_csum_item);
-	ret = 0;
+	item_end = (struct btrfs_csum_item *)((unsigned char *)item +
+				      btrfs_item_size_nr(leaf, path->slots[0]));
 	item = (struct btrfs_csum_item *)((unsigned char *)item +
 					  csum_offset * csum_size);
 found:
-	item_end = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_csum_item);
-	item_end = (struct btrfs_csum_item *)((unsigned char *)item_end +
-				      btrfs_item_size_nr(leaf, path->slots[0]));
-	eb_token = NULL;
-next_sector:
-
-	if (!eb_token ||
-	   (unsigned long)item + csum_size >= map_start + map_len) {
-		int err;
-
-		if (eb_token)
-			unmap_extent_buffer(leaf, eb_token, KM_USER1);
-		eb_token = NULL;
-		err = map_private_extent_buffer(leaf, (unsigned long)item,
-						csum_size,
-						&eb_token, &eb_map,
-						&map_start, &map_len, KM_USER1);
-		if (err)
-			eb_token = NULL;
-	}
-	if (eb_token) {
-		memcpy(eb_token + ((unsigned long)item & (PAGE_CACHE_SIZE - 1)),
-		       &sector_sum->sum, csum_size);
-	} else {
-		write_extent_buffer(leaf, &sector_sum->sum,
-				    (unsigned long)item, csum_size);
-	}
+	ins_size = (u32)(sums->len - total_bytes) >>
+		   root->fs_info->sb->s_blocksize_bits;
+	ins_size *= csum_size;
+	ins_size = min_t(u32, (unsigned long)item_end - (unsigned long)item,
+			      ins_size);
+	write_extent_buffer(leaf, sums->sums + index, (unsigned long)item,
+			    ins_size);
+
+	ins_size /= csum_size;
+	total_bytes += ins_size * root->sectorsize;
+	index += ins_size;
 
-	total_bytes += root->sectorsize;
-	sector_sum++;
-	if (total_bytes < sums->len) {
-		item = (struct btrfs_csum_item *)((char *)item +
-						  csum_size);
-		if (item < item_end && bytenr + PAGE_CACHE_SIZE ==
-		    sector_sum->bytenr) {
-			bytenr = sector_sum->bytenr;
-			goto next_sector;
-		}
-	}
-	if (eb_token) {
-		unmap_extent_buffer(leaf, eb_token, KM_USER1);
-		eb_token = NULL;
-	}
 	btrfs_mark_buffer_dirty(path->nodes[0]);
 	if (total_bytes < sums->len) {
-		btrfs_release_path(root, path);
+		btrfs_release_path(path);
 		cond_resched();
 		goto again;
 	}
@@ -833,3 +885,79 @@ out:
 fail_unlock:
 	goto out;
 }
+
+void btrfs_extent_item_to_extent_map(struct inode *inode,
+				     const struct btrfs_path *path,
+				     struct btrfs_file_extent_item *fi,
+				     const bool new_inline,
+				     struct extent_map *em)
+{
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct extent_buffer *leaf = path->nodes[0];
+	const int slot = path->slots[0];
+	struct btrfs_key key;
+	u64 extent_start, extent_end;
+	u64 bytenr;
+	u8 type = btrfs_file_extent_type(leaf, fi);
+	int compress_type = btrfs_file_extent_compression(leaf, fi);
+
+	em->bdev = root->fs_info->fs_devices->latest_bdev;
+	btrfs_item_key_to_cpu(leaf, &key, slot);
+	extent_start = key.offset;
+
+	if (type == BTRFS_FILE_EXTENT_REG ||
+	    type == BTRFS_FILE_EXTENT_PREALLOC) {
+		extent_end = extent_start +
+			btrfs_file_extent_num_bytes(leaf, fi);
+	} else if (type == BTRFS_FILE_EXTENT_INLINE) {
+		size_t size;
+		size = btrfs_file_extent_inline_len(leaf, slot, fi);
+		extent_end = ALIGN(extent_start + size, root->sectorsize);
+	}
+
+	em->ram_bytes = btrfs_file_extent_ram_bytes(leaf, fi);
+	if (type == BTRFS_FILE_EXTENT_REG ||
+	    type == BTRFS_FILE_EXTENT_PREALLOC) {
+		em->start = extent_start;
+		em->len = extent_end - extent_start;
+		em->orig_start = extent_start -
+			btrfs_file_extent_offset(leaf, fi);
+		em->orig_block_len = btrfs_file_extent_disk_num_bytes(leaf, fi);
+		bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
+		if (bytenr == 0) {
+			em->block_start = EXTENT_MAP_HOLE;
+			return;
+		}
+		if (compress_type != BTRFS_COMPRESS_NONE) {
+			set_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
+			em->compress_type = compress_type;
+			em->block_start = bytenr;
+			em->block_len = em->orig_block_len;
+		} else {
+			bytenr += btrfs_file_extent_offset(leaf, fi);
+			em->block_start = bytenr;
+			em->block_len = em->len;
+			if (type == BTRFS_FILE_EXTENT_PREALLOC)
+				set_bit(EXTENT_FLAG_PREALLOC, &em->flags);
+		}
+	} else if (type == BTRFS_FILE_EXTENT_INLINE) {
+		em->block_start = EXTENT_MAP_INLINE;
+		em->start = extent_start;
+		em->len = extent_end - extent_start;
+		/*
+		 * Initialize orig_start and block_len with the same values
+		 * as in inode.c:btrfs_get_extent().
+		 */
+		em->orig_start = EXTENT_MAP_HOLE;
+		em->block_len = (u64)-1;
+		if (!new_inline && compress_type != BTRFS_COMPRESS_NONE) {
+			set_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
+			em->compress_type = compress_type;
+		}
+	} else {
+		btrfs_err(root->fs_info,
+			  "unknown file extent item type %d, inode %llu, offset %llu, root %llu",
+			  type, btrfs_ino(inode), extent_start,
+			  root->root_key.objectid);
+	}
+}
diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index 29ff749ff4c..1f2b99cb55e 100644
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -24,72 +24,459 @@
 #include <linux/string.h>
 #include <linux/backing-dev.h>
 #include <linux/mpage.h>
+#include <linux/aio.h>
+#include <linux/falloc.h>
 #include <linux/swap.h>
 #include <linux/writeback.h>
 #include <linux/statfs.h>
 #include <linux/compat.h>
 #include <linux/slab.h>
+#include <linux/btrfs.h>
 #include "ctree.h"
 #include "disk-io.h"
 #include "transaction.h"
 #include "btrfs_inode.h"
-#include "ioctl.h"
 #include "print-tree.h"
 #include "tree-log.h"
 #include "locking.h"
-#include "compat.h"
+#include "volumes.h"
+#include "qgroup.h"
 
+static struct kmem_cache *btrfs_inode_defrag_cachep;
+/*
+ * when auto defrag is enabled we
+ * queue up these defrag structs to remember which
+ * inodes need defragging passes
+ */
+struct inode_defrag {
+	struct rb_node rb_node;
+	/* objectid */
+	u64 ino;
+	/*
+	 * transid where the defrag was added, we search for
+	 * extents newer than this
+	 */
+	u64 transid;
+
+	/* root objectid */
+	u64 root;
+
+	/* last offset we were able to defrag */
+	u64 last_offset;
+
+	/* if we've wrapped around back to zero once already */
+	int cycled;
+};
+
+static int __compare_inode_defrag(struct inode_defrag *defrag1,
+				  struct inode_defrag *defrag2)
+{
+	if (defrag1->root > defrag2->root)
+		return 1;
+	else if (defrag1->root < defrag2->root)
+		return -1;
+	else if (defrag1->ino > defrag2->ino)
+		return 1;
+	else if (defrag1->ino < defrag2->ino)
+		return -1;
+	else
+		return 0;
+}
+
+/* pop a record for an inode into the defrag tree.  The lock
+ * must be held already
+ *
+ * If you're inserting a record for an older transid than an
+ * existing record, the transid already in the tree is lowered
+ *
+ * If an existing record is found the defrag item you
+ * pass in is freed
+ */
+static int __btrfs_add_inode_defrag(struct inode *inode,
+				    struct inode_defrag *defrag)
+{
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct inode_defrag *entry;
+	struct rb_node **p;
+	struct rb_node *parent = NULL;
+	int ret;
+
+	p = &root->fs_info->defrag_inodes.rb_node;
+	while (*p) {
+		parent = *p;
+		entry = rb_entry(parent, struct inode_defrag, rb_node);
+
+		ret = __compare_inode_defrag(defrag, entry);
+		if (ret < 0)
+			p = &parent->rb_left;
+		else if (ret > 0)
+			p = &parent->rb_right;
+		else {
+			/* if we're reinserting an entry for
+			 * an old defrag run, make sure to
+			 * lower the transid of our existing record
+			 */
+			if (defrag->transid < entry->transid)
+				entry->transid = defrag->transid;
+			if (defrag->last_offset > entry->last_offset)
+				entry->last_offset = defrag->last_offset;
+			return -EEXIST;
+		}
+	}
+	set_bit(BTRFS_INODE_IN_DEFRAG, &BTRFS_I(inode)->runtime_flags);
+	rb_link_node(&defrag->rb_node, parent, p);
+	rb_insert_color(&defrag->rb_node, &root->fs_info->defrag_inodes);
+	return 0;
+}
+
+static inline int __need_auto_defrag(struct btrfs_root *root)
+{
+	if (!btrfs_test_opt(root, AUTO_DEFRAG))
+		return 0;
+
+	if (btrfs_fs_closing(root->fs_info))
+		return 0;
+
+	return 1;
+}
+
+/*
+ * insert a defrag record for this inode if auto defrag is
+ * enabled
+ */
+int btrfs_add_inode_defrag(struct btrfs_trans_handle *trans,
+			   struct inode *inode)
+{
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct inode_defrag *defrag;
+	u64 transid;
+	int ret;
+
+	if (!__need_auto_defrag(root))
+		return 0;
+
+	if (test_bit(BTRFS_INODE_IN_DEFRAG, &BTRFS_I(inode)->runtime_flags))
+		return 0;
+
+	if (trans)
+		transid = trans->transid;
+	else
+		transid = BTRFS_I(inode)->root->last_trans;
+
+	defrag = kmem_cache_zalloc(btrfs_inode_defrag_cachep, GFP_NOFS);
+	if (!defrag)
+		return -ENOMEM;
+
+	defrag->ino = btrfs_ino(inode);
+	defrag->transid = transid;
+	defrag->root = root->root_key.objectid;
+
+	spin_lock(&root->fs_info->defrag_inodes_lock);
+	if (!test_bit(BTRFS_INODE_IN_DEFRAG, &BTRFS_I(inode)->runtime_flags)) {
+		/*
+		 * If we set IN_DEFRAG flag and evict the inode from memory,
+		 * and then re-read this inode, this new inode doesn't have
+		 * IN_DEFRAG flag. At the case, we may find the existed defrag.
+		 */
+		ret = __btrfs_add_inode_defrag(inode, defrag);
+		if (ret)
+			kmem_cache_free(btrfs_inode_defrag_cachep, defrag);
+	} else {
+		kmem_cache_free(btrfs_inode_defrag_cachep, defrag);
+	}
+	spin_unlock(&root->fs_info->defrag_inodes_lock);
+	return 0;
+}
+
+/*
+ * Requeue the defrag object. If there is a defrag object that points to
+ * the same inode in the tree, we will merge them together (by
+ * __btrfs_add_inode_defrag()) and free the one that we want to requeue.
+ */
+static void btrfs_requeue_inode_defrag(struct inode *inode,
+				       struct inode_defrag *defrag)
+{
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	int ret;
+
+	if (!__need_auto_defrag(root))
+		goto out;
+
+	/*
+	 * Here we don't check the IN_DEFRAG flag, because we need merge
+	 * them together.
+	 */
+	spin_lock(&root->fs_info->defrag_inodes_lock);
+	ret = __btrfs_add_inode_defrag(inode, defrag);
+	spin_unlock(&root->fs_info->defrag_inodes_lock);
+	if (ret)
+		goto out;
+	return;
+out:
+	kmem_cache_free(btrfs_inode_defrag_cachep, defrag);
+}
+
+/*
+ * pick the defragable inode that we want, if it doesn't exist, we will get
+ * the next one.
+ */
+static struct inode_defrag *
+btrfs_pick_defrag_inode(struct btrfs_fs_info *fs_info, u64 root, u64 ino)
+{
+	struct inode_defrag *entry = NULL;
+	struct inode_defrag tmp;
+	struct rb_node *p;
+	struct rb_node *parent = NULL;
+	int ret;
+
+	tmp.ino = ino;
+	tmp.root = root;
+
+	spin_lock(&fs_info->defrag_inodes_lock);
+	p = fs_info->defrag_inodes.rb_node;
+	while (p) {
+		parent = p;
+		entry = rb_entry(parent, struct inode_defrag, rb_node);
+
+		ret = __compare_inode_defrag(&tmp, entry);
+		if (ret < 0)
+			p = parent->rb_left;
+		else if (ret > 0)
+			p = parent->rb_right;
+		else
+			goto out;
+	}
+
+	if (parent && __compare_inode_defrag(&tmp, entry) > 0) {
+		parent = rb_next(parent);
+		if (parent)
+			entry = rb_entry(parent, struct inode_defrag, rb_node);
+		else
+			entry = NULL;
+	}
+out:
+	if (entry)
+		rb_erase(parent, &fs_info->defrag_inodes);
+	spin_unlock(&fs_info->defrag_inodes_lock);
+	return entry;
+}
+
+void btrfs_cleanup_defrag_inodes(struct btrfs_fs_info *fs_info)
+{
+	struct inode_defrag *defrag;
+	struct rb_node *node;
+
+	spin_lock(&fs_info->defrag_inodes_lock);
+	node = rb_first(&fs_info->defrag_inodes);
+	while (node) {
+		rb_erase(node, &fs_info->defrag_inodes);
+		defrag = rb_entry(node, struct inode_defrag, rb_node);
+		kmem_cache_free(btrfs_inode_defrag_cachep, defrag);
+
+		if (need_resched()) {
+			spin_unlock(&fs_info->defrag_inodes_lock);
+			cond_resched();
+			spin_lock(&fs_info->defrag_inodes_lock);
+		}
+
+		node = rb_first(&fs_info->defrag_inodes);
+	}
+	spin_unlock(&fs_info->defrag_inodes_lock);
+}
+
+#define BTRFS_DEFRAG_BATCH	1024
+
+static int __btrfs_run_defrag_inode(struct btrfs_fs_info *fs_info,
+				    struct inode_defrag *defrag)
+{
+	struct btrfs_root *inode_root;
+	struct inode *inode;
+	struct btrfs_key key;
+	struct btrfs_ioctl_defrag_range_args range;
+	int num_defrag;
+	int index;
+	int ret;
+
+	/* get the inode */
+	key.objectid = defrag->root;
+	btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
+	key.offset = (u64)-1;
+
+	index = srcu_read_lock(&fs_info->subvol_srcu);
+
+	inode_root = btrfs_read_fs_root_no_name(fs_info, &key);
+	if (IS_ERR(inode_root)) {
+		ret = PTR_ERR(inode_root);
+		goto cleanup;
+	}
+
+	key.objectid = defrag->ino;
+	btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
+	key.offset = 0;
+	inode = btrfs_iget(fs_info->sb, &key, inode_root, NULL);
+	if (IS_ERR(inode)) {
+		ret = PTR_ERR(inode);
+		goto cleanup;
+	}
+	srcu_read_unlock(&fs_info->subvol_srcu, index);
+
+	/* do a chunk of defrag */
+	clear_bit(BTRFS_INODE_IN_DEFRAG, &BTRFS_I(inode)->runtime_flags);
+	memset(&range, 0, sizeof(range));
+	range.len = (u64)-1;
+	range.start = defrag->last_offset;
+
+	sb_start_write(fs_info->sb);
+	num_defrag = btrfs_defrag_file(inode, NULL, &range, defrag->transid,
+				       BTRFS_DEFRAG_BATCH);
+	sb_end_write(fs_info->sb);
+	/*
+	 * if we filled the whole defrag batch, there
+	 * must be more work to do.  Queue this defrag
+	 * again
+	 */
+	if (num_defrag == BTRFS_DEFRAG_BATCH) {
+		defrag->last_offset = range.start;
+		btrfs_requeue_inode_defrag(inode, defrag);
+	} else if (defrag->last_offset && !defrag->cycled) {
+		/*
+		 * we didn't fill our defrag batch, but
+		 * we didn't start at zero.  Make sure we loop
+		 * around to the start of the file.
+		 */
+		defrag->last_offset = 0;
+		defrag->cycled = 1;
+		btrfs_requeue_inode_defrag(inode, defrag);
+	} else {
+		kmem_cache_free(btrfs_inode_defrag_cachep, defrag);
+	}
+
+	iput(inode);
+	return 0;
+cleanup:
+	srcu_read_unlock(&fs_info->subvol_srcu, index);
+	kmem_cache_free(btrfs_inode_defrag_cachep, defrag);
+	return ret;
+}
+
+/*
+ * run through the list of inodes in the FS that need
+ * defragging
+ */
+int btrfs_run_defrag_inodes(struct btrfs_fs_info *fs_info)
+{
+	struct inode_defrag *defrag;
+	u64 first_ino = 0;
+	u64 root_objectid = 0;
+
+	atomic_inc(&fs_info->defrag_running);
+	while (1) {
+		/* Pause the auto defragger. */
+		if (test_bit(BTRFS_FS_STATE_REMOUNTING,
+			     &fs_info->fs_state))
+			break;
+
+		if (!__need_auto_defrag(fs_info->tree_root))
+			break;
+
+		/* find an inode to defrag */
+		defrag = btrfs_pick_defrag_inode(fs_info, root_objectid,
+						 first_ino);
+		if (!defrag) {
+			if (root_objectid || first_ino) {
+				root_objectid = 0;
+				first_ino = 0;
+				continue;
+			} else {
+				break;
+			}
+		}
+
+		first_ino = defrag->ino + 1;
+		root_objectid = defrag->root;
+
+		__btrfs_run_defrag_inode(fs_info, defrag);
+	}
+	atomic_dec(&fs_info->defrag_running);
+
+	/*
+	 * during unmount, we use the transaction_wait queue to
+	 * wait for the defragger to stop
+	 */
+	wake_up(&fs_info->transaction_wait);
+	return 0;
+}
 
 /* simple helper to fault in pages and copy.  This should go away
  * and be replaced with calls into generic code.
  */
 static noinline int btrfs_copy_from_user(loff_t pos, int num_pages,
-					 int write_bytes,
+					 size_t write_bytes,
 					 struct page **prepared_pages,
-					 const char __user *buf)
+					 struct iov_iter *i)
 {
-	long page_fault = 0;
-	int i;
+	size_t copied = 0;
+	size_t total_copied = 0;
+	int pg = 0;
 	int offset = pos & (PAGE_CACHE_SIZE - 1);
 
-	for (i = 0; i < num_pages && write_bytes > 0; i++, offset = 0) {
+	while (write_bytes > 0) {
 		size_t count = min_t(size_t,
 				     PAGE_CACHE_SIZE - offset, write_bytes);
-		struct page *page = prepared_pages[i];
-		fault_in_pages_readable(buf, count);
+		struct page *page = prepared_pages[pg];
+		/*
+		 * Copy data from userspace to the current page
+		 */
+		copied = iov_iter_copy_from_user_atomic(page, i, offset, count);
 
-		/* Copy data from userspace to the current page */
-		kmap(page);
-		page_fault = __copy_from_user(page_address(page) + offset,
-					      buf, count);
 		/* Flush processor's dcache for this page */
 		flush_dcache_page(page);
-		kunmap(page);
-		buf += count;
-		write_bytes -= count;
 
-		if (page_fault)
+		/*
+		 * if we get a partial write, we can end up with
+		 * partially up to date pages.  These add
+		 * a lot of complexity, so make sure they don't
+		 * happen by forcing this copy to be retried.
+		 *
+		 * The rest of the btrfs_file_write code will fall
+		 * back to page at a time copies after we return 0.
+		 */
+		if (!PageUptodate(page) && copied < count)
+			copied = 0;
+
+		iov_iter_advance(i, copied);
+		write_bytes -= copied;
+		total_copied += copied;
+
+		/* Return to btrfs_file_write_iter to fault page */
+		if (unlikely(copied == 0))
 			break;
+
+		if (unlikely(copied < PAGE_CACHE_SIZE - offset)) {
+			offset += copied;
+		} else {
+			pg++;
+			offset = 0;
+		}
 	}
-	return page_fault ? -EFAULT : 0;
+	return total_copied;
 }
 
 /*
  * unlocks pages after btrfs_file_write is done with them
  */
-static noinline void btrfs_drop_pages(struct page **pages, size_t num_pages)
+static void btrfs_drop_pages(struct page **pages, size_t num_pages)
 {
 	size_t i;
 	for (i = 0; i < num_pages; i++) {
-		if (!pages[i])
-			break;
 		/* page checked is some magic around finding pages that
 		 * have been modified without going through btrfs_set_page_dirty
-		 * clear it here
+		 * clear it here. There should be no need to mark the pages
+		 * accessed as prepare_pages should have marked them accessed
+		 * in prepare_pages via find_or_create_page()
 		 */
 		ClearPageChecked(pages[i]);
 		unlock_page(pages[i]);
-		mark_page_accessed(pages[i]);
 		page_cache_release(pages[i]);
 	}
 }
@@ -102,17 +489,13 @@ static noinline void btrfs_drop_pages(struct page **pages, size_t num_pages)
  * this also makes the decision about creating an inline extent vs
  * doing real data extents, marking pages dirty and delalloc as required.
  */
-static noinline int dirty_and_release_pages(struct btrfs_trans_handle *trans,
-				   struct btrfs_root *root,
-				   struct file *file,
-				   struct page **pages,
-				   size_t num_pages,
-				   loff_t pos,
-				   size_t write_bytes)
+int btrfs_dirty_pages(struct btrfs_root *root, struct inode *inode,
+			     struct page **pages, size_t num_pages,
+			     loff_t pos, size_t write_bytes,
+			     struct extent_state **cached)
 {
 	int err = 0;
 	int i;
-	struct inode *inode = fdentry(file)->d_inode;
 	u64 num_bytes;
 	u64 start_pos;
 	u64 end_of_last_block;
@@ -120,12 +503,11 @@ static noinline int dirty_and_release_pages(struct btrfs_trans_handle *trans,
 	loff_t isize = i_size_read(inode);
 
 	start_pos = pos & ~((u64)root->sectorsize - 1);
-	num_bytes = (write_bytes + pos - start_pos +
-		    root->sectorsize - 1) & ~((u64)root->sectorsize - 1);
+	num_bytes = ALIGN(write_bytes + pos - start_pos, root->sectorsize);
 
 	end_of_last_block = start_pos + num_bytes - 1;
 	err = btrfs_set_extent_delalloc(inode, start_pos, end_of_last_block,
-					NULL);
+					cached);
 	if (err)
 		return err;
 
@@ -135,32 +517,35 @@ static noinline int dirty_and_release_pages(struct btrfs_trans_handle *trans,
 		ClearPageChecked(p);
 		set_page_dirty(p);
 	}
-	if (end_pos > isize) {
+
+	/*
+	 * we've only changed i_size in ram, and we haven't updated
+	 * the disk i_size.  There is no need to log the inode
+	 * at this time.
+	 */
+	if (end_pos > isize)
 		i_size_write(inode, end_pos);
-		/* we've only changed i_size in ram, and we haven't updated
-		 * the disk i_size.  There is no need to log the inode
-		 * at this time.
-		 */
-	}
-	return err;
+	return 0;
 }
 
 /*
  * this drops all the extents in the cache that intersect the range
  * [start, end].  Existing extents are split as required.
  */
-int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end,
-			    int skip_pinned)
+void btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end,
+			     int skip_pinned)
 {
 	struct extent_map *em;
 	struct extent_map *split = NULL;
 	struct extent_map *split2 = NULL;
 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
 	u64 len = end - start + 1;
+	u64 gen;
 	int ret;
 	int testend = 1;
 	unsigned long flags;
 	int compressed = 0;
+	bool modified;
 
 	WARN_ON(end < start);
 	if (end == (u64)-1) {
@@ -168,10 +553,15 @@ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end,
 		testend = 0;
 	}
 	while (1) {
+		int no_splits = 0;
+
+		modified = false;
 		if (!split)
-			split = alloc_extent_map(GFP_NOFS);
+			split = alloc_extent_map();
 		if (!split2)
-			split2 = alloc_extent_map(GFP_NOFS);
+			split2 = alloc_extent_map();
+		if (!split || !split2)
+			no_splits = 1;
 
 		write_lock(&em_tree->lock);
 		em = lookup_extent_mapping(em_tree, start, len);
@@ -180,6 +570,7 @@ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end,
 			break;
 		}
 		flags = em->flags;
+		gen = em->generation;
 		if (skip_pinned && test_bit(EXTENT_FLAG_PINNED, &em->flags)) {
 			if (testend && em->start + em->len >= start + len) {
 				free_extent_map(em);
@@ -195,52 +586,90 @@ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end,
 		}
 		compressed = test_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
 		clear_bit(EXTENT_FLAG_PINNED, &em->flags);
-		remove_extent_mapping(em_tree, em);
+		clear_bit(EXTENT_FLAG_LOGGING, &flags);
+		modified = !list_empty(&em->list);
+		if (no_splits)
+			goto next;
 
-		if (em->block_start < EXTENT_MAP_LAST_BYTE &&
-		    em->start < start) {
+		if (em->start < start) {
 			split->start = em->start;
 			split->len = start - em->start;
-			split->orig_start = em->orig_start;
-			split->block_start = em->block_start;
 
-			if (compressed)
-				split->block_len = em->block_len;
-			else
-				split->block_len = split->len;
+			if (em->block_start < EXTENT_MAP_LAST_BYTE) {
+				split->orig_start = em->orig_start;
+				split->block_start = em->block_start;
 
+				if (compressed)
+					split->block_len = em->block_len;
+				else
+					split->block_len = split->len;
+				split->orig_block_len = max(split->block_len,
+						em->orig_block_len);
+				split->ram_bytes = em->ram_bytes;
+			} else {
+				split->orig_start = split->start;
+				split->block_len = 0;
+				split->block_start = em->block_start;
+				split->orig_block_len = 0;
+				split->ram_bytes = split->len;
+			}
+
+			split->generation = gen;
 			split->bdev = em->bdev;
 			split->flags = flags;
-			ret = add_extent_mapping(em_tree, split);
-			BUG_ON(ret);
+			split->compress_type = em->compress_type;
+			replace_extent_mapping(em_tree, em, split, modified);
 			free_extent_map(split);
 			split = split2;
 			split2 = NULL;
 		}
-		if (em->block_start < EXTENT_MAP_LAST_BYTE &&
-		    testend && em->start + em->len > start + len) {
+		if (testend && em->start + em->len > start + len) {
 			u64 diff = start + len - em->start;
 
 			split->start = start + len;
 			split->len = em->start + em->len - (start + len);
 			split->bdev = em->bdev;
 			split->flags = flags;
-
-			if (compressed) {
-				split->block_len = em->block_len;
-				split->block_start = em->block_start;
-				split->orig_start = em->orig_start;
+			split->compress_type = em->compress_type;
+			split->generation = gen;
+
+			if (em->block_start < EXTENT_MAP_LAST_BYTE) {
+				split->orig_block_len = max(em->block_len,
+						    em->orig_block_len);
+
+				split->ram_bytes = em->ram_bytes;
+				if (compressed) {
+					split->block_len = em->block_len;
+					split->block_start = em->block_start;
+					split->orig_start = em->orig_start;
+				} else {
+					split->block_len = split->len;
+					split->block_start = em->block_start
+						+ diff;
+					split->orig_start = em->orig_start;
+				}
 			} else {
-				split->block_len = split->len;
-				split->block_start = em->block_start + diff;
+				split->ram_bytes = split->len;
 				split->orig_start = split->start;
+				split->block_len = 0;
+				split->block_start = em->block_start;
+				split->orig_block_len = 0;
 			}
 
-			ret = add_extent_mapping(em_tree, split);
-			BUG_ON(ret);
+			if (extent_map_in_tree(em)) {
+				replace_extent_mapping(em_tree, em, split,
+						       modified);
+			} else {
+				ret = add_extent_mapping(em_tree, split,
+							 modified);
+				ASSERT(ret == 0); /* Logic error */
+			}
 			free_extent_map(split);
 			split = NULL;
 		}
+next:
+		if (extent_map_in_tree(em))
+			remove_extent_mapping(em_tree, em);
 		write_unlock(&em_tree->lock);
 
 		/* once for us */
@@ -252,7 +681,6 @@ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end,
 		free_extent_map(split);
 	if (split2)
 		free_extent_map(split2);
-	return 0;
 }
 
 /*
@@ -264,15 +692,19 @@ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end,
  * it is either truncated or split.  Anything entirely inside the range
  * is deleted from the tree.
  */
-int btrfs_drop_extents(struct btrfs_trans_handle *trans, struct inode *inode,
-		       u64 start, u64 end, u64 *hint_byte, int drop_cache)
+int __btrfs_drop_extents(struct btrfs_trans_handle *trans,
+			 struct btrfs_root *root, struct inode *inode,
+			 struct btrfs_path *path, u64 start, u64 end,
+			 u64 *drop_end, int drop_cache,
+			 int replace_extent,
+			 u32 extent_item_size,
+			 int *key_inserted)
 {
-	struct btrfs_root *root = BTRFS_I(inode)->root;
 	struct extent_buffer *leaf;
 	struct btrfs_file_extent_item *fi;
-	struct btrfs_path *path;
 	struct btrfs_key key;
 	struct btrfs_key new_key;
+	u64 ino = btrfs_ino(inode);
 	u64 search_start = start;
 	u64 disk_bytenr = 0;
 	u64 num_bytes = 0;
@@ -283,28 +715,34 @@ int btrfs_drop_extents(struct btrfs_trans_handle *trans, struct inode *inode,
 	int extent_type;
 	int recow;
 	int ret;
+	int modify_tree = -1;
+	int update_refs;
+	int found = 0;
+	int leafs_visited = 0;
 
 	if (drop_cache)
 		btrfs_drop_extent_cache(inode, start, end - 1, 0);
 
-	path = btrfs_alloc_path();
-	if (!path)
-		return -ENOMEM;
+	if (start >= BTRFS_I(inode)->disk_i_size && !replace_extent)
+		modify_tree = 0;
 
+	update_refs = (test_bit(BTRFS_ROOT_REF_COWS, &root->state) ||
+		       root == root->fs_info->tree_root);
 	while (1) {
 		recow = 0;
-		ret = btrfs_lookup_file_extent(trans, root, path, inode->i_ino,
-					       search_start, -1);
+		ret = btrfs_lookup_file_extent(trans, root, path, ino,
+					       search_start, modify_tree);
 		if (ret < 0)
 			break;
 		if (ret > 0 && path->slots[0] > 0 && search_start == start) {
 			leaf = path->nodes[0];
 			btrfs_item_key_to_cpu(leaf, &key, path->slots[0] - 1);
-			if (key.objectid == inode->i_ino &&
+			if (key.objectid == ino &&
 			    key.type == BTRFS_EXTENT_DATA_KEY)
 				path->slots[0]--;
 		}
 		ret = 0;
+		leafs_visited++;
 next_slot:
 		leaf = path->nodes[0];
 		if (path->slots[0] >= btrfs_header_nritems(leaf)) {
@@ -316,12 +754,13 @@ next_slot:
 				ret = 0;
 				break;
 			}
+			leafs_visited++;
 			leaf = path->nodes[0];
 			recow = 1;
 		}
 
 		btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
-		if (key.objectid > inode->i_ino ||
+		if (key.objectid > ino ||
 		    key.type > BTRFS_EXTENT_DATA_KEY || key.offset >= end)
 			break;
 
@@ -338,20 +777,35 @@ next_slot:
 				btrfs_file_extent_num_bytes(leaf, fi);
 		} else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
 			extent_end = key.offset +
-				btrfs_file_extent_inline_len(leaf, fi);
+				btrfs_file_extent_inline_len(leaf,
+						     path->slots[0], fi);
 		} else {
 			WARN_ON(1);
 			extent_end = search_start;
 		}
 
+		/*
+		 * Don't skip extent items representing 0 byte lengths. They
+		 * used to be created (bug) if while punching holes we hit
+		 * -ENOSPC condition. So if we find one here, just ensure we
+		 * delete it, otherwise we would insert a new file extent item
+		 * with the same key (offset) as that 0 bytes length file
+		 * extent item in the call to setup_items_for_insert() later
+		 * in this function.
+		 */
+		if (extent_end == key.offset && extent_end >= search_start)
+			goto delete_extent_item;
+
 		if (extent_end <= search_start) {
 			path->slots[0]++;
 			goto next_slot;
 		}
 
+		found = 1;
 		search_start = max(key.offset, start);
-		if (recow) {
-			btrfs_release_path(root, path);
+		if (recow || !modify_tree) {
+			modify_tree = -1;
+			btrfs_release_path(path);
 			continue;
 		}
 
@@ -361,14 +815,17 @@ next_slot:
 		 */
 		if (start > key.offset && end < extent_end) {
 			BUG_ON(del_nr > 0);
-			BUG_ON(extent_type == BTRFS_FILE_EXTENT_INLINE);
+			if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
+				ret = -EOPNOTSUPP;
+				break;
+			}
 
 			memcpy(&new_key, &key, sizeof(new_key));
 			new_key.offset = start;
 			ret = btrfs_duplicate_item(trans, root, path,
 						   &new_key);
 			if (ret == -EAGAIN) {
-				btrfs_release_path(root, path);
+				btrfs_release_path(path);
 				continue;
 			}
 			if (ret < 0)
@@ -389,14 +846,13 @@ next_slot:
 							extent_end - start);
 			btrfs_mark_buffer_dirty(leaf);
 
-			if (disk_bytenr > 0) {
+			if (update_refs && disk_bytenr > 0) {
 				ret = btrfs_inc_extent_ref(trans, root,
 						disk_bytenr, num_bytes, 0,
 						root->root_key.objectid,
 						new_key.objectid,
-						start - extent_offset);
-				BUG_ON(ret);
-				*hint_byte = disk_bytenr;
+						start - extent_offset, 1);
+				BUG_ON(ret); /* -ENOMEM */
 			}
 			key.offset = start;
 		}
@@ -405,21 +861,22 @@ next_slot:
 		 *      | -------- extent -------- |
 		 */
 		if (start <= key.offset && end < extent_end) {
-			BUG_ON(extent_type == BTRFS_FILE_EXTENT_INLINE);
+			if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
+				ret = -EOPNOTSUPP;
+				break;
+			}
 
 			memcpy(&new_key, &key, sizeof(new_key));
 			new_key.offset = end;
-			btrfs_set_item_key_safe(trans, root, path, &new_key);
+			btrfs_set_item_key_safe(root, path, &new_key);
 
 			extent_offset += end - key.offset;
 			btrfs_set_file_extent_offset(leaf, fi, extent_offset);
 			btrfs_set_file_extent_num_bytes(leaf, fi,
 							extent_end - end);
 			btrfs_mark_buffer_dirty(leaf);
-			if (disk_bytenr > 0) {
+			if (update_refs && disk_bytenr > 0)
 				inode_sub_bytes(inode, end - key.offset);
-				*hint_byte = disk_bytenr;
-			}
 			break;
 		}
 
@@ -430,15 +887,16 @@ next_slot:
 		 */
 		if (start > key.offset && end >= extent_end) {
 			BUG_ON(del_nr > 0);
-			BUG_ON(extent_type == BTRFS_FILE_EXTENT_INLINE);
+			if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
+				ret = -EOPNOTSUPP;
+				break;
+			}
 
 			btrfs_set_file_extent_num_bytes(leaf, fi,
 							start - key.offset);
 			btrfs_mark_buffer_dirty(leaf);
-			if (disk_bytenr > 0) {
+			if (update_refs && disk_bytenr > 0)
 				inode_sub_bytes(inode, extent_end - start);
-				*hint_byte = disk_bytenr;
-			}
 			if (end == extent_end)
 				break;
 
@@ -451,6 +909,7 @@ next_slot:
 		 *    | ------ extent ------ |
 		 */
 		if (start <= key.offset && end >= extent_end) {
+delete_extent_item:
 			if (del_nr == 0) {
 				del_slot = path->slots[0];
 				del_nr = 1;
@@ -459,21 +918,21 @@ next_slot:
 				del_nr++;
 			}
 
-			if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
+			if (update_refs &&
+			    extent_type == BTRFS_FILE_EXTENT_INLINE) {
 				inode_sub_bytes(inode,
 						extent_end - key.offset);
 				extent_end = ALIGN(extent_end,
 						   root->sectorsize);
-			} else if (disk_bytenr > 0) {
+			} else if (update_refs && disk_bytenr > 0) {
 				ret = btrfs_free_extent(trans, root,
 						disk_bytenr, num_bytes, 0,
 						root->root_key.objectid,
 						key.objectid, key.offset -
-						extent_offset);
-				BUG_ON(ret);
+						extent_offset, 0);
+				BUG_ON(ret); /* -ENOMEM */
 				inode_sub_bytes(inode,
 						extent_end - key.offset);
-				*hint_byte = disk_bytenr;
 			}
 
 			if (end == extent_end)
@@ -486,23 +945,83 @@ next_slot:
 
 			ret = btrfs_del_items(trans, root, path, del_slot,
 					      del_nr);
-			BUG_ON(ret);
+			if (ret) {
+				btrfs_abort_transaction(trans, root, ret);
+				break;
+			}
 
 			del_nr = 0;
 			del_slot = 0;
 
-			btrfs_release_path(root, path);
+			btrfs_release_path(path);
 			continue;
 		}
 
 		BUG_ON(1);
 	}
 
-	if (del_nr > 0) {
+	if (!ret && del_nr > 0) {
+		/*
+		 * Set path->slots[0] to first slot, so that after the delete
+		 * if items are move off from our leaf to its immediate left or
+		 * right neighbor leafs, we end up with a correct and adjusted
+		 * path->slots[0] for our insertion (if replace_extent != 0).
+		 */
+		path->slots[0] = del_slot;
 		ret = btrfs_del_items(trans, root, path, del_slot, del_nr);
-		BUG_ON(ret);
+		if (ret)
+			btrfs_abort_transaction(trans, root, ret);
 	}
 
+	leaf = path->nodes[0];
+	/*
+	 * If btrfs_del_items() was called, it might have deleted a leaf, in
+	 * which case it unlocked our path, so check path->locks[0] matches a
+	 * write lock.
+	 */
+	if (!ret && replace_extent && leafs_visited == 1 &&
+	    (path->locks[0] == BTRFS_WRITE_LOCK_BLOCKING ||
+	     path->locks[0] == BTRFS_WRITE_LOCK) &&
+	    btrfs_leaf_free_space(root, leaf) >=
+	    sizeof(struct btrfs_item) + extent_item_size) {
+
+		key.objectid = ino;
+		key.type = BTRFS_EXTENT_DATA_KEY;
+		key.offset = start;
+		if (!del_nr && path->slots[0] < btrfs_header_nritems(leaf)) {
+			struct btrfs_key slot_key;
+
+			btrfs_item_key_to_cpu(leaf, &slot_key, path->slots[0]);
+			if (btrfs_comp_cpu_keys(&key, &slot_key) > 0)
+				path->slots[0]++;
+		}
+		setup_items_for_insert(root, path, &key,
+				       &extent_item_size,
+				       extent_item_size,
+				       sizeof(struct btrfs_item) +
+				       extent_item_size, 1);
+		*key_inserted = 1;
+	}
+
+	if (!replace_extent || !(*key_inserted))
+		btrfs_release_path(path);
+	if (drop_end)
+		*drop_end = found ? min(end, extent_end) : end;
+	return ret;
+}
+
+int btrfs_drop_extents(struct btrfs_trans_handle *trans,
+		       struct btrfs_root *root, struct inode *inode, u64 start,
+		       u64 end, int drop_cache)
+{
+	struct btrfs_path *path;
+	int ret;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+	ret = __btrfs_drop_extents(trans, root, inode, path, start, end, NULL,
+				   drop_cache, 0, 0, NULL);
 	btrfs_free_path(path);
 	return ret;
 }
@@ -567,26 +1086,27 @@ int btrfs_mark_extent_written(struct btrfs_trans_handle *trans,
 	int del_slot = 0;
 	int recow;
 	int ret;
-
-	btrfs_drop_extent_cache(inode, start, end - 1, 0);
+	u64 ino = btrfs_ino(inode);
 
 	path = btrfs_alloc_path();
-	BUG_ON(!path);
+	if (!path)
+		return -ENOMEM;
 again:
 	recow = 0;
 	split = start;
-	key.objectid = inode->i_ino;
+	key.objectid = ino;
 	key.type = BTRFS_EXTENT_DATA_KEY;
 	key.offset = split;
 
 	ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
+	if (ret < 0)
+		goto out;
 	if (ret > 0 && path->slots[0] > 0)
 		path->slots[0]--;
 
 	leaf = path->nodes[0];
 	btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
-	BUG_ON(key.objectid != inode->i_ino ||
-	       key.type != BTRFS_EXTENT_DATA_KEY);
+	BUG_ON(key.objectid != ino || key.type != BTRFS_EXTENT_DATA_KEY);
 	fi = btrfs_item_ptr(leaf, path->slots[0],
 			    struct btrfs_file_extent_item);
 	BUG_ON(btrfs_file_extent_type(leaf, fi) !=
@@ -603,18 +1123,22 @@ again:
 		other_start = 0;
 		other_end = start;
 		if (extent_mergeable(leaf, path->slots[0] - 1,
-				     inode->i_ino, bytenr, orig_offset,
+				     ino, bytenr, orig_offset,
 				     &other_start, &other_end)) {
 			new_key.offset = end;
-			btrfs_set_item_key_safe(trans, root, path, &new_key);
+			btrfs_set_item_key_safe(root, path, &new_key);
 			fi = btrfs_item_ptr(leaf, path->slots[0],
 					    struct btrfs_file_extent_item);
+			btrfs_set_file_extent_generation(leaf, fi,
+							 trans->transid);
 			btrfs_set_file_extent_num_bytes(leaf, fi,
 							extent_end - end);
 			btrfs_set_file_extent_offset(leaf, fi,
 						     end - orig_offset);
 			fi = btrfs_item_ptr(leaf, path->slots[0] - 1,
 					    struct btrfs_file_extent_item);
+			btrfs_set_file_extent_generation(leaf, fi,
+							 trans->transid);
 			btrfs_set_file_extent_num_bytes(leaf, fi,
 							end - other_start);
 			btrfs_mark_buffer_dirty(leaf);
@@ -626,18 +1150,22 @@ again:
 		other_start = end;
 		other_end = 0;
 		if (extent_mergeable(leaf, path->slots[0] + 1,
-				     inode->i_ino, bytenr, orig_offset,
+				     ino, bytenr, orig_offset,
 				     &other_start, &other_end)) {
 			fi = btrfs_item_ptr(leaf, path->slots[0],
 					    struct btrfs_file_extent_item);
 			btrfs_set_file_extent_num_bytes(leaf, fi,
 							start - key.offset);
+			btrfs_set_file_extent_generation(leaf, fi,
+							 trans->transid);
 			path->slots[0]++;
 			new_key.offset = start;
-			btrfs_set_item_key_safe(trans, root, path, &new_key);
+			btrfs_set_item_key_safe(root, path, &new_key);
 
 			fi = btrfs_item_ptr(leaf, path->slots[0],
 					    struct btrfs_file_extent_item);
+			btrfs_set_file_extent_generation(leaf, fi,
+							 trans->transid);
 			btrfs_set_file_extent_num_bytes(leaf, fi,
 							other_end - start);
 			btrfs_set_file_extent_offset(leaf, fi,
@@ -654,20 +1182,25 @@ again:
 		new_key.offset = split;
 		ret = btrfs_duplicate_item(trans, root, path, &new_key);
 		if (ret == -EAGAIN) {
-			btrfs_release_path(root, path);
+			btrfs_release_path(path);
 			goto again;
 		}
-		BUG_ON(ret < 0);
+		if (ret < 0) {
+			btrfs_abort_transaction(trans, root, ret);
+			goto out;
+		}
 
 		leaf = path->nodes[0];
 		fi = btrfs_item_ptr(leaf, path->slots[0] - 1,
 				    struct btrfs_file_extent_item);
+		btrfs_set_file_extent_generation(leaf, fi, trans->transid);
 		btrfs_set_file_extent_num_bytes(leaf, fi,
 						split - key.offset);
 
 		fi = btrfs_item_ptr(leaf, path->slots[0],
 				    struct btrfs_file_extent_item);
 
+		btrfs_set_file_extent_generation(leaf, fi, trans->transid);
 		btrfs_set_file_extent_offset(leaf, fi, split - orig_offset);
 		btrfs_set_file_extent_num_bytes(leaf, fi,
 						extent_end - split);
@@ -675,8 +1208,8 @@ again:
 
 		ret = btrfs_inc_extent_ref(trans, root, bytenr, num_bytes, 0,
 					   root->root_key.objectid,
-					   inode->i_ino, orig_offset);
-		BUG_ON(ret);
+					   ino, orig_offset, 1);
+		BUG_ON(ret); /* -ENOMEM */
 
 		if (split == start) {
 			key.offset = start;
@@ -691,10 +1224,10 @@ again:
 	other_start = end;
 	other_end = 0;
 	if (extent_mergeable(leaf, path->slots[0] + 1,
-			     inode->i_ino, bytenr, orig_offset,
+			     ino, bytenr, orig_offset,
 			     &other_start, &other_end)) {
 		if (recow) {
-			btrfs_release_path(root, path);
+			btrfs_release_path(path);
 			goto again;
 		}
 		extent_end = other_end;
@@ -702,16 +1235,16 @@ again:
 		del_nr++;
 		ret = btrfs_free_extent(trans, root, bytenr, num_bytes,
 					0, root->root_key.objectid,
-					inode->i_ino, orig_offset);
-		BUG_ON(ret);
+					ino, orig_offset, 0);
+		BUG_ON(ret); /* -ENOMEM */
 	}
 	other_start = 0;
 	other_end = start;
 	if (extent_mergeable(leaf, path->slots[0] - 1,
-			     inode->i_ino, bytenr, orig_offset,
+			     ino, bytenr, orig_offset,
 			     &other_start, &other_end)) {
 		if (recow) {
-			btrfs_release_path(root, path);
+			btrfs_release_path(path);
 			goto again;
 		}
 		key.offset = other_start;
@@ -719,26 +1252,31 @@ again:
 		del_nr++;
 		ret = btrfs_free_extent(trans, root, bytenr, num_bytes,
 					0, root->root_key.objectid,
-					inode->i_ino, orig_offset);
-		BUG_ON(ret);
+					ino, orig_offset, 0);
+		BUG_ON(ret); /* -ENOMEM */
 	}
 	if (del_nr == 0) {
 		fi = btrfs_item_ptr(leaf, path->slots[0],
 			   struct btrfs_file_extent_item);
 		btrfs_set_file_extent_type(leaf, fi,
 					   BTRFS_FILE_EXTENT_REG);
+		btrfs_set_file_extent_generation(leaf, fi, trans->transid);
 		btrfs_mark_buffer_dirty(leaf);
 	} else {
 		fi = btrfs_item_ptr(leaf, del_slot - 1,
 			   struct btrfs_file_extent_item);
 		btrfs_set_file_extent_type(leaf, fi,
 					   BTRFS_FILE_EXTENT_REG);
+		btrfs_set_file_extent_generation(leaf, fi, trans->transid);
 		btrfs_set_file_extent_num_bytes(leaf, fi,
 						extent_end - key.offset);
 		btrfs_mark_buffer_dirty(leaf);
 
 		ret = btrfs_del_items(trans, root, path, del_slot, del_nr);
-		BUG_ON(ret);
+		if (ret < 0) {
+			btrfs_abort_transaction(trans, root, ret);
+			goto out;
+		}
 	}
 out:
 	btrfs_free_path(path);
@@ -746,245 +1284,526 @@ out:
 }
 
 /*
- * this gets pages into the page cache and locks them down, it also properly
- * waits for data=ordered extents to finish before allowing the pages to be
- * modified.
+ * on error we return an unlocked page and the error value
+ * on success we return a locked page and 0
  */
-static noinline int prepare_pages(struct btrfs_root *root, struct file *file,
-			 struct page **pages, size_t num_pages,
-			 loff_t pos, unsigned long first_index,
-			 unsigned long last_index, size_t write_bytes)
+static int prepare_uptodate_page(struct page *page, u64 pos,
+				 bool force_uptodate)
+{
+	int ret = 0;
+
+	if (((pos & (PAGE_CACHE_SIZE - 1)) || force_uptodate) &&
+	    !PageUptodate(page)) {
+		ret = btrfs_readpage(NULL, page);
+		if (ret)
+			return ret;
+		lock_page(page);
+		if (!PageUptodate(page)) {
+			unlock_page(page);
+			return -EIO;
+		}
+	}
+	return 0;
+}
+
+/*
+ * this just gets pages into the page cache and locks them down.
+ */
+static noinline int prepare_pages(struct inode *inode, struct page **pages,
+				  size_t num_pages, loff_t pos,
+				  size_t write_bytes, bool force_uptodate)
 {
-	struct extent_state *cached_state = NULL;
 	int i;
 	unsigned long index = pos >> PAGE_CACHE_SHIFT;
-	struct inode *inode = fdentry(file)->d_inode;
+	gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
 	int err = 0;
-	u64 start_pos;
-	u64 last_pos;
+	int faili;
 
-	start_pos = pos & ~((u64)root->sectorsize - 1);
-	last_pos = ((u64)index + num_pages) << PAGE_CACHE_SHIFT;
-
-	if (start_pos > inode->i_size) {
-		err = btrfs_cont_expand(inode, start_pos);
-		if (err)
-			return err;
-	}
-
-	memset(pages, 0, num_pages * sizeof(struct page *));
-again:
 	for (i = 0; i < num_pages; i++) {
-		pages[i] = grab_cache_page(inode->i_mapping, index + i);
+		pages[i] = find_or_create_page(inode->i_mapping, index + i,
+					       mask | __GFP_WRITE);
 		if (!pages[i]) {
+			faili = i - 1;
 			err = -ENOMEM;
-			BUG_ON(1);
+			goto fail;
+		}
+
+		if (i == 0)
+			err = prepare_uptodate_page(pages[i], pos,
+						    force_uptodate);
+		if (i == num_pages - 1)
+			err = prepare_uptodate_page(pages[i],
+						    pos + write_bytes, false);
+		if (err) {
+			page_cache_release(pages[i]);
+			faili = i - 1;
+			goto fail;
 		}
 		wait_on_page_writeback(pages[i]);
 	}
+
+	return 0;
+fail:
+	while (faili >= 0) {
+		unlock_page(pages[faili]);
+		page_cache_release(pages[faili]);
+		faili--;
+	}
+	return err;
+
+}
+
+/*
+ * This function locks the extent and properly waits for data=ordered extents
+ * to finish before allowing the pages to be modified if need.
+ *
+ * The return value:
+ * 1 - the extent is locked
+ * 0 - the extent is not locked, and everything is OK
+ * -EAGAIN - need re-prepare the pages
+ * the other < 0 number - Something wrong happens
+ */
+static noinline int
+lock_and_cleanup_extent_if_need(struct inode *inode, struct page **pages,
+				size_t num_pages, loff_t pos,
+				u64 *lockstart, u64 *lockend,
+				struct extent_state **cached_state)
+{
+	u64 start_pos;
+	u64 last_pos;
+	int i;
+	int ret = 0;
+
+	start_pos = pos & ~((u64)PAGE_CACHE_SIZE - 1);
+	last_pos = start_pos + ((u64)num_pages << PAGE_CACHE_SHIFT) - 1;
+
 	if (start_pos < inode->i_size) {
 		struct btrfs_ordered_extent *ordered;
 		lock_extent_bits(&BTRFS_I(inode)->io_tree,
-				 start_pos, last_pos - 1, 0, &cached_state,
-				 GFP_NOFS);
-		ordered = btrfs_lookup_first_ordered_extent(inode,
-							    last_pos - 1);
+				 start_pos, last_pos, 0, cached_state);
+		ordered = btrfs_lookup_ordered_range(inode, start_pos,
+						     last_pos - start_pos + 1);
 		if (ordered &&
 		    ordered->file_offset + ordered->len > start_pos &&
-		    ordered->file_offset < last_pos) {
-			btrfs_put_ordered_extent(ordered);
+		    ordered->file_offset <= last_pos) {
 			unlock_extent_cached(&BTRFS_I(inode)->io_tree,
-					     start_pos, last_pos - 1,
-					     &cached_state, GFP_NOFS);
+					     start_pos, last_pos,
+					     cached_state, GFP_NOFS);
 			for (i = 0; i < num_pages; i++) {
 				unlock_page(pages[i]);
 				page_cache_release(pages[i]);
 			}
-			btrfs_wait_ordered_range(inode, start_pos,
-						 last_pos - start_pos);
-			goto again;
+			btrfs_start_ordered_extent(inode, ordered, 1);
+			btrfs_put_ordered_extent(ordered);
+			return -EAGAIN;
 		}
 		if (ordered)
 			btrfs_put_ordered_extent(ordered);
 
 		clear_extent_bit(&BTRFS_I(inode)->io_tree, start_pos,
-				  last_pos - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
-				  EXTENT_DO_ACCOUNTING, 0, 0, &cached_state,
-				  GFP_NOFS);
-		unlock_extent_cached(&BTRFS_I(inode)->io_tree,
-				     start_pos, last_pos - 1, &cached_state,
-				     GFP_NOFS);
+				  last_pos, EXTENT_DIRTY | EXTENT_DELALLOC |
+				  EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG,
+				  0, 0, cached_state, GFP_NOFS);
+		*lockstart = start_pos;
+		*lockend = last_pos;
+		ret = 1;
 	}
+
 	for (i = 0; i < num_pages; i++) {
-		clear_page_dirty_for_io(pages[i]);
+		if (clear_page_dirty_for_io(pages[i]))
+			account_page_redirty(pages[i]);
 		set_page_extent_mapped(pages[i]);
 		WARN_ON(!PageLocked(pages[i]));
 	}
-	return 0;
+
+	return ret;
 }
 
-static ssize_t btrfs_file_write(struct file *file, const char __user *buf,
-				size_t count, loff_t *ppos)
+static noinline int check_can_nocow(struct inode *inode, loff_t pos,
+				    size_t *write_bytes)
 {
-	loff_t pos;
-	loff_t start_pos;
-	ssize_t num_written = 0;
-	ssize_t err = 0;
-	int ret = 0;
-	struct inode *inode = fdentry(file)->d_inode;
 	struct btrfs_root *root = BTRFS_I(inode)->root;
-	struct page **pages = NULL;
-	int nrptrs;
-	struct page *pinned[2];
-	unsigned long first_index;
-	unsigned long last_index;
-	int will_write;
+	struct btrfs_ordered_extent *ordered;
+	u64 lockstart, lockend;
+	u64 num_bytes;
+	int ret;
 
-	will_write = ((file->f_flags & O_DSYNC) || IS_SYNC(inode) ||
-		      (file->f_flags & O_DIRECT));
+	ret = btrfs_start_nocow_write(root);
+	if (!ret)
+		return -ENOSPC;
 
-	nrptrs = min((count + PAGE_CACHE_SIZE - 1) / PAGE_CACHE_SIZE,
-		     PAGE_CACHE_SIZE / (sizeof(struct page *)));
-	pinned[0] = NULL;
-	pinned[1] = NULL;
+	lockstart = round_down(pos, root->sectorsize);
+	lockend = round_up(pos + *write_bytes, root->sectorsize) - 1;
 
-	pos = *ppos;
-	start_pos = pos;
+	while (1) {
+		lock_extent(&BTRFS_I(inode)->io_tree, lockstart, lockend);
+		ordered = btrfs_lookup_ordered_range(inode, lockstart,
+						     lockend - lockstart + 1);
+		if (!ordered) {
+			break;
+		}
+		unlock_extent(&BTRFS_I(inode)->io_tree, lockstart, lockend);
+		btrfs_start_ordered_extent(inode, ordered, 1);
+		btrfs_put_ordered_extent(ordered);
+	}
 
-	vfs_check_frozen(inode->i_sb, SB_FREEZE_WRITE);
+	num_bytes = lockend - lockstart + 1;
+	ret = can_nocow_extent(inode, lockstart, &num_bytes, NULL, NULL, NULL);
+	if (ret <= 0) {
+		ret = 0;
+		btrfs_end_nocow_write(root);
+	} else {
+		*write_bytes = min_t(size_t, *write_bytes ,
+				     num_bytes - pos + lockstart);
+	}
 
-	/* do the reserve before the mutex lock in case we have to do some
-	 * flushing.  We wouldn't deadlock, but this is more polite.
-	 */
-	err = btrfs_reserve_metadata_for_delalloc(root, inode, 1);
-	if (err)
-		goto out_nolock;
+	unlock_extent(&BTRFS_I(inode)->io_tree, lockstart, lockend);
 
-	mutex_lock(&inode->i_mutex);
+	return ret;
+}
 
-	current->backing_dev_info = inode->i_mapping->backing_dev_info;
-	err = generic_write_checks(file, &pos, &count, S_ISBLK(inode->i_mode));
-	if (err)
-		goto out;
+static noinline ssize_t __btrfs_buffered_write(struct file *file,
+					       struct iov_iter *i,
+					       loff_t pos)
+{
+	struct inode *inode = file_inode(file);
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct page **pages = NULL;
+	struct extent_state *cached_state = NULL;
+	u64 release_bytes = 0;
+	u64 lockstart;
+	u64 lockend;
+	unsigned long first_index;
+	size_t num_written = 0;
+	int nrptrs;
+	int ret = 0;
+	bool only_release_metadata = false;
+	bool force_page_uptodate = false;
+	bool need_unlock;
+
+	nrptrs = min((iov_iter_count(i) + PAGE_CACHE_SIZE - 1) /
+		     PAGE_CACHE_SIZE, PAGE_CACHE_SIZE /
+		     (sizeof(struct page *)));
+	nrptrs = min(nrptrs, current->nr_dirtied_pause - current->nr_dirtied);
+	nrptrs = max(nrptrs, 8);
+	pages = kmalloc(nrptrs * sizeof(struct page *), GFP_KERNEL);
+	if (!pages)
+		return -ENOMEM;
 
-	if (count == 0)
-		goto out;
+	first_index = pos >> PAGE_CACHE_SHIFT;
 
-	err = file_remove_suid(file);
-	if (err)
-		goto out;
+	while (iov_iter_count(i) > 0) {
+		size_t offset = pos & (PAGE_CACHE_SIZE - 1);
+		size_t write_bytes = min(iov_iter_count(i),
+					 nrptrs * (size_t)PAGE_CACHE_SIZE -
+					 offset);
+		size_t num_pages = (write_bytes + offset +
+				    PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
+		size_t reserve_bytes;
+		size_t dirty_pages;
+		size_t copied;
 
-	file_update_time(file);
+		WARN_ON(num_pages > nrptrs);
 
-	pages = kmalloc(nrptrs * sizeof(struct page *), GFP_KERNEL);
+		/*
+		 * Fault pages before locking them in prepare_pages
+		 * to avoid recursive lock
+		 */
+		if (unlikely(iov_iter_fault_in_readable(i, write_bytes))) {
+			ret = -EFAULT;
+			break;
+		}
 
-	/* generic_write_checks can change our pos */
-	start_pos = pos;
+		reserve_bytes = num_pages << PAGE_CACHE_SHIFT;
+		ret = btrfs_check_data_free_space(inode, reserve_bytes);
+		if (ret == -ENOSPC &&
+		    (BTRFS_I(inode)->flags & (BTRFS_INODE_NODATACOW |
+					      BTRFS_INODE_PREALLOC))) {
+			ret = check_can_nocow(inode, pos, &write_bytes);
+			if (ret > 0) {
+				only_release_metadata = true;
+				/*
+				 * our prealloc extent may be smaller than
+				 * write_bytes, so scale down.
+				 */
+				num_pages = (write_bytes + offset +
+					     PAGE_CACHE_SIZE - 1) >>
+					PAGE_CACHE_SHIFT;
+				reserve_bytes = num_pages << PAGE_CACHE_SHIFT;
+				ret = 0;
+			} else {
+				ret = -ENOSPC;
+			}
+		}
 
-	BTRFS_I(inode)->sequence++;
-	first_index = pos >> PAGE_CACHE_SHIFT;
-	last_index = (pos + count) >> PAGE_CACHE_SHIFT;
+		if (ret)
+			break;
 
-	/*
-	 * there are lots of better ways to do this, but this code
-	 * makes sure the first and last page in the file range are
-	 * up to date and ready for cow
-	 */
-	if ((pos & (PAGE_CACHE_SIZE - 1))) {
-		pinned[0] = grab_cache_page(inode->i_mapping, first_index);
-		if (!PageUptodate(pinned[0])) {
-			ret = btrfs_readpage(NULL, pinned[0]);
-			BUG_ON(ret);
-			wait_on_page_locked(pinned[0]);
-		} else {
-			unlock_page(pinned[0]);
+		ret = btrfs_delalloc_reserve_metadata(inode, reserve_bytes);
+		if (ret) {
+			if (!only_release_metadata)
+				btrfs_free_reserved_data_space(inode,
+							       reserve_bytes);
+			else
+				btrfs_end_nocow_write(root);
+			break;
 		}
-	}
-	if ((pos + count) & (PAGE_CACHE_SIZE - 1)) {
-		pinned[1] = grab_cache_page(inode->i_mapping, last_index);
-		if (!PageUptodate(pinned[1])) {
-			ret = btrfs_readpage(NULL, pinned[1]);
-			BUG_ON(ret);
-			wait_on_page_locked(pinned[1]);
-		} else {
-			unlock_page(pinned[1]);
+
+		release_bytes = reserve_bytes;
+		need_unlock = false;
+again:
+		/*
+		 * This is going to setup the pages array with the number of
+		 * pages we want, so we don't really need to worry about the
+		 * contents of pages from loop to loop
+		 */
+		ret = prepare_pages(inode, pages, num_pages,
+				    pos, write_bytes,
+				    force_page_uptodate);
+		if (ret)
+			break;
+
+		ret = lock_and_cleanup_extent_if_need(inode, pages, num_pages,
+						      pos, &lockstart, &lockend,
+						      &cached_state);
+		if (ret < 0) {
+			if (ret == -EAGAIN)
+				goto again;
+			break;
+		} else if (ret > 0) {
+			need_unlock = true;
+			ret = 0;
 		}
-	}
 
-	while (count > 0) {
-		size_t offset = pos & (PAGE_CACHE_SIZE - 1);
-		size_t write_bytes = min(count, nrptrs *
-					(size_t)PAGE_CACHE_SIZE -
-					 offset);
-		size_t num_pages = (write_bytes + PAGE_CACHE_SIZE - 1) >>
-					PAGE_CACHE_SHIFT;
+		copied = btrfs_copy_from_user(pos, num_pages,
+					   write_bytes, pages, i);
 
-		WARN_ON(num_pages > nrptrs);
-		memset(pages, 0, sizeof(struct page *) * nrptrs);
+		/*
+		 * if we have trouble faulting in the pages, fall
+		 * back to one page at a time
+		 */
+		if (copied < write_bytes)
+			nrptrs = 1;
 
-		ret = btrfs_check_data_free_space(root, inode, write_bytes);
-		if (ret)
-			goto out;
+		if (copied == 0) {
+			force_page_uptodate = true;
+			dirty_pages = 0;
+		} else {
+			force_page_uptodate = false;
+			dirty_pages = (copied + offset +
+				       PAGE_CACHE_SIZE - 1) >>
+				       PAGE_CACHE_SHIFT;
+		}
 
-		ret = prepare_pages(root, file, pages, num_pages,
-				    pos, first_index, last_index,
-				    write_bytes);
-		if (ret) {
-			btrfs_free_reserved_data_space(root, inode,
-						       write_bytes);
-			goto out;
+		/*
+		 * If we had a short copy we need to release the excess delaloc
+		 * bytes we reserved.  We need to increment outstanding_extents
+		 * because btrfs_delalloc_release_space will decrement it, but
+		 * we still have an outstanding extent for the chunk we actually
+		 * managed to copy.
+		 */
+		if (num_pages > dirty_pages) {
+			release_bytes = (num_pages - dirty_pages) <<
+				PAGE_CACHE_SHIFT;
+			if (copied > 0) {
+				spin_lock(&BTRFS_I(inode)->lock);
+				BTRFS_I(inode)->outstanding_extents++;
+				spin_unlock(&BTRFS_I(inode)->lock);
+			}
+			if (only_release_metadata)
+				btrfs_delalloc_release_metadata(inode,
+								release_bytes);
+			else
+				btrfs_delalloc_release_space(inode,
+							     release_bytes);
 		}
 
-		ret = btrfs_copy_from_user(pos, num_pages,
-					   write_bytes, pages, buf);
+		release_bytes = dirty_pages << PAGE_CACHE_SHIFT;
+
+		if (copied > 0)
+			ret = btrfs_dirty_pages(root, inode, pages,
+						dirty_pages, pos, copied,
+						NULL);
+		if (need_unlock)
+			unlock_extent_cached(&BTRFS_I(inode)->io_tree,
+					     lockstart, lockend, &cached_state,
+					     GFP_NOFS);
 		if (ret) {
-			btrfs_free_reserved_data_space(root, inode,
-						       write_bytes);
 			btrfs_drop_pages(pages, num_pages);
-			goto out;
+			break;
 		}
 
-		ret = dirty_and_release_pages(NULL, root, file, pages,
-					      num_pages, pos, write_bytes);
-		btrfs_drop_pages(pages, num_pages);
-		if (ret) {
-			btrfs_free_reserved_data_space(root, inode,
-						       write_bytes);
-			goto out;
+		release_bytes = 0;
+		if (only_release_metadata)
+			btrfs_end_nocow_write(root);
+
+		if (only_release_metadata && copied > 0) {
+			u64 lockstart = round_down(pos, root->sectorsize);
+			u64 lockend = lockstart +
+				(dirty_pages << PAGE_CACHE_SHIFT) - 1;
+
+			set_extent_bit(&BTRFS_I(inode)->io_tree, lockstart,
+				       lockend, EXTENT_NORESERVE, NULL,
+				       NULL, GFP_NOFS);
+			only_release_metadata = false;
 		}
 
-		if (will_write) {
-			filemap_fdatawrite_range(inode->i_mapping, pos,
-						 pos + write_bytes - 1);
+		btrfs_drop_pages(pages, num_pages);
+
+		cond_resched();
+
+		balance_dirty_pages_ratelimited(inode->i_mapping);
+		if (dirty_pages < (root->leafsize >> PAGE_CACHE_SHIFT) + 1)
+			btrfs_btree_balance_dirty(root);
+
+		pos += copied;
+		num_written += copied;
+	}
+
+	kfree(pages);
+
+	if (release_bytes) {
+		if (only_release_metadata) {
+			btrfs_end_nocow_write(root);
+			btrfs_delalloc_release_metadata(inode, release_bytes);
 		} else {
-			balance_dirty_pages_ratelimited_nr(inode->i_mapping,
-							   num_pages);
-			if (num_pages <
-			    (root->leafsize >> PAGE_CACHE_SHIFT) + 1)
-				btrfs_btree_balance_dirty(root, 1);
-			btrfs_throttle(root);
+			btrfs_delalloc_release_space(inode, release_bytes);
 		}
+	}
 
-		buf += write_bytes;
-		count -= write_bytes;
-		pos += write_bytes;
-		num_written += write_bytes;
+	return num_written ? num_written : ret;
+}
 
-		cond_resched();
+static ssize_t __btrfs_direct_write(struct kiocb *iocb,
+				    struct iov_iter *from,
+				    loff_t pos)
+{
+	struct file *file = iocb->ki_filp;
+	ssize_t written;
+	ssize_t written_buffered;
+	loff_t endbyte;
+	int err;
+
+	written = generic_file_direct_write(iocb, from, pos);
+
+	if (written < 0 || !iov_iter_count(from))
+		return written;
+
+	pos += written;
+	written_buffered = __btrfs_buffered_write(file, from, pos);
+	if (written_buffered < 0) {
+		err = written_buffered;
+		goto out;
 	}
+	endbyte = pos + written_buffered - 1;
+	err = filemap_write_and_wait_range(file->f_mapping, pos, endbyte);
+	if (err)
+		goto out;
+	written += written_buffered;
+	iocb->ki_pos = pos + written_buffered;
+	invalidate_mapping_pages(file->f_mapping, pos >> PAGE_CACHE_SHIFT,
+				 endbyte >> PAGE_CACHE_SHIFT);
 out:
-	mutex_unlock(&inode->i_mutex);
-	if (ret)
-		err = ret;
-	btrfs_unreserve_metadata_for_delalloc(root, inode, 1);
+	return written ? written : err;
+}
 
-out_nolock:
-	kfree(pages);
-	if (pinned[0])
-		page_cache_release(pinned[0]);
-	if (pinned[1])
-		page_cache_release(pinned[1]);
-	*ppos = pos;
+static void update_time_for_write(struct inode *inode)
+{
+	struct timespec now;
+
+	if (IS_NOCMTIME(inode))
+		return;
+
+	now = current_fs_time(inode->i_sb);
+	if (!timespec_equal(&inode->i_mtime, &now))
+		inode->i_mtime = now;
+
+	if (!timespec_equal(&inode->i_ctime, &now))
+		inode->i_ctime = now;
+
+	if (IS_I_VERSION(inode))
+		inode_inc_iversion(inode);
+}
+
+static ssize_t btrfs_file_write_iter(struct kiocb *iocb,
+				    struct iov_iter *from)
+{
+	struct file *file = iocb->ki_filp;
+	struct inode *inode = file_inode(file);
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	u64 start_pos;
+	u64 end_pos;
+	ssize_t num_written = 0;
+	ssize_t err = 0;
+	size_t count = iov_iter_count(from);
+	bool sync = (file->f_flags & O_DSYNC) || IS_SYNC(file->f_mapping->host);
+	loff_t pos = iocb->ki_pos;
+
+	mutex_lock(&inode->i_mutex);
+
+	current->backing_dev_info = inode->i_mapping->backing_dev_info;
+	err = generic_write_checks(file, &pos, &count, S_ISBLK(inode->i_mode));
+	if (err) {
+		mutex_unlock(&inode->i_mutex);
+		goto out;
+	}
+
+	if (count == 0) {
+		mutex_unlock(&inode->i_mutex);
+		goto out;
+	}
+
+	iov_iter_truncate(from, count);
+
+	err = file_remove_suid(file);
+	if (err) {
+		mutex_unlock(&inode->i_mutex);
+		goto out;
+	}
+
+	/*
+	 * If BTRFS flips readonly due to some impossible error
+	 * (fs_info->fs_state now has BTRFS_SUPER_FLAG_ERROR),
+	 * although we have opened a file as writable, we have
+	 * to stop this write operation to ensure FS consistency.
+	 */
+	if (test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state)) {
+		mutex_unlock(&inode->i_mutex);
+		err = -EROFS;
+		goto out;
+	}
+
+	/*
+	 * We reserve space for updating the inode when we reserve space for the
+	 * extent we are going to write, so we will enospc out there.  We don't
+	 * need to start yet another transaction to update the inode as we will
+	 * update the inode when we finish writing whatever data we write.
+	 */
+	update_time_for_write(inode);
+
+	start_pos = round_down(pos, root->sectorsize);
+	if (start_pos > i_size_read(inode)) {
+		/* Expand hole size to cover write data, preventing empty gap */
+		end_pos = round_up(pos + count, root->sectorsize);
+		err = btrfs_cont_expand(inode, i_size_read(inode), end_pos);
+		if (err) {
+			mutex_unlock(&inode->i_mutex);
+			goto out;
+		}
+	}
+
+	if (sync)
+		atomic_inc(&BTRFS_I(inode)->sync_writers);
+
+	if (unlikely(file->f_flags & O_DIRECT)) {
+		num_written = __btrfs_direct_write(iocb, from, pos);
+	} else {
+		num_written = __btrfs_buffered_write(file, from, pos);
+		if (num_written > 0)
+			iocb->ki_pos = pos + num_written;
+	}
+
+	mutex_unlock(&inode->i_mutex);
 
 	/*
 	 * we want to make sure fsync finds this change
@@ -997,38 +1816,22 @@ out_nolock:
 	 * this will either be one more than the running transaction
 	 * or the generation used for the next transaction if there isn't
 	 * one running right now.
+	 *
+	 * We also have to set last_sub_trans to the current log transid,
+	 * otherwise subsequent syncs to a file that's been synced in this
+	 * transaction will appear to have already occured.
 	 */
 	BTRFS_I(inode)->last_trans = root->fs_info->generation + 1;
-
-	if (num_written > 0 && will_write) {
-		struct btrfs_trans_handle *trans;
-
-		err = btrfs_wait_ordered_range(inode, start_pos, num_written);
-		if (err)
+	BTRFS_I(inode)->last_sub_trans = root->log_transid;
+	if (num_written > 0) {
+		err = generic_write_sync(file, pos, num_written);
+		if (err < 0)
 			num_written = err;
-
-		if ((file->f_flags & O_DSYNC) || IS_SYNC(inode)) {
-			trans = btrfs_start_transaction(root, 1);
-			ret = btrfs_log_dentry_safe(trans, root,
-						    file->f_dentry);
-			if (ret == 0) {
-				ret = btrfs_sync_log(trans, root);
-				if (ret == 0)
-					btrfs_end_transaction(trans, root);
-				else
-					btrfs_commit_transaction(trans, root);
-			} else if (ret != BTRFS_NO_LOG_SYNC) {
-				btrfs_commit_transaction(trans, root);
-			} else {
-				btrfs_end_transaction(trans, root);
-			}
-		}
-		if (file->f_flags & O_DIRECT) {
-			invalidate_mapping_pages(inode->i_mapping,
-			      start_pos >> PAGE_CACHE_SHIFT,
-			     (start_pos + num_written - 1) >> PAGE_CACHE_SHIFT);
-		}
 	}
+
+	if (sync)
+		atomic_dec(&BTRFS_I(inode)->sync_writers);
+out:
 	current->backing_dev_info = NULL;
 	return num_written ? num_written : err;
 }
@@ -1041,9 +1844,22 @@ int btrfs_release_file(struct inode *inode, struct file *filp)
 	 * flush down new bytes that may have been written if the
 	 * application were using truncate to replace a file in place.
 	 */
-	if (BTRFS_I(inode)->ordered_data_close) {
-		BTRFS_I(inode)->ordered_data_close = 0;
-		btrfs_add_ordered_operation(NULL, BTRFS_I(inode)->root, inode);
+	if (test_and_clear_bit(BTRFS_INODE_ORDERED_DATA_CLOSE,
+			       &BTRFS_I(inode)->runtime_flags)) {
+		struct btrfs_trans_handle *trans;
+		struct btrfs_root *root = BTRFS_I(inode)->root;
+
+		/*
+		 * We need to block on a committing transaction to keep us from
+		 * throwing a ordered operation on to the list and causing
+		 * something like sync to deadlock trying to flush out this
+		 * inode.
+		 */
+		trans = btrfs_start_transaction(root, 0);
+		if (IS_ERR(trans))
+			return PTR_ERR(trans);
+		btrfs_add_ordered_operation(trans, BTRFS_I(inode)->root, inode);
+		btrfs_end_transaction(trans, root);
 		if (inode->i_size > BTRFS_ORDERED_OPERATIONS_FLUSH_LIMIT)
 			filemap_flush(inode->i_mapping);
 	}
@@ -1063,56 +1879,114 @@ int btrfs_release_file(struct inode *inode, struct file *filp)
  * important optimization for directories because holding the mutex prevents
  * new operations on the dir while we write to disk.
  */
-int btrfs_sync_file(struct file *file, struct dentry *dentry, int datasync)
+int btrfs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
 {
+	struct dentry *dentry = file->f_path.dentry;
 	struct inode *inode = dentry->d_inode;
 	struct btrfs_root *root = BTRFS_I(inode)->root;
-	int ret = 0;
 	struct btrfs_trans_handle *trans;
+	struct btrfs_log_ctx ctx;
+	int ret = 0;
+	bool full_sync = 0;
 
+	trace_btrfs_sync_file(file, datasync);
+
+	/*
+	 * We write the dirty pages in the range and wait until they complete
+	 * out of the ->i_mutex. If so, we can flush the dirty pages by
+	 * multi-task, and make the performance up.  See
+	 * btrfs_wait_ordered_range for an explanation of the ASYNC check.
+	 */
+	atomic_inc(&BTRFS_I(inode)->sync_writers);
+	ret = filemap_fdatawrite_range(inode->i_mapping, start, end);
+	if (!ret && test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
+			     &BTRFS_I(inode)->runtime_flags))
+		ret = filemap_fdatawrite_range(inode->i_mapping, start, end);
+	atomic_dec(&BTRFS_I(inode)->sync_writers);
+	if (ret)
+		return ret;
 
-	/* we wait first, since the writeback may change the inode */
-	root->log_batch++;
-	/* the VFS called filemap_fdatawrite for us */
-	btrfs_wait_ordered_range(inode, 0, (u64)-1);
-	root->log_batch++;
+	mutex_lock(&inode->i_mutex);
+
+	/*
+	 * We flush the dirty pages again to avoid some dirty pages in the
+	 * range being left.
+	 */
+	atomic_inc(&root->log_batch);
+	full_sync = test_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
+			     &BTRFS_I(inode)->runtime_flags);
+	if (full_sync) {
+		ret = btrfs_wait_ordered_range(inode, start, end - start + 1);
+		if (ret) {
+			mutex_unlock(&inode->i_mutex);
+			goto out;
+		}
+	}
+	atomic_inc(&root->log_batch);
 
 	/*
 	 * check the transaction that last modified this inode
 	 * and see if its already been committed
 	 */
-	if (!BTRFS_I(inode)->last_trans)
+	if (!BTRFS_I(inode)->last_trans) {
+		mutex_unlock(&inode->i_mutex);
 		goto out;
+	}
 
 	/*
 	 * if the last transaction that changed this file was before
 	 * the current transaction, we can bail out now without any
 	 * syncing
 	 */
-	mutex_lock(&root->fs_info->trans_mutex);
-	if (BTRFS_I(inode)->last_trans <=
+	smp_mb();
+	if (btrfs_inode_in_log(inode, root->fs_info->generation) ||
+	    BTRFS_I(inode)->last_trans <=
 	    root->fs_info->last_trans_committed) {
 		BTRFS_I(inode)->last_trans = 0;
-		mutex_unlock(&root->fs_info->trans_mutex);
+
+		/*
+		 * We'v had everything committed since the last time we were
+		 * modified so clear this flag in case it was set for whatever
+		 * reason, it's no longer relevant.
+		 */
+		clear_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
+			  &BTRFS_I(inode)->runtime_flags);
+		mutex_unlock(&inode->i_mutex);
 		goto out;
 	}
-	mutex_unlock(&root->fs_info->trans_mutex);
 
 	/*
 	 * ok we haven't committed the transaction yet, lets do a commit
 	 */
-	if (file && file->private_data)
+	if (file->private_data)
 		btrfs_ioctl_trans_end(file);
 
-	trans = btrfs_start_transaction(root, 1);
-	if (!trans) {
-		ret = -ENOMEM;
+	/*
+	 * We use start here because we will need to wait on the IO to complete
+	 * in btrfs_sync_log, which could require joining a transaction (for
+	 * example checking cross references in the nocow path).  If we use join
+	 * here we could get into a situation where we're waiting on IO to
+	 * happen that is blocked on a transaction trying to commit.  With start
+	 * we inc the extwriter counter, so we wait for all extwriters to exit
+	 * before we start blocking join'ers.  This comment is to keep somebody
+	 * from thinking they are super smart and changing this to
+	 * btrfs_join_transaction *cough*Josef*cough*.
+	 */
+	trans = btrfs_start_transaction(root, 0);
+	if (IS_ERR(trans)) {
+		ret = PTR_ERR(trans);
+		mutex_unlock(&inode->i_mutex);
 		goto out;
 	}
+	trans->sync = true;
 
-	ret = btrfs_log_dentry_safe(trans, root, dentry);
-	if (ret < 0)
-		goto out;
+	btrfs_init_log_ctx(&ctx);
+
+	ret = btrfs_log_dentry_safe(trans, root, dentry, &ctx);
+	if (ret < 0) {
+		/* Fallthrough and commit/free transaction. */
+		ret = 1;
+	}
 
 	/* we've logged all the items and now have a consistent
 	 * version of the file in the log.  It is possible that
@@ -1124,50 +1998,763 @@ int btrfs_sync_file(struct file *file, struct dentry *dentry, int datasync)
 	 * file again, but that will end up using the synchronization
 	 * inside btrfs_sync_log to keep things safe.
 	 */
-	mutex_unlock(&dentry->d_inode->i_mutex);
+	mutex_unlock(&inode->i_mutex);
 
 	if (ret != BTRFS_NO_LOG_SYNC) {
-		if (ret > 0) {
-			ret = btrfs_commit_transaction(trans, root);
-		} else {
-			ret = btrfs_sync_log(trans, root);
-			if (ret == 0)
+		if (!ret) {
+			ret = btrfs_sync_log(trans, root, &ctx);
+			if (!ret) {
 				ret = btrfs_end_transaction(trans, root);
-			else
-				ret = btrfs_commit_transaction(trans, root);
+				goto out;
+			}
 		}
+		if (!full_sync) {
+			ret = btrfs_wait_ordered_range(inode, start,
+						       end - start + 1);
+			if (ret) {
+				btrfs_end_transaction(trans, root);
+				goto out;
+			}
+		}
+		ret = btrfs_commit_transaction(trans, root);
 	} else {
 		ret = btrfs_end_transaction(trans, root);
 	}
-	mutex_lock(&dentry->d_inode->i_mutex);
 out:
 	return ret > 0 ? -EIO : ret;
 }
 
 static const struct vm_operations_struct btrfs_file_vm_ops = {
 	.fault		= filemap_fault,
+	.map_pages	= filemap_map_pages,
 	.page_mkwrite	= btrfs_page_mkwrite,
+	.remap_pages	= generic_file_remap_pages,
 };
 
 static int btrfs_file_mmap(struct file	*filp, struct vm_area_struct *vma)
 {
-	vma->vm_ops = &btrfs_file_vm_ops;
+	struct address_space *mapping = filp->f_mapping;
+
+	if (!mapping->a_ops->readpage)
+		return -ENOEXEC;
+
 	file_accessed(filp);
+	vma->vm_ops = &btrfs_file_vm_ops;
+
 	return 0;
 }
 
+static int hole_mergeable(struct inode *inode, struct extent_buffer *leaf,
+			  int slot, u64 start, u64 end)
+{
+	struct btrfs_file_extent_item *fi;
+	struct btrfs_key key;
+
+	if (slot < 0 || slot >= btrfs_header_nritems(leaf))
+		return 0;
+
+	btrfs_item_key_to_cpu(leaf, &key, slot);
+	if (key.objectid != btrfs_ino(inode) ||
+	    key.type != BTRFS_EXTENT_DATA_KEY)
+		return 0;
+
+	fi = btrfs_item_ptr(leaf, slot, struct btrfs_file_extent_item);
+
+	if (btrfs_file_extent_type(leaf, fi) != BTRFS_FILE_EXTENT_REG)
+		return 0;
+
+	if (btrfs_file_extent_disk_bytenr(leaf, fi))
+		return 0;
+
+	if (key.offset == end)
+		return 1;
+	if (key.offset + btrfs_file_extent_num_bytes(leaf, fi) == start)
+		return 1;
+	return 0;
+}
+
+static int fill_holes(struct btrfs_trans_handle *trans, struct inode *inode,
+		      struct btrfs_path *path, u64 offset, u64 end)
+{
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct extent_buffer *leaf;
+	struct btrfs_file_extent_item *fi;
+	struct extent_map *hole_em;
+	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
+	struct btrfs_key key;
+	int ret;
+
+	if (btrfs_fs_incompat(root->fs_info, NO_HOLES))
+		goto out;
+
+	key.objectid = btrfs_ino(inode);
+	key.type = BTRFS_EXTENT_DATA_KEY;
+	key.offset = offset;
+
+	ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
+	if (ret < 0)
+		return ret;
+	BUG_ON(!ret);
+
+	leaf = path->nodes[0];
+	if (hole_mergeable(inode, leaf, path->slots[0]-1, offset, end)) {
+		u64 num_bytes;
+
+		path->slots[0]--;
+		fi = btrfs_item_ptr(leaf, path->slots[0],
+				    struct btrfs_file_extent_item);
+		num_bytes = btrfs_file_extent_num_bytes(leaf, fi) +
+			end - offset;
+		btrfs_set_file_extent_num_bytes(leaf, fi, num_bytes);
+		btrfs_set_file_extent_ram_bytes(leaf, fi, num_bytes);
+		btrfs_set_file_extent_offset(leaf, fi, 0);
+		btrfs_mark_buffer_dirty(leaf);
+		goto out;
+	}
+
+	if (hole_mergeable(inode, leaf, path->slots[0]+1, offset, end)) {
+		u64 num_bytes;
+
+		path->slots[0]++;
+		key.offset = offset;
+		btrfs_set_item_key_safe(root, path, &key);
+		fi = btrfs_item_ptr(leaf, path->slots[0],
+				    struct btrfs_file_extent_item);
+		num_bytes = btrfs_file_extent_num_bytes(leaf, fi) + end -
+			offset;
+		btrfs_set_file_extent_num_bytes(leaf, fi, num_bytes);
+		btrfs_set_file_extent_ram_bytes(leaf, fi, num_bytes);
+		btrfs_set_file_extent_offset(leaf, fi, 0);
+		btrfs_mark_buffer_dirty(leaf);
+		goto out;
+	}
+	btrfs_release_path(path);
+
+	ret = btrfs_insert_file_extent(trans, root, btrfs_ino(inode), offset,
+				       0, 0, end - offset, 0, end - offset,
+				       0, 0, 0);
+	if (ret)
+		return ret;
+
+out:
+	btrfs_release_path(path);
+
+	hole_em = alloc_extent_map();
+	if (!hole_em) {
+		btrfs_drop_extent_cache(inode, offset, end - 1, 0);
+		set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
+			&BTRFS_I(inode)->runtime_flags);
+	} else {
+		hole_em->start = offset;
+		hole_em->len = end - offset;
+		hole_em->ram_bytes = hole_em->len;
+		hole_em->orig_start = offset;
+
+		hole_em->block_start = EXTENT_MAP_HOLE;
+		hole_em->block_len = 0;
+		hole_em->orig_block_len = 0;
+		hole_em->bdev = root->fs_info->fs_devices->latest_bdev;
+		hole_em->compress_type = BTRFS_COMPRESS_NONE;
+		hole_em->generation = trans->transid;
+
+		do {
+			btrfs_drop_extent_cache(inode, offset, end - 1, 0);
+			write_lock(&em_tree->lock);
+			ret = add_extent_mapping(em_tree, hole_em, 1);
+			write_unlock(&em_tree->lock);
+		} while (ret == -EEXIST);
+		free_extent_map(hole_em);
+		if (ret)
+			set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
+				&BTRFS_I(inode)->runtime_flags);
+	}
+
+	return 0;
+}
+
+/*
+ * Find a hole extent on given inode and change start/len to the end of hole
+ * extent.(hole/vacuum extent whose em->start <= start &&
+ *	   em->start + em->len > start)
+ * When a hole extent is found, return 1 and modify start/len.
+ */
+static int find_first_non_hole(struct inode *inode, u64 *start, u64 *len)
+{
+	struct extent_map *em;
+	int ret = 0;
+
+	em = btrfs_get_extent(inode, NULL, 0, *start, *len, 0);
+	if (IS_ERR_OR_NULL(em)) {
+		if (!em)
+			ret = -ENOMEM;
+		else
+			ret = PTR_ERR(em);
+		return ret;
+	}
+
+	/* Hole or vacuum extent(only exists in no-hole mode) */
+	if (em->block_start == EXTENT_MAP_HOLE) {
+		ret = 1;
+		*len = em->start + em->len > *start + *len ?
+		       0 : *start + *len - em->start - em->len;
+		*start = em->start + em->len;
+	}
+	free_extent_map(em);
+	return ret;
+}
+
+static int btrfs_punch_hole(struct inode *inode, loff_t offset, loff_t len)
+{
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct extent_state *cached_state = NULL;
+	struct btrfs_path *path;
+	struct btrfs_block_rsv *rsv;
+	struct btrfs_trans_handle *trans;
+	u64 lockstart;
+	u64 lockend;
+	u64 tail_start;
+	u64 tail_len;
+	u64 orig_start = offset;
+	u64 cur_offset;
+	u64 min_size = btrfs_calc_trunc_metadata_size(root, 1);
+	u64 drop_end;
+	int ret = 0;
+	int err = 0;
+	int rsv_count;
+	bool same_page;
+	bool no_holes = btrfs_fs_incompat(root->fs_info, NO_HOLES);
+	u64 ino_size;
+
+	ret = btrfs_wait_ordered_range(inode, offset, len);
+	if (ret)
+		return ret;
+
+	mutex_lock(&inode->i_mutex);
+	ino_size = round_up(inode->i_size, PAGE_CACHE_SIZE);
+	ret = find_first_non_hole(inode, &offset, &len);
+	if (ret < 0)
+		goto out_only_mutex;
+	if (ret && !len) {
+		/* Already in a large hole */
+		ret = 0;
+		goto out_only_mutex;
+	}
+
+	lockstart = round_up(offset , BTRFS_I(inode)->root->sectorsize);
+	lockend = round_down(offset + len,
+			     BTRFS_I(inode)->root->sectorsize) - 1;
+	same_page = ((offset >> PAGE_CACHE_SHIFT) ==
+		    ((offset + len - 1) >> PAGE_CACHE_SHIFT));
+
+	/*
+	 * We needn't truncate any page which is beyond the end of the file
+	 * because we are sure there is no data there.
+	 */
+	/*
+	 * Only do this if we are in the same page and we aren't doing the
+	 * entire page.
+	 */
+	if (same_page && len < PAGE_CACHE_SIZE) {
+		if (offset < ino_size)
+			ret = btrfs_truncate_page(inode, offset, len, 0);
+		goto out_only_mutex;
+	}
+
+	/* zero back part of the first page */
+	if (offset < ino_size) {
+		ret = btrfs_truncate_page(inode, offset, 0, 0);
+		if (ret) {
+			mutex_unlock(&inode->i_mutex);
+			return ret;
+		}
+	}
+
+	/* Check the aligned pages after the first unaligned page,
+	 * if offset != orig_start, which means the first unaligned page
+	 * including serveral following pages are already in holes,
+	 * the extra check can be skipped */
+	if (offset == orig_start) {
+		/* after truncate page, check hole again */
+		len = offset + len - lockstart;
+		offset = lockstart;
+		ret = find_first_non_hole(inode, &offset, &len);
+		if (ret < 0)
+			goto out_only_mutex;
+		if (ret && !len) {
+			ret = 0;
+			goto out_only_mutex;
+		}
+		lockstart = offset;
+	}
+
+	/* Check the tail unaligned part is in a hole */
+	tail_start = lockend + 1;
+	tail_len = offset + len - tail_start;
+	if (tail_len) {
+		ret = find_first_non_hole(inode, &tail_start, &tail_len);
+		if (unlikely(ret < 0))
+			goto out_only_mutex;
+		if (!ret) {
+			/* zero the front end of the last page */
+			if (tail_start + tail_len < ino_size) {
+				ret = btrfs_truncate_page(inode,
+						tail_start + tail_len, 0, 1);
+				if (ret)
+					goto out_only_mutex;
+				}
+		}
+	}
+
+	if (lockend < lockstart) {
+		mutex_unlock(&inode->i_mutex);
+		return 0;
+	}
+
+	while (1) {
+		struct btrfs_ordered_extent *ordered;
+
+		truncate_pagecache_range(inode, lockstart, lockend);
+
+		lock_extent_bits(&BTRFS_I(inode)->io_tree, lockstart, lockend,
+				 0, &cached_state);
+		ordered = btrfs_lookup_first_ordered_extent(inode, lockend);
+
+		/*
+		 * We need to make sure we have no ordered extents in this range
+		 * and nobody raced in and read a page in this range, if we did
+		 * we need to try again.
+		 */
+		if ((!ordered ||
+		    (ordered->file_offset + ordered->len <= lockstart ||
+		     ordered->file_offset > lockend)) &&
+		     !btrfs_page_exists_in_range(inode, lockstart, lockend)) {
+			if (ordered)
+				btrfs_put_ordered_extent(ordered);
+			break;
+		}
+		if (ordered)
+			btrfs_put_ordered_extent(ordered);
+		unlock_extent_cached(&BTRFS_I(inode)->io_tree, lockstart,
+				     lockend, &cached_state, GFP_NOFS);
+		ret = btrfs_wait_ordered_range(inode, lockstart,
+					       lockend - lockstart + 1);
+		if (ret) {
+			mutex_unlock(&inode->i_mutex);
+			return ret;
+		}
+	}
+
+	path = btrfs_alloc_path();
+	if (!path) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	rsv = btrfs_alloc_block_rsv(root, BTRFS_BLOCK_RSV_TEMP);
+	if (!rsv) {
+		ret = -ENOMEM;
+		goto out_free;
+	}
+	rsv->size = btrfs_calc_trunc_metadata_size(root, 1);
+	rsv->failfast = 1;
+
+	/*
+	 * 1 - update the inode
+	 * 1 - removing the extents in the range
+	 * 1 - adding the hole extent if no_holes isn't set
+	 */
+	rsv_count = no_holes ? 2 : 3;
+	trans = btrfs_start_transaction(root, rsv_count);
+	if (IS_ERR(trans)) {
+		err = PTR_ERR(trans);
+		goto out_free;
+	}
+
+	ret = btrfs_block_rsv_migrate(&root->fs_info->trans_block_rsv, rsv,
+				      min_size);
+	BUG_ON(ret);
+	trans->block_rsv = rsv;
+
+	cur_offset = lockstart;
+	len = lockend - cur_offset;
+	while (cur_offset < lockend) {
+		ret = __btrfs_drop_extents(trans, root, inode, path,
+					   cur_offset, lockend + 1,
+					   &drop_end, 1, 0, 0, NULL);
+		if (ret != -ENOSPC)
+			break;
+
+		trans->block_rsv = &root->fs_info->trans_block_rsv;
+
+		if (cur_offset < ino_size) {
+			ret = fill_holes(trans, inode, path, cur_offset,
+					 drop_end);
+			if (ret) {
+				err = ret;
+				break;
+			}
+		}
+
+		cur_offset = drop_end;
+
+		ret = btrfs_update_inode(trans, root, inode);
+		if (ret) {
+			err = ret;
+			break;
+		}
+
+		btrfs_end_transaction(trans, root);
+		btrfs_btree_balance_dirty(root);
+
+		trans = btrfs_start_transaction(root, rsv_count);
+		if (IS_ERR(trans)) {
+			ret = PTR_ERR(trans);
+			trans = NULL;
+			break;
+		}
+
+		ret = btrfs_block_rsv_migrate(&root->fs_info->trans_block_rsv,
+					      rsv, min_size);
+		BUG_ON(ret);	/* shouldn't happen */
+		trans->block_rsv = rsv;
+
+		ret = find_first_non_hole(inode, &cur_offset, &len);
+		if (unlikely(ret < 0))
+			break;
+		if (ret && !len) {
+			ret = 0;
+			break;
+		}
+	}
+
+	if (ret) {
+		err = ret;
+		goto out_trans;
+	}
+
+	trans->block_rsv = &root->fs_info->trans_block_rsv;
+	/*
+	 * Don't insert file hole extent item if it's for a range beyond eof
+	 * (because it's useless) or if it represents a 0 bytes range (when
+	 * cur_offset == drop_end).
+	 */
+	if (cur_offset < ino_size && cur_offset < drop_end) {
+		ret = fill_holes(trans, inode, path, cur_offset, drop_end);
+		if (ret) {
+			err = ret;
+			goto out_trans;
+		}
+	}
+
+out_trans:
+	if (!trans)
+		goto out_free;
+
+	inode_inc_iversion(inode);
+	inode->i_mtime = inode->i_ctime = CURRENT_TIME;
+
+	trans->block_rsv = &root->fs_info->trans_block_rsv;
+	ret = btrfs_update_inode(trans, root, inode);
+	btrfs_end_transaction(trans, root);
+	btrfs_btree_balance_dirty(root);
+out_free:
+	btrfs_free_path(path);
+	btrfs_free_block_rsv(root, rsv);
+out:
+	unlock_extent_cached(&BTRFS_I(inode)->io_tree, lockstart, lockend,
+			     &cached_state, GFP_NOFS);
+out_only_mutex:
+	mutex_unlock(&inode->i_mutex);
+	if (ret && !err)
+		err = ret;
+	return err;
+}
+
+static long btrfs_fallocate(struct file *file, int mode,
+			    loff_t offset, loff_t len)
+{
+	struct inode *inode = file_inode(file);
+	struct extent_state *cached_state = NULL;
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	u64 cur_offset;
+	u64 last_byte;
+	u64 alloc_start;
+	u64 alloc_end;
+	u64 alloc_hint = 0;
+	u64 locked_end;
+	struct extent_map *em;
+	int blocksize = BTRFS_I(inode)->root->sectorsize;
+	int ret;
+
+	alloc_start = round_down(offset, blocksize);
+	alloc_end = round_up(offset + len, blocksize);
+
+	/* Make sure we aren't being give some crap mode */
+	if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
+		return -EOPNOTSUPP;
+
+	if (mode & FALLOC_FL_PUNCH_HOLE)
+		return btrfs_punch_hole(inode, offset, len);
+
+	/*
+	 * Make sure we have enough space before we do the
+	 * allocation.
+	 */
+	ret = btrfs_check_data_free_space(inode, alloc_end - alloc_start);
+	if (ret)
+		return ret;
+	if (root->fs_info->quota_enabled) {
+		ret = btrfs_qgroup_reserve(root, alloc_end - alloc_start);
+		if (ret)
+			goto out_reserve_fail;
+	}
+
+	mutex_lock(&inode->i_mutex);
+	ret = inode_newsize_ok(inode, alloc_end);
+	if (ret)
+		goto out;
+
+	if (alloc_start > inode->i_size) {
+		ret = btrfs_cont_expand(inode, i_size_read(inode),
+					alloc_start);
+		if (ret)
+			goto out;
+	} else {
+		/*
+		 * If we are fallocating from the end of the file onward we
+		 * need to zero out the end of the page if i_size lands in the
+		 * middle of a page.
+		 */
+		ret = btrfs_truncate_page(inode, inode->i_size, 0, 0);
+		if (ret)
+			goto out;
+	}
+
+	/*
+	 * wait for ordered IO before we have any locks.  We'll loop again
+	 * below with the locks held.
+	 */
+	ret = btrfs_wait_ordered_range(inode, alloc_start,
+				       alloc_end - alloc_start);
+	if (ret)
+		goto out;
+
+	locked_end = alloc_end - 1;
+	while (1) {
+		struct btrfs_ordered_extent *ordered;
+
+		/* the extent lock is ordered inside the running
+		 * transaction
+		 */
+		lock_extent_bits(&BTRFS_I(inode)->io_tree, alloc_start,
+				 locked_end, 0, &cached_state);
+		ordered = btrfs_lookup_first_ordered_extent(inode,
+							    alloc_end - 1);
+		if (ordered &&
+		    ordered->file_offset + ordered->len > alloc_start &&
+		    ordered->file_offset < alloc_end) {
+			btrfs_put_ordered_extent(ordered);
+			unlock_extent_cached(&BTRFS_I(inode)->io_tree,
+					     alloc_start, locked_end,
+					     &cached_state, GFP_NOFS);
+			/*
+			 * we can't wait on the range with the transaction
+			 * running or with the extent lock held
+			 */
+			ret = btrfs_wait_ordered_range(inode, alloc_start,
+						       alloc_end - alloc_start);
+			if (ret)
+				goto out;
+		} else {
+			if (ordered)
+				btrfs_put_ordered_extent(ordered);
+			break;
+		}
+	}
+
+	cur_offset = alloc_start;
+	while (1) {
+		u64 actual_end;
+
+		em = btrfs_get_extent(inode, NULL, 0, cur_offset,
+				      alloc_end - cur_offset, 0);
+		if (IS_ERR_OR_NULL(em)) {
+			if (!em)
+				ret = -ENOMEM;
+			else
+				ret = PTR_ERR(em);
+			break;
+		}
+		last_byte = min(extent_map_end(em), alloc_end);
+		actual_end = min_t(u64, extent_map_end(em), offset + len);
+		last_byte = ALIGN(last_byte, blocksize);
+
+		if (em->block_start == EXTENT_MAP_HOLE ||
+		    (cur_offset >= inode->i_size &&
+		     !test_bit(EXTENT_FLAG_PREALLOC, &em->flags))) {
+			ret = btrfs_prealloc_file_range(inode, mode, cur_offset,
+							last_byte - cur_offset,
+							1 << inode->i_blkbits,
+							offset + len,
+							&alloc_hint);
+
+			if (ret < 0) {
+				free_extent_map(em);
+				break;
+			}
+		} else if (actual_end > inode->i_size &&
+			   !(mode & FALLOC_FL_KEEP_SIZE)) {
+			/*
+			 * We didn't need to allocate any more space, but we
+			 * still extended the size of the file so we need to
+			 * update i_size.
+			 */
+			inode->i_ctime = CURRENT_TIME;
+			i_size_write(inode, actual_end);
+			btrfs_ordered_update_i_size(inode, actual_end, NULL);
+		}
+		free_extent_map(em);
+
+		cur_offset = last_byte;
+		if (cur_offset >= alloc_end) {
+			ret = 0;
+			break;
+		}
+	}
+	unlock_extent_cached(&BTRFS_I(inode)->io_tree, alloc_start, locked_end,
+			     &cached_state, GFP_NOFS);
+out:
+	mutex_unlock(&inode->i_mutex);
+	if (root->fs_info->quota_enabled)
+		btrfs_qgroup_free(root, alloc_end - alloc_start);
+out_reserve_fail:
+	/* Let go of our reservation. */
+	btrfs_free_reserved_data_space(inode, alloc_end - alloc_start);
+	return ret;
+}
+
+static int find_desired_extent(struct inode *inode, loff_t *offset, int whence)
+{
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct extent_map *em = NULL;
+	struct extent_state *cached_state = NULL;
+	u64 lockstart = *offset;
+	u64 lockend = i_size_read(inode);
+	u64 start = *offset;
+	u64 len = i_size_read(inode);
+	int ret = 0;
+
+	lockend = max_t(u64, root->sectorsize, lockend);
+	if (lockend <= lockstart)
+		lockend = lockstart + root->sectorsize;
+
+	lockend--;
+	len = lockend - lockstart + 1;
+
+	len = max_t(u64, len, root->sectorsize);
+	if (inode->i_size == 0)
+		return -ENXIO;
+
+	lock_extent_bits(&BTRFS_I(inode)->io_tree, lockstart, lockend, 0,
+			 &cached_state);
+
+	while (start < inode->i_size) {
+		em = btrfs_get_extent_fiemap(inode, NULL, 0, start, len, 0);
+		if (IS_ERR(em)) {
+			ret = PTR_ERR(em);
+			em = NULL;
+			break;
+		}
+
+		if (whence == SEEK_HOLE &&
+		    (em->block_start == EXTENT_MAP_HOLE ||
+		     test_bit(EXTENT_FLAG_PREALLOC, &em->flags)))
+			break;
+		else if (whence == SEEK_DATA &&
+			   (em->block_start != EXTENT_MAP_HOLE &&
+			    !test_bit(EXTENT_FLAG_PREALLOC, &em->flags)))
+			break;
+
+		start = em->start + em->len;
+		free_extent_map(em);
+		em = NULL;
+		cond_resched();
+	}
+	free_extent_map(em);
+	if (!ret) {
+		if (whence == SEEK_DATA && start >= inode->i_size)
+			ret = -ENXIO;
+		else
+			*offset = min_t(loff_t, start, inode->i_size);
+	}
+	unlock_extent_cached(&BTRFS_I(inode)->io_tree, lockstart, lockend,
+			     &cached_state, GFP_NOFS);
+	return ret;
+}
+
+static loff_t btrfs_file_llseek(struct file *file, loff_t offset, int whence)
+{
+	struct inode *inode = file->f_mapping->host;
+	int ret;
+
+	mutex_lock(&inode->i_mutex);
+	switch (whence) {
+	case SEEK_END:
+	case SEEK_CUR:
+		offset = generic_file_llseek(file, offset, whence);
+		goto out;
+	case SEEK_DATA:
+	case SEEK_HOLE:
+		if (offset >= i_size_read(inode)) {
+			mutex_unlock(&inode->i_mutex);
+			return -ENXIO;
+		}
+
+		ret = find_desired_extent(inode, &offset, whence);
+		if (ret) {
+			mutex_unlock(&inode->i_mutex);
+			return ret;
+		}
+	}
+
+	offset = vfs_setpos(file, offset, inode->i_sb->s_maxbytes);
+out:
+	mutex_unlock(&inode->i_mutex);
+	return offset;
+}
+
 const struct file_operations btrfs_file_operations = {
-	.llseek		= generic_file_llseek,
-	.read		= do_sync_read,
-	.aio_read       = generic_file_aio_read,
+	.llseek		= btrfs_file_llseek,
+	.read		= new_sync_read,
+	.write		= new_sync_write,
+	.read_iter      = generic_file_read_iter,
 	.splice_read	= generic_file_splice_read,
-	.write		= btrfs_file_write,
+	.write_iter	= btrfs_file_write_iter,
 	.mmap		= btrfs_file_mmap,
 	.open		= generic_file_open,
 	.release	= btrfs_release_file,
 	.fsync		= btrfs_sync_file,
+	.fallocate	= btrfs_fallocate,
 	.unlocked_ioctl	= btrfs_ioctl,
 #ifdef CONFIG_COMPAT
 	.compat_ioctl	= btrfs_ioctl,
 #endif
 };
+
+void btrfs_auto_defrag_exit(void)
+{
+	if (btrfs_inode_defrag_cachep)
+		kmem_cache_destroy(btrfs_inode_defrag_cachep);
+}
+
+int btrfs_auto_defrag_init(void)
+{
+	btrfs_inode_defrag_cachep = kmem_cache_create("btrfs_inode_defrag",
+					sizeof(struct inode_defrag), 0,
+					SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD,
+					NULL);
+	if (!btrfs_inode_defrag_cachep)
+		return -ENOMEM;
+
+	return 0;
+}
diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c
index f488fac04d9..2b0a627cb5f 100644
--- a/fs/btrfs/free-space-cache.c
+++ b/fs/btrfs/free-space-cache.c
@@ -20,37 +20,1259 @@
 #include <linux/sched.h>
 #include <linux/slab.h>
 #include <linux/math64.h>
+#include <linux/ratelimit.h>
 #include "ctree.h"
 #include "free-space-cache.h"
 #include "transaction.h"
+#include "disk-io.h"
+#include "extent_io.h"
+#include "inode-map.h"
 
 #define BITS_PER_BITMAP		(PAGE_CACHE_SIZE * 8)
 #define MAX_CACHE_BYTES_PER_GIG	(32 * 1024)
 
-static inline unsigned long offset_to_bit(u64 bitmap_start, u64 sectorsize,
+static int link_free_space(struct btrfs_free_space_ctl *ctl,
+			   struct btrfs_free_space *info);
+static void unlink_free_space(struct btrfs_free_space_ctl *ctl,
+			      struct btrfs_free_space *info);
+
+static struct inode *__lookup_free_space_inode(struct btrfs_root *root,
+					       struct btrfs_path *path,
+					       u64 offset)
+{
+	struct btrfs_key key;
+	struct btrfs_key location;
+	struct btrfs_disk_key disk_key;
+	struct btrfs_free_space_header *header;
+	struct extent_buffer *leaf;
+	struct inode *inode = NULL;
+	int ret;
+
+	key.objectid = BTRFS_FREE_SPACE_OBJECTID;
+	key.offset = offset;
+	key.type = 0;
+
+	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+	if (ret < 0)
+		return ERR_PTR(ret);
+	if (ret > 0) {
+		btrfs_release_path(path);
+		return ERR_PTR(-ENOENT);
+	}
+
+	leaf = path->nodes[0];
+	header = btrfs_item_ptr(leaf, path->slots[0],
+				struct btrfs_free_space_header);
+	btrfs_free_space_key(leaf, header, &disk_key);
+	btrfs_disk_key_to_cpu(&location, &disk_key);
+	btrfs_release_path(path);
+
+	inode = btrfs_iget(root->fs_info->sb, &location, root, NULL);
+	if (!inode)
+		return ERR_PTR(-ENOENT);
+	if (IS_ERR(inode))
+		return inode;
+	if (is_bad_inode(inode)) {
+		iput(inode);
+		return ERR_PTR(-ENOENT);
+	}
+
+	mapping_set_gfp_mask(inode->i_mapping,
+			mapping_gfp_mask(inode->i_mapping) & ~__GFP_FS);
+
+	return inode;
+}
+
+struct inode *lookup_free_space_inode(struct btrfs_root *root,
+				      struct btrfs_block_group_cache
+				      *block_group, struct btrfs_path *path)
+{
+	struct inode *inode = NULL;
+	u32 flags = BTRFS_INODE_NODATASUM | BTRFS_INODE_NODATACOW;
+
+	spin_lock(&block_group->lock);
+	if (block_group->inode)
+		inode = igrab(block_group->inode);
+	spin_unlock(&block_group->lock);
+	if (inode)
+		return inode;
+
+	inode = __lookup_free_space_inode(root, path,
+					  block_group->key.objectid);
+	if (IS_ERR(inode))
+		return inode;
+
+	spin_lock(&block_group->lock);
+	if (!((BTRFS_I(inode)->flags & flags) == flags)) {
+		btrfs_info(root->fs_info,
+			"Old style space inode found, converting.");
+		BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM |
+			BTRFS_INODE_NODATACOW;
+		block_group->disk_cache_state = BTRFS_DC_CLEAR;
+	}
+
+	if (!block_group->iref) {
+		block_group->inode = igrab(inode);
+		block_group->iref = 1;
+	}
+	spin_unlock(&block_group->lock);
+
+	return inode;
+}
+
+static int __create_free_space_inode(struct btrfs_root *root,
+				     struct btrfs_trans_handle *trans,
+				     struct btrfs_path *path,
+				     u64 ino, u64 offset)
+{
+	struct btrfs_key key;
+	struct btrfs_disk_key disk_key;
+	struct btrfs_free_space_header *header;
+	struct btrfs_inode_item *inode_item;
+	struct extent_buffer *leaf;
+	u64 flags = BTRFS_INODE_NOCOMPRESS | BTRFS_INODE_PREALLOC;
+	int ret;
+
+	ret = btrfs_insert_empty_inode(trans, root, path, ino);
+	if (ret)
+		return ret;
+
+	/* We inline crc's for the free disk space cache */
+	if (ino != BTRFS_FREE_INO_OBJECTID)
+		flags |= BTRFS_INODE_NODATASUM | BTRFS_INODE_NODATACOW;
+
+	leaf = path->nodes[0];
+	inode_item = btrfs_item_ptr(leaf, path->slots[0],
+				    struct btrfs_inode_item);
+	btrfs_item_key(leaf, &disk_key, path->slots[0]);
+	memset_extent_buffer(leaf, 0, (unsigned long)inode_item,
+			     sizeof(*inode_item));
+	btrfs_set_inode_generation(leaf, inode_item, trans->transid);
+	btrfs_set_inode_size(leaf, inode_item, 0);
+	btrfs_set_inode_nbytes(leaf, inode_item, 0);
+	btrfs_set_inode_uid(leaf, inode_item, 0);
+	btrfs_set_inode_gid(leaf, inode_item, 0);
+	btrfs_set_inode_mode(leaf, inode_item, S_IFREG | 0600);
+	btrfs_set_inode_flags(leaf, inode_item, flags);
+	btrfs_set_inode_nlink(leaf, inode_item, 1);
+	btrfs_set_inode_transid(leaf, inode_item, trans->transid);
+	btrfs_set_inode_block_group(leaf, inode_item, offset);
+	btrfs_mark_buffer_dirty(leaf);
+	btrfs_release_path(path);
+
+	key.objectid = BTRFS_FREE_SPACE_OBJECTID;
+	key.offset = offset;
+	key.type = 0;
+
+	ret = btrfs_insert_empty_item(trans, root, path, &key,
+				      sizeof(struct btrfs_free_space_header));
+	if (ret < 0) {
+		btrfs_release_path(path);
+		return ret;
+	}
+	leaf = path->nodes[0];
+	header = btrfs_item_ptr(leaf, path->slots[0],
+				struct btrfs_free_space_header);
+	memset_extent_buffer(leaf, 0, (unsigned long)header, sizeof(*header));
+	btrfs_set_free_space_key(leaf, header, &disk_key);
+	btrfs_mark_buffer_dirty(leaf);
+	btrfs_release_path(path);
+
+	return 0;
+}
+
+int create_free_space_inode(struct btrfs_root *root,
+			    struct btrfs_trans_handle *trans,
+			    struct btrfs_block_group_cache *block_group,
+			    struct btrfs_path *path)
+{
+	int ret;
+	u64 ino;
+
+	ret = btrfs_find_free_objectid(root, &ino);
+	if (ret < 0)
+		return ret;
+
+	return __create_free_space_inode(root, trans, path, ino,
+					 block_group->key.objectid);
+}
+
+int btrfs_check_trunc_cache_free_space(struct btrfs_root *root,
+				       struct btrfs_block_rsv *rsv)
+{
+	u64 needed_bytes;
+	int ret;
+
+	/* 1 for slack space, 1 for updating the inode */
+	needed_bytes = btrfs_calc_trunc_metadata_size(root, 1) +
+		btrfs_calc_trans_metadata_size(root, 1);
+
+	spin_lock(&rsv->lock);
+	if (rsv->reserved < needed_bytes)
+		ret = -ENOSPC;
+	else
+		ret = 0;
+	spin_unlock(&rsv->lock);
+	return ret;
+}
+
+int btrfs_truncate_free_space_cache(struct btrfs_root *root,
+				    struct btrfs_trans_handle *trans,
+				    struct inode *inode)
+{
+	int ret = 0;
+
+	btrfs_i_size_write(inode, 0);
+	truncate_pagecache(inode, 0);
+
+	/*
+	 * We don't need an orphan item because truncating the free space cache
+	 * will never be split across transactions.
+	 */
+	ret = btrfs_truncate_inode_items(trans, root, inode,
+					 0, BTRFS_EXTENT_DATA_KEY);
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
+		return ret;
+	}
+
+	ret = btrfs_update_inode(trans, root, inode);
+	if (ret)
+		btrfs_abort_transaction(trans, root, ret);
+
+	return ret;
+}
+
+static int readahead_cache(struct inode *inode)
+{
+	struct file_ra_state *ra;
+	unsigned long last_index;
+
+	ra = kzalloc(sizeof(*ra), GFP_NOFS);
+	if (!ra)
+		return -ENOMEM;
+
+	file_ra_state_init(ra, inode->i_mapping);
+	last_index = (i_size_read(inode) - 1) >> PAGE_CACHE_SHIFT;
+
+	page_cache_sync_readahead(inode->i_mapping, ra, NULL, 0, last_index);
+
+	kfree(ra);
+
+	return 0;
+}
+
+struct io_ctl {
+	void *cur, *orig;
+	struct page *page;
+	struct page **pages;
+	struct btrfs_root *root;
+	unsigned long size;
+	int index;
+	int num_pages;
+	unsigned check_crcs:1;
+};
+
+static int io_ctl_init(struct io_ctl *io_ctl, struct inode *inode,
+		       struct btrfs_root *root, int write)
+{
+	int num_pages;
+	int check_crcs = 0;
+
+	num_pages = (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
+		    PAGE_CACHE_SHIFT;
+
+	if (btrfs_ino(inode) != BTRFS_FREE_INO_OBJECTID)
+		check_crcs = 1;
+
+	/* Make sure we can fit our crcs into the first page */
+	if (write && check_crcs &&
+	    (num_pages * sizeof(u32)) >= PAGE_CACHE_SIZE)
+		return -ENOSPC;
+
+	memset(io_ctl, 0, sizeof(struct io_ctl));
+
+	io_ctl->pages = kzalloc(sizeof(struct page *) * num_pages, GFP_NOFS);
+	if (!io_ctl->pages)
+		return -ENOMEM;
+
+	io_ctl->num_pages = num_pages;
+	io_ctl->root = root;
+	io_ctl->check_crcs = check_crcs;
+
+	return 0;
+}
+
+static void io_ctl_free(struct io_ctl *io_ctl)
+{
+	kfree(io_ctl->pages);
+}
+
+static void io_ctl_unmap_page(struct io_ctl *io_ctl)
+{
+	if (io_ctl->cur) {
+		kunmap(io_ctl->page);
+		io_ctl->cur = NULL;
+		io_ctl->orig = NULL;
+	}
+}
+
+static void io_ctl_map_page(struct io_ctl *io_ctl, int clear)
+{
+	ASSERT(io_ctl->index < io_ctl->num_pages);
+	io_ctl->page = io_ctl->pages[io_ctl->index++];
+	io_ctl->cur = kmap(io_ctl->page);
+	io_ctl->orig = io_ctl->cur;
+	io_ctl->size = PAGE_CACHE_SIZE;
+	if (clear)
+		memset(io_ctl->cur, 0, PAGE_CACHE_SIZE);
+}
+
+static void io_ctl_drop_pages(struct io_ctl *io_ctl)
+{
+	int i;
+
+	io_ctl_unmap_page(io_ctl);
+
+	for (i = 0; i < io_ctl->num_pages; i++) {
+		if (io_ctl->pages[i]) {
+			ClearPageChecked(io_ctl->pages[i]);
+			unlock_page(io_ctl->pages[i]);
+			page_cache_release(io_ctl->pages[i]);
+		}
+	}
+}
+
+static int io_ctl_prepare_pages(struct io_ctl *io_ctl, struct inode *inode,
+				int uptodate)
+{
+	struct page *page;
+	gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
+	int i;
+
+	for (i = 0; i < io_ctl->num_pages; i++) {
+		page = find_or_create_page(inode->i_mapping, i, mask);
+		if (!page) {
+			io_ctl_drop_pages(io_ctl);
+			return -ENOMEM;
+		}
+		io_ctl->pages[i] = page;
+		if (uptodate && !PageUptodate(page)) {
+			btrfs_readpage(NULL, page);
+			lock_page(page);
+			if (!PageUptodate(page)) {
+				btrfs_err(BTRFS_I(inode)->root->fs_info,
+					   "error reading free space cache");
+				io_ctl_drop_pages(io_ctl);
+				return -EIO;
+			}
+		}
+	}
+
+	for (i = 0; i < io_ctl->num_pages; i++) {
+		clear_page_dirty_for_io(io_ctl->pages[i]);
+		set_page_extent_mapped(io_ctl->pages[i]);
+	}
+
+	return 0;
+}
+
+static void io_ctl_set_generation(struct io_ctl *io_ctl, u64 generation)
+{
+	__le64 *val;
+
+	io_ctl_map_page(io_ctl, 1);
+
+	/*
+	 * Skip the csum areas.  If we don't check crcs then we just have a
+	 * 64bit chunk at the front of the first page.
+	 */
+	if (io_ctl->check_crcs) {
+		io_ctl->cur += (sizeof(u32) * io_ctl->num_pages);
+		io_ctl->size -= sizeof(u64) + (sizeof(u32) * io_ctl->num_pages);
+	} else {
+		io_ctl->cur += sizeof(u64);
+		io_ctl->size -= sizeof(u64) * 2;
+	}
+
+	val = io_ctl->cur;
+	*val = cpu_to_le64(generation);
+	io_ctl->cur += sizeof(u64);
+}
+
+static int io_ctl_check_generation(struct io_ctl *io_ctl, u64 generation)
+{
+	__le64 *gen;
+
+	/*
+	 * Skip the crc area.  If we don't check crcs then we just have a 64bit
+	 * chunk at the front of the first page.
+	 */
+	if (io_ctl->check_crcs) {
+		io_ctl->cur += sizeof(u32) * io_ctl->num_pages;
+		io_ctl->size -= sizeof(u64) +
+			(sizeof(u32) * io_ctl->num_pages);
+	} else {
+		io_ctl->cur += sizeof(u64);
+		io_ctl->size -= sizeof(u64) * 2;
+	}
+
+	gen = io_ctl->cur;
+	if (le64_to_cpu(*gen) != generation) {
+		printk_ratelimited(KERN_ERR "BTRFS: space cache generation "
+				   "(%Lu) does not match inode (%Lu)\n", *gen,
+				   generation);
+		io_ctl_unmap_page(io_ctl);
+		return -EIO;
+	}
+	io_ctl->cur += sizeof(u64);
+	return 0;
+}
+
+static void io_ctl_set_crc(struct io_ctl *io_ctl, int index)
+{
+	u32 *tmp;
+	u32 crc = ~(u32)0;
+	unsigned offset = 0;
+
+	if (!io_ctl->check_crcs) {
+		io_ctl_unmap_page(io_ctl);
+		return;
+	}
+
+	if (index == 0)
+		offset = sizeof(u32) * io_ctl->num_pages;
+
+	crc = btrfs_csum_data(io_ctl->orig + offset, crc,
+			      PAGE_CACHE_SIZE - offset);
+	btrfs_csum_final(crc, (char *)&crc);
+	io_ctl_unmap_page(io_ctl);
+	tmp = kmap(io_ctl->pages[0]);
+	tmp += index;
+	*tmp = crc;
+	kunmap(io_ctl->pages[0]);
+}
+
+static int io_ctl_check_crc(struct io_ctl *io_ctl, int index)
+{
+	u32 *tmp, val;
+	u32 crc = ~(u32)0;
+	unsigned offset = 0;
+
+	if (!io_ctl->check_crcs) {
+		io_ctl_map_page(io_ctl, 0);
+		return 0;
+	}
+
+	if (index == 0)
+		offset = sizeof(u32) * io_ctl->num_pages;
+
+	tmp = kmap(io_ctl->pages[0]);
+	tmp += index;
+	val = *tmp;
+	kunmap(io_ctl->pages[0]);
+
+	io_ctl_map_page(io_ctl, 0);
+	crc = btrfs_csum_data(io_ctl->orig + offset, crc,
+			      PAGE_CACHE_SIZE - offset);
+	btrfs_csum_final(crc, (char *)&crc);
+	if (val != crc) {
+		printk_ratelimited(KERN_ERR "BTRFS: csum mismatch on free "
+				   "space cache\n");
+		io_ctl_unmap_page(io_ctl);
+		return -EIO;
+	}
+
+	return 0;
+}
+
+static int io_ctl_add_entry(struct io_ctl *io_ctl, u64 offset, u64 bytes,
+			    void *bitmap)
+{
+	struct btrfs_free_space_entry *entry;
+
+	if (!io_ctl->cur)
+		return -ENOSPC;
+
+	entry = io_ctl->cur;
+	entry->offset = cpu_to_le64(offset);
+	entry->bytes = cpu_to_le64(bytes);
+	entry->type = (bitmap) ? BTRFS_FREE_SPACE_BITMAP :
+		BTRFS_FREE_SPACE_EXTENT;
+	io_ctl->cur += sizeof(struct btrfs_free_space_entry);
+	io_ctl->size -= sizeof(struct btrfs_free_space_entry);
+
+	if (io_ctl->size >= sizeof(struct btrfs_free_space_entry))
+		return 0;
+
+	io_ctl_set_crc(io_ctl, io_ctl->index - 1);
+
+	/* No more pages to map */
+	if (io_ctl->index >= io_ctl->num_pages)
+		return 0;
+
+	/* map the next page */
+	io_ctl_map_page(io_ctl, 1);
+	return 0;
+}
+
+static int io_ctl_add_bitmap(struct io_ctl *io_ctl, void *bitmap)
+{
+	if (!io_ctl->cur)
+		return -ENOSPC;
+
+	/*
+	 * If we aren't at the start of the current page, unmap this one and
+	 * map the next one if there is any left.
+	 */
+	if (io_ctl->cur != io_ctl->orig) {
+		io_ctl_set_crc(io_ctl, io_ctl->index - 1);
+		if (io_ctl->index >= io_ctl->num_pages)
+			return -ENOSPC;
+		io_ctl_map_page(io_ctl, 0);
+	}
+
+	memcpy(io_ctl->cur, bitmap, PAGE_CACHE_SIZE);
+	io_ctl_set_crc(io_ctl, io_ctl->index - 1);
+	if (io_ctl->index < io_ctl->num_pages)
+		io_ctl_map_page(io_ctl, 0);
+	return 0;
+}
+
+static void io_ctl_zero_remaining_pages(struct io_ctl *io_ctl)
+{
+	/*
+	 * If we're not on the boundary we know we've modified the page and we
+	 * need to crc the page.
+	 */
+	if (io_ctl->cur != io_ctl->orig)
+		io_ctl_set_crc(io_ctl, io_ctl->index - 1);
+	else
+		io_ctl_unmap_page(io_ctl);
+
+	while (io_ctl->index < io_ctl->num_pages) {
+		io_ctl_map_page(io_ctl, 1);
+		io_ctl_set_crc(io_ctl, io_ctl->index - 1);
+	}
+}
+
+static int io_ctl_read_entry(struct io_ctl *io_ctl,
+			    struct btrfs_free_space *entry, u8 *type)
+{
+	struct btrfs_free_space_entry *e;
+	int ret;
+
+	if (!io_ctl->cur) {
+		ret = io_ctl_check_crc(io_ctl, io_ctl->index);
+		if (ret)
+			return ret;
+	}
+
+	e = io_ctl->cur;
+	entry->offset = le64_to_cpu(e->offset);
+	entry->bytes = le64_to_cpu(e->bytes);
+	*type = e->type;
+	io_ctl->cur += sizeof(struct btrfs_free_space_entry);
+	io_ctl->size -= sizeof(struct btrfs_free_space_entry);
+
+	if (io_ctl->size >= sizeof(struct btrfs_free_space_entry))
+		return 0;
+
+	io_ctl_unmap_page(io_ctl);
+
+	return 0;
+}
+
+static int io_ctl_read_bitmap(struct io_ctl *io_ctl,
+			      struct btrfs_free_space *entry)
+{
+	int ret;
+
+	ret = io_ctl_check_crc(io_ctl, io_ctl->index);
+	if (ret)
+		return ret;
+
+	memcpy(entry->bitmap, io_ctl->cur, PAGE_CACHE_SIZE);
+	io_ctl_unmap_page(io_ctl);
+
+	return 0;
+}
+
+/*
+ * Since we attach pinned extents after the fact we can have contiguous sections
+ * of free space that are split up in entries.  This poses a problem with the
+ * tree logging stuff since it could have allocated across what appears to be 2
+ * entries since we would have merged the entries when adding the pinned extents
+ * back to the free space cache.  So run through the space cache that we just
+ * loaded and merge contiguous entries.  This will make the log replay stuff not
+ * blow up and it will make for nicer allocator behavior.
+ */
+static void merge_space_tree(struct btrfs_free_space_ctl *ctl)
+{
+	struct btrfs_free_space *e, *prev = NULL;
+	struct rb_node *n;
+
+again:
+	spin_lock(&ctl->tree_lock);
+	for (n = rb_first(&ctl->free_space_offset); n; n = rb_next(n)) {
+		e = rb_entry(n, struct btrfs_free_space, offset_index);
+		if (!prev)
+			goto next;
+		if (e->bitmap || prev->bitmap)
+			goto next;
+		if (prev->offset + prev->bytes == e->offset) {
+			unlink_free_space(ctl, prev);
+			unlink_free_space(ctl, e);
+			prev->bytes += e->bytes;
+			kmem_cache_free(btrfs_free_space_cachep, e);
+			link_free_space(ctl, prev);
+			prev = NULL;
+			spin_unlock(&ctl->tree_lock);
+			goto again;
+		}
+next:
+		prev = e;
+	}
+	spin_unlock(&ctl->tree_lock);
+}
+
+static int __load_free_space_cache(struct btrfs_root *root, struct inode *inode,
+				   struct btrfs_free_space_ctl *ctl,
+				   struct btrfs_path *path, u64 offset)
+{
+	struct btrfs_free_space_header *header;
+	struct extent_buffer *leaf;
+	struct io_ctl io_ctl;
+	struct btrfs_key key;
+	struct btrfs_free_space *e, *n;
+	struct list_head bitmaps;
+	u64 num_entries;
+	u64 num_bitmaps;
+	u64 generation;
+	u8 type;
+	int ret = 0;
+
+	INIT_LIST_HEAD(&bitmaps);
+
+	/* Nothing in the space cache, goodbye */
+	if (!i_size_read(inode))
+		return 0;
+
+	key.objectid = BTRFS_FREE_SPACE_OBJECTID;
+	key.offset = offset;
+	key.type = 0;
+
+	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+	if (ret < 0)
+		return 0;
+	else if (ret > 0) {
+		btrfs_release_path(path);
+		return 0;
+	}
+
+	ret = -1;
+
+	leaf = path->nodes[0];
+	header = btrfs_item_ptr(leaf, path->slots[0],
+				struct btrfs_free_space_header);
+	num_entries = btrfs_free_space_entries(leaf, header);
+	num_bitmaps = btrfs_free_space_bitmaps(leaf, header);
+	generation = btrfs_free_space_generation(leaf, header);
+	btrfs_release_path(path);
+
+	if (!BTRFS_I(inode)->generation) {
+		btrfs_info(root->fs_info,
+			   "The free space cache file (%llu) is invalid. skip it\n",
+			   offset);
+		return 0;
+	}
+
+	if (BTRFS_I(inode)->generation != generation) {
+		btrfs_err(root->fs_info,
+			"free space inode generation (%llu) "
+			"did not match free space cache generation (%llu)",
+			BTRFS_I(inode)->generation, generation);
+		return 0;
+	}
+
+	if (!num_entries)
+		return 0;
+
+	ret = io_ctl_init(&io_ctl, inode, root, 0);
+	if (ret)
+		return ret;
+
+	ret = readahead_cache(inode);
+	if (ret)
+		goto out;
+
+	ret = io_ctl_prepare_pages(&io_ctl, inode, 1);
+	if (ret)
+		goto out;
+
+	ret = io_ctl_check_crc(&io_ctl, 0);
+	if (ret)
+		goto free_cache;
+
+	ret = io_ctl_check_generation(&io_ctl, generation);
+	if (ret)
+		goto free_cache;
+
+	while (num_entries) {
+		e = kmem_cache_zalloc(btrfs_free_space_cachep,
+				      GFP_NOFS);
+		if (!e)
+			goto free_cache;
+
+		ret = io_ctl_read_entry(&io_ctl, e, &type);
+		if (ret) {
+			kmem_cache_free(btrfs_free_space_cachep, e);
+			goto free_cache;
+		}
+
+		if (!e->bytes) {
+			kmem_cache_free(btrfs_free_space_cachep, e);
+			goto free_cache;
+		}
+
+		if (type == BTRFS_FREE_SPACE_EXTENT) {
+			spin_lock(&ctl->tree_lock);
+			ret = link_free_space(ctl, e);
+			spin_unlock(&ctl->tree_lock);
+			if (ret) {
+				btrfs_err(root->fs_info,
+					"Duplicate entries in free space cache, dumping");
+				kmem_cache_free(btrfs_free_space_cachep, e);
+				goto free_cache;
+			}
+		} else {
+			ASSERT(num_bitmaps);
+			num_bitmaps--;
+			e->bitmap = kzalloc(PAGE_CACHE_SIZE, GFP_NOFS);
+			if (!e->bitmap) {
+				kmem_cache_free(
+					btrfs_free_space_cachep, e);
+				goto free_cache;
+			}
+			spin_lock(&ctl->tree_lock);
+			ret = link_free_space(ctl, e);
+			ctl->total_bitmaps++;
+			ctl->op->recalc_thresholds(ctl);
+			spin_unlock(&ctl->tree_lock);
+			if (ret) {
+				btrfs_err(root->fs_info,
+					"Duplicate entries in free space cache, dumping");
+				kmem_cache_free(btrfs_free_space_cachep, e);
+				goto free_cache;
+			}
+			list_add_tail(&e->list, &bitmaps);
+		}
+
+		num_entries--;
+	}
+
+	io_ctl_unmap_page(&io_ctl);
+
+	/*
+	 * We add the bitmaps at the end of the entries in order that
+	 * the bitmap entries are added to the cache.
+	 */
+	list_for_each_entry_safe(e, n, &bitmaps, list) {
+		list_del_init(&e->list);
+		ret = io_ctl_read_bitmap(&io_ctl, e);
+		if (ret)
+			goto free_cache;
+	}
+
+	io_ctl_drop_pages(&io_ctl);
+	merge_space_tree(ctl);
+	ret = 1;
+out:
+	io_ctl_free(&io_ctl);
+	return ret;
+free_cache:
+	io_ctl_drop_pages(&io_ctl);
+	__btrfs_remove_free_space_cache(ctl);
+	goto out;
+}
+
+int load_free_space_cache(struct btrfs_fs_info *fs_info,
+			  struct btrfs_block_group_cache *block_group)
+{
+	struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
+	struct btrfs_root *root = fs_info->tree_root;
+	struct inode *inode;
+	struct btrfs_path *path;
+	int ret = 0;
+	bool matched;
+	u64 used = btrfs_block_group_used(&block_group->item);
+
+	/*
+	 * If this block group has been marked to be cleared for one reason or
+	 * another then we can't trust the on disk cache, so just return.
+	 */
+	spin_lock(&block_group->lock);
+	if (block_group->disk_cache_state != BTRFS_DC_WRITTEN) {
+		spin_unlock(&block_group->lock);
+		return 0;
+	}
+	spin_unlock(&block_group->lock);
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return 0;
+	path->search_commit_root = 1;
+	path->skip_locking = 1;
+
+	inode = lookup_free_space_inode(root, block_group, path);
+	if (IS_ERR(inode)) {
+		btrfs_free_path(path);
+		return 0;
+	}
+
+	/* We may have converted the inode and made the cache invalid. */
+	spin_lock(&block_group->lock);
+	if (block_group->disk_cache_state != BTRFS_DC_WRITTEN) {
+		spin_unlock(&block_group->lock);
+		btrfs_free_path(path);
+		goto out;
+	}
+	spin_unlock(&block_group->lock);
+
+	ret = __load_free_space_cache(fs_info->tree_root, inode, ctl,
+				      path, block_group->key.objectid);
+	btrfs_free_path(path);
+	if (ret <= 0)
+		goto out;
+
+	spin_lock(&ctl->tree_lock);
+	matched = (ctl->free_space == (block_group->key.offset - used -
+				       block_group->bytes_super));
+	spin_unlock(&ctl->tree_lock);
+
+	if (!matched) {
+		__btrfs_remove_free_space_cache(ctl);
+		btrfs_warn(fs_info, "block group %llu has wrong amount of free space",
+			block_group->key.objectid);
+		ret = -1;
+	}
+out:
+	if (ret < 0) {
+		/* This cache is bogus, make sure it gets cleared */
+		spin_lock(&block_group->lock);
+		block_group->disk_cache_state = BTRFS_DC_CLEAR;
+		spin_unlock(&block_group->lock);
+		ret = 0;
+
+		btrfs_warn(fs_info, "failed to load free space cache for block group %llu, rebuild it now",
+			block_group->key.objectid);
+	}
+
+	iput(inode);
+	return ret;
+}
+
+static noinline_for_stack
+int write_cache_extent_entries(struct io_ctl *io_ctl,
+			      struct btrfs_free_space_ctl *ctl,
+			      struct btrfs_block_group_cache *block_group,
+			      int *entries, int *bitmaps,
+			      struct list_head *bitmap_list)
+{
+	int ret;
+	struct btrfs_free_cluster *cluster = NULL;
+	struct rb_node *node = rb_first(&ctl->free_space_offset);
+
+	/* Get the cluster for this block_group if it exists */
+	if (block_group && !list_empty(&block_group->cluster_list)) {
+		cluster = list_entry(block_group->cluster_list.next,
+				     struct btrfs_free_cluster,
+				     block_group_list);
+	}
+
+	if (!node && cluster) {
+		node = rb_first(&cluster->root);
+		cluster = NULL;
+	}
+
+	/* Write out the extent entries */
+	while (node) {
+		struct btrfs_free_space *e;
+
+		e = rb_entry(node, struct btrfs_free_space, offset_index);
+		*entries += 1;
+
+		ret = io_ctl_add_entry(io_ctl, e->offset, e->bytes,
+				       e->bitmap);
+		if (ret)
+			goto fail;
+
+		if (e->bitmap) {
+			list_add_tail(&e->list, bitmap_list);
+			*bitmaps += 1;
+		}
+		node = rb_next(node);
+		if (!node && cluster) {
+			node = rb_first(&cluster->root);
+			cluster = NULL;
+		}
+	}
+	return 0;
+fail:
+	return -ENOSPC;
+}
+
+static noinline_for_stack int
+update_cache_item(struct btrfs_trans_handle *trans,
+		  struct btrfs_root *root,
+		  struct inode *inode,
+		  struct btrfs_path *path, u64 offset,
+		  int entries, int bitmaps)
+{
+	struct btrfs_key key;
+	struct btrfs_free_space_header *header;
+	struct extent_buffer *leaf;
+	int ret;
+
+	key.objectid = BTRFS_FREE_SPACE_OBJECTID;
+	key.offset = offset;
+	key.type = 0;
+
+	ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
+	if (ret < 0) {
+		clear_extent_bit(&BTRFS_I(inode)->io_tree, 0, inode->i_size - 1,
+				 EXTENT_DIRTY | EXTENT_DELALLOC, 0, 0, NULL,
+				 GFP_NOFS);
+		goto fail;
+	}
+	leaf = path->nodes[0];
+	if (ret > 0) {
+		struct btrfs_key found_key;
+		ASSERT(path->slots[0]);
+		path->slots[0]--;
+		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
+		if (found_key.objectid != BTRFS_FREE_SPACE_OBJECTID ||
+		    found_key.offset != offset) {
+			clear_extent_bit(&BTRFS_I(inode)->io_tree, 0,
+					 inode->i_size - 1,
+					 EXTENT_DIRTY | EXTENT_DELALLOC, 0, 0,
+					 NULL, GFP_NOFS);
+			btrfs_release_path(path);
+			goto fail;
+		}
+	}
+
+	BTRFS_I(inode)->generation = trans->transid;
+	header = btrfs_item_ptr(leaf, path->slots[0],
+				struct btrfs_free_space_header);
+	btrfs_set_free_space_entries(leaf, header, entries);
+	btrfs_set_free_space_bitmaps(leaf, header, bitmaps);
+	btrfs_set_free_space_generation(leaf, header, trans->transid);
+	btrfs_mark_buffer_dirty(leaf);
+	btrfs_release_path(path);
+
+	return 0;
+
+fail:
+	return -1;
+}
+
+static noinline_for_stack int
+write_pinned_extent_entries(struct btrfs_root *root,
+			    struct btrfs_block_group_cache *block_group,
+			    struct io_ctl *io_ctl,
+			    int *entries)
+{
+	u64 start, extent_start, extent_end, len;
+	struct extent_io_tree *unpin = NULL;
+	int ret;
+
+	if (!block_group)
+		return 0;
+
+	/*
+	 * We want to add any pinned extents to our free space cache
+	 * so we don't leak the space
+	 *
+	 * We shouldn't have switched the pinned extents yet so this is the
+	 * right one
+	 */
+	unpin = root->fs_info->pinned_extents;
+
+	start = block_group->key.objectid;
+
+	while (start < block_group->key.objectid + block_group->key.offset) {
+		ret = find_first_extent_bit(unpin, start,
+					    &extent_start, &extent_end,
+					    EXTENT_DIRTY, NULL);
+		if (ret)
+			return 0;
+
+		/* This pinned extent is out of our range */
+		if (extent_start >= block_group->key.objectid +
+		    block_group->key.offset)
+			return 0;
+
+		extent_start = max(extent_start, start);
+		extent_end = min(block_group->key.objectid +
+				 block_group->key.offset, extent_end + 1);
+		len = extent_end - extent_start;
+
+		*entries += 1;
+		ret = io_ctl_add_entry(io_ctl, extent_start, len, NULL);
+		if (ret)
+			return -ENOSPC;
+
+		start = extent_end;
+	}
+
+	return 0;
+}
+
+static noinline_for_stack int
+write_bitmap_entries(struct io_ctl *io_ctl, struct list_head *bitmap_list)
+{
+	struct list_head *pos, *n;
+	int ret;
+
+	/* Write out the bitmaps */
+	list_for_each_safe(pos, n, bitmap_list) {
+		struct btrfs_free_space *entry =
+			list_entry(pos, struct btrfs_free_space, list);
+
+		ret = io_ctl_add_bitmap(io_ctl, entry->bitmap);
+		if (ret)
+			return -ENOSPC;
+		list_del_init(&entry->list);
+	}
+
+	return 0;
+}
+
+static int flush_dirty_cache(struct inode *inode)
+{
+	int ret;
+
+	ret = btrfs_wait_ordered_range(inode, 0, (u64)-1);
+	if (ret)
+		clear_extent_bit(&BTRFS_I(inode)->io_tree, 0, inode->i_size - 1,
+				 EXTENT_DIRTY | EXTENT_DELALLOC, 0, 0, NULL,
+				 GFP_NOFS);
+
+	return ret;
+}
+
+static void noinline_for_stack
+cleanup_write_cache_enospc(struct inode *inode,
+			   struct io_ctl *io_ctl,
+			   struct extent_state **cached_state,
+			   struct list_head *bitmap_list)
+{
+	struct list_head *pos, *n;
+
+	list_for_each_safe(pos, n, bitmap_list) {
+		struct btrfs_free_space *entry =
+			list_entry(pos, struct btrfs_free_space, list);
+		list_del_init(&entry->list);
+	}
+	io_ctl_drop_pages(io_ctl);
+	unlock_extent_cached(&BTRFS_I(inode)->io_tree, 0,
+			     i_size_read(inode) - 1, cached_state,
+			     GFP_NOFS);
+}
+
+/**
+ * __btrfs_write_out_cache - write out cached info to an inode
+ * @root - the root the inode belongs to
+ * @ctl - the free space cache we are going to write out
+ * @block_group - the block_group for this cache if it belongs to a block_group
+ * @trans - the trans handle
+ * @path - the path to use
+ * @offset - the offset for the key we'll insert
+ *
+ * This function writes out a free space cache struct to disk for quick recovery
+ * on mount.  This will return 0 if it was successfull in writing the cache out,
+ * and -1 if it was not.
+ */
+static int __btrfs_write_out_cache(struct btrfs_root *root, struct inode *inode,
+				   struct btrfs_free_space_ctl *ctl,
+				   struct btrfs_block_group_cache *block_group,
+				   struct btrfs_trans_handle *trans,
+				   struct btrfs_path *path, u64 offset)
+{
+	struct extent_state *cached_state = NULL;
+	struct io_ctl io_ctl;
+	LIST_HEAD(bitmap_list);
+	int entries = 0;
+	int bitmaps = 0;
+	int ret;
+
+	if (!i_size_read(inode))
+		return -1;
+
+	ret = io_ctl_init(&io_ctl, inode, root, 1);
+	if (ret)
+		return -1;
+
+	if (block_group && (block_group->flags & BTRFS_BLOCK_GROUP_DATA)) {
+		down_write(&block_group->data_rwsem);
+		spin_lock(&block_group->lock);
+		if (block_group->delalloc_bytes) {
+			block_group->disk_cache_state = BTRFS_DC_WRITTEN;
+			spin_unlock(&block_group->lock);
+			up_write(&block_group->data_rwsem);
+			BTRFS_I(inode)->generation = 0;
+			ret = 0;
+			goto out;
+		}
+		spin_unlock(&block_group->lock);
+	}
+
+	/* Lock all pages first so we can lock the extent safely. */
+	io_ctl_prepare_pages(&io_ctl, inode, 0);
+
+	lock_extent_bits(&BTRFS_I(inode)->io_tree, 0, i_size_read(inode) - 1,
+			 0, &cached_state);
+
+	io_ctl_set_generation(&io_ctl, trans->transid);
+
+	/* Write out the extent entries in the free space cache */
+	ret = write_cache_extent_entries(&io_ctl, ctl,
+					 block_group, &entries, &bitmaps,
+					 &bitmap_list);
+	if (ret)
+		goto out_nospc;
+
+	/*
+	 * Some spaces that are freed in the current transaction are pinned,
+	 * they will be added into free space cache after the transaction is
+	 * committed, we shouldn't lose them.
+	 */
+	ret = write_pinned_extent_entries(root, block_group, &io_ctl, &entries);
+	if (ret)
+		goto out_nospc;
+
+	/* At last, we write out all the bitmaps. */
+	ret = write_bitmap_entries(&io_ctl, &bitmap_list);
+	if (ret)
+		goto out_nospc;
+
+	/* Zero out the rest of the pages just to make sure */
+	io_ctl_zero_remaining_pages(&io_ctl);
+
+	/* Everything is written out, now we dirty the pages in the file. */
+	ret = btrfs_dirty_pages(root, inode, io_ctl.pages, io_ctl.num_pages,
+				0, i_size_read(inode), &cached_state);
+	if (ret)
+		goto out_nospc;
+
+	if (block_group && (block_group->flags & BTRFS_BLOCK_GROUP_DATA))
+		up_write(&block_group->data_rwsem);
+	/*
+	 * Release the pages and unlock the extent, we will flush
+	 * them out later
+	 */
+	io_ctl_drop_pages(&io_ctl);
+
+	unlock_extent_cached(&BTRFS_I(inode)->io_tree, 0,
+			     i_size_read(inode) - 1, &cached_state, GFP_NOFS);
+
+	/* Flush the dirty pages in the cache file. */
+	ret = flush_dirty_cache(inode);
+	if (ret)
+		goto out;
+
+	/* Update the cache item to tell everyone this cache file is valid. */
+	ret = update_cache_item(trans, root, inode, path, offset,
+				entries, bitmaps);
+out:
+	io_ctl_free(&io_ctl);
+	if (ret) {
+		invalidate_inode_pages2(inode->i_mapping);
+		BTRFS_I(inode)->generation = 0;
+	}
+	btrfs_update_inode(trans, root, inode);
+	return ret;
+
+out_nospc:
+	cleanup_write_cache_enospc(inode, &io_ctl, &cached_state, &bitmap_list);
+
+	if (block_group && (block_group->flags & BTRFS_BLOCK_GROUP_DATA))
+		up_write(&block_group->data_rwsem);
+
+	goto out;
+}
+
+int btrfs_write_out_cache(struct btrfs_root *root,
+			  struct btrfs_trans_handle *trans,
+			  struct btrfs_block_group_cache *block_group,
+			  struct btrfs_path *path)
+{
+	struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
+	struct inode *inode;
+	int ret = 0;
+
+	root = root->fs_info->tree_root;
+
+	spin_lock(&block_group->lock);
+	if (block_group->disk_cache_state < BTRFS_DC_SETUP) {
+		spin_unlock(&block_group->lock);
+		return 0;
+	}
+
+	if (block_group->delalloc_bytes) {
+		block_group->disk_cache_state = BTRFS_DC_WRITTEN;
+		spin_unlock(&block_group->lock);
+		return 0;
+	}
+	spin_unlock(&block_group->lock);
+
+	inode = lookup_free_space_inode(root, block_group, path);
+	if (IS_ERR(inode))
+		return 0;
+
+	ret = __btrfs_write_out_cache(root, inode, ctl, block_group, trans,
+				      path, block_group->key.objectid);
+	if (ret) {
+		spin_lock(&block_group->lock);
+		block_group->disk_cache_state = BTRFS_DC_ERROR;
+		spin_unlock(&block_group->lock);
+		ret = 0;
+#ifdef DEBUG
+		btrfs_err(root->fs_info,
+			"failed to write free space cache for block group %llu",
+			block_group->key.objectid);
+#endif
+	}
+
+	iput(inode);
+	return ret;
+}
+
+static inline unsigned long offset_to_bit(u64 bitmap_start, u32 unit,
 					  u64 offset)
 {
-	BUG_ON(offset < bitmap_start);
+	ASSERT(offset >= bitmap_start);
 	offset -= bitmap_start;
-	return (unsigned long)(div64_u64(offset, sectorsize));
+	return (unsigned long)(div_u64(offset, unit));
 }
 
-static inline unsigned long bytes_to_bits(u64 bytes, u64 sectorsize)
+static inline unsigned long bytes_to_bits(u64 bytes, u32 unit)
 {
-	return (unsigned long)(div64_u64(bytes, sectorsize));
+	return (unsigned long)(div_u64(bytes, unit));
 }
 
-static inline u64 offset_to_bitmap(struct btrfs_block_group_cache *block_group,
+static inline u64 offset_to_bitmap(struct btrfs_free_space_ctl *ctl,
 				   u64 offset)
 {
 	u64 bitmap_start;
 	u64 bytes_per_bitmap;
 
-	bytes_per_bitmap = BITS_PER_BITMAP * block_group->sectorsize;
-	bitmap_start = offset - block_group->key.objectid;
+	bytes_per_bitmap = BITS_PER_BITMAP * ctl->unit;
+	bitmap_start = offset - ctl->start;
 	bitmap_start = div64_u64(bitmap_start, bytes_per_bitmap);
 	bitmap_start *= bytes_per_bitmap;
-	bitmap_start += block_group->key.objectid;
+	bitmap_start += ctl->start;
 
 	return bitmap_start;
 }
@@ -85,10 +1307,16 @@ static int tree_insert_offset(struct rb_root *root, u64 offset,
 			 * logically.
 			 */
 			if (bitmap) {
-				WARN_ON(info->bitmap);
+				if (info->bitmap) {
+					WARN_ON_ONCE(1);
+					return -EEXIST;
+				}
 				p = &(*p)->rb_right;
 			} else {
-				WARN_ON(!info->bitmap);
+				if (!info->bitmap) {
+					WARN_ON_ONCE(1);
+					return -EEXIST;
+				}
 				p = &(*p)->rb_left;
 			}
 		}
@@ -108,10 +1336,10 @@ static int tree_insert_offset(struct rb_root *root, u64 offset,
  * offset.
  */
 static struct btrfs_free_space *
-tree_search_offset(struct btrfs_block_group_cache *block_group,
+tree_search_offset(struct btrfs_free_space_ctl *ctl,
 		   u64 offset, int bitmap_only, int fuzzy)
 {
-	struct rb_node *n = block_group->free_space_offset.rb_node;
+	struct rb_node *n = ctl->free_space_offset.rb_node;
 	struct btrfs_free_space *entry, *prev = NULL;
 
 	/* find entry that is closest to the 'offset' */
@@ -157,18 +1385,13 @@ tree_search_offset(struct btrfs_block_group_cache *block_group,
 			 * if previous extent entry covers the offset,
 			 * we should return it instead of the bitmap entry
 			 */
-			n = &entry->offset_index;
-			while (1) {
-				n = rb_prev(n);
-				if (!n)
-					break;
+			n = rb_prev(&entry->offset_index);
+			if (n) {
 				prev = rb_entry(n, struct btrfs_free_space,
 						offset_index);
-				if (!prev->bitmap) {
-					if (prev->offset + prev->bytes > offset)
-						entry = prev;
-					break;
-				}
+				if (!prev->bitmap &&
+				    prev->offset + prev->bytes > offset)
+					entry = prev;
 			}
 		}
 		return entry;
@@ -184,7 +1407,7 @@ tree_search_offset(struct btrfs_block_group_cache *block_group,
 		if (n) {
 			entry = rb_entry(n, struct btrfs_free_space,
 					offset_index);
-			BUG_ON(entry->offset > offset);
+			ASSERT(entry->offset <= offset);
 		} else {
 			if (fuzzy)
 				return entry;
@@ -194,21 +1417,15 @@ tree_search_offset(struct btrfs_block_group_cache *block_group,
 	}
 
 	if (entry->bitmap) {
-		n = &entry->offset_index;
-		while (1) {
-			n = rb_prev(n);
-			if (!n)
-				break;
+		n = rb_prev(&entry->offset_index);
+		if (n) {
 			prev = rb_entry(n, struct btrfs_free_space,
 					offset_index);
-			if (!prev->bitmap) {
-				if (prev->offset + prev->bytes > offset)
-					return prev;
-				break;
-			}
+			if (!prev->bitmap &&
+			    prev->offset + prev->bytes > offset)
+				return prev;
 		}
-		if (entry->offset + BITS_PER_BITMAP *
-		    block_group->sectorsize > offset)
+		if (entry->offset + BITS_PER_BITMAP * ctl->unit > offset)
 			return entry;
 	} else if (entry->offset + entry->bytes > offset)
 		return entry;
@@ -219,7 +1436,7 @@ tree_search_offset(struct btrfs_block_group_cache *block_group,
 	while (1) {
 		if (entry->bitmap) {
 			if (entry->offset + BITS_PER_BITMAP *
-			    block_group->sectorsize > offset)
+			    ctl->unit > offset)
 				break;
 		} else {
 			if (entry->offset + entry->bytes > offset)
@@ -234,53 +1451,71 @@ tree_search_offset(struct btrfs_block_group_cache *block_group,
 	return entry;
 }
 
-static void unlink_free_space(struct btrfs_block_group_cache *block_group,
+static inline void
+__unlink_free_space(struct btrfs_free_space_ctl *ctl,
+		    struct btrfs_free_space *info)
+{
+	rb_erase(&info->offset_index, &ctl->free_space_offset);
+	ctl->free_extents--;
+}
+
+static void unlink_free_space(struct btrfs_free_space_ctl *ctl,
 			      struct btrfs_free_space *info)
 {
-	rb_erase(&info->offset_index, &block_group->free_space_offset);
-	block_group->free_extents--;
-	block_group->free_space -= info->bytes;
+	__unlink_free_space(ctl, info);
+	ctl->free_space -= info->bytes;
 }
 
-static int link_free_space(struct btrfs_block_group_cache *block_group,
+static int link_free_space(struct btrfs_free_space_ctl *ctl,
 			   struct btrfs_free_space *info)
 {
 	int ret = 0;
 
-	BUG_ON(!info->bitmap && !info->bytes);
-	ret = tree_insert_offset(&block_group->free_space_offset, info->offset,
+	ASSERT(info->bytes || info->bitmap);
+	ret = tree_insert_offset(&ctl->free_space_offset, info->offset,
 				 &info->offset_index, (info->bitmap != NULL));
 	if (ret)
 		return ret;
 
-	block_group->free_space += info->bytes;
-	block_group->free_extents++;
+	ctl->free_space += info->bytes;
+	ctl->free_extents++;
 	return ret;
 }
 
-static void recalculate_thresholds(struct btrfs_block_group_cache *block_group)
+static void recalculate_thresholds(struct btrfs_free_space_ctl *ctl)
 {
+	struct btrfs_block_group_cache *block_group = ctl->private;
 	u64 max_bytes;
 	u64 bitmap_bytes;
 	u64 extent_bytes;
+	u64 size = block_group->key.offset;
+	u64 bytes_per_bg = BITS_PER_BITMAP * ctl->unit;
+	int max_bitmaps = div64_u64(size + bytes_per_bg - 1, bytes_per_bg);
+
+	max_bitmaps = max(max_bitmaps, 1);
+
+	ASSERT(ctl->total_bitmaps <= max_bitmaps);
 
 	/*
 	 * The goal is to keep the total amount of memory used per 1gb of space
 	 * at or below 32k, so we need to adjust how much memory we allow to be
 	 * used by extent based free space tracking
 	 */
-	max_bytes = MAX_CACHE_BYTES_PER_GIG *
-		(div64_u64(block_group->key.offset, 1024 * 1024 * 1024));
+	if (size < 1024 * 1024 * 1024)
+		max_bytes = MAX_CACHE_BYTES_PER_GIG;
+	else
+		max_bytes = MAX_CACHE_BYTES_PER_GIG *
+			div64_u64(size, 1024 * 1024 * 1024);
 
 	/*
 	 * we want to account for 1 more bitmap than what we have so we can make
 	 * sure we don't go over our overall goal of MAX_CACHE_BYTES_PER_GIG as
 	 * we add more bitmaps.
 	 */
-	bitmap_bytes = (block_group->total_bitmaps + 1) * PAGE_CACHE_SIZE;
+	bitmap_bytes = (ctl->total_bitmaps + 1) * PAGE_CACHE_SIZE;
 
 	if (bitmap_bytes >= max_bytes) {
-		block_group->extents_thresh = 0;
+		ctl->extents_thresh = 0;
 		return;
 	}
 
@@ -291,134 +1526,180 @@ static void recalculate_thresholds(struct btrfs_block_group_cache *block_group)
 	extent_bytes = max_bytes - bitmap_bytes;
 	extent_bytes = min_t(u64, extent_bytes, div64_u64(max_bytes, 2));
 
-	block_group->extents_thresh =
+	ctl->extents_thresh =
 		div64_u64(extent_bytes, (sizeof(struct btrfs_free_space)));
 }
 
-static void bitmap_clear_bits(struct btrfs_block_group_cache *block_group,
-			      struct btrfs_free_space *info, u64 offset,
-			      u64 bytes)
+static inline void __bitmap_clear_bits(struct btrfs_free_space_ctl *ctl,
+				       struct btrfs_free_space *info,
+				       u64 offset, u64 bytes)
 {
-	unsigned long start, end;
-	unsigned long i;
+	unsigned long start, count;
 
-	start = offset_to_bit(info->offset, block_group->sectorsize, offset);
-	end = start + bytes_to_bits(bytes, block_group->sectorsize);
-	BUG_ON(end > BITS_PER_BITMAP);
+	start = offset_to_bit(info->offset, ctl->unit, offset);
+	count = bytes_to_bits(bytes, ctl->unit);
+	ASSERT(start + count <= BITS_PER_BITMAP);
 
-	for (i = start; i < end; i++)
-		clear_bit(i, info->bitmap);
+	bitmap_clear(info->bitmap, start, count);
 
 	info->bytes -= bytes;
-	block_group->free_space -= bytes;
 }
 
-static void bitmap_set_bits(struct btrfs_block_group_cache *block_group,
+static void bitmap_clear_bits(struct btrfs_free_space_ctl *ctl,
+			      struct btrfs_free_space *info, u64 offset,
+			      u64 bytes)
+{
+	__bitmap_clear_bits(ctl, info, offset, bytes);
+	ctl->free_space -= bytes;
+}
+
+static void bitmap_set_bits(struct btrfs_free_space_ctl *ctl,
 			    struct btrfs_free_space *info, u64 offset,
 			    u64 bytes)
 {
-	unsigned long start, end;
-	unsigned long i;
+	unsigned long start, count;
 
-	start = offset_to_bit(info->offset, block_group->sectorsize, offset);
-	end = start + bytes_to_bits(bytes, block_group->sectorsize);
-	BUG_ON(end > BITS_PER_BITMAP);
+	start = offset_to_bit(info->offset, ctl->unit, offset);
+	count = bytes_to_bits(bytes, ctl->unit);
+	ASSERT(start + count <= BITS_PER_BITMAP);
 
-	for (i = start; i < end; i++)
-		set_bit(i, info->bitmap);
+	bitmap_set(info->bitmap, start, count);
 
 	info->bytes += bytes;
-	block_group->free_space += bytes;
+	ctl->free_space += bytes;
 }
 
-static int search_bitmap(struct btrfs_block_group_cache *block_group,
+/*
+ * If we can not find suitable extent, we will use bytes to record
+ * the size of the max extent.
+ */
+static int search_bitmap(struct btrfs_free_space_ctl *ctl,
 			 struct btrfs_free_space *bitmap_info, u64 *offset,
 			 u64 *bytes)
 {
 	unsigned long found_bits = 0;
+	unsigned long max_bits = 0;
 	unsigned long bits, i;
 	unsigned long next_zero;
+	unsigned long extent_bits;
 
-	i = offset_to_bit(bitmap_info->offset, block_group->sectorsize,
+	i = offset_to_bit(bitmap_info->offset, ctl->unit,
 			  max_t(u64, *offset, bitmap_info->offset));
-	bits = bytes_to_bits(*bytes, block_group->sectorsize);
+	bits = bytes_to_bits(*bytes, ctl->unit);
 
-	for (i = find_next_bit(bitmap_info->bitmap, BITS_PER_BITMAP, i);
-	     i < BITS_PER_BITMAP;
-	     i = find_next_bit(bitmap_info->bitmap, BITS_PER_BITMAP, i + 1)) {
+	for_each_set_bit_from(i, bitmap_info->bitmap, BITS_PER_BITMAP) {
 		next_zero = find_next_zero_bit(bitmap_info->bitmap,
 					       BITS_PER_BITMAP, i);
-		if ((next_zero - i) >= bits) {
-			found_bits = next_zero - i;
+		extent_bits = next_zero - i;
+		if (extent_bits >= bits) {
+			found_bits = extent_bits;
 			break;
+		} else if (extent_bits > max_bits) {
+			max_bits = extent_bits;
 		}
 		i = next_zero;
 	}
 
 	if (found_bits) {
-		*offset = (u64)(i * block_group->sectorsize) +
-			bitmap_info->offset;
-		*bytes = (u64)(found_bits) * block_group->sectorsize;
+		*offset = (u64)(i * ctl->unit) + bitmap_info->offset;
+		*bytes = (u64)(found_bits) * ctl->unit;
 		return 0;
 	}
 
+	*bytes = (u64)(max_bits) * ctl->unit;
 	return -1;
 }
 
-static struct btrfs_free_space *find_free_space(struct btrfs_block_group_cache
-						*block_group, u64 *offset,
-						u64 *bytes, int debug)
+/* Cache the size of the max extent in bytes */
+static struct btrfs_free_space *
+find_free_space(struct btrfs_free_space_ctl *ctl, u64 *offset, u64 *bytes,
+		unsigned long align, u64 *max_extent_size)
 {
 	struct btrfs_free_space *entry;
 	struct rb_node *node;
+	u64 tmp;
+	u64 align_off;
 	int ret;
 
-	if (!block_group->free_space_offset.rb_node)
-		return NULL;
+	if (!ctl->free_space_offset.rb_node)
+		goto out;
 
-	entry = tree_search_offset(block_group,
-				   offset_to_bitmap(block_group, *offset),
-				   0, 1);
+	entry = tree_search_offset(ctl, offset_to_bitmap(ctl, *offset), 0, 1);
 	if (!entry)
-		return NULL;
+		goto out;
 
 	for (node = &entry->offset_index; node; node = rb_next(node)) {
 		entry = rb_entry(node, struct btrfs_free_space, offset_index);
-		if (entry->bytes < *bytes)
+		if (entry->bytes < *bytes) {
+			if (entry->bytes > *max_extent_size)
+				*max_extent_size = entry->bytes;
+			continue;
+		}
+
+		/* make sure the space returned is big enough
+		 * to match our requested alignment
+		 */
+		if (*bytes >= align) {
+			tmp = entry->offset - ctl->start + align - 1;
+			do_div(tmp, align);
+			tmp = tmp * align + ctl->start;
+			align_off = tmp - entry->offset;
+		} else {
+			align_off = 0;
+			tmp = entry->offset;
+		}
+
+		if (entry->bytes < *bytes + align_off) {
+			if (entry->bytes > *max_extent_size)
+				*max_extent_size = entry->bytes;
 			continue;
+		}
 
 		if (entry->bitmap) {
-			ret = search_bitmap(block_group, entry, offset, bytes);
-			if (!ret)
+			u64 size = *bytes;
+
+			ret = search_bitmap(ctl, entry, &tmp, &size);
+			if (!ret) {
+				*offset = tmp;
+				*bytes = size;
 				return entry;
+			} else if (size > *max_extent_size) {
+				*max_extent_size = size;
+			}
 			continue;
 		}
 
-		*offset = entry->offset;
-		*bytes = entry->bytes;
+		*offset = tmp;
+		*bytes = entry->bytes - align_off;
 		return entry;
 	}
-
+out:
 	return NULL;
 }
 
-static void add_new_bitmap(struct btrfs_block_group_cache *block_group,
+static void add_new_bitmap(struct btrfs_free_space_ctl *ctl,
 			   struct btrfs_free_space *info, u64 offset)
 {
-	u64 bytes_per_bg = BITS_PER_BITMAP * block_group->sectorsize;
-	int max_bitmaps = (int)div64_u64(block_group->key.offset +
-					 bytes_per_bg - 1, bytes_per_bg);
-	BUG_ON(block_group->total_bitmaps >= max_bitmaps);
-
-	info->offset = offset_to_bitmap(block_group, offset);
+	info->offset = offset_to_bitmap(ctl, offset);
 	info->bytes = 0;
-	link_free_space(block_group, info);
-	block_group->total_bitmaps++;
+	INIT_LIST_HEAD(&info->list);
+	link_free_space(ctl, info);
+	ctl->total_bitmaps++;
+
+	ctl->op->recalc_thresholds(ctl);
+}
 
-	recalculate_thresholds(block_group);
+static void free_bitmap(struct btrfs_free_space_ctl *ctl,
+			struct btrfs_free_space *bitmap_info)
+{
+	unlink_free_space(ctl, bitmap_info);
+	kfree(bitmap_info->bitmap);
+	kmem_cache_free(btrfs_free_space_cachep, bitmap_info);
+	ctl->total_bitmaps--;
+	ctl->op->recalc_thresholds(ctl);
 }
 
-static noinline int remove_from_bitmap(struct btrfs_block_group_cache *block_group,
+static noinline int remove_from_bitmap(struct btrfs_free_space_ctl *ctl,
 			      struct btrfs_free_space *bitmap_info,
 			      u64 *offset, u64 *bytes)
 {
@@ -427,44 +1708,35 @@ static noinline int remove_from_bitmap(struct btrfs_block_group_cache *block_gro
 	int ret;
 
 again:
-	end = bitmap_info->offset +
-		(u64)(BITS_PER_BITMAP * block_group->sectorsize) - 1;
+	end = bitmap_info->offset + (u64)(BITS_PER_BITMAP * ctl->unit) - 1;
 
 	/*
-	 * XXX - this can go away after a few releases.
-	 *
-	 * since the only user of btrfs_remove_free_space is the tree logging
-	 * stuff, and the only way to test that is under crash conditions, we
-	 * want to have this debug stuff here just in case somethings not
-	 * working.  Search the bitmap for the space we are trying to use to
-	 * make sure its actually there.  If its not there then we need to stop
-	 * because something has gone wrong.
+	 * We need to search for bits in this bitmap.  We could only cover some
+	 * of the extent in this bitmap thanks to how we add space, so we need
+	 * to search for as much as it as we can and clear that amount, and then
+	 * go searching for the next bit.
 	 */
 	search_start = *offset;
-	search_bytes = *bytes;
-	ret = search_bitmap(block_group, bitmap_info, &search_start,
-			    &search_bytes);
-	BUG_ON(ret < 0 || search_start != *offset);
+	search_bytes = ctl->unit;
+	search_bytes = min(search_bytes, end - search_start + 1);
+	ret = search_bitmap(ctl, bitmap_info, &search_start, &search_bytes);
+	if (ret < 0 || search_start != *offset)
+		return -EINVAL;
 
-	if (*offset > bitmap_info->offset && *offset + *bytes > end) {
-		bitmap_clear_bits(block_group, bitmap_info, *offset,
-				  end - *offset + 1);
-		*bytes -= end - *offset + 1;
-		*offset = end + 1;
-	} else if (*offset >= bitmap_info->offset && *offset + *bytes <= end) {
-		bitmap_clear_bits(block_group, bitmap_info, *offset, *bytes);
-		*bytes = 0;
-	}
+	/* We may have found more bits than what we need */
+	search_bytes = min(search_bytes, *bytes);
+
+	/* Cannot clear past the end of the bitmap */
+	search_bytes = min(search_bytes, end - search_start + 1);
+
+	bitmap_clear_bits(ctl, bitmap_info, search_start, search_bytes);
+	*offset += search_bytes;
+	*bytes -= search_bytes;
 
 	if (*bytes) {
 		struct rb_node *next = rb_next(&bitmap_info->offset_index);
-		if (!bitmap_info->bytes) {
-			unlink_free_space(block_group, bitmap_info);
-			kfree(bitmap_info->bitmap);
-			kfree(bitmap_info);
-			block_group->total_bitmaps--;
-			recalculate_thresholds(block_group);
-		}
+		if (!bitmap_info->bytes)
+			free_bitmap(ctl, bitmap_info);
 
 		/*
 		 * no entry after this bitmap, but we still have bytes to
@@ -490,75 +1762,149 @@ again:
 		 * everything over again.
 		 */
 		search_start = *offset;
-		search_bytes = *bytes;
-		ret = search_bitmap(block_group, bitmap_info, &search_start,
+		search_bytes = ctl->unit;
+		ret = search_bitmap(ctl, bitmap_info, &search_start,
 				    &search_bytes);
 		if (ret < 0 || search_start != *offset)
 			return -EAGAIN;
 
 		goto again;
-	} else if (!bitmap_info->bytes) {
-		unlink_free_space(block_group, bitmap_info);
-		kfree(bitmap_info->bitmap);
-		kfree(bitmap_info);
-		block_group->total_bitmaps--;
-		recalculate_thresholds(block_group);
-	}
+	} else if (!bitmap_info->bytes)
+		free_bitmap(ctl, bitmap_info);
 
 	return 0;
 }
 
-static int insert_into_bitmap(struct btrfs_block_group_cache *block_group,
-			      struct btrfs_free_space *info)
+static u64 add_bytes_to_bitmap(struct btrfs_free_space_ctl *ctl,
+			       struct btrfs_free_space *info, u64 offset,
+			       u64 bytes)
 {
-	struct btrfs_free_space *bitmap_info;
-	int added = 0;
-	u64 bytes, offset, end;
-	int ret;
+	u64 bytes_to_set = 0;
+	u64 end;
+
+	end = info->offset + (u64)(BITS_PER_BITMAP * ctl->unit);
+
+	bytes_to_set = min(end - offset, bytes);
+
+	bitmap_set_bits(ctl, info, offset, bytes_to_set);
+
+	return bytes_to_set;
+
+}
+
+static bool use_bitmap(struct btrfs_free_space_ctl *ctl,
+		      struct btrfs_free_space *info)
+{
+	struct btrfs_block_group_cache *block_group = ctl->private;
 
 	/*
 	 * If we are below the extents threshold then we can add this as an
 	 * extent, and don't have to deal with the bitmap
 	 */
-	if (block_group->free_extents < block_group->extents_thresh &&
-	    info->bytes > block_group->sectorsize * 4)
-		return 0;
+	if (ctl->free_extents < ctl->extents_thresh) {
+		/*
+		 * If this block group has some small extents we don't want to
+		 * use up all of our free slots in the cache with them, we want
+		 * to reserve them to larger extents, however if we have plent
+		 * of cache left then go ahead an dadd them, no sense in adding
+		 * the overhead of a bitmap if we don't have to.
+		 */
+		if (info->bytes <= block_group->sectorsize * 4) {
+			if (ctl->free_extents * 2 <= ctl->extents_thresh)
+				return false;
+		} else {
+			return false;
+		}
+	}
 
 	/*
-	 * some block groups are so tiny they can't be enveloped by a bitmap, so
-	 * don't even bother to create a bitmap for this
+	 * The original block groups from mkfs can be really small, like 8
+	 * megabytes, so don't bother with a bitmap for those entries.  However
+	 * some block groups can be smaller than what a bitmap would cover but
+	 * are still large enough that they could overflow the 32k memory limit,
+	 * so allow those block groups to still be allowed to have a bitmap
+	 * entry.
 	 */
-	if (BITS_PER_BITMAP * block_group->sectorsize >
-	    block_group->key.offset)
-		return 0;
+	if (((BITS_PER_BITMAP * ctl->unit) >> 1) > block_group->key.offset)
+		return false;
+
+	return true;
+}
+
+static struct btrfs_free_space_op free_space_op = {
+	.recalc_thresholds	= recalculate_thresholds,
+	.use_bitmap		= use_bitmap,
+};
+
+static int insert_into_bitmap(struct btrfs_free_space_ctl *ctl,
+			      struct btrfs_free_space *info)
+{
+	struct btrfs_free_space *bitmap_info;
+	struct btrfs_block_group_cache *block_group = NULL;
+	int added = 0;
+	u64 bytes, offset, bytes_added;
+	int ret;
 
 	bytes = info->bytes;
 	offset = info->offset;
 
+	if (!ctl->op->use_bitmap(ctl, info))
+		return 0;
+
+	if (ctl->op == &free_space_op)
+		block_group = ctl->private;
 again:
-	bitmap_info = tree_search_offset(block_group,
-					 offset_to_bitmap(block_group, offset),
+	/*
+	 * Since we link bitmaps right into the cluster we need to see if we
+	 * have a cluster here, and if so and it has our bitmap we need to add
+	 * the free space to that bitmap.
+	 */
+	if (block_group && !list_empty(&block_group->cluster_list)) {
+		struct btrfs_free_cluster *cluster;
+		struct rb_node *node;
+		struct btrfs_free_space *entry;
+
+		cluster = list_entry(block_group->cluster_list.next,
+				     struct btrfs_free_cluster,
+				     block_group_list);
+		spin_lock(&cluster->lock);
+		node = rb_first(&cluster->root);
+		if (!node) {
+			spin_unlock(&cluster->lock);
+			goto no_cluster_bitmap;
+		}
+
+		entry = rb_entry(node, struct btrfs_free_space, offset_index);
+		if (!entry->bitmap) {
+			spin_unlock(&cluster->lock);
+			goto no_cluster_bitmap;
+		}
+
+		if (entry->offset == offset_to_bitmap(ctl, offset)) {
+			bytes_added = add_bytes_to_bitmap(ctl, entry,
+							  offset, bytes);
+			bytes -= bytes_added;
+			offset += bytes_added;
+		}
+		spin_unlock(&cluster->lock);
+		if (!bytes) {
+			ret = 1;
+			goto out;
+		}
+	}
+
+no_cluster_bitmap:
+	bitmap_info = tree_search_offset(ctl, offset_to_bitmap(ctl, offset),
 					 1, 0);
 	if (!bitmap_info) {
-		BUG_ON(added);
+		ASSERT(added == 0);
 		goto new_bitmap;
 	}
 
-	end = bitmap_info->offset +
-		(u64)(BITS_PER_BITMAP * block_group->sectorsize);
-
-	if (offset >= bitmap_info->offset && offset + bytes > end) {
-		bitmap_set_bits(block_group, bitmap_info, offset,
-				end - offset);
-		bytes -= end - offset;
-		offset = end;
-		added = 0;
-	} else if (offset >= bitmap_info->offset && offset + bytes <= end) {
-		bitmap_set_bits(block_group, bitmap_info, offset, bytes);
-		bytes = 0;
-	} else {
-		BUG();
-	}
+	bytes_added = add_bytes_to_bitmap(ctl, bitmap_info, offset, bytes);
+	bytes -= bytes_added;
+	offset += bytes_added;
+	added = 0;
 
 	if (!bytes) {
 		ret = 1;
@@ -568,19 +1914,19 @@ again:
 
 new_bitmap:
 	if (info && info->bitmap) {
-		add_new_bitmap(block_group, info, offset);
+		add_new_bitmap(ctl, info, offset);
 		added = 1;
 		info = NULL;
 		goto again;
 	} else {
-		spin_unlock(&block_group->tree_lock);
+		spin_unlock(&ctl->tree_lock);
 
 		/* no pre-allocated info, allocate a new one */
 		if (!info) {
-			info = kzalloc(sizeof(struct btrfs_free_space),
-				       GFP_NOFS);
+			info = kmem_cache_zalloc(btrfs_free_space_cachep,
+						 GFP_NOFS);
 			if (!info) {
-				spin_lock(&block_group->tree_lock);
+				spin_lock(&ctl->tree_lock);
 				ret = -ENOMEM;
 				goto out;
 			}
@@ -588,7 +1934,7 @@ new_bitmap:
 
 		/* allocate the bitmap */
 		info->bitmap = kzalloc(PAGE_CACHE_SIZE, GFP_NOFS);
-		spin_lock(&block_group->tree_lock);
+		spin_lock(&ctl->tree_lock);
 		if (!info->bitmap) {
 			ret = -ENOMEM;
 			goto out;
@@ -600,81 +1946,98 @@ out:
 	if (info) {
 		if (info->bitmap)
 			kfree(info->bitmap);
-		kfree(info);
+		kmem_cache_free(btrfs_free_space_cachep, info);
 	}
 
 	return ret;
 }
 
-int btrfs_add_free_space(struct btrfs_block_group_cache *block_group,
-			 u64 offset, u64 bytes)
+static bool try_merge_free_space(struct btrfs_free_space_ctl *ctl,
+			  struct btrfs_free_space *info, bool update_stat)
 {
-	struct btrfs_free_space *right_info = NULL;
-	struct btrfs_free_space *left_info = NULL;
-	struct btrfs_free_space *info = NULL;
-	int ret = 0;
-
-	info = kzalloc(sizeof(struct btrfs_free_space), GFP_NOFS);
-	if (!info)
-		return -ENOMEM;
-
-	info->offset = offset;
-	info->bytes = bytes;
-
-	spin_lock(&block_group->tree_lock);
+	struct btrfs_free_space *left_info;
+	struct btrfs_free_space *right_info;
+	bool merged = false;
+	u64 offset = info->offset;
+	u64 bytes = info->bytes;
 
 	/*
 	 * first we want to see if there is free space adjacent to the range we
 	 * are adding, if there is remove that struct and add a new one to
 	 * cover the entire range
 	 */
-	right_info = tree_search_offset(block_group, offset + bytes, 0, 0);
+	right_info = tree_search_offset(ctl, offset + bytes, 0, 0);
 	if (right_info && rb_prev(&right_info->offset_index))
 		left_info = rb_entry(rb_prev(&right_info->offset_index),
 				     struct btrfs_free_space, offset_index);
 	else
-		left_info = tree_search_offset(block_group, offset - 1, 0, 0);
-
-	/*
-	 * If there was no extent directly to the left or right of this new
-	 * extent then we know we're going to have to allocate a new extent, so
-	 * before we do that see if we need to drop this into a bitmap
-	 */
-	if ((!left_info || left_info->bitmap) &&
-	    (!right_info || right_info->bitmap)) {
-		ret = insert_into_bitmap(block_group, info);
-
-		if (ret < 0) {
-			goto out;
-		} else if (ret) {
-			ret = 0;
-			goto out;
-		}
-	}
+		left_info = tree_search_offset(ctl, offset - 1, 0, 0);
 
 	if (right_info && !right_info->bitmap) {
-		unlink_free_space(block_group, right_info);
+		if (update_stat)
+			unlink_free_space(ctl, right_info);
+		else
+			__unlink_free_space(ctl, right_info);
 		info->bytes += right_info->bytes;
-		kfree(right_info);
+		kmem_cache_free(btrfs_free_space_cachep, right_info);
+		merged = true;
 	}
 
 	if (left_info && !left_info->bitmap &&
 	    left_info->offset + left_info->bytes == offset) {
-		unlink_free_space(block_group, left_info);
+		if (update_stat)
+			unlink_free_space(ctl, left_info);
+		else
+			__unlink_free_space(ctl, left_info);
 		info->offset = left_info->offset;
 		info->bytes += left_info->bytes;
-		kfree(left_info);
+		kmem_cache_free(btrfs_free_space_cachep, left_info);
+		merged = true;
 	}
 
-	ret = link_free_space(block_group, info);
+	return merged;
+}
+
+int __btrfs_add_free_space(struct btrfs_free_space_ctl *ctl,
+			   u64 offset, u64 bytes)
+{
+	struct btrfs_free_space *info;
+	int ret = 0;
+
+	info = kmem_cache_zalloc(btrfs_free_space_cachep, GFP_NOFS);
+	if (!info)
+		return -ENOMEM;
+
+	info->offset = offset;
+	info->bytes = bytes;
+
+	spin_lock(&ctl->tree_lock);
+
+	if (try_merge_free_space(ctl, info, true))
+		goto link;
+
+	/*
+	 * There was no extent directly to the left or right of this new
+	 * extent then we know we're going to have to allocate a new extent, so
+	 * before we do that see if we need to drop this into a bitmap
+	 */
+	ret = insert_into_bitmap(ctl, info);
+	if (ret < 0) {
+		goto out;
+	} else if (ret) {
+		ret = 0;
+		goto out;
+	}
+link:
+	ret = link_free_space(ctl, info);
 	if (ret)
-		kfree(info);
+		kmem_cache_free(btrfs_free_space_cachep, info);
 out:
-	spin_unlock(&block_group->tree_lock);
+	spin_unlock(&ctl->tree_lock);
 
 	if (ret) {
-		printk(KERN_CRIT "btrfs: unable to add free space :%d\n", ret);
-		BUG_ON(ret == -EEXIST);
+		printk(KERN_CRIT "BTRFS: unable to add free space :%d\n", ret);
+		ASSERT(ret != -EEXIST);
 	}
 
 	return ret;
@@ -683,116 +2046,89 @@ out:
 int btrfs_remove_free_space(struct btrfs_block_group_cache *block_group,
 			    u64 offset, u64 bytes)
 {
+	struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
 	struct btrfs_free_space *info;
-	struct btrfs_free_space *next_info = NULL;
-	int ret = 0;
+	int ret;
+	bool re_search = false;
 
-	spin_lock(&block_group->tree_lock);
+	spin_lock(&ctl->tree_lock);
 
 again:
-	info = tree_search_offset(block_group, offset, 0, 0);
+	ret = 0;
+	if (!bytes)
+		goto out_lock;
+
+	info = tree_search_offset(ctl, offset, 0, 0);
 	if (!info) {
 		/*
 		 * oops didn't find an extent that matched the space we wanted
 		 * to remove, look for a bitmap instead
 		 */
-		info = tree_search_offset(block_group,
-					  offset_to_bitmap(block_group, offset),
+		info = tree_search_offset(ctl, offset_to_bitmap(ctl, offset),
 					  1, 0);
 		if (!info) {
-			WARN_ON(1);
-			goto out_lock;
-		}
-	}
-
-	if (info->bytes < bytes && rb_next(&info->offset_index)) {
-		u64 end;
-		next_info = rb_entry(rb_next(&info->offset_index),
-					     struct btrfs_free_space,
-					     offset_index);
-
-		if (next_info->bitmap)
-			end = next_info->offset + BITS_PER_BITMAP *
-				block_group->sectorsize - 1;
-		else
-			end = next_info->offset + next_info->bytes;
-
-		if (next_info->bytes < bytes ||
-		    next_info->offset > offset || offset > end) {
-			printk(KERN_CRIT "Found free space at %llu, size %llu,"
-			      " trying to use %llu\n",
-			      (unsigned long long)info->offset,
-			      (unsigned long long)info->bytes,
-			      (unsigned long long)bytes);
-			WARN_ON(1);
-			ret = -EINVAL;
+			/*
+			 * If we found a partial bit of our free space in a
+			 * bitmap but then couldn't find the other part this may
+			 * be a problem, so WARN about it.
+			 */
+			WARN_ON(re_search);
 			goto out_lock;
 		}
-
-		info = next_info;
 	}
 
-	if (info->bytes == bytes) {
-		unlink_free_space(block_group, info);
-		if (info->bitmap) {
-			kfree(info->bitmap);
-			block_group->total_bitmaps--;
-		}
-		kfree(info);
-		goto out_lock;
-	}
-
-	if (!info->bitmap && info->offset == offset) {
-		unlink_free_space(block_group, info);
-		info->offset += bytes;
-		info->bytes -= bytes;
-		link_free_space(block_group, info);
-		goto out_lock;
-	}
+	re_search = false;
+	if (!info->bitmap) {
+		unlink_free_space(ctl, info);
+		if (offset == info->offset) {
+			u64 to_free = min(bytes, info->bytes);
+
+			info->bytes -= to_free;
+			info->offset += to_free;
+			if (info->bytes) {
+				ret = link_free_space(ctl, info);
+				WARN_ON(ret);
+			} else {
+				kmem_cache_free(btrfs_free_space_cachep, info);
+			}
 
-	if (!info->bitmap && info->offset <= offset &&
-	    info->offset + info->bytes >= offset + bytes) {
-		u64 old_start = info->offset;
-		/*
-		 * we're freeing space in the middle of the info,
-		 * this can happen during tree log replay
-		 *
-		 * first unlink the old info and then
-		 * insert it again after the hole we're creating
-		 */
-		unlink_free_space(block_group, info);
-		if (offset + bytes < info->offset + info->bytes) {
-			u64 old_end = info->offset + info->bytes;
+			offset += to_free;
+			bytes -= to_free;
+			goto again;
+		} else {
+			u64 old_end = info->bytes + info->offset;
 
-			info->offset = offset + bytes;
-			info->bytes = old_end - info->offset;
-			ret = link_free_space(block_group, info);
+			info->bytes = offset - info->offset;
+			ret = link_free_space(ctl, info);
 			WARN_ON(ret);
 			if (ret)
 				goto out_lock;
-		} else {
-			/* the hole we're creating ends at the end
-			 * of the info struct, just free the info
-			 */
-			kfree(info);
-		}
-		spin_unlock(&block_group->tree_lock);
 
-		/* step two, insert a new info struct to cover
-		 * anything before the hole
-		 */
-		ret = btrfs_add_free_space(block_group, old_start,
-					   offset - old_start);
-		WARN_ON(ret);
-		goto out;
+			/* Not enough bytes in this entry to satisfy us */
+			if (old_end < offset + bytes) {
+				bytes -= old_end - offset;
+				offset = old_end;
+				goto again;
+			} else if (old_end == offset + bytes) {
+				/* all done */
+				goto out_lock;
+			}
+			spin_unlock(&ctl->tree_lock);
+
+			ret = btrfs_add_free_space(block_group, offset + bytes,
+						   old_end - (offset + bytes));
+			WARN_ON(ret);
+			goto out;
+		}
 	}
 
-	ret = remove_from_bitmap(block_group, info, &offset, &bytes);
-	if (ret == -EAGAIN)
+	ret = remove_from_bitmap(ctl, info, &offset, &bytes);
+	if (ret == -EAGAIN) {
+		re_search = true;
 		goto again;
-	BUG_ON(ret);
+	}
 out_lock:
-	spin_unlock(&block_group->tree_lock);
+	spin_unlock(&ctl->tree_lock);
 out:
 	return ret;
 }
@@ -800,38 +2136,43 @@ out:
 void btrfs_dump_free_space(struct btrfs_block_group_cache *block_group,
 			   u64 bytes)
 {
+	struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
 	struct btrfs_free_space *info;
 	struct rb_node *n;
 	int count = 0;
 
-	for (n = rb_first(&block_group->free_space_offset); n; n = rb_next(n)) {
+	for (n = rb_first(&ctl->free_space_offset); n; n = rb_next(n)) {
 		info = rb_entry(n, struct btrfs_free_space, offset_index);
-		if (info->bytes >= bytes)
+		if (info->bytes >= bytes && !block_group->ro)
 			count++;
-		printk(KERN_CRIT "entry offset %llu, bytes %llu, bitmap %s\n",
-		       (unsigned long long)info->offset,
-		       (unsigned long long)info->bytes,
+		btrfs_crit(block_group->fs_info,
+			   "entry offset %llu, bytes %llu, bitmap %s",
+			   info->offset, info->bytes,
 		       (info->bitmap) ? "yes" : "no");
 	}
-	printk(KERN_INFO "block group has cluster?: %s\n",
+	btrfs_info(block_group->fs_info, "block group has cluster?: %s",
 	       list_empty(&block_group->cluster_list) ? "no" : "yes");
-	printk(KERN_INFO "%d blocks of free space at or bigger than bytes is"
-	       "\n", count);
+	btrfs_info(block_group->fs_info,
+		   "%d blocks of free space at or bigger than bytes is", count);
 }
 
-u64 btrfs_block_group_free_space(struct btrfs_block_group_cache *block_group)
+void btrfs_init_free_space_ctl(struct btrfs_block_group_cache *block_group)
 {
-	struct btrfs_free_space *info;
-	struct rb_node *n;
-	u64 ret = 0;
+	struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
 
-	for (n = rb_first(&block_group->free_space_offset); n;
-	     n = rb_next(n)) {
-		info = rb_entry(n, struct btrfs_free_space, offset_index);
-		ret += info->bytes;
-	}
+	spin_lock_init(&ctl->tree_lock);
+	ctl->unit = block_group->sectorsize;
+	ctl->start = block_group->key.objectid;
+	ctl->private = block_group;
+	ctl->op = &free_space_op;
 
-	return ret;
+	/*
+	 * we only want to have 32k of ram per block group for keeping
+	 * track of free space, and if we pass 1/2 of that we want to
+	 * start converting things over to using bitmaps
+	 */
+	ctl->extents_thresh = ((1024 * 32) / 2) /
+				sizeof(struct btrfs_free_space);
 }
 
 /*
@@ -845,31 +2186,31 @@ __btrfs_return_cluster_to_free_space(
 			     struct btrfs_block_group_cache *block_group,
 			     struct btrfs_free_cluster *cluster)
 {
+	struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
 	struct btrfs_free_space *entry;
 	struct rb_node *node;
-	bool bitmap;
 
 	spin_lock(&cluster->lock);
 	if (cluster->block_group != block_group)
 		goto out;
 
-	bitmap = cluster->points_to_bitmap;
 	cluster->block_group = NULL;
 	cluster->window_start = 0;
 	list_del_init(&cluster->block_group_list);
-	cluster->points_to_bitmap = false;
-
-	if (bitmap)
-		goto out;
 
 	node = rb_first(&cluster->root);
 	while (node) {
+		bool bitmap;
+
 		entry = rb_entry(node, struct btrfs_free_space, offset_index);
 		node = rb_next(&entry->offset_index);
 		rb_erase(&entry->offset_index, &cluster->root);
-		BUG_ON(entry->bitmap);
-		tree_insert_offset(&block_group->free_space_offset,
-				   entry->offset, &entry->offset_index, 0);
+
+		bitmap = (entry->bitmap != NULL);
+		if (!bitmap)
+			try_merge_free_space(ctl, entry, false);
+		tree_insert_offset(&ctl->free_space_offset,
+				   entry->offset, &entry->offset_index, bitmap);
 	}
 	cluster->root = RB_ROOT;
 
@@ -879,14 +2220,42 @@ out:
 	return 0;
 }
 
-void btrfs_remove_free_space_cache(struct btrfs_block_group_cache *block_group)
+static void __btrfs_remove_free_space_cache_locked(
+				struct btrfs_free_space_ctl *ctl)
 {
 	struct btrfs_free_space *info;
 	struct rb_node *node;
+
+	while ((node = rb_last(&ctl->free_space_offset)) != NULL) {
+		info = rb_entry(node, struct btrfs_free_space, offset_index);
+		if (!info->bitmap) {
+			unlink_free_space(ctl, info);
+			kmem_cache_free(btrfs_free_space_cachep, info);
+		} else {
+			free_bitmap(ctl, info);
+		}
+		if (need_resched()) {
+			spin_unlock(&ctl->tree_lock);
+			cond_resched();
+			spin_lock(&ctl->tree_lock);
+		}
+	}
+}
+
+void __btrfs_remove_free_space_cache(struct btrfs_free_space_ctl *ctl)
+{
+	spin_lock(&ctl->tree_lock);
+	__btrfs_remove_free_space_cache_locked(ctl);
+	spin_unlock(&ctl->tree_lock);
+}
+
+void btrfs_remove_free_space_cache(struct btrfs_block_group_cache *block_group)
+{
+	struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
 	struct btrfs_free_cluster *cluster;
 	struct list_head *head;
 
-	spin_lock(&block_group->tree_lock);
+	spin_lock(&ctl->tree_lock);
 	while ((head = block_group->cluster_list.next) !=
 	       &block_group->cluster_list) {
 		cluster = list_entry(head, struct btrfs_free_cluster,
@@ -895,63 +2264,57 @@ void btrfs_remove_free_space_cache(struct btrfs_block_group_cache *block_group)
 		WARN_ON(cluster->block_group != block_group);
 		__btrfs_return_cluster_to_free_space(block_group, cluster);
 		if (need_resched()) {
-			spin_unlock(&block_group->tree_lock);
+			spin_unlock(&ctl->tree_lock);
 			cond_resched();
-			spin_lock(&block_group->tree_lock);
+			spin_lock(&ctl->tree_lock);
 		}
 	}
+	__btrfs_remove_free_space_cache_locked(ctl);
+	spin_unlock(&ctl->tree_lock);
 
-	while ((node = rb_last(&block_group->free_space_offset)) != NULL) {
-		info = rb_entry(node, struct btrfs_free_space, offset_index);
-		unlink_free_space(block_group, info);
-		if (info->bitmap)
-			kfree(info->bitmap);
-		kfree(info);
-		if (need_resched()) {
-			spin_unlock(&block_group->tree_lock);
-			cond_resched();
-			spin_lock(&block_group->tree_lock);
-		}
-	}
-
-	spin_unlock(&block_group->tree_lock);
 }
 
 u64 btrfs_find_space_for_alloc(struct btrfs_block_group_cache *block_group,
-			       u64 offset, u64 bytes, u64 empty_size)
+			       u64 offset, u64 bytes, u64 empty_size,
+			       u64 *max_extent_size)
 {
+	struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
 	struct btrfs_free_space *entry = NULL;
 	u64 bytes_search = bytes + empty_size;
 	u64 ret = 0;
+	u64 align_gap = 0;
+	u64 align_gap_len = 0;
 
-	spin_lock(&block_group->tree_lock);
-	entry = find_free_space(block_group, &offset, &bytes_search, 0);
+	spin_lock(&ctl->tree_lock);
+	entry = find_free_space(ctl, &offset, &bytes_search,
+				block_group->full_stripe_len, max_extent_size);
 	if (!entry)
 		goto out;
 
 	ret = offset;
 	if (entry->bitmap) {
-		bitmap_clear_bits(block_group, entry, offset, bytes);
-		if (!entry->bytes) {
-			unlink_free_space(block_group, entry);
-			kfree(entry->bitmap);
-			kfree(entry);
-			block_group->total_bitmaps--;
-			recalculate_thresholds(block_group);
-		}
+		bitmap_clear_bits(ctl, entry, offset, bytes);
+		if (!entry->bytes)
+			free_bitmap(ctl, entry);
 	} else {
-		unlink_free_space(block_group, entry);
-		entry->offset += bytes;
-		entry->bytes -= bytes;
+		unlink_free_space(ctl, entry);
+		align_gap_len = offset - entry->offset;
+		align_gap = entry->offset;
+
+		entry->offset = offset + bytes;
+		WARN_ON(entry->bytes < bytes + align_gap_len);
+
+		entry->bytes -= bytes + align_gap_len;
 		if (!entry->bytes)
-			kfree(entry);
+			kmem_cache_free(btrfs_free_space_cachep, entry);
 		else
-			link_free_space(block_group, entry);
+			link_free_space(ctl, entry);
 	}
-
 out:
-	spin_unlock(&block_group->tree_lock);
+	spin_unlock(&ctl->tree_lock);
 
+	if (align_gap_len)
+		__btrfs_add_free_space(ctl, align_gap, align_gap_len);
 	return ret;
 }
 
@@ -967,6 +2330,7 @@ int btrfs_return_cluster_to_free_space(
 			       struct btrfs_block_group_cache *block_group,
 			       struct btrfs_free_cluster *cluster)
 {
+	struct btrfs_free_space_ctl *ctl;
 	int ret;
 
 	/* first, get a safe pointer to the block group */
@@ -985,10 +2349,12 @@ int btrfs_return_cluster_to_free_space(
 	atomic_inc(&block_group->count);
 	spin_unlock(&cluster->lock);
 
+	ctl = block_group->free_space_ctl;
+
 	/* now return any extents the cluster had on it */
-	spin_lock(&block_group->tree_lock);
+	spin_lock(&ctl->tree_lock);
 	ret = __btrfs_return_cluster_to_free_space(block_group, cluster);
-	spin_unlock(&block_group->tree_lock);
+	spin_unlock(&ctl->tree_lock);
 
 	/* finally drop our ref */
 	btrfs_put_block_group(block_group);
@@ -997,48 +2363,28 @@ int btrfs_return_cluster_to_free_space(
 
 static u64 btrfs_alloc_from_bitmap(struct btrfs_block_group_cache *block_group,
 				   struct btrfs_free_cluster *cluster,
-				   u64 bytes, u64 min_start)
+				   struct btrfs_free_space *entry,
+				   u64 bytes, u64 min_start,
+				   u64 *max_extent_size)
 {
-	struct btrfs_free_space *entry;
+	struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
 	int err;
 	u64 search_start = cluster->window_start;
 	u64 search_bytes = bytes;
 	u64 ret = 0;
 
-	spin_lock(&block_group->tree_lock);
-	spin_lock(&cluster->lock);
-
-	if (!cluster->points_to_bitmap)
-		goto out;
-
-	if (cluster->block_group != block_group)
-		goto out;
-
-	/*
-	 * search_start is the beginning of the bitmap, but at some point it may
-	 * be a good idea to point to the actual start of the free area in the
-	 * bitmap, so do the offset_to_bitmap trick anyway, and set bitmap_only
-	 * to 1 to make sure we get the bitmap entry
-	 */
-	entry = tree_search_offset(block_group,
-				   offset_to_bitmap(block_group, search_start),
-				   1, 0);
-	if (!entry || !entry->bitmap)
-		goto out;
-
 	search_start = min_start;
 	search_bytes = bytes;
 
-	err = search_bitmap(block_group, entry, &search_start,
-			    &search_bytes);
-	if (err)
-		goto out;
+	err = search_bitmap(ctl, entry, &search_start, &search_bytes);
+	if (err) {
+		if (search_bytes > *max_extent_size)
+			*max_extent_size = search_bytes;
+		return 0;
+	}
 
 	ret = search_start;
-	bitmap_clear_bits(block_group, entry, ret, bytes);
-out:
-	spin_unlock(&cluster->lock);
-	spin_unlock(&block_group->tree_lock);
+	__bitmap_clear_bits(ctl, entry, ret, bytes);
 
 	return ret;
 }
@@ -1050,16 +2396,13 @@ out:
  */
 u64 btrfs_alloc_from_cluster(struct btrfs_block_group_cache *block_group,
 			     struct btrfs_free_cluster *cluster, u64 bytes,
-			     u64 min_start)
+			     u64 min_start, u64 *max_extent_size)
 {
+	struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
 	struct btrfs_free_space *entry = NULL;
 	struct rb_node *node;
 	u64 ret = 0;
 
-	if (cluster->points_to_bitmap)
-		return btrfs_alloc_from_bitmap(block_group, cluster, bytes,
-					       min_start);
-
 	spin_lock(&cluster->lock);
 	if (bytes > cluster->max_size)
 		goto out;
@@ -1072,11 +2415,12 @@ u64 btrfs_alloc_from_cluster(struct btrfs_block_group_cache *block_group,
 		goto out;
 
 	entry = rb_entry(node, struct btrfs_free_space, offset_index);
+	while (1) {
+		if (entry->bytes < bytes && entry->bytes > *max_extent_size)
+			*max_extent_size = entry->bytes;
 
-	while(1) {
-		if (entry->bytes < bytes || entry->offset < min_start) {
-			struct rb_node *node;
-
+		if (entry->bytes < bytes ||
+		    (!entry->bitmap && entry->offset < min_start)) {
 			node = rb_next(&entry->offset_index);
 			if (!node)
 				break;
@@ -1084,50 +2428,83 @@ u64 btrfs_alloc_from_cluster(struct btrfs_block_group_cache *block_group,
 					 offset_index);
 			continue;
 		}
-		ret = entry->offset;
 
-		entry->offset += bytes;
-		entry->bytes -= bytes;
+		if (entry->bitmap) {
+			ret = btrfs_alloc_from_bitmap(block_group,
+						      cluster, entry, bytes,
+						      cluster->window_start,
+						      max_extent_size);
+			if (ret == 0) {
+				node = rb_next(&entry->offset_index);
+				if (!node)
+					break;
+				entry = rb_entry(node, struct btrfs_free_space,
+						 offset_index);
+				continue;
+			}
+			cluster->window_start += bytes;
+		} else {
+			ret = entry->offset;
+
+			entry->offset += bytes;
+			entry->bytes -= bytes;
+		}
 
-		if (entry->bytes == 0) {
+		if (entry->bytes == 0)
 			rb_erase(&entry->offset_index, &cluster->root);
-			kfree(entry);
-		}
 		break;
 	}
 out:
 	spin_unlock(&cluster->lock);
 
+	if (!ret)
+		return 0;
+
+	spin_lock(&ctl->tree_lock);
+
+	ctl->free_space -= bytes;
+	if (entry->bytes == 0) {
+		ctl->free_extents--;
+		if (entry->bitmap) {
+			kfree(entry->bitmap);
+			ctl->total_bitmaps--;
+			ctl->op->recalc_thresholds(ctl);
+		}
+		kmem_cache_free(btrfs_free_space_cachep, entry);
+	}
+
+	spin_unlock(&ctl->tree_lock);
+
 	return ret;
 }
 
 static int btrfs_bitmap_cluster(struct btrfs_block_group_cache *block_group,
 				struct btrfs_free_space *entry,
 				struct btrfs_free_cluster *cluster,
-				u64 offset, u64 bytes, u64 min_bytes)
+				u64 offset, u64 bytes,
+				u64 cont1_bytes, u64 min_bytes)
 {
+	struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
 	unsigned long next_zero;
 	unsigned long i;
-	unsigned long search_bits;
-	unsigned long total_bits;
+	unsigned long want_bits;
+	unsigned long min_bits;
 	unsigned long found_bits;
 	unsigned long start = 0;
 	unsigned long total_found = 0;
-	bool found = false;
+	int ret;
 
-	i = offset_to_bit(entry->offset, block_group->sectorsize,
+	i = offset_to_bit(entry->offset, ctl->unit,
 			  max_t(u64, offset, entry->offset));
-	search_bits = bytes_to_bits(min_bytes, block_group->sectorsize);
-	total_bits = bytes_to_bits(bytes, block_group->sectorsize);
+	want_bits = bytes_to_bits(bytes, ctl->unit);
+	min_bits = bytes_to_bits(min_bytes, ctl->unit);
 
 again:
 	found_bits = 0;
-	for (i = find_next_bit(entry->bitmap, BITS_PER_BITMAP, i);
-	     i < BITS_PER_BITMAP;
-	     i = find_next_bit(entry->bitmap, BITS_PER_BITMAP, i + 1)) {
+	for_each_set_bit_from(i, entry->bitmap, BITS_PER_BITMAP) {
 		next_zero = find_next_zero_bit(entry->bitmap,
 					       BITS_PER_BITMAP, i);
-		if (next_zero - i >= search_bits) {
+		if (next_zero - i >= min_bits) {
 			found_bits = next_zero - i;
 			break;
 		}
@@ -1135,77 +2512,218 @@ again:
 	}
 
 	if (!found_bits)
-		return -1;
+		return -ENOSPC;
 
-	if (!found) {
+	if (!total_found) {
 		start = i;
-		found = true;
+		cluster->max_size = 0;
 	}
 
 	total_found += found_bits;
 
-	if (cluster->max_size < found_bits * block_group->sectorsize)
-		cluster->max_size = found_bits * block_group->sectorsize;
+	if (cluster->max_size < found_bits * ctl->unit)
+		cluster->max_size = found_bits * ctl->unit;
 
-	if (total_found < total_bits) {
-		i = find_next_bit(entry->bitmap, BITS_PER_BITMAP, next_zero);
-		if (i - start > total_bits * 2) {
-			total_found = 0;
-			cluster->max_size = 0;
-			found = false;
-		}
+	if (total_found < want_bits || cluster->max_size < cont1_bytes) {
+		i = next_zero + 1;
 		goto again;
 	}
 
-	cluster->window_start = start * block_group->sectorsize +
-		entry->offset;
-	cluster->points_to_bitmap = true;
+	cluster->window_start = start * ctl->unit + entry->offset;
+	rb_erase(&entry->offset_index, &ctl->free_space_offset);
+	ret = tree_insert_offset(&cluster->root, entry->offset,
+				 &entry->offset_index, 1);
+	ASSERT(!ret); /* -EEXIST; Logic error */
 
+	trace_btrfs_setup_cluster(block_group, cluster,
+				  total_found * ctl->unit, 1);
 	return 0;
 }
 
 /*
+ * This searches the block group for just extents to fill the cluster with.
+ * Try to find a cluster with at least bytes total bytes, at least one
+ * extent of cont1_bytes, and other clusters of at least min_bytes.
+ */
+static noinline int
+setup_cluster_no_bitmap(struct btrfs_block_group_cache *block_group,
+			struct btrfs_free_cluster *cluster,
+			struct list_head *bitmaps, u64 offset, u64 bytes,
+			u64 cont1_bytes, u64 min_bytes)
+{
+	struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
+	struct btrfs_free_space *first = NULL;
+	struct btrfs_free_space *entry = NULL;
+	struct btrfs_free_space *last;
+	struct rb_node *node;
+	u64 window_free;
+	u64 max_extent;
+	u64 total_size = 0;
+
+	entry = tree_search_offset(ctl, offset, 0, 1);
+	if (!entry)
+		return -ENOSPC;
+
+	/*
+	 * We don't want bitmaps, so just move along until we find a normal
+	 * extent entry.
+	 */
+	while (entry->bitmap || entry->bytes < min_bytes) {
+		if (entry->bitmap && list_empty(&entry->list))
+			list_add_tail(&entry->list, bitmaps);
+		node = rb_next(&entry->offset_index);
+		if (!node)
+			return -ENOSPC;
+		entry = rb_entry(node, struct btrfs_free_space, offset_index);
+	}
+
+	window_free = entry->bytes;
+	max_extent = entry->bytes;
+	first = entry;
+	last = entry;
+
+	for (node = rb_next(&entry->offset_index); node;
+	     node = rb_next(&entry->offset_index)) {
+		entry = rb_entry(node, struct btrfs_free_space, offset_index);
+
+		if (entry->bitmap) {
+			if (list_empty(&entry->list))
+				list_add_tail(&entry->list, bitmaps);
+			continue;
+		}
+
+		if (entry->bytes < min_bytes)
+			continue;
+
+		last = entry;
+		window_free += entry->bytes;
+		if (entry->bytes > max_extent)
+			max_extent = entry->bytes;
+	}
+
+	if (window_free < bytes || max_extent < cont1_bytes)
+		return -ENOSPC;
+
+	cluster->window_start = first->offset;
+
+	node = &first->offset_index;
+
+	/*
+	 * now we've found our entries, pull them out of the free space
+	 * cache and put them into the cluster rbtree
+	 */
+	do {
+		int ret;
+
+		entry = rb_entry(node, struct btrfs_free_space, offset_index);
+		node = rb_next(&entry->offset_index);
+		if (entry->bitmap || entry->bytes < min_bytes)
+			continue;
+
+		rb_erase(&entry->offset_index, &ctl->free_space_offset);
+		ret = tree_insert_offset(&cluster->root, entry->offset,
+					 &entry->offset_index, 0);
+		total_size += entry->bytes;
+		ASSERT(!ret); /* -EEXIST; Logic error */
+	} while (node && entry != last);
+
+	cluster->max_size = max_extent;
+	trace_btrfs_setup_cluster(block_group, cluster, total_size, 0);
+	return 0;
+}
+
+/*
+ * This specifically looks for bitmaps that may work in the cluster, we assume
+ * that we have already failed to find extents that will work.
+ */
+static noinline int
+setup_cluster_bitmap(struct btrfs_block_group_cache *block_group,
+		     struct btrfs_free_cluster *cluster,
+		     struct list_head *bitmaps, u64 offset, u64 bytes,
+		     u64 cont1_bytes, u64 min_bytes)
+{
+	struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
+	struct btrfs_free_space *entry;
+	int ret = -ENOSPC;
+	u64 bitmap_offset = offset_to_bitmap(ctl, offset);
+
+	if (ctl->total_bitmaps == 0)
+		return -ENOSPC;
+
+	/*
+	 * The bitmap that covers offset won't be in the list unless offset
+	 * is just its start offset.
+	 */
+	entry = list_first_entry(bitmaps, struct btrfs_free_space, list);
+	if (entry->offset != bitmap_offset) {
+		entry = tree_search_offset(ctl, bitmap_offset, 1, 0);
+		if (entry && list_empty(&entry->list))
+			list_add(&entry->list, bitmaps);
+	}
+
+	list_for_each_entry(entry, bitmaps, list) {
+		if (entry->bytes < bytes)
+			continue;
+		ret = btrfs_bitmap_cluster(block_group, entry, cluster, offset,
+					   bytes, cont1_bytes, min_bytes);
+		if (!ret)
+			return 0;
+	}
+
+	/*
+	 * The bitmaps list has all the bitmaps that record free space
+	 * starting after offset, so no more search is required.
+	 */
+	return -ENOSPC;
+}
+
+/*
  * here we try to find a cluster of blocks in a block group.  The goal
- * is to find at least bytes free and up to empty_size + bytes free.
+ * is to find at least bytes+empty_size.
  * We might not find them all in one contiguous area.
  *
  * returns zero and sets up cluster if things worked out, otherwise
  * it returns -enospc
  */
-int btrfs_find_space_cluster(struct btrfs_trans_handle *trans,
-			     struct btrfs_root *root,
+int btrfs_find_space_cluster(struct btrfs_root *root,
 			     struct btrfs_block_group_cache *block_group,
 			     struct btrfs_free_cluster *cluster,
 			     u64 offset, u64 bytes, u64 empty_size)
 {
-	struct btrfs_free_space *entry = NULL;
-	struct rb_node *node;
-	struct btrfs_free_space *next;
-	struct btrfs_free_space *last = NULL;
+	struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
+	struct btrfs_free_space *entry, *tmp;
+	LIST_HEAD(bitmaps);
 	u64 min_bytes;
-	u64 window_start;
-	u64 window_free;
-	u64 max_extent = 0;
-	bool found_bitmap = false;
+	u64 cont1_bytes;
 	int ret;
 
-	/* for metadata, allow allocates with more holes */
+	/*
+	 * Choose the minimum extent size we'll require for this
+	 * cluster.  For SSD_SPREAD, don't allow any fragmentation.
+	 * For metadata, allow allocates with smaller extents.  For
+	 * data, keep it dense.
+	 */
 	if (btrfs_test_opt(root, SSD_SPREAD)) {
-		min_bytes = bytes + empty_size;
+		cont1_bytes = min_bytes = bytes + empty_size;
 	} else if (block_group->flags & BTRFS_BLOCK_GROUP_METADATA) {
-		/*
-		 * we want to do larger allocations when we are
-		 * flushing out the delayed refs, it helps prevent
-		 * making more work as we go along.
-		 */
-		if (trans->transaction->delayed_refs.flushing)
-			min_bytes = max(bytes, (bytes + empty_size) >> 1);
-		else
-			min_bytes = max(bytes, (bytes + empty_size) >> 4);
-	} else
-		min_bytes = max(bytes, (bytes + empty_size) >> 2);
+		cont1_bytes = bytes;
+		min_bytes = block_group->sectorsize;
+	} else {
+		cont1_bytes = max(bytes, (bytes + empty_size) >> 2);
+		min_bytes = block_group->sectorsize;
+	}
+
+	spin_lock(&ctl->tree_lock);
+
+	/*
+	 * If we know we don't have enough space to make a cluster don't even
+	 * bother doing all the work to try and find one.
+	 */
+	if (ctl->free_space < bytes) {
+		spin_unlock(&ctl->tree_lock);
+		return -ENOSPC;
+	}
 
-	spin_lock(&block_group->tree_lock);
 	spin_lock(&cluster->lock);
 
 	/* someone already found a cluster, hooray */
@@ -1213,153 +2731,547 @@ int btrfs_find_space_cluster(struct btrfs_trans_handle *trans,
 		ret = 0;
 		goto out;
 	}
-again:
-	entry = tree_search_offset(block_group, offset, found_bitmap, 1);
-	if (!entry) {
-		ret = -ENOSPC;
-		goto out;
+
+	trace_btrfs_find_cluster(block_group, offset, bytes, empty_size,
+				 min_bytes);
+
+	INIT_LIST_HEAD(&bitmaps);
+	ret = setup_cluster_no_bitmap(block_group, cluster, &bitmaps, offset,
+				      bytes + empty_size,
+				      cont1_bytes, min_bytes);
+	if (ret)
+		ret = setup_cluster_bitmap(block_group, cluster, &bitmaps,
+					   offset, bytes + empty_size,
+					   cont1_bytes, min_bytes);
+
+	/* Clear our temporary list */
+	list_for_each_entry_safe(entry, tmp, &bitmaps, list)
+		list_del_init(&entry->list);
+
+	if (!ret) {
+		atomic_inc(&block_group->count);
+		list_add_tail(&cluster->block_group_list,
+			      &block_group->cluster_list);
+		cluster->block_group = block_group;
+	} else {
+		trace_btrfs_failed_cluster_setup(block_group);
 	}
+out:
+	spin_unlock(&cluster->lock);
+	spin_unlock(&ctl->tree_lock);
 
-	/*
-	 * If found_bitmap is true, we exhausted our search for extent entries,
-	 * and we just want to search all of the bitmaps that we can find, and
-	 * ignore any extent entries we find.
-	 */
-	while (entry->bitmap || found_bitmap ||
-	       (!entry->bitmap && entry->bytes < min_bytes)) {
-		struct rb_node *node = rb_next(&entry->offset_index);
-
-		if (entry->bitmap && entry->bytes > bytes + empty_size) {
-			ret = btrfs_bitmap_cluster(block_group, entry, cluster,
-						   offset, bytes + empty_size,
-						   min_bytes);
-			if (!ret)
-				goto got_it;
+	return ret;
+}
+
+/*
+ * simple code to zero out a cluster
+ */
+void btrfs_init_free_cluster(struct btrfs_free_cluster *cluster)
+{
+	spin_lock_init(&cluster->lock);
+	spin_lock_init(&cluster->refill_lock);
+	cluster->root = RB_ROOT;
+	cluster->max_size = 0;
+	INIT_LIST_HEAD(&cluster->block_group_list);
+	cluster->block_group = NULL;
+}
+
+static int do_trimming(struct btrfs_block_group_cache *block_group,
+		       u64 *total_trimmed, u64 start, u64 bytes,
+		       u64 reserved_start, u64 reserved_bytes)
+{
+	struct btrfs_space_info *space_info = block_group->space_info;
+	struct btrfs_fs_info *fs_info = block_group->fs_info;
+	int ret;
+	int update = 0;
+	u64 trimmed = 0;
+
+	spin_lock(&space_info->lock);
+	spin_lock(&block_group->lock);
+	if (!block_group->ro) {
+		block_group->reserved += reserved_bytes;
+		space_info->bytes_reserved += reserved_bytes;
+		update = 1;
+	}
+	spin_unlock(&block_group->lock);
+	spin_unlock(&space_info->lock);
+
+	ret = btrfs_error_discard_extent(fs_info->extent_root,
+					 start, bytes, &trimmed);
+	if (!ret)
+		*total_trimmed += trimmed;
+
+	btrfs_add_free_space(block_group, reserved_start, reserved_bytes);
+
+	if (update) {
+		spin_lock(&space_info->lock);
+		spin_lock(&block_group->lock);
+		if (block_group->ro)
+			space_info->bytes_readonly += reserved_bytes;
+		block_group->reserved -= reserved_bytes;
+		space_info->bytes_reserved -= reserved_bytes;
+		spin_unlock(&space_info->lock);
+		spin_unlock(&block_group->lock);
+	}
+
+	return ret;
+}
+
+static int trim_no_bitmap(struct btrfs_block_group_cache *block_group,
+			  u64 *total_trimmed, u64 start, u64 end, u64 minlen)
+{
+	struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
+	struct btrfs_free_space *entry;
+	struct rb_node *node;
+	int ret = 0;
+	u64 extent_start;
+	u64 extent_bytes;
+	u64 bytes;
+
+	while (start < end) {
+		spin_lock(&ctl->tree_lock);
+
+		if (ctl->free_space < minlen) {
+			spin_unlock(&ctl->tree_lock);
+			break;
 		}
 
-		if (!node) {
-			ret = -ENOSPC;
-			goto out;
+		entry = tree_search_offset(ctl, start, 0, 1);
+		if (!entry) {
+			spin_unlock(&ctl->tree_lock);
+			break;
 		}
-		entry = rb_entry(node, struct btrfs_free_space, offset_index);
-	}
 
-	/*
-	 * We already searched all the extent entries from the passed in offset
-	 * to the end and didn't find enough space for the cluster, and we also
-	 * didn't find any bitmaps that met our criteria, just go ahead and exit
-	 */
-	if (found_bitmap) {
-		ret = -ENOSPC;
-		goto out;
+		/* skip bitmaps */
+		while (entry->bitmap) {
+			node = rb_next(&entry->offset_index);
+			if (!node) {
+				spin_unlock(&ctl->tree_lock);
+				goto out;
+			}
+			entry = rb_entry(node, struct btrfs_free_space,
+					 offset_index);
+		}
+
+		if (entry->offset >= end) {
+			spin_unlock(&ctl->tree_lock);
+			break;
+		}
+
+		extent_start = entry->offset;
+		extent_bytes = entry->bytes;
+		start = max(start, extent_start);
+		bytes = min(extent_start + extent_bytes, end) - start;
+		if (bytes < minlen) {
+			spin_unlock(&ctl->tree_lock);
+			goto next;
+		}
+
+		unlink_free_space(ctl, entry);
+		kmem_cache_free(btrfs_free_space_cachep, entry);
+
+		spin_unlock(&ctl->tree_lock);
+
+		ret = do_trimming(block_group, total_trimmed, start, bytes,
+				  extent_start, extent_bytes);
+		if (ret)
+			break;
+next:
+		start += bytes;
+
+		if (fatal_signal_pending(current)) {
+			ret = -ERESTARTSYS;
+			break;
+		}
+
+		cond_resched();
 	}
+out:
+	return ret;
+}
 
-	cluster->points_to_bitmap = false;
-	window_start = entry->offset;
-	window_free = entry->bytes;
-	last = entry;
-	max_extent = entry->bytes;
+static int trim_bitmaps(struct btrfs_block_group_cache *block_group,
+			u64 *total_trimmed, u64 start, u64 end, u64 minlen)
+{
+	struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
+	struct btrfs_free_space *entry;
+	int ret = 0;
+	int ret2;
+	u64 bytes;
+	u64 offset = offset_to_bitmap(ctl, start);
 
-	while (1) {
-		/* out window is just right, lets fill it */
-		if (window_free >= bytes + empty_size)
+	while (offset < end) {
+		bool next_bitmap = false;
+
+		spin_lock(&ctl->tree_lock);
+
+		if (ctl->free_space < minlen) {
+			spin_unlock(&ctl->tree_lock);
 			break;
+		}
 
-		node = rb_next(&last->offset_index);
-		if (!node) {
-			if (found_bitmap)
-				goto again;
-			ret = -ENOSPC;
-			goto out;
+		entry = tree_search_offset(ctl, offset, 1, 0);
+		if (!entry) {
+			spin_unlock(&ctl->tree_lock);
+			next_bitmap = true;
+			goto next;
 		}
-		next = rb_entry(node, struct btrfs_free_space, offset_index);
 
-		/*
-		 * we found a bitmap, so if this search doesn't result in a
-		 * cluster, we know to go and search again for the bitmaps and
-		 * start looking for space there
-		 */
-		if (next->bitmap) {
-			if (!found_bitmap)
-				offset = next->offset;
-			found_bitmap = true;
-			last = next;
-			continue;
+		bytes = minlen;
+		ret2 = search_bitmap(ctl, entry, &start, &bytes);
+		if (ret2 || start >= end) {
+			spin_unlock(&ctl->tree_lock);
+			next_bitmap = true;
+			goto next;
 		}
 
-		/*
-		 * we haven't filled the empty size and the window is
-		 * very large.  reset and try again
-		 */
-		if (next->offset - (last->offset + last->bytes) > 128 * 1024 ||
-		    next->offset - window_start > (bytes + empty_size) * 2) {
-			entry = next;
-			window_start = entry->offset;
-			window_free = entry->bytes;
-			last = entry;
-			max_extent = entry->bytes;
+		bytes = min(bytes, end - start);
+		if (bytes < minlen) {
+			spin_unlock(&ctl->tree_lock);
+			goto next;
+		}
+
+		bitmap_clear_bits(ctl, entry, start, bytes);
+		if (entry->bytes == 0)
+			free_bitmap(ctl, entry);
+
+		spin_unlock(&ctl->tree_lock);
+
+		ret = do_trimming(block_group, total_trimmed, start, bytes,
+				  start, bytes);
+		if (ret)
+			break;
+next:
+		if (next_bitmap) {
+			offset += BITS_PER_BITMAP * ctl->unit;
 		} else {
-			last = next;
-			window_free += next->bytes;
-			if (entry->bytes > max_extent)
-				max_extent = entry->bytes;
+			start += bytes;
+			if (start >= offset + BITS_PER_BITMAP * ctl->unit)
+				offset += BITS_PER_BITMAP * ctl->unit;
 		}
+
+		if (fatal_signal_pending(current)) {
+			ret = -ERESTARTSYS;
+			break;
+		}
+
+		cond_resched();
+	}
+
+	return ret;
+}
+
+int btrfs_trim_block_group(struct btrfs_block_group_cache *block_group,
+			   u64 *trimmed, u64 start, u64 end, u64 minlen)
+{
+	int ret;
+
+	*trimmed = 0;
+
+	ret = trim_no_bitmap(block_group, trimmed, start, end, minlen);
+	if (ret)
+		return ret;
+
+	ret = trim_bitmaps(block_group, trimmed, start, end, minlen);
+
+	return ret;
+}
+
+/*
+ * Find the left-most item in the cache tree, and then return the
+ * smallest inode number in the item.
+ *
+ * Note: the returned inode number may not be the smallest one in
+ * the tree, if the left-most item is a bitmap.
+ */
+u64 btrfs_find_ino_for_alloc(struct btrfs_root *fs_root)
+{
+	struct btrfs_free_space_ctl *ctl = fs_root->free_ino_ctl;
+	struct btrfs_free_space *entry = NULL;
+	u64 ino = 0;
+
+	spin_lock(&ctl->tree_lock);
+
+	if (RB_EMPTY_ROOT(&ctl->free_space_offset))
+		goto out;
+
+	entry = rb_entry(rb_first(&ctl->free_space_offset),
+			 struct btrfs_free_space, offset_index);
+
+	if (!entry->bitmap) {
+		ino = entry->offset;
+
+		unlink_free_space(ctl, entry);
+		entry->offset++;
+		entry->bytes--;
+		if (!entry->bytes)
+			kmem_cache_free(btrfs_free_space_cachep, entry);
+		else
+			link_free_space(ctl, entry);
+	} else {
+		u64 offset = 0;
+		u64 count = 1;
+		int ret;
+
+		ret = search_bitmap(ctl, entry, &offset, &count);
+		/* Logic error; Should be empty if it can't find anything */
+		ASSERT(!ret);
+
+		ino = offset;
+		bitmap_clear_bits(ctl, entry, offset, 1);
+		if (entry->bytes == 0)
+			free_bitmap(ctl, entry);
 	}
+out:
+	spin_unlock(&ctl->tree_lock);
 
-	cluster->window_start = entry->offset;
+	return ino;
+}
+
+struct inode *lookup_free_ino_inode(struct btrfs_root *root,
+				    struct btrfs_path *path)
+{
+	struct inode *inode = NULL;
+
+	spin_lock(&root->cache_lock);
+	if (root->cache_inode)
+		inode = igrab(root->cache_inode);
+	spin_unlock(&root->cache_lock);
+	if (inode)
+		return inode;
+
+	inode = __lookup_free_space_inode(root, path, 0);
+	if (IS_ERR(inode))
+		return inode;
+
+	spin_lock(&root->cache_lock);
+	if (!btrfs_fs_closing(root->fs_info))
+		root->cache_inode = igrab(inode);
+	spin_unlock(&root->cache_lock);
+
+	return inode;
+}
+
+int create_free_ino_inode(struct btrfs_root *root,
+			  struct btrfs_trans_handle *trans,
+			  struct btrfs_path *path)
+{
+	return __create_free_space_inode(root, trans, path,
+					 BTRFS_FREE_INO_OBJECTID, 0);
+}
+
+int load_free_ino_cache(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
+{
+	struct btrfs_free_space_ctl *ctl = root->free_ino_ctl;
+	struct btrfs_path *path;
+	struct inode *inode;
+	int ret = 0;
+	u64 root_gen = btrfs_root_generation(&root->root_item);
+
+	if (!btrfs_test_opt(root, INODE_MAP_CACHE))
+		return 0;
 
 	/*
-	 * now we've found our entries, pull them out of the free space
-	 * cache and put them into the cluster rbtree
-	 *
-	 * The cluster includes an rbtree, but only uses the offset index
-	 * of each free space cache entry.
+	 * If we're unmounting then just return, since this does a search on the
+	 * normal root and not the commit root and we could deadlock.
 	 */
-	while (1) {
-		node = rb_next(&entry->offset_index);
-		if (entry->bitmap && node) {
-			entry = rb_entry(node, struct btrfs_free_space,
-					 offset_index);
-			continue;
-		} else if (entry->bitmap && !node) {
-			break;
-		}
+	if (btrfs_fs_closing(fs_info))
+		return 0;
 
-		rb_erase(&entry->offset_index, &block_group->free_space_offset);
-		ret = tree_insert_offset(&cluster->root, entry->offset,
-					 &entry->offset_index, 0);
-		BUG_ON(ret);
+	path = btrfs_alloc_path();
+	if (!path)
+		return 0;
 
-		if (!node || entry == last)
-			break;
+	inode = lookup_free_ino_inode(root, path);
+	if (IS_ERR(inode))
+		goto out;
 
-		entry = rb_entry(node, struct btrfs_free_space, offset_index);
-	}
+	if (root_gen != BTRFS_I(inode)->generation)
+		goto out_put;
 
-	cluster->max_size = max_extent;
-got_it:
-	ret = 0;
-	atomic_inc(&block_group->count);
-	list_add_tail(&cluster->block_group_list, &block_group->cluster_list);
-	cluster->block_group = block_group;
+	ret = __load_free_space_cache(root, inode, ctl, path, 0);
+
+	if (ret < 0)
+		btrfs_err(fs_info,
+			"failed to load free ino cache for root %llu",
+			root->root_key.objectid);
+out_put:
+	iput(inode);
 out:
-	spin_unlock(&cluster->lock);
-	spin_unlock(&block_group->tree_lock);
+	btrfs_free_path(path);
+	return ret;
+}
+
+int btrfs_write_out_ino_cache(struct btrfs_root *root,
+			      struct btrfs_trans_handle *trans,
+			      struct btrfs_path *path,
+			      struct inode *inode)
+{
+	struct btrfs_free_space_ctl *ctl = root->free_ino_ctl;
+	int ret;
+
+	if (!btrfs_test_opt(root, INODE_MAP_CACHE))
+		return 0;
+
+	ret = __btrfs_write_out_cache(root, inode, ctl, NULL, trans, path, 0);
+	if (ret) {
+		btrfs_delalloc_release_metadata(inode, inode->i_size);
+#ifdef DEBUG
+		btrfs_err(root->fs_info,
+			"failed to write free ino cache for root %llu",
+			root->root_key.objectid);
+#endif
+	}
 
 	return ret;
 }
 
+#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
 /*
- * simple code to zero out a cluster
+ * Use this if you need to make a bitmap or extent entry specifically, it
+ * doesn't do any of the merging that add_free_space does, this acts a lot like
+ * how the free space cache loading stuff works, so you can get really weird
+ * configurations.
  */
-void btrfs_init_free_cluster(struct btrfs_free_cluster *cluster)
+int test_add_free_space_entry(struct btrfs_block_group_cache *cache,
+			      u64 offset, u64 bytes, bool bitmap)
 {
-	spin_lock_init(&cluster->lock);
-	spin_lock_init(&cluster->refill_lock);
-	cluster->root = RB_ROOT;
-	cluster->max_size = 0;
-	cluster->points_to_bitmap = false;
-	INIT_LIST_HEAD(&cluster->block_group_list);
-	cluster->block_group = NULL;
+	struct btrfs_free_space_ctl *ctl = cache->free_space_ctl;
+	struct btrfs_free_space *info = NULL, *bitmap_info;
+	void *map = NULL;
+	u64 bytes_added;
+	int ret;
+
+again:
+	if (!info) {
+		info = kmem_cache_zalloc(btrfs_free_space_cachep, GFP_NOFS);
+		if (!info)
+			return -ENOMEM;
+	}
+
+	if (!bitmap) {
+		spin_lock(&ctl->tree_lock);
+		info->offset = offset;
+		info->bytes = bytes;
+		ret = link_free_space(ctl, info);
+		spin_unlock(&ctl->tree_lock);
+		if (ret)
+			kmem_cache_free(btrfs_free_space_cachep, info);
+		return ret;
+	}
+
+	if (!map) {
+		map = kzalloc(PAGE_CACHE_SIZE, GFP_NOFS);
+		if (!map) {
+			kmem_cache_free(btrfs_free_space_cachep, info);
+			return -ENOMEM;
+		}
+	}
+
+	spin_lock(&ctl->tree_lock);
+	bitmap_info = tree_search_offset(ctl, offset_to_bitmap(ctl, offset),
+					 1, 0);
+	if (!bitmap_info) {
+		info->bitmap = map;
+		map = NULL;
+		add_new_bitmap(ctl, info, offset);
+		bitmap_info = info;
+	}
+
+	bytes_added = add_bytes_to_bitmap(ctl, bitmap_info, offset, bytes);
+	bytes -= bytes_added;
+	offset += bytes_added;
+	spin_unlock(&ctl->tree_lock);
+
+	if (bytes)
+		goto again;
+
+	if (map)
+		kfree(map);
+	return 0;
 }
 
+/*
+ * Checks to see if the given range is in the free space cache.  This is really
+ * just used to check the absence of space, so if there is free space in the
+ * range at all we will return 1.
+ */
+int test_check_exists(struct btrfs_block_group_cache *cache,
+		      u64 offset, u64 bytes)
+{
+	struct btrfs_free_space_ctl *ctl = cache->free_space_ctl;
+	struct btrfs_free_space *info;
+	int ret = 0;
+
+	spin_lock(&ctl->tree_lock);
+	info = tree_search_offset(ctl, offset, 0, 0);
+	if (!info) {
+		info = tree_search_offset(ctl, offset_to_bitmap(ctl, offset),
+					  1, 0);
+		if (!info)
+			goto out;
+	}
+
+have_info:
+	if (info->bitmap) {
+		u64 bit_off, bit_bytes;
+		struct rb_node *n;
+		struct btrfs_free_space *tmp;
+
+		bit_off = offset;
+		bit_bytes = ctl->unit;
+		ret = search_bitmap(ctl, info, &bit_off, &bit_bytes);
+		if (!ret) {
+			if (bit_off == offset) {
+				ret = 1;
+				goto out;
+			} else if (bit_off > offset &&
+				   offset + bytes > bit_off) {
+				ret = 1;
+				goto out;
+			}
+		}
+
+		n = rb_prev(&info->offset_index);
+		while (n) {
+			tmp = rb_entry(n, struct btrfs_free_space,
+				       offset_index);
+			if (tmp->offset + tmp->bytes < offset)
+				break;
+			if (offset + bytes < tmp->offset) {
+				n = rb_prev(&info->offset_index);
+				continue;
+			}
+			info = tmp;
+			goto have_info;
+		}
+
+		n = rb_next(&info->offset_index);
+		while (n) {
+			tmp = rb_entry(n, struct btrfs_free_space,
+				       offset_index);
+			if (offset + bytes < tmp->offset)
+				break;
+			if (tmp->offset + tmp->bytes < offset) {
+				n = rb_next(&info->offset_index);
+				continue;
+			}
+			info = tmp;
+			goto have_info;
+		}
+
+		goto out;
+	}
+
+	if (info->offset == offset) {
+		ret = 1;
+		goto out;
+	}
+
+	if (offset > info->offset && offset < info->offset + info->bytes)
+		ret = 1;
+out:
+	spin_unlock(&ctl->tree_lock);
+	return ret;
+}
+#endif /* CONFIG_BTRFS_FS_RUN_SANITY_TESTS */
diff --git a/fs/btrfs/free-space-cache.h b/fs/btrfs/free-space-cache.h
index 890a8e79011..0cf4977ef70 100644
--- a/fs/btrfs/free-space-cache.h
+++ b/fs/btrfs/free-space-cache.h
@@ -27,27 +27,98 @@ struct btrfs_free_space {
 	struct list_head list;
 };
 
-int btrfs_add_free_space(struct btrfs_block_group_cache *block_group,
-			 u64 bytenr, u64 size);
+struct btrfs_free_space_ctl {
+	spinlock_t tree_lock;
+	struct rb_root free_space_offset;
+	u64 free_space;
+	int extents_thresh;
+	int free_extents;
+	int total_bitmaps;
+	int unit;
+	u64 start;
+	struct btrfs_free_space_op *op;
+	void *private;
+};
+
+struct btrfs_free_space_op {
+	void (*recalc_thresholds)(struct btrfs_free_space_ctl *ctl);
+	bool (*use_bitmap)(struct btrfs_free_space_ctl *ctl,
+			   struct btrfs_free_space *info);
+};
+
+struct inode *lookup_free_space_inode(struct btrfs_root *root,
+				      struct btrfs_block_group_cache
+				      *block_group, struct btrfs_path *path);
+int create_free_space_inode(struct btrfs_root *root,
+			    struct btrfs_trans_handle *trans,
+			    struct btrfs_block_group_cache *block_group,
+			    struct btrfs_path *path);
+
+int btrfs_check_trunc_cache_free_space(struct btrfs_root *root,
+				       struct btrfs_block_rsv *rsv);
+int btrfs_truncate_free_space_cache(struct btrfs_root *root,
+				    struct btrfs_trans_handle *trans,
+				    struct inode *inode);
+int load_free_space_cache(struct btrfs_fs_info *fs_info,
+			  struct btrfs_block_group_cache *block_group);
+int btrfs_write_out_cache(struct btrfs_root *root,
+			  struct btrfs_trans_handle *trans,
+			  struct btrfs_block_group_cache *block_group,
+			  struct btrfs_path *path);
+
+struct inode *lookup_free_ino_inode(struct btrfs_root *root,
+				    struct btrfs_path *path);
+int create_free_ino_inode(struct btrfs_root *root,
+			  struct btrfs_trans_handle *trans,
+			  struct btrfs_path *path);
+int load_free_ino_cache(struct btrfs_fs_info *fs_info,
+			struct btrfs_root *root);
+int btrfs_write_out_ino_cache(struct btrfs_root *root,
+			      struct btrfs_trans_handle *trans,
+			      struct btrfs_path *path,
+			      struct inode *inode);
+
+void btrfs_init_free_space_ctl(struct btrfs_block_group_cache *block_group);
+int __btrfs_add_free_space(struct btrfs_free_space_ctl *ctl,
+			   u64 bytenr, u64 size);
+static inline int
+btrfs_add_free_space(struct btrfs_block_group_cache *block_group,
+		     u64 bytenr, u64 size)
+{
+	return __btrfs_add_free_space(block_group->free_space_ctl,
+				      bytenr, size);
+}
 int btrfs_remove_free_space(struct btrfs_block_group_cache *block_group,
 			    u64 bytenr, u64 size);
+void __btrfs_remove_free_space_cache(struct btrfs_free_space_ctl *ctl);
 void btrfs_remove_free_space_cache(struct btrfs_block_group_cache
-				   *block_group);
+				     *block_group);
 u64 btrfs_find_space_for_alloc(struct btrfs_block_group_cache *block_group,
-			       u64 offset, u64 bytes, u64 empty_size);
+			       u64 offset, u64 bytes, u64 empty_size,
+			       u64 *max_extent_size);
+u64 btrfs_find_ino_for_alloc(struct btrfs_root *fs_root);
 void btrfs_dump_free_space(struct btrfs_block_group_cache *block_group,
 			   u64 bytes);
-u64 btrfs_block_group_free_space(struct btrfs_block_group_cache *block_group);
-int btrfs_find_space_cluster(struct btrfs_trans_handle *trans,
-			     struct btrfs_root *root,
+int btrfs_find_space_cluster(struct btrfs_root *root,
 			     struct btrfs_block_group_cache *block_group,
 			     struct btrfs_free_cluster *cluster,
 			     u64 offset, u64 bytes, u64 empty_size);
 void btrfs_init_free_cluster(struct btrfs_free_cluster *cluster);
 u64 btrfs_alloc_from_cluster(struct btrfs_block_group_cache *block_group,
 			     struct btrfs_free_cluster *cluster, u64 bytes,
-			     u64 min_start);
+			     u64 min_start, u64 *max_extent_size);
 int btrfs_return_cluster_to_free_space(
 			       struct btrfs_block_group_cache *block_group,
 			       struct btrfs_free_cluster *cluster);
+int btrfs_trim_block_group(struct btrfs_block_group_cache *block_group,
+			   u64 *trimmed, u64 start, u64 end, u64 minlen);
+
+/* Support functions for runnint our sanity tests */
+#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
+int test_add_free_space_entry(struct btrfs_block_group_cache *cache,
+			      u64 offset, u64 bytes, bool bitmap);
+int test_check_exists(struct btrfs_block_group_cache *cache,
+		      u64 offset, u64 bytes);
+#endif
+
 #endif
diff --git a/fs/btrfs/hash.c b/fs/btrfs/hash.c
new file mode 100644
index 00000000000..85889aa82c6
--- /dev/null
+++ b/fs/btrfs/hash.c
@@ -0,0 +1,50 @@
+/*
+ * Copyright (C) 2014 Filipe David Borba Manana <fdmanana@gmail.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ */
+
+#include <crypto/hash.h>
+#include <linux/err.h>
+#include "hash.h"
+
+static struct crypto_shash *tfm;
+
+int __init btrfs_hash_init(void)
+{
+	tfm = crypto_alloc_shash("crc32c", 0, 0);
+	if (IS_ERR(tfm))
+		return PTR_ERR(tfm);
+
+	return 0;
+}
+
+void btrfs_hash_exit(void)
+{
+	crypto_free_shash(tfm);
+}
+
+u32 btrfs_crc32c(u32 crc, const void *address, unsigned int length)
+{
+	struct {
+		struct shash_desc shash;
+		char ctx[crypto_shash_descsize(tfm)];
+	} desc;
+	int err;
+
+	desc.shash.tfm = tfm;
+	desc.shash.flags = 0;
+	*(u32 *)desc.ctx = crc;
+
+	err = crypto_shash_update(&desc.shash, address, length);
+	BUG_ON(err);
+
+	return *(u32 *)desc.ctx;
+}
diff --git a/fs/btrfs/hash.h b/fs/btrfs/hash.h
index db2ff9773b9..118a2316e5d 100644
--- a/fs/btrfs/hash.h
+++ b/fs/btrfs/hash.h
@@ -19,9 +19,24 @@
 #ifndef __HASH__
 #define __HASH__
 
-#include <linux/crc32c.h>
+int __init btrfs_hash_init(void);
+
+void btrfs_hash_exit(void);
+
+u32 btrfs_crc32c(u32 crc, const void *address, unsigned int length);
+
 static inline u64 btrfs_name_hash(const char *name, int len)
 {
-	return crc32c((u32)~1, name, len);
+	return btrfs_crc32c((u32)~1, name, len);
 }
+
+/*
+ * Figure the key offset of an extended inode ref
+ */
+static inline u64 btrfs_extref_hash(u64 parent_objectid, const char *name,
+				    int len)
+{
+	return (u64) btrfs_crc32c(parent_objectid, name, len);
+}
+
 #endif
diff --git a/fs/btrfs/inode-item.c b/fs/btrfs/inode-item.c
index 72ce3c173d6..2be38df703c 100644
--- a/fs/btrfs/inode-item.c
+++ b/fs/btrfs/inode-item.c
@@ -18,7 +18,9 @@
 
 #include "ctree.h"
 #include "disk-io.h"
+#include "hash.h"
 #include "transaction.h"
+#include "print-tree.h"
 
 static int find_name_in_backref(struct btrfs_path *path, const char *name,
 			 int name_len, struct btrfs_inode_ref **ref_ret)
@@ -49,10 +51,149 @@ static int find_name_in_backref(struct btrfs_path *path, const char *name,
 	return 0;
 }
 
+int btrfs_find_name_in_ext_backref(struct btrfs_path *path, u64 ref_objectid,
+				   const char *name, int name_len,
+				   struct btrfs_inode_extref **extref_ret)
+{
+	struct extent_buffer *leaf;
+	struct btrfs_inode_extref *extref;
+	unsigned long ptr;
+	unsigned long name_ptr;
+	u32 item_size;
+	u32 cur_offset = 0;
+	int ref_name_len;
+
+	leaf = path->nodes[0];
+	item_size = btrfs_item_size_nr(leaf, path->slots[0]);
+	ptr = btrfs_item_ptr_offset(leaf, path->slots[0]);
+
+	/*
+	 * Search all extended backrefs in this item. We're only
+	 * looking through any collisions so most of the time this is
+	 * just going to compare against one buffer. If all is well,
+	 * we'll return success and the inode ref object.
+	 */
+	while (cur_offset < item_size) {
+		extref = (struct btrfs_inode_extref *) (ptr + cur_offset);
+		name_ptr = (unsigned long)(&extref->name);
+		ref_name_len = btrfs_inode_extref_name_len(leaf, extref);
+
+		if (ref_name_len == name_len &&
+		    btrfs_inode_extref_parent(leaf, extref) == ref_objectid &&
+		    (memcmp_extent_buffer(leaf, name, name_ptr, name_len) == 0)) {
+			if (extref_ret)
+				*extref_ret = extref;
+			return 1;
+		}
+
+		cur_offset += ref_name_len + sizeof(*extref);
+	}
+	return 0;
+}
+
+/* Returns NULL if no extref found */
+struct btrfs_inode_extref *
+btrfs_lookup_inode_extref(struct btrfs_trans_handle *trans,
+			  struct btrfs_root *root,
+			  struct btrfs_path *path,
+			  const char *name, int name_len,
+			  u64 inode_objectid, u64 ref_objectid, int ins_len,
+			  int cow)
+{
+	int ret;
+	struct btrfs_key key;
+	struct btrfs_inode_extref *extref;
+
+	key.objectid = inode_objectid;
+	key.type = BTRFS_INODE_EXTREF_KEY;
+	key.offset = btrfs_extref_hash(ref_objectid, name, name_len);
+
+	ret = btrfs_search_slot(trans, root, &key, path, ins_len, cow);
+	if (ret < 0)
+		return ERR_PTR(ret);
+	if (ret > 0)
+		return NULL;
+	if (!btrfs_find_name_in_ext_backref(path, ref_objectid, name, name_len, &extref))
+		return NULL;
+	return extref;
+}
+
+static int btrfs_del_inode_extref(struct btrfs_trans_handle *trans,
+				  struct btrfs_root *root,
+				  const char *name, int name_len,
+				  u64 inode_objectid, u64 ref_objectid,
+				  u64 *index)
+{
+	struct btrfs_path *path;
+	struct btrfs_key key;
+	struct btrfs_inode_extref *extref;
+	struct extent_buffer *leaf;
+	int ret;
+	int del_len = name_len + sizeof(*extref);
+	unsigned long ptr;
+	unsigned long item_start;
+	u32 item_size;
+
+	key.objectid = inode_objectid;
+	btrfs_set_key_type(&key, BTRFS_INODE_EXTREF_KEY);
+	key.offset = btrfs_extref_hash(ref_objectid, name, name_len);
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	path->leave_spinning = 1;
+
+	ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
+	if (ret > 0)
+		ret = -ENOENT;
+	if (ret < 0)
+		goto out;
+
+	/*
+	 * Sanity check - did we find the right item for this name?
+	 * This should always succeed so error here will make the FS
+	 * readonly.
+	 */
+	if (!btrfs_find_name_in_ext_backref(path, ref_objectid,
+					    name, name_len, &extref)) {
+		btrfs_std_error(root->fs_info, -ENOENT);
+		ret = -EROFS;
+		goto out;
+	}
+
+	leaf = path->nodes[0];
+	item_size = btrfs_item_size_nr(leaf, path->slots[0]);
+	if (index)
+		*index = btrfs_inode_extref_index(leaf, extref);
+
+	if (del_len == item_size) {
+		/*
+		 * Common case only one ref in the item, remove the
+		 * whole item.
+		 */
+		ret = btrfs_del_item(trans, root, path);
+		goto out;
+	}
+
+	ptr = (unsigned long)extref;
+	item_start = btrfs_item_ptr_offset(leaf, path->slots[0]);
+
+	memmove_extent_buffer(leaf, ptr, ptr + del_len,
+			      item_size - (ptr + del_len - item_start));
+
+	btrfs_truncate_item(root, path, item_size - del_len, 1);
+
+out:
+	btrfs_free_path(path);
+
+	return ret;
+}
+
 int btrfs_del_inode_ref(struct btrfs_trans_handle *trans,
-			   struct btrfs_root *root,
-			   const char *name, int name_len,
-			   u64 inode_objectid, u64 ref_objectid, u64 *index)
+			struct btrfs_root *root,
+			const char *name, int name_len,
+			u64 inode_objectid, u64 ref_objectid, u64 *index)
 {
 	struct btrfs_path *path;
 	struct btrfs_key key;
@@ -63,6 +204,7 @@ int btrfs_del_inode_ref(struct btrfs_trans_handle *trans,
 	u32 item_size;
 	u32 sub_item_len;
 	int ret;
+	int search_ext_refs = 0;
 	int del_len = name_len + sizeof(*ref);
 
 	key.objectid = inode_objectid;
@@ -78,12 +220,14 @@ int btrfs_del_inode_ref(struct btrfs_trans_handle *trans,
 	ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
 	if (ret > 0) {
 		ret = -ENOENT;
+		search_ext_refs = 1;
 		goto out;
 	} else if (ret < 0) {
 		goto out;
 	}
 	if (!find_name_in_backref(path, name, name_len, &ref)) {
 		ret = -ENOENT;
+		search_ext_refs = 1;
 		goto out;
 	}
 	leaf = path->nodes[0];
@@ -101,14 +245,84 @@ int btrfs_del_inode_ref(struct btrfs_trans_handle *trans,
 	item_start = btrfs_item_ptr_offset(leaf, path->slots[0]);
 	memmove_extent_buffer(leaf, ptr, ptr + sub_item_len,
 			      item_size - (ptr + sub_item_len - item_start));
-	ret = btrfs_truncate_item(trans, root, path,
-				  item_size - sub_item_len, 1);
-	BUG_ON(ret);
+	btrfs_truncate_item(root, path, item_size - sub_item_len, 1);
+out:
+	btrfs_free_path(path);
+
+	if (search_ext_refs) {
+		/*
+		 * No refs were found, or we could not find the
+		 * name in our ref array. Find and remove the extended
+		 * inode ref then.
+		 */
+		return btrfs_del_inode_extref(trans, root, name, name_len,
+					      inode_objectid, ref_objectid, index);
+	}
+
+	return ret;
+}
+
+/*
+ * btrfs_insert_inode_extref() - Inserts an extended inode ref into a tree.
+ *
+ * The caller must have checked against BTRFS_LINK_MAX already.
+ */
+static int btrfs_insert_inode_extref(struct btrfs_trans_handle *trans,
+				     struct btrfs_root *root,
+				     const char *name, int name_len,
+				     u64 inode_objectid, u64 ref_objectid, u64 index)
+{
+	struct btrfs_inode_extref *extref;
+	int ret;
+	int ins_len = name_len + sizeof(*extref);
+	unsigned long ptr;
+	struct btrfs_path *path;
+	struct btrfs_key key;
+	struct extent_buffer *leaf;
+	struct btrfs_item *item;
+
+	key.objectid = inode_objectid;
+	key.type = BTRFS_INODE_EXTREF_KEY;
+	key.offset = btrfs_extref_hash(ref_objectid, name, name_len);
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	path->leave_spinning = 1;
+	ret = btrfs_insert_empty_item(trans, root, path, &key,
+				      ins_len);
+	if (ret == -EEXIST) {
+		if (btrfs_find_name_in_ext_backref(path, ref_objectid,
+						   name, name_len, NULL))
+			goto out;
+
+		btrfs_extend_item(root, path, ins_len);
+		ret = 0;
+	}
+	if (ret < 0)
+		goto out;
+
+	leaf = path->nodes[0];
+	item = btrfs_item_nr(path->slots[0]);
+	ptr = (unsigned long)btrfs_item_ptr(leaf, path->slots[0], char);
+	ptr += btrfs_item_size(leaf, item) - ins_len;
+	extref = (struct btrfs_inode_extref *)ptr;
+
+	btrfs_set_inode_extref_name_len(path->nodes[0], extref, name_len);
+	btrfs_set_inode_extref_index(path->nodes[0], extref, index);
+	btrfs_set_inode_extref_parent(path->nodes[0], extref, ref_objectid);
+
+	ptr = (unsigned long)&extref->name;
+	write_extent_buffer(path->nodes[0], name, ptr, name_len);
+	btrfs_mark_buffer_dirty(path->nodes[0]);
+
 out:
 	btrfs_free_path(path);
 	return ret;
 }
 
+/* Will return 0, -ENOMEM, -EMLINK, or -EEXIST or anything from the CoW path */
 int btrfs_insert_inode_ref(struct btrfs_trans_handle *trans,
 			   struct btrfs_root *root,
 			   const char *name, int name_len,
@@ -139,8 +353,7 @@ int btrfs_insert_inode_ref(struct btrfs_trans_handle *trans,
 			goto out;
 
 		old_size = btrfs_item_size_nr(path->nodes[0], path->slots[0]);
-		ret = btrfs_extend_item(trans, root, path, ins_len);
-		BUG_ON(ret);
+		btrfs_extend_item(root, path, ins_len);
 		ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
 				     struct btrfs_inode_ref);
 		ref = (struct btrfs_inode_ref *)((unsigned long)ref + old_size);
@@ -164,6 +377,19 @@ int btrfs_insert_inode_ref(struct btrfs_trans_handle *trans,
 
 out:
 	btrfs_free_path(path);
+
+	if (ret == -EMLINK) {
+		struct btrfs_super_block *disk_super = root->fs_info->super_copy;
+		/* We ran out of space in the ref array. Need to
+		 * add an extended ref. */
+		if (btrfs_super_incompat_flags(disk_super)
+		    & BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF)
+			ret = btrfs_insert_inode_extref(trans, root, name,
+							name_len,
+							inode_objectid,
+							ref_objectid, index);
+	}
+
 	return ret;
 }
 
diff --git a/fs/btrfs/inode-map.c b/fs/btrfs/inode-map.c
index c56eb590917..888fbe19079 100644
--- a/fs/btrfs/inode-map.c
+++ b/fs/btrfs/inode-map.c
@@ -16,11 +16,497 @@
  * Boston, MA 021110-1307, USA.
  */
 
+#include <linux/delay.h>
+#include <linux/kthread.h>
+#include <linux/pagemap.h>
+
 #include "ctree.h"
 #include "disk-io.h"
+#include "free-space-cache.h"
+#include "inode-map.h"
 #include "transaction.h"
 
-int btrfs_find_highest_inode(struct btrfs_root *root, u64 *objectid)
+static int caching_kthread(void *data)
+{
+	struct btrfs_root *root = data;
+	struct btrfs_fs_info *fs_info = root->fs_info;
+	struct btrfs_free_space_ctl *ctl = root->free_ino_ctl;
+	struct btrfs_key key;
+	struct btrfs_path *path;
+	struct extent_buffer *leaf;
+	u64 last = (u64)-1;
+	int slot;
+	int ret;
+
+	if (!btrfs_test_opt(root, INODE_MAP_CACHE))
+		return 0;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	/* Since the commit root is read-only, we can safely skip locking. */
+	path->skip_locking = 1;
+	path->search_commit_root = 1;
+	path->reada = 2;
+
+	key.objectid = BTRFS_FIRST_FREE_OBJECTID;
+	key.offset = 0;
+	key.type = BTRFS_INODE_ITEM_KEY;
+again:
+	/* need to make sure the commit_root doesn't disappear */
+	down_read(&fs_info->commit_root_sem);
+
+	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+	if (ret < 0)
+		goto out;
+
+	while (1) {
+		if (btrfs_fs_closing(fs_info))
+			goto out;
+
+		leaf = path->nodes[0];
+		slot = path->slots[0];
+		if (slot >= btrfs_header_nritems(leaf)) {
+			ret = btrfs_next_leaf(root, path);
+			if (ret < 0)
+				goto out;
+			else if (ret > 0)
+				break;
+
+			if (need_resched() ||
+			    btrfs_transaction_in_commit(fs_info)) {
+				leaf = path->nodes[0];
+
+				if (WARN_ON(btrfs_header_nritems(leaf) == 0))
+					break;
+
+				/*
+				 * Save the key so we can advances forward
+				 * in the next search.
+				 */
+				btrfs_item_key_to_cpu(leaf, &key, 0);
+				btrfs_release_path(path);
+				root->cache_progress = last;
+				up_read(&fs_info->commit_root_sem);
+				schedule_timeout(1);
+				goto again;
+			} else
+				continue;
+		}
+
+		btrfs_item_key_to_cpu(leaf, &key, slot);
+
+		if (key.type != BTRFS_INODE_ITEM_KEY)
+			goto next;
+
+		if (key.objectid >= root->highest_objectid)
+			break;
+
+		if (last != (u64)-1 && last + 1 != key.objectid) {
+			__btrfs_add_free_space(ctl, last + 1,
+					       key.objectid - last - 1);
+			wake_up(&root->cache_wait);
+		}
+
+		last = key.objectid;
+next:
+		path->slots[0]++;
+	}
+
+	if (last < root->highest_objectid - 1) {
+		__btrfs_add_free_space(ctl, last + 1,
+				       root->highest_objectid - last - 1);
+	}
+
+	spin_lock(&root->cache_lock);
+	root->cached = BTRFS_CACHE_FINISHED;
+	spin_unlock(&root->cache_lock);
+
+	root->cache_progress = (u64)-1;
+	btrfs_unpin_free_ino(root);
+out:
+	wake_up(&root->cache_wait);
+	up_read(&fs_info->commit_root_sem);
+
+	btrfs_free_path(path);
+
+	return ret;
+}
+
+static void start_caching(struct btrfs_root *root)
+{
+	struct btrfs_free_space_ctl *ctl = root->free_ino_ctl;
+	struct task_struct *tsk;
+	int ret;
+	u64 objectid;
+
+	if (!btrfs_test_opt(root, INODE_MAP_CACHE))
+		return;
+
+	spin_lock(&root->cache_lock);
+	if (root->cached != BTRFS_CACHE_NO) {
+		spin_unlock(&root->cache_lock);
+		return;
+	}
+
+	root->cached = BTRFS_CACHE_STARTED;
+	spin_unlock(&root->cache_lock);
+
+	ret = load_free_ino_cache(root->fs_info, root);
+	if (ret == 1) {
+		spin_lock(&root->cache_lock);
+		root->cached = BTRFS_CACHE_FINISHED;
+		spin_unlock(&root->cache_lock);
+		return;
+	}
+
+	/*
+	 * It can be quite time-consuming to fill the cache by searching
+	 * through the extent tree, and this can keep ino allocation path
+	 * waiting. Therefore at start we quickly find out the highest
+	 * inode number and we know we can use inode numbers which fall in
+	 * [highest_ino + 1, BTRFS_LAST_FREE_OBJECTID].
+	 */
+	ret = btrfs_find_free_objectid(root, &objectid);
+	if (!ret && objectid <= BTRFS_LAST_FREE_OBJECTID) {
+		__btrfs_add_free_space(ctl, objectid,
+				       BTRFS_LAST_FREE_OBJECTID - objectid + 1);
+	}
+
+	tsk = kthread_run(caching_kthread, root, "btrfs-ino-cache-%llu",
+			  root->root_key.objectid);
+	if (IS_ERR(tsk)) {
+		btrfs_warn(root->fs_info, "failed to start inode caching task");
+		btrfs_clear_and_info(root, CHANGE_INODE_CACHE,
+				"disabling inode map caching");
+	}
+}
+
+int btrfs_find_free_ino(struct btrfs_root *root, u64 *objectid)
+{
+	if (!btrfs_test_opt(root, INODE_MAP_CACHE))
+		return btrfs_find_free_objectid(root, objectid);
+
+again:
+	*objectid = btrfs_find_ino_for_alloc(root);
+
+	if (*objectid != 0)
+		return 0;
+
+	start_caching(root);
+
+	wait_event(root->cache_wait,
+		   root->cached == BTRFS_CACHE_FINISHED ||
+		   root->free_ino_ctl->free_space > 0);
+
+	if (root->cached == BTRFS_CACHE_FINISHED &&
+	    root->free_ino_ctl->free_space == 0)
+		return -ENOSPC;
+	else
+		goto again;
+}
+
+void btrfs_return_ino(struct btrfs_root *root, u64 objectid)
+{
+	struct btrfs_free_space_ctl *pinned = root->free_ino_pinned;
+
+	if (!btrfs_test_opt(root, INODE_MAP_CACHE))
+		return;
+again:
+	if (root->cached == BTRFS_CACHE_FINISHED) {
+		__btrfs_add_free_space(pinned, objectid, 1);
+	} else {
+		down_write(&root->fs_info->commit_root_sem);
+		spin_lock(&root->cache_lock);
+		if (root->cached == BTRFS_CACHE_FINISHED) {
+			spin_unlock(&root->cache_lock);
+			up_write(&root->fs_info->commit_root_sem);
+			goto again;
+		}
+		spin_unlock(&root->cache_lock);
+
+		start_caching(root);
+
+		__btrfs_add_free_space(pinned, objectid, 1);
+
+		up_write(&root->fs_info->commit_root_sem);
+	}
+}
+
+/*
+ * When a transaction is committed, we'll move those inode numbers which
+ * are smaller than root->cache_progress from pinned tree to free_ino tree,
+ * and others will just be dropped, because the commit root we were
+ * searching has changed.
+ *
+ * Must be called with root->fs_info->commit_root_sem held
+ */
+void btrfs_unpin_free_ino(struct btrfs_root *root)
+{
+	struct btrfs_free_space_ctl *ctl = root->free_ino_ctl;
+	struct rb_root *rbroot = &root->free_ino_pinned->free_space_offset;
+	struct btrfs_free_space *info;
+	struct rb_node *n;
+	u64 count;
+
+	if (!btrfs_test_opt(root, INODE_MAP_CACHE))
+		return;
+
+	while (1) {
+		n = rb_first(rbroot);
+		if (!n)
+			break;
+
+		info = rb_entry(n, struct btrfs_free_space, offset_index);
+		BUG_ON(info->bitmap); /* Logic error */
+
+		if (info->offset > root->cache_progress)
+			goto free;
+		else if (info->offset + info->bytes > root->cache_progress)
+			count = root->cache_progress - info->offset + 1;
+		else
+			count = info->bytes;
+
+		__btrfs_add_free_space(ctl, info->offset, count);
+free:
+		rb_erase(&info->offset_index, rbroot);
+		kfree(info);
+	}
+}
+
+#define INIT_THRESHOLD	(((1024 * 32) / 2) / sizeof(struct btrfs_free_space))
+#define INODES_PER_BITMAP (PAGE_CACHE_SIZE * 8)
+
+/*
+ * The goal is to keep the memory used by the free_ino tree won't
+ * exceed the memory if we use bitmaps only.
+ */
+static void recalculate_thresholds(struct btrfs_free_space_ctl *ctl)
+{
+	struct btrfs_free_space *info;
+	struct rb_node *n;
+	int max_ino;
+	int max_bitmaps;
+
+	n = rb_last(&ctl->free_space_offset);
+	if (!n) {
+		ctl->extents_thresh = INIT_THRESHOLD;
+		return;
+	}
+	info = rb_entry(n, struct btrfs_free_space, offset_index);
+
+	/*
+	 * Find the maximum inode number in the filesystem. Note we
+	 * ignore the fact that this can be a bitmap, because we are
+	 * not doing precise calculation.
+	 */
+	max_ino = info->bytes - 1;
+
+	max_bitmaps = ALIGN(max_ino, INODES_PER_BITMAP) / INODES_PER_BITMAP;
+	if (max_bitmaps <= ctl->total_bitmaps) {
+		ctl->extents_thresh = 0;
+		return;
+	}
+
+	ctl->extents_thresh = (max_bitmaps - ctl->total_bitmaps) *
+				PAGE_CACHE_SIZE / sizeof(*info);
+}
+
+/*
+ * We don't fall back to bitmap, if we are below the extents threshold
+ * or this chunk of inode numbers is a big one.
+ */
+static bool use_bitmap(struct btrfs_free_space_ctl *ctl,
+		       struct btrfs_free_space *info)
+{
+	if (ctl->free_extents < ctl->extents_thresh ||
+	    info->bytes > INODES_PER_BITMAP / 10)
+		return false;
+
+	return true;
+}
+
+static struct btrfs_free_space_op free_ino_op = {
+	.recalc_thresholds	= recalculate_thresholds,
+	.use_bitmap		= use_bitmap,
+};
+
+static void pinned_recalc_thresholds(struct btrfs_free_space_ctl *ctl)
+{
+}
+
+static bool pinned_use_bitmap(struct btrfs_free_space_ctl *ctl,
+			      struct btrfs_free_space *info)
+{
+	/*
+	 * We always use extents for two reasons:
+	 *
+	 * - The pinned tree is only used during the process of caching
+	 *   work.
+	 * - Make code simpler. See btrfs_unpin_free_ino().
+	 */
+	return false;
+}
+
+static struct btrfs_free_space_op pinned_free_ino_op = {
+	.recalc_thresholds	= pinned_recalc_thresholds,
+	.use_bitmap		= pinned_use_bitmap,
+};
+
+void btrfs_init_free_ino_ctl(struct btrfs_root *root)
+{
+	struct btrfs_free_space_ctl *ctl = root->free_ino_ctl;
+	struct btrfs_free_space_ctl *pinned = root->free_ino_pinned;
+
+	spin_lock_init(&ctl->tree_lock);
+	ctl->unit = 1;
+	ctl->start = 0;
+	ctl->private = NULL;
+	ctl->op = &free_ino_op;
+
+	/*
+	 * Initially we allow to use 16K of ram to cache chunks of
+	 * inode numbers before we resort to bitmaps. This is somewhat
+	 * arbitrary, but it will be adjusted in runtime.
+	 */
+	ctl->extents_thresh = INIT_THRESHOLD;
+
+	spin_lock_init(&pinned->tree_lock);
+	pinned->unit = 1;
+	pinned->start = 0;
+	pinned->private = NULL;
+	pinned->extents_thresh = 0;
+	pinned->op = &pinned_free_ino_op;
+}
+
+int btrfs_save_ino_cache(struct btrfs_root *root,
+			 struct btrfs_trans_handle *trans)
+{
+	struct btrfs_free_space_ctl *ctl = root->free_ino_ctl;
+	struct btrfs_path *path;
+	struct inode *inode;
+	struct btrfs_block_rsv *rsv;
+	u64 num_bytes;
+	u64 alloc_hint = 0;
+	int ret;
+	int prealloc;
+	bool retry = false;
+
+	/* only fs tree and subvol/snap needs ino cache */
+	if (root->root_key.objectid != BTRFS_FS_TREE_OBJECTID &&
+	    (root->root_key.objectid < BTRFS_FIRST_FREE_OBJECTID ||
+	     root->root_key.objectid > BTRFS_LAST_FREE_OBJECTID))
+		return 0;
+
+	/* Don't save inode cache if we are deleting this root */
+	if (btrfs_root_refs(&root->root_item) == 0)
+		return 0;
+
+	if (!btrfs_test_opt(root, INODE_MAP_CACHE))
+		return 0;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	rsv = trans->block_rsv;
+	trans->block_rsv = &root->fs_info->trans_block_rsv;
+
+	num_bytes = trans->bytes_reserved;
+	/*
+	 * 1 item for inode item insertion if need
+	 * 4 items for inode item update (in the worst case)
+	 * 1 items for slack space if we need do truncation
+	 * 1 item for free space object
+	 * 3 items for pre-allocation
+	 */
+	trans->bytes_reserved = btrfs_calc_trans_metadata_size(root, 10);
+	ret = btrfs_block_rsv_add(root, trans->block_rsv,
+				  trans->bytes_reserved,
+				  BTRFS_RESERVE_NO_FLUSH);
+	if (ret)
+		goto out;
+	trace_btrfs_space_reservation(root->fs_info, "ino_cache",
+				      trans->transid, trans->bytes_reserved, 1);
+again:
+	inode = lookup_free_ino_inode(root, path);
+	if (IS_ERR(inode) && (PTR_ERR(inode) != -ENOENT || retry)) {
+		ret = PTR_ERR(inode);
+		goto out_release;
+	}
+
+	if (IS_ERR(inode)) {
+		BUG_ON(retry); /* Logic error */
+		retry = true;
+
+		ret = create_free_ino_inode(root, trans, path);
+		if (ret)
+			goto out_release;
+		goto again;
+	}
+
+	BTRFS_I(inode)->generation = 0;
+	ret = btrfs_update_inode(trans, root, inode);
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
+		goto out_put;
+	}
+
+	if (i_size_read(inode) > 0) {
+		ret = btrfs_truncate_free_space_cache(root, trans, inode);
+		if (ret) {
+			if (ret != -ENOSPC)
+				btrfs_abort_transaction(trans, root, ret);
+			goto out_put;
+		}
+	}
+
+	spin_lock(&root->cache_lock);
+	if (root->cached != BTRFS_CACHE_FINISHED) {
+		ret = -1;
+		spin_unlock(&root->cache_lock);
+		goto out_put;
+	}
+	spin_unlock(&root->cache_lock);
+
+	spin_lock(&ctl->tree_lock);
+	prealloc = sizeof(struct btrfs_free_space) * ctl->free_extents;
+	prealloc = ALIGN(prealloc, PAGE_CACHE_SIZE);
+	prealloc += ctl->total_bitmaps * PAGE_CACHE_SIZE;
+	spin_unlock(&ctl->tree_lock);
+
+	/* Just to make sure we have enough space */
+	prealloc += 8 * PAGE_CACHE_SIZE;
+
+	ret = btrfs_delalloc_reserve_space(inode, prealloc);
+	if (ret)
+		goto out_put;
+
+	ret = btrfs_prealloc_file_range_trans(inode, trans, 0, 0, prealloc,
+					      prealloc, prealloc, &alloc_hint);
+	if (ret) {
+		btrfs_delalloc_release_space(inode, prealloc);
+		goto out_put;
+	}
+	btrfs_free_reserved_data_space(inode, prealloc);
+
+	ret = btrfs_write_out_ino_cache(root, trans, path, inode);
+out_put:
+	iput(inode);
+out_release:
+	trace_btrfs_space_reservation(root->fs_info, "ino_cache",
+				      trans->transid, trans->bytes_reserved, 0);
+	btrfs_block_rsv_release(root, trans->block_rsv, trans->bytes_reserved);
+out:
+	trans->block_rsv = rsv;
+	trans->bytes_reserved = num_bytes;
+
+	btrfs_free_path(path);
+	return ret;
+}
+
+static int btrfs_find_highest_objectid(struct btrfs_root *root, u64 *objectid)
 {
 	struct btrfs_path *path;
 	int ret;
@@ -30,7 +516,8 @@ int btrfs_find_highest_inode(struct btrfs_root *root, u64 *objectid)
 	int slot;
 
 	path = btrfs_alloc_path();
-	BUG_ON(!path);
+	if (!path)
+		return -ENOMEM;
 
 	search_key.objectid = BTRFS_LAST_FREE_OBJECTID;
 	search_key.type = -1;
@@ -38,7 +525,7 @@ int btrfs_find_highest_inode(struct btrfs_root *root, u64 *objectid)
 	ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0);
 	if (ret < 0)
 		goto error;
-	BUG_ON(ret == 0);
+	BUG_ON(ret == 0); /* Corruption */
 	if (path->slots[0] > 0) {
 		slot = path->slots[0] - 1;
 		l = path->nodes[0];
@@ -54,15 +541,14 @@ error:
 	return ret;
 }
 
-int btrfs_find_free_objectid(struct btrfs_trans_handle *trans,
-			     struct btrfs_root *root,
-			     u64 dirid, u64 *objectid)
+int btrfs_find_free_objectid(struct btrfs_root *root, u64 *objectid)
 {
 	int ret;
 	mutex_lock(&root->objectid_mutex);
 
 	if (unlikely(root->highest_objectid < BTRFS_FIRST_FREE_OBJECTID)) {
-		ret = btrfs_find_highest_inode(root, &root->highest_objectid);
+		ret = btrfs_find_highest_objectid(root,
+						  &root->highest_objectid);
 		if (ret)
 			goto out;
 	}
diff --git a/fs/btrfs/inode-map.h b/fs/btrfs/inode-map.h
new file mode 100644
index 00000000000..ddb347bfee2
--- /dev/null
+++ b/fs/btrfs/inode-map.h
@@ -0,0 +1,13 @@
+#ifndef __BTRFS_INODE_MAP
+#define __BTRFS_INODE_MAP
+
+void btrfs_init_free_ino_ctl(struct btrfs_root *root);
+void btrfs_unpin_free_ino(struct btrfs_root *root);
+void btrfs_return_ino(struct btrfs_root *root, u64 objectid);
+int btrfs_find_free_ino(struct btrfs_root *root, u64 *objectid);
+int btrfs_save_ino_cache(struct btrfs_root *root,
+			 struct btrfs_trans_handle *trans);
+
+int btrfs_find_free_objectid(struct btrfs_root *root, u64 *objectid);
+
+#endif
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 2bfdc641d4e..3668048e16f 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -32,27 +32,36 @@
 #include <linux/writeback.h>
 #include <linux/statfs.h>
 #include <linux/compat.h>
+#include <linux/aio.h>
 #include <linux/bit_spinlock.h>
 #include <linux/xattr.h>
 #include <linux/posix_acl.h>
 #include <linux/falloc.h>
 #include <linux/slab.h>
-#include "compat.h"
+#include <linux/ratelimit.h>
+#include <linux/mount.h>
+#include <linux/btrfs.h>
+#include <linux/blkdev.h>
+#include <linux/posix_acl_xattr.h>
 #include "ctree.h"
 #include "disk-io.h"
 #include "transaction.h"
 #include "btrfs_inode.h"
-#include "ioctl.h"
 #include "print-tree.h"
-#include "volumes.h"
 #include "ordered-data.h"
 #include "xattr.h"
 #include "tree-log.h"
+#include "volumes.h"
 #include "compression.h"
 #include "locking.h"
+#include "free-space-cache.h"
+#include "inode-map.h"
+#include "backref.h"
+#include "hash.h"
+#include "props.h"
 
 struct btrfs_iget_args {
-	u64 ino;
+	struct btrfs_key *location;
 	struct btrfs_root *root;
 };
 
@@ -67,9 +76,11 @@ static const struct file_operations btrfs_dir_file_operations;
 static struct extent_io_ops btrfs_extent_io_ops;
 
 static struct kmem_cache *btrfs_inode_cachep;
+static struct kmem_cache *btrfs_delalloc_work_cachep;
 struct kmem_cache *btrfs_trans_handle_cachep;
 struct kmem_cache *btrfs_transaction_cachep;
 struct kmem_cache *btrfs_path_cachep;
+struct kmem_cache *btrfs_free_space_cachep;
 
 #define S_SHIFT 12
 static unsigned char btrfs_type_by_mode[S_IFMT >> S_SHIFT] = {
@@ -82,21 +93,30 @@ static unsigned char btrfs_type_by_mode[S_IFMT >> S_SHIFT] = {
 	[S_IFLNK >> S_SHIFT]	= BTRFS_FT_SYMLINK,
 };
 
-static void btrfs_truncate(struct inode *inode);
-static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end);
+static int btrfs_setsize(struct inode *inode, struct iattr *attr);
+static int btrfs_truncate(struct inode *inode);
+static int btrfs_finish_ordered_io(struct btrfs_ordered_extent *ordered_extent);
 static noinline int cow_file_range(struct inode *inode,
 				   struct page *locked_page,
 				   u64 start, u64 end, int *page_started,
 				   unsigned long *nr_written, int unlock);
+static struct extent_map *create_pinned_em(struct inode *inode, u64 start,
+					   u64 len, u64 orig_start,
+					   u64 block_start, u64 block_len,
+					   u64 orig_block_len, u64 ram_bytes,
+					   int type);
+
+static int btrfs_dirty_inode(struct inode *inode);
 
 static int btrfs_init_inode_security(struct btrfs_trans_handle *trans,
-				     struct inode *inode,  struct inode *dir)
+				     struct inode *inode,  struct inode *dir,
+				     const struct qstr *qstr)
 {
 	int err;
 
 	err = btrfs_init_acl(trans, inode, dir);
 	if (!err)
-		err = btrfs_xattr_security_init(trans, inode, dir);
+		err = btrfs_xattr_security_init(trans, inode, dir, qstr);
 	return err;
 }
 
@@ -105,13 +125,13 @@ static int btrfs_init_inode_security(struct btrfs_trans_handle *trans,
  * the btree.  The caller should have done a btrfs_drop_extents so that
  * no overlapping inline items exist in the btree
  */
-static noinline int insert_inline_extent(struct btrfs_trans_handle *trans,
+static int insert_inline_extent(struct btrfs_trans_handle *trans,
+				struct btrfs_path *path, int extent_inserted,
 				struct btrfs_root *root, struct inode *inode,
 				u64 start, size_t size, size_t compressed_size,
+				int compress_type,
 				struct page **compressed_pages)
 {
-	struct btrfs_key key;
-	struct btrfs_path *path;
 	struct extent_buffer *leaf;
 	struct page *page = NULL;
 	char *kaddr;
@@ -120,34 +140,29 @@ static noinline int insert_inline_extent(struct btrfs_trans_handle *trans,
 	int err = 0;
 	int ret;
 	size_t cur_size = size;
-	size_t datasize;
 	unsigned long offset;
-	int use_compress = 0;
 
-	if (compressed_size && compressed_pages) {
-		use_compress = 1;
+	if (compressed_size && compressed_pages)
 		cur_size = compressed_size;
-	}
 
-	path = btrfs_alloc_path();
-	if (!path)
-		return -ENOMEM;
+	inode_add_bytes(inode, size);
 
-	path->leave_spinning = 1;
-	btrfs_set_trans_block_group(trans, inode);
+	if (!extent_inserted) {
+		struct btrfs_key key;
+		size_t datasize;
 
-	key.objectid = inode->i_ino;
-	key.offset = start;
-	btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
-	datasize = btrfs_file_extent_calc_inline_size(cur_size);
+		key.objectid = btrfs_ino(inode);
+		key.offset = start;
+		btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
 
-	inode_add_bytes(inode, size);
-	ret = btrfs_insert_empty_item(trans, root, path, &key,
-				      datasize);
-	BUG_ON(ret);
-	if (ret) {
-		err = ret;
-		goto fail;
+		datasize = btrfs_file_extent_calc_inline_size(cur_size);
+		path->leave_spinning = 1;
+		ret = btrfs_insert_empty_item(trans, root, path, &key,
+					      datasize);
+		if (ret) {
+			err = ret;
+			goto fail;
+		}
 	}
 	leaf = path->nodes[0];
 	ei = btrfs_item_ptr(leaf, path->slots[0],
@@ -159,7 +174,7 @@ static noinline int insert_inline_extent(struct btrfs_trans_handle *trans,
 	btrfs_set_file_extent_ram_bytes(leaf, ei, size);
 	ptr = btrfs_file_extent_inline_start(ei);
 
-	if (use_compress) {
+	if (compress_type != BTRFS_COMPRESS_NONE) {
 		struct page *cpage;
 		int i = 0;
 		while (compressed_size > 0) {
@@ -167,28 +182,28 @@ static noinline int insert_inline_extent(struct btrfs_trans_handle *trans,
 			cur_size = min_t(unsigned long, compressed_size,
 				       PAGE_CACHE_SIZE);
 
-			kaddr = kmap_atomic(cpage, KM_USER0);
+			kaddr = kmap_atomic(cpage);
 			write_extent_buffer(leaf, kaddr, ptr, cur_size);
-			kunmap_atomic(kaddr, KM_USER0);
+			kunmap_atomic(kaddr);
 
 			i++;
 			ptr += cur_size;
 			compressed_size -= cur_size;
 		}
 		btrfs_set_file_extent_compression(leaf, ei,
-						  BTRFS_COMPRESS_ZLIB);
+						  compress_type);
 	} else {
 		page = find_get_page(inode->i_mapping,
 				     start >> PAGE_CACHE_SHIFT);
 		btrfs_set_file_extent_compression(leaf, ei, 0);
-		kaddr = kmap_atomic(page, KM_USER0);
+		kaddr = kmap_atomic(page);
 		offset = start & (PAGE_CACHE_SIZE - 1);
 		write_extent_buffer(leaf, kaddr + offset, ptr, size);
-		kunmap_atomic(kaddr, KM_USER0);
+		kunmap_atomic(kaddr);
 		page_cache_release(page);
 	}
 	btrfs_mark_buffer_dirty(leaf);
-	btrfs_free_path(path);
+	btrfs_release_path(path);
 
 	/*
 	 * we're an inline extent, so nobody can
@@ -200,11 +215,10 @@ static noinline int insert_inline_extent(struct btrfs_trans_handle *trans,
 	 * could end up racing with unlink.
 	 */
 	BTRFS_I(inode)->disk_i_size = inode->i_size;
-	btrfs_update_inode(trans, root, inode);
+	ret = btrfs_update_inode(trans, root, inode);
 
-	return 0;
+	return ret;
 fail:
-	btrfs_free_path(path);
 	return err;
 }
 
@@ -214,20 +228,22 @@ fail:
  * does the checks required to make sure the data is small enough
  * to fit as an inline extent.
  */
-static noinline int cow_file_range_inline(struct btrfs_trans_handle *trans,
-				 struct btrfs_root *root,
-				 struct inode *inode, u64 start, u64 end,
-				 size_t compressed_size,
-				 struct page **compressed_pages)
+static noinline int cow_file_range_inline(struct btrfs_root *root,
+					  struct inode *inode, u64 start,
+					  u64 end, size_t compressed_size,
+					  int compress_type,
+					  struct page **compressed_pages)
 {
+	struct btrfs_trans_handle *trans;
 	u64 isize = i_size_read(inode);
 	u64 actual_end = min(end + 1, isize);
 	u64 inline_len = actual_end - start;
-	u64 aligned_end = (end + root->sectorsize - 1) &
-			~((u64)root->sectorsize - 1);
-	u64 hint_byte;
+	u64 aligned_end = ALIGN(end, root->sectorsize);
 	u64 data_len = inline_len;
 	int ret;
+	struct btrfs_path *path;
+	int extent_inserted = 0;
+	u32 extent_item_size;
 
 	if (compressed_size)
 		data_len = compressed_size;
@@ -242,18 +258,53 @@ static noinline int cow_file_range_inline(struct btrfs_trans_handle *trans,
 		return 1;
 	}
 
-	ret = btrfs_drop_extents(trans, inode, start, aligned_end,
-				 &hint_byte, 1);
-	BUG_ON(ret);
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	trans = btrfs_join_transaction(root);
+	if (IS_ERR(trans)) {
+		btrfs_free_path(path);
+		return PTR_ERR(trans);
+	}
+	trans->block_rsv = &root->fs_info->delalloc_block_rsv;
+
+	if (compressed_size && compressed_pages)
+		extent_item_size = btrfs_file_extent_calc_inline_size(
+		   compressed_size);
+	else
+		extent_item_size = btrfs_file_extent_calc_inline_size(
+		    inline_len);
+
+	ret = __btrfs_drop_extents(trans, root, inode, path,
+				   start, aligned_end, NULL,
+				   1, 1, extent_item_size, &extent_inserted);
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
+		goto out;
+	}
 
 	if (isize > actual_end)
 		inline_len = min_t(u64, isize, actual_end);
-	ret = insert_inline_extent(trans, root, inode, start,
+	ret = insert_inline_extent(trans, path, extent_inserted,
+				   root, inode, start,
 				   inline_len, compressed_size,
-				   compressed_pages);
-	BUG_ON(ret);
+				   compress_type, compressed_pages);
+	if (ret && ret != -ENOSPC) {
+		btrfs_abort_transaction(trans, root, ret);
+		goto out;
+	} else if (ret == -ENOSPC) {
+		ret = 1;
+		goto out;
+	}
+
+	set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &BTRFS_I(inode)->runtime_flags);
+	btrfs_delalloc_release_metadata(inode, end + 1 - start);
 	btrfs_drop_extent_cache(inode, start, aligned_end - 1, 0);
-	return 0;
+out:
+	btrfs_free_path(path);
+	btrfs_end_transaction(trans, root);
+	return ret;
 }
 
 struct async_extent {
@@ -262,6 +313,7 @@ struct async_extent {
 	u64 compressed_size;
 	struct page **pages;
 	unsigned long nr_pages;
+	int compress_type;
 	struct list_head list;
 };
 
@@ -279,16 +331,19 @@ static noinline int add_async_extent(struct async_cow *cow,
 				     u64 start, u64 ram_size,
 				     u64 compressed_size,
 				     struct page **pages,
-				     unsigned long nr_pages)
+				     unsigned long nr_pages,
+				     int compress_type)
 {
 	struct async_extent *async_extent;
 
 	async_extent = kmalloc(sizeof(*async_extent), GFP_NOFS);
+	BUG_ON(!async_extent); /* -ENOMEM */
 	async_extent->start = start;
 	async_extent->ram_size = ram_size;
 	async_extent->compressed_size = compressed_size;
 	async_extent->pages = pages;
 	async_extent->nr_pages = nr_pages;
+	async_extent->compress_type = compress_type;
 	list_add_tail(&async_extent->list, &cow->extents);
 	return 0;
 }
@@ -307,7 +362,8 @@ static noinline int add_async_extent(struct async_cow *cow,
  * If this code finds it can't get good compression, it puts an
  * entry onto the work queue to write the uncompressed bytes.  This
  * makes sure that both compressed inodes and uncompressed inodes
- * are written in the same order that pdflush sent them down.
+ * are written in the same order that the flusher thread sent them
+ * down.
  */
 static noinline int compress_file_range(struct inode *inode,
 					struct page *locked_page,
@@ -316,10 +372,7 @@ static noinline int compress_file_range(struct inode *inode,
 					int *num_added)
 {
 	struct btrfs_root *root = BTRFS_I(inode)->root;
-	struct btrfs_trans_handle *trans;
 	u64 num_bytes;
-	u64 orig_start;
-	u64 disk_num_bytes;
 	u64 blocksize = root->sectorsize;
 	u64 actual_end;
 	u64 isize = i_size_read(inode);
@@ -333,8 +386,21 @@ static noinline int compress_file_range(struct inode *inode,
 	unsigned long max_uncompressed = 128 * 1024;
 	int i;
 	int will_compress;
+	int compress_type = root->fs_info->compress_type;
+	int redirty = 0;
+
+	/* if this is a small write inside eof, kick off a defrag */
+	if ((end - start + 1) < 16 * 1024 &&
+	    (start > 0 || end + 1 < BTRFS_I(inode)->disk_i_size))
+		btrfs_add_inode_defrag(NULL, inode);
 
-	orig_start = start;
+	/*
+	 * skip compression for a small file range(<=blocksize) that
+	 * isn't an inline extent, since it dosen't save disk space at all.
+	 */
+	if ((end - start + 1) <= blocksize &&
+	    (start > 0 || end + 1 < BTRFS_I(inode)->disk_i_size))
+		goto cleanup_and_bail_uncompressed;
 
 	actual_end = min_t(u64, isize, end + 1);
 again:
@@ -368,9 +434,8 @@ again:
 	 * a compressed extent to 128k.
 	 */
 	total_compressed = min(total_compressed, max_uncompressed);
-	num_bytes = (end - start + blocksize) & ~(blocksize - 1);
+	num_bytes = ALIGN(end - start + 1, blocksize);
 	num_bytes = max(blocksize,  num_bytes);
-	disk_num_bytes = num_bytes;
 	total_in = 0;
 	ret = 0;
 
@@ -381,16 +446,36 @@ again:
 	 */
 	if (!(BTRFS_I(inode)->flags & BTRFS_INODE_NOCOMPRESS) &&
 	    (btrfs_test_opt(root, COMPRESS) ||
-	     (BTRFS_I(inode)->force_compress))) {
+	     (BTRFS_I(inode)->force_compress) ||
+	     (BTRFS_I(inode)->flags & BTRFS_INODE_COMPRESS))) {
 		WARN_ON(pages);
 		pages = kzalloc(sizeof(struct page *) * nr_pages, GFP_NOFS);
+		if (!pages) {
+			/* just bail out to the uncompressed code */
+			goto cont;
+		}
 
-		ret = btrfs_zlib_compress_pages(inode->i_mapping, start,
-						total_compressed, pages,
-						nr_pages, &nr_pages_ret,
-						&total_in,
-						&total_compressed,
-						max_compressed);
+		if (BTRFS_I(inode)->force_compress)
+			compress_type = BTRFS_I(inode)->force_compress;
+
+		/*
+		 * we need to call clear_page_dirty_for_io on each
+		 * page in the range.  Otherwise applications with the file
+		 * mmap'd can wander in and change the page contents while
+		 * we are compressing them.
+		 *
+		 * If the compression fails for any reason, we set the pages
+		 * dirty again later on.
+		 */
+		extent_range_clear_dirty_for_io(inode, start, end);
+		redirty = 1;
+		ret = btrfs_compress_pages(compress_type,
+					   inode->i_mapping, start,
+					   total_compressed, pages,
+					   nr_pages, &nr_pages_ret,
+					   &total_in,
+					   &total_compressed,
+					   max_compressed);
 
 		if (!ret) {
 			unsigned long offset = total_compressed &
@@ -402,50 +487,46 @@ again:
 			 * sending it down to disk
 			 */
 			if (offset) {
-				kaddr = kmap_atomic(page, KM_USER0);
+				kaddr = kmap_atomic(page);
 				memset(kaddr + offset, 0,
 				       PAGE_CACHE_SIZE - offset);
-				kunmap_atomic(kaddr, KM_USER0);
+				kunmap_atomic(kaddr);
 			}
 			will_compress = 1;
 		}
 	}
+cont:
 	if (start == 0) {
-		trans = btrfs_join_transaction(root, 1);
-		BUG_ON(!trans);
-		btrfs_set_trans_block_group(trans, inode);
-
 		/* lets try to make an inline extent */
 		if (ret || total_in < (actual_end - start)) {
 			/* we didn't compress the entire range, try
 			 * to make an uncompressed inline extent.
 			 */
-			ret = cow_file_range_inline(trans, root, inode,
-						    start, end, 0, NULL);
+			ret = cow_file_range_inline(root, inode, start, end,
+						    0, 0, NULL);
 		} else {
 			/* try making a compressed inline extent */
-			ret = cow_file_range_inline(trans, root, inode,
-						    start, end,
-						    total_compressed, pages);
+			ret = cow_file_range_inline(root, inode, start, end,
+						    total_compressed,
+						    compress_type, pages);
 		}
-		if (ret == 0) {
+		if (ret <= 0) {
+			unsigned long clear_flags = EXTENT_DELALLOC |
+				EXTENT_DEFRAG;
+			clear_flags |= (ret < 0) ? EXTENT_DO_ACCOUNTING : 0;
+
 			/*
-			 * inline extent creation worked, we don't need
-			 * to create any more async work items.  Unlock
-			 * and free up our temp pages.
+			 * inline extent creation worked or returned error,
+			 * we don't need to create any more async work items.
+			 * Unlock and free up our temp pages.
 			 */
-			extent_clear_unlock_delalloc(inode,
-			     &BTRFS_I(inode)->io_tree,
-			     start, end, NULL,
-			     EXTENT_CLEAR_UNLOCK_PAGE | EXTENT_CLEAR_DIRTY |
-			     EXTENT_CLEAR_DELALLOC |
-			     EXTENT_CLEAR_ACCOUNTING |
-			     EXTENT_SET_WRITEBACK | EXTENT_END_WRITEBACK);
-
-			btrfs_end_transaction(trans, root);
+			extent_clear_unlock_delalloc(inode, start, end, NULL,
+						     clear_flags, PAGE_UNLOCK |
+						     PAGE_CLEAR_DIRTY |
+						     PAGE_SET_WRITEBACK |
+						     PAGE_END_WRITEBACK);
 			goto free_pages_out;
 		}
-		btrfs_end_transaction(trans, root);
 	}
 
 	if (will_compress) {
@@ -454,19 +535,16 @@ again:
 		 * up to a block size boundary so the allocator does sane
 		 * things
 		 */
-		total_compressed = (total_compressed + blocksize - 1) &
-			~(blocksize - 1);
+		total_compressed = ALIGN(total_compressed, blocksize);
 
 		/*
 		 * one last check to make sure the compression is really a
 		 * win, compare the page count read with the blocks on disk
 		 */
-		total_in = (total_in + PAGE_CACHE_SIZE - 1) &
-			~(PAGE_CACHE_SIZE - 1);
+		total_in = ALIGN(total_in, PAGE_CACHE_SIZE);
 		if (total_compressed >= total_in) {
 			will_compress = 0;
 		} else {
-			disk_num_bytes = total_compressed;
 			num_bytes = total_in;
 		}
 	}
@@ -498,9 +576,10 @@ again:
 		 * and will submit them to the elevator.
 		 */
 		add_async_extent(async_cow, start, num_bytes,
-				 total_compressed, pages, nr_pages_ret);
+				 total_compressed, pages, nr_pages_ret,
+				 compress_type);
 
-		if (start + num_bytes < end && start + num_bytes < actual_end) {
+		if (start + num_bytes < end) {
 			start += num_bytes;
 			pages = NULL;
 			cond_resched();
@@ -520,12 +599,15 @@ cleanup_and_bail_uncompressed:
 			__set_page_dirty_nobuffers(locked_page);
 			/* unlocked later on in the async handlers */
 		}
-		add_async_extent(async_cow, start, end - start + 1, 0, NULL, 0);
+		if (redirty)
+			extent_range_redirty_for_io(inode, start, end);
+		add_async_extent(async_cow, start, end - start + 1,
+				 0, NULL, 0, BTRFS_COMPRESS_NONE);
 		*num_added += 1;
 	}
 
 out:
-	return 0;
+	return ret;
 
 free_pages_out:
 	for (i = 0; i < nr_pages_ret; i++) {
@@ -548,7 +630,6 @@ static noinline int submit_compressed_extents(struct inode *inode,
 {
 	struct async_extent *async_extent;
 	u64 alloc_hint = 0;
-	struct btrfs_trans_handle *trans;
 	struct btrfs_key ins;
 	struct extent_map *em;
 	struct btrfs_root *root = BTRFS_I(inode)->root;
@@ -559,7 +640,7 @@ static noinline int submit_compressed_extents(struct inode *inode,
 	if (list_empty(&async_cow->extents))
 		return 0;
 
-
+again:
 	while (!list_empty(&async_cow->extents)) {
 		async_extent = list_entry(async_cow->extents.next,
 					  struct async_extent, list);
@@ -575,7 +656,7 @@ retry:
 
 			lock_extent(io_tree, async_extent->start,
 					 async_extent->start +
-					 async_extent->ram_size - 1, GFP_NOFS);
+					 async_extent->ram_size - 1);
 
 			/* allocate blocks */
 			ret = cow_file_range(inode, async_cow->locked_page,
@@ -584,6 +665,8 @@ retry:
 					     async_extent->ram_size - 1,
 					     &page_started, &nr_written, 0);
 
+			/* JDM XXX */
+
 			/*
 			 * if page_started, cow_file_range inserted an
 			 * inline extent and took care of all the unlocking
@@ -597,25 +680,23 @@ retry:
 						  async_extent->ram_size - 1,
 						  btrfs_get_extent,
 						  WB_SYNC_ALL);
+			else if (ret)
+				unlock_page(async_cow->locked_page);
 			kfree(async_extent);
 			cond_resched();
 			continue;
 		}
 
 		lock_extent(io_tree, async_extent->start,
-			    async_extent->start + async_extent->ram_size - 1,
-			    GFP_NOFS);
+			    async_extent->start + async_extent->ram_size - 1);
 
-		trans = btrfs_join_transaction(root, 1);
-		ret = btrfs_reserve_extent(trans, root,
+		ret = btrfs_reserve_extent(root,
 					   async_extent->compressed_size,
 					   async_extent->compressed_size,
-					   0, alloc_hint,
-					   (u64)-1, &ins, 1);
-		btrfs_end_transaction(trans, root);
-
+					   0, alloc_hint, &ins, 1, 1);
 		if (ret) {
 			int i;
+
 			for (i = 0; i < async_extent->nr_pages; i++) {
 				WARN_ON(async_extent->pages[i]->mapping);
 				page_cache_release(async_extent->pages[i]);
@@ -623,10 +704,14 @@ retry:
 			kfree(async_extent->pages);
 			async_extent->nr_pages = 0;
 			async_extent->pages = NULL;
-			unlock_extent(io_tree, async_extent->start,
-				      async_extent->start +
-				      async_extent->ram_size - 1, GFP_NOFS);
-			goto retry;
+
+			if (ret == -ENOSPC) {
+				unlock_extent(io_tree, async_extent->start,
+					      async_extent->start +
+					      async_extent->ram_size - 1);
+				goto retry;
+			}
+			goto out_free;
 		}
 
 		/*
@@ -637,20 +722,30 @@ retry:
 					async_extent->start +
 					async_extent->ram_size - 1, 0);
 
-		em = alloc_extent_map(GFP_NOFS);
+		em = alloc_extent_map();
+		if (!em) {
+			ret = -ENOMEM;
+			goto out_free_reserve;
+		}
 		em->start = async_extent->start;
 		em->len = async_extent->ram_size;
 		em->orig_start = em->start;
+		em->mod_start = em->start;
+		em->mod_len = em->len;
 
 		em->block_start = ins.objectid;
 		em->block_len = ins.offset;
+		em->orig_block_len = ins.offset;
+		em->ram_bytes = async_extent->ram_size;
 		em->bdev = root->fs_info->fs_devices->latest_bdev;
+		em->compress_type = async_extent->compress_type;
 		set_bit(EXTENT_FLAG_PINNED, &em->flags);
 		set_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
+		em->generation = -1;
 
 		while (1) {
 			write_lock(&em_tree->lock);
-			ret = add_extent_mapping(em_tree, em);
+			ret = add_extent_mapping(em_tree, em, 1);
 			write_unlock(&em_tree->lock);
 			if (ret != -EEXIST) {
 				free_extent_map(em);
@@ -661,40 +756,87 @@ retry:
 						async_extent->ram_size - 1, 0);
 		}
 
-		ret = btrfs_add_ordered_extent(inode, async_extent->start,
-					       ins.objectid,
-					       async_extent->ram_size,
-					       ins.offset,
-					       BTRFS_ORDERED_COMPRESSED);
-		BUG_ON(ret);
+		if (ret)
+			goto out_free_reserve;
+
+		ret = btrfs_add_ordered_extent_compress(inode,
+						async_extent->start,
+						ins.objectid,
+						async_extent->ram_size,
+						ins.offset,
+						BTRFS_ORDERED_COMPRESSED,
+						async_extent->compress_type);
+		if (ret)
+			goto out_free_reserve;
 
 		/*
 		 * clear dirty, set writeback and unlock the pages.
 		 */
-		extent_clear_unlock_delalloc(inode,
-				&BTRFS_I(inode)->io_tree,
-				async_extent->start,
+		extent_clear_unlock_delalloc(inode, async_extent->start,
 				async_extent->start +
 				async_extent->ram_size - 1,
-				NULL, EXTENT_CLEAR_UNLOCK_PAGE |
-				EXTENT_CLEAR_UNLOCK |
-				EXTENT_CLEAR_DELALLOC |
-				EXTENT_CLEAR_DIRTY | EXTENT_SET_WRITEBACK);
-
+				NULL, EXTENT_LOCKED | EXTENT_DELALLOC,
+				PAGE_UNLOCK | PAGE_CLEAR_DIRTY |
+				PAGE_SET_WRITEBACK);
 		ret = btrfs_submit_compressed_write(inode,
 				    async_extent->start,
 				    async_extent->ram_size,
 				    ins.objectid,
 				    ins.offset, async_extent->pages,
 				    async_extent->nr_pages);
-
-		BUG_ON(ret);
 		alloc_hint = ins.objectid + ins.offset;
 		kfree(async_extent);
+		if (ret)
+			goto out;
 		cond_resched();
 	}
+	ret = 0;
+out:
+	return ret;
+out_free_reserve:
+	btrfs_free_reserved_extent(root, ins.objectid, ins.offset, 1);
+out_free:
+	extent_clear_unlock_delalloc(inode, async_extent->start,
+				     async_extent->start +
+				     async_extent->ram_size - 1,
+				     NULL, EXTENT_LOCKED | EXTENT_DELALLOC |
+				     EXTENT_DEFRAG | EXTENT_DO_ACCOUNTING,
+				     PAGE_UNLOCK | PAGE_CLEAR_DIRTY |
+				     PAGE_SET_WRITEBACK | PAGE_END_WRITEBACK);
+	kfree(async_extent);
+	goto again;
+}
 
-	return 0;
+static u64 get_extent_allocation_hint(struct inode *inode, u64 start,
+				      u64 num_bytes)
+{
+	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
+	struct extent_map *em;
+	u64 alloc_hint = 0;
+
+	read_lock(&em_tree->lock);
+	em = search_extent_mapping(em_tree, start, num_bytes);
+	if (em) {
+		/*
+		 * if block start isn't an actual block number then find the
+		 * first block in this inode and use that as a hint.  If that
+		 * block is also bogus then just don't worry about it.
+		 */
+		if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
+			free_extent_map(em);
+			em = search_extent_mapping(em_tree, 0, 0);
+			if (em && em->block_start < EXTENT_MAP_LAST_BYTE)
+				alloc_hint = em->block_start;
+			if (em)
+				free_extent_map(em);
+		} else {
+			alloc_hint = em->block_start;
+			free_extent_map(em);
+		}
+	}
+	read_unlock(&em_tree->lock);
+
+	return alloc_hint;
 }
 
 /*
@@ -717,106 +859,91 @@ static noinline int cow_file_range(struct inode *inode,
 				   int unlock)
 {
 	struct btrfs_root *root = BTRFS_I(inode)->root;
-	struct btrfs_trans_handle *trans;
 	u64 alloc_hint = 0;
 	u64 num_bytes;
 	unsigned long ram_size;
 	u64 disk_num_bytes;
 	u64 cur_alloc_size;
 	u64 blocksize = root->sectorsize;
-	u64 actual_end;
-	u64 isize = i_size_read(inode);
 	struct btrfs_key ins;
 	struct extent_map *em;
 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
 	int ret = 0;
 
-	trans = btrfs_join_transaction(root, 1);
-	BUG_ON(!trans);
-	btrfs_set_trans_block_group(trans, inode);
-
-	actual_end = min_t(u64, isize, end + 1);
+	if (btrfs_is_free_space_inode(inode)) {
+		WARN_ON_ONCE(1);
+		ret = -EINVAL;
+		goto out_unlock;
+	}
 
-	num_bytes = (end - start + blocksize) & ~(blocksize - 1);
+	num_bytes = ALIGN(end - start + 1, blocksize);
 	num_bytes = max(blocksize,  num_bytes);
 	disk_num_bytes = num_bytes;
-	ret = 0;
+
+	/* if this is a small write inside eof, kick off defrag */
+	if (num_bytes < 64 * 1024 &&
+	    (start > 0 || end + 1 < BTRFS_I(inode)->disk_i_size))
+		btrfs_add_inode_defrag(NULL, inode);
 
 	if (start == 0) {
 		/* lets try to make an inline extent */
-		ret = cow_file_range_inline(trans, root, inode,
-					    start, end, 0, NULL);
+		ret = cow_file_range_inline(root, inode, start, end, 0, 0,
+					    NULL);
 		if (ret == 0) {
-			extent_clear_unlock_delalloc(inode,
-				     &BTRFS_I(inode)->io_tree,
-				     start, end, NULL,
-				     EXTENT_CLEAR_UNLOCK_PAGE |
-				     EXTENT_CLEAR_UNLOCK |
-				     EXTENT_CLEAR_DELALLOC |
-				     EXTENT_CLEAR_ACCOUNTING |
-				     EXTENT_CLEAR_DIRTY |
-				     EXTENT_SET_WRITEBACK |
-				     EXTENT_END_WRITEBACK);
+			extent_clear_unlock_delalloc(inode, start, end, NULL,
+				     EXTENT_LOCKED | EXTENT_DELALLOC |
+				     EXTENT_DEFRAG, PAGE_UNLOCK |
+				     PAGE_CLEAR_DIRTY | PAGE_SET_WRITEBACK |
+				     PAGE_END_WRITEBACK);
 
 			*nr_written = *nr_written +
 			     (end - start + PAGE_CACHE_SIZE) / PAGE_CACHE_SIZE;
 			*page_started = 1;
-			ret = 0;
 			goto out;
+		} else if (ret < 0) {
+			goto out_unlock;
 		}
 	}
 
 	BUG_ON(disk_num_bytes >
-	       btrfs_super_total_bytes(&root->fs_info->super_copy));
-
+	       btrfs_super_total_bytes(root->fs_info->super_copy));
 
-	read_lock(&BTRFS_I(inode)->extent_tree.lock);
-	em = search_extent_mapping(&BTRFS_I(inode)->extent_tree,
-				   start, num_bytes);
-	if (em) {
-		/*
-		 * if block start isn't an actual block number then find the
-		 * first block in this inode and use that as a hint.  If that
-		 * block is also bogus then just don't worry about it.
-		 */
-		if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
-			free_extent_map(em);
-			em = search_extent_mapping(em_tree, 0, 0);
-			if (em && em->block_start < EXTENT_MAP_LAST_BYTE)
-				alloc_hint = em->block_start;
-			if (em)
-				free_extent_map(em);
-		} else {
-			alloc_hint = em->block_start;
-			free_extent_map(em);
-		}
-	}
-	read_unlock(&BTRFS_I(inode)->extent_tree.lock);
+	alloc_hint = get_extent_allocation_hint(inode, start, num_bytes);
 	btrfs_drop_extent_cache(inode, start, start + num_bytes - 1, 0);
 
 	while (disk_num_bytes > 0) {
 		unsigned long op;
 
 		cur_alloc_size = disk_num_bytes;
-		ret = btrfs_reserve_extent(trans, root, cur_alloc_size,
+		ret = btrfs_reserve_extent(root, cur_alloc_size,
 					   root->sectorsize, 0, alloc_hint,
-					   (u64)-1, &ins, 1);
-		BUG_ON(ret);
+					   &ins, 1, 1);
+		if (ret < 0)
+			goto out_unlock;
 
-		em = alloc_extent_map(GFP_NOFS);
+		em = alloc_extent_map();
+		if (!em) {
+			ret = -ENOMEM;
+			goto out_reserve;
+		}
 		em->start = start;
 		em->orig_start = em->start;
 		ram_size = ins.offset;
 		em->len = ins.offset;
+		em->mod_start = em->start;
+		em->mod_len = em->len;
 
 		em->block_start = ins.objectid;
 		em->block_len = ins.offset;
+		em->orig_block_len = ins.offset;
+		em->ram_bytes = ram_size;
 		em->bdev = root->fs_info->fs_devices->latest_bdev;
 		set_bit(EXTENT_FLAG_PINNED, &em->flags);
+		em->generation = -1;
 
 		while (1) {
 			write_lock(&em_tree->lock);
-			ret = add_extent_mapping(em_tree, em);
+			ret = add_extent_mapping(em_tree, em, 1);
 			write_unlock(&em_tree->lock);
 			if (ret != -EEXIST) {
 				free_extent_map(em);
@@ -825,17 +952,21 @@ static noinline int cow_file_range(struct inode *inode,
 			btrfs_drop_extent_cache(inode, start,
 						start + ram_size - 1, 0);
 		}
+		if (ret)
+			goto out_reserve;
 
 		cur_alloc_size = ins.offset;
 		ret = btrfs_add_ordered_extent(inode, start, ins.objectid,
 					       ram_size, cur_alloc_size, 0);
-		BUG_ON(ret);
+		if (ret)
+			goto out_reserve;
 
 		if (root->root_key.objectid ==
 		    BTRFS_DATA_RELOC_TREE_OBJECTID) {
 			ret = btrfs_reloc_clone_csums(inode, start,
 						      cur_alloc_size);
-			BUG_ON(ret);
+			if (ret)
+				goto out_reserve;
 		}
 
 		if (disk_num_bytes < cur_alloc_size)
@@ -848,23 +979,30 @@ static noinline int cow_file_range(struct inode *inode,
 		 * Do set the Private2 bit so we know this page was properly
 		 * setup for writepage
 		 */
-		op = unlock ? EXTENT_CLEAR_UNLOCK_PAGE : 0;
-		op |= EXTENT_CLEAR_UNLOCK | EXTENT_CLEAR_DELALLOC |
-			EXTENT_SET_PRIVATE2;
+		op = unlock ? PAGE_UNLOCK : 0;
+		op |= PAGE_SET_PRIVATE2;
 
-		extent_clear_unlock_delalloc(inode, &BTRFS_I(inode)->io_tree,
-					     start, start + ram_size - 1,
-					     locked_page, op);
+		extent_clear_unlock_delalloc(inode, start,
+					     start + ram_size - 1, locked_page,
+					     EXTENT_LOCKED | EXTENT_DELALLOC,
+					     op);
 		disk_num_bytes -= cur_alloc_size;
 		num_bytes -= cur_alloc_size;
 		alloc_hint = ins.objectid + ins.offset;
 		start += cur_alloc_size;
 	}
 out:
-	ret = 0;
-	btrfs_end_transaction(trans, root);
-
 	return ret;
+
+out_reserve:
+	btrfs_free_reserved_extent(root, ins.objectid, ins.offset, 1);
+out_unlock:
+	extent_clear_unlock_delalloc(inode, start, end, locked_page,
+				     EXTENT_LOCKED | EXTENT_DO_ACCOUNTING |
+				     EXTENT_DELALLOC | EXTENT_DEFRAG,
+				     PAGE_UNLOCK | PAGE_CLEAR_DIRTY |
+				     PAGE_SET_WRITEBACK | PAGE_END_WRITEBACK);
+	goto out;
 }
 
 /*
@@ -879,8 +1017,10 @@ static noinline void async_cow_start(struct btrfs_work *work)
 	compress_file_range(async_cow->inode, async_cow->locked_page,
 			    async_cow->start, async_cow->end, async_cow,
 			    &num_added);
-	if (num_added == 0)
+	if (num_added == 0) {
+		btrfs_add_delayed_iput(async_cow->inode);
 		async_cow->inode = NULL;
+	}
 }
 
 /*
@@ -898,10 +1038,8 @@ static noinline void async_cow_submit(struct btrfs_work *work)
 	nr_pages = (async_cow->end - async_cow->start + PAGE_CACHE_SIZE) >>
 		PAGE_CACHE_SHIFT;
 
-	atomic_sub(nr_pages, &root->fs_info->async_delalloc_pages);
-
-	if (atomic_read(&root->fs_info->async_delalloc_pages) <
-	    5 * 1042 * 1024 &&
+	if (atomic_sub_return(nr_pages, &root->fs_info->async_delalloc_pages) <
+	    5 * 1024 * 1024 &&
 	    waitqueue_active(&root->fs_info->async_submit_wait))
 		wake_up(&root->fs_info->async_submit_wait);
 
@@ -913,6 +1051,8 @@ static noinline void async_cow_free(struct btrfs_work *work)
 {
 	struct async_cow *async_cow;
 	async_cow = container_of(work, struct async_cow, work);
+	if (async_cow->inode)
+		btrfs_add_delayed_iput(async_cow->inode);
 	kfree(async_cow);
 }
 
@@ -924,13 +1064,14 @@ static int cow_file_range_async(struct inode *inode, struct page *locked_page,
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	unsigned long nr_pages;
 	u64 cur_end;
-	int limit = 10 * 1024 * 1042;
+	int limit = 10 * 1024 * 1024;
 
 	clear_extent_bit(&BTRFS_I(inode)->io_tree, start, end, EXTENT_LOCKED,
 			 1, 0, NULL, GFP_NOFS);
 	while (start < end) {
 		async_cow = kmalloc(sizeof(*async_cow), GFP_NOFS);
-		async_cow->inode = inode;
+		BUG_ON(!async_cow); /* -ENOMEM */
+		async_cow->inode = igrab(inode);
 		async_cow->root = root;
 		async_cow->locked_page = locked_page;
 		async_cow->start = start;
@@ -943,17 +1084,15 @@ static int cow_file_range_async(struct inode *inode, struct page *locked_page,
 		async_cow->end = cur_end;
 		INIT_LIST_HEAD(&async_cow->extents);
 
-		async_cow->work.func = async_cow_start;
-		async_cow->work.ordered_func = async_cow_submit;
-		async_cow->work.ordered_free = async_cow_free;
-		async_cow->work.flags = 0;
+		btrfs_init_work(&async_cow->work, async_cow_start,
+				async_cow_submit, async_cow_free);
 
 		nr_pages = (cur_end - start + PAGE_CACHE_SIZE) >>
 			PAGE_CACHE_SHIFT;
 		atomic_add(nr_pages, &root->fs_info->async_delalloc_pages);
 
-		btrfs_queue_worker(&root->fs_info->delalloc_workers,
-				   &async_cow->work);
+		btrfs_queue_work(root->fs_info->delalloc_workers,
+				 &async_cow->work);
 
 		if (atomic_read(&root->fs_info->async_delalloc_pages) > limit) {
 			wait_event(root->fs_info->async_submit_wait,
@@ -983,7 +1122,7 @@ static noinline int csum_exist_in_range(struct btrfs_root *root,
 	LIST_HEAD(list);
 
 	ret = btrfs_lookup_csums_range(root->fs_info->csum_root, bytenr,
-				       bytenr + num_bytes - 1, &list);
+				       bytenr + num_bytes - 1, &list, 0);
 	if (ret == 0 && list_empty(&list))
 		return 0;
 
@@ -1019,28 +1158,61 @@ static noinline int run_delalloc_nocow(struct inode *inode,
 	u64 extent_offset;
 	u64 disk_bytenr;
 	u64 num_bytes;
+	u64 disk_num_bytes;
+	u64 ram_bytes;
 	int extent_type;
-	int ret;
+	int ret, err;
 	int type;
 	int nocow;
 	int check_prev = 1;
+	bool nolock;
+	u64 ino = btrfs_ino(inode);
 
 	path = btrfs_alloc_path();
-	BUG_ON(!path);
-	trans = btrfs_join_transaction(root, 1);
-	BUG_ON(!trans);
+	if (!path) {
+		extent_clear_unlock_delalloc(inode, start, end, locked_page,
+					     EXTENT_LOCKED | EXTENT_DELALLOC |
+					     EXTENT_DO_ACCOUNTING |
+					     EXTENT_DEFRAG, PAGE_UNLOCK |
+					     PAGE_CLEAR_DIRTY |
+					     PAGE_SET_WRITEBACK |
+					     PAGE_END_WRITEBACK);
+		return -ENOMEM;
+	}
+
+	nolock = btrfs_is_free_space_inode(inode);
+
+	if (nolock)
+		trans = btrfs_join_transaction_nolock(root);
+	else
+		trans = btrfs_join_transaction(root);
+
+	if (IS_ERR(trans)) {
+		extent_clear_unlock_delalloc(inode, start, end, locked_page,
+					     EXTENT_LOCKED | EXTENT_DELALLOC |
+					     EXTENT_DO_ACCOUNTING |
+					     EXTENT_DEFRAG, PAGE_UNLOCK |
+					     PAGE_CLEAR_DIRTY |
+					     PAGE_SET_WRITEBACK |
+					     PAGE_END_WRITEBACK);
+		btrfs_free_path(path);
+		return PTR_ERR(trans);
+	}
+
+	trans->block_rsv = &root->fs_info->delalloc_block_rsv;
 
 	cow_start = (u64)-1;
 	cur_offset = start;
 	while (1) {
-		ret = btrfs_lookup_file_extent(trans, root, path, inode->i_ino,
+		ret = btrfs_lookup_file_extent(trans, root, path, ino,
 					       cur_offset, 0);
-		BUG_ON(ret < 0);
+		if (ret < 0)
+			goto error;
 		if (ret > 0 && path->slots[0] > 0 && check_prev) {
 			leaf = path->nodes[0];
 			btrfs_item_key_to_cpu(leaf, &found_key,
 					      path->slots[0] - 1);
-			if (found_key.objectid == inode->i_ino &&
+			if (found_key.objectid == ino &&
 			    found_key.type == BTRFS_EXTENT_DATA_KEY)
 				path->slots[0]--;
 		}
@@ -1050,7 +1222,7 @@ next_slot:
 		if (path->slots[0] >= btrfs_header_nritems(leaf)) {
 			ret = btrfs_next_leaf(root, path);
 			if (ret < 0)
-				BUG_ON(1);
+				goto error;
 			if (ret > 0)
 				break;
 			leaf = path->nodes[0];
@@ -1061,7 +1233,7 @@ next_slot:
 		num_bytes = 0;
 		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
 
-		if (found_key.objectid > inode->i_ino ||
+		if (found_key.objectid > ino ||
 		    found_key.type > BTRFS_EXTENT_DATA_KEY ||
 		    found_key.offset > end)
 			break;
@@ -1076,12 +1248,15 @@ next_slot:
 				    struct btrfs_file_extent_item);
 		extent_type = btrfs_file_extent_type(leaf, fi);
 
+		ram_bytes = btrfs_file_extent_ram_bytes(leaf, fi);
 		if (extent_type == BTRFS_FILE_EXTENT_REG ||
 		    extent_type == BTRFS_FILE_EXTENT_PREALLOC) {
 			disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
 			extent_offset = btrfs_file_extent_offset(leaf, fi);
 			extent_end = found_key.offset +
 				btrfs_file_extent_num_bytes(leaf, fi);
+			disk_num_bytes =
+				btrfs_file_extent_disk_num_bytes(leaf, fi);
 			if (extent_end <= start) {
 				path->slots[0]++;
 				goto next_slot;
@@ -1096,7 +1271,7 @@ next_slot:
 				goto out_check;
 			if (btrfs_extent_readonly(root, disk_bytenr))
 				goto out_check;
-			if (btrfs_cross_ref_exist(trans, root, inode->i_ino,
+			if (btrfs_cross_ref_exist(trans, root, ino,
 						  found_key.offset -
 						  extent_offset, disk_bytenr))
 				goto out_check;
@@ -1104,6 +1279,15 @@ next_slot:
 			disk_bytenr += cur_offset - found_key.offset;
 			num_bytes = min(end + 1, extent_end) - cur_offset;
 			/*
+			 * if there are pending snapshots for this root,
+			 * we fall into common COW way.
+			 */
+			if (!nolock) {
+				err = btrfs_start_nocow_write(root);
+				if (!err)
+					goto out_check;
+			}
+			/*
 			 * force cow if csum exists in the range.
 			 * this ensure that csum for a given extent are
 			 * either valid or do not exist.
@@ -1113,7 +1297,8 @@ next_slot:
 			nocow = 1;
 		} else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
 			extent_end = found_key.offset +
-				btrfs_file_extent_inline_len(leaf, fi);
+				btrfs_file_extent_inline_len(leaf,
+						     path->slots[0], fi);
 			extent_end = ALIGN(extent_end, root->sectorsize);
 		} else {
 			BUG_ON(1);
@@ -1121,6 +1306,8 @@ next_slot:
 out_check:
 		if (extent_end <= start) {
 			path->slots[0]++;
+			if (!nolock && nocow)
+				btrfs_end_nocow_write(root);
 			goto next_slot;
 		}
 		if (!nocow) {
@@ -1133,12 +1320,16 @@ out_check:
 			goto next_slot;
 		}
 
-		btrfs_release_path(root, path);
+		btrfs_release_path(path);
 		if (cow_start != (u64)-1) {
-			ret = cow_file_range(inode, locked_page, cow_start,
-					found_key.offset - 1, page_started,
-					nr_written, 1);
-			BUG_ON(ret);
+			ret = cow_file_range(inode, locked_page,
+					     cow_start, found_key.offset - 1,
+					     page_started, nr_written, 1);
+			if (ret) {
+				if (!nolock && nocow)
+					btrfs_end_nocow_write(root);
+				goto error;
+			}
 			cow_start = (u64)-1;
 		}
 
@@ -1146,17 +1337,24 @@ out_check:
 			struct extent_map *em;
 			struct extent_map_tree *em_tree;
 			em_tree = &BTRFS_I(inode)->extent_tree;
-			em = alloc_extent_map(GFP_NOFS);
+			em = alloc_extent_map();
+			BUG_ON(!em); /* -ENOMEM */
 			em->start = cur_offset;
-			em->orig_start = em->start;
+			em->orig_start = found_key.offset - extent_offset;
 			em->len = num_bytes;
 			em->block_len = num_bytes;
 			em->block_start = disk_bytenr;
+			em->orig_block_len = disk_num_bytes;
+			em->ram_bytes = ram_bytes;
 			em->bdev = root->fs_info->fs_devices->latest_bdev;
+			em->mod_start = em->start;
+			em->mod_len = em->len;
 			set_bit(EXTENT_FLAG_PINNED, &em->flags);
+			set_bit(EXTENT_FLAG_FILLING, &em->flags);
+			em->generation = -1;
 			while (1) {
 				write_lock(&em_tree->lock);
-				ret = add_extent_mapping(em_tree, em);
+				ret = add_extent_mapping(em_tree, em, 1);
 				write_unlock(&em_tree->lock);
 				if (ret != -EEXIST) {
 					free_extent_map(em);
@@ -1172,31 +1370,59 @@ out_check:
 
 		ret = btrfs_add_ordered_extent(inode, cur_offset, disk_bytenr,
 					       num_bytes, num_bytes, type);
-		BUG_ON(ret);
+		BUG_ON(ret); /* -ENOMEM */
+
+		if (root->root_key.objectid ==
+		    BTRFS_DATA_RELOC_TREE_OBJECTID) {
+			ret = btrfs_reloc_clone_csums(inode, cur_offset,
+						      num_bytes);
+			if (ret) {
+				if (!nolock && nocow)
+					btrfs_end_nocow_write(root);
+				goto error;
+			}
+		}
 
-		extent_clear_unlock_delalloc(inode, &BTRFS_I(inode)->io_tree,
-				cur_offset, cur_offset + num_bytes - 1,
-				locked_page, EXTENT_CLEAR_UNLOCK_PAGE |
-				EXTENT_CLEAR_UNLOCK | EXTENT_CLEAR_DELALLOC |
-				EXTENT_SET_PRIVATE2);
+		extent_clear_unlock_delalloc(inode, cur_offset,
+					     cur_offset + num_bytes - 1,
+					     locked_page, EXTENT_LOCKED |
+					     EXTENT_DELALLOC, PAGE_UNLOCK |
+					     PAGE_SET_PRIVATE2);
+		if (!nolock && nocow)
+			btrfs_end_nocow_write(root);
 		cur_offset = extent_end;
 		if (cur_offset > end)
 			break;
 	}
-	btrfs_release_path(root, path);
+	btrfs_release_path(path);
 
-	if (cur_offset <= end && cow_start == (u64)-1)
+	if (cur_offset <= end && cow_start == (u64)-1) {
 		cow_start = cur_offset;
+		cur_offset = end;
+	}
+
 	if (cow_start != (u64)-1) {
 		ret = cow_file_range(inode, locked_page, cow_start, end,
 				     page_started, nr_written, 1);
-		BUG_ON(ret);
+		if (ret)
+			goto error;
 	}
 
-	ret = btrfs_end_transaction(trans, root);
-	BUG_ON(ret);
+error:
+	err = btrfs_end_transaction(trans, root);
+	if (!ret)
+		ret = err;
+
+	if (ret && cur_offset < end)
+		extent_clear_unlock_delalloc(inode, cur_offset, end,
+					     locked_page, EXTENT_LOCKED |
+					     EXTENT_DELALLOC | EXTENT_DEFRAG |
+					     EXTENT_DO_ACCOUNTING, PAGE_UNLOCK |
+					     PAGE_CLEAR_DIRTY |
+					     PAGE_SET_WRITEBACK |
+					     PAGE_END_WRITEBACK);
 	btrfs_free_path(path);
-	return 0;
+	return ret;
 }
 
 /*
@@ -1209,33 +1435,36 @@ static int run_delalloc_range(struct inode *inode, struct page *locked_page,
 	int ret;
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 
-	if (BTRFS_I(inode)->flags & BTRFS_INODE_NODATACOW)
+	if (BTRFS_I(inode)->flags & BTRFS_INODE_NODATACOW) {
 		ret = run_delalloc_nocow(inode, locked_page, start, end,
 					 page_started, 1, nr_written);
-	else if (BTRFS_I(inode)->flags & BTRFS_INODE_PREALLOC)
+	} else if (BTRFS_I(inode)->flags & BTRFS_INODE_PREALLOC) {
 		ret = run_delalloc_nocow(inode, locked_page, start, end,
 					 page_started, 0, nr_written);
-	else if (!btrfs_test_opt(root, COMPRESS) &&
-		 !(BTRFS_I(inode)->force_compress))
+	} else if (!btrfs_test_opt(root, COMPRESS) &&
+		   !(BTRFS_I(inode)->force_compress) &&
+		   !(BTRFS_I(inode)->flags & BTRFS_INODE_COMPRESS)) {
 		ret = cow_file_range(inode, locked_page, start, end,
 				      page_started, nr_written, 1);
-	else
+	} else {
+		set_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
+			&BTRFS_I(inode)->runtime_flags);
 		ret = cow_file_range_async(inode, locked_page, start, end,
 					   page_started, nr_written);
+	}
 	return ret;
 }
 
-static int btrfs_split_extent_hook(struct inode *inode,
+static void btrfs_split_extent_hook(struct inode *inode,
 				    struct extent_state *orig, u64 split)
 {
+	/* not delalloc, ignore it */
 	if (!(orig->state & EXTENT_DELALLOC))
-		return 0;
+		return;
 
-	spin_lock(&BTRFS_I(inode)->accounting_lock);
+	spin_lock(&BTRFS_I(inode)->lock);
 	BTRFS_I(inode)->outstanding_extents++;
-	spin_unlock(&BTRFS_I(inode)->accounting_lock);
-
-	return 0;
+	spin_unlock(&BTRFS_I(inode)->lock);
 }
 
 /*
@@ -1244,19 +1473,57 @@ static int btrfs_split_extent_hook(struct inode *inode,
  * extents, such as when we are doing sequential writes, so we can properly
  * account for the metadata space we'll need.
  */
-static int btrfs_merge_extent_hook(struct inode *inode,
-				   struct extent_state *new,
-				   struct extent_state *other)
+static void btrfs_merge_extent_hook(struct inode *inode,
+				    struct extent_state *new,
+				    struct extent_state *other)
 {
 	/* not delalloc, ignore it */
 	if (!(other->state & EXTENT_DELALLOC))
-		return 0;
+		return;
 
-	spin_lock(&BTRFS_I(inode)->accounting_lock);
+	spin_lock(&BTRFS_I(inode)->lock);
 	BTRFS_I(inode)->outstanding_extents--;
-	spin_unlock(&BTRFS_I(inode)->accounting_lock);
+	spin_unlock(&BTRFS_I(inode)->lock);
+}
 
-	return 0;
+static void btrfs_add_delalloc_inodes(struct btrfs_root *root,
+				      struct inode *inode)
+{
+	spin_lock(&root->delalloc_lock);
+	if (list_empty(&BTRFS_I(inode)->delalloc_inodes)) {
+		list_add_tail(&BTRFS_I(inode)->delalloc_inodes,
+			      &root->delalloc_inodes);
+		set_bit(BTRFS_INODE_IN_DELALLOC_LIST,
+			&BTRFS_I(inode)->runtime_flags);
+		root->nr_delalloc_inodes++;
+		if (root->nr_delalloc_inodes == 1) {
+			spin_lock(&root->fs_info->delalloc_root_lock);
+			BUG_ON(!list_empty(&root->delalloc_root));
+			list_add_tail(&root->delalloc_root,
+				      &root->fs_info->delalloc_roots);
+			spin_unlock(&root->fs_info->delalloc_root_lock);
+		}
+	}
+	spin_unlock(&root->delalloc_lock);
+}
+
+static void btrfs_del_delalloc_inode(struct btrfs_root *root,
+				     struct inode *inode)
+{
+	spin_lock(&root->delalloc_lock);
+	if (!list_empty(&BTRFS_I(inode)->delalloc_inodes)) {
+		list_del_init(&BTRFS_I(inode)->delalloc_inodes);
+		clear_bit(BTRFS_INODE_IN_DELALLOC_LIST,
+			  &BTRFS_I(inode)->runtime_flags);
+		root->nr_delalloc_inodes--;
+		if (!root->nr_delalloc_inodes) {
+			spin_lock(&root->fs_info->delalloc_root_lock);
+			BUG_ON(list_empty(&root->delalloc_root));
+			list_del_init(&root->delalloc_root);
+			spin_unlock(&root->fs_info->delalloc_root_lock);
+		}
+	}
+	spin_unlock(&root->delalloc_lock);
 }
 
 /*
@@ -1264,98 +1531,99 @@ static int btrfs_merge_extent_hook(struct inode *inode,
  * bytes in this file, and to maintain the list of inodes that
  * have pending delalloc work to be done.
  */
-static int btrfs_set_bit_hook(struct inode *inode, u64 start, u64 end,
-		       unsigned long old, unsigned long bits)
+static void btrfs_set_bit_hook(struct inode *inode,
+			       struct extent_state *state, unsigned long *bits)
 {
 
 	/*
 	 * set_bit and clear bit hooks normally require _irqsave/restore
-	 * but in this case, we are only testeing for the DELALLOC
+	 * but in this case, we are only testing for the DELALLOC
 	 * bit, which is only set or cleared with irqs on
 	 */
-	if (!(old & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) {
+	if (!(state->state & EXTENT_DELALLOC) && (*bits & EXTENT_DELALLOC)) {
 		struct btrfs_root *root = BTRFS_I(inode)->root;
+		u64 len = state->end + 1 - state->start;
+		bool do_list = !btrfs_is_free_space_inode(inode);
 
-		spin_lock(&BTRFS_I(inode)->accounting_lock);
-		BTRFS_I(inode)->outstanding_extents++;
-		spin_unlock(&BTRFS_I(inode)->accounting_lock);
-		btrfs_delalloc_reserve_space(root, inode, end - start + 1);
-
-		spin_lock(&root->fs_info->delalloc_lock);
-		BTRFS_I(inode)->delalloc_bytes += end - start + 1;
-		root->fs_info->delalloc_bytes += end - start + 1;
-		if (list_empty(&BTRFS_I(inode)->delalloc_inodes)) {
-			list_add_tail(&BTRFS_I(inode)->delalloc_inodes,
-				      &root->fs_info->delalloc_inodes);
+		if (*bits & EXTENT_FIRST_DELALLOC) {
+			*bits &= ~EXTENT_FIRST_DELALLOC;
+		} else {
+			spin_lock(&BTRFS_I(inode)->lock);
+			BTRFS_I(inode)->outstanding_extents++;
+			spin_unlock(&BTRFS_I(inode)->lock);
 		}
-		spin_unlock(&root->fs_info->delalloc_lock);
+
+		__percpu_counter_add(&root->fs_info->delalloc_bytes, len,
+				     root->fs_info->delalloc_batch);
+		spin_lock(&BTRFS_I(inode)->lock);
+		BTRFS_I(inode)->delalloc_bytes += len;
+		if (do_list && !test_bit(BTRFS_INODE_IN_DELALLOC_LIST,
+					 &BTRFS_I(inode)->runtime_flags))
+			btrfs_add_delalloc_inodes(root, inode);
+		spin_unlock(&BTRFS_I(inode)->lock);
 	}
-	return 0;
 }
 
 /*
  * extent_io.c clear_bit_hook, see set_bit_hook for why
  */
-static int btrfs_clear_bit_hook(struct inode *inode,
-				struct extent_state *state, unsigned long bits)
+static void btrfs_clear_bit_hook(struct inode *inode,
+				 struct extent_state *state,
+				 unsigned long *bits)
 {
 	/*
 	 * set_bit and clear bit hooks normally require _irqsave/restore
-	 * but in this case, we are only testeing for the DELALLOC
+	 * but in this case, we are only testing for the DELALLOC
 	 * bit, which is only set or cleared with irqs on
 	 */
-	if ((state->state & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) {
+	if ((state->state & EXTENT_DELALLOC) && (*bits & EXTENT_DELALLOC)) {
 		struct btrfs_root *root = BTRFS_I(inode)->root;
+		u64 len = state->end + 1 - state->start;
+		bool do_list = !btrfs_is_free_space_inode(inode);
 
-		if (bits & EXTENT_DO_ACCOUNTING) {
-			spin_lock(&BTRFS_I(inode)->accounting_lock);
-			WARN_ON(!BTRFS_I(inode)->outstanding_extents);
+		if (*bits & EXTENT_FIRST_DELALLOC) {
+			*bits &= ~EXTENT_FIRST_DELALLOC;
+		} else if (!(*bits & EXTENT_DO_ACCOUNTING)) {
+			spin_lock(&BTRFS_I(inode)->lock);
 			BTRFS_I(inode)->outstanding_extents--;
-			spin_unlock(&BTRFS_I(inode)->accounting_lock);
-			btrfs_unreserve_metadata_for_delalloc(root, inode, 1);
+			spin_unlock(&BTRFS_I(inode)->lock);
 		}
 
-		spin_lock(&root->fs_info->delalloc_lock);
-		if (state->end - state->start + 1 >
-		    root->fs_info->delalloc_bytes) {
-			printk(KERN_INFO "btrfs warning: delalloc account "
-			       "%llu %llu\n",
-			       (unsigned long long)
-			       state->end - state->start + 1,
-			       (unsigned long long)
-			       root->fs_info->delalloc_bytes);
-			btrfs_delalloc_free_space(root, inode, (u64)-1);
-			root->fs_info->delalloc_bytes = 0;
-			BTRFS_I(inode)->delalloc_bytes = 0;
-		} else {
-			btrfs_delalloc_free_space(root, inode,
-						  state->end -
-						  state->start + 1);
-			root->fs_info->delalloc_bytes -= state->end -
-				state->start + 1;
-			BTRFS_I(inode)->delalloc_bytes -= state->end -
-				state->start + 1;
-		}
-		if (BTRFS_I(inode)->delalloc_bytes == 0 &&
-		    !list_empty(&BTRFS_I(inode)->delalloc_inodes)) {
-			list_del_init(&BTRFS_I(inode)->delalloc_inodes);
-		}
-		spin_unlock(&root->fs_info->delalloc_lock);
+		/*
+		 * We don't reserve metadata space for space cache inodes so we
+		 * don't need to call dellalloc_release_metadata if there is an
+		 * error.
+		 */
+		if (*bits & EXTENT_DO_ACCOUNTING &&
+		    root != root->fs_info->tree_root)
+			btrfs_delalloc_release_metadata(inode, len);
+
+		if (root->root_key.objectid != BTRFS_DATA_RELOC_TREE_OBJECTID
+		    && do_list && !(state->state & EXTENT_NORESERVE))
+			btrfs_free_reserved_data_space(inode, len);
+
+		__percpu_counter_add(&root->fs_info->delalloc_bytes, -len,
+				     root->fs_info->delalloc_batch);
+		spin_lock(&BTRFS_I(inode)->lock);
+		BTRFS_I(inode)->delalloc_bytes -= len;
+		if (do_list && BTRFS_I(inode)->delalloc_bytes == 0 &&
+		    test_bit(BTRFS_INODE_IN_DELALLOC_LIST,
+			     &BTRFS_I(inode)->runtime_flags))
+			btrfs_del_delalloc_inode(root, inode);
+		spin_unlock(&BTRFS_I(inode)->lock);
 	}
-	return 0;
 }
 
 /*
  * extent_io.c merge_bio_hook, this must check the chunk tree to make sure
  * we don't create bios that span stripes or chunks
  */
-int btrfs_merge_bio_hook(struct page *page, unsigned long offset,
+int btrfs_merge_bio_hook(int rw, struct page *page, unsigned long offset,
 			 size_t size, struct bio *bio,
 			 unsigned long bio_flags)
 {
 	struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
-	struct btrfs_mapping_tree *map_tree;
-	u64 logical = (u64)bio->bi_sector << 9;
+	u64 logical = (u64)bio->bi_iter.bi_sector << 9;
 	u64 length = 0;
 	u64 map_length;
 	int ret;
@@ -1363,12 +1631,12 @@ int btrfs_merge_bio_hook(struct page *page, unsigned long offset,
 	if (bio_flags & EXTENT_BIO_COMPRESSED)
 		return 0;
 
-	length = bio->bi_size;
-	map_tree = &root->fs_info->mapping_tree;
+	length = bio->bi_iter.bi_size;
 	map_length = length;
-	ret = btrfs_map_block(map_tree, READ, logical,
+	ret = btrfs_map_block(root->fs_info, rw, logical,
 			      &map_length, NULL, 0);
-
+	/* Will always return 0 with map_multi == NULL */
+	BUG_ON(ret < 0);
 	if (map_length < length + size)
 		return 1;
 	return 0;
@@ -1384,13 +1652,14 @@ int btrfs_merge_bio_hook(struct page *page, unsigned long offset,
  */
 static int __btrfs_submit_bio_start(struct inode *inode, int rw,
 				    struct bio *bio, int mirror_num,
-				    unsigned long bio_flags)
+				    unsigned long bio_flags,
+				    u64 bio_offset)
 {
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	int ret = 0;
 
 	ret = btrfs_csum_one_bio(root, inode, bio, 0, 0);
-	BUG_ON(ret);
+	BUG_ON(ret); /* -ENOMEM */
 	return 0;
 }
 
@@ -1403,10 +1672,16 @@ static int __btrfs_submit_bio_start(struct inode *inode, int rw,
  * are inserted into the btree
  */
 static int __btrfs_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
-			  int mirror_num, unsigned long bio_flags)
+			  int mirror_num, unsigned long bio_flags,
+			  u64 bio_offset)
 {
 	struct btrfs_root *root = BTRFS_I(inode)->root;
-	return btrfs_map_bio(root, rw, bio, mirror_num, 1);
+	int ret;
+
+	ret = btrfs_map_bio(root, rw, bio, mirror_num, 1);
+	if (ret)
+		bio_endio(bio, ret);
+	return ret;
 }
 
 /*
@@ -1414,37 +1689,60 @@ static int __btrfs_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
  * on write, or reading the csums from the tree before a read
  */
 static int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
-			  int mirror_num, unsigned long bio_flags)
+			  int mirror_num, unsigned long bio_flags,
+			  u64 bio_offset)
 {
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	int ret = 0;
 	int skip_sum;
+	int metadata = 0;
+	int async = !atomic_read(&BTRFS_I(inode)->sync_writers);
 
 	skip_sum = BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM;
 
-	ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0);
-	BUG_ON(ret);
+	if (btrfs_is_free_space_inode(inode))
+		metadata = 2;
+
+	if (!(rw & REQ_WRITE)) {
+		ret = btrfs_bio_wq_end_io(root->fs_info, bio, metadata);
+		if (ret)
+			goto out;
 
-	if (!(rw & (1 << BIO_RW))) {
 		if (bio_flags & EXTENT_BIO_COMPRESSED) {
-			return btrfs_submit_compressed_read(inode, bio,
-						    mirror_num, bio_flags);
-		} else if (!skip_sum)
-			btrfs_lookup_bio_sums(root, inode, bio, NULL);
+			ret = btrfs_submit_compressed_read(inode, bio,
+							   mirror_num,
+							   bio_flags);
+			goto out;
+		} else if (!skip_sum) {
+			ret = btrfs_lookup_bio_sums(root, inode, bio, NULL);
+			if (ret)
+				goto out;
+		}
 		goto mapit;
-	} else if (!skip_sum) {
+	} else if (async && !skip_sum) {
 		/* csum items have already been cloned */
 		if (root->root_key.objectid == BTRFS_DATA_RELOC_TREE_OBJECTID)
 			goto mapit;
 		/* we're doing a write, do the async checksumming */
-		return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
+		ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
 				   inode, rw, bio, mirror_num,
-				   bio_flags, __btrfs_submit_bio_start,
+				   bio_flags, bio_offset,
+				   __btrfs_submit_bio_start,
 				   __btrfs_submit_bio_done);
+		goto out;
+	} else if (!skip_sum) {
+		ret = btrfs_csum_one_bio(root, inode, bio, 0, 0);
+		if (ret)
+			goto out;
 	}
 
 mapit:
-	return btrfs_map_bio(root, rw, bio, mirror_num, 0);
+	ret = btrfs_map_bio(root, rw, bio, mirror_num, 0);
+
+out:
+	if (ret < 0)
+		bio_endio(bio, ret);
+	return ret;
 }
 
 /*
@@ -1457,11 +1755,11 @@ static noinline int add_pending_csums(struct btrfs_trans_handle *trans,
 {
 	struct btrfs_ordered_sum *sum;
 
-	btrfs_set_trans_block_group(trans, inode);
-
 	list_for_each_entry(sum, list, list) {
+		trans->adding_csums = 1;
 		btrfs_csum_file_blocks(trans,
 		       BTRFS_I(inode)->root->fs_info->csum_root, sum);
+		trans->adding_csums = 0;
 	}
 	return 0;
 }
@@ -1469,8 +1767,7 @@ static noinline int add_pending_csums(struct btrfs_trans_handle *trans,
 int btrfs_set_extent_delalloc(struct inode *inode, u64 start, u64 end,
 			      struct extent_state **cached_state)
 {
-	if ((end & (PAGE_CACHE_SIZE - 1)) == 0)
-		WARN_ON(1);
+	WARN_ON((end & (PAGE_CACHE_SIZE - 1)) == 0);
 	return set_extent_delalloc(&BTRFS_I(inode)->io_tree, start, end,
 				   cached_state, GFP_NOFS);
 }
@@ -1490,6 +1787,7 @@ static void btrfs_writepage_fixup_worker(struct btrfs_work *work)
 	struct inode *inode;
 	u64 page_start;
 	u64 page_end;
+	int ret;
 
 	fixup = container_of(work, struct btrfs_writepage_fixup, work);
 	page = fixup->page;
@@ -1505,7 +1803,7 @@ again:
 	page_end = page_offset(page) + PAGE_CACHE_SIZE - 1;
 
 	lock_extent_bits(&BTRFS_I(inode)->io_tree, page_start, page_end, 0,
-			 &cached_state, GFP_NOFS);
+			 &cached_state);
 
 	/* already ordered? We're done */
 	if (PagePrivate2(page))
@@ -1517,17 +1815,28 @@ again:
 				     page_end, &cached_state, GFP_NOFS);
 		unlock_page(page);
 		btrfs_start_ordered_extent(inode, ordered, 1);
+		btrfs_put_ordered_extent(ordered);
 		goto again;
 	}
 
+	ret = btrfs_delalloc_reserve_space(inode, PAGE_CACHE_SIZE);
+	if (ret) {
+		mapping_set_error(page->mapping, ret);
+		end_extent_writepage(page, ret, page_start, page_end);
+		ClearPageChecked(page);
+		goto out;
+	 }
+
 	btrfs_set_extent_delalloc(inode, page_start, page_end, &cached_state);
 	ClearPageChecked(page);
+	set_page_dirty(page);
 out:
 	unlock_extent_cached(&BTRFS_I(inode)->io_tree, page_start, page_end,
 			     &cached_state, GFP_NOFS);
 out_page:
 	unlock_page(page);
 	page_cache_release(page);
+	kfree(fixup);
 }
 
 /*
@@ -1560,10 +1869,10 @@ static int btrfs_writepage_start_hook(struct page *page, u64 start, u64 end)
 
 	SetPageChecked(page);
 	page_cache_get(page);
-	fixup->work.func = btrfs_writepage_fixup_worker;
+	btrfs_init_work(&fixup->work, btrfs_writepage_fixup_worker, NULL, NULL);
 	fixup->page = page;
-	btrfs_queue_worker(&root->fs_info->fixup_workers, &fixup->work);
-	return -EAGAIN;
+	btrfs_queue_work(root->fs_info->fixup_workers, &fixup->work);
+	return -EBUSY;
 }
 
 static int insert_reserved_file_extent(struct btrfs_trans_handle *trans,
@@ -1578,13 +1887,12 @@ static int insert_reserved_file_extent(struct btrfs_trans_handle *trans,
 	struct btrfs_path *path;
 	struct extent_buffer *leaf;
 	struct btrfs_key ins;
-	u64 hint;
+	int extent_inserted = 0;
 	int ret;
 
 	path = btrfs_alloc_path();
-	BUG_ON(!path);
-
-	path->leave_spinning = 1;
+	if (!path)
+		return -ENOMEM;
 
 	/*
 	 * we may be replacing one extent in the tree with another.
@@ -1595,15 +1903,23 @@ static int insert_reserved_file_extent(struct btrfs_trans_handle *trans,
 	 * the caller is expected to unpin it and allow it to be merged
 	 * with the others.
 	 */
-	ret = btrfs_drop_extents(trans, inode, file_pos, file_pos + num_bytes,
-				 &hint, 0);
-	BUG_ON(ret);
+	ret = __btrfs_drop_extents(trans, root, inode, path, file_pos,
+				   file_pos + num_bytes, NULL, 0,
+				   1, sizeof(*fi), &extent_inserted);
+	if (ret)
+		goto out;
 
-	ins.objectid = inode->i_ino;
-	ins.offset = file_pos;
-	ins.type = BTRFS_EXTENT_DATA_KEY;
-	ret = btrfs_insert_empty_item(trans, root, path, &ins, sizeof(*fi));
-	BUG_ON(ret);
+	if (!extent_inserted) {
+		ins.objectid = btrfs_ino(inode);
+		ins.offset = file_pos;
+		ins.type = BTRFS_EXTENT_DATA_KEY;
+
+		path->leave_spinning = 1;
+		ret = btrfs_insert_empty_item(trans, root, path, &ins,
+					      sizeof(*fi));
+		if (ret)
+			goto out;
+	}
 	leaf = path->nodes[0];
 	fi = btrfs_item_ptr(leaf, path->slots[0],
 			    struct btrfs_file_extent_item);
@@ -1618,10 +1934,8 @@ static int insert_reserved_file_extent(struct btrfs_trans_handle *trans,
 	btrfs_set_file_extent_encryption(leaf, fi, encryption);
 	btrfs_set_file_extent_other_encoding(leaf, fi, other_encoding);
 
-	btrfs_unlock_up_safe(path, 1);
-	btrfs_set_lock_blocking(leaf);
-
 	btrfs_mark_buffer_dirty(leaf);
+	btrfs_release_path(path);
 
 	inode_add_bytes(inode, num_bytes);
 
@@ -1630,266 +1944,910 @@ static int insert_reserved_file_extent(struct btrfs_trans_handle *trans,
 	ins.type = BTRFS_EXTENT_ITEM_KEY;
 	ret = btrfs_alloc_reserved_file_extent(trans, root,
 					root->root_key.objectid,
-					inode->i_ino, file_pos, &ins);
-	BUG_ON(ret);
+					btrfs_ino(inode), file_pos, &ins);
+out:
 	btrfs_free_path(path);
 
+	return ret;
+}
+
+/* snapshot-aware defrag */
+struct sa_defrag_extent_backref {
+	struct rb_node node;
+	struct old_sa_defrag_extent *old;
+	u64 root_id;
+	u64 inum;
+	u64 file_pos;
+	u64 extent_offset;
+	u64 num_bytes;
+	u64 generation;
+};
+
+struct old_sa_defrag_extent {
+	struct list_head list;
+	struct new_sa_defrag_extent *new;
+
+	u64 extent_offset;
+	u64 bytenr;
+	u64 offset;
+	u64 len;
+	int count;
+};
+
+struct new_sa_defrag_extent {
+	struct rb_root root;
+	struct list_head head;
+	struct btrfs_path *path;
+	struct inode *inode;
+	u64 file_pos;
+	u64 len;
+	u64 bytenr;
+	u64 disk_len;
+	u8 compress_type;
+};
+
+static int backref_comp(struct sa_defrag_extent_backref *b1,
+			struct sa_defrag_extent_backref *b2)
+{
+	if (b1->root_id < b2->root_id)
+		return -1;
+	else if (b1->root_id > b2->root_id)
+		return 1;
+
+	if (b1->inum < b2->inum)
+		return -1;
+	else if (b1->inum > b2->inum)
+		return 1;
+
+	if (b1->file_pos < b2->file_pos)
+		return -1;
+	else if (b1->file_pos > b2->file_pos)
+		return 1;
+
+	/*
+	 * [------------------------------] ===> (a range of space)
+	 *     |<--->|   |<---->| =============> (fs/file tree A)
+	 * |<---------------------------->| ===> (fs/file tree B)
+	 *
+	 * A range of space can refer to two file extents in one tree while
+	 * refer to only one file extent in another tree.
+	 *
+	 * So we may process a disk offset more than one time(two extents in A)
+	 * and locate at the same extent(one extent in B), then insert two same
+	 * backrefs(both refer to the extent in B).
+	 */
 	return 0;
 }
 
+static void backref_insert(struct rb_root *root,
+			   struct sa_defrag_extent_backref *backref)
+{
+	struct rb_node **p = &root->rb_node;
+	struct rb_node *parent = NULL;
+	struct sa_defrag_extent_backref *entry;
+	int ret;
+
+	while (*p) {
+		parent = *p;
+		entry = rb_entry(parent, struct sa_defrag_extent_backref, node);
+
+		ret = backref_comp(backref, entry);
+		if (ret < 0)
+			p = &(*p)->rb_left;
+		else
+			p = &(*p)->rb_right;
+	}
+
+	rb_link_node(&backref->node, parent, p);
+	rb_insert_color(&backref->node, root);
+}
+
+/*
+ * Note the backref might has changed, and in this case we just return 0.
+ */
+static noinline int record_one_backref(u64 inum, u64 offset, u64 root_id,
+				       void *ctx)
+{
+	struct btrfs_file_extent_item *extent;
+	struct btrfs_fs_info *fs_info;
+	struct old_sa_defrag_extent *old = ctx;
+	struct new_sa_defrag_extent *new = old->new;
+	struct btrfs_path *path = new->path;
+	struct btrfs_key key;
+	struct btrfs_root *root;
+	struct sa_defrag_extent_backref *backref;
+	struct extent_buffer *leaf;
+	struct inode *inode = new->inode;
+	int slot;
+	int ret;
+	u64 extent_offset;
+	u64 num_bytes;
+
+	if (BTRFS_I(inode)->root->root_key.objectid == root_id &&
+	    inum == btrfs_ino(inode))
+		return 0;
+
+	key.objectid = root_id;
+	key.type = BTRFS_ROOT_ITEM_KEY;
+	key.offset = (u64)-1;
+
+	fs_info = BTRFS_I(inode)->root->fs_info;
+	root = btrfs_read_fs_root_no_name(fs_info, &key);
+	if (IS_ERR(root)) {
+		if (PTR_ERR(root) == -ENOENT)
+			return 0;
+		WARN_ON(1);
+		pr_debug("inum=%llu, offset=%llu, root_id=%llu\n",
+			 inum, offset, root_id);
+		return PTR_ERR(root);
+	}
+
+	key.objectid = inum;
+	key.type = BTRFS_EXTENT_DATA_KEY;
+	if (offset > (u64)-1 << 32)
+		key.offset = 0;
+	else
+		key.offset = offset;
+
+	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+	if (WARN_ON(ret < 0))
+		return ret;
+	ret = 0;
+
+	while (1) {
+		cond_resched();
+
+		leaf = path->nodes[0];
+		slot = path->slots[0];
+
+		if (slot >= btrfs_header_nritems(leaf)) {
+			ret = btrfs_next_leaf(root, path);
+			if (ret < 0) {
+				goto out;
+			} else if (ret > 0) {
+				ret = 0;
+				goto out;
+			}
+			continue;
+		}
+
+		path->slots[0]++;
+
+		btrfs_item_key_to_cpu(leaf, &key, slot);
+
+		if (key.objectid > inum)
+			goto out;
+
+		if (key.objectid < inum || key.type != BTRFS_EXTENT_DATA_KEY)
+			continue;
+
+		extent = btrfs_item_ptr(leaf, slot,
+					struct btrfs_file_extent_item);
+
+		if (btrfs_file_extent_disk_bytenr(leaf, extent) != old->bytenr)
+			continue;
+
+		/*
+		 * 'offset' refers to the exact key.offset,
+		 * NOT the 'offset' field in btrfs_extent_data_ref, ie.
+		 * (key.offset - extent_offset).
+		 */
+		if (key.offset != offset)
+			continue;
+
+		extent_offset = btrfs_file_extent_offset(leaf, extent);
+		num_bytes = btrfs_file_extent_num_bytes(leaf, extent);
+
+		if (extent_offset >= old->extent_offset + old->offset +
+		    old->len || extent_offset + num_bytes <=
+		    old->extent_offset + old->offset)
+			continue;
+		break;
+	}
+
+	backref = kmalloc(sizeof(*backref), GFP_NOFS);
+	if (!backref) {
+		ret = -ENOENT;
+		goto out;
+	}
+
+	backref->root_id = root_id;
+	backref->inum = inum;
+	backref->file_pos = offset;
+	backref->num_bytes = num_bytes;
+	backref->extent_offset = extent_offset;
+	backref->generation = btrfs_file_extent_generation(leaf, extent);
+	backref->old = old;
+	backref_insert(&new->root, backref);
+	old->count++;
+out:
+	btrfs_release_path(path);
+	WARN_ON(ret);
+	return ret;
+}
+
+static noinline bool record_extent_backrefs(struct btrfs_path *path,
+				   struct new_sa_defrag_extent *new)
+{
+	struct btrfs_fs_info *fs_info = BTRFS_I(new->inode)->root->fs_info;
+	struct old_sa_defrag_extent *old, *tmp;
+	int ret;
+
+	new->path = path;
+
+	list_for_each_entry_safe(old, tmp, &new->head, list) {
+		ret = iterate_inodes_from_logical(old->bytenr +
+						  old->extent_offset, fs_info,
+						  path, record_one_backref,
+						  old);
+		if (ret < 0 && ret != -ENOENT)
+			return false;
+
+		/* no backref to be processed for this extent */
+		if (!old->count) {
+			list_del(&old->list);
+			kfree(old);
+		}
+	}
+
+	if (list_empty(&new->head))
+		return false;
+
+	return true;
+}
+
+static int relink_is_mergable(struct extent_buffer *leaf,
+			      struct btrfs_file_extent_item *fi,
+			      struct new_sa_defrag_extent *new)
+{
+	if (btrfs_file_extent_disk_bytenr(leaf, fi) != new->bytenr)
+		return 0;
+
+	if (btrfs_file_extent_type(leaf, fi) != BTRFS_FILE_EXTENT_REG)
+		return 0;
+
+	if (btrfs_file_extent_compression(leaf, fi) != new->compress_type)
+		return 0;
+
+	if (btrfs_file_extent_encryption(leaf, fi) ||
+	    btrfs_file_extent_other_encoding(leaf, fi))
+		return 0;
+
+	return 1;
+}
+
 /*
- * helper function for btrfs_finish_ordered_io, this
- * just reads in some of the csum leaves to prime them into ram
- * before we start the transaction.  It limits the amount of btree
- * reads required while inside the transaction.
+ * Note the backref might has changed, and in this case we just return 0.
  */
+static noinline int relink_extent_backref(struct btrfs_path *path,
+				 struct sa_defrag_extent_backref *prev,
+				 struct sa_defrag_extent_backref *backref)
+{
+	struct btrfs_file_extent_item *extent;
+	struct btrfs_file_extent_item *item;
+	struct btrfs_ordered_extent *ordered;
+	struct btrfs_trans_handle *trans;
+	struct btrfs_fs_info *fs_info;
+	struct btrfs_root *root;
+	struct btrfs_key key;
+	struct extent_buffer *leaf;
+	struct old_sa_defrag_extent *old = backref->old;
+	struct new_sa_defrag_extent *new = old->new;
+	struct inode *src_inode = new->inode;
+	struct inode *inode;
+	struct extent_state *cached = NULL;
+	int ret = 0;
+	u64 start;
+	u64 len;
+	u64 lock_start;
+	u64 lock_end;
+	bool merge = false;
+	int index;
+
+	if (prev && prev->root_id == backref->root_id &&
+	    prev->inum == backref->inum &&
+	    prev->file_pos + prev->num_bytes == backref->file_pos)
+		merge = true;
+
+	/* step 1: get root */
+	key.objectid = backref->root_id;
+	key.type = BTRFS_ROOT_ITEM_KEY;
+	key.offset = (u64)-1;
+
+	fs_info = BTRFS_I(src_inode)->root->fs_info;
+	index = srcu_read_lock(&fs_info->subvol_srcu);
+
+	root = btrfs_read_fs_root_no_name(fs_info, &key);
+	if (IS_ERR(root)) {
+		srcu_read_unlock(&fs_info->subvol_srcu, index);
+		if (PTR_ERR(root) == -ENOENT)
+			return 0;
+		return PTR_ERR(root);
+	}
+
+	if (btrfs_root_readonly(root)) {
+		srcu_read_unlock(&fs_info->subvol_srcu, index);
+		return 0;
+	}
+
+	/* step 2: get inode */
+	key.objectid = backref->inum;
+	key.type = BTRFS_INODE_ITEM_KEY;
+	key.offset = 0;
+
+	inode = btrfs_iget(fs_info->sb, &key, root, NULL);
+	if (IS_ERR(inode)) {
+		srcu_read_unlock(&fs_info->subvol_srcu, index);
+		return 0;
+	}
+
+	srcu_read_unlock(&fs_info->subvol_srcu, index);
+
+	/* step 3: relink backref */
+	lock_start = backref->file_pos;
+	lock_end = backref->file_pos + backref->num_bytes - 1;
+	lock_extent_bits(&BTRFS_I(inode)->io_tree, lock_start, lock_end,
+			 0, &cached);
+
+	ordered = btrfs_lookup_first_ordered_extent(inode, lock_end);
+	if (ordered) {
+		btrfs_put_ordered_extent(ordered);
+		goto out_unlock;
+	}
+
+	trans = btrfs_join_transaction(root);
+	if (IS_ERR(trans)) {
+		ret = PTR_ERR(trans);
+		goto out_unlock;
+	}
+
+	key.objectid = backref->inum;
+	key.type = BTRFS_EXTENT_DATA_KEY;
+	key.offset = backref->file_pos;
+
+	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+	if (ret < 0) {
+		goto out_free_path;
+	} else if (ret > 0) {
+		ret = 0;
+		goto out_free_path;
+	}
+
+	extent = btrfs_item_ptr(path->nodes[0], path->slots[0],
+				struct btrfs_file_extent_item);
+
+	if (btrfs_file_extent_generation(path->nodes[0], extent) !=
+	    backref->generation)
+		goto out_free_path;
+
+	btrfs_release_path(path);
+
+	start = backref->file_pos;
+	if (backref->extent_offset < old->extent_offset + old->offset)
+		start += old->extent_offset + old->offset -
+			 backref->extent_offset;
+
+	len = min(backref->extent_offset + backref->num_bytes,
+		  old->extent_offset + old->offset + old->len);
+	len -= max(backref->extent_offset, old->extent_offset + old->offset);
+
+	ret = btrfs_drop_extents(trans, root, inode, start,
+				 start + len, 1);
+	if (ret)
+		goto out_free_path;
+again:
+	key.objectid = btrfs_ino(inode);
+	key.type = BTRFS_EXTENT_DATA_KEY;
+	key.offset = start;
+
+	path->leave_spinning = 1;
+	if (merge) {
+		struct btrfs_file_extent_item *fi;
+		u64 extent_len;
+		struct btrfs_key found_key;
+
+		ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
+		if (ret < 0)
+			goto out_free_path;
+
+		path->slots[0]--;
+		leaf = path->nodes[0];
+		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
+
+		fi = btrfs_item_ptr(leaf, path->slots[0],
+				    struct btrfs_file_extent_item);
+		extent_len = btrfs_file_extent_num_bytes(leaf, fi);
+
+		if (extent_len + found_key.offset == start &&
+		    relink_is_mergable(leaf, fi, new)) {
+			btrfs_set_file_extent_num_bytes(leaf, fi,
+							extent_len + len);
+			btrfs_mark_buffer_dirty(leaf);
+			inode_add_bytes(inode, len);
+
+			ret = 1;
+			goto out_free_path;
+		} else {
+			merge = false;
+			btrfs_release_path(path);
+			goto again;
+		}
+	}
+
+	ret = btrfs_insert_empty_item(trans, root, path, &key,
+					sizeof(*extent));
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
+		goto out_free_path;
+	}
+
+	leaf = path->nodes[0];
+	item = btrfs_item_ptr(leaf, path->slots[0],
+				struct btrfs_file_extent_item);
+	btrfs_set_file_extent_disk_bytenr(leaf, item, new->bytenr);
+	btrfs_set_file_extent_disk_num_bytes(leaf, item, new->disk_len);
+	btrfs_set_file_extent_offset(leaf, item, start - new->file_pos);
+	btrfs_set_file_extent_num_bytes(leaf, item, len);
+	btrfs_set_file_extent_ram_bytes(leaf, item, new->len);
+	btrfs_set_file_extent_generation(leaf, item, trans->transid);
+	btrfs_set_file_extent_type(leaf, item, BTRFS_FILE_EXTENT_REG);
+	btrfs_set_file_extent_compression(leaf, item, new->compress_type);
+	btrfs_set_file_extent_encryption(leaf, item, 0);
+	btrfs_set_file_extent_other_encoding(leaf, item, 0);
+
+	btrfs_mark_buffer_dirty(leaf);
+	inode_add_bytes(inode, len);
+	btrfs_release_path(path);
+
+	ret = btrfs_inc_extent_ref(trans, root, new->bytenr,
+			new->disk_len, 0,
+			backref->root_id, backref->inum,
+			new->file_pos, 0);	/* start - extent_offset */
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
+		goto out_free_path;
+	}
+
+	ret = 1;
+out_free_path:
+	btrfs_release_path(path);
+	path->leave_spinning = 0;
+	btrfs_end_transaction(trans, root);
+out_unlock:
+	unlock_extent_cached(&BTRFS_I(inode)->io_tree, lock_start, lock_end,
+			     &cached, GFP_NOFS);
+	iput(inode);
+	return ret;
+}
+
+static void free_sa_defrag_extent(struct new_sa_defrag_extent *new)
+{
+	struct old_sa_defrag_extent *old, *tmp;
+
+	if (!new)
+		return;
+
+	list_for_each_entry_safe(old, tmp, &new->head, list) {
+		list_del(&old->list);
+		kfree(old);
+	}
+	kfree(new);
+}
+
+static void relink_file_extents(struct new_sa_defrag_extent *new)
+{
+	struct btrfs_path *path;
+	struct sa_defrag_extent_backref *backref;
+	struct sa_defrag_extent_backref *prev = NULL;
+	struct inode *inode;
+	struct btrfs_root *root;
+	struct rb_node *node;
+	int ret;
+
+	inode = new->inode;
+	root = BTRFS_I(inode)->root;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return;
+
+	if (!record_extent_backrefs(path, new)) {
+		btrfs_free_path(path);
+		goto out;
+	}
+	btrfs_release_path(path);
+
+	while (1) {
+		node = rb_first(&new->root);
+		if (!node)
+			break;
+		rb_erase(node, &new->root);
+
+		backref = rb_entry(node, struct sa_defrag_extent_backref, node);
+
+		ret = relink_extent_backref(path, prev, backref);
+		WARN_ON(ret < 0);
+
+		kfree(prev);
+
+		if (ret == 1)
+			prev = backref;
+		else
+			prev = NULL;
+		cond_resched();
+	}
+	kfree(prev);
+
+	btrfs_free_path(path);
+out:
+	free_sa_defrag_extent(new);
+
+	atomic_dec(&root->fs_info->defrag_running);
+	wake_up(&root->fs_info->transaction_wait);
+}
+
+static struct new_sa_defrag_extent *
+record_old_file_extents(struct inode *inode,
+			struct btrfs_ordered_extent *ordered)
+{
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct btrfs_path *path;
+	struct btrfs_key key;
+	struct old_sa_defrag_extent *old;
+	struct new_sa_defrag_extent *new;
+	int ret;
+
+	new = kmalloc(sizeof(*new), GFP_NOFS);
+	if (!new)
+		return NULL;
+
+	new->inode = inode;
+	new->file_pos = ordered->file_offset;
+	new->len = ordered->len;
+	new->bytenr = ordered->start;
+	new->disk_len = ordered->disk_len;
+	new->compress_type = ordered->compress_type;
+	new->root = RB_ROOT;
+	INIT_LIST_HEAD(&new->head);
+
+	path = btrfs_alloc_path();
+	if (!path)
+		goto out_kfree;
+
+	key.objectid = btrfs_ino(inode);
+	key.type = BTRFS_EXTENT_DATA_KEY;
+	key.offset = new->file_pos;
+
+	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+	if (ret < 0)
+		goto out_free_path;
+	if (ret > 0 && path->slots[0] > 0)
+		path->slots[0]--;
+
+	/* find out all the old extents for the file range */
+	while (1) {
+		struct btrfs_file_extent_item *extent;
+		struct extent_buffer *l;
+		int slot;
+		u64 num_bytes;
+		u64 offset;
+		u64 end;
+		u64 disk_bytenr;
+		u64 extent_offset;
+
+		l = path->nodes[0];
+		slot = path->slots[0];
+
+		if (slot >= btrfs_header_nritems(l)) {
+			ret = btrfs_next_leaf(root, path);
+			if (ret < 0)
+				goto out_free_path;
+			else if (ret > 0)
+				break;
+			continue;
+		}
+
+		btrfs_item_key_to_cpu(l, &key, slot);
+
+		if (key.objectid != btrfs_ino(inode))
+			break;
+		if (key.type != BTRFS_EXTENT_DATA_KEY)
+			break;
+		if (key.offset >= new->file_pos + new->len)
+			break;
+
+		extent = btrfs_item_ptr(l, slot, struct btrfs_file_extent_item);
+
+		num_bytes = btrfs_file_extent_num_bytes(l, extent);
+		if (key.offset + num_bytes < new->file_pos)
+			goto next;
+
+		disk_bytenr = btrfs_file_extent_disk_bytenr(l, extent);
+		if (!disk_bytenr)
+			goto next;
+
+		extent_offset = btrfs_file_extent_offset(l, extent);
+
+		old = kmalloc(sizeof(*old), GFP_NOFS);
+		if (!old)
+			goto out_free_path;
+
+		offset = max(new->file_pos, key.offset);
+		end = min(new->file_pos + new->len, key.offset + num_bytes);
+
+		old->bytenr = disk_bytenr;
+		old->extent_offset = extent_offset;
+		old->offset = offset - key.offset;
+		old->len = end - offset;
+		old->new = new;
+		old->count = 0;
+		list_add_tail(&old->list, &new->head);
+next:
+		path->slots[0]++;
+		cond_resched();
+	}
+
+	btrfs_free_path(path);
+	atomic_inc(&root->fs_info->defrag_running);
+
+	return new;
+
+out_free_path:
+	btrfs_free_path(path);
+out_kfree:
+	free_sa_defrag_extent(new);
+	return NULL;
+}
+
+static void btrfs_release_delalloc_bytes(struct btrfs_root *root,
+					 u64 start, u64 len)
+{
+	struct btrfs_block_group_cache *cache;
+
+	cache = btrfs_lookup_block_group(root->fs_info, start);
+	ASSERT(cache);
+
+	spin_lock(&cache->lock);
+	cache->delalloc_bytes -= len;
+	spin_unlock(&cache->lock);
+
+	btrfs_put_block_group(cache);
+}
+
 /* as ordered data IO finishes, this gets called so we can finish
  * an ordered extent if the range of bytes in the file it covers are
  * fully written.
  */
-static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end)
+static int btrfs_finish_ordered_io(struct btrfs_ordered_extent *ordered_extent)
 {
+	struct inode *inode = ordered_extent->inode;
 	struct btrfs_root *root = BTRFS_I(inode)->root;
-	struct btrfs_trans_handle *trans;
-	struct btrfs_ordered_extent *ordered_extent = NULL;
+	struct btrfs_trans_handle *trans = NULL;
 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
 	struct extent_state *cached_state = NULL;
-	int compressed = 0;
-	int ret;
+	struct new_sa_defrag_extent *new = NULL;
+	int compress_type = 0;
+	int ret = 0;
+	u64 logical_len = ordered_extent->len;
+	bool nolock;
+	bool truncated = false;
 
-	ret = btrfs_dec_test_ordered_pending(inode, &ordered_extent, start,
-					     end - start + 1);
-	if (!ret)
-		return 0;
-	BUG_ON(!ordered_extent);
+	nolock = btrfs_is_free_space_inode(inode);
+
+	if (test_bit(BTRFS_ORDERED_IOERR, &ordered_extent->flags)) {
+		ret = -EIO;
+		goto out;
+	}
+
+	if (test_bit(BTRFS_ORDERED_TRUNCATED, &ordered_extent->flags)) {
+		truncated = true;
+		logical_len = ordered_extent->truncated_len;
+		/* Truncated the entire extent, don't bother adding */
+		if (!logical_len)
+			goto out;
+	}
 
 	if (test_bit(BTRFS_ORDERED_NOCOW, &ordered_extent->flags)) {
-		BUG_ON(!list_empty(&ordered_extent->list));
-		ret = btrfs_ordered_update_i_size(inode, 0, ordered_extent);
-		if (!ret) {
-			trans = btrfs_join_transaction(root, 1);
-			ret = btrfs_update_inode(trans, root, inode);
-			BUG_ON(ret);
-			btrfs_end_transaction(trans, root);
+		BUG_ON(!list_empty(&ordered_extent->list)); /* Logic error */
+		btrfs_ordered_update_i_size(inode, 0, ordered_extent);
+		if (nolock)
+			trans = btrfs_join_transaction_nolock(root);
+		else
+			trans = btrfs_join_transaction(root);
+		if (IS_ERR(trans)) {
+			ret = PTR_ERR(trans);
+			trans = NULL;
+			goto out;
 		}
+		trans->block_rsv = &root->fs_info->delalloc_block_rsv;
+		ret = btrfs_update_inode_fallback(trans, root, inode);
+		if (ret) /* -ENOMEM or corruption */
+			btrfs_abort_transaction(trans, root, ret);
 		goto out;
 	}
 
 	lock_extent_bits(io_tree, ordered_extent->file_offset,
 			 ordered_extent->file_offset + ordered_extent->len - 1,
-			 0, &cached_state, GFP_NOFS);
+			 0, &cached_state);
+
+	ret = test_range_bit(io_tree, ordered_extent->file_offset,
+			ordered_extent->file_offset + ordered_extent->len - 1,
+			EXTENT_DEFRAG, 1, cached_state);
+	if (ret) {
+		u64 last_snapshot = btrfs_root_last_snapshot(&root->root_item);
+		if (0 && last_snapshot >= BTRFS_I(inode)->generation)
+			/* the inode is shared */
+			new = record_old_file_extents(inode, ordered_extent);
 
-	trans = btrfs_join_transaction(root, 1);
+		clear_extent_bit(io_tree, ordered_extent->file_offset,
+			ordered_extent->file_offset + ordered_extent->len - 1,
+			EXTENT_DEFRAG, 0, 0, &cached_state, GFP_NOFS);
+	}
+
+	if (nolock)
+		trans = btrfs_join_transaction_nolock(root);
+	else
+		trans = btrfs_join_transaction(root);
+	if (IS_ERR(trans)) {
+		ret = PTR_ERR(trans);
+		trans = NULL;
+		goto out_unlock;
+	}
+
+	trans->block_rsv = &root->fs_info->delalloc_block_rsv;
 
 	if (test_bit(BTRFS_ORDERED_COMPRESSED, &ordered_extent->flags))
-		compressed = 1;
+		compress_type = ordered_extent->compress_type;
 	if (test_bit(BTRFS_ORDERED_PREALLOC, &ordered_extent->flags)) {
-		BUG_ON(compressed);
+		BUG_ON(compress_type);
 		ret = btrfs_mark_extent_written(trans, inode,
 						ordered_extent->file_offset,
 						ordered_extent->file_offset +
-						ordered_extent->len);
-		BUG_ON(ret);
+						logical_len);
 	} else {
+		BUG_ON(root == root->fs_info->tree_root);
 		ret = insert_reserved_file_extent(trans, inode,
 						ordered_extent->file_offset,
 						ordered_extent->start,
 						ordered_extent->disk_len,
-						ordered_extent->len,
-						ordered_extent->len,
-						compressed, 0, 0,
+						logical_len, logical_len,
+						compress_type, 0, 0,
 						BTRFS_FILE_EXTENT_REG);
-		unpin_extent_cache(&BTRFS_I(inode)->extent_tree,
-				   ordered_extent->file_offset,
-				   ordered_extent->len);
-		BUG_ON(ret);
+		if (!ret)
+			btrfs_release_delalloc_bytes(root,
+						     ordered_extent->start,
+						     ordered_extent->disk_len);
+	}
+	unpin_extent_cache(&BTRFS_I(inode)->extent_tree,
+			   ordered_extent->file_offset, ordered_extent->len,
+			   trans->transid);
+	if (ret < 0) {
+		btrfs_abort_transaction(trans, root, ret);
+		goto out_unlock;
 	}
-	unlock_extent_cached(io_tree, ordered_extent->file_offset,
-			     ordered_extent->file_offset +
-			     ordered_extent->len - 1, &cached_state, GFP_NOFS);
 
 	add_pending_csums(trans, inode, ordered_extent->file_offset,
 			  &ordered_extent->list);
 
-	/* this also removes the ordered extent from the tree */
 	btrfs_ordered_update_i_size(inode, 0, ordered_extent);
-	ret = btrfs_update_inode(trans, root, inode);
-	BUG_ON(ret);
-	btrfs_end_transaction(trans, root);
+	ret = btrfs_update_inode_fallback(trans, root, inode);
+	if (ret) { /* -ENOMEM or corruption */
+		btrfs_abort_transaction(trans, root, ret);
+		goto out_unlock;
+	}
+	ret = 0;
+out_unlock:
+	unlock_extent_cached(io_tree, ordered_extent->file_offset,
+			     ordered_extent->file_offset +
+			     ordered_extent->len - 1, &cached_state, GFP_NOFS);
 out:
+	if (root != root->fs_info->tree_root)
+		btrfs_delalloc_release_metadata(inode, ordered_extent->len);
+	if (trans)
+		btrfs_end_transaction(trans, root);
+
+	if (ret || truncated) {
+		u64 start, end;
+
+		if (truncated)
+			start = ordered_extent->file_offset + logical_len;
+		else
+			start = ordered_extent->file_offset;
+		end = ordered_extent->file_offset + ordered_extent->len - 1;
+		clear_extent_uptodate(io_tree, start, end, NULL, GFP_NOFS);
+
+		/* Drop the cache for the part of the extent we didn't write. */
+		btrfs_drop_extent_cache(inode, start, end, 0);
+
+		/*
+		 * If the ordered extent had an IOERR or something else went
+		 * wrong we need to return the space for this ordered extent
+		 * back to the allocator.  We only free the extent in the
+		 * truncated case if we didn't write out the extent at all.
+		 */
+		if ((ret || !logical_len) &&
+		    !test_bit(BTRFS_ORDERED_NOCOW, &ordered_extent->flags) &&
+		    !test_bit(BTRFS_ORDERED_PREALLOC, &ordered_extent->flags))
+			btrfs_free_reserved_extent(root, ordered_extent->start,
+						   ordered_extent->disk_len, 1);
+	}
+
+
+	/*
+	 * This needs to be done to make sure anybody waiting knows we are done
+	 * updating everything for this ordered extent.
+	 */
+	btrfs_remove_ordered_extent(inode, ordered_extent);
+
+	/* for snapshot-aware defrag */
+	if (new) {
+		if (ret) {
+			free_sa_defrag_extent(new);
+			atomic_dec(&root->fs_info->defrag_running);
+		} else {
+			relink_file_extents(new);
+		}
+	}
+
 	/* once for us */
 	btrfs_put_ordered_extent(ordered_extent);
 	/* once for the tree */
 	btrfs_put_ordered_extent(ordered_extent);
 
-	return 0;
+	return ret;
 }
 
-static int btrfs_writepage_end_io_hook(struct page *page, u64 start, u64 end,
-				struct extent_state *state, int uptodate)
+static void finish_ordered_fn(struct btrfs_work *work)
 {
-	ClearPagePrivate2(page);
-	return btrfs_finish_ordered_io(page->mapping->host, start, end);
+	struct btrfs_ordered_extent *ordered_extent;
+	ordered_extent = container_of(work, struct btrfs_ordered_extent, work);
+	btrfs_finish_ordered_io(ordered_extent);
 }
 
-/*
- * When IO fails, either with EIO or csum verification fails, we
- * try other mirrors that might have a good copy of the data.  This
- * io_failure_record is used to record state as we go through all the
- * mirrors.  If another mirror has good data, the page is set up to date
- * and things continue.  If a good mirror can't be found, the original
- * bio end_io callback is called to indicate things have failed.
- */
-struct io_failure_record {
-	struct page *page;
-	u64 start;
-	u64 len;
-	u64 logical;
-	unsigned long bio_flags;
-	int last_mirror;
-};
-
-static int btrfs_io_failed_hook(struct bio *failed_bio,
-			 struct page *page, u64 start, u64 end,
-			 struct extent_state *state)
+static int btrfs_writepage_end_io_hook(struct page *page, u64 start, u64 end,
+				struct extent_state *state, int uptodate)
 {
-	struct io_failure_record *failrec = NULL;
-	u64 private;
-	struct extent_map *em;
 	struct inode *inode = page->mapping->host;
-	struct extent_io_tree *failure_tree = &BTRFS_I(inode)->io_failure_tree;
-	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
-	struct bio *bio;
-	int num_copies;
-	int ret;
-	int rw;
-	u64 logical;
-
-	ret = get_state_private(failure_tree, start, &private);
-	if (ret) {
-		failrec = kmalloc(sizeof(*failrec), GFP_NOFS);
-		if (!failrec)
-			return -ENOMEM;
-		failrec->start = start;
-		failrec->len = end - start + 1;
-		failrec->last_mirror = 0;
-		failrec->bio_flags = 0;
-
-		read_lock(&em_tree->lock);
-		em = lookup_extent_mapping(em_tree, start, failrec->len);
-		if (em->start > start || em->start + em->len < start) {
-			free_extent_map(em);
-			em = NULL;
-		}
-		read_unlock(&em_tree->lock);
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct btrfs_ordered_extent *ordered_extent = NULL;
+	struct btrfs_workqueue *workers;
 
-		if (!em || IS_ERR(em)) {
-			kfree(failrec);
-			return -EIO;
-		}
-		logical = start - em->start;
-		logical = em->block_start + logical;
-		if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) {
-			logical = em->block_start;
-			failrec->bio_flags = EXTENT_BIO_COMPRESSED;
-		}
-		failrec->logical = logical;
-		free_extent_map(em);
-		set_extent_bits(failure_tree, start, end, EXTENT_LOCKED |
-				EXTENT_DIRTY, GFP_NOFS);
-		set_state_private(failure_tree, start,
-				 (u64)(unsigned long)failrec);
-	} else {
-		failrec = (struct io_failure_record *)(unsigned long)private;
-	}
-	num_copies = btrfs_num_copies(
-			      &BTRFS_I(inode)->root->fs_info->mapping_tree,
-			      failrec->logical, failrec->len);
-	failrec->last_mirror++;
-	if (!state) {
-		spin_lock(&BTRFS_I(inode)->io_tree.lock);
-		state = find_first_extent_bit_state(&BTRFS_I(inode)->io_tree,
-						    failrec->start,
-						    EXTENT_LOCKED);
-		if (state && state->start != failrec->start)
-			state = NULL;
-		spin_unlock(&BTRFS_I(inode)->io_tree.lock);
-	}
-	if (!state || failrec->last_mirror > num_copies) {
-		set_state_private(failure_tree, failrec->start, 0);
-		clear_extent_bits(failure_tree, failrec->start,
-				  failrec->start + failrec->len - 1,
-				  EXTENT_LOCKED | EXTENT_DIRTY, GFP_NOFS);
-		kfree(failrec);
-		return -EIO;
-	}
-	bio = bio_alloc(GFP_NOFS, 1);
-	bio->bi_private = state;
-	bio->bi_end_io = failed_bio->bi_end_io;
-	bio->bi_sector = failrec->logical >> 9;
-	bio->bi_bdev = failed_bio->bi_bdev;
-	bio->bi_size = 0;
+	trace_btrfs_writepage_end_io_hook(page, start, end, uptodate);
 
-	bio_add_page(bio, page, failrec->len, start - page_offset(page));
-	if (failed_bio->bi_rw & (1 << BIO_RW))
-		rw = WRITE;
-	else
-		rw = READ;
+	ClearPagePrivate2(page);
+	if (!btrfs_dec_test_ordered_pending(inode, &ordered_extent, start,
+					    end - start + 1, uptodate))
+		return 0;
 
-	BTRFS_I(inode)->io_tree.ops->submit_bio_hook(inode, rw, bio,
-						      failrec->last_mirror,
-						      failrec->bio_flags);
-	return 0;
-}
+	btrfs_init_work(&ordered_extent->work, finish_ordered_fn, NULL, NULL);
 
-/*
- * each time an IO finishes, we do a fast check in the IO failure tree
- * to see if we need to process or clean up an io_failure_record
- */
-static int btrfs_clean_io_failures(struct inode *inode, u64 start)
-{
-	u64 private;
-	u64 private_failure;
-	struct io_failure_record *failure;
-	int ret;
+	if (btrfs_is_free_space_inode(inode))
+		workers = root->fs_info->endio_freespace_worker;
+	else
+		workers = root->fs_info->endio_write_workers;
+	btrfs_queue_work(workers, &ordered_extent->work);
 
-	private = 0;
-	if (count_range_bits(&BTRFS_I(inode)->io_failure_tree, &private,
-			     (u64)-1, 1, EXTENT_DIRTY)) {
-		ret = get_state_private(&BTRFS_I(inode)->io_failure_tree,
-					start, &private_failure);
-		if (ret == 0) {
-			failure = (struct io_failure_record *)(unsigned long)
-				   private_failure;
-			set_state_private(&BTRFS_I(inode)->io_failure_tree,
-					  failure->start, 0);
-			clear_extent_bits(&BTRFS_I(inode)->io_failure_tree,
-					  failure->start,
-					  failure->start + failure->len - 1,
-					  EXTENT_DIRTY | EXTENT_LOCKED,
-					  GFP_NOFS);
-			kfree(failure);
-		}
-	}
 	return 0;
 }
 
 /*
  * when reads are done, we need to check csums to verify the data is correct
- * if there's a match, we allow the bio to finish.  If not, we go through
- * the io_failure_record routines to find good copies
+ * if there's a match, we allow the bio to finish.  If not, the code in
+ * extent_io.c will try to find good copies for us.
  */
-static int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end,
-			       struct extent_state *state)
+static int btrfs_readpage_end_io_hook(struct btrfs_io_bio *io_bio,
+				      u64 phy_offset, struct page *page,
+				      u64 start, u64 end, int mirror)
 {
-	size_t offset = start - ((u64)page->index << PAGE_CACHE_SHIFT);
+	size_t offset = start - page_offset(page);
 	struct inode *inode = page->mapping->host;
 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
 	char *kaddr;
-	u64 private = ~(u32)0;
-	int ret;
 	struct btrfs_root *root = BTRFS_I(inode)->root;
+	u32 csum_expected;
 	u32 csum = ~(u32)0;
+	static DEFINE_RATELIMIT_STATE(_rs, DEFAULT_RATELIMIT_INTERVAL,
+	                              DEFAULT_RATELIMIT_BURST);
 
 	if (PageChecked(page)) {
 		ClearPageChecked(page);
@@ -1897,7 +2855,7 @@ static int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end,
 	}
 
 	if (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)
-		return 0;
+		goto good;
 
 	if (root->root_key.objectid == BTRFS_DATA_RELOC_TREE_OBJECTID &&
 	    test_range_bit(io_tree, start, end, EXTENT_NODATASUM, 1, NULL)) {
@@ -1906,40 +2864,27 @@ static int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end,
 		return 0;
 	}
 
-	if (state && state->start == start) {
-		private = state->private;
-		ret = 0;
-	} else {
-		ret = get_state_private(io_tree, start, &private);
-	}
-	kaddr = kmap_atomic(page, KM_USER0);
-	if (ret)
-		goto zeroit;
+	phy_offset >>= inode->i_sb->s_blocksize_bits;
+	csum_expected = *(((u32 *)io_bio->csum) + phy_offset);
 
-	csum = btrfs_csum_data(root, kaddr + offset, csum,  end - start + 1);
+	kaddr = kmap_atomic(page);
+	csum = btrfs_csum_data(kaddr + offset, csum,  end - start + 1);
 	btrfs_csum_final(csum, (char *)&csum);
-	if (csum != private)
+	if (csum != csum_expected)
 		goto zeroit;
 
-	kunmap_atomic(kaddr, KM_USER0);
+	kunmap_atomic(kaddr);
 good:
-	/* if the io failure tree for this inode is non-empty,
-	 * check to see if we've recovered from a failed IO
-	 */
-	btrfs_clean_io_failures(inode, start);
 	return 0;
 
 zeroit:
-	if (printk_ratelimit()) {
-		printk(KERN_INFO "btrfs csum failed ino %lu off %llu csum %u "
-		       "private %llu\n", page->mapping->host->i_ino,
-		       (unsigned long long)start, csum,
-		       (unsigned long long)private);
-	}
+	if (__ratelimit(&_rs))
+		btrfs_info(root->fs_info, "csum failed ino %llu off %llu csum %u expected csum %u",
+			btrfs_ino(page->mapping->host), start, csum, csum_expected);
 	memset(kaddr + offset, 1, end - start + 1);
 	flush_dcache_page(page);
-	kunmap_atomic(kaddr, KM_USER0);
-	if (private == 0)
+	kunmap_atomic(kaddr);
+	if (csum_expected == 0)
 		return 0;
 	return -EIO;
 }
@@ -1949,6 +2894,8 @@ struct delayed_iput {
 	struct inode *inode;
 };
 
+/* JDM: If this is fs-wide, why can't we add a pointer to
+ * btrfs_inode instead and avoid the allocation? */
 void btrfs_add_delayed_iput(struct inode *inode)
 {
 	struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
@@ -1978,7 +2925,6 @@ void btrfs_run_delayed_iputs(struct btrfs_root *root)
 	if (empty)
 		return;
 
-	down_read(&root->fs_info->cleanup_work_sem);
 	spin_lock(&fs_info->delayed_iput_lock);
 	list_splice_init(&fs_info->delayed_iputs, &list);
 	spin_unlock(&fs_info->delayed_iput_lock);
@@ -1989,63 +2935,175 @@ void btrfs_run_delayed_iputs(struct btrfs_root *root)
 		iput(delayed->inode);
 		kfree(delayed);
 	}
-	up_read(&root->fs_info->cleanup_work_sem);
+}
+
+/*
+ * This is called in transaction commit time. If there are no orphan
+ * files in the subvolume, it removes orphan item and frees block_rsv
+ * structure.
+ */
+void btrfs_orphan_commit_root(struct btrfs_trans_handle *trans,
+			      struct btrfs_root *root)
+{
+	struct btrfs_block_rsv *block_rsv;
+	int ret;
+
+	if (atomic_read(&root->orphan_inodes) ||
+	    root->orphan_cleanup_state != ORPHAN_CLEANUP_DONE)
+		return;
+
+	spin_lock(&root->orphan_lock);
+	if (atomic_read(&root->orphan_inodes)) {
+		spin_unlock(&root->orphan_lock);
+		return;
+	}
+
+	if (root->orphan_cleanup_state != ORPHAN_CLEANUP_DONE) {
+		spin_unlock(&root->orphan_lock);
+		return;
+	}
+
+	block_rsv = root->orphan_block_rsv;
+	root->orphan_block_rsv = NULL;
+	spin_unlock(&root->orphan_lock);
+
+	if (test_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state) &&
+	    btrfs_root_refs(&root->root_item) > 0) {
+		ret = btrfs_del_orphan_item(trans, root->fs_info->tree_root,
+					    root->root_key.objectid);
+		if (ret)
+			btrfs_abort_transaction(trans, root, ret);
+		else
+			clear_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED,
+				  &root->state);
+	}
+
+	if (block_rsv) {
+		WARN_ON(block_rsv->size > 0);
+		btrfs_free_block_rsv(root, block_rsv);
+	}
 }
 
 /*
  * This creates an orphan entry for the given inode in case something goes
  * wrong in the middle of an unlink/truncate.
+ *
+ * NOTE: caller of this function should reserve 5 units of metadata for
+ *	 this function.
  */
 int btrfs_orphan_add(struct btrfs_trans_handle *trans, struct inode *inode)
 {
 	struct btrfs_root *root = BTRFS_I(inode)->root;
-	int ret = 0;
+	struct btrfs_block_rsv *block_rsv = NULL;
+	int reserve = 0;
+	int insert = 0;
+	int ret;
 
-	spin_lock(&root->list_lock);
+	if (!root->orphan_block_rsv) {
+		block_rsv = btrfs_alloc_block_rsv(root, BTRFS_BLOCK_RSV_TEMP);
+		if (!block_rsv)
+			return -ENOMEM;
+	}
 
-	/* already on the orphan list, we're good */
-	if (!list_empty(&BTRFS_I(inode)->i_orphan)) {
-		spin_unlock(&root->list_lock);
-		return 0;
+	spin_lock(&root->orphan_lock);
+	if (!root->orphan_block_rsv) {
+		root->orphan_block_rsv = block_rsv;
+	} else if (block_rsv) {
+		btrfs_free_block_rsv(root, block_rsv);
+		block_rsv = NULL;
 	}
 
-	list_add(&BTRFS_I(inode)->i_orphan, &root->orphan_list);
+	if (!test_and_set_bit(BTRFS_INODE_HAS_ORPHAN_ITEM,
+			      &BTRFS_I(inode)->runtime_flags)) {
+#if 0
+		/*
+		 * For proper ENOSPC handling, we should do orphan
+		 * cleanup when mounting. But this introduces backward
+		 * compatibility issue.
+		 */
+		if (!xchg(&root->orphan_item_inserted, 1))
+			insert = 2;
+		else
+			insert = 1;
+#endif
+		insert = 1;
+		atomic_inc(&root->orphan_inodes);
+	}
 
-	spin_unlock(&root->list_lock);
+	if (!test_and_set_bit(BTRFS_INODE_ORPHAN_META_RESERVED,
+			      &BTRFS_I(inode)->runtime_flags))
+		reserve = 1;
+	spin_unlock(&root->orphan_lock);
 
-	/*
-	 * insert an orphan item to track this unlinked/truncated file
-	 */
-	ret = btrfs_insert_orphan_item(trans, root, inode->i_ino);
+	/* grab metadata reservation from transaction handle */
+	if (reserve) {
+		ret = btrfs_orphan_reserve_metadata(trans, inode);
+		BUG_ON(ret); /* -ENOSPC in reservation; Logic error? JDM */
+	}
 
-	return ret;
+	/* insert an orphan item to track this unlinked/truncated file */
+	if (insert >= 1) {
+		ret = btrfs_insert_orphan_item(trans, root, btrfs_ino(inode));
+		if (ret) {
+			atomic_dec(&root->orphan_inodes);
+			if (reserve) {
+				clear_bit(BTRFS_INODE_ORPHAN_META_RESERVED,
+					  &BTRFS_I(inode)->runtime_flags);
+				btrfs_orphan_release_metadata(inode);
+			}
+			if (ret != -EEXIST) {
+				clear_bit(BTRFS_INODE_HAS_ORPHAN_ITEM,
+					  &BTRFS_I(inode)->runtime_flags);
+				btrfs_abort_transaction(trans, root, ret);
+				return ret;
+			}
+		}
+		ret = 0;
+	}
+
+	/* insert an orphan item to track subvolume contains orphan files */
+	if (insert >= 2) {
+		ret = btrfs_insert_orphan_item(trans, root->fs_info->tree_root,
+					       root->root_key.objectid);
+		if (ret && ret != -EEXIST) {
+			btrfs_abort_transaction(trans, root, ret);
+			return ret;
+		}
+	}
+	return 0;
 }
 
 /*
  * We have done the truncate/delete so we can go ahead and remove the orphan
  * item for this particular inode.
  */
-int btrfs_orphan_del(struct btrfs_trans_handle *trans, struct inode *inode)
+static int btrfs_orphan_del(struct btrfs_trans_handle *trans,
+			    struct inode *inode)
 {
 	struct btrfs_root *root = BTRFS_I(inode)->root;
+	int delete_item = 0;
+	int release_rsv = 0;
 	int ret = 0;
 
-	spin_lock(&root->list_lock);
+	spin_lock(&root->orphan_lock);
+	if (test_and_clear_bit(BTRFS_INODE_HAS_ORPHAN_ITEM,
+			       &BTRFS_I(inode)->runtime_flags))
+		delete_item = 1;
 
-	if (list_empty(&BTRFS_I(inode)->i_orphan)) {
-		spin_unlock(&root->list_lock);
-		return 0;
-	}
+	if (test_and_clear_bit(BTRFS_INODE_ORPHAN_META_RESERVED,
+			       &BTRFS_I(inode)->runtime_flags))
+		release_rsv = 1;
+	spin_unlock(&root->orphan_lock);
 
-	list_del_init(&BTRFS_I(inode)->i_orphan);
-	if (!trans) {
-		spin_unlock(&root->list_lock);
-		return 0;
+	if (delete_item) {
+		atomic_dec(&root->orphan_inodes);
+		if (trans)
+			ret = btrfs_del_orphan_item(trans, root,
+						    btrfs_ino(inode));
 	}
 
-	spin_unlock(&root->list_lock);
-
-	ret = btrfs_del_orphan_item(trans, root, inode->i_ino);
+	if (release_rsv)
+		btrfs_orphan_release_metadata(inode);
 
 	return ret;
 }
@@ -2054,21 +3112,24 @@ int btrfs_orphan_del(struct btrfs_trans_handle *trans, struct inode *inode)
  * this cleans up any orphans that may be left on the list from the last use
  * of this root.
  */
-void btrfs_orphan_cleanup(struct btrfs_root *root)
+int btrfs_orphan_cleanup(struct btrfs_root *root)
 {
 	struct btrfs_path *path;
 	struct extent_buffer *leaf;
-	struct btrfs_item *item;
 	struct btrfs_key key, found_key;
 	struct btrfs_trans_handle *trans;
 	struct inode *inode;
+	u64 last_objectid = 0;
 	int ret = 0, nr_unlink = 0, nr_truncate = 0;
 
-	if (!xchg(&root->clean_orphans, 0))
-		return;
+	if (cmpxchg(&root->orphan_cleanup_state, 0, ORPHAN_CLEANUP_STARTED))
+		return 0;
 
 	path = btrfs_alloc_path();
-	BUG_ON(!path);
+	if (!path) {
+		ret = -ENOMEM;
+		goto out;
+	}
 	path->reada = -1;
 
 	key.objectid = BTRFS_ORPHAN_OBJECTID;
@@ -2077,18 +3138,16 @@ void btrfs_orphan_cleanup(struct btrfs_root *root)
 
 	while (1) {
 		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
-		if (ret < 0) {
-			printk(KERN_ERR "Error searching slot for orphan: %d"
-			       "\n", ret);
-			break;
-		}
+		if (ret < 0)
+			goto out;
 
 		/*
 		 * if ret == 0 means we found what we were searching for, which
-		 * is weird, but possible, so only screw with path if we didnt
+		 * is weird, but possible, so only screw with path if we didn't
 		 * find the key and see if we have stuff that matches
 		 */
 		if (ret > 0) {
+			ret = 0;
 			if (path->slots[0] == 0)
 				break;
 			path->slots[0]--;
@@ -2096,7 +3155,6 @@ void btrfs_orphan_cleanup(struct btrfs_root *root)
 
 		/* pull out the item */
 		leaf = path->nodes[0];
-		item = btrfs_item_nr(leaf, path->slots[0]);
 		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
 
 		/* make sure the item matches what we want */
@@ -2106,60 +3164,152 @@ void btrfs_orphan_cleanup(struct btrfs_root *root)
 			break;
 
 		/* release the path since we're done with it */
-		btrfs_release_path(root, path);
+		btrfs_release_path(path);
 
 		/*
 		 * this is where we are basically btrfs_lookup, without the
 		 * crossing root thing.  we store the inode number in the
 		 * offset of the orphan item.
 		 */
+
+		if (found_key.offset == last_objectid) {
+			btrfs_err(root->fs_info,
+				"Error removing orphan entry, stopping orphan cleanup");
+			ret = -EINVAL;
+			goto out;
+		}
+
+		last_objectid = found_key.offset;
+
 		found_key.objectid = found_key.offset;
 		found_key.type = BTRFS_INODE_ITEM_KEY;
 		found_key.offset = 0;
 		inode = btrfs_iget(root->fs_info->sb, &found_key, root, NULL);
-		if (IS_ERR(inode))
-			break;
+		ret = PTR_ERR_OR_ZERO(inode);
+		if (ret && ret != -ESTALE)
+			goto out;
 
-		/*
-		 * add this inode to the orphan list so btrfs_orphan_del does
-		 * the proper thing when we hit it
-		 */
-		spin_lock(&root->list_lock);
-		list_add(&BTRFS_I(inode)->i_orphan, &root->orphan_list);
-		spin_unlock(&root->list_lock);
+		if (ret == -ESTALE && root == root->fs_info->tree_root) {
+			struct btrfs_root *dead_root;
+			struct btrfs_fs_info *fs_info = root->fs_info;
+			int is_dead_root = 0;
 
+			/*
+			 * this is an orphan in the tree root. Currently these
+			 * could come from 2 sources:
+			 *  a) a snapshot deletion in progress
+			 *  b) a free space cache inode
+			 * We need to distinguish those two, as the snapshot
+			 * orphan must not get deleted.
+			 * find_dead_roots already ran before us, so if this
+			 * is a snapshot deletion, we should find the root
+			 * in the dead_roots list
+			 */
+			spin_lock(&fs_info->trans_lock);
+			list_for_each_entry(dead_root, &fs_info->dead_roots,
+					    root_list) {
+				if (dead_root->root_key.objectid ==
+				    found_key.objectid) {
+					is_dead_root = 1;
+					break;
+				}
+			}
+			spin_unlock(&fs_info->trans_lock);
+			if (is_dead_root) {
+				/* prevent this orphan from being found again */
+				key.offset = found_key.objectid - 1;
+				continue;
+			}
+		}
 		/*
-		 * if this is a bad inode, means we actually succeeded in
-		 * removing the inode, but not the orphan record, which means
-		 * we need to manually delete the orphan since iput will just
-		 * do a destroy_inode
+		 * Inode is already gone but the orphan item is still there,
+		 * kill the orphan item.
 		 */
-		if (is_bad_inode(inode)) {
+		if (ret == -ESTALE) {
 			trans = btrfs_start_transaction(root, 1);
-			btrfs_orphan_del(trans, inode);
+			if (IS_ERR(trans)) {
+				ret = PTR_ERR(trans);
+				goto out;
+			}
+			btrfs_debug(root->fs_info, "auto deleting %Lu",
+				found_key.objectid);
+			ret = btrfs_del_orphan_item(trans, root,
+						    found_key.objectid);
 			btrfs_end_transaction(trans, root);
-			iput(inode);
+			if (ret)
+				goto out;
 			continue;
 		}
 
+		/*
+		 * add this inode to the orphan list so btrfs_orphan_del does
+		 * the proper thing when we hit it
+		 */
+		set_bit(BTRFS_INODE_HAS_ORPHAN_ITEM,
+			&BTRFS_I(inode)->runtime_flags);
+		atomic_inc(&root->orphan_inodes);
+
 		/* if we have links, this was a truncate, lets do that */
 		if (inode->i_nlink) {
+			if (WARN_ON(!S_ISREG(inode->i_mode))) {
+				iput(inode);
+				continue;
+			}
 			nr_truncate++;
-			btrfs_truncate(inode);
+
+			/* 1 for the orphan item deletion. */
+			trans = btrfs_start_transaction(root, 1);
+			if (IS_ERR(trans)) {
+				iput(inode);
+				ret = PTR_ERR(trans);
+				goto out;
+			}
+			ret = btrfs_orphan_add(trans, inode);
+			btrfs_end_transaction(trans, root);
+			if (ret) {
+				iput(inode);
+				goto out;
+			}
+
+			ret = btrfs_truncate(inode);
+			if (ret)
+				btrfs_orphan_del(NULL, inode);
 		} else {
 			nr_unlink++;
 		}
 
 		/* this will do delete_inode and everything for us */
 		iput(inode);
+		if (ret)
+			goto out;
+	}
+	/* release the path since we're done with it */
+	btrfs_release_path(path);
+
+	root->orphan_cleanup_state = ORPHAN_CLEANUP_DONE;
+
+	if (root->orphan_block_rsv)
+		btrfs_block_rsv_release(root, root->orphan_block_rsv,
+					(u64)-1);
+
+	if (root->orphan_block_rsv ||
+	    test_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state)) {
+		trans = btrfs_join_transaction(root);
+		if (!IS_ERR(trans))
+			btrfs_end_transaction(trans, root);
 	}
 
 	if (nr_unlink)
-		printk(KERN_INFO "btrfs: unlinked %d orphans\n", nr_unlink);
+		btrfs_debug(root->fs_info, "unlinked %d orphans", nr_unlink);
 	if (nr_truncate)
-		printk(KERN_INFO "btrfs: truncated %d orphans\n", nr_truncate);
+		btrfs_debug(root->fs_info, "truncated %d orphans", nr_truncate);
 
+out:
+	if (ret)
+		btrfs_crit(root->fs_info,
+			"could not do orphan cleanup %d", ret);
 	btrfs_free_path(path);
+	return ret;
 }
 
 /*
@@ -2169,13 +3319,24 @@ void btrfs_orphan_cleanup(struct btrfs_root *root)
  * slot is the slot the inode is in, objectid is the objectid of the inode
  */
 static noinline int acls_after_inode_item(struct extent_buffer *leaf,
-					  int slot, u64 objectid)
+					  int slot, u64 objectid,
+					  int *first_xattr_slot)
 {
 	u32 nritems = btrfs_header_nritems(leaf);
 	struct btrfs_key found_key;
+	static u64 xattr_access = 0;
+	static u64 xattr_default = 0;
 	int scanned = 0;
 
+	if (!xattr_access) {
+		xattr_access = btrfs_name_hash(POSIX_ACL_XATTR_ACCESS,
+					strlen(POSIX_ACL_XATTR_ACCESS));
+		xattr_default = btrfs_name_hash(POSIX_ACL_XATTR_DEFAULT,
+					strlen(POSIX_ACL_XATTR_DEFAULT));
+	}
+
 	slot++;
+	*first_xattr_slot = -1;
 	while (slot < nritems) {
 		btrfs_item_key_to_cpu(leaf, &found_key, slot);
 
@@ -2184,8 +3345,13 @@ static noinline int acls_after_inode_item(struct extent_buffer *leaf,
 			return 0;
 
 		/* we found an xattr, assume we've got an acl */
-		if (found_key.type == BTRFS_XATTR_ITEM_KEY)
-			return 1;
+		if (found_key.type == BTRFS_XATTR_ITEM_KEY) {
+			if (*first_xattr_slot == -1)
+				*first_xattr_slot = slot;
+			if (found_key.offset == xattr_access ||
+			    found_key.offset == xattr_default)
+				return 1;
+		}
 
 		/*
 		 * we found a key greater than an xattr key, there can't
@@ -2210,6 +3376,8 @@ static noinline int acls_after_inode_item(struct extent_buffer *leaf,
 	 * something larger than an xattr.  We have to assume the inode
 	 * has acls
 	 */
+	if (*first_xattr_slot == -1)
+		*first_xattr_slot = slot;
 	return 1;
 }
 
@@ -2224,13 +3392,21 @@ static void btrfs_read_locked_inode(struct inode *inode)
 	struct btrfs_timespec *tspec;
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	struct btrfs_key location;
+	unsigned long ptr;
 	int maybe_acls;
-	u64 alloc_group_block;
 	u32 rdev;
 	int ret;
+	bool filled = false;
+	int first_xattr_slot;
+
+	ret = btrfs_fill_inode(inode, &rdev);
+	if (!ret)
+		filled = true;
 
 	path = btrfs_alloc_path();
-	BUG_ON(!path);
+	if (!path)
+		goto make_bad;
+
 	memcpy(&location, &BTRFS_I(inode)->location, sizeof(location));
 
 	ret = btrfs_lookup_inode(NULL, root, path, &location, 0);
@@ -2238,13 +3414,16 @@ static void btrfs_read_locked_inode(struct inode *inode)
 		goto make_bad;
 
 	leaf = path->nodes[0];
+
+	if (filled)
+		goto cache_index;
+
 	inode_item = btrfs_item_ptr(leaf, path->slots[0],
 				    struct btrfs_inode_item);
-
 	inode->i_mode = btrfs_inode_mode(leaf, inode_item);
-	inode->i_nlink = btrfs_inode_nlink(leaf, inode_item);
-	inode->i_uid = btrfs_inode_uid(leaf, inode_item);
-	inode->i_gid = btrfs_inode_gid(leaf, inode_item);
+	set_nlink(inode, btrfs_inode_nlink(leaf, inode_item));
+	i_uid_write(inode, btrfs_inode_uid(leaf, inode_item));
+	i_gid_write(inode, btrfs_inode_gid(leaf, inode_item));
 	btrfs_i_size_write(inode, btrfs_inode_size(leaf, inode_item));
 
 	tspec = btrfs_inode_atime(inode_item);
@@ -2261,7 +3440,19 @@ static void btrfs_read_locked_inode(struct inode *inode)
 
 	inode_set_bytes(inode, btrfs_inode_nbytes(leaf, inode_item));
 	BTRFS_I(inode)->generation = btrfs_inode_generation(leaf, inode_item);
-	BTRFS_I(inode)->sequence = btrfs_inode_sequence(leaf, inode_item);
+	BTRFS_I(inode)->last_trans = btrfs_inode_transid(leaf, inode_item);
+
+	/*
+	 * If we were modified in the current generation and evicted from memory
+	 * and then re-read we need to do a full sync since we don't have any
+	 * idea about which extents were modified before we were evicted from
+	 * cache.
+	 */
+	if (BTRFS_I(inode)->last_trans == root->fs_info->generation)
+		set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
+			&BTRFS_I(inode)->runtime_flags);
+
+	inode->i_version = btrfs_inode_sequence(leaf, inode_item);
 	inode->i_generation = BTRFS_I(inode)->generation;
 	inode->i_rdev = 0;
 	rdev = btrfs_inode_rdev(leaf, inode_item);
@@ -2269,21 +3460,50 @@ static void btrfs_read_locked_inode(struct inode *inode)
 	BTRFS_I(inode)->index_cnt = (u64)-1;
 	BTRFS_I(inode)->flags = btrfs_inode_flags(leaf, inode_item);
 
-	alloc_group_block = btrfs_inode_block_group(leaf, inode_item);
+cache_index:
+	path->slots[0]++;
+	if (inode->i_nlink != 1 ||
+	    path->slots[0] >= btrfs_header_nritems(leaf))
+		goto cache_acl;
+
+	btrfs_item_key_to_cpu(leaf, &location, path->slots[0]);
+	if (location.objectid != btrfs_ino(inode))
+		goto cache_acl;
+
+	ptr = btrfs_item_ptr_offset(leaf, path->slots[0]);
+	if (location.type == BTRFS_INODE_REF_KEY) {
+		struct btrfs_inode_ref *ref;
+
+		ref = (struct btrfs_inode_ref *)ptr;
+		BTRFS_I(inode)->dir_index = btrfs_inode_ref_index(leaf, ref);
+	} else if (location.type == BTRFS_INODE_EXTREF_KEY) {
+		struct btrfs_inode_extref *extref;
 
+		extref = (struct btrfs_inode_extref *)ptr;
+		BTRFS_I(inode)->dir_index = btrfs_inode_extref_index(leaf,
+								     extref);
+	}
+cache_acl:
 	/*
 	 * try to precache a NULL acl entry for files that don't have
 	 * any xattrs or acls
 	 */
-	maybe_acls = acls_after_inode_item(leaf, path->slots[0], inode->i_ino);
+	maybe_acls = acls_after_inode_item(leaf, path->slots[0],
+					   btrfs_ino(inode), &first_xattr_slot);
+	if (first_xattr_slot != -1) {
+		path->slots[0] = first_xattr_slot;
+		ret = btrfs_load_inode_props(inode, path);
+		if (ret)
+			btrfs_err(root->fs_info,
+				  "error loading props for ino %llu (root %llu): %d",
+				  btrfs_ino(inode),
+				  root->root_key.objectid, ret);
+	}
+	btrfs_free_path(path);
+
 	if (!maybe_acls)
 		cache_no_acl(inode);
 
-	BTRFS_I(inode)->block_group = btrfs_find_block_group(root, 0,
-						alloc_group_block, 0);
-	btrfs_free_path(path);
-	inode_item = NULL;
-
 	switch (inode->i_mode & S_IFMT) {
 	case S_IFREG:
 		inode->i_mapping->a_ops = &btrfs_aops;
@@ -2326,40 +3546,47 @@ static void fill_inode_item(struct btrfs_trans_handle *trans,
 			    struct btrfs_inode_item *item,
 			    struct inode *inode)
 {
-	btrfs_set_inode_uid(leaf, item, inode->i_uid);
-	btrfs_set_inode_gid(leaf, item, inode->i_gid);
-	btrfs_set_inode_size(leaf, item, BTRFS_I(inode)->disk_i_size);
-	btrfs_set_inode_mode(leaf, item, inode->i_mode);
-	btrfs_set_inode_nlink(leaf, item, inode->i_nlink);
-
-	btrfs_set_timespec_sec(leaf, btrfs_inode_atime(item),
-			       inode->i_atime.tv_sec);
-	btrfs_set_timespec_nsec(leaf, btrfs_inode_atime(item),
-				inode->i_atime.tv_nsec);
-
-	btrfs_set_timespec_sec(leaf, btrfs_inode_mtime(item),
-			       inode->i_mtime.tv_sec);
-	btrfs_set_timespec_nsec(leaf, btrfs_inode_mtime(item),
-				inode->i_mtime.tv_nsec);
-
-	btrfs_set_timespec_sec(leaf, btrfs_inode_ctime(item),
-			       inode->i_ctime.tv_sec);
-	btrfs_set_timespec_nsec(leaf, btrfs_inode_ctime(item),
-				inode->i_ctime.tv_nsec);
-
-	btrfs_set_inode_nbytes(leaf, item, inode_get_bytes(inode));
-	btrfs_set_inode_generation(leaf, item, BTRFS_I(inode)->generation);
-	btrfs_set_inode_sequence(leaf, item, BTRFS_I(inode)->sequence);
-	btrfs_set_inode_transid(leaf, item, trans->transid);
-	btrfs_set_inode_rdev(leaf, item, inode->i_rdev);
-	btrfs_set_inode_flags(leaf, item, BTRFS_I(inode)->flags);
-	btrfs_set_inode_block_group(leaf, item, BTRFS_I(inode)->block_group);
+	struct btrfs_map_token token;
+
+	btrfs_init_map_token(&token);
+
+	btrfs_set_token_inode_uid(leaf, item, i_uid_read(inode), &token);
+	btrfs_set_token_inode_gid(leaf, item, i_gid_read(inode), &token);
+	btrfs_set_token_inode_size(leaf, item, BTRFS_I(inode)->disk_i_size,
+				   &token);
+	btrfs_set_token_inode_mode(leaf, item, inode->i_mode, &token);
+	btrfs_set_token_inode_nlink(leaf, item, inode->i_nlink, &token);
+
+	btrfs_set_token_timespec_sec(leaf, btrfs_inode_atime(item),
+				     inode->i_atime.tv_sec, &token);
+	btrfs_set_token_timespec_nsec(leaf, btrfs_inode_atime(item),
+				      inode->i_atime.tv_nsec, &token);
+
+	btrfs_set_token_timespec_sec(leaf, btrfs_inode_mtime(item),
+				     inode->i_mtime.tv_sec, &token);
+	btrfs_set_token_timespec_nsec(leaf, btrfs_inode_mtime(item),
+				      inode->i_mtime.tv_nsec, &token);
+
+	btrfs_set_token_timespec_sec(leaf, btrfs_inode_ctime(item),
+				     inode->i_ctime.tv_sec, &token);
+	btrfs_set_token_timespec_nsec(leaf, btrfs_inode_ctime(item),
+				      inode->i_ctime.tv_nsec, &token);
+
+	btrfs_set_token_inode_nbytes(leaf, item, inode_get_bytes(inode),
+				     &token);
+	btrfs_set_token_inode_generation(leaf, item, BTRFS_I(inode)->generation,
+					 &token);
+	btrfs_set_token_inode_sequence(leaf, item, inode->i_version, &token);
+	btrfs_set_token_inode_transid(leaf, item, trans->transid, &token);
+	btrfs_set_token_inode_rdev(leaf, item, inode->i_rdev, &token);
+	btrfs_set_token_inode_flags(leaf, item, BTRFS_I(inode)->flags, &token);
+	btrfs_set_token_inode_block_group(leaf, item, 0, &token);
 }
 
 /*
  * copy everything in the in-memory inode into the btree.
  */
-noinline int btrfs_update_inode(struct btrfs_trans_handle *trans,
+static noinline int btrfs_update_inode_item(struct btrfs_trans_handle *trans,
 				struct btrfs_root *root, struct inode *inode)
 {
 	struct btrfs_inode_item *inode_item;
@@ -2368,20 +3595,21 @@ noinline int btrfs_update_inode(struct btrfs_trans_handle *trans,
 	int ret;
 
 	path = btrfs_alloc_path();
-	BUG_ON(!path);
+	if (!path)
+		return -ENOMEM;
+
 	path->leave_spinning = 1;
-	ret = btrfs_lookup_inode(trans, root, path,
-				 &BTRFS_I(inode)->location, 1);
+	ret = btrfs_lookup_inode(trans, root, path, &BTRFS_I(inode)->location,
+				 1);
 	if (ret) {
 		if (ret > 0)
 			ret = -ENOENT;
 		goto failed;
 	}
 
-	btrfs_unlock_up_safe(path, 1);
 	leaf = path->nodes[0];
 	inode_item = btrfs_item_ptr(leaf, path->slots[0],
-				  struct btrfs_inode_item);
+				    struct btrfs_inode_item);
 
 	fill_inode_item(trans, leaf, inode_item, inode);
 	btrfs_mark_buffer_dirty(leaf);
@@ -2392,16 +3620,55 @@ failed:
 	return ret;
 }
 
+/*
+ * copy everything in the in-memory inode into the btree.
+ */
+noinline int btrfs_update_inode(struct btrfs_trans_handle *trans,
+				struct btrfs_root *root, struct inode *inode)
+{
+	int ret;
+
+	/*
+	 * If the inode is a free space inode, we can deadlock during commit
+	 * if we put it into the delayed code.
+	 *
+	 * The data relocation inode should also be directly updated
+	 * without delay
+	 */
+	if (!btrfs_is_free_space_inode(inode)
+	    && root->root_key.objectid != BTRFS_DATA_RELOC_TREE_OBJECTID) {
+		btrfs_update_root_times(trans, root);
+
+		ret = btrfs_delayed_update_inode(trans, root, inode);
+		if (!ret)
+			btrfs_set_inode_last_trans(trans, inode);
+		return ret;
+	}
+
+	return btrfs_update_inode_item(trans, root, inode);
+}
+
+noinline int btrfs_update_inode_fallback(struct btrfs_trans_handle *trans,
+					 struct btrfs_root *root,
+					 struct inode *inode)
+{
+	int ret;
+
+	ret = btrfs_update_inode(trans, root, inode);
+	if (ret == -ENOSPC)
+		return btrfs_update_inode_item(trans, root, inode);
+	return ret;
+}
 
 /*
  * unlink helper that gets used here in inode.c and in the tree logging
  * recovery code.  It remove a link in a directory with a given name, and
  * also drops the back refs in the inode to the directory
  */
-int btrfs_unlink_inode(struct btrfs_trans_handle *trans,
-		       struct btrfs_root *root,
-		       struct inode *dir, struct inode *inode,
-		       const char *name, int name_len)
+static int __btrfs_unlink_inode(struct btrfs_trans_handle *trans,
+				struct btrfs_root *root,
+				struct inode *dir, struct inode *inode,
+				const char *name, int name_len)
 {
 	struct btrfs_path *path;
 	int ret = 0;
@@ -2409,15 +3676,17 @@ int btrfs_unlink_inode(struct btrfs_trans_handle *trans,
 	struct btrfs_dir_item *di;
 	struct btrfs_key key;
 	u64 index;
+	u64 ino = btrfs_ino(inode);
+	u64 dir_ino = btrfs_ino(dir);
 
 	path = btrfs_alloc_path();
 	if (!path) {
 		ret = -ENOMEM;
-		goto err;
+		goto out;
 	}
 
 	path->leave_spinning = 1;
-	di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino,
+	di = btrfs_lookup_dir_item(trans, root, path, dir_ino,
 				    name, name_len, -1);
 	if (IS_ERR(di)) {
 		ret = PTR_ERR(di);
@@ -2432,91 +3701,154 @@ int btrfs_unlink_inode(struct btrfs_trans_handle *trans,
 	ret = btrfs_delete_one_dir_name(trans, root, path, di);
 	if (ret)
 		goto err;
-	btrfs_release_path(root, path);
+	btrfs_release_path(path);
 
-	ret = btrfs_del_inode_ref(trans, root, name, name_len,
-				  inode->i_ino,
-				  dir->i_ino, &index);
-	if (ret) {
-		printk(KERN_INFO "btrfs failed to delete reference to %.*s, "
-		       "inode %lu parent %lu\n", name_len, name,
-		       inode->i_ino, dir->i_ino);
-		goto err;
+	/*
+	 * If we don't have dir index, we have to get it by looking up
+	 * the inode ref, since we get the inode ref, remove it directly,
+	 * it is unnecessary to do delayed deletion.
+	 *
+	 * But if we have dir index, needn't search inode ref to get it.
+	 * Since the inode ref is close to the inode item, it is better
+	 * that we delay to delete it, and just do this deletion when
+	 * we update the inode item.
+	 */
+	if (BTRFS_I(inode)->dir_index) {
+		ret = btrfs_delayed_delete_inode_ref(inode);
+		if (!ret) {
+			index = BTRFS_I(inode)->dir_index;
+			goto skip_backref;
+		}
 	}
 
-	di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino,
-					 index, name, name_len, -1);
-	if (IS_ERR(di)) {
-		ret = PTR_ERR(di);
+	ret = btrfs_del_inode_ref(trans, root, name, name_len, ino,
+				  dir_ino, &index);
+	if (ret) {
+		btrfs_info(root->fs_info,
+			"failed to delete reference to %.*s, inode %llu parent %llu",
+			name_len, name, ino, dir_ino);
+		btrfs_abort_transaction(trans, root, ret);
 		goto err;
 	}
-	if (!di) {
-		ret = -ENOENT;
+skip_backref:
+	ret = btrfs_delete_delayed_dir_index(trans, root, dir, index);
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
 		goto err;
 	}
-	ret = btrfs_delete_one_dir_name(trans, root, path, di);
-	btrfs_release_path(root, path);
 
 	ret = btrfs_del_inode_ref_in_log(trans, root, name, name_len,
-					 inode, dir->i_ino);
-	BUG_ON(ret != 0 && ret != -ENOENT);
+					 inode, dir_ino);
+	if (ret != 0 && ret != -ENOENT) {
+		btrfs_abort_transaction(trans, root, ret);
+		goto err;
+	}
 
 	ret = btrfs_del_dir_entries_in_log(trans, root, name, name_len,
 					   dir, index);
-	BUG_ON(ret);
+	if (ret == -ENOENT)
+		ret = 0;
+	else if (ret)
+		btrfs_abort_transaction(trans, root, ret);
 err:
 	btrfs_free_path(path);
 	if (ret)
 		goto out;
 
 	btrfs_i_size_write(dir, dir->i_size - name_len * 2);
+	inode_inc_iversion(inode);
+	inode_inc_iversion(dir);
 	inode->i_ctime = dir->i_mtime = dir->i_ctime = CURRENT_TIME;
-	btrfs_update_inode(trans, root, dir);
-	btrfs_drop_nlink(inode);
-	ret = btrfs_update_inode(trans, root, inode);
+	ret = btrfs_update_inode(trans, root, dir);
 out:
 	return ret;
 }
 
-static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
+int btrfs_unlink_inode(struct btrfs_trans_handle *trans,
+		       struct btrfs_root *root,
+		       struct inode *dir, struct inode *inode,
+		       const char *name, int name_len)
 {
-	struct btrfs_root *root;
-	struct btrfs_trans_handle *trans;
-	struct inode *inode = dentry->d_inode;
 	int ret;
-	unsigned long nr = 0;
+	ret = __btrfs_unlink_inode(trans, root, dir, inode, name, name_len);
+	if (!ret) {
+		drop_nlink(inode);
+		ret = btrfs_update_inode(trans, root, inode);
+	}
+	return ret;
+}
 
-	root = BTRFS_I(dir)->root;
+/*
+ * helper to start transaction for unlink and rmdir.
+ *
+ * unlink and rmdir are special in btrfs, they do not always free space, so
+ * if we cannot make our reservations the normal way try and see if there is
+ * plenty of slack room in the global reserve to migrate, otherwise we cannot
+ * allow the unlink to occur.
+ */
+static struct btrfs_trans_handle *__unlink_start_trans(struct inode *dir)
+{
+	struct btrfs_trans_handle *trans;
+	struct btrfs_root *root = BTRFS_I(dir)->root;
+	int ret;
 
 	/*
-	 * 5 items for unlink inode
-	 * 1 for orphan
+	 * 1 for the possible orphan item
+	 * 1 for the dir item
+	 * 1 for the dir index
+	 * 1 for the inode ref
+	 * 1 for the inode
 	 */
-	ret = btrfs_reserve_metadata_space(root, 6);
-	if (ret)
-		return ret;
-
-	trans = btrfs_start_transaction(root, 1);
-	if (IS_ERR(trans)) {
-		btrfs_unreserve_metadata_space(root, 6);
-		return PTR_ERR(trans);
+	trans = btrfs_start_transaction(root, 5);
+	if (!IS_ERR(trans) || PTR_ERR(trans) != -ENOSPC)
+		return trans;
+
+	if (PTR_ERR(trans) == -ENOSPC) {
+		u64 num_bytes = btrfs_calc_trans_metadata_size(root, 5);
+
+		trans = btrfs_start_transaction(root, 0);
+		if (IS_ERR(trans))
+			return trans;
+		ret = btrfs_cond_migrate_bytes(root->fs_info,
+					       &root->fs_info->trans_block_rsv,
+					       num_bytes, 5);
+		if (ret) {
+			btrfs_end_transaction(trans, root);
+			return ERR_PTR(ret);
+		}
+		trans->block_rsv = &root->fs_info->trans_block_rsv;
+		trans->bytes_reserved = num_bytes;
 	}
+	return trans;
+}
 
-	btrfs_set_trans_block_group(trans, dir);
+static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
+{
+	struct btrfs_root *root = BTRFS_I(dir)->root;
+	struct btrfs_trans_handle *trans;
+	struct inode *inode = dentry->d_inode;
+	int ret;
+
+	trans = __unlink_start_trans(dir);
+	if (IS_ERR(trans))
+		return PTR_ERR(trans);
 
 	btrfs_record_unlink_dir(trans, dir, dentry->d_inode, 0);
 
 	ret = btrfs_unlink_inode(trans, root, dir, dentry->d_inode,
 				 dentry->d_name.name, dentry->d_name.len);
+	if (ret)
+		goto out;
 
-	if (inode->i_nlink == 0)
+	if (inode->i_nlink == 0) {
 		ret = btrfs_orphan_add(trans, inode);
+		if (ret)
+			goto out;
+	}
 
-	nr = trans->blocks_used;
-
-	btrfs_end_transaction_throttle(trans, root);
-	btrfs_unreserve_metadata_space(root, 6);
-	btrfs_btree_balance_dirty(root, nr);
+out:
+	btrfs_end_transaction(trans, root);
+	btrfs_btree_balance_dirty(root);
 	return ret;
 }
 
@@ -2531,84 +3863,92 @@ int btrfs_unlink_subvol(struct btrfs_trans_handle *trans,
 	struct btrfs_key key;
 	u64 index;
 	int ret;
+	u64 dir_ino = btrfs_ino(dir);
 
 	path = btrfs_alloc_path();
 	if (!path)
 		return -ENOMEM;
 
-	di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino,
+	di = btrfs_lookup_dir_item(trans, root, path, dir_ino,
 				   name, name_len, -1);
-	BUG_ON(!di || IS_ERR(di));
+	if (IS_ERR_OR_NULL(di)) {
+		if (!di)
+			ret = -ENOENT;
+		else
+			ret = PTR_ERR(di);
+		goto out;
+	}
 
 	leaf = path->nodes[0];
 	btrfs_dir_item_key_to_cpu(leaf, di, &key);
 	WARN_ON(key.type != BTRFS_ROOT_ITEM_KEY || key.objectid != objectid);
 	ret = btrfs_delete_one_dir_name(trans, root, path, di);
-	BUG_ON(ret);
-	btrfs_release_path(root, path);
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
+		goto out;
+	}
+	btrfs_release_path(path);
 
 	ret = btrfs_del_root_ref(trans, root->fs_info->tree_root,
 				 objectid, root->root_key.objectid,
-				 dir->i_ino, &index, name, name_len);
+				 dir_ino, &index, name, name_len);
 	if (ret < 0) {
-		BUG_ON(ret != -ENOENT);
-		di = btrfs_search_dir_index_item(root, path, dir->i_ino,
+		if (ret != -ENOENT) {
+			btrfs_abort_transaction(trans, root, ret);
+			goto out;
+		}
+		di = btrfs_search_dir_index_item(root, path, dir_ino,
 						 name, name_len);
-		BUG_ON(!di || IS_ERR(di));
+		if (IS_ERR_OR_NULL(di)) {
+			if (!di)
+				ret = -ENOENT;
+			else
+				ret = PTR_ERR(di);
+			btrfs_abort_transaction(trans, root, ret);
+			goto out;
+		}
 
 		leaf = path->nodes[0];
 		btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
-		btrfs_release_path(root, path);
+		btrfs_release_path(path);
 		index = key.offset;
 	}
+	btrfs_release_path(path);
 
-	di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino,
-					 index, name, name_len, -1);
-	BUG_ON(!di || IS_ERR(di));
-
-	leaf = path->nodes[0];
-	btrfs_dir_item_key_to_cpu(leaf, di, &key);
-	WARN_ON(key.type != BTRFS_ROOT_ITEM_KEY || key.objectid != objectid);
-	ret = btrfs_delete_one_dir_name(trans, root, path, di);
-	BUG_ON(ret);
-	btrfs_release_path(root, path);
+	ret = btrfs_delete_delayed_dir_index(trans, root, dir, index);
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
+		goto out;
+	}
 
 	btrfs_i_size_write(dir, dir->i_size - name_len * 2);
+	inode_inc_iversion(dir);
 	dir->i_mtime = dir->i_ctime = CURRENT_TIME;
-	ret = btrfs_update_inode(trans, root, dir);
-	BUG_ON(ret);
-	dir->i_sb->s_dirt = 1;
-
+	ret = btrfs_update_inode_fallback(trans, root, dir);
+	if (ret)
+		btrfs_abort_transaction(trans, root, ret);
+out:
 	btrfs_free_path(path);
-	return 0;
+	return ret;
 }
 
 static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
 {
 	struct inode *inode = dentry->d_inode;
 	int err = 0;
-	int ret;
 	struct btrfs_root *root = BTRFS_I(dir)->root;
 	struct btrfs_trans_handle *trans;
-	unsigned long nr = 0;
 
-	if (inode->i_size > BTRFS_EMPTY_DIR_SIZE ||
-	    inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
+	if (inode->i_size > BTRFS_EMPTY_DIR_SIZE)
 		return -ENOTEMPTY;
+	if (btrfs_ino(inode) == BTRFS_FIRST_FREE_OBJECTID)
+		return -EPERM;
 
-	ret = btrfs_reserve_metadata_space(root, 5);
-	if (ret)
-		return ret;
-
-	trans = btrfs_start_transaction(root, 1);
-	if (IS_ERR(trans)) {
-		btrfs_unreserve_metadata_space(root, 5);
+	trans = __unlink_start_trans(dir);
+	if (IS_ERR(trans))
 		return PTR_ERR(trans);
-	}
 
-	btrfs_set_trans_block_group(trans, dir);
-
-	if (unlikely(inode->i_ino == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID)) {
+	if (unlikely(btrfs_ino(inode) == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID)) {
 		err = btrfs_unlink_subvol(trans, root, dir,
 					  BTRFS_I(inode)->location.objectid,
 					  dentry->d_name.name,
@@ -2626,188 +3966,12 @@ static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
 	if (!err)
 		btrfs_i_size_write(inode, 0);
 out:
-	nr = trans->blocks_used;
-	ret = btrfs_end_transaction_throttle(trans, root);
-	btrfs_unreserve_metadata_space(root, 5);
-	btrfs_btree_balance_dirty(root, nr);
+	btrfs_end_transaction(trans, root);
+	btrfs_btree_balance_dirty(root);
 
-	if (ret && !err)
-		err = ret;
 	return err;
 }
 
-#if 0
-/*
- * when truncating bytes in a file, it is possible to avoid reading
- * the leaves that contain only checksum items.  This can be the
- * majority of the IO required to delete a large file, but it must
- * be done carefully.
- *
- * The keys in the level just above the leaves are checked to make sure
- * the lowest key in a given leaf is a csum key, and starts at an offset
- * after the new  size.
- *
- * Then the key for the next leaf is checked to make sure it also has
- * a checksum item for the same file.  If it does, we know our target leaf
- * contains only checksum items, and it can be safely freed without reading
- * it.
- *
- * This is just an optimization targeted at large files.  It may do
- * nothing.  It will return 0 unless things went badly.
- */
-static noinline int drop_csum_leaves(struct btrfs_trans_handle *trans,
-				     struct btrfs_root *root,
-				     struct btrfs_path *path,
-				     struct inode *inode, u64 new_size)
-{
-	struct btrfs_key key;
-	int ret;
-	int nritems;
-	struct btrfs_key found_key;
-	struct btrfs_key other_key;
-	struct btrfs_leaf_ref *ref;
-	u64 leaf_gen;
-	u64 leaf_start;
-
-	path->lowest_level = 1;
-	key.objectid = inode->i_ino;
-	key.type = BTRFS_CSUM_ITEM_KEY;
-	key.offset = new_size;
-again:
-	ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
-	if (ret < 0)
-		goto out;
-
-	if (path->nodes[1] == NULL) {
-		ret = 0;
-		goto out;
-	}
-	ret = 0;
-	btrfs_node_key_to_cpu(path->nodes[1], &found_key, path->slots[1]);
-	nritems = btrfs_header_nritems(path->nodes[1]);
-
-	if (!nritems)
-		goto out;
-
-	if (path->slots[1] >= nritems)
-		goto next_node;
-
-	/* did we find a key greater than anything we want to delete? */
-	if (found_key.objectid > inode->i_ino ||
-	   (found_key.objectid == inode->i_ino && found_key.type > key.type))
-		goto out;
-
-	/* we check the next key in the node to make sure the leave contains
-	 * only checksum items.  This comparison doesn't work if our
-	 * leaf is the last one in the node
-	 */
-	if (path->slots[1] + 1 >= nritems) {
-next_node:
-		/* search forward from the last key in the node, this
-		 * will bring us into the next node in the tree
-		 */
-		btrfs_node_key_to_cpu(path->nodes[1], &found_key, nritems - 1);
-
-		/* unlikely, but we inc below, so check to be safe */
-		if (found_key.offset == (u64)-1)
-			goto out;
-
-		/* search_forward needs a path with locks held, do the
-		 * search again for the original key.  It is possible
-		 * this will race with a balance and return a path that
-		 * we could modify, but this drop is just an optimization
-		 * and is allowed to miss some leaves.
-		 */
-		btrfs_release_path(root, path);
-		found_key.offset++;
-
-		/* setup a max key for search_forward */
-		other_key.offset = (u64)-1;
-		other_key.type = key.type;
-		other_key.objectid = key.objectid;
-
-		path->keep_locks = 1;
-		ret = btrfs_search_forward(root, &found_key, &other_key,
-					   path, 0, 0);
-		path->keep_locks = 0;
-		if (ret || found_key.objectid != key.objectid ||
-		    found_key.type != key.type) {
-			ret = 0;
-			goto out;
-		}
-
-		key.offset = found_key.offset;
-		btrfs_release_path(root, path);
-		cond_resched();
-		goto again;
-	}
-
-	/* we know there's one more slot after us in the tree,
-	 * read that key so we can verify it is also a checksum item
-	 */
-	btrfs_node_key_to_cpu(path->nodes[1], &other_key, path->slots[1] + 1);
-
-	if (found_key.objectid < inode->i_ino)
-		goto next_key;
-
-	if (found_key.type != key.type || found_key.offset < new_size)
-		goto next_key;
-
-	/*
-	 * if the key for the next leaf isn't a csum key from this objectid,
-	 * we can't be sure there aren't good items inside this leaf.
-	 * Bail out
-	 */
-	if (other_key.objectid != inode->i_ino || other_key.type != key.type)
-		goto out;
-
-	leaf_start = btrfs_node_blockptr(path->nodes[1], path->slots[1]);
-	leaf_gen = btrfs_node_ptr_generation(path->nodes[1], path->slots[1]);
-	/*
-	 * it is safe to delete this leaf, it contains only
-	 * csum items from this inode at an offset >= new_size
-	 */
-	ret = btrfs_del_leaf(trans, root, path, leaf_start);
-	BUG_ON(ret);
-
-	if (root->ref_cows && leaf_gen < trans->transid) {
-		ref = btrfs_alloc_leaf_ref(root, 0);
-		if (ref) {
-			ref->root_gen = root->root_key.offset;
-			ref->bytenr = leaf_start;
-			ref->owner = 0;
-			ref->generation = leaf_gen;
-			ref->nritems = 0;
-
-			btrfs_sort_leaf_ref(ref);
-
-			ret = btrfs_add_leaf_ref(root, ref, 0);
-			WARN_ON(ret);
-			btrfs_free_leaf_ref(root, ref);
-		} else {
-			WARN_ON(1);
-		}
-	}
-next_key:
-	btrfs_release_path(root, path);
-
-	if (other_key.objectid == inode->i_ino &&
-	    other_key.type == key.type && other_key.offset > key.offset) {
-		key.offset = other_key.offset;
-		cond_resched();
-		goto again;
-	}
-	ret = 0;
-out:
-	/* fixup any changes we've made to the path */
-	path->lowest_level = 0;
-	path->keep_locks = 0;
-	btrfs_release_path(root, path);
-	return ret;
-}
-
-#endif
-
 /*
  * this can truncate away extent items, csum items and directory items.
  * It starts at a high offset and removes keys until it can't find
@@ -2833,27 +3997,44 @@ int btrfs_truncate_inode_items(struct btrfs_trans_handle *trans,
 	u64 extent_num_bytes = 0;
 	u64 extent_offset = 0;
 	u64 item_end = 0;
-	u64 mask = root->sectorsize - 1;
+	u64 last_size = (u64)-1;
 	u32 found_type = (u8)-1;
 	int found_extent;
 	int del_item;
 	int pending_del_nr = 0;
 	int pending_del_slot = 0;
 	int extent_type = -1;
-	int encoding;
 	int ret;
 	int err = 0;
+	u64 ino = btrfs_ino(inode);
 
 	BUG_ON(new_size > 0 && min_type != BTRFS_EXTENT_DATA_KEY);
 
-	if (root->ref_cows)
-		btrfs_drop_extent_cache(inode, new_size & (~mask), (u64)-1, 0);
-
 	path = btrfs_alloc_path();
-	BUG_ON(!path);
+	if (!path)
+		return -ENOMEM;
 	path->reada = -1;
 
-	key.objectid = inode->i_ino;
+	/*
+	 * We want to drop from the next block forward in case this new size is
+	 * not block aligned since we will be keeping the last block of the
+	 * extent just the way it is.
+	 */
+	if (test_bit(BTRFS_ROOT_REF_COWS, &root->state) ||
+	    root == root->fs_info->tree_root)
+		btrfs_drop_extent_cache(inode, ALIGN(new_size,
+					root->sectorsize), (u64)-1, 0);
+
+	/*
+	 * This function is also used to drop the items in the log tree before
+	 * we relog the inode, so if root != BTRFS_I(inode)->root, it means
+	 * it is used to drop the loged items. So we shouldn't kill the delayed
+	 * items.
+	 */
+	if (min_type == 0 && root == BTRFS_I(inode)->root)
+		btrfs_kill_delayed_inode_items(inode);
+
+	key.objectid = ino;
 	key.offset = (u64)-1;
 	key.type = (u8)-1;
 
@@ -2879,9 +4060,8 @@ search_again:
 		leaf = path->nodes[0];
 		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
 		found_type = btrfs_key_type(&found_key);
-		encoding = 0;
 
-		if (found_key.objectid != inode->i_ino)
+		if (found_key.objectid != ino)
 			break;
 
 		if (found_type < min_type)
@@ -2892,16 +4072,12 @@ search_again:
 			fi = btrfs_item_ptr(leaf, path->slots[0],
 					    struct btrfs_file_extent_item);
 			extent_type = btrfs_file_extent_type(leaf, fi);
-			encoding = btrfs_file_extent_compression(leaf, fi);
-			encoding |= btrfs_file_extent_encryption(leaf, fi);
-			encoding |= btrfs_file_extent_other_encoding(leaf, fi);
-
 			if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
 				item_end +=
 				    btrfs_file_extent_num_bytes(leaf, fi);
 			} else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
 				item_end += btrfs_file_extent_inline_len(leaf,
-									 fi);
+							 path->slots[0], fi);
 			}
 			item_end--;
 		}
@@ -2920,21 +4096,27 @@ search_again:
 		if (found_type != BTRFS_EXTENT_DATA_KEY)
 			goto delete;
 
+		if (del_item)
+			last_size = found_key.offset;
+		else
+			last_size = new_size;
+
 		if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
 			u64 num_dec;
 			extent_start = btrfs_file_extent_disk_bytenr(leaf, fi);
-			if (!del_item && !encoding) {
+			if (!del_item) {
 				u64 orig_num_bytes =
 					btrfs_file_extent_num_bytes(leaf, fi);
-				extent_num_bytes = new_size -
-					found_key.offset + root->sectorsize - 1;
-				extent_num_bytes = extent_num_bytes &
-					~((u64)root->sectorsize - 1);
+				extent_num_bytes = ALIGN(new_size -
+						found_key.offset,
+						root->sectorsize);
 				btrfs_set_file_extent_num_bytes(leaf, fi,
 							 extent_num_bytes);
 				num_dec = (orig_num_bytes -
 					   extent_num_bytes);
-				if (root->ref_cows && extent_start != 0)
+				if (test_bit(BTRFS_ROOT_REF_COWS,
+					     &root->state) &&
+				    extent_start != 0)
 					inode_sub_bytes(inode, num_dec);
 				btrfs_mark_buffer_dirty(leaf);
 			} else {
@@ -2948,7 +4130,8 @@ search_again:
 				num_dec = btrfs_file_extent_num_bytes(leaf, fi);
 				if (extent_start != 0) {
 					found_extent = 1;
-					if (root->ref_cows)
+					if (test_bit(BTRFS_ROOT_REF_COWS,
+						     &root->state))
 						inode_sub_bytes(inode, num_dec);
 				}
 			}
@@ -2963,16 +4146,20 @@ search_again:
 			    btrfs_file_extent_other_encoding(leaf, fi) == 0) {
 				u32 size = new_size - found_key.offset;
 
-				if (root->ref_cows) {
+				if (test_bit(BTRFS_ROOT_REF_COWS, &root->state))
 					inode_sub_bytes(inode, item_end + 1 -
 							new_size);
-				}
+
+				/*
+				 * update the ram bytes to properly reflect
+				 * the new size of our item
+				 */
+				btrfs_set_file_extent_ram_bytes(leaf, fi, size);
 				size =
 				    btrfs_file_extent_calc_inline_size(size);
-				ret = btrfs_truncate_item(trans, root, path,
-							  size, 1);
-				BUG_ON(ret);
-			} else if (root->ref_cows) {
+				btrfs_truncate_item(root, path, size, 1);
+			} else if (test_bit(BTRFS_ROOT_REF_COWS,
+					    &root->state)) {
 				inode_sub_bytes(inode, item_end + 1 -
 						found_key.offset);
 			}
@@ -2994,12 +4181,14 @@ delete:
 		} else {
 			break;
 		}
-		if (found_extent && root->ref_cows) {
+		if (found_extent &&
+		    (test_bit(BTRFS_ROOT_REF_COWS, &root->state) ||
+		     root == root->fs_info->tree_root)) {
 			btrfs_set_path_blocking(path);
 			ret = btrfs_free_extent(trans, root, extent_start,
 						extent_num_bytes, 0,
 						btrfs_header_owner(leaf),
-						inode->i_ino, extent_offset);
+						ino, extent_offset, 0);
 			BUG_ON(ret);
 		}
 
@@ -3008,18 +4197,18 @@ delete:
 
 		if (path->slots[0] == 0 ||
 		    path->slots[0] != pending_del_slot) {
-			if (root->ref_cows) {
-				err = -EAGAIN;
-				goto out;
-			}
 			if (pending_del_nr) {
 				ret = btrfs_del_items(trans, root, path,
 						pending_del_slot,
 						pending_del_nr);
-				BUG_ON(ret);
+				if (ret) {
+					btrfs_abort_transaction(trans,
+								root, ret);
+					goto error;
+				}
 				pending_del_nr = 0;
 			}
-			btrfs_release_path(root, path);
+			btrfs_release_path(path);
 			goto search_again;
 		} else {
 			path->slots[0]--;
@@ -3029,18 +4218,31 @@ out:
 	if (pending_del_nr) {
 		ret = btrfs_del_items(trans, root, path, pending_del_slot,
 				      pending_del_nr);
+		if (ret)
+			btrfs_abort_transaction(trans, root, ret);
 	}
+error:
+	if (last_size != (u64)-1)
+		btrfs_ordered_update_i_size(inode, last_size, NULL);
 	btrfs_free_path(path);
 	return err;
 }
 
 /*
- * taken from block_truncate_page, but does cow as it zeros out
- * any bytes left in the last page in the file.
+ * btrfs_truncate_page - read, zero a chunk and write a page
+ * @inode - inode that we're zeroing
+ * @from - the offset to start zeroing
+ * @len - the length to zero, 0 to zero the entire range respective to the
+ *	offset
+ * @front - zero up to the offset instead of from the offset on
+ *
+ * This will find the page for the "from" offset and cow the page and zero the
+ * part we want to zero.  This is used with truncate and hole punching.
  */
-static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
+int btrfs_truncate_page(struct inode *inode, loff_t from, loff_t len,
+			int front)
 {
-	struct inode *inode = mapping->host;
+	struct address_space *mapping = inode->i_mapping;
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
 	struct btrfs_ordered_extent *ordered;
@@ -3050,26 +4252,23 @@ static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
 	pgoff_t index = from >> PAGE_CACHE_SHIFT;
 	unsigned offset = from & (PAGE_CACHE_SIZE-1);
 	struct page *page;
+	gfp_t mask = btrfs_alloc_write_mask(mapping);
 	int ret = 0;
 	u64 page_start;
 	u64 page_end;
 
-	if ((offset & (blocksize - 1)) == 0)
+	if ((offset & (blocksize - 1)) == 0 &&
+	    (!len || ((len & (blocksize - 1)) == 0)))
 		goto out;
-	ret = btrfs_check_data_free_space(root, inode, PAGE_CACHE_SIZE);
-	if (ret)
-		goto out;
-
-	ret = btrfs_reserve_metadata_for_delalloc(root, inode, 1);
+	ret = btrfs_delalloc_reserve_space(inode, PAGE_CACHE_SIZE);
 	if (ret)
 		goto out;
 
-	ret = -ENOMEM;
 again:
-	page = grab_cache_page(mapping, index);
+	page = find_or_create_page(mapping, index, mask);
 	if (!page) {
-		btrfs_free_reserved_data_space(root, inode, PAGE_CACHE_SIZE);
-		btrfs_unreserve_metadata_for_delalloc(root, inode, 1);
+		btrfs_delalloc_release_space(inode, PAGE_CACHE_SIZE);
+		ret = -ENOMEM;
 		goto out;
 	}
 
@@ -3091,8 +4290,7 @@ again:
 	}
 	wait_on_page_writeback(page);
 
-	lock_extent_bits(io_tree, page_start, page_end, 0, &cached_state,
-			 GFP_NOFS);
+	lock_extent_bits(io_tree, page_start, page_end, 0, &cached_state);
 	set_page_extent_mapped(page);
 
 	ordered = btrfs_lookup_ordered_extent(inode, page_start);
@@ -3107,7 +4305,8 @@ again:
 	}
 
 	clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, page_end,
-			  EXTENT_DIRTY | EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING,
+			  EXTENT_DIRTY | EXTENT_DELALLOC |
+			  EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG,
 			  0, 0, &cached_state, GFP_NOFS);
 
 	ret = btrfs_set_extent_delalloc(inode, page_start, page_end,
@@ -3118,10 +4317,14 @@ again:
 		goto out_unlock;
 	}
 
-	ret = 0;
 	if (offset != PAGE_CACHE_SIZE) {
+		if (!len)
+			len = PAGE_CACHE_SIZE - offset;
 		kaddr = kmap(page);
-		memset(kaddr + offset, 0, PAGE_CACHE_SIZE - offset);
+		if (front)
+			memset(kaddr, 0, offset);
+		else
+			memset(kaddr + offset, 0, len);
 		flush_dcache_page(page);
 		kunmap(page);
 	}
@@ -3132,43 +4335,100 @@ again:
 
 out_unlock:
 	if (ret)
-		btrfs_free_reserved_data_space(root, inode, PAGE_CACHE_SIZE);
-	btrfs_unreserve_metadata_for_delalloc(root, inode, 1);
+		btrfs_delalloc_release_space(inode, PAGE_CACHE_SIZE);
 	unlock_page(page);
 	page_cache_release(page);
 out:
 	return ret;
 }
 
-int btrfs_cont_expand(struct inode *inode, loff_t size)
+static int maybe_insert_hole(struct btrfs_root *root, struct inode *inode,
+			     u64 offset, u64 len)
 {
 	struct btrfs_trans_handle *trans;
+	int ret;
+
+	/*
+	 * Still need to make sure the inode looks like it's been updated so
+	 * that any holes get logged if we fsync.
+	 */
+	if (btrfs_fs_incompat(root->fs_info, NO_HOLES)) {
+		BTRFS_I(inode)->last_trans = root->fs_info->generation;
+		BTRFS_I(inode)->last_sub_trans = root->log_transid;
+		BTRFS_I(inode)->last_log_commit = root->last_log_commit;
+		return 0;
+	}
+
+	/*
+	 * 1 - for the one we're dropping
+	 * 1 - for the one we're adding
+	 * 1 - for updating the inode.
+	 */
+	trans = btrfs_start_transaction(root, 3);
+	if (IS_ERR(trans))
+		return PTR_ERR(trans);
+
+	ret = btrfs_drop_extents(trans, root, inode, offset, offset + len, 1);
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
+		btrfs_end_transaction(trans, root);
+		return ret;
+	}
+
+	ret = btrfs_insert_file_extent(trans, root, btrfs_ino(inode), offset,
+				       0, 0, len, 0, len, 0, 0, 0);
+	if (ret)
+		btrfs_abort_transaction(trans, root, ret);
+	else
+		btrfs_update_inode(trans, root, inode);
+	btrfs_end_transaction(trans, root);
+	return ret;
+}
+
+/*
+ * This function puts in dummy file extents for the area we're creating a hole
+ * for.  So if we are truncating this file to a larger size we need to insert
+ * these file extents so that btrfs_get_extent will return a EXTENT_MAP_HOLE for
+ * the range between oldsize and size
+ */
+int btrfs_cont_expand(struct inode *inode, loff_t oldsize, loff_t size)
+{
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
-	struct extent_map *em;
+	struct extent_map *em = NULL;
 	struct extent_state *cached_state = NULL;
-	u64 mask = root->sectorsize - 1;
-	u64 hole_start = (inode->i_size + mask) & ~mask;
-	u64 block_end = (size + mask) & ~mask;
+	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
+	u64 hole_start = ALIGN(oldsize, root->sectorsize);
+	u64 block_end = ALIGN(size, root->sectorsize);
 	u64 last_byte;
 	u64 cur_offset;
 	u64 hole_size;
 	int err = 0;
 
+	/*
+	 * If our size started in the middle of a page we need to zero out the
+	 * rest of the page before we expand the i_size, otherwise we could
+	 * expose stale data.
+	 */
+	err = btrfs_truncate_page(inode, oldsize, 0, 0);
+	if (err)
+		return err;
+
 	if (size <= hole_start)
 		return 0;
 
 	while (1) {
 		struct btrfs_ordered_extent *ordered;
-		btrfs_wait_ordered_range(inode, hole_start,
-					 block_end - hole_start);
+
 		lock_extent_bits(io_tree, hole_start, block_end - 1, 0,
-				 &cached_state, GFP_NOFS);
-		ordered = btrfs_lookup_ordered_extent(inode, hole_start);
+				 &cached_state);
+		ordered = btrfs_lookup_ordered_range(inode, hole_start,
+						     block_end - hole_start);
 		if (!ordered)
 			break;
 		unlock_extent_cached(io_tree, hole_start, block_end - 1,
 				     &cached_state, GFP_NOFS);
+		btrfs_start_ordered_extent(inode, ordered, 1);
 		btrfs_put_ordered_extent(ordered);
 	}
 
@@ -3176,197 +4436,381 @@ int btrfs_cont_expand(struct inode *inode, loff_t size)
 	while (1) {
 		em = btrfs_get_extent(inode, NULL, 0, cur_offset,
 				block_end - cur_offset, 0);
-		BUG_ON(IS_ERR(em) || !em);
+		if (IS_ERR(em)) {
+			err = PTR_ERR(em);
+			em = NULL;
+			break;
+		}
 		last_byte = min(extent_map_end(em), block_end);
-		last_byte = (last_byte + mask) & ~mask;
+		last_byte = ALIGN(last_byte , root->sectorsize);
 		if (!test_bit(EXTENT_FLAG_PREALLOC, &em->flags)) {
-			u64 hint_byte = 0;
+			struct extent_map *hole_em;
 			hole_size = last_byte - cur_offset;
 
-			err = btrfs_reserve_metadata_space(root, 2);
+			err = maybe_insert_hole(root, inode, cur_offset,
+						hole_size);
 			if (err)
 				break;
+			btrfs_drop_extent_cache(inode, cur_offset,
+						cur_offset + hole_size - 1, 0);
+			hole_em = alloc_extent_map();
+			if (!hole_em) {
+				set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
+					&BTRFS_I(inode)->runtime_flags);
+				goto next;
+			}
+			hole_em->start = cur_offset;
+			hole_em->len = hole_size;
+			hole_em->orig_start = cur_offset;
+
+			hole_em->block_start = EXTENT_MAP_HOLE;
+			hole_em->block_len = 0;
+			hole_em->orig_block_len = 0;
+			hole_em->ram_bytes = hole_size;
+			hole_em->bdev = root->fs_info->fs_devices->latest_bdev;
+			hole_em->compress_type = BTRFS_COMPRESS_NONE;
+			hole_em->generation = root->fs_info->generation;
 
-			trans = btrfs_start_transaction(root, 1);
-			btrfs_set_trans_block_group(trans, inode);
-
-			err = btrfs_drop_extents(trans, inode, cur_offset,
-						 cur_offset + hole_size,
-						 &hint_byte, 1);
-			BUG_ON(err);
-
-			err = btrfs_insert_file_extent(trans, root,
-					inode->i_ino, cur_offset, 0,
-					0, hole_size, 0, hole_size,
-					0, 0, 0);
-			BUG_ON(err);
-
-			btrfs_drop_extent_cache(inode, hole_start,
-					last_byte - 1, 0);
-
-			btrfs_end_transaction(trans, root);
-			btrfs_unreserve_metadata_space(root, 2);
+			while (1) {
+				write_lock(&em_tree->lock);
+				err = add_extent_mapping(em_tree, hole_em, 1);
+				write_unlock(&em_tree->lock);
+				if (err != -EEXIST)
+					break;
+				btrfs_drop_extent_cache(inode, cur_offset,
+							cur_offset +
+							hole_size - 1, 0);
+			}
+			free_extent_map(hole_em);
 		}
+next:
 		free_extent_map(em);
+		em = NULL;
 		cur_offset = last_byte;
 		if (cur_offset >= block_end)
 			break;
 	}
-
+	free_extent_map(em);
 	unlock_extent_cached(io_tree, hole_start, block_end - 1, &cached_state,
 			     GFP_NOFS);
 	return err;
 }
 
-static int btrfs_setattr_size(struct inode *inode, struct iattr *attr)
+static int btrfs_setsize(struct inode *inode, struct iattr *attr)
 {
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	struct btrfs_trans_handle *trans;
-	unsigned long nr;
+	loff_t oldsize = i_size_read(inode);
+	loff_t newsize = attr->ia_size;
+	int mask = attr->ia_valid;
 	int ret;
 
-	if (attr->ia_size == inode->i_size)
-		return 0;
-
-	if (attr->ia_size > inode->i_size) {
-		unsigned long limit;
-		limit = current->signal->rlim[RLIMIT_FSIZE].rlim_cur;
-		if (attr->ia_size > inode->i_sb->s_maxbytes)
-			return -EFBIG;
-		if (limit != RLIM_INFINITY && attr->ia_size > limit) {
-			send_sig(SIGXFSZ, current, 0);
-			return -EFBIG;
-		}
+	/*
+	 * The regular truncate() case without ATTR_CTIME and ATTR_MTIME is a
+	 * special case where we need to update the times despite not having
+	 * these flags set.  For all other operations the VFS set these flags
+	 * explicitly if it wants a timestamp update.
+	 */
+	if (newsize != oldsize) {
+		inode_inc_iversion(inode);
+		if (!(mask & (ATTR_CTIME | ATTR_MTIME)))
+			inode->i_ctime = inode->i_mtime =
+				current_fs_time(inode->i_sb);
 	}
 
-	ret = btrfs_reserve_metadata_space(root, 1);
-	if (ret)
-		return ret;
+	if (newsize > oldsize) {
+		truncate_pagecache(inode, newsize);
+		ret = btrfs_cont_expand(inode, oldsize, newsize);
+		if (ret)
+			return ret;
 
-	trans = btrfs_start_transaction(root, 1);
-	btrfs_set_trans_block_group(trans, inode);
+		trans = btrfs_start_transaction(root, 1);
+		if (IS_ERR(trans))
+			return PTR_ERR(trans);
 
-	ret = btrfs_orphan_add(trans, inode);
-	BUG_ON(ret);
+		i_size_write(inode, newsize);
+		btrfs_ordered_update_i_size(inode, i_size_read(inode), NULL);
+		ret = btrfs_update_inode(trans, root, inode);
+		btrfs_end_transaction(trans, root);
+	} else {
 
-	nr = trans->blocks_used;
-	btrfs_end_transaction(trans, root);
-	btrfs_unreserve_metadata_space(root, 1);
-	btrfs_btree_balance_dirty(root, nr);
+		/*
+		 * We're truncating a file that used to have good data down to
+		 * zero. Make sure it gets into the ordered flush list so that
+		 * any new writes get down to disk quickly.
+		 */
+		if (newsize == 0)
+			set_bit(BTRFS_INODE_ORDERED_DATA_CLOSE,
+				&BTRFS_I(inode)->runtime_flags);
 
-	if (attr->ia_size > inode->i_size) {
-		ret = btrfs_cont_expand(inode, attr->ia_size);
-		if (ret) {
-			btrfs_truncate(inode);
+		/*
+		 * 1 for the orphan item we're going to add
+		 * 1 for the orphan item deletion.
+		 */
+		trans = btrfs_start_transaction(root, 2);
+		if (IS_ERR(trans))
+			return PTR_ERR(trans);
+
+		/*
+		 * We need to do this in case we fail at _any_ point during the
+		 * actual truncate.  Once we do the truncate_setsize we could
+		 * invalidate pages which forces any outstanding ordered io to
+		 * be instantly completed which will give us extents that need
+		 * to be truncated.  If we fail to get an orphan inode down we
+		 * could have left over extents that were never meant to live,
+		 * so we need to garuntee from this point on that everything
+		 * will be consistent.
+		 */
+		ret = btrfs_orphan_add(trans, inode);
+		btrfs_end_transaction(trans, root);
+		if (ret)
 			return ret;
-		}
 
-		i_size_write(inode, attr->ia_size);
-		btrfs_ordered_update_i_size(inode, inode->i_size, NULL);
+		/* we don't support swapfiles, so vmtruncate shouldn't fail */
+		truncate_setsize(inode, newsize);
 
-		trans = btrfs_start_transaction(root, 1);
-		btrfs_set_trans_block_group(trans, inode);
+		/* Disable nonlocked read DIO to avoid the end less truncate */
+		btrfs_inode_block_unlocked_dio(inode);
+		inode_dio_wait(inode);
+		btrfs_inode_resume_unlocked_dio(inode);
 
-		ret = btrfs_update_inode(trans, root, inode);
-		BUG_ON(ret);
-		if (inode->i_nlink > 0) {
-			ret = btrfs_orphan_del(trans, inode);
-			BUG_ON(ret);
+		ret = btrfs_truncate(inode);
+		if (ret && inode->i_nlink) {
+			int err;
+
+			/*
+			 * failed to truncate, disk_i_size is only adjusted down
+			 * as we remove extents, so it should represent the true
+			 * size of the inode, so reset the in memory size and
+			 * delete our orphan entry.
+			 */
+			trans = btrfs_join_transaction(root);
+			if (IS_ERR(trans)) {
+				btrfs_orphan_del(NULL, inode);
+				return ret;
+			}
+			i_size_write(inode, BTRFS_I(inode)->disk_i_size);
+			err = btrfs_orphan_del(trans, inode);
+			if (err)
+				btrfs_abort_transaction(trans, root, err);
+			btrfs_end_transaction(trans, root);
 		}
-		nr = trans->blocks_used;
-		btrfs_end_transaction(trans, root);
-		btrfs_btree_balance_dirty(root, nr);
-		return 0;
 	}
 
-	/*
-	 * We're truncating a file that used to have good data down to
-	 * zero. Make sure it gets into the ordered flush list so that
-	 * any new writes get down to disk quickly.
-	 */
-	if (attr->ia_size == 0)
-		BTRFS_I(inode)->ordered_data_close = 1;
-
-	/* we don't support swapfiles, so vmtruncate shouldn't fail */
-	ret = vmtruncate(inode, attr->ia_size);
-	BUG_ON(ret);
-
-	return 0;
+	return ret;
 }
 
 static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
 {
 	struct inode *inode = dentry->d_inode;
+	struct btrfs_root *root = BTRFS_I(inode)->root;
 	int err;
 
+	if (btrfs_root_readonly(root))
+		return -EROFS;
+
 	err = inode_change_ok(inode, attr);
 	if (err)
 		return err;
 
 	if (S_ISREG(inode->i_mode) && (attr->ia_valid & ATTR_SIZE)) {
-		err = btrfs_setattr_size(inode, attr);
+		err = btrfs_setsize(inode, attr);
 		if (err)
 			return err;
 	}
-	attr->ia_valid &= ~ATTR_SIZE;
 
-	if (attr->ia_valid)
-		err = inode_setattr(inode, attr);
+	if (attr->ia_valid) {
+		setattr_copy(inode, attr);
+		inode_inc_iversion(inode);
+		err = btrfs_dirty_inode(inode);
+
+		if (!err && attr->ia_valid & ATTR_MODE)
+			err = posix_acl_chmod(inode, inode->i_mode);
+	}
 
-	if (!err && ((attr->ia_valid & ATTR_MODE)))
-		err = btrfs_acl_chmod(inode);
 	return err;
 }
 
-void btrfs_delete_inode(struct inode *inode)
+/*
+ * While truncating the inode pages during eviction, we get the VFS calling
+ * btrfs_invalidatepage() against each page of the inode. This is slow because
+ * the calls to btrfs_invalidatepage() result in a huge amount of calls to
+ * lock_extent_bits() and clear_extent_bit(), which keep merging and splitting
+ * extent_state structures over and over, wasting lots of time.
+ *
+ * Therefore if the inode is being evicted, let btrfs_invalidatepage() skip all
+ * those expensive operations on a per page basis and do only the ordered io
+ * finishing, while we release here the extent_map and extent_state structures,
+ * without the excessive merging and splitting.
+ */
+static void evict_inode_truncate_pages(struct inode *inode)
+{
+	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
+	struct extent_map_tree *map_tree = &BTRFS_I(inode)->extent_tree;
+	struct rb_node *node;
+
+	ASSERT(inode->i_state & I_FREEING);
+	truncate_inode_pages_final(&inode->i_data);
+
+	write_lock(&map_tree->lock);
+	while (!RB_EMPTY_ROOT(&map_tree->map)) {
+		struct extent_map *em;
+
+		node = rb_first(&map_tree->map);
+		em = rb_entry(node, struct extent_map, rb_node);
+		clear_bit(EXTENT_FLAG_PINNED, &em->flags);
+		clear_bit(EXTENT_FLAG_LOGGING, &em->flags);
+		remove_extent_mapping(map_tree, em);
+		free_extent_map(em);
+	}
+	write_unlock(&map_tree->lock);
+
+	spin_lock(&io_tree->lock);
+	while (!RB_EMPTY_ROOT(&io_tree->state)) {
+		struct extent_state *state;
+		struct extent_state *cached_state = NULL;
+
+		node = rb_first(&io_tree->state);
+		state = rb_entry(node, struct extent_state, rb_node);
+		atomic_inc(&state->refs);
+		spin_unlock(&io_tree->lock);
+
+		lock_extent_bits(io_tree, state->start, state->end,
+				 0, &cached_state);
+		clear_extent_bit(io_tree, state->start, state->end,
+				 EXTENT_LOCKED | EXTENT_DIRTY |
+				 EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING |
+				 EXTENT_DEFRAG, 1, 1,
+				 &cached_state, GFP_NOFS);
+		free_extent_state(state);
+
+		spin_lock(&io_tree->lock);
+	}
+	spin_unlock(&io_tree->lock);
+}
+
+void btrfs_evict_inode(struct inode *inode)
 {
 	struct btrfs_trans_handle *trans;
 	struct btrfs_root *root = BTRFS_I(inode)->root;
-	unsigned long nr;
+	struct btrfs_block_rsv *rsv, *global_rsv;
+	u64 min_size = btrfs_calc_trunc_metadata_size(root, 1);
 	int ret;
 
-	truncate_inode_pages(&inode->i_data, 0);
+	trace_btrfs_inode_evict(inode);
+
+	evict_inode_truncate_pages(inode);
+
+	if (inode->i_nlink &&
+	    ((btrfs_root_refs(&root->root_item) != 0 &&
+	      root->root_key.objectid != BTRFS_ROOT_TREE_OBJECTID) ||
+	     btrfs_is_free_space_inode(inode)))
+		goto no_delete;
+
 	if (is_bad_inode(inode)) {
 		btrfs_orphan_del(NULL, inode);
 		goto no_delete;
 	}
+	/* do we really want it for ->i_nlink > 0 and zero btrfs_root_refs? */
 	btrfs_wait_ordered_range(inode, 0, (u64)-1);
 
 	if (root->fs_info->log_root_recovering) {
-		BUG_ON(!list_empty(&BTRFS_I(inode)->i_orphan));
+		BUG_ON(test_bit(BTRFS_INODE_HAS_ORPHAN_ITEM,
+				 &BTRFS_I(inode)->runtime_flags));
 		goto no_delete;
 	}
 
 	if (inode->i_nlink > 0) {
-		BUG_ON(btrfs_root_refs(&root->root_item) != 0);
+		BUG_ON(btrfs_root_refs(&root->root_item) != 0 &&
+		       root->root_key.objectid != BTRFS_ROOT_TREE_OBJECTID);
 		goto no_delete;
 	}
 
+	ret = btrfs_commit_inode_delayed_inode(inode);
+	if (ret) {
+		btrfs_orphan_del(NULL, inode);
+		goto no_delete;
+	}
+
+	rsv = btrfs_alloc_block_rsv(root, BTRFS_BLOCK_RSV_TEMP);
+	if (!rsv) {
+		btrfs_orphan_del(NULL, inode);
+		goto no_delete;
+	}
+	rsv->size = min_size;
+	rsv->failfast = 1;
+	global_rsv = &root->fs_info->global_block_rsv;
+
 	btrfs_i_size_write(inode, 0);
 
+	/*
+	 * This is a bit simpler than btrfs_truncate since we've already
+	 * reserved our space for our orphan item in the unlink, so we just
+	 * need to reserve some slack space in case we add bytes and update
+	 * inode item when doing the truncate.
+	 */
 	while (1) {
-		trans = btrfs_start_transaction(root, 1);
-		btrfs_set_trans_block_group(trans, inode);
-		ret = btrfs_truncate_inode_items(trans, root, inode, 0, 0);
+		ret = btrfs_block_rsv_refill(root, rsv, min_size,
+					     BTRFS_RESERVE_FLUSH_LIMIT);
+
+		/*
+		 * Try and steal from the global reserve since we will
+		 * likely not use this space anyway, we want to try as
+		 * hard as possible to get this to work.
+		 */
+		if (ret)
+			ret = btrfs_block_rsv_migrate(global_rsv, rsv, min_size);
 
-		if (ret != -EAGAIN)
+		if (ret) {
+			btrfs_warn(root->fs_info,
+				"Could not get space for a delete, will truncate on mount %d",
+				ret);
+			btrfs_orphan_del(NULL, inode);
+			btrfs_free_block_rsv(root, rsv);
+			goto no_delete;
+		}
+
+		trans = btrfs_join_transaction(root);
+		if (IS_ERR(trans)) {
+			btrfs_orphan_del(NULL, inode);
+			btrfs_free_block_rsv(root, rsv);
+			goto no_delete;
+		}
+
+		trans->block_rsv = rsv;
+
+		ret = btrfs_truncate_inode_items(trans, root, inode, 0, 0);
+		if (ret != -ENOSPC)
 			break;
 
-		nr = trans->blocks_used;
+		trans->block_rsv = &root->fs_info->trans_block_rsv;
 		btrfs_end_transaction(trans, root);
 		trans = NULL;
-		btrfs_btree_balance_dirty(root, nr);
+		btrfs_btree_balance_dirty(root);
 	}
 
+	btrfs_free_block_rsv(root, rsv);
+
+	/*
+	 * Errors here aren't a big deal, it just means we leave orphan items
+	 * in the tree.  They will be cleaned up on the next mount.
+	 */
 	if (ret == 0) {
-		ret = btrfs_orphan_del(trans, inode);
-		BUG_ON(ret);
+		trans->block_rsv = root->orphan_block_rsv;
+		btrfs_orphan_del(trans, inode);
+	} else {
+		btrfs_orphan_del(NULL, inode);
 	}
 
-	nr = trans->blocks_used;
+	trans->block_rsv = &root->fs_info->trans_block_rsv;
+	if (!(root == root->fs_info->tree_root ||
+	      root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID))
+		btrfs_return_ino(root, btrfs_ino(inode));
+
 	btrfs_end_transaction(trans, root);
-	btrfs_btree_balance_dirty(root, nr);
+	btrfs_btree_balance_dirty(root);
 no_delete:
+	btrfs_remove_delayed_node(inode);
 	clear_inode(inode);
 	return;
 }
@@ -3386,14 +4830,15 @@ static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry,
 	int ret = 0;
 
 	path = btrfs_alloc_path();
-	BUG_ON(!path);
+	if (!path)
+		return -ENOMEM;
 
-	di = btrfs_lookup_dir_item(NULL, root, path, dir->i_ino, name,
+	di = btrfs_lookup_dir_item(NULL, root, path, btrfs_ino(dir), name,
 				    namelen, 0);
 	if (IS_ERR(di))
 		ret = PTR_ERR(di);
 
-	if (!di || IS_ERR(di))
+	if (IS_ERR_OR_NULL(di))
 		goto out_err;
 
 	btrfs_dir_item_key_to_cpu(path->nodes[0], di, location);
@@ -3430,9 +4875,9 @@ static int fixup_tree_root_location(struct btrfs_root *root,
 	}
 
 	err = -ENOENT;
-	ret = btrfs_find_root_ref(root->fs_info->tree_root, path,
-				  BTRFS_I(dir)->root->root_key.objectid,
-				  location->objectid);
+	ret = btrfs_find_item(root->fs_info->tree_root, path,
+				BTRFS_I(dir)->root->root_key.objectid,
+				location->objectid, BTRFS_ROOT_REF_KEY, NULL);
 	if (ret) {
 		if (ret < 0)
 			err = ret;
@@ -3441,7 +4886,7 @@ static int fixup_tree_root_location(struct btrfs_root *root,
 
 	leaf = path->nodes[0];
 	ref = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_root_ref);
-	if (btrfs_root_ref_dirid(leaf, ref) != dir->i_ino ||
+	if (btrfs_root_ref_dirid(leaf, ref) != btrfs_ino(dir) ||
 	    btrfs_root_ref_name_len(leaf, ref) != dentry->d_name.len)
 		goto out;
 
@@ -3451,7 +4896,7 @@ static int fixup_tree_root_location(struct btrfs_root *root,
 	if (ret)
 		goto out;
 
-	btrfs_release_path(root->fs_info->tree_root, path);
+	btrfs_release_path(path);
 
 	new_root = btrfs_read_fs_root_no_name(root->fs_info, location);
 	if (IS_ERR(new_root)) {
@@ -3459,11 +4904,6 @@ static int fixup_tree_root_location(struct btrfs_root *root,
 		goto out;
 	}
 
-	if (btrfs_root_refs(&new_root->root_item) == 0) {
-		err = -ENOENT;
-		goto out;
-	}
-
 	*sub_root = new_root;
 	location->objectid = btrfs_root_dirid(&new_root->root_item);
 	location->type = BTRFS_INODE_ITEM_KEY;
@@ -3480,33 +4920,33 @@ static void inode_tree_add(struct inode *inode)
 	struct btrfs_inode *entry;
 	struct rb_node **p;
 	struct rb_node *parent;
-again:
-	p = &root->inode_tree.rb_node;
-	parent = NULL;
+	struct rb_node *new = &BTRFS_I(inode)->rb_node;
+	u64 ino = btrfs_ino(inode);
 
-	if (hlist_unhashed(&inode->i_hash))
+	if (inode_unhashed(inode))
 		return;
-
+	parent = NULL;
 	spin_lock(&root->inode_lock);
+	p = &root->inode_tree.rb_node;
 	while (*p) {
 		parent = *p;
 		entry = rb_entry(parent, struct btrfs_inode, rb_node);
 
-		if (inode->i_ino < entry->vfs_inode.i_ino)
+		if (ino < btrfs_ino(&entry->vfs_inode))
 			p = &parent->rb_left;
-		else if (inode->i_ino > entry->vfs_inode.i_ino)
+		else if (ino > btrfs_ino(&entry->vfs_inode))
 			p = &parent->rb_right;
 		else {
 			WARN_ON(!(entry->vfs_inode.i_state &
-				  (I_WILL_FREE | I_FREEING | I_CLEAR)));
-			rb_erase(parent, &root->inode_tree);
+				  (I_WILL_FREE | I_FREEING)));
+			rb_replace_node(parent, new, &root->inode_tree);
 			RB_CLEAR_NODE(parent);
 			spin_unlock(&root->inode_lock);
-			goto again;
+			return;
 		}
 	}
-	rb_link_node(&BTRFS_I(inode)->rb_node, parent, p);
-	rb_insert_color(&BTRFS_I(inode)->rb_node, &root->inode_tree);
+	rb_link_node(new, parent, p);
+	rb_insert_color(new, &root->inode_tree);
 	spin_unlock(&root->inode_lock);
 }
 
@@ -3533,7 +4973,7 @@ static void inode_tree_del(struct inode *inode)
 	}
 }
 
-int btrfs_invalidate_inodes(struct btrfs_root *root)
+void btrfs_invalidate_inodes(struct btrfs_root *root)
 {
 	struct rb_node *node;
 	struct rb_node *prev;
@@ -3541,7 +4981,8 @@ int btrfs_invalidate_inodes(struct btrfs_root *root)
 	struct inode *inode;
 	u64 objectid = 0;
 
-	WARN_ON(btrfs_root_refs(&root->root_item) != 0);
+	if (!test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state))
+		WARN_ON(btrfs_root_refs(&root->root_item) != 0);
 
 	spin_lock(&root->inode_lock);
 again:
@@ -3551,9 +4992,9 @@ again:
 		prev = node;
 		entry = rb_entry(node, struct btrfs_inode, rb_node);
 
-		if (objectid < entry->vfs_inode.i_ino)
+		if (objectid < btrfs_ino(&entry->vfs_inode))
 			node = node->rb_left;
-		else if (objectid > entry->vfs_inode.i_ino)
+		else if (objectid > btrfs_ino(&entry->vfs_inode))
 			node = node->rb_right;
 		else
 			break;
@@ -3561,7 +5002,7 @@ again:
 	if (!node) {
 		while (prev) {
 			entry = rb_entry(prev, struct btrfs_inode, rb_node);
-			if (objectid <= entry->vfs_inode.i_ino) {
+			if (objectid <= btrfs_ino(&entry->vfs_inode)) {
 				node = prev;
 				break;
 			}
@@ -3570,14 +5011,14 @@ again:
 	}
 	while (node) {
 		entry = rb_entry(node, struct btrfs_inode, rb_node);
-		objectid = entry->vfs_inode.i_ino + 1;
+		objectid = btrfs_ino(&entry->vfs_inode) + 1;
 		inode = igrab(&entry->vfs_inode);
 		if (inode) {
 			spin_unlock(&root->inode_lock);
 			if (atomic_read(&inode->i_count) > 1)
 				d_prune_aliases(inode);
 			/*
-			 * btrfs_drop_inode will remove it from
+			 * btrfs_drop_inode will have it removed from
 			 * the inode cache when its usage count
 			 * hits zero.
 			 */
@@ -3593,65 +5034,37 @@ again:
 		node = rb_next(node);
 	}
 	spin_unlock(&root->inode_lock);
-	return 0;
-}
-
-static noinline void init_btrfs_i(struct inode *inode)
-{
-	struct btrfs_inode *bi = BTRFS_I(inode);
-
-	bi->generation = 0;
-	bi->sequence = 0;
-	bi->last_trans = 0;
-	bi->last_sub_trans = 0;
-	bi->logged_trans = 0;
-	bi->delalloc_bytes = 0;
-	bi->reserved_bytes = 0;
-	bi->disk_i_size = 0;
-	bi->flags = 0;
-	bi->index_cnt = (u64)-1;
-	bi->last_unlink_trans = 0;
-	bi->ordered_data_close = 0;
-	bi->force_compress = 0;
-	extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
-	extent_io_tree_init(&BTRFS_I(inode)->io_tree,
-			     inode->i_mapping, GFP_NOFS);
-	extent_io_tree_init(&BTRFS_I(inode)->io_failure_tree,
-			     inode->i_mapping, GFP_NOFS);
-	INIT_LIST_HEAD(&BTRFS_I(inode)->delalloc_inodes);
-	INIT_LIST_HEAD(&BTRFS_I(inode)->ordered_operations);
-	RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node);
-	btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree);
-	mutex_init(&BTRFS_I(inode)->log_mutex);
 }
 
 static int btrfs_init_locked_inode(struct inode *inode, void *p)
 {
 	struct btrfs_iget_args *args = p;
-	inode->i_ino = args->ino;
-	init_btrfs_i(inode);
+	inode->i_ino = args->location->objectid;
+	memcpy(&BTRFS_I(inode)->location, args->location,
+	       sizeof(*args->location));
 	BTRFS_I(inode)->root = args->root;
-	btrfs_set_inode_space_info(args->root, inode);
 	return 0;
 }
 
 static int btrfs_find_actor(struct inode *inode, void *opaque)
 {
 	struct btrfs_iget_args *args = opaque;
-	return args->ino == inode->i_ino &&
+	return args->location->objectid == BTRFS_I(inode)->location.objectid &&
 		args->root == BTRFS_I(inode)->root;
 }
 
 static struct inode *btrfs_iget_locked(struct super_block *s,
-				       u64 objectid,
+				       struct btrfs_key *location,
 				       struct btrfs_root *root)
 {
 	struct inode *inode;
 	struct btrfs_iget_args args;
-	args.ino = objectid;
+	unsigned long hashval = btrfs_inode_hash(location->objectid, root);
+
+	args.location = location;
 	args.root = root;
 
-	inode = iget5_locked(s, objectid, btrfs_find_actor,
+	inode = iget5_locked(s, hashval, btrfs_find_actor,
 			     btrfs_init_locked_inode,
 			     (void *)&args);
 	return inode;
@@ -3665,19 +5078,22 @@ struct inode *btrfs_iget(struct super_block *s, struct btrfs_key *location,
 {
 	struct inode *inode;
 
-	inode = btrfs_iget_locked(s, location->objectid, root);
+	inode = btrfs_iget_locked(s, location, root);
 	if (!inode)
 		return ERR_PTR(-ENOMEM);
 
 	if (inode->i_state & I_NEW) {
-		BTRFS_I(inode)->root = root;
-		memcpy(&BTRFS_I(inode)->location, location, sizeof(*location));
 		btrfs_read_locked_inode(inode);
-
-		inode_tree_add(inode);
-		unlock_new_inode(inode);
-		if (new)
-			*new = 1;
+		if (!is_bad_inode(inode)) {
+			inode_tree_add(inode);
+			unlock_new_inode(inode);
+			if (new)
+				*new = 1;
+		} else {
+			unlock_new_inode(inode);
+			iput(inode);
+			inode = ERR_PTR(-ESTALE);
+		}
 	}
 
 	return inode;
@@ -3692,14 +5108,12 @@ static struct inode *new_simple_dir(struct super_block *s,
 	if (!inode)
 		return ERR_PTR(-ENOMEM);
 
-	init_btrfs_i(inode);
-
 	BTRFS_I(inode)->root = root;
 	memcpy(&BTRFS_I(inode)->location, key, sizeof(*key));
-	BTRFS_I(inode)->dummy_inode = 1;
+	set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags);
 
 	inode->i_ino = BTRFS_EMPTY_SUBVOL_DIR_OBJECTID;
-	inode->i_op = &simple_dir_inode_operations;
+	inode->i_op = &btrfs_dir_ro_inode_operations;
 	inode->i_fop = &simple_dir_operations;
 	inode->i_mode = S_IFDIR | S_IRUGO | S_IWUSR | S_IXUGO;
 	inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
@@ -3714,20 +5128,17 @@ struct inode *btrfs_lookup_dentry(struct inode *dir, struct dentry *dentry)
 	struct btrfs_root *sub_root = root;
 	struct btrfs_key location;
 	int index;
-	int ret;
-
-	dentry->d_op = &btrfs_dentry_operations;
+	int ret = 0;
 
 	if (dentry->d_name.len > BTRFS_NAME_LEN)
 		return ERR_PTR(-ENAMETOOLONG);
 
 	ret = btrfs_inode_by_name(dir, dentry, &location);
-
 	if (ret < 0)
 		return ERR_PTR(ret);
 
 	if (location.objectid == 0)
-		return NULL;
+		return ERR_PTR(-ENOENT);
 
 	if (location.type == BTRFS_INODE_ITEM_KEY) {
 		inode = btrfs_iget(dir->i_sb, &location, root, NULL);
@@ -3749,62 +5160,78 @@ struct inode *btrfs_lookup_dentry(struct inode *dir, struct dentry *dentry)
 	}
 	srcu_read_unlock(&root->fs_info->subvol_srcu, index);
 
-	if (root != sub_root) {
+	if (!IS_ERR(inode) && root != sub_root) {
 		down_read(&root->fs_info->cleanup_work_sem);
 		if (!(inode->i_sb->s_flags & MS_RDONLY))
-			btrfs_orphan_cleanup(sub_root);
+			ret = btrfs_orphan_cleanup(sub_root);
 		up_read(&root->fs_info->cleanup_work_sem);
+		if (ret) {
+			iput(inode);
+			inode = ERR_PTR(ret);
+		}
 	}
 
 	return inode;
 }
 
-static int btrfs_dentry_delete(struct dentry *dentry)
+static int btrfs_dentry_delete(const struct dentry *dentry)
 {
 	struct btrfs_root *root;
+	struct inode *inode = dentry->d_inode;
 
-	if (!dentry->d_inode && !IS_ROOT(dentry))
-		dentry = dentry->d_parent;
+	if (!inode && !IS_ROOT(dentry))
+		inode = dentry->d_parent->d_inode;
 
-	if (dentry->d_inode) {
-		root = BTRFS_I(dentry->d_inode)->root;
+	if (inode) {
+		root = BTRFS_I(inode)->root;
 		if (btrfs_root_refs(&root->root_item) == 0)
 			return 1;
+
+		if (btrfs_ino(inode) == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID)
+			return 1;
 	}
 	return 0;
 }
 
+static void btrfs_dentry_release(struct dentry *dentry)
+{
+	kfree(dentry->d_fsdata);
+}
+
 static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
-				   struct nameidata *nd)
+				   unsigned int flags)
 {
 	struct inode *inode;
 
 	inode = btrfs_lookup_dentry(dir, dentry);
-	if (IS_ERR(inode))
-		return ERR_CAST(inode);
+	if (IS_ERR(inode)) {
+		if (PTR_ERR(inode) == -ENOENT)
+			inode = NULL;
+		else
+			return ERR_CAST(inode);
+	}
 
-	return d_splice_alias(inode, dentry);
+	return d_materialise_unique(dentry, inode);
 }
 
-static unsigned char btrfs_filetype_table[] = {
+unsigned char btrfs_filetype_table[] = {
 	DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
 };
 
-static int btrfs_real_readdir(struct file *filp, void *dirent,
-			      filldir_t filldir)
+static int btrfs_real_readdir(struct file *file, struct dir_context *ctx)
 {
-	struct inode *inode = filp->f_dentry->d_inode;
+	struct inode *inode = file_inode(file);
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	struct btrfs_item *item;
 	struct btrfs_dir_item *di;
 	struct btrfs_key key;
 	struct btrfs_key found_key;
 	struct btrfs_path *path;
+	struct list_head ins_list;
+	struct list_head del_list;
 	int ret;
-	u32 nritems;
 	struct extent_buffer *leaf;
 	int slot;
-	int advance;
 	unsigned char d_type;
 	int over = 0;
 	u32 di_cur;
@@ -3814,71 +5241,63 @@ static int btrfs_real_readdir(struct file *filp, void *dirent,
 	char tmp_name[32];
 	char *name_ptr;
 	int name_len;
+	int is_curr = 0;	/* ctx->pos points to the current index? */
 
 	/* FIXME, use a real flag for deciding about the key type */
 	if (root->fs_info->tree_root == root)
 		key_type = BTRFS_DIR_ITEM_KEY;
 
-	/* special case for "." */
-	if (filp->f_pos == 0) {
-		over = filldir(dirent, ".", 1,
-			       1, inode->i_ino,
-			       DT_DIR);
-		if (over)
-			return 0;
-		filp->f_pos = 1;
-	}
-	/* special case for .., just use the back ref */
-	if (filp->f_pos == 1) {
-		u64 pino = parent_ino(filp->f_path.dentry);
-		over = filldir(dirent, "..", 2,
-			       2, pino, DT_DIR);
-		if (over)
-			return 0;
-		filp->f_pos = 2;
-	}
+	if (!dir_emit_dots(file, ctx))
+		return 0;
+
 	path = btrfs_alloc_path();
-	path->reada = 2;
+	if (!path)
+		return -ENOMEM;
+
+	path->reada = 1;
+
+	if (key_type == BTRFS_DIR_INDEX_KEY) {
+		INIT_LIST_HEAD(&ins_list);
+		INIT_LIST_HEAD(&del_list);
+		btrfs_get_delayed_items(inode, &ins_list, &del_list);
+	}
 
 	btrfs_set_key_type(&key, key_type);
-	key.offset = filp->f_pos;
-	key.objectid = inode->i_ino;
+	key.offset = ctx->pos;
+	key.objectid = btrfs_ino(inode);
 
 	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
 	if (ret < 0)
 		goto err;
-	advance = 0;
 
 	while (1) {
 		leaf = path->nodes[0];
-		nritems = btrfs_header_nritems(leaf);
 		slot = path->slots[0];
-		if (advance || slot >= nritems) {
-			if (slot >= nritems - 1) {
-				ret = btrfs_next_leaf(root, path);
-				if (ret)
-					break;
-				leaf = path->nodes[0];
-				nritems = btrfs_header_nritems(leaf);
-				slot = path->slots[0];
-			} else {
-				slot++;
-				path->slots[0]++;
-			}
+		if (slot >= btrfs_header_nritems(leaf)) {
+			ret = btrfs_next_leaf(root, path);
+			if (ret < 0)
+				goto err;
+			else if (ret > 0)
+				break;
+			continue;
 		}
 
-		advance = 1;
-		item = btrfs_item_nr(leaf, slot);
+		item = btrfs_item_nr(slot);
 		btrfs_item_key_to_cpu(leaf, &found_key, slot);
 
 		if (found_key.objectid != key.objectid)
 			break;
 		if (btrfs_key_type(&found_key) != key_type)
 			break;
-		if (found_key.offset < filp->f_pos)
-			continue;
+		if (found_key.offset < ctx->pos)
+			goto next;
+		if (key_type == BTRFS_DIR_INDEX_KEY &&
+		    btrfs_should_delete_dir_index(&del_list,
+						  found_key.offset))
+			goto next;
 
-		filp->f_pos = found_key.offset;
+		ctx->pos = found_key.offset;
+		is_curr = 1;
 
 		di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
 		di_cur = 0;
@@ -3887,6 +5306,9 @@ static int btrfs_real_readdir(struct file *filp, void *dirent,
 		while (di_cur < di_total) {
 			struct btrfs_key location;
 
+			if (verify_dir_item(root, leaf, di))
+				break;
+
 			name_len = btrfs_dir_name_len(leaf, di);
 			if (name_len <= sizeof(tmp_name)) {
 				name_ptr = tmp_name;
@@ -3903,17 +5325,23 @@ static int btrfs_real_readdir(struct file *filp, void *dirent,
 			d_type = btrfs_filetype_table[btrfs_dir_type(leaf, di)];
 			btrfs_dir_item_key_to_cpu(leaf, di, &location);
 
+
 			/* is this a reference to our own snapshot? If so
-			 * skip it
+			 * skip it.
+			 *
+			 * In contrast to old kernels, we insert the snapshot's
+			 * dir item and dir index after it has been created, so
+			 * we won't find a reference to our own snapshot. We
+			 * still keep the following code for backward
+			 * compatibility.
 			 */
 			if (location.type == BTRFS_ROOT_ITEM_KEY &&
 			    location.objectid == root->root_key.objectid) {
 				over = 0;
 				goto skip;
 			}
-			over = filldir(dirent, name_ptr, name_len,
-				       found_key.offset, location.objectid,
-				       d_type);
+			over = !dir_emit(ctx, name_ptr, name_len,
+				       location.objectid, d_type);
 
 skip:
 			if (name_ptr != tmp_name)
@@ -3926,20 +5354,49 @@ skip:
 			di_cur += di_len;
 			di = (struct btrfs_dir_item *)((char *)di + di_len);
 		}
+next:
+		path->slots[0]++;
+	}
+
+	if (key_type == BTRFS_DIR_INDEX_KEY) {
+		if (is_curr)
+			ctx->pos++;
+		ret = btrfs_readdir_delayed_dir_index(ctx, &ins_list);
+		if (ret)
+			goto nopos;
 	}
 
 	/* Reached end of directory/root. Bump pos past the last item. */
-	if (key_type == BTRFS_DIR_INDEX_KEY)
-		/*
-		 * 32-bit glibc will use getdents64, but then strtol -
-		 * so the last number we can serve is this.
-		 */
-		filp->f_pos = 0x7fffffff;
-	else
-		filp->f_pos++;
+	ctx->pos++;
+
+	/*
+	 * Stop new entries from being returned after we return the last
+	 * entry.
+	 *
+	 * New directory entries are assigned a strictly increasing
+	 * offset.  This means that new entries created during readdir
+	 * are *guaranteed* to be seen in the future by that readdir.
+	 * This has broken buggy programs which operate on names as
+	 * they're returned by readdir.  Until we re-use freed offsets
+	 * we have this hack to stop new entries from being returned
+	 * under the assumption that they'll never reach this huge
+	 * offset.
+	 *
+	 * This is being careful not to overflow 32bit loff_t unless the
+	 * last entry requires it because doing so has broken 32bit apps
+	 * in the past.
+	 */
+	if (key_type == BTRFS_DIR_INDEX_KEY) {
+		if (ctx->pos >= INT_MAX)
+			ctx->pos = LLONG_MAX;
+		else
+			ctx->pos = INT_MAX;
+	}
 nopos:
 	ret = 0;
 err:
+	if (key_type == BTRFS_DIR_INDEX_KEY)
+		btrfs_put_delayed_items(&ins_list, &del_list);
 	btrfs_free_path(path);
 	return ret;
 }
@@ -3949,13 +5406,21 @@ int btrfs_write_inode(struct inode *inode, struct writeback_control *wbc)
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	struct btrfs_trans_handle *trans;
 	int ret = 0;
+	bool nolock = false;
 
-	if (root->fs_info->btree_inode == inode)
+	if (test_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags))
 		return 0;
 
+	if (btrfs_fs_closing(root->fs_info) && btrfs_is_free_space_inode(inode))
+		nolock = true;
+
 	if (wbc->sync_mode == WB_SYNC_ALL) {
-		trans = btrfs_join_transaction(root, 1);
-		btrfs_set_trans_block_group(trans, inode);
+		if (nolock)
+			trans = btrfs_join_transaction_nolock(root);
+		else
+			trans = btrfs_join_transaction(root);
+		if (IS_ERR(trans))
+			return PTR_ERR(trans);
 		ret = btrfs_commit_transaction(trans, root);
 	}
 	return ret;
@@ -3967,15 +5432,57 @@ int btrfs_write_inode(struct inode *inode, struct writeback_control *wbc)
  * FIXME, needs more benchmarking...there are no reasons other than performance
  * to keep or drop this code.
  */
-void btrfs_dirty_inode(struct inode *inode)
+static int btrfs_dirty_inode(struct inode *inode)
 {
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	struct btrfs_trans_handle *trans;
+	int ret;
+
+	if (test_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags))
+		return 0;
 
-	trans = btrfs_join_transaction(root, 1);
-	btrfs_set_trans_block_group(trans, inode);
-	btrfs_update_inode(trans, root, inode);
+	trans = btrfs_join_transaction(root);
+	if (IS_ERR(trans))
+		return PTR_ERR(trans);
+
+	ret = btrfs_update_inode(trans, root, inode);
+	if (ret && ret == -ENOSPC) {
+		/* whoops, lets try again with the full transaction */
+		btrfs_end_transaction(trans, root);
+		trans = btrfs_start_transaction(root, 1);
+		if (IS_ERR(trans))
+			return PTR_ERR(trans);
+
+		ret = btrfs_update_inode(trans, root, inode);
+	}
 	btrfs_end_transaction(trans, root);
+	if (BTRFS_I(inode)->delayed_node)
+		btrfs_balance_delayed_items(root);
+
+	return ret;
+}
+
+/*
+ * This is a copy of file_update_time.  We need this so we can return error on
+ * ENOSPC for updating the inode in the case of file write and mmap writes.
+ */
+static int btrfs_update_time(struct inode *inode, struct timespec *now,
+			     int flags)
+{
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+
+	if (btrfs_root_readonly(root))
+		return -EROFS;
+
+	if (flags & S_VERSION)
+		inode_inc_iversion(inode);
+	if (flags & S_CTIME)
+		inode->i_ctime = *now;
+	if (flags & S_MTIME)
+		inode->i_mtime = *now;
+	if (flags & S_ATIME)
+		inode->i_atime = *now;
+	return btrfs_dirty_inode(inode);
 }
 
 /*
@@ -3991,7 +5498,7 @@ static int btrfs_set_inode_index_count(struct inode *inode)
 	struct extent_buffer *leaf;
 	int ret;
 
-	key.objectid = inode->i_ino;
+	key.objectid = btrfs_ino(inode);
 	btrfs_set_key_type(&key, BTRFS_DIR_INDEX_KEY);
 	key.offset = (u64)-1;
 
@@ -4023,7 +5530,7 @@ static int btrfs_set_inode_index_count(struct inode *inode)
 	leaf = path->nodes[0];
 	btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
 
-	if (found_key.objectid != inode->i_ino ||
+	if (found_key.objectid != btrfs_ino(inode) ||
 	    btrfs_key_type(&found_key) != BTRFS_DIR_INDEX_KEY) {
 		BTRFS_I(inode)->index_cnt = 2;
 		goto out;
@@ -4044,9 +5551,12 @@ int btrfs_set_inode_index(struct inode *dir, u64 *index)
 	int ret = 0;
 
 	if (BTRFS_I(dir)->index_cnt == (u64)-1) {
-		ret = btrfs_set_inode_index_count(dir);
-		if (ret)
-			return ret;
+		ret = btrfs_inode_delayed_dir_index_count(dir);
+		if (ret) {
+			ret = btrfs_set_inode_index_count(dir);
+			if (ret)
+				return ret;
+		}
 	}
 
 	*index = BTRFS_I(dir)->index_cnt;
@@ -4060,7 +5570,7 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
 				     struct inode *dir,
 				     const char *name, int name_len,
 				     u64 ref_objectid, u64 objectid,
-				     u64 alloc_hint, int mode, u64 *index)
+				     umode_t mode, u64 *index)
 {
 	struct inode *inode;
 	struct btrfs_inode_item *inode_item;
@@ -4069,81 +5579,99 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
 	struct btrfs_inode_ref *ref;
 	struct btrfs_key key[2];
 	u32 sizes[2];
+	int nitems = name ? 2 : 1;
 	unsigned long ptr;
 	int ret;
-	int owner;
 
 	path = btrfs_alloc_path();
-	BUG_ON(!path);
+	if (!path)
+		return ERR_PTR(-ENOMEM);
 
 	inode = new_inode(root->fs_info->sb);
-	if (!inode)
+	if (!inode) {
+		btrfs_free_path(path);
 		return ERR_PTR(-ENOMEM);
+	}
+
+	/*
+	 * we have to initialize this early, so we can reclaim the inode
+	 * number if we fail afterwards in this function.
+	 */
+	inode->i_ino = objectid;
+
+	if (dir && name) {
+		trace_btrfs_inode_request(dir);
 
-	if (dir) {
 		ret = btrfs_set_inode_index(dir, index);
 		if (ret) {
+			btrfs_free_path(path);
 			iput(inode);
 			return ERR_PTR(ret);
 		}
+	} else if (dir) {
+		*index = 0;
 	}
 	/*
 	 * index_cnt is ignored for everything but a dir,
 	 * btrfs_get_inode_index_count has an explanation for the magic
 	 * number
 	 */
-	init_btrfs_i(inode);
 	BTRFS_I(inode)->index_cnt = 2;
+	BTRFS_I(inode)->dir_index = *index;
 	BTRFS_I(inode)->root = root;
 	BTRFS_I(inode)->generation = trans->transid;
-	btrfs_set_inode_space_info(root, inode);
+	inode->i_generation = BTRFS_I(inode)->generation;
 
-	if (mode & S_IFDIR)
-		owner = 0;
-	else
-		owner = 1;
-	BTRFS_I(inode)->block_group =
-			btrfs_find_block_group(root, 0, alloc_hint, owner);
+	/*
+	 * We could have gotten an inode number from somebody who was fsynced
+	 * and then removed in this same transaction, so let's just set full
+	 * sync since it will be a full sync anyway and this will blow away the
+	 * old info in the log.
+	 */
+	set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &BTRFS_I(inode)->runtime_flags);
 
 	key[0].objectid = objectid;
 	btrfs_set_key_type(&key[0], BTRFS_INODE_ITEM_KEY);
 	key[0].offset = 0;
 
-	key[1].objectid = objectid;
-	btrfs_set_key_type(&key[1], BTRFS_INODE_REF_KEY);
-	key[1].offset = ref_objectid;
-
 	sizes[0] = sizeof(struct btrfs_inode_item);
-	sizes[1] = name_len + sizeof(*ref);
+
+	if (name) {
+		/*
+		 * Start new inodes with an inode_ref. This is slightly more
+		 * efficient for small numbers of hard links since they will
+		 * be packed into one item. Extended refs will kick in if we
+		 * add more hard links than can fit in the ref item.
+		 */
+		key[1].objectid = objectid;
+		btrfs_set_key_type(&key[1], BTRFS_INODE_REF_KEY);
+		key[1].offset = ref_objectid;
+
+		sizes[1] = name_len + sizeof(*ref);
+	}
 
 	path->leave_spinning = 1;
-	ret = btrfs_insert_empty_items(trans, root, path, key, sizes, 2);
+	ret = btrfs_insert_empty_items(trans, root, path, key, sizes, nitems);
 	if (ret != 0)
 		goto fail;
 
-	inode->i_uid = current_fsuid();
-
-	if (dir && (dir->i_mode & S_ISGID)) {
-		inode->i_gid = dir->i_gid;
-		if (S_ISDIR(mode))
-			mode |= S_ISGID;
-	} else
-		inode->i_gid = current_fsgid();
-
-	inode->i_mode = mode;
-	inode->i_ino = objectid;
+	inode_init_owner(inode, dir, mode);
 	inode_set_bytes(inode, 0);
 	inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
 	inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
 				  struct btrfs_inode_item);
+	memset_extent_buffer(path->nodes[0], 0, (unsigned long)inode_item,
+			     sizeof(*inode_item));
 	fill_inode_item(trans, path->nodes[0], inode_item, inode);
 
-	ref = btrfs_item_ptr(path->nodes[0], path->slots[0] + 1,
-			     struct btrfs_inode_ref);
-	btrfs_set_inode_ref_name_len(path->nodes[0], ref, name_len);
-	btrfs_set_inode_ref_index(path->nodes[0], ref, *index);
-	ptr = (unsigned long)(ref + 1);
-	write_extent_buffer(path->nodes[0], name, ptr, name_len);
+	if (name) {
+		ref = btrfs_item_ptr(path->nodes[0], path->slots[0] + 1,
+				     struct btrfs_inode_ref);
+		btrfs_set_inode_ref_name_len(path->nodes[0], ref, name_len);
+		btrfs_set_inode_ref_index(path->nodes[0], ref, *index);
+		ptr = (unsigned long)(ref + 1);
+		write_extent_buffer(path->nodes[0], name, ptr, name_len);
+	}
 
 	btrfs_mark_buffer_dirty(path->nodes[0]);
 	btrfs_free_path(path);
@@ -4155,18 +5683,31 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
 
 	btrfs_inherit_iflags(inode, dir);
 
-	if ((mode & S_IFREG)) {
+	if (S_ISREG(mode)) {
 		if (btrfs_test_opt(root, NODATASUM))
 			BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
 		if (btrfs_test_opt(root, NODATACOW))
-			BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
+			BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW |
+				BTRFS_INODE_NODATASUM;
 	}
 
-	insert_inode_hash(inode);
+	btrfs_insert_inode_hash(inode);
 	inode_tree_add(inode);
+
+	trace_btrfs_inode_new(inode);
+	btrfs_set_inode_last_trans(trans, inode);
+
+	btrfs_update_root_times(trans, root);
+
+	ret = btrfs_inode_inherit_props(trans, inode, dir);
+	if (ret)
+		btrfs_err(root->fs_info,
+			  "error inheriting props for ino %llu (root %llu): %d",
+			  btrfs_ino(inode), root->root_key.objectid, ret);
+
 	return inode;
 fail:
-	if (dir)
+	if (dir && name)
 		BTRFS_I(dir)->index_cnt--;
 	btrfs_free_path(path);
 	iput(inode);
@@ -4191,58 +5732,81 @@ int btrfs_add_link(struct btrfs_trans_handle *trans,
 	int ret = 0;
 	struct btrfs_key key;
 	struct btrfs_root *root = BTRFS_I(parent_inode)->root;
+	u64 ino = btrfs_ino(inode);
+	u64 parent_ino = btrfs_ino(parent_inode);
 
-	if (unlikely(inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)) {
+	if (unlikely(ino == BTRFS_FIRST_FREE_OBJECTID)) {
 		memcpy(&key, &BTRFS_I(inode)->root->root_key, sizeof(key));
 	} else {
-		key.objectid = inode->i_ino;
+		key.objectid = ino;
 		btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
 		key.offset = 0;
 	}
 
-	if (unlikely(inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)) {
+	if (unlikely(ino == BTRFS_FIRST_FREE_OBJECTID)) {
 		ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
 					 key.objectid, root->root_key.objectid,
-					 parent_inode->i_ino,
-					 index, name, name_len);
+					 parent_ino, index, name, name_len);
 	} else if (add_backref) {
-		ret = btrfs_insert_inode_ref(trans, root,
-					     name, name_len, inode->i_ino,
-					     parent_inode->i_ino, index);
+		ret = btrfs_insert_inode_ref(trans, root, name, name_len, ino,
+					     parent_ino, index);
 	}
 
-	if (ret == 0) {
-		ret = btrfs_insert_dir_item(trans, root, name, name_len,
-					    parent_inode->i_ino, &key,
-					    btrfs_inode_type(inode), index);
-		BUG_ON(ret);
+	/* Nothing to clean up yet */
+	if (ret)
+		return ret;
 
-		btrfs_i_size_write(parent_inode, parent_inode->i_size +
-				   name_len * 2);
-		parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME;
-		ret = btrfs_update_inode(trans, root, parent_inode);
+	ret = btrfs_insert_dir_item(trans, root, name, name_len,
+				    parent_inode, &key,
+				    btrfs_inode_type(inode), index);
+	if (ret == -EEXIST || ret == -EOVERFLOW)
+		goto fail_dir_item;
+	else if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
+		return ret;
+	}
+
+	btrfs_i_size_write(parent_inode, parent_inode->i_size +
+			   name_len * 2);
+	inode_inc_iversion(parent_inode);
+	parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME;
+	ret = btrfs_update_inode(trans, root, parent_inode);
+	if (ret)
+		btrfs_abort_transaction(trans, root, ret);
+	return ret;
+
+fail_dir_item:
+	if (unlikely(ino == BTRFS_FIRST_FREE_OBJECTID)) {
+		u64 local_index;
+		int err;
+		err = btrfs_del_root_ref(trans, root->fs_info->tree_root,
+				 key.objectid, root->root_key.objectid,
+				 parent_ino, &local_index, name, name_len);
+
+	} else if (add_backref) {
+		u64 local_index;
+		int err;
+
+		err = btrfs_del_inode_ref(trans, root, name, name_len,
+					  ino, parent_ino, &local_index);
 	}
 	return ret;
 }
 
 static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
-			    struct dentry *dentry, struct inode *inode,
-			    int backref, u64 index)
+			    struct inode *dir, struct dentry *dentry,
+			    struct inode *inode, int backref, u64 index)
 {
-	int err = btrfs_add_link(trans, dentry->d_parent->d_inode,
-				 inode, dentry->d_name.name,
-				 dentry->d_name.len, backref, index);
-	if (!err) {
-		d_instantiate(dentry, inode);
-		return 0;
-	}
+	int err = btrfs_add_link(trans, dir, inode,
+				 dentry->d_name.name, dentry->d_name.len,
+				 backref, index);
 	if (err > 0)
 		err = -EEXIST;
 	return err;
 }
 
 static int btrfs_mknod(struct inode *dir, struct dentry *dentry,
-			int mode, dev_t rdev)
+			umode_t mode, dev_t rdev)
 {
 	struct btrfs_trans_handle *trans;
 	struct btrfs_root *root = BTRFS_I(dir)->root;
@@ -4250,7 +5814,6 @@ static int btrfs_mknod(struct inode *dir, struct dentry *dentry,
 	int err;
 	int drop_inode = 0;
 	u64 objectid;
-	unsigned long nr = 0;
 	u64 index = 0;
 
 	if (!new_valid_dev(rdev))
@@ -4261,68 +5824,63 @@ static int btrfs_mknod(struct inode *dir, struct dentry *dentry,
 	 * 2 for dir items
 	 * 1 for xattr if selinux is on
 	 */
-	err = btrfs_reserve_metadata_space(root, 5);
-	if (err)
-		return err;
-
-	trans = btrfs_start_transaction(root, 1);
-	if (!trans)
-		goto fail;
-	btrfs_set_trans_block_group(trans, dir);
+	trans = btrfs_start_transaction(root, 5);
+	if (IS_ERR(trans))
+		return PTR_ERR(trans);
 
-	err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
-	if (err) {
-		err = -ENOSPC;
+	err = btrfs_find_free_ino(root, &objectid);
+	if (err)
 		goto out_unlock;
-	}
 
 	inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
-				dentry->d_name.len,
-				dentry->d_parent->d_inode->i_ino, objectid,
-				BTRFS_I(dir)->block_group, mode, &index);
-	err = PTR_ERR(inode);
-	if (IS_ERR(inode))
+				dentry->d_name.len, btrfs_ino(dir), objectid,
+				mode, &index);
+	if (IS_ERR(inode)) {
+		err = PTR_ERR(inode);
 		goto out_unlock;
+	}
 
-	err = btrfs_init_inode_security(trans, inode, dir);
+	err = btrfs_init_inode_security(trans, inode, dir, &dentry->d_name);
 	if (err) {
 		drop_inode = 1;
 		goto out_unlock;
 	}
 
-	btrfs_set_trans_block_group(trans, inode);
-	err = btrfs_add_nondir(trans, dentry, inode, 0, index);
+	/*
+	* If the active LSM wants to access the inode during
+	* d_instantiate it needs these. Smack checks to see
+	* if the filesystem supports xattrs by looking at the
+	* ops vector.
+	*/
+
+	inode->i_op = &btrfs_special_inode_operations;
+	err = btrfs_add_nondir(trans, dir, dentry, inode, 0, index);
 	if (err)
 		drop_inode = 1;
 	else {
-		inode->i_op = &btrfs_special_inode_operations;
 		init_special_inode(inode, inode->i_mode, rdev);
 		btrfs_update_inode(trans, root, inode);
+		d_instantiate(dentry, inode);
 	}
-	btrfs_update_inode_block_group(trans, inode);
-	btrfs_update_inode_block_group(trans, dir);
 out_unlock:
-	nr = trans->blocks_used;
-	btrfs_end_transaction_throttle(trans, root);
-fail:
-	btrfs_unreserve_metadata_space(root, 5);
+	btrfs_end_transaction(trans, root);
+	btrfs_balance_delayed_items(root);
+	btrfs_btree_balance_dirty(root);
 	if (drop_inode) {
 		inode_dec_link_count(inode);
 		iput(inode);
 	}
-	btrfs_btree_balance_dirty(root, nr);
 	return err;
 }
 
 static int btrfs_create(struct inode *dir, struct dentry *dentry,
-			int mode, struct nameidata *nd)
+			umode_t mode, bool excl)
 {
 	struct btrfs_trans_handle *trans;
 	struct btrfs_root *root = BTRFS_I(dir)->root;
 	struct inode *inode = NULL;
+	int drop_inode_on_err = 0;
 	int err;
-	int drop_inode = 0;
-	unsigned long nr = 0;
 	u64 objectid;
 	u64 index = 0;
 
@@ -4331,59 +5889,57 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry,
 	 * 2 for dir items
 	 * 1 for xattr if selinux is on
 	 */
-	err = btrfs_reserve_metadata_space(root, 5);
-	if (err)
-		return err;
+	trans = btrfs_start_transaction(root, 5);
+	if (IS_ERR(trans))
+		return PTR_ERR(trans);
 
-	trans = btrfs_start_transaction(root, 1);
-	if (!trans)
-		goto fail;
-	btrfs_set_trans_block_group(trans, dir);
+	err = btrfs_find_free_ino(root, &objectid);
+	if (err)
+		goto out_unlock;
 
-	err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
-	if (err) {
-		err = -ENOSPC;
+	inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
+				dentry->d_name.len, btrfs_ino(dir), objectid,
+				mode, &index);
+	if (IS_ERR(inode)) {
+		err = PTR_ERR(inode);
 		goto out_unlock;
 	}
+	drop_inode_on_err = 1;
 
-	inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
-				dentry->d_name.len,
-				dentry->d_parent->d_inode->i_ino,
-				objectid, BTRFS_I(dir)->block_group, mode,
-				&index);
-	err = PTR_ERR(inode);
-	if (IS_ERR(inode))
+	err = btrfs_init_inode_security(trans, inode, dir, &dentry->d_name);
+	if (err)
 		goto out_unlock;
 
-	err = btrfs_init_inode_security(trans, inode, dir);
-	if (err) {
-		drop_inode = 1;
+	err = btrfs_update_inode(trans, root, inode);
+	if (err)
 		goto out_unlock;
-	}
 
-	btrfs_set_trans_block_group(trans, inode);
-	err = btrfs_add_nondir(trans, dentry, inode, 0, index);
+	/*
+	* If the active LSM wants to access the inode during
+	* d_instantiate it needs these. Smack checks to see
+	* if the filesystem supports xattrs by looking at the
+	* ops vector.
+	*/
+	inode->i_fop = &btrfs_file_operations;
+	inode->i_op = &btrfs_file_inode_operations;
+
+	err = btrfs_add_nondir(trans, dir, dentry, inode, 0, index);
 	if (err)
-		drop_inode = 1;
-	else {
-		inode->i_mapping->a_ops = &btrfs_aops;
-		inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
-		inode->i_fop = &btrfs_file_operations;
-		inode->i_op = &btrfs_file_inode_operations;
-		BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
-	}
-	btrfs_update_inode_block_group(trans, inode);
-	btrfs_update_inode_block_group(trans, dir);
+		goto out_unlock;
+
+	inode->i_mapping->a_ops = &btrfs_aops;
+	inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
+	BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
+	d_instantiate(dentry, inode);
+
 out_unlock:
-	nr = trans->blocks_used;
-	btrfs_end_transaction_throttle(trans, root);
-fail:
-	btrfs_unreserve_metadata_space(root, 5);
-	if (drop_inode) {
+	btrfs_end_transaction(trans, root);
+	if (err && drop_inode_on_err) {
 		inode_dec_link_count(inode);
 		iput(inode);
 	}
-	btrfs_btree_balance_dirty(root, nr);
+	btrfs_balance_delayed_items(root);
+	btrfs_btree_balance_dirty(root);
 	return err;
 }
 
@@ -4394,60 +5950,73 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
 	struct btrfs_root *root = BTRFS_I(dir)->root;
 	struct inode *inode = old_dentry->d_inode;
 	u64 index;
-	unsigned long nr = 0;
 	int err;
 	int drop_inode = 0;
 
-	if (inode->i_nlink == 0)
-		return -ENOENT;
-
 	/* do not allow sys_link's with other subvols of the same device */
 	if (root->objectid != BTRFS_I(inode)->root->objectid)
-		return -EPERM;
-
-	/*
-	 * 1 item for inode ref
-	 * 2 items for dir items
-	 */
-	err = btrfs_reserve_metadata_space(root, 3);
-	if (err)
-		return err;
+		return -EXDEV;
 
-	btrfs_inc_nlink(inode);
+	if (inode->i_nlink >= BTRFS_LINK_MAX)
+		return -EMLINK;
 
 	err = btrfs_set_inode_index(dir, &index);
 	if (err)
 		goto fail;
 
-	trans = btrfs_start_transaction(root, 1);
+	/*
+	 * 2 items for inode and inode ref
+	 * 2 items for dir items
+	 * 1 item for parent inode
+	 */
+	trans = btrfs_start_transaction(root, 5);
+	if (IS_ERR(trans)) {
+		err = PTR_ERR(trans);
+		goto fail;
+	}
 
-	btrfs_set_trans_block_group(trans, dir);
-	atomic_inc(&inode->i_count);
+	/* There are several dir indexes for this inode, clear the cache. */
+	BTRFS_I(inode)->dir_index = 0ULL;
+	inc_nlink(inode);
+	inode_inc_iversion(inode);
+	inode->i_ctime = CURRENT_TIME;
+	ihold(inode);
+	set_bit(BTRFS_INODE_COPY_EVERYTHING, &BTRFS_I(inode)->runtime_flags);
 
-	err = btrfs_add_nondir(trans, dentry, inode, 1, index);
+	err = btrfs_add_nondir(trans, dir, dentry, inode, 1, index);
 
 	if (err) {
 		drop_inode = 1;
 	} else {
-		btrfs_update_inode_block_group(trans, dir);
+		struct dentry *parent = dentry->d_parent;
 		err = btrfs_update_inode(trans, root, inode);
-		BUG_ON(err);
-		btrfs_log_new_name(trans, inode, NULL, dentry->d_parent);
+		if (err)
+			goto fail;
+		if (inode->i_nlink == 1) {
+			/*
+			 * If new hard link count is 1, it's a file created
+			 * with open(2) O_TMPFILE flag.
+			 */
+			err = btrfs_orphan_del(trans, inode);
+			if (err)
+				goto fail;
+		}
+		d_instantiate(dentry, inode);
+		btrfs_log_new_name(trans, inode, NULL, parent);
 	}
 
-	nr = trans->blocks_used;
-	btrfs_end_transaction_throttle(trans, root);
+	btrfs_end_transaction(trans, root);
+	btrfs_balance_delayed_items(root);
 fail:
-	btrfs_unreserve_metadata_space(root, 3);
 	if (drop_inode) {
 		inode_dec_link_count(inode);
 		iput(inode);
 	}
-	btrfs_btree_balance_dirty(root, nr);
+	btrfs_btree_balance_dirty(root);
 	return err;
 }
 
-static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
+static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
 {
 	struct inode *inode = NULL;
 	struct btrfs_trans_handle *trans;
@@ -4456,35 +6025,23 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
 	int drop_on_err = 0;
 	u64 objectid = 0;
 	u64 index = 0;
-	unsigned long nr = 1;
 
 	/*
 	 * 2 items for inode and ref
 	 * 2 items for dir items
 	 * 1 for xattr if selinux is on
 	 */
-	err = btrfs_reserve_metadata_space(root, 5);
-	if (err)
-		return err;
-
-	trans = btrfs_start_transaction(root, 1);
-	if (!trans) {
-		err = -ENOMEM;
-		goto out_unlock;
-	}
-	btrfs_set_trans_block_group(trans, dir);
+	trans = btrfs_start_transaction(root, 5);
+	if (IS_ERR(trans))
+		return PTR_ERR(trans);
 
-	err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
-	if (err) {
-		err = -ENOSPC;
+	err = btrfs_find_free_ino(root, &objectid);
+	if (err)
 		goto out_fail;
-	}
 
 	inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
-				dentry->d_name.len,
-				dentry->d_parent->d_inode->i_ino, objectid,
-				BTRFS_I(dir)->block_group, S_IFDIR | mode,
-				&index);
+				dentry->d_name.len, btrfs_ino(dir), objectid,
+				S_IFDIR | mode, &index);
 	if (IS_ERR(inode)) {
 		err = PTR_ERR(inode);
 		goto out_fail;
@@ -4492,39 +6049,32 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
 
 	drop_on_err = 1;
 
-	err = btrfs_init_inode_security(trans, inode, dir);
+	err = btrfs_init_inode_security(trans, inode, dir, &dentry->d_name);
 	if (err)
 		goto out_fail;
 
 	inode->i_op = &btrfs_dir_inode_operations;
 	inode->i_fop = &btrfs_dir_file_operations;
-	btrfs_set_trans_block_group(trans, inode);
 
 	btrfs_i_size_write(inode, 0);
 	err = btrfs_update_inode(trans, root, inode);
 	if (err)
 		goto out_fail;
 
-	err = btrfs_add_link(trans, dentry->d_parent->d_inode,
-				 inode, dentry->d_name.name,
-				 dentry->d_name.len, 0, index);
+	err = btrfs_add_link(trans, dir, inode, dentry->d_name.name,
+			     dentry->d_name.len, 0, index);
 	if (err)
 		goto out_fail;
 
 	d_instantiate(dentry, inode);
 	drop_on_err = 0;
-	btrfs_update_inode_block_group(trans, inode);
-	btrfs_update_inode_block_group(trans, dir);
 
 out_fail:
-	nr = trans->blocks_used;
-	btrfs_end_transaction_throttle(trans, root);
-
-out_unlock:
-	btrfs_unreserve_metadata_space(root, 5);
+	btrfs_end_transaction(trans, root);
 	if (drop_on_err)
 		iput(inode);
-	btrfs_btree_balance_dirty(root, nr);
+	btrfs_balance_delayed_items(root);
+	btrfs_btree_balance_dirty(root);
 	return err;
 }
 
@@ -4548,7 +6098,7 @@ static int merge_extent_mapping(struct extent_map_tree *em_tree,
 		em->block_start += start_diff;
 		em->block_len -= start_diff;
 	}
-	return add_extent_mapping(em_tree, em);
+	return add_extent_mapping(em_tree, em, 0);
 }
 
 static noinline int uncompress_inline(struct btrfs_path *path,
@@ -4562,29 +6112,25 @@ static noinline int uncompress_inline(struct btrfs_path *path,
 	size_t max_size;
 	unsigned long inline_size;
 	unsigned long ptr;
+	int compress_type;
 
 	WARN_ON(pg_offset != 0);
+	compress_type = btrfs_file_extent_compression(leaf, item);
 	max_size = btrfs_file_extent_ram_bytes(leaf, item);
 	inline_size = btrfs_file_extent_inline_item_len(leaf,
-					btrfs_item_nr(leaf, path->slots[0]));
+					btrfs_item_nr(path->slots[0]));
 	tmp = kmalloc(inline_size, GFP_NOFS);
+	if (!tmp)
+		return -ENOMEM;
 	ptr = btrfs_file_extent_inline_start(item);
 
 	read_extent_buffer(leaf, tmp, ptr, inline_size);
 
 	max_size = min_t(unsigned long, PAGE_CACHE_SIZE, max_size);
-	ret = btrfs_zlib_decompress(tmp, page, extent_offset,
-				    inline_size, max_size);
-	if (ret) {
-		char *kaddr = kmap_atomic(page, KM_USER0);
-		unsigned long copy_size = min_t(u64,
-				  PAGE_CACHE_SIZE - pg_offset,
-				  max_size - extent_offset);
-		memset(kaddr + pg_offset, 0, copy_size);
-		kunmap_atomic(kaddr, KM_USER0);
-	}
+	ret = btrfs_decompress(compress_type, tmp, page,
+			       extent_offset, inline_size, max_size);
 	kfree(tmp);
-	return 0;
+	return ret;
 }
 
 /*
@@ -4602,10 +6148,9 @@ struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
 {
 	int ret;
 	int err = 0;
-	u64 bytenr;
 	u64 extent_start = 0;
 	u64 extent_end = 0;
-	u64 objectid = inode->i_ino;
+	u64 objectid = btrfs_ino(inode);
 	u32 found_type;
 	struct btrfs_path *path = NULL;
 	struct btrfs_root *root = BTRFS_I(inode)->root;
@@ -4616,7 +6161,7 @@ struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
 	struct btrfs_trans_handle *trans = NULL;
-	int compressed;
+	const bool new_inline = !page || create;
 
 again:
 	read_lock(&em_tree->lock);
@@ -4633,7 +6178,7 @@ again:
 		else
 			goto out;
 	}
-	em = alloc_extent_map(GFP_NOFS);
+	em = alloc_extent_map();
 	if (!em) {
 		err = -ENOMEM;
 		goto out;
@@ -4646,7 +6191,15 @@ again:
 
 	if (!path) {
 		path = btrfs_alloc_path();
-		BUG_ON(!path);
+		if (!path) {
+			err = -ENOMEM;
+			goto out;
+		}
+		/*
+		 * Chances are we'll be called again, so go ahead and do
+		 * readahead
+		 */
+		path->reada = 1;
 	}
 
 	ret = btrfs_lookup_file_extent(trans, root, path,
@@ -4670,23 +6223,28 @@ again:
 	found_type = btrfs_key_type(&found_key);
 	if (found_key.objectid != objectid ||
 	    found_type != BTRFS_EXTENT_DATA_KEY) {
-		goto not_found;
+		/*
+		 * If we backup past the first extent we want to move forward
+		 * and see if there is an extent in front of us, otherwise we'll
+		 * say there is a hole for our whole search range which can
+		 * cause problems.
+		 */
+		extent_end = start;
+		goto next;
 	}
 
 	found_type = btrfs_file_extent_type(leaf, item);
 	extent_start = found_key.offset;
-	compressed = btrfs_file_extent_compression(leaf, item);
 	if (found_type == BTRFS_FILE_EXTENT_REG ||
 	    found_type == BTRFS_FILE_EXTENT_PREALLOC) {
 		extent_end = extent_start +
 		       btrfs_file_extent_num_bytes(leaf, item);
 	} else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
 		size_t size;
-		size = btrfs_file_extent_inline_len(leaf, item);
-		extent_end = (extent_start + size + root->sectorsize - 1) &
-			~((u64)root->sectorsize - 1);
+		size = btrfs_file_extent_inline_len(leaf, path->slots[0], item);
+		extent_end = ALIGN(extent_start + size, root->sectorsize);
 	}
-
+next:
 	if (start >= extent_end) {
 		path->slots[0]++;
 		if (path->slots[0] >= btrfs_header_nritems(leaf)) {
@@ -4706,33 +6264,15 @@ again:
 		if (start + len <= found_key.offset)
 			goto not_found;
 		em->start = start;
+		em->orig_start = start;
 		em->len = found_key.offset - start;
 		goto not_found_em;
 	}
 
+	btrfs_extent_item_to_extent_map(inode, path, item, new_inline, em);
+
 	if (found_type == BTRFS_FILE_EXTENT_REG ||
 	    found_type == BTRFS_FILE_EXTENT_PREALLOC) {
-		em->start = extent_start;
-		em->len = extent_end - extent_start;
-		em->orig_start = extent_start -
-				 btrfs_file_extent_offset(leaf, item);
-		bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
-		if (bytenr == 0) {
-			em->block_start = EXTENT_MAP_HOLE;
-			goto insert;
-		}
-		if (compressed) {
-			set_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
-			em->block_start = bytenr;
-			em->block_len = btrfs_file_extent_disk_num_bytes(leaf,
-									 item);
-		} else {
-			bytenr += btrfs_file_extent_offset(leaf, item);
-			em->block_start = bytenr;
-			em->block_len = em->len;
-			if (found_type == BTRFS_FILE_EXTENT_PREALLOC)
-				set_bit(EXTENT_FLAG_PREALLOC, &em->flags);
-		}
 		goto insert;
 	} else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
 		unsigned long ptr;
@@ -4741,31 +6281,28 @@ again:
 		size_t extent_offset;
 		size_t copy_size;
 
-		em->block_start = EXTENT_MAP_INLINE;
-		if (!page || create) {
-			em->start = extent_start;
-			em->len = extent_end - extent_start;
+		if (new_inline)
 			goto out;
-		}
 
-		size = btrfs_file_extent_inline_len(leaf, item);
+		size = btrfs_file_extent_inline_len(leaf, path->slots[0], item);
 		extent_offset = page_offset(page) + pg_offset - extent_start;
 		copy_size = min_t(u64, PAGE_CACHE_SIZE - pg_offset,
 				size - extent_offset);
 		em->start = extent_start + extent_offset;
-		em->len = (copy_size + root->sectorsize - 1) &
-			~((u64)root->sectorsize - 1);
-		em->orig_start = EXTENT_MAP_INLINE;
-		if (compressed)
-			set_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
+		em->len = ALIGN(copy_size, root->sectorsize);
+		em->orig_block_len = em->len;
+		em->orig_start = em->start;
 		ptr = btrfs_file_extent_inline_start(item) + extent_offset;
 		if (create == 0 && !PageUptodate(page)) {
-			if (btrfs_file_extent_compression(leaf, item) ==
-			    BTRFS_COMPRESS_ZLIB) {
+			if (btrfs_file_extent_compression(leaf, item) !=
+			    BTRFS_COMPRESS_NONE) {
 				ret = uncompress_inline(path, inode, page,
 							pg_offset,
 							extent_offset, item);
-				BUG_ON(ret);
+				if (ret) {
+					err = ret;
+					goto out;
+				}
 			} else {
 				map = kmap(page);
 				read_extent_buffer(leaf, map + pg_offset, ptr,
@@ -4779,12 +6316,17 @@ again:
 			}
 			flush_dcache_page(page);
 		} else if (create && PageUptodate(page)) {
+			BUG();
 			if (!trans) {
 				kunmap(page);
 				free_extent_map(em);
 				em = NULL;
-				btrfs_release_path(root, path);
-				trans = btrfs_join_transaction(root, 1);
+
+				btrfs_release_path(path);
+				trans = btrfs_join_transaction(root);
+
+				if (IS_ERR(trans))
+					return ERR_CAST(trans);
 				goto again;
 			}
 			map = kmap(page);
@@ -4794,33 +6336,28 @@ again:
 			btrfs_mark_buffer_dirty(leaf);
 		}
 		set_extent_uptodate(io_tree, em->start,
-				    extent_map_end(em) - 1, GFP_NOFS);
+				    extent_map_end(em) - 1, NULL, GFP_NOFS);
 		goto insert;
-	} else {
-		printk(KERN_ERR "btrfs unknown found_type %d\n", found_type);
-		WARN_ON(1);
 	}
 not_found:
 	em->start = start;
+	em->orig_start = start;
 	em->len = len;
 not_found_em:
 	em->block_start = EXTENT_MAP_HOLE;
 	set_bit(EXTENT_FLAG_VACANCY, &em->flags);
 insert:
-	btrfs_release_path(root, path);
+	btrfs_release_path(path);
 	if (em->start > start || extent_map_end(em) <= start) {
-		printk(KERN_ERR "Btrfs: bad extent! em: [%llu %llu] passed "
-		       "[%llu %llu]\n", (unsigned long long)em->start,
-		       (unsigned long long)em->len,
-		       (unsigned long long)start,
-		       (unsigned long long)len);
+		btrfs_err(root->fs_info, "bad extent! em: [%llu %llu] passed [%llu %llu]",
+			em->start, em->len, start, len);
 		err = -EIO;
 		goto out;
 	}
 
 	err = 0;
 	write_lock(&em_tree->lock);
-	ret = add_extent_mapping(em_tree, em);
+	ret = add_extent_mapping(em_tree, em, 0);
 	/* it is possible that someone inserted the extent into the tree
 	 * while we had the lock dropped.  It is also possible that
 	 * an overlapping map exists in the tree
@@ -4861,6 +6398,9 @@ insert:
 	}
 	write_unlock(&em_tree->lock);
 out:
+
+	trace_btrfs_get_extent(root, em);
+
 	if (path)
 		btrfs_free_path(path);
 	if (trans) {
@@ -4872,27 +6412,1179 @@ out:
 		free_extent_map(em);
 		return ERR_PTR(err);
 	}
+	BUG_ON(!em); /* Error is always set */
+	return em;
+}
+
+struct extent_map *btrfs_get_extent_fiemap(struct inode *inode, struct page *page,
+					   size_t pg_offset, u64 start, u64 len,
+					   int create)
+{
+	struct extent_map *em;
+	struct extent_map *hole_em = NULL;
+	u64 range_start = start;
+	u64 end;
+	u64 found;
+	u64 found_end;
+	int err = 0;
+
+	em = btrfs_get_extent(inode, page, pg_offset, start, len, create);
+	if (IS_ERR(em))
+		return em;
+	if (em) {
+		/*
+		 * if our em maps to
+		 * -  a hole or
+		 * -  a pre-alloc extent,
+		 * there might actually be delalloc bytes behind it.
+		 */
+		if (em->block_start != EXTENT_MAP_HOLE &&
+		    !test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
+			return em;
+		else
+			hole_em = em;
+	}
+
+	/* check to see if we've wrapped (len == -1 or similar) */
+	end = start + len;
+	if (end < start)
+		end = (u64)-1;
+	else
+		end -= 1;
+
+	em = NULL;
+
+	/* ok, we didn't find anything, lets look for delalloc */
+	found = count_range_bits(&BTRFS_I(inode)->io_tree, &range_start,
+				 end, len, EXTENT_DELALLOC, 1);
+	found_end = range_start + found;
+	if (found_end < range_start)
+		found_end = (u64)-1;
+
+	/*
+	 * we didn't find anything useful, return
+	 * the original results from get_extent()
+	 */
+	if (range_start > end || found_end <= start) {
+		em = hole_em;
+		hole_em = NULL;
+		goto out;
+	}
+
+	/* adjust the range_start to make sure it doesn't
+	 * go backwards from the start they passed in
+	 */
+	range_start = max(start, range_start);
+	found = found_end - range_start;
+
+	if (found > 0) {
+		u64 hole_start = start;
+		u64 hole_len = len;
+
+		em = alloc_extent_map();
+		if (!em) {
+			err = -ENOMEM;
+			goto out;
+		}
+		/*
+		 * when btrfs_get_extent can't find anything it
+		 * returns one huge hole
+		 *
+		 * make sure what it found really fits our range, and
+		 * adjust to make sure it is based on the start from
+		 * the caller
+		 */
+		if (hole_em) {
+			u64 calc_end = extent_map_end(hole_em);
+
+			if (calc_end <= start || (hole_em->start > end)) {
+				free_extent_map(hole_em);
+				hole_em = NULL;
+			} else {
+				hole_start = max(hole_em->start, start);
+				hole_len = calc_end - hole_start;
+			}
+		}
+		em->bdev = NULL;
+		if (hole_em && range_start > hole_start) {
+			/* our hole starts before our delalloc, so we
+			 * have to return just the parts of the hole
+			 * that go until  the delalloc starts
+			 */
+			em->len = min(hole_len,
+				      range_start - hole_start);
+			em->start = hole_start;
+			em->orig_start = hole_start;
+			/*
+			 * don't adjust block start at all,
+			 * it is fixed at EXTENT_MAP_HOLE
+			 */
+			em->block_start = hole_em->block_start;
+			em->block_len = hole_len;
+			if (test_bit(EXTENT_FLAG_PREALLOC, &hole_em->flags))
+				set_bit(EXTENT_FLAG_PREALLOC, &em->flags);
+		} else {
+			em->start = range_start;
+			em->len = found;
+			em->orig_start = range_start;
+			em->block_start = EXTENT_MAP_DELALLOC;
+			em->block_len = found;
+		}
+	} else if (hole_em) {
+		return hole_em;
+	}
+out:
+
+	free_extent_map(hole_em);
+	if (err) {
+		free_extent_map(em);
+		return ERR_PTR(err);
+	}
+	return em;
+}
+
+static struct extent_map *btrfs_new_extent_direct(struct inode *inode,
+						  u64 start, u64 len)
+{
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct extent_map *em;
+	struct btrfs_key ins;
+	u64 alloc_hint;
+	int ret;
+
+	alloc_hint = get_extent_allocation_hint(inode, start, len);
+	ret = btrfs_reserve_extent(root, len, root->sectorsize, 0,
+				   alloc_hint, &ins, 1, 1);
+	if (ret)
+		return ERR_PTR(ret);
+
+	em = create_pinned_em(inode, start, ins.offset, start, ins.objectid,
+			      ins.offset, ins.offset, ins.offset, 0);
+	if (IS_ERR(em)) {
+		btrfs_free_reserved_extent(root, ins.objectid, ins.offset, 1);
+		return em;
+	}
+
+	ret = btrfs_add_ordered_extent_dio(inode, start, ins.objectid,
+					   ins.offset, ins.offset, 0);
+	if (ret) {
+		btrfs_free_reserved_extent(root, ins.objectid, ins.offset, 1);
+		free_extent_map(em);
+		return ERR_PTR(ret);
+	}
+
 	return em;
 }
 
+/*
+ * returns 1 when the nocow is safe, < 1 on error, 0 if the
+ * block must be cow'd
+ */
+noinline int can_nocow_extent(struct inode *inode, u64 offset, u64 *len,
+			      u64 *orig_start, u64 *orig_block_len,
+			      u64 *ram_bytes)
+{
+	struct btrfs_trans_handle *trans;
+	struct btrfs_path *path;
+	int ret;
+	struct extent_buffer *leaf;
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
+	struct btrfs_file_extent_item *fi;
+	struct btrfs_key key;
+	u64 disk_bytenr;
+	u64 backref_offset;
+	u64 extent_end;
+	u64 num_bytes;
+	int slot;
+	int found_type;
+	bool nocow = (BTRFS_I(inode)->flags & BTRFS_INODE_NODATACOW);
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	ret = btrfs_lookup_file_extent(NULL, root, path, btrfs_ino(inode),
+				       offset, 0);
+	if (ret < 0)
+		goto out;
+
+	slot = path->slots[0];
+	if (ret == 1) {
+		if (slot == 0) {
+			/* can't find the item, must cow */
+			ret = 0;
+			goto out;
+		}
+		slot--;
+	}
+	ret = 0;
+	leaf = path->nodes[0];
+	btrfs_item_key_to_cpu(leaf, &key, slot);
+	if (key.objectid != btrfs_ino(inode) ||
+	    key.type != BTRFS_EXTENT_DATA_KEY) {
+		/* not our file or wrong item type, must cow */
+		goto out;
+	}
+
+	if (key.offset > offset) {
+		/* Wrong offset, must cow */
+		goto out;
+	}
+
+	fi = btrfs_item_ptr(leaf, slot, struct btrfs_file_extent_item);
+	found_type = btrfs_file_extent_type(leaf, fi);
+	if (found_type != BTRFS_FILE_EXTENT_REG &&
+	    found_type != BTRFS_FILE_EXTENT_PREALLOC) {
+		/* not a regular extent, must cow */
+		goto out;
+	}
+
+	if (!nocow && found_type == BTRFS_FILE_EXTENT_REG)
+		goto out;
+
+	extent_end = key.offset + btrfs_file_extent_num_bytes(leaf, fi);
+	if (extent_end <= offset)
+		goto out;
+
+	disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
+	if (disk_bytenr == 0)
+		goto out;
+
+	if (btrfs_file_extent_compression(leaf, fi) ||
+	    btrfs_file_extent_encryption(leaf, fi) ||
+	    btrfs_file_extent_other_encoding(leaf, fi))
+		goto out;
+
+	backref_offset = btrfs_file_extent_offset(leaf, fi);
+
+	if (orig_start) {
+		*orig_start = key.offset - backref_offset;
+		*orig_block_len = btrfs_file_extent_disk_num_bytes(leaf, fi);
+		*ram_bytes = btrfs_file_extent_ram_bytes(leaf, fi);
+	}
+
+	if (btrfs_extent_readonly(root, disk_bytenr))
+		goto out;
+
+	num_bytes = min(offset + *len, extent_end) - offset;
+	if (!nocow && found_type == BTRFS_FILE_EXTENT_PREALLOC) {
+		u64 range_end;
+
+		range_end = round_up(offset + num_bytes, root->sectorsize) - 1;
+		ret = test_range_bit(io_tree, offset, range_end,
+				     EXTENT_DELALLOC, 0, NULL);
+		if (ret) {
+			ret = -EAGAIN;
+			goto out;
+		}
+	}
+
+	btrfs_release_path(path);
+
+	/*
+	 * look for other files referencing this extent, if we
+	 * find any we must cow
+	 */
+	trans = btrfs_join_transaction(root);
+	if (IS_ERR(trans)) {
+		ret = 0;
+		goto out;
+	}
+
+	ret = btrfs_cross_ref_exist(trans, root, btrfs_ino(inode),
+				    key.offset - backref_offset, disk_bytenr);
+	btrfs_end_transaction(trans, root);
+	if (ret) {
+		ret = 0;
+		goto out;
+	}
+
+	/*
+	 * adjust disk_bytenr and num_bytes to cover just the bytes
+	 * in this extent we are about to write.  If there
+	 * are any csums in that range we have to cow in order
+	 * to keep the csums correct
+	 */
+	disk_bytenr += backref_offset;
+	disk_bytenr += offset - key.offset;
+	if (csum_exist_in_range(root, disk_bytenr, num_bytes))
+				goto out;
+	/*
+	 * all of the above have passed, it is safe to overwrite this extent
+	 * without cow
+	 */
+	*len = num_bytes;
+	ret = 1;
+out:
+	btrfs_free_path(path);
+	return ret;
+}
+
+bool btrfs_page_exists_in_range(struct inode *inode, loff_t start, loff_t end)
+{
+	struct radix_tree_root *root = &inode->i_mapping->page_tree;
+	int found = false;
+	void **pagep = NULL;
+	struct page *page = NULL;
+	int start_idx;
+	int end_idx;
+
+	start_idx = start >> PAGE_CACHE_SHIFT;
+
+	/*
+	 * end is the last byte in the last page.  end == start is legal
+	 */
+	end_idx = end >> PAGE_CACHE_SHIFT;
+
+	rcu_read_lock();
+
+	/* Most of the code in this while loop is lifted from
+	 * find_get_page.  It's been modified to begin searching from a
+	 * page and return just the first page found in that range.  If the
+	 * found idx is less than or equal to the end idx then we know that
+	 * a page exists.  If no pages are found or if those pages are
+	 * outside of the range then we're fine (yay!) */
+	while (page == NULL &&
+	       radix_tree_gang_lookup_slot(root, &pagep, NULL, start_idx, 1)) {
+		page = radix_tree_deref_slot(pagep);
+		if (unlikely(!page))
+			break;
+
+		if (radix_tree_exception(page)) {
+			if (radix_tree_deref_retry(page)) {
+				page = NULL;
+				continue;
+			}
+			/*
+			 * Otherwise, shmem/tmpfs must be storing a swap entry
+			 * here as an exceptional entry: so return it without
+			 * attempting to raise page count.
+			 */
+			page = NULL;
+			break; /* TODO: Is this relevant for this use case? */
+		}
+
+		if (!page_cache_get_speculative(page)) {
+			page = NULL;
+			continue;
+		}
+
+		/*
+		 * Has the page moved?
+		 * This is part of the lockless pagecache protocol. See
+		 * include/linux/pagemap.h for details.
+		 */
+		if (unlikely(page != *pagep)) {
+			page_cache_release(page);
+			page = NULL;
+		}
+	}
+
+	if (page) {
+		if (page->index <= end_idx)
+			found = true;
+		page_cache_release(page);
+	}
+
+	rcu_read_unlock();
+	return found;
+}
+
+static int lock_extent_direct(struct inode *inode, u64 lockstart, u64 lockend,
+			      struct extent_state **cached_state, int writing)
+{
+	struct btrfs_ordered_extent *ordered;
+	int ret = 0;
+
+	while (1) {
+		lock_extent_bits(&BTRFS_I(inode)->io_tree, lockstart, lockend,
+				 0, cached_state);
+		/*
+		 * We're concerned with the entire range that we're going to be
+		 * doing DIO to, so we need to make sure theres no ordered
+		 * extents in this range.
+		 */
+		ordered = btrfs_lookup_ordered_range(inode, lockstart,
+						     lockend - lockstart + 1);
+
+		/*
+		 * We need to make sure there are no buffered pages in this
+		 * range either, we could have raced between the invalidate in
+		 * generic_file_direct_write and locking the extent.  The
+		 * invalidate needs to happen so that reads after a write do not
+		 * get stale data.
+		 */
+		if (!ordered &&
+		    (!writing ||
+		     !btrfs_page_exists_in_range(inode, lockstart, lockend)))
+			break;
+
+		unlock_extent_cached(&BTRFS_I(inode)->io_tree, lockstart, lockend,
+				     cached_state, GFP_NOFS);
+
+		if (ordered) {
+			btrfs_start_ordered_extent(inode, ordered, 1);
+			btrfs_put_ordered_extent(ordered);
+		} else {
+			/* Screw you mmap */
+			ret = filemap_write_and_wait_range(inode->i_mapping,
+							   lockstart,
+							   lockend);
+			if (ret)
+				break;
+
+			/*
+			 * If we found a page that couldn't be invalidated just
+			 * fall back to buffered.
+			 */
+			ret = invalidate_inode_pages2_range(inode->i_mapping,
+					lockstart >> PAGE_CACHE_SHIFT,
+					lockend >> PAGE_CACHE_SHIFT);
+			if (ret)
+				break;
+		}
+
+		cond_resched();
+	}
+
+	return ret;
+}
+
+static struct extent_map *create_pinned_em(struct inode *inode, u64 start,
+					   u64 len, u64 orig_start,
+					   u64 block_start, u64 block_len,
+					   u64 orig_block_len, u64 ram_bytes,
+					   int type)
+{
+	struct extent_map_tree *em_tree;
+	struct extent_map *em;
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	int ret;
+
+	em_tree = &BTRFS_I(inode)->extent_tree;
+	em = alloc_extent_map();
+	if (!em)
+		return ERR_PTR(-ENOMEM);
+
+	em->start = start;
+	em->orig_start = orig_start;
+	em->mod_start = start;
+	em->mod_len = len;
+	em->len = len;
+	em->block_len = block_len;
+	em->block_start = block_start;
+	em->bdev = root->fs_info->fs_devices->latest_bdev;
+	em->orig_block_len = orig_block_len;
+	em->ram_bytes = ram_bytes;
+	em->generation = -1;
+	set_bit(EXTENT_FLAG_PINNED, &em->flags);
+	if (type == BTRFS_ORDERED_PREALLOC)
+		set_bit(EXTENT_FLAG_FILLING, &em->flags);
+
+	do {
+		btrfs_drop_extent_cache(inode, em->start,
+				em->start + em->len - 1, 0);
+		write_lock(&em_tree->lock);
+		ret = add_extent_mapping(em_tree, em, 1);
+		write_unlock(&em_tree->lock);
+	} while (ret == -EEXIST);
+
+	if (ret) {
+		free_extent_map(em);
+		return ERR_PTR(ret);
+	}
+
+	return em;
+}
+
+
+static int btrfs_get_blocks_direct(struct inode *inode, sector_t iblock,
+				   struct buffer_head *bh_result, int create)
+{
+	struct extent_map *em;
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct extent_state *cached_state = NULL;
+	u64 start = iblock << inode->i_blkbits;
+	u64 lockstart, lockend;
+	u64 len = bh_result->b_size;
+	int unlock_bits = EXTENT_LOCKED;
+	int ret = 0;
+
+	if (create)
+		unlock_bits |= EXTENT_DELALLOC | EXTENT_DIRTY;
+	else
+		len = min_t(u64, len, root->sectorsize);
+
+	lockstart = start;
+	lockend = start + len - 1;
+
+	/*
+	 * If this errors out it's because we couldn't invalidate pagecache for
+	 * this range and we need to fallback to buffered.
+	 */
+	if (lock_extent_direct(inode, lockstart, lockend, &cached_state, create))
+		return -ENOTBLK;
+
+	em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
+	if (IS_ERR(em)) {
+		ret = PTR_ERR(em);
+		goto unlock_err;
+	}
+
+	/*
+	 * Ok for INLINE and COMPRESSED extents we need to fallback on buffered
+	 * io.  INLINE is special, and we could probably kludge it in here, but
+	 * it's still buffered so for safety lets just fall back to the generic
+	 * buffered path.
+	 *
+	 * For COMPRESSED we _have_ to read the entire extent in so we can
+	 * decompress it, so there will be buffering required no matter what we
+	 * do, so go ahead and fallback to buffered.
+	 *
+	 * We return -ENOTBLK because thats what makes DIO go ahead and go back
+	 * to buffered IO.  Don't blame me, this is the price we pay for using
+	 * the generic code.
+	 */
+	if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags) ||
+	    em->block_start == EXTENT_MAP_INLINE) {
+		free_extent_map(em);
+		ret = -ENOTBLK;
+		goto unlock_err;
+	}
+
+	/* Just a good old fashioned hole, return */
+	if (!create && (em->block_start == EXTENT_MAP_HOLE ||
+			test_bit(EXTENT_FLAG_PREALLOC, &em->flags))) {
+		free_extent_map(em);
+		goto unlock_err;
+	}
+
+	/*
+	 * We don't allocate a new extent in the following cases
+	 *
+	 * 1) The inode is marked as NODATACOW.  In this case we'll just use the
+	 * existing extent.
+	 * 2) The extent is marked as PREALLOC.  We're good to go here and can
+	 * just use the extent.
+	 *
+	 */
+	if (!create) {
+		len = min(len, em->len - (start - em->start));
+		lockstart = start + len;
+		goto unlock;
+	}
+
+	if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags) ||
+	    ((BTRFS_I(inode)->flags & BTRFS_INODE_NODATACOW) &&
+	     em->block_start != EXTENT_MAP_HOLE)) {
+		int type;
+		int ret;
+		u64 block_start, orig_start, orig_block_len, ram_bytes;
+
+		if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
+			type = BTRFS_ORDERED_PREALLOC;
+		else
+			type = BTRFS_ORDERED_NOCOW;
+		len = min(len, em->len - (start - em->start));
+		block_start = em->block_start + (start - em->start);
+
+		if (can_nocow_extent(inode, start, &len, &orig_start,
+				     &orig_block_len, &ram_bytes) == 1) {
+			if (type == BTRFS_ORDERED_PREALLOC) {
+				free_extent_map(em);
+				em = create_pinned_em(inode, start, len,
+						       orig_start,
+						       block_start, len,
+						       orig_block_len,
+						       ram_bytes, type);
+				if (IS_ERR(em))
+					goto unlock_err;
+			}
+
+			ret = btrfs_add_ordered_extent_dio(inode, start,
+					   block_start, len, len, type);
+			if (ret) {
+				free_extent_map(em);
+				goto unlock_err;
+			}
+			goto unlock;
+		}
+	}
+
+	/*
+	 * this will cow the extent, reset the len in case we changed
+	 * it above
+	 */
+	len = bh_result->b_size;
+	free_extent_map(em);
+	em = btrfs_new_extent_direct(inode, start, len);
+	if (IS_ERR(em)) {
+		ret = PTR_ERR(em);
+		goto unlock_err;
+	}
+	len = min(len, em->len - (start - em->start));
+unlock:
+	bh_result->b_blocknr = (em->block_start + (start - em->start)) >>
+		inode->i_blkbits;
+	bh_result->b_size = len;
+	bh_result->b_bdev = em->bdev;
+	set_buffer_mapped(bh_result);
+	if (create) {
+		if (!test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
+			set_buffer_new(bh_result);
+
+		/*
+		 * Need to update the i_size under the extent lock so buffered
+		 * readers will get the updated i_size when we unlock.
+		 */
+		if (start + len > i_size_read(inode))
+			i_size_write(inode, start + len);
+
+		spin_lock(&BTRFS_I(inode)->lock);
+		BTRFS_I(inode)->outstanding_extents++;
+		spin_unlock(&BTRFS_I(inode)->lock);
+
+		ret = set_extent_bit(&BTRFS_I(inode)->io_tree, lockstart,
+				     lockstart + len - 1, EXTENT_DELALLOC, NULL,
+				     &cached_state, GFP_NOFS);
+		BUG_ON(ret);
+	}
+
+	/*
+	 * In the case of write we need to clear and unlock the entire range,
+	 * in the case of read we need to unlock only the end area that we
+	 * aren't using if there is any left over space.
+	 */
+	if (lockstart < lockend) {
+		clear_extent_bit(&BTRFS_I(inode)->io_tree, lockstart,
+				 lockend, unlock_bits, 1, 0,
+				 &cached_state, GFP_NOFS);
+	} else {
+		free_extent_state(cached_state);
+	}
+
+	free_extent_map(em);
+
+	return 0;
+
+unlock_err:
+	clear_extent_bit(&BTRFS_I(inode)->io_tree, lockstart, lockend,
+			 unlock_bits, 1, 0, &cached_state, GFP_NOFS);
+	return ret;
+}
+
+static void btrfs_endio_direct_read(struct bio *bio, int err)
+{
+	struct btrfs_dio_private *dip = bio->bi_private;
+	struct bio_vec *bvec;
+	struct inode *inode = dip->inode;
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct bio *dio_bio;
+	u32 *csums = (u32 *)dip->csum;
+	u64 start;
+	int i;
+
+	start = dip->logical_offset;
+	bio_for_each_segment_all(bvec, bio, i) {
+		if (!(BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)) {
+			struct page *page = bvec->bv_page;
+			char *kaddr;
+			u32 csum = ~(u32)0;
+			unsigned long flags;
+
+			local_irq_save(flags);
+			kaddr = kmap_atomic(page);
+			csum = btrfs_csum_data(kaddr + bvec->bv_offset,
+					       csum, bvec->bv_len);
+			btrfs_csum_final(csum, (char *)&csum);
+			kunmap_atomic(kaddr);
+			local_irq_restore(flags);
+
+			flush_dcache_page(bvec->bv_page);
+			if (csum != csums[i]) {
+				btrfs_err(root->fs_info, "csum failed ino %llu off %llu csum %u expected csum %u",
+					  btrfs_ino(inode), start, csum,
+					  csums[i]);
+				err = -EIO;
+			}
+		}
+
+		start += bvec->bv_len;
+	}
+
+	unlock_extent(&BTRFS_I(inode)->io_tree, dip->logical_offset,
+		      dip->logical_offset + dip->bytes - 1);
+	dio_bio = dip->dio_bio;
+
+	kfree(dip);
+
+	/* If we had a csum failure make sure to clear the uptodate flag */
+	if (err)
+		clear_bit(BIO_UPTODATE, &dio_bio->bi_flags);
+	dio_end_io(dio_bio, err);
+	bio_put(bio);
+}
+
+static void btrfs_endio_direct_write(struct bio *bio, int err)
+{
+	struct btrfs_dio_private *dip = bio->bi_private;
+	struct inode *inode = dip->inode;
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct btrfs_ordered_extent *ordered = NULL;
+	u64 ordered_offset = dip->logical_offset;
+	u64 ordered_bytes = dip->bytes;
+	struct bio *dio_bio;
+	int ret;
+
+	if (err)
+		goto out_done;
+again:
+	ret = btrfs_dec_test_first_ordered_pending(inode, &ordered,
+						   &ordered_offset,
+						   ordered_bytes, !err);
+	if (!ret)
+		goto out_test;
+
+	btrfs_init_work(&ordered->work, finish_ordered_fn, NULL, NULL);
+	btrfs_queue_work(root->fs_info->endio_write_workers,
+			 &ordered->work);
+out_test:
+	/*
+	 * our bio might span multiple ordered extents.  If we haven't
+	 * completed the accounting for the whole dio, go back and try again
+	 */
+	if (ordered_offset < dip->logical_offset + dip->bytes) {
+		ordered_bytes = dip->logical_offset + dip->bytes -
+			ordered_offset;
+		ordered = NULL;
+		goto again;
+	}
+out_done:
+	dio_bio = dip->dio_bio;
+
+	kfree(dip);
+
+	/* If we had an error make sure to clear the uptodate flag */
+	if (err)
+		clear_bit(BIO_UPTODATE, &dio_bio->bi_flags);
+	dio_end_io(dio_bio, err);
+	bio_put(bio);
+}
+
+static int __btrfs_submit_bio_start_direct_io(struct inode *inode, int rw,
+				    struct bio *bio, int mirror_num,
+				    unsigned long bio_flags, u64 offset)
+{
+	int ret;
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	ret = btrfs_csum_one_bio(root, inode, bio, offset, 1);
+	BUG_ON(ret); /* -ENOMEM */
+	return 0;
+}
+
+static void btrfs_end_dio_bio(struct bio *bio, int err)
+{
+	struct btrfs_dio_private *dip = bio->bi_private;
+
+	if (err) {
+		btrfs_err(BTRFS_I(dip->inode)->root->fs_info,
+			  "direct IO failed ino %llu rw %lu sector %#Lx len %u err no %d",
+		      btrfs_ino(dip->inode), bio->bi_rw,
+		      (unsigned long long)bio->bi_iter.bi_sector,
+		      bio->bi_iter.bi_size, err);
+		dip->errors = 1;
+
+		/*
+		 * before atomic variable goto zero, we must make sure
+		 * dip->errors is perceived to be set.
+		 */
+		smp_mb__before_atomic();
+	}
+
+	/* if there are more bios still pending for this dio, just exit */
+	if (!atomic_dec_and_test(&dip->pending_bios))
+		goto out;
+
+	if (dip->errors) {
+		bio_io_error(dip->orig_bio);
+	} else {
+		set_bit(BIO_UPTODATE, &dip->dio_bio->bi_flags);
+		bio_endio(dip->orig_bio, 0);
+	}
+out:
+	bio_put(bio);
+}
+
+static struct bio *btrfs_dio_bio_alloc(struct block_device *bdev,
+				       u64 first_sector, gfp_t gfp_flags)
+{
+	int nr_vecs = bio_get_nr_vecs(bdev);
+	return btrfs_bio_alloc(bdev, first_sector, nr_vecs, gfp_flags);
+}
+
+static inline int __btrfs_submit_dio_bio(struct bio *bio, struct inode *inode,
+					 int rw, u64 file_offset, int skip_sum,
+					 int async_submit)
+{
+	struct btrfs_dio_private *dip = bio->bi_private;
+	int write = rw & REQ_WRITE;
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	int ret;
+
+	if (async_submit)
+		async_submit = !atomic_read(&BTRFS_I(inode)->sync_writers);
+
+	bio_get(bio);
+
+	if (!write) {
+		ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0);
+		if (ret)
+			goto err;
+	}
+
+	if (skip_sum)
+		goto map;
+
+	if (write && async_submit) {
+		ret = btrfs_wq_submit_bio(root->fs_info,
+				   inode, rw, bio, 0, 0,
+				   file_offset,
+				   __btrfs_submit_bio_start_direct_io,
+				   __btrfs_submit_bio_done);
+		goto err;
+	} else if (write) {
+		/*
+		 * If we aren't doing async submit, calculate the csum of the
+		 * bio now.
+		 */
+		ret = btrfs_csum_one_bio(root, inode, bio, file_offset, 1);
+		if (ret)
+			goto err;
+	} else if (!skip_sum) {
+		ret = btrfs_lookup_bio_sums_dio(root, inode, dip, bio,
+						file_offset);
+		if (ret)
+			goto err;
+	}
+
+map:
+	ret = btrfs_map_bio(root, rw, bio, 0, async_submit);
+err:
+	bio_put(bio);
+	return ret;
+}
+
+static int btrfs_submit_direct_hook(int rw, struct btrfs_dio_private *dip,
+				    int skip_sum)
+{
+	struct inode *inode = dip->inode;
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct bio *bio;
+	struct bio *orig_bio = dip->orig_bio;
+	struct bio_vec *bvec = orig_bio->bi_io_vec;
+	u64 start_sector = orig_bio->bi_iter.bi_sector;
+	u64 file_offset = dip->logical_offset;
+	u64 submit_len = 0;
+	u64 map_length;
+	int nr_pages = 0;
+	int ret = 0;
+	int async_submit = 0;
+
+	map_length = orig_bio->bi_iter.bi_size;
+	ret = btrfs_map_block(root->fs_info, rw, start_sector << 9,
+			      &map_length, NULL, 0);
+	if (ret) {
+		bio_put(orig_bio);
+		return -EIO;
+	}
+
+	if (map_length >= orig_bio->bi_iter.bi_size) {
+		bio = orig_bio;
+		goto submit;
+	}
+
+	/* async crcs make it difficult to collect full stripe writes. */
+	if (btrfs_get_alloc_profile(root, 1) &
+	    (BTRFS_BLOCK_GROUP_RAID5 | BTRFS_BLOCK_GROUP_RAID6))
+		async_submit = 0;
+	else
+		async_submit = 1;
+
+	bio = btrfs_dio_bio_alloc(orig_bio->bi_bdev, start_sector, GFP_NOFS);
+	if (!bio)
+		return -ENOMEM;
+	bio->bi_private = dip;
+	bio->bi_end_io = btrfs_end_dio_bio;
+	atomic_inc(&dip->pending_bios);
+
+	while (bvec <= (orig_bio->bi_io_vec + orig_bio->bi_vcnt - 1)) {
+		if (unlikely(map_length < submit_len + bvec->bv_len ||
+		    bio_add_page(bio, bvec->bv_page, bvec->bv_len,
+				 bvec->bv_offset) < bvec->bv_len)) {
+			/*
+			 * inc the count before we submit the bio so
+			 * we know the end IO handler won't happen before
+			 * we inc the count. Otherwise, the dip might get freed
+			 * before we're done setting it up
+			 */
+			atomic_inc(&dip->pending_bios);
+			ret = __btrfs_submit_dio_bio(bio, inode, rw,
+						     file_offset, skip_sum,
+						     async_submit);
+			if (ret) {
+				bio_put(bio);
+				atomic_dec(&dip->pending_bios);
+				goto out_err;
+			}
+
+			start_sector += submit_len >> 9;
+			file_offset += submit_len;
+
+			submit_len = 0;
+			nr_pages = 0;
+
+			bio = btrfs_dio_bio_alloc(orig_bio->bi_bdev,
+						  start_sector, GFP_NOFS);
+			if (!bio)
+				goto out_err;
+			bio->bi_private = dip;
+			bio->bi_end_io = btrfs_end_dio_bio;
+
+			map_length = orig_bio->bi_iter.bi_size;
+			ret = btrfs_map_block(root->fs_info, rw,
+					      start_sector << 9,
+					      &map_length, NULL, 0);
+			if (ret) {
+				bio_put(bio);
+				goto out_err;
+			}
+		} else {
+			submit_len += bvec->bv_len;
+			nr_pages++;
+			bvec++;
+		}
+	}
+
+submit:
+	ret = __btrfs_submit_dio_bio(bio, inode, rw, file_offset, skip_sum,
+				     async_submit);
+	if (!ret)
+		return 0;
+
+	bio_put(bio);
+out_err:
+	dip->errors = 1;
+	/*
+	 * before atomic variable goto zero, we must
+	 * make sure dip->errors is perceived to be set.
+	 */
+	smp_mb__before_atomic();
+	if (atomic_dec_and_test(&dip->pending_bios))
+		bio_io_error(dip->orig_bio);
+
+	/* bio_end_io() will handle error, so we needn't return it */
+	return 0;
+}
+
+static void btrfs_submit_direct(int rw, struct bio *dio_bio,
+				struct inode *inode, loff_t file_offset)
+{
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct btrfs_dio_private *dip;
+	struct bio *io_bio;
+	int skip_sum;
+	int sum_len;
+	int write = rw & REQ_WRITE;
+	int ret = 0;
+	u16 csum_size;
+
+	skip_sum = BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM;
+
+	io_bio = btrfs_bio_clone(dio_bio, GFP_NOFS);
+	if (!io_bio) {
+		ret = -ENOMEM;
+		goto free_ordered;
+	}
+
+	if (!skip_sum && !write) {
+		csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
+		sum_len = dio_bio->bi_iter.bi_size >>
+			inode->i_sb->s_blocksize_bits;
+		sum_len *= csum_size;
+	} else {
+		sum_len = 0;
+	}
+
+	dip = kmalloc(sizeof(*dip) + sum_len, GFP_NOFS);
+	if (!dip) {
+		ret = -ENOMEM;
+		goto free_io_bio;
+	}
+
+	dip->private = dio_bio->bi_private;
+	dip->inode = inode;
+	dip->logical_offset = file_offset;
+	dip->bytes = dio_bio->bi_iter.bi_size;
+	dip->disk_bytenr = (u64)dio_bio->bi_iter.bi_sector << 9;
+	io_bio->bi_private = dip;
+	dip->errors = 0;
+	dip->orig_bio = io_bio;
+	dip->dio_bio = dio_bio;
+	atomic_set(&dip->pending_bios, 0);
+
+	if (write)
+		io_bio->bi_end_io = btrfs_endio_direct_write;
+	else
+		io_bio->bi_end_io = btrfs_endio_direct_read;
+
+	ret = btrfs_submit_direct_hook(rw, dip, skip_sum);
+	if (!ret)
+		return;
+
+free_io_bio:
+	bio_put(io_bio);
+
+free_ordered:
+	/*
+	 * If this is a write, we need to clean up the reserved space and kill
+	 * the ordered extent.
+	 */
+	if (write) {
+		struct btrfs_ordered_extent *ordered;
+		ordered = btrfs_lookup_ordered_extent(inode, file_offset);
+		if (!test_bit(BTRFS_ORDERED_PREALLOC, &ordered->flags) &&
+		    !test_bit(BTRFS_ORDERED_NOCOW, &ordered->flags))
+			btrfs_free_reserved_extent(root, ordered->start,
+						   ordered->disk_len, 1);
+		btrfs_put_ordered_extent(ordered);
+		btrfs_put_ordered_extent(ordered);
+	}
+	bio_endio(dio_bio, ret);
+}
+
+static ssize_t check_direct_IO(struct btrfs_root *root, int rw, struct kiocb *iocb,
+			const struct iov_iter *iter, loff_t offset)
+{
+	int seg;
+	int i;
+	unsigned blocksize_mask = root->sectorsize - 1;
+	ssize_t retval = -EINVAL;
+
+	if (offset & blocksize_mask)
+		goto out;
+
+	if (iov_iter_alignment(iter) & blocksize_mask)
+		goto out;
+
+	/* If this is a write we don't need to check anymore */
+	if (rw & WRITE)
+		return 0;
+	/*
+	 * Check to make sure we don't have duplicate iov_base's in this
+	 * iovec, if so return EINVAL, otherwise we'll get csum errors
+	 * when reading back.
+	 */
+	for (seg = 0; seg < iter->nr_segs; seg++) {
+		for (i = seg + 1; i < iter->nr_segs; i++) {
+			if (iter->iov[seg].iov_base == iter->iov[i].iov_base)
+				goto out;
+		}
+	}
+	retval = 0;
+out:
+	return retval;
+}
+
 static ssize_t btrfs_direct_IO(int rw, struct kiocb *iocb,
-			const struct iovec *iov, loff_t offset,
-			unsigned long nr_segs)
+			struct iov_iter *iter, loff_t offset)
 {
-	return -EINVAL;
+	struct file *file = iocb->ki_filp;
+	struct inode *inode = file->f_mapping->host;
+	size_t count = 0;
+	int flags = 0;
+	bool wakeup = true;
+	bool relock = false;
+	ssize_t ret;
+
+	if (check_direct_IO(BTRFS_I(inode)->root, rw, iocb, iter, offset))
+		return 0;
+
+	atomic_inc(&inode->i_dio_count);
+	smp_mb__after_atomic();
+
+	/*
+	 * The generic stuff only does filemap_write_and_wait_range, which
+	 * isn't enough if we've written compressed pages to this area, so
+	 * we need to flush the dirty pages again to make absolutely sure
+	 * that any outstanding dirty pages are on disk.
+	 */
+	count = iov_iter_count(iter);
+	if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
+		     &BTRFS_I(inode)->runtime_flags))
+		filemap_fdatawrite_range(inode->i_mapping, offset, count);
+
+	if (rw & WRITE) {
+		/*
+		 * If the write DIO is beyond the EOF, we need update
+		 * the isize, but it is protected by i_mutex. So we can
+		 * not unlock the i_mutex at this case.
+		 */
+		if (offset + count <= inode->i_size) {
+			mutex_unlock(&inode->i_mutex);
+			relock = true;
+		}
+		ret = btrfs_delalloc_reserve_space(inode, count);
+		if (ret)
+			goto out;
+	} else if (unlikely(test_bit(BTRFS_INODE_READDIO_NEED_LOCK,
+				     &BTRFS_I(inode)->runtime_flags))) {
+		inode_dio_done(inode);
+		flags = DIO_LOCKING | DIO_SKIP_HOLES;
+		wakeup = false;
+	}
+
+	ret = __blockdev_direct_IO(rw, iocb, inode,
+			BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev,
+			iter, offset, btrfs_get_blocks_direct, NULL,
+			btrfs_submit_direct, flags);
+	if (rw & WRITE) {
+		if (ret < 0 && ret != -EIOCBQUEUED)
+			btrfs_delalloc_release_space(inode, count);
+		else if (ret >= 0 && (size_t)ret < count)
+			btrfs_delalloc_release_space(inode,
+						     count - (size_t)ret);
+		else
+			btrfs_delalloc_release_metadata(inode, 0);
+	}
+out:
+	if (wakeup)
+		inode_dio_done(inode);
+	if (relock)
+		mutex_lock(&inode->i_mutex);
+
+	return ret;
 }
 
+#define BTRFS_FIEMAP_FLAGS	(FIEMAP_FLAG_SYNC)
+
 static int btrfs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
 		__u64 start, __u64 len)
 {
-	return extent_fiemap(inode, fieinfo, start, len, btrfs_get_extent);
+	int	ret;
+
+	ret = fiemap_check_flags(fieinfo, BTRFS_FIEMAP_FLAGS);
+	if (ret)
+		return ret;
+
+	return extent_fiemap(inode, fieinfo, start, len, btrfs_get_extent_fiemap);
 }
 
 int btrfs_readpage(struct file *file, struct page *page)
 {
 	struct extent_io_tree *tree;
 	tree = &BTRFS_I(page->mapping->host)->io_tree;
-	return extent_read_full_page(tree, page, btrfs_get_extent);
+	return extent_read_full_page(tree, page, btrfs_get_extent, 0);
 }
 
 static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
@@ -4909,8 +7601,8 @@ static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
 	return extent_write_full_page(tree, page, btrfs_get_extent, wbc);
 }
 
-int btrfs_writepages(struct address_space *mapping,
-		     struct writeback_control *wbc)
+static int btrfs_writepages(struct address_space *mapping,
+			    struct writeback_control *wbc)
 {
 	struct extent_io_tree *tree;
 
@@ -4951,14 +7643,16 @@ static int btrfs_releasepage(struct page *page, gfp_t gfp_flags)
 	return __btrfs_releasepage(page, gfp_flags & GFP_NOFS);
 }
 
-static void btrfs_invalidatepage(struct page *page, unsigned long offset)
+static void btrfs_invalidatepage(struct page *page, unsigned int offset,
+				 unsigned int length)
 {
+	struct inode *inode = page->mapping->host;
 	struct extent_io_tree *tree;
 	struct btrfs_ordered_extent *ordered;
 	struct extent_state *cached_state = NULL;
 	u64 page_start = page_offset(page);
 	u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
-
+	int inode_evicting = inode->i_state & I_FREEING;
 
 	/*
 	 * we have the page locked, so new writeback can't start,
@@ -4969,41 +7663,65 @@ static void btrfs_invalidatepage(struct page *page, unsigned long offset)
 	 */
 	wait_on_page_writeback(page);
 
-	tree = &BTRFS_I(page->mapping->host)->io_tree;
+	tree = &BTRFS_I(inode)->io_tree;
 	if (offset) {
 		btrfs_releasepage(page, GFP_NOFS);
 		return;
 	}
-	lock_extent_bits(tree, page_start, page_end, 0, &cached_state,
-			 GFP_NOFS);
-	ordered = btrfs_lookup_ordered_extent(page->mapping->host,
-					   page_offset(page));
+
+	if (!inode_evicting)
+		lock_extent_bits(tree, page_start, page_end, 0, &cached_state);
+	ordered = btrfs_lookup_ordered_extent(inode, page_start);
 	if (ordered) {
 		/*
 		 * IO on this page will never be started, so we need
 		 * to account for any ordered extents now
 		 */
-		clear_extent_bit(tree, page_start, page_end,
-				 EXTENT_DIRTY | EXTENT_DELALLOC |
-				 EXTENT_LOCKED | EXTENT_DO_ACCOUNTING, 1, 0,
-				 &cached_state, GFP_NOFS);
+		if (!inode_evicting)
+			clear_extent_bit(tree, page_start, page_end,
+					 EXTENT_DIRTY | EXTENT_DELALLOC |
+					 EXTENT_LOCKED | EXTENT_DO_ACCOUNTING |
+					 EXTENT_DEFRAG, 1, 0, &cached_state,
+					 GFP_NOFS);
 		/*
 		 * whoever cleared the private bit is responsible
 		 * for the finish_ordered_io
 		 */
 		if (TestClearPagePrivate2(page)) {
-			btrfs_finish_ordered_io(page->mapping->host,
-						page_start, page_end);
+			struct btrfs_ordered_inode_tree *tree;
+			u64 new_len;
+
+			tree = &BTRFS_I(inode)->ordered_tree;
+
+			spin_lock_irq(&tree->lock);
+			set_bit(BTRFS_ORDERED_TRUNCATED, &ordered->flags);
+			new_len = page_start - ordered->file_offset;
+			if (new_len < ordered->truncated_len)
+				ordered->truncated_len = new_len;
+			spin_unlock_irq(&tree->lock);
+
+			if (btrfs_dec_test_ordered_pending(inode, &ordered,
+							   page_start,
+							   PAGE_CACHE_SIZE, 1))
+				btrfs_finish_ordered_io(ordered);
 		}
 		btrfs_put_ordered_extent(ordered);
-		cached_state = NULL;
-		lock_extent_bits(tree, page_start, page_end, 0, &cached_state,
-				 GFP_NOFS);
+		if (!inode_evicting) {
+			cached_state = NULL;
+			lock_extent_bits(tree, page_start, page_end, 0,
+					 &cached_state);
+		}
+	}
+
+	if (!inode_evicting) {
+		clear_extent_bit(tree, page_start, page_end,
+				 EXTENT_LOCKED | EXTENT_DIRTY |
+				 EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING |
+				 EXTENT_DEFRAG, 1, 1,
+				 &cached_state, GFP_NOFS);
+
+		__btrfs_releasepage(page, GFP_NOFS);
 	}
-	clear_extent_bit(tree, page_start, page_end,
-		 EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC |
-		 EXTENT_DO_ACCOUNTING, 1, 1, &cached_state, GFP_NOFS);
-	__btrfs_releasepage(page, GFP_NOFS);
 
 	ClearPageChecked(page);
 	if (PagePrivate(page)) {
@@ -5031,7 +7749,7 @@ static void btrfs_invalidatepage(struct page *page, unsigned long offset)
 int btrfs_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
 {
 	struct page *page = vmf->page;
-	struct inode *inode = fdentry(vma->vm_file)->d_inode;
+	struct inode *inode = file_inode(vma->vm_file);
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
 	struct btrfs_ordered_extent *ordered;
@@ -5040,23 +7758,24 @@ int btrfs_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
 	unsigned long zero_start;
 	loff_t size;
 	int ret;
+	int reserved = 0;
 	u64 page_start;
 	u64 page_end;
 
-	ret = btrfs_check_data_free_space(root, inode, PAGE_CACHE_SIZE);
+	sb_start_pagefault(inode->i_sb);
+	ret  = btrfs_delalloc_reserve_space(inode, PAGE_CACHE_SIZE);
+	if (!ret) {
+		ret = file_update_time(vma->vm_file);
+		reserved = 1;
+	}
 	if (ret) {
 		if (ret == -ENOMEM)
 			ret = VM_FAULT_OOM;
 		else /* -ENOSPC, -EIO, etc */
 			ret = VM_FAULT_SIGBUS;
-		goto out;
-	}
-
-	ret = btrfs_reserve_metadata_for_delalloc(root, inode, 1);
-	if (ret) {
-		btrfs_free_reserved_data_space(root, inode, PAGE_CACHE_SIZE);
-		ret = VM_FAULT_SIGBUS;
-		goto out;
+		if (reserved)
+			goto out;
+		goto out_noreserve;
 	}
 
 	ret = VM_FAULT_NOPAGE; /* make the VM retry the fault */
@@ -5068,14 +7787,12 @@ again:
 
 	if ((page->mapping != inode->i_mapping) ||
 	    (page_start >= size)) {
-		btrfs_free_reserved_data_space(root, inode, PAGE_CACHE_SIZE);
 		/* page got truncated out from underneath us */
 		goto out_unlock;
 	}
 	wait_on_page_writeback(page);
 
-	lock_extent_bits(io_tree, page_start, page_end, 0, &cached_state,
-			 GFP_NOFS);
+	lock_extent_bits(io_tree, page_start, page_end, 0, &cached_state);
 	set_page_extent_mapped(page);
 
 	/*
@@ -5100,7 +7817,8 @@ again:
 	 * prepare_pages in the normal write path.
 	 */
 	clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, page_end,
-			  EXTENT_DIRTY | EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING,
+			  EXTENT_DIRTY | EXTENT_DELALLOC |
+			  EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG,
 			  0, 0, &cached_state, GFP_NOFS);
 
 	ret = btrfs_set_extent_delalloc(inode, page_start, page_end,
@@ -5109,7 +7827,6 @@ again:
 		unlock_extent_cached(io_tree, page_start, page_end,
 				     &cached_state, GFP_NOFS);
 		ret = VM_FAULT_SIGBUS;
-		btrfs_free_reserved_data_space(root, inode, PAGE_CACHE_SIZE);
 		goto out_unlock;
 	}
 	ret = 0;
@@ -5132,40 +7849,94 @@ again:
 
 	BTRFS_I(inode)->last_trans = root->fs_info->generation;
 	BTRFS_I(inode)->last_sub_trans = BTRFS_I(inode)->root->log_transid;
+	BTRFS_I(inode)->last_log_commit = BTRFS_I(inode)->root->last_log_commit;
 
 	unlock_extent_cached(io_tree, page_start, page_end, &cached_state, GFP_NOFS);
 
 out_unlock:
-	btrfs_unreserve_metadata_for_delalloc(root, inode, 1);
-	if (!ret)
+	if (!ret) {
+		sb_end_pagefault(inode->i_sb);
 		return VM_FAULT_LOCKED;
+	}
 	unlock_page(page);
 out:
+	btrfs_delalloc_release_space(inode, PAGE_CACHE_SIZE);
+out_noreserve:
+	sb_end_pagefault(inode->i_sb);
 	return ret;
 }
 
-static void btrfs_truncate(struct inode *inode)
+static int btrfs_truncate(struct inode *inode)
 {
 	struct btrfs_root *root = BTRFS_I(inode)->root;
-	int ret;
+	struct btrfs_block_rsv *rsv;
+	int ret = 0;
+	int err = 0;
 	struct btrfs_trans_handle *trans;
-	unsigned long nr;
 	u64 mask = root->sectorsize - 1;
+	u64 min_size = btrfs_calc_trunc_metadata_size(root, 1);
 
-	if (!S_ISREG(inode->i_mode)) {
-		WARN_ON(1);
-		return;
-	}
-
-	ret = btrfs_truncate_page(inode->i_mapping, inode->i_size);
+	ret = btrfs_wait_ordered_range(inode, inode->i_size & (~mask),
+				       (u64)-1);
 	if (ret)
-		return;
+		return ret;
 
-	btrfs_wait_ordered_range(inode, inode->i_size & (~mask), (u64)-1);
-	btrfs_ordered_update_i_size(inode, inode->i_size, NULL);
+	/*
+	 * Yes ladies and gentelment, this is indeed ugly.  The fact is we have
+	 * 3 things going on here
+	 *
+	 * 1) We need to reserve space for our orphan item and the space to
+	 * delete our orphan item.  Lord knows we don't want to have a dangling
+	 * orphan item because we didn't reserve space to remove it.
+	 *
+	 * 2) We need to reserve space to update our inode.
+	 *
+	 * 3) We need to have something to cache all the space that is going to
+	 * be free'd up by the truncate operation, but also have some slack
+	 * space reserved in case it uses space during the truncate (thank you
+	 * very much snapshotting).
+	 *
+	 * And we need these to all be seperate.  The fact is we can use alot of
+	 * space doing the truncate, and we have no earthly idea how much space
+	 * we will use, so we need the truncate reservation to be seperate so it
+	 * doesn't end up using space reserved for updating the inode or
+	 * removing the orphan item.  We also need to be able to stop the
+	 * transaction and start a new one, which means we need to be able to
+	 * update the inode several times, and we have no idea of knowing how
+	 * many times that will be, so we can't just reserve 1 item for the
+	 * entirety of the opration, so that has to be done seperately as well.
+	 * Then there is the orphan item, which does indeed need to be held on
+	 * to for the whole operation, and we need nobody to touch this reserved
+	 * space except the orphan code.
+	 *
+	 * So that leaves us with
+	 *
+	 * 1) root->orphan_block_rsv - for the orphan deletion.
+	 * 2) rsv - for the truncate reservation, which we will steal from the
+	 * transaction reservation.
+	 * 3) fs_info->trans_block_rsv - this will have 1 items worth left for
+	 * updating the inode.
+	 */
+	rsv = btrfs_alloc_block_rsv(root, BTRFS_BLOCK_RSV_TEMP);
+	if (!rsv)
+		return -ENOMEM;
+	rsv->size = min_size;
+	rsv->failfast = 1;
+
+	/*
+	 * 1 for the truncate slack space
+	 * 1 for updating the inode.
+	 */
+	trans = btrfs_start_transaction(root, 2);
+	if (IS_ERR(trans)) {
+		err = PTR_ERR(trans);
+		goto out;
+	}
 
-	trans = btrfs_start_transaction(root, 1);
-	btrfs_set_trans_block_group(trans, inode);
+	/* Migrate the slack space for the truncate to our reserve */
+	ret = btrfs_block_rsv_migrate(&root->fs_info->trans_block_rsv, rsv,
+				      min_size);
+	BUG_ON(ret);
 
 	/*
 	 * setattr is responsible for setting the ordered_data_close flag,
@@ -5184,39 +7955,76 @@ static void btrfs_truncate(struct inode *inode)
 	 * using truncate to replace the contents of the file will
 	 * end up with a zero length file after a crash.
 	 */
-	if (inode->i_size == 0 && BTRFS_I(inode)->ordered_data_close)
+	if (inode->i_size == 0 && test_bit(BTRFS_INODE_ORDERED_DATA_CLOSE,
+					   &BTRFS_I(inode)->runtime_flags))
 		btrfs_add_ordered_operation(trans, root, inode);
 
+	/*
+	 * So if we truncate and then write and fsync we normally would just
+	 * write the extents that changed, which is a problem if we need to
+	 * first truncate that entire inode.  So set this flag so we write out
+	 * all of the extents in the inode to the sync log so we're completely
+	 * safe.
+	 */
+	set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &BTRFS_I(inode)->runtime_flags);
+	trans->block_rsv = rsv;
+
 	while (1) {
 		ret = btrfs_truncate_inode_items(trans, root, inode,
 						 inode->i_size,
 						 BTRFS_EXTENT_DATA_KEY);
-		if (ret != -EAGAIN)
+		if (ret != -ENOSPC) {
+			err = ret;
 			break;
+		}
 
+		trans->block_rsv = &root->fs_info->trans_block_rsv;
 		ret = btrfs_update_inode(trans, root, inode);
-		BUG_ON(ret);
+		if (ret) {
+			err = ret;
+			break;
+		}
 
-		nr = trans->blocks_used;
 		btrfs_end_transaction(trans, root);
-		btrfs_btree_balance_dirty(root, nr);
+		btrfs_btree_balance_dirty(root);
 
-		trans = btrfs_start_transaction(root, 1);
-		btrfs_set_trans_block_group(trans, inode);
+		trans = btrfs_start_transaction(root, 2);
+		if (IS_ERR(trans)) {
+			ret = err = PTR_ERR(trans);
+			trans = NULL;
+			break;
+		}
+
+		ret = btrfs_block_rsv_migrate(&root->fs_info->trans_block_rsv,
+					      rsv, min_size);
+		BUG_ON(ret);	/* shouldn't happen */
+		trans->block_rsv = rsv;
 	}
 
 	if (ret == 0 && inode->i_nlink > 0) {
+		trans->block_rsv = root->orphan_block_rsv;
 		ret = btrfs_orphan_del(trans, inode);
-		BUG_ON(ret);
+		if (ret)
+			err = ret;
 	}
 
-	ret = btrfs_update_inode(trans, root, inode);
-	BUG_ON(ret);
+	if (trans) {
+		trans->block_rsv = &root->fs_info->trans_block_rsv;
+		ret = btrfs_update_inode(trans, root, inode);
+		if (ret && !err)
+			err = ret;
 
-	nr = trans->blocks_used;
-	ret = btrfs_end_transaction_throttle(trans, root);
-	BUG_ON(ret);
-	btrfs_btree_balance_dirty(root, nr);
+		ret = btrfs_end_transaction(trans, root);
+		btrfs_btree_balance_dirty(root);
+	}
+
+out:
+	btrfs_free_block_rsv(root, rsv);
+
+	if (ret && !err)
+		err = ret;
+
+	return err;
 }
 
 /*
@@ -5224,60 +8032,98 @@ static void btrfs_truncate(struct inode *inode)
  */
 int btrfs_create_subvol_root(struct btrfs_trans_handle *trans,
 			     struct btrfs_root *new_root,
-			     u64 new_dirid, u64 alloc_hint)
+			     struct btrfs_root *parent_root,
+			     u64 new_dirid)
 {
 	struct inode *inode;
 	int err;
 	u64 index = 0;
 
-	inode = btrfs_new_inode(trans, new_root, NULL, "..", 2, new_dirid,
-				new_dirid, alloc_hint, S_IFDIR | 0700, &index);
+	inode = btrfs_new_inode(trans, new_root, NULL, "..", 2,
+				new_dirid, new_dirid,
+				S_IFDIR | (~current_umask() & S_IRWXUGO),
+				&index);
 	if (IS_ERR(inode))
 		return PTR_ERR(inode);
 	inode->i_op = &btrfs_dir_inode_operations;
 	inode->i_fop = &btrfs_dir_file_operations;
 
-	inode->i_nlink = 1;
+	set_nlink(inode, 1);
 	btrfs_i_size_write(inode, 0);
 
+	err = btrfs_subvol_inherit_props(trans, new_root, parent_root);
+	if (err)
+		btrfs_err(new_root->fs_info,
+			  "error inheriting subvolume %llu properties: %d",
+			  new_root->root_key.objectid, err);
+
 	err = btrfs_update_inode(trans, new_root, inode);
-	BUG_ON(err);
 
 	iput(inode);
-	return 0;
-}
-
-/* helper function for file defrag and space balancing.  This
- * forces readahead on a given range of bytes in an inode
- */
-unsigned long btrfs_force_ra(struct address_space *mapping,
-			      struct file_ra_state *ra, struct file *file,
-			      pgoff_t offset, pgoff_t last_index)
-{
-	pgoff_t req_size = last_index - offset + 1;
-
-	page_cache_sync_readahead(mapping, ra, file, offset, req_size);
-	return offset + req_size;
+	return err;
 }
 
 struct inode *btrfs_alloc_inode(struct super_block *sb)
 {
 	struct btrfs_inode *ei;
+	struct inode *inode;
 
 	ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS);
 	if (!ei)
 		return NULL;
+
+	ei->root = NULL;
+	ei->generation = 0;
 	ei->last_trans = 0;
 	ei->last_sub_trans = 0;
 	ei->logged_trans = 0;
+	ei->delalloc_bytes = 0;
+	ei->disk_i_size = 0;
+	ei->flags = 0;
+	ei->csum_bytes = 0;
+	ei->index_cnt = (u64)-1;
+	ei->dir_index = 0;
+	ei->last_unlink_trans = 0;
+	ei->last_log_commit = 0;
+
+	spin_lock_init(&ei->lock);
 	ei->outstanding_extents = 0;
 	ei->reserved_extents = 0;
-	ei->root = NULL;
-	spin_lock_init(&ei->accounting_lock);
+
+	ei->runtime_flags = 0;
+	ei->force_compress = BTRFS_COMPRESS_NONE;
+
+	ei->delayed_node = NULL;
+
+	inode = &ei->vfs_inode;
+	extent_map_tree_init(&ei->extent_tree);
+	extent_io_tree_init(&ei->io_tree, &inode->i_data);
+	extent_io_tree_init(&ei->io_failure_tree, &inode->i_data);
+	ei->io_tree.track_uptodate = 1;
+	ei->io_failure_tree.track_uptodate = 1;
+	atomic_set(&ei->sync_writers, 0);
+	mutex_init(&ei->log_mutex);
+	mutex_init(&ei->delalloc_mutex);
 	btrfs_ordered_inode_tree_init(&ei->ordered_tree);
-	INIT_LIST_HEAD(&ei->i_orphan);
+	INIT_LIST_HEAD(&ei->delalloc_inodes);
 	INIT_LIST_HEAD(&ei->ordered_operations);
-	return &ei->vfs_inode;
+	RB_CLEAR_NODE(&ei->rb_node);
+
+	return inode;
+}
+
+#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
+void btrfs_test_destroy_inode(struct inode *inode)
+{
+	btrfs_drop_extent_cache(inode, 0, (u64)-1, 0);
+	kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
+}
+#endif
+
+static void btrfs_i_callback(struct rcu_head *head)
+{
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
 }
 
 void btrfs_destroy_inode(struct inode *inode)
@@ -5285,8 +8131,12 @@ void btrfs_destroy_inode(struct inode *inode)
 	struct btrfs_ordered_extent *ordered;
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 
-	WARN_ON(!list_empty(&inode->i_dentry));
+	WARN_ON(!hlist_empty(&inode->i_dentry));
 	WARN_ON(inode->i_data.nrpages);
+	WARN_ON(BTRFS_I(inode)->outstanding_extents);
+	WARN_ON(BTRFS_I(inode)->reserved_extents);
+	WARN_ON(BTRFS_I(inode)->delalloc_bytes);
+	WARN_ON(BTRFS_I(inode)->csum_bytes);
 
 	/*
 	 * This can happen where we create an inode, but somebody else also
@@ -5302,28 +8152,25 @@ void btrfs_destroy_inode(struct inode *inode)
 	 */
 	smp_mb();
 	if (!list_empty(&BTRFS_I(inode)->ordered_operations)) {
-		spin_lock(&root->fs_info->ordered_extent_lock);
+		spin_lock(&root->fs_info->ordered_root_lock);
 		list_del_init(&BTRFS_I(inode)->ordered_operations);
-		spin_unlock(&root->fs_info->ordered_extent_lock);
+		spin_unlock(&root->fs_info->ordered_root_lock);
 	}
 
-	spin_lock(&root->list_lock);
-	if (!list_empty(&BTRFS_I(inode)->i_orphan)) {
-		printk(KERN_INFO "BTRFS: inode %lu still on the orphan list\n",
-		       inode->i_ino);
-		list_del_init(&BTRFS_I(inode)->i_orphan);
+	if (test_bit(BTRFS_INODE_HAS_ORPHAN_ITEM,
+		     &BTRFS_I(inode)->runtime_flags)) {
+		btrfs_info(root->fs_info, "inode %llu still on the orphan list",
+			btrfs_ino(inode));
+		atomic_dec(&root->orphan_inodes);
 	}
-	spin_unlock(&root->list_lock);
 
 	while (1) {
 		ordered = btrfs_lookup_first_ordered_extent(inode, (u64)-1);
 		if (!ordered)
 			break;
 		else {
-			printk(KERN_ERR "btrfs found ordered "
-			       "extent %llu %llu on inode cleanup\n",
-			       (unsigned long long)ordered->file_offset,
-			       (unsigned long long)ordered->len);
+			btrfs_err(root->fs_info, "found ordered extent %llu %llu on inode cleanup",
+				ordered->file_offset, ordered->len);
 			btrfs_remove_ordered_extent(inode, ordered);
 			btrfs_put_ordered_extent(ordered);
 			btrfs_put_ordered_extent(ordered);
@@ -5332,16 +8179,21 @@ void btrfs_destroy_inode(struct inode *inode)
 	inode_tree_del(inode);
 	btrfs_drop_extent_cache(inode, 0, (u64)-1, 0);
 free:
-	kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
+	call_rcu(&inode->i_rcu, btrfs_i_callback);
 }
 
-void btrfs_drop_inode(struct inode *inode)
+int btrfs_drop_inode(struct inode *inode)
 {
 	struct btrfs_root *root = BTRFS_I(inode)->root;
-	if (inode->i_nlink > 0 && btrfs_root_refs(&root->root_item) == 0)
-		generic_delete_inode(inode);
+
+	if (root == NULL)
+		return 1;
+
+	/* the snap/subvol tree is on deleting */
+	if (btrfs_root_refs(&root->root_item) == 0)
+		return 1;
 	else
-		generic_drop_inode(inode);
+		return generic_drop_inode(inode);
 }
 
 static void init_once(void *foo)
@@ -5353,6 +8205,11 @@ static void init_once(void *foo)
 
 void btrfs_destroy_cachep(void)
 {
+	/*
+	 * Make sure all delayed rcu free inodes are flushed before we
+	 * destroy cache.
+	 */
+	rcu_barrier();
 	if (btrfs_inode_cachep)
 		kmem_cache_destroy(btrfs_inode_cachep);
 	if (btrfs_trans_handle_cachep)
@@ -5361,34 +8218,51 @@ void btrfs_destroy_cachep(void)
 		kmem_cache_destroy(btrfs_transaction_cachep);
 	if (btrfs_path_cachep)
 		kmem_cache_destroy(btrfs_path_cachep);
+	if (btrfs_free_space_cachep)
+		kmem_cache_destroy(btrfs_free_space_cachep);
+	if (btrfs_delalloc_work_cachep)
+		kmem_cache_destroy(btrfs_delalloc_work_cachep);
 }
 
 int btrfs_init_cachep(void)
 {
-	btrfs_inode_cachep = kmem_cache_create("btrfs_inode_cache",
+	btrfs_inode_cachep = kmem_cache_create("btrfs_inode",
 			sizeof(struct btrfs_inode), 0,
 			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, init_once);
 	if (!btrfs_inode_cachep)
 		goto fail;
 
-	btrfs_trans_handle_cachep = kmem_cache_create("btrfs_trans_handle_cache",
+	btrfs_trans_handle_cachep = kmem_cache_create("btrfs_trans_handle",
 			sizeof(struct btrfs_trans_handle), 0,
 			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
 	if (!btrfs_trans_handle_cachep)
 		goto fail;
 
-	btrfs_transaction_cachep = kmem_cache_create("btrfs_transaction_cache",
+	btrfs_transaction_cachep = kmem_cache_create("btrfs_transaction",
 			sizeof(struct btrfs_transaction), 0,
 			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
 	if (!btrfs_transaction_cachep)
 		goto fail;
 
-	btrfs_path_cachep = kmem_cache_create("btrfs_path_cache",
+	btrfs_path_cachep = kmem_cache_create("btrfs_path",
 			sizeof(struct btrfs_path), 0,
 			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
 	if (!btrfs_path_cachep)
 		goto fail;
 
+	btrfs_free_space_cachep = kmem_cache_create("btrfs_free_space",
+			sizeof(struct btrfs_free_space), 0,
+			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
+	if (!btrfs_free_space_cachep)
+		goto fail;
+
+	btrfs_delalloc_work_cachep = kmem_cache_create("btrfs_delalloc_work",
+			sizeof(struct btrfs_delalloc_work), 0,
+			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD,
+			NULL);
+	if (!btrfs_delalloc_work_cachep)
+		goto fail;
+
 	return 0;
 fail:
 	btrfs_destroy_cachep();
@@ -5398,12 +8272,19 @@ fail:
 static int btrfs_getattr(struct vfsmount *mnt,
 			 struct dentry *dentry, struct kstat *stat)
 {
+	u64 delalloc_bytes;
 	struct inode *inode = dentry->d_inode;
+	u32 blocksize = inode->i_sb->s_blocksize;
+
 	generic_fillattr(inode, stat);
-	stat->dev = BTRFS_I(inode)->root->anon_super.s_dev;
+	stat->dev = BTRFS_I(inode)->root->anon_dev;
 	stat->blksize = PAGE_CACHE_SIZE;
-	stat->blocks = (inode_get_bytes(inode) +
-			BTRFS_I(inode)->delalloc_bytes) >> 9;
+
+	spin_lock(&BTRFS_I(inode)->lock);
+	delalloc_bytes = BTRFS_I(inode)->delalloc_bytes;
+	spin_unlock(&BTRFS_I(inode)->lock);
+	stat->blocks = (ALIGN(inode_get_bytes(inode), blocksize) +
+			ALIGN(delalloc_bytes, blocksize)) >> 9;
 	return 0;
 }
 
@@ -5419,33 +8300,42 @@ static int btrfs_rename(struct inode *old_dir, struct dentry *old_dentry,
 	u64 index = 0;
 	u64 root_objectid;
 	int ret;
+	u64 old_ino = btrfs_ino(old_inode);
 
-	if (new_dir->i_ino == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID)
+	if (btrfs_ino(new_dir) == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID)
 		return -EPERM;
 
 	/* we only allow rename subvolume link between subvolumes */
-	if (old_inode->i_ino != BTRFS_FIRST_FREE_OBJECTID && root != dest)
+	if (old_ino != BTRFS_FIRST_FREE_OBJECTID && root != dest)
 		return -EXDEV;
 
-	if (old_inode->i_ino == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID ||
-	    (new_inode && new_inode->i_ino == BTRFS_FIRST_FREE_OBJECTID))
+	if (old_ino == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID ||
+	    (new_inode && btrfs_ino(new_inode) == BTRFS_FIRST_FREE_OBJECTID))
 		return -ENOTEMPTY;
 
 	if (S_ISDIR(old_inode->i_mode) && new_inode &&
 	    new_inode->i_size > BTRFS_EMPTY_DIR_SIZE)
 		return -ENOTEMPTY;
 
-	/*
-	 * We want to reserve the absolute worst case amount of items.  So if
-	 * both inodes are subvols and we need to unlink them then that would
-	 * require 4 item modifications, but if they are both normal inodes it
-	 * would require 5 item modifications, so we'll assume their normal
-	 * inodes.  So 5 * 2 is 10, plus 1 for the new link, so 11 total items
-	 * should cover the worst case number of items we'll modify.
-	 */
-	ret = btrfs_reserve_metadata_space(root, 11);
-	if (ret)
-		return ret;
+
+	/* check for collisions, even if the  name isn't there */
+	ret = btrfs_check_dir_item_collision(dest, new_dir->i_ino,
+			     new_dentry->d_name.name,
+			     new_dentry->d_name.len);
+
+	if (ret) {
+		if (ret == -EEXIST) {
+			/* we shouldn't get
+			 * eexist without a new_inode */
+			if (WARN_ON(!new_inode)) {
+				return ret;
+			}
+		} else {
+			/* maybe -EOVERFLOW */
+			return ret;
+		}
+	}
+	ret = 0;
 
 	/*
 	 * we're using rename to replace one file with another.
@@ -5457,11 +8347,21 @@ static int btrfs_rename(struct inode *old_dir, struct dentry *old_dentry,
 		filemap_flush(old_inode->i_mapping);
 
 	/* close the racy window with snapshot create/destroy ioctl */
-	if (old_inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
+	if (old_ino == BTRFS_FIRST_FREE_OBJECTID)
 		down_read(&root->fs_info->subvol_sem);
-
-	trans = btrfs_start_transaction(root, 1);
-	btrfs_set_trans_block_group(trans, new_dir);
+	/*
+	 * We want to reserve the absolute worst case amount of items.  So if
+	 * both inodes are subvols and we need to unlink them then that would
+	 * require 4 item modifications, but if they are both normal inodes it
+	 * would require 5 item modifications, so we'll assume their normal
+	 * inodes.  So 5 * 2 is 10, plus 1 for the new link, so 11 total items
+	 * should cover the worst case number of items we'll modify.
+	 */
+	trans = btrfs_start_transaction(root, 11);
+	if (IS_ERR(trans)) {
+                ret = PTR_ERR(trans);
+                goto out_notrans;
+        }
 
 	if (dest != root)
 		btrfs_record_root_in_trans(trans, dest);
@@ -5470,15 +8370,16 @@ static int btrfs_rename(struct inode *old_dir, struct dentry *old_dentry,
 	if (ret)
 		goto out_fail;
 
-	if (unlikely(old_inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)) {
+	BTRFS_I(old_inode)->dir_index = 0ULL;
+	if (unlikely(old_ino == BTRFS_FIRST_FREE_OBJECTID)) {
 		/* force full log commit if subvolume involved. */
-		root->fs_info->last_trans_log_full_commit = trans->transid;
+		btrfs_set_log_full_commit(root->fs_info, trans);
 	} else {
 		ret = btrfs_insert_inode_ref(trans, dest,
 					     new_dentry->d_name.name,
 					     new_dentry->d_name.len,
-					     old_inode->i_ino,
-					     new_dir->i_ino, index);
+					     old_ino,
+					     btrfs_ino(new_dir), index);
 		if (ret)
 			goto out_fail;
 		/*
@@ -5494,11 +8395,12 @@ static int btrfs_rename(struct inode *old_dir, struct dentry *old_dentry,
 	 * make sure the inode gets flushed if it is replacing
 	 * something.
 	 */
-	if (new_inode && new_inode->i_size &&
-	    old_inode && S_ISREG(old_inode->i_mode)) {
+	if (new_inode && new_inode->i_size && S_ISREG(old_inode->i_mode))
 		btrfs_add_ordered_operation(trans, root, old_inode);
-	}
 
+	inode_inc_iversion(old_dir);
+	inode_inc_iversion(new_dir);
+	inode_inc_iversion(old_inode);
 	old_dir->i_ctime = old_dir->i_mtime = ctime;
 	new_dir->i_ctime = new_dir->i_mtime = ctime;
 	old_inode->i_ctime = ctime;
@@ -5506,23 +8408,28 @@ static int btrfs_rename(struct inode *old_dir, struct dentry *old_dentry,
 	if (old_dentry->d_parent != new_dentry->d_parent)
 		btrfs_record_unlink_dir(trans, old_dir, old_inode, 1);
 
-	if (unlikely(old_inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)) {
+	if (unlikely(old_ino == BTRFS_FIRST_FREE_OBJECTID)) {
 		root_objectid = BTRFS_I(old_inode)->root->root_key.objectid;
 		ret = btrfs_unlink_subvol(trans, root, old_dir, root_objectid,
 					old_dentry->d_name.name,
 					old_dentry->d_name.len);
 	} else {
-		btrfs_inc_nlink(old_dentry->d_inode);
-		ret = btrfs_unlink_inode(trans, root, old_dir,
-					 old_dentry->d_inode,
-					 old_dentry->d_name.name,
-					 old_dentry->d_name.len);
+		ret = __btrfs_unlink_inode(trans, root, old_dir,
+					old_dentry->d_inode,
+					old_dentry->d_name.name,
+					old_dentry->d_name.len);
+		if (!ret)
+			ret = btrfs_update_inode(trans, root, old_inode);
+	}
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
+		goto out_fail;
 	}
-	BUG_ON(ret);
 
 	if (new_inode) {
+		inode_inc_iversion(new_inode);
 		new_inode->i_ctime = CURRENT_TIME;
-		if (unlikely(new_inode->i_ino ==
+		if (unlikely(btrfs_ino(new_inode) ==
 			     BTRFS_EMPTY_SUBVOL_DIR_OBJECTID)) {
 			root_objectid = BTRFS_I(new_inode)->location.objectid;
 			ret = btrfs_unlink_subvol(trans, dest, new_dir,
@@ -5536,67 +8443,168 @@ static int btrfs_rename(struct inode *old_dir, struct dentry *old_dentry,
 						 new_dentry->d_name.name,
 						 new_dentry->d_name.len);
 		}
-		BUG_ON(ret);
-		if (new_inode->i_nlink == 0) {
+		if (!ret && new_inode->i_nlink == 0)
 			ret = btrfs_orphan_add(trans, new_dentry->d_inode);
-			BUG_ON(ret);
+		if (ret) {
+			btrfs_abort_transaction(trans, root, ret);
+			goto out_fail;
 		}
 	}
 
 	ret = btrfs_add_link(trans, new_dir, old_inode,
 			     new_dentry->d_name.name,
 			     new_dentry->d_name.len, 0, index);
-	BUG_ON(ret);
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
+		goto out_fail;
+	}
+
+	if (old_inode->i_nlink == 1)
+		BTRFS_I(old_inode)->dir_index = index;
 
-	if (old_inode->i_ino != BTRFS_FIRST_FREE_OBJECTID) {
-		btrfs_log_new_name(trans, old_inode, old_dir,
-				   new_dentry->d_parent);
+	if (old_ino != BTRFS_FIRST_FREE_OBJECTID) {
+		struct dentry *parent = new_dentry->d_parent;
+		btrfs_log_new_name(trans, old_inode, old_dir, parent);
 		btrfs_end_log_trans(root);
 	}
 out_fail:
-	btrfs_end_transaction_throttle(trans, root);
-
-	if (old_inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
+	btrfs_end_transaction(trans, root);
+out_notrans:
+	if (old_ino == BTRFS_FIRST_FREE_OBJECTID)
 		up_read(&root->fs_info->subvol_sem);
 
-	btrfs_unreserve_metadata_space(root, 11);
 	return ret;
 }
 
+static void btrfs_run_delalloc_work(struct btrfs_work *work)
+{
+	struct btrfs_delalloc_work *delalloc_work;
+	struct inode *inode;
+
+	delalloc_work = container_of(work, struct btrfs_delalloc_work,
+				     work);
+	inode = delalloc_work->inode;
+	if (delalloc_work->wait) {
+		btrfs_wait_ordered_range(inode, 0, (u64)-1);
+	} else {
+		filemap_flush(inode->i_mapping);
+		if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
+			     &BTRFS_I(inode)->runtime_flags))
+			filemap_flush(inode->i_mapping);
+	}
+
+	if (delalloc_work->delay_iput)
+		btrfs_add_delayed_iput(inode);
+	else
+		iput(inode);
+	complete(&delalloc_work->completion);
+}
+
+struct btrfs_delalloc_work *btrfs_alloc_delalloc_work(struct inode *inode,
+						    int wait, int delay_iput)
+{
+	struct btrfs_delalloc_work *work;
+
+	work = kmem_cache_zalloc(btrfs_delalloc_work_cachep, GFP_NOFS);
+	if (!work)
+		return NULL;
+
+	init_completion(&work->completion);
+	INIT_LIST_HEAD(&work->list);
+	work->inode = inode;
+	work->wait = wait;
+	work->delay_iput = delay_iput;
+	btrfs_init_work(&work->work, btrfs_run_delalloc_work, NULL, NULL);
+
+	return work;
+}
+
+void btrfs_wait_and_free_delalloc_work(struct btrfs_delalloc_work *work)
+{
+	wait_for_completion(&work->completion);
+	kmem_cache_free(btrfs_delalloc_work_cachep, work);
+}
+
 /*
  * some fairly slow code that needs optimization. This walks the list
  * of all the inodes with pending delalloc and forces them to disk.
  */
-int btrfs_start_delalloc_inodes(struct btrfs_root *root, int delay_iput)
+static int __start_delalloc_inodes(struct btrfs_root *root, int delay_iput,
+				   int nr)
 {
-	struct list_head *head = &root->fs_info->delalloc_inodes;
 	struct btrfs_inode *binode;
 	struct inode *inode;
+	struct btrfs_delalloc_work *work, *next;
+	struct list_head works;
+	struct list_head splice;
+	int ret = 0;
 
-	if (root->fs_info->sb->s_flags & MS_RDONLY)
-		return -EROFS;
+	INIT_LIST_HEAD(&works);
+	INIT_LIST_HEAD(&splice);
 
-	spin_lock(&root->fs_info->delalloc_lock);
-	while (!list_empty(head)) {
-		binode = list_entry(head->next, struct btrfs_inode,
+	mutex_lock(&root->delalloc_mutex);
+	spin_lock(&root->delalloc_lock);
+	list_splice_init(&root->delalloc_inodes, &splice);
+	while (!list_empty(&splice)) {
+		binode = list_entry(splice.next, struct btrfs_inode,
 				    delalloc_inodes);
+
+		list_move_tail(&binode->delalloc_inodes,
+			       &root->delalloc_inodes);
 		inode = igrab(&binode->vfs_inode);
-		if (!inode)
-			list_del_init(&binode->delalloc_inodes);
-		spin_unlock(&root->fs_info->delalloc_lock);
-		if (inode) {
-			filemap_flush(inode->i_mapping);
+		if (!inode) {
+			cond_resched_lock(&root->delalloc_lock);
+			continue;
+		}
+		spin_unlock(&root->delalloc_lock);
+
+		work = btrfs_alloc_delalloc_work(inode, 0, delay_iput);
+		if (unlikely(!work)) {
 			if (delay_iput)
 				btrfs_add_delayed_iput(inode);
 			else
 				iput(inode);
+			ret = -ENOMEM;
+			goto out;
 		}
+		list_add_tail(&work->list, &works);
+		btrfs_queue_work(root->fs_info->flush_workers,
+				 &work->work);
+		ret++;
+		if (nr != -1 && ret >= nr)
+			goto out;
 		cond_resched();
-		spin_lock(&root->fs_info->delalloc_lock);
+		spin_lock(&root->delalloc_lock);
+	}
+	spin_unlock(&root->delalloc_lock);
+
+out:
+	list_for_each_entry_safe(work, next, &works, list) {
+		list_del_init(&work->list);
+		btrfs_wait_and_free_delalloc_work(work);
 	}
-	spin_unlock(&root->fs_info->delalloc_lock);
 
-	/* the filemap_flush will queue IO into the worker threads, but
+	if (!list_empty_careful(&splice)) {
+		spin_lock(&root->delalloc_lock);
+		list_splice_tail(&splice, &root->delalloc_inodes);
+		spin_unlock(&root->delalloc_lock);
+	}
+	mutex_unlock(&root->delalloc_mutex);
+	return ret;
+}
+
+int btrfs_start_delalloc_inodes(struct btrfs_root *root, int delay_iput)
+{
+	int ret;
+
+	if (test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state))
+		return -EROFS;
+
+	ret = __start_delalloc_inodes(root, delay_iput, -1);
+	if (ret > 0)
+		ret = 0;
+	/*
+	 * the filemap_flush will queue IO into the worker threads, but
 	 * we have to make sure the IO is actually started and that
 	 * ordered extents get created before we return
 	 */
@@ -5608,7 +8616,63 @@ int btrfs_start_delalloc_inodes(struct btrfs_root *root, int delay_iput)
 		    atomic_read(&root->fs_info->async_delalloc_pages) == 0));
 	}
 	atomic_dec(&root->fs_info->async_submit_draining);
-	return 0;
+	return ret;
+}
+
+int btrfs_start_delalloc_roots(struct btrfs_fs_info *fs_info, int delay_iput,
+			       int nr)
+{
+	struct btrfs_root *root;
+	struct list_head splice;
+	int ret;
+
+	if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
+		return -EROFS;
+
+	INIT_LIST_HEAD(&splice);
+
+	mutex_lock(&fs_info->delalloc_root_mutex);
+	spin_lock(&fs_info->delalloc_root_lock);
+	list_splice_init(&fs_info->delalloc_roots, &splice);
+	while (!list_empty(&splice) && nr) {
+		root = list_first_entry(&splice, struct btrfs_root,
+					delalloc_root);
+		root = btrfs_grab_fs_root(root);
+		BUG_ON(!root);
+		list_move_tail(&root->delalloc_root,
+			       &fs_info->delalloc_roots);
+		spin_unlock(&fs_info->delalloc_root_lock);
+
+		ret = __start_delalloc_inodes(root, delay_iput, nr);
+		btrfs_put_fs_root(root);
+		if (ret < 0)
+			goto out;
+
+		if (nr != -1) {
+			nr -= ret;
+			WARN_ON(nr < 0);
+		}
+		spin_lock(&fs_info->delalloc_root_lock);
+	}
+	spin_unlock(&fs_info->delalloc_root_lock);
+
+	ret = 0;
+	atomic_inc(&fs_info->async_submit_draining);
+	while (atomic_read(&fs_info->nr_async_submits) ||
+	      atomic_read(&fs_info->async_delalloc_pages)) {
+		wait_event(fs_info->async_submit_wait,
+		   (atomic_read(&fs_info->nr_async_submits) == 0 &&
+		    atomic_read(&fs_info->async_delalloc_pages) == 0));
+	}
+	atomic_dec(&fs_info->async_submit_draining);
+out:
+	if (!list_empty_careful(&splice)) {
+		spin_lock(&fs_info->delalloc_root_lock);
+		list_splice_tail(&splice, &fs_info->delalloc_roots);
+		spin_unlock(&fs_info->delalloc_root_lock);
+	}
+	mutex_unlock(&fs_info->delalloc_root_mutex);
+	return ret;
 }
 
 static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
@@ -5622,15 +8686,14 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
 	int err;
 	int drop_inode = 0;
 	u64 objectid;
-	u64 index = 0 ;
+	u64 index = 0;
 	int name_len;
 	int datasize;
 	unsigned long ptr;
 	struct btrfs_file_extent_item *ei;
 	struct extent_buffer *leaf;
-	unsigned long nr = 0;
 
-	name_len = strlen(symname) + 1;
+	name_len = strlen(symname);
 	if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root))
 		return -ENAMETOOLONG;
 
@@ -5639,55 +8702,55 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
 	 * 2 items for dir items
 	 * 1 item for xattr if selinux is on
 	 */
-	err = btrfs_reserve_metadata_space(root, 5);
-	if (err)
-		return err;
-
-	trans = btrfs_start_transaction(root, 1);
-	if (!trans)
-		goto out_fail;
-	btrfs_set_trans_block_group(trans, dir);
+	trans = btrfs_start_transaction(root, 5);
+	if (IS_ERR(trans))
+		return PTR_ERR(trans);
 
-	err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
-	if (err) {
-		err = -ENOSPC;
+	err = btrfs_find_free_ino(root, &objectid);
+	if (err)
 		goto out_unlock;
-	}
 
 	inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
-				dentry->d_name.len,
-				dentry->d_parent->d_inode->i_ino, objectid,
-				BTRFS_I(dir)->block_group, S_IFLNK|S_IRWXUGO,
-				&index);
-	err = PTR_ERR(inode);
-	if (IS_ERR(inode))
+				dentry->d_name.len, btrfs_ino(dir), objectid,
+				S_IFLNK|S_IRWXUGO, &index);
+	if (IS_ERR(inode)) {
+		err = PTR_ERR(inode);
 		goto out_unlock;
+	}
 
-	err = btrfs_init_inode_security(trans, inode, dir);
+	err = btrfs_init_inode_security(trans, inode, dir, &dentry->d_name);
 	if (err) {
 		drop_inode = 1;
 		goto out_unlock;
 	}
 
-	btrfs_set_trans_block_group(trans, inode);
-	err = btrfs_add_nondir(trans, dentry, inode, 0, index);
+	/*
+	* If the active LSM wants to access the inode during
+	* d_instantiate it needs these. Smack checks to see
+	* if the filesystem supports xattrs by looking at the
+	* ops vector.
+	*/
+	inode->i_fop = &btrfs_file_operations;
+	inode->i_op = &btrfs_file_inode_operations;
+
+	err = btrfs_add_nondir(trans, dir, dentry, inode, 0, index);
 	if (err)
 		drop_inode = 1;
 	else {
 		inode->i_mapping->a_ops = &btrfs_aops;
 		inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
-		inode->i_fop = &btrfs_file_operations;
-		inode->i_op = &btrfs_file_inode_operations;
 		BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
 	}
-	btrfs_update_inode_block_group(trans, inode);
-	btrfs_update_inode_block_group(trans, dir);
 	if (drop_inode)
 		goto out_unlock;
 
 	path = btrfs_alloc_path();
-	BUG_ON(!path);
-	key.objectid = inode->i_ino;
+	if (!path) {
+		err = -ENOMEM;
+		drop_inode = 1;
+		goto out_unlock;
+	}
+	key.objectid = btrfs_ino(inode);
 	key.offset = 0;
 	btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
 	datasize = btrfs_file_extent_calc_inline_size(name_len);
@@ -5695,6 +8758,7 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
 				      datasize);
 	if (err) {
 		drop_inode = 1;
+		btrfs_free_path(path);
 		goto out_unlock;
 	}
 	leaf = path->nodes[0];
@@ -5717,51 +8781,57 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
 	inode->i_mapping->a_ops = &btrfs_symlink_aops;
 	inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
 	inode_set_bytes(inode, name_len);
-	btrfs_i_size_write(inode, name_len - 1);
+	btrfs_i_size_write(inode, name_len);
 	err = btrfs_update_inode(trans, root, inode);
 	if (err)
 		drop_inode = 1;
 
 out_unlock:
-	nr = trans->blocks_used;
-	btrfs_end_transaction_throttle(trans, root);
-out_fail:
-	btrfs_unreserve_metadata_space(root, 5);
+	if (!err)
+		d_instantiate(dentry, inode);
+	btrfs_end_transaction(trans, root);
 	if (drop_inode) {
 		inode_dec_link_count(inode);
 		iput(inode);
 	}
-	btrfs_btree_balance_dirty(root, nr);
+	btrfs_btree_balance_dirty(root);
 	return err;
 }
 
-static int prealloc_file_range(struct inode *inode, u64 start, u64 end,
-			u64 alloc_hint, int mode, loff_t actual_len)
+static int __btrfs_prealloc_file_range(struct inode *inode, int mode,
+				       u64 start, u64 num_bytes, u64 min_size,
+				       loff_t actual_len, u64 *alloc_hint,
+				       struct btrfs_trans_handle *trans)
 {
-	struct btrfs_trans_handle *trans;
+	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
+	struct extent_map *em;
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	struct btrfs_key ins;
 	u64 cur_offset = start;
-	u64 num_bytes = end - start;
-	int ret = 0;
 	u64 i_size;
+	u64 cur_bytes;
+	int ret = 0;
+	bool own_trans = true;
 
+	if (trans)
+		own_trans = false;
 	while (num_bytes > 0) {
-		trans = btrfs_start_transaction(root, 1);
-
-		ret = btrfs_reserve_extent(trans, root, num_bytes,
-					   root->sectorsize, 0, alloc_hint,
-					   (u64)-1, &ins, 1);
-		if (ret) {
-			WARN_ON(1);
-			goto stop_trans;
+		if (own_trans) {
+			trans = btrfs_start_transaction(root, 3);
+			if (IS_ERR(trans)) {
+				ret = PTR_ERR(trans);
+				break;
+			}
 		}
 
-		ret = btrfs_reserve_metadata_space(root, 3);
+		cur_bytes = min(num_bytes, 256ULL * 1024 * 1024);
+		cur_bytes = max(cur_bytes, min_size);
+		ret = btrfs_reserve_extent(root, cur_bytes, min_size, 0,
+					   *alloc_hint, &ins, 1, 0);
 		if (ret) {
-			btrfs_free_reserved_extent(root, ins.objectid,
-						   ins.offset);
-			goto stop_trans;
+			if (own_trans)
+				btrfs_end_transaction(trans, root);
+			break;
 		}
 
 		ret = insert_reserved_file_extent(trans, inode,
@@ -5769,22 +8839,59 @@ static int prealloc_file_range(struct inode *inode, u64 start, u64 end,
 						  ins.offset, ins.offset,
 						  ins.offset, 0, 0, 0,
 						  BTRFS_FILE_EXTENT_PREALLOC);
-		BUG_ON(ret);
+		if (ret) {
+			btrfs_free_reserved_extent(root, ins.objectid,
+						   ins.offset, 0);
+			btrfs_abort_transaction(trans, root, ret);
+			if (own_trans)
+				btrfs_end_transaction(trans, root);
+			break;
+		}
 		btrfs_drop_extent_cache(inode, cur_offset,
 					cur_offset + ins.offset -1, 0);
 
+		em = alloc_extent_map();
+		if (!em) {
+			set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
+				&BTRFS_I(inode)->runtime_flags);
+			goto next;
+		}
+
+		em->start = cur_offset;
+		em->orig_start = cur_offset;
+		em->len = ins.offset;
+		em->block_start = ins.objectid;
+		em->block_len = ins.offset;
+		em->orig_block_len = ins.offset;
+		em->ram_bytes = ins.offset;
+		em->bdev = root->fs_info->fs_devices->latest_bdev;
+		set_bit(EXTENT_FLAG_PREALLOC, &em->flags);
+		em->generation = trans->transid;
+
+		while (1) {
+			write_lock(&em_tree->lock);
+			ret = add_extent_mapping(em_tree, em, 1);
+			write_unlock(&em_tree->lock);
+			if (ret != -EEXIST)
+				break;
+			btrfs_drop_extent_cache(inode, cur_offset,
+						cur_offset + ins.offset - 1,
+						0);
+		}
+		free_extent_map(em);
+next:
 		num_bytes -= ins.offset;
 		cur_offset += ins.offset;
-		alloc_hint = ins.objectid + ins.offset;
+		*alloc_hint = ins.objectid + ins.offset;
 
+		inode_inc_iversion(inode);
 		inode->i_ctime = CURRENT_TIME;
 		BTRFS_I(inode)->flags |= BTRFS_INODE_PREALLOC;
 		if (!(mode & FALLOC_FL_KEEP_SIZE) &&
-			(actual_len > inode->i_size) &&
-			(cur_offset > inode->i_size)) {
-
+		    (actual_len > inode->i_size) &&
+		    (cur_offset > inode->i_size)) {
 			if (cur_offset > actual_len)
-				i_size  = actual_len;
+				i_size = actual_len;
 			else
 				i_size = cur_offset;
 			i_size_write(inode, i_size);
@@ -5792,133 +8899,116 @@ static int prealloc_file_range(struct inode *inode, u64 start, u64 end,
 		}
 
 		ret = btrfs_update_inode(trans, root, inode);
-		BUG_ON(ret);
 
-		btrfs_end_transaction(trans, root);
-		btrfs_unreserve_metadata_space(root, 3);
+		if (ret) {
+			btrfs_abort_transaction(trans, root, ret);
+			if (own_trans)
+				btrfs_end_transaction(trans, root);
+			break;
+		}
+
+		if (own_trans)
+			btrfs_end_transaction(trans, root);
 	}
 	return ret;
+}
 
-stop_trans:
-	btrfs_end_transaction(trans, root);
-	return ret;
+int btrfs_prealloc_file_range(struct inode *inode, int mode,
+			      u64 start, u64 num_bytes, u64 min_size,
+			      loff_t actual_len, u64 *alloc_hint)
+{
+	return __btrfs_prealloc_file_range(inode, mode, start, num_bytes,
+					   min_size, actual_len, alloc_hint,
+					   NULL);
+}
 
+int btrfs_prealloc_file_range_trans(struct inode *inode,
+				    struct btrfs_trans_handle *trans, int mode,
+				    u64 start, u64 num_bytes, u64 min_size,
+				    loff_t actual_len, u64 *alloc_hint)
+{
+	return __btrfs_prealloc_file_range(inode, mode, start, num_bytes,
+					   min_size, actual_len, alloc_hint, trans);
 }
 
-static long btrfs_fallocate(struct inode *inode, int mode,
-			    loff_t offset, loff_t len)
+static int btrfs_set_page_dirty(struct page *page)
 {
-	struct extent_state *cached_state = NULL;
-	u64 cur_offset;
-	u64 last_byte;
-	u64 alloc_start;
-	u64 alloc_end;
-	u64 alloc_hint = 0;
-	u64 locked_end;
-	u64 mask = BTRFS_I(inode)->root->sectorsize - 1;
-	struct extent_map *em;
-	int ret;
+	return __set_page_dirty_nobuffers(page);
+}
+
+static int btrfs_permission(struct inode *inode, int mask)
+{
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	umode_t mode = inode->i_mode;
 
-	alloc_start = offset & ~mask;
-	alloc_end =  (offset + len + mask) & ~mask;
+	if (mask & MAY_WRITE &&
+	    (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))) {
+		if (btrfs_root_readonly(root))
+			return -EROFS;
+		if (BTRFS_I(inode)->flags & BTRFS_INODE_READONLY)
+			return -EACCES;
+	}
+	return generic_permission(inode, mask);
+}
+
+static int btrfs_tmpfile(struct inode *dir, struct dentry *dentry, umode_t mode)
+{
+	struct btrfs_trans_handle *trans;
+	struct btrfs_root *root = BTRFS_I(dir)->root;
+	struct inode *inode = NULL;
+	u64 objectid;
+	u64 index;
+	int ret = 0;
 
 	/*
-	 * wait for ordered IO before we have any locks.  We'll loop again
-	 * below with the locks held.
+	 * 5 units required for adding orphan entry
 	 */
-	btrfs_wait_ordered_range(inode, alloc_start, alloc_end - alloc_start);
+	trans = btrfs_start_transaction(root, 5);
+	if (IS_ERR(trans))
+		return PTR_ERR(trans);
 
-	mutex_lock(&inode->i_mutex);
-	if (alloc_start > inode->i_size) {
-		ret = btrfs_cont_expand(inode, alloc_start);
-		if (ret)
-			goto out;
+	ret = btrfs_find_free_ino(root, &objectid);
+	if (ret)
+		goto out;
+
+	inode = btrfs_new_inode(trans, root, dir, NULL, 0,
+				btrfs_ino(dir), objectid, mode, &index);
+	if (IS_ERR(inode)) {
+		ret = PTR_ERR(inode);
+		inode = NULL;
+		goto out;
 	}
 
-	ret = btrfs_check_data_free_space(BTRFS_I(inode)->root, inode,
-					  alloc_end - alloc_start);
+	ret = btrfs_init_inode_security(trans, inode, dir, NULL);
 	if (ret)
 		goto out;
 
-	locked_end = alloc_end - 1;
-	while (1) {
-		struct btrfs_ordered_extent *ordered;
+	ret = btrfs_update_inode(trans, root, inode);
+	if (ret)
+		goto out;
 
-		/* the extent lock is ordered inside the running
-		 * transaction
-		 */
-		lock_extent_bits(&BTRFS_I(inode)->io_tree, alloc_start,
-				 locked_end, 0, &cached_state, GFP_NOFS);
-		ordered = btrfs_lookup_first_ordered_extent(inode,
-							    alloc_end - 1);
-		if (ordered &&
-		    ordered->file_offset + ordered->len > alloc_start &&
-		    ordered->file_offset < alloc_end) {
-			btrfs_put_ordered_extent(ordered);
-			unlock_extent_cached(&BTRFS_I(inode)->io_tree,
-					     alloc_start, locked_end,
-					     &cached_state, GFP_NOFS);
-			/*
-			 * we can't wait on the range with the transaction
-			 * running or with the extent lock held
-			 */
-			btrfs_wait_ordered_range(inode, alloc_start,
-						 alloc_end - alloc_start);
-		} else {
-			if (ordered)
-				btrfs_put_ordered_extent(ordered);
-			break;
-		}
-	}
+	inode->i_fop = &btrfs_file_operations;
+	inode->i_op = &btrfs_file_inode_operations;
 
-	cur_offset = alloc_start;
-	while (1) {
-		em = btrfs_get_extent(inode, NULL, 0, cur_offset,
-				      alloc_end - cur_offset, 0);
-		BUG_ON(IS_ERR(em) || !em);
-		last_byte = min(extent_map_end(em), alloc_end);
-		last_byte = (last_byte + mask) & ~mask;
-		if (em->block_start == EXTENT_MAP_HOLE ||
-		    (cur_offset >= inode->i_size &&
-		     !test_bit(EXTENT_FLAG_PREALLOC, &em->flags))) {
-			ret = prealloc_file_range(inode,
-						  cur_offset, last_byte,
-						alloc_hint, mode, offset+len);
-			if (ret < 0) {
-				free_extent_map(em);
-				break;
-			}
-		}
-		if (em->block_start <= EXTENT_MAP_LAST_BYTE)
-			alloc_hint = em->block_start;
-		free_extent_map(em);
+	inode->i_mapping->a_ops = &btrfs_aops;
+	inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
+	BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
 
-		cur_offset = last_byte;
-		if (cur_offset >= alloc_end) {
-			ret = 0;
-			break;
-		}
-	}
-	unlock_extent_cached(&BTRFS_I(inode)->io_tree, alloc_start, locked_end,
-			     &cached_state, GFP_NOFS);
+	ret = btrfs_orphan_add(trans, inode);
+	if (ret)
+		goto out;
 
-	btrfs_free_reserved_data_space(BTRFS_I(inode)->root, inode,
-				       alloc_end - alloc_start);
-out:
-	mutex_unlock(&inode->i_mutex);
-	return ret;
-}
+	d_tmpfile(dentry, inode);
+	mark_inode_dirty(inode);
 
-static int btrfs_set_page_dirty(struct page *page)
-{
-	return __set_page_dirty_nobuffers(page);
-}
+out:
+	btrfs_end_transaction(trans, root);
+	if (ret)
+		iput(inode);
+	btrfs_balance_delayed_items(root);
+	btrfs_btree_balance_dirty(root);
 
-static int btrfs_permission(struct inode *inode, int mask)
-{
-	if ((BTRFS_I(inode)->flags & BTRFS_INODE_READONLY) && (mask & MAY_WRITE))
-		return -EACCES;
-	return generic_permission(inode, mask, btrfs_check_acl);
+	return ret;
 }
 
 static const struct inode_operations btrfs_dir_inode_operations = {
@@ -5938,16 +9028,23 @@ static const struct inode_operations btrfs_dir_inode_operations = {
 	.listxattr	= btrfs_listxattr,
 	.removexattr	= btrfs_removexattr,
 	.permission	= btrfs_permission,
+	.get_acl	= btrfs_get_acl,
+	.set_acl	= btrfs_set_acl,
+	.update_time	= btrfs_update_time,
+	.tmpfile        = btrfs_tmpfile,
 };
 static const struct inode_operations btrfs_dir_ro_inode_operations = {
 	.lookup		= btrfs_lookup,
 	.permission	= btrfs_permission,
+	.get_acl	= btrfs_get_acl,
+	.set_acl	= btrfs_set_acl,
+	.update_time	= btrfs_update_time,
 };
 
 static const struct file_operations btrfs_dir_file_operations = {
 	.llseek		= generic_file_llseek,
 	.read		= generic_read_dir,
-	.readdir	= btrfs_real_readdir,
+	.iterate	= btrfs_real_readdir,
 	.unlocked_ioctl	= btrfs_ioctl,
 #ifdef CONFIG_COMPAT
 	.compat_ioctl	= btrfs_ioctl,
@@ -5963,7 +9060,6 @@ static struct extent_io_ops btrfs_extent_io_ops = {
 	.readpage_end_io_hook = btrfs_readpage_end_io_hook,
 	.writepage_end_io_hook = btrfs_writepage_end_io_hook,
 	.writepage_start_hook = btrfs_writepage_start_hook,
-	.readpage_io_failed_hook = btrfs_io_failed_hook,
 	.set_bit_hook = btrfs_set_bit_hook,
 	.clear_bit_hook = btrfs_clear_bit_hook,
 	.merge_extent_hook = btrfs_merge_extent_hook,
@@ -5987,7 +9083,6 @@ static const struct address_space_operations btrfs_aops = {
 	.writepage	= btrfs_writepage,
 	.writepages	= btrfs_writepages,
 	.readpages	= btrfs_readpages,
-	.sync_page	= block_sync_page,
 	.direct_IO	= btrfs_direct_IO,
 	.invalidatepage = btrfs_invalidatepage,
 	.releasepage	= btrfs_releasepage,
@@ -6003,7 +9098,6 @@ static const struct address_space_operations btrfs_symlink_aops = {
 };
 
 static const struct inode_operations btrfs_file_inode_operations = {
-	.truncate	= btrfs_truncate,
 	.getattr	= btrfs_getattr,
 	.setattr	= btrfs_setattr,
 	.setxattr	= btrfs_setxattr,
@@ -6011,8 +9105,10 @@ static const struct inode_operations btrfs_file_inode_operations = {
 	.listxattr      = btrfs_listxattr,
 	.removexattr	= btrfs_removexattr,
 	.permission	= btrfs_permission,
-	.fallocate	= btrfs_fallocate,
 	.fiemap		= btrfs_fiemap,
+	.get_acl	= btrfs_get_acl,
+	.set_acl	= btrfs_set_acl,
+	.update_time	= btrfs_update_time,
 };
 static const struct inode_operations btrfs_special_inode_operations = {
 	.getattr	= btrfs_getattr,
@@ -6022,18 +9118,25 @@ static const struct inode_operations btrfs_special_inode_operations = {
 	.getxattr	= btrfs_getxattr,
 	.listxattr	= btrfs_listxattr,
 	.removexattr	= btrfs_removexattr,
+	.get_acl	= btrfs_get_acl,
+	.set_acl	= btrfs_set_acl,
+	.update_time	= btrfs_update_time,
 };
 static const struct inode_operations btrfs_symlink_inode_operations = {
 	.readlink	= generic_readlink,
 	.follow_link	= page_follow_link_light,
 	.put_link	= page_put_link,
+	.getattr	= btrfs_getattr,
+	.setattr	= btrfs_setattr,
 	.permission	= btrfs_permission,
 	.setxattr	= btrfs_setxattr,
 	.getxattr	= btrfs_getxattr,
 	.listxattr	= btrfs_listxattr,
 	.removexattr	= btrfs_removexattr,
+	.update_time	= btrfs_update_time,
 };
 
 const struct dentry_operations btrfs_dentry_operations = {
 	.d_delete	= btrfs_dentry_delete,
+	.d_release	= btrfs_dentry_release,
 };
diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c
index e84ef60ffe3..47aceb494d1 100644
--- a/fs/btrfs/ioctl.c
+++ b/fs/btrfs/ioctl.c
@@ -40,15 +40,54 @@
 #include <linux/xattr.h>
 #include <linux/vmalloc.h>
 #include <linux/slab.h>
-#include "compat.h"
+#include <linux/blkdev.h>
+#include <linux/uuid.h>
+#include <linux/btrfs.h>
+#include <linux/uaccess.h>
 #include "ctree.h"
 #include "disk-io.h"
 #include "transaction.h"
 #include "btrfs_inode.h"
-#include "ioctl.h"
 #include "print-tree.h"
 #include "volumes.h"
 #include "locking.h"
+#include "inode-map.h"
+#include "backref.h"
+#include "rcu-string.h"
+#include "send.h"
+#include "dev-replace.h"
+#include "props.h"
+#include "sysfs.h"
+#include "qgroup.h"
+
+#ifdef CONFIG_64BIT
+/* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
+ * structures are incorrect, as the timespec structure from userspace
+ * is 4 bytes too small. We define these alternatives here to teach
+ * the kernel about the 32-bit struct packing.
+ */
+struct btrfs_ioctl_timespec_32 {
+	__u64 sec;
+	__u32 nsec;
+} __attribute__ ((__packed__));
+
+struct btrfs_ioctl_received_subvol_args_32 {
+	char	uuid[BTRFS_UUID_SIZE];	/* in */
+	__u64	stransid;		/* in */
+	__u64	rtransid;		/* out */
+	struct btrfs_ioctl_timespec_32 stime; /* in */
+	struct btrfs_ioctl_timespec_32 rtime; /* out */
+	__u64	flags;			/* in */
+	__u64	reserved[16];		/* in */
+} __attribute__ ((__packed__));
+
+#define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
+				struct btrfs_ioctl_received_subvol_args_32)
+#endif
+
+
+static int btrfs_clone(struct inode *src, struct inode *inode,
+		       u64 off, u64 olen, u64 olen_aligned, u64 destoff);
 
 /* Mask out flags that are inappropriate for the given type of inode. */
 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
@@ -80,6 +119,13 @@ static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
 		iflags |= FS_NOATIME_FL;
 	if (flags & BTRFS_INODE_DIRSYNC)
 		iflags |= FS_DIRSYNC_FL;
+	if (flags & BTRFS_INODE_NODATACOW)
+		iflags |= FS_NOCOW_FL;
+
+	if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
+		iflags |= FS_COMPR_FL;
+	else if (flags & BTRFS_INODE_NOCOMPRESS)
+		iflags |= FS_NOCOMP_FL;
 
 	return iflags;
 }
@@ -90,25 +136,28 @@ static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
 void btrfs_update_iflags(struct inode *inode)
 {
 	struct btrfs_inode *ip = BTRFS_I(inode);
-
-	inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
+	unsigned int new_fl = 0;
 
 	if (ip->flags & BTRFS_INODE_SYNC)
-		inode->i_flags |= S_SYNC;
+		new_fl |= S_SYNC;
 	if (ip->flags & BTRFS_INODE_IMMUTABLE)
-		inode->i_flags |= S_IMMUTABLE;
+		new_fl |= S_IMMUTABLE;
 	if (ip->flags & BTRFS_INODE_APPEND)
-		inode->i_flags |= S_APPEND;
+		new_fl |= S_APPEND;
 	if (ip->flags & BTRFS_INODE_NOATIME)
-		inode->i_flags |= S_NOATIME;
+		new_fl |= S_NOATIME;
 	if (ip->flags & BTRFS_INODE_DIRSYNC)
-		inode->i_flags |= S_DIRSYNC;
+		new_fl |= S_DIRSYNC;
+
+	set_mask_bits(&inode->i_flags,
+		      S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC,
+		      new_fl);
 }
 
 /*
  * Inherit flags from the parent inode.
  *
- * Unlike extN we don't have any flags we don't want to inherit currently.
+ * Currently only the compression flags and the cow flags are inherited.
  */
 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
 {
@@ -119,18 +168,26 @@ void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
 
 	flags = BTRFS_I(dir)->flags;
 
-	if (S_ISREG(inode->i_mode))
-		flags &= ~BTRFS_INODE_DIRSYNC;
-	else if (!S_ISDIR(inode->i_mode))
-		flags &= (BTRFS_INODE_NODUMP | BTRFS_INODE_NOATIME);
+	if (flags & BTRFS_INODE_NOCOMPRESS) {
+		BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
+		BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
+	} else if (flags & BTRFS_INODE_COMPRESS) {
+		BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
+		BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
+	}
+
+	if (flags & BTRFS_INODE_NODATACOW) {
+		BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
+		if (S_ISREG(inode->i_mode))
+			BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
+	}
 
-	BTRFS_I(inode)->flags = flags;
 	btrfs_update_iflags(inode);
 }
 
 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
 {
-	struct btrfs_inode *ip = BTRFS_I(file->f_path.dentry->d_inode);
+	struct btrfs_inode *ip = BTRFS_I(file_inode(file));
 	unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
 
 	if (copy_to_user(arg, &flags, sizeof(flags)))
@@ -138,28 +195,56 @@ static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
 	return 0;
 }
 
+static int check_flags(unsigned int flags)
+{
+	if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
+		      FS_NOATIME_FL | FS_NODUMP_FL | \
+		      FS_SYNC_FL | FS_DIRSYNC_FL | \
+		      FS_NOCOMP_FL | FS_COMPR_FL |
+		      FS_NOCOW_FL))
+		return -EOPNOTSUPP;
+
+	if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
+		return -EINVAL;
+
+	return 0;
+}
+
 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
 {
-	struct inode *inode = file->f_path.dentry->d_inode;
+	struct inode *inode = file_inode(file);
 	struct btrfs_inode *ip = BTRFS_I(inode);
 	struct btrfs_root *root = ip->root;
 	struct btrfs_trans_handle *trans;
 	unsigned int flags, oldflags;
 	int ret;
+	u64 ip_oldflags;
+	unsigned int i_oldflags;
+	umode_t mode;
+
+	if (!inode_owner_or_capable(inode))
+		return -EPERM;
+
+	if (btrfs_root_readonly(root))
+		return -EROFS;
 
 	if (copy_from_user(&flags, arg, sizeof(flags)))
 		return -EFAULT;
 
-	if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
-		      FS_NOATIME_FL | FS_NODUMP_FL | \
-		      FS_SYNC_FL | FS_DIRSYNC_FL))
-		return -EOPNOTSUPP;
+	ret = check_flags(flags);
+	if (ret)
+		return ret;
 
-	if (!is_owner_or_cap(inode))
-		return -EACCES;
+	ret = mnt_want_write_file(file);
+	if (ret)
+		return ret;
 
 	mutex_lock(&inode->i_mutex);
 
+	ip_oldflags = ip->flags;
+	i_oldflags = inode->i_flags;
+	mode = inode->i_mode;
+
 	flags = btrfs_mask_flags(inode->i_mode, flags);
 	oldflags = btrfs_flags_to_ioctl(ip->flags);
 	if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
@@ -169,10 +254,6 @@ static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
 		}
 	}
 
-	ret = mnt_want_write(file->f_path.mnt);
-	if (ret)
-		goto out_unlock;
-
 	if (flags & FS_SYNC_FL)
 		ip->flags |= BTRFS_INODE_SYNC;
 	else
@@ -197,63 +278,202 @@ static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
 		ip->flags |= BTRFS_INODE_DIRSYNC;
 	else
 		ip->flags &= ~BTRFS_INODE_DIRSYNC;
+	if (flags & FS_NOCOW_FL) {
+		if (S_ISREG(mode)) {
+			/*
+			 * It's safe to turn csums off here, no extents exist.
+			 * Otherwise we want the flag to reflect the real COW
+			 * status of the file and will not set it.
+			 */
+			if (inode->i_size == 0)
+				ip->flags |= BTRFS_INODE_NODATACOW
+					   | BTRFS_INODE_NODATASUM;
+		} else {
+			ip->flags |= BTRFS_INODE_NODATACOW;
+		}
+	} else {
+		/*
+		 * Revert back under same assuptions as above
+		 */
+		if (S_ISREG(mode)) {
+			if (inode->i_size == 0)
+				ip->flags &= ~(BTRFS_INODE_NODATACOW
+				             | BTRFS_INODE_NODATASUM);
+		} else {
+			ip->flags &= ~BTRFS_INODE_NODATACOW;
+		}
+	}
 
+	/*
+	 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
+	 * flag may be changed automatically if compression code won't make
+	 * things smaller.
+	 */
+	if (flags & FS_NOCOMP_FL) {
+		ip->flags &= ~BTRFS_INODE_COMPRESS;
+		ip->flags |= BTRFS_INODE_NOCOMPRESS;
+
+		ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
+		if (ret && ret != -ENODATA)
+			goto out_drop;
+	} else if (flags & FS_COMPR_FL) {
+		const char *comp;
+
+		ip->flags |= BTRFS_INODE_COMPRESS;
+		ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
+
+		if (root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
+			comp = "lzo";
+		else
+			comp = "zlib";
+		ret = btrfs_set_prop(inode, "btrfs.compression",
+				     comp, strlen(comp), 0);
+		if (ret)
+			goto out_drop;
 
-	trans = btrfs_join_transaction(root, 1);
-	BUG_ON(!trans);
+	} else {
+		ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
+	}
 
-	ret = btrfs_update_inode(trans, root, inode);
-	BUG_ON(ret);
+	trans = btrfs_start_transaction(root, 1);
+	if (IS_ERR(trans)) {
+		ret = PTR_ERR(trans);
+		goto out_drop;
+	}
 
 	btrfs_update_iflags(inode);
+	inode_inc_iversion(inode);
 	inode->i_ctime = CURRENT_TIME;
+	ret = btrfs_update_inode(trans, root, inode);
+
 	btrfs_end_transaction(trans, root);
+ out_drop:
+	if (ret) {
+		ip->flags = ip_oldflags;
+		inode->i_flags = i_oldflags;
+	}
 
-	mnt_drop_write(file->f_path.mnt);
  out_unlock:
 	mutex_unlock(&inode->i_mutex);
-	return 0;
+	mnt_drop_write_file(file);
+	return ret;
 }
 
 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
 {
-	struct inode *inode = file->f_path.dentry->d_inode;
+	struct inode *inode = file_inode(file);
 
 	return put_user(inode->i_generation, arg);
 }
 
-static noinline int create_subvol(struct btrfs_root *root,
+static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
+{
+	struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
+	struct btrfs_device *device;
+	struct request_queue *q;
+	struct fstrim_range range;
+	u64 minlen = ULLONG_MAX;
+	u64 num_devices = 0;
+	u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
+	int ret;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return -EPERM;
+
+	rcu_read_lock();
+	list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
+				dev_list) {
+		if (!device->bdev)
+			continue;
+		q = bdev_get_queue(device->bdev);
+		if (blk_queue_discard(q)) {
+			num_devices++;
+			minlen = min((u64)q->limits.discard_granularity,
+				     minlen);
+		}
+	}
+	rcu_read_unlock();
+
+	if (!num_devices)
+		return -EOPNOTSUPP;
+	if (copy_from_user(&range, arg, sizeof(range)))
+		return -EFAULT;
+	if (range.start > total_bytes ||
+	    range.len < fs_info->sb->s_blocksize)
+		return -EINVAL;
+
+	range.len = min(range.len, total_bytes - range.start);
+	range.minlen = max(range.minlen, minlen);
+	ret = btrfs_trim_fs(fs_info->tree_root, &range);
+	if (ret < 0)
+		return ret;
+
+	if (copy_to_user(arg, &range, sizeof(range)))
+		return -EFAULT;
+
+	return 0;
+}
+
+int btrfs_is_empty_uuid(u8 *uuid)
+{
+	int i;
+
+	for (i = 0; i < BTRFS_UUID_SIZE; i++) {
+		if (uuid[i])
+			return 0;
+	}
+	return 1;
+}
+
+static noinline int create_subvol(struct inode *dir,
 				  struct dentry *dentry,
-				  char *name, int namelen)
+				  char *name, int namelen,
+				  u64 *async_transid,
+				  struct btrfs_qgroup_inherit *inherit)
 {
 	struct btrfs_trans_handle *trans;
 	struct btrfs_key key;
 	struct btrfs_root_item root_item;
 	struct btrfs_inode_item *inode_item;
 	struct extent_buffer *leaf;
+	struct btrfs_root *root = BTRFS_I(dir)->root;
 	struct btrfs_root *new_root;
-	struct inode *dir = dentry->d_parent->d_inode;
+	struct btrfs_block_rsv block_rsv;
+	struct timespec cur_time = CURRENT_TIME;
+	struct inode *inode;
 	int ret;
 	int err;
 	u64 objectid;
 	u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
 	u64 index = 0;
+	u64 qgroup_reserved;
+	uuid_le new_uuid;
 
+	ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
+	if (ret)
+		return ret;
+
+	btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
 	/*
-	 * 1 - inode item
-	 * 2 - refs
-	 * 1 - root item
-	 * 2 - dir items
+	 * The same as the snapshot creation, please see the comment
+	 * of create_snapshot().
 	 */
-	ret = btrfs_reserve_metadata_space(root, 6);
+	ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
+					       8, &qgroup_reserved, false);
 	if (ret)
 		return ret;
 
-	trans = btrfs_start_transaction(root, 1);
-	BUG_ON(!trans);
+	trans = btrfs_start_transaction(root, 0);
+	if (IS_ERR(trans)) {
+		ret = PTR_ERR(trans);
+		btrfs_subvolume_release_metadata(root, &block_rsv,
+						 qgroup_reserved);
+		return ret;
+	}
+	trans->block_rsv = &block_rsv;
+	trans->bytes_reserved = block_rsv.size;
 
-	ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
-				       0, &objectid);
+	ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit);
 	if (ret)
 		goto fail;
 
@@ -270,21 +490,25 @@ static noinline int create_subvol(struct btrfs_root *root,
 	btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
 	btrfs_set_header_owner(leaf, objectid);
 
-	write_extent_buffer(leaf, root->fs_info->fsid,
-			    (unsigned long)btrfs_header_fsid(leaf),
+	write_extent_buffer(leaf, root->fs_info->fsid, btrfs_header_fsid(),
 			    BTRFS_FSID_SIZE);
 	write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
-			    (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
+			    btrfs_header_chunk_tree_uuid(leaf),
 			    BTRFS_UUID_SIZE);
 	btrfs_mark_buffer_dirty(leaf);
 
+	memset(&root_item, 0, sizeof(root_item));
+
 	inode_item = &root_item.inode;
-	memset(inode_item, 0, sizeof(*inode_item));
-	inode_item->generation = cpu_to_le64(1);
-	inode_item->size = cpu_to_le64(3);
-	inode_item->nlink = cpu_to_le32(1);
-	inode_item->nbytes = cpu_to_le64(root->leafsize);
-	inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
+	btrfs_set_stack_inode_generation(inode_item, 1);
+	btrfs_set_stack_inode_size(inode_item, 3);
+	btrfs_set_stack_inode_nlink(inode_item, 1);
+	btrfs_set_stack_inode_nbytes(inode_item, root->leafsize);
+	btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
+
+	btrfs_set_root_flags(&root_item, 0);
+	btrfs_set_root_limit(&root_item, 0);
+	btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
 
 	btrfs_set_root_bytenr(&root_item, leaf->start);
 	btrfs_set_root_generation(&root_item, trans->transid);
@@ -293,8 +517,15 @@ static noinline int create_subvol(struct btrfs_root *root,
 	btrfs_set_root_used(&root_item, leaf->len);
 	btrfs_set_root_last_snapshot(&root_item, 0);
 
-	memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
-	root_item.drop_level = 0;
+	btrfs_set_root_generation_v2(&root_item,
+			btrfs_root_generation(&root_item));
+	uuid_le_gen(&new_uuid);
+	memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
+	btrfs_set_stack_timespec_sec(&root_item.otime, cur_time.tv_sec);
+	btrfs_set_stack_timespec_nsec(&root_item.otime, cur_time.tv_nsec);
+	root_item.ctime = root_item.otime;
+	btrfs_set_root_ctransid(&root_item, trans->transid);
+	btrfs_set_root_otransid(&root_item, trans->transid);
 
 	btrfs_tree_unlock(leaf);
 	free_extent_buffer(leaf);
@@ -312,23 +543,37 @@ static noinline int create_subvol(struct btrfs_root *root,
 
 	key.offset = (u64)-1;
 	new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
-	BUG_ON(IS_ERR(new_root));
+	if (IS_ERR(new_root)) {
+		btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
+		ret = PTR_ERR(new_root);
+		goto fail;
+	}
 
 	btrfs_record_root_in_trans(trans, new_root);
 
-	ret = btrfs_create_subvol_root(trans, new_root, new_dirid,
-				       BTRFS_I(dir)->block_group);
+	ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
+	if (ret) {
+		/* We potentially lose an unused inode item here */
+		btrfs_abort_transaction(trans, root, ret);
+		goto fail;
+	}
+
 	/*
 	 * insert the directory item
 	 */
 	ret = btrfs_set_inode_index(dir, &index);
-	BUG_ON(ret);
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
+		goto fail;
+	}
 
 	ret = btrfs_insert_dir_item(trans, root,
-				    name, namelen, dir->i_ino, &key,
+				    name, namelen, dir, &key,
 				    BTRFS_FT_DIR, index);
-	if (ret)
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
 		goto fail;
+	}
 
 	btrfs_i_size_write(dir, dir->i_size + namelen * 2);
 	ret = btrfs_update_inode(trans, root, dir);
@@ -336,78 +581,258 @@ static noinline int create_subvol(struct btrfs_root *root,
 
 	ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
 				 objectid, root->root_key.objectid,
-				 dir->i_ino, index, name, namelen);
-
+				 btrfs_ino(dir), index, name, namelen);
 	BUG_ON(ret);
 
-	d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
+	ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
+				  root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
+				  objectid);
+	if (ret)
+		btrfs_abort_transaction(trans, root, ret);
+
 fail:
-	err = btrfs_commit_transaction(trans, root);
+	trans->block_rsv = NULL;
+	trans->bytes_reserved = 0;
+	btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
+
+	if (async_transid) {
+		*async_transid = trans->transid;
+		err = btrfs_commit_transaction_async(trans, root, 1);
+		if (err)
+			err = btrfs_commit_transaction(trans, root);
+	} else {
+		err = btrfs_commit_transaction(trans, root);
+	}
 	if (err && !ret)
 		ret = err;
 
-	btrfs_unreserve_metadata_space(root, 6);
+	if (!ret) {
+		inode = btrfs_lookup_dentry(dir, dentry);
+		if (IS_ERR(inode))
+			return PTR_ERR(inode);
+		d_instantiate(dentry, inode);
+	}
 	return ret;
 }
 
-static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
-			   char *name, int namelen)
+static void btrfs_wait_nocow_write(struct btrfs_root *root)
+{
+	s64 writers;
+	DEFINE_WAIT(wait);
+
+	do {
+		prepare_to_wait(&root->subv_writers->wait, &wait,
+				TASK_UNINTERRUPTIBLE);
+
+		writers = percpu_counter_sum(&root->subv_writers->counter);
+		if (writers)
+			schedule();
+
+		finish_wait(&root->subv_writers->wait, &wait);
+	} while (writers);
+}
+
+static int create_snapshot(struct btrfs_root *root, struct inode *dir,
+			   struct dentry *dentry, char *name, int namelen,
+			   u64 *async_transid, bool readonly,
+			   struct btrfs_qgroup_inherit *inherit)
 {
 	struct inode *inode;
 	struct btrfs_pending_snapshot *pending_snapshot;
 	struct btrfs_trans_handle *trans;
 	int ret;
 
-	if (!root->ref_cows)
+	if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
 		return -EINVAL;
 
-	/*
-	 * 1 - inode item
-	 * 2 - refs
-	 * 1 - root item
-	 * 2 - dir items
-	 */
-	ret = btrfs_reserve_metadata_space(root, 6);
+	atomic_inc(&root->will_be_snapshoted);
+	smp_mb__after_atomic();
+	btrfs_wait_nocow_write(root);
+
+	ret = btrfs_start_delalloc_inodes(root, 0);
 	if (ret)
-		goto fail;
+		goto out;
+
+	btrfs_wait_ordered_extents(root, -1);
 
 	pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
 	if (!pending_snapshot) {
 		ret = -ENOMEM;
-		btrfs_unreserve_metadata_space(root, 6);
-		goto fail;
-	}
-	pending_snapshot->name = kmalloc(namelen + 1, GFP_NOFS);
-	if (!pending_snapshot->name) {
-		ret = -ENOMEM;
-		kfree(pending_snapshot);
-		btrfs_unreserve_metadata_space(root, 6);
-		goto fail;
+		goto out;
 	}
-	memcpy(pending_snapshot->name, name, namelen);
-	pending_snapshot->name[namelen] = '\0';
+
+	btrfs_init_block_rsv(&pending_snapshot->block_rsv,
+			     BTRFS_BLOCK_RSV_TEMP);
+	/*
+	 * 1 - parent dir inode
+	 * 2 - dir entries
+	 * 1 - root item
+	 * 2 - root ref/backref
+	 * 1 - root of snapshot
+	 * 1 - UUID item
+	 */
+	ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
+					&pending_snapshot->block_rsv, 8,
+					&pending_snapshot->qgroup_reserved,
+					false);
+	if (ret)
+		goto free;
+
 	pending_snapshot->dentry = dentry;
-	trans = btrfs_start_transaction(root, 1);
-	BUG_ON(!trans);
 	pending_snapshot->root = root;
+	pending_snapshot->readonly = readonly;
+	pending_snapshot->dir = dir;
+	pending_snapshot->inherit = inherit;
+
+	trans = btrfs_start_transaction(root, 0);
+	if (IS_ERR(trans)) {
+		ret = PTR_ERR(trans);
+		goto fail;
+	}
+
+	spin_lock(&root->fs_info->trans_lock);
 	list_add(&pending_snapshot->list,
 		 &trans->transaction->pending_snapshots);
-	ret = btrfs_commit_transaction(trans, root);
-	BUG_ON(ret);
-	btrfs_unreserve_metadata_space(root, 6);
+	spin_unlock(&root->fs_info->trans_lock);
+	if (async_transid) {
+		*async_transid = trans->transid;
+		ret = btrfs_commit_transaction_async(trans,
+				     root->fs_info->extent_root, 1);
+		if (ret)
+			ret = btrfs_commit_transaction(trans, root);
+	} else {
+		ret = btrfs_commit_transaction(trans,
+					       root->fs_info->extent_root);
+	}
+	if (ret)
+		goto fail;
+
+	ret = pending_snapshot->error;
+	if (ret)
+		goto fail;
+
+	ret = btrfs_orphan_cleanup(pending_snapshot->snap);
+	if (ret)
+		goto fail;
+
+	/*
+	 * If orphan cleanup did remove any orphans, it means the tree was
+	 * modified and therefore the commit root is not the same as the
+	 * current root anymore. This is a problem, because send uses the
+	 * commit root and therefore can see inode items that don't exist
+	 * in the current root anymore, and for example make calls to
+	 * btrfs_iget, which will do tree lookups based on the current root
+	 * and not on the commit root. Those lookups will fail, returning a
+	 * -ESTALE error, and making send fail with that error. So make sure
+	 * a send does not see any orphans we have just removed, and that it
+	 * will see the same inodes regardless of whether a transaction
+	 * commit happened before it started (meaning that the commit root
+	 * will be the same as the current root) or not.
+	 */
+	if (readonly && pending_snapshot->snap->node !=
+	    pending_snapshot->snap->commit_root) {
+		trans = btrfs_join_transaction(pending_snapshot->snap);
+		if (IS_ERR(trans) && PTR_ERR(trans) != -ENOENT) {
+			ret = PTR_ERR(trans);
+			goto fail;
+		}
+		if (!IS_ERR(trans)) {
+			ret = btrfs_commit_transaction(trans,
+						       pending_snapshot->snap);
+			if (ret)
+				goto fail;
+		}
+	}
 
 	inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
 	if (IS_ERR(inode)) {
 		ret = PTR_ERR(inode);
 		goto fail;
 	}
-	BUG_ON(!inode);
+
 	d_instantiate(dentry, inode);
 	ret = 0;
 fail:
+	btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
+					 &pending_snapshot->block_rsv,
+					 pending_snapshot->qgroup_reserved);
+free:
+	kfree(pending_snapshot);
+out:
+	atomic_dec(&root->will_be_snapshoted);
 	return ret;
 }
 
+/*  copy of check_sticky in fs/namei.c()
+* It's inline, so penalty for filesystems that don't use sticky bit is
+* minimal.
+*/
+static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
+{
+	kuid_t fsuid = current_fsuid();
+
+	if (!(dir->i_mode & S_ISVTX))
+		return 0;
+	if (uid_eq(inode->i_uid, fsuid))
+		return 0;
+	if (uid_eq(dir->i_uid, fsuid))
+		return 0;
+	return !capable(CAP_FOWNER);
+}
+
+/*  copy of may_delete in fs/namei.c()
+ *	Check whether we can remove a link victim from directory dir, check
+ *  whether the type of victim is right.
+ *  1. We can't do it if dir is read-only (done in permission())
+ *  2. We should have write and exec permissions on dir
+ *  3. We can't remove anything from append-only dir
+ *  4. We can't do anything with immutable dir (done in permission())
+ *  5. If the sticky bit on dir is set we should either
+ *	a. be owner of dir, or
+ *	b. be owner of victim, or
+ *	c. have CAP_FOWNER capability
+ *  6. If the victim is append-only or immutable we can't do antyhing with
+ *     links pointing to it.
+ *  7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
+ *  8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
+ *  9. We can't remove a root or mountpoint.
+ * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
+ *     nfs_async_unlink().
+ */
+
+static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
+{
+	int error;
+
+	if (!victim->d_inode)
+		return -ENOENT;
+
+	BUG_ON(victim->d_parent->d_inode != dir);
+	audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
+
+	error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
+	if (error)
+		return error;
+	if (IS_APPEND(dir))
+		return -EPERM;
+	if (btrfs_check_sticky(dir, victim->d_inode)||
+		IS_APPEND(victim->d_inode)||
+	    IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
+		return -EPERM;
+	if (isdir) {
+		if (!S_ISDIR(victim->d_inode->i_mode))
+			return -ENOTDIR;
+		if (IS_ROOT(victim))
+			return -EBUSY;
+	} else if (S_ISDIR(victim->d_inode->i_mode))
+		return -EISDIR;
+	if (IS_DEADDIR(dir))
+		return -ENOENT;
+	if (victim->d_flags & DCACHE_NFSFS_RENAMED)
+		return -EBUSY;
+	return 0;
+}
+
 /* copy of may_create in fs/namei.c() */
 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
 {
@@ -425,13 +850,17 @@ static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
  */
 static noinline int btrfs_mksubvol(struct path *parent,
 				   char *name, int namelen,
-				   struct btrfs_root *snap_src)
+				   struct btrfs_root *snap_src,
+				   u64 *async_transid, bool readonly,
+				   struct btrfs_qgroup_inherit *inherit)
 {
 	struct inode *dir  = parent->dentry->d_inode;
 	struct dentry *dentry;
 	int error;
 
-	mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
+	error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
+	if (error == -EINTR)
+		return error;
 
 	dentry = lookup_one_len(name, parent->dentry, namelen);
 	error = PTR_ERR(dentry);
@@ -442,13 +871,19 @@ static noinline int btrfs_mksubvol(struct path *parent,
 	if (dentry->d_inode)
 		goto out_dput;
 
-	error = mnt_want_write(parent->mnt);
+	error = btrfs_may_create(dir, dentry);
 	if (error)
 		goto out_dput;
 
-	error = btrfs_may_create(dir, dentry);
+	/*
+	 * even if this name doesn't exist, we may get hash collisions.
+	 * check for them now when we can safely fail
+	 */
+	error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
+					       dir->i_ino, name,
+					       namelen);
 	if (error)
-		goto out_drop_write;
+		goto out_dput;
 
 	down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
 
@@ -456,18 +891,16 @@ static noinline int btrfs_mksubvol(struct path *parent,
 		goto out_up_read;
 
 	if (snap_src) {
-		error = create_snapshot(snap_src, dentry,
-					name, namelen);
+		error = create_snapshot(snap_src, dir, dentry, name, namelen,
+					async_transid, readonly, inherit);
 	} else {
-		error = create_subvol(BTRFS_I(dir)->root, dentry,
-				      name, namelen);
+		error = create_subvol(dir, dentry, name, namelen,
+				      async_transid, inherit);
 	}
 	if (!error)
 		fsnotify_mkdir(dir, dentry);
 out_up_read:
 	up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
-out_drop_write:
-	mnt_drop_write(parent->mnt);
 out_dput:
 	dput(dentry);
 out_unlock:
@@ -475,27 +908,115 @@ out_unlock:
 	return error;
 }
 
-static int should_defrag_range(struct inode *inode, u64 start, u64 len,
-			       int thresh, u64 *last_len, u64 *skip,
-			       u64 *defrag_end)
+/*
+ * When we're defragging a range, we don't want to kick it off again
+ * if it is really just waiting for delalloc to send it down.
+ * If we find a nice big extent or delalloc range for the bytes in the
+ * file you want to defrag, we return 0 to let you know to skip this
+ * part of the file
+ */
+static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
 {
 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
 	struct extent_map *em = NULL;
 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
-	int ret = 1;
+	u64 end;
 
+	read_lock(&em_tree->lock);
+	em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
+	read_unlock(&em_tree->lock);
 
-	if (thresh == 0)
-		thresh = 256 * 1024;
+	if (em) {
+		end = extent_map_end(em);
+		free_extent_map(em);
+		if (end - offset > thresh)
+			return 0;
+	}
+	/* if we already have a nice delalloc here, just stop */
+	thresh /= 2;
+	end = count_range_bits(io_tree, &offset, offset + thresh,
+			       thresh, EXTENT_DELALLOC, 1);
+	if (end >= thresh)
+		return 0;
+	return 1;
+}
 
-	/*
-	 * make sure that once we start defragging and extent, we keep on
-	 * defragging it
-	 */
-	if (start < *defrag_end)
-		return 1;
+/*
+ * helper function to walk through a file and find extents
+ * newer than a specific transid, and smaller than thresh.
+ *
+ * This is used by the defragging code to find new and small
+ * extents
+ */
+static int find_new_extents(struct btrfs_root *root,
+			    struct inode *inode, u64 newer_than,
+			    u64 *off, int thresh)
+{
+	struct btrfs_path *path;
+	struct btrfs_key min_key;
+	struct extent_buffer *leaf;
+	struct btrfs_file_extent_item *extent;
+	int type;
+	int ret;
+	u64 ino = btrfs_ino(inode);
 
-	*skip = 0;
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	min_key.objectid = ino;
+	min_key.type = BTRFS_EXTENT_DATA_KEY;
+	min_key.offset = *off;
+
+	while (1) {
+		path->keep_locks = 1;
+		ret = btrfs_search_forward(root, &min_key, path, newer_than);
+		if (ret != 0)
+			goto none;
+		path->keep_locks = 0;
+		btrfs_unlock_up_safe(path, 1);
+process_slot:
+		if (min_key.objectid != ino)
+			goto none;
+		if (min_key.type != BTRFS_EXTENT_DATA_KEY)
+			goto none;
+
+		leaf = path->nodes[0];
+		extent = btrfs_item_ptr(leaf, path->slots[0],
+					struct btrfs_file_extent_item);
+
+		type = btrfs_file_extent_type(leaf, extent);
+		if (type == BTRFS_FILE_EXTENT_REG &&
+		    btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
+		    check_defrag_in_cache(inode, min_key.offset, thresh)) {
+			*off = min_key.offset;
+			btrfs_free_path(path);
+			return 0;
+		}
+
+		path->slots[0]++;
+		if (path->slots[0] < btrfs_header_nritems(leaf)) {
+			btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
+			goto process_slot;
+		}
+
+		if (min_key.offset == (u64)-1)
+			goto none;
+
+		min_key.offset++;
+		btrfs_release_path(path);
+	}
+none:
+	btrfs_free_path(path);
+	return -ENOENT;
+}
+
+static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
+{
+	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
+	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
+	struct extent_map *em;
+	u64 len = PAGE_CACHE_SIZE;
 
 	/*
 	 * hopefully we have this extent in the tree already, try without
@@ -506,25 +1027,76 @@ static int should_defrag_range(struct inode *inode, u64 start, u64 len,
 	read_unlock(&em_tree->lock);
 
 	if (!em) {
+		struct extent_state *cached = NULL;
+		u64 end = start + len - 1;
+
 		/* get the big lock and read metadata off disk */
-		lock_extent(io_tree, start, start + len - 1, GFP_NOFS);
+		lock_extent_bits(io_tree, start, end, 0, &cached);
 		em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
-		unlock_extent(io_tree, start, start + len - 1, GFP_NOFS);
+		unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS);
 
 		if (IS_ERR(em))
-			return 0;
+			return NULL;
 	}
 
+	return em;
+}
+
+static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
+{
+	struct extent_map *next;
+	bool ret = true;
+
+	/* this is the last extent */
+	if (em->start + em->len >= i_size_read(inode))
+		return false;
+
+	next = defrag_lookup_extent(inode, em->start + em->len);
+	if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE ||
+	    (em->block_start + em->block_len == next->block_start))
+		ret = false;
+
+	free_extent_map(next);
+	return ret;
+}
+
+static int should_defrag_range(struct inode *inode, u64 start, int thresh,
+			       u64 *last_len, u64 *skip, u64 *defrag_end,
+			       int compress)
+{
+	struct extent_map *em;
+	int ret = 1;
+	bool next_mergeable = true;
+
+	/*
+	 * make sure that once we start defragging an extent, we keep on
+	 * defragging it
+	 */
+	if (start < *defrag_end)
+		return 1;
+
+	*skip = 0;
+
+	em = defrag_lookup_extent(inode, start);
+	if (!em)
+		return 0;
+
 	/* this will cover holes, and inline extents */
-	if (em->block_start >= EXTENT_MAP_LAST_BYTE)
+	if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
 		ret = 0;
+		goto out;
+	}
+
+	next_mergeable = defrag_check_next_extent(inode, em);
 
 	/*
-	 * we hit a real extent, if it is big don't bother defragging it again
+	 * we hit a real extent, if it is big or the next extent is not a
+	 * real extent, don't bother defragging it
 	 */
-	if ((*last_len == 0 || *last_len >= thresh) && em->len >= thresh)
+	if (!compress && (*last_len == 0 || *last_len >= thresh) &&
+	    (em->len >= thresh || !next_mergeable))
 		ret = 0;
-
+out:
 	/*
 	 * last_len ends up being a counter of how many bytes we've defragged.
 	 * every time we choose not to defrag an extent, we reset *last_len
@@ -534,7 +1106,6 @@ static int should_defrag_range(struct inode *inode, u64 start, u64 len,
 	 * extent will force at least part of that big extent to be defragged.
 	 */
 	if (ret) {
-		*last_len += len;
 		*defrag_end = extent_map_end(em);
 	} else {
 		*last_len = 0;
@@ -546,84 +1117,84 @@ static int should_defrag_range(struct inode *inode, u64 start, u64 len,
 	return ret;
 }
 
-static int btrfs_defrag_file(struct file *file,
-			     struct btrfs_ioctl_defrag_range_args *range)
+/*
+ * it doesn't do much good to defrag one or two pages
+ * at a time.  This pulls in a nice chunk of pages
+ * to COW and defrag.
+ *
+ * It also makes sure the delalloc code has enough
+ * dirty data to avoid making new small extents as part
+ * of the defrag
+ *
+ * It's a good idea to start RA on this range
+ * before calling this.
+ */
+static int cluster_pages_for_defrag(struct inode *inode,
+				    struct page **pages,
+				    unsigned long start_index,
+				    unsigned long num_pages)
 {
-	struct inode *inode = fdentry(file)->d_inode;
-	struct btrfs_root *root = BTRFS_I(inode)->root;
-	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
-	struct btrfs_ordered_extent *ordered;
-	struct page *page;
-	unsigned long last_index;
-	unsigned long ra_pages = root->fs_info->bdi.ra_pages;
-	unsigned long total_read = 0;
+	unsigned long file_end;
+	u64 isize = i_size_read(inode);
 	u64 page_start;
 	u64 page_end;
-	u64 last_len = 0;
-	u64 skip = 0;
-	u64 defrag_end = 0;
-	unsigned long i;
+	u64 page_cnt;
 	int ret;
+	int i;
+	int i_done;
+	struct btrfs_ordered_extent *ordered;
+	struct extent_state *cached_state = NULL;
+	struct extent_io_tree *tree;
+	gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
 
-	if (inode->i_size == 0)
+	file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
+	if (!isize || start_index > file_end)
 		return 0;
 
-	if (range->start + range->len > range->start) {
-		last_index = min_t(u64, inode->i_size - 1,
-			 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
-	} else {
-		last_index = (inode->i_size - 1) >> PAGE_CACHE_SHIFT;
-	}
+	page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
 
-	i = range->start >> PAGE_CACHE_SHIFT;
-	while (i <= last_index) {
-		if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
-					PAGE_CACHE_SIZE,
-					range->extent_thresh,
-					&last_len, &skip,
-					&defrag_end)) {
-			unsigned long next;
-			/*
-			 * the should_defrag function tells us how much to skip
-			 * bump our counter by the suggested amount
-			 */
-			next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
-			i = max(i + 1, next);
-			continue;
-		}
-
-		if (total_read % ra_pages == 0) {
-			btrfs_force_ra(inode->i_mapping, &file->f_ra, file, i,
-				       min(last_index, i + ra_pages - 1));
-		}
-		total_read++;
-		mutex_lock(&inode->i_mutex);
-		if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
-			BTRFS_I(inode)->force_compress = 1;
+	ret = btrfs_delalloc_reserve_space(inode,
+					   page_cnt << PAGE_CACHE_SHIFT);
+	if (ret)
+		return ret;
+	i_done = 0;
+	tree = &BTRFS_I(inode)->io_tree;
 
-		ret = btrfs_check_data_free_space(root, inode, PAGE_CACHE_SIZE);
-		if (ret) {
-			ret = -ENOSPC;
+	/* step one, lock all the pages */
+	for (i = 0; i < page_cnt; i++) {
+		struct page *page;
+again:
+		page = find_or_create_page(inode->i_mapping,
+					   start_index + i, mask);
+		if (!page)
 			break;
-		}
 
-		ret = btrfs_reserve_metadata_for_delalloc(root, inode, 1);
-		if (ret) {
-			btrfs_free_reserved_data_space(root, inode,
-						       PAGE_CACHE_SIZE);
-			ret = -ENOSPC;
-			break;
-		}
-again:
-		if (inode->i_size == 0 ||
-		    i > ((inode->i_size - 1) >> PAGE_CACHE_SHIFT)) {
-			ret = 0;
-			goto err_reservations;
-		}
+		page_start = page_offset(page);
+		page_end = page_start + PAGE_CACHE_SIZE - 1;
+		while (1) {
+			lock_extent_bits(tree, page_start, page_end,
+					 0, &cached_state);
+			ordered = btrfs_lookup_ordered_extent(inode,
+							      page_start);
+			unlock_extent_cached(tree, page_start, page_end,
+					     &cached_state, GFP_NOFS);
+			if (!ordered)
+				break;
 
-		page = grab_cache_page(inode->i_mapping, i);
-		if (!page)
-			goto err_reservations;
+			unlock_page(page);
+			btrfs_start_ordered_extent(inode, ordered, 1);
+			btrfs_put_ordered_extent(ordered);
+			lock_page(page);
+			/*
+			 * we unlocked the page above, so we need check if
+			 * it was released or not.
+			 */
+			if (page->mapping != inode->i_mapping) {
+				unlock_page(page);
+				page_cache_release(page);
+				goto again;
+			}
+		}
 
 		if (!PageUptodate(page)) {
 			btrfs_readpage(NULL, page);
@@ -631,7 +1202,8 @@ again:
 			if (!PageUptodate(page)) {
 				unlock_page(page);
 				page_cache_release(page);
-				goto err_reservations;
+				ret = -EIO;
+				break;
 			}
 		}
 
@@ -641,55 +1213,253 @@ again:
 			goto again;
 		}
 
-		wait_on_page_writeback(page);
+		pages[i] = page;
+		i_done++;
+	}
+	if (!i_done || ret)
+		goto out;
 
-		if (PageDirty(page)) {
-			btrfs_free_reserved_data_space(root, inode,
-						       PAGE_CACHE_SIZE);
-			goto loop_unlock;
-		}
+	if (!(inode->i_sb->s_flags & MS_ACTIVE))
+		goto out;
 
-		page_start = (u64)page->index << PAGE_CACHE_SHIFT;
-		page_end = page_start + PAGE_CACHE_SIZE - 1;
-		lock_extent(io_tree, page_start, page_end, GFP_NOFS);
+	/*
+	 * so now we have a nice long stream of locked
+	 * and up to date pages, lets wait on them
+	 */
+	for (i = 0; i < i_done; i++)
+		wait_on_page_writeback(pages[i]);
+
+	page_start = page_offset(pages[0]);
+	page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
+
+	lock_extent_bits(&BTRFS_I(inode)->io_tree,
+			 page_start, page_end - 1, 0, &cached_state);
+	clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
+			  page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
+			  EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
+			  &cached_state, GFP_NOFS);
+
+	if (i_done != page_cnt) {
+		spin_lock(&BTRFS_I(inode)->lock);
+		BTRFS_I(inode)->outstanding_extents++;
+		spin_unlock(&BTRFS_I(inode)->lock);
+		btrfs_delalloc_release_space(inode,
+				     (page_cnt - i_done) << PAGE_CACHE_SHIFT);
+	}
 
-		ordered = btrfs_lookup_ordered_extent(inode, page_start);
-		if (ordered) {
-			unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
-			unlock_page(page);
-			page_cache_release(page);
-			btrfs_start_ordered_extent(inode, ordered, 1);
-			btrfs_put_ordered_extent(ordered);
-			goto again;
-		}
-		set_page_extent_mapped(page);
 
+	set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
+			  &cached_state, GFP_NOFS);
+
+	unlock_extent_cached(&BTRFS_I(inode)->io_tree,
+			     page_start, page_end - 1, &cached_state,
+			     GFP_NOFS);
+
+	for (i = 0; i < i_done; i++) {
+		clear_page_dirty_for_io(pages[i]);
+		ClearPageChecked(pages[i]);
+		set_page_extent_mapped(pages[i]);
+		set_page_dirty(pages[i]);
+		unlock_page(pages[i]);
+		page_cache_release(pages[i]);
+	}
+	return i_done;
+out:
+	for (i = 0; i < i_done; i++) {
+		unlock_page(pages[i]);
+		page_cache_release(pages[i]);
+	}
+	btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
+	return ret;
+
+}
+
+int btrfs_defrag_file(struct inode *inode, struct file *file,
+		      struct btrfs_ioctl_defrag_range_args *range,
+		      u64 newer_than, unsigned long max_to_defrag)
+{
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct file_ra_state *ra = NULL;
+	unsigned long last_index;
+	u64 isize = i_size_read(inode);
+	u64 last_len = 0;
+	u64 skip = 0;
+	u64 defrag_end = 0;
+	u64 newer_off = range->start;
+	unsigned long i;
+	unsigned long ra_index = 0;
+	int ret;
+	int defrag_count = 0;
+	int compress_type = BTRFS_COMPRESS_ZLIB;
+	int extent_thresh = range->extent_thresh;
+	unsigned long max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
+	unsigned long cluster = max_cluster;
+	u64 new_align = ~((u64)128 * 1024 - 1);
+	struct page **pages = NULL;
+
+	if (isize == 0)
+		return 0;
+
+	if (range->start >= isize)
+		return -EINVAL;
+
+	if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
+		if (range->compress_type > BTRFS_COMPRESS_TYPES)
+			return -EINVAL;
+		if (range->compress_type)
+			compress_type = range->compress_type;
+	}
+
+	if (extent_thresh == 0)
+		extent_thresh = 256 * 1024;
+
+	/*
+	 * if we were not given a file, allocate a readahead
+	 * context
+	 */
+	if (!file) {
+		ra = kzalloc(sizeof(*ra), GFP_NOFS);
+		if (!ra)
+			return -ENOMEM;
+		file_ra_state_init(ra, inode->i_mapping);
+	} else {
+		ra = &file->f_ra;
+	}
+
+	pages = kmalloc_array(max_cluster, sizeof(struct page *),
+			GFP_NOFS);
+	if (!pages) {
+		ret = -ENOMEM;
+		goto out_ra;
+	}
+
+	/* find the last page to defrag */
+	if (range->start + range->len > range->start) {
+		last_index = min_t(u64, isize - 1,
+			 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
+	} else {
+		last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
+	}
+
+	if (newer_than) {
+		ret = find_new_extents(root, inode, newer_than,
+				       &newer_off, 64 * 1024);
+		if (!ret) {
+			range->start = newer_off;
+			/*
+			 * we always align our defrag to help keep
+			 * the extents in the file evenly spaced
+			 */
+			i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
+		} else
+			goto out_ra;
+	} else {
+		i = range->start >> PAGE_CACHE_SHIFT;
+	}
+	if (!max_to_defrag)
+		max_to_defrag = last_index + 1;
+
+	/*
+	 * make writeback starts from i, so the defrag range can be
+	 * written sequentially.
+	 */
+	if (i < inode->i_mapping->writeback_index)
+		inode->i_mapping->writeback_index = i;
+
+	while (i <= last_index && defrag_count < max_to_defrag &&
+	       (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
+		PAGE_CACHE_SHIFT)) {
 		/*
-		 * this makes sure page_mkwrite is called on the
-		 * page if it is dirtied again later
+		 * make sure we stop running if someone unmounts
+		 * the FS
 		 */
-		clear_page_dirty_for_io(page);
-		clear_extent_bits(&BTRFS_I(inode)->io_tree, page_start,
-				  page_end, EXTENT_DIRTY | EXTENT_DELALLOC |
-				  EXTENT_DO_ACCOUNTING, GFP_NOFS);
-
-		btrfs_set_extent_delalloc(inode, page_start, page_end, NULL);
-		ClearPageChecked(page);
-		set_page_dirty(page);
-		unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
-
-loop_unlock:
-		unlock_page(page);
-		page_cache_release(page);
+		if (!(inode->i_sb->s_flags & MS_ACTIVE))
+			break;
+
+		if (btrfs_defrag_cancelled(root->fs_info)) {
+			printk(KERN_DEBUG "BTRFS: defrag_file cancelled\n");
+			ret = -EAGAIN;
+			break;
+		}
+
+		if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
+					 extent_thresh, &last_len, &skip,
+					 &defrag_end, range->flags &
+					 BTRFS_DEFRAG_RANGE_COMPRESS)) {
+			unsigned long next;
+			/*
+			 * the should_defrag function tells us how much to skip
+			 * bump our counter by the suggested amount
+			 */
+			next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
+			i = max(i + 1, next);
+			continue;
+		}
+
+		if (!newer_than) {
+			cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
+				   PAGE_CACHE_SHIFT) - i;
+			cluster = min(cluster, max_cluster);
+		} else {
+			cluster = max_cluster;
+		}
+
+		if (i + cluster > ra_index) {
+			ra_index = max(i, ra_index);
+			btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
+				       cluster);
+			ra_index += max_cluster;
+		}
+
+		mutex_lock(&inode->i_mutex);
+		if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
+			BTRFS_I(inode)->force_compress = compress_type;
+		ret = cluster_pages_for_defrag(inode, pages, i, cluster);
+		if (ret < 0) {
+			mutex_unlock(&inode->i_mutex);
+			goto out_ra;
+		}
+
+		defrag_count += ret;
+		balance_dirty_pages_ratelimited(inode->i_mapping);
 		mutex_unlock(&inode->i_mutex);
 
-		btrfs_unreserve_metadata_for_delalloc(root, inode, 1);
-		balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1);
-		i++;
+		if (newer_than) {
+			if (newer_off == (u64)-1)
+				break;
+
+			if (ret > 0)
+				i += ret;
+
+			newer_off = max(newer_off + 1,
+					(u64)i << PAGE_CACHE_SHIFT);
+
+			ret = find_new_extents(root, inode,
+					       newer_than, &newer_off,
+					       64 * 1024);
+			if (!ret) {
+				range->start = newer_off;
+				i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
+			} else {
+				break;
+			}
+		} else {
+			if (ret > 0) {
+				i += ret;
+				last_len += ret << PAGE_CACHE_SHIFT;
+			} else {
+				i++;
+				last_len = 0;
+			}
+		}
 	}
 
-	if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
+	if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
 		filemap_flush(inode->i_mapping);
+		if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
+			     &BTRFS_I(inode)->runtime_flags))
+			filemap_flush(inode->i_mapping);
+	}
 
 	if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
 		/* the filemap_flush will queue IO into the worker threads, but
@@ -704,68 +1474,96 @@ loop_unlock:
 			    atomic_read(&root->fs_info->async_delalloc_pages) == 0));
 		}
 		atomic_dec(&root->fs_info->async_submit_draining);
+	}
 
-		mutex_lock(&inode->i_mutex);
-		BTRFS_I(inode)->force_compress = 0;
-		mutex_unlock(&inode->i_mutex);
+	if (range->compress_type == BTRFS_COMPRESS_LZO) {
+		btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
 	}
 
-	return 0;
+	ret = defrag_count;
 
-err_reservations:
-	mutex_unlock(&inode->i_mutex);
-	btrfs_free_reserved_data_space(root, inode, PAGE_CACHE_SIZE);
-	btrfs_unreserve_metadata_for_delalloc(root, inode, 1);
+out_ra:
+	if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
+		mutex_lock(&inode->i_mutex);
+		BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
+		mutex_unlock(&inode->i_mutex);
+	}
+	if (!file)
+		kfree(ra);
+	kfree(pages);
 	return ret;
 }
 
-static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
+static noinline int btrfs_ioctl_resize(struct file *file,
 					void __user *arg)
 {
 	u64 new_size;
 	u64 old_size;
 	u64 devid = 1;
+	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
 	struct btrfs_ioctl_vol_args *vol_args;
 	struct btrfs_trans_handle *trans;
 	struct btrfs_device *device = NULL;
 	char *sizestr;
+	char *retptr;
 	char *devstr = NULL;
 	int ret = 0;
-	int namelen;
 	int mod = 0;
 
-	if (root->fs_info->sb->s_flags & MS_RDONLY)
-		return -EROFS;
-
 	if (!capable(CAP_SYS_ADMIN))
 		return -EPERM;
 
+	ret = mnt_want_write_file(file);
+	if (ret)
+		return ret;
+
+	if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
+			1)) {
+		mnt_drop_write_file(file);
+		return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
+	}
+
+	mutex_lock(&root->fs_info->volume_mutex);
 	vol_args = memdup_user(arg, sizeof(*vol_args));
-	if (IS_ERR(vol_args))
-		return PTR_ERR(vol_args);
+	if (IS_ERR(vol_args)) {
+		ret = PTR_ERR(vol_args);
+		goto out;
+	}
 
 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
-	namelen = strlen(vol_args->name);
 
-	mutex_lock(&root->fs_info->volume_mutex);
 	sizestr = vol_args->name;
 	devstr = strchr(sizestr, ':');
 	if (devstr) {
-		char *end;
 		sizestr = devstr + 1;
 		*devstr = '\0';
 		devstr = vol_args->name;
-		devid = simple_strtoull(devstr, &end, 10);
-		printk(KERN_INFO "resizing devid %llu\n",
-		       (unsigned long long)devid);
+		ret = kstrtoull(devstr, 10, &devid);
+		if (ret)
+			goto out_free;
+		if (!devid) {
+			ret = -EINVAL;
+			goto out_free;
+		}
+		btrfs_info(root->fs_info, "resizing devid %llu", devid);
 	}
-	device = btrfs_find_device(root, devid, NULL, NULL);
+
+	device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
 	if (!device) {
-		printk(KERN_INFO "resizer unable to find device %llu\n",
-		       (unsigned long long)devid);
-		ret = -EINVAL;
-		goto out_unlock;
+		btrfs_info(root->fs_info, "resizer unable to find device %llu",
+		       devid);
+		ret = -ENODEV;
+		goto out_free;
+	}
+
+	if (!device->writeable) {
+		btrfs_info(root->fs_info,
+			   "resizer unable to apply on readonly device %llu",
+		       devid);
+		ret = -EPERM;
+		goto out_free;
 	}
+
 	if (!strcmp(sizestr, "max"))
 		new_size = device->bdev->bd_inode->i_size;
 	else {
@@ -776,102 +1574,308 @@ static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
 			mod = 1;
 			sizestr++;
 		}
-		new_size = memparse(sizestr, NULL);
-		if (new_size == 0) {
+		new_size = memparse(sizestr, &retptr);
+		if (*retptr != '\0' || new_size == 0) {
 			ret = -EINVAL;
-			goto out_unlock;
+			goto out_free;
 		}
 	}
 
+	if (device->is_tgtdev_for_dev_replace) {
+		ret = -EPERM;
+		goto out_free;
+	}
+
 	old_size = device->total_bytes;
 
 	if (mod < 0) {
 		if (new_size > old_size) {
 			ret = -EINVAL;
-			goto out_unlock;
+			goto out_free;
 		}
 		new_size = old_size - new_size;
 	} else if (mod > 0) {
+		if (new_size > ULLONG_MAX - old_size) {
+			ret = -ERANGE;
+			goto out_free;
+		}
 		new_size = old_size + new_size;
 	}
 
 	if (new_size < 256 * 1024 * 1024) {
 		ret = -EINVAL;
-		goto out_unlock;
+		goto out_free;
 	}
 	if (new_size > device->bdev->bd_inode->i_size) {
 		ret = -EFBIG;
-		goto out_unlock;
+		goto out_free;
 	}
 
 	do_div(new_size, root->sectorsize);
 	new_size *= root->sectorsize;
 
-	printk(KERN_INFO "new size for %s is %llu\n",
-		device->name, (unsigned long long)new_size);
+	printk_in_rcu(KERN_INFO "BTRFS: new size for %s is %llu\n",
+		      rcu_str_deref(device->name), new_size);
 
 	if (new_size > old_size) {
-		trans = btrfs_start_transaction(root, 1);
+		trans = btrfs_start_transaction(root, 0);
+		if (IS_ERR(trans)) {
+			ret = PTR_ERR(trans);
+			goto out_free;
+		}
 		ret = btrfs_grow_device(trans, device, new_size);
 		btrfs_commit_transaction(trans, root);
-	} else {
+	} else if (new_size < old_size) {
 		ret = btrfs_shrink_device(device, new_size);
-	}
+	} /* equal, nothing need to do */
 
-out_unlock:
-	mutex_unlock(&root->fs_info->volume_mutex);
+out_free:
 	kfree(vol_args);
+out:
+	mutex_unlock(&root->fs_info->volume_mutex);
+	atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
+	mnt_drop_write_file(file);
+	return ret;
+}
+
+static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
+				char *name, unsigned long fd, int subvol,
+				u64 *transid, bool readonly,
+				struct btrfs_qgroup_inherit *inherit)
+{
+	int namelen;
+	int ret = 0;
+
+	ret = mnt_want_write_file(file);
+	if (ret)
+		goto out;
+
+	namelen = strlen(name);
+	if (strchr(name, '/')) {
+		ret = -EINVAL;
+		goto out_drop_write;
+	}
+
+	if (name[0] == '.' &&
+	   (namelen == 1 || (name[1] == '.' && namelen == 2))) {
+		ret = -EEXIST;
+		goto out_drop_write;
+	}
+
+	if (subvol) {
+		ret = btrfs_mksubvol(&file->f_path, name, namelen,
+				     NULL, transid, readonly, inherit);
+	} else {
+		struct fd src = fdget(fd);
+		struct inode *src_inode;
+		if (!src.file) {
+			ret = -EINVAL;
+			goto out_drop_write;
+		}
+
+		src_inode = file_inode(src.file);
+		if (src_inode->i_sb != file_inode(file)->i_sb) {
+			btrfs_info(BTRFS_I(src_inode)->root->fs_info,
+				   "Snapshot src from another FS");
+			ret = -EXDEV;
+		} else if (!inode_owner_or_capable(src_inode)) {
+			/*
+			 * Subvolume creation is not restricted, but snapshots
+			 * are limited to own subvolumes only
+			 */
+			ret = -EPERM;
+		} else {
+			ret = btrfs_mksubvol(&file->f_path, name, namelen,
+					     BTRFS_I(src_inode)->root,
+					     transid, readonly, inherit);
+		}
+		fdput(src);
+	}
+out_drop_write:
+	mnt_drop_write_file(file);
+out:
 	return ret;
 }
 
 static noinline int btrfs_ioctl_snap_create(struct file *file,
 					    void __user *arg, int subvol)
 {
-	struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
 	struct btrfs_ioctl_vol_args *vol_args;
-	struct file *src_file;
-	int namelen;
-	int ret = 0;
+	int ret;
 
-	if (root->fs_info->sb->s_flags & MS_RDONLY)
-		return -EROFS;
+	vol_args = memdup_user(arg, sizeof(*vol_args));
+	if (IS_ERR(vol_args))
+		return PTR_ERR(vol_args);
+	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
+
+	ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
+					      vol_args->fd, subvol,
+					      NULL, false, NULL);
+
+	kfree(vol_args);
+	return ret;
+}
+
+static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
+					       void __user *arg, int subvol)
+{
+	struct btrfs_ioctl_vol_args_v2 *vol_args;
+	int ret;
+	u64 transid = 0;
+	u64 *ptr = NULL;
+	bool readonly = false;
+	struct btrfs_qgroup_inherit *inherit = NULL;
 
 	vol_args = memdup_user(arg, sizeof(*vol_args));
 	if (IS_ERR(vol_args))
 		return PTR_ERR(vol_args);
+	vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
 
-	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
-	namelen = strlen(vol_args->name);
-	if (strchr(vol_args->name, '/')) {
-		ret = -EINVAL;
+	if (vol_args->flags &
+	    ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
+	      BTRFS_SUBVOL_QGROUP_INHERIT)) {
+		ret = -EOPNOTSUPP;
 		goto out;
 	}
 
-	if (subvol) {
-		ret = btrfs_mksubvol(&file->f_path, vol_args->name, namelen,
-				     NULL);
-	} else {
-		struct inode *src_inode;
-		src_file = fget(vol_args->fd);
-		if (!src_file) {
+	if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
+		ptr = &transid;
+	if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
+		readonly = true;
+	if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
+		if (vol_args->size > PAGE_CACHE_SIZE) {
 			ret = -EINVAL;
 			goto out;
 		}
-
-		src_inode = src_file->f_path.dentry->d_inode;
-		if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
-			printk(KERN_INFO "btrfs: Snapshot src from "
-			       "another FS\n");
-			ret = -EINVAL;
-			fput(src_file);
+		inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
+		if (IS_ERR(inherit)) {
+			ret = PTR_ERR(inherit);
 			goto out;
 		}
-		ret = btrfs_mksubvol(&file->f_path, vol_args->name, namelen,
-				     BTRFS_I(src_inode)->root);
-		fput(src_file);
 	}
+
+	ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
+					      vol_args->fd, subvol, ptr,
+					      readonly, inherit);
+
+	if (ret == 0 && ptr &&
+	    copy_to_user(arg +
+			 offsetof(struct btrfs_ioctl_vol_args_v2,
+				  transid), ptr, sizeof(*ptr)))
+		ret = -EFAULT;
 out:
 	kfree(vol_args);
+	kfree(inherit);
+	return ret;
+}
+
+static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
+						void __user *arg)
+{
+	struct inode *inode = file_inode(file);
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	int ret = 0;
+	u64 flags = 0;
+
+	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
+		return -EINVAL;
+
+	down_read(&root->fs_info->subvol_sem);
+	if (btrfs_root_readonly(root))
+		flags |= BTRFS_SUBVOL_RDONLY;
+	up_read(&root->fs_info->subvol_sem);
+
+	if (copy_to_user(arg, &flags, sizeof(flags)))
+		ret = -EFAULT;
+
+	return ret;
+}
+
+static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
+					      void __user *arg)
+{
+	struct inode *inode = file_inode(file);
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct btrfs_trans_handle *trans;
+	u64 root_flags;
+	u64 flags;
+	int ret = 0;
+
+	if (!inode_owner_or_capable(inode))
+		return -EPERM;
+
+	ret = mnt_want_write_file(file);
+	if (ret)
+		goto out;
+
+	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
+		ret = -EINVAL;
+		goto out_drop_write;
+	}
+
+	if (copy_from_user(&flags, arg, sizeof(flags))) {
+		ret = -EFAULT;
+		goto out_drop_write;
+	}
+
+	if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
+		ret = -EINVAL;
+		goto out_drop_write;
+	}
+
+	if (flags & ~BTRFS_SUBVOL_RDONLY) {
+		ret = -EOPNOTSUPP;
+		goto out_drop_write;
+	}
+
+	down_write(&root->fs_info->subvol_sem);
+
+	/* nothing to do */
+	if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
+		goto out_drop_sem;
+
+	root_flags = btrfs_root_flags(&root->root_item);
+	if (flags & BTRFS_SUBVOL_RDONLY) {
+		btrfs_set_root_flags(&root->root_item,
+				     root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
+	} else {
+		/*
+		 * Block RO -> RW transition if this subvolume is involved in
+		 * send
+		 */
+		spin_lock(&root->root_item_lock);
+		if (root->send_in_progress == 0) {
+			btrfs_set_root_flags(&root->root_item,
+				     root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
+			spin_unlock(&root->root_item_lock);
+		} else {
+			spin_unlock(&root->root_item_lock);
+			btrfs_warn(root->fs_info,
+			"Attempt to set subvolume %llu read-write during send",
+					root->root_key.objectid);
+			ret = -EPERM;
+			goto out_drop_sem;
+		}
+	}
+
+	trans = btrfs_start_transaction(root, 1);
+	if (IS_ERR(trans)) {
+		ret = PTR_ERR(trans);
+		goto out_reset;
+	}
+
+	ret = btrfs_update_root(trans, root->fs_info->tree_root,
+				&root->root_key, &root->root_item);
+
+	btrfs_commit_transaction(trans, root);
+out_reset:
+	if (ret)
+		btrfs_set_root_flags(&root->root_item, root_flags);
+out_drop_sem:
+	up_write(&root->fs_info->subvol_sem);
+out_drop_write:
+	mnt_drop_write_file(file);
+out:
 	return ret;
 }
 
@@ -881,13 +1885,30 @@ out:
 static noinline int may_destroy_subvol(struct btrfs_root *root)
 {
 	struct btrfs_path *path;
+	struct btrfs_dir_item *di;
 	struct btrfs_key key;
+	u64 dir_id;
 	int ret;
 
 	path = btrfs_alloc_path();
 	if (!path)
 		return -ENOMEM;
 
+	/* Make sure this root isn't set as the default subvol */
+	dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
+	di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path,
+				   dir_id, "default", 7, 0);
+	if (di && !IS_ERR(di)) {
+		btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
+		if (key.objectid == root->root_key.objectid) {
+			ret = -EPERM;
+			btrfs_err(root->fs_info, "deleting default subvolume "
+				  "%llu is not allowed", key.objectid);
+			goto out;
+		}
+		btrfs_release_path(path);
+	}
+
 	key.objectid = root->root_key.objectid;
 	key.type = BTRFS_ROOT_REF_KEY;
 	key.offset = (u64)-1;
@@ -939,7 +1960,8 @@ static noinline int copy_to_sk(struct btrfs_root *root,
 			       struct btrfs_path *path,
 			       struct btrfs_key *key,
 			       struct btrfs_ioctl_search_key *sk,
-			       char *buf,
+			       size_t *buf_size,
+			       char __user *ubuf,
 			       unsigned long *sk_offset,
 			       int *num_found)
 {
@@ -951,7 +1973,6 @@ static noinline int copy_to_sk(struct btrfs_root *root,
 	int nritems;
 	int i;
 	int slot;
-	int found = 0;
 	int ret = 0;
 
 	leaf = path->nodes[0];
@@ -968,19 +1989,31 @@ static noinline int copy_to_sk(struct btrfs_root *root,
 		item_off = btrfs_item_ptr_offset(leaf, i);
 		item_len = btrfs_item_size_nr(leaf, i);
 
-		if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
+		btrfs_item_key_to_cpu(leaf, key, i);
+		if (!key_in_sk(key, sk))
+			continue;
+
+		if (sizeof(sh) + item_len > *buf_size) {
+			if (*num_found) {
+				ret = 1;
+				goto out;
+			}
+
+			/*
+			 * return one empty item back for v1, which does not
+			 * handle -EOVERFLOW
+			 */
+
+			*buf_size = sizeof(sh) + item_len;
 			item_len = 0;
+			ret = -EOVERFLOW;
+		}
 
-		if (sizeof(sh) + item_len + *sk_offset >
-		    BTRFS_SEARCH_ARGS_BUFSIZE) {
+		if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
 			ret = 1;
-			goto overflow;
+			goto out;
 		}
 
-		btrfs_item_key_to_cpu(leaf, key, i);
-		if (!key_in_sk(key, sk))
-			continue;
-
 		sh.objectid = key->objectid;
 		sh.offset = key->offset;
 		sh.type = key->type;
@@ -988,20 +2021,33 @@ static noinline int copy_to_sk(struct btrfs_root *root,
 		sh.transid = found_transid;
 
 		/* copy search result header */
-		memcpy(buf + *sk_offset, &sh, sizeof(sh));
+		if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
+			ret = -EFAULT;
+			goto out;
+		}
+
 		*sk_offset += sizeof(sh);
 
 		if (item_len) {
-			char *p = buf + *sk_offset;
+			char __user *up = ubuf + *sk_offset;
 			/* copy the item */
-			read_extent_buffer(leaf, p,
-					   item_off, item_len);
+			if (read_extent_buffer_to_user(leaf, up,
+						       item_off, item_len)) {
+				ret = -EFAULT;
+				goto out;
+			}
+
 			*sk_offset += item_len;
 		}
-		found++;
+		(*num_found)++;
 
-		if (*num_found >= sk->nr_items)
-			break;
+		if (ret) /* -EOVERFLOW from above */
+			goto out;
+
+		if (*num_found >= sk->nr_items) {
+			ret = 1;
+			goto out;
+		}
 	}
 advance_key:
 	ret = 0;
@@ -1016,24 +2062,37 @@ advance_key:
 		key->objectid++;
 	} else
 		ret = 1;
-overflow:
-	*num_found += found;
+out:
+	/*
+	 *  0: all items from this leaf copied, continue with next
+	 *  1: * more items can be copied, but unused buffer is too small
+	 *     * all items were found
+	 *     Either way, it will stops the loop which iterates to the next
+	 *     leaf
+	 *  -EOVERFLOW: item was to large for buffer
+	 *  -EFAULT: could not copy extent buffer back to userspace
+	 */
 	return ret;
 }
 
 static noinline int search_ioctl(struct inode *inode,
-				 struct btrfs_ioctl_search_args *args)
+				 struct btrfs_ioctl_search_key *sk,
+				 size_t *buf_size,
+				 char __user *ubuf)
 {
 	struct btrfs_root *root;
 	struct btrfs_key key;
-	struct btrfs_key max_key;
 	struct btrfs_path *path;
-	struct btrfs_ioctl_search_key *sk = &args->key;
 	struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
 	int ret;
 	int num_found = 0;
 	unsigned long sk_offset = 0;
 
+	if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
+		*buf_size = sizeof(struct btrfs_ioctl_search_header);
+		return -EOVERFLOW;
+	}
+
 	path = btrfs_alloc_path();
 	if (!path)
 		return -ENOMEM;
@@ -1047,7 +2106,7 @@ static noinline int search_ioctl(struct inode *inode,
 		key.offset = (u64)-1;
 		root = btrfs_read_fs_root_no_name(info, &key);
 		if (IS_ERR(root)) {
-			printk(KERN_ERR "could not find root %llu\n",
+			printk(KERN_ERR "BTRFS: could not find root %llu\n",
 			       sk->tree_id);
 			btrfs_free_path(path);
 			return -ENOENT;
@@ -1058,28 +2117,24 @@ static noinline int search_ioctl(struct inode *inode,
 	key.type = sk->min_type;
 	key.offset = sk->min_offset;
 
-	max_key.objectid = sk->max_objectid;
-	max_key.type = sk->max_type;
-	max_key.offset = sk->max_offset;
-
 	path->keep_locks = 1;
 
-	while(1) {
-		ret = btrfs_search_forward(root, &key, &max_key, path, 0,
-					   sk->min_transid);
+	while (1) {
+		ret = btrfs_search_forward(root, &key, path, sk->min_transid);
 		if (ret != 0) {
 			if (ret > 0)
 				ret = 0;
 			goto err;
 		}
-		ret = copy_to_sk(root, path, &key, sk, args->buf,
+		ret = copy_to_sk(root, path, &key, sk, buf_size, ubuf,
 				 &sk_offset, &num_found);
-		btrfs_release_path(root, path);
-		if (ret || num_found >= sk->nr_items)
+		btrfs_release_path(path);
+		if (ret)
 			break;
 
 	}
-	ret = 0;
+	if (ret > 0)
+		ret = 0;
 err:
 	sk->nr_items = num_found;
 	btrfs_free_path(path);
@@ -1089,26 +2144,73 @@ err:
 static noinline int btrfs_ioctl_tree_search(struct file *file,
 					   void __user *argp)
 {
-	 struct btrfs_ioctl_search_args *args;
-	 struct inode *inode;
-	 int ret;
+	struct btrfs_ioctl_search_args __user *uargs;
+	struct btrfs_ioctl_search_key sk;
+	struct inode *inode;
+	int ret;
+	size_t buf_size;
 
 	if (!capable(CAP_SYS_ADMIN))
 		return -EPERM;
 
-	args = kmalloc(sizeof(*args), GFP_KERNEL);
-	if (!args)
-		return -ENOMEM;
+	uargs = (struct btrfs_ioctl_search_args __user *)argp;
 
-	if (copy_from_user(args, argp, sizeof(*args))) {
-		kfree(args);
+	if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
 		return -EFAULT;
-	}
-	inode = fdentry(file)->d_inode;
-	ret = search_ioctl(inode, args);
-	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
+
+	buf_size = sizeof(uargs->buf);
+
+	inode = file_inode(file);
+	ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
+
+	/*
+	 * In the origin implementation an overflow is handled by returning a
+	 * search header with a len of zero, so reset ret.
+	 */
+	if (ret == -EOVERFLOW)
+		ret = 0;
+
+	if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
 		ret = -EFAULT;
-	kfree(args);
+	return ret;
+}
+
+static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
+					       void __user *argp)
+{
+	struct btrfs_ioctl_search_args_v2 __user *uarg;
+	struct btrfs_ioctl_search_args_v2 args;
+	struct inode *inode;
+	int ret;
+	size_t buf_size;
+	const size_t buf_limit = 16 * 1024 * 1024;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return -EPERM;
+
+	/* copy search header and buffer size */
+	uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
+	if (copy_from_user(&args, uarg, sizeof(args)))
+		return -EFAULT;
+
+	buf_size = args.buf_size;
+
+	if (buf_size < sizeof(struct btrfs_ioctl_search_header))
+		return -EOVERFLOW;
+
+	/* limit result size to 16MB */
+	if (buf_size > buf_limit)
+		buf_size = buf_limit;
+
+	inode = file_inode(file);
+	ret = search_ioctl(inode, &args.key, &buf_size,
+			   (char *)(&uarg->buf[0]));
+	if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
+		ret = -EFAULT;
+	else if (ret == -EOVERFLOW &&
+		copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
+		ret = -EFAULT;
+
 	return ret;
 }
 
@@ -1146,7 +2248,7 @@ static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
 	key.offset = (u64)-1;
 	root = btrfs_read_fs_root_no_name(info, &key);
 	if (IS_ERR(root)) {
-		printk(KERN_ERR "could not find root %llu\n", tree_id);
+		printk(KERN_ERR "BTRFS: could not find root %llu\n", tree_id);
 		ret = -ENOENT;
 		goto out;
 	}
@@ -1155,46 +2257,47 @@ static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
 	key.type = BTRFS_INODE_REF_KEY;
 	key.offset = (u64)-1;
 
-	while(1) {
+	while (1) {
 		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
 		if (ret < 0)
 			goto out;
+		else if (ret > 0) {
+			ret = btrfs_previous_item(root, path, dirid,
+						  BTRFS_INODE_REF_KEY);
+			if (ret < 0)
+				goto out;
+			else if (ret > 0) {
+				ret = -ENOENT;
+				goto out;
+			}
+		}
 
 		l = path->nodes[0];
 		slot = path->slots[0];
-		if (ret > 0 && slot > 0)
-			slot--;
 		btrfs_item_key_to_cpu(l, &key, slot);
 
-		if (ret > 0 && (key.objectid != dirid ||
-				key.type != BTRFS_INODE_REF_KEY)) {
-			ret = -ENOENT;
-			goto out;
-		}
-
 		iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
 		len = btrfs_inode_ref_name_len(l, iref);
 		ptr -= len + 1;
 		total_len += len + 1;
-		if (ptr < name)
+		if (ptr < name) {
+			ret = -ENAMETOOLONG;
 			goto out;
+		}
 
 		*(ptr + len) = '/';
-		read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
+		read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
 
 		if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
 			break;
 
-		btrfs_release_path(root, path);
+		btrfs_release_path(path);
 		key.objectid = key.offset;
 		key.offset = (u64)-1;
 		dirid = key.objectid;
-
 	}
-	if (ptr < name)
-		goto out;
-	memcpy(name, ptr, total_len);
-	name[total_len]='\0';
+	memmove(name, ptr, total_len);
+	name[total_len] = '\0';
 	ret = 0;
 out:
 	btrfs_free_path(path);
@@ -1211,15 +2314,11 @@ static noinline int btrfs_ioctl_ino_lookup(struct file *file,
 	if (!capable(CAP_SYS_ADMIN))
 		return -EPERM;
 
-	args = kmalloc(sizeof(*args), GFP_KERNEL);
-	if (!args)
-		return -ENOMEM;
+	args = memdup_user(argp, sizeof(*args));
+	if (IS_ERR(args))
+		return PTR_ERR(args);
 
-	if (copy_from_user(args, argp, sizeof(*args))) {
-		kfree(args);
-		return -EFAULT;
-	}
-	inode = fdentry(file)->d_inode;
+	inode = file_inode(file);
 
 	if (args->treeid == 0)
 		args->treeid = BTRFS_I(inode)->root->root_key.objectid;
@@ -1238,7 +2337,7 @@ static noinline int btrfs_ioctl_ino_lookup(struct file *file,
 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
 					     void __user *arg)
 {
-	struct dentry *parent = fdentry(file);
+	struct dentry *parent = file->f_path.dentry;
 	struct dentry *dentry;
 	struct inode *dir = parent->d_inode;
 	struct inode *inode;
@@ -1246,13 +2345,13 @@ static noinline int btrfs_ioctl_snap_destroy(struct file *file,
 	struct btrfs_root *dest = NULL;
 	struct btrfs_ioctl_vol_args *vol_args;
 	struct btrfs_trans_handle *trans;
+	struct btrfs_block_rsv block_rsv;
+	u64 root_flags;
+	u64 qgroup_reserved;
 	int namelen;
 	int ret;
 	int err = 0;
 
-	if (!capable(CAP_SYS_ADMIN))
-		return -EPERM;
-
 	vol_args = memdup_user(arg, sizeof(*vol_args));
 	if (IS_ERR(vol_args))
 		return PTR_ERR(vol_args);
@@ -1265,11 +2364,14 @@ static noinline int btrfs_ioctl_snap_destroy(struct file *file,
 		goto out;
 	}
 
-	err = mnt_want_write(file->f_path.mnt);
+	err = mnt_want_write_file(file);
 	if (err)
 		goto out;
 
-	mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
+
+	err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
+	if (err == -EINTR)
+		goto out_drop_write;
 	dentry = lookup_one_len(vol_args->name, parent, namelen);
 	if (IS_ERR(dentry)) {
 		err = PTR_ERR(dentry);
@@ -1282,14 +2384,73 @@ static noinline int btrfs_ioctl_snap_destroy(struct file *file,
 	}
 
 	inode = dentry->d_inode;
-	if (inode->i_ino != BTRFS_FIRST_FREE_OBJECTID) {
+	dest = BTRFS_I(inode)->root;
+	if (!capable(CAP_SYS_ADMIN)) {
+		/*
+		 * Regular user.  Only allow this with a special mount
+		 * option, when the user has write+exec access to the
+		 * subvol root, and when rmdir(2) would have been
+		 * allowed.
+		 *
+		 * Note that this is _not_ check that the subvol is
+		 * empty or doesn't contain data that we wouldn't
+		 * otherwise be able to delete.
+		 *
+		 * Users who want to delete empty subvols should try
+		 * rmdir(2).
+		 */
+		err = -EPERM;
+		if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
+			goto out_dput;
+
+		/*
+		 * Do not allow deletion if the parent dir is the same
+		 * as the dir to be deleted.  That means the ioctl
+		 * must be called on the dentry referencing the root
+		 * of the subvol, not a random directory contained
+		 * within it.
+		 */
 		err = -EINVAL;
-		goto out_dput;
+		if (root == dest)
+			goto out_dput;
+
+		err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
+		if (err)
+			goto out_dput;
 	}
 
-	dest = BTRFS_I(inode)->root;
+	/* check if subvolume may be deleted by a user */
+	err = btrfs_may_delete(dir, dentry, 1);
+	if (err)
+		goto out_dput;
+
+	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
+		err = -EINVAL;
+		goto out_dput;
+	}
 
 	mutex_lock(&inode->i_mutex);
+
+	/*
+	 * Don't allow to delete a subvolume with send in progress. This is
+	 * inside the i_mutex so the error handling that has to drop the bit
+	 * again is not run concurrently.
+	 */
+	spin_lock(&dest->root_item_lock);
+	root_flags = btrfs_root_flags(&dest->root_item);
+	if (dest->send_in_progress == 0) {
+		btrfs_set_root_flags(&dest->root_item,
+				root_flags | BTRFS_ROOT_SUBVOL_DEAD);
+		spin_unlock(&dest->root_item_lock);
+	} else {
+		spin_unlock(&dest->root_item_lock);
+		btrfs_warn(root->fs_info,
+			"Attempt to delete subvolume %llu during send",
+			dest->root_key.objectid);
+		err = -EPERM;
+		goto out_dput;
+	}
+
 	err = d_invalidate(dentry);
 	if (err)
 		goto out_unlock;
@@ -1300,12 +2461,33 @@ static noinline int btrfs_ioctl_snap_destroy(struct file *file,
 	if (err)
 		goto out_up_write;
 
-	trans = btrfs_start_transaction(root, 1);
+	btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
+	/*
+	 * One for dir inode, two for dir entries, two for root
+	 * ref/backref.
+	 */
+	err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
+					       5, &qgroup_reserved, true);
+	if (err)
+		goto out_up_write;
+
+	trans = btrfs_start_transaction(root, 0);
+	if (IS_ERR(trans)) {
+		err = PTR_ERR(trans);
+		goto out_release;
+	}
+	trans->block_rsv = &block_rsv;
+	trans->bytes_reserved = block_rsv.size;
+
 	ret = btrfs_unlink_subvol(trans, root, dir,
 				dest->root_key.objectid,
 				dentry->d_name.name,
 				dentry->d_name.len);
-	BUG_ON(ret);
+	if (ret) {
+		err = ret;
+		btrfs_abort_transaction(trans, root, ret);
+		goto out_end_trans;
+	}
 
 	btrfs_record_root_in_trans(trans, dest);
 
@@ -1314,28 +2496,75 @@ static noinline int btrfs_ioctl_snap_destroy(struct file *file,
 	dest->root_item.drop_level = 0;
 	btrfs_set_root_refs(&dest->root_item, 0);
 
-	ret = btrfs_insert_orphan_item(trans,
-				root->fs_info->tree_root,
-				dest->root_key.objectid);
-	BUG_ON(ret);
+	if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) {
+		ret = btrfs_insert_orphan_item(trans,
+					root->fs_info->tree_root,
+					dest->root_key.objectid);
+		if (ret) {
+			btrfs_abort_transaction(trans, root, ret);
+			err = ret;
+			goto out_end_trans;
+		}
+	}
 
-	ret = btrfs_commit_transaction(trans, root);
-	BUG_ON(ret);
+	ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
+				  dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
+				  dest->root_key.objectid);
+	if (ret && ret != -ENOENT) {
+		btrfs_abort_transaction(trans, root, ret);
+		err = ret;
+		goto out_end_trans;
+	}
+	if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
+		ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
+					  dest->root_item.received_uuid,
+					  BTRFS_UUID_KEY_RECEIVED_SUBVOL,
+					  dest->root_key.objectid);
+		if (ret && ret != -ENOENT) {
+			btrfs_abort_transaction(trans, root, ret);
+			err = ret;
+			goto out_end_trans;
+		}
+	}
+
+out_end_trans:
+	trans->block_rsv = NULL;
+	trans->bytes_reserved = 0;
+	ret = btrfs_end_transaction(trans, root);
+	if (ret && !err)
+		err = ret;
 	inode->i_flags |= S_DEAD;
+out_release:
+	btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
 out_up_write:
 	up_write(&root->fs_info->subvol_sem);
 out_unlock:
+	if (err) {
+		spin_lock(&dest->root_item_lock);
+		root_flags = btrfs_root_flags(&dest->root_item);
+		btrfs_set_root_flags(&dest->root_item,
+				root_flags & ~BTRFS_ROOT_SUBVOL_DEAD);
+		spin_unlock(&dest->root_item_lock);
+	}
 	mutex_unlock(&inode->i_mutex);
 	if (!err) {
 		shrink_dcache_sb(root->fs_info->sb);
 		btrfs_invalidate_inodes(dest);
 		d_delete(dentry);
+		ASSERT(dest->send_in_progress == 0);
+
+		/* the last ref */
+		if (dest->cache_inode) {
+			iput(dest->cache_inode);
+			dest->cache_inode = NULL;
+		}
 	}
 out_dput:
 	dput(dentry);
 out_unlock_dir:
 	mutex_unlock(&dir->i_mutex);
-	mnt_drop_write(file->f_path.mnt);
+out_drop_write:
+	mnt_drop_write_file(file);
 out:
 	kfree(vol_args);
 	return err;
@@ -1343,23 +2572,30 @@ out:
 
 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
 {
-	struct inode *inode = fdentry(file)->d_inode;
+	struct inode *inode = file_inode(file);
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	struct btrfs_ioctl_defrag_range_args *range;
 	int ret;
 
-	ret = mnt_want_write(file->f_path.mnt);
+	ret = mnt_want_write_file(file);
 	if (ret)
 		return ret;
 
+	if (btrfs_root_readonly(root)) {
+		ret = -EROFS;
+		goto out;
+	}
+
 	switch (inode->i_mode & S_IFMT) {
 	case S_IFDIR:
 		if (!capable(CAP_SYS_ADMIN)) {
 			ret = -EPERM;
 			goto out;
 		}
-		btrfs_defrag_root(root, 0);
-		btrfs_defrag_root(root->fs_info->extent_root, 0);
+		ret = btrfs_defrag_root(root);
+		if (ret)
+			goto out;
+		ret = btrfs_defrag_root(root->fs_info->extent_root);
 		break;
 	case S_IFREG:
 		if (!(file->f_mode & FMODE_WRITE)) {
@@ -1389,12 +2625,17 @@ static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
 			/* the rest are all set to zero by kzalloc */
 			range->len = (u64)-1;
 		}
-		btrfs_defrag_file(file, range);
+		ret = btrfs_defrag_file(file_inode(file), file,
+					range, 0, 0);
+		if (ret > 0)
+			ret = 0;
 		kfree(range);
 		break;
+	default:
+		ret = -EINVAL;
 	}
 out:
-	mnt_drop_write(file->f_path.mnt);
+	mnt_drop_write_file(file);
 	return ret;
 }
 
@@ -1406,168 +2647,628 @@ static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
 	if (!capable(CAP_SYS_ADMIN))
 		return -EPERM;
 
+	if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
+			1)) {
+		return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
+	}
+
+	mutex_lock(&root->fs_info->volume_mutex);
 	vol_args = memdup_user(arg, sizeof(*vol_args));
-	if (IS_ERR(vol_args))
-		return PTR_ERR(vol_args);
+	if (IS_ERR(vol_args)) {
+		ret = PTR_ERR(vol_args);
+		goto out;
+	}
 
 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
 	ret = btrfs_init_new_device(root, vol_args->name);
 
 	kfree(vol_args);
+out:
+	mutex_unlock(&root->fs_info->volume_mutex);
+	atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
 	return ret;
 }
 
-static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
+static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
 {
+	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
 	struct btrfs_ioctl_vol_args *vol_args;
 	int ret;
 
 	if (!capable(CAP_SYS_ADMIN))
 		return -EPERM;
 
-	if (root->fs_info->sb->s_flags & MS_RDONLY)
-		return -EROFS;
+	ret = mnt_want_write_file(file);
+	if (ret)
+		return ret;
 
 	vol_args = memdup_user(arg, sizeof(*vol_args));
-	if (IS_ERR(vol_args))
-		return PTR_ERR(vol_args);
+	if (IS_ERR(vol_args)) {
+		ret = PTR_ERR(vol_args);
+		goto out;
+	}
 
 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
+
+	if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
+			1)) {
+		ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
+		goto out;
+	}
+
+	mutex_lock(&root->fs_info->volume_mutex);
 	ret = btrfs_rm_device(root, vol_args->name);
+	mutex_unlock(&root->fs_info->volume_mutex);
+	atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
 
+out:
 	kfree(vol_args);
+	mnt_drop_write_file(file);
 	return ret;
 }
 
-static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
-				       u64 off, u64 olen, u64 destoff)
+static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
+{
+	struct btrfs_ioctl_fs_info_args *fi_args;
+	struct btrfs_device *device;
+	struct btrfs_device *next;
+	struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
+	int ret = 0;
+
+	fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
+	if (!fi_args)
+		return -ENOMEM;
+
+	mutex_lock(&fs_devices->device_list_mutex);
+	fi_args->num_devices = fs_devices->num_devices;
+	memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
+
+	list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
+		if (device->devid > fi_args->max_id)
+			fi_args->max_id = device->devid;
+	}
+	mutex_unlock(&fs_devices->device_list_mutex);
+
+	fi_args->nodesize = root->fs_info->super_copy->nodesize;
+	fi_args->sectorsize = root->fs_info->super_copy->sectorsize;
+	fi_args->clone_alignment = root->fs_info->super_copy->sectorsize;
+
+	if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
+		ret = -EFAULT;
+
+	kfree(fi_args);
+	return ret;
+}
+
+static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
+{
+	struct btrfs_ioctl_dev_info_args *di_args;
+	struct btrfs_device *dev;
+	struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
+	int ret = 0;
+	char *s_uuid = NULL;
+
+	di_args = memdup_user(arg, sizeof(*di_args));
+	if (IS_ERR(di_args))
+		return PTR_ERR(di_args);
+
+	if (!btrfs_is_empty_uuid(di_args->uuid))
+		s_uuid = di_args->uuid;
+
+	mutex_lock(&fs_devices->device_list_mutex);
+	dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
+
+	if (!dev) {
+		ret = -ENODEV;
+		goto out;
+	}
+
+	di_args->devid = dev->devid;
+	di_args->bytes_used = dev->bytes_used;
+	di_args->total_bytes = dev->total_bytes;
+	memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
+	if (dev->name) {
+		struct rcu_string *name;
+
+		rcu_read_lock();
+		name = rcu_dereference(dev->name);
+		strncpy(di_args->path, name->str, sizeof(di_args->path));
+		rcu_read_unlock();
+		di_args->path[sizeof(di_args->path) - 1] = 0;
+	} else {
+		di_args->path[0] = '\0';
+	}
+
+out:
+	mutex_unlock(&fs_devices->device_list_mutex);
+	if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
+		ret = -EFAULT;
+
+	kfree(di_args);
+	return ret;
+}
+
+static struct page *extent_same_get_page(struct inode *inode, u64 off)
+{
+	struct page *page;
+	pgoff_t index;
+	struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
+
+	index = off >> PAGE_CACHE_SHIFT;
+
+	page = grab_cache_page(inode->i_mapping, index);
+	if (!page)
+		return NULL;
+
+	if (!PageUptodate(page)) {
+		if (extent_read_full_page_nolock(tree, page, btrfs_get_extent,
+						 0))
+			return NULL;
+		lock_page(page);
+		if (!PageUptodate(page)) {
+			unlock_page(page);
+			page_cache_release(page);
+			return NULL;
+		}
+	}
+	unlock_page(page);
+
+	return page;
+}
+
+static inline void lock_extent_range(struct inode *inode, u64 off, u64 len)
+{
+	/* do any pending delalloc/csum calc on src, one way or
+	   another, and lock file content */
+	while (1) {
+		struct btrfs_ordered_extent *ordered;
+		lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
+		ordered = btrfs_lookup_first_ordered_extent(inode,
+							    off + len - 1);
+		if ((!ordered ||
+		     ordered->file_offset + ordered->len <= off ||
+		     ordered->file_offset >= off + len) &&
+		    !test_range_bit(&BTRFS_I(inode)->io_tree, off,
+				    off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
+			if (ordered)
+				btrfs_put_ordered_extent(ordered);
+			break;
+		}
+		unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
+		if (ordered)
+			btrfs_put_ordered_extent(ordered);
+		btrfs_wait_ordered_range(inode, off, len);
+	}
+}
+
+static void btrfs_double_unlock(struct inode *inode1, u64 loff1,
+				struct inode *inode2, u64 loff2, u64 len)
+{
+	unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
+	unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
+
+	mutex_unlock(&inode1->i_mutex);
+	mutex_unlock(&inode2->i_mutex);
+}
+
+static void btrfs_double_lock(struct inode *inode1, u64 loff1,
+			      struct inode *inode2, u64 loff2, u64 len)
+{
+	if (inode1 < inode2) {
+		swap(inode1, inode2);
+		swap(loff1, loff2);
+	}
+
+	mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT);
+	lock_extent_range(inode1, loff1, len);
+	if (inode1 != inode2) {
+		mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD);
+		lock_extent_range(inode2, loff2, len);
+	}
+}
+
+static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst,
+			  u64 dst_loff, u64 len)
+{
+	int ret = 0;
+	struct page *src_page, *dst_page;
+	unsigned int cmp_len = PAGE_CACHE_SIZE;
+	void *addr, *dst_addr;
+
+	while (len) {
+		if (len < PAGE_CACHE_SIZE)
+			cmp_len = len;
+
+		src_page = extent_same_get_page(src, loff);
+		if (!src_page)
+			return -EINVAL;
+		dst_page = extent_same_get_page(dst, dst_loff);
+		if (!dst_page) {
+			page_cache_release(src_page);
+			return -EINVAL;
+		}
+		addr = kmap_atomic(src_page);
+		dst_addr = kmap_atomic(dst_page);
+
+		flush_dcache_page(src_page);
+		flush_dcache_page(dst_page);
+
+		if (memcmp(addr, dst_addr, cmp_len))
+			ret = BTRFS_SAME_DATA_DIFFERS;
+
+		kunmap_atomic(addr);
+		kunmap_atomic(dst_addr);
+		page_cache_release(src_page);
+		page_cache_release(dst_page);
+
+		if (ret)
+			break;
+
+		loff += cmp_len;
+		dst_loff += cmp_len;
+		len -= cmp_len;
+	}
+
+	return ret;
+}
+
+static int extent_same_check_offsets(struct inode *inode, u64 off, u64 len)
+{
+	u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
+
+	if (off + len > inode->i_size || off + len < off)
+		return -EINVAL;
+	/* Check that we are block aligned - btrfs_clone() requires this */
+	if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
+		return -EINVAL;
+
+	return 0;
+}
+
+static int btrfs_extent_same(struct inode *src, u64 loff, u64 len,
+			     struct inode *dst, u64 dst_loff)
 {
-	struct inode *inode = fdentry(file)->d_inode;
-	struct btrfs_root *root = BTRFS_I(inode)->root;
-	struct file *src_file;
-	struct inode *src;
-	struct btrfs_trans_handle *trans;
-	struct btrfs_path *path;
-	struct extent_buffer *leaf;
-	char *buf;
-	struct btrfs_key key;
-	u32 nritems;
-	int slot;
 	int ret;
-	u64 len = olen;
-	u64 bs = root->fs_info->sb->s_blocksize;
-	u64 hint_byte;
 
 	/*
-	 * TODO:
-	 * - split compressed inline extents.  annoying: we need to
-	 *   decompress into destination's address_space (the file offset
-	 *   may change, so source mapping won't do), then recompress (or
-	 *   otherwise reinsert) a subrange.
-	 * - allow ranges within the same file to be cloned (provided
-	 *   they don't overlap)?
+	 * btrfs_clone() can't handle extents in the same file
+	 * yet. Once that works, we can drop this check and replace it
+	 * with a check for the same inode, but overlapping extents.
 	 */
+	if (src == dst)
+		return -EINVAL;
 
-	/* the destination must be opened for writing */
-	if (!(file->f_mode & FMODE_WRITE))
+	btrfs_double_lock(src, loff, dst, dst_loff, len);
+
+	ret = extent_same_check_offsets(src, loff, len);
+	if (ret)
+		goto out_unlock;
+
+	ret = extent_same_check_offsets(dst, dst_loff, len);
+	if (ret)
+		goto out_unlock;
+
+	/* don't make the dst file partly checksummed */
+	if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
+	    (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
+		ret = -EINVAL;
+		goto out_unlock;
+	}
+
+	ret = btrfs_cmp_data(src, loff, dst, dst_loff, len);
+	if (ret == 0)
+		ret = btrfs_clone(src, dst, loff, len, len, dst_loff);
+
+out_unlock:
+	btrfs_double_unlock(src, loff, dst, dst_loff, len);
+
+	return ret;
+}
+
+#define BTRFS_MAX_DEDUPE_LEN	(16 * 1024 * 1024)
+
+static long btrfs_ioctl_file_extent_same(struct file *file,
+			struct btrfs_ioctl_same_args __user *argp)
+{
+	struct btrfs_ioctl_same_args *same;
+	struct btrfs_ioctl_same_extent_info *info;
+	struct inode *src = file_inode(file);
+	u64 off;
+	u64 len;
+	int i;
+	int ret;
+	unsigned long size;
+	u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
+	bool is_admin = capable(CAP_SYS_ADMIN);
+	u16 count;
+
+	if (!(file->f_mode & FMODE_READ))
 		return -EINVAL;
 
-	ret = mnt_want_write(file->f_path.mnt);
+	ret = mnt_want_write_file(file);
 	if (ret)
 		return ret;
 
-	src_file = fget(srcfd);
-	if (!src_file) {
-		ret = -EBADF;
-		goto out_drop_write;
+	if (get_user(count, &argp->dest_count)) {
+		ret = -EFAULT;
+		goto out;
 	}
-	src = src_file->f_dentry->d_inode;
 
-	ret = -EINVAL;
-	if (src == inode)
-		goto out_fput;
+	size = offsetof(struct btrfs_ioctl_same_args __user, info[count]);
+
+	same = memdup_user(argp, size);
+
+	if (IS_ERR(same)) {
+		ret = PTR_ERR(same);
+		goto out;
+	}
+
+	off = same->logical_offset;
+	len = same->length;
+
+	/*
+	 * Limit the total length we will dedupe for each operation.
+	 * This is intended to bound the total time spent in this
+	 * ioctl to something sane.
+	 */
+	if (len > BTRFS_MAX_DEDUPE_LEN)
+		len = BTRFS_MAX_DEDUPE_LEN;
+
+	if (WARN_ON_ONCE(bs < PAGE_CACHE_SIZE)) {
+		/*
+		 * Btrfs does not support blocksize < page_size. As a
+		 * result, btrfs_cmp_data() won't correctly handle
+		 * this situation without an update.
+		 */
+		ret = -EINVAL;
+		goto out;
+	}
 
 	ret = -EISDIR;
-	if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
-		goto out_fput;
+	if (S_ISDIR(src->i_mode))
+		goto out;
 
-	ret = -EXDEV;
-	if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root)
-		goto out_fput;
+	ret = -EACCES;
+	if (!S_ISREG(src->i_mode))
+		goto out;
 
-	ret = -ENOMEM;
-	buf = vmalloc(btrfs_level_size(root, 0));
-	if (!buf)
-		goto out_fput;
+	/* pre-format output fields to sane values */
+	for (i = 0; i < count; i++) {
+		same->info[i].bytes_deduped = 0ULL;
+		same->info[i].status = 0;
+	}
 
-	path = btrfs_alloc_path();
-	if (!path) {
-		vfree(buf);
-		goto out_fput;
+	for (i = 0, info = same->info; i < count; i++, info++) {
+		struct inode *dst;
+		struct fd dst_file = fdget(info->fd);
+		if (!dst_file.file) {
+			info->status = -EBADF;
+			continue;
+		}
+		dst = file_inode(dst_file.file);
+
+		if (!(is_admin || (dst_file.file->f_mode & FMODE_WRITE))) {
+			info->status = -EINVAL;
+		} else if (file->f_path.mnt != dst_file.file->f_path.mnt) {
+			info->status = -EXDEV;
+		} else if (S_ISDIR(dst->i_mode)) {
+			info->status = -EISDIR;
+		} else if (!S_ISREG(dst->i_mode)) {
+			info->status = -EACCES;
+		} else {
+			info->status = btrfs_extent_same(src, off, len, dst,
+							info->logical_offset);
+			if (info->status == 0)
+				info->bytes_deduped += len;
+		}
+		fdput(dst_file);
 	}
-	path->reada = 2;
 
-	if (inode < src) {
-		mutex_lock(&inode->i_mutex);
-		mutex_lock(&src->i_mutex);
-	} else {
-		mutex_lock(&src->i_mutex);
-		mutex_lock(&inode->i_mutex);
+	ret = copy_to_user(argp, same, size);
+	if (ret)
+		ret = -EFAULT;
+
+out:
+	mnt_drop_write_file(file);
+	return ret;
+}
+
+/* Helper to check and see if this root currently has a ref on the given disk
+ * bytenr.  If it does then we need to update the quota for this root.  This
+ * doesn't do anything if quotas aren't enabled.
+ */
+static int check_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+		     u64 disko)
+{
+	struct seq_list tree_mod_seq_elem = {};
+	struct ulist *roots;
+	struct ulist_iterator uiter;
+	struct ulist_node *root_node = NULL;
+	int ret;
+
+	if (!root->fs_info->quota_enabled)
+		return 1;
+
+	btrfs_get_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
+	ret = btrfs_find_all_roots(trans, root->fs_info, disko,
+				   tree_mod_seq_elem.seq, &roots);
+	if (ret < 0)
+		goto out;
+	ret = 0;
+	ULIST_ITER_INIT(&uiter);
+	while ((root_node = ulist_next(roots, &uiter))) {
+		if (root_node->val == root->objectid) {
+			ret = 1;
+			break;
+		}
 	}
+	ulist_free(roots);
+out:
+	btrfs_put_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
+	return ret;
+}
 
-	/* determine range to clone */
-	ret = -EINVAL;
-	if (off >= src->i_size || off + len > src->i_size)
-		goto out_unlock;
-	if (len == 0)
-		olen = len = src->i_size - off;
-	/* if we extend to eof, continue to block boundary */
-	if (off + len == src->i_size)
-		len = ((src->i_size + bs-1) & ~(bs-1))
-			- off;
+static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
+				     struct inode *inode,
+				     u64 endoff,
+				     const u64 destoff,
+				     const u64 olen)
+{
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	int ret;
 
-	/* verify the end result is block aligned */
-	if ((off & (bs-1)) ||
-	    ((off + len) & (bs-1)))
-		goto out_unlock;
+	inode_inc_iversion(inode);
+	inode->i_mtime = inode->i_ctime = CURRENT_TIME;
+	/*
+	 * We round up to the block size at eof when determining which
+	 * extents to clone above, but shouldn't round up the file size.
+	 */
+	if (endoff > destoff + olen)
+		endoff = destoff + olen;
+	if (endoff > inode->i_size)
+		btrfs_i_size_write(inode, endoff);
+
+	ret = btrfs_update_inode(trans, root, inode);
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
+		btrfs_end_transaction(trans, root);
+		goto out;
+	}
+	ret = btrfs_end_transaction(trans, root);
+out:
+	return ret;
+}
+
+static void clone_update_extent_map(struct inode *inode,
+				    const struct btrfs_trans_handle *trans,
+				    const struct btrfs_path *path,
+				    const u64 hole_offset,
+				    const u64 hole_len)
+{
+	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
+	struct extent_map *em;
+	int ret;
+
+	em = alloc_extent_map();
+	if (!em) {
+		set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
+			&BTRFS_I(inode)->runtime_flags);
+		return;
+	}
+
+	if (path) {
+		struct btrfs_file_extent_item *fi;
+
+		fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
+				    struct btrfs_file_extent_item);
+		btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
+		em->generation = -1;
+		if (btrfs_file_extent_type(path->nodes[0], fi) ==
+		    BTRFS_FILE_EXTENT_INLINE)
+			set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
+				&BTRFS_I(inode)->runtime_flags);
+	} else {
+		em->start = hole_offset;
+		em->len = hole_len;
+		em->ram_bytes = em->len;
+		em->orig_start = hole_offset;
+		em->block_start = EXTENT_MAP_HOLE;
+		em->block_len = 0;
+		em->orig_block_len = 0;
+		em->compress_type = BTRFS_COMPRESS_NONE;
+		em->generation = trans->transid;
+	}
 
-	/* do any pending delalloc/csum calc on src, one way or
-	   another, and lock file content */
 	while (1) {
-		struct btrfs_ordered_extent *ordered;
-		lock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
-		ordered = btrfs_lookup_first_ordered_extent(inode, off+len);
-		if (BTRFS_I(src)->delalloc_bytes == 0 && !ordered)
+		write_lock(&em_tree->lock);
+		ret = add_extent_mapping(em_tree, em, 1);
+		write_unlock(&em_tree->lock);
+		if (ret != -EEXIST) {
+			free_extent_map(em);
 			break;
-		unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
-		if (ordered)
-			btrfs_put_ordered_extent(ordered);
-		btrfs_wait_ordered_range(src, off, off+len);
+		}
+		btrfs_drop_extent_cache(inode, em->start,
+					em->start + em->len - 1, 0);
 	}
 
-	trans = btrfs_start_transaction(root, 1);
-	BUG_ON(!trans);
+	if (unlikely(ret))
+		set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
+			&BTRFS_I(inode)->runtime_flags);
+}
+
+/**
+ * btrfs_clone() - clone a range from inode file to another
+ *
+ * @src: Inode to clone from
+ * @inode: Inode to clone to
+ * @off: Offset within source to start clone from
+ * @olen: Original length, passed by user, of range to clone
+ * @olen_aligned: Block-aligned value of olen, extent_same uses
+ *               identical values here
+ * @destoff: Offset within @inode to start clone
+ */
+static int btrfs_clone(struct inode *src, struct inode *inode,
+		       const u64 off, const u64 olen, const u64 olen_aligned,
+		       const u64 destoff)
+{
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct btrfs_path *path = NULL;
+	struct extent_buffer *leaf;
+	struct btrfs_trans_handle *trans;
+	char *buf = NULL;
+	struct btrfs_key key;
+	u32 nritems;
+	int slot;
+	int ret;
+	int no_quota;
+	const u64 len = olen_aligned;
+	u64 last_disko = 0;
+	u64 last_dest_end = destoff;
 
-	/* punch hole in destination first */
-	btrfs_drop_extents(trans, inode, off, off + len, &hint_byte, 1);
+	ret = -ENOMEM;
+	buf = vmalloc(btrfs_level_size(root, 0));
+	if (!buf)
+		return ret;
 
+	path = btrfs_alloc_path();
+	if (!path) {
+		vfree(buf);
+		return ret;
+	}
+
+	path->reada = 2;
 	/* clone data */
-	key.objectid = src->i_ino;
+	key.objectid = btrfs_ino(src);
 	key.type = BTRFS_EXTENT_DATA_KEY;
-	key.offset = 0;
+	key.offset = off;
 
 	while (1) {
 		/*
 		 * note the key will change type as we walk through the
 		 * tree.
 		 */
-		ret = btrfs_search_slot(trans, root, &key, path, 0, 0);
+		path->leave_spinning = 1;
+		ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
+				0, 0);
 		if (ret < 0)
 			goto out;
+		/*
+		 * First search, if no extent item that starts at offset off was
+		 * found but the previous item is an extent item, it's possible
+		 * it might overlap our target range, therefore process it.
+		 */
+		if (key.offset == off && ret > 0 && path->slots[0] > 0) {
+			btrfs_item_key_to_cpu(path->nodes[0], &key,
+					      path->slots[0] - 1);
+			if (key.type == BTRFS_EXTENT_DATA_KEY)
+				path->slots[0]--;
+		}
 
 		nritems = btrfs_header_nritems(path->nodes[0]);
+process_slot:
+		no_quota = 1;
 		if (path->slots[0] >= nritems) {
-			ret = btrfs_next_leaf(root, path);
+			ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
 			if (ret < 0)
 				goto out;
 			if (ret > 0)
@@ -1579,7 +3280,7 @@ static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
 
 		btrfs_item_key_to_cpu(leaf, &key, slot);
 		if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
-		    key.objectid != src->i_ino)
+		    key.objectid != btrfs_ino(src))
 			break;
 
 		if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
@@ -1590,11 +3291,7 @@ static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
 			u64 disko = 0, diskl = 0;
 			u64 datao = 0, datal = 0;
 			u8 comp;
-
-			size = btrfs_item_size_nr(leaf, slot);
-			read_extent_buffer(leaf, buf,
-					   btrfs_item_ptr_offset(leaf, slot),
-					   size);
+			u64 drop_start;
 
 			extent = btrfs_item_ptr(leaf, slot,
 						struct btrfs_file_extent_item);
@@ -1614,22 +3311,94 @@ static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
 				datal = btrfs_file_extent_ram_bytes(leaf,
 								    extent);
 			}
-			btrfs_release_path(root, path);
 
-			if (key.offset + datal < off ||
-			    key.offset >= off+len)
-				goto next;
+			/*
+			 * The first search might have left us at an extent
+			 * item that ends before our target range's start, can
+			 * happen if we have holes and NO_HOLES feature enabled.
+			 */
+			if (key.offset + datal <= off) {
+				path->slots[0]++;
+				goto process_slot;
+			} else if (key.offset >= off + len) {
+				break;
+			}
+
+			size = btrfs_item_size_nr(leaf, slot);
+			read_extent_buffer(leaf, buf,
+					   btrfs_item_ptr_offset(leaf, slot),
+					   size);
+
+			btrfs_release_path(path);
+			path->leave_spinning = 0;
 
 			memcpy(&new_key, &key, sizeof(new_key));
-			new_key.objectid = inode->i_ino;
-			new_key.offset = key.offset + destoff - off;
+			new_key.objectid = btrfs_ino(inode);
+			if (off <= key.offset)
+				new_key.offset = key.offset + destoff - off;
+			else
+				new_key.offset = destoff;
+
+			/*
+			 * Deal with a hole that doesn't have an extent item
+			 * that represents it (NO_HOLES feature enabled).
+			 * This hole is either in the middle of the cloning
+			 * range or at the beginning (fully overlaps it or
+			 * partially overlaps it).
+			 */
+			if (new_key.offset != last_dest_end)
+				drop_start = last_dest_end;
+			else
+				drop_start = new_key.offset;
+
+			/*
+			 * 1 - adjusting old extent (we may have to split it)
+			 * 1 - add new extent
+			 * 1 - inode update
+			 */
+			trans = btrfs_start_transaction(root, 3);
+			if (IS_ERR(trans)) {
+				ret = PTR_ERR(trans);
+				goto out;
+			}
 
 			if (type == BTRFS_FILE_EXTENT_REG ||
 			    type == BTRFS_FILE_EXTENT_PREALLOC) {
+				/*
+				 *    a  | --- range to clone ---|  b
+				 * | ------------- extent ------------- |
+				 */
+
+				/* subtract range b */
+				if (key.offset + datal > off + len)
+					datal = off + len - key.offset;
+
+				/* subtract range a */
+				if (off > key.offset) {
+					datao += off - key.offset;
+					datal -= off - key.offset;
+				}
+
+				ret = btrfs_drop_extents(trans, root, inode,
+							 drop_start,
+							 new_key.offset + datal,
+							 1);
+				if (ret) {
+					if (ret != -EOPNOTSUPP)
+						btrfs_abort_transaction(trans,
+								root, ret);
+					btrfs_end_transaction(trans, root);
+					goto out;
+				}
+
 				ret = btrfs_insert_empty_item(trans, root, path,
 							      &new_key, size);
-				if (ret)
+				if (ret) {
+					btrfs_abort_transaction(trans, root,
+								ret);
+					btrfs_end_transaction(trans, root);
 					goto out;
+				}
 
 				leaf = path->nodes[0];
 				slot = path->slots[0];
@@ -1640,14 +3409,6 @@ static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
 				extent = btrfs_item_ptr(leaf, slot,
 						struct btrfs_file_extent_item);
 
-				if (off > key.offset) {
-					datao += off - key.offset;
-					datal -= off - key.offset;
-				}
-
-				if (key.offset + datal > off + len)
-					datal = off + len - key.offset;
-
 				/* disko == 0 means it's a hole */
 				if (!disko)
 					datao = 0;
@@ -1656,36 +3417,89 @@ static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
 							     datao);
 				btrfs_set_file_extent_num_bytes(leaf, extent,
 								datal);
+
+				/*
+				 * We need to look up the roots that point at
+				 * this bytenr and see if the new root does.  If
+				 * it does not we need to make sure we update
+				 * quotas appropriately.
+				 */
+				if (disko && root != BTRFS_I(src)->root &&
+				    disko != last_disko) {
+					no_quota = check_ref(trans, root,
+							     disko);
+					if (no_quota < 0) {
+						btrfs_abort_transaction(trans,
+									root,
+									ret);
+						btrfs_end_transaction(trans,
+								      root);
+						ret = no_quota;
+						goto out;
+					}
+				}
+
 				if (disko) {
 					inode_add_bytes(inode, datal);
 					ret = btrfs_inc_extent_ref(trans, root,
 							disko, diskl, 0,
 							root->root_key.objectid,
-							inode->i_ino,
-							new_key.offset - datao);
-					BUG_ON(ret);
+							btrfs_ino(inode),
+							new_key.offset - datao,
+							no_quota);
+					if (ret) {
+						btrfs_abort_transaction(trans,
+									root,
+									ret);
+						btrfs_end_transaction(trans,
+								      root);
+						goto out;
+
+					}
 				}
 			} else if (type == BTRFS_FILE_EXTENT_INLINE) {
 				u64 skip = 0;
 				u64 trim = 0;
+				u64 aligned_end = 0;
+
 				if (off > key.offset) {
 					skip = off - key.offset;
 					new_key.offset += skip;
 				}
 
-				if (key.offset + datal > off+len)
-					trim = key.offset + datal - (off+len);
+				if (key.offset + datal > off + len)
+					trim = key.offset + datal - (off + len);
 
 				if (comp && (skip || trim)) {
 					ret = -EINVAL;
+					btrfs_end_transaction(trans, root);
 					goto out;
 				}
 				size -= skip + trim;
 				datal -= skip + trim;
+
+				aligned_end = ALIGN(new_key.offset + datal,
+						    root->sectorsize);
+				ret = btrfs_drop_extents(trans, root, inode,
+							 drop_start,
+							 aligned_end,
+							 1);
+				if (ret) {
+					if (ret != -EOPNOTSUPP)
+						btrfs_abort_transaction(trans,
+							root, ret);
+					btrfs_end_transaction(trans, root);
+					goto out;
+				}
+
 				ret = btrfs_insert_empty_item(trans, root, path,
 							      &new_key, size);
-				if (ret)
+				if (ret) {
+					btrfs_abort_transaction(trans, root,
+								ret);
+					btrfs_end_transaction(trans, root);
 					goto out;
+				}
 
 				if (skip) {
 					u32 start =
@@ -1702,36 +3516,224 @@ static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
 				inode_add_bytes(inode, datal);
 			}
 
+			/* If we have an implicit hole (NO_HOLES feature). */
+			if (drop_start < new_key.offset)
+				clone_update_extent_map(inode, trans,
+						NULL, drop_start,
+						new_key.offset - drop_start);
+
+			clone_update_extent_map(inode, trans, path, 0, 0);
+
 			btrfs_mark_buffer_dirty(leaf);
-		}
+			btrfs_release_path(path);
 
-next:
-		btrfs_release_path(root, path);
+			last_dest_end = new_key.offset + datal;
+			ret = clone_finish_inode_update(trans, inode,
+							last_dest_end,
+							destoff, olen);
+			if (ret)
+				goto out;
+			if (new_key.offset + datal >= destoff + len)
+				break;
+		}
+		btrfs_release_path(path);
 		key.offset++;
 	}
 	ret = 0;
-out:
-	btrfs_release_path(root, path);
-	if (ret == 0) {
-		inode->i_mtime = inode->i_ctime = CURRENT_TIME;
-		if (destoff + olen > inode->i_size)
-			btrfs_i_size_write(inode, destoff + olen);
-		BTRFS_I(inode)->flags = BTRFS_I(src)->flags;
-		ret = btrfs_update_inode(trans, root, inode);
+
+	if (last_dest_end < destoff + len) {
+		/*
+		 * We have an implicit hole (NO_HOLES feature is enabled) that
+		 * fully or partially overlaps our cloning range at its end.
+		 */
+		btrfs_release_path(path);
+
+		/*
+		 * 1 - remove extent(s)
+		 * 1 - inode update
+		 */
+		trans = btrfs_start_transaction(root, 2);
+		if (IS_ERR(trans)) {
+			ret = PTR_ERR(trans);
+			goto out;
+		}
+		ret = btrfs_drop_extents(trans, root, inode,
+					 last_dest_end, destoff + len, 1);
+		if (ret) {
+			if (ret != -EOPNOTSUPP)
+				btrfs_abort_transaction(trans, root, ret);
+			btrfs_end_transaction(trans, root);
+			goto out;
+		}
+		clone_update_extent_map(inode, trans, NULL, last_dest_end,
+					destoff + len - last_dest_end);
+		ret = clone_finish_inode_update(trans, inode, destoff + len,
+						destoff, olen);
 	}
-	btrfs_end_transaction(trans, root);
-	unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
+
+out:
+	btrfs_free_path(path);
+	vfree(buf);
+	return ret;
+}
+
+static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
+				       u64 off, u64 olen, u64 destoff)
+{
+	struct inode *inode = file_inode(file);
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct fd src_file;
+	struct inode *src;
+	int ret;
+	u64 len = olen;
+	u64 bs = root->fs_info->sb->s_blocksize;
+	int same_inode = 0;
+
+	/*
+	 * TODO:
+	 * - split compressed inline extents.  annoying: we need to
+	 *   decompress into destination's address_space (the file offset
+	 *   may change, so source mapping won't do), then recompress (or
+	 *   otherwise reinsert) a subrange.
+	 *
+	 * - split destination inode's inline extents.  The inline extents can
+	 *   be either compressed or non-compressed.
+	 */
+
+	/* the destination must be opened for writing */
+	if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
+		return -EINVAL;
+
+	if (btrfs_root_readonly(root))
+		return -EROFS;
+
+	ret = mnt_want_write_file(file);
 	if (ret)
-		vmtruncate(inode, 0);
+		return ret;
+
+	src_file = fdget(srcfd);
+	if (!src_file.file) {
+		ret = -EBADF;
+		goto out_drop_write;
+	}
+
+	ret = -EXDEV;
+	if (src_file.file->f_path.mnt != file->f_path.mnt)
+		goto out_fput;
+
+	src = file_inode(src_file.file);
+
+	ret = -EINVAL;
+	if (src == inode)
+		same_inode = 1;
+
+	/* the src must be open for reading */
+	if (!(src_file.file->f_mode & FMODE_READ))
+		goto out_fput;
+
+	/* don't make the dst file partly checksummed */
+	if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
+	    (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
+		goto out_fput;
+
+	ret = -EISDIR;
+	if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
+		goto out_fput;
+
+	ret = -EXDEV;
+	if (src->i_sb != inode->i_sb)
+		goto out_fput;
+
+	if (!same_inode) {
+		if (inode < src) {
+			mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
+			mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
+		} else {
+			mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
+			mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
+		}
+	} else {
+		mutex_lock(&src->i_mutex);
+	}
+
+	/* determine range to clone */
+	ret = -EINVAL;
+	if (off + len > src->i_size || off + len < off)
+		goto out_unlock;
+	if (len == 0)
+		olen = len = src->i_size - off;
+	/* if we extend to eof, continue to block boundary */
+	if (off + len == src->i_size)
+		len = ALIGN(src->i_size, bs) - off;
+
+	/* verify the end result is block aligned */
+	if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
+	    !IS_ALIGNED(destoff, bs))
+		goto out_unlock;
+
+	/* verify if ranges are overlapped within the same file */
+	if (same_inode) {
+		if (destoff + len > off && destoff < off + len)
+			goto out_unlock;
+	}
+
+	if (destoff > inode->i_size) {
+		ret = btrfs_cont_expand(inode, inode->i_size, destoff);
+		if (ret)
+			goto out_unlock;
+	}
+
+	/*
+	 * Lock the target range too. Right after we replace the file extent
+	 * items in the fs tree (which now point to the cloned data), we might
+	 * have a worker replace them with extent items relative to a write
+	 * operation that was issued before this clone operation (i.e. confront
+	 * with inode.c:btrfs_finish_ordered_io).
+	 */
+	if (same_inode) {
+		u64 lock_start = min_t(u64, off, destoff);
+		u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
+
+		lock_extent_range(src, lock_start, lock_len);
+	} else {
+		lock_extent_range(src, off, len);
+		lock_extent_range(inode, destoff, len);
+	}
+
+	ret = btrfs_clone(src, inode, off, olen, len, destoff);
+
+	if (same_inode) {
+		u64 lock_start = min_t(u64, off, destoff);
+		u64 lock_end = max_t(u64, off, destoff) + len - 1;
+
+		unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
+	} else {
+		unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
+		unlock_extent(&BTRFS_I(inode)->io_tree, destoff,
+			      destoff + len - 1);
+	}
+	/*
+	 * Truncate page cache pages so that future reads will see the cloned
+	 * data immediately and not the previous data.
+	 */
+	truncate_inode_pages_range(&inode->i_data, destoff,
+				   PAGE_CACHE_ALIGN(destoff + len) - 1);
 out_unlock:
-	mutex_unlock(&src->i_mutex);
-	mutex_unlock(&inode->i_mutex);
-	vfree(buf);
-	btrfs_free_path(path);
+	if (!same_inode) {
+		if (inode < src) {
+			mutex_unlock(&src->i_mutex);
+			mutex_unlock(&inode->i_mutex);
+		} else {
+			mutex_unlock(&inode->i_mutex);
+			mutex_unlock(&src->i_mutex);
+		}
+	} else {
+		mutex_unlock(&src->i_mutex);
+	}
 out_fput:
-	fput(src_file);
+	fdput(src_file);
 out_drop_write:
-	mnt_drop_write(file->f_path.mnt);
+	mnt_drop_write_file(file);
 	return ret;
 }
 
@@ -1753,7 +3755,7 @@ static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
  */
 static long btrfs_ioctl_trans_start(struct file *file)
 {
-	struct inode *inode = fdentry(file)->d_inode;
+	struct inode *inode = file_inode(file);
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	struct btrfs_trans_handle *trans;
 	int ret;
@@ -1766,34 +3768,34 @@ static long btrfs_ioctl_trans_start(struct file *file)
 	if (file->private_data)
 		goto out;
 
-	ret = mnt_want_write(file->f_path.mnt);
+	ret = -EROFS;
+	if (btrfs_root_readonly(root))
+		goto out;
+
+	ret = mnt_want_write_file(file);
 	if (ret)
 		goto out;
 
-	mutex_lock(&root->fs_info->trans_mutex);
-	root->fs_info->open_ioctl_trans++;
-	mutex_unlock(&root->fs_info->trans_mutex);
+	atomic_inc(&root->fs_info->open_ioctl_trans);
 
 	ret = -ENOMEM;
-	trans = btrfs_start_ioctl_transaction(root, 0);
-	if (!trans)
+	trans = btrfs_start_ioctl_transaction(root);
+	if (IS_ERR(trans))
 		goto out_drop;
 
 	file->private_data = trans;
 	return 0;
 
 out_drop:
-	mutex_lock(&root->fs_info->trans_mutex);
-	root->fs_info->open_ioctl_trans--;
-	mutex_unlock(&root->fs_info->trans_mutex);
-	mnt_drop_write(file->f_path.mnt);
+	atomic_dec(&root->fs_info->open_ioctl_trans);
+	mnt_drop_write_file(file);
 out:
 	return ret;
 }
 
 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
 {
-	struct inode *inode = fdentry(file)->d_inode;
+	struct inode *inode = file_inode(file);
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	struct btrfs_root *new_root;
 	struct btrfs_dir_item *di;
@@ -1801,51 +3803,59 @@ static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
 	struct btrfs_path *path;
 	struct btrfs_key location;
 	struct btrfs_disk_key disk_key;
-	struct btrfs_super_block *disk_super;
-	u64 features;
 	u64 objectid = 0;
 	u64 dir_id;
+	int ret;
 
 	if (!capable(CAP_SYS_ADMIN))
 		return -EPERM;
 
-	if (copy_from_user(&objectid, argp, sizeof(objectid)))
-		return -EFAULT;
+	ret = mnt_want_write_file(file);
+	if (ret)
+		return ret;
+
+	if (copy_from_user(&objectid, argp, sizeof(objectid))) {
+		ret = -EFAULT;
+		goto out;
+	}
 
 	if (!objectid)
-		objectid = root->root_key.objectid;
+		objectid = BTRFS_FS_TREE_OBJECTID;
 
 	location.objectid = objectid;
 	location.type = BTRFS_ROOT_ITEM_KEY;
 	location.offset = (u64)-1;
 
 	new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
-	if (IS_ERR(new_root))
-		return PTR_ERR(new_root);
-
-	if (btrfs_root_refs(&new_root->root_item) == 0)
-		return -ENOENT;
+	if (IS_ERR(new_root)) {
+		ret = PTR_ERR(new_root);
+		goto out;
+	}
 
 	path = btrfs_alloc_path();
-	if (!path)
-		return -ENOMEM;
+	if (!path) {
+		ret = -ENOMEM;
+		goto out;
+	}
 	path->leave_spinning = 1;
 
 	trans = btrfs_start_transaction(root, 1);
-	if (!trans) {
+	if (IS_ERR(trans)) {
 		btrfs_free_path(path);
-		return -ENOMEM;
+		ret = PTR_ERR(trans);
+		goto out;
 	}
 
-	dir_id = btrfs_super_root_dir(&root->fs_info->super_copy);
+	dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
 	di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
 				   dir_id, "default", 7, 1);
-	if (!di) {
+	if (IS_ERR_OR_NULL(di)) {
 		btrfs_free_path(path);
 		btrfs_end_transaction(trans, root);
-		printk(KERN_ERR "Umm, you don't have the default dir item, "
-		       "this isn't going to work\n");
-		return -ENOENT;
+		btrfs_err(new_root->fs_info, "Umm, you don't have the default dir"
+			   "item, this isn't going to work");
+		ret = -ENOENT;
+		goto out;
 	}
 
 	btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
@@ -1853,46 +3863,92 @@ static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
 	btrfs_mark_buffer_dirty(path->nodes[0]);
 	btrfs_free_path(path);
 
-	disk_super = &root->fs_info->super_copy;
-	features = btrfs_super_incompat_flags(disk_super);
-	if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) {
-		features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL;
-		btrfs_set_super_incompat_flags(disk_super, features);
-	}
+	btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
 	btrfs_end_transaction(trans, root);
+out:
+	mnt_drop_write_file(file);
+	return ret;
+}
 
-	return 0;
+void btrfs_get_block_group_info(struct list_head *groups_list,
+				struct btrfs_ioctl_space_info *space)
+{
+	struct btrfs_block_group_cache *block_group;
+
+	space->total_bytes = 0;
+	space->used_bytes = 0;
+	space->flags = 0;
+	list_for_each_entry(block_group, groups_list, list) {
+		space->flags = block_group->flags;
+		space->total_bytes += block_group->key.offset;
+		space->used_bytes +=
+			btrfs_block_group_used(&block_group->item);
+	}
 }
 
-long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
+static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
 {
 	struct btrfs_ioctl_space_args space_args;
 	struct btrfs_ioctl_space_info space;
 	struct btrfs_ioctl_space_info *dest;
 	struct btrfs_ioctl_space_info *dest_orig;
-	struct btrfs_ioctl_space_info *user_dest;
+	struct btrfs_ioctl_space_info __user *user_dest;
 	struct btrfs_space_info *info;
+	u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
+		       BTRFS_BLOCK_GROUP_SYSTEM,
+		       BTRFS_BLOCK_GROUP_METADATA,
+		       BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
+	int num_types = 4;
 	int alloc_size;
 	int ret = 0;
-	int slot_count = 0;
+	u64 slot_count = 0;
+	int i, c;
 
 	if (copy_from_user(&space_args,
 			   (struct btrfs_ioctl_space_args __user *)arg,
 			   sizeof(space_args)))
 		return -EFAULT;
 
-	/* first we count slots */
-	rcu_read_lock();
-	list_for_each_entry_rcu(info, &root->fs_info->space_info, list)
-		slot_count++;
-	rcu_read_unlock();
+	for (i = 0; i < num_types; i++) {
+		struct btrfs_space_info *tmp;
+
+		info = NULL;
+		rcu_read_lock();
+		list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
+					list) {
+			if (tmp->flags == types[i]) {
+				info = tmp;
+				break;
+			}
+		}
+		rcu_read_unlock();
+
+		if (!info)
+			continue;
+
+		down_read(&info->groups_sem);
+		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
+			if (!list_empty(&info->block_groups[c]))
+				slot_count++;
+		}
+		up_read(&info->groups_sem);
+	}
+
+	/*
+	 * Global block reserve, exported as a space_info
+	 */
+	slot_count++;
 
 	/* space_slots == 0 means they are asking for a count */
 	if (space_args.space_slots == 0) {
 		space_args.total_spaces = slot_count;
 		goto out;
 	}
+
+	slot_count = min_t(u64, space_args.space_slots, slot_count);
+
 	alloc_size = sizeof(*dest) * slot_count;
+
 	/* we generally have at most 6 or so space infos, one for each raid
 	 * level.  So, a whole page should be more than enough for everyone
 	 */
@@ -1906,29 +3962,57 @@ long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
 	dest_orig = dest;
 
 	/* now we have a buffer to copy into */
-	rcu_read_lock();
-	list_for_each_entry_rcu(info, &root->fs_info->space_info, list) {
-		/* make sure we don't copy more than we allocated
-		 * in our buffer
-		 */
-		if (slot_count == 0)
-			break;
-		slot_count--;
+	for (i = 0; i < num_types; i++) {
+		struct btrfs_space_info *tmp;
 
-		/* make sure userland has enough room in their buffer */
-		if (space_args.total_spaces >= space_args.space_slots)
+		if (!slot_count)
 			break;
 
-		space.flags = info->flags;
-		space.total_bytes = info->total_bytes;
-		space.used_bytes = info->bytes_used;
+		info = NULL;
+		rcu_read_lock();
+		list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
+					list) {
+			if (tmp->flags == types[i]) {
+				info = tmp;
+				break;
+			}
+		}
+		rcu_read_unlock();
+
+		if (!info)
+			continue;
+		down_read(&info->groups_sem);
+		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
+			if (!list_empty(&info->block_groups[c])) {
+				btrfs_get_block_group_info(
+					&info->block_groups[c], &space);
+				memcpy(dest, &space, sizeof(space));
+				dest++;
+				space_args.total_spaces++;
+				slot_count--;
+			}
+			if (!slot_count)
+				break;
+		}
+		up_read(&info->groups_sem);
+	}
+
+	/*
+	 * Add global block reserve
+	 */
+	if (slot_count) {
+		struct btrfs_block_rsv *block_rsv = &root->fs_info->global_block_rsv;
+
+		spin_lock(&block_rsv->lock);
+		space.total_bytes = block_rsv->size;
+		space.used_bytes = block_rsv->size - block_rsv->reserved;
+		spin_unlock(&block_rsv->lock);
+		space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
 		memcpy(dest, &space, sizeof(space));
-		dest++;
 		space_args.total_spaces++;
 	}
-	rcu_read_unlock();
 
-	user_dest = (struct btrfs_ioctl_space_info *)
+	user_dest = (struct btrfs_ioctl_space_info __user *)
 		(arg + sizeof(struct btrfs_ioctl_space_args));
 
 	if (copy_to_user(user_dest, dest_orig, alloc_size))
@@ -1950,7 +4034,7 @@ out:
  */
 long btrfs_ioctl_trans_end(struct file *file)
 {
-	struct inode *inode = fdentry(file)->d_inode;
+	struct inode *inode = file_inode(file);
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	struct btrfs_trans_handle *trans;
 
@@ -1961,18 +4045,1200 @@ long btrfs_ioctl_trans_end(struct file *file)
 
 	btrfs_end_transaction(trans, root);
 
-	mutex_lock(&root->fs_info->trans_mutex);
-	root->fs_info->open_ioctl_trans--;
-	mutex_unlock(&root->fs_info->trans_mutex);
+	atomic_dec(&root->fs_info->open_ioctl_trans);
+
+	mnt_drop_write_file(file);
+	return 0;
+}
+
+static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
+					    void __user *argp)
+{
+	struct btrfs_trans_handle *trans;
+	u64 transid;
+	int ret;
+
+	trans = btrfs_attach_transaction_barrier(root);
+	if (IS_ERR(trans)) {
+		if (PTR_ERR(trans) != -ENOENT)
+			return PTR_ERR(trans);
 
-	mnt_drop_write(file->f_path.mnt);
+		/* No running transaction, don't bother */
+		transid = root->fs_info->last_trans_committed;
+		goto out;
+	}
+	transid = trans->transid;
+	ret = btrfs_commit_transaction_async(trans, root, 0);
+	if (ret) {
+		btrfs_end_transaction(trans, root);
+		return ret;
+	}
+out:
+	if (argp)
+		if (copy_to_user(argp, &transid, sizeof(transid)))
+			return -EFAULT;
 	return 0;
 }
 
+static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
+					   void __user *argp)
+{
+	u64 transid;
+
+	if (argp) {
+		if (copy_from_user(&transid, argp, sizeof(transid)))
+			return -EFAULT;
+	} else {
+		transid = 0;  /* current trans */
+	}
+	return btrfs_wait_for_commit(root, transid);
+}
+
+static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
+{
+	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
+	struct btrfs_ioctl_scrub_args *sa;
+	int ret;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return -EPERM;
+
+	sa = memdup_user(arg, sizeof(*sa));
+	if (IS_ERR(sa))
+		return PTR_ERR(sa);
+
+	if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
+		ret = mnt_want_write_file(file);
+		if (ret)
+			goto out;
+	}
+
+	ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
+			      &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
+			      0);
+
+	if (copy_to_user(arg, sa, sizeof(*sa)))
+		ret = -EFAULT;
+
+	if (!(sa->flags & BTRFS_SCRUB_READONLY))
+		mnt_drop_write_file(file);
+out:
+	kfree(sa);
+	return ret;
+}
+
+static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
+{
+	if (!capable(CAP_SYS_ADMIN))
+		return -EPERM;
+
+	return btrfs_scrub_cancel(root->fs_info);
+}
+
+static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
+				       void __user *arg)
+{
+	struct btrfs_ioctl_scrub_args *sa;
+	int ret;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return -EPERM;
+
+	sa = memdup_user(arg, sizeof(*sa));
+	if (IS_ERR(sa))
+		return PTR_ERR(sa);
+
+	ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
+
+	if (copy_to_user(arg, sa, sizeof(*sa)))
+		ret = -EFAULT;
+
+	kfree(sa);
+	return ret;
+}
+
+static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
+				      void __user *arg)
+{
+	struct btrfs_ioctl_get_dev_stats *sa;
+	int ret;
+
+	sa = memdup_user(arg, sizeof(*sa));
+	if (IS_ERR(sa))
+		return PTR_ERR(sa);
+
+	if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
+		kfree(sa);
+		return -EPERM;
+	}
+
+	ret = btrfs_get_dev_stats(root, sa);
+
+	if (copy_to_user(arg, sa, sizeof(*sa)))
+		ret = -EFAULT;
+
+	kfree(sa);
+	return ret;
+}
+
+static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
+{
+	struct btrfs_ioctl_dev_replace_args *p;
+	int ret;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return -EPERM;
+
+	p = memdup_user(arg, sizeof(*p));
+	if (IS_ERR(p))
+		return PTR_ERR(p);
+
+	switch (p->cmd) {
+	case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
+		if (root->fs_info->sb->s_flags & MS_RDONLY) {
+			ret = -EROFS;
+			goto out;
+		}
+		if (atomic_xchg(
+			&root->fs_info->mutually_exclusive_operation_running,
+			1)) {
+			ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
+		} else {
+			ret = btrfs_dev_replace_start(root, p);
+			atomic_set(
+			 &root->fs_info->mutually_exclusive_operation_running,
+			 0);
+		}
+		break;
+	case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
+		btrfs_dev_replace_status(root->fs_info, p);
+		ret = 0;
+		break;
+	case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
+		ret = btrfs_dev_replace_cancel(root->fs_info, p);
+		break;
+	default:
+		ret = -EINVAL;
+		break;
+	}
+
+	if (copy_to_user(arg, p, sizeof(*p)))
+		ret = -EFAULT;
+out:
+	kfree(p);
+	return ret;
+}
+
+static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
+{
+	int ret = 0;
+	int i;
+	u64 rel_ptr;
+	int size;
+	struct btrfs_ioctl_ino_path_args *ipa = NULL;
+	struct inode_fs_paths *ipath = NULL;
+	struct btrfs_path *path;
+
+	if (!capable(CAP_DAC_READ_SEARCH))
+		return -EPERM;
+
+	path = btrfs_alloc_path();
+	if (!path) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	ipa = memdup_user(arg, sizeof(*ipa));
+	if (IS_ERR(ipa)) {
+		ret = PTR_ERR(ipa);
+		ipa = NULL;
+		goto out;
+	}
+
+	size = min_t(u32, ipa->size, 4096);
+	ipath = init_ipath(size, root, path);
+	if (IS_ERR(ipath)) {
+		ret = PTR_ERR(ipath);
+		ipath = NULL;
+		goto out;
+	}
+
+	ret = paths_from_inode(ipa->inum, ipath);
+	if (ret < 0)
+		goto out;
+
+	for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
+		rel_ptr = ipath->fspath->val[i] -
+			  (u64)(unsigned long)ipath->fspath->val;
+		ipath->fspath->val[i] = rel_ptr;
+	}
+
+	ret = copy_to_user((void *)(unsigned long)ipa->fspath,
+			   (void *)(unsigned long)ipath->fspath, size);
+	if (ret) {
+		ret = -EFAULT;
+		goto out;
+	}
+
+out:
+	btrfs_free_path(path);
+	free_ipath(ipath);
+	kfree(ipa);
+
+	return ret;
+}
+
+static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
+{
+	struct btrfs_data_container *inodes = ctx;
+	const size_t c = 3 * sizeof(u64);
+
+	if (inodes->bytes_left >= c) {
+		inodes->bytes_left -= c;
+		inodes->val[inodes->elem_cnt] = inum;
+		inodes->val[inodes->elem_cnt + 1] = offset;
+		inodes->val[inodes->elem_cnt + 2] = root;
+		inodes->elem_cnt += 3;
+	} else {
+		inodes->bytes_missing += c - inodes->bytes_left;
+		inodes->bytes_left = 0;
+		inodes->elem_missed += 3;
+	}
+
+	return 0;
+}
+
+static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
+					void __user *arg)
+{
+	int ret = 0;
+	int size;
+	struct btrfs_ioctl_logical_ino_args *loi;
+	struct btrfs_data_container *inodes = NULL;
+	struct btrfs_path *path = NULL;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return -EPERM;
+
+	loi = memdup_user(arg, sizeof(*loi));
+	if (IS_ERR(loi)) {
+		ret = PTR_ERR(loi);
+		loi = NULL;
+		goto out;
+	}
+
+	path = btrfs_alloc_path();
+	if (!path) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	size = min_t(u32, loi->size, 64 * 1024);
+	inodes = init_data_container(size);
+	if (IS_ERR(inodes)) {
+		ret = PTR_ERR(inodes);
+		inodes = NULL;
+		goto out;
+	}
+
+	ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
+					  build_ino_list, inodes);
+	if (ret == -EINVAL)
+		ret = -ENOENT;
+	if (ret < 0)
+		goto out;
+
+	ret = copy_to_user((void *)(unsigned long)loi->inodes,
+			   (void *)(unsigned long)inodes, size);
+	if (ret)
+		ret = -EFAULT;
+
+out:
+	btrfs_free_path(path);
+	vfree(inodes);
+	kfree(loi);
+
+	return ret;
+}
+
+void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
+			       struct btrfs_ioctl_balance_args *bargs)
+{
+	struct btrfs_balance_control *bctl = fs_info->balance_ctl;
+
+	bargs->flags = bctl->flags;
+
+	if (atomic_read(&fs_info->balance_running))
+		bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
+	if (atomic_read(&fs_info->balance_pause_req))
+		bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
+	if (atomic_read(&fs_info->balance_cancel_req))
+		bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
+
+	memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
+	memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
+	memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
+
+	if (lock) {
+		spin_lock(&fs_info->balance_lock);
+		memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
+		spin_unlock(&fs_info->balance_lock);
+	} else {
+		memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
+	}
+}
+
+static long btrfs_ioctl_balance(struct file *file, void __user *arg)
+{
+	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
+	struct btrfs_fs_info *fs_info = root->fs_info;
+	struct btrfs_ioctl_balance_args *bargs;
+	struct btrfs_balance_control *bctl;
+	bool need_unlock; /* for mut. excl. ops lock */
+	int ret;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return -EPERM;
+
+	ret = mnt_want_write_file(file);
+	if (ret)
+		return ret;
+
+again:
+	if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
+		mutex_lock(&fs_info->volume_mutex);
+		mutex_lock(&fs_info->balance_mutex);
+		need_unlock = true;
+		goto locked;
+	}
+
+	/*
+	 * mut. excl. ops lock is locked.  Three possibilites:
+	 *   (1) some other op is running
+	 *   (2) balance is running
+	 *   (3) balance is paused -- special case (think resume)
+	 */
+	mutex_lock(&fs_info->balance_mutex);
+	if (fs_info->balance_ctl) {
+		/* this is either (2) or (3) */
+		if (!atomic_read(&fs_info->balance_running)) {
+			mutex_unlock(&fs_info->balance_mutex);
+			if (!mutex_trylock(&fs_info->volume_mutex))
+				goto again;
+			mutex_lock(&fs_info->balance_mutex);
+
+			if (fs_info->balance_ctl &&
+			    !atomic_read(&fs_info->balance_running)) {
+				/* this is (3) */
+				need_unlock = false;
+				goto locked;
+			}
+
+			mutex_unlock(&fs_info->balance_mutex);
+			mutex_unlock(&fs_info->volume_mutex);
+			goto again;
+		} else {
+			/* this is (2) */
+			mutex_unlock(&fs_info->balance_mutex);
+			ret = -EINPROGRESS;
+			goto out;
+		}
+	} else {
+		/* this is (1) */
+		mutex_unlock(&fs_info->balance_mutex);
+		ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
+		goto out;
+	}
+
+locked:
+	BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
+
+	if (arg) {
+		bargs = memdup_user(arg, sizeof(*bargs));
+		if (IS_ERR(bargs)) {
+			ret = PTR_ERR(bargs);
+			goto out_unlock;
+		}
+
+		if (bargs->flags & BTRFS_BALANCE_RESUME) {
+			if (!fs_info->balance_ctl) {
+				ret = -ENOTCONN;
+				goto out_bargs;
+			}
+
+			bctl = fs_info->balance_ctl;
+			spin_lock(&fs_info->balance_lock);
+			bctl->flags |= BTRFS_BALANCE_RESUME;
+			spin_unlock(&fs_info->balance_lock);
+
+			goto do_balance;
+		}
+	} else {
+		bargs = NULL;
+	}
+
+	if (fs_info->balance_ctl) {
+		ret = -EINPROGRESS;
+		goto out_bargs;
+	}
+
+	bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
+	if (!bctl) {
+		ret = -ENOMEM;
+		goto out_bargs;
+	}
+
+	bctl->fs_info = fs_info;
+	if (arg) {
+		memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
+		memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
+		memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
+
+		bctl->flags = bargs->flags;
+	} else {
+		/* balance everything - no filters */
+		bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
+	}
+
+do_balance:
+	/*
+	 * Ownership of bctl and mutually_exclusive_operation_running
+	 * goes to to btrfs_balance.  bctl is freed in __cancel_balance,
+	 * or, if restriper was paused all the way until unmount, in
+	 * free_fs_info.  mutually_exclusive_operation_running is
+	 * cleared in __cancel_balance.
+	 */
+	need_unlock = false;
+
+	ret = btrfs_balance(bctl, bargs);
+
+	if (arg) {
+		if (copy_to_user(arg, bargs, sizeof(*bargs)))
+			ret = -EFAULT;
+	}
+
+out_bargs:
+	kfree(bargs);
+out_unlock:
+	mutex_unlock(&fs_info->balance_mutex);
+	mutex_unlock(&fs_info->volume_mutex);
+	if (need_unlock)
+		atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
+out:
+	mnt_drop_write_file(file);
+	return ret;
+}
+
+static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
+{
+	if (!capable(CAP_SYS_ADMIN))
+		return -EPERM;
+
+	switch (cmd) {
+	case BTRFS_BALANCE_CTL_PAUSE:
+		return btrfs_pause_balance(root->fs_info);
+	case BTRFS_BALANCE_CTL_CANCEL:
+		return btrfs_cancel_balance(root->fs_info);
+	}
+
+	return -EINVAL;
+}
+
+static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
+					 void __user *arg)
+{
+	struct btrfs_fs_info *fs_info = root->fs_info;
+	struct btrfs_ioctl_balance_args *bargs;
+	int ret = 0;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return -EPERM;
+
+	mutex_lock(&fs_info->balance_mutex);
+	if (!fs_info->balance_ctl) {
+		ret = -ENOTCONN;
+		goto out;
+	}
+
+	bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
+	if (!bargs) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	update_ioctl_balance_args(fs_info, 1, bargs);
+
+	if (copy_to_user(arg, bargs, sizeof(*bargs)))
+		ret = -EFAULT;
+
+	kfree(bargs);
+out:
+	mutex_unlock(&fs_info->balance_mutex);
+	return ret;
+}
+
+static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
+{
+	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
+	struct btrfs_ioctl_quota_ctl_args *sa;
+	struct btrfs_trans_handle *trans = NULL;
+	int ret;
+	int err;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return -EPERM;
+
+	ret = mnt_want_write_file(file);
+	if (ret)
+		return ret;
+
+	sa = memdup_user(arg, sizeof(*sa));
+	if (IS_ERR(sa)) {
+		ret = PTR_ERR(sa);
+		goto drop_write;
+	}
+
+	down_write(&root->fs_info->subvol_sem);
+	trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
+	if (IS_ERR(trans)) {
+		ret = PTR_ERR(trans);
+		goto out;
+	}
+
+	switch (sa->cmd) {
+	case BTRFS_QUOTA_CTL_ENABLE:
+		ret = btrfs_quota_enable(trans, root->fs_info);
+		break;
+	case BTRFS_QUOTA_CTL_DISABLE:
+		ret = btrfs_quota_disable(trans, root->fs_info);
+		break;
+	default:
+		ret = -EINVAL;
+		break;
+	}
+
+	err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
+	if (err && !ret)
+		ret = err;
+out:
+	kfree(sa);
+	up_write(&root->fs_info->subvol_sem);
+drop_write:
+	mnt_drop_write_file(file);
+	return ret;
+}
+
+static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
+{
+	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
+	struct btrfs_ioctl_qgroup_assign_args *sa;
+	struct btrfs_trans_handle *trans;
+	int ret;
+	int err;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return -EPERM;
+
+	ret = mnt_want_write_file(file);
+	if (ret)
+		return ret;
+
+	sa = memdup_user(arg, sizeof(*sa));
+	if (IS_ERR(sa)) {
+		ret = PTR_ERR(sa);
+		goto drop_write;
+	}
+
+	trans = btrfs_join_transaction(root);
+	if (IS_ERR(trans)) {
+		ret = PTR_ERR(trans);
+		goto out;
+	}
+
+	/* FIXME: check if the IDs really exist */
+	if (sa->assign) {
+		ret = btrfs_add_qgroup_relation(trans, root->fs_info,
+						sa->src, sa->dst);
+	} else {
+		ret = btrfs_del_qgroup_relation(trans, root->fs_info,
+						sa->src, sa->dst);
+	}
+
+	err = btrfs_end_transaction(trans, root);
+	if (err && !ret)
+		ret = err;
+
+out:
+	kfree(sa);
+drop_write:
+	mnt_drop_write_file(file);
+	return ret;
+}
+
+static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
+{
+	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
+	struct btrfs_ioctl_qgroup_create_args *sa;
+	struct btrfs_trans_handle *trans;
+	int ret;
+	int err;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return -EPERM;
+
+	ret = mnt_want_write_file(file);
+	if (ret)
+		return ret;
+
+	sa = memdup_user(arg, sizeof(*sa));
+	if (IS_ERR(sa)) {
+		ret = PTR_ERR(sa);
+		goto drop_write;
+	}
+
+	if (!sa->qgroupid) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	trans = btrfs_join_transaction(root);
+	if (IS_ERR(trans)) {
+		ret = PTR_ERR(trans);
+		goto out;
+	}
+
+	/* FIXME: check if the IDs really exist */
+	if (sa->create) {
+		ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid,
+					  NULL);
+	} else {
+		ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
+	}
+
+	err = btrfs_end_transaction(trans, root);
+	if (err && !ret)
+		ret = err;
+
+out:
+	kfree(sa);
+drop_write:
+	mnt_drop_write_file(file);
+	return ret;
+}
+
+static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
+{
+	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
+	struct btrfs_ioctl_qgroup_limit_args *sa;
+	struct btrfs_trans_handle *trans;
+	int ret;
+	int err;
+	u64 qgroupid;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return -EPERM;
+
+	ret = mnt_want_write_file(file);
+	if (ret)
+		return ret;
+
+	sa = memdup_user(arg, sizeof(*sa));
+	if (IS_ERR(sa)) {
+		ret = PTR_ERR(sa);
+		goto drop_write;
+	}
+
+	trans = btrfs_join_transaction(root);
+	if (IS_ERR(trans)) {
+		ret = PTR_ERR(trans);
+		goto out;
+	}
+
+	qgroupid = sa->qgroupid;
+	if (!qgroupid) {
+		/* take the current subvol as qgroup */
+		qgroupid = root->root_key.objectid;
+	}
+
+	/* FIXME: check if the IDs really exist */
+	ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
+
+	err = btrfs_end_transaction(trans, root);
+	if (err && !ret)
+		ret = err;
+
+out:
+	kfree(sa);
+drop_write:
+	mnt_drop_write_file(file);
+	return ret;
+}
+
+static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
+{
+	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
+	struct btrfs_ioctl_quota_rescan_args *qsa;
+	int ret;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return -EPERM;
+
+	ret = mnt_want_write_file(file);
+	if (ret)
+		return ret;
+
+	qsa = memdup_user(arg, sizeof(*qsa));
+	if (IS_ERR(qsa)) {
+		ret = PTR_ERR(qsa);
+		goto drop_write;
+	}
+
+	if (qsa->flags) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	ret = btrfs_qgroup_rescan(root->fs_info);
+
+out:
+	kfree(qsa);
+drop_write:
+	mnt_drop_write_file(file);
+	return ret;
+}
+
+static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
+{
+	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
+	struct btrfs_ioctl_quota_rescan_args *qsa;
+	int ret = 0;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return -EPERM;
+
+	qsa = kzalloc(sizeof(*qsa), GFP_NOFS);
+	if (!qsa)
+		return -ENOMEM;
+
+	if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
+		qsa->flags = 1;
+		qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
+	}
+
+	if (copy_to_user(arg, qsa, sizeof(*qsa)))
+		ret = -EFAULT;
+
+	kfree(qsa);
+	return ret;
+}
+
+static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
+{
+	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return -EPERM;
+
+	return btrfs_qgroup_wait_for_completion(root->fs_info);
+}
+
+static long _btrfs_ioctl_set_received_subvol(struct file *file,
+					    struct btrfs_ioctl_received_subvol_args *sa)
+{
+	struct inode *inode = file_inode(file);
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct btrfs_root_item *root_item = &root->root_item;
+	struct btrfs_trans_handle *trans;
+	struct timespec ct = CURRENT_TIME;
+	int ret = 0;
+	int received_uuid_changed;
+
+	if (!inode_owner_or_capable(inode))
+		return -EPERM;
+
+	ret = mnt_want_write_file(file);
+	if (ret < 0)
+		return ret;
+
+	down_write(&root->fs_info->subvol_sem);
+
+	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	if (btrfs_root_readonly(root)) {
+		ret = -EROFS;
+		goto out;
+	}
+
+	/*
+	 * 1 - root item
+	 * 2 - uuid items (received uuid + subvol uuid)
+	 */
+	trans = btrfs_start_transaction(root, 3);
+	if (IS_ERR(trans)) {
+		ret = PTR_ERR(trans);
+		trans = NULL;
+		goto out;
+	}
+
+	sa->rtransid = trans->transid;
+	sa->rtime.sec = ct.tv_sec;
+	sa->rtime.nsec = ct.tv_nsec;
+
+	received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
+				       BTRFS_UUID_SIZE);
+	if (received_uuid_changed &&
+	    !btrfs_is_empty_uuid(root_item->received_uuid))
+		btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
+				    root_item->received_uuid,
+				    BTRFS_UUID_KEY_RECEIVED_SUBVOL,
+				    root->root_key.objectid);
+	memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
+	btrfs_set_root_stransid(root_item, sa->stransid);
+	btrfs_set_root_rtransid(root_item, sa->rtransid);
+	btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
+	btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
+	btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
+	btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
+
+	ret = btrfs_update_root(trans, root->fs_info->tree_root,
+				&root->root_key, &root->root_item);
+	if (ret < 0) {
+		btrfs_end_transaction(trans, root);
+		goto out;
+	}
+	if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
+		ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
+					  sa->uuid,
+					  BTRFS_UUID_KEY_RECEIVED_SUBVOL,
+					  root->root_key.objectid);
+		if (ret < 0 && ret != -EEXIST) {
+			btrfs_abort_transaction(trans, root, ret);
+			goto out;
+		}
+	}
+	ret = btrfs_commit_transaction(trans, root);
+	if (ret < 0) {
+		btrfs_abort_transaction(trans, root, ret);
+		goto out;
+	}
+
+out:
+	up_write(&root->fs_info->subvol_sem);
+	mnt_drop_write_file(file);
+	return ret;
+}
+
+#ifdef CONFIG_64BIT
+static long btrfs_ioctl_set_received_subvol_32(struct file *file,
+						void __user *arg)
+{
+	struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
+	struct btrfs_ioctl_received_subvol_args *args64 = NULL;
+	int ret = 0;
+
+	args32 = memdup_user(arg, sizeof(*args32));
+	if (IS_ERR(args32)) {
+		ret = PTR_ERR(args32);
+		args32 = NULL;
+		goto out;
+	}
+
+	args64 = kmalloc(sizeof(*args64), GFP_NOFS);
+	if (!args64) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
+	args64->stransid = args32->stransid;
+	args64->rtransid = args32->rtransid;
+	args64->stime.sec = args32->stime.sec;
+	args64->stime.nsec = args32->stime.nsec;
+	args64->rtime.sec = args32->rtime.sec;
+	args64->rtime.nsec = args32->rtime.nsec;
+	args64->flags = args32->flags;
+
+	ret = _btrfs_ioctl_set_received_subvol(file, args64);
+	if (ret)
+		goto out;
+
+	memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
+	args32->stransid = args64->stransid;
+	args32->rtransid = args64->rtransid;
+	args32->stime.sec = args64->stime.sec;
+	args32->stime.nsec = args64->stime.nsec;
+	args32->rtime.sec = args64->rtime.sec;
+	args32->rtime.nsec = args64->rtime.nsec;
+	args32->flags = args64->flags;
+
+	ret = copy_to_user(arg, args32, sizeof(*args32));
+	if (ret)
+		ret = -EFAULT;
+
+out:
+	kfree(args32);
+	kfree(args64);
+	return ret;
+}
+#endif
+
+static long btrfs_ioctl_set_received_subvol(struct file *file,
+					    void __user *arg)
+{
+	struct btrfs_ioctl_received_subvol_args *sa = NULL;
+	int ret = 0;
+
+	sa = memdup_user(arg, sizeof(*sa));
+	if (IS_ERR(sa)) {
+		ret = PTR_ERR(sa);
+		sa = NULL;
+		goto out;
+	}
+
+	ret = _btrfs_ioctl_set_received_subvol(file, sa);
+
+	if (ret)
+		goto out;
+
+	ret = copy_to_user(arg, sa, sizeof(*sa));
+	if (ret)
+		ret = -EFAULT;
+
+out:
+	kfree(sa);
+	return ret;
+}
+
+static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
+{
+	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
+	size_t len;
+	int ret;
+	char label[BTRFS_LABEL_SIZE];
+
+	spin_lock(&root->fs_info->super_lock);
+	memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
+	spin_unlock(&root->fs_info->super_lock);
+
+	len = strnlen(label, BTRFS_LABEL_SIZE);
+
+	if (len == BTRFS_LABEL_SIZE) {
+		btrfs_warn(root->fs_info,
+			"label is too long, return the first %zu bytes", --len);
+	}
+
+	ret = copy_to_user(arg, label, len);
+
+	return ret ? -EFAULT : 0;
+}
+
+static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
+{
+	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
+	struct btrfs_super_block *super_block = root->fs_info->super_copy;
+	struct btrfs_trans_handle *trans;
+	char label[BTRFS_LABEL_SIZE];
+	int ret;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return -EPERM;
+
+	if (copy_from_user(label, arg, sizeof(label)))
+		return -EFAULT;
+
+	if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
+		btrfs_err(root->fs_info, "unable to set label with more than %d bytes",
+		       BTRFS_LABEL_SIZE - 1);
+		return -EINVAL;
+	}
+
+	ret = mnt_want_write_file(file);
+	if (ret)
+		return ret;
+
+	trans = btrfs_start_transaction(root, 0);
+	if (IS_ERR(trans)) {
+		ret = PTR_ERR(trans);
+		goto out_unlock;
+	}
+
+	spin_lock(&root->fs_info->super_lock);
+	strcpy(super_block->label, label);
+	spin_unlock(&root->fs_info->super_lock);
+	ret = btrfs_commit_transaction(trans, root);
+
+out_unlock:
+	mnt_drop_write_file(file);
+	return ret;
+}
+
+#define INIT_FEATURE_FLAGS(suffix) \
+	{ .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
+	  .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
+	  .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
+
+static int btrfs_ioctl_get_supported_features(struct file *file,
+					      void __user *arg)
+{
+	static struct btrfs_ioctl_feature_flags features[3] = {
+		INIT_FEATURE_FLAGS(SUPP),
+		INIT_FEATURE_FLAGS(SAFE_SET),
+		INIT_FEATURE_FLAGS(SAFE_CLEAR)
+	};
+
+	if (copy_to_user(arg, &features, sizeof(features)))
+		return -EFAULT;
+
+	return 0;
+}
+
+static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
+{
+	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
+	struct btrfs_super_block *super_block = root->fs_info->super_copy;
+	struct btrfs_ioctl_feature_flags features;
+
+	features.compat_flags = btrfs_super_compat_flags(super_block);
+	features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
+	features.incompat_flags = btrfs_super_incompat_flags(super_block);
+
+	if (copy_to_user(arg, &features, sizeof(features)))
+		return -EFAULT;
+
+	return 0;
+}
+
+static int check_feature_bits(struct btrfs_root *root,
+			      enum btrfs_feature_set set,
+			      u64 change_mask, u64 flags, u64 supported_flags,
+			      u64 safe_set, u64 safe_clear)
+{
+	const char *type = btrfs_feature_set_names[set];
+	char *names;
+	u64 disallowed, unsupported;
+	u64 set_mask = flags & change_mask;
+	u64 clear_mask = ~flags & change_mask;
+
+	unsupported = set_mask & ~supported_flags;
+	if (unsupported) {
+		names = btrfs_printable_features(set, unsupported);
+		if (names) {
+			btrfs_warn(root->fs_info,
+			   "this kernel does not support the %s feature bit%s",
+			   names, strchr(names, ',') ? "s" : "");
+			kfree(names);
+		} else
+			btrfs_warn(root->fs_info,
+			   "this kernel does not support %s bits 0x%llx",
+			   type, unsupported);
+		return -EOPNOTSUPP;
+	}
+
+	disallowed = set_mask & ~safe_set;
+	if (disallowed) {
+		names = btrfs_printable_features(set, disallowed);
+		if (names) {
+			btrfs_warn(root->fs_info,
+			   "can't set the %s feature bit%s while mounted",
+			   names, strchr(names, ',') ? "s" : "");
+			kfree(names);
+		} else
+			btrfs_warn(root->fs_info,
+			   "can't set %s bits 0x%llx while mounted",
+			   type, disallowed);
+		return -EPERM;
+	}
+
+	disallowed = clear_mask & ~safe_clear;
+	if (disallowed) {
+		names = btrfs_printable_features(set, disallowed);
+		if (names) {
+			btrfs_warn(root->fs_info,
+			   "can't clear the %s feature bit%s while mounted",
+			   names, strchr(names, ',') ? "s" : "");
+			kfree(names);
+		} else
+			btrfs_warn(root->fs_info,
+			   "can't clear %s bits 0x%llx while mounted",
+			   type, disallowed);
+		return -EPERM;
+	}
+
+	return 0;
+}
+
+#define check_feature(root, change_mask, flags, mask_base)	\
+check_feature_bits(root, FEAT_##mask_base, change_mask, flags,	\
+		   BTRFS_FEATURE_ ## mask_base ## _SUPP,	\
+		   BTRFS_FEATURE_ ## mask_base ## _SAFE_SET,	\
+		   BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
+
+static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
+{
+	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
+	struct btrfs_super_block *super_block = root->fs_info->super_copy;
+	struct btrfs_ioctl_feature_flags flags[2];
+	struct btrfs_trans_handle *trans;
+	u64 newflags;
+	int ret;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return -EPERM;
+
+	if (copy_from_user(flags, arg, sizeof(flags)))
+		return -EFAULT;
+
+	/* Nothing to do */
+	if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
+	    !flags[0].incompat_flags)
+		return 0;
+
+	ret = check_feature(root, flags[0].compat_flags,
+			    flags[1].compat_flags, COMPAT);
+	if (ret)
+		return ret;
+
+	ret = check_feature(root, flags[0].compat_ro_flags,
+			    flags[1].compat_ro_flags, COMPAT_RO);
+	if (ret)
+		return ret;
+
+	ret = check_feature(root, flags[0].incompat_flags,
+			    flags[1].incompat_flags, INCOMPAT);
+	if (ret)
+		return ret;
+
+	trans = btrfs_start_transaction(root, 0);
+	if (IS_ERR(trans))
+		return PTR_ERR(trans);
+
+	spin_lock(&root->fs_info->super_lock);
+	newflags = btrfs_super_compat_flags(super_block);
+	newflags |= flags[0].compat_flags & flags[1].compat_flags;
+	newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
+	btrfs_set_super_compat_flags(super_block, newflags);
+
+	newflags = btrfs_super_compat_ro_flags(super_block);
+	newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
+	newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
+	btrfs_set_super_compat_ro_flags(super_block, newflags);
+
+	newflags = btrfs_super_incompat_flags(super_block);
+	newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
+	newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
+	btrfs_set_super_incompat_flags(super_block, newflags);
+	spin_unlock(&root->fs_info->super_lock);
+
+	return btrfs_commit_transaction(trans, root);
+}
+
 long btrfs_ioctl(struct file *file, unsigned int
 		cmd, unsigned long arg)
 {
-	struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
+	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
 	void __user *argp = (void __user *)arg;
 
 	switch (cmd) {
@@ -1982,12 +5248,22 @@ long btrfs_ioctl(struct file *file, unsigned int
 		return btrfs_ioctl_setflags(file, argp);
 	case FS_IOC_GETVERSION:
 		return btrfs_ioctl_getversion(file, argp);
+	case FITRIM:
+		return btrfs_ioctl_fitrim(file, argp);
 	case BTRFS_IOC_SNAP_CREATE:
 		return btrfs_ioctl_snap_create(file, argp, 0);
+	case BTRFS_IOC_SNAP_CREATE_V2:
+		return btrfs_ioctl_snap_create_v2(file, argp, 0);
 	case BTRFS_IOC_SUBVOL_CREATE:
 		return btrfs_ioctl_snap_create(file, argp, 1);
+	case BTRFS_IOC_SUBVOL_CREATE_V2:
+		return btrfs_ioctl_snap_create_v2(file, argp, 1);
 	case BTRFS_IOC_SNAP_DESTROY:
 		return btrfs_ioctl_snap_destroy(file, argp);
+	case BTRFS_IOC_SUBVOL_GETFLAGS:
+		return btrfs_ioctl_subvol_getflags(file, argp);
+	case BTRFS_IOC_SUBVOL_SETFLAGS:
+		return btrfs_ioctl_subvol_setflags(file, argp);
 	case BTRFS_IOC_DEFAULT_SUBVOL:
 		return btrfs_ioctl_default_subvol(file, argp);
 	case BTRFS_IOC_DEFRAG:
@@ -1995,13 +5271,17 @@ long btrfs_ioctl(struct file *file, unsigned int
 	case BTRFS_IOC_DEFRAG_RANGE:
 		return btrfs_ioctl_defrag(file, argp);
 	case BTRFS_IOC_RESIZE:
-		return btrfs_ioctl_resize(root, argp);
+		return btrfs_ioctl_resize(file, argp);
 	case BTRFS_IOC_ADD_DEV:
 		return btrfs_ioctl_add_dev(root, argp);
 	case BTRFS_IOC_RM_DEV:
-		return btrfs_ioctl_rm_dev(root, argp);
+		return btrfs_ioctl_rm_dev(file, argp);
+	case BTRFS_IOC_FS_INFO:
+		return btrfs_ioctl_fs_info(root, argp);
+	case BTRFS_IOC_DEV_INFO:
+		return btrfs_ioctl_dev_info(root, argp);
 	case BTRFS_IOC_BALANCE:
-		return btrfs_balance(root->fs_info->dev_root);
+		return btrfs_ioctl_balance(file, NULL);
 	case BTRFS_IOC_CLONE:
 		return btrfs_ioctl_clone(file, arg, 0, 0, 0);
 	case BTRFS_IOC_CLONE_RANGE:
@@ -2012,13 +5292,79 @@ long btrfs_ioctl(struct file *file, unsigned int
 		return btrfs_ioctl_trans_end(file);
 	case BTRFS_IOC_TREE_SEARCH:
 		return btrfs_ioctl_tree_search(file, argp);
+	case BTRFS_IOC_TREE_SEARCH_V2:
+		return btrfs_ioctl_tree_search_v2(file, argp);
 	case BTRFS_IOC_INO_LOOKUP:
 		return btrfs_ioctl_ino_lookup(file, argp);
+	case BTRFS_IOC_INO_PATHS:
+		return btrfs_ioctl_ino_to_path(root, argp);
+	case BTRFS_IOC_LOGICAL_INO:
+		return btrfs_ioctl_logical_to_ino(root, argp);
 	case BTRFS_IOC_SPACE_INFO:
 		return btrfs_ioctl_space_info(root, argp);
-	case BTRFS_IOC_SYNC:
-		btrfs_sync_fs(file->f_dentry->d_sb, 1);
-		return 0;
+	case BTRFS_IOC_SYNC: {
+		int ret;
+
+		ret = btrfs_start_delalloc_roots(root->fs_info, 0, -1);
+		if (ret)
+			return ret;
+		ret = btrfs_sync_fs(file->f_dentry->d_sb, 1);
+		return ret;
+	}
+	case BTRFS_IOC_START_SYNC:
+		return btrfs_ioctl_start_sync(root, argp);
+	case BTRFS_IOC_WAIT_SYNC:
+		return btrfs_ioctl_wait_sync(root, argp);
+	case BTRFS_IOC_SCRUB:
+		return btrfs_ioctl_scrub(file, argp);
+	case BTRFS_IOC_SCRUB_CANCEL:
+		return btrfs_ioctl_scrub_cancel(root, argp);
+	case BTRFS_IOC_SCRUB_PROGRESS:
+		return btrfs_ioctl_scrub_progress(root, argp);
+	case BTRFS_IOC_BALANCE_V2:
+		return btrfs_ioctl_balance(file, argp);
+	case BTRFS_IOC_BALANCE_CTL:
+		return btrfs_ioctl_balance_ctl(root, arg);
+	case BTRFS_IOC_BALANCE_PROGRESS:
+		return btrfs_ioctl_balance_progress(root, argp);
+	case BTRFS_IOC_SET_RECEIVED_SUBVOL:
+		return btrfs_ioctl_set_received_subvol(file, argp);
+#ifdef CONFIG_64BIT
+	case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
+		return btrfs_ioctl_set_received_subvol_32(file, argp);
+#endif
+	case BTRFS_IOC_SEND:
+		return btrfs_ioctl_send(file, argp);
+	case BTRFS_IOC_GET_DEV_STATS:
+		return btrfs_ioctl_get_dev_stats(root, argp);
+	case BTRFS_IOC_QUOTA_CTL:
+		return btrfs_ioctl_quota_ctl(file, argp);
+	case BTRFS_IOC_QGROUP_ASSIGN:
+		return btrfs_ioctl_qgroup_assign(file, argp);
+	case BTRFS_IOC_QGROUP_CREATE:
+		return btrfs_ioctl_qgroup_create(file, argp);
+	case BTRFS_IOC_QGROUP_LIMIT:
+		return btrfs_ioctl_qgroup_limit(file, argp);
+	case BTRFS_IOC_QUOTA_RESCAN:
+		return btrfs_ioctl_quota_rescan(file, argp);
+	case BTRFS_IOC_QUOTA_RESCAN_STATUS:
+		return btrfs_ioctl_quota_rescan_status(file, argp);
+	case BTRFS_IOC_QUOTA_RESCAN_WAIT:
+		return btrfs_ioctl_quota_rescan_wait(file, argp);
+	case BTRFS_IOC_DEV_REPLACE:
+		return btrfs_ioctl_dev_replace(root, argp);
+	case BTRFS_IOC_GET_FSLABEL:
+		return btrfs_ioctl_get_fslabel(file, argp);
+	case BTRFS_IOC_SET_FSLABEL:
+		return btrfs_ioctl_set_fslabel(file, argp);
+	case BTRFS_IOC_FILE_EXTENT_SAME:
+		return btrfs_ioctl_file_extent_same(file, argp);
+	case BTRFS_IOC_GET_SUPPORTED_FEATURES:
+		return btrfs_ioctl_get_supported_features(file, argp);
+	case BTRFS_IOC_GET_FEATURES:
+		return btrfs_ioctl_get_features(file, argp);
+	case BTRFS_IOC_SET_FEATURES:
+		return btrfs_ioctl_set_features(file, argp);
 	}
 
 	return -ENOTTY;
diff --git a/fs/btrfs/ioctl.h b/fs/btrfs/ioctl.h
deleted file mode 100644
index 424694aa517..00000000000
--- a/fs/btrfs/ioctl.h
+++ /dev/null
@@ -1,181 +0,0 @@
-/*
- * Copyright (C) 2007 Oracle.  All rights reserved.
- *
- * This program is free software; you can redistribute it and/or
- * modify it under the terms of the GNU General Public
- * License v2 as published by the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful,
- * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * General Public License for more details.
- *
- * You should have received a copy of the GNU General Public
- * License along with this program; if not, write to the
- * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
- * Boston, MA 021110-1307, USA.
- */
-
-#ifndef __IOCTL_
-#define __IOCTL_
-#include <linux/ioctl.h>
-
-#define BTRFS_IOCTL_MAGIC 0x94
-#define BTRFS_VOL_NAME_MAX 255
-#define BTRFS_PATH_NAME_MAX 4087
-
-/* this should be 4k */
-struct btrfs_ioctl_vol_args {
-	__s64 fd;
-	char name[BTRFS_PATH_NAME_MAX + 1];
-};
-
-#define BTRFS_INO_LOOKUP_PATH_MAX 4080
-struct btrfs_ioctl_ino_lookup_args {
-	__u64 treeid;
-	__u64 objectid;
-	char name[BTRFS_INO_LOOKUP_PATH_MAX];
-};
-
-struct btrfs_ioctl_search_key {
-	/* which root are we searching.  0 is the tree of tree roots */
-	__u64 tree_id;
-
-	/* keys returned will be >= min and <= max */
-	__u64 min_objectid;
-	__u64 max_objectid;
-
-	/* keys returned will be >= min and <= max */
-	__u64 min_offset;
-	__u64 max_offset;
-
-	/* max and min transids to search for */
-	__u64 min_transid;
-	__u64 max_transid;
-
-	/* keys returned will be >= min and <= max */
-	__u32 min_type;
-	__u32 max_type;
-
-	/*
-	 * how many items did userland ask for, and how many are we
-	 * returning
-	 */
-	__u32 nr_items;
-
-	/* align to 64 bits */
-	__u32 unused;
-
-	/* some extra for later */
-	__u64 unused1;
-	__u64 unused2;
-	__u64 unused3;
-	__u64 unused4;
-};
-
-struct btrfs_ioctl_search_header {
-	__u64 transid;
-	__u64 objectid;
-	__u64 offset;
-	__u32 type;
-	__u32 len;
-};
-
-#define BTRFS_SEARCH_ARGS_BUFSIZE (4096 - sizeof(struct btrfs_ioctl_search_key))
-/*
- * the buf is an array of search headers where
- * each header is followed by the actual item
- * the type field is expanded to 32 bits for alignment
- */
-struct btrfs_ioctl_search_args {
-	struct btrfs_ioctl_search_key key;
-	char buf[BTRFS_SEARCH_ARGS_BUFSIZE];
-};
-
-struct btrfs_ioctl_clone_range_args {
-  __s64 src_fd;
-  __u64 src_offset, src_length;
-  __u64 dest_offset;
-};
-
-/* flags for the defrag range ioctl */
-#define BTRFS_DEFRAG_RANGE_COMPRESS 1
-#define BTRFS_DEFRAG_RANGE_START_IO 2
-
-struct btrfs_ioctl_defrag_range_args {
-	/* start of the defrag operation */
-	__u64 start;
-
-	/* number of bytes to defrag, use (u64)-1 to say all */
-	__u64 len;
-
-	/*
-	 * flags for the operation, which can include turning
-	 * on compression for this one defrag
-	 */
-	__u64 flags;
-
-	/*
-	 * any extent bigger than this will be considered
-	 * already defragged.  Use 0 to take the kernel default
-	 * Use 1 to say every single extent must be rewritten
-	 */
-	__u32 extent_thresh;
-
-	/* spare for later */
-	__u32 unused[5];
-};
-
-struct btrfs_ioctl_space_info {
-	__u64 flags;
-	__u64 total_bytes;
-	__u64 used_bytes;
-};
-
-struct btrfs_ioctl_space_args {
-	__u64 space_slots;
-	__u64 total_spaces;
-	struct btrfs_ioctl_space_info spaces[0];
-};
-
-#define BTRFS_IOC_SNAP_CREATE _IOW(BTRFS_IOCTL_MAGIC, 1, \
-				   struct btrfs_ioctl_vol_args)
-#define BTRFS_IOC_DEFRAG _IOW(BTRFS_IOCTL_MAGIC, 2, \
-				   struct btrfs_ioctl_vol_args)
-#define BTRFS_IOC_RESIZE _IOW(BTRFS_IOCTL_MAGIC, 3, \
-				   struct btrfs_ioctl_vol_args)
-#define BTRFS_IOC_SCAN_DEV _IOW(BTRFS_IOCTL_MAGIC, 4, \
-				   struct btrfs_ioctl_vol_args)
-/* trans start and trans end are dangerous, and only for
- * use by applications that know how to avoid the
- * resulting deadlocks
- */
-#define BTRFS_IOC_TRANS_START  _IO(BTRFS_IOCTL_MAGIC, 6)
-#define BTRFS_IOC_TRANS_END    _IO(BTRFS_IOCTL_MAGIC, 7)
-#define BTRFS_IOC_SYNC         _IO(BTRFS_IOCTL_MAGIC, 8)
-
-#define BTRFS_IOC_CLONE        _IOW(BTRFS_IOCTL_MAGIC, 9, int)
-#define BTRFS_IOC_ADD_DEV _IOW(BTRFS_IOCTL_MAGIC, 10, \
-				   struct btrfs_ioctl_vol_args)
-#define BTRFS_IOC_RM_DEV _IOW(BTRFS_IOCTL_MAGIC, 11, \
-				   struct btrfs_ioctl_vol_args)
-#define BTRFS_IOC_BALANCE _IOW(BTRFS_IOCTL_MAGIC, 12, \
-				   struct btrfs_ioctl_vol_args)
-
-#define BTRFS_IOC_CLONE_RANGE _IOW(BTRFS_IOCTL_MAGIC, 13, \
-				  struct btrfs_ioctl_clone_range_args)
-
-#define BTRFS_IOC_SUBVOL_CREATE _IOW(BTRFS_IOCTL_MAGIC, 14, \
-				   struct btrfs_ioctl_vol_args)
-#define BTRFS_IOC_SNAP_DESTROY _IOW(BTRFS_IOCTL_MAGIC, 15, \
-				struct btrfs_ioctl_vol_args)
-#define BTRFS_IOC_DEFRAG_RANGE _IOW(BTRFS_IOCTL_MAGIC, 16, \
-				struct btrfs_ioctl_defrag_range_args)
-#define BTRFS_IOC_TREE_SEARCH _IOWR(BTRFS_IOCTL_MAGIC, 17, \
-				   struct btrfs_ioctl_search_args)
-#define BTRFS_IOC_INO_LOOKUP _IOWR(BTRFS_IOCTL_MAGIC, 18, \
-				   struct btrfs_ioctl_ino_lookup_args)
-#define BTRFS_IOC_DEFAULT_SUBVOL _IOW(BTRFS_IOCTL_MAGIC, 19, u64)
-#define BTRFS_IOC_SPACE_INFO _IOWR(BTRFS_IOCTL_MAGIC, 20, \
-				    struct btrfs_ioctl_space_args)
-#endif
diff --git a/fs/btrfs/locking.c b/fs/btrfs/locking.c
index 6151f2ea38b..5665d214924 100644
--- a/fs/btrfs/locking.c
+++ b/fs/btrfs/locking.c
@@ -24,210 +24,259 @@
 #include "extent_io.h"
 #include "locking.h"
 
-static inline void spin_nested(struct extent_buffer *eb)
-{
-	spin_lock(&eb->lock);
-}
+static void btrfs_assert_tree_read_locked(struct extent_buffer *eb);
 
 /*
- * Setting a lock to blocking will drop the spinlock and set the
- * flag that forces other procs who want the lock to wait.  After
- * this you can safely schedule with the lock held.
+ * if we currently have a spinning reader or writer lock
+ * (indicated by the rw flag) this will bump the count
+ * of blocking holders and drop the spinlock.
  */
-void btrfs_set_lock_blocking(struct extent_buffer *eb)
+void btrfs_set_lock_blocking_rw(struct extent_buffer *eb, int rw)
 {
-	if (!test_bit(EXTENT_BUFFER_BLOCKING, &eb->bflags)) {
-		set_bit(EXTENT_BUFFER_BLOCKING, &eb->bflags);
-		spin_unlock(&eb->lock);
+	/*
+	 * no lock is required.  The lock owner may change if
+	 * we have a read lock, but it won't change to or away
+	 * from us.  If we have the write lock, we are the owner
+	 * and it'll never change.
+	 */
+	if (eb->lock_nested && current->pid == eb->lock_owner)
+		return;
+	if (rw == BTRFS_WRITE_LOCK) {
+		if (atomic_read(&eb->blocking_writers) == 0) {
+			WARN_ON(atomic_read(&eb->spinning_writers) != 1);
+			atomic_dec(&eb->spinning_writers);
+			btrfs_assert_tree_locked(eb);
+			atomic_inc(&eb->blocking_writers);
+			write_unlock(&eb->lock);
+		}
+	} else if (rw == BTRFS_READ_LOCK) {
+		btrfs_assert_tree_read_locked(eb);
+		atomic_inc(&eb->blocking_readers);
+		WARN_ON(atomic_read(&eb->spinning_readers) == 0);
+		atomic_dec(&eb->spinning_readers);
+		read_unlock(&eb->lock);
 	}
-	/* exit with the spin lock released and the bit set */
+	return;
 }
 
 /*
- * clearing the blocking flag will take the spinlock again.
- * After this you can't safely schedule
+ * if we currently have a blocking lock, take the spinlock
+ * and drop our blocking count
  */
-void btrfs_clear_lock_blocking(struct extent_buffer *eb)
+void btrfs_clear_lock_blocking_rw(struct extent_buffer *eb, int rw)
 {
-	if (test_bit(EXTENT_BUFFER_BLOCKING, &eb->bflags)) {
-		spin_nested(eb);
-		clear_bit(EXTENT_BUFFER_BLOCKING, &eb->bflags);
-		smp_mb__after_clear_bit();
+	/*
+	 * no lock is required.  The lock owner may change if
+	 * we have a read lock, but it won't change to or away
+	 * from us.  If we have the write lock, we are the owner
+	 * and it'll never change.
+	 */
+	if (eb->lock_nested && current->pid == eb->lock_owner)
+		return;
+
+	if (rw == BTRFS_WRITE_LOCK_BLOCKING) {
+		BUG_ON(atomic_read(&eb->blocking_writers) != 1);
+		write_lock(&eb->lock);
+		WARN_ON(atomic_read(&eb->spinning_writers));
+		atomic_inc(&eb->spinning_writers);
+		if (atomic_dec_and_test(&eb->blocking_writers) &&
+		    waitqueue_active(&eb->write_lock_wq))
+			wake_up(&eb->write_lock_wq);
+	} else if (rw == BTRFS_READ_LOCK_BLOCKING) {
+		BUG_ON(atomic_read(&eb->blocking_readers) == 0);
+		read_lock(&eb->lock);
+		atomic_inc(&eb->spinning_readers);
+		if (atomic_dec_and_test(&eb->blocking_readers) &&
+		    waitqueue_active(&eb->read_lock_wq))
+			wake_up(&eb->read_lock_wq);
 	}
-	/* exit with the spin lock held */
+	return;
 }
 
 /*
- * unfortunately, many of the places that currently set a lock to blocking
- * don't end up blocking for very long, and often they don't block
- * at all.  For a dbench 50 run, if we don't spin on the blocking bit
- * at all, the context switch rate can jump up to 400,000/sec or more.
- *
- * So, we're still stuck with this crummy spin on the blocking bit,
- * at least until the most common causes of the short blocks
- * can be dealt with.
+ * take a spinning read lock.  This will wait for any blocking
+ * writers
  */
-static int btrfs_spin_on_block(struct extent_buffer *eb)
+void btrfs_tree_read_lock(struct extent_buffer *eb)
 {
-	int i;
-
-	for (i = 0; i < 512; i++) {
-		if (!test_bit(EXTENT_BUFFER_BLOCKING, &eb->bflags))
-			return 1;
-		if (need_resched())
-			break;
-		cpu_relax();
+again:
+	BUG_ON(!atomic_read(&eb->blocking_writers) &&
+	       current->pid == eb->lock_owner);
+
+	read_lock(&eb->lock);
+	if (atomic_read(&eb->blocking_writers) &&
+	    current->pid == eb->lock_owner) {
+		/*
+		 * This extent is already write-locked by our thread. We allow
+		 * an additional read lock to be added because it's for the same
+		 * thread. btrfs_find_all_roots() depends on this as it may be
+		 * called on a partly (write-)locked tree.
+		 */
+		BUG_ON(eb->lock_nested);
+		eb->lock_nested = 1;
+		read_unlock(&eb->lock);
+		return;
 	}
-	return 0;
+	if (atomic_read(&eb->blocking_writers)) {
+		read_unlock(&eb->lock);
+		wait_event(eb->write_lock_wq,
+			   atomic_read(&eb->blocking_writers) == 0);
+		goto again;
+	}
+	atomic_inc(&eb->read_locks);
+	atomic_inc(&eb->spinning_readers);
 }
 
 /*
- * This is somewhat different from trylock.  It will take the
- * spinlock but if it finds the lock is set to blocking, it will
- * return without the lock held.
- *
- * returns 1 if it was able to take the lock and zero otherwise
- *
- * After this call, scheduling is not safe without first calling
- * btrfs_set_lock_blocking()
+ * returns 1 if we get the read lock and 0 if we don't
+ * this won't wait for blocking writers
  */
-int btrfs_try_spin_lock(struct extent_buffer *eb)
+int btrfs_try_tree_read_lock(struct extent_buffer *eb)
 {
-	int i;
+	if (atomic_read(&eb->blocking_writers))
+		return 0;
 
-	if (btrfs_spin_on_block(eb)) {
-		spin_nested(eb);
-		if (!test_bit(EXTENT_BUFFER_BLOCKING, &eb->bflags))
-			return 1;
-		spin_unlock(&eb->lock);
-	}
-	/* spin for a bit on the BLOCKING flag */
-	for (i = 0; i < 2; i++) {
-		cpu_relax();
-		if (!btrfs_spin_on_block(eb))
-			break;
-
-		spin_nested(eb);
-		if (!test_bit(EXTENT_BUFFER_BLOCKING, &eb->bflags))
-			return 1;
-		spin_unlock(&eb->lock);
+	if (!read_trylock(&eb->lock))
+		return 0;
+
+	if (atomic_read(&eb->blocking_writers)) {
+		read_unlock(&eb->lock);
+		return 0;
 	}
-	return 0;
+	atomic_inc(&eb->read_locks);
+	atomic_inc(&eb->spinning_readers);
+	return 1;
 }
 
 /*
- * the autoremove wake function will return 0 if it tried to wake up
- * a process that was already awake, which means that process won't
- * count as an exclusive wakeup.  The waitq code will continue waking
- * procs until it finds one that was actually sleeping.
- *
- * For btrfs, this isn't quite what we want.  We want a single proc
- * to be notified that the lock is ready for taking.  If that proc
- * already happen to be awake, great, it will loop around and try for
- * the lock.
- *
- * So, btrfs_wake_function always returns 1, even when the proc that we
- * tried to wake up was already awake.
+ * returns 1 if we get the read lock and 0 if we don't
+ * this won't wait for blocking writers or readers
  */
-static int btrfs_wake_function(wait_queue_t *wait, unsigned mode,
-			       int sync, void *key)
+int btrfs_try_tree_write_lock(struct extent_buffer *eb)
 {
-	autoremove_wake_function(wait, mode, sync, key);
+	if (atomic_read(&eb->blocking_writers) ||
+	    atomic_read(&eb->blocking_readers))
+		return 0;
+
+	if (!write_trylock(&eb->lock))
+		return 0;
+
+	if (atomic_read(&eb->blocking_writers) ||
+	    atomic_read(&eb->blocking_readers)) {
+		write_unlock(&eb->lock);
+		return 0;
+	}
+	atomic_inc(&eb->write_locks);
+	atomic_inc(&eb->spinning_writers);
+	eb->lock_owner = current->pid;
 	return 1;
 }
 
 /*
- * returns with the extent buffer spinlocked.
- *
- * This will spin and/or wait as required to take the lock, and then
- * return with the spinlock held.
- *
- * After this call, scheduling is not safe without first calling
- * btrfs_set_lock_blocking()
+ * drop a spinning read lock
  */
-int btrfs_tree_lock(struct extent_buffer *eb)
+void btrfs_tree_read_unlock(struct extent_buffer *eb)
 {
-	DEFINE_WAIT(wait);
-	wait.func = btrfs_wake_function;
-
-	if (!btrfs_spin_on_block(eb))
-		goto sleep;
-
-	while(1) {
-		spin_nested(eb);
-
-		/* nobody is blocking, exit with the spinlock held */
-		if (!test_bit(EXTENT_BUFFER_BLOCKING, &eb->bflags))
-			return 0;
-
-		/*
-		 * we have the spinlock, but the real owner is blocking.
-		 * wait for them
-		 */
-		spin_unlock(&eb->lock);
-
-		/*
-		 * spin for a bit, and if the blocking flag goes away,
-		 * loop around
-		 */
-		cpu_relax();
-		if (btrfs_spin_on_block(eb))
-			continue;
-sleep:
-		prepare_to_wait_exclusive(&eb->lock_wq, &wait,
-					  TASK_UNINTERRUPTIBLE);
-
-		if (test_bit(EXTENT_BUFFER_BLOCKING, &eb->bflags))
-			schedule();
+	/*
+	 * if we're nested, we have the write lock.  No new locking
+	 * is needed as long as we are the lock owner.
+	 * The write unlock will do a barrier for us, and the lock_nested
+	 * field only matters to the lock owner.
+	 */
+	if (eb->lock_nested && current->pid == eb->lock_owner) {
+		eb->lock_nested = 0;
+		return;
+	}
+	btrfs_assert_tree_read_locked(eb);
+	WARN_ON(atomic_read(&eb->spinning_readers) == 0);
+	atomic_dec(&eb->spinning_readers);
+	atomic_dec(&eb->read_locks);
+	read_unlock(&eb->lock);
+}
 
-		finish_wait(&eb->lock_wq, &wait);
+/*
+ * drop a blocking read lock
+ */
+void btrfs_tree_read_unlock_blocking(struct extent_buffer *eb)
+{
+	/*
+	 * if we're nested, we have the write lock.  No new locking
+	 * is needed as long as we are the lock owner.
+	 * The write unlock will do a barrier for us, and the lock_nested
+	 * field only matters to the lock owner.
+	 */
+	if (eb->lock_nested && current->pid == eb->lock_owner) {
+		eb->lock_nested = 0;
+		return;
 	}
-	return 0;
+	btrfs_assert_tree_read_locked(eb);
+	WARN_ON(atomic_read(&eb->blocking_readers) == 0);
+	if (atomic_dec_and_test(&eb->blocking_readers) &&
+	    waitqueue_active(&eb->read_lock_wq))
+		wake_up(&eb->read_lock_wq);
+	atomic_dec(&eb->read_locks);
 }
 
 /*
- * Very quick trylock, this does not spin or schedule.  It returns
- * 1 with the spinlock held if it was able to take the lock, or it
- * returns zero if it was unable to take the lock.
- *
- * After this call, scheduling is not safe without first calling
- * btrfs_set_lock_blocking()
+ * take a spinning write lock.  This will wait for both
+ * blocking readers or writers
  */
-int btrfs_try_tree_lock(struct extent_buffer *eb)
+void btrfs_tree_lock(struct extent_buffer *eb)
 {
-	if (spin_trylock(&eb->lock)) {
-		if (test_bit(EXTENT_BUFFER_BLOCKING, &eb->bflags)) {
-			/*
-			 * we've got the spinlock, but the real owner is
-			 * blocking.  Drop the spinlock and return failure
-			 */
-			spin_unlock(&eb->lock);
-			return 0;
-		}
-		return 1;
+again:
+	wait_event(eb->read_lock_wq, atomic_read(&eb->blocking_readers) == 0);
+	wait_event(eb->write_lock_wq, atomic_read(&eb->blocking_writers) == 0);
+	write_lock(&eb->lock);
+	if (atomic_read(&eb->blocking_readers)) {
+		write_unlock(&eb->lock);
+		wait_event(eb->read_lock_wq,
+			   atomic_read(&eb->blocking_readers) == 0);
+		goto again;
 	}
-	/* someone else has the spinlock giveup */
-	return 0;
+	if (atomic_read(&eb->blocking_writers)) {
+		write_unlock(&eb->lock);
+		wait_event(eb->write_lock_wq,
+			   atomic_read(&eb->blocking_writers) == 0);
+		goto again;
+	}
+	WARN_ON(atomic_read(&eb->spinning_writers));
+	atomic_inc(&eb->spinning_writers);
+	atomic_inc(&eb->write_locks);
+	eb->lock_owner = current->pid;
 }
 
-int btrfs_tree_unlock(struct extent_buffer *eb)
+/*
+ * drop a spinning or a blocking write lock.
+ */
+void btrfs_tree_unlock(struct extent_buffer *eb)
 {
-	/*
-	 * if we were a blocking owner, we don't have the spinlock held
-	 * just clear the bit and look for waiters
-	 */
-	if (test_and_clear_bit(EXTENT_BUFFER_BLOCKING, &eb->bflags))
-		smp_mb__after_clear_bit();
-	else
-		spin_unlock(&eb->lock);
-
-	if (waitqueue_active(&eb->lock_wq))
-		wake_up(&eb->lock_wq);
-	return 0;
+	int blockers = atomic_read(&eb->blocking_writers);
+
+	BUG_ON(blockers > 1);
+
+	btrfs_assert_tree_locked(eb);
+	eb->lock_owner = 0;
+	atomic_dec(&eb->write_locks);
+
+	if (blockers) {
+		WARN_ON(atomic_read(&eb->spinning_writers));
+		atomic_dec(&eb->blocking_writers);
+		smp_mb();
+		if (waitqueue_active(&eb->write_lock_wq))
+			wake_up(&eb->write_lock_wq);
+	} else {
+		WARN_ON(atomic_read(&eb->spinning_writers) != 1);
+		atomic_dec(&eb->spinning_writers);
+		write_unlock(&eb->lock);
+	}
 }
 
 void btrfs_assert_tree_locked(struct extent_buffer *eb)
 {
-	if (!test_bit(EXTENT_BUFFER_BLOCKING, &eb->bflags))
-		assert_spin_locked(&eb->lock);
+	BUG_ON(!atomic_read(&eb->write_locks));
+}
+
+static void btrfs_assert_tree_read_locked(struct extent_buffer *eb)
+{
+	BUG_ON(!atomic_read(&eb->read_locks));
 }
diff --git a/fs/btrfs/locking.h b/fs/btrfs/locking.h
index 6c4ce457168..b81e0e9a489 100644
--- a/fs/btrfs/locking.h
+++ b/fs/btrfs/locking.h
@@ -19,13 +19,42 @@
 #ifndef __BTRFS_LOCKING_
 #define __BTRFS_LOCKING_
 
-int btrfs_tree_lock(struct extent_buffer *eb);
-int btrfs_tree_unlock(struct extent_buffer *eb);
+#define BTRFS_WRITE_LOCK 1
+#define BTRFS_READ_LOCK 2
+#define BTRFS_WRITE_LOCK_BLOCKING 3
+#define BTRFS_READ_LOCK_BLOCKING 4
 
-int btrfs_try_tree_lock(struct extent_buffer *eb);
-int btrfs_try_spin_lock(struct extent_buffer *eb);
+void btrfs_tree_lock(struct extent_buffer *eb);
+void btrfs_tree_unlock(struct extent_buffer *eb);
 
-void btrfs_set_lock_blocking(struct extent_buffer *eb);
-void btrfs_clear_lock_blocking(struct extent_buffer *eb);
+void btrfs_tree_read_lock(struct extent_buffer *eb);
+void btrfs_tree_read_unlock(struct extent_buffer *eb);
+void btrfs_tree_read_unlock_blocking(struct extent_buffer *eb);
+void btrfs_set_lock_blocking_rw(struct extent_buffer *eb, int rw);
+void btrfs_clear_lock_blocking_rw(struct extent_buffer *eb, int rw);
 void btrfs_assert_tree_locked(struct extent_buffer *eb);
+int btrfs_try_tree_read_lock(struct extent_buffer *eb);
+int btrfs_try_tree_write_lock(struct extent_buffer *eb);
+
+static inline void btrfs_tree_unlock_rw(struct extent_buffer *eb, int rw)
+{
+	if (rw == BTRFS_WRITE_LOCK || rw == BTRFS_WRITE_LOCK_BLOCKING)
+		btrfs_tree_unlock(eb);
+	else if (rw == BTRFS_READ_LOCK_BLOCKING)
+		btrfs_tree_read_unlock_blocking(eb);
+	else if (rw == BTRFS_READ_LOCK)
+		btrfs_tree_read_unlock(eb);
+	else
+		BUG();
+}
+
+static inline void btrfs_set_lock_blocking(struct extent_buffer *eb)
+{
+	btrfs_set_lock_blocking_rw(eb, BTRFS_WRITE_LOCK);
+}
+
+static inline void btrfs_clear_lock_blocking(struct extent_buffer *eb)
+{
+	btrfs_clear_lock_blocking_rw(eb, BTRFS_WRITE_LOCK_BLOCKING);
+}
 #endif
diff --git a/fs/btrfs/lzo.c b/fs/btrfs/lzo.c
new file mode 100644
index 00000000000..dfad8514f0d
--- /dev/null
+++ b/fs/btrfs/lzo.c
@@ -0,0 +1,429 @@
+/*
+ * Copyright (C) 2008 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/kernel.h>
+#include <linux/slab.h>
+#include <linux/vmalloc.h>
+#include <linux/init.h>
+#include <linux/err.h>
+#include <linux/sched.h>
+#include <linux/pagemap.h>
+#include <linux/bio.h>
+#include <linux/lzo.h>
+#include "compression.h"
+
+#define LZO_LEN	4
+
+struct workspace {
+	void *mem;
+	void *buf;	/* where decompressed data goes */
+	void *cbuf;	/* where compressed data goes */
+	struct list_head list;
+};
+
+static void lzo_free_workspace(struct list_head *ws)
+{
+	struct workspace *workspace = list_entry(ws, struct workspace, list);
+
+	vfree(workspace->buf);
+	vfree(workspace->cbuf);
+	vfree(workspace->mem);
+	kfree(workspace);
+}
+
+static struct list_head *lzo_alloc_workspace(void)
+{
+	struct workspace *workspace;
+
+	workspace = kzalloc(sizeof(*workspace), GFP_NOFS);
+	if (!workspace)
+		return ERR_PTR(-ENOMEM);
+
+	workspace->mem = vmalloc(LZO1X_MEM_COMPRESS);
+	workspace->buf = vmalloc(lzo1x_worst_compress(PAGE_CACHE_SIZE));
+	workspace->cbuf = vmalloc(lzo1x_worst_compress(PAGE_CACHE_SIZE));
+	if (!workspace->mem || !workspace->buf || !workspace->cbuf)
+		goto fail;
+
+	INIT_LIST_HEAD(&workspace->list);
+
+	return &workspace->list;
+fail:
+	lzo_free_workspace(&workspace->list);
+	return ERR_PTR(-ENOMEM);
+}
+
+static inline void write_compress_length(char *buf, size_t len)
+{
+	__le32 dlen;
+
+	dlen = cpu_to_le32(len);
+	memcpy(buf, &dlen, LZO_LEN);
+}
+
+static inline size_t read_compress_length(char *buf)
+{
+	__le32 dlen;
+
+	memcpy(&dlen, buf, LZO_LEN);
+	return le32_to_cpu(dlen);
+}
+
+static int lzo_compress_pages(struct list_head *ws,
+			      struct address_space *mapping,
+			      u64 start, unsigned long len,
+			      struct page **pages,
+			      unsigned long nr_dest_pages,
+			      unsigned long *out_pages,
+			      unsigned long *total_in,
+			      unsigned long *total_out,
+			      unsigned long max_out)
+{
+	struct workspace *workspace = list_entry(ws, struct workspace, list);
+	int ret = 0;
+	char *data_in;
+	char *cpage_out;
+	int nr_pages = 0;
+	struct page *in_page = NULL;
+	struct page *out_page = NULL;
+	unsigned long bytes_left;
+
+	size_t in_len;
+	size_t out_len;
+	char *buf;
+	unsigned long tot_in = 0;
+	unsigned long tot_out = 0;
+	unsigned long pg_bytes_left;
+	unsigned long out_offset;
+	unsigned long bytes;
+
+	*out_pages = 0;
+	*total_out = 0;
+	*total_in = 0;
+
+	in_page = find_get_page(mapping, start >> PAGE_CACHE_SHIFT);
+	data_in = kmap(in_page);
+
+	/*
+	 * store the size of all chunks of compressed data in
+	 * the first 4 bytes
+	 */
+	out_page = alloc_page(GFP_NOFS | __GFP_HIGHMEM);
+	if (out_page == NULL) {
+		ret = -ENOMEM;
+		goto out;
+	}
+	cpage_out = kmap(out_page);
+	out_offset = LZO_LEN;
+	tot_out = LZO_LEN;
+	pages[0] = out_page;
+	nr_pages = 1;
+	pg_bytes_left = PAGE_CACHE_SIZE - LZO_LEN;
+
+	/* compress at most one page of data each time */
+	in_len = min(len, PAGE_CACHE_SIZE);
+	while (tot_in < len) {
+		ret = lzo1x_1_compress(data_in, in_len, workspace->cbuf,
+				       &out_len, workspace->mem);
+		if (ret != LZO_E_OK) {
+			printk(KERN_DEBUG "BTRFS: deflate in loop returned %d\n",
+			       ret);
+			ret = -EIO;
+			goto out;
+		}
+
+		/* store the size of this chunk of compressed data */
+		write_compress_length(cpage_out + out_offset, out_len);
+		tot_out += LZO_LEN;
+		out_offset += LZO_LEN;
+		pg_bytes_left -= LZO_LEN;
+
+		tot_in += in_len;
+		tot_out += out_len;
+
+		/* copy bytes from the working buffer into the pages */
+		buf = workspace->cbuf;
+		while (out_len) {
+			bytes = min_t(unsigned long, pg_bytes_left, out_len);
+
+			memcpy(cpage_out + out_offset, buf, bytes);
+
+			out_len -= bytes;
+			pg_bytes_left -= bytes;
+			buf += bytes;
+			out_offset += bytes;
+
+			/*
+			 * we need another page for writing out.
+			 *
+			 * Note if there's less than 4 bytes left, we just
+			 * skip to a new page.
+			 */
+			if ((out_len == 0 && pg_bytes_left < LZO_LEN) ||
+			    pg_bytes_left == 0) {
+				if (pg_bytes_left) {
+					memset(cpage_out + out_offset, 0,
+					       pg_bytes_left);
+					tot_out += pg_bytes_left;
+				}
+
+				/* we're done, don't allocate new page */
+				if (out_len == 0 && tot_in >= len)
+					break;
+
+				kunmap(out_page);
+				if (nr_pages == nr_dest_pages) {
+					out_page = NULL;
+					ret = -E2BIG;
+					goto out;
+				}
+
+				out_page = alloc_page(GFP_NOFS | __GFP_HIGHMEM);
+				if (out_page == NULL) {
+					ret = -ENOMEM;
+					goto out;
+				}
+				cpage_out = kmap(out_page);
+				pages[nr_pages++] = out_page;
+
+				pg_bytes_left = PAGE_CACHE_SIZE;
+				out_offset = 0;
+			}
+		}
+
+		/* we're making it bigger, give up */
+		if (tot_in > 8192 && tot_in < tot_out) {
+			ret = -E2BIG;
+			goto out;
+		}
+
+		/* we're all done */
+		if (tot_in >= len)
+			break;
+
+		if (tot_out > max_out)
+			break;
+
+		bytes_left = len - tot_in;
+		kunmap(in_page);
+		page_cache_release(in_page);
+
+		start += PAGE_CACHE_SIZE;
+		in_page = find_get_page(mapping, start >> PAGE_CACHE_SHIFT);
+		data_in = kmap(in_page);
+		in_len = min(bytes_left, PAGE_CACHE_SIZE);
+	}
+
+	if (tot_out > tot_in)
+		goto out;
+
+	/* store the size of all chunks of compressed data */
+	cpage_out = kmap(pages[0]);
+	write_compress_length(cpage_out, tot_out);
+
+	kunmap(pages[0]);
+
+	ret = 0;
+	*total_out = tot_out;
+	*total_in = tot_in;
+out:
+	*out_pages = nr_pages;
+	if (out_page)
+		kunmap(out_page);
+
+	if (in_page) {
+		kunmap(in_page);
+		page_cache_release(in_page);
+	}
+
+	return ret;
+}
+
+static int lzo_decompress_biovec(struct list_head *ws,
+				 struct page **pages_in,
+				 u64 disk_start,
+				 struct bio_vec *bvec,
+				 int vcnt,
+				 size_t srclen)
+{
+	struct workspace *workspace = list_entry(ws, struct workspace, list);
+	int ret = 0, ret2;
+	char *data_in;
+	unsigned long page_in_index = 0;
+	unsigned long page_out_index = 0;
+	unsigned long total_pages_in = (srclen + PAGE_CACHE_SIZE - 1) /
+					PAGE_CACHE_SIZE;
+	unsigned long buf_start;
+	unsigned long buf_offset = 0;
+	unsigned long bytes;
+	unsigned long working_bytes;
+	unsigned long pg_offset;
+
+	size_t in_len;
+	size_t out_len;
+	unsigned long in_offset;
+	unsigned long in_page_bytes_left;
+	unsigned long tot_in;
+	unsigned long tot_out;
+	unsigned long tot_len;
+	char *buf;
+	bool may_late_unmap, need_unmap;
+
+	data_in = kmap(pages_in[0]);
+	tot_len = read_compress_length(data_in);
+
+	tot_in = LZO_LEN;
+	in_offset = LZO_LEN;
+	tot_len = min_t(size_t, srclen, tot_len);
+	in_page_bytes_left = PAGE_CACHE_SIZE - LZO_LEN;
+
+	tot_out = 0;
+	pg_offset = 0;
+
+	while (tot_in < tot_len) {
+		in_len = read_compress_length(data_in + in_offset);
+		in_page_bytes_left -= LZO_LEN;
+		in_offset += LZO_LEN;
+		tot_in += LZO_LEN;
+
+		tot_in += in_len;
+		working_bytes = in_len;
+		may_late_unmap = need_unmap = false;
+
+		/* fast path: avoid using the working buffer */
+		if (in_page_bytes_left >= in_len) {
+			buf = data_in + in_offset;
+			bytes = in_len;
+			may_late_unmap = true;
+			goto cont;
+		}
+
+		/* copy bytes from the pages into the working buffer */
+		buf = workspace->cbuf;
+		buf_offset = 0;
+		while (working_bytes) {
+			bytes = min(working_bytes, in_page_bytes_left);
+
+			memcpy(buf + buf_offset, data_in + in_offset, bytes);
+			buf_offset += bytes;
+cont:
+			working_bytes -= bytes;
+			in_page_bytes_left -= bytes;
+			in_offset += bytes;
+
+			/* check if we need to pick another page */
+			if ((working_bytes == 0 && in_page_bytes_left < LZO_LEN)
+			    || in_page_bytes_left == 0) {
+				tot_in += in_page_bytes_left;
+
+				if (working_bytes == 0 && tot_in >= tot_len)
+					break;
+
+				if (page_in_index + 1 >= total_pages_in) {
+					ret = -EIO;
+					goto done;
+				}
+
+				if (may_late_unmap)
+					need_unmap = true;
+				else
+					kunmap(pages_in[page_in_index]);
+
+				data_in = kmap(pages_in[++page_in_index]);
+
+				in_page_bytes_left = PAGE_CACHE_SIZE;
+				in_offset = 0;
+			}
+		}
+
+		out_len = lzo1x_worst_compress(PAGE_CACHE_SIZE);
+		ret = lzo1x_decompress_safe(buf, in_len, workspace->buf,
+					    &out_len);
+		if (need_unmap)
+			kunmap(pages_in[page_in_index - 1]);
+		if (ret != LZO_E_OK) {
+			printk(KERN_WARNING "BTRFS: decompress failed\n");
+			ret = -EIO;
+			break;
+		}
+
+		buf_start = tot_out;
+		tot_out += out_len;
+
+		ret2 = btrfs_decompress_buf2page(workspace->buf, buf_start,
+						 tot_out, disk_start,
+						 bvec, vcnt,
+						 &page_out_index, &pg_offset);
+		if (ret2 == 0)
+			break;
+	}
+done:
+	kunmap(pages_in[page_in_index]);
+	return ret;
+}
+
+static int lzo_decompress(struct list_head *ws, unsigned char *data_in,
+			  struct page *dest_page,
+			  unsigned long start_byte,
+			  size_t srclen, size_t destlen)
+{
+	struct workspace *workspace = list_entry(ws, struct workspace, list);
+	size_t in_len;
+	size_t out_len;
+	size_t tot_len;
+	int ret = 0;
+	char *kaddr;
+	unsigned long bytes;
+
+	BUG_ON(srclen < LZO_LEN);
+
+	tot_len = read_compress_length(data_in);
+	data_in += LZO_LEN;
+
+	in_len = read_compress_length(data_in);
+	data_in += LZO_LEN;
+
+	out_len = PAGE_CACHE_SIZE;
+	ret = lzo1x_decompress_safe(data_in, in_len, workspace->buf, &out_len);
+	if (ret != LZO_E_OK) {
+		printk(KERN_WARNING "BTRFS: decompress failed!\n");
+		ret = -EIO;
+		goto out;
+	}
+
+	if (out_len < start_byte) {
+		ret = -EIO;
+		goto out;
+	}
+
+	bytes = min_t(unsigned long, destlen, out_len - start_byte);
+
+	kaddr = kmap_atomic(dest_page);
+	memcpy(kaddr, workspace->buf + start_byte, bytes);
+	kunmap_atomic(kaddr);
+out:
+	return ret;
+}
+
+struct btrfs_compress_op btrfs_lzo_compress = {
+	.alloc_workspace	= lzo_alloc_workspace,
+	.free_workspace		= lzo_free_workspace,
+	.compress_pages		= lzo_compress_pages,
+	.decompress_biovec	= lzo_decompress_biovec,
+	.decompress		= lzo_decompress,
+};
diff --git a/fs/btrfs/math.h b/fs/btrfs/math.h
new file mode 100644
index 00000000000..b7816cefbd1
--- /dev/null
+++ b/fs/btrfs/math.h
@@ -0,0 +1,44 @@
+
+/*
+ * Copyright (C) 2012 Fujitsu.  All rights reserved.
+ * Written by Miao Xie <miaox@cn.fujitsu.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef __BTRFS_MATH_H
+#define __BTRFS_MATH_H
+
+#include <asm/div64.h>
+
+static inline u64 div_factor(u64 num, int factor)
+{
+	if (factor == 10)
+		return num;
+	num *= factor;
+	do_div(num, 10);
+	return num;
+}
+
+static inline u64 div_factor_fine(u64 num, int factor)
+{
+	if (factor == 100)
+		return num;
+	num *= factor;
+	do_div(num, 100);
+	return num;
+}
+
+#endif
diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c
index a127c0ebb2d..7187b14faa6 100644
--- a/fs/btrfs/ordered-data.c
+++ b/fs/btrfs/ordered-data.c
@@ -24,6 +24,9 @@
 #include "transaction.h"
 #include "btrfs_inode.h"
 #include "extent_io.h"
+#include "disk-io.h"
+
+static struct kmem_cache *btrfs_ordered_extent_cache;
 
 static u64 entry_end(struct btrfs_ordered_extent *entry)
 {
@@ -59,6 +62,14 @@ static struct rb_node *tree_insert(struct rb_root *root, u64 file_offset,
 	return NULL;
 }
 
+static void ordered_data_tree_panic(struct inode *inode, int errno,
+					       u64 offset)
+{
+	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
+	btrfs_panic(fs_info, errno, "Inconsistency in ordered tree at offset "
+		    "%llu", offset);
+}
+
 /*
  * look for a given offset in the tree, and if it can't be found return the
  * first lesser offset
@@ -124,6 +135,15 @@ static int offset_in_entry(struct btrfs_ordered_extent *entry, u64 file_offset)
 	return 1;
 }
 
+static int range_overlaps(struct btrfs_ordered_extent *entry, u64 file_offset,
+			  u64 len)
+{
+	if (file_offset + len <= entry->file_offset ||
+	    entry->file_offset + entry->len <= file_offset)
+		return 0;
+	return 1;
+}
+
 /*
  * look find the first ordered struct that has this offset, otherwise
  * the first one less than this offset
@@ -132,7 +152,7 @@ static inline struct rb_node *tree_search(struct btrfs_ordered_inode_tree *tree,
 					  u64 file_offset)
 {
 	struct rb_root *root = &tree->tree;
-	struct rb_node *prev;
+	struct rb_node *prev = NULL;
 	struct rb_node *ret;
 	struct btrfs_ordered_extent *entry;
 
@@ -161,64 +181,188 @@ static inline struct rb_node *tree_search(struct btrfs_ordered_inode_tree *tree,
  * The tree is given a single reference on the ordered extent that was
  * inserted.
  */
-int btrfs_add_ordered_extent(struct inode *inode, u64 file_offset,
-			     u64 start, u64 len, u64 disk_len, int type)
+static int __btrfs_add_ordered_extent(struct inode *inode, u64 file_offset,
+				      u64 start, u64 len, u64 disk_len,
+				      int type, int dio, int compress_type)
 {
+	struct btrfs_root *root = BTRFS_I(inode)->root;
 	struct btrfs_ordered_inode_tree *tree;
 	struct rb_node *node;
 	struct btrfs_ordered_extent *entry;
 
 	tree = &BTRFS_I(inode)->ordered_tree;
-	entry = kzalloc(sizeof(*entry), GFP_NOFS);
+	entry = kmem_cache_zalloc(btrfs_ordered_extent_cache, GFP_NOFS);
 	if (!entry)
 		return -ENOMEM;
 
 	entry->file_offset = file_offset;
 	entry->start = start;
 	entry->len = len;
+	if (!(BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM) &&
+	    !(type == BTRFS_ORDERED_NOCOW))
+		entry->csum_bytes_left = disk_len;
 	entry->disk_len = disk_len;
 	entry->bytes_left = len;
-	entry->inode = inode;
+	entry->inode = igrab(inode);
+	entry->compress_type = compress_type;
+	entry->truncated_len = (u64)-1;
 	if (type != BTRFS_ORDERED_IO_DONE && type != BTRFS_ORDERED_COMPLETE)
 		set_bit(type, &entry->flags);
 
+	if (dio)
+		set_bit(BTRFS_ORDERED_DIRECT, &entry->flags);
+
 	/* one ref for the tree */
 	atomic_set(&entry->refs, 1);
 	init_waitqueue_head(&entry->wait);
 	INIT_LIST_HEAD(&entry->list);
 	INIT_LIST_HEAD(&entry->root_extent_list);
+	INIT_LIST_HEAD(&entry->work_list);
+	init_completion(&entry->completion);
+	INIT_LIST_HEAD(&entry->log_list);
+
+	trace_btrfs_ordered_extent_add(inode, entry);
 
-	spin_lock(&tree->lock);
+	spin_lock_irq(&tree->lock);
 	node = tree_insert(&tree->tree, file_offset,
 			   &entry->rb_node);
-	BUG_ON(node);
-	spin_unlock(&tree->lock);
+	if (node)
+		ordered_data_tree_panic(inode, -EEXIST, file_offset);
+	spin_unlock_irq(&tree->lock);
 
-	spin_lock(&BTRFS_I(inode)->root->fs_info->ordered_extent_lock);
+	spin_lock(&root->ordered_extent_lock);
 	list_add_tail(&entry->root_extent_list,
-		      &BTRFS_I(inode)->root->fs_info->ordered_extents);
-	spin_unlock(&BTRFS_I(inode)->root->fs_info->ordered_extent_lock);
+		      &root->ordered_extents);
+	root->nr_ordered_extents++;
+	if (root->nr_ordered_extents == 1) {
+		spin_lock(&root->fs_info->ordered_root_lock);
+		BUG_ON(!list_empty(&root->ordered_root));
+		list_add_tail(&root->ordered_root,
+			      &root->fs_info->ordered_roots);
+		spin_unlock(&root->fs_info->ordered_root_lock);
+	}
+	spin_unlock(&root->ordered_extent_lock);
 
-	BUG_ON(node);
 	return 0;
 }
 
+int btrfs_add_ordered_extent(struct inode *inode, u64 file_offset,
+			     u64 start, u64 len, u64 disk_len, int type)
+{
+	return __btrfs_add_ordered_extent(inode, file_offset, start, len,
+					  disk_len, type, 0,
+					  BTRFS_COMPRESS_NONE);
+}
+
+int btrfs_add_ordered_extent_dio(struct inode *inode, u64 file_offset,
+				 u64 start, u64 len, u64 disk_len, int type)
+{
+	return __btrfs_add_ordered_extent(inode, file_offset, start, len,
+					  disk_len, type, 1,
+					  BTRFS_COMPRESS_NONE);
+}
+
+int btrfs_add_ordered_extent_compress(struct inode *inode, u64 file_offset,
+				      u64 start, u64 len, u64 disk_len,
+				      int type, int compress_type)
+{
+	return __btrfs_add_ordered_extent(inode, file_offset, start, len,
+					  disk_len, type, 0,
+					  compress_type);
+}
+
 /*
  * Add a struct btrfs_ordered_sum into the list of checksums to be inserted
  * when an ordered extent is finished.  If the list covers more than one
  * ordered extent, it is split across multiples.
  */
-int btrfs_add_ordered_sum(struct inode *inode,
-			  struct btrfs_ordered_extent *entry,
-			  struct btrfs_ordered_sum *sum)
+void btrfs_add_ordered_sum(struct inode *inode,
+			   struct btrfs_ordered_extent *entry,
+			   struct btrfs_ordered_sum *sum)
 {
 	struct btrfs_ordered_inode_tree *tree;
 
 	tree = &BTRFS_I(inode)->ordered_tree;
-	spin_lock(&tree->lock);
+	spin_lock_irq(&tree->lock);
 	list_add_tail(&sum->list, &entry->list);
-	spin_unlock(&tree->lock);
-	return 0;
+	WARN_ON(entry->csum_bytes_left < sum->len);
+	entry->csum_bytes_left -= sum->len;
+	if (entry->csum_bytes_left == 0)
+		wake_up(&entry->wait);
+	spin_unlock_irq(&tree->lock);
+}
+
+/*
+ * this is used to account for finished IO across a given range
+ * of the file.  The IO may span ordered extents.  If
+ * a given ordered_extent is completely done, 1 is returned, otherwise
+ * 0.
+ *
+ * test_and_set_bit on a flag in the struct btrfs_ordered_extent is used
+ * to make sure this function only returns 1 once for a given ordered extent.
+ *
+ * file_offset is updated to one byte past the range that is recorded as
+ * complete.  This allows you to walk forward in the file.
+ */
+int btrfs_dec_test_first_ordered_pending(struct inode *inode,
+				   struct btrfs_ordered_extent **cached,
+				   u64 *file_offset, u64 io_size, int uptodate)
+{
+	struct btrfs_ordered_inode_tree *tree;
+	struct rb_node *node;
+	struct btrfs_ordered_extent *entry = NULL;
+	int ret;
+	unsigned long flags;
+	u64 dec_end;
+	u64 dec_start;
+	u64 to_dec;
+
+	tree = &BTRFS_I(inode)->ordered_tree;
+	spin_lock_irqsave(&tree->lock, flags);
+	node = tree_search(tree, *file_offset);
+	if (!node) {
+		ret = 1;
+		goto out;
+	}
+
+	entry = rb_entry(node, struct btrfs_ordered_extent, rb_node);
+	if (!offset_in_entry(entry, *file_offset)) {
+		ret = 1;
+		goto out;
+	}
+
+	dec_start = max(*file_offset, entry->file_offset);
+	dec_end = min(*file_offset + io_size, entry->file_offset +
+		      entry->len);
+	*file_offset = dec_end;
+	if (dec_start > dec_end) {
+		btrfs_crit(BTRFS_I(inode)->root->fs_info,
+			"bad ordering dec_start %llu end %llu", dec_start, dec_end);
+	}
+	to_dec = dec_end - dec_start;
+	if (to_dec > entry->bytes_left) {
+		btrfs_crit(BTRFS_I(inode)->root->fs_info,
+			"bad ordered accounting left %llu size %llu",
+			entry->bytes_left, to_dec);
+	}
+	entry->bytes_left -= to_dec;
+	if (!uptodate)
+		set_bit(BTRFS_ORDERED_IOERR, &entry->flags);
+
+	if (entry->bytes_left == 0) {
+		ret = test_and_set_bit(BTRFS_ORDERED_IO_DONE, &entry->flags);
+		if (waitqueue_active(&entry->wait))
+			wake_up(&entry->wait);
+	} else {
+		ret = 1;
+	}
+out:
+	if (!ret && cached && entry) {
+		*cached = entry;
+		atomic_inc(&entry->refs);
+	}
+	spin_unlock_irqrestore(&tree->lock, flags);
+	return ret == 0;
 }
 
 /*
@@ -232,15 +376,21 @@ int btrfs_add_ordered_sum(struct inode *inode,
  */
 int btrfs_dec_test_ordered_pending(struct inode *inode,
 				   struct btrfs_ordered_extent **cached,
-				   u64 file_offset, u64 io_size)
+				   u64 file_offset, u64 io_size, int uptodate)
 {
 	struct btrfs_ordered_inode_tree *tree;
 	struct rb_node *node;
 	struct btrfs_ordered_extent *entry = NULL;
+	unsigned long flags;
 	int ret;
 
 	tree = &BTRFS_I(inode)->ordered_tree;
-	spin_lock(&tree->lock);
+	spin_lock_irqsave(&tree->lock, flags);
+	if (cached && *cached) {
+		entry = *cached;
+		goto have_entry;
+	}
+
 	node = tree_search(tree, file_offset);
 	if (!node) {
 		ret = 1;
@@ -248,78 +398,178 @@ int btrfs_dec_test_ordered_pending(struct inode *inode,
 	}
 
 	entry = rb_entry(node, struct btrfs_ordered_extent, rb_node);
+have_entry:
 	if (!offset_in_entry(entry, file_offset)) {
 		ret = 1;
 		goto out;
 	}
 
 	if (io_size > entry->bytes_left) {
-		printk(KERN_CRIT "bad ordered accounting left %llu size %llu\n",
-		       (unsigned long long)entry->bytes_left,
-		       (unsigned long long)io_size);
+		btrfs_crit(BTRFS_I(inode)->root->fs_info,
+			   "bad ordered accounting left %llu size %llu",
+		       entry->bytes_left, io_size);
 	}
 	entry->bytes_left -= io_size;
-	if (entry->bytes_left == 0)
+	if (!uptodate)
+		set_bit(BTRFS_ORDERED_IOERR, &entry->flags);
+
+	if (entry->bytes_left == 0) {
 		ret = test_and_set_bit(BTRFS_ORDERED_IO_DONE, &entry->flags);
-	else
+		if (waitqueue_active(&entry->wait))
+			wake_up(&entry->wait);
+	} else {
 		ret = 1;
+	}
 out:
 	if (!ret && cached && entry) {
 		*cached = entry;
 		atomic_inc(&entry->refs);
 	}
-	spin_unlock(&tree->lock);
+	spin_unlock_irqrestore(&tree->lock, flags);
 	return ret == 0;
 }
 
+/* Needs to either be called under a log transaction or the log_mutex */
+void btrfs_get_logged_extents(struct inode *inode,
+			      struct list_head *logged_list)
+{
+	struct btrfs_ordered_inode_tree *tree;
+	struct btrfs_ordered_extent *ordered;
+	struct rb_node *n;
+
+	tree = &BTRFS_I(inode)->ordered_tree;
+	spin_lock_irq(&tree->lock);
+	for (n = rb_first(&tree->tree); n; n = rb_next(n)) {
+		ordered = rb_entry(n, struct btrfs_ordered_extent, rb_node);
+		if (!list_empty(&ordered->log_list))
+			continue;
+		list_add_tail(&ordered->log_list, logged_list);
+		atomic_inc(&ordered->refs);
+	}
+	spin_unlock_irq(&tree->lock);
+}
+
+void btrfs_put_logged_extents(struct list_head *logged_list)
+{
+	struct btrfs_ordered_extent *ordered;
+
+	while (!list_empty(logged_list)) {
+		ordered = list_first_entry(logged_list,
+					   struct btrfs_ordered_extent,
+					   log_list);
+		list_del_init(&ordered->log_list);
+		btrfs_put_ordered_extent(ordered);
+	}
+}
+
+void btrfs_submit_logged_extents(struct list_head *logged_list,
+				 struct btrfs_root *log)
+{
+	int index = log->log_transid % 2;
+
+	spin_lock_irq(&log->log_extents_lock[index]);
+	list_splice_tail(logged_list, &log->logged_list[index]);
+	spin_unlock_irq(&log->log_extents_lock[index]);
+}
+
+void btrfs_wait_logged_extents(struct btrfs_root *log, u64 transid)
+{
+	struct btrfs_ordered_extent *ordered;
+	int index = transid % 2;
+
+	spin_lock_irq(&log->log_extents_lock[index]);
+	while (!list_empty(&log->logged_list[index])) {
+		ordered = list_first_entry(&log->logged_list[index],
+					   struct btrfs_ordered_extent,
+					   log_list);
+		list_del_init(&ordered->log_list);
+		spin_unlock_irq(&log->log_extents_lock[index]);
+
+		if (!test_bit(BTRFS_ORDERED_IO_DONE, &ordered->flags) &&
+		    !test_bit(BTRFS_ORDERED_DIRECT, &ordered->flags)) {
+			struct inode *inode = ordered->inode;
+			u64 start = ordered->file_offset;
+			u64 end = ordered->file_offset + ordered->len - 1;
+
+			WARN_ON(!inode);
+			filemap_fdatawrite_range(inode->i_mapping, start, end);
+		}
+		wait_event(ordered->wait, test_bit(BTRFS_ORDERED_IO_DONE,
+						   &ordered->flags));
+
+		btrfs_put_ordered_extent(ordered);
+		spin_lock_irq(&log->log_extents_lock[index]);
+	}
+	spin_unlock_irq(&log->log_extents_lock[index]);
+}
+
+void btrfs_free_logged_extents(struct btrfs_root *log, u64 transid)
+{
+	struct btrfs_ordered_extent *ordered;
+	int index = transid % 2;
+
+	spin_lock_irq(&log->log_extents_lock[index]);
+	while (!list_empty(&log->logged_list[index])) {
+		ordered = list_first_entry(&log->logged_list[index],
+					   struct btrfs_ordered_extent,
+					   log_list);
+		list_del_init(&ordered->log_list);
+		spin_unlock_irq(&log->log_extents_lock[index]);
+		btrfs_put_ordered_extent(ordered);
+		spin_lock_irq(&log->log_extents_lock[index]);
+	}
+	spin_unlock_irq(&log->log_extents_lock[index]);
+}
+
 /*
  * used to drop a reference on an ordered extent.  This will free
  * the extent if the last reference is dropped
  */
-int btrfs_put_ordered_extent(struct btrfs_ordered_extent *entry)
+void btrfs_put_ordered_extent(struct btrfs_ordered_extent *entry)
 {
 	struct list_head *cur;
 	struct btrfs_ordered_sum *sum;
 
+	trace_btrfs_ordered_extent_put(entry->inode, entry);
+
 	if (atomic_dec_and_test(&entry->refs)) {
+		if (entry->inode)
+			btrfs_add_delayed_iput(entry->inode);
 		while (!list_empty(&entry->list)) {
 			cur = entry->list.next;
 			sum = list_entry(cur, struct btrfs_ordered_sum, list);
 			list_del(&sum->list);
 			kfree(sum);
 		}
-		kfree(entry);
+		kmem_cache_free(btrfs_ordered_extent_cache, entry);
 	}
-	return 0;
 }
 
 /*
  * remove an ordered extent from the tree.  No references are dropped
- * and you must wake_up entry->wait.  You must hold the tree lock
- * while you call this function.
+ * and waiters are woken up.
  */
-static int __btrfs_remove_ordered_extent(struct inode *inode,
-				struct btrfs_ordered_extent *entry)
+void btrfs_remove_ordered_extent(struct inode *inode,
+				 struct btrfs_ordered_extent *entry)
 {
 	struct btrfs_ordered_inode_tree *tree;
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	struct rb_node *node;
 
 	tree = &BTRFS_I(inode)->ordered_tree;
+	spin_lock_irq(&tree->lock);
 	node = &entry->rb_node;
 	rb_erase(node, &tree->tree);
-	tree->last = NULL;
+	if (tree->last == node)
+		tree->last = NULL;
 	set_bit(BTRFS_ORDERED_COMPLETE, &entry->flags);
+	spin_unlock_irq(&tree->lock);
 
-	spin_lock(&BTRFS_I(inode)->accounting_lock);
-	WARN_ON(!BTRFS_I(inode)->outstanding_extents);
-	BTRFS_I(inode)->outstanding_extents--;
-	spin_unlock(&BTRFS_I(inode)->accounting_lock);
-	btrfs_unreserve_metadata_for_delalloc(BTRFS_I(inode)->root,
-					      inode, 1);
-
-	spin_lock(&root->fs_info->ordered_extent_lock);
+	spin_lock(&root->ordered_extent_lock);
 	list_del_init(&entry->root_extent_list);
+	root->nr_ordered_extents--;
+
+	trace_btrfs_ordered_extent_remove(inode, entry);
 
 	/*
 	 * we have no more ordered extents for this inode and
@@ -328,86 +578,112 @@ static int __btrfs_remove_ordered_extent(struct inode *inode,
 	 */
 	if (RB_EMPTY_ROOT(&tree->tree) &&
 	    !mapping_tagged(inode->i_mapping, PAGECACHE_TAG_DIRTY)) {
+		spin_lock(&root->fs_info->ordered_root_lock);
 		list_del_init(&BTRFS_I(inode)->ordered_operations);
+		spin_unlock(&root->fs_info->ordered_root_lock);
 	}
-	spin_unlock(&root->fs_info->ordered_extent_lock);
 
-	return 0;
+	if (!root->nr_ordered_extents) {
+		spin_lock(&root->fs_info->ordered_root_lock);
+		BUG_ON(list_empty(&root->ordered_root));
+		list_del_init(&root->ordered_root);
+		spin_unlock(&root->fs_info->ordered_root_lock);
+	}
+	spin_unlock(&root->ordered_extent_lock);
+	wake_up(&entry->wait);
 }
 
-/*
- * remove an ordered extent from the tree.  No references are dropped
- * but any waiters are woken.
- */
-int btrfs_remove_ordered_extent(struct inode *inode,
-				struct btrfs_ordered_extent *entry)
+static void btrfs_run_ordered_extent_work(struct btrfs_work *work)
 {
-	struct btrfs_ordered_inode_tree *tree;
-	int ret;
-
-	tree = &BTRFS_I(inode)->ordered_tree;
-	spin_lock(&tree->lock);
-	ret = __btrfs_remove_ordered_extent(inode, entry);
-	spin_unlock(&tree->lock);
-	wake_up(&entry->wait);
+	struct btrfs_ordered_extent *ordered;
 
-	return ret;
+	ordered = container_of(work, struct btrfs_ordered_extent, flush_work);
+	btrfs_start_ordered_extent(ordered->inode, ordered, 1);
+	complete(&ordered->completion);
 }
 
 /*
  * wait for all the ordered extents in a root.  This is done when balancing
  * space between drives.
  */
-int btrfs_wait_ordered_extents(struct btrfs_root *root,
-			       int nocow_only, int delay_iput)
+int btrfs_wait_ordered_extents(struct btrfs_root *root, int nr)
 {
-	struct list_head splice;
-	struct list_head *cur;
-	struct btrfs_ordered_extent *ordered;
-	struct inode *inode;
+	struct list_head splice, works;
+	struct btrfs_ordered_extent *ordered, *next;
+	int count = 0;
 
 	INIT_LIST_HEAD(&splice);
+	INIT_LIST_HEAD(&works);
+
+	mutex_lock(&root->ordered_extent_mutex);
+	spin_lock(&root->ordered_extent_lock);
+	list_splice_init(&root->ordered_extents, &splice);
+	while (!list_empty(&splice) && nr) {
+		ordered = list_first_entry(&splice, struct btrfs_ordered_extent,
+					   root_extent_list);
+		list_move_tail(&ordered->root_extent_list,
+			       &root->ordered_extents);
+		atomic_inc(&ordered->refs);
+		spin_unlock(&root->ordered_extent_lock);
 
-	spin_lock(&root->fs_info->ordered_extent_lock);
-	list_splice_init(&root->fs_info->ordered_extents, &splice);
-	while (!list_empty(&splice)) {
-		cur = splice.next;
-		ordered = list_entry(cur, struct btrfs_ordered_extent,
-				     root_extent_list);
-		if (nocow_only &&
-		    !test_bit(BTRFS_ORDERED_NOCOW, &ordered->flags) &&
-		    !test_bit(BTRFS_ORDERED_PREALLOC, &ordered->flags)) {
-			list_move(&ordered->root_extent_list,
-				  &root->fs_info->ordered_extents);
-			cond_resched_lock(&root->fs_info->ordered_extent_lock);
-			continue;
-		}
+		btrfs_init_work(&ordered->flush_work,
+				btrfs_run_ordered_extent_work, NULL, NULL);
+		list_add_tail(&ordered->work_list, &works);
+		btrfs_queue_work(root->fs_info->flush_workers,
+				 &ordered->flush_work);
 
-		list_del_init(&ordered->root_extent_list);
-		atomic_inc(&ordered->refs);
+		cond_resched();
+		spin_lock(&root->ordered_extent_lock);
+		if (nr != -1)
+			nr--;
+		count++;
+	}
+	list_splice_tail(&splice, &root->ordered_extents);
+	spin_unlock(&root->ordered_extent_lock);
 
-		/*
-		 * the inode may be getting freed (in sys_unlink path).
-		 */
-		inode = igrab(ordered->inode);
+	list_for_each_entry_safe(ordered, next, &works, work_list) {
+		list_del_init(&ordered->work_list);
+		wait_for_completion(&ordered->completion);
+		btrfs_put_ordered_extent(ordered);
+		cond_resched();
+	}
+	mutex_unlock(&root->ordered_extent_mutex);
 
-		spin_unlock(&root->fs_info->ordered_extent_lock);
+	return count;
+}
 
-		if (inode) {
-			btrfs_start_ordered_extent(inode, ordered, 1);
-			btrfs_put_ordered_extent(ordered);
-			if (delay_iput)
-				btrfs_add_delayed_iput(inode);
-			else
-				iput(inode);
-		} else {
-			btrfs_put_ordered_extent(ordered);
-		}
+void btrfs_wait_ordered_roots(struct btrfs_fs_info *fs_info, int nr)
+{
+	struct btrfs_root *root;
+	struct list_head splice;
+	int done;
+
+	INIT_LIST_HEAD(&splice);
 
-		spin_lock(&root->fs_info->ordered_extent_lock);
+	mutex_lock(&fs_info->ordered_operations_mutex);
+	spin_lock(&fs_info->ordered_root_lock);
+	list_splice_init(&fs_info->ordered_roots, &splice);
+	while (!list_empty(&splice) && nr) {
+		root = list_first_entry(&splice, struct btrfs_root,
+					ordered_root);
+		root = btrfs_grab_fs_root(root);
+		BUG_ON(!root);
+		list_move_tail(&root->ordered_root,
+			       &fs_info->ordered_roots);
+		spin_unlock(&fs_info->ordered_root_lock);
+
+		done = btrfs_wait_ordered_extents(root, nr);
+		btrfs_put_fs_root(root);
+
+		spin_lock(&fs_info->ordered_root_lock);
+		if (nr != -1) {
+			nr -= done;
+			WARN_ON(nr < 0);
+		}
 	}
-	spin_unlock(&root->fs_info->ordered_extent_lock);
-	return 0;
+	list_splice_tail(&splice, &fs_info->ordered_roots);
+	spin_unlock(&fs_info->ordered_root_lock);
+	mutex_unlock(&fs_info->ordered_operations_mutex);
 }
 
 /*
@@ -420,23 +696,26 @@ int btrfs_wait_ordered_extents(struct btrfs_root *root,
  * extra check to make sure the ordered operation list really is empty
  * before we return
  */
-int btrfs_run_ordered_operations(struct btrfs_root *root, int wait)
+int btrfs_run_ordered_operations(struct btrfs_trans_handle *trans,
+				 struct btrfs_root *root, int wait)
 {
 	struct btrfs_inode *btrfs_inode;
 	struct inode *inode;
+	struct btrfs_transaction *cur_trans = trans->transaction;
 	struct list_head splice;
+	struct list_head works;
+	struct btrfs_delalloc_work *work, *next;
+	int ret = 0;
 
 	INIT_LIST_HEAD(&splice);
+	INIT_LIST_HEAD(&works);
 
-	mutex_lock(&root->fs_info->ordered_operations_mutex);
-	spin_lock(&root->fs_info->ordered_extent_lock);
-again:
-	list_splice_init(&root->fs_info->ordered_operations, &splice);
-
+	mutex_lock(&root->fs_info->ordered_extent_flush_mutex);
+	spin_lock(&root->fs_info->ordered_root_lock);
+	list_splice_init(&cur_trans->ordered_operations, &splice);
 	while (!list_empty(&splice)) {
 		btrfs_inode = list_entry(splice.next, struct btrfs_inode,
 				   ordered_operations);
-
 		inode = &btrfs_inode->vfs_inode;
 
 		list_del_init(&btrfs_inode->ordered_operations);
@@ -445,31 +724,41 @@ again:
 		 * the inode may be getting freed (in sys_unlink path).
 		 */
 		inode = igrab(inode);
+		if (!inode)
+			continue;
 
-		if (!wait && inode) {
+		if (!wait)
 			list_add_tail(&BTRFS_I(inode)->ordered_operations,
-			      &root->fs_info->ordered_operations);
-		}
-		spin_unlock(&root->fs_info->ordered_extent_lock);
-
-		if (inode) {
-			if (wait)
-				btrfs_wait_ordered_range(inode, 0, (u64)-1);
-			else
-				filemap_flush(inode->i_mapping);
-			btrfs_add_delayed_iput(inode);
+				      &cur_trans->ordered_operations);
+		spin_unlock(&root->fs_info->ordered_root_lock);
+
+		work = btrfs_alloc_delalloc_work(inode, wait, 1);
+		if (!work) {
+			spin_lock(&root->fs_info->ordered_root_lock);
+			if (list_empty(&BTRFS_I(inode)->ordered_operations))
+				list_add_tail(&btrfs_inode->ordered_operations,
+					      &splice);
+			list_splice_tail(&splice,
+					 &cur_trans->ordered_operations);
+			spin_unlock(&root->fs_info->ordered_root_lock);
+			ret = -ENOMEM;
+			goto out;
 		}
+		list_add_tail(&work->list, &works);
+		btrfs_queue_work(root->fs_info->flush_workers,
+				 &work->work);
 
 		cond_resched();
-		spin_lock(&root->fs_info->ordered_extent_lock);
+		spin_lock(&root->fs_info->ordered_root_lock);
 	}
-	if (wait && !list_empty(&root->fs_info->ordered_operations))
-		goto again;
-
-	spin_unlock(&root->fs_info->ordered_extent_lock);
-	mutex_unlock(&root->fs_info->ordered_operations_mutex);
-
-	return 0;
+	spin_unlock(&root->fs_info->ordered_root_lock);
+out:
+	list_for_each_entry_safe(work, next, &works, list) {
+		list_del_init(&work->list);
+		btrfs_wait_and_free_delalloc_work(work);
+	}
+	mutex_unlock(&root->fs_info->ordered_extent_flush_mutex);
+	return ret;
 }
 
 /*
@@ -486,12 +775,15 @@ void btrfs_start_ordered_extent(struct inode *inode,
 	u64 start = entry->file_offset;
 	u64 end = start + entry->len - 1;
 
+	trace_btrfs_ordered_extent_start(inode, entry);
+
 	/*
 	 * pages in the range can be dirty, clean or writeback.  We
 	 * start IO on any dirty ones so the wait doesn't stall waiting
-	 * for pdflush to find them
+	 * for the flusher thread to find them
 	 */
-	filemap_fdatawrite_range(inode->i_mapping, start, end);
+	if (!test_bit(BTRFS_ORDERED_DIRECT, &entry->flags))
+		filemap_fdatawrite_range(inode->i_mapping, start, end);
 	if (wait) {
 		wait_event(entry->wait, test_bit(BTRFS_ORDERED_COMPLETE,
 						 &entry->flags));
@@ -503,11 +795,10 @@ void btrfs_start_ordered_extent(struct inode *inode,
  */
 int btrfs_wait_ordered_range(struct inode *inode, u64 start, u64 len)
 {
+	int ret = 0;
 	u64 end;
 	u64 orig_end;
-	u64 wait_end;
 	struct btrfs_ordered_extent *ordered;
-	int found;
 
 	if (start + len < start) {
 		orig_end = INT_LIMIT(loff_t);
@@ -516,23 +807,39 @@ int btrfs_wait_ordered_range(struct inode *inode, u64 start, u64 len)
 		if (orig_end > INT_LIMIT(loff_t))
 			orig_end = INT_LIMIT(loff_t);
 	}
-	wait_end = orig_end;
-again:
+
 	/* start IO across the range first to instantiate any delalloc
 	 * extents
 	 */
-	filemap_fdatawrite_range(inode->i_mapping, start, orig_end);
-
-	/* The compression code will leave pages locked but return from
-	 * writepage without setting the page writeback.  Starting again
-	 * with WB_SYNC_ALL will end up waiting for the IO to actually start.
+	ret = filemap_fdatawrite_range(inode->i_mapping, start, orig_end);
+	if (ret)
+		return ret;
+	/*
+	 * So with compression we will find and lock a dirty page and clear the
+	 * first one as dirty, setup an async extent, and immediately return
+	 * with the entire range locked but with nobody actually marked with
+	 * writeback.  So we can't just filemap_write_and_wait_range() and
+	 * expect it to work since it will just kick off a thread to do the
+	 * actual work.  So we need to call filemap_fdatawrite_range _again_
+	 * since it will wait on the page lock, which won't be unlocked until
+	 * after the pages have been marked as writeback and so we're good to go
+	 * from there.  We have to do this otherwise we'll miss the ordered
+	 * extents and that results in badness.  Please Josef, do not think you
+	 * know better and pull this out at some point in the future, it is
+	 * right and you are wrong.
 	 */
-	filemap_fdatawrite_range(inode->i_mapping, start, orig_end);
-
-	filemap_fdatawait_range(inode->i_mapping, start, orig_end);
+	if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
+		     &BTRFS_I(inode)->runtime_flags)) {
+		ret = filemap_fdatawrite_range(inode->i_mapping, start,
+					       orig_end);
+		if (ret)
+			return ret;
+	}
+	ret = filemap_fdatawait_range(inode->i_mapping, start, orig_end);
+	if (ret)
+		return ret;
 
 	end = orig_end;
-	found = 0;
 	while (1) {
 		ordered = btrfs_lookup_first_ordered_extent(inode, end);
 		if (!ordered)
@@ -541,24 +848,20 @@ again:
 			btrfs_put_ordered_extent(ordered);
 			break;
 		}
-		if (ordered->file_offset + ordered->len < start) {
+		if (ordered->file_offset + ordered->len <= start) {
 			btrfs_put_ordered_extent(ordered);
 			break;
 		}
-		found++;
 		btrfs_start_ordered_extent(inode, ordered, 1);
 		end = ordered->file_offset;
+		if (test_bit(BTRFS_ORDERED_IOERR, &ordered->flags))
+			ret = -EIO;
 		btrfs_put_ordered_extent(ordered);
-		if (end == 0 || end == start)
+		if (ret || end == 0 || end == start)
 			break;
 		end--;
 	}
-	if (found || test_range_bit(&BTRFS_I(inode)->io_tree, start, orig_end,
-			   EXTENT_DELALLOC, 0, NULL)) {
-		schedule_timeout(1);
-		goto again;
-	}
-	return 0;
+	return ret;
 }
 
 /*
@@ -573,7 +876,7 @@ struct btrfs_ordered_extent *btrfs_lookup_ordered_extent(struct inode *inode,
 	struct btrfs_ordered_extent *entry = NULL;
 
 	tree = &BTRFS_I(inode)->ordered_tree;
-	spin_lock(&tree->lock);
+	spin_lock_irq(&tree->lock);
 	node = tree_search(tree, file_offset);
 	if (!node)
 		goto out;
@@ -584,7 +887,48 @@ struct btrfs_ordered_extent *btrfs_lookup_ordered_extent(struct inode *inode,
 	if (entry)
 		atomic_inc(&entry->refs);
 out:
-	spin_unlock(&tree->lock);
+	spin_unlock_irq(&tree->lock);
+	return entry;
+}
+
+/* Since the DIO code tries to lock a wide area we need to look for any ordered
+ * extents that exist in the range, rather than just the start of the range.
+ */
+struct btrfs_ordered_extent *btrfs_lookup_ordered_range(struct inode *inode,
+							u64 file_offset,
+							u64 len)
+{
+	struct btrfs_ordered_inode_tree *tree;
+	struct rb_node *node;
+	struct btrfs_ordered_extent *entry = NULL;
+
+	tree = &BTRFS_I(inode)->ordered_tree;
+	spin_lock_irq(&tree->lock);
+	node = tree_search(tree, file_offset);
+	if (!node) {
+		node = tree_search(tree, file_offset + len);
+		if (!node)
+			goto out;
+	}
+
+	while (1) {
+		entry = rb_entry(node, struct btrfs_ordered_extent, rb_node);
+		if (range_overlaps(entry, file_offset, len))
+			break;
+
+		if (entry->file_offset >= file_offset + len) {
+			entry = NULL;
+			break;
+		}
+		entry = NULL;
+		node = rb_next(node);
+		if (!node)
+			break;
+	}
+out:
+	if (entry)
+		atomic_inc(&entry->refs);
+	spin_unlock_irq(&tree->lock);
 	return entry;
 }
 
@@ -600,7 +944,7 @@ btrfs_lookup_first_ordered_extent(struct inode *inode, u64 file_offset)
 	struct btrfs_ordered_extent *entry = NULL;
 
 	tree = &BTRFS_I(inode)->ordered_tree;
-	spin_lock(&tree->lock);
+	spin_lock_irq(&tree->lock);
 	node = tree_search(tree, file_offset);
 	if (!node)
 		goto out;
@@ -608,7 +952,7 @@ btrfs_lookup_first_ordered_extent(struct inode *inode, u64 file_offset)
 	entry = rb_entry(node, struct btrfs_ordered_extent, rb_node);
 	atomic_inc(&entry->refs);
 out:
-	spin_unlock(&tree->lock);
+	spin_unlock_irq(&tree->lock);
 	return entry;
 }
 
@@ -620,22 +964,24 @@ int btrfs_ordered_update_i_size(struct inode *inode, u64 offset,
 				struct btrfs_ordered_extent *ordered)
 {
 	struct btrfs_ordered_inode_tree *tree = &BTRFS_I(inode)->ordered_tree;
-	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
 	u64 disk_i_size;
 	u64 new_i_size;
-	u64 i_size_test;
 	u64 i_size = i_size_read(inode);
 	struct rb_node *node;
 	struct rb_node *prev = NULL;
 	struct btrfs_ordered_extent *test;
 	int ret = 1;
 
-	if (ordered)
+	spin_lock_irq(&tree->lock);
+	if (ordered) {
 		offset = entry_end(ordered);
-	else
+		if (test_bit(BTRFS_ORDERED_TRUNCATED, &ordered->flags))
+			offset = min(offset,
+				     ordered->file_offset +
+				     ordered->truncated_len);
+	} else {
 		offset = ALIGN(offset, BTRFS_I(inode)->root->sectorsize);
-
-	spin_lock(&tree->lock);
+	}
 	disk_i_size = BTRFS_I(inode)->disk_i_size;
 
 	/* truncate file */
@@ -649,18 +995,17 @@ int btrfs_ordered_update_i_size(struct inode *inode, u64 offset,
 	 * if the disk i_size is already at the inode->i_size, or
 	 * this ordered extent is inside the disk i_size, we're done
 	 */
-	if (disk_i_size == i_size || offset <= disk_i_size) {
+	if (disk_i_size == i_size)
 		goto out;
-	}
 
 	/*
-	 * we can't update the disk_isize if there are delalloc bytes
-	 * between disk_i_size and  this ordered extent
+	 * We still need to update disk_i_size if outstanding_isize is greater
+	 * than disk_i_size.
 	 */
-	if (test_range_bit(io_tree, disk_i_size, offset - 1,
-			   EXTENT_DELALLOC, 0, NULL)) {
+	if (offset <= disk_i_size &&
+	    (!ordered || ordered->outstanding_isize <= disk_i_size))
 		goto out;
-	}
+
 	/*
 	 * walk backward from this ordered extent to disk_i_size.
 	 * if we find an ordered extent then we can't update disk i_size
@@ -681,69 +1026,53 @@ int btrfs_ordered_update_i_size(struct inode *inode, u64 offset,
 		}
 		node = prev;
 	}
-	while (node) {
+	for (; node; node = rb_prev(node)) {
 		test = rb_entry(node, struct btrfs_ordered_extent, rb_node);
+
+		/* We treat this entry as if it doesnt exist */
+		if (test_bit(BTRFS_ORDERED_UPDATED_ISIZE, &test->flags))
+			continue;
 		if (test->file_offset + test->len <= disk_i_size)
 			break;
 		if (test->file_offset >= i_size)
 			break;
-		if (test->file_offset >= disk_i_size)
+		if (entry_end(test) > disk_i_size) {
+			/*
+			 * we don't update disk_i_size now, so record this
+			 * undealt i_size. Or we will not know the real
+			 * i_size.
+			 */
+			if (test->outstanding_isize < offset)
+				test->outstanding_isize = offset;
+			if (ordered &&
+			    ordered->outstanding_isize >
+			    test->outstanding_isize)
+				test->outstanding_isize =
+						ordered->outstanding_isize;
 			goto out;
-		node = rb_prev(node);
+		}
 	}
 	new_i_size = min_t(u64, offset, i_size);
 
 	/*
-	 * at this point, we know we can safely update i_size to at least
-	 * the offset from this ordered extent.  But, we need to
-	 * walk forward and see if ios from higher up in the file have
-	 * finished.
+	 * Some ordered extents may completed before the current one, and
+	 * we hold the real i_size in ->outstanding_isize.
 	 */
-	if (ordered) {
-		node = rb_next(&ordered->rb_node);
-	} else {
-		if (prev)
-			node = rb_next(prev);
-		else
-			node = rb_first(&tree->tree);
-	}
-	i_size_test = 0;
-	if (node) {
-		/*
-		 * do we have an area where IO might have finished
-		 * between our ordered extent and the next one.
-		 */
-		test = rb_entry(node, struct btrfs_ordered_extent, rb_node);
-		if (test->file_offset > offset)
-			i_size_test = test->file_offset;
-	} else {
-		i_size_test = i_size;
-	}
-
-	/*
-	 * i_size_test is the end of a region after this ordered
-	 * extent where there are no ordered extents.  As long as there
-	 * are no delalloc bytes in this area, it is safe to update
-	 * disk_i_size to the end of the region.
-	 */
-	if (i_size_test > offset &&
-	    !test_range_bit(io_tree, offset, i_size_test - 1,
-			    EXTENT_DELALLOC, 0, NULL)) {
-		new_i_size = min_t(u64, i_size_test, i_size);
-	}
+	if (ordered && ordered->outstanding_isize > new_i_size)
+		new_i_size = min_t(u64, ordered->outstanding_isize, i_size);
 	BTRFS_I(inode)->disk_i_size = new_i_size;
 	ret = 0;
 out:
 	/*
-	 * we need to remove the ordered extent with the tree lock held
-	 * so that other people calling this function don't find our fully
-	 * processed ordered entry and skip updating the i_size
+	 * We need to do this because we can't remove ordered extents until
+	 * after the i_disk_size has been updated and then the inode has been
+	 * updated to reflect the change, so we need to tell anybody who finds
+	 * this ordered extent that we've already done all the real work, we
+	 * just haven't completed all the other work.
 	 */
 	if (ordered)
-		__btrfs_remove_ordered_extent(inode, ordered);
-	spin_unlock(&tree->lock);
-	if (ordered)
-		wake_up(&ordered->wait);
+		set_bit(BTRFS_ORDERED_UPDATED_ISIZE, &ordered->flags);
+	spin_unlock_irq(&tree->lock);
 	return ret;
 }
 
@@ -753,39 +1082,42 @@ out:
  * be reclaimed before their checksum is actually put into the btree
  */
 int btrfs_find_ordered_sum(struct inode *inode, u64 offset, u64 disk_bytenr,
-			   u32 *sum)
+			   u32 *sum, int len)
 {
 	struct btrfs_ordered_sum *ordered_sum;
-	struct btrfs_sector_sum *sector_sums;
 	struct btrfs_ordered_extent *ordered;
 	struct btrfs_ordered_inode_tree *tree = &BTRFS_I(inode)->ordered_tree;
 	unsigned long num_sectors;
 	unsigned long i;
 	u32 sectorsize = BTRFS_I(inode)->root->sectorsize;
-	int ret = 1;
+	int index = 0;
 
 	ordered = btrfs_lookup_ordered_extent(inode, offset);
 	if (!ordered)
-		return 1;
+		return 0;
 
-	spin_lock(&tree->lock);
+	spin_lock_irq(&tree->lock);
 	list_for_each_entry_reverse(ordered_sum, &ordered->list, list) {
-		if (disk_bytenr >= ordered_sum->bytenr) {
-			num_sectors = ordered_sum->len / sectorsize;
-			sector_sums = ordered_sum->sums;
-			for (i = 0; i < num_sectors; i++) {
-				if (sector_sums[i].bytenr == disk_bytenr) {
-					*sum = sector_sums[i].sum;
-					ret = 0;
-					goto out;
-				}
-			}
+		if (disk_bytenr >= ordered_sum->bytenr &&
+		    disk_bytenr < ordered_sum->bytenr + ordered_sum->len) {
+			i = (disk_bytenr - ordered_sum->bytenr) >>
+			    inode->i_sb->s_blocksize_bits;
+			num_sectors = ordered_sum->len >>
+				      inode->i_sb->s_blocksize_bits;
+			num_sectors = min_t(int, len - index, num_sectors - i);
+			memcpy(sum + index, ordered_sum->sums + i,
+			       num_sectors);
+
+			index += (int)num_sectors;
+			if (index == len)
+				goto out;
+			disk_bytenr += num_sectors * sectorsize;
 		}
 	}
 out:
-	spin_unlock(&tree->lock);
+	spin_unlock_irq(&tree->lock);
 	btrfs_put_ordered_extent(ordered);
-	return ret;
+	return index;
 }
 
 
@@ -801,10 +1133,10 @@ out:
  * If trans is not null, we'll do a friendly check for a transaction that
  * is already flushing things and force the IO down ourselves.
  */
-int btrfs_add_ordered_operation(struct btrfs_trans_handle *trans,
-				struct btrfs_root *root,
-				struct inode *inode)
+void btrfs_add_ordered_operation(struct btrfs_trans_handle *trans,
+				 struct btrfs_root *root, struct inode *inode)
 {
+	struct btrfs_transaction *cur_trans = trans->transaction;
 	u64 last_mod;
 
 	last_mod = max(BTRFS_I(inode)->generation, BTRFS_I(inode)->last_trans);
@@ -813,24 +1145,31 @@ int btrfs_add_ordered_operation(struct btrfs_trans_handle *trans,
 	 * if this file hasn't been changed since the last transaction
 	 * commit, we can safely return without doing anything
 	 */
-	if (last_mod < root->fs_info->last_trans_committed)
-		return 0;
-
-	/*
-	 * the transaction is already committing.  Just start the IO and
-	 * don't bother with all of this list nonsense
-	 */
-	if (trans && root->fs_info->running_transaction->blocked) {
-		btrfs_wait_ordered_range(inode, 0, (u64)-1);
-		return 0;
-	}
+	if (last_mod <= root->fs_info->last_trans_committed)
+		return;
 
-	spin_lock(&root->fs_info->ordered_extent_lock);
+	spin_lock(&root->fs_info->ordered_root_lock);
 	if (list_empty(&BTRFS_I(inode)->ordered_operations)) {
 		list_add_tail(&BTRFS_I(inode)->ordered_operations,
-			      &root->fs_info->ordered_operations);
+			      &cur_trans->ordered_operations);
 	}
-	spin_unlock(&root->fs_info->ordered_extent_lock);
+	spin_unlock(&root->fs_info->ordered_root_lock);
+}
+
+int __init ordered_data_init(void)
+{
+	btrfs_ordered_extent_cache = kmem_cache_create("btrfs_ordered_extent",
+				     sizeof(struct btrfs_ordered_extent), 0,
+				     SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD,
+				     NULL);
+	if (!btrfs_ordered_extent_cache)
+		return -ENOMEM;
 
 	return 0;
 }
+
+void ordered_data_exit(void)
+{
+	if (btrfs_ordered_extent_cache)
+		kmem_cache_destroy(btrfs_ordered_extent_cache);
+}
diff --git a/fs/btrfs/ordered-data.h b/fs/btrfs/ordered-data.h
index c82f76a9f04..246897058ef 100644
--- a/fs/btrfs/ordered-data.h
+++ b/fs/btrfs/ordered-data.h
@@ -26,18 +26,6 @@ struct btrfs_ordered_inode_tree {
 	struct rb_node *last;
 };
 
-/*
- * these are used to collect checksums done just before bios submission.
- * They are attached via a list into the ordered extent, and
- * checksum items are inserted into the tree after all the blocks in
- * the ordered extent are on disk
- */
-struct btrfs_sector_sum {
-	/* bytenr on disk */
-	u64 bytenr;
-	u32 sum;
-};
-
 struct btrfs_ordered_sum {
 	/* bytenr is the start of this extent on disk */
 	u64 bytenr;
@@ -45,10 +33,10 @@ struct btrfs_ordered_sum {
 	/*
 	 * this is the length in bytes covered by the sums array below.
 	 */
-	unsigned long len;
+	int len;
 	struct list_head list;
-	/* last field is a variable length array of btrfs_sector_sums */
-	struct btrfs_sector_sum sums[];
+	/* last field is a variable length array of csums */
+	u32 sums[];
 };
 
 /*
@@ -68,10 +56,21 @@ struct btrfs_ordered_sum {
 
 #define BTRFS_ORDERED_NOCOW 2 /* set when we want to write in place */
 
-#define BTRFS_ORDERED_COMPRESSED 3 /* writing a compressed extent */
+#define BTRFS_ORDERED_COMPRESSED 3 /* writing a zlib compressed extent */
 
 #define BTRFS_ORDERED_PREALLOC 4 /* set when writing to prealloced extent */
 
+#define BTRFS_ORDERED_DIRECT 5 /* set when we're doing DIO with this extent */
+
+#define BTRFS_ORDERED_IOERR 6 /* We had an io error when writing this out */
+
+#define BTRFS_ORDERED_UPDATED_ISIZE 7 /* indicates whether this ordered extent
+				       * has done its due diligence in updating
+				       * the isize. */
+#define BTRFS_ORDERED_LOGGED_CSUM 8 /* We've logged the csums on this ordered
+				       ordered extent */
+#define BTRFS_ORDERED_TRUNCATED 9 /* Set when we have to truncate an extent */
+
 struct btrfs_ordered_extent {
 	/* logical offset in the file */
 	u64 file_offset;
@@ -88,9 +87,28 @@ struct btrfs_ordered_extent {
 	/* number of bytes that still need writing */
 	u64 bytes_left;
 
+	/* number of bytes that still need csumming */
+	u64 csum_bytes_left;
+
+	/*
+	 * the end of the ordered extent which is behind it but
+	 * didn't update disk_i_size. Please see the comment of
+	 * btrfs_ordered_update_i_size();
+	 */
+	u64 outstanding_isize;
+
+	/*
+	 * If we get truncated we need to adjust the file extent we enter for
+	 * this ordered extent so that we do not expose stale data.
+	 */
+	u64 truncated_len;
+
 	/* flags (described above) */
 	unsigned long flags;
 
+	/* compression algorithm */
+	int compress_type;
+
 	/* reference count */
 	atomic_t refs;
 
@@ -100,6 +118,9 @@ struct btrfs_ordered_extent {
 	/* list of checksums for insertion when the extent io is done */
 	struct list_head list;
 
+	/* If we need to wait on this to be done */
+	struct list_head log_list;
+
 	/* used to wait for the BTRFS_ORDERED_COMPLETE bit */
 	wait_queue_head_t wait;
 
@@ -108,8 +129,13 @@ struct btrfs_ordered_extent {
 
 	/* a per root list of all the pending ordered extents */
 	struct list_head root_extent_list;
-};
 
+	struct btrfs_work work;
+
+	struct completion completion;
+	struct btrfs_work flush_work;
+	struct list_head work_list;
+};
 
 /*
  * calculates the total size you need to allocate for an ordered sum
@@ -118,11 +144,8 @@ struct btrfs_ordered_extent {
 static inline int btrfs_ordered_sum_size(struct btrfs_root *root,
 					 unsigned long bytes)
 {
-	unsigned long num_sectors = (bytes + root->sectorsize - 1) /
-		root->sectorsize;
-	num_sectors++;
-	return sizeof(struct btrfs_ordered_sum) +
-		num_sectors * sizeof(struct btrfs_sector_sum);
+	int num_sectors = (int)DIV_ROUND_UP(bytes, root->sectorsize);
+	return sizeof(struct btrfs_ordered_sum) + num_sectors * sizeof(u32);
 }
 
 static inline void
@@ -133,17 +156,26 @@ btrfs_ordered_inode_tree_init(struct btrfs_ordered_inode_tree *t)
 	t->last = NULL;
 }
 
-int btrfs_put_ordered_extent(struct btrfs_ordered_extent *entry);
-int btrfs_remove_ordered_extent(struct inode *inode,
+void btrfs_put_ordered_extent(struct btrfs_ordered_extent *entry);
+void btrfs_remove_ordered_extent(struct inode *inode,
 				struct btrfs_ordered_extent *entry);
 int btrfs_dec_test_ordered_pending(struct inode *inode,
 				   struct btrfs_ordered_extent **cached,
-				   u64 file_offset, u64 io_size);
+				   u64 file_offset, u64 io_size, int uptodate);
+int btrfs_dec_test_first_ordered_pending(struct inode *inode,
+				   struct btrfs_ordered_extent **cached,
+				   u64 *file_offset, u64 io_size,
+				   int uptodate);
 int btrfs_add_ordered_extent(struct inode *inode, u64 file_offset,
-			     u64 start, u64 len, u64 disk_len, int tyep);
-int btrfs_add_ordered_sum(struct inode *inode,
-			  struct btrfs_ordered_extent *entry,
-			  struct btrfs_ordered_sum *sum);
+			     u64 start, u64 len, u64 disk_len, int type);
+int btrfs_add_ordered_extent_dio(struct inode *inode, u64 file_offset,
+				 u64 start, u64 len, u64 disk_len, int type);
+int btrfs_add_ordered_extent_compress(struct inode *inode, u64 file_offset,
+				      u64 start, u64 len, u64 disk_len,
+				      int type, int compress_type);
+void btrfs_add_ordered_sum(struct inode *inode,
+			   struct btrfs_ordered_extent *entry,
+			   struct btrfs_ordered_sum *sum);
 struct btrfs_ordered_extent *btrfs_lookup_ordered_extent(struct inode *inode,
 							 u64 file_offset);
 void btrfs_start_ordered_extent(struct inode *inode,
@@ -151,13 +183,27 @@ void btrfs_start_ordered_extent(struct inode *inode,
 int btrfs_wait_ordered_range(struct inode *inode, u64 start, u64 len);
 struct btrfs_ordered_extent *
 btrfs_lookup_first_ordered_extent(struct inode * inode, u64 file_offset);
+struct btrfs_ordered_extent *btrfs_lookup_ordered_range(struct inode *inode,
+							u64 file_offset,
+							u64 len);
 int btrfs_ordered_update_i_size(struct inode *inode, u64 offset,
 				struct btrfs_ordered_extent *ordered);
-int btrfs_find_ordered_sum(struct inode *inode, u64 offset, u64 disk_bytenr, u32 *sum);
-int btrfs_run_ordered_operations(struct btrfs_root *root, int wait);
-int btrfs_add_ordered_operation(struct btrfs_trans_handle *trans,
-				struct btrfs_root *root,
-				struct inode *inode);
-int btrfs_wait_ordered_extents(struct btrfs_root *root,
-			       int nocow_only, int delay_iput);
+int btrfs_find_ordered_sum(struct inode *inode, u64 offset, u64 disk_bytenr,
+			   u32 *sum, int len);
+int btrfs_run_ordered_operations(struct btrfs_trans_handle *trans,
+				 struct btrfs_root *root, int wait);
+void btrfs_add_ordered_operation(struct btrfs_trans_handle *trans,
+				 struct btrfs_root *root,
+				 struct inode *inode);
+int btrfs_wait_ordered_extents(struct btrfs_root *root, int nr);
+void btrfs_wait_ordered_roots(struct btrfs_fs_info *fs_info, int nr);
+void btrfs_get_logged_extents(struct inode *inode,
+			      struct list_head *logged_list);
+void btrfs_put_logged_extents(struct list_head *logged_list);
+void btrfs_submit_logged_extents(struct list_head *logged_list,
+				 struct btrfs_root *log);
+void btrfs_wait_logged_extents(struct btrfs_root *log, u64 transid);
+void btrfs_free_logged_extents(struct btrfs_root *log, u64 transid);
+int __init ordered_data_init(void);
+void ordered_data_exit(void);
 #endif
diff --git a/fs/btrfs/orphan.c b/fs/btrfs/orphan.c
index 79cba5fbc28..65793edb38c 100644
--- a/fs/btrfs/orphan.c
+++ b/fs/btrfs/orphan.c
@@ -56,8 +56,12 @@ int btrfs_del_orphan_item(struct btrfs_trans_handle *trans,
 		return -ENOMEM;
 
 	ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
-	if (ret)
+	if (ret < 0)
 		goto out;
+	if (ret) { /* JDM: Really? */
+		ret = -ENOENT;
+		goto out;
+	}
 
 	ret = btrfs_del_item(trans, root, path);
 
@@ -65,23 +69,3 @@ out:
 	btrfs_free_path(path);
 	return ret;
 }
-
-int btrfs_find_orphan_item(struct btrfs_root *root, u64 offset)
-{
-	struct btrfs_path *path;
-	struct btrfs_key key;
-	int ret;
-
-	key.objectid = BTRFS_ORPHAN_OBJECTID;
-	key.type = BTRFS_ORPHAN_ITEM_KEY;
-	key.offset = offset;
-
-	path = btrfs_alloc_path();
-	if (!path)
-		return -ENOMEM;
-
-	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
-
-	btrfs_free_path(path);
-	return ret;
-}
diff --git a/fs/btrfs/print-tree.c b/fs/btrfs/print-tree.c
index 0d126be22b6..9626b4ad3b9 100644
--- a/fs/btrfs/print-tree.c
+++ b/fs/btrfs/print-tree.c
@@ -26,14 +26,12 @@ static void print_chunk(struct extent_buffer *eb, struct btrfs_chunk *chunk)
 	int i;
 	printk(KERN_INFO "\t\tchunk length %llu owner %llu type %llu "
 	       "num_stripes %d\n",
-	       (unsigned long long)btrfs_chunk_length(eb, chunk),
-	       (unsigned long long)btrfs_chunk_owner(eb, chunk),
-	       (unsigned long long)btrfs_chunk_type(eb, chunk),
-	       num_stripes);
+	       btrfs_chunk_length(eb, chunk), btrfs_chunk_owner(eb, chunk),
+	       btrfs_chunk_type(eb, chunk), num_stripes);
 	for (i = 0 ; i < num_stripes ; i++) {
 		printk(KERN_INFO "\t\t\tstripe %d devid %llu offset %llu\n", i,
-		      (unsigned long long)btrfs_stripe_devid_nr(eb, chunk, i),
-		      (unsigned long long)btrfs_stripe_offset_nr(eb, chunk, i));
+		      btrfs_stripe_devid_nr(eb, chunk, i),
+		      btrfs_stripe_offset_nr(eb, chunk, i));
 	}
 }
 static void print_dev_item(struct extent_buffer *eb,
@@ -41,22 +39,22 @@ static void print_dev_item(struct extent_buffer *eb,
 {
 	printk(KERN_INFO "\t\tdev item devid %llu "
 	       "total_bytes %llu bytes used %llu\n",
-	       (unsigned long long)btrfs_device_id(eb, dev_item),
-	       (unsigned long long)btrfs_device_total_bytes(eb, dev_item),
-	       (unsigned long long)btrfs_device_bytes_used(eb, dev_item));
+	       btrfs_device_id(eb, dev_item),
+	       btrfs_device_total_bytes(eb, dev_item),
+	       btrfs_device_bytes_used(eb, dev_item));
 }
 static void print_extent_data_ref(struct extent_buffer *eb,
 				  struct btrfs_extent_data_ref *ref)
 {
 	printk(KERN_INFO "\t\textent data backref root %llu "
 	       "objectid %llu offset %llu count %u\n",
-	       (unsigned long long)btrfs_extent_data_ref_root(eb, ref),
-	       (unsigned long long)btrfs_extent_data_ref_objectid(eb, ref),
-	       (unsigned long long)btrfs_extent_data_ref_offset(eb, ref),
+	       btrfs_extent_data_ref_root(eb, ref),
+	       btrfs_extent_data_ref_objectid(eb, ref),
+	       btrfs_extent_data_ref_offset(eb, ref),
 	       btrfs_extent_data_ref_count(eb, ref));
 }
 
-static void print_extent_item(struct extent_buffer *eb, int slot)
+static void print_extent_item(struct extent_buffer *eb, int slot, int type)
 {
 	struct btrfs_extent_item *ei;
 	struct btrfs_extent_inline_ref *iref;
@@ -65,7 +63,6 @@ static void print_extent_item(struct extent_buffer *eb, int slot)
 	struct btrfs_disk_key key;
 	unsigned long end;
 	unsigned long ptr;
-	int type;
 	u32 item_size = btrfs_item_size_nr(eb, slot);
 	u64 flags;
 	u64 offset;
@@ -87,19 +84,18 @@ static void print_extent_item(struct extent_buffer *eb, int slot)
 	flags = btrfs_extent_flags(eb, ei);
 
 	printk(KERN_INFO "\t\textent refs %llu gen %llu flags %llu\n",
-	       (unsigned long long)btrfs_extent_refs(eb, ei),
-	       (unsigned long long)btrfs_extent_generation(eb, ei),
-	       (unsigned long long)flags);
+	       btrfs_extent_refs(eb, ei), btrfs_extent_generation(eb, ei),
+	       flags);
 
-	if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
+	if ((type == BTRFS_EXTENT_ITEM_KEY) &&
+	    flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
 		struct btrfs_tree_block_info *info;
 		info = (struct btrfs_tree_block_info *)(ei + 1);
 		btrfs_tree_block_key(eb, info, &key);
-		printk(KERN_INFO "\t\ttree block key (%llu %x %llu) "
+		printk(KERN_INFO "\t\ttree block key (%llu %u %llu) "
 		       "level %d\n",
-		       (unsigned long long)btrfs_disk_key_objectid(&key),
-		       key.type,
-		       (unsigned long long)btrfs_disk_key_offset(&key),
+		       btrfs_disk_key_objectid(&key), key.type,
+		       btrfs_disk_key_offset(&key),
 		       btrfs_tree_block_level(eb, info));
 		iref = (struct btrfs_extent_inline_ref *)(info + 1);
 	} else {
@@ -115,11 +111,11 @@ static void print_extent_item(struct extent_buffer *eb, int slot)
 		switch (type) {
 		case BTRFS_TREE_BLOCK_REF_KEY:
 			printk(KERN_INFO "\t\ttree block backref "
-				"root %llu\n", (unsigned long long)offset);
+				"root %llu\n", offset);
 			break;
 		case BTRFS_SHARED_BLOCK_REF_KEY:
 			printk(KERN_INFO "\t\tshared block backref "
-				"parent %llu\n", (unsigned long long)offset);
+				"parent %llu\n", offset);
 			break;
 		case BTRFS_EXTENT_DATA_REF_KEY:
 			dref = (struct btrfs_extent_data_ref *)(&iref->offset);
@@ -129,8 +125,7 @@ static void print_extent_item(struct extent_buffer *eb, int slot)
 			sref = (struct btrfs_shared_data_ref *)(iref + 1);
 			printk(KERN_INFO "\t\tshared data backref "
 			       "parent %llu count %u\n",
-			       (unsigned long long)offset,
-			       btrfs_shared_data_ref_count(eb, sref));
+			       offset, btrfs_shared_data_ref_count(eb, sref));
 			break;
 		default:
 			BUG();
@@ -148,18 +143,36 @@ static void print_extent_ref_v0(struct extent_buffer *eb, int slot)
 	ref0 = btrfs_item_ptr(eb, slot, struct btrfs_extent_ref_v0);
 	printk("\t\textent back ref root %llu gen %llu "
 		"owner %llu num_refs %lu\n",
-		(unsigned long long)btrfs_ref_root_v0(eb, ref0),
-		(unsigned long long)btrfs_ref_generation_v0(eb, ref0),
-		(unsigned long long)btrfs_ref_objectid_v0(eb, ref0),
+		btrfs_ref_root_v0(eb, ref0),
+		btrfs_ref_generation_v0(eb, ref0),
+		btrfs_ref_objectid_v0(eb, ref0),
 		(unsigned long)btrfs_ref_count_v0(eb, ref0));
 }
 #endif
 
+static void print_uuid_item(struct extent_buffer *l, unsigned long offset,
+			    u32 item_size)
+{
+	if (!IS_ALIGNED(item_size, sizeof(u64))) {
+		pr_warn("BTRFS: uuid item with illegal size %lu!\n",
+			(unsigned long)item_size);
+		return;
+	}
+	while (item_size) {
+		__le64 subvol_id;
+
+		read_extent_buffer(l, &subvol_id, offset, sizeof(subvol_id));
+		printk(KERN_INFO "\t\tsubvol_id %llu\n",
+		       (unsigned long long)le64_to_cpu(subvol_id));
+		item_size -= sizeof(u64);
+		offset += sizeof(u64);
+	}
+}
+
 void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l)
 {
 	int i;
-	u32 type;
-	u32 nr = btrfs_header_nritems(l);
+	u32 type, nr;
 	struct btrfs_item *item;
 	struct btrfs_root_item *ri;
 	struct btrfs_dir_item *di;
@@ -172,45 +185,46 @@ void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l)
 	struct btrfs_key key;
 	struct btrfs_key found_key;
 
-	printk(KERN_INFO "leaf %llu total ptrs %d free space %d\n",
-		(unsigned long long)btrfs_header_bytenr(l), nr,
-		btrfs_leaf_free_space(root, l));
+	if (!l)
+		return;
+
+	nr = btrfs_header_nritems(l);
+
+	btrfs_info(root->fs_info, "leaf %llu total ptrs %d free space %d",
+		   btrfs_header_bytenr(l), nr, btrfs_leaf_free_space(root, l));
 	for (i = 0 ; i < nr ; i++) {
-		item = btrfs_item_nr(l, i);
+		item = btrfs_item_nr(i);
 		btrfs_item_key_to_cpu(l, &key, i);
 		type = btrfs_key_type(&key);
-		printk(KERN_INFO "\titem %d key (%llu %x %llu) itemoff %d "
+		printk(KERN_INFO "\titem %d key (%llu %u %llu) itemoff %d "
 		       "itemsize %d\n",
-			i,
-			(unsigned long long)key.objectid, type,
-			(unsigned long long)key.offset,
+			i, key.objectid, type, key.offset,
 			btrfs_item_offset(l, item), btrfs_item_size(l, item));
 		switch (type) {
 		case BTRFS_INODE_ITEM_KEY:
 			ii = btrfs_item_ptr(l, i, struct btrfs_inode_item);
 			printk(KERN_INFO "\t\tinode generation %llu size %llu "
 			       "mode %o\n",
-			       (unsigned long long)
 			       btrfs_inode_generation(l, ii),
-			      (unsigned long long)btrfs_inode_size(l, ii),
+			       btrfs_inode_size(l, ii),
 			       btrfs_inode_mode(l, ii));
 			break;
 		case BTRFS_DIR_ITEM_KEY:
 			di = btrfs_item_ptr(l, i, struct btrfs_dir_item);
 			btrfs_dir_item_key_to_cpu(l, di, &found_key);
 			printk(KERN_INFO "\t\tdir oid %llu type %u\n",
-				(unsigned long long)found_key.objectid,
+				found_key.objectid,
 				btrfs_dir_type(l, di));
 			break;
 		case BTRFS_ROOT_ITEM_KEY:
 			ri = btrfs_item_ptr(l, i, struct btrfs_root_item);
 			printk(KERN_INFO "\t\troot data bytenr %llu refs %u\n",
-				(unsigned long long)
 				btrfs_disk_root_bytenr(l, ri),
 				btrfs_disk_root_refs(l, ri));
 			break;
 		case BTRFS_EXTENT_ITEM_KEY:
-			print_extent_item(l, i);
+		case BTRFS_METADATA_ITEM_KEY:
+			print_extent_item(l, i, type);
 			break;
 		case BTRFS_TREE_BLOCK_REF_KEY:
 			printk(KERN_INFO "\t\ttree block backref\n");
@@ -236,22 +250,17 @@ void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l)
 			    BTRFS_FILE_EXTENT_INLINE) {
 				printk(KERN_INFO "\t\tinline extent data "
 				       "size %u\n",
-				       btrfs_file_extent_inline_len(l, fi));
+				       btrfs_file_extent_inline_len(l, i, fi));
 				break;
 			}
 			printk(KERN_INFO "\t\textent data disk bytenr %llu "
 			       "nr %llu\n",
-			       (unsigned long long)
 			       btrfs_file_extent_disk_bytenr(l, fi),
-			       (unsigned long long)
 			       btrfs_file_extent_disk_num_bytes(l, fi));
 			printk(KERN_INFO "\t\textent data offset %llu "
 			       "nr %llu ram %llu\n",
-			       (unsigned long long)
 			       btrfs_file_extent_offset(l, fi),
-			       (unsigned long long)
 			       btrfs_file_extent_num_bytes(l, fi),
-			       (unsigned long long)
 			       btrfs_file_extent_ram_bytes(l, fi));
 			break;
 		case BTRFS_EXTENT_REF_V0_KEY:
@@ -260,11 +269,11 @@ void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l)
 #else
 			BUG();
 #endif
+			break;
 		case BTRFS_BLOCK_GROUP_ITEM_KEY:
 			bi = btrfs_item_ptr(l, i,
 					    struct btrfs_block_group_item);
 			printk(KERN_INFO "\t\tblock group used %llu\n",
-			       (unsigned long long)
 			       btrfs_disk_block_group_used(l, bi));
 			break;
 		case BTRFS_CHUNK_ITEM_KEY:
@@ -281,14 +290,22 @@ void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l)
 			printk(KERN_INFO "\t\tdev extent chunk_tree %llu\n"
 			       "\t\tchunk objectid %llu chunk offset %llu "
 			       "length %llu\n",
-			       (unsigned long long)
 			       btrfs_dev_extent_chunk_tree(l, dev_extent),
-			       (unsigned long long)
 			       btrfs_dev_extent_chunk_objectid(l, dev_extent),
-			       (unsigned long long)
 			       btrfs_dev_extent_chunk_offset(l, dev_extent),
-			       (unsigned long long)
 			       btrfs_dev_extent_length(l, dev_extent));
+			break;
+		case BTRFS_DEV_STATS_KEY:
+			printk(KERN_INFO "\t\tdevice stats\n");
+			break;
+		case BTRFS_DEV_REPLACE_KEY:
+			printk(KERN_INFO "\t\tdev replace\n");
+			break;
+		case BTRFS_UUID_KEY_SUBVOL:
+		case BTRFS_UUID_KEY_RECEIVED_SUBVOL:
+			print_uuid_item(l, btrfs_item_ptr_offset(l, i),
+					btrfs_item_size_nr(l, i));
+			break;
 		};
 	}
 }
@@ -307,18 +324,14 @@ void btrfs_print_tree(struct btrfs_root *root, struct extent_buffer *c)
 		btrfs_print_leaf(root, c);
 		return;
 	}
-	printk(KERN_INFO "node %llu level %d total ptrs %d free spc %u\n",
-	       (unsigned long long)btrfs_header_bytenr(c),
-	      level, nr,
-	       (u32)BTRFS_NODEPTRS_PER_BLOCK(root) - nr);
+	btrfs_info(root->fs_info, "node %llu level %d total ptrs %d free spc %u",
+		btrfs_header_bytenr(c), level, nr,
+		(u32)BTRFS_NODEPTRS_PER_BLOCK(root) - nr);
 	for (i = 0; i < nr; i++) {
 		btrfs_node_key_to_cpu(c, &key, i);
 		printk(KERN_INFO "\tkey %d (%llu %u %llu) block %llu\n",
-		       i,
-		       (unsigned long long)key.objectid,
-		       key.type,
-		       (unsigned long long)key.offset,
-		       (unsigned long long)btrfs_node_blockptr(c, i));
+		       i, key.objectid, key.type, key.offset,
+		       btrfs_node_blockptr(c, i));
 	}
 	for (i = 0; i < nr; i++) {
 		struct extent_buffer *next = read_tree_block(root,
diff --git a/fs/btrfs/print-tree.h b/fs/btrfs/print-tree.h
index da75efe534d..7faddfacc5b 100644
--- a/fs/btrfs/print-tree.h
+++ b/fs/btrfs/print-tree.h
@@ -19,5 +19,5 @@
 #ifndef __PRINT_TREE_
 #define __PRINT_TREE_
 void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l);
-void btrfs_print_tree(struct btrfs_root *root, struct extent_buffer *t);
+void btrfs_print_tree(struct btrfs_root *root, struct extent_buffer *c);
 #endif
diff --git a/fs/btrfs/props.c b/fs/btrfs/props.c
new file mode 100644
index 00000000000..129b1dd2852
--- /dev/null
+++ b/fs/btrfs/props.c
@@ -0,0 +1,427 @@
+/*
+ * Copyright (C) 2014 Filipe David Borba Manana <fdmanana@gmail.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/hashtable.h>
+#include "props.h"
+#include "btrfs_inode.h"
+#include "hash.h"
+#include "transaction.h"
+#include "xattr.h"
+
+#define BTRFS_PROP_HANDLERS_HT_BITS 8
+static DEFINE_HASHTABLE(prop_handlers_ht, BTRFS_PROP_HANDLERS_HT_BITS);
+
+struct prop_handler {
+	struct hlist_node node;
+	const char *xattr_name;
+	int (*validate)(const char *value, size_t len);
+	int (*apply)(struct inode *inode, const char *value, size_t len);
+	const char *(*extract)(struct inode *inode);
+	int inheritable;
+};
+
+static int prop_compression_validate(const char *value, size_t len);
+static int prop_compression_apply(struct inode *inode,
+				  const char *value,
+				  size_t len);
+static const char *prop_compression_extract(struct inode *inode);
+
+static struct prop_handler prop_handlers[] = {
+	{
+		.xattr_name = XATTR_BTRFS_PREFIX "compression",
+		.validate = prop_compression_validate,
+		.apply = prop_compression_apply,
+		.extract = prop_compression_extract,
+		.inheritable = 1
+	},
+	{
+		.xattr_name = NULL
+	}
+};
+
+void __init btrfs_props_init(void)
+{
+	struct prop_handler *p;
+
+	hash_init(prop_handlers_ht);
+
+	for (p = &prop_handlers[0]; p->xattr_name; p++) {
+		u64 h = btrfs_name_hash(p->xattr_name, strlen(p->xattr_name));
+
+		hash_add(prop_handlers_ht, &p->node, h);
+	}
+}
+
+static const struct hlist_head *find_prop_handlers_by_hash(const u64 hash)
+{
+	struct hlist_head *h;
+
+	h = &prop_handlers_ht[hash_min(hash, BTRFS_PROP_HANDLERS_HT_BITS)];
+	if (hlist_empty(h))
+		return NULL;
+
+	return h;
+}
+
+static const struct prop_handler *
+find_prop_handler(const char *name,
+		  const struct hlist_head *handlers)
+{
+	struct prop_handler *h;
+
+	if (!handlers) {
+		u64 hash = btrfs_name_hash(name, strlen(name));
+
+		handlers = find_prop_handlers_by_hash(hash);
+		if (!handlers)
+			return NULL;
+	}
+
+	hlist_for_each_entry(h, handlers, node)
+		if (!strcmp(h->xattr_name, name))
+			return h;
+
+	return NULL;
+}
+
+static int __btrfs_set_prop(struct btrfs_trans_handle *trans,
+			    struct inode *inode,
+			    const char *name,
+			    const char *value,
+			    size_t value_len,
+			    int flags)
+{
+	const struct prop_handler *handler;
+	int ret;
+
+	if (strlen(name) <= XATTR_BTRFS_PREFIX_LEN)
+		return -EINVAL;
+
+	handler = find_prop_handler(name, NULL);
+	if (!handler)
+		return -EINVAL;
+
+	if (value_len == 0) {
+		ret = __btrfs_setxattr(trans, inode, handler->xattr_name,
+				       NULL, 0, flags);
+		if (ret)
+			return ret;
+
+		ret = handler->apply(inode, NULL, 0);
+		ASSERT(ret == 0);
+
+		return ret;
+	}
+
+	ret = handler->validate(value, value_len);
+	if (ret)
+		return ret;
+	ret = __btrfs_setxattr(trans, inode, handler->xattr_name,
+			       value, value_len, flags);
+	if (ret)
+		return ret;
+	ret = handler->apply(inode, value, value_len);
+	if (ret) {
+		__btrfs_setxattr(trans, inode, handler->xattr_name,
+				 NULL, 0, flags);
+		return ret;
+	}
+
+	set_bit(BTRFS_INODE_HAS_PROPS, &BTRFS_I(inode)->runtime_flags);
+
+	return 0;
+}
+
+int btrfs_set_prop(struct inode *inode,
+		   const char *name,
+		   const char *value,
+		   size_t value_len,
+		   int flags)
+{
+	return __btrfs_set_prop(NULL, inode, name, value, value_len, flags);
+}
+
+static int iterate_object_props(struct btrfs_root *root,
+				struct btrfs_path *path,
+				u64 objectid,
+				void (*iterator)(void *,
+						 const struct prop_handler *,
+						 const char *,
+						 size_t),
+				void *ctx)
+{
+	int ret;
+	char *name_buf = NULL;
+	char *value_buf = NULL;
+	int name_buf_len = 0;
+	int value_buf_len = 0;
+
+	while (1) {
+		struct btrfs_key key;
+		struct btrfs_dir_item *di;
+		struct extent_buffer *leaf;
+		u32 total_len, cur, this_len;
+		int slot;
+		const struct hlist_head *handlers;
+
+		slot = path->slots[0];
+		leaf = path->nodes[0];
+
+		if (slot >= btrfs_header_nritems(leaf)) {
+			ret = btrfs_next_leaf(root, path);
+			if (ret < 0)
+				goto out;
+			else if (ret > 0)
+				break;
+			continue;
+		}
+
+		btrfs_item_key_to_cpu(leaf, &key, slot);
+		if (key.objectid != objectid)
+			break;
+		if (key.type != BTRFS_XATTR_ITEM_KEY)
+			break;
+
+		handlers = find_prop_handlers_by_hash(key.offset);
+		if (!handlers)
+			goto next_slot;
+
+		di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
+		cur = 0;
+		total_len = btrfs_item_size_nr(leaf, slot);
+
+		while (cur < total_len) {
+			u32 name_len = btrfs_dir_name_len(leaf, di);
+			u32 data_len = btrfs_dir_data_len(leaf, di);
+			unsigned long name_ptr, data_ptr;
+			const struct prop_handler *handler;
+
+			this_len = sizeof(*di) + name_len + data_len;
+			name_ptr = (unsigned long)(di + 1);
+			data_ptr = name_ptr + name_len;
+
+			if (name_len <= XATTR_BTRFS_PREFIX_LEN ||
+			    memcmp_extent_buffer(leaf, XATTR_BTRFS_PREFIX,
+						 name_ptr,
+						 XATTR_BTRFS_PREFIX_LEN))
+				goto next_dir_item;
+
+			if (name_len >= name_buf_len) {
+				kfree(name_buf);
+				name_buf_len = name_len + 1;
+				name_buf = kmalloc(name_buf_len, GFP_NOFS);
+				if (!name_buf) {
+					ret = -ENOMEM;
+					goto out;
+				}
+			}
+			read_extent_buffer(leaf, name_buf, name_ptr, name_len);
+			name_buf[name_len] = '\0';
+
+			handler = find_prop_handler(name_buf, handlers);
+			if (!handler)
+				goto next_dir_item;
+
+			if (data_len > value_buf_len) {
+				kfree(value_buf);
+				value_buf_len = data_len;
+				value_buf = kmalloc(data_len, GFP_NOFS);
+				if (!value_buf) {
+					ret = -ENOMEM;
+					goto out;
+				}
+			}
+			read_extent_buffer(leaf, value_buf, data_ptr, data_len);
+
+			iterator(ctx, handler, value_buf, data_len);
+next_dir_item:
+			cur += this_len;
+			di = (struct btrfs_dir_item *)((char *) di + this_len);
+		}
+
+next_slot:
+		path->slots[0]++;
+	}
+
+	ret = 0;
+out:
+	btrfs_release_path(path);
+	kfree(name_buf);
+	kfree(value_buf);
+
+	return ret;
+}
+
+static void inode_prop_iterator(void *ctx,
+				const struct prop_handler *handler,
+				const char *value,
+				size_t len)
+{
+	struct inode *inode = ctx;
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	int ret;
+
+	ret = handler->apply(inode, value, len);
+	if (unlikely(ret))
+		btrfs_warn(root->fs_info,
+			   "error applying prop %s to ino %llu (root %llu): %d",
+			   handler->xattr_name, btrfs_ino(inode),
+			   root->root_key.objectid, ret);
+	else
+		set_bit(BTRFS_INODE_HAS_PROPS, &BTRFS_I(inode)->runtime_flags);
+}
+
+int btrfs_load_inode_props(struct inode *inode, struct btrfs_path *path)
+{
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	u64 ino = btrfs_ino(inode);
+	int ret;
+
+	ret = iterate_object_props(root, path, ino, inode_prop_iterator, inode);
+
+	return ret;
+}
+
+static int inherit_props(struct btrfs_trans_handle *trans,
+			 struct inode *inode,
+			 struct inode *parent)
+{
+	const struct prop_handler *h;
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	int ret;
+
+	if (!test_bit(BTRFS_INODE_HAS_PROPS,
+		      &BTRFS_I(parent)->runtime_flags))
+		return 0;
+
+	for (h = &prop_handlers[0]; h->xattr_name; h++) {
+		const char *value;
+		u64 num_bytes;
+
+		if (!h->inheritable)
+			continue;
+
+		value = h->extract(parent);
+		if (!value)
+			continue;
+
+		num_bytes = btrfs_calc_trans_metadata_size(root, 1);
+		ret = btrfs_block_rsv_add(root, trans->block_rsv,
+					  num_bytes, BTRFS_RESERVE_NO_FLUSH);
+		if (ret)
+			goto out;
+		ret = __btrfs_set_prop(trans, inode, h->xattr_name,
+				       value, strlen(value), 0);
+		btrfs_block_rsv_release(root, trans->block_rsv, num_bytes);
+		if (ret)
+			goto out;
+	}
+	ret = 0;
+out:
+	return ret;
+}
+
+int btrfs_inode_inherit_props(struct btrfs_trans_handle *trans,
+			      struct inode *inode,
+			      struct inode *dir)
+{
+	if (!dir)
+		return 0;
+
+	return inherit_props(trans, inode, dir);
+}
+
+int btrfs_subvol_inherit_props(struct btrfs_trans_handle *trans,
+			       struct btrfs_root *root,
+			       struct btrfs_root *parent_root)
+{
+	struct btrfs_key key;
+	struct inode *parent_inode, *child_inode;
+	int ret;
+
+	key.objectid = BTRFS_FIRST_FREE_OBJECTID;
+	key.type = BTRFS_INODE_ITEM_KEY;
+	key.offset = 0;
+
+	parent_inode = btrfs_iget(parent_root->fs_info->sb, &key,
+				  parent_root, NULL);
+	if (IS_ERR(parent_inode))
+		return PTR_ERR(parent_inode);
+
+	child_inode = btrfs_iget(root->fs_info->sb, &key, root, NULL);
+	if (IS_ERR(child_inode)) {
+		iput(parent_inode);
+		return PTR_ERR(child_inode);
+	}
+
+	ret = inherit_props(trans, child_inode, parent_inode);
+	iput(child_inode);
+	iput(parent_inode);
+
+	return ret;
+}
+
+static int prop_compression_validate(const char *value, size_t len)
+{
+	if (!strncmp("lzo", value, len))
+		return 0;
+	else if (!strncmp("zlib", value, len))
+		return 0;
+
+	return -EINVAL;
+}
+
+static int prop_compression_apply(struct inode *inode,
+				  const char *value,
+				  size_t len)
+{
+	int type;
+
+	if (len == 0) {
+		BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
+		BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
+		BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
+
+		return 0;
+	}
+
+	if (!strncmp("lzo", value, len))
+		type = BTRFS_COMPRESS_LZO;
+	else if (!strncmp("zlib", value, len))
+		type = BTRFS_COMPRESS_ZLIB;
+	else
+		return -EINVAL;
+
+	BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
+	BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
+	BTRFS_I(inode)->force_compress = type;
+
+	return 0;
+}
+
+static const char *prop_compression_extract(struct inode *inode)
+{
+	switch (BTRFS_I(inode)->force_compress) {
+	case BTRFS_COMPRESS_ZLIB:
+		return "zlib";
+	case BTRFS_COMPRESS_LZO:
+		return "lzo";
+	}
+
+	return NULL;
+}
diff --git a/fs/btrfs/props.h b/fs/btrfs/props.h
new file mode 100644
index 00000000000..100f18829d5
--- /dev/null
+++ b/fs/btrfs/props.h
@@ -0,0 +1,42 @@
+/*
+ * Copyright (C) 2014 Filipe David Borba Manana <fdmanana@gmail.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef __BTRFS_PROPS_H
+#define __BTRFS_PROPS_H
+
+#include "ctree.h"
+
+void __init btrfs_props_init(void);
+
+int btrfs_set_prop(struct inode *inode,
+		   const char *name,
+		   const char *value,
+		   size_t value_len,
+		   int flags);
+
+int btrfs_load_inode_props(struct inode *inode, struct btrfs_path *path);
+
+int btrfs_inode_inherit_props(struct btrfs_trans_handle *trans,
+			      struct inode *inode,
+			      struct inode *dir);
+
+int btrfs_subvol_inherit_props(struct btrfs_trans_handle *trans,
+			       struct btrfs_root *root,
+			       struct btrfs_root *parent_root);
+
+#endif
diff --git a/fs/btrfs/qgroup.c b/fs/btrfs/qgroup.c
new file mode 100644
index 00000000000..98cb6b2630f
--- /dev/null
+++ b/fs/btrfs/qgroup.c
@@ -0,0 +1,2651 @@
+/*
+ * Copyright (C) 2011 STRATO.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/sched.h>
+#include <linux/pagemap.h>
+#include <linux/writeback.h>
+#include <linux/blkdev.h>
+#include <linux/rbtree.h>
+#include <linux/slab.h>
+#include <linux/workqueue.h>
+#include <linux/btrfs.h>
+
+#include "ctree.h"
+#include "transaction.h"
+#include "disk-io.h"
+#include "locking.h"
+#include "ulist.h"
+#include "backref.h"
+#include "extent_io.h"
+#include "qgroup.h"
+
+/* TODO XXX FIXME
+ *  - subvol delete -> delete when ref goes to 0? delete limits also?
+ *  - reorganize keys
+ *  - compressed
+ *  - sync
+ *  - copy also limits on subvol creation
+ *  - limit
+ *  - caches fuer ulists
+ *  - performance benchmarks
+ *  - check all ioctl parameters
+ */
+
+/*
+ * one struct for each qgroup, organized in fs_info->qgroup_tree.
+ */
+struct btrfs_qgroup {
+	u64 qgroupid;
+
+	/*
+	 * state
+	 */
+	u64 rfer;	/* referenced */
+	u64 rfer_cmpr;	/* referenced compressed */
+	u64 excl;	/* exclusive */
+	u64 excl_cmpr;	/* exclusive compressed */
+
+	/*
+	 * limits
+	 */
+	u64 lim_flags;	/* which limits are set */
+	u64 max_rfer;
+	u64 max_excl;
+	u64 rsv_rfer;
+	u64 rsv_excl;
+
+	/*
+	 * reservation tracking
+	 */
+	u64 reserved;
+
+	/*
+	 * lists
+	 */
+	struct list_head groups;  /* groups this group is member of */
+	struct list_head members; /* groups that are members of this group */
+	struct list_head dirty;   /* dirty groups */
+	struct rb_node node;	  /* tree of qgroups */
+
+	/*
+	 * temp variables for accounting operations
+	 */
+	u64 old_refcnt;
+	u64 new_refcnt;
+};
+
+/*
+ * glue structure to represent the relations between qgroups.
+ */
+struct btrfs_qgroup_list {
+	struct list_head next_group;
+	struct list_head next_member;
+	struct btrfs_qgroup *group;
+	struct btrfs_qgroup *member;
+};
+
+#define ptr_to_u64(x) ((u64)(uintptr_t)x)
+#define u64_to_ptr(x) ((struct btrfs_qgroup *)(uintptr_t)x)
+
+static int
+qgroup_rescan_init(struct btrfs_fs_info *fs_info, u64 progress_objectid,
+		   int init_flags);
+static void qgroup_rescan_zero_tracking(struct btrfs_fs_info *fs_info);
+
+/* must be called with qgroup_ioctl_lock held */
+static struct btrfs_qgroup *find_qgroup_rb(struct btrfs_fs_info *fs_info,
+					   u64 qgroupid)
+{
+	struct rb_node *n = fs_info->qgroup_tree.rb_node;
+	struct btrfs_qgroup *qgroup;
+
+	while (n) {
+		qgroup = rb_entry(n, struct btrfs_qgroup, node);
+		if (qgroup->qgroupid < qgroupid)
+			n = n->rb_left;
+		else if (qgroup->qgroupid > qgroupid)
+			n = n->rb_right;
+		else
+			return qgroup;
+	}
+	return NULL;
+}
+
+/* must be called with qgroup_lock held */
+static struct btrfs_qgroup *add_qgroup_rb(struct btrfs_fs_info *fs_info,
+					  u64 qgroupid)
+{
+	struct rb_node **p = &fs_info->qgroup_tree.rb_node;
+	struct rb_node *parent = NULL;
+	struct btrfs_qgroup *qgroup;
+
+	while (*p) {
+		parent = *p;
+		qgroup = rb_entry(parent, struct btrfs_qgroup, node);
+
+		if (qgroup->qgroupid < qgroupid)
+			p = &(*p)->rb_left;
+		else if (qgroup->qgroupid > qgroupid)
+			p = &(*p)->rb_right;
+		else
+			return qgroup;
+	}
+
+	qgroup = kzalloc(sizeof(*qgroup), GFP_ATOMIC);
+	if (!qgroup)
+		return ERR_PTR(-ENOMEM);
+
+	qgroup->qgroupid = qgroupid;
+	INIT_LIST_HEAD(&qgroup->groups);
+	INIT_LIST_HEAD(&qgroup->members);
+	INIT_LIST_HEAD(&qgroup->dirty);
+
+	rb_link_node(&qgroup->node, parent, p);
+	rb_insert_color(&qgroup->node, &fs_info->qgroup_tree);
+
+	return qgroup;
+}
+
+static void __del_qgroup_rb(struct btrfs_qgroup *qgroup)
+{
+	struct btrfs_qgroup_list *list;
+
+	list_del(&qgroup->dirty);
+	while (!list_empty(&qgroup->groups)) {
+		list = list_first_entry(&qgroup->groups,
+					struct btrfs_qgroup_list, next_group);
+		list_del(&list->next_group);
+		list_del(&list->next_member);
+		kfree(list);
+	}
+
+	while (!list_empty(&qgroup->members)) {
+		list = list_first_entry(&qgroup->members,
+					struct btrfs_qgroup_list, next_member);
+		list_del(&list->next_group);
+		list_del(&list->next_member);
+		kfree(list);
+	}
+	kfree(qgroup);
+}
+
+/* must be called with qgroup_lock held */
+static int del_qgroup_rb(struct btrfs_fs_info *fs_info, u64 qgroupid)
+{
+	struct btrfs_qgroup *qgroup = find_qgroup_rb(fs_info, qgroupid);
+
+	if (!qgroup)
+		return -ENOENT;
+
+	rb_erase(&qgroup->node, &fs_info->qgroup_tree);
+	__del_qgroup_rb(qgroup);
+	return 0;
+}
+
+/* must be called with qgroup_lock held */
+static int add_relation_rb(struct btrfs_fs_info *fs_info,
+			   u64 memberid, u64 parentid)
+{
+	struct btrfs_qgroup *member;
+	struct btrfs_qgroup *parent;
+	struct btrfs_qgroup_list *list;
+
+	member = find_qgroup_rb(fs_info, memberid);
+	parent = find_qgroup_rb(fs_info, parentid);
+	if (!member || !parent)
+		return -ENOENT;
+
+	list = kzalloc(sizeof(*list), GFP_ATOMIC);
+	if (!list)
+		return -ENOMEM;
+
+	list->group = parent;
+	list->member = member;
+	list_add_tail(&list->next_group, &member->groups);
+	list_add_tail(&list->next_member, &parent->members);
+
+	return 0;
+}
+
+/* must be called with qgroup_lock held */
+static int del_relation_rb(struct btrfs_fs_info *fs_info,
+			   u64 memberid, u64 parentid)
+{
+	struct btrfs_qgroup *member;
+	struct btrfs_qgroup *parent;
+	struct btrfs_qgroup_list *list;
+
+	member = find_qgroup_rb(fs_info, memberid);
+	parent = find_qgroup_rb(fs_info, parentid);
+	if (!member || !parent)
+		return -ENOENT;
+
+	list_for_each_entry(list, &member->groups, next_group) {
+		if (list->group == parent) {
+			list_del(&list->next_group);
+			list_del(&list->next_member);
+			kfree(list);
+			return 0;
+		}
+	}
+	return -ENOENT;
+}
+
+#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
+int btrfs_verify_qgroup_counts(struct btrfs_fs_info *fs_info, u64 qgroupid,
+			       u64 rfer, u64 excl)
+{
+	struct btrfs_qgroup *qgroup;
+
+	qgroup = find_qgroup_rb(fs_info, qgroupid);
+	if (!qgroup)
+		return -EINVAL;
+	if (qgroup->rfer != rfer || qgroup->excl != excl)
+		return -EINVAL;
+	return 0;
+}
+#endif
+
+/*
+ * The full config is read in one go, only called from open_ctree()
+ * It doesn't use any locking, as at this point we're still single-threaded
+ */
+int btrfs_read_qgroup_config(struct btrfs_fs_info *fs_info)
+{
+	struct btrfs_key key;
+	struct btrfs_key found_key;
+	struct btrfs_root *quota_root = fs_info->quota_root;
+	struct btrfs_path *path = NULL;
+	struct extent_buffer *l;
+	int slot;
+	int ret = 0;
+	u64 flags = 0;
+	u64 rescan_progress = 0;
+
+	if (!fs_info->quota_enabled)
+		return 0;
+
+	fs_info->qgroup_ulist = ulist_alloc(GFP_NOFS);
+	if (!fs_info->qgroup_ulist) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	path = btrfs_alloc_path();
+	if (!path) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	/* default this to quota off, in case no status key is found */
+	fs_info->qgroup_flags = 0;
+
+	/*
+	 * pass 1: read status, all qgroup infos and limits
+	 */
+	key.objectid = 0;
+	key.type = 0;
+	key.offset = 0;
+	ret = btrfs_search_slot_for_read(quota_root, &key, path, 1, 1);
+	if (ret)
+		goto out;
+
+	while (1) {
+		struct btrfs_qgroup *qgroup;
+
+		slot = path->slots[0];
+		l = path->nodes[0];
+		btrfs_item_key_to_cpu(l, &found_key, slot);
+
+		if (found_key.type == BTRFS_QGROUP_STATUS_KEY) {
+			struct btrfs_qgroup_status_item *ptr;
+
+			ptr = btrfs_item_ptr(l, slot,
+					     struct btrfs_qgroup_status_item);
+
+			if (btrfs_qgroup_status_version(l, ptr) !=
+			    BTRFS_QGROUP_STATUS_VERSION) {
+				btrfs_err(fs_info,
+				 "old qgroup version, quota disabled");
+				goto out;
+			}
+			if (btrfs_qgroup_status_generation(l, ptr) !=
+			    fs_info->generation) {
+				flags |= BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT;
+				btrfs_err(fs_info,
+					"qgroup generation mismatch, "
+					"marked as inconsistent");
+			}
+			fs_info->qgroup_flags = btrfs_qgroup_status_flags(l,
+									  ptr);
+			rescan_progress = btrfs_qgroup_status_rescan(l, ptr);
+			goto next1;
+		}
+
+		if (found_key.type != BTRFS_QGROUP_INFO_KEY &&
+		    found_key.type != BTRFS_QGROUP_LIMIT_KEY)
+			goto next1;
+
+		qgroup = find_qgroup_rb(fs_info, found_key.offset);
+		if ((qgroup && found_key.type == BTRFS_QGROUP_INFO_KEY) ||
+		    (!qgroup && found_key.type == BTRFS_QGROUP_LIMIT_KEY)) {
+			btrfs_err(fs_info, "inconsitent qgroup config");
+			flags |= BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT;
+		}
+		if (!qgroup) {
+			qgroup = add_qgroup_rb(fs_info, found_key.offset);
+			if (IS_ERR(qgroup)) {
+				ret = PTR_ERR(qgroup);
+				goto out;
+			}
+		}
+		switch (found_key.type) {
+		case BTRFS_QGROUP_INFO_KEY: {
+			struct btrfs_qgroup_info_item *ptr;
+
+			ptr = btrfs_item_ptr(l, slot,
+					     struct btrfs_qgroup_info_item);
+			qgroup->rfer = btrfs_qgroup_info_rfer(l, ptr);
+			qgroup->rfer_cmpr = btrfs_qgroup_info_rfer_cmpr(l, ptr);
+			qgroup->excl = btrfs_qgroup_info_excl(l, ptr);
+			qgroup->excl_cmpr = btrfs_qgroup_info_excl_cmpr(l, ptr);
+			/* generation currently unused */
+			break;
+		}
+		case BTRFS_QGROUP_LIMIT_KEY: {
+			struct btrfs_qgroup_limit_item *ptr;
+
+			ptr = btrfs_item_ptr(l, slot,
+					     struct btrfs_qgroup_limit_item);
+			qgroup->lim_flags = btrfs_qgroup_limit_flags(l, ptr);
+			qgroup->max_rfer = btrfs_qgroup_limit_max_rfer(l, ptr);
+			qgroup->max_excl = btrfs_qgroup_limit_max_excl(l, ptr);
+			qgroup->rsv_rfer = btrfs_qgroup_limit_rsv_rfer(l, ptr);
+			qgroup->rsv_excl = btrfs_qgroup_limit_rsv_excl(l, ptr);
+			break;
+		}
+		}
+next1:
+		ret = btrfs_next_item(quota_root, path);
+		if (ret < 0)
+			goto out;
+		if (ret)
+			break;
+	}
+	btrfs_release_path(path);
+
+	/*
+	 * pass 2: read all qgroup relations
+	 */
+	key.objectid = 0;
+	key.type = BTRFS_QGROUP_RELATION_KEY;
+	key.offset = 0;
+	ret = btrfs_search_slot_for_read(quota_root, &key, path, 1, 0);
+	if (ret)
+		goto out;
+	while (1) {
+		slot = path->slots[0];
+		l = path->nodes[0];
+		btrfs_item_key_to_cpu(l, &found_key, slot);
+
+		if (found_key.type != BTRFS_QGROUP_RELATION_KEY)
+			goto next2;
+
+		if (found_key.objectid > found_key.offset) {
+			/* parent <- member, not needed to build config */
+			/* FIXME should we omit the key completely? */
+			goto next2;
+		}
+
+		ret = add_relation_rb(fs_info, found_key.objectid,
+				      found_key.offset);
+		if (ret == -ENOENT) {
+			btrfs_warn(fs_info,
+				"orphan qgroup relation 0x%llx->0x%llx",
+				found_key.objectid, found_key.offset);
+			ret = 0;	/* ignore the error */
+		}
+		if (ret)
+			goto out;
+next2:
+		ret = btrfs_next_item(quota_root, path);
+		if (ret < 0)
+			goto out;
+		if (ret)
+			break;
+	}
+out:
+	fs_info->qgroup_flags |= flags;
+	if (!(fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_ON)) {
+		fs_info->quota_enabled = 0;
+		fs_info->pending_quota_state = 0;
+	} else if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN &&
+		   ret >= 0) {
+		ret = qgroup_rescan_init(fs_info, rescan_progress, 0);
+	}
+	btrfs_free_path(path);
+
+	if (ret < 0) {
+		ulist_free(fs_info->qgroup_ulist);
+		fs_info->qgroup_ulist = NULL;
+		fs_info->qgroup_flags &= ~BTRFS_QGROUP_STATUS_FLAG_RESCAN;
+	}
+
+	return ret < 0 ? ret : 0;
+}
+
+/*
+ * This is called from close_ctree() or open_ctree() or btrfs_quota_disable(),
+ * first two are in single-threaded paths.And for the third one, we have set
+ * quota_root to be null with qgroup_lock held before, so it is safe to clean
+ * up the in-memory structures without qgroup_lock held.
+ */
+void btrfs_free_qgroup_config(struct btrfs_fs_info *fs_info)
+{
+	struct rb_node *n;
+	struct btrfs_qgroup *qgroup;
+
+	while ((n = rb_first(&fs_info->qgroup_tree))) {
+		qgroup = rb_entry(n, struct btrfs_qgroup, node);
+		rb_erase(n, &fs_info->qgroup_tree);
+		__del_qgroup_rb(qgroup);
+	}
+	/*
+	 * we call btrfs_free_qgroup_config() when umounting
+	 * filesystem and disabling quota, so we set qgroup_ulit
+	 * to be null here to avoid double free.
+	 */
+	ulist_free(fs_info->qgroup_ulist);
+	fs_info->qgroup_ulist = NULL;
+}
+
+static int add_qgroup_relation_item(struct btrfs_trans_handle *trans,
+				    struct btrfs_root *quota_root,
+				    u64 src, u64 dst)
+{
+	int ret;
+	struct btrfs_path *path;
+	struct btrfs_key key;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	key.objectid = src;
+	key.type = BTRFS_QGROUP_RELATION_KEY;
+	key.offset = dst;
+
+	ret = btrfs_insert_empty_item(trans, quota_root, path, &key, 0);
+
+	btrfs_mark_buffer_dirty(path->nodes[0]);
+
+	btrfs_free_path(path);
+	return ret;
+}
+
+static int del_qgroup_relation_item(struct btrfs_trans_handle *trans,
+				    struct btrfs_root *quota_root,
+				    u64 src, u64 dst)
+{
+	int ret;
+	struct btrfs_path *path;
+	struct btrfs_key key;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	key.objectid = src;
+	key.type = BTRFS_QGROUP_RELATION_KEY;
+	key.offset = dst;
+
+	ret = btrfs_search_slot(trans, quota_root, &key, path, -1, 1);
+	if (ret < 0)
+		goto out;
+
+	if (ret > 0) {
+		ret = -ENOENT;
+		goto out;
+	}
+
+	ret = btrfs_del_item(trans, quota_root, path);
+out:
+	btrfs_free_path(path);
+	return ret;
+}
+
+static int add_qgroup_item(struct btrfs_trans_handle *trans,
+			   struct btrfs_root *quota_root, u64 qgroupid)
+{
+	int ret;
+	struct btrfs_path *path;
+	struct btrfs_qgroup_info_item *qgroup_info;
+	struct btrfs_qgroup_limit_item *qgroup_limit;
+	struct extent_buffer *leaf;
+	struct btrfs_key key;
+
+#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
+	if (unlikely(test_bit(BTRFS_ROOT_DUMMY_ROOT, &quota_root->state)))
+		return 0;
+#endif
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	key.objectid = 0;
+	key.type = BTRFS_QGROUP_INFO_KEY;
+	key.offset = qgroupid;
+
+	ret = btrfs_insert_empty_item(trans, quota_root, path, &key,
+				      sizeof(*qgroup_info));
+	if (ret)
+		goto out;
+
+	leaf = path->nodes[0];
+	qgroup_info = btrfs_item_ptr(leaf, path->slots[0],
+				 struct btrfs_qgroup_info_item);
+	btrfs_set_qgroup_info_generation(leaf, qgroup_info, trans->transid);
+	btrfs_set_qgroup_info_rfer(leaf, qgroup_info, 0);
+	btrfs_set_qgroup_info_rfer_cmpr(leaf, qgroup_info, 0);
+	btrfs_set_qgroup_info_excl(leaf, qgroup_info, 0);
+	btrfs_set_qgroup_info_excl_cmpr(leaf, qgroup_info, 0);
+
+	btrfs_mark_buffer_dirty(leaf);
+
+	btrfs_release_path(path);
+
+	key.type = BTRFS_QGROUP_LIMIT_KEY;
+	ret = btrfs_insert_empty_item(trans, quota_root, path, &key,
+				      sizeof(*qgroup_limit));
+	if (ret)
+		goto out;
+
+	leaf = path->nodes[0];
+	qgroup_limit = btrfs_item_ptr(leaf, path->slots[0],
+				  struct btrfs_qgroup_limit_item);
+	btrfs_set_qgroup_limit_flags(leaf, qgroup_limit, 0);
+	btrfs_set_qgroup_limit_max_rfer(leaf, qgroup_limit, 0);
+	btrfs_set_qgroup_limit_max_excl(leaf, qgroup_limit, 0);
+	btrfs_set_qgroup_limit_rsv_rfer(leaf, qgroup_limit, 0);
+	btrfs_set_qgroup_limit_rsv_excl(leaf, qgroup_limit, 0);
+
+	btrfs_mark_buffer_dirty(leaf);
+
+	ret = 0;
+out:
+	btrfs_free_path(path);
+	return ret;
+}
+
+static int del_qgroup_item(struct btrfs_trans_handle *trans,
+			   struct btrfs_root *quota_root, u64 qgroupid)
+{
+	int ret;
+	struct btrfs_path *path;
+	struct btrfs_key key;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	key.objectid = 0;
+	key.type = BTRFS_QGROUP_INFO_KEY;
+	key.offset = qgroupid;
+	ret = btrfs_search_slot(trans, quota_root, &key, path, -1, 1);
+	if (ret < 0)
+		goto out;
+
+	if (ret > 0) {
+		ret = -ENOENT;
+		goto out;
+	}
+
+	ret = btrfs_del_item(trans, quota_root, path);
+	if (ret)
+		goto out;
+
+	btrfs_release_path(path);
+
+	key.type = BTRFS_QGROUP_LIMIT_KEY;
+	ret = btrfs_search_slot(trans, quota_root, &key, path, -1, 1);
+	if (ret < 0)
+		goto out;
+
+	if (ret > 0) {
+		ret = -ENOENT;
+		goto out;
+	}
+
+	ret = btrfs_del_item(trans, quota_root, path);
+
+out:
+	btrfs_free_path(path);
+	return ret;
+}
+
+static int update_qgroup_limit_item(struct btrfs_trans_handle *trans,
+				    struct btrfs_root *root, u64 qgroupid,
+				    u64 flags, u64 max_rfer, u64 max_excl,
+				    u64 rsv_rfer, u64 rsv_excl)
+{
+	struct btrfs_path *path;
+	struct btrfs_key key;
+	struct extent_buffer *l;
+	struct btrfs_qgroup_limit_item *qgroup_limit;
+	int ret;
+	int slot;
+
+	key.objectid = 0;
+	key.type = BTRFS_QGROUP_LIMIT_KEY;
+	key.offset = qgroupid;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
+	if (ret > 0)
+		ret = -ENOENT;
+
+	if (ret)
+		goto out;
+
+	l = path->nodes[0];
+	slot = path->slots[0];
+	qgroup_limit = btrfs_item_ptr(l, slot, struct btrfs_qgroup_limit_item);
+	btrfs_set_qgroup_limit_flags(l, qgroup_limit, flags);
+	btrfs_set_qgroup_limit_max_rfer(l, qgroup_limit, max_rfer);
+	btrfs_set_qgroup_limit_max_excl(l, qgroup_limit, max_excl);
+	btrfs_set_qgroup_limit_rsv_rfer(l, qgroup_limit, rsv_rfer);
+	btrfs_set_qgroup_limit_rsv_excl(l, qgroup_limit, rsv_excl);
+
+	btrfs_mark_buffer_dirty(l);
+
+out:
+	btrfs_free_path(path);
+	return ret;
+}
+
+static int update_qgroup_info_item(struct btrfs_trans_handle *trans,
+				   struct btrfs_root *root,
+				   struct btrfs_qgroup *qgroup)
+{
+	struct btrfs_path *path;
+	struct btrfs_key key;
+	struct extent_buffer *l;
+	struct btrfs_qgroup_info_item *qgroup_info;
+	int ret;
+	int slot;
+
+#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
+	if (unlikely(test_bit(BTRFS_ROOT_DUMMY_ROOT, &root->state)))
+		return 0;
+#endif
+	key.objectid = 0;
+	key.type = BTRFS_QGROUP_INFO_KEY;
+	key.offset = qgroup->qgroupid;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
+	if (ret > 0)
+		ret = -ENOENT;
+
+	if (ret)
+		goto out;
+
+	l = path->nodes[0];
+	slot = path->slots[0];
+	qgroup_info = btrfs_item_ptr(l, slot, struct btrfs_qgroup_info_item);
+	btrfs_set_qgroup_info_generation(l, qgroup_info, trans->transid);
+	btrfs_set_qgroup_info_rfer(l, qgroup_info, qgroup->rfer);
+	btrfs_set_qgroup_info_rfer_cmpr(l, qgroup_info, qgroup->rfer_cmpr);
+	btrfs_set_qgroup_info_excl(l, qgroup_info, qgroup->excl);
+	btrfs_set_qgroup_info_excl_cmpr(l, qgroup_info, qgroup->excl_cmpr);
+
+	btrfs_mark_buffer_dirty(l);
+
+out:
+	btrfs_free_path(path);
+	return ret;
+}
+
+static int update_qgroup_status_item(struct btrfs_trans_handle *trans,
+				     struct btrfs_fs_info *fs_info,
+				    struct btrfs_root *root)
+{
+	struct btrfs_path *path;
+	struct btrfs_key key;
+	struct extent_buffer *l;
+	struct btrfs_qgroup_status_item *ptr;
+	int ret;
+	int slot;
+
+	key.objectid = 0;
+	key.type = BTRFS_QGROUP_STATUS_KEY;
+	key.offset = 0;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
+	if (ret > 0)
+		ret = -ENOENT;
+
+	if (ret)
+		goto out;
+
+	l = path->nodes[0];
+	slot = path->slots[0];
+	ptr = btrfs_item_ptr(l, slot, struct btrfs_qgroup_status_item);
+	btrfs_set_qgroup_status_flags(l, ptr, fs_info->qgroup_flags);
+	btrfs_set_qgroup_status_generation(l, ptr, trans->transid);
+	btrfs_set_qgroup_status_rescan(l, ptr,
+				fs_info->qgroup_rescan_progress.objectid);
+
+	btrfs_mark_buffer_dirty(l);
+
+out:
+	btrfs_free_path(path);
+	return ret;
+}
+
+/*
+ * called with qgroup_lock held
+ */
+static int btrfs_clean_quota_tree(struct btrfs_trans_handle *trans,
+				  struct btrfs_root *root)
+{
+	struct btrfs_path *path;
+	struct btrfs_key key;
+	struct extent_buffer *leaf = NULL;
+	int ret;
+	int nr = 0;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	path->leave_spinning = 1;
+
+	key.objectid = 0;
+	key.offset = 0;
+	key.type = 0;
+
+	while (1) {
+		ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
+		if (ret < 0)
+			goto out;
+		leaf = path->nodes[0];
+		nr = btrfs_header_nritems(leaf);
+		if (!nr)
+			break;
+		/*
+		 * delete the leaf one by one
+		 * since the whole tree is going
+		 * to be deleted.
+		 */
+		path->slots[0] = 0;
+		ret = btrfs_del_items(trans, root, path, 0, nr);
+		if (ret)
+			goto out;
+
+		btrfs_release_path(path);
+	}
+	ret = 0;
+out:
+	root->fs_info->pending_quota_state = 0;
+	btrfs_free_path(path);
+	return ret;
+}
+
+int btrfs_quota_enable(struct btrfs_trans_handle *trans,
+		       struct btrfs_fs_info *fs_info)
+{
+	struct btrfs_root *quota_root;
+	struct btrfs_root *tree_root = fs_info->tree_root;
+	struct btrfs_path *path = NULL;
+	struct btrfs_qgroup_status_item *ptr;
+	struct extent_buffer *leaf;
+	struct btrfs_key key;
+	struct btrfs_key found_key;
+	struct btrfs_qgroup *qgroup = NULL;
+	int ret = 0;
+	int slot;
+
+	mutex_lock(&fs_info->qgroup_ioctl_lock);
+	if (fs_info->quota_root) {
+		fs_info->pending_quota_state = 1;
+		goto out;
+	}
+
+	fs_info->qgroup_ulist = ulist_alloc(GFP_NOFS);
+	if (!fs_info->qgroup_ulist) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	/*
+	 * initially create the quota tree
+	 */
+	quota_root = btrfs_create_tree(trans, fs_info,
+				       BTRFS_QUOTA_TREE_OBJECTID);
+	if (IS_ERR(quota_root)) {
+		ret =  PTR_ERR(quota_root);
+		goto out;
+	}
+
+	path = btrfs_alloc_path();
+	if (!path) {
+		ret = -ENOMEM;
+		goto out_free_root;
+	}
+
+	key.objectid = 0;
+	key.type = BTRFS_QGROUP_STATUS_KEY;
+	key.offset = 0;
+
+	ret = btrfs_insert_empty_item(trans, quota_root, path, &key,
+				      sizeof(*ptr));
+	if (ret)
+		goto out_free_path;
+
+	leaf = path->nodes[0];
+	ptr = btrfs_item_ptr(leaf, path->slots[0],
+				 struct btrfs_qgroup_status_item);
+	btrfs_set_qgroup_status_generation(leaf, ptr, trans->transid);
+	btrfs_set_qgroup_status_version(leaf, ptr, BTRFS_QGROUP_STATUS_VERSION);
+	fs_info->qgroup_flags = BTRFS_QGROUP_STATUS_FLAG_ON |
+				BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT;
+	btrfs_set_qgroup_status_flags(leaf, ptr, fs_info->qgroup_flags);
+	btrfs_set_qgroup_status_rescan(leaf, ptr, 0);
+
+	btrfs_mark_buffer_dirty(leaf);
+
+	key.objectid = 0;
+	key.type = BTRFS_ROOT_REF_KEY;
+	key.offset = 0;
+
+	btrfs_release_path(path);
+	ret = btrfs_search_slot_for_read(tree_root, &key, path, 1, 0);
+	if (ret > 0)
+		goto out_add_root;
+	if (ret < 0)
+		goto out_free_path;
+
+
+	while (1) {
+		slot = path->slots[0];
+		leaf = path->nodes[0];
+		btrfs_item_key_to_cpu(leaf, &found_key, slot);
+
+		if (found_key.type == BTRFS_ROOT_REF_KEY) {
+			ret = add_qgroup_item(trans, quota_root,
+					      found_key.offset);
+			if (ret)
+				goto out_free_path;
+
+			qgroup = add_qgroup_rb(fs_info, found_key.offset);
+			if (IS_ERR(qgroup)) {
+				ret = PTR_ERR(qgroup);
+				goto out_free_path;
+			}
+		}
+		ret = btrfs_next_item(tree_root, path);
+		if (ret < 0)
+			goto out_free_path;
+		if (ret)
+			break;
+	}
+
+out_add_root:
+	btrfs_release_path(path);
+	ret = add_qgroup_item(trans, quota_root, BTRFS_FS_TREE_OBJECTID);
+	if (ret)
+		goto out_free_path;
+
+	qgroup = add_qgroup_rb(fs_info, BTRFS_FS_TREE_OBJECTID);
+	if (IS_ERR(qgroup)) {
+		ret = PTR_ERR(qgroup);
+		goto out_free_path;
+	}
+	spin_lock(&fs_info->qgroup_lock);
+	fs_info->quota_root = quota_root;
+	fs_info->pending_quota_state = 1;
+	spin_unlock(&fs_info->qgroup_lock);
+out_free_path:
+	btrfs_free_path(path);
+out_free_root:
+	if (ret) {
+		free_extent_buffer(quota_root->node);
+		free_extent_buffer(quota_root->commit_root);
+		kfree(quota_root);
+	}
+out:
+	if (ret) {
+		ulist_free(fs_info->qgroup_ulist);
+		fs_info->qgroup_ulist = NULL;
+	}
+	mutex_unlock(&fs_info->qgroup_ioctl_lock);
+	return ret;
+}
+
+int btrfs_quota_disable(struct btrfs_trans_handle *trans,
+			struct btrfs_fs_info *fs_info)
+{
+	struct btrfs_root *tree_root = fs_info->tree_root;
+	struct btrfs_root *quota_root;
+	int ret = 0;
+
+	mutex_lock(&fs_info->qgroup_ioctl_lock);
+	if (!fs_info->quota_root)
+		goto out;
+	spin_lock(&fs_info->qgroup_lock);
+	fs_info->quota_enabled = 0;
+	fs_info->pending_quota_state = 0;
+	quota_root = fs_info->quota_root;
+	fs_info->quota_root = NULL;
+	spin_unlock(&fs_info->qgroup_lock);
+
+	btrfs_free_qgroup_config(fs_info);
+
+	ret = btrfs_clean_quota_tree(trans, quota_root);
+	if (ret)
+		goto out;
+
+	ret = btrfs_del_root(trans, tree_root, &quota_root->root_key);
+	if (ret)
+		goto out;
+
+	list_del(&quota_root->dirty_list);
+
+	btrfs_tree_lock(quota_root->node);
+	clean_tree_block(trans, tree_root, quota_root->node);
+	btrfs_tree_unlock(quota_root->node);
+	btrfs_free_tree_block(trans, quota_root, quota_root->node, 0, 1);
+
+	free_extent_buffer(quota_root->node);
+	free_extent_buffer(quota_root->commit_root);
+	kfree(quota_root);
+out:
+	mutex_unlock(&fs_info->qgroup_ioctl_lock);
+	return ret;
+}
+
+static void qgroup_dirty(struct btrfs_fs_info *fs_info,
+			 struct btrfs_qgroup *qgroup)
+{
+	if (list_empty(&qgroup->dirty))
+		list_add(&qgroup->dirty, &fs_info->dirty_qgroups);
+}
+
+int btrfs_add_qgroup_relation(struct btrfs_trans_handle *trans,
+			      struct btrfs_fs_info *fs_info, u64 src, u64 dst)
+{
+	struct btrfs_root *quota_root;
+	struct btrfs_qgroup *parent;
+	struct btrfs_qgroup *member;
+	struct btrfs_qgroup_list *list;
+	int ret = 0;
+
+	mutex_lock(&fs_info->qgroup_ioctl_lock);
+	quota_root = fs_info->quota_root;
+	if (!quota_root) {
+		ret = -EINVAL;
+		goto out;
+	}
+	member = find_qgroup_rb(fs_info, src);
+	parent = find_qgroup_rb(fs_info, dst);
+	if (!member || !parent) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	/* check if such qgroup relation exist firstly */
+	list_for_each_entry(list, &member->groups, next_group) {
+		if (list->group == parent) {
+			ret = -EEXIST;
+			goto out;
+		}
+	}
+
+	ret = add_qgroup_relation_item(trans, quota_root, src, dst);
+	if (ret)
+		goto out;
+
+	ret = add_qgroup_relation_item(trans, quota_root, dst, src);
+	if (ret) {
+		del_qgroup_relation_item(trans, quota_root, src, dst);
+		goto out;
+	}
+
+	spin_lock(&fs_info->qgroup_lock);
+	ret = add_relation_rb(quota_root->fs_info, src, dst);
+	spin_unlock(&fs_info->qgroup_lock);
+out:
+	mutex_unlock(&fs_info->qgroup_ioctl_lock);
+	return ret;
+}
+
+int btrfs_del_qgroup_relation(struct btrfs_trans_handle *trans,
+			      struct btrfs_fs_info *fs_info, u64 src, u64 dst)
+{
+	struct btrfs_root *quota_root;
+	struct btrfs_qgroup *parent;
+	struct btrfs_qgroup *member;
+	struct btrfs_qgroup_list *list;
+	int ret = 0;
+	int err;
+
+	mutex_lock(&fs_info->qgroup_ioctl_lock);
+	quota_root = fs_info->quota_root;
+	if (!quota_root) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	member = find_qgroup_rb(fs_info, src);
+	parent = find_qgroup_rb(fs_info, dst);
+	if (!member || !parent) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	/* check if such qgroup relation exist firstly */
+	list_for_each_entry(list, &member->groups, next_group) {
+		if (list->group == parent)
+			goto exist;
+	}
+	ret = -ENOENT;
+	goto out;
+exist:
+	ret = del_qgroup_relation_item(trans, quota_root, src, dst);
+	err = del_qgroup_relation_item(trans, quota_root, dst, src);
+	if (err && !ret)
+		ret = err;
+
+	spin_lock(&fs_info->qgroup_lock);
+	del_relation_rb(fs_info, src, dst);
+	spin_unlock(&fs_info->qgroup_lock);
+out:
+	mutex_unlock(&fs_info->qgroup_ioctl_lock);
+	return ret;
+}
+
+int btrfs_create_qgroup(struct btrfs_trans_handle *trans,
+			struct btrfs_fs_info *fs_info, u64 qgroupid, char *name)
+{
+	struct btrfs_root *quota_root;
+	struct btrfs_qgroup *qgroup;
+	int ret = 0;
+
+	mutex_lock(&fs_info->qgroup_ioctl_lock);
+	quota_root = fs_info->quota_root;
+	if (!quota_root) {
+		ret = -EINVAL;
+		goto out;
+	}
+	qgroup = find_qgroup_rb(fs_info, qgroupid);
+	if (qgroup) {
+		ret = -EEXIST;
+		goto out;
+	}
+
+	ret = add_qgroup_item(trans, quota_root, qgroupid);
+	if (ret)
+		goto out;
+
+	spin_lock(&fs_info->qgroup_lock);
+	qgroup = add_qgroup_rb(fs_info, qgroupid);
+	spin_unlock(&fs_info->qgroup_lock);
+
+	if (IS_ERR(qgroup))
+		ret = PTR_ERR(qgroup);
+out:
+	mutex_unlock(&fs_info->qgroup_ioctl_lock);
+	return ret;
+}
+
+int btrfs_remove_qgroup(struct btrfs_trans_handle *trans,
+			struct btrfs_fs_info *fs_info, u64 qgroupid)
+{
+	struct btrfs_root *quota_root;
+	struct btrfs_qgroup *qgroup;
+	int ret = 0;
+
+	mutex_lock(&fs_info->qgroup_ioctl_lock);
+	quota_root = fs_info->quota_root;
+	if (!quota_root) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	qgroup = find_qgroup_rb(fs_info, qgroupid);
+	if (!qgroup) {
+		ret = -ENOENT;
+		goto out;
+	} else {
+		/* check if there are no relations to this qgroup */
+		if (!list_empty(&qgroup->groups) ||
+		    !list_empty(&qgroup->members)) {
+			ret = -EBUSY;
+			goto out;
+		}
+	}
+	ret = del_qgroup_item(trans, quota_root, qgroupid);
+
+	spin_lock(&fs_info->qgroup_lock);
+	del_qgroup_rb(quota_root->fs_info, qgroupid);
+	spin_unlock(&fs_info->qgroup_lock);
+out:
+	mutex_unlock(&fs_info->qgroup_ioctl_lock);
+	return ret;
+}
+
+int btrfs_limit_qgroup(struct btrfs_trans_handle *trans,
+		       struct btrfs_fs_info *fs_info, u64 qgroupid,
+		       struct btrfs_qgroup_limit *limit)
+{
+	struct btrfs_root *quota_root;
+	struct btrfs_qgroup *qgroup;
+	int ret = 0;
+
+	mutex_lock(&fs_info->qgroup_ioctl_lock);
+	quota_root = fs_info->quota_root;
+	if (!quota_root) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	qgroup = find_qgroup_rb(fs_info, qgroupid);
+	if (!qgroup) {
+		ret = -ENOENT;
+		goto out;
+	}
+	ret = update_qgroup_limit_item(trans, quota_root, qgroupid,
+				       limit->flags, limit->max_rfer,
+				       limit->max_excl, limit->rsv_rfer,
+				       limit->rsv_excl);
+	if (ret) {
+		fs_info->qgroup_flags |= BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT;
+		btrfs_info(fs_info, "unable to update quota limit for %llu",
+		       qgroupid);
+	}
+
+	spin_lock(&fs_info->qgroup_lock);
+	qgroup->lim_flags = limit->flags;
+	qgroup->max_rfer = limit->max_rfer;
+	qgroup->max_excl = limit->max_excl;
+	qgroup->rsv_rfer = limit->rsv_rfer;
+	qgroup->rsv_excl = limit->rsv_excl;
+	spin_unlock(&fs_info->qgroup_lock);
+out:
+	mutex_unlock(&fs_info->qgroup_ioctl_lock);
+	return ret;
+}
+static int comp_oper(struct btrfs_qgroup_operation *oper1,
+		     struct btrfs_qgroup_operation *oper2)
+{
+	if (oper1->bytenr < oper2->bytenr)
+		return -1;
+	if (oper1->bytenr > oper2->bytenr)
+		return 1;
+	if (oper1->seq < oper2->seq)
+		return -1;
+	if (oper1->seq > oper2->seq)
+		return -1;
+	if (oper1->ref_root < oper2->ref_root)
+		return -1;
+	if (oper1->ref_root > oper2->ref_root)
+		return 1;
+	if (oper1->type < oper2->type)
+		return -1;
+	if (oper1->type > oper2->type)
+		return 1;
+	return 0;
+}
+
+static int insert_qgroup_oper(struct btrfs_fs_info *fs_info,
+			      struct btrfs_qgroup_operation *oper)
+{
+	struct rb_node **p;
+	struct rb_node *parent = NULL;
+	struct btrfs_qgroup_operation *cur;
+	int cmp;
+
+	spin_lock(&fs_info->qgroup_op_lock);
+	p = &fs_info->qgroup_op_tree.rb_node;
+	while (*p) {
+		parent = *p;
+		cur = rb_entry(parent, struct btrfs_qgroup_operation, n);
+		cmp = comp_oper(cur, oper);
+		if (cmp < 0) {
+			p = &(*p)->rb_right;
+		} else if (cmp) {
+			p = &(*p)->rb_left;
+		} else {
+			spin_unlock(&fs_info->qgroup_op_lock);
+			return -EEXIST;
+		}
+	}
+	rb_link_node(&oper->n, parent, p);
+	rb_insert_color(&oper->n, &fs_info->qgroup_op_tree);
+	spin_unlock(&fs_info->qgroup_op_lock);
+	return 0;
+}
+
+/*
+ * Record a quota operation for processing later on.
+ * @trans: the transaction we are adding the delayed op to.
+ * @fs_info: the fs_info for this fs.
+ * @ref_root: the root of the reference we are acting on,
+ * @bytenr: the bytenr we are acting on.
+ * @num_bytes: the number of bytes in the reference.
+ * @type: the type of operation this is.
+ * @mod_seq: do we need to get a sequence number for looking up roots.
+ *
+ * We just add it to our trans qgroup_ref_list and carry on and process these
+ * operations in order at some later point.  If the reference root isn't a fs
+ * root then we don't bother with doing anything.
+ *
+ * MUST BE HOLDING THE REF LOCK.
+ */
+int btrfs_qgroup_record_ref(struct btrfs_trans_handle *trans,
+			    struct btrfs_fs_info *fs_info, u64 ref_root,
+			    u64 bytenr, u64 num_bytes,
+			    enum btrfs_qgroup_operation_type type, int mod_seq)
+{
+	struct btrfs_qgroup_operation *oper;
+	int ret;
+
+	if (!is_fstree(ref_root) || !fs_info->quota_enabled)
+		return 0;
+
+	oper = kmalloc(sizeof(*oper), GFP_NOFS);
+	if (!oper)
+		return -ENOMEM;
+
+	oper->ref_root = ref_root;
+	oper->bytenr = bytenr;
+	oper->num_bytes = num_bytes;
+	oper->type = type;
+	oper->seq = atomic_inc_return(&fs_info->qgroup_op_seq);
+	INIT_LIST_HEAD(&oper->elem.list);
+	oper->elem.seq = 0;
+	ret = insert_qgroup_oper(fs_info, oper);
+	if (ret) {
+		/* Shouldn't happen so have an assert for developers */
+		ASSERT(0);
+		kfree(oper);
+		return ret;
+	}
+	list_add_tail(&oper->list, &trans->qgroup_ref_list);
+
+	if (mod_seq)
+		btrfs_get_tree_mod_seq(fs_info, &oper->elem);
+
+	return 0;
+}
+
+/*
+ * The easy accounting, if we are adding/removing the only ref for an extent
+ * then this qgroup and all of the parent qgroups get their refrence and
+ * exclusive counts adjusted.
+ */
+static int qgroup_excl_accounting(struct btrfs_fs_info *fs_info,
+				  struct btrfs_qgroup_operation *oper)
+{
+	struct btrfs_qgroup *qgroup;
+	struct ulist *tmp;
+	struct btrfs_qgroup_list *glist;
+	struct ulist_node *unode;
+	struct ulist_iterator uiter;
+	int sign = 0;
+	int ret = 0;
+
+	tmp = ulist_alloc(GFP_NOFS);
+	if (!tmp)
+		return -ENOMEM;
+
+	spin_lock(&fs_info->qgroup_lock);
+	if (!fs_info->quota_root)
+		goto out;
+	qgroup = find_qgroup_rb(fs_info, oper->ref_root);
+	if (!qgroup)
+		goto out;
+	switch (oper->type) {
+	case BTRFS_QGROUP_OPER_ADD_EXCL:
+		sign = 1;
+		break;
+	case BTRFS_QGROUP_OPER_SUB_EXCL:
+		sign = -1;
+		break;
+	default:
+		ASSERT(0);
+	}
+	qgroup->rfer += sign * oper->num_bytes;
+	qgroup->rfer_cmpr += sign * oper->num_bytes;
+
+	WARN_ON(sign < 0 && qgroup->excl < oper->num_bytes);
+	qgroup->excl += sign * oper->num_bytes;
+	qgroup->excl_cmpr += sign * oper->num_bytes;
+
+	qgroup_dirty(fs_info, qgroup);
+
+	/* Get all of the parent groups that contain this qgroup */
+	list_for_each_entry(glist, &qgroup->groups, next_group) {
+		ret = ulist_add(tmp, glist->group->qgroupid,
+				ptr_to_u64(glist->group), GFP_ATOMIC);
+		if (ret < 0)
+			goto out;
+	}
+
+	/* Iterate all of the parents and adjust their reference counts */
+	ULIST_ITER_INIT(&uiter);
+	while ((unode = ulist_next(tmp, &uiter))) {
+		qgroup = u64_to_ptr(unode->aux);
+		qgroup->rfer += sign * oper->num_bytes;
+		qgroup->rfer_cmpr += sign * oper->num_bytes;
+		qgroup->excl += sign * oper->num_bytes;
+		if (sign < 0)
+			WARN_ON(qgroup->excl < oper->num_bytes);
+		qgroup->excl_cmpr += sign * oper->num_bytes;
+		qgroup_dirty(fs_info, qgroup);
+
+		/* Add any parents of the parents */
+		list_for_each_entry(glist, &qgroup->groups, next_group) {
+			ret = ulist_add(tmp, glist->group->qgroupid,
+					ptr_to_u64(glist->group), GFP_ATOMIC);
+			if (ret < 0)
+				goto out;
+		}
+	}
+	ret = 0;
+out:
+	spin_unlock(&fs_info->qgroup_lock);
+	ulist_free(tmp);
+	return ret;
+}
+
+/*
+ * Walk all of the roots that pointed to our bytenr and adjust their refcnts as
+ * properly.
+ */
+static int qgroup_calc_old_refcnt(struct btrfs_fs_info *fs_info,
+				  u64 root_to_skip, struct ulist *tmp,
+				  struct ulist *roots, struct ulist *qgroups,
+				  u64 seq, int *old_roots, int rescan)
+{
+	struct ulist_node *unode;
+	struct ulist_iterator uiter;
+	struct ulist_node *tmp_unode;
+	struct ulist_iterator tmp_uiter;
+	struct btrfs_qgroup *qg;
+	int ret;
+
+	ULIST_ITER_INIT(&uiter);
+	while ((unode = ulist_next(roots, &uiter))) {
+		/* We don't count our current root here */
+		if (unode->val == root_to_skip)
+			continue;
+		qg = find_qgroup_rb(fs_info, unode->val);
+		if (!qg)
+			continue;
+		/*
+		 * We could have a pending removal of this same ref so we may
+		 * not have actually found our ref root when doing
+		 * btrfs_find_all_roots, so we need to keep track of how many
+		 * old roots we find in case we removed ours and added a
+		 * different one at the same time.  I don't think this could
+		 * happen in practice but that sort of thinking leads to pain
+		 * and suffering and to the dark side.
+		 */
+		(*old_roots)++;
+
+		ulist_reinit(tmp);
+		ret = ulist_add(qgroups, qg->qgroupid, ptr_to_u64(qg),
+				GFP_ATOMIC);
+		if (ret < 0)
+			return ret;
+		ret = ulist_add(tmp, qg->qgroupid, ptr_to_u64(qg), GFP_ATOMIC);
+		if (ret < 0)
+			return ret;
+		ULIST_ITER_INIT(&tmp_uiter);
+		while ((tmp_unode = ulist_next(tmp, &tmp_uiter))) {
+			struct btrfs_qgroup_list *glist;
+
+			qg = u64_to_ptr(tmp_unode->aux);
+			/*
+			 * We use this sequence number to keep from having to
+			 * run the whole list and 0 out the refcnt every time.
+			 * We basically use sequnce as the known 0 count and
+			 * then add 1 everytime we see a qgroup.  This is how we
+			 * get how many of the roots actually point up to the
+			 * upper level qgroups in order to determine exclusive
+			 * counts.
+			 *
+			 * For rescan we want to set old_refcnt to seq so our
+			 * exclusive calculations end up correct.
+			 */
+			if (rescan)
+				qg->old_refcnt = seq;
+			else if (qg->old_refcnt < seq)
+				qg->old_refcnt = seq + 1;
+			else
+				qg->old_refcnt++;
+
+			if (qg->new_refcnt < seq)
+				qg->new_refcnt = seq + 1;
+			else
+				qg->new_refcnt++;
+			list_for_each_entry(glist, &qg->groups, next_group) {
+				ret = ulist_add(qgroups, glist->group->qgroupid,
+						ptr_to_u64(glist->group),
+						GFP_ATOMIC);
+				if (ret < 0)
+					return ret;
+				ret = ulist_add(tmp, glist->group->qgroupid,
+						ptr_to_u64(glist->group),
+						GFP_ATOMIC);
+				if (ret < 0)
+					return ret;
+			}
+		}
+	}
+	return 0;
+}
+
+/*
+ * We need to walk forward in our operation tree and account for any roots that
+ * were deleted after we made this operation.
+ */
+static int qgroup_account_deleted_refs(struct btrfs_fs_info *fs_info,
+				       struct btrfs_qgroup_operation *oper,
+				       struct ulist *tmp,
+				       struct ulist *qgroups, u64 seq,
+				       int *old_roots)
+{
+	struct ulist_node *unode;
+	struct ulist_iterator uiter;
+	struct btrfs_qgroup *qg;
+	struct btrfs_qgroup_operation *tmp_oper;
+	struct rb_node *n;
+	int ret;
+
+	ulist_reinit(tmp);
+
+	/*
+	 * We only walk forward in the tree since we're only interested in
+	 * removals that happened _after_  our operation.
+	 */
+	spin_lock(&fs_info->qgroup_op_lock);
+	n = rb_next(&oper->n);
+	spin_unlock(&fs_info->qgroup_op_lock);
+	if (!n)
+		return 0;
+	tmp_oper = rb_entry(n, struct btrfs_qgroup_operation, n);
+	while (tmp_oper->bytenr == oper->bytenr) {
+		/*
+		 * If it's not a removal we don't care, additions work out
+		 * properly with our refcnt tracking.
+		 */
+		if (tmp_oper->type != BTRFS_QGROUP_OPER_SUB_SHARED &&
+		    tmp_oper->type != BTRFS_QGROUP_OPER_SUB_EXCL)
+			goto next;
+		qg = find_qgroup_rb(fs_info, tmp_oper->ref_root);
+		if (!qg)
+			goto next;
+		ret = ulist_add(qgroups, qg->qgroupid, ptr_to_u64(qg),
+				GFP_ATOMIC);
+		if (ret) {
+			if (ret < 0)
+				return ret;
+			/*
+			 * We only want to increase old_roots if this qgroup is
+			 * not already in the list of qgroups.  If it is already
+			 * there then that means it must have been re-added or
+			 * the delete will be discarded because we had an
+			 * existing ref that we haven't looked up yet.  In this
+			 * case we don't want to increase old_roots.  So if ret
+			 * == 1 then we know that this is the first time we've
+			 * seen this qgroup and we can bump the old_roots.
+			 */
+			(*old_roots)++;
+			ret = ulist_add(tmp, qg->qgroupid, ptr_to_u64(qg),
+					GFP_ATOMIC);
+			if (ret < 0)
+				return ret;
+		}
+next:
+		spin_lock(&fs_info->qgroup_op_lock);
+		n = rb_next(&tmp_oper->n);
+		spin_unlock(&fs_info->qgroup_op_lock);
+		if (!n)
+			break;
+		tmp_oper = rb_entry(n, struct btrfs_qgroup_operation, n);
+	}
+
+	/* Ok now process the qgroups we found */
+	ULIST_ITER_INIT(&uiter);
+	while ((unode = ulist_next(tmp, &uiter))) {
+		struct btrfs_qgroup_list *glist;
+
+		qg = u64_to_ptr(unode->aux);
+		if (qg->old_refcnt < seq)
+			qg->old_refcnt = seq + 1;
+		else
+			qg->old_refcnt++;
+		if (qg->new_refcnt < seq)
+			qg->new_refcnt = seq + 1;
+		else
+			qg->new_refcnt++;
+		list_for_each_entry(glist, &qg->groups, next_group) {
+			ret = ulist_add(qgroups, glist->group->qgroupid,
+					ptr_to_u64(glist->group), GFP_ATOMIC);
+			if (ret < 0)
+				return ret;
+			ret = ulist_add(tmp, glist->group->qgroupid,
+					ptr_to_u64(glist->group), GFP_ATOMIC);
+			if (ret < 0)
+				return ret;
+		}
+	}
+	return 0;
+}
+
+/* Add refcnt for the newly added reference. */
+static int qgroup_calc_new_refcnt(struct btrfs_fs_info *fs_info,
+				  struct btrfs_qgroup_operation *oper,
+				  struct btrfs_qgroup *qgroup,
+				  struct ulist *tmp, struct ulist *qgroups,
+				  u64 seq)
+{
+	struct ulist_node *unode;
+	struct ulist_iterator uiter;
+	struct btrfs_qgroup *qg;
+	int ret;
+
+	ulist_reinit(tmp);
+	ret = ulist_add(qgroups, qgroup->qgroupid, ptr_to_u64(qgroup),
+			GFP_ATOMIC);
+	if (ret < 0)
+		return ret;
+	ret = ulist_add(tmp, qgroup->qgroupid, ptr_to_u64(qgroup),
+			GFP_ATOMIC);
+	if (ret < 0)
+		return ret;
+	ULIST_ITER_INIT(&uiter);
+	while ((unode = ulist_next(tmp, &uiter))) {
+		struct btrfs_qgroup_list *glist;
+
+		qg = u64_to_ptr(unode->aux);
+		if (oper->type == BTRFS_QGROUP_OPER_ADD_SHARED) {
+			if (qg->new_refcnt < seq)
+				qg->new_refcnt = seq + 1;
+			else
+				qg->new_refcnt++;
+		} else {
+			if (qg->old_refcnt < seq)
+				qg->old_refcnt = seq + 1;
+			else
+				qg->old_refcnt++;
+		}
+		list_for_each_entry(glist, &qg->groups, next_group) {
+			ret = ulist_add(tmp, glist->group->qgroupid,
+					ptr_to_u64(glist->group), GFP_ATOMIC);
+			if (ret < 0)
+				return ret;
+			ret = ulist_add(qgroups, glist->group->qgroupid,
+					ptr_to_u64(glist->group), GFP_ATOMIC);
+			if (ret < 0)
+				return ret;
+		}
+	}
+	return 0;
+}
+
+/*
+ * This adjusts the counters for all referenced qgroups if need be.
+ */
+static int qgroup_adjust_counters(struct btrfs_fs_info *fs_info,
+				  u64 root_to_skip, u64 num_bytes,
+				  struct ulist *qgroups, u64 seq,
+				  int old_roots, int new_roots, int rescan)
+{
+	struct ulist_node *unode;
+	struct ulist_iterator uiter;
+	struct btrfs_qgroup *qg;
+	u64 cur_new_count, cur_old_count;
+
+	ULIST_ITER_INIT(&uiter);
+	while ((unode = ulist_next(qgroups, &uiter))) {
+		bool dirty = false;
+
+		qg = u64_to_ptr(unode->aux);
+		/*
+		 * Wasn't referenced before but is now, add to the reference
+		 * counters.
+		 */
+		if (qg->old_refcnt <= seq && qg->new_refcnt > seq) {
+			qg->rfer += num_bytes;
+			qg->rfer_cmpr += num_bytes;
+			dirty = true;
+		}
+
+		/*
+		 * Was referenced before but isn't now, subtract from the
+		 * reference counters.
+		 */
+		if (qg->old_refcnt > seq && qg->new_refcnt <= seq) {
+			qg->rfer -= num_bytes;
+			qg->rfer_cmpr -= num_bytes;
+			dirty = true;
+		}
+
+		if (qg->old_refcnt < seq)
+			cur_old_count = 0;
+		else
+			cur_old_count = qg->old_refcnt - seq;
+		if (qg->new_refcnt < seq)
+			cur_new_count = 0;
+		else
+			cur_new_count = qg->new_refcnt - seq;
+
+		/*
+		 * If our refcount was the same as the roots previously but our
+		 * new count isn't the same as the number of roots now then we
+		 * went from having a exclusive reference on this range to not.
+		 */
+		if (old_roots && cur_old_count == old_roots &&
+		    (cur_new_count != new_roots || new_roots == 0)) {
+			WARN_ON(cur_new_count != new_roots && new_roots == 0);
+			qg->excl -= num_bytes;
+			qg->excl_cmpr -= num_bytes;
+			dirty = true;
+		}
+
+		/*
+		 * If we didn't reference all the roots before but now we do we
+		 * have an exclusive reference to this range.
+		 */
+		if ((!old_roots || (old_roots && cur_old_count != old_roots))
+		    && cur_new_count == new_roots) {
+			qg->excl += num_bytes;
+			qg->excl_cmpr += num_bytes;
+			dirty = true;
+		}
+
+		if (dirty)
+			qgroup_dirty(fs_info, qg);
+	}
+	return 0;
+}
+
+/*
+ * If we removed a data extent and there were other references for that bytenr
+ * then we need to lookup all referenced roots to make sure we still don't
+ * reference this bytenr.  If we do then we can just discard this operation.
+ */
+static int check_existing_refs(struct btrfs_trans_handle *trans,
+			       struct btrfs_fs_info *fs_info,
+			       struct btrfs_qgroup_operation *oper)
+{
+	struct ulist *roots = NULL;
+	struct ulist_node *unode;
+	struct ulist_iterator uiter;
+	int ret = 0;
+
+	ret = btrfs_find_all_roots(trans, fs_info, oper->bytenr,
+				   oper->elem.seq, &roots);
+	if (ret < 0)
+		return ret;
+	ret = 0;
+
+	ULIST_ITER_INIT(&uiter);
+	while ((unode = ulist_next(roots, &uiter))) {
+		if (unode->val == oper->ref_root) {
+			ret = 1;
+			break;
+		}
+	}
+	ulist_free(roots);
+	btrfs_put_tree_mod_seq(fs_info, &oper->elem);
+
+	return ret;
+}
+
+/*
+ * If we share a reference across multiple roots then we may need to adjust
+ * various qgroups referenced and exclusive counters.  The basic premise is this
+ *
+ * 1) We have seq to represent a 0 count.  Instead of looping through all of the
+ * qgroups and resetting their refcount to 0 we just constantly bump this
+ * sequence number to act as the base reference count.  This means that if
+ * anybody is equal to or below this sequence they were never referenced.  We
+ * jack this sequence up by the number of roots we found each time in order to
+ * make sure we don't have any overlap.
+ *
+ * 2) We first search all the roots that reference the area _except_ the root
+ * we're acting on currently.  This makes up the old_refcnt of all the qgroups
+ * before.
+ *
+ * 3) We walk all of the qgroups referenced by the root we are currently acting
+ * on, and will either adjust old_refcnt in the case of a removal or the
+ * new_refcnt in the case of an addition.
+ *
+ * 4) Finally we walk all the qgroups that are referenced by this range
+ * including the root we are acting on currently.  We will adjust the counters
+ * based on the number of roots we had and will have after this operation.
+ *
+ * Take this example as an illustration
+ *
+ *			[qgroup 1/0]
+ *		     /         |          \
+ *		[qg 0/0]   [qg 0/1]	[qg 0/2]
+ *		   \          |            /
+ *		  [	   extent	    ]
+ *
+ * Say we are adding a reference that is covered by qg 0/0.  The first step
+ * would give a refcnt of 1 to qg 0/1 and 0/2 and a refcnt of 2 to qg 1/0 with
+ * old_roots being 2.  Because it is adding new_roots will be 1.  We then go
+ * through qg 0/0 which will get the new_refcnt set to 1 and add 1 to qg 1/0's
+ * new_refcnt, bringing it to 3.  We then walk through all of the qgroups, we
+ * notice that the old refcnt for qg 0/0 < the new refcnt, so we added a
+ * reference and thus must add the size to the referenced bytes.  Everything
+ * else is the same so nothing else changes.
+ */
+static int qgroup_shared_accounting(struct btrfs_trans_handle *trans,
+				    struct btrfs_fs_info *fs_info,
+				    struct btrfs_qgroup_operation *oper)
+{
+	struct ulist *roots = NULL;
+	struct ulist *qgroups, *tmp;
+	struct btrfs_qgroup *qgroup;
+	struct seq_list elem = {};
+	u64 seq;
+	int old_roots = 0;
+	int new_roots = 0;
+	int ret = 0;
+
+	if (oper->elem.seq) {
+		ret = check_existing_refs(trans, fs_info, oper);
+		if (ret < 0)
+			return ret;
+		if (ret)
+			return 0;
+	}
+
+	qgroups = ulist_alloc(GFP_NOFS);
+	if (!qgroups)
+		return -ENOMEM;
+
+	tmp = ulist_alloc(GFP_NOFS);
+	if (!tmp) {
+		ulist_free(qgroups);
+		return -ENOMEM;
+	}
+
+	btrfs_get_tree_mod_seq(fs_info, &elem);
+	ret = btrfs_find_all_roots(trans, fs_info, oper->bytenr, elem.seq,
+				   &roots);
+	btrfs_put_tree_mod_seq(fs_info, &elem);
+	if (ret < 0) {
+		ulist_free(qgroups);
+		ulist_free(tmp);
+		return ret;
+	}
+	spin_lock(&fs_info->qgroup_lock);
+	qgroup = find_qgroup_rb(fs_info, oper->ref_root);
+	if (!qgroup)
+		goto out;
+	seq = fs_info->qgroup_seq;
+
+	/*
+	 * So roots is the list of all the roots currently pointing at the
+	 * bytenr, including the ref we are adding if we are adding, or not if
+	 * we are removing a ref.  So we pass in the ref_root to skip that root
+	 * in our calculations.  We set old_refnct and new_refcnt cause who the
+	 * hell knows what everything looked like before, and it doesn't matter
+	 * except...
+	 */
+	ret = qgroup_calc_old_refcnt(fs_info, oper->ref_root, tmp, roots, qgroups,
+				     seq, &old_roots, 0);
+	if (ret < 0)
+		goto out;
+
+	/*
+	 * Now adjust the refcounts of the qgroups that care about this
+	 * reference, either the old_count in the case of removal or new_count
+	 * in the case of an addition.
+	 */
+	ret = qgroup_calc_new_refcnt(fs_info, oper, qgroup, tmp, qgroups,
+				     seq);
+	if (ret < 0)
+		goto out;
+
+	/*
+	 * ...in the case of removals.  If we had a removal before we got around
+	 * to processing this operation then we need to find that guy and count
+	 * his references as if they really existed so we don't end up screwing
+	 * up the exclusive counts.  Then whenever we go to process the delete
+	 * everything will be grand and we can account for whatever exclusive
+	 * changes need to be made there.  We also have to pass in old_roots so
+	 * we have an accurate count of the roots as it pertains to this
+	 * operations view of the world.
+	 */
+	ret = qgroup_account_deleted_refs(fs_info, oper, tmp, qgroups, seq,
+					  &old_roots);
+	if (ret < 0)
+		goto out;
+
+	/*
+	 * We are adding our root, need to adjust up the number of roots,
+	 * otherwise old_roots is the number of roots we want.
+	 */
+	if (oper->type == BTRFS_QGROUP_OPER_ADD_SHARED) {
+		new_roots = old_roots + 1;
+	} else {
+		new_roots = old_roots;
+		old_roots++;
+	}
+	fs_info->qgroup_seq += old_roots + 1;
+
+
+	/*
+	 * And now the magic happens, bless Arne for having a pretty elegant
+	 * solution for this.
+	 */
+	qgroup_adjust_counters(fs_info, oper->ref_root, oper->num_bytes,
+			       qgroups, seq, old_roots, new_roots, 0);
+out:
+	spin_unlock(&fs_info->qgroup_lock);
+	ulist_free(qgroups);
+	ulist_free(roots);
+	ulist_free(tmp);
+	return ret;
+}
+
+/*
+ * btrfs_qgroup_account_ref is called for every ref that is added to or deleted
+ * from the fs. First, all roots referencing the extent are searched, and
+ * then the space is accounted accordingly to the different roots. The
+ * accounting algorithm works in 3 steps documented inline.
+ */
+static int btrfs_qgroup_account(struct btrfs_trans_handle *trans,
+				struct btrfs_fs_info *fs_info,
+				struct btrfs_qgroup_operation *oper)
+{
+	int ret = 0;
+
+	if (!fs_info->quota_enabled)
+		return 0;
+
+	BUG_ON(!fs_info->quota_root);
+
+	mutex_lock(&fs_info->qgroup_rescan_lock);
+	if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
+		if (fs_info->qgroup_rescan_progress.objectid <= oper->bytenr) {
+			mutex_unlock(&fs_info->qgroup_rescan_lock);
+			return 0;
+		}
+	}
+	mutex_unlock(&fs_info->qgroup_rescan_lock);
+
+	ASSERT(is_fstree(oper->ref_root));
+
+	switch (oper->type) {
+	case BTRFS_QGROUP_OPER_ADD_EXCL:
+	case BTRFS_QGROUP_OPER_SUB_EXCL:
+		ret = qgroup_excl_accounting(fs_info, oper);
+		break;
+	case BTRFS_QGROUP_OPER_ADD_SHARED:
+	case BTRFS_QGROUP_OPER_SUB_SHARED:
+		ret = qgroup_shared_accounting(trans, fs_info, oper);
+		break;
+	default:
+		ASSERT(0);
+	}
+	return ret;
+}
+
+/*
+ * Needs to be called everytime we run delayed refs, even if there is an error
+ * in order to cleanup outstanding operations.
+ */
+int btrfs_delayed_qgroup_accounting(struct btrfs_trans_handle *trans,
+				    struct btrfs_fs_info *fs_info)
+{
+	struct btrfs_qgroup_operation *oper;
+	int ret = 0;
+
+	while (!list_empty(&trans->qgroup_ref_list)) {
+		oper = list_first_entry(&trans->qgroup_ref_list,
+					struct btrfs_qgroup_operation, list);
+		list_del_init(&oper->list);
+		if (!ret || !trans->aborted)
+			ret = btrfs_qgroup_account(trans, fs_info, oper);
+		spin_lock(&fs_info->qgroup_op_lock);
+		rb_erase(&oper->n, &fs_info->qgroup_op_tree);
+		spin_unlock(&fs_info->qgroup_op_lock);
+		btrfs_put_tree_mod_seq(fs_info, &oper->elem);
+		kfree(oper);
+	}
+	return ret;
+}
+
+/*
+ * called from commit_transaction. Writes all changed qgroups to disk.
+ */
+int btrfs_run_qgroups(struct btrfs_trans_handle *trans,
+		      struct btrfs_fs_info *fs_info)
+{
+	struct btrfs_root *quota_root = fs_info->quota_root;
+	int ret = 0;
+	int start_rescan_worker = 0;
+
+	if (!quota_root)
+		goto out;
+
+	if (!fs_info->quota_enabled && fs_info->pending_quota_state)
+		start_rescan_worker = 1;
+
+	fs_info->quota_enabled = fs_info->pending_quota_state;
+
+	spin_lock(&fs_info->qgroup_lock);
+	while (!list_empty(&fs_info->dirty_qgroups)) {
+		struct btrfs_qgroup *qgroup;
+		qgroup = list_first_entry(&fs_info->dirty_qgroups,
+					  struct btrfs_qgroup, dirty);
+		list_del_init(&qgroup->dirty);
+		spin_unlock(&fs_info->qgroup_lock);
+		ret = update_qgroup_info_item(trans, quota_root, qgroup);
+		if (ret)
+			fs_info->qgroup_flags |=
+					BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT;
+		spin_lock(&fs_info->qgroup_lock);
+	}
+	if (fs_info->quota_enabled)
+		fs_info->qgroup_flags |= BTRFS_QGROUP_STATUS_FLAG_ON;
+	else
+		fs_info->qgroup_flags &= ~BTRFS_QGROUP_STATUS_FLAG_ON;
+	spin_unlock(&fs_info->qgroup_lock);
+
+	ret = update_qgroup_status_item(trans, fs_info, quota_root);
+	if (ret)
+		fs_info->qgroup_flags |= BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT;
+
+	if (!ret && start_rescan_worker) {
+		ret = qgroup_rescan_init(fs_info, 0, 1);
+		if (!ret) {
+			qgroup_rescan_zero_tracking(fs_info);
+			btrfs_queue_work(fs_info->qgroup_rescan_workers,
+					 &fs_info->qgroup_rescan_work);
+		}
+		ret = 0;
+	}
+
+out:
+
+	return ret;
+}
+
+/*
+ * copy the acounting information between qgroups. This is necessary when a
+ * snapshot or a subvolume is created
+ */
+int btrfs_qgroup_inherit(struct btrfs_trans_handle *trans,
+			 struct btrfs_fs_info *fs_info, u64 srcid, u64 objectid,
+			 struct btrfs_qgroup_inherit *inherit)
+{
+	int ret = 0;
+	int i;
+	u64 *i_qgroups;
+	struct btrfs_root *quota_root = fs_info->quota_root;
+	struct btrfs_qgroup *srcgroup;
+	struct btrfs_qgroup *dstgroup;
+	u32 level_size = 0;
+	u64 nums;
+
+	mutex_lock(&fs_info->qgroup_ioctl_lock);
+	if (!fs_info->quota_enabled)
+		goto out;
+
+	if (!quota_root) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	if (inherit) {
+		i_qgroups = (u64 *)(inherit + 1);
+		nums = inherit->num_qgroups + 2 * inherit->num_ref_copies +
+		       2 * inherit->num_excl_copies;
+		for (i = 0; i < nums; ++i) {
+			srcgroup = find_qgroup_rb(fs_info, *i_qgroups);
+			if (!srcgroup) {
+				ret = -EINVAL;
+				goto out;
+			}
+			++i_qgroups;
+		}
+	}
+
+	/*
+	 * create a tracking group for the subvol itself
+	 */
+	ret = add_qgroup_item(trans, quota_root, objectid);
+	if (ret)
+		goto out;
+
+	if (inherit && inherit->flags & BTRFS_QGROUP_INHERIT_SET_LIMITS) {
+		ret = update_qgroup_limit_item(trans, quota_root, objectid,
+					       inherit->lim.flags,
+					       inherit->lim.max_rfer,
+					       inherit->lim.max_excl,
+					       inherit->lim.rsv_rfer,
+					       inherit->lim.rsv_excl);
+		if (ret)
+			goto out;
+	}
+
+	if (srcid) {
+		struct btrfs_root *srcroot;
+		struct btrfs_key srckey;
+		int srcroot_level;
+
+		srckey.objectid = srcid;
+		srckey.type = BTRFS_ROOT_ITEM_KEY;
+		srckey.offset = (u64)-1;
+		srcroot = btrfs_read_fs_root_no_name(fs_info, &srckey);
+		if (IS_ERR(srcroot)) {
+			ret = PTR_ERR(srcroot);
+			goto out;
+		}
+
+		rcu_read_lock();
+		srcroot_level = btrfs_header_level(srcroot->node);
+		level_size = btrfs_level_size(srcroot, srcroot_level);
+		rcu_read_unlock();
+	}
+
+	/*
+	 * add qgroup to all inherited groups
+	 */
+	if (inherit) {
+		i_qgroups = (u64 *)(inherit + 1);
+		for (i = 0; i < inherit->num_qgroups; ++i) {
+			ret = add_qgroup_relation_item(trans, quota_root,
+						       objectid, *i_qgroups);
+			if (ret)
+				goto out;
+			ret = add_qgroup_relation_item(trans, quota_root,
+						       *i_qgroups, objectid);
+			if (ret)
+				goto out;
+			++i_qgroups;
+		}
+	}
+
+
+	spin_lock(&fs_info->qgroup_lock);
+
+	dstgroup = add_qgroup_rb(fs_info, objectid);
+	if (IS_ERR(dstgroup)) {
+		ret = PTR_ERR(dstgroup);
+		goto unlock;
+	}
+
+	if (srcid) {
+		srcgroup = find_qgroup_rb(fs_info, srcid);
+		if (!srcgroup)
+			goto unlock;
+
+		/*
+		 * We call inherit after we clone the root in order to make sure
+		 * our counts don't go crazy, so at this point the only
+		 * difference between the two roots should be the root node.
+		 */
+		dstgroup->rfer = srcgroup->rfer;
+		dstgroup->rfer_cmpr = srcgroup->rfer_cmpr;
+		dstgroup->excl = level_size;
+		dstgroup->excl_cmpr = level_size;
+		srcgroup->excl = level_size;
+		srcgroup->excl_cmpr = level_size;
+		qgroup_dirty(fs_info, dstgroup);
+		qgroup_dirty(fs_info, srcgroup);
+	}
+
+	if (!inherit)
+		goto unlock;
+
+	i_qgroups = (u64 *)(inherit + 1);
+	for (i = 0; i < inherit->num_qgroups; ++i) {
+		ret = add_relation_rb(quota_root->fs_info, objectid,
+				      *i_qgroups);
+		if (ret)
+			goto unlock;
+		++i_qgroups;
+	}
+
+	for (i = 0; i <  inherit->num_ref_copies; ++i) {
+		struct btrfs_qgroup *src;
+		struct btrfs_qgroup *dst;
+
+		src = find_qgroup_rb(fs_info, i_qgroups[0]);
+		dst = find_qgroup_rb(fs_info, i_qgroups[1]);
+
+		if (!src || !dst) {
+			ret = -EINVAL;
+			goto unlock;
+		}
+
+		dst->rfer = src->rfer - level_size;
+		dst->rfer_cmpr = src->rfer_cmpr - level_size;
+		i_qgroups += 2;
+	}
+	for (i = 0; i <  inherit->num_excl_copies; ++i) {
+		struct btrfs_qgroup *src;
+		struct btrfs_qgroup *dst;
+
+		src = find_qgroup_rb(fs_info, i_qgroups[0]);
+		dst = find_qgroup_rb(fs_info, i_qgroups[1]);
+
+		if (!src || !dst) {
+			ret = -EINVAL;
+			goto unlock;
+		}
+
+		dst->excl = src->excl + level_size;
+		dst->excl_cmpr = src->excl_cmpr + level_size;
+		i_qgroups += 2;
+	}
+
+unlock:
+	spin_unlock(&fs_info->qgroup_lock);
+out:
+	mutex_unlock(&fs_info->qgroup_ioctl_lock);
+	return ret;
+}
+
+/*
+ * reserve some space for a qgroup and all its parents. The reservation takes
+ * place with start_transaction or dealloc_reserve, similar to ENOSPC
+ * accounting. If not enough space is available, EDQUOT is returned.
+ * We assume that the requested space is new for all qgroups.
+ */
+int btrfs_qgroup_reserve(struct btrfs_root *root, u64 num_bytes)
+{
+	struct btrfs_root *quota_root;
+	struct btrfs_qgroup *qgroup;
+	struct btrfs_fs_info *fs_info = root->fs_info;
+	u64 ref_root = root->root_key.objectid;
+	int ret = 0;
+	struct ulist_node *unode;
+	struct ulist_iterator uiter;
+
+	if (!is_fstree(ref_root))
+		return 0;
+
+	if (num_bytes == 0)
+		return 0;
+
+	spin_lock(&fs_info->qgroup_lock);
+	quota_root = fs_info->quota_root;
+	if (!quota_root)
+		goto out;
+
+	qgroup = find_qgroup_rb(fs_info, ref_root);
+	if (!qgroup)
+		goto out;
+
+	/*
+	 * in a first step, we check all affected qgroups if any limits would
+	 * be exceeded
+	 */
+	ulist_reinit(fs_info->qgroup_ulist);
+	ret = ulist_add(fs_info->qgroup_ulist, qgroup->qgroupid,
+			(uintptr_t)qgroup, GFP_ATOMIC);
+	if (ret < 0)
+		goto out;
+	ULIST_ITER_INIT(&uiter);
+	while ((unode = ulist_next(fs_info->qgroup_ulist, &uiter))) {
+		struct btrfs_qgroup *qg;
+		struct btrfs_qgroup_list *glist;
+
+		qg = u64_to_ptr(unode->aux);
+
+		if ((qg->lim_flags & BTRFS_QGROUP_LIMIT_MAX_RFER) &&
+		    qg->reserved + (s64)qg->rfer + num_bytes >
+		    qg->max_rfer) {
+			ret = -EDQUOT;
+			goto out;
+		}
+
+		if ((qg->lim_flags & BTRFS_QGROUP_LIMIT_MAX_EXCL) &&
+		    qg->reserved + (s64)qg->excl + num_bytes >
+		    qg->max_excl) {
+			ret = -EDQUOT;
+			goto out;
+		}
+
+		list_for_each_entry(glist, &qg->groups, next_group) {
+			ret = ulist_add(fs_info->qgroup_ulist,
+					glist->group->qgroupid,
+					(uintptr_t)glist->group, GFP_ATOMIC);
+			if (ret < 0)
+				goto out;
+		}
+	}
+	ret = 0;
+	/*
+	 * no limits exceeded, now record the reservation into all qgroups
+	 */
+	ULIST_ITER_INIT(&uiter);
+	while ((unode = ulist_next(fs_info->qgroup_ulist, &uiter))) {
+		struct btrfs_qgroup *qg;
+
+		qg = u64_to_ptr(unode->aux);
+
+		qg->reserved += num_bytes;
+	}
+
+out:
+	spin_unlock(&fs_info->qgroup_lock);
+	return ret;
+}
+
+void btrfs_qgroup_free(struct btrfs_root *root, u64 num_bytes)
+{
+	struct btrfs_root *quota_root;
+	struct btrfs_qgroup *qgroup;
+	struct btrfs_fs_info *fs_info = root->fs_info;
+	struct ulist_node *unode;
+	struct ulist_iterator uiter;
+	u64 ref_root = root->root_key.objectid;
+	int ret = 0;
+
+	if (!is_fstree(ref_root))
+		return;
+
+	if (num_bytes == 0)
+		return;
+
+	spin_lock(&fs_info->qgroup_lock);
+
+	quota_root = fs_info->quota_root;
+	if (!quota_root)
+		goto out;
+
+	qgroup = find_qgroup_rb(fs_info, ref_root);
+	if (!qgroup)
+		goto out;
+
+	ulist_reinit(fs_info->qgroup_ulist);
+	ret = ulist_add(fs_info->qgroup_ulist, qgroup->qgroupid,
+			(uintptr_t)qgroup, GFP_ATOMIC);
+	if (ret < 0)
+		goto out;
+	ULIST_ITER_INIT(&uiter);
+	while ((unode = ulist_next(fs_info->qgroup_ulist, &uiter))) {
+		struct btrfs_qgroup *qg;
+		struct btrfs_qgroup_list *glist;
+
+		qg = u64_to_ptr(unode->aux);
+
+		qg->reserved -= num_bytes;
+
+		list_for_each_entry(glist, &qg->groups, next_group) {
+			ret = ulist_add(fs_info->qgroup_ulist,
+					glist->group->qgroupid,
+					(uintptr_t)glist->group, GFP_ATOMIC);
+			if (ret < 0)
+				goto out;
+		}
+	}
+
+out:
+	spin_unlock(&fs_info->qgroup_lock);
+}
+
+void assert_qgroups_uptodate(struct btrfs_trans_handle *trans)
+{
+	if (list_empty(&trans->qgroup_ref_list) && !trans->delayed_ref_elem.seq)
+		return;
+	btrfs_err(trans->root->fs_info,
+		"qgroups not uptodate in trans handle %p:  list is%s empty, "
+		"seq is %#x.%x",
+		trans, list_empty(&trans->qgroup_ref_list) ? "" : " not",
+		(u32)(trans->delayed_ref_elem.seq >> 32),
+		(u32)trans->delayed_ref_elem.seq);
+	BUG();
+}
+
+/*
+ * returns < 0 on error, 0 when more leafs are to be scanned.
+ * returns 1 when done, 2 when done and FLAG_INCONSISTENT was cleared.
+ */
+static int
+qgroup_rescan_leaf(struct btrfs_fs_info *fs_info, struct btrfs_path *path,
+		   struct btrfs_trans_handle *trans, struct ulist *qgroups,
+		   struct ulist *tmp, struct extent_buffer *scratch_leaf)
+{
+	struct btrfs_key found;
+	struct ulist *roots = NULL;
+	struct seq_list tree_mod_seq_elem = {};
+	u64 num_bytes;
+	u64 seq;
+	int new_roots;
+	int slot;
+	int ret;
+
+	path->leave_spinning = 1;
+	mutex_lock(&fs_info->qgroup_rescan_lock);
+	ret = btrfs_search_slot_for_read(fs_info->extent_root,
+					 &fs_info->qgroup_rescan_progress,
+					 path, 1, 0);
+
+	pr_debug("current progress key (%llu %u %llu), search_slot ret %d\n",
+		 fs_info->qgroup_rescan_progress.objectid,
+		 fs_info->qgroup_rescan_progress.type,
+		 fs_info->qgroup_rescan_progress.offset, ret);
+
+	if (ret) {
+		/*
+		 * The rescan is about to end, we will not be scanning any
+		 * further blocks. We cannot unset the RESCAN flag here, because
+		 * we want to commit the transaction if everything went well.
+		 * To make the live accounting work in this phase, we set our
+		 * scan progress pointer such that every real extent objectid
+		 * will be smaller.
+		 */
+		fs_info->qgroup_rescan_progress.objectid = (u64)-1;
+		btrfs_release_path(path);
+		mutex_unlock(&fs_info->qgroup_rescan_lock);
+		return ret;
+	}
+
+	btrfs_item_key_to_cpu(path->nodes[0], &found,
+			      btrfs_header_nritems(path->nodes[0]) - 1);
+	fs_info->qgroup_rescan_progress.objectid = found.objectid + 1;
+
+	btrfs_get_tree_mod_seq(fs_info, &tree_mod_seq_elem);
+	memcpy(scratch_leaf, path->nodes[0], sizeof(*scratch_leaf));
+	slot = path->slots[0];
+	btrfs_release_path(path);
+	mutex_unlock(&fs_info->qgroup_rescan_lock);
+
+	for (; slot < btrfs_header_nritems(scratch_leaf); ++slot) {
+		btrfs_item_key_to_cpu(scratch_leaf, &found, slot);
+		if (found.type != BTRFS_EXTENT_ITEM_KEY &&
+		    found.type != BTRFS_METADATA_ITEM_KEY)
+			continue;
+		if (found.type == BTRFS_METADATA_ITEM_KEY)
+			num_bytes = fs_info->extent_root->leafsize;
+		else
+			num_bytes = found.offset;
+
+		ulist_reinit(qgroups);
+		ret = btrfs_find_all_roots(NULL, fs_info, found.objectid, 0,
+					   &roots);
+		if (ret < 0)
+			goto out;
+		spin_lock(&fs_info->qgroup_lock);
+		seq = fs_info->qgroup_seq;
+		fs_info->qgroup_seq += roots->nnodes + 1; /* max refcnt */
+
+		new_roots = 0;
+		ret = qgroup_calc_old_refcnt(fs_info, 0, tmp, roots, qgroups,
+					     seq, &new_roots, 1);
+		if (ret < 0) {
+			spin_unlock(&fs_info->qgroup_lock);
+			ulist_free(roots);
+			goto out;
+		}
+
+		ret = qgroup_adjust_counters(fs_info, 0, num_bytes, qgroups,
+					     seq, 0, new_roots, 1);
+		if (ret < 0) {
+			spin_unlock(&fs_info->qgroup_lock);
+			ulist_free(roots);
+			goto out;
+		}
+		spin_unlock(&fs_info->qgroup_lock);
+		ulist_free(roots);
+	}
+out:
+	btrfs_put_tree_mod_seq(fs_info, &tree_mod_seq_elem);
+
+	return ret;
+}
+
+static void btrfs_qgroup_rescan_worker(struct btrfs_work *work)
+{
+	struct btrfs_fs_info *fs_info = container_of(work, struct btrfs_fs_info,
+						     qgroup_rescan_work);
+	struct btrfs_path *path;
+	struct btrfs_trans_handle *trans = NULL;
+	struct ulist *tmp = NULL, *qgroups = NULL;
+	struct extent_buffer *scratch_leaf = NULL;
+	int err = -ENOMEM;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		goto out;
+	qgroups = ulist_alloc(GFP_NOFS);
+	if (!qgroups)
+		goto out;
+	tmp = ulist_alloc(GFP_NOFS);
+	if (!tmp)
+		goto out;
+	scratch_leaf = kmalloc(sizeof(*scratch_leaf), GFP_NOFS);
+	if (!scratch_leaf)
+		goto out;
+
+	err = 0;
+	while (!err) {
+		trans = btrfs_start_transaction(fs_info->fs_root, 0);
+		if (IS_ERR(trans)) {
+			err = PTR_ERR(trans);
+			break;
+		}
+		if (!fs_info->quota_enabled) {
+			err = -EINTR;
+		} else {
+			err = qgroup_rescan_leaf(fs_info, path, trans,
+						 qgroups, tmp, scratch_leaf);
+		}
+		if (err > 0)
+			btrfs_commit_transaction(trans, fs_info->fs_root);
+		else
+			btrfs_end_transaction(trans, fs_info->fs_root);
+	}
+
+out:
+	kfree(scratch_leaf);
+	ulist_free(qgroups);
+	ulist_free(tmp);
+	btrfs_free_path(path);
+
+	mutex_lock(&fs_info->qgroup_rescan_lock);
+	fs_info->qgroup_flags &= ~BTRFS_QGROUP_STATUS_FLAG_RESCAN;
+
+	if (err == 2 &&
+	    fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT) {
+		fs_info->qgroup_flags &= ~BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT;
+	} else if (err < 0) {
+		fs_info->qgroup_flags |= BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT;
+	}
+	mutex_unlock(&fs_info->qgroup_rescan_lock);
+
+	if (err >= 0) {
+		btrfs_info(fs_info, "qgroup scan completed%s",
+			err == 2 ? " (inconsistency flag cleared)" : "");
+	} else {
+		btrfs_err(fs_info, "qgroup scan failed with %d", err);
+	}
+
+	complete_all(&fs_info->qgroup_rescan_completion);
+}
+
+/*
+ * Checks that (a) no rescan is running and (b) quota is enabled. Allocates all
+ * memory required for the rescan context.
+ */
+static int
+qgroup_rescan_init(struct btrfs_fs_info *fs_info, u64 progress_objectid,
+		   int init_flags)
+{
+	int ret = 0;
+
+	if (!init_flags &&
+	    (!(fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) ||
+	     !(fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_ON))) {
+		ret = -EINVAL;
+		goto err;
+	}
+
+	mutex_lock(&fs_info->qgroup_rescan_lock);
+	spin_lock(&fs_info->qgroup_lock);
+
+	if (init_flags) {
+		if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN)
+			ret = -EINPROGRESS;
+		else if (!(fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_ON))
+			ret = -EINVAL;
+
+		if (ret) {
+			spin_unlock(&fs_info->qgroup_lock);
+			mutex_unlock(&fs_info->qgroup_rescan_lock);
+			goto err;
+		}
+
+		fs_info->qgroup_flags |= BTRFS_QGROUP_STATUS_FLAG_RESCAN;
+	}
+
+	memset(&fs_info->qgroup_rescan_progress, 0,
+		sizeof(fs_info->qgroup_rescan_progress));
+	fs_info->qgroup_rescan_progress.objectid = progress_objectid;
+
+	spin_unlock(&fs_info->qgroup_lock);
+	mutex_unlock(&fs_info->qgroup_rescan_lock);
+
+	init_completion(&fs_info->qgroup_rescan_completion);
+
+	memset(&fs_info->qgroup_rescan_work, 0,
+	       sizeof(fs_info->qgroup_rescan_work));
+	btrfs_init_work(&fs_info->qgroup_rescan_work,
+			btrfs_qgroup_rescan_worker, NULL, NULL);
+
+	if (ret) {
+err:
+		btrfs_info(fs_info, "qgroup_rescan_init failed with %d", ret);
+		return ret;
+	}
+
+	return 0;
+}
+
+static void
+qgroup_rescan_zero_tracking(struct btrfs_fs_info *fs_info)
+{
+	struct rb_node *n;
+	struct btrfs_qgroup *qgroup;
+
+	spin_lock(&fs_info->qgroup_lock);
+	/* clear all current qgroup tracking information */
+	for (n = rb_first(&fs_info->qgroup_tree); n; n = rb_next(n)) {
+		qgroup = rb_entry(n, struct btrfs_qgroup, node);
+		qgroup->rfer = 0;
+		qgroup->rfer_cmpr = 0;
+		qgroup->excl = 0;
+		qgroup->excl_cmpr = 0;
+	}
+	spin_unlock(&fs_info->qgroup_lock);
+}
+
+int
+btrfs_qgroup_rescan(struct btrfs_fs_info *fs_info)
+{
+	int ret = 0;
+	struct btrfs_trans_handle *trans;
+
+	ret = qgroup_rescan_init(fs_info, 0, 1);
+	if (ret)
+		return ret;
+
+	/*
+	 * We have set the rescan_progress to 0, which means no more
+	 * delayed refs will be accounted by btrfs_qgroup_account_ref.
+	 * However, btrfs_qgroup_account_ref may be right after its call
+	 * to btrfs_find_all_roots, in which case it would still do the
+	 * accounting.
+	 * To solve this, we're committing the transaction, which will
+	 * ensure we run all delayed refs and only after that, we are
+	 * going to clear all tracking information for a clean start.
+	 */
+
+	trans = btrfs_join_transaction(fs_info->fs_root);
+	if (IS_ERR(trans)) {
+		fs_info->qgroup_flags &= ~BTRFS_QGROUP_STATUS_FLAG_RESCAN;
+		return PTR_ERR(trans);
+	}
+	ret = btrfs_commit_transaction(trans, fs_info->fs_root);
+	if (ret) {
+		fs_info->qgroup_flags &= ~BTRFS_QGROUP_STATUS_FLAG_RESCAN;
+		return ret;
+	}
+
+	qgroup_rescan_zero_tracking(fs_info);
+
+	btrfs_queue_work(fs_info->qgroup_rescan_workers,
+			 &fs_info->qgroup_rescan_work);
+
+	return 0;
+}
+
+int btrfs_qgroup_wait_for_completion(struct btrfs_fs_info *fs_info)
+{
+	int running;
+	int ret = 0;
+
+	mutex_lock(&fs_info->qgroup_rescan_lock);
+	spin_lock(&fs_info->qgroup_lock);
+	running = fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN;
+	spin_unlock(&fs_info->qgroup_lock);
+	mutex_unlock(&fs_info->qgroup_rescan_lock);
+
+	if (running)
+		ret = wait_for_completion_interruptible(
+					&fs_info->qgroup_rescan_completion);
+
+	return ret;
+}
+
+/*
+ * this is only called from open_ctree where we're still single threaded, thus
+ * locking is omitted here.
+ */
+void
+btrfs_qgroup_rescan_resume(struct btrfs_fs_info *fs_info)
+{
+	if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN)
+		btrfs_queue_work(fs_info->qgroup_rescan_workers,
+				 &fs_info->qgroup_rescan_work);
+}
diff --git a/fs/btrfs/qgroup.h b/fs/btrfs/qgroup.h
new file mode 100644
index 00000000000..5952ff1fbd7
--- /dev/null
+++ b/fs/btrfs/qgroup.h
@@ -0,0 +1,107 @@
+/*
+ * Copyright (C) 2014 Facebook.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef __BTRFS_QGROUP__
+#define __BTRFS_QGROUP__
+
+/*
+ * A description of the operations, all of these operations only happen when we
+ * are adding the 1st reference for that subvolume in the case of adding space
+ * or on the last reference delete in the case of subtraction.  The only
+ * exception is the last one, which is added for confusion.
+ *
+ * BTRFS_QGROUP_OPER_ADD_EXCL: adding bytes where this subvolume is the only
+ * one pointing at the bytes we are adding.  This is called on the first
+ * allocation.
+ *
+ * BTRFS_QGROUP_OPER_ADD_SHARED: adding bytes where this bytenr is going to be
+ * shared between subvols.  This is called on the creation of a ref that already
+ * has refs from a different subvolume, so basically reflink.
+ *
+ * BTRFS_QGROUP_OPER_SUB_EXCL: removing bytes where this subvolume is the only
+ * one referencing the range.
+ *
+ * BTRFS_QGROUP_OPER_SUB_SHARED: removing bytes where this subvolume shares with
+ * refs with other subvolumes.
+ */
+enum btrfs_qgroup_operation_type {
+	BTRFS_QGROUP_OPER_ADD_EXCL,
+	BTRFS_QGROUP_OPER_ADD_SHARED,
+	BTRFS_QGROUP_OPER_SUB_EXCL,
+	BTRFS_QGROUP_OPER_SUB_SHARED,
+};
+
+struct btrfs_qgroup_operation {
+	u64 ref_root;
+	u64 bytenr;
+	u64 num_bytes;
+	u64 seq;
+	enum btrfs_qgroup_operation_type type;
+	struct seq_list elem;
+	struct rb_node n;
+	struct list_head list;
+};
+
+int btrfs_quota_enable(struct btrfs_trans_handle *trans,
+		       struct btrfs_fs_info *fs_info);
+int btrfs_quota_disable(struct btrfs_trans_handle *trans,
+			struct btrfs_fs_info *fs_info);
+int btrfs_qgroup_rescan(struct btrfs_fs_info *fs_info);
+void btrfs_qgroup_rescan_resume(struct btrfs_fs_info *fs_info);
+int btrfs_qgroup_wait_for_completion(struct btrfs_fs_info *fs_info);
+int btrfs_add_qgroup_relation(struct btrfs_trans_handle *trans,
+			      struct btrfs_fs_info *fs_info, u64 src, u64 dst);
+int btrfs_del_qgroup_relation(struct btrfs_trans_handle *trans,
+			      struct btrfs_fs_info *fs_info, u64 src, u64 dst);
+int btrfs_create_qgroup(struct btrfs_trans_handle *trans,
+			struct btrfs_fs_info *fs_info, u64 qgroupid,
+			char *name);
+int btrfs_remove_qgroup(struct btrfs_trans_handle *trans,
+			      struct btrfs_fs_info *fs_info, u64 qgroupid);
+int btrfs_limit_qgroup(struct btrfs_trans_handle *trans,
+		       struct btrfs_fs_info *fs_info, u64 qgroupid,
+		       struct btrfs_qgroup_limit *limit);
+int btrfs_read_qgroup_config(struct btrfs_fs_info *fs_info);
+void btrfs_free_qgroup_config(struct btrfs_fs_info *fs_info);
+struct btrfs_delayed_extent_op;
+int btrfs_qgroup_record_ref(struct btrfs_trans_handle *trans,
+			    struct btrfs_fs_info *fs_info, u64 ref_root,
+			    u64 bytenr, u64 num_bytes,
+			    enum btrfs_qgroup_operation_type type,
+			    int mod_seq);
+int btrfs_delayed_qgroup_accounting(struct btrfs_trans_handle *trans,
+				    struct btrfs_fs_info *fs_info);
+void btrfs_remove_qgroup_operation(struct btrfs_trans_handle *trans,
+				   struct btrfs_fs_info *fs_info,
+				   struct btrfs_qgroup_operation *oper);
+int btrfs_run_qgroups(struct btrfs_trans_handle *trans,
+		      struct btrfs_fs_info *fs_info);
+int btrfs_qgroup_inherit(struct btrfs_trans_handle *trans,
+			 struct btrfs_fs_info *fs_info, u64 srcid, u64 objectid,
+			 struct btrfs_qgroup_inherit *inherit);
+int btrfs_qgroup_reserve(struct btrfs_root *root, u64 num_bytes);
+void btrfs_qgroup_free(struct btrfs_root *root, u64 num_bytes);
+
+void assert_qgroups_uptodate(struct btrfs_trans_handle *trans);
+
+#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
+int btrfs_verify_qgroup_counts(struct btrfs_fs_info *fs_info, u64 qgroupid,
+			       u64 rfer, u64 excl);
+#endif
+
+#endif /* __BTRFS_QGROUP__ */
diff --git a/fs/btrfs/raid56.c b/fs/btrfs/raid56.c
new file mode 100644
index 00000000000..4a88f073fdd
--- /dev/null
+++ b/fs/btrfs/raid56.c
@@ -0,0 +1,2097 @@
+/*
+ * Copyright (C) 2012 Fusion-io  All rights reserved.
+ * Copyright (C) 2012 Intel Corp. All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+#include <linux/sched.h>
+#include <linux/wait.h>
+#include <linux/bio.h>
+#include <linux/slab.h>
+#include <linux/buffer_head.h>
+#include <linux/blkdev.h>
+#include <linux/random.h>
+#include <linux/iocontext.h>
+#include <linux/capability.h>
+#include <linux/ratelimit.h>
+#include <linux/kthread.h>
+#include <linux/raid/pq.h>
+#include <linux/hash.h>
+#include <linux/list_sort.h>
+#include <linux/raid/xor.h>
+#include <linux/vmalloc.h>
+#include <asm/div64.h>
+#include "ctree.h"
+#include "extent_map.h"
+#include "disk-io.h"
+#include "transaction.h"
+#include "print-tree.h"
+#include "volumes.h"
+#include "raid56.h"
+#include "async-thread.h"
+#include "check-integrity.h"
+#include "rcu-string.h"
+
+/* set when additional merges to this rbio are not allowed */
+#define RBIO_RMW_LOCKED_BIT	1
+
+/*
+ * set when this rbio is sitting in the hash, but it is just a cache
+ * of past RMW
+ */
+#define RBIO_CACHE_BIT		2
+
+/*
+ * set when it is safe to trust the stripe_pages for caching
+ */
+#define RBIO_CACHE_READY_BIT	3
+
+
+#define RBIO_CACHE_SIZE 1024
+
+struct btrfs_raid_bio {
+	struct btrfs_fs_info *fs_info;
+	struct btrfs_bio *bbio;
+
+	/*
+	 * logical block numbers for the start of each stripe
+	 * The last one or two are p/q.  These are sorted,
+	 * so raid_map[0] is the start of our full stripe
+	 */
+	u64 *raid_map;
+
+	/* while we're doing rmw on a stripe
+	 * we put it into a hash table so we can
+	 * lock the stripe and merge more rbios
+	 * into it.
+	 */
+	struct list_head hash_list;
+
+	/*
+	 * LRU list for the stripe cache
+	 */
+	struct list_head stripe_cache;
+
+	/*
+	 * for scheduling work in the helper threads
+	 */
+	struct btrfs_work work;
+
+	/*
+	 * bio list and bio_list_lock are used
+	 * to add more bios into the stripe
+	 * in hopes of avoiding the full rmw
+	 */
+	struct bio_list bio_list;
+	spinlock_t bio_list_lock;
+
+	/* also protected by the bio_list_lock, the
+	 * plug list is used by the plugging code
+	 * to collect partial bios while plugged.  The
+	 * stripe locking code also uses it to hand off
+	 * the stripe lock to the next pending IO
+	 */
+	struct list_head plug_list;
+
+	/*
+	 * flags that tell us if it is safe to
+	 * merge with this bio
+	 */
+	unsigned long flags;
+
+	/* size of each individual stripe on disk */
+	int stripe_len;
+
+	/* number of data stripes (no p/q) */
+	int nr_data;
+
+	/*
+	 * set if we're doing a parity rebuild
+	 * for a read from higher up, which is handled
+	 * differently from a parity rebuild as part of
+	 * rmw
+	 */
+	int read_rebuild;
+
+	/* first bad stripe */
+	int faila;
+
+	/* second bad stripe (for raid6 use) */
+	int failb;
+
+	/*
+	 * number of pages needed to represent the full
+	 * stripe
+	 */
+	int nr_pages;
+
+	/*
+	 * size of all the bios in the bio_list.  This
+	 * helps us decide if the rbio maps to a full
+	 * stripe or not
+	 */
+	int bio_list_bytes;
+
+	atomic_t refs;
+
+	/*
+	 * these are two arrays of pointers.  We allocate the
+	 * rbio big enough to hold them both and setup their
+	 * locations when the rbio is allocated
+	 */
+
+	/* pointers to pages that we allocated for
+	 * reading/writing stripes directly from the disk (including P/Q)
+	 */
+	struct page **stripe_pages;
+
+	/*
+	 * pointers to the pages in the bio_list.  Stored
+	 * here for faster lookup
+	 */
+	struct page **bio_pages;
+};
+
+static int __raid56_parity_recover(struct btrfs_raid_bio *rbio);
+static noinline void finish_rmw(struct btrfs_raid_bio *rbio);
+static void rmw_work(struct btrfs_work *work);
+static void read_rebuild_work(struct btrfs_work *work);
+static void async_rmw_stripe(struct btrfs_raid_bio *rbio);
+static void async_read_rebuild(struct btrfs_raid_bio *rbio);
+static int fail_bio_stripe(struct btrfs_raid_bio *rbio, struct bio *bio);
+static int fail_rbio_index(struct btrfs_raid_bio *rbio, int failed);
+static void __free_raid_bio(struct btrfs_raid_bio *rbio);
+static void index_rbio_pages(struct btrfs_raid_bio *rbio);
+static int alloc_rbio_pages(struct btrfs_raid_bio *rbio);
+
+/*
+ * the stripe hash table is used for locking, and to collect
+ * bios in hopes of making a full stripe
+ */
+int btrfs_alloc_stripe_hash_table(struct btrfs_fs_info *info)
+{
+	struct btrfs_stripe_hash_table *table;
+	struct btrfs_stripe_hash_table *x;
+	struct btrfs_stripe_hash *cur;
+	struct btrfs_stripe_hash *h;
+	int num_entries = 1 << BTRFS_STRIPE_HASH_TABLE_BITS;
+	int i;
+	int table_size;
+
+	if (info->stripe_hash_table)
+		return 0;
+
+	/*
+	 * The table is large, starting with order 4 and can go as high as
+	 * order 7 in case lock debugging is turned on.
+	 *
+	 * Try harder to allocate and fallback to vmalloc to lower the chance
+	 * of a failing mount.
+	 */
+	table_size = sizeof(*table) + sizeof(*h) * num_entries;
+	table = kzalloc(table_size, GFP_KERNEL | __GFP_NOWARN | __GFP_REPEAT);
+	if (!table) {
+		table = vzalloc(table_size);
+		if (!table)
+			return -ENOMEM;
+	}
+
+	spin_lock_init(&table->cache_lock);
+	INIT_LIST_HEAD(&table->stripe_cache);
+
+	h = table->table;
+
+	for (i = 0; i < num_entries; i++) {
+		cur = h + i;
+		INIT_LIST_HEAD(&cur->hash_list);
+		spin_lock_init(&cur->lock);
+		init_waitqueue_head(&cur->wait);
+	}
+
+	x = cmpxchg(&info->stripe_hash_table, NULL, table);
+	if (x) {
+		if (is_vmalloc_addr(x))
+			vfree(x);
+		else
+			kfree(x);
+	}
+	return 0;
+}
+
+/*
+ * caching an rbio means to copy anything from the
+ * bio_pages array into the stripe_pages array.  We
+ * use the page uptodate bit in the stripe cache array
+ * to indicate if it has valid data
+ *
+ * once the caching is done, we set the cache ready
+ * bit.
+ */
+static void cache_rbio_pages(struct btrfs_raid_bio *rbio)
+{
+	int i;
+	char *s;
+	char *d;
+	int ret;
+
+	ret = alloc_rbio_pages(rbio);
+	if (ret)
+		return;
+
+	for (i = 0; i < rbio->nr_pages; i++) {
+		if (!rbio->bio_pages[i])
+			continue;
+
+		s = kmap(rbio->bio_pages[i]);
+		d = kmap(rbio->stripe_pages[i]);
+
+		memcpy(d, s, PAGE_CACHE_SIZE);
+
+		kunmap(rbio->bio_pages[i]);
+		kunmap(rbio->stripe_pages[i]);
+		SetPageUptodate(rbio->stripe_pages[i]);
+	}
+	set_bit(RBIO_CACHE_READY_BIT, &rbio->flags);
+}
+
+/*
+ * we hash on the first logical address of the stripe
+ */
+static int rbio_bucket(struct btrfs_raid_bio *rbio)
+{
+	u64 num = rbio->raid_map[0];
+
+	/*
+	 * we shift down quite a bit.  We're using byte
+	 * addressing, and most of the lower bits are zeros.
+	 * This tends to upset hash_64, and it consistently
+	 * returns just one or two different values.
+	 *
+	 * shifting off the lower bits fixes things.
+	 */
+	return hash_64(num >> 16, BTRFS_STRIPE_HASH_TABLE_BITS);
+}
+
+/*
+ * stealing an rbio means taking all the uptodate pages from the stripe
+ * array in the source rbio and putting them into the destination rbio
+ */
+static void steal_rbio(struct btrfs_raid_bio *src, struct btrfs_raid_bio *dest)
+{
+	int i;
+	struct page *s;
+	struct page *d;
+
+	if (!test_bit(RBIO_CACHE_READY_BIT, &src->flags))
+		return;
+
+	for (i = 0; i < dest->nr_pages; i++) {
+		s = src->stripe_pages[i];
+		if (!s || !PageUptodate(s)) {
+			continue;
+		}
+
+		d = dest->stripe_pages[i];
+		if (d)
+			__free_page(d);
+
+		dest->stripe_pages[i] = s;
+		src->stripe_pages[i] = NULL;
+	}
+}
+
+/*
+ * merging means we take the bio_list from the victim and
+ * splice it into the destination.  The victim should
+ * be discarded afterwards.
+ *
+ * must be called with dest->rbio_list_lock held
+ */
+static void merge_rbio(struct btrfs_raid_bio *dest,
+		       struct btrfs_raid_bio *victim)
+{
+	bio_list_merge(&dest->bio_list, &victim->bio_list);
+	dest->bio_list_bytes += victim->bio_list_bytes;
+	bio_list_init(&victim->bio_list);
+}
+
+/*
+ * used to prune items that are in the cache.  The caller
+ * must hold the hash table lock.
+ */
+static void __remove_rbio_from_cache(struct btrfs_raid_bio *rbio)
+{
+	int bucket = rbio_bucket(rbio);
+	struct btrfs_stripe_hash_table *table;
+	struct btrfs_stripe_hash *h;
+	int freeit = 0;
+
+	/*
+	 * check the bit again under the hash table lock.
+	 */
+	if (!test_bit(RBIO_CACHE_BIT, &rbio->flags))
+		return;
+
+	table = rbio->fs_info->stripe_hash_table;
+	h = table->table + bucket;
+
+	/* hold the lock for the bucket because we may be
+	 * removing it from the hash table
+	 */
+	spin_lock(&h->lock);
+
+	/*
+	 * hold the lock for the bio list because we need
+	 * to make sure the bio list is empty
+	 */
+	spin_lock(&rbio->bio_list_lock);
+
+	if (test_and_clear_bit(RBIO_CACHE_BIT, &rbio->flags)) {
+		list_del_init(&rbio->stripe_cache);
+		table->cache_size -= 1;
+		freeit = 1;
+
+		/* if the bio list isn't empty, this rbio is
+		 * still involved in an IO.  We take it out
+		 * of the cache list, and drop the ref that
+		 * was held for the list.
+		 *
+		 * If the bio_list was empty, we also remove
+		 * the rbio from the hash_table, and drop
+		 * the corresponding ref
+		 */
+		if (bio_list_empty(&rbio->bio_list)) {
+			if (!list_empty(&rbio->hash_list)) {
+				list_del_init(&rbio->hash_list);
+				atomic_dec(&rbio->refs);
+				BUG_ON(!list_empty(&rbio->plug_list));
+			}
+		}
+	}
+
+	spin_unlock(&rbio->bio_list_lock);
+	spin_unlock(&h->lock);
+
+	if (freeit)
+		__free_raid_bio(rbio);
+}
+
+/*
+ * prune a given rbio from the cache
+ */
+static void remove_rbio_from_cache(struct btrfs_raid_bio *rbio)
+{
+	struct btrfs_stripe_hash_table *table;
+	unsigned long flags;
+
+	if (!test_bit(RBIO_CACHE_BIT, &rbio->flags))
+		return;
+
+	table = rbio->fs_info->stripe_hash_table;
+
+	spin_lock_irqsave(&table->cache_lock, flags);
+	__remove_rbio_from_cache(rbio);
+	spin_unlock_irqrestore(&table->cache_lock, flags);
+}
+
+/*
+ * remove everything in the cache
+ */
+static void btrfs_clear_rbio_cache(struct btrfs_fs_info *info)
+{
+	struct btrfs_stripe_hash_table *table;
+	unsigned long flags;
+	struct btrfs_raid_bio *rbio;
+
+	table = info->stripe_hash_table;
+
+	spin_lock_irqsave(&table->cache_lock, flags);
+	while (!list_empty(&table->stripe_cache)) {
+		rbio = list_entry(table->stripe_cache.next,
+				  struct btrfs_raid_bio,
+				  stripe_cache);
+		__remove_rbio_from_cache(rbio);
+	}
+	spin_unlock_irqrestore(&table->cache_lock, flags);
+}
+
+/*
+ * remove all cached entries and free the hash table
+ * used by unmount
+ */
+void btrfs_free_stripe_hash_table(struct btrfs_fs_info *info)
+{
+	if (!info->stripe_hash_table)
+		return;
+	btrfs_clear_rbio_cache(info);
+	if (is_vmalloc_addr(info->stripe_hash_table))
+		vfree(info->stripe_hash_table);
+	else
+		kfree(info->stripe_hash_table);
+	info->stripe_hash_table = NULL;
+}
+
+/*
+ * insert an rbio into the stripe cache.  It
+ * must have already been prepared by calling
+ * cache_rbio_pages
+ *
+ * If this rbio was already cached, it gets
+ * moved to the front of the lru.
+ *
+ * If the size of the rbio cache is too big, we
+ * prune an item.
+ */
+static void cache_rbio(struct btrfs_raid_bio *rbio)
+{
+	struct btrfs_stripe_hash_table *table;
+	unsigned long flags;
+
+	if (!test_bit(RBIO_CACHE_READY_BIT, &rbio->flags))
+		return;
+
+	table = rbio->fs_info->stripe_hash_table;
+
+	spin_lock_irqsave(&table->cache_lock, flags);
+	spin_lock(&rbio->bio_list_lock);
+
+	/* bump our ref if we were not in the list before */
+	if (!test_and_set_bit(RBIO_CACHE_BIT, &rbio->flags))
+		atomic_inc(&rbio->refs);
+
+	if (!list_empty(&rbio->stripe_cache)){
+		list_move(&rbio->stripe_cache, &table->stripe_cache);
+	} else {
+		list_add(&rbio->stripe_cache, &table->stripe_cache);
+		table->cache_size += 1;
+	}
+
+	spin_unlock(&rbio->bio_list_lock);
+
+	if (table->cache_size > RBIO_CACHE_SIZE) {
+		struct btrfs_raid_bio *found;
+
+		found = list_entry(table->stripe_cache.prev,
+				  struct btrfs_raid_bio,
+				  stripe_cache);
+
+		if (found != rbio)
+			__remove_rbio_from_cache(found);
+	}
+
+	spin_unlock_irqrestore(&table->cache_lock, flags);
+	return;
+}
+
+/*
+ * helper function to run the xor_blocks api.  It is only
+ * able to do MAX_XOR_BLOCKS at a time, so we need to
+ * loop through.
+ */
+static void run_xor(void **pages, int src_cnt, ssize_t len)
+{
+	int src_off = 0;
+	int xor_src_cnt = 0;
+	void *dest = pages[src_cnt];
+
+	while(src_cnt > 0) {
+		xor_src_cnt = min(src_cnt, MAX_XOR_BLOCKS);
+		xor_blocks(xor_src_cnt, len, dest, pages + src_off);
+
+		src_cnt -= xor_src_cnt;
+		src_off += xor_src_cnt;
+	}
+}
+
+/*
+ * returns true if the bio list inside this rbio
+ * covers an entire stripe (no rmw required).
+ * Must be called with the bio list lock held, or
+ * at a time when you know it is impossible to add
+ * new bios into the list
+ */
+static int __rbio_is_full(struct btrfs_raid_bio *rbio)
+{
+	unsigned long size = rbio->bio_list_bytes;
+	int ret = 1;
+
+	if (size != rbio->nr_data * rbio->stripe_len)
+		ret = 0;
+
+	BUG_ON(size > rbio->nr_data * rbio->stripe_len);
+	return ret;
+}
+
+static int rbio_is_full(struct btrfs_raid_bio *rbio)
+{
+	unsigned long flags;
+	int ret;
+
+	spin_lock_irqsave(&rbio->bio_list_lock, flags);
+	ret = __rbio_is_full(rbio);
+	spin_unlock_irqrestore(&rbio->bio_list_lock, flags);
+	return ret;
+}
+
+/*
+ * returns 1 if it is safe to merge two rbios together.
+ * The merging is safe if the two rbios correspond to
+ * the same stripe and if they are both going in the same
+ * direction (read vs write), and if neither one is
+ * locked for final IO
+ *
+ * The caller is responsible for locking such that
+ * rmw_locked is safe to test
+ */
+static int rbio_can_merge(struct btrfs_raid_bio *last,
+			  struct btrfs_raid_bio *cur)
+{
+	if (test_bit(RBIO_RMW_LOCKED_BIT, &last->flags) ||
+	    test_bit(RBIO_RMW_LOCKED_BIT, &cur->flags))
+		return 0;
+
+	/*
+	 * we can't merge with cached rbios, since the
+	 * idea is that when we merge the destination
+	 * rbio is going to run our IO for us.  We can
+	 * steal from cached rbio's though, other functions
+	 * handle that.
+	 */
+	if (test_bit(RBIO_CACHE_BIT, &last->flags) ||
+	    test_bit(RBIO_CACHE_BIT, &cur->flags))
+		return 0;
+
+	if (last->raid_map[0] !=
+	    cur->raid_map[0])
+		return 0;
+
+	/* reads can't merge with writes */
+	if (last->read_rebuild !=
+	    cur->read_rebuild) {
+		return 0;
+	}
+
+	return 1;
+}
+
+/*
+ * helper to index into the pstripe
+ */
+static struct page *rbio_pstripe_page(struct btrfs_raid_bio *rbio, int index)
+{
+	index += (rbio->nr_data * rbio->stripe_len) >> PAGE_CACHE_SHIFT;
+	return rbio->stripe_pages[index];
+}
+
+/*
+ * helper to index into the qstripe, returns null
+ * if there is no qstripe
+ */
+static struct page *rbio_qstripe_page(struct btrfs_raid_bio *rbio, int index)
+{
+	if (rbio->nr_data + 1 == rbio->bbio->num_stripes)
+		return NULL;
+
+	index += ((rbio->nr_data + 1) * rbio->stripe_len) >>
+		PAGE_CACHE_SHIFT;
+	return rbio->stripe_pages[index];
+}
+
+/*
+ * The first stripe in the table for a logical address
+ * has the lock.  rbios are added in one of three ways:
+ *
+ * 1) Nobody has the stripe locked yet.  The rbio is given
+ * the lock and 0 is returned.  The caller must start the IO
+ * themselves.
+ *
+ * 2) Someone has the stripe locked, but we're able to merge
+ * with the lock owner.  The rbio is freed and the IO will
+ * start automatically along with the existing rbio.  1 is returned.
+ *
+ * 3) Someone has the stripe locked, but we're not able to merge.
+ * The rbio is added to the lock owner's plug list, or merged into
+ * an rbio already on the plug list.  When the lock owner unlocks,
+ * the next rbio on the list is run and the IO is started automatically.
+ * 1 is returned
+ *
+ * If we return 0, the caller still owns the rbio and must continue with
+ * IO submission.  If we return 1, the caller must assume the rbio has
+ * already been freed.
+ */
+static noinline int lock_stripe_add(struct btrfs_raid_bio *rbio)
+{
+	int bucket = rbio_bucket(rbio);
+	struct btrfs_stripe_hash *h = rbio->fs_info->stripe_hash_table->table + bucket;
+	struct btrfs_raid_bio *cur;
+	struct btrfs_raid_bio *pending;
+	unsigned long flags;
+	DEFINE_WAIT(wait);
+	struct btrfs_raid_bio *freeit = NULL;
+	struct btrfs_raid_bio *cache_drop = NULL;
+	int ret = 0;
+	int walk = 0;
+
+	spin_lock_irqsave(&h->lock, flags);
+	list_for_each_entry(cur, &h->hash_list, hash_list) {
+		walk++;
+		if (cur->raid_map[0] == rbio->raid_map[0]) {
+			spin_lock(&cur->bio_list_lock);
+
+			/* can we steal this cached rbio's pages? */
+			if (bio_list_empty(&cur->bio_list) &&
+			    list_empty(&cur->plug_list) &&
+			    test_bit(RBIO_CACHE_BIT, &cur->flags) &&
+			    !test_bit(RBIO_RMW_LOCKED_BIT, &cur->flags)) {
+				list_del_init(&cur->hash_list);
+				atomic_dec(&cur->refs);
+
+				steal_rbio(cur, rbio);
+				cache_drop = cur;
+				spin_unlock(&cur->bio_list_lock);
+
+				goto lockit;
+			}
+
+			/* can we merge into the lock owner? */
+			if (rbio_can_merge(cur, rbio)) {
+				merge_rbio(cur, rbio);
+				spin_unlock(&cur->bio_list_lock);
+				freeit = rbio;
+				ret = 1;
+				goto out;
+			}
+
+
+			/*
+			 * we couldn't merge with the running
+			 * rbio, see if we can merge with the
+			 * pending ones.  We don't have to
+			 * check for rmw_locked because there
+			 * is no way they are inside finish_rmw
+			 * right now
+			 */
+			list_for_each_entry(pending, &cur->plug_list,
+					    plug_list) {
+				if (rbio_can_merge(pending, rbio)) {
+					merge_rbio(pending, rbio);
+					spin_unlock(&cur->bio_list_lock);
+					freeit = rbio;
+					ret = 1;
+					goto out;
+				}
+			}
+
+			/* no merging, put us on the tail of the plug list,
+			 * our rbio will be started with the currently
+			 * running rbio unlocks
+			 */
+			list_add_tail(&rbio->plug_list, &cur->plug_list);
+			spin_unlock(&cur->bio_list_lock);
+			ret = 1;
+			goto out;
+		}
+	}
+lockit:
+	atomic_inc(&rbio->refs);
+	list_add(&rbio->hash_list, &h->hash_list);
+out:
+	spin_unlock_irqrestore(&h->lock, flags);
+	if (cache_drop)
+		remove_rbio_from_cache(cache_drop);
+	if (freeit)
+		__free_raid_bio(freeit);
+	return ret;
+}
+
+/*
+ * called as rmw or parity rebuild is completed.  If the plug list has more
+ * rbios waiting for this stripe, the next one on the list will be started
+ */
+static noinline void unlock_stripe(struct btrfs_raid_bio *rbio)
+{
+	int bucket;
+	struct btrfs_stripe_hash *h;
+	unsigned long flags;
+	int keep_cache = 0;
+
+	bucket = rbio_bucket(rbio);
+	h = rbio->fs_info->stripe_hash_table->table + bucket;
+
+	if (list_empty(&rbio->plug_list))
+		cache_rbio(rbio);
+
+	spin_lock_irqsave(&h->lock, flags);
+	spin_lock(&rbio->bio_list_lock);
+
+	if (!list_empty(&rbio->hash_list)) {
+		/*
+		 * if we're still cached and there is no other IO
+		 * to perform, just leave this rbio here for others
+		 * to steal from later
+		 */
+		if (list_empty(&rbio->plug_list) &&
+		    test_bit(RBIO_CACHE_BIT, &rbio->flags)) {
+			keep_cache = 1;
+			clear_bit(RBIO_RMW_LOCKED_BIT, &rbio->flags);
+			BUG_ON(!bio_list_empty(&rbio->bio_list));
+			goto done;
+		}
+
+		list_del_init(&rbio->hash_list);
+		atomic_dec(&rbio->refs);
+
+		/*
+		 * we use the plug list to hold all the rbios
+		 * waiting for the chance to lock this stripe.
+		 * hand the lock over to one of them.
+		 */
+		if (!list_empty(&rbio->plug_list)) {
+			struct btrfs_raid_bio *next;
+			struct list_head *head = rbio->plug_list.next;
+
+			next = list_entry(head, struct btrfs_raid_bio,
+					  plug_list);
+
+			list_del_init(&rbio->plug_list);
+
+			list_add(&next->hash_list, &h->hash_list);
+			atomic_inc(&next->refs);
+			spin_unlock(&rbio->bio_list_lock);
+			spin_unlock_irqrestore(&h->lock, flags);
+
+			if (next->read_rebuild)
+				async_read_rebuild(next);
+			else {
+				steal_rbio(rbio, next);
+				async_rmw_stripe(next);
+			}
+
+			goto done_nolock;
+		} else  if (waitqueue_active(&h->wait)) {
+			spin_unlock(&rbio->bio_list_lock);
+			spin_unlock_irqrestore(&h->lock, flags);
+			wake_up(&h->wait);
+			goto done_nolock;
+		}
+	}
+done:
+	spin_unlock(&rbio->bio_list_lock);
+	spin_unlock_irqrestore(&h->lock, flags);
+
+done_nolock:
+	if (!keep_cache)
+		remove_rbio_from_cache(rbio);
+}
+
+static void __free_raid_bio(struct btrfs_raid_bio *rbio)
+{
+	int i;
+
+	WARN_ON(atomic_read(&rbio->refs) < 0);
+	if (!atomic_dec_and_test(&rbio->refs))
+		return;
+
+	WARN_ON(!list_empty(&rbio->stripe_cache));
+	WARN_ON(!list_empty(&rbio->hash_list));
+	WARN_ON(!bio_list_empty(&rbio->bio_list));
+
+	for (i = 0; i < rbio->nr_pages; i++) {
+		if (rbio->stripe_pages[i]) {
+			__free_page(rbio->stripe_pages[i]);
+			rbio->stripe_pages[i] = NULL;
+		}
+	}
+	kfree(rbio->raid_map);
+	kfree(rbio->bbio);
+	kfree(rbio);
+}
+
+static void free_raid_bio(struct btrfs_raid_bio *rbio)
+{
+	unlock_stripe(rbio);
+	__free_raid_bio(rbio);
+}
+
+/*
+ * this frees the rbio and runs through all the bios in the
+ * bio_list and calls end_io on them
+ */
+static void rbio_orig_end_io(struct btrfs_raid_bio *rbio, int err, int uptodate)
+{
+	struct bio *cur = bio_list_get(&rbio->bio_list);
+	struct bio *next;
+	free_raid_bio(rbio);
+
+	while (cur) {
+		next = cur->bi_next;
+		cur->bi_next = NULL;
+		if (uptodate)
+			set_bit(BIO_UPTODATE, &cur->bi_flags);
+		bio_endio(cur, err);
+		cur = next;
+	}
+}
+
+/*
+ * end io function used by finish_rmw.  When we finally
+ * get here, we've written a full stripe
+ */
+static void raid_write_end_io(struct bio *bio, int err)
+{
+	struct btrfs_raid_bio *rbio = bio->bi_private;
+
+	if (err)
+		fail_bio_stripe(rbio, bio);
+
+	bio_put(bio);
+
+	if (!atomic_dec_and_test(&rbio->bbio->stripes_pending))
+		return;
+
+	err = 0;
+
+	/* OK, we have read all the stripes we need to. */
+	if (atomic_read(&rbio->bbio->error) > rbio->bbio->max_errors)
+		err = -EIO;
+
+	rbio_orig_end_io(rbio, err, 0);
+	return;
+}
+
+/*
+ * the read/modify/write code wants to use the original bio for
+ * any pages it included, and then use the rbio for everything
+ * else.  This function decides if a given index (stripe number)
+ * and page number in that stripe fall inside the original bio
+ * or the rbio.
+ *
+ * if you set bio_list_only, you'll get a NULL back for any ranges
+ * that are outside the bio_list
+ *
+ * This doesn't take any refs on anything, you get a bare page pointer
+ * and the caller must bump refs as required.
+ *
+ * You must call index_rbio_pages once before you can trust
+ * the answers from this function.
+ */
+static struct page *page_in_rbio(struct btrfs_raid_bio *rbio,
+				 int index, int pagenr, int bio_list_only)
+{
+	int chunk_page;
+	struct page *p = NULL;
+
+	chunk_page = index * (rbio->stripe_len >> PAGE_SHIFT) + pagenr;
+
+	spin_lock_irq(&rbio->bio_list_lock);
+	p = rbio->bio_pages[chunk_page];
+	spin_unlock_irq(&rbio->bio_list_lock);
+
+	if (p || bio_list_only)
+		return p;
+
+	return rbio->stripe_pages[chunk_page];
+}
+
+/*
+ * number of pages we need for the entire stripe across all the
+ * drives
+ */
+static unsigned long rbio_nr_pages(unsigned long stripe_len, int nr_stripes)
+{
+	unsigned long nr = stripe_len * nr_stripes;
+	return (nr + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
+}
+
+/*
+ * allocation and initial setup for the btrfs_raid_bio.  Not
+ * this does not allocate any pages for rbio->pages.
+ */
+static struct btrfs_raid_bio *alloc_rbio(struct btrfs_root *root,
+			  struct btrfs_bio *bbio, u64 *raid_map,
+			  u64 stripe_len)
+{
+	struct btrfs_raid_bio *rbio;
+	int nr_data = 0;
+	int num_pages = rbio_nr_pages(stripe_len, bbio->num_stripes);
+	void *p;
+
+	rbio = kzalloc(sizeof(*rbio) + num_pages * sizeof(struct page *) * 2,
+			GFP_NOFS);
+	if (!rbio) {
+		kfree(raid_map);
+		kfree(bbio);
+		return ERR_PTR(-ENOMEM);
+	}
+
+	bio_list_init(&rbio->bio_list);
+	INIT_LIST_HEAD(&rbio->plug_list);
+	spin_lock_init(&rbio->bio_list_lock);
+	INIT_LIST_HEAD(&rbio->stripe_cache);
+	INIT_LIST_HEAD(&rbio->hash_list);
+	rbio->bbio = bbio;
+	rbio->raid_map = raid_map;
+	rbio->fs_info = root->fs_info;
+	rbio->stripe_len = stripe_len;
+	rbio->nr_pages = num_pages;
+	rbio->faila = -1;
+	rbio->failb = -1;
+	atomic_set(&rbio->refs, 1);
+
+	/*
+	 * the stripe_pages and bio_pages array point to the extra
+	 * memory we allocated past the end of the rbio
+	 */
+	p = rbio + 1;
+	rbio->stripe_pages = p;
+	rbio->bio_pages = p + sizeof(struct page *) * num_pages;
+
+	if (raid_map[bbio->num_stripes - 1] == RAID6_Q_STRIPE)
+		nr_data = bbio->num_stripes - 2;
+	else
+		nr_data = bbio->num_stripes - 1;
+
+	rbio->nr_data = nr_data;
+	return rbio;
+}
+
+/* allocate pages for all the stripes in the bio, including parity */
+static int alloc_rbio_pages(struct btrfs_raid_bio *rbio)
+{
+	int i;
+	struct page *page;
+
+	for (i = 0; i < rbio->nr_pages; i++) {
+		if (rbio->stripe_pages[i])
+			continue;
+		page = alloc_page(GFP_NOFS | __GFP_HIGHMEM);
+		if (!page)
+			return -ENOMEM;
+		rbio->stripe_pages[i] = page;
+		ClearPageUptodate(page);
+	}
+	return 0;
+}
+
+/* allocate pages for just the p/q stripes */
+static int alloc_rbio_parity_pages(struct btrfs_raid_bio *rbio)
+{
+	int i;
+	struct page *page;
+
+	i = (rbio->nr_data * rbio->stripe_len) >> PAGE_CACHE_SHIFT;
+
+	for (; i < rbio->nr_pages; i++) {
+		if (rbio->stripe_pages[i])
+			continue;
+		page = alloc_page(GFP_NOFS | __GFP_HIGHMEM);
+		if (!page)
+			return -ENOMEM;
+		rbio->stripe_pages[i] = page;
+	}
+	return 0;
+}
+
+/*
+ * add a single page from a specific stripe into our list of bios for IO
+ * this will try to merge into existing bios if possible, and returns
+ * zero if all went well.
+ */
+static int rbio_add_io_page(struct btrfs_raid_bio *rbio,
+			    struct bio_list *bio_list,
+			    struct page *page,
+			    int stripe_nr,
+			    unsigned long page_index,
+			    unsigned long bio_max_len)
+{
+	struct bio *last = bio_list->tail;
+	u64 last_end = 0;
+	int ret;
+	struct bio *bio;
+	struct btrfs_bio_stripe *stripe;
+	u64 disk_start;
+
+	stripe = &rbio->bbio->stripes[stripe_nr];
+	disk_start = stripe->physical + (page_index << PAGE_CACHE_SHIFT);
+
+	/* if the device is missing, just fail this stripe */
+	if (!stripe->dev->bdev)
+		return fail_rbio_index(rbio, stripe_nr);
+
+	/* see if we can add this page onto our existing bio */
+	if (last) {
+		last_end = (u64)last->bi_iter.bi_sector << 9;
+		last_end += last->bi_iter.bi_size;
+
+		/*
+		 * we can't merge these if they are from different
+		 * devices or if they are not contiguous
+		 */
+		if (last_end == disk_start && stripe->dev->bdev &&
+		    test_bit(BIO_UPTODATE, &last->bi_flags) &&
+		    last->bi_bdev == stripe->dev->bdev) {
+			ret = bio_add_page(last, page, PAGE_CACHE_SIZE, 0);
+			if (ret == PAGE_CACHE_SIZE)
+				return 0;
+		}
+	}
+
+	/* put a new bio on the list */
+	bio = btrfs_io_bio_alloc(GFP_NOFS, bio_max_len >> PAGE_SHIFT?:1);
+	if (!bio)
+		return -ENOMEM;
+
+	bio->bi_iter.bi_size = 0;
+	bio->bi_bdev = stripe->dev->bdev;
+	bio->bi_iter.bi_sector = disk_start >> 9;
+	set_bit(BIO_UPTODATE, &bio->bi_flags);
+
+	bio_add_page(bio, page, PAGE_CACHE_SIZE, 0);
+	bio_list_add(bio_list, bio);
+	return 0;
+}
+
+/*
+ * while we're doing the read/modify/write cycle, we could
+ * have errors in reading pages off the disk.  This checks
+ * for errors and if we're not able to read the page it'll
+ * trigger parity reconstruction.  The rmw will be finished
+ * after we've reconstructed the failed stripes
+ */
+static void validate_rbio_for_rmw(struct btrfs_raid_bio *rbio)
+{
+	if (rbio->faila >= 0 || rbio->failb >= 0) {
+		BUG_ON(rbio->faila == rbio->bbio->num_stripes - 1);
+		__raid56_parity_recover(rbio);
+	} else {
+		finish_rmw(rbio);
+	}
+}
+
+/*
+ * these are just the pages from the rbio array, not from anything
+ * the FS sent down to us
+ */
+static struct page *rbio_stripe_page(struct btrfs_raid_bio *rbio, int stripe, int page)
+{
+	int index;
+	index = stripe * (rbio->stripe_len >> PAGE_CACHE_SHIFT);
+	index += page;
+	return rbio->stripe_pages[index];
+}
+
+/*
+ * helper function to walk our bio list and populate the bio_pages array with
+ * the result.  This seems expensive, but it is faster than constantly
+ * searching through the bio list as we setup the IO in finish_rmw or stripe
+ * reconstruction.
+ *
+ * This must be called before you trust the answers from page_in_rbio
+ */
+static void index_rbio_pages(struct btrfs_raid_bio *rbio)
+{
+	struct bio *bio;
+	u64 start;
+	unsigned long stripe_offset;
+	unsigned long page_index;
+	struct page *p;
+	int i;
+
+	spin_lock_irq(&rbio->bio_list_lock);
+	bio_list_for_each(bio, &rbio->bio_list) {
+		start = (u64)bio->bi_iter.bi_sector << 9;
+		stripe_offset = start - rbio->raid_map[0];
+		page_index = stripe_offset >> PAGE_CACHE_SHIFT;
+
+		for (i = 0; i < bio->bi_vcnt; i++) {
+			p = bio->bi_io_vec[i].bv_page;
+			rbio->bio_pages[page_index + i] = p;
+		}
+	}
+	spin_unlock_irq(&rbio->bio_list_lock);
+}
+
+/*
+ * this is called from one of two situations.  We either
+ * have a full stripe from the higher layers, or we've read all
+ * the missing bits off disk.
+ *
+ * This will calculate the parity and then send down any
+ * changed blocks.
+ */
+static noinline void finish_rmw(struct btrfs_raid_bio *rbio)
+{
+	struct btrfs_bio *bbio = rbio->bbio;
+	void *pointers[bbio->num_stripes];
+	int stripe_len = rbio->stripe_len;
+	int nr_data = rbio->nr_data;
+	int stripe;
+	int pagenr;
+	int p_stripe = -1;
+	int q_stripe = -1;
+	struct bio_list bio_list;
+	struct bio *bio;
+	int pages_per_stripe = stripe_len >> PAGE_CACHE_SHIFT;
+	int ret;
+
+	bio_list_init(&bio_list);
+
+	if (bbio->num_stripes - rbio->nr_data == 1) {
+		p_stripe = bbio->num_stripes - 1;
+	} else if (bbio->num_stripes - rbio->nr_data == 2) {
+		p_stripe = bbio->num_stripes - 2;
+		q_stripe = bbio->num_stripes - 1;
+	} else {
+		BUG();
+	}
+
+	/* at this point we either have a full stripe,
+	 * or we've read the full stripe from the drive.
+	 * recalculate the parity and write the new results.
+	 *
+	 * We're not allowed to add any new bios to the
+	 * bio list here, anyone else that wants to
+	 * change this stripe needs to do their own rmw.
+	 */
+	spin_lock_irq(&rbio->bio_list_lock);
+	set_bit(RBIO_RMW_LOCKED_BIT, &rbio->flags);
+	spin_unlock_irq(&rbio->bio_list_lock);
+
+	atomic_set(&rbio->bbio->error, 0);
+
+	/*
+	 * now that we've set rmw_locked, run through the
+	 * bio list one last time and map the page pointers
+	 *
+	 * We don't cache full rbios because we're assuming
+	 * the higher layers are unlikely to use this area of
+	 * the disk again soon.  If they do use it again,
+	 * hopefully they will send another full bio.
+	 */
+	index_rbio_pages(rbio);
+	if (!rbio_is_full(rbio))
+		cache_rbio_pages(rbio);
+	else
+		clear_bit(RBIO_CACHE_READY_BIT, &rbio->flags);
+
+	for (pagenr = 0; pagenr < pages_per_stripe; pagenr++) {
+		struct page *p;
+		/* first collect one page from each data stripe */
+		for (stripe = 0; stripe < nr_data; stripe++) {
+			p = page_in_rbio(rbio, stripe, pagenr, 0);
+			pointers[stripe] = kmap(p);
+		}
+
+		/* then add the parity stripe */
+		p = rbio_pstripe_page(rbio, pagenr);
+		SetPageUptodate(p);
+		pointers[stripe++] = kmap(p);
+
+		if (q_stripe != -1) {
+
+			/*
+			 * raid6, add the qstripe and call the
+			 * library function to fill in our p/q
+			 */
+			p = rbio_qstripe_page(rbio, pagenr);
+			SetPageUptodate(p);
+			pointers[stripe++] = kmap(p);
+
+			raid6_call.gen_syndrome(bbio->num_stripes, PAGE_SIZE,
+						pointers);
+		} else {
+			/* raid5 */
+			memcpy(pointers[nr_data], pointers[0], PAGE_SIZE);
+			run_xor(pointers + 1, nr_data - 1, PAGE_CACHE_SIZE);
+		}
+
+
+		for (stripe = 0; stripe < bbio->num_stripes; stripe++)
+			kunmap(page_in_rbio(rbio, stripe, pagenr, 0));
+	}
+
+	/*
+	 * time to start writing.  Make bios for everything from the
+	 * higher layers (the bio_list in our rbio) and our p/q.  Ignore
+	 * everything else.
+	 */
+	for (stripe = 0; stripe < bbio->num_stripes; stripe++) {
+		for (pagenr = 0; pagenr < pages_per_stripe; pagenr++) {
+			struct page *page;
+			if (stripe < rbio->nr_data) {
+				page = page_in_rbio(rbio, stripe, pagenr, 1);
+				if (!page)
+					continue;
+			} else {
+			       page = rbio_stripe_page(rbio, stripe, pagenr);
+			}
+
+			ret = rbio_add_io_page(rbio, &bio_list,
+				       page, stripe, pagenr, rbio->stripe_len);
+			if (ret)
+				goto cleanup;
+		}
+	}
+
+	atomic_set(&bbio->stripes_pending, bio_list_size(&bio_list));
+	BUG_ON(atomic_read(&bbio->stripes_pending) == 0);
+
+	while (1) {
+		bio = bio_list_pop(&bio_list);
+		if (!bio)
+			break;
+
+		bio->bi_private = rbio;
+		bio->bi_end_io = raid_write_end_io;
+		BUG_ON(!test_bit(BIO_UPTODATE, &bio->bi_flags));
+		submit_bio(WRITE, bio);
+	}
+	return;
+
+cleanup:
+	rbio_orig_end_io(rbio, -EIO, 0);
+}
+
+/*
+ * helper to find the stripe number for a given bio.  Used to figure out which
+ * stripe has failed.  This expects the bio to correspond to a physical disk,
+ * so it looks up based on physical sector numbers.
+ */
+static int find_bio_stripe(struct btrfs_raid_bio *rbio,
+			   struct bio *bio)
+{
+	u64 physical = bio->bi_iter.bi_sector;
+	u64 stripe_start;
+	int i;
+	struct btrfs_bio_stripe *stripe;
+
+	physical <<= 9;
+
+	for (i = 0; i < rbio->bbio->num_stripes; i++) {
+		stripe = &rbio->bbio->stripes[i];
+		stripe_start = stripe->physical;
+		if (physical >= stripe_start &&
+		    physical < stripe_start + rbio->stripe_len) {
+			return i;
+		}
+	}
+	return -1;
+}
+
+/*
+ * helper to find the stripe number for a given
+ * bio (before mapping).  Used to figure out which stripe has
+ * failed.  This looks up based on logical block numbers.
+ */
+static int find_logical_bio_stripe(struct btrfs_raid_bio *rbio,
+				   struct bio *bio)
+{
+	u64 logical = bio->bi_iter.bi_sector;
+	u64 stripe_start;
+	int i;
+
+	logical <<= 9;
+
+	for (i = 0; i < rbio->nr_data; i++) {
+		stripe_start = rbio->raid_map[i];
+		if (logical >= stripe_start &&
+		    logical < stripe_start + rbio->stripe_len) {
+			return i;
+		}
+	}
+	return -1;
+}
+
+/*
+ * returns -EIO if we had too many failures
+ */
+static int fail_rbio_index(struct btrfs_raid_bio *rbio, int failed)
+{
+	unsigned long flags;
+	int ret = 0;
+
+	spin_lock_irqsave(&rbio->bio_list_lock, flags);
+
+	/* we already know this stripe is bad, move on */
+	if (rbio->faila == failed || rbio->failb == failed)
+		goto out;
+
+	if (rbio->faila == -1) {
+		/* first failure on this rbio */
+		rbio->faila = failed;
+		atomic_inc(&rbio->bbio->error);
+	} else if (rbio->failb == -1) {
+		/* second failure on this rbio */
+		rbio->failb = failed;
+		atomic_inc(&rbio->bbio->error);
+	} else {
+		ret = -EIO;
+	}
+out:
+	spin_unlock_irqrestore(&rbio->bio_list_lock, flags);
+
+	return ret;
+}
+
+/*
+ * helper to fail a stripe based on a physical disk
+ * bio.
+ */
+static int fail_bio_stripe(struct btrfs_raid_bio *rbio,
+			   struct bio *bio)
+{
+	int failed = find_bio_stripe(rbio, bio);
+
+	if (failed < 0)
+		return -EIO;
+
+	return fail_rbio_index(rbio, failed);
+}
+
+/*
+ * this sets each page in the bio uptodate.  It should only be used on private
+ * rbio pages, nothing that comes in from the higher layers
+ */
+static void set_bio_pages_uptodate(struct bio *bio)
+{
+	int i;
+	struct page *p;
+
+	for (i = 0; i < bio->bi_vcnt; i++) {
+		p = bio->bi_io_vec[i].bv_page;
+		SetPageUptodate(p);
+	}
+}
+
+/*
+ * end io for the read phase of the rmw cycle.  All the bios here are physical
+ * stripe bios we've read from the disk so we can recalculate the parity of the
+ * stripe.
+ *
+ * This will usually kick off finish_rmw once all the bios are read in, but it
+ * may trigger parity reconstruction if we had any errors along the way
+ */
+static void raid_rmw_end_io(struct bio *bio, int err)
+{
+	struct btrfs_raid_bio *rbio = bio->bi_private;
+
+	if (err)
+		fail_bio_stripe(rbio, bio);
+	else
+		set_bio_pages_uptodate(bio);
+
+	bio_put(bio);
+
+	if (!atomic_dec_and_test(&rbio->bbio->stripes_pending))
+		return;
+
+	err = 0;
+	if (atomic_read(&rbio->bbio->error) > rbio->bbio->max_errors)
+		goto cleanup;
+
+	/*
+	 * this will normally call finish_rmw to start our write
+	 * but if there are any failed stripes we'll reconstruct
+	 * from parity first
+	 */
+	validate_rbio_for_rmw(rbio);
+	return;
+
+cleanup:
+
+	rbio_orig_end_io(rbio, -EIO, 0);
+}
+
+static void async_rmw_stripe(struct btrfs_raid_bio *rbio)
+{
+	btrfs_init_work(&rbio->work, rmw_work, NULL, NULL);
+
+	btrfs_queue_work(rbio->fs_info->rmw_workers,
+			 &rbio->work);
+}
+
+static void async_read_rebuild(struct btrfs_raid_bio *rbio)
+{
+	btrfs_init_work(&rbio->work, read_rebuild_work, NULL, NULL);
+
+	btrfs_queue_work(rbio->fs_info->rmw_workers,
+			 &rbio->work);
+}
+
+/*
+ * the stripe must be locked by the caller.  It will
+ * unlock after all the writes are done
+ */
+static int raid56_rmw_stripe(struct btrfs_raid_bio *rbio)
+{
+	int bios_to_read = 0;
+	struct btrfs_bio *bbio = rbio->bbio;
+	struct bio_list bio_list;
+	int ret;
+	int nr_pages = (rbio->stripe_len + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
+	int pagenr;
+	int stripe;
+	struct bio *bio;
+
+	bio_list_init(&bio_list);
+
+	ret = alloc_rbio_pages(rbio);
+	if (ret)
+		goto cleanup;
+
+	index_rbio_pages(rbio);
+
+	atomic_set(&rbio->bbio->error, 0);
+	/*
+	 * build a list of bios to read all the missing parts of this
+	 * stripe
+	 */
+	for (stripe = 0; stripe < rbio->nr_data; stripe++) {
+		for (pagenr = 0; pagenr < nr_pages; pagenr++) {
+			struct page *page;
+			/*
+			 * we want to find all the pages missing from
+			 * the rbio and read them from the disk.  If
+			 * page_in_rbio finds a page in the bio list
+			 * we don't need to read it off the stripe.
+			 */
+			page = page_in_rbio(rbio, stripe, pagenr, 1);
+			if (page)
+				continue;
+
+			page = rbio_stripe_page(rbio, stripe, pagenr);
+			/*
+			 * the bio cache may have handed us an uptodate
+			 * page.  If so, be happy and use it
+			 */
+			if (PageUptodate(page))
+				continue;
+
+			ret = rbio_add_io_page(rbio, &bio_list, page,
+				       stripe, pagenr, rbio->stripe_len);
+			if (ret)
+				goto cleanup;
+		}
+	}
+
+	bios_to_read = bio_list_size(&bio_list);
+	if (!bios_to_read) {
+		/*
+		 * this can happen if others have merged with
+		 * us, it means there is nothing left to read.
+		 * But if there are missing devices it may not be
+		 * safe to do the full stripe write yet.
+		 */
+		goto finish;
+	}
+
+	/*
+	 * the bbio may be freed once we submit the last bio.  Make sure
+	 * not to touch it after that
+	 */
+	atomic_set(&bbio->stripes_pending, bios_to_read);
+	while (1) {
+		bio = bio_list_pop(&bio_list);
+		if (!bio)
+			break;
+
+		bio->bi_private = rbio;
+		bio->bi_end_io = raid_rmw_end_io;
+
+		btrfs_bio_wq_end_io(rbio->fs_info, bio,
+				    BTRFS_WQ_ENDIO_RAID56);
+
+		BUG_ON(!test_bit(BIO_UPTODATE, &bio->bi_flags));
+		submit_bio(READ, bio);
+	}
+	/* the actual write will happen once the reads are done */
+	return 0;
+
+cleanup:
+	rbio_orig_end_io(rbio, -EIO, 0);
+	return -EIO;
+
+finish:
+	validate_rbio_for_rmw(rbio);
+	return 0;
+}
+
+/*
+ * if the upper layers pass in a full stripe, we thank them by only allocating
+ * enough pages to hold the parity, and sending it all down quickly.
+ */
+static int full_stripe_write(struct btrfs_raid_bio *rbio)
+{
+	int ret;
+
+	ret = alloc_rbio_parity_pages(rbio);
+	if (ret) {
+		__free_raid_bio(rbio);
+		return ret;
+	}
+
+	ret = lock_stripe_add(rbio);
+	if (ret == 0)
+		finish_rmw(rbio);
+	return 0;
+}
+
+/*
+ * partial stripe writes get handed over to async helpers.
+ * We're really hoping to merge a few more writes into this
+ * rbio before calculating new parity
+ */
+static int partial_stripe_write(struct btrfs_raid_bio *rbio)
+{
+	int ret;
+
+	ret = lock_stripe_add(rbio);
+	if (ret == 0)
+		async_rmw_stripe(rbio);
+	return 0;
+}
+
+/*
+ * sometimes while we were reading from the drive to
+ * recalculate parity, enough new bios come into create
+ * a full stripe.  So we do a check here to see if we can
+ * go directly to finish_rmw
+ */
+static int __raid56_parity_write(struct btrfs_raid_bio *rbio)
+{
+	/* head off into rmw land if we don't have a full stripe */
+	if (!rbio_is_full(rbio))
+		return partial_stripe_write(rbio);
+	return full_stripe_write(rbio);
+}
+
+/*
+ * We use plugging call backs to collect full stripes.
+ * Any time we get a partial stripe write while plugged
+ * we collect it into a list.  When the unplug comes down,
+ * we sort the list by logical block number and merge
+ * everything we can into the same rbios
+ */
+struct btrfs_plug_cb {
+	struct blk_plug_cb cb;
+	struct btrfs_fs_info *info;
+	struct list_head rbio_list;
+	struct btrfs_work work;
+};
+
+/*
+ * rbios on the plug list are sorted for easier merging.
+ */
+static int plug_cmp(void *priv, struct list_head *a, struct list_head *b)
+{
+	struct btrfs_raid_bio *ra = container_of(a, struct btrfs_raid_bio,
+						 plug_list);
+	struct btrfs_raid_bio *rb = container_of(b, struct btrfs_raid_bio,
+						 plug_list);
+	u64 a_sector = ra->bio_list.head->bi_iter.bi_sector;
+	u64 b_sector = rb->bio_list.head->bi_iter.bi_sector;
+
+	if (a_sector < b_sector)
+		return -1;
+	if (a_sector > b_sector)
+		return 1;
+	return 0;
+}
+
+static void run_plug(struct btrfs_plug_cb *plug)
+{
+	struct btrfs_raid_bio *cur;
+	struct btrfs_raid_bio *last = NULL;
+
+	/*
+	 * sort our plug list then try to merge
+	 * everything we can in hopes of creating full
+	 * stripes.
+	 */
+	list_sort(NULL, &plug->rbio_list, plug_cmp);
+	while (!list_empty(&plug->rbio_list)) {
+		cur = list_entry(plug->rbio_list.next,
+				 struct btrfs_raid_bio, plug_list);
+		list_del_init(&cur->plug_list);
+
+		if (rbio_is_full(cur)) {
+			/* we have a full stripe, send it down */
+			full_stripe_write(cur);
+			continue;
+		}
+		if (last) {
+			if (rbio_can_merge(last, cur)) {
+				merge_rbio(last, cur);
+				__free_raid_bio(cur);
+				continue;
+
+			}
+			__raid56_parity_write(last);
+		}
+		last = cur;
+	}
+	if (last) {
+		__raid56_parity_write(last);
+	}
+	kfree(plug);
+}
+
+/*
+ * if the unplug comes from schedule, we have to push the
+ * work off to a helper thread
+ */
+static void unplug_work(struct btrfs_work *work)
+{
+	struct btrfs_plug_cb *plug;
+	plug = container_of(work, struct btrfs_plug_cb, work);
+	run_plug(plug);
+}
+
+static void btrfs_raid_unplug(struct blk_plug_cb *cb, bool from_schedule)
+{
+	struct btrfs_plug_cb *plug;
+	plug = container_of(cb, struct btrfs_plug_cb, cb);
+
+	if (from_schedule) {
+		btrfs_init_work(&plug->work, unplug_work, NULL, NULL);
+		btrfs_queue_work(plug->info->rmw_workers,
+				 &plug->work);
+		return;
+	}
+	run_plug(plug);
+}
+
+/*
+ * our main entry point for writes from the rest of the FS.
+ */
+int raid56_parity_write(struct btrfs_root *root, struct bio *bio,
+			struct btrfs_bio *bbio, u64 *raid_map,
+			u64 stripe_len)
+{
+	struct btrfs_raid_bio *rbio;
+	struct btrfs_plug_cb *plug = NULL;
+	struct blk_plug_cb *cb;
+
+	rbio = alloc_rbio(root, bbio, raid_map, stripe_len);
+	if (IS_ERR(rbio))
+		return PTR_ERR(rbio);
+	bio_list_add(&rbio->bio_list, bio);
+	rbio->bio_list_bytes = bio->bi_iter.bi_size;
+
+	/*
+	 * don't plug on full rbios, just get them out the door
+	 * as quickly as we can
+	 */
+	if (rbio_is_full(rbio))
+		return full_stripe_write(rbio);
+
+	cb = blk_check_plugged(btrfs_raid_unplug, root->fs_info,
+			       sizeof(*plug));
+	if (cb) {
+		plug = container_of(cb, struct btrfs_plug_cb, cb);
+		if (!plug->info) {
+			plug->info = root->fs_info;
+			INIT_LIST_HEAD(&plug->rbio_list);
+		}
+		list_add_tail(&rbio->plug_list, &plug->rbio_list);
+	} else {
+		return __raid56_parity_write(rbio);
+	}
+	return 0;
+}
+
+/*
+ * all parity reconstruction happens here.  We've read in everything
+ * we can find from the drives and this does the heavy lifting of
+ * sorting the good from the bad.
+ */
+static void __raid_recover_end_io(struct btrfs_raid_bio *rbio)
+{
+	int pagenr, stripe;
+	void **pointers;
+	int faila = -1, failb = -1;
+	int nr_pages = (rbio->stripe_len + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
+	struct page *page;
+	int err;
+	int i;
+
+	pointers = kzalloc(rbio->bbio->num_stripes * sizeof(void *),
+			   GFP_NOFS);
+	if (!pointers) {
+		err = -ENOMEM;
+		goto cleanup_io;
+	}
+
+	faila = rbio->faila;
+	failb = rbio->failb;
+
+	if (rbio->read_rebuild) {
+		spin_lock_irq(&rbio->bio_list_lock);
+		set_bit(RBIO_RMW_LOCKED_BIT, &rbio->flags);
+		spin_unlock_irq(&rbio->bio_list_lock);
+	}
+
+	index_rbio_pages(rbio);
+
+	for (pagenr = 0; pagenr < nr_pages; pagenr++) {
+		/* setup our array of pointers with pages
+		 * from each stripe
+		 */
+		for (stripe = 0; stripe < rbio->bbio->num_stripes; stripe++) {
+			/*
+			 * if we're rebuilding a read, we have to use
+			 * pages from the bio list
+			 */
+			if (rbio->read_rebuild &&
+			    (stripe == faila || stripe == failb)) {
+				page = page_in_rbio(rbio, stripe, pagenr, 0);
+			} else {
+				page = rbio_stripe_page(rbio, stripe, pagenr);
+			}
+			pointers[stripe] = kmap(page);
+		}
+
+		/* all raid6 handling here */
+		if (rbio->raid_map[rbio->bbio->num_stripes - 1] ==
+		    RAID6_Q_STRIPE) {
+
+			/*
+			 * single failure, rebuild from parity raid5
+			 * style
+			 */
+			if (failb < 0) {
+				if (faila == rbio->nr_data) {
+					/*
+					 * Just the P stripe has failed, without
+					 * a bad data or Q stripe.
+					 * TODO, we should redo the xor here.
+					 */
+					err = -EIO;
+					goto cleanup;
+				}
+				/*
+				 * a single failure in raid6 is rebuilt
+				 * in the pstripe code below
+				 */
+				goto pstripe;
+			}
+
+			/* make sure our ps and qs are in order */
+			if (faila > failb) {
+				int tmp = failb;
+				failb = faila;
+				faila = tmp;
+			}
+
+			/* if the q stripe is failed, do a pstripe reconstruction
+			 * from the xors.
+			 * If both the q stripe and the P stripe are failed, we're
+			 * here due to a crc mismatch and we can't give them the
+			 * data they want
+			 */
+			if (rbio->raid_map[failb] == RAID6_Q_STRIPE) {
+				if (rbio->raid_map[faila] == RAID5_P_STRIPE) {
+					err = -EIO;
+					goto cleanup;
+				}
+				/*
+				 * otherwise we have one bad data stripe and
+				 * a good P stripe.  raid5!
+				 */
+				goto pstripe;
+			}
+
+			if (rbio->raid_map[failb] == RAID5_P_STRIPE) {
+				raid6_datap_recov(rbio->bbio->num_stripes,
+						  PAGE_SIZE, faila, pointers);
+			} else {
+				raid6_2data_recov(rbio->bbio->num_stripes,
+						  PAGE_SIZE, faila, failb,
+						  pointers);
+			}
+		} else {
+			void *p;
+
+			/* rebuild from P stripe here (raid5 or raid6) */
+			BUG_ON(failb != -1);
+pstripe:
+			/* Copy parity block into failed block to start with */
+			memcpy(pointers[faila],
+			       pointers[rbio->nr_data],
+			       PAGE_CACHE_SIZE);
+
+			/* rearrange the pointer array */
+			p = pointers[faila];
+			for (stripe = faila; stripe < rbio->nr_data - 1; stripe++)
+				pointers[stripe] = pointers[stripe + 1];
+			pointers[rbio->nr_data - 1] = p;
+
+			/* xor in the rest */
+			run_xor(pointers, rbio->nr_data - 1, PAGE_CACHE_SIZE);
+		}
+		/* if we're doing this rebuild as part of an rmw, go through
+		 * and set all of our private rbio pages in the
+		 * failed stripes as uptodate.  This way finish_rmw will
+		 * know they can be trusted.  If this was a read reconstruction,
+		 * other endio functions will fiddle the uptodate bits
+		 */
+		if (!rbio->read_rebuild) {
+			for (i = 0;  i < nr_pages; i++) {
+				if (faila != -1) {
+					page = rbio_stripe_page(rbio, faila, i);
+					SetPageUptodate(page);
+				}
+				if (failb != -1) {
+					page = rbio_stripe_page(rbio, failb, i);
+					SetPageUptodate(page);
+				}
+			}
+		}
+		for (stripe = 0; stripe < rbio->bbio->num_stripes; stripe++) {
+			/*
+			 * if we're rebuilding a read, we have to use
+			 * pages from the bio list
+			 */
+			if (rbio->read_rebuild &&
+			    (stripe == faila || stripe == failb)) {
+				page = page_in_rbio(rbio, stripe, pagenr, 0);
+			} else {
+				page = rbio_stripe_page(rbio, stripe, pagenr);
+			}
+			kunmap(page);
+		}
+	}
+
+	err = 0;
+cleanup:
+	kfree(pointers);
+
+cleanup_io:
+
+	if (rbio->read_rebuild) {
+		if (err == 0)
+			cache_rbio_pages(rbio);
+		else
+			clear_bit(RBIO_CACHE_READY_BIT, &rbio->flags);
+
+		rbio_orig_end_io(rbio, err, err == 0);
+	} else if (err == 0) {
+		rbio->faila = -1;
+		rbio->failb = -1;
+		finish_rmw(rbio);
+	} else {
+		rbio_orig_end_io(rbio, err, 0);
+	}
+}
+
+/*
+ * This is called only for stripes we've read from disk to
+ * reconstruct the parity.
+ */
+static void raid_recover_end_io(struct bio *bio, int err)
+{
+	struct btrfs_raid_bio *rbio = bio->bi_private;
+
+	/*
+	 * we only read stripe pages off the disk, set them
+	 * up to date if there were no errors
+	 */
+	if (err)
+		fail_bio_stripe(rbio, bio);
+	else
+		set_bio_pages_uptodate(bio);
+	bio_put(bio);
+
+	if (!atomic_dec_and_test(&rbio->bbio->stripes_pending))
+		return;
+
+	if (atomic_read(&rbio->bbio->error) > rbio->bbio->max_errors)
+		rbio_orig_end_io(rbio, -EIO, 0);
+	else
+		__raid_recover_end_io(rbio);
+}
+
+/*
+ * reads everything we need off the disk to reconstruct
+ * the parity. endio handlers trigger final reconstruction
+ * when the IO is done.
+ *
+ * This is used both for reads from the higher layers and for
+ * parity construction required to finish a rmw cycle.
+ */
+static int __raid56_parity_recover(struct btrfs_raid_bio *rbio)
+{
+	int bios_to_read = 0;
+	struct btrfs_bio *bbio = rbio->bbio;
+	struct bio_list bio_list;
+	int ret;
+	int nr_pages = (rbio->stripe_len + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
+	int pagenr;
+	int stripe;
+	struct bio *bio;
+
+	bio_list_init(&bio_list);
+
+	ret = alloc_rbio_pages(rbio);
+	if (ret)
+		goto cleanup;
+
+	atomic_set(&rbio->bbio->error, 0);
+
+	/*
+	 * read everything that hasn't failed.  Thanks to the
+	 * stripe cache, it is possible that some or all of these
+	 * pages are going to be uptodate.
+	 */
+	for (stripe = 0; stripe < bbio->num_stripes; stripe++) {
+		if (rbio->faila == stripe || rbio->failb == stripe) {
+			atomic_inc(&rbio->bbio->error);
+			continue;
+		}
+
+		for (pagenr = 0; pagenr < nr_pages; pagenr++) {
+			struct page *p;
+
+			/*
+			 * the rmw code may have already read this
+			 * page in
+			 */
+			p = rbio_stripe_page(rbio, stripe, pagenr);
+			if (PageUptodate(p))
+				continue;
+
+			ret = rbio_add_io_page(rbio, &bio_list,
+				       rbio_stripe_page(rbio, stripe, pagenr),
+				       stripe, pagenr, rbio->stripe_len);
+			if (ret < 0)
+				goto cleanup;
+		}
+	}
+
+	bios_to_read = bio_list_size(&bio_list);
+	if (!bios_to_read) {
+		/*
+		 * we might have no bios to read just because the pages
+		 * were up to date, or we might have no bios to read because
+		 * the devices were gone.
+		 */
+		if (atomic_read(&rbio->bbio->error) <= rbio->bbio->max_errors) {
+			__raid_recover_end_io(rbio);
+			goto out;
+		} else {
+			goto cleanup;
+		}
+	}
+
+	/*
+	 * the bbio may be freed once we submit the last bio.  Make sure
+	 * not to touch it after that
+	 */
+	atomic_set(&bbio->stripes_pending, bios_to_read);
+	while (1) {
+		bio = bio_list_pop(&bio_list);
+		if (!bio)
+			break;
+
+		bio->bi_private = rbio;
+		bio->bi_end_io = raid_recover_end_io;
+
+		btrfs_bio_wq_end_io(rbio->fs_info, bio,
+				    BTRFS_WQ_ENDIO_RAID56);
+
+		BUG_ON(!test_bit(BIO_UPTODATE, &bio->bi_flags));
+		submit_bio(READ, bio);
+	}
+out:
+	return 0;
+
+cleanup:
+	if (rbio->read_rebuild)
+		rbio_orig_end_io(rbio, -EIO, 0);
+	return -EIO;
+}
+
+/*
+ * the main entry point for reads from the higher layers.  This
+ * is really only called when the normal read path had a failure,
+ * so we assume the bio they send down corresponds to a failed part
+ * of the drive.
+ */
+int raid56_parity_recover(struct btrfs_root *root, struct bio *bio,
+			  struct btrfs_bio *bbio, u64 *raid_map,
+			  u64 stripe_len, int mirror_num)
+{
+	struct btrfs_raid_bio *rbio;
+	int ret;
+
+	rbio = alloc_rbio(root, bbio, raid_map, stripe_len);
+	if (IS_ERR(rbio))
+		return PTR_ERR(rbio);
+
+	rbio->read_rebuild = 1;
+	bio_list_add(&rbio->bio_list, bio);
+	rbio->bio_list_bytes = bio->bi_iter.bi_size;
+
+	rbio->faila = find_logical_bio_stripe(rbio, bio);
+	if (rbio->faila == -1) {
+		BUG();
+		kfree(raid_map);
+		kfree(bbio);
+		kfree(rbio);
+		return -EIO;
+	}
+
+	/*
+	 * reconstruct from the q stripe if they are
+	 * asking for mirror 3
+	 */
+	if (mirror_num == 3)
+		rbio->failb = bbio->num_stripes - 2;
+
+	ret = lock_stripe_add(rbio);
+
+	/*
+	 * __raid56_parity_recover will end the bio with
+	 * any errors it hits.  We don't want to return
+	 * its error value up the stack because our caller
+	 * will end up calling bio_endio with any nonzero
+	 * return
+	 */
+	if (ret == 0)
+		__raid56_parity_recover(rbio);
+	/*
+	 * our rbio has been added to the list of
+	 * rbios that will be handled after the
+	 * currently lock owner is done
+	 */
+	return 0;
+
+}
+
+static void rmw_work(struct btrfs_work *work)
+{
+	struct btrfs_raid_bio *rbio;
+
+	rbio = container_of(work, struct btrfs_raid_bio, work);
+	raid56_rmw_stripe(rbio);
+}
+
+static void read_rebuild_work(struct btrfs_work *work)
+{
+	struct btrfs_raid_bio *rbio;
+
+	rbio = container_of(work, struct btrfs_raid_bio, work);
+	__raid56_parity_recover(rbio);
+}
diff --git a/fs/btrfs/raid56.h b/fs/btrfs/raid56.h
new file mode 100644
index 00000000000..ea5d73bfdfb
--- /dev/null
+++ b/fs/btrfs/raid56.h
@@ -0,0 +1,51 @@
+/*
+ * Copyright (C) 2012 Fusion-io  All rights reserved.
+ * Copyright (C) 2012 Intel Corp. All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef __BTRFS_RAID56__
+#define __BTRFS_RAID56__
+static inline int nr_parity_stripes(struct map_lookup *map)
+{
+	if (map->type & BTRFS_BLOCK_GROUP_RAID5)
+		return 1;
+	else if (map->type & BTRFS_BLOCK_GROUP_RAID6)
+		return 2;
+	else
+		return 0;
+}
+
+static inline int nr_data_stripes(struct map_lookup *map)
+{
+	return map->num_stripes - nr_parity_stripes(map);
+}
+#define RAID5_P_STRIPE ((u64)-2)
+#define RAID6_Q_STRIPE ((u64)-1)
+
+#define is_parity_stripe(x) (((x) == RAID5_P_STRIPE) ||		\
+			     ((x) == RAID6_Q_STRIPE))
+
+int raid56_parity_recover(struct btrfs_root *root, struct bio *bio,
+				 struct btrfs_bio *bbio, u64 *raid_map,
+				 u64 stripe_len, int mirror_num);
+int raid56_parity_write(struct btrfs_root *root, struct bio *bio,
+			       struct btrfs_bio *bbio, u64 *raid_map,
+			       u64 stripe_len);
+
+int btrfs_alloc_stripe_hash_table(struct btrfs_fs_info *info);
+void btrfs_free_stripe_hash_table(struct btrfs_fs_info *info);
+#endif
diff --git a/fs/btrfs/rcu-string.h b/fs/btrfs/rcu-string.h
new file mode 100644
index 00000000000..9e111e4576d
--- /dev/null
+++ b/fs/btrfs/rcu-string.h
@@ -0,0 +1,56 @@
+/*
+ * Copyright (C) 2012 Red Hat.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+struct rcu_string {
+	struct rcu_head rcu;
+	char str[0];
+};
+
+static inline struct rcu_string *rcu_string_strdup(const char *src, gfp_t mask)
+{
+	size_t len = strlen(src) + 1;
+	struct rcu_string *ret = kzalloc(sizeof(struct rcu_string) +
+					 (len * sizeof(char)), mask);
+	if (!ret)
+		return ret;
+	strncpy(ret->str, src, len);
+	return ret;
+}
+
+static inline void rcu_string_free(struct rcu_string *str)
+{
+	if (str)
+		kfree_rcu(str, rcu);
+}
+
+#define printk_in_rcu(fmt, ...) do {	\
+	rcu_read_lock();		\
+	printk(fmt, __VA_ARGS__);	\
+	rcu_read_unlock();		\
+} while (0)
+
+#define printk_ratelimited_in_rcu(fmt, ...) do {	\
+	rcu_read_lock();				\
+	printk_ratelimited(fmt, __VA_ARGS__);		\
+	rcu_read_unlock();				\
+} while (0)
+
+#define rcu_str_deref(rcu_str) ({				\
+	struct rcu_string *__str = rcu_dereference(rcu_str);	\
+	__str->str;						\
+})
diff --git a/fs/btrfs/reada.c b/fs/btrfs/reada.c
new file mode 100644
index 00000000000..09230cf3a24
--- /dev/null
+++ b/fs/btrfs/reada.c
@@ -0,0 +1,994 @@
+/*
+ * Copyright (C) 2011 STRATO.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/sched.h>
+#include <linux/pagemap.h>
+#include <linux/writeback.h>
+#include <linux/blkdev.h>
+#include <linux/rbtree.h>
+#include <linux/slab.h>
+#include <linux/workqueue.h>
+#include "ctree.h"
+#include "volumes.h"
+#include "disk-io.h"
+#include "transaction.h"
+#include "dev-replace.h"
+
+#undef DEBUG
+
+/*
+ * This is the implementation for the generic read ahead framework.
+ *
+ * To trigger a readahead, btrfs_reada_add must be called. It will start
+ * a read ahead for the given range [start, end) on tree root. The returned
+ * handle can either be used to wait on the readahead to finish
+ * (btrfs_reada_wait), or to send it to the background (btrfs_reada_detach).
+ *
+ * The read ahead works as follows:
+ * On btrfs_reada_add, the root of the tree is inserted into a radix_tree.
+ * reada_start_machine will then search for extents to prefetch and trigger
+ * some reads. When a read finishes for a node, all contained node/leaf
+ * pointers that lie in the given range will also be enqueued. The reads will
+ * be triggered in sequential order, thus giving a big win over a naive
+ * enumeration. It will also make use of multi-device layouts. Each disk
+ * will have its on read pointer and all disks will by utilized in parallel.
+ * Also will no two disks read both sides of a mirror simultaneously, as this
+ * would waste seeking capacity. Instead both disks will read different parts
+ * of the filesystem.
+ * Any number of readaheads can be started in parallel. The read order will be
+ * determined globally, i.e. 2 parallel readaheads will normally finish faster
+ * than the 2 started one after another.
+ */
+
+#define MAX_IN_FLIGHT 6
+
+struct reada_extctl {
+	struct list_head	list;
+	struct reada_control	*rc;
+	u64			generation;
+};
+
+struct reada_extent {
+	u64			logical;
+	struct btrfs_key	top;
+	u32			blocksize;
+	int			err;
+	struct list_head	extctl;
+	int 			refcnt;
+	spinlock_t		lock;
+	struct reada_zone	*zones[BTRFS_MAX_MIRRORS];
+	int			nzones;
+	struct btrfs_device	*scheduled_for;
+};
+
+struct reada_zone {
+	u64			start;
+	u64			end;
+	u64			elems;
+	struct list_head	list;
+	spinlock_t		lock;
+	int			locked;
+	struct btrfs_device	*device;
+	struct btrfs_device	*devs[BTRFS_MAX_MIRRORS]; /* full list, incl
+							   * self */
+	int			ndevs;
+	struct kref		refcnt;
+};
+
+struct reada_machine_work {
+	struct btrfs_work	work;
+	struct btrfs_fs_info	*fs_info;
+};
+
+static void reada_extent_put(struct btrfs_fs_info *, struct reada_extent *);
+static void reada_control_release(struct kref *kref);
+static void reada_zone_release(struct kref *kref);
+static void reada_start_machine(struct btrfs_fs_info *fs_info);
+static void __reada_start_machine(struct btrfs_fs_info *fs_info);
+
+static int reada_add_block(struct reada_control *rc, u64 logical,
+			   struct btrfs_key *top, int level, u64 generation);
+
+/* recurses */
+/* in case of err, eb might be NULL */
+static int __readahead_hook(struct btrfs_root *root, struct extent_buffer *eb,
+			    u64 start, int err)
+{
+	int level = 0;
+	int nritems;
+	int i;
+	u64 bytenr;
+	u64 generation;
+	struct reada_extent *re;
+	struct btrfs_fs_info *fs_info = root->fs_info;
+	struct list_head list;
+	unsigned long index = start >> PAGE_CACHE_SHIFT;
+	struct btrfs_device *for_dev;
+
+	if (eb)
+		level = btrfs_header_level(eb);
+
+	/* find extent */
+	spin_lock(&fs_info->reada_lock);
+	re = radix_tree_lookup(&fs_info->reada_tree, index);
+	if (re)
+		re->refcnt++;
+	spin_unlock(&fs_info->reada_lock);
+
+	if (!re)
+		return -1;
+
+	spin_lock(&re->lock);
+	/*
+	 * just take the full list from the extent. afterwards we
+	 * don't need the lock anymore
+	 */
+	list_replace_init(&re->extctl, &list);
+	for_dev = re->scheduled_for;
+	re->scheduled_for = NULL;
+	spin_unlock(&re->lock);
+
+	if (err == 0) {
+		nritems = level ? btrfs_header_nritems(eb) : 0;
+		generation = btrfs_header_generation(eb);
+		/*
+		 * FIXME: currently we just set nritems to 0 if this is a leaf,
+		 * effectively ignoring the content. In a next step we could
+		 * trigger more readahead depending from the content, e.g.
+		 * fetch the checksums for the extents in the leaf.
+		 */
+	} else {
+		/*
+		 * this is the error case, the extent buffer has not been
+		 * read correctly. We won't access anything from it and
+		 * just cleanup our data structures. Effectively this will
+		 * cut the branch below this node from read ahead.
+		 */
+		nritems = 0;
+		generation = 0;
+	}
+
+	for (i = 0; i < nritems; i++) {
+		struct reada_extctl *rec;
+		u64 n_gen;
+		struct btrfs_key key;
+		struct btrfs_key next_key;
+
+		btrfs_node_key_to_cpu(eb, &key, i);
+		if (i + 1 < nritems)
+			btrfs_node_key_to_cpu(eb, &next_key, i + 1);
+		else
+			next_key = re->top;
+		bytenr = btrfs_node_blockptr(eb, i);
+		n_gen = btrfs_node_ptr_generation(eb, i);
+
+		list_for_each_entry(rec, &list, list) {
+			struct reada_control *rc = rec->rc;
+
+			/*
+			 * if the generation doesn't match, just ignore this
+			 * extctl. This will probably cut off a branch from
+			 * prefetch. Alternatively one could start a new (sub-)
+			 * prefetch for this branch, starting again from root.
+			 * FIXME: move the generation check out of this loop
+			 */
+#ifdef DEBUG
+			if (rec->generation != generation) {
+				btrfs_debug(root->fs_info,
+					   "generation mismatch for (%llu,%d,%llu) %llu != %llu",
+				       key.objectid, key.type, key.offset,
+				       rec->generation, generation);
+			}
+#endif
+			if (rec->generation == generation &&
+			    btrfs_comp_cpu_keys(&key, &rc->key_end) < 0 &&
+			    btrfs_comp_cpu_keys(&next_key, &rc->key_start) > 0)
+				reada_add_block(rc, bytenr, &next_key,
+						level - 1, n_gen);
+		}
+	}
+	/*
+	 * free extctl records
+	 */
+	while (!list_empty(&list)) {
+		struct reada_control *rc;
+		struct reada_extctl *rec;
+
+		rec = list_first_entry(&list, struct reada_extctl, list);
+		list_del(&rec->list);
+		rc = rec->rc;
+		kfree(rec);
+
+		kref_get(&rc->refcnt);
+		if (atomic_dec_and_test(&rc->elems)) {
+			kref_put(&rc->refcnt, reada_control_release);
+			wake_up(&rc->wait);
+		}
+		kref_put(&rc->refcnt, reada_control_release);
+
+		reada_extent_put(fs_info, re);	/* one ref for each entry */
+	}
+	reada_extent_put(fs_info, re);	/* our ref */
+	if (for_dev)
+		atomic_dec(&for_dev->reada_in_flight);
+
+	return 0;
+}
+
+/*
+ * start is passed separately in case eb in NULL, which may be the case with
+ * failed I/O
+ */
+int btree_readahead_hook(struct btrfs_root *root, struct extent_buffer *eb,
+			 u64 start, int err)
+{
+	int ret;
+
+	ret = __readahead_hook(root, eb, start, err);
+
+	reada_start_machine(root->fs_info);
+
+	return ret;
+}
+
+static struct reada_zone *reada_find_zone(struct btrfs_fs_info *fs_info,
+					  struct btrfs_device *dev, u64 logical,
+					  struct btrfs_bio *bbio)
+{
+	int ret;
+	struct reada_zone *zone;
+	struct btrfs_block_group_cache *cache = NULL;
+	u64 start;
+	u64 end;
+	int i;
+
+	zone = NULL;
+	spin_lock(&fs_info->reada_lock);
+	ret = radix_tree_gang_lookup(&dev->reada_zones, (void **)&zone,
+				     logical >> PAGE_CACHE_SHIFT, 1);
+	if (ret == 1)
+		kref_get(&zone->refcnt);
+	spin_unlock(&fs_info->reada_lock);
+
+	if (ret == 1) {
+		if (logical >= zone->start && logical < zone->end)
+			return zone;
+		spin_lock(&fs_info->reada_lock);
+		kref_put(&zone->refcnt, reada_zone_release);
+		spin_unlock(&fs_info->reada_lock);
+	}
+
+	cache = btrfs_lookup_block_group(fs_info, logical);
+	if (!cache)
+		return NULL;
+
+	start = cache->key.objectid;
+	end = start + cache->key.offset - 1;
+	btrfs_put_block_group(cache);
+
+	zone = kzalloc(sizeof(*zone), GFP_NOFS);
+	if (!zone)
+		return NULL;
+
+	zone->start = start;
+	zone->end = end;
+	INIT_LIST_HEAD(&zone->list);
+	spin_lock_init(&zone->lock);
+	zone->locked = 0;
+	kref_init(&zone->refcnt);
+	zone->elems = 0;
+	zone->device = dev; /* our device always sits at index 0 */
+	for (i = 0; i < bbio->num_stripes; ++i) {
+		/* bounds have already been checked */
+		zone->devs[i] = bbio->stripes[i].dev;
+	}
+	zone->ndevs = bbio->num_stripes;
+
+	spin_lock(&fs_info->reada_lock);
+	ret = radix_tree_insert(&dev->reada_zones,
+				(unsigned long)(zone->end >> PAGE_CACHE_SHIFT),
+				zone);
+
+	if (ret == -EEXIST) {
+		kfree(zone);
+		ret = radix_tree_gang_lookup(&dev->reada_zones, (void **)&zone,
+					     logical >> PAGE_CACHE_SHIFT, 1);
+		if (ret == 1)
+			kref_get(&zone->refcnt);
+	}
+	spin_unlock(&fs_info->reada_lock);
+
+	return zone;
+}
+
+static struct reada_extent *reada_find_extent(struct btrfs_root *root,
+					      u64 logical,
+					      struct btrfs_key *top, int level)
+{
+	int ret;
+	struct reada_extent *re = NULL;
+	struct reada_extent *re_exist = NULL;
+	struct btrfs_fs_info *fs_info = root->fs_info;
+	struct btrfs_bio *bbio = NULL;
+	struct btrfs_device *dev;
+	struct btrfs_device *prev_dev;
+	u32 blocksize;
+	u64 length;
+	int nzones = 0;
+	int i;
+	unsigned long index = logical >> PAGE_CACHE_SHIFT;
+	int dev_replace_is_ongoing;
+
+	spin_lock(&fs_info->reada_lock);
+	re = radix_tree_lookup(&fs_info->reada_tree, index);
+	if (re)
+		re->refcnt++;
+	spin_unlock(&fs_info->reada_lock);
+
+	if (re)
+		return re;
+
+	re = kzalloc(sizeof(*re), GFP_NOFS);
+	if (!re)
+		return NULL;
+
+	blocksize = btrfs_level_size(root, level);
+	re->logical = logical;
+	re->blocksize = blocksize;
+	re->top = *top;
+	INIT_LIST_HEAD(&re->extctl);
+	spin_lock_init(&re->lock);
+	re->refcnt = 1;
+
+	/*
+	 * map block
+	 */
+	length = blocksize;
+	ret = btrfs_map_block(fs_info, REQ_GET_READ_MIRRORS, logical, &length,
+			      &bbio, 0);
+	if (ret || !bbio || length < blocksize)
+		goto error;
+
+	if (bbio->num_stripes > BTRFS_MAX_MIRRORS) {
+		btrfs_err(root->fs_info,
+			   "readahead: more than %d copies not supported",
+			   BTRFS_MAX_MIRRORS);
+		goto error;
+	}
+
+	for (nzones = 0; nzones < bbio->num_stripes; ++nzones) {
+		struct reada_zone *zone;
+
+		dev = bbio->stripes[nzones].dev;
+		zone = reada_find_zone(fs_info, dev, logical, bbio);
+		if (!zone)
+			break;
+
+		re->zones[nzones] = zone;
+		spin_lock(&zone->lock);
+		if (!zone->elems)
+			kref_get(&zone->refcnt);
+		++zone->elems;
+		spin_unlock(&zone->lock);
+		spin_lock(&fs_info->reada_lock);
+		kref_put(&zone->refcnt, reada_zone_release);
+		spin_unlock(&fs_info->reada_lock);
+	}
+	re->nzones = nzones;
+	if (nzones == 0) {
+		/* not a single zone found, error and out */
+		goto error;
+	}
+
+	/* insert extent in reada_tree + all per-device trees, all or nothing */
+	btrfs_dev_replace_lock(&fs_info->dev_replace);
+	spin_lock(&fs_info->reada_lock);
+	ret = radix_tree_insert(&fs_info->reada_tree, index, re);
+	if (ret == -EEXIST) {
+		re_exist = radix_tree_lookup(&fs_info->reada_tree, index);
+		BUG_ON(!re_exist);
+		re_exist->refcnt++;
+		spin_unlock(&fs_info->reada_lock);
+		btrfs_dev_replace_unlock(&fs_info->dev_replace);
+		goto error;
+	}
+	if (ret) {
+		spin_unlock(&fs_info->reada_lock);
+		btrfs_dev_replace_unlock(&fs_info->dev_replace);
+		goto error;
+	}
+	prev_dev = NULL;
+	dev_replace_is_ongoing = btrfs_dev_replace_is_ongoing(
+			&fs_info->dev_replace);
+	for (i = 0; i < nzones; ++i) {
+		dev = bbio->stripes[i].dev;
+		if (dev == prev_dev) {
+			/*
+			 * in case of DUP, just add the first zone. As both
+			 * are on the same device, there's nothing to gain
+			 * from adding both.
+			 * Also, it wouldn't work, as the tree is per device
+			 * and adding would fail with EEXIST
+			 */
+			continue;
+		}
+		if (!dev->bdev) {
+			/*
+			 * cannot read ahead on missing device, but for RAID5/6,
+			 * REQ_GET_READ_MIRRORS return 1. So don't skip missing
+			 * device for such case.
+			 */
+			if (nzones > 1)
+				continue;
+		}
+		if (dev_replace_is_ongoing &&
+		    dev == fs_info->dev_replace.tgtdev) {
+			/*
+			 * as this device is selected for reading only as
+			 * a last resort, skip it for read ahead.
+			 */
+			continue;
+		}
+		prev_dev = dev;
+		ret = radix_tree_insert(&dev->reada_extents, index, re);
+		if (ret) {
+			while (--i >= 0) {
+				dev = bbio->stripes[i].dev;
+				BUG_ON(dev == NULL);
+				/* ignore whether the entry was inserted */
+				radix_tree_delete(&dev->reada_extents, index);
+			}
+			BUG_ON(fs_info == NULL);
+			radix_tree_delete(&fs_info->reada_tree, index);
+			spin_unlock(&fs_info->reada_lock);
+			btrfs_dev_replace_unlock(&fs_info->dev_replace);
+			goto error;
+		}
+	}
+	spin_unlock(&fs_info->reada_lock);
+	btrfs_dev_replace_unlock(&fs_info->dev_replace);
+
+	kfree(bbio);
+	return re;
+
+error:
+	while (nzones) {
+		struct reada_zone *zone;
+
+		--nzones;
+		zone = re->zones[nzones];
+		kref_get(&zone->refcnt);
+		spin_lock(&zone->lock);
+		--zone->elems;
+		if (zone->elems == 0) {
+			/*
+			 * no fs_info->reada_lock needed, as this can't be
+			 * the last ref
+			 */
+			kref_put(&zone->refcnt, reada_zone_release);
+		}
+		spin_unlock(&zone->lock);
+
+		spin_lock(&fs_info->reada_lock);
+		kref_put(&zone->refcnt, reada_zone_release);
+		spin_unlock(&fs_info->reada_lock);
+	}
+	kfree(bbio);
+	kfree(re);
+	return re_exist;
+}
+
+static void reada_extent_put(struct btrfs_fs_info *fs_info,
+			     struct reada_extent *re)
+{
+	int i;
+	unsigned long index = re->logical >> PAGE_CACHE_SHIFT;
+
+	spin_lock(&fs_info->reada_lock);
+	if (--re->refcnt) {
+		spin_unlock(&fs_info->reada_lock);
+		return;
+	}
+
+	radix_tree_delete(&fs_info->reada_tree, index);
+	for (i = 0; i < re->nzones; ++i) {
+		struct reada_zone *zone = re->zones[i];
+
+		radix_tree_delete(&zone->device->reada_extents, index);
+	}
+
+	spin_unlock(&fs_info->reada_lock);
+
+	for (i = 0; i < re->nzones; ++i) {
+		struct reada_zone *zone = re->zones[i];
+
+		kref_get(&zone->refcnt);
+		spin_lock(&zone->lock);
+		--zone->elems;
+		if (zone->elems == 0) {
+			/* no fs_info->reada_lock needed, as this can't be
+			 * the last ref */
+			kref_put(&zone->refcnt, reada_zone_release);
+		}
+		spin_unlock(&zone->lock);
+
+		spin_lock(&fs_info->reada_lock);
+		kref_put(&zone->refcnt, reada_zone_release);
+		spin_unlock(&fs_info->reada_lock);
+	}
+	if (re->scheduled_for)
+		atomic_dec(&re->scheduled_for->reada_in_flight);
+
+	kfree(re);
+}
+
+static void reada_zone_release(struct kref *kref)
+{
+	struct reada_zone *zone = container_of(kref, struct reada_zone, refcnt);
+
+	radix_tree_delete(&zone->device->reada_zones,
+			  zone->end >> PAGE_CACHE_SHIFT);
+
+	kfree(zone);
+}
+
+static void reada_control_release(struct kref *kref)
+{
+	struct reada_control *rc = container_of(kref, struct reada_control,
+						refcnt);
+
+	kfree(rc);
+}
+
+static int reada_add_block(struct reada_control *rc, u64 logical,
+			   struct btrfs_key *top, int level, u64 generation)
+{
+	struct btrfs_root *root = rc->root;
+	struct reada_extent *re;
+	struct reada_extctl *rec;
+
+	re = reada_find_extent(root, logical, top, level); /* takes one ref */
+	if (!re)
+		return -1;
+
+	rec = kzalloc(sizeof(*rec), GFP_NOFS);
+	if (!rec) {
+		reada_extent_put(root->fs_info, re);
+		return -1;
+	}
+
+	rec->rc = rc;
+	rec->generation = generation;
+	atomic_inc(&rc->elems);
+
+	spin_lock(&re->lock);
+	list_add_tail(&rec->list, &re->extctl);
+	spin_unlock(&re->lock);
+
+	/* leave the ref on the extent */
+
+	return 0;
+}
+
+/*
+ * called with fs_info->reada_lock held
+ */
+static void reada_peer_zones_set_lock(struct reada_zone *zone, int lock)
+{
+	int i;
+	unsigned long index = zone->end >> PAGE_CACHE_SHIFT;
+
+	for (i = 0; i < zone->ndevs; ++i) {
+		struct reada_zone *peer;
+		peer = radix_tree_lookup(&zone->devs[i]->reada_zones, index);
+		if (peer && peer->device != zone->device)
+			peer->locked = lock;
+	}
+}
+
+/*
+ * called with fs_info->reada_lock held
+ */
+static int reada_pick_zone(struct btrfs_device *dev)
+{
+	struct reada_zone *top_zone = NULL;
+	struct reada_zone *top_locked_zone = NULL;
+	u64 top_elems = 0;
+	u64 top_locked_elems = 0;
+	unsigned long index = 0;
+	int ret;
+
+	if (dev->reada_curr_zone) {
+		reada_peer_zones_set_lock(dev->reada_curr_zone, 0);
+		kref_put(&dev->reada_curr_zone->refcnt, reada_zone_release);
+		dev->reada_curr_zone = NULL;
+	}
+	/* pick the zone with the most elements */
+	while (1) {
+		struct reada_zone *zone;
+
+		ret = radix_tree_gang_lookup(&dev->reada_zones,
+					     (void **)&zone, index, 1);
+		if (ret == 0)
+			break;
+		index = (zone->end >> PAGE_CACHE_SHIFT) + 1;
+		if (zone->locked) {
+			if (zone->elems > top_locked_elems) {
+				top_locked_elems = zone->elems;
+				top_locked_zone = zone;
+			}
+		} else {
+			if (zone->elems > top_elems) {
+				top_elems = zone->elems;
+				top_zone = zone;
+			}
+		}
+	}
+	if (top_zone)
+		dev->reada_curr_zone = top_zone;
+	else if (top_locked_zone)
+		dev->reada_curr_zone = top_locked_zone;
+	else
+		return 0;
+
+	dev->reada_next = dev->reada_curr_zone->start;
+	kref_get(&dev->reada_curr_zone->refcnt);
+	reada_peer_zones_set_lock(dev->reada_curr_zone, 1);
+
+	return 1;
+}
+
+static int reada_start_machine_dev(struct btrfs_fs_info *fs_info,
+				   struct btrfs_device *dev)
+{
+	struct reada_extent *re = NULL;
+	int mirror_num = 0;
+	struct extent_buffer *eb = NULL;
+	u64 logical;
+	u32 blocksize;
+	int ret;
+	int i;
+	int need_kick = 0;
+
+	spin_lock(&fs_info->reada_lock);
+	if (dev->reada_curr_zone == NULL) {
+		ret = reada_pick_zone(dev);
+		if (!ret) {
+			spin_unlock(&fs_info->reada_lock);
+			return 0;
+		}
+	}
+	/*
+	 * FIXME currently we issue the reads one extent at a time. If we have
+	 * a contiguous block of extents, we could also coagulate them or use
+	 * plugging to speed things up
+	 */
+	ret = radix_tree_gang_lookup(&dev->reada_extents, (void **)&re,
+				     dev->reada_next >> PAGE_CACHE_SHIFT, 1);
+	if (ret == 0 || re->logical >= dev->reada_curr_zone->end) {
+		ret = reada_pick_zone(dev);
+		if (!ret) {
+			spin_unlock(&fs_info->reada_lock);
+			return 0;
+		}
+		re = NULL;
+		ret = radix_tree_gang_lookup(&dev->reada_extents, (void **)&re,
+					dev->reada_next >> PAGE_CACHE_SHIFT, 1);
+	}
+	if (ret == 0) {
+		spin_unlock(&fs_info->reada_lock);
+		return 0;
+	}
+	dev->reada_next = re->logical + re->blocksize;
+	re->refcnt++;
+
+	spin_unlock(&fs_info->reada_lock);
+
+	/*
+	 * find mirror num
+	 */
+	for (i = 0; i < re->nzones; ++i) {
+		if (re->zones[i]->device == dev) {
+			mirror_num = i + 1;
+			break;
+		}
+	}
+	logical = re->logical;
+	blocksize = re->blocksize;
+
+	spin_lock(&re->lock);
+	if (re->scheduled_for == NULL) {
+		re->scheduled_for = dev;
+		need_kick = 1;
+	}
+	spin_unlock(&re->lock);
+
+	reada_extent_put(fs_info, re);
+
+	if (!need_kick)
+		return 0;
+
+	atomic_inc(&dev->reada_in_flight);
+	ret = reada_tree_block_flagged(fs_info->extent_root, logical, blocksize,
+			 mirror_num, &eb);
+	if (ret)
+		__readahead_hook(fs_info->extent_root, NULL, logical, ret);
+	else if (eb)
+		__readahead_hook(fs_info->extent_root, eb, eb->start, ret);
+
+	if (eb)
+		free_extent_buffer(eb);
+
+	return 1;
+
+}
+
+static void reada_start_machine_worker(struct btrfs_work *work)
+{
+	struct reada_machine_work *rmw;
+	struct btrfs_fs_info *fs_info;
+	int old_ioprio;
+
+	rmw = container_of(work, struct reada_machine_work, work);
+	fs_info = rmw->fs_info;
+
+	kfree(rmw);
+
+	old_ioprio = IOPRIO_PRIO_VALUE(task_nice_ioclass(current),
+				       task_nice_ioprio(current));
+	set_task_ioprio(current, BTRFS_IOPRIO_READA);
+	__reada_start_machine(fs_info);
+	set_task_ioprio(current, old_ioprio);
+}
+
+static void __reada_start_machine(struct btrfs_fs_info *fs_info)
+{
+	struct btrfs_device *device;
+	struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
+	u64 enqueued;
+	u64 total = 0;
+	int i;
+
+	do {
+		enqueued = 0;
+		list_for_each_entry(device, &fs_devices->devices, dev_list) {
+			if (atomic_read(&device->reada_in_flight) <
+			    MAX_IN_FLIGHT)
+				enqueued += reada_start_machine_dev(fs_info,
+								    device);
+		}
+		total += enqueued;
+	} while (enqueued && total < 10000);
+
+	if (enqueued == 0)
+		return;
+
+	/*
+	 * If everything is already in the cache, this is effectively single
+	 * threaded. To a) not hold the caller for too long and b) to utilize
+	 * more cores, we broke the loop above after 10000 iterations and now
+	 * enqueue to workers to finish it. This will distribute the load to
+	 * the cores.
+	 */
+	for (i = 0; i < 2; ++i)
+		reada_start_machine(fs_info);
+}
+
+static void reada_start_machine(struct btrfs_fs_info *fs_info)
+{
+	struct reada_machine_work *rmw;
+
+	rmw = kzalloc(sizeof(*rmw), GFP_NOFS);
+	if (!rmw) {
+		/* FIXME we cannot handle this properly right now */
+		BUG();
+	}
+	btrfs_init_work(&rmw->work, reada_start_machine_worker, NULL, NULL);
+	rmw->fs_info = fs_info;
+
+	btrfs_queue_work(fs_info->readahead_workers, &rmw->work);
+}
+
+#ifdef DEBUG
+static void dump_devs(struct btrfs_fs_info *fs_info, int all)
+{
+	struct btrfs_device *device;
+	struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
+	unsigned long index;
+	int ret;
+	int i;
+	int j;
+	int cnt;
+
+	spin_lock(&fs_info->reada_lock);
+	list_for_each_entry(device, &fs_devices->devices, dev_list) {
+		printk(KERN_DEBUG "dev %lld has %d in flight\n", device->devid,
+			atomic_read(&device->reada_in_flight));
+		index = 0;
+		while (1) {
+			struct reada_zone *zone;
+			ret = radix_tree_gang_lookup(&device->reada_zones,
+						     (void **)&zone, index, 1);
+			if (ret == 0)
+				break;
+			printk(KERN_DEBUG "  zone %llu-%llu elems %llu locked "
+				"%d devs", zone->start, zone->end, zone->elems,
+				zone->locked);
+			for (j = 0; j < zone->ndevs; ++j) {
+				printk(KERN_CONT " %lld",
+					zone->devs[j]->devid);
+			}
+			if (device->reada_curr_zone == zone)
+				printk(KERN_CONT " curr off %llu",
+					device->reada_next - zone->start);
+			printk(KERN_CONT "\n");
+			index = (zone->end >> PAGE_CACHE_SHIFT) + 1;
+		}
+		cnt = 0;
+		index = 0;
+		while (all) {
+			struct reada_extent *re = NULL;
+
+			ret = radix_tree_gang_lookup(&device->reada_extents,
+						     (void **)&re, index, 1);
+			if (ret == 0)
+				break;
+			printk(KERN_DEBUG
+				"  re: logical %llu size %u empty %d for %lld",
+				re->logical, re->blocksize,
+				list_empty(&re->extctl), re->scheduled_for ?
+				re->scheduled_for->devid : -1);
+
+			for (i = 0; i < re->nzones; ++i) {
+				printk(KERN_CONT " zone %llu-%llu devs",
+					re->zones[i]->start,
+					re->zones[i]->end);
+				for (j = 0; j < re->zones[i]->ndevs; ++j) {
+					printk(KERN_CONT " %lld",
+						re->zones[i]->devs[j]->devid);
+				}
+			}
+			printk(KERN_CONT "\n");
+			index = (re->logical >> PAGE_CACHE_SHIFT) + 1;
+			if (++cnt > 15)
+				break;
+		}
+	}
+
+	index = 0;
+	cnt = 0;
+	while (all) {
+		struct reada_extent *re = NULL;
+
+		ret = radix_tree_gang_lookup(&fs_info->reada_tree, (void **)&re,
+					     index, 1);
+		if (ret == 0)
+			break;
+		if (!re->scheduled_for) {
+			index = (re->logical >> PAGE_CACHE_SHIFT) + 1;
+			continue;
+		}
+		printk(KERN_DEBUG
+			"re: logical %llu size %u list empty %d for %lld",
+			re->logical, re->blocksize, list_empty(&re->extctl),
+			re->scheduled_for ? re->scheduled_for->devid : -1);
+		for (i = 0; i < re->nzones; ++i) {
+			printk(KERN_CONT " zone %llu-%llu devs",
+				re->zones[i]->start,
+				re->zones[i]->end);
+			for (i = 0; i < re->nzones; ++i) {
+				printk(KERN_CONT " zone %llu-%llu devs",
+					re->zones[i]->start,
+					re->zones[i]->end);
+				for (j = 0; j < re->zones[i]->ndevs; ++j) {
+					printk(KERN_CONT " %lld",
+						re->zones[i]->devs[j]->devid);
+				}
+			}
+		}
+		printk(KERN_CONT "\n");
+		index = (re->logical >> PAGE_CACHE_SHIFT) + 1;
+	}
+	spin_unlock(&fs_info->reada_lock);
+}
+#endif
+
+/*
+ * interface
+ */
+struct reada_control *btrfs_reada_add(struct btrfs_root *root,
+			struct btrfs_key *key_start, struct btrfs_key *key_end)
+{
+	struct reada_control *rc;
+	u64 start;
+	u64 generation;
+	int level;
+	struct extent_buffer *node;
+	static struct btrfs_key max_key = {
+		.objectid = (u64)-1,
+		.type = (u8)-1,
+		.offset = (u64)-1
+	};
+
+	rc = kzalloc(sizeof(*rc), GFP_NOFS);
+	if (!rc)
+		return ERR_PTR(-ENOMEM);
+
+	rc->root = root;
+	rc->key_start = *key_start;
+	rc->key_end = *key_end;
+	atomic_set(&rc->elems, 0);
+	init_waitqueue_head(&rc->wait);
+	kref_init(&rc->refcnt);
+	kref_get(&rc->refcnt); /* one ref for having elements */
+
+	node = btrfs_root_node(root);
+	start = node->start;
+	level = btrfs_header_level(node);
+	generation = btrfs_header_generation(node);
+	free_extent_buffer(node);
+
+	if (reada_add_block(rc, start, &max_key, level, generation)) {
+		kfree(rc);
+		return ERR_PTR(-ENOMEM);
+	}
+
+	reada_start_machine(root->fs_info);
+
+	return rc;
+}
+
+#ifdef DEBUG
+int btrfs_reada_wait(void *handle)
+{
+	struct reada_control *rc = handle;
+
+	while (atomic_read(&rc->elems)) {
+		wait_event_timeout(rc->wait, atomic_read(&rc->elems) == 0,
+				   5 * HZ);
+		dump_devs(rc->root->fs_info,
+			  atomic_read(&rc->elems) < 10 ? 1 : 0);
+	}
+
+	dump_devs(rc->root->fs_info, atomic_read(&rc->elems) < 10 ? 1 : 0);
+
+	kref_put(&rc->refcnt, reada_control_release);
+
+	return 0;
+}
+#else
+int btrfs_reada_wait(void *handle)
+{
+	struct reada_control *rc = handle;
+
+	while (atomic_read(&rc->elems)) {
+		wait_event(rc->wait, atomic_read(&rc->elems) == 0);
+	}
+
+	kref_put(&rc->refcnt, reada_control_release);
+
+	return 0;
+}
+#endif
+
+void btrfs_reada_detach(void *handle)
+{
+	struct reada_control *rc = handle;
+
+	kref_put(&rc->refcnt, reada_control_release);
+}
diff --git a/fs/btrfs/ref-cache.c b/fs/btrfs/ref-cache.c
deleted file mode 100644
index a97314cf6bd..00000000000
--- a/fs/btrfs/ref-cache.c
+++ /dev/null
@@ -1,232 +0,0 @@
-/*
- * Copyright (C) 2008 Oracle.  All rights reserved.
- *
- * This program is free software; you can redistribute it and/or
- * modify it under the terms of the GNU General Public
- * License v2 as published by the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful,
- * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * General Public License for more details.
- *
- * You should have received a copy of the GNU General Public
- * License along with this program; if not, write to the
- * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
- * Boston, MA 021110-1307, USA.
- */
-
-#include <linux/sched.h>
-#include <linux/slab.h>
-#include <linux/sort.h>
-#include "ctree.h"
-#include "ref-cache.h"
-#include "transaction.h"
-
-/*
- * leaf refs are used to cache the information about which extents
- * a given leaf has references on.  This allows us to process that leaf
- * in btrfs_drop_snapshot without needing to read it back from disk.
- */
-
-/*
- * kmalloc a leaf reference struct and update the counters for the
- * total ref cache size
- */
-struct btrfs_leaf_ref *btrfs_alloc_leaf_ref(struct btrfs_root *root,
-					    int nr_extents)
-{
-	struct btrfs_leaf_ref *ref;
-	size_t size = btrfs_leaf_ref_size(nr_extents);
-
-	ref = kmalloc(size, GFP_NOFS);
-	if (ref) {
-		spin_lock(&root->fs_info->ref_cache_lock);
-		root->fs_info->total_ref_cache_size += size;
-		spin_unlock(&root->fs_info->ref_cache_lock);
-
-		memset(ref, 0, sizeof(*ref));
-		atomic_set(&ref->usage, 1);
-		INIT_LIST_HEAD(&ref->list);
-	}
-	return ref;
-}
-
-/*
- * free a leaf reference struct and update the counters for the
- * total ref cache size
- */
-void btrfs_free_leaf_ref(struct btrfs_root *root, struct btrfs_leaf_ref *ref)
-{
-	if (!ref)
-		return;
-	WARN_ON(atomic_read(&ref->usage) == 0);
-	if (atomic_dec_and_test(&ref->usage)) {
-		size_t size = btrfs_leaf_ref_size(ref->nritems);
-
-		BUG_ON(ref->in_tree);
-		kfree(ref);
-
-		spin_lock(&root->fs_info->ref_cache_lock);
-		root->fs_info->total_ref_cache_size -= size;
-		spin_unlock(&root->fs_info->ref_cache_lock);
-	}
-}
-
-static struct rb_node *tree_insert(struct rb_root *root, u64 bytenr,
-				   struct rb_node *node)
-{
-	struct rb_node **p = &root->rb_node;
-	struct rb_node *parent = NULL;
-	struct btrfs_leaf_ref *entry;
-
-	while (*p) {
-		parent = *p;
-		entry = rb_entry(parent, struct btrfs_leaf_ref, rb_node);
-
-		if (bytenr < entry->bytenr)
-			p = &(*p)->rb_left;
-		else if (bytenr > entry->bytenr)
-			p = &(*p)->rb_right;
-		else
-			return parent;
-	}
-
-	entry = rb_entry(node, struct btrfs_leaf_ref, rb_node);
-	rb_link_node(node, parent, p);
-	rb_insert_color(node, root);
-	return NULL;
-}
-
-static struct rb_node *tree_search(struct rb_root *root, u64 bytenr)
-{
-	struct rb_node *n = root->rb_node;
-	struct btrfs_leaf_ref *entry;
-
-	while (n) {
-		entry = rb_entry(n, struct btrfs_leaf_ref, rb_node);
-		WARN_ON(!entry->in_tree);
-
-		if (bytenr < entry->bytenr)
-			n = n->rb_left;
-		else if (bytenr > entry->bytenr)
-			n = n->rb_right;
-		else
-			return n;
-	}
-	return NULL;
-}
-
-int btrfs_remove_leaf_refs(struct btrfs_root *root, u64 max_root_gen,
-			   int shared)
-{
-	struct btrfs_leaf_ref *ref = NULL;
-	struct btrfs_leaf_ref_tree *tree = root->ref_tree;
-
-	if (shared)
-		tree = &root->fs_info->shared_ref_tree;
-	if (!tree)
-		return 0;
-
-	spin_lock(&tree->lock);
-	while (!list_empty(&tree->list)) {
-		ref = list_entry(tree->list.next, struct btrfs_leaf_ref, list);
-		BUG_ON(ref->tree != tree);
-		if (ref->root_gen > max_root_gen)
-			break;
-		if (!xchg(&ref->in_tree, 0)) {
-			cond_resched_lock(&tree->lock);
-			continue;
-		}
-
-		rb_erase(&ref->rb_node, &tree->root);
-		list_del_init(&ref->list);
-
-		spin_unlock(&tree->lock);
-		btrfs_free_leaf_ref(root, ref);
-		cond_resched();
-		spin_lock(&tree->lock);
-	}
-	spin_unlock(&tree->lock);
-	return 0;
-}
-
-/*
- * find the leaf ref for a given extent.  This returns the ref struct with
- * a usage reference incremented
- */
-struct btrfs_leaf_ref *btrfs_lookup_leaf_ref(struct btrfs_root *root,
-					     u64 bytenr)
-{
-	struct rb_node *rb;
-	struct btrfs_leaf_ref *ref = NULL;
-	struct btrfs_leaf_ref_tree *tree = root->ref_tree;
-again:
-	if (tree) {
-		spin_lock(&tree->lock);
-		rb = tree_search(&tree->root, bytenr);
-		if (rb)
-			ref = rb_entry(rb, struct btrfs_leaf_ref, rb_node);
-		if (ref)
-			atomic_inc(&ref->usage);
-		spin_unlock(&tree->lock);
-		if (ref)
-			return ref;
-	}
-	if (tree != &root->fs_info->shared_ref_tree) {
-		tree = &root->fs_info->shared_ref_tree;
-		goto again;
-	}
-	return NULL;
-}
-
-/*
- * add a fully filled in leaf ref struct
- * remove all the refs older than a given root generation
- */
-int btrfs_add_leaf_ref(struct btrfs_root *root, struct btrfs_leaf_ref *ref,
-		       int shared)
-{
-	int ret = 0;
-	struct rb_node *rb;
-	struct btrfs_leaf_ref_tree *tree = root->ref_tree;
-
-	if (shared)
-		tree = &root->fs_info->shared_ref_tree;
-
-	spin_lock(&tree->lock);
-	rb = tree_insert(&tree->root, ref->bytenr, &ref->rb_node);
-	if (rb) {
-		ret = -EEXIST;
-	} else {
-		atomic_inc(&ref->usage);
-		ref->tree = tree;
-		ref->in_tree = 1;
-		list_add_tail(&ref->list, &tree->list);
-	}
-	spin_unlock(&tree->lock);
-	return ret;
-}
-
-/*
- * remove a single leaf ref from the tree.  This drops the ref held by the tree
- * only
- */
-int btrfs_remove_leaf_ref(struct btrfs_root *root, struct btrfs_leaf_ref *ref)
-{
-	struct btrfs_leaf_ref_tree *tree;
-
-	if (!xchg(&ref->in_tree, 0))
-		return 0;
-
-	tree = ref->tree;
-	spin_lock(&tree->lock);
-
-	rb_erase(&ref->rb_node, &tree->root);
-	list_del_init(&ref->list);
-
-	spin_unlock(&tree->lock);
-
-	btrfs_free_leaf_ref(root, ref);
-	return 0;
-}
diff --git a/fs/btrfs/ref-cache.h b/fs/btrfs/ref-cache.h
deleted file mode 100644
index e2a55cb2072..00000000000
--- a/fs/btrfs/ref-cache.h
+++ /dev/null
@@ -1,76 +0,0 @@
-/*
- * Copyright (C) 2008 Oracle.  All rights reserved.
- *
- * This program is free software; you can redistribute it and/or
- * modify it under the terms of the GNU General Public
- * License v2 as published by the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful,
- * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * General Public License for more details.
- *
- * You should have received a copy of the GNU General Public
- * License along with this program; if not, write to the
- * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
- * Boston, MA 021110-1307, USA.
- */
-#ifndef __REFCACHE__
-#define __REFCACHE__
-
-struct btrfs_extent_info {
-	/* bytenr and num_bytes find the extent in the extent allocation tree */
-	u64 bytenr;
-	u64 num_bytes;
-
-	/* objectid and offset find the back reference for the file */
-	u64 objectid;
-	u64 offset;
-};
-
-struct btrfs_leaf_ref {
-	struct rb_node rb_node;
-	struct btrfs_leaf_ref_tree *tree;
-	int in_tree;
-	atomic_t usage;
-
-	u64 root_gen;
-	u64 bytenr;
-	u64 owner;
-	u64 generation;
-	int nritems;
-
-	struct list_head list;
-	struct btrfs_extent_info extents[];
-};
-
-static inline size_t btrfs_leaf_ref_size(int nr_extents)
-{
-	return sizeof(struct btrfs_leaf_ref) +
-	       sizeof(struct btrfs_extent_info) * nr_extents;
-}
-
-static inline void btrfs_leaf_ref_tree_init(struct btrfs_leaf_ref_tree *tree)
-{
-	tree->root = RB_ROOT;
-	INIT_LIST_HEAD(&tree->list);
-	spin_lock_init(&tree->lock);
-}
-
-static inline int btrfs_leaf_ref_tree_empty(struct btrfs_leaf_ref_tree *tree)
-{
-	return RB_EMPTY_ROOT(&tree->root);
-}
-
-void btrfs_leaf_ref_tree_init(struct btrfs_leaf_ref_tree *tree);
-struct btrfs_leaf_ref *btrfs_alloc_leaf_ref(struct btrfs_root *root,
-					    int nr_extents);
-void btrfs_free_leaf_ref(struct btrfs_root *root, struct btrfs_leaf_ref *ref);
-struct btrfs_leaf_ref *btrfs_lookup_leaf_ref(struct btrfs_root *root,
-					     u64 bytenr);
-int btrfs_add_leaf_ref(struct btrfs_root *root, struct btrfs_leaf_ref *ref,
-		       int shared);
-int btrfs_remove_leaf_refs(struct btrfs_root *root, u64 max_root_gen,
-			   int shared);
-int btrfs_remove_leaf_ref(struct btrfs_root *root, struct btrfs_leaf_ref *ref);
-#endif
diff --git a/fs/btrfs/relocation.c b/fs/btrfs/relocation.c
index e558dd941de..65245a07275 100644
--- a/fs/btrfs/relocation.c
+++ b/fs/btrfs/relocation.c
@@ -29,6 +29,8 @@
 #include "locking.h"
 #include "btrfs_inode.h"
 #include "async-thread.h"
+#include "free-space-cache.h"
+#include "inode-map.h"
 
 /*
  * backref_node, mapping_node and tree_block start with this
@@ -44,8 +46,12 @@ struct tree_entry {
 struct backref_node {
 	struct rb_node rb_node;
 	u64 bytenr;
-	/* objectid tree block owner */
+
+	u64 new_bytenr;
+	/* objectid of tree block owner, can be not uptodate */
 	u64 owner;
+	/* link to pending, changed or detached list */
+	struct list_head list;
 	/* list of upper level blocks reference this block */
 	struct list_head upper;
 	/* list of child blocks in the cache */
@@ -56,9 +62,9 @@ struct backref_node {
 	struct extent_buffer *eb;
 	/* level of tree block */
 	unsigned int level:8;
-	/* 1 if the block is root of old snapshot */
-	unsigned int old_root:1;
-	/* 1 if no child blocks in the cache */
+	/* is the block in non-reference counted tree */
+	unsigned int cowonly:1;
+	/* 1 if no child node in the cache */
 	unsigned int lowest:1;
 	/* is the extent buffer locked */
 	unsigned int locked:1;
@@ -66,6 +72,16 @@ struct backref_node {
 	unsigned int processed:1;
 	/* have backrefs of this block been checked */
 	unsigned int checked:1;
+	/*
+	 * 1 if corresponding block has been cowed but some upper
+	 * level block pointers may not point to the new location
+	 */
+	unsigned int pending:1;
+	/*
+	 * 1 if the backref node isn't connected to any other
+	 * backref node.
+	 */
+	unsigned int detached:1;
 };
 
 /*
@@ -74,18 +90,34 @@ struct backref_node {
 struct backref_edge {
 	struct list_head list[2];
 	struct backref_node *node[2];
-	u64 blockptr;
 };
 
 #define LOWER	0
 #define UPPER	1
+#define RELOCATION_RESERVED_NODES	256
 
 struct backref_cache {
 	/* red black tree of all backref nodes in the cache */
 	struct rb_root rb_root;
-	/* list of backref nodes with no child block in the cache */
+	/* for passing backref nodes to btrfs_reloc_cow_block */
+	struct backref_node *path[BTRFS_MAX_LEVEL];
+	/*
+	 * list of blocks that have been cowed but some block
+	 * pointers in upper level blocks may not reflect the
+	 * new location
+	 */
 	struct list_head pending[BTRFS_MAX_LEVEL];
-	spinlock_t lock;
+	/* list of backref nodes with no child node */
+	struct list_head leaves;
+	/* list of blocks that have been cowed in current transaction */
+	struct list_head changed;
+	/* list of detached backref node. */
+	struct list_head detached;
+
+	u64 last_trans;
+
+	int nr_nodes;
+	int nr_edges;
 };
 
 /*
@@ -113,15 +145,6 @@ struct tree_block {
 	unsigned int key_ready:1;
 };
 
-/* inode vector */
-#define INODEVEC_SIZE 16
-
-struct inodevec {
-	struct list_head list;
-	struct inode *inode[INODEVEC_SIZE];
-	int nr;
-};
-
 #define MAX_EXTENTS 128
 
 struct file_extent_cluster {
@@ -138,36 +161,42 @@ struct reloc_control {
 	struct btrfs_root *extent_root;
 	/* inode for moving data */
 	struct inode *data_inode;
-	struct btrfs_workers workers;
+
+	struct btrfs_block_rsv *block_rsv;
+
+	struct backref_cache backref_cache;
+
+	struct file_extent_cluster cluster;
 	/* tree blocks have been processed */
 	struct extent_io_tree processed_blocks;
 	/* map start of tree root to corresponding reloc tree */
 	struct mapping_tree reloc_root_tree;
 	/* list of reloc trees */
 	struct list_head reloc_roots;
+	/* size of metadata reservation for merging reloc trees */
+	u64 merging_rsv_size;
+	/* size of relocated tree nodes */
+	u64 nodes_relocated;
+	/* reserved size for block group relocation*/
+	u64 reserved_bytes;
+
 	u64 search_start;
 	u64 extents_found;
-	u64 extents_skipped;
-	int stage;
-	int create_reloc_root;
+
+	unsigned int stage:8;
+	unsigned int create_reloc_tree:1;
+	unsigned int merge_reloc_tree:1;
 	unsigned int found_file_extent:1;
-	unsigned int found_old_snapshot:1;
 };
 
 /* stages of data relocation */
 #define MOVE_DATA_EXTENTS	0
 #define UPDATE_DATA_PTRS	1
 
-/*
- * merge reloc tree to corresponding fs tree in worker threads
- */
-struct async_merge {
-	struct btrfs_work work;
-	struct reloc_control *rc;
-	struct btrfs_root *root;
-	struct completion *done;
-	atomic_t *num_pending;
-};
+static void remove_backref_node(struct backref_cache *cache,
+				struct backref_node *node);
+static void __mark_block_processed(struct reloc_control *rc,
+				   struct backref_node *node);
 
 static void mapping_tree_init(struct mapping_tree *tree)
 {
@@ -181,15 +210,80 @@ static void backref_cache_init(struct backref_cache *cache)
 	cache->rb_root = RB_ROOT;
 	for (i = 0; i < BTRFS_MAX_LEVEL; i++)
 		INIT_LIST_HEAD(&cache->pending[i]);
-	spin_lock_init(&cache->lock);
+	INIT_LIST_HEAD(&cache->changed);
+	INIT_LIST_HEAD(&cache->detached);
+	INIT_LIST_HEAD(&cache->leaves);
+}
+
+static void backref_cache_cleanup(struct backref_cache *cache)
+{
+	struct backref_node *node;
+	int i;
+
+	while (!list_empty(&cache->detached)) {
+		node = list_entry(cache->detached.next,
+				  struct backref_node, list);
+		remove_backref_node(cache, node);
+	}
+
+	while (!list_empty(&cache->leaves)) {
+		node = list_entry(cache->leaves.next,
+				  struct backref_node, lower);
+		remove_backref_node(cache, node);
+	}
+
+	cache->last_trans = 0;
+
+	for (i = 0; i < BTRFS_MAX_LEVEL; i++)
+		BUG_ON(!list_empty(&cache->pending[i]));
+	BUG_ON(!list_empty(&cache->changed));
+	BUG_ON(!list_empty(&cache->detached));
+	BUG_ON(!RB_EMPTY_ROOT(&cache->rb_root));
+	BUG_ON(cache->nr_nodes);
+	BUG_ON(cache->nr_edges);
+}
+
+static struct backref_node *alloc_backref_node(struct backref_cache *cache)
+{
+	struct backref_node *node;
+
+	node = kzalloc(sizeof(*node), GFP_NOFS);
+	if (node) {
+		INIT_LIST_HEAD(&node->list);
+		INIT_LIST_HEAD(&node->upper);
+		INIT_LIST_HEAD(&node->lower);
+		RB_CLEAR_NODE(&node->rb_node);
+		cache->nr_nodes++;
+	}
+	return node;
+}
+
+static void free_backref_node(struct backref_cache *cache,
+			      struct backref_node *node)
+{
+	if (node) {
+		cache->nr_nodes--;
+		kfree(node);
+	}
 }
 
-static void backref_node_init(struct backref_node *node)
+static struct backref_edge *alloc_backref_edge(struct backref_cache *cache)
 {
-	memset(node, 0, sizeof(*node));
-	INIT_LIST_HEAD(&node->upper);
-	INIT_LIST_HEAD(&node->lower);
-	RB_CLEAR_NODE(&node->rb_node);
+	struct backref_edge *edge;
+
+	edge = kzalloc(sizeof(*edge), GFP_NOFS);
+	if (edge)
+		cache->nr_edges++;
+	return edge;
+}
+
+static void free_backref_edge(struct backref_cache *cache,
+			      struct backref_edge *edge)
+{
+	if (edge) {
+		cache->nr_edges--;
+		kfree(edge);
+	}
 }
 
 static struct rb_node *tree_insert(struct rb_root *root, u64 bytenr,
@@ -234,6 +328,18 @@ static struct rb_node *tree_search(struct rb_root *root, u64 bytenr)
 	return NULL;
 }
 
+static void backref_tree_panic(struct rb_node *rb_node, int errno, u64 bytenr)
+{
+
+	struct btrfs_fs_info *fs_info = NULL;
+	struct backref_node *bnode = rb_entry(rb_node, struct backref_node,
+					      rb_node);
+	if (bnode->root)
+		fs_info = bnode->root->fs_info;
+	btrfs_panic(fs_info, errno, "Inconsistency in backref cache "
+		    "found at offset %llu", bytenr);
+}
+
 /*
  * walk up backref nodes until reach node presents tree root
  */
@@ -250,6 +356,7 @@ static struct backref_node *walk_up_backref(struct backref_node *node,
 		edges[idx++] = edge;
 		node = edge->node[UPPER];
 	}
+	BUG_ON(node->detached);
 	*index = idx;
 	return node;
 }
@@ -281,13 +388,18 @@ static struct backref_node *walk_down_backref(struct backref_edge *edges[],
 	return NULL;
 }
 
+static void unlock_node_buffer(struct backref_node *node)
+{
+	if (node->locked) {
+		btrfs_tree_unlock(node->eb);
+		node->locked = 0;
+	}
+}
+
 static void drop_node_buffer(struct backref_node *node)
 {
 	if (node->eb) {
-		if (node->locked) {
-			btrfs_tree_unlock(node->eb);
-			node->locked = 0;
-		}
+		unlock_node_buffer(node);
 		free_extent_buffer(node->eb);
 		node->eb = NULL;
 	}
@@ -296,14 +408,14 @@ static void drop_node_buffer(struct backref_node *node)
 static void drop_backref_node(struct backref_cache *tree,
 			      struct backref_node *node)
 {
-	BUG_ON(!node->lowest);
 	BUG_ON(!list_empty(&node->upper));
 
 	drop_node_buffer(node);
+	list_del(&node->list);
 	list_del(&node->lower);
-
-	rb_erase(&node->rb_node, &tree->rb_root);
-	kfree(node);
+	if (!RB_EMPTY_NODE(&node->rb_node))
+		rb_erase(&node->rb_node, &tree->rb_root);
+	free_backref_node(tree, node);
 }
 
 /*
@@ -318,27 +430,122 @@ static void remove_backref_node(struct backref_cache *cache,
 	if (!node)
 		return;
 
-	BUG_ON(!node->lowest);
+	BUG_ON(!node->lowest && !node->detached);
 	while (!list_empty(&node->upper)) {
 		edge = list_entry(node->upper.next, struct backref_edge,
 				  list[LOWER]);
 		upper = edge->node[UPPER];
 		list_del(&edge->list[LOWER]);
 		list_del(&edge->list[UPPER]);
-		kfree(edge);
+		free_backref_edge(cache, edge);
+
+		if (RB_EMPTY_NODE(&upper->rb_node)) {
+			BUG_ON(!list_empty(&node->upper));
+			drop_backref_node(cache, node);
+			node = upper;
+			node->lowest = 1;
+			continue;
+		}
 		/*
-		 * add the node to pending list if no other
+		 * add the node to leaf node list if no other
 		 * child block cached.
 		 */
 		if (list_empty(&upper->lower)) {
-			list_add_tail(&upper->lower,
-				      &cache->pending[upper->level]);
+			list_add_tail(&upper->lower, &cache->leaves);
 			upper->lowest = 1;
 		}
 	}
+
 	drop_backref_node(cache, node);
 }
 
+static void update_backref_node(struct backref_cache *cache,
+				struct backref_node *node, u64 bytenr)
+{
+	struct rb_node *rb_node;
+	rb_erase(&node->rb_node, &cache->rb_root);
+	node->bytenr = bytenr;
+	rb_node = tree_insert(&cache->rb_root, node->bytenr, &node->rb_node);
+	if (rb_node)
+		backref_tree_panic(rb_node, -EEXIST, bytenr);
+}
+
+/*
+ * update backref cache after a transaction commit
+ */
+static int update_backref_cache(struct btrfs_trans_handle *trans,
+				struct backref_cache *cache)
+{
+	struct backref_node *node;
+	int level = 0;
+
+	if (cache->last_trans == 0) {
+		cache->last_trans = trans->transid;
+		return 0;
+	}
+
+	if (cache->last_trans == trans->transid)
+		return 0;
+
+	/*
+	 * detached nodes are used to avoid unnecessary backref
+	 * lookup. transaction commit changes the extent tree.
+	 * so the detached nodes are no longer useful.
+	 */
+	while (!list_empty(&cache->detached)) {
+		node = list_entry(cache->detached.next,
+				  struct backref_node, list);
+		remove_backref_node(cache, node);
+	}
+
+	while (!list_empty(&cache->changed)) {
+		node = list_entry(cache->changed.next,
+				  struct backref_node, list);
+		list_del_init(&node->list);
+		BUG_ON(node->pending);
+		update_backref_node(cache, node, node->new_bytenr);
+	}
+
+	/*
+	 * some nodes can be left in the pending list if there were
+	 * errors during processing the pending nodes.
+	 */
+	for (level = 0; level < BTRFS_MAX_LEVEL; level++) {
+		list_for_each_entry(node, &cache->pending[level], list) {
+			BUG_ON(!node->pending);
+			if (node->bytenr == node->new_bytenr)
+				continue;
+			update_backref_node(cache, node, node->new_bytenr);
+		}
+	}
+
+	cache->last_trans = 0;
+	return 1;
+}
+
+
+static int should_ignore_root(struct btrfs_root *root)
+{
+	struct btrfs_root *reloc_root;
+
+	if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
+		return 0;
+
+	reloc_root = root->reloc_root;
+	if (!reloc_root)
+		return 0;
+
+	if (btrfs_root_last_snapshot(&reloc_root->root_item) ==
+	    root->fs_info->running_transaction->transid - 1)
+		return 0;
+	/*
+	 * if there is reloc tree and it was created in previous
+	 * transaction backref lookup can find the reloc tree,
+	 * so backref node for the fs tree root is useless for
+	 * relocation.
+	 */
+	return 1;
+}
 /*
  * find reloc tree by address of tree root
  */
@@ -366,7 +573,9 @@ static int is_cowonly_root(u64 root_objectid)
 	    root_objectid == BTRFS_CHUNK_TREE_OBJECTID ||
 	    root_objectid == BTRFS_DEV_TREE_OBJECTID ||
 	    root_objectid == BTRFS_TREE_LOG_OBJECTID ||
-	    root_objectid == BTRFS_CSUM_TREE_OBJECTID)
+	    root_objectid == BTRFS_CSUM_TREE_OBJECTID ||
+	    root_objectid == BTRFS_UUID_TREE_OBJECTID ||
+	    root_objectid == BTRFS_QUOTA_TREE_OBJECTID)
 		return 1;
 	return 0;
 }
@@ -383,7 +592,7 @@ static struct btrfs_root *read_fs_root(struct btrfs_fs_info *fs_info,
 	else
 		key.offset = (u64)-1;
 
-	return btrfs_read_fs_root_no_name(fs_info, &key);
+	return btrfs_get_fs_root(fs_info, &key, false);
 }
 
 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
@@ -401,7 +610,7 @@ struct btrfs_root *find_tree_root(struct reloc_control *rc,
 	root = read_fs_root(rc->extent_root->fs_info, root_objectid);
 	BUG_ON(IS_ERR(root));
 
-	if (root->ref_cows &&
+	if (test_bit(BTRFS_ROOT_REF_COWS, &root->state) &&
 	    generation != btrfs_root_generation(&root->root_item))
 		return NULL;
 
@@ -413,10 +622,13 @@ static noinline_for_stack
 int find_inline_backref(struct extent_buffer *leaf, int slot,
 			unsigned long *ptr, unsigned long *end)
 {
+	struct btrfs_key key;
 	struct btrfs_extent_item *ei;
 	struct btrfs_tree_block_info *bi;
 	u32 item_size;
 
+	btrfs_item_key_to_cpu(leaf, &key, slot);
+
 	item_size = btrfs_item_size_nr(leaf, slot);
 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
 	if (item_size < sizeof(*ei)) {
@@ -428,13 +640,23 @@ int find_inline_backref(struct extent_buffer *leaf, int slot,
 	WARN_ON(!(btrfs_extent_flags(leaf, ei) &
 		  BTRFS_EXTENT_FLAG_TREE_BLOCK));
 
-	if (item_size <= sizeof(*ei) + sizeof(*bi)) {
+	if (key.type == BTRFS_EXTENT_ITEM_KEY &&
+	    item_size <= sizeof(*ei) + sizeof(*bi)) {
 		WARN_ON(item_size < sizeof(*ei) + sizeof(*bi));
 		return 1;
 	}
+	if (key.type == BTRFS_METADATA_ITEM_KEY &&
+	    item_size <= sizeof(*ei)) {
+		WARN_ON(item_size < sizeof(*ei));
+		return 1;
+	}
 
-	bi = (struct btrfs_tree_block_info *)(ei + 1);
-	*ptr = (unsigned long)(bi + 1);
+	if (key.type == BTRFS_EXTENT_ITEM_KEY) {
+		bi = (struct btrfs_tree_block_info *)(ei + 1);
+		*ptr = (unsigned long)(bi + 1);
+	} else {
+		*ptr = (unsigned long)(ei + 1);
+	}
 	*end = (unsigned long)ei + item_size;
 	return 0;
 }
@@ -453,11 +675,12 @@ int find_inline_backref(struct extent_buffer *leaf, int slot,
  * for all upper level blocks that directly/indirectly reference the
  * block are also cached.
  */
-static struct backref_node *build_backref_tree(struct reloc_control *rc,
-					       struct backref_cache *cache,
-					       struct btrfs_key *node_key,
-					       int level, u64 bytenr)
+static noinline_for_stack
+struct backref_node *build_backref_tree(struct reloc_control *rc,
+					struct btrfs_key *node_key,
+					int level, u64 bytenr)
 {
+	struct backref_cache *cache = &rc->backref_cache;
 	struct btrfs_path *path1;
 	struct btrfs_path *path2;
 	struct extent_buffer *eb;
@@ -473,8 +696,11 @@ static struct backref_node *build_backref_tree(struct reloc_control *rc,
 	unsigned long end;
 	unsigned long ptr;
 	LIST_HEAD(list);
+	LIST_HEAD(useless);
+	int cowonly;
 	int ret;
 	int err = 0;
+	bool need_check = true;
 
 	path1 = btrfs_alloc_path();
 	path2 = btrfs_alloc_path();
@@ -482,16 +708,16 @@ static struct backref_node *build_backref_tree(struct reloc_control *rc,
 		err = -ENOMEM;
 		goto out;
 	}
+	path1->reada = 1;
+	path2->reada = 2;
 
-	node = kmalloc(sizeof(*node), GFP_NOFS);
+	node = alloc_backref_node(cache);
 	if (!node) {
 		err = -ENOMEM;
 		goto out;
 	}
 
-	backref_node_init(node);
 	node->bytenr = bytenr;
-	node->owner = 0;
 	node->level = level;
 	node->lowest = 1;
 	cur = node;
@@ -499,7 +725,7 @@ again:
 	end = 0;
 	ptr = 0;
 	key.objectid = cur->bytenr;
-	key.type = BTRFS_EXTENT_ITEM_KEY;
+	key.type = BTRFS_METADATA_ITEM_KEY;
 	key.offset = (u64)-1;
 
 	path1->search_commit_root = 1;
@@ -517,7 +743,7 @@ again:
 	WARN_ON(cur->checked);
 	if (!list_empty(&cur->upper)) {
 		/*
-		 * the backref was added previously when processsing
+		 * the backref was added previously when processing
 		 * backref of type BTRFS_TREE_BLOCK_REF_KEY
 		 */
 		BUG_ON(!list_is_singular(&cur->upper));
@@ -557,7 +783,8 @@ again:
 				break;
 			}
 
-			if (key.type == BTRFS_EXTENT_ITEM_KEY) {
+			if (key.type == BTRFS_EXTENT_ITEM_KEY ||
+			    key.type == BTRFS_METADATA_ITEM_KEY) {
 				ret = find_inline_backref(eb, path1->slots[0],
 							  &ptr, &end);
 				if (ret)
@@ -587,17 +814,21 @@ again:
 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
 		if (key.type == BTRFS_SHARED_BLOCK_REF_KEY ||
 		    key.type == BTRFS_EXTENT_REF_V0_KEY) {
-			if (key.objectid == key.offset &&
-			    key.type == BTRFS_EXTENT_REF_V0_KEY) {
+			if (key.type == BTRFS_EXTENT_REF_V0_KEY) {
 				struct btrfs_extent_ref_v0 *ref0;
 				ref0 = btrfs_item_ptr(eb, path1->slots[0],
 						struct btrfs_extent_ref_v0);
-				root = find_tree_root(rc, eb, ref0);
-				if (root)
-					cur->root = root;
-				else
-					cur->old_root = 1;
-				break;
+				if (key.objectid == key.offset) {
+					root = find_tree_root(rc, eb, ref0);
+					if (root && !should_ignore_root(root))
+						cur->root = root;
+					else
+						list_add(&cur->list, &useless);
+					break;
+				}
+				if (is_cowonly_root(btrfs_ref_root_v0(eb,
+								      ref0)))
+					cur->cowonly = 1;
 			}
 #else
 		BUG_ON(key.type == BTRFS_EXTENT_REF_V0_KEY);
@@ -614,22 +845,20 @@ again:
 				break;
 			}
 
-			edge = kzalloc(sizeof(*edge), GFP_NOFS);
+			edge = alloc_backref_edge(cache);
 			if (!edge) {
 				err = -ENOMEM;
 				goto out;
 			}
 			rb_node = tree_search(&cache->rb_root, key.offset);
 			if (!rb_node) {
-				upper = kmalloc(sizeof(*upper), GFP_NOFS);
+				upper = alloc_backref_node(cache);
 				if (!upper) {
-					kfree(edge);
+					free_backref_edge(cache, edge);
 					err = -ENOMEM;
 					goto out;
 				}
-				backref_node_init(upper);
 				upper->bytenr = key.offset;
-				upper->owner = 0;
 				upper->level = cur->level + 1;
 				/*
 				 *  backrefs for the upper level block isn't
@@ -639,11 +868,12 @@ again:
 			} else {
 				upper = rb_entry(rb_node, struct backref_node,
 						 rb_node);
+				BUG_ON(!upper->checked);
 				INIT_LIST_HEAD(&edge->list[UPPER]);
 			}
-			list_add(&edge->list[LOWER], &cur->upper);
-			edge->node[UPPER] = upper;
+			list_add_tail(&edge->list[LOWER], &cur->upper);
 			edge->node[LOWER] = cur;
+			edge->node[UPPER] = upper;
 
 			goto next;
 		} else if (key.type != BTRFS_TREE_BLOCK_REF_KEY) {
@@ -657,11 +887,17 @@ again:
 			goto out;
 		}
 
+		if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
+			cur->cowonly = 1;
+
 		if (btrfs_root_level(&root->root_item) == cur->level) {
 			/* tree root */
 			BUG_ON(btrfs_root_bytenr(&root->root_item) !=
 			       cur->bytenr);
-			cur->root = root;
+			if (should_ignore_root(root))
+				list_add(&cur->list, &useless);
+			else
+				cur->root = root;
 			break;
 		}
 
@@ -688,15 +924,19 @@ again:
 			cur->bytenr);
 
 		lower = cur;
+		need_check = true;
 		for (; level < BTRFS_MAX_LEVEL; level++) {
 			if (!path2->nodes[level]) {
 				BUG_ON(btrfs_root_bytenr(&root->root_item) !=
 				       lower->bytenr);
-				lower->root = root;
+				if (should_ignore_root(root))
+					list_add(&lower->list, &useless);
+				else
+					lower->root = root;
 				break;
 			}
 
-			edge = kzalloc(sizeof(*edge), GFP_NOFS);
+			edge = alloc_backref_edge(cache);
 			if (!edge) {
 				err = -ENOMEM;
 				goto out;
@@ -705,16 +945,18 @@ again:
 			eb = path2->nodes[level];
 			rb_node = tree_search(&cache->rb_root, eb->start);
 			if (!rb_node) {
-				upper = kmalloc(sizeof(*upper), GFP_NOFS);
+				upper = alloc_backref_node(cache);
 				if (!upper) {
-					kfree(edge);
+					free_backref_edge(cache, edge);
 					err = -ENOMEM;
 					goto out;
 				}
-				backref_node_init(upper);
 				upper->bytenr = eb->start;
 				upper->owner = btrfs_header_owner(eb);
 				upper->level = lower->level + 1;
+				if (!test_bit(BTRFS_ROOT_REF_COWS,
+					      &root->state))
+					upper->cowonly = 1;
 
 				/*
 				 * if we know the block isn't shared
@@ -727,14 +969,12 @@ again:
 
 				/*
 				 * add the block to pending list if we
-				 * need check its backrefs. only block
-				 * at 'cur->level + 1' is added to the
-				 * tail of pending list. this guarantees
-				 * we check backrefs from lower level
-				 * blocks to upper level blocks.
+				 * need check its backrefs, we only do this once
+				 * while walking up a tree as we will catch
+				 * anything else later on.
 				 */
-				if (!upper->checked &&
-				    level == cur->level + 1) {
+				if (!upper->checked && need_check) {
+					need_check = false;
 					list_add_tail(&edge->list[UPPER],
 						      &list);
 				} else
@@ -744,17 +984,19 @@ again:
 						 rb_node);
 				BUG_ON(!upper->checked);
 				INIT_LIST_HEAD(&edge->list[UPPER]);
+				if (!upper->owner)
+					upper->owner = btrfs_header_owner(eb);
 			}
 			list_add_tail(&edge->list[LOWER], &lower->upper);
-			edge->node[UPPER] = upper;
 			edge->node[LOWER] = lower;
+			edge->node[UPPER] = upper;
 
 			if (rb_node)
 				break;
 			lower = upper;
 			upper = NULL;
 		}
-		btrfs_release_path(root, path2);
+		btrfs_release_path(path2);
 next:
 		if (ptr < end) {
 			ptr += btrfs_extent_inline_ref_size(key.type);
@@ -767,7 +1009,7 @@ next:
 		if (ptr >= end)
 			path1->slots[0]++;
 	}
-	btrfs_release_path(rc->extent_root, path1);
+	btrfs_release_path(path1);
 
 	cur->checked = 1;
 	WARN_ON(exist);
@@ -785,8 +1027,14 @@ next:
 	 * into the cache.
 	 */
 	BUG_ON(!node->checked);
-	rb_node = tree_insert(&cache->rb_root, node->bytenr, &node->rb_node);
-	BUG_ON(rb_node);
+	cowonly = node->cowonly;
+	if (!cowonly) {
+		rb_node = tree_insert(&cache->rb_root, node->bytenr,
+				      &node->rb_node);
+		if (rb_node)
+			backref_tree_panic(rb_node, -EEXIST, node->bytenr);
+		list_add_tail(&node->lower, &cache->leaves);
+	}
 
 	list_for_each_entry(edge, &node->upper, list[LOWER])
 		list_add_tail(&edge->list[UPPER], &list);
@@ -795,6 +1043,14 @@ next:
 		edge = list_entry(list.next, struct backref_edge, list[UPPER]);
 		list_del_init(&edge->list[UPPER]);
 		upper = edge->node[UPPER];
+		if (upper->detached) {
+			list_del(&edge->list[LOWER]);
+			lower = edge->node[LOWER];
+			free_backref_edge(cache, edge);
+			if (list_empty(&lower->upper))
+				list_add(&lower->list, &useless);
+			continue;
+		}
 
 		if (!RB_EMPTY_NODE(&upper->rb_node)) {
 			if (upper->lowest) {
@@ -807,25 +1063,71 @@ next:
 		}
 
 		BUG_ON(!upper->checked);
-		rb_node = tree_insert(&cache->rb_root, upper->bytenr,
-				      &upper->rb_node);
-		BUG_ON(rb_node);
+		BUG_ON(cowonly != upper->cowonly);
+		if (!cowonly) {
+			rb_node = tree_insert(&cache->rb_root, upper->bytenr,
+					      &upper->rb_node);
+			if (rb_node)
+				backref_tree_panic(rb_node, -EEXIST,
+						   upper->bytenr);
+		}
 
 		list_add_tail(&edge->list[UPPER], &upper->lower);
 
 		list_for_each_entry(edge, &upper->upper, list[LOWER])
 			list_add_tail(&edge->list[UPPER], &list);
 	}
+	/*
+	 * process useless backref nodes. backref nodes for tree leaves
+	 * are deleted from the cache. backref nodes for upper level
+	 * tree blocks are left in the cache to avoid unnecessary backref
+	 * lookup.
+	 */
+	while (!list_empty(&useless)) {
+		upper = list_entry(useless.next, struct backref_node, list);
+		list_del_init(&upper->list);
+		BUG_ON(!list_empty(&upper->upper));
+		if (upper == node)
+			node = NULL;
+		if (upper->lowest) {
+			list_del_init(&upper->lower);
+			upper->lowest = 0;
+		}
+		while (!list_empty(&upper->lower)) {
+			edge = list_entry(upper->lower.next,
+					  struct backref_edge, list[UPPER]);
+			list_del(&edge->list[UPPER]);
+			list_del(&edge->list[LOWER]);
+			lower = edge->node[LOWER];
+			free_backref_edge(cache, edge);
+
+			if (list_empty(&lower->upper))
+				list_add(&lower->list, &useless);
+		}
+		__mark_block_processed(rc, upper);
+		if (upper->level > 0) {
+			list_add(&upper->list, &cache->detached);
+			upper->detached = 1;
+		} else {
+			rb_erase(&upper->rb_node, &cache->rb_root);
+			free_backref_node(cache, upper);
+		}
+	}
 out:
 	btrfs_free_path(path1);
 	btrfs_free_path(path2);
 	if (err) {
-		INIT_LIST_HEAD(&list);
+		while (!list_empty(&useless)) {
+			lower = list_entry(useless.next,
+					   struct backref_node, upper);
+			list_del_init(&lower->upper);
+		}
 		upper = node;
+		INIT_LIST_HEAD(&list);
 		while (upper) {
 			if (RB_EMPTY_NODE(&upper->rb_node)) {
 				list_splice_tail(&upper->upper, &list);
-				kfree(upper);
+				free_backref_node(cache, upper);
 			}
 
 			if (list_empty(&list))
@@ -833,25 +1135,119 @@ out:
 
 			edge = list_entry(list.next, struct backref_edge,
 					  list[LOWER]);
+			list_del(&edge->list[LOWER]);
 			upper = edge->node[UPPER];
-			kfree(edge);
+			free_backref_edge(cache, edge);
 		}
 		return ERR_PTR(err);
 	}
+	BUG_ON(node && node->detached);
 	return node;
 }
 
 /*
+ * helper to add backref node for the newly created snapshot.
+ * the backref node is created by cloning backref node that
+ * corresponds to root of source tree
+ */
+static int clone_backref_node(struct btrfs_trans_handle *trans,
+			      struct reloc_control *rc,
+			      struct btrfs_root *src,
+			      struct btrfs_root *dest)
+{
+	struct btrfs_root *reloc_root = src->reloc_root;
+	struct backref_cache *cache = &rc->backref_cache;
+	struct backref_node *node = NULL;
+	struct backref_node *new_node;
+	struct backref_edge *edge;
+	struct backref_edge *new_edge;
+	struct rb_node *rb_node;
+
+	if (cache->last_trans > 0)
+		update_backref_cache(trans, cache);
+
+	rb_node = tree_search(&cache->rb_root, src->commit_root->start);
+	if (rb_node) {
+		node = rb_entry(rb_node, struct backref_node, rb_node);
+		if (node->detached)
+			node = NULL;
+		else
+			BUG_ON(node->new_bytenr != reloc_root->node->start);
+	}
+
+	if (!node) {
+		rb_node = tree_search(&cache->rb_root,
+				      reloc_root->commit_root->start);
+		if (rb_node) {
+			node = rb_entry(rb_node, struct backref_node,
+					rb_node);
+			BUG_ON(node->detached);
+		}
+	}
+
+	if (!node)
+		return 0;
+
+	new_node = alloc_backref_node(cache);
+	if (!new_node)
+		return -ENOMEM;
+
+	new_node->bytenr = dest->node->start;
+	new_node->level = node->level;
+	new_node->lowest = node->lowest;
+	new_node->checked = 1;
+	new_node->root = dest;
+
+	if (!node->lowest) {
+		list_for_each_entry(edge, &node->lower, list[UPPER]) {
+			new_edge = alloc_backref_edge(cache);
+			if (!new_edge)
+				goto fail;
+
+			new_edge->node[UPPER] = new_node;
+			new_edge->node[LOWER] = edge->node[LOWER];
+			list_add_tail(&new_edge->list[UPPER],
+				      &new_node->lower);
+		}
+	} else {
+		list_add_tail(&new_node->lower, &cache->leaves);
+	}
+
+	rb_node = tree_insert(&cache->rb_root, new_node->bytenr,
+			      &new_node->rb_node);
+	if (rb_node)
+		backref_tree_panic(rb_node, -EEXIST, new_node->bytenr);
+
+	if (!new_node->lowest) {
+		list_for_each_entry(new_edge, &new_node->lower, list[UPPER]) {
+			list_add_tail(&new_edge->list[LOWER],
+				      &new_edge->node[LOWER]->upper);
+		}
+	}
+	return 0;
+fail:
+	while (!list_empty(&new_node->lower)) {
+		new_edge = list_entry(new_node->lower.next,
+				      struct backref_edge, list[UPPER]);
+		list_del(&new_edge->list[UPPER]);
+		free_backref_edge(cache, new_edge);
+	}
+	free_backref_node(cache, new_node);
+	return -ENOMEM;
+}
+
+/*
  * helper to add 'address of tree root -> reloc tree' mapping
  */
-static int __add_reloc_root(struct btrfs_root *root)
+static int __must_check __add_reloc_root(struct btrfs_root *root)
 {
 	struct rb_node *rb_node;
 	struct mapping_node *node;
 	struct reloc_control *rc = root->fs_info->reloc_ctl;
 
 	node = kmalloc(sizeof(*node), GFP_NOFS);
-	BUG_ON(!node);
+	if (!node)
+		return -ENOMEM;
 
 	node->bytenr = root->node->start;
 	node->data = root;
@@ -860,17 +1256,23 @@ static int __add_reloc_root(struct btrfs_root *root)
 	rb_node = tree_insert(&rc->reloc_root_tree.rb_root,
 			      node->bytenr, &node->rb_node);
 	spin_unlock(&rc->reloc_root_tree.lock);
-	BUG_ON(rb_node);
+	if (rb_node) {
+		btrfs_panic(root->fs_info, -EEXIST, "Duplicate root found "
+			    "for start=%llu while inserting into relocation "
+			    "tree", node->bytenr);
+		kfree(node);
+		return -EEXIST;
+	}
 
 	list_add_tail(&root->root_list, &rc->reloc_roots);
 	return 0;
 }
 
 /*
- * helper to update/delete the 'address of tree root -> reloc tree'
+ * helper to delete the 'address of tree root -> reloc tree'
  * mapping
  */
-static int __update_reloc_root(struct btrfs_root *root, int del)
+static void __del_reloc_root(struct btrfs_root *root)
 {
 	struct rb_node *rb_node;
 	struct mapping_node *node = NULL;
@@ -878,72 +1280,112 @@ static int __update_reloc_root(struct btrfs_root *root, int del)
 
 	spin_lock(&rc->reloc_root_tree.lock);
 	rb_node = tree_search(&rc->reloc_root_tree.rb_root,
-			      root->commit_root->start);
+			      root->node->start);
 	if (rb_node) {
 		node = rb_entry(rb_node, struct mapping_node, rb_node);
 		rb_erase(&node->rb_node, &rc->reloc_root_tree.rb_root);
 	}
 	spin_unlock(&rc->reloc_root_tree.lock);
 
+	if (!node)
+		return;
 	BUG_ON((struct btrfs_root *)node->data != root);
 
-	if (!del) {
-		spin_lock(&rc->reloc_root_tree.lock);
-		node->bytenr = root->node->start;
-		rb_node = tree_insert(&rc->reloc_root_tree.rb_root,
-				      node->bytenr, &node->rb_node);
-		spin_unlock(&rc->reloc_root_tree.lock);
-		BUG_ON(rb_node);
-	} else {
-		list_del_init(&root->root_list);
-		kfree(node);
-	}
-	return 0;
+	spin_lock(&root->fs_info->trans_lock);
+	list_del_init(&root->root_list);
+	spin_unlock(&root->fs_info->trans_lock);
+	kfree(node);
 }
 
 /*
- * create reloc tree for a given fs tree. reloc tree is just a
- * snapshot of the fs tree with special root objectid.
+ * helper to update the 'address of tree root -> reloc tree'
+ * mapping
  */
-int btrfs_init_reloc_root(struct btrfs_trans_handle *trans,
-			  struct btrfs_root *root)
+static int __update_reloc_root(struct btrfs_root *root, u64 new_bytenr)
+{
+	struct rb_node *rb_node;
+	struct mapping_node *node = NULL;
+	struct reloc_control *rc = root->fs_info->reloc_ctl;
+
+	spin_lock(&rc->reloc_root_tree.lock);
+	rb_node = tree_search(&rc->reloc_root_tree.rb_root,
+			      root->node->start);
+	if (rb_node) {
+		node = rb_entry(rb_node, struct mapping_node, rb_node);
+		rb_erase(&node->rb_node, &rc->reloc_root_tree.rb_root);
+	}
+	spin_unlock(&rc->reloc_root_tree.lock);
+
+	if (!node)
+		return 0;
+	BUG_ON((struct btrfs_root *)node->data != root);
+
+	spin_lock(&rc->reloc_root_tree.lock);
+	node->bytenr = new_bytenr;
+	rb_node = tree_insert(&rc->reloc_root_tree.rb_root,
+			      node->bytenr, &node->rb_node);
+	spin_unlock(&rc->reloc_root_tree.lock);
+	if (rb_node)
+		backref_tree_panic(rb_node, -EEXIST, node->bytenr);
+	return 0;
+}
+
+static struct btrfs_root *create_reloc_root(struct btrfs_trans_handle *trans,
+					struct btrfs_root *root, u64 objectid)
 {
 	struct btrfs_root *reloc_root;
 	struct extent_buffer *eb;
 	struct btrfs_root_item *root_item;
 	struct btrfs_key root_key;
+	u64 last_snap = 0;
 	int ret;
 
-	if (root->reloc_root) {
-		reloc_root = root->reloc_root;
-		reloc_root->last_trans = trans->transid;
-		return 0;
-	}
-
-	if (!root->fs_info->reloc_ctl ||
-	    !root->fs_info->reloc_ctl->create_reloc_root ||
-	    root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID)
-		return 0;
-
 	root_item = kmalloc(sizeof(*root_item), GFP_NOFS);
 	BUG_ON(!root_item);
 
 	root_key.objectid = BTRFS_TREE_RELOC_OBJECTID;
 	root_key.type = BTRFS_ROOT_ITEM_KEY;
-	root_key.offset = root->root_key.objectid;
+	root_key.offset = objectid;
 
-	ret = btrfs_copy_root(trans, root, root->commit_root, &eb,
-			      BTRFS_TREE_RELOC_OBJECTID);
-	BUG_ON(ret);
+	if (root->root_key.objectid == objectid) {
+		/* called by btrfs_init_reloc_root */
+		ret = btrfs_copy_root(trans, root, root->commit_root, &eb,
+				      BTRFS_TREE_RELOC_OBJECTID);
+		BUG_ON(ret);
+
+		last_snap = btrfs_root_last_snapshot(&root->root_item);
+		btrfs_set_root_last_snapshot(&root->root_item,
+					     trans->transid - 1);
+	} else {
+		/*
+		 * called by btrfs_reloc_post_snapshot_hook.
+		 * the source tree is a reloc tree, all tree blocks
+		 * modified after it was created have RELOC flag
+		 * set in their headers. so it's OK to not update
+		 * the 'last_snapshot'.
+		 */
+		ret = btrfs_copy_root(trans, root, root->node, &eb,
+				      BTRFS_TREE_RELOC_OBJECTID);
+		BUG_ON(ret);
+	}
 
-	btrfs_set_root_last_snapshot(&root->root_item, trans->transid - 1);
 	memcpy(root_item, &root->root_item, sizeof(*root_item));
-	btrfs_set_root_refs(root_item, 1);
 	btrfs_set_root_bytenr(root_item, eb->start);
 	btrfs_set_root_level(root_item, btrfs_header_level(eb));
 	btrfs_set_root_generation(root_item, trans->transid);
-	memset(&root_item->drop_progress, 0, sizeof(struct btrfs_disk_key));
-	root_item->drop_level = 0;
+
+	if (root->root_key.objectid == objectid) {
+		btrfs_set_root_refs(root_item, 0);
+		memset(&root_item->drop_progress, 0,
+		       sizeof(struct btrfs_disk_key));
+		root_item->drop_level = 0;
+		/*
+		 * abuse rtransid, it is safe because it is impossible to
+		 * receive data into a relocation tree.
+		 */
+		btrfs_set_root_rtransid(root_item, last_snap);
+		btrfs_set_root_otransid(root_item, trans->transid);
+	}
 
 	btrfs_tree_unlock(eb);
 	free_extent_buffer(eb);
@@ -953,12 +1395,46 @@ int btrfs_init_reloc_root(struct btrfs_trans_handle *trans,
 	BUG_ON(ret);
 	kfree(root_item);
 
-	reloc_root = btrfs_read_fs_root_no_radix(root->fs_info->tree_root,
-						 &root_key);
+	reloc_root = btrfs_read_fs_root(root->fs_info->tree_root, &root_key);
 	BUG_ON(IS_ERR(reloc_root));
 	reloc_root->last_trans = trans->transid;
+	return reloc_root;
+}
+
+/*
+ * create reloc tree for a given fs tree. reloc tree is just a
+ * snapshot of the fs tree with special root objectid.
+ */
+int btrfs_init_reloc_root(struct btrfs_trans_handle *trans,
+			  struct btrfs_root *root)
+{
+	struct btrfs_root *reloc_root;
+	struct reloc_control *rc = root->fs_info->reloc_ctl;
+	struct btrfs_block_rsv *rsv;
+	int clear_rsv = 0;
+	int ret;
 
-	__add_reloc_root(reloc_root);
+	if (root->reloc_root) {
+		reloc_root = root->reloc_root;
+		reloc_root->last_trans = trans->transid;
+		return 0;
+	}
+
+	if (!rc || !rc->create_reloc_tree ||
+	    root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID)
+		return 0;
+
+	if (!trans->reloc_reserved) {
+		rsv = trans->block_rsv;
+		trans->block_rsv = rc->block_rsv;
+		clear_rsv = 1;
+	}
+	reloc_root = create_reloc_root(trans, root, root->root_key.objectid);
+	if (clear_rsv)
+		trans->block_rsv = rsv;
+
+	ret = __add_reloc_root(reloc_root);
+	BUG_ON(ret < 0);
 	root->reloc_root = reloc_root;
 	return 0;
 }
@@ -971,22 +1447,20 @@ int btrfs_update_reloc_root(struct btrfs_trans_handle *trans,
 {
 	struct btrfs_root *reloc_root;
 	struct btrfs_root_item *root_item;
-	int del = 0;
 	int ret;
 
 	if (!root->reloc_root)
-		return 0;
+		goto out;
 
 	reloc_root = root->reloc_root;
 	root_item = &reloc_root->root_item;
 
-	if (btrfs_root_refs(root_item) == 0) {
+	if (root->fs_info->reloc_ctl->merge_reloc_tree &&
+	    btrfs_root_refs(root_item) == 0) {
 		root->reloc_root = NULL;
-		del = 1;
+		__del_reloc_root(reloc_root);
 	}
 
-	__update_reloc_root(reloc_root, del);
-
 	if (reloc_root->commit_root != reloc_root->node) {
 		btrfs_set_root_node(root_item, reloc_root->node);
 		free_extent_buffer(reloc_root->commit_root);
@@ -996,6 +1470,8 @@ int btrfs_update_reloc_root(struct btrfs_trans_handle *trans,
 	ret = btrfs_update_root(trans, root->fs_info->tree_root,
 				&reloc_root->root_key, root_item);
 	BUG_ON(ret);
+
+out:
 	return 0;
 }
 
@@ -1018,9 +1494,9 @@ again:
 		prev = node;
 		entry = rb_entry(node, struct btrfs_inode, rb_node);
 
-		if (objectid < entry->vfs_inode.i_ino)
+		if (objectid < btrfs_ino(&entry->vfs_inode))
 			node = node->rb_left;
-		else if (objectid > entry->vfs_inode.i_ino)
+		else if (objectid > btrfs_ino(&entry->vfs_inode))
 			node = node->rb_right;
 		else
 			break;
@@ -1028,7 +1504,7 @@ again:
 	if (!node) {
 		while (prev) {
 			entry = rb_entry(prev, struct btrfs_inode, rb_node);
-			if (objectid <= entry->vfs_inode.i_ino) {
+			if (objectid <= btrfs_ino(&entry->vfs_inode)) {
 				node = prev;
 				break;
 			}
@@ -1043,7 +1519,7 @@ again:
 			return inode;
 		}
 
-		objectid = entry->vfs_inode.i_ino + 1;
+		objectid = btrfs_ino(&entry->vfs_inode) + 1;
 		if (cond_resched_lock(&root->inode_lock))
 			goto again;
 
@@ -1079,7 +1555,7 @@ static int get_new_location(struct inode *reloc_inode, u64 *new_bytenr,
 		return -ENOMEM;
 
 	bytenr -= BTRFS_I(reloc_inode)->index_cnt;
-	ret = btrfs_lookup_file_extent(NULL, root, path, reloc_inode->i_ino,
+	ret = btrfs_lookup_file_extent(NULL, root, path, btrfs_ino(reloc_inode),
 				       bytenr, 0);
 	if (ret < 0)
 		goto out;
@@ -1098,12 +1574,11 @@ static int get_new_location(struct inode *reloc_inode, u64 *new_bytenr,
 	       btrfs_file_extent_other_encoding(leaf, fi));
 
 	if (num_bytes != btrfs_file_extent_disk_num_bytes(leaf, fi)) {
-		ret = 1;
+		ret = -EINVAL;
 		goto out;
 	}
 
-	if (new_bytenr)
-		*new_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
+	*new_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
 	ret = 0;
 out:
 	btrfs_free_path(path);
@@ -1114,24 +1589,23 @@ out:
  * update file extent items in the tree leaf to point to
  * the new locations.
  */
-static int replace_file_extents(struct btrfs_trans_handle *trans,
-				struct reloc_control *rc,
-				struct btrfs_root *root,
-				struct extent_buffer *leaf,
-				struct list_head *inode_list)
+static noinline_for_stack
+int replace_file_extents(struct btrfs_trans_handle *trans,
+			 struct reloc_control *rc,
+			 struct btrfs_root *root,
+			 struct extent_buffer *leaf)
 {
 	struct btrfs_key key;
 	struct btrfs_file_extent_item *fi;
 	struct inode *inode = NULL;
-	struct inodevec *ivec = NULL;
 	u64 parent;
 	u64 bytenr;
-	u64 new_bytenr;
+	u64 new_bytenr = 0;
 	u64 num_bytes;
 	u64 end;
 	u32 nritems;
 	u32 i;
-	int ret;
+	int ret = 0;
 	int first = 1;
 	int dirty = 0;
 
@@ -1166,23 +1640,14 @@ static int replace_file_extents(struct btrfs_trans_handle *trans,
 		 * to complete and drop the extent cache
 		 */
 		if (root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID) {
-			if (!ivec || ivec->nr == INODEVEC_SIZE) {
-				ivec = kmalloc(sizeof(*ivec), GFP_NOFS);
-				BUG_ON(!ivec);
-				ivec->nr = 0;
-				list_add_tail(&ivec->list, inode_list);
-			}
 			if (first) {
 				inode = find_next_inode(root, key.objectid);
-				if (inode)
-					ivec->inode[ivec->nr++] = inode;
 				first = 0;
-			} else if (inode && inode->i_ino < key.objectid) {
+			} else if (inode && btrfs_ino(inode) < key.objectid) {
+				btrfs_add_delayed_iput(inode);
 				inode = find_next_inode(root, key.objectid);
-				if (inode)
-					ivec->inode[ivec->nr++] = inode;
 			}
-			if (inode && inode->i_ino == key.objectid) {
+			if (inode && btrfs_ino(inode) == key.objectid) {
 				end = key.offset +
 				      btrfs_file_extent_num_bytes(leaf, fi);
 				WARN_ON(!IS_ALIGNED(key.offset,
@@ -1190,23 +1655,26 @@ static int replace_file_extents(struct btrfs_trans_handle *trans,
 				WARN_ON(!IS_ALIGNED(end, root->sectorsize));
 				end--;
 				ret = try_lock_extent(&BTRFS_I(inode)->io_tree,
-						      key.offset, end,
-						      GFP_NOFS);
+						      key.offset, end);
 				if (!ret)
 					continue;
 
 				btrfs_drop_extent_cache(inode, key.offset, end,
 							1);
 				unlock_extent(&BTRFS_I(inode)->io_tree,
-					      key.offset, end, GFP_NOFS);
+					      key.offset, end);
 			}
 		}
 
 		ret = get_new_location(rc->data_inode, &new_bytenr,
 				       bytenr, num_bytes);
-		if (ret > 0)
-			continue;
-		BUG_ON(ret < 0);
+		if (ret) {
+			/*
+			 * Don't have to abort since we've not changed anything
+			 * in the file extent yet.
+			 */
+			break;
+		}
 
 		btrfs_set_file_extent_disk_bytenr(leaf, fi, new_bytenr);
 		dirty = 1;
@@ -1215,17 +1683,25 @@ static int replace_file_extents(struct btrfs_trans_handle *trans,
 		ret = btrfs_inc_extent_ref(trans, root, new_bytenr,
 					   num_bytes, parent,
 					   btrfs_header_owner(leaf),
-					   key.objectid, key.offset);
-		BUG_ON(ret);
+					   key.objectid, key.offset, 1);
+		if (ret) {
+			btrfs_abort_transaction(trans, root, ret);
+			break;
+		}
 
 		ret = btrfs_free_extent(trans, root, bytenr, num_bytes,
 					parent, btrfs_header_owner(leaf),
-					key.objectid, key.offset);
-		BUG_ON(ret);
+					key.objectid, key.offset, 1);
+		if (ret) {
+			btrfs_abort_transaction(trans, root, ret);
+			break;
+		}
 	}
 	if (dirty)
 		btrfs_mark_buffer_dirty(leaf);
-	return 0;
+	if (inode)
+		btrfs_add_delayed_iput(inode);
+	return ret;
 }
 
 static noinline_for_stack
@@ -1248,11 +1724,11 @@ int memcmp_node_keys(struct extent_buffer *eb, int slot,
  * if no block got replaced, 0 is returned. if there are other
  * errors, a negative error number is returned.
  */
-static int replace_path(struct btrfs_trans_handle *trans,
-			struct btrfs_root *dest, struct btrfs_root *src,
-			struct btrfs_path *path, struct btrfs_key *next_key,
-			struct extent_buffer **leaf,
-			int lowest_level, int max_level)
+static noinline_for_stack
+int replace_path(struct btrfs_trans_handle *trans,
+		 struct btrfs_root *dest, struct btrfs_root *src,
+		 struct btrfs_path *path, struct btrfs_key *next_key,
+		 int lowest_level, int max_level)
 {
 	struct extent_buffer *eb;
 	struct extent_buffer *parent;
@@ -1263,16 +1739,16 @@ static int replace_path(struct btrfs_trans_handle *trans,
 	u64 new_ptr_gen;
 	u64 last_snapshot;
 	u32 blocksize;
+	int cow = 0;
 	int level;
 	int ret;
 	int slot;
 
 	BUG_ON(src->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID);
 	BUG_ON(dest->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID);
-	BUG_ON(lowest_level > 1 && leaf);
 
 	last_snapshot = btrfs_root_last_snapshot(&src->root_item);
-
+again:
 	slot = path->slots[lowest_level];
 	btrfs_node_key_to_cpu(path->nodes[lowest_level], &key, slot);
 
@@ -1286,8 +1762,10 @@ static int replace_path(struct btrfs_trans_handle *trans,
 		return 0;
 	}
 
-	ret = btrfs_cow_block(trans, dest, eb, NULL, 0, &eb);
-	BUG_ON(ret);
+	if (cow) {
+		ret = btrfs_cow_block(trans, dest, eb, NULL, 0, &eb);
+		BUG_ON(ret);
+	}
 	btrfs_set_lock_blocking(eb);
 
 	if (next_key) {
@@ -1323,32 +1801,32 @@ static int replace_path(struct btrfs_trans_handle *trans,
 			new_ptr_gen = 0;
 		}
 
-		if (new_bytenr > 0 && new_bytenr == old_bytenr) {
-			WARN_ON(1);
+		if (WARN_ON(new_bytenr > 0 && new_bytenr == old_bytenr)) {
 			ret = level;
 			break;
 		}
 
 		if (new_bytenr == 0 || old_ptr_gen > last_snapshot ||
 		    memcmp_node_keys(parent, slot, path, level)) {
-			if (level <= lowest_level && !leaf) {
+			if (level <= lowest_level) {
 				ret = 0;
 				break;
 			}
 
 			eb = read_tree_block(dest, old_bytenr, blocksize,
 					     old_ptr_gen);
-			btrfs_tree_lock(eb);
-			ret = btrfs_cow_block(trans, dest, eb, parent,
-					      slot, &eb);
-			BUG_ON(ret);
-			btrfs_set_lock_blocking(eb);
-
-			if (level <= lowest_level) {
-				*leaf = eb;
-				ret = 0;
+			if (!eb || !extent_buffer_uptodate(eb)) {
+				ret = (!eb) ? -ENOMEM : -EIO;
+				free_extent_buffer(eb);
 				break;
 			}
+			btrfs_tree_lock(eb);
+			if (cow) {
+				ret = btrfs_cow_block(trans, dest, eb, parent,
+						      slot, &eb);
+				BUG_ON(ret);
+			}
+			btrfs_set_lock_blocking(eb);
 
 			btrfs_tree_unlock(parent);
 			free_extent_buffer(parent);
@@ -1357,9 +1835,16 @@ static int replace_path(struct btrfs_trans_handle *trans,
 			continue;
 		}
 
+		if (!cow) {
+			btrfs_tree_unlock(parent);
+			free_extent_buffer(parent);
+			cow = 1;
+			goto again;
+		}
+
 		btrfs_node_key_to_cpu(path->nodes[level], &key,
 				      path->slots[level]);
-		btrfs_release_path(src, path);
+		btrfs_release_path(path);
 
 		path->lowest_level = level;
 		ret = btrfs_search_slot(trans, src, &key, path, 0, 1);
@@ -1381,21 +1866,23 @@ static int replace_path(struct btrfs_trans_handle *trans,
 
 		ret = btrfs_inc_extent_ref(trans, src, old_bytenr, blocksize,
 					path->nodes[level]->start,
-					src->root_key.objectid, level - 1, 0);
+					src->root_key.objectid, level - 1, 0,
+					1);
 		BUG_ON(ret);
 		ret = btrfs_inc_extent_ref(trans, dest, new_bytenr, blocksize,
 					0, dest->root_key.objectid, level - 1,
-					0);
+					0, 1);
 		BUG_ON(ret);
 
 		ret = btrfs_free_extent(trans, src, new_bytenr, blocksize,
 					path->nodes[level]->start,
-					src->root_key.objectid, level - 1, 0);
+					src->root_key.objectid, level - 1, 0,
+					1);
 		BUG_ON(ret);
 
 		ret = btrfs_free_extent(trans, dest, old_bytenr, blocksize,
 					0, dest->root_key.objectid, level - 1,
-					0);
+					0, 1);
 		BUG_ON(ret);
 
 		btrfs_unlock_up_safe(path, 0);
@@ -1485,6 +1972,10 @@ int walk_down_reloc_tree(struct btrfs_root *root, struct btrfs_path *path,
 		bytenr = btrfs_node_blockptr(eb, path->slots[i]);
 		blocksize = btrfs_level_size(root, i - 1);
 		eb = read_tree_block(root, bytenr, blocksize, ptr_gen);
+		if (!eb || !extent_buffer_uptodate(eb)) {
+			free_extent_buffer(eb);
+			return -EIO;
+		}
 		BUG_ON(btrfs_header_level(eb) != i - 1);
 		path->nodes[i - 1] = eb;
 		path->slots[i - 1] = 0;
@@ -1503,6 +1994,7 @@ static int invalidate_extent_cache(struct btrfs_root *root,
 	struct inode *inode = NULL;
 	u64 objectid;
 	u64 start, end;
+	u64 ino;
 
 	objectid = min_key->objectid;
 	while (1) {
@@ -1515,17 +2007,18 @@ static int invalidate_extent_cache(struct btrfs_root *root,
 		inode = find_next_inode(root, objectid);
 		if (!inode)
 			break;
+		ino = btrfs_ino(inode);
 
-		if (inode->i_ino > max_key->objectid) {
+		if (ino > max_key->objectid) {
 			iput(inode);
 			break;
 		}
 
-		objectid = inode->i_ino + 1;
+		objectid = ino + 1;
 		if (!S_ISREG(inode->i_mode))
 			continue;
 
-		if (unlikely(min_key->objectid == inode->i_ino)) {
+		if (unlikely(min_key->objectid == ino)) {
 			if (min_key->type > BTRFS_EXTENT_DATA_KEY)
 				continue;
 			if (min_key->type < BTRFS_EXTENT_DATA_KEY)
@@ -1538,7 +2031,7 @@ static int invalidate_extent_cache(struct btrfs_root *root,
 			start = 0;
 		}
 
-		if (unlikely(max_key->objectid == inode->i_ino)) {
+		if (unlikely(max_key->objectid == ino)) {
 			if (max_key->type < BTRFS_EXTENT_DATA_KEY)
 				continue;
 			if (max_key->type > BTRFS_EXTENT_DATA_KEY) {
@@ -1555,27 +2048,13 @@ static int invalidate_extent_cache(struct btrfs_root *root,
 		}
 
 		/* the lock_extent waits for readpage to complete */
-		lock_extent(&BTRFS_I(inode)->io_tree, start, end, GFP_NOFS);
+		lock_extent(&BTRFS_I(inode)->io_tree, start, end);
 		btrfs_drop_extent_cache(inode, start, end, 1);
-		unlock_extent(&BTRFS_I(inode)->io_tree, start, end, GFP_NOFS);
+		unlock_extent(&BTRFS_I(inode)->io_tree, start, end);
 	}
 	return 0;
 }
 
-static void put_inodes(struct list_head *list)
-{
-	struct inodevec *ivec;
-	while (!list_empty(list)) {
-		ivec = list_entry(list->next, struct inodevec, list);
-		list_del(&ivec->list);
-		while (ivec->nr > 0) {
-			ivec->nr--;
-			iput(ivec->inode[ivec->nr]);
-		}
-		kfree(ivec);
-	}
-}
-
 static int find_next_key(struct btrfs_path *path, int level,
 			 struct btrfs_key *key)
 
@@ -1604,21 +2083,22 @@ static noinline_for_stack int merge_reloc_root(struct reloc_control *rc,
 	LIST_HEAD(inode_list);
 	struct btrfs_key key;
 	struct btrfs_key next_key;
-	struct btrfs_trans_handle *trans;
+	struct btrfs_trans_handle *trans = NULL;
 	struct btrfs_root *reloc_root;
 	struct btrfs_root_item *root_item;
 	struct btrfs_path *path;
-	struct extent_buffer *leaf = NULL;
-	unsigned long nr;
+	struct extent_buffer *leaf;
 	int level;
 	int max_level;
 	int replaced = 0;
 	int ret;
 	int err = 0;
+	u32 min_reserved;
 
 	path = btrfs_alloc_path();
 	if (!path)
 		return -ENOMEM;
+	path->reada = 1;
 
 	reloc_root = root->reloc_root;
 	root_item = &reloc_root->root_item;
@@ -1648,34 +2128,25 @@ static noinline_for_stack int merge_reloc_root(struct reloc_control *rc,
 		btrfs_unlock_up_safe(path, 0);
 	}
 
-	if (level == 0 && rc->stage == UPDATE_DATA_PTRS) {
-		trans = btrfs_start_transaction(root, 1);
-
-		leaf = path->nodes[0];
-		btrfs_item_key_to_cpu(leaf, &key, 0);
-		btrfs_release_path(reloc_root, path);
+	min_reserved = root->nodesize * (BTRFS_MAX_LEVEL - 1) * 2;
+	memset(&next_key, 0, sizeof(next_key));
 
-		ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
-		if (ret < 0) {
+	while (1) {
+		ret = btrfs_block_rsv_refill(root, rc->block_rsv, min_reserved,
+					     BTRFS_RESERVE_FLUSH_ALL);
+		if (ret) {
 			err = ret;
 			goto out;
 		}
+		trans = btrfs_start_transaction(root, 0);
+		if (IS_ERR(trans)) {
+			err = PTR_ERR(trans);
+			trans = NULL;
+			goto out;
+		}
+		trans->block_rsv = rc->block_rsv;
 
-		leaf = path->nodes[0];
-		btrfs_unlock_up_safe(path, 1);
-		ret = replace_file_extents(trans, rc, root, leaf,
-					   &inode_list);
-		if (ret < 0)
-			err = ret;
-		goto out;
-	}
-
-	memset(&next_key, 0, sizeof(next_key));
-
-	while (1) {
-		leaf = NULL;
 		replaced = 0;
-		trans = btrfs_start_transaction(root, 1);
 		max_level = level;
 
 		ret = walk_down_reloc_tree(reloc_root, path, &level);
@@ -1689,14 +2160,9 @@ static noinline_for_stack int merge_reloc_root(struct reloc_control *rc,
 		if (!find_next_key(path, level, &key) &&
 		    btrfs_comp_cpu_keys(&next_key, &key) >= 0) {
 			ret = 0;
-		} else if (level == 1 && rc->stage == UPDATE_DATA_PTRS) {
-			ret = replace_path(trans, root, reloc_root,
-					   path, &next_key, &leaf,
-					   level, max_level);
 		} else {
-			ret = replace_path(trans, root, reloc_root,
-					   path, &next_key, NULL,
-					   level, max_level);
+			ret = replace_path(trans, root, reloc_root, path,
+					   &next_key, level, max_level);
 		}
 		if (ret < 0) {
 			err = ret;
@@ -1708,16 +2174,6 @@ static noinline_for_stack int merge_reloc_root(struct reloc_control *rc,
 			btrfs_node_key_to_cpu(path->nodes[level], &key,
 					      path->slots[level]);
 			replaced = 1;
-		} else if (leaf) {
-			/*
-			 * no block got replaced, try replacing file extents
-			 */
-			btrfs_item_key_to_cpu(leaf, &key, 0);
-			ret = replace_file_extents(trans, rc, root, leaf,
-						   &inode_list);
-			btrfs_tree_unlock(leaf);
-			free_extent_buffer(leaf);
-			BUG_ON(ret < 0);
 		}
 
 		ret = walk_up_reloc_tree(reloc_root, path, &level);
@@ -1733,15 +2189,10 @@ static noinline_for_stack int merge_reloc_root(struct reloc_control *rc,
 			       path->slots[level]);
 		root_item->drop_level = level;
 
-		nr = trans->blocks_used;
-		btrfs_end_transaction(trans, root);
-
-		btrfs_btree_balance_dirty(root, nr);
+		btrfs_end_transaction_throttle(trans, root);
+		trans = NULL;
 
-		/*
-		 * put inodes outside transaction, otherwise we may deadlock.
-		 */
-		put_inodes(&inode_list);
+		btrfs_btree_balance_dirty(root);
 
 		if (replaced && rc->stage == UPDATE_DATA_PTRS)
 			invalidate_extent_cache(root, &key, &next_key);
@@ -1765,14 +2216,13 @@ out:
 		       sizeof(root_item->drop_progress));
 		root_item->drop_level = 0;
 		btrfs_set_root_refs(root_item, 0);
+		btrfs_update_reloc_root(trans, root);
 	}
 
-	nr = trans->blocks_used;
-	btrfs_end_transaction(trans, root);
-
-	btrfs_btree_balance_dirty(root, nr);
+	if (trans)
+		btrfs_end_transaction_throttle(trans, root);
 
-	put_inodes(&inode_list);
+	btrfs_btree_balance_dirty(root);
 
 	if (replaced && rc->stage == UPDATE_DATA_PTRS)
 		invalidate_extent_cache(root, &key, &next_key);
@@ -1780,74 +2230,174 @@ out:
 	return err;
 }
 
-/*
- * callback for the work threads.
- * this function merges reloc tree with corresponding fs tree,
- * and then drops the reloc tree.
- */
-static void merge_func(struct btrfs_work *work)
+static noinline_for_stack
+int prepare_to_merge(struct reloc_control *rc, int err)
 {
-	struct btrfs_trans_handle *trans;
-	struct btrfs_root *root;
+	struct btrfs_root *root = rc->extent_root;
 	struct btrfs_root *reloc_root;
-	struct async_merge *async;
+	struct btrfs_trans_handle *trans;
+	LIST_HEAD(reloc_roots);
+	u64 num_bytes = 0;
+	int ret;
+
+	mutex_lock(&root->fs_info->reloc_mutex);
+	rc->merging_rsv_size += root->nodesize * (BTRFS_MAX_LEVEL - 1) * 2;
+	rc->merging_rsv_size += rc->nodes_relocated * 2;
+	mutex_unlock(&root->fs_info->reloc_mutex);
+
+again:
+	if (!err) {
+		num_bytes = rc->merging_rsv_size;
+		ret = btrfs_block_rsv_add(root, rc->block_rsv, num_bytes,
+					  BTRFS_RESERVE_FLUSH_ALL);
+		if (ret)
+			err = ret;
+	}
+
+	trans = btrfs_join_transaction(rc->extent_root);
+	if (IS_ERR(trans)) {
+		if (!err)
+			btrfs_block_rsv_release(rc->extent_root,
+						rc->block_rsv, num_bytes);
+		return PTR_ERR(trans);
+	}
+
+	if (!err) {
+		if (num_bytes != rc->merging_rsv_size) {
+			btrfs_end_transaction(trans, rc->extent_root);
+			btrfs_block_rsv_release(rc->extent_root,
+						rc->block_rsv, num_bytes);
+			goto again;
+		}
+	}
 
-	async = container_of(work, struct async_merge, work);
-	reloc_root = async->root;
+	rc->merge_reloc_tree = 1;
+
+	while (!list_empty(&rc->reloc_roots)) {
+		reloc_root = list_entry(rc->reloc_roots.next,
+					struct btrfs_root, root_list);
+		list_del_init(&reloc_root->root_list);
 
-	if (btrfs_root_refs(&reloc_root->root_item) > 0) {
 		root = read_fs_root(reloc_root->fs_info,
 				    reloc_root->root_key.offset);
 		BUG_ON(IS_ERR(root));
 		BUG_ON(root->reloc_root != reloc_root);
 
-		merge_reloc_root(async->rc, root);
-
-		trans = btrfs_start_transaction(root, 1);
+		/*
+		 * set reference count to 1, so btrfs_recover_relocation
+		 * knows it should resumes merging
+		 */
+		if (!err)
+			btrfs_set_root_refs(&reloc_root->root_item, 1);
 		btrfs_update_reloc_root(trans, root);
-		btrfs_end_transaction(trans, root);
+
+		list_add(&reloc_root->root_list, &reloc_roots);
 	}
 
-	btrfs_drop_snapshot(reloc_root, 0);
+	list_splice(&reloc_roots, &rc->reloc_roots);
 
-	if (atomic_dec_and_test(async->num_pending))
-		complete(async->done);
+	if (!err)
+		btrfs_commit_transaction(trans, rc->extent_root);
+	else
+		btrfs_end_transaction(trans, rc->extent_root);
+	return err;
+}
 
-	kfree(async);
+static noinline_for_stack
+void free_reloc_roots(struct list_head *list)
+{
+	struct btrfs_root *reloc_root;
+
+	while (!list_empty(list)) {
+		reloc_root = list_entry(list->next, struct btrfs_root,
+					root_list);
+		__del_reloc_root(reloc_root);
+	}
 }
 
-static int merge_reloc_roots(struct reloc_control *rc)
+static noinline_for_stack
+int merge_reloc_roots(struct reloc_control *rc)
 {
-	struct async_merge *async;
 	struct btrfs_root *root;
-	struct completion done;
-	atomic_t num_pending;
+	struct btrfs_root *reloc_root;
+	u64 last_snap;
+	u64 otransid;
+	u64 objectid;
+	LIST_HEAD(reloc_roots);
+	int found = 0;
+	int ret = 0;
+again:
+	root = rc->extent_root;
 
-	init_completion(&done);
-	atomic_set(&num_pending, 1);
+	/*
+	 * this serializes us with btrfs_record_root_in_transaction,
+	 * we have to make sure nobody is in the middle of
+	 * adding their roots to the list while we are
+	 * doing this splice
+	 */
+	mutex_lock(&root->fs_info->reloc_mutex);
+	list_splice_init(&rc->reloc_roots, &reloc_roots);
+	mutex_unlock(&root->fs_info->reloc_mutex);
 
-	while (!list_empty(&rc->reloc_roots)) {
-		root = list_entry(rc->reloc_roots.next,
-				  struct btrfs_root, root_list);
-		list_del_init(&root->root_list);
+	while (!list_empty(&reloc_roots)) {
+		found = 1;
+		reloc_root = list_entry(reloc_roots.next,
+					struct btrfs_root, root_list);
 
-		async = kmalloc(sizeof(*async), GFP_NOFS);
-		BUG_ON(!async);
-		async->work.func = merge_func;
-		async->work.flags = 0;
-		async->rc = rc;
-		async->root = root;
-		async->done = &done;
-		async->num_pending = &num_pending;
-		atomic_inc(&num_pending);
-		btrfs_queue_worker(&rc->workers, &async->work);
+		if (btrfs_root_refs(&reloc_root->root_item) > 0) {
+			root = read_fs_root(reloc_root->fs_info,
+					    reloc_root->root_key.offset);
+			BUG_ON(IS_ERR(root));
+			BUG_ON(root->reloc_root != reloc_root);
+
+			ret = merge_reloc_root(rc, root);
+			if (ret) {
+				if (list_empty(&reloc_root->root_list))
+					list_add_tail(&reloc_root->root_list,
+						      &reloc_roots);
+				goto out;
+			}
+		} else {
+			list_del_init(&reloc_root->root_list);
+		}
+
+		/*
+		 * we keep the old last snapshod transid in rtranid when we
+		 * created the relocation tree.
+		 */
+		last_snap = btrfs_root_rtransid(&reloc_root->root_item);
+		otransid = btrfs_root_otransid(&reloc_root->root_item);
+		objectid = reloc_root->root_key.offset;
+
+		ret = btrfs_drop_snapshot(reloc_root, rc->block_rsv, 0, 1);
+		if (ret < 0) {
+			if (list_empty(&reloc_root->root_list))
+				list_add_tail(&reloc_root->root_list,
+					      &reloc_roots);
+			goto out;
+		}
 	}
 
-	if (!atomic_dec_and_test(&num_pending))
-		wait_for_completion(&done);
+	if (found) {
+		found = 0;
+		goto again;
+	}
+out:
+	if (ret) {
+		btrfs_std_error(root->fs_info, ret);
+		if (!list_empty(&reloc_roots))
+			free_reloc_roots(&reloc_roots);
+
+		/* new reloc root may be added */
+		mutex_lock(&root->fs_info->reloc_mutex);
+		list_splice_init(&rc->reloc_roots, &reloc_roots);
+		mutex_unlock(&root->fs_info->reloc_mutex);
+		if (!list_empty(&reloc_roots))
+			free_reloc_roots(&reloc_roots);
+	}
 
 	BUG_ON(!RB_EMPTY_ROOT(&rc->reloc_root_tree.rb_root));
-	return 0;
+	return ret;
 }
 
 static void free_block_list(struct rb_root *blocks)
@@ -1876,119 +2426,175 @@ static int record_reloc_root_in_trans(struct btrfs_trans_handle *trans,
 	return btrfs_record_root_in_trans(trans, root);
 }
 
-/*
- * select one tree from trees that references the block.
- * for blocks in refernce counted trees, we preper reloc tree.
- * if no reloc tree found and reloc_only is true, NULL is returned.
- */
-static struct btrfs_root *__select_one_root(struct btrfs_trans_handle *trans,
-					    struct backref_node *node,
-					    struct backref_edge *edges[],
-					    int *nr, int reloc_only)
+static noinline_for_stack
+struct btrfs_root *select_reloc_root(struct btrfs_trans_handle *trans,
+				     struct reloc_control *rc,
+				     struct backref_node *node,
+				     struct backref_edge *edges[])
 {
 	struct backref_node *next;
 	struct btrfs_root *root;
-	int index;
-	int loop = 0;
-again:
-	index = 0;
+	int index = 0;
+
 	next = node;
 	while (1) {
 		cond_resched();
 		next = walk_up_backref(next, edges, &index);
 		root = next->root;
-		if (!root) {
-			BUG_ON(!node->old_root);
-			goto skip;
-		}
-
-		/* no other choice for non-refernce counted tree */
-		if (!root->ref_cows) {
-			BUG_ON(reloc_only);
-			break;
-		}
+		BUG_ON(!root);
+		BUG_ON(!test_bit(BTRFS_ROOT_REF_COWS, &root->state));
 
 		if (root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID) {
 			record_reloc_root_in_trans(trans, root);
 			break;
 		}
 
-		if (loop) {
-			btrfs_record_root_in_trans(trans, root);
+		btrfs_record_root_in_trans(trans, root);
+		root = root->reloc_root;
+
+		if (next->new_bytenr != root->node->start) {
+			BUG_ON(next->new_bytenr);
+			BUG_ON(!list_empty(&next->list));
+			next->new_bytenr = root->node->start;
+			next->root = root;
+			list_add_tail(&next->list,
+				      &rc->backref_cache.changed);
+			__mark_block_processed(rc, next);
 			break;
 		}
 
-		if (reloc_only || next != node) {
-			if (!root->reloc_root)
-				btrfs_record_root_in_trans(trans, root);
-			root = root->reloc_root;
-			/*
-			 * if the reloc tree was created in current
-			 * transation, there is no node in backref tree
-			 * corresponds to the root of the reloc tree.
-			 */
-			if (btrfs_root_last_snapshot(&root->root_item) ==
-			    trans->transid - 1)
-				break;
-		}
-skip:
+		WARN_ON(1);
 		root = NULL;
 		next = walk_down_backref(edges, &index);
 		if (!next || next->level <= node->level)
 			break;
 	}
+	if (!root)
+		return NULL;
 
-	if (!root && !loop && !reloc_only) {
-		loop = 1;
-		goto again;
+	next = node;
+	/* setup backref node path for btrfs_reloc_cow_block */
+	while (1) {
+		rc->backref_cache.path[next->level] = next;
+		if (--index < 0)
+			break;
+		next = edges[index]->node[UPPER];
 	}
-
-	if (root)
-		*nr = index;
-	else
-		*nr = 0;
-
 	return root;
 }
 
+/*
+ * select a tree root for relocation. return NULL if the block
+ * is reference counted. we should use do_relocation() in this
+ * case. return a tree root pointer if the block isn't reference
+ * counted. return -ENOENT if the block is root of reloc tree.
+ */
 static noinline_for_stack
 struct btrfs_root *select_one_root(struct btrfs_trans_handle *trans,
 				   struct backref_node *node)
 {
+	struct backref_node *next;
+	struct btrfs_root *root;
+	struct btrfs_root *fs_root = NULL;
 	struct backref_edge *edges[BTRFS_MAX_LEVEL - 1];
-	int nr;
-	return __select_one_root(trans, node, edges, &nr, 0);
+	int index = 0;
+
+	next = node;
+	while (1) {
+		cond_resched();
+		next = walk_up_backref(next, edges, &index);
+		root = next->root;
+		BUG_ON(!root);
+
+		/* no other choice for non-references counted tree */
+		if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
+			return root;
+
+		if (root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID)
+			fs_root = root;
+
+		if (next != node)
+			return NULL;
+
+		next = walk_down_backref(edges, &index);
+		if (!next || next->level <= node->level)
+			break;
+	}
+
+	if (!fs_root)
+		return ERR_PTR(-ENOENT);
+	return fs_root;
 }
 
 static noinline_for_stack
-struct btrfs_root *select_reloc_root(struct btrfs_trans_handle *trans,
-				     struct backref_node *node,
-				     struct backref_edge *edges[], int *nr)
+u64 calcu_metadata_size(struct reloc_control *rc,
+			struct backref_node *node, int reserve)
 {
-	return __select_one_root(trans, node, edges, nr, 1);
+	struct backref_node *next = node;
+	struct backref_edge *edge;
+	struct backref_edge *edges[BTRFS_MAX_LEVEL - 1];
+	u64 num_bytes = 0;
+	int index = 0;
+
+	BUG_ON(reserve && node->processed);
+
+	while (next) {
+		cond_resched();
+		while (1) {
+			if (next->processed && (reserve || next != node))
+				break;
+
+			num_bytes += btrfs_level_size(rc->extent_root,
+						      next->level);
+
+			if (list_empty(&next->upper))
+				break;
+
+			edge = list_entry(next->upper.next,
+					  struct backref_edge, list[LOWER]);
+			edges[index++] = edge;
+			next = edge->node[UPPER];
+		}
+		next = walk_down_backref(edges, &index);
+	}
+	return num_bytes;
 }
 
-static void grab_path_buffers(struct btrfs_path *path,
-			      struct backref_node *node,
-			      struct backref_edge *edges[], int nr)
+static int reserve_metadata_space(struct btrfs_trans_handle *trans,
+				  struct reloc_control *rc,
+				  struct backref_node *node)
 {
-	int i = 0;
-	while (1) {
-		drop_node_buffer(node);
-		node->eb = path->nodes[node->level];
-		BUG_ON(!node->eb);
-		if (path->locks[node->level])
-			node->locked = 1;
-		path->nodes[node->level] = NULL;
-		path->locks[node->level] = 0;
-
-		if (i >= nr)
-			break;
-
-		edges[i]->blockptr = node->eb->start;
-		node = edges[i]->node[UPPER];
-		i++;
+	struct btrfs_root *root = rc->extent_root;
+	u64 num_bytes;
+	int ret;
+	u64 tmp;
+
+	num_bytes = calcu_metadata_size(rc, node, 1) * 2;
+
+	trans->block_rsv = rc->block_rsv;
+	rc->reserved_bytes += num_bytes;
+	ret = btrfs_block_rsv_refill(root, rc->block_rsv, num_bytes,
+				BTRFS_RESERVE_FLUSH_ALL);
+	if (ret) {
+		if (ret == -EAGAIN) {
+			tmp = rc->extent_root->nodesize *
+				RELOCATION_RESERVED_NODES;
+			while (tmp <= rc->reserved_bytes)
+				tmp <<= 1;
+			/*
+			 * only one thread can access block_rsv at this point,
+			 * so we don't need hold lock to protect block_rsv.
+			 * we expand more reservation size here to allow enough
+			 * space for relocation and we will return eailer in
+			 * enospc case.
+			 */
+			rc->block_rsv->size = tmp + rc->extent_root->nodesize *
+					      RELOCATION_RESERVED_NODES;
+		}
+		return ret;
 	}
+
+	return 0;
 }
 
 /*
@@ -1999,6 +2605,7 @@ static void grab_path_buffers(struct btrfs_path *path,
  * in that case this function just updates pointers.
  */
 static int do_relocation(struct btrfs_trans_handle *trans,
+			 struct reloc_control *rc,
 			 struct backref_node *node,
 			 struct btrfs_key *key,
 			 struct btrfs_path *path, int lowest)
@@ -2011,7 +2618,6 @@ static int do_relocation(struct btrfs_trans_handle *trans,
 	u32 blocksize;
 	u64 bytenr;
 	u64 generation;
-	int nr;
 	int slot;
 	int ret;
 	int err = 0;
@@ -2019,18 +2625,25 @@ static int do_relocation(struct btrfs_trans_handle *trans,
 	BUG_ON(lowest && node->eb);
 
 	path->lowest_level = node->level + 1;
+	rc->backref_cache.path[node->level] = node;
 	list_for_each_entry(edge, &node->upper, list[LOWER]) {
 		cond_resched();
-		if (node->eb && node->eb->start == edge->blockptr)
-			continue;
 
 		upper = edge->node[UPPER];
-		root = select_reloc_root(trans, upper, edges, &nr);
-		if (!root)
-			continue;
-
-		if (upper->eb && !upper->locked)
+		root = select_reloc_root(trans, rc, upper, edges);
+		BUG_ON(!root);
+
+		if (upper->eb && !upper->locked) {
+			if (!lowest) {
+				ret = btrfs_bin_search(upper->eb, key,
+						       upper->level, &slot);
+				BUG_ON(ret);
+				bytenr = btrfs_node_blockptr(upper->eb, slot);
+				if (node->eb->start == bytenr)
+					goto next;
+			}
 			drop_node_buffer(upper);
+		}
 
 		if (!upper->eb) {
 			ret = btrfs_search_slot(trans, root, key, path, 0, 1);
@@ -2040,12 +2653,18 @@ static int do_relocation(struct btrfs_trans_handle *trans,
 			}
 			BUG_ON(ret > 0);
 
-			slot = path->slots[upper->level];
+			if (!upper->eb) {
+				upper->eb = path->nodes[upper->level];
+				path->nodes[upper->level] = NULL;
+			} else {
+				BUG_ON(upper->eb != path->nodes[upper->level]);
+			}
 
-			btrfs_unlock_up_safe(path, upper->level + 1);
-			grab_path_buffers(path, upper, edges, nr);
+			upper->locked = 1;
+			path->locks[upper->level] = 0;
 
-			btrfs_release_path(NULL, path);
+			slot = path->slots[upper->level];
+			btrfs_release_path(path);
 		} else {
 			ret = btrfs_bin_search(upper->eb, key, upper->level,
 					       &slot);
@@ -2053,32 +2672,34 @@ static int do_relocation(struct btrfs_trans_handle *trans,
 		}
 
 		bytenr = btrfs_node_blockptr(upper->eb, slot);
-		if (!lowest) {
-			if (node->eb->start == bytenr) {
-				btrfs_tree_unlock(upper->eb);
-				upper->locked = 0;
-				continue;
-			}
+		if (lowest) {
+			BUG_ON(bytenr != node->bytenr);
 		} else {
-			BUG_ON(node->bytenr != bytenr);
+			if (node->eb->start == bytenr)
+				goto next;
 		}
 
 		blocksize = btrfs_level_size(root, node->level);
 		generation = btrfs_node_ptr_generation(upper->eb, slot);
 		eb = read_tree_block(root, bytenr, blocksize, generation);
+		if (!eb || !extent_buffer_uptodate(eb)) {
+			free_extent_buffer(eb);
+			err = -EIO;
+			goto next;
+		}
 		btrfs_tree_lock(eb);
 		btrfs_set_lock_blocking(eb);
 
 		if (!node->eb) {
 			ret = btrfs_cow_block(trans, root, eb, upper->eb,
 					      slot, &eb);
+			btrfs_tree_unlock(eb);
+			free_extent_buffer(eb);
 			if (ret < 0) {
 				err = ret;
-				break;
+				goto next;
 			}
-			btrfs_set_lock_blocking(eb);
-			node->eb = eb;
-			node->locked = 1;
+			BUG_ON(node->eb != eb);
 		} else {
 			btrfs_set_node_blockptr(upper->eb, slot,
 						node->eb->start);
@@ -2090,73 +2711,86 @@ static int do_relocation(struct btrfs_trans_handle *trans,
 						node->eb->start, blocksize,
 						upper->eb->start,
 						btrfs_header_owner(upper->eb),
-						node->level, 0);
+						node->level, 0, 1);
 			BUG_ON(ret);
 
 			ret = btrfs_drop_subtree(trans, root, eb, upper->eb);
 			BUG_ON(ret);
 		}
-		if (!lowest) {
-			btrfs_tree_unlock(upper->eb);
-			upper->locked = 0;
-		}
+next:
+		if (!upper->pending)
+			drop_node_buffer(upper);
+		else
+			unlock_node_buffer(upper);
+		if (err)
+			break;
 	}
+
+	if (!err && node->pending) {
+		drop_node_buffer(node);
+		list_move_tail(&node->list, &rc->backref_cache.changed);
+		node->pending = 0;
+	}
+
 	path->lowest_level = 0;
+	BUG_ON(err == -ENOSPC);
 	return err;
 }
 
 static int link_to_upper(struct btrfs_trans_handle *trans,
+			 struct reloc_control *rc,
 			 struct backref_node *node,
 			 struct btrfs_path *path)
 {
 	struct btrfs_key key;
-	if (!node->eb || list_empty(&node->upper))
-		return 0;
 
 	btrfs_node_key_to_cpu(node->eb, &key, 0);
-	return do_relocation(trans, node, &key, path, 0);
+	return do_relocation(trans, rc, node, &key, path, 0);
 }
 
 static int finish_pending_nodes(struct btrfs_trans_handle *trans,
-				struct backref_cache *cache,
-				struct btrfs_path *path)
+				struct reloc_control *rc,
+				struct btrfs_path *path, int err)
 {
+	LIST_HEAD(list);
+	struct backref_cache *cache = &rc->backref_cache;
 	struct backref_node *node;
 	int level;
 	int ret;
-	int err = 0;
 
 	for (level = 0; level < BTRFS_MAX_LEVEL; level++) {
 		while (!list_empty(&cache->pending[level])) {
 			node = list_entry(cache->pending[level].next,
-					  struct backref_node, lower);
-			BUG_ON(node->level != level);
+					  struct backref_node, list);
+			list_move_tail(&node->list, &list);
+			BUG_ON(!node->pending);
 
-			ret = link_to_upper(trans, node, path);
-			if (ret < 0)
-				err = ret;
-			/*
-			 * this remove the node from the pending list and
-			 * may add some other nodes to the level + 1
-			 * pending list
-			 */
-			remove_backref_node(cache, node);
+			if (!err) {
+				ret = link_to_upper(trans, rc, node, path);
+				if (ret < 0)
+					err = ret;
+			}
 		}
+		list_splice_init(&list, &cache->pending[level]);
 	}
-	BUG_ON(!RB_EMPTY_ROOT(&cache->rb_root));
 	return err;
 }
 
 static void mark_block_processed(struct reloc_control *rc,
-				 struct backref_node *node)
+				 u64 bytenr, u32 blocksize)
+{
+	set_extent_bits(&rc->processed_blocks, bytenr, bytenr + blocksize - 1,
+			EXTENT_DIRTY, GFP_NOFS);
+}
+
+static void __mark_block_processed(struct reloc_control *rc,
+				   struct backref_node *node)
 {
 	u32 blocksize;
 	if (node->level == 0 ||
 	    in_block_group(node->bytenr, rc->block_group)) {
 		blocksize = btrfs_level_size(rc->extent_root, node->level);
-		set_extent_bits(&rc->processed_blocks, node->bytenr,
-				node->bytenr + blocksize - 1, EXTENT_DIRTY,
-				GFP_NOFS);
+		mark_block_processed(rc, node->bytenr, blocksize);
 	}
 	node->processed = 1;
 }
@@ -2179,7 +2813,7 @@ static void update_processed_blocks(struct reloc_control *rc,
 			if (next->processed)
 				break;
 
-			mark_block_processed(rc, next);
+			__mark_block_processed(rc, next);
 
 			if (list_empty(&next->upper))
 				break;
@@ -2202,138 +2836,6 @@ static int tree_block_processed(u64 bytenr, u32 blocksize,
 	return 0;
 }
 
-/*
- * check if there are any file extent pointers in the leaf point to
- * data require processing
- */
-static int check_file_extents(struct reloc_control *rc,
-			      u64 bytenr, u32 blocksize, u64 ptr_gen)
-{
-	struct btrfs_key found_key;
-	struct btrfs_file_extent_item *fi;
-	struct extent_buffer *leaf;
-	u32 nritems;
-	int i;
-	int ret = 0;
-
-	leaf = read_tree_block(rc->extent_root, bytenr, blocksize, ptr_gen);
-
-	nritems = btrfs_header_nritems(leaf);
-	for (i = 0; i < nritems; i++) {
-		cond_resched();
-		btrfs_item_key_to_cpu(leaf, &found_key, i);
-		if (found_key.type != BTRFS_EXTENT_DATA_KEY)
-			continue;
-		fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
-		if (btrfs_file_extent_type(leaf, fi) ==
-		    BTRFS_FILE_EXTENT_INLINE)
-			continue;
-		bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
-		if (bytenr == 0)
-			continue;
-		if (in_block_group(bytenr, rc->block_group)) {
-			ret = 1;
-			break;
-		}
-	}
-	free_extent_buffer(leaf);
-	return ret;
-}
-
-/*
- * scan child blocks of a given block to find blocks require processing
- */
-static int add_child_blocks(struct btrfs_trans_handle *trans,
-			    struct reloc_control *rc,
-			    struct backref_node *node,
-			    struct rb_root *blocks)
-{
-	struct tree_block *block;
-	struct rb_node *rb_node;
-	u64 bytenr;
-	u64 ptr_gen;
-	u32 blocksize;
-	u32 nritems;
-	int i;
-	int err = 0;
-
-	nritems = btrfs_header_nritems(node->eb);
-	blocksize = btrfs_level_size(rc->extent_root, node->level - 1);
-	for (i = 0; i < nritems; i++) {
-		cond_resched();
-		bytenr = btrfs_node_blockptr(node->eb, i);
-		ptr_gen = btrfs_node_ptr_generation(node->eb, i);
-		if (ptr_gen == trans->transid)
-			continue;
-		if (!in_block_group(bytenr, rc->block_group) &&
-		    (node->level > 1 || rc->stage == MOVE_DATA_EXTENTS))
-			continue;
-		if (tree_block_processed(bytenr, blocksize, rc))
-			continue;
-
-		readahead_tree_block(rc->extent_root,
-				     bytenr, blocksize, ptr_gen);
-	}
-
-	for (i = 0; i < nritems; i++) {
-		cond_resched();
-		bytenr = btrfs_node_blockptr(node->eb, i);
-		ptr_gen = btrfs_node_ptr_generation(node->eb, i);
-		if (ptr_gen == trans->transid)
-			continue;
-		if (!in_block_group(bytenr, rc->block_group) &&
-		    (node->level > 1 || rc->stage == MOVE_DATA_EXTENTS))
-			continue;
-		if (tree_block_processed(bytenr, blocksize, rc))
-			continue;
-		if (!in_block_group(bytenr, rc->block_group) &&
-		    !check_file_extents(rc, bytenr, blocksize, ptr_gen))
-			continue;
-
-		block = kmalloc(sizeof(*block), GFP_NOFS);
-		if (!block) {
-			err = -ENOMEM;
-			break;
-		}
-		block->bytenr = bytenr;
-		btrfs_node_key_to_cpu(node->eb, &block->key, i);
-		block->level = node->level - 1;
-		block->key_ready = 1;
-		rb_node = tree_insert(blocks, block->bytenr, &block->rb_node);
-		BUG_ON(rb_node);
-	}
-	if (err)
-		free_block_list(blocks);
-	return err;
-}
-
-/*
- * find adjacent blocks require processing
- */
-static noinline_for_stack
-int add_adjacent_blocks(struct btrfs_trans_handle *trans,
-			struct reloc_control *rc,
-			struct backref_cache *cache,
-			struct rb_root *blocks, int level,
-			struct backref_node **upper)
-{
-	struct backref_node *node;
-	int ret = 0;
-
-	WARN_ON(!list_empty(&cache->pending[level]));
-
-	if (list_empty(&cache->pending[level + 1]))
-		return 1;
-
-	node = list_entry(cache->pending[level + 1].next,
-			  struct backref_node, lower);
-	if (node->eb)
-		ret = add_child_blocks(trans, rc, node, blocks);
-
-	*upper = node;
-	return ret;
-}
-
 static int get_tree_block_key(struct reloc_control *rc,
 			      struct tree_block *block)
 {
@@ -2342,6 +2844,10 @@ static int get_tree_block_key(struct reloc_control *rc,
 	BUG_ON(block->key_ready);
 	eb = read_tree_block(rc->extent_root, block->bytenr,
 			     block->key.objectid, block->key.offset);
+	if (!eb || !extent_buffer_uptodate(eb)) {
+		free_extent_buffer(eb);
+		return -EIO;
+	}
 	WARN_ON(btrfs_header_level(eb) != block->level);
 	if (block->level == 0)
 		btrfs_item_key_to_cpu(eb, &block->key, 0);
@@ -2356,8 +2862,13 @@ static int reada_tree_block(struct reloc_control *rc,
 			    struct tree_block *block)
 {
 	BUG_ON(block->key_ready);
-	readahead_tree_block(rc->extent_root, block->bytenr,
-			     block->key.objectid, block->key.offset);
+	if (block->key.type == BTRFS_METADATA_ITEM_KEY)
+		readahead_tree_block(rc->extent_root, block->bytenr,
+				     block->key.objectid,
+				     rc->extent_root->leafsize);
+	else
+		readahead_tree_block(rc->extent_root, block->bytenr,
+				     block->key.objectid, block->key.offset);
 	return 0;
 }
 
@@ -2371,40 +2882,48 @@ static int relocate_tree_block(struct btrfs_trans_handle *trans,
 				struct btrfs_path *path)
 {
 	struct btrfs_root *root;
-	int ret;
+	int ret = 0;
 
+	if (!node)
+		return 0;
+
+	BUG_ON(node->processed);
 	root = select_one_root(trans, node);
-	if (unlikely(!root)) {
-		rc->found_old_snapshot = 1;
+	if (root == ERR_PTR(-ENOENT)) {
 		update_processed_blocks(rc, node);
-		return 0;
+		goto out;
 	}
 
-	if (root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID) {
-		ret = do_relocation(trans, node, key, path, 1);
-		if (ret < 0)
-			goto out;
-		if (node->level == 0 && rc->stage == UPDATE_DATA_PTRS) {
-			ret = replace_file_extents(trans, rc, root,
-						   node->eb, NULL);
-			if (ret < 0)
-				goto out;
-		}
-		drop_node_buffer(node);
-	} else if (!root->ref_cows) {
-		path->lowest_level = node->level;
-		ret = btrfs_search_slot(trans, root, key, path, 0, 1);
-		btrfs_release_path(root, path);
-		if (ret < 0)
+	if (!root || test_bit(BTRFS_ROOT_REF_COWS, &root->state)) {
+		ret = reserve_metadata_space(trans, rc, node);
+		if (ret)
 			goto out;
-	} else if (root != node->root) {
-		WARN_ON(node->level > 0 || rc->stage != UPDATE_DATA_PTRS);
 	}
 
-	update_processed_blocks(rc, node);
-	ret = 0;
+	if (root) {
+		if (test_bit(BTRFS_ROOT_REF_COWS, &root->state)) {
+			BUG_ON(node->new_bytenr);
+			BUG_ON(!list_empty(&node->list));
+			btrfs_record_root_in_trans(trans, root);
+			root = root->reloc_root;
+			node->new_bytenr = root->node->start;
+			node->root = root;
+			list_add_tail(&node->list, &rc->backref_cache.changed);
+		} else {
+			path->lowest_level = node->level;
+			ret = btrfs_search_slot(trans, root, key, path, 0, 1);
+			btrfs_release_path(path);
+			if (ret > 0)
+				ret = 0;
+		}
+		if (!ret)
+			update_processed_blocks(rc, node);
+	} else {
+		ret = do_relocation(trans, rc, node, key, path, 1);
+	}
 out:
-	drop_node_buffer(node);
+	if (ret || node->level == 0 || node->cowonly)
+		remove_backref_node(&rc->backref_cache, node);
 	return ret;
 }
 
@@ -2415,34 +2934,22 @@ static noinline_for_stack
 int relocate_tree_blocks(struct btrfs_trans_handle *trans,
 			 struct reloc_control *rc, struct rb_root *blocks)
 {
-	struct backref_cache *cache;
 	struct backref_node *node;
 	struct btrfs_path *path;
 	struct tree_block *block;
 	struct rb_node *rb_node;
-	int level = -1;
 	int ret;
 	int err = 0;
 
 	path = btrfs_alloc_path();
-	if (!path)
-		return -ENOMEM;
-
-	cache = kmalloc(sizeof(*cache), GFP_NOFS);
-	if (!cache) {
-		btrfs_free_path(path);
-		return -ENOMEM;
+	if (!path) {
+		err = -ENOMEM;
+		goto out_free_blocks;
 	}
 
-	backref_cache_init(cache);
-
 	rb_node = rb_first(blocks);
 	while (rb_node) {
 		block = rb_entry(rb_node, struct tree_block, rb_node);
-		if (level == -1)
-			level = block->level;
-		else
-			BUG_ON(level != block->level);
 		if (!block->key_ready)
 			reada_tree_block(rc, block);
 		rb_node = rb_next(rb_node);
@@ -2451,8 +2958,11 @@ int relocate_tree_blocks(struct btrfs_trans_handle *trans,
 	rb_node = rb_first(blocks);
 	while (rb_node) {
 		block = rb_entry(rb_node, struct tree_block, rb_node);
-		if (!block->key_ready)
-			get_tree_block_key(rc, block);
+		if (!block->key_ready) {
+			err = get_tree_block_key(rc, block);
+			if (err)
+				goto out_free_path;
+		}
 		rb_node = rb_next(rb_node);
 	}
 
@@ -2460,7 +2970,7 @@ int relocate_tree_blocks(struct btrfs_trans_handle *trans,
 	while (rb_node) {
 		block = rb_entry(rb_node, struct tree_block, rb_node);
 
-		node = build_backref_tree(rc, cache, &block->key,
+		node = build_backref_tree(rc, &block->key,
 					  block->level, block->bytenr);
 		if (IS_ERR(node)) {
 			err = PTR_ERR(node);
@@ -2470,79 +2980,64 @@ int relocate_tree_blocks(struct btrfs_trans_handle *trans,
 		ret = relocate_tree_block(trans, rc, node, &block->key,
 					  path);
 		if (ret < 0) {
-			err = ret;
+			if (ret != -EAGAIN || rb_node == rb_first(blocks))
+				err = ret;
 			goto out;
 		}
-		remove_backref_node(cache, node);
 		rb_node = rb_next(rb_node);
 	}
+out:
+	err = finish_pending_nodes(trans, rc, path, err);
 
-	if (level > 0)
-		goto out;
-
+out_free_path:
+	btrfs_free_path(path);
+out_free_blocks:
 	free_block_list(blocks);
+	return err;
+}
 
-	/*
-	 * now backrefs of some upper level tree blocks have been cached,
-	 * try relocating blocks referenced by these upper level blocks.
-	 */
-	while (1) {
-		struct backref_node *upper = NULL;
-		if (trans->transaction->in_commit ||
-		    trans->transaction->delayed_refs.flushing)
-			break;
-
-		ret = add_adjacent_blocks(trans, rc, cache, blocks, level,
-					  &upper);
-		if (ret < 0)
-			err = ret;
-		if (ret != 0)
-			break;
+static noinline_for_stack
+int prealloc_file_extent_cluster(struct inode *inode,
+				 struct file_extent_cluster *cluster)
+{
+	u64 alloc_hint = 0;
+	u64 start;
+	u64 end;
+	u64 offset = BTRFS_I(inode)->index_cnt;
+	u64 num_bytes;
+	int nr = 0;
+	int ret = 0;
 
-		rb_node = rb_first(blocks);
-		while (rb_node) {
-			block = rb_entry(rb_node, struct tree_block, rb_node);
-			if (trans->transaction->in_commit ||
-			    trans->transaction->delayed_refs.flushing)
-				goto out;
-			BUG_ON(!block->key_ready);
-			node = build_backref_tree(rc, cache, &block->key,
-						  level, block->bytenr);
-			if (IS_ERR(node)) {
-				err = PTR_ERR(node);
-				goto out;
-			}
+	BUG_ON(cluster->start != cluster->boundary[0]);
+	mutex_lock(&inode->i_mutex);
 
-			ret = relocate_tree_block(trans, rc, node,
-						  &block->key, path);
-			if (ret < 0) {
-				err = ret;
-				goto out;
-			}
-			remove_backref_node(cache, node);
-			rb_node = rb_next(rb_node);
-		}
-		free_block_list(blocks);
+	ret = btrfs_check_data_free_space(inode, cluster->end +
+					  1 - cluster->start);
+	if (ret)
+		goto out;
 
-		if (upper) {
-			ret = link_to_upper(trans, upper, path);
-			if (ret < 0) {
-				err = ret;
-				break;
-			}
-			remove_backref_node(cache, upper);
-		}
+	while (nr < cluster->nr) {
+		start = cluster->boundary[nr] - offset;
+		if (nr + 1 < cluster->nr)
+			end = cluster->boundary[nr + 1] - 1 - offset;
+		else
+			end = cluster->end - offset;
+
+		lock_extent(&BTRFS_I(inode)->io_tree, start, end);
+		num_bytes = end + 1 - start;
+		ret = btrfs_prealloc_file_range(inode, 0, start,
+						num_bytes, num_bytes,
+						end + 1, &alloc_hint);
+		unlock_extent(&BTRFS_I(inode)->io_tree, start, end);
+		if (ret)
+			break;
+		nr++;
 	}
+	btrfs_free_reserved_data_space(inode, cluster->end +
+				       1 - cluster->start);
 out:
-	free_block_list(blocks);
-
-	ret = finish_pending_nodes(trans, cache, path);
-	if (ret < 0)
-		err = ret;
-
-	kfree(cache);
-	btrfs_free_path(path);
-	return err;
+	mutex_unlock(&inode->i_mutex);
+	return ret;
 }
 
 static noinline_for_stack
@@ -2554,7 +3049,7 @@ int setup_extent_mapping(struct inode *inode, u64 start, u64 end,
 	struct extent_map *em;
 	int ret = 0;
 
-	em = alloc_extent_map(GFP_NOFS);
+	em = alloc_extent_map();
 	if (!em)
 		return -ENOMEM;
 
@@ -2565,10 +3060,10 @@ int setup_extent_mapping(struct inode *inode, u64 start, u64 end,
 	em->bdev = root->fs_info->fs_devices->latest_bdev;
 	set_bit(EXTENT_FLAG_PINNED, &em->flags);
 
-	lock_extent(&BTRFS_I(inode)->io_tree, start, end, GFP_NOFS);
+	lock_extent(&BTRFS_I(inode)->io_tree, start, end);
 	while (1) {
 		write_lock(&em_tree->lock);
-		ret = add_extent_mapping(em_tree, em);
+		ret = add_extent_mapping(em_tree, em, 0);
 		write_unlock(&em_tree->lock);
 		if (ret != -EEXIST) {
 			free_extent_map(em);
@@ -2576,7 +3071,7 @@ int setup_extent_mapping(struct inode *inode, u64 start, u64 end,
 		}
 		btrfs_drop_extent_cache(inode, start, end, 0);
 	}
-	unlock_extent(&BTRFS_I(inode)->io_tree, start, end, GFP_NOFS);
+	unlock_extent(&BTRFS_I(inode)->io_tree, start, end);
 	return ret;
 }
 
@@ -2588,9 +3083,9 @@ static int relocate_file_extent_cluster(struct inode *inode,
 	u64 offset = BTRFS_I(inode)->index_cnt;
 	unsigned long index;
 	unsigned long last_index;
-	unsigned int dirty_page = 0;
 	struct page *page;
 	struct file_ra_state *ra;
+	gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
 	int nr = 0;
 	int ret = 0;
 
@@ -2601,30 +3096,36 @@ static int relocate_file_extent_cluster(struct inode *inode,
 	if (!ra)
 		return -ENOMEM;
 
-	index = (cluster->start - offset) >> PAGE_CACHE_SHIFT;
-	last_index = (cluster->end - offset) >> PAGE_CACHE_SHIFT;
+	ret = prealloc_file_extent_cluster(inode, cluster);
+	if (ret)
+		goto out;
 
-	mutex_lock(&inode->i_mutex);
+	file_ra_state_init(ra, inode->i_mapping);
 
-	i_size_write(inode, cluster->end + 1 - offset);
 	ret = setup_extent_mapping(inode, cluster->start - offset,
 				   cluster->end - offset, cluster->start);
 	if (ret)
-		goto out_unlock;
-
-	file_ra_state_init(ra, inode->i_mapping);
+		goto out;
 
-	WARN_ON(cluster->start != cluster->boundary[0]);
+	index = (cluster->start - offset) >> PAGE_CACHE_SHIFT;
+	last_index = (cluster->end - offset) >> PAGE_CACHE_SHIFT;
 	while (index <= last_index) {
+		ret = btrfs_delalloc_reserve_metadata(inode, PAGE_CACHE_SIZE);
+		if (ret)
+			goto out;
+
 		page = find_lock_page(inode->i_mapping, index);
 		if (!page) {
 			page_cache_sync_readahead(inode->i_mapping,
 						  ra, NULL, index,
 						  last_index + 1 - index);
-			page = grab_cache_page(inode->i_mapping, index);
+			page = find_or_create_page(inode->i_mapping, index,
+						   mask);
 			if (!page) {
+				btrfs_delalloc_release_metadata(inode,
+							PAGE_CACHE_SIZE);
 				ret = -ENOMEM;
-				goto out_unlock;
+				goto out;
 			}
 		}
 
@@ -2640,16 +3141,17 @@ static int relocate_file_extent_cluster(struct inode *inode,
 			if (!PageUptodate(page)) {
 				unlock_page(page);
 				page_cache_release(page);
+				btrfs_delalloc_release_metadata(inode,
+							PAGE_CACHE_SIZE);
 				ret = -EIO;
-				goto out_unlock;
+				goto out;
 			}
 		}
 
-		page_start = (u64)page->index << PAGE_CACHE_SHIFT;
+		page_start = page_offset(page);
 		page_end = page_start + PAGE_CACHE_SIZE - 1;
 
-		lock_extent(&BTRFS_I(inode)->io_tree,
-			    page_start, page_end, GFP_NOFS);
+		lock_extent(&BTRFS_I(inode)->io_tree, page_start, page_end);
 
 		set_page_extent_mapped(page);
 
@@ -2660,31 +3162,21 @@ static int relocate_file_extent_cluster(struct inode *inode,
 					EXTENT_BOUNDARY, GFP_NOFS);
 			nr++;
 		}
-		btrfs_set_extent_delalloc(inode, page_start, page_end, NULL);
 
+		btrfs_set_extent_delalloc(inode, page_start, page_end, NULL);
 		set_page_dirty(page);
-		dirty_page++;
 
 		unlock_extent(&BTRFS_I(inode)->io_tree,
-			      page_start, page_end, GFP_NOFS);
+			      page_start, page_end);
 		unlock_page(page);
 		page_cache_release(page);
 
 		index++;
-		if (nr < cluster->nr &&
-		    page_end + 1 + offset == cluster->boundary[nr]) {
-			balance_dirty_pages_ratelimited_nr(inode->i_mapping,
-							   dirty_page);
-			dirty_page = 0;
-		}
-	}
-	if (dirty_page) {
-		balance_dirty_pages_ratelimited_nr(inode->i_mapping,
-						   dirty_page);
+		balance_dirty_pages_ratelimited(inode->i_mapping);
+		btrfs_throttle(BTRFS_I(inode)->root);
 	}
 	WARN_ON(nr != cluster->nr);
-out_unlock:
-	mutex_unlock(&inode->i_mutex);
+out:
 	kfree(ra);
 	return ret;
 }
@@ -2777,17 +3269,22 @@ static int add_tree_block(struct reloc_control *rc,
 	struct rb_node *rb_node;
 	u32 item_size;
 	int level = -1;
-	int generation;
+	u64 generation;
 
 	eb =  path->nodes[0];
 	item_size = btrfs_item_size_nr(eb, path->slots[0]);
 
-	if (item_size >= sizeof(*ei) + sizeof(*bi)) {
+	if (extent_key->type == BTRFS_METADATA_ITEM_KEY ||
+	    item_size >= sizeof(*ei) + sizeof(*bi)) {
 		ei = btrfs_item_ptr(eb, path->slots[0],
 				struct btrfs_extent_item);
-		bi = (struct btrfs_tree_block_info *)(ei + 1);
+		if (extent_key->type == BTRFS_EXTENT_ITEM_KEY) {
+			bi = (struct btrfs_tree_block_info *)(ei + 1);
+			level = btrfs_tree_block_level(eb, bi);
+		} else {
+			level = (int)extent_key->offset;
+		}
 		generation = btrfs_extent_generation(eb, ei);
-		level = btrfs_tree_block_level(eb, bi);
 	} else {
 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
 		u64 ref_owner;
@@ -2796,6 +3293,8 @@ static int add_tree_block(struct reloc_control *rc,
 		BUG_ON(item_size != sizeof(struct btrfs_extent_item_v0));
 		ret = get_ref_objectid_v0(rc, path, extent_key,
 					  &ref_owner, NULL);
+		if (ret < 0)
+			return ret;
 		BUG_ON(ref_owner >= BTRFS_MAX_LEVEL);
 		level = (int)ref_owner;
 		/* FIXME: get real generation */
@@ -2805,7 +3304,7 @@ static int add_tree_block(struct reloc_control *rc,
 #endif
 	}
 
-	btrfs_release_path(rc->extent_root, path);
+	btrfs_release_path(path);
 
 	BUG_ON(level == -1);
 
@@ -2814,13 +3313,14 @@ static int add_tree_block(struct reloc_control *rc,
 		return -ENOMEM;
 
 	block->bytenr = extent_key->objectid;
-	block->key.objectid = extent_key->offset;
+	block->key.objectid = rc->extent_root->leafsize;
 	block->key.offset = generation;
 	block->level = level;
 	block->key_ready = 0;
 
 	rb_node = tree_insert(blocks, block->bytenr, &block->rb_node);
-	BUG_ON(rb_node);
+	if (rb_node)
+		backref_tree_panic(rb_node, -EEXIST, block->bytenr);
 
 	return 0;
 }
@@ -2835,6 +3335,8 @@ static int __add_tree_block(struct reloc_control *rc,
 	struct btrfs_path *path;
 	struct btrfs_key key;
 	int ret;
+	bool skinny = btrfs_fs_incompat(rc->extent_root->fs_info,
+					SKINNY_METADATA);
 
 	if (tree_block_processed(bytenr, blocksize, rc))
 		return 0;
@@ -2845,19 +3347,42 @@ static int __add_tree_block(struct reloc_control *rc,
 	path = btrfs_alloc_path();
 	if (!path)
 		return -ENOMEM;
-
+again:
 	key.objectid = bytenr;
-	key.type = BTRFS_EXTENT_ITEM_KEY;
-	key.offset = blocksize;
+	if (skinny) {
+		key.type = BTRFS_METADATA_ITEM_KEY;
+		key.offset = (u64)-1;
+	} else {
+		key.type = BTRFS_EXTENT_ITEM_KEY;
+		key.offset = blocksize;
+	}
 
 	path->search_commit_root = 1;
 	path->skip_locking = 1;
 	ret = btrfs_search_slot(NULL, rc->extent_root, &key, path, 0, 0);
 	if (ret < 0)
 		goto out;
+
+	if (ret > 0 && skinny) {
+		if (path->slots[0]) {
+			path->slots[0]--;
+			btrfs_item_key_to_cpu(path->nodes[0], &key,
+					      path->slots[0]);
+			if (key.objectid == bytenr &&
+			    (key.type == BTRFS_METADATA_ITEM_KEY ||
+			     (key.type == BTRFS_EXTENT_ITEM_KEY &&
+			      key.offset == blocksize)))
+				ret = 0;
+		}
+
+		if (ret) {
+			skinny = false;
+			btrfs_release_path(path);
+			goto again;
+		}
+	}
 	BUG_ON(ret);
 
-	btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
 	ret = add_tree_block(rc, &key, path, blocks);
 out:
 	btrfs_free_path(path);
@@ -2870,9 +3395,6 @@ out:
 static int block_use_full_backref(struct reloc_control *rc,
 				  struct extent_buffer *eb)
 {
-	struct btrfs_path *path;
-	struct btrfs_extent_item *ei;
-	struct btrfs_key key;
 	u64 flags;
 	int ret;
 
@@ -2880,28 +3402,58 @@ static int block_use_full_backref(struct reloc_control *rc,
 	    btrfs_header_backref_rev(eb) < BTRFS_MIXED_BACKREF_REV)
 		return 1;
 
-	path = btrfs_alloc_path();
-	BUG_ON(!path);
-
-	key.objectid = eb->start;
-	key.type = BTRFS_EXTENT_ITEM_KEY;
-	key.offset = eb->len;
-
-	path->search_commit_root = 1;
-	path->skip_locking = 1;
-	ret = btrfs_search_slot(NULL, rc->extent_root,
-				&key, path, 0, 0);
+	ret = btrfs_lookup_extent_info(NULL, rc->extent_root,
+				       eb->start, btrfs_header_level(eb), 1,
+				       NULL, &flags);
 	BUG_ON(ret);
 
-	ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
-			    struct btrfs_extent_item);
-	flags = btrfs_extent_flags(path->nodes[0], ei);
-	BUG_ON(!(flags & BTRFS_EXTENT_FLAG_TREE_BLOCK));
 	if (flags & BTRFS_BLOCK_FLAG_FULL_BACKREF)
 		ret = 1;
 	else
 		ret = 0;
-	btrfs_free_path(path);
+	return ret;
+}
+
+static int delete_block_group_cache(struct btrfs_fs_info *fs_info,
+				    struct inode *inode, u64 ino)
+{
+	struct btrfs_key key;
+	struct btrfs_root *root = fs_info->tree_root;
+	struct btrfs_trans_handle *trans;
+	int ret = 0;
+
+	if (inode)
+		goto truncate;
+
+	key.objectid = ino;
+	key.type = BTRFS_INODE_ITEM_KEY;
+	key.offset = 0;
+
+	inode = btrfs_iget(fs_info->sb, &key, root, NULL);
+	if (IS_ERR(inode) || is_bad_inode(inode)) {
+		if (!IS_ERR(inode))
+			iput(inode);
+		return -ENOENT;
+	}
+
+truncate:
+	ret = btrfs_check_trunc_cache_free_space(root,
+						 &fs_info->global_block_rsv);
+	if (ret)
+		goto out;
+
+	trans = btrfs_join_transaction(root);
+	if (IS_ERR(trans)) {
+		ret = PTR_ERR(trans);
+		goto out;
+	}
+
+	ret = btrfs_truncate_free_space_cache(root, trans, inode);
+
+	btrfs_end_transaction(trans, root);
+	btrfs_btree_balance_dirty(root);
+out:
+	iput(inode);
 	return ret;
 }
 
@@ -2931,15 +3483,28 @@ static int find_data_references(struct reloc_control *rc,
 	int counted;
 	int ret;
 
-	path = btrfs_alloc_path();
-	if (!path)
-		return -ENOMEM;
-
 	ref_root = btrfs_extent_data_ref_root(leaf, ref);
 	ref_objectid = btrfs_extent_data_ref_objectid(leaf, ref);
 	ref_offset = btrfs_extent_data_ref_offset(leaf, ref);
 	ref_count = btrfs_extent_data_ref_count(leaf, ref);
 
+	/*
+	 * This is an extent belonging to the free space cache, lets just delete
+	 * it and redo the search.
+	 */
+	if (ref_root == BTRFS_ROOT_TREE_OBJECTID) {
+		ret = delete_block_group_cache(rc->extent_root->fs_info,
+					       NULL, ref_objectid);
+		if (ret != -ENOENT)
+			return ret;
+		ret = 0;
+	}
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+	path->reada = 1;
+
 	root = read_fs_root(rc->extent_root->fs_info, ref_root);
 	if (IS_ERR(root)) {
 		err = PTR_ERR(root);
@@ -2947,8 +3512,11 @@ static int find_data_references(struct reloc_control *rc,
 	}
 
 	key.objectid = ref_objectid;
-	key.offset = ref_offset;
 	key.type = BTRFS_EXTENT_DATA_KEY;
+	if (ref_offset > ((u64)-1 << 32))
+		key.offset = 0;
+	else
+		key.offset = ref_offset;
 
 	path->search_commit_root = 1;
 	path->skip_locking = 1;
@@ -2983,10 +3551,8 @@ static int find_data_references(struct reloc_control *rc,
 				err = ret;
 				goto out;
 			}
-			if (ret > 0) {
-				WARN_ON(1);
+			if (WARN_ON(ret > 0))
 				goto out;
-			}
 
 			leaf = path->nodes[0];
 			nritems = btrfs_header_nritems(leaf);
@@ -3006,11 +3572,9 @@ static int find_data_references(struct reloc_control *rc,
 		}
 
 		btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
-		if (key.objectid != ref_objectid ||
-		    key.type != BTRFS_EXTENT_DATA_KEY) {
-			WARN_ON(1);
+		if (WARN_ON(key.objectid != ref_objectid ||
+		    key.type != BTRFS_EXTENT_DATA_KEY))
 			break;
-		}
 
 		fi = btrfs_item_ptr(leaf, path->slots[0],
 				    struct btrfs_file_extent_item);
@@ -3044,7 +3608,9 @@ static int find_data_references(struct reloc_control *rc,
 			block->key_ready = 1;
 			rb_node = tree_insert(blocks, block->bytenr,
 					      &block->rb_node);
-			BUG_ON(rb_node);
+			if (rb_node)
+				backref_tree_panic(rb_node, -EEXIST,
+						   block->bytenr);
 		}
 		if (counted)
 			added = 1;
@@ -3060,7 +3626,7 @@ out:
 }
 
 /*
- * hepler to find all tree blocks that reference a given data extent
+ * helper to find all tree blocks that reference a given data extent
  */
 static noinline_for_stack
 int add_data_references(struct reloc_control *rc,
@@ -3074,22 +3640,10 @@ int add_data_references(struct reloc_control *rc,
 	struct btrfs_extent_inline_ref *iref;
 	unsigned long ptr;
 	unsigned long end;
-	u32 blocksize;
-	int ret;
+	u32 blocksize = btrfs_level_size(rc->extent_root, 0);
+	int ret = 0;
 	int err = 0;
 
-	ret = get_new_location(rc->data_inode, NULL, extent_key->objectid,
-			       extent_key->offset);
-	BUG_ON(ret < 0);
-	if (ret > 0) {
-		/* the relocated data is fragmented */
-		rc->extents_skipped++;
-		btrfs_release_path(rc->extent_root, path);
-		return 0;
-	}
-
-	blocksize = btrfs_level_size(rc->extent_root, 0);
-
 	eb = path->nodes[0];
 	ptr = btrfs_item_ptr_offset(eb, path->slots[0]);
 	end = ptr + btrfs_item_size_nr(eb, path->slots[0]);
@@ -3114,6 +3668,10 @@ int add_data_references(struct reloc_control *rc,
 		} else {
 			BUG();
 		}
+		if (ret) {
+			err = ret;
+			goto out;
+		}
 		ptr += btrfs_extent_inline_ref_size(key.type);
 	}
 	WARN_ON(ptr > end);
@@ -3159,18 +3717,20 @@ int add_data_references(struct reloc_control *rc,
 		}
 		path->slots[0]++;
 	}
-	btrfs_release_path(rc->extent_root, path);
+out:
+	btrfs_release_path(path);
 	if (err)
 		free_block_list(blocks);
 	return err;
 }
 
 /*
- * hepler to find next unprocessed extent
+ * helper to find next unprocessed extent
  */
 static noinline_for_stack
 int find_next_extent(struct btrfs_trans_handle *trans,
-		     struct reloc_control *rc, struct btrfs_path *path)
+		     struct reloc_control *rc, struct btrfs_path *path,
+		     struct btrfs_key *extent_key)
 {
 	struct btrfs_key key;
 	struct extent_buffer *leaf;
@@ -3210,42 +3770,62 @@ next:
 			break;
 		}
 
-		if (key.type != BTRFS_EXTENT_ITEM_KEY ||
+		if (key.type != BTRFS_EXTENT_ITEM_KEY &&
+		    key.type != BTRFS_METADATA_ITEM_KEY) {
+			path->slots[0]++;
+			goto next;
+		}
+
+		if (key.type == BTRFS_EXTENT_ITEM_KEY &&
 		    key.objectid + key.offset <= rc->search_start) {
 			path->slots[0]++;
 			goto next;
 		}
 
+		if (key.type == BTRFS_METADATA_ITEM_KEY &&
+		    key.objectid + rc->extent_root->leafsize <=
+		    rc->search_start) {
+			path->slots[0]++;
+			goto next;
+		}
+
 		ret = find_first_extent_bit(&rc->processed_blocks,
 					    key.objectid, &start, &end,
-					    EXTENT_DIRTY);
+					    EXTENT_DIRTY, NULL);
 
 		if (ret == 0 && start <= key.objectid) {
-			btrfs_release_path(rc->extent_root, path);
+			btrfs_release_path(path);
 			rc->search_start = end + 1;
 		} else {
-			rc->search_start = key.objectid + key.offset;
+			if (key.type == BTRFS_EXTENT_ITEM_KEY)
+				rc->search_start = key.objectid + key.offset;
+			else
+				rc->search_start = key.objectid +
+					rc->extent_root->leafsize;
+			memcpy(extent_key, &key, sizeof(key));
 			return 0;
 		}
 	}
-	btrfs_release_path(rc->extent_root, path);
+	btrfs_release_path(path);
 	return ret;
 }
 
 static void set_reloc_control(struct reloc_control *rc)
 {
 	struct btrfs_fs_info *fs_info = rc->extent_root->fs_info;
-	mutex_lock(&fs_info->trans_mutex);
+
+	mutex_lock(&fs_info->reloc_mutex);
 	fs_info->reloc_ctl = rc;
-	mutex_unlock(&fs_info->trans_mutex);
+	mutex_unlock(&fs_info->reloc_mutex);
 }
 
 static void unset_reloc_control(struct reloc_control *rc)
 {
 	struct btrfs_fs_info *fs_info = rc->extent_root->fs_info;
-	mutex_lock(&fs_info->trans_mutex);
+
+	mutex_lock(&fs_info->reloc_mutex);
 	fs_info->reloc_ctl = NULL;
-	mutex_unlock(&fs_info->trans_mutex);
+	mutex_unlock(&fs_info->reloc_mutex);
 }
 
 static int check_extent_flags(u64 flags)
@@ -3262,48 +3842,89 @@ static int check_extent_flags(u64 flags)
 	return 0;
 }
 
+static noinline_for_stack
+int prepare_to_relocate(struct reloc_control *rc)
+{
+	struct btrfs_trans_handle *trans;
+
+	rc->block_rsv = btrfs_alloc_block_rsv(rc->extent_root,
+					      BTRFS_BLOCK_RSV_TEMP);
+	if (!rc->block_rsv)
+		return -ENOMEM;
+
+	memset(&rc->cluster, 0, sizeof(rc->cluster));
+	rc->search_start = rc->block_group->key.objectid;
+	rc->extents_found = 0;
+	rc->nodes_relocated = 0;
+	rc->merging_rsv_size = 0;
+	rc->reserved_bytes = 0;
+	rc->block_rsv->size = rc->extent_root->nodesize *
+			      RELOCATION_RESERVED_NODES;
+
+	rc->create_reloc_tree = 1;
+	set_reloc_control(rc);
+
+	trans = btrfs_join_transaction(rc->extent_root);
+	if (IS_ERR(trans)) {
+		unset_reloc_control(rc);
+		/*
+		 * extent tree is not a ref_cow tree and has no reloc_root to
+		 * cleanup.  And callers are responsible to free the above
+		 * block rsv.
+		 */
+		return PTR_ERR(trans);
+	}
+	btrfs_commit_transaction(trans, rc->extent_root);
+	return 0;
+}
 
 static noinline_for_stack int relocate_block_group(struct reloc_control *rc)
 {
 	struct rb_root blocks = RB_ROOT;
 	struct btrfs_key key;
-	struct file_extent_cluster *cluster;
 	struct btrfs_trans_handle *trans = NULL;
 	struct btrfs_path *path;
 	struct btrfs_extent_item *ei;
-	unsigned long nr;
 	u64 flags;
 	u32 item_size;
 	int ret;
 	int err = 0;
-
-	cluster = kzalloc(sizeof(*cluster), GFP_NOFS);
-	if (!cluster)
-		return -ENOMEM;
+	int progress = 0;
 
 	path = btrfs_alloc_path();
-	if (!path) {
-		kfree(cluster);
+	if (!path)
 		return -ENOMEM;
-	}
-
-	rc->extents_found = 0;
-	rc->extents_skipped = 0;
-
-	rc->search_start = rc->block_group->key.objectid;
-	clear_extent_bits(&rc->processed_blocks, 0, (u64)-1, EXTENT_DIRTY,
-			  GFP_NOFS);
-
-	rc->create_reloc_root = 1;
-	set_reloc_control(rc);
+	path->reada = 1;
 
-	trans = btrfs_start_transaction(rc->extent_root, 1);
-	btrfs_commit_transaction(trans, rc->extent_root);
+	ret = prepare_to_relocate(rc);
+	if (ret) {
+		err = ret;
+		goto out_free;
+	}
 
 	while (1) {
-		trans = btrfs_start_transaction(rc->extent_root, 1);
+		rc->reserved_bytes = 0;
+		ret = btrfs_block_rsv_refill(rc->extent_root,
+					rc->block_rsv, rc->block_rsv->size,
+					BTRFS_RESERVE_FLUSH_ALL);
+		if (ret) {
+			err = ret;
+			break;
+		}
+		progress++;
+		trans = btrfs_start_transaction(rc->extent_root, 0);
+		if (IS_ERR(trans)) {
+			err = PTR_ERR(trans);
+			trans = NULL;
+			break;
+		}
+restart:
+		if (update_backref_cache(trans, &rc->backref_cache)) {
+			btrfs_end_transaction(trans, rc->extent_root);
+			continue;
+		}
 
-		ret = find_next_extent(trans, rc, path);
+		ret = find_next_extent(trans, rc, path, &key);
 		if (ret < 0)
 			err = ret;
 		if (ret != 0)
@@ -3313,9 +3934,7 @@ static noinline_for_stack int relocate_block_group(struct reloc_control *rc)
 
 		ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
 				    struct btrfs_extent_item);
-		btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
-		item_size = btrfs_item_size_nr(path->nodes[0],
-					       path->slots[0]);
+		item_size = btrfs_item_size_nr(path->nodes[0], path->slots[0]);
 		if (item_size >= sizeof(*ei)) {
 			flags = btrfs_extent_flags(path->nodes[0], ei);
 			ret = check_extent_flags(flags);
@@ -3336,7 +3955,7 @@ static noinline_for_stack int relocate_block_group(struct reloc_control *rc)
 				flags = BTRFS_EXTENT_FLAG_DATA;
 
 			if (path_change) {
-				btrfs_release_path(rc->extent_root, path);
+				btrfs_release_path(path);
 
 				path->search_commit_root = 1;
 				path->skip_locking = 1;
@@ -3356,73 +3975,99 @@ static noinline_for_stack int relocate_block_group(struct reloc_control *rc)
 		if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
 			ret = add_tree_block(rc, &key, path, &blocks);
 		} else if (rc->stage == UPDATE_DATA_PTRS &&
-			 (flags & BTRFS_EXTENT_FLAG_DATA)) {
+			   (flags & BTRFS_EXTENT_FLAG_DATA)) {
 			ret = add_data_references(rc, &key, path, &blocks);
 		} else {
-			btrfs_release_path(rc->extent_root, path);
+			btrfs_release_path(path);
 			ret = 0;
 		}
 		if (ret < 0) {
-			err = 0;
+			err = ret;
 			break;
 		}
 
 		if (!RB_EMPTY_ROOT(&blocks)) {
 			ret = relocate_tree_blocks(trans, rc, &blocks);
 			if (ret < 0) {
-				err = ret;
-				break;
+				/*
+				 * if we fail to relocate tree blocks, force to update
+				 * backref cache when committing transaction.
+				 */
+				rc->backref_cache.last_trans = trans->transid - 1;
+
+				if (ret != -EAGAIN) {
+					err = ret;
+					break;
+				}
+				rc->extents_found--;
+				rc->search_start = key.objectid;
 			}
 		}
 
-		nr = trans->blocks_used;
-		btrfs_end_transaction(trans, rc->extent_root);
+		btrfs_end_transaction_throttle(trans, rc->extent_root);
+		btrfs_btree_balance_dirty(rc->extent_root);
 		trans = NULL;
-		btrfs_btree_balance_dirty(rc->extent_root, nr);
 
 		if (rc->stage == MOVE_DATA_EXTENTS &&
 		    (flags & BTRFS_EXTENT_FLAG_DATA)) {
 			rc->found_file_extent = 1;
 			ret = relocate_data_extent(rc->data_inode,
-						   &key, cluster);
+						   &key, &rc->cluster);
 			if (ret < 0) {
 				err = ret;
 				break;
 			}
 		}
 	}
-	btrfs_free_path(path);
+	if (trans && progress && err == -ENOSPC) {
+		ret = btrfs_force_chunk_alloc(trans, rc->extent_root,
+					      rc->block_group->flags);
+		if (ret == 0) {
+			err = 0;
+			progress = 0;
+			goto restart;
+		}
+	}
+
+	btrfs_release_path(path);
+	clear_extent_bits(&rc->processed_blocks, 0, (u64)-1, EXTENT_DIRTY,
+			  GFP_NOFS);
 
 	if (trans) {
-		nr = trans->blocks_used;
-		btrfs_end_transaction(trans, rc->extent_root);
-		btrfs_btree_balance_dirty(rc->extent_root, nr);
+		btrfs_end_transaction_throttle(trans, rc->extent_root);
+		btrfs_btree_balance_dirty(rc->extent_root);
 	}
 
 	if (!err) {
-		ret = relocate_file_extent_cluster(rc->data_inode, cluster);
+		ret = relocate_file_extent_cluster(rc->data_inode,
+						   &rc->cluster);
 		if (ret < 0)
 			err = ret;
 	}
 
-	kfree(cluster);
+	rc->create_reloc_tree = 0;
+	set_reloc_control(rc);
 
-	rc->create_reloc_root = 0;
-	smp_mb();
+	backref_cache_cleanup(&rc->backref_cache);
+	btrfs_block_rsv_release(rc->extent_root, rc->block_rsv, (u64)-1);
 
-	if (rc->extents_found > 0) {
-		trans = btrfs_start_transaction(rc->extent_root, 1);
-		btrfs_commit_transaction(trans, rc->extent_root);
-	}
+	err = prepare_to_merge(rc, err);
 
 	merge_reloc_roots(rc);
 
+	rc->merge_reloc_tree = 0;
 	unset_reloc_control(rc);
+	btrfs_block_rsv_release(rc->extent_root, rc->block_rsv, (u64)-1);
 
 	/* get rid of pinned extents */
-	trans = btrfs_start_transaction(rc->extent_root, 1);
-	btrfs_commit_transaction(trans, rc->extent_root);
-
+	trans = btrfs_join_transaction(rc->extent_root);
+	if (IS_ERR(trans))
+		err = PTR_ERR(trans);
+	else
+		btrfs_commit_transaction(trans, rc->extent_root);
+out_free:
+	btrfs_free_block_rsv(rc->extent_root, rc->block_rsv);
+	btrfs_free_path(path);
 	return err;
 }
 
@@ -3448,9 +4093,10 @@ static int __insert_orphan_inode(struct btrfs_trans_handle *trans,
 	btrfs_set_inode_generation(leaf, item, 1);
 	btrfs_set_inode_size(leaf, item, 0);
 	btrfs_set_inode_mode(leaf, item, S_IFREG | 0600);
-	btrfs_set_inode_flags(leaf, item, BTRFS_INODE_NOCOMPRESS);
+	btrfs_set_inode_flags(leaf, item, BTRFS_INODE_NOCOMPRESS |
+					  BTRFS_INODE_PREALLOC);
 	btrfs_mark_buffer_dirty(leaf);
-	btrfs_release_path(root, path);
+	btrfs_release_path(path);
 out:
 	btrfs_free_path(path);
 	return ret;
@@ -3460,14 +4106,14 @@ out:
  * helper to create inode for data relocation.
  * the inode is in data relocation tree and its link count is 0
  */
-static struct inode *create_reloc_inode(struct btrfs_fs_info *fs_info,
-					struct btrfs_block_group_cache *group)
+static noinline_for_stack
+struct inode *create_reloc_inode(struct btrfs_fs_info *fs_info,
+				 struct btrfs_block_group_cache *group)
 {
 	struct inode *inode = NULL;
 	struct btrfs_trans_handle *trans;
 	struct btrfs_root *root;
 	struct btrfs_key key;
-	unsigned long nr;
 	u64 objectid = BTRFS_FIRST_FREE_OBJECTID;
 	int err = 0;
 
@@ -3475,10 +4121,11 @@ static struct inode *create_reloc_inode(struct btrfs_fs_info *fs_info,
 	if (IS_ERR(root))
 		return ERR_CAST(root);
 
-	trans = btrfs_start_transaction(root, 1);
-	BUG_ON(!trans);
+	trans = btrfs_start_transaction(root, 6);
+	if (IS_ERR(trans))
+		return ERR_CAST(trans);
 
-	err = btrfs_find_free_objectid(trans, root, objectid, &objectid);
+	err = btrfs_find_free_objectid(root, &objectid);
 	if (err)
 		goto out;
 
@@ -3494,10 +4141,8 @@ static struct inode *create_reloc_inode(struct btrfs_fs_info *fs_info,
 
 	err = btrfs_orphan_add(trans, inode);
 out:
-	nr = trans->blocks_used;
 	btrfs_end_transaction(trans, root);
-
-	btrfs_btree_balance_dirty(root, nr);
+	btrfs_btree_balance_dirty(root);
 	if (err) {
 		if (inode)
 			iput(inode);
@@ -3506,6 +4151,22 @@ out:
 	return inode;
 }
 
+static struct reloc_control *alloc_reloc_control(struct btrfs_fs_info *fs_info)
+{
+	struct reloc_control *rc;
+
+	rc = kzalloc(sizeof(*rc), GFP_NOFS);
+	if (!rc)
+		return NULL;
+
+	INIT_LIST_HEAD(&rc->reloc_roots);
+	backref_cache_init(&rc->backref_cache);
+	mapping_tree_init(&rc->reloc_root_tree);
+	extent_io_tree_init(&rc->processed_blocks,
+			    fs_info->btree_inode->i_mapping);
+	return rc;
+}
+
 /*
  * function to relocate all extents in a block group.
  */
@@ -3513,25 +4174,49 @@ int btrfs_relocate_block_group(struct btrfs_root *extent_root, u64 group_start)
 {
 	struct btrfs_fs_info *fs_info = extent_root->fs_info;
 	struct reloc_control *rc;
+	struct inode *inode;
+	struct btrfs_path *path;
 	int ret;
+	int rw = 0;
 	int err = 0;
 
-	rc = kzalloc(sizeof(*rc), GFP_NOFS);
+	rc = alloc_reloc_control(fs_info);
 	if (!rc)
 		return -ENOMEM;
 
-	mapping_tree_init(&rc->reloc_root_tree);
-	extent_io_tree_init(&rc->processed_blocks, NULL, GFP_NOFS);
-	INIT_LIST_HEAD(&rc->reloc_roots);
+	rc->extent_root = extent_root;
 
 	rc->block_group = btrfs_lookup_block_group(fs_info, group_start);
 	BUG_ON(!rc->block_group);
 
-	btrfs_init_workers(&rc->workers, "relocate",
-			   fs_info->thread_pool_size, NULL);
+	if (!rc->block_group->ro) {
+		ret = btrfs_set_block_group_ro(extent_root, rc->block_group);
+		if (ret) {
+			err = ret;
+			goto out;
+		}
+		rw = 1;
+	}
 
-	rc->extent_root = extent_root;
-	btrfs_prepare_block_group_relocation(extent_root, rc->block_group);
+	path = btrfs_alloc_path();
+	if (!path) {
+		err = -ENOMEM;
+		goto out;
+	}
+
+	inode = lookup_free_space_inode(fs_info->tree_root, rc->block_group,
+					path);
+	btrfs_free_path(path);
+
+	if (!IS_ERR(inode))
+		ret = delete_block_group_cache(fs_info, inode, 0);
+	else
+		ret = PTR_ERR(inode);
+
+	if (ret && ret != -ENOENT) {
+		err = ret;
+		goto out;
+	}
 
 	rc->data_inode = create_reloc_inode(fs_info, rc->block_group);
 	if (IS_ERR(rc->data_inode)) {
@@ -3540,65 +4225,51 @@ int btrfs_relocate_block_group(struct btrfs_root *extent_root, u64 group_start)
 		goto out;
 	}
 
-	printk(KERN_INFO "btrfs: relocating block group %llu flags %llu\n",
-	       (unsigned long long)rc->block_group->key.objectid,
-	       (unsigned long long)rc->block_group->flags);
+	btrfs_info(extent_root->fs_info, "relocating block group %llu flags %llu",
+	       rc->block_group->key.objectid, rc->block_group->flags);
 
-	btrfs_start_delalloc_inodes(fs_info->tree_root, 0);
-	btrfs_wait_ordered_extents(fs_info->tree_root, 0, 0);
+	ret = btrfs_start_delalloc_roots(fs_info, 0, -1);
+	if (ret < 0) {
+		err = ret;
+		goto out;
+	}
+	btrfs_wait_ordered_roots(fs_info, -1);
 
 	while (1) {
-		rc->extents_found = 0;
-		rc->extents_skipped = 0;
-
 		mutex_lock(&fs_info->cleaner_mutex);
-
-		btrfs_clean_old_snapshots(fs_info->tree_root);
 		ret = relocate_block_group(rc);
-
 		mutex_unlock(&fs_info->cleaner_mutex);
 		if (ret < 0) {
 			err = ret;
-			break;
+			goto out;
 		}
 
 		if (rc->extents_found == 0)
 			break;
 
-		printk(KERN_INFO "btrfs: found %llu extents\n",
-			(unsigned long long)rc->extents_found);
+		btrfs_info(extent_root->fs_info, "found %llu extents",
+			rc->extents_found);
 
 		if (rc->stage == MOVE_DATA_EXTENTS && rc->found_file_extent) {
-			btrfs_wait_ordered_range(rc->data_inode, 0, (u64)-1);
+			ret = btrfs_wait_ordered_range(rc->data_inode, 0,
+						       (u64)-1);
+			if (ret) {
+				err = ret;
+				goto out;
+			}
 			invalidate_mapping_pages(rc->data_inode->i_mapping,
 						 0, -1);
 			rc->stage = UPDATE_DATA_PTRS;
-		} else if (rc->stage == UPDATE_DATA_PTRS &&
-			   rc->extents_skipped >= rc->extents_found) {
-			iput(rc->data_inode);
-			rc->data_inode = create_reloc_inode(fs_info,
-							    rc->block_group);
-			if (IS_ERR(rc->data_inode)) {
-				err = PTR_ERR(rc->data_inode);
-				rc->data_inode = NULL;
-				break;
-			}
-			rc->stage = MOVE_DATA_EXTENTS;
-			rc->found_file_extent = 0;
 		}
 	}
 
-	filemap_write_and_wait_range(fs_info->btree_inode->i_mapping,
-				     rc->block_group->key.objectid,
-				     rc->block_group->key.objectid +
-				     rc->block_group->key.offset - 1);
-
 	WARN_ON(rc->block_group->pinned > 0);
 	WARN_ON(rc->block_group->reserved > 0);
 	WARN_ON(btrfs_block_group_used(&rc->block_group->item) > 0);
 out:
+	if (err && rw)
+		btrfs_set_block_group_rw(extent_root, rc->block_group);
 	iput(rc->data_inode);
-	btrfs_stop_workers(&rc->workers);
 	btrfs_put_block_group(rc->block_group);
 	kfree(rc);
 	return err;
@@ -3607,9 +4278,11 @@ out:
 static noinline_for_stack int mark_garbage_root(struct btrfs_root *root)
 {
 	struct btrfs_trans_handle *trans;
-	int ret;
+	int ret, err;
 
-	trans = btrfs_start_transaction(root->fs_info->tree_root, 1);
+	trans = btrfs_start_transaction(root->fs_info->tree_root, 0);
+	if (IS_ERR(trans))
+		return PTR_ERR(trans);
 
 	memset(&root->root_item.drop_progress, 0,
 		sizeof(root->root_item.drop_progress));
@@ -3617,11 +4290,11 @@ static noinline_for_stack int mark_garbage_root(struct btrfs_root *root)
 	btrfs_set_root_refs(&root->root_item, 0);
 	ret = btrfs_update_root(trans, root->fs_info->tree_root,
 				&root->root_key, &root->root_item);
-	BUG_ON(ret);
 
-	ret = btrfs_end_transaction(trans, root->fs_info->tree_root);
-	BUG_ON(ret);
-	return 0;
+	err = btrfs_end_transaction(trans, root->fs_info->tree_root);
+	if (err)
+		return err;
+	return ret;
 }
 
 /*
@@ -3646,6 +4319,7 @@ int btrfs_recover_relocation(struct btrfs_root *root)
 	path = btrfs_alloc_path();
 	if (!path)
 		return -ENOMEM;
+	path->reada = -1;
 
 	key.objectid = BTRFS_TREE_RELOC_OBJECTID;
 	key.type = BTRFS_ROOT_ITEM_KEY;
@@ -3665,13 +4339,13 @@ int btrfs_recover_relocation(struct btrfs_root *root)
 		}
 		leaf = path->nodes[0];
 		btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
-		btrfs_release_path(root->fs_info->tree_root, path);
+		btrfs_release_path(path);
 
 		if (key.objectid != BTRFS_TREE_RELOC_OBJECTID ||
 		    key.type != BTRFS_ROOT_ITEM_KEY)
 			break;
 
-		reloc_root = btrfs_read_fs_root_no_radix(root, &key);
+		reloc_root = btrfs_read_fs_root(root, &key);
 		if (IS_ERR(reloc_root)) {
 			err = PTR_ERR(reloc_root);
 			goto out;
@@ -3688,7 +4362,11 @@ int btrfs_recover_relocation(struct btrfs_root *root)
 					err = ret;
 					goto out;
 				}
-				mark_garbage_root(reloc_root);
+				ret = mark_garbage_root(reloc_root);
+				if (ret < 0) {
+					err = ret;
+					goto out;
+				}
 			}
 		}
 
@@ -3697,25 +4375,30 @@ int btrfs_recover_relocation(struct btrfs_root *root)
 
 		key.offset--;
 	}
-	btrfs_release_path(root->fs_info->tree_root, path);
+	btrfs_release_path(path);
 
 	if (list_empty(&reloc_roots))
 		goto out;
 
-	rc = kzalloc(sizeof(*rc), GFP_NOFS);
+	rc = alloc_reloc_control(root->fs_info);
 	if (!rc) {
 		err = -ENOMEM;
 		goto out;
 	}
 
-	mapping_tree_init(&rc->reloc_root_tree);
-	INIT_LIST_HEAD(&rc->reloc_roots);
-	btrfs_init_workers(&rc->workers, "relocate",
-			   root->fs_info->thread_pool_size, NULL);
 	rc->extent_root = root->fs_info->extent_root;
 
 	set_reloc_control(rc);
 
+	trans = btrfs_join_transaction(rc->extent_root);
+	if (IS_ERR(trans)) {
+		unset_reloc_control(rc);
+		err = PTR_ERR(trans);
+		goto out_free;
+	}
+
+	rc->merge_reloc_tree = 1;
+
 	while (!list_empty(&reloc_roots)) {
 		reloc_root = list_entry(reloc_roots.next,
 					struct btrfs_root, root_list);
@@ -3729,34 +4412,35 @@ int btrfs_recover_relocation(struct btrfs_root *root)
 
 		fs_root = read_fs_root(root->fs_info,
 				       reloc_root->root_key.offset);
-		BUG_ON(IS_ERR(fs_root));
+		if (IS_ERR(fs_root)) {
+			err = PTR_ERR(fs_root);
+			goto out_free;
+		}
 
-		__add_reloc_root(reloc_root);
+		err = __add_reloc_root(reloc_root);
+		BUG_ON(err < 0); /* -ENOMEM or logic error */
 		fs_root->reloc_root = reloc_root;
 	}
 
-	trans = btrfs_start_transaction(rc->extent_root, 1);
-	btrfs_commit_transaction(trans, rc->extent_root);
+	err = btrfs_commit_transaction(trans, rc->extent_root);
+	if (err)
+		goto out_free;
 
 	merge_reloc_roots(rc);
 
 	unset_reloc_control(rc);
 
-	trans = btrfs_start_transaction(rc->extent_root, 1);
-	btrfs_commit_transaction(trans, rc->extent_root);
+	trans = btrfs_join_transaction(rc->extent_root);
+	if (IS_ERR(trans))
+		err = PTR_ERR(trans);
+	else
+		err = btrfs_commit_transaction(trans, rc->extent_root);
+out_free:
+	kfree(rc);
 out:
-	if (rc) {
-		btrfs_stop_workers(&rc->workers);
-		kfree(rc);
-	}
-	while (!list_empty(&reloc_roots)) {
-		reloc_root = list_entry(reloc_roots.next,
-					struct btrfs_root, root_list);
-		list_del(&reloc_root->root_list);
-		free_extent_buffer(reloc_root->node);
-		free_extent_buffer(reloc_root->commit_root);
-		kfree(reloc_root);
-	}
+	if (!list_empty(&reloc_roots))
+		free_reloc_roots(&reloc_roots);
+
 	btrfs_free_path(path);
 
 	if (err == 0) {
@@ -3766,7 +4450,7 @@ out:
 		if (IS_ERR(fs_root))
 			err = PTR_ERR(fs_root);
 		else
-			btrfs_orphan_cleanup(fs_root);
+			err = btrfs_orphan_cleanup(fs_root);
 	}
 	return err;
 }
@@ -3780,12 +4464,11 @@ out:
 int btrfs_reloc_clone_csums(struct inode *inode, u64 file_pos, u64 len)
 {
 	struct btrfs_ordered_sum *sums;
-	struct btrfs_sector_sum *sector_sum;
 	struct btrfs_ordered_extent *ordered;
 	struct btrfs_root *root = BTRFS_I(inode)->root;
-	size_t offset;
 	int ret;
 	u64 disk_bytenr;
+	u64 new_bytenr;
 	LIST_HEAD(list);
 
 	ordered = btrfs_lookup_ordered_extent(inode, file_pos);
@@ -3793,24 +4476,166 @@ int btrfs_reloc_clone_csums(struct inode *inode, u64 file_pos, u64 len)
 
 	disk_bytenr = file_pos + BTRFS_I(inode)->index_cnt;
 	ret = btrfs_lookup_csums_range(root->fs_info->csum_root, disk_bytenr,
-				       disk_bytenr + len - 1, &list);
+				       disk_bytenr + len - 1, &list, 0);
+	if (ret)
+		goto out;
 
 	while (!list_empty(&list)) {
 		sums = list_entry(list.next, struct btrfs_ordered_sum, list);
 		list_del_init(&sums->list);
 
-		sector_sum = sums->sums;
-		sums->bytenr = ordered->start;
-
-		offset = 0;
-		while (offset < sums->len) {
-			sector_sum->bytenr += ordered->start - disk_bytenr;
-			sector_sum++;
-			offset += root->sectorsize;
-		}
+		/*
+		 * We need to offset the new_bytenr based on where the csum is.
+		 * We need to do this because we will read in entire prealloc
+		 * extents but we may have written to say the middle of the
+		 * prealloc extent, so we need to make sure the csum goes with
+		 * the right disk offset.
+		 *
+		 * We can do this because the data reloc inode refers strictly
+		 * to the on disk bytes, so we don't have to worry about
+		 * disk_len vs real len like with real inodes since it's all
+		 * disk length.
+		 */
+		new_bytenr = ordered->start + (sums->bytenr - disk_bytenr);
+		sums->bytenr = new_bytenr;
 
 		btrfs_add_ordered_sum(inode, ordered, sums);
 	}
+out:
 	btrfs_put_ordered_extent(ordered);
-	return 0;
+	return ret;
+}
+
+int btrfs_reloc_cow_block(struct btrfs_trans_handle *trans,
+			  struct btrfs_root *root, struct extent_buffer *buf,
+			  struct extent_buffer *cow)
+{
+	struct reloc_control *rc;
+	struct backref_node *node;
+	int first_cow = 0;
+	int level;
+	int ret = 0;
+
+	rc = root->fs_info->reloc_ctl;
+	if (!rc)
+		return 0;
+
+	BUG_ON(rc->stage == UPDATE_DATA_PTRS &&
+	       root->root_key.objectid == BTRFS_DATA_RELOC_TREE_OBJECTID);
+
+	if (root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID) {
+		if (buf == root->node)
+			__update_reloc_root(root, cow->start);
+	}
+
+	level = btrfs_header_level(buf);
+	if (btrfs_header_generation(buf) <=
+	    btrfs_root_last_snapshot(&root->root_item))
+		first_cow = 1;
+
+	if (root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID &&
+	    rc->create_reloc_tree) {
+		WARN_ON(!first_cow && level == 0);
+
+		node = rc->backref_cache.path[level];
+		BUG_ON(node->bytenr != buf->start &&
+		       node->new_bytenr != buf->start);
+
+		drop_node_buffer(node);
+		extent_buffer_get(cow);
+		node->eb = cow;
+		node->new_bytenr = cow->start;
+
+		if (!node->pending) {
+			list_move_tail(&node->list,
+				       &rc->backref_cache.pending[level]);
+			node->pending = 1;
+		}
+
+		if (first_cow)
+			__mark_block_processed(rc, node);
+
+		if (first_cow && level > 0)
+			rc->nodes_relocated += buf->len;
+	}
+
+	if (level == 0 && first_cow && rc->stage == UPDATE_DATA_PTRS)
+		ret = replace_file_extents(trans, rc, root, cow);
+	return ret;
+}
+
+/*
+ * called before creating snapshot. it calculates metadata reservation
+ * requried for relocating tree blocks in the snapshot
+ */
+void btrfs_reloc_pre_snapshot(struct btrfs_trans_handle *trans,
+			      struct btrfs_pending_snapshot *pending,
+			      u64 *bytes_to_reserve)
+{
+	struct btrfs_root *root;
+	struct reloc_control *rc;
+
+	root = pending->root;
+	if (!root->reloc_root)
+		return;
+
+	rc = root->fs_info->reloc_ctl;
+	if (!rc->merge_reloc_tree)
+		return;
+
+	root = root->reloc_root;
+	BUG_ON(btrfs_root_refs(&root->root_item) == 0);
+	/*
+	 * relocation is in the stage of merging trees. the space
+	 * used by merging a reloc tree is twice the size of
+	 * relocated tree nodes in the worst case. half for cowing
+	 * the reloc tree, half for cowing the fs tree. the space
+	 * used by cowing the reloc tree will be freed after the
+	 * tree is dropped. if we create snapshot, cowing the fs
+	 * tree may use more space than it frees. so we need
+	 * reserve extra space.
+	 */
+	*bytes_to_reserve += rc->nodes_relocated;
+}
+
+/*
+ * called after snapshot is created. migrate block reservation
+ * and create reloc root for the newly created snapshot
+ */
+int btrfs_reloc_post_snapshot(struct btrfs_trans_handle *trans,
+			       struct btrfs_pending_snapshot *pending)
+{
+	struct btrfs_root *root = pending->root;
+	struct btrfs_root *reloc_root;
+	struct btrfs_root *new_root;
+	struct reloc_control *rc;
+	int ret;
+
+	if (!root->reloc_root)
+		return 0;
+
+	rc = root->fs_info->reloc_ctl;
+	rc->merging_rsv_size += rc->nodes_relocated;
+
+	if (rc->merge_reloc_tree) {
+		ret = btrfs_block_rsv_migrate(&pending->block_rsv,
+					      rc->block_rsv,
+					      rc->nodes_relocated);
+		if (ret)
+			return ret;
+	}
+
+	new_root = pending->snap;
+	reloc_root = create_reloc_root(trans, root->reloc_root,
+				       new_root->root_key.objectid);
+	if (IS_ERR(reloc_root))
+		return PTR_ERR(reloc_root);
+
+	ret = __add_reloc_root(reloc_root);
+	BUG_ON(ret < 0);
+	new_root->reloc_root = reloc_root;
+
+	if (rc->create_reloc_tree)
+		ret = clone_backref_node(trans, rc, root, reloc_root);
+	return ret;
 }
diff --git a/fs/btrfs/root-tree.c b/fs/btrfs/root-tree.c
index 67fa2d29d66..360a728a639 100644
--- a/fs/btrfs/root-tree.c
+++ b/fs/btrfs/root-tree.c
@@ -16,114 +16,117 @@
  * Boston, MA 021110-1307, USA.
  */
 
+#include <linux/err.h>
+#include <linux/uuid.h>
 #include "ctree.h"
 #include "transaction.h"
 #include "disk-io.h"
 #include "print-tree.h"
 
 /*
- *  search forward for a root, starting with objectid 'search_start'
- *  if a root key is found, the objectid we find is filled into 'found_objectid'
- *  and 0 is returned.  < 0 is returned on error, 1 if there is nothing
- *  left in the tree.
+ * Read a root item from the tree. In case we detect a root item smaller then
+ * sizeof(root_item), we know it's an old version of the root structure and
+ * initialize all new fields to zero. The same happens if we detect mismatching
+ * generation numbers as then we know the root was once mounted with an older
+ * kernel that was not aware of the root item structure change.
  */
-int btrfs_search_root(struct btrfs_root *root, u64 search_start,
-		      u64 *found_objectid)
+static void btrfs_read_root_item(struct extent_buffer *eb, int slot,
+				struct btrfs_root_item *item)
 {
-	struct btrfs_path *path;
-	struct btrfs_key search_key;
-	int ret;
-
-	root = root->fs_info->tree_root;
-	search_key.objectid = search_start;
-	search_key.type = (u8)-1;
-	search_key.offset = (u64)-1;
-
-	path = btrfs_alloc_path();
-	BUG_ON(!path);
-again:
-	ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0);
-	if (ret < 0)
-		goto out;
-	if (ret == 0) {
-		ret = 1;
-		goto out;
-	}
-	if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
-		ret = btrfs_next_leaf(root, path);
-		if (ret)
-			goto out;
-	}
-	btrfs_item_key_to_cpu(path->nodes[0], &search_key, path->slots[0]);
-	if (search_key.type != BTRFS_ROOT_ITEM_KEY) {
-		search_key.offset++;
-		btrfs_release_path(root, path);
-		goto again;
+	uuid_le uuid;
+	int len;
+	int need_reset = 0;
+
+	len = btrfs_item_size_nr(eb, slot);
+	read_extent_buffer(eb, item, btrfs_item_ptr_offset(eb, slot),
+			min_t(int, len, (int)sizeof(*item)));
+	if (len < sizeof(*item))
+		need_reset = 1;
+	if (!need_reset && btrfs_root_generation(item)
+		!= btrfs_root_generation_v2(item)) {
+		if (btrfs_root_generation_v2(item) != 0) {
+			printk(KERN_WARNING "BTRFS: mismatching "
+					"generation and generation_v2 "
+					"found in root item. This root "
+					"was probably mounted with an "
+					"older kernel. Resetting all "
+					"new fields.\n");
+		}
+		need_reset = 1;
 	}
-	ret = 0;
-	*found_objectid = search_key.objectid;
+	if (need_reset) {
+		memset(&item->generation_v2, 0,
+			sizeof(*item) - offsetof(struct btrfs_root_item,
+					generation_v2));
 
-out:
-	btrfs_free_path(path);
-	return ret;
+		uuid_le_gen(&uuid);
+		memcpy(item->uuid, uuid.b, BTRFS_UUID_SIZE);
+	}
 }
 
 /*
- * lookup the root with the highest offset for a given objectid.  The key we do
- * find is copied into 'key'.  If we find something return 0, otherwise 1, < 0
- * on error.
+ * btrfs_find_root - lookup the root by the key.
+ * root: the root of the root tree
+ * search_key: the key to search
+ * path: the path we search
+ * root_item: the root item of the tree we look for
+ * root_key: the reak key of the tree we look for
+ *
+ * If ->offset of 'seach_key' is -1ULL, it means we are not sure the offset
+ * of the search key, just lookup the root with the highest offset for a
+ * given objectid.
+ *
+ * If we find something return 0, otherwise > 0, < 0 on error.
  */
-int btrfs_find_last_root(struct btrfs_root *root, u64 objectid,
-			struct btrfs_root_item *item, struct btrfs_key *key)
+int btrfs_find_root(struct btrfs_root *root, struct btrfs_key *search_key,
+		    struct btrfs_path *path, struct btrfs_root_item *root_item,
+		    struct btrfs_key *root_key)
 {
-	struct btrfs_path *path;
-	struct btrfs_key search_key;
 	struct btrfs_key found_key;
 	struct extent_buffer *l;
 	int ret;
 	int slot;
 
-	search_key.objectid = objectid;
-	search_key.type = BTRFS_ROOT_ITEM_KEY;
-	search_key.offset = (u64)-1;
-
-	path = btrfs_alloc_path();
-	BUG_ON(!path);
-	ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0);
+	ret = btrfs_search_slot(NULL, root, search_key, path, 0, 0);
 	if (ret < 0)
-		goto out;
+		return ret;
 
-	BUG_ON(ret == 0);
-	if (path->slots[0] == 0) {
-		ret = 1;
-		goto out;
+	if (search_key->offset != -1ULL) {	/* the search key is exact */
+		if (ret > 0)
+			goto out;
+	} else {
+		BUG_ON(ret == 0);		/* Logical error */
+		if (path->slots[0] == 0)
+			goto out;
+		path->slots[0]--;
+		ret = 0;
 	}
+
 	l = path->nodes[0];
-	slot = path->slots[0] - 1;
+	slot = path->slots[0];
+
 	btrfs_item_key_to_cpu(l, &found_key, slot);
-	if (found_key.objectid != objectid ||
+	if (found_key.objectid != search_key->objectid ||
 	    found_key.type != BTRFS_ROOT_ITEM_KEY) {
 		ret = 1;
 		goto out;
 	}
-	if (item)
-		read_extent_buffer(l, item, btrfs_item_ptr_offset(l, slot),
-				   sizeof(*item));
-	if (key)
-		memcpy(key, &found_key, sizeof(found_key));
-	ret = 0;
+
+	if (root_item)
+		btrfs_read_root_item(l, slot, root_item);
+	if (root_key)
+		memcpy(root_key, &found_key, sizeof(found_key));
 out:
-	btrfs_free_path(path);
+	btrfs_release_path(path);
 	return ret;
 }
 
-int btrfs_set_root_node(struct btrfs_root_item *item,
-			struct extent_buffer *node)
+void btrfs_set_root_node(struct btrfs_root_item *item,
+			 struct extent_buffer *node)
 {
 	btrfs_set_root_bytenr(item, node->start);
 	btrfs_set_root_level(item, btrfs_header_level(node));
 	btrfs_set_root_generation(item, btrfs_header_generation(node));
-	return 0;
 }
 
 /*
@@ -138,24 +141,67 @@ int btrfs_update_root(struct btrfs_trans_handle *trans, struct btrfs_root
 	int ret;
 	int slot;
 	unsigned long ptr;
+	int old_len;
 
 	path = btrfs_alloc_path();
-	BUG_ON(!path);
+	if (!path)
+		return -ENOMEM;
+
 	ret = btrfs_search_slot(trans, root, key, path, 0, 1);
-	if (ret < 0)
+	if (ret < 0) {
+		btrfs_abort_transaction(trans, root, ret);
 		goto out;
+	}
 
 	if (ret != 0) {
 		btrfs_print_leaf(root, path->nodes[0]);
-		printk(KERN_CRIT "unable to update root key %llu %u %llu\n",
-		       (unsigned long long)key->objectid, key->type,
-		       (unsigned long long)key->offset);
+		btrfs_crit(root->fs_info, "unable to update root key %llu %u %llu",
+		       key->objectid, key->type, key->offset);
 		BUG_ON(1);
 	}
 
 	l = path->nodes[0];
 	slot = path->slots[0];
 	ptr = btrfs_item_ptr_offset(l, slot);
+	old_len = btrfs_item_size_nr(l, slot);
+
+	/*
+	 * If this is the first time we update the root item which originated
+	 * from an older kernel, we need to enlarge the item size to make room
+	 * for the added fields.
+	 */
+	if (old_len < sizeof(*item)) {
+		btrfs_release_path(path);
+		ret = btrfs_search_slot(trans, root, key, path,
+				-1, 1);
+		if (ret < 0) {
+			btrfs_abort_transaction(trans, root, ret);
+			goto out;
+		}
+
+		ret = btrfs_del_item(trans, root, path);
+		if (ret < 0) {
+			btrfs_abort_transaction(trans, root, ret);
+			goto out;
+		}
+		btrfs_release_path(path);
+		ret = btrfs_insert_empty_item(trans, root, path,
+				key, sizeof(*item));
+		if (ret < 0) {
+			btrfs_abort_transaction(trans, root, ret);
+			goto out;
+		}
+		l = path->nodes[0];
+		slot = path->slots[0];
+		ptr = btrfs_item_ptr_offset(l, slot);
+	}
+
+	/*
+	 * Update generation_v2 so at the next mount we know the new root
+	 * fields are valid.
+	 */
+	btrfs_set_root_generation_v2(item, btrfs_root_generation(item));
+
 	write_extent_buffer(l, item, ptr, sizeof(*item));
 	btrfs_mark_buffer_dirty(path->nodes[0]);
 out:
@@ -163,95 +209,14 @@ out:
 	return ret;
 }
 
-int btrfs_insert_root(struct btrfs_trans_handle *trans, struct btrfs_root
-		      *root, struct btrfs_key *key, struct btrfs_root_item
-		      *item)
+int btrfs_insert_root(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+		      struct btrfs_key *key, struct btrfs_root_item *item)
 {
-	int ret;
-	ret = btrfs_insert_item(trans, root, key, item, sizeof(*item));
-	return ret;
-}
-
-/*
- * at mount time we want to find all the old transaction snapshots that were in
- * the process of being deleted if we crashed.  This is any root item with an
- * offset lower than the latest root.  They need to be queued for deletion to
- * finish what was happening when we crashed.
- */
-int btrfs_find_dead_roots(struct btrfs_root *root, u64 objectid)
-{
-	struct btrfs_root *dead_root;
-	struct btrfs_item *item;
-	struct btrfs_root_item *ri;
-	struct btrfs_key key;
-	struct btrfs_key found_key;
-	struct btrfs_path *path;
-	int ret;
-	u32 nritems;
-	struct extent_buffer *leaf;
-	int slot;
-
-	key.objectid = objectid;
-	btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
-	key.offset = 0;
-	path = btrfs_alloc_path();
-	if (!path)
-		return -ENOMEM;
-
-again:
-	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
-	if (ret < 0)
-		goto err;
-	while (1) {
-		leaf = path->nodes[0];
-		nritems = btrfs_header_nritems(leaf);
-		slot = path->slots[0];
-		if (slot >= nritems) {
-			ret = btrfs_next_leaf(root, path);
-			if (ret)
-				break;
-			leaf = path->nodes[0];
-			nritems = btrfs_header_nritems(leaf);
-			slot = path->slots[0];
-		}
-		item = btrfs_item_nr(leaf, slot);
-		btrfs_item_key_to_cpu(leaf, &key, slot);
-		if (btrfs_key_type(&key) != BTRFS_ROOT_ITEM_KEY)
-			goto next;
-
-		if (key.objectid < objectid)
-			goto next;
-
-		if (key.objectid > objectid)
-			break;
-
-		ri = btrfs_item_ptr(leaf, slot, struct btrfs_root_item);
-		if (btrfs_disk_root_refs(leaf, ri) != 0)
-			goto next;
-
-		memcpy(&found_key, &key, sizeof(key));
-		key.offset++;
-		btrfs_release_path(root, path);
-		dead_root =
-			btrfs_read_fs_root_no_radix(root->fs_info->tree_root,
-						    &found_key);
-		if (IS_ERR(dead_root)) {
-			ret = PTR_ERR(dead_root);
-			goto err;
-		}
-
-		ret = btrfs_add_dead_root(dead_root);
-		if (ret)
-			goto err;
-		goto again;
-next:
-		slot++;
-		path->slots[0]++;
-	}
-	ret = 0;
-err:
-	btrfs_free_path(path);
-	return ret;
+	/*
+	 * Make sure generation v1 and v2 match. See update_root for details.
+	 */
+	btrfs_set_root_generation_v2(item, btrfs_root_generation(item));
+	return btrfs_insert_item(trans, root, key, item, sizeof(*item));
 }
 
 int btrfs_find_orphan_roots(struct btrfs_root *tree_root)
@@ -259,8 +224,14 @@ int btrfs_find_orphan_roots(struct btrfs_root *tree_root)
 	struct extent_buffer *leaf;
 	struct btrfs_path *path;
 	struct btrfs_key key;
+	struct btrfs_key root_key;
+	struct btrfs_root *root;
 	int err = 0;
 	int ret;
+	bool can_recover = true;
+
+	if (tree_root->fs_info->sb->s_flags & MS_RDONLY)
+		can_recover = false;
 
 	path = btrfs_alloc_path();
 	if (!path)
@@ -270,6 +241,9 @@ int btrfs_find_orphan_roots(struct btrfs_root *tree_root)
 	key.type = BTRFS_ORPHAN_ITEM_KEY;
 	key.offset = 0;
 
+	root_key.type = BTRFS_ROOT_ITEM_KEY;
+	root_key.offset = (u64)-1;
+
 	while (1) {
 		ret = btrfs_search_slot(NULL, tree_root, &key, path, 0, 0);
 		if (ret < 0) {
@@ -288,19 +262,61 @@ int btrfs_find_orphan_roots(struct btrfs_root *tree_root)
 		}
 
 		btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
-		btrfs_release_path(tree_root, path);
+		btrfs_release_path(path);
 
 		if (key.objectid != BTRFS_ORPHAN_OBJECTID ||
 		    key.type != BTRFS_ORPHAN_ITEM_KEY)
 			break;
 
-		ret = btrfs_find_dead_roots(tree_root, key.offset);
-		if (ret) {
-			err = ret;
+		root_key.objectid = key.offset;
+		key.offset++;
+
+		root = btrfs_read_fs_root(tree_root, &root_key);
+		err = PTR_ERR_OR_ZERO(root);
+		if (err && err != -ENOENT) {
 			break;
+		} else if (err == -ENOENT) {
+			struct btrfs_trans_handle *trans;
+
+			btrfs_release_path(path);
+
+			trans = btrfs_join_transaction(tree_root);
+			if (IS_ERR(trans)) {
+				err = PTR_ERR(trans);
+				btrfs_error(tree_root->fs_info, err,
+					    "Failed to start trans to delete "
+					    "orphan item");
+				break;
+			}
+			err = btrfs_del_orphan_item(trans, tree_root,
+						    root_key.objectid);
+			btrfs_end_transaction(trans, tree_root);
+			if (err) {
+				btrfs_error(tree_root->fs_info, err,
+					    "Failed to delete root orphan "
+					    "item");
+				break;
+			}
+			continue;
 		}
 
-		key.offset++;
+		err = btrfs_init_fs_root(root);
+		if (err) {
+			btrfs_free_fs_root(root);
+			break;
+		}
+
+		set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state);
+
+		err = btrfs_insert_fs_root(root->fs_info, root);
+		if (err) {
+			BUG_ON(err == -EEXIST);
+			btrfs_free_fs_root(root);
+			break;
+		}
+
+		if (btrfs_root_refs(&root->root_item) == 0)
+			btrfs_add_dead_root(root);
 	}
 
 	btrfs_free_path(path);
@@ -313,22 +329,16 @@ int btrfs_del_root(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 {
 	struct btrfs_path *path;
 	int ret;
-	u32 refs;
-	struct btrfs_root_item *ri;
-	struct extent_buffer *leaf;
 
 	path = btrfs_alloc_path();
-	BUG_ON(!path);
+	if (!path)
+		return -ENOMEM;
 	ret = btrfs_search_slot(trans, root, key, path, -1, 1);
 	if (ret < 0)
 		goto out;
 
 	BUG_ON(ret != 0);
-	leaf = path->nodes[0];
-	ri = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_root_item);
 
-	refs = btrfs_disk_root_refs(leaf, ri);
-	BUG_ON(refs != 0);
 	ret = btrfs_del_item(trans, root, path);
 out:
 	btrfs_free_path(path);
@@ -371,37 +381,26 @@ again:
 		*sequence = btrfs_root_ref_sequence(leaf, ref);
 
 		ret = btrfs_del_item(trans, tree_root, path);
-		BUG_ON(ret);
+		if (ret) {
+			err = ret;
+			goto out;
+		}
 	} else
 		err = -ENOENT;
 
 	if (key.type == BTRFS_ROOT_BACKREF_KEY) {
-		btrfs_release_path(tree_root, path);
+		btrfs_release_path(path);
 		key.objectid = ref_id;
 		key.type = BTRFS_ROOT_REF_KEY;
 		key.offset = root_id;
 		goto again;
 	}
 
+out:
 	btrfs_free_path(path);
 	return err;
 }
 
-int btrfs_find_root_ref(struct btrfs_root *tree_root,
-		   struct btrfs_path *path,
-		   u64 root_id, u64 ref_id)
-{
-	struct btrfs_key key;
-	int ret;
-
-	key.objectid = root_id;
-	key.type = BTRFS_ROOT_REF_KEY;
-	key.offset = ref_id;
-
-	ret = btrfs_search_slot(NULL, tree_root, &key, path, 0, 0);
-	return ret;
-}
-
 /*
  * add a btrfs_root_ref item.  type is either BTRFS_ROOT_REF_KEY
  * or BTRFS_ROOT_BACKREF_KEY.
@@ -414,6 +413,8 @@ int btrfs_find_root_ref(struct btrfs_root *tree_root,
  *
  * For a back ref the root_id is the id of the subvol or snapshot and
  * ref_id is the id of the tree referencing it.
+ *
+ * Will return 0, -ENOMEM, or anything from the CoW path
  */
 int btrfs_add_root_ref(struct btrfs_trans_handle *trans,
 		       struct btrfs_root *tree_root,
@@ -437,7 +438,11 @@ int btrfs_add_root_ref(struct btrfs_trans_handle *trans,
 again:
 	ret = btrfs_insert_empty_item(trans, tree_root, path, &key,
 				      sizeof(*ref) + name_len);
-	BUG_ON(ret);
+	if (ret) {
+		btrfs_abort_transaction(trans, tree_root, ret);
+		btrfs_free_path(path);
+		return ret;
+	}
 
 	leaf = path->nodes[0];
 	ref = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_root_ref);
@@ -449,7 +454,7 @@ again:
 	btrfs_mark_buffer_dirty(leaf);
 
 	if (key.type == BTRFS_ROOT_BACKREF_KEY) {
-		btrfs_release_path(tree_root, path);
+		btrfs_release_path(path);
 		key.objectid = ref_id;
 		key.type = BTRFS_ROOT_REF_KEY;
 		key.offset = root_id;
@@ -459,3 +464,34 @@ again:
 	btrfs_free_path(path);
 	return 0;
 }
+
+/*
+ * Old btrfs forgets to init root_item->flags and root_item->byte_limit
+ * for subvolumes. To work around this problem, we steal a bit from
+ * root_item->inode_item->flags, and use it to indicate if those fields
+ * have been properly initialized.
+ */
+void btrfs_check_and_init_root_item(struct btrfs_root_item *root_item)
+{
+	u64 inode_flags = btrfs_stack_inode_flags(&root_item->inode);
+
+	if (!(inode_flags & BTRFS_INODE_ROOT_ITEM_INIT)) {
+		inode_flags |= BTRFS_INODE_ROOT_ITEM_INIT;
+		btrfs_set_stack_inode_flags(&root_item->inode, inode_flags);
+		btrfs_set_root_flags(root_item, 0);
+		btrfs_set_root_limit(root_item, 0);
+	}
+}
+
+void btrfs_update_root_times(struct btrfs_trans_handle *trans,
+			     struct btrfs_root *root)
+{
+	struct btrfs_root_item *item = &root->root_item;
+	struct timespec ct = CURRENT_TIME;
+
+	spin_lock(&root->root_item_lock);
+	btrfs_set_root_ctransid(item, trans->transid);
+	btrfs_set_stack_timespec_sec(&item->ctime, ct.tv_sec);
+	btrfs_set_stack_timespec_nsec(&item->ctime, ct.tv_nsec);
+	spin_unlock(&root->root_item_lock);
+}
diff --git a/fs/btrfs/scrub.c b/fs/btrfs/scrub.c
new file mode 100644
index 00000000000..b6d198f5181
--- /dev/null
+++ b/fs/btrfs/scrub.c
@@ -0,0 +1,3489 @@
+/*
+ * Copyright (C) 2011, 2012 STRATO.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/blkdev.h>
+#include <linux/ratelimit.h>
+#include "ctree.h"
+#include "volumes.h"
+#include "disk-io.h"
+#include "ordered-data.h"
+#include "transaction.h"
+#include "backref.h"
+#include "extent_io.h"
+#include "dev-replace.h"
+#include "check-integrity.h"
+#include "rcu-string.h"
+#include "raid56.h"
+
+/*
+ * This is only the first step towards a full-features scrub. It reads all
+ * extent and super block and verifies the checksums. In case a bad checksum
+ * is found or the extent cannot be read, good data will be written back if
+ * any can be found.
+ *
+ * Future enhancements:
+ *  - In case an unrepairable extent is encountered, track which files are
+ *    affected and report them
+ *  - track and record media errors, throw out bad devices
+ *  - add a mode to also read unallocated space
+ */
+
+struct scrub_block;
+struct scrub_ctx;
+
+/*
+ * the following three values only influence the performance.
+ * The last one configures the number of parallel and outstanding I/O
+ * operations. The first two values configure an upper limit for the number
+ * of (dynamically allocated) pages that are added to a bio.
+ */
+#define SCRUB_PAGES_PER_RD_BIO	32	/* 128k per bio */
+#define SCRUB_PAGES_PER_WR_BIO	32	/* 128k per bio */
+#define SCRUB_BIOS_PER_SCTX	64	/* 8MB per device in flight */
+
+/*
+ * the following value times PAGE_SIZE needs to be large enough to match the
+ * largest node/leaf/sector size that shall be supported.
+ * Values larger than BTRFS_STRIPE_LEN are not supported.
+ */
+#define SCRUB_MAX_PAGES_PER_BLOCK	16	/* 64k per node/leaf/sector */
+
+struct scrub_page {
+	struct scrub_block	*sblock;
+	struct page		*page;
+	struct btrfs_device	*dev;
+	u64			flags;  /* extent flags */
+	u64			generation;
+	u64			logical;
+	u64			physical;
+	u64			physical_for_dev_replace;
+	atomic_t		ref_count;
+	struct {
+		unsigned int	mirror_num:8;
+		unsigned int	have_csum:1;
+		unsigned int	io_error:1;
+	};
+	u8			csum[BTRFS_CSUM_SIZE];
+};
+
+struct scrub_bio {
+	int			index;
+	struct scrub_ctx	*sctx;
+	struct btrfs_device	*dev;
+	struct bio		*bio;
+	int			err;
+	u64			logical;
+	u64			physical;
+#if SCRUB_PAGES_PER_WR_BIO >= SCRUB_PAGES_PER_RD_BIO
+	struct scrub_page	*pagev[SCRUB_PAGES_PER_WR_BIO];
+#else
+	struct scrub_page	*pagev[SCRUB_PAGES_PER_RD_BIO];
+#endif
+	int			page_count;
+	int			next_free;
+	struct btrfs_work	work;
+};
+
+struct scrub_block {
+	struct scrub_page	*pagev[SCRUB_MAX_PAGES_PER_BLOCK];
+	int			page_count;
+	atomic_t		outstanding_pages;
+	atomic_t		ref_count; /* free mem on transition to zero */
+	struct scrub_ctx	*sctx;
+	struct {
+		unsigned int	header_error:1;
+		unsigned int	checksum_error:1;
+		unsigned int	no_io_error_seen:1;
+		unsigned int	generation_error:1; /* also sets header_error */
+	};
+};
+
+struct scrub_wr_ctx {
+	struct scrub_bio *wr_curr_bio;
+	struct btrfs_device *tgtdev;
+	int pages_per_wr_bio; /* <= SCRUB_PAGES_PER_WR_BIO */
+	atomic_t flush_all_writes;
+	struct mutex wr_lock;
+};
+
+struct scrub_ctx {
+	struct scrub_bio	*bios[SCRUB_BIOS_PER_SCTX];
+	struct btrfs_root	*dev_root;
+	int			first_free;
+	int			curr;
+	atomic_t		bios_in_flight;
+	atomic_t		workers_pending;
+	spinlock_t		list_lock;
+	wait_queue_head_t	list_wait;
+	u16			csum_size;
+	struct list_head	csum_list;
+	atomic_t		cancel_req;
+	int			readonly;
+	int			pages_per_rd_bio;
+	u32			sectorsize;
+	u32			nodesize;
+	u32			leafsize;
+
+	int			is_dev_replace;
+	struct scrub_wr_ctx	wr_ctx;
+
+	/*
+	 * statistics
+	 */
+	struct btrfs_scrub_progress stat;
+	spinlock_t		stat_lock;
+};
+
+struct scrub_fixup_nodatasum {
+	struct scrub_ctx	*sctx;
+	struct btrfs_device	*dev;
+	u64			logical;
+	struct btrfs_root	*root;
+	struct btrfs_work	work;
+	int			mirror_num;
+};
+
+struct scrub_nocow_inode {
+	u64			inum;
+	u64			offset;
+	u64			root;
+	struct list_head	list;
+};
+
+struct scrub_copy_nocow_ctx {
+	struct scrub_ctx	*sctx;
+	u64			logical;
+	u64			len;
+	int			mirror_num;
+	u64			physical_for_dev_replace;
+	struct list_head	inodes;
+	struct btrfs_work	work;
+};
+
+struct scrub_warning {
+	struct btrfs_path	*path;
+	u64			extent_item_size;
+	char			*scratch_buf;
+	char			*msg_buf;
+	const char		*errstr;
+	sector_t		sector;
+	u64			logical;
+	struct btrfs_device	*dev;
+	int			msg_bufsize;
+	int			scratch_bufsize;
+};
+
+
+static void scrub_pending_bio_inc(struct scrub_ctx *sctx);
+static void scrub_pending_bio_dec(struct scrub_ctx *sctx);
+static void scrub_pending_trans_workers_inc(struct scrub_ctx *sctx);
+static void scrub_pending_trans_workers_dec(struct scrub_ctx *sctx);
+static int scrub_handle_errored_block(struct scrub_block *sblock_to_check);
+static int scrub_setup_recheck_block(struct scrub_ctx *sctx,
+				     struct btrfs_fs_info *fs_info,
+				     struct scrub_block *original_sblock,
+				     u64 length, u64 logical,
+				     struct scrub_block *sblocks_for_recheck);
+static void scrub_recheck_block(struct btrfs_fs_info *fs_info,
+				struct scrub_block *sblock, int is_metadata,
+				int have_csum, u8 *csum, u64 generation,
+				u16 csum_size);
+static void scrub_recheck_block_checksum(struct btrfs_fs_info *fs_info,
+					 struct scrub_block *sblock,
+					 int is_metadata, int have_csum,
+					 const u8 *csum, u64 generation,
+					 u16 csum_size);
+static int scrub_repair_block_from_good_copy(struct scrub_block *sblock_bad,
+					     struct scrub_block *sblock_good,
+					     int force_write);
+static int scrub_repair_page_from_good_copy(struct scrub_block *sblock_bad,
+					    struct scrub_block *sblock_good,
+					    int page_num, int force_write);
+static void scrub_write_block_to_dev_replace(struct scrub_block *sblock);
+static int scrub_write_page_to_dev_replace(struct scrub_block *sblock,
+					   int page_num);
+static int scrub_checksum_data(struct scrub_block *sblock);
+static int scrub_checksum_tree_block(struct scrub_block *sblock);
+static int scrub_checksum_super(struct scrub_block *sblock);
+static void scrub_block_get(struct scrub_block *sblock);
+static void scrub_block_put(struct scrub_block *sblock);
+static void scrub_page_get(struct scrub_page *spage);
+static void scrub_page_put(struct scrub_page *spage);
+static int scrub_add_page_to_rd_bio(struct scrub_ctx *sctx,
+				    struct scrub_page *spage);
+static int scrub_pages(struct scrub_ctx *sctx, u64 logical, u64 len,
+		       u64 physical, struct btrfs_device *dev, u64 flags,
+		       u64 gen, int mirror_num, u8 *csum, int force,
+		       u64 physical_for_dev_replace);
+static void scrub_bio_end_io(struct bio *bio, int err);
+static void scrub_bio_end_io_worker(struct btrfs_work *work);
+static void scrub_block_complete(struct scrub_block *sblock);
+static void scrub_remap_extent(struct btrfs_fs_info *fs_info,
+			       u64 extent_logical, u64 extent_len,
+			       u64 *extent_physical,
+			       struct btrfs_device **extent_dev,
+			       int *extent_mirror_num);
+static int scrub_setup_wr_ctx(struct scrub_ctx *sctx,
+			      struct scrub_wr_ctx *wr_ctx,
+			      struct btrfs_fs_info *fs_info,
+			      struct btrfs_device *dev,
+			      int is_dev_replace);
+static void scrub_free_wr_ctx(struct scrub_wr_ctx *wr_ctx);
+static int scrub_add_page_to_wr_bio(struct scrub_ctx *sctx,
+				    struct scrub_page *spage);
+static void scrub_wr_submit(struct scrub_ctx *sctx);
+static void scrub_wr_bio_end_io(struct bio *bio, int err);
+static void scrub_wr_bio_end_io_worker(struct btrfs_work *work);
+static int write_page_nocow(struct scrub_ctx *sctx,
+			    u64 physical_for_dev_replace, struct page *page);
+static int copy_nocow_pages_for_inode(u64 inum, u64 offset, u64 root,
+				      struct scrub_copy_nocow_ctx *ctx);
+static int copy_nocow_pages(struct scrub_ctx *sctx, u64 logical, u64 len,
+			    int mirror_num, u64 physical_for_dev_replace);
+static void copy_nocow_pages_worker(struct btrfs_work *work);
+static void __scrub_blocked_if_needed(struct btrfs_fs_info *fs_info);
+static void scrub_blocked_if_needed(struct btrfs_fs_info *fs_info);
+
+
+static void scrub_pending_bio_inc(struct scrub_ctx *sctx)
+{
+	atomic_inc(&sctx->bios_in_flight);
+}
+
+static void scrub_pending_bio_dec(struct scrub_ctx *sctx)
+{
+	atomic_dec(&sctx->bios_in_flight);
+	wake_up(&sctx->list_wait);
+}
+
+static void __scrub_blocked_if_needed(struct btrfs_fs_info *fs_info)
+{
+	while (atomic_read(&fs_info->scrub_pause_req)) {
+		mutex_unlock(&fs_info->scrub_lock);
+		wait_event(fs_info->scrub_pause_wait,
+		   atomic_read(&fs_info->scrub_pause_req) == 0);
+		mutex_lock(&fs_info->scrub_lock);
+	}
+}
+
+static void scrub_blocked_if_needed(struct btrfs_fs_info *fs_info)
+{
+	atomic_inc(&fs_info->scrubs_paused);
+	wake_up(&fs_info->scrub_pause_wait);
+
+	mutex_lock(&fs_info->scrub_lock);
+	__scrub_blocked_if_needed(fs_info);
+	atomic_dec(&fs_info->scrubs_paused);
+	mutex_unlock(&fs_info->scrub_lock);
+
+	wake_up(&fs_info->scrub_pause_wait);
+}
+
+/*
+ * used for workers that require transaction commits (i.e., for the
+ * NOCOW case)
+ */
+static void scrub_pending_trans_workers_inc(struct scrub_ctx *sctx)
+{
+	struct btrfs_fs_info *fs_info = sctx->dev_root->fs_info;
+
+	/*
+	 * increment scrubs_running to prevent cancel requests from
+	 * completing as long as a worker is running. we must also
+	 * increment scrubs_paused to prevent deadlocking on pause
+	 * requests used for transactions commits (as the worker uses a
+	 * transaction context). it is safe to regard the worker
+	 * as paused for all matters practical. effectively, we only
+	 * avoid cancellation requests from completing.
+	 */
+	mutex_lock(&fs_info->scrub_lock);
+	atomic_inc(&fs_info->scrubs_running);
+	atomic_inc(&fs_info->scrubs_paused);
+	mutex_unlock(&fs_info->scrub_lock);
+
+	/*
+	 * check if @scrubs_running=@scrubs_paused condition
+	 * inside wait_event() is not an atomic operation.
+	 * which means we may inc/dec @scrub_running/paused
+	 * at any time. Let's wake up @scrub_pause_wait as
+	 * much as we can to let commit transaction blocked less.
+	 */
+	wake_up(&fs_info->scrub_pause_wait);
+
+	atomic_inc(&sctx->workers_pending);
+}
+
+/* used for workers that require transaction commits */
+static void scrub_pending_trans_workers_dec(struct scrub_ctx *sctx)
+{
+	struct btrfs_fs_info *fs_info = sctx->dev_root->fs_info;
+
+	/*
+	 * see scrub_pending_trans_workers_inc() why we're pretending
+	 * to be paused in the scrub counters
+	 */
+	mutex_lock(&fs_info->scrub_lock);
+	atomic_dec(&fs_info->scrubs_running);
+	atomic_dec(&fs_info->scrubs_paused);
+	mutex_unlock(&fs_info->scrub_lock);
+	atomic_dec(&sctx->workers_pending);
+	wake_up(&fs_info->scrub_pause_wait);
+	wake_up(&sctx->list_wait);
+}
+
+static void scrub_free_csums(struct scrub_ctx *sctx)
+{
+	while (!list_empty(&sctx->csum_list)) {
+		struct btrfs_ordered_sum *sum;
+		sum = list_first_entry(&sctx->csum_list,
+				       struct btrfs_ordered_sum, list);
+		list_del(&sum->list);
+		kfree(sum);
+	}
+}
+
+static noinline_for_stack void scrub_free_ctx(struct scrub_ctx *sctx)
+{
+	int i;
+
+	if (!sctx)
+		return;
+
+	scrub_free_wr_ctx(&sctx->wr_ctx);
+
+	/* this can happen when scrub is cancelled */
+	if (sctx->curr != -1) {
+		struct scrub_bio *sbio = sctx->bios[sctx->curr];
+
+		for (i = 0; i < sbio->page_count; i++) {
+			WARN_ON(!sbio->pagev[i]->page);
+			scrub_block_put(sbio->pagev[i]->sblock);
+		}
+		bio_put(sbio->bio);
+	}
+
+	for (i = 0; i < SCRUB_BIOS_PER_SCTX; ++i) {
+		struct scrub_bio *sbio = sctx->bios[i];
+
+		if (!sbio)
+			break;
+		kfree(sbio);
+	}
+
+	scrub_free_csums(sctx);
+	kfree(sctx);
+}
+
+static noinline_for_stack
+struct scrub_ctx *scrub_setup_ctx(struct btrfs_device *dev, int is_dev_replace)
+{
+	struct scrub_ctx *sctx;
+	int		i;
+	struct btrfs_fs_info *fs_info = dev->dev_root->fs_info;
+	int pages_per_rd_bio;
+	int ret;
+
+	/*
+	 * the setting of pages_per_rd_bio is correct for scrub but might
+	 * be wrong for the dev_replace code where we might read from
+	 * different devices in the initial huge bios. However, that
+	 * code is able to correctly handle the case when adding a page
+	 * to a bio fails.
+	 */
+	if (dev->bdev)
+		pages_per_rd_bio = min_t(int, SCRUB_PAGES_PER_RD_BIO,
+					 bio_get_nr_vecs(dev->bdev));
+	else
+		pages_per_rd_bio = SCRUB_PAGES_PER_RD_BIO;
+	sctx = kzalloc(sizeof(*sctx), GFP_NOFS);
+	if (!sctx)
+		goto nomem;
+	sctx->is_dev_replace = is_dev_replace;
+	sctx->pages_per_rd_bio = pages_per_rd_bio;
+	sctx->curr = -1;
+	sctx->dev_root = dev->dev_root;
+	for (i = 0; i < SCRUB_BIOS_PER_SCTX; ++i) {
+		struct scrub_bio *sbio;
+
+		sbio = kzalloc(sizeof(*sbio), GFP_NOFS);
+		if (!sbio)
+			goto nomem;
+		sctx->bios[i] = sbio;
+
+		sbio->index = i;
+		sbio->sctx = sctx;
+		sbio->page_count = 0;
+		btrfs_init_work(&sbio->work, scrub_bio_end_io_worker,
+				NULL, NULL);
+
+		if (i != SCRUB_BIOS_PER_SCTX - 1)
+			sctx->bios[i]->next_free = i + 1;
+		else
+			sctx->bios[i]->next_free = -1;
+	}
+	sctx->first_free = 0;
+	sctx->nodesize = dev->dev_root->nodesize;
+	sctx->leafsize = dev->dev_root->leafsize;
+	sctx->sectorsize = dev->dev_root->sectorsize;
+	atomic_set(&sctx->bios_in_flight, 0);
+	atomic_set(&sctx->workers_pending, 0);
+	atomic_set(&sctx->cancel_req, 0);
+	sctx->csum_size = btrfs_super_csum_size(fs_info->super_copy);
+	INIT_LIST_HEAD(&sctx->csum_list);
+
+	spin_lock_init(&sctx->list_lock);
+	spin_lock_init(&sctx->stat_lock);
+	init_waitqueue_head(&sctx->list_wait);
+
+	ret = scrub_setup_wr_ctx(sctx, &sctx->wr_ctx, fs_info,
+				 fs_info->dev_replace.tgtdev, is_dev_replace);
+	if (ret) {
+		scrub_free_ctx(sctx);
+		return ERR_PTR(ret);
+	}
+	return sctx;
+
+nomem:
+	scrub_free_ctx(sctx);
+	return ERR_PTR(-ENOMEM);
+}
+
+static int scrub_print_warning_inode(u64 inum, u64 offset, u64 root,
+				     void *warn_ctx)
+{
+	u64 isize;
+	u32 nlink;
+	int ret;
+	int i;
+	struct extent_buffer *eb;
+	struct btrfs_inode_item *inode_item;
+	struct scrub_warning *swarn = warn_ctx;
+	struct btrfs_fs_info *fs_info = swarn->dev->dev_root->fs_info;
+	struct inode_fs_paths *ipath = NULL;
+	struct btrfs_root *local_root;
+	struct btrfs_key root_key;
+
+	root_key.objectid = root;
+	root_key.type = BTRFS_ROOT_ITEM_KEY;
+	root_key.offset = (u64)-1;
+	local_root = btrfs_read_fs_root_no_name(fs_info, &root_key);
+	if (IS_ERR(local_root)) {
+		ret = PTR_ERR(local_root);
+		goto err;
+	}
+
+	ret = inode_item_info(inum, 0, local_root, swarn->path);
+	if (ret) {
+		btrfs_release_path(swarn->path);
+		goto err;
+	}
+
+	eb = swarn->path->nodes[0];
+	inode_item = btrfs_item_ptr(eb, swarn->path->slots[0],
+					struct btrfs_inode_item);
+	isize = btrfs_inode_size(eb, inode_item);
+	nlink = btrfs_inode_nlink(eb, inode_item);
+	btrfs_release_path(swarn->path);
+
+	ipath = init_ipath(4096, local_root, swarn->path);
+	if (IS_ERR(ipath)) {
+		ret = PTR_ERR(ipath);
+		ipath = NULL;
+		goto err;
+	}
+	ret = paths_from_inode(inum, ipath);
+
+	if (ret < 0)
+		goto err;
+
+	/*
+	 * we deliberately ignore the bit ipath might have been too small to
+	 * hold all of the paths here
+	 */
+	for (i = 0; i < ipath->fspath->elem_cnt; ++i)
+		printk_in_rcu(KERN_WARNING "BTRFS: %s at logical %llu on dev "
+			"%s, sector %llu, root %llu, inode %llu, offset %llu, "
+			"length %llu, links %u (path: %s)\n", swarn->errstr,
+			swarn->logical, rcu_str_deref(swarn->dev->name),
+			(unsigned long long)swarn->sector, root, inum, offset,
+			min(isize - offset, (u64)PAGE_SIZE), nlink,
+			(char *)(unsigned long)ipath->fspath->val[i]);
+
+	free_ipath(ipath);
+	return 0;
+
+err:
+	printk_in_rcu(KERN_WARNING "BTRFS: %s at logical %llu on dev "
+		"%s, sector %llu, root %llu, inode %llu, offset %llu: path "
+		"resolving failed with ret=%d\n", swarn->errstr,
+		swarn->logical, rcu_str_deref(swarn->dev->name),
+		(unsigned long long)swarn->sector, root, inum, offset, ret);
+
+	free_ipath(ipath);
+	return 0;
+}
+
+static void scrub_print_warning(const char *errstr, struct scrub_block *sblock)
+{
+	struct btrfs_device *dev;
+	struct btrfs_fs_info *fs_info;
+	struct btrfs_path *path;
+	struct btrfs_key found_key;
+	struct extent_buffer *eb;
+	struct btrfs_extent_item *ei;
+	struct scrub_warning swarn;
+	unsigned long ptr = 0;
+	u64 extent_item_pos;
+	u64 flags = 0;
+	u64 ref_root;
+	u32 item_size;
+	u8 ref_level;
+	const int bufsize = 4096;
+	int ret;
+
+	WARN_ON(sblock->page_count < 1);
+	dev = sblock->pagev[0]->dev;
+	fs_info = sblock->sctx->dev_root->fs_info;
+
+	path = btrfs_alloc_path();
+
+	swarn.scratch_buf = kmalloc(bufsize, GFP_NOFS);
+	swarn.msg_buf = kmalloc(bufsize, GFP_NOFS);
+	swarn.sector = (sblock->pagev[0]->physical) >> 9;
+	swarn.logical = sblock->pagev[0]->logical;
+	swarn.errstr = errstr;
+	swarn.dev = NULL;
+	swarn.msg_bufsize = bufsize;
+	swarn.scratch_bufsize = bufsize;
+
+	if (!path || !swarn.scratch_buf || !swarn.msg_buf)
+		goto out;
+
+	ret = extent_from_logical(fs_info, swarn.logical, path, &found_key,
+				  &flags);
+	if (ret < 0)
+		goto out;
+
+	extent_item_pos = swarn.logical - found_key.objectid;
+	swarn.extent_item_size = found_key.offset;
+
+	eb = path->nodes[0];
+	ei = btrfs_item_ptr(eb, path->slots[0], struct btrfs_extent_item);
+	item_size = btrfs_item_size_nr(eb, path->slots[0]);
+
+	if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
+		do {
+			ret = tree_backref_for_extent(&ptr, eb, &found_key, ei,
+						      item_size, &ref_root,
+						      &ref_level);
+			printk_in_rcu(KERN_WARNING
+				"BTRFS: %s at logical %llu on dev %s, "
+				"sector %llu: metadata %s (level %d) in tree "
+				"%llu\n", errstr, swarn.logical,
+				rcu_str_deref(dev->name),
+				(unsigned long long)swarn.sector,
+				ref_level ? "node" : "leaf",
+				ret < 0 ? -1 : ref_level,
+				ret < 0 ? -1 : ref_root);
+		} while (ret != 1);
+		btrfs_release_path(path);
+	} else {
+		btrfs_release_path(path);
+		swarn.path = path;
+		swarn.dev = dev;
+		iterate_extent_inodes(fs_info, found_key.objectid,
+					extent_item_pos, 1,
+					scrub_print_warning_inode, &swarn);
+	}
+
+out:
+	btrfs_free_path(path);
+	kfree(swarn.scratch_buf);
+	kfree(swarn.msg_buf);
+}
+
+static int scrub_fixup_readpage(u64 inum, u64 offset, u64 root, void *fixup_ctx)
+{
+	struct page *page = NULL;
+	unsigned long index;
+	struct scrub_fixup_nodatasum *fixup = fixup_ctx;
+	int ret;
+	int corrected = 0;
+	struct btrfs_key key;
+	struct inode *inode = NULL;
+	struct btrfs_fs_info *fs_info;
+	u64 end = offset + PAGE_SIZE - 1;
+	struct btrfs_root *local_root;
+	int srcu_index;
+
+	key.objectid = root;
+	key.type = BTRFS_ROOT_ITEM_KEY;
+	key.offset = (u64)-1;
+
+	fs_info = fixup->root->fs_info;
+	srcu_index = srcu_read_lock(&fs_info->subvol_srcu);
+
+	local_root = btrfs_read_fs_root_no_name(fs_info, &key);
+	if (IS_ERR(local_root)) {
+		srcu_read_unlock(&fs_info->subvol_srcu, srcu_index);
+		return PTR_ERR(local_root);
+	}
+
+	key.type = BTRFS_INODE_ITEM_KEY;
+	key.objectid = inum;
+	key.offset = 0;
+	inode = btrfs_iget(fs_info->sb, &key, local_root, NULL);
+	srcu_read_unlock(&fs_info->subvol_srcu, srcu_index);
+	if (IS_ERR(inode))
+		return PTR_ERR(inode);
+
+	index = offset >> PAGE_CACHE_SHIFT;
+
+	page = find_or_create_page(inode->i_mapping, index, GFP_NOFS);
+	if (!page) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	if (PageUptodate(page)) {
+		if (PageDirty(page)) {
+			/*
+			 * we need to write the data to the defect sector. the
+			 * data that was in that sector is not in memory,
+			 * because the page was modified. we must not write the
+			 * modified page to that sector.
+			 *
+			 * TODO: what could be done here: wait for the delalloc
+			 *       runner to write out that page (might involve
+			 *       COW) and see whether the sector is still
+			 *       referenced afterwards.
+			 *
+			 * For the meantime, we'll treat this error
+			 * incorrectable, although there is a chance that a
+			 * later scrub will find the bad sector again and that
+			 * there's no dirty page in memory, then.
+			 */
+			ret = -EIO;
+			goto out;
+		}
+		fs_info = BTRFS_I(inode)->root->fs_info;
+		ret = repair_io_failure(fs_info, offset, PAGE_SIZE,
+					fixup->logical, page,
+					fixup->mirror_num);
+		unlock_page(page);
+		corrected = !ret;
+	} else {
+		/*
+		 * we need to get good data first. the general readpage path
+		 * will call repair_io_failure for us, we just have to make
+		 * sure we read the bad mirror.
+		 */
+		ret = set_extent_bits(&BTRFS_I(inode)->io_tree, offset, end,
+					EXTENT_DAMAGED, GFP_NOFS);
+		if (ret) {
+			/* set_extent_bits should give proper error */
+			WARN_ON(ret > 0);
+			if (ret > 0)
+				ret = -EFAULT;
+			goto out;
+		}
+
+		ret = extent_read_full_page(&BTRFS_I(inode)->io_tree, page,
+						btrfs_get_extent,
+						fixup->mirror_num);
+		wait_on_page_locked(page);
+
+		corrected = !test_range_bit(&BTRFS_I(inode)->io_tree, offset,
+						end, EXTENT_DAMAGED, 0, NULL);
+		if (!corrected)
+			clear_extent_bits(&BTRFS_I(inode)->io_tree, offset, end,
+						EXTENT_DAMAGED, GFP_NOFS);
+	}
+
+out:
+	if (page)
+		put_page(page);
+
+	iput(inode);
+
+	if (ret < 0)
+		return ret;
+
+	if (ret == 0 && corrected) {
+		/*
+		 * we only need to call readpage for one of the inodes belonging
+		 * to this extent. so make iterate_extent_inodes stop
+		 */
+		return 1;
+	}
+
+	return -EIO;
+}
+
+static void scrub_fixup_nodatasum(struct btrfs_work *work)
+{
+	int ret;
+	struct scrub_fixup_nodatasum *fixup;
+	struct scrub_ctx *sctx;
+	struct btrfs_trans_handle *trans = NULL;
+	struct btrfs_path *path;
+	int uncorrectable = 0;
+
+	fixup = container_of(work, struct scrub_fixup_nodatasum, work);
+	sctx = fixup->sctx;
+
+	path = btrfs_alloc_path();
+	if (!path) {
+		spin_lock(&sctx->stat_lock);
+		++sctx->stat.malloc_errors;
+		spin_unlock(&sctx->stat_lock);
+		uncorrectable = 1;
+		goto out;
+	}
+
+	trans = btrfs_join_transaction(fixup->root);
+	if (IS_ERR(trans)) {
+		uncorrectable = 1;
+		goto out;
+	}
+
+	/*
+	 * the idea is to trigger a regular read through the standard path. we
+	 * read a page from the (failed) logical address by specifying the
+	 * corresponding copynum of the failed sector. thus, that readpage is
+	 * expected to fail.
+	 * that is the point where on-the-fly error correction will kick in
+	 * (once it's finished) and rewrite the failed sector if a good copy
+	 * can be found.
+	 */
+	ret = iterate_inodes_from_logical(fixup->logical, fixup->root->fs_info,
+						path, scrub_fixup_readpage,
+						fixup);
+	if (ret < 0) {
+		uncorrectable = 1;
+		goto out;
+	}
+	WARN_ON(ret != 1);
+
+	spin_lock(&sctx->stat_lock);
+	++sctx->stat.corrected_errors;
+	spin_unlock(&sctx->stat_lock);
+
+out:
+	if (trans && !IS_ERR(trans))
+		btrfs_end_transaction(trans, fixup->root);
+	if (uncorrectable) {
+		spin_lock(&sctx->stat_lock);
+		++sctx->stat.uncorrectable_errors;
+		spin_unlock(&sctx->stat_lock);
+		btrfs_dev_replace_stats_inc(
+			&sctx->dev_root->fs_info->dev_replace.
+			num_uncorrectable_read_errors);
+		printk_ratelimited_in_rcu(KERN_ERR "BTRFS: "
+		    "unable to fixup (nodatasum) error at logical %llu on dev %s\n",
+			fixup->logical, rcu_str_deref(fixup->dev->name));
+	}
+
+	btrfs_free_path(path);
+	kfree(fixup);
+
+	scrub_pending_trans_workers_dec(sctx);
+}
+
+/*
+ * scrub_handle_errored_block gets called when either verification of the
+ * pages failed or the bio failed to read, e.g. with EIO. In the latter
+ * case, this function handles all pages in the bio, even though only one
+ * may be bad.
+ * The goal of this function is to repair the errored block by using the
+ * contents of one of the mirrors.
+ */
+static int scrub_handle_errored_block(struct scrub_block *sblock_to_check)
+{
+	struct scrub_ctx *sctx = sblock_to_check->sctx;
+	struct btrfs_device *dev;
+	struct btrfs_fs_info *fs_info;
+	u64 length;
+	u64 logical;
+	u64 generation;
+	unsigned int failed_mirror_index;
+	unsigned int is_metadata;
+	unsigned int have_csum;
+	u8 *csum;
+	struct scrub_block *sblocks_for_recheck; /* holds one for each mirror */
+	struct scrub_block *sblock_bad;
+	int ret;
+	int mirror_index;
+	int page_num;
+	int success;
+	static DEFINE_RATELIMIT_STATE(_rs, DEFAULT_RATELIMIT_INTERVAL,
+				      DEFAULT_RATELIMIT_BURST);
+
+	BUG_ON(sblock_to_check->page_count < 1);
+	fs_info = sctx->dev_root->fs_info;
+	if (sblock_to_check->pagev[0]->flags & BTRFS_EXTENT_FLAG_SUPER) {
+		/*
+		 * if we find an error in a super block, we just report it.
+		 * They will get written with the next transaction commit
+		 * anyway
+		 */
+		spin_lock(&sctx->stat_lock);
+		++sctx->stat.super_errors;
+		spin_unlock(&sctx->stat_lock);
+		return 0;
+	}
+	length = sblock_to_check->page_count * PAGE_SIZE;
+	logical = sblock_to_check->pagev[0]->logical;
+	generation = sblock_to_check->pagev[0]->generation;
+	BUG_ON(sblock_to_check->pagev[0]->mirror_num < 1);
+	failed_mirror_index = sblock_to_check->pagev[0]->mirror_num - 1;
+	is_metadata = !(sblock_to_check->pagev[0]->flags &
+			BTRFS_EXTENT_FLAG_DATA);
+	have_csum = sblock_to_check->pagev[0]->have_csum;
+	csum = sblock_to_check->pagev[0]->csum;
+	dev = sblock_to_check->pagev[0]->dev;
+
+	if (sctx->is_dev_replace && !is_metadata && !have_csum) {
+		sblocks_for_recheck = NULL;
+		goto nodatasum_case;
+	}
+
+	/*
+	 * read all mirrors one after the other. This includes to
+	 * re-read the extent or metadata block that failed (that was
+	 * the cause that this fixup code is called) another time,
+	 * page by page this time in order to know which pages
+	 * caused I/O errors and which ones are good (for all mirrors).
+	 * It is the goal to handle the situation when more than one
+	 * mirror contains I/O errors, but the errors do not
+	 * overlap, i.e. the data can be repaired by selecting the
+	 * pages from those mirrors without I/O error on the
+	 * particular pages. One example (with blocks >= 2 * PAGE_SIZE)
+	 * would be that mirror #1 has an I/O error on the first page,
+	 * the second page is good, and mirror #2 has an I/O error on
+	 * the second page, but the first page is good.
+	 * Then the first page of the first mirror can be repaired by
+	 * taking the first page of the second mirror, and the
+	 * second page of the second mirror can be repaired by
+	 * copying the contents of the 2nd page of the 1st mirror.
+	 * One more note: if the pages of one mirror contain I/O
+	 * errors, the checksum cannot be verified. In order to get
+	 * the best data for repairing, the first attempt is to find
+	 * a mirror without I/O errors and with a validated checksum.
+	 * Only if this is not possible, the pages are picked from
+	 * mirrors with I/O errors without considering the checksum.
+	 * If the latter is the case, at the end, the checksum of the
+	 * repaired area is verified in order to correctly maintain
+	 * the statistics.
+	 */
+
+	sblocks_for_recheck = kzalloc(BTRFS_MAX_MIRRORS *
+				     sizeof(*sblocks_for_recheck),
+				     GFP_NOFS);
+	if (!sblocks_for_recheck) {
+		spin_lock(&sctx->stat_lock);
+		sctx->stat.malloc_errors++;
+		sctx->stat.read_errors++;
+		sctx->stat.uncorrectable_errors++;
+		spin_unlock(&sctx->stat_lock);
+		btrfs_dev_stat_inc_and_print(dev, BTRFS_DEV_STAT_READ_ERRS);
+		goto out;
+	}
+
+	/* setup the context, map the logical blocks and alloc the pages */
+	ret = scrub_setup_recheck_block(sctx, fs_info, sblock_to_check, length,
+					logical, sblocks_for_recheck);
+	if (ret) {
+		spin_lock(&sctx->stat_lock);
+		sctx->stat.read_errors++;
+		sctx->stat.uncorrectable_errors++;
+		spin_unlock(&sctx->stat_lock);
+		btrfs_dev_stat_inc_and_print(dev, BTRFS_DEV_STAT_READ_ERRS);
+		goto out;
+	}
+	BUG_ON(failed_mirror_index >= BTRFS_MAX_MIRRORS);
+	sblock_bad = sblocks_for_recheck + failed_mirror_index;
+
+	/* build and submit the bios for the failed mirror, check checksums */
+	scrub_recheck_block(fs_info, sblock_bad, is_metadata, have_csum,
+			    csum, generation, sctx->csum_size);
+
+	if (!sblock_bad->header_error && !sblock_bad->checksum_error &&
+	    sblock_bad->no_io_error_seen) {
+		/*
+		 * the error disappeared after reading page by page, or
+		 * the area was part of a huge bio and other parts of the
+		 * bio caused I/O errors, or the block layer merged several
+		 * read requests into one and the error is caused by a
+		 * different bio (usually one of the two latter cases is
+		 * the cause)
+		 */
+		spin_lock(&sctx->stat_lock);
+		sctx->stat.unverified_errors++;
+		spin_unlock(&sctx->stat_lock);
+
+		if (sctx->is_dev_replace)
+			scrub_write_block_to_dev_replace(sblock_bad);
+		goto out;
+	}
+
+	if (!sblock_bad->no_io_error_seen) {
+		spin_lock(&sctx->stat_lock);
+		sctx->stat.read_errors++;
+		spin_unlock(&sctx->stat_lock);
+		if (__ratelimit(&_rs))
+			scrub_print_warning("i/o error", sblock_to_check);
+		btrfs_dev_stat_inc_and_print(dev, BTRFS_DEV_STAT_READ_ERRS);
+	} else if (sblock_bad->checksum_error) {
+		spin_lock(&sctx->stat_lock);
+		sctx->stat.csum_errors++;
+		spin_unlock(&sctx->stat_lock);
+		if (__ratelimit(&_rs))
+			scrub_print_warning("checksum error", sblock_to_check);
+		btrfs_dev_stat_inc_and_print(dev,
+					     BTRFS_DEV_STAT_CORRUPTION_ERRS);
+	} else if (sblock_bad->header_error) {
+		spin_lock(&sctx->stat_lock);
+		sctx->stat.verify_errors++;
+		spin_unlock(&sctx->stat_lock);
+		if (__ratelimit(&_rs))
+			scrub_print_warning("checksum/header error",
+					    sblock_to_check);
+		if (sblock_bad->generation_error)
+			btrfs_dev_stat_inc_and_print(dev,
+				BTRFS_DEV_STAT_GENERATION_ERRS);
+		else
+			btrfs_dev_stat_inc_and_print(dev,
+				BTRFS_DEV_STAT_CORRUPTION_ERRS);
+	}
+
+	if (sctx->readonly) {
+		ASSERT(!sctx->is_dev_replace);
+		goto out;
+	}
+
+	if (!is_metadata && !have_csum) {
+		struct scrub_fixup_nodatasum *fixup_nodatasum;
+
+nodatasum_case:
+		WARN_ON(sctx->is_dev_replace);
+
+		/*
+		 * !is_metadata and !have_csum, this means that the data
+		 * might not be COW'ed, that it might be modified
+		 * concurrently. The general strategy to work on the
+		 * commit root does not help in the case when COW is not
+		 * used.
+		 */
+		fixup_nodatasum = kzalloc(sizeof(*fixup_nodatasum), GFP_NOFS);
+		if (!fixup_nodatasum)
+			goto did_not_correct_error;
+		fixup_nodatasum->sctx = sctx;
+		fixup_nodatasum->dev = dev;
+		fixup_nodatasum->logical = logical;
+		fixup_nodatasum->root = fs_info->extent_root;
+		fixup_nodatasum->mirror_num = failed_mirror_index + 1;
+		scrub_pending_trans_workers_inc(sctx);
+		btrfs_init_work(&fixup_nodatasum->work, scrub_fixup_nodatasum,
+				NULL, NULL);
+		btrfs_queue_work(fs_info->scrub_workers,
+				 &fixup_nodatasum->work);
+		goto out;
+	}
+
+	/*
+	 * now build and submit the bios for the other mirrors, check
+	 * checksums.
+	 * First try to pick the mirror which is completely without I/O
+	 * errors and also does not have a checksum error.
+	 * If one is found, and if a checksum is present, the full block
+	 * that is known to contain an error is rewritten. Afterwards
+	 * the block is known to be corrected.
+	 * If a mirror is found which is completely correct, and no
+	 * checksum is present, only those pages are rewritten that had
+	 * an I/O error in the block to be repaired, since it cannot be
+	 * determined, which copy of the other pages is better (and it
+	 * could happen otherwise that a correct page would be
+	 * overwritten by a bad one).
+	 */
+	for (mirror_index = 0;
+	     mirror_index < BTRFS_MAX_MIRRORS &&
+	     sblocks_for_recheck[mirror_index].page_count > 0;
+	     mirror_index++) {
+		struct scrub_block *sblock_other;
+
+		if (mirror_index == failed_mirror_index)
+			continue;
+		sblock_other = sblocks_for_recheck + mirror_index;
+
+		/* build and submit the bios, check checksums */
+		scrub_recheck_block(fs_info, sblock_other, is_metadata,
+				    have_csum, csum, generation,
+				    sctx->csum_size);
+
+		if (!sblock_other->header_error &&
+		    !sblock_other->checksum_error &&
+		    sblock_other->no_io_error_seen) {
+			if (sctx->is_dev_replace) {
+				scrub_write_block_to_dev_replace(sblock_other);
+			} else {
+				int force_write = is_metadata || have_csum;
+
+				ret = scrub_repair_block_from_good_copy(
+						sblock_bad, sblock_other,
+						force_write);
+			}
+			if (0 == ret)
+				goto corrected_error;
+		}
+	}
+
+	/*
+	 * for dev_replace, pick good pages and write to the target device.
+	 */
+	if (sctx->is_dev_replace) {
+		success = 1;
+		for (page_num = 0; page_num < sblock_bad->page_count;
+		     page_num++) {
+			int sub_success;
+
+			sub_success = 0;
+			for (mirror_index = 0;
+			     mirror_index < BTRFS_MAX_MIRRORS &&
+			     sblocks_for_recheck[mirror_index].page_count > 0;
+			     mirror_index++) {
+				struct scrub_block *sblock_other =
+					sblocks_for_recheck + mirror_index;
+				struct scrub_page *page_other =
+					sblock_other->pagev[page_num];
+
+				if (!page_other->io_error) {
+					ret = scrub_write_page_to_dev_replace(
+							sblock_other, page_num);
+					if (ret == 0) {
+						/* succeeded for this page */
+						sub_success = 1;
+						break;
+					} else {
+						btrfs_dev_replace_stats_inc(
+							&sctx->dev_root->
+							fs_info->dev_replace.
+							num_write_errors);
+					}
+				}
+			}
+
+			if (!sub_success) {
+				/*
+				 * did not find a mirror to fetch the page
+				 * from. scrub_write_page_to_dev_replace()
+				 * handles this case (page->io_error), by
+				 * filling the block with zeros before
+				 * submitting the write request
+				 */
+				success = 0;
+				ret = scrub_write_page_to_dev_replace(
+						sblock_bad, page_num);
+				if (ret)
+					btrfs_dev_replace_stats_inc(
+						&sctx->dev_root->fs_info->
+						dev_replace.num_write_errors);
+			}
+		}
+
+		goto out;
+	}
+
+	/*
+	 * for regular scrub, repair those pages that are errored.
+	 * In case of I/O errors in the area that is supposed to be
+	 * repaired, continue by picking good copies of those pages.
+	 * Select the good pages from mirrors to rewrite bad pages from
+	 * the area to fix. Afterwards verify the checksum of the block
+	 * that is supposed to be repaired. This verification step is
+	 * only done for the purpose of statistic counting and for the
+	 * final scrub report, whether errors remain.
+	 * A perfect algorithm could make use of the checksum and try
+	 * all possible combinations of pages from the different mirrors
+	 * until the checksum verification succeeds. For example, when
+	 * the 2nd page of mirror #1 faces I/O errors, and the 2nd page
+	 * of mirror #2 is readable but the final checksum test fails,
+	 * then the 2nd page of mirror #3 could be tried, whether now
+	 * the final checksum succeedes. But this would be a rare
+	 * exception and is therefore not implemented. At least it is
+	 * avoided that the good copy is overwritten.
+	 * A more useful improvement would be to pick the sectors
+	 * without I/O error based on sector sizes (512 bytes on legacy
+	 * disks) instead of on PAGE_SIZE. Then maybe 512 byte of one
+	 * mirror could be repaired by taking 512 byte of a different
+	 * mirror, even if other 512 byte sectors in the same PAGE_SIZE
+	 * area are unreadable.
+	 */
+
+	/* can only fix I/O errors from here on */
+	if (sblock_bad->no_io_error_seen)
+		goto did_not_correct_error;
+
+	success = 1;
+	for (page_num = 0; page_num < sblock_bad->page_count; page_num++) {
+		struct scrub_page *page_bad = sblock_bad->pagev[page_num];
+
+		if (!page_bad->io_error)
+			continue;
+
+		for (mirror_index = 0;
+		     mirror_index < BTRFS_MAX_MIRRORS &&
+		     sblocks_for_recheck[mirror_index].page_count > 0;
+		     mirror_index++) {
+			struct scrub_block *sblock_other = sblocks_for_recheck +
+							   mirror_index;
+			struct scrub_page *page_other = sblock_other->pagev[
+							page_num];
+
+			if (!page_other->io_error) {
+				ret = scrub_repair_page_from_good_copy(
+					sblock_bad, sblock_other, page_num, 0);
+				if (0 == ret) {
+					page_bad->io_error = 0;
+					break; /* succeeded for this page */
+				}
+			}
+		}
+
+		if (page_bad->io_error) {
+			/* did not find a mirror to copy the page from */
+			success = 0;
+		}
+	}
+
+	if (success) {
+		if (is_metadata || have_csum) {
+			/*
+			 * need to verify the checksum now that all
+			 * sectors on disk are repaired (the write
+			 * request for data to be repaired is on its way).
+			 * Just be lazy and use scrub_recheck_block()
+			 * which re-reads the data before the checksum
+			 * is verified, but most likely the data comes out
+			 * of the page cache.
+			 */
+			scrub_recheck_block(fs_info, sblock_bad,
+					    is_metadata, have_csum, csum,
+					    generation, sctx->csum_size);
+			if (!sblock_bad->header_error &&
+			    !sblock_bad->checksum_error &&
+			    sblock_bad->no_io_error_seen)
+				goto corrected_error;
+			else
+				goto did_not_correct_error;
+		} else {
+corrected_error:
+			spin_lock(&sctx->stat_lock);
+			sctx->stat.corrected_errors++;
+			spin_unlock(&sctx->stat_lock);
+			printk_ratelimited_in_rcu(KERN_ERR
+				"BTRFS: fixed up error at logical %llu on dev %s\n",
+				logical, rcu_str_deref(dev->name));
+		}
+	} else {
+did_not_correct_error:
+		spin_lock(&sctx->stat_lock);
+		sctx->stat.uncorrectable_errors++;
+		spin_unlock(&sctx->stat_lock);
+		printk_ratelimited_in_rcu(KERN_ERR
+			"BTRFS: unable to fixup (regular) error at logical %llu on dev %s\n",
+			logical, rcu_str_deref(dev->name));
+	}
+
+out:
+	if (sblocks_for_recheck) {
+		for (mirror_index = 0; mirror_index < BTRFS_MAX_MIRRORS;
+		     mirror_index++) {
+			struct scrub_block *sblock = sblocks_for_recheck +
+						     mirror_index;
+			int page_index;
+
+			for (page_index = 0; page_index < sblock->page_count;
+			     page_index++) {
+				sblock->pagev[page_index]->sblock = NULL;
+				scrub_page_put(sblock->pagev[page_index]);
+			}
+		}
+		kfree(sblocks_for_recheck);
+	}
+
+	return 0;
+}
+
+static int scrub_setup_recheck_block(struct scrub_ctx *sctx,
+				     struct btrfs_fs_info *fs_info,
+				     struct scrub_block *original_sblock,
+				     u64 length, u64 logical,
+				     struct scrub_block *sblocks_for_recheck)
+{
+	int page_index;
+	int mirror_index;
+	int ret;
+
+	/*
+	 * note: the two members ref_count and outstanding_pages
+	 * are not used (and not set) in the blocks that are used for
+	 * the recheck procedure
+	 */
+
+	page_index = 0;
+	while (length > 0) {
+		u64 sublen = min_t(u64, length, PAGE_SIZE);
+		u64 mapped_length = sublen;
+		struct btrfs_bio *bbio = NULL;
+
+		/*
+		 * with a length of PAGE_SIZE, each returned stripe
+		 * represents one mirror
+		 */
+		ret = btrfs_map_block(fs_info, REQ_GET_READ_MIRRORS, logical,
+				      &mapped_length, &bbio, 0);
+		if (ret || !bbio || mapped_length < sublen) {
+			kfree(bbio);
+			return -EIO;
+		}
+
+		BUG_ON(page_index >= SCRUB_PAGES_PER_RD_BIO);
+		for (mirror_index = 0; mirror_index < (int)bbio->num_stripes;
+		     mirror_index++) {
+			struct scrub_block *sblock;
+			struct scrub_page *page;
+
+			if (mirror_index >= BTRFS_MAX_MIRRORS)
+				continue;
+
+			sblock = sblocks_for_recheck + mirror_index;
+			sblock->sctx = sctx;
+			page = kzalloc(sizeof(*page), GFP_NOFS);
+			if (!page) {
+leave_nomem:
+				spin_lock(&sctx->stat_lock);
+				sctx->stat.malloc_errors++;
+				spin_unlock(&sctx->stat_lock);
+				kfree(bbio);
+				return -ENOMEM;
+			}
+			scrub_page_get(page);
+			sblock->pagev[page_index] = page;
+			page->logical = logical;
+			page->physical = bbio->stripes[mirror_index].physical;
+			BUG_ON(page_index >= original_sblock->page_count);
+			page->physical_for_dev_replace =
+				original_sblock->pagev[page_index]->
+				physical_for_dev_replace;
+			/* for missing devices, dev->bdev is NULL */
+			page->dev = bbio->stripes[mirror_index].dev;
+			page->mirror_num = mirror_index + 1;
+			sblock->page_count++;
+			page->page = alloc_page(GFP_NOFS);
+			if (!page->page)
+				goto leave_nomem;
+		}
+		kfree(bbio);
+		length -= sublen;
+		logical += sublen;
+		page_index++;
+	}
+
+	return 0;
+}
+
+/*
+ * this function will check the on disk data for checksum errors, header
+ * errors and read I/O errors. If any I/O errors happen, the exact pages
+ * which are errored are marked as being bad. The goal is to enable scrub
+ * to take those pages that are not errored from all the mirrors so that
+ * the pages that are errored in the just handled mirror can be repaired.
+ */
+static void scrub_recheck_block(struct btrfs_fs_info *fs_info,
+				struct scrub_block *sblock, int is_metadata,
+				int have_csum, u8 *csum, u64 generation,
+				u16 csum_size)
+{
+	int page_num;
+
+	sblock->no_io_error_seen = 1;
+	sblock->header_error = 0;
+	sblock->checksum_error = 0;
+
+	for (page_num = 0; page_num < sblock->page_count; page_num++) {
+		struct bio *bio;
+		struct scrub_page *page = sblock->pagev[page_num];
+
+		if (page->dev->bdev == NULL) {
+			page->io_error = 1;
+			sblock->no_io_error_seen = 0;
+			continue;
+		}
+
+		WARN_ON(!page->page);
+		bio = btrfs_io_bio_alloc(GFP_NOFS, 1);
+		if (!bio) {
+			page->io_error = 1;
+			sblock->no_io_error_seen = 0;
+			continue;
+		}
+		bio->bi_bdev = page->dev->bdev;
+		bio->bi_iter.bi_sector = page->physical >> 9;
+
+		bio_add_page(bio, page->page, PAGE_SIZE, 0);
+		if (btrfsic_submit_bio_wait(READ, bio))
+			sblock->no_io_error_seen = 0;
+
+		bio_put(bio);
+	}
+
+	if (sblock->no_io_error_seen)
+		scrub_recheck_block_checksum(fs_info, sblock, is_metadata,
+					     have_csum, csum, generation,
+					     csum_size);
+
+	return;
+}
+
+static void scrub_recheck_block_checksum(struct btrfs_fs_info *fs_info,
+					 struct scrub_block *sblock,
+					 int is_metadata, int have_csum,
+					 const u8 *csum, u64 generation,
+					 u16 csum_size)
+{
+	int page_num;
+	u8 calculated_csum[BTRFS_CSUM_SIZE];
+	u32 crc = ~(u32)0;
+	void *mapped_buffer;
+
+	WARN_ON(!sblock->pagev[0]->page);
+	if (is_metadata) {
+		struct btrfs_header *h;
+
+		mapped_buffer = kmap_atomic(sblock->pagev[0]->page);
+		h = (struct btrfs_header *)mapped_buffer;
+
+		if (sblock->pagev[0]->logical != btrfs_stack_header_bytenr(h) ||
+		    memcmp(h->fsid, fs_info->fsid, BTRFS_UUID_SIZE) ||
+		    memcmp(h->chunk_tree_uuid, fs_info->chunk_tree_uuid,
+			   BTRFS_UUID_SIZE)) {
+			sblock->header_error = 1;
+		} else if (generation != btrfs_stack_header_generation(h)) {
+			sblock->header_error = 1;
+			sblock->generation_error = 1;
+		}
+		csum = h->csum;
+	} else {
+		if (!have_csum)
+			return;
+
+		mapped_buffer = kmap_atomic(sblock->pagev[0]->page);
+	}
+
+	for (page_num = 0;;) {
+		if (page_num == 0 && is_metadata)
+			crc = btrfs_csum_data(
+				((u8 *)mapped_buffer) + BTRFS_CSUM_SIZE,
+				crc, PAGE_SIZE - BTRFS_CSUM_SIZE);
+		else
+			crc = btrfs_csum_data(mapped_buffer, crc, PAGE_SIZE);
+
+		kunmap_atomic(mapped_buffer);
+		page_num++;
+		if (page_num >= sblock->page_count)
+			break;
+		WARN_ON(!sblock->pagev[page_num]->page);
+
+		mapped_buffer = kmap_atomic(sblock->pagev[page_num]->page);
+	}
+
+	btrfs_csum_final(crc, calculated_csum);
+	if (memcmp(calculated_csum, csum, csum_size))
+		sblock->checksum_error = 1;
+}
+
+static int scrub_repair_block_from_good_copy(struct scrub_block *sblock_bad,
+					     struct scrub_block *sblock_good,
+					     int force_write)
+{
+	int page_num;
+	int ret = 0;
+
+	for (page_num = 0; page_num < sblock_bad->page_count; page_num++) {
+		int ret_sub;
+
+		ret_sub = scrub_repair_page_from_good_copy(sblock_bad,
+							   sblock_good,
+							   page_num,
+							   force_write);
+		if (ret_sub)
+			ret = ret_sub;
+	}
+
+	return ret;
+}
+
+static int scrub_repair_page_from_good_copy(struct scrub_block *sblock_bad,
+					    struct scrub_block *sblock_good,
+					    int page_num, int force_write)
+{
+	struct scrub_page *page_bad = sblock_bad->pagev[page_num];
+	struct scrub_page *page_good = sblock_good->pagev[page_num];
+
+	BUG_ON(page_bad->page == NULL);
+	BUG_ON(page_good->page == NULL);
+	if (force_write || sblock_bad->header_error ||
+	    sblock_bad->checksum_error || page_bad->io_error) {
+		struct bio *bio;
+		int ret;
+
+		if (!page_bad->dev->bdev) {
+			printk_ratelimited(KERN_WARNING "BTRFS: "
+				"scrub_repair_page_from_good_copy(bdev == NULL) "
+				"is unexpected!\n");
+			return -EIO;
+		}
+
+		bio = btrfs_io_bio_alloc(GFP_NOFS, 1);
+		if (!bio)
+			return -EIO;
+		bio->bi_bdev = page_bad->dev->bdev;
+		bio->bi_iter.bi_sector = page_bad->physical >> 9;
+
+		ret = bio_add_page(bio, page_good->page, PAGE_SIZE, 0);
+		if (PAGE_SIZE != ret) {
+			bio_put(bio);
+			return -EIO;
+		}
+
+		if (btrfsic_submit_bio_wait(WRITE, bio)) {
+			btrfs_dev_stat_inc_and_print(page_bad->dev,
+				BTRFS_DEV_STAT_WRITE_ERRS);
+			btrfs_dev_replace_stats_inc(
+				&sblock_bad->sctx->dev_root->fs_info->
+				dev_replace.num_write_errors);
+			bio_put(bio);
+			return -EIO;
+		}
+		bio_put(bio);
+	}
+
+	return 0;
+}
+
+static void scrub_write_block_to_dev_replace(struct scrub_block *sblock)
+{
+	int page_num;
+
+	for (page_num = 0; page_num < sblock->page_count; page_num++) {
+		int ret;
+
+		ret = scrub_write_page_to_dev_replace(sblock, page_num);
+		if (ret)
+			btrfs_dev_replace_stats_inc(
+				&sblock->sctx->dev_root->fs_info->dev_replace.
+				num_write_errors);
+	}
+}
+
+static int scrub_write_page_to_dev_replace(struct scrub_block *sblock,
+					   int page_num)
+{
+	struct scrub_page *spage = sblock->pagev[page_num];
+
+	BUG_ON(spage->page == NULL);
+	if (spage->io_error) {
+		void *mapped_buffer = kmap_atomic(spage->page);
+
+		memset(mapped_buffer, 0, PAGE_CACHE_SIZE);
+		flush_dcache_page(spage->page);
+		kunmap_atomic(mapped_buffer);
+	}
+	return scrub_add_page_to_wr_bio(sblock->sctx, spage);
+}
+
+static int scrub_add_page_to_wr_bio(struct scrub_ctx *sctx,
+				    struct scrub_page *spage)
+{
+	struct scrub_wr_ctx *wr_ctx = &sctx->wr_ctx;
+	struct scrub_bio *sbio;
+	int ret;
+
+	mutex_lock(&wr_ctx->wr_lock);
+again:
+	if (!wr_ctx->wr_curr_bio) {
+		wr_ctx->wr_curr_bio = kzalloc(sizeof(*wr_ctx->wr_curr_bio),
+					      GFP_NOFS);
+		if (!wr_ctx->wr_curr_bio) {
+			mutex_unlock(&wr_ctx->wr_lock);
+			return -ENOMEM;
+		}
+		wr_ctx->wr_curr_bio->sctx = sctx;
+		wr_ctx->wr_curr_bio->page_count = 0;
+	}
+	sbio = wr_ctx->wr_curr_bio;
+	if (sbio->page_count == 0) {
+		struct bio *bio;
+
+		sbio->physical = spage->physical_for_dev_replace;
+		sbio->logical = spage->logical;
+		sbio->dev = wr_ctx->tgtdev;
+		bio = sbio->bio;
+		if (!bio) {
+			bio = btrfs_io_bio_alloc(GFP_NOFS, wr_ctx->pages_per_wr_bio);
+			if (!bio) {
+				mutex_unlock(&wr_ctx->wr_lock);
+				return -ENOMEM;
+			}
+			sbio->bio = bio;
+		}
+
+		bio->bi_private = sbio;
+		bio->bi_end_io = scrub_wr_bio_end_io;
+		bio->bi_bdev = sbio->dev->bdev;
+		bio->bi_iter.bi_sector = sbio->physical >> 9;
+		sbio->err = 0;
+	} else if (sbio->physical + sbio->page_count * PAGE_SIZE !=
+		   spage->physical_for_dev_replace ||
+		   sbio->logical + sbio->page_count * PAGE_SIZE !=
+		   spage->logical) {
+		scrub_wr_submit(sctx);
+		goto again;
+	}
+
+	ret = bio_add_page(sbio->bio, spage->page, PAGE_SIZE, 0);
+	if (ret != PAGE_SIZE) {
+		if (sbio->page_count < 1) {
+			bio_put(sbio->bio);
+			sbio->bio = NULL;
+			mutex_unlock(&wr_ctx->wr_lock);
+			return -EIO;
+		}
+		scrub_wr_submit(sctx);
+		goto again;
+	}
+
+	sbio->pagev[sbio->page_count] = spage;
+	scrub_page_get(spage);
+	sbio->page_count++;
+	if (sbio->page_count == wr_ctx->pages_per_wr_bio)
+		scrub_wr_submit(sctx);
+	mutex_unlock(&wr_ctx->wr_lock);
+
+	return 0;
+}
+
+static void scrub_wr_submit(struct scrub_ctx *sctx)
+{
+	struct scrub_wr_ctx *wr_ctx = &sctx->wr_ctx;
+	struct scrub_bio *sbio;
+
+	if (!wr_ctx->wr_curr_bio)
+		return;
+
+	sbio = wr_ctx->wr_curr_bio;
+	wr_ctx->wr_curr_bio = NULL;
+	WARN_ON(!sbio->bio->bi_bdev);
+	scrub_pending_bio_inc(sctx);
+	/* process all writes in a single worker thread. Then the block layer
+	 * orders the requests before sending them to the driver which
+	 * doubled the write performance on spinning disks when measured
+	 * with Linux 3.5 */
+	btrfsic_submit_bio(WRITE, sbio->bio);
+}
+
+static void scrub_wr_bio_end_io(struct bio *bio, int err)
+{
+	struct scrub_bio *sbio = bio->bi_private;
+	struct btrfs_fs_info *fs_info = sbio->dev->dev_root->fs_info;
+
+	sbio->err = err;
+	sbio->bio = bio;
+
+	btrfs_init_work(&sbio->work, scrub_wr_bio_end_io_worker, NULL, NULL);
+	btrfs_queue_work(fs_info->scrub_wr_completion_workers, &sbio->work);
+}
+
+static void scrub_wr_bio_end_io_worker(struct btrfs_work *work)
+{
+	struct scrub_bio *sbio = container_of(work, struct scrub_bio, work);
+	struct scrub_ctx *sctx = sbio->sctx;
+	int i;
+
+	WARN_ON(sbio->page_count > SCRUB_PAGES_PER_WR_BIO);
+	if (sbio->err) {
+		struct btrfs_dev_replace *dev_replace =
+			&sbio->sctx->dev_root->fs_info->dev_replace;
+
+		for (i = 0; i < sbio->page_count; i++) {
+			struct scrub_page *spage = sbio->pagev[i];
+
+			spage->io_error = 1;
+			btrfs_dev_replace_stats_inc(&dev_replace->
+						    num_write_errors);
+		}
+	}
+
+	for (i = 0; i < sbio->page_count; i++)
+		scrub_page_put(sbio->pagev[i]);
+
+	bio_put(sbio->bio);
+	kfree(sbio);
+	scrub_pending_bio_dec(sctx);
+}
+
+static int scrub_checksum(struct scrub_block *sblock)
+{
+	u64 flags;
+	int ret;
+
+	WARN_ON(sblock->page_count < 1);
+	flags = sblock->pagev[0]->flags;
+	ret = 0;
+	if (flags & BTRFS_EXTENT_FLAG_DATA)
+		ret = scrub_checksum_data(sblock);
+	else if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK)
+		ret = scrub_checksum_tree_block(sblock);
+	else if (flags & BTRFS_EXTENT_FLAG_SUPER)
+		(void)scrub_checksum_super(sblock);
+	else
+		WARN_ON(1);
+	if (ret)
+		scrub_handle_errored_block(sblock);
+
+	return ret;
+}
+
+static int scrub_checksum_data(struct scrub_block *sblock)
+{
+	struct scrub_ctx *sctx = sblock->sctx;
+	u8 csum[BTRFS_CSUM_SIZE];
+	u8 *on_disk_csum;
+	struct page *page;
+	void *buffer;
+	u32 crc = ~(u32)0;
+	int fail = 0;
+	u64 len;
+	int index;
+
+	BUG_ON(sblock->page_count < 1);
+	if (!sblock->pagev[0]->have_csum)
+		return 0;
+
+	on_disk_csum = sblock->pagev[0]->csum;
+	page = sblock->pagev[0]->page;
+	buffer = kmap_atomic(page);
+
+	len = sctx->sectorsize;
+	index = 0;
+	for (;;) {
+		u64 l = min_t(u64, len, PAGE_SIZE);
+
+		crc = btrfs_csum_data(buffer, crc, l);
+		kunmap_atomic(buffer);
+		len -= l;
+		if (len == 0)
+			break;
+		index++;
+		BUG_ON(index >= sblock->page_count);
+		BUG_ON(!sblock->pagev[index]->page);
+		page = sblock->pagev[index]->page;
+		buffer = kmap_atomic(page);
+	}
+
+	btrfs_csum_final(crc, csum);
+	if (memcmp(csum, on_disk_csum, sctx->csum_size))
+		fail = 1;
+
+	return fail;
+}
+
+static int scrub_checksum_tree_block(struct scrub_block *sblock)
+{
+	struct scrub_ctx *sctx = sblock->sctx;
+	struct btrfs_header *h;
+	struct btrfs_root *root = sctx->dev_root;
+	struct btrfs_fs_info *fs_info = root->fs_info;
+	u8 calculated_csum[BTRFS_CSUM_SIZE];
+	u8 on_disk_csum[BTRFS_CSUM_SIZE];
+	struct page *page;
+	void *mapped_buffer;
+	u64 mapped_size;
+	void *p;
+	u32 crc = ~(u32)0;
+	int fail = 0;
+	int crc_fail = 0;
+	u64 len;
+	int index;
+
+	BUG_ON(sblock->page_count < 1);
+	page = sblock->pagev[0]->page;
+	mapped_buffer = kmap_atomic(page);
+	h = (struct btrfs_header *)mapped_buffer;
+	memcpy(on_disk_csum, h->csum, sctx->csum_size);
+
+	/*
+	 * we don't use the getter functions here, as we
+	 * a) don't have an extent buffer and
+	 * b) the page is already kmapped
+	 */
+
+	if (sblock->pagev[0]->logical != btrfs_stack_header_bytenr(h))
+		++fail;
+
+	if (sblock->pagev[0]->generation != btrfs_stack_header_generation(h))
+		++fail;
+
+	if (memcmp(h->fsid, fs_info->fsid, BTRFS_UUID_SIZE))
+		++fail;
+
+	if (memcmp(h->chunk_tree_uuid, fs_info->chunk_tree_uuid,
+		   BTRFS_UUID_SIZE))
+		++fail;
+
+	WARN_ON(sctx->nodesize != sctx->leafsize);
+	len = sctx->nodesize - BTRFS_CSUM_SIZE;
+	mapped_size = PAGE_SIZE - BTRFS_CSUM_SIZE;
+	p = ((u8 *)mapped_buffer) + BTRFS_CSUM_SIZE;
+	index = 0;
+	for (;;) {
+		u64 l = min_t(u64, len, mapped_size);
+
+		crc = btrfs_csum_data(p, crc, l);
+		kunmap_atomic(mapped_buffer);
+		len -= l;
+		if (len == 0)
+			break;
+		index++;
+		BUG_ON(index >= sblock->page_count);
+		BUG_ON(!sblock->pagev[index]->page);
+		page = sblock->pagev[index]->page;
+		mapped_buffer = kmap_atomic(page);
+		mapped_size = PAGE_SIZE;
+		p = mapped_buffer;
+	}
+
+	btrfs_csum_final(crc, calculated_csum);
+	if (memcmp(calculated_csum, on_disk_csum, sctx->csum_size))
+		++crc_fail;
+
+	return fail || crc_fail;
+}
+
+static int scrub_checksum_super(struct scrub_block *sblock)
+{
+	struct btrfs_super_block *s;
+	struct scrub_ctx *sctx = sblock->sctx;
+	struct btrfs_root *root = sctx->dev_root;
+	struct btrfs_fs_info *fs_info = root->fs_info;
+	u8 calculated_csum[BTRFS_CSUM_SIZE];
+	u8 on_disk_csum[BTRFS_CSUM_SIZE];
+	struct page *page;
+	void *mapped_buffer;
+	u64 mapped_size;
+	void *p;
+	u32 crc = ~(u32)0;
+	int fail_gen = 0;
+	int fail_cor = 0;
+	u64 len;
+	int index;
+
+	BUG_ON(sblock->page_count < 1);
+	page = sblock->pagev[0]->page;
+	mapped_buffer = kmap_atomic(page);
+	s = (struct btrfs_super_block *)mapped_buffer;
+	memcpy(on_disk_csum, s->csum, sctx->csum_size);
+
+	if (sblock->pagev[0]->logical != btrfs_super_bytenr(s))
+		++fail_cor;
+
+	if (sblock->pagev[0]->generation != btrfs_super_generation(s))
+		++fail_gen;
+
+	if (memcmp(s->fsid, fs_info->fsid, BTRFS_UUID_SIZE))
+		++fail_cor;
+
+	len = BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE;
+	mapped_size = PAGE_SIZE - BTRFS_CSUM_SIZE;
+	p = ((u8 *)mapped_buffer) + BTRFS_CSUM_SIZE;
+	index = 0;
+	for (;;) {
+		u64 l = min_t(u64, len, mapped_size);
+
+		crc = btrfs_csum_data(p, crc, l);
+		kunmap_atomic(mapped_buffer);
+		len -= l;
+		if (len == 0)
+			break;
+		index++;
+		BUG_ON(index >= sblock->page_count);
+		BUG_ON(!sblock->pagev[index]->page);
+		page = sblock->pagev[index]->page;
+		mapped_buffer = kmap_atomic(page);
+		mapped_size = PAGE_SIZE;
+		p = mapped_buffer;
+	}
+
+	btrfs_csum_final(crc, calculated_csum);
+	if (memcmp(calculated_csum, on_disk_csum, sctx->csum_size))
+		++fail_cor;
+
+	if (fail_cor + fail_gen) {
+		/*
+		 * if we find an error in a super block, we just report it.
+		 * They will get written with the next transaction commit
+		 * anyway
+		 */
+		spin_lock(&sctx->stat_lock);
+		++sctx->stat.super_errors;
+		spin_unlock(&sctx->stat_lock);
+		if (fail_cor)
+			btrfs_dev_stat_inc_and_print(sblock->pagev[0]->dev,
+				BTRFS_DEV_STAT_CORRUPTION_ERRS);
+		else
+			btrfs_dev_stat_inc_and_print(sblock->pagev[0]->dev,
+				BTRFS_DEV_STAT_GENERATION_ERRS);
+	}
+
+	return fail_cor + fail_gen;
+}
+
+static void scrub_block_get(struct scrub_block *sblock)
+{
+	atomic_inc(&sblock->ref_count);
+}
+
+static void scrub_block_put(struct scrub_block *sblock)
+{
+	if (atomic_dec_and_test(&sblock->ref_count)) {
+		int i;
+
+		for (i = 0; i < sblock->page_count; i++)
+			scrub_page_put(sblock->pagev[i]);
+		kfree(sblock);
+	}
+}
+
+static void scrub_page_get(struct scrub_page *spage)
+{
+	atomic_inc(&spage->ref_count);
+}
+
+static void scrub_page_put(struct scrub_page *spage)
+{
+	if (atomic_dec_and_test(&spage->ref_count)) {
+		if (spage->page)
+			__free_page(spage->page);
+		kfree(spage);
+	}
+}
+
+static void scrub_submit(struct scrub_ctx *sctx)
+{
+	struct scrub_bio *sbio;
+
+	if (sctx->curr == -1)
+		return;
+
+	sbio = sctx->bios[sctx->curr];
+	sctx->curr = -1;
+	scrub_pending_bio_inc(sctx);
+
+	if (!sbio->bio->bi_bdev) {
+		/*
+		 * this case should not happen. If btrfs_map_block() is
+		 * wrong, it could happen for dev-replace operations on
+		 * missing devices when no mirrors are available, but in
+		 * this case it should already fail the mount.
+		 * This case is handled correctly (but _very_ slowly).
+		 */
+		printk_ratelimited(KERN_WARNING
+			"BTRFS: scrub_submit(bio bdev == NULL) is unexpected!\n");
+		bio_endio(sbio->bio, -EIO);
+	} else {
+		btrfsic_submit_bio(READ, sbio->bio);
+	}
+}
+
+static int scrub_add_page_to_rd_bio(struct scrub_ctx *sctx,
+				    struct scrub_page *spage)
+{
+	struct scrub_block *sblock = spage->sblock;
+	struct scrub_bio *sbio;
+	int ret;
+
+again:
+	/*
+	 * grab a fresh bio or wait for one to become available
+	 */
+	while (sctx->curr == -1) {
+		spin_lock(&sctx->list_lock);
+		sctx->curr = sctx->first_free;
+		if (sctx->curr != -1) {
+			sctx->first_free = sctx->bios[sctx->curr]->next_free;
+			sctx->bios[sctx->curr]->next_free = -1;
+			sctx->bios[sctx->curr]->page_count = 0;
+			spin_unlock(&sctx->list_lock);
+		} else {
+			spin_unlock(&sctx->list_lock);
+			wait_event(sctx->list_wait, sctx->first_free != -1);
+		}
+	}
+	sbio = sctx->bios[sctx->curr];
+	if (sbio->page_count == 0) {
+		struct bio *bio;
+
+		sbio->physical = spage->physical;
+		sbio->logical = spage->logical;
+		sbio->dev = spage->dev;
+		bio = sbio->bio;
+		if (!bio) {
+			bio = btrfs_io_bio_alloc(GFP_NOFS, sctx->pages_per_rd_bio);
+			if (!bio)
+				return -ENOMEM;
+			sbio->bio = bio;
+		}
+
+		bio->bi_private = sbio;
+		bio->bi_end_io = scrub_bio_end_io;
+		bio->bi_bdev = sbio->dev->bdev;
+		bio->bi_iter.bi_sector = sbio->physical >> 9;
+		sbio->err = 0;
+	} else if (sbio->physical + sbio->page_count * PAGE_SIZE !=
+		   spage->physical ||
+		   sbio->logical + sbio->page_count * PAGE_SIZE !=
+		   spage->logical ||
+		   sbio->dev != spage->dev) {
+		scrub_submit(sctx);
+		goto again;
+	}
+
+	sbio->pagev[sbio->page_count] = spage;
+	ret = bio_add_page(sbio->bio, spage->page, PAGE_SIZE, 0);
+	if (ret != PAGE_SIZE) {
+		if (sbio->page_count < 1) {
+			bio_put(sbio->bio);
+			sbio->bio = NULL;
+			return -EIO;
+		}
+		scrub_submit(sctx);
+		goto again;
+	}
+
+	scrub_block_get(sblock); /* one for the page added to the bio */
+	atomic_inc(&sblock->outstanding_pages);
+	sbio->page_count++;
+	if (sbio->page_count == sctx->pages_per_rd_bio)
+		scrub_submit(sctx);
+
+	return 0;
+}
+
+static int scrub_pages(struct scrub_ctx *sctx, u64 logical, u64 len,
+		       u64 physical, struct btrfs_device *dev, u64 flags,
+		       u64 gen, int mirror_num, u8 *csum, int force,
+		       u64 physical_for_dev_replace)
+{
+	struct scrub_block *sblock;
+	int index;
+
+	sblock = kzalloc(sizeof(*sblock), GFP_NOFS);
+	if (!sblock) {
+		spin_lock(&sctx->stat_lock);
+		sctx->stat.malloc_errors++;
+		spin_unlock(&sctx->stat_lock);
+		return -ENOMEM;
+	}
+
+	/* one ref inside this function, plus one for each page added to
+	 * a bio later on */
+	atomic_set(&sblock->ref_count, 1);
+	sblock->sctx = sctx;
+	sblock->no_io_error_seen = 1;
+
+	for (index = 0; len > 0; index++) {
+		struct scrub_page *spage;
+		u64 l = min_t(u64, len, PAGE_SIZE);
+
+		spage = kzalloc(sizeof(*spage), GFP_NOFS);
+		if (!spage) {
+leave_nomem:
+			spin_lock(&sctx->stat_lock);
+			sctx->stat.malloc_errors++;
+			spin_unlock(&sctx->stat_lock);
+			scrub_block_put(sblock);
+			return -ENOMEM;
+		}
+		BUG_ON(index >= SCRUB_MAX_PAGES_PER_BLOCK);
+		scrub_page_get(spage);
+		sblock->pagev[index] = spage;
+		spage->sblock = sblock;
+		spage->dev = dev;
+		spage->flags = flags;
+		spage->generation = gen;
+		spage->logical = logical;
+		spage->physical = physical;
+		spage->physical_for_dev_replace = physical_for_dev_replace;
+		spage->mirror_num = mirror_num;
+		if (csum) {
+			spage->have_csum = 1;
+			memcpy(spage->csum, csum, sctx->csum_size);
+		} else {
+			spage->have_csum = 0;
+		}
+		sblock->page_count++;
+		spage->page = alloc_page(GFP_NOFS);
+		if (!spage->page)
+			goto leave_nomem;
+		len -= l;
+		logical += l;
+		physical += l;
+		physical_for_dev_replace += l;
+	}
+
+	WARN_ON(sblock->page_count == 0);
+	for (index = 0; index < sblock->page_count; index++) {
+		struct scrub_page *spage = sblock->pagev[index];
+		int ret;
+
+		ret = scrub_add_page_to_rd_bio(sctx, spage);
+		if (ret) {
+			scrub_block_put(sblock);
+			return ret;
+		}
+	}
+
+	if (force)
+		scrub_submit(sctx);
+
+	/* last one frees, either here or in bio completion for last page */
+	scrub_block_put(sblock);
+	return 0;
+}
+
+static void scrub_bio_end_io(struct bio *bio, int err)
+{
+	struct scrub_bio *sbio = bio->bi_private;
+	struct btrfs_fs_info *fs_info = sbio->dev->dev_root->fs_info;
+
+	sbio->err = err;
+	sbio->bio = bio;
+
+	btrfs_queue_work(fs_info->scrub_workers, &sbio->work);
+}
+
+static void scrub_bio_end_io_worker(struct btrfs_work *work)
+{
+	struct scrub_bio *sbio = container_of(work, struct scrub_bio, work);
+	struct scrub_ctx *sctx = sbio->sctx;
+	int i;
+
+	BUG_ON(sbio->page_count > SCRUB_PAGES_PER_RD_BIO);
+	if (sbio->err) {
+		for (i = 0; i < sbio->page_count; i++) {
+			struct scrub_page *spage = sbio->pagev[i];
+
+			spage->io_error = 1;
+			spage->sblock->no_io_error_seen = 0;
+		}
+	}
+
+	/* now complete the scrub_block items that have all pages completed */
+	for (i = 0; i < sbio->page_count; i++) {
+		struct scrub_page *spage = sbio->pagev[i];
+		struct scrub_block *sblock = spage->sblock;
+
+		if (atomic_dec_and_test(&sblock->outstanding_pages))
+			scrub_block_complete(sblock);
+		scrub_block_put(sblock);
+	}
+
+	bio_put(sbio->bio);
+	sbio->bio = NULL;
+	spin_lock(&sctx->list_lock);
+	sbio->next_free = sctx->first_free;
+	sctx->first_free = sbio->index;
+	spin_unlock(&sctx->list_lock);
+
+	if (sctx->is_dev_replace &&
+	    atomic_read(&sctx->wr_ctx.flush_all_writes)) {
+		mutex_lock(&sctx->wr_ctx.wr_lock);
+		scrub_wr_submit(sctx);
+		mutex_unlock(&sctx->wr_ctx.wr_lock);
+	}
+
+	scrub_pending_bio_dec(sctx);
+}
+
+static void scrub_block_complete(struct scrub_block *sblock)
+{
+	if (!sblock->no_io_error_seen) {
+		scrub_handle_errored_block(sblock);
+	} else {
+		/*
+		 * if has checksum error, write via repair mechanism in
+		 * dev replace case, otherwise write here in dev replace
+		 * case.
+		 */
+		if (!scrub_checksum(sblock) && sblock->sctx->is_dev_replace)
+			scrub_write_block_to_dev_replace(sblock);
+	}
+}
+
+static int scrub_find_csum(struct scrub_ctx *sctx, u64 logical, u64 len,
+			   u8 *csum)
+{
+	struct btrfs_ordered_sum *sum = NULL;
+	unsigned long index;
+	unsigned long num_sectors;
+
+	while (!list_empty(&sctx->csum_list)) {
+		sum = list_first_entry(&sctx->csum_list,
+				       struct btrfs_ordered_sum, list);
+		if (sum->bytenr > logical)
+			return 0;
+		if (sum->bytenr + sum->len > logical)
+			break;
+
+		++sctx->stat.csum_discards;
+		list_del(&sum->list);
+		kfree(sum);
+		sum = NULL;
+	}
+	if (!sum)
+		return 0;
+
+	index = ((u32)(logical - sum->bytenr)) / sctx->sectorsize;
+	num_sectors = sum->len / sctx->sectorsize;
+	memcpy(csum, sum->sums + index, sctx->csum_size);
+	if (index == num_sectors - 1) {
+		list_del(&sum->list);
+		kfree(sum);
+	}
+	return 1;
+}
+
+/* scrub extent tries to collect up to 64 kB for each bio */
+static int scrub_extent(struct scrub_ctx *sctx, u64 logical, u64 len,
+			u64 physical, struct btrfs_device *dev, u64 flags,
+			u64 gen, int mirror_num, u64 physical_for_dev_replace)
+{
+	int ret;
+	u8 csum[BTRFS_CSUM_SIZE];
+	u32 blocksize;
+
+	if (flags & BTRFS_EXTENT_FLAG_DATA) {
+		blocksize = sctx->sectorsize;
+		spin_lock(&sctx->stat_lock);
+		sctx->stat.data_extents_scrubbed++;
+		sctx->stat.data_bytes_scrubbed += len;
+		spin_unlock(&sctx->stat_lock);
+	} else if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
+		WARN_ON(sctx->nodesize != sctx->leafsize);
+		blocksize = sctx->nodesize;
+		spin_lock(&sctx->stat_lock);
+		sctx->stat.tree_extents_scrubbed++;
+		sctx->stat.tree_bytes_scrubbed += len;
+		spin_unlock(&sctx->stat_lock);
+	} else {
+		blocksize = sctx->sectorsize;
+		WARN_ON(1);
+	}
+
+	while (len) {
+		u64 l = min_t(u64, len, blocksize);
+		int have_csum = 0;
+
+		if (flags & BTRFS_EXTENT_FLAG_DATA) {
+			/* push csums to sbio */
+			have_csum = scrub_find_csum(sctx, logical, l, csum);
+			if (have_csum == 0)
+				++sctx->stat.no_csum;
+			if (sctx->is_dev_replace && !have_csum) {
+				ret = copy_nocow_pages(sctx, logical, l,
+						       mirror_num,
+						      physical_for_dev_replace);
+				goto behind_scrub_pages;
+			}
+		}
+		ret = scrub_pages(sctx, logical, l, physical, dev, flags, gen,
+				  mirror_num, have_csum ? csum : NULL, 0,
+				  physical_for_dev_replace);
+behind_scrub_pages:
+		if (ret)
+			return ret;
+		len -= l;
+		logical += l;
+		physical += l;
+		physical_for_dev_replace += l;
+	}
+	return 0;
+}
+
+/*
+ * Given a physical address, this will calculate it's
+ * logical offset. if this is a parity stripe, it will return
+ * the most left data stripe's logical offset.
+ *
+ * return 0 if it is a data stripe, 1 means parity stripe.
+ */
+static int get_raid56_logic_offset(u64 physical, int num,
+				   struct map_lookup *map, u64 *offset)
+{
+	int i;
+	int j = 0;
+	u64 stripe_nr;
+	u64 last_offset;
+	int stripe_index;
+	int rot;
+
+	last_offset = (physical - map->stripes[num].physical) *
+		      nr_data_stripes(map);
+	*offset = last_offset;
+	for (i = 0; i < nr_data_stripes(map); i++) {
+		*offset = last_offset + i * map->stripe_len;
+
+		stripe_nr = *offset;
+		do_div(stripe_nr, map->stripe_len);
+		do_div(stripe_nr, nr_data_stripes(map));
+
+		/* Work out the disk rotation on this stripe-set */
+		rot = do_div(stripe_nr, map->num_stripes);
+		/* calculate which stripe this data locates */
+		rot += i;
+		stripe_index = rot % map->num_stripes;
+		if (stripe_index == num)
+			return 0;
+		if (stripe_index < num)
+			j++;
+	}
+	*offset = last_offset + j * map->stripe_len;
+	return 1;
+}
+
+static noinline_for_stack int scrub_stripe(struct scrub_ctx *sctx,
+					   struct map_lookup *map,
+					   struct btrfs_device *scrub_dev,
+					   int num, u64 base, u64 length,
+					   int is_dev_replace)
+{
+	struct btrfs_path *path;
+	struct btrfs_fs_info *fs_info = sctx->dev_root->fs_info;
+	struct btrfs_root *root = fs_info->extent_root;
+	struct btrfs_root *csum_root = fs_info->csum_root;
+	struct btrfs_extent_item *extent;
+	struct blk_plug plug;
+	u64 flags;
+	int ret;
+	int slot;
+	u64 nstripes;
+	struct extent_buffer *l;
+	struct btrfs_key key;
+	u64 physical;
+	u64 logical;
+	u64 logic_end;
+	u64 physical_end;
+	u64 generation;
+	int mirror_num;
+	struct reada_control *reada1;
+	struct reada_control *reada2;
+	struct btrfs_key key_start;
+	struct btrfs_key key_end;
+	u64 increment = map->stripe_len;
+	u64 offset;
+	u64 extent_logical;
+	u64 extent_physical;
+	u64 extent_len;
+	struct btrfs_device *extent_dev;
+	int extent_mirror_num;
+	int stop_loop = 0;
+
+	nstripes = length;
+	physical = map->stripes[num].physical;
+	offset = 0;
+	do_div(nstripes, map->stripe_len);
+	if (map->type & BTRFS_BLOCK_GROUP_RAID0) {
+		offset = map->stripe_len * num;
+		increment = map->stripe_len * map->num_stripes;
+		mirror_num = 1;
+	} else if (map->type & BTRFS_BLOCK_GROUP_RAID10) {
+		int factor = map->num_stripes / map->sub_stripes;
+		offset = map->stripe_len * (num / map->sub_stripes);
+		increment = map->stripe_len * factor;
+		mirror_num = num % map->sub_stripes + 1;
+	} else if (map->type & BTRFS_BLOCK_GROUP_RAID1) {
+		increment = map->stripe_len;
+		mirror_num = num % map->num_stripes + 1;
+	} else if (map->type & BTRFS_BLOCK_GROUP_DUP) {
+		increment = map->stripe_len;
+		mirror_num = num % map->num_stripes + 1;
+	} else if (map->type & (BTRFS_BLOCK_GROUP_RAID5 |
+				BTRFS_BLOCK_GROUP_RAID6)) {
+		get_raid56_logic_offset(physical, num, map, &offset);
+		increment = map->stripe_len * nr_data_stripes(map);
+		mirror_num = 1;
+	} else {
+		increment = map->stripe_len;
+		mirror_num = 1;
+	}
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	/*
+	 * work on commit root. The related disk blocks are static as
+	 * long as COW is applied. This means, it is save to rewrite
+	 * them to repair disk errors without any race conditions
+	 */
+	path->search_commit_root = 1;
+	path->skip_locking = 1;
+
+	/*
+	 * trigger the readahead for extent tree csum tree and wait for
+	 * completion. During readahead, the scrub is officially paused
+	 * to not hold off transaction commits
+	 */
+	logical = base + offset;
+	physical_end = physical + nstripes * map->stripe_len;
+	if (map->type & (BTRFS_BLOCK_GROUP_RAID5 |
+			 BTRFS_BLOCK_GROUP_RAID6)) {
+		get_raid56_logic_offset(physical_end, num,
+					map, &logic_end);
+		logic_end += base;
+	} else {
+		logic_end = logical + increment * nstripes;
+	}
+	wait_event(sctx->list_wait,
+		   atomic_read(&sctx->bios_in_flight) == 0);
+	scrub_blocked_if_needed(fs_info);
+
+	/* FIXME it might be better to start readahead at commit root */
+	key_start.objectid = logical;
+	key_start.type = BTRFS_EXTENT_ITEM_KEY;
+	key_start.offset = (u64)0;
+	key_end.objectid = logic_end;
+	key_end.type = BTRFS_METADATA_ITEM_KEY;
+	key_end.offset = (u64)-1;
+	reada1 = btrfs_reada_add(root, &key_start, &key_end);
+
+	key_start.objectid = BTRFS_EXTENT_CSUM_OBJECTID;
+	key_start.type = BTRFS_EXTENT_CSUM_KEY;
+	key_start.offset = logical;
+	key_end.objectid = BTRFS_EXTENT_CSUM_OBJECTID;
+	key_end.type = BTRFS_EXTENT_CSUM_KEY;
+	key_end.offset = logic_end;
+	reada2 = btrfs_reada_add(csum_root, &key_start, &key_end);
+
+	if (!IS_ERR(reada1))
+		btrfs_reada_wait(reada1);
+	if (!IS_ERR(reada2))
+		btrfs_reada_wait(reada2);
+
+
+	/*
+	 * collect all data csums for the stripe to avoid seeking during
+	 * the scrub. This might currently (crc32) end up to be about 1MB
+	 */
+	blk_start_plug(&plug);
+
+	/*
+	 * now find all extents for each stripe and scrub them
+	 */
+	ret = 0;
+	while (physical < physical_end) {
+		/* for raid56, we skip parity stripe */
+		if (map->type & (BTRFS_BLOCK_GROUP_RAID5 |
+				BTRFS_BLOCK_GROUP_RAID6)) {
+			ret = get_raid56_logic_offset(physical, num,
+					map, &logical);
+			logical += base;
+			if (ret)
+				goto skip;
+		}
+		/*
+		 * canceled?
+		 */
+		if (atomic_read(&fs_info->scrub_cancel_req) ||
+		    atomic_read(&sctx->cancel_req)) {
+			ret = -ECANCELED;
+			goto out;
+		}
+		/*
+		 * check to see if we have to pause
+		 */
+		if (atomic_read(&fs_info->scrub_pause_req)) {
+			/* push queued extents */
+			atomic_set(&sctx->wr_ctx.flush_all_writes, 1);
+			scrub_submit(sctx);
+			mutex_lock(&sctx->wr_ctx.wr_lock);
+			scrub_wr_submit(sctx);
+			mutex_unlock(&sctx->wr_ctx.wr_lock);
+			wait_event(sctx->list_wait,
+				   atomic_read(&sctx->bios_in_flight) == 0);
+			atomic_set(&sctx->wr_ctx.flush_all_writes, 0);
+			scrub_blocked_if_needed(fs_info);
+		}
+
+		if (btrfs_fs_incompat(fs_info, SKINNY_METADATA))
+			key.type = BTRFS_METADATA_ITEM_KEY;
+		else
+			key.type = BTRFS_EXTENT_ITEM_KEY;
+		key.objectid = logical;
+		key.offset = (u64)-1;
+
+		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+		if (ret < 0)
+			goto out;
+
+		if (ret > 0) {
+			ret = btrfs_previous_extent_item(root, path, 0);
+			if (ret < 0)
+				goto out;
+			if (ret > 0) {
+				/* there's no smaller item, so stick with the
+				 * larger one */
+				btrfs_release_path(path);
+				ret = btrfs_search_slot(NULL, root, &key,
+							path, 0, 0);
+				if (ret < 0)
+					goto out;
+			}
+		}
+
+		stop_loop = 0;
+		while (1) {
+			u64 bytes;
+
+			l = path->nodes[0];
+			slot = path->slots[0];
+			if (slot >= btrfs_header_nritems(l)) {
+				ret = btrfs_next_leaf(root, path);
+				if (ret == 0)
+					continue;
+				if (ret < 0)
+					goto out;
+
+				stop_loop = 1;
+				break;
+			}
+			btrfs_item_key_to_cpu(l, &key, slot);
+
+			if (key.type == BTRFS_METADATA_ITEM_KEY)
+				bytes = root->leafsize;
+			else
+				bytes = key.offset;
+
+			if (key.objectid + bytes <= logical)
+				goto next;
+
+			if (key.type != BTRFS_EXTENT_ITEM_KEY &&
+			    key.type != BTRFS_METADATA_ITEM_KEY)
+				goto next;
+
+			if (key.objectid >= logical + map->stripe_len) {
+				/* out of this device extent */
+				if (key.objectid >= logic_end)
+					stop_loop = 1;
+				break;
+			}
+
+			extent = btrfs_item_ptr(l, slot,
+						struct btrfs_extent_item);
+			flags = btrfs_extent_flags(l, extent);
+			generation = btrfs_extent_generation(l, extent);
+
+			if (key.objectid < logical &&
+			    (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK)) {
+				btrfs_err(fs_info,
+					   "scrub: tree block %llu spanning "
+					   "stripes, ignored. logical=%llu",
+				       key.objectid, logical);
+				goto next;
+			}
+
+again:
+			extent_logical = key.objectid;
+			extent_len = bytes;
+
+			/*
+			 * trim extent to this stripe
+			 */
+			if (extent_logical < logical) {
+				extent_len -= logical - extent_logical;
+				extent_logical = logical;
+			}
+			if (extent_logical + extent_len >
+			    logical + map->stripe_len) {
+				extent_len = logical + map->stripe_len -
+					     extent_logical;
+			}
+
+			extent_physical = extent_logical - logical + physical;
+			extent_dev = scrub_dev;
+			extent_mirror_num = mirror_num;
+			if (is_dev_replace)
+				scrub_remap_extent(fs_info, extent_logical,
+						   extent_len, &extent_physical,
+						   &extent_dev,
+						   &extent_mirror_num);
+
+			ret = btrfs_lookup_csums_range(csum_root, logical,
+						logical + map->stripe_len - 1,
+						&sctx->csum_list, 1);
+			if (ret)
+				goto out;
+
+			ret = scrub_extent(sctx, extent_logical, extent_len,
+					   extent_physical, extent_dev, flags,
+					   generation, extent_mirror_num,
+					   extent_logical - logical + physical);
+			if (ret)
+				goto out;
+
+			scrub_free_csums(sctx);
+			if (extent_logical + extent_len <
+			    key.objectid + bytes) {
+				if (map->type & (BTRFS_BLOCK_GROUP_RAID5 |
+					BTRFS_BLOCK_GROUP_RAID6)) {
+					/*
+					 * loop until we find next data stripe
+					 * or we have finished all stripes.
+					 */
+					do {
+						physical += map->stripe_len;
+						ret = get_raid56_logic_offset(
+								physical, num,
+								map, &logical);
+						logical += base;
+					} while (physical < physical_end && ret);
+				} else {
+					physical += map->stripe_len;
+					logical += increment;
+				}
+				if (logical < key.objectid + bytes) {
+					cond_resched();
+					goto again;
+				}
+
+				if (physical >= physical_end) {
+					stop_loop = 1;
+					break;
+				}
+			}
+next:
+			path->slots[0]++;
+		}
+		btrfs_release_path(path);
+skip:
+		logical += increment;
+		physical += map->stripe_len;
+		spin_lock(&sctx->stat_lock);
+		if (stop_loop)
+			sctx->stat.last_physical = map->stripes[num].physical +
+						   length;
+		else
+			sctx->stat.last_physical = physical;
+		spin_unlock(&sctx->stat_lock);
+		if (stop_loop)
+			break;
+	}
+out:
+	/* push queued extents */
+	scrub_submit(sctx);
+	mutex_lock(&sctx->wr_ctx.wr_lock);
+	scrub_wr_submit(sctx);
+	mutex_unlock(&sctx->wr_ctx.wr_lock);
+
+	blk_finish_plug(&plug);
+	btrfs_free_path(path);
+	return ret < 0 ? ret : 0;
+}
+
+static noinline_for_stack int scrub_chunk(struct scrub_ctx *sctx,
+					  struct btrfs_device *scrub_dev,
+					  u64 chunk_tree, u64 chunk_objectid,
+					  u64 chunk_offset, u64 length,
+					  u64 dev_offset, int is_dev_replace)
+{
+	struct btrfs_mapping_tree *map_tree =
+		&sctx->dev_root->fs_info->mapping_tree;
+	struct map_lookup *map;
+	struct extent_map *em;
+	int i;
+	int ret = 0;
+
+	read_lock(&map_tree->map_tree.lock);
+	em = lookup_extent_mapping(&map_tree->map_tree, chunk_offset, 1);
+	read_unlock(&map_tree->map_tree.lock);
+
+	if (!em)
+		return -EINVAL;
+
+	map = (struct map_lookup *)em->bdev;
+	if (em->start != chunk_offset)
+		goto out;
+
+	if (em->len < length)
+		goto out;
+
+	for (i = 0; i < map->num_stripes; ++i) {
+		if (map->stripes[i].dev->bdev == scrub_dev->bdev &&
+		    map->stripes[i].physical == dev_offset) {
+			ret = scrub_stripe(sctx, map, scrub_dev, i,
+					   chunk_offset, length,
+					   is_dev_replace);
+			if (ret)
+				goto out;
+		}
+	}
+out:
+	free_extent_map(em);
+
+	return ret;
+}
+
+static noinline_for_stack
+int scrub_enumerate_chunks(struct scrub_ctx *sctx,
+			   struct btrfs_device *scrub_dev, u64 start, u64 end,
+			   int is_dev_replace)
+{
+	struct btrfs_dev_extent *dev_extent = NULL;
+	struct btrfs_path *path;
+	struct btrfs_root *root = sctx->dev_root;
+	struct btrfs_fs_info *fs_info = root->fs_info;
+	u64 length;
+	u64 chunk_tree;
+	u64 chunk_objectid;
+	u64 chunk_offset;
+	int ret;
+	int slot;
+	struct extent_buffer *l;
+	struct btrfs_key key;
+	struct btrfs_key found_key;
+	struct btrfs_block_group_cache *cache;
+	struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	path->reada = 2;
+	path->search_commit_root = 1;
+	path->skip_locking = 1;
+
+	key.objectid = scrub_dev->devid;
+	key.offset = 0ull;
+	key.type = BTRFS_DEV_EXTENT_KEY;
+
+	while (1) {
+		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+		if (ret < 0)
+			break;
+		if (ret > 0) {
+			if (path->slots[0] >=
+			    btrfs_header_nritems(path->nodes[0])) {
+				ret = btrfs_next_leaf(root, path);
+				if (ret)
+					break;
+			}
+		}
+
+		l = path->nodes[0];
+		slot = path->slots[0];
+
+		btrfs_item_key_to_cpu(l, &found_key, slot);
+
+		if (found_key.objectid != scrub_dev->devid)
+			break;
+
+		if (btrfs_key_type(&found_key) != BTRFS_DEV_EXTENT_KEY)
+			break;
+
+		if (found_key.offset >= end)
+			break;
+
+		if (found_key.offset < key.offset)
+			break;
+
+		dev_extent = btrfs_item_ptr(l, slot, struct btrfs_dev_extent);
+		length = btrfs_dev_extent_length(l, dev_extent);
+
+		if (found_key.offset + length <= start)
+			goto skip;
+
+		chunk_tree = btrfs_dev_extent_chunk_tree(l, dev_extent);
+		chunk_objectid = btrfs_dev_extent_chunk_objectid(l, dev_extent);
+		chunk_offset = btrfs_dev_extent_chunk_offset(l, dev_extent);
+
+		/*
+		 * get a reference on the corresponding block group to prevent
+		 * the chunk from going away while we scrub it
+		 */
+		cache = btrfs_lookup_block_group(fs_info, chunk_offset);
+
+		/* some chunks are removed but not committed to disk yet,
+		 * continue scrubbing */
+		if (!cache)
+			goto skip;
+
+		dev_replace->cursor_right = found_key.offset + length;
+		dev_replace->cursor_left = found_key.offset;
+		dev_replace->item_needs_writeback = 1;
+		ret = scrub_chunk(sctx, scrub_dev, chunk_tree, chunk_objectid,
+				  chunk_offset, length, found_key.offset,
+				  is_dev_replace);
+
+		/*
+		 * flush, submit all pending read and write bios, afterwards
+		 * wait for them.
+		 * Note that in the dev replace case, a read request causes
+		 * write requests that are submitted in the read completion
+		 * worker. Therefore in the current situation, it is required
+		 * that all write requests are flushed, so that all read and
+		 * write requests are really completed when bios_in_flight
+		 * changes to 0.
+		 */
+		atomic_set(&sctx->wr_ctx.flush_all_writes, 1);
+		scrub_submit(sctx);
+		mutex_lock(&sctx->wr_ctx.wr_lock);
+		scrub_wr_submit(sctx);
+		mutex_unlock(&sctx->wr_ctx.wr_lock);
+
+		wait_event(sctx->list_wait,
+			   atomic_read(&sctx->bios_in_flight) == 0);
+		atomic_inc(&fs_info->scrubs_paused);
+		wake_up(&fs_info->scrub_pause_wait);
+
+		/*
+		 * must be called before we decrease @scrub_paused.
+		 * make sure we don't block transaction commit while
+		 * we are waiting pending workers finished.
+		 */
+		wait_event(sctx->list_wait,
+			   atomic_read(&sctx->workers_pending) == 0);
+		atomic_set(&sctx->wr_ctx.flush_all_writes, 0);
+
+		mutex_lock(&fs_info->scrub_lock);
+		__scrub_blocked_if_needed(fs_info);
+		atomic_dec(&fs_info->scrubs_paused);
+		mutex_unlock(&fs_info->scrub_lock);
+		wake_up(&fs_info->scrub_pause_wait);
+
+		btrfs_put_block_group(cache);
+		if (ret)
+			break;
+		if (is_dev_replace &&
+		    atomic64_read(&dev_replace->num_write_errors) > 0) {
+			ret = -EIO;
+			break;
+		}
+		if (sctx->stat.malloc_errors > 0) {
+			ret = -ENOMEM;
+			break;
+		}
+
+		dev_replace->cursor_left = dev_replace->cursor_right;
+		dev_replace->item_needs_writeback = 1;
+skip:
+		key.offset = found_key.offset + length;
+		btrfs_release_path(path);
+	}
+
+	btrfs_free_path(path);
+
+	/*
+	 * ret can still be 1 from search_slot or next_leaf,
+	 * that's not an error
+	 */
+	return ret < 0 ? ret : 0;
+}
+
+static noinline_for_stack int scrub_supers(struct scrub_ctx *sctx,
+					   struct btrfs_device *scrub_dev)
+{
+	int	i;
+	u64	bytenr;
+	u64	gen;
+	int	ret;
+	struct btrfs_root *root = sctx->dev_root;
+
+	if (test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state))
+		return -EIO;
+
+	gen = root->fs_info->last_trans_committed;
+
+	for (i = 0; i < BTRFS_SUPER_MIRROR_MAX; i++) {
+		bytenr = btrfs_sb_offset(i);
+		if (bytenr + BTRFS_SUPER_INFO_SIZE > scrub_dev->total_bytes)
+			break;
+
+		ret = scrub_pages(sctx, bytenr, BTRFS_SUPER_INFO_SIZE, bytenr,
+				  scrub_dev, BTRFS_EXTENT_FLAG_SUPER, gen, i,
+				  NULL, 1, bytenr);
+		if (ret)
+			return ret;
+	}
+	wait_event(sctx->list_wait, atomic_read(&sctx->bios_in_flight) == 0);
+
+	return 0;
+}
+
+/*
+ * get a reference count on fs_info->scrub_workers. start worker if necessary
+ */
+static noinline_for_stack int scrub_workers_get(struct btrfs_fs_info *fs_info,
+						int is_dev_replace)
+{
+	int ret = 0;
+	int flags = WQ_FREEZABLE | WQ_UNBOUND;
+	int max_active = fs_info->thread_pool_size;
+
+	if (fs_info->scrub_workers_refcnt == 0) {
+		if (is_dev_replace)
+			fs_info->scrub_workers =
+				btrfs_alloc_workqueue("btrfs-scrub", flags,
+						      1, 4);
+		else
+			fs_info->scrub_workers =
+				btrfs_alloc_workqueue("btrfs-scrub", flags,
+						      max_active, 4);
+		if (!fs_info->scrub_workers) {
+			ret = -ENOMEM;
+			goto out;
+		}
+		fs_info->scrub_wr_completion_workers =
+			btrfs_alloc_workqueue("btrfs-scrubwrc", flags,
+					      max_active, 2);
+		if (!fs_info->scrub_wr_completion_workers) {
+			ret = -ENOMEM;
+			goto out;
+		}
+		fs_info->scrub_nocow_workers =
+			btrfs_alloc_workqueue("btrfs-scrubnc", flags, 1, 0);
+		if (!fs_info->scrub_nocow_workers) {
+			ret = -ENOMEM;
+			goto out;
+		}
+	}
+	++fs_info->scrub_workers_refcnt;
+out:
+	return ret;
+}
+
+static noinline_for_stack void scrub_workers_put(struct btrfs_fs_info *fs_info)
+{
+	if (--fs_info->scrub_workers_refcnt == 0) {
+		btrfs_destroy_workqueue(fs_info->scrub_workers);
+		btrfs_destroy_workqueue(fs_info->scrub_wr_completion_workers);
+		btrfs_destroy_workqueue(fs_info->scrub_nocow_workers);
+	}
+	WARN_ON(fs_info->scrub_workers_refcnt < 0);
+}
+
+int btrfs_scrub_dev(struct btrfs_fs_info *fs_info, u64 devid, u64 start,
+		    u64 end, struct btrfs_scrub_progress *progress,
+		    int readonly, int is_dev_replace)
+{
+	struct scrub_ctx *sctx;
+	int ret;
+	struct btrfs_device *dev;
+
+	if (btrfs_fs_closing(fs_info))
+		return -EINVAL;
+
+	/*
+	 * check some assumptions
+	 */
+	if (fs_info->chunk_root->nodesize != fs_info->chunk_root->leafsize) {
+		btrfs_err(fs_info,
+			   "scrub: size assumption nodesize == leafsize (%d == %d) fails",
+		       fs_info->chunk_root->nodesize,
+		       fs_info->chunk_root->leafsize);
+		return -EINVAL;
+	}
+
+	if (fs_info->chunk_root->nodesize > BTRFS_STRIPE_LEN) {
+		/*
+		 * in this case scrub is unable to calculate the checksum
+		 * the way scrub is implemented. Do not handle this
+		 * situation at all because it won't ever happen.
+		 */
+		btrfs_err(fs_info,
+			   "scrub: size assumption nodesize <= BTRFS_STRIPE_LEN (%d <= %d) fails",
+		       fs_info->chunk_root->nodesize, BTRFS_STRIPE_LEN);
+		return -EINVAL;
+	}
+
+	if (fs_info->chunk_root->sectorsize != PAGE_SIZE) {
+		/* not supported for data w/o checksums */
+		btrfs_err(fs_info,
+			   "scrub: size assumption sectorsize != PAGE_SIZE "
+			   "(%d != %lu) fails",
+		       fs_info->chunk_root->sectorsize, PAGE_SIZE);
+		return -EINVAL;
+	}
+
+	if (fs_info->chunk_root->nodesize >
+	    PAGE_SIZE * SCRUB_MAX_PAGES_PER_BLOCK ||
+	    fs_info->chunk_root->sectorsize >
+	    PAGE_SIZE * SCRUB_MAX_PAGES_PER_BLOCK) {
+		/*
+		 * would exhaust the array bounds of pagev member in
+		 * struct scrub_block
+		 */
+		btrfs_err(fs_info, "scrub: size assumption nodesize and sectorsize "
+			   "<= SCRUB_MAX_PAGES_PER_BLOCK (%d <= %d && %d <= %d) fails",
+		       fs_info->chunk_root->nodesize,
+		       SCRUB_MAX_PAGES_PER_BLOCK,
+		       fs_info->chunk_root->sectorsize,
+		       SCRUB_MAX_PAGES_PER_BLOCK);
+		return -EINVAL;
+	}
+
+
+	mutex_lock(&fs_info->fs_devices->device_list_mutex);
+	dev = btrfs_find_device(fs_info, devid, NULL, NULL);
+	if (!dev || (dev->missing && !is_dev_replace)) {
+		mutex_unlock(&fs_info->fs_devices->device_list_mutex);
+		return -ENODEV;
+	}
+
+	mutex_lock(&fs_info->scrub_lock);
+	if (!dev->in_fs_metadata || dev->is_tgtdev_for_dev_replace) {
+		mutex_unlock(&fs_info->scrub_lock);
+		mutex_unlock(&fs_info->fs_devices->device_list_mutex);
+		return -EIO;
+	}
+
+	btrfs_dev_replace_lock(&fs_info->dev_replace);
+	if (dev->scrub_device ||
+	    (!is_dev_replace &&
+	     btrfs_dev_replace_is_ongoing(&fs_info->dev_replace))) {
+		btrfs_dev_replace_unlock(&fs_info->dev_replace);
+		mutex_unlock(&fs_info->scrub_lock);
+		mutex_unlock(&fs_info->fs_devices->device_list_mutex);
+		return -EINPROGRESS;
+	}
+	btrfs_dev_replace_unlock(&fs_info->dev_replace);
+
+	ret = scrub_workers_get(fs_info, is_dev_replace);
+	if (ret) {
+		mutex_unlock(&fs_info->scrub_lock);
+		mutex_unlock(&fs_info->fs_devices->device_list_mutex);
+		return ret;
+	}
+
+	sctx = scrub_setup_ctx(dev, is_dev_replace);
+	if (IS_ERR(sctx)) {
+		mutex_unlock(&fs_info->scrub_lock);
+		mutex_unlock(&fs_info->fs_devices->device_list_mutex);
+		scrub_workers_put(fs_info);
+		return PTR_ERR(sctx);
+	}
+	sctx->readonly = readonly;
+	dev->scrub_device = sctx;
+	mutex_unlock(&fs_info->fs_devices->device_list_mutex);
+
+	/*
+	 * checking @scrub_pause_req here, we can avoid
+	 * race between committing transaction and scrubbing.
+	 */
+	__scrub_blocked_if_needed(fs_info);
+	atomic_inc(&fs_info->scrubs_running);
+	mutex_unlock(&fs_info->scrub_lock);
+
+	if (!is_dev_replace) {
+		/*
+		 * by holding device list mutex, we can
+		 * kick off writing super in log tree sync.
+		 */
+		mutex_lock(&fs_info->fs_devices->device_list_mutex);
+		ret = scrub_supers(sctx, dev);
+		mutex_unlock(&fs_info->fs_devices->device_list_mutex);
+	}
+
+	if (!ret)
+		ret = scrub_enumerate_chunks(sctx, dev, start, end,
+					     is_dev_replace);
+
+	wait_event(sctx->list_wait, atomic_read(&sctx->bios_in_flight) == 0);
+	atomic_dec(&fs_info->scrubs_running);
+	wake_up(&fs_info->scrub_pause_wait);
+
+	wait_event(sctx->list_wait, atomic_read(&sctx->workers_pending) == 0);
+
+	if (progress)
+		memcpy(progress, &sctx->stat, sizeof(*progress));
+
+	mutex_lock(&fs_info->scrub_lock);
+	dev->scrub_device = NULL;
+	scrub_workers_put(fs_info);
+	mutex_unlock(&fs_info->scrub_lock);
+
+	scrub_free_ctx(sctx);
+
+	return ret;
+}
+
+void btrfs_scrub_pause(struct btrfs_root *root)
+{
+	struct btrfs_fs_info *fs_info = root->fs_info;
+
+	mutex_lock(&fs_info->scrub_lock);
+	atomic_inc(&fs_info->scrub_pause_req);
+	while (atomic_read(&fs_info->scrubs_paused) !=
+	       atomic_read(&fs_info->scrubs_running)) {
+		mutex_unlock(&fs_info->scrub_lock);
+		wait_event(fs_info->scrub_pause_wait,
+			   atomic_read(&fs_info->scrubs_paused) ==
+			   atomic_read(&fs_info->scrubs_running));
+		mutex_lock(&fs_info->scrub_lock);
+	}
+	mutex_unlock(&fs_info->scrub_lock);
+}
+
+void btrfs_scrub_continue(struct btrfs_root *root)
+{
+	struct btrfs_fs_info *fs_info = root->fs_info;
+
+	atomic_dec(&fs_info->scrub_pause_req);
+	wake_up(&fs_info->scrub_pause_wait);
+}
+
+int btrfs_scrub_cancel(struct btrfs_fs_info *fs_info)
+{
+	mutex_lock(&fs_info->scrub_lock);
+	if (!atomic_read(&fs_info->scrubs_running)) {
+		mutex_unlock(&fs_info->scrub_lock);
+		return -ENOTCONN;
+	}
+
+	atomic_inc(&fs_info->scrub_cancel_req);
+	while (atomic_read(&fs_info->scrubs_running)) {
+		mutex_unlock(&fs_info->scrub_lock);
+		wait_event(fs_info->scrub_pause_wait,
+			   atomic_read(&fs_info->scrubs_running) == 0);
+		mutex_lock(&fs_info->scrub_lock);
+	}
+	atomic_dec(&fs_info->scrub_cancel_req);
+	mutex_unlock(&fs_info->scrub_lock);
+
+	return 0;
+}
+
+int btrfs_scrub_cancel_dev(struct btrfs_fs_info *fs_info,
+			   struct btrfs_device *dev)
+{
+	struct scrub_ctx *sctx;
+
+	mutex_lock(&fs_info->scrub_lock);
+	sctx = dev->scrub_device;
+	if (!sctx) {
+		mutex_unlock(&fs_info->scrub_lock);
+		return -ENOTCONN;
+	}
+	atomic_inc(&sctx->cancel_req);
+	while (dev->scrub_device) {
+		mutex_unlock(&fs_info->scrub_lock);
+		wait_event(fs_info->scrub_pause_wait,
+			   dev->scrub_device == NULL);
+		mutex_lock(&fs_info->scrub_lock);
+	}
+	mutex_unlock(&fs_info->scrub_lock);
+
+	return 0;
+}
+
+int btrfs_scrub_progress(struct btrfs_root *root, u64 devid,
+			 struct btrfs_scrub_progress *progress)
+{
+	struct btrfs_device *dev;
+	struct scrub_ctx *sctx = NULL;
+
+	mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
+	dev = btrfs_find_device(root->fs_info, devid, NULL, NULL);
+	if (dev)
+		sctx = dev->scrub_device;
+	if (sctx)
+		memcpy(progress, &sctx->stat, sizeof(*progress));
+	mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
+
+	return dev ? (sctx ? 0 : -ENOTCONN) : -ENODEV;
+}
+
+static void scrub_remap_extent(struct btrfs_fs_info *fs_info,
+			       u64 extent_logical, u64 extent_len,
+			       u64 *extent_physical,
+			       struct btrfs_device **extent_dev,
+			       int *extent_mirror_num)
+{
+	u64 mapped_length;
+	struct btrfs_bio *bbio = NULL;
+	int ret;
+
+	mapped_length = extent_len;
+	ret = btrfs_map_block(fs_info, READ, extent_logical,
+			      &mapped_length, &bbio, 0);
+	if (ret || !bbio || mapped_length < extent_len ||
+	    !bbio->stripes[0].dev->bdev) {
+		kfree(bbio);
+		return;
+	}
+
+	*extent_physical = bbio->stripes[0].physical;
+	*extent_mirror_num = bbio->mirror_num;
+	*extent_dev = bbio->stripes[0].dev;
+	kfree(bbio);
+}
+
+static int scrub_setup_wr_ctx(struct scrub_ctx *sctx,
+			      struct scrub_wr_ctx *wr_ctx,
+			      struct btrfs_fs_info *fs_info,
+			      struct btrfs_device *dev,
+			      int is_dev_replace)
+{
+	WARN_ON(wr_ctx->wr_curr_bio != NULL);
+
+	mutex_init(&wr_ctx->wr_lock);
+	wr_ctx->wr_curr_bio = NULL;
+	if (!is_dev_replace)
+		return 0;
+
+	WARN_ON(!dev->bdev);
+	wr_ctx->pages_per_wr_bio = min_t(int, SCRUB_PAGES_PER_WR_BIO,
+					 bio_get_nr_vecs(dev->bdev));
+	wr_ctx->tgtdev = dev;
+	atomic_set(&wr_ctx->flush_all_writes, 0);
+	return 0;
+}
+
+static void scrub_free_wr_ctx(struct scrub_wr_ctx *wr_ctx)
+{
+	mutex_lock(&wr_ctx->wr_lock);
+	kfree(wr_ctx->wr_curr_bio);
+	wr_ctx->wr_curr_bio = NULL;
+	mutex_unlock(&wr_ctx->wr_lock);
+}
+
+static int copy_nocow_pages(struct scrub_ctx *sctx, u64 logical, u64 len,
+			    int mirror_num, u64 physical_for_dev_replace)
+{
+	struct scrub_copy_nocow_ctx *nocow_ctx;
+	struct btrfs_fs_info *fs_info = sctx->dev_root->fs_info;
+
+	nocow_ctx = kzalloc(sizeof(*nocow_ctx), GFP_NOFS);
+	if (!nocow_ctx) {
+		spin_lock(&sctx->stat_lock);
+		sctx->stat.malloc_errors++;
+		spin_unlock(&sctx->stat_lock);
+		return -ENOMEM;
+	}
+
+	scrub_pending_trans_workers_inc(sctx);
+
+	nocow_ctx->sctx = sctx;
+	nocow_ctx->logical = logical;
+	nocow_ctx->len = len;
+	nocow_ctx->mirror_num = mirror_num;
+	nocow_ctx->physical_for_dev_replace = physical_for_dev_replace;
+	btrfs_init_work(&nocow_ctx->work, copy_nocow_pages_worker, NULL, NULL);
+	INIT_LIST_HEAD(&nocow_ctx->inodes);
+	btrfs_queue_work(fs_info->scrub_nocow_workers,
+			 &nocow_ctx->work);
+
+	return 0;
+}
+
+static int record_inode_for_nocow(u64 inum, u64 offset, u64 root, void *ctx)
+{
+	struct scrub_copy_nocow_ctx *nocow_ctx = ctx;
+	struct scrub_nocow_inode *nocow_inode;
+
+	nocow_inode = kzalloc(sizeof(*nocow_inode), GFP_NOFS);
+	if (!nocow_inode)
+		return -ENOMEM;
+	nocow_inode->inum = inum;
+	nocow_inode->offset = offset;
+	nocow_inode->root = root;
+	list_add_tail(&nocow_inode->list, &nocow_ctx->inodes);
+	return 0;
+}
+
+#define COPY_COMPLETE 1
+
+static void copy_nocow_pages_worker(struct btrfs_work *work)
+{
+	struct scrub_copy_nocow_ctx *nocow_ctx =
+		container_of(work, struct scrub_copy_nocow_ctx, work);
+	struct scrub_ctx *sctx = nocow_ctx->sctx;
+	u64 logical = nocow_ctx->logical;
+	u64 len = nocow_ctx->len;
+	int mirror_num = nocow_ctx->mirror_num;
+	u64 physical_for_dev_replace = nocow_ctx->physical_for_dev_replace;
+	int ret;
+	struct btrfs_trans_handle *trans = NULL;
+	struct btrfs_fs_info *fs_info;
+	struct btrfs_path *path;
+	struct btrfs_root *root;
+	int not_written = 0;
+
+	fs_info = sctx->dev_root->fs_info;
+	root = fs_info->extent_root;
+
+	path = btrfs_alloc_path();
+	if (!path) {
+		spin_lock(&sctx->stat_lock);
+		sctx->stat.malloc_errors++;
+		spin_unlock(&sctx->stat_lock);
+		not_written = 1;
+		goto out;
+	}
+
+	trans = btrfs_join_transaction(root);
+	if (IS_ERR(trans)) {
+		not_written = 1;
+		goto out;
+	}
+
+	ret = iterate_inodes_from_logical(logical, fs_info, path,
+					  record_inode_for_nocow, nocow_ctx);
+	if (ret != 0 && ret != -ENOENT) {
+		btrfs_warn(fs_info, "iterate_inodes_from_logical() failed: log %llu, "
+			"phys %llu, len %llu, mir %u, ret %d",
+			logical, physical_for_dev_replace, len, mirror_num,
+			ret);
+		not_written = 1;
+		goto out;
+	}
+
+	btrfs_end_transaction(trans, root);
+	trans = NULL;
+	while (!list_empty(&nocow_ctx->inodes)) {
+		struct scrub_nocow_inode *entry;
+		entry = list_first_entry(&nocow_ctx->inodes,
+					 struct scrub_nocow_inode,
+					 list);
+		list_del_init(&entry->list);
+		ret = copy_nocow_pages_for_inode(entry->inum, entry->offset,
+						 entry->root, nocow_ctx);
+		kfree(entry);
+		if (ret == COPY_COMPLETE) {
+			ret = 0;
+			break;
+		} else if (ret) {
+			break;
+		}
+	}
+out:
+	while (!list_empty(&nocow_ctx->inodes)) {
+		struct scrub_nocow_inode *entry;
+		entry = list_first_entry(&nocow_ctx->inodes,
+					 struct scrub_nocow_inode,
+					 list);
+		list_del_init(&entry->list);
+		kfree(entry);
+	}
+	if (trans && !IS_ERR(trans))
+		btrfs_end_transaction(trans, root);
+	if (not_written)
+		btrfs_dev_replace_stats_inc(&fs_info->dev_replace.
+					    num_uncorrectable_read_errors);
+
+	btrfs_free_path(path);
+	kfree(nocow_ctx);
+
+	scrub_pending_trans_workers_dec(sctx);
+}
+
+static int copy_nocow_pages_for_inode(u64 inum, u64 offset, u64 root,
+				      struct scrub_copy_nocow_ctx *nocow_ctx)
+{
+	struct btrfs_fs_info *fs_info = nocow_ctx->sctx->dev_root->fs_info;
+	struct btrfs_key key;
+	struct inode *inode;
+	struct page *page;
+	struct btrfs_root *local_root;
+	struct btrfs_ordered_extent *ordered;
+	struct extent_map *em;
+	struct extent_state *cached_state = NULL;
+	struct extent_io_tree *io_tree;
+	u64 physical_for_dev_replace;
+	u64 len = nocow_ctx->len;
+	u64 lockstart = offset, lockend = offset + len - 1;
+	unsigned long index;
+	int srcu_index;
+	int ret = 0;
+	int err = 0;
+
+	key.objectid = root;
+	key.type = BTRFS_ROOT_ITEM_KEY;
+	key.offset = (u64)-1;
+
+	srcu_index = srcu_read_lock(&fs_info->subvol_srcu);
+
+	local_root = btrfs_read_fs_root_no_name(fs_info, &key);
+	if (IS_ERR(local_root)) {
+		srcu_read_unlock(&fs_info->subvol_srcu, srcu_index);
+		return PTR_ERR(local_root);
+	}
+
+	key.type = BTRFS_INODE_ITEM_KEY;
+	key.objectid = inum;
+	key.offset = 0;
+	inode = btrfs_iget(fs_info->sb, &key, local_root, NULL);
+	srcu_read_unlock(&fs_info->subvol_srcu, srcu_index);
+	if (IS_ERR(inode))
+		return PTR_ERR(inode);
+
+	/* Avoid truncate/dio/punch hole.. */
+	mutex_lock(&inode->i_mutex);
+	inode_dio_wait(inode);
+
+	physical_for_dev_replace = nocow_ctx->physical_for_dev_replace;
+	io_tree = &BTRFS_I(inode)->io_tree;
+
+	lock_extent_bits(io_tree, lockstart, lockend, 0, &cached_state);
+	ordered = btrfs_lookup_ordered_range(inode, lockstart, len);
+	if (ordered) {
+		btrfs_put_ordered_extent(ordered);
+		goto out_unlock;
+	}
+
+	em = btrfs_get_extent(inode, NULL, 0, lockstart, len, 0);
+	if (IS_ERR(em)) {
+		ret = PTR_ERR(em);
+		goto out_unlock;
+	}
+
+	/*
+	 * This extent does not actually cover the logical extent anymore,
+	 * move on to the next inode.
+	 */
+	if (em->block_start > nocow_ctx->logical ||
+	    em->block_start + em->block_len < nocow_ctx->logical + len) {
+		free_extent_map(em);
+		goto out_unlock;
+	}
+	free_extent_map(em);
+
+	while (len >= PAGE_CACHE_SIZE) {
+		index = offset >> PAGE_CACHE_SHIFT;
+again:
+		page = find_or_create_page(inode->i_mapping, index, GFP_NOFS);
+		if (!page) {
+			btrfs_err(fs_info, "find_or_create_page() failed");
+			ret = -ENOMEM;
+			goto out;
+		}
+
+		if (PageUptodate(page)) {
+			if (PageDirty(page))
+				goto next_page;
+		} else {
+			ClearPageError(page);
+			err = extent_read_full_page_nolock(io_tree, page,
+							   btrfs_get_extent,
+							   nocow_ctx->mirror_num);
+			if (err) {
+				ret = err;
+				goto next_page;
+			}
+
+			lock_page(page);
+			/*
+			 * If the page has been remove from the page cache,
+			 * the data on it is meaningless, because it may be
+			 * old one, the new data may be written into the new
+			 * page in the page cache.
+			 */
+			if (page->mapping != inode->i_mapping) {
+				unlock_page(page);
+				page_cache_release(page);
+				goto again;
+			}
+			if (!PageUptodate(page)) {
+				ret = -EIO;
+				goto next_page;
+			}
+		}
+		err = write_page_nocow(nocow_ctx->sctx,
+				       physical_for_dev_replace, page);
+		if (err)
+			ret = err;
+next_page:
+		unlock_page(page);
+		page_cache_release(page);
+
+		if (ret)
+			break;
+
+		offset += PAGE_CACHE_SIZE;
+		physical_for_dev_replace += PAGE_CACHE_SIZE;
+		len -= PAGE_CACHE_SIZE;
+	}
+	ret = COPY_COMPLETE;
+out_unlock:
+	unlock_extent_cached(io_tree, lockstart, lockend, &cached_state,
+			     GFP_NOFS);
+out:
+	mutex_unlock(&inode->i_mutex);
+	iput(inode);
+	return ret;
+}
+
+static int write_page_nocow(struct scrub_ctx *sctx,
+			    u64 physical_for_dev_replace, struct page *page)
+{
+	struct bio *bio;
+	struct btrfs_device *dev;
+	int ret;
+
+	dev = sctx->wr_ctx.tgtdev;
+	if (!dev)
+		return -EIO;
+	if (!dev->bdev) {
+		printk_ratelimited(KERN_WARNING
+			"BTRFS: scrub write_page_nocow(bdev == NULL) is unexpected!\n");
+		return -EIO;
+	}
+	bio = btrfs_io_bio_alloc(GFP_NOFS, 1);
+	if (!bio) {
+		spin_lock(&sctx->stat_lock);
+		sctx->stat.malloc_errors++;
+		spin_unlock(&sctx->stat_lock);
+		return -ENOMEM;
+	}
+	bio->bi_iter.bi_size = 0;
+	bio->bi_iter.bi_sector = physical_for_dev_replace >> 9;
+	bio->bi_bdev = dev->bdev;
+	ret = bio_add_page(bio, page, PAGE_CACHE_SIZE, 0);
+	if (ret != PAGE_CACHE_SIZE) {
+leave_with_eio:
+		bio_put(bio);
+		btrfs_dev_stat_inc_and_print(dev, BTRFS_DEV_STAT_WRITE_ERRS);
+		return -EIO;
+	}
+
+	if (btrfsic_submit_bio_wait(WRITE_SYNC, bio))
+		goto leave_with_eio;
+
+	bio_put(bio);
+	return 0;
+}
diff --git a/fs/btrfs/send.c b/fs/btrfs/send.c
new file mode 100644
index 00000000000..6528aa66218
--- /dev/null
+++ b/fs/btrfs/send.c
@@ -0,0 +1,5791 @@
+/*
+ * Copyright (C) 2012 Alexander Block.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/bsearch.h>
+#include <linux/fs.h>
+#include <linux/file.h>
+#include <linux/sort.h>
+#include <linux/mount.h>
+#include <linux/xattr.h>
+#include <linux/posix_acl_xattr.h>
+#include <linux/radix-tree.h>
+#include <linux/vmalloc.h>
+#include <linux/string.h>
+
+#include "send.h"
+#include "backref.h"
+#include "hash.h"
+#include "locking.h"
+#include "disk-io.h"
+#include "btrfs_inode.h"
+#include "transaction.h"
+
+static int g_verbose = 0;
+
+#define verbose_printk(...) if (g_verbose) printk(__VA_ARGS__)
+
+/*
+ * A fs_path is a helper to dynamically build path names with unknown size.
+ * It reallocates the internal buffer on demand.
+ * It allows fast adding of path elements on the right side (normal path) and
+ * fast adding to the left side (reversed path). A reversed path can also be
+ * unreversed if needed.
+ */
+struct fs_path {
+	union {
+		struct {
+			char *start;
+			char *end;
+
+			char *buf;
+			unsigned short buf_len:15;
+			unsigned short reversed:1;
+			char inline_buf[];
+		};
+		/*
+		 * Average path length does not exceed 200 bytes, we'll have
+		 * better packing in the slab and higher chance to satisfy
+		 * a allocation later during send.
+		 */
+		char pad[256];
+	};
+};
+#define FS_PATH_INLINE_SIZE \
+	(sizeof(struct fs_path) - offsetof(struct fs_path, inline_buf))
+
+
+/* reused for each extent */
+struct clone_root {
+	struct btrfs_root *root;
+	u64 ino;
+	u64 offset;
+
+	u64 found_refs;
+};
+
+#define SEND_CTX_MAX_NAME_CACHE_SIZE 128
+#define SEND_CTX_NAME_CACHE_CLEAN_SIZE (SEND_CTX_MAX_NAME_CACHE_SIZE * 2)
+
+struct send_ctx {
+	struct file *send_filp;
+	loff_t send_off;
+	char *send_buf;
+	u32 send_size;
+	u32 send_max_size;
+	u64 total_send_size;
+	u64 cmd_send_size[BTRFS_SEND_C_MAX + 1];
+	u64 flags;	/* 'flags' member of btrfs_ioctl_send_args is u64 */
+
+	struct btrfs_root *send_root;
+	struct btrfs_root *parent_root;
+	struct clone_root *clone_roots;
+	int clone_roots_cnt;
+
+	/* current state of the compare_tree call */
+	struct btrfs_path *left_path;
+	struct btrfs_path *right_path;
+	struct btrfs_key *cmp_key;
+
+	/*
+	 * infos of the currently processed inode. In case of deleted inodes,
+	 * these are the values from the deleted inode.
+	 */
+	u64 cur_ino;
+	u64 cur_inode_gen;
+	int cur_inode_new;
+	int cur_inode_new_gen;
+	int cur_inode_deleted;
+	u64 cur_inode_size;
+	u64 cur_inode_mode;
+	u64 cur_inode_rdev;
+	u64 cur_inode_last_extent;
+
+	u64 send_progress;
+
+	struct list_head new_refs;
+	struct list_head deleted_refs;
+
+	struct radix_tree_root name_cache;
+	struct list_head name_cache_list;
+	int name_cache_size;
+
+	struct file_ra_state ra;
+
+	char *read_buf;
+
+	/*
+	 * We process inodes by their increasing order, so if before an
+	 * incremental send we reverse the parent/child relationship of
+	 * directories such that a directory with a lower inode number was
+	 * the parent of a directory with a higher inode number, and the one
+	 * becoming the new parent got renamed too, we can't rename/move the
+	 * directory with lower inode number when we finish processing it - we
+	 * must process the directory with higher inode number first, then
+	 * rename/move it and then rename/move the directory with lower inode
+	 * number. Example follows.
+	 *
+	 * Tree state when the first send was performed:
+	 *
+	 * .
+	 * |-- a                   (ino 257)
+	 *     |-- b               (ino 258)
+	 *         |
+	 *         |
+	 *         |-- c           (ino 259)
+	 *         |   |-- d       (ino 260)
+	 *         |
+	 *         |-- c2          (ino 261)
+	 *
+	 * Tree state when the second (incremental) send is performed:
+	 *
+	 * .
+	 * |-- a                   (ino 257)
+	 *     |-- b               (ino 258)
+	 *         |-- c2          (ino 261)
+	 *             |-- d2      (ino 260)
+	 *                 |-- cc  (ino 259)
+	 *
+	 * The sequence of steps that lead to the second state was:
+	 *
+	 * mv /a/b/c/d /a/b/c2/d2
+	 * mv /a/b/c /a/b/c2/d2/cc
+	 *
+	 * "c" has lower inode number, but we can't move it (2nd mv operation)
+	 * before we move "d", which has higher inode number.
+	 *
+	 * So we just memorize which move/rename operations must be performed
+	 * later when their respective parent is processed and moved/renamed.
+	 */
+
+	/* Indexed by parent directory inode number. */
+	struct rb_root pending_dir_moves;
+
+	/*
+	 * Reverse index, indexed by the inode number of a directory that
+	 * is waiting for the move/rename of its immediate parent before its
+	 * own move/rename can be performed.
+	 */
+	struct rb_root waiting_dir_moves;
+
+	/*
+	 * A directory that is going to be rm'ed might have a child directory
+	 * which is in the pending directory moves index above. In this case,
+	 * the directory can only be removed after the move/rename of its child
+	 * is performed. Example:
+	 *
+	 * Parent snapshot:
+	 *
+	 * .                        (ino 256)
+	 * |-- a/                   (ino 257)
+	 *     |-- b/               (ino 258)
+	 *         |-- c/           (ino 259)
+	 *         |   |-- x/       (ino 260)
+	 *         |
+	 *         |-- y/           (ino 261)
+	 *
+	 * Send snapshot:
+	 *
+	 * .                        (ino 256)
+	 * |-- a/                   (ino 257)
+	 *     |-- b/               (ino 258)
+	 *         |-- YY/          (ino 261)
+	 *              |-- x/      (ino 260)
+	 *
+	 * Sequence of steps that lead to the send snapshot:
+	 * rm -f /a/b/c/foo.txt
+	 * mv /a/b/y /a/b/YY
+	 * mv /a/b/c/x /a/b/YY
+	 * rmdir /a/b/c
+	 *
+	 * When the child is processed, its move/rename is delayed until its
+	 * parent is processed (as explained above), but all other operations
+	 * like update utimes, chown, chgrp, etc, are performed and the paths
+	 * that it uses for those operations must use the orphanized name of
+	 * its parent (the directory we're going to rm later), so we need to
+	 * memorize that name.
+	 *
+	 * Indexed by the inode number of the directory to be deleted.
+	 */
+	struct rb_root orphan_dirs;
+};
+
+struct pending_dir_move {
+	struct rb_node node;
+	struct list_head list;
+	u64 parent_ino;
+	u64 ino;
+	u64 gen;
+	struct list_head update_refs;
+};
+
+struct waiting_dir_move {
+	struct rb_node node;
+	u64 ino;
+	/*
+	 * There might be some directory that could not be removed because it
+	 * was waiting for this directory inode to be moved first. Therefore
+	 * after this directory is moved, we can try to rmdir the ino rmdir_ino.
+	 */
+	u64 rmdir_ino;
+};
+
+struct orphan_dir_info {
+	struct rb_node node;
+	u64 ino;
+	u64 gen;
+};
+
+struct name_cache_entry {
+	struct list_head list;
+	/*
+	 * radix_tree has only 32bit entries but we need to handle 64bit inums.
+	 * We use the lower 32bit of the 64bit inum to store it in the tree. If
+	 * more then one inum would fall into the same entry, we use radix_list
+	 * to store the additional entries. radix_list is also used to store
+	 * entries where two entries have the same inum but different
+	 * generations.
+	 */
+	struct list_head radix_list;
+	u64 ino;
+	u64 gen;
+	u64 parent_ino;
+	u64 parent_gen;
+	int ret;
+	int need_later_update;
+	int name_len;
+	char name[];
+};
+
+static int is_waiting_for_move(struct send_ctx *sctx, u64 ino);
+
+static struct waiting_dir_move *
+get_waiting_dir_move(struct send_ctx *sctx, u64 ino);
+
+static int is_waiting_for_rm(struct send_ctx *sctx, u64 dir_ino);
+
+static int need_send_hole(struct send_ctx *sctx)
+{
+	return (sctx->parent_root && !sctx->cur_inode_new &&
+		!sctx->cur_inode_new_gen && !sctx->cur_inode_deleted &&
+		S_ISREG(sctx->cur_inode_mode));
+}
+
+static void fs_path_reset(struct fs_path *p)
+{
+	if (p->reversed) {
+		p->start = p->buf + p->buf_len - 1;
+		p->end = p->start;
+		*p->start = 0;
+	} else {
+		p->start = p->buf;
+		p->end = p->start;
+		*p->start = 0;
+	}
+}
+
+static struct fs_path *fs_path_alloc(void)
+{
+	struct fs_path *p;
+
+	p = kmalloc(sizeof(*p), GFP_NOFS);
+	if (!p)
+		return NULL;
+	p->reversed = 0;
+	p->buf = p->inline_buf;
+	p->buf_len = FS_PATH_INLINE_SIZE;
+	fs_path_reset(p);
+	return p;
+}
+
+static struct fs_path *fs_path_alloc_reversed(void)
+{
+	struct fs_path *p;
+
+	p = fs_path_alloc();
+	if (!p)
+		return NULL;
+	p->reversed = 1;
+	fs_path_reset(p);
+	return p;
+}
+
+static void fs_path_free(struct fs_path *p)
+{
+	if (!p)
+		return;
+	if (p->buf != p->inline_buf)
+		kfree(p->buf);
+	kfree(p);
+}
+
+static int fs_path_len(struct fs_path *p)
+{
+	return p->end - p->start;
+}
+
+static int fs_path_ensure_buf(struct fs_path *p, int len)
+{
+	char *tmp_buf;
+	int path_len;
+	int old_buf_len;
+
+	len++;
+
+	if (p->buf_len >= len)
+		return 0;
+
+	if (len > PATH_MAX) {
+		WARN_ON(1);
+		return -ENOMEM;
+	}
+
+	path_len = p->end - p->start;
+	old_buf_len = p->buf_len;
+
+	/*
+	 * First time the inline_buf does not suffice
+	 */
+	if (p->buf == p->inline_buf) {
+		tmp_buf = kmalloc(len, GFP_NOFS);
+		if (tmp_buf)
+			memcpy(tmp_buf, p->buf, old_buf_len);
+	} else {
+		tmp_buf = krealloc(p->buf, len, GFP_NOFS);
+	}
+	if (!tmp_buf)
+		return -ENOMEM;
+	p->buf = tmp_buf;
+	/*
+	 * The real size of the buffer is bigger, this will let the fast path
+	 * happen most of the time
+	 */
+	p->buf_len = ksize(p->buf);
+
+	if (p->reversed) {
+		tmp_buf = p->buf + old_buf_len - path_len - 1;
+		p->end = p->buf + p->buf_len - 1;
+		p->start = p->end - path_len;
+		memmove(p->start, tmp_buf, path_len + 1);
+	} else {
+		p->start = p->buf;
+		p->end = p->start + path_len;
+	}
+	return 0;
+}
+
+static int fs_path_prepare_for_add(struct fs_path *p, int name_len,
+				   char **prepared)
+{
+	int ret;
+	int new_len;
+
+	new_len = p->end - p->start + name_len;
+	if (p->start != p->end)
+		new_len++;
+	ret = fs_path_ensure_buf(p, new_len);
+	if (ret < 0)
+		goto out;
+
+	if (p->reversed) {
+		if (p->start != p->end)
+			*--p->start = '/';
+		p->start -= name_len;
+		*prepared = p->start;
+	} else {
+		if (p->start != p->end)
+			*p->end++ = '/';
+		*prepared = p->end;
+		p->end += name_len;
+		*p->end = 0;
+	}
+
+out:
+	return ret;
+}
+
+static int fs_path_add(struct fs_path *p, const char *name, int name_len)
+{
+	int ret;
+	char *prepared;
+
+	ret = fs_path_prepare_for_add(p, name_len, &prepared);
+	if (ret < 0)
+		goto out;
+	memcpy(prepared, name, name_len);
+
+out:
+	return ret;
+}
+
+static int fs_path_add_path(struct fs_path *p, struct fs_path *p2)
+{
+	int ret;
+	char *prepared;
+
+	ret = fs_path_prepare_for_add(p, p2->end - p2->start, &prepared);
+	if (ret < 0)
+		goto out;
+	memcpy(prepared, p2->start, p2->end - p2->start);
+
+out:
+	return ret;
+}
+
+static int fs_path_add_from_extent_buffer(struct fs_path *p,
+					  struct extent_buffer *eb,
+					  unsigned long off, int len)
+{
+	int ret;
+	char *prepared;
+
+	ret = fs_path_prepare_for_add(p, len, &prepared);
+	if (ret < 0)
+		goto out;
+
+	read_extent_buffer(eb, prepared, off, len);
+
+out:
+	return ret;
+}
+
+static int fs_path_copy(struct fs_path *p, struct fs_path *from)
+{
+	int ret;
+
+	p->reversed = from->reversed;
+	fs_path_reset(p);
+
+	ret = fs_path_add_path(p, from);
+
+	return ret;
+}
+
+
+static void fs_path_unreverse(struct fs_path *p)
+{
+	char *tmp;
+	int len;
+
+	if (!p->reversed)
+		return;
+
+	tmp = p->start;
+	len = p->end - p->start;
+	p->start = p->buf;
+	p->end = p->start + len;
+	memmove(p->start, tmp, len + 1);
+	p->reversed = 0;
+}
+
+static struct btrfs_path *alloc_path_for_send(void)
+{
+	struct btrfs_path *path;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return NULL;
+	path->search_commit_root = 1;
+	path->skip_locking = 1;
+	path->need_commit_sem = 1;
+	return path;
+}
+
+static int write_buf(struct file *filp, const void *buf, u32 len, loff_t *off)
+{
+	int ret;
+	mm_segment_t old_fs;
+	u32 pos = 0;
+
+	old_fs = get_fs();
+	set_fs(KERNEL_DS);
+
+	while (pos < len) {
+		ret = vfs_write(filp, (char *)buf + pos, len - pos, off);
+		/* TODO handle that correctly */
+		/*if (ret == -ERESTARTSYS) {
+			continue;
+		}*/
+		if (ret < 0)
+			goto out;
+		if (ret == 0) {
+			ret = -EIO;
+			goto out;
+		}
+		pos += ret;
+	}
+
+	ret = 0;
+
+out:
+	set_fs(old_fs);
+	return ret;
+}
+
+static int tlv_put(struct send_ctx *sctx, u16 attr, const void *data, int len)
+{
+	struct btrfs_tlv_header *hdr;
+	int total_len = sizeof(*hdr) + len;
+	int left = sctx->send_max_size - sctx->send_size;
+
+	if (unlikely(left < total_len))
+		return -EOVERFLOW;
+
+	hdr = (struct btrfs_tlv_header *) (sctx->send_buf + sctx->send_size);
+	hdr->tlv_type = cpu_to_le16(attr);
+	hdr->tlv_len = cpu_to_le16(len);
+	memcpy(hdr + 1, data, len);
+	sctx->send_size += total_len;
+
+	return 0;
+}
+
+#define TLV_PUT_DEFINE_INT(bits) \
+	static int tlv_put_u##bits(struct send_ctx *sctx,	 	\
+			u##bits attr, u##bits value)			\
+	{								\
+		__le##bits __tmp = cpu_to_le##bits(value);		\
+		return tlv_put(sctx, attr, &__tmp, sizeof(__tmp));	\
+	}
+
+TLV_PUT_DEFINE_INT(64)
+
+static int tlv_put_string(struct send_ctx *sctx, u16 attr,
+			  const char *str, int len)
+{
+	if (len == -1)
+		len = strlen(str);
+	return tlv_put(sctx, attr, str, len);
+}
+
+static int tlv_put_uuid(struct send_ctx *sctx, u16 attr,
+			const u8 *uuid)
+{
+	return tlv_put(sctx, attr, uuid, BTRFS_UUID_SIZE);
+}
+
+static int tlv_put_btrfs_timespec(struct send_ctx *sctx, u16 attr,
+				  struct extent_buffer *eb,
+				  struct btrfs_timespec *ts)
+{
+	struct btrfs_timespec bts;
+	read_extent_buffer(eb, &bts, (unsigned long)ts, sizeof(bts));
+	return tlv_put(sctx, attr, &bts, sizeof(bts));
+}
+
+
+#define TLV_PUT(sctx, attrtype, attrlen, data) \
+	do { \
+		ret = tlv_put(sctx, attrtype, attrlen, data); \
+		if (ret < 0) \
+			goto tlv_put_failure; \
+	} while (0)
+
+#define TLV_PUT_INT(sctx, attrtype, bits, value) \
+	do { \
+		ret = tlv_put_u##bits(sctx, attrtype, value); \
+		if (ret < 0) \
+			goto tlv_put_failure; \
+	} while (0)
+
+#define TLV_PUT_U8(sctx, attrtype, data) TLV_PUT_INT(sctx, attrtype, 8, data)
+#define TLV_PUT_U16(sctx, attrtype, data) TLV_PUT_INT(sctx, attrtype, 16, data)
+#define TLV_PUT_U32(sctx, attrtype, data) TLV_PUT_INT(sctx, attrtype, 32, data)
+#define TLV_PUT_U64(sctx, attrtype, data) TLV_PUT_INT(sctx, attrtype, 64, data)
+#define TLV_PUT_STRING(sctx, attrtype, str, len) \
+	do { \
+		ret = tlv_put_string(sctx, attrtype, str, len); \
+		if (ret < 0) \
+			goto tlv_put_failure; \
+	} while (0)
+#define TLV_PUT_PATH(sctx, attrtype, p) \
+	do { \
+		ret = tlv_put_string(sctx, attrtype, p->start, \
+			p->end - p->start); \
+		if (ret < 0) \
+			goto tlv_put_failure; \
+	} while(0)
+#define TLV_PUT_UUID(sctx, attrtype, uuid) \
+	do { \
+		ret = tlv_put_uuid(sctx, attrtype, uuid); \
+		if (ret < 0) \
+			goto tlv_put_failure; \
+	} while (0)
+#define TLV_PUT_BTRFS_TIMESPEC(sctx, attrtype, eb, ts) \
+	do { \
+		ret = tlv_put_btrfs_timespec(sctx, attrtype, eb, ts); \
+		if (ret < 0) \
+			goto tlv_put_failure; \
+	} while (0)
+
+static int send_header(struct send_ctx *sctx)
+{
+	struct btrfs_stream_header hdr;
+
+	strcpy(hdr.magic, BTRFS_SEND_STREAM_MAGIC);
+	hdr.version = cpu_to_le32(BTRFS_SEND_STREAM_VERSION);
+
+	return write_buf(sctx->send_filp, &hdr, sizeof(hdr),
+					&sctx->send_off);
+}
+
+/*
+ * For each command/item we want to send to userspace, we call this function.
+ */
+static int begin_cmd(struct send_ctx *sctx, int cmd)
+{
+	struct btrfs_cmd_header *hdr;
+
+	if (WARN_ON(!sctx->send_buf))
+		return -EINVAL;
+
+	BUG_ON(sctx->send_size);
+
+	sctx->send_size += sizeof(*hdr);
+	hdr = (struct btrfs_cmd_header *)sctx->send_buf;
+	hdr->cmd = cpu_to_le16(cmd);
+
+	return 0;
+}
+
+static int send_cmd(struct send_ctx *sctx)
+{
+	int ret;
+	struct btrfs_cmd_header *hdr;
+	u32 crc;
+
+	hdr = (struct btrfs_cmd_header *)sctx->send_buf;
+	hdr->len = cpu_to_le32(sctx->send_size - sizeof(*hdr));
+	hdr->crc = 0;
+
+	crc = btrfs_crc32c(0, (unsigned char *)sctx->send_buf, sctx->send_size);
+	hdr->crc = cpu_to_le32(crc);
+
+	ret = write_buf(sctx->send_filp, sctx->send_buf, sctx->send_size,
+					&sctx->send_off);
+
+	sctx->total_send_size += sctx->send_size;
+	sctx->cmd_send_size[le16_to_cpu(hdr->cmd)] += sctx->send_size;
+	sctx->send_size = 0;
+
+	return ret;
+}
+
+/*
+ * Sends a move instruction to user space
+ */
+static int send_rename(struct send_ctx *sctx,
+		     struct fs_path *from, struct fs_path *to)
+{
+	int ret;
+
+verbose_printk("btrfs: send_rename %s -> %s\n", from->start, to->start);
+
+	ret = begin_cmd(sctx, BTRFS_SEND_C_RENAME);
+	if (ret < 0)
+		goto out;
+
+	TLV_PUT_PATH(sctx, BTRFS_SEND_A_PATH, from);
+	TLV_PUT_PATH(sctx, BTRFS_SEND_A_PATH_TO, to);
+
+	ret = send_cmd(sctx);
+
+tlv_put_failure:
+out:
+	return ret;
+}
+
+/*
+ * Sends a link instruction to user space
+ */
+static int send_link(struct send_ctx *sctx,
+		     struct fs_path *path, struct fs_path *lnk)
+{
+	int ret;
+
+verbose_printk("btrfs: send_link %s -> %s\n", path->start, lnk->start);
+
+	ret = begin_cmd(sctx, BTRFS_SEND_C_LINK);
+	if (ret < 0)
+		goto out;
+
+	TLV_PUT_PATH(sctx, BTRFS_SEND_A_PATH, path);
+	TLV_PUT_PATH(sctx, BTRFS_SEND_A_PATH_LINK, lnk);
+
+	ret = send_cmd(sctx);
+
+tlv_put_failure:
+out:
+	return ret;
+}
+
+/*
+ * Sends an unlink instruction to user space
+ */
+static int send_unlink(struct send_ctx *sctx, struct fs_path *path)
+{
+	int ret;
+
+verbose_printk("btrfs: send_unlink %s\n", path->start);
+
+	ret = begin_cmd(sctx, BTRFS_SEND_C_UNLINK);
+	if (ret < 0)
+		goto out;
+
+	TLV_PUT_PATH(sctx, BTRFS_SEND_A_PATH, path);
+
+	ret = send_cmd(sctx);
+
+tlv_put_failure:
+out:
+	return ret;
+}
+
+/*
+ * Sends a rmdir instruction to user space
+ */
+static int send_rmdir(struct send_ctx *sctx, struct fs_path *path)
+{
+	int ret;
+
+verbose_printk("btrfs: send_rmdir %s\n", path->start);
+
+	ret = begin_cmd(sctx, BTRFS_SEND_C_RMDIR);
+	if (ret < 0)
+		goto out;
+
+	TLV_PUT_PATH(sctx, BTRFS_SEND_A_PATH, path);
+
+	ret = send_cmd(sctx);
+
+tlv_put_failure:
+out:
+	return ret;
+}
+
+/*
+ * Helper function to retrieve some fields from an inode item.
+ */
+static int __get_inode_info(struct btrfs_root *root, struct btrfs_path *path,
+			  u64 ino, u64 *size, u64 *gen, u64 *mode, u64 *uid,
+			  u64 *gid, u64 *rdev)
+{
+	int ret;
+	struct btrfs_inode_item *ii;
+	struct btrfs_key key;
+
+	key.objectid = ino;
+	key.type = BTRFS_INODE_ITEM_KEY;
+	key.offset = 0;
+	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+	if (ret) {
+		if (ret > 0)
+			ret = -ENOENT;
+		return ret;
+	}
+
+	ii = btrfs_item_ptr(path->nodes[0], path->slots[0],
+			struct btrfs_inode_item);
+	if (size)
+		*size = btrfs_inode_size(path->nodes[0], ii);
+	if (gen)
+		*gen = btrfs_inode_generation(path->nodes[0], ii);
+	if (mode)
+		*mode = btrfs_inode_mode(path->nodes[0], ii);
+	if (uid)
+		*uid = btrfs_inode_uid(path->nodes[0], ii);
+	if (gid)
+		*gid = btrfs_inode_gid(path->nodes[0], ii);
+	if (rdev)
+		*rdev = btrfs_inode_rdev(path->nodes[0], ii);
+
+	return ret;
+}
+
+static int get_inode_info(struct btrfs_root *root,
+			  u64 ino, u64 *size, u64 *gen,
+			  u64 *mode, u64 *uid, u64 *gid,
+			  u64 *rdev)
+{
+	struct btrfs_path *path;
+	int ret;
+
+	path = alloc_path_for_send();
+	if (!path)
+		return -ENOMEM;
+	ret = __get_inode_info(root, path, ino, size, gen, mode, uid, gid,
+			       rdev);
+	btrfs_free_path(path);
+	return ret;
+}
+
+typedef int (*iterate_inode_ref_t)(int num, u64 dir, int index,
+				   struct fs_path *p,
+				   void *ctx);
+
+/*
+ * Helper function to iterate the entries in ONE btrfs_inode_ref or
+ * btrfs_inode_extref.
+ * The iterate callback may return a non zero value to stop iteration. This can
+ * be a negative value for error codes or 1 to simply stop it.
+ *
+ * path must point to the INODE_REF or INODE_EXTREF when called.
+ */
+static int iterate_inode_ref(struct btrfs_root *root, struct btrfs_path *path,
+			     struct btrfs_key *found_key, int resolve,
+			     iterate_inode_ref_t iterate, void *ctx)
+{
+	struct extent_buffer *eb = path->nodes[0];
+	struct btrfs_item *item;
+	struct btrfs_inode_ref *iref;
+	struct btrfs_inode_extref *extref;
+	struct btrfs_path *tmp_path;
+	struct fs_path *p;
+	u32 cur = 0;
+	u32 total;
+	int slot = path->slots[0];
+	u32 name_len;
+	char *start;
+	int ret = 0;
+	int num = 0;
+	int index;
+	u64 dir;
+	unsigned long name_off;
+	unsigned long elem_size;
+	unsigned long ptr;
+
+	p = fs_path_alloc_reversed();
+	if (!p)
+		return -ENOMEM;
+
+	tmp_path = alloc_path_for_send();
+	if (!tmp_path) {
+		fs_path_free(p);
+		return -ENOMEM;
+	}
+
+
+	if (found_key->type == BTRFS_INODE_REF_KEY) {
+		ptr = (unsigned long)btrfs_item_ptr(eb, slot,
+						    struct btrfs_inode_ref);
+		item = btrfs_item_nr(slot);
+		total = btrfs_item_size(eb, item);
+		elem_size = sizeof(*iref);
+	} else {
+		ptr = btrfs_item_ptr_offset(eb, slot);
+		total = btrfs_item_size_nr(eb, slot);
+		elem_size = sizeof(*extref);
+	}
+
+	while (cur < total) {
+		fs_path_reset(p);
+
+		if (found_key->type == BTRFS_INODE_REF_KEY) {
+			iref = (struct btrfs_inode_ref *)(ptr + cur);
+			name_len = btrfs_inode_ref_name_len(eb, iref);
+			name_off = (unsigned long)(iref + 1);
+			index = btrfs_inode_ref_index(eb, iref);
+			dir = found_key->offset;
+		} else {
+			extref = (struct btrfs_inode_extref *)(ptr + cur);
+			name_len = btrfs_inode_extref_name_len(eb, extref);
+			name_off = (unsigned long)&extref->name;
+			index = btrfs_inode_extref_index(eb, extref);
+			dir = btrfs_inode_extref_parent(eb, extref);
+		}
+
+		if (resolve) {
+			start = btrfs_ref_to_path(root, tmp_path, name_len,
+						  name_off, eb, dir,
+						  p->buf, p->buf_len);
+			if (IS_ERR(start)) {
+				ret = PTR_ERR(start);
+				goto out;
+			}
+			if (start < p->buf) {
+				/* overflow , try again with larger buffer */
+				ret = fs_path_ensure_buf(p,
+						p->buf_len + p->buf - start);
+				if (ret < 0)
+					goto out;
+				start = btrfs_ref_to_path(root, tmp_path,
+							  name_len, name_off,
+							  eb, dir,
+							  p->buf, p->buf_len);
+				if (IS_ERR(start)) {
+					ret = PTR_ERR(start);
+					goto out;
+				}
+				BUG_ON(start < p->buf);
+			}
+			p->start = start;
+		} else {
+			ret = fs_path_add_from_extent_buffer(p, eb, name_off,
+							     name_len);
+			if (ret < 0)
+				goto out;
+		}
+
+		cur += elem_size + name_len;
+		ret = iterate(num, dir, index, p, ctx);
+		if (ret)
+			goto out;
+		num++;
+	}
+
+out:
+	btrfs_free_path(tmp_path);
+	fs_path_free(p);
+	return ret;
+}
+
+typedef int (*iterate_dir_item_t)(int num, struct btrfs_key *di_key,
+				  const char *name, int name_len,
+				  const char *data, int data_len,
+				  u8 type, void *ctx);
+
+/*
+ * Helper function to iterate the entries in ONE btrfs_dir_item.
+ * The iterate callback may return a non zero value to stop iteration. This can
+ * be a negative value for error codes or 1 to simply stop it.
+ *
+ * path must point to the dir item when called.
+ */
+static int iterate_dir_item(struct btrfs_root *root, struct btrfs_path *path,
+			    struct btrfs_key *found_key,
+			    iterate_dir_item_t iterate, void *ctx)
+{
+	int ret = 0;
+	struct extent_buffer *eb;
+	struct btrfs_item *item;
+	struct btrfs_dir_item *di;
+	struct btrfs_key di_key;
+	char *buf = NULL;
+	int buf_len;
+	u32 name_len;
+	u32 data_len;
+	u32 cur;
+	u32 len;
+	u32 total;
+	int slot;
+	int num;
+	u8 type;
+
+	if (found_key->type == BTRFS_XATTR_ITEM_KEY)
+		buf_len = BTRFS_MAX_XATTR_SIZE(root);
+	else
+		buf_len = PATH_MAX;
+
+	buf = kmalloc(buf_len, GFP_NOFS);
+	if (!buf) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	eb = path->nodes[0];
+	slot = path->slots[0];
+	item = btrfs_item_nr(slot);
+	di = btrfs_item_ptr(eb, slot, struct btrfs_dir_item);
+	cur = 0;
+	len = 0;
+	total = btrfs_item_size(eb, item);
+
+	num = 0;
+	while (cur < total) {
+		name_len = btrfs_dir_name_len(eb, di);
+		data_len = btrfs_dir_data_len(eb, di);
+		type = btrfs_dir_type(eb, di);
+		btrfs_dir_item_key_to_cpu(eb, di, &di_key);
+
+		if (type == BTRFS_FT_XATTR) {
+			if (name_len > XATTR_NAME_MAX) {
+				ret = -ENAMETOOLONG;
+				goto out;
+			}
+			if (name_len + data_len > buf_len) {
+				ret = -E2BIG;
+				goto out;
+			}
+		} else {
+			/*
+			 * Path too long
+			 */
+			if (name_len + data_len > buf_len) {
+				ret = -ENAMETOOLONG;
+				goto out;
+			}
+		}
+
+		read_extent_buffer(eb, buf, (unsigned long)(di + 1),
+				name_len + data_len);
+
+		len = sizeof(*di) + name_len + data_len;
+		di = (struct btrfs_dir_item *)((char *)di + len);
+		cur += len;
+
+		ret = iterate(num, &di_key, buf, name_len, buf + name_len,
+				data_len, type, ctx);
+		if (ret < 0)
+			goto out;
+		if (ret) {
+			ret = 0;
+			goto out;
+		}
+
+		num++;
+	}
+
+out:
+	kfree(buf);
+	return ret;
+}
+
+static int __copy_first_ref(int num, u64 dir, int index,
+			    struct fs_path *p, void *ctx)
+{
+	int ret;
+	struct fs_path *pt = ctx;
+
+	ret = fs_path_copy(pt, p);
+	if (ret < 0)
+		return ret;
+
+	/* we want the first only */
+	return 1;
+}
+
+/*
+ * Retrieve the first path of an inode. If an inode has more then one
+ * ref/hardlink, this is ignored.
+ */
+static int get_inode_path(struct btrfs_root *root,
+			  u64 ino, struct fs_path *path)
+{
+	int ret;
+	struct btrfs_key key, found_key;
+	struct btrfs_path *p;
+
+	p = alloc_path_for_send();
+	if (!p)
+		return -ENOMEM;
+
+	fs_path_reset(path);
+
+	key.objectid = ino;
+	key.type = BTRFS_INODE_REF_KEY;
+	key.offset = 0;
+
+	ret = btrfs_search_slot_for_read(root, &key, p, 1, 0);
+	if (ret < 0)
+		goto out;
+	if (ret) {
+		ret = 1;
+		goto out;
+	}
+	btrfs_item_key_to_cpu(p->nodes[0], &found_key, p->slots[0]);
+	if (found_key.objectid != ino ||
+	    (found_key.type != BTRFS_INODE_REF_KEY &&
+	     found_key.type != BTRFS_INODE_EXTREF_KEY)) {
+		ret = -ENOENT;
+		goto out;
+	}
+
+	ret = iterate_inode_ref(root, p, &found_key, 1,
+				__copy_first_ref, path);
+	if (ret < 0)
+		goto out;
+	ret = 0;
+
+out:
+	btrfs_free_path(p);
+	return ret;
+}
+
+struct backref_ctx {
+	struct send_ctx *sctx;
+
+	struct btrfs_path *path;
+	/* number of total found references */
+	u64 found;
+
+	/*
+	 * used for clones found in send_root. clones found behind cur_objectid
+	 * and cur_offset are not considered as allowed clones.
+	 */
+	u64 cur_objectid;
+	u64 cur_offset;
+
+	/* may be truncated in case it's the last extent in a file */
+	u64 extent_len;
+
+	/* Just to check for bugs in backref resolving */
+	int found_itself;
+};
+
+static int __clone_root_cmp_bsearch(const void *key, const void *elt)
+{
+	u64 root = (u64)(uintptr_t)key;
+	struct clone_root *cr = (struct clone_root *)elt;
+
+	if (root < cr->root->objectid)
+		return -1;
+	if (root > cr->root->objectid)
+		return 1;
+	return 0;
+}
+
+static int __clone_root_cmp_sort(const void *e1, const void *e2)
+{
+	struct clone_root *cr1 = (struct clone_root *)e1;
+	struct clone_root *cr2 = (struct clone_root *)e2;
+
+	if (cr1->root->objectid < cr2->root->objectid)
+		return -1;
+	if (cr1->root->objectid > cr2->root->objectid)
+		return 1;
+	return 0;
+}
+
+/*
+ * Called for every backref that is found for the current extent.
+ * Results are collected in sctx->clone_roots->ino/offset/found_refs
+ */
+static int __iterate_backrefs(u64 ino, u64 offset, u64 root, void *ctx_)
+{
+	struct backref_ctx *bctx = ctx_;
+	struct clone_root *found;
+	int ret;
+	u64 i_size;
+
+	/* First check if the root is in the list of accepted clone sources */
+	found = bsearch((void *)(uintptr_t)root, bctx->sctx->clone_roots,
+			bctx->sctx->clone_roots_cnt,
+			sizeof(struct clone_root),
+			__clone_root_cmp_bsearch);
+	if (!found)
+		return 0;
+
+	if (found->root == bctx->sctx->send_root &&
+	    ino == bctx->cur_objectid &&
+	    offset == bctx->cur_offset) {
+		bctx->found_itself = 1;
+	}
+
+	/*
+	 * There are inodes that have extents that lie behind its i_size. Don't
+	 * accept clones from these extents.
+	 */
+	ret = __get_inode_info(found->root, bctx->path, ino, &i_size, NULL, NULL,
+			       NULL, NULL, NULL);
+	btrfs_release_path(bctx->path);
+	if (ret < 0)
+		return ret;
+
+	if (offset + bctx->extent_len > i_size)
+		return 0;
+
+	/*
+	 * Make sure we don't consider clones from send_root that are
+	 * behind the current inode/offset.
+	 */
+	if (found->root == bctx->sctx->send_root) {
+		/*
+		 * TODO for the moment we don't accept clones from the inode
+		 * that is currently send. We may change this when
+		 * BTRFS_IOC_CLONE_RANGE supports cloning from and to the same
+		 * file.
+		 */
+		if (ino >= bctx->cur_objectid)
+			return 0;
+#if 0
+		if (ino > bctx->cur_objectid)
+			return 0;
+		if (offset + bctx->extent_len > bctx->cur_offset)
+			return 0;
+#endif
+	}
+
+	bctx->found++;
+	found->found_refs++;
+	if (ino < found->ino) {
+		found->ino = ino;
+		found->offset = offset;
+	} else if (found->ino == ino) {
+		/*
+		 * same extent found more then once in the same file.
+		 */
+		if (found->offset > offset + bctx->extent_len)
+			found->offset = offset;
+	}
+
+	return 0;
+}
+
+/*
+ * Given an inode, offset and extent item, it finds a good clone for a clone
+ * instruction. Returns -ENOENT when none could be found. The function makes
+ * sure that the returned clone is usable at the point where sending is at the
+ * moment. This means, that no clones are accepted which lie behind the current
+ * inode+offset.
+ *
+ * path must point to the extent item when called.
+ */
+static int find_extent_clone(struct send_ctx *sctx,
+			     struct btrfs_path *path,
+			     u64 ino, u64 data_offset,
+			     u64 ino_size,
+			     struct clone_root **found)
+{
+	int ret;
+	int extent_type;
+	u64 logical;
+	u64 disk_byte;
+	u64 num_bytes;
+	u64 extent_item_pos;
+	u64 flags = 0;
+	struct btrfs_file_extent_item *fi;
+	struct extent_buffer *eb = path->nodes[0];
+	struct backref_ctx *backref_ctx = NULL;
+	struct clone_root *cur_clone_root;
+	struct btrfs_key found_key;
+	struct btrfs_path *tmp_path;
+	int compressed;
+	u32 i;
+
+	tmp_path = alloc_path_for_send();
+	if (!tmp_path)
+		return -ENOMEM;
+
+	/* We only use this path under the commit sem */
+	tmp_path->need_commit_sem = 0;
+
+	backref_ctx = kmalloc(sizeof(*backref_ctx), GFP_NOFS);
+	if (!backref_ctx) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	backref_ctx->path = tmp_path;
+
+	if (data_offset >= ino_size) {
+		/*
+		 * There may be extents that lie behind the file's size.
+		 * I at least had this in combination with snapshotting while
+		 * writing large files.
+		 */
+		ret = 0;
+		goto out;
+	}
+
+	fi = btrfs_item_ptr(eb, path->slots[0],
+			struct btrfs_file_extent_item);
+	extent_type = btrfs_file_extent_type(eb, fi);
+	if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
+		ret = -ENOENT;
+		goto out;
+	}
+	compressed = btrfs_file_extent_compression(eb, fi);
+
+	num_bytes = btrfs_file_extent_num_bytes(eb, fi);
+	disk_byte = btrfs_file_extent_disk_bytenr(eb, fi);
+	if (disk_byte == 0) {
+		ret = -ENOENT;
+		goto out;
+	}
+	logical = disk_byte + btrfs_file_extent_offset(eb, fi);
+
+	down_read(&sctx->send_root->fs_info->commit_root_sem);
+	ret = extent_from_logical(sctx->send_root->fs_info, disk_byte, tmp_path,
+				  &found_key, &flags);
+	up_read(&sctx->send_root->fs_info->commit_root_sem);
+	btrfs_release_path(tmp_path);
+
+	if (ret < 0)
+		goto out;
+	if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
+		ret = -EIO;
+		goto out;
+	}
+
+	/*
+	 * Setup the clone roots.
+	 */
+	for (i = 0; i < sctx->clone_roots_cnt; i++) {
+		cur_clone_root = sctx->clone_roots + i;
+		cur_clone_root->ino = (u64)-1;
+		cur_clone_root->offset = 0;
+		cur_clone_root->found_refs = 0;
+	}
+
+	backref_ctx->sctx = sctx;
+	backref_ctx->found = 0;
+	backref_ctx->cur_objectid = ino;
+	backref_ctx->cur_offset = data_offset;
+	backref_ctx->found_itself = 0;
+	backref_ctx->extent_len = num_bytes;
+
+	/*
+	 * The last extent of a file may be too large due to page alignment.
+	 * We need to adjust extent_len in this case so that the checks in
+	 * __iterate_backrefs work.
+	 */
+	if (data_offset + num_bytes >= ino_size)
+		backref_ctx->extent_len = ino_size - data_offset;
+
+	/*
+	 * Now collect all backrefs.
+	 */
+	if (compressed == BTRFS_COMPRESS_NONE)
+		extent_item_pos = logical - found_key.objectid;
+	else
+		extent_item_pos = 0;
+	ret = iterate_extent_inodes(sctx->send_root->fs_info,
+					found_key.objectid, extent_item_pos, 1,
+					__iterate_backrefs, backref_ctx);
+
+	if (ret < 0)
+		goto out;
+
+	if (!backref_ctx->found_itself) {
+		/* found a bug in backref code? */
+		ret = -EIO;
+		btrfs_err(sctx->send_root->fs_info, "did not find backref in "
+				"send_root. inode=%llu, offset=%llu, "
+				"disk_byte=%llu found extent=%llu",
+				ino, data_offset, disk_byte, found_key.objectid);
+		goto out;
+	}
+
+verbose_printk(KERN_DEBUG "btrfs: find_extent_clone: data_offset=%llu, "
+		"ino=%llu, "
+		"num_bytes=%llu, logical=%llu\n",
+		data_offset, ino, num_bytes, logical);
+
+	if (!backref_ctx->found)
+		verbose_printk("btrfs:    no clones found\n");
+
+	cur_clone_root = NULL;
+	for (i = 0; i < sctx->clone_roots_cnt; i++) {
+		if (sctx->clone_roots[i].found_refs) {
+			if (!cur_clone_root)
+				cur_clone_root = sctx->clone_roots + i;
+			else if (sctx->clone_roots[i].root == sctx->send_root)
+				/* prefer clones from send_root over others */
+				cur_clone_root = sctx->clone_roots + i;
+		}
+
+	}
+
+	if (cur_clone_root) {
+		if (compressed != BTRFS_COMPRESS_NONE) {
+			/*
+			 * Offsets given by iterate_extent_inodes() are relative
+			 * to the start of the extent, we need to add logical
+			 * offset from the file extent item.
+			 * (See why at backref.c:check_extent_in_eb())
+			 */
+			cur_clone_root->offset += btrfs_file_extent_offset(eb,
+									   fi);
+		}
+		*found = cur_clone_root;
+		ret = 0;
+	} else {
+		ret = -ENOENT;
+	}
+
+out:
+	btrfs_free_path(tmp_path);
+	kfree(backref_ctx);
+	return ret;
+}
+
+static int read_symlink(struct btrfs_root *root,
+			u64 ino,
+			struct fs_path *dest)
+{
+	int ret;
+	struct btrfs_path *path;
+	struct btrfs_key key;
+	struct btrfs_file_extent_item *ei;
+	u8 type;
+	u8 compression;
+	unsigned long off;
+	int len;
+
+	path = alloc_path_for_send();
+	if (!path)
+		return -ENOMEM;
+
+	key.objectid = ino;
+	key.type = BTRFS_EXTENT_DATA_KEY;
+	key.offset = 0;
+	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+	if (ret < 0)
+		goto out;
+	BUG_ON(ret);
+
+	ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
+			struct btrfs_file_extent_item);
+	type = btrfs_file_extent_type(path->nodes[0], ei);
+	compression = btrfs_file_extent_compression(path->nodes[0], ei);
+	BUG_ON(type != BTRFS_FILE_EXTENT_INLINE);
+	BUG_ON(compression);
+
+	off = btrfs_file_extent_inline_start(ei);
+	len = btrfs_file_extent_inline_len(path->nodes[0], path->slots[0], ei);
+
+	ret = fs_path_add_from_extent_buffer(dest, path->nodes[0], off, len);
+
+out:
+	btrfs_free_path(path);
+	return ret;
+}
+
+/*
+ * Helper function to generate a file name that is unique in the root of
+ * send_root and parent_root. This is used to generate names for orphan inodes.
+ */
+static int gen_unique_name(struct send_ctx *sctx,
+			   u64 ino, u64 gen,
+			   struct fs_path *dest)
+{
+	int ret = 0;
+	struct btrfs_path *path;
+	struct btrfs_dir_item *di;
+	char tmp[64];
+	int len;
+	u64 idx = 0;
+
+	path = alloc_path_for_send();
+	if (!path)
+		return -ENOMEM;
+
+	while (1) {
+		len = snprintf(tmp, sizeof(tmp), "o%llu-%llu-%llu",
+				ino, gen, idx);
+		ASSERT(len < sizeof(tmp));
+
+		di = btrfs_lookup_dir_item(NULL, sctx->send_root,
+				path, BTRFS_FIRST_FREE_OBJECTID,
+				tmp, strlen(tmp), 0);
+		btrfs_release_path(path);
+		if (IS_ERR(di)) {
+			ret = PTR_ERR(di);
+			goto out;
+		}
+		if (di) {
+			/* not unique, try again */
+			idx++;
+			continue;
+		}
+
+		if (!sctx->parent_root) {
+			/* unique */
+			ret = 0;
+			break;
+		}
+
+		di = btrfs_lookup_dir_item(NULL, sctx->parent_root,
+				path, BTRFS_FIRST_FREE_OBJECTID,
+				tmp, strlen(tmp), 0);
+		btrfs_release_path(path);
+		if (IS_ERR(di)) {
+			ret = PTR_ERR(di);
+			goto out;
+		}
+		if (di) {
+			/* not unique, try again */
+			idx++;
+			continue;
+		}
+		/* unique */
+		break;
+	}
+
+	ret = fs_path_add(dest, tmp, strlen(tmp));
+
+out:
+	btrfs_free_path(path);
+	return ret;
+}
+
+enum inode_state {
+	inode_state_no_change,
+	inode_state_will_create,
+	inode_state_did_create,
+	inode_state_will_delete,
+	inode_state_did_delete,
+};
+
+static int get_cur_inode_state(struct send_ctx *sctx, u64 ino, u64 gen)
+{
+	int ret;
+	int left_ret;
+	int right_ret;
+	u64 left_gen;
+	u64 right_gen;
+
+	ret = get_inode_info(sctx->send_root, ino, NULL, &left_gen, NULL, NULL,
+			NULL, NULL);
+	if (ret < 0 && ret != -ENOENT)
+		goto out;
+	left_ret = ret;
+
+	if (!sctx->parent_root) {
+		right_ret = -ENOENT;
+	} else {
+		ret = get_inode_info(sctx->parent_root, ino, NULL, &right_gen,
+				NULL, NULL, NULL, NULL);
+		if (ret < 0 && ret != -ENOENT)
+			goto out;
+		right_ret = ret;
+	}
+
+	if (!left_ret && !right_ret) {
+		if (left_gen == gen && right_gen == gen) {
+			ret = inode_state_no_change;
+		} else if (left_gen == gen) {
+			if (ino < sctx->send_progress)
+				ret = inode_state_did_create;
+			else
+				ret = inode_state_will_create;
+		} else if (right_gen == gen) {
+			if (ino < sctx->send_progress)
+				ret = inode_state_did_delete;
+			else
+				ret = inode_state_will_delete;
+		} else  {
+			ret = -ENOENT;
+		}
+	} else if (!left_ret) {
+		if (left_gen == gen) {
+			if (ino < sctx->send_progress)
+				ret = inode_state_did_create;
+			else
+				ret = inode_state_will_create;
+		} else {
+			ret = -ENOENT;
+		}
+	} else if (!right_ret) {
+		if (right_gen == gen) {
+			if (ino < sctx->send_progress)
+				ret = inode_state_did_delete;
+			else
+				ret = inode_state_will_delete;
+		} else {
+			ret = -ENOENT;
+		}
+	} else {
+		ret = -ENOENT;
+	}
+
+out:
+	return ret;
+}
+
+static int is_inode_existent(struct send_ctx *sctx, u64 ino, u64 gen)
+{
+	int ret;
+
+	ret = get_cur_inode_state(sctx, ino, gen);
+	if (ret < 0)
+		goto out;
+
+	if (ret == inode_state_no_change ||
+	    ret == inode_state_did_create ||
+	    ret == inode_state_will_delete)
+		ret = 1;
+	else
+		ret = 0;
+
+out:
+	return ret;
+}
+
+/*
+ * Helper function to lookup a dir item in a dir.
+ */
+static int lookup_dir_item_inode(struct btrfs_root *root,
+				 u64 dir, const char *name, int name_len,
+				 u64 *found_inode,
+				 u8 *found_type)
+{
+	int ret = 0;
+	struct btrfs_dir_item *di;
+	struct btrfs_key key;
+	struct btrfs_path *path;
+
+	path = alloc_path_for_send();
+	if (!path)
+		return -ENOMEM;
+
+	di = btrfs_lookup_dir_item(NULL, root, path,
+			dir, name, name_len, 0);
+	if (!di) {
+		ret = -ENOENT;
+		goto out;
+	}
+	if (IS_ERR(di)) {
+		ret = PTR_ERR(di);
+		goto out;
+	}
+	btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
+	if (key.type == BTRFS_ROOT_ITEM_KEY) {
+		ret = -ENOENT;
+		goto out;
+	}
+	*found_inode = key.objectid;
+	*found_type = btrfs_dir_type(path->nodes[0], di);
+
+out:
+	btrfs_free_path(path);
+	return ret;
+}
+
+/*
+ * Looks up the first btrfs_inode_ref of a given ino. It returns the parent dir,
+ * generation of the parent dir and the name of the dir entry.
+ */
+static int get_first_ref(struct btrfs_root *root, u64 ino,
+			 u64 *dir, u64 *dir_gen, struct fs_path *name)
+{
+	int ret;
+	struct btrfs_key key;
+	struct btrfs_key found_key;
+	struct btrfs_path *path;
+	int len;
+	u64 parent_dir;
+
+	path = alloc_path_for_send();
+	if (!path)
+		return -ENOMEM;
+
+	key.objectid = ino;
+	key.type = BTRFS_INODE_REF_KEY;
+	key.offset = 0;
+
+	ret = btrfs_search_slot_for_read(root, &key, path, 1, 0);
+	if (ret < 0)
+		goto out;
+	if (!ret)
+		btrfs_item_key_to_cpu(path->nodes[0], &found_key,
+				path->slots[0]);
+	if (ret || found_key.objectid != ino ||
+	    (found_key.type != BTRFS_INODE_REF_KEY &&
+	     found_key.type != BTRFS_INODE_EXTREF_KEY)) {
+		ret = -ENOENT;
+		goto out;
+	}
+
+	if (found_key.type == BTRFS_INODE_REF_KEY) {
+		struct btrfs_inode_ref *iref;
+		iref = btrfs_item_ptr(path->nodes[0], path->slots[0],
+				      struct btrfs_inode_ref);
+		len = btrfs_inode_ref_name_len(path->nodes[0], iref);
+		ret = fs_path_add_from_extent_buffer(name, path->nodes[0],
+						     (unsigned long)(iref + 1),
+						     len);
+		parent_dir = found_key.offset;
+	} else {
+		struct btrfs_inode_extref *extref;
+		extref = btrfs_item_ptr(path->nodes[0], path->slots[0],
+					struct btrfs_inode_extref);
+		len = btrfs_inode_extref_name_len(path->nodes[0], extref);
+		ret = fs_path_add_from_extent_buffer(name, path->nodes[0],
+					(unsigned long)&extref->name, len);
+		parent_dir = btrfs_inode_extref_parent(path->nodes[0], extref);
+	}
+	if (ret < 0)
+		goto out;
+	btrfs_release_path(path);
+
+	if (dir_gen) {
+		ret = get_inode_info(root, parent_dir, NULL, dir_gen, NULL,
+				     NULL, NULL, NULL);
+		if (ret < 0)
+			goto out;
+	}
+
+	*dir = parent_dir;
+
+out:
+	btrfs_free_path(path);
+	return ret;
+}
+
+static int is_first_ref(struct btrfs_root *root,
+			u64 ino, u64 dir,
+			const char *name, int name_len)
+{
+	int ret;
+	struct fs_path *tmp_name;
+	u64 tmp_dir;
+
+	tmp_name = fs_path_alloc();
+	if (!tmp_name)
+		return -ENOMEM;
+
+	ret = get_first_ref(root, ino, &tmp_dir, NULL, tmp_name);
+	if (ret < 0)
+		goto out;
+
+	if (dir != tmp_dir || name_len != fs_path_len(tmp_name)) {
+		ret = 0;
+		goto out;
+	}
+
+	ret = !memcmp(tmp_name->start, name, name_len);
+
+out:
+	fs_path_free(tmp_name);
+	return ret;
+}
+
+/*
+ * Used by process_recorded_refs to determine if a new ref would overwrite an
+ * already existing ref. In case it detects an overwrite, it returns the
+ * inode/gen in who_ino/who_gen.
+ * When an overwrite is detected, process_recorded_refs does proper orphanizing
+ * to make sure later references to the overwritten inode are possible.
+ * Orphanizing is however only required for the first ref of an inode.
+ * process_recorded_refs does an additional is_first_ref check to see if
+ * orphanizing is really required.
+ */
+static int will_overwrite_ref(struct send_ctx *sctx, u64 dir, u64 dir_gen,
+			      const char *name, int name_len,
+			      u64 *who_ino, u64 *who_gen)
+{
+	int ret = 0;
+	u64 gen;
+	u64 other_inode = 0;
+	u8 other_type = 0;
+
+	if (!sctx->parent_root)
+		goto out;
+
+	ret = is_inode_existent(sctx, dir, dir_gen);
+	if (ret <= 0)
+		goto out;
+
+	/*
+	 * If we have a parent root we need to verify that the parent dir was
+	 * not delted and then re-created, if it was then we have no overwrite
+	 * and we can just unlink this entry.
+	 */
+	if (sctx->parent_root) {
+		ret = get_inode_info(sctx->parent_root, dir, NULL, &gen, NULL,
+				     NULL, NULL, NULL);
+		if (ret < 0 && ret != -ENOENT)
+			goto out;
+		if (ret) {
+			ret = 0;
+			goto out;
+		}
+		if (gen != dir_gen)
+			goto out;
+	}
+
+	ret = lookup_dir_item_inode(sctx->parent_root, dir, name, name_len,
+			&other_inode, &other_type);
+	if (ret < 0 && ret != -ENOENT)
+		goto out;
+	if (ret) {
+		ret = 0;
+		goto out;
+	}
+
+	/*
+	 * Check if the overwritten ref was already processed. If yes, the ref
+	 * was already unlinked/moved, so we can safely assume that we will not
+	 * overwrite anything at this point in time.
+	 */
+	if (other_inode > sctx->send_progress) {
+		ret = get_inode_info(sctx->parent_root, other_inode, NULL,
+				who_gen, NULL, NULL, NULL, NULL);
+		if (ret < 0)
+			goto out;
+
+		ret = 1;
+		*who_ino = other_inode;
+	} else {
+		ret = 0;
+	}
+
+out:
+	return ret;
+}
+
+/*
+ * Checks if the ref was overwritten by an already processed inode. This is
+ * used by __get_cur_name_and_parent to find out if the ref was orphanized and
+ * thus the orphan name needs be used.
+ * process_recorded_refs also uses it to avoid unlinking of refs that were
+ * overwritten.
+ */
+static int did_overwrite_ref(struct send_ctx *sctx,
+			    u64 dir, u64 dir_gen,
+			    u64 ino, u64 ino_gen,
+			    const char *name, int name_len)
+{
+	int ret = 0;
+	u64 gen;
+	u64 ow_inode;
+	u8 other_type;
+
+	if (!sctx->parent_root)
+		goto out;
+
+	ret = is_inode_existent(sctx, dir, dir_gen);
+	if (ret <= 0)
+		goto out;
+
+	/* check if the ref was overwritten by another ref */
+	ret = lookup_dir_item_inode(sctx->send_root, dir, name, name_len,
+			&ow_inode, &other_type);
+	if (ret < 0 && ret != -ENOENT)
+		goto out;
+	if (ret) {
+		/* was never and will never be overwritten */
+		ret = 0;
+		goto out;
+	}
+
+	ret = get_inode_info(sctx->send_root, ow_inode, NULL, &gen, NULL, NULL,
+			NULL, NULL);
+	if (ret < 0)
+		goto out;
+
+	if (ow_inode == ino && gen == ino_gen) {
+		ret = 0;
+		goto out;
+	}
+
+	/* we know that it is or will be overwritten. check this now */
+	if (ow_inode < sctx->send_progress)
+		ret = 1;
+	else
+		ret = 0;
+
+out:
+	return ret;
+}
+
+/*
+ * Same as did_overwrite_ref, but also checks if it is the first ref of an inode
+ * that got overwritten. This is used by process_recorded_refs to determine
+ * if it has to use the path as returned by get_cur_path or the orphan name.
+ */
+static int did_overwrite_first_ref(struct send_ctx *sctx, u64 ino, u64 gen)
+{
+	int ret = 0;
+	struct fs_path *name = NULL;
+	u64 dir;
+	u64 dir_gen;
+
+	if (!sctx->parent_root)
+		goto out;
+
+	name = fs_path_alloc();
+	if (!name)
+		return -ENOMEM;
+
+	ret = get_first_ref(sctx->parent_root, ino, &dir, &dir_gen, name);
+	if (ret < 0)
+		goto out;
+
+	ret = did_overwrite_ref(sctx, dir, dir_gen, ino, gen,
+			name->start, fs_path_len(name));
+
+out:
+	fs_path_free(name);
+	return ret;
+}
+
+/*
+ * Insert a name cache entry. On 32bit kernels the radix tree index is 32bit,
+ * so we need to do some special handling in case we have clashes. This function
+ * takes care of this with the help of name_cache_entry::radix_list.
+ * In case of error, nce is kfreed.
+ */
+static int name_cache_insert(struct send_ctx *sctx,
+			     struct name_cache_entry *nce)
+{
+	int ret = 0;
+	struct list_head *nce_head;
+
+	nce_head = radix_tree_lookup(&sctx->name_cache,
+			(unsigned long)nce->ino);
+	if (!nce_head) {
+		nce_head = kmalloc(sizeof(*nce_head), GFP_NOFS);
+		if (!nce_head) {
+			kfree(nce);
+			return -ENOMEM;
+		}
+		INIT_LIST_HEAD(nce_head);
+
+		ret = radix_tree_insert(&sctx->name_cache, nce->ino, nce_head);
+		if (ret < 0) {
+			kfree(nce_head);
+			kfree(nce);
+			return ret;
+		}
+	}
+	list_add_tail(&nce->radix_list, nce_head);
+	list_add_tail(&nce->list, &sctx->name_cache_list);
+	sctx->name_cache_size++;
+
+	return ret;
+}
+
+static void name_cache_delete(struct send_ctx *sctx,
+			      struct name_cache_entry *nce)
+{
+	struct list_head *nce_head;
+
+	nce_head = radix_tree_lookup(&sctx->name_cache,
+			(unsigned long)nce->ino);
+	if (!nce_head) {
+		btrfs_err(sctx->send_root->fs_info,
+	      "name_cache_delete lookup failed ino %llu cache size %d, leaking memory",
+			nce->ino, sctx->name_cache_size);
+	}
+
+	list_del(&nce->radix_list);
+	list_del(&nce->list);
+	sctx->name_cache_size--;
+
+	/*
+	 * We may not get to the final release of nce_head if the lookup fails
+	 */
+	if (nce_head && list_empty(nce_head)) {
+		radix_tree_delete(&sctx->name_cache, (unsigned long)nce->ino);
+		kfree(nce_head);
+	}
+}
+
+static struct name_cache_entry *name_cache_search(struct send_ctx *sctx,
+						    u64 ino, u64 gen)
+{
+	struct list_head *nce_head;
+	struct name_cache_entry *cur;
+
+	nce_head = radix_tree_lookup(&sctx->name_cache, (unsigned long)ino);
+	if (!nce_head)
+		return NULL;
+
+	list_for_each_entry(cur, nce_head, radix_list) {
+		if (cur->ino == ino && cur->gen == gen)
+			return cur;
+	}
+	return NULL;
+}
+
+/*
+ * Removes the entry from the list and adds it back to the end. This marks the
+ * entry as recently used so that name_cache_clean_unused does not remove it.
+ */
+static void name_cache_used(struct send_ctx *sctx, struct name_cache_entry *nce)
+{
+	list_del(&nce->list);
+	list_add_tail(&nce->list, &sctx->name_cache_list);
+}
+
+/*
+ * Remove some entries from the beginning of name_cache_list.
+ */
+static void name_cache_clean_unused(struct send_ctx *sctx)
+{
+	struct name_cache_entry *nce;
+
+	if (sctx->name_cache_size < SEND_CTX_NAME_CACHE_CLEAN_SIZE)
+		return;
+
+	while (sctx->name_cache_size > SEND_CTX_MAX_NAME_CACHE_SIZE) {
+		nce = list_entry(sctx->name_cache_list.next,
+				struct name_cache_entry, list);
+		name_cache_delete(sctx, nce);
+		kfree(nce);
+	}
+}
+
+static void name_cache_free(struct send_ctx *sctx)
+{
+	struct name_cache_entry *nce;
+
+	while (!list_empty(&sctx->name_cache_list)) {
+		nce = list_entry(sctx->name_cache_list.next,
+				struct name_cache_entry, list);
+		name_cache_delete(sctx, nce);
+		kfree(nce);
+	}
+}
+
+/*
+ * Used by get_cur_path for each ref up to the root.
+ * Returns 0 if it succeeded.
+ * Returns 1 if the inode is not existent or got overwritten. In that case, the
+ * name is an orphan name. This instructs get_cur_path to stop iterating. If 1
+ * is returned, parent_ino/parent_gen are not guaranteed to be valid.
+ * Returns <0 in case of error.
+ */
+static int __get_cur_name_and_parent(struct send_ctx *sctx,
+				     u64 ino, u64 gen,
+				     u64 *parent_ino,
+				     u64 *parent_gen,
+				     struct fs_path *dest)
+{
+	int ret;
+	int nce_ret;
+	struct name_cache_entry *nce = NULL;
+
+	/*
+	 * First check if we already did a call to this function with the same
+	 * ino/gen. If yes, check if the cache entry is still up-to-date. If yes
+	 * return the cached result.
+	 */
+	nce = name_cache_search(sctx, ino, gen);
+	if (nce) {
+		if (ino < sctx->send_progress && nce->need_later_update) {
+			name_cache_delete(sctx, nce);
+			kfree(nce);
+			nce = NULL;
+		} else {
+			name_cache_used(sctx, nce);
+			*parent_ino = nce->parent_ino;
+			*parent_gen = nce->parent_gen;
+			ret = fs_path_add(dest, nce->name, nce->name_len);
+			if (ret < 0)
+				goto out;
+			ret = nce->ret;
+			goto out;
+		}
+	}
+
+	/*
+	 * If the inode is not existent yet, add the orphan name and return 1.
+	 * This should only happen for the parent dir that we determine in
+	 * __record_new_ref
+	 */
+	ret = is_inode_existent(sctx, ino, gen);
+	if (ret < 0)
+		goto out;
+
+	if (!ret) {
+		ret = gen_unique_name(sctx, ino, gen, dest);
+		if (ret < 0)
+			goto out;
+		ret = 1;
+		goto out_cache;
+	}
+
+	/*
+	 * Depending on whether the inode was already processed or not, use
+	 * send_root or parent_root for ref lookup.
+	 */
+	if (ino < sctx->send_progress)
+		ret = get_first_ref(sctx->send_root, ino,
+				    parent_ino, parent_gen, dest);
+	else
+		ret = get_first_ref(sctx->parent_root, ino,
+				    parent_ino, parent_gen, dest);
+	if (ret < 0)
+		goto out;
+
+	/*
+	 * Check if the ref was overwritten by an inode's ref that was processed
+	 * earlier. If yes, treat as orphan and return 1.
+	 */
+	ret = did_overwrite_ref(sctx, *parent_ino, *parent_gen, ino, gen,
+			dest->start, dest->end - dest->start);
+	if (ret < 0)
+		goto out;
+	if (ret) {
+		fs_path_reset(dest);
+		ret = gen_unique_name(sctx, ino, gen, dest);
+		if (ret < 0)
+			goto out;
+		ret = 1;
+	}
+
+out_cache:
+	/*
+	 * Store the result of the lookup in the name cache.
+	 */
+	nce = kmalloc(sizeof(*nce) + fs_path_len(dest) + 1, GFP_NOFS);
+	if (!nce) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	nce->ino = ino;
+	nce->gen = gen;
+	nce->parent_ino = *parent_ino;
+	nce->parent_gen = *parent_gen;
+	nce->name_len = fs_path_len(dest);
+	nce->ret = ret;
+	strcpy(nce->name, dest->start);
+
+	if (ino < sctx->send_progress)
+		nce->need_later_update = 0;
+	else
+		nce->need_later_update = 1;
+
+	nce_ret = name_cache_insert(sctx, nce);
+	if (nce_ret < 0)
+		ret = nce_ret;
+	name_cache_clean_unused(sctx);
+
+out:
+	return ret;
+}
+
+/*
+ * Magic happens here. This function returns the first ref to an inode as it
+ * would look like while receiving the stream at this point in time.
+ * We walk the path up to the root. For every inode in between, we check if it
+ * was already processed/sent. If yes, we continue with the parent as found
+ * in send_root. If not, we continue with the parent as found in parent_root.
+ * If we encounter an inode that was deleted at this point in time, we use the
+ * inodes "orphan" name instead of the real name and stop. Same with new inodes
+ * that were not created yet and overwritten inodes/refs.
+ *
+ * When do we have have orphan inodes:
+ * 1. When an inode is freshly created and thus no valid refs are available yet
+ * 2. When a directory lost all it's refs (deleted) but still has dir items
+ *    inside which were not processed yet (pending for move/delete). If anyone
+ *    tried to get the path to the dir items, it would get a path inside that
+ *    orphan directory.
+ * 3. When an inode is moved around or gets new links, it may overwrite the ref
+ *    of an unprocessed inode. If in that case the first ref would be
+ *    overwritten, the overwritten inode gets "orphanized". Later when we
+ *    process this overwritten inode, it is restored at a new place by moving
+ *    the orphan inode.
+ *
+ * sctx->send_progress tells this function at which point in time receiving
+ * would be.
+ */
+static int get_cur_path(struct send_ctx *sctx, u64 ino, u64 gen,
+			struct fs_path *dest)
+{
+	int ret = 0;
+	struct fs_path *name = NULL;
+	u64 parent_inode = 0;
+	u64 parent_gen = 0;
+	int stop = 0;
+
+	name = fs_path_alloc();
+	if (!name) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	dest->reversed = 1;
+	fs_path_reset(dest);
+
+	while (!stop && ino != BTRFS_FIRST_FREE_OBJECTID) {
+		fs_path_reset(name);
+
+		if (is_waiting_for_rm(sctx, ino)) {
+			ret = gen_unique_name(sctx, ino, gen, name);
+			if (ret < 0)
+				goto out;
+			ret = fs_path_add_path(dest, name);
+			break;
+		}
+
+		if (is_waiting_for_move(sctx, ino)) {
+			ret = get_first_ref(sctx->parent_root, ino,
+					    &parent_inode, &parent_gen, name);
+		} else {
+			ret = __get_cur_name_and_parent(sctx, ino, gen,
+							&parent_inode,
+							&parent_gen, name);
+			if (ret)
+				stop = 1;
+		}
+
+		if (ret < 0)
+			goto out;
+
+		ret = fs_path_add_path(dest, name);
+		if (ret < 0)
+			goto out;
+
+		ino = parent_inode;
+		gen = parent_gen;
+	}
+
+out:
+	fs_path_free(name);
+	if (!ret)
+		fs_path_unreverse(dest);
+	return ret;
+}
+
+/*
+ * Sends a BTRFS_SEND_C_SUBVOL command/item to userspace
+ */
+static int send_subvol_begin(struct send_ctx *sctx)
+{
+	int ret;
+	struct btrfs_root *send_root = sctx->send_root;
+	struct btrfs_root *parent_root = sctx->parent_root;
+	struct btrfs_path *path;
+	struct btrfs_key key;
+	struct btrfs_root_ref *ref;
+	struct extent_buffer *leaf;
+	char *name = NULL;
+	int namelen;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	name = kmalloc(BTRFS_PATH_NAME_MAX, GFP_NOFS);
+	if (!name) {
+		btrfs_free_path(path);
+		return -ENOMEM;
+	}
+
+	key.objectid = send_root->objectid;
+	key.type = BTRFS_ROOT_BACKREF_KEY;
+	key.offset = 0;
+
+	ret = btrfs_search_slot_for_read(send_root->fs_info->tree_root,
+				&key, path, 1, 0);
+	if (ret < 0)
+		goto out;
+	if (ret) {
+		ret = -ENOENT;
+		goto out;
+	}
+
+	leaf = path->nodes[0];
+	btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
+	if (key.type != BTRFS_ROOT_BACKREF_KEY ||
+	    key.objectid != send_root->objectid) {
+		ret = -ENOENT;
+		goto out;
+	}
+	ref = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_root_ref);
+	namelen = btrfs_root_ref_name_len(leaf, ref);
+	read_extent_buffer(leaf, name, (unsigned long)(ref + 1), namelen);
+	btrfs_release_path(path);
+
+	if (parent_root) {
+		ret = begin_cmd(sctx, BTRFS_SEND_C_SNAPSHOT);
+		if (ret < 0)
+			goto out;
+	} else {
+		ret = begin_cmd(sctx, BTRFS_SEND_C_SUBVOL);
+		if (ret < 0)
+			goto out;
+	}
+
+	TLV_PUT_STRING(sctx, BTRFS_SEND_A_PATH, name, namelen);
+	TLV_PUT_UUID(sctx, BTRFS_SEND_A_UUID,
+			sctx->send_root->root_item.uuid);
+	TLV_PUT_U64(sctx, BTRFS_SEND_A_CTRANSID,
+		    le64_to_cpu(sctx->send_root->root_item.ctransid));
+	if (parent_root) {
+		TLV_PUT_UUID(sctx, BTRFS_SEND_A_CLONE_UUID,
+				sctx->parent_root->root_item.uuid);
+		TLV_PUT_U64(sctx, BTRFS_SEND_A_CLONE_CTRANSID,
+			    le64_to_cpu(sctx->parent_root->root_item.ctransid));
+	}
+
+	ret = send_cmd(sctx);
+
+tlv_put_failure:
+out:
+	btrfs_free_path(path);
+	kfree(name);
+	return ret;
+}
+
+static int send_truncate(struct send_ctx *sctx, u64 ino, u64 gen, u64 size)
+{
+	int ret = 0;
+	struct fs_path *p;
+
+verbose_printk("btrfs: send_truncate %llu size=%llu\n", ino, size);
+
+	p = fs_path_alloc();
+	if (!p)
+		return -ENOMEM;
+
+	ret = begin_cmd(sctx, BTRFS_SEND_C_TRUNCATE);
+	if (ret < 0)
+		goto out;
+
+	ret = get_cur_path(sctx, ino, gen, p);
+	if (ret < 0)
+		goto out;
+	TLV_PUT_PATH(sctx, BTRFS_SEND_A_PATH, p);
+	TLV_PUT_U64(sctx, BTRFS_SEND_A_SIZE, size);
+
+	ret = send_cmd(sctx);
+
+tlv_put_failure:
+out:
+	fs_path_free(p);
+	return ret;
+}
+
+static int send_chmod(struct send_ctx *sctx, u64 ino, u64 gen, u64 mode)
+{
+	int ret = 0;
+	struct fs_path *p;
+
+verbose_printk("btrfs: send_chmod %llu mode=%llu\n", ino, mode);
+
+	p = fs_path_alloc();
+	if (!p)
+		return -ENOMEM;
+
+	ret = begin_cmd(sctx, BTRFS_SEND_C_CHMOD);
+	if (ret < 0)
+		goto out;
+
+	ret = get_cur_path(sctx, ino, gen, p);
+	if (ret < 0)
+		goto out;
+	TLV_PUT_PATH(sctx, BTRFS_SEND_A_PATH, p);
+	TLV_PUT_U64(sctx, BTRFS_SEND_A_MODE, mode & 07777);
+
+	ret = send_cmd(sctx);
+
+tlv_put_failure:
+out:
+	fs_path_free(p);
+	return ret;
+}
+
+static int send_chown(struct send_ctx *sctx, u64 ino, u64 gen, u64 uid, u64 gid)
+{
+	int ret = 0;
+	struct fs_path *p;
+
+verbose_printk("btrfs: send_chown %llu uid=%llu, gid=%llu\n", ino, uid, gid);
+
+	p = fs_path_alloc();
+	if (!p)
+		return -ENOMEM;
+
+	ret = begin_cmd(sctx, BTRFS_SEND_C_CHOWN);
+	if (ret < 0)
+		goto out;
+
+	ret = get_cur_path(sctx, ino, gen, p);
+	if (ret < 0)
+		goto out;
+	TLV_PUT_PATH(sctx, BTRFS_SEND_A_PATH, p);
+	TLV_PUT_U64(sctx, BTRFS_SEND_A_UID, uid);
+	TLV_PUT_U64(sctx, BTRFS_SEND_A_GID, gid);
+
+	ret = send_cmd(sctx);
+
+tlv_put_failure:
+out:
+	fs_path_free(p);
+	return ret;
+}
+
+static int send_utimes(struct send_ctx *sctx, u64 ino, u64 gen)
+{
+	int ret = 0;
+	struct fs_path *p = NULL;
+	struct btrfs_inode_item *ii;
+	struct btrfs_path *path = NULL;
+	struct extent_buffer *eb;
+	struct btrfs_key key;
+	int slot;
+
+verbose_printk("btrfs: send_utimes %llu\n", ino);
+
+	p = fs_path_alloc();
+	if (!p)
+		return -ENOMEM;
+
+	path = alloc_path_for_send();
+	if (!path) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	key.objectid = ino;
+	key.type = BTRFS_INODE_ITEM_KEY;
+	key.offset = 0;
+	ret = btrfs_search_slot(NULL, sctx->send_root, &key, path, 0, 0);
+	if (ret < 0)
+		goto out;
+
+	eb = path->nodes[0];
+	slot = path->slots[0];
+	ii = btrfs_item_ptr(eb, slot, struct btrfs_inode_item);
+
+	ret = begin_cmd(sctx, BTRFS_SEND_C_UTIMES);
+	if (ret < 0)
+		goto out;
+
+	ret = get_cur_path(sctx, ino, gen, p);
+	if (ret < 0)
+		goto out;
+	TLV_PUT_PATH(sctx, BTRFS_SEND_A_PATH, p);
+	TLV_PUT_BTRFS_TIMESPEC(sctx, BTRFS_SEND_A_ATIME, eb,
+			btrfs_inode_atime(ii));
+	TLV_PUT_BTRFS_TIMESPEC(sctx, BTRFS_SEND_A_MTIME, eb,
+			btrfs_inode_mtime(ii));
+	TLV_PUT_BTRFS_TIMESPEC(sctx, BTRFS_SEND_A_CTIME, eb,
+			btrfs_inode_ctime(ii));
+	/* TODO Add otime support when the otime patches get into upstream */
+
+	ret = send_cmd(sctx);
+
+tlv_put_failure:
+out:
+	fs_path_free(p);
+	btrfs_free_path(path);
+	return ret;
+}
+
+/*
+ * Sends a BTRFS_SEND_C_MKXXX or SYMLINK command to user space. We don't have
+ * a valid path yet because we did not process the refs yet. So, the inode
+ * is created as orphan.
+ */
+static int send_create_inode(struct send_ctx *sctx, u64 ino)
+{
+	int ret = 0;
+	struct fs_path *p;
+	int cmd;
+	u64 gen;
+	u64 mode;
+	u64 rdev;
+
+verbose_printk("btrfs: send_create_inode %llu\n", ino);
+
+	p = fs_path_alloc();
+	if (!p)
+		return -ENOMEM;
+
+	if (ino != sctx->cur_ino) {
+		ret = get_inode_info(sctx->send_root, ino, NULL, &gen, &mode,
+				     NULL, NULL, &rdev);
+		if (ret < 0)
+			goto out;
+	} else {
+		gen = sctx->cur_inode_gen;
+		mode = sctx->cur_inode_mode;
+		rdev = sctx->cur_inode_rdev;
+	}
+
+	if (S_ISREG(mode)) {
+		cmd = BTRFS_SEND_C_MKFILE;
+	} else if (S_ISDIR(mode)) {
+		cmd = BTRFS_SEND_C_MKDIR;
+	} else if (S_ISLNK(mode)) {
+		cmd = BTRFS_SEND_C_SYMLINK;
+	} else if (S_ISCHR(mode) || S_ISBLK(mode)) {
+		cmd = BTRFS_SEND_C_MKNOD;
+	} else if (S_ISFIFO(mode)) {
+		cmd = BTRFS_SEND_C_MKFIFO;
+	} else if (S_ISSOCK(mode)) {
+		cmd = BTRFS_SEND_C_MKSOCK;
+	} else {
+		printk(KERN_WARNING "btrfs: unexpected inode type %o",
+				(int)(mode & S_IFMT));
+		ret = -ENOTSUPP;
+		goto out;
+	}
+
+	ret = begin_cmd(sctx, cmd);
+	if (ret < 0)
+		goto out;
+
+	ret = gen_unique_name(sctx, ino, gen, p);
+	if (ret < 0)
+		goto out;
+
+	TLV_PUT_PATH(sctx, BTRFS_SEND_A_PATH, p);
+	TLV_PUT_U64(sctx, BTRFS_SEND_A_INO, ino);
+
+	if (S_ISLNK(mode)) {
+		fs_path_reset(p);
+		ret = read_symlink(sctx->send_root, ino, p);
+		if (ret < 0)
+			goto out;
+		TLV_PUT_PATH(sctx, BTRFS_SEND_A_PATH_LINK, p);
+	} else if (S_ISCHR(mode) || S_ISBLK(mode) ||
+		   S_ISFIFO(mode) || S_ISSOCK(mode)) {
+		TLV_PUT_U64(sctx, BTRFS_SEND_A_RDEV, new_encode_dev(rdev));
+		TLV_PUT_U64(sctx, BTRFS_SEND_A_MODE, mode);
+	}
+
+	ret = send_cmd(sctx);
+	if (ret < 0)
+		goto out;
+
+
+tlv_put_failure:
+out:
+	fs_path_free(p);
+	return ret;
+}
+
+/*
+ * We need some special handling for inodes that get processed before the parent
+ * directory got created. See process_recorded_refs for details.
+ * This function does the check if we already created the dir out of order.
+ */
+static int did_create_dir(struct send_ctx *sctx, u64 dir)
+{
+	int ret = 0;
+	struct btrfs_path *path = NULL;
+	struct btrfs_key key;
+	struct btrfs_key found_key;
+	struct btrfs_key di_key;
+	struct extent_buffer *eb;
+	struct btrfs_dir_item *di;
+	int slot;
+
+	path = alloc_path_for_send();
+	if (!path) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	key.objectid = dir;
+	key.type = BTRFS_DIR_INDEX_KEY;
+	key.offset = 0;
+	ret = btrfs_search_slot(NULL, sctx->send_root, &key, path, 0, 0);
+	if (ret < 0)
+		goto out;
+
+	while (1) {
+		eb = path->nodes[0];
+		slot = path->slots[0];
+		if (slot >= btrfs_header_nritems(eb)) {
+			ret = btrfs_next_leaf(sctx->send_root, path);
+			if (ret < 0) {
+				goto out;
+			} else if (ret > 0) {
+				ret = 0;
+				break;
+			}
+			continue;
+		}
+
+		btrfs_item_key_to_cpu(eb, &found_key, slot);
+		if (found_key.objectid != key.objectid ||
+		    found_key.type != key.type) {
+			ret = 0;
+			goto out;
+		}
+
+		di = btrfs_item_ptr(eb, slot, struct btrfs_dir_item);
+		btrfs_dir_item_key_to_cpu(eb, di, &di_key);
+
+		if (di_key.type != BTRFS_ROOT_ITEM_KEY &&
+		    di_key.objectid < sctx->send_progress) {
+			ret = 1;
+			goto out;
+		}
+
+		path->slots[0]++;
+	}
+
+out:
+	btrfs_free_path(path);
+	return ret;
+}
+
+/*
+ * Only creates the inode if it is:
+ * 1. Not a directory
+ * 2. Or a directory which was not created already due to out of order
+ *    directories. See did_create_dir and process_recorded_refs for details.
+ */
+static int send_create_inode_if_needed(struct send_ctx *sctx)
+{
+	int ret;
+
+	if (S_ISDIR(sctx->cur_inode_mode)) {
+		ret = did_create_dir(sctx, sctx->cur_ino);
+		if (ret < 0)
+			goto out;
+		if (ret) {
+			ret = 0;
+			goto out;
+		}
+	}
+
+	ret = send_create_inode(sctx, sctx->cur_ino);
+	if (ret < 0)
+		goto out;
+
+out:
+	return ret;
+}
+
+struct recorded_ref {
+	struct list_head list;
+	char *dir_path;
+	char *name;
+	struct fs_path *full_path;
+	u64 dir;
+	u64 dir_gen;
+	int dir_path_len;
+	int name_len;
+};
+
+/*
+ * We need to process new refs before deleted refs, but compare_tree gives us
+ * everything mixed. So we first record all refs and later process them.
+ * This function is a helper to record one ref.
+ */
+static int __record_ref(struct list_head *head, u64 dir,
+		      u64 dir_gen, struct fs_path *path)
+{
+	struct recorded_ref *ref;
+
+	ref = kmalloc(sizeof(*ref), GFP_NOFS);
+	if (!ref)
+		return -ENOMEM;
+
+	ref->dir = dir;
+	ref->dir_gen = dir_gen;
+	ref->full_path = path;
+
+	ref->name = (char *)kbasename(ref->full_path->start);
+	ref->name_len = ref->full_path->end - ref->name;
+	ref->dir_path = ref->full_path->start;
+	if (ref->name == ref->full_path->start)
+		ref->dir_path_len = 0;
+	else
+		ref->dir_path_len = ref->full_path->end -
+				ref->full_path->start - 1 - ref->name_len;
+
+	list_add_tail(&ref->list, head);
+	return 0;
+}
+
+static int dup_ref(struct recorded_ref *ref, struct list_head *list)
+{
+	struct recorded_ref *new;
+
+	new = kmalloc(sizeof(*ref), GFP_NOFS);
+	if (!new)
+		return -ENOMEM;
+
+	new->dir = ref->dir;
+	new->dir_gen = ref->dir_gen;
+	new->full_path = NULL;
+	INIT_LIST_HEAD(&new->list);
+	list_add_tail(&new->list, list);
+	return 0;
+}
+
+static void __free_recorded_refs(struct list_head *head)
+{
+	struct recorded_ref *cur;
+
+	while (!list_empty(head)) {
+		cur = list_entry(head->next, struct recorded_ref, list);
+		fs_path_free(cur->full_path);
+		list_del(&cur->list);
+		kfree(cur);
+	}
+}
+
+static void free_recorded_refs(struct send_ctx *sctx)
+{
+	__free_recorded_refs(&sctx->new_refs);
+	__free_recorded_refs(&sctx->deleted_refs);
+}
+
+/*
+ * Renames/moves a file/dir to its orphan name. Used when the first
+ * ref of an unprocessed inode gets overwritten and for all non empty
+ * directories.
+ */
+static int orphanize_inode(struct send_ctx *sctx, u64 ino, u64 gen,
+			  struct fs_path *path)
+{
+	int ret;
+	struct fs_path *orphan;
+
+	orphan = fs_path_alloc();
+	if (!orphan)
+		return -ENOMEM;
+
+	ret = gen_unique_name(sctx, ino, gen, orphan);
+	if (ret < 0)
+		goto out;
+
+	ret = send_rename(sctx, path, orphan);
+
+out:
+	fs_path_free(orphan);
+	return ret;
+}
+
+static struct orphan_dir_info *
+add_orphan_dir_info(struct send_ctx *sctx, u64 dir_ino)
+{
+	struct rb_node **p = &sctx->orphan_dirs.rb_node;
+	struct rb_node *parent = NULL;
+	struct orphan_dir_info *entry, *odi;
+
+	odi = kmalloc(sizeof(*odi), GFP_NOFS);
+	if (!odi)
+		return ERR_PTR(-ENOMEM);
+	odi->ino = dir_ino;
+	odi->gen = 0;
+
+	while (*p) {
+		parent = *p;
+		entry = rb_entry(parent, struct orphan_dir_info, node);
+		if (dir_ino < entry->ino) {
+			p = &(*p)->rb_left;
+		} else if (dir_ino > entry->ino) {
+			p = &(*p)->rb_right;
+		} else {
+			kfree(odi);
+			return entry;
+		}
+	}
+
+	rb_link_node(&odi->node, parent, p);
+	rb_insert_color(&odi->node, &sctx->orphan_dirs);
+	return odi;
+}
+
+static struct orphan_dir_info *
+get_orphan_dir_info(struct send_ctx *sctx, u64 dir_ino)
+{
+	struct rb_node *n = sctx->orphan_dirs.rb_node;
+	struct orphan_dir_info *entry;
+
+	while (n) {
+		entry = rb_entry(n, struct orphan_dir_info, node);
+		if (dir_ino < entry->ino)
+			n = n->rb_left;
+		else if (dir_ino > entry->ino)
+			n = n->rb_right;
+		else
+			return entry;
+	}
+	return NULL;
+}
+
+static int is_waiting_for_rm(struct send_ctx *sctx, u64 dir_ino)
+{
+	struct orphan_dir_info *odi = get_orphan_dir_info(sctx, dir_ino);
+
+	return odi != NULL;
+}
+
+static void free_orphan_dir_info(struct send_ctx *sctx,
+				 struct orphan_dir_info *odi)
+{
+	if (!odi)
+		return;
+	rb_erase(&odi->node, &sctx->orphan_dirs);
+	kfree(odi);
+}
+
+/*
+ * Returns 1 if a directory can be removed at this point in time.
+ * We check this by iterating all dir items and checking if the inode behind
+ * the dir item was already processed.
+ */
+static int can_rmdir(struct send_ctx *sctx, u64 dir, u64 dir_gen,
+		     u64 send_progress)
+{
+	int ret = 0;
+	struct btrfs_root *root = sctx->parent_root;
+	struct btrfs_path *path;
+	struct btrfs_key key;
+	struct btrfs_key found_key;
+	struct btrfs_key loc;
+	struct btrfs_dir_item *di;
+
+	/*
+	 * Don't try to rmdir the top/root subvolume dir.
+	 */
+	if (dir == BTRFS_FIRST_FREE_OBJECTID)
+		return 0;
+
+	path = alloc_path_for_send();
+	if (!path)
+		return -ENOMEM;
+
+	key.objectid = dir;
+	key.type = BTRFS_DIR_INDEX_KEY;
+	key.offset = 0;
+	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+	if (ret < 0)
+		goto out;
+
+	while (1) {
+		struct waiting_dir_move *dm;
+
+		if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
+			ret = btrfs_next_leaf(root, path);
+			if (ret < 0)
+				goto out;
+			else if (ret > 0)
+				break;
+			continue;
+		}
+		btrfs_item_key_to_cpu(path->nodes[0], &found_key,
+				      path->slots[0]);
+		if (found_key.objectid != key.objectid ||
+		    found_key.type != key.type)
+			break;
+
+		di = btrfs_item_ptr(path->nodes[0], path->slots[0],
+				struct btrfs_dir_item);
+		btrfs_dir_item_key_to_cpu(path->nodes[0], di, &loc);
+
+		dm = get_waiting_dir_move(sctx, loc.objectid);
+		if (dm) {
+			struct orphan_dir_info *odi;
+
+			odi = add_orphan_dir_info(sctx, dir);
+			if (IS_ERR(odi)) {
+				ret = PTR_ERR(odi);
+				goto out;
+			}
+			odi->gen = dir_gen;
+			dm->rmdir_ino = dir;
+			ret = 0;
+			goto out;
+		}
+
+		if (loc.objectid > send_progress) {
+			ret = 0;
+			goto out;
+		}
+
+		path->slots[0]++;
+	}
+
+	ret = 1;
+
+out:
+	btrfs_free_path(path);
+	return ret;
+}
+
+static int is_waiting_for_move(struct send_ctx *sctx, u64 ino)
+{
+	struct waiting_dir_move *entry = get_waiting_dir_move(sctx, ino);
+
+	return entry != NULL;
+}
+
+static int add_waiting_dir_move(struct send_ctx *sctx, u64 ino)
+{
+	struct rb_node **p = &sctx->waiting_dir_moves.rb_node;
+	struct rb_node *parent = NULL;
+	struct waiting_dir_move *entry, *dm;
+
+	dm = kmalloc(sizeof(*dm), GFP_NOFS);
+	if (!dm)
+		return -ENOMEM;
+	dm->ino = ino;
+	dm->rmdir_ino = 0;
+
+	while (*p) {
+		parent = *p;
+		entry = rb_entry(parent, struct waiting_dir_move, node);
+		if (ino < entry->ino) {
+			p = &(*p)->rb_left;
+		} else if (ino > entry->ino) {
+			p = &(*p)->rb_right;
+		} else {
+			kfree(dm);
+			return -EEXIST;
+		}
+	}
+
+	rb_link_node(&dm->node, parent, p);
+	rb_insert_color(&dm->node, &sctx->waiting_dir_moves);
+	return 0;
+}
+
+static struct waiting_dir_move *
+get_waiting_dir_move(struct send_ctx *sctx, u64 ino)
+{
+	struct rb_node *n = sctx->waiting_dir_moves.rb_node;
+	struct waiting_dir_move *entry;
+
+	while (n) {
+		entry = rb_entry(n, struct waiting_dir_move, node);
+		if (ino < entry->ino)
+			n = n->rb_left;
+		else if (ino > entry->ino)
+			n = n->rb_right;
+		else
+			return entry;
+	}
+	return NULL;
+}
+
+static void free_waiting_dir_move(struct send_ctx *sctx,
+				  struct waiting_dir_move *dm)
+{
+	if (!dm)
+		return;
+	rb_erase(&dm->node, &sctx->waiting_dir_moves);
+	kfree(dm);
+}
+
+static int add_pending_dir_move(struct send_ctx *sctx,
+				u64 ino,
+				u64 ino_gen,
+				u64 parent_ino,
+				struct list_head *new_refs,
+				struct list_head *deleted_refs)
+{
+	struct rb_node **p = &sctx->pending_dir_moves.rb_node;
+	struct rb_node *parent = NULL;
+	struct pending_dir_move *entry = NULL, *pm;
+	struct recorded_ref *cur;
+	int exists = 0;
+	int ret;
+
+	pm = kmalloc(sizeof(*pm), GFP_NOFS);
+	if (!pm)
+		return -ENOMEM;
+	pm->parent_ino = parent_ino;
+	pm->ino = ino;
+	pm->gen = ino_gen;
+	INIT_LIST_HEAD(&pm->list);
+	INIT_LIST_HEAD(&pm->update_refs);
+	RB_CLEAR_NODE(&pm->node);
+
+	while (*p) {
+		parent = *p;
+		entry = rb_entry(parent, struct pending_dir_move, node);
+		if (parent_ino < entry->parent_ino) {
+			p = &(*p)->rb_left;
+		} else if (parent_ino > entry->parent_ino) {
+			p = &(*p)->rb_right;
+		} else {
+			exists = 1;
+			break;
+		}
+	}
+
+	list_for_each_entry(cur, deleted_refs, list) {
+		ret = dup_ref(cur, &pm->update_refs);
+		if (ret < 0)
+			goto out;
+	}
+	list_for_each_entry(cur, new_refs, list) {
+		ret = dup_ref(cur, &pm->update_refs);
+		if (ret < 0)
+			goto out;
+	}
+
+	ret = add_waiting_dir_move(sctx, pm->ino);
+	if (ret)
+		goto out;
+
+	if (exists) {
+		list_add_tail(&pm->list, &entry->list);
+	} else {
+		rb_link_node(&pm->node, parent, p);
+		rb_insert_color(&pm->node, &sctx->pending_dir_moves);
+	}
+	ret = 0;
+out:
+	if (ret) {
+		__free_recorded_refs(&pm->update_refs);
+		kfree(pm);
+	}
+	return ret;
+}
+
+static struct pending_dir_move *get_pending_dir_moves(struct send_ctx *sctx,
+						      u64 parent_ino)
+{
+	struct rb_node *n = sctx->pending_dir_moves.rb_node;
+	struct pending_dir_move *entry;
+
+	while (n) {
+		entry = rb_entry(n, struct pending_dir_move, node);
+		if (parent_ino < entry->parent_ino)
+			n = n->rb_left;
+		else if (parent_ino > entry->parent_ino)
+			n = n->rb_right;
+		else
+			return entry;
+	}
+	return NULL;
+}
+
+static int path_loop(struct send_ctx *sctx, struct fs_path *name,
+		     u64 ino, u64 gen, u64 *ancestor_ino)
+{
+	int ret = 0;
+	u64 parent_inode = 0;
+	u64 parent_gen = 0;
+	u64 start_ino = ino;
+
+	*ancestor_ino = 0;
+	while (ino != BTRFS_FIRST_FREE_OBJECTID) {
+		fs_path_reset(name);
+
+		if (is_waiting_for_rm(sctx, ino))
+			break;
+		if (is_waiting_for_move(sctx, ino)) {
+			if (*ancestor_ino == 0)
+				*ancestor_ino = ino;
+			ret = get_first_ref(sctx->parent_root, ino,
+					    &parent_inode, &parent_gen, name);
+		} else {
+			ret = __get_cur_name_and_parent(sctx, ino, gen,
+							&parent_inode,
+							&parent_gen, name);
+			if (ret > 0) {
+				ret = 0;
+				break;
+			}
+		}
+		if (ret < 0)
+			break;
+		if (parent_inode == start_ino) {
+			ret = 1;
+			if (*ancestor_ino == 0)
+				*ancestor_ino = ino;
+			break;
+		}
+		ino = parent_inode;
+		gen = parent_gen;
+	}
+	return ret;
+}
+
+static int apply_dir_move(struct send_ctx *sctx, struct pending_dir_move *pm)
+{
+	struct fs_path *from_path = NULL;
+	struct fs_path *to_path = NULL;
+	struct fs_path *name = NULL;
+	u64 orig_progress = sctx->send_progress;
+	struct recorded_ref *cur;
+	u64 parent_ino, parent_gen;
+	struct waiting_dir_move *dm = NULL;
+	u64 rmdir_ino = 0;
+	int ret;
+	u64 ancestor = 0;
+
+	name = fs_path_alloc();
+	from_path = fs_path_alloc();
+	if (!name || !from_path) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	dm = get_waiting_dir_move(sctx, pm->ino);
+	ASSERT(dm);
+	rmdir_ino = dm->rmdir_ino;
+	free_waiting_dir_move(sctx, dm);
+
+	ret = get_first_ref(sctx->parent_root, pm->ino,
+			    &parent_ino, &parent_gen, name);
+	if (ret < 0)
+		goto out;
+
+	ret = get_cur_path(sctx, parent_ino, parent_gen,
+			   from_path);
+	if (ret < 0)
+		goto out;
+	ret = fs_path_add_path(from_path, name);
+	if (ret < 0)
+		goto out;
+
+	sctx->send_progress = sctx->cur_ino + 1;
+	ret = path_loop(sctx, name, pm->ino, pm->gen, &ancestor);
+	if (ret) {
+		LIST_HEAD(deleted_refs);
+		ASSERT(ancestor > BTRFS_FIRST_FREE_OBJECTID);
+		ret = add_pending_dir_move(sctx, pm->ino, pm->gen, ancestor,
+					   &pm->update_refs, &deleted_refs);
+		if (ret < 0)
+			goto out;
+		if (rmdir_ino) {
+			dm = get_waiting_dir_move(sctx, pm->ino);
+			ASSERT(dm);
+			dm->rmdir_ino = rmdir_ino;
+		}
+		goto out;
+	}
+	fs_path_reset(name);
+	to_path = name;
+	name = NULL;
+	ret = get_cur_path(sctx, pm->ino, pm->gen, to_path);
+	if (ret < 0)
+		goto out;
+
+	ret = send_rename(sctx, from_path, to_path);
+	if (ret < 0)
+		goto out;
+
+	if (rmdir_ino) {
+		struct orphan_dir_info *odi;
+
+		odi = get_orphan_dir_info(sctx, rmdir_ino);
+		if (!odi) {
+			/* already deleted */
+			goto finish;
+		}
+		ret = can_rmdir(sctx, rmdir_ino, odi->gen, sctx->cur_ino + 1);
+		if (ret < 0)
+			goto out;
+		if (!ret)
+			goto finish;
+
+		name = fs_path_alloc();
+		if (!name) {
+			ret = -ENOMEM;
+			goto out;
+		}
+		ret = get_cur_path(sctx, rmdir_ino, odi->gen, name);
+		if (ret < 0)
+			goto out;
+		ret = send_rmdir(sctx, name);
+		if (ret < 0)
+			goto out;
+		free_orphan_dir_info(sctx, odi);
+	}
+
+finish:
+	ret = send_utimes(sctx, pm->ino, pm->gen);
+	if (ret < 0)
+		goto out;
+
+	/*
+	 * After rename/move, need to update the utimes of both new parent(s)
+	 * and old parent(s).
+	 */
+	list_for_each_entry(cur, &pm->update_refs, list) {
+		if (cur->dir == rmdir_ino)
+			continue;
+		ret = send_utimes(sctx, cur->dir, cur->dir_gen);
+		if (ret < 0)
+			goto out;
+	}
+
+out:
+	fs_path_free(name);
+	fs_path_free(from_path);
+	fs_path_free(to_path);
+	sctx->send_progress = orig_progress;
+
+	return ret;
+}
+
+static void free_pending_move(struct send_ctx *sctx, struct pending_dir_move *m)
+{
+	if (!list_empty(&m->list))
+		list_del(&m->list);
+	if (!RB_EMPTY_NODE(&m->node))
+		rb_erase(&m->node, &sctx->pending_dir_moves);
+	__free_recorded_refs(&m->update_refs);
+	kfree(m);
+}
+
+static void tail_append_pending_moves(struct pending_dir_move *moves,
+				      struct list_head *stack)
+{
+	if (list_empty(&moves->list)) {
+		list_add_tail(&moves->list, stack);
+	} else {
+		LIST_HEAD(list);
+		list_splice_init(&moves->list, &list);
+		list_add_tail(&moves->list, stack);
+		list_splice_tail(&list, stack);
+	}
+}
+
+static int apply_children_dir_moves(struct send_ctx *sctx)
+{
+	struct pending_dir_move *pm;
+	struct list_head stack;
+	u64 parent_ino = sctx->cur_ino;
+	int ret = 0;
+
+	pm = get_pending_dir_moves(sctx, parent_ino);
+	if (!pm)
+		return 0;
+
+	INIT_LIST_HEAD(&stack);
+	tail_append_pending_moves(pm, &stack);
+
+	while (!list_empty(&stack)) {
+		pm = list_first_entry(&stack, struct pending_dir_move, list);
+		parent_ino = pm->ino;
+		ret = apply_dir_move(sctx, pm);
+		free_pending_move(sctx, pm);
+		if (ret)
+			goto out;
+		pm = get_pending_dir_moves(sctx, parent_ino);
+		if (pm)
+			tail_append_pending_moves(pm, &stack);
+	}
+	return 0;
+
+out:
+	while (!list_empty(&stack)) {
+		pm = list_first_entry(&stack, struct pending_dir_move, list);
+		free_pending_move(sctx, pm);
+	}
+	return ret;
+}
+
+static int wait_for_parent_move(struct send_ctx *sctx,
+				struct recorded_ref *parent_ref)
+{
+	int ret = 0;
+	u64 ino = parent_ref->dir;
+	u64 parent_ino_before, parent_ino_after;
+	struct fs_path *path_before = NULL;
+	struct fs_path *path_after = NULL;
+	int len1, len2;
+
+	path_after = fs_path_alloc();
+	path_before = fs_path_alloc();
+	if (!path_after || !path_before) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	/*
+	 * Our current directory inode may not yet be renamed/moved because some
+	 * ancestor (immediate or not) has to be renamed/moved first. So find if
+	 * such ancestor exists and make sure our own rename/move happens after
+	 * that ancestor is processed.
+	 */
+	while (ino > BTRFS_FIRST_FREE_OBJECTID) {
+		if (is_waiting_for_move(sctx, ino)) {
+			ret = 1;
+			break;
+		}
+
+		fs_path_reset(path_before);
+		fs_path_reset(path_after);
+
+		ret = get_first_ref(sctx->send_root, ino, &parent_ino_after,
+				    NULL, path_after);
+		if (ret < 0)
+			goto out;
+		ret = get_first_ref(sctx->parent_root, ino, &parent_ino_before,
+				    NULL, path_before);
+		if (ret < 0 && ret != -ENOENT) {
+			goto out;
+		} else if (ret == -ENOENT) {
+			ret = 1;
+			break;
+		}
+
+		len1 = fs_path_len(path_before);
+		len2 = fs_path_len(path_after);
+		if (ino > sctx->cur_ino &&
+		    (parent_ino_before != parent_ino_after || len1 != len2 ||
+		     memcmp(path_before->start, path_after->start, len1))) {
+			ret = 1;
+			break;
+		}
+		ino = parent_ino_after;
+	}
+
+out:
+	fs_path_free(path_before);
+	fs_path_free(path_after);
+
+	if (ret == 1) {
+		ret = add_pending_dir_move(sctx,
+					   sctx->cur_ino,
+					   sctx->cur_inode_gen,
+					   ino,
+					   &sctx->new_refs,
+					   &sctx->deleted_refs);
+		if (!ret)
+			ret = 1;
+	}
+
+	return ret;
+}
+
+/*
+ * This does all the move/link/unlink/rmdir magic.
+ */
+static int process_recorded_refs(struct send_ctx *sctx, int *pending_move)
+{
+	int ret = 0;
+	struct recorded_ref *cur;
+	struct recorded_ref *cur2;
+	struct list_head check_dirs;
+	struct fs_path *valid_path = NULL;
+	u64 ow_inode = 0;
+	u64 ow_gen;
+	int did_overwrite = 0;
+	int is_orphan = 0;
+	u64 last_dir_ino_rm = 0;
+
+verbose_printk("btrfs: process_recorded_refs %llu\n", sctx->cur_ino);
+
+	/*
+	 * This should never happen as the root dir always has the same ref
+	 * which is always '..'
+	 */
+	BUG_ON(sctx->cur_ino <= BTRFS_FIRST_FREE_OBJECTID);
+	INIT_LIST_HEAD(&check_dirs);
+
+	valid_path = fs_path_alloc();
+	if (!valid_path) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	/*
+	 * First, check if the first ref of the current inode was overwritten
+	 * before. If yes, we know that the current inode was already orphanized
+	 * and thus use the orphan name. If not, we can use get_cur_path to
+	 * get the path of the first ref as it would like while receiving at
+	 * this point in time.
+	 * New inodes are always orphan at the beginning, so force to use the
+	 * orphan name in this case.
+	 * The first ref is stored in valid_path and will be updated if it
+	 * gets moved around.
+	 */
+	if (!sctx->cur_inode_new) {
+		ret = did_overwrite_first_ref(sctx, sctx->cur_ino,
+				sctx->cur_inode_gen);
+		if (ret < 0)
+			goto out;
+		if (ret)
+			did_overwrite = 1;
+	}
+	if (sctx->cur_inode_new || did_overwrite) {
+		ret = gen_unique_name(sctx, sctx->cur_ino,
+				sctx->cur_inode_gen, valid_path);
+		if (ret < 0)
+			goto out;
+		is_orphan = 1;
+	} else {
+		ret = get_cur_path(sctx, sctx->cur_ino, sctx->cur_inode_gen,
+				valid_path);
+		if (ret < 0)
+			goto out;
+	}
+
+	list_for_each_entry(cur, &sctx->new_refs, list) {
+		/*
+		 * We may have refs where the parent directory does not exist
+		 * yet. This happens if the parent directories inum is higher
+		 * the the current inum. To handle this case, we create the
+		 * parent directory out of order. But we need to check if this
+		 * did already happen before due to other refs in the same dir.
+		 */
+		ret = get_cur_inode_state(sctx, cur->dir, cur->dir_gen);
+		if (ret < 0)
+			goto out;
+		if (ret == inode_state_will_create) {
+			ret = 0;
+			/*
+			 * First check if any of the current inodes refs did
+			 * already create the dir.
+			 */
+			list_for_each_entry(cur2, &sctx->new_refs, list) {
+				if (cur == cur2)
+					break;
+				if (cur2->dir == cur->dir) {
+					ret = 1;
+					break;
+				}
+			}
+
+			/*
+			 * If that did not happen, check if a previous inode
+			 * did already create the dir.
+			 */
+			if (!ret)
+				ret = did_create_dir(sctx, cur->dir);
+			if (ret < 0)
+				goto out;
+			if (!ret) {
+				ret = send_create_inode(sctx, cur->dir);
+				if (ret < 0)
+					goto out;
+			}
+		}
+
+		/*
+		 * Check if this new ref would overwrite the first ref of
+		 * another unprocessed inode. If yes, orphanize the
+		 * overwritten inode. If we find an overwritten ref that is
+		 * not the first ref, simply unlink it.
+		 */
+		ret = will_overwrite_ref(sctx, cur->dir, cur->dir_gen,
+				cur->name, cur->name_len,
+				&ow_inode, &ow_gen);
+		if (ret < 0)
+			goto out;
+		if (ret) {
+			ret = is_first_ref(sctx->parent_root,
+					   ow_inode, cur->dir, cur->name,
+					   cur->name_len);
+			if (ret < 0)
+				goto out;
+			if (ret) {
+				ret = orphanize_inode(sctx, ow_inode, ow_gen,
+						cur->full_path);
+				if (ret < 0)
+					goto out;
+			} else {
+				ret = send_unlink(sctx, cur->full_path);
+				if (ret < 0)
+					goto out;
+			}
+		}
+
+		/*
+		 * link/move the ref to the new place. If we have an orphan
+		 * inode, move it and update valid_path. If not, link or move
+		 * it depending on the inode mode.
+		 */
+		if (is_orphan) {
+			ret = send_rename(sctx, valid_path, cur->full_path);
+			if (ret < 0)
+				goto out;
+			is_orphan = 0;
+			ret = fs_path_copy(valid_path, cur->full_path);
+			if (ret < 0)
+				goto out;
+		} else {
+			if (S_ISDIR(sctx->cur_inode_mode)) {
+				/*
+				 * Dirs can't be linked, so move it. For moved
+				 * dirs, we always have one new and one deleted
+				 * ref. The deleted ref is ignored later.
+				 */
+				ret = wait_for_parent_move(sctx, cur);
+				if (ret < 0)
+					goto out;
+				if (ret) {
+					*pending_move = 1;
+				} else {
+					ret = send_rename(sctx, valid_path,
+							  cur->full_path);
+					if (!ret)
+						ret = fs_path_copy(valid_path,
+							       cur->full_path);
+				}
+				if (ret < 0)
+					goto out;
+			} else {
+				ret = send_link(sctx, cur->full_path,
+						valid_path);
+				if (ret < 0)
+					goto out;
+			}
+		}
+		ret = dup_ref(cur, &check_dirs);
+		if (ret < 0)
+			goto out;
+	}
+
+	if (S_ISDIR(sctx->cur_inode_mode) && sctx->cur_inode_deleted) {
+		/*
+		 * Check if we can already rmdir the directory. If not,
+		 * orphanize it. For every dir item inside that gets deleted
+		 * later, we do this check again and rmdir it then if possible.
+		 * See the use of check_dirs for more details.
+		 */
+		ret = can_rmdir(sctx, sctx->cur_ino, sctx->cur_inode_gen,
+				sctx->cur_ino);
+		if (ret < 0)
+			goto out;
+		if (ret) {
+			ret = send_rmdir(sctx, valid_path);
+			if (ret < 0)
+				goto out;
+		} else if (!is_orphan) {
+			ret = orphanize_inode(sctx, sctx->cur_ino,
+					sctx->cur_inode_gen, valid_path);
+			if (ret < 0)
+				goto out;
+			is_orphan = 1;
+		}
+
+		list_for_each_entry(cur, &sctx->deleted_refs, list) {
+			ret = dup_ref(cur, &check_dirs);
+			if (ret < 0)
+				goto out;
+		}
+	} else if (S_ISDIR(sctx->cur_inode_mode) &&
+		   !list_empty(&sctx->deleted_refs)) {
+		/*
+		 * We have a moved dir. Add the old parent to check_dirs
+		 */
+		cur = list_entry(sctx->deleted_refs.next, struct recorded_ref,
+				list);
+		ret = dup_ref(cur, &check_dirs);
+		if (ret < 0)
+			goto out;
+	} else if (!S_ISDIR(sctx->cur_inode_mode)) {
+		/*
+		 * We have a non dir inode. Go through all deleted refs and
+		 * unlink them if they were not already overwritten by other
+		 * inodes.
+		 */
+		list_for_each_entry(cur, &sctx->deleted_refs, list) {
+			ret = did_overwrite_ref(sctx, cur->dir, cur->dir_gen,
+					sctx->cur_ino, sctx->cur_inode_gen,
+					cur->name, cur->name_len);
+			if (ret < 0)
+				goto out;
+			if (!ret) {
+				ret = send_unlink(sctx, cur->full_path);
+				if (ret < 0)
+					goto out;
+			}
+			ret = dup_ref(cur, &check_dirs);
+			if (ret < 0)
+				goto out;
+		}
+		/*
+		 * If the inode is still orphan, unlink the orphan. This may
+		 * happen when a previous inode did overwrite the first ref
+		 * of this inode and no new refs were added for the current
+		 * inode. Unlinking does not mean that the inode is deleted in
+		 * all cases. There may still be links to this inode in other
+		 * places.
+		 */
+		if (is_orphan) {
+			ret = send_unlink(sctx, valid_path);
+			if (ret < 0)
+				goto out;
+		}
+	}
+
+	/*
+	 * We did collect all parent dirs where cur_inode was once located. We
+	 * now go through all these dirs and check if they are pending for
+	 * deletion and if it's finally possible to perform the rmdir now.
+	 * We also update the inode stats of the parent dirs here.
+	 */
+	list_for_each_entry(cur, &check_dirs, list) {
+		/*
+		 * In case we had refs into dirs that were not processed yet,
+		 * we don't need to do the utime and rmdir logic for these dirs.
+		 * The dir will be processed later.
+		 */
+		if (cur->dir > sctx->cur_ino)
+			continue;
+
+		ret = get_cur_inode_state(sctx, cur->dir, cur->dir_gen);
+		if (ret < 0)
+			goto out;
+
+		if (ret == inode_state_did_create ||
+		    ret == inode_state_no_change) {
+			/* TODO delayed utimes */
+			ret = send_utimes(sctx, cur->dir, cur->dir_gen);
+			if (ret < 0)
+				goto out;
+		} else if (ret == inode_state_did_delete &&
+			   cur->dir != last_dir_ino_rm) {
+			ret = can_rmdir(sctx, cur->dir, cur->dir_gen,
+					sctx->cur_ino);
+			if (ret < 0)
+				goto out;
+			if (ret) {
+				ret = get_cur_path(sctx, cur->dir,
+						   cur->dir_gen, valid_path);
+				if (ret < 0)
+					goto out;
+				ret = send_rmdir(sctx, valid_path);
+				if (ret < 0)
+					goto out;
+				last_dir_ino_rm = cur->dir;
+			}
+		}
+	}
+
+	ret = 0;
+
+out:
+	__free_recorded_refs(&check_dirs);
+	free_recorded_refs(sctx);
+	fs_path_free(valid_path);
+	return ret;
+}
+
+static int record_ref(struct btrfs_root *root, int num, u64 dir, int index,
+		      struct fs_path *name, void *ctx, struct list_head *refs)
+{
+	int ret = 0;
+	struct send_ctx *sctx = ctx;
+	struct fs_path *p;
+	u64 gen;
+
+	p = fs_path_alloc();
+	if (!p)
+		return -ENOMEM;
+
+	ret = get_inode_info(root, dir, NULL, &gen, NULL, NULL,
+			NULL, NULL);
+	if (ret < 0)
+		goto out;
+
+	ret = get_cur_path(sctx, dir, gen, p);
+	if (ret < 0)
+		goto out;
+	ret = fs_path_add_path(p, name);
+	if (ret < 0)
+		goto out;
+
+	ret = __record_ref(refs, dir, gen, p);
+
+out:
+	if (ret)
+		fs_path_free(p);
+	return ret;
+}
+
+static int __record_new_ref(int num, u64 dir, int index,
+			    struct fs_path *name,
+			    void *ctx)
+{
+	struct send_ctx *sctx = ctx;
+	return record_ref(sctx->send_root, num, dir, index, name,
+			  ctx, &sctx->new_refs);
+}
+
+
+static int __record_deleted_ref(int num, u64 dir, int index,
+				struct fs_path *name,
+				void *ctx)
+{
+	struct send_ctx *sctx = ctx;
+	return record_ref(sctx->parent_root, num, dir, index, name,
+			  ctx, &sctx->deleted_refs);
+}
+
+static int record_new_ref(struct send_ctx *sctx)
+{
+	int ret;
+
+	ret = iterate_inode_ref(sctx->send_root, sctx->left_path,
+				sctx->cmp_key, 0, __record_new_ref, sctx);
+	if (ret < 0)
+		goto out;
+	ret = 0;
+
+out:
+	return ret;
+}
+
+static int record_deleted_ref(struct send_ctx *sctx)
+{
+	int ret;
+
+	ret = iterate_inode_ref(sctx->parent_root, sctx->right_path,
+				sctx->cmp_key, 0, __record_deleted_ref, sctx);
+	if (ret < 0)
+		goto out;
+	ret = 0;
+
+out:
+	return ret;
+}
+
+struct find_ref_ctx {
+	u64 dir;
+	u64 dir_gen;
+	struct btrfs_root *root;
+	struct fs_path *name;
+	int found_idx;
+};
+
+static int __find_iref(int num, u64 dir, int index,
+		       struct fs_path *name,
+		       void *ctx_)
+{
+	struct find_ref_ctx *ctx = ctx_;
+	u64 dir_gen;
+	int ret;
+
+	if (dir == ctx->dir && fs_path_len(name) == fs_path_len(ctx->name) &&
+	    strncmp(name->start, ctx->name->start, fs_path_len(name)) == 0) {
+		/*
+		 * To avoid doing extra lookups we'll only do this if everything
+		 * else matches.
+		 */
+		ret = get_inode_info(ctx->root, dir, NULL, &dir_gen, NULL,
+				     NULL, NULL, NULL);
+		if (ret)
+			return ret;
+		if (dir_gen != ctx->dir_gen)
+			return 0;
+		ctx->found_idx = num;
+		return 1;
+	}
+	return 0;
+}
+
+static int find_iref(struct btrfs_root *root,
+		     struct btrfs_path *path,
+		     struct btrfs_key *key,
+		     u64 dir, u64 dir_gen, struct fs_path *name)
+{
+	int ret;
+	struct find_ref_ctx ctx;
+
+	ctx.dir = dir;
+	ctx.name = name;
+	ctx.dir_gen = dir_gen;
+	ctx.found_idx = -1;
+	ctx.root = root;
+
+	ret = iterate_inode_ref(root, path, key, 0, __find_iref, &ctx);
+	if (ret < 0)
+		return ret;
+
+	if (ctx.found_idx == -1)
+		return -ENOENT;
+
+	return ctx.found_idx;
+}
+
+static int __record_changed_new_ref(int num, u64 dir, int index,
+				    struct fs_path *name,
+				    void *ctx)
+{
+	u64 dir_gen;
+	int ret;
+	struct send_ctx *sctx = ctx;
+
+	ret = get_inode_info(sctx->send_root, dir, NULL, &dir_gen, NULL,
+			     NULL, NULL, NULL);
+	if (ret)
+		return ret;
+
+	ret = find_iref(sctx->parent_root, sctx->right_path,
+			sctx->cmp_key, dir, dir_gen, name);
+	if (ret == -ENOENT)
+		ret = __record_new_ref(num, dir, index, name, sctx);
+	else if (ret > 0)
+		ret = 0;
+
+	return ret;
+}
+
+static int __record_changed_deleted_ref(int num, u64 dir, int index,
+					struct fs_path *name,
+					void *ctx)
+{
+	u64 dir_gen;
+	int ret;
+	struct send_ctx *sctx = ctx;
+
+	ret = get_inode_info(sctx->parent_root, dir, NULL, &dir_gen, NULL,
+			     NULL, NULL, NULL);
+	if (ret)
+		return ret;
+
+	ret = find_iref(sctx->send_root, sctx->left_path, sctx->cmp_key,
+			dir, dir_gen, name);
+	if (ret == -ENOENT)
+		ret = __record_deleted_ref(num, dir, index, name, sctx);
+	else if (ret > 0)
+		ret = 0;
+
+	return ret;
+}
+
+static int record_changed_ref(struct send_ctx *sctx)
+{
+	int ret = 0;
+
+	ret = iterate_inode_ref(sctx->send_root, sctx->left_path,
+			sctx->cmp_key, 0, __record_changed_new_ref, sctx);
+	if (ret < 0)
+		goto out;
+	ret = iterate_inode_ref(sctx->parent_root, sctx->right_path,
+			sctx->cmp_key, 0, __record_changed_deleted_ref, sctx);
+	if (ret < 0)
+		goto out;
+	ret = 0;
+
+out:
+	return ret;
+}
+
+/*
+ * Record and process all refs at once. Needed when an inode changes the
+ * generation number, which means that it was deleted and recreated.
+ */
+static int process_all_refs(struct send_ctx *sctx,
+			    enum btrfs_compare_tree_result cmd)
+{
+	int ret;
+	struct btrfs_root *root;
+	struct btrfs_path *path;
+	struct btrfs_key key;
+	struct btrfs_key found_key;
+	struct extent_buffer *eb;
+	int slot;
+	iterate_inode_ref_t cb;
+	int pending_move = 0;
+
+	path = alloc_path_for_send();
+	if (!path)
+		return -ENOMEM;
+
+	if (cmd == BTRFS_COMPARE_TREE_NEW) {
+		root = sctx->send_root;
+		cb = __record_new_ref;
+	} else if (cmd == BTRFS_COMPARE_TREE_DELETED) {
+		root = sctx->parent_root;
+		cb = __record_deleted_ref;
+	} else {
+		btrfs_err(sctx->send_root->fs_info,
+				"Wrong command %d in process_all_refs", cmd);
+		ret = -EINVAL;
+		goto out;
+	}
+
+	key.objectid = sctx->cmp_key->objectid;
+	key.type = BTRFS_INODE_REF_KEY;
+	key.offset = 0;
+	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+	if (ret < 0)
+		goto out;
+
+	while (1) {
+		eb = path->nodes[0];
+		slot = path->slots[0];
+		if (slot >= btrfs_header_nritems(eb)) {
+			ret = btrfs_next_leaf(root, path);
+			if (ret < 0)
+				goto out;
+			else if (ret > 0)
+				break;
+			continue;
+		}
+
+		btrfs_item_key_to_cpu(eb, &found_key, slot);
+
+		if (found_key.objectid != key.objectid ||
+		    (found_key.type != BTRFS_INODE_REF_KEY &&
+		     found_key.type != BTRFS_INODE_EXTREF_KEY))
+			break;
+
+		ret = iterate_inode_ref(root, path, &found_key, 0, cb, sctx);
+		if (ret < 0)
+			goto out;
+
+		path->slots[0]++;
+	}
+	btrfs_release_path(path);
+
+	ret = process_recorded_refs(sctx, &pending_move);
+	/* Only applicable to an incremental send. */
+	ASSERT(pending_move == 0);
+
+out:
+	btrfs_free_path(path);
+	return ret;
+}
+
+static int send_set_xattr(struct send_ctx *sctx,
+			  struct fs_path *path,
+			  const char *name, int name_len,
+			  const char *data, int data_len)
+{
+	int ret = 0;
+
+	ret = begin_cmd(sctx, BTRFS_SEND_C_SET_XATTR);
+	if (ret < 0)
+		goto out;
+
+	TLV_PUT_PATH(sctx, BTRFS_SEND_A_PATH, path);
+	TLV_PUT_STRING(sctx, BTRFS_SEND_A_XATTR_NAME, name, name_len);
+	TLV_PUT(sctx, BTRFS_SEND_A_XATTR_DATA, data, data_len);
+
+	ret = send_cmd(sctx);
+
+tlv_put_failure:
+out:
+	return ret;
+}
+
+static int send_remove_xattr(struct send_ctx *sctx,
+			  struct fs_path *path,
+			  const char *name, int name_len)
+{
+	int ret = 0;
+
+	ret = begin_cmd(sctx, BTRFS_SEND_C_REMOVE_XATTR);
+	if (ret < 0)
+		goto out;
+
+	TLV_PUT_PATH(sctx, BTRFS_SEND_A_PATH, path);
+	TLV_PUT_STRING(sctx, BTRFS_SEND_A_XATTR_NAME, name, name_len);
+
+	ret = send_cmd(sctx);
+
+tlv_put_failure:
+out:
+	return ret;
+}
+
+static int __process_new_xattr(int num, struct btrfs_key *di_key,
+			       const char *name, int name_len,
+			       const char *data, int data_len,
+			       u8 type, void *ctx)
+{
+	int ret;
+	struct send_ctx *sctx = ctx;
+	struct fs_path *p;
+	posix_acl_xattr_header dummy_acl;
+
+	p = fs_path_alloc();
+	if (!p)
+		return -ENOMEM;
+
+	/*
+	 * This hack is needed because empty acl's are stored as zero byte
+	 * data in xattrs. Problem with that is, that receiving these zero byte
+	 * acl's will fail later. To fix this, we send a dummy acl list that
+	 * only contains the version number and no entries.
+	 */
+	if (!strncmp(name, XATTR_NAME_POSIX_ACL_ACCESS, name_len) ||
+	    !strncmp(name, XATTR_NAME_POSIX_ACL_DEFAULT, name_len)) {
+		if (data_len == 0) {
+			dummy_acl.a_version =
+					cpu_to_le32(POSIX_ACL_XATTR_VERSION);
+			data = (char *)&dummy_acl;
+			data_len = sizeof(dummy_acl);
+		}
+	}
+
+	ret = get_cur_path(sctx, sctx->cur_ino, sctx->cur_inode_gen, p);
+	if (ret < 0)
+		goto out;
+
+	ret = send_set_xattr(sctx, p, name, name_len, data, data_len);
+
+out:
+	fs_path_free(p);
+	return ret;
+}
+
+static int __process_deleted_xattr(int num, struct btrfs_key *di_key,
+				   const char *name, int name_len,
+				   const char *data, int data_len,
+				   u8 type, void *ctx)
+{
+	int ret;
+	struct send_ctx *sctx = ctx;
+	struct fs_path *p;
+
+	p = fs_path_alloc();
+	if (!p)
+		return -ENOMEM;
+
+	ret = get_cur_path(sctx, sctx->cur_ino, sctx->cur_inode_gen, p);
+	if (ret < 0)
+		goto out;
+
+	ret = send_remove_xattr(sctx, p, name, name_len);
+
+out:
+	fs_path_free(p);
+	return ret;
+}
+
+static int process_new_xattr(struct send_ctx *sctx)
+{
+	int ret = 0;
+
+	ret = iterate_dir_item(sctx->send_root, sctx->left_path,
+			       sctx->cmp_key, __process_new_xattr, sctx);
+
+	return ret;
+}
+
+static int process_deleted_xattr(struct send_ctx *sctx)
+{
+	int ret;
+
+	ret = iterate_dir_item(sctx->parent_root, sctx->right_path,
+			       sctx->cmp_key, __process_deleted_xattr, sctx);
+
+	return ret;
+}
+
+struct find_xattr_ctx {
+	const char *name;
+	int name_len;
+	int found_idx;
+	char *found_data;
+	int found_data_len;
+};
+
+static int __find_xattr(int num, struct btrfs_key *di_key,
+			const char *name, int name_len,
+			const char *data, int data_len,
+			u8 type, void *vctx)
+{
+	struct find_xattr_ctx *ctx = vctx;
+
+	if (name_len == ctx->name_len &&
+	    strncmp(name, ctx->name, name_len) == 0) {
+		ctx->found_idx = num;
+		ctx->found_data_len = data_len;
+		ctx->found_data = kmemdup(data, data_len, GFP_NOFS);
+		if (!ctx->found_data)
+			return -ENOMEM;
+		return 1;
+	}
+	return 0;
+}
+
+static int find_xattr(struct btrfs_root *root,
+		      struct btrfs_path *path,
+		      struct btrfs_key *key,
+		      const char *name, int name_len,
+		      char **data, int *data_len)
+{
+	int ret;
+	struct find_xattr_ctx ctx;
+
+	ctx.name = name;
+	ctx.name_len = name_len;
+	ctx.found_idx = -1;
+	ctx.found_data = NULL;
+	ctx.found_data_len = 0;
+
+	ret = iterate_dir_item(root, path, key, __find_xattr, &ctx);
+	if (ret < 0)
+		return ret;
+
+	if (ctx.found_idx == -1)
+		return -ENOENT;
+	if (data) {
+		*data = ctx.found_data;
+		*data_len = ctx.found_data_len;
+	} else {
+		kfree(ctx.found_data);
+	}
+	return ctx.found_idx;
+}
+
+
+static int __process_changed_new_xattr(int num, struct btrfs_key *di_key,
+				       const char *name, int name_len,
+				       const char *data, int data_len,
+				       u8 type, void *ctx)
+{
+	int ret;
+	struct send_ctx *sctx = ctx;
+	char *found_data = NULL;
+	int found_data_len  = 0;
+
+	ret = find_xattr(sctx->parent_root, sctx->right_path,
+			 sctx->cmp_key, name, name_len, &found_data,
+			 &found_data_len);
+	if (ret == -ENOENT) {
+		ret = __process_new_xattr(num, di_key, name, name_len, data,
+				data_len, type, ctx);
+	} else if (ret >= 0) {
+		if (data_len != found_data_len ||
+		    memcmp(data, found_data, data_len)) {
+			ret = __process_new_xattr(num, di_key, name, name_len,
+					data, data_len, type, ctx);
+		} else {
+			ret = 0;
+		}
+	}
+
+	kfree(found_data);
+	return ret;
+}
+
+static int __process_changed_deleted_xattr(int num, struct btrfs_key *di_key,
+					   const char *name, int name_len,
+					   const char *data, int data_len,
+					   u8 type, void *ctx)
+{
+	int ret;
+	struct send_ctx *sctx = ctx;
+
+	ret = find_xattr(sctx->send_root, sctx->left_path, sctx->cmp_key,
+			 name, name_len, NULL, NULL);
+	if (ret == -ENOENT)
+		ret = __process_deleted_xattr(num, di_key, name, name_len, data,
+				data_len, type, ctx);
+	else if (ret >= 0)
+		ret = 0;
+
+	return ret;
+}
+
+static int process_changed_xattr(struct send_ctx *sctx)
+{
+	int ret = 0;
+
+	ret = iterate_dir_item(sctx->send_root, sctx->left_path,
+			sctx->cmp_key, __process_changed_new_xattr, sctx);
+	if (ret < 0)
+		goto out;
+	ret = iterate_dir_item(sctx->parent_root, sctx->right_path,
+			sctx->cmp_key, __process_changed_deleted_xattr, sctx);
+
+out:
+	return ret;
+}
+
+static int process_all_new_xattrs(struct send_ctx *sctx)
+{
+	int ret;
+	struct btrfs_root *root;
+	struct btrfs_path *path;
+	struct btrfs_key key;
+	struct btrfs_key found_key;
+	struct extent_buffer *eb;
+	int slot;
+
+	path = alloc_path_for_send();
+	if (!path)
+		return -ENOMEM;
+
+	root = sctx->send_root;
+
+	key.objectid = sctx->cmp_key->objectid;
+	key.type = BTRFS_XATTR_ITEM_KEY;
+	key.offset = 0;
+	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+	if (ret < 0)
+		goto out;
+
+	while (1) {
+		eb = path->nodes[0];
+		slot = path->slots[0];
+		if (slot >= btrfs_header_nritems(eb)) {
+			ret = btrfs_next_leaf(root, path);
+			if (ret < 0) {
+				goto out;
+			} else if (ret > 0) {
+				ret = 0;
+				break;
+			}
+			continue;
+		}
+
+		btrfs_item_key_to_cpu(eb, &found_key, slot);
+		if (found_key.objectid != key.objectid ||
+		    found_key.type != key.type) {
+			ret = 0;
+			goto out;
+		}
+
+		ret = iterate_dir_item(root, path, &found_key,
+				       __process_new_xattr, sctx);
+		if (ret < 0)
+			goto out;
+
+		path->slots[0]++;
+	}
+
+out:
+	btrfs_free_path(path);
+	return ret;
+}
+
+static ssize_t fill_read_buf(struct send_ctx *sctx, u64 offset, u32 len)
+{
+	struct btrfs_root *root = sctx->send_root;
+	struct btrfs_fs_info *fs_info = root->fs_info;
+	struct inode *inode;
+	struct page *page;
+	char *addr;
+	struct btrfs_key key;
+	pgoff_t index = offset >> PAGE_CACHE_SHIFT;
+	pgoff_t last_index;
+	unsigned pg_offset = offset & ~PAGE_CACHE_MASK;
+	ssize_t ret = 0;
+
+	key.objectid = sctx->cur_ino;
+	key.type = BTRFS_INODE_ITEM_KEY;
+	key.offset = 0;
+
+	inode = btrfs_iget(fs_info->sb, &key, root, NULL);
+	if (IS_ERR(inode))
+		return PTR_ERR(inode);
+
+	if (offset + len > i_size_read(inode)) {
+		if (offset > i_size_read(inode))
+			len = 0;
+		else
+			len = offset - i_size_read(inode);
+	}
+	if (len == 0)
+		goto out;
+
+	last_index = (offset + len - 1) >> PAGE_CACHE_SHIFT;
+
+	/* initial readahead */
+	memset(&sctx->ra, 0, sizeof(struct file_ra_state));
+	file_ra_state_init(&sctx->ra, inode->i_mapping);
+	btrfs_force_ra(inode->i_mapping, &sctx->ra, NULL, index,
+		       last_index - index + 1);
+
+	while (index <= last_index) {
+		unsigned cur_len = min_t(unsigned, len,
+					 PAGE_CACHE_SIZE - pg_offset);
+		page = find_or_create_page(inode->i_mapping, index, GFP_NOFS);
+		if (!page) {
+			ret = -ENOMEM;
+			break;
+		}
+
+		if (!PageUptodate(page)) {
+			btrfs_readpage(NULL, page);
+			lock_page(page);
+			if (!PageUptodate(page)) {
+				unlock_page(page);
+				page_cache_release(page);
+				ret = -EIO;
+				break;
+			}
+		}
+
+		addr = kmap(page);
+		memcpy(sctx->read_buf + ret, addr + pg_offset, cur_len);
+		kunmap(page);
+		unlock_page(page);
+		page_cache_release(page);
+		index++;
+		pg_offset = 0;
+		len -= cur_len;
+		ret += cur_len;
+	}
+out:
+	iput(inode);
+	return ret;
+}
+
+/*
+ * Read some bytes from the current inode/file and send a write command to
+ * user space.
+ */
+static int send_write(struct send_ctx *sctx, u64 offset, u32 len)
+{
+	int ret = 0;
+	struct fs_path *p;
+	ssize_t num_read = 0;
+
+	p = fs_path_alloc();
+	if (!p)
+		return -ENOMEM;
+
+verbose_printk("btrfs: send_write offset=%llu, len=%d\n", offset, len);
+
+	num_read = fill_read_buf(sctx, offset, len);
+	if (num_read <= 0) {
+		if (num_read < 0)
+			ret = num_read;
+		goto out;
+	}
+
+	ret = begin_cmd(sctx, BTRFS_SEND_C_WRITE);
+	if (ret < 0)
+		goto out;
+
+	ret = get_cur_path(sctx, sctx->cur_ino, sctx->cur_inode_gen, p);
+	if (ret < 0)
+		goto out;
+
+	TLV_PUT_PATH(sctx, BTRFS_SEND_A_PATH, p);
+	TLV_PUT_U64(sctx, BTRFS_SEND_A_FILE_OFFSET, offset);
+	TLV_PUT(sctx, BTRFS_SEND_A_DATA, sctx->read_buf, num_read);
+
+	ret = send_cmd(sctx);
+
+tlv_put_failure:
+out:
+	fs_path_free(p);
+	if (ret < 0)
+		return ret;
+	return num_read;
+}
+
+/*
+ * Send a clone command to user space.
+ */
+static int send_clone(struct send_ctx *sctx,
+		      u64 offset, u32 len,
+		      struct clone_root *clone_root)
+{
+	int ret = 0;
+	struct fs_path *p;
+	u64 gen;
+
+verbose_printk("btrfs: send_clone offset=%llu, len=%d, clone_root=%llu, "
+	       "clone_inode=%llu, clone_offset=%llu\n", offset, len,
+		clone_root->root->objectid, clone_root->ino,
+		clone_root->offset);
+
+	p = fs_path_alloc();
+	if (!p)
+		return -ENOMEM;
+
+	ret = begin_cmd(sctx, BTRFS_SEND_C_CLONE);
+	if (ret < 0)
+		goto out;
+
+	ret = get_cur_path(sctx, sctx->cur_ino, sctx->cur_inode_gen, p);
+	if (ret < 0)
+		goto out;
+
+	TLV_PUT_U64(sctx, BTRFS_SEND_A_FILE_OFFSET, offset);
+	TLV_PUT_U64(sctx, BTRFS_SEND_A_CLONE_LEN, len);
+	TLV_PUT_PATH(sctx, BTRFS_SEND_A_PATH, p);
+
+	if (clone_root->root == sctx->send_root) {
+		ret = get_inode_info(sctx->send_root, clone_root->ino, NULL,
+				&gen, NULL, NULL, NULL, NULL);
+		if (ret < 0)
+			goto out;
+		ret = get_cur_path(sctx, clone_root->ino, gen, p);
+	} else {
+		ret = get_inode_path(clone_root->root, clone_root->ino, p);
+	}
+	if (ret < 0)
+		goto out;
+
+	TLV_PUT_UUID(sctx, BTRFS_SEND_A_CLONE_UUID,
+			clone_root->root->root_item.uuid);
+	TLV_PUT_U64(sctx, BTRFS_SEND_A_CLONE_CTRANSID,
+		    le64_to_cpu(clone_root->root->root_item.ctransid));
+	TLV_PUT_PATH(sctx, BTRFS_SEND_A_CLONE_PATH, p);
+	TLV_PUT_U64(sctx, BTRFS_SEND_A_CLONE_OFFSET,
+			clone_root->offset);
+
+	ret = send_cmd(sctx);
+
+tlv_put_failure:
+out:
+	fs_path_free(p);
+	return ret;
+}
+
+/*
+ * Send an update extent command to user space.
+ */
+static int send_update_extent(struct send_ctx *sctx,
+			      u64 offset, u32 len)
+{
+	int ret = 0;
+	struct fs_path *p;
+
+	p = fs_path_alloc();
+	if (!p)
+		return -ENOMEM;
+
+	ret = begin_cmd(sctx, BTRFS_SEND_C_UPDATE_EXTENT);
+	if (ret < 0)
+		goto out;
+
+	ret = get_cur_path(sctx, sctx->cur_ino, sctx->cur_inode_gen, p);
+	if (ret < 0)
+		goto out;
+
+	TLV_PUT_PATH(sctx, BTRFS_SEND_A_PATH, p);
+	TLV_PUT_U64(sctx, BTRFS_SEND_A_FILE_OFFSET, offset);
+	TLV_PUT_U64(sctx, BTRFS_SEND_A_SIZE, len);
+
+	ret = send_cmd(sctx);
+
+tlv_put_failure:
+out:
+	fs_path_free(p);
+	return ret;
+}
+
+static int send_hole(struct send_ctx *sctx, u64 end)
+{
+	struct fs_path *p = NULL;
+	u64 offset = sctx->cur_inode_last_extent;
+	u64 len;
+	int ret = 0;
+
+	p = fs_path_alloc();
+	if (!p)
+		return -ENOMEM;
+	ret = get_cur_path(sctx, sctx->cur_ino, sctx->cur_inode_gen, p);
+	if (ret < 0)
+		goto tlv_put_failure;
+	memset(sctx->read_buf, 0, BTRFS_SEND_READ_SIZE);
+	while (offset < end) {
+		len = min_t(u64, end - offset, BTRFS_SEND_READ_SIZE);
+
+		ret = begin_cmd(sctx, BTRFS_SEND_C_WRITE);
+		if (ret < 0)
+			break;
+		TLV_PUT_PATH(sctx, BTRFS_SEND_A_PATH, p);
+		TLV_PUT_U64(sctx, BTRFS_SEND_A_FILE_OFFSET, offset);
+		TLV_PUT(sctx, BTRFS_SEND_A_DATA, sctx->read_buf, len);
+		ret = send_cmd(sctx);
+		if (ret < 0)
+			break;
+		offset += len;
+	}
+tlv_put_failure:
+	fs_path_free(p);
+	return ret;
+}
+
+static int send_write_or_clone(struct send_ctx *sctx,
+			       struct btrfs_path *path,
+			       struct btrfs_key *key,
+			       struct clone_root *clone_root)
+{
+	int ret = 0;
+	struct btrfs_file_extent_item *ei;
+	u64 offset = key->offset;
+	u64 pos = 0;
+	u64 len;
+	u32 l;
+	u8 type;
+	u64 bs = sctx->send_root->fs_info->sb->s_blocksize;
+
+	ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
+			struct btrfs_file_extent_item);
+	type = btrfs_file_extent_type(path->nodes[0], ei);
+	if (type == BTRFS_FILE_EXTENT_INLINE) {
+		len = btrfs_file_extent_inline_len(path->nodes[0],
+						   path->slots[0], ei);
+		/*
+		 * it is possible the inline item won't cover the whole page,
+		 * but there may be items after this page.  Make
+		 * sure to send the whole thing
+		 */
+		len = PAGE_CACHE_ALIGN(len);
+	} else {
+		len = btrfs_file_extent_num_bytes(path->nodes[0], ei);
+	}
+
+	if (offset + len > sctx->cur_inode_size)
+		len = sctx->cur_inode_size - offset;
+	if (len == 0) {
+		ret = 0;
+		goto out;
+	}
+
+	if (clone_root && IS_ALIGNED(offset + len, bs)) {
+		ret = send_clone(sctx, offset, len, clone_root);
+	} else if (sctx->flags & BTRFS_SEND_FLAG_NO_FILE_DATA) {
+		ret = send_update_extent(sctx, offset, len);
+	} else {
+		while (pos < len) {
+			l = len - pos;
+			if (l > BTRFS_SEND_READ_SIZE)
+				l = BTRFS_SEND_READ_SIZE;
+			ret = send_write(sctx, pos + offset, l);
+			if (ret < 0)
+				goto out;
+			if (!ret)
+				break;
+			pos += ret;
+		}
+		ret = 0;
+	}
+out:
+	return ret;
+}
+
+static int is_extent_unchanged(struct send_ctx *sctx,
+			       struct btrfs_path *left_path,
+			       struct btrfs_key *ekey)
+{
+	int ret = 0;
+	struct btrfs_key key;
+	struct btrfs_path *path = NULL;
+	struct extent_buffer *eb;
+	int slot;
+	struct btrfs_key found_key;
+	struct btrfs_file_extent_item *ei;
+	u64 left_disknr;
+	u64 right_disknr;
+	u64 left_offset;
+	u64 right_offset;
+	u64 left_offset_fixed;
+	u64 left_len;
+	u64 right_len;
+	u64 left_gen;
+	u64 right_gen;
+	u8 left_type;
+	u8 right_type;
+
+	path = alloc_path_for_send();
+	if (!path)
+		return -ENOMEM;
+
+	eb = left_path->nodes[0];
+	slot = left_path->slots[0];
+	ei = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item);
+	left_type = btrfs_file_extent_type(eb, ei);
+
+	if (left_type != BTRFS_FILE_EXTENT_REG) {
+		ret = 0;
+		goto out;
+	}
+	left_disknr = btrfs_file_extent_disk_bytenr(eb, ei);
+	left_len = btrfs_file_extent_num_bytes(eb, ei);
+	left_offset = btrfs_file_extent_offset(eb, ei);
+	left_gen = btrfs_file_extent_generation(eb, ei);
+
+	/*
+	 * Following comments will refer to these graphics. L is the left
+	 * extents which we are checking at the moment. 1-8 are the right
+	 * extents that we iterate.
+	 *
+	 *       |-----L-----|
+	 * |-1-|-2a-|-3-|-4-|-5-|-6-|
+	 *
+	 *       |-----L-----|
+	 * |--1--|-2b-|...(same as above)
+	 *
+	 * Alternative situation. Happens on files where extents got split.
+	 *       |-----L-----|
+	 * |-----------7-----------|-6-|
+	 *
+	 * Alternative situation. Happens on files which got larger.
+	 *       |-----L-----|
+	 * |-8-|
+	 * Nothing follows after 8.
+	 */
+
+	key.objectid = ekey->objectid;
+	key.type = BTRFS_EXTENT_DATA_KEY;
+	key.offset = ekey->offset;
+	ret = btrfs_search_slot_for_read(sctx->parent_root, &key, path, 0, 0);
+	if (ret < 0)
+		goto out;
+	if (ret) {
+		ret = 0;
+		goto out;
+	}
+
+	/*
+	 * Handle special case where the right side has no extents at all.
+	 */
+	eb = path->nodes[0];
+	slot = path->slots[0];
+	btrfs_item_key_to_cpu(eb, &found_key, slot);
+	if (found_key.objectid != key.objectid ||
+	    found_key.type != key.type) {
+		/* If we're a hole then just pretend nothing changed */
+		ret = (left_disknr) ? 0 : 1;
+		goto out;
+	}
+
+	/*
+	 * We're now on 2a, 2b or 7.
+	 */
+	key = found_key;
+	while (key.offset < ekey->offset + left_len) {
+		ei = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item);
+		right_type = btrfs_file_extent_type(eb, ei);
+		if (right_type != BTRFS_FILE_EXTENT_REG) {
+			ret = 0;
+			goto out;
+		}
+
+		right_disknr = btrfs_file_extent_disk_bytenr(eb, ei);
+		right_len = btrfs_file_extent_num_bytes(eb, ei);
+		right_offset = btrfs_file_extent_offset(eb, ei);
+		right_gen = btrfs_file_extent_generation(eb, ei);
+
+		/*
+		 * Are we at extent 8? If yes, we know the extent is changed.
+		 * This may only happen on the first iteration.
+		 */
+		if (found_key.offset + right_len <= ekey->offset) {
+			/* If we're a hole just pretend nothing changed */
+			ret = (left_disknr) ? 0 : 1;
+			goto out;
+		}
+
+		left_offset_fixed = left_offset;
+		if (key.offset < ekey->offset) {
+			/* Fix the right offset for 2a and 7. */
+			right_offset += ekey->offset - key.offset;
+		} else {
+			/* Fix the left offset for all behind 2a and 2b */
+			left_offset_fixed += key.offset - ekey->offset;
+		}
+
+		/*
+		 * Check if we have the same extent.
+		 */
+		if (left_disknr != right_disknr ||
+		    left_offset_fixed != right_offset ||
+		    left_gen != right_gen) {
+			ret = 0;
+			goto out;
+		}
+
+		/*
+		 * Go to the next extent.
+		 */
+		ret = btrfs_next_item(sctx->parent_root, path);
+		if (ret < 0)
+			goto out;
+		if (!ret) {
+			eb = path->nodes[0];
+			slot = path->slots[0];
+			btrfs_item_key_to_cpu(eb, &found_key, slot);
+		}
+		if (ret || found_key.objectid != key.objectid ||
+		    found_key.type != key.type) {
+			key.offset += right_len;
+			break;
+		}
+		if (found_key.offset != key.offset + right_len) {
+			ret = 0;
+			goto out;
+		}
+		key = found_key;
+	}
+
+	/*
+	 * We're now behind the left extent (treat as unchanged) or at the end
+	 * of the right side (treat as changed).
+	 */
+	if (key.offset >= ekey->offset + left_len)
+		ret = 1;
+	else
+		ret = 0;
+
+
+out:
+	btrfs_free_path(path);
+	return ret;
+}
+
+static int get_last_extent(struct send_ctx *sctx, u64 offset)
+{
+	struct btrfs_path *path;
+	struct btrfs_root *root = sctx->send_root;
+	struct btrfs_file_extent_item *fi;
+	struct btrfs_key key;
+	u64 extent_end;
+	u8 type;
+	int ret;
+
+	path = alloc_path_for_send();
+	if (!path)
+		return -ENOMEM;
+
+	sctx->cur_inode_last_extent = 0;
+
+	key.objectid = sctx->cur_ino;
+	key.type = BTRFS_EXTENT_DATA_KEY;
+	key.offset = offset;
+	ret = btrfs_search_slot_for_read(root, &key, path, 0, 1);
+	if (ret < 0)
+		goto out;
+	ret = 0;
+	btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
+	if (key.objectid != sctx->cur_ino || key.type != BTRFS_EXTENT_DATA_KEY)
+		goto out;
+
+	fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
+			    struct btrfs_file_extent_item);
+	type = btrfs_file_extent_type(path->nodes[0], fi);
+	if (type == BTRFS_FILE_EXTENT_INLINE) {
+		u64 size = btrfs_file_extent_inline_len(path->nodes[0],
+							path->slots[0], fi);
+		extent_end = ALIGN(key.offset + size,
+				   sctx->send_root->sectorsize);
+	} else {
+		extent_end = key.offset +
+			btrfs_file_extent_num_bytes(path->nodes[0], fi);
+	}
+	sctx->cur_inode_last_extent = extent_end;
+out:
+	btrfs_free_path(path);
+	return ret;
+}
+
+static int maybe_send_hole(struct send_ctx *sctx, struct btrfs_path *path,
+			   struct btrfs_key *key)
+{
+	struct btrfs_file_extent_item *fi;
+	u64 extent_end;
+	u8 type;
+	int ret = 0;
+
+	if (sctx->cur_ino != key->objectid || !need_send_hole(sctx))
+		return 0;
+
+	if (sctx->cur_inode_last_extent == (u64)-1) {
+		ret = get_last_extent(sctx, key->offset - 1);
+		if (ret)
+			return ret;
+	}
+
+	fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
+			    struct btrfs_file_extent_item);
+	type = btrfs_file_extent_type(path->nodes[0], fi);
+	if (type == BTRFS_FILE_EXTENT_INLINE) {
+		u64 size = btrfs_file_extent_inline_len(path->nodes[0],
+							path->slots[0], fi);
+		extent_end = ALIGN(key->offset + size,
+				   sctx->send_root->sectorsize);
+	} else {
+		extent_end = key->offset +
+			btrfs_file_extent_num_bytes(path->nodes[0], fi);
+	}
+
+	if (path->slots[0] == 0 &&
+	    sctx->cur_inode_last_extent < key->offset) {
+		/*
+		 * We might have skipped entire leafs that contained only
+		 * file extent items for our current inode. These leafs have
+		 * a generation number smaller (older) than the one in the
+		 * current leaf and the leaf our last extent came from, and
+		 * are located between these 2 leafs.
+		 */
+		ret = get_last_extent(sctx, key->offset - 1);
+		if (ret)
+			return ret;
+	}
+
+	if (sctx->cur_inode_last_extent < key->offset)
+		ret = send_hole(sctx, key->offset);
+	sctx->cur_inode_last_extent = extent_end;
+	return ret;
+}
+
+static int process_extent(struct send_ctx *sctx,
+			  struct btrfs_path *path,
+			  struct btrfs_key *key)
+{
+	struct clone_root *found_clone = NULL;
+	int ret = 0;
+
+	if (S_ISLNK(sctx->cur_inode_mode))
+		return 0;
+
+	if (sctx->parent_root && !sctx->cur_inode_new) {
+		ret = is_extent_unchanged(sctx, path, key);
+		if (ret < 0)
+			goto out;
+		if (ret) {
+			ret = 0;
+			goto out_hole;
+		}
+	} else {
+		struct btrfs_file_extent_item *ei;
+		u8 type;
+
+		ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
+				    struct btrfs_file_extent_item);
+		type = btrfs_file_extent_type(path->nodes[0], ei);
+		if (type == BTRFS_FILE_EXTENT_PREALLOC ||
+		    type == BTRFS_FILE_EXTENT_REG) {
+			/*
+			 * The send spec does not have a prealloc command yet,
+			 * so just leave a hole for prealloc'ed extents until
+			 * we have enough commands queued up to justify rev'ing
+			 * the send spec.
+			 */
+			if (type == BTRFS_FILE_EXTENT_PREALLOC) {
+				ret = 0;
+				goto out;
+			}
+
+			/* Have a hole, just skip it. */
+			if (btrfs_file_extent_disk_bytenr(path->nodes[0], ei) == 0) {
+				ret = 0;
+				goto out;
+			}
+		}
+	}
+
+	ret = find_extent_clone(sctx, path, key->objectid, key->offset,
+			sctx->cur_inode_size, &found_clone);
+	if (ret != -ENOENT && ret < 0)
+		goto out;
+
+	ret = send_write_or_clone(sctx, path, key, found_clone);
+	if (ret)
+		goto out;
+out_hole:
+	ret = maybe_send_hole(sctx, path, key);
+out:
+	return ret;
+}
+
+static int process_all_extents(struct send_ctx *sctx)
+{
+	int ret;
+	struct btrfs_root *root;
+	struct btrfs_path *path;
+	struct btrfs_key key;
+	struct btrfs_key found_key;
+	struct extent_buffer *eb;
+	int slot;
+
+	root = sctx->send_root;
+	path = alloc_path_for_send();
+	if (!path)
+		return -ENOMEM;
+
+	key.objectid = sctx->cmp_key->objectid;
+	key.type = BTRFS_EXTENT_DATA_KEY;
+	key.offset = 0;
+	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+	if (ret < 0)
+		goto out;
+
+	while (1) {
+		eb = path->nodes[0];
+		slot = path->slots[0];
+
+		if (slot >= btrfs_header_nritems(eb)) {
+			ret = btrfs_next_leaf(root, path);
+			if (ret < 0) {
+				goto out;
+			} else if (ret > 0) {
+				ret = 0;
+				break;
+			}
+			continue;
+		}
+
+		btrfs_item_key_to_cpu(eb, &found_key, slot);
+
+		if (found_key.objectid != key.objectid ||
+		    found_key.type != key.type) {
+			ret = 0;
+			goto out;
+		}
+
+		ret = process_extent(sctx, path, &found_key);
+		if (ret < 0)
+			goto out;
+
+		path->slots[0]++;
+	}
+
+out:
+	btrfs_free_path(path);
+	return ret;
+}
+
+static int process_recorded_refs_if_needed(struct send_ctx *sctx, int at_end,
+					   int *pending_move,
+					   int *refs_processed)
+{
+	int ret = 0;
+
+	if (sctx->cur_ino == 0)
+		goto out;
+	if (!at_end && sctx->cur_ino == sctx->cmp_key->objectid &&
+	    sctx->cmp_key->type <= BTRFS_INODE_EXTREF_KEY)
+		goto out;
+	if (list_empty(&sctx->new_refs) && list_empty(&sctx->deleted_refs))
+		goto out;
+
+	ret = process_recorded_refs(sctx, pending_move);
+	if (ret < 0)
+		goto out;
+
+	*refs_processed = 1;
+out:
+	return ret;
+}
+
+static int finish_inode_if_needed(struct send_ctx *sctx, int at_end)
+{
+	int ret = 0;
+	u64 left_mode;
+	u64 left_uid;
+	u64 left_gid;
+	u64 right_mode;
+	u64 right_uid;
+	u64 right_gid;
+	int need_chmod = 0;
+	int need_chown = 0;
+	int pending_move = 0;
+	int refs_processed = 0;
+
+	ret = process_recorded_refs_if_needed(sctx, at_end, &pending_move,
+					      &refs_processed);
+	if (ret < 0)
+		goto out;
+
+	/*
+	 * We have processed the refs and thus need to advance send_progress.
+	 * Now, calls to get_cur_xxx will take the updated refs of the current
+	 * inode into account.
+	 *
+	 * On the other hand, if our current inode is a directory and couldn't
+	 * be moved/renamed because its parent was renamed/moved too and it has
+	 * a higher inode number, we can only move/rename our current inode
+	 * after we moved/renamed its parent. Therefore in this case operate on
+	 * the old path (pre move/rename) of our current inode, and the
+	 * move/rename will be performed later.
+	 */
+	if (refs_processed && !pending_move)
+		sctx->send_progress = sctx->cur_ino + 1;
+
+	if (sctx->cur_ino == 0 || sctx->cur_inode_deleted)
+		goto out;
+	if (!at_end && sctx->cmp_key->objectid == sctx->cur_ino)
+		goto out;
+
+	ret = get_inode_info(sctx->send_root, sctx->cur_ino, NULL, NULL,
+			&left_mode, &left_uid, &left_gid, NULL);
+	if (ret < 0)
+		goto out;
+
+	if (!sctx->parent_root || sctx->cur_inode_new) {
+		need_chown = 1;
+		if (!S_ISLNK(sctx->cur_inode_mode))
+			need_chmod = 1;
+	} else {
+		ret = get_inode_info(sctx->parent_root, sctx->cur_ino,
+				NULL, NULL, &right_mode, &right_uid,
+				&right_gid, NULL);
+		if (ret < 0)
+			goto out;
+
+		if (left_uid != right_uid || left_gid != right_gid)
+			need_chown = 1;
+		if (!S_ISLNK(sctx->cur_inode_mode) && left_mode != right_mode)
+			need_chmod = 1;
+	}
+
+	if (S_ISREG(sctx->cur_inode_mode)) {
+		if (need_send_hole(sctx)) {
+			if (sctx->cur_inode_last_extent == (u64)-1 ||
+			    sctx->cur_inode_last_extent <
+			    sctx->cur_inode_size) {
+				ret = get_last_extent(sctx, (u64)-1);
+				if (ret)
+					goto out;
+			}
+			if (sctx->cur_inode_last_extent <
+			    sctx->cur_inode_size) {
+				ret = send_hole(sctx, sctx->cur_inode_size);
+				if (ret)
+					goto out;
+			}
+		}
+		ret = send_truncate(sctx, sctx->cur_ino, sctx->cur_inode_gen,
+				sctx->cur_inode_size);
+		if (ret < 0)
+			goto out;
+	}
+
+	if (need_chown) {
+		ret = send_chown(sctx, sctx->cur_ino, sctx->cur_inode_gen,
+				left_uid, left_gid);
+		if (ret < 0)
+			goto out;
+	}
+	if (need_chmod) {
+		ret = send_chmod(sctx, sctx->cur_ino, sctx->cur_inode_gen,
+				left_mode);
+		if (ret < 0)
+			goto out;
+	}
+
+	/*
+	 * If other directory inodes depended on our current directory
+	 * inode's move/rename, now do their move/rename operations.
+	 */
+	if (!is_waiting_for_move(sctx, sctx->cur_ino)) {
+		ret = apply_children_dir_moves(sctx);
+		if (ret)
+			goto out;
+		/*
+		 * Need to send that every time, no matter if it actually
+		 * changed between the two trees as we have done changes to
+		 * the inode before. If our inode is a directory and it's
+		 * waiting to be moved/renamed, we will send its utimes when
+		 * it's moved/renamed, therefore we don't need to do it here.
+		 */
+		sctx->send_progress = sctx->cur_ino + 1;
+		ret = send_utimes(sctx, sctx->cur_ino, sctx->cur_inode_gen);
+		if (ret < 0)
+			goto out;
+	}
+
+out:
+	return ret;
+}
+
+static int changed_inode(struct send_ctx *sctx,
+			 enum btrfs_compare_tree_result result)
+{
+	int ret = 0;
+	struct btrfs_key *key = sctx->cmp_key;
+	struct btrfs_inode_item *left_ii = NULL;
+	struct btrfs_inode_item *right_ii = NULL;
+	u64 left_gen = 0;
+	u64 right_gen = 0;
+
+	sctx->cur_ino = key->objectid;
+	sctx->cur_inode_new_gen = 0;
+	sctx->cur_inode_last_extent = (u64)-1;
+
+	/*
+	 * Set send_progress to current inode. This will tell all get_cur_xxx
+	 * functions that the current inode's refs are not updated yet. Later,
+	 * when process_recorded_refs is finished, it is set to cur_ino + 1.
+	 */
+	sctx->send_progress = sctx->cur_ino;
+
+	if (result == BTRFS_COMPARE_TREE_NEW ||
+	    result == BTRFS_COMPARE_TREE_CHANGED) {
+		left_ii = btrfs_item_ptr(sctx->left_path->nodes[0],
+				sctx->left_path->slots[0],
+				struct btrfs_inode_item);
+		left_gen = btrfs_inode_generation(sctx->left_path->nodes[0],
+				left_ii);
+	} else {
+		right_ii = btrfs_item_ptr(sctx->right_path->nodes[0],
+				sctx->right_path->slots[0],
+				struct btrfs_inode_item);
+		right_gen = btrfs_inode_generation(sctx->right_path->nodes[0],
+				right_ii);
+	}
+	if (result == BTRFS_COMPARE_TREE_CHANGED) {
+		right_ii = btrfs_item_ptr(sctx->right_path->nodes[0],
+				sctx->right_path->slots[0],
+				struct btrfs_inode_item);
+
+		right_gen = btrfs_inode_generation(sctx->right_path->nodes[0],
+				right_ii);
+
+		/*
+		 * The cur_ino = root dir case is special here. We can't treat
+		 * the inode as deleted+reused because it would generate a
+		 * stream that tries to delete/mkdir the root dir.
+		 */
+		if (left_gen != right_gen &&
+		    sctx->cur_ino != BTRFS_FIRST_FREE_OBJECTID)
+			sctx->cur_inode_new_gen = 1;
+	}
+
+	if (result == BTRFS_COMPARE_TREE_NEW) {
+		sctx->cur_inode_gen = left_gen;
+		sctx->cur_inode_new = 1;
+		sctx->cur_inode_deleted = 0;
+		sctx->cur_inode_size = btrfs_inode_size(
+				sctx->left_path->nodes[0], left_ii);
+		sctx->cur_inode_mode = btrfs_inode_mode(
+				sctx->left_path->nodes[0], left_ii);
+		sctx->cur_inode_rdev = btrfs_inode_rdev(
+				sctx->left_path->nodes[0], left_ii);
+		if (sctx->cur_ino != BTRFS_FIRST_FREE_OBJECTID)
+			ret = send_create_inode_if_needed(sctx);
+	} else if (result == BTRFS_COMPARE_TREE_DELETED) {
+		sctx->cur_inode_gen = right_gen;
+		sctx->cur_inode_new = 0;
+		sctx->cur_inode_deleted = 1;
+		sctx->cur_inode_size = btrfs_inode_size(
+				sctx->right_path->nodes[0], right_ii);
+		sctx->cur_inode_mode = btrfs_inode_mode(
+				sctx->right_path->nodes[0], right_ii);
+	} else if (result == BTRFS_COMPARE_TREE_CHANGED) {
+		/*
+		 * We need to do some special handling in case the inode was
+		 * reported as changed with a changed generation number. This
+		 * means that the original inode was deleted and new inode
+		 * reused the same inum. So we have to treat the old inode as
+		 * deleted and the new one as new.
+		 */
+		if (sctx->cur_inode_new_gen) {
+			/*
+			 * First, process the inode as if it was deleted.
+			 */
+			sctx->cur_inode_gen = right_gen;
+			sctx->cur_inode_new = 0;
+			sctx->cur_inode_deleted = 1;
+			sctx->cur_inode_size = btrfs_inode_size(
+					sctx->right_path->nodes[0], right_ii);
+			sctx->cur_inode_mode = btrfs_inode_mode(
+					sctx->right_path->nodes[0], right_ii);
+			ret = process_all_refs(sctx,
+					BTRFS_COMPARE_TREE_DELETED);
+			if (ret < 0)
+				goto out;
+
+			/*
+			 * Now process the inode as if it was new.
+			 */
+			sctx->cur_inode_gen = left_gen;
+			sctx->cur_inode_new = 1;
+			sctx->cur_inode_deleted = 0;
+			sctx->cur_inode_size = btrfs_inode_size(
+					sctx->left_path->nodes[0], left_ii);
+			sctx->cur_inode_mode = btrfs_inode_mode(
+					sctx->left_path->nodes[0], left_ii);
+			sctx->cur_inode_rdev = btrfs_inode_rdev(
+					sctx->left_path->nodes[0], left_ii);
+			ret = send_create_inode_if_needed(sctx);
+			if (ret < 0)
+				goto out;
+
+			ret = process_all_refs(sctx, BTRFS_COMPARE_TREE_NEW);
+			if (ret < 0)
+				goto out;
+			/*
+			 * Advance send_progress now as we did not get into
+			 * process_recorded_refs_if_needed in the new_gen case.
+			 */
+			sctx->send_progress = sctx->cur_ino + 1;
+
+			/*
+			 * Now process all extents and xattrs of the inode as if
+			 * they were all new.
+			 */
+			ret = process_all_extents(sctx);
+			if (ret < 0)
+				goto out;
+			ret = process_all_new_xattrs(sctx);
+			if (ret < 0)
+				goto out;
+		} else {
+			sctx->cur_inode_gen = left_gen;
+			sctx->cur_inode_new = 0;
+			sctx->cur_inode_new_gen = 0;
+			sctx->cur_inode_deleted = 0;
+			sctx->cur_inode_size = btrfs_inode_size(
+					sctx->left_path->nodes[0], left_ii);
+			sctx->cur_inode_mode = btrfs_inode_mode(
+					sctx->left_path->nodes[0], left_ii);
+		}
+	}
+
+out:
+	return ret;
+}
+
+/*
+ * We have to process new refs before deleted refs, but compare_trees gives us
+ * the new and deleted refs mixed. To fix this, we record the new/deleted refs
+ * first and later process them in process_recorded_refs.
+ * For the cur_inode_new_gen case, we skip recording completely because
+ * changed_inode did already initiate processing of refs. The reason for this is
+ * that in this case, compare_tree actually compares the refs of 2 different
+ * inodes. To fix this, process_all_refs is used in changed_inode to handle all
+ * refs of the right tree as deleted and all refs of the left tree as new.
+ */
+static int changed_ref(struct send_ctx *sctx,
+		       enum btrfs_compare_tree_result result)
+{
+	int ret = 0;
+
+	BUG_ON(sctx->cur_ino != sctx->cmp_key->objectid);
+
+	if (!sctx->cur_inode_new_gen &&
+	    sctx->cur_ino != BTRFS_FIRST_FREE_OBJECTID) {
+		if (result == BTRFS_COMPARE_TREE_NEW)
+			ret = record_new_ref(sctx);
+		else if (result == BTRFS_COMPARE_TREE_DELETED)
+			ret = record_deleted_ref(sctx);
+		else if (result == BTRFS_COMPARE_TREE_CHANGED)
+			ret = record_changed_ref(sctx);
+	}
+
+	return ret;
+}
+
+/*
+ * Process new/deleted/changed xattrs. We skip processing in the
+ * cur_inode_new_gen case because changed_inode did already initiate processing
+ * of xattrs. The reason is the same as in changed_ref
+ */
+static int changed_xattr(struct send_ctx *sctx,
+			 enum btrfs_compare_tree_result result)
+{
+	int ret = 0;
+
+	BUG_ON(sctx->cur_ino != sctx->cmp_key->objectid);
+
+	if (!sctx->cur_inode_new_gen && !sctx->cur_inode_deleted) {
+		if (result == BTRFS_COMPARE_TREE_NEW)
+			ret = process_new_xattr(sctx);
+		else if (result == BTRFS_COMPARE_TREE_DELETED)
+			ret = process_deleted_xattr(sctx);
+		else if (result == BTRFS_COMPARE_TREE_CHANGED)
+			ret = process_changed_xattr(sctx);
+	}
+
+	return ret;
+}
+
+/*
+ * Process new/deleted/changed extents. We skip processing in the
+ * cur_inode_new_gen case because changed_inode did already initiate processing
+ * of extents. The reason is the same as in changed_ref
+ */
+static int changed_extent(struct send_ctx *sctx,
+			  enum btrfs_compare_tree_result result)
+{
+	int ret = 0;
+
+	BUG_ON(sctx->cur_ino != sctx->cmp_key->objectid);
+
+	if (!sctx->cur_inode_new_gen && !sctx->cur_inode_deleted) {
+		if (result != BTRFS_COMPARE_TREE_DELETED)
+			ret = process_extent(sctx, sctx->left_path,
+					sctx->cmp_key);
+	}
+
+	return ret;
+}
+
+static int dir_changed(struct send_ctx *sctx, u64 dir)
+{
+	u64 orig_gen, new_gen;
+	int ret;
+
+	ret = get_inode_info(sctx->send_root, dir, NULL, &new_gen, NULL, NULL,
+			     NULL, NULL);
+	if (ret)
+		return ret;
+
+	ret = get_inode_info(sctx->parent_root, dir, NULL, &orig_gen, NULL,
+			     NULL, NULL, NULL);
+	if (ret)
+		return ret;
+
+	return (orig_gen != new_gen) ? 1 : 0;
+}
+
+static int compare_refs(struct send_ctx *sctx, struct btrfs_path *path,
+			struct btrfs_key *key)
+{
+	struct btrfs_inode_extref *extref;
+	struct extent_buffer *leaf;
+	u64 dirid = 0, last_dirid = 0;
+	unsigned long ptr;
+	u32 item_size;
+	u32 cur_offset = 0;
+	int ref_name_len;
+	int ret = 0;
+
+	/* Easy case, just check this one dirid */
+	if (key->type == BTRFS_INODE_REF_KEY) {
+		dirid = key->offset;
+
+		ret = dir_changed(sctx, dirid);
+		goto out;
+	}
+
+	leaf = path->nodes[0];
+	item_size = btrfs_item_size_nr(leaf, path->slots[0]);
+	ptr = btrfs_item_ptr_offset(leaf, path->slots[0]);
+	while (cur_offset < item_size) {
+		extref = (struct btrfs_inode_extref *)(ptr +
+						       cur_offset);
+		dirid = btrfs_inode_extref_parent(leaf, extref);
+		ref_name_len = btrfs_inode_extref_name_len(leaf, extref);
+		cur_offset += ref_name_len + sizeof(*extref);
+		if (dirid == last_dirid)
+			continue;
+		ret = dir_changed(sctx, dirid);
+		if (ret)
+			break;
+		last_dirid = dirid;
+	}
+out:
+	return ret;
+}
+
+/*
+ * Updates compare related fields in sctx and simply forwards to the actual
+ * changed_xxx functions.
+ */
+static int changed_cb(struct btrfs_root *left_root,
+		      struct btrfs_root *right_root,
+		      struct btrfs_path *left_path,
+		      struct btrfs_path *right_path,
+		      struct btrfs_key *key,
+		      enum btrfs_compare_tree_result result,
+		      void *ctx)
+{
+	int ret = 0;
+	struct send_ctx *sctx = ctx;
+
+	if (result == BTRFS_COMPARE_TREE_SAME) {
+		if (key->type == BTRFS_INODE_REF_KEY ||
+		    key->type == BTRFS_INODE_EXTREF_KEY) {
+			ret = compare_refs(sctx, left_path, key);
+			if (!ret)
+				return 0;
+			if (ret < 0)
+				return ret;
+		} else if (key->type == BTRFS_EXTENT_DATA_KEY) {
+			return maybe_send_hole(sctx, left_path, key);
+		} else {
+			return 0;
+		}
+		result = BTRFS_COMPARE_TREE_CHANGED;
+		ret = 0;
+	}
+
+	sctx->left_path = left_path;
+	sctx->right_path = right_path;
+	sctx->cmp_key = key;
+
+	ret = finish_inode_if_needed(sctx, 0);
+	if (ret < 0)
+		goto out;
+
+	/* Ignore non-FS objects */
+	if (key->objectid == BTRFS_FREE_INO_OBJECTID ||
+	    key->objectid == BTRFS_FREE_SPACE_OBJECTID)
+		goto out;
+
+	if (key->type == BTRFS_INODE_ITEM_KEY)
+		ret = changed_inode(sctx, result);
+	else if (key->type == BTRFS_INODE_REF_KEY ||
+		 key->type == BTRFS_INODE_EXTREF_KEY)
+		ret = changed_ref(sctx, result);
+	else if (key->type == BTRFS_XATTR_ITEM_KEY)
+		ret = changed_xattr(sctx, result);
+	else if (key->type == BTRFS_EXTENT_DATA_KEY)
+		ret = changed_extent(sctx, result);
+
+out:
+	return ret;
+}
+
+static int full_send_tree(struct send_ctx *sctx)
+{
+	int ret;
+	struct btrfs_root *send_root = sctx->send_root;
+	struct btrfs_key key;
+	struct btrfs_key found_key;
+	struct btrfs_path *path;
+	struct extent_buffer *eb;
+	int slot;
+
+	path = alloc_path_for_send();
+	if (!path)
+		return -ENOMEM;
+
+	key.objectid = BTRFS_FIRST_FREE_OBJECTID;
+	key.type = BTRFS_INODE_ITEM_KEY;
+	key.offset = 0;
+
+	ret = btrfs_search_slot_for_read(send_root, &key, path, 1, 0);
+	if (ret < 0)
+		goto out;
+	if (ret)
+		goto out_finish;
+
+	while (1) {
+		eb = path->nodes[0];
+		slot = path->slots[0];
+		btrfs_item_key_to_cpu(eb, &found_key, slot);
+
+		ret = changed_cb(send_root, NULL, path, NULL,
+				&found_key, BTRFS_COMPARE_TREE_NEW, sctx);
+		if (ret < 0)
+			goto out;
+
+		key.objectid = found_key.objectid;
+		key.type = found_key.type;
+		key.offset = found_key.offset + 1;
+
+		ret = btrfs_next_item(send_root, path);
+		if (ret < 0)
+			goto out;
+		if (ret) {
+			ret  = 0;
+			break;
+		}
+	}
+
+out_finish:
+	ret = finish_inode_if_needed(sctx, 1);
+
+out:
+	btrfs_free_path(path);
+	return ret;
+}
+
+static int send_subvol(struct send_ctx *sctx)
+{
+	int ret;
+
+	if (!(sctx->flags & BTRFS_SEND_FLAG_OMIT_STREAM_HEADER)) {
+		ret = send_header(sctx);
+		if (ret < 0)
+			goto out;
+	}
+
+	ret = send_subvol_begin(sctx);
+	if (ret < 0)
+		goto out;
+
+	if (sctx->parent_root) {
+		ret = btrfs_compare_trees(sctx->send_root, sctx->parent_root,
+				changed_cb, sctx);
+		if (ret < 0)
+			goto out;
+		ret = finish_inode_if_needed(sctx, 1);
+		if (ret < 0)
+			goto out;
+	} else {
+		ret = full_send_tree(sctx);
+		if (ret < 0)
+			goto out;
+	}
+
+out:
+	free_recorded_refs(sctx);
+	return ret;
+}
+
+static void btrfs_root_dec_send_in_progress(struct btrfs_root* root)
+{
+	spin_lock(&root->root_item_lock);
+	root->send_in_progress--;
+	/*
+	 * Not much left to do, we don't know why it's unbalanced and
+	 * can't blindly reset it to 0.
+	 */
+	if (root->send_in_progress < 0)
+		btrfs_err(root->fs_info,
+			"send_in_progres unbalanced %d root %llu",
+			root->send_in_progress, root->root_key.objectid);
+	spin_unlock(&root->root_item_lock);
+}
+
+long btrfs_ioctl_send(struct file *mnt_file, void __user *arg_)
+{
+	int ret = 0;
+	struct btrfs_root *send_root;
+	struct btrfs_root *clone_root;
+	struct btrfs_fs_info *fs_info;
+	struct btrfs_ioctl_send_args *arg = NULL;
+	struct btrfs_key key;
+	struct send_ctx *sctx = NULL;
+	u32 i;
+	u64 *clone_sources_tmp = NULL;
+	int clone_sources_to_rollback = 0;
+	int sort_clone_roots = 0;
+	int index;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return -EPERM;
+
+	send_root = BTRFS_I(file_inode(mnt_file))->root;
+	fs_info = send_root->fs_info;
+
+	/*
+	 * The subvolume must remain read-only during send, protect against
+	 * making it RW. This also protects against deletion.
+	 */
+	spin_lock(&send_root->root_item_lock);
+	send_root->send_in_progress++;
+	spin_unlock(&send_root->root_item_lock);
+
+	/*
+	 * This is done when we lookup the root, it should already be complete
+	 * by the time we get here.
+	 */
+	WARN_ON(send_root->orphan_cleanup_state != ORPHAN_CLEANUP_DONE);
+
+	/*
+	 * Userspace tools do the checks and warn the user if it's
+	 * not RO.
+	 */
+	if (!btrfs_root_readonly(send_root)) {
+		ret = -EPERM;
+		goto out;
+	}
+
+	arg = memdup_user(arg_, sizeof(*arg));
+	if (IS_ERR(arg)) {
+		ret = PTR_ERR(arg);
+		arg = NULL;
+		goto out;
+	}
+
+	if (!access_ok(VERIFY_READ, arg->clone_sources,
+			sizeof(*arg->clone_sources) *
+			arg->clone_sources_count)) {
+		ret = -EFAULT;
+		goto out;
+	}
+
+	if (arg->flags & ~BTRFS_SEND_FLAG_MASK) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	sctx = kzalloc(sizeof(struct send_ctx), GFP_NOFS);
+	if (!sctx) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	INIT_LIST_HEAD(&sctx->new_refs);
+	INIT_LIST_HEAD(&sctx->deleted_refs);
+	INIT_RADIX_TREE(&sctx->name_cache, GFP_NOFS);
+	INIT_LIST_HEAD(&sctx->name_cache_list);
+
+	sctx->flags = arg->flags;
+
+	sctx->send_filp = fget(arg->send_fd);
+	if (!sctx->send_filp) {
+		ret = -EBADF;
+		goto out;
+	}
+
+	sctx->send_root = send_root;
+	/*
+	 * Unlikely but possible, if the subvolume is marked for deletion but
+	 * is slow to remove the directory entry, send can still be started
+	 */
+	if (btrfs_root_dead(sctx->send_root)) {
+		ret = -EPERM;
+		goto out;
+	}
+
+	sctx->clone_roots_cnt = arg->clone_sources_count;
+
+	sctx->send_max_size = BTRFS_SEND_BUF_SIZE;
+	sctx->send_buf = vmalloc(sctx->send_max_size);
+	if (!sctx->send_buf) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	sctx->read_buf = vmalloc(BTRFS_SEND_READ_SIZE);
+	if (!sctx->read_buf) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	sctx->pending_dir_moves = RB_ROOT;
+	sctx->waiting_dir_moves = RB_ROOT;
+	sctx->orphan_dirs = RB_ROOT;
+
+	sctx->clone_roots = vzalloc(sizeof(struct clone_root) *
+			(arg->clone_sources_count + 1));
+	if (!sctx->clone_roots) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	if (arg->clone_sources_count) {
+		clone_sources_tmp = vmalloc(arg->clone_sources_count *
+				sizeof(*arg->clone_sources));
+		if (!clone_sources_tmp) {
+			ret = -ENOMEM;
+			goto out;
+		}
+
+		ret = copy_from_user(clone_sources_tmp, arg->clone_sources,
+				arg->clone_sources_count *
+				sizeof(*arg->clone_sources));
+		if (ret) {
+			ret = -EFAULT;
+			goto out;
+		}
+
+		for (i = 0; i < arg->clone_sources_count; i++) {
+			key.objectid = clone_sources_tmp[i];
+			key.type = BTRFS_ROOT_ITEM_KEY;
+			key.offset = (u64)-1;
+
+			index = srcu_read_lock(&fs_info->subvol_srcu);
+
+			clone_root = btrfs_read_fs_root_no_name(fs_info, &key);
+			if (IS_ERR(clone_root)) {
+				srcu_read_unlock(&fs_info->subvol_srcu, index);
+				ret = PTR_ERR(clone_root);
+				goto out;
+			}
+			clone_sources_to_rollback = i + 1;
+			spin_lock(&clone_root->root_item_lock);
+			clone_root->send_in_progress++;
+			if (!btrfs_root_readonly(clone_root)) {
+				spin_unlock(&clone_root->root_item_lock);
+				srcu_read_unlock(&fs_info->subvol_srcu, index);
+				ret = -EPERM;
+				goto out;
+			}
+			spin_unlock(&clone_root->root_item_lock);
+			srcu_read_unlock(&fs_info->subvol_srcu, index);
+
+			sctx->clone_roots[i].root = clone_root;
+		}
+		vfree(clone_sources_tmp);
+		clone_sources_tmp = NULL;
+	}
+
+	if (arg->parent_root) {
+		key.objectid = arg->parent_root;
+		key.type = BTRFS_ROOT_ITEM_KEY;
+		key.offset = (u64)-1;
+
+		index = srcu_read_lock(&fs_info->subvol_srcu);
+
+		sctx->parent_root = btrfs_read_fs_root_no_name(fs_info, &key);
+		if (IS_ERR(sctx->parent_root)) {
+			srcu_read_unlock(&fs_info->subvol_srcu, index);
+			ret = PTR_ERR(sctx->parent_root);
+			goto out;
+		}
+
+		spin_lock(&sctx->parent_root->root_item_lock);
+		sctx->parent_root->send_in_progress++;
+		if (!btrfs_root_readonly(sctx->parent_root) ||
+				btrfs_root_dead(sctx->parent_root)) {
+			spin_unlock(&sctx->parent_root->root_item_lock);
+			srcu_read_unlock(&fs_info->subvol_srcu, index);
+			ret = -EPERM;
+			goto out;
+		}
+		spin_unlock(&sctx->parent_root->root_item_lock);
+
+		srcu_read_unlock(&fs_info->subvol_srcu, index);
+	}
+
+	/*
+	 * Clones from send_root are allowed, but only if the clone source
+	 * is behind the current send position. This is checked while searching
+	 * for possible clone sources.
+	 */
+	sctx->clone_roots[sctx->clone_roots_cnt++].root = sctx->send_root;
+
+	/* We do a bsearch later */
+	sort(sctx->clone_roots, sctx->clone_roots_cnt,
+			sizeof(*sctx->clone_roots), __clone_root_cmp_sort,
+			NULL);
+	sort_clone_roots = 1;
+
+	current->journal_info = (void *)BTRFS_SEND_TRANS_STUB;
+	ret = send_subvol(sctx);
+	current->journal_info = NULL;
+	if (ret < 0)
+		goto out;
+
+	if (!(sctx->flags & BTRFS_SEND_FLAG_OMIT_END_CMD)) {
+		ret = begin_cmd(sctx, BTRFS_SEND_C_END);
+		if (ret < 0)
+			goto out;
+		ret = send_cmd(sctx);
+		if (ret < 0)
+			goto out;
+	}
+
+out:
+	WARN_ON(sctx && !ret && !RB_EMPTY_ROOT(&sctx->pending_dir_moves));
+	while (sctx && !RB_EMPTY_ROOT(&sctx->pending_dir_moves)) {
+		struct rb_node *n;
+		struct pending_dir_move *pm;
+
+		n = rb_first(&sctx->pending_dir_moves);
+		pm = rb_entry(n, struct pending_dir_move, node);
+		while (!list_empty(&pm->list)) {
+			struct pending_dir_move *pm2;
+
+			pm2 = list_first_entry(&pm->list,
+					       struct pending_dir_move, list);
+			free_pending_move(sctx, pm2);
+		}
+		free_pending_move(sctx, pm);
+	}
+
+	WARN_ON(sctx && !ret && !RB_EMPTY_ROOT(&sctx->waiting_dir_moves));
+	while (sctx && !RB_EMPTY_ROOT(&sctx->waiting_dir_moves)) {
+		struct rb_node *n;
+		struct waiting_dir_move *dm;
+
+		n = rb_first(&sctx->waiting_dir_moves);
+		dm = rb_entry(n, struct waiting_dir_move, node);
+		rb_erase(&dm->node, &sctx->waiting_dir_moves);
+		kfree(dm);
+	}
+
+	WARN_ON(sctx && !ret && !RB_EMPTY_ROOT(&sctx->orphan_dirs));
+	while (sctx && !RB_EMPTY_ROOT(&sctx->orphan_dirs)) {
+		struct rb_node *n;
+		struct orphan_dir_info *odi;
+
+		n = rb_first(&sctx->orphan_dirs);
+		odi = rb_entry(n, struct orphan_dir_info, node);
+		free_orphan_dir_info(sctx, odi);
+	}
+
+	if (sort_clone_roots) {
+		for (i = 0; i < sctx->clone_roots_cnt; i++)
+			btrfs_root_dec_send_in_progress(
+					sctx->clone_roots[i].root);
+	} else {
+		for (i = 0; sctx && i < clone_sources_to_rollback; i++)
+			btrfs_root_dec_send_in_progress(
+					sctx->clone_roots[i].root);
+
+		btrfs_root_dec_send_in_progress(send_root);
+	}
+	if (sctx && !IS_ERR_OR_NULL(sctx->parent_root))
+		btrfs_root_dec_send_in_progress(sctx->parent_root);
+
+	kfree(arg);
+	vfree(clone_sources_tmp);
+
+	if (sctx) {
+		if (sctx->send_filp)
+			fput(sctx->send_filp);
+
+		vfree(sctx->clone_roots);
+		vfree(sctx->send_buf);
+		vfree(sctx->read_buf);
+
+		name_cache_free(sctx);
+
+		kfree(sctx);
+	}
+
+	return ret;
+}
diff --git a/fs/btrfs/send.h b/fs/btrfs/send.h
new file mode 100644
index 00000000000..48d425aef05
--- /dev/null
+++ b/fs/btrfs/send.h
@@ -0,0 +1,134 @@
+/*
+ * Copyright (C) 2012 Alexander Block.  All rights reserved.
+ * Copyright (C) 2012 STRATO.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include "ctree.h"
+
+#define BTRFS_SEND_STREAM_MAGIC "btrfs-stream"
+#define BTRFS_SEND_STREAM_VERSION 1
+
+#define BTRFS_SEND_BUF_SIZE (1024 * 64)
+#define BTRFS_SEND_READ_SIZE (1024 * 48)
+
+enum btrfs_tlv_type {
+	BTRFS_TLV_U8,
+	BTRFS_TLV_U16,
+	BTRFS_TLV_U32,
+	BTRFS_TLV_U64,
+	BTRFS_TLV_BINARY,
+	BTRFS_TLV_STRING,
+	BTRFS_TLV_UUID,
+	BTRFS_TLV_TIMESPEC,
+};
+
+struct btrfs_stream_header {
+	char magic[sizeof(BTRFS_SEND_STREAM_MAGIC)];
+	__le32 version;
+} __attribute__ ((__packed__));
+
+struct btrfs_cmd_header {
+	/* len excluding the header */
+	__le32 len;
+	__le16 cmd;
+	/* crc including the header with zero crc field */
+	__le32 crc;
+} __attribute__ ((__packed__));
+
+struct btrfs_tlv_header {
+	__le16 tlv_type;
+	/* len excluding the header */
+	__le16 tlv_len;
+} __attribute__ ((__packed__));
+
+/* commands */
+enum btrfs_send_cmd {
+	BTRFS_SEND_C_UNSPEC,
+
+	BTRFS_SEND_C_SUBVOL,
+	BTRFS_SEND_C_SNAPSHOT,
+
+	BTRFS_SEND_C_MKFILE,
+	BTRFS_SEND_C_MKDIR,
+	BTRFS_SEND_C_MKNOD,
+	BTRFS_SEND_C_MKFIFO,
+	BTRFS_SEND_C_MKSOCK,
+	BTRFS_SEND_C_SYMLINK,
+
+	BTRFS_SEND_C_RENAME,
+	BTRFS_SEND_C_LINK,
+	BTRFS_SEND_C_UNLINK,
+	BTRFS_SEND_C_RMDIR,
+
+	BTRFS_SEND_C_SET_XATTR,
+	BTRFS_SEND_C_REMOVE_XATTR,
+
+	BTRFS_SEND_C_WRITE,
+	BTRFS_SEND_C_CLONE,
+
+	BTRFS_SEND_C_TRUNCATE,
+	BTRFS_SEND_C_CHMOD,
+	BTRFS_SEND_C_CHOWN,
+	BTRFS_SEND_C_UTIMES,
+
+	BTRFS_SEND_C_END,
+	BTRFS_SEND_C_UPDATE_EXTENT,
+	__BTRFS_SEND_C_MAX,
+};
+#define BTRFS_SEND_C_MAX (__BTRFS_SEND_C_MAX - 1)
+
+/* attributes in send stream */
+enum {
+	BTRFS_SEND_A_UNSPEC,
+
+	BTRFS_SEND_A_UUID,
+	BTRFS_SEND_A_CTRANSID,
+
+	BTRFS_SEND_A_INO,
+	BTRFS_SEND_A_SIZE,
+	BTRFS_SEND_A_MODE,
+	BTRFS_SEND_A_UID,
+	BTRFS_SEND_A_GID,
+	BTRFS_SEND_A_RDEV,
+	BTRFS_SEND_A_CTIME,
+	BTRFS_SEND_A_MTIME,
+	BTRFS_SEND_A_ATIME,
+	BTRFS_SEND_A_OTIME,
+
+	BTRFS_SEND_A_XATTR_NAME,
+	BTRFS_SEND_A_XATTR_DATA,
+
+	BTRFS_SEND_A_PATH,
+	BTRFS_SEND_A_PATH_TO,
+	BTRFS_SEND_A_PATH_LINK,
+
+	BTRFS_SEND_A_FILE_OFFSET,
+	BTRFS_SEND_A_DATA,
+
+	BTRFS_SEND_A_CLONE_UUID,
+	BTRFS_SEND_A_CLONE_CTRANSID,
+	BTRFS_SEND_A_CLONE_PATH,
+	BTRFS_SEND_A_CLONE_OFFSET,
+	BTRFS_SEND_A_CLONE_LEN,
+
+	__BTRFS_SEND_A_MAX,
+};
+#define BTRFS_SEND_A_MAX (__BTRFS_SEND_A_MAX - 1)
+
+#ifdef __KERNEL__
+long btrfs_ioctl_send(struct file *mnt_file, void __user *arg);
+#endif
diff --git a/fs/btrfs/struct-funcs.c b/fs/btrfs/struct-funcs.c
index c0f7ecaf1e7..b976597b072 100644
--- a/fs/btrfs/struct-funcs.c
+++ b/fs/btrfs/struct-funcs.c
@@ -17,15 +17,27 @@
  */
 
 #include <linux/highmem.h>
+#include <asm/unaligned.h>
 
-/* this is some deeply nasty code.  ctree.h has a different
- * definition for this BTRFS_SETGET_FUNCS macro, behind a #ifndef
+#include "ctree.h"
+
+static inline u8 get_unaligned_le8(const void *p)
+{
+       return *(u8 *)p;
+}
+
+static inline void put_unaligned_le8(u8 val, void *p)
+{
+       *(u8 *)p = val;
+}
+
+/*
+ * this is some deeply nasty code.
  *
  * The end result is that anyone who #includes ctree.h gets a
- * declaration for the btrfs_set_foo functions and btrfs_foo functions
- *
- * This file declares the macros and then #includes ctree.h, which results
- * in cpp creating the function here based on the template below.
+ * declaration for the btrfs_set_foo functions and btrfs_foo functions,
+ * which are wappers of btrfs_set_token_#bits functions and
+ * btrfs_get_token_#bits functions, which are defined in this file.
  *
  * These setget functions do all the extent_buffer related mapping
  * required to efficiently read and write specific fields in the extent
@@ -33,107 +45,98 @@
  * an unsigned long offset into the extent buffer which has been
  * cast to a specific type.  This gives us all the gcc type checking.
  *
- * The extent buffer api is used to do all the kmapping and page
- * spanning work required to get extent buffers in highmem and have
- * a metadata blocksize different from the page size.
- *
- * The macro starts with a simple function prototype declaration so that
- * sparse won't complain about it being static.
+ * The extent buffer api is used to do the page spanning work required to
+ * have a metadata blocksize different from the page size.
  */
 
-#define BTRFS_SETGET_FUNCS(name, type, member, bits)			\
-u##bits btrfs_##name(struct extent_buffer *eb, type *s);		\
-void btrfs_set_##name(struct extent_buffer *eb, type *s, u##bits val);	\
-u##bits btrfs_##name(struct extent_buffer *eb,				\
-				   type *s)				\
+#define DEFINE_BTRFS_SETGET_BITS(bits)					\
+u##bits btrfs_get_token_##bits(struct extent_buffer *eb, void *ptr,	\
+			       unsigned long off,			\
+			       struct btrfs_map_token *token)		\
 {									\
-	unsigned long part_offset = (unsigned long)s;			\
-	unsigned long offset = part_offset + offsetof(type, member);	\
-	type *p;							\
-	/* ugly, but we want the fast path here */			\
-	if (eb->map_token && offset >= eb->map_start &&			\
-	    offset + sizeof(((type *)0)->member) <= eb->map_start +	\
-	    eb->map_len) {						\
-		p = (type *)(eb->kaddr + part_offset - eb->map_start);	\
-		return le##bits##_to_cpu(p->member);			\
-	}								\
-	{								\
-		int err;						\
-		char *map_token;					\
-		char *kaddr;						\
-		int unmap_on_exit = (eb->map_token == NULL);		\
-		unsigned long map_start;				\
-		unsigned long map_len;					\
-		u##bits res;						\
-		err = map_extent_buffer(eb, offset,			\
-				sizeof(((type *)0)->member),		\
-				&map_token, &kaddr,			\
-				&map_start, &map_len, KM_USER1);	\
-		if (err) {						\
-			__le##bits leres;				\
-			read_eb_member(eb, s, type, member, &leres);	\
-			return le##bits##_to_cpu(leres);		\
-		}							\
-		p = (type *)(kaddr + part_offset - map_start);		\
-		res = le##bits##_to_cpu(p->member);			\
-		if (unmap_on_exit)					\
-			unmap_extent_buffer(eb, map_token, KM_USER1);	\
+	unsigned long part_offset = (unsigned long)ptr;			\
+	unsigned long offset = part_offset + off;			\
+	void *p;							\
+	int err;							\
+	char *kaddr;							\
+	unsigned long map_start;					\
+	unsigned long map_len;						\
+	int size = sizeof(u##bits);					\
+	u##bits res;							\
+									\
+	if (token && token->kaddr && token->offset <= offset &&		\
+	    token->eb == eb &&						\
+	   (token->offset + PAGE_CACHE_SIZE >= offset + size)) {	\
+		kaddr = token->kaddr;					\
+		p = kaddr + part_offset - token->offset;		\
+		res = get_unaligned_le##bits(p + off);			\
 		return res;						\
 	}								\
+	err = map_private_extent_buffer(eb, offset, size,		\
+					&kaddr, &map_start, &map_len);	\
+	if (err) {							\
+		__le##bits leres;					\
+									\
+		read_extent_buffer(eb, &leres, offset, size);		\
+		return le##bits##_to_cpu(leres);			\
+	}								\
+	p = kaddr + part_offset - map_start;				\
+	res = get_unaligned_le##bits(p + off);				\
+	if (token) {							\
+		token->kaddr = kaddr;					\
+		token->offset = map_start;				\
+		token->eb = eb;						\
+	}								\
+	return res;							\
 }									\
-void btrfs_set_##name(struct extent_buffer *eb,				\
-				    type *s, u##bits val)		\
+void btrfs_set_token_##bits(struct extent_buffer *eb,			\
+			    void *ptr, unsigned long off, u##bits val,	\
+			    struct btrfs_map_token *token)		\
 {									\
-	unsigned long part_offset = (unsigned long)s;			\
-	unsigned long offset = part_offset + offsetof(type, member);	\
-	type *p;							\
-	/* ugly, but we want the fast path here */			\
-	if (eb->map_token && offset >= eb->map_start &&			\
-	    offset + sizeof(((type *)0)->member) <= eb->map_start +	\
-	    eb->map_len) {						\
-		p = (type *)(eb->kaddr + part_offset - eb->map_start);	\
-		p->member = cpu_to_le##bits(val);			\
+	unsigned long part_offset = (unsigned long)ptr;			\
+	unsigned long offset = part_offset + off;			\
+	void *p;							\
+	int err;							\
+	char *kaddr;							\
+	unsigned long map_start;					\
+	unsigned long map_len;						\
+	int size = sizeof(u##bits);					\
+									\
+	if (token && token->kaddr && token->offset <= offset &&		\
+	    token->eb == eb &&						\
+	   (token->offset + PAGE_CACHE_SIZE >= offset + size)) {	\
+		kaddr = token->kaddr;					\
+		p = kaddr + part_offset - token->offset;		\
+		put_unaligned_le##bits(val, p + off);			\
+		return;							\
+	}								\
+	err = map_private_extent_buffer(eb, offset, size,		\
+			&kaddr, &map_start, &map_len);			\
+	if (err) {							\
+		__le##bits val2;					\
+									\
+		val2 = cpu_to_le##bits(val);				\
+		write_extent_buffer(eb, &val2, offset, size);		\
 		return;							\
 	}								\
-	{								\
-		int err;						\
-		char *map_token;					\
-		char *kaddr;						\
-		int unmap_on_exit = (eb->map_token == NULL);		\
-		unsigned long map_start;				\
-		unsigned long map_len;					\
-		err = map_extent_buffer(eb, offset,			\
-				sizeof(((type *)0)->member),		\
-				&map_token, &kaddr,			\
-				&map_start, &map_len, KM_USER1);	\
-		if (err) {						\
-			__le##bits val2;				\
-			val2 = cpu_to_le##bits(val);			\
-			write_eb_member(eb, s, type, member, &val2);	\
-			return;						\
-		}							\
-		p = (type *)(kaddr + part_offset - map_start);		\
-		p->member = cpu_to_le##bits(val);			\
-		if (unmap_on_exit)					\
-			unmap_extent_buffer(eb, map_token, KM_USER1);	\
+	p = kaddr + part_offset - map_start;				\
+	put_unaligned_le##bits(val, p + off);				\
+	if (token) {							\
+		token->kaddr = kaddr;					\
+		token->offset = map_start;				\
+		token->eb = eb;						\
 	}								\
 }
 
-#include "ctree.h"
+DEFINE_BTRFS_SETGET_BITS(8)
+DEFINE_BTRFS_SETGET_BITS(16)
+DEFINE_BTRFS_SETGET_BITS(32)
+DEFINE_BTRFS_SETGET_BITS(64)
 
 void btrfs_node_key(struct extent_buffer *eb,
 		    struct btrfs_disk_key *disk_key, int nr)
 {
 	unsigned long ptr = btrfs_node_key_ptr_offset(nr);
-	if (eb->map_token && ptr >= eb->map_start &&
-	    ptr + sizeof(*disk_key) <= eb->map_start + eb->map_len) {
-		memcpy(disk_key, eb->kaddr + ptr - eb->map_start,
-			sizeof(*disk_key));
-		return;
-	} else if (eb->map_token) {
-		unmap_extent_buffer(eb, eb->map_token, KM_USER1);
-		eb->map_token = NULL;
-	}
 	read_eb_member(eb, (struct btrfs_key_ptr *)ptr,
 		       struct btrfs_key_ptr, key, disk_key);
 }
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 1866dff0538..8e16bca69c5 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -39,76 +39,375 @@
 #include <linux/miscdevice.h>
 #include <linux/magic.h>
 #include <linux/slab.h>
-#include "compat.h"
+#include <linux/cleancache.h>
+#include <linux/ratelimit.h>
+#include <linux/btrfs.h>
+#include "delayed-inode.h"
 #include "ctree.h"
 #include "disk-io.h"
 #include "transaction.h"
 #include "btrfs_inode.h"
-#include "ioctl.h"
 #include "print-tree.h"
+#include "hash.h"
+#include "props.h"
 #include "xattr.h"
 #include "volumes.h"
-#include "version.h"
 #include "export.h"
 #include "compression.h"
+#include "rcu-string.h"
+#include "dev-replace.h"
+#include "free-space-cache.h"
+#include "backref.h"
+#include "tests/btrfs-tests.h"
+
+#define CREATE_TRACE_POINTS
+#include <trace/events/btrfs.h>
 
 static const struct super_operations btrfs_super_ops;
+static struct file_system_type btrfs_fs_type;
 
-static void btrfs_put_super(struct super_block *sb)
+static int btrfs_remount(struct super_block *sb, int *flags, char *data);
+
+static const char *btrfs_decode_error(int errno)
 {
-	struct btrfs_root *root = btrfs_sb(sb);
-	int ret;
+	char *errstr = "unknown";
+
+	switch (errno) {
+	case -EIO:
+		errstr = "IO failure";
+		break;
+	case -ENOMEM:
+		errstr = "Out of memory";
+		break;
+	case -EROFS:
+		errstr = "Readonly filesystem";
+		break;
+	case -EEXIST:
+		errstr = "Object already exists";
+		break;
+	case -ENOSPC:
+		errstr = "No space left";
+		break;
+	case -ENOENT:
+		errstr = "No such entry";
+		break;
+	}
+
+	return errstr;
+}
+
+static void save_error_info(struct btrfs_fs_info *fs_info)
+{
+	/*
+	 * today we only save the error info into ram.  Long term we'll
+	 * also send it down to the disk
+	 */
+	set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
+}
+
+/* btrfs handle error by forcing the filesystem readonly */
+static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
+{
+	struct super_block *sb = fs_info->sb;
+
+	if (sb->s_flags & MS_RDONLY)
+		return;
+
+	if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
+		sb->s_flags |= MS_RDONLY;
+		btrfs_info(fs_info, "forced readonly");
+		/*
+		 * Note that a running device replace operation is not
+		 * canceled here although there is no way to update
+		 * the progress. It would add the risk of a deadlock,
+		 * therefore the canceling is ommited. The only penalty
+		 * is that some I/O remains active until the procedure
+		 * completes. The next time when the filesystem is
+		 * mounted writeable again, the device replace
+		 * operation continues.
+		 */
+	}
+}
+
+#ifdef CONFIG_PRINTK
+/*
+ * __btrfs_std_error decodes expected errors from the caller and
+ * invokes the approciate error response.
+ */
+void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
+		       unsigned int line, int errno, const char *fmt, ...)
+{
+	struct super_block *sb = fs_info->sb;
+	const char *errstr;
+
+	/*
+	 * Special case: if the error is EROFS, and we're already
+	 * under MS_RDONLY, then it is safe here.
+	 */
+	if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
+  		return;
+
+	errstr = btrfs_decode_error(errno);
+	if (fmt) {
+		struct va_format vaf;
+		va_list args;
+
+		va_start(args, fmt);
+		vaf.fmt = fmt;
+		vaf.va = &args;
+
+		printk(KERN_CRIT
+			"BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
+			sb->s_id, function, line, errno, errstr, &vaf);
+		va_end(args);
+	} else {
+		printk(KERN_CRIT "BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
+			sb->s_id, function, line, errno, errstr);
+	}
+
+	/* Don't go through full error handling during mount */
+	save_error_info(fs_info);
+	if (sb->s_flags & MS_BORN)
+		btrfs_handle_error(fs_info);
+}
+
+static const char * const logtypes[] = {
+	"emergency",
+	"alert",
+	"critical",
+	"error",
+	"warning",
+	"notice",
+	"info",
+	"debug",
+};
+
+void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
+{
+	struct super_block *sb = fs_info->sb;
+	char lvl[4];
+	struct va_format vaf;
+	va_list args;
+	const char *type = logtypes[4];
+	int kern_level;
+
+	va_start(args, fmt);
+
+	kern_level = printk_get_level(fmt);
+	if (kern_level) {
+		size_t size = printk_skip_level(fmt) - fmt;
+		memcpy(lvl, fmt,  size);
+		lvl[size] = '\0';
+		fmt += size;
+		type = logtypes[kern_level - '0'];
+	} else
+		*lvl = '\0';
+
+	vaf.fmt = fmt;
+	vaf.va = &args;
+
+	printk("%sBTRFS %s (device %s): %pV\n", lvl, type, sb->s_id, &vaf);
+
+	va_end(args);
+}
+
+#else
+
+void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
+		       unsigned int line, int errno, const char *fmt, ...)
+{
+	struct super_block *sb = fs_info->sb;
+
+	/*
+	 * Special case: if the error is EROFS, and we're already
+	 * under MS_RDONLY, then it is safe here.
+	 */
+	if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
+		return;
+
+	/* Don't go through full error handling during mount */
+	if (sb->s_flags & MS_BORN) {
+		save_error_info(fs_info);
+		btrfs_handle_error(fs_info);
+	}
+}
+#endif
+
+/*
+ * We only mark the transaction aborted and then set the file system read-only.
+ * This will prevent new transactions from starting or trying to join this
+ * one.
+ *
+ * This means that error recovery at the call site is limited to freeing
+ * any local memory allocations and passing the error code up without
+ * further cleanup. The transaction should complete as it normally would
+ * in the call path but will return -EIO.
+ *
+ * We'll complete the cleanup in btrfs_end_transaction and
+ * btrfs_commit_transaction.
+ */
+void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
+			       struct btrfs_root *root, const char *function,
+			       unsigned int line, int errno)
+{
+	/*
+	 * Report first abort since mount
+	 */
+	if (!test_and_set_bit(BTRFS_FS_STATE_TRANS_ABORTED,
+				&root->fs_info->fs_state)) {
+		WARN(1, KERN_DEBUG "BTRFS: Transaction aborted (error %d)\n",
+				errno);
+	}
+	trans->aborted = errno;
+	/* Nothing used. The other threads that have joined this
+	 * transaction may be able to continue. */
+	if (!trans->blocks_used) {
+		const char *errstr;
+
+		errstr = btrfs_decode_error(errno);
+		btrfs_warn(root->fs_info,
+		           "%s:%d: Aborting unused transaction(%s).",
+		           function, line, errstr);
+		return;
+	}
+	ACCESS_ONCE(trans->transaction->aborted) = errno;
+	/* Wake up anybody who may be waiting on this transaction */
+	wake_up(&root->fs_info->transaction_wait);
+	wake_up(&root->fs_info->transaction_blocked_wait);
+	__btrfs_std_error(root->fs_info, function, line, errno, NULL);
+}
+/*
+ * __btrfs_panic decodes unexpected, fatal errors from the caller,
+ * issues an alert, and either panics or BUGs, depending on mount options.
+ */
+void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
+		   unsigned int line, int errno, const char *fmt, ...)
+{
+	char *s_id = "<unknown>";
+	const char *errstr;
+	struct va_format vaf = { .fmt = fmt };
+	va_list args;
+
+	if (fs_info)
+		s_id = fs_info->sb->s_id;
+
+	va_start(args, fmt);
+	vaf.va = &args;
+
+	errstr = btrfs_decode_error(errno);
+	if (fs_info && (fs_info->mount_opt & BTRFS_MOUNT_PANIC_ON_FATAL_ERROR))
+		panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
+			s_id, function, line, &vaf, errno, errstr);
+
+	btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
+		   function, line, &vaf, errno, errstr);
+	va_end(args);
+	/* Caller calls BUG() */
+}
 
-	ret = close_ctree(root);
-	sb->s_fs_info = NULL;
+static void btrfs_put_super(struct super_block *sb)
+{
+	(void)close_ctree(btrfs_sb(sb)->tree_root);
+	/* FIXME: need to fix VFS to return error? */
+	/* AV: return it _where_?  ->put_super() can be triggered by any number
+	 * of async events, up to and including delivery of SIGKILL to the
+	 * last process that kept it busy.  Or segfault in the aforementioned
+	 * process...  Whom would you report that to?
+	 */
 }
 
 enum {
 	Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
 	Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
 	Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
-	Opt_compress_force, Opt_notreelog, Opt_ratio, Opt_flushoncommit,
-	Opt_discard, Opt_err,
+	Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
+	Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
+	Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed,
+	Opt_enospc_debug, Opt_subvolrootid, Opt_defrag, Opt_inode_cache,
+	Opt_no_space_cache, Opt_recovery, Opt_skip_balance,
+	Opt_check_integrity, Opt_check_integrity_including_extent_data,
+	Opt_check_integrity_print_mask, Opt_fatal_errors, Opt_rescan_uuid_tree,
+	Opt_commit_interval, Opt_barrier, Opt_nodefrag, Opt_nodiscard,
+	Opt_noenospc_debug, Opt_noflushoncommit, Opt_acl, Opt_datacow,
+	Opt_datasum, Opt_treelog, Opt_noinode_cache,
+	Opt_err,
 };
 
 static match_table_t tokens = {
 	{Opt_degraded, "degraded"},
 	{Opt_subvol, "subvol=%s"},
-	{Opt_subvolid, "subvolid=%d"},
+	{Opt_subvolid, "subvolid=%s"},
 	{Opt_device, "device=%s"},
 	{Opt_nodatasum, "nodatasum"},
+	{Opt_datasum, "datasum"},
 	{Opt_nodatacow, "nodatacow"},
+	{Opt_datacow, "datacow"},
 	{Opt_nobarrier, "nobarrier"},
+	{Opt_barrier, "barrier"},
 	{Opt_max_inline, "max_inline=%s"},
 	{Opt_alloc_start, "alloc_start=%s"},
 	{Opt_thread_pool, "thread_pool=%d"},
 	{Opt_compress, "compress"},
+	{Opt_compress_type, "compress=%s"},
 	{Opt_compress_force, "compress-force"},
+	{Opt_compress_force_type, "compress-force=%s"},
 	{Opt_ssd, "ssd"},
 	{Opt_ssd_spread, "ssd_spread"},
 	{Opt_nossd, "nossd"},
+	{Opt_acl, "acl"},
 	{Opt_noacl, "noacl"},
 	{Opt_notreelog, "notreelog"},
+	{Opt_treelog, "treelog"},
 	{Opt_flushoncommit, "flushoncommit"},
+	{Opt_noflushoncommit, "noflushoncommit"},
 	{Opt_ratio, "metadata_ratio=%d"},
 	{Opt_discard, "discard"},
+	{Opt_nodiscard, "nodiscard"},
+	{Opt_space_cache, "space_cache"},
+	{Opt_clear_cache, "clear_cache"},
+	{Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
+	{Opt_enospc_debug, "enospc_debug"},
+	{Opt_noenospc_debug, "noenospc_debug"},
+	{Opt_subvolrootid, "subvolrootid=%d"},
+	{Opt_defrag, "autodefrag"},
+	{Opt_nodefrag, "noautodefrag"},
+	{Opt_inode_cache, "inode_cache"},
+	{Opt_noinode_cache, "noinode_cache"},
+	{Opt_no_space_cache, "nospace_cache"},
+	{Opt_recovery, "recovery"},
+	{Opt_skip_balance, "skip_balance"},
+	{Opt_check_integrity, "check_int"},
+	{Opt_check_integrity_including_extent_data, "check_int_data"},
+	{Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
+	{Opt_rescan_uuid_tree, "rescan_uuid_tree"},
+	{Opt_fatal_errors, "fatal_errors=%s"},
+	{Opt_commit_interval, "commit=%d"},
 	{Opt_err, NULL},
 };
 
 /*
  * Regular mount options parser.  Everything that is needed only when
  * reading in a new superblock is parsed here.
+ * XXX JDM: This needs to be cleaned up for remount.
  */
 int btrfs_parse_options(struct btrfs_root *root, char *options)
 {
 	struct btrfs_fs_info *info = root->fs_info;
 	substring_t args[MAX_OPT_ARGS];
-	char *p, *num, *orig;
+	char *p, *num, *orig = NULL;
+	u64 cache_gen;
 	int intarg;
 	int ret = 0;
+	char *compress_type;
+	bool compress_force = false;
+	bool compress = false;
+
+	cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
+	if (cache_gen)
+		btrfs_set_opt(info->mount_opt, SPACE_CACHE);
 
 	if (!options)
-		return 0;
+		goto out;
 
 	/*
 	 * strsep changes the string, duplicate it because parse_options
@@ -128,11 +427,12 @@ int btrfs_parse_options(struct btrfs_root *root, char *options)
 		token = match_token(p, tokens, args);
 		switch (token) {
 		case Opt_degraded:
-			printk(KERN_INFO "btrfs: allowing degraded mounts\n");
+			btrfs_info(root->fs_info, "allowing degraded mounts");
 			btrfs_set_opt(info->mount_opt, DEGRADED);
 			break;
 		case Opt_subvol:
 		case Opt_subvolid:
+		case Opt_subvolrootid:
 		case Opt_device:
 			/*
 			 * These are parsed by btrfs_parse_early_options
@@ -140,51 +440,112 @@ int btrfs_parse_options(struct btrfs_root *root, char *options)
 			 */
 			break;
 		case Opt_nodatasum:
-			printk(KERN_INFO "btrfs: setting nodatasum\n");
-			btrfs_set_opt(info->mount_opt, NODATASUM);
+			btrfs_set_and_info(root, NODATASUM,
+					   "setting nodatasum");
+			break;
+		case Opt_datasum:
+			if (btrfs_test_opt(root, NODATASUM)) {
+				if (btrfs_test_opt(root, NODATACOW))
+					btrfs_info(root->fs_info, "setting datasum, datacow enabled");
+				else
+					btrfs_info(root->fs_info, "setting datasum");
+			}
+			btrfs_clear_opt(info->mount_opt, NODATACOW);
+			btrfs_clear_opt(info->mount_opt, NODATASUM);
 			break;
 		case Opt_nodatacow:
-			printk(KERN_INFO "btrfs: setting nodatacow\n");
+			if (!btrfs_test_opt(root, NODATACOW)) {
+				if (!btrfs_test_opt(root, COMPRESS) ||
+				    !btrfs_test_opt(root, FORCE_COMPRESS)) {
+					btrfs_info(root->fs_info,
+						   "setting nodatacow, compression disabled");
+				} else {
+					btrfs_info(root->fs_info, "setting nodatacow");
+				}
+			}
+			btrfs_clear_opt(info->mount_opt, COMPRESS);
+			btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
 			btrfs_set_opt(info->mount_opt, NODATACOW);
 			btrfs_set_opt(info->mount_opt, NODATASUM);
 			break;
-		case Opt_compress:
-			printk(KERN_INFO "btrfs: use compression\n");
-			btrfs_set_opt(info->mount_opt, COMPRESS);
+		case Opt_datacow:
+			btrfs_clear_and_info(root, NODATACOW,
+					     "setting datacow");
 			break;
 		case Opt_compress_force:
-			printk(KERN_INFO "btrfs: forcing compression\n");
-			btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
-			btrfs_set_opt(info->mount_opt, COMPRESS);
+		case Opt_compress_force_type:
+			compress_force = true;
+			/* Fallthrough */
+		case Opt_compress:
+		case Opt_compress_type:
+			compress = true;
+			if (token == Opt_compress ||
+			    token == Opt_compress_force ||
+			    strcmp(args[0].from, "zlib") == 0) {
+				compress_type = "zlib";
+				info->compress_type = BTRFS_COMPRESS_ZLIB;
+				btrfs_set_opt(info->mount_opt, COMPRESS);
+				btrfs_clear_opt(info->mount_opt, NODATACOW);
+				btrfs_clear_opt(info->mount_opt, NODATASUM);
+			} else if (strcmp(args[0].from, "lzo") == 0) {
+				compress_type = "lzo";
+				info->compress_type = BTRFS_COMPRESS_LZO;
+				btrfs_set_opt(info->mount_opt, COMPRESS);
+				btrfs_clear_opt(info->mount_opt, NODATACOW);
+				btrfs_clear_opt(info->mount_opt, NODATASUM);
+				btrfs_set_fs_incompat(info, COMPRESS_LZO);
+			} else if (strncmp(args[0].from, "no", 2) == 0) {
+				compress_type = "no";
+				btrfs_clear_opt(info->mount_opt, COMPRESS);
+				btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
+				compress_force = false;
+			} else {
+				ret = -EINVAL;
+				goto out;
+			}
+
+			if (compress_force) {
+				btrfs_set_and_info(root, FORCE_COMPRESS,
+						   "force %s compression",
+						   compress_type);
+			} else if (compress) {
+				if (!btrfs_test_opt(root, COMPRESS))
+					btrfs_info(root->fs_info,
+						   "btrfs: use %s compression",
+						   compress_type);
+			}
 			break;
 		case Opt_ssd:
-			printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
-			btrfs_set_opt(info->mount_opt, SSD);
+			btrfs_set_and_info(root, SSD,
+					   "use ssd allocation scheme");
 			break;
 		case Opt_ssd_spread:
-			printk(KERN_INFO "btrfs: use spread ssd "
-			       "allocation scheme\n");
+			btrfs_set_and_info(root, SSD_SPREAD,
+					   "use spread ssd allocation scheme");
 			btrfs_set_opt(info->mount_opt, SSD);
-			btrfs_set_opt(info->mount_opt, SSD_SPREAD);
 			break;
 		case Opt_nossd:
-			printk(KERN_INFO "btrfs: not using ssd allocation "
-			       "scheme\n");
-			btrfs_set_opt(info->mount_opt, NOSSD);
+			btrfs_set_and_info(root, NOSSD,
+					     "not using ssd allocation scheme");
 			btrfs_clear_opt(info->mount_opt, SSD);
-			btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
+			break;
+		case Opt_barrier:
+			btrfs_clear_and_info(root, NOBARRIER,
+					     "turning on barriers");
 			break;
 		case Opt_nobarrier:
-			printk(KERN_INFO "btrfs: turning off barriers\n");
-			btrfs_set_opt(info->mount_opt, NOBARRIER);
+			btrfs_set_and_info(root, NOBARRIER,
+					   "turning off barriers");
 			break;
 		case Opt_thread_pool:
-			intarg = 0;
-			match_int(&args[0], &intarg);
-			if (intarg) {
+			ret = match_int(&args[0], &intarg);
+			if (ret) {
+				goto out;
+			} else if (intarg > 0) {
 				info->thread_pool_size = intarg;
-				printk(KERN_INFO "btrfs: thread pool %d\n",
-				       info->thread_pool_size);
+			} else {
+				ret = -EINVAL;
+				goto out;
 			}
 			break;
 		case Opt_max_inline:
@@ -194,50 +555,196 @@ int btrfs_parse_options(struct btrfs_root *root, char *options)
 				kfree(num);
 
 				if (info->max_inline) {
-					info->max_inline = max_t(u64,
+					info->max_inline = min_t(u64,
 						info->max_inline,
 						root->sectorsize);
 				}
-				printk(KERN_INFO "btrfs: max_inline at %llu\n",
-					(unsigned long long)info->max_inline);
+				btrfs_info(root->fs_info, "max_inline at %llu",
+					info->max_inline);
+			} else {
+				ret = -ENOMEM;
+				goto out;
 			}
 			break;
 		case Opt_alloc_start:
 			num = match_strdup(&args[0]);
 			if (num) {
+				mutex_lock(&info->chunk_mutex);
 				info->alloc_start = memparse(num, NULL);
+				mutex_unlock(&info->chunk_mutex);
 				kfree(num);
-				printk(KERN_INFO
-					"btrfs: allocations start at %llu\n",
-					(unsigned long long)info->alloc_start);
+				btrfs_info(root->fs_info, "allocations start at %llu",
+					info->alloc_start);
+			} else {
+				ret = -ENOMEM;
+				goto out;
 			}
 			break;
+		case Opt_acl:
+#ifdef CONFIG_BTRFS_FS_POSIX_ACL
+			root->fs_info->sb->s_flags |= MS_POSIXACL;
+			break;
+#else
+			btrfs_err(root->fs_info,
+				"support for ACL not compiled in!");
+			ret = -EINVAL;
+			goto out;
+#endif
 		case Opt_noacl:
 			root->fs_info->sb->s_flags &= ~MS_POSIXACL;
 			break;
 		case Opt_notreelog:
-			printk(KERN_INFO "btrfs: disabling tree log\n");
-			btrfs_set_opt(info->mount_opt, NOTREELOG);
+			btrfs_set_and_info(root, NOTREELOG,
+					   "disabling tree log");
+			break;
+		case Opt_treelog:
+			btrfs_clear_and_info(root, NOTREELOG,
+					     "enabling tree log");
 			break;
 		case Opt_flushoncommit:
-			printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
-			btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
+			btrfs_set_and_info(root, FLUSHONCOMMIT,
+					   "turning on flush-on-commit");
+			break;
+		case Opt_noflushoncommit:
+			btrfs_clear_and_info(root, FLUSHONCOMMIT,
+					     "turning off flush-on-commit");
 			break;
 		case Opt_ratio:
-			intarg = 0;
-			match_int(&args[0], &intarg);
-			if (intarg) {
+			ret = match_int(&args[0], &intarg);
+			if (ret) {
+				goto out;
+			} else if (intarg >= 0) {
 				info->metadata_ratio = intarg;
-				printk(KERN_INFO "btrfs: metadata ratio %d\n",
+				btrfs_info(root->fs_info, "metadata ratio %d",
 				       info->metadata_ratio);
+			} else {
+				ret = -EINVAL;
+				goto out;
 			}
 			break;
 		case Opt_discard:
-			btrfs_set_opt(info->mount_opt, DISCARD);
+			btrfs_set_and_info(root, DISCARD,
+					   "turning on discard");
+			break;
+		case Opt_nodiscard:
+			btrfs_clear_and_info(root, DISCARD,
+					     "turning off discard");
+			break;
+		case Opt_space_cache:
+			btrfs_set_and_info(root, SPACE_CACHE,
+					   "enabling disk space caching");
+			break;
+		case Opt_rescan_uuid_tree:
+			btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
+			break;
+		case Opt_no_space_cache:
+			btrfs_clear_and_info(root, SPACE_CACHE,
+					     "disabling disk space caching");
+			break;
+		case Opt_inode_cache:
+			btrfs_set_and_info(root, CHANGE_INODE_CACHE,
+					   "enabling inode map caching");
+			break;
+		case Opt_noinode_cache:
+			btrfs_clear_and_info(root, CHANGE_INODE_CACHE,
+					     "disabling inode map caching");
+			break;
+		case Opt_clear_cache:
+			btrfs_set_and_info(root, CLEAR_CACHE,
+					   "force clearing of disk cache");
+			break;
+		case Opt_user_subvol_rm_allowed:
+			btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
+			break;
+		case Opt_enospc_debug:
+			btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
+			break;
+		case Opt_noenospc_debug:
+			btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
+			break;
+		case Opt_defrag:
+			btrfs_set_and_info(root, AUTO_DEFRAG,
+					   "enabling auto defrag");
+			break;
+		case Opt_nodefrag:
+			btrfs_clear_and_info(root, AUTO_DEFRAG,
+					     "disabling auto defrag");
+			break;
+		case Opt_recovery:
+			btrfs_info(root->fs_info, "enabling auto recovery");
+			btrfs_set_opt(info->mount_opt, RECOVERY);
+			break;
+		case Opt_skip_balance:
+			btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
+			break;
+#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
+		case Opt_check_integrity_including_extent_data:
+			btrfs_info(root->fs_info,
+				   "enabling check integrity including extent data");
+			btrfs_set_opt(info->mount_opt,
+				      CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
+			btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
+			break;
+		case Opt_check_integrity:
+			btrfs_info(root->fs_info, "enabling check integrity");
+			btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
+			break;
+		case Opt_check_integrity_print_mask:
+			ret = match_int(&args[0], &intarg);
+			if (ret) {
+				goto out;
+			} else if (intarg >= 0) {
+				info->check_integrity_print_mask = intarg;
+				btrfs_info(root->fs_info, "check_integrity_print_mask 0x%x",
+				       info->check_integrity_print_mask);
+			} else {
+				ret = -EINVAL;
+				goto out;
+			}
+			break;
+#else
+		case Opt_check_integrity_including_extent_data:
+		case Opt_check_integrity:
+		case Opt_check_integrity_print_mask:
+			btrfs_err(root->fs_info,
+				"support for check_integrity* not compiled in!");
+			ret = -EINVAL;
+			goto out;
+#endif
+		case Opt_fatal_errors:
+			if (strcmp(args[0].from, "panic") == 0)
+				btrfs_set_opt(info->mount_opt,
+					      PANIC_ON_FATAL_ERROR);
+			else if (strcmp(args[0].from, "bug") == 0)
+				btrfs_clear_opt(info->mount_opt,
+					      PANIC_ON_FATAL_ERROR);
+			else {
+				ret = -EINVAL;
+				goto out;
+			}
+			break;
+		case Opt_commit_interval:
+			intarg = 0;
+			ret = match_int(&args[0], &intarg);
+			if (ret < 0) {
+				btrfs_err(root->fs_info, "invalid commit interval");
+				ret = -EINVAL;
+				goto out;
+			}
+			if (intarg > 0) {
+				if (intarg > 300) {
+					btrfs_warn(root->fs_info, "excessive commit interval %d",
+							intarg);
+				}
+				info->commit_interval = intarg;
+			} else {
+				btrfs_info(root->fs_info, "using default commit interval %ds",
+				    BTRFS_DEFAULT_COMMIT_INTERVAL);
+				info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
+			}
 			break;
 		case Opt_err:
-			printk(KERN_INFO "btrfs: unrecognized mount option "
-			       "'%s'\n", p);
+			btrfs_info(root->fs_info, "unrecognized mount option '%s'", p);
 			ret = -EINVAL;
 			goto out;
 		default:
@@ -245,6 +752,8 @@ int btrfs_parse_options(struct btrfs_root *root, char *options)
 		}
 	}
 out:
+	if (!ret && btrfs_test_opt(root, SPACE_CACHE))
+		btrfs_info(root->fs_info, "disk space caching is enabled");
 	kfree(orig);
 	return ret;
 }
@@ -260,12 +769,12 @@ static int btrfs_parse_early_options(const char *options, fmode_t flags,
 		struct btrfs_fs_devices **fs_devices)
 {
 	substring_t args[MAX_OPT_ARGS];
-	char *opts, *p;
+	char *device_name, *opts, *orig, *p;
+	char *num = NULL;
 	int error = 0;
-	int intarg;
 
 	if (!options)
-		goto out;
+		return 0;
 
 	/*
 	 * strsep changes the string, duplicate it because parse_options
@@ -274,6 +783,7 @@ static int btrfs_parse_early_options(const char *options, fmode_t flags,
 	opts = kstrdup(options, GFP_KERNEL);
 	if (!opts)
 		return -ENOMEM;
+	orig = opts;
 
 	while ((p = strsep(&opts, ",")) != NULL) {
 		int token;
@@ -283,51 +793,59 @@ static int btrfs_parse_early_options(const char *options, fmode_t flags,
 		token = match_token(p, tokens, args);
 		switch (token) {
 		case Opt_subvol:
+			kfree(*subvol_name);
 			*subvol_name = match_strdup(&args[0]);
+			if (!*subvol_name) {
+				error = -ENOMEM;
+				goto out;
+			}
 			break;
 		case Opt_subvolid:
-			intarg = 0;
-			error = match_int(&args[0], &intarg);
-			if (!error) {
+			num = match_strdup(&args[0]);
+			if (num) {
+				*subvol_objectid = memparse(num, NULL);
+				kfree(num);
 				/* we want the original fs_tree */
-				if (!intarg)
+				if (!*subvol_objectid)
 					*subvol_objectid =
 						BTRFS_FS_TREE_OBJECTID;
-				else
-					*subvol_objectid = intarg;
+			} else {
+				error = -EINVAL;
+				goto out;
 			}
 			break;
+		case Opt_subvolrootid:
+			printk(KERN_WARNING
+				"BTRFS: 'subvolrootid' mount option is deprecated and has "
+				"no effect\n");
+			break;
 		case Opt_device:
-			error = btrfs_scan_one_device(match_strdup(&args[0]),
+			device_name = match_strdup(&args[0]);
+			if (!device_name) {
+				error = -ENOMEM;
+				goto out;
+			}
+			error = btrfs_scan_one_device(device_name,
 					flags, holder, fs_devices);
+			kfree(device_name);
 			if (error)
-				goto out_free_opts;
+				goto out;
 			break;
 		default:
 			break;
 		}
 	}
 
- out_free_opts:
-	kfree(opts);
- out:
-	/*
-	 * If no subvolume name is specified we use the default one.  Allocate
-	 * a copy of the string "." here so that code later in the
-	 * mount path doesn't care if it's the default volume or another one.
-	 */
-	if (!*subvol_name) {
-		*subvol_name = kstrdup(".", GFP_KERNEL);
-		if (!*subvol_name)
-			return -ENOMEM;
-	}
+out:
+	kfree(orig);
 	return error;
 }
 
 static struct dentry *get_default_root(struct super_block *sb,
 				       u64 subvol_objectid)
 {
-	struct btrfs_root *root = sb->s_fs_info;
+	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
+	struct btrfs_root *root = fs_info->tree_root;
 	struct btrfs_root *new_root;
 	struct btrfs_dir_item *di;
 	struct btrfs_path *path;
@@ -358,8 +876,12 @@ static struct dentry *get_default_root(struct super_block *sb,
 	 * will mount by default if we haven't been given a specific subvolume
 	 * to mount.
 	 */
-	dir_id = btrfs_super_root_dir(&root->fs_info->super_copy);
+	dir_id = btrfs_super_root_dir(fs_info->super_copy);
 	di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
+	if (IS_ERR(di)) {
+		btrfs_free_path(path);
+		return ERR_CAST(di);
+	}
 	if (!di) {
 		/*
 		 * Ok the default dir item isn't there.  This is weird since
@@ -368,7 +890,7 @@ static struct dentry *get_default_root(struct super_block *sb,
 		 */
 		btrfs_free_path(path);
 		dir_id = BTRFS_FIRST_FREE_OBJECTID;
-		new_root = root->fs_info->fs_root;
+		new_root = fs_info->fs_root;
 		goto setup_root;
 	}
 
@@ -376,12 +898,9 @@ static struct dentry *get_default_root(struct super_block *sb,
 	btrfs_free_path(path);
 
 find_root:
-	new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
+	new_root = btrfs_read_fs_root_no_name(fs_info, &location);
 	if (IS_ERR(new_root))
-		return ERR_PTR(PTR_ERR(new_root));
-
-	if (btrfs_root_refs(&new_root->root_item) == 0)
-		return ERR_PTR(-ENOENT);
+		return ERR_CAST(new_root);
 
 	dir_id = btrfs_root_dirid(&new_root->root_item);
 setup_root:
@@ -390,8 +909,8 @@ setup_root:
 	location.offset = 0;
 
 	inode = btrfs_iget(sb, &location, new_root, &new);
-	if (!inode)
-		return ERR_PTR(-ENOMEM);
+	if (IS_ERR(inode))
+		return ERR_CAST(inode);
 
 	/*
 	 * If we're just mounting the root most subvol put the inode and return
@@ -403,28 +922,12 @@ setup_root:
 		return dget(sb->s_root);
 	}
 
-	if (new) {
-		const struct qstr name = { .name = "/", .len = 1 };
-
-		/*
-		 * New inode, we need to make the dentry a sibling of s_root so
-		 * everything gets cleaned up properly on unmount.
-		 */
-		dentry = d_alloc(sb->s_root, &name);
-		if (!dentry) {
-			iput(inode);
-			return ERR_PTR(-ENOMEM);
-		}
-		d_splice_alias(inode, dentry);
-	} else {
-		/*
-		 * We found the inode in cache, just find a dentry for it and
-		 * put the reference to the inode we just got.
-		 */
-		dentry = d_find_alias(inode);
-		iput(inode);
+	dentry = d_obtain_alias(inode);
+	if (!IS_ERR(dentry)) {
+		spin_lock(&dentry->d_lock);
+		dentry->d_flags &= ~DCACHE_DISCONNECTED;
+		spin_unlock(&dentry->d_lock);
 	}
-
 	return dentry;
 }
 
@@ -433,80 +936,82 @@ static int btrfs_fill_super(struct super_block *sb,
 			    void *data, int silent)
 {
 	struct inode *inode;
-	struct dentry *root_dentry;
-	struct btrfs_super_block *disk_super;
-	struct btrfs_root *tree_root;
+	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
 	struct btrfs_key key;
 	int err;
 
 	sb->s_maxbytes = MAX_LFS_FILESIZE;
 	sb->s_magic = BTRFS_SUPER_MAGIC;
 	sb->s_op = &btrfs_super_ops;
+	sb->s_d_op = &btrfs_dentry_operations;
 	sb->s_export_op = &btrfs_export_ops;
 	sb->s_xattr = btrfs_xattr_handlers;
 	sb->s_time_gran = 1;
 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
 	sb->s_flags |= MS_POSIXACL;
 #endif
-
-	tree_root = open_ctree(sb, fs_devices, (char *)data);
-
-	if (IS_ERR(tree_root)) {
-		printk("btrfs: open_ctree failed\n");
-		return PTR_ERR(tree_root);
+	sb->s_flags |= MS_I_VERSION;
+	err = open_ctree(sb, fs_devices, (char *)data);
+	if (err) {
+		printk(KERN_ERR "BTRFS: open_ctree failed\n");
+		return err;
 	}
-	sb->s_fs_info = tree_root;
-	disk_super = &tree_root->fs_info->super_copy;
 
 	key.objectid = BTRFS_FIRST_FREE_OBJECTID;
 	key.type = BTRFS_INODE_ITEM_KEY;
 	key.offset = 0;
-	inode = btrfs_iget(sb, &key, tree_root->fs_info->fs_root, NULL);
+	inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
 	if (IS_ERR(inode)) {
 		err = PTR_ERR(inode);
 		goto fail_close;
 	}
 
-	root_dentry = d_alloc_root(inode);
-	if (!root_dentry) {
-		iput(inode);
+	sb->s_root = d_make_root(inode);
+	if (!sb->s_root) {
 		err = -ENOMEM;
 		goto fail_close;
 	}
 
-	sb->s_root = root_dentry;
-
 	save_mount_options(sb, data);
+	cleancache_init_fs(sb);
+	sb->s_flags |= MS_ACTIVE;
 	return 0;
 
 fail_close:
-	close_ctree(tree_root);
+	close_ctree(fs_info->tree_root);
 	return err;
 }
 
 int btrfs_sync_fs(struct super_block *sb, int wait)
 {
 	struct btrfs_trans_handle *trans;
-	struct btrfs_root *root = btrfs_sb(sb);
-	int ret;
+	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
+	struct btrfs_root *root = fs_info->tree_root;
+
+	trace_btrfs_sync_fs(wait);
 
 	if (!wait) {
-		filemap_flush(root->fs_info->btree_inode->i_mapping);
+		filemap_flush(fs_info->btree_inode->i_mapping);
 		return 0;
 	}
 
-	btrfs_start_delalloc_inodes(root, 0);
-	btrfs_wait_ordered_extents(root, 0, 0);
+	btrfs_wait_ordered_roots(fs_info, -1);
 
-	trans = btrfs_start_transaction(root, 1);
-	ret = btrfs_commit_transaction(trans, root);
-	return ret;
+	trans = btrfs_attach_transaction_barrier(root);
+	if (IS_ERR(trans)) {
+		/* no transaction, don't bother */
+		if (PTR_ERR(trans) == -ENOENT)
+			return 0;
+		return PTR_ERR(trans);
+	}
+	return btrfs_commit_transaction(trans, root);
 }
 
-static int btrfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
+static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
 {
-	struct btrfs_root *root = btrfs_sb(vfs->mnt_sb);
-	struct btrfs_fs_info *info = root->fs_info;
+	struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
+	struct btrfs_root *root = info->tree_root;
+	char *compress_type;
 
 	if (btrfs_test_opt(root, DEGRADED))
 		seq_puts(seq, ",degraded");
@@ -517,16 +1022,22 @@ static int btrfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
 	if (btrfs_test_opt(root, NOBARRIER))
 		seq_puts(seq, ",nobarrier");
 	if (info->max_inline != 8192 * 1024)
-		seq_printf(seq, ",max_inline=%llu",
-			   (unsigned long long)info->max_inline);
+		seq_printf(seq, ",max_inline=%llu", info->max_inline);
 	if (info->alloc_start != 0)
-		seq_printf(seq, ",alloc_start=%llu",
-			   (unsigned long long)info->alloc_start);
+		seq_printf(seq, ",alloc_start=%llu", info->alloc_start);
 	if (info->thread_pool_size !=  min_t(unsigned long,
 					     num_online_cpus() + 2, 8))
 		seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
-	if (btrfs_test_opt(root, COMPRESS))
-		seq_puts(seq, ",compress");
+	if (btrfs_test_opt(root, COMPRESS)) {
+		if (info->compress_type == BTRFS_COMPRESS_ZLIB)
+			compress_type = "zlib";
+		else
+			compress_type = "lzo";
+		if (btrfs_test_opt(root, FORCE_COMPRESS))
+			seq_printf(seq, ",compress-force=%s", compress_type);
+		else
+			seq_printf(seq, ",compress=%s", compress_type);
+	}
 	if (btrfs_test_opt(root, NOSSD))
 		seq_puts(seq, ",nossd");
 	if (btrfs_test_opt(root, SSD_SPREAD))
@@ -541,15 +1052,174 @@ static int btrfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
 		seq_puts(seq, ",discard");
 	if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
 		seq_puts(seq, ",noacl");
+	if (btrfs_test_opt(root, SPACE_CACHE))
+		seq_puts(seq, ",space_cache");
+	else
+		seq_puts(seq, ",nospace_cache");
+	if (btrfs_test_opt(root, RESCAN_UUID_TREE))
+		seq_puts(seq, ",rescan_uuid_tree");
+	if (btrfs_test_opt(root, CLEAR_CACHE))
+		seq_puts(seq, ",clear_cache");
+	if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
+		seq_puts(seq, ",user_subvol_rm_allowed");
+	if (btrfs_test_opt(root, ENOSPC_DEBUG))
+		seq_puts(seq, ",enospc_debug");
+	if (btrfs_test_opt(root, AUTO_DEFRAG))
+		seq_puts(seq, ",autodefrag");
+	if (btrfs_test_opt(root, INODE_MAP_CACHE))
+		seq_puts(seq, ",inode_cache");
+	if (btrfs_test_opt(root, SKIP_BALANCE))
+		seq_puts(seq, ",skip_balance");
+	if (btrfs_test_opt(root, RECOVERY))
+		seq_puts(seq, ",recovery");
+#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
+	if (btrfs_test_opt(root, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
+		seq_puts(seq, ",check_int_data");
+	else if (btrfs_test_opt(root, CHECK_INTEGRITY))
+		seq_puts(seq, ",check_int");
+	if (info->check_integrity_print_mask)
+		seq_printf(seq, ",check_int_print_mask=%d",
+				info->check_integrity_print_mask);
+#endif
+	if (info->metadata_ratio)
+		seq_printf(seq, ",metadata_ratio=%d",
+				info->metadata_ratio);
+	if (btrfs_test_opt(root, PANIC_ON_FATAL_ERROR))
+		seq_puts(seq, ",fatal_errors=panic");
+	if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
+		seq_printf(seq, ",commit=%d", info->commit_interval);
 	return 0;
 }
 
 static int btrfs_test_super(struct super_block *s, void *data)
 {
-	struct btrfs_fs_devices *test_fs_devices = data;
-	struct btrfs_root *root = btrfs_sb(s);
+	struct btrfs_fs_info *p = data;
+	struct btrfs_fs_info *fs_info = btrfs_sb(s);
+
+	return fs_info->fs_devices == p->fs_devices;
+}
+
+static int btrfs_set_super(struct super_block *s, void *data)
+{
+	int err = set_anon_super(s, data);
+	if (!err)
+		s->s_fs_info = data;
+	return err;
+}
 
-	return root->fs_info->fs_devices == test_fs_devices;
+/*
+ * subvolumes are identified by ino 256
+ */
+static inline int is_subvolume_inode(struct inode *inode)
+{
+	if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
+		return 1;
+	return 0;
+}
+
+/*
+ * This will strip out the subvol=%s argument for an argument string and add
+ * subvolid=0 to make sure we get the actual tree root for path walking to the
+ * subvol we want.
+ */
+static char *setup_root_args(char *args)
+{
+	unsigned len = strlen(args) + 2 + 1;
+	char *src, *dst, *buf;
+
+	/*
+	 * We need the same args as before, but with this substitution:
+	 * s!subvol=[^,]+!subvolid=0!
+	 *
+	 * Since the replacement string is up to 2 bytes longer than the
+	 * original, allocate strlen(args) + 2 + 1 bytes.
+	 */
+
+	src = strstr(args, "subvol=");
+	/* This shouldn't happen, but just in case.. */
+	if (!src)
+		return NULL;
+
+	buf = dst = kmalloc(len, GFP_NOFS);
+	if (!buf)
+		return NULL;
+
+	/*
+	 * If the subvol= arg is not at the start of the string,
+	 * copy whatever precedes it into buf.
+	 */
+	if (src != args) {
+		*src++ = '\0';
+		strcpy(buf, args);
+		dst += strlen(args);
+	}
+
+	strcpy(dst, "subvolid=0");
+	dst += strlen("subvolid=0");
+
+	/*
+	 * If there is a "," after the original subvol=... string,
+	 * copy that suffix into our buffer.  Otherwise, we're done.
+	 */
+	src = strchr(src, ',');
+	if (src)
+		strcpy(dst, src);
+
+	return buf;
+}
+
+static struct dentry *mount_subvol(const char *subvol_name, int flags,
+				   const char *device_name, char *data)
+{
+	struct dentry *root;
+	struct vfsmount *mnt;
+	char *newargs;
+
+	newargs = setup_root_args(data);
+	if (!newargs)
+		return ERR_PTR(-ENOMEM);
+	mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name,
+			     newargs);
+
+	if (PTR_RET(mnt) == -EBUSY) {
+		if (flags & MS_RDONLY) {
+			mnt = vfs_kern_mount(&btrfs_fs_type, flags & ~MS_RDONLY, device_name,
+					     newargs);
+		} else {
+			int r;
+			mnt = vfs_kern_mount(&btrfs_fs_type, flags | MS_RDONLY, device_name,
+					     newargs);
+			if (IS_ERR(mnt)) {
+				kfree(newargs);
+				return ERR_CAST(mnt);
+			}
+
+			r = btrfs_remount(mnt->mnt_sb, &flags, NULL);
+			if (r < 0) {
+				/* FIXME: release vfsmount mnt ??*/
+				kfree(newargs);
+				return ERR_PTR(r);
+			}
+		}
+	}
+
+	kfree(newargs);
+
+	if (IS_ERR(mnt))
+		return ERR_CAST(mnt);
+
+	root = mount_subtree(mnt, subvol_name);
+
+	if (!IS_ERR(root) && !is_subvolume_inode(root->d_inode)) {
+		struct super_block *s = root->d_sb;
+		dput(root);
+		root = ERR_PTR(-EINVAL);
+		deactivate_locked_super(s);
+		printk(KERN_ERR "BTRFS: '%s' is not a valid subvolume\n",
+				subvol_name);
+	}
+
+	return root;
 }
 
 /*
@@ -558,18 +1228,18 @@ static int btrfs_test_super(struct super_block *s, void *data)
  * Note:  This is based on get_sb_bdev from fs/super.c with a few additions
  *	  for multiple device setup.  Make sure to keep it in sync.
  */
-static int btrfs_get_sb(struct file_system_type *fs_type, int flags,
-		const char *dev_name, void *data, struct vfsmount *mnt)
+static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
+		const char *device_name, void *data)
 {
 	struct block_device *bdev = NULL;
 	struct super_block *s;
 	struct dentry *root;
 	struct btrfs_fs_devices *fs_devices = NULL;
+	struct btrfs_fs_info *fs_info = NULL;
 	fmode_t mode = FMODE_READ;
 	char *subvol_name = NULL;
 	u64 subvol_objectid = 0;
 	int error = 0;
-	int found = 0;
 
 	if (!(flags & MS_RDONLY))
 		mode |= FMODE_WRITE;
@@ -577,16 +1247,43 @@ static int btrfs_get_sb(struct file_system_type *fs_type, int flags,
 	error = btrfs_parse_early_options(data, mode, fs_type,
 					  &subvol_name, &subvol_objectid,
 					  &fs_devices);
-	if (error)
-		return error;
+	if (error) {
+		kfree(subvol_name);
+		return ERR_PTR(error);
+	}
+
+	if (subvol_name) {
+		root = mount_subvol(subvol_name, flags, device_name, data);
+		kfree(subvol_name);
+		return root;
+	}
 
-	error = btrfs_scan_one_device(dev_name, mode, fs_type, &fs_devices);
+	error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
 	if (error)
-		goto error_free_subvol_name;
+		return ERR_PTR(error);
+
+	/*
+	 * Setup a dummy root and fs_info for test/set super.  This is because
+	 * we don't actually fill this stuff out until open_ctree, but we need
+	 * it for searching for existing supers, so this lets us do that and
+	 * then open_ctree will properly initialize everything later.
+	 */
+	fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
+	if (!fs_info)
+		return ERR_PTR(-ENOMEM);
+
+	fs_info->fs_devices = fs_devices;
+
+	fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
+	fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
+	if (!fs_info->super_copy || !fs_info->super_for_commit) {
+		error = -ENOMEM;
+		goto error_fs_info;
+	}
 
 	error = btrfs_open_devices(fs_devices, mode, fs_type);
 	if (error)
-		goto error_free_subvol_name;
+		goto error_fs_info;
 
 	if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
 		error = -EACCES;
@@ -594,156 +1291,427 @@ static int btrfs_get_sb(struct file_system_type *fs_type, int flags,
 	}
 
 	bdev = fs_devices->latest_bdev;
-	s = sget(fs_type, btrfs_test_super, set_anon_super, fs_devices);
-	if (IS_ERR(s))
-		goto error_s;
+	s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | MS_NOSEC,
+		 fs_info);
+	if (IS_ERR(s)) {
+		error = PTR_ERR(s);
+		goto error_close_devices;
+	}
 
 	if (s->s_root) {
-		if ((flags ^ s->s_flags) & MS_RDONLY) {
-			deactivate_locked_super(s);
-			error = -EBUSY;
-			goto error_close_devices;
-		}
-
-		found = 1;
 		btrfs_close_devices(fs_devices);
+		free_fs_info(fs_info);
+		if ((flags ^ s->s_flags) & MS_RDONLY)
+			error = -EBUSY;
 	} else {
 		char b[BDEVNAME_SIZE];
 
-		s->s_flags = flags;
 		strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
+		btrfs_sb(s)->bdev_holder = fs_type;
 		error = btrfs_fill_super(s, fs_devices, data,
 					 flags & MS_SILENT ? 1 : 0);
-		if (error) {
-			deactivate_locked_super(s);
-			goto error_free_subvol_name;
-		}
-
-		btrfs_sb(s)->fs_info->bdev_holder = fs_type;
-		s->s_flags |= MS_ACTIVE;
 	}
 
-	root = get_default_root(s, subvol_objectid);
-	if (IS_ERR(root)) {
-		error = PTR_ERR(root);
+	root = !error ? get_default_root(s, subvol_objectid) : ERR_PTR(error);
+	if (IS_ERR(root))
 		deactivate_locked_super(s);
-		goto error;
-	}
-	/* if they gave us a subvolume name bind mount into that */
-	if (strcmp(subvol_name, ".")) {
-		struct dentry *new_root;
-		mutex_lock(&root->d_inode->i_mutex);
-		new_root = lookup_one_len(subvol_name, root,
-				      strlen(subvol_name));
-		mutex_unlock(&root->d_inode->i_mutex);
-
-		if (IS_ERR(new_root)) {
-			deactivate_locked_super(s);
-			error = PTR_ERR(new_root);
-			dput(root);
-			goto error_close_devices;
-		}
-		if (!new_root->d_inode) {
-			dput(root);
-			dput(new_root);
-			deactivate_locked_super(s);
-			error = -ENXIO;
-			goto error_close_devices;
-		}
-		dput(root);
-		root = new_root;
-	}
 
-	mnt->mnt_sb = s;
-	mnt->mnt_root = root;
+	return root;
 
-	kfree(subvol_name);
-	return 0;
-
-error_s:
-	error = PTR_ERR(s);
 error_close_devices:
 	btrfs_close_devices(fs_devices);
-error_free_subvol_name:
-	kfree(subvol_name);
-error:
-	return error;
+error_fs_info:
+	free_fs_info(fs_info);
+	return ERR_PTR(error);
+}
+
+static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
+				     int new_pool_size, int old_pool_size)
+{
+	if (new_pool_size == old_pool_size)
+		return;
+
+	fs_info->thread_pool_size = new_pool_size;
+
+	btrfs_info(fs_info, "resize thread pool %d -> %d",
+	       old_pool_size, new_pool_size);
+
+	btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
+	btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
+	btrfs_workqueue_set_max(fs_info->submit_workers, new_pool_size);
+	btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
+	btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
+	btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
+	btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
+				new_pool_size);
+	btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
+	btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
+	btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
+	btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
+	btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
+				new_pool_size);
+}
+
+static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
+{
+	set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
+}
+
+static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
+				       unsigned long old_opts, int flags)
+{
+	if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
+	    (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
+	     (flags & MS_RDONLY))) {
+		/* wait for any defraggers to finish */
+		wait_event(fs_info->transaction_wait,
+			   (atomic_read(&fs_info->defrag_running) == 0));
+		if (flags & MS_RDONLY)
+			sync_filesystem(fs_info->sb);
+	}
+}
+
+static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
+					 unsigned long old_opts)
+{
+	/*
+	 * We need cleanup all defragable inodes if the autodefragment is
+	 * close or the fs is R/O.
+	 */
+	if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
+	    (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
+	     (fs_info->sb->s_flags & MS_RDONLY))) {
+		btrfs_cleanup_defrag_inodes(fs_info);
+	}
+
+	clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
 }
 
 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
 {
-	struct btrfs_root *root = btrfs_sb(sb);
+	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
+	struct btrfs_root *root = fs_info->tree_root;
+	unsigned old_flags = sb->s_flags;
+	unsigned long old_opts = fs_info->mount_opt;
+	unsigned long old_compress_type = fs_info->compress_type;
+	u64 old_max_inline = fs_info->max_inline;
+	u64 old_alloc_start = fs_info->alloc_start;
+	int old_thread_pool_size = fs_info->thread_pool_size;
+	unsigned int old_metadata_ratio = fs_info->metadata_ratio;
 	int ret;
 
+	sync_filesystem(sb);
+	btrfs_remount_prepare(fs_info);
+
 	ret = btrfs_parse_options(root, data);
-	if (ret)
-		return -EINVAL;
+	if (ret) {
+		ret = -EINVAL;
+		goto restore;
+	}
+
+	btrfs_remount_begin(fs_info, old_opts, *flags);
+	btrfs_resize_thread_pool(fs_info,
+		fs_info->thread_pool_size, old_thread_pool_size);
 
 	if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
-		return 0;
+		goto out;
 
 	if (*flags & MS_RDONLY) {
+		/*
+		 * this also happens on 'umount -rf' or on shutdown, when
+		 * the filesystem is busy.
+		 */
+		cancel_work_sync(&fs_info->async_reclaim_work);
+
+		/* wait for the uuid_scan task to finish */
+		down(&fs_info->uuid_tree_rescan_sem);
+		/* avoid complains from lockdep et al. */
+		up(&fs_info->uuid_tree_rescan_sem);
+
 		sb->s_flags |= MS_RDONLY;
 
-		ret =  btrfs_commit_super(root);
-		WARN_ON(ret);
+		btrfs_dev_replace_suspend_for_unmount(fs_info);
+		btrfs_scrub_cancel(fs_info);
+		btrfs_pause_balance(fs_info);
+
+		ret = btrfs_commit_super(root);
+		if (ret)
+			goto restore;
 	} else {
-		if (root->fs_info->fs_devices->rw_devices == 0)
-			return -EACCES;
+		if (test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state)) {
+			btrfs_err(fs_info,
+				"Remounting read-write after error is not allowed");
+			ret = -EINVAL;
+			goto restore;
+		}
+		if (fs_info->fs_devices->rw_devices == 0) {
+			ret = -EACCES;
+			goto restore;
+		}
 
-		if (btrfs_super_log_root(&root->fs_info->super_copy) != 0)
-			return -EINVAL;
+		if (fs_info->fs_devices->missing_devices >
+		     fs_info->num_tolerated_disk_barrier_failures &&
+		    !(*flags & MS_RDONLY)) {
+			btrfs_warn(fs_info,
+				"too many missing devices, writeable remount is not allowed");
+			ret = -EACCES;
+			goto restore;
+		}
+
+		if (btrfs_super_log_root(fs_info->super_copy) != 0) {
+			ret = -EINVAL;
+			goto restore;
+		}
+
+		ret = btrfs_cleanup_fs_roots(fs_info);
+		if (ret)
+			goto restore;
 
 		/* recover relocation */
+		mutex_lock(&fs_info->cleaner_mutex);
 		ret = btrfs_recover_relocation(root);
-		WARN_ON(ret);
-
-		ret = btrfs_cleanup_fs_roots(root->fs_info);
-		WARN_ON(ret);
+		mutex_unlock(&fs_info->cleaner_mutex);
+		if (ret)
+			goto restore;
+
+		ret = btrfs_resume_balance_async(fs_info);
+		if (ret)
+			goto restore;
+
+		ret = btrfs_resume_dev_replace_async(fs_info);
+		if (ret) {
+			btrfs_warn(fs_info, "failed to resume dev_replace");
+			goto restore;
+		}
 
+		if (!fs_info->uuid_root) {
+			btrfs_info(fs_info, "creating UUID tree");
+			ret = btrfs_create_uuid_tree(fs_info);
+			if (ret) {
+				btrfs_warn(fs_info, "failed to create the UUID tree %d", ret);
+				goto restore;
+			}
+		}
 		sb->s_flags &= ~MS_RDONLY;
 	}
+out:
+	wake_up_process(fs_info->transaction_kthread);
+	btrfs_remount_cleanup(fs_info, old_opts);
+	return 0;
+
+restore:
+	/* We've hit an error - don't reset MS_RDONLY */
+	if (sb->s_flags & MS_RDONLY)
+		old_flags |= MS_RDONLY;
+	sb->s_flags = old_flags;
+	fs_info->mount_opt = old_opts;
+	fs_info->compress_type = old_compress_type;
+	fs_info->max_inline = old_max_inline;
+	mutex_lock(&fs_info->chunk_mutex);
+	fs_info->alloc_start = old_alloc_start;
+	mutex_unlock(&fs_info->chunk_mutex);
+	btrfs_resize_thread_pool(fs_info,
+		old_thread_pool_size, fs_info->thread_pool_size);
+	fs_info->metadata_ratio = old_metadata_ratio;
+	btrfs_remount_cleanup(fs_info, old_opts);
+	return ret;
+}
+
+/* Used to sort the devices by max_avail(descending sort) */
+static int btrfs_cmp_device_free_bytes(const void *dev_info1,
+				       const void *dev_info2)
+{
+	if (((struct btrfs_device_info *)dev_info1)->max_avail >
+	    ((struct btrfs_device_info *)dev_info2)->max_avail)
+		return -1;
+	else if (((struct btrfs_device_info *)dev_info1)->max_avail <
+		 ((struct btrfs_device_info *)dev_info2)->max_avail)
+		return 1;
+	else
+	return 0;
+}
+
+/*
+ * sort the devices by max_avail, in which max free extent size of each device
+ * is stored.(Descending Sort)
+ */
+static inline void btrfs_descending_sort_devices(
+					struct btrfs_device_info *devices,
+					size_t nr_devices)
+{
+	sort(devices, nr_devices, sizeof(struct btrfs_device_info),
+	     btrfs_cmp_device_free_bytes, NULL);
+}
+
+/*
+ * The helper to calc the free space on the devices that can be used to store
+ * file data.
+ */
+static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
+{
+	struct btrfs_fs_info *fs_info = root->fs_info;
+	struct btrfs_device_info *devices_info;
+	struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
+	struct btrfs_device *device;
+	u64 skip_space;
+	u64 type;
+	u64 avail_space;
+	u64 used_space;
+	u64 min_stripe_size;
+	int min_stripes = 1, num_stripes = 1;
+	int i = 0, nr_devices;
+	int ret;
+
+	nr_devices = fs_info->fs_devices->open_devices;
+	BUG_ON(!nr_devices);
+
+	devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
+			       GFP_NOFS);
+	if (!devices_info)
+		return -ENOMEM;
+
+	/* calc min stripe number for data space alloction */
+	type = btrfs_get_alloc_profile(root, 1);
+	if (type & BTRFS_BLOCK_GROUP_RAID0) {
+		min_stripes = 2;
+		num_stripes = nr_devices;
+	} else if (type & BTRFS_BLOCK_GROUP_RAID1) {
+		min_stripes = 2;
+		num_stripes = 2;
+	} else if (type & BTRFS_BLOCK_GROUP_RAID10) {
+		min_stripes = 4;
+		num_stripes = 4;
+	}
+
+	if (type & BTRFS_BLOCK_GROUP_DUP)
+		min_stripe_size = 2 * BTRFS_STRIPE_LEN;
+	else
+		min_stripe_size = BTRFS_STRIPE_LEN;
+
+	list_for_each_entry(device, &fs_devices->devices, dev_list) {
+		if (!device->in_fs_metadata || !device->bdev ||
+		    device->is_tgtdev_for_dev_replace)
+			continue;
+
+		avail_space = device->total_bytes - device->bytes_used;
+
+		/* align with stripe_len */
+		do_div(avail_space, BTRFS_STRIPE_LEN);
+		avail_space *= BTRFS_STRIPE_LEN;
+
+		/*
+		 * In order to avoid overwritting the superblock on the drive,
+		 * btrfs starts at an offset of at least 1MB when doing chunk
+		 * allocation.
+		 */
+		skip_space = 1024 * 1024;
+
+		/* user can set the offset in fs_info->alloc_start. */
+		if (fs_info->alloc_start + BTRFS_STRIPE_LEN <=
+		    device->total_bytes)
+			skip_space = max(fs_info->alloc_start, skip_space);
+
+		/*
+		 * btrfs can not use the free space in [0, skip_space - 1],
+		 * we must subtract it from the total. In order to implement
+		 * it, we account the used space in this range first.
+		 */
+		ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1,
+						     &used_space);
+		if (ret) {
+			kfree(devices_info);
+			return ret;
+		}
+
+		/* calc the free space in [0, skip_space - 1] */
+		skip_space -= used_space;
+
+		/*
+		 * we can use the free space in [0, skip_space - 1], subtract
+		 * it from the total.
+		 */
+		if (avail_space && avail_space >= skip_space)
+			avail_space -= skip_space;
+		else
+			avail_space = 0;
+
+		if (avail_space < min_stripe_size)
+			continue;
+
+		devices_info[i].dev = device;
+		devices_info[i].max_avail = avail_space;
+
+		i++;
+	}
+
+	nr_devices = i;
 
+	btrfs_descending_sort_devices(devices_info, nr_devices);
+
+	i = nr_devices - 1;
+	avail_space = 0;
+	while (nr_devices >= min_stripes) {
+		if (num_stripes > nr_devices)
+			num_stripes = nr_devices;
+
+		if (devices_info[i].max_avail >= min_stripe_size) {
+			int j;
+			u64 alloc_size;
+
+			avail_space += devices_info[i].max_avail * num_stripes;
+			alloc_size = devices_info[i].max_avail;
+			for (j = i + 1 - num_stripes; j <= i; j++)
+				devices_info[j].max_avail -= alloc_size;
+		}
+		i--;
+		nr_devices--;
+	}
+
+	kfree(devices_info);
+	*free_bytes = avail_space;
 	return 0;
 }
 
 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
 {
-	struct btrfs_root *root = btrfs_sb(dentry->d_sb);
-	struct btrfs_super_block *disk_super = &root->fs_info->super_copy;
-	struct list_head *head = &root->fs_info->space_info;
+	struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
+	struct btrfs_super_block *disk_super = fs_info->super_copy;
+	struct list_head *head = &fs_info->space_info;
 	struct btrfs_space_info *found;
 	u64 total_used = 0;
-	u64 data_used = 0;
+	u64 total_free_data = 0;
 	int bits = dentry->d_sb->s_blocksize_bits;
-	__be32 *fsid = (__be32 *)root->fs_info->fsid;
+	__be32 *fsid = (__be32 *)fs_info->fsid;
+	int ret;
 
+	/* holding chunk_muext to avoid allocating new chunks */
+	mutex_lock(&fs_info->chunk_mutex);
 	rcu_read_lock();
 	list_for_each_entry_rcu(found, head, list) {
-		if (found->flags & (BTRFS_BLOCK_GROUP_DUP|
-				    BTRFS_BLOCK_GROUP_RAID10|
-				    BTRFS_BLOCK_GROUP_RAID1)) {
-			total_used += found->bytes_used;
-			if (found->flags & BTRFS_BLOCK_GROUP_DATA)
-				data_used += found->bytes_used;
-			else
-				data_used += found->total_bytes;
+		if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
+			total_free_data += found->disk_total - found->disk_used;
+			total_free_data -=
+				btrfs_account_ro_block_groups_free_space(found);
 		}
 
-		total_used += found->bytes_used;
-		if (found->flags & BTRFS_BLOCK_GROUP_DATA)
-			data_used += found->bytes_used;
-		else
-			data_used += found->total_bytes;
+		total_used += found->disk_used;
 	}
 	rcu_read_unlock();
 
 	buf->f_namelen = BTRFS_NAME_LEN;
 	buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
 	buf->f_bfree = buf->f_blocks - (total_used >> bits);
-	buf->f_bavail = buf->f_blocks - (data_used >> bits);
 	buf->f_bsize = dentry->d_sb->s_blocksize;
 	buf->f_type = BTRFS_SUPER_MAGIC;
+	buf->f_bavail = total_free_data;
+	ret = btrfs_calc_avail_data_space(fs_info->tree_root, &total_free_data);
+	if (ret) {
+		mutex_unlock(&fs_info->chunk_mutex);
+		return ret;
+	}
+	buf->f_bavail += total_free_data;
+	buf->f_bavail = buf->f_bavail >> bits;
+	mutex_unlock(&fs_info->chunk_mutex);
 
 	/* We treat it as constant endianness (it doesn't matter _which_)
 	   because we want the fsid to come out the same whether mounted
@@ -757,13 +1725,21 @@ static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
 	return 0;
 }
 
+static void btrfs_kill_super(struct super_block *sb)
+{
+	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
+	kill_anon_super(sb);
+	free_fs_info(fs_info);
+}
+
 static struct file_system_type btrfs_fs_type = {
 	.owner		= THIS_MODULE,
 	.name		= "btrfs",
-	.get_sb		= btrfs_get_sb,
-	.kill_sb	= kill_anon_super,
+	.mount		= btrfs_mount,
+	.kill_sb	= btrfs_kill_super,
 	.fs_flags	= FS_REQUIRES_DEV,
 };
+MODULE_ALIAS_FS("btrfs");
 
 /*
  * used by btrfsctl to scan devices when no FS is mounted
@@ -787,6 +1763,13 @@ static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
 		ret = btrfs_scan_one_device(vol->name, FMODE_READ,
 					    &btrfs_fs_type, &fs_devices);
 		break;
+	case BTRFS_IOC_DEVICES_READY:
+		ret = btrfs_scan_one_device(vol->name, FMODE_READ,
+					    &btrfs_fs_type, &fs_devices);
+		if (ret)
+			break;
+		ret = !(fs_devices->num_devices == fs_devices->total_devices);
+		break;
 	}
 
 	kfree(vol);
@@ -795,28 +1778,67 @@ static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
 
 static int btrfs_freeze(struct super_block *sb)
 {
-	struct btrfs_root *root = btrfs_sb(sb);
-	mutex_lock(&root->fs_info->transaction_kthread_mutex);
-	mutex_lock(&root->fs_info->cleaner_mutex);
-	return 0;
+	struct btrfs_trans_handle *trans;
+	struct btrfs_root *root = btrfs_sb(sb)->tree_root;
+
+	trans = btrfs_attach_transaction_barrier(root);
+	if (IS_ERR(trans)) {
+		/* no transaction, don't bother */
+		if (PTR_ERR(trans) == -ENOENT)
+			return 0;
+		return PTR_ERR(trans);
+	}
+	return btrfs_commit_transaction(trans, root);
 }
 
 static int btrfs_unfreeze(struct super_block *sb)
 {
-	struct btrfs_root *root = btrfs_sb(sb);
-	mutex_unlock(&root->fs_info->cleaner_mutex);
-	mutex_unlock(&root->fs_info->transaction_kthread_mutex);
+	return 0;
+}
+
+static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
+{
+	struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
+	struct btrfs_fs_devices *cur_devices;
+	struct btrfs_device *dev, *first_dev = NULL;
+	struct list_head *head;
+	struct rcu_string *name;
+
+	mutex_lock(&fs_info->fs_devices->device_list_mutex);
+	cur_devices = fs_info->fs_devices;
+	while (cur_devices) {
+		head = &cur_devices->devices;
+		list_for_each_entry(dev, head, dev_list) {
+			if (dev->missing)
+				continue;
+			if (!dev->name)
+				continue;
+			if (!first_dev || dev->devid < first_dev->devid)
+				first_dev = dev;
+		}
+		cur_devices = cur_devices->seed;
+	}
+
+	if (first_dev) {
+		rcu_read_lock();
+		name = rcu_dereference(first_dev->name);
+		seq_escape(m, name->str, " \t\n\\");
+		rcu_read_unlock();
+	} else {
+		WARN_ON(1);
+	}
+	mutex_unlock(&fs_info->fs_devices->device_list_mutex);
 	return 0;
 }
 
 static const struct super_operations btrfs_super_ops = {
 	.drop_inode	= btrfs_drop_inode,
-	.delete_inode	= btrfs_delete_inode,
+	.evict_inode	= btrfs_evict_inode,
 	.put_super	= btrfs_put_super,
 	.sync_fs	= btrfs_sync_fs,
 	.show_options	= btrfs_show_options,
+	.show_devname	= btrfs_show_devname,
 	.write_inode	= btrfs_write_inode,
-	.dirty_inode	= btrfs_dirty_inode,
 	.alloc_inode	= btrfs_alloc_inode,
 	.destroy_inode	= btrfs_destroy_inode,
 	.statfs		= btrfs_statfs,
@@ -829,14 +1851,18 @@ static const struct file_operations btrfs_ctl_fops = {
 	.unlocked_ioctl	 = btrfs_control_ioctl,
 	.compat_ioctl = btrfs_control_ioctl,
 	.owner	 = THIS_MODULE,
+	.llseek = noop_llseek,
 };
 
 static struct miscdevice btrfs_misc = {
-	.minor		= MISC_DYNAMIC_MINOR,
+	.minor		= BTRFS_MINOR,
 	.name		= "btrfs-control",
 	.fops		= &btrfs_ctl_fops
 };
 
+MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
+MODULE_ALIAS("devname:btrfs-control");
+
 static int btrfs_interface_init(void)
 {
 	return misc_register(&btrfs_misc);
@@ -845,20 +1871,69 @@ static int btrfs_interface_init(void)
 static void btrfs_interface_exit(void)
 {
 	if (misc_deregister(&btrfs_misc) < 0)
-		printk(KERN_INFO "misc_deregister failed for control device");
+		printk(KERN_INFO "BTRFS: misc_deregister failed for control device\n");
+}
+
+static void btrfs_print_info(void)
+{
+	printk(KERN_INFO "Btrfs loaded"
+#ifdef CONFIG_BTRFS_DEBUG
+			", debug=on"
+#endif
+#ifdef CONFIG_BTRFS_ASSERT
+			", assert=on"
+#endif
+#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
+			", integrity-checker=on"
+#endif
+			"\n");
+}
+
+static int btrfs_run_sanity_tests(void)
+{
+	int ret;
+
+	ret = btrfs_init_test_fs();
+	if (ret)
+		return ret;
+
+	ret = btrfs_test_free_space_cache();
+	if (ret)
+		goto out;
+	ret = btrfs_test_extent_buffer_operations();
+	if (ret)
+		goto out;
+	ret = btrfs_test_extent_io();
+	if (ret)
+		goto out;
+	ret = btrfs_test_inodes();
+	if (ret)
+		goto out;
+	ret = btrfs_test_qgroups();
+out:
+	btrfs_destroy_test_fs();
+	return ret;
 }
 
 static int __init init_btrfs_fs(void)
 {
 	int err;
 
-	err = btrfs_init_sysfs();
+	err = btrfs_hash_init();
 	if (err)
 		return err;
 
+	btrfs_props_init();
+
+	err = btrfs_init_sysfs();
+	if (err)
+		goto free_hash;
+
+	btrfs_init_compress();
+
 	err = btrfs_init_cachep();
 	if (err)
-		goto free_sysfs;
+		goto free_compress;
 
 	err = extent_io_init();
 	if (err)
@@ -868,43 +1943,89 @@ static int __init init_btrfs_fs(void)
 	if (err)
 		goto free_extent_io;
 
-	err = btrfs_interface_init();
+	err = ordered_data_init();
 	if (err)
 		goto free_extent_map;
 
+	err = btrfs_delayed_inode_init();
+	if (err)
+		goto free_ordered_data;
+
+	err = btrfs_auto_defrag_init();
+	if (err)
+		goto free_delayed_inode;
+
+	err = btrfs_delayed_ref_init();
+	if (err)
+		goto free_auto_defrag;
+
+	err = btrfs_prelim_ref_init();
+	if (err)
+		goto free_prelim_ref;
+
+	err = btrfs_interface_init();
+	if (err)
+		goto free_delayed_ref;
+
+	btrfs_init_lockdep();
+
+	btrfs_print_info();
+
+	err = btrfs_run_sanity_tests();
+	if (err)
+		goto unregister_ioctl;
+
 	err = register_filesystem(&btrfs_fs_type);
 	if (err)
 		goto unregister_ioctl;
 
-	printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION);
 	return 0;
 
 unregister_ioctl:
 	btrfs_interface_exit();
+free_prelim_ref:
+	btrfs_prelim_ref_exit();
+free_delayed_ref:
+	btrfs_delayed_ref_exit();
+free_auto_defrag:
+	btrfs_auto_defrag_exit();
+free_delayed_inode:
+	btrfs_delayed_inode_exit();
+free_ordered_data:
+	ordered_data_exit();
 free_extent_map:
 	extent_map_exit();
 free_extent_io:
 	extent_io_exit();
 free_cachep:
 	btrfs_destroy_cachep();
-free_sysfs:
+free_compress:
+	btrfs_exit_compress();
 	btrfs_exit_sysfs();
+free_hash:
+	btrfs_hash_exit();
 	return err;
 }
 
 static void __exit exit_btrfs_fs(void)
 {
 	btrfs_destroy_cachep();
+	btrfs_delayed_ref_exit();
+	btrfs_auto_defrag_exit();
+	btrfs_delayed_inode_exit();
+	btrfs_prelim_ref_exit();
+	ordered_data_exit();
 	extent_map_exit();
 	extent_io_exit();
 	btrfs_interface_exit();
 	unregister_filesystem(&btrfs_fs_type);
 	btrfs_exit_sysfs();
 	btrfs_cleanup_fs_uuids();
-	btrfs_zlib_exit();
+	btrfs_exit_compress();
+	btrfs_hash_exit();
 }
 
-module_init(init_btrfs_fs)
+late_initcall(init_btrfs_fs);
 module_exit(exit_btrfs_fs)
 
 MODULE_LICENSE("GPL");
diff --git a/fs/btrfs/sysfs.c b/fs/btrfs/sysfs.c
index 4ce16ef702a..78699364f53 100644
--- a/fs/btrfs/sysfs.c
+++ b/fs/btrfs/sysfs.c
@@ -21,249 +21,733 @@
 #include <linux/spinlock.h>
 #include <linux/completion.h>
 #include <linux/buffer_head.h>
-#include <linux/module.h>
 #include <linux/kobject.h>
+#include <linux/bug.h>
+#include <linux/genhd.h>
+#include <linux/debugfs.h>
 
 #include "ctree.h"
 #include "disk-io.h"
 #include "transaction.h"
+#include "sysfs.h"
+#include "volumes.h"
 
-static ssize_t root_blocks_used_show(struct btrfs_root *root, char *buf)
+static inline struct btrfs_fs_info *to_fs_info(struct kobject *kobj);
+
+static u64 get_features(struct btrfs_fs_info *fs_info,
+			enum btrfs_feature_set set)
 {
-	return snprintf(buf, PAGE_SIZE, "%llu\n",
-		(unsigned long long)btrfs_root_used(&root->root_item));
+	struct btrfs_super_block *disk_super = fs_info->super_copy;
+	if (set == FEAT_COMPAT)
+		return btrfs_super_compat_flags(disk_super);
+	else if (set == FEAT_COMPAT_RO)
+		return btrfs_super_compat_ro_flags(disk_super);
+	else
+		return btrfs_super_incompat_flags(disk_super);
 }
 
-static ssize_t root_block_limit_show(struct btrfs_root *root, char *buf)
+static void set_features(struct btrfs_fs_info *fs_info,
+			 enum btrfs_feature_set set, u64 features)
 {
-	return snprintf(buf, PAGE_SIZE, "%llu\n",
-		(unsigned long long)btrfs_root_limit(&root->root_item));
+	struct btrfs_super_block *disk_super = fs_info->super_copy;
+	if (set == FEAT_COMPAT)
+		btrfs_set_super_compat_flags(disk_super, features);
+	else if (set == FEAT_COMPAT_RO)
+		btrfs_set_super_compat_ro_flags(disk_super, features);
+	else
+		btrfs_set_super_incompat_flags(disk_super, features);
 }
 
-static ssize_t super_blocks_used_show(struct btrfs_fs_info *fs, char *buf)
+static int can_modify_feature(struct btrfs_feature_attr *fa)
 {
+	int val = 0;
+	u64 set, clear;
+	switch (fa->feature_set) {
+	case FEAT_COMPAT:
+		set = BTRFS_FEATURE_COMPAT_SAFE_SET;
+		clear = BTRFS_FEATURE_COMPAT_SAFE_CLEAR;
+		break;
+	case FEAT_COMPAT_RO:
+		set = BTRFS_FEATURE_COMPAT_RO_SAFE_SET;
+		clear = BTRFS_FEATURE_COMPAT_RO_SAFE_CLEAR;
+		break;
+	case FEAT_INCOMPAT:
+		set = BTRFS_FEATURE_INCOMPAT_SAFE_SET;
+		clear = BTRFS_FEATURE_INCOMPAT_SAFE_CLEAR;
+		break;
+	default:
+		printk(KERN_WARNING "btrfs: sysfs: unknown feature set %d\n",
+				fa->feature_set);
+		return 0;
+	}
+
+	if (set & fa->feature_bit)
+		val |= 1;
+	if (clear & fa->feature_bit)
+		val |= 2;
 
-	return snprintf(buf, PAGE_SIZE, "%llu\n",
-		(unsigned long long)btrfs_super_bytes_used(&fs->super_copy));
+	return val;
 }
 
-static ssize_t super_total_blocks_show(struct btrfs_fs_info *fs, char *buf)
+static ssize_t btrfs_feature_attr_show(struct kobject *kobj,
+				       struct kobj_attribute *a, char *buf)
 {
-	return snprintf(buf, PAGE_SIZE, "%llu\n",
-		(unsigned long long)btrfs_super_total_bytes(&fs->super_copy));
+	int val = 0;
+	struct btrfs_fs_info *fs_info = to_fs_info(kobj);
+	struct btrfs_feature_attr *fa = to_btrfs_feature_attr(a);
+	if (fs_info) {
+		u64 features = get_features(fs_info, fa->feature_set);
+		if (features & fa->feature_bit)
+			val = 1;
+	} else
+		val = can_modify_feature(fa);
+
+	return snprintf(buf, PAGE_SIZE, "%d\n", val);
 }
 
-static ssize_t super_blocksize_show(struct btrfs_fs_info *fs, char *buf)
+static ssize_t btrfs_feature_attr_store(struct kobject *kobj,
+					struct kobj_attribute *a,
+					const char *buf, size_t count)
 {
-	return snprintf(buf, PAGE_SIZE, "%llu\n",
-		(unsigned long long)btrfs_super_sectorsize(&fs->super_copy));
+	struct btrfs_fs_info *fs_info;
+	struct btrfs_feature_attr *fa = to_btrfs_feature_attr(a);
+	struct btrfs_trans_handle *trans;
+	u64 features, set, clear;
+	unsigned long val;
+	int ret;
+
+	fs_info = to_fs_info(kobj);
+	if (!fs_info)
+		return -EPERM;
+
+	ret = kstrtoul(skip_spaces(buf), 0, &val);
+	if (ret)
+		return ret;
+
+	if (fa->feature_set == FEAT_COMPAT) {
+		set = BTRFS_FEATURE_COMPAT_SAFE_SET;
+		clear = BTRFS_FEATURE_COMPAT_SAFE_CLEAR;
+	} else if (fa->feature_set == FEAT_COMPAT_RO) {
+		set = BTRFS_FEATURE_COMPAT_RO_SAFE_SET;
+		clear = BTRFS_FEATURE_COMPAT_RO_SAFE_CLEAR;
+	} else {
+		set = BTRFS_FEATURE_INCOMPAT_SAFE_SET;
+		clear = BTRFS_FEATURE_INCOMPAT_SAFE_CLEAR;
+	}
+
+	features = get_features(fs_info, fa->feature_set);
+
+	/* Nothing to do */
+	if ((val && (features & fa->feature_bit)) ||
+	    (!val && !(features & fa->feature_bit)))
+		return count;
+
+	if ((val && !(set & fa->feature_bit)) ||
+	    (!val && !(clear & fa->feature_bit))) {
+		btrfs_info(fs_info,
+			"%sabling feature %s on mounted fs is not supported.",
+			val ? "En" : "Dis", fa->kobj_attr.attr.name);
+		return -EPERM;
+	}
+
+	btrfs_info(fs_info, "%s %s feature flag",
+		   val ? "Setting" : "Clearing", fa->kobj_attr.attr.name);
+
+	trans = btrfs_start_transaction(fs_info->fs_root, 0);
+	if (IS_ERR(trans))
+		return PTR_ERR(trans);
+
+	spin_lock(&fs_info->super_lock);
+	features = get_features(fs_info, fa->feature_set);
+	if (val)
+		features |= fa->feature_bit;
+	else
+		features &= ~fa->feature_bit;
+	set_features(fs_info, fa->feature_set, features);
+	spin_unlock(&fs_info->super_lock);
+
+	ret = btrfs_commit_transaction(trans, fs_info->fs_root);
+	if (ret)
+		return ret;
+
+	return count;
 }
 
-/* this is for root attrs (subvols/snapshots) */
-struct btrfs_root_attr {
-	struct attribute attr;
-	ssize_t (*show)(struct btrfs_root *, char *);
-	ssize_t (*store)(struct btrfs_root *, const char *, size_t);
-};
+static umode_t btrfs_feature_visible(struct kobject *kobj,
+				     struct attribute *attr, int unused)
+{
+	struct btrfs_fs_info *fs_info = to_fs_info(kobj);
+	umode_t mode = attr->mode;
 
-#define ROOT_ATTR(name, mode, show, store) \
-static struct btrfs_root_attr btrfs_root_attr_##name = __ATTR(name, mode, \
-							      show, store)
+	if (fs_info) {
+		struct btrfs_feature_attr *fa;
+		u64 features;
 
-ROOT_ATTR(blocks_used,	0444,	root_blocks_used_show,	NULL);
-ROOT_ATTR(block_limit,	0644,	root_block_limit_show,	NULL);
+		fa = attr_to_btrfs_feature_attr(attr);
+		features = get_features(fs_info, fa->feature_set);
 
-static struct attribute *btrfs_root_attrs[] = {
-	&btrfs_root_attr_blocks_used.attr,
-	&btrfs_root_attr_block_limit.attr,
-	NULL,
+		if (can_modify_feature(fa))
+			mode |= S_IWUSR;
+		else if (!(features & fa->feature_bit))
+			mode = 0;
+	}
+
+	return mode;
+}
+
+BTRFS_FEAT_ATTR_INCOMPAT(mixed_backref, MIXED_BACKREF);
+BTRFS_FEAT_ATTR_INCOMPAT(default_subvol, DEFAULT_SUBVOL);
+BTRFS_FEAT_ATTR_INCOMPAT(mixed_groups, MIXED_GROUPS);
+BTRFS_FEAT_ATTR_INCOMPAT(compress_lzo, COMPRESS_LZO);
+BTRFS_FEAT_ATTR_INCOMPAT(big_metadata, BIG_METADATA);
+BTRFS_FEAT_ATTR_INCOMPAT(extended_iref, EXTENDED_IREF);
+BTRFS_FEAT_ATTR_INCOMPAT(raid56, RAID56);
+BTRFS_FEAT_ATTR_INCOMPAT(skinny_metadata, SKINNY_METADATA);
+BTRFS_FEAT_ATTR_INCOMPAT(no_holes, NO_HOLES);
+
+static struct attribute *btrfs_supported_feature_attrs[] = {
+	BTRFS_FEAT_ATTR_PTR(mixed_backref),
+	BTRFS_FEAT_ATTR_PTR(default_subvol),
+	BTRFS_FEAT_ATTR_PTR(mixed_groups),
+	BTRFS_FEAT_ATTR_PTR(compress_lzo),
+	BTRFS_FEAT_ATTR_PTR(big_metadata),
+	BTRFS_FEAT_ATTR_PTR(extended_iref),
+	BTRFS_FEAT_ATTR_PTR(raid56),
+	BTRFS_FEAT_ATTR_PTR(skinny_metadata),
+	BTRFS_FEAT_ATTR_PTR(no_holes),
+	NULL
 };
 
-/* this is for super attrs (actual full fs) */
-struct btrfs_super_attr {
-	struct attribute attr;
-	ssize_t (*show)(struct btrfs_fs_info *, char *);
-	ssize_t (*store)(struct btrfs_fs_info *, const char *, size_t);
+static const struct attribute_group btrfs_feature_attr_group = {
+	.name = "features",
+	.is_visible = btrfs_feature_visible,
+	.attrs = btrfs_supported_feature_attrs,
 };
 
-#define SUPER_ATTR(name, mode, show, store) \
-static struct btrfs_super_attr btrfs_super_attr_##name = __ATTR(name, mode, \
-								show, store)
+static ssize_t btrfs_show_u64(u64 *value_ptr, spinlock_t *lock, char *buf)
+{
+	u64 val;
+	if (lock)
+		spin_lock(lock);
+	val = *value_ptr;
+	if (lock)
+		spin_unlock(lock);
+	return snprintf(buf, PAGE_SIZE, "%llu\n", val);
+}
 
-SUPER_ATTR(blocks_used,		0444,	super_blocks_used_show,		NULL);
-SUPER_ATTR(total_blocks,	0444,	super_total_blocks_show,	NULL);
-SUPER_ATTR(blocksize,		0444,	super_blocksize_show,		NULL);
+static ssize_t global_rsv_size_show(struct kobject *kobj,
+				    struct kobj_attribute *ka, char *buf)
+{
+	struct btrfs_fs_info *fs_info = to_fs_info(kobj->parent);
+	struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
+	return btrfs_show_u64(&block_rsv->size, &block_rsv->lock, buf);
+}
+BTRFS_ATTR(global_rsv_size, 0444, global_rsv_size_show);
 
-static struct attribute *btrfs_super_attrs[] = {
-	&btrfs_super_attr_blocks_used.attr,
-	&btrfs_super_attr_total_blocks.attr,
-	&btrfs_super_attr_blocksize.attr,
-	NULL,
+static ssize_t global_rsv_reserved_show(struct kobject *kobj,
+					struct kobj_attribute *a, char *buf)
+{
+	struct btrfs_fs_info *fs_info = to_fs_info(kobj->parent);
+	struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
+	return btrfs_show_u64(&block_rsv->reserved, &block_rsv->lock, buf);
+}
+BTRFS_ATTR(global_rsv_reserved, 0444, global_rsv_reserved_show);
+
+#define to_space_info(_kobj) container_of(_kobj, struct btrfs_space_info, kobj)
+#define to_raid_kobj(_kobj) container_of(_kobj, struct raid_kobject, kobj)
+
+static ssize_t raid_bytes_show(struct kobject *kobj,
+			       struct kobj_attribute *attr, char *buf);
+BTRFS_RAID_ATTR(total_bytes, raid_bytes_show);
+BTRFS_RAID_ATTR(used_bytes, raid_bytes_show);
+
+static ssize_t raid_bytes_show(struct kobject *kobj,
+			       struct kobj_attribute *attr, char *buf)
+
+{
+	struct btrfs_space_info *sinfo = to_space_info(kobj->parent);
+	struct btrfs_block_group_cache *block_group;
+	int index = to_raid_kobj(kobj)->raid_type;
+	u64 val = 0;
+
+	down_read(&sinfo->groups_sem);
+	list_for_each_entry(block_group, &sinfo->block_groups[index], list) {
+		if (&attr->attr == BTRFS_RAID_ATTR_PTR(total_bytes))
+			val += block_group->key.offset;
+		else
+			val += btrfs_block_group_used(&block_group->item);
+	}
+	up_read(&sinfo->groups_sem);
+	return snprintf(buf, PAGE_SIZE, "%llu\n", val);
+}
+
+static struct attribute *raid_attributes[] = {
+	BTRFS_RAID_ATTR_PTR(total_bytes),
+	BTRFS_RAID_ATTR_PTR(used_bytes),
+	NULL
 };
 
-static ssize_t btrfs_super_attr_show(struct kobject *kobj,
-				    struct attribute *attr, char *buf)
+static void release_raid_kobj(struct kobject *kobj)
 {
-	struct btrfs_fs_info *fs = container_of(kobj, struct btrfs_fs_info,
-						super_kobj);
-	struct btrfs_super_attr *a = container_of(attr,
-						  struct btrfs_super_attr,
-						  attr);
+	kfree(to_raid_kobj(kobj));
+}
 
-	return a->show ? a->show(fs, buf) : 0;
+struct kobj_type btrfs_raid_ktype = {
+	.sysfs_ops = &kobj_sysfs_ops,
+	.release = release_raid_kobj,
+	.default_attrs = raid_attributes,
+};
+
+#define SPACE_INFO_ATTR(field)						\
+static ssize_t btrfs_space_info_show_##field(struct kobject *kobj,	\
+					     struct kobj_attribute *a,	\
+					     char *buf)			\
+{									\
+	struct btrfs_space_info *sinfo = to_space_info(kobj);		\
+	return btrfs_show_u64(&sinfo->field, &sinfo->lock, buf);	\
+}									\
+BTRFS_ATTR(field, 0444, btrfs_space_info_show_##field)
+
+static ssize_t btrfs_space_info_show_total_bytes_pinned(struct kobject *kobj,
+						       struct kobj_attribute *a,
+						       char *buf)
+{
+	struct btrfs_space_info *sinfo = to_space_info(kobj);
+	s64 val = percpu_counter_sum(&sinfo->total_bytes_pinned);
+	return snprintf(buf, PAGE_SIZE, "%lld\n", val);
 }
 
-static ssize_t btrfs_super_attr_store(struct kobject *kobj,
-				     struct attribute *attr,
-				     const char *buf, size_t len)
+SPACE_INFO_ATTR(flags);
+SPACE_INFO_ATTR(total_bytes);
+SPACE_INFO_ATTR(bytes_used);
+SPACE_INFO_ATTR(bytes_pinned);
+SPACE_INFO_ATTR(bytes_reserved);
+SPACE_INFO_ATTR(bytes_may_use);
+SPACE_INFO_ATTR(disk_used);
+SPACE_INFO_ATTR(disk_total);
+BTRFS_ATTR(total_bytes_pinned, 0444, btrfs_space_info_show_total_bytes_pinned);
+
+static struct attribute *space_info_attrs[] = {
+	BTRFS_ATTR_PTR(flags),
+	BTRFS_ATTR_PTR(total_bytes),
+	BTRFS_ATTR_PTR(bytes_used),
+	BTRFS_ATTR_PTR(bytes_pinned),
+	BTRFS_ATTR_PTR(bytes_reserved),
+	BTRFS_ATTR_PTR(bytes_may_use),
+	BTRFS_ATTR_PTR(disk_used),
+	BTRFS_ATTR_PTR(disk_total),
+	BTRFS_ATTR_PTR(total_bytes_pinned),
+	NULL,
+};
+
+static void space_info_release(struct kobject *kobj)
 {
-	struct btrfs_fs_info *fs = container_of(kobj, struct btrfs_fs_info,
-						super_kobj);
-	struct btrfs_super_attr *a = container_of(attr,
-						  struct btrfs_super_attr,
-						  attr);
+	struct btrfs_space_info *sinfo = to_space_info(kobj);
+	percpu_counter_destroy(&sinfo->total_bytes_pinned);
+	kfree(sinfo);
+}
+
+struct kobj_type space_info_ktype = {
+	.sysfs_ops = &kobj_sysfs_ops,
+	.release = space_info_release,
+	.default_attrs = space_info_attrs,
+};
 
-	return a->store ? a->store(fs, buf, len) : 0;
+static const struct attribute *allocation_attrs[] = {
+	BTRFS_ATTR_PTR(global_rsv_reserved),
+	BTRFS_ATTR_PTR(global_rsv_size),
+	NULL,
+};
+
+static ssize_t btrfs_label_show(struct kobject *kobj,
+				struct kobj_attribute *a, char *buf)
+{
+	struct btrfs_fs_info *fs_info = to_fs_info(kobj);
+	return snprintf(buf, PAGE_SIZE, "%s\n", fs_info->super_copy->label);
 }
 
-static ssize_t btrfs_root_attr_show(struct kobject *kobj,
-				    struct attribute *attr, char *buf)
+static ssize_t btrfs_label_store(struct kobject *kobj,
+				 struct kobj_attribute *a,
+				 const char *buf, size_t len)
 {
-	struct btrfs_root *root = container_of(kobj, struct btrfs_root,
-						root_kobj);
-	struct btrfs_root_attr *a = container_of(attr,
-						 struct btrfs_root_attr,
-						 attr);
+	struct btrfs_fs_info *fs_info = to_fs_info(kobj);
+	struct btrfs_trans_handle *trans;
+	struct btrfs_root *root = fs_info->fs_root;
+	int ret;
+
+	if (len >= BTRFS_LABEL_SIZE)
+		return -EINVAL;
+
+	trans = btrfs_start_transaction(root, 0);
+	if (IS_ERR(trans))
+		return PTR_ERR(trans);
+
+	spin_lock(&root->fs_info->super_lock);
+	strcpy(fs_info->super_copy->label, buf);
+	spin_unlock(&root->fs_info->super_lock);
+	ret = btrfs_commit_transaction(trans, root);
+
+	if (!ret)
+		return len;
+
+	return ret;
+}
+BTRFS_ATTR_RW(label, 0644, btrfs_label_show, btrfs_label_store);
 
-	return a->show ? a->show(root, buf) : 0;
+static ssize_t btrfs_no_store(struct kobject *kobj,
+				 struct kobj_attribute *a,
+				 const char *buf, size_t len)
+{
+	return -EPERM;
 }
 
-static ssize_t btrfs_root_attr_store(struct kobject *kobj,
-				     struct attribute *attr,
-				     const char *buf, size_t len)
+static ssize_t btrfs_nodesize_show(struct kobject *kobj,
+				struct kobj_attribute *a, char *buf)
 {
-	struct btrfs_root *root = container_of(kobj, struct btrfs_root,
-						root_kobj);
-	struct btrfs_root_attr *a = container_of(attr,
-						 struct btrfs_root_attr,
-						 attr);
-	return a->store ? a->store(root, buf, len) : 0;
+	struct btrfs_fs_info *fs_info = to_fs_info(kobj);
+
+	return snprintf(buf, PAGE_SIZE, "%u\n", fs_info->super_copy->nodesize);
 }
 
-static void btrfs_super_release(struct kobject *kobj)
+BTRFS_ATTR_RW(nodesize, 0444, btrfs_nodesize_show, btrfs_no_store);
+
+static ssize_t btrfs_sectorsize_show(struct kobject *kobj,
+				struct kobj_attribute *a, char *buf)
 {
-	struct btrfs_fs_info *fs = container_of(kobj, struct btrfs_fs_info,
-						super_kobj);
-	complete(&fs->kobj_unregister);
+	struct btrfs_fs_info *fs_info = to_fs_info(kobj);
+
+	return snprintf(buf, PAGE_SIZE, "%u\n", fs_info->super_copy->sectorsize);
 }
 
-static void btrfs_root_release(struct kobject *kobj)
+BTRFS_ATTR_RW(sectorsize, 0444, btrfs_sectorsize_show, btrfs_no_store);
+
+static ssize_t btrfs_clone_alignment_show(struct kobject *kobj,
+				struct kobj_attribute *a, char *buf)
 {
-	struct btrfs_root *root = container_of(kobj, struct btrfs_root,
-						root_kobj);
-	complete(&root->kobj_unregister);
+	struct btrfs_fs_info *fs_info = to_fs_info(kobj);
+
+	return snprintf(buf, PAGE_SIZE, "%u\n", fs_info->super_copy->sectorsize);
 }
 
-static const struct sysfs_ops btrfs_super_attr_ops = {
-	.show	= btrfs_super_attr_show,
-	.store	= btrfs_super_attr_store,
-};
+BTRFS_ATTR_RW(clone_alignment, 0444, btrfs_clone_alignment_show, btrfs_no_store);
 
-static const struct sysfs_ops btrfs_root_attr_ops = {
-	.show	= btrfs_root_attr_show,
-	.store	= btrfs_root_attr_store,
+static struct attribute *btrfs_attrs[] = {
+	BTRFS_ATTR_PTR(label),
+	BTRFS_ATTR_PTR(nodesize),
+	BTRFS_ATTR_PTR(sectorsize),
+	BTRFS_ATTR_PTR(clone_alignment),
+	NULL,
 };
 
-static struct kobj_type btrfs_root_ktype = {
-	.default_attrs	= btrfs_root_attrs,
-	.sysfs_ops	= &btrfs_root_attr_ops,
-	.release	= btrfs_root_release,
+static void btrfs_release_super_kobj(struct kobject *kobj)
+{
+	struct btrfs_fs_info *fs_info = to_fs_info(kobj);
+	complete(&fs_info->kobj_unregister);
+}
+
+static struct kobj_type btrfs_ktype = {
+	.sysfs_ops	= &kobj_sysfs_ops,
+	.release	= btrfs_release_super_kobj,
+	.default_attrs	= btrfs_attrs,
 };
 
-static struct kobj_type btrfs_super_ktype = {
-	.default_attrs	= btrfs_super_attrs,
-	.sysfs_ops	= &btrfs_super_attr_ops,
-	.release	= btrfs_super_release,
+static inline struct btrfs_fs_info *to_fs_info(struct kobject *kobj)
+{
+	if (kobj->ktype != &btrfs_ktype)
+		return NULL;
+	return container_of(kobj, struct btrfs_fs_info, super_kobj);
+}
+
+#define NUM_FEATURE_BITS 64
+static char btrfs_unknown_feature_names[3][NUM_FEATURE_BITS][13];
+static struct btrfs_feature_attr btrfs_feature_attrs[3][NUM_FEATURE_BITS];
+
+static u64 supported_feature_masks[3] = {
+	[FEAT_COMPAT]    = BTRFS_FEATURE_COMPAT_SUPP,
+	[FEAT_COMPAT_RO] = BTRFS_FEATURE_COMPAT_RO_SUPP,
+	[FEAT_INCOMPAT]  = BTRFS_FEATURE_INCOMPAT_SUPP,
 };
 
-/* /sys/fs/btrfs/ entry */
-static struct kset *btrfs_kset;
+static int addrm_unknown_feature_attrs(struct btrfs_fs_info *fs_info, bool add)
+{
+	int set;
+
+	for (set = 0; set < FEAT_MAX; set++) {
+		int i;
+		struct attribute *attrs[2];
+		struct attribute_group agroup = {
+			.name = "features",
+			.attrs = attrs,
+		};
+		u64 features = get_features(fs_info, set);
+		features &= ~supported_feature_masks[set];
+
+		if (!features)
+			continue;
+
+		attrs[1] = NULL;
+		for (i = 0; i < NUM_FEATURE_BITS; i++) {
+			struct btrfs_feature_attr *fa;
+
+			if (!(features & (1ULL << i)))
+				continue;
+
+			fa = &btrfs_feature_attrs[set][i];
+			attrs[0] = &fa->kobj_attr.attr;
+			if (add) {
+				int ret;
+				ret = sysfs_merge_group(&fs_info->super_kobj,
+							&agroup);
+				if (ret)
+					return ret;
+			} else
+				sysfs_unmerge_group(&fs_info->super_kobj,
+						    &agroup);
+		}
+
+	}
+	return 0;
+}
 
-int btrfs_sysfs_add_super(struct btrfs_fs_info *fs)
+static void __btrfs_sysfs_remove_one(struct btrfs_fs_info *fs_info)
 {
-	int error;
-	char *name;
-	char c;
-	int len = strlen(fs->sb->s_id) + 1;
+	kobject_del(&fs_info->super_kobj);
+	kobject_put(&fs_info->super_kobj);
+	wait_for_completion(&fs_info->kobj_unregister);
+}
+
+void btrfs_sysfs_remove_one(struct btrfs_fs_info *fs_info)
+{
+	if (fs_info->space_info_kobj) {
+		sysfs_remove_files(fs_info->space_info_kobj, allocation_attrs);
+		kobject_del(fs_info->space_info_kobj);
+		kobject_put(fs_info->space_info_kobj);
+	}
+	kobject_del(fs_info->device_dir_kobj);
+	kobject_put(fs_info->device_dir_kobj);
+	addrm_unknown_feature_attrs(fs_info, false);
+	sysfs_remove_group(&fs_info->super_kobj, &btrfs_feature_attr_group);
+	__btrfs_sysfs_remove_one(fs_info);
+}
+
+const char * const btrfs_feature_set_names[3] = {
+	[FEAT_COMPAT]	 = "compat",
+	[FEAT_COMPAT_RO] = "compat_ro",
+	[FEAT_INCOMPAT]	 = "incompat",
+};
+
+char *btrfs_printable_features(enum btrfs_feature_set set, u64 flags)
+{
+	size_t bufsize = 4096; /* safe max, 64 names * 64 bytes */
+	int len = 0;
 	int i;
+	char *str;
 
-	name = kmalloc(len, GFP_NOFS);
-	if (!name) {
-		error = -ENOMEM;
-		goto fail;
+	str = kmalloc(bufsize, GFP_KERNEL);
+	if (!str)
+		return str;
+
+	for (i = 0; i < ARRAY_SIZE(btrfs_feature_attrs[set]); i++) {
+		const char *name;
+
+		if (!(flags & (1ULL << i)))
+			continue;
+
+		name = btrfs_feature_attrs[set][i].kobj_attr.attr.name;
+		len += snprintf(str + len, bufsize - len, "%s%s",
+				len ? "," : "", name);
 	}
 
-	for (i = 0; i < len; i++) {
-		c = fs->sb->s_id[i];
-		if (c == '/' || c == '\\')
-			c = '!';
-		name[i] = c;
+	return str;
+}
+
+static void init_feature_attrs(void)
+{
+	struct btrfs_feature_attr *fa;
+	int set, i;
+
+	BUILD_BUG_ON(ARRAY_SIZE(btrfs_unknown_feature_names) !=
+		     ARRAY_SIZE(btrfs_feature_attrs));
+	BUILD_BUG_ON(ARRAY_SIZE(btrfs_unknown_feature_names[0]) !=
+		     ARRAY_SIZE(btrfs_feature_attrs[0]));
+
+	memset(btrfs_feature_attrs, 0, sizeof(btrfs_feature_attrs));
+	memset(btrfs_unknown_feature_names, 0,
+	       sizeof(btrfs_unknown_feature_names));
+
+	for (i = 0; btrfs_supported_feature_attrs[i]; i++) {
+		struct btrfs_feature_attr *sfa;
+		struct attribute *a = btrfs_supported_feature_attrs[i];
+		int bit;
+		sfa = attr_to_btrfs_feature_attr(a);
+		bit = ilog2(sfa->feature_bit);
+		fa = &btrfs_feature_attrs[sfa->feature_set][bit];
+
+		fa->kobj_attr.attr.name = sfa->kobj_attr.attr.name;
 	}
-	name[len] = '\0';
 
-	fs->super_kobj.kset = btrfs_kset;
-	error = kobject_init_and_add(&fs->super_kobj, &btrfs_super_ktype,
-				     NULL, "%s", name);
-	kfree(name);
-	if (error)
-		goto fail;
+	for (set = 0; set < FEAT_MAX; set++) {
+		for (i = 0; i < ARRAY_SIZE(btrfs_feature_attrs[set]); i++) {
+			char *name = btrfs_unknown_feature_names[set][i];
+			fa = &btrfs_feature_attrs[set][i];
+
+			if (fa->kobj_attr.attr.name)
+				continue;
+
+			snprintf(name, 13, "%s:%u",
+				 btrfs_feature_set_names[set], i);
+
+			fa->kobj_attr.attr.name = name;
+			fa->kobj_attr.attr.mode = S_IRUGO;
+			fa->feature_set = set;
+			fa->feature_bit = 1ULL << i;
+		}
+	}
+}
+
+int btrfs_kobj_rm_device(struct btrfs_fs_info *fs_info,
+		struct btrfs_device *one_device)
+{
+	struct hd_struct *disk;
+	struct kobject *disk_kobj;
+
+	if (!fs_info->device_dir_kobj)
+		return -EINVAL;
+
+	if (one_device) {
+		disk = one_device->bdev->bd_part;
+		disk_kobj = &part_to_dev(disk)->kobj;
+
+		sysfs_remove_link(fs_info->device_dir_kobj,
+						disk_kobj->name);
+	}
 
 	return 0;
+}
+
+int btrfs_kobj_add_device(struct btrfs_fs_info *fs_info,
+		struct btrfs_device *one_device)
+{
+	int error = 0;
+	struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
+	struct btrfs_device *dev;
+
+	if (!fs_info->device_dir_kobj)
+		fs_info->device_dir_kobj = kobject_create_and_add("devices",
+						&fs_info->super_kobj);
+
+	if (!fs_info->device_dir_kobj)
+		return -ENOMEM;
+
+	list_for_each_entry(dev, &fs_devices->devices, dev_list) {
+		struct hd_struct *disk;
+		struct kobject *disk_kobj;
+
+		if (!dev->bdev)
+			continue;
+
+		if (one_device && one_device != dev)
+			continue;
+
+		disk = dev->bdev->bd_part;
+		disk_kobj = &part_to_dev(disk)->kobj;
+
+		error = sysfs_create_link(fs_info->device_dir_kobj,
+					  disk_kobj, disk_kobj->name);
+		if (error)
+			break;
+	}
 
-fail:
-	printk(KERN_ERR "btrfs: sysfs creation for super failed\n");
 	return error;
 }
 
-int btrfs_sysfs_add_root(struct btrfs_root *root)
+/* /sys/fs/btrfs/ entry */
+static struct kset *btrfs_kset;
+
+/* /sys/kernel/debug/btrfs */
+static struct dentry *btrfs_debugfs_root_dentry;
+
+/* Debugging tunables and exported data */
+u64 btrfs_debugfs_test;
+
+int btrfs_sysfs_add_one(struct btrfs_fs_info *fs_info)
 {
 	int error;
 
-	error = kobject_init_and_add(&root->root_kobj, &btrfs_root_ktype,
-				     &root->fs_info->super_kobj,
-				     "%s", root->name);
+	init_completion(&fs_info->kobj_unregister);
+	fs_info->super_kobj.kset = btrfs_kset;
+	error = kobject_init_and_add(&fs_info->super_kobj, &btrfs_ktype, NULL,
+				     "%pU", fs_info->fsid);
 	if (error)
-		goto fail;
+		return error;
 
-	return 0;
+	error = sysfs_create_group(&fs_info->super_kobj,
+				   &btrfs_feature_attr_group);
+	if (error) {
+		__btrfs_sysfs_remove_one(fs_info);
+		return error;
+	}
+
+	error = addrm_unknown_feature_attrs(fs_info, true);
+	if (error)
+		goto failure;
+
+	error = btrfs_kobj_add_device(fs_info, NULL);
+	if (error)
+		goto failure;
 
-fail:
-	printk(KERN_ERR "btrfs: sysfs creation for root failed\n");
+	fs_info->space_info_kobj = kobject_create_and_add("allocation",
+						  &fs_info->super_kobj);
+	if (!fs_info->space_info_kobj) {
+		error = -ENOMEM;
+		goto failure;
+	}
+
+	error = sysfs_create_files(fs_info->space_info_kobj, allocation_attrs);
+	if (error)
+		goto failure;
+
+	return 0;
+failure:
+	btrfs_sysfs_remove_one(fs_info);
 	return error;
 }
 
-void btrfs_sysfs_del_root(struct btrfs_root *root)
+static int btrfs_init_debugfs(void)
 {
-	kobject_put(&root->root_kobj);
-	wait_for_completion(&root->kobj_unregister);
-}
+#ifdef CONFIG_DEBUG_FS
+	btrfs_debugfs_root_dentry = debugfs_create_dir("btrfs", NULL);
+	if (!btrfs_debugfs_root_dentry)
+		return -ENOMEM;
 
-void btrfs_sysfs_del_super(struct btrfs_fs_info *fs)
-{
-	kobject_put(&fs->super_kobj);
-	wait_for_completion(&fs->kobj_unregister);
+	debugfs_create_u64("test", S_IRUGO | S_IWUGO, btrfs_debugfs_root_dentry,
+			&btrfs_debugfs_test);
+#endif
+	return 0;
 }
 
 int btrfs_init_sysfs(void)
 {
+	int ret;
+
 	btrfs_kset = kset_create_and_add("btrfs", NULL, fs_kobj);
 	if (!btrfs_kset)
 		return -ENOMEM;
-	return 0;
+
+	ret = btrfs_init_debugfs();
+	if (ret)
+		return ret;
+
+	init_feature_attrs();
+	ret = sysfs_create_group(&btrfs_kset->kobj, &btrfs_feature_attr_group);
+
+	return ret;
 }
 
 void btrfs_exit_sysfs(void)
 {
+	sysfs_remove_group(&btrfs_kset->kobj, &btrfs_feature_attr_group);
 	kset_unregister(btrfs_kset);
+	debugfs_remove_recursive(btrfs_debugfs_root_dentry);
 }
 
diff --git a/fs/btrfs/sysfs.h b/fs/btrfs/sysfs.h
new file mode 100644
index 00000000000..ac46df37504
--- /dev/null
+++ b/fs/btrfs/sysfs.h
@@ -0,0 +1,73 @@
+#ifndef _BTRFS_SYSFS_H_
+#define _BTRFS_SYSFS_H_
+
+/*
+ * Data exported through sysfs
+ */
+extern u64 btrfs_debugfs_test;
+
+enum btrfs_feature_set {
+	FEAT_COMPAT,
+	FEAT_COMPAT_RO,
+	FEAT_INCOMPAT,
+	FEAT_MAX
+};
+
+#define __INIT_KOBJ_ATTR(_name, _mode, _show, _store)			\
+{									\
+	.attr	= { .name = __stringify(_name), .mode = _mode },	\
+	.show	= _show,						\
+	.store	= _store,						\
+}
+
+#define BTRFS_ATTR_RW(_name, _mode, _show, _store)			\
+static struct kobj_attribute btrfs_attr_##_name =			\
+			__INIT_KOBJ_ATTR(_name, _mode, _show, _store)
+#define BTRFS_ATTR(_name, _mode, _show)					\
+	BTRFS_ATTR_RW(_name, _mode, _show, NULL)
+#define BTRFS_ATTR_PTR(_name)    (&btrfs_attr_##_name.attr)
+
+#define BTRFS_RAID_ATTR(_name, _show)					\
+static struct kobj_attribute btrfs_raid_attr_##_name =			\
+			__INIT_KOBJ_ATTR(_name, 0444, _show, NULL)
+#define BTRFS_RAID_ATTR_PTR(_name)    (&btrfs_raid_attr_##_name.attr)
+
+
+struct btrfs_feature_attr {
+	struct kobj_attribute kobj_attr;
+	enum btrfs_feature_set feature_set;
+	u64 feature_bit;
+};
+
+#define BTRFS_FEAT_ATTR(_name, _feature_set, _prefix, _feature_bit)	     \
+static struct btrfs_feature_attr btrfs_attr_##_name = {			     \
+	.kobj_attr = __INIT_KOBJ_ATTR(_name, S_IRUGO,			     \
+				      btrfs_feature_attr_show,		     \
+				      btrfs_feature_attr_store),	     \
+	.feature_set	= _feature_set,					     \
+	.feature_bit	= _prefix ##_## _feature_bit,			     \
+}
+#define BTRFS_FEAT_ATTR_PTR(_name)    (&btrfs_attr_##_name.kobj_attr.attr)
+
+#define BTRFS_FEAT_ATTR_COMPAT(name, feature) \
+	BTRFS_FEAT_ATTR(name, FEAT_COMPAT, BTRFS_FEATURE_COMPAT, feature)
+#define BTRFS_FEAT_ATTR_COMPAT_RO(name, feature) \
+	BTRFS_FEAT_ATTR(name, FEAT_COMPAT_RO, BTRFS_FEATURE_COMPAT, feature)
+#define BTRFS_FEAT_ATTR_INCOMPAT(name, feature) \
+	BTRFS_FEAT_ATTR(name, FEAT_INCOMPAT, BTRFS_FEATURE_INCOMPAT, feature)
+
+/* convert from attribute */
+#define to_btrfs_feature_attr(a) \
+			container_of(a, struct btrfs_feature_attr, kobj_attr)
+#define attr_to_btrfs_attr(a) container_of(a, struct kobj_attribute, attr)
+#define attr_to_btrfs_feature_attr(a) \
+			to_btrfs_feature_attr(attr_to_btrfs_attr(a))
+char *btrfs_printable_features(enum btrfs_feature_set set, u64 flags);
+extern const char * const btrfs_feature_set_names[3];
+extern struct kobj_type space_info_ktype;
+extern struct kobj_type btrfs_raid_ktype;
+int btrfs_kobj_add_device(struct btrfs_fs_info *fs_info,
+		struct btrfs_device *one_device);
+int btrfs_kobj_rm_device(struct btrfs_fs_info *fs_info,
+                struct btrfs_device *one_device);
+#endif /* _BTRFS_SYSFS_H_ */
diff --git a/fs/btrfs/tests/btrfs-tests.c b/fs/btrfs/tests/btrfs-tests.c
new file mode 100644
index 00000000000..9626252ee6b
--- /dev/null
+++ b/fs/btrfs/tests/btrfs-tests.c
@@ -0,0 +1,171 @@
+/*
+ * Copyright (C) 2013 Fusion IO.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/fs.h>
+#include <linux/mount.h>
+#include <linux/magic.h>
+#include "btrfs-tests.h"
+#include "../ctree.h"
+#include "../volumes.h"
+#include "../disk-io.h"
+#include "../qgroup.h"
+
+static struct vfsmount *test_mnt = NULL;
+
+static const struct super_operations btrfs_test_super_ops = {
+	.alloc_inode	= btrfs_alloc_inode,
+	.destroy_inode	= btrfs_test_destroy_inode,
+};
+
+static struct dentry *btrfs_test_mount(struct file_system_type *fs_type,
+				       int flags, const char *dev_name,
+				       void *data)
+{
+	return mount_pseudo(fs_type, "btrfs_test:", &btrfs_test_super_ops,
+			    NULL, BTRFS_TEST_MAGIC);
+}
+
+static struct file_system_type test_type = {
+	.name		= "btrfs_test_fs",
+	.mount		= btrfs_test_mount,
+	.kill_sb	= kill_anon_super,
+};
+
+struct inode *btrfs_new_test_inode(void)
+{
+	return new_inode(test_mnt->mnt_sb);
+}
+
+int btrfs_init_test_fs(void)
+{
+	int ret;
+
+	ret = register_filesystem(&test_type);
+	if (ret) {
+		printk(KERN_ERR "btrfs: cannot register test file system\n");
+		return ret;
+	}
+
+	test_mnt = kern_mount(&test_type);
+	if (IS_ERR(test_mnt)) {
+		printk(KERN_ERR "btrfs: cannot mount test file system\n");
+		unregister_filesystem(&test_type);
+		return ret;
+	}
+	return 0;
+}
+
+void btrfs_destroy_test_fs(void)
+{
+	kern_unmount(test_mnt);
+	unregister_filesystem(&test_type);
+}
+
+struct btrfs_fs_info *btrfs_alloc_dummy_fs_info(void)
+{
+	struct btrfs_fs_info *fs_info = kzalloc(sizeof(struct btrfs_fs_info),
+						GFP_NOFS);
+
+	if (!fs_info)
+		return fs_info;
+	fs_info->fs_devices = kzalloc(sizeof(struct btrfs_fs_devices),
+				      GFP_NOFS);
+	if (!fs_info->fs_devices) {
+		kfree(fs_info);
+		return NULL;
+	}
+	fs_info->super_copy = kzalloc(sizeof(struct btrfs_super_block),
+				      GFP_NOFS);
+	if (!fs_info->super_copy) {
+		kfree(fs_info->fs_devices);
+		kfree(fs_info);
+		return NULL;
+	}
+
+	if (init_srcu_struct(&fs_info->subvol_srcu)) {
+		kfree(fs_info->fs_devices);
+		kfree(fs_info->super_copy);
+		kfree(fs_info);
+		return NULL;
+	}
+
+	spin_lock_init(&fs_info->buffer_lock);
+	spin_lock_init(&fs_info->qgroup_lock);
+	spin_lock_init(&fs_info->qgroup_op_lock);
+	spin_lock_init(&fs_info->super_lock);
+	spin_lock_init(&fs_info->fs_roots_radix_lock);
+	spin_lock_init(&fs_info->tree_mod_seq_lock);
+	mutex_init(&fs_info->qgroup_ioctl_lock);
+	mutex_init(&fs_info->qgroup_rescan_lock);
+	rwlock_init(&fs_info->tree_mod_log_lock);
+	fs_info->running_transaction = NULL;
+	fs_info->qgroup_tree = RB_ROOT;
+	fs_info->qgroup_ulist = NULL;
+	atomic64_set(&fs_info->tree_mod_seq, 0);
+	INIT_LIST_HEAD(&fs_info->dirty_qgroups);
+	INIT_LIST_HEAD(&fs_info->dead_roots);
+	INIT_LIST_HEAD(&fs_info->tree_mod_seq_list);
+	INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC);
+	INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
+	return fs_info;
+}
+
+static void btrfs_free_dummy_fs_info(struct btrfs_fs_info *fs_info)
+{
+	struct radix_tree_iter iter;
+	void **slot;
+
+	spin_lock(&fs_info->buffer_lock);
+restart:
+	radix_tree_for_each_slot(slot, &fs_info->buffer_radix, &iter, 0) {
+		struct extent_buffer *eb;
+
+		eb = radix_tree_deref_slot_protected(slot, &fs_info->buffer_lock);
+		if (!eb)
+			continue;
+		/* Shouldn't happen but that kind of thinking creates CVE's */
+		if (radix_tree_exception(eb)) {
+			if (radix_tree_deref_retry(eb))
+				goto restart;
+			continue;
+		}
+		spin_unlock(&fs_info->buffer_lock);
+		free_extent_buffer_stale(eb);
+		spin_lock(&fs_info->buffer_lock);
+	}
+	spin_unlock(&fs_info->buffer_lock);
+
+	btrfs_free_qgroup_config(fs_info);
+	btrfs_free_fs_roots(fs_info);
+	cleanup_srcu_struct(&fs_info->subvol_srcu);
+	kfree(fs_info->super_copy);
+	kfree(fs_info->fs_devices);
+	kfree(fs_info);
+}
+
+void btrfs_free_dummy_root(struct btrfs_root *root)
+{
+	if (!root)
+		return;
+	if (root->node)
+		free_extent_buffer(root->node);
+	if (root->fs_info)
+		btrfs_free_dummy_fs_info(root->fs_info);
+	kfree(root);
+}
+
diff --git a/fs/btrfs/tests/btrfs-tests.h b/fs/btrfs/tests/btrfs-tests.h
new file mode 100644
index 00000000000..fd395422448
--- /dev/null
+++ b/fs/btrfs/tests/btrfs-tests.h
@@ -0,0 +1,68 @@
+/*
+ * Copyright (C) 2013 Fusion IO.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef __BTRFS_TESTS
+#define __BTRFS_TESTS
+
+#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
+
+#define test_msg(fmt, ...) pr_info("BTRFS: selftest: " fmt, ##__VA_ARGS__)
+
+struct btrfs_root;
+
+int btrfs_test_free_space_cache(void);
+int btrfs_test_extent_buffer_operations(void);
+int btrfs_test_extent_io(void);
+int btrfs_test_inodes(void);
+int btrfs_test_qgroups(void);
+int btrfs_init_test_fs(void);
+void btrfs_destroy_test_fs(void);
+struct inode *btrfs_new_test_inode(void);
+struct btrfs_fs_info *btrfs_alloc_dummy_fs_info(void);
+void btrfs_free_dummy_root(struct btrfs_root *root);
+#else
+static inline int btrfs_test_free_space_cache(void)
+{
+	return 0;
+}
+static inline int btrfs_test_extent_buffer_operations(void)
+{
+	return 0;
+}
+static inline int btrfs_init_test_fs(void)
+{
+	return 0;
+}
+static inline void btrfs_destroy_test_fs(void)
+{
+}
+static inline int btrfs_test_extent_io(void)
+{
+	return 0;
+}
+static inline int btrfs_test_inodes(void)
+{
+	return 0;
+}
+static inline int btrfs_test_qgroups(void)
+{
+	return 0;
+}
+#endif
+
+#endif
diff --git a/fs/btrfs/tests/extent-buffer-tests.c b/fs/btrfs/tests/extent-buffer-tests.c
new file mode 100644
index 00000000000..cc286ce97d1
--- /dev/null
+++ b/fs/btrfs/tests/extent-buffer-tests.c
@@ -0,0 +1,229 @@
+/*
+ * Copyright (C) 2013 Fusion IO.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/slab.h>
+#include "btrfs-tests.h"
+#include "../ctree.h"
+#include "../extent_io.h"
+#include "../disk-io.h"
+
+static int test_btrfs_split_item(void)
+{
+	struct btrfs_path *path;
+	struct btrfs_root *root;
+	struct extent_buffer *eb;
+	struct btrfs_item *item;
+	char *value = "mary had a little lamb";
+	char *split1 = "mary had a little";
+	char *split2 = " lamb";
+	char *split3 = "mary";
+	char *split4 = " had a little";
+	char buf[32];
+	struct btrfs_key key;
+	u32 value_len = strlen(value);
+	int ret = 0;
+
+	test_msg("Running btrfs_split_item tests\n");
+
+	root = btrfs_alloc_dummy_root();
+	if (IS_ERR(root)) {
+		test_msg("Could not allocate root\n");
+		return PTR_ERR(root);
+	}
+
+	path = btrfs_alloc_path();
+	if (!path) {
+		test_msg("Could not allocate path\n");
+		kfree(root);
+		return -ENOMEM;
+	}
+
+	path->nodes[0] = eb = alloc_dummy_extent_buffer(0, 4096);
+	if (!eb) {
+		test_msg("Could not allocate dummy buffer\n");
+		ret = -ENOMEM;
+		goto out;
+	}
+	path->slots[0] = 0;
+
+	key.objectid = 0;
+	key.type = BTRFS_EXTENT_CSUM_KEY;
+	key.offset = 0;
+
+	setup_items_for_insert(root, path, &key, &value_len, value_len,
+			       value_len + sizeof(struct btrfs_item), 1);
+	item = btrfs_item_nr(0);
+	write_extent_buffer(eb, value, btrfs_item_ptr_offset(eb, 0),
+			    value_len);
+
+	key.offset = 3;
+
+	/*
+	 * Passing NULL trans here should be safe because we have plenty of
+	 * space in this leaf to split the item without having to split the
+	 * leaf.
+	 */
+	ret = btrfs_split_item(NULL, root, path, &key, 17);
+	if (ret) {
+		test_msg("Split item failed %d\n", ret);
+		goto out;
+	}
+
+	/*
+	 * Read the first slot, it should have the original key and contain only
+	 * 'mary had a little'
+	 */
+	btrfs_item_key_to_cpu(eb, &key, 0);
+	if (key.objectid != 0 || key.type != BTRFS_EXTENT_CSUM_KEY ||
+	    key.offset != 0) {
+		test_msg("Invalid key at slot 0\n");
+		ret = -EINVAL;
+		goto out;
+	}
+
+	item = btrfs_item_nr(0);
+	if (btrfs_item_size(eb, item) != strlen(split1)) {
+		test_msg("Invalid len in the first split\n");
+		ret = -EINVAL;
+		goto out;
+	}
+
+	read_extent_buffer(eb, buf, btrfs_item_ptr_offset(eb, 0),
+			   strlen(split1));
+	if (memcmp(buf, split1, strlen(split1))) {
+		test_msg("Data in the buffer doesn't match what it should "
+			 "in the first split have='%.*s' want '%s'\n",
+			 (int)strlen(split1), buf, split1);
+		ret = -EINVAL;
+		goto out;
+	}
+
+	btrfs_item_key_to_cpu(eb, &key, 1);
+	if (key.objectid != 0 || key.type != BTRFS_EXTENT_CSUM_KEY ||
+	    key.offset != 3) {
+		test_msg("Invalid key at slot 1\n");
+		ret = -EINVAL;
+		goto out;
+	}
+
+	item = btrfs_item_nr(1);
+	if (btrfs_item_size(eb, item) != strlen(split2)) {
+		test_msg("Invalid len in the second split\n");
+		ret = -EINVAL;
+		goto out;
+	}
+
+	read_extent_buffer(eb, buf, btrfs_item_ptr_offset(eb, 1),
+			   strlen(split2));
+	if (memcmp(buf, split2, strlen(split2))) {
+		test_msg("Data in the buffer doesn't match what it should "
+			 "in the second split\n");
+		ret = -EINVAL;
+		goto out;
+	}
+
+	key.offset = 1;
+	/* Do it again so we test memmoving the other items in the leaf */
+	ret = btrfs_split_item(NULL, root, path, &key, 4);
+	if (ret) {
+		test_msg("Second split item failed %d\n", ret);
+		goto out;
+	}
+
+	btrfs_item_key_to_cpu(eb, &key, 0);
+	if (key.objectid != 0 || key.type != BTRFS_EXTENT_CSUM_KEY ||
+	    key.offset != 0) {
+		test_msg("Invalid key at slot 0\n");
+		ret = -EINVAL;
+		goto out;
+	}
+
+	item = btrfs_item_nr(0);
+	if (btrfs_item_size(eb, item) != strlen(split3)) {
+		test_msg("Invalid len in the first split\n");
+		ret = -EINVAL;
+		goto out;
+	}
+
+	read_extent_buffer(eb, buf, btrfs_item_ptr_offset(eb, 0),
+			   strlen(split3));
+	if (memcmp(buf, split3, strlen(split3))) {
+		test_msg("Data in the buffer doesn't match what it should "
+			 "in the third split");
+		ret = -EINVAL;
+		goto out;
+	}
+
+	btrfs_item_key_to_cpu(eb, &key, 1);
+	if (key.objectid != 0 || key.type != BTRFS_EXTENT_CSUM_KEY ||
+	    key.offset != 1) {
+		test_msg("Invalid key at slot 1\n");
+		ret = -EINVAL;
+		goto out;
+	}
+
+	item = btrfs_item_nr(1);
+	if (btrfs_item_size(eb, item) != strlen(split4)) {
+		test_msg("Invalid len in the second split\n");
+		ret = -EINVAL;
+		goto out;
+	}
+
+	read_extent_buffer(eb, buf, btrfs_item_ptr_offset(eb, 1),
+			   strlen(split4));
+	if (memcmp(buf, split4, strlen(split4))) {
+		test_msg("Data in the buffer doesn't match what it should "
+			 "in the fourth split\n");
+		ret = -EINVAL;
+		goto out;
+	}
+
+	btrfs_item_key_to_cpu(eb, &key, 2);
+	if (key.objectid != 0 || key.type != BTRFS_EXTENT_CSUM_KEY ||
+	    key.offset != 3) {
+		test_msg("Invalid key at slot 2\n");
+		ret = -EINVAL;
+		goto out;
+	}
+
+	item = btrfs_item_nr(2);
+	if (btrfs_item_size(eb, item) != strlen(split2)) {
+		test_msg("Invalid len in the second split\n");
+		ret = -EINVAL;
+		goto out;
+	}
+
+	read_extent_buffer(eb, buf, btrfs_item_ptr_offset(eb, 2),
+			   strlen(split2));
+	if (memcmp(buf, split2, strlen(split2))) {
+		test_msg("Data in the buffer doesn't match what it should "
+			 "in the last chunk\n");
+		ret = -EINVAL;
+		goto out;
+	}
+out:
+	btrfs_free_path(path);
+	kfree(root);
+	return ret;
+}
+
+int btrfs_test_extent_buffer_operations(void)
+{
+	test_msg("Running extent buffer operation tests");
+	return test_btrfs_split_item();
+}
diff --git a/fs/btrfs/tests/extent-io-tests.c b/fs/btrfs/tests/extent-io-tests.c
new file mode 100644
index 00000000000..7e99c2f98dd
--- /dev/null
+++ b/fs/btrfs/tests/extent-io-tests.c
@@ -0,0 +1,276 @@
+/*
+ * Copyright (C) 2013 Fusion IO.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/pagemap.h>
+#include <linux/sched.h>
+#include "btrfs-tests.h"
+#include "../extent_io.h"
+
+#define PROCESS_UNLOCK		(1 << 0)
+#define PROCESS_RELEASE		(1 << 1)
+#define PROCESS_TEST_LOCKED	(1 << 2)
+
+static noinline int process_page_range(struct inode *inode, u64 start, u64 end,
+				       unsigned long flags)
+{
+	int ret;
+	struct page *pages[16];
+	unsigned long index = start >> PAGE_CACHE_SHIFT;
+	unsigned long end_index = end >> PAGE_CACHE_SHIFT;
+	unsigned long nr_pages = end_index - index + 1;
+	int i;
+	int count = 0;
+	int loops = 0;
+
+	while (nr_pages > 0) {
+		ret = find_get_pages_contig(inode->i_mapping, index,
+				     min_t(unsigned long, nr_pages,
+				     ARRAY_SIZE(pages)), pages);
+		for (i = 0; i < ret; i++) {
+			if (flags & PROCESS_TEST_LOCKED &&
+			    !PageLocked(pages[i]))
+				count++;
+			if (flags & PROCESS_UNLOCK && PageLocked(pages[i]))
+				unlock_page(pages[i]);
+			page_cache_release(pages[i]);
+			if (flags & PROCESS_RELEASE)
+				page_cache_release(pages[i]);
+		}
+		nr_pages -= ret;
+		index += ret;
+		cond_resched();
+		loops++;
+		if (loops > 100000) {
+			printk(KERN_ERR "stuck in a loop, start %Lu, end %Lu, nr_pages %lu, ret %d\n", start, end, nr_pages, ret);
+			break;
+		}
+	}
+	return count;
+}
+
+static int test_find_delalloc(void)
+{
+	struct inode *inode;
+	struct extent_io_tree tmp;
+	struct page *page;
+	struct page *locked_page = NULL;
+	unsigned long index = 0;
+	u64 total_dirty = 256 * 1024 * 1024;
+	u64 max_bytes = 128 * 1024 * 1024;
+	u64 start, end, test_start;
+	u64 found;
+	int ret = -EINVAL;
+
+	inode = btrfs_new_test_inode();
+	if (!inode) {
+		test_msg("Failed to allocate test inode\n");
+		return -ENOMEM;
+	}
+
+	extent_io_tree_init(&tmp, &inode->i_data);
+
+	/*
+	 * First go through and create and mark all of our pages dirty, we pin
+	 * everything to make sure our pages don't get evicted and screw up our
+	 * test.
+	 */
+	for (index = 0; index < (total_dirty >> PAGE_CACHE_SHIFT); index++) {
+		page = find_or_create_page(inode->i_mapping, index, GFP_NOFS);
+		if (!page) {
+			test_msg("Failed to allocate test page\n");
+			ret = -ENOMEM;
+			goto out;
+		}
+		SetPageDirty(page);
+		if (index) {
+			unlock_page(page);
+		} else {
+			page_cache_get(page);
+			locked_page = page;
+		}
+	}
+
+	/* Test this scenario
+	 * |--- delalloc ---|
+	 * |---  search  ---|
+	 */
+	set_extent_delalloc(&tmp, 0, 4095, NULL, GFP_NOFS);
+	start = 0;
+	end = 0;
+	found = find_lock_delalloc_range(inode, &tmp, locked_page, &start,
+					 &end, max_bytes);
+	if (!found) {
+		test_msg("Should have found at least one delalloc\n");
+		goto out_bits;
+	}
+	if (start != 0 || end != 4095) {
+		test_msg("Expected start 0 end 4095, got start %Lu end %Lu\n",
+			 start, end);
+		goto out_bits;
+	}
+	unlock_extent(&tmp, start, end);
+	unlock_page(locked_page);
+	page_cache_release(locked_page);
+
+	/*
+	 * Test this scenario
+	 *
+	 * |--- delalloc ---|
+	 *           |--- search ---|
+	 */
+	test_start = 64 * 1024 * 1024;
+	locked_page = find_lock_page(inode->i_mapping,
+				     test_start >> PAGE_CACHE_SHIFT);
+	if (!locked_page) {
+		test_msg("Couldn't find the locked page\n");
+		goto out_bits;
+	}
+	set_extent_delalloc(&tmp, 4096, max_bytes - 1, NULL, GFP_NOFS);
+	start = test_start;
+	end = 0;
+	found = find_lock_delalloc_range(inode, &tmp, locked_page, &start,
+					 &end, max_bytes);
+	if (!found) {
+		test_msg("Couldn't find delalloc in our range\n");
+		goto out_bits;
+	}
+	if (start != test_start || end != max_bytes - 1) {
+		test_msg("Expected start %Lu end %Lu, got start %Lu, end "
+			 "%Lu\n", test_start, max_bytes - 1, start, end);
+		goto out_bits;
+	}
+	if (process_page_range(inode, start, end,
+			       PROCESS_TEST_LOCKED | PROCESS_UNLOCK)) {
+		test_msg("There were unlocked pages in the range\n");
+		goto out_bits;
+	}
+	unlock_extent(&tmp, start, end);
+	/* locked_page was unlocked above */
+	page_cache_release(locked_page);
+
+	/*
+	 * Test this scenario
+	 * |--- delalloc ---|
+	 *                    |--- search ---|
+	 */
+	test_start = max_bytes + 4096;
+	locked_page = find_lock_page(inode->i_mapping, test_start >>
+				     PAGE_CACHE_SHIFT);
+	if (!locked_page) {
+		test_msg("Could'nt find the locked page\n");
+		goto out_bits;
+	}
+	start = test_start;
+	end = 0;
+	found = find_lock_delalloc_range(inode, &tmp, locked_page, &start,
+					 &end, max_bytes);
+	if (found) {
+		test_msg("Found range when we shouldn't have\n");
+		goto out_bits;
+	}
+	if (end != (u64)-1) {
+		test_msg("Did not return the proper end offset\n");
+		goto out_bits;
+	}
+
+	/*
+	 * Test this scenario
+	 * [------- delalloc -------|
+	 * [max_bytes]|-- search--|
+	 *
+	 * We are re-using our test_start from above since it works out well.
+	 */
+	set_extent_delalloc(&tmp, max_bytes, total_dirty - 1, NULL, GFP_NOFS);
+	start = test_start;
+	end = 0;
+	found = find_lock_delalloc_range(inode, &tmp, locked_page, &start,
+					 &end, max_bytes);
+	if (!found) {
+		test_msg("Didn't find our range\n");
+		goto out_bits;
+	}
+	if (start != test_start || end != total_dirty - 1) {
+		test_msg("Expected start %Lu end %Lu, got start %Lu end %Lu\n",
+			 test_start, total_dirty - 1, start, end);
+		goto out_bits;
+	}
+	if (process_page_range(inode, start, end,
+			       PROCESS_TEST_LOCKED | PROCESS_UNLOCK)) {
+		test_msg("Pages in range were not all locked\n");
+		goto out_bits;
+	}
+	unlock_extent(&tmp, start, end);
+
+	/*
+	 * Now to test where we run into a page that is no longer dirty in the
+	 * range we want to find.
+	 */
+	page = find_get_page(inode->i_mapping, (max_bytes + (1 * 1024 * 1024))
+			     >> PAGE_CACHE_SHIFT);
+	if (!page) {
+		test_msg("Couldn't find our page\n");
+		goto out_bits;
+	}
+	ClearPageDirty(page);
+	page_cache_release(page);
+
+	/* We unlocked it in the previous test */
+	lock_page(locked_page);
+	start = test_start;
+	end = 0;
+	/*
+	 * Currently if we fail to find dirty pages in the delalloc range we
+	 * will adjust max_bytes down to PAGE_CACHE_SIZE and then re-search.  If
+	 * this changes at any point in the future we will need to fix this
+	 * tests expected behavior.
+	 */
+	found = find_lock_delalloc_range(inode, &tmp, locked_page, &start,
+					 &end, max_bytes);
+	if (!found) {
+		test_msg("Didn't find our range\n");
+		goto out_bits;
+	}
+	if (start != test_start && end != test_start + PAGE_CACHE_SIZE - 1) {
+		test_msg("Expected start %Lu end %Lu, got start %Lu end %Lu\n",
+			 test_start, test_start + PAGE_CACHE_SIZE - 1, start,
+			 end);
+		goto out_bits;
+	}
+	if (process_page_range(inode, start, end, PROCESS_TEST_LOCKED |
+			       PROCESS_UNLOCK)) {
+		test_msg("Pages in range were not all locked\n");
+		goto out_bits;
+	}
+	ret = 0;
+out_bits:
+	clear_extent_bits(&tmp, 0, total_dirty - 1,
+			  (unsigned long)-1, GFP_NOFS);
+out:
+	if (locked_page)
+		page_cache_release(locked_page);
+	process_page_range(inode, 0, total_dirty - 1,
+			   PROCESS_UNLOCK | PROCESS_RELEASE);
+	iput(inode);
+	return ret;
+}
+
+int btrfs_test_extent_io(void)
+{
+	test_msg("Running find delalloc tests\n");
+	return test_find_delalloc();
+}
diff --git a/fs/btrfs/tests/free-space-tests.c b/fs/btrfs/tests/free-space-tests.c
new file mode 100644
index 00000000000..c8d9ddf84c6
--- /dev/null
+++ b/fs/btrfs/tests/free-space-tests.c
@@ -0,0 +1,395 @@
+/*
+ * Copyright (C) 2013 Fusion IO.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/slab.h>
+#include "btrfs-tests.h"
+#include "../ctree.h"
+#include "../free-space-cache.h"
+
+#define BITS_PER_BITMAP		(PAGE_CACHE_SIZE * 8)
+static struct btrfs_block_group_cache *init_test_block_group(void)
+{
+	struct btrfs_block_group_cache *cache;
+
+	cache = kzalloc(sizeof(*cache), GFP_NOFS);
+	if (!cache)
+		return NULL;
+	cache->free_space_ctl = kzalloc(sizeof(*cache->free_space_ctl),
+					GFP_NOFS);
+	if (!cache->free_space_ctl) {
+		kfree(cache);
+		return NULL;
+	}
+
+	cache->key.objectid = 0;
+	cache->key.offset = 1024 * 1024 * 1024;
+	cache->key.type = BTRFS_BLOCK_GROUP_ITEM_KEY;
+	cache->sectorsize = 4096;
+
+	spin_lock_init(&cache->lock);
+	INIT_LIST_HEAD(&cache->list);
+	INIT_LIST_HEAD(&cache->cluster_list);
+	INIT_LIST_HEAD(&cache->new_bg_list);
+
+	btrfs_init_free_space_ctl(cache);
+
+	return cache;
+}
+
+/*
+ * This test just does basic sanity checking, making sure we can add an exten
+ * entry and remove space from either end and the middle, and make sure we can
+ * remove space that covers adjacent extent entries.
+ */
+static int test_extents(struct btrfs_block_group_cache *cache)
+{
+	int ret = 0;
+
+	test_msg("Running extent only tests\n");
+
+	/* First just make sure we can remove an entire entry */
+	ret = btrfs_add_free_space(cache, 0, 4 * 1024 * 1024);
+	if (ret) {
+		test_msg("Error adding initial extents %d\n", ret);
+		return ret;
+	}
+
+	ret = btrfs_remove_free_space(cache, 0, 4 * 1024 * 1024);
+	if (ret) {
+		test_msg("Error removing extent %d\n", ret);
+		return ret;
+	}
+
+	if (test_check_exists(cache, 0, 4 * 1024 * 1024)) {
+		test_msg("Full remove left some lingering space\n");
+		return -1;
+	}
+
+	/* Ok edge and middle cases now */
+	ret = btrfs_add_free_space(cache, 0, 4 * 1024 * 1024);
+	if (ret) {
+		test_msg("Error adding half extent %d\n", ret);
+		return ret;
+	}
+
+	ret = btrfs_remove_free_space(cache, 3 * 1024 * 1024, 1 * 1024 * 1024);
+	if (ret) {
+		test_msg("Error removing tail end %d\n", ret);
+		return ret;
+	}
+
+	ret = btrfs_remove_free_space(cache, 0, 1 * 1024 * 1024);
+	if (ret) {
+		test_msg("Error removing front end %d\n", ret);
+		return ret;
+	}
+
+	ret = btrfs_remove_free_space(cache, 2 * 1024 * 1024, 4096);
+	if (ret) {
+		test_msg("Error removing middle piece %d\n", ret);
+		return ret;
+	}
+
+	if (test_check_exists(cache, 0, 1 * 1024 * 1024)) {
+		test_msg("Still have space at the front\n");
+		return -1;
+	}
+
+	if (test_check_exists(cache, 2 * 1024 * 1024, 4096)) {
+		test_msg("Still have space in the middle\n");
+		return -1;
+	}
+
+	if (test_check_exists(cache, 3 * 1024 * 1024, 1 * 1024 * 1024)) {
+		test_msg("Still have space at the end\n");
+		return -1;
+	}
+
+	/* Cleanup */
+	__btrfs_remove_free_space_cache(cache->free_space_ctl);
+
+	return 0;
+}
+
+static int test_bitmaps(struct btrfs_block_group_cache *cache)
+{
+	u64 next_bitmap_offset;
+	int ret;
+
+	test_msg("Running bitmap only tests\n");
+
+	ret = test_add_free_space_entry(cache, 0, 4 * 1024 * 1024, 1);
+	if (ret) {
+		test_msg("Couldn't create a bitmap entry %d\n", ret);
+		return ret;
+	}
+
+	ret = btrfs_remove_free_space(cache, 0, 4 * 1024 * 1024);
+	if (ret) {
+		test_msg("Error removing bitmap full range %d\n", ret);
+		return ret;
+	}
+
+	if (test_check_exists(cache, 0, 4 * 1024 * 1024)) {
+		test_msg("Left some space in bitmap\n");
+		return -1;
+	}
+
+	ret = test_add_free_space_entry(cache, 0, 4 * 1024 * 1024, 1);
+	if (ret) {
+		test_msg("Couldn't add to our bitmap entry %d\n", ret);
+		return ret;
+	}
+
+	ret = btrfs_remove_free_space(cache, 1 * 1024 * 1024, 2 * 1024 * 1024);
+	if (ret) {
+		test_msg("Couldn't remove middle chunk %d\n", ret);
+		return ret;
+	}
+
+	/*
+	 * The first bitmap we have starts at offset 0 so the next one is just
+	 * at the end of the first bitmap.
+	 */
+	next_bitmap_offset = (u64)(BITS_PER_BITMAP * 4096);
+
+	/* Test a bit straddling two bitmaps */
+	ret = test_add_free_space_entry(cache, next_bitmap_offset -
+				   (2 * 1024 * 1024), 4 * 1024 * 1024, 1);
+	if (ret) {
+		test_msg("Couldn't add space that straddles two bitmaps %d\n",
+				ret);
+		return ret;
+	}
+
+	ret = btrfs_remove_free_space(cache, next_bitmap_offset -
+				      (1 * 1024 * 1024), 2 * 1024 * 1024);
+	if (ret) {
+		test_msg("Couldn't remove overlapping space %d\n", ret);
+		return ret;
+	}
+
+	if (test_check_exists(cache, next_bitmap_offset - (1 * 1024 * 1024),
+			 2 * 1024 * 1024)) {
+		test_msg("Left some space when removing overlapping\n");
+		return -1;
+	}
+
+	__btrfs_remove_free_space_cache(cache->free_space_ctl);
+
+	return 0;
+}
+
+/* This is the high grade jackassery */
+static int test_bitmaps_and_extents(struct btrfs_block_group_cache *cache)
+{
+	u64 bitmap_offset = (u64)(BITS_PER_BITMAP * 4096);
+	int ret;
+
+	test_msg("Running bitmap and extent tests\n");
+
+	/*
+	 * First let's do something simple, an extent at the same offset as the
+	 * bitmap, but the free space completely in the extent and then
+	 * completely in the bitmap.
+	 */
+	ret = test_add_free_space_entry(cache, 4 * 1024 * 1024, 1 * 1024 * 1024, 1);
+	if (ret) {
+		test_msg("Couldn't create bitmap entry %d\n", ret);
+		return ret;
+	}
+
+	ret = test_add_free_space_entry(cache, 0, 1 * 1024 * 1024, 0);
+	if (ret) {
+		test_msg("Couldn't add extent entry %d\n", ret);
+		return ret;
+	}
+
+	ret = btrfs_remove_free_space(cache, 0, 1 * 1024 * 1024);
+	if (ret) {
+		test_msg("Couldn't remove extent entry %d\n", ret);
+		return ret;
+	}
+
+	if (test_check_exists(cache, 0, 1 * 1024 * 1024)) {
+		test_msg("Left remnants after our remove\n");
+		return -1;
+	}
+
+	/* Now to add back the extent entry and remove from the bitmap */
+	ret = test_add_free_space_entry(cache, 0, 1 * 1024 * 1024, 0);
+	if (ret) {
+		test_msg("Couldn't re-add extent entry %d\n", ret);
+		return ret;
+	}
+
+	ret = btrfs_remove_free_space(cache, 4 * 1024 * 1024, 1 * 1024 * 1024);
+	if (ret) {
+		test_msg("Couldn't remove from bitmap %d\n", ret);
+		return ret;
+	}
+
+	if (test_check_exists(cache, 4 * 1024 * 1024, 1 * 1024 * 1024)) {
+		test_msg("Left remnants in the bitmap\n");
+		return -1;
+	}
+
+	/*
+	 * Ok so a little more evil, extent entry and bitmap at the same offset,
+	 * removing an overlapping chunk.
+	 */
+	ret = test_add_free_space_entry(cache, 1 * 1024 * 1024, 4 * 1024 * 1024, 1);
+	if (ret) {
+		test_msg("Couldn't add to a bitmap %d\n", ret);
+		return ret;
+	}
+
+	ret = btrfs_remove_free_space(cache, 512 * 1024, 3 * 1024 * 1024);
+	if (ret) {
+		test_msg("Couldn't remove overlapping space %d\n", ret);
+		return ret;
+	}
+
+	if (test_check_exists(cache, 512 * 1024, 3 * 1024 * 1024)) {
+		test_msg("Left over pieces after removing overlapping\n");
+		return -1;
+	}
+
+	__btrfs_remove_free_space_cache(cache->free_space_ctl);
+
+	/* Now with the extent entry offset into the bitmap */
+	ret = test_add_free_space_entry(cache, 4 * 1024 * 1024, 4 * 1024 * 1024, 1);
+	if (ret) {
+		test_msg("Couldn't add space to the bitmap %d\n", ret);
+		return ret;
+	}
+
+	ret = test_add_free_space_entry(cache, 2 * 1024 * 1024, 2 * 1024 * 1024, 0);
+	if (ret) {
+		test_msg("Couldn't add extent to the cache %d\n", ret);
+		return ret;
+	}
+
+	ret = btrfs_remove_free_space(cache, 3 * 1024 * 1024, 4 * 1024 * 1024);
+	if (ret) {
+		test_msg("Problem removing overlapping space %d\n", ret);
+		return ret;
+	}
+
+	if (test_check_exists(cache, 3 * 1024 * 1024, 4 * 1024 * 1024)) {
+		test_msg("Left something behind when removing space");
+		return -1;
+	}
+
+	/*
+	 * This has blown up in the past, the extent entry starts before the
+	 * bitmap entry, but we're trying to remove an offset that falls
+	 * completely within the bitmap range and is in both the extent entry
+	 * and the bitmap entry, looks like this
+	 *
+	 *   [ extent ]
+	 *      [ bitmap ]
+	 *        [ del ]
+	 */
+	__btrfs_remove_free_space_cache(cache->free_space_ctl);
+	ret = test_add_free_space_entry(cache, bitmap_offset + 4 * 1024 * 1024,
+				   4 * 1024 * 1024, 1);
+	if (ret) {
+		test_msg("Couldn't add bitmap %d\n", ret);
+		return ret;
+	}
+
+	ret = test_add_free_space_entry(cache, bitmap_offset - 1 * 1024 * 1024,
+				   5 * 1024 * 1024, 0);
+	if (ret) {
+		test_msg("Couldn't add extent entry %d\n", ret);
+		return ret;
+	}
+
+	ret = btrfs_remove_free_space(cache, bitmap_offset + 1 * 1024 * 1024,
+				      5 * 1024 * 1024);
+	if (ret) {
+		test_msg("Failed to free our space %d\n", ret);
+		return ret;
+	}
+
+	if (test_check_exists(cache, bitmap_offset + 1 * 1024 * 1024,
+			 5 * 1024 * 1024)) {
+		test_msg("Left stuff over\n");
+		return -1;
+	}
+
+	__btrfs_remove_free_space_cache(cache->free_space_ctl);
+
+	/*
+	 * This blew up before, we have part of the free space in a bitmap and
+	 * then the entirety of the rest of the space in an extent.  This used
+	 * to return -EAGAIN back from btrfs_remove_extent, make sure this
+	 * doesn't happen.
+	 */
+	ret = test_add_free_space_entry(cache, 1 * 1024 * 1024, 2 * 1024 * 1024, 1);
+	if (ret) {
+		test_msg("Couldn't add bitmap entry %d\n", ret);
+		return ret;
+	}
+
+	ret = test_add_free_space_entry(cache, 3 * 1024 * 1024, 1 * 1024 * 1024, 0);
+	if (ret) {
+		test_msg("Couldn't add extent entry %d\n", ret);
+		return ret;
+	}
+
+	ret = btrfs_remove_free_space(cache, 1 * 1024 * 1024, 3 * 1024 * 1024);
+	if (ret) {
+		test_msg("Error removing bitmap and extent overlapping %d\n", ret);
+		return ret;
+	}
+
+	__btrfs_remove_free_space_cache(cache->free_space_ctl);
+	return 0;
+}
+
+int btrfs_test_free_space_cache(void)
+{
+	struct btrfs_block_group_cache *cache;
+	int ret;
+
+	test_msg("Running btrfs free space cache tests\n");
+
+	cache = init_test_block_group();
+	if (!cache) {
+		test_msg("Couldn't run the tests\n");
+		return 0;
+	}
+
+	ret = test_extents(cache);
+	if (ret)
+		goto out;
+	ret = test_bitmaps(cache);
+	if (ret)
+		goto out;
+	ret = test_bitmaps_and_extents(cache);
+	if (ret)
+		goto out;
+out:
+	__btrfs_remove_free_space_cache(cache->free_space_ctl);
+	kfree(cache->free_space_ctl);
+	kfree(cache);
+	test_msg("Free space cache tests finished\n");
+	return ret;
+}
diff --git a/fs/btrfs/tests/inode-tests.c b/fs/btrfs/tests/inode-tests.c
new file mode 100644
index 00000000000..3ae0f5b8bb8
--- /dev/null
+++ b/fs/btrfs/tests/inode-tests.c
@@ -0,0 +1,928 @@
+/*
+ * Copyright (C) 2013 Fusion IO.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include "btrfs-tests.h"
+#include "../ctree.h"
+#include "../btrfs_inode.h"
+#include "../disk-io.h"
+#include "../extent_io.h"
+#include "../volumes.h"
+
+static void insert_extent(struct btrfs_root *root, u64 start, u64 len,
+			  u64 ram_bytes, u64 offset, u64 disk_bytenr,
+			  u64 disk_len, u32 type, u8 compression, int slot)
+{
+	struct btrfs_path path;
+	struct btrfs_file_extent_item *fi;
+	struct extent_buffer *leaf = root->node;
+	struct btrfs_key key;
+	u32 value_len = sizeof(struct btrfs_file_extent_item);
+
+	if (type == BTRFS_FILE_EXTENT_INLINE)
+		value_len += len;
+	memset(&path, 0, sizeof(path));
+
+	path.nodes[0] = leaf;
+	path.slots[0] = slot;
+
+	key.objectid = BTRFS_FIRST_FREE_OBJECTID;
+	key.type = BTRFS_EXTENT_DATA_KEY;
+	key.offset = start;
+
+	setup_items_for_insert(root, &path, &key, &value_len, value_len,
+			       value_len + sizeof(struct btrfs_item), 1);
+	fi = btrfs_item_ptr(leaf, slot, struct btrfs_file_extent_item);
+	btrfs_set_file_extent_generation(leaf, fi, 1);
+	btrfs_set_file_extent_type(leaf, fi, type);
+	btrfs_set_file_extent_disk_bytenr(leaf, fi, disk_bytenr);
+	btrfs_set_file_extent_disk_num_bytes(leaf, fi, disk_len);
+	btrfs_set_file_extent_offset(leaf, fi, offset);
+	btrfs_set_file_extent_num_bytes(leaf, fi, len);
+	btrfs_set_file_extent_ram_bytes(leaf, fi, ram_bytes);
+	btrfs_set_file_extent_compression(leaf, fi, compression);
+	btrfs_set_file_extent_encryption(leaf, fi, 0);
+	btrfs_set_file_extent_other_encoding(leaf, fi, 0);
+}
+
+static void insert_inode_item_key(struct btrfs_root *root)
+{
+	struct btrfs_path path;
+	struct extent_buffer *leaf = root->node;
+	struct btrfs_key key;
+	u32 value_len = 0;
+
+	memset(&path, 0, sizeof(path));
+
+	path.nodes[0] = leaf;
+	path.slots[0] = 0;
+
+	key.objectid = BTRFS_INODE_ITEM_KEY;
+	key.type = BTRFS_INODE_ITEM_KEY;
+	key.offset = 0;
+
+	setup_items_for_insert(root, &path, &key, &value_len, value_len,
+			       value_len + sizeof(struct btrfs_item), 1);
+}
+
+/*
+ * Build the most complicated map of extents the earth has ever seen.  We want
+ * this so we can test all of the corner cases of btrfs_get_extent.  Here is a
+ * diagram of how the extents will look though this may not be possible we still
+ * want to make sure everything acts normally (the last number is not inclusive)
+ *
+ * [0 - 5][5 -  6][6 - 10][10 - 4096][  4096 - 8192 ][8192 - 12288]
+ * [hole ][inline][ hole ][ regular ][regular1 split][    hole    ]
+ *
+ * [ 12288 - 20480][20480 - 24576][  24576 - 28672  ][28672 - 36864][36864 - 45056]
+ * [regular1 split][   prealloc1 ][prealloc1 written][   prealloc1 ][ compressed  ]
+ *
+ * [45056 - 49152][49152-53248][53248-61440][61440-65536][     65536+81920   ]
+ * [ compressed1 ][  regular  ][compressed1][  regular  ][ hole but no extent]
+ *
+ * [81920-86016]
+ * [  regular  ]
+ */
+static void setup_file_extents(struct btrfs_root *root)
+{
+	int slot = 0;
+	u64 disk_bytenr = 1 * 1024 * 1024;
+	u64 offset = 0;
+
+	/* First we want a hole */
+	insert_extent(root, offset, 5, 5, 0, 0, 0, BTRFS_FILE_EXTENT_REG, 0,
+		      slot);
+	slot++;
+	offset += 5;
+
+	/*
+	 * Now we want an inline extent, I don't think this is possible but hey
+	 * why not?  Also keep in mind if we have an inline extent it counts as
+	 * the whole first page.  If we were to expand it we would have to cow
+	 * and we wouldn't have an inline extent anymore.
+	 */
+	insert_extent(root, offset, 1, 1, 0, 0, 0, BTRFS_FILE_EXTENT_INLINE, 0,
+		      slot);
+	slot++;
+	offset = 4096;
+
+	/* Now another hole */
+	insert_extent(root, offset, 4, 4, 0, 0, 0, BTRFS_FILE_EXTENT_REG, 0,
+		      slot);
+	slot++;
+	offset += 4;
+
+	/* Now for a regular extent */
+	insert_extent(root, offset, 4095, 4095, 0, disk_bytenr, 4096,
+		      BTRFS_FILE_EXTENT_REG, 0, slot);
+	slot++;
+	disk_bytenr += 4096;
+	offset += 4095;
+
+	/*
+	 * Now for 3 extents that were split from a hole punch so we test
+	 * offsets properly.
+	 */
+	insert_extent(root, offset, 4096, 16384, 0, disk_bytenr, 16384,
+		      BTRFS_FILE_EXTENT_REG, 0, slot);
+	slot++;
+	offset += 4096;
+	insert_extent(root, offset, 4096, 4096, 0, 0, 0, BTRFS_FILE_EXTENT_REG,
+		      0, slot);
+	slot++;
+	offset += 4096;
+	insert_extent(root, offset, 8192, 16384, 8192, disk_bytenr, 16384,
+		      BTRFS_FILE_EXTENT_REG, 0, slot);
+	slot++;
+	offset += 8192;
+	disk_bytenr += 16384;
+
+	/* Now for a unwritten prealloc extent */
+	insert_extent(root, offset, 4096, 4096, 0, disk_bytenr, 4096,
+		      BTRFS_FILE_EXTENT_PREALLOC, 0, slot);
+	slot++;
+	offset += 4096;
+
+	/*
+	 * We want to jack up disk_bytenr a little more so the em stuff doesn't
+	 * merge our records.
+	 */
+	disk_bytenr += 8192;
+
+	/*
+	 * Now for a partially written prealloc extent, basically the same as
+	 * the hole punch example above.  Ram_bytes never changes when you mark
+	 * extents written btw.
+	 */
+	insert_extent(root, offset, 4096, 16384, 0, disk_bytenr, 16384,
+		      BTRFS_FILE_EXTENT_PREALLOC, 0, slot);
+	slot++;
+	offset += 4096;
+	insert_extent(root, offset, 4096, 16384, 4096, disk_bytenr, 16384,
+		      BTRFS_FILE_EXTENT_REG, 0, slot);
+	slot++;
+	offset += 4096;
+	insert_extent(root, offset, 8192, 16384, 8192, disk_bytenr, 16384,
+		      BTRFS_FILE_EXTENT_PREALLOC, 0, slot);
+	slot++;
+	offset += 8192;
+	disk_bytenr += 16384;
+
+	/* Now a normal compressed extent */
+	insert_extent(root, offset, 8192, 8192, 0, disk_bytenr, 4096,
+		      BTRFS_FILE_EXTENT_REG, BTRFS_COMPRESS_ZLIB, slot);
+	slot++;
+	offset += 8192;
+	/* No merges */
+	disk_bytenr += 8192;
+
+	/* Now a split compressed extent */
+	insert_extent(root, offset, 4096, 16384, 0, disk_bytenr, 4096,
+		      BTRFS_FILE_EXTENT_REG, BTRFS_COMPRESS_ZLIB, slot);
+	slot++;
+	offset += 4096;
+	insert_extent(root, offset, 4096, 4096, 0, disk_bytenr + 4096, 4096,
+		      BTRFS_FILE_EXTENT_REG, 0, slot);
+	slot++;
+	offset += 4096;
+	insert_extent(root, offset, 8192, 16384, 8192, disk_bytenr, 4096,
+		      BTRFS_FILE_EXTENT_REG, BTRFS_COMPRESS_ZLIB, slot);
+	slot++;
+	offset += 8192;
+	disk_bytenr += 8192;
+
+	/* Now extents that have a hole but no hole extent */
+	insert_extent(root, offset, 4096, 4096, 0, disk_bytenr, 4096,
+		      BTRFS_FILE_EXTENT_REG, 0, slot);
+	slot++;
+	offset += 16384;
+	disk_bytenr += 4096;
+	insert_extent(root, offset, 4096, 4096, 0, disk_bytenr, 4096,
+		      BTRFS_FILE_EXTENT_REG, 0, slot);
+}
+
+static unsigned long prealloc_only = 0;
+static unsigned long compressed_only = 0;
+static unsigned long vacancy_only = 0;
+
+static noinline int test_btrfs_get_extent(void)
+{
+	struct inode *inode = NULL;
+	struct btrfs_root *root = NULL;
+	struct extent_map *em = NULL;
+	u64 orig_start;
+	u64 disk_bytenr;
+	u64 offset;
+	int ret = -ENOMEM;
+
+	inode = btrfs_new_test_inode();
+	if (!inode) {
+		test_msg("Couldn't allocate inode\n");
+		return ret;
+	}
+
+	BTRFS_I(inode)->location.type = BTRFS_INODE_ITEM_KEY;
+	BTRFS_I(inode)->location.objectid = BTRFS_FIRST_FREE_OBJECTID;
+	BTRFS_I(inode)->location.offset = 0;
+
+	root = btrfs_alloc_dummy_root();
+	if (IS_ERR(root)) {
+		test_msg("Couldn't allocate root\n");
+		goto out;
+	}
+
+	/*
+	 * We do this since btrfs_get_extent wants to assign em->bdev to
+	 * root->fs_info->fs_devices->latest_bdev.
+	 */
+	root->fs_info = btrfs_alloc_dummy_fs_info();
+	if (!root->fs_info) {
+		test_msg("Couldn't allocate dummy fs info\n");
+		goto out;
+	}
+
+	root->node = alloc_dummy_extent_buffer(0, 4096);
+	if (!root->node) {
+		test_msg("Couldn't allocate dummy buffer\n");
+		goto out;
+	}
+
+	/*
+	 * We will just free a dummy node if it's ref count is 2 so we need an
+	 * extra ref so our searches don't accidently release our page.
+	 */
+	extent_buffer_get(root->node);
+	btrfs_set_header_nritems(root->node, 0);
+	btrfs_set_header_level(root->node, 0);
+	ret = -EINVAL;
+
+	/* First with no extents */
+	BTRFS_I(inode)->root = root;
+	em = btrfs_get_extent(inode, NULL, 0, 0, 4096, 0);
+	if (IS_ERR(em)) {
+		em = NULL;
+		test_msg("Got an error when we shouldn't have\n");
+		goto out;
+	}
+	if (em->block_start != EXTENT_MAP_HOLE) {
+		test_msg("Expected a hole, got %llu\n", em->block_start);
+		goto out;
+	}
+	if (!test_bit(EXTENT_FLAG_VACANCY, &em->flags)) {
+		test_msg("Vacancy flag wasn't set properly\n");
+		goto out;
+	}
+	free_extent_map(em);
+	btrfs_drop_extent_cache(inode, 0, (u64)-1, 0);
+
+	/*
+	 * All of the magic numbers are based on the mapping setup in
+	 * setup_file_extents, so if you change anything there you need to
+	 * update the comment and update the expected values below.
+	 */
+	setup_file_extents(root);
+
+	em = btrfs_get_extent(inode, NULL, 0, 0, (u64)-1, 0);
+	if (IS_ERR(em)) {
+		test_msg("Got an error when we shouldn't have\n");
+		goto out;
+	}
+	if (em->block_start != EXTENT_MAP_HOLE) {
+		test_msg("Expected a hole, got %llu\n", em->block_start);
+		goto out;
+	}
+	if (em->start != 0 || em->len != 5) {
+		test_msg("Unexpected extent wanted start 0 len 5, got start "
+			 "%llu len %llu\n", em->start, em->len);
+		goto out;
+	}
+	if (em->flags != 0) {
+		test_msg("Unexpected flags set, want 0 have %lu\n", em->flags);
+		goto out;
+	}
+	offset = em->start + em->len;
+	free_extent_map(em);
+
+	em = btrfs_get_extent(inode, NULL, 0, offset, 4096, 0);
+	if (IS_ERR(em)) {
+		test_msg("Got an error when we shouldn't have\n");
+		goto out;
+	}
+	if (em->block_start != EXTENT_MAP_INLINE) {
+		test_msg("Expected an inline, got %llu\n", em->block_start);
+		goto out;
+	}
+	if (em->start != offset || em->len != 4091) {
+		test_msg("Unexpected extent wanted start %llu len 1, got start "
+			 "%llu len %llu\n", offset, em->start, em->len);
+		goto out;
+	}
+	if (em->flags != 0) {
+		test_msg("Unexpected flags set, want 0 have %lu\n", em->flags);
+		goto out;
+	}
+	/*
+	 * We don't test anything else for inline since it doesn't get set
+	 * unless we have a page for it to write into.  Maybe we should change
+	 * this?
+	 */
+	offset = em->start + em->len;
+	free_extent_map(em);
+
+	em = btrfs_get_extent(inode, NULL, 0, offset, 4096, 0);
+	if (IS_ERR(em)) {
+		test_msg("Got an error when we shouldn't have\n");
+		goto out;
+	}
+	if (em->block_start != EXTENT_MAP_HOLE) {
+		test_msg("Expected a hole, got %llu\n", em->block_start);
+		goto out;
+	}
+	if (em->start != offset || em->len != 4) {
+		test_msg("Unexpected extent wanted start %llu len 4, got start "
+			 "%llu len %llu\n", offset, em->start, em->len);
+		goto out;
+	}
+	if (em->flags != 0) {
+		test_msg("Unexpected flags set, want 0 have %lu\n", em->flags);
+		goto out;
+	}
+	offset = em->start + em->len;
+	free_extent_map(em);
+
+	/* Regular extent */
+	em = btrfs_get_extent(inode, NULL, 0, offset, 4096, 0);
+	if (IS_ERR(em)) {
+		test_msg("Got an error when we shouldn't have\n");
+		goto out;
+	}
+	if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
+		test_msg("Expected a real extent, got %llu\n", em->block_start);
+		goto out;
+	}
+	if (em->start != offset || em->len != 4095) {
+		test_msg("Unexpected extent wanted start %llu len 4095, got "
+			 "start %llu len %llu\n", offset, em->start, em->len);
+		goto out;
+	}
+	if (em->flags != 0) {
+		test_msg("Unexpected flags set, want 0 have %lu\n", em->flags);
+		goto out;
+	}
+	if (em->orig_start != em->start) {
+		test_msg("Wrong orig offset, want %llu, have %llu\n", em->start,
+			 em->orig_start);
+		goto out;
+	}
+	offset = em->start + em->len;
+	free_extent_map(em);
+
+	/* The next 3 are split extents */
+	em = btrfs_get_extent(inode, NULL, 0, offset, 4096, 0);
+	if (IS_ERR(em)) {
+		test_msg("Got an error when we shouldn't have\n");
+		goto out;
+	}
+	if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
+		test_msg("Expected a real extent, got %llu\n", em->block_start);
+		goto out;
+	}
+	if (em->start != offset || em->len != 4096) {
+		test_msg("Unexpected extent wanted start %llu len 4096, got "
+			 "start %llu len %llu\n", offset, em->start, em->len);
+		goto out;
+	}
+	if (em->flags != 0) {
+		test_msg("Unexpected flags set, want 0 have %lu\n", em->flags);
+		goto out;
+	}
+	if (em->orig_start != em->start) {
+		test_msg("Wrong orig offset, want %llu, have %llu\n", em->start,
+			 em->orig_start);
+		goto out;
+	}
+	disk_bytenr = em->block_start;
+	orig_start = em->start;
+	offset = em->start + em->len;
+	free_extent_map(em);
+
+	em = btrfs_get_extent(inode, NULL, 0, offset, 4096, 0);
+	if (IS_ERR(em)) {
+		test_msg("Got an error when we shouldn't have\n");
+		goto out;
+	}
+	if (em->block_start != EXTENT_MAP_HOLE) {
+		test_msg("Expected a hole, got %llu\n", em->block_start);
+		goto out;
+	}
+	if (em->start != offset || em->len != 4096) {
+		test_msg("Unexpected extent wanted start %llu len 4096, got "
+			 "start %llu len %llu\n", offset, em->start, em->len);
+		goto out;
+	}
+	if (em->flags != 0) {
+		test_msg("Unexpected flags set, want 0 have %lu\n", em->flags);
+		goto out;
+	}
+	offset = em->start + em->len;
+	free_extent_map(em);
+
+	em = btrfs_get_extent(inode, NULL, 0, offset, 4096, 0);
+	if (IS_ERR(em)) {
+		test_msg("Got an error when we shouldn't have\n");
+		goto out;
+	}
+	if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
+		test_msg("Expected a real extent, got %llu\n", em->block_start);
+		goto out;
+	}
+	if (em->start != offset || em->len != 8192) {
+		test_msg("Unexpected extent wanted start %llu len 8192, got "
+			 "start %llu len %llu\n", offset, em->start, em->len);
+		goto out;
+	}
+	if (em->flags != 0) {
+		test_msg("Unexpected flags set, want 0 have %lu\n", em->flags);
+		goto out;
+	}
+	if (em->orig_start != orig_start) {
+		test_msg("Wrong orig offset, want %llu, have %llu\n",
+			 orig_start, em->orig_start);
+		goto out;
+	}
+	disk_bytenr += (em->start - orig_start);
+	if (em->block_start != disk_bytenr) {
+		test_msg("Wrong block start, want %llu, have %llu\n",
+			 disk_bytenr, em->block_start);
+		goto out;
+	}
+	offset = em->start + em->len;
+	free_extent_map(em);
+
+	/* Prealloc extent */
+	em = btrfs_get_extent(inode, NULL, 0, offset, 4096, 0);
+	if (IS_ERR(em)) {
+		test_msg("Got an error when we shouldn't have\n");
+		goto out;
+	}
+	if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
+		test_msg("Expected a real extent, got %llu\n", em->block_start);
+		goto out;
+	}
+	if (em->start != offset || em->len != 4096) {
+		test_msg("Unexpected extent wanted start %llu len 4096, got "
+			 "start %llu len %llu\n", offset, em->start, em->len);
+		goto out;
+	}
+	if (em->flags != prealloc_only) {
+		test_msg("Unexpected flags set, want %lu have %lu\n",
+			 prealloc_only, em->flags);
+		goto out;
+	}
+	if (em->orig_start != em->start) {
+		test_msg("Wrong orig offset, want %llu, have %llu\n", em->start,
+			 em->orig_start);
+		goto out;
+	}
+	offset = em->start + em->len;
+	free_extent_map(em);
+
+	/* The next 3 are a half written prealloc extent */
+	em = btrfs_get_extent(inode, NULL, 0, offset, 4096, 0);
+	if (IS_ERR(em)) {
+		test_msg("Got an error when we shouldn't have\n");
+		goto out;
+	}
+	if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
+		test_msg("Expected a real extent, got %llu\n", em->block_start);
+		goto out;
+	}
+	if (em->start != offset || em->len != 4096) {
+		test_msg("Unexpected extent wanted start %llu len 4096, got "
+			 "start %llu len %llu\n", offset, em->start, em->len);
+		goto out;
+	}
+	if (em->flags != prealloc_only) {
+		test_msg("Unexpected flags set, want %lu have %lu\n",
+			 prealloc_only, em->flags);
+		goto out;
+	}
+	if (em->orig_start != em->start) {
+		test_msg("Wrong orig offset, want %llu, have %llu\n", em->start,
+			 em->orig_start);
+		goto out;
+	}
+	disk_bytenr = em->block_start;
+	orig_start = em->start;
+	offset = em->start + em->len;
+	free_extent_map(em);
+
+	em = btrfs_get_extent(inode, NULL, 0, offset, 4096, 0);
+	if (IS_ERR(em)) {
+		test_msg("Got an error when we shouldn't have\n");
+		goto out;
+	}
+	if (em->block_start >= EXTENT_MAP_HOLE) {
+		test_msg("Expected a real extent, got %llu\n", em->block_start);
+		goto out;
+	}
+	if (em->start != offset || em->len != 4096) {
+		test_msg("Unexpected extent wanted start %llu len 4096, got "
+			 "start %llu len %llu\n", offset, em->start, em->len);
+		goto out;
+	}
+	if (em->flags != 0) {
+		test_msg("Unexpected flags set, want 0 have %lu\n", em->flags);
+		goto out;
+	}
+	if (em->orig_start != orig_start) {
+		test_msg("Unexpected orig offset, wanted %llu, have %llu\n",
+			 orig_start, em->orig_start);
+		goto out;
+	}
+	if (em->block_start != (disk_bytenr + (em->start - em->orig_start))) {
+		test_msg("Unexpected block start, wanted %llu, have %llu\n",
+			 disk_bytenr + (em->start - em->orig_start),
+			 em->block_start);
+		goto out;
+	}
+	offset = em->start + em->len;
+	free_extent_map(em);
+
+	em = btrfs_get_extent(inode, NULL, 0, offset, 4096, 0);
+	if (IS_ERR(em)) {
+		test_msg("Got an error when we shouldn't have\n");
+		goto out;
+	}
+	if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
+		test_msg("Expected a real extent, got %llu\n", em->block_start);
+		goto out;
+	}
+	if (em->start != offset || em->len != 8192) {
+		test_msg("Unexpected extent wanted start %llu len 8192, got "
+			 "start %llu len %llu\n", offset, em->start, em->len);
+		goto out;
+	}
+	if (em->flags != prealloc_only) {
+		test_msg("Unexpected flags set, want %lu have %lu\n",
+			 prealloc_only, em->flags);
+		goto out;
+	}
+	if (em->orig_start != orig_start) {
+		test_msg("Wrong orig offset, want %llu, have %llu\n", orig_start,
+			 em->orig_start);
+		goto out;
+	}
+	if (em->block_start != (disk_bytenr + (em->start - em->orig_start))) {
+		test_msg("Unexpected block start, wanted %llu, have %llu\n",
+			 disk_bytenr + (em->start - em->orig_start),
+			 em->block_start);
+		goto out;
+	}
+	offset = em->start + em->len;
+	free_extent_map(em);
+
+	/* Now for the compressed extent */
+	em = btrfs_get_extent(inode, NULL, 0, offset, 4096, 0);
+	if (IS_ERR(em)) {
+		test_msg("Got an error when we shouldn't have\n");
+		goto out;
+	}
+	if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
+		test_msg("Expected a real extent, got %llu\n", em->block_start);
+		goto out;
+	}
+	if (em->start != offset || em->len != 8192) {
+		test_msg("Unexpected extent wanted start %llu len 8192, got "
+			 "start %llu len %llu\n", offset, em->start, em->len);
+		goto out;
+	}
+	if (em->flags != compressed_only) {
+		test_msg("Unexpected flags set, want %lu have %lu\n",
+			 compressed_only, em->flags);
+		goto out;
+	}
+	if (em->orig_start != em->start) {
+		test_msg("Wrong orig offset, want %llu, have %llu\n",
+			 em->start, em->orig_start);
+		goto out;
+	}
+	if (em->compress_type != BTRFS_COMPRESS_ZLIB) {
+		test_msg("Unexpected compress type, wanted %d, got %d\n",
+			 BTRFS_COMPRESS_ZLIB, em->compress_type);
+		goto out;
+	}
+	offset = em->start + em->len;
+	free_extent_map(em);
+
+	/* Split compressed extent */
+	em = btrfs_get_extent(inode, NULL, 0, offset, 4096, 0);
+	if (IS_ERR(em)) {
+		test_msg("Got an error when we shouldn't have\n");
+		goto out;
+	}
+	if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
+		test_msg("Expected a real extent, got %llu\n", em->block_start);
+		goto out;
+	}
+	if (em->start != offset || em->len != 4096) {
+		test_msg("Unexpected extent wanted start %llu len 4096, got "
+			 "start %llu len %llu\n", offset, em->start, em->len);
+		goto out;
+	}
+	if (em->flags != compressed_only) {
+		test_msg("Unexpected flags set, want %lu have %lu\n",
+			 compressed_only, em->flags);
+		goto out;
+	}
+	if (em->orig_start != em->start) {
+		test_msg("Wrong orig offset, want %llu, have %llu\n",
+			 em->start, em->orig_start);
+		goto out;
+	}
+	if (em->compress_type != BTRFS_COMPRESS_ZLIB) {
+		test_msg("Unexpected compress type, wanted %d, got %d\n",
+			 BTRFS_COMPRESS_ZLIB, em->compress_type);
+		goto out;
+	}
+	disk_bytenr = em->block_start;
+	orig_start = em->start;
+	offset = em->start + em->len;
+	free_extent_map(em);
+
+	em = btrfs_get_extent(inode, NULL, 0, offset, 4096, 0);
+	if (IS_ERR(em)) {
+		test_msg("Got an error when we shouldn't have\n");
+		goto out;
+	}
+	if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
+		test_msg("Expected a real extent, got %llu\n", em->block_start);
+		goto out;
+	}
+	if (em->start != offset || em->len != 4096) {
+		test_msg("Unexpected extent wanted start %llu len 4096, got "
+			 "start %llu len %llu\n", offset, em->start, em->len);
+		goto out;
+	}
+	if (em->flags != 0) {
+		test_msg("Unexpected flags set, want 0 have %lu\n", em->flags);
+		goto out;
+	}
+	if (em->orig_start != em->start) {
+		test_msg("Wrong orig offset, want %llu, have %llu\n", em->start,
+			 em->orig_start);
+		goto out;
+	}
+	offset = em->start + em->len;
+	free_extent_map(em);
+
+	em = btrfs_get_extent(inode, NULL, 0, offset, 4096, 0);
+	if (IS_ERR(em)) {
+		test_msg("Got an error when we shouldn't have\n");
+		goto out;
+	}
+	if (em->block_start != disk_bytenr) {
+		test_msg("Block start does not match, want %llu got %llu\n",
+			 disk_bytenr, em->block_start);
+		goto out;
+	}
+	if (em->start != offset || em->len != 8192) {
+		test_msg("Unexpected extent wanted start %llu len 8192, got "
+			 "start %llu len %llu\n", offset, em->start, em->len);
+		goto out;
+	}
+	if (em->flags != compressed_only) {
+		test_msg("Unexpected flags set, want %lu have %lu\n",
+			 compressed_only, em->flags);
+		goto out;
+	}
+	if (em->orig_start != orig_start) {
+		test_msg("Wrong orig offset, want %llu, have %llu\n",
+			 em->start, orig_start);
+		goto out;
+	}
+	if (em->compress_type != BTRFS_COMPRESS_ZLIB) {
+		test_msg("Unexpected compress type, wanted %d, got %d\n",
+			 BTRFS_COMPRESS_ZLIB, em->compress_type);
+		goto out;
+	}
+	offset = em->start + em->len;
+	free_extent_map(em);
+
+	/* A hole between regular extents but no hole extent */
+	em = btrfs_get_extent(inode, NULL, 0, offset + 6, 4096, 0);
+	if (IS_ERR(em)) {
+		test_msg("Got an error when we shouldn't have\n");
+		goto out;
+	}
+	if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
+		test_msg("Expected a real extent, got %llu\n", em->block_start);
+		goto out;
+	}
+	if (em->start != offset || em->len != 4096) {
+		test_msg("Unexpected extent wanted start %llu len 4096, got "
+			 "start %llu len %llu\n", offset, em->start, em->len);
+		goto out;
+	}
+	if (em->flags != 0) {
+		test_msg("Unexpected flags set, want 0 have %lu\n", em->flags);
+		goto out;
+	}
+	if (em->orig_start != em->start) {
+		test_msg("Wrong orig offset, want %llu, have %llu\n", em->start,
+			 em->orig_start);
+		goto out;
+	}
+	offset = em->start + em->len;
+	free_extent_map(em);
+
+	em = btrfs_get_extent(inode, NULL, 0, offset, 4096 * 1024, 0);
+	if (IS_ERR(em)) {
+		test_msg("Got an error when we shouldn't have\n");
+		goto out;
+	}
+	if (em->block_start != EXTENT_MAP_HOLE) {
+		test_msg("Expected a hole extent, got %llu\n", em->block_start);
+		goto out;
+	}
+	/*
+	 * Currently we just return a length that we requested rather than the
+	 * length of the actual hole, if this changes we'll have to change this
+	 * test.
+	 */
+	if (em->start != offset || em->len != 12288) {
+		test_msg("Unexpected extent wanted start %llu len 12288, got "
+			 "start %llu len %llu\n", offset, em->start, em->len);
+		goto out;
+	}
+	if (em->flags != vacancy_only) {
+		test_msg("Unexpected flags set, want %lu have %lu\n",
+			 vacancy_only, em->flags);
+		goto out;
+	}
+	if (em->orig_start != em->start) {
+		test_msg("Wrong orig offset, want %llu, have %llu\n", em->start,
+			 em->orig_start);
+		goto out;
+	}
+	offset = em->start + em->len;
+	free_extent_map(em);
+
+	em = btrfs_get_extent(inode, NULL, 0, offset, 4096, 0);
+	if (IS_ERR(em)) {
+		test_msg("Got an error when we shouldn't have\n");
+		goto out;
+	}
+	if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
+		test_msg("Expected a real extent, got %llu\n", em->block_start);
+		goto out;
+	}
+	if (em->start != offset || em->len != 4096) {
+		test_msg("Unexpected extent wanted start %llu len 4096, got "
+			 "start %llu len %llu\n", offset, em->start, em->len);
+		goto out;
+	}
+	if (em->flags != 0) {
+		test_msg("Unexpected flags set, want 0 have %lu\n", em->flags);
+		goto out;
+	}
+	if (em->orig_start != em->start) {
+		test_msg("Wrong orig offset, want %llu, have %llu\n", em->start,
+			 em->orig_start);
+		goto out;
+	}
+	ret = 0;
+out:
+	if (!IS_ERR(em))
+		free_extent_map(em);
+	iput(inode);
+	btrfs_free_dummy_root(root);
+	return ret;
+}
+
+static int test_hole_first(void)
+{
+	struct inode *inode = NULL;
+	struct btrfs_root *root = NULL;
+	struct extent_map *em = NULL;
+	int ret = -ENOMEM;
+
+	inode = btrfs_new_test_inode();
+	if (!inode) {
+		test_msg("Couldn't allocate inode\n");
+		return ret;
+	}
+
+	BTRFS_I(inode)->location.type = BTRFS_INODE_ITEM_KEY;
+	BTRFS_I(inode)->location.objectid = BTRFS_FIRST_FREE_OBJECTID;
+	BTRFS_I(inode)->location.offset = 0;
+
+	root = btrfs_alloc_dummy_root();
+	if (IS_ERR(root)) {
+		test_msg("Couldn't allocate root\n");
+		goto out;
+	}
+
+	root->fs_info = btrfs_alloc_dummy_fs_info();
+	if (!root->fs_info) {
+		test_msg("Couldn't allocate dummy fs info\n");
+		goto out;
+	}
+
+	root->node = alloc_dummy_extent_buffer(0, 4096);
+	if (!root->node) {
+		test_msg("Couldn't allocate dummy buffer\n");
+		goto out;
+	}
+
+	extent_buffer_get(root->node);
+	btrfs_set_header_nritems(root->node, 0);
+	btrfs_set_header_level(root->node, 0);
+	BTRFS_I(inode)->root = root;
+	ret = -EINVAL;
+
+	/*
+	 * Need a blank inode item here just so we don't confuse
+	 * btrfs_get_extent.
+	 */
+	insert_inode_item_key(root);
+	insert_extent(root, 4096, 4096, 4096, 0, 4096, 4096,
+		      BTRFS_FILE_EXTENT_REG, 0, 1);
+	em = btrfs_get_extent(inode, NULL, 0, 0, 8192, 0);
+	if (IS_ERR(em)) {
+		test_msg("Got an error when we shouldn't have\n");
+		goto out;
+	}
+	if (em->block_start != EXTENT_MAP_HOLE) {
+		test_msg("Expected a hole, got %llu\n", em->block_start);
+		goto out;
+	}
+	if (em->start != 0 || em->len != 4096) {
+		test_msg("Unexpected extent wanted start 0 len 4096, got start "
+			 "%llu len %llu\n", em->start, em->len);
+		goto out;
+	}
+	if (em->flags != vacancy_only) {
+		test_msg("Wrong flags, wanted %lu, have %lu\n", vacancy_only,
+			 em->flags);
+		goto out;
+	}
+	free_extent_map(em);
+
+	em = btrfs_get_extent(inode, NULL, 0, 4096, 8192, 0);
+	if (IS_ERR(em)) {
+		test_msg("Got an error when we shouldn't have\n");
+		goto out;
+	}
+	if (em->block_start != 4096) {
+		test_msg("Expected a real extent, got %llu\n", em->block_start);
+		goto out;
+	}
+	if (em->start != 4096 || em->len != 4096) {
+		test_msg("Unexpected extent wanted start 4096 len 4096, got "
+			 "start %llu len %llu\n", em->start, em->len);
+		goto out;
+	}
+	if (em->flags != 0) {
+		test_msg("Unexpected flags set, wanted 0 got %lu\n",
+			 em->flags);
+		goto out;
+	}
+	ret = 0;
+out:
+	if (!IS_ERR(em))
+		free_extent_map(em);
+	iput(inode);
+	btrfs_free_dummy_root(root);
+	return ret;
+}
+
+int btrfs_test_inodes(void)
+{
+	int ret;
+
+	set_bit(EXTENT_FLAG_COMPRESSED, &compressed_only);
+	set_bit(EXTENT_FLAG_VACANCY, &vacancy_only);
+	set_bit(EXTENT_FLAG_PREALLOC, &prealloc_only);
+
+	test_msg("Running btrfs_get_extent tests\n");
+	ret = test_btrfs_get_extent();
+	if (ret)
+		return ret;
+	test_msg("Running hole first btrfs_get_extent test\n");
+	return test_hole_first();
+}
diff --git a/fs/btrfs/tests/qgroup-tests.c b/fs/btrfs/tests/qgroup-tests.c
new file mode 100644
index 00000000000..ec3dcb20235
--- /dev/null
+++ b/fs/btrfs/tests/qgroup-tests.c
@@ -0,0 +1,470 @@
+/*
+ * Copyright (C) 2013 Facebook.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include "btrfs-tests.h"
+#include "../ctree.h"
+#include "../transaction.h"
+#include "../disk-io.h"
+#include "../qgroup.h"
+
+static void init_dummy_trans(struct btrfs_trans_handle *trans)
+{
+	memset(trans, 0, sizeof(*trans));
+	trans->transid = 1;
+	INIT_LIST_HEAD(&trans->qgroup_ref_list);
+	trans->type = __TRANS_DUMMY;
+}
+
+static int insert_normal_tree_ref(struct btrfs_root *root, u64 bytenr,
+				  u64 num_bytes, u64 parent, u64 root_objectid)
+{
+	struct btrfs_trans_handle trans;
+	struct btrfs_extent_item *item;
+	struct btrfs_extent_inline_ref *iref;
+	struct btrfs_tree_block_info *block_info;
+	struct btrfs_path *path;
+	struct extent_buffer *leaf;
+	struct btrfs_key ins;
+	u32 size = sizeof(*item) + sizeof(*iref) + sizeof(*block_info);
+	int ret;
+
+	init_dummy_trans(&trans);
+
+	ins.objectid = bytenr;
+	ins.type = BTRFS_EXTENT_ITEM_KEY;
+	ins.offset = num_bytes;
+
+	path = btrfs_alloc_path();
+	if (!path) {
+		test_msg("Couldn't allocate path\n");
+		return -ENOMEM;
+	}
+
+	path->leave_spinning = 1;
+	ret = btrfs_insert_empty_item(&trans, root, path, &ins, size);
+	if (ret) {
+		test_msg("Couldn't insert ref %d\n", ret);
+		btrfs_free_path(path);
+		return ret;
+	}
+
+	leaf = path->nodes[0];
+	item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
+	btrfs_set_extent_refs(leaf, item, 1);
+	btrfs_set_extent_generation(leaf, item, 1);
+	btrfs_set_extent_flags(leaf, item, BTRFS_EXTENT_FLAG_TREE_BLOCK);
+	block_info = (struct btrfs_tree_block_info *)(item + 1);
+	btrfs_set_tree_block_level(leaf, block_info, 1);
+	iref = (struct btrfs_extent_inline_ref *)(block_info + 1);
+	if (parent > 0) {
+		btrfs_set_extent_inline_ref_type(leaf, iref,
+						 BTRFS_SHARED_BLOCK_REF_KEY);
+		btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
+	} else {
+		btrfs_set_extent_inline_ref_type(leaf, iref, BTRFS_TREE_BLOCK_REF_KEY);
+		btrfs_set_extent_inline_ref_offset(leaf, iref, root_objectid);
+	}
+	btrfs_free_path(path);
+	return 0;
+}
+
+static int add_tree_ref(struct btrfs_root *root, u64 bytenr, u64 num_bytes,
+			u64 parent, u64 root_objectid)
+{
+	struct btrfs_trans_handle trans;
+	struct btrfs_extent_item *item;
+	struct btrfs_path *path;
+	struct btrfs_key key;
+	u64 refs;
+	int ret;
+
+	init_dummy_trans(&trans);
+
+	key.objectid = bytenr;
+	key.type = BTRFS_EXTENT_ITEM_KEY;
+	key.offset = num_bytes;
+
+	path = btrfs_alloc_path();
+	if (!path) {
+		test_msg("Couldn't allocate path\n");
+		return -ENOMEM;
+	}
+
+	path->leave_spinning = 1;
+	ret = btrfs_search_slot(&trans, root, &key, path, 0, 1);
+	if (ret) {
+		test_msg("Couldn't find extent ref\n");
+		btrfs_free_path(path);
+		return ret;
+	}
+
+	item = btrfs_item_ptr(path->nodes[0], path->slots[0],
+			      struct btrfs_extent_item);
+	refs = btrfs_extent_refs(path->nodes[0], item);
+	btrfs_set_extent_refs(path->nodes[0], item, refs + 1);
+	btrfs_release_path(path);
+
+	key.objectid = bytenr;
+	if (parent) {
+		key.type = BTRFS_SHARED_BLOCK_REF_KEY;
+		key.offset = parent;
+	} else {
+		key.type = BTRFS_TREE_BLOCK_REF_KEY;
+		key.offset = root_objectid;
+	}
+
+	ret = btrfs_insert_empty_item(&trans, root, path, &key, 0);
+	if (ret)
+		test_msg("Failed to insert backref\n");
+	btrfs_free_path(path);
+	return ret;
+}
+
+static int remove_extent_item(struct btrfs_root *root, u64 bytenr,
+			      u64 num_bytes)
+{
+	struct btrfs_trans_handle trans;
+	struct btrfs_key key;
+	struct btrfs_path *path;
+	int ret;
+
+	init_dummy_trans(&trans);
+
+	key.objectid = bytenr;
+	key.type = BTRFS_EXTENT_ITEM_KEY;
+	key.offset = num_bytes;
+
+	path = btrfs_alloc_path();
+	if (!path) {
+		test_msg("Couldn't allocate path\n");
+		return -ENOMEM;
+	}
+	path->leave_spinning = 1;
+
+	ret = btrfs_search_slot(&trans, root, &key, path, -1, 1);
+	if (ret) {
+		test_msg("Didn't find our key %d\n", ret);
+		btrfs_free_path(path);
+		return ret;
+	}
+	btrfs_del_item(&trans, root, path);
+	btrfs_free_path(path);
+	return 0;
+}
+
+static int remove_extent_ref(struct btrfs_root *root, u64 bytenr,
+			     u64 num_bytes, u64 parent, u64 root_objectid)
+{
+	struct btrfs_trans_handle trans;
+	struct btrfs_extent_item *item;
+	struct btrfs_path *path;
+	struct btrfs_key key;
+	u64 refs;
+	int ret;
+
+	init_dummy_trans(&trans);
+
+	key.objectid = bytenr;
+	key.type = BTRFS_EXTENT_ITEM_KEY;
+	key.offset = num_bytes;
+
+	path = btrfs_alloc_path();
+	if (!path) {
+		test_msg("Couldn't allocate path\n");
+		return -ENOMEM;
+	}
+
+	path->leave_spinning = 1;
+	ret = btrfs_search_slot(&trans, root, &key, path, 0, 1);
+	if (ret) {
+		test_msg("Couldn't find extent ref\n");
+		btrfs_free_path(path);
+		return ret;
+	}
+
+	item = btrfs_item_ptr(path->nodes[0], path->slots[0],
+			      struct btrfs_extent_item);
+	refs = btrfs_extent_refs(path->nodes[0], item);
+	btrfs_set_extent_refs(path->nodes[0], item, refs - 1);
+	btrfs_release_path(path);
+
+	key.objectid = bytenr;
+	if (parent) {
+		key.type = BTRFS_SHARED_BLOCK_REF_KEY;
+		key.offset = parent;
+	} else {
+		key.type = BTRFS_TREE_BLOCK_REF_KEY;
+		key.offset = root_objectid;
+	}
+
+	ret = btrfs_search_slot(&trans, root, &key, path, -1, 1);
+	if (ret) {
+		test_msg("Couldn't find backref %d\n", ret);
+		btrfs_free_path(path);
+		return ret;
+	}
+	btrfs_del_item(&trans, root, path);
+	btrfs_free_path(path);
+	return ret;
+}
+
+static int test_no_shared_qgroup(struct btrfs_root *root)
+{
+	struct btrfs_trans_handle trans;
+	struct btrfs_fs_info *fs_info = root->fs_info;
+	int ret;
+
+	init_dummy_trans(&trans);
+
+	test_msg("Qgroup basic add\n");
+	ret = btrfs_create_qgroup(NULL, fs_info, 5, NULL);
+	if (ret) {
+		test_msg("Couldn't create a qgroup %d\n", ret);
+		return ret;
+	}
+
+	ret = btrfs_qgroup_record_ref(&trans, fs_info, 5, 4096, 4096,
+				      BTRFS_QGROUP_OPER_ADD_EXCL, 0);
+	if (ret) {
+		test_msg("Couldn't add space to a qgroup %d\n", ret);
+		return ret;
+	}
+
+	ret = insert_normal_tree_ref(root, 4096, 4096, 0, 5);
+	if (ret)
+		return ret;
+
+	ret = btrfs_delayed_qgroup_accounting(&trans, fs_info);
+	if (ret) {
+		test_msg("Delayed qgroup accounting failed %d\n", ret);
+		return ret;
+	}
+
+	if (btrfs_verify_qgroup_counts(fs_info, 5, 4096, 4096)) {
+		test_msg("Qgroup counts didn't match expected values\n");
+		return -EINVAL;
+	}
+
+	ret = remove_extent_item(root, 4096, 4096);
+	if (ret)
+		return -EINVAL;
+
+	ret = btrfs_qgroup_record_ref(&trans, fs_info, 5, 4096, 4096,
+				      BTRFS_QGROUP_OPER_SUB_EXCL, 0);
+	if (ret) {
+		test_msg("Couldn't remove space from the qgroup %d\n", ret);
+		return -EINVAL;
+	}
+
+	ret = btrfs_delayed_qgroup_accounting(&trans, fs_info);
+	if (ret) {
+		test_msg("Qgroup accounting failed %d\n", ret);
+		return -EINVAL;
+	}
+
+	if (btrfs_verify_qgroup_counts(fs_info, 5, 0, 0)) {
+		test_msg("Qgroup counts didn't match expected values\n");
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+/*
+ * Add a ref for two different roots to make sure the shared value comes out
+ * right, also remove one of the roots and make sure the exclusive count is
+ * adjusted properly.
+ */
+static int test_multiple_refs(struct btrfs_root *root)
+{
+	struct btrfs_trans_handle trans;
+	struct btrfs_fs_info *fs_info = root->fs_info;
+	int ret;
+
+	init_dummy_trans(&trans);
+
+	test_msg("Qgroup multiple refs test\n");
+
+	/* We have 5 created already from the previous test */
+	ret = btrfs_create_qgroup(NULL, fs_info, 256, NULL);
+	if (ret) {
+		test_msg("Couldn't create a qgroup %d\n", ret);
+		return ret;
+	}
+
+	ret = insert_normal_tree_ref(root, 4096, 4096, 0, 5);
+	if (ret)
+		return ret;
+
+	ret = btrfs_qgroup_record_ref(&trans, fs_info, 5, 4096, 4096,
+				      BTRFS_QGROUP_OPER_ADD_EXCL, 0);
+	if (ret) {
+		test_msg("Couldn't add space to a qgroup %d\n", ret);
+		return ret;
+	}
+
+	ret = btrfs_delayed_qgroup_accounting(&trans, fs_info);
+	if (ret) {
+		test_msg("Delayed qgroup accounting failed %d\n", ret);
+		return ret;
+	}
+
+	if (btrfs_verify_qgroup_counts(fs_info, 5, 4096, 4096)) {
+		test_msg("Qgroup counts didn't match expected values\n");
+		return -EINVAL;
+	}
+
+	ret = add_tree_ref(root, 4096, 4096, 0, 256);
+	if (ret)
+		return ret;
+
+	ret = btrfs_qgroup_record_ref(&trans, fs_info, 256, 4096, 4096,
+				      BTRFS_QGROUP_OPER_ADD_SHARED, 0);
+	if (ret) {
+		test_msg("Qgroup record ref failed %d\n", ret);
+		return ret;
+	}
+
+	ret = btrfs_delayed_qgroup_accounting(&trans, fs_info);
+	if (ret) {
+		test_msg("Qgroup accounting failed %d\n", ret);
+		return ret;
+	}
+
+	if (btrfs_verify_qgroup_counts(fs_info, 5, 4096, 0)) {
+		test_msg("Qgroup counts didn't match expected values\n");
+		return -EINVAL;
+	}
+
+	if (btrfs_verify_qgroup_counts(fs_info, 256, 4096, 0)) {
+		test_msg("Qgroup counts didn't match expected values\n");
+		return -EINVAL;
+	}
+
+	ret = remove_extent_ref(root, 4096, 4096, 0, 256);
+	if (ret)
+		return ret;
+
+	ret = btrfs_qgroup_record_ref(&trans, fs_info, 256, 4096, 4096,
+				      BTRFS_QGROUP_OPER_SUB_SHARED, 0);
+	if (ret) {
+		test_msg("Qgroup record ref failed %d\n", ret);
+		return ret;
+	}
+
+	ret = btrfs_delayed_qgroup_accounting(&trans, fs_info);
+	if (ret) {
+		test_msg("Qgroup accounting failed %d\n", ret);
+		return ret;
+	}
+
+	if (btrfs_verify_qgroup_counts(fs_info, 256, 0, 0)) {
+		test_msg("Qgroup counts didn't match expected values\n");
+		return -EINVAL;
+	}
+
+	if (btrfs_verify_qgroup_counts(fs_info, 5, 4096, 4096)) {
+		test_msg("Qgroup counts didn't match expected values\n");
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+int btrfs_test_qgroups(void)
+{
+	struct btrfs_root *root;
+	struct btrfs_root *tmp_root;
+	int ret = 0;
+
+	root = btrfs_alloc_dummy_root();
+	if (IS_ERR(root)) {
+		test_msg("Couldn't allocate root\n");
+		return PTR_ERR(root);
+	}
+
+	root->fs_info = btrfs_alloc_dummy_fs_info();
+	if (!root->fs_info) {
+		test_msg("Couldn't allocate dummy fs info\n");
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	/*
+	 * Can't use bytenr 0, some things freak out
+	 * *cough*backref walking code*cough*
+	 */
+	root->node = alloc_test_extent_buffer(root->fs_info, 4096, 4096);
+	if (!root->node) {
+		test_msg("Couldn't allocate dummy buffer\n");
+		ret = -ENOMEM;
+		goto out;
+	}
+	btrfs_set_header_level(root->node, 0);
+	btrfs_set_header_nritems(root->node, 0);
+	root->alloc_bytenr += 8192;
+
+	tmp_root = btrfs_alloc_dummy_root();
+	if (IS_ERR(tmp_root)) {
+		test_msg("Couldn't allocate a fs root\n");
+		ret = PTR_ERR(tmp_root);
+		goto out;
+	}
+
+	tmp_root->root_key.objectid = 5;
+	root->fs_info->fs_root = tmp_root;
+	ret = btrfs_insert_fs_root(root->fs_info, tmp_root);
+	if (ret) {
+		test_msg("Couldn't insert fs root %d\n", ret);
+		goto out;
+	}
+
+	tmp_root = btrfs_alloc_dummy_root();
+	if (IS_ERR(tmp_root)) {
+		test_msg("Couldn't allocate a fs root\n");
+		ret = PTR_ERR(tmp_root);
+		goto out;
+	}
+
+	tmp_root->root_key.objectid = 256;
+	ret = btrfs_insert_fs_root(root->fs_info, tmp_root);
+	if (ret) {
+		test_msg("Couldn't insert fs root %d\n", ret);
+		goto out;
+	}
+
+	/* We are using this root as our extent root */
+	root->fs_info->extent_root = root;
+
+	/*
+	 * Some of the paths we test assume we have a filled out fs_info, so we
+	 * just need to addt he root in there so we don't panic.
+	 */
+	root->fs_info->tree_root = root;
+	root->fs_info->quota_root = root;
+	root->fs_info->quota_enabled = 1;
+
+	test_msg("Running qgroup tests\n");
+	ret = test_no_shared_qgroup(root);
+	if (ret)
+		goto out;
+	ret = test_multiple_refs(root);
+out:
+	btrfs_free_dummy_root(root);
+	return ret;
+}
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index 2cb116099b9..5f379affdf2 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -22,75 +22,214 @@
 #include <linux/writeback.h>
 #include <linux/pagemap.h>
 #include <linux/blkdev.h>
+#include <linux/uuid.h>
 #include "ctree.h"
 #include "disk-io.h"
 #include "transaction.h"
 #include "locking.h"
 #include "tree-log.h"
+#include "inode-map.h"
+#include "volumes.h"
+#include "dev-replace.h"
+#include "qgroup.h"
 
 #define BTRFS_ROOT_TRANS_TAG 0
 
-static noinline void put_transaction(struct btrfs_transaction *transaction)
+static unsigned int btrfs_blocked_trans_types[TRANS_STATE_MAX] = {
+	[TRANS_STATE_RUNNING]		= 0U,
+	[TRANS_STATE_BLOCKED]		= (__TRANS_USERSPACE |
+					   __TRANS_START),
+	[TRANS_STATE_COMMIT_START]	= (__TRANS_USERSPACE |
+					   __TRANS_START |
+					   __TRANS_ATTACH),
+	[TRANS_STATE_COMMIT_DOING]	= (__TRANS_USERSPACE |
+					   __TRANS_START |
+					   __TRANS_ATTACH |
+					   __TRANS_JOIN),
+	[TRANS_STATE_UNBLOCKED]		= (__TRANS_USERSPACE |
+					   __TRANS_START |
+					   __TRANS_ATTACH |
+					   __TRANS_JOIN |
+					   __TRANS_JOIN_NOLOCK),
+	[TRANS_STATE_COMPLETED]		= (__TRANS_USERSPACE |
+					   __TRANS_START |
+					   __TRANS_ATTACH |
+					   __TRANS_JOIN |
+					   __TRANS_JOIN_NOLOCK),
+};
+
+void btrfs_put_transaction(struct btrfs_transaction *transaction)
 {
-	WARN_ON(transaction->use_count == 0);
-	transaction->use_count--;
-	if (transaction->use_count == 0) {
-		list_del_init(&transaction->list);
-		memset(transaction, 0, sizeof(*transaction));
+	WARN_ON(atomic_read(&transaction->use_count) == 0);
+	if (atomic_dec_and_test(&transaction->use_count)) {
+		BUG_ON(!list_empty(&transaction->list));
+		WARN_ON(!RB_EMPTY_ROOT(&transaction->delayed_refs.href_root));
+		while (!list_empty(&transaction->pending_chunks)) {
+			struct extent_map *em;
+
+			em = list_first_entry(&transaction->pending_chunks,
+					      struct extent_map, list);
+			list_del_init(&em->list);
+			free_extent_map(em);
+		}
 		kmem_cache_free(btrfs_transaction_cachep, transaction);
 	}
 }
 
-static noinline void switch_commit_root(struct btrfs_root *root)
+static noinline void switch_commit_roots(struct btrfs_transaction *trans,
+					 struct btrfs_fs_info *fs_info)
+{
+	struct btrfs_root *root, *tmp;
+
+	down_write(&fs_info->commit_root_sem);
+	list_for_each_entry_safe(root, tmp, &trans->switch_commits,
+				 dirty_list) {
+		list_del_init(&root->dirty_list);
+		free_extent_buffer(root->commit_root);
+		root->commit_root = btrfs_root_node(root);
+		if (is_fstree(root->objectid))
+			btrfs_unpin_free_ino(root);
+	}
+	up_write(&fs_info->commit_root_sem);
+}
+
+static inline void extwriter_counter_inc(struct btrfs_transaction *trans,
+					 unsigned int type)
+{
+	if (type & TRANS_EXTWRITERS)
+		atomic_inc(&trans->num_extwriters);
+}
+
+static inline void extwriter_counter_dec(struct btrfs_transaction *trans,
+					 unsigned int type)
 {
-	free_extent_buffer(root->commit_root);
-	root->commit_root = btrfs_root_node(root);
+	if (type & TRANS_EXTWRITERS)
+		atomic_dec(&trans->num_extwriters);
+}
+
+static inline void extwriter_counter_init(struct btrfs_transaction *trans,
+					  unsigned int type)
+{
+	atomic_set(&trans->num_extwriters, ((type & TRANS_EXTWRITERS) ? 1 : 0));
+}
+
+static inline int extwriter_counter_read(struct btrfs_transaction *trans)
+{
+	return atomic_read(&trans->num_extwriters);
 }
 
 /*
  * either allocate a new transaction or hop into the existing one
  */
-static noinline int join_transaction(struct btrfs_root *root)
+static noinline int join_transaction(struct btrfs_root *root, unsigned int type)
 {
 	struct btrfs_transaction *cur_trans;
-	cur_trans = root->fs_info->running_transaction;
-	if (!cur_trans) {
-		cur_trans = kmem_cache_alloc(btrfs_transaction_cachep,
-					     GFP_NOFS);
-		BUG_ON(!cur_trans);
-		root->fs_info->generation++;
-		cur_trans->num_writers = 1;
-		cur_trans->num_joined = 0;
-		cur_trans->transid = root->fs_info->generation;
-		init_waitqueue_head(&cur_trans->writer_wait);
-		init_waitqueue_head(&cur_trans->commit_wait);
-		cur_trans->in_commit = 0;
-		cur_trans->blocked = 0;
-		cur_trans->use_count = 1;
-		cur_trans->commit_done = 0;
-		cur_trans->start_time = get_seconds();
-
-		cur_trans->delayed_refs.root = RB_ROOT;
-		cur_trans->delayed_refs.num_entries = 0;
-		cur_trans->delayed_refs.num_heads_ready = 0;
-		cur_trans->delayed_refs.num_heads = 0;
-		cur_trans->delayed_refs.flushing = 0;
-		cur_trans->delayed_refs.run_delayed_start = 0;
-		spin_lock_init(&cur_trans->delayed_refs.lock);
-
-		INIT_LIST_HEAD(&cur_trans->pending_snapshots);
-		list_add_tail(&cur_trans->list, &root->fs_info->trans_list);
-		extent_io_tree_init(&cur_trans->dirty_pages,
-				     root->fs_info->btree_inode->i_mapping,
-				     GFP_NOFS);
-		spin_lock(&root->fs_info->new_trans_lock);
-		root->fs_info->running_transaction = cur_trans;
-		spin_unlock(&root->fs_info->new_trans_lock);
-	} else {
-		cur_trans->num_writers++;
-		cur_trans->num_joined++;
+	struct btrfs_fs_info *fs_info = root->fs_info;
+
+	spin_lock(&fs_info->trans_lock);
+loop:
+	/* The file system has been taken offline. No new transactions. */
+	if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
+		spin_unlock(&fs_info->trans_lock);
+		return -EROFS;
 	}
 
+	cur_trans = fs_info->running_transaction;
+	if (cur_trans) {
+		if (cur_trans->aborted) {
+			spin_unlock(&fs_info->trans_lock);
+			return cur_trans->aborted;
+		}
+		if (btrfs_blocked_trans_types[cur_trans->state] & type) {
+			spin_unlock(&fs_info->trans_lock);
+			return -EBUSY;
+		}
+		atomic_inc(&cur_trans->use_count);
+		atomic_inc(&cur_trans->num_writers);
+		extwriter_counter_inc(cur_trans, type);
+		spin_unlock(&fs_info->trans_lock);
+		return 0;
+	}
+	spin_unlock(&fs_info->trans_lock);
+
+	/*
+	 * If we are ATTACH, we just want to catch the current transaction,
+	 * and commit it. If there is no transaction, just return ENOENT.
+	 */
+	if (type == TRANS_ATTACH)
+		return -ENOENT;
+
+	/*
+	 * JOIN_NOLOCK only happens during the transaction commit, so
+	 * it is impossible that ->running_transaction is NULL
+	 */
+	BUG_ON(type == TRANS_JOIN_NOLOCK);
+
+	cur_trans = kmem_cache_alloc(btrfs_transaction_cachep, GFP_NOFS);
+	if (!cur_trans)
+		return -ENOMEM;
+
+	spin_lock(&fs_info->trans_lock);
+	if (fs_info->running_transaction) {
+		/*
+		 * someone started a transaction after we unlocked.  Make sure
+		 * to redo the checks above
+		 */
+		kmem_cache_free(btrfs_transaction_cachep, cur_trans);
+		goto loop;
+	} else if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
+		spin_unlock(&fs_info->trans_lock);
+		kmem_cache_free(btrfs_transaction_cachep, cur_trans);
+		return -EROFS;
+	}
+
+	atomic_set(&cur_trans->num_writers, 1);
+	extwriter_counter_init(cur_trans, type);
+	init_waitqueue_head(&cur_trans->writer_wait);
+	init_waitqueue_head(&cur_trans->commit_wait);
+	cur_trans->state = TRANS_STATE_RUNNING;
+	/*
+	 * One for this trans handle, one so it will live on until we
+	 * commit the transaction.
+	 */
+	atomic_set(&cur_trans->use_count, 2);
+	cur_trans->start_time = get_seconds();
+
+	cur_trans->delayed_refs.href_root = RB_ROOT;
+	atomic_set(&cur_trans->delayed_refs.num_entries, 0);
+	cur_trans->delayed_refs.num_heads_ready = 0;
+	cur_trans->delayed_refs.num_heads = 0;
+	cur_trans->delayed_refs.flushing = 0;
+	cur_trans->delayed_refs.run_delayed_start = 0;
+
+	/*
+	 * although the tree mod log is per file system and not per transaction,
+	 * the log must never go across transaction boundaries.
+	 */
+	smp_mb();
+	if (!list_empty(&fs_info->tree_mod_seq_list))
+		WARN(1, KERN_ERR "BTRFS: tree_mod_seq_list not empty when "
+			"creating a fresh transaction\n");
+	if (!RB_EMPTY_ROOT(&fs_info->tree_mod_log))
+		WARN(1, KERN_ERR "BTRFS: tree_mod_log rb tree not empty when "
+			"creating a fresh transaction\n");
+	atomic64_set(&fs_info->tree_mod_seq, 0);
+
+	spin_lock_init(&cur_trans->delayed_refs.lock);
+
+	INIT_LIST_HEAD(&cur_trans->pending_snapshots);
+	INIT_LIST_HEAD(&cur_trans->ordered_operations);
+	INIT_LIST_HEAD(&cur_trans->pending_chunks);
+	INIT_LIST_HEAD(&cur_trans->switch_commits);
+	list_add_tail(&cur_trans->list, &fs_info->trans_list);
+	extent_io_tree_init(&cur_trans->dirty_pages,
+			     fs_info->btree_inode->i_mapping);
+	fs_info->generation++;
+	cur_trans->transid = fs_info->generation;
+	fs_info->running_transaction = cur_trans;
+	cur_trans->aborted = 0;
+	spin_unlock(&fs_info->trans_lock);
+
 	return 0;
 }
 
@@ -100,39 +239,93 @@ static noinline int join_transaction(struct btrfs_root *root)
  * to make sure the old root from before we joined the transaction is deleted
  * when the transaction commits
  */
-static noinline int record_root_in_trans(struct btrfs_trans_handle *trans,
-					 struct btrfs_root *root)
+static int record_root_in_trans(struct btrfs_trans_handle *trans,
+			       struct btrfs_root *root)
 {
-	if (root->ref_cows && root->last_trans < trans->transid) {
+	if (test_bit(BTRFS_ROOT_REF_COWS, &root->state) &&
+	    root->last_trans < trans->transid) {
 		WARN_ON(root == root->fs_info->extent_root);
 		WARN_ON(root->commit_root != root->node);
 
+		/*
+		 * see below for IN_TRANS_SETUP usage rules
+		 * we have the reloc mutex held now, so there
+		 * is only one writer in this function
+		 */
+		set_bit(BTRFS_ROOT_IN_TRANS_SETUP, &root->state);
+
+		/* make sure readers find IN_TRANS_SETUP before
+		 * they find our root->last_trans update
+		 */
+		smp_wmb();
+
+		spin_lock(&root->fs_info->fs_roots_radix_lock);
+		if (root->last_trans == trans->transid) {
+			spin_unlock(&root->fs_info->fs_roots_radix_lock);
+			return 0;
+		}
 		radix_tree_tag_set(&root->fs_info->fs_roots_radix,
 			   (unsigned long)root->root_key.objectid,
 			   BTRFS_ROOT_TRANS_TAG);
+		spin_unlock(&root->fs_info->fs_roots_radix_lock);
 		root->last_trans = trans->transid;
+
+		/* this is pretty tricky.  We don't want to
+		 * take the relocation lock in btrfs_record_root_in_trans
+		 * unless we're really doing the first setup for this root in
+		 * this transaction.
+		 *
+		 * Normally we'd use root->last_trans as a flag to decide
+		 * if we want to take the expensive mutex.
+		 *
+		 * But, we have to set root->last_trans before we
+		 * init the relocation root, otherwise, we trip over warnings
+		 * in ctree.c.  The solution used here is to flag ourselves
+		 * with root IN_TRANS_SETUP.  When this is 1, we're still
+		 * fixing up the reloc trees and everyone must wait.
+		 *
+		 * When this is zero, they can trust root->last_trans and fly
+		 * through btrfs_record_root_in_trans without having to take the
+		 * lock.  smp_wmb() makes sure that all the writes above are
+		 * done before we pop in the zero below
+		 */
 		btrfs_init_reloc_root(trans, root);
+		smp_mb__before_atomic();
+		clear_bit(BTRFS_ROOT_IN_TRANS_SETUP, &root->state);
 	}
 	return 0;
 }
 
+
 int btrfs_record_root_in_trans(struct btrfs_trans_handle *trans,
 			       struct btrfs_root *root)
 {
-	if (!root->ref_cows)
+	if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
 		return 0;
 
-	mutex_lock(&root->fs_info->trans_mutex);
-	if (root->last_trans == trans->transid) {
-		mutex_unlock(&root->fs_info->trans_mutex);
+	/*
+	 * see record_root_in_trans for comments about IN_TRANS_SETUP usage
+	 * and barriers
+	 */
+	smp_rmb();
+	if (root->last_trans == trans->transid &&
+	    !test_bit(BTRFS_ROOT_IN_TRANS_SETUP, &root->state))
 		return 0;
-	}
 
+	mutex_lock(&root->fs_info->reloc_mutex);
 	record_root_in_trans(trans, root);
-	mutex_unlock(&root->fs_info->trans_mutex);
+	mutex_unlock(&root->fs_info->reloc_mutex);
+
 	return 0;
 }
 
+static inline int is_transaction_blocked(struct btrfs_transaction *trans)
+{
+	return (trans->state >= TRANS_STATE_BLOCKED &&
+		trans->state < TRANS_STATE_UNBLOCKED &&
+		!trans->aborted);
+}
+
 /* wait for commit against the current transaction to become unblocked
  * when this is done, it is safe to start a new transaction, but the current
  * transaction might not be fully on disk.
@@ -141,198 +334,464 @@ static void wait_current_trans(struct btrfs_root *root)
 {
 	struct btrfs_transaction *cur_trans;
 
+	spin_lock(&root->fs_info->trans_lock);
 	cur_trans = root->fs_info->running_transaction;
-	if (cur_trans && cur_trans->blocked) {
-		DEFINE_WAIT(wait);
-		cur_trans->use_count++;
-		while (1) {
-			prepare_to_wait(&root->fs_info->transaction_wait, &wait,
-					TASK_UNINTERRUPTIBLE);
-			if (!cur_trans->blocked)
-				break;
-			mutex_unlock(&root->fs_info->trans_mutex);
-			schedule();
-			mutex_lock(&root->fs_info->trans_mutex);
-		}
-		finish_wait(&root->fs_info->transaction_wait, &wait);
-		put_transaction(cur_trans);
+	if (cur_trans && is_transaction_blocked(cur_trans)) {
+		atomic_inc(&cur_trans->use_count);
+		spin_unlock(&root->fs_info->trans_lock);
+
+		wait_event(root->fs_info->transaction_wait,
+			   cur_trans->state >= TRANS_STATE_UNBLOCKED ||
+			   cur_trans->aborted);
+		btrfs_put_transaction(cur_trans);
+	} else {
+		spin_unlock(&root->fs_info->trans_lock);
 	}
 }
 
-enum btrfs_trans_type {
-	TRANS_START,
-	TRANS_JOIN,
-	TRANS_USERSPACE,
-};
+static int may_wait_transaction(struct btrfs_root *root, int type)
+{
+	if (root->fs_info->log_root_recovering)
+		return 0;
+
+	if (type == TRANS_USERSPACE)
+		return 1;
 
-static struct btrfs_trans_handle *start_transaction(struct btrfs_root *root,
-					     int num_blocks, int type)
+	if (type == TRANS_START &&
+	    !atomic_read(&root->fs_info->open_ioctl_trans))
+		return 1;
+
+	return 0;
+}
+
+static inline bool need_reserve_reloc_root(struct btrfs_root *root)
 {
-	struct btrfs_trans_handle *h =
-		kmem_cache_alloc(btrfs_trans_handle_cachep, GFP_NOFS);
+	if (!root->fs_info->reloc_ctl ||
+	    !test_bit(BTRFS_ROOT_REF_COWS, &root->state) ||
+	    root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID ||
+	    root->reloc_root)
+		return false;
+
+	return true;
+}
+
+static struct btrfs_trans_handle *
+start_transaction(struct btrfs_root *root, u64 num_items, unsigned int type,
+		  enum btrfs_reserve_flush_enum flush)
+{
+	struct btrfs_trans_handle *h;
+	struct btrfs_transaction *cur_trans;
+	u64 num_bytes = 0;
+	u64 qgroup_reserved = 0;
+	bool reloc_reserved = false;
 	int ret;
 
-	mutex_lock(&root->fs_info->trans_mutex);
-	if (!root->fs_info->log_root_recovering &&
-	    ((type == TRANS_START && !root->fs_info->open_ioctl_trans) ||
-	     type == TRANS_USERSPACE))
+	/* Send isn't supposed to start transactions. */
+	ASSERT(current->journal_info != (void *)BTRFS_SEND_TRANS_STUB);
+
+	if (test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state))
+		return ERR_PTR(-EROFS);
+
+	if (current->journal_info) {
+		WARN_ON(type & TRANS_EXTWRITERS);
+		h = current->journal_info;
+		h->use_count++;
+		WARN_ON(h->use_count > 2);
+		h->orig_rsv = h->block_rsv;
+		h->block_rsv = NULL;
+		goto got_it;
+	}
+
+	/*
+	 * Do the reservation before we join the transaction so we can do all
+	 * the appropriate flushing if need be.
+	 */
+	if (num_items > 0 && root != root->fs_info->chunk_root) {
+		if (root->fs_info->quota_enabled &&
+		    is_fstree(root->root_key.objectid)) {
+			qgroup_reserved = num_items * root->leafsize;
+			ret = btrfs_qgroup_reserve(root, qgroup_reserved);
+			if (ret)
+				return ERR_PTR(ret);
+		}
+
+		num_bytes = btrfs_calc_trans_metadata_size(root, num_items);
+		/*
+		 * Do the reservation for the relocation root creation
+		 */
+		if (unlikely(need_reserve_reloc_root(root))) {
+			num_bytes += root->nodesize;
+			reloc_reserved = true;
+		}
+
+		ret = btrfs_block_rsv_add(root,
+					  &root->fs_info->trans_block_rsv,
+					  num_bytes, flush);
+		if (ret)
+			goto reserve_fail;
+	}
+again:
+	h = kmem_cache_alloc(btrfs_trans_handle_cachep, GFP_NOFS);
+	if (!h) {
+		ret = -ENOMEM;
+		goto alloc_fail;
+	}
+
+	/*
+	 * If we are JOIN_NOLOCK we're already committing a transaction and
+	 * waiting on this guy, so we don't need to do the sb_start_intwrite
+	 * because we're already holding a ref.  We need this because we could
+	 * have raced in and did an fsync() on a file which can kick a commit
+	 * and then we deadlock with somebody doing a freeze.
+	 *
+	 * If we are ATTACH, it means we just want to catch the current
+	 * transaction and commit it, so we needn't do sb_start_intwrite(). 
+	 */
+	if (type & __TRANS_FREEZABLE)
+		sb_start_intwrite(root->fs_info->sb);
+
+	if (may_wait_transaction(root, type))
 		wait_current_trans(root);
-	ret = join_transaction(root);
-	BUG_ON(ret);
 
-	h->transid = root->fs_info->running_transaction->transid;
-	h->transaction = root->fs_info->running_transaction;
-	h->blocks_reserved = num_blocks;
+	do {
+		ret = join_transaction(root, type);
+		if (ret == -EBUSY) {
+			wait_current_trans(root);
+			if (unlikely(type == TRANS_ATTACH))
+				ret = -ENOENT;
+		}
+	} while (ret == -EBUSY);
+
+	if (ret < 0) {
+		/* We must get the transaction if we are JOIN_NOLOCK. */
+		BUG_ON(type == TRANS_JOIN_NOLOCK);
+		goto join_fail;
+	}
+
+	cur_trans = root->fs_info->running_transaction;
+
+	h->transid = cur_trans->transid;
+	h->transaction = cur_trans;
 	h->blocks_used = 0;
-	h->block_group = 0;
-	h->alloc_exclude_nr = 0;
-	h->alloc_exclude_start = 0;
+	h->bytes_reserved = 0;
+	h->root = root;
 	h->delayed_ref_updates = 0;
+	h->use_count = 1;
+	h->adding_csums = 0;
+	h->block_rsv = NULL;
+	h->orig_rsv = NULL;
+	h->aborted = 0;
+	h->qgroup_reserved = 0;
+	h->delayed_ref_elem.seq = 0;
+	h->type = type;
+	h->allocating_chunk = false;
+	h->reloc_reserved = false;
+	h->sync = false;
+	INIT_LIST_HEAD(&h->qgroup_ref_list);
+	INIT_LIST_HEAD(&h->new_bgs);
 
-	if (!current->journal_info && type != TRANS_USERSPACE)
+	smp_mb();
+	if (cur_trans->state >= TRANS_STATE_BLOCKED &&
+	    may_wait_transaction(root, type)) {
 		current->journal_info = h;
+		btrfs_commit_transaction(h, root);
+		goto again;
+	}
+
+	if (num_bytes) {
+		trace_btrfs_space_reservation(root->fs_info, "transaction",
+					      h->transid, num_bytes, 1);
+		h->block_rsv = &root->fs_info->trans_block_rsv;
+		h->bytes_reserved = num_bytes;
+		h->reloc_reserved = reloc_reserved;
+	}
+	h->qgroup_reserved = qgroup_reserved;
+
+got_it:
+	btrfs_record_root_in_trans(h, root);
 
-	root->fs_info->running_transaction->use_count++;
-	record_root_in_trans(h, root);
-	mutex_unlock(&root->fs_info->trans_mutex);
+	if (!current->journal_info && type != TRANS_USERSPACE)
+		current->journal_info = h;
 	return h;
+
+join_fail:
+	if (type & __TRANS_FREEZABLE)
+		sb_end_intwrite(root->fs_info->sb);
+	kmem_cache_free(btrfs_trans_handle_cachep, h);
+alloc_fail:
+	if (num_bytes)
+		btrfs_block_rsv_release(root, &root->fs_info->trans_block_rsv,
+					num_bytes);
+reserve_fail:
+	if (qgroup_reserved)
+		btrfs_qgroup_free(root, qgroup_reserved);
+	return ERR_PTR(ret);
 }
 
 struct btrfs_trans_handle *btrfs_start_transaction(struct btrfs_root *root,
-						   int num_blocks)
+						   int num_items)
 {
-	return start_transaction(root, num_blocks, TRANS_START);
+	return start_transaction(root, num_items, TRANS_START,
+				 BTRFS_RESERVE_FLUSH_ALL);
 }
-struct btrfs_trans_handle *btrfs_join_transaction(struct btrfs_root *root,
-						   int num_blocks)
+
+struct btrfs_trans_handle *btrfs_start_transaction_lflush(
+					struct btrfs_root *root, int num_items)
 {
-	return start_transaction(root, num_blocks, TRANS_JOIN);
+	return start_transaction(root, num_items, TRANS_START,
+				 BTRFS_RESERVE_FLUSH_LIMIT);
 }
 
-struct btrfs_trans_handle *btrfs_start_ioctl_transaction(struct btrfs_root *r,
-							 int num_blocks)
+struct btrfs_trans_handle *btrfs_join_transaction(struct btrfs_root *root)
 {
-	return start_transaction(r, num_blocks, TRANS_USERSPACE);
+	return start_transaction(root, 0, TRANS_JOIN, 0);
 }
 
-/* wait for a transaction commit to be fully complete */
-static noinline int wait_for_commit(struct btrfs_root *root,
-				    struct btrfs_transaction *commit)
+struct btrfs_trans_handle *btrfs_join_transaction_nolock(struct btrfs_root *root)
 {
-	DEFINE_WAIT(wait);
-	mutex_lock(&root->fs_info->trans_mutex);
-	while (!commit->commit_done) {
-		prepare_to_wait(&commit->commit_wait, &wait,
-				TASK_UNINTERRUPTIBLE);
-		if (commit->commit_done)
-			break;
-		mutex_unlock(&root->fs_info->trans_mutex);
-		schedule();
-		mutex_lock(&root->fs_info->trans_mutex);
-	}
-	mutex_unlock(&root->fs_info->trans_mutex);
-	finish_wait(&commit->commit_wait, &wait);
-	return 0;
+	return start_transaction(root, 0, TRANS_JOIN_NOLOCK, 0);
+}
+
+struct btrfs_trans_handle *btrfs_start_ioctl_transaction(struct btrfs_root *root)
+{
+	return start_transaction(root, 0, TRANS_USERSPACE, 0);
 }
 
-#if 0
 /*
- * rate limit against the drop_snapshot code.  This helps to slow down new
- * operations if the drop_snapshot code isn't able to keep up.
+ * btrfs_attach_transaction() - catch the running transaction
+ *
+ * It is used when we want to commit the current the transaction, but
+ * don't want to start a new one.
+ *
+ * Note: If this function return -ENOENT, it just means there is no
+ * running transaction. But it is possible that the inactive transaction
+ * is still in the memory, not fully on disk. If you hope there is no
+ * inactive transaction in the fs when -ENOENT is returned, you should
+ * invoke
+ *     btrfs_attach_transaction_barrier()
  */
-static void throttle_on_drops(struct btrfs_root *root)
+struct btrfs_trans_handle *btrfs_attach_transaction(struct btrfs_root *root)
 {
-	struct btrfs_fs_info *info = root->fs_info;
-	int harder_count = 0;
-
-harder:
-	if (atomic_read(&info->throttles)) {
-		DEFINE_WAIT(wait);
-		int thr;
-		thr = atomic_read(&info->throttle_gen);
-
-		do {
-			prepare_to_wait(&info->transaction_throttle,
-					&wait, TASK_UNINTERRUPTIBLE);
-			if (!atomic_read(&info->throttles)) {
-				finish_wait(&info->transaction_throttle, &wait);
-				break;
-			}
-			schedule();
-			finish_wait(&info->transaction_throttle, &wait);
-		} while (thr == atomic_read(&info->throttle_gen));
-		harder_count++;
+	return start_transaction(root, 0, TRANS_ATTACH, 0);
+}
+
+/*
+ * btrfs_attach_transaction_barrier() - catch the running transaction
+ *
+ * It is similar to the above function, the differentia is this one
+ * will wait for all the inactive transactions until they fully
+ * complete.
+ */
+struct btrfs_trans_handle *
+btrfs_attach_transaction_barrier(struct btrfs_root *root)
+{
+	struct btrfs_trans_handle *trans;
 
-		if (root->fs_info->total_ref_cache_size > 1 * 1024 * 1024 &&
-		    harder_count < 2)
-			goto harder;
+	trans = start_transaction(root, 0, TRANS_ATTACH, 0);
+	if (IS_ERR(trans) && PTR_ERR(trans) == -ENOENT)
+		btrfs_wait_for_commit(root, 0);
 
-		if (root->fs_info->total_ref_cache_size > 5 * 1024 * 1024 &&
-		    harder_count < 10)
-			goto harder;
+	return trans;
+}
+
+/* wait for a transaction commit to be fully complete */
+static noinline void wait_for_commit(struct btrfs_root *root,
+				    struct btrfs_transaction *commit)
+{
+	wait_event(commit->commit_wait, commit->state == TRANS_STATE_COMPLETED);
+}
+
+int btrfs_wait_for_commit(struct btrfs_root *root, u64 transid)
+{
+	struct btrfs_transaction *cur_trans = NULL, *t;
+	int ret = 0;
 
-		if (root->fs_info->total_ref_cache_size > 10 * 1024 * 1024 &&
-		    harder_count < 20)
-			goto harder;
+	if (transid) {
+		if (transid <= root->fs_info->last_trans_committed)
+			goto out;
+
+		ret = -EINVAL;
+		/* find specified transaction */
+		spin_lock(&root->fs_info->trans_lock);
+		list_for_each_entry(t, &root->fs_info->trans_list, list) {
+			if (t->transid == transid) {
+				cur_trans = t;
+				atomic_inc(&cur_trans->use_count);
+				ret = 0;
+				break;
+			}
+			if (t->transid > transid) {
+				ret = 0;
+				break;
+			}
+		}
+		spin_unlock(&root->fs_info->trans_lock);
+		/* The specified transaction doesn't exist */
+		if (!cur_trans)
+			goto out;
+	} else {
+		/* find newest transaction that is committing | committed */
+		spin_lock(&root->fs_info->trans_lock);
+		list_for_each_entry_reverse(t, &root->fs_info->trans_list,
+					    list) {
+			if (t->state >= TRANS_STATE_COMMIT_START) {
+				if (t->state == TRANS_STATE_COMPLETED)
+					break;
+				cur_trans = t;
+				atomic_inc(&cur_trans->use_count);
+				break;
+			}
+		}
+		spin_unlock(&root->fs_info->trans_lock);
+		if (!cur_trans)
+			goto out;  /* nothing committing|committed */
 	}
+
+	wait_for_commit(root, cur_trans);
+	btrfs_put_transaction(cur_trans);
+out:
+	return ret;
 }
-#endif
 
 void btrfs_throttle(struct btrfs_root *root)
 {
-	mutex_lock(&root->fs_info->trans_mutex);
-	if (!root->fs_info->open_ioctl_trans)
+	if (!atomic_read(&root->fs_info->open_ioctl_trans))
 		wait_current_trans(root);
-	mutex_unlock(&root->fs_info->trans_mutex);
+}
+
+static int should_end_transaction(struct btrfs_trans_handle *trans,
+				  struct btrfs_root *root)
+{
+	if (root->fs_info->global_block_rsv.space_info->full &&
+	    btrfs_check_space_for_delayed_refs(trans, root))
+		return 1;
+
+	return !!btrfs_block_rsv_check(root, &root->fs_info->global_block_rsv, 5);
+}
+
+int btrfs_should_end_transaction(struct btrfs_trans_handle *trans,
+				 struct btrfs_root *root)
+{
+	struct btrfs_transaction *cur_trans = trans->transaction;
+	int updates;
+	int err;
+
+	smp_mb();
+	if (cur_trans->state >= TRANS_STATE_BLOCKED ||
+	    cur_trans->delayed_refs.flushing)
+		return 1;
+
+	updates = trans->delayed_ref_updates;
+	trans->delayed_ref_updates = 0;
+	if (updates) {
+		err = btrfs_run_delayed_refs(trans, root, updates);
+		if (err) /* Error code will also eval true */
+			return err;
+	}
+
+	return should_end_transaction(trans, root);
 }
 
 static int __btrfs_end_transaction(struct btrfs_trans_handle *trans,
 			  struct btrfs_root *root, int throttle)
 {
-	struct btrfs_transaction *cur_trans;
+	struct btrfs_transaction *cur_trans = trans->transaction;
 	struct btrfs_fs_info *info = root->fs_info;
-	int count = 0;
-
-	while (count < 4) {
-		unsigned long cur = trans->delayed_ref_updates;
-		trans->delayed_ref_updates = 0;
-		if (cur &&
-		    trans->transaction->delayed_refs.num_heads_ready > 64) {
-			trans->delayed_ref_updates = 0;
-
-			/*
-			 * do a full flush if the transaction is trying
-			 * to close
-			 */
-			if (trans->transaction->delayed_refs.flushing)
-				cur = 0;
-			btrfs_run_delayed_refs(trans, root, cur);
-		} else {
-			break;
-		}
-		count++;
+	unsigned long cur = trans->delayed_ref_updates;
+	int lock = (trans->type != TRANS_JOIN_NOLOCK);
+	int err = 0;
+	int must_run_delayed_refs = 0;
+
+	if (trans->use_count > 1) {
+		trans->use_count--;
+		trans->block_rsv = trans->orig_rsv;
+		return 0;
 	}
 
-	mutex_lock(&info->trans_mutex);
-	cur_trans = info->running_transaction;
-	WARN_ON(cur_trans != trans->transaction);
-	WARN_ON(cur_trans->num_writers < 1);
-	cur_trans->num_writers--;
+	btrfs_trans_release_metadata(trans, root);
+	trans->block_rsv = NULL;
+
+	if (!list_empty(&trans->new_bgs))
+		btrfs_create_pending_block_groups(trans, root);
+
+	trans->delayed_ref_updates = 0;
+	if (!trans->sync) {
+		must_run_delayed_refs =
+			btrfs_should_throttle_delayed_refs(trans, root);
+		cur = max_t(unsigned long, cur, 32);
+
+		/*
+		 * don't make the caller wait if they are from a NOLOCK
+		 * or ATTACH transaction, it will deadlock with commit
+		 */
+		if (must_run_delayed_refs == 1 &&
+		    (trans->type & (__TRANS_JOIN_NOLOCK | __TRANS_ATTACH)))
+			must_run_delayed_refs = 2;
+	}
+
+	if (trans->qgroup_reserved) {
+		/*
+		 * the same root has to be passed here between start_transaction
+		 * and end_transaction. Subvolume quota depends on this.
+		 */
+		btrfs_qgroup_free(trans->root, trans->qgroup_reserved);
+		trans->qgroup_reserved = 0;
+	}
+
+	btrfs_trans_release_metadata(trans, root);
+	trans->block_rsv = NULL;
+
+	if (!list_empty(&trans->new_bgs))
+		btrfs_create_pending_block_groups(trans, root);
+
+	if (lock && !atomic_read(&root->fs_info->open_ioctl_trans) &&
+	    should_end_transaction(trans, root) &&
+	    ACCESS_ONCE(cur_trans->state) == TRANS_STATE_RUNNING) {
+		spin_lock(&info->trans_lock);
+		if (cur_trans->state == TRANS_STATE_RUNNING)
+			cur_trans->state = TRANS_STATE_BLOCKED;
+		spin_unlock(&info->trans_lock);
+	}
+
+	if (lock && ACCESS_ONCE(cur_trans->state) == TRANS_STATE_BLOCKED) {
+		if (throttle)
+			return btrfs_commit_transaction(trans, root);
+		else
+			wake_up_process(info->transaction_kthread);
+	}
+
+	if (trans->type & __TRANS_FREEZABLE)
+		sb_end_intwrite(root->fs_info->sb);
 
+	WARN_ON(cur_trans != info->running_transaction);
+	WARN_ON(atomic_read(&cur_trans->num_writers) < 1);
+	atomic_dec(&cur_trans->num_writers);
+	extwriter_counter_dec(cur_trans, trans->type);
+
+	smp_mb();
 	if (waitqueue_active(&cur_trans->writer_wait))
 		wake_up(&cur_trans->writer_wait);
-	put_transaction(cur_trans);
-	mutex_unlock(&info->trans_mutex);
+	btrfs_put_transaction(cur_trans);
 
 	if (current->journal_info == trans)
 		current->journal_info = NULL;
-	memset(trans, 0, sizeof(*trans));
-	kmem_cache_free(btrfs_trans_handle_cachep, trans);
 
 	if (throttle)
 		btrfs_run_delayed_iputs(root);
 
-	return 0;
+	if (trans->aborted ||
+	    test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state)) {
+		wake_up_process(info->transaction_kthread);
+		err = -EIO;
+	}
+	assert_qgroups_uptodate(trans);
+
+	kmem_cache_free(btrfs_trans_handle_cachep, trans);
+	if (must_run_delayed_refs) {
+		btrfs_async_run_delayed_refs(root, cur,
+					     must_run_delayed_refs == 1);
+	}
+	return err;
 }
 
 int btrfs_end_transaction(struct btrfs_trans_handle *trans,
@@ -355,50 +814,23 @@ int btrfs_end_transaction_throttle(struct btrfs_trans_handle *trans,
 int btrfs_write_marked_extents(struct btrfs_root *root,
 			       struct extent_io_tree *dirty_pages, int mark)
 {
-	int ret;
 	int err = 0;
 	int werr = 0;
-	struct page *page;
-	struct inode *btree_inode = root->fs_info->btree_inode;
+	struct address_space *mapping = root->fs_info->btree_inode->i_mapping;
+	struct extent_state *cached_state = NULL;
 	u64 start = 0;
 	u64 end;
-	unsigned long index;
 
-	while (1) {
-		ret = find_first_extent_bit(dirty_pages, start, &start, &end,
-					    mark);
-		if (ret)
-			break;
-		while (start <= end) {
-			cond_resched();
-
-			index = start >> PAGE_CACHE_SHIFT;
-			start = (u64)(index + 1) << PAGE_CACHE_SHIFT;
-			page = find_get_page(btree_inode->i_mapping, index);
-			if (!page)
-				continue;
-
-			btree_lock_page_hook(page);
-			if (!page->mapping) {
-				unlock_page(page);
-				page_cache_release(page);
-				continue;
-			}
-
-			if (PageWriteback(page)) {
-				if (PageDirty(page))
-					wait_on_page_writeback(page);
-				else {
-					unlock_page(page);
-					page_cache_release(page);
-					continue;
-				}
-			}
-			err = write_one_page(page, 0);
-			if (err)
-				werr = err;
-			page_cache_release(page);
-		}
+	while (!find_first_extent_bit(dirty_pages, start, &start, &end,
+				      mark, &cached_state)) {
+		convert_extent_bit(dirty_pages, start, end, EXTENT_NEED_WAIT,
+				   mark, &cached_state, GFP_NOFS);
+		cached_state = NULL;
+		err = filemap_fdatawrite_range(mapping, start, end);
+		if (err)
+			werr = err;
+		cond_resched();
+		start = end + 1;
 	}
 	if (err)
 		werr = err;
@@ -414,39 +846,22 @@ int btrfs_write_marked_extents(struct btrfs_root *root,
 int btrfs_wait_marked_extents(struct btrfs_root *root,
 			      struct extent_io_tree *dirty_pages, int mark)
 {
-	int ret;
 	int err = 0;
 	int werr = 0;
-	struct page *page;
-	struct inode *btree_inode = root->fs_info->btree_inode;
+	struct address_space *mapping = root->fs_info->btree_inode->i_mapping;
+	struct extent_state *cached_state = NULL;
 	u64 start = 0;
 	u64 end;
-	unsigned long index;
 
-	while (1) {
-		ret = find_first_extent_bit(dirty_pages, start, &start, &end,
-					    mark);
-		if (ret)
-			break;
-
-		clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS);
-		while (start <= end) {
-			index = start >> PAGE_CACHE_SHIFT;
-			start = (u64)(index + 1) << PAGE_CACHE_SHIFT;
-			page = find_get_page(btree_inode->i_mapping, index);
-			if (!page)
-				continue;
-			if (PageDirty(page)) {
-				btree_lock_page_hook(page);
-				wait_on_page_writeback(page);
-				err = write_one_page(page, 0);
-				if (err)
-					werr = err;
-			}
-			wait_on_page_writeback(page);
-			page_cache_release(page);
-			cond_resched();
-		}
+	while (!find_first_extent_bit(dirty_pages, start, &start, &end,
+				      EXTENT_NEED_WAIT, &cached_state)) {
+		clear_extent_bit(dirty_pages, start, end, EXTENT_NEED_WAIT,
+				 0, 0, &cached_state, GFP_NOFS);
+		err = filemap_fdatawait_range(mapping, start, end);
+		if (err)
+			werr = err;
+		cond_resched();
+		start = end + 1;
 	}
 	if (err)
 		werr = err;
@@ -458,15 +873,23 @@ int btrfs_wait_marked_extents(struct btrfs_root *root,
  * them in one of two extent_io trees.  This is used to make sure all of
  * those extents are on disk for transaction or log commit
  */
-int btrfs_write_and_wait_marked_extents(struct btrfs_root *root,
+static int btrfs_write_and_wait_marked_extents(struct btrfs_root *root,
 				struct extent_io_tree *dirty_pages, int mark)
 {
 	int ret;
 	int ret2;
+	struct blk_plug plug;
 
+	blk_start_plug(&plug);
 	ret = btrfs_write_marked_extents(root, dirty_pages, mark);
+	blk_finish_plug(&plug);
 	ret2 = btrfs_wait_marked_extents(root, dirty_pages, mark);
-	return ret || ret2;
+
+	if (ret)
+		return ret;
+	if (ret2)
+		return ret2;
+	return 0;
 }
 
 int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans,
@@ -513,21 +936,24 @@ static int update_cowonly_root(struct btrfs_trans_handle *trans,
 		ret = btrfs_update_root(trans, tree_root,
 					&root->root_key,
 					&root->root_item);
-		BUG_ON(ret);
+		if (ret)
+			return ret;
 
 		old_root_used = btrfs_root_used(&root->root_item);
 		ret = btrfs_write_dirty_block_groups(trans, root);
-		BUG_ON(ret);
+		if (ret)
+			return ret;
 	}
 
-	if (root != root->fs_info->extent_root)
-		switch_commit_root(root);
-
 	return 0;
 }
 
 /*
  * update all the cowonly tree roots on disk
+ *
+ * The error handling in this function may not be obvious. Any of the
+ * failures will cause the file system to go offline. We still need
+ * to clean up the delayed refs.
  */
 static noinline int commit_cowonly_roots(struct btrfs_trans_handle *trans,
 					 struct btrfs_root *root)
@@ -538,27 +964,53 @@ static noinline int commit_cowonly_roots(struct btrfs_trans_handle *trans,
 	int ret;
 
 	ret = btrfs_run_delayed_refs(trans, root, (unsigned long)-1);
-	BUG_ON(ret);
+	if (ret)
+		return ret;
 
 	eb = btrfs_lock_root_node(fs_info->tree_root);
-	btrfs_cow_block(trans, fs_info->tree_root, eb, NULL, 0, &eb);
+	ret = btrfs_cow_block(trans, fs_info->tree_root, eb, NULL,
+			      0, &eb);
 	btrfs_tree_unlock(eb);
 	free_extent_buffer(eb);
 
+	if (ret)
+		return ret;
+
 	ret = btrfs_run_delayed_refs(trans, root, (unsigned long)-1);
-	BUG_ON(ret);
+	if (ret)
+		return ret;
+
+	ret = btrfs_run_dev_stats(trans, root->fs_info);
+	if (ret)
+		return ret;
+	ret = btrfs_run_dev_replace(trans, root->fs_info);
+	if (ret)
+		return ret;
+	ret = btrfs_run_qgroups(trans, root->fs_info);
+	if (ret)
+		return ret;
+
+	/* run_qgroups might have added some more refs */
+	ret = btrfs_run_delayed_refs(trans, root, (unsigned long)-1);
+	if (ret)
+		return ret;
 
 	while (!list_empty(&fs_info->dirty_cowonly_roots)) {
 		next = fs_info->dirty_cowonly_roots.next;
 		list_del_init(next);
 		root = list_entry(next, struct btrfs_root, dirty_list);
 
-		update_cowonly_root(trans, root);
+		if (root != fs_info->extent_root)
+			list_add_tail(&root->dirty_list,
+				      &trans->transaction->switch_commits);
+		ret = update_cowonly_root(trans, root);
+		if (ret)
+			return ret;
 	}
 
-	down_write(&fs_info->extent_commit_sem);
-	switch_commit_root(fs_info->extent_root);
-	up_write(&fs_info->extent_commit_sem);
+	list_add_tail(&fs_info->extent_root->dirty_list,
+		      &trans->transaction->switch_commits);
+	btrfs_after_dev_replace_commit(fs_info);
 
 	return 0;
 }
@@ -568,12 +1020,12 @@ static noinline int commit_cowonly_roots(struct btrfs_trans_handle *trans,
  * a dirty root struct and adds it into the list of dead roots that need to
  * be deleted
  */
-int btrfs_add_dead_root(struct btrfs_root *root)
+void btrfs_add_dead_root(struct btrfs_root *root)
 {
-	mutex_lock(&root->fs_info->trans_mutex);
-	list_add(&root->root_list, &root->fs_info->dead_roots);
-	mutex_unlock(&root->fs_info->trans_mutex);
-	return 0;
+	spin_lock(&root->fs_info->trans_lock);
+	if (list_empty(&root->root_list))
+		list_add_tail(&root->root_list, &root->fs_info->dead_roots);
+	spin_unlock(&root->fs_info->trans_lock);
 }
 
 /*
@@ -588,6 +1040,7 @@ static noinline int commit_fs_roots(struct btrfs_trans_handle *trans,
 	int ret;
 	int err = 0;
 
+	spin_lock(&fs_info->fs_roots_radix_lock);
 	while (1) {
 		ret = radix_tree_gang_lookup_tag(&fs_info->fs_roots_radix,
 						 (void **)gang, 0,
@@ -600,12 +1053,21 @@ static noinline int commit_fs_roots(struct btrfs_trans_handle *trans,
 			radix_tree_tag_clear(&fs_info->fs_roots_radix,
 					(unsigned long)root->root_key.objectid,
 					BTRFS_ROOT_TRANS_TAG);
+			spin_unlock(&fs_info->fs_roots_radix_lock);
 
 			btrfs_free_log(trans, root);
 			btrfs_update_reloc_root(trans, root);
+			btrfs_orphan_commit_root(trans, root);
+
+			btrfs_save_ino_cache(root, trans);
+
+			/* see comments in should_cow_block() */
+			clear_bit(BTRFS_ROOT_FORCE_COW, &root->state);
+			smp_mb__after_atomic();
 
 			if (root->commit_root != root->node) {
-				switch_commit_root(root);
+				list_add_tail(&root->dirty_list,
+					&trans->transaction->switch_commits);
 				btrfs_set_root_node(&root->root_item,
 						    root->node);
 			}
@@ -613,140 +1075,60 @@ static noinline int commit_fs_roots(struct btrfs_trans_handle *trans,
 			err = btrfs_update_root(trans, fs_info->tree_root,
 						&root->root_key,
 						&root->root_item);
+			spin_lock(&fs_info->fs_roots_radix_lock);
 			if (err)
 				break;
 		}
 	}
+	spin_unlock(&fs_info->fs_roots_radix_lock);
 	return err;
 }
 
 /*
- * defrag a given btree.  If cacheonly == 1, this won't read from the disk,
- * otherwise every leaf in the btree is read and defragged.
+ * defrag a given btree.
+ * Every leaf in the btree is read and defragged.
  */
-int btrfs_defrag_root(struct btrfs_root *root, int cacheonly)
+int btrfs_defrag_root(struct btrfs_root *root)
 {
 	struct btrfs_fs_info *info = root->fs_info;
-	int ret;
 	struct btrfs_trans_handle *trans;
-	unsigned long nr;
+	int ret;
 
-	smp_mb();
-	if (root->defrag_running)
+	if (test_and_set_bit(BTRFS_ROOT_DEFRAG_RUNNING, &root->state))
 		return 0;
-	trans = btrfs_start_transaction(root, 1);
-	while (1) {
-		root->defrag_running = 1;
-		ret = btrfs_defrag_leaves(trans, root, cacheonly);
-		nr = trans->blocks_used;
-		btrfs_end_transaction(trans, root);
-		btrfs_btree_balance_dirty(info->tree_root, nr);
-		cond_resched();
-
-		trans = btrfs_start_transaction(root, 1);
-		if (root->fs_info->closing || ret != -EAGAIN)
-			break;
-	}
-	root->defrag_running = 0;
-	smp_mb();
-	btrfs_end_transaction(trans, root);
-	return 0;
-}
-
-#if 0
-/*
- * when dropping snapshots, we generate a ton of delayed refs, and it makes
- * sense not to join the transaction while it is trying to flush the current
- * queue of delayed refs out.
- *
- * This is used by the drop snapshot code only
- */
-static noinline int wait_transaction_pre_flush(struct btrfs_fs_info *info)
-{
-	DEFINE_WAIT(wait);
-
-	mutex_lock(&info->trans_mutex);
-	while (info->running_transaction &&
-	       info->running_transaction->delayed_refs.flushing) {
-		prepare_to_wait(&info->transaction_wait, &wait,
-				TASK_UNINTERRUPTIBLE);
-		mutex_unlock(&info->trans_mutex);
-
-		schedule();
-
-		mutex_lock(&info->trans_mutex);
-		finish_wait(&info->transaction_wait, &wait);
-	}
-	mutex_unlock(&info->trans_mutex);
-	return 0;
-}
-
-/*
- * Given a list of roots that need to be deleted, call btrfs_drop_snapshot on
- * all of them
- */
-int btrfs_drop_dead_root(struct btrfs_root *root)
-{
-	struct btrfs_trans_handle *trans;
-	struct btrfs_root *tree_root = root->fs_info->tree_root;
-	unsigned long nr;
-	int ret;
 
 	while (1) {
-		/*
-		 * we don't want to jump in and create a bunch of
-		 * delayed refs if the transaction is starting to close
-		 */
-		wait_transaction_pre_flush(tree_root->fs_info);
-		trans = btrfs_start_transaction(tree_root, 1);
+		trans = btrfs_start_transaction(root, 0);
+		if (IS_ERR(trans))
+			return PTR_ERR(trans);
 
-		/*
-		 * we've joined a transaction, make sure it isn't
-		 * closing right now
-		 */
-		if (trans->transaction->delayed_refs.flushing) {
-			btrfs_end_transaction(trans, tree_root);
-			continue;
-		}
+		ret = btrfs_defrag_leaves(trans, root);
 
-		ret = btrfs_drop_snapshot(trans, root);
-		if (ret != -EAGAIN)
-			break;
+		btrfs_end_transaction(trans, root);
+		btrfs_btree_balance_dirty(info->tree_root);
+		cond_resched();
 
-		ret = btrfs_update_root(trans, tree_root,
-					&root->root_key,
-					&root->root_item);
-		if (ret)
+		if (btrfs_fs_closing(root->fs_info) || ret != -EAGAIN)
 			break;
 
-		nr = trans->blocks_used;
-		ret = btrfs_end_transaction(trans, tree_root);
-		BUG_ON(ret);
-
-		btrfs_btree_balance_dirty(tree_root, nr);
-		cond_resched();
+		if (btrfs_defrag_cancelled(root->fs_info)) {
+			pr_debug("BTRFS: defrag_root cancelled\n");
+			ret = -EAGAIN;
+			break;
+		}
 	}
-	BUG_ON(ret);
-
-	ret = btrfs_del_root(trans, tree_root, &root->root_key);
-	BUG_ON(ret);
-
-	nr = trans->blocks_used;
-	ret = btrfs_end_transaction(trans, tree_root);
-	BUG_ON(ret);
-
-	free_extent_buffer(root->node);
-	free_extent_buffer(root->commit_root);
-	kfree(root);
-
-	btrfs_btree_balance_dirty(tree_root, nr);
+	clear_bit(BTRFS_ROOT_DEFRAG_RUNNING, &root->state);
 	return ret;
 }
-#endif
 
 /*
  * new snapshots need to be created at a very specific time in the
- * transaction commit.  This does the actual creation
+ * transaction commit.  This does the actual creation.
+ *
+ * Note:
+ * If the error which may affect the commitment of the current transaction
+ * happens, we should return the error number. If the error which just affect
+ * the creation of the pending snapshots, just return 0.
  */
 static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans,
 				   struct btrfs_fs_info *fs_info,
@@ -757,79 +1139,253 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans,
 	struct btrfs_root *tree_root = fs_info->tree_root;
 	struct btrfs_root *root = pending->root;
 	struct btrfs_root *parent_root;
+	struct btrfs_block_rsv *rsv;
 	struct inode *parent_inode;
+	struct btrfs_path *path;
+	struct btrfs_dir_item *dir_item;
+	struct dentry *dentry;
 	struct extent_buffer *tmp;
 	struct extent_buffer *old;
-	int ret;
-	u64 objectid;
-	int namelen;
+	struct timespec cur_time = CURRENT_TIME;
+	int ret = 0;
+	u64 to_reserve = 0;
 	u64 index = 0;
+	u64 objectid;
+	u64 root_flags;
+	uuid_le new_uuid;
 
-	parent_inode = pending->dentry->d_parent->d_inode;
-	parent_root = BTRFS_I(parent_inode)->root;
+	path = btrfs_alloc_path();
+	if (!path) {
+		pending->error = -ENOMEM;
+		return 0;
+	}
 
 	new_root_item = kmalloc(sizeof(*new_root_item), GFP_NOFS);
 	if (!new_root_item) {
-		ret = -ENOMEM;
-		goto fail;
+		pending->error = -ENOMEM;
+		goto root_item_alloc_fail;
+	}
+
+	pending->error = btrfs_find_free_objectid(tree_root, &objectid);
+	if (pending->error)
+		goto no_free_objectid;
+
+	btrfs_reloc_pre_snapshot(trans, pending, &to_reserve);
+
+	if (to_reserve > 0) {
+		pending->error = btrfs_block_rsv_add(root,
+						     &pending->block_rsv,
+						     to_reserve,
+						     BTRFS_RESERVE_NO_FLUSH);
+		if (pending->error)
+			goto no_free_objectid;
 	}
-	ret = btrfs_find_free_objectid(trans, tree_root, 0, &objectid);
-	if (ret)
-		goto fail;
 
 	key.objectid = objectid;
-	/* record when the snapshot was created in key.offset */
-	key.offset = trans->transid;
-	btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
+	key.offset = (u64)-1;
+	key.type = BTRFS_ROOT_ITEM_KEY;
 
-	memcpy(&pending->root_key, &key, sizeof(key));
-	pending->root_key.offset = (u64)-1;
+	rsv = trans->block_rsv;
+	trans->block_rsv = &pending->block_rsv;
+	trans->bytes_reserved = trans->block_rsv->reserved;
 
+	dentry = pending->dentry;
+	parent_inode = pending->dir;
+	parent_root = BTRFS_I(parent_inode)->root;
 	record_root_in_trans(trans, parent_root);
+
 	/*
 	 * insert the directory item
 	 */
-	namelen = strlen(pending->name);
 	ret = btrfs_set_inode_index(parent_inode, &index);
-	BUG_ON(ret);
-	ret = btrfs_insert_dir_item(trans, parent_root,
-			    pending->name, namelen,
-			    parent_inode->i_ino,
-			    &pending->root_key, BTRFS_FT_DIR, index);
-	BUG_ON(ret);
+	BUG_ON(ret); /* -ENOMEM */
+
+	/* check if there is a file/dir which has the same name. */
+	dir_item = btrfs_lookup_dir_item(NULL, parent_root, path,
+					 btrfs_ino(parent_inode),
+					 dentry->d_name.name,
+					 dentry->d_name.len, 0);
+	if (dir_item != NULL && !IS_ERR(dir_item)) {
+		pending->error = -EEXIST;
+		goto dir_item_existed;
+	} else if (IS_ERR(dir_item)) {
+		ret = PTR_ERR(dir_item);
+		btrfs_abort_transaction(trans, root, ret);
+		goto fail;
+	}
+	btrfs_release_path(path);
 
-	btrfs_i_size_write(parent_inode, parent_inode->i_size + namelen * 2);
-	ret = btrfs_update_inode(trans, parent_root, parent_inode);
-	BUG_ON(ret);
+	/*
+	 * pull in the delayed directory update
+	 * and the delayed inode item
+	 * otherwise we corrupt the FS during
+	 * snapshot
+	 */
+	ret = btrfs_run_delayed_items(trans, root);
+	if (ret) {	/* Transaction aborted */
+		btrfs_abort_transaction(trans, root, ret);
+		goto fail;
+	}
 
 	record_root_in_trans(trans, root);
 	btrfs_set_root_last_snapshot(&root->root_item, trans->transid);
 	memcpy(new_root_item, &root->root_item, sizeof(*new_root_item));
+	btrfs_check_and_init_root_item(new_root_item);
+
+	root_flags = btrfs_root_flags(new_root_item);
+	if (pending->readonly)
+		root_flags |= BTRFS_ROOT_SUBVOL_RDONLY;
+	else
+		root_flags &= ~BTRFS_ROOT_SUBVOL_RDONLY;
+	btrfs_set_root_flags(new_root_item, root_flags);
+
+	btrfs_set_root_generation_v2(new_root_item,
+			trans->transid);
+	uuid_le_gen(&new_uuid);
+	memcpy(new_root_item->uuid, new_uuid.b, BTRFS_UUID_SIZE);
+	memcpy(new_root_item->parent_uuid, root->root_item.uuid,
+			BTRFS_UUID_SIZE);
+	if (!(root_flags & BTRFS_ROOT_SUBVOL_RDONLY)) {
+		memset(new_root_item->received_uuid, 0,
+		       sizeof(new_root_item->received_uuid));
+		memset(&new_root_item->stime, 0, sizeof(new_root_item->stime));
+		memset(&new_root_item->rtime, 0, sizeof(new_root_item->rtime));
+		btrfs_set_root_stransid(new_root_item, 0);
+		btrfs_set_root_rtransid(new_root_item, 0);
+	}
+	btrfs_set_stack_timespec_sec(&new_root_item->otime, cur_time.tv_sec);
+	btrfs_set_stack_timespec_nsec(&new_root_item->otime, cur_time.tv_nsec);
+	btrfs_set_root_otransid(new_root_item, trans->transid);
 
 	old = btrfs_lock_root_node(root);
-	btrfs_cow_block(trans, root, old, NULL, 0, &old);
+	ret = btrfs_cow_block(trans, root, old, NULL, 0, &old);
+	if (ret) {
+		btrfs_tree_unlock(old);
+		free_extent_buffer(old);
+		btrfs_abort_transaction(trans, root, ret);
+		goto fail;
+	}
+
 	btrfs_set_lock_blocking(old);
 
-	btrfs_copy_root(trans, root, old, &tmp, objectid);
+	ret = btrfs_copy_root(trans, root, old, &tmp, objectid);
+	/* clean up in any case */
 	btrfs_tree_unlock(old);
 	free_extent_buffer(old);
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
+		goto fail;
+	}
+
+	/*
+	 * We need to flush delayed refs in order to make sure all of our quota
+	 * operations have been done before we call btrfs_qgroup_inherit.
+	 */
+	ret = btrfs_run_delayed_refs(trans, root, (unsigned long)-1);
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
+		goto fail;
+	}
+
+	ret = btrfs_qgroup_inherit(trans, fs_info,
+				   root->root_key.objectid,
+				   objectid, pending->inherit);
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
+		goto fail;
+	}
+
+	/* see comments in should_cow_block() */
+	set_bit(BTRFS_ROOT_FORCE_COW, &root->state);
+	smp_wmb();
 
 	btrfs_set_root_node(new_root_item, tmp);
-	ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
-				new_root_item);
-	BUG_ON(ret);
+	/* record when the snapshot was created in key.offset */
+	key.offset = trans->transid;
+	ret = btrfs_insert_root(trans, tree_root, &key, new_root_item);
 	btrfs_tree_unlock(tmp);
 	free_extent_buffer(tmp);
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
+		goto fail;
+	}
 
-	ret = btrfs_add_root_ref(trans, parent_root->fs_info->tree_root,
-				 pending->root_key.objectid,
+	/*
+	 * insert root back/forward references
+	 */
+	ret = btrfs_add_root_ref(trans, tree_root, objectid,
 				 parent_root->root_key.objectid,
-				 parent_inode->i_ino, index, pending->name,
-				 namelen);
-	BUG_ON(ret);
+				 btrfs_ino(parent_inode), index,
+				 dentry->d_name.name, dentry->d_name.len);
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
+		goto fail;
+	}
+
+	key.offset = (u64)-1;
+	pending->snap = btrfs_read_fs_root_no_name(root->fs_info, &key);
+	if (IS_ERR(pending->snap)) {
+		ret = PTR_ERR(pending->snap);
+		btrfs_abort_transaction(trans, root, ret);
+		goto fail;
+	}
+
+	ret = btrfs_reloc_post_snapshot(trans, pending);
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
+		goto fail;
+	}
+
+	ret = btrfs_run_delayed_refs(trans, root, (unsigned long)-1);
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
+		goto fail;
+	}
+
+	ret = btrfs_insert_dir_item(trans, parent_root,
+				    dentry->d_name.name, dentry->d_name.len,
+				    parent_inode, &key,
+				    BTRFS_FT_DIR, index);
+	/* We have check then name at the beginning, so it is impossible. */
+	BUG_ON(ret == -EEXIST || ret == -EOVERFLOW);
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
+		goto fail;
+	}
 
+	btrfs_i_size_write(parent_inode, parent_inode->i_size +
+					 dentry->d_name.len * 2);
+	parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME;
+	ret = btrfs_update_inode_fallback(trans, parent_root, parent_inode);
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
+		goto fail;
+	}
+	ret = btrfs_uuid_tree_add(trans, fs_info->uuid_root, new_uuid.b,
+				  BTRFS_UUID_KEY_SUBVOL, objectid);
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
+		goto fail;
+	}
+	if (!btrfs_is_empty_uuid(new_root_item->received_uuid)) {
+		ret = btrfs_uuid_tree_add(trans, fs_info->uuid_root,
+					  new_root_item->received_uuid,
+					  BTRFS_UUID_KEY_RECEIVED_SUBVOL,
+					  objectid);
+		if (ret && ret != -EEXIST) {
+			btrfs_abort_transaction(trans, root, ret);
+			goto fail;
+		}
+	}
 fail:
+	pending->error = ret;
+dir_item_existed:
+	trans->block_rsv = rsv;
+	trans->bytes_reserved = 0;
+no_free_objectid:
 	kfree(new_root_item);
+root_item_alloc_fail:
+	btrfs_free_path(path);
 	return ret;
 }
 
@@ -839,15 +1395,17 @@ fail:
 static noinline int create_pending_snapshots(struct btrfs_trans_handle *trans,
 					     struct btrfs_fs_info *fs_info)
 {
-	struct btrfs_pending_snapshot *pending;
+	struct btrfs_pending_snapshot *pending, *next;
 	struct list_head *head = &trans->transaction->pending_snapshots;
-	int ret;
+	int ret = 0;
 
-	list_for_each_entry(pending, head, list) {
+	list_for_each_entry_safe(pending, next, head, list) {
+		list_del(&pending->list);
 		ret = create_pending_snapshot(trans, fs_info, pending);
-		BUG_ON(ret);
+		if (ret)
+			break;
 	}
-	return 0;
+	return ret;
 }
 
 static void update_super_roots(struct btrfs_root *root)
@@ -855,7 +1413,7 @@ static void update_super_roots(struct btrfs_root *root)
 	struct btrfs_root_item *root_item;
 	struct btrfs_super_block *super;
 
-	super = &root->fs_info->super_copy;
+	super = root->fs_info->super_copy;
 
 	root_item = &root->fs_info->chunk_root->root_item;
 	super->chunk_root = root_item->bytenr;
@@ -866,130 +1424,398 @@ static void update_super_roots(struct btrfs_root *root)
 	super->root = root_item->bytenr;
 	super->generation = root_item->generation;
 	super->root_level = root_item->level;
+	if (btrfs_test_opt(root, SPACE_CACHE))
+		super->cache_generation = root_item->generation;
+	if (root->fs_info->update_uuid_tree_gen)
+		super->uuid_tree_generation = root_item->generation;
 }
 
 int btrfs_transaction_in_commit(struct btrfs_fs_info *info)
 {
+	struct btrfs_transaction *trans;
+	int ret = 0;
+
+	spin_lock(&info->trans_lock);
+	trans = info->running_transaction;
+	if (trans)
+		ret = (trans->state >= TRANS_STATE_COMMIT_START);
+	spin_unlock(&info->trans_lock);
+	return ret;
+}
+
+int btrfs_transaction_blocked(struct btrfs_fs_info *info)
+{
+	struct btrfs_transaction *trans;
 	int ret = 0;
-	spin_lock(&info->new_trans_lock);
-	if (info->running_transaction)
-		ret = info->running_transaction->in_commit;
-	spin_unlock(&info->new_trans_lock);
+
+	spin_lock(&info->trans_lock);
+	trans = info->running_transaction;
+	if (trans)
+		ret = is_transaction_blocked(trans);
+	spin_unlock(&info->trans_lock);
 	return ret;
 }
 
+/*
+ * wait for the current transaction commit to start and block subsequent
+ * transaction joins
+ */
+static void wait_current_trans_commit_start(struct btrfs_root *root,
+					    struct btrfs_transaction *trans)
+{
+	wait_event(root->fs_info->transaction_blocked_wait,
+		   trans->state >= TRANS_STATE_COMMIT_START ||
+		   trans->aborted);
+}
+
+/*
+ * wait for the current transaction to start and then become unblocked.
+ * caller holds ref.
+ */
+static void wait_current_trans_commit_start_and_unblock(struct btrfs_root *root,
+					 struct btrfs_transaction *trans)
+{
+	wait_event(root->fs_info->transaction_wait,
+		   trans->state >= TRANS_STATE_UNBLOCKED ||
+		   trans->aborted);
+}
+
+/*
+ * commit transactions asynchronously. once btrfs_commit_transaction_async
+ * returns, any subsequent transaction will not be allowed to join.
+ */
+struct btrfs_async_commit {
+	struct btrfs_trans_handle *newtrans;
+	struct btrfs_root *root;
+	struct work_struct work;
+};
+
+static void do_async_commit(struct work_struct *work)
+{
+	struct btrfs_async_commit *ac =
+		container_of(work, struct btrfs_async_commit, work);
+
+	/*
+	 * We've got freeze protection passed with the transaction.
+	 * Tell lockdep about it.
+	 */
+	if (ac->newtrans->type & __TRANS_FREEZABLE)
+		rwsem_acquire_read(
+		     &ac->root->fs_info->sb->s_writers.lock_map[SB_FREEZE_FS-1],
+		     0, 1, _THIS_IP_);
+
+	current->journal_info = ac->newtrans;
+
+	btrfs_commit_transaction(ac->newtrans, ac->root);
+	kfree(ac);
+}
+
+int btrfs_commit_transaction_async(struct btrfs_trans_handle *trans,
+				   struct btrfs_root *root,
+				   int wait_for_unblock)
+{
+	struct btrfs_async_commit *ac;
+	struct btrfs_transaction *cur_trans;
+
+	ac = kmalloc(sizeof(*ac), GFP_NOFS);
+	if (!ac)
+		return -ENOMEM;
+
+	INIT_WORK(&ac->work, do_async_commit);
+	ac->root = root;
+	ac->newtrans = btrfs_join_transaction(root);
+	if (IS_ERR(ac->newtrans)) {
+		int err = PTR_ERR(ac->newtrans);
+		kfree(ac);
+		return err;
+	}
+
+	/* take transaction reference */
+	cur_trans = trans->transaction;
+	atomic_inc(&cur_trans->use_count);
+
+	btrfs_end_transaction(trans, root);
+
+	/*
+	 * Tell lockdep we've released the freeze rwsem, since the
+	 * async commit thread will be the one to unlock it.
+	 */
+	if (ac->newtrans->type & __TRANS_FREEZABLE)
+		rwsem_release(
+			&root->fs_info->sb->s_writers.lock_map[SB_FREEZE_FS-1],
+			1, _THIS_IP_);
+
+	schedule_work(&ac->work);
+
+	/* wait for transaction to start and unblock */
+	if (wait_for_unblock)
+		wait_current_trans_commit_start_and_unblock(root, cur_trans);
+	else
+		wait_current_trans_commit_start(root, cur_trans);
+
+	if (current->journal_info == trans)
+		current->journal_info = NULL;
+
+	btrfs_put_transaction(cur_trans);
+	return 0;
+}
+
+
+static void cleanup_transaction(struct btrfs_trans_handle *trans,
+				struct btrfs_root *root, int err)
+{
+	struct btrfs_transaction *cur_trans = trans->transaction;
+	DEFINE_WAIT(wait);
+
+	WARN_ON(trans->use_count > 1);
+
+	btrfs_abort_transaction(trans, root, err);
+
+	spin_lock(&root->fs_info->trans_lock);
+
+	/*
+	 * If the transaction is removed from the list, it means this
+	 * transaction has been committed successfully, so it is impossible
+	 * to call the cleanup function.
+	 */
+	BUG_ON(list_empty(&cur_trans->list));
+
+	list_del_init(&cur_trans->list);
+	if (cur_trans == root->fs_info->running_transaction) {
+		cur_trans->state = TRANS_STATE_COMMIT_DOING;
+		spin_unlock(&root->fs_info->trans_lock);
+		wait_event(cur_trans->writer_wait,
+			   atomic_read(&cur_trans->num_writers) == 1);
+
+		spin_lock(&root->fs_info->trans_lock);
+	}
+	spin_unlock(&root->fs_info->trans_lock);
+
+	btrfs_cleanup_one_transaction(trans->transaction, root);
+
+	spin_lock(&root->fs_info->trans_lock);
+	if (cur_trans == root->fs_info->running_transaction)
+		root->fs_info->running_transaction = NULL;
+	spin_unlock(&root->fs_info->trans_lock);
+
+	if (trans->type & __TRANS_FREEZABLE)
+		sb_end_intwrite(root->fs_info->sb);
+	btrfs_put_transaction(cur_trans);
+	btrfs_put_transaction(cur_trans);
+
+	trace_btrfs_transaction_commit(root);
+
+	if (current->journal_info == trans)
+		current->journal_info = NULL;
+	btrfs_scrub_cancel(root->fs_info);
+
+	kmem_cache_free(btrfs_trans_handle_cachep, trans);
+}
+
+static int btrfs_flush_all_pending_stuffs(struct btrfs_trans_handle *trans,
+					  struct btrfs_root *root)
+{
+	int ret;
+
+	ret = btrfs_run_delayed_items(trans, root);
+	if (ret)
+		return ret;
+
+	/*
+	 * rename don't use btrfs_join_transaction, so, once we
+	 * set the transaction to blocked above, we aren't going
+	 * to get any new ordered operations.  We can safely run
+	 * it here and no for sure that nothing new will be added
+	 * to the list
+	 */
+	ret = btrfs_run_ordered_operations(trans, root, 1);
+
+	return ret;
+}
+
+static inline int btrfs_start_delalloc_flush(struct btrfs_fs_info *fs_info)
+{
+	if (btrfs_test_opt(fs_info->tree_root, FLUSHONCOMMIT))
+		return btrfs_start_delalloc_roots(fs_info, 1, -1);
+	return 0;
+}
+
+static inline void btrfs_wait_delalloc_flush(struct btrfs_fs_info *fs_info)
+{
+	if (btrfs_test_opt(fs_info->tree_root, FLUSHONCOMMIT))
+		btrfs_wait_ordered_roots(fs_info, -1);
+}
+
 int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 			     struct btrfs_root *root)
 {
-	unsigned long joined = 0;
-	unsigned long timeout = 1;
-	struct btrfs_transaction *cur_trans;
+	struct btrfs_transaction *cur_trans = trans->transaction;
 	struct btrfs_transaction *prev_trans = NULL;
-	DEFINE_WAIT(wait);
 	int ret;
-	int should_grow = 0;
-	unsigned long now = get_seconds();
-	int flush_on_commit = btrfs_test_opt(root, FLUSHONCOMMIT);
 
-	btrfs_run_ordered_operations(root, 0);
+	ret = btrfs_run_ordered_operations(trans, root, 0);
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
+		btrfs_end_transaction(trans, root);
+		return ret;
+	}
+
+	/* Stop the commit early if ->aborted is set */
+	if (unlikely(ACCESS_ONCE(cur_trans->aborted))) {
+		ret = cur_trans->aborted;
+		btrfs_end_transaction(trans, root);
+		return ret;
+	}
 
 	/* make a pass through all the delayed refs we have so far
 	 * any runnings procs may add more while we are here
 	 */
 	ret = btrfs_run_delayed_refs(trans, root, 0);
-	BUG_ON(ret);
+	if (ret) {
+		btrfs_end_transaction(trans, root);
+		return ret;
+	}
+
+	btrfs_trans_release_metadata(trans, root);
+	trans->block_rsv = NULL;
+	if (trans->qgroup_reserved) {
+		btrfs_qgroup_free(root, trans->qgroup_reserved);
+		trans->qgroup_reserved = 0;
+	}
 
 	cur_trans = trans->transaction;
+
 	/*
 	 * set the flushing flag so procs in this transaction have to
 	 * start sending their work down.
 	 */
 	cur_trans->delayed_refs.flushing = 1;
+	smp_wmb();
 
-	ret = btrfs_run_delayed_refs(trans, root, 0);
-	BUG_ON(ret);
+	if (!list_empty(&trans->new_bgs))
+		btrfs_create_pending_block_groups(trans, root);
 
-	mutex_lock(&root->fs_info->trans_mutex);
-	if (cur_trans->in_commit) {
-		cur_trans->use_count++;
-		mutex_unlock(&root->fs_info->trans_mutex);
+	ret = btrfs_run_delayed_refs(trans, root, 0);
+	if (ret) {
 		btrfs_end_transaction(trans, root);
+		return ret;
+	}
 
-		ret = wait_for_commit(root, cur_trans);
-		BUG_ON(ret);
+	spin_lock(&root->fs_info->trans_lock);
+	if (cur_trans->state >= TRANS_STATE_COMMIT_START) {
+		spin_unlock(&root->fs_info->trans_lock);
+		atomic_inc(&cur_trans->use_count);
+		ret = btrfs_end_transaction(trans, root);
 
-		mutex_lock(&root->fs_info->trans_mutex);
-		put_transaction(cur_trans);
-		mutex_unlock(&root->fs_info->trans_mutex);
+		wait_for_commit(root, cur_trans);
 
-		return 0;
+		btrfs_put_transaction(cur_trans);
+
+		return ret;
 	}
 
-	trans->transaction->in_commit = 1;
-	trans->transaction->blocked = 1;
+	cur_trans->state = TRANS_STATE_COMMIT_START;
+	wake_up(&root->fs_info->transaction_blocked_wait);
+
 	if (cur_trans->list.prev != &root->fs_info->trans_list) {
 		prev_trans = list_entry(cur_trans->list.prev,
 					struct btrfs_transaction, list);
-		if (!prev_trans->commit_done) {
-			prev_trans->use_count++;
-			mutex_unlock(&root->fs_info->trans_mutex);
+		if (prev_trans->state != TRANS_STATE_COMPLETED) {
+			atomic_inc(&prev_trans->use_count);
+			spin_unlock(&root->fs_info->trans_lock);
 
 			wait_for_commit(root, prev_trans);
 
-			mutex_lock(&root->fs_info->trans_mutex);
-			put_transaction(prev_trans);
+			btrfs_put_transaction(prev_trans);
+		} else {
+			spin_unlock(&root->fs_info->trans_lock);
 		}
+	} else {
+		spin_unlock(&root->fs_info->trans_lock);
 	}
 
-	if (now < cur_trans->start_time || now - cur_trans->start_time < 1)
-		should_grow = 1;
+	extwriter_counter_dec(cur_trans, trans->type);
 
-	do {
-		int snap_pending = 0;
-		joined = cur_trans->num_joined;
-		if (!list_empty(&trans->transaction->pending_snapshots))
-			snap_pending = 1;
-
-		WARN_ON(cur_trans != trans->transaction);
-		prepare_to_wait(&cur_trans->writer_wait, &wait,
-				TASK_UNINTERRUPTIBLE);
-
-		if (cur_trans->num_writers > 1)
-			timeout = MAX_SCHEDULE_TIMEOUT;
-		else if (should_grow)
-			timeout = 1;
-
-		mutex_unlock(&root->fs_info->trans_mutex);
-
-		if (flush_on_commit || snap_pending) {
-			btrfs_start_delalloc_inodes(root, 1);
-			ret = btrfs_wait_ordered_extents(root, 0, 1);
-			BUG_ON(ret);
-		}
+	ret = btrfs_start_delalloc_flush(root->fs_info);
+	if (ret)
+		goto cleanup_transaction;
 
-		/*
-		 * rename don't use btrfs_join_transaction, so, once we
-		 * set the transaction to blocked above, we aren't going
-		 * to get any new ordered operations.  We can safely run
-		 * it here and no for sure that nothing new will be added
-		 * to the list
-		 */
-		btrfs_run_ordered_operations(root, 1);
+	ret = btrfs_flush_all_pending_stuffs(trans, root);
+	if (ret)
+		goto cleanup_transaction;
 
-		smp_mb();
-		if (cur_trans->num_writers > 1 || should_grow)
-			schedule_timeout(timeout);
+	wait_event(cur_trans->writer_wait,
+		   extwriter_counter_read(cur_trans) == 0);
 
-		mutex_lock(&root->fs_info->trans_mutex);
-		finish_wait(&cur_trans->writer_wait, &wait);
-	} while (cur_trans->num_writers > 1 ||
-		 (should_grow && cur_trans->num_joined != joined));
+	/* some pending stuffs might be added after the previous flush. */
+	ret = btrfs_flush_all_pending_stuffs(trans, root);
+	if (ret)
+		goto cleanup_transaction;
+
+	btrfs_wait_delalloc_flush(root->fs_info);
+
+	btrfs_scrub_pause(root);
+	/*
+	 * Ok now we need to make sure to block out any other joins while we
+	 * commit the transaction.  We could have started a join before setting
+	 * COMMIT_DOING so make sure to wait for num_writers to == 1 again.
+	 */
+	spin_lock(&root->fs_info->trans_lock);
+	cur_trans->state = TRANS_STATE_COMMIT_DOING;
+	spin_unlock(&root->fs_info->trans_lock);
+	wait_event(cur_trans->writer_wait,
+		   atomic_read(&cur_trans->num_writers) == 1);
+
+	/* ->aborted might be set after the previous check, so check it */
+	if (unlikely(ACCESS_ONCE(cur_trans->aborted))) {
+		ret = cur_trans->aborted;
+		goto scrub_continue;
+	}
+	/*
+	 * the reloc mutex makes sure that we stop
+	 * the balancing code from coming in and moving
+	 * extents around in the middle of the commit
+	 */
+	mutex_lock(&root->fs_info->reloc_mutex);
 
+	/*
+	 * We needn't worry about the delayed items because we will
+	 * deal with them in create_pending_snapshot(), which is the
+	 * core function of the snapshot creation.
+	 */
 	ret = create_pending_snapshots(trans, root->fs_info);
-	BUG_ON(ret);
+	if (ret) {
+		mutex_unlock(&root->fs_info->reloc_mutex);
+		goto scrub_continue;
+	}
+
+	/*
+	 * We insert the dir indexes of the snapshots and update the inode
+	 * of the snapshots' parents after the snapshot creation, so there
+	 * are some delayed items which are not dealt with. Now deal with
+	 * them.
+	 *
+	 * We needn't worry that this operation will corrupt the snapshots,
+	 * because all the tree which are snapshoted will be forced to COW
+	 * the nodes and leaves.
+	 */
+	ret = btrfs_run_delayed_items(trans, root);
+	if (ret) {
+		mutex_unlock(&root->fs_info->reloc_mutex);
+		goto scrub_continue;
+	}
 
 	ret = btrfs_run_delayed_refs(trans, root, (unsigned long)-1);
-	BUG_ON(ret);
+	if (ret) {
+		mutex_unlock(&root->fs_info->reloc_mutex);
+		goto scrub_continue;
+	}
+
+	/*
+	 * make sure none of the code above managed to slip in a
+	 * delayed item
+	 */
+	btrfs_assert_delayed_root_empty(root);
 
 	WARN_ON(cur_trans != trans->transaction);
 
@@ -1009,7 +1835,20 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 	mutex_lock(&root->fs_info->tree_log_mutex);
 
 	ret = commit_fs_roots(trans, root);
-	BUG_ON(ret);
+	if (ret) {
+		mutex_unlock(&root->fs_info->tree_log_mutex);
+		mutex_unlock(&root->fs_info->reloc_mutex);
+		goto scrub_continue;
+	}
+
+	/*
+	 * Since the transaction is done, we should set the inode map cache flag
+	 * before any other comming transaction.
+	 */
+	if (btrfs_test_opt(root, CHANGE_INODE_CACHE))
+		btrfs_set_opt(root->fs_info->mount_opt, INODE_MAP_CACHE);
+	else
+		btrfs_clear_opt(root->fs_info->mount_opt, INODE_MAP_CACHE);
 
 	/* commit_fs_roots gets rid of all the tree log roots, it is now
 	 * safe to free the root of tree log roots
@@ -1017,41 +1856,68 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 	btrfs_free_log_root_tree(trans, root->fs_info);
 
 	ret = commit_cowonly_roots(trans, root);
-	BUG_ON(ret);
+	if (ret) {
+		mutex_unlock(&root->fs_info->tree_log_mutex);
+		mutex_unlock(&root->fs_info->reloc_mutex);
+		goto scrub_continue;
+	}
+
+	/*
+	 * The tasks which save the space cache and inode cache may also
+	 * update ->aborted, check it.
+	 */
+	if (unlikely(ACCESS_ONCE(cur_trans->aborted))) {
+		ret = cur_trans->aborted;
+		mutex_unlock(&root->fs_info->tree_log_mutex);
+		mutex_unlock(&root->fs_info->reloc_mutex);
+		goto scrub_continue;
+	}
 
 	btrfs_prepare_extent_commit(trans, root);
 
 	cur_trans = root->fs_info->running_transaction;
-	spin_lock(&root->fs_info->new_trans_lock);
-	root->fs_info->running_transaction = NULL;
-	spin_unlock(&root->fs_info->new_trans_lock);
 
 	btrfs_set_root_node(&root->fs_info->tree_root->root_item,
 			    root->fs_info->tree_root->node);
-	switch_commit_root(root->fs_info->tree_root);
+	list_add_tail(&root->fs_info->tree_root->dirty_list,
+		      &cur_trans->switch_commits);
 
 	btrfs_set_root_node(&root->fs_info->chunk_root->root_item,
 			    root->fs_info->chunk_root->node);
-	switch_commit_root(root->fs_info->chunk_root);
+	list_add_tail(&root->fs_info->chunk_root->dirty_list,
+		      &cur_trans->switch_commits);
 
-	update_super_roots(root);
+	switch_commit_roots(cur_trans, root->fs_info);
 
-	if (!root->fs_info->log_root_recovering) {
-		btrfs_set_super_log_root(&root->fs_info->super_copy, 0);
-		btrfs_set_super_log_root_level(&root->fs_info->super_copy, 0);
-	}
+	assert_qgroups_uptodate(trans);
+	update_super_roots(root);
 
-	memcpy(&root->fs_info->super_for_commit, &root->fs_info->super_copy,
-	       sizeof(root->fs_info->super_copy));
+	btrfs_set_super_log_root(root->fs_info->super_copy, 0);
+	btrfs_set_super_log_root_level(root->fs_info->super_copy, 0);
+	memcpy(root->fs_info->super_for_commit, root->fs_info->super_copy,
+	       sizeof(*root->fs_info->super_copy));
 
-	trans->transaction->blocked = 0;
+	spin_lock(&root->fs_info->trans_lock);
+	cur_trans->state = TRANS_STATE_UNBLOCKED;
+	root->fs_info->running_transaction = NULL;
+	spin_unlock(&root->fs_info->trans_lock);
+	mutex_unlock(&root->fs_info->reloc_mutex);
 
 	wake_up(&root->fs_info->transaction_wait);
 
-	mutex_unlock(&root->fs_info->trans_mutex);
 	ret = btrfs_write_and_wait_transaction(trans, root);
-	BUG_ON(ret);
-	write_ctree_super(trans, root, 0);
+	if (ret) {
+		btrfs_error(root->fs_info, ret,
+			    "Error while writing out transaction");
+		mutex_unlock(&root->fs_info->tree_log_mutex);
+		goto scrub_continue;
+	}
+
+	ret = write_ctree_super(trans, root, 0);
+	if (ret) {
+		mutex_unlock(&root->fs_info->tree_log_mutex);
+		goto scrub_continue;
+	}
 
 	/*
 	 * the super is written, we can safely allow the tree-loggers
@@ -1061,18 +1927,27 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 
 	btrfs_finish_extent_commit(trans, root);
 
-	mutex_lock(&root->fs_info->trans_mutex);
+	root->fs_info->last_trans_committed = cur_trans->transid;
+	/*
+	 * We needn't acquire the lock here because there is no other task
+	 * which can change it.
+	 */
+	cur_trans->state = TRANS_STATE_COMPLETED;
+	wake_up(&cur_trans->commit_wait);
 
-	cur_trans->commit_done = 1;
+	spin_lock(&root->fs_info->trans_lock);
+	list_del_init(&cur_trans->list);
+	spin_unlock(&root->fs_info->trans_lock);
 
-	root->fs_info->last_trans_committed = cur_trans->transid;
+	btrfs_put_transaction(cur_trans);
+	btrfs_put_transaction(cur_trans);
 
-	wake_up(&cur_trans->commit_wait);
+	if (trans->type & __TRANS_FREEZABLE)
+		sb_end_intwrite(root->fs_info->sb);
 
-	put_transaction(cur_trans);
-	put_transaction(cur_trans);
+	trace_btrfs_transaction_commit(root);
 
-	mutex_unlock(&root->fs_info->trans_mutex);
+	btrfs_scrub_continue(root);
 
 	if (current->journal_info == trans)
 		current->journal_info = NULL;
@@ -1083,29 +1958,61 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 		btrfs_run_delayed_iputs(root);
 
 	return ret;
+
+scrub_continue:
+	btrfs_scrub_continue(root);
+cleanup_transaction:
+	btrfs_trans_release_metadata(trans, root);
+	trans->block_rsv = NULL;
+	if (trans->qgroup_reserved) {
+		btrfs_qgroup_free(root, trans->qgroup_reserved);
+		trans->qgroup_reserved = 0;
+	}
+	btrfs_warn(root->fs_info, "Skipping commit of aborted transaction.");
+	if (current->journal_info == trans)
+		current->journal_info = NULL;
+	cleanup_transaction(trans, root, ret);
+
+	return ret;
 }
 
 /*
- * interface function to delete all the snapshots we have scheduled for deletion
+ * return < 0 if error
+ * 0 if there are no more dead_roots at the time of call
+ * 1 there are more to be processed, call me again
+ *
+ * The return value indicates there are certainly more snapshots to delete, but
+ * if there comes a new one during processing, it may return 0. We don't mind,
+ * because btrfs_commit_super will poke cleaner thread and it will process it a
+ * few seconds later.
  */
-int btrfs_clean_old_snapshots(struct btrfs_root *root)
+int btrfs_clean_one_deleted_snapshot(struct btrfs_root *root)
 {
-	LIST_HEAD(list);
+	int ret;
 	struct btrfs_fs_info *fs_info = root->fs_info;
 
-	mutex_lock(&fs_info->trans_mutex);
-	list_splice_init(&fs_info->dead_roots, &list);
-	mutex_unlock(&fs_info->trans_mutex);
+	spin_lock(&fs_info->trans_lock);
+	if (list_empty(&fs_info->dead_roots)) {
+		spin_unlock(&fs_info->trans_lock);
+		return 0;
+	}
+	root = list_first_entry(&fs_info->dead_roots,
+			struct btrfs_root, root_list);
+	list_del_init(&root->root_list);
+	spin_unlock(&fs_info->trans_lock);
 
-	while (!list_empty(&list)) {
-		root = list_entry(list.next, struct btrfs_root, root_list);
-		list_del(&root->root_list);
+	pr_debug("BTRFS: cleaner removing %llu\n", root->objectid);
 
-		if (btrfs_header_backref_rev(root->node) <
-		    BTRFS_MIXED_BACKREF_REV)
-			btrfs_drop_snapshot(root, 0);
-		else
-			btrfs_drop_snapshot(root, 1);
-	}
-	return 0;
+	btrfs_kill_all_delayed_nodes(root);
+
+	if (btrfs_header_backref_rev(root->node) <
+			BTRFS_MIXED_BACKREF_REV)
+		ret = btrfs_drop_snapshot(root, NULL, 0, 0);
+	else
+		ret = btrfs_drop_snapshot(root, NULL, 1, 0);
+	/*
+	 * If we encounter a transaction abort during snapshot cleaning, we
+	 * don't want to crash here
+	 */
+	return (ret < 0) ? 0 : 1;
 }
diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h
index 93c7ccb3311..7dd558ed071 100644
--- a/fs/btrfs/transaction.h
+++ b/fs/btrfs/transaction.h
@@ -20,97 +20,155 @@
 #define __BTRFS_TRANSACTION__
 #include "btrfs_inode.h"
 #include "delayed-ref.h"
+#include "ctree.h"
+
+enum btrfs_trans_state {
+	TRANS_STATE_RUNNING		= 0,
+	TRANS_STATE_BLOCKED		= 1,
+	TRANS_STATE_COMMIT_START	= 2,
+	TRANS_STATE_COMMIT_DOING	= 3,
+	TRANS_STATE_UNBLOCKED		= 4,
+	TRANS_STATE_COMPLETED		= 5,
+	TRANS_STATE_MAX			= 6,
+};
 
 struct btrfs_transaction {
 	u64 transid;
 	/*
+	 * total external writers(USERSPACE/START/ATTACH) in this
+	 * transaction, it must be zero before the transaction is
+	 * being committed
+	 */
+	atomic_t num_extwriters;
+	/*
 	 * total writers in this transaction, it must be zero before the
 	 * transaction can end
 	 */
-	unsigned long num_writers;
+	atomic_t num_writers;
+	atomic_t use_count;
 
-	unsigned long num_joined;
-	int in_commit;
-	int use_count;
-	int commit_done;
-	int blocked;
+	/* Be protected by fs_info->trans_lock when we want to change it. */
+	enum btrfs_trans_state state;
 	struct list_head list;
 	struct extent_io_tree dirty_pages;
 	unsigned long start_time;
 	wait_queue_head_t writer_wait;
 	wait_queue_head_t commit_wait;
 	struct list_head pending_snapshots;
+	struct list_head ordered_operations;
+	struct list_head pending_chunks;
+	struct list_head switch_commits;
 	struct btrfs_delayed_ref_root delayed_refs;
+	int aborted;
 };
 
+#define __TRANS_FREEZABLE	(1U << 0)
+
+#define __TRANS_USERSPACE	(1U << 8)
+#define __TRANS_START		(1U << 9)
+#define __TRANS_ATTACH		(1U << 10)
+#define __TRANS_JOIN		(1U << 11)
+#define __TRANS_JOIN_NOLOCK	(1U << 12)
+#define __TRANS_DUMMY		(1U << 13)
+
+#define TRANS_USERSPACE		(__TRANS_USERSPACE | __TRANS_FREEZABLE)
+#define TRANS_START		(__TRANS_START | __TRANS_FREEZABLE)
+#define TRANS_ATTACH		(__TRANS_ATTACH)
+#define TRANS_JOIN		(__TRANS_JOIN | __TRANS_FREEZABLE)
+#define TRANS_JOIN_NOLOCK	(__TRANS_JOIN_NOLOCK)
+
+#define TRANS_EXTWRITERS	(__TRANS_USERSPACE | __TRANS_START |	\
+				 __TRANS_ATTACH)
+
+#define BTRFS_SEND_TRANS_STUB	1
+
 struct btrfs_trans_handle {
 	u64 transid;
+	u64 bytes_reserved;
+	u64 qgroup_reserved;
+	unsigned long use_count;
 	unsigned long blocks_reserved;
 	unsigned long blocks_used;
-	struct btrfs_transaction *transaction;
-	u64 block_group;
-	u64 alloc_exclude_start;
-	u64 alloc_exclude_nr;
 	unsigned long delayed_ref_updates;
+	struct btrfs_transaction *transaction;
+	struct btrfs_block_rsv *block_rsv;
+	struct btrfs_block_rsv *orig_rsv;
+	short aborted;
+	short adding_csums;
+	bool allocating_chunk;
+	bool reloc_reserved;
+	bool sync;
+	unsigned int type;
+	/*
+	 * this root is only needed to validate that the root passed to
+	 * start_transaction is the same as the one passed to end_transaction.
+	 * Subvolume quota depends on this
+	 */
+	struct btrfs_root *root;
+	struct seq_list delayed_ref_elem;
+	struct list_head qgroup_ref_list;
+	struct list_head new_bgs;
 };
 
 struct btrfs_pending_snapshot {
 	struct dentry *dentry;
+	struct inode *dir;
 	struct btrfs_root *root;
-	char *name;
-	struct btrfs_key root_key;
+	struct btrfs_root *snap;
+	struct btrfs_qgroup_inherit *inherit;
+	/* block reservation for the operation */
+	struct btrfs_block_rsv block_rsv;
+	u64 qgroup_reserved;
+	/* extra metadata reseration for relocation */
+	int error;
+	bool readonly;
 	struct list_head list;
 };
 
-static inline void btrfs_set_trans_block_group(struct btrfs_trans_handle *trans,
-					       struct inode *inode)
-{
-	trans->block_group = BTRFS_I(inode)->block_group;
-}
-
-static inline void btrfs_update_inode_block_group(
-					  struct btrfs_trans_handle *trans,
-					  struct inode *inode)
-{
-	BTRFS_I(inode)->block_group = trans->block_group;
-}
-
 static inline void btrfs_set_inode_last_trans(struct btrfs_trans_handle *trans,
 					      struct inode *inode)
 {
 	BTRFS_I(inode)->last_trans = trans->transaction->transid;
 	BTRFS_I(inode)->last_sub_trans = BTRFS_I(inode)->root->log_transid;
+	BTRFS_I(inode)->last_log_commit = BTRFS_I(inode)->root->last_log_commit;
 }
 
 int btrfs_end_transaction(struct btrfs_trans_handle *trans,
 			  struct btrfs_root *root);
 struct btrfs_trans_handle *btrfs_start_transaction(struct btrfs_root *root,
-						   int num_blocks);
-struct btrfs_trans_handle *btrfs_join_transaction(struct btrfs_root *root,
-						   int num_blocks);
-struct btrfs_trans_handle *btrfs_start_ioctl_transaction(struct btrfs_root *r,
-						   int num_blocks);
+						   int num_items);
+struct btrfs_trans_handle *btrfs_start_transaction_lflush(
+					struct btrfs_root *root, int num_items);
+struct btrfs_trans_handle *btrfs_join_transaction(struct btrfs_root *root);
+struct btrfs_trans_handle *btrfs_join_transaction_nolock(struct btrfs_root *root);
+struct btrfs_trans_handle *btrfs_attach_transaction(struct btrfs_root *root);
+struct btrfs_trans_handle *btrfs_attach_transaction_barrier(
+					struct btrfs_root *root);
+struct btrfs_trans_handle *btrfs_start_ioctl_transaction(struct btrfs_root *root);
+int btrfs_wait_for_commit(struct btrfs_root *root, u64 transid);
 int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans,
 				     struct btrfs_root *root);
-int btrfs_commit_tree_roots(struct btrfs_trans_handle *trans,
-			    struct btrfs_root *root);
 
-int btrfs_add_dead_root(struct btrfs_root *root);
-int btrfs_drop_dead_root(struct btrfs_root *root);
-int btrfs_defrag_root(struct btrfs_root *root, int cacheonly);
-int btrfs_clean_old_snapshots(struct btrfs_root *root);
+void btrfs_add_dead_root(struct btrfs_root *root);
+int btrfs_defrag_root(struct btrfs_root *root);
+int btrfs_clean_one_deleted_snapshot(struct btrfs_root *root);
 int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 			     struct btrfs_root *root);
+int btrfs_commit_transaction_async(struct btrfs_trans_handle *trans,
+				   struct btrfs_root *root,
+				   int wait_for_unblock);
 int btrfs_end_transaction_throttle(struct btrfs_trans_handle *trans,
 				   struct btrfs_root *root);
+int btrfs_should_end_transaction(struct btrfs_trans_handle *trans,
+				 struct btrfs_root *root);
 void btrfs_throttle(struct btrfs_root *root);
 int btrfs_record_root_in_trans(struct btrfs_trans_handle *trans,
 				struct btrfs_root *root);
-int btrfs_write_and_wait_marked_extents(struct btrfs_root *root,
-				struct extent_io_tree *dirty_pages, int mark);
 int btrfs_write_marked_extents(struct btrfs_root *root,
 				struct extent_io_tree *dirty_pages, int mark);
 int btrfs_wait_marked_extents(struct btrfs_root *root,
 				struct extent_io_tree *dirty_pages, int mark);
+int btrfs_transaction_blocked(struct btrfs_fs_info *info);
 int btrfs_transaction_in_commit(struct btrfs_fs_info *info);
+void btrfs_put_transaction(struct btrfs_transaction *transaction);
 #endif
diff --git a/fs/btrfs/tree-defrag.c b/fs/btrfs/tree-defrag.c
index b10eacdb162..a63719cc957 100644
--- a/fs/btrfs/tree-defrag.c
+++ b/fs/btrfs/tree-defrag.c
@@ -23,28 +23,24 @@
 #include "transaction.h"
 #include "locking.h"
 
-/* defrag all the leaves in a given btree.  If cache_only == 1, don't read
- * things from disk, otherwise read all the leaves and try to get key order to
+/*
+ * Defrag all the leaves in a given btree.
+ * Read all the leaves and try to get key order to
  * better reflect disk order
  */
 
 int btrfs_defrag_leaves(struct btrfs_trans_handle *trans,
-			struct btrfs_root *root, int cache_only)
+			struct btrfs_root *root)
 {
 	struct btrfs_path *path = NULL;
 	struct btrfs_key key;
 	int ret = 0;
 	int wret;
 	int level;
-	int orig_level;
-	int is_extent = 0;
 	int next_key_ret = 0;
 	u64 last_ret = 0;
 	u64 min_trans = 0;
 
-	if (cache_only)
-		goto out;
-
 	if (root->fs_info->extent_root == root) {
 		/*
 		 * there's recursion here right now in the tree locking,
@@ -53,7 +49,7 @@ int btrfs_defrag_leaves(struct btrfs_trans_handle *trans,
 		goto out;
 	}
 
-	if (root->ref_cows == 0 && !is_extent)
+	if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
 		goto out;
 
 	if (btrfs_test_opt(root, SSD))
@@ -64,7 +60,6 @@ int btrfs_defrag_leaves(struct btrfs_trans_handle *trans,
 		return -ENOMEM;
 
 	level = btrfs_header_level(root->node);
-	orig_level = level;
 
 	if (level == 0)
 		goto out;
@@ -88,18 +83,15 @@ int btrfs_defrag_leaves(struct btrfs_trans_handle *trans,
 	}
 
 	path->keep_locks = 1;
-	if (cache_only)
-		min_trans = root->defrag_trans_start;
 
-	ret = btrfs_search_forward(root, &key, NULL, path,
-				   cache_only, min_trans);
+	ret = btrfs_search_forward(root, &key, path, min_trans);
 	if (ret < 0)
 		goto out;
 	if (ret > 0) {
 		ret = 0;
 		goto out;
 	}
-	btrfs_release_path(root, path);
+	btrfs_release_path(path);
 	wret = btrfs_search_slot(trans, root, &key, path, 0, 1);
 
 	if (wret < 0) {
@@ -111,19 +103,20 @@ int btrfs_defrag_leaves(struct btrfs_trans_handle *trans,
 		goto out;
 	}
 	path->slots[1] = btrfs_header_nritems(path->nodes[1]);
-	next_key_ret = btrfs_find_next_key(root, path, &key, 1, cache_only,
+	next_key_ret = btrfs_find_next_key(root, path, &key, 1,
 					   min_trans);
 	ret = btrfs_realloc_node(trans, root,
 				 path->nodes[1], 0,
-				 cache_only, &last_ret,
+				 &last_ret,
 				 &root->defrag_progress);
-	WARN_ON(ret && ret != -EAGAIN);
+	if (ret) {
+		WARN_ON(ret == -EAGAIN);
+		goto out;
+	}
 	if (next_key_ret == 0) {
 		memcpy(&root->defrag_progress, &key, sizeof(key));
 		ret = -EAGAIN;
 	}
-
-	btrfs_release_path(root, path);
 out:
 	if (path)
 		btrfs_free_path(path);
diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c
index af57dd2b43d..9e1f2cd5e67 100644
--- a/fs/btrfs/tree-log.c
+++ b/fs/btrfs/tree-log.c
@@ -18,13 +18,14 @@
 
 #include <linux/sched.h>
 #include <linux/slab.h>
-#include "ctree.h"
-#include "transaction.h"
+#include <linux/blkdev.h>
+#include <linux/list_sort.h>
+#include "tree-log.h"
 #include "disk-io.h"
 #include "locking.h"
 #include "print-tree.h"
-#include "compat.h"
-#include "tree-log.h"
+#include "backref.h"
+#include "hash.h"
 
 /* magic values for the inode_only field in btrfs_log_inode:
  *
@@ -89,7 +90,8 @@
  */
 #define LOG_WALK_PIN_ONLY 0
 #define LOG_WALK_REPLAY_INODES 1
-#define LOG_WALK_REPLAY_ALL 2
+#define LOG_WALK_REPLAY_DIR_INDEX 2
+#define LOG_WALK_REPLAY_ALL 3
 
 static int btrfs_log_inode(struct btrfs_trans_handle *trans,
 			     struct btrfs_root *root, struct inode *inode,
@@ -132,40 +134,61 @@ static noinline int replay_dir_deletes(struct btrfs_trans_handle *trans,
  * syncing the tree wait for us to finish
  */
 static int start_log_trans(struct btrfs_trans_handle *trans,
-			   struct btrfs_root *root)
+			   struct btrfs_root *root,
+			   struct btrfs_log_ctx *ctx)
 {
+	int index;
 	int ret;
 
 	mutex_lock(&root->log_mutex);
 	if (root->log_root) {
+		if (btrfs_need_log_full_commit(root->fs_info, trans)) {
+			ret = -EAGAIN;
+			goto out;
+		}
 		if (!root->log_start_pid) {
 			root->log_start_pid = current->pid;
-			root->log_multiple_pids = false;
+			clear_bit(BTRFS_ROOT_MULTI_LOG_TASKS, &root->state);
 		} else if (root->log_start_pid != current->pid) {
-			root->log_multiple_pids = true;
+			set_bit(BTRFS_ROOT_MULTI_LOG_TASKS, &root->state);
 		}
 
-		root->log_batch++;
+		atomic_inc(&root->log_batch);
 		atomic_inc(&root->log_writers);
+		if (ctx) {
+			index = root->log_transid % 2;
+			list_add_tail(&ctx->list, &root->log_ctxs[index]);
+			ctx->log_transid = root->log_transid;
+		}
 		mutex_unlock(&root->log_mutex);
 		return 0;
 	}
-	root->log_multiple_pids = false;
-	root->log_start_pid = current->pid;
+
+	ret = 0;
 	mutex_lock(&root->fs_info->tree_log_mutex);
-	if (!root->fs_info->log_root_tree) {
+	if (!root->fs_info->log_root_tree)
 		ret = btrfs_init_log_root_tree(trans, root->fs_info);
-		BUG_ON(ret);
-	}
+	mutex_unlock(&root->fs_info->tree_log_mutex);
+	if (ret)
+		goto out;
+
 	if (!root->log_root) {
 		ret = btrfs_add_log_tree(trans, root);
-		BUG_ON(ret);
+		if (ret)
+			goto out;
 	}
-	mutex_unlock(&root->fs_info->tree_log_mutex);
-	root->log_batch++;
+	clear_bit(BTRFS_ROOT_MULTI_LOG_TASKS, &root->state);
+	root->log_start_pid = current->pid;
+	atomic_inc(&root->log_batch);
 	atomic_inc(&root->log_writers);
+	if (ctx) {
+		index = root->log_transid % 2;
+		list_add_tail(&ctx->list, &root->log_ctxs[index]);
+		ctx->log_transid = root->log_transid;
+	}
+out:
 	mutex_unlock(&root->log_mutex);
-	return 0;
+	return ret;
 }
 
 /*
@@ -209,14 +232,13 @@ int btrfs_pin_log_trans(struct btrfs_root *root)
  * indicate we're done making changes to the log tree
  * and wake up anyone waiting to do a sync
  */
-int btrfs_end_log_trans(struct btrfs_root *root)
+void btrfs_end_log_trans(struct btrfs_root *root)
 {
 	if (atomic_dec_and_test(&root->log_writers)) {
 		smp_mb();
 		if (waitqueue_active(&root->log_writer_wait))
 			wake_up(&root->log_writer_wait);
 	}
-	return 0;
 }
 
 
@@ -272,17 +294,31 @@ static int process_one_buffer(struct btrfs_root *log,
 			      struct extent_buffer *eb,
 			      struct walk_control *wc, u64 gen)
 {
+	int ret = 0;
+
+	/*
+	 * If this fs is mixed then we need to be able to process the leaves to
+	 * pin down any logged extents, so we have to read the block.
+	 */
+	if (btrfs_fs_incompat(log->fs_info, MIXED_GROUPS)) {
+		ret = btrfs_read_buffer(eb, gen);
+		if (ret)
+			return ret;
+	}
+
 	if (wc->pin)
-		btrfs_pin_extent(log->fs_info->extent_root,
-				 eb->start, eb->len, 0);
+		ret = btrfs_pin_extent_for_log_replay(log->fs_info->extent_root,
+						      eb->start, eb->len);
 
-	if (btrfs_buffer_uptodate(eb, gen)) {
+	if (!ret && btrfs_buffer_uptodate(eb, gen, 0)) {
+		if (wc->pin && btrfs_header_level(eb) == 0)
+			ret = btrfs_exclude_logged_extents(log, eb);
 		if (wc->write)
 			btrfs_write_tree_block(eb);
 		if (wc->wait)
 			btrfs_wait_tree_block_writeback(eb);
 	}
-	return 0;
+	return ret;
 }
 
 /*
@@ -312,6 +348,7 @@ static noinline int overwrite_item(struct btrfs_trans_handle *trans,
 	unsigned long src_ptr;
 	unsigned long dst_ptr;
 	int overwrite_root = 0;
+	bool inode_item = key->type == BTRFS_INODE_ITEM_KEY;
 
 	if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID)
 		overwrite_root = 1;
@@ -321,6 +358,9 @@ static noinline int overwrite_item(struct btrfs_trans_handle *trans,
 
 	/* look for the key in the destination tree */
 	ret = btrfs_search_slot(NULL, root, key, path, 0, 0);
+	if (ret < 0)
+		return ret;
+
 	if (ret == 0) {
 		char *src_copy;
 		char *dst_copy;
@@ -330,11 +370,17 @@ static noinline int overwrite_item(struct btrfs_trans_handle *trans,
 			goto insert;
 
 		if (item_size == 0) {
-			btrfs_release_path(root, path);
+			btrfs_release_path(path);
 			return 0;
 		}
 		dst_copy = kmalloc(item_size, GFP_NOFS);
 		src_copy = kmalloc(item_size, GFP_NOFS);
+		if (!dst_copy || !src_copy) {
+			btrfs_release_path(path);
+			kfree(dst_copy);
+			kfree(src_copy);
+			return -ENOMEM;
+		}
 
 		read_extent_buffer(eb, src_copy, src_ptr, item_size);
 
@@ -352,13 +398,57 @@ static noinline int overwrite_item(struct btrfs_trans_handle *trans,
 		 * sync
 		 */
 		if (ret == 0) {
-			btrfs_release_path(root, path);
+			btrfs_release_path(path);
 			return 0;
 		}
 
+		/*
+		 * We need to load the old nbytes into the inode so when we
+		 * replay the extents we've logged we get the right nbytes.
+		 */
+		if (inode_item) {
+			struct btrfs_inode_item *item;
+			u64 nbytes;
+			u32 mode;
+
+			item = btrfs_item_ptr(path->nodes[0], path->slots[0],
+					      struct btrfs_inode_item);
+			nbytes = btrfs_inode_nbytes(path->nodes[0], item);
+			item = btrfs_item_ptr(eb, slot,
+					      struct btrfs_inode_item);
+			btrfs_set_inode_nbytes(eb, item, nbytes);
+
+			/*
+			 * If this is a directory we need to reset the i_size to
+			 * 0 so that we can set it up properly when replaying
+			 * the rest of the items in this log.
+			 */
+			mode = btrfs_inode_mode(eb, item);
+			if (S_ISDIR(mode))
+				btrfs_set_inode_size(eb, item, 0);
+		}
+	} else if (inode_item) {
+		struct btrfs_inode_item *item;
+		u32 mode;
+
+		/*
+		 * New inode, set nbytes to 0 so that the nbytes comes out
+		 * properly when we replay the extents.
+		 */
+		item = btrfs_item_ptr(eb, slot, struct btrfs_inode_item);
+		btrfs_set_inode_nbytes(eb, item, 0);
+
+		/*
+		 * If this is a directory we need to reset the i_size to 0 so
+		 * that we can set it up properly when replaying the rest of
+		 * the items in this log.
+		 */
+		mode = btrfs_inode_mode(eb, item);
+		if (S_ISDIR(mode))
+			btrfs_set_inode_size(eb, item, 0);
 	}
 insert:
-	btrfs_release_path(root, path);
+	btrfs_release_path(path);
 	/* try to insert the key into the destination tree */
 	ret = btrfs_insert_empty_item(trans, root, path,
 				      key, item_size);
@@ -368,15 +458,13 @@ insert:
 		u32 found_size;
 		found_size = btrfs_item_size_nr(path->nodes[0],
 						path->slots[0]);
-		if (found_size > item_size) {
-			btrfs_truncate_item(trans, root, path, item_size, 1);
-		} else if (found_size < item_size) {
-			ret = btrfs_extend_item(trans, root, path,
-						item_size - found_size);
-			BUG_ON(ret);
-		}
+		if (found_size > item_size)
+			btrfs_truncate_item(root, path, item_size, 1);
+		else if (found_size < item_size)
+			btrfs_extend_item(root, path,
+					  item_size - found_size);
 	} else if (ret) {
-		BUG();
+		return ret;
 	}
 	dst_ptr = btrfs_item_ptr_offset(path->nodes[0],
 					path->slots[0]);
@@ -429,7 +517,7 @@ insert:
 	}
 no_copy:
 	btrfs_mark_buffer_dirty(path->nodes[0]);
-	btrfs_release_path(root, path);
+	btrfs_release_path(path);
 	return 0;
 }
 
@@ -475,11 +563,9 @@ static noinline int replay_one_extent(struct btrfs_trans_handle *trans,
 				      struct btrfs_key *key)
 {
 	int found_type;
-	u64 mask = root->sectorsize - 1;
 	u64 extent_end;
-	u64 alloc_hint;
 	u64 start = key->offset;
-	u64 saved_nbytes;
+	u64 nbytes = 0;
 	struct btrfs_file_extent_item *item;
 	struct inode *inode = NULL;
 	unsigned long size;
@@ -489,11 +575,20 @@ static noinline int replay_one_extent(struct btrfs_trans_handle *trans,
 	found_type = btrfs_file_extent_type(eb, item);
 
 	if (found_type == BTRFS_FILE_EXTENT_REG ||
-	    found_type == BTRFS_FILE_EXTENT_PREALLOC)
-		extent_end = start + btrfs_file_extent_num_bytes(eb, item);
-	else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
-		size = btrfs_file_extent_inline_len(eb, item);
-		extent_end = (start + size + mask) & ~mask;
+	    found_type == BTRFS_FILE_EXTENT_PREALLOC) {
+		nbytes = btrfs_file_extent_num_bytes(eb, item);
+		extent_end = start + nbytes;
+
+		/*
+		 * We don't add to the inodes nbytes if we are prealloc or a
+		 * hole.
+		 */
+		if (btrfs_file_extent_disk_bytenr(eb, item) == 0)
+			nbytes = 0;
+	} else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
+		size = btrfs_file_extent_inline_len(eb, slot, item);
+		nbytes = btrfs_file_extent_ram_bytes(eb, item);
+		extent_end = ALIGN(start + size, root->sectorsize);
 	} else {
 		ret = 0;
 		goto out;
@@ -510,7 +605,7 @@ static noinline int replay_one_extent(struct btrfs_trans_handle *trans,
 	 * file.  This must be done before the btrfs_drop_extents run
 	 * so we don't try to drop this extent.
 	 */
-	ret = btrfs_lookup_file_extent(trans, root, path, inode->i_ino,
+	ret = btrfs_lookup_file_extent(trans, root, path, btrfs_ino(inode),
 				       start, 0);
 
 	if (ret == 0 &&
@@ -535,17 +630,16 @@ static noinline int replay_one_extent(struct btrfs_trans_handle *trans,
 		 * we don't have to do anything
 		 */
 		if (memcmp(&cmp1, &cmp2, sizeof(cmp1)) == 0) {
-			btrfs_release_path(root, path);
+			btrfs_release_path(path);
 			goto out;
 		}
 	}
-	btrfs_release_path(root, path);
+	btrfs_release_path(path);
 
-	saved_nbytes = inode_get_bytes(inode);
 	/* drop any overlapping extents */
-	ret = btrfs_drop_extents(trans, inode, start, extent_end,
-				 &alloc_hint, 1);
-	BUG_ON(ret);
+	ret = btrfs_drop_extents(trans, root, inode, start, extent_end, 1);
+	if (ret)
+		goto out;
 
 	if (found_type == BTRFS_FILE_EXTENT_REG ||
 	    found_type == BTRFS_FILE_EXTENT_PREALLOC) {
@@ -555,7 +649,8 @@ static noinline int replay_one_extent(struct btrfs_trans_handle *trans,
 
 		ret = btrfs_insert_empty_item(trans, root, path, key,
 					      sizeof(*item));
-		BUG_ON(ret);
+		if (ret)
+			goto out;
 		dest_offset = btrfs_item_ptr_offset(path->nodes[0],
 						    path->slots[0]);
 		copy_extent_buffer(path->nodes[0], eb, dest_offset,
@@ -580,7 +675,9 @@ static noinline int replay_one_extent(struct btrfs_trans_handle *trans,
 				ret = btrfs_inc_extent_ref(trans, root,
 						ins.objectid, ins.offset,
 						0, root->root_key.objectid,
-						key->objectid, offset);
+						key->objectid, offset, 0);
+				if (ret)
+					goto out;
 			} else {
 				/*
 				 * insert the extent pointer in the extent
@@ -589,9 +686,10 @@ static noinline int replay_one_extent(struct btrfs_trans_handle *trans,
 				ret = btrfs_alloc_logged_file_extent(trans,
 						root, root->root_key.objectid,
 						key->objectid, offset, &ins);
-				BUG_ON(ret);
+				if (ret)
+					goto out;
 			}
-			btrfs_release_path(root, path);
+			btrfs_release_path(path);
 
 			if (btrfs_file_extent_compression(eb, item)) {
 				csum_start = ins.objectid;
@@ -605,31 +703,35 @@ static noinline int replay_one_extent(struct btrfs_trans_handle *trans,
 
 			ret = btrfs_lookup_csums_range(root->log_root,
 						csum_start, csum_end - 1,
-						&ordered_sums);
-			BUG_ON(ret);
+						&ordered_sums, 0);
+			if (ret)
+				goto out;
 			while (!list_empty(&ordered_sums)) {
 				struct btrfs_ordered_sum *sums;
 				sums = list_entry(ordered_sums.next,
 						struct btrfs_ordered_sum,
 						list);
-				ret = btrfs_csum_file_blocks(trans,
+				if (!ret)
+					ret = btrfs_csum_file_blocks(trans,
 						root->fs_info->csum_root,
 						sums);
-				BUG_ON(ret);
 				list_del(&sums->list);
 				kfree(sums);
 			}
+			if (ret)
+				goto out;
 		} else {
-			btrfs_release_path(root, path);
+			btrfs_release_path(path);
 		}
 	} else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
 		/* inline extents are easy, we just overwrite them */
 		ret = overwrite_item(trans, root, path, eb, slot, key);
-		BUG_ON(ret);
+		if (ret)
+			goto out;
 	}
 
-	inode_set_bytes(inode, saved_nbytes);
-	btrfs_update_inode(trans, root, inode);
+	inode_add_bytes(inode, nbytes);
+	ret = btrfs_update_inode(trans, root, inode);
 out:
 	if (inode)
 		iput(inode);
@@ -662,19 +764,29 @@ static noinline int drop_one_dir_item(struct btrfs_trans_handle *trans,
 	btrfs_dir_item_key_to_cpu(leaf, di, &location);
 	name_len = btrfs_dir_name_len(leaf, di);
 	name = kmalloc(name_len, GFP_NOFS);
+	if (!name)
+		return -ENOMEM;
+
 	read_extent_buffer(leaf, name, (unsigned long)(di + 1), name_len);
-	btrfs_release_path(root, path);
+	btrfs_release_path(path);
 
 	inode = read_one_inode(root, location.objectid);
-	BUG_ON(!inode);
+	if (!inode) {
+		ret = -EIO;
+		goto out;
+	}
 
 	ret = link_to_fixup_dir(trans, root, path, location.objectid);
-	BUG_ON(ret);
+	if (ret)
+		goto out;
 
 	ret = btrfs_unlink_inode(trans, root, dir, inode, name, name_len);
-	BUG_ON(ret);
+	if (ret)
+		goto out;
+	else
+		ret = btrfs_run_delayed_items(trans, root);
+out:
 	kfree(name);
-
 	iput(inode);
 	return ret;
 }
@@ -701,7 +813,7 @@ static noinline int inode_in_dir(struct btrfs_root *root,
 			goto out;
 	} else
 		goto out;
-	btrfs_release_path(root, path);
+	btrfs_release_path(path);
 
 	di = btrfs_lookup_dir_item(NULL, root, path, dirid, name, name_len, 0);
 	if (di && !IS_ERR(di)) {
@@ -712,7 +824,7 @@ static noinline int inode_in_dir(struct btrfs_root *root,
 		goto out;
 	match = 1;
 out:
-	btrfs_release_path(root, path);
+	btrfs_release_path(path);
 	return match;
 }
 
@@ -728,6 +840,7 @@ out:
  */
 static noinline int backref_in_log(struct btrfs_root *log,
 				   struct btrfs_key *key,
+				   u64 ref_objectid,
 				   char *name, int namelen)
 {
 	struct btrfs_path *path;
@@ -741,12 +854,24 @@ static noinline int backref_in_log(struct btrfs_root *log,
 	int match = 0;
 
 	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
 	ret = btrfs_search_slot(NULL, log, key, path, 0, 0);
 	if (ret != 0)
 		goto out;
 
-	item_size = btrfs_item_size_nr(path->nodes[0], path->slots[0]);
 	ptr = btrfs_item_ptr_offset(path->nodes[0], path->slots[0]);
+
+	if (key->type == BTRFS_INODE_EXTREF_KEY) {
+		if (btrfs_find_name_in_ext_backref(path, ref_objectid,
+						   name, namelen, NULL))
+			match = 1;
+
+		goto out;
+	}
+
+	item_size = btrfs_item_size_nr(path->nodes[0], path->slots[0]);
 	ptr_end = ptr + item_size;
 	while (ptr < ptr_end) {
 		ref = (struct btrfs_inode_ref *)ptr;
@@ -767,89 +892,42 @@ out:
 	return match;
 }
 
-
-/*
- * replay one inode back reference item found in the log tree.
- * eb, slot and key refer to the buffer and key found in the log tree.
- * root is the destination we are replaying into, and path is for temp
- * use by this function.  (it should be released on return).
- */
-static noinline int add_inode_ref(struct btrfs_trans_handle *trans,
+static inline int __add_inode_ref(struct btrfs_trans_handle *trans,
 				  struct btrfs_root *root,
-				  struct btrfs_root *log,
 				  struct btrfs_path *path,
-				  struct extent_buffer *eb, int slot,
-				  struct btrfs_key *key)
+				  struct btrfs_root *log_root,
+				  struct inode *dir, struct inode *inode,
+				  struct extent_buffer *eb,
+				  u64 inode_objectid, u64 parent_objectid,
+				  u64 ref_index, char *name, int namelen,
+				  int *search_done)
 {
-	struct inode *dir;
 	int ret;
-	struct btrfs_key location;
-	struct btrfs_inode_ref *ref;
+	char *victim_name;
+	int victim_name_len;
+	struct extent_buffer *leaf;
 	struct btrfs_dir_item *di;
-	struct inode *inode;
-	char *name;
-	int namelen;
-	unsigned long ref_ptr;
-	unsigned long ref_end;
-
-	location.objectid = key->objectid;
-	location.type = BTRFS_INODE_ITEM_KEY;
-	location.offset = 0;
-
-	/*
-	 * it is possible that we didn't log all the parent directories
-	 * for a given inode.  If we don't find the dir, just don't
-	 * copy the back ref in.  The link count fixup code will take
-	 * care of the rest
-	 */
-	dir = read_one_inode(root, key->offset);
-	if (!dir)
-		return -ENOENT;
-
-	inode = read_one_inode(root, key->objectid);
-	BUG_ON(!inode);
-
-	ref_ptr = btrfs_item_ptr_offset(eb, slot);
-	ref_end = ref_ptr + btrfs_item_size_nr(eb, slot);
+	struct btrfs_key search_key;
+	struct btrfs_inode_extref *extref;
 
 again:
-	ref = (struct btrfs_inode_ref *)ref_ptr;
-
-	namelen = btrfs_inode_ref_name_len(eb, ref);
-	name = kmalloc(namelen, GFP_NOFS);
-	BUG_ON(!name);
-
-	read_extent_buffer(eb, name, (unsigned long)(ref + 1), namelen);
-
-	/* if we already have a perfect match, we're done */
-	if (inode_in_dir(root, path, dir->i_ino, inode->i_ino,
-			 btrfs_inode_ref_index(eb, ref),
-			 name, namelen)) {
-		goto out;
-	}
-
-	/*
-	 * look for a conflicting back reference in the metadata.
-	 * if we find one we have to unlink that name of the file
-	 * before we add our new link.  Later on, we overwrite any
-	 * existing back reference, and we don't want to create
-	 * dangling pointers in the directory.
-	 */
-conflict_again:
-	ret = btrfs_search_slot(NULL, root, key, path, 0, 0);
+	/* Search old style refs */
+	search_key.objectid = inode_objectid;
+	search_key.type = BTRFS_INODE_REF_KEY;
+	search_key.offset = parent_objectid;
+	ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0);
 	if (ret == 0) {
-		char *victim_name;
-		int victim_name_len;
 		struct btrfs_inode_ref *victim_ref;
 		unsigned long ptr;
 		unsigned long ptr_end;
-		struct extent_buffer *leaf = path->nodes[0];
+
+		leaf = path->nodes[0];
 
 		/* are we trying to overwrite a back ref for the root directory
 		 * if so, just jump out, we're done
 		 */
-		if (key->objectid == key->offset)
-			goto out_nowrite;
+		if (search_key.objectid == search_key.offset)
+			return 1;
 
 		/* check all the names in this back reference to see
 		 * if they are in the log.  if so, we allow them to stay
@@ -862,114 +940,385 @@ conflict_again:
 			victim_name_len = btrfs_inode_ref_name_len(leaf,
 								   victim_ref);
 			victim_name = kmalloc(victim_name_len, GFP_NOFS);
-			BUG_ON(!victim_name);
+			if (!victim_name)
+				return -ENOMEM;
 
 			read_extent_buffer(leaf, victim_name,
 					   (unsigned long)(victim_ref + 1),
 					   victim_name_len);
 
-			if (!backref_in_log(log, key, victim_name,
+			if (!backref_in_log(log_root, &search_key,
+					    parent_objectid,
+					    victim_name,
 					    victim_name_len)) {
-				btrfs_inc_nlink(inode);
-				btrfs_release_path(root, path);
+				inc_nlink(inode);
+				btrfs_release_path(path);
 
 				ret = btrfs_unlink_inode(trans, root, dir,
 							 inode, victim_name,
 							 victim_name_len);
 				kfree(victim_name);
-				btrfs_release_path(root, path);
-				goto conflict_again;
+				if (ret)
+					return ret;
+				ret = btrfs_run_delayed_items(trans, root);
+				if (ret)
+					return ret;
+				*search_done = 1;
+				goto again;
 			}
 			kfree(victim_name);
+
 			ptr = (unsigned long)(victim_ref + 1) + victim_name_len;
 		}
-		BUG_ON(ret);
+
+		/*
+		 * NOTE: we have searched root tree and checked the
+		 * coresponding ref, it does not need to check again.
+		 */
+		*search_done = 1;
 	}
-	btrfs_release_path(root, path);
+	btrfs_release_path(path);
+
+	/* Same search but for extended refs */
+	extref = btrfs_lookup_inode_extref(NULL, root, path, name, namelen,
+					   inode_objectid, parent_objectid, 0,
+					   0);
+	if (!IS_ERR_OR_NULL(extref)) {
+		u32 item_size;
+		u32 cur_offset = 0;
+		unsigned long base;
+		struct inode *victim_parent;
+
+		leaf = path->nodes[0];
+
+		item_size = btrfs_item_size_nr(leaf, path->slots[0]);
+		base = btrfs_item_ptr_offset(leaf, path->slots[0]);
+
+		while (cur_offset < item_size) {
+			extref = (struct btrfs_inode_extref *)base + cur_offset;
+
+			victim_name_len = btrfs_inode_extref_name_len(leaf, extref);
+
+			if (btrfs_inode_extref_parent(leaf, extref) != parent_objectid)
+				goto next;
+
+			victim_name = kmalloc(victim_name_len, GFP_NOFS);
+			if (!victim_name)
+				return -ENOMEM;
+			read_extent_buffer(leaf, victim_name, (unsigned long)&extref->name,
+					   victim_name_len);
+
+			search_key.objectid = inode_objectid;
+			search_key.type = BTRFS_INODE_EXTREF_KEY;
+			search_key.offset = btrfs_extref_hash(parent_objectid,
+							      victim_name,
+							      victim_name_len);
+			ret = 0;
+			if (!backref_in_log(log_root, &search_key,
+					    parent_objectid, victim_name,
+					    victim_name_len)) {
+				ret = -ENOENT;
+				victim_parent = read_one_inode(root,
+							       parent_objectid);
+				if (victim_parent) {
+					inc_nlink(inode);
+					btrfs_release_path(path);
+
+					ret = btrfs_unlink_inode(trans, root,
+								 victim_parent,
+								 inode,
+								 victim_name,
+								 victim_name_len);
+					if (!ret)
+						ret = btrfs_run_delayed_items(
+								  trans, root);
+				}
+				iput(victim_parent);
+				kfree(victim_name);
+				if (ret)
+					return ret;
+				*search_done = 1;
+				goto again;
+			}
+			kfree(victim_name);
+			if (ret)
+				return ret;
+next:
+			cur_offset += victim_name_len + sizeof(*extref);
+		}
+		*search_done = 1;
+	}
+	btrfs_release_path(path);
 
 	/* look for a conflicting sequence number */
-	di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino,
-					 btrfs_inode_ref_index(eb, ref),
-					 name, namelen, 0);
+	di = btrfs_lookup_dir_index_item(trans, root, path, btrfs_ino(dir),
+					 ref_index, name, namelen, 0);
 	if (di && !IS_ERR(di)) {
 		ret = drop_one_dir_item(trans, root, path, dir, di);
-		BUG_ON(ret);
+		if (ret)
+			return ret;
 	}
-	btrfs_release_path(root, path);
+	btrfs_release_path(path);
 
-
-	/* look for a conflicting name */
-	di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino,
+	/* look for a conflicing name */
+	di = btrfs_lookup_dir_item(trans, root, path, btrfs_ino(dir),
 				   name, namelen, 0);
 	if (di && !IS_ERR(di)) {
 		ret = drop_one_dir_item(trans, root, path, dir, di);
-		BUG_ON(ret);
+		if (ret)
+			return ret;
 	}
-	btrfs_release_path(root, path);
+	btrfs_release_path(path);
 
-	/* insert our name */
-	ret = btrfs_add_link(trans, dir, inode, name, namelen, 0,
-			     btrfs_inode_ref_index(eb, ref));
-	BUG_ON(ret);
+	return 0;
+}
 
-	btrfs_update_inode(trans, root, inode);
+static int extref_get_fields(struct extent_buffer *eb, unsigned long ref_ptr,
+			     u32 *namelen, char **name, u64 *index,
+			     u64 *parent_objectid)
+{
+	struct btrfs_inode_extref *extref;
 
-out:
-	ref_ptr = (unsigned long)(ref + 1) + namelen;
-	kfree(name);
-	if (ref_ptr < ref_end)
-		goto again;
+	extref = (struct btrfs_inode_extref *)ref_ptr;
+
+	*namelen = btrfs_inode_extref_name_len(eb, extref);
+	*name = kmalloc(*namelen, GFP_NOFS);
+	if (*name == NULL)
+		return -ENOMEM;
+
+	read_extent_buffer(eb, *name, (unsigned long)&extref->name,
+			   *namelen);
+
+	*index = btrfs_inode_extref_index(eb, extref);
+	if (parent_objectid)
+		*parent_objectid = btrfs_inode_extref_parent(eb, extref);
+
+	return 0;
+}
+
+static int ref_get_fields(struct extent_buffer *eb, unsigned long ref_ptr,
+			  u32 *namelen, char **name, u64 *index)
+{
+	struct btrfs_inode_ref *ref;
+
+	ref = (struct btrfs_inode_ref *)ref_ptr;
+
+	*namelen = btrfs_inode_ref_name_len(eb, ref);
+	*name = kmalloc(*namelen, GFP_NOFS);
+	if (*name == NULL)
+		return -ENOMEM;
+
+	read_extent_buffer(eb, *name, (unsigned long)(ref + 1), *namelen);
+
+	*index = btrfs_inode_ref_index(eb, ref);
+
+	return 0;
+}
+
+/*
+ * replay one inode back reference item found in the log tree.
+ * eb, slot and key refer to the buffer and key found in the log tree.
+ * root is the destination we are replaying into, and path is for temp
+ * use by this function.  (it should be released on return).
+ */
+static noinline int add_inode_ref(struct btrfs_trans_handle *trans,
+				  struct btrfs_root *root,
+				  struct btrfs_root *log,
+				  struct btrfs_path *path,
+				  struct extent_buffer *eb, int slot,
+				  struct btrfs_key *key)
+{
+	struct inode *dir = NULL;
+	struct inode *inode = NULL;
+	unsigned long ref_ptr;
+	unsigned long ref_end;
+	char *name = NULL;
+	int namelen;
+	int ret;
+	int search_done = 0;
+	int log_ref_ver = 0;
+	u64 parent_objectid;
+	u64 inode_objectid;
+	u64 ref_index = 0;
+	int ref_struct_size;
+
+	ref_ptr = btrfs_item_ptr_offset(eb, slot);
+	ref_end = ref_ptr + btrfs_item_size_nr(eb, slot);
+
+	if (key->type == BTRFS_INODE_EXTREF_KEY) {
+		struct btrfs_inode_extref *r;
+
+		ref_struct_size = sizeof(struct btrfs_inode_extref);
+		log_ref_ver = 1;
+		r = (struct btrfs_inode_extref *)ref_ptr;
+		parent_objectid = btrfs_inode_extref_parent(eb, r);
+	} else {
+		ref_struct_size = sizeof(struct btrfs_inode_ref);
+		parent_objectid = key->offset;
+	}
+	inode_objectid = key->objectid;
+
+	/*
+	 * it is possible that we didn't log all the parent directories
+	 * for a given inode.  If we don't find the dir, just don't
+	 * copy the back ref in.  The link count fixup code will take
+	 * care of the rest
+	 */
+	dir = read_one_inode(root, parent_objectid);
+	if (!dir) {
+		ret = -ENOENT;
+		goto out;
+	}
+
+	inode = read_one_inode(root, inode_objectid);
+	if (!inode) {
+		ret = -EIO;
+		goto out;
+	}
+
+	while (ref_ptr < ref_end) {
+		if (log_ref_ver) {
+			ret = extref_get_fields(eb, ref_ptr, &namelen, &name,
+						&ref_index, &parent_objectid);
+			/*
+			 * parent object can change from one array
+			 * item to another.
+			 */
+			if (!dir)
+				dir = read_one_inode(root, parent_objectid);
+			if (!dir) {
+				ret = -ENOENT;
+				goto out;
+			}
+		} else {
+			ret = ref_get_fields(eb, ref_ptr, &namelen, &name,
+					     &ref_index);
+		}
+		if (ret)
+			goto out;
+
+		/* if we already have a perfect match, we're done */
+		if (!inode_in_dir(root, path, btrfs_ino(dir), btrfs_ino(inode),
+				  ref_index, name, namelen)) {
+			/*
+			 * look for a conflicting back reference in the
+			 * metadata. if we find one we have to unlink that name
+			 * of the file before we add our new link.  Later on, we
+			 * overwrite any existing back reference, and we don't
+			 * want to create dangling pointers in the directory.
+			 */
+
+			if (!search_done) {
+				ret = __add_inode_ref(trans, root, path, log,
+						      dir, inode, eb,
+						      inode_objectid,
+						      parent_objectid,
+						      ref_index, name, namelen,
+						      &search_done);
+				if (ret) {
+					if (ret == 1)
+						ret = 0;
+					goto out;
+				}
+			}
+
+			/* insert our name */
+			ret = btrfs_add_link(trans, dir, inode, name, namelen,
+					     0, ref_index);
+			if (ret)
+				goto out;
+
+			btrfs_update_inode(trans, root, inode);
+		}
+
+		ref_ptr = (unsigned long)(ref_ptr + ref_struct_size) + namelen;
+		kfree(name);
+		name = NULL;
+		if (log_ref_ver) {
+			iput(dir);
+			dir = NULL;
+		}
+	}
 
 	/* finally write the back reference in the inode */
 	ret = overwrite_item(trans, root, path, eb, slot, key);
-	BUG_ON(ret);
-
-out_nowrite:
-	btrfs_release_path(root, path);
+out:
+	btrfs_release_path(path);
+	kfree(name);
 	iput(dir);
 	iput(inode);
-	return 0;
+	return ret;
 }
 
 static int insert_orphan_item(struct btrfs_trans_handle *trans,
 			      struct btrfs_root *root, u64 offset)
 {
 	int ret;
-	ret = btrfs_find_orphan_item(root, offset);
+	ret = btrfs_find_item(root, NULL, BTRFS_ORPHAN_OBJECTID,
+			offset, BTRFS_ORPHAN_ITEM_KEY, NULL);
 	if (ret > 0)
 		ret = btrfs_insert_orphan_item(trans, root, offset);
 	return ret;
 }
 
+static int count_inode_extrefs(struct btrfs_root *root,
+			       struct inode *inode, struct btrfs_path *path)
+{
+	int ret = 0;
+	int name_len;
+	unsigned int nlink = 0;
+	u32 item_size;
+	u32 cur_offset = 0;
+	u64 inode_objectid = btrfs_ino(inode);
+	u64 offset = 0;
+	unsigned long ptr;
+	struct btrfs_inode_extref *extref;
+	struct extent_buffer *leaf;
 
-/*
- * There are a few corners where the link count of the file can't
- * be properly maintained during replay.  So, instead of adding
- * lots of complexity to the log code, we just scan the backrefs
- * for any file that has been through replay.
- *
- * The scan will update the link count on the inode to reflect the
- * number of back refs found.  If it goes down to zero, the iput
- * will free the inode.
- */
-static noinline int fixup_inode_link_count(struct btrfs_trans_handle *trans,
-					   struct btrfs_root *root,
-					   struct inode *inode)
+	while (1) {
+		ret = btrfs_find_one_extref(root, inode_objectid, offset, path,
+					    &extref, &offset);
+		if (ret)
+			break;
+
+		leaf = path->nodes[0];
+		item_size = btrfs_item_size_nr(leaf, path->slots[0]);
+		ptr = btrfs_item_ptr_offset(leaf, path->slots[0]);
+
+		while (cur_offset < item_size) {
+			extref = (struct btrfs_inode_extref *) (ptr + cur_offset);
+			name_len = btrfs_inode_extref_name_len(leaf, extref);
+
+			nlink++;
+
+			cur_offset += name_len + sizeof(*extref);
+		}
+
+		offset++;
+		btrfs_release_path(path);
+	}
+	btrfs_release_path(path);
+
+	if (ret < 0)
+		return ret;
+	return nlink;
+}
+
+static int count_inode_refs(struct btrfs_root *root,
+			       struct inode *inode, struct btrfs_path *path)
 {
-	struct btrfs_path *path;
 	int ret;
 	struct btrfs_key key;
-	u64 nlink = 0;
+	unsigned int nlink = 0;
 	unsigned long ptr;
 	unsigned long ptr_end;
 	int name_len;
+	u64 ino = btrfs_ino(inode);
 
-	key.objectid = inode->i_ino;
+	key.objectid = ino;
 	key.type = BTRFS_INODE_REF_KEY;
 	key.offset = (u64)-1;
 
-	path = btrfs_alloc_path();
-
 	while (1) {
 		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
 		if (ret < 0)
@@ -979,9 +1328,10 @@ static noinline int fixup_inode_link_count(struct btrfs_trans_handle *trans,
 				break;
 			path->slots[0]--;
 		}
+process_slot:
 		btrfs_item_key_to_cpu(path->nodes[0], &key,
 				      path->slots[0]);
-		if (key.objectid != inode->i_ino ||
+		if (key.objectid != ino ||
 		    key.type != BTRFS_INODE_REF_KEY)
 			break;
 		ptr = btrfs_item_ptr_offset(path->nodes[0], path->slots[0]);
@@ -999,12 +1349,60 @@ static noinline int fixup_inode_link_count(struct btrfs_trans_handle *trans,
 
 		if (key.offset == 0)
 			break;
+		if (path->slots[0] > 0) {
+			path->slots[0]--;
+			goto process_slot;
+		}
 		key.offset--;
-		btrfs_release_path(root, path);
+		btrfs_release_path(path);
 	}
-	btrfs_release_path(root, path);
+	btrfs_release_path(path);
+
+	return nlink;
+}
+
+/*
+ * There are a few corners where the link count of the file can't
+ * be properly maintained during replay.  So, instead of adding
+ * lots of complexity to the log code, we just scan the backrefs
+ * for any file that has been through replay.
+ *
+ * The scan will update the link count on the inode to reflect the
+ * number of back refs found.  If it goes down to zero, the iput
+ * will free the inode.
+ */
+static noinline int fixup_inode_link_count(struct btrfs_trans_handle *trans,
+					   struct btrfs_root *root,
+					   struct inode *inode)
+{
+	struct btrfs_path *path;
+	int ret;
+	u64 nlink = 0;
+	u64 ino = btrfs_ino(inode);
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	ret = count_inode_refs(root, inode, path);
+	if (ret < 0)
+		goto out;
+
+	nlink = ret;
+
+	ret = count_inode_extrefs(root, inode, path);
+	if (ret == -ENOENT)
+		ret = 0;
+
+	if (ret < 0)
+		goto out;
+
+	nlink += ret;
+
+	ret = 0;
+
 	if (nlink != inode->i_nlink) {
-		inode->i_nlink = nlink;
+		set_nlink(inode, nlink);
 		btrfs_update_inode(trans, root, inode);
 	}
 	BTRFS_I(inode)->index_cnt = (u64)-1;
@@ -1012,15 +1410,16 @@ static noinline int fixup_inode_link_count(struct btrfs_trans_handle *trans,
 	if (inode->i_nlink == 0) {
 		if (S_ISDIR(inode->i_mode)) {
 			ret = replay_dir_deletes(trans, root, NULL, path,
-						 inode->i_ino, 1);
-			BUG_ON(ret);
+						 ino, 1);
+			if (ret)
+				goto out;
 		}
-		ret = insert_orphan_item(trans, root, inode->i_ino);
-		BUG_ON(ret);
+		ret = insert_orphan_item(trans, root, ino);
 	}
-	btrfs_free_path(path);
 
-	return 0;
+out:
+	btrfs_free_path(path);
+	return ret;
 }
 
 static noinline int fixup_inode_link_counts(struct btrfs_trans_handle *trans,
@@ -1051,16 +1450,18 @@ static noinline int fixup_inode_link_counts(struct btrfs_trans_handle *trans,
 			break;
 
 		ret = btrfs_del_item(trans, root, path);
-		BUG_ON(ret);
+		if (ret)
+			goto out;
 
-		btrfs_release_path(root, path);
+		btrfs_release_path(path);
 		inode = read_one_inode(root, key.offset);
-		BUG_ON(!inode);
+		if (!inode)
+			return -EIO;
 
 		ret = fixup_inode_link_count(trans, root, inode);
-		BUG_ON(ret);
-
 		iput(inode);
+		if (ret)
+			goto out;
 
 		/*
 		 * fixup on a directory may create new entries,
@@ -1069,8 +1470,10 @@ static noinline int fixup_inode_link_counts(struct btrfs_trans_handle *trans,
 		 */
 		key.offset = (u64)-1;
 	}
-	btrfs_release_path(root, path);
-	return 0;
+	ret = 0;
+out:
+	btrfs_release_path(path);
+	return ret;
 }
 
 
@@ -1089,7 +1492,8 @@ static noinline int link_to_fixup_dir(struct btrfs_trans_handle *trans,
 	struct inode *inode;
 
 	inode = read_one_inode(root, objectid);
-	BUG_ON(!inode);
+	if (!inode)
+		return -EIO;
 
 	key.objectid = BTRFS_TREE_LOG_FIXUP_OBJECTID;
 	btrfs_set_key_type(&key, BTRFS_ORPHAN_ITEM_KEY);
@@ -1097,14 +1501,17 @@ static noinline int link_to_fixup_dir(struct btrfs_trans_handle *trans,
 
 	ret = btrfs_insert_empty_item(trans, root, path, &key, 0);
 
-	btrfs_release_path(root, path);
+	btrfs_release_path(path);
 	if (ret == 0) {
-		btrfs_inc_nlink(inode);
-		btrfs_update_inode(trans, root, inode);
+		if (!inode->i_nlink)
+			set_nlink(inode, 1);
+		else
+			inc_nlink(inode);
+		ret = btrfs_update_inode(trans, root, inode);
 	} else if (ret == -EEXIST) {
 		ret = 0;
 	} else {
-		BUG();
+		BUG(); /* Logic Error */
 	}
 	iput(inode);
 
@@ -1136,6 +1543,7 @@ static noinline int insert_one_name(struct btrfs_trans_handle *trans,
 		iput(inode);
 		return -EIO;
 	}
+
 	ret = btrfs_add_link(trans, dir, inode, name, name_len, 1, index);
 
 	/* FIXME, put inode into FIXUP list */
@@ -1173,13 +1581,20 @@ static noinline int replay_one_name(struct btrfs_trans_handle *trans,
 	struct inode *dir;
 	u8 log_type;
 	int exists;
-	int ret;
+	int ret = 0;
+	bool update_size = (key->type == BTRFS_DIR_INDEX_KEY);
 
 	dir = read_one_inode(root, key->objectid);
-	BUG_ON(!dir);
+	if (!dir)
+		return -EIO;
 
 	name_len = btrfs_dir_name_len(eb, di);
 	name = kmalloc(name_len, GFP_NOFS);
+	if (!name) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
 	log_type = btrfs_dir_type(eb, di);
 	read_extent_buffer(eb, name, (unsigned long)(di + 1),
 		   name_len);
@@ -1190,7 +1605,7 @@ static noinline int replay_one_name(struct btrfs_trans_handle *trans,
 		exists = 1;
 	else
 		exists = 0;
-	btrfs_release_path(root, path);
+	btrfs_release_path(path);
 
 	if (key->type == BTRFS_DIR_ITEM_KEY) {
 		dst_di = btrfs_lookup_dir_item(trans, root, path, key->objectid,
@@ -1201,9 +1616,11 @@ static noinline int replay_one_name(struct btrfs_trans_handle *trans,
 						     key->offset, name,
 						     name_len, 1);
 	} else {
-		BUG();
+		/* Corruption */
+		ret = -EINVAL;
+		goto out;
 	}
-	if (!dst_di || IS_ERR(dst_di)) {
+	if (IS_ERR_OR_NULL(dst_di)) {
 		/* we need a sequence number to insert, so we only
 		 * do inserts for the BTRFS_DIR_INDEX_KEY types
 		 */
@@ -1229,22 +1646,29 @@ static noinline int replay_one_name(struct btrfs_trans_handle *trans,
 		goto out;
 
 	ret = drop_one_dir_item(trans, root, path, dir, dst_di);
-	BUG_ON(ret);
+	if (ret)
+		goto out;
 
 	if (key->type == BTRFS_DIR_INDEX_KEY)
 		goto insert;
 out:
-	btrfs_release_path(root, path);
+	btrfs_release_path(path);
+	if (!ret && update_size) {
+		btrfs_i_size_write(dir, dir->i_size + name_len * 2);
+		ret = btrfs_update_inode(trans, root, dir);
+	}
 	kfree(name);
 	iput(dir);
-	return 0;
+	return ret;
 
 insert:
-	btrfs_release_path(root, path);
+	btrfs_release_path(path);
 	ret = insert_one_name(trans, root, path, key->objectid, key->offset,
 			      name, name_len, log_type, &log_key);
-
-	BUG_ON(ret && ret != -ENOENT);
+	if (ret && ret != -ENOENT)
+		goto out;
+	update_size = false;
+	ret = 0;
 	goto out;
 }
 
@@ -1271,9 +1695,12 @@ static noinline int replay_one_dir_item(struct btrfs_trans_handle *trans,
 	ptr_end = ptr + item_size;
 	while (ptr < ptr_end) {
 		di = (struct btrfs_dir_item *)ptr;
+		if (verify_dir_item(root, eb, di))
+			return -EIO;
 		name_len = btrfs_dir_name_len(eb, di);
 		ret = replay_one_name(trans, root, path, eb, di, key);
-		BUG_ON(ret);
+		if (ret)
+			return ret;
 		ptr = (unsigned long)(di + 1);
 		ptr += name_len;
 	}
@@ -1359,7 +1786,7 @@ next:
 	*end_ret = found_end;
 	ret = 0;
 out:
-	btrfs_release_path(root, path);
+	btrfs_release_path(path);
 	return ret;
 }
 
@@ -1397,6 +1824,11 @@ again:
 	ptr_end = ptr + item_size;
 	while (ptr < ptr_end) {
 		di = (struct btrfs_dir_item *)ptr;
+		if (verify_dir_item(root, eb, di)) {
+			ret = -EIO;
+			goto out;
+		}
+
 		name_len = btrfs_dir_name_len(eb, di);
 		name = kmalloc(name_len, GFP_NOFS);
 		if (!name) {
@@ -1417,22 +1849,33 @@ again:
 						     dir_key->offset,
 						     name, name_len, 0);
 		}
-		if (!log_di || IS_ERR(log_di)) {
+		if (!log_di || (IS_ERR(log_di) && PTR_ERR(log_di) == -ENOENT)) {
 			btrfs_dir_item_key_to_cpu(eb, di, &location);
-			btrfs_release_path(root, path);
-			btrfs_release_path(log, log_path);
+			btrfs_release_path(path);
+			btrfs_release_path(log_path);
 			inode = read_one_inode(root, location.objectid);
-			BUG_ON(!inode);
+			if (!inode) {
+				kfree(name);
+				return -EIO;
+			}
 
 			ret = link_to_fixup_dir(trans, root,
 						path, location.objectid);
-			BUG_ON(ret);
-			btrfs_inc_nlink(inode);
+			if (ret) {
+				kfree(name);
+				iput(inode);
+				goto out;
+			}
+
+			inc_nlink(inode);
 			ret = btrfs_unlink_inode(trans, root, dir, inode,
 						 name, name_len);
-			BUG_ON(ret);
+			if (!ret)
+				ret = btrfs_run_delayed_items(trans, root);
 			kfree(name);
 			iput(inode);
+			if (ret)
+				goto out;
 
 			/* there might still be more names under this key
 			 * check and repeat if required
@@ -1443,8 +1886,11 @@ again:
 				goto again;
 			ret = 0;
 			goto out;
+		} else if (IS_ERR(log_di)) {
+			kfree(name);
+			return PTR_ERR(log_di);
 		}
-		btrfs_release_path(log, log_path);
+		btrfs_release_path(log_path);
 		kfree(name);
 
 		ptr = (unsigned long)(di + 1);
@@ -1452,8 +1898,8 @@ again:
 	}
 	ret = 0;
 out:
-	btrfs_release_path(root, path);
-	btrfs_release_path(log, log_path);
+	btrfs_release_path(path);
+	btrfs_release_path(log_path);
 	return ret;
 }
 
@@ -1536,12 +1982,13 @@ again:
 			ret = check_item_in_log(trans, root, log, path,
 						log_path, dir,
 						&found_key);
-			BUG_ON(ret);
+			if (ret)
+				goto out;
 			if (found_key.offset == (u64)-1)
 				break;
 			dir_key.offset = found_key.offset + 1;
 		}
-		btrfs_release_path(root, path);
+		btrfs_release_path(path);
 		if (range_end == (u64)-1)
 			break;
 		range_start = range_end + 1;
@@ -1552,11 +1999,11 @@ next_type:
 	if (key_type == BTRFS_DIR_LOG_ITEM_KEY) {
 		key_type = BTRFS_DIR_LOG_INDEX_KEY;
 		dir_key.type = BTRFS_DIR_INDEX_KEY;
-		btrfs_release_path(root, path);
+		btrfs_release_path(path);
 		goto again;
 	}
 out:
-	btrfs_release_path(root, path);
+	btrfs_release_path(path);
 	btrfs_free_path(log_path);
 	iput(dir);
 	return ret;
@@ -1580,12 +2027,13 @@ static int replay_one_buffer(struct btrfs_root *log, struct extent_buffer *eb,
 	struct btrfs_path *path;
 	struct btrfs_root *root = wc->replay_dest;
 	struct btrfs_key key;
-	u32 item_size;
 	int level;
 	int i;
 	int ret;
 
-	btrfs_read_buffer(eb, gen);
+	ret = btrfs_read_buffer(eb, gen);
+	if (ret)
+		return ret;
 
 	level = btrfs_header_level(eb);
 
@@ -1593,12 +2041,12 @@ static int replay_one_buffer(struct btrfs_root *log, struct extent_buffer *eb,
 		return 0;
 
 	path = btrfs_alloc_path();
-	BUG_ON(!path);
+	if (!path)
+		return -ENOMEM;
 
 	nritems = btrfs_header_nritems(eb);
 	for (i = 0; i < nritems; i++) {
 		btrfs_item_key_to_cpu(eb, &key, i);
-		item_size = btrfs_item_size_nr(eb, i);
 
 		/* inode keys are done during the first stage */
 		if (key.type == BTRFS_INODE_ITEM_KEY &&
@@ -1612,11 +2060,13 @@ static int replay_one_buffer(struct btrfs_root *log, struct extent_buffer *eb,
 			if (S_ISDIR(mode)) {
 				ret = replay_dir_deletes(wc->trans,
 					 root, log, path, key.objectid, 0);
-				BUG_ON(ret);
+				if (ret)
+					break;
 			}
 			ret = overwrite_item(wc->trans, root, path,
 					     eb, i, &key);
-			BUG_ON(ret);
+			if (ret)
+				break;
 
 			/* for regular files, make sure corresponding
 			 * orhpan item exist. extents past the new EOF
@@ -1625,13 +2075,24 @@ static int replay_one_buffer(struct btrfs_root *log, struct extent_buffer *eb,
 			if (S_ISREG(mode)) {
 				ret = insert_orphan_item(wc->trans, root,
 							 key.objectid);
-				BUG_ON(ret);
+				if (ret)
+					break;
 			}
 
 			ret = link_to_fixup_dir(wc->trans, root,
 						path, key.objectid);
-			BUG_ON(ret);
+			if (ret)
+				break;
 		}
+
+		if (key.type == BTRFS_DIR_INDEX_KEY &&
+		    wc->stage == LOG_WALK_REPLAY_DIR_INDEX) {
+			ret = replay_one_dir_item(wc->trans, root, path,
+						  eb, i, &key);
+			if (ret)
+				break;
+		}
+
 		if (wc->stage < LOG_WALK_REPLAY_ALL)
 			continue;
 
@@ -1639,24 +2100,29 @@ static int replay_one_buffer(struct btrfs_root *log, struct extent_buffer *eb,
 		if (key.type == BTRFS_XATTR_ITEM_KEY) {
 			ret = overwrite_item(wc->trans, root, path,
 					     eb, i, &key);
-			BUG_ON(ret);
-		} else if (key.type == BTRFS_INODE_REF_KEY) {
+			if (ret)
+				break;
+		} else if (key.type == BTRFS_INODE_REF_KEY ||
+			   key.type == BTRFS_INODE_EXTREF_KEY) {
 			ret = add_inode_ref(wc->trans, root, log, path,
 					    eb, i, &key);
-			BUG_ON(ret && ret != -ENOENT);
+			if (ret && ret != -ENOENT)
+				break;
+			ret = 0;
 		} else if (key.type == BTRFS_EXTENT_DATA_KEY) {
 			ret = replay_one_extent(wc->trans, root, path,
 						eb, i, &key);
-			BUG_ON(ret);
-		} else if (key.type == BTRFS_DIR_ITEM_KEY ||
-			   key.type == BTRFS_DIR_INDEX_KEY) {
+			if (ret)
+				break;
+		} else if (key.type == BTRFS_DIR_ITEM_KEY) {
 			ret = replay_one_dir_item(wc->trans, root, path,
 						  eb, i, &key);
-			BUG_ON(ret);
+			if (ret)
+				break;
 		}
 	}
 	btrfs_free_path(path);
-	return 0;
+	return ret;
 }
 
 static noinline int walk_down_log_tree(struct btrfs_trans_handle *trans,
@@ -1665,7 +2131,6 @@ static noinline int walk_down_log_tree(struct btrfs_trans_handle *trans,
 				   struct walk_control *wc)
 {
 	u64 root_owner;
-	u64 root_gen;
 	u64 bytenr;
 	u64 ptr_gen;
 	struct extent_buffer *next;
@@ -1682,8 +2147,7 @@ static noinline int walk_down_log_tree(struct btrfs_trans_handle *trans,
 		WARN_ON(*level >= BTRFS_MAX_LEVEL);
 		cur = path->nodes[*level];
 
-		if (btrfs_header_level(cur) != *level)
-			WARN_ON(1);
+		WARN_ON(btrfs_header_level(cur) != *level);
 
 		if (path->slots[*level] >=
 		    btrfs_header_nritems(cur))
@@ -1695,33 +2159,51 @@ static noinline int walk_down_log_tree(struct btrfs_trans_handle *trans,
 
 		parent = path->nodes[*level];
 		root_owner = btrfs_header_owner(parent);
-		root_gen = btrfs_header_generation(parent);
 
 		next = btrfs_find_create_tree_block(root, bytenr, blocksize);
-
-		wc->process_func(root, next, wc, ptr_gen);
+		if (!next)
+			return -ENOMEM;
 
 		if (*level == 1) {
+			ret = wc->process_func(root, next, wc, ptr_gen);
+			if (ret) {
+				free_extent_buffer(next);
+				return ret;
+			}
+
 			path->slots[*level]++;
 			if (wc->free) {
-				btrfs_read_buffer(next, ptr_gen);
+				ret = btrfs_read_buffer(next, ptr_gen);
+				if (ret) {
+					free_extent_buffer(next);
+					return ret;
+				}
 
-				btrfs_tree_lock(next);
-				clean_tree_block(trans, root, next);
-				btrfs_set_lock_blocking(next);
-				btrfs_wait_tree_block_writeback(next);
-				btrfs_tree_unlock(next);
+				if (trans) {
+					btrfs_tree_lock(next);
+					btrfs_set_lock_blocking(next);
+					clean_tree_block(trans, root, next);
+					btrfs_wait_tree_block_writeback(next);
+					btrfs_tree_unlock(next);
+				}
 
 				WARN_ON(root_owner !=
 					BTRFS_TREE_LOG_OBJECTID);
-				ret = btrfs_free_reserved_extent(root,
+				ret = btrfs_free_and_pin_reserved_extent(root,
 							 bytenr, blocksize);
-				BUG_ON(ret);
+				if (ret) {
+					free_extent_buffer(next);
+					return ret;
+				}
 			}
 			free_extent_buffer(next);
 			continue;
 		}
-		btrfs_read_buffer(next, ptr_gen);
+		ret = btrfs_read_buffer(next, ptr_gen);
+		if (ret) {
+			free_extent_buffer(next);
+			return ret;
+		}
 
 		WARN_ON(*level <= 0);
 		if (path->nodes[*level-1])
@@ -1734,35 +2216,7 @@ static noinline int walk_down_log_tree(struct btrfs_trans_handle *trans,
 	WARN_ON(*level < 0);
 	WARN_ON(*level >= BTRFS_MAX_LEVEL);
 
-	if (path->nodes[*level] == root->node)
-		parent = path->nodes[*level];
-	else
-		parent = path->nodes[*level + 1];
-
-	bytenr = path->nodes[*level]->start;
-
-	blocksize = btrfs_level_size(root, *level);
-	root_owner = btrfs_header_owner(parent);
-	root_gen = btrfs_header_generation(parent);
-
-	wc->process_func(root, path->nodes[*level], wc,
-			 btrfs_header_generation(path->nodes[*level]));
-
-	if (wc->free) {
-		next = path->nodes[*level];
-		btrfs_tree_lock(next);
-		clean_tree_block(trans, root, next);
-		btrfs_set_lock_blocking(next);
-		btrfs_wait_tree_block_writeback(next);
-		btrfs_tree_unlock(next);
-
-		WARN_ON(root_owner != BTRFS_TREE_LOG_OBJECTID);
-		ret = btrfs_free_reserved_extent(root, bytenr, blocksize);
-		BUG_ON(ret);
-	}
-	free_extent_buffer(path->nodes[*level]);
-	path->nodes[*level] = NULL;
-	*level += 1;
+	path->slots[*level] = btrfs_header_nritems(path->nodes[*level]);
 
 	cond_resched();
 	return 0;
@@ -1774,16 +2228,13 @@ static noinline int walk_up_log_tree(struct btrfs_trans_handle *trans,
 				 struct walk_control *wc)
 {
 	u64 root_owner;
-	u64 root_gen;
 	int i;
 	int slot;
 	int ret;
 
 	for (i = *level; i < BTRFS_MAX_LEVEL - 1 && path->nodes[i]; i++) {
 		slot = path->slots[i];
-		if (slot < btrfs_header_nritems(path->nodes[i]) - 1) {
-			struct extent_buffer *node;
-			node = path->nodes[i];
+		if (slot + 1 < btrfs_header_nritems(path->nodes[i])) {
 			path->slots[i]++;
 			*level = i;
 			WARN_ON(*level == 0);
@@ -1796,25 +2247,30 @@ static noinline int walk_up_log_tree(struct btrfs_trans_handle *trans,
 				parent = path->nodes[*level + 1];
 
 			root_owner = btrfs_header_owner(parent);
-			root_gen = btrfs_header_generation(parent);
-			wc->process_func(root, path->nodes[*level], wc,
+			ret = wc->process_func(root, path->nodes[*level], wc,
 				 btrfs_header_generation(path->nodes[*level]));
+			if (ret)
+				return ret;
+
 			if (wc->free) {
 				struct extent_buffer *next;
 
 				next = path->nodes[*level];
 
-				btrfs_tree_lock(next);
-				clean_tree_block(trans, root, next);
-				btrfs_set_lock_blocking(next);
-				btrfs_wait_tree_block_writeback(next);
-				btrfs_tree_unlock(next);
+				if (trans) {
+					btrfs_tree_lock(next);
+					btrfs_set_lock_blocking(next);
+					clean_tree_block(trans, root, next);
+					btrfs_wait_tree_block_writeback(next);
+					btrfs_tree_unlock(next);
+				}
 
 				WARN_ON(root_owner != BTRFS_TREE_LOG_OBJECTID);
-				ret = btrfs_free_reserved_extent(root,
+				ret = btrfs_free_and_pin_reserved_extent(root,
 						path->nodes[*level]->start,
 						path->nodes[*level]->len);
-				BUG_ON(ret);
+				if (ret)
+					return ret;
 			}
 			free_extent_buffer(path->nodes[*level]);
 			path->nodes[*level] = NULL;
@@ -1836,11 +2292,11 @@ static int walk_log_tree(struct btrfs_trans_handle *trans,
 	int wret;
 	int level;
 	struct btrfs_path *path;
-	int i;
 	int orig_level;
 
 	path = btrfs_alloc_path();
-	BUG_ON(!path);
+	if (!path)
+		return -ENOMEM;
 
 	level = btrfs_header_level(log->node);
 	orig_level = level;
@@ -1852,45 +2308,49 @@ static int walk_log_tree(struct btrfs_trans_handle *trans,
 		wret = walk_down_log_tree(trans, log, path, &level, wc);
 		if (wret > 0)
 			break;
-		if (wret < 0)
+		if (wret < 0) {
 			ret = wret;
+			goto out;
+		}
 
 		wret = walk_up_log_tree(trans, log, path, &level, wc);
 		if (wret > 0)
 			break;
-		if (wret < 0)
+		if (wret < 0) {
 			ret = wret;
+			goto out;
+		}
 	}
 
 	/* was the root node processed? if not, catch it here */
 	if (path->nodes[orig_level]) {
-		wc->process_func(log, path->nodes[orig_level], wc,
+		ret = wc->process_func(log, path->nodes[orig_level], wc,
 			 btrfs_header_generation(path->nodes[orig_level]));
+		if (ret)
+			goto out;
 		if (wc->free) {
 			struct extent_buffer *next;
 
 			next = path->nodes[orig_level];
 
-			btrfs_tree_lock(next);
-			clean_tree_block(trans, log, next);
-			btrfs_set_lock_blocking(next);
-			btrfs_wait_tree_block_writeback(next);
-			btrfs_tree_unlock(next);
+			if (trans) {
+				btrfs_tree_lock(next);
+				btrfs_set_lock_blocking(next);
+				clean_tree_block(trans, log, next);
+				btrfs_wait_tree_block_writeback(next);
+				btrfs_tree_unlock(next);
+			}
 
 			WARN_ON(log->root_key.objectid !=
 				BTRFS_TREE_LOG_OBJECTID);
-			ret = btrfs_free_reserved_extent(log, next->start,
+			ret = btrfs_free_and_pin_reserved_extent(log, next->start,
 							 next->len);
-			BUG_ON(ret);
+			if (ret)
+				goto out;
 		}
 	}
 
-	for (i = 0; i <= orig_level; i++) {
-		if (path->nodes[i]) {
-			free_extent_buffer(path->nodes[i]);
-			path->nodes[i] = NULL;
-		}
-	}
+out:
 	btrfs_free_path(path);
 	return ret;
 }
@@ -1915,8 +2375,8 @@ static int update_log_root(struct btrfs_trans_handle *trans,
 	return ret;
 }
 
-static int wait_log_commit(struct btrfs_trans_handle *trans,
-			   struct btrfs_root *root, unsigned long transid)
+static void wait_log_commit(struct btrfs_trans_handle *trans,
+			    struct btrfs_root *root, int transid)
 {
 	DEFINE_WAIT(wait);
 	int index = transid % 2;
@@ -1931,33 +2391,61 @@ static int wait_log_commit(struct btrfs_trans_handle *trans,
 				&wait, TASK_UNINTERRUPTIBLE);
 		mutex_unlock(&root->log_mutex);
 
-		if (root->fs_info->last_trans_log_full_commit !=
-		    trans->transid && root->log_transid < transid + 2 &&
+		if (root->log_transid_committed < transid &&
 		    atomic_read(&root->log_commit[index]))
 			schedule();
 
 		finish_wait(&root->log_commit_wait[index], &wait);
 		mutex_lock(&root->log_mutex);
-	} while (root->log_transid < transid + 2 &&
+	} while (root->log_transid_committed < transid &&
 		 atomic_read(&root->log_commit[index]));
-	return 0;
 }
 
-static int wait_for_writer(struct btrfs_trans_handle *trans,
-			   struct btrfs_root *root)
+static void wait_for_writer(struct btrfs_trans_handle *trans,
+			    struct btrfs_root *root)
 {
 	DEFINE_WAIT(wait);
+
 	while (atomic_read(&root->log_writers)) {
 		prepare_to_wait(&root->log_writer_wait,
 				&wait, TASK_UNINTERRUPTIBLE);
 		mutex_unlock(&root->log_mutex);
-		if (root->fs_info->last_trans_log_full_commit !=
-		    trans->transid && atomic_read(&root->log_writers))
+		if (atomic_read(&root->log_writers))
 			schedule();
 		mutex_lock(&root->log_mutex);
 		finish_wait(&root->log_writer_wait, &wait);
 	}
-	return 0;
+}
+
+static inline void btrfs_remove_log_ctx(struct btrfs_root *root,
+					struct btrfs_log_ctx *ctx)
+{
+	if (!ctx)
+		return;
+
+	mutex_lock(&root->log_mutex);
+	list_del_init(&ctx->list);
+	mutex_unlock(&root->log_mutex);
+}
+
+/* 
+ * Invoked in log mutex context, or be sure there is no other task which
+ * can access the list.
+ */
+static inline void btrfs_remove_all_log_ctxs(struct btrfs_root *root,
+					     int index, int error)
+{
+	struct btrfs_log_ctx *ctx;
+
+	if (!error) {
+		INIT_LIST_HEAD(&root->log_ctxs[index]);
+		return;
+	}
+
+	list_for_each_entry(ctx, &root->log_ctxs[index], list)
+		ctx->log_ret = error;
+
+	INIT_LIST_HEAD(&root->log_ctxs[index]);
 }
 
 /*
@@ -1973,7 +2461,7 @@ static int wait_for_writer(struct btrfs_trans_handle *trans,
  * that has happened.
  */
 int btrfs_sync_log(struct btrfs_trans_handle *trans,
-		   struct btrfs_root *root)
+		   struct btrfs_root *root, struct btrfs_log_ctx *ctx)
 {
 	int index1;
 	int index2;
@@ -1981,41 +2469,52 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
 	int ret;
 	struct btrfs_root *log = root->log_root;
 	struct btrfs_root *log_root_tree = root->fs_info->log_root_tree;
-	unsigned long log_transid = 0;
+	int log_transid = 0;
+	struct btrfs_log_ctx root_log_ctx;
+	struct blk_plug plug;
 
 	mutex_lock(&root->log_mutex);
-	index1 = root->log_transid % 2;
+	log_transid = ctx->log_transid;
+	if (root->log_transid_committed >= log_transid) {
+		mutex_unlock(&root->log_mutex);
+		return ctx->log_ret;
+	}
+
+	index1 = log_transid % 2;
 	if (atomic_read(&root->log_commit[index1])) {
-		wait_log_commit(trans, root, root->log_transid);
+		wait_log_commit(trans, root, log_transid);
 		mutex_unlock(&root->log_mutex);
-		return 0;
+		return ctx->log_ret;
 	}
+	ASSERT(log_transid == root->log_transid);
 	atomic_set(&root->log_commit[index1], 1);
 
 	/* wait for previous tree log sync to complete */
 	if (atomic_read(&root->log_commit[(index1 + 1) % 2]))
-		wait_log_commit(trans, root, root->log_transid - 1);
+		wait_log_commit(trans, root, log_transid - 1);
 
 	while (1) {
-		unsigned long batch = root->log_batch;
-		if (root->log_multiple_pids) {
+		int batch = atomic_read(&root->log_batch);
+		/* when we're on an ssd, just kick the log commit out */
+		if (!btrfs_test_opt(root, SSD) &&
+		    test_bit(BTRFS_ROOT_MULTI_LOG_TASKS, &root->state)) {
 			mutex_unlock(&root->log_mutex);
 			schedule_timeout_uninterruptible(1);
 			mutex_lock(&root->log_mutex);
 		}
 		wait_for_writer(trans, root);
-		if (batch == root->log_batch)
+		if (batch == atomic_read(&root->log_batch))
 			break;
 	}
 
 	/* bail out if we need to do a full commit */
-	if (root->fs_info->last_trans_log_full_commit == trans->transid) {
+	if (btrfs_need_log_full_commit(root->fs_info, trans)) {
 		ret = -EAGAIN;
+		btrfs_free_logged_extents(log, log_transid);
 		mutex_unlock(&root->log_mutex);
 		goto out;
 	}
 
-	log_transid = root->log_transid;
 	if (log_transid % 2 == 0)
 		mark = EXTENT_DIRTY;
 	else
@@ -2024,16 +2523,22 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
 	/* we start IO on  all the marked extents here, but we don't actually
 	 * wait for them until later.
 	 */
+	blk_start_plug(&plug);
 	ret = btrfs_write_marked_extents(log, &log->dirty_log_pages, mark);
-	BUG_ON(ret);
+	if (ret) {
+		blk_finish_plug(&plug);
+		btrfs_abort_transaction(trans, root, ret);
+		btrfs_free_logged_extents(log, log_transid);
+		btrfs_set_log_full_commit(root->fs_info, trans);
+		mutex_unlock(&root->log_mutex);
+		goto out;
+	}
 
 	btrfs_set_root_node(&log->root_item, log->node);
 
-	root->log_batch = 0;
 	root->log_transid++;
 	log->log_transid = root->log_transid;
 	root->log_start_pid = 0;
-	smp_mb();
 	/*
 	 * IO has been started, blocks of the log tree have WRITTEN flag set
 	 * in their headers. new modifications of the log will be written to
@@ -2041,13 +2546,19 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
 	 */
 	mutex_unlock(&root->log_mutex);
 
+	btrfs_init_log_ctx(&root_log_ctx);
+
 	mutex_lock(&log_root_tree->log_mutex);
-	log_root_tree->log_batch++;
+	atomic_inc(&log_root_tree->log_batch);
 	atomic_inc(&log_root_tree->log_writers);
+
+	index2 = log_root_tree->log_transid % 2;
+	list_add_tail(&root_log_ctx.list, &log_root_tree->log_ctxs[index2]);
+	root_log_ctx.log_transid = log_root_tree->log_transid;
+
 	mutex_unlock(&log_root_tree->log_mutex);
 
 	ret = update_log_root(trans, log);
-	BUG_ON(ret);
 
 	mutex_lock(&log_root_tree->log_mutex);
 	if (atomic_dec_and_test(&log_root_tree->log_writers)) {
@@ -2056,19 +2567,48 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
 			wake_up(&log_root_tree->log_writer_wait);
 	}
 
-	index2 = log_root_tree->log_transid % 2;
+	if (ret) {
+		if (!list_empty(&root_log_ctx.list))
+			list_del_init(&root_log_ctx.list);
+
+		blk_finish_plug(&plug);
+		btrfs_set_log_full_commit(root->fs_info, trans);
+
+		if (ret != -ENOSPC) {
+			btrfs_abort_transaction(trans, root, ret);
+			mutex_unlock(&log_root_tree->log_mutex);
+			goto out;
+		}
+		btrfs_wait_marked_extents(log, &log->dirty_log_pages, mark);
+		btrfs_free_logged_extents(log, log_transid);
+		mutex_unlock(&log_root_tree->log_mutex);
+		ret = -EAGAIN;
+		goto out;
+	}
+
+	if (log_root_tree->log_transid_committed >= root_log_ctx.log_transid) {
+		mutex_unlock(&log_root_tree->log_mutex);
+		ret = root_log_ctx.log_ret;
+		goto out;
+	}
+
+	index2 = root_log_ctx.log_transid % 2;
 	if (atomic_read(&log_root_tree->log_commit[index2])) {
+		blk_finish_plug(&plug);
 		btrfs_wait_marked_extents(log, &log->dirty_log_pages, mark);
 		wait_log_commit(trans, log_root_tree,
-				log_root_tree->log_transid);
+				root_log_ctx.log_transid);
+		btrfs_free_logged_extents(log, log_transid);
 		mutex_unlock(&log_root_tree->log_mutex);
+		ret = root_log_ctx.log_ret;
 		goto out;
 	}
+	ASSERT(root_log_ctx.log_transid == log_root_tree->log_transid);
 	atomic_set(&log_root_tree->log_commit[index2], 1);
 
 	if (atomic_read(&log_root_tree->log_commit[(index2 + 1) % 2])) {
 		wait_log_commit(trans, log_root_tree,
-				log_root_tree->log_transid - 1);
+				root_log_ctx.log_transid - 1);
 	}
 
 	wait_for_writer(trans, log_root_tree);
@@ -2077,28 +2617,38 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
 	 * now that we've moved on to the tree of log tree roots,
 	 * check the full commit flag again
 	 */
-	if (root->fs_info->last_trans_log_full_commit == trans->transid) {
+	if (btrfs_need_log_full_commit(root->fs_info, trans)) {
+		blk_finish_plug(&plug);
 		btrfs_wait_marked_extents(log, &log->dirty_log_pages, mark);
+		btrfs_free_logged_extents(log, log_transid);
 		mutex_unlock(&log_root_tree->log_mutex);
 		ret = -EAGAIN;
 		goto out_wake_log_root;
 	}
 
-	ret = btrfs_write_and_wait_marked_extents(log_root_tree,
-				&log_root_tree->dirty_log_pages,
-				EXTENT_DIRTY | EXTENT_NEW);
-	BUG_ON(ret);
+	ret = btrfs_write_marked_extents(log_root_tree,
+					 &log_root_tree->dirty_log_pages,
+					 EXTENT_DIRTY | EXTENT_NEW);
+	blk_finish_plug(&plug);
+	if (ret) {
+		btrfs_set_log_full_commit(root->fs_info, trans);
+		btrfs_abort_transaction(trans, root, ret);
+		btrfs_free_logged_extents(log, log_transid);
+		mutex_unlock(&log_root_tree->log_mutex);
+		goto out_wake_log_root;
+	}
 	btrfs_wait_marked_extents(log, &log->dirty_log_pages, mark);
+	btrfs_wait_marked_extents(log_root_tree,
+				  &log_root_tree->dirty_log_pages,
+				  EXTENT_NEW | EXTENT_DIRTY);
+	btrfs_wait_logged_extents(log, log_transid);
 
-	btrfs_set_super_log_root(&root->fs_info->super_for_commit,
+	btrfs_set_super_log_root(root->fs_info->super_for_commit,
 				log_root_tree->node->start);
-	btrfs_set_super_log_root_level(&root->fs_info->super_for_commit,
+	btrfs_set_super_log_root_level(root->fs_info->super_for_commit,
 				btrfs_header_level(log_root_tree->node));
 
-	log_root_tree->log_batch = 0;
 	log_root_tree->log_transid++;
-	smp_mb();
-
 	mutex_unlock(&log_root_tree->log_mutex);
 
 	/*
@@ -2108,8 +2658,12 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
 	 * the running transaction open, so a full commit can't hop
 	 * in and cause problems either.
 	 */
-	write_ctree_super(trans, root->fs_info->tree_root, 1);
-	ret = 0;
+	ret = write_ctree_super(trans, root->fs_info->tree_root, 1);
+	if (ret) {
+		btrfs_set_log_full_commit(root->fs_info, trans);
+		btrfs_abort_transaction(trans, root, ret);
+		goto out_wake_log_root;
+	}
 
 	mutex_lock(&root->log_mutex);
 	if (root->last_log_commit < log_transid)
@@ -2117,27 +2671,37 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
 	mutex_unlock(&root->log_mutex);
 
 out_wake_log_root:
+	/*
+	 * We needn't get log_mutex here because we are sure all
+	 * the other tasks are blocked.
+	 */
+	btrfs_remove_all_log_ctxs(log_root_tree, index2, ret);
+
+	mutex_lock(&log_root_tree->log_mutex);
+	log_root_tree->log_transid_committed++;
 	atomic_set(&log_root_tree->log_commit[index2], 0);
-	smp_mb();
+	mutex_unlock(&log_root_tree->log_mutex);
+
 	if (waitqueue_active(&log_root_tree->log_commit_wait[index2]))
 		wake_up(&log_root_tree->log_commit_wait[index2]);
 out:
+	/* See above. */
+	btrfs_remove_all_log_ctxs(root, index1, ret);
+
+	mutex_lock(&root->log_mutex);
+	root->log_transid_committed++;
 	atomic_set(&root->log_commit[index1], 0);
-	smp_mb();
+	mutex_unlock(&root->log_mutex);
+
 	if (waitqueue_active(&root->log_commit_wait[index1]))
 		wake_up(&root->log_commit_wait[index1]);
-	return 0;
+	return ret;
 }
 
-/*
- * free all the extents used by the tree log.  This should be called
- * at commit time of the full transaction
- */
-int btrfs_free_log(struct btrfs_trans_handle *trans, struct btrfs_root *root)
+static void free_log_tree(struct btrfs_trans_handle *trans,
+			  struct btrfs_root *log)
 {
 	int ret;
-	struct btrfs_root *log;
-	struct key;
 	u64 start;
 	u64 end;
 	struct walk_control wc = {
@@ -2145,16 +2709,15 @@ int btrfs_free_log(struct btrfs_trans_handle *trans, struct btrfs_root *root)
 		.process_func = process_one_buffer
 	};
 
-	if (!root->log_root || root->fs_info->log_root_recovering)
-		return 0;
-
-	log = root->log_root;
 	ret = walk_log_tree(trans, log, &wc);
-	BUG_ON(ret);
+	/* I don't think this can happen but just in case */
+	if (ret)
+		btrfs_abort_transaction(trans, log, ret);
 
 	while (1) {
 		ret = find_first_extent_bit(&log->dirty_log_pages,
-				0, &start, &end, EXTENT_DIRTY | EXTENT_NEW);
+				0, &start, &end, EXTENT_DIRTY | EXTENT_NEW,
+				NULL);
 		if (ret)
 			break;
 
@@ -2162,14 +2725,38 @@ int btrfs_free_log(struct btrfs_trans_handle *trans, struct btrfs_root *root)
 				  EXTENT_DIRTY | EXTENT_NEW, GFP_NOFS);
 	}
 
-	if (log->log_transid > 0) {
-		ret = btrfs_del_root(trans, root->fs_info->log_root_tree,
-				     &log->root_key);
-		BUG_ON(ret);
-	}
-	root->log_root = NULL;
+	/*
+	 * We may have short-circuited the log tree with the full commit logic
+	 * and left ordered extents on our list, so clear these out to keep us
+	 * from leaking inodes and memory.
+	 */
+	btrfs_free_logged_extents(log, 0);
+	btrfs_free_logged_extents(log, 1);
+
 	free_extent_buffer(log->node);
 	kfree(log);
+}
+
+/*
+ * free all the extents used by the tree log.  This should be called
+ * at commit time of the full transaction
+ */
+int btrfs_free_log(struct btrfs_trans_handle *trans, struct btrfs_root *root)
+{
+	if (root->log_root) {
+		free_log_tree(trans, root->log_root);
+		root->log_root = NULL;
+	}
+	return 0;
+}
+
+int btrfs_free_log_root_tree(struct btrfs_trans_handle *trans,
+			     struct btrfs_fs_info *fs_info)
+{
+	if (fs_info->log_root_tree) {
+		free_log_tree(trans, fs_info->log_root_tree);
+		fs_info->log_root_tree = NULL;
+	}
 	return 0;
 }
 
@@ -2203,7 +2790,9 @@ int btrfs_del_dir_entries_in_log(struct btrfs_trans_handle *trans,
 	struct btrfs_dir_item *di;
 	struct btrfs_path *path;
 	int ret;
+	int err = 0;
 	int bytes_del = 0;
+	u64 dir_ino = btrfs_ino(dir);
 
 	if (BTRFS_I(dir)->logged_trans < trans->transid)
 		return 0;
@@ -2216,20 +2805,39 @@ int btrfs_del_dir_entries_in_log(struct btrfs_trans_handle *trans,
 
 	log = root->log_root;
 	path = btrfs_alloc_path();
-	di = btrfs_lookup_dir_item(trans, log, path, dir->i_ino,
+	if (!path) {
+		err = -ENOMEM;
+		goto out_unlock;
+	}
+
+	di = btrfs_lookup_dir_item(trans, log, path, dir_ino,
 				   name, name_len, -1);
-	if (di && !IS_ERR(di)) {
+	if (IS_ERR(di)) {
+		err = PTR_ERR(di);
+		goto fail;
+	}
+	if (di) {
 		ret = btrfs_delete_one_dir_name(trans, log, path, di);
 		bytes_del += name_len;
-		BUG_ON(ret);
+		if (ret) {
+			err = ret;
+			goto fail;
+		}
 	}
-	btrfs_release_path(log, path);
-	di = btrfs_lookup_dir_index_item(trans, log, path, dir->i_ino,
+	btrfs_release_path(path);
+	di = btrfs_lookup_dir_index_item(trans, log, path, dir_ino,
 					 index, name, name_len, -1);
-	if (di && !IS_ERR(di)) {
+	if (IS_ERR(di)) {
+		err = PTR_ERR(di);
+		goto fail;
+	}
+	if (di) {
 		ret = btrfs_delete_one_dir_name(trans, log, path, di);
 		bytes_del += name_len;
-		BUG_ON(ret);
+		if (ret) {
+			err = ret;
+			goto fail;
+		}
 	}
 
 	/* update the directory size in the log to reflect the names
@@ -2238,12 +2846,16 @@ int btrfs_del_dir_entries_in_log(struct btrfs_trans_handle *trans,
 	if (bytes_del) {
 		struct btrfs_key key;
 
-		key.objectid = dir->i_ino;
+		key.objectid = dir_ino;
 		key.offset = 0;
 		key.type = BTRFS_INODE_ITEM_KEY;
-		btrfs_release_path(log, path);
+		btrfs_release_path(path);
 
 		ret = btrfs_search_slot(trans, log, &key, path, 0, 1);
+		if (ret < 0) {
+			err = ret;
+			goto fail;
+		}
 		if (ret == 0) {
 			struct btrfs_inode_item *item;
 			u64 i_size;
@@ -2259,14 +2871,21 @@ int btrfs_del_dir_entries_in_log(struct btrfs_trans_handle *trans,
 			btrfs_mark_buffer_dirty(path->nodes[0]);
 		} else
 			ret = 0;
-		btrfs_release_path(log, path);
+		btrfs_release_path(path);
 	}
-
+fail:
 	btrfs_free_path(path);
+out_unlock:
 	mutex_unlock(&BTRFS_I(dir)->log_mutex);
+	if (ret == -ENOSPC) {
+		btrfs_set_log_full_commit(root->fs_info, trans);
+		ret = 0;
+	} else if (ret < 0)
+		btrfs_abort_transaction(trans, root, ret);
+
 	btrfs_end_log_trans(root);
 
-	return 0;
+	return err;
 }
 
 /* see comments for btrfs_del_dir_entries_in_log */
@@ -2288,9 +2907,14 @@ int btrfs_del_inode_ref_in_log(struct btrfs_trans_handle *trans,
 	log = root->log_root;
 	mutex_lock(&BTRFS_I(inode)->log_mutex);
 
-	ret = btrfs_del_inode_ref(trans, log, name, name_len, inode->i_ino,
+	ret = btrfs_del_inode_ref(trans, log, name, name_len, btrfs_ino(inode),
 				  dirid, &index);
 	mutex_unlock(&BTRFS_I(inode)->log_mutex);
+	if (ret == -ENOSPC) {
+		btrfs_set_log_full_commit(root->fs_info, trans);
+		ret = 0;
+	} else if (ret < 0 && ret != -ENOENT)
+		btrfs_abort_transaction(trans, root, ret);
 	btrfs_end_log_trans(root);
 
 	return ret;
@@ -2318,13 +2942,14 @@ static noinline int insert_dir_log_key(struct btrfs_trans_handle *trans,
 	else
 		key.type = BTRFS_DIR_LOG_INDEX_KEY;
 	ret = btrfs_insert_empty_item(trans, log, path, &key, sizeof(*item));
-	BUG_ON(ret);
+	if (ret)
+		return ret;
 
 	item = btrfs_item_ptr(path->nodes[0], path->slots[0],
 			      struct btrfs_dir_log_item);
 	btrfs_set_dir_log_end(path->nodes[0], item, last_offset);
 	btrfs_mark_buffer_dirty(path->nodes[0]);
-	btrfs_release_path(log, path);
+	btrfs_release_path(path);
 	return 0;
 }
 
@@ -2340,45 +2965,41 @@ static noinline int log_dir_items(struct btrfs_trans_handle *trans,
 			  u64 min_offset, u64 *last_offset_ret)
 {
 	struct btrfs_key min_key;
-	struct btrfs_key max_key;
 	struct btrfs_root *log = root->log_root;
 	struct extent_buffer *src;
+	int err = 0;
 	int ret;
 	int i;
 	int nritems;
 	u64 first_offset = min_offset;
 	u64 last_offset = (u64)-1;
+	u64 ino = btrfs_ino(inode);
 
 	log = root->log_root;
-	max_key.objectid = inode->i_ino;
-	max_key.offset = (u64)-1;
-	max_key.type = key_type;
 
-	min_key.objectid = inode->i_ino;
+	min_key.objectid = ino;
 	min_key.type = key_type;
 	min_key.offset = min_offset;
 
 	path->keep_locks = 1;
 
-	ret = btrfs_search_forward(root, &min_key, &max_key,
-				   path, 0, trans->transid);
+	ret = btrfs_search_forward(root, &min_key, path, trans->transid);
 
 	/*
 	 * we didn't find anything from this transaction, see if there
 	 * is anything at all
 	 */
-	if (ret != 0 || min_key.objectid != inode->i_ino ||
-	    min_key.type != key_type) {
-		min_key.objectid = inode->i_ino;
+	if (ret != 0 || min_key.objectid != ino || min_key.type != key_type) {
+		min_key.objectid = ino;
 		min_key.type = key_type;
 		min_key.offset = (u64)-1;
-		btrfs_release_path(root, path);
+		btrfs_release_path(path);
 		ret = btrfs_search_slot(NULL, root, &min_key, path, 0, 0);
 		if (ret < 0) {
-			btrfs_release_path(root, path);
+			btrfs_release_path(path);
 			return ret;
 		}
-		ret = btrfs_previous_item(root, path, inode->i_ino, key_type);
+		ret = btrfs_previous_item(root, path, ino, key_type);
 
 		/* if ret == 0 there are items for this type,
 		 * create a range to tell us the last key of this type.
@@ -2396,7 +3017,7 @@ static noinline int log_dir_items(struct btrfs_trans_handle *trans,
 	}
 
 	/* go backward to find any previous key */
-	ret = btrfs_previous_item(root, path, inode->i_ino, key_type);
+	ret = btrfs_previous_item(root, path, ino, key_type);
 	if (ret == 0) {
 		struct btrfs_key tmp;
 		btrfs_item_key_to_cpu(path->nodes[0], &tmp, path->slots[0]);
@@ -2405,16 +3026,18 @@ static noinline int log_dir_items(struct btrfs_trans_handle *trans,
 			ret = overwrite_item(trans, log, dst_path,
 					     path->nodes[0], path->slots[0],
 					     &tmp);
+			if (ret) {
+				err = ret;
+				goto done;
+			}
 		}
 	}
-	btrfs_release_path(root, path);
+	btrfs_release_path(path);
 
 	/* find the first key from this transaction again */
 	ret = btrfs_search_slot(NULL, root, &min_key, path, 0, 0);
-	if (ret != 0) {
-		WARN_ON(1);
+	if (WARN_ON(ret != 0))
 		goto done;
-	}
 
 	/*
 	 * we have a block from this transaction, log every item in it
@@ -2427,12 +3050,14 @@ static noinline int log_dir_items(struct btrfs_trans_handle *trans,
 		for (i = path->slots[0]; i < nritems; i++) {
 			btrfs_item_key_to_cpu(src, &min_key, i);
 
-			if (min_key.objectid != inode->i_ino ||
-			    min_key.type != key_type)
+			if (min_key.objectid != ino || min_key.type != key_type)
 				goto done;
 			ret = overwrite_item(trans, log, dst_path, src, i,
 					     &min_key);
-			BUG_ON(ret);
+			if (ret) {
+				err = ret;
+				goto done;
+			}
 		}
 		path->slots[0] = nritems;
 
@@ -2446,7 +3071,7 @@ static noinline int log_dir_items(struct btrfs_trans_handle *trans,
 			goto done;
 		}
 		btrfs_item_key_to_cpu(path->nodes[0], &tmp, path->slots[0]);
-		if (tmp.objectid != inode->i_ino || tmp.type != key_type) {
+		if (tmp.objectid != ino || tmp.type != key_type) {
 			last_offset = (u64)-1;
 			goto done;
 		}
@@ -2454,22 +3079,29 @@ static noinline int log_dir_items(struct btrfs_trans_handle *trans,
 			ret = overwrite_item(trans, log, dst_path,
 					     path->nodes[0], path->slots[0],
 					     &tmp);
-
-			BUG_ON(ret);
-			last_offset = tmp.offset;
+			if (ret)
+				err = ret;
+			else
+				last_offset = tmp.offset;
 			goto done;
 		}
 	}
 done:
-	*last_offset_ret = last_offset;
-	btrfs_release_path(root, path);
-	btrfs_release_path(log, dst_path);
-
-	/* insert the log range keys to indicate where the log is valid */
-	ret = insert_dir_log_key(trans, log, path, key_type, inode->i_ino,
-				 first_offset, last_offset);
-	BUG_ON(ret);
-	return 0;
+	btrfs_release_path(path);
+	btrfs_release_path(dst_path);
+
+	if (err == 0) {
+		*last_offset_ret = last_offset;
+		/*
+		 * insert the log range keys to indicate where the log
+		 * is valid
+		 */
+		ret = insert_dir_log_key(trans, log, path, key_type,
+					 ino, first_offset, last_offset);
+		if (ret)
+			err = ret;
+	}
+	return err;
 }
 
 /*
@@ -2501,7 +3133,8 @@ again:
 		ret = log_dir_items(trans, root, inode, path,
 				    dst_path, key_type, min_key,
 				    &max_key);
-		BUG_ON(ret);
+		if (ret)
+			return ret;
 		if (max_key == (u64)-1)
 			break;
 		min_key = max_key + 1;
@@ -2528,6 +3161,7 @@ static int drop_objectid_items(struct btrfs_trans_handle *trans,
 	int ret;
 	struct btrfs_key key;
 	struct btrfs_key found_key;
+	int start_slot;
 
 	key.objectid = objectid;
 	key.type = max_key_type;
@@ -2535,8 +3169,8 @@ static int drop_objectid_items(struct btrfs_trans_handle *trans,
 
 	while (1) {
 		ret = btrfs_search_slot(trans, log, &key, path, -1, 1);
-
-		if (ret != 1)
+		BUG_ON(ret == 0); /* Logic error */
+		if (ret < 0)
 			break;
 
 		if (path->slots[0] == 0)
@@ -2549,35 +3183,133 @@ static int drop_objectid_items(struct btrfs_trans_handle *trans,
 		if (found_key.objectid != objectid)
 			break;
 
-		ret = btrfs_del_item(trans, log, path);
-		BUG_ON(ret);
-		btrfs_release_path(log, path);
+		found_key.offset = 0;
+		found_key.type = 0;
+		ret = btrfs_bin_search(path->nodes[0], &found_key, 0,
+				       &start_slot);
+
+		ret = btrfs_del_items(trans, log, path, start_slot,
+				      path->slots[0] - start_slot + 1);
+		/*
+		 * If start slot isn't 0 then we don't need to re-search, we've
+		 * found the last guy with the objectid in this tree.
+		 */
+		if (ret || start_slot != 0)
+			break;
+		btrfs_release_path(path);
 	}
-	btrfs_release_path(log, path);
+	btrfs_release_path(path);
+	if (ret > 0)
+		ret = 0;
+	return ret;
+}
+
+static void fill_inode_item(struct btrfs_trans_handle *trans,
+			    struct extent_buffer *leaf,
+			    struct btrfs_inode_item *item,
+			    struct inode *inode, int log_inode_only)
+{
+	struct btrfs_map_token token;
+
+	btrfs_init_map_token(&token);
+
+	if (log_inode_only) {
+		/* set the generation to zero so the recover code
+		 * can tell the difference between an logging
+		 * just to say 'this inode exists' and a logging
+		 * to say 'update this inode with these values'
+		 */
+		btrfs_set_token_inode_generation(leaf, item, 0, &token);
+		btrfs_set_token_inode_size(leaf, item, 0, &token);
+	} else {
+		btrfs_set_token_inode_generation(leaf, item,
+						 BTRFS_I(inode)->generation,
+						 &token);
+		btrfs_set_token_inode_size(leaf, item, inode->i_size, &token);
+	}
+
+	btrfs_set_token_inode_uid(leaf, item, i_uid_read(inode), &token);
+	btrfs_set_token_inode_gid(leaf, item, i_gid_read(inode), &token);
+	btrfs_set_token_inode_mode(leaf, item, inode->i_mode, &token);
+	btrfs_set_token_inode_nlink(leaf, item, inode->i_nlink, &token);
+
+	btrfs_set_token_timespec_sec(leaf, btrfs_inode_atime(item),
+				     inode->i_atime.tv_sec, &token);
+	btrfs_set_token_timespec_nsec(leaf, btrfs_inode_atime(item),
+				      inode->i_atime.tv_nsec, &token);
+
+	btrfs_set_token_timespec_sec(leaf, btrfs_inode_mtime(item),
+				     inode->i_mtime.tv_sec, &token);
+	btrfs_set_token_timespec_nsec(leaf, btrfs_inode_mtime(item),
+				      inode->i_mtime.tv_nsec, &token);
+
+	btrfs_set_token_timespec_sec(leaf, btrfs_inode_ctime(item),
+				     inode->i_ctime.tv_sec, &token);
+	btrfs_set_token_timespec_nsec(leaf, btrfs_inode_ctime(item),
+				      inode->i_ctime.tv_nsec, &token);
+
+	btrfs_set_token_inode_nbytes(leaf, item, inode_get_bytes(inode),
+				     &token);
+
+	btrfs_set_token_inode_sequence(leaf, item, inode->i_version, &token);
+	btrfs_set_token_inode_transid(leaf, item, trans->transid, &token);
+	btrfs_set_token_inode_rdev(leaf, item, inode->i_rdev, &token);
+	btrfs_set_token_inode_flags(leaf, item, BTRFS_I(inode)->flags, &token);
+	btrfs_set_token_inode_block_group(leaf, item, 0, &token);
+}
+
+static int log_inode_item(struct btrfs_trans_handle *trans,
+			  struct btrfs_root *log, struct btrfs_path *path,
+			  struct inode *inode)
+{
+	struct btrfs_inode_item *inode_item;
+	int ret;
+
+	ret = btrfs_insert_empty_item(trans, log, path,
+				      &BTRFS_I(inode)->location,
+				      sizeof(*inode_item));
+	if (ret && ret != -EEXIST)
+		return ret;
+	inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
+				    struct btrfs_inode_item);
+	fill_inode_item(trans, path->nodes[0], inode_item, inode, 0);
+	btrfs_release_path(path);
 	return 0;
 }
 
 static noinline int copy_items(struct btrfs_trans_handle *trans,
-			       struct btrfs_root *log,
+			       struct inode *inode,
 			       struct btrfs_path *dst_path,
-			       struct extent_buffer *src,
+			       struct btrfs_path *src_path, u64 *last_extent,
 			       int start_slot, int nr, int inode_only)
 {
 	unsigned long src_offset;
 	unsigned long dst_offset;
+	struct btrfs_root *log = BTRFS_I(inode)->root->log_root;
 	struct btrfs_file_extent_item *extent;
 	struct btrfs_inode_item *inode_item;
+	struct extent_buffer *src = src_path->nodes[0];
+	struct btrfs_key first_key, last_key, key;
 	int ret;
 	struct btrfs_key *ins_keys;
 	u32 *ins_sizes;
 	char *ins_data;
 	int i;
 	struct list_head ordered_sums;
+	int skip_csum = BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM;
+	bool has_extents = false;
+	bool need_find_last_extent = (*last_extent == 0);
+	bool done = false;
 
 	INIT_LIST_HEAD(&ordered_sums);
 
 	ins_data = kmalloc(nr * sizeof(struct btrfs_key) +
 			   nr * sizeof(u32), GFP_NOFS);
+	if (!ins_data)
+		return -ENOMEM;
+
+	first_key.objectid = (u64)-1;
+
 	ins_sizes = (u32 *)ins_data;
 	ins_keys = (struct btrfs_key *)(ins_data + nr * sizeof(u32));
 
@@ -2587,7 +3319,10 @@ static noinline int copy_items(struct btrfs_trans_handle *trans,
 	}
 	ret = btrfs_insert_empty_items(trans, log, dst_path,
 				       ins_keys, ins_sizes, nr);
-	BUG_ON(ret);
+	if (ret) {
+		kfree(ins_data);
+		return ret;
+	}
 
 	for (i = 0; i < nr; i++, dst_path->slots[0]++) {
 		dst_offset = btrfs_item_ptr_offset(dst_path->nodes[0],
@@ -2595,36 +3330,50 @@ static noinline int copy_items(struct btrfs_trans_handle *trans,
 
 		src_offset = btrfs_item_ptr_offset(src, start_slot + i);
 
-		copy_extent_buffer(dst_path->nodes[0], src, dst_offset,
-				   src_offset, ins_sizes[i]);
+		if ((i == (nr - 1)))
+			last_key = ins_keys[i];
 
-		if (inode_only == LOG_INODE_EXISTS &&
-		    ins_keys[i].type == BTRFS_INODE_ITEM_KEY) {
+		if (ins_keys[i].type == BTRFS_INODE_ITEM_KEY) {
 			inode_item = btrfs_item_ptr(dst_path->nodes[0],
 						    dst_path->slots[0],
 						    struct btrfs_inode_item);
-			btrfs_set_inode_size(dst_path->nodes[0], inode_item, 0);
+			fill_inode_item(trans, dst_path->nodes[0], inode_item,
+					inode, inode_only == LOG_INODE_EXISTS);
+		} else {
+			copy_extent_buffer(dst_path->nodes[0], src, dst_offset,
+					   src_offset, ins_sizes[i]);
+		}
 
-			/* set the generation to zero so the recover code
-			 * can tell the difference between an logging
-			 * just to say 'this inode exists' and a logging
-			 * to say 'update this inode with these values'
-			 */
-			btrfs_set_inode_generation(dst_path->nodes[0],
-						   inode_item, 0);
+		/*
+		 * We set need_find_last_extent here in case we know we were
+		 * processing other items and then walk into the first extent in
+		 * the inode.  If we don't hit an extent then nothing changes,
+		 * we'll do the last search the next time around.
+		 */
+		if (ins_keys[i].type == BTRFS_EXTENT_DATA_KEY) {
+			has_extents = true;
+			if (need_find_last_extent &&
+			    first_key.objectid == (u64)-1)
+				first_key = ins_keys[i];
+		} else {
+			need_find_last_extent = false;
 		}
+
 		/* take a reference on file data extents so that truncates
 		 * or deletes of this inode don't have to relog the inode
 		 * again
 		 */
-		if (btrfs_key_type(ins_keys + i) == BTRFS_EXTENT_DATA_KEY) {
+		if (btrfs_key_type(ins_keys + i) == BTRFS_EXTENT_DATA_KEY &&
+		    !skip_csum) {
 			int found_type;
 			extent = btrfs_item_ptr(src, start_slot + i,
 						struct btrfs_file_extent_item);
 
+			if (btrfs_file_extent_generation(src, extent) < trans->transid)
+				continue;
+
 			found_type = btrfs_file_extent_type(src, extent);
-			if (found_type == BTRFS_FILE_EXTENT_REG ||
-			    found_type == BTRFS_FILE_EXTENT_PREALLOC) {
+			if (found_type == BTRFS_FILE_EXTENT_REG) {
 				u64 ds, dl, cs, cl;
 				ds = btrfs_file_extent_disk_bytenr(src,
 								extent);
@@ -2646,32 +3395,445 @@ static noinline int copy_items(struct btrfs_trans_handle *trans,
 				ret = btrfs_lookup_csums_range(
 						log->fs_info->csum_root,
 						ds + cs, ds + cs + cl - 1,
-						&ordered_sums);
-				BUG_ON(ret);
+						&ordered_sums, 0);
+				if (ret) {
+					btrfs_release_path(dst_path);
+					kfree(ins_data);
+					return ret;
+				}
 			}
 		}
 	}
 
 	btrfs_mark_buffer_dirty(dst_path->nodes[0]);
-	btrfs_release_path(log, dst_path);
+	btrfs_release_path(dst_path);
 	kfree(ins_data);
 
 	/*
 	 * we have to do this after the loop above to avoid changing the
 	 * log tree while trying to change the log tree.
 	 */
+	ret = 0;
 	while (!list_empty(&ordered_sums)) {
 		struct btrfs_ordered_sum *sums = list_entry(ordered_sums.next,
 						   struct btrfs_ordered_sum,
 						   list);
-		ret = btrfs_csum_file_blocks(trans, log, sums);
-		BUG_ON(ret);
+		if (!ret)
+			ret = btrfs_csum_file_blocks(trans, log, sums);
 		list_del(&sums->list);
 		kfree(sums);
 	}
+
+	if (!has_extents)
+		return ret;
+
+	/*
+	 * Because we use btrfs_search_forward we could skip leaves that were
+	 * not modified and then assume *last_extent is valid when it really
+	 * isn't.  So back up to the previous leaf and read the end of the last
+	 * extent before we go and fill in holes.
+	 */
+	if (need_find_last_extent) {
+		u64 len;
+
+		ret = btrfs_prev_leaf(BTRFS_I(inode)->root, src_path);
+		if (ret < 0)
+			return ret;
+		if (ret)
+			goto fill_holes;
+		if (src_path->slots[0])
+			src_path->slots[0]--;
+		src = src_path->nodes[0];
+		btrfs_item_key_to_cpu(src, &key, src_path->slots[0]);
+		if (key.objectid != btrfs_ino(inode) ||
+		    key.type != BTRFS_EXTENT_DATA_KEY)
+			goto fill_holes;
+		extent = btrfs_item_ptr(src, src_path->slots[0],
+					struct btrfs_file_extent_item);
+		if (btrfs_file_extent_type(src, extent) ==
+		    BTRFS_FILE_EXTENT_INLINE) {
+			len = btrfs_file_extent_inline_len(src,
+							   src_path->slots[0],
+							   extent);
+			*last_extent = ALIGN(key.offset + len,
+					     log->sectorsize);
+		} else {
+			len = btrfs_file_extent_num_bytes(src, extent);
+			*last_extent = key.offset + len;
+		}
+	}
+fill_holes:
+	/* So we did prev_leaf, now we need to move to the next leaf, but a few
+	 * things could have happened
+	 *
+	 * 1) A merge could have happened, so we could currently be on a leaf
+	 * that holds what we were copying in the first place.
+	 * 2) A split could have happened, and now not all of the items we want
+	 * are on the same leaf.
+	 *
+	 * So we need to adjust how we search for holes, we need to drop the
+	 * path and re-search for the first extent key we found, and then walk
+	 * forward until we hit the last one we copied.
+	 */
+	if (need_find_last_extent) {
+		/* btrfs_prev_leaf could return 1 without releasing the path */
+		btrfs_release_path(src_path);
+		ret = btrfs_search_slot(NULL, BTRFS_I(inode)->root, &first_key,
+					src_path, 0, 0);
+		if (ret < 0)
+			return ret;
+		ASSERT(ret == 0);
+		src = src_path->nodes[0];
+		i = src_path->slots[0];
+	} else {
+		i = start_slot;
+	}
+
+	/*
+	 * Ok so here we need to go through and fill in any holes we may have
+	 * to make sure that holes are punched for those areas in case they had
+	 * extents previously.
+	 */
+	while (!done) {
+		u64 offset, len;
+		u64 extent_end;
+
+		if (i >= btrfs_header_nritems(src_path->nodes[0])) {
+			ret = btrfs_next_leaf(BTRFS_I(inode)->root, src_path);
+			if (ret < 0)
+				return ret;
+			ASSERT(ret == 0);
+			src = src_path->nodes[0];
+			i = 0;
+		}
+
+		btrfs_item_key_to_cpu(src, &key, i);
+		if (!btrfs_comp_cpu_keys(&key, &last_key))
+			done = true;
+		if (key.objectid != btrfs_ino(inode) ||
+		    key.type != BTRFS_EXTENT_DATA_KEY) {
+			i++;
+			continue;
+		}
+		extent = btrfs_item_ptr(src, i, struct btrfs_file_extent_item);
+		if (btrfs_file_extent_type(src, extent) ==
+		    BTRFS_FILE_EXTENT_INLINE) {
+			len = btrfs_file_extent_inline_len(src, i, extent);
+			extent_end = ALIGN(key.offset + len, log->sectorsize);
+		} else {
+			len = btrfs_file_extent_num_bytes(src, extent);
+			extent_end = key.offset + len;
+		}
+		i++;
+
+		if (*last_extent == key.offset) {
+			*last_extent = extent_end;
+			continue;
+		}
+		offset = *last_extent;
+		len = key.offset - *last_extent;
+		ret = btrfs_insert_file_extent(trans, log, btrfs_ino(inode),
+					       offset, 0, 0, len, 0, len, 0,
+					       0, 0);
+		if (ret)
+			break;
+		*last_extent = offset + len;
+	}
+	/*
+	 * Need to let the callers know we dropped the path so they should
+	 * re-search.
+	 */
+	if (!ret && need_find_last_extent)
+		ret = 1;
+	return ret;
+}
+
+static int extent_cmp(void *priv, struct list_head *a, struct list_head *b)
+{
+	struct extent_map *em1, *em2;
+
+	em1 = list_entry(a, struct extent_map, list);
+	em2 = list_entry(b, struct extent_map, list);
+
+	if (em1->start < em2->start)
+		return -1;
+	else if (em1->start > em2->start)
+		return 1;
 	return 0;
 }
 
+static int log_one_extent(struct btrfs_trans_handle *trans,
+			  struct inode *inode, struct btrfs_root *root,
+			  struct extent_map *em, struct btrfs_path *path,
+			  struct list_head *logged_list)
+{
+	struct btrfs_root *log = root->log_root;
+	struct btrfs_file_extent_item *fi;
+	struct extent_buffer *leaf;
+	struct btrfs_ordered_extent *ordered;
+	struct list_head ordered_sums;
+	struct btrfs_map_token token;
+	struct btrfs_key key;
+	u64 mod_start = em->mod_start;
+	u64 mod_len = em->mod_len;
+	u64 csum_offset;
+	u64 csum_len;
+	u64 extent_offset = em->start - em->orig_start;
+	u64 block_len;
+	int ret;
+	bool skip_csum = BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM;
+	int extent_inserted = 0;
+
+	INIT_LIST_HEAD(&ordered_sums);
+	btrfs_init_map_token(&token);
+
+	ret = __btrfs_drop_extents(trans, log, inode, path, em->start,
+				   em->start + em->len, NULL, 0, 1,
+				   sizeof(*fi), &extent_inserted);
+	if (ret)
+		return ret;
+
+	if (!extent_inserted) {
+		key.objectid = btrfs_ino(inode);
+		key.type = BTRFS_EXTENT_DATA_KEY;
+		key.offset = em->start;
+
+		ret = btrfs_insert_empty_item(trans, log, path, &key,
+					      sizeof(*fi));
+		if (ret)
+			return ret;
+	}
+	leaf = path->nodes[0];
+	fi = btrfs_item_ptr(leaf, path->slots[0],
+			    struct btrfs_file_extent_item);
+
+	btrfs_set_token_file_extent_generation(leaf, fi, em->generation,
+					       &token);
+	if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags)) {
+		skip_csum = true;
+		btrfs_set_token_file_extent_type(leaf, fi,
+						 BTRFS_FILE_EXTENT_PREALLOC,
+						 &token);
+	} else {
+		btrfs_set_token_file_extent_type(leaf, fi,
+						 BTRFS_FILE_EXTENT_REG,
+						 &token);
+		if (em->block_start == EXTENT_MAP_HOLE)
+			skip_csum = true;
+	}
+
+	block_len = max(em->block_len, em->orig_block_len);
+	if (em->compress_type != BTRFS_COMPRESS_NONE) {
+		btrfs_set_token_file_extent_disk_bytenr(leaf, fi,
+							em->block_start,
+							&token);
+		btrfs_set_token_file_extent_disk_num_bytes(leaf, fi, block_len,
+							   &token);
+	} else if (em->block_start < EXTENT_MAP_LAST_BYTE) {
+		btrfs_set_token_file_extent_disk_bytenr(leaf, fi,
+							em->block_start -
+							extent_offset, &token);
+		btrfs_set_token_file_extent_disk_num_bytes(leaf, fi, block_len,
+							   &token);
+	} else {
+		btrfs_set_token_file_extent_disk_bytenr(leaf, fi, 0, &token);
+		btrfs_set_token_file_extent_disk_num_bytes(leaf, fi, 0,
+							   &token);
+	}
+
+	btrfs_set_token_file_extent_offset(leaf, fi,
+					   em->start - em->orig_start,
+					   &token);
+	btrfs_set_token_file_extent_num_bytes(leaf, fi, em->len, &token);
+	btrfs_set_token_file_extent_ram_bytes(leaf, fi, em->ram_bytes, &token);
+	btrfs_set_token_file_extent_compression(leaf, fi, em->compress_type,
+						&token);
+	btrfs_set_token_file_extent_encryption(leaf, fi, 0, &token);
+	btrfs_set_token_file_extent_other_encoding(leaf, fi, 0, &token);
+	btrfs_mark_buffer_dirty(leaf);
+
+	btrfs_release_path(path);
+	if (ret) {
+		return ret;
+	}
+
+	if (skip_csum)
+		return 0;
+
+	/*
+	 * First check and see if our csums are on our outstanding ordered
+	 * extents.
+	 */
+	list_for_each_entry(ordered, logged_list, log_list) {
+		struct btrfs_ordered_sum *sum;
+
+		if (!mod_len)
+			break;
+
+		if (ordered->file_offset + ordered->len <= mod_start ||
+		    mod_start + mod_len <= ordered->file_offset)
+			continue;
+
+		/*
+		 * We are going to copy all the csums on this ordered extent, so
+		 * go ahead and adjust mod_start and mod_len in case this
+		 * ordered extent has already been logged.
+		 */
+		if (ordered->file_offset > mod_start) {
+			if (ordered->file_offset + ordered->len >=
+			    mod_start + mod_len)
+				mod_len = ordered->file_offset - mod_start;
+			/*
+			 * If we have this case
+			 *
+			 * |--------- logged extent ---------|
+			 *       |----- ordered extent ----|
+			 *
+			 * Just don't mess with mod_start and mod_len, we'll
+			 * just end up logging more csums than we need and it
+			 * will be ok.
+			 */
+		} else {
+			if (ordered->file_offset + ordered->len <
+			    mod_start + mod_len) {
+				mod_len = (mod_start + mod_len) -
+					(ordered->file_offset + ordered->len);
+				mod_start = ordered->file_offset +
+					ordered->len;
+			} else {
+				mod_len = 0;
+			}
+		}
+
+		/*
+		 * To keep us from looping for the above case of an ordered
+		 * extent that falls inside of the logged extent.
+		 */
+		if (test_and_set_bit(BTRFS_ORDERED_LOGGED_CSUM,
+				     &ordered->flags))
+			continue;
+
+		if (ordered->csum_bytes_left) {
+			btrfs_start_ordered_extent(inode, ordered, 0);
+			wait_event(ordered->wait,
+				   ordered->csum_bytes_left == 0);
+		}
+
+		list_for_each_entry(sum, &ordered->list, list) {
+			ret = btrfs_csum_file_blocks(trans, log, sum);
+			if (ret)
+				goto unlocked;
+		}
+
+	}
+unlocked:
+
+	if (!mod_len || ret)
+		return ret;
+
+	if (em->compress_type) {
+		csum_offset = 0;
+		csum_len = block_len;
+	} else {
+		csum_offset = mod_start - em->start;
+		csum_len = mod_len;
+	}
+
+	/* block start is already adjusted for the file extent offset. */
+	ret = btrfs_lookup_csums_range(log->fs_info->csum_root,
+				       em->block_start + csum_offset,
+				       em->block_start + csum_offset +
+				       csum_len - 1, &ordered_sums, 0);
+	if (ret)
+		return ret;
+
+	while (!list_empty(&ordered_sums)) {
+		struct btrfs_ordered_sum *sums = list_entry(ordered_sums.next,
+						   struct btrfs_ordered_sum,
+						   list);
+		if (!ret)
+			ret = btrfs_csum_file_blocks(trans, log, sums);
+		list_del(&sums->list);
+		kfree(sums);
+	}
+
+	return ret;
+}
+
+static int btrfs_log_changed_extents(struct btrfs_trans_handle *trans,
+				     struct btrfs_root *root,
+				     struct inode *inode,
+				     struct btrfs_path *path,
+				     struct list_head *logged_list)
+{
+	struct extent_map *em, *n;
+	struct list_head extents;
+	struct extent_map_tree *tree = &BTRFS_I(inode)->extent_tree;
+	u64 test_gen;
+	int ret = 0;
+	int num = 0;
+
+	INIT_LIST_HEAD(&extents);
+
+	write_lock(&tree->lock);
+	test_gen = root->fs_info->last_trans_committed;
+
+	list_for_each_entry_safe(em, n, &tree->modified_extents, list) {
+		list_del_init(&em->list);
+
+		/*
+		 * Just an arbitrary number, this can be really CPU intensive
+		 * once we start getting a lot of extents, and really once we
+		 * have a bunch of extents we just want to commit since it will
+		 * be faster.
+		 */
+		if (++num > 32768) {
+			list_del_init(&tree->modified_extents);
+			ret = -EFBIG;
+			goto process;
+		}
+
+		if (em->generation <= test_gen)
+			continue;
+		/* Need a ref to keep it from getting evicted from cache */
+		atomic_inc(&em->refs);
+		set_bit(EXTENT_FLAG_LOGGING, &em->flags);
+		list_add_tail(&em->list, &extents);
+		num++;
+	}
+
+	list_sort(NULL, &extents, extent_cmp);
+
+process:
+	while (!list_empty(&extents)) {
+		em = list_entry(extents.next, struct extent_map, list);
+
+		list_del_init(&em->list);
+
+		/*
+		 * If we had an error we just need to delete everybody from our
+		 * private list.
+		 */
+		if (ret) {
+			clear_em_logging(tree, em);
+			free_extent_map(em);
+			continue;
+		}
+
+		write_unlock(&tree->lock);
+
+		ret = log_one_extent(trans, inode, root, em, path, logged_list);
+		write_lock(&tree->lock);
+		clear_em_logging(tree, em);
+		free_extent_map(em);
+	}
+	WARN_ON(!list_empty(&extents));
+	write_unlock(&tree->lock);
+
+	btrfs_release_path(path);
+	return ret;
+}
+
 /* log a single inode in the tree log.
  * At least one parent directory for this inode must exist in the tree
  * or be logged already.
@@ -2696,35 +3858,57 @@ static int btrfs_log_inode(struct btrfs_trans_handle *trans,
 	struct btrfs_key max_key;
 	struct btrfs_root *log = root->log_root;
 	struct extent_buffer *src = NULL;
-	u32 size;
+	LIST_HEAD(logged_list);
+	u64 last_extent = 0;
+	int err = 0;
 	int ret;
 	int nritems;
 	int ins_start_slot = 0;
 	int ins_nr;
-
-	log = root->log_root;
+	bool fast_search = false;
+	u64 ino = btrfs_ino(inode);
 
 	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
 	dst_path = btrfs_alloc_path();
+	if (!dst_path) {
+		btrfs_free_path(path);
+		return -ENOMEM;
+	}
 
-	min_key.objectid = inode->i_ino;
+	min_key.objectid = ino;
 	min_key.type = BTRFS_INODE_ITEM_KEY;
 	min_key.offset = 0;
 
-	max_key.objectid = inode->i_ino;
+	max_key.objectid = ino;
 
-	/* today the code can only do partial logging of directories */
-	if (!S_ISDIR(inode->i_mode))
-	    inode_only = LOG_INODE_ALL;
 
-	if (inode_only == LOG_INODE_EXISTS || S_ISDIR(inode->i_mode))
+	/* today the code can only do partial logging of directories */
+	if (S_ISDIR(inode->i_mode) ||
+	    (!test_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
+		       &BTRFS_I(inode)->runtime_flags) &&
+	     inode_only == LOG_INODE_EXISTS))
 		max_key.type = BTRFS_XATTR_ITEM_KEY;
 	else
 		max_key.type = (u8)-1;
 	max_key.offset = (u64)-1;
 
+	/* Only run delayed items if we are a dir or a new file */
+	if (S_ISDIR(inode->i_mode) ||
+	    BTRFS_I(inode)->generation > root->fs_info->last_trans_committed) {
+		ret = btrfs_commit_inode_delayed_items(trans, inode);
+		if (ret) {
+			btrfs_free_path(path);
+			btrfs_free_path(dst_path);
+			return ret;
+		}
+	}
+
 	mutex_lock(&BTRFS_I(inode)->log_mutex);
 
+	btrfs_get_logged_extents(inode, &logged_list);
+
 	/*
 	 * a brute force approach to making sure we get the most uptodate
 	 * copies of everything.
@@ -2734,29 +3918,54 @@ static int btrfs_log_inode(struct btrfs_trans_handle *trans,
 
 		if (inode_only == LOG_INODE_EXISTS)
 			max_key_type = BTRFS_XATTR_ITEM_KEY;
-		ret = drop_objectid_items(trans, log, path,
-					  inode->i_ino, max_key_type);
+		ret = drop_objectid_items(trans, log, path, ino, max_key_type);
 	} else {
-		ret = btrfs_truncate_inode_items(trans, log, inode, 0, 0);
+		if (test_and_clear_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
+				       &BTRFS_I(inode)->runtime_flags)) {
+			clear_bit(BTRFS_INODE_COPY_EVERYTHING,
+				  &BTRFS_I(inode)->runtime_flags);
+			ret = btrfs_truncate_inode_items(trans, log,
+							 inode, 0, 0);
+		} else if (test_and_clear_bit(BTRFS_INODE_COPY_EVERYTHING,
+					      &BTRFS_I(inode)->runtime_flags) ||
+			   inode_only == LOG_INODE_EXISTS) {
+			if (inode_only == LOG_INODE_ALL)
+				fast_search = true;
+			max_key.type = BTRFS_XATTR_ITEM_KEY;
+			ret = drop_objectid_items(trans, log, path, ino,
+						  max_key.type);
+		} else {
+			if (inode_only == LOG_INODE_ALL)
+				fast_search = true;
+			ret = log_inode_item(trans, log, dst_path, inode);
+			if (ret) {
+				err = ret;
+				goto out_unlock;
+			}
+			goto log_extents;
+		}
+
+	}
+	if (ret) {
+		err = ret;
+		goto out_unlock;
 	}
-	BUG_ON(ret);
 	path->keep_locks = 1;
 
 	while (1) {
 		ins_nr = 0;
-		ret = btrfs_search_forward(root, &min_key, &max_key,
-					   path, 0, trans->transid);
+		ret = btrfs_search_forward(root, &min_key,
+					   path, trans->transid);
 		if (ret != 0)
 			break;
 again:
 		/* note, ins_nr might be > 0 here, cleanup outside the loop */
-		if (min_key.objectid != inode->i_ino)
+		if (min_key.objectid != ino)
 			break;
 		if (min_key.type > max_key.type)
 			break;
 
 		src = path->nodes[0];
-		size = btrfs_item_size_nr(src, path->slots[0]);
 		if (ins_nr && ins_start_slot + ins_nr == path->slots[0]) {
 			ins_nr++;
 			goto next_slot;
@@ -2766,9 +3975,16 @@ again:
 			goto next_slot;
 		}
 
-		ret = copy_items(trans, log, dst_path, src, ins_start_slot,
-				 ins_nr, inode_only);
-		BUG_ON(ret);
+		ret = copy_items(trans, inode, dst_path, path, &last_extent,
+				 ins_start_slot, ins_nr, inode_only);
+		if (ret < 0) {
+			err = ret;
+			goto out_unlock;
+		} if (ret) {
+			ins_nr = 0;
+			btrfs_release_path(path);
+			continue;
+		}
 		ins_nr = 1;
 		ins_start_slot = path->slots[0];
 next_slot:
@@ -2781,43 +3997,77 @@ next_slot:
 			goto again;
 		}
 		if (ins_nr) {
-			ret = copy_items(trans, log, dst_path, src,
-					 ins_start_slot,
+			ret = copy_items(trans, inode, dst_path, path,
+					 &last_extent, ins_start_slot,
 					 ins_nr, inode_only);
-			BUG_ON(ret);
+			if (ret < 0) {
+				err = ret;
+				goto out_unlock;
+			}
+			ret = 0;
 			ins_nr = 0;
 		}
-		btrfs_release_path(root, path);
+		btrfs_release_path(path);
 
-		if (min_key.offset < (u64)-1)
+		if (min_key.offset < (u64)-1) {
 			min_key.offset++;
-		else if (min_key.type < (u8)-1)
+		} else if (min_key.type < max_key.type) {
 			min_key.type++;
-		else if (min_key.objectid < (u64)-1)
-			min_key.objectid++;
-		else
+			min_key.offset = 0;
+		} else {
 			break;
+		}
 	}
 	if (ins_nr) {
-		ret = copy_items(trans, log, dst_path, src,
-				 ins_start_slot,
-				 ins_nr, inode_only);
-		BUG_ON(ret);
+		ret = copy_items(trans, inode, dst_path, path, &last_extent,
+				 ins_start_slot, ins_nr, inode_only);
+		if (ret < 0) {
+			err = ret;
+			goto out_unlock;
+		}
+		ret = 0;
 		ins_nr = 0;
 	}
-	WARN_ON(ins_nr);
+
+log_extents:
+	btrfs_release_path(path);
+	btrfs_release_path(dst_path);
+	if (fast_search) {
+		ret = btrfs_log_changed_extents(trans, root, inode, dst_path,
+						&logged_list);
+		if (ret) {
+			err = ret;
+			goto out_unlock;
+		}
+	} else if (inode_only == LOG_INODE_ALL) {
+		struct extent_map_tree *tree = &BTRFS_I(inode)->extent_tree;
+		struct extent_map *em, *n;
+
+		write_lock(&tree->lock);
+		list_for_each_entry_safe(em, n, &tree->modified_extents, list)
+			list_del_init(&em->list);
+		write_unlock(&tree->lock);
+	}
+
 	if (inode_only == LOG_INODE_ALL && S_ISDIR(inode->i_mode)) {
-		btrfs_release_path(root, path);
-		btrfs_release_path(log, dst_path);
 		ret = log_directory_changes(trans, root, inode, path, dst_path);
-		BUG_ON(ret);
+		if (ret) {
+			err = ret;
+			goto out_unlock;
+		}
 	}
 	BTRFS_I(inode)->logged_trans = trans->transid;
+	BTRFS_I(inode)->last_log_commit = BTRFS_I(inode)->last_sub_trans;
+out_unlock:
+	if (unlikely(err))
+		btrfs_put_logged_extents(&logged_list);
+	else
+		btrfs_submit_logged_extents(&logged_list, log);
 	mutex_unlock(&BTRFS_I(inode)->log_mutex);
 
 	btrfs_free_path(path);
 	btrfs_free_path(dst_path);
-	return 0;
+	return err;
 }
 
 /*
@@ -2834,6 +4084,8 @@ static noinline int check_parent_dirs_for_sync(struct btrfs_trans_handle *trans,
 {
 	int ret = 0;
 	struct btrfs_root *root;
+	struct dentry *old_parent = NULL;
+	struct inode *orig_inode = inode;
 
 	/*
 	 * for regular files, if its inode is already on disk, we don't
@@ -2853,7 +4105,14 @@ static noinline int check_parent_dirs_for_sync(struct btrfs_trans_handle *trans,
 	}
 
 	while (1) {
-		BTRFS_I(inode)->logged_trans = trans->transid;
+		/*
+		 * If we are logging a directory then we start with our inode,
+		 * not our parents inode, so we need to skipp setting the
+		 * logged_trans so that further down in the log code we don't
+		 * think this inode has already been logged.
+		 */
+		if (inode != orig_inode)
+			BTRFS_I(inode)->logged_trans = trans->transid;
 		smp_mb();
 
 		if (BTRFS_I(inode)->last_unlink_trans > last_committed) {
@@ -2863,8 +4122,7 @@ static noinline int check_parent_dirs_for_sync(struct btrfs_trans_handle *trans,
 			 * make sure any commits to the log are forced
 			 * to be full commits
 			 */
-			root->fs_info->last_trans_log_full_commit =
-				trans->transid;
+			btrfs_set_log_full_commit(root->fs_info, trans);
 			ret = 1;
 			break;
 		}
@@ -2875,41 +4133,31 @@ static noinline int check_parent_dirs_for_sync(struct btrfs_trans_handle *trans,
 		if (IS_ROOT(parent))
 			break;
 
-		parent = parent->d_parent;
+		parent = dget_parent(parent);
+		dput(old_parent);
+		old_parent = parent;
 		inode = parent->d_inode;
 
 	}
+	dput(old_parent);
 out:
 	return ret;
 }
 
-static int inode_in_log(struct btrfs_trans_handle *trans,
-		 struct inode *inode)
-{
-	struct btrfs_root *root = BTRFS_I(inode)->root;
-	int ret = 0;
-
-	mutex_lock(&root->log_mutex);
-	if (BTRFS_I(inode)->logged_trans == trans->transid &&
-	    BTRFS_I(inode)->last_sub_trans <= root->last_log_commit)
-		ret = 1;
-	mutex_unlock(&root->log_mutex);
-	return ret;
-}
-
-
 /*
  * helper function around btrfs_log_inode to make sure newly created
  * parent directories also end up in the log.  A minimal inode and backref
  * only logging is done of any parent directories that are older than
  * the last committed transaction
  */
-int btrfs_log_inode_parent(struct btrfs_trans_handle *trans,
-		    struct btrfs_root *root, struct inode *inode,
-		    struct dentry *parent, int exists_only)
+static int btrfs_log_inode_parent(struct btrfs_trans_handle *trans,
+			    	  struct btrfs_root *root, struct inode *inode,
+			    	  struct dentry *parent, int exists_only,
+				  struct btrfs_log_ctx *ctx)
 {
 	int inode_only = exists_only ? LOG_INODE_EXISTS : LOG_INODE_ALL;
 	struct super_block *sb;
+	struct dentry *old_parent = NULL;
 	int ret = 0;
 	u64 last_committed = root->fs_info->last_trans_committed;
 
@@ -2920,6 +4168,10 @@ int btrfs_log_inode_parent(struct btrfs_trans_handle *trans,
 		goto end_no_trans;
 	}
 
+	/*
+	 * The prev transaction commit doesn't complete, we need do
+	 * full commit by ourselves.
+	 */
 	if (root->fs_info->last_trans_log_full_commit >
 	    root->fs_info->last_trans_committed) {
 		ret = 1;
@@ -2937,15 +4189,18 @@ int btrfs_log_inode_parent(struct btrfs_trans_handle *trans,
 	if (ret)
 		goto end_no_trans;
 
-	if (inode_in_log(trans, inode)) {
+	if (btrfs_inode_in_log(inode, trans->transid)) {
 		ret = BTRFS_NO_LOG_SYNC;
 		goto end_no_trans;
 	}
 
-	start_log_trans(trans, root);
+	ret = start_log_trans(trans, root, ctx);
+	if (ret)
+		goto end_no_trans;
 
 	ret = btrfs_log_inode(trans, root, inode, inode_only);
-	BUG_ON(ret);
+	if (ret)
+		goto end_trans;
 
 	/*
 	 * for regular files, if its inode is already on disk, we don't
@@ -2955,8 +4210,10 @@ int btrfs_log_inode_parent(struct btrfs_trans_handle *trans,
 	 */
 	if (S_ISREG(inode->i_mode) &&
 	    BTRFS_I(inode)->generation <= last_committed &&
-	    BTRFS_I(inode)->last_unlink_trans <= last_committed)
-			goto no_parent;
+	    BTRFS_I(inode)->last_unlink_trans <= last_committed) {
+		ret = 0;
+		goto end_trans;
+	}
 
 	inode_only = LOG_INODE_EXISTS;
 	while (1) {
@@ -2970,15 +4227,26 @@ int btrfs_log_inode_parent(struct btrfs_trans_handle *trans,
 		if (BTRFS_I(inode)->generation >
 		    root->fs_info->last_trans_committed) {
 			ret = btrfs_log_inode(trans, root, inode, inode_only);
-			BUG_ON(ret);
+			if (ret)
+				goto end_trans;
 		}
 		if (IS_ROOT(parent))
 			break;
 
-		parent = parent->d_parent;
+		parent = dget_parent(parent);
+		dput(old_parent);
+		old_parent = parent;
 	}
-no_parent:
 	ret = 0;
+end_trans:
+	dput(old_parent);
+	if (ret < 0) {
+		btrfs_set_log_full_commit(root->fs_info, trans);
+		ret = 1;
+	}
+
+	if (ret)
+		btrfs_remove_log_ctx(root, ctx);
 	btrfs_end_log_trans(root);
 end_no_trans:
 	return ret;
@@ -2991,10 +4259,17 @@ end_no_trans:
  * data on disk.
  */
 int btrfs_log_dentry_safe(struct btrfs_trans_handle *trans,
-			  struct btrfs_root *root, struct dentry *dentry)
+			  struct btrfs_root *root, struct dentry *dentry,
+			  struct btrfs_log_ctx *ctx)
 {
-	return btrfs_log_inode_parent(trans, root, dentry->d_inode,
-				      dentry->d_parent, 0);
+	struct dentry *parent = dget_parent(dentry);
+	int ret;
+
+	ret = btrfs_log_inode_parent(trans, root, dentry->d_inode, parent,
+				     0, ctx);
+	dput(parent);
+
+	return ret;
 }
 
 /*
@@ -3016,16 +4291,27 @@ int btrfs_recover_log_trees(struct btrfs_root *log_root_tree)
 		.stage = 0,
 	};
 
-	fs_info->log_root_recovering = 1;
 	path = btrfs_alloc_path();
-	BUG_ON(!path);
+	if (!path)
+		return -ENOMEM;
 
-	trans = btrfs_start_transaction(fs_info->tree_root, 1);
+	fs_info->log_root_recovering = 1;
+
+	trans = btrfs_start_transaction(fs_info->tree_root, 0);
+	if (IS_ERR(trans)) {
+		ret = PTR_ERR(trans);
+		goto error;
+	}
 
 	wc.trans = trans;
 	wc.pin = 1;
 
-	walk_log_tree(trans, log_root_tree, &wc);
+	ret = walk_log_tree(trans, log_root_tree, &wc);
+	if (ret) {
+		btrfs_error(fs_info, ret, "Failed to pin buffers while "
+			    "recovering log root tree.");
+		goto error;
+	}
 
 again:
 	key.objectid = BTRFS_TREE_LOG_OBJECTID;
@@ -3034,8 +4320,12 @@ again:
 
 	while (1) {
 		ret = btrfs_search_slot(NULL, log_root_tree, &key, path, 0, 0);
-		if (ret < 0)
-			break;
+
+		if (ret < 0) {
+			btrfs_error(fs_info, ret,
+				    "Couldn't find tree log root.");
+			goto error;
+		}
 		if (ret > 0) {
 			if (path->slots[0] == 0)
 				break;
@@ -3043,31 +4333,40 @@ again:
 		}
 		btrfs_item_key_to_cpu(path->nodes[0], &found_key,
 				      path->slots[0]);
-		btrfs_release_path(log_root_tree, path);
+		btrfs_release_path(path);
 		if (found_key.objectid != BTRFS_TREE_LOG_OBJECTID)
 			break;
 
-		log = btrfs_read_fs_root_no_radix(log_root_tree,
-						  &found_key);
-		BUG_ON(!log);
-
+		log = btrfs_read_fs_root(log_root_tree, &found_key);
+		if (IS_ERR(log)) {
+			ret = PTR_ERR(log);
+			btrfs_error(fs_info, ret,
+				    "Couldn't read tree log root.");
+			goto error;
+		}
 
 		tmp_key.objectid = found_key.offset;
 		tmp_key.type = BTRFS_ROOT_ITEM_KEY;
 		tmp_key.offset = (u64)-1;
 
 		wc.replay_dest = btrfs_read_fs_root_no_name(fs_info, &tmp_key);
-		BUG_ON(!wc.replay_dest);
+		if (IS_ERR(wc.replay_dest)) {
+			ret = PTR_ERR(wc.replay_dest);
+			free_extent_buffer(log->node);
+			free_extent_buffer(log->commit_root);
+			kfree(log);
+			btrfs_error(fs_info, ret, "Couldn't read target root "
+				    "for tree log recovery.");
+			goto error;
+		}
 
 		wc.replay_dest->log_root = log;
 		btrfs_record_root_in_trans(trans, wc.replay_dest);
 		ret = walk_log_tree(trans, log, &wc);
-		BUG_ON(ret);
 
-		if (wc.stage == LOG_WALK_REPLAY_ALL) {
+		if (!ret && wc.stage == LOG_WALK_REPLAY_ALL) {
 			ret = fixup_inode_link_counts(trans, wc.replay_dest,
 						      path);
-			BUG_ON(ret);
 		}
 
 		key.offset = found_key.offset - 1;
@@ -3076,10 +4375,13 @@ again:
 		free_extent_buffer(log->commit_root);
 		kfree(log);
 
+		if (ret)
+			goto error;
+
 		if (found_key.offset == 0)
 			break;
 	}
-	btrfs_release_path(log_root_tree, path);
+	btrfs_release_path(path);
 
 	/* step one is to pin it all, step two is to replay just inodes */
 	if (wc.pin) {
@@ -3096,15 +4398,22 @@ again:
 
 	btrfs_free_path(path);
 
+	/* step 4: commit the transaction, which also unpins the blocks */
+	ret = btrfs_commit_transaction(trans, fs_info->tree_root);
+	if (ret)
+		return ret;
+
 	free_extent_buffer(log_root_tree->node);
 	log_root_tree->log_root = NULL;
 	fs_info->log_root_recovering = 0;
-
-	/* step 4: commit the transaction, which also unpins the blocks */
-	btrfs_commit_transaction(trans, fs_info->tree_root);
-
 	kfree(log_root_tree);
+
 	return 0;
+error:
+	if (wc.trans)
+		btrfs_end_transaction(wc.trans, fs_info->tree_root);
+	btrfs_free_path(path);
+	return ret;
 }
 
 /*
@@ -3194,6 +4503,6 @@ int btrfs_log_new_name(struct btrfs_trans_handle *trans,
 		    root->fs_info->last_trans_committed))
 		return 0;
 
-	return btrfs_log_inode_parent(trans, root, inode, parent, 1);
+	return btrfs_log_inode_parent(trans, root, inode, parent, 1, NULL);
 }
 
diff --git a/fs/btrfs/tree-log.h b/fs/btrfs/tree-log.h
index 0776eacb508..7f5b41bd537 100644
--- a/fs/btrfs/tree-log.h
+++ b/fs/btrfs/tree-log.h
@@ -19,15 +19,47 @@
 #ifndef __TREE_LOG_
 #define __TREE_LOG_
 
+#include "ctree.h"
+#include "transaction.h"
+
 /* return value for btrfs_log_dentry_safe that means we don't need to log it at all */
 #define BTRFS_NO_LOG_SYNC 256
 
+struct btrfs_log_ctx {
+	int log_ret;
+	int log_transid;
+	struct list_head list;
+};
+
+static inline void btrfs_init_log_ctx(struct btrfs_log_ctx *ctx)
+{
+	ctx->log_ret = 0;
+	ctx->log_transid = 0;
+	INIT_LIST_HEAD(&ctx->list);
+}
+
+static inline void btrfs_set_log_full_commit(struct btrfs_fs_info *fs_info,
+					     struct btrfs_trans_handle *trans)
+{
+	ACCESS_ONCE(fs_info->last_trans_log_full_commit) = trans->transid;
+}
+
+static inline int btrfs_need_log_full_commit(struct btrfs_fs_info *fs_info,
+					     struct btrfs_trans_handle *trans)
+{
+	return ACCESS_ONCE(fs_info->last_trans_log_full_commit) ==
+		trans->transid;
+}
+
 int btrfs_sync_log(struct btrfs_trans_handle *trans,
-		   struct btrfs_root *root);
+		   struct btrfs_root *root, struct btrfs_log_ctx *ctx);
 int btrfs_free_log(struct btrfs_trans_handle *trans, struct btrfs_root *root);
+int btrfs_free_log_root_tree(struct btrfs_trans_handle *trans,
+			     struct btrfs_fs_info *fs_info);
 int btrfs_recover_log_trees(struct btrfs_root *tree_root);
 int btrfs_log_dentry_safe(struct btrfs_trans_handle *trans,
-			  struct btrfs_root *root, struct dentry *dentry);
+			  struct btrfs_root *root, struct dentry *dentry,
+			  struct btrfs_log_ctx *ctx);
 int btrfs_del_dir_entries_in_log(struct btrfs_trans_handle *trans,
 				 struct btrfs_root *root,
 				 const char *name, int name_len,
@@ -36,12 +68,8 @@ int btrfs_del_inode_ref_in_log(struct btrfs_trans_handle *trans,
 			       struct btrfs_root *root,
 			       const char *name, int name_len,
 			       struct inode *inode, u64 dirid);
-int btrfs_join_running_log_trans(struct btrfs_root *root);
-int btrfs_end_log_trans(struct btrfs_root *root);
+void btrfs_end_log_trans(struct btrfs_root *root);
 int btrfs_pin_log_trans(struct btrfs_root *root);
-int btrfs_log_inode_parent(struct btrfs_trans_handle *trans,
-		    struct btrfs_root *root, struct inode *inode,
-		    struct dentry *parent, int exists_only);
 void btrfs_record_unlink_dir(struct btrfs_trans_handle *trans,
 			     struct inode *dir, struct inode *inode,
 			     int for_rename);
diff --git a/fs/btrfs/ulist.c b/fs/btrfs/ulist.c
new file mode 100644
index 00000000000..840a38b2778
--- /dev/null
+++ b/fs/btrfs/ulist.c
@@ -0,0 +1,251 @@
+/*
+ * Copyright (C) 2011 STRATO AG
+ * written by Arne Jansen <sensille@gmx.net>
+ * Distributed under the GNU GPL license version 2.
+ */
+
+#include <linux/slab.h>
+#include "ulist.h"
+#include "ctree.h"
+
+/*
+ * ulist is a generic data structure to hold a collection of unique u64
+ * values. The only operations it supports is adding to the list and
+ * enumerating it.
+ * It is possible to store an auxiliary value along with the key.
+ *
+ * A sample usage for ulists is the enumeration of directed graphs without
+ * visiting a node twice. The pseudo-code could look like this:
+ *
+ * ulist = ulist_alloc();
+ * ulist_add(ulist, root);
+ * ULIST_ITER_INIT(&uiter);
+ *
+ * while ((elem = ulist_next(ulist, &uiter)) {
+ * 	for (all child nodes n in elem)
+ *		ulist_add(ulist, n);
+ *	do something useful with the node;
+ * }
+ * ulist_free(ulist);
+ *
+ * This assumes the graph nodes are adressable by u64. This stems from the
+ * usage for tree enumeration in btrfs, where the logical addresses are
+ * 64 bit.
+ *
+ * It is also useful for tree enumeration which could be done elegantly
+ * recursively, but is not possible due to kernel stack limitations. The
+ * loop would be similar to the above.
+ */
+
+/**
+ * ulist_init - freshly initialize a ulist
+ * @ulist:	the ulist to initialize
+ *
+ * Note: don't use this function to init an already used ulist, use
+ * ulist_reinit instead.
+ */
+void ulist_init(struct ulist *ulist)
+{
+	INIT_LIST_HEAD(&ulist->nodes);
+	ulist->root = RB_ROOT;
+	ulist->nnodes = 0;
+}
+
+/**
+ * ulist_fini - free up additionally allocated memory for the ulist
+ * @ulist:	the ulist from which to free the additional memory
+ *
+ * This is useful in cases where the base 'struct ulist' has been statically
+ * allocated.
+ */
+static void ulist_fini(struct ulist *ulist)
+{
+	struct ulist_node *node;
+	struct ulist_node *next;
+
+	list_for_each_entry_safe(node, next, &ulist->nodes, list) {
+		kfree(node);
+	}
+	ulist->root = RB_ROOT;
+	INIT_LIST_HEAD(&ulist->nodes);
+}
+
+/**
+ * ulist_reinit - prepare a ulist for reuse
+ * @ulist:	ulist to be reused
+ *
+ * Free up all additional memory allocated for the list elements and reinit
+ * the ulist.
+ */
+void ulist_reinit(struct ulist *ulist)
+{
+	ulist_fini(ulist);
+	ulist_init(ulist);
+}
+
+/**
+ * ulist_alloc - dynamically allocate a ulist
+ * @gfp_mask:	allocation flags to for base allocation
+ *
+ * The allocated ulist will be returned in an initialized state.
+ */
+struct ulist *ulist_alloc(gfp_t gfp_mask)
+{
+	struct ulist *ulist = kmalloc(sizeof(*ulist), gfp_mask);
+
+	if (!ulist)
+		return NULL;
+
+	ulist_init(ulist);
+
+	return ulist;
+}
+
+/**
+ * ulist_free - free dynamically allocated ulist
+ * @ulist:	ulist to free
+ *
+ * It is not necessary to call ulist_fini before.
+ */
+void ulist_free(struct ulist *ulist)
+{
+	if (!ulist)
+		return;
+	ulist_fini(ulist);
+	kfree(ulist);
+}
+
+static struct ulist_node *ulist_rbtree_search(struct ulist *ulist, u64 val)
+{
+	struct rb_node *n = ulist->root.rb_node;
+	struct ulist_node *u = NULL;
+
+	while (n) {
+		u = rb_entry(n, struct ulist_node, rb_node);
+		if (u->val < val)
+			n = n->rb_right;
+		else if (u->val > val)
+			n = n->rb_left;
+		else
+			return u;
+	}
+	return NULL;
+}
+
+static int ulist_rbtree_insert(struct ulist *ulist, struct ulist_node *ins)
+{
+	struct rb_node **p = &ulist->root.rb_node;
+	struct rb_node *parent = NULL;
+	struct ulist_node *cur = NULL;
+
+	while (*p) {
+		parent = *p;
+		cur = rb_entry(parent, struct ulist_node, rb_node);
+
+		if (cur->val < ins->val)
+			p = &(*p)->rb_right;
+		else if (cur->val > ins->val)
+			p = &(*p)->rb_left;
+		else
+			return -EEXIST;
+	}
+	rb_link_node(&ins->rb_node, parent, p);
+	rb_insert_color(&ins->rb_node, &ulist->root);
+	return 0;
+}
+
+/**
+ * ulist_add - add an element to the ulist
+ * @ulist:	ulist to add the element to
+ * @val:	value to add to ulist
+ * @aux:	auxiliary value to store along with val
+ * @gfp_mask:	flags to use for allocation
+ *
+ * Note: locking must be provided by the caller. In case of rwlocks write
+ *       locking is needed
+ *
+ * Add an element to a ulist. The @val will only be added if it doesn't
+ * already exist. If it is added, the auxiliary value @aux is stored along with
+ * it. In case @val already exists in the ulist, @aux is ignored, even if
+ * it differs from the already stored value.
+ *
+ * ulist_add returns 0 if @val already exists in ulist and 1 if @val has been
+ * inserted.
+ * In case of allocation failure -ENOMEM is returned and the ulist stays
+ * unaltered.
+ */
+int ulist_add(struct ulist *ulist, u64 val, u64 aux, gfp_t gfp_mask)
+{
+	return ulist_add_merge(ulist, val, aux, NULL, gfp_mask);
+}
+
+int ulist_add_merge(struct ulist *ulist, u64 val, u64 aux,
+		    u64 *old_aux, gfp_t gfp_mask)
+{
+	int ret;
+	struct ulist_node *node;
+
+	node = ulist_rbtree_search(ulist, val);
+	if (node) {
+		if (old_aux)
+			*old_aux = node->aux;
+		return 0;
+	}
+	node = kmalloc(sizeof(*node), gfp_mask);
+	if (!node)
+		return -ENOMEM;
+
+	node->val = val;
+	node->aux = aux;
+#ifdef CONFIG_BTRFS_DEBUG
+	node->seqnum = ulist->nnodes;
+#endif
+
+	ret = ulist_rbtree_insert(ulist, node);
+	ASSERT(!ret);
+	list_add_tail(&node->list, &ulist->nodes);
+	ulist->nnodes++;
+
+	return 1;
+}
+
+/**
+ * ulist_next - iterate ulist
+ * @ulist:	ulist to iterate
+ * @uiter:	iterator variable, initialized with ULIST_ITER_INIT(&iterator)
+ *
+ * Note: locking must be provided by the caller. In case of rwlocks only read
+ *       locking is needed
+ *
+ * This function is used to iterate an ulist.
+ * It returns the next element from the ulist or %NULL when the
+ * end is reached. No guarantee is made with respect to the order in which
+ * the elements are returned. They might neither be returned in order of
+ * addition nor in ascending order.
+ * It is allowed to call ulist_add during an enumeration. Newly added items
+ * are guaranteed to show up in the running enumeration.
+ */
+struct ulist_node *ulist_next(struct ulist *ulist, struct ulist_iterator *uiter)
+{
+	struct ulist_node *node;
+
+	if (list_empty(&ulist->nodes))
+		return NULL;
+	if (uiter->cur_list && uiter->cur_list->next == &ulist->nodes)
+		return NULL;
+	if (uiter->cur_list) {
+		uiter->cur_list = uiter->cur_list->next;
+	} else {
+		uiter->cur_list = ulist->nodes.next;
+#ifdef CONFIG_BTRFS_DEBUG
+		uiter->i = 0;
+#endif
+	}
+	node = list_entry(uiter->cur_list, struct ulist_node, list);
+#ifdef CONFIG_BTRFS_DEBUG
+	ASSERT(node->seqnum == uiter->i);
+	ASSERT(uiter->i >= 0 && uiter->i < ulist->nnodes);
+	uiter->i++;
+#endif
+	return node;
+}
diff --git a/fs/btrfs/ulist.h b/fs/btrfs/ulist.h
new file mode 100644
index 00000000000..7f78cbf5cf4
--- /dev/null
+++ b/fs/btrfs/ulist.h
@@ -0,0 +1,65 @@
+/*
+ * Copyright (C) 2011 STRATO AG
+ * written by Arne Jansen <sensille@gmx.net>
+ * Distributed under the GNU GPL license version 2.
+ *
+ */
+
+#ifndef __ULIST__
+#define __ULIST__
+
+#include <linux/list.h>
+#include <linux/rbtree.h>
+
+/*
+ * ulist is a generic data structure to hold a collection of unique u64
+ * values. The only operations it supports is adding to the list and
+ * enumerating it.
+ * It is possible to store an auxiliary value along with the key.
+ *
+ */
+struct ulist_iterator {
+#ifdef CONFIG_BTRFS_DEBUG
+	int i;
+#endif
+	struct list_head *cur_list;  /* hint to start search */
+};
+
+/*
+ * element of the list
+ */
+struct ulist_node {
+	u64 val;		/* value to store */
+	u64 aux;		/* auxiliary value saved along with the val */
+
+#ifdef CONFIG_BTRFS_DEBUG
+	int seqnum;		/* sequence number this node is added */
+#endif
+
+	struct list_head list;  /* used to link node */
+	struct rb_node rb_node;	/* used to speed up search */
+};
+
+struct ulist {
+	/*
+	 * number of elements stored in list
+	 */
+	unsigned long nnodes;
+
+	struct list_head nodes;
+	struct rb_root root;
+};
+
+void ulist_init(struct ulist *ulist);
+void ulist_reinit(struct ulist *ulist);
+struct ulist *ulist_alloc(gfp_t gfp_mask);
+void ulist_free(struct ulist *ulist);
+int ulist_add(struct ulist *ulist, u64 val, u64 aux, gfp_t gfp_mask);
+int ulist_add_merge(struct ulist *ulist, u64 val, u64 aux,
+		    u64 *old_aux, gfp_t gfp_mask);
+struct ulist_node *ulist_next(struct ulist *ulist,
+			      struct ulist_iterator *uiter);
+
+#define ULIST_ITER_INIT(uiter) ((uiter)->cur_list = NULL)
+
+#endif
diff --git a/fs/btrfs/uuid-tree.c b/fs/btrfs/uuid-tree.c
new file mode 100644
index 00000000000..f6a4c03ee7d
--- /dev/null
+++ b/fs/btrfs/uuid-tree.c
@@ -0,0 +1,355 @@
+/*
+ * Copyright (C) STRATO AG 2013.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+#include <linux/uuid.h>
+#include <asm/unaligned.h>
+#include "ctree.h"
+#include "transaction.h"
+#include "disk-io.h"
+#include "print-tree.h"
+
+
+static void btrfs_uuid_to_key(u8 *uuid, u8 type, struct btrfs_key *key)
+{
+	key->type = type;
+	key->objectid = get_unaligned_le64(uuid);
+	key->offset = get_unaligned_le64(uuid + sizeof(u64));
+}
+
+/* return -ENOENT for !found, < 0 for errors, or 0 if an item was found */
+static int btrfs_uuid_tree_lookup(struct btrfs_root *uuid_root, u8 *uuid,
+				  u8 type, u64 subid)
+{
+	int ret;
+	struct btrfs_path *path = NULL;
+	struct extent_buffer *eb;
+	int slot;
+	u32 item_size;
+	unsigned long offset;
+	struct btrfs_key key;
+
+	if (WARN_ON_ONCE(!uuid_root)) {
+		ret = -ENOENT;
+		goto out;
+	}
+
+	path = btrfs_alloc_path();
+	if (!path) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	btrfs_uuid_to_key(uuid, type, &key);
+	ret = btrfs_search_slot(NULL, uuid_root, &key, path, 0, 0);
+	if (ret < 0) {
+		goto out;
+	} else if (ret > 0) {
+		ret = -ENOENT;
+		goto out;
+	}
+
+	eb = path->nodes[0];
+	slot = path->slots[0];
+	item_size = btrfs_item_size_nr(eb, slot);
+	offset = btrfs_item_ptr_offset(eb, slot);
+	ret = -ENOENT;
+
+	if (!IS_ALIGNED(item_size, sizeof(u64))) {
+		btrfs_warn(uuid_root->fs_info, "uuid item with illegal size %lu!",
+			(unsigned long)item_size);
+		goto out;
+	}
+	while (item_size) {
+		__le64 data;
+
+		read_extent_buffer(eb, &data, offset, sizeof(data));
+		if (le64_to_cpu(data) == subid) {
+			ret = 0;
+			break;
+		}
+		offset += sizeof(data);
+		item_size -= sizeof(data);
+	}
+
+out:
+	btrfs_free_path(path);
+	return ret;
+}
+
+int btrfs_uuid_tree_add(struct btrfs_trans_handle *trans,
+			struct btrfs_root *uuid_root, u8 *uuid, u8 type,
+			u64 subid_cpu)
+{
+	int ret;
+	struct btrfs_path *path = NULL;
+	struct btrfs_key key;
+	struct extent_buffer *eb;
+	int slot;
+	unsigned long offset;
+	__le64 subid_le;
+
+	ret = btrfs_uuid_tree_lookup(uuid_root, uuid, type, subid_cpu);
+	if (ret != -ENOENT)
+		return ret;
+
+	if (WARN_ON_ONCE(!uuid_root)) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	btrfs_uuid_to_key(uuid, type, &key);
+
+	path = btrfs_alloc_path();
+	if (!path) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	ret = btrfs_insert_empty_item(trans, uuid_root, path, &key,
+				      sizeof(subid_le));
+	if (ret >= 0) {
+		/* Add an item for the type for the first time */
+		eb = path->nodes[0];
+		slot = path->slots[0];
+		offset = btrfs_item_ptr_offset(eb, slot);
+	} else if (ret == -EEXIST) {
+		/*
+		 * An item with that type already exists.
+		 * Extend the item and store the new subid at the end.
+		 */
+		btrfs_extend_item(uuid_root, path, sizeof(subid_le));
+		eb = path->nodes[0];
+		slot = path->slots[0];
+		offset = btrfs_item_ptr_offset(eb, slot);
+		offset += btrfs_item_size_nr(eb, slot) - sizeof(subid_le);
+	} else if (ret < 0) {
+		btrfs_warn(uuid_root->fs_info, "insert uuid item failed %d "
+			"(0x%016llx, 0x%016llx) type %u!",
+			ret, (unsigned long long)key.objectid,
+			(unsigned long long)key.offset, type);
+		goto out;
+	}
+
+	ret = 0;
+	subid_le = cpu_to_le64(subid_cpu);
+	write_extent_buffer(eb, &subid_le, offset, sizeof(subid_le));
+	btrfs_mark_buffer_dirty(eb);
+
+out:
+	btrfs_free_path(path);
+	return ret;
+}
+
+int btrfs_uuid_tree_rem(struct btrfs_trans_handle *trans,
+			struct btrfs_root *uuid_root, u8 *uuid, u8 type,
+			u64 subid)
+{
+	int ret;
+	struct btrfs_path *path = NULL;
+	struct btrfs_key key;
+	struct extent_buffer *eb;
+	int slot;
+	unsigned long offset;
+	u32 item_size;
+	unsigned long move_dst;
+	unsigned long move_src;
+	unsigned long move_len;
+
+	if (WARN_ON_ONCE(!uuid_root)) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	btrfs_uuid_to_key(uuid, type, &key);
+
+	path = btrfs_alloc_path();
+	if (!path) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	ret = btrfs_search_slot(trans, uuid_root, &key, path, -1, 1);
+	if (ret < 0) {
+		btrfs_warn(uuid_root->fs_info, "error %d while searching for uuid item!",
+			ret);
+		goto out;
+	}
+	if (ret > 0) {
+		ret = -ENOENT;
+		goto out;
+	}
+
+	eb = path->nodes[0];
+	slot = path->slots[0];
+	offset = btrfs_item_ptr_offset(eb, slot);
+	item_size = btrfs_item_size_nr(eb, slot);
+	if (!IS_ALIGNED(item_size, sizeof(u64))) {
+		btrfs_warn(uuid_root->fs_info, "uuid item with illegal size %lu!",
+			(unsigned long)item_size);
+		ret = -ENOENT;
+		goto out;
+	}
+	while (item_size) {
+		__le64 read_subid;
+
+		read_extent_buffer(eb, &read_subid, offset, sizeof(read_subid));
+		if (le64_to_cpu(read_subid) == subid)
+			break;
+		offset += sizeof(read_subid);
+		item_size -= sizeof(read_subid);
+	}
+
+	if (!item_size) {
+		ret = -ENOENT;
+		goto out;
+	}
+
+	item_size = btrfs_item_size_nr(eb, slot);
+	if (item_size == sizeof(subid)) {
+		ret = btrfs_del_item(trans, uuid_root, path);
+		goto out;
+	}
+
+	move_dst = offset;
+	move_src = offset + sizeof(subid);
+	move_len = item_size - (move_src - btrfs_item_ptr_offset(eb, slot));
+	memmove_extent_buffer(eb, move_dst, move_src, move_len);
+	btrfs_truncate_item(uuid_root, path, item_size - sizeof(subid), 1);
+
+out:
+	btrfs_free_path(path);
+	return ret;
+}
+
+static int btrfs_uuid_iter_rem(struct btrfs_root *uuid_root, u8 *uuid, u8 type,
+			       u64 subid)
+{
+	struct btrfs_trans_handle *trans;
+	int ret;
+
+	/* 1 - for the uuid item */
+	trans = btrfs_start_transaction(uuid_root, 1);
+	if (IS_ERR(trans)) {
+		ret = PTR_ERR(trans);
+		goto out;
+	}
+
+	ret = btrfs_uuid_tree_rem(trans, uuid_root, uuid, type, subid);
+	btrfs_end_transaction(trans, uuid_root);
+
+out:
+	return ret;
+}
+
+int btrfs_uuid_tree_iterate(struct btrfs_fs_info *fs_info,
+			    int (*check_func)(struct btrfs_fs_info *, u8 *, u8,
+					      u64))
+{
+	struct btrfs_root *root = fs_info->uuid_root;
+	struct btrfs_key key;
+	struct btrfs_path *path;
+	int ret = 0;
+	struct extent_buffer *leaf;
+	int slot;
+	u32 item_size;
+	unsigned long offset;
+
+	path = btrfs_alloc_path();
+	if (!path) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	key.objectid = 0;
+	key.type = 0;
+	key.offset = 0;
+
+again_search_slot:
+	path->keep_locks = 1;
+	ret = btrfs_search_forward(root, &key, path, 0);
+	if (ret) {
+		if (ret > 0)
+			ret = 0;
+		goto out;
+	}
+
+	while (1) {
+		cond_resched();
+		leaf = path->nodes[0];
+		slot = path->slots[0];
+		btrfs_item_key_to_cpu(leaf, &key, slot);
+
+		if (key.type != BTRFS_UUID_KEY_SUBVOL &&
+		    key.type != BTRFS_UUID_KEY_RECEIVED_SUBVOL)
+			goto skip;
+
+		offset = btrfs_item_ptr_offset(leaf, slot);
+		item_size = btrfs_item_size_nr(leaf, slot);
+		if (!IS_ALIGNED(item_size, sizeof(u64))) {
+			btrfs_warn(fs_info, "uuid item with illegal size %lu!",
+				(unsigned long)item_size);
+			goto skip;
+		}
+		while (item_size) {
+			u8 uuid[BTRFS_UUID_SIZE];
+			__le64 subid_le;
+			u64 subid_cpu;
+
+			put_unaligned_le64(key.objectid, uuid);
+			put_unaligned_le64(key.offset, uuid + sizeof(u64));
+			read_extent_buffer(leaf, &subid_le, offset,
+					   sizeof(subid_le));
+			subid_cpu = le64_to_cpu(subid_le);
+			ret = check_func(fs_info, uuid, key.type, subid_cpu);
+			if (ret < 0)
+				goto out;
+			if (ret > 0) {
+				btrfs_release_path(path);
+				ret = btrfs_uuid_iter_rem(root, uuid, key.type,
+							  subid_cpu);
+				if (ret == 0) {
+					/*
+					 * this might look inefficient, but the
+					 * justification is that it is an
+					 * exception that check_func returns 1,
+					 * and that in the regular case only one
+					 * entry per UUID exists.
+					 */
+					goto again_search_slot;
+				}
+				if (ret < 0 && ret != -ENOENT)
+					goto out;
+			}
+			item_size -= sizeof(subid_le);
+			offset += sizeof(subid_le);
+		}
+
+skip:
+		ret = btrfs_next_item(root, path);
+		if (ret == 0)
+			continue;
+		else if (ret > 0)
+			ret = 0;
+		break;
+	}
+
+out:
+	btrfs_free_path(path);
+	if (ret)
+		btrfs_warn(fs_info, "btrfs_uuid_tree_iterate failed %d", ret);
+	return 0;
+}
diff --git a/fs/btrfs/version.h b/fs/btrfs/version.h
deleted file mode 100644
index 9bf3946d5ef..00000000000
--- a/fs/btrfs/version.h
+++ /dev/null
@@ -1,4 +0,0 @@
-#ifndef __BTRFS_VERSION_H
-#define __BTRFS_VERSION_H
-#define BTRFS_BUILD_VERSION "Btrfs"
-#endif
diff --git a/fs/btrfs/version.sh b/fs/btrfs/version.sh
deleted file mode 100644
index 1ca1952fd91..00000000000
--- a/fs/btrfs/version.sh
+++ /dev/null
@@ -1,43 +0,0 @@
-#!/bin/bash
-#
-# determine-version -- report a useful version for releases
-#
-# Copyright 2008, Aron Griffis <agriffis@n01se.net>
-# Copyright 2008, Oracle
-# Released under the GNU GPLv2
- 
-v="v0.16"
-
-which git &> /dev/null
-if [ $? == 0 ]; then
-    git branch >& /dev/null
-    if [ $? == 0 ]; then
-	    if head=`git rev-parse --verify HEAD 2>/dev/null`; then
-		if tag=`git describe --tags 2>/dev/null`; then
-		    v="$tag"
-		fi
-
-		# Are there uncommitted changes?
-		git update-index --refresh --unmerged > /dev/null
-		if git diff-index --name-only HEAD | \
-		    grep -v "^scripts/package" \
-		    | read dummy; then
-		    v="$v"-dirty
-		fi
-	    fi
-    fi
-fi
- 
-echo "#ifndef __BUILD_VERSION" > .build-version.h
-echo "#define __BUILD_VERSION" >> .build-version.h
-echo "#define BTRFS_BUILD_VERSION \"Btrfs $v\"" >> .build-version.h
-echo "#endif" >> .build-version.h
-
-diff -q version.h .build-version.h >& /dev/null
-
-if [ $? == 0 ]; then
-    rm .build-version.h
-    exit 0
-fi
-
-mv .build-version.h version.h
diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
index aa7dc36dac7..6cb82f62cb7 100644
--- a/fs/btrfs/volumes.c
+++ b/fs/btrfs/volumes.c
@@ -22,56 +22,87 @@
 #include <linux/blkdev.h>
 #include <linux/random.h>
 #include <linux/iocontext.h>
+#include <linux/capability.h>
+#include <linux/ratelimit.h>
+#include <linux/kthread.h>
+#include <linux/raid/pq.h>
+#include <linux/semaphore.h>
 #include <asm/div64.h>
-#include "compat.h"
 #include "ctree.h"
 #include "extent_map.h"
 #include "disk-io.h"
 #include "transaction.h"
 #include "print-tree.h"
 #include "volumes.h"
+#include "raid56.h"
 #include "async-thread.h"
-
-struct map_lookup {
-	u64 type;
-	int io_align;
-	int io_width;
-	int stripe_len;
-	int sector_size;
-	int num_stripes;
-	int sub_stripes;
-	struct btrfs_bio_stripe stripes[];
-};
+#include "check-integrity.h"
+#include "rcu-string.h"
+#include "math.h"
+#include "dev-replace.h"
+#include "sysfs.h"
 
 static int init_first_rw_device(struct btrfs_trans_handle *trans,
 				struct btrfs_root *root,
 				struct btrfs_device *device);
 static int btrfs_relocate_sys_chunks(struct btrfs_root *root);
-
-#define map_lookup_size(n) (sizeof(struct map_lookup) + \
-			    (sizeof(struct btrfs_bio_stripe) * (n)))
+static void __btrfs_reset_dev_stats(struct btrfs_device *dev);
+static void btrfs_dev_stat_print_on_error(struct btrfs_device *dev);
+static void btrfs_dev_stat_print_on_load(struct btrfs_device *device);
 
 static DEFINE_MUTEX(uuid_mutex);
 static LIST_HEAD(fs_uuids);
 
-void btrfs_lock_volumes(void)
+static void lock_chunks(struct btrfs_root *root)
 {
-	mutex_lock(&uuid_mutex);
+	mutex_lock(&root->fs_info->chunk_mutex);
 }
 
-void btrfs_unlock_volumes(void)
+static void unlock_chunks(struct btrfs_root *root)
 {
-	mutex_unlock(&uuid_mutex);
+	mutex_unlock(&root->fs_info->chunk_mutex);
 }
 
-static void lock_chunks(struct btrfs_root *root)
+static struct btrfs_fs_devices *__alloc_fs_devices(void)
 {
-	mutex_lock(&root->fs_info->chunk_mutex);
+	struct btrfs_fs_devices *fs_devs;
+
+	fs_devs = kzalloc(sizeof(*fs_devs), GFP_NOFS);
+	if (!fs_devs)
+		return ERR_PTR(-ENOMEM);
+
+	mutex_init(&fs_devs->device_list_mutex);
+
+	INIT_LIST_HEAD(&fs_devs->devices);
+	INIT_LIST_HEAD(&fs_devs->alloc_list);
+	INIT_LIST_HEAD(&fs_devs->list);
+
+	return fs_devs;
 }
 
-static void unlock_chunks(struct btrfs_root *root)
+/**
+ * alloc_fs_devices - allocate struct btrfs_fs_devices
+ * @fsid:	a pointer to UUID for this FS.  If NULL a new UUID is
+ *		generated.
+ *
+ * Return: a pointer to a new &struct btrfs_fs_devices on success;
+ * ERR_PTR() on error.  Returned struct is not linked onto any lists and
+ * can be destroyed with kfree() right away.
+ */
+static struct btrfs_fs_devices *alloc_fs_devices(const u8 *fsid)
 {
-	mutex_unlock(&root->fs_info->chunk_mutex);
+	struct btrfs_fs_devices *fs_devs;
+
+	fs_devs = __alloc_fs_devices();
+	if (IS_ERR(fs_devs))
+		return fs_devs;
+
+	if (fsid)
+		memcpy(fs_devs->fsid, fsid, BTRFS_FSID_SIZE);
+	else
+		generate_random_uuid(fs_devs->fsid);
+
+	return fs_devs;
 }
 
 static void free_fs_devices(struct btrfs_fs_devices *fs_devices)
@@ -82,13 +113,26 @@ static void free_fs_devices(struct btrfs_fs_devices *fs_devices)
 		device = list_entry(fs_devices->devices.next,
 				    struct btrfs_device, dev_list);
 		list_del(&device->dev_list);
-		kfree(device->name);
+		rcu_string_free(device->name);
 		kfree(device);
 	}
 	kfree(fs_devices);
 }
 
-int btrfs_cleanup_fs_uuids(void)
+static void btrfs_kobject_uevent(struct block_device *bdev,
+				 enum kobject_action action)
+{
+	int ret;
+
+	ret = kobject_uevent(&disk_to_dev(bdev->bd_disk)->kobj, action);
+	if (ret)
+		pr_warn("BTRFS: Sending event '%d' to kobject: '%s' (%p): failed\n",
+			action,
+			kobject_name(&disk_to_dev(bdev->bd_disk)->kobj),
+			&disk_to_dev(bdev->bd_disk)->kobj);
+}
+
+void btrfs_cleanup_fs_uuids(void)
 {
 	struct btrfs_fs_devices *fs_devices;
 
@@ -98,7 +142,27 @@ int btrfs_cleanup_fs_uuids(void)
 		list_del(&fs_devices->list);
 		free_fs_devices(fs_devices);
 	}
-	return 0;
+}
+
+static struct btrfs_device *__alloc_device(void)
+{
+	struct btrfs_device *dev;
+
+	dev = kzalloc(sizeof(*dev), GFP_NOFS);
+	if (!dev)
+		return ERR_PTR(-ENOMEM);
+
+	INIT_LIST_HEAD(&dev->dev_list);
+	INIT_LIST_HEAD(&dev->dev_alloc_list);
+
+	spin_lock_init(&dev->io_lock);
+
+	spin_lock_init(&dev->reada_lock);
+	atomic_set(&dev->reada_in_flight, 0);
+	INIT_RADIX_TREE(&dev->reada_zones, GFP_NOFS & ~__GFP_WAIT);
+	INIT_RADIX_TREE(&dev->reada_extents, GFP_NOFS & ~__GFP_WAIT);
+
+	return dev;
 }
 
 static noinline struct btrfs_device *__find_device(struct list_head *head,
@@ -126,6 +190,44 @@ static noinline struct btrfs_fs_devices *find_fsid(u8 *fsid)
 	return NULL;
 }
 
+static int
+btrfs_get_bdev_and_sb(const char *device_path, fmode_t flags, void *holder,
+		      int flush, struct block_device **bdev,
+		      struct buffer_head **bh)
+{
+	int ret;
+
+	*bdev = blkdev_get_by_path(device_path, flags, holder);
+
+	if (IS_ERR(*bdev)) {
+		ret = PTR_ERR(*bdev);
+		printk(KERN_INFO "BTRFS: open %s failed\n", device_path);
+		goto error;
+	}
+
+	if (flush)
+		filemap_write_and_wait((*bdev)->bd_inode->i_mapping);
+	ret = set_blocksize(*bdev, 4096);
+	if (ret) {
+		blkdev_put(*bdev, flags);
+		goto error;
+	}
+	invalidate_bdev(*bdev);
+	*bh = btrfs_read_dev_super(*bdev);
+	if (!*bh) {
+		ret = -EINVAL;
+		blkdev_put(*bdev, flags);
+		goto error;
+	}
+
+	return 0;
+
+error:
+	*bdev = NULL;
+	*bh = NULL;
+	return ret;
+}
+
 static void requeue_list(struct btrfs_pending_bios *pending_bios,
 			struct bio *head, struct bio *tail)
 {
@@ -151,7 +253,7 @@ static void requeue_list(struct btrfs_pending_bios *pending_bios,
  * the list if the block device is congested.  This way, multiple devices
  * can make progress from a single worker thread.
  */
-static noinline int run_scheduled_bios(struct btrfs_device *device)
+static noinline void run_scheduled_bios(struct btrfs_device *device)
 {
 	struct bio *pending;
 	struct backing_dev_info *bdi;
@@ -161,22 +263,26 @@ static noinline int run_scheduled_bios(struct btrfs_device *device)
 	struct bio *cur;
 	int again = 0;
 	unsigned long num_run;
-	unsigned long num_sync_run;
 	unsigned long batch_run = 0;
 	unsigned long limit;
 	unsigned long last_waited = 0;
 	int force_reg = 0;
+	int sync_pending = 0;
+	struct blk_plug plug;
+
+	/*
+	 * this function runs all the bios we've collected for
+	 * a particular device.  We don't want to wander off to
+	 * another device without first sending all of these down.
+	 * So, setup a plug here and finish it off before we return
+	 */
+	blk_start_plug(&plug);
 
 	bdi = blk_get_backing_dev_info(device->bdev);
 	fs_info = device->dev_root->fs_info;
 	limit = btrfs_async_submit_limit(fs_info);
 	limit = limit * 2 / 3;
 
-	/* we want to make sure that every time we switch from the sync
-	 * list to the normal list, we unplug
-	 */
-	num_sync_run = 0;
-
 loop:
 	spin_lock(&device->io_lock);
 
@@ -222,15 +328,6 @@ loop_lock:
 
 	spin_unlock(&device->io_lock);
 
-	/*
-	 * if we're doing the regular priority list, make sure we unplug
-	 * for any high prio bios we've sent down
-	 */
-	if (pending_bios == &device->pending_bios && num_sync_run > 0) {
-		num_sync_run = 0;
-		blk_run_backing_dev(bdi, NULL);
-	}
-
 	while (pending) {
 
 		rmb();
@@ -250,27 +347,34 @@ loop_lock:
 		cur = pending;
 		pending = pending->bi_next;
 		cur->bi_next = NULL;
-		atomic_dec(&fs_info->nr_async_bios);
 
-		if (atomic_read(&fs_info->nr_async_bios) < limit &&
+		if (atomic_dec_return(&fs_info->nr_async_bios) < limit &&
 		    waitqueue_active(&fs_info->async_submit_wait))
 			wake_up(&fs_info->async_submit_wait);
 
 		BUG_ON(atomic_read(&cur->bi_cnt) == 0);
 
-		if (bio_rw_flagged(cur, BIO_RW_SYNCIO))
-			num_sync_run++;
+		/*
+		 * if we're doing the sync list, record that our
+		 * plug has some sync requests on it
+		 *
+		 * If we're doing the regular list and there are
+		 * sync requests sitting around, unplug before
+		 * we add more
+		 */
+		if (pending_bios == &device->pending_sync_bios) {
+			sync_pending = 1;
+		} else if (sync_pending) {
+			blk_finish_plug(&plug);
+			blk_start_plug(&plug);
+			sync_pending = 0;
+		}
 
-		submit_bio(cur->bi_rw, cur);
+		btrfsic_submit_bio(cur->bi_rw, cur);
 		num_run++;
 		batch_run++;
-		if (need_resched()) {
-			if (num_sync_run) {
-				blk_run_backing_dev(bdi, NULL);
-				num_sync_run = 0;
-			}
+		if (need_resched())
 			cond_resched();
-		}
 
 		/*
 		 * we made progress, there is more work to do and the bdi
@@ -303,13 +407,8 @@ loop_lock:
 				 * against it before looping
 				 */
 				last_waited = ioc->last_waited;
-				if (need_resched()) {
-					if (num_sync_run) {
-						blk_run_backing_dev(bdi, NULL);
-						num_sync_run = 0;
-					}
+				if (need_resched())
 					cond_resched();
-				}
 				continue;
 			}
 			spin_lock(&device->io_lock);
@@ -317,27 +416,18 @@ loop_lock:
 			device->running_pending = 1;
 
 			spin_unlock(&device->io_lock);
-			btrfs_requeue_work(&device->work);
+			btrfs_queue_work(fs_info->submit_workers,
+					 &device->work);
 			goto done;
 		}
+		/* unplug every 64 requests just for good measure */
+		if (batch_run % 64 == 0) {
+			blk_finish_plug(&plug);
+			blk_start_plug(&plug);
+			sync_pending = 0;
+		}
 	}
 
-	if (num_sync_run) {
-		num_sync_run = 0;
-		blk_run_backing_dev(bdi, NULL);
-	}
-	/*
-	 * IO has already been through a long path to get here.  Checksumming,
-	 * async helper threads, perhaps compression.  We've done a pretty
-	 * good job of collecting a batch of IO and should just unplug
-	 * the device right away.
-	 *
-	 * This will help anyone who is waiting on the IO, they might have
-	 * already unplugged, but managed to do so before the bio they
-	 * cared about found its way down here.
-	 */
-	blk_run_backing_dev(bdi, NULL);
-
 	cond_resched();
 	if (again)
 		goto loop;
@@ -348,7 +438,7 @@ loop_lock:
 	spin_unlock(&device->io_lock);
 
 done:
-	return 0;
+	blk_finish_plug(&plug);
 }
 
 static void pending_bios_fn(struct btrfs_work *work)
@@ -359,27 +449,34 @@ static void pending_bios_fn(struct btrfs_work *work)
 	run_scheduled_bios(device);
 }
 
+/*
+ * Add new device to list of registered devices
+ *
+ * Returns:
+ * 1   - first time device is seen
+ * 0   - device already known
+ * < 0 - error
+ */
 static noinline int device_list_add(const char *path,
 			   struct btrfs_super_block *disk_super,
 			   u64 devid, struct btrfs_fs_devices **fs_devices_ret)
 {
 	struct btrfs_device *device;
 	struct btrfs_fs_devices *fs_devices;
+	struct rcu_string *name;
+	int ret = 0;
 	u64 found_transid = btrfs_super_generation(disk_super);
-	char *name;
 
 	fs_devices = find_fsid(disk_super->fsid);
 	if (!fs_devices) {
-		fs_devices = kzalloc(sizeof(*fs_devices), GFP_NOFS);
-		if (!fs_devices)
-			return -ENOMEM;
-		INIT_LIST_HEAD(&fs_devices->devices);
-		INIT_LIST_HEAD(&fs_devices->alloc_list);
+		fs_devices = alloc_fs_devices(disk_super->fsid);
+		if (IS_ERR(fs_devices))
+			return PTR_ERR(fs_devices);
+
 		list_add(&fs_devices->list, &fs_uuids);
-		memcpy(fs_devices->fsid, disk_super->fsid, BTRFS_FSID_SIZE);
 		fs_devices->latest_devid = devid;
 		fs_devices->latest_trans = found_transid;
-		mutex_init(&fs_devices->device_list_mutex);
+
 		device = NULL;
 	} else {
 		device = __find_device(&fs_devices->devices, devid,
@@ -389,36 +486,37 @@ static noinline int device_list_add(const char *path,
 		if (fs_devices->opened)
 			return -EBUSY;
 
-		device = kzalloc(sizeof(*device), GFP_NOFS);
-		if (!device) {
+		device = btrfs_alloc_device(NULL, &devid,
+					    disk_super->dev_item.uuid);
+		if (IS_ERR(device)) {
 			/* we can safely leave the fs_devices entry around */
-			return -ENOMEM;
+			return PTR_ERR(device);
 		}
-		device->devid = devid;
-		device->work.func = pending_bios_fn;
-		memcpy(device->uuid, disk_super->dev_item.uuid,
-		       BTRFS_UUID_SIZE);
-		device->barriers = 1;
-		spin_lock_init(&device->io_lock);
-		device->name = kstrdup(path, GFP_NOFS);
-		if (!device->name) {
+
+		name = rcu_string_strdup(path, GFP_NOFS);
+		if (!name) {
 			kfree(device);
 			return -ENOMEM;
 		}
-		INIT_LIST_HEAD(&device->dev_alloc_list);
+		rcu_assign_pointer(device->name, name);
 
 		mutex_lock(&fs_devices->device_list_mutex);
-		list_add(&device->dev_list, &fs_devices->devices);
+		list_add_rcu(&device->dev_list, &fs_devices->devices);
+		fs_devices->num_devices++;
 		mutex_unlock(&fs_devices->device_list_mutex);
 
+		ret = 1;
 		device->fs_devices = fs_devices;
-		fs_devices->num_devices++;
-	} else if (strcmp(device->name, path)) {
-		name = kstrdup(path, GFP_NOFS);
+	} else if (!device->name || strcmp(device->name->str, path)) {
+		name = rcu_string_strdup(path, GFP_NOFS);
 		if (!name)
 			return -ENOMEM;
-		kfree(device->name);
-		device->name = name;
+		rcu_string_free(device->name);
+		rcu_assign_pointer(device->name, name);
+		if (device->missing) {
+			fs_devices->missing_devices--;
+			device->missing = 0;
+		}
 	}
 
 	if (found_transid > fs_devices->latest_trans) {
@@ -426,7 +524,8 @@ static noinline int device_list_add(const char *path,
 		fs_devices->latest_trans = found_transid;
 	}
 	*fs_devices_ret = fs_devices;
-	return 0;
+
+	return ret;
 }
 
 static struct btrfs_fs_devices *clone_fs_devices(struct btrfs_fs_devices *orig)
@@ -435,85 +534,135 @@ static struct btrfs_fs_devices *clone_fs_devices(struct btrfs_fs_devices *orig)
 	struct btrfs_device *device;
 	struct btrfs_device *orig_dev;
 
-	fs_devices = kzalloc(sizeof(*fs_devices), GFP_NOFS);
-	if (!fs_devices)
-		return ERR_PTR(-ENOMEM);
+	fs_devices = alloc_fs_devices(orig->fsid);
+	if (IS_ERR(fs_devices))
+		return fs_devices;
 
-	INIT_LIST_HEAD(&fs_devices->devices);
-	INIT_LIST_HEAD(&fs_devices->alloc_list);
-	INIT_LIST_HEAD(&fs_devices->list);
-	mutex_init(&fs_devices->device_list_mutex);
 	fs_devices->latest_devid = orig->latest_devid;
 	fs_devices->latest_trans = orig->latest_trans;
-	memcpy(fs_devices->fsid, orig->fsid, sizeof(fs_devices->fsid));
+	fs_devices->total_devices = orig->total_devices;
 
-	mutex_lock(&orig->device_list_mutex);
+	/* We have held the volume lock, it is safe to get the devices. */
 	list_for_each_entry(orig_dev, &orig->devices, dev_list) {
-		device = kzalloc(sizeof(*device), GFP_NOFS);
-		if (!device)
-			goto error;
+		struct rcu_string *name;
 
-		device->name = kstrdup(orig_dev->name, GFP_NOFS);
-		if (!device->name) {
-			kfree(device);
+		device = btrfs_alloc_device(NULL, &orig_dev->devid,
+					    orig_dev->uuid);
+		if (IS_ERR(device))
 			goto error;
-		}
 
-		device->devid = orig_dev->devid;
-		device->work.func = pending_bios_fn;
-		memcpy(device->uuid, orig_dev->uuid, sizeof(device->uuid));
-		device->barriers = 1;
-		spin_lock_init(&device->io_lock);
-		INIT_LIST_HEAD(&device->dev_list);
-		INIT_LIST_HEAD(&device->dev_alloc_list);
+		/*
+		 * This is ok to do without rcu read locked because we hold the
+		 * uuid mutex so nothing we touch in here is going to disappear.
+		 */
+		if (orig_dev->name) {
+			name = rcu_string_strdup(orig_dev->name->str, GFP_NOFS);
+			if (!name) {
+				kfree(device);
+				goto error;
+			}
+			rcu_assign_pointer(device->name, name);
+		}
 
 		list_add(&device->dev_list, &fs_devices->devices);
 		device->fs_devices = fs_devices;
 		fs_devices->num_devices++;
 	}
-	mutex_unlock(&orig->device_list_mutex);
 	return fs_devices;
 error:
-	mutex_unlock(&orig->device_list_mutex);
 	free_fs_devices(fs_devices);
 	return ERR_PTR(-ENOMEM);
 }
 
-int btrfs_close_extra_devices(struct btrfs_fs_devices *fs_devices)
+void btrfs_close_extra_devices(struct btrfs_fs_info *fs_info,
+			       struct btrfs_fs_devices *fs_devices, int step)
 {
 	struct btrfs_device *device, *next;
 
+	struct block_device *latest_bdev = NULL;
+	u64 latest_devid = 0;
+	u64 latest_transid = 0;
+
 	mutex_lock(&uuid_mutex);
 again:
-	mutex_lock(&fs_devices->device_list_mutex);
+	/* This is the initialized path, it is safe to release the devices. */
 	list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
-		if (device->in_fs_metadata)
+		if (device->in_fs_metadata) {
+			if (!device->is_tgtdev_for_dev_replace &&
+			    (!latest_transid ||
+			     device->generation > latest_transid)) {
+				latest_devid = device->devid;
+				latest_transid = device->generation;
+				latest_bdev = device->bdev;
+			}
 			continue;
+		}
 
+		if (device->devid == BTRFS_DEV_REPLACE_DEVID) {
+			/*
+			 * In the first step, keep the device which has
+			 * the correct fsid and the devid that is used
+			 * for the dev_replace procedure.
+			 * In the second step, the dev_replace state is
+			 * read from the device tree and it is known
+			 * whether the procedure is really active or
+			 * not, which means whether this device is
+			 * used or whether it should be removed.
+			 */
+			if (step == 0 || device->is_tgtdev_for_dev_replace) {
+				continue;
+			}
+		}
 		if (device->bdev) {
-			close_bdev_exclusive(device->bdev, device->mode);
+			blkdev_put(device->bdev, device->mode);
 			device->bdev = NULL;
 			fs_devices->open_devices--;
 		}
 		if (device->writeable) {
 			list_del_init(&device->dev_alloc_list);
 			device->writeable = 0;
-			fs_devices->rw_devices--;
+			if (!device->is_tgtdev_for_dev_replace)
+				fs_devices->rw_devices--;
 		}
 		list_del_init(&device->dev_list);
 		fs_devices->num_devices--;
-		kfree(device->name);
+		rcu_string_free(device->name);
 		kfree(device);
 	}
-	mutex_unlock(&fs_devices->device_list_mutex);
 
 	if (fs_devices->seed) {
 		fs_devices = fs_devices->seed;
 		goto again;
 	}
 
+	fs_devices->latest_bdev = latest_bdev;
+	fs_devices->latest_devid = latest_devid;
+	fs_devices->latest_trans = latest_transid;
+
 	mutex_unlock(&uuid_mutex);
-	return 0;
+}
+
+static void __free_device(struct work_struct *work)
+{
+	struct btrfs_device *device;
+
+	device = container_of(work, struct btrfs_device, rcu_work);
+
+	if (device->bdev)
+		blkdev_put(device->bdev, device->mode);
+
+	rcu_string_free(device->name);
+	kfree(device);
+}
+
+static void free_device(struct rcu_head *head)
+{
+	struct btrfs_device *device;
+
+	device = container_of(head, struct btrfs_device, rcu);
+
+	INIT_WORK(&device->rcu_work, __free_device);
+	schedule_work(&device->rcu_work);
 }
 
 static int __btrfs_close_devices(struct btrfs_fs_devices *fs_devices)
@@ -523,20 +672,43 @@ static int __btrfs_close_devices(struct btrfs_fs_devices *fs_devices)
 	if (--fs_devices->opened > 0)
 		return 0;
 
+	mutex_lock(&fs_devices->device_list_mutex);
 	list_for_each_entry(device, &fs_devices->devices, dev_list) {
-		if (device->bdev) {
-			close_bdev_exclusive(device->bdev, device->mode);
+		struct btrfs_device *new_device;
+		struct rcu_string *name;
+
+		if (device->bdev)
 			fs_devices->open_devices--;
-		}
-		if (device->writeable) {
+
+		if (device->writeable &&
+		    device->devid != BTRFS_DEV_REPLACE_DEVID) {
 			list_del_init(&device->dev_alloc_list);
 			fs_devices->rw_devices--;
 		}
 
-		device->bdev = NULL;
-		device->writeable = 0;
-		device->in_fs_metadata = 0;
+		if (device->can_discard)
+			fs_devices->num_can_discard--;
+		if (device->missing)
+			fs_devices->missing_devices--;
+
+		new_device = btrfs_alloc_device(NULL, &device->devid,
+						device->uuid);
+		BUG_ON(IS_ERR(new_device)); /* -ENOMEM */
+
+		/* Safe because we are under uuid_mutex */
+		if (device->name) {
+			name = rcu_string_strdup(device->name->str, GFP_NOFS);
+			BUG_ON(!name); /* -ENOMEM */
+			rcu_assign_pointer(new_device->name, name);
+		}
+
+		list_replace_rcu(&device->dev_list, &new_device->dev_list);
+		new_device->fs_devices = device->fs_devices;
+
+		call_rcu(&device->rcu, free_device);
 	}
+	mutex_unlock(&fs_devices->device_list_mutex);
+
 	WARN_ON(fs_devices->open_devices);
 	WARN_ON(fs_devices->rw_devices);
 	fs_devices->opened = 0;
@@ -564,12 +736,19 @@ int btrfs_close_devices(struct btrfs_fs_devices *fs_devices)
 		__btrfs_close_devices(fs_devices);
 		free_fs_devices(fs_devices);
 	}
+	/*
+	 * Wait for rcu kworkers under __btrfs_close_devices
+	 * to finish all blkdev_puts so device is really
+	 * free when umount is done.
+	 */
+	rcu_barrier();
 	return ret;
 }
 
 static int __btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
 				fmode_t flags, void *holder)
 {
+	struct request_queue *q;
 	struct block_device *bdev;
 	struct list_head *head = &fs_devices->devices;
 	struct btrfs_device *device;
@@ -582,22 +761,18 @@ static int __btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
 	int seeding = 1;
 	int ret = 0;
 
+	flags |= FMODE_EXCL;
+
 	list_for_each_entry(device, head, dev_list) {
 		if (device->bdev)
 			continue;
 		if (!device->name)
 			continue;
 
-		bdev = open_bdev_exclusive(device->name, flags, holder);
-		if (IS_ERR(bdev)) {
-			printk(KERN_INFO "open %s failed\n", device->name);
-			goto error;
-		}
-		set_blocksize(bdev, 4096);
-
-		bh = btrfs_read_dev_super(bdev);
-		if (!bh)
-			goto error_close;
+		/* Just open everything we can; ignore failures here */
+		if (btrfs_get_bdev_and_sb(device->name->str, flags, holder, 1,
+					    &bdev, &bh))
+			continue;
 
 		disk_super = (struct btrfs_super_block *)bh->b_data;
 		devid = btrfs_stack_device_id(&disk_super->dev_item);
@@ -622,6 +797,12 @@ static int __btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
 			seeding = 0;
 		}
 
+		q = bdev_get_queue(bdev);
+		if (blk_queue_discard(q)) {
+			device->can_discard = 1;
+			fs_devices->num_can_discard++;
+		}
+
 		device->bdev = bdev;
 		device->in_fs_metadata = 0;
 		device->mode = flags;
@@ -630,22 +811,22 @@ static int __btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
 			fs_devices->rotating = 1;
 
 		fs_devices->open_devices++;
-		if (device->writeable) {
+		if (device->writeable &&
+		    device->devid != BTRFS_DEV_REPLACE_DEVID) {
 			fs_devices->rw_devices++;
 			list_add(&device->dev_alloc_list,
 				 &fs_devices->alloc_list);
 		}
+		brelse(bh);
 		continue;
 
 error_brelse:
 		brelse(bh);
-error_close:
-		close_bdev_exclusive(bdev, FMODE_READ);
-error:
+		blkdev_put(bdev, flags);
 		continue;
 	}
 	if (fs_devices->open_devices == 0) {
-		ret = -EIO;
+		ret = -EINVAL;
 		goto out;
 	}
 	fs_devices->seeding = seeding;
@@ -674,109 +855,290 @@ int btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
 	return ret;
 }
 
+/*
+ * Look for a btrfs signature on a device. This may be called out of the mount path
+ * and we are not allowed to call set_blocksize during the scan. The superblock
+ * is read via pagecache
+ */
 int btrfs_scan_one_device(const char *path, fmode_t flags, void *holder,
 			  struct btrfs_fs_devices **fs_devices_ret)
 {
 	struct btrfs_super_block *disk_super;
 	struct block_device *bdev;
-	struct buffer_head *bh;
-	int ret;
+	struct page *page;
+	void *p;
+	int ret = -EINVAL;
 	u64 devid;
 	u64 transid;
+	u64 total_devices;
+	u64 bytenr;
+	pgoff_t index;
 
+	/*
+	 * we would like to check all the supers, but that would make
+	 * a btrfs mount succeed after a mkfs from a different FS.
+	 * So, we need to add a special mount option to scan for
+	 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
+	 */
+	bytenr = btrfs_sb_offset(0);
+	flags |= FMODE_EXCL;
 	mutex_lock(&uuid_mutex);
 
-	bdev = open_bdev_exclusive(path, flags, holder);
+	bdev = blkdev_get_by_path(path, flags, holder);
 
 	if (IS_ERR(bdev)) {
 		ret = PTR_ERR(bdev);
 		goto error;
 	}
 
-	ret = set_blocksize(bdev, 4096);
-	if (ret)
-		goto error_close;
-	bh = btrfs_read_dev_super(bdev);
-	if (!bh) {
-		ret = -EIO;
-		goto error_close;
-	}
-	disk_super = (struct btrfs_super_block *)bh->b_data;
+	/* make sure our super fits in the device */
+	if (bytenr + PAGE_CACHE_SIZE >= i_size_read(bdev->bd_inode))
+		goto error_bdev_put;
+
+	/* make sure our super fits in the page */
+	if (sizeof(*disk_super) > PAGE_CACHE_SIZE)
+		goto error_bdev_put;
+
+	/* make sure our super doesn't straddle pages on disk */
+	index = bytenr >> PAGE_CACHE_SHIFT;
+	if ((bytenr + sizeof(*disk_super) - 1) >> PAGE_CACHE_SHIFT != index)
+		goto error_bdev_put;
+
+	/* pull in the page with our super */
+	page = read_cache_page_gfp(bdev->bd_inode->i_mapping,
+				   index, GFP_NOFS);
+
+	if (IS_ERR_OR_NULL(page))
+		goto error_bdev_put;
+
+	p = kmap(page);
+
+	/* align our pointer to the offset of the super block */
+	disk_super = p + (bytenr & ~PAGE_CACHE_MASK);
+
+	if (btrfs_super_bytenr(disk_super) != bytenr ||
+	    btrfs_super_magic(disk_super) != BTRFS_MAGIC)
+		goto error_unmap;
+
 	devid = btrfs_stack_device_id(&disk_super->dev_item);
 	transid = btrfs_super_generation(disk_super);
-	if (disk_super->label[0])
-		printk(KERN_INFO "device label %s ", disk_super->label);
-	else {
-		/* FIXME, make a readl uuid parser */
-		printk(KERN_INFO "device fsid %llx-%llx ",
-		       *(unsigned long long *)disk_super->fsid,
-		       *(unsigned long long *)(disk_super->fsid + 8));
-	}
-	printk(KERN_CONT "devid %llu transid %llu %s\n",
-	       (unsigned long long)devid, (unsigned long long)transid, path);
+	total_devices = btrfs_super_num_devices(disk_super);
+
 	ret = device_list_add(path, disk_super, devid, fs_devices_ret);
+	if (ret > 0) {
+		if (disk_super->label[0]) {
+			if (disk_super->label[BTRFS_LABEL_SIZE - 1])
+				disk_super->label[BTRFS_LABEL_SIZE - 1] = '\0';
+			printk(KERN_INFO "BTRFS: device label %s ", disk_super->label);
+		} else {
+			printk(KERN_INFO "BTRFS: device fsid %pU ", disk_super->fsid);
+		}
 
-	brelse(bh);
-error_close:
-	close_bdev_exclusive(bdev, flags);
+		printk(KERN_CONT "devid %llu transid %llu %s\n", devid, transid, path);
+		ret = 0;
+	}
+	if (!ret && fs_devices_ret)
+		(*fs_devices_ret)->total_devices = total_devices;
+
+error_unmap:
+	kunmap(page);
+	page_cache_release(page);
+
+error_bdev_put:
+	blkdev_put(bdev, flags);
 error:
 	mutex_unlock(&uuid_mutex);
 	return ret;
 }
 
+/* helper to account the used device space in the range */
+int btrfs_account_dev_extents_size(struct btrfs_device *device, u64 start,
+				   u64 end, u64 *length)
+{
+	struct btrfs_key key;
+	struct btrfs_root *root = device->dev_root;
+	struct btrfs_dev_extent *dev_extent;
+	struct btrfs_path *path;
+	u64 extent_end;
+	int ret;
+	int slot;
+	struct extent_buffer *l;
+
+	*length = 0;
+
+	if (start >= device->total_bytes || device->is_tgtdev_for_dev_replace)
+		return 0;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+	path->reada = 2;
+
+	key.objectid = device->devid;
+	key.offset = start;
+	key.type = BTRFS_DEV_EXTENT_KEY;
+
+	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+	if (ret < 0)
+		goto out;
+	if (ret > 0) {
+		ret = btrfs_previous_item(root, path, key.objectid, key.type);
+		if (ret < 0)
+			goto out;
+	}
+
+	while (1) {
+		l = path->nodes[0];
+		slot = path->slots[0];
+		if (slot >= btrfs_header_nritems(l)) {
+			ret = btrfs_next_leaf(root, path);
+			if (ret == 0)
+				continue;
+			if (ret < 0)
+				goto out;
+
+			break;
+		}
+		btrfs_item_key_to_cpu(l, &key, slot);
+
+		if (key.objectid < device->devid)
+			goto next;
+
+		if (key.objectid > device->devid)
+			break;
+
+		if (btrfs_key_type(&key) != BTRFS_DEV_EXTENT_KEY)
+			goto next;
+
+		dev_extent = btrfs_item_ptr(l, slot, struct btrfs_dev_extent);
+		extent_end = key.offset + btrfs_dev_extent_length(l,
+								  dev_extent);
+		if (key.offset <= start && extent_end > end) {
+			*length = end - start + 1;
+			break;
+		} else if (key.offset <= start && extent_end > start)
+			*length += extent_end - start;
+		else if (key.offset > start && extent_end <= end)
+			*length += extent_end - key.offset;
+		else if (key.offset > start && key.offset <= end) {
+			*length += end - key.offset + 1;
+			break;
+		} else if (key.offset > end)
+			break;
+
+next:
+		path->slots[0]++;
+	}
+	ret = 0;
+out:
+	btrfs_free_path(path);
+	return ret;
+}
+
+static int contains_pending_extent(struct btrfs_trans_handle *trans,
+				   struct btrfs_device *device,
+				   u64 *start, u64 len)
+{
+	struct extent_map *em;
+	int ret = 0;
+
+	list_for_each_entry(em, &trans->transaction->pending_chunks, list) {
+		struct map_lookup *map;
+		int i;
+
+		map = (struct map_lookup *)em->bdev;
+		for (i = 0; i < map->num_stripes; i++) {
+			if (map->stripes[i].dev != device)
+				continue;
+			if (map->stripes[i].physical >= *start + len ||
+			    map->stripes[i].physical + em->orig_block_len <=
+			    *start)
+				continue;
+			*start = map->stripes[i].physical +
+				em->orig_block_len;
+			ret = 1;
+		}
+	}
+
+	return ret;
+}
+
+
 /*
+ * find_free_dev_extent - find free space in the specified device
+ * @device:	the device which we search the free space in
+ * @num_bytes:	the size of the free space that we need
+ * @start:	store the start of the free space.
+ * @len:	the size of the free space. that we find, or the size of the max
+ * 		free space if we don't find suitable free space
+ *
  * this uses a pretty simple search, the expectation is that it is
  * called very infrequently and that a given device has a small number
  * of extents
+ *
+ * @start is used to store the start of the free space if we find. But if we
+ * don't find suitable free space, it will be used to store the start position
+ * of the max free space.
+ *
+ * @len is used to store the size of the free space that we find.
+ * But if we don't find suitable free space, it is used to store the size of
+ * the max free space.
  */
 int find_free_dev_extent(struct btrfs_trans_handle *trans,
 			 struct btrfs_device *device, u64 num_bytes,
-			 u64 *start, u64 *max_avail)
+			 u64 *start, u64 *len)
 {
 	struct btrfs_key key;
 	struct btrfs_root *root = device->dev_root;
-	struct btrfs_dev_extent *dev_extent = NULL;
+	struct btrfs_dev_extent *dev_extent;
 	struct btrfs_path *path;
-	u64 hole_size = 0;
-	u64 last_byte = 0;
-	u64 search_start = 0;
+	u64 hole_size;
+	u64 max_hole_start;
+	u64 max_hole_size;
+	u64 extent_end;
+	u64 search_start;
 	u64 search_end = device->total_bytes;
 	int ret;
-	int slot = 0;
-	int start_found;
+	int slot;
 	struct extent_buffer *l;
 
-	path = btrfs_alloc_path();
-	if (!path)
-		return -ENOMEM;
-	path->reada = 2;
-	start_found = 0;
-
 	/* FIXME use last free of some kind */
 
 	/* we don't want to overwrite the superblock on the drive,
 	 * so we make sure to start at an offset of at least 1MB
 	 */
-	search_start = max((u64)1024 * 1024, search_start);
+	search_start = max(root->fs_info->alloc_start, 1024ull * 1024);
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+again:
+	max_hole_start = search_start;
+	max_hole_size = 0;
+	hole_size = 0;
+
+	if (search_start >= search_end || device->is_tgtdev_for_dev_replace) {
+		ret = -ENOSPC;
+		goto out;
+	}
 
-	if (root->fs_info->alloc_start + num_bytes <= device->total_bytes)
-		search_start = max(root->fs_info->alloc_start, search_start);
+	path->reada = 2;
+	path->search_commit_root = 1;
+	path->skip_locking = 1;
 
 	key.objectid = device->devid;
 	key.offset = search_start;
 	key.type = BTRFS_DEV_EXTENT_KEY;
-	ret = btrfs_search_slot(trans, root, &key, path, 0, 0);
+
+	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
 	if (ret < 0)
-		goto error;
+		goto out;
 	if (ret > 0) {
 		ret = btrfs_previous_item(root, path, key.objectid, key.type);
 		if (ret < 0)
-			goto error;
-		if (ret > 0)
-			start_found = 1;
+			goto out;
 	}
-	l = path->nodes[0];
-	btrfs_item_key_to_cpu(l, &key, path->slots[0]);
+
 	while (1) {
 		l = path->nodes[0];
 		slot = path->slots[0];
@@ -785,24 +1147,9 @@ int find_free_dev_extent(struct btrfs_trans_handle *trans,
 			if (ret == 0)
 				continue;
 			if (ret < 0)
-				goto error;
-no_more_items:
-			if (!start_found) {
-				if (search_start >= search_end) {
-					ret = -ENOSPC;
-					goto error;
-				}
-				*start = search_start;
-				start_found = 1;
-				goto check_pending;
-			}
-			*start = last_byte > search_start ?
-				last_byte : search_start;
-			if (search_end <= *start) {
-				ret = -ENOSPC;
-				goto error;
-			}
-			goto check_pending;
+				goto out;
+
+			break;
 		}
 		btrfs_item_key_to_cpu(l, &key, slot);
 
@@ -810,48 +1157,82 @@ no_more_items:
 			goto next;
 
 		if (key.objectid > device->devid)
-			goto no_more_items;
+			break;
 
-		if (key.offset >= search_start && key.offset > last_byte &&
-		    start_found) {
-			if (last_byte < search_start)
-				last_byte = search_start;
-			hole_size = key.offset - last_byte;
+		if (btrfs_key_type(&key) != BTRFS_DEV_EXTENT_KEY)
+			goto next;
+
+		if (key.offset > search_start) {
+			hole_size = key.offset - search_start;
 
-			if (hole_size > *max_avail)
-				*max_avail = hole_size;
+			/*
+			 * Have to check before we set max_hole_start, otherwise
+			 * we could end up sending back this offset anyway.
+			 */
+			if (contains_pending_extent(trans, device,
+						    &search_start,
+						    hole_size))
+				hole_size = 0;
+
+			if (hole_size > max_hole_size) {
+				max_hole_start = search_start;
+				max_hole_size = hole_size;
+			}
 
-			if (key.offset > last_byte &&
-			    hole_size >= num_bytes) {
-				*start = last_byte;
-				goto check_pending;
+			/*
+			 * If this free space is greater than which we need,
+			 * it must be the max free space that we have found
+			 * until now, so max_hole_start must point to the start
+			 * of this free space and the length of this free space
+			 * is stored in max_hole_size. Thus, we return
+			 * max_hole_start and max_hole_size and go back to the
+			 * caller.
+			 */
+			if (hole_size >= num_bytes) {
+				ret = 0;
+				goto out;
 			}
 		}
-		if (btrfs_key_type(&key) != BTRFS_DEV_EXTENT_KEY)
-			goto next;
 
-		start_found = 1;
 		dev_extent = btrfs_item_ptr(l, slot, struct btrfs_dev_extent);
-		last_byte = key.offset + btrfs_dev_extent_length(l, dev_extent);
+		extent_end = key.offset + btrfs_dev_extent_length(l,
+								  dev_extent);
+		if (extent_end > search_start)
+			search_start = extent_end;
 next:
 		path->slots[0]++;
 		cond_resched();
 	}
-check_pending:
-	/* we have to make sure we didn't find an extent that has already
-	 * been allocated by the map tree or the original allocation
+
+	/*
+	 * At this point, search_start should be the end of
+	 * allocated dev extents, and when shrinking the device,
+	 * search_end may be smaller than search_start.
 	 */
-	BUG_ON(*start < search_start);
+	if (search_end > search_start)
+		hole_size = search_end - search_start;
 
-	if (*start + num_bytes > search_end) {
-		ret = -ENOSPC;
-		goto error;
+	if (hole_size > max_hole_size) {
+		max_hole_start = search_start;
+		max_hole_size = hole_size;
 	}
-	/* check for pending inserts here */
-	ret = 0;
 
-error:
+	if (contains_pending_extent(trans, device, &search_start, hole_size)) {
+		btrfs_release_path(path);
+		goto again;
+	}
+
+	/* See above. */
+	if (hole_size < num_bytes)
+		ret = -ENOSPC;
+	else
+		ret = 0;
+
+out:
 	btrfs_free_path(path);
+	*start = max_hole_start;
+	if (len)
+		*len = max_hole_size;
 	return ret;
 }
 
@@ -874,39 +1255,52 @@ static int btrfs_free_dev_extent(struct btrfs_trans_handle *trans,
 	key.objectid = device->devid;
 	key.offset = start;
 	key.type = BTRFS_DEV_EXTENT_KEY;
-
+again:
 	ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
 	if (ret > 0) {
 		ret = btrfs_previous_item(root, path, key.objectid,
 					  BTRFS_DEV_EXTENT_KEY);
-		BUG_ON(ret);
+		if (ret)
+			goto out;
 		leaf = path->nodes[0];
 		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
 		extent = btrfs_item_ptr(leaf, path->slots[0],
 					struct btrfs_dev_extent);
 		BUG_ON(found_key.offset > start || found_key.offset +
 		       btrfs_dev_extent_length(leaf, extent) < start);
-		ret = 0;
+		key = found_key;
+		btrfs_release_path(path);
+		goto again;
 	} else if (ret == 0) {
 		leaf = path->nodes[0];
 		extent = btrfs_item_ptr(leaf, path->slots[0],
 					struct btrfs_dev_extent);
+	} else {
+		btrfs_error(root->fs_info, ret, "Slot search failed");
+		goto out;
 	}
-	BUG_ON(ret);
 
-	if (device->bytes_used > 0)
-		device->bytes_used -= btrfs_dev_extent_length(leaf, extent);
+	if (device->bytes_used > 0) {
+		u64 len = btrfs_dev_extent_length(leaf, extent);
+		device->bytes_used -= len;
+		spin_lock(&root->fs_info->free_chunk_lock);
+		root->fs_info->free_chunk_space += len;
+		spin_unlock(&root->fs_info->free_chunk_lock);
+	}
 	ret = btrfs_del_item(trans, root, path);
-	BUG_ON(ret);
-
+	if (ret) {
+		btrfs_error(root->fs_info, ret,
+			    "Failed to remove dev extent item");
+	}
+out:
 	btrfs_free_path(path);
 	return ret;
 }
 
-int btrfs_alloc_dev_extent(struct btrfs_trans_handle *trans,
-			   struct btrfs_device *device,
-			   u64 chunk_tree, u64 chunk_objectid,
-			   u64 chunk_offset, u64 start, u64 num_bytes)
+static int btrfs_alloc_dev_extent(struct btrfs_trans_handle *trans,
+				  struct btrfs_device *device,
+				  u64 chunk_tree, u64 chunk_objectid,
+				  u64 chunk_offset, u64 start, u64 num_bytes)
 {
 	int ret;
 	struct btrfs_path *path;
@@ -916,6 +1310,7 @@ int btrfs_alloc_dev_extent(struct btrfs_trans_handle *trans,
 	struct btrfs_key key;
 
 	WARN_ON(!device->in_fs_metadata);
+	WARN_ON(device->is_tgtdev_for_dev_replace);
 	path = btrfs_alloc_path();
 	if (!path)
 		return -ENOMEM;
@@ -925,7 +1320,8 @@ int btrfs_alloc_dev_extent(struct btrfs_trans_handle *trans,
 	key.type = BTRFS_DEV_EXTENT_KEY;
 	ret = btrfs_insert_empty_item(trans, root, path, &key,
 				      sizeof(*extent));
-	BUG_ON(ret);
+	if (ret)
+		goto out;
 
 	leaf = path->nodes[0];
 	extent = btrfs_item_ptr(leaf, path->slots[0],
@@ -935,67 +1331,42 @@ int btrfs_alloc_dev_extent(struct btrfs_trans_handle *trans,
 	btrfs_set_dev_extent_chunk_offset(leaf, extent, chunk_offset);
 
 	write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
-		    (unsigned long)btrfs_dev_extent_chunk_tree_uuid(extent),
-		    BTRFS_UUID_SIZE);
+		    btrfs_dev_extent_chunk_tree_uuid(extent), BTRFS_UUID_SIZE);
 
 	btrfs_set_dev_extent_length(leaf, extent, num_bytes);
 	btrfs_mark_buffer_dirty(leaf);
+out:
 	btrfs_free_path(path);
 	return ret;
 }
 
-static noinline int find_next_chunk(struct btrfs_root *root,
-				    u64 objectid, u64 *offset)
+static u64 find_next_chunk(struct btrfs_fs_info *fs_info)
 {
-	struct btrfs_path *path;
-	int ret;
-	struct btrfs_key key;
-	struct btrfs_chunk *chunk;
-	struct btrfs_key found_key;
-
-	path = btrfs_alloc_path();
-	BUG_ON(!path);
-
-	key.objectid = objectid;
-	key.offset = (u64)-1;
-	key.type = BTRFS_CHUNK_ITEM_KEY;
-
-	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
-	if (ret < 0)
-		goto error;
-
-	BUG_ON(ret == 0);
+	struct extent_map_tree *em_tree;
+	struct extent_map *em;
+	struct rb_node *n;
+	u64 ret = 0;
 
-	ret = btrfs_previous_item(root, path, 0, BTRFS_CHUNK_ITEM_KEY);
-	if (ret) {
-		*offset = 0;
-	} else {
-		btrfs_item_key_to_cpu(path->nodes[0], &found_key,
-				      path->slots[0]);
-		if (found_key.objectid != objectid)
-			*offset = 0;
-		else {
-			chunk = btrfs_item_ptr(path->nodes[0], path->slots[0],
-					       struct btrfs_chunk);
-			*offset = found_key.offset +
-				btrfs_chunk_length(path->nodes[0], chunk);
-		}
+	em_tree = &fs_info->mapping_tree.map_tree;
+	read_lock(&em_tree->lock);
+	n = rb_last(&em_tree->map);
+	if (n) {
+		em = rb_entry(n, struct extent_map, rb_node);
+		ret = em->start + em->len;
 	}
-	ret = 0;
-error:
-	btrfs_free_path(path);
+	read_unlock(&em_tree->lock);
+
 	return ret;
 }
 
-static noinline int find_next_devid(struct btrfs_root *root, u64 *objectid)
+static noinline int find_next_devid(struct btrfs_fs_info *fs_info,
+				    u64 *devid_ret)
 {
 	int ret;
 	struct btrfs_key key;
 	struct btrfs_key found_key;
 	struct btrfs_path *path;
 
-	root = root->fs_info->chunk_root;
-
 	path = btrfs_alloc_path();
 	if (!path)
 		return -ENOMEM;
@@ -1004,20 +1375,21 @@ static noinline int find_next_devid(struct btrfs_root *root, u64 *objectid)
 	key.type = BTRFS_DEV_ITEM_KEY;
 	key.offset = (u64)-1;
 
-	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+	ret = btrfs_search_slot(NULL, fs_info->chunk_root, &key, path, 0, 0);
 	if (ret < 0)
 		goto error;
 
-	BUG_ON(ret == 0);
+	BUG_ON(ret == 0); /* Corruption */
 
-	ret = btrfs_previous_item(root, path, BTRFS_DEV_ITEMS_OBJECTID,
+	ret = btrfs_previous_item(fs_info->chunk_root, path,
+				  BTRFS_DEV_ITEMS_OBJECTID,
 				  BTRFS_DEV_ITEM_KEY);
 	if (ret) {
-		*objectid = 1;
+		*devid_ret = 1;
 	} else {
 		btrfs_item_key_to_cpu(path->nodes[0], &found_key,
 				      path->slots[0]);
-		*objectid = found_key.offset + 1;
+		*devid_ret = found_key.offset + 1;
 	}
 	ret = 0;
 error:
@@ -1029,9 +1401,9 @@ error:
  * the device information is stored in the chunk root
  * the btrfs_device struct should be fully filled in
  */
-int btrfs_add_device(struct btrfs_trans_handle *trans,
-		     struct btrfs_root *root,
-		     struct btrfs_device *device)
+static int btrfs_add_device(struct btrfs_trans_handle *trans,
+			    struct btrfs_root *root,
+			    struct btrfs_device *device)
 {
 	int ret;
 	struct btrfs_path *path;
@@ -1071,9 +1443,9 @@ int btrfs_add_device(struct btrfs_trans_handle *trans,
 	btrfs_set_device_bandwidth(leaf, dev_item, 0);
 	btrfs_set_device_start_offset(leaf, dev_item, 0);
 
-	ptr = (unsigned long)btrfs_device_uuid(dev_item);
+	ptr = btrfs_device_uuid(dev_item);
 	write_extent_buffer(leaf, device->uuid, ptr, BTRFS_UUID_SIZE);
-	ptr = (unsigned long)btrfs_device_fsid(dev_item);
+	ptr = btrfs_device_fsid(dev_item);
 	write_extent_buffer(leaf, root->fs_info->fsid, ptr, BTRFS_UUID_SIZE);
 	btrfs_mark_buffer_dirty(leaf);
 
@@ -1083,6 +1455,22 @@ out:
 	return ret;
 }
 
+/*
+ * Function to update ctime/mtime for a given device path.
+ * Mainly used for ctime/mtime based probe like libblkid.
+ */
+static void update_dev_time(char *path_name)
+{
+	struct file *filp;
+
+	filp = filp_open(path_name, O_RDWR, 0);
+	if (!filp)
+		return;
+	file_update_time(filp);
+	filp_close(filp, NULL);
+	return;
+}
+
 static int btrfs_rm_dev_item(struct btrfs_root *root,
 			     struct btrfs_device *device)
 {
@@ -1097,7 +1485,11 @@ static int btrfs_rm_dev_item(struct btrfs_root *root,
 	if (!path)
 		return -ENOMEM;
 
-	trans = btrfs_start_transaction(root, 1);
+	trans = btrfs_start_transaction(root, 0);
+	if (IS_ERR(trans)) {
+		btrfs_free_path(path);
+		return PTR_ERR(trans);
+	}
 	key.objectid = BTRFS_DEV_ITEMS_OBJECTID;
 	key.type = BTRFS_DEV_ITEM_KEY;
 	key.offset = device->devid;
@@ -1129,32 +1521,51 @@ int btrfs_rm_device(struct btrfs_root *root, char *device_path)
 	struct block_device *bdev;
 	struct buffer_head *bh = NULL;
 	struct btrfs_super_block *disk_super;
+	struct btrfs_fs_devices *cur_devices;
 	u64 all_avail;
 	u64 devid;
 	u64 num_devices;
 	u8 *dev_uuid;
+	unsigned seq;
 	int ret = 0;
+	bool clear_super = false;
 
 	mutex_lock(&uuid_mutex);
-	mutex_lock(&root->fs_info->volume_mutex);
 
-	all_avail = root->fs_info->avail_data_alloc_bits |
-		root->fs_info->avail_system_alloc_bits |
-		root->fs_info->avail_metadata_alloc_bits;
+	do {
+		seq = read_seqbegin(&root->fs_info->profiles_lock);
 
-	if ((all_avail & BTRFS_BLOCK_GROUP_RAID10) &&
-	    root->fs_info->fs_devices->num_devices <= 4) {
-		printk(KERN_ERR "btrfs: unable to go below four devices "
-		       "on raid10\n");
-		ret = -EINVAL;
+		all_avail = root->fs_info->avail_data_alloc_bits |
+			    root->fs_info->avail_system_alloc_bits |
+			    root->fs_info->avail_metadata_alloc_bits;
+	} while (read_seqretry(&root->fs_info->profiles_lock, seq));
+
+	num_devices = root->fs_info->fs_devices->num_devices;
+	btrfs_dev_replace_lock(&root->fs_info->dev_replace);
+	if (btrfs_dev_replace_is_ongoing(&root->fs_info->dev_replace)) {
+		WARN_ON(num_devices < 1);
+		num_devices--;
+	}
+	btrfs_dev_replace_unlock(&root->fs_info->dev_replace);
+
+	if ((all_avail & BTRFS_BLOCK_GROUP_RAID10) && num_devices <= 4) {
+		ret = BTRFS_ERROR_DEV_RAID10_MIN_NOT_MET;
 		goto out;
 	}
 
-	if ((all_avail & BTRFS_BLOCK_GROUP_RAID1) &&
-	    root->fs_info->fs_devices->num_devices <= 2) {
-		printk(KERN_ERR "btrfs: unable to go below two "
-		       "devices on raid1\n");
-		ret = -EINVAL;
+	if ((all_avail & BTRFS_BLOCK_GROUP_RAID1) && num_devices <= 2) {
+		ret = BTRFS_ERROR_DEV_RAID1_MIN_NOT_MET;
+		goto out;
+	}
+
+	if ((all_avail & BTRFS_BLOCK_GROUP_RAID5) &&
+	    root->fs_info->fs_devices->rw_devices <= 2) {
+		ret = BTRFS_ERROR_DEV_RAID5_MIN_NOT_MET;
+		goto out;
+	}
+	if ((all_avail & BTRFS_BLOCK_GROUP_RAID6) &&
+	    root->fs_info->fs_devices->rw_devices <= 3) {
+		ret = BTRFS_ERROR_DEV_RAID6_MIN_NOT_MET;
 		goto out;
 	}
 
@@ -1164,40 +1575,36 @@ int btrfs_rm_device(struct btrfs_root *root, char *device_path)
 
 		device = NULL;
 		devices = &root->fs_info->fs_devices->devices;
-		mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
+		/*
+		 * It is safe to read the devices since the volume_mutex
+		 * is held.
+		 */
 		list_for_each_entry(tmp, devices, dev_list) {
-			if (tmp->in_fs_metadata && !tmp->bdev) {
+			if (tmp->in_fs_metadata &&
+			    !tmp->is_tgtdev_for_dev_replace &&
+			    !tmp->bdev) {
 				device = tmp;
 				break;
 			}
 		}
-		mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
 		bdev = NULL;
 		bh = NULL;
 		disk_super = NULL;
 		if (!device) {
-			printk(KERN_ERR "btrfs: no missing devices found to "
-			       "remove\n");
+			ret = BTRFS_ERROR_DEV_MISSING_NOT_FOUND;
 			goto out;
 		}
 	} else {
-		bdev = open_bdev_exclusive(device_path, FMODE_READ,
-				      root->fs_info->bdev_holder);
-		if (IS_ERR(bdev)) {
-			ret = PTR_ERR(bdev);
+		ret = btrfs_get_bdev_and_sb(device_path,
+					    FMODE_WRITE | FMODE_EXCL,
+					    root->fs_info->bdev_holder, 0,
+					    &bdev, &bh);
+		if (ret)
 			goto out;
-		}
-
-		set_blocksize(bdev, 4096);
-		bh = btrfs_read_dev_super(bdev);
-		if (!bh) {
-			ret = -EIO;
-			goto error_close;
-		}
 		disk_super = (struct btrfs_super_block *)bh->b_data;
 		devid = btrfs_stack_device_id(&disk_super->dev_item);
 		dev_uuid = disk_super->dev_item.uuid;
-		device = btrfs_find_device(root, devid, dev_uuid,
+		device = btrfs_find_device(root->fs_info, devid, dev_uuid,
 					   disk_super->fsid);
 		if (!device) {
 			ret = -ENOENT;
@@ -1205,38 +1612,66 @@ int btrfs_rm_device(struct btrfs_root *root, char *device_path)
 		}
 	}
 
+	if (device->is_tgtdev_for_dev_replace) {
+		ret = BTRFS_ERROR_DEV_TGT_REPLACE;
+		goto error_brelse;
+	}
+
 	if (device->writeable && root->fs_info->fs_devices->rw_devices == 1) {
-		printk(KERN_ERR "btrfs: unable to remove the only writeable "
-		       "device\n");
-		ret = -EINVAL;
+		ret = BTRFS_ERROR_DEV_ONLY_WRITABLE;
 		goto error_brelse;
 	}
 
 	if (device->writeable) {
+		lock_chunks(root);
 		list_del_init(&device->dev_alloc_list);
+		unlock_chunks(root);
 		root->fs_info->fs_devices->rw_devices--;
+		clear_super = true;
 	}
 
+	mutex_unlock(&uuid_mutex);
 	ret = btrfs_shrink_device(device, 0);
+	mutex_lock(&uuid_mutex);
 	if (ret)
-		goto error_brelse;
+		goto error_undo;
 
+	/*
+	 * TODO: the superblock still includes this device in its num_devices
+	 * counter although write_all_supers() is not locked out. This
+	 * could give a filesystem state which requires a degraded mount.
+	 */
 	ret = btrfs_rm_dev_item(root->fs_info->chunk_root, device);
 	if (ret)
-		goto error_brelse;
+		goto error_undo;
+
+	spin_lock(&root->fs_info->free_chunk_lock);
+	root->fs_info->free_chunk_space = device->total_bytes -
+		device->bytes_used;
+	spin_unlock(&root->fs_info->free_chunk_lock);
 
 	device->in_fs_metadata = 0;
+	btrfs_scrub_cancel_dev(root->fs_info, device);
 
 	/*
 	 * the device list mutex makes sure that we don't change
 	 * the device list while someone else is writing out all
-	 * the device supers.
+	 * the device supers. Whoever is writing all supers, should
+	 * lock the device list mutex before getting the number of
+	 * devices in the super block (super_copy). Conversely,
+	 * whoever updates the number of devices in the super block
+	 * (super_copy) should hold the device list mutex.
 	 */
+
+	cur_devices = device->fs_devices;
 	mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
-	list_del_init(&device->dev_list);
-	mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
+	list_del_rcu(&device->dev_list);
 
 	device->fs_devices->num_devices--;
+	device->fs_devices->total_devices--;
+
+	if (device->missing)
+		root->fs_info->fs_devices->missing_devices--;
 
 	next_device = list_entry(root->fs_info->fs_devices->devices.next,
 				 struct btrfs_device, dev_list);
@@ -1246,66 +1681,229 @@ int btrfs_rm_device(struct btrfs_root *root, char *device_path)
 		root->fs_info->fs_devices->latest_bdev = next_device->bdev;
 
 	if (device->bdev) {
-		close_bdev_exclusive(device->bdev, device->mode);
-		device->bdev = NULL;
 		device->fs_devices->open_devices--;
+		/* remove sysfs entry */
+		btrfs_kobj_rm_device(root->fs_info, device);
 	}
 
-	num_devices = btrfs_super_num_devices(&root->fs_info->super_copy) - 1;
-	btrfs_set_super_num_devices(&root->fs_info->super_copy, num_devices);
+	call_rcu(&device->rcu, free_device);
+
+	num_devices = btrfs_super_num_devices(root->fs_info->super_copy) - 1;
+	btrfs_set_super_num_devices(root->fs_info->super_copy, num_devices);
+	mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
 
-	if (device->fs_devices->open_devices == 0) {
+	if (cur_devices->open_devices == 0) {
 		struct btrfs_fs_devices *fs_devices;
 		fs_devices = root->fs_info->fs_devices;
 		while (fs_devices) {
-			if (fs_devices->seed == device->fs_devices)
+			if (fs_devices->seed == cur_devices) {
+				fs_devices->seed = cur_devices->seed;
 				break;
+			}
 			fs_devices = fs_devices->seed;
 		}
-		fs_devices->seed = device->fs_devices->seed;
-		device->fs_devices->seed = NULL;
-		__btrfs_close_devices(device->fs_devices);
-		free_fs_devices(device->fs_devices);
+		cur_devices->seed = NULL;
+		lock_chunks(root);
+		__btrfs_close_devices(cur_devices);
+		unlock_chunks(root);
+		free_fs_devices(cur_devices);
 	}
 
+	root->fs_info->num_tolerated_disk_barrier_failures =
+		btrfs_calc_num_tolerated_disk_barrier_failures(root->fs_info);
+
 	/*
 	 * at this point, the device is zero sized.  We want to
 	 * remove it from the devices list and zero out the old super
 	 */
-	if (device->writeable) {
+	if (clear_super && disk_super) {
+		u64 bytenr;
+		int i;
+
 		/* make sure this device isn't detected as part of
 		 * the FS anymore
 		 */
 		memset(&disk_super->magic, 0, sizeof(disk_super->magic));
 		set_buffer_dirty(bh);
 		sync_dirty_buffer(bh);
+
+		/* clear the mirror copies of super block on the disk
+		 * being removed, 0th copy is been taken care above and
+		 * the below would take of the rest
+		 */
+		for (i = 1; i < BTRFS_SUPER_MIRROR_MAX; i++) {
+			bytenr = btrfs_sb_offset(i);
+			if (bytenr + BTRFS_SUPER_INFO_SIZE >=
+					i_size_read(bdev->bd_inode))
+				break;
+
+			brelse(bh);
+			bh = __bread(bdev, bytenr / 4096,
+					BTRFS_SUPER_INFO_SIZE);
+			if (!bh)
+				continue;
+
+			disk_super = (struct btrfs_super_block *)bh->b_data;
+
+			if (btrfs_super_bytenr(disk_super) != bytenr ||
+				btrfs_super_magic(disk_super) != BTRFS_MAGIC) {
+				continue;
+			}
+			memset(&disk_super->magic, 0,
+						sizeof(disk_super->magic));
+			set_buffer_dirty(bh);
+			sync_dirty_buffer(bh);
+		}
 	}
 
-	kfree(device->name);
-	kfree(device);
 	ret = 0;
 
+	if (bdev) {
+		/* Notify udev that device has changed */
+		btrfs_kobject_uevent(bdev, KOBJ_CHANGE);
+
+		/* Update ctime/mtime for device path for libblkid */
+		update_dev_time(device_path);
+	}
+
 error_brelse:
 	brelse(bh);
-error_close:
 	if (bdev)
-		close_bdev_exclusive(bdev, FMODE_READ);
+		blkdev_put(bdev, FMODE_READ | FMODE_EXCL);
 out:
-	mutex_unlock(&root->fs_info->volume_mutex);
 	mutex_unlock(&uuid_mutex);
 	return ret;
+error_undo:
+	if (device->writeable) {
+		lock_chunks(root);
+		list_add(&device->dev_alloc_list,
+			 &root->fs_info->fs_devices->alloc_list);
+		unlock_chunks(root);
+		root->fs_info->fs_devices->rw_devices++;
+	}
+	goto error_brelse;
+}
+
+void btrfs_rm_dev_replace_srcdev(struct btrfs_fs_info *fs_info,
+				 struct btrfs_device *srcdev)
+{
+	WARN_ON(!mutex_is_locked(&fs_info->fs_devices->device_list_mutex));
+
+	list_del_rcu(&srcdev->dev_list);
+	list_del_rcu(&srcdev->dev_alloc_list);
+	fs_info->fs_devices->num_devices--;
+	if (srcdev->missing) {
+		fs_info->fs_devices->missing_devices--;
+		fs_info->fs_devices->rw_devices++;
+	}
+	if (srcdev->can_discard)
+		fs_info->fs_devices->num_can_discard--;
+	if (srcdev->bdev) {
+		fs_info->fs_devices->open_devices--;
+
+		/* zero out the old super */
+		btrfs_scratch_superblock(srcdev);
+	}
+
+	call_rcu(&srcdev->rcu, free_device);
+}
+
+void btrfs_destroy_dev_replace_tgtdev(struct btrfs_fs_info *fs_info,
+				      struct btrfs_device *tgtdev)
+{
+	struct btrfs_device *next_device;
+
+	WARN_ON(!tgtdev);
+	mutex_lock(&fs_info->fs_devices->device_list_mutex);
+	if (tgtdev->bdev) {
+		btrfs_scratch_superblock(tgtdev);
+		fs_info->fs_devices->open_devices--;
+	}
+	fs_info->fs_devices->num_devices--;
+	if (tgtdev->can_discard)
+		fs_info->fs_devices->num_can_discard++;
+
+	next_device = list_entry(fs_info->fs_devices->devices.next,
+				 struct btrfs_device, dev_list);
+	if (tgtdev->bdev == fs_info->sb->s_bdev)
+		fs_info->sb->s_bdev = next_device->bdev;
+	if (tgtdev->bdev == fs_info->fs_devices->latest_bdev)
+		fs_info->fs_devices->latest_bdev = next_device->bdev;
+	list_del_rcu(&tgtdev->dev_list);
+
+	call_rcu(&tgtdev->rcu, free_device);
+
+	mutex_unlock(&fs_info->fs_devices->device_list_mutex);
+}
+
+static int btrfs_find_device_by_path(struct btrfs_root *root, char *device_path,
+				     struct btrfs_device **device)
+{
+	int ret = 0;
+	struct btrfs_super_block *disk_super;
+	u64 devid;
+	u8 *dev_uuid;
+	struct block_device *bdev;
+	struct buffer_head *bh;
+
+	*device = NULL;
+	ret = btrfs_get_bdev_and_sb(device_path, FMODE_READ,
+				    root->fs_info->bdev_holder, 0, &bdev, &bh);
+	if (ret)
+		return ret;
+	disk_super = (struct btrfs_super_block *)bh->b_data;
+	devid = btrfs_stack_device_id(&disk_super->dev_item);
+	dev_uuid = disk_super->dev_item.uuid;
+	*device = btrfs_find_device(root->fs_info, devid, dev_uuid,
+				    disk_super->fsid);
+	brelse(bh);
+	if (!*device)
+		ret = -ENOENT;
+	blkdev_put(bdev, FMODE_READ);
+	return ret;
+}
+
+int btrfs_find_device_missing_or_by_path(struct btrfs_root *root,
+					 char *device_path,
+					 struct btrfs_device **device)
+{
+	*device = NULL;
+	if (strcmp(device_path, "missing") == 0) {
+		struct list_head *devices;
+		struct btrfs_device *tmp;
+
+		devices = &root->fs_info->fs_devices->devices;
+		/*
+		 * It is safe to read the devices since the volume_mutex
+		 * is held by the caller.
+		 */
+		list_for_each_entry(tmp, devices, dev_list) {
+			if (tmp->in_fs_metadata && !tmp->bdev) {
+				*device = tmp;
+				break;
+			}
+		}
+
+		if (!*device) {
+			btrfs_err(root->fs_info, "no missing device found");
+			return -ENOENT;
+		}
+
+		return 0;
+	} else {
+		return btrfs_find_device_by_path(root, device_path, device);
+	}
 }
 
 /*
  * does all the dirty work required for changing file system's UUID.
  */
-static int btrfs_prepare_sprout(struct btrfs_trans_handle *trans,
-				struct btrfs_root *root)
+static int btrfs_prepare_sprout(struct btrfs_root *root)
 {
 	struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
 	struct btrfs_fs_devices *old_devices;
 	struct btrfs_fs_devices *seed_devices;
-	struct btrfs_super_block *disk_super = &root->fs_info->super_copy;
+	struct btrfs_super_block *disk_super = root->fs_info->super_copy;
 	struct btrfs_device *device;
 	u64 super_flags;
 
@@ -1313,9 +1911,9 @@ static int btrfs_prepare_sprout(struct btrfs_trans_handle *trans,
 	if (!fs_devices->seeding)
 		return -EINVAL;
 
-	seed_devices = kzalloc(sizeof(*fs_devices), GFP_NOFS);
-	if (!seed_devices)
-		return -ENOMEM;
+	seed_devices = __alloc_fs_devices();
+	if (IS_ERR(seed_devices))
+		return PTR_ERR(seed_devices);
 
 	old_devices = clone_fs_devices(fs_devices);
 	if (IS_ERR(old_devices)) {
@@ -1330,7 +1928,11 @@ static int btrfs_prepare_sprout(struct btrfs_trans_handle *trans,
 	INIT_LIST_HEAD(&seed_devices->devices);
 	INIT_LIST_HEAD(&seed_devices->alloc_list);
 	mutex_init(&seed_devices->device_list_mutex);
-	list_splice_init(&fs_devices->devices, &seed_devices->devices);
+
+	mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
+	list_splice_init_rcu(&fs_devices->devices, &seed_devices->devices,
+			      synchronize_rcu);
+
 	list_splice_init(&fs_devices->alloc_list, &seed_devices->alloc_list);
 	list_for_each_entry(device, &seed_devices->devices, dev_list) {
 		device->fs_devices = seed_devices;
@@ -1344,6 +1946,8 @@ static int btrfs_prepare_sprout(struct btrfs_trans_handle *trans,
 	generate_random_uuid(fs_devices->fsid);
 	memcpy(root->fs_info->fsid, fs_devices->fsid, BTRFS_FSID_SIZE);
 	memcpy(disk_super->fsid, fs_devices->fsid, BTRFS_FSID_SIZE);
+	mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
+
 	super_flags = btrfs_super_flags(disk_super) &
 		      ~BTRFS_SUPER_FLAG_SEEDING;
 	btrfs_set_super_flags(disk_super, super_flags);
@@ -1391,7 +1995,7 @@ next_slot:
 				goto error;
 			leaf = path->nodes[0];
 			btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
-			btrfs_release_path(root, path);
+			btrfs_release_path(path);
 			continue;
 		}
 
@@ -1403,14 +2007,13 @@ next_slot:
 		dev_item = btrfs_item_ptr(leaf, path->slots[0],
 					  struct btrfs_dev_item);
 		devid = btrfs_device_id(leaf, dev_item);
-		read_extent_buffer(leaf, dev_uuid,
-				   (unsigned long)btrfs_device_uuid(dev_item),
+		read_extent_buffer(leaf, dev_uuid, btrfs_device_uuid(dev_item),
 				   BTRFS_UUID_SIZE);
-		read_extent_buffer(leaf, fs_uuid,
-				   (unsigned long)btrfs_device_fsid(dev_item),
+		read_extent_buffer(leaf, fs_uuid, btrfs_device_fsid(dev_item),
 				   BTRFS_UUID_SIZE);
-		device = btrfs_find_device(root, devid, dev_uuid, fs_uuid);
-		BUG_ON(!device);
+		device = btrfs_find_device(root->fs_info, devid, dev_uuid,
+					   fs_uuid);
+		BUG_ON(!device); /* Logic error */
 
 		if (device->fs_devices->seeding) {
 			btrfs_set_device_generation(leaf, dev_item,
@@ -1429,19 +2032,22 @@ error:
 
 int btrfs_init_new_device(struct btrfs_root *root, char *device_path)
 {
+	struct request_queue *q;
 	struct btrfs_trans_handle *trans;
 	struct btrfs_device *device;
 	struct block_device *bdev;
 	struct list_head *devices;
 	struct super_block *sb = root->fs_info->sb;
+	struct rcu_string *name;
 	u64 total_bytes;
 	int seeding_dev = 0;
 	int ret = 0;
 
 	if ((sb->s_flags & MS_RDONLY) && !root->fs_info->fs_devices->seeding)
-		return -EINVAL;
+		return -EROFS;
 
-	bdev = open_bdev_exclusive(device_path, 0, root->fs_info->bdev_holder);
+	bdev = blkdev_get_by_path(device_path, FMODE_WRITE | FMODE_EXCL,
+				  root->fs_info->bdev_holder);
 	if (IS_ERR(bdev))
 		return PTR_ERR(bdev);
 
@@ -1452,48 +2058,49 @@ int btrfs_init_new_device(struct btrfs_root *root, char *device_path)
 	}
 
 	filemap_write_and_wait(bdev->bd_inode->i_mapping);
-	mutex_lock(&root->fs_info->volume_mutex);
 
 	devices = &root->fs_info->fs_devices->devices;
-	/*
-	 * we have the volume lock, so we don't need the extra
-	 * device list mutex while reading the list here.
-	 */
+
+	mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
 	list_for_each_entry(device, devices, dev_list) {
 		if (device->bdev == bdev) {
 			ret = -EEXIST;
+			mutex_unlock(
+				&root->fs_info->fs_devices->device_list_mutex);
 			goto error;
 		}
 	}
+	mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
 
-	device = kzalloc(sizeof(*device), GFP_NOFS);
-	if (!device) {
+	device = btrfs_alloc_device(root->fs_info, NULL, NULL);
+	if (IS_ERR(device)) {
 		/* we can safely leave the fs_devices entry around */
-		ret = -ENOMEM;
+		ret = PTR_ERR(device);
 		goto error;
 	}
 
-	device->name = kstrdup(device_path, GFP_NOFS);
-	if (!device->name) {
+	name = rcu_string_strdup(device_path, GFP_NOFS);
+	if (!name) {
 		kfree(device);
 		ret = -ENOMEM;
 		goto error;
 	}
+	rcu_assign_pointer(device->name, name);
 
-	ret = find_next_devid(root, &device->devid);
-	if (ret) {
+	trans = btrfs_start_transaction(root, 0);
+	if (IS_ERR(trans)) {
+		rcu_string_free(device->name);
 		kfree(device);
+		ret = PTR_ERR(trans);
 		goto error;
 	}
 
-	trans = btrfs_start_transaction(root, 1);
 	lock_chunks(root);
 
-	device->barriers = 1;
+	q = bdev_get_queue(bdev);
+	if (blk_queue_discard(q))
+		device->can_discard = 1;
 	device->writeable = 1;
-	device->work.func = pending_bios_fn;
-	generate_random_uuid(device->uuid);
-	spin_lock_init(&device->io_lock);
 	device->generation = trans->transid;
 	device->io_width = root->sectorsize;
 	device->io_align = root->sectorsize;
@@ -1503,49 +2110,77 @@ int btrfs_init_new_device(struct btrfs_root *root, char *device_path)
 	device->dev_root = root->fs_info->dev_root;
 	device->bdev = bdev;
 	device->in_fs_metadata = 1;
-	device->mode = 0;
+	device->is_tgtdev_for_dev_replace = 0;
+	device->mode = FMODE_EXCL;
+	device->dev_stats_valid = 1;
 	set_blocksize(device->bdev, 4096);
 
 	if (seeding_dev) {
 		sb->s_flags &= ~MS_RDONLY;
-		ret = btrfs_prepare_sprout(trans, root);
-		BUG_ON(ret);
+		ret = btrfs_prepare_sprout(root);
+		BUG_ON(ret); /* -ENOMEM */
 	}
 
 	device->fs_devices = root->fs_info->fs_devices;
 
-	/*
-	 * we don't want write_supers to jump in here with our device
-	 * half setup
-	 */
 	mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
-	list_add(&device->dev_list, &root->fs_info->fs_devices->devices);
+	list_add_rcu(&device->dev_list, &root->fs_info->fs_devices->devices);
 	list_add(&device->dev_alloc_list,
 		 &root->fs_info->fs_devices->alloc_list);
 	root->fs_info->fs_devices->num_devices++;
 	root->fs_info->fs_devices->open_devices++;
 	root->fs_info->fs_devices->rw_devices++;
+	root->fs_info->fs_devices->total_devices++;
+	if (device->can_discard)
+		root->fs_info->fs_devices->num_can_discard++;
 	root->fs_info->fs_devices->total_rw_bytes += device->total_bytes;
 
+	spin_lock(&root->fs_info->free_chunk_lock);
+	root->fs_info->free_chunk_space += device->total_bytes;
+	spin_unlock(&root->fs_info->free_chunk_lock);
+
 	if (!blk_queue_nonrot(bdev_get_queue(bdev)))
 		root->fs_info->fs_devices->rotating = 1;
 
-	total_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy);
-	btrfs_set_super_total_bytes(&root->fs_info->super_copy,
+	total_bytes = btrfs_super_total_bytes(root->fs_info->super_copy);
+	btrfs_set_super_total_bytes(root->fs_info->super_copy,
 				    total_bytes + device->total_bytes);
 
-	total_bytes = btrfs_super_num_devices(&root->fs_info->super_copy);
-	btrfs_set_super_num_devices(&root->fs_info->super_copy,
+	total_bytes = btrfs_super_num_devices(root->fs_info->super_copy);
+	btrfs_set_super_num_devices(root->fs_info->super_copy,
 				    total_bytes + 1);
+
+	/* add sysfs device entry */
+	btrfs_kobj_add_device(root->fs_info, device);
+
 	mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
 
 	if (seeding_dev) {
+		char fsid_buf[BTRFS_UUID_UNPARSED_SIZE];
 		ret = init_first_rw_device(trans, root, device);
-		BUG_ON(ret);
+		if (ret) {
+			btrfs_abort_transaction(trans, root, ret);
+			goto error_trans;
+		}
 		ret = btrfs_finish_sprout(trans, root);
-		BUG_ON(ret);
+		if (ret) {
+			btrfs_abort_transaction(trans, root, ret);
+			goto error_trans;
+		}
+
+		/* Sprouting would change fsid of the mounted root,
+		 * so rename the fsid on the sysfs
+		 */
+		snprintf(fsid_buf, BTRFS_UUID_UNPARSED_SIZE, "%pU",
+						root->fs_info->fsid);
+		if (kobject_rename(&root->fs_info->super_kobj, fsid_buf))
+			goto error_trans;
 	} else {
 		ret = btrfs_add_device(trans, root, device);
+		if (ret) {
+			btrfs_abort_transaction(trans, root, ret);
+			goto error_trans;
+		}
 	}
 
 	/*
@@ -1555,25 +2190,139 @@ int btrfs_init_new_device(struct btrfs_root *root, char *device_path)
 	btrfs_clear_space_info_full(root->fs_info);
 
 	unlock_chunks(root);
-	btrfs_commit_transaction(trans, root);
+	root->fs_info->num_tolerated_disk_barrier_failures =
+		btrfs_calc_num_tolerated_disk_barrier_failures(root->fs_info);
+	ret = btrfs_commit_transaction(trans, root);
 
 	if (seeding_dev) {
 		mutex_unlock(&uuid_mutex);
 		up_write(&sb->s_umount);
 
+		if (ret) /* transaction commit */
+			return ret;
+
 		ret = btrfs_relocate_sys_chunks(root);
-		BUG_ON(ret);
+		if (ret < 0)
+			btrfs_error(root->fs_info, ret,
+				    "Failed to relocate sys chunks after "
+				    "device initialization. This can be fixed "
+				    "using the \"btrfs balance\" command.");
+		trans = btrfs_attach_transaction(root);
+		if (IS_ERR(trans)) {
+			if (PTR_ERR(trans) == -ENOENT)
+				return 0;
+			return PTR_ERR(trans);
+		}
+		ret = btrfs_commit_transaction(trans, root);
 	}
-out:
-	mutex_unlock(&root->fs_info->volume_mutex);
+
+	/* Update ctime/mtime for libblkid */
+	update_dev_time(device_path);
 	return ret;
+
+error_trans:
+	unlock_chunks(root);
+	btrfs_end_transaction(trans, root);
+	rcu_string_free(device->name);
+	btrfs_kobj_rm_device(root->fs_info, device);
+	kfree(device);
 error:
-	close_bdev_exclusive(bdev, 0);
+	blkdev_put(bdev, FMODE_EXCL);
 	if (seeding_dev) {
 		mutex_unlock(&uuid_mutex);
 		up_write(&sb->s_umount);
 	}
-	goto out;
+	return ret;
+}
+
+int btrfs_init_dev_replace_tgtdev(struct btrfs_root *root, char *device_path,
+				  struct btrfs_device **device_out)
+{
+	struct request_queue *q;
+	struct btrfs_device *device;
+	struct block_device *bdev;
+	struct btrfs_fs_info *fs_info = root->fs_info;
+	struct list_head *devices;
+	struct rcu_string *name;
+	u64 devid = BTRFS_DEV_REPLACE_DEVID;
+	int ret = 0;
+
+	*device_out = NULL;
+	if (fs_info->fs_devices->seeding)
+		return -EINVAL;
+
+	bdev = blkdev_get_by_path(device_path, FMODE_WRITE | FMODE_EXCL,
+				  fs_info->bdev_holder);
+	if (IS_ERR(bdev))
+		return PTR_ERR(bdev);
+
+	filemap_write_and_wait(bdev->bd_inode->i_mapping);
+
+	devices = &fs_info->fs_devices->devices;
+	list_for_each_entry(device, devices, dev_list) {
+		if (device->bdev == bdev) {
+			ret = -EEXIST;
+			goto error;
+		}
+	}
+
+	device = btrfs_alloc_device(NULL, &devid, NULL);
+	if (IS_ERR(device)) {
+		ret = PTR_ERR(device);
+		goto error;
+	}
+
+	name = rcu_string_strdup(device_path, GFP_NOFS);
+	if (!name) {
+		kfree(device);
+		ret = -ENOMEM;
+		goto error;
+	}
+	rcu_assign_pointer(device->name, name);
+
+	q = bdev_get_queue(bdev);
+	if (blk_queue_discard(q))
+		device->can_discard = 1;
+	mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
+	device->writeable = 1;
+	device->generation = 0;
+	device->io_width = root->sectorsize;
+	device->io_align = root->sectorsize;
+	device->sector_size = root->sectorsize;
+	device->total_bytes = i_size_read(bdev->bd_inode);
+	device->disk_total_bytes = device->total_bytes;
+	device->dev_root = fs_info->dev_root;
+	device->bdev = bdev;
+	device->in_fs_metadata = 1;
+	device->is_tgtdev_for_dev_replace = 1;
+	device->mode = FMODE_EXCL;
+	device->dev_stats_valid = 1;
+	set_blocksize(device->bdev, 4096);
+	device->fs_devices = fs_info->fs_devices;
+	list_add(&device->dev_list, &fs_info->fs_devices->devices);
+	fs_info->fs_devices->num_devices++;
+	fs_info->fs_devices->open_devices++;
+	if (device->can_discard)
+		fs_info->fs_devices->num_can_discard++;
+	mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
+
+	*device_out = device;
+	return ret;
+
+error:
+	blkdev_put(bdev, FMODE_EXCL);
+	return ret;
+}
+
+void btrfs_init_dev_replace_tgtdev_for_resume(struct btrfs_fs_info *fs_info,
+					      struct btrfs_device *tgtdev)
+{
+	WARN_ON(fs_info->fs_devices->rw_devices == 0);
+	tgtdev->io_width = fs_info->dev_root->sectorsize;
+	tgtdev->io_align = fs_info->dev_root->sectorsize;
+	tgtdev->sector_size = fs_info->dev_root->sectorsize;
+	tgtdev->dev_root = fs_info->dev_root;
+	tgtdev->in_fs_metadata = 1;
 }
 
 static noinline int btrfs_update_device(struct btrfs_trans_handle *trans,
@@ -1626,13 +2375,14 @@ static int __btrfs_grow_device(struct btrfs_trans_handle *trans,
 		      struct btrfs_device *device, u64 new_size)
 {
 	struct btrfs_super_block *super_copy =
-		&device->dev_root->fs_info->super_copy;
+		device->dev_root->fs_info->super_copy;
 	u64 old_total = btrfs_super_total_bytes(super_copy);
 	u64 diff = new_size - device->total_bytes;
 
 	if (!device->writeable)
 		return -EACCES;
-	if (new_size <= device->total_bytes)
+	if (new_size <= device->total_bytes ||
+	    device->is_tgtdev_for_dev_replace)
 		return -EINVAL;
 
 	btrfs_set_super_total_bytes(super_copy, old_total + diff);
@@ -1674,19 +2424,28 @@ static int btrfs_free_chunk(struct btrfs_trans_handle *trans,
 	key.type = BTRFS_CHUNK_ITEM_KEY;
 
 	ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
-	BUG_ON(ret);
+	if (ret < 0)
+		goto out;
+	else if (ret > 0) { /* Logic error or corruption */
+		btrfs_error(root->fs_info, -ENOENT,
+			    "Failed lookup while freeing chunk.");
+		ret = -ENOENT;
+		goto out;
+	}
 
 	ret = btrfs_del_item(trans, root, path);
-	BUG_ON(ret);
-
+	if (ret < 0)
+		btrfs_error(root->fs_info, ret,
+			    "Failed to delete chunk item.");
+out:
 	btrfs_free_path(path);
-	return 0;
+	return ret;
 }
 
 static int btrfs_del_sys_chunk(struct btrfs_root *root, u64 chunk_objectid, u64
 			chunk_offset)
 {
-	struct btrfs_super_block *super_copy = &root->fs_info->super_copy;
+	struct btrfs_super_block *super_copy = root->fs_info->super_copy;
 	struct btrfs_disk_key *disk_key;
 	struct btrfs_chunk *chunk;
 	u8 *ptr;
@@ -1751,10 +2510,15 @@ static int btrfs_relocate_chunk(struct btrfs_root *root,
 
 	/* step one, relocate all the extents inside this chunk */
 	ret = btrfs_relocate_block_group(extent_root, chunk_offset);
-	BUG_ON(ret);
+	if (ret)
+		return ret;
 
-	trans = btrfs_start_transaction(root, 1);
-	BUG_ON(!trans);
+	trans = btrfs_start_transaction(root, 0);
+	if (IS_ERR(trans)) {
+		ret = PTR_ERR(trans);
+		btrfs_std_error(root->fs_info, ret);
+		return ret;
+	}
 
 	lock_chunks(root);
 
@@ -1766,7 +2530,7 @@ static int btrfs_relocate_chunk(struct btrfs_root *root,
 	em = lookup_extent_mapping(em_tree, chunk_offset, 1);
 	read_unlock(&em_tree->lock);
 
-	BUG_ON(em->start > chunk_offset ||
+	BUG_ON(!em || em->start > chunk_offset ||
 	       em->start + em->len < chunk_offset);
 	map = (struct map_lookup *)em->bdev;
 
@@ -1785,6 +2549,8 @@ static int btrfs_relocate_chunk(struct btrfs_root *root,
 
 	BUG_ON(ret);
 
+	trace_btrfs_chunk_free(root, map, chunk_offset, em->len);
+
 	if (map->type & BTRFS_BLOCK_GROUP_SYSTEM) {
 		ret = btrfs_del_sys_chunk(root, chunk_objectid, chunk_offset);
 		BUG_ON(ret);
@@ -1797,9 +2563,6 @@ static int btrfs_relocate_chunk(struct btrfs_root *root,
 	remove_extent_mapping(em_tree, em);
 	write_unlock(&em_tree->lock);
 
-	kfree(map);
-	em->bdev = NULL;
-
 	/* once for the tree */
 	free_extent_map(em);
 	/* once for us */
@@ -1837,7 +2600,7 @@ again:
 		ret = btrfs_search_slot(NULL, chunk_root, &key, path, 0, 0);
 		if (ret < 0)
 			goto error;
-		BUG_ON(ret == 0);
+		BUG_ON(ret == 0); /* Corruption */
 
 		ret = btrfs_previous_item(chunk_root, path, key.objectid,
 					  key.type);
@@ -1852,7 +2615,7 @@ again:
 		chunk = btrfs_item_ptr(leaf, path->slots[0],
 				       struct btrfs_chunk);
 		chunk_type = btrfs_chunk_type(leaf, chunk);
-		btrfs_release_path(chunk_root, path);
+		btrfs_release_path(path);
 
 		if (chunk_type & BTRFS_BLOCK_GROUP_SYSTEM) {
 			ret = btrfs_relocate_chunk(chunk_root, chunk_tree,
@@ -1873,8 +2636,7 @@ again:
 		failed = 0;
 		retried = true;
 		goto again;
-	} else if (failed && retried) {
-		WARN_ON(1);
+	} else if (WARN_ON(failed && retried)) {
 		ret = -ENOSPC;
 	}
 error:
@@ -1882,42 +2644,399 @@ error:
 	return ret;
 }
 
-static u64 div_factor(u64 num, int factor)
+static int insert_balance_item(struct btrfs_root *root,
+			       struct btrfs_balance_control *bctl)
 {
-	if (factor == 10)
-		return num;
-	num *= factor;
-	do_div(num, 10);
-	return num;
+	struct btrfs_trans_handle *trans;
+	struct btrfs_balance_item *item;
+	struct btrfs_disk_balance_args disk_bargs;
+	struct btrfs_path *path;
+	struct extent_buffer *leaf;
+	struct btrfs_key key;
+	int ret, err;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	trans = btrfs_start_transaction(root, 0);
+	if (IS_ERR(trans)) {
+		btrfs_free_path(path);
+		return PTR_ERR(trans);
+	}
+
+	key.objectid = BTRFS_BALANCE_OBJECTID;
+	key.type = BTRFS_BALANCE_ITEM_KEY;
+	key.offset = 0;
+
+	ret = btrfs_insert_empty_item(trans, root, path, &key,
+				      sizeof(*item));
+	if (ret)
+		goto out;
+
+	leaf = path->nodes[0];
+	item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_balance_item);
+
+	memset_extent_buffer(leaf, 0, (unsigned long)item, sizeof(*item));
+
+	btrfs_cpu_balance_args_to_disk(&disk_bargs, &bctl->data);
+	btrfs_set_balance_data(leaf, item, &disk_bargs);
+	btrfs_cpu_balance_args_to_disk(&disk_bargs, &bctl->meta);
+	btrfs_set_balance_meta(leaf, item, &disk_bargs);
+	btrfs_cpu_balance_args_to_disk(&disk_bargs, &bctl->sys);
+	btrfs_set_balance_sys(leaf, item, &disk_bargs);
+
+	btrfs_set_balance_flags(leaf, item, bctl->flags);
+
+	btrfs_mark_buffer_dirty(leaf);
+out:
+	btrfs_free_path(path);
+	err = btrfs_commit_transaction(trans, root);
+	if (err && !ret)
+		ret = err;
+	return ret;
 }
 
-int btrfs_balance(struct btrfs_root *dev_root)
+static int del_balance_item(struct btrfs_root *root)
 {
-	int ret;
-	struct list_head *devices = &dev_root->fs_info->fs_devices->devices;
+	struct btrfs_trans_handle *trans;
+	struct btrfs_path *path;
+	struct btrfs_key key;
+	int ret, err;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	trans = btrfs_start_transaction(root, 0);
+	if (IS_ERR(trans)) {
+		btrfs_free_path(path);
+		return PTR_ERR(trans);
+	}
+
+	key.objectid = BTRFS_BALANCE_OBJECTID;
+	key.type = BTRFS_BALANCE_ITEM_KEY;
+	key.offset = 0;
+
+	ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
+	if (ret < 0)
+		goto out;
+	if (ret > 0) {
+		ret = -ENOENT;
+		goto out;
+	}
+
+	ret = btrfs_del_item(trans, root, path);
+out:
+	btrfs_free_path(path);
+	err = btrfs_commit_transaction(trans, root);
+	if (err && !ret)
+		ret = err;
+	return ret;
+}
+
+/*
+ * This is a heuristic used to reduce the number of chunks balanced on
+ * resume after balance was interrupted.
+ */
+static void update_balance_args(struct btrfs_balance_control *bctl)
+{
+	/*
+	 * Turn on soft mode for chunk types that were being converted.
+	 */
+	if (bctl->data.flags & BTRFS_BALANCE_ARGS_CONVERT)
+		bctl->data.flags |= BTRFS_BALANCE_ARGS_SOFT;
+	if (bctl->sys.flags & BTRFS_BALANCE_ARGS_CONVERT)
+		bctl->sys.flags |= BTRFS_BALANCE_ARGS_SOFT;
+	if (bctl->meta.flags & BTRFS_BALANCE_ARGS_CONVERT)
+		bctl->meta.flags |= BTRFS_BALANCE_ARGS_SOFT;
+
+	/*
+	 * Turn on usage filter if is not already used.  The idea is
+	 * that chunks that we have already balanced should be
+	 * reasonably full.  Don't do it for chunks that are being
+	 * converted - that will keep us from relocating unconverted
+	 * (albeit full) chunks.
+	 */
+	if (!(bctl->data.flags & BTRFS_BALANCE_ARGS_USAGE) &&
+	    !(bctl->data.flags & BTRFS_BALANCE_ARGS_CONVERT)) {
+		bctl->data.flags |= BTRFS_BALANCE_ARGS_USAGE;
+		bctl->data.usage = 90;
+	}
+	if (!(bctl->sys.flags & BTRFS_BALANCE_ARGS_USAGE) &&
+	    !(bctl->sys.flags & BTRFS_BALANCE_ARGS_CONVERT)) {
+		bctl->sys.flags |= BTRFS_BALANCE_ARGS_USAGE;
+		bctl->sys.usage = 90;
+	}
+	if (!(bctl->meta.flags & BTRFS_BALANCE_ARGS_USAGE) &&
+	    !(bctl->meta.flags & BTRFS_BALANCE_ARGS_CONVERT)) {
+		bctl->meta.flags |= BTRFS_BALANCE_ARGS_USAGE;
+		bctl->meta.usage = 90;
+	}
+}
+
+/*
+ * Should be called with both balance and volume mutexes held to
+ * serialize other volume operations (add_dev/rm_dev/resize) with
+ * restriper.  Same goes for unset_balance_control.
+ */
+static void set_balance_control(struct btrfs_balance_control *bctl)
+{
+	struct btrfs_fs_info *fs_info = bctl->fs_info;
+
+	BUG_ON(fs_info->balance_ctl);
+
+	spin_lock(&fs_info->balance_lock);
+	fs_info->balance_ctl = bctl;
+	spin_unlock(&fs_info->balance_lock);
+}
+
+static void unset_balance_control(struct btrfs_fs_info *fs_info)
+{
+	struct btrfs_balance_control *bctl = fs_info->balance_ctl;
+
+	BUG_ON(!fs_info->balance_ctl);
+
+	spin_lock(&fs_info->balance_lock);
+	fs_info->balance_ctl = NULL;
+	spin_unlock(&fs_info->balance_lock);
+
+	kfree(bctl);
+}
+
+/*
+ * Balance filters.  Return 1 if chunk should be filtered out
+ * (should not be balanced).
+ */
+static int chunk_profiles_filter(u64 chunk_type,
+				 struct btrfs_balance_args *bargs)
+{
+	chunk_type = chunk_to_extended(chunk_type) &
+				BTRFS_EXTENDED_PROFILE_MASK;
+
+	if (bargs->profiles & chunk_type)
+		return 0;
+
+	return 1;
+}
+
+static int chunk_usage_filter(struct btrfs_fs_info *fs_info, u64 chunk_offset,
+			      struct btrfs_balance_args *bargs)
+{
+	struct btrfs_block_group_cache *cache;
+	u64 chunk_used, user_thresh;
+	int ret = 1;
+
+	cache = btrfs_lookup_block_group(fs_info, chunk_offset);
+	chunk_used = btrfs_block_group_used(&cache->item);
+
+	if (bargs->usage == 0)
+		user_thresh = 1;
+	else if (bargs->usage > 100)
+		user_thresh = cache->key.offset;
+	else
+		user_thresh = div_factor_fine(cache->key.offset,
+					      bargs->usage);
+
+	if (chunk_used < user_thresh)
+		ret = 0;
+
+	btrfs_put_block_group(cache);
+	return ret;
+}
+
+static int chunk_devid_filter(struct extent_buffer *leaf,
+			      struct btrfs_chunk *chunk,
+			      struct btrfs_balance_args *bargs)
+{
+	struct btrfs_stripe *stripe;
+	int num_stripes = btrfs_chunk_num_stripes(leaf, chunk);
+	int i;
+
+	for (i = 0; i < num_stripes; i++) {
+		stripe = btrfs_stripe_nr(chunk, i);
+		if (btrfs_stripe_devid(leaf, stripe) == bargs->devid)
+			return 0;
+	}
+
+	return 1;
+}
+
+/* [pstart, pend) */
+static int chunk_drange_filter(struct extent_buffer *leaf,
+			       struct btrfs_chunk *chunk,
+			       u64 chunk_offset,
+			       struct btrfs_balance_args *bargs)
+{
+	struct btrfs_stripe *stripe;
+	int num_stripes = btrfs_chunk_num_stripes(leaf, chunk);
+	u64 stripe_offset;
+	u64 stripe_length;
+	int factor;
+	int i;
+
+	if (!(bargs->flags & BTRFS_BALANCE_ARGS_DEVID))
+		return 0;
+
+	if (btrfs_chunk_type(leaf, chunk) & (BTRFS_BLOCK_GROUP_DUP |
+	     BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID10)) {
+		factor = num_stripes / 2;
+	} else if (btrfs_chunk_type(leaf, chunk) & BTRFS_BLOCK_GROUP_RAID5) {
+		factor = num_stripes - 1;
+	} else if (btrfs_chunk_type(leaf, chunk) & BTRFS_BLOCK_GROUP_RAID6) {
+		factor = num_stripes - 2;
+	} else {
+		factor = num_stripes;
+	}
+
+	for (i = 0; i < num_stripes; i++) {
+		stripe = btrfs_stripe_nr(chunk, i);
+		if (btrfs_stripe_devid(leaf, stripe) != bargs->devid)
+			continue;
+
+		stripe_offset = btrfs_stripe_offset(leaf, stripe);
+		stripe_length = btrfs_chunk_length(leaf, chunk);
+		do_div(stripe_length, factor);
+
+		if (stripe_offset < bargs->pend &&
+		    stripe_offset + stripe_length > bargs->pstart)
+			return 0;
+	}
+
+	return 1;
+}
+
+/* [vstart, vend) */
+static int chunk_vrange_filter(struct extent_buffer *leaf,
+			       struct btrfs_chunk *chunk,
+			       u64 chunk_offset,
+			       struct btrfs_balance_args *bargs)
+{
+	if (chunk_offset < bargs->vend &&
+	    chunk_offset + btrfs_chunk_length(leaf, chunk) > bargs->vstart)
+		/* at least part of the chunk is inside this vrange */
+		return 0;
+
+	return 1;
+}
+
+static int chunk_soft_convert_filter(u64 chunk_type,
+				     struct btrfs_balance_args *bargs)
+{
+	if (!(bargs->flags & BTRFS_BALANCE_ARGS_CONVERT))
+		return 0;
+
+	chunk_type = chunk_to_extended(chunk_type) &
+				BTRFS_EXTENDED_PROFILE_MASK;
+
+	if (bargs->target == chunk_type)
+		return 1;
+
+	return 0;
+}
+
+static int should_balance_chunk(struct btrfs_root *root,
+				struct extent_buffer *leaf,
+				struct btrfs_chunk *chunk, u64 chunk_offset)
+{
+	struct btrfs_balance_control *bctl = root->fs_info->balance_ctl;
+	struct btrfs_balance_args *bargs = NULL;
+	u64 chunk_type = btrfs_chunk_type(leaf, chunk);
+
+	/* type filter */
+	if (!((chunk_type & BTRFS_BLOCK_GROUP_TYPE_MASK) &
+	      (bctl->flags & BTRFS_BALANCE_TYPE_MASK))) {
+		return 0;
+	}
+
+	if (chunk_type & BTRFS_BLOCK_GROUP_DATA)
+		bargs = &bctl->data;
+	else if (chunk_type & BTRFS_BLOCK_GROUP_SYSTEM)
+		bargs = &bctl->sys;
+	else if (chunk_type & BTRFS_BLOCK_GROUP_METADATA)
+		bargs = &bctl->meta;
+
+	/* profiles filter */
+	if ((bargs->flags & BTRFS_BALANCE_ARGS_PROFILES) &&
+	    chunk_profiles_filter(chunk_type, bargs)) {
+		return 0;
+	}
+
+	/* usage filter */
+	if ((bargs->flags & BTRFS_BALANCE_ARGS_USAGE) &&
+	    chunk_usage_filter(bctl->fs_info, chunk_offset, bargs)) {
+		return 0;
+	}
+
+	/* devid filter */
+	if ((bargs->flags & BTRFS_BALANCE_ARGS_DEVID) &&
+	    chunk_devid_filter(leaf, chunk, bargs)) {
+		return 0;
+	}
+
+	/* drange filter, makes sense only with devid filter */
+	if ((bargs->flags & BTRFS_BALANCE_ARGS_DRANGE) &&
+	    chunk_drange_filter(leaf, chunk, chunk_offset, bargs)) {
+		return 0;
+	}
+
+	/* vrange filter */
+	if ((bargs->flags & BTRFS_BALANCE_ARGS_VRANGE) &&
+	    chunk_vrange_filter(leaf, chunk, chunk_offset, bargs)) {
+		return 0;
+	}
+
+	/* soft profile changing mode */
+	if ((bargs->flags & BTRFS_BALANCE_ARGS_SOFT) &&
+	    chunk_soft_convert_filter(chunk_type, bargs)) {
+		return 0;
+	}
+
+	/*
+	 * limited by count, must be the last filter
+	 */
+	if ((bargs->flags & BTRFS_BALANCE_ARGS_LIMIT)) {
+		if (bargs->limit == 0)
+			return 0;
+		else
+			bargs->limit--;
+	}
+
+	return 1;
+}
+
+static int __btrfs_balance(struct btrfs_fs_info *fs_info)
+{
+	struct btrfs_balance_control *bctl = fs_info->balance_ctl;
+	struct btrfs_root *chunk_root = fs_info->chunk_root;
+	struct btrfs_root *dev_root = fs_info->dev_root;
+	struct list_head *devices;
 	struct btrfs_device *device;
 	u64 old_size;
 	u64 size_to_free;
+	struct btrfs_chunk *chunk;
 	struct btrfs_path *path;
 	struct btrfs_key key;
-	struct btrfs_chunk *chunk;
-	struct btrfs_root *chunk_root = dev_root->fs_info->chunk_root;
-	struct btrfs_trans_handle *trans;
 	struct btrfs_key found_key;
-
-	if (dev_root->fs_info->sb->s_flags & MS_RDONLY)
-		return -EROFS;
-
-	mutex_lock(&dev_root->fs_info->volume_mutex);
-	dev_root = dev_root->fs_info->dev_root;
+	struct btrfs_trans_handle *trans;
+	struct extent_buffer *leaf;
+	int slot;
+	int ret;
+	int enospc_errors = 0;
+	bool counting = true;
+	u64 limit_data = bctl->data.limit;
+	u64 limit_meta = bctl->meta.limit;
+	u64 limit_sys = bctl->sys.limit;
 
 	/* step one make some room on all the devices */
+	devices = &fs_info->fs_devices->devices;
 	list_for_each_entry(device, devices, dev_list) {
 		old_size = device->total_bytes;
 		size_to_free = div_factor(old_size, 1);
 		size_to_free = min(size_to_free, (u64)1 * 1024 * 1024);
 		if (!device->writeable ||
-		    device->total_bytes - device->bytes_used > size_to_free)
+		    device->total_bytes - device->bytes_used > size_to_free ||
+		    device->is_tgtdev_for_dev_replace)
 			continue;
 
 		ret = btrfs_shrink_device(device, old_size - size_to_free);
@@ -1925,8 +3044,8 @@ int btrfs_balance(struct btrfs_root *dev_root)
 			break;
 		BUG_ON(ret);
 
-		trans = btrfs_start_transaction(dev_root, 1);
-		BUG_ON(!trans);
+		trans = btrfs_start_transaction(dev_root, 0);
+		BUG_ON(IS_ERR(trans));
 
 		ret = btrfs_grow_device(trans, device, old_size);
 		BUG_ON(ret);
@@ -1936,13 +3055,32 @@ int btrfs_balance(struct btrfs_root *dev_root)
 
 	/* step two, relocate all the chunks */
 	path = btrfs_alloc_path();
-	BUG_ON(!path);
+	if (!path) {
+		ret = -ENOMEM;
+		goto error;
+	}
 
+	/* zero out stat counters */
+	spin_lock(&fs_info->balance_lock);
+	memset(&bctl->stat, 0, sizeof(bctl->stat));
+	spin_unlock(&fs_info->balance_lock);
+again:
+	if (!counting) {
+		bctl->data.limit = limit_data;
+		bctl->meta.limit = limit_meta;
+		bctl->sys.limit = limit_sys;
+	}
 	key.objectid = BTRFS_FIRST_CHUNK_TREE_OBJECTID;
 	key.offset = (u64)-1;
 	key.type = BTRFS_CHUNK_ITEM_KEY;
 
 	while (1) {
+		if ((!counting && atomic_read(&fs_info->balance_pause_req)) ||
+		    atomic_read(&fs_info->balance_cancel_req)) {
+			ret = -ECANCELED;
+			goto error;
+		}
+
 		ret = btrfs_search_slot(NULL, chunk_root, &key, path, 0, 0);
 		if (ret < 0)
 			goto error;
@@ -1952,40 +3090,736 @@ int btrfs_balance(struct btrfs_root *dev_root)
 		 * failed
 		 */
 		if (ret == 0)
-			break;
+			BUG(); /* FIXME break ? */
 
 		ret = btrfs_previous_item(chunk_root, path, 0,
 					  BTRFS_CHUNK_ITEM_KEY);
-		if (ret)
+		if (ret) {
+			ret = 0;
 			break;
+		}
+
+		leaf = path->nodes[0];
+		slot = path->slots[0];
+		btrfs_item_key_to_cpu(leaf, &found_key, slot);
 
-		btrfs_item_key_to_cpu(path->nodes[0], &found_key,
-				      path->slots[0]);
 		if (found_key.objectid != key.objectid)
 			break;
 
-		chunk = btrfs_item_ptr(path->nodes[0],
-				       path->slots[0],
-				       struct btrfs_chunk);
-		/* chunk zero is special */
-		if (found_key.offset == 0)
-			break;
+		chunk = btrfs_item_ptr(leaf, slot, struct btrfs_chunk);
+
+		if (!counting) {
+			spin_lock(&fs_info->balance_lock);
+			bctl->stat.considered++;
+			spin_unlock(&fs_info->balance_lock);
+		}
+
+		ret = should_balance_chunk(chunk_root, leaf, chunk,
+					   found_key.offset);
+		btrfs_release_path(path);
+		if (!ret)
+			goto loop;
+
+		if (counting) {
+			spin_lock(&fs_info->balance_lock);
+			bctl->stat.expected++;
+			spin_unlock(&fs_info->balance_lock);
+			goto loop;
+		}
 
-		btrfs_release_path(chunk_root, path);
 		ret = btrfs_relocate_chunk(chunk_root,
 					   chunk_root->root_key.objectid,
 					   found_key.objectid,
 					   found_key.offset);
-		BUG_ON(ret && ret != -ENOSPC);
+		if (ret && ret != -ENOSPC)
+			goto error;
+		if (ret == -ENOSPC) {
+			enospc_errors++;
+		} else {
+			spin_lock(&fs_info->balance_lock);
+			bctl->stat.completed++;
+			spin_unlock(&fs_info->balance_lock);
+		}
+loop:
+		if (found_key.offset == 0)
+			break;
 		key.offset = found_key.offset - 1;
 	}
-	ret = 0;
+
+	if (counting) {
+		btrfs_release_path(path);
+		counting = false;
+		goto again;
+	}
 error:
 	btrfs_free_path(path);
-	mutex_unlock(&dev_root->fs_info->volume_mutex);
+	if (enospc_errors) {
+		btrfs_info(fs_info, "%d enospc errors during balance",
+		       enospc_errors);
+		if (!ret)
+			ret = -ENOSPC;
+	}
+
 	return ret;
 }
 
+/**
+ * alloc_profile_is_valid - see if a given profile is valid and reduced
+ * @flags: profile to validate
+ * @extended: if true @flags is treated as an extended profile
+ */
+static int alloc_profile_is_valid(u64 flags, int extended)
+{
+	u64 mask = (extended ? BTRFS_EXTENDED_PROFILE_MASK :
+			       BTRFS_BLOCK_GROUP_PROFILE_MASK);
+
+	flags &= ~BTRFS_BLOCK_GROUP_TYPE_MASK;
+
+	/* 1) check that all other bits are zeroed */
+	if (flags & ~mask)
+		return 0;
+
+	/* 2) see if profile is reduced */
+	if (flags == 0)
+		return !extended; /* "0" is valid for usual profiles */
+
+	/* true if exactly one bit set */
+	return (flags & (flags - 1)) == 0;
+}
+
+static inline int balance_need_close(struct btrfs_fs_info *fs_info)
+{
+	/* cancel requested || normal exit path */
+	return atomic_read(&fs_info->balance_cancel_req) ||
+		(atomic_read(&fs_info->balance_pause_req) == 0 &&
+		 atomic_read(&fs_info->balance_cancel_req) == 0);
+}
+
+static void __cancel_balance(struct btrfs_fs_info *fs_info)
+{
+	int ret;
+
+	unset_balance_control(fs_info);
+	ret = del_balance_item(fs_info->tree_root);
+	if (ret)
+		btrfs_std_error(fs_info, ret);
+
+	atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
+}
+
+/*
+ * Should be called with both balance and volume mutexes held
+ */
+int btrfs_balance(struct btrfs_balance_control *bctl,
+		  struct btrfs_ioctl_balance_args *bargs)
+{
+	struct btrfs_fs_info *fs_info = bctl->fs_info;
+	u64 allowed;
+	int mixed = 0;
+	int ret;
+	u64 num_devices;
+	unsigned seq;
+
+	if (btrfs_fs_closing(fs_info) ||
+	    atomic_read(&fs_info->balance_pause_req) ||
+	    atomic_read(&fs_info->balance_cancel_req)) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	allowed = btrfs_super_incompat_flags(fs_info->super_copy);
+	if (allowed & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS)
+		mixed = 1;
+
+	/*
+	 * In case of mixed groups both data and meta should be picked,
+	 * and identical options should be given for both of them.
+	 */
+	allowed = BTRFS_BALANCE_DATA | BTRFS_BALANCE_METADATA;
+	if (mixed && (bctl->flags & allowed)) {
+		if (!(bctl->flags & BTRFS_BALANCE_DATA) ||
+		    !(bctl->flags & BTRFS_BALANCE_METADATA) ||
+		    memcmp(&bctl->data, &bctl->meta, sizeof(bctl->data))) {
+			btrfs_err(fs_info, "with mixed groups data and "
+				   "metadata balance options must be the same");
+			ret = -EINVAL;
+			goto out;
+		}
+	}
+
+	num_devices = fs_info->fs_devices->num_devices;
+	btrfs_dev_replace_lock(&fs_info->dev_replace);
+	if (btrfs_dev_replace_is_ongoing(&fs_info->dev_replace)) {
+		BUG_ON(num_devices < 1);
+		num_devices--;
+	}
+	btrfs_dev_replace_unlock(&fs_info->dev_replace);
+	allowed = BTRFS_AVAIL_ALLOC_BIT_SINGLE;
+	if (num_devices == 1)
+		allowed |= BTRFS_BLOCK_GROUP_DUP;
+	else if (num_devices > 1)
+		allowed |= (BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1);
+	if (num_devices > 2)
+		allowed |= BTRFS_BLOCK_GROUP_RAID5;
+	if (num_devices > 3)
+		allowed |= (BTRFS_BLOCK_GROUP_RAID10 |
+			    BTRFS_BLOCK_GROUP_RAID6);
+	if ((bctl->data.flags & BTRFS_BALANCE_ARGS_CONVERT) &&
+	    (!alloc_profile_is_valid(bctl->data.target, 1) ||
+	     (bctl->data.target & ~allowed))) {
+		btrfs_err(fs_info, "unable to start balance with target "
+			   "data profile %llu",
+		       bctl->data.target);
+		ret = -EINVAL;
+		goto out;
+	}
+	if ((bctl->meta.flags & BTRFS_BALANCE_ARGS_CONVERT) &&
+	    (!alloc_profile_is_valid(bctl->meta.target, 1) ||
+	     (bctl->meta.target & ~allowed))) {
+		btrfs_err(fs_info,
+			   "unable to start balance with target metadata profile %llu",
+		       bctl->meta.target);
+		ret = -EINVAL;
+		goto out;
+	}
+	if ((bctl->sys.flags & BTRFS_BALANCE_ARGS_CONVERT) &&
+	    (!alloc_profile_is_valid(bctl->sys.target, 1) ||
+	     (bctl->sys.target & ~allowed))) {
+		btrfs_err(fs_info,
+			   "unable to start balance with target system profile %llu",
+		       bctl->sys.target);
+		ret = -EINVAL;
+		goto out;
+	}
+
+	/* allow dup'ed data chunks only in mixed mode */
+	if (!mixed && (bctl->data.flags & BTRFS_BALANCE_ARGS_CONVERT) &&
+	    (bctl->data.target & BTRFS_BLOCK_GROUP_DUP)) {
+		btrfs_err(fs_info, "dup for data is not allowed");
+		ret = -EINVAL;
+		goto out;
+	}
+
+	/* allow to reduce meta or sys integrity only if force set */
+	allowed = BTRFS_BLOCK_GROUP_DUP | BTRFS_BLOCK_GROUP_RAID1 |
+			BTRFS_BLOCK_GROUP_RAID10 |
+			BTRFS_BLOCK_GROUP_RAID5 |
+			BTRFS_BLOCK_GROUP_RAID6;
+	do {
+		seq = read_seqbegin(&fs_info->profiles_lock);
+
+		if (((bctl->sys.flags & BTRFS_BALANCE_ARGS_CONVERT) &&
+		     (fs_info->avail_system_alloc_bits & allowed) &&
+		     !(bctl->sys.target & allowed)) ||
+		    ((bctl->meta.flags & BTRFS_BALANCE_ARGS_CONVERT) &&
+		     (fs_info->avail_metadata_alloc_bits & allowed) &&
+		     !(bctl->meta.target & allowed))) {
+			if (bctl->flags & BTRFS_BALANCE_FORCE) {
+				btrfs_info(fs_info, "force reducing metadata integrity");
+			} else {
+				btrfs_err(fs_info, "balance will reduce metadata "
+					   "integrity, use force if you want this");
+				ret = -EINVAL;
+				goto out;
+			}
+		}
+	} while (read_seqretry(&fs_info->profiles_lock, seq));
+
+	if (bctl->sys.flags & BTRFS_BALANCE_ARGS_CONVERT) {
+		int num_tolerated_disk_barrier_failures;
+		u64 target = bctl->sys.target;
+
+		num_tolerated_disk_barrier_failures =
+			btrfs_calc_num_tolerated_disk_barrier_failures(fs_info);
+		if (num_tolerated_disk_barrier_failures > 0 &&
+		    (target &
+		     (BTRFS_BLOCK_GROUP_DUP | BTRFS_BLOCK_GROUP_RAID0 |
+		      BTRFS_AVAIL_ALLOC_BIT_SINGLE)))
+			num_tolerated_disk_barrier_failures = 0;
+		else if (num_tolerated_disk_barrier_failures > 1 &&
+			 (target &
+			  (BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID10)))
+			num_tolerated_disk_barrier_failures = 1;
+
+		fs_info->num_tolerated_disk_barrier_failures =
+			num_tolerated_disk_barrier_failures;
+	}
+
+	ret = insert_balance_item(fs_info->tree_root, bctl);
+	if (ret && ret != -EEXIST)
+		goto out;
+
+	if (!(bctl->flags & BTRFS_BALANCE_RESUME)) {
+		BUG_ON(ret == -EEXIST);
+		set_balance_control(bctl);
+	} else {
+		BUG_ON(ret != -EEXIST);
+		spin_lock(&fs_info->balance_lock);
+		update_balance_args(bctl);
+		spin_unlock(&fs_info->balance_lock);
+	}
+
+	atomic_inc(&fs_info->balance_running);
+	mutex_unlock(&fs_info->balance_mutex);
+
+	ret = __btrfs_balance(fs_info);
+
+	mutex_lock(&fs_info->balance_mutex);
+	atomic_dec(&fs_info->balance_running);
+
+	if (bctl->sys.flags & BTRFS_BALANCE_ARGS_CONVERT) {
+		fs_info->num_tolerated_disk_barrier_failures =
+			btrfs_calc_num_tolerated_disk_barrier_failures(fs_info);
+	}
+
+	if (bargs) {
+		memset(bargs, 0, sizeof(*bargs));
+		update_ioctl_balance_args(fs_info, 0, bargs);
+	}
+
+	if ((ret && ret != -ECANCELED && ret != -ENOSPC) ||
+	    balance_need_close(fs_info)) {
+		__cancel_balance(fs_info);
+	}
+
+	wake_up(&fs_info->balance_wait_q);
+
+	return ret;
+out:
+	if (bctl->flags & BTRFS_BALANCE_RESUME)
+		__cancel_balance(fs_info);
+	else {
+		kfree(bctl);
+		atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
+	}
+	return ret;
+}
+
+static int balance_kthread(void *data)
+{
+	struct btrfs_fs_info *fs_info = data;
+	int ret = 0;
+
+	mutex_lock(&fs_info->volume_mutex);
+	mutex_lock(&fs_info->balance_mutex);
+
+	if (fs_info->balance_ctl) {
+		btrfs_info(fs_info, "continuing balance");
+		ret = btrfs_balance(fs_info->balance_ctl, NULL);
+	}
+
+	mutex_unlock(&fs_info->balance_mutex);
+	mutex_unlock(&fs_info->volume_mutex);
+
+	return ret;
+}
+
+int btrfs_resume_balance_async(struct btrfs_fs_info *fs_info)
+{
+	struct task_struct *tsk;
+
+	spin_lock(&fs_info->balance_lock);
+	if (!fs_info->balance_ctl) {
+		spin_unlock(&fs_info->balance_lock);
+		return 0;
+	}
+	spin_unlock(&fs_info->balance_lock);
+
+	if (btrfs_test_opt(fs_info->tree_root, SKIP_BALANCE)) {
+		btrfs_info(fs_info, "force skipping balance");
+		return 0;
+	}
+
+	tsk = kthread_run(balance_kthread, fs_info, "btrfs-balance");
+	return PTR_ERR_OR_ZERO(tsk);
+}
+
+int btrfs_recover_balance(struct btrfs_fs_info *fs_info)
+{
+	struct btrfs_balance_control *bctl;
+	struct btrfs_balance_item *item;
+	struct btrfs_disk_balance_args disk_bargs;
+	struct btrfs_path *path;
+	struct extent_buffer *leaf;
+	struct btrfs_key key;
+	int ret;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	key.objectid = BTRFS_BALANCE_OBJECTID;
+	key.type = BTRFS_BALANCE_ITEM_KEY;
+	key.offset = 0;
+
+	ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
+	if (ret < 0)
+		goto out;
+	if (ret > 0) { /* ret = -ENOENT; */
+		ret = 0;
+		goto out;
+	}
+
+	bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
+	if (!bctl) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	leaf = path->nodes[0];
+	item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_balance_item);
+
+	bctl->fs_info = fs_info;
+	bctl->flags = btrfs_balance_flags(leaf, item);
+	bctl->flags |= BTRFS_BALANCE_RESUME;
+
+	btrfs_balance_data(leaf, item, &disk_bargs);
+	btrfs_disk_balance_args_to_cpu(&bctl->data, &disk_bargs);
+	btrfs_balance_meta(leaf, item, &disk_bargs);
+	btrfs_disk_balance_args_to_cpu(&bctl->meta, &disk_bargs);
+	btrfs_balance_sys(leaf, item, &disk_bargs);
+	btrfs_disk_balance_args_to_cpu(&bctl->sys, &disk_bargs);
+
+	WARN_ON(atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1));
+
+	mutex_lock(&fs_info->volume_mutex);
+	mutex_lock(&fs_info->balance_mutex);
+
+	set_balance_control(bctl);
+
+	mutex_unlock(&fs_info->balance_mutex);
+	mutex_unlock(&fs_info->volume_mutex);
+out:
+	btrfs_free_path(path);
+	return ret;
+}
+
+int btrfs_pause_balance(struct btrfs_fs_info *fs_info)
+{
+	int ret = 0;
+
+	mutex_lock(&fs_info->balance_mutex);
+	if (!fs_info->balance_ctl) {
+		mutex_unlock(&fs_info->balance_mutex);
+		return -ENOTCONN;
+	}
+
+	if (atomic_read(&fs_info->balance_running)) {
+		atomic_inc(&fs_info->balance_pause_req);
+		mutex_unlock(&fs_info->balance_mutex);
+
+		wait_event(fs_info->balance_wait_q,
+			   atomic_read(&fs_info->balance_running) == 0);
+
+		mutex_lock(&fs_info->balance_mutex);
+		/* we are good with balance_ctl ripped off from under us */
+		BUG_ON(atomic_read(&fs_info->balance_running));
+		atomic_dec(&fs_info->balance_pause_req);
+	} else {
+		ret = -ENOTCONN;
+	}
+
+	mutex_unlock(&fs_info->balance_mutex);
+	return ret;
+}
+
+int btrfs_cancel_balance(struct btrfs_fs_info *fs_info)
+{
+	if (fs_info->sb->s_flags & MS_RDONLY)
+		return -EROFS;
+
+	mutex_lock(&fs_info->balance_mutex);
+	if (!fs_info->balance_ctl) {
+		mutex_unlock(&fs_info->balance_mutex);
+		return -ENOTCONN;
+	}
+
+	atomic_inc(&fs_info->balance_cancel_req);
+	/*
+	 * if we are running just wait and return, balance item is
+	 * deleted in btrfs_balance in this case
+	 */
+	if (atomic_read(&fs_info->balance_running)) {
+		mutex_unlock(&fs_info->balance_mutex);
+		wait_event(fs_info->balance_wait_q,
+			   atomic_read(&fs_info->balance_running) == 0);
+		mutex_lock(&fs_info->balance_mutex);
+	} else {
+		/* __cancel_balance needs volume_mutex */
+		mutex_unlock(&fs_info->balance_mutex);
+		mutex_lock(&fs_info->volume_mutex);
+		mutex_lock(&fs_info->balance_mutex);
+
+		if (fs_info->balance_ctl)
+			__cancel_balance(fs_info);
+
+		mutex_unlock(&fs_info->volume_mutex);
+	}
+
+	BUG_ON(fs_info->balance_ctl || atomic_read(&fs_info->balance_running));
+	atomic_dec(&fs_info->balance_cancel_req);
+	mutex_unlock(&fs_info->balance_mutex);
+	return 0;
+}
+
+static int btrfs_uuid_scan_kthread(void *data)
+{
+	struct btrfs_fs_info *fs_info = data;
+	struct btrfs_root *root = fs_info->tree_root;
+	struct btrfs_key key;
+	struct btrfs_key max_key;
+	struct btrfs_path *path = NULL;
+	int ret = 0;
+	struct extent_buffer *eb;
+	int slot;
+	struct btrfs_root_item root_item;
+	u32 item_size;
+	struct btrfs_trans_handle *trans = NULL;
+
+	path = btrfs_alloc_path();
+	if (!path) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	key.objectid = 0;
+	key.type = BTRFS_ROOT_ITEM_KEY;
+	key.offset = 0;
+
+	max_key.objectid = (u64)-1;
+	max_key.type = BTRFS_ROOT_ITEM_KEY;
+	max_key.offset = (u64)-1;
+
+	path->keep_locks = 1;
+
+	while (1) {
+		ret = btrfs_search_forward(root, &key, path, 0);
+		if (ret) {
+			if (ret > 0)
+				ret = 0;
+			break;
+		}
+
+		if (key.type != BTRFS_ROOT_ITEM_KEY ||
+		    (key.objectid < BTRFS_FIRST_FREE_OBJECTID &&
+		     key.objectid != BTRFS_FS_TREE_OBJECTID) ||
+		    key.objectid > BTRFS_LAST_FREE_OBJECTID)
+			goto skip;
+
+		eb = path->nodes[0];
+		slot = path->slots[0];
+		item_size = btrfs_item_size_nr(eb, slot);
+		if (item_size < sizeof(root_item))
+			goto skip;
+
+		read_extent_buffer(eb, &root_item,
+				   btrfs_item_ptr_offset(eb, slot),
+				   (int)sizeof(root_item));
+		if (btrfs_root_refs(&root_item) == 0)
+			goto skip;
+
+		if (!btrfs_is_empty_uuid(root_item.uuid) ||
+		    !btrfs_is_empty_uuid(root_item.received_uuid)) {
+			if (trans)
+				goto update_tree;
+
+			btrfs_release_path(path);
+			/*
+			 * 1 - subvol uuid item
+			 * 1 - received_subvol uuid item
+			 */
+			trans = btrfs_start_transaction(fs_info->uuid_root, 2);
+			if (IS_ERR(trans)) {
+				ret = PTR_ERR(trans);
+				break;
+			}
+			continue;
+		} else {
+			goto skip;
+		}
+update_tree:
+		if (!btrfs_is_empty_uuid(root_item.uuid)) {
+			ret = btrfs_uuid_tree_add(trans, fs_info->uuid_root,
+						  root_item.uuid,
+						  BTRFS_UUID_KEY_SUBVOL,
+						  key.objectid);
+			if (ret < 0) {
+				btrfs_warn(fs_info, "uuid_tree_add failed %d",
+					ret);
+				break;
+			}
+		}
+
+		if (!btrfs_is_empty_uuid(root_item.received_uuid)) {
+			ret = btrfs_uuid_tree_add(trans, fs_info->uuid_root,
+						  root_item.received_uuid,
+						 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
+						  key.objectid);
+			if (ret < 0) {
+				btrfs_warn(fs_info, "uuid_tree_add failed %d",
+					ret);
+				break;
+			}
+		}
+
+skip:
+		if (trans) {
+			ret = btrfs_end_transaction(trans, fs_info->uuid_root);
+			trans = NULL;
+			if (ret)
+				break;
+		}
+
+		btrfs_release_path(path);
+		if (key.offset < (u64)-1) {
+			key.offset++;
+		} else if (key.type < BTRFS_ROOT_ITEM_KEY) {
+			key.offset = 0;
+			key.type = BTRFS_ROOT_ITEM_KEY;
+		} else if (key.objectid < (u64)-1) {
+			key.offset = 0;
+			key.type = BTRFS_ROOT_ITEM_KEY;
+			key.objectid++;
+		} else {
+			break;
+		}
+		cond_resched();
+	}
+
+out:
+	btrfs_free_path(path);
+	if (trans && !IS_ERR(trans))
+		btrfs_end_transaction(trans, fs_info->uuid_root);
+	if (ret)
+		btrfs_warn(fs_info, "btrfs_uuid_scan_kthread failed %d", ret);
+	else
+		fs_info->update_uuid_tree_gen = 1;
+	up(&fs_info->uuid_tree_rescan_sem);
+	return 0;
+}
+
+/*
+ * Callback for btrfs_uuid_tree_iterate().
+ * returns:
+ * 0	check succeeded, the entry is not outdated.
+ * < 0	if an error occured.
+ * > 0	if the check failed, which means the caller shall remove the entry.
+ */
+static int btrfs_check_uuid_tree_entry(struct btrfs_fs_info *fs_info,
+				       u8 *uuid, u8 type, u64 subid)
+{
+	struct btrfs_key key;
+	int ret = 0;
+	struct btrfs_root *subvol_root;
+
+	if (type != BTRFS_UUID_KEY_SUBVOL &&
+	    type != BTRFS_UUID_KEY_RECEIVED_SUBVOL)
+		goto out;
+
+	key.objectid = subid;
+	key.type = BTRFS_ROOT_ITEM_KEY;
+	key.offset = (u64)-1;
+	subvol_root = btrfs_read_fs_root_no_name(fs_info, &key);
+	if (IS_ERR(subvol_root)) {
+		ret = PTR_ERR(subvol_root);
+		if (ret == -ENOENT)
+			ret = 1;
+		goto out;
+	}
+
+	switch (type) {
+	case BTRFS_UUID_KEY_SUBVOL:
+		if (memcmp(uuid, subvol_root->root_item.uuid, BTRFS_UUID_SIZE))
+			ret = 1;
+		break;
+	case BTRFS_UUID_KEY_RECEIVED_SUBVOL:
+		if (memcmp(uuid, subvol_root->root_item.received_uuid,
+			   BTRFS_UUID_SIZE))
+			ret = 1;
+		break;
+	}
+
+out:
+	return ret;
+}
+
+static int btrfs_uuid_rescan_kthread(void *data)
+{
+	struct btrfs_fs_info *fs_info = (struct btrfs_fs_info *)data;
+	int ret;
+
+	/*
+	 * 1st step is to iterate through the existing UUID tree and
+	 * to delete all entries that contain outdated data.
+	 * 2nd step is to add all missing entries to the UUID tree.
+	 */
+	ret = btrfs_uuid_tree_iterate(fs_info, btrfs_check_uuid_tree_entry);
+	if (ret < 0) {
+		btrfs_warn(fs_info, "iterating uuid_tree failed %d", ret);
+		up(&fs_info->uuid_tree_rescan_sem);
+		return ret;
+	}
+	return btrfs_uuid_scan_kthread(data);
+}
+
+int btrfs_create_uuid_tree(struct btrfs_fs_info *fs_info)
+{
+	struct btrfs_trans_handle *trans;
+	struct btrfs_root *tree_root = fs_info->tree_root;
+	struct btrfs_root *uuid_root;
+	struct task_struct *task;
+	int ret;
+
+	/*
+	 * 1 - root node
+	 * 1 - root item
+	 */
+	trans = btrfs_start_transaction(tree_root, 2);
+	if (IS_ERR(trans))
+		return PTR_ERR(trans);
+
+	uuid_root = btrfs_create_tree(trans, fs_info,
+				      BTRFS_UUID_TREE_OBJECTID);
+	if (IS_ERR(uuid_root)) {
+		btrfs_abort_transaction(trans, tree_root,
+					PTR_ERR(uuid_root));
+		return PTR_ERR(uuid_root);
+	}
+
+	fs_info->uuid_root = uuid_root;
+
+	ret = btrfs_commit_transaction(trans, tree_root);
+	if (ret)
+		return ret;
+
+	down(&fs_info->uuid_tree_rescan_sem);
+	task = kthread_run(btrfs_uuid_scan_kthread, fs_info, "btrfs-uuid");
+	if (IS_ERR(task)) {
+		/* fs_info->update_uuid_tree_gen remains 0 in all error case */
+		btrfs_warn(fs_info, "failed to start uuid_scan task");
+		up(&fs_info->uuid_tree_rescan_sem);
+		return PTR_ERR(task);
+	}
+
+	return 0;
+}
+
+int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info)
+{
+	struct task_struct *task;
+
+	down(&fs_info->uuid_tree_rescan_sem);
+	task = kthread_run(btrfs_uuid_rescan_kthread, fs_info, "btrfs-uuid");
+	if (IS_ERR(task)) {
+		/* fs_info->update_uuid_tree_gen remains 0 in all error case */
+		btrfs_warn(fs_info, "failed to start uuid_rescan task");
+		up(&fs_info->uuid_tree_rescan_sem);
+		return PTR_ERR(task);
+	}
+
+	return 0;
+}
+
 /*
  * shrinking a device means finding all of the device extents past
  * the new size, and then following the back refs to the chunks.
@@ -2007,12 +3841,12 @@ int btrfs_shrink_device(struct btrfs_device *device, u64 new_size)
 	bool retried = false;
 	struct extent_buffer *l;
 	struct btrfs_key key;
-	struct btrfs_super_block *super_copy = &root->fs_info->super_copy;
+	struct btrfs_super_block *super_copy = root->fs_info->super_copy;
 	u64 old_total = btrfs_super_total_bytes(super_copy);
 	u64 old_size = device->total_bytes;
 	u64 diff = device->total_bytes - new_size;
 
-	if (new_size >= device->total_bytes)
+	if (device->is_tgtdev_for_dev_replace)
 		return -EINVAL;
 
 	path = btrfs_alloc_path();
@@ -2024,8 +3858,12 @@ int btrfs_shrink_device(struct btrfs_device *device, u64 new_size)
 	lock_chunks(root);
 
 	device->total_bytes = new_size;
-	if (device->writeable)
+	if (device->writeable) {
 		device->fs_devices->total_rw_bytes -= diff;
+		spin_lock(&root->fs_info->free_chunk_lock);
+		root->fs_info->free_chunk_space -= diff;
+		spin_unlock(&root->fs_info->free_chunk_lock);
+	}
 	unlock_chunks(root);
 
 again:
@@ -2033,7 +3871,7 @@ again:
 	key.offset = (u64)-1;
 	key.type = BTRFS_DEV_EXTENT_KEY;
 
-	while (1) {
+	do {
 		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
 		if (ret < 0)
 			goto done;
@@ -2043,7 +3881,7 @@ again:
 			goto done;
 		if (ret) {
 			ret = 0;
-			btrfs_release_path(root, path);
+			btrfs_release_path(path);
 			break;
 		}
 
@@ -2052,7 +3890,7 @@ again:
 		btrfs_item_key_to_cpu(l, &key, path->slots[0]);
 
 		if (key.objectid != device->devid) {
-			btrfs_release_path(root, path);
+			btrfs_release_path(path);
 			break;
 		}
 
@@ -2060,14 +3898,14 @@ again:
 		length = btrfs_dev_extent_length(l, dev_extent);
 
 		if (key.offset + length <= new_size) {
-			btrfs_release_path(root, path);
+			btrfs_release_path(path);
 			break;
 		}
 
 		chunk_tree = btrfs_dev_extent_chunk_tree(l, dev_extent);
 		chunk_objectid = btrfs_dev_extent_chunk_objectid(l, dev_extent);
 		chunk_offset = btrfs_dev_extent_chunk_offset(l, dev_extent);
-		btrfs_release_path(root, path);
+		btrfs_release_path(path);
 
 		ret = btrfs_relocate_chunk(root, chunk_tree, chunk_objectid,
 					   chunk_offset);
@@ -2075,8 +3913,7 @@ again:
 			goto done;
 		if (ret == -ENOSPC)
 			failed++;
-		key.offset -= 1;
-	}
+	} while (key.offset-- > 0);
 
 	if (failed && !retried) {
 		failed = 0;
@@ -2089,16 +3926,20 @@ again:
 		device->total_bytes = old_size;
 		if (device->writeable)
 			device->fs_devices->total_rw_bytes += diff;
+		spin_lock(&root->fs_info->free_chunk_lock);
+		root->fs_info->free_chunk_space += diff;
+		spin_unlock(&root->fs_info->free_chunk_lock);
 		unlock_chunks(root);
 		goto done;
 	}
 
 	/* Shrinking succeeded, else we would be at "done". */
-	trans = btrfs_start_transaction(root, 1);
-	if (!trans) {
-		ret = -ENOMEM;
+	trans = btrfs_start_transaction(root, 0);
+	if (IS_ERR(trans)) {
+		ret = PTR_ERR(trans);
 		goto done;
 	}
+
 	lock_chunks(root);
 
 	device->disk_total_bytes = new_size;
@@ -2118,18 +3959,18 @@ done:
 	return ret;
 }
 
-static int btrfs_add_system_chunk(struct btrfs_trans_handle *trans,
-			   struct btrfs_root *root,
+static int btrfs_add_system_chunk(struct btrfs_root *root,
 			   struct btrfs_key *key,
 			   struct btrfs_chunk *chunk, int item_size)
 {
-	struct btrfs_super_block *super_copy = &root->fs_info->super_copy;
+	struct btrfs_super_block *super_copy = root->fs_info->super_copy;
 	struct btrfs_disk_key disk_key;
 	u32 array_size;
 	u8 *ptr;
 
 	array_size = btrfs_super_sys_array_size(super_copy);
-	if (array_size + item_size > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE)
+	if (array_size + item_size + sizeof(disk_key)
+			> BTRFS_SYSTEM_CHUNK_ARRAY_SIZE)
 		return -EFBIG;
 
 	ptr = super_copy->sys_chunk_array + array_size;
@@ -2142,275 +3983,470 @@ static int btrfs_add_system_chunk(struct btrfs_trans_handle *trans,
 	return 0;
 }
 
-static noinline u64 chunk_bytes_by_type(u64 type, u64 calc_size,
-					int num_stripes, int sub_stripes)
+/*
+ * sort the devices in descending order by max_avail, total_avail
+ */
+static int btrfs_cmp_device_info(const void *a, const void *b)
 {
-	if (type & (BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_DUP))
-		return calc_size;
-	else if (type & BTRFS_BLOCK_GROUP_RAID10)
-		return calc_size * (num_stripes / sub_stripes);
-	else
-		return calc_size * num_stripes;
+	const struct btrfs_device_info *di_a = a;
+	const struct btrfs_device_info *di_b = b;
+
+	if (di_a->max_avail > di_b->max_avail)
+		return -1;
+	if (di_a->max_avail < di_b->max_avail)
+		return 1;
+	if (di_a->total_avail > di_b->total_avail)
+		return -1;
+	if (di_a->total_avail < di_b->total_avail)
+		return 1;
+	return 0;
+}
+
+static struct btrfs_raid_attr btrfs_raid_array[BTRFS_NR_RAID_TYPES] = {
+	[BTRFS_RAID_RAID10] = {
+		.sub_stripes	= 2,
+		.dev_stripes	= 1,
+		.devs_max	= 0,	/* 0 == as many as possible */
+		.devs_min	= 4,
+		.devs_increment	= 2,
+		.ncopies	= 2,
+	},
+	[BTRFS_RAID_RAID1] = {
+		.sub_stripes	= 1,
+		.dev_stripes	= 1,
+		.devs_max	= 2,
+		.devs_min	= 2,
+		.devs_increment	= 2,
+		.ncopies	= 2,
+	},
+	[BTRFS_RAID_DUP] = {
+		.sub_stripes	= 1,
+		.dev_stripes	= 2,
+		.devs_max	= 1,
+		.devs_min	= 1,
+		.devs_increment	= 1,
+		.ncopies	= 2,
+	},
+	[BTRFS_RAID_RAID0] = {
+		.sub_stripes	= 1,
+		.dev_stripes	= 1,
+		.devs_max	= 0,
+		.devs_min	= 2,
+		.devs_increment	= 1,
+		.ncopies	= 1,
+	},
+	[BTRFS_RAID_SINGLE] = {
+		.sub_stripes	= 1,
+		.dev_stripes	= 1,
+		.devs_max	= 1,
+		.devs_min	= 1,
+		.devs_increment	= 1,
+		.ncopies	= 1,
+	},
+	[BTRFS_RAID_RAID5] = {
+		.sub_stripes	= 1,
+		.dev_stripes	= 1,
+		.devs_max	= 0,
+		.devs_min	= 2,
+		.devs_increment	= 1,
+		.ncopies	= 2,
+	},
+	[BTRFS_RAID_RAID6] = {
+		.sub_stripes	= 1,
+		.dev_stripes	= 1,
+		.devs_max	= 0,
+		.devs_min	= 3,
+		.devs_increment	= 1,
+		.ncopies	= 3,
+	},
+};
+
+static u32 find_raid56_stripe_len(u32 data_devices, u32 dev_stripe_target)
+{
+	/* TODO allow them to set a preferred stripe size */
+	return 64 * 1024;
+}
+
+static void check_raid56_incompat_flag(struct btrfs_fs_info *info, u64 type)
+{
+	if (!(type & (BTRFS_BLOCK_GROUP_RAID5 | BTRFS_BLOCK_GROUP_RAID6)))
+		return;
+
+	btrfs_set_fs_incompat(info, RAID56);
 }
 
+#define BTRFS_MAX_DEVS(r) ((BTRFS_LEAF_DATA_SIZE(r)		\
+			- sizeof(struct btrfs_item)		\
+			- sizeof(struct btrfs_chunk))		\
+			/ sizeof(struct btrfs_stripe) + 1)
+
+#define BTRFS_MAX_DEVS_SYS_CHUNK ((BTRFS_SYSTEM_CHUNK_ARRAY_SIZE	\
+				- 2 * sizeof(struct btrfs_disk_key)	\
+				- 2 * sizeof(struct btrfs_chunk))	\
+				/ sizeof(struct btrfs_stripe) + 1)
+
 static int __btrfs_alloc_chunk(struct btrfs_trans_handle *trans,
-			       struct btrfs_root *extent_root,
-			       struct map_lookup **map_ret,
-			       u64 *num_bytes, u64 *stripe_size,
-			       u64 start, u64 type)
+			       struct btrfs_root *extent_root, u64 start,
+			       u64 type)
 {
 	struct btrfs_fs_info *info = extent_root->fs_info;
-	struct btrfs_device *device = NULL;
 	struct btrfs_fs_devices *fs_devices = info->fs_devices;
 	struct list_head *cur;
 	struct map_lookup *map = NULL;
 	struct extent_map_tree *em_tree;
 	struct extent_map *em;
-	struct list_head private_devs;
-	int min_stripe_size = 1 * 1024 * 1024;
-	u64 calc_size = 1024 * 1024 * 1024;
-	u64 max_chunk_size = calc_size;
-	u64 min_free;
-	u64 avail;
-	u64 max_avail = 0;
-	u64 dev_offset;
-	int num_stripes = 1;
-	int min_stripes = 1;
-	int sub_stripes = 0;
-	int looped = 0;
+	struct btrfs_device_info *devices_info = NULL;
+	u64 total_avail;
+	int num_stripes;	/* total number of stripes to allocate */
+	int data_stripes;	/* number of stripes that count for
+				   block group size */
+	int sub_stripes;	/* sub_stripes info for map */
+	int dev_stripes;	/* stripes per dev */
+	int devs_max;		/* max devs to use */
+	int devs_min;		/* min devs needed */
+	int devs_increment;	/* ndevs has to be a multiple of this */
+	int ncopies;		/* how many copies to data has */
 	int ret;
+	u64 max_stripe_size;
+	u64 max_chunk_size;
+	u64 stripe_size;
+	u64 num_bytes;
+	u64 raid_stripe_len = BTRFS_STRIPE_LEN;
+	int ndevs;
+	int i;
+	int j;
 	int index;
-	int stripe_len = 64 * 1024;
 
-	if ((type & BTRFS_BLOCK_GROUP_RAID1) &&
-	    (type & BTRFS_BLOCK_GROUP_DUP)) {
-		WARN_ON(1);
-		type &= ~BTRFS_BLOCK_GROUP_DUP;
-	}
+	BUG_ON(!alloc_profile_is_valid(type, 0));
+
 	if (list_empty(&fs_devices->alloc_list))
 		return -ENOSPC;
 
-	if (type & (BTRFS_BLOCK_GROUP_RAID0)) {
-		num_stripes = fs_devices->rw_devices;
-		min_stripes = 2;
-	}
-	if (type & (BTRFS_BLOCK_GROUP_DUP)) {
-		num_stripes = 2;
-		min_stripes = 2;
-	}
-	if (type & (BTRFS_BLOCK_GROUP_RAID1)) {
-		if (fs_devices->rw_devices < 2)
-			return -ENOSPC;
-		num_stripes = 2;
-		min_stripes = 2;
-	}
-	if (type & (BTRFS_BLOCK_GROUP_RAID10)) {
-		num_stripes = fs_devices->rw_devices;
-		if (num_stripes < 4)
-			return -ENOSPC;
-		num_stripes &= ~(u32)1;
-		sub_stripes = 2;
-		min_stripes = 4;
-	}
+	index = __get_raid_index(type);
+
+	sub_stripes = btrfs_raid_array[index].sub_stripes;
+	dev_stripes = btrfs_raid_array[index].dev_stripes;
+	devs_max = btrfs_raid_array[index].devs_max;
+	devs_min = btrfs_raid_array[index].devs_min;
+	devs_increment = btrfs_raid_array[index].devs_increment;
+	ncopies = btrfs_raid_array[index].ncopies;
 
 	if (type & BTRFS_BLOCK_GROUP_DATA) {
-		max_chunk_size = 10 * calc_size;
-		min_stripe_size = 64 * 1024 * 1024;
+		max_stripe_size = 1024 * 1024 * 1024;
+		max_chunk_size = 10 * max_stripe_size;
+		if (!devs_max)
+			devs_max = BTRFS_MAX_DEVS(info->chunk_root);
 	} else if (type & BTRFS_BLOCK_GROUP_METADATA) {
-		max_chunk_size = 256 * 1024 * 1024;
-		min_stripe_size = 32 * 1024 * 1024;
+		/* for larger filesystems, use larger metadata chunks */
+		if (fs_devices->total_rw_bytes > 50ULL * 1024 * 1024 * 1024)
+			max_stripe_size = 1024 * 1024 * 1024;
+		else
+			max_stripe_size = 256 * 1024 * 1024;
+		max_chunk_size = max_stripe_size;
+		if (!devs_max)
+			devs_max = BTRFS_MAX_DEVS(info->chunk_root);
 	} else if (type & BTRFS_BLOCK_GROUP_SYSTEM) {
-		calc_size = 8 * 1024 * 1024;
-		max_chunk_size = calc_size * 2;
-		min_stripe_size = 1 * 1024 * 1024;
+		max_stripe_size = 32 * 1024 * 1024;
+		max_chunk_size = 2 * max_stripe_size;
+		if (!devs_max)
+			devs_max = BTRFS_MAX_DEVS_SYS_CHUNK;
+	} else {
+		btrfs_err(info, "invalid chunk type 0x%llx requested",
+		       type);
+		BUG_ON(1);
 	}
 
 	/* we don't want a chunk larger than 10% of writeable space */
 	max_chunk_size = min(div_factor(fs_devices->total_rw_bytes, 1),
 			     max_chunk_size);
 
-again:
-	max_avail = 0;
-	if (!map || map->num_stripes != num_stripes) {
-		kfree(map);
-		map = kmalloc(map_lookup_size(num_stripes), GFP_NOFS);
-		if (!map)
-			return -ENOMEM;
-		map->num_stripes = num_stripes;
-	}
-
-	if (calc_size * num_stripes > max_chunk_size) {
-		calc_size = max_chunk_size;
-		do_div(calc_size, num_stripes);
-		do_div(calc_size, stripe_len);
-		calc_size *= stripe_len;
-	}
-
-	/* we don't want tiny stripes */
-	if (!looped)
-		calc_size = max_t(u64, min_stripe_size, calc_size);
-
-	do_div(calc_size, stripe_len);
-	calc_size *= stripe_len;
+	devices_info = kzalloc(sizeof(*devices_info) * fs_devices->rw_devices,
+			       GFP_NOFS);
+	if (!devices_info)
+		return -ENOMEM;
 
 	cur = fs_devices->alloc_list.next;
-	index = 0;
-
-	if (type & BTRFS_BLOCK_GROUP_DUP)
-		min_free = calc_size * 2;
-	else
-		min_free = calc_size;
 
 	/*
-	 * we add 1MB because we never use the first 1MB of the device, unless
-	 * we've looped, then we are likely allocating the maximum amount of
-	 * space left already
+	 * in the first pass through the devices list, we gather information
+	 * about the available holes on each device.
 	 */
-	if (!looped)
-		min_free += 1024 * 1024;
+	ndevs = 0;
+	while (cur != &fs_devices->alloc_list) {
+		struct btrfs_device *device;
+		u64 max_avail;
+		u64 dev_offset;
 
-	INIT_LIST_HEAD(&private_devs);
-	while (index < num_stripes) {
 		device = list_entry(cur, struct btrfs_device, dev_alloc_list);
-		BUG_ON(!device->writeable);
+
+		cur = cur->next;
+
+		if (!device->writeable) {
+			WARN(1, KERN_ERR
+			       "BTRFS: read-only device in alloc_list\n");
+			continue;
+		}
+
+		if (!device->in_fs_metadata ||
+		    device->is_tgtdev_for_dev_replace)
+			continue;
+
 		if (device->total_bytes > device->bytes_used)
-			avail = device->total_bytes - device->bytes_used;
+			total_avail = device->total_bytes - device->bytes_used;
 		else
-			avail = 0;
-		cur = cur->next;
+			total_avail = 0;
 
-		if (device->in_fs_metadata && avail >= min_free) {
-			ret = find_free_dev_extent(trans, device,
-						   min_free, &dev_offset,
-						   &max_avail);
-			if (ret == 0) {
-				list_move_tail(&device->dev_alloc_list,
-					       &private_devs);
-				map->stripes[index].dev = device;
-				map->stripes[index].physical = dev_offset;
-				index++;
-				if (type & BTRFS_BLOCK_GROUP_DUP) {
-					map->stripes[index].dev = device;
-					map->stripes[index].physical =
-						dev_offset + calc_size;
-					index++;
-				}
-			}
-		} else if (device->in_fs_metadata && avail > max_avail)
-			max_avail = avail;
-		if (cur == &fs_devices->alloc_list)
+		/* If there is no space on this device, skip it. */
+		if (total_avail == 0)
+			continue;
+
+		ret = find_free_dev_extent(trans, device,
+					   max_stripe_size * dev_stripes,
+					   &dev_offset, &max_avail);
+		if (ret && ret != -ENOSPC)
+			goto error;
+
+		if (ret == 0)
+			max_avail = max_stripe_size * dev_stripes;
+
+		if (max_avail < BTRFS_STRIPE_LEN * dev_stripes)
+			continue;
+
+		if (ndevs == fs_devices->rw_devices) {
+			WARN(1, "%s: found more than %llu devices\n",
+			     __func__, fs_devices->rw_devices);
 			break;
-	}
-	list_splice(&private_devs, &fs_devices->alloc_list);
-	if (index < num_stripes) {
-		if (index >= min_stripes) {
-			num_stripes = index;
-			if (type & (BTRFS_BLOCK_GROUP_RAID10)) {
-				num_stripes /= sub_stripes;
-				num_stripes *= sub_stripes;
-			}
-			looped = 1;
-			goto again;
 		}
-		if (!looped && max_avail > 0) {
-			looped = 1;
-			calc_size = max_avail;
-			goto again;
+		devices_info[ndevs].dev_offset = dev_offset;
+		devices_info[ndevs].max_avail = max_avail;
+		devices_info[ndevs].total_avail = total_avail;
+		devices_info[ndevs].dev = device;
+		++ndevs;
+	}
+
+	/*
+	 * now sort the devices by hole size / available space
+	 */
+	sort(devices_info, ndevs, sizeof(struct btrfs_device_info),
+	     btrfs_cmp_device_info, NULL);
+
+	/* round down to number of usable stripes */
+	ndevs -= ndevs % devs_increment;
+
+	if (ndevs < devs_increment * sub_stripes || ndevs < devs_min) {
+		ret = -ENOSPC;
+		goto error;
+	}
+
+	if (devs_max && ndevs > devs_max)
+		ndevs = devs_max;
+	/*
+	 * the primary goal is to maximize the number of stripes, so use as many
+	 * devices as possible, even if the stripes are not maximum sized.
+	 */
+	stripe_size = devices_info[ndevs-1].max_avail;
+	num_stripes = ndevs * dev_stripes;
+
+	/*
+	 * this will have to be fixed for RAID1 and RAID10 over
+	 * more drives
+	 */
+	data_stripes = num_stripes / ncopies;
+
+	if (type & BTRFS_BLOCK_GROUP_RAID5) {
+		raid_stripe_len = find_raid56_stripe_len(ndevs - 1,
+				 btrfs_super_stripesize(info->super_copy));
+		data_stripes = num_stripes - 1;
+	}
+	if (type & BTRFS_BLOCK_GROUP_RAID6) {
+		raid_stripe_len = find_raid56_stripe_len(ndevs - 2,
+				 btrfs_super_stripesize(info->super_copy));
+		data_stripes = num_stripes - 2;
+	}
+
+	/*
+	 * Use the number of data stripes to figure out how big this chunk
+	 * is really going to be in terms of logical address space,
+	 * and compare that answer with the max chunk size
+	 */
+	if (stripe_size * data_stripes > max_chunk_size) {
+		u64 mask = (1ULL << 24) - 1;
+		stripe_size = max_chunk_size;
+		do_div(stripe_size, data_stripes);
+
+		/* bump the answer up to a 16MB boundary */
+		stripe_size = (stripe_size + mask) & ~mask;
+
+		/* but don't go higher than the limits we found
+		 * while searching for free extents
+		 */
+		if (stripe_size > devices_info[ndevs-1].max_avail)
+			stripe_size = devices_info[ndevs-1].max_avail;
+	}
+
+	do_div(stripe_size, dev_stripes);
+
+	/* align to BTRFS_STRIPE_LEN */
+	do_div(stripe_size, raid_stripe_len);
+	stripe_size *= raid_stripe_len;
+
+	map = kmalloc(map_lookup_size(num_stripes), GFP_NOFS);
+	if (!map) {
+		ret = -ENOMEM;
+		goto error;
+	}
+	map->num_stripes = num_stripes;
+
+	for (i = 0; i < ndevs; ++i) {
+		for (j = 0; j < dev_stripes; ++j) {
+			int s = i * dev_stripes + j;
+			map->stripes[s].dev = devices_info[i].dev;
+			map->stripes[s].physical = devices_info[i].dev_offset +
+						   j * stripe_size;
 		}
-		kfree(map);
-		return -ENOSPC;
 	}
 	map->sector_size = extent_root->sectorsize;
-	map->stripe_len = stripe_len;
-	map->io_align = stripe_len;
-	map->io_width = stripe_len;
+	map->stripe_len = raid_stripe_len;
+	map->io_align = raid_stripe_len;
+	map->io_width = raid_stripe_len;
 	map->type = type;
-	map->num_stripes = num_stripes;
 	map->sub_stripes = sub_stripes;
 
-	*map_ret = map;
-	*stripe_size = calc_size;
-	*num_bytes = chunk_bytes_by_type(type, calc_size,
-					 num_stripes, sub_stripes);
+	num_bytes = stripe_size * data_stripes;
 
-	em = alloc_extent_map(GFP_NOFS);
+	trace_btrfs_chunk_alloc(info->chunk_root, map, start, num_bytes);
+
+	em = alloc_extent_map();
 	if (!em) {
 		kfree(map);
-		return -ENOMEM;
+		ret = -ENOMEM;
+		goto error;
 	}
+	set_bit(EXTENT_FLAG_FS_MAPPING, &em->flags);
 	em->bdev = (struct block_device *)map;
 	em->start = start;
-	em->len = *num_bytes;
+	em->len = num_bytes;
 	em->block_start = 0;
 	em->block_len = em->len;
+	em->orig_block_len = stripe_size;
 
 	em_tree = &extent_root->fs_info->mapping_tree.map_tree;
 	write_lock(&em_tree->lock);
-	ret = add_extent_mapping(em_tree, em);
+	ret = add_extent_mapping(em_tree, em, 0);
+	if (!ret) {
+		list_add_tail(&em->list, &trans->transaction->pending_chunks);
+		atomic_inc(&em->refs);
+	}
 	write_unlock(&em_tree->lock);
-	BUG_ON(ret);
-	free_extent_map(em);
+	if (ret) {
+		free_extent_map(em);
+		goto error;
+	}
 
 	ret = btrfs_make_block_group(trans, extent_root, 0, type,
 				     BTRFS_FIRST_CHUNK_TREE_OBJECTID,
-				     start, *num_bytes);
-	BUG_ON(ret);
-
-	index = 0;
-	while (index < map->num_stripes) {
-		device = map->stripes[index].dev;
-		dev_offset = map->stripes[index].physical;
+				     start, num_bytes);
+	if (ret)
+		goto error_del_extent;
 
-		ret = btrfs_alloc_dev_extent(trans, device,
-				info->chunk_root->root_key.objectid,
-				BTRFS_FIRST_CHUNK_TREE_OBJECTID,
-				start, dev_offset, calc_size);
-		BUG_ON(ret);
-		index++;
-	}
+	free_extent_map(em);
+	check_raid56_incompat_flag(extent_root->fs_info, type);
 
+	kfree(devices_info);
 	return 0;
+
+error_del_extent:
+	write_lock(&em_tree->lock);
+	remove_extent_mapping(em_tree, em);
+	write_unlock(&em_tree->lock);
+
+	/* One for our allocation */
+	free_extent_map(em);
+	/* One for the tree reference */
+	free_extent_map(em);
+error:
+	kfree(devices_info);
+	return ret;
 }
 
-static int __finish_chunk_alloc(struct btrfs_trans_handle *trans,
+int btrfs_finish_chunk_alloc(struct btrfs_trans_handle *trans,
 				struct btrfs_root *extent_root,
-				struct map_lookup *map, u64 chunk_offset,
-				u64 chunk_size, u64 stripe_size)
+				u64 chunk_offset, u64 chunk_size)
 {
-	u64 dev_offset;
 	struct btrfs_key key;
 	struct btrfs_root *chunk_root = extent_root->fs_info->chunk_root;
 	struct btrfs_device *device;
 	struct btrfs_chunk *chunk;
 	struct btrfs_stripe *stripe;
-	size_t item_size = btrfs_chunk_item_size(map->num_stripes);
-	int index = 0;
+	struct extent_map_tree *em_tree;
+	struct extent_map *em;
+	struct map_lookup *map;
+	size_t item_size;
+	u64 dev_offset;
+	u64 stripe_size;
+	int i = 0;
 	int ret;
 
+	em_tree = &extent_root->fs_info->mapping_tree.map_tree;
+	read_lock(&em_tree->lock);
+	em = lookup_extent_mapping(em_tree, chunk_offset, chunk_size);
+	read_unlock(&em_tree->lock);
+
+	if (!em) {
+		btrfs_crit(extent_root->fs_info, "unable to find logical "
+			   "%Lu len %Lu", chunk_offset, chunk_size);
+		return -EINVAL;
+	}
+
+	if (em->start != chunk_offset || em->len != chunk_size) {
+		btrfs_crit(extent_root->fs_info, "found a bad mapping, wanted"
+			  " %Lu-%Lu, found %Lu-%Lu", chunk_offset,
+			  chunk_size, em->start, em->len);
+		free_extent_map(em);
+		return -EINVAL;
+	}
+
+	map = (struct map_lookup *)em->bdev;
+	item_size = btrfs_chunk_item_size(map->num_stripes);
+	stripe_size = em->orig_block_len;
+
 	chunk = kzalloc(item_size, GFP_NOFS);
-	if (!chunk)
-		return -ENOMEM;
+	if (!chunk) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	for (i = 0; i < map->num_stripes; i++) {
+		device = map->stripes[i].dev;
+		dev_offset = map->stripes[i].physical;
 
-	index = 0;
-	while (index < map->num_stripes) {
-		device = map->stripes[index].dev;
 		device->bytes_used += stripe_size;
 		ret = btrfs_update_device(trans, device);
-		BUG_ON(ret);
-		index++;
+		if (ret)
+			goto out;
+		ret = btrfs_alloc_dev_extent(trans, device,
+					     chunk_root->root_key.objectid,
+					     BTRFS_FIRST_CHUNK_TREE_OBJECTID,
+					     chunk_offset, dev_offset,
+					     stripe_size);
+		if (ret)
+			goto out;
 	}
 
-	index = 0;
+	spin_lock(&extent_root->fs_info->free_chunk_lock);
+	extent_root->fs_info->free_chunk_space -= (stripe_size *
+						   map->num_stripes);
+	spin_unlock(&extent_root->fs_info->free_chunk_lock);
+
 	stripe = &chunk->stripe;
-	while (index < map->num_stripes) {
-		device = map->stripes[index].dev;
-		dev_offset = map->stripes[index].physical;
+	for (i = 0; i < map->num_stripes; i++) {
+		device = map->stripes[i].dev;
+		dev_offset = map->stripes[i].physical;
 
 		btrfs_set_stack_stripe_devid(stripe, device->devid);
 		btrfs_set_stack_stripe_offset(stripe, dev_offset);
 		memcpy(stripe->dev_uuid, device->uuid, BTRFS_UUID_SIZE);
 		stripe++;
-		index++;
 	}
 
 	btrfs_set_stack_chunk_length(chunk, chunk_size);
@@ -2428,15 +4464,19 @@ static int __finish_chunk_alloc(struct btrfs_trans_handle *trans,
 	key.offset = chunk_offset;
 
 	ret = btrfs_insert_item(trans, chunk_root, &key, chunk, item_size);
-	BUG_ON(ret);
-
-	if (map->type & BTRFS_BLOCK_GROUP_SYSTEM) {
-		ret = btrfs_add_system_chunk(trans, chunk_root, &key, chunk,
+	if (ret == 0 && map->type & BTRFS_BLOCK_GROUP_SYSTEM) {
+		/*
+		 * TODO: Cleanup of inserted chunk root in case of
+		 * failure.
+		 */
+		ret = btrfs_add_system_chunk(chunk_root, &key, chunk,
 					     item_size);
-		BUG_ON(ret);
 	}
+
+out:
 	kfree(chunk);
-	return 0;
+	free_extent_map(em);
+	return ret;
 }
 
 /*
@@ -2450,26 +4490,9 @@ int btrfs_alloc_chunk(struct btrfs_trans_handle *trans,
 		      struct btrfs_root *extent_root, u64 type)
 {
 	u64 chunk_offset;
-	u64 chunk_size;
-	u64 stripe_size;
-	struct map_lookup *map;
-	struct btrfs_root *chunk_root = extent_root->fs_info->chunk_root;
-	int ret;
 
-	ret = find_next_chunk(chunk_root, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
-			      &chunk_offset);
-	if (ret)
-		return ret;
-
-	ret = __btrfs_alloc_chunk(trans, extent_root, &map, &chunk_size,
-				  &stripe_size, chunk_offset, type);
-	if (ret)
-		return ret;
-
-	ret = __finish_chunk_alloc(trans, extent_root, map, chunk_offset,
-				   chunk_size, stripe_size);
-	BUG_ON(ret);
-	return 0;
+	chunk_offset = find_next_chunk(extent_root->fs_info);
+	return __btrfs_alloc_chunk(trans, extent_root, chunk_offset, type);
 }
 
 static noinline int init_first_rw_device(struct btrfs_trans_handle *trans,
@@ -2478,60 +4501,32 @@ static noinline int init_first_rw_device(struct btrfs_trans_handle *trans,
 {
 	u64 chunk_offset;
 	u64 sys_chunk_offset;
-	u64 chunk_size;
-	u64 sys_chunk_size;
-	u64 stripe_size;
-	u64 sys_stripe_size;
 	u64 alloc_profile;
-	struct map_lookup *map;
-	struct map_lookup *sys_map;
 	struct btrfs_fs_info *fs_info = root->fs_info;
 	struct btrfs_root *extent_root = fs_info->extent_root;
 	int ret;
 
-	ret = find_next_chunk(fs_info->chunk_root,
-			      BTRFS_FIRST_CHUNK_TREE_OBJECTID, &chunk_offset);
-	BUG_ON(ret);
-
-	alloc_profile = BTRFS_BLOCK_GROUP_METADATA |
-			(fs_info->metadata_alloc_profile &
-			 fs_info->avail_metadata_alloc_bits);
-	alloc_profile = btrfs_reduce_alloc_profile(root, alloc_profile);
-
-	ret = __btrfs_alloc_chunk(trans, extent_root, &map, &chunk_size,
-				  &stripe_size, chunk_offset, alloc_profile);
-	BUG_ON(ret);
-
-	sys_chunk_offset = chunk_offset + chunk_size;
-
-	alloc_profile = BTRFS_BLOCK_GROUP_SYSTEM |
-			(fs_info->system_alloc_profile &
-			 fs_info->avail_system_alloc_bits);
-	alloc_profile = btrfs_reduce_alloc_profile(root, alloc_profile);
+	chunk_offset = find_next_chunk(fs_info);
+	alloc_profile = btrfs_get_alloc_profile(extent_root, 0);
+	ret = __btrfs_alloc_chunk(trans, extent_root, chunk_offset,
+				  alloc_profile);
+	if (ret)
+		return ret;
 
-	ret = __btrfs_alloc_chunk(trans, extent_root, &sys_map,
-				  &sys_chunk_size, &sys_stripe_size,
-				  sys_chunk_offset, alloc_profile);
-	BUG_ON(ret);
+	sys_chunk_offset = find_next_chunk(root->fs_info);
+	alloc_profile = btrfs_get_alloc_profile(fs_info->chunk_root, 0);
+	ret = __btrfs_alloc_chunk(trans, extent_root, sys_chunk_offset,
+				  alloc_profile);
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
+		goto out;
+	}
 
 	ret = btrfs_add_device(trans, fs_info->chunk_root, device);
-	BUG_ON(ret);
-
-	/*
-	 * Modifying chunk tree needs allocating new blocks from both
-	 * system block group and metadata block group. So we only can
-	 * do operations require modifying the chunk tree after both
-	 * block groups were created.
-	 */
-	ret = __finish_chunk_alloc(trans, extent_root, map, chunk_offset,
-				   chunk_size, stripe_size);
-	BUG_ON(ret);
-
-	ret = __finish_chunk_alloc(trans, extent_root, sys_map,
-				   sys_chunk_offset, sys_chunk_size,
-				   sys_stripe_size);
-	BUG_ON(ret);
-	return 0;
+	if (ret)
+		btrfs_abort_transaction(trans, root, ret);
+out:
+	return ret;
 }
 
 int btrfs_chunk_readonly(struct btrfs_root *root, u64 chunk_offset)
@@ -2566,7 +4561,7 @@ int btrfs_chunk_readonly(struct btrfs_root *root, u64 chunk_offset)
 
 void btrfs_mapping_init(struct btrfs_mapping_tree *tree)
 {
-	extent_map_tree_init(&tree->map_tree, GFP_NOFS);
+	extent_map_tree_init(&tree->map_tree);
 }
 
 void btrfs_mapping_tree_free(struct btrfs_mapping_tree *tree)
@@ -2581,7 +4576,6 @@ void btrfs_mapping_tree_free(struct btrfs_mapping_tree *tree)
 		write_unlock(&tree->map_tree.lock);
 		if (!em)
 			break;
-		kfree(em->bdev);
 		/* once for us */
 		free_extent_map(em);
 		/* once for the tree */
@@ -2589,8 +4583,9 @@ void btrfs_mapping_tree_free(struct btrfs_mapping_tree *tree)
 	}
 }
 
-int btrfs_num_copies(struct btrfs_mapping_tree *map_tree, u64 logical, u64 len)
+int btrfs_num_copies(struct btrfs_fs_info *fs_info, u64 logical, u64 len)
 {
+	struct btrfs_mapping_tree *map_tree = &fs_info->mapping_tree;
 	struct extent_map *em;
 	struct map_lookup *map;
 	struct extent_map_tree *em_tree = &map_tree->map_tree;
@@ -2599,152 +4594,388 @@ int btrfs_num_copies(struct btrfs_mapping_tree *map_tree, u64 logical, u64 len)
 	read_lock(&em_tree->lock);
 	em = lookup_extent_mapping(em_tree, logical, len);
 	read_unlock(&em_tree->lock);
-	BUG_ON(!em);
 
-	BUG_ON(em->start > logical || em->start + em->len < logical);
+	/*
+	 * We could return errors for these cases, but that could get ugly and
+	 * we'd probably do the same thing which is just not do anything else
+	 * and exit, so return 1 so the callers don't try to use other copies.
+	 */
+	if (!em) {
+		btrfs_crit(fs_info, "No mapping for %Lu-%Lu", logical,
+			    logical+len);
+		return 1;
+	}
+
+	if (em->start > logical || em->start + em->len < logical) {
+		btrfs_crit(fs_info, "Invalid mapping for %Lu-%Lu, got "
+			    "%Lu-%Lu", logical, logical+len, em->start,
+			    em->start + em->len);
+		free_extent_map(em);
+		return 1;
+	}
+
 	map = (struct map_lookup *)em->bdev;
 	if (map->type & (BTRFS_BLOCK_GROUP_DUP | BTRFS_BLOCK_GROUP_RAID1))
 		ret = map->num_stripes;
 	else if (map->type & BTRFS_BLOCK_GROUP_RAID10)
 		ret = map->sub_stripes;
+	else if (map->type & BTRFS_BLOCK_GROUP_RAID5)
+		ret = 2;
+	else if (map->type & BTRFS_BLOCK_GROUP_RAID6)
+		ret = 3;
 	else
 		ret = 1;
 	free_extent_map(em);
+
+	btrfs_dev_replace_lock(&fs_info->dev_replace);
+	if (btrfs_dev_replace_is_ongoing(&fs_info->dev_replace))
+		ret++;
+	btrfs_dev_replace_unlock(&fs_info->dev_replace);
+
 	return ret;
 }
 
-static int find_live_mirror(struct map_lookup *map, int first, int num,
-			    int optimal)
+unsigned long btrfs_full_stripe_len(struct btrfs_root *root,
+				    struct btrfs_mapping_tree *map_tree,
+				    u64 logical)
+{
+	struct extent_map *em;
+	struct map_lookup *map;
+	struct extent_map_tree *em_tree = &map_tree->map_tree;
+	unsigned long len = root->sectorsize;
+
+	read_lock(&em_tree->lock);
+	em = lookup_extent_mapping(em_tree, logical, len);
+	read_unlock(&em_tree->lock);
+	BUG_ON(!em);
+
+	BUG_ON(em->start > logical || em->start + em->len < logical);
+	map = (struct map_lookup *)em->bdev;
+	if (map->type & (BTRFS_BLOCK_GROUP_RAID5 |
+			 BTRFS_BLOCK_GROUP_RAID6)) {
+		len = map->stripe_len * nr_data_stripes(map);
+	}
+	free_extent_map(em);
+	return len;
+}
+
+int btrfs_is_parity_mirror(struct btrfs_mapping_tree *map_tree,
+			   u64 logical, u64 len, int mirror_num)
+{
+	struct extent_map *em;
+	struct map_lookup *map;
+	struct extent_map_tree *em_tree = &map_tree->map_tree;
+	int ret = 0;
+
+	read_lock(&em_tree->lock);
+	em = lookup_extent_mapping(em_tree, logical, len);
+	read_unlock(&em_tree->lock);
+	BUG_ON(!em);
+
+	BUG_ON(em->start > logical || em->start + em->len < logical);
+	map = (struct map_lookup *)em->bdev;
+	if (map->type & (BTRFS_BLOCK_GROUP_RAID5 |
+			 BTRFS_BLOCK_GROUP_RAID6))
+		ret = 1;
+	free_extent_map(em);
+	return ret;
+}
+
+static int find_live_mirror(struct btrfs_fs_info *fs_info,
+			    struct map_lookup *map, int first, int num,
+			    int optimal, int dev_replace_is_ongoing)
 {
 	int i;
-	if (map->stripes[optimal].dev->bdev)
-		return optimal;
-	for (i = first; i < first + num; i++) {
-		if (map->stripes[i].dev->bdev)
-			return i;
+	int tolerance;
+	struct btrfs_device *srcdev;
+
+	if (dev_replace_is_ongoing &&
+	    fs_info->dev_replace.cont_reading_from_srcdev_mode ==
+	     BTRFS_DEV_REPLACE_ITEM_CONT_READING_FROM_SRCDEV_MODE_AVOID)
+		srcdev = fs_info->dev_replace.srcdev;
+	else
+		srcdev = NULL;
+
+	/*
+	 * try to avoid the drive that is the source drive for a
+	 * dev-replace procedure, only choose it if no other non-missing
+	 * mirror is available
+	 */
+	for (tolerance = 0; tolerance < 2; tolerance++) {
+		if (map->stripes[optimal].dev->bdev &&
+		    (tolerance || map->stripes[optimal].dev != srcdev))
+			return optimal;
+		for (i = first; i < first + num; i++) {
+			if (map->stripes[i].dev->bdev &&
+			    (tolerance || map->stripes[i].dev != srcdev))
+				return i;
+		}
 	}
+
 	/* we couldn't find one that doesn't fail.  Just return something
 	 * and the io error handling code will clean up eventually
 	 */
 	return optimal;
 }
 
-static int __btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw,
+static inline int parity_smaller(u64 a, u64 b)
+{
+	return a > b;
+}
+
+/* Bubble-sort the stripe set to put the parity/syndrome stripes last */
+static void sort_parity_stripes(struct btrfs_bio *bbio, u64 *raid_map)
+{
+	struct btrfs_bio_stripe s;
+	int i;
+	u64 l;
+	int again = 1;
+
+	while (again) {
+		again = 0;
+		for (i = 0; i < bbio->num_stripes - 1; i++) {
+			if (parity_smaller(raid_map[i], raid_map[i+1])) {
+				s = bbio->stripes[i];
+				l = raid_map[i];
+				bbio->stripes[i] = bbio->stripes[i+1];
+				raid_map[i] = raid_map[i+1];
+				bbio->stripes[i+1] = s;
+				raid_map[i+1] = l;
+				again = 1;
+			}
+		}
+	}
+}
+
+static int __btrfs_map_block(struct btrfs_fs_info *fs_info, int rw,
 			     u64 logical, u64 *length,
-			     struct btrfs_multi_bio **multi_ret,
-			     int mirror_num, struct page *unplug_page)
+			     struct btrfs_bio **bbio_ret,
+			     int mirror_num, u64 **raid_map_ret)
 {
 	struct extent_map *em;
 	struct map_lookup *map;
+	struct btrfs_mapping_tree *map_tree = &fs_info->mapping_tree;
 	struct extent_map_tree *em_tree = &map_tree->map_tree;
 	u64 offset;
 	u64 stripe_offset;
+	u64 stripe_end_offset;
 	u64 stripe_nr;
-	int stripes_allocated = 8;
-	int stripes_required = 1;
+	u64 stripe_nr_orig;
+	u64 stripe_nr_end;
+	u64 stripe_len;
+	u64 *raid_map = NULL;
 	int stripe_index;
 	int i;
+	int ret = 0;
 	int num_stripes;
 	int max_errors = 0;
-	struct btrfs_multi_bio *multi = NULL;
-
-	if (multi_ret && !(rw & (1 << BIO_RW)))
-		stripes_allocated = 1;
-again:
-	if (multi_ret) {
-		multi = kzalloc(btrfs_multi_bio_size(stripes_allocated),
-				GFP_NOFS);
-		if (!multi)
-			return -ENOMEM;
-
-		atomic_set(&multi->error, 0);
-	}
+	struct btrfs_bio *bbio = NULL;
+	struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
+	int dev_replace_is_ongoing = 0;
+	int num_alloc_stripes;
+	int patch_the_first_stripe_for_dev_replace = 0;
+	u64 physical_to_patch_in_first_stripe = 0;
+	u64 raid56_full_stripe_start = (u64)-1;
 
 	read_lock(&em_tree->lock);
 	em = lookup_extent_mapping(em_tree, logical, *length);
 	read_unlock(&em_tree->lock);
 
-	if (!em && unplug_page) {
-		kfree(multi);
-		return 0;
+	if (!em) {
+		btrfs_crit(fs_info, "unable to find logical %llu len %llu",
+			logical, *length);
+		return -EINVAL;
 	}
 
-	if (!em) {
-		printk(KERN_CRIT "unable to find logical %llu len %llu\n",
-		       (unsigned long long)logical,
-		       (unsigned long long)*length);
-		BUG();
+	if (em->start > logical || em->start + em->len < logical) {
+		btrfs_crit(fs_info, "found a bad mapping, wanted %Lu, "
+			   "found %Lu-%Lu", logical, em->start,
+			   em->start + em->len);
+		free_extent_map(em);
+		return -EINVAL;
 	}
 
-	BUG_ON(em->start > logical || em->start + em->len < logical);
 	map = (struct map_lookup *)em->bdev;
 	offset = logical - em->start;
 
-	if (mirror_num > map->num_stripes)
-		mirror_num = 0;
-
-	/* if our multi bio struct is too small, back off and try again */
-	if (rw & (1 << BIO_RW)) {
-		if (map->type & (BTRFS_BLOCK_GROUP_RAID1 |
-				 BTRFS_BLOCK_GROUP_DUP)) {
-			stripes_required = map->num_stripes;
-			max_errors = 1;
-		} else if (map->type & BTRFS_BLOCK_GROUP_RAID10) {
-			stripes_required = map->sub_stripes;
-			max_errors = 1;
-		}
-	}
-	if (multi_ret && (rw & (1 << BIO_RW)) &&
-	    stripes_allocated < stripes_required) {
-		stripes_allocated = map->num_stripes;
-		free_extent_map(em);
-		kfree(multi);
-		goto again;
-	}
+	stripe_len = map->stripe_len;
 	stripe_nr = offset;
 	/*
 	 * stripe_nr counts the total number of stripes we have to stride
 	 * to get to this block
 	 */
-	do_div(stripe_nr, map->stripe_len);
+	do_div(stripe_nr, stripe_len);
 
-	stripe_offset = stripe_nr * map->stripe_len;
+	stripe_offset = stripe_nr * stripe_len;
 	BUG_ON(offset < stripe_offset);
 
 	/* stripe_offset is the offset of this block in its stripe*/
 	stripe_offset = offset - stripe_offset;
 
-	if (map->type & (BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 |
-			 BTRFS_BLOCK_GROUP_RAID10 |
-			 BTRFS_BLOCK_GROUP_DUP)) {
-		/* we limit the length of each bio to what fits in a stripe */
-		*length = min_t(u64, em->len - offset,
-			      map->stripe_len - stripe_offset);
+	/* if we're here for raid56, we need to know the stripe aligned start */
+	if (map->type & (BTRFS_BLOCK_GROUP_RAID5 | BTRFS_BLOCK_GROUP_RAID6)) {
+		unsigned long full_stripe_len = stripe_len * nr_data_stripes(map);
+		raid56_full_stripe_start = offset;
+
+		/* allow a write of a full stripe, but make sure we don't
+		 * allow straddling of stripes
+		 */
+		do_div(raid56_full_stripe_start, full_stripe_len);
+		raid56_full_stripe_start *= full_stripe_len;
+	}
+
+	if (rw & REQ_DISCARD) {
+		/* we don't discard raid56 yet */
+		if (map->type &
+		    (BTRFS_BLOCK_GROUP_RAID5 | BTRFS_BLOCK_GROUP_RAID6)) {
+			ret = -EOPNOTSUPP;
+			goto out;
+		}
+		*length = min_t(u64, em->len - offset, *length);
+	} else if (map->type & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
+		u64 max_len;
+		/* For writes to RAID[56], allow a full stripeset across all disks.
+		   For other RAID types and for RAID[56] reads, just allow a single
+		   stripe (on a single disk). */
+		if (map->type & (BTRFS_BLOCK_GROUP_RAID5 | BTRFS_BLOCK_GROUP_RAID6) &&
+		    (rw & REQ_WRITE)) {
+			max_len = stripe_len * nr_data_stripes(map) -
+				(offset - raid56_full_stripe_start);
+		} else {
+			/* we limit the length of each bio to what fits in a stripe */
+			max_len = stripe_len - stripe_offset;
+		}
+		*length = min_t(u64, em->len - offset, max_len);
 	} else {
 		*length = em->len - offset;
 	}
 
-	if (!multi_ret && !unplug_page)
+	/* This is for when we're called from btrfs_merge_bio_hook() and all
+	   it cares about is the length */
+	if (!bbio_ret)
 		goto out;
 
+	btrfs_dev_replace_lock(dev_replace);
+	dev_replace_is_ongoing = btrfs_dev_replace_is_ongoing(dev_replace);
+	if (!dev_replace_is_ongoing)
+		btrfs_dev_replace_unlock(dev_replace);
+
+	if (dev_replace_is_ongoing && mirror_num == map->num_stripes + 1 &&
+	    !(rw & (REQ_WRITE | REQ_DISCARD | REQ_GET_READ_MIRRORS)) &&
+	    dev_replace->tgtdev != NULL) {
+		/*
+		 * in dev-replace case, for repair case (that's the only
+		 * case where the mirror is selected explicitly when
+		 * calling btrfs_map_block), blocks left of the left cursor
+		 * can also be read from the target drive.
+		 * For REQ_GET_READ_MIRRORS, the target drive is added as
+		 * the last one to the array of stripes. For READ, it also
+		 * needs to be supported using the same mirror number.
+		 * If the requested block is not left of the left cursor,
+		 * EIO is returned. This can happen because btrfs_num_copies()
+		 * returns one more in the dev-replace case.
+		 */
+		u64 tmp_length = *length;
+		struct btrfs_bio *tmp_bbio = NULL;
+		int tmp_num_stripes;
+		u64 srcdev_devid = dev_replace->srcdev->devid;
+		int index_srcdev = 0;
+		int found = 0;
+		u64 physical_of_found = 0;
+
+		ret = __btrfs_map_block(fs_info, REQ_GET_READ_MIRRORS,
+			     logical, &tmp_length, &tmp_bbio, 0, NULL);
+		if (ret) {
+			WARN_ON(tmp_bbio != NULL);
+			goto out;
+		}
+
+		tmp_num_stripes = tmp_bbio->num_stripes;
+		if (mirror_num > tmp_num_stripes) {
+			/*
+			 * REQ_GET_READ_MIRRORS does not contain this
+			 * mirror, that means that the requested area
+			 * is not left of the left cursor
+			 */
+			ret = -EIO;
+			kfree(tmp_bbio);
+			goto out;
+		}
+
+		/*
+		 * process the rest of the function using the mirror_num
+		 * of the source drive. Therefore look it up first.
+		 * At the end, patch the device pointer to the one of the
+		 * target drive.
+		 */
+		for (i = 0; i < tmp_num_stripes; i++) {
+			if (tmp_bbio->stripes[i].dev->devid == srcdev_devid) {
+				/*
+				 * In case of DUP, in order to keep it
+				 * simple, only add the mirror with the
+				 * lowest physical address
+				 */
+				if (found &&
+				    physical_of_found <=
+				     tmp_bbio->stripes[i].physical)
+					continue;
+				index_srcdev = i;
+				found = 1;
+				physical_of_found =
+					tmp_bbio->stripes[i].physical;
+			}
+		}
+
+		if (found) {
+			mirror_num = index_srcdev + 1;
+			patch_the_first_stripe_for_dev_replace = 1;
+			physical_to_patch_in_first_stripe = physical_of_found;
+		} else {
+			WARN_ON(1);
+			ret = -EIO;
+			kfree(tmp_bbio);
+			goto out;
+		}
+
+		kfree(tmp_bbio);
+	} else if (mirror_num > map->num_stripes) {
+		mirror_num = 0;
+	}
+
 	num_stripes = 1;
 	stripe_index = 0;
-	if (map->type & BTRFS_BLOCK_GROUP_RAID1) {
-		if (unplug_page || (rw & (1 << BIO_RW)))
+	stripe_nr_orig = stripe_nr;
+	stripe_nr_end = ALIGN(offset + *length, map->stripe_len);
+	do_div(stripe_nr_end, map->stripe_len);
+	stripe_end_offset = stripe_nr_end * map->stripe_len -
+			    (offset + *length);
+
+	if (map->type & BTRFS_BLOCK_GROUP_RAID0) {
+		if (rw & REQ_DISCARD)
+			num_stripes = min_t(u64, map->num_stripes,
+					    stripe_nr_end - stripe_nr_orig);
+		stripe_index = do_div(stripe_nr, map->num_stripes);
+	} else if (map->type & BTRFS_BLOCK_GROUP_RAID1) {
+		if (rw & (REQ_WRITE | REQ_DISCARD | REQ_GET_READ_MIRRORS))
 			num_stripes = map->num_stripes;
 		else if (mirror_num)
 			stripe_index = mirror_num - 1;
 		else {
-			stripe_index = find_live_mirror(map, 0,
+			stripe_index = find_live_mirror(fs_info, map, 0,
 					    map->num_stripes,
-					    current->pid % map->num_stripes);
+					    current->pid % map->num_stripes,
+					    dev_replace_is_ongoing);
+			mirror_num = stripe_index + 1;
 		}
 
 	} else if (map->type & BTRFS_BLOCK_GROUP_DUP) {
-		if (rw & (1 << BIO_RW))
+		if (rw & (REQ_WRITE | REQ_DISCARD | REQ_GET_READ_MIRRORS)) {
 			num_stripes = map->num_stripes;
-		else if (mirror_num)
+		} else if (mirror_num) {
 			stripe_index = mirror_num - 1;
+		} else {
+			mirror_num = 1;
+		}
 
 	} else if (map->type & BTRFS_BLOCK_GROUP_RAID10) {
 		int factor = map->num_stripes / map->sub_stripes;
@@ -2752,14 +4983,81 @@ again:
 		stripe_index = do_div(stripe_nr, factor);
 		stripe_index *= map->sub_stripes;
 
-		if (unplug_page || (rw & (1 << BIO_RW)))
+		if (rw & (REQ_WRITE | REQ_GET_READ_MIRRORS))
 			num_stripes = map->sub_stripes;
+		else if (rw & REQ_DISCARD)
+			num_stripes = min_t(u64, map->sub_stripes *
+					    (stripe_nr_end - stripe_nr_orig),
+					    map->num_stripes);
 		else if (mirror_num)
 			stripe_index += mirror_num - 1;
 		else {
-			stripe_index = find_live_mirror(map, stripe_index,
+			int old_stripe_index = stripe_index;
+			stripe_index = find_live_mirror(fs_info, map,
+					      stripe_index,
 					      map->sub_stripes, stripe_index +
-					      current->pid % map->sub_stripes);
+					      current->pid % map->sub_stripes,
+					      dev_replace_is_ongoing);
+			mirror_num = stripe_index - old_stripe_index + 1;
+		}
+
+	} else if (map->type & (BTRFS_BLOCK_GROUP_RAID5 |
+				BTRFS_BLOCK_GROUP_RAID6)) {
+		u64 tmp;
+
+		if (bbio_ret && ((rw & REQ_WRITE) || mirror_num > 1)
+		    && raid_map_ret) {
+			int i, rot;
+
+			/* push stripe_nr back to the start of the full stripe */
+			stripe_nr = raid56_full_stripe_start;
+			do_div(stripe_nr, stripe_len);
+
+			stripe_index = do_div(stripe_nr, nr_data_stripes(map));
+
+			/* RAID[56] write or recovery. Return all stripes */
+			num_stripes = map->num_stripes;
+			max_errors = nr_parity_stripes(map);
+
+			raid_map = kmalloc_array(num_stripes, sizeof(u64),
+					   GFP_NOFS);
+			if (!raid_map) {
+				ret = -ENOMEM;
+				goto out;
+			}
+
+			/* Work out the disk rotation on this stripe-set */
+			tmp = stripe_nr;
+			rot = do_div(tmp, num_stripes);
+
+			/* Fill in the logical address of each stripe */
+			tmp = stripe_nr * nr_data_stripes(map);
+			for (i = 0; i < nr_data_stripes(map); i++)
+				raid_map[(i+rot) % num_stripes] =
+					em->start + (tmp + i) * map->stripe_len;
+
+			raid_map[(i+rot) % map->num_stripes] = RAID5_P_STRIPE;
+			if (map->type & BTRFS_BLOCK_GROUP_RAID6)
+				raid_map[(i+rot+1) % num_stripes] =
+					RAID6_Q_STRIPE;
+
+			*length = map->stripe_len;
+			stripe_index = 0;
+			stripe_offset = 0;
+		} else {
+			/*
+			 * Mirror #0 or #1 means the original data block.
+			 * Mirror #2 is RAID5 parity block.
+			 * Mirror #3 is RAID6 Q block.
+			 */
+			stripe_index = do_div(stripe_nr, nr_data_stripes(map));
+			if (mirror_num > 1)
+				stripe_index = nr_data_stripes(map) +
+						mirror_num - 2;
+
+			/* We distribute the parity blocks across stripes */
+			tmp = stripe_nr + stripe_index;
+			stripe_index = do_div(tmp, map->num_stripes);
 		}
 	} else {
 		/*
@@ -2768,43 +5066,229 @@ again:
 		 * stripe_index is the number of our device in the stripe array
 		 */
 		stripe_index = do_div(stripe_nr, map->num_stripes);
+		mirror_num = stripe_index + 1;
 	}
 	BUG_ON(stripe_index >= map->num_stripes);
 
-	for (i = 0; i < num_stripes; i++) {
-		if (unplug_page) {
-			struct btrfs_device *device;
-			struct backing_dev_info *bdi;
-
-			device = map->stripes[stripe_index].dev;
-			if (device->bdev) {
-				bdi = blk_get_backing_dev_info(device->bdev);
-				if (bdi->unplug_io_fn)
-					bdi->unplug_io_fn(bdi, unplug_page);
-			}
-		} else {
-			multi->stripes[i].physical =
+	num_alloc_stripes = num_stripes;
+	if (dev_replace_is_ongoing) {
+		if (rw & (REQ_WRITE | REQ_DISCARD))
+			num_alloc_stripes <<= 1;
+		if (rw & REQ_GET_READ_MIRRORS)
+			num_alloc_stripes++;
+	}
+	bbio = kzalloc(btrfs_bio_size(num_alloc_stripes), GFP_NOFS);
+	if (!bbio) {
+		kfree(raid_map);
+		ret = -ENOMEM;
+		goto out;
+	}
+	atomic_set(&bbio->error, 0);
+
+	if (rw & REQ_DISCARD) {
+		int factor = 0;
+		int sub_stripes = 0;
+		u64 stripes_per_dev = 0;
+		u32 remaining_stripes = 0;
+		u32 last_stripe = 0;
+
+		if (map->type &
+		    (BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID10)) {
+			if (map->type & BTRFS_BLOCK_GROUP_RAID0)
+				sub_stripes = 1;
+			else
+				sub_stripes = map->sub_stripes;
+
+			factor = map->num_stripes / sub_stripes;
+			stripes_per_dev = div_u64_rem(stripe_nr_end -
+						      stripe_nr_orig,
+						      factor,
+						      &remaining_stripes);
+			div_u64_rem(stripe_nr_end - 1, factor, &last_stripe);
+			last_stripe *= sub_stripes;
+		}
+
+		for (i = 0; i < num_stripes; i++) {
+			bbio->stripes[i].physical =
 				map->stripes[stripe_index].physical +
 				stripe_offset + stripe_nr * map->stripe_len;
-			multi->stripes[i].dev = map->stripes[stripe_index].dev;
+			bbio->stripes[i].dev = map->stripes[stripe_index].dev;
+
+			if (map->type & (BTRFS_BLOCK_GROUP_RAID0 |
+					 BTRFS_BLOCK_GROUP_RAID10)) {
+				bbio->stripes[i].length = stripes_per_dev *
+							  map->stripe_len;
+
+				if (i / sub_stripes < remaining_stripes)
+					bbio->stripes[i].length +=
+						map->stripe_len;
+
+				/*
+				 * Special for the first stripe and
+				 * the last stripe:
+				 *
+				 * |-------|...|-------|
+				 *     |----------|
+				 *    off     end_off
+				 */
+				if (i < sub_stripes)
+					bbio->stripes[i].length -=
+						stripe_offset;
+
+				if (stripe_index >= last_stripe &&
+				    stripe_index <= (last_stripe +
+						     sub_stripes - 1))
+					bbio->stripes[i].length -=
+						stripe_end_offset;
+
+				if (i == sub_stripes - 1)
+					stripe_offset = 0;
+			} else
+				bbio->stripes[i].length = *length;
+
+			stripe_index++;
+			if (stripe_index == map->num_stripes) {
+				/* This could only happen for RAID0/10 */
+				stripe_index = 0;
+				stripe_nr++;
+			}
+		}
+	} else {
+		for (i = 0; i < num_stripes; i++) {
+			bbio->stripes[i].physical =
+				map->stripes[stripe_index].physical +
+				stripe_offset +
+				stripe_nr * map->stripe_len;
+			bbio->stripes[i].dev =
+				map->stripes[stripe_index].dev;
+			stripe_index++;
+		}
+	}
+
+	if (rw & (REQ_WRITE | REQ_GET_READ_MIRRORS)) {
+		if (map->type & (BTRFS_BLOCK_GROUP_RAID1 |
+				 BTRFS_BLOCK_GROUP_RAID10 |
+				 BTRFS_BLOCK_GROUP_RAID5 |
+				 BTRFS_BLOCK_GROUP_DUP)) {
+			max_errors = 1;
+		} else if (map->type & BTRFS_BLOCK_GROUP_RAID6) {
+			max_errors = 2;
 		}
-		stripe_index++;
 	}
-	if (multi_ret) {
-		*multi_ret = multi;
-		multi->num_stripes = num_stripes;
-		multi->max_errors = max_errors;
+
+	if (dev_replace_is_ongoing && (rw & (REQ_WRITE | REQ_DISCARD)) &&
+	    dev_replace->tgtdev != NULL) {
+		int index_where_to_add;
+		u64 srcdev_devid = dev_replace->srcdev->devid;
+
+		/*
+		 * duplicate the write operations while the dev replace
+		 * procedure is running. Since the copying of the old disk
+		 * to the new disk takes place at run time while the
+		 * filesystem is mounted writable, the regular write
+		 * operations to the old disk have to be duplicated to go
+		 * to the new disk as well.
+		 * Note that device->missing is handled by the caller, and
+		 * that the write to the old disk is already set up in the
+		 * stripes array.
+		 */
+		index_where_to_add = num_stripes;
+		for (i = 0; i < num_stripes; i++) {
+			if (bbio->stripes[i].dev->devid == srcdev_devid) {
+				/* write to new disk, too */
+				struct btrfs_bio_stripe *new =
+					bbio->stripes + index_where_to_add;
+				struct btrfs_bio_stripe *old =
+					bbio->stripes + i;
+
+				new->physical = old->physical;
+				new->length = old->length;
+				new->dev = dev_replace->tgtdev;
+				index_where_to_add++;
+				max_errors++;
+			}
+		}
+		num_stripes = index_where_to_add;
+	} else if (dev_replace_is_ongoing && (rw & REQ_GET_READ_MIRRORS) &&
+		   dev_replace->tgtdev != NULL) {
+		u64 srcdev_devid = dev_replace->srcdev->devid;
+		int index_srcdev = 0;
+		int found = 0;
+		u64 physical_of_found = 0;
+
+		/*
+		 * During the dev-replace procedure, the target drive can
+		 * also be used to read data in case it is needed to repair
+		 * a corrupt block elsewhere. This is possible if the
+		 * requested area is left of the left cursor. In this area,
+		 * the target drive is a full copy of the source drive.
+		 */
+		for (i = 0; i < num_stripes; i++) {
+			if (bbio->stripes[i].dev->devid == srcdev_devid) {
+				/*
+				 * In case of DUP, in order to keep it
+				 * simple, only add the mirror with the
+				 * lowest physical address
+				 */
+				if (found &&
+				    physical_of_found <=
+				     bbio->stripes[i].physical)
+					continue;
+				index_srcdev = i;
+				found = 1;
+				physical_of_found = bbio->stripes[i].physical;
+			}
+		}
+		if (found) {
+			u64 length = map->stripe_len;
+
+			if (physical_of_found + length <=
+			    dev_replace->cursor_left) {
+				struct btrfs_bio_stripe *tgtdev_stripe =
+					bbio->stripes + num_stripes;
+
+				tgtdev_stripe->physical = physical_of_found;
+				tgtdev_stripe->length =
+					bbio->stripes[index_srcdev].length;
+				tgtdev_stripe->dev = dev_replace->tgtdev;
+
+				num_stripes++;
+			}
+		}
+	}
+
+	*bbio_ret = bbio;
+	bbio->num_stripes = num_stripes;
+	bbio->max_errors = max_errors;
+	bbio->mirror_num = mirror_num;
+
+	/*
+	 * this is the case that REQ_READ && dev_replace_is_ongoing &&
+	 * mirror_num == num_stripes + 1 && dev_replace target drive is
+	 * available as a mirror
+	 */
+	if (patch_the_first_stripe_for_dev_replace && num_stripes > 0) {
+		WARN_ON(num_stripes > 1);
+		bbio->stripes[0].dev = dev_replace->tgtdev;
+		bbio->stripes[0].physical = physical_to_patch_in_first_stripe;
+		bbio->mirror_num = map->num_stripes + 1;
+	}
+	if (raid_map) {
+		sort_parity_stripes(bbio, raid_map);
+		*raid_map_ret = raid_map;
 	}
 out:
+	if (dev_replace_is_ongoing)
+		btrfs_dev_replace_unlock(dev_replace);
 	free_extent_map(em);
-	return 0;
+	return ret;
 }
 
-int btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw,
+int btrfs_map_block(struct btrfs_fs_info *fs_info, int rw,
 		      u64 logical, u64 *length,
-		      struct btrfs_multi_bio **multi_ret, int mirror_num)
+		      struct btrfs_bio **bbio_ret, int mirror_num)
 {
-	return __btrfs_map_block(map_tree, rw, logical, length, multi_ret,
+	return __btrfs_map_block(fs_info, rw, logical, length, bbio_ret,
 				 mirror_num, NULL);
 }
 
@@ -2819,23 +5303,42 @@ int btrfs_rmap_block(struct btrfs_mapping_tree *map_tree,
 	u64 bytenr;
 	u64 length;
 	u64 stripe_nr;
+	u64 rmap_len;
 	int i, j, nr = 0;
 
 	read_lock(&em_tree->lock);
 	em = lookup_extent_mapping(em_tree, chunk_start, 1);
 	read_unlock(&em_tree->lock);
 
-	BUG_ON(!em || em->start != chunk_start);
+	if (!em) {
+		printk(KERN_ERR "BTRFS: couldn't find em for chunk %Lu\n",
+		       chunk_start);
+		return -EIO;
+	}
+
+	if (em->start != chunk_start) {
+		printk(KERN_ERR "BTRFS: bad chunk start, em=%Lu, wanted=%Lu\n",
+		       em->start, chunk_start);
+		free_extent_map(em);
+		return -EIO;
+	}
 	map = (struct map_lookup *)em->bdev;
 
 	length = em->len;
+	rmap_len = map->stripe_len;
+
 	if (map->type & BTRFS_BLOCK_GROUP_RAID10)
 		do_div(length, map->num_stripes / map->sub_stripes);
 	else if (map->type & BTRFS_BLOCK_GROUP_RAID0)
 		do_div(length, map->num_stripes);
+	else if (map->type & (BTRFS_BLOCK_GROUP_RAID5 |
+			      BTRFS_BLOCK_GROUP_RAID6)) {
+		do_div(length, nr_data_stripes(map));
+		rmap_len = map->stripe_len * nr_data_stripes(map);
+	}
 
 	buf = kzalloc(sizeof(u64) * map->num_stripes, GFP_NOFS);
-	BUG_ON(!buf);
+	BUG_ON(!buf); /* -ENOMEM */
 
 	for (i = 0; i < map->num_stripes; i++) {
 		if (devid && map->stripes[i].dev->devid != devid)
@@ -2852,8 +5355,11 @@ int btrfs_rmap_block(struct btrfs_mapping_tree *map_tree,
 			do_div(stripe_nr, map->sub_stripes);
 		} else if (map->type & BTRFS_BLOCK_GROUP_RAID0) {
 			stripe_nr = stripe_nr * map->num_stripes + i;
-		}
-		bytenr = chunk_start + stripe_nr * map->stripe_len;
+		} /* else if RAID[56], multiply by nr_data_stripes().
+		   * Alternatively, just use rmap_len below instead of
+		   * map->stripe_len */
+
+		bytenr = chunk_start + stripe_nr * rmap_len;
 		WARN_ON(nr >= map->num_stripes);
 		for (j = 0; j < nr; j++) {
 			if (buf[j] == bytenr)
@@ -2867,44 +5373,70 @@ int btrfs_rmap_block(struct btrfs_mapping_tree *map_tree,
 
 	*logical = buf;
 	*naddrs = nr;
-	*stripe_len = map->stripe_len;
+	*stripe_len = rmap_len;
 
 	free_extent_map(em);
 	return 0;
 }
 
-int btrfs_unplug_page(struct btrfs_mapping_tree *map_tree,
-		      u64 logical, struct page *page)
+static inline void btrfs_end_bbio(struct btrfs_bio *bbio, struct bio *bio, int err)
 {
-	u64 length = PAGE_CACHE_SIZE;
-	return __btrfs_map_block(map_tree, READ, logical, &length,
-				 NULL, 0, page);
+	if (likely(bbio->flags & BTRFS_BIO_ORIG_BIO_SUBMITTED))
+		bio_endio_nodec(bio, err);
+	else
+		bio_endio(bio, err);
+	kfree(bbio);
 }
 
-static void end_bio_multi_stripe(struct bio *bio, int err)
+static void btrfs_end_bio(struct bio *bio, int err)
 {
-	struct btrfs_multi_bio *multi = bio->bi_private;
+	struct btrfs_bio *bbio = bio->bi_private;
+	struct btrfs_device *dev = bbio->stripes[0].dev;
 	int is_orig_bio = 0;
 
-	if (err)
-		atomic_inc(&multi->error);
+	if (err) {
+		atomic_inc(&bbio->error);
+		if (err == -EIO || err == -EREMOTEIO) {
+			unsigned int stripe_index =
+				btrfs_io_bio(bio)->stripe_index;
+
+			BUG_ON(stripe_index >= bbio->num_stripes);
+			dev = bbio->stripes[stripe_index].dev;
+			if (dev->bdev) {
+				if (bio->bi_rw & WRITE)
+					btrfs_dev_stat_inc(dev,
+						BTRFS_DEV_STAT_WRITE_ERRS);
+				else
+					btrfs_dev_stat_inc(dev,
+						BTRFS_DEV_STAT_READ_ERRS);
+				if ((bio->bi_rw & WRITE_FLUSH) == WRITE_FLUSH)
+					btrfs_dev_stat_inc(dev,
+						BTRFS_DEV_STAT_FLUSH_ERRS);
+				btrfs_dev_stat_print_on_error(dev);
+			}
+		}
+	}
 
-	if (bio == multi->orig_bio)
+	if (bio == bbio->orig_bio)
 		is_orig_bio = 1;
 
-	if (atomic_dec_and_test(&multi->stripes_pending)) {
+	btrfs_bio_counter_dec(bbio->fs_info);
+
+	if (atomic_dec_and_test(&bbio->stripes_pending)) {
 		if (!is_orig_bio) {
 			bio_put(bio);
-			bio = multi->orig_bio;
+			bio = bbio->orig_bio;
 		}
-		bio->bi_private = multi->private;
-		bio->bi_end_io = multi->end_io;
+
+		bio->bi_private = bbio->private;
+		bio->bi_end_io = bbio->end_io;
+		btrfs_io_bio(bio)->mirror_num = bbio->mirror_num;
 		/* only send an error to the higher layers if it is
-		 * beyond the tolerance of the multi-bio
+		 * beyond the tolerance of the btrfs bio
 		 */
-		if (atomic_read(&multi->error) > multi->max_errors) {
+		if (atomic_read(&bbio->error) > bbio->max_errors) {
 			err = -EIO;
-		} else if (err) {
+		} else {
 			/*
 			 * this bio is actually up to date, we didn't
 			 * go over the max number of errors
@@ -2912,21 +5444,13 @@ static void end_bio_multi_stripe(struct bio *bio, int err)
 			set_bit(BIO_UPTODATE, &bio->bi_flags);
 			err = 0;
 		}
-		kfree(multi);
 
-		bio_endio(bio, err);
+		btrfs_end_bbio(bbio, bio, err);
 	} else if (!is_orig_bio) {
 		bio_put(bio);
 	}
 }
 
-struct async_sched {
-	struct bio *bio;
-	int rw;
-	struct btrfs_fs_info *info;
-	struct btrfs_work work;
-};
-
 /*
  * see run_scheduled_bios for a description of why bios are collected for
  * async submit.
@@ -2934,19 +5458,24 @@ struct async_sched {
  * This will add one bio to the pending list for a device and make sure
  * the work struct is scheduled.
  */
-static noinline int schedule_bio(struct btrfs_root *root,
-				 struct btrfs_device *device,
-				 int rw, struct bio *bio)
+static noinline void btrfs_schedule_bio(struct btrfs_root *root,
+					struct btrfs_device *device,
+					int rw, struct bio *bio)
 {
 	int should_queue = 1;
 	struct btrfs_pending_bios *pending_bios;
 
+	if (device->missing || !device->bdev) {
+		bio_endio(bio, -EIO);
+		return;
+	}
+
 	/* don't bother with additional async steps for reads, right now */
-	if (!(rw & (1 << BIO_RW))) {
+	if (!(rw & REQ_WRITE)) {
 		bio_get(bio);
-		submit_bio(rw, bio);
+		btrfsic_submit_bio(rw, bio);
 		bio_put(bio);
-		return 0;
+		return;
 	}
 
 	/*
@@ -2961,7 +5490,7 @@ static noinline int schedule_bio(struct btrfs_root *root,
 	bio->bi_rw |= rw;
 
 	spin_lock(&device->io_lock);
-	if (bio_rw_flagged(bio, BIO_RW_SYNCIO))
+	if (bio->bi_rw & REQ_SYNC)
 		pending_bios = &device->pending_sync_bios;
 	else
 		pending_bios = &device->pending_bios;
@@ -2978,85 +5507,221 @@ static noinline int schedule_bio(struct btrfs_root *root,
 	spin_unlock(&device->io_lock);
 
 	if (should_queue)
-		btrfs_queue_worker(&root->fs_info->submit_workers,
-				   &device->work);
+		btrfs_queue_work(root->fs_info->submit_workers,
+				 &device->work);
+}
+
+static int bio_size_ok(struct block_device *bdev, struct bio *bio,
+		       sector_t sector)
+{
+	struct bio_vec *prev;
+	struct request_queue *q = bdev_get_queue(bdev);
+	unsigned int max_sectors = queue_max_sectors(q);
+	struct bvec_merge_data bvm = {
+		.bi_bdev = bdev,
+		.bi_sector = sector,
+		.bi_rw = bio->bi_rw,
+	};
+
+	if (WARN_ON(bio->bi_vcnt == 0))
+		return 1;
+
+	prev = &bio->bi_io_vec[bio->bi_vcnt - 1];
+	if (bio_sectors(bio) > max_sectors)
+		return 0;
+
+	if (!q->merge_bvec_fn)
+		return 1;
+
+	bvm.bi_size = bio->bi_iter.bi_size - prev->bv_len;
+	if (q->merge_bvec_fn(q, &bvm, prev) < prev->bv_len)
+		return 0;
+	return 1;
+}
+
+static void submit_stripe_bio(struct btrfs_root *root, struct btrfs_bio *bbio,
+			      struct bio *bio, u64 physical, int dev_nr,
+			      int rw, int async)
+{
+	struct btrfs_device *dev = bbio->stripes[dev_nr].dev;
+
+	bio->bi_private = bbio;
+	btrfs_io_bio(bio)->stripe_index = dev_nr;
+	bio->bi_end_io = btrfs_end_bio;
+	bio->bi_iter.bi_sector = physical >> 9;
+#ifdef DEBUG
+	{
+		struct rcu_string *name;
+
+		rcu_read_lock();
+		name = rcu_dereference(dev->name);
+		pr_debug("btrfs_map_bio: rw %d, sector=%llu, dev=%lu "
+			 "(%s id %llu), size=%u\n", rw,
+			 (u64)bio->bi_sector, (u_long)dev->bdev->bd_dev,
+			 name->str, dev->devid, bio->bi_size);
+		rcu_read_unlock();
+	}
+#endif
+	bio->bi_bdev = dev->bdev;
+
+	btrfs_bio_counter_inc_noblocked(root->fs_info);
+
+	if (async)
+		btrfs_schedule_bio(root, dev, rw, bio);
+	else
+		btrfsic_submit_bio(rw, bio);
+}
+
+static int breakup_stripe_bio(struct btrfs_root *root, struct btrfs_bio *bbio,
+			      struct bio *first_bio, struct btrfs_device *dev,
+			      int dev_nr, int rw, int async)
+{
+	struct bio_vec *bvec = first_bio->bi_io_vec;
+	struct bio *bio;
+	int nr_vecs = bio_get_nr_vecs(dev->bdev);
+	u64 physical = bbio->stripes[dev_nr].physical;
+
+again:
+	bio = btrfs_bio_alloc(dev->bdev, physical >> 9, nr_vecs, GFP_NOFS);
+	if (!bio)
+		return -ENOMEM;
+
+	while (bvec <= (first_bio->bi_io_vec + first_bio->bi_vcnt - 1)) {
+		if (bio_add_page(bio, bvec->bv_page, bvec->bv_len,
+				 bvec->bv_offset) < bvec->bv_len) {
+			u64 len = bio->bi_iter.bi_size;
+
+			atomic_inc(&bbio->stripes_pending);
+			submit_stripe_bio(root, bbio, bio, physical, dev_nr,
+					  rw, async);
+			physical += len;
+			goto again;
+		}
+		bvec++;
+	}
+
+	submit_stripe_bio(root, bbio, bio, physical, dev_nr, rw, async);
 	return 0;
 }
 
+static void bbio_error(struct btrfs_bio *bbio, struct bio *bio, u64 logical)
+{
+	atomic_inc(&bbio->error);
+	if (atomic_dec_and_test(&bbio->stripes_pending)) {
+		/* Shoud be the original bio. */
+		WARN_ON(bio != bbio->orig_bio);
+
+		bio->bi_private = bbio->private;
+		bio->bi_end_io = bbio->end_io;
+		btrfs_io_bio(bio)->mirror_num = bbio->mirror_num;
+		bio->bi_iter.bi_sector = logical >> 9;
+
+		btrfs_end_bbio(bbio, bio, -EIO);
+	}
+}
+
 int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio,
 		  int mirror_num, int async_submit)
 {
-	struct btrfs_mapping_tree *map_tree;
 	struct btrfs_device *dev;
 	struct bio *first_bio = bio;
-	u64 logical = (u64)bio->bi_sector << 9;
+	u64 logical = (u64)bio->bi_iter.bi_sector << 9;
 	u64 length = 0;
 	u64 map_length;
-	struct btrfs_multi_bio *multi = NULL;
+	u64 *raid_map = NULL;
 	int ret;
 	int dev_nr = 0;
 	int total_devs = 1;
+	struct btrfs_bio *bbio = NULL;
 
-	length = bio->bi_size;
-	map_tree = &root->fs_info->mapping_tree;
+	length = bio->bi_iter.bi_size;
 	map_length = length;
 
-	ret = btrfs_map_block(map_tree, rw, logical, &map_length, &multi,
-			      mirror_num);
-	BUG_ON(ret);
+	btrfs_bio_counter_inc_blocked(root->fs_info);
+	ret = __btrfs_map_block(root->fs_info, rw, logical, &map_length, &bbio,
+			      mirror_num, &raid_map);
+	if (ret) {
+		btrfs_bio_counter_dec(root->fs_info);
+		return ret;
+	}
+
+	total_devs = bbio->num_stripes;
+	bbio->orig_bio = first_bio;
+	bbio->private = first_bio->bi_private;
+	bbio->end_io = first_bio->bi_end_io;
+	bbio->fs_info = root->fs_info;
+	atomic_set(&bbio->stripes_pending, bbio->num_stripes);
+
+	if (raid_map) {
+		/* In this case, map_length has been set to the length of
+		   a single stripe; not the whole write */
+		if (rw & WRITE) {
+			ret = raid56_parity_write(root, bio, bbio,
+						  raid_map, map_length);
+		} else {
+			ret = raid56_parity_recover(root, bio, bbio,
+						    raid_map, map_length,
+						    mirror_num);
+		}
+		/*
+		 * FIXME, replace dosen't support raid56 yet, please fix
+		 * it in the future.
+		 */
+		btrfs_bio_counter_dec(root->fs_info);
+		return ret;
+	}
 
-	total_devs = multi->num_stripes;
 	if (map_length < length) {
-		printk(KERN_CRIT "mapping failed logical %llu bio len %llu "
-		       "len %llu\n", (unsigned long long)logical,
-		       (unsigned long long)length,
-		       (unsigned long long)map_length);
+		btrfs_crit(root->fs_info, "mapping failed logical %llu bio len %llu len %llu",
+			logical, length, map_length);
 		BUG();
 	}
-	multi->end_io = first_bio->bi_end_io;
-	multi->private = first_bio->bi_private;
-	multi->orig_bio = first_bio;
-	atomic_set(&multi->stripes_pending, multi->num_stripes);
 
 	while (dev_nr < total_devs) {
-		if (total_devs > 1) {
-			if (dev_nr < total_devs - 1) {
-				bio = bio_clone(first_bio, GFP_NOFS);
-				BUG_ON(!bio);
-			} else {
-				bio = first_bio;
-			}
-			bio->bi_private = multi;
-			bio->bi_end_io = end_bio_multi_stripe;
-		}
-		bio->bi_sector = multi->stripes[dev_nr].physical >> 9;
-		dev = multi->stripes[dev_nr].dev;
-		BUG_ON(rw == WRITE && !dev->writeable);
-		if (dev && dev->bdev) {
-			bio->bi_bdev = dev->bdev;
-			if (async_submit)
-				schedule_bio(root, dev, rw, bio);
-			else
-				submit_bio(rw, bio);
+		dev = bbio->stripes[dev_nr].dev;
+		if (!dev || !dev->bdev || (rw & WRITE && !dev->writeable)) {
+			bbio_error(bbio, first_bio, logical);
+			dev_nr++;
+			continue;
+		}
+
+		/*
+		 * Check and see if we're ok with this bio based on it's size
+		 * and offset with the given device.
+		 */
+		if (!bio_size_ok(dev->bdev, first_bio,
+				 bbio->stripes[dev_nr].physical >> 9)) {
+			ret = breakup_stripe_bio(root, bbio, first_bio, dev,
+						 dev_nr, rw, async_submit);
+			BUG_ON(ret);
+			dev_nr++;
+			continue;
+		}
+
+		if (dev_nr < total_devs - 1) {
+			bio = btrfs_bio_clone(first_bio, GFP_NOFS);
+			BUG_ON(!bio); /* -ENOMEM */
 		} else {
-			bio->bi_bdev = root->fs_info->fs_devices->latest_bdev;
-			bio->bi_sector = logical >> 9;
-			bio_endio(bio, -EIO);
+			bio = first_bio;
+			bbio->flags |= BTRFS_BIO_ORIG_BIO_SUBMITTED;
 		}
+
+		submit_stripe_bio(root, bbio, bio,
+				  bbio->stripes[dev_nr].physical, dev_nr, rw,
+				  async_submit);
 		dev_nr++;
 	}
-	if (total_devs == 1)
-		kfree(multi);
+	btrfs_bio_counter_dec(root->fs_info);
 	return 0;
 }
 
-struct btrfs_device *btrfs_find_device(struct btrfs_root *root, u64 devid,
+struct btrfs_device *btrfs_find_device(struct btrfs_fs_info *fs_info, u64 devid,
 				       u8 *uuid, u8 *fsid)
 {
 	struct btrfs_device *device;
 	struct btrfs_fs_devices *cur_devices;
 
-	cur_devices = root->fs_info->fs_devices;
+	cur_devices = fs_info->fs_devices;
 	while (cur_devices) {
 		if (!fsid ||
 		    !memcmp(cur_devices->fsid, fsid, BTRFS_UUID_SIZE)) {
@@ -3076,23 +5741,70 @@ static struct btrfs_device *add_missing_dev(struct btrfs_root *root,
 	struct btrfs_device *device;
 	struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
 
-	device = kzalloc(sizeof(*device), GFP_NOFS);
-	if (!device)
+	device = btrfs_alloc_device(NULL, &devid, dev_uuid);
+	if (IS_ERR(device))
 		return NULL;
-	list_add(&device->dev_list,
-		 &fs_devices->devices);
-	device->barriers = 1;
-	device->dev_root = root->fs_info->dev_root;
-	device->devid = devid;
-	device->work.func = pending_bios_fn;
+
+	list_add(&device->dev_list, &fs_devices->devices);
 	device->fs_devices = fs_devices;
 	fs_devices->num_devices++;
-	spin_lock_init(&device->io_lock);
-	INIT_LIST_HEAD(&device->dev_alloc_list);
-	memcpy(device->uuid, dev_uuid, BTRFS_UUID_SIZE);
+
+	device->missing = 1;
+	fs_devices->missing_devices++;
+
 	return device;
 }
 
+/**
+ * btrfs_alloc_device - allocate struct btrfs_device
+ * @fs_info:	used only for generating a new devid, can be NULL if
+ *		devid is provided (i.e. @devid != NULL).
+ * @devid:	a pointer to devid for this device.  If NULL a new devid
+ *		is generated.
+ * @uuid:	a pointer to UUID for this device.  If NULL a new UUID
+ *		is generated.
+ *
+ * Return: a pointer to a new &struct btrfs_device on success; ERR_PTR()
+ * on error.  Returned struct is not linked onto any lists and can be
+ * destroyed with kfree() right away.
+ */
+struct btrfs_device *btrfs_alloc_device(struct btrfs_fs_info *fs_info,
+					const u64 *devid,
+					const u8 *uuid)
+{
+	struct btrfs_device *dev;
+	u64 tmp;
+
+	if (WARN_ON(!devid && !fs_info))
+		return ERR_PTR(-EINVAL);
+
+	dev = __alloc_device();
+	if (IS_ERR(dev))
+		return dev;
+
+	if (devid)
+		tmp = *devid;
+	else {
+		int ret;
+
+		ret = find_next_devid(fs_info, &tmp);
+		if (ret) {
+			kfree(dev);
+			return ERR_PTR(ret);
+		}
+	}
+	dev->devid = tmp;
+
+	if (uuid)
+		memcpy(dev->uuid, uuid, BTRFS_UUID_SIZE);
+	else
+		generate_random_uuid(dev->uuid);
+
+	btrfs_init_work(&dev->work, pending_bios_fn, NULL, NULL);
+
+	return dev;
+}
+
 static int read_one_chunk(struct btrfs_root *root, struct btrfs_key *key,
 			  struct extent_buffer *leaf,
 			  struct btrfs_chunk *chunk)
@@ -3123,7 +5835,7 @@ static int read_one_chunk(struct btrfs_root *root, struct btrfs_key *key,
 		free_extent_map(em);
 	}
 
-	em = alloc_extent_map(GFP_NOFS);
+	em = alloc_extent_map();
 	if (!em)
 		return -ENOMEM;
 	num_stripes = btrfs_chunk_num_stripes(leaf, chunk);
@@ -3133,9 +5845,11 @@ static int read_one_chunk(struct btrfs_root *root, struct btrfs_key *key,
 		return -ENOMEM;
 	}
 
+	set_bit(EXTENT_FLAG_FS_MAPPING, &em->flags);
 	em->bdev = (struct block_device *)map;
 	em->start = logical;
 	em->len = length;
+	em->orig_start = 0;
 	em->block_start = 0;
 	em->block_len = em->len;
 
@@ -3153,10 +5867,9 @@ static int read_one_chunk(struct btrfs_root *root, struct btrfs_key *key,
 		read_extent_buffer(leaf, uuid, (unsigned long)
 				   btrfs_stripe_dev_uuid_nr(chunk, i),
 				   BTRFS_UUID_SIZE);
-		map->stripes[i].dev = btrfs_find_device(root, devid, uuid,
-							NULL);
+		map->stripes[i].dev = btrfs_find_device(root->fs_info, devid,
+							uuid, NULL);
 		if (!map->stripes[i].dev && !btrfs_test_opt(root, DEGRADED)) {
-			kfree(map);
 			free_extent_map(em);
 			return -EIO;
 		}
@@ -3164,7 +5877,6 @@ static int read_one_chunk(struct btrfs_root *root, struct btrfs_key *key,
 			map->stripes[i].dev =
 				add_missing_dev(root, devid, uuid);
 			if (!map->stripes[i].dev) {
-				kfree(map);
 				free_extent_map(em);
 				return -EIO;
 			}
@@ -3173,15 +5885,15 @@ static int read_one_chunk(struct btrfs_root *root, struct btrfs_key *key,
 	}
 
 	write_lock(&map_tree->map_tree.lock);
-	ret = add_extent_mapping(&map_tree->map_tree, em);
+	ret = add_extent_mapping(&map_tree->map_tree, em, 0);
 	write_unlock(&map_tree->map_tree.lock);
-	BUG_ON(ret);
+	BUG_ON(ret); /* Tree corruption */
 	free_extent_map(em);
 
 	return 0;
 }
 
-static int fill_device_from_item(struct extent_buffer *leaf,
+static void fill_device_from_item(struct extent_buffer *leaf,
 				 struct btrfs_dev_item *dev_item,
 				 struct btrfs_device *device)
 {
@@ -3195,11 +5907,11 @@ static int fill_device_from_item(struct extent_buffer *leaf,
 	device->io_align = btrfs_device_io_align(leaf, dev_item);
 	device->io_width = btrfs_device_io_width(leaf, dev_item);
 	device->sector_size = btrfs_device_sector_size(leaf, dev_item);
+	WARN_ON(device->devid == BTRFS_DEV_REPLACE_DEVID);
+	device->is_tgtdev_for_dev_replace = 0;
 
-	ptr = (unsigned long)btrfs_device_uuid(dev_item);
+	ptr = btrfs_device_uuid(dev_item);
 	read_extent_buffer(leaf, device->uuid, ptr, BTRFS_UUID_SIZE);
-
-	return 0;
 }
 
 static int open_seed_devices(struct btrfs_root *root, u8 *fsid)
@@ -3207,7 +5919,7 @@ static int open_seed_devices(struct btrfs_root *root, u8 *fsid)
 	struct btrfs_fs_devices *fs_devices;
 	int ret;
 
-	mutex_lock(&uuid_mutex);
+	BUG_ON(!mutex_is_locked(&uuid_mutex));
 
 	fs_devices = root->fs_info->fs_devices->seed;
 	while (fs_devices) {
@@ -3232,8 +5944,10 @@ static int open_seed_devices(struct btrfs_root *root, u8 *fsid)
 
 	ret = __btrfs_open_devices(fs_devices, FMODE_READ,
 				   root->fs_info->bdev_holder);
-	if (ret)
+	if (ret) {
+		free_fs_devices(fs_devices);
 		goto out;
+	}
 
 	if (!fs_devices->seeding) {
 		__btrfs_close_devices(fs_devices);
@@ -3245,7 +5959,6 @@ static int open_seed_devices(struct btrfs_root *root, u8 *fsid)
 	fs_devices->seed = root->fs_info->fs_devices->seed;
 	root->fs_info->fs_devices->seed = fs_devices;
 out:
-	mutex_unlock(&uuid_mutex);
 	return ret;
 }
 
@@ -3260,11 +5973,9 @@ static int read_one_dev(struct btrfs_root *root,
 	u8 dev_uuid[BTRFS_UUID_SIZE];
 
 	devid = btrfs_device_id(leaf, dev_item);
-	read_extent_buffer(leaf, dev_uuid,
-			   (unsigned long)btrfs_device_uuid(dev_item),
+	read_extent_buffer(leaf, dev_uuid, btrfs_device_uuid(dev_item),
 			   BTRFS_UUID_SIZE);
-	read_extent_buffer(leaf, fs_uuid,
-			   (unsigned long)btrfs_device_fsid(dev_item),
+	read_extent_buffer(leaf, fs_uuid, btrfs_device_fsid(dev_item),
 			   BTRFS_UUID_SIZE);
 
 	if (memcmp(fs_uuid, root->fs_info->fsid, BTRFS_UUID_SIZE)) {
@@ -3273,17 +5984,25 @@ static int read_one_dev(struct btrfs_root *root,
 			return ret;
 	}
 
-	device = btrfs_find_device(root, devid, dev_uuid, fs_uuid);
+	device = btrfs_find_device(root->fs_info, devid, dev_uuid, fs_uuid);
 	if (!device || !device->bdev) {
 		if (!btrfs_test_opt(root, DEGRADED))
 			return -EIO;
 
 		if (!device) {
-			printk(KERN_WARNING "warning devid %llu missing\n",
-			       (unsigned long long)devid);
+			btrfs_warn(root->fs_info, "devid %llu missing", devid);
 			device = add_missing_dev(root, devid, dev_uuid);
 			if (!device)
 				return -ENOMEM;
+		} else if (!device->missing) {
+			/*
+			 * this happens when a device that was properly setup
+			 * in the device info lists suddenly goes bad.
+			 * device->bdev is NULL, and so we have to set
+			 * device->missing to one here
+			 */
+			root->fs_info->fs_devices->missing_devices++;
+			device->missing = 1;
 		}
 	}
 
@@ -3295,26 +6014,21 @@ static int read_one_dev(struct btrfs_root *root,
 	}
 
 	fill_device_from_item(leaf, dev_item, device);
-	device->dev_root = root->fs_info->dev_root;
 	device->in_fs_metadata = 1;
-	if (device->writeable)
+	if (device->writeable && !device->is_tgtdev_for_dev_replace) {
 		device->fs_devices->total_rw_bytes += device->total_bytes;
+		spin_lock(&root->fs_info->free_chunk_lock);
+		root->fs_info->free_chunk_space += device->total_bytes -
+			device->bytes_used;
+		spin_unlock(&root->fs_info->free_chunk_lock);
+	}
 	ret = 0;
 	return ret;
 }
 
-int btrfs_read_super_device(struct btrfs_root *root, struct extent_buffer *buf)
-{
-	struct btrfs_dev_item *dev_item;
-
-	dev_item = (struct btrfs_dev_item *)offsetof(struct btrfs_super_block,
-						     dev_item);
-	return read_one_dev(root, buf, dev_item);
-}
-
 int btrfs_read_sys_array(struct btrfs_root *root)
 {
-	struct btrfs_super_block *super_copy = &root->fs_info->super_copy;
+	struct btrfs_super_block *super_copy = root->fs_info->super_copy;
 	struct extent_buffer *sb;
 	struct btrfs_disk_key *disk_key;
 	struct btrfs_chunk *chunk;
@@ -3332,7 +6046,21 @@ int btrfs_read_sys_array(struct btrfs_root *root)
 	if (!sb)
 		return -ENOMEM;
 	btrfs_set_buffer_uptodate(sb);
-	btrfs_set_buffer_lockdep_class(sb, 0);
+	btrfs_set_buffer_lockdep_class(root->root_key.objectid, sb, 0);
+	/*
+	 * The sb extent buffer is artifical and just used to read the system array.
+	 * btrfs_set_buffer_uptodate() call does not properly mark all it's
+	 * pages up-to-date when the page is larger: extent does not cover the
+	 * whole page and consequently check_page_uptodate does not find all
+	 * the page's extents up-to-date (the hole beyond sb),
+	 * write_extent_buffer then triggers a WARN_ON.
+	 *
+	 * Regular short extents go through mark_extent_buffer_dirty/writeback cycle,
+	 * but sb spans only this function. Add an explicit SetPageUptodate call
+	 * to silence the warning eg. on PowerPC 64.
+	 */
+	if (PAGE_CACHE_SIZE > BTRFS_SUPER_INFO_SIZE)
+		SetPageUptodate(sb->pages[0]);
 
 	write_extent_buffer(sb, super_copy, 0, BTRFS_SUPER_INFO_SIZE);
 	array_size = btrfs_super_sys_array_size(super_copy);
@@ -3383,14 +6111,18 @@ int btrfs_read_chunk_tree(struct btrfs_root *root)
 	if (!path)
 		return -ENOMEM;
 
-	/* first we search for all of the device items, and then we
-	 * read in all of the chunk items.  This way we can create chunk
-	 * mappings that reference all of the devices that are afound
+	mutex_lock(&uuid_mutex);
+	lock_chunks(root);
+
+	/*
+	 * Read all device items, and then all the chunk items. All
+	 * device items are found before any chunk item (their object id
+	 * is smaller than the lowest possible object id for a chunk
+	 * item - BTRFS_FIRST_CHUNK_TREE_OBJECTID).
 	 */
 	key.objectid = BTRFS_DEV_ITEMS_OBJECTID;
 	key.offset = 0;
 	key.type = 0;
-again:
 	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
 	if (ret < 0)
 		goto error;
@@ -3406,17 +6138,13 @@ again:
 			break;
 		}
 		btrfs_item_key_to_cpu(leaf, &found_key, slot);
-		if (key.objectid == BTRFS_DEV_ITEMS_OBJECTID) {
-			if (found_key.objectid != BTRFS_DEV_ITEMS_OBJECTID)
-				break;
-			if (found_key.type == BTRFS_DEV_ITEM_KEY) {
-				struct btrfs_dev_item *dev_item;
-				dev_item = btrfs_item_ptr(leaf, slot,
+		if (found_key.type == BTRFS_DEV_ITEM_KEY) {
+			struct btrfs_dev_item *dev_item;
+			dev_item = btrfs_item_ptr(leaf, slot,
 						  struct btrfs_dev_item);
-				ret = read_one_dev(root, leaf, dev_item);
-				if (ret)
-					goto error;
-			}
+			ret = read_one_dev(root, leaf, dev_item);
+			if (ret)
+				goto error;
 		} else if (found_key.type == BTRFS_CHUNK_ITEM_KEY) {
 			struct btrfs_chunk *chunk;
 			chunk = btrfs_item_ptr(leaf, slot, struct btrfs_chunk);
@@ -3426,13 +6154,276 @@ again:
 		}
 		path->slots[0]++;
 	}
-	if (key.objectid == BTRFS_DEV_ITEMS_OBJECTID) {
-		key.objectid = 0;
-		btrfs_release_path(root, path);
-		goto again;
-	}
 	ret = 0;
 error:
+	unlock_chunks(root);
+	mutex_unlock(&uuid_mutex);
+
 	btrfs_free_path(path);
 	return ret;
 }
+
+void btrfs_init_devices_late(struct btrfs_fs_info *fs_info)
+{
+	struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
+	struct btrfs_device *device;
+
+	while (fs_devices) {
+		mutex_lock(&fs_devices->device_list_mutex);
+		list_for_each_entry(device, &fs_devices->devices, dev_list)
+			device->dev_root = fs_info->dev_root;
+		mutex_unlock(&fs_devices->device_list_mutex);
+
+		fs_devices = fs_devices->seed;
+	}
+}
+
+static void __btrfs_reset_dev_stats(struct btrfs_device *dev)
+{
+	int i;
+
+	for (i = 0; i < BTRFS_DEV_STAT_VALUES_MAX; i++)
+		btrfs_dev_stat_reset(dev, i);
+}
+
+int btrfs_init_dev_stats(struct btrfs_fs_info *fs_info)
+{
+	struct btrfs_key key;
+	struct btrfs_key found_key;
+	struct btrfs_root *dev_root = fs_info->dev_root;
+	struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
+	struct extent_buffer *eb;
+	int slot;
+	int ret = 0;
+	struct btrfs_device *device;
+	struct btrfs_path *path = NULL;
+	int i;
+
+	path = btrfs_alloc_path();
+	if (!path) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	mutex_lock(&fs_devices->device_list_mutex);
+	list_for_each_entry(device, &fs_devices->devices, dev_list) {
+		int item_size;
+		struct btrfs_dev_stats_item *ptr;
+
+		key.objectid = 0;
+		key.type = BTRFS_DEV_STATS_KEY;
+		key.offset = device->devid;
+		ret = btrfs_search_slot(NULL, dev_root, &key, path, 0, 0);
+		if (ret) {
+			__btrfs_reset_dev_stats(device);
+			device->dev_stats_valid = 1;
+			btrfs_release_path(path);
+			continue;
+		}
+		slot = path->slots[0];
+		eb = path->nodes[0];
+		btrfs_item_key_to_cpu(eb, &found_key, slot);
+		item_size = btrfs_item_size_nr(eb, slot);
+
+		ptr = btrfs_item_ptr(eb, slot,
+				     struct btrfs_dev_stats_item);
+
+		for (i = 0; i < BTRFS_DEV_STAT_VALUES_MAX; i++) {
+			if (item_size >= (1 + i) * sizeof(__le64))
+				btrfs_dev_stat_set(device, i,
+					btrfs_dev_stats_value(eb, ptr, i));
+			else
+				btrfs_dev_stat_reset(device, i);
+		}
+
+		device->dev_stats_valid = 1;
+		btrfs_dev_stat_print_on_load(device);
+		btrfs_release_path(path);
+	}
+	mutex_unlock(&fs_devices->device_list_mutex);
+
+out:
+	btrfs_free_path(path);
+	return ret < 0 ? ret : 0;
+}
+
+static int update_dev_stat_item(struct btrfs_trans_handle *trans,
+				struct btrfs_root *dev_root,
+				struct btrfs_device *device)
+{
+	struct btrfs_path *path;
+	struct btrfs_key key;
+	struct extent_buffer *eb;
+	struct btrfs_dev_stats_item *ptr;
+	int ret;
+	int i;
+
+	key.objectid = 0;
+	key.type = BTRFS_DEV_STATS_KEY;
+	key.offset = device->devid;
+
+	path = btrfs_alloc_path();
+	BUG_ON(!path);
+	ret = btrfs_search_slot(trans, dev_root, &key, path, -1, 1);
+	if (ret < 0) {
+		printk_in_rcu(KERN_WARNING "BTRFS: "
+			"error %d while searching for dev_stats item for device %s!\n",
+			      ret, rcu_str_deref(device->name));
+		goto out;
+	}
+
+	if (ret == 0 &&
+	    btrfs_item_size_nr(path->nodes[0], path->slots[0]) < sizeof(*ptr)) {
+		/* need to delete old one and insert a new one */
+		ret = btrfs_del_item(trans, dev_root, path);
+		if (ret != 0) {
+			printk_in_rcu(KERN_WARNING "BTRFS: "
+				"delete too small dev_stats item for device %s failed %d!\n",
+				      rcu_str_deref(device->name), ret);
+			goto out;
+		}
+		ret = 1;
+	}
+
+	if (ret == 1) {
+		/* need to insert a new item */
+		btrfs_release_path(path);
+		ret = btrfs_insert_empty_item(trans, dev_root, path,
+					      &key, sizeof(*ptr));
+		if (ret < 0) {
+			printk_in_rcu(KERN_WARNING "BTRFS: "
+					  "insert dev_stats item for device %s failed %d!\n",
+				      rcu_str_deref(device->name), ret);
+			goto out;
+		}
+	}
+
+	eb = path->nodes[0];
+	ptr = btrfs_item_ptr(eb, path->slots[0], struct btrfs_dev_stats_item);
+	for (i = 0; i < BTRFS_DEV_STAT_VALUES_MAX; i++)
+		btrfs_set_dev_stats_value(eb, ptr, i,
+					  btrfs_dev_stat_read(device, i));
+	btrfs_mark_buffer_dirty(eb);
+
+out:
+	btrfs_free_path(path);
+	return ret;
+}
+
+/*
+ * called from commit_transaction. Writes all changed device stats to disk.
+ */
+int btrfs_run_dev_stats(struct btrfs_trans_handle *trans,
+			struct btrfs_fs_info *fs_info)
+{
+	struct btrfs_root *dev_root = fs_info->dev_root;
+	struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
+	struct btrfs_device *device;
+	int ret = 0;
+
+	mutex_lock(&fs_devices->device_list_mutex);
+	list_for_each_entry(device, &fs_devices->devices, dev_list) {
+		if (!device->dev_stats_valid || !device->dev_stats_dirty)
+			continue;
+
+		ret = update_dev_stat_item(trans, dev_root, device);
+		if (!ret)
+			device->dev_stats_dirty = 0;
+	}
+	mutex_unlock(&fs_devices->device_list_mutex);
+
+	return ret;
+}
+
+void btrfs_dev_stat_inc_and_print(struct btrfs_device *dev, int index)
+{
+	btrfs_dev_stat_inc(dev, index);
+	btrfs_dev_stat_print_on_error(dev);
+}
+
+static void btrfs_dev_stat_print_on_error(struct btrfs_device *dev)
+{
+	if (!dev->dev_stats_valid)
+		return;
+	printk_ratelimited_in_rcu(KERN_ERR "BTRFS: "
+			   "bdev %s errs: wr %u, rd %u, flush %u, corrupt %u, gen %u\n",
+			   rcu_str_deref(dev->name),
+			   btrfs_dev_stat_read(dev, BTRFS_DEV_STAT_WRITE_ERRS),
+			   btrfs_dev_stat_read(dev, BTRFS_DEV_STAT_READ_ERRS),
+			   btrfs_dev_stat_read(dev, BTRFS_DEV_STAT_FLUSH_ERRS),
+			   btrfs_dev_stat_read(dev, BTRFS_DEV_STAT_CORRUPTION_ERRS),
+			   btrfs_dev_stat_read(dev, BTRFS_DEV_STAT_GENERATION_ERRS));
+}
+
+static void btrfs_dev_stat_print_on_load(struct btrfs_device *dev)
+{
+	int i;
+
+	for (i = 0; i < BTRFS_DEV_STAT_VALUES_MAX; i++)
+		if (btrfs_dev_stat_read(dev, i) != 0)
+			break;
+	if (i == BTRFS_DEV_STAT_VALUES_MAX)
+		return; /* all values == 0, suppress message */
+
+	printk_in_rcu(KERN_INFO "BTRFS: "
+		   "bdev %s errs: wr %u, rd %u, flush %u, corrupt %u, gen %u\n",
+	       rcu_str_deref(dev->name),
+	       btrfs_dev_stat_read(dev, BTRFS_DEV_STAT_WRITE_ERRS),
+	       btrfs_dev_stat_read(dev, BTRFS_DEV_STAT_READ_ERRS),
+	       btrfs_dev_stat_read(dev, BTRFS_DEV_STAT_FLUSH_ERRS),
+	       btrfs_dev_stat_read(dev, BTRFS_DEV_STAT_CORRUPTION_ERRS),
+	       btrfs_dev_stat_read(dev, BTRFS_DEV_STAT_GENERATION_ERRS));
+}
+
+int btrfs_get_dev_stats(struct btrfs_root *root,
+			struct btrfs_ioctl_get_dev_stats *stats)
+{
+	struct btrfs_device *dev;
+	struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
+	int i;
+
+	mutex_lock(&fs_devices->device_list_mutex);
+	dev = btrfs_find_device(root->fs_info, stats->devid, NULL, NULL);
+	mutex_unlock(&fs_devices->device_list_mutex);
+
+	if (!dev) {
+		btrfs_warn(root->fs_info, "get dev_stats failed, device not found");
+		return -ENODEV;
+	} else if (!dev->dev_stats_valid) {
+		btrfs_warn(root->fs_info, "get dev_stats failed, not yet valid");
+		return -ENODEV;
+	} else if (stats->flags & BTRFS_DEV_STATS_RESET) {
+		for (i = 0; i < BTRFS_DEV_STAT_VALUES_MAX; i++) {
+			if (stats->nr_items > i)
+				stats->values[i] =
+					btrfs_dev_stat_read_and_reset(dev, i);
+			else
+				btrfs_dev_stat_reset(dev, i);
+		}
+	} else {
+		for (i = 0; i < BTRFS_DEV_STAT_VALUES_MAX; i++)
+			if (stats->nr_items > i)
+				stats->values[i] = btrfs_dev_stat_read(dev, i);
+	}
+	if (stats->nr_items > BTRFS_DEV_STAT_VALUES_MAX)
+		stats->nr_items = BTRFS_DEV_STAT_VALUES_MAX;
+	return 0;
+}
+
+int btrfs_scratch_superblock(struct btrfs_device *device)
+{
+	struct buffer_head *bh;
+	struct btrfs_super_block *disk_super;
+
+	bh = btrfs_read_dev_super(device->bdev);
+	if (!bh)
+		return -EINVAL;
+	disk_super = (struct btrfs_super_block *)bh->b_data;
+
+	memset(&disk_super->magic, 0, sizeof(disk_super->magic));
+	set_buffer_dirty(bh);
+	sync_dirty_buffer(bh);
+	brelse(bh);
+
+	return 0;
+}
diff --git a/fs/btrfs/volumes.h b/fs/btrfs/volumes.h
index 31b0fabdd2e..2aaa00c4781 100644
--- a/fs/btrfs/volumes.h
+++ b/fs/btrfs/volumes.h
@@ -20,8 +20,12 @@
 #define __BTRFS_VOLUMES_
 
 #include <linux/bio.h>
+#include <linux/sort.h>
+#include <linux/btrfs.h>
 #include "async-thread.h"
 
+#define BTRFS_STRIPE_LEN	(64 * 1024)
+
 struct buffer_head;
 struct btrfs_pending_bios {
 	struct bio *head;
@@ -39,21 +43,22 @@ struct btrfs_device {
 	/* WRITE_SYNC bios */
 	struct btrfs_pending_bios pending_sync_bios;
 
-	int running_pending;
 	u64 generation;
-
-	int barriers;
+	int running_pending;
 	int writeable;
 	int in_fs_metadata;
+	int missing;
+	int can_discard;
+	int is_tgtdev_for_dev_replace;
 
 	spinlock_t io_lock;
+	/* the mode sent to blkdev_get */
+	fmode_t mode;
 
 	struct block_device *bdev;
 
-	/* the mode sent to open_bdev_exclusive */
-	fmode_t mode;
 
-	char *name;
+	struct rcu_string *name;
 
 	/* the internal btrfs device id */
 	u64 devid;
@@ -72,17 +77,42 @@ struct btrfs_device {
 
 	/* optimal io width for this device */
 	u32 io_width;
+	/* type and info about this device */
+	u64 type;
 
 	/* minimal io size for this device */
 	u32 sector_size;
 
-	/* type and info about this device */
-	u64 type;
 
 	/* physical drive uuid (or lvm uuid) */
 	u8 uuid[BTRFS_UUID_SIZE];
 
+	/* for sending down flush barriers */
+	int nobarriers;
+	struct bio *flush_bio;
+	struct completion flush_wait;
+
+	/* per-device scrub information */
+	struct scrub_ctx *scrub_device;
+
 	struct btrfs_work work;
+	struct rcu_head rcu;
+	struct work_struct rcu_work;
+
+	/* readahead state */
+	spinlock_t reada_lock;
+	atomic_t reada_in_flight;
+	u64 reada_next;
+	struct reada_zone *reada_curr_zone;
+	struct radix_tree_root reada_zones;
+	struct radix_tree_root reada_extents;
+
+
+	/* disk I/O failure stats. For detailed description refer to
+	 * enum btrfs_dev_stat_values in ioctl.h */
+	int dev_stats_valid;
+	int dev_stats_dirty; /* counters need to be written to disk */
+	atomic_t dev_stat_values[BTRFS_DEV_STAT_VALUES_MAX];
 };
 
 struct btrfs_fs_devices {
@@ -94,12 +124,17 @@ struct btrfs_fs_devices {
 	u64 num_devices;
 	u64 open_devices;
 	u64 rw_devices;
+	u64 missing_devices;
 	u64 total_rw_bytes;
+	u64 num_can_discard;
+	u64 total_devices;
 	struct block_device *latest_bdev;
 
 	/* all of the devices in the FS, protected by a mutex
 	 * so we can safely walk it to write out the supers without
-	 * worrying about add/remove by the multi-device code
+	 * worrying about add/remove by the multi-device code.
+	 * Scrubbing super can kick off supers writing by holding
+	 * this mutex lock.
 	 */
 	struct mutex device_list_mutex;
 	struct list_head devices;
@@ -119,32 +154,143 @@ struct btrfs_fs_devices {
 	int rotating;
 };
 
+#define BTRFS_BIO_INLINE_CSUM_SIZE	64
+
+/*
+ * we need the mirror number and stripe index to be passed around
+ * the call chain while we are processing end_io (especially errors).
+ * Really, what we need is a btrfs_bio structure that has this info
+ * and is properly sized with its stripe array, but we're not there
+ * quite yet.  We have our own btrfs bioset, and all of the bios
+ * we allocate are actually btrfs_io_bios.  We'll cram as much of
+ * struct btrfs_bio as we can into this over time.
+ */
+typedef void (btrfs_io_bio_end_io_t) (struct btrfs_io_bio *bio, int err);
+struct btrfs_io_bio {
+	unsigned long mirror_num;
+	unsigned long stripe_index;
+	u8 *csum;
+	u8 csum_inline[BTRFS_BIO_INLINE_CSUM_SIZE];
+	u8 *csum_allocated;
+	btrfs_io_bio_end_io_t *end_io;
+	struct bio bio;
+};
+
+static inline struct btrfs_io_bio *btrfs_io_bio(struct bio *bio)
+{
+	return container_of(bio, struct btrfs_io_bio, bio);
+}
+
 struct btrfs_bio_stripe {
 	struct btrfs_device *dev;
 	u64 physical;
+	u64 length; /* only used for discard mappings */
 };
 
-struct btrfs_multi_bio {
+struct btrfs_bio;
+typedef void (btrfs_bio_end_io_t) (struct btrfs_bio *bio, int err);
+
+#define BTRFS_BIO_ORIG_BIO_SUBMITTED	0x1
+
+struct btrfs_bio {
 	atomic_t stripes_pending;
+	struct btrfs_fs_info *fs_info;
 	bio_end_io_t *end_io;
 	struct bio *orig_bio;
+	unsigned long flags;
 	void *private;
 	atomic_t error;
 	int max_errors;
 	int num_stripes;
+	int mirror_num;
 	struct btrfs_bio_stripe stripes[];
 };
 
-#define btrfs_multi_bio_size(n) (sizeof(struct btrfs_multi_bio) + \
+struct btrfs_device_info {
+	struct btrfs_device *dev;
+	u64 dev_offset;
+	u64 max_avail;
+	u64 total_avail;
+};
+
+struct btrfs_raid_attr {
+	int sub_stripes;	/* sub_stripes info for map */
+	int dev_stripes;	/* stripes per dev */
+	int devs_max;		/* max devs to use */
+	int devs_min;		/* min devs needed */
+	int devs_increment;	/* ndevs has to be a multiple of this */
+	int ncopies;		/* how many copies to data has */
+};
+
+struct map_lookup {
+	u64 type;
+	int io_align;
+	int io_width;
+	int stripe_len;
+	int sector_size;
+	int num_stripes;
+	int sub_stripes;
+	struct btrfs_bio_stripe stripes[];
+};
+
+#define map_lookup_size(n) (sizeof(struct map_lookup) + \
+			    (sizeof(struct btrfs_bio_stripe) * (n)))
+
+/*
+ * Restriper's general type filter
+ */
+#define BTRFS_BALANCE_DATA		(1ULL << 0)
+#define BTRFS_BALANCE_SYSTEM		(1ULL << 1)
+#define BTRFS_BALANCE_METADATA		(1ULL << 2)
+
+#define BTRFS_BALANCE_TYPE_MASK		(BTRFS_BALANCE_DATA |	    \
+					 BTRFS_BALANCE_SYSTEM |	    \
+					 BTRFS_BALANCE_METADATA)
+
+#define BTRFS_BALANCE_FORCE		(1ULL << 3)
+#define BTRFS_BALANCE_RESUME		(1ULL << 4)
+
+/*
+ * Balance filters
+ */
+#define BTRFS_BALANCE_ARGS_PROFILES	(1ULL << 0)
+#define BTRFS_BALANCE_ARGS_USAGE	(1ULL << 1)
+#define BTRFS_BALANCE_ARGS_DEVID	(1ULL << 2)
+#define BTRFS_BALANCE_ARGS_DRANGE	(1ULL << 3)
+#define BTRFS_BALANCE_ARGS_VRANGE	(1ULL << 4)
+#define BTRFS_BALANCE_ARGS_LIMIT	(1ULL << 5)
+
+/*
+ * Profile changing flags.  When SOFT is set we won't relocate chunk if
+ * it already has the target profile (even though it may be
+ * half-filled).
+ */
+#define BTRFS_BALANCE_ARGS_CONVERT	(1ULL << 8)
+#define BTRFS_BALANCE_ARGS_SOFT		(1ULL << 9)
+
+struct btrfs_balance_args;
+struct btrfs_balance_progress;
+struct btrfs_balance_control {
+	struct btrfs_fs_info *fs_info;
+
+	struct btrfs_balance_args data;
+	struct btrfs_balance_args meta;
+	struct btrfs_balance_args sys;
+
+	u64 flags;
+
+	struct btrfs_balance_progress stat;
+};
+
+int btrfs_account_dev_extents_size(struct btrfs_device *device, u64 start,
+				   u64 end, u64 *length);
+
+#define btrfs_bio_size(n) (sizeof(struct btrfs_bio) + \
 			    (sizeof(struct btrfs_bio_stripe) * (n)))
 
-int btrfs_alloc_dev_extent(struct btrfs_trans_handle *trans,
-			   struct btrfs_device *device,
-			   u64 chunk_tree, u64 chunk_objectid,
-			   u64 chunk_offset, u64 start, u64 num_bytes);
-int btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw,
+int btrfs_map_block(struct btrfs_fs_info *fs_info, int rw,
 		    u64 logical, u64 *length,
-		    struct btrfs_multi_bio **multi_ret, int mirror_num);
+		    struct btrfs_bio **bbio_ret, int mirror_num);
 int btrfs_rmap_block(struct btrfs_mapping_tree *map_tree,
 		     u64 chunk_start, u64 physical, u64 devid,
 		     u64 **logical, int *naddrs, int *stripe_len);
@@ -156,32 +302,97 @@ void btrfs_mapping_init(struct btrfs_mapping_tree *tree);
 void btrfs_mapping_tree_free(struct btrfs_mapping_tree *tree);
 int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio,
 		  int mirror_num, int async_submit);
-int btrfs_read_super_device(struct btrfs_root *root, struct extent_buffer *buf);
 int btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
 		       fmode_t flags, void *holder);
 int btrfs_scan_one_device(const char *path, fmode_t flags, void *holder,
 			  struct btrfs_fs_devices **fs_devices_ret);
 int btrfs_close_devices(struct btrfs_fs_devices *fs_devices);
-int btrfs_close_extra_devices(struct btrfs_fs_devices *fs_devices);
-int btrfs_add_device(struct btrfs_trans_handle *trans,
-		     struct btrfs_root *root,
-		     struct btrfs_device *device);
+void btrfs_close_extra_devices(struct btrfs_fs_info *fs_info,
+			       struct btrfs_fs_devices *fs_devices, int step);
+int btrfs_find_device_missing_or_by_path(struct btrfs_root *root,
+					 char *device_path,
+					 struct btrfs_device **device);
+struct btrfs_device *btrfs_alloc_device(struct btrfs_fs_info *fs_info,
+					const u64 *devid,
+					const u8 *uuid);
 int btrfs_rm_device(struct btrfs_root *root, char *device_path);
-int btrfs_cleanup_fs_uuids(void);
-int btrfs_num_copies(struct btrfs_mapping_tree *map_tree, u64 logical, u64 len);
-int btrfs_unplug_page(struct btrfs_mapping_tree *map_tree,
-		      u64 logical, struct page *page);
+void btrfs_cleanup_fs_uuids(void);
+int btrfs_num_copies(struct btrfs_fs_info *fs_info, u64 logical, u64 len);
 int btrfs_grow_device(struct btrfs_trans_handle *trans,
 		      struct btrfs_device *device, u64 new_size);
-struct btrfs_device *btrfs_find_device(struct btrfs_root *root, u64 devid,
+struct btrfs_device *btrfs_find_device(struct btrfs_fs_info *fs_info, u64 devid,
 				       u8 *uuid, u8 *fsid);
 int btrfs_shrink_device(struct btrfs_device *device, u64 new_size);
 int btrfs_init_new_device(struct btrfs_root *root, char *path);
-int btrfs_balance(struct btrfs_root *dev_root);
-void btrfs_unlock_volumes(void);
-void btrfs_lock_volumes(void);
+int btrfs_init_dev_replace_tgtdev(struct btrfs_root *root, char *device_path,
+				  struct btrfs_device **device_out);
+int btrfs_balance(struct btrfs_balance_control *bctl,
+		  struct btrfs_ioctl_balance_args *bargs);
+int btrfs_resume_balance_async(struct btrfs_fs_info *fs_info);
+int btrfs_recover_balance(struct btrfs_fs_info *fs_info);
+int btrfs_pause_balance(struct btrfs_fs_info *fs_info);
+int btrfs_cancel_balance(struct btrfs_fs_info *fs_info);
+int btrfs_create_uuid_tree(struct btrfs_fs_info *fs_info);
+int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info);
 int btrfs_chunk_readonly(struct btrfs_root *root, u64 chunk_offset);
 int find_free_dev_extent(struct btrfs_trans_handle *trans,
 			 struct btrfs_device *device, u64 num_bytes,
 			 u64 *start, u64 *max_avail);
+void btrfs_dev_stat_inc_and_print(struct btrfs_device *dev, int index);
+int btrfs_get_dev_stats(struct btrfs_root *root,
+			struct btrfs_ioctl_get_dev_stats *stats);
+void btrfs_init_devices_late(struct btrfs_fs_info *fs_info);
+int btrfs_init_dev_stats(struct btrfs_fs_info *fs_info);
+int btrfs_run_dev_stats(struct btrfs_trans_handle *trans,
+			struct btrfs_fs_info *fs_info);
+void btrfs_rm_dev_replace_srcdev(struct btrfs_fs_info *fs_info,
+				 struct btrfs_device *srcdev);
+void btrfs_destroy_dev_replace_tgtdev(struct btrfs_fs_info *fs_info,
+				      struct btrfs_device *tgtdev);
+void btrfs_init_dev_replace_tgtdev_for_resume(struct btrfs_fs_info *fs_info,
+					      struct btrfs_device *tgtdev);
+int btrfs_scratch_superblock(struct btrfs_device *device);
+int btrfs_is_parity_mirror(struct btrfs_mapping_tree *map_tree,
+			   u64 logical, u64 len, int mirror_num);
+unsigned long btrfs_full_stripe_len(struct btrfs_root *root,
+				    struct btrfs_mapping_tree *map_tree,
+				    u64 logical);
+int btrfs_finish_chunk_alloc(struct btrfs_trans_handle *trans,
+				struct btrfs_root *extent_root,
+				u64 chunk_offset, u64 chunk_size);
+static inline void btrfs_dev_stat_inc(struct btrfs_device *dev,
+				      int index)
+{
+	atomic_inc(dev->dev_stat_values + index);
+	dev->dev_stats_dirty = 1;
+}
+
+static inline int btrfs_dev_stat_read(struct btrfs_device *dev,
+				      int index)
+{
+	return atomic_read(dev->dev_stat_values + index);
+}
+
+static inline int btrfs_dev_stat_read_and_reset(struct btrfs_device *dev,
+						int index)
+{
+	int ret;
+
+	ret = atomic_xchg(dev->dev_stat_values + index, 0);
+	dev->dev_stats_dirty = 1;
+	return ret;
+}
+
+static inline void btrfs_dev_stat_set(struct btrfs_device *dev,
+				      int index, unsigned long val)
+{
+	atomic_set(dev->dev_stat_values + index, val);
+	dev->dev_stats_dirty = 1;
+}
+
+static inline void btrfs_dev_stat_reset(struct btrfs_device *dev,
+					int index)
+{
+	btrfs_dev_stat_set(dev, index, 0);
+}
 #endif
diff --git a/fs/btrfs/xattr.c b/fs/btrfs/xattr.c
index 193b58f7d3f..ad8328d797e 100644
--- a/fs/btrfs/xattr.c
+++ b/fs/btrfs/xattr.c
@@ -22,11 +22,13 @@
 #include <linux/rwsem.h>
 #include <linux/xattr.h>
 #include <linux/security.h>
+#include <linux/posix_acl_xattr.h>
 #include "ctree.h"
 #include "btrfs_inode.h"
 #include "transaction.h"
 #include "xattr.h"
 #include "disk-io.h"
+#include "props.h"
 
 
 ssize_t __btrfs_getxattr(struct inode *inode, const char *name,
@@ -44,7 +46,7 @@ ssize_t __btrfs_getxattr(struct inode *inode, const char *name,
 		return -ENOMEM;
 
 	/* lookup the xattr by name */
-	di = btrfs_lookup_xattr(NULL, root, path, inode->i_ino, name,
+	di = btrfs_lookup_xattr(NULL, root, path, btrfs_ino(inode), name,
 				strlen(name), 0);
 	if (!di) {
 		ret = -ENODATA;
@@ -102,48 +104,92 @@ static int do_setxattr(struct btrfs_trans_handle *trans,
 	if (!path)
 		return -ENOMEM;
 
-	/* first lets see if we already have this xattr */
-	di = btrfs_lookup_xattr(trans, root, path, inode->i_ino, name,
-				strlen(name), -1);
-	if (IS_ERR(di)) {
-		ret = PTR_ERR(di);
-		goto out;
-	}
-
-	/* ok we already have this xattr, lets remove it */
-	if (di) {
-		/* if we want create only exit */
-		if (flags & XATTR_CREATE) {
-			ret = -EEXIST;
+	if (flags & XATTR_REPLACE) {
+		di = btrfs_lookup_xattr(trans, root, path, btrfs_ino(inode), name,
+					name_len, -1);
+		if (IS_ERR(di)) {
+			ret = PTR_ERR(di);
+			goto out;
+		} else if (!di) {
+			ret = -ENODATA;
 			goto out;
 		}
-
 		ret = btrfs_delete_one_dir_name(trans, root, path, di);
-		BUG_ON(ret);
-		btrfs_release_path(root, path);
+		if (ret)
+			goto out;
+		btrfs_release_path(path);
 
-		/* if we don't have a value then we are removing the xattr */
+		/*
+		 * remove the attribute
+		 */
 		if (!value)
 			goto out;
 	} else {
-		btrfs_release_path(root, path);
-
-		if (flags & XATTR_REPLACE) {
-			/* we couldn't find the attr to replace */
-			ret = -ENODATA;
+		di = btrfs_lookup_xattr(NULL, root, path, btrfs_ino(inode),
+					name, name_len, 0);
+		if (IS_ERR(di)) {
+			ret = PTR_ERR(di);
 			goto out;
 		}
+		if (!di && !value)
+			goto out;
+		btrfs_release_path(path);
 	}
 
-	/* ok we have to create a completely new xattr */
-	ret = btrfs_insert_xattr_item(trans, root, path, inode->i_ino,
+again:
+	ret = btrfs_insert_xattr_item(trans, root, path, btrfs_ino(inode),
 				      name, name_len, value, size);
-	BUG_ON(ret);
+	/*
+	 * If we're setting an xattr to a new value but the new value is say
+	 * exactly BTRFS_MAX_XATTR_SIZE, we could end up with EOVERFLOW getting
+	 * back from split_leaf.  This is because it thinks we'll be extending
+	 * the existing item size, but we're asking for enough space to add the
+	 * item itself.  So if we get EOVERFLOW just set ret to EEXIST and let
+	 * the rest of the function figure it out.
+	 */
+	if (ret == -EOVERFLOW)
+		ret = -EEXIST;
+
+	if (ret == -EEXIST) {
+		if (flags & XATTR_CREATE)
+			goto out;
+		/*
+		 * We can't use the path we already have since we won't have the
+		 * proper locking for a delete, so release the path and
+		 * re-lookup to delete the thing.
+		 */
+		btrfs_release_path(path);
+		di = btrfs_lookup_xattr(trans, root, path, btrfs_ino(inode),
+					name, name_len, -1);
+		if (IS_ERR(di)) {
+			ret = PTR_ERR(di);
+			goto out;
+		} else if (!di) {
+			/* Shouldn't happen but just in case... */
+			btrfs_release_path(path);
+			goto again;
+		}
+
+		ret = btrfs_delete_one_dir_name(trans, root, path, di);
+		if (ret)
+			goto out;
+
+		/*
+		 * We have a value to set, so go back and try to insert it now.
+		 */
+		if (value) {
+			btrfs_release_path(path);
+			goto again;
+		}
+	}
 out:
 	btrfs_free_path(path);
 	return ret;
 }
 
+/*
+ * @value: "" makes the attribute to empty, NULL removes it
+ */
 int __btrfs_setxattr(struct btrfs_trans_handle *trans,
 		     struct inode *inode, const char *name,
 		     const void *value, size_t size, int flags)
@@ -154,27 +200,21 @@ int __btrfs_setxattr(struct btrfs_trans_handle *trans,
 	if (trans)
 		return do_setxattr(trans, inode, name, value, size, flags);
 
-	ret = btrfs_reserve_metadata_space(root, 2);
-	if (ret)
-		return ret;
-
-	trans = btrfs_start_transaction(root, 1);
-	if (!trans) {
-		ret = -ENOMEM;
-		goto out;
-	}
-	btrfs_set_trans_block_group(trans, inode);
+	trans = btrfs_start_transaction(root, 2);
+	if (IS_ERR(trans))
+		return PTR_ERR(trans);
 
 	ret = do_setxattr(trans, inode, name, value, size, flags);
 	if (ret)
 		goto out;
 
+	inode_inc_iversion(inode);
 	inode->i_ctime = CURRENT_TIME;
+	set_bit(BTRFS_INODE_COPY_EVERYTHING, &BTRFS_I(inode)->runtime_flags);
 	ret = btrfs_update_inode(trans, root, inode);
 	BUG_ON(ret);
 out:
-	btrfs_end_transaction_throttle(trans, root);
-	btrfs_unreserve_metadata_space(root, 2);
+	btrfs_end_transaction(trans, root);
 	return ret;
 }
 
@@ -184,21 +224,19 @@ ssize_t btrfs_listxattr(struct dentry *dentry, char *buffer, size_t size)
 	struct inode *inode = dentry->d_inode;
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	struct btrfs_path *path;
-	struct btrfs_item *item;
 	struct extent_buffer *leaf;
 	struct btrfs_dir_item *di;
-	int ret = 0, slot, advance;
+	int ret = 0, slot;
 	size_t total_size = 0, size_left = size;
 	unsigned long name_ptr;
 	size_t name_len;
-	u32 nritems;
 
 	/*
 	 * ok we want all objects associated with this id.
 	 * NOTE: we set key.offset = 0; because we want to start with the
 	 * first xattr that we find and walk forward
 	 */
-	key.objectid = inode->i_ino;
+	key.objectid = btrfs_ino(inode);
 	btrfs_set_key_type(&key, BTRFS_XATTR_ITEM_KEY);
 	key.offset = 0;
 
@@ -211,36 +249,25 @@ ssize_t btrfs_listxattr(struct dentry *dentry, char *buffer, size_t size)
 	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
 	if (ret < 0)
 		goto err;
-	advance = 0;
+
 	while (1) {
 		leaf = path->nodes[0];
-		nritems = btrfs_header_nritems(leaf);
 		slot = path->slots[0];
 
 		/* this is where we start walking through the path */
-		if (advance || slot >= nritems) {
+		if (slot >= btrfs_header_nritems(leaf)) {
 			/*
 			 * if we've reached the last slot in this leaf we need
 			 * to go to the next leaf and reset everything
 			 */
-			if (slot >= nritems-1) {
-				ret = btrfs_next_leaf(root, path);
-				if (ret)
-					break;
-				leaf = path->nodes[0];
-				nritems = btrfs_header_nritems(leaf);
-				slot = path->slots[0];
-			} else {
-				/*
-				 * just walking through the slots on this leaf
-				 */
-				slot++;
-				path->slots[0]++;
-			}
+			ret = btrfs_next_leaf(root, path);
+			if (ret < 0)
+				goto err;
+			else if (ret > 0)
+				break;
+			continue;
 		}
-		advance = 1;
 
-		item = btrfs_item_nr(leaf, slot);
 		btrfs_item_key_to_cpu(leaf, &found_key, slot);
 
 		/* check to make sure this item is what we want */
@@ -250,13 +277,15 @@ ssize_t btrfs_listxattr(struct dentry *dentry, char *buffer, size_t size)
 			break;
 
 		di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
+		if (verify_dir_item(root, leaf, di))
+			goto next;
 
 		name_len = btrfs_dir_name_len(leaf, di);
 		total_size += name_len + 1;
 
 		/* we are just looking for how big our buffer needs to be */
 		if (!size)
-			continue;
+			goto next;
 
 		if (!buffer || (name_len + 1) > size_left) {
 			ret = -ERANGE;
@@ -269,6 +298,8 @@ ssize_t btrfs_listxattr(struct dentry *dentry, char *buffer, size_t size)
 
 		size_left -= name_len + 1;
 		buffer += name_len + 1;
+next:
+		path->slots[0]++;
 	}
 	ret = total_size;
 
@@ -282,10 +313,10 @@ err:
  * List of handlers for synthetic system.* attributes.  All real ondisk
  * attributes are handled directly.
  */
-struct xattr_handler *btrfs_xattr_handlers[] = {
+const struct xattr_handler *btrfs_xattr_handlers[] = {
 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
-	&btrfs_xattr_acl_access_handler,
-	&btrfs_xattr_acl_default_handler,
+	&posix_acl_access_xattr_handler,
+	&posix_acl_default_xattr_handler,
 #endif
 	NULL,
 };
@@ -302,7 +333,8 @@ static bool btrfs_is_valid_xattr(const char *name)
 			XATTR_SECURITY_PREFIX_LEN) ||
 	       !strncmp(name, XATTR_SYSTEM_PREFIX, XATTR_SYSTEM_PREFIX_LEN) ||
 	       !strncmp(name, XATTR_TRUSTED_PREFIX, XATTR_TRUSTED_PREFIX_LEN) ||
-	       !strncmp(name, XATTR_USER_PREFIX, XATTR_USER_PREFIX_LEN);
+	       !strncmp(name, XATTR_USER_PREFIX, XATTR_USER_PREFIX_LEN) ||
+		!strncmp(name, XATTR_BTRFS_PREFIX, XATTR_BTRFS_PREFIX_LEN);
 }
 
 ssize_t btrfs_getxattr(struct dentry *dentry, const char *name,
@@ -324,6 +356,15 @@ ssize_t btrfs_getxattr(struct dentry *dentry, const char *name,
 int btrfs_setxattr(struct dentry *dentry, const char *name, const void *value,
 		   size_t size, int flags)
 {
+	struct btrfs_root *root = BTRFS_I(dentry->d_inode)->root;
+
+	/*
+	 * The permission on security.* and system.* is not checked
+	 * in permission().
+	 */
+	if (btrfs_root_readonly(root))
+		return -EROFS;
+
 	/*
 	 * If this is a request for a synthetic attribute in the system.*
 	 * namespace use the generic infrastructure to resolve a handler
@@ -335,6 +376,10 @@ int btrfs_setxattr(struct dentry *dentry, const char *name, const void *value,
 	if (!btrfs_is_valid_xattr(name))
 		return -EOPNOTSUPP;
 
+	if (!strncmp(name, XATTR_BTRFS_PREFIX, XATTR_BTRFS_PREFIX_LEN))
+		return btrfs_set_prop(dentry->d_inode, name,
+				      value, size, flags);
+
 	if (size == 0)
 		value = "";  /* empty EA, do not remove */
 
@@ -344,6 +389,15 @@ int btrfs_setxattr(struct dentry *dentry, const char *name, const void *value,
 
 int btrfs_removexattr(struct dentry *dentry, const char *name)
 {
+	struct btrfs_root *root = BTRFS_I(dentry->d_inode)->root;
+
+	/*
+	 * The permission on security.* and system.* is not checked
+	 * in permission().
+	 */
+	if (btrfs_root_readonly(root))
+		return -EROFS;
+
 	/*
 	 * If this is a request for a synthetic attribute in the system.*
 	 * namespace use the generic infrastructure to resolve a handler
@@ -355,38 +409,44 @@ int btrfs_removexattr(struct dentry *dentry, const char *name)
 	if (!btrfs_is_valid_xattr(name))
 		return -EOPNOTSUPP;
 
+	if (!strncmp(name, XATTR_BTRFS_PREFIX, XATTR_BTRFS_PREFIX_LEN))
+		return btrfs_set_prop(dentry->d_inode, name,
+				      NULL, 0, XATTR_REPLACE);
+
 	return __btrfs_setxattr(NULL, dentry->d_inode, name, NULL, 0,
 				XATTR_REPLACE);
 }
 
-int btrfs_xattr_security_init(struct btrfs_trans_handle *trans,
-			      struct inode *inode, struct inode *dir)
+static int btrfs_initxattrs(struct inode *inode,
+			    const struct xattr *xattr_array, void *fs_info)
 {
-	int err;
-	size_t len;
-	void *value;
-	char *suffix;
+	const struct xattr *xattr;
+	struct btrfs_trans_handle *trans = fs_info;
 	char *name;
+	int err = 0;
 
-	err = security_inode_init_security(inode, dir, &suffix, &value, &len);
-	if (err) {
-		if (err == -EOPNOTSUPP)
-			return 0;
-		return err;
-	}
-
-	name = kmalloc(XATTR_SECURITY_PREFIX_LEN + strlen(suffix) + 1,
-		       GFP_NOFS);
-	if (!name) {
-		err = -ENOMEM;
-	} else {
+	for (xattr = xattr_array; xattr->name != NULL; xattr++) {
+		name = kmalloc(XATTR_SECURITY_PREFIX_LEN +
+			       strlen(xattr->name) + 1, GFP_NOFS);
+		if (!name) {
+			err = -ENOMEM;
+			break;
+		}
 		strcpy(name, XATTR_SECURITY_PREFIX);
-		strcpy(name + XATTR_SECURITY_PREFIX_LEN, suffix);
-		err = __btrfs_setxattr(trans, inode, name, value, len, 0);
+		strcpy(name + XATTR_SECURITY_PREFIX_LEN, xattr->name);
+		err = __btrfs_setxattr(trans, inode, name,
+				       xattr->value, xattr->value_len, 0);
 		kfree(name);
+		if (err < 0)
+			break;
 	}
-
-	kfree(suffix);
-	kfree(value);
 	return err;
 }
+
+int btrfs_xattr_security_init(struct btrfs_trans_handle *trans,
+			      struct inode *inode, struct inode *dir,
+			      const struct qstr *qstr)
+{
+	return security_inode_init_security(inode, dir, qstr,
+					    &btrfs_initxattrs, trans);
+}
diff --git a/fs/btrfs/xattr.h b/fs/btrfs/xattr.h
index 721efa0346e..5049608d138 100644
--- a/fs/btrfs/xattr.h
+++ b/fs/btrfs/xattr.h
@@ -21,9 +21,7 @@
 
 #include <linux/xattr.h>
 
-extern struct xattr_handler btrfs_xattr_acl_access_handler;
-extern struct xattr_handler btrfs_xattr_acl_default_handler;
-extern struct xattr_handler *btrfs_xattr_handlers[];
+extern const struct xattr_handler *btrfs_xattr_handlers[];
 
 extern ssize_t __btrfs_getxattr(struct inode *inode, const char *name,
 		void *buffer, size_t size);
@@ -37,6 +35,7 @@ extern int btrfs_setxattr(struct dentry *dentry, const char *name,
 extern int btrfs_removexattr(struct dentry *dentry, const char *name);
 
 extern int btrfs_xattr_security_init(struct btrfs_trans_handle *trans,
-				     struct inode *inode, struct inode *dir);
+				     struct inode *inode, struct inode *dir,
+				     const struct qstr *qstr);
 
 #endif /* __XATTR__ */
diff --git a/fs/btrfs/zlib.c b/fs/btrfs/zlib.c
index 3e2b90eaa23..b67d8fc8127 100644
--- a/fs/btrfs/zlib.c
+++ b/fs/btrfs/zlib.c
@@ -32,15 +32,6 @@
 #include <linux/bio.h>
 #include "compression.h"
 
-/* Plan: call deflate() with avail_in == *sourcelen,
-	avail_out = *dstlen - 12 and flush == Z_FINISH.
-	If it doesn't manage to finish,	call it again with
-	avail_in == 0 and avail_out set to the remaining 12
-	bytes for it to clean up.
-   Q: Is 12 bytes sufficient?
-*/
-#define STREAM_END_SPACE 12
-
 struct workspace {
 	z_stream inf_strm;
 	z_stream def_strm;
@@ -48,172 +39,66 @@ struct workspace {
 	struct list_head list;
 };
 
-static LIST_HEAD(idle_workspace);
-static DEFINE_SPINLOCK(workspace_lock);
-static unsigned long num_workspace;
-static atomic_t alloc_workspace = ATOMIC_INIT(0);
-static DECLARE_WAIT_QUEUE_HEAD(workspace_wait);
-
-/*
- * this finds an available zlib workspace or allocates a new one
- * NULL or an ERR_PTR is returned if things go bad.
- */
-static struct workspace *find_zlib_workspace(void)
+static void zlib_free_workspace(struct list_head *ws)
 {
-	struct workspace *workspace;
-	int ret;
-	int cpus = num_online_cpus();
-
-again:
-	spin_lock(&workspace_lock);
-	if (!list_empty(&idle_workspace)) {
-		workspace = list_entry(idle_workspace.next, struct workspace,
-				       list);
-		list_del(&workspace->list);
-		num_workspace--;
-		spin_unlock(&workspace_lock);
-		return workspace;
-
-	}
-	spin_unlock(&workspace_lock);
-	if (atomic_read(&alloc_workspace) > cpus) {
-		DEFINE_WAIT(wait);
-		prepare_to_wait(&workspace_wait, &wait, TASK_UNINTERRUPTIBLE);
-		if (atomic_read(&alloc_workspace) > cpus)
-			schedule();
-		finish_wait(&workspace_wait, &wait);
-		goto again;
-	}
-	atomic_inc(&alloc_workspace);
-	workspace = kzalloc(sizeof(*workspace), GFP_NOFS);
-	if (!workspace) {
-		ret = -ENOMEM;
-		goto fail;
-	}
-
-	workspace->def_strm.workspace = vmalloc(zlib_deflate_workspacesize());
-	if (!workspace->def_strm.workspace) {
-		ret = -ENOMEM;
-		goto fail;
-	}
-	workspace->inf_strm.workspace = vmalloc(zlib_inflate_workspacesize());
-	if (!workspace->inf_strm.workspace) {
-		ret = -ENOMEM;
-		goto fail_inflate;
-	}
-	workspace->buf = kmalloc(PAGE_CACHE_SIZE, GFP_NOFS);
-	if (!workspace->buf) {
-		ret = -ENOMEM;
-		goto fail_kmalloc;
-	}
-	return workspace;
-
-fail_kmalloc:
-	vfree(workspace->inf_strm.workspace);
-fail_inflate:
-	vfree(workspace->def_strm.workspace);
-fail:
-	kfree(workspace);
-	atomic_dec(&alloc_workspace);
-	wake_up(&workspace_wait);
-	return ERR_PTR(ret);
-}
+	struct workspace *workspace = list_entry(ws, struct workspace, list);
 
-/*
- * put a workspace struct back on the list or free it if we have enough
- * idle ones sitting around
- */
-static int free_workspace(struct workspace *workspace)
-{
-	spin_lock(&workspace_lock);
-	if (num_workspace < num_online_cpus()) {
-		list_add_tail(&workspace->list, &idle_workspace);
-		num_workspace++;
-		spin_unlock(&workspace_lock);
-		if (waitqueue_active(&workspace_wait))
-			wake_up(&workspace_wait);
-		return 0;
-	}
-	spin_unlock(&workspace_lock);
 	vfree(workspace->def_strm.workspace);
 	vfree(workspace->inf_strm.workspace);
 	kfree(workspace->buf);
 	kfree(workspace);
-
-	atomic_dec(&alloc_workspace);
-	if (waitqueue_active(&workspace_wait))
-		wake_up(&workspace_wait);
-	return 0;
 }
 
-/*
- * cleanup function for module exit
- */
-static void free_workspaces(void)
+static struct list_head *zlib_alloc_workspace(void)
 {
 	struct workspace *workspace;
-	while (!list_empty(&idle_workspace)) {
-		workspace = list_entry(idle_workspace.next, struct workspace,
-				       list);
-		list_del(&workspace->list);
-		vfree(workspace->def_strm.workspace);
-		vfree(workspace->inf_strm.workspace);
-		kfree(workspace->buf);
-		kfree(workspace);
-		atomic_dec(&alloc_workspace);
-	}
+
+	workspace = kzalloc(sizeof(*workspace), GFP_NOFS);
+	if (!workspace)
+		return ERR_PTR(-ENOMEM);
+
+	workspace->def_strm.workspace = vmalloc(zlib_deflate_workspacesize(
+						MAX_WBITS, MAX_MEM_LEVEL));
+	workspace->inf_strm.workspace = vmalloc(zlib_inflate_workspacesize());
+	workspace->buf = kmalloc(PAGE_CACHE_SIZE, GFP_NOFS);
+	if (!workspace->def_strm.workspace ||
+	    !workspace->inf_strm.workspace || !workspace->buf)
+		goto fail;
+
+	INIT_LIST_HEAD(&workspace->list);
+
+	return &workspace->list;
+fail:
+	zlib_free_workspace(&workspace->list);
+	return ERR_PTR(-ENOMEM);
 }
 
-/*
- * given an address space and start/len, compress the bytes.
- *
- * pages are allocated to hold the compressed result and stored
- * in 'pages'
- *
- * out_pages is used to return the number of pages allocated.  There
- * may be pages allocated even if we return an error
- *
- * total_in is used to return the number of bytes actually read.  It
- * may be smaller then len if we had to exit early because we
- * ran out of room in the pages array or because we cross the
- * max_out threshold.
- *
- * total_out is used to return the total number of compressed bytes
- *
- * max_out tells us the max number of bytes that we're allowed to
- * stuff into pages
- */
-int btrfs_zlib_compress_pages(struct address_space *mapping,
-			      u64 start, unsigned long len,
-			      struct page **pages,
-			      unsigned long nr_dest_pages,
-			      unsigned long *out_pages,
-			      unsigned long *total_in,
-			      unsigned long *total_out,
-			      unsigned long max_out)
+static int zlib_compress_pages(struct list_head *ws,
+			       struct address_space *mapping,
+			       u64 start, unsigned long len,
+			       struct page **pages,
+			       unsigned long nr_dest_pages,
+			       unsigned long *out_pages,
+			       unsigned long *total_in,
+			       unsigned long *total_out,
+			       unsigned long max_out)
 {
+	struct workspace *workspace = list_entry(ws, struct workspace, list);
 	int ret;
-	struct workspace *workspace;
 	char *data_in;
 	char *cpage_out;
 	int nr_pages = 0;
 	struct page *in_page = NULL;
 	struct page *out_page = NULL;
-	int out_written = 0;
-	int in_read = 0;
 	unsigned long bytes_left;
 
 	*out_pages = 0;
 	*total_out = 0;
 	*total_in = 0;
 
-	workspace = find_zlib_workspace();
-	if (IS_ERR(workspace))
-		return -1;
-
 	if (Z_OK != zlib_deflateInit(&workspace->def_strm, 3)) {
-		printk(KERN_WARNING "deflateInit failed\n");
-		ret = -1;
+		printk(KERN_WARNING "BTRFS: deflateInit failed\n");
+		ret = -EIO;
 		goto out;
 	}
 
@@ -224,6 +109,10 @@ int btrfs_zlib_compress_pages(struct address_space *mapping,
 	data_in = kmap(in_page);
 
 	out_page = alloc_page(GFP_NOFS | __GFP_HIGHMEM);
+	if (out_page == NULL) {
+		ret = -ENOMEM;
+		goto out;
+	}
 	cpage_out = kmap(out_page);
 	pages[0] = out_page;
 	nr_pages = 1;
@@ -233,16 +122,13 @@ int btrfs_zlib_compress_pages(struct address_space *mapping,
 	workspace->def_strm.avail_out = PAGE_CACHE_SIZE;
 	workspace->def_strm.avail_in = min(len, PAGE_CACHE_SIZE);
 
-	out_written = 0;
-	in_read = 0;
-
 	while (workspace->def_strm.total_in < len) {
 		ret = zlib_deflate(&workspace->def_strm, Z_SYNC_FLUSH);
 		if (ret != Z_OK) {
-			printk(KERN_DEBUG "btrfs deflate in loop returned %d\n",
+			printk(KERN_DEBUG "BTRFS: deflate in loop returned %d\n",
 			       ret);
 			zlib_deflateEnd(&workspace->def_strm);
-			ret = -1;
+			ret = -EIO;
 			goto out;
 		}
 
@@ -250,7 +136,7 @@ int btrfs_zlib_compress_pages(struct address_space *mapping,
 		if (workspace->def_strm.total_in > 8192 &&
 		    workspace->def_strm.total_in <
 		    workspace->def_strm.total_out) {
-			ret = -1;
+			ret = -E2BIG;
 			goto out;
 		}
 		/* we need another page for writing out.  Test this
@@ -261,10 +147,14 @@ int btrfs_zlib_compress_pages(struct address_space *mapping,
 			kunmap(out_page);
 			if (nr_pages == nr_dest_pages) {
 				out_page = NULL;
-				ret = -1;
+				ret = -E2BIG;
 				goto out;
 			}
 			out_page = alloc_page(GFP_NOFS | __GFP_HIGHMEM);
+			if (out_page == NULL) {
+				ret = -ENOMEM;
+				goto out;
+			}
 			cpage_out = kmap(out_page);
 			pages[nr_pages] = out_page;
 			nr_pages++;
@@ -298,12 +188,12 @@ int btrfs_zlib_compress_pages(struct address_space *mapping,
 	zlib_deflateEnd(&workspace->def_strm);
 
 	if (ret != Z_STREAM_END) {
-		ret = -1;
+		ret = -EIO;
 		goto out;
 	}
 
 	if (workspace->def_strm.total_out >= workspace->def_strm.total_in) {
-		ret = -1;
+		ret = -E2BIG;
 		goto out;
 	}
 
@@ -319,55 +209,26 @@ out:
 		kunmap(in_page);
 		page_cache_release(in_page);
 	}
-	free_workspace(workspace);
 	return ret;
 }
 
-/*
- * pages_in is an array of pages with compressed data.
- *
- * disk_start is the starting logical offset of this array in the file
- *
- * bvec is a bio_vec of pages from the file that we want to decompress into
- *
- * vcnt is the count of pages in the biovec
- *
- * srclen is the number of bytes in pages_in
- *
- * The basic idea is that we have a bio that was created by readpages.
- * The pages in the bio are for the uncompressed data, and they may not
- * be contiguous.  They all correspond to the range of bytes covered by
- * the compressed extent.
- */
-int btrfs_zlib_decompress_biovec(struct page **pages_in,
-			      u64 disk_start,
-			      struct bio_vec *bvec,
-			      int vcnt,
-			      size_t srclen)
+static int zlib_decompress_biovec(struct list_head *ws, struct page **pages_in,
+				  u64 disk_start,
+				  struct bio_vec *bvec,
+				  int vcnt,
+				  size_t srclen)
 {
-	int ret = 0;
+	struct workspace *workspace = list_entry(ws, struct workspace, list);
+	int ret = 0, ret2;
 	int wbits = MAX_WBITS;
-	struct workspace *workspace;
 	char *data_in;
 	size_t total_out = 0;
-	unsigned long page_bytes_left;
 	unsigned long page_in_index = 0;
 	unsigned long page_out_index = 0;
-	struct page *page_out;
 	unsigned long total_pages_in = (srclen + PAGE_CACHE_SIZE - 1) /
 					PAGE_CACHE_SIZE;
 	unsigned long buf_start;
-	unsigned long buf_offset;
-	unsigned long bytes;
-	unsigned long working_bytes;
 	unsigned long pg_offset;
-	unsigned long start_byte;
-	unsigned long current_buf_start;
-	char *kaddr;
-
-	workspace = find_zlib_workspace();
-	if (IS_ERR(workspace))
-		return -ENOMEM;
 
 	data_in = kmap(pages_in[page_in_index]);
 	workspace->inf_strm.next_in = data_in;
@@ -377,8 +238,6 @@ int btrfs_zlib_decompress_biovec(struct page **pages_in,
 	workspace->inf_strm.total_out = 0;
 	workspace->inf_strm.next_out = workspace->buf;
 	workspace->inf_strm.avail_out = PAGE_CACHE_SIZE;
-	page_out = bvec[page_out_index].bv_page;
-	page_bytes_left = PAGE_CACHE_SIZE;
 	pg_offset = 0;
 
 	/* If it's deflate, and it's got no preset dictionary, then
@@ -393,108 +252,30 @@ int btrfs_zlib_decompress_biovec(struct page **pages_in,
 	}
 
 	if (Z_OK != zlib_inflateInit2(&workspace->inf_strm, wbits)) {
-		printk(KERN_WARNING "inflateInit failed\n");
-		ret = -1;
-		goto out;
+		printk(KERN_WARNING "BTRFS: inflateInit failed\n");
+		return -EIO;
 	}
 	while (workspace->inf_strm.total_in < srclen) {
 		ret = zlib_inflate(&workspace->inf_strm, Z_NO_FLUSH);
 		if (ret != Z_OK && ret != Z_STREAM_END)
 			break;
-		/*
-		 * buf start is the byte offset we're of the start of
-		 * our workspace buffer
-		 */
-		buf_start = total_out;
 
-		/* total_out is the last byte of the workspace buffer */
+		buf_start = total_out;
 		total_out = workspace->inf_strm.total_out;
 
-		working_bytes = total_out - buf_start;
-
-		/*
-		 * start byte is the first byte of the page we're currently
-		 * copying into relative to the start of the compressed data.
-		 */
-		start_byte = page_offset(page_out) - disk_start;
-
-		if (working_bytes == 0) {
-			/* we didn't make progress in this inflate
-			 * call, we're done
-			 */
-			if (ret != Z_STREAM_END)
-				ret = -1;
+		/* we didn't make progress in this inflate call, we're done */
+		if (buf_start == total_out)
 			break;
-		}
-
-		/* we haven't yet hit data corresponding to this page */
-		if (total_out <= start_byte)
-			goto next;
 
-		/*
-		 * the start of the data we care about is offset into
-		 * the middle of our working buffer
-		 */
-		if (total_out > start_byte && buf_start < start_byte) {
-			buf_offset = start_byte - buf_start;
-			working_bytes -= buf_offset;
-		} else {
-			buf_offset = 0;
+		ret2 = btrfs_decompress_buf2page(workspace->buf, buf_start,
+						 total_out, disk_start,
+						 bvec, vcnt,
+						 &page_out_index, &pg_offset);
+		if (ret2 == 0) {
+			ret = 0;
+			goto done;
 		}
-		current_buf_start = buf_start;
-
-		/* copy bytes from the working buffer into the pages */
-		while (working_bytes > 0) {
-			bytes = min(PAGE_CACHE_SIZE - pg_offset,
-				    PAGE_CACHE_SIZE - buf_offset);
-			bytes = min(bytes, working_bytes);
-			kaddr = kmap_atomic(page_out, KM_USER0);
-			memcpy(kaddr + pg_offset, workspace->buf + buf_offset,
-			       bytes);
-			kunmap_atomic(kaddr, KM_USER0);
-			flush_dcache_page(page_out);
-
-			pg_offset += bytes;
-			page_bytes_left -= bytes;
-			buf_offset += bytes;
-			working_bytes -= bytes;
-			current_buf_start += bytes;
-
-			/* check if we need to pick another page */
-			if (page_bytes_left == 0) {
-				page_out_index++;
-				if (page_out_index >= vcnt) {
-					ret = 0;
-					goto done;
-				}
-
-				page_out = bvec[page_out_index].bv_page;
-				pg_offset = 0;
-				page_bytes_left = PAGE_CACHE_SIZE;
-				start_byte = page_offset(page_out) - disk_start;
-
-				/*
-				 * make sure our new page is covered by this
-				 * working buffer
-				 */
-				if (total_out <= start_byte)
-					goto next;
-
-				/* the next page in the biovec might not
-				 * be adjacent to the last page, but it
-				 * might still be found inside this working
-				 * buffer.  bump our offset pointer
-				 */
-				if (total_out > start_byte &&
-				    current_buf_start < start_byte) {
-					buf_offset = start_byte - buf_start;
-					working_bytes = total_out - start_byte;
-					current_buf_start = buf_start +
-						buf_offset;
-				}
-			}
-		}
-next:
+
 		workspace->inf_strm.next_out = workspace->buf;
 		workspace->inf_strm.avail_out = PAGE_CACHE_SIZE;
 
@@ -514,42 +295,28 @@ next:
 		}
 	}
 	if (ret != Z_STREAM_END)
-		ret = -1;
+		ret = -EIO;
 	else
 		ret = 0;
 done:
 	zlib_inflateEnd(&workspace->inf_strm);
 	if (data_in)
 		kunmap(pages_in[page_in_index]);
-out:
-	free_workspace(workspace);
 	return ret;
 }
 
-/*
- * a less complex decompression routine.  Our compressed data fits in a
- * single page, and we want to read a single page out of it.
- * start_byte tells us the offset into the compressed data we're interested in
- */
-int btrfs_zlib_decompress(unsigned char *data_in,
-			  struct page *dest_page,
-			  unsigned long start_byte,
-			  size_t srclen, size_t destlen)
+static int zlib_decompress(struct list_head *ws, unsigned char *data_in,
+			   struct page *dest_page,
+			   unsigned long start_byte,
+			   size_t srclen, size_t destlen)
 {
+	struct workspace *workspace = list_entry(ws, struct workspace, list);
 	int ret = 0;
 	int wbits = MAX_WBITS;
-	struct workspace *workspace;
 	unsigned long bytes_left = destlen;
 	unsigned long total_out = 0;
 	char *kaddr;
 
-	if (destlen > PAGE_CACHE_SIZE)
-		return -ENOMEM;
-
-	workspace = find_zlib_workspace();
-	if (IS_ERR(workspace))
-		return -ENOMEM;
-
 	workspace->inf_strm.next_in = data_in;
 	workspace->inf_strm.avail_in = srclen;
 	workspace->inf_strm.total_in = 0;
@@ -569,9 +336,8 @@ int btrfs_zlib_decompress(unsigned char *data_in,
 	}
 
 	if (Z_OK != zlib_inflateInit2(&workspace->inf_strm, wbits)) {
-		printk(KERN_WARNING "inflateInit failed\n");
-		ret = -1;
-		goto out;
+		printk(KERN_WARNING "BTRFS: inflateInit failed\n");
+		return -EIO;
 	}
 
 	while (bytes_left > 0) {
@@ -588,7 +354,7 @@ int btrfs_zlib_decompress(unsigned char *data_in,
 		total_out = workspace->inf_strm.total_out;
 
 		if (total_out == buf_start) {
-			ret = -1;
+			ret = -EIO;
 			break;
 		}
 
@@ -604,9 +370,9 @@ int btrfs_zlib_decompress(unsigned char *data_in,
 			    PAGE_CACHE_SIZE - buf_offset);
 		bytes = min(bytes, bytes_left);
 
-		kaddr = kmap_atomic(dest_page, KM_USER0);
+		kaddr = kmap_atomic(dest_page);
 		memcpy(kaddr + pg_offset, workspace->buf + buf_offset, bytes);
-		kunmap_atomic(kaddr, KM_USER0);
+		kunmap_atomic(kaddr);
 
 		pg_offset += bytes;
 		bytes_left -= bytes;
@@ -616,17 +382,18 @@ next:
 	}
 
 	if (ret != Z_STREAM_END && bytes_left != 0)
-		ret = -1;
+		ret = -EIO;
 	else
 		ret = 0;
 
 	zlib_inflateEnd(&workspace->inf_strm);
-out:
-	free_workspace(workspace);
 	return ret;
 }
 
-void btrfs_zlib_exit(void)
-{
-    free_workspaces();
-}
+struct btrfs_compress_op btrfs_zlib_compress = {
+	.alloc_workspace	= zlib_alloc_workspace,
+	.free_workspace		= zlib_free_workspace,
+	.compress_pages		= zlib_compress_pages,
+	.decompress_biovec	= zlib_decompress_biovec,
+	.decompress		= zlib_decompress,
+};