Btrfs: Add support for multiple devices per filesystem

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

13 files changed, 1563 insertions, 380 deletions

diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile
index 9c61609f371..5ac6f2946c2 100644
--- a/fs/btrfs/Makefile
+++ b/fs/btrfs/Makefile
@@ -6,7 +6,7 @@ btrfs-y := super.o ctree.o extent-tree.o print-tree.o root-tree.o dir-item.o \
 	   hash.o file-item.o inode-item.o inode-map.o disk-io.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
+	   extent_io.o volumes.o
 
 ifeq ($(CONFIG_FS_POSIX_ACL),y)
 btrfs-y += acl.o
diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index ef8809c1e69..70b6ddfe15a 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -70,6 +70,14 @@ void btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p)
 	memset(p, 0, sizeof(*p));
 }
 
+static void add_root_to_dirty_list(struct btrfs_root *root)
+{
+	if (root->track_dirty && list_empty(&root->dirty_list)) {
+		list_add(&root->dirty_list,
+			 &root->fs_info->dirty_cowonly_roots);
+	}
+}
+
 int btrfs_copy_root(struct btrfs_trans_handle *trans,
 		      struct btrfs_root *root,
 		      struct extent_buffer *buf,
@@ -196,6 +204,7 @@ int __btrfs_cow_block(struct btrfs_trans_handle *trans,
 					  root_gen, 0, 0, 1);
 		}
 		free_extent_buffer(buf);
+		add_root_to_dirty_list(root);
 	} else {
 		root_gen = btrfs_header_generation(parent);
 		btrfs_set_node_blockptr(parent, parent_slot,
@@ -241,7 +250,7 @@ int btrfs_cow_block(struct btrfs_trans_handle *trans,
 		return 0;
 	}
 
-	search_start = buf->start & ~((u64)BTRFS_BLOCK_GROUP_SIZE - 1);
+	search_start = buf->start & ~((u64)(1024 * 1024 * 1024) - 1);
 	ret = __btrfs_cow_block(trans, root, buf, parent,
 				 parent_slot, cow_ret, search_start, 0);
 	return ret;
@@ -724,6 +733,7 @@ static int balance_level(struct btrfs_trans_handle *trans,
 		BUG_ON(ret);
 
 		root->node = child;
+		add_root_to_dirty_list(root);
 		path->nodes[level] = NULL;
 		clean_tree_block(trans, root, mid);
 		wait_on_tree_block_writeback(root, mid);
@@ -1369,6 +1379,7 @@ static int noinline insert_new_root(struct btrfs_trans_handle *trans,
 	/* the super has an extra ref to root->node */
 	free_extent_buffer(root->node);
 	root->node = c;
+	add_root_to_dirty_list(root);
 	extent_buffer_get(c);
 	path->nodes[level] = c;
 	path->slots[level] = 0;
@@ -2777,3 +2788,28 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path)
 	}
 	return 0;
 }
+
+int btrfs_previous_item(struct btrfs_root *root,
+			struct btrfs_path *path, u64 min_objectid,
+			int type)
+{
+	struct btrfs_key found_key;
+	struct extent_buffer *leaf;
+	int ret;
+
+	while(1) {
+		if (path->slots[0] == 0) {
+			ret = btrfs_prev_leaf(root, path);
+			if (ret != 0)
+				return ret;
+		} else {
+			path->slots[0]--;
+		}
+		leaf = path->nodes[0];
+		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
+		if (found_key.type == type)
+			return 0;
+	}
+	return 1;
+}
+
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 92d892f9207..1453d995fef 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -40,12 +40,44 @@ extern struct kmem_cache *btrfs_path_cachep;
 #define BTRFS_MAGIC "_B4RfS_M"
 
 #define BTRFS_MAX_LEVEL 8
+
+/* holds pointers to all of the tree roots */
 #define BTRFS_ROOT_TREE_OBJECTID 1ULL
+
+/* stores information about which extents are in use, and reference counts */
 #define BTRFS_EXTENT_TREE_OBJECTID 2ULL
+
+/* one per subvolume, storing files and directories */
 #define BTRFS_FS_TREE_OBJECTID 3ULL
+
+/* directory objectid inside the root tree */
 #define BTRFS_ROOT_TREE_DIR_OBJECTID 4ULL
+
+
+/*
+ * chunk tree stores translations from logical -> physical block numbering
+ * the super block points to the chunk tree
+ */
+#define BTRFS_CHUNK_TREE_OBJECTID 5ULL
+
+/*
+ * stores information about which areas of a given device are in use.
+ * one per device.  The tree of tree roots points to the device tree
+ */
+#define BTRFS_DEV_TREE_OBJECTID 6ULL
+
+/*
+ * All files have objectids higher than this.
+ */
 #define BTRFS_FIRST_FREE_OBJECTID 256ULL
 
+
+/*
+ * the device items go into the chunk tree.  The key is in the form
+ * [ 1 BTRFS_DEV_ITEM_KEY device_id ]
+ */
+#define BTRFS_DEV_ITEMS_OBJECTID 1ULL
+
 /*
  * we can actually store much bigger names, but lets not confuse the rest
  * of linux
@@ -95,6 +127,81 @@ struct btrfs_key {
 	u64 offset;
 } __attribute__ ((__packed__));
 
+struct btrfs_mapping_tree {
+	struct extent_map_tree map_tree;
+};
+
+#define BTRFS_DEV_UUID_SIZE 16
+struct btrfs_dev_item {
+	/* the internal btrfs device id */
+	__le64 devid;
+
+	/* size of the device */
+	__le64 total_bytes;
+
+	/* bytes used */
+	__le64 bytes_used;
+
+	/* optimal io alignment for this device */
+	__le32 io_align;
+
+	/* optimal io width for this device */
+	__le32 io_width;
+
+	/* minimal io size for this device */
+	__le32 sector_size;
+
+	/* the kernel device number */
+	__le64 rdev;
+
+	/* type and info about this device */
+	__le64 type;
+
+	/* partition number, 0 for whole dev */
+	__le32 partition;
+
+	/* length of the name data at the end of the item */
+	__le16 name_len;
+
+	/* physical drive uuid (or lvm uuid) */
+	u8 uuid[BTRFS_DEV_UUID_SIZE];
+	/* name goes here */
+} __attribute__ ((__packed__));
+
+struct btrfs_stripe {
+	__le64 devid;
+	__le64 offset;
+} __attribute__ ((__packed__));
+
+struct btrfs_chunk {
+	__le64 owner;
+	__le64 stripe_len;
+	__le64 type;
+
+	/* optimal io alignment for this chunk */
+	__le32 io_align;
+
+	/* optimal io width for this chunk */
+	__le32 io_width;
+
+	/* minimal io size for this chunk */
+	__le32 sector_size;
+
+	/* 2^16 stripes is quite a lot, a second limit is the size of a single
+	 * item in the btree
+	 */
+	__le16 num_stripes;
+	struct btrfs_stripe stripe;
+	/* additional stripes go here */
+} __attribute__ ((__packed__));
+
+static inline unsigned long btrfs_chunk_item_size(int num_stripes)
+{
+	BUG_ON(num_stripes == 0);
+	return sizeof(struct btrfs_chunk) +
+		sizeof(struct btrfs_stripe) * (num_stripes - 1);
+}
+
 #define BTRFS_FSID_SIZE 16
 /*
  * every tree block (leaf or node) starts with this header.
@@ -119,6 +226,13 @@ struct btrfs_header {
 					sizeof(struct btrfs_item) - \
 					sizeof(struct btrfs_file_extent_item))
 
+
+/*
+ * this is a very generous portion of the super block, giving us
+ * room to translate 14 chunks with 3 stripes each.
+ */
+#define BTRFS_SYSTEM_CHUNK_ARRAY_SIZE 2048
+
 /*
  * the super block basically lists the main trees of the FS
  * it currently lacks any block count etc etc
@@ -131,6 +245,7 @@ struct btrfs_super_block {
 	__le64 magic;
 	__le64 generation;
 	__le64 root;
+	__le64 chunk_root;
 	__le64 total_bytes;
 	__le64 bytes_used;
 	__le64 root_dir_objectid;
@@ -138,7 +253,10 @@ struct btrfs_super_block {
 	__le32 nodesize;
 	__le32 leafsize;
 	__le32 stripesize;
+	__le32 sys_chunk_array_size;
 	u8 root_level;
+	u8 chunk_root_level;
+	u8 sys_chunk_array[BTRFS_SYSTEM_CHUNK_ARRAY_SIZE];
 } __attribute__ ((__packed__));
 
 /*
@@ -208,12 +326,22 @@ struct btrfs_extent_ref {
 	__le64 offset;
 } __attribute__ ((__packed__));
 
+/* dev extents record free space on individual devices.  The owner
+ * field points back to the chunk allocation mapping tree that allocated
+ * the extent
+ */
+struct btrfs_dev_extent {
+	__le64 owner;
+	__le64 length;
+} __attribute__ ((__packed__));
+
+
 struct btrfs_inode_ref {
 	__le16 name_len;
 	/* name goes here */
 } __attribute__ ((__packed__));
 
-struct btrfs_inode_timespec {
+struct btrfs_timespec {
 	__le64 sec;
 	__le32 nsec;
 } __attribute__ ((__packed__));
@@ -231,13 +359,13 @@ struct btrfs_inode_item {
 	__le32 uid;
 	__le32 gid;
 	__le32 mode;
-	__le32 rdev;
+	__le64 rdev;
 	__le16 flags;
 	__le16 compat_flags;
-	struct btrfs_inode_timespec atime;
-	struct btrfs_inode_timespec ctime;
-	struct btrfs_inode_timespec mtime;
-	struct btrfs_inode_timespec otime;
+	struct btrfs_timespec atime;
+	struct btrfs_timespec ctime;
+	struct btrfs_timespec mtime;
+	struct btrfs_timespec otime;
 } __attribute__ ((__packed__));
 
 struct btrfs_dir_item {
@@ -290,29 +418,34 @@ struct btrfs_csum_item {
 	u8 csum;
 } __attribute__ ((__packed__));
 
-/* tag for the radix tree of block groups in ram */
-#define BTRFS_BLOCK_GROUP_SIZE (256 * 1024 * 1024)
-
+/* 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_DATA 1
-#define BTRFS_BLOCK_GROUP_MIXED 2
 
 struct btrfs_block_group_item {
 	__le64 used;
-	u8 flags;
+	__le64 chunk_tree;
+	__le64 chunk_objectid;
+	__le64 flags;
 } __attribute__ ((__packed__));
 
 struct btrfs_block_group_cache {
 	struct btrfs_key key;
 	struct btrfs_block_group_item item;
-	int data;
-	int cached;
 	u64 pinned;
+	u64 flags;
+	int cached;
 };
+
+struct btrfs_device;
 struct btrfs_fs_info {
 	u8 fsid[BTRFS_FSID_SIZE];
 	struct btrfs_root *extent_root;
 	struct btrfs_root *tree_root;
+	struct btrfs_root *chunk_root;
+	struct btrfs_root *dev_root;
 	struct radix_tree_root fs_roots_radix;
 
 	struct extent_io_tree free_space_cache;
@@ -321,6 +454,9 @@ struct btrfs_fs_info {
 	struct extent_io_tree pending_del;
 	struct extent_io_tree extent_ins;
 
+	/* logical->physical extent mapping */
+	struct btrfs_mapping_tree mapping_tree;
+
 	u64 generation;
 	u64 last_trans_committed;
 	unsigned long mount_opt;
@@ -330,6 +466,7 @@ struct btrfs_fs_info {
 	struct btrfs_transaction *running_transaction;
 	struct btrfs_super_block super_copy;
 	struct extent_buffer *sb_buffer;
+	struct block_device *__bdev;
 	struct super_block *sb;
 	struct inode *btree_inode;
 	spinlock_t hash_lock;
@@ -350,12 +487,17 @@ struct btrfs_fs_info {
 	unsigned long throttles;
 
 	u64 total_pinned;
+	struct list_head dirty_cowonly_roots;
+
+	struct list_head devices;
+	struct list_head *last_device;
 	spinlock_t delalloc_lock;
 	spinlock_t new_trans_lock;
 	u64 delalloc_bytes;
 	u64 last_alloc;
 	u64 last_data_alloc;
 };
+
 /*
  * in ram representation of the tree.  extent_root is used for all allocations
  * and for the extent tree extent_root root.
@@ -387,14 +529,19 @@ struct btrfs_root {
 	u64 highest_inode;
 	u64 last_inode_alloc;
 	int ref_cows;
+	int track_dirty;
 	struct btrfs_key defrag_progress;
 	int defrag_running;
 	int defrag_level;
 	char *name;
 	int in_sysfs;
+
+	/* the dirty list is only used by non-reference counted roots */
+	struct list_head dirty_list;
 };
 
 /*
+
  * 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
  * the FS
@@ -439,6 +586,10 @@ struct btrfs_root {
  */
 #define BTRFS_BLOCK_GROUP_ITEM_KEY 50
 
+#define BTRFS_DEV_EXTENT_KEY	75
+#define BTRFS_DEV_ITEM_KEY	76
+#define BTRFS_CHUNK_ITEM_KEY	77
+
 /*
  * string items are for debugging.  They just store a short string of
  * data in the FS
@@ -518,13 +669,104 @@ static inline void btrfs_set_##name(type *s, u##bits val)		\
 	s->member = cpu_to_le##bits(val);				\
 }
 
+BTRFS_SETGET_FUNCS(device_type, struct btrfs_dev_item, type, 64);
+BTRFS_SETGET_FUNCS(device_total_bytes, struct btrfs_dev_item, total_bytes, 64);
+BTRFS_SETGET_FUNCS(device_bytes_used, struct btrfs_dev_item, bytes_used, 64);
+BTRFS_SETGET_FUNCS(device_io_align, struct btrfs_dev_item, io_align, 32);
+BTRFS_SETGET_FUNCS(device_io_width, struct btrfs_dev_item, io_width, 32);
+BTRFS_SETGET_FUNCS(device_sector_size, struct btrfs_dev_item, sector_size, 32);
+BTRFS_SETGET_FUNCS(device_id, struct btrfs_dev_item, devid, 64);
+BTRFS_SETGET_FUNCS(device_rdev, struct btrfs_dev_item, rdev, 64);
+BTRFS_SETGET_FUNCS(device_partition, struct btrfs_dev_item, partition, 32);
+BTRFS_SETGET_FUNCS(device_name_len, struct btrfs_dev_item, name_len, 16);
+
+static inline char *btrfs_device_uuid(struct btrfs_dev_item *d)
+{
+	return (char *)d + offsetof(struct btrfs_dev_item, uuid);
+}
+
+static inline char *btrfs_device_name(struct btrfs_dev_item *d)
+{
+	return (char *)(d + 1);
+}
+
+BTRFS_SETGET_FUNCS(chunk_owner, struct btrfs_chunk, owner, 64);
+BTRFS_SETGET_FUNCS(chunk_stripe_len, struct btrfs_chunk, stripe_len, 64);
+BTRFS_SETGET_FUNCS(chunk_io_align, struct btrfs_chunk, io_align, 32);
+BTRFS_SETGET_FUNCS(chunk_io_width, struct btrfs_chunk, io_width, 32);
+BTRFS_SETGET_FUNCS(chunk_sector_size, struct btrfs_chunk, sector_size, 32);
+BTRFS_SETGET_FUNCS(chunk_type, struct btrfs_chunk, type, 64);
+BTRFS_SETGET_FUNCS(chunk_num_stripes, struct btrfs_chunk, num_stripes, 16);
+BTRFS_SETGET_FUNCS(stripe_devid, struct btrfs_stripe, devid, 64);
+BTRFS_SETGET_FUNCS(stripe_offset, struct btrfs_stripe, offset, 64);
+
+BTRFS_SETGET_STACK_FUNCS(stack_chunk_owner, struct btrfs_chunk, owner, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_chunk_stripe_len, struct btrfs_chunk,
+			 stripe_len, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_chunk_io_align, struct btrfs_chunk,
+			 io_align, 32);
+BTRFS_SETGET_STACK_FUNCS(stack_chunk_io_width, struct btrfs_chunk,
+			 io_width, 32);
+BTRFS_SETGET_STACK_FUNCS(stack_chunk_sector_size, struct btrfs_chunk,
+			 sector_size, 32);
+BTRFS_SETGET_STACK_FUNCS(stack_chunk_type, struct btrfs_chunk, type, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_chunk_num_stripes, struct btrfs_chunk,
+			 num_stripes, 16);
+BTRFS_SETGET_STACK_FUNCS(stack_stripe_devid, struct btrfs_stripe, devid, 64);
+BTRFS_SETGET_STACK_FUNCS(stack_stripe_offset, struct btrfs_stripe, offset, 64);
+
+static inline struct btrfs_stripe *btrfs_stripe_nr(struct btrfs_chunk *c,
+						   int nr)
+{
+	unsigned long offset = (unsigned long)c;
+	offset += offsetof(struct btrfs_chunk, stripe);
+	offset += nr * sizeof(struct btrfs_stripe);
+	return (struct btrfs_stripe *)offset;
+}
+
+static inline u64 btrfs_stripe_offset_nr(struct extent_buffer *eb,
+					 struct btrfs_chunk *c, int nr)
+{
+	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);
 BTRFS_SETGET_FUNCS(disk_block_group_used, struct btrfs_block_group_item,
 			 used, 64);
-BTRFS_SETGET_FUNCS(disk_block_group_flags, struct btrfs_block_group_item,
-		   flags, 8);
+BTRFS_SETGET_STACK_FUNCS(block_group_chunk_tree, struct btrfs_block_group_item,
+			 chunk_tree, 64);
+BTRFS_SETGET_FUNCS(disk_block_group_chunk_tree, struct btrfs_block_group_item,
+			 chunk_tree, 64);
+BTRFS_SETGET_STACK_FUNCS(block_group_chunk_objectid,
+			struct btrfs_block_group_item, chunk_objectid, 64);
+BTRFS_SETGET_FUNCS(disk_block_group_chunk_objecitd,
+		   struct btrfs_block_group_item, chunk_objectid, 64);
+BTRFS_SETGET_FUNCS(disk_block_group_flags,
+		   struct btrfs_block_group_item, flags, 64);
+BTRFS_SETGET_STACK_FUNCS(block_group_flags,
+			struct btrfs_block_group_item, flags, 64);
 
 /* struct btrfs_inode_ref */
 BTRFS_SETGET_FUNCS(inode_ref_name_len, struct btrfs_inode_ref, name_len, 16);
@@ -538,49 +780,53 @@ BTRFS_SETGET_FUNCS(inode_nlink, struct btrfs_inode_item, nlink, 32);
 BTRFS_SETGET_FUNCS(inode_uid, struct btrfs_inode_item, uid, 32);
 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, 32);
+BTRFS_SETGET_FUNCS(inode_rdev, struct btrfs_inode_item, rdev, 64);
 BTRFS_SETGET_FUNCS(inode_flags, struct btrfs_inode_item, flags, 16);
 BTRFS_SETGET_FUNCS(inode_compat_flags, struct btrfs_inode_item,
 		   compat_flags, 16);
 
-static inline struct btrfs_inode_timespec *
+static inline struct btrfs_timespec *
 btrfs_inode_atime(struct btrfs_inode_item *inode_item)
 {
 	unsigned long ptr = (unsigned long)inode_item;
 	ptr += offsetof(struct btrfs_inode_item, atime);
-	return (struct btrfs_inode_timespec *)ptr;
+	return (struct btrfs_timespec *)ptr;
 }
 
-static inline struct btrfs_inode_timespec *
+static inline struct btrfs_timespec *
 btrfs_inode_mtime(struct btrfs_inode_item *inode_item)
 {
 	unsigned long ptr = (unsigned long)inode_item;
 	ptr += offsetof(struct btrfs_inode_item, mtime);
-	return (struct btrfs_inode_timespec *)ptr;
+	return (struct btrfs_timespec *)ptr;
 }
 
-static inline struct btrfs_inode_timespec *
+static inline struct btrfs_timespec *
 btrfs_inode_ctime(struct btrfs_inode_item *inode_item)
 {
 	unsigned long ptr = (unsigned long)inode_item;
 	ptr += offsetof(struct btrfs_inode_item, ctime);
-	return (struct btrfs_inode_timespec *)ptr;
+	return (struct btrfs_timespec *)ptr;
 }
 
-static inline struct btrfs_inode_timespec *
+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_inode_timespec *)ptr;
+	return (struct btrfs_timespec *)ptr;
 }
 
-BTRFS_SETGET_FUNCS(timespec_sec, struct btrfs_inode_timespec, sec, 64);
-BTRFS_SETGET_FUNCS(timespec_nsec, struct btrfs_inode_timespec, nsec, 32);
+BTRFS_SETGET_FUNCS(timespec_sec, struct btrfs_timespec, sec, 64);
+BTRFS_SETGET_FUNCS(timespec_nsec, struct btrfs_timespec, nsec, 32);
 
 /* struct btrfs_extent_item */
 BTRFS_SETGET_FUNCS(extent_refs, struct btrfs_extent_item, refs, 32);
 
+/* struct btrfs_dev_extent */
+BTRFS_SETGET_FUNCS(dev_extent_owner, struct btrfs_dev_extent, owner, 64);
+BTRFS_SETGET_FUNCS(dev_extent_length, struct btrfs_dev_extent, length, 64);
+
 /* struct btrfs_extent_ref */
 BTRFS_SETGET_FUNCS(ref_root, struct btrfs_extent_ref, root, 64);
 BTRFS_SETGET_FUNCS(ref_generation, struct btrfs_extent_ref, generation, 64);
@@ -846,8 +1092,14 @@ BTRFS_SETGET_STACK_FUNCS(super_bytenr, struct btrfs_super_block, bytenr, 64);
 BTRFS_SETGET_STACK_FUNCS(super_generation, struct btrfs_super_block,
 			 generation, 64);
 BTRFS_SETGET_STACK_FUNCS(super_root, struct btrfs_super_block, root, 64);
+BTRFS_SETGET_STACK_FUNCS(super_sys_array_size,
+			 struct btrfs_super_block, sys_chunk_array_size, 32);
 BTRFS_SETGET_STACK_FUNCS(super_root_level, struct btrfs_super_block,
 			 root_level, 8);
+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);
 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,
@@ -1009,7 +1261,14 @@ int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans,
 				    struct btrfs_root *root);
 int btrfs_free_block_groups(struct btrfs_fs_info *info);
 int btrfs_read_block_groups(struct btrfs_root *root);
+int btrfs_make_block_group(struct btrfs_trans_handle *trans,
+			   struct btrfs_root *root, u64 bytes_used,
+			   u64 type, u64 chunk_tree, u64 chunk_objectid,
+			   u64 size);
 /* ctree.c */
+int btrfs_previous_item(struct btrfs_root *root,
+			struct btrfs_path *path, u64 min_objectid,
+			int type);
 int btrfs_cow_block(struct btrfs_trans_handle *trans,
 		    struct btrfs_root *root, struct extent_buffer *buf,
 		    struct extent_buffer *parent, int parent_slot,
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 88e21bdbc47..8e37fa120cc 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -28,6 +28,7 @@
 #include "disk-io.h"
 #include "transaction.h"
 #include "btrfs_inode.h"
+#include "volumes.h"
 #include "print-tree.h"
 
 #if 0
@@ -234,6 +235,19 @@ static int btree_writepage_io_hook(struct page *page, u64 start, u64 end)
 	return 0;
 }
 
+static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio)
+{
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	u64 offset;
+	offset = bio->bi_sector << 9;
+	if (offset == BTRFS_SUPER_INFO_OFFSET) {
+		bio->bi_bdev = root->fs_info->sb->s_bdev;
+		submit_bio(rw, bio);
+		return 0;
+	}
+	return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio);
+}
+
 static int btree_writepage(struct page *page, struct writeback_control *wbc)
 {
 	struct extent_io_tree *tree;
@@ -345,6 +359,23 @@ int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize)
 	return ret;
 }
 
+static int close_all_devices(struct btrfs_fs_info *fs_info)
+{
+	struct list_head *list;
+	struct list_head *next;
+	struct btrfs_device *device;
+
+	list = &fs_info->devices;
+	while(!list_empty(list)) {
+		next = list->next;
+		list_del(next);
+		device = list_entry(next, struct btrfs_device, dev_list);
+		kfree(device->name);
+		kfree(device);
+	}
+	return 0;
+}
+
 struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
 				      u32 blocksize)
 {
@@ -420,6 +451,8 @@ static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
 	root->leafsize = leafsize;
 	root->stripesize = stripesize;
 	root->ref_cows = 0;
+	root->track_dirty = 0;
+
 	root->fs_info = fs_info;
 	root->objectid = objectid;
 	root->last_trans = 0;
@@ -427,6 +460,8 @@ static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
 	root->last_inode_alloc = 0;
 	root->name = NULL;
 	root->in_sysfs = 0;
+
+	INIT_LIST_HEAD(&root->dirty_list);
 	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));
@@ -634,6 +669,10 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 					       GFP_NOFS);
 	struct btrfs_fs_info *fs_info = kmalloc(sizeof(*fs_info),
 						GFP_NOFS);
+	struct btrfs_root *chunk_root = kmalloc(sizeof(struct btrfs_root),
+						GFP_NOFS);
+	struct btrfs_root *dev_root = kmalloc(sizeof(struct btrfs_root),
+					      GFP_NOFS);
 	int ret;
 	int err = -EIO;
 	struct btrfs_super_block *disk_super;
@@ -657,6 +696,12 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 	fs_info->last_trans_committed = 0;
 	fs_info->tree_root = tree_root;
 	fs_info->extent_root = extent_root;
+	fs_info->chunk_root = chunk_root;
+	fs_info->dev_root = dev_root;
+	INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
+	INIT_LIST_HEAD(&fs_info->devices);
+	btrfs_mapping_init(&fs_info->mapping_tree);
+	fs_info->last_device = &fs_info->devices;
 	fs_info->sb = sb;
 	fs_info->throttles = 0;
 	fs_info->mount_opt = 0;
@@ -714,12 +759,12 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 		goto fail_iput;
 	}
 #endif
-	__setup_root(512, 512, 512, 512, tree_root,
+	__setup_root(4096, 4096, 4096, 4096, tree_root,
 		     fs_info, BTRFS_ROOT_TREE_OBJECTID);
 
 	fs_info->sb_buffer = read_tree_block(tree_root,
 					     BTRFS_SUPER_INFO_OFFSET,
-					     512);
+					     4096);
 
 	if (!fs_info->sb_buffer)
 		goto fail_iput;
@@ -730,6 +775,7 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 	read_extent_buffer(fs_info->sb_buffer, fs_info->fsid,
 			   (unsigned long)btrfs_super_fsid(fs_info->sb_buffer),
 			   BTRFS_FSID_SIZE);
+
 	disk_super = &fs_info->super_copy;
 	if (!btrfs_super_root(disk_super))
 		goto fail_sb_buffer;
@@ -753,23 +799,47 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 		goto fail_sb_buffer;
 	}
 
+	mutex_lock(&fs_info->fs_mutex);
+	ret = btrfs_read_sys_array(tree_root);
+	BUG_ON(ret);
+
+	blocksize = btrfs_level_size(tree_root,
+				     btrfs_super_chunk_root_level(disk_super));
+
+	__setup_root(nodesize, leafsize, sectorsize, stripesize,
+		     chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
+
+	chunk_root->node = read_tree_block(chunk_root,
+					   btrfs_super_chunk_root(disk_super),
+					   blocksize);
+	BUG_ON(!chunk_root->node);
+
+	ret = btrfs_read_chunk_tree(chunk_root);
+	BUG_ON(ret);
+
 	blocksize = btrfs_level_size(tree_root,
 				     btrfs_super_root_level(disk_super));
 
+
 	tree_root->node = read_tree_block(tree_root,
 					  btrfs_super_root(disk_super),
 					  blocksize);
 	if (!tree_root->node)
 		goto fail_sb_buffer;
 
-	mutex_lock(&fs_info->fs_mutex);
 
 	ret = find_and_setup_root(tree_root, fs_info,
 				  BTRFS_EXTENT_TREE_OBJECTID, extent_root);
-	if (ret) {
-		mutex_unlock(&fs_info->fs_mutex);
+	if (ret)
 		goto fail_tree_root;
-	}
+	extent_root->track_dirty = 1;
+
+	ret = find_and_setup_root(tree_root, fs_info,
+				  BTRFS_DEV_TREE_OBJECTID, dev_root);
+	dev_root->track_dirty = 1;
+
+	if (ret)
+		goto fail_extent_root;
 
 	btrfs_read_block_groups(extent_root);
 
@@ -777,7 +847,10 @@ struct btrfs_root *open_ctree(struct super_block *sb)
 	mutex_unlock(&fs_info->fs_mutex);
 	return tree_root;
 
+fail_extent_root:
+	free_extent_buffer(extent_root->node);
 fail_tree_root:
+	mutex_unlock(&fs_info->fs_mutex);
 	free_extent_buffer(tree_root->node);
 fail_sb_buffer:
 	free_extent_buffer(fs_info->sb_buffer);
@@ -874,6 +947,12 @@ int close_ctree(struct btrfs_root *root)
 	if (fs_info->tree_root->node)
 		free_extent_buffer(fs_info->tree_root->node);
 
+	if (root->fs_info->chunk_root->node);
+		free_extent_buffer(root->fs_info->chunk_root->node);
+
+	if (root->fs_info->dev_root->node);
+		free_extent_buffer(root->fs_info->dev_root->node);
+
 	free_extent_buffer(fs_info->sb_buffer);
 
 	btrfs_free_block_groups(root->fs_info);
@@ -901,8 +980,13 @@ int close_ctree(struct btrfs_root *root)
 		kfree(hasher);
 	}
 #endif
+	close_all_devices(fs_info);
+	btrfs_mapping_tree_free(&fs_info->mapping_tree);
+
 	kfree(fs_info->extent_root);
 	kfree(fs_info->tree_root);
+	kfree(fs_info->chunk_root);
+	kfree(fs_info->dev_root);
 	return 0;
 }
 
@@ -1016,4 +1100,5 @@ int btrfs_read_buffer(struct extent_buffer *buf)
 
 static struct extent_io_ops btree_extent_io_ops = {
 	.writepage_io_hook = btree_writepage_io_hook,
+	.submit_bio_hook = btree_submit_bio_hook,
 };
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 828f3a2081b..206cb48638f 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -20,6 +20,7 @@
 #define __DISKIO__
 
 #define BTRFS_SUPER_INFO_OFFSET (16 * 1024)
+struct btrfs_device;
 
 struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
 				      u32 blocksize);
@@ -65,4 +66,5 @@ int btrfs_read_buffer(struct extent_buffer *buf);
 u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len);
 void btrfs_csum_final(u32 crc, char *result);
 void btrfs_throttle(struct btrfs_root *root);
+int btrfs_open_device(struct btrfs_device *dev);
 #endif
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index ebfd304138c..2cd957d6e8d 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -24,37 +24,19 @@
 #include "disk-io.h"
 #include "print-tree.h"
 #include "transaction.h"
+#include "volumes.h"
 
-#define BLOCK_GROUP_DATA EXTENT_WRITEBACK
+#define BLOCK_GROUP_DATA     EXTENT_WRITEBACK
 #define BLOCK_GROUP_METADATA EXTENT_UPTODATE
+#define BLOCK_GROUP_SYSTEM   EXTENT_NEW
+
 #define BLOCK_GROUP_DIRTY EXTENT_DIRTY
 
 static int finish_current_insert(struct btrfs_trans_handle *trans, struct
 				 btrfs_root *extent_root);
 static int del_pending_extents(struct btrfs_trans_handle *trans, struct
 			       btrfs_root *extent_root);
-static int find_previous_extent(struct btrfs_root *root,
-				struct btrfs_path *path)
-{
-	struct btrfs_key found_key;
-	struct extent_buffer *leaf;
-	int ret;
 
-	while(1) {
-		if (path->slots[0] == 0) {
-			ret = btrfs_prev_leaf(root, path);
-			if (ret != 0)
-				return ret;
-		} else {
-			path->slots[0]--;
-		}
-		leaf = path->nodes[0];
-		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
-		if (found_key.type == BTRFS_EXTENT_ITEM_KEY)
-			return 0;
-	}
-	return 1;
-}
 
 static int cache_block_group(struct btrfs_root *root,
 			     struct btrfs_block_group_cache *block_group)
@@ -91,7 +73,7 @@ static int cache_block_group(struct btrfs_root *root,
 	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
 	if (ret < 0)
 		return ret;
-	ret = find_previous_extent(root, path);
+	ret = btrfs_previous_item(root, path, 0, BTRFS_EXTENT_ITEM_KEY);
 	if (ret < 0)
 		return ret;
 	if (ret == 0) {
@@ -168,7 +150,8 @@ struct btrfs_block_group_cache *btrfs_lookup_block_group(struct
 	block_group_cache = &info->block_group_cache;
 	ret = find_first_extent_bit(block_group_cache,
 				    bytenr, &start, &end,
-				    BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA);
+				    BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA |
+				    BLOCK_GROUP_SYSTEM);
 	if (ret) {
 		return NULL;
 	}
@@ -182,23 +165,38 @@ struct btrfs_block_group_cache *btrfs_lookup_block_group(struct
 		return block_group;
 	return NULL;
 }
-static u64 noinline find_search_start(struct btrfs_root *root,
+
+static int block_group_bits(struct btrfs_block_group_cache *cache, u64 bits)
+{
+	if ((bits & BLOCK_GROUP_DATA) &&
+	    (cache->flags & BTRFS_BLOCK_GROUP_DATA))
+		return 1;
+	if ((bits & BLOCK_GROUP_METADATA) &&
+	     (cache->flags & BTRFS_BLOCK_GROUP_METADATA))
+		return 1;
+	if ((bits & BLOCK_GROUP_SYSTEM) &&
+	     (cache->flags & BTRFS_BLOCK_GROUP_SYSTEM))
+		return 1;
+	return 0;
+}
+
+static int noinline find_search_start(struct btrfs_root *root,
 			      struct btrfs_block_group_cache **cache_ret,
-			      u64 search_start, int num, int data)
+			      u64 *start_ret, int num, int data)
 {
 	int ret;
 	struct btrfs_block_group_cache *cache = *cache_ret;
 	struct extent_io_tree *free_space_cache;
-	struct extent_state *state;
 	u64 last;
 	u64 start = 0;
+	u64 end = 0;
 	u64 cache_miss = 0;
 	u64 total_fs_bytes;
+	u64 search_start = *start_ret;
 	int wrapped = 0;
 
-	if (!cache) {
+	if (!cache)
 		goto out;
-	}
 	total_fs_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy);
 	free_space_cache = &root->fs_info->free_space_cache;
 
@@ -208,6 +206,9 @@ again:
 		goto out;
 
 	last = max(search_start, cache->key.objectid);
+	if (!block_group_bits(cache, data)) {
+		goto new_group;
+	}
 
 	while(1) {
 		ret = find_first_extent_bit(&root->fs_info->free_space_cache,
@@ -225,22 +226,20 @@ again:
 				cache_miss = start;
 			continue;
 		}
-		if (data != BTRFS_BLOCK_GROUP_MIXED &&
-		    start + num > cache->key.objectid + cache->key.offset)
+		if (start + num > cache->key.objectid + cache->key.offset)
 			goto new_group;
 		if (start + num  > total_fs_bytes)
 			goto new_group;
-		return start;
+		*start_ret = start;
+		return 0;
 	}
 out:
 	cache = btrfs_lookup_block_group(root->fs_info, search_start);
 	if (!cache) {
-		printk("Unable to find block group for %Lu\n",
-		       search_start);
+		printk("Unable to find block group for %Lu\n", search_start);
 		WARN_ON(1);
-		return search_start;
 	}
-	return search_start