Btrfs: Add a write ahead tree log to optimize synchronous operations

File syncs and directory syncs are optimized by copying their
items into a special (copy-on-write) log tree.  There is one log tree per
subvolume and the btrfs super block points to a tree of log tree roots.

After a crash, items are copied out of the log tree and back into the
subvolume.  See tree-log.c for all the details.

Signed-off-by: Chris Mason <chris.mason@oracle.com>

17 files changed, 3408 insertions, 205 deletions

diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile
index 3a01065d4ef..b7addbfd8c2 100644
--- a/fs/btrfs/Makefile
+++ b/fs/btrfs/Makefile
@@ -7,8 +7,7 @@ btrfs-y := super.o ctree.o extent-tree.o print-tree.o root-tree.o dir-item.o \
 	   transaction.o bit-radix.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 \
-	   ref-cache.o acl.o export.o
-
+	   ref-cache.o export.o tree-log.o acl.o
 else
 
 # Normal Makefile
diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h
index 111f90524ae..fcc8cf27e90 100644
--- a/fs/btrfs/btrfs_inode.h
+++ b/fs/btrfs/btrfs_inode.h
@@ -33,6 +33,7 @@ struct btrfs_inode {
 	struct extent_io_tree io_failure_tree;
 	struct mutex csum_mutex;
 	struct mutex extent_mutex;
+	struct mutex log_mutex;
 	struct inode vfs_inode;
 	struct btrfs_ordered_inode_tree ordered_tree;
 
@@ -44,10 +45,17 @@ struct btrfs_inode {
 
 	struct list_head delalloc_inodes;
 
+	/* full 64 bit generation number */
+	u64 generation;
+
 	/*
 	 * transid of the trans_handle that last modified this inode
 	 */
 	u64 last_trans;
+	/*
+	 * transid that last logged this inode
+	 */
+	u64 logged_trans;
 	u64 delalloc_bytes;
 	u64 disk_i_size;
 	u32 flags;
diff --git a/fs/btrfs/compat.h b/fs/btrfs/compat.h
index d45fb37887b..b0ed1887d9b 100644
--- a/fs/btrfs/compat.h
+++ b/fs/btrfs/compat.h
@@ -22,6 +22,21 @@ static inline struct dentry *d_obtain_alias(struct inode *inode)
 }
 #endif
 
+#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
+static inline void btrfs_drop_nlink(struct inode *inode)
+{
+	inode->i_nlink--;
+}
+
+static inline void btrfs_inc_nlink(struct inode *inode)
+{
+	inode->i_nlink++;
+}
+#else
+# define btrfs_drop_nlink(inode) drop_nlink(inode)
+# define btrfs_inc_nlink(inode)	inc_nlink(inode)
+#endif
+
 /*
  * Even if AppArmor isn't enabled, it still has different prototypes.
  * Add more distro/version pairs here to declare which has AppArmor applied.
diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index 7114faafa9d..579124043d9 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -60,7 +60,7 @@ void btrfs_free_path(struct btrfs_path *p)
 	kmem_cache_free(btrfs_path_cachep, p);
 }
 
-void btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p)
+void noinline btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p)
 {
 	int i;
 
@@ -176,7 +176,7 @@ int btrfs_copy_root(struct btrfs_trans_handle *trans,
 	return 0;
 }
 
-int __btrfs_cow_block(struct btrfs_trans_handle *trans,
+int noinline __btrfs_cow_block(struct btrfs_trans_handle *trans,
 			     struct btrfs_root *root,
 			     struct extent_buffer *buf,
 			     struct extent_buffer *parent, int parent_slot,
@@ -294,7 +294,7 @@ int __btrfs_cow_block(struct btrfs_trans_handle *trans,
 	return 0;
 }
 
-int btrfs_cow_block(struct btrfs_trans_handle *trans,
+int noinline btrfs_cow_block(struct btrfs_trans_handle *trans,
 		    struct btrfs_root *root, struct extent_buffer *buf,
 		    struct extent_buffer *parent, int parent_slot,
 		    struct extent_buffer **cow_ret, u64 prealloc_dest)
@@ -677,9 +677,10 @@ static int noinline check_block(struct btrfs_root *root,
  *
  * slot may point to max if the key is bigger than all of the keys
  */
-static int generic_bin_search(struct extent_buffer *eb, unsigned long p,
-			      int item_size, struct btrfs_key *key,
-			      int max, int *slot)
+static noinline int generic_bin_search(struct extent_buffer *eb,
+				       unsigned long p,
+				       int item_size, struct btrfs_key *key,
+				       int max, int *slot)
 {
 	int low = 0;
 	int high = max;
@@ -765,7 +766,7 @@ static int bin_search(struct extent_buffer *eb, struct btrfs_key *key,
 	return -1;
 }
 
-static struct extent_buffer *read_node_slot(struct btrfs_root *root,
+static noinline struct extent_buffer *read_node_slot(struct btrfs_root *root,
 				   struct extent_buffer *parent, int slot)
 {
 	int level = btrfs_header_level(parent);
@@ -781,7 +782,7 @@ static struct extent_buffer *read_node_slot(struct btrfs_root *root,
 		       btrfs_node_ptr_generation(parent, slot));
 }
 
-static int balance_level(struct btrfs_trans_handle *trans,
+static noinline int balance_level(struct btrfs_trans_handle *trans,
 			 struct btrfs_root *root,
 			 struct btrfs_path *path, int level)
 {
@@ -1128,8 +1129,9 @@ static int noinline push_nodes_for_insert(struct btrfs_trans_handle *trans,
 /*
  * readahead one full node of leaves
  */
-static void reada_for_search(struct btrfs_root *root, struct btrfs_path *path,
-			     int level, int slot, u64 objectid)
+static noinline void reada_for_search(struct btrfs_root *root,
+				      struct btrfs_path *path,
+				      int level, int slot, u64 objectid)
 {
 	struct extent_buffer *node;
 	struct btrfs_disk_key disk_key;
@@ -1201,7 +1203,8 @@ static void reada_for_search(struct btrfs_root *root, struct btrfs_path *path,
 	}
 }
 
-static void unlock_up(struct btrfs_path *path, int level, int lowest_unlock)
+static noinline void unlock_up(struct btrfs_path *path, int level,
+			       int lowest_unlock)
 {
 	int i;
 	int skip_level = level;
@@ -1759,8 +1762,9 @@ static int insert_ptr(struct btrfs_trans_handle *trans, struct btrfs_root
  *
  * returns 0 on success and < 0 on failure
  */
-static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root
-		      *root, struct btrfs_path *path, int level)
+static noinline int split_node(struct btrfs_trans_handle *trans,
+			       struct btrfs_root *root,
+			       struct btrfs_path *path, int level)
 {
 	u64 root_gen;
 	struct extent_buffer *c;
@@ -1874,7 +1878,8 @@ static int leaf_space_used(struct extent_buffer *l, int start, int nr)
  * the start of the leaf data.  IOW, how much room
  * the leaf has left for both items and data
  */
-int btrfs_leaf_free_space(struct btrfs_root *root, struct extent_buffer *leaf)
+int noinline btrfs_leaf_free_space(struct btrfs_root *root,
+				   struct extent_buffer *leaf)
 {
 	int nritems = btrfs_header_nritems(leaf);
 	int ret;
@@ -2283,9 +2288,11 @@ out:
  *
  * returns 0 if all went well and < 0 on failure.
  */
-static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root
-		      *root, struct btrfs_key *ins_key,
-		      struct btrfs_path *path, int data_size, int extend)
+static noinline int split_leaf(struct btrfs_trans_handle *trans,
+			       struct btrfs_root *root,
+			       struct btrfs_key *ins_key,
+			       struct btrfs_path *path, int data_size,
+			       int extend)
 {
 	u64 root_gen;
 	struct extent_buffer *l;
@@ -3079,6 +3086,7 @@ 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,
 			 u64 min_trans)
 {
@@ -3093,6 +3101,7 @@ int btrfs_search_forward(struct btrfs_root *root, struct btrfs_key *min_key,
 again:
 	cur = btrfs_lock_root_node(root);
 	level = btrfs_header_level(cur);
+	WARN_ON(path->nodes[level]);
 	path->nodes[level] = cur;
 	path->locks[level] = 1;
 
@@ -3107,6 +3116,8 @@ again:
 
 		/* at level = 0, we're done, setup the path and exit */
 		if (level == 0) {
+			if (slot >= nritems)
+				goto find_next_key;
 			ret = 0;
 			path->slots[level] = slot;
 			btrfs_item_key_to_cpu(cur, &found_key, slot);
@@ -3123,6 +3134,8 @@ again:
 			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) {
@@ -3132,6 +3145,14 @@ again:
 			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));
 
@@ -3143,14 +3164,16 @@ again:
 				free_extent_buffer(tmp);
 			slot++;
 		}
+find_next_key:
 		/*
 		 * we didn't find a candidate key in this node, walk forward
 		 * and find another one
 		 */
 		if (slot >= nritems) {
-			ret = btrfs_find_next_key(root, path, min_key, level,
+			path->slots[level] = slot;
+			sret = btrfs_find_next_key(root, path, min_key, level,
 						  cache_only, min_trans);
-			if (ret == 0) {
+			if (sret == 0) {
 				btrfs_release_path(root, path);
 				goto again;
 			} else {
@@ -3351,6 +3374,7 @@ int btrfs_previous_item(struct btrfs_root *root,
 {
 	struct btrfs_key found_key;
 	struct extent_buffer *leaf;
+	u32 nritems;
 	int ret;
 
 	while(1) {
@@ -3362,9 +3386,20 @@ int btrfs_previous_item(struct btrfs_root *root,
 			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.type == type)
 			return 0;
+		if (found_key.objectid < min_objectid)
+			break;
+		if (found_key.objectid == min_objectid &&
+		    found_key.type < type)
+			break;
 	}
 	return 1;
 }
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index b305ae7e10b..6532b60683e 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -77,6 +77,10 @@ struct btrfs_ordered_sum;
 /* orhpan objectid for tracking unlinked/truncated files */
 #define BTRFS_ORPHAN_OBJECTID -5ULL
 
+/* does write ahead logging to speed up fsyncs */
+#define BTRFS_TREE_LOG_OBJECTID -6ULL
+#define BTRFS_TREE_LOG_FIXUP_OBJECTID -7ULL
+
 /*
  * All files have objectids higher than this.
  */
@@ -276,6 +280,7 @@ struct btrfs_super_block {
 	__le64 generation;
 	__le64 root;
 	__le64 chunk_root;
+	__le64 log_root;
 	__le64 total_bytes;
 	__le64 bytes_used;
 	__le64 root_dir_objectid;
@@ -287,6 +292,7 @@ struct btrfs_super_block {
 	__le32 sys_chunk_array_size;
 	u8 root_level;
 	u8 chunk_root_level;
+	u8 log_root_level;
 	struct btrfs_dev_item dev_item;
 	char label[BTRFS_LABEL_SIZE];
 	u8 sys_chunk_array[BTRFS_SYSTEM_CHUNK_ARRAY_SIZE];
@@ -392,7 +398,10 @@ struct btrfs_timespec {
  * make a new item type
  */
 struct btrfs_inode_item {
+	/* nfs style generation number */
 	__le64 generation;
+	/* transid that last touched this inode */
+	__le64 transid;
 	__le64 size;
 	__le64 nblocks;
 	__le64 block_group;
@@ -409,8 +418,13 @@ struct btrfs_inode_item {
 	struct btrfs_timespec otime;
 } __attribute__ ((__packed__));
 
+struct btrfs_dir_log_item {
+	__le64 end;
+} __attribute__ ((__packed__));
+
 struct btrfs_dir_item {
 	struct btrfs_disk_key location;
+	__le64 transid;
 	__le16 data_len;
 	__le16 name_len;
 	u8 type;
@@ -505,6 +519,9 @@ struct btrfs_fs_info {
 	struct btrfs_root *tree_root;
 	struct btrfs_root *chunk_root;
 	struct btrfs_root *dev_root;
+
+	/* the log root tree is a directory of all the other log roots */
+	struct btrfs_root *log_root_tree;
 	struct radix_tree_root fs_roots_radix;
 
 	struct extent_io_tree free_space_cache;
@@ -518,6 +535,7 @@ struct btrfs_fs_info {
 
 	u64 generation;
 	u64 last_trans_committed;
+	u64 last_trans_new_blockgroup;
 	u64 open_ioctl_trans;
 	unsigned long mount_opt;
 	u64 max_extent;
@@ -527,6 +545,9 @@ struct btrfs_fs_info {
 	wait_queue_head_t transaction_throttle;
 	wait_queue_head_t transaction_wait;
 	wait_queue_head_t async_submit_wait;
+
+	wait_queue_head_t tree_log_wait;
+
 	struct btrfs_super_block super_copy;
 	struct btrfs_super_block super_for_commit;
 	struct block_device *__bdev;
@@ -535,6 +556,7 @@ struct btrfs_fs_info {
 	struct backing_dev_info bdi;
 	spinlock_t hash_lock;
 	struct mutex trans_mutex;
+	struct mutex tree_log_mutex;
 	struct mutex transaction_kthread_mutex;
 	struct mutex cleaner_mutex;
 	struct mutex alloc_mutex;
@@ -544,8 +566,13 @@ struct btrfs_fs_info {
 	struct list_head trans_list;
 	struct list_head hashers;
 	struct list_head dead_roots;
+
 	atomic_t nr_async_submits;
 	atomic_t nr_async_bios;
+	atomic_t tree_log_writers;
+	atomic_t tree_log_commit;
+	unsigned long tree_log_batch;
+	u64 tree_log_transid;
 
 	/*
 	 * this is used by the balancing code to wait for all the pending
@@ -583,6 +610,7 @@ struct btrfs_fs_info {
 	struct completion kobj_unregister;
 	int do_barriers;
 	int closing;
+	int log_root_recovering;
 	atomic_t throttles;
 	atomic_t throttle_gen;
 
@@ -596,6 +624,7 @@ struct btrfs_fs_info {
 	u64 delalloc_bytes;
 	u64 last_alloc;
 	u64 last_data_alloc;
+	u64 last_log_alloc;
 
 	spinlock_t ref_cache_lock;
 	u64 total_ref_cache_size;
@@ -632,6 +661,7 @@ struct btrfs_root {
 	struct btrfs_leaf_ref_tree *ref_tree;
 	struct btrfs_leaf_ref_tree ref_tree_struct;
 	struct btrfs_dirty_root *dirty_root;
+	struct btrfs_root *log_root;
 
 	struct btrfs_root_item root_item;
 	struct btrfs_key root_key;
@@ -640,6 +670,7 @@ struct btrfs_root {
 	struct kobject root_kobj;
 	struct completion kobj_unregister;
 	struct mutex objectid_mutex;
+	struct mutex log_mutex;
 
 	u64 objectid;
 	u64 last_trans;
@@ -692,6 +723,8 @@ struct btrfs_root {
  * dir items are the name -> inode pointers in a directory.  There is one
  * for every name in a directory.
  */
+#define BTRFS_DIR_LOG_ITEM_KEY  14
+#define BTRFS_DIR_LOG_INDEX_KEY 15
 #define BTRFS_DIR_ITEM_KEY	16
 #define BTRFS_DIR_INDEX_KEY	17
 /*
@@ -703,7 +736,8 @@ struct btrfs_root {
  */
 #define BTRFS_CSUM_ITEM_KEY	19
 
-/* reserve 20-31 for other file stuff */
+
+/* reserve 21-31 for other file/dir stuff */
 
 /*
  * root items point to tree roots.  There are typically in the root
@@ -938,6 +972,7 @@ BTRFS_SETGET_FUNCS(inode_ref_index, struct btrfs_inode_ref, index, 64);
 
 /* struct btrfs_inode_item */
 BTRFS_SETGET_FUNCS(inode_generation, struct btrfs_inode_item, generation, 64);
+BTRFS_SETGET_FUNCS(inode_transid, struct btrfs_inode_item, transid, 64);
 BTRFS_SETGET_FUNCS(inode_size, struct btrfs_inode_item, size, 64);
 BTRFS_SETGET_FUNCS(inode_nblocks, struct btrfs_inode_item, nblocks, 64);
 BTRFS_SETGET_FUNCS(inode_block_group, struct btrfs_inode_item, block_group, 64);
@@ -1126,10 +1161,13 @@ static inline void btrfs_set_item_key(struct extent_buffer *eb,
 	write_eb_member(eb, item, struct btrfs_item, key, disk_key);
 }
 
+BTRFS_SETGET_FUNCS(dir_log_end, struct btrfs_dir_log_item, end, 64);
+
 /* struct btrfs_dir_item */
 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);
 
 static inline void btrfs_dir_item_key(struct extent_buffer *eb,
 				      struct btrfs_dir_item *item,
@@ -1301,7 +1339,11 @@ BTRFS_SETGET_STACK_FUNCS(super_root_level, struct btrfs_super_block,
 BTRFS_SETGET_STACK_FUNCS(super_chunk_root, struct btrfs_super_block,
 			 chunk_root, 64);
 BTRFS_SETGET_STACK_FUNCS(super_chunk_root_level, struct btrfs_super_block,
-			 chunk_root_level, 64);
+			 chunk_root_level, 8);
+BTRFS_SETGET_STACK_FUNCS(super_log_root, struct btrfs_super_block,
+			 log_root, 64);
+BTRFS_SETGET_STACK_FUNCS(super_log_root_level, struct btrfs_super_block,
+			 log_root_level, 8);
 BTRFS_SETGET_STACK_FUNCS(super_total_bytes, struct btrfs_super_block,
 			 total_bytes, 64);
 BTRFS_SETGET_STACK_FUNCS(super_bytes_used, struct btrfs_super_block,
@@ -1405,6 +1447,12 @@ static inline struct dentry *fdentry(struct file *file) {
 }
 
 /* extent-tree.c */
+int btrfs_lookup_extent(struct btrfs_root *root, struct btrfs_path *path,
+			u64 start, u64 len);
+int btrfs_update_pinned_extents(struct btrfs_root *root,
+				u64 bytenr, u64 num, int pin);
+int btrfs_drop_leaf_ref(struct btrfs_trans_handle *trans,
+			struct btrfs_root *root, struct extent_buffer *leaf);
 int btrfs_cross_ref_exists(struct btrfs_trans_handle *trans,
 			   struct btrfs_root *root,
 			   struct btrfs_key *key, u64 bytenr);
@@ -1448,6 +1496,11 @@ int btrfs_alloc_reserved_extent(struct btrfs_trans_handle *trans,
 				u64 root_objectid, u64 ref_generation,
 				u64 owner, u64 owner_offset,
 				struct btrfs_key *ins);
+int btrfs_alloc_logged_extent(struct btrfs_trans_handle *trans,
+				struct btrfs_root *root,
+				u64 root_objectid, u64 ref_generation,
+				u64 owner, u64 owner_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,
@@ -1488,9 +1541,9 @@ 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);
 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,
 			 u64 min_trans);
-
 int btrfs_cow_block(struct btrfs_trans_handle *trans,
 		    struct btrfs_root *root, struct extent_buffer *buf,
 		    struct extent_buffer *parent, int parent_slot,
@@ -1656,6 +1709,18 @@ int btrfs_csum_truncate(struct btrfs_trans_handle *trans,
 #define PageChecked PageFsMisc
 #endif
 
+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);
+int btrfs_add_link(struct btrfs_trans_handle *trans,
+		   struct inode *parent_inode, struct inode *inode,
+		   const char *name, int name_len, int add_backref, u64 index);
+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 btrfs_set_extent_delalloc(struct inode *inode, u64 start, u64 end);
 int btrfs_writepages(struct address_space *mapping,
@@ -1715,6 +1780,7 @@ int btrfs_update_inode(struct btrfs_trans_handle *trans,
 long btrfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
 
 /* 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 btrfs_check_file(struct btrfs_root *root, struct inode *inode);
 extern struct file_operations btrfs_file_operations;
diff --git a/fs/btrfs/dir-item.c b/fs/btrfs/dir-item.c
index 125094617fe..e4f30090d64 100644
--- a/fs/btrfs/dir-item.c
+++ b/fs/btrfs/dir-item.c
@@ -96,6 +96,7 @@ int btrfs_insert_xattr_item(struct btrfs_trans_handle *trans,
 	btrfs_set_dir_item_key(leaf, dir_item, &disk_key);
 	btrfs_set_dir_type(leaf, dir_item, BTRFS_FT_XATTR);
 	btrfs_set_dir_name_len(leaf, dir_item, name_len);
+	btrfs_set_dir_transid(leaf, dir_item, trans->transid);
 	btrfs_set_dir_data_len(leaf, dir_item, data_len);
 	name_ptr = (unsigned long)(dir_item + 1);
 	data_ptr = (unsigned long)((char *)name_ptr + name_len);
@@ -142,6 +143,7 @@ int btrfs_insert_dir_item(struct btrfs_trans_handle *trans, struct btrfs_root
 	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);
@@ -169,6 +171,7 @@ second_insert:
 	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);
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 8e7a938bfbc..a4373db5967 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -41,6 +41,7 @@
 #include "async-thread.h"
 #include "locking.h"
 #include "ref-cache.h"
+#include "tree-log.h"
 
 #if 0
 static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf)
@@ -694,6 +695,18 @@ struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
 }
 
 
+int btrfs_write_tree_block(struct extent_buffer *buf)
+{
+	return btrfs_fdatawrite_range(buf->first_page->mapping, buf->start,
+				      buf->start + buf->len - 1, WB_SYNC_NONE);
+}
+
+int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
+{
+	return btrfs_wait_on_page_writeback_range(buf->first_page->mapping,
+				  buf->start, buf->start + buf->len -1);
+}
+
 struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
 				      u32 blocksize, u64 parent_transid)
 {
@@ -732,15 +745,6 @@ int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 	return 0;
 }
 
-int wait_on_tree_block_writeback(struct btrfs_root *root,
-				 struct extent_buffer *buf)
-{
-	struct inode *btree_inode = root->fs_info->btree_inode;
-	wait_on_extent_buffer_writeback(&BTRFS_I(btree_inode)->io_tree,
-					buf);
-	return 0;
-}
-
 static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
 			u32 stripesize, struct btrfs_root *root,
 			struct btrfs_fs_info *fs_info,
@@ -771,6 +775,7 @@ static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
 	spin_lock_init(&root->node_lock);
 	spin_lock_init(&root->list_lock);
 	mutex_init(&root->objectid_mutex);
+	mutex_init(&root->log_mutex);
 
 	btrfs_leaf_ref_tree_init(&root->ref_tree_struct);
 	root->ref_tree = &root->ref_tree_struct;
@@ -809,11 +814,74 @@ static int find_and_setup_root(struct btrfs_root *tree_root,
 	return 0;
 }
 
-struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info,
-					       struct btrfs_key *location)
+int btrfs_free_log_root_tree(struct btrfs_trans_handle *trans,
+			     struct btrfs_fs_info *fs_info)
+{
+	struct extent_buffer *eb;
+	int ret;
+
+	if (!fs_info->log_root_tree)
+		return 0;
+
+	eb = fs_info->log_root_tree->node;
+
+	WARN_ON(btrfs_header_level(eb) != 0);
+	WARN_ON(btrfs_header_nritems(eb) != 0);
+
+	ret = btrfs_free_extent(trans, fs_info->tree_root,
+				eb->start, eb->len,
+				BTRFS_TREE_LOG_OBJECTID, 0, 0, 0, 1);
+	BUG_ON(ret);
+
+	free_extent_buffer(eb);
+	kfree(fs_info->log_root_tree);
+	fs_info->log_root_tree = NULL;
+	return 0;
+}
+
+int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
+			     struct btrfs_fs_info *fs_info)
 {
 	struct btrfs_root *root;
 	struct btrfs_root *tree_root = fs_info->tree_root;
+
+	root = kzalloc(sizeof(*root), GFP_NOFS);
+	if (!root)
+		return -ENOMEM;
+
+	__setup_root(tree_root->nodesize, tree_root->leafsize,
+		     tree_root->sectorsize, tree_root->stripesize,
+		     root, fs_info, BTRFS_TREE_LOG_OBJECTID);
+
+	root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
+	root->root_key.type = BTRFS_ROOT_ITEM_KEY;
+	root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
+	root->ref_cows = 0;
+
+	root->node = btrfs_alloc_free_block(trans, root, root->leafsize,
+					    BTRFS_TREE_LOG_OBJECTID,
+					    0, 0, 0, 0, 0);
+
+	btrfs_set_header_nritems(root->node, 0);
+	btrfs_set_header_level(root->node, 0);
+	btrfs_set_header_bytenr(root->node, root->node->start);
+	btrfs_set_header_generation(root->node, trans->transid);
+	btrfs_set_header_owner(root->node, BTRFS_TREE_LOG_OBJECTID);
+
+	write_extent_buffer(root->node, root->fs_info->fsid,
+			    (unsigned long)btrfs_header_fsid(root->node),
+			    BTRFS_FSID_SIZE);
+	btrfs_mark_buffer_dirty(root->node);
+	btrfs_tree_unlock(root->node);
+	fs_info->log_root_tree = root;
+	return 0;
+}
+
+struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root,
+					       struct btrfs_key *location)
+{
+	struct btrfs_root *root;
+	struct btrfs_fs_info *fs_info = tree_root->fs_info;
 	struct btrfs_path *path;
 	struct extent_buffer *l;
 	u64 highest_inode;
@@ -863,11 +931,13 @@ out:
 				     blocksize, 0);
 	BUG_ON(!root->node);
 insert:
-	root->ref_cows = 1;
-	ret = btrfs_find_highest_inode(root, &highest_inode);
-	if (ret == 0) {
-		root->highest_inode = highest_inode;
-		root->last_inode_alloc = highest_inode;
+	if (location->objectid != BTRFS_TREE_LOG_OBJECTID) {
+		root->ref_cows = 1;
+		ret = btrfs_find_highest_inode(root, &highest_inode);
+		if (ret == 0) {
+			root->highest_inode = highest_inode;
+			root->last_inode_alloc = highest_inode;
+		}
 	}
 	return root;
 }
@@ -907,7 +977,7 @@ struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
 	if (root)
 		return root;
 
-	root = btrfs_read_fs_root_no_radix(fs_info, location);
+	root = btrfs_read_fs_root_no_radix(fs_info->tree_root, location);
 	if (IS_ERR(root))
 		return root;
 	ret = radix_tree_insert(&fs_info->fs_roots_radix,
@@ -1250,16 +1320,18 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	u32 blocksize;
 	u32 stripesize;
 	struct buffer_head *bh;
-	struct btrfs_root *extent_root = kmalloc(sizeof(struct btrfs_root),
+	struct btrfs_root *extent_root = kzalloc(sizeof(struct btrfs_root),
 						 GFP_NOFS);
-	struct btrfs_root *tree_root = kmalloc(sizeof(struct btrfs_root),
+	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 = kmalloc(sizeof(struct btrfs_root),
+	struct btrfs_root *chunk_root = kzalloc(sizeof(struct btrfs_root),
 						GFP_NOFS);
-	struct btrfs_root *dev_root = kmalloc(sizeof(struct btrfs_root),
+	struct btrfs_root *dev_root = kzalloc(sizeof(struct btrfs_root),
 					      GFP_NOFS);
+	struct btrfs_root *log_tree_root;
+
 	int ret;
 	int err = -EINVAL;
 
@@ -1343,6 +1415,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
 
 	mutex_init(&fs_info->trans_mutex);
+	mutex_init(&fs_info->tree_log_mutex);
 	mutex_init(&fs_info->drop_mutex);
 	mutex_init(&fs_info->alloc_mutex);
 	mutex_init(&fs_info->chunk_mutex);
@@ -1352,6 +1425,10 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	init_waitqueue_head(&fs_info->transaction_throttle);
 	init_waitqueue_head(&fs_info->transaction_wait);
 	init_waitqueue_head(&fs_info->async_submit_wait);
+	init_waitqueue_head(&fs_info->tree_log_wait);
+	atomic_set(&fs_info->tree_log_commit, 0);
+	atomic_set(&fs_info->tree_log_writers, 0);
+	fs_info->tree_log_transid = 0;
 
 #if 0
 	ret = add_hasher(fs_info, "crc32c");
@@ -1532,7 +1609,26 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	if (!fs_info->transaction_kthread)
 		goto fail_cleaner;
 
+	if (btrfs_super_log_root(disk_super) != 0) {
+		u32 blocksize;
+		u64 bytenr = btrfs_super_log_root(disk_super);
+
+		blocksize =
+		     btrfs_level_size(tree_root,
+				      btrfs_super_log_root_level(disk_super));
 
+		log_tree_root = kzalloc(sizeof(struct btrfs_root),
+						      GFP_NOFS);
+
+		__setup_root(nodesize, leafsize, sectorsize, stripesize,
+			     log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
+
+		log_tree_root->node = read_tree_block(tree_root, bytenr,
+						      blocksize, 0);
+		ret = btrfs_recover_log_trees(log_tree_root);
+		BUG_ON(ret);
+	}
+	fs_info->last_trans_committed = btrfs_super_generation(disk_super);
 	return tree_root;
 
 fail_cleaner:
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 2562a273ae1..6b6fdc697f3 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -45,7 +45,7 @@ struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
 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_fs_info *fs_info,
+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);
@@ -74,4 +74,10 @@ int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
 			extent_submit_bio_hook_t *submit_bio_hook);
 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);
 #endif
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index e63b3b4bed7..646b9148ca2 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -496,6 +496,23 @@ static int match_extent_ref(struct extent_buffer *leaf,
 	return ret == 0;
 }
 
+/* simple helper to search for an existing extent at a given offset */
+int btrfs_lookup_extent(struct btrfs_root *root, struct btrfs_path *path,
+			u64 start, u64 len)
+{
+	int ret;
+	struct btrfs_key key;
+
+	maybe_lock_mutex(root);
+	key.objectid = start;
+	key.offset = len;
+	btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
+	ret = btrfs_search_slot(NULL, root->fs_info->extent_root, &key, path,
+				0, 0);
+	maybe_unlock_mutex(root);
+	return ret;
+}
+
 static int noinline lookup_extent_backref(struct btrfs_trans_handle *trans,
 					  struct btrfs_root *root,
 					  struct btrfs_path *path, u64 bytenr,
@@ -1409,7 +1426,7 @@ static u64 first_logical_byte(struct btrfs_root *root, u64 search_start)
 }
 
 
-static int update_pinned_extents(struct btrfs_root *root,
+int btrfs_update_pinned_extents(struct btrfs_root *root,
 				u64 bytenr, u64 num, int pin)
 {
 	u64 len;
@@ -1492,7 +1509,7 @@ int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans,
 					    EXTENT_DIRTY);
 		if (ret)
 			break;
-		update_pinned_extents(root, start, end + 1 - start, 0);
+		btrfs_update_pinned_extents(root, start, end + 1 - start, 0);
 		clear_extent_dirty(unpin, start, end, GFP_NOFS);
 		set_extent_dirty(free_space_cache, start, end, GFP_NOFS);
 		if (need_resched()) {
@@ -1538,14 +1555,11 @@ static int finish_current_insert(struct btrfs_trans_handle *trans,
 		clear_extent_bits(&info->extent_ins, start, end, EXTENT_LOCKED,
 				  GFP_NOFS);
 
-		eb = btrfs_find_tree_block(extent_root, ins.objectid,
+		eb = btrfs_find_create_tree_block(extent_root, ins.objectid,
 					   ins.offset);
 
-		if (!btrfs_buffer_uptodate(eb, trans->transid)) {
-			mutex_unlock(&extent_root->fs_info->alloc_mutex);
+		if (!btrfs_buffer_uptodate(eb, trans->transid))
 			btrfs_read_buffer(eb, trans->transid);
-			mutex_lock(&extent_root->fs_info->alloc_mutex);
-		}
 
 		btrfs_tree_lock(eb);
 		level = btrfs_header_level(eb);
@@ -1585,13 +1599,20 @@ static int pin_down_bytes(struct btrfs_root *root, u64 bytenr, u32 num_bytes,
 		struct extent_buffer *buf;
 		buf = btrfs_find_tree_block(root, bytenr, num_bytes);
 		if (buf) {
+			/* 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) &&