Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/btrfs-unstable

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/btrfs-unstable: (39 commits)
Btrfs: deal with errors from updating the tree log
Btrfs: allow subvol deletion by unprivileged user with -o user_subvol_rm_allowed
Btrfs: make SNAP_DESTROY async
Btrfs: add SNAP_CREATE_ASYNC ioctl
Btrfs: add START_SYNC, WAIT_SYNC ioctls
Btrfs: async transaction commit
Btrfs: fix deadlock in btrfs_commit_transaction
Btrfs: fix lockdep warning on clone ioctl
Btrfs: fix clone ioctl where range is adjacent to extent
Btrfs: fix delalloc checks in clone ioctl
Btrfs: drop unused variable in block_alloc_rsv
Btrfs: cleanup warnings from gcc 4.6 (nonbugs)
Btrfs: Fix variables set but not read (bugs found by gcc 4.6)
Btrfs: Use ERR_CAST helpers
Btrfs: use memdup_user helpers
Btrfs: fix raid code for removing missing drives
Btrfs: Switch the extent buffer rbtree into a radix tree
Btrfs: restructure try_release_extent_buffer()
Btrfs: use the flusher threads for delalloc throttling
Btrfs: tune the chunk allocation to 5% of the FS as metadata
...

Fix up trivial conflicts in fs/btrfs/super.c and fs/fs-writeback.c, and
remove use of INIT_RCU_HEAD in fs/btrfs/extent_io.c (that init macro was
useless and removed in commit 5e8067adfdba: "rcu head remove init")

fs/btrfs/compression.c

···
  */
 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;
···
 	/* 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,

fs/btrfs/ctree.c

···
 		      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->last_trans);
 
 	level = btrfs_header_level(buf);
-	nritems = btrfs_header_nritems(buf);
 	if (level == 0)
 		btrfs_item_key(buf, &disk_key, 0);
 	else
···
 	int wret;
 	int pslot;
 	int orig_slot = path->slots[level];
-	int err_on_enospc = 0;
 	u64 orig_ptr;
 
 	if (level == 0)
···
 	    BTRFS_NODEPTRS_PER_BLOCK(root) / 4)
 		return 0;
 
-	if (btrfs_header_nritems(mid) < 2)
-		err_on_enospc = 1;
+	btrfs_header_nritems(mid);
 
 	left = read_node_slot(root, parent, pslot - 1);
 	if (left) {
···
 		wret = push_node_left(trans, root, left, mid, 1);
 		if (wret < 0)
 			ret = wret;
-		if (btrfs_header_nritems(mid) < 2)
-			err_on_enospc = 1;
+		btrfs_header_nritems(mid);
 	}
 
 	/*
···
 	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)
 		parent = path->nodes[level + 1];
···
 	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
-		 */
-		*eb_ret = tmp;
-		return 0;
+	if (tmp) {
+		if (btrfs_buffer_uptodate(tmp, 0)) {
+			if (btrfs_buffer_uptodate(tmp, gen)) {
+				/*
+				 * we found an up to date block without
+				 * sleeping, return
+				 * right away
+				 */
+				*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
+			 */
+			free_extent_buffer(tmp);
+			tmp = read_tree_block(root, blocknr, blocksize, gen);
+			if (tmp && btrfs_buffer_uptodate(tmp, gen)) {
+				*eb_ret = tmp;
+				return 0;
+			}
+			free_extent_buffer(tmp);
+			btrfs_release_path(NULL, p);
+			return -EIO;
+		}
 	}
 
 	/*
···
 	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);
 
···
 {
 	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;
···
 	int wret;
 	u32 this_item_size;
 	u32 old_left_item_size;
-
-	slot = path->slots[1];
 
 	if (empty)
 		nr = min(right_nritems, max_slot);
···
 {
 	int ret = 0;
 	int slot;
-	int slot_orig;
 	struct extent_buffer *leaf;
 	struct btrfs_item *item;
 	u32 nritems;
···
 	unsigned int size_diff;
 	int i;
 
-	slot_orig = path->slots[0];
 	leaf = path->nodes[0];
 	slot = path->slots[0];
 
···
 {
 	int ret = 0;
 	int slot;
-	int slot_orig;
 	struct extent_buffer *leaf;
 	struct btrfs_item *item;
 	u32 nritems;
···
 	unsigned int old_size;
 	int i;
 
-	slot_orig = path->slots[0];
 	leaf = path->nodes[0];
 
 	nritems = btrfs_header_nritems(leaf);
···
 			    struct btrfs_key *cpu_key, u32 *data_size,
 			    int nr)
 {
-	struct extent_buffer *leaf;
 	int ret = 0;
 	int slot;
 	int i;
···
 	if (ret < 0)
 		goto out;
 
-	leaf = path->nodes[0];
 	slot = path->slots[0];
 	BUG_ON(slot < 0);
 

fs/btrfs/ctree.h

···
  */
 #define BTRFS_EXTENT_CSUM_OBJECTID -10ULL
 
+/* For storing free space cache */
+#define BTRFS_FREE_SPACE_OBJECTID -11ULL
+
 /* dummy objectid represents multiple objectids */
 #define BTRFS_MULTIPLE_OBJECTIDS -255ULL
 
···
 	/* 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);
···
 
 	char label[BTRFS_LABEL_SIZE];
 
+	__le64 cache_generation;
+
 	/* future expansion */
-	__le64 reserved[32];
+	__le64 reserved[31];
 	u8 sys_chunk_array[BTRFS_SYSTEM_CHUNK_ARRAY_SIZE];
 } __attribute__ ((__packed__));
 
···
  * 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_COMPAT_SUPP		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_INCOMPAT_SUPP			\
+	(BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF |		\
+	 BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL |	\
+	 BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS)
 
 /*
  * A leaf is full of items. offset and size tell us where to find
···
 struct btrfs_space_info {
 	u64 flags;
 
-	u64 total_bytes;	/* total bytes in the space */
+	u64 total_bytes;	/* total bytes in the space,
+				   this doesn't take mirrors into account */
 	u64 bytes_used;		/* total bytes used,
 				   this does't take mirrors into account */
 	u64 bytes_pinned;	/* total bytes pinned, will be freed when the
···
 	u64 bytes_may_use;	/* number of bytes that may be used for
 				   delalloc/allocations */
 	u64 disk_used;		/* total bytes used on disk */
+	u64 disk_total;		/* total bytes on disk, takes mirrors into
+				   account */
 
 	int full;		/* indicates that we cannot allocate any more
 				   chunks for this space */
···
 	BTRFS_CACHE_FINISHED	= 2,
 };
 
+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;
···
 	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;
···
 	int extents_thresh;
 	int free_extents;
 	int total_bitmaps;
-	int ro;
-	int dirty;
+	int ro:1;
+	int dirty:1;
+	int iref:1;
+
+	int disk_cache_state;
 
 	/* cache tracking stuff */
 	int cached;
···
 	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_workers endio_meta_workers;
 	struct btrfs_workers endio_meta_write_workers;
 	struct btrfs_workers endio_write_workers;
+	struct btrfs_workers endio_freespace_worker;
 	struct btrfs_workers submit_workers;
 	/*
 	 * fixup workers take dirty pages that didn't properly go through
···
 #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_clear_opt(o, opt)		((o) &= ~BTRFS_MOUNT_##opt)
 #define btrfs_set_opt(o, opt)		((o) |= BTRFS_MOUNT_##opt)
···
 	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);
···
 			 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);
 
 static inline int btrfs_super_csum_size(struct btrfs_super_block *s)
 {
···
 	return file->f_path.dentry;
 }
 
+static inline bool btrfs_mixed_space_info(struct btrfs_space_info *space_info)
+{
+	return ((space_info->flags & BTRFS_BLOCK_GROUP_METADATA) &&
+		(space_info->flags & BTRFS_BLOCK_GROUP_DATA));
+}
+
 /* extent-tree.c */
 void btrfs_put_block_group(struct btrfs_block_group_cache *cache);
 int btrfs_run_delayed_refs(struct btrfs_trans_handle *trans,
···
 void btrfs_free_reserved_data_space(struct inode *inode, u64 bytes);
 int btrfs_trans_reserve_metadata(struct btrfs_trans_handle *trans,
 				struct btrfs_root *root,
-				int num_items, int *retries);
+				int num_items);
 void btrfs_trans_release_metadata(struct btrfs_trans_handle *trans,
 				struct btrfs_root *root);
 int btrfs_orphan_reserve_metadata(struct btrfs_trans_handle *trans,
···
 int btrfs_block_rsv_add(struct btrfs_trans_handle *trans,
 			struct btrfs_root *root,
 			struct btrfs_block_rsv *block_rsv,
-			u64 num_bytes, int *retries);
+			u64 num_bytes);
 int btrfs_block_rsv_check(struct btrfs_trans_handle *trans,
 			  struct btrfs_root *root,
 			  struct btrfs_block_rsv *block_rsv,
···
 			     struct btrfs_block_group_cache *cache);
 int 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);
 /* ctree.c */
 int btrfs_bin_search(struct extent_buffer *eb, struct btrfs_key *key,
 		     int level, int *slot);
···
 			       u32 min_type);
 
 int btrfs_start_delalloc_inodes(struct btrfs_root *root, int delay_iput);
-int btrfs_start_one_delalloc_inode(struct btrfs_root *root, int delay_iput);
+int btrfs_start_one_delalloc_inode(struct btrfs_root *root, int delay_iput,
+				   int sync);
 int btrfs_set_extent_delalloc(struct inode *inode, u64 start, u64 end,
 			      struct extent_state **cached_state);
 int btrfs_writepages(struct address_space *mapping,
···
 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 */

fs/btrfs/dir-item.c

···
 		ret = btrfs_truncate_item(trans, root, path,
 					  item_len - sub_item_len, 1);
 	}
-	return 0;
+	return ret;
 }

fs/btrfs/disk-io.c

···
 	struct extent_io_tree *tree;
 	u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
 	u64 found_start;
-	int found_level;
 	unsigned long len;
 	struct extent_buffer *eb;
 	int ret;
···
 		WARN_ON(1);
 		goto err;
 	}
-	found_level = btrfs_header_level(eb);
-
 	csum_tree_block(root, eb, 0);
 err:
 	free_extent_buffer(eb);
···
 	end_io_wq->work.flags = 0;
 
 	if (bio->bi_rw & REQ_WRITE) {
-		if (end_io_wq->metadata)
+		if (end_io_wq->metadata == 1)
 			btrfs_queue_worker(&fs_info->endio_meta_write_workers,
+					   &end_io_wq->work);
+		else if (end_io_wq->metadata == 2)
+			btrfs_queue_worker(&fs_info->endio_freespace_worker,
 					   &end_io_wq->work);
 		else
 			btrfs_queue_worker(&fs_info->endio_write_workers,
···
 	}
 }
 
+/*
+ * For the metadata arg you want
+ *
+ * 0 - if data
+ * 1 - if normal metadta
+ * 2 - if writing to the free space cache area
+ */
 int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
 			int metadata)
 {
···
 
 static void run_one_async_start(struct btrfs_work *work)
 {
-	struct btrfs_fs_info *fs_info;
 	struct async_submit_bio *async;
 
 	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,
 			       async->bio_offset);
···
 				      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)
···
 	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;
···
 		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)
···
 
 	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);
 
 	__setup_root(4096, 4096, 4096, 4096, tree_root,
 		     fs_info, BTRFS_ROOT_TREE_OBJECTID);
-
 
 	bh = btrfs_read_dev_super(fs_devices->latest_bdev);
 	if (!bh)
···
 	btrfs_init_workers(&fs_info->endio_write_workers, "endio-write",
 			   fs_info->thread_pool_size,
 			   &fs_info->generic_worker);
+	btrfs_init_workers(&fs_info->endio_freespace_worker, "freespace-write",
+			   1, &fs_info->generic_worker);
 
 	/*
 	 * endios are largely parallel and should have a very
···
 	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->endio_freespace_worker, 1);
 
 	fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
 	fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
···
 	if (!(sb->s_flags & MS_RDONLY)) {
 		down_read(&fs_info->cleanup_work_sem);
 		btrfs_orphan_cleanup(fs_info->fs_root);
+		btrfs_orphan_cleanup(fs_info->tree_root);
 		up_read(&fs_info->cleanup_work_sem);
 	}
 
···
 	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->endio_freespace_worker);
 	btrfs_stop_workers(&fs_info->submit_workers);
 fail_iput:
 	invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
···
 	fs_info->closing = 1;
 	smp_mb();
 
+	btrfs_put_block_group_cache(fs_info);
 	if (!(fs_info->sb->s_flags & MS_RDONLY)) {
 		ret =  btrfs_commit_super(root);
 		if (ret)
···
 	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->endio_freespace_worker);
 	btrfs_stop_workers(&fs_info->submit_workers);
 
 	btrfs_close_devices(fs_info->fs_devices);

fs/btrfs/extent-tree.c

···
 		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);
···
 	return 0;
 }
 
-static int cache_block_group(struct btrfs_block_group_cache *cache)
+static int cache_block_group(struct btrfs_block_group_cache *cache,
+			     struct btrfs_trans_handle *trans,
+			     int load_cache_only)
 {
 	struct btrfs_fs_info *fs_info = cache->fs_info;
 	struct btrfs_caching_control *caching_ctl;
···
 
 	smp_mb();
 	if (cache->cached != BTRFS_CACHE_NO)
+		return 0;
+
+	/*
+	 * We can't do the read from on-disk cache during a commit since we need
+	 * to have the normal tree locking.
+	 */
+	if (!trans->transaction->in_commit) {
+		spin_lock(&cache->lock);
+		if (cache->cached != BTRFS_CACHE_NO) {
+			spin_unlock(&cache->lock);
+			return 0;
+		}
+		cache->cached = BTRFS_CACHE_STARTED;
+		spin_unlock(&cache->lock);
+
+		ret = load_free_space_cache(fs_info, cache);
+
+		spin_lock(&cache->lock);
+		if (ret == 1) {
+			cache->cached = BTRFS_CACHE_FINISHED;
+			cache->last_byte_to_unpin = (u64)-1;
+		} else {
+			cache->cached = BTRFS_CACHE_NO;
+		}
+		spin_unlock(&cache->lock);
+		if (ret == 1)
+			return 0;
+	}
+
+	if (load_cache_only)
 		return 0;
 
 	caching_ctl = kzalloc(sizeof(*caching_ctl), GFP_KERNEL);
···
 
 	rcu_read_lock();
 	list_for_each_entry_rcu(found, head, list) {
-		if (found->flags == flags) {
+		if (found->flags & flags) {
 			rcu_read_unlock();
 			return found;
 		}
···
 		return num;
 	num *= factor;
 	do_div(num, 10);
+	return num;
+}
+
+static u64 div_factor_fine(u64 num, int factor)
+{
+	if (factor == 100)
+		return num;
+	num *= factor;
+	do_div(num, 100);
 	return num;
 }
 
···
 	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 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(root, 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 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_truncate_free_space_cache(root, trans, path,
+						      inode);
+		if (ret)
+			goto out_put;
+	}
+
+	spin_lock(&block_group->lock);
+	if (block_group->cached != BTRFS_CACHE_FINISHED) {
+		spin_unlock(&block_group->lock);
+		goto out_put;
+	}
+	spin_unlock(&block_group->lock);
+
+	num_pages = (int)div64_u64(block_group->key.offset, 1024 * 1024 * 1024);
+	if (!num_pages)
+		num_pages = 1;
+
+	/*
+	 * Just to make absolutely sure we have enough space, we're going to
+	 * preallocate 12 pages worth of space for each block group.  In
+	 * practice we ought to use at most 8, but we need extra space so we can
+	 * add our header and have a terminator between the extents and the
+	 * bitmaps.
+	 */
+	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);
+	btrfs_free_reserved_data_space(inode, num_pages);
+out_put:
+	iput(inode);
+out_free:
+	btrfs_release_path(root, path);
+out:
+	spin_lock(&block_group->lock);
+	if (ret)
+		block_group->disk_cache_state = BTRFS_DC_ERROR;
+	else
+		block_group->disk_cache_state = BTRFS_DC_SETUP;
+	spin_unlock(&block_group->lock);
+
+	return ret;
+}
+
 int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans,
 				   struct btrfs_root *root)
 {
···
 	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,
···
 
 		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);
···
 			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);
+	}
+
+	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);
+			BUG_ON(err);
+		}
+
+		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;
+		}
+
+		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 (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);
 	}
 
···
 	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;
···
 				BTRFS_BLOCK_GROUP_SYSTEM |
 				BTRFS_BLOCK_GROUP_METADATA);
 	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;
···
 	struct btrfs_space_info *data_sinfo;
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	u64 used;
-	int ret = 0, committed = 0;
+	int ret = 0, committed = 0, alloc_chunk = 1;
 
 	/* make sure bytes are sectorsize aligned */
 	bytes = (bytes + root->sectorsize - 1) & ~((u64)root->sectorsize - 1);
+
+	if (root == root->fs_info->tree_root) {
+		alloc_chunk = 0;
+		committed = 1;
+	}
 
 	data_sinfo = BTRFS_I(inode)->space_info;
 	if (!data_sinfo)
···
 		 * if we don't have enough free bytes in this space then we need
 		 * to alloc a new chunk.
 		 */
-		if (!data_sinfo->full) {
+		if (!data_sinfo->full && alloc_chunk) {
 			u64 alloc_target;
 
 			data_sinfo->force_alloc = 1;
···
 	rcu_read_unlock();
 }
 
-static int should_alloc_chunk(struct btrfs_space_info *sinfo,
-			      u64 alloc_bytes)
+static int should_alloc_chunk(struct btrfs_root *root,
+			      struct btrfs_space_info *sinfo, u64 alloc_bytes)
 {
 	u64 num_bytes = sinfo->total_bytes - sinfo->bytes_readonly;
+	u64 thresh;
 
 	if (sinfo->bytes_used + sinfo->bytes_reserved +
 	    alloc_bytes + 256 * 1024 * 1024 < num_bytes)
···
 
 	if (sinfo->bytes_used + sinfo->bytes_reserved +
 	    alloc_bytes < div_factor(num_bytes, 8))
+		return 0;
+
+	thresh = btrfs_super_total_bytes(&root->fs_info->super_copy);
+	thresh = max_t(u64, 256 * 1024 * 1024, div_factor_fine(thresh, 5));
+
+	if (num_bytes > thresh && sinfo->bytes_used < div_factor(num_bytes, 3))
 		return 0;
 
 	return 1;
···
 		goto out;
 	}
 
-	if (!force && !should_alloc_chunk(space_info, alloc_bytes)) {
+	if (!force && !should_alloc_chunk(extent_root, space_info,
+					  alloc_bytes)) {
 		spin_unlock(&space_info->lock);
 		goto out;
 	}
 	spin_unlock(&space_info->lock);
+
+	/*
+	 * 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
···
 	return ret;
 }
 
-static int maybe_allocate_chunk(struct btrfs_trans_handle *trans,
-				struct btrfs_root *root,
-				struct btrfs_space_info *sinfo, u64 num_bytes)
-{
-	int ret;
-	int end_trans = 0;
-
-	if (sinfo->full)
-		return 0;
-
-	spin_lock(&sinfo->lock);
-	ret = should_alloc_chunk(sinfo, num_bytes + 2 * 1024 * 1024);
-	spin_unlock(&sinfo->lock);
-	if (!ret)
-		return 0;
-
-	if (!trans) {
-		trans = btrfs_join_transaction(root, 1);
-		BUG_ON(IS_ERR(trans));
-		end_trans = 1;
-	}
-
-	ret = do_chunk_alloc(trans, root->fs_info->extent_root,
-			     num_bytes + 2 * 1024 * 1024,
-			     get_alloc_profile(root, sinfo->flags), 0);
-
-	if (end_trans)
-		btrfs_end_transaction(trans, root);
-
-	return ret == 1 ? 1 : 0;
-}
-
 /*
  * shrink metadata reservation for delalloc
  */
 static int shrink_delalloc(struct btrfs_trans_handle *trans,
-			   struct btrfs_root *root, u64 to_reclaim)
+			   struct btrfs_root *root, u64 to_reclaim, int sync)
 {
 	struct btrfs_block_rsv *block_rsv;
+	struct btrfs_space_info *space_info;
 	u64 reserved;
 	u64 max_reclaim;
 	u64 reclaimed = 0;
 	int pause = 1;
-	int ret;
+	int nr_pages = (2 * 1024 * 1024) >> PAGE_CACHE_SHIFT;
 
 	block_rsv = &root->fs_info->delalloc_block_rsv;
-	spin_lock(&block_rsv->lock);
-	reserved = block_rsv->reserved;
-	spin_unlock(&block_rsv->lock);
+	space_info = block_rsv->space_info;
+
+	smp_mb();
+	reserved = space_info->bytes_reserved;
 
 	if (reserved == 0)
 		return 0;
···
 	max_reclaim = min(reserved, to_reclaim);
 
 	while (1) {
-		ret = btrfs_start_one_delalloc_inode(root, trans ? 1 : 0);
-		if (!ret) {
-			__set_current_state(TASK_INTERRUPTIBLE);
-			schedule_timeout(pause);
-			pause <<= 1;
-			if (pause > HZ / 10)
-				pause = HZ / 10;
-		} else {
-			pause = 1;
-		}
+		/* have the flusher threads jump in and do some IO */
+		smp_mb();
+		nr_pages = min_t(unsigned long, nr_pages,
+		       root->fs_info->delalloc_bytes >> PAGE_CACHE_SHIFT);
+		writeback_inodes_sb_nr_if_idle(root->fs_info->sb, nr_pages);
 
-		spin_lock(&block_rsv->lock);
-		if (reserved > block_rsv->reserved)
-			reclaimed = reserved - block_rsv->reserved;
-		reserved = block_rsv->reserved;
-		spin_unlock(&block_rsv->lock);
+		spin_lock(&space_info->lock);
+		if (reserved > space_info->bytes_reserved)
+			reclaimed += reserved - space_info->bytes_reserved;
+		reserved = space_info->bytes_reserved;
+		spin_unlock(&space_info->lock);
 
 		if (reserved == 0 || reclaimed >= max_reclaim)
 			break;
 
 		if (trans && trans->transaction->blocked)
 			return -EAGAIN;
+
+		__set_current_state(TASK_INTERRUPTIBLE);
+		schedule_timeout(pause);
+		pause <<= 1;
+		if (pause > HZ / 10)
+			pause = HZ / 10;
+
 	}
 	return reclaimed >= to_reclaim;
 }
 
-static int should_retry_reserve(struct btrfs_trans_handle *trans,
-				struct btrfs_root *root,
-				struct btrfs_block_rsv *block_rsv,
-				u64 num_bytes, int *retries)
-{
-	struct btrfs_space_info *space_info = block_rsv->space_info;
-	int ret;
-
-	if ((*retries) > 2)
-		return -ENOSPC;
-
-	ret = maybe_allocate_chunk(trans, root, space_info, num_bytes);
-	if (ret)
-		return 1;
-
-	if (trans && trans->transaction->in_commit)
-		return -ENOSPC;
-
-	ret = shrink_delalloc(trans, root, num_bytes);
-	if (ret)
-		return ret;
-
-	spin_lock(&space_info->lock);
-	if (space_info->bytes_pinned < num_bytes)
-		ret = 1;
-	spin_unlock(&space_info->lock);
-	if (ret)
-		return -ENOSPC;
-
-	(*retries)++;
-
-	if (trans)
-		return -EAGAIN;
-
-	trans = btrfs_join_transaction(root, 1);
-	BUG_ON(IS_ERR(trans));
-	ret = btrfs_commit_transaction(trans, root);
-	BUG_ON(ret);
-
-	return 1;
-}
-
-static int reserve_metadata_bytes(struct btrfs_block_rsv *block_rsv,
-				  u64 num_bytes)
+/*
+ * Retries tells us how many times we've called reserve_metadata_bytes.  The
+ * idea is if this is the first call (retries == 0) then we will add to our
+ * reserved count if we can't make the allocation in order to hold our place
+ * while we go and try and free up space.  That way for retries > 1 we don't try
+ * and add space, we just check to see if the amount of unused space is >= the
+ * total space, meaning that our reservation is valid.
+ *
+ * However if we don't intend to retry this reservation, pass -1 as retries so
+ * that it short circuits this logic.
+ */
+static int reserve_metadata_bytes(struct btrfs_trans_handle *trans,
+				  struct btrfs_root *root,
+				  struct btrfs_block_rsv *block_rsv,
+				  u64 orig_bytes, int flush)
 {
 	struct btrfs_space_info *space_info = block_rsv->space_info;
 	u64 unused;
-	int ret = -ENOSPC;
+	u64 num_bytes = orig_bytes;
+	int retries = 0;
+	int ret = 0;
+	bool reserved = false;
+	bool committed = false;
+
+again:
+	ret = -ENOSPC;
+	if (reserved)
+		num_bytes = 0;
 
 	spin_lock(&space_info->lock);
 	unused = space_info->bytes_used + space_info->bytes_reserved +
-		 space_info->bytes_pinned + space_info->bytes_readonly;
+		 space_info->bytes_pinned + space_info->bytes_readonly +
+		 space_info->bytes_may_use;
 
-	if (unused < space_info->total_bytes)
-		unused = space_info->total_bytes - unused;
-	else
-		unused = 0;
-
-	if (unused >= num_bytes) {
-		if (block_rsv->priority >= 10) {
-			space_info->bytes_reserved += num_bytes;
+	/*
+	 * 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 (unused <= space_info->total_bytes) {
+		unused -= space_info->total_bytes;
+		if (unused >= num_bytes) {
+			if (!reserved)
+				space_info->bytes_reserved += orig_bytes;
 			ret = 0;
 		} else {
-			if ((unused + block_rsv->reserved) *
-			    block_rsv->priority >=
-			    (num_bytes + block_rsv->reserved) * 10) {
-				space_info->bytes_reserved += num_bytes;
-				ret = 0;
-			}
+			/*
+			 * 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 {
+		/*
+		 * 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 = unused - space_info->total_bytes +
+			(orig_bytes * (retries + 1));
 	}
+
+	/*
+	 * 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.
+	 */
+	if (ret && !reserved) {
+		space_info->bytes_reserved += orig_bytes;
+		reserved = true;
+	}
+
 	spin_unlock(&space_info->lock);
+
+	if (!ret)
+		return 0;
+
+	if (!flush)
+		goto out;
+
+	/*
+	 * We do synchronous shrinking since we don't actually unreserve
+	 * metadata until after the IO is completed.
+	 */
+	ret = shrink_delalloc(trans, root, num_bytes, 1);
+	if (ret > 0)
+		return 0;
+	else if (ret < 0)
+		goto out;
+
+	/*
+	 * So if we were overcommitted it's possible that somebody else flushed
+	 * out enough space and we simply didn't have enough space to reclaim,
+	 * so go back around and try again.
+	 */
+	if (retries < 2) {
+		retries++;
+		goto again;
+	}
+
+	spin_lock(&space_info->lock);
+	/*
+	 * Not enough space to be reclaimed, don't bother committing the
+	 * transaction.
+	 */
+	if (space_info->bytes_pinned < orig_bytes)
+		ret = -ENOSPC;
+	spin_unlock(&space_info->lock);
+	if (ret)
+		goto out;
+
+	ret = -EAGAIN;
+	if (trans || committed)
+		goto out;
+
+	ret = -ENOSPC;
+	trans = btrfs_join_transaction(root, 1);
+	if (IS_ERR(trans))
+		goto out;
+	ret = btrfs_commit_transaction(trans, root);
+	if (!ret) {
+		trans = NULL;
+		committed = true;
+		goto again;
+	}
+
+out:
+	if (reserved) {
+		spin_lock(&space_info->lock);
+		space_info->bytes_reserved -= orig_bytes;
+		spin_unlock(&space_info->lock);
+	}
 
 	return ret;
 }
···
 {
 	struct btrfs_block_rsv *block_rsv;
 	struct btrfs_fs_info *fs_info = root->fs_info;
-	u64 alloc_target;
 
 	block_rsv = kmalloc(sizeof(*block_rsv), GFP_NOFS);
 	if (!block_rsv)
 		return NULL;
 
 	btrfs_init_block_rsv(block_rsv);
-
-	alloc_target = btrfs_get_alloc_profile(root, 0);
 	block_rsv->space_info = __find_space_info(fs_info,
 						  BTRFS_BLOCK_GROUP_METADATA);
-
 	return block_rsv;
 }
 
···
 int btrfs_block_rsv_add(struct btrfs_trans_handle *trans,
 			struct btrfs_root *root,
 			struct btrfs_block_rsv *block_rsv,
-			u64 num_bytes, int *retries)
+			u64 num_bytes)
 {
 	int ret;
 
 	if (num_bytes == 0)
 		return 0;
-again:
-	ret = reserve_metadata_bytes(block_rsv, num_bytes);
+
+	ret = reserve_metadata_bytes(trans, root, block_rsv, num_bytes, 1);
 	if (!ret) {
 		block_rsv_add_bytes(block_rsv, num_bytes, 1);
 		return 0;
 	}
-
-	ret = should_retry_reserve(trans, root, block_rsv, num_bytes, retries);
-	if (ret > 0)
-		goto again;
 
 	return ret;
 }
···
 		return 0;
 
 	if (block_rsv->refill_used) {
-		ret = reserve_metadata_bytes(block_rsv, num_bytes);
+		ret = reserve_metadata_bytes(trans, root, block_rsv,
+					     num_bytes, 0);
 		if (!ret) {
 			block_rsv_add_bytes(block_rsv, num_bytes, 0);
 			return 0;
···
 
 	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);
 
···
 	block_rsv->size = num_bytes;
 
 	num_bytes = sinfo->bytes_used + sinfo->bytes_pinned +
-		    sinfo->bytes_reserved + sinfo->bytes_readonly;
+		    sinfo->bytes_reserved + sinfo->bytes_readonly +
+		    sinfo->bytes_may_use;
 
 	if (sinfo->total_bytes > num_bytes) {
 		num_bytes = sinfo->total_bytes - num_bytes;
···
 
 int btrfs_trans_reserve_metadata(struct btrfs_trans_handle *trans,
 				 struct btrfs_root *root,
-				 int num_items, int *retries)
+				 int num_items)
 {
 	u64 num_bytes;
 	int ret;
···
 
 	num_bytes = calc_trans_metadata_size(root, num_items);
 	ret = btrfs_block_rsv_add(trans, root, &root->fs_info->trans_block_rsv,
-				  num_bytes, retries);
+				  num_bytes);
 	if (!ret) {
 		trans->bytes_reserved += num_bytes;
 		trans->block_rsv = &root->fs_info->trans_block_rsv;
···
 	struct btrfs_block_rsv *block_rsv = &root->fs_info->delalloc_block_rsv;
 	u64 to_reserve;
 	int nr_extents;
-	int retries = 0;
 	int ret;
 
 	if (btrfs_transaction_in_commit(root->fs_info))
 		schedule_timeout(1);
 
 	num_bytes = ALIGN(num_bytes, root->sectorsize);
-again:
+
 	spin_lock(&BTRFS_I(inode)->accounting_lock);
 	nr_extents = atomic_read(&BTRFS_I(inode)->outstanding_extents) + 1;
 	if (nr_extents > BTRFS_I(inode)->reserved_extents) {
···
 		nr_extents = 0;
 		to_reserve = 0;
 	}
+	spin_unlock(&BTRFS_I(inode)->accounting_lock);
 
 	to_reserve += calc_csum_metadata_size(inode, num_bytes);
-	ret = reserve_metadata_bytes(block_rsv, to_reserve);
-	if (ret) {
-		spin_unlock(&BTRFS_I(inode)->accounting_lock);
-		ret = should_retry_reserve(NULL, root, block_rsv, to_reserve,
-					   &retries);
-		if (ret > 0)
-			goto again;
+	ret = reserve_metadata_bytes(NULL, root, block_rsv, to_reserve, 1);
+	if (ret)
 		return ret;
-	}
 
+	spin_lock(&BTRFS_I(inode)->accounting_lock);
 	BTRFS_I(inode)->reserved_extents += nr_extents;
 	atomic_inc(&BTRFS_I(inode)->outstanding_extents);
 	spin_unlock(&BTRFS_I(inode)->accounting_lock);
···
 	block_rsv_add_bytes(block_rsv, to_reserve, 1);
 
 	if (block_rsv->size > 512 * 1024 * 1024)
-		shrink_delalloc(NULL, root, to_reserve);
+		shrink_delalloc(NULL, root, to_reserve, 0);
 
 	return 0;
 }
···
 			      struct btrfs_root *root,
 			      u64 bytenr, u64 num_bytes, int alloc)
 {
-	struct btrfs_block_group_cache *cache;
+	struct btrfs_block_group_cache *cache = NULL;
 	struct btrfs_fs_info *info = root->fs_info;
-	int factor;
 	u64 total = num_bytes;
 	u64 old_val;
 	u64 byte_in_group;
+	int factor;
 
 	/* block accounting for super block */
 	spin_lock(&info->delalloc_lock);
···
 			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, trans, 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_super_cache_generation(&info->super_copy) != 0 &&
+		    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);
···
 	bool found_uncached_bg = false;
 	bool failed_cluster_refill = false;
 	bool failed_alloc = false;
+	bool use_cluster = true;
 	u64 ideal_cache_percent = 0;
 	u64 ideal_cache_offset = 0;
 
···
 		return -ENOSPC;
 	}
 
+	/*
+	 * 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 (orig_root->ref_cows || empty_size)
 		allowed_chunk_alloc = 1;
 
-	if (data & BTRFS_BLOCK_GROUP_METADATA) {
+	if (data & 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 ((data & BTRFS_BLOCK_GROUP_DATA) && use_cluster &&
+	    btrfs_test_opt(root, SSD)) {
 		last_ptr = &root->fs_info->data_alloc_cluster;
 	}
 
···
 		if (unlikely(block_group->cached == BTRFS_CACHE_NO)) {
 			u64 free_percent;
 
+			ret = cache_block_group(block_group, trans, 1);
+			if (block_group->cached == BTRFS_CACHE_FINISHED)
+				goto have_block_group;
+
 			free_percent = btrfs_block_group_used(&block_group->item);
 			free_percent *= 100;
 			free_percent = div64_u64(free_percent,
···
 			if (loop > LOOP_CACHING_NOWAIT ||
 			    (loop > LOOP_FIND_IDEAL &&
 			     atomic_read(&space_info->caching_threads) < 2)) {
-				ret = cache_block_group(block_group);
+				ret = cache_block_group(block_group, trans, 0);
 				BUG_ON(ret);
 			}
 			found_uncached_bg = true;
···
 	u64 num_bytes = ins->offset;
 
 	block_group = btrfs_lookup_block_group(root->fs_info, ins->objectid);
-	cache_block_group(block_group);
+	cache_block_group(block_group, trans, 0);
 	caching_ctl = get_caching_control(block_group);
 
 	if (!caching_ctl) {
···
 	block_rsv = get_block_rsv(trans, root);
 
 	if (block_rsv->size == 0) {
-		ret = reserve_metadata_bytes(block_rsv, blocksize);
+		ret = reserve_metadata_bytes(trans, root, block_rsv,
+					     blocksize, 0);
 		if (ret)
 			return ERR_PTR(ret);
 		return block_rsv;
···
 	ret = block_rsv_use_bytes(block_rsv, blocksize);
 	if (!ret)
 		return block_rsv;
-
-	WARN_ON(1);
-	printk(KERN_INFO"block_rsv size %llu reserved %llu freed %llu %llu\n",
-		block_rsv->size, block_rsv->reserved,
-		block_rsv->freed[0], block_rsv->freed[1]);
 
 	return ERR_PTR(-ENOSPC);
 }
···
 	u64 generation;
 	u64 refs;
 	u64 flags;
-	u64 last = 0;
 	u32 nritems;
 	u32 blocksize;
 	struct btrfs_key key;
···
 					   generation);
 		if (ret)
 			break;
-		last = bytenr + blocksize;
 		nread++;
 	}
 	wc->reada_slot = slot;
···
 	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;
···
 	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;
···
 	path = btrfs_alloc_path();
 	if (!path)
 		return -ENOMEM;
+
+	cache_gen = btrfs_super_cache_generation(&root->fs_info->super_copy);
+	if (cache_gen != 0 &&
+	    btrfs_super_generation(&root->fs_info->super_copy) != cache_gen)
+		need_clear = 1;
+	if (btrfs_test_opt(root, CLEAR_CACHE))
+		need_clear = 1;
+	if (!btrfs_test_opt(root, SPACE_CACHE) && cache_gen)
+		printk(KERN_INFO "btrfs: disk space caching is enabled\n");
 
 	while (1) {
 		ret = find_first_block_group(root, path, &key);
···
 		cache->fs_info = info;
 		INIT_LIST_HEAD(&cache->list);
 		INIT_LIST_HEAD(&cache->cluster_list);
+
+		if (need_clear)
+			cache->disk_cache_state = BTRFS_DC_CLEAR;
 
 		/*
 		 * we only want to have 32k of ram per block group for keeping
···
 	cache->key.offset = size;
 	cache->key.type = BTRFS_BLOCK_GROUP_ITEM_KEY;
 	cache->sectorsize = root->sectorsize;
+	cache->fs_info = root->fs_info;
 
 	/*
 	 * we only want to have 32k of ram per block group for keeping track
···
 	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;
 	int ret;
+	int factor;
 
 	root = root->fs_info->extent_root;
 
···
 	BUG_ON(!block_group->ro);
 
 	memcpy(&key, &block_group->key, sizeof(key));
+	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;
···
 
 	path = btrfs_alloc_path();
 	BUG_ON(!path);
+
+	inode = lookup_free_space_inode(root, block_group, path);
+	if (!IS_ERR(inode)) {
+		btrfs_orphan_add(trans, inode);
+		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 */
+		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(tree_root, path);
+	if (ret == 0) {
+		ret = btrfs_del_item(trans, tree_root, path);
+		if (ret)
+			goto out;
+		btrfs_release_path(tree_root, path);
+	}
 
 	spin_lock(&root->fs_info->block_group_cache_lock);
 	rb_erase(&block_group->cache_node,
···
 	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);
 

fs/btrfs/extent_io.c

···
 			  struct address_space *mapping, gfp_t mask)
 {
 	tree->state = RB_ROOT;
-	tree->buffer = RB_ROOT;
+	INIT_RADIX_TREE(&tree->buffer, GFP_ATOMIC);
 	tree->ops = NULL;
 	tree->dirty_bytes = 0;
 	spin_lock_init(&tree->lock);
···
 	if (!ret)
 		return prev;
 	return ret;
-}
-
-static struct extent_buffer *buffer_tree_insert(struct extent_io_tree *tree,
-					  u64 offset, struct rb_node *node)
-{
-	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;
 }
 
 static void merge_cb(struct extent_io_tree *tree, struct extent_state *new,
···
 	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;
 
 	bio->bi_private = NULL;
 
···
 	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;
···
 		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;
 	}
 
···
 			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,
···
 
 			cur += iosize;
 			pg_offset += iosize;
-			unlock_start = cur;
 			continue;
 		}
 		/* leave this out until we have a page_mkwrite call */
···
 	pgoff_t index;
 	pgoff_t end;		/* Inclusive */
 	int scanned = 0;
-	int range_whole = 0;
 
 	pagevec_init(&pvec, 0);
 	if (wbc->range_cyclic) {
···
 	} 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;
 	}
 retry:
···
 					 NULL, 1,
 					 end_bio_extent_preparewrite, 0,
 					 0, 0);
+			if (ret && !err)
+				err = ret;
 			iocount++;
 			block_start = block_start + iosize;
 		} else {
···
 	kmem_cache_free(extent_buffer_cache, eb);
 }
 
+/*
+ * Helper for releasing extent buffer page.
+ */
+static void btrfs_release_extent_buffer_page(struct extent_buffer *eb,
+						unsigned long start_idx)
+{
+	unsigned long index;
+	struct page *page;
+
+	if (!eb->first_page)
+		return;
+
+	index = num_extent_pages(eb->start, eb->len);
+	if (start_idx >= index)
+		return;
+
+	do {
+		index--;
+		page = extent_buffer_page(eb, index);
+		if (page)
+			page_cache_release(page);
+	} while (index != start_idx);
+}
+
+/*
+ * Helper for releasing the extent buffer.
+ */
+static inline void btrfs_release_extent_buffer(struct extent_buffer *eb)
+{
+	btrfs_release_extent_buffer_page(eb, 0);
+	__free_extent_buffer(eb);
+}
+
 struct extent_buffer *alloc_extent_buffer(struct extent_io_tree *tree,
 					  u64 start, unsigned long len,
 					  struct page *page0,
···
 	struct page *p;
 	struct address_space *mapping = tree->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);
+	rcu_read_lock();
+	eb = radix_tree_lookup(&tree->buffer, start >> PAGE_CACHE_SHIFT);
+	if (eb && atomic_inc_not_zero(&eb->refs)) {
+		rcu_read_unlock();
 		mark_page_accessed(eb->first_page);
 		return eb;
 	}
-	spin_unlock(&tree->buffer_lock);
+	rcu_read_unlock();
 
 	eb = __alloc_extent_buffer(tree, start, len, mask);
 	if (!eb)
···
 	if (uptodate)
 		set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
 
+	ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
+	if (ret)
+		goto free_eb;
+
 	spin_lock(&tree->buffer_lock);
-	exists = buffer_tree_insert(tree, start, &eb->rb_node);
-	if (exists) {
+	ret = radix_tree_insert(&tree->buffer, start >> PAGE_CACHE_SHIFT, eb);
+	if (ret == -EEXIST) {
+		exists = radix_tree_lookup(&tree->buffer,
+						start >> PAGE_CACHE_SHIFT);
 		/* add one reference for the caller */
 		atomic_inc(&exists->refs);
 		spin_unlock(&tree->buffer_lock);
+		radix_tree_preload_end();
 		goto free_eb;
 	}
 	/* add one reference for the tree */
 	atomic_inc(&eb->refs);
 	spin_unlock(&tree->buffer_lock);
+	radix_tree_preload_end();
 	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);
+	btrfs_release_extent_buffer(eb);
 	return exists;
 }
 
···
 {
 	struct extent_buffer *eb;
 
-	spin_lock(&tree->buffer_lock);
-	eb = buffer_search(tree, start);
-	if (eb)
-		atomic_inc(&eb->refs);
-	spin_unlock(&tree->buffer_lock);
-
-	if (eb)
+	rcu_read_lock();
+	eb = radix_tree_lookup(&tree->buffer, start >> PAGE_CACHE_SHIFT);
+	if (eb && atomic_inc_not_zero(&eb->refs)) {
+		rcu_read_unlock();
 		mark_page_accessed(eb->first_page);
+		return eb;
+	}
+	rcu_read_unlock();
 
-	return eb;
+	return NULL;
 }
 
 void free_extent_buffer(struct extent_buffer *eb)
···
 	}
 }
 
+static inline void btrfs_release_extent_buffer_rcu(struct rcu_head *head)
+{
+	struct extent_buffer *eb =
+			container_of(head, struct extent_buffer, rcu_head);
+
+	btrfs_release_extent_buffer(eb);
+}
+
 int try_release_extent_buffer(struct extent_io_tree *tree, 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);
+	eb = radix_tree_lookup(&tree->buffer, start >> PAGE_CACHE_SHIFT);
 	if (!eb)
 		goto out;
 
-	if (atomic_read(&eb->refs) > 1) {
-		ret = 0;
-		goto out;
-	}
 	if (test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) {
 		ret = 0;
 		goto out;
 	}
-	/* 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);
+
+	/*
+	 * set @eb->refs to 0 if it is already 1, and then release the @eb.
+	 * Or go back.
+	 */
+	if (atomic_cmpxchg(&eb->refs, 1, 0) != 1) {
+		ret = 0;
+		goto out;
+	}
+
+	radix_tree_delete(&tree->buffer, start >> PAGE_CACHE_SHIFT);
 out:
 	spin_unlock(&tree->buffer_lock);
+
+	/* at this point we can safely release the extent buffer */
+	if (atomic_read(&eb->refs) == 0)
+		call_rcu(&eb->rcu_head, btrfs_release_extent_buffer_rcu);
 	return ret;
 }

fs/btrfs/extent_io.h

···
 
 struct extent_io_tree {
 	struct rb_root state;
-	struct rb_root buffer;
+	struct radix_tree_root buffer;
 	struct address_space *mapping;
 	u64 dirty_bytes;
 	spinlock_t lock;
···
 	unsigned long bflags;
 	atomic_t refs;
 	struct list_head leak_list;
-	struct rb_node rb_node;
+	struct rcu_head rcu_head;
 
 	/* the spinlock is used to protect most operations */
 	spinlock_t lock;

fs/btrfs/extent_map.c

···
 		goto out;
 	}
 	if (IS_ERR(rb_node)) {
-		em = ERR_PTR(PTR_ERR(rb_node));
+		em = ERR_CAST(rb_node);
 		goto out;
 	}
 	em = rb_entry(rb_node, struct extent_map, rb_node);
···
 		goto out;
 	}
 	if (IS_ERR(rb_node)) {
-		em = ERR_PTR(PTR_ERR(rb_node));
+		em = ERR_CAST(rb_node);
 		goto out;
 	}
 	em = rb_entry(rb_node, struct extent_map, rb_node);

fs/btrfs/free-space-cache.c

···
 #include "ctree.h"
 #include "free-space-cache.h"
 #include "transaction.h"
+#include "disk-io.h"
 
 #define BITS_PER_BITMAP		(PAGE_CACHE_SIZE * 8)
 #define MAX_CACHE_BYTES_PER_GIG	(32 * 1024)
+
+static void recalculate_thresholds(struct btrfs_block_group_cache
+				   *block_group);
+static int link_free_space(struct btrfs_block_group_cache *block_group,
+			   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)
+{
+	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;
+
+	spin_lock(&block_group->lock);
+	if (block_group->inode)
+		inode = igrab(block_group->inode);
+	spin_unlock(&block_group->lock);
+	if (inode)
+		return inode;
+
+	key.objectid = BTRFS_FREE_SPACE_OBJECTID;
+	key.offset = block_group->key.objectid;
+	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(root, 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(root, 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);
+	}
+
+	spin_lock(&block_group->lock);
+	if (!root->fs_info->closing) {
+		block_group->inode = igrab(inode);
+		block_group->iref = 1;
+	}
+	spin_unlock(&block_group->lock);
+
+	return inode;
+}
+
+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)
+{
+	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 objectid;
+	int ret;
+
+	ret = btrfs_find_free_objectid(trans, root, 0, &objectid);
+	if (ret < 0)
+		return ret;
+
+	ret = btrfs_insert_empty_inode(trans, root, path, objectid);
+	if (ret)
+		return ret;
+
+	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, BTRFS_INODE_NOCOMPRESS |
+			      BTRFS_INODE_PREALLOC | BTRFS_INODE_NODATASUM);
+	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,
+				    block_group->key.objectid);
+	btrfs_mark_buffer_dirty(leaf);
+	btrfs_release_path(root, path);
+
+	key.objectid = BTRFS_FREE_SPACE_OBJECTID;
+	key.offset = block_group->key.objectid;
+	key.type = 0;
+
+	ret = btrfs_insert_empty_item(trans, root, path, &key,
+				      sizeof(struct btrfs_free_space_header));
+	if (ret < 0) {
+		btrfs_release_path(root, 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(root, path);
+
+	return 0;
+}
+
+int btrfs_truncate_free_space_cache(struct btrfs_root *root,
+				    struct btrfs_trans_handle *trans,
+				    struct btrfs_path *path,
+				    struct inode *inode)
+{
+	loff_t oldsize;
+	int ret = 0;
+
+	trans->block_rsv = root->orphan_block_rsv;
+	ret = btrfs_block_rsv_check(trans, root,
+				    root->orphan_block_rsv,
+				    0, 5);
+	if (ret)
+		return ret;
+
+	oldsize = i_size_read(inode);
+	btrfs_i_size_write(inode, 0);
+	truncate_pagecache(inode, oldsize, 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) {
+		WARN_ON(1);
+		return ret;
+	}
+
+	return btrfs_update_inode(trans, root, inode);
+}
+
+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;
+}
+
+int load_free_space_cache(struct btrfs_fs_info *fs_info,
+			  struct btrfs_block_group_cache *block_group)
+{
+	struct btrfs_root *root = fs_info->tree_root;
+	struct inode *inode;
+	struct btrfs_free_space_header *header;
+	struct extent_buffer *leaf;
+	struct page *page;
+	struct btrfs_path *path;
+	u32 *checksums = NULL, *crc;
+	char *disk_crcs = NULL;
+	struct btrfs_key key;
+	struct list_head bitmaps;
+	u64 num_entries;
+	u64 num_bitmaps;
+	u64 generation;
+	u32 cur_crc = ~(u32)0;
+	pgoff_t index = 0;
+	unsigned long first_page_offset;
+	int num_checksums;
+	int ret = 0;
+
+	/*
+	 * 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.
+	 */
+	smp_mb();
+	if (fs_info->closing)
+		return 0;
+
+	/*
+	 * 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);
+
+	INIT_LIST_HEAD(&bitmaps);
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return 0;
+
+	inode = lookup_free_space_inode(root, block_group, path);
+	if (IS_ERR(inode)) {
+		btrfs_free_path(path);
+		return 0;
+	}
+
+	/* Nothing in the space cache, goodbye */
+	if (!i_size_read(inode)) {
+		btrfs_free_path(path);
+		goto out;
+	}
+
+	key.objectid = BTRFS_FREE_SPACE_OBJECTID;
+	key.offset = block_group->key.objectid;
+	key.type = 0;
+
+	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+	if (ret) {
+		btrfs_free_path(path);
+		goto out;
+	}
+
+	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_free_path(path);
+
+	if (BTRFS_I(inode)->generation != generation) {
+		printk(KERN_ERR "btrfs: free space inode generation (%llu) did"
+		       " not match free space cache generation (%llu) for "
+		       "block group %llu\n",
+		       (unsigned long long)BTRFS_I(inode)->generation,
+		       (unsigned long long)generation,
+		       (unsigned long long)block_group->key.objectid);
+		goto out;
+	}
+
+	if (!num_entries)
+		goto out;
+
+	/* Setup everything for doing checksumming */
+	num_checksums = i_size_read(inode) / PAGE_CACHE_SIZE;
+	checksums = crc = kzalloc(sizeof(u32) * num_checksums, GFP_NOFS);
+	if (!checksums)
+		goto out;
+	first_page_offset = (sizeof(u32) * num_checksums) + sizeof(u64);
+	disk_crcs = kzalloc(first_page_offset, GFP_NOFS);
+	if (!disk_crcs)
+		goto out;
+
+	ret = readahead_cache(inode);
+	if (ret) {
+		ret = 0;
+		goto out;
+	}
+
+	while (1) {
+		struct btrfs_free_space_entry *entry;
+		struct btrfs_free_space *e;
+		void *addr;
+		unsigned long offset = 0;
+		unsigned long start_offset = 0;
+		int need_loop = 0;
+
+		if (!num_entries && !num_bitmaps)
+			break;
+
+		if (index == 0) {
+			start_offset = first_page_offset;
+			offset = start_offset;
+		}
+
+		page = grab_cache_page(inode->i_mapping, index);
+		if (!page) {
+			ret = 0;
+			goto free_cache;
+		}
+
+		if (!PageUptodate(page)) {
+			btrfs_readpage(NULL, page);
+			lock_page(page);
+			if (!PageUptodate(page)) {
+				unlock_page(page);
+				page_cache_release(page);
+				printk(KERN_ERR "btrfs: error reading free "
+				       "space cache: %llu\n",
+				       (unsigned long long)
+				       block_group->key.objectid);
+				goto free_cache;
+			}
+		}
+		addr = kmap(page);
+
+		if (index == 0) {
+			u64 *gen;
+
+			memcpy(disk_crcs, addr, first_page_offset);
+			gen = addr + (sizeof(u32) * num_checksums);
+			if (*gen != BTRFS_I(inode)->generation) {
+				printk(KERN_ERR "btrfs: space cache generation"
+				       " (%llu) does not match inode (%llu) "
+				       "for block group %llu\n",
+				       (unsigned long long)*gen,
+				       (unsigned long long)
+				       BTRFS_I(inode)->generation,
+				       (unsigned long long)
+				       block_group->key.objectid);
+				kunmap(page);
+				unlock_page(page);
+				page_cache_release(page);
+				goto free_cache;
+			}
+			crc = (u32 *)disk_crcs;
+		}
+		entry = addr + start_offset;
+
+		/* First lets check our crc before we do anything fun */
+		cur_crc = ~(u32)0;
+		cur_crc = btrfs_csum_data(root, addr + start_offset, cur_crc,
+					  PAGE_CACHE_SIZE - start_offset);
+		btrfs_csum_final(cur_crc, (char *)&cur_crc);
+		if (cur_crc != *crc) {
+			printk(KERN_ERR "btrfs: crc mismatch for page %lu in "
+			       "block group %llu\n", index,
+			       (unsigned long long)block_group->key.objectid);
+			kunmap(page);
+			unlock_page(page);
+			page_cache_release(page);
+			goto free_cache;
+		}
+		crc++;
+
+		while (1) {
+			if (!num_entries)
+				break;
+
+			need_loop = 1;
+			e = kzalloc(sizeof(struct btrfs_free_space), GFP_NOFS);
+			if (!e) {
+				kunmap(page);
+				unlock_page(page);
+				page_cache_release(page);
+				goto free_cache;
+			}
+
+			e->offset = le64_to_cpu(entry->offset);
+			e->bytes = le64_to_cpu(entry->bytes);
+			if (!e->bytes) {
+				kunmap(page);
+				kfree(e);
+				unlock_page(page);
+				page_cache_release(page);
+				goto free_cache;
+			}
+
+			if (entry->type == BTRFS_FREE_SPACE_EXTENT) {
+				spin_lock(&block_group->tree_lock);
+				ret = link_free_space(block_group, e);
+				spin_unlock(&block_group->tree_lock);
+				BUG_ON(ret);
+			} else {
+				e->bitmap = kzalloc(PAGE_CACHE_SIZE, GFP_NOFS);
+				if (!e->bitmap) {
+					kunmap(page);
+					kfree(e);
+					unlock_page(page);
+					page_cache_release(page);
+					goto free_cache;
+				}
+				spin_lock(&block_group->tree_lock);
+				ret = link_free_space(block_group, e);
+				block_group->total_bitmaps++;
+				recalculate_thresholds(block_group);
+				spin_unlock(&block_group->tree_lock);
+				list_add_tail(&e->list, &bitmaps);
+			}
+
+			num_entries--;
+			offset += sizeof(struct btrfs_free_space_entry);
+			if (offset + sizeof(struct btrfs_free_space_entry) >=
+			    PAGE_CACHE_SIZE)
+				break;
+			entry++;
+		}
+
+		/*
+		 * We read an entry out of this page, we need to move on to the
+		 * next page.
+		 */
+		if (need_loop) {
+			kunmap(page);
+			goto next;
+		}
+
+		/*
+		 * We add the bitmaps at the end of the entries in order that
+		 * the bitmap entries are added to the cache.
+		 */
+		e = list_entry(bitmaps.next, struct btrfs_free_space, list);
+		list_del_init(&e->list);
+		memcpy(e->bitmap, addr, PAGE_CACHE_SIZE);
+		kunmap(page);
+		num_bitmaps--;
+next:
+		unlock_page(page);
+		page_cache_release(page);
+		index++;
+	}
+
+	ret = 1;
+out:
+	kfree(checksums);
+	kfree(disk_crcs);
+	iput(inode);
+	return ret;
+
+free_cache:
+	/* 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);
+	btrfs_remove_free_space_cache(block_group);
+	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_header *header;
+	struct extent_buffer *leaf;
+	struct inode *inode;
+	struct rb_node *node;
+	struct list_head *pos, *n;
+	struct page *page;
+	struct extent_state *cached_state = NULL;
+	struct list_head bitmap_list;
+	struct btrfs_key key;
+	u64 bytes = 0;
+	u32 *crc, *checksums;
+	pgoff_t index = 0, last_index = 0;
+	unsigned long first_page_offset;
+	int num_checksums;
+	int entries = 0;
+	int bitmaps = 0;
+	int ret = 0;
+
+	root = root->fs_info->tree_root;
+
+	INIT_LIST_HEAD(&bitmap_list);
+
+	spin_lock(&block_group->lock);
+	if (block_group->disk_cache_state < BTRFS_DC_SETUP) {
+		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;
+
+	if (!i_size_read(inode)) {
+		iput(inode);
+		return 0;
+	}
+
+	last_index = (i_size_read(inode) - 1) >> PAGE_CACHE_SHIFT;
+	filemap_write_and_wait(inode->i_mapping);
+	btrfs_wait_ordered_range(inode, inode->i_size &
+				 ~(root->sectorsize - 1), (u64)-1);
+
+	/* We need a checksum per page. */
+	num_checksums = i_size_read(inode) / PAGE_CACHE_SIZE;
+	crc = checksums  = kzalloc(sizeof(u32) * num_checksums, GFP_NOFS);
+	if (!crc) {
+		iput(inode);
+		return 0;
+	}
+
+	/* Since the first page has all of our checksums and our generation we
+	 * need to calculate the offset into the page that we can start writing
+	 * our entries.
+	 */
+	first_page_offset = (sizeof(u32) * num_checksums) + sizeof(u64);
+
+	node = rb_first(&block_group->free_space_offset);
+	if (!node)
+		goto out_free;
+
+	/*
+	 * Lock all pages first so we can lock the extent safely.
+	 *
+	 * NOTE: Because we hold the ref the entire time we're going to write to
+	 * the page find_get_page should never fail, so we don't do a check
+	 * after find_get_page at this point.  Just putting this here so people
+	 * know and don't freak out.
+	 */
+	while (index <= last_index) {
+		page = grab_cache_page(inode->i_mapping, index);
+		if (!page) {
+			pgoff_t i = 0;
+
+			while (i < index) {
+				page = find_get_page(inode->i_mapping, i);
+				unlock_page(page);
+				page_cache_release(page);
+				page_cache_release(page);
+				i++;
+			}
+			goto out_free;
+		}
+		index++;
+	}
+
+	index = 0;
+	lock_extent_bits(&BTRFS_I(inode)->io_tree, 0, i_size_read(inode) - 1,
+			 0, &cached_state, GFP_NOFS);
+
+	/* Write out the extent entries */
+	do {
+		struct btrfs_free_space_entry *entry;
+		void *addr;
+		unsigned long offset = 0;
+		unsigned long start_offset = 0;
+
+		if (index == 0) {
+			start_offset = first_page_offset;
+			offset = start_offset;
+		}
+
+		page = find_get_page(inode->i_mapping, index);
+
+		addr = kmap(page);
+		entry = addr + start_offset;
+
+		memset(addr, 0, PAGE_CACHE_SIZE);
+		while (1) {
+			struct btrfs_free_space *e;
+
+			e = rb_entry(node, struct btrfs_free_space, offset_index);
+			entries++;
+
+			entry->offset = cpu_to_le64(e->offset);
+			entry->bytes = cpu_to_le64(e->bytes);
+			if (e->bitmap) {
+				entry->type = BTRFS_FREE_SPACE_BITMAP;
+				list_add_tail(&e->list, &bitmap_list);
+				bitmaps++;
+			} else {
+				entry->type = BTRFS_FREE_SPACE_EXTENT;
+			}
+			node = rb_next(node);
+			if (!node)
+				break;
+			offset += sizeof(struct btrfs_free_space_entry);
+			if (offset + sizeof(struct btrfs_free_space_entry) >=
+			    PAGE_CACHE_SIZE)
+				break;
+			entry++;
+		}
+		*crc = ~(u32)0;
+		*crc = btrfs_csum_data(root, addr + start_offset, *crc,
+				       PAGE_CACHE_SIZE - start_offset);
+		kunmap(page);
+
+		btrfs_csum_final(*crc, (char *)crc);
+		crc++;
+
+		bytes += PAGE_CACHE_SIZE;
+
+		ClearPageChecked(page);
+		set_page_extent_mapped(page);
+		SetPageUptodate(page);
+		set_page_dirty(page);
+
+		/*
+		 * We need to release our reference we got for grab_cache_page,
+		 * except for the first page which will hold our checksums, we
+		 * do that below.
+		 */
+		if (index != 0) {
+			unlock_page(page);
+			page_cache_release(page);
+		}
+
+		page_cache_release(page);
+
+		index++;
+	} while (node);
+
+	/* Write out the bitmaps */
+	list_for_each_safe(pos, n, &bitmap_list) {
+		void *addr;
+		struct btrfs_free_space *entry =
+			list_entry(pos, struct btrfs_free_space, list);
+
+		page = find_get_page(inode->i_mapping, index);
+
+		addr = kmap(page);
+		memcpy(addr, entry->bitmap, PAGE_CACHE_SIZE);
+		*crc = ~(u32)0;
+		*crc = btrfs_csum_data(root, addr, *crc, PAGE_CACHE_SIZE);
+		kunmap(page);
+		btrfs_csum_final(*crc, (char *)crc);
+		crc++;
+		bytes += PAGE_CACHE_SIZE;
+
+		ClearPageChecked(page);
+		set_page_extent_mapped(page);
+		SetPageUptodate(page);
+		set_page_dirty(page);
+		unlock_page(page);
+		page_cache_release(page);
+		page_cache_release(page);
+		list_del_init(&entry->list);
+		index++;
+	}
+
+	/* Zero out the rest of the pages just to make sure */
+	while (index <= last_index) {
+		void *addr;
+
+		page = find_get_page(inode->i_mapping, index);
+
+		addr = kmap(page);
+		memset(addr, 0, PAGE_CACHE_SIZE);
+		kunmap(page);
+		ClearPageChecked(page);
+		set_page_extent_mapped(page);
+		SetPageUptodate(page);
+		set_page_dirty(page);
+		unlock_page(page);
+		page_cache_release(page);
+		page_cache_release(page);
+		bytes += PAGE_CACHE_SIZE;
+		index++;
+	}
+
+	btrfs_set_extent_delalloc(inode, 0, bytes - 1, &cached_state);
+
+	/* Write the checksums and trans id to the first page */
+	{
+		void *addr;
+		u64 *gen;
+
+		page = find_get_page(inode->i_mapping, 0);
+
+		addr = kmap(page);
+		memcpy(addr, checksums, sizeof(u32) * num_checksums);
+		gen = addr + (sizeof(u32) * num_checksums);
+		*gen = trans->transid;
+		kunmap(page);
+		ClearPageChecked(page);
+		set_page_extent_mapped(page);
+		SetPageUptodate(page);
+		set_page_dirty(page);
+		unlock_page(page);
+		page_cache_release(page);
+		page_cache_release(page);
+	}
+	BTRFS_I(inode)->generation = trans->transid;
+
+	unlock_extent_cached(&BTRFS_I(inode)->io_tree, 0,
+			     i_size_read(inode) - 1, &cached_state, GFP_NOFS);
+
+	filemap_write_and_wait(inode->i_mapping);
+
+	key.objectid = BTRFS_FREE_SPACE_OBJECTID;
+	key.offset = block_group->key.objectid;
+	key.type = 0;
+
+	ret = btrfs_search_slot(trans, root, &key, path, 1, 1);
+	if (ret < 0) {
+		ret = 0;
+		clear_extent_bit(&BTRFS_I(inode)->io_tree, 0, bytes - 1,
+				 EXTENT_DIRTY | EXTENT_DELALLOC |
+				 EXTENT_DO_ACCOUNTING, 0, 0, NULL, GFP_NOFS);
+		goto out_free;
+	}
+	leaf = path->nodes[0];
+	if (ret > 0) {
+		struct btrfs_key found_key;
+		BUG_ON(!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 != block_group->key.objectid) {
+			ret = 0;
+			clear_extent_bit(&BTRFS_I(inode)->io_tree, 0, bytes - 1,
+					 EXTENT_DIRTY | EXTENT_DELALLOC |
+					 EXTENT_DO_ACCOUNTING, 0, 0, NULL,
+					 GFP_NOFS);
+			btrfs_release_path(root, path);
+			goto out_free;
+		}
+	}
+	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(root, path);
+
+	ret = 1;
+
+out_free:
+	if (ret == 0) {
+		invalidate_inode_pages2_range(inode->i_mapping, 0, index);
+		spin_lock(&block_group->lock);
+		block_group->disk_cache_state = BTRFS_DC_ERROR;
+		spin_unlock(&block_group->lock);
+		BTRFS_I(inode)->generation = 0;
+	}
+	kfree(checksums);
+	btrfs_update_inode(trans, root, inode);
+	iput(inode);
+	return ret;
+}
 
 static inline unsigned long offset_to_bit(u64 bitmap_start, u64 sectorsize,
 					  u64 offset)

fs/btrfs/free-space-cache.h

···
 	struct list_head list;
 };
 
+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_truncate_free_space_cache(struct btrfs_root *root,
+				    struct btrfs_trans_handle *trans,
+				    struct btrfs_path *path,
+				    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);
 int btrfs_add_free_space(struct btrfs_block_group_cache *block_group,
 			 u64 bytenr, u64 size);
 int btrfs_remove_free_space(struct btrfs_block_group_cache *block_group,

fs/btrfs/inode.c

···
 	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);
···
 	unsigned long max_uncompressed = 128 * 1024;
 	int i;
 	int will_compress;
-
-	orig_start = start;
 
 	actual_end = min_t(u64, isize, end + 1);
 again:
···
 	total_compressed = min(total_compressed, max_uncompressed);
 	num_bytes = (end - start + blocksize) & ~(blocksize - 1);
 	num_bytes = max(blocksize,  num_bytes);
-	disk_num_bytes = num_bytes;
 	total_in = 0;
 	ret = 0;
 
···
 		if (total_compressed >= total_in) {
 			will_compress = 0;
 		} else {
-			disk_num_bytes = total_compressed;
 			num_bytes = total_in;
 		}
 	}
···
 	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;
 
+	BUG_ON(root == root->fs_info->tree_root);
 	trans = btrfs_join_transaction(root, 1);
 	BUG_ON(!trans);
 	btrfs_set_trans_block_group(trans, inode);
 	trans->block_rsv = &root->fs_info->delalloc_block_rsv;
-
-	actual_end = min_t(u64, isize, end + 1);
 
 	num_bytes = (end - start + blocksize) & ~(blocksize - 1);
 	num_bytes = max(blocksize,  num_bytes);
···
 	int type;
 	int nocow;
 	int check_prev = 1;
+	bool nolock = false;
 
 	path = btrfs_alloc_path();
 	BUG_ON(!path);
-	trans = btrfs_join_transaction(root, 1);
+	if (root == root->fs_info->tree_root) {
+		nolock = true;
+		trans = btrfs_join_transaction_nolock(root, 1);
+	} else {
+		trans = btrfs_join_transaction(root, 1);
+	}
 	BUG_ON(!trans);
 
 	cow_start = (u64)-1;
···
 		BUG_ON(ret);
 	}
 
-	ret = btrfs_end_transaction(trans, root);
-	BUG_ON(ret);
+	if (nolock) {
+		ret = btrfs_end_transaction_nolock(trans, root);
+		BUG_ON(ret);
+	} else {
+		ret = btrfs_end_transaction(trans, root);
+		BUG_ON(ret);
+	}
 	btrfs_free_path(path);
 	return 0;
 }
···
 	if (!(state->state & EXTENT_DELALLOC) && (*bits & EXTENT_DELALLOC)) {
 		struct btrfs_root *root = BTRFS_I(inode)->root;
 		u64 len = state->end + 1 - state->start;
+		int do_list = (root->root_key.objectid !=
+			       BTRFS_ROOT_TREE_OBJECTID);
 
 		if (*bits & EXTENT_FIRST_DELALLOC)
 			*bits &= ~EXTENT_FIRST_DELALLOC;
···
 		spin_lock(&root->fs_info->delalloc_lock);
 		BTRFS_I(inode)->delalloc_bytes += len;
 		root->fs_info->delalloc_bytes += len;
-		if (list_empty(&BTRFS_I(inode)->delalloc_inodes)) {
+		if (do_list && list_empty(&BTRFS_I(inode)->delalloc_inodes)) {
 			list_add_tail(&BTRFS_I(inode)->delalloc_inodes,
 				      &root->fs_info->delalloc_inodes);
 		}
···
 	if ((state->state & EXTENT_DELALLOC) && (*bits & EXTENT_DELALLOC)) {
 		struct btrfs_root *root = BTRFS_I(inode)->root;
 		u64 len = state->end + 1 - state->start;
+		int do_list = (root->root_key.objectid !=
+			       BTRFS_ROOT_TREE_OBJECTID);
 
 		if (*bits & EXTENT_FIRST_DELALLOC)
 			*bits &= ~EXTENT_FIRST_DELALLOC;
···
 		if (*bits & EXTENT_DO_ACCOUNTING)
 			btrfs_delalloc_release_metadata(inode, len);
 
-		if (root->root_key.objectid != BTRFS_DATA_RELOC_TREE_OBJECTID)
+		if (root->root_key.objectid != BTRFS_DATA_RELOC_TREE_OBJECTID
+		    && do_list)
 			btrfs_free_reserved_data_space(inode, len);
 
 		spin_lock(&root->fs_info->delalloc_lock);
 		root->fs_info->delalloc_bytes -= len;
 		BTRFS_I(inode)->delalloc_bytes -= len;
 
-		if (BTRFS_I(inode)->delalloc_bytes == 0 &&
+		if (do_list && BTRFS_I(inode)->delalloc_bytes == 0 &&
 		    !list_empty(&BTRFS_I(inode)->delalloc_inodes)) {
 			list_del_init(&BTRFS_I(inode)->delalloc_inodes);
 		}
···
 
 	if (map_length < length + size)
 		return 1;
-	return 0;
+	return ret;
 }
 
 /*
···
 
 	skip_sum = BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM;
 
-	ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0);
+	if (root == root->fs_info->tree_root)
+		ret = btrfs_bio_wq_end_io(root->fs_info, bio, 2);
+	else
+		ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0);
 	BUG_ON(ret);
 
 	if (!(rw & REQ_WRITE)) {
···
 	struct extent_state *cached_state = NULL;
 	int compressed = 0;
 	int ret;
+	bool nolock = false;
 
 	ret = btrfs_dec_test_ordered_pending(inode, &ordered_extent, start,
 					     end - start + 1);
···
 		return 0;
 	BUG_ON(!ordered_extent);
 
+	nolock = (root == root->fs_info->tree_root);
+
 	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);
+			if (nolock)
+				trans = btrfs_join_transaction_nolock(root, 1);
+			else
+				trans = btrfs_join_transaction(root, 1);
+			BUG_ON(!trans);
 			btrfs_set_trans_block_group(trans, inode);
 			trans->block_rsv = &root->fs_info->delalloc_block_rsv;
 			ret = btrfs_update_inode(trans, root, inode);
···
 			 ordered_extent->file_offset + ordered_extent->len - 1,
 			 0, &cached_state, GFP_NOFS);
 
-	trans = btrfs_join_transaction(root, 1);
+	if (nolock)
+		trans = btrfs_join_transaction_nolock(root, 1);
+	else
+		trans = btrfs_join_transaction(root, 1);
 	btrfs_set_trans_block_group(trans, inode);
 	trans->block_rsv = &root->fs_info->delalloc_block_rsv;
 
···
 						ordered_extent->len);
 		BUG_ON(ret);
 	} else {
+		BUG_ON(root == root->fs_info->tree_root);
 		ret = insert_reserved_file_extent(trans, inode,
 						ordered_extent->file_offset,
 						ordered_extent->start,
···
 	ret = btrfs_update_inode(trans, root, inode);
 	BUG_ON(ret);
 out:
-	btrfs_delalloc_release_metadata(inode, ordered_extent->len);
-	if (trans)
-		btrfs_end_transaction(trans, root);
+	if (nolock) {
+		if (trans)
+			btrfs_end_transaction_nolock(trans, root);
+	} else {
+		btrfs_delalloc_release_metadata(inode, ordered_extent->len);
+		if (trans)
+			btrfs_end_transaction(trans, root);
+	}
+
 	/* once for us */
 	btrfs_put_ordered_extent(ordered_extent);
 	/* once for the tree */
···
 {
 	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;
···
 
 		/* 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 */
···
 
 	ret = btrfs_del_dir_entries_in_log(trans, root, name, name_len,
 					   dir, index);
-	BUG_ON(ret);
+	if (ret == -ENOENT)
+		ret = 0;
 err:
 	btrfs_free_path(path);
 	if (ret)
···
 {
 	struct extent_buffer *eb;
 	int level;
-	int ret;
 	u64 refs = 1;
+	int uninitialized_var(ret);
 
 	for (level = 0; level < BTRFS_MAX_LEVEL; level++) {
 		if (!path->nodes[level])
···
 		if (refs > 1)
 			return 1;
 	}
-	return 0;
+	return ret; /* XXX callers? */
 }
 
 /*
···
 
 	BUG_ON(new_size > 0 && min_type != BTRFS_EXTENT_DATA_KEY);
 
-	if (root->ref_cows)
+	if (root->ref_cows || root == root->fs_info->tree_root)
 		btrfs_drop_extent_cache(inode, new_size & (~mask), (u64)-1, 0);
 
 	path = btrfs_alloc_path();
···
 		} else {
 			break;
 		}
-		if (found_extent && root->ref_cows) {
+		if (found_extent && (root->ref_cows ||
+				     root == root->fs_info->tree_root)) {
 			btrfs_set_path_blocking(path);
 			ret = btrfs_free_extent(trans, root, extent_start,
 						extent_num_bytes, 0,
···
 	int ret;
 
 	truncate_inode_pages(&inode->i_data, 0);
-	if (inode->i_nlink && btrfs_root_refs(&root->root_item) != 0)
+	if (inode->i_nlink && (btrfs_root_refs(&root->root_item) != 0 ||
+			       root == root->fs_info->tree_root))
 		goto no_delete;
 
 	if (is_bad_inode(inode)) {
···
 	}
 	spin_unlock(&root->inode_lock);
 
-	if (empty && btrfs_root_refs(&root->root_item) == 0) {
+	/*
+	 * Free space cache has inodes in the tree root, but the tree root has a
+	 * root_refs of 0, so this could end up dropping the tree root as a
+	 * snapshot, so we need the extra !root->fs_info->tree_root check to
+	 * make sure we don't drop it.
+	 */
+	if (empty && btrfs_root_refs(&root->root_item) == 0 &&
+	    root != root->fs_info->tree_root) {
 		synchronize_srcu(&root->fs_info->subvol_srcu);
 		spin_lock(&root->inode_lock);
 		empty = RB_EMPTY_ROOT(&root->inode_tree);
···
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	struct btrfs_trans_handle *trans;
 	int ret = 0;
+	bool nolock = false;
 
 	if (BTRFS_I(inode)->dummy_inode)
 		return 0;
 
+	smp_mb();
+	nolock = (root->fs_info->closing && root == root->fs_info->tree_root);
+
 	if (wbc->sync_mode == WB_SYNC_ALL) {
-		trans = btrfs_join_transaction(root, 1);
+		if (nolock)
+			trans = btrfs_join_transaction_nolock(root, 1);
+		else
+			trans = btrfs_join_transaction(root, 1);
 		btrfs_set_trans_block_group(trans, inode);
-		ret = btrfs_commit_transaction(trans, root);
+		if (nolock)
+			ret = btrfs_end_transaction_nolock(trans, root);
+		else
+			ret = btrfs_commit_transaction(trans, root);
 	}
 	return ret;
 }
···
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	struct btrfs_dio_private *dip;
 	struct bio_vec *bvec = bio->bi_io_vec;
-	u64 start;
 	int skip_sum;
 	int write = rw & REQ_WRITE;
 	int ret = 0;
···
 	dip->inode = inode;
 	dip->logical_offset = file_offset;
 
-	start = dip->logical_offset;
 	dip->bytes = 0;
 	do {
 		dip->bytes += bvec->bv_len;
···
 		spin_unlock(&root->fs_info->ordered_extent_lock);
 	}
 
+	if (root == root->fs_info->tree_root) {
+		struct btrfs_block_group_cache *block_group;
+
+		block_group = btrfs_lookup_block_group(root->fs_info,
+						BTRFS_I(inode)->block_group);
+		if (block_group && block_group->inode == inode) {
+			spin_lock(&block_group->lock);
+			block_group->inode = NULL;
+			spin_unlock(&block_group->lock);
+			btrfs_put_block_group(block_group);
+		} else if (block_group) {
+			btrfs_put_block_group(block_group);
+		}
+	}
+
 	spin_lock(&root->orphan_lock);
 	if (!list_empty(&BTRFS_I(inode)->i_orphan)) {
 		printk(KERN_INFO "BTRFS: inode %lu still on the orphan list\n",
···
 {
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 
-	if (btrfs_root_refs(&root->root_item) == 0)
+	if (btrfs_root_refs(&root->root_item) == 0 &&
+	    root != root->fs_info->tree_root)
 		return 1;
 	else
 		return generic_drop_inode(inode);
···
 	return 0;
 }
 
-int btrfs_start_one_delalloc_inode(struct btrfs_root *root, int delay_iput)
+int btrfs_start_one_delalloc_inode(struct btrfs_root *root, int delay_iput,
+				   int sync)
 {
 	struct btrfs_inode *binode;
 	struct inode *inode = NULL;
···
 	spin_unlock(&root->fs_info->delalloc_lock);
 
 	if (inode) {
-		write_inode_now(inode, 0);
+		if (sync) {
+			filemap_write_and_wait(inode->i_mapping);
+			/*
+			 * We have to do this because compression doesn't
+			 * actually set PG_writeback until it submits the pages
+			 * for IO, which happens in an async thread, so we could
+			 * race and not actually wait for any writeback pages
+			 * because they've not been submitted yet.  Technically
+			 * this could still be the case for the ordered stuff
+			 * since the async thread may not have started to do its
+			 * work yet.  If this becomes the case then we need to
+			 * figure out a way to make sure that in writepage we
+			 * wait for any async pages to be submitted before
+			 * returning so that fdatawait does what its supposed to
+			 * do.
+			 */
+			btrfs_wait_ordered_range(inode, 0, (u64)-1);
+		} else {
+			filemap_flush(inode->i_mapping);
+		}
 		if (delay_iput)
 			btrfs_add_delayed_iput(inode);
 		else
···
 	return err;
 }
 
-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)
+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 btrfs_root *root = BTRFS_I(inode)->root;
 	struct btrfs_key ins;
 	u64 cur_offset = start;
 	int ret = 0;
+	bool own_trans = true;
 
+	if (trans)
+		own_trans = false;
 	while (num_bytes > 0) {
-		trans = btrfs_start_transaction(root, 3);
-		if (IS_ERR(trans)) {
-			ret = PTR_ERR(trans);
-			break;
+		if (own_trans) {
+			trans = btrfs_start_transaction(root, 3);
+			if (IS_ERR(trans)) {
+				ret = PTR_ERR(trans);
+				break;
+			}
 		}
 
 		ret = btrfs_reserve_extent(trans, root, num_bytes, min_size,
 					   0, *alloc_hint, (u64)-1, &ins, 1);
 		if (ret) {
-			btrfs_end_transaction(trans, root);
+			if (own_trans)
+				btrfs_end_transaction(trans, root);
 			break;
 		}
 
···
 		ret = btrfs_update_inode(trans, root, inode);
 		BUG_ON(ret);
 
-		btrfs_end_transaction(trans, root);
+		if (own_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,

fs/btrfs/ioctl.c

···
 
 static noinline int create_subvol(struct btrfs_root *root,
 				  struct dentry *dentry,
-				  char *name, int namelen)
+				  char *name, int namelen,
+				  u64 *async_transid)
 {
 	struct btrfs_trans_handle *trans;
 	struct btrfs_key key;
···
 
 	d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
 fail:
-	err = btrfs_commit_transaction(trans, root);
+	if (async_transid) {
+		*async_transid = trans->transid;
+		err = btrfs_commit_transaction_async(trans, root, 1);
+	} else {
+		err = btrfs_commit_transaction(trans, root);
+	}
 	if (err && !ret)
 		ret = err;
 	return ret;
 }
 
-static int create_snapshot(struct btrfs_root *root, struct dentry *dentry)
+static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
+			   char *name, int namelen, u64 *async_transid)
 {
 	struct inode *inode;
 	struct btrfs_pending_snapshot *pending_snapshot;
···
 
 	list_add(&pending_snapshot->list,
 		 &trans->transaction->pending_snapshots);
-	ret = btrfs_commit_transaction(trans, root->fs_info->extent_root);
+	if (async_transid) {
+		*async_transid = trans->transid;
+		ret = btrfs_commit_transaction_async(trans,
+				     root->fs_info->extent_root, 1);
+	} else {
+		ret = btrfs_commit_transaction(trans,
+					       root->fs_info->extent_root);
+	}
 	BUG_ON(ret);
 
 	ret = pending_snapshot->error;
···
 	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)
+{
+	uid_t fsuid = current_fsuid();
+
+	if (!(dir->i_mode & S_ISVTX))
+		return 0;
+	if (inode->i_uid == fsuid)
+		return 0;
+	if (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(victim, dir);
+
+	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)
 {
···
  */
 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)
 {
 	struct inode *dir  = parent->dentry->d_inode;
 	struct dentry *dentry;
···
 		goto out_up_read;
 
 	if (snap_src) {
-		error = create_snapshot(snap_src, dentry);
+		error = create_snapshot(snap_src, dentry,
+					name, namelen, async_transid);
 	} else {
 		error = create_subvol(BTRFS_I(dir)->root, dentry,
-				      name, namelen);
+				      name, namelen, async_transid);
 	}
 	if (!error)
 		fsnotify_mkdir(dir, dentry);
···
 	char *sizestr;
 	char *devstr = NULL;
 	int ret = 0;
-	int namelen;
 	int mod = 0;
 
 	if (root->fs_info->sb->s_flags & MS_RDONLY)
···
 		return PTR_ERR(vol_args);
 
 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
-	namelen = strlen(vol_args->name);
 
 	mutex_lock(&root->fs_info->volume_mutex);
 	sizestr = vol_args->name;
···
 	return ret;
 }
 
-static noinline int btrfs_ioctl_snap_create(struct file *file,
-					    void __user *arg, int subvol)
+static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
+						    char *name,
+						    unsigned long fd,
+						    int subvol,
+						    u64 *transid)
 {
 	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;
···
 	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';
-	namelen = strlen(vol_args->name);
-	if (strchr(vol_args->name, '/')) {
+	namelen = strlen(name);
+	if (strchr(name, '/')) {
 		ret = -EINVAL;
 		goto out;
 	}
 
 	if (subvol) {
-		ret = btrfs_mksubvol(&file->f_path, vol_args->name, namelen,
-				     NULL);
+		ret = btrfs_mksubvol(&file->f_path, name, namelen,
+				     NULL, transid);
 	} else {
 		struct inode *src_inode;
-		src_file = fget(vol_args->fd);
+		src_file = fget(fd);
 		if (!src_file) {
 			ret = -EINVAL;
 			goto out;
···
 			fput(src_file);
 			goto out;
 		}
-		ret = btrfs_mksubvol(&file->f_path, vol_args->name, namelen,
-				     BTRFS_I(src_inode)->root);
+		ret = btrfs_mksubvol(&file->f_path, name, namelen,
+				     BTRFS_I(src_inode)->root,
+				     transid);
 		fput(src_file);
 	}
 out:
+	return ret;
+}
+
+static noinline int btrfs_ioctl_snap_create(struct file *file,
+					    void __user *arg, int subvol,
+					    int async)
+{
+	struct btrfs_ioctl_vol_args *vol_args = NULL;
+	struct btrfs_ioctl_async_vol_args *async_vol_args = NULL;
+	char *name;
+	u64 fd;
+	u64 transid = 0;
+	int ret;
+
+	if (async) {
+		async_vol_args = memdup_user(arg, sizeof(*async_vol_args));
+		if (IS_ERR(async_vol_args))
+			return PTR_ERR(async_vol_args);
+
+		name = async_vol_args->name;
+		fd = async_vol_args->fd;
+		async_vol_args->name[BTRFS_SNAPSHOT_NAME_MAX] = '\0';
+	} else {
+		vol_args = memdup_user(arg, sizeof(*vol_args));
+		if (IS_ERR(vol_args))
+			return PTR_ERR(vol_args);
+		name = vol_args->name;
+		fd = vol_args->fd;
+		vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
+	}
+
+	ret = btrfs_ioctl_snap_create_transid(file, name, fd,
+					      subvol, &transid);
+
+	if (!ret && async) {
+		if (copy_to_user(arg +
+				offsetof(struct btrfs_ioctl_async_vol_args,
+				transid), &transid, sizeof(transid)))
+			return -EFAULT;
+	}
+
 	kfree(vol_args);
+	kfree(async_vol_args);
+
 	return ret;
 }
 
···
 	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;
 	ret = search_ioctl(inode, args);
 	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
···
 	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;
 
 	if (args->treeid == 0)
···
 	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);
···
 	}
 
 	inode = dentry->d_inode;
+	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;
+		if (root == dest)
+			goto out_dput;
+
+		err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
+		if (err)
+			goto out_dput;
+
+		/* check if subvolume may be deleted by a non-root user */
+		err = btrfs_may_delete(dir, dentry, 1);
+		if (err)
+			goto out_dput;
+	}
+
 	if (inode->i_ino != BTRFS_FIRST_FREE_OBJECTID) {
 		err = -EINVAL;
 		goto out_dput;
 	}
-
-	dest = BTRFS_I(inode)->root;
 
 	mutex_lock(&inode->i_mutex);
 	err = d_invalidate(dentry);
···
 		BUG_ON(ret);
 	}
 
-	ret = btrfs_commit_transaction(trans, root);
+	ret = btrfs_end_transaction(trans, root);
 	BUG_ON(ret);
 	inode->i_flags |= S_DEAD;
 out_up_write:
···
 	path->reada = 2;
 
 	if (inode < src) {
-		mutex_lock(&inode->i_mutex);
-		mutex_lock(&src->i_mutex);
+		mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
+		mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
 	} else {
-		mutex_lock(&src->i_mutex);
-		mutex_lock(&inode->i_mutex);
+		mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
+		mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
 	}
 
 	/* determine range to clone */
···
 	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)
+		ordered = btrfs_lookup_first_ordered_extent(src, off+len);
+		if (!ordered &&
+		    !test_range_bit(&BTRFS_I(src)->io_tree, off, off+len,
+				   EXTENT_DELALLOC, 0, NULL))
 			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_wait_ordered_range(src, off, len);
 	}
 
 	/* clone data */
···
 			}
 			btrfs_release_path(root, path);
 
-			if (key.offset + datal < off ||
+			if (key.offset + datal <= off ||
 			    key.offset >= off+len)
 				goto next;
 
···
 	return 0;
 }
 
+static void 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)
 {
 	struct btrfs_ioctl_space_args space_args;
···
 	struct btrfs_ioctl_space_info *dest_orig;
 	struct btrfs_ioctl_space_info *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;
+	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);
+	}
 
 	/* 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(int, 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
 	 */
···
 	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)
-			break;
+		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();
 
-		space.flags = info->flags;
-		space.total_bytes = info->total_bytes;
-		space.used_bytes = info->bytes_used;
-		memcpy(dest, &space, sizeof(space));
-		dest++;
-		space_args.total_spaces++;
+		if (!info)
+			continue;
+		down_read(&info->groups_sem);
+		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
+			if (!list_empty(&info->block_groups[c])) {
+				get_block_group_info(&info->block_groups[c],
+						     &space);
+				memcpy(dest, &space, sizeof(space));
+				dest++;
+				space_args.total_spaces++;
+			}
+		}
+		up_read(&info->groups_sem);
 	}
-	rcu_read_unlock();
 
 	user_dest = (struct btrfs_ioctl_space_info *)
 		(arg + sizeof(struct btrfs_ioctl_space_args));
···
 	return 0;
 }
 
+static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp)
+{
+	struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
+	struct btrfs_trans_handle *trans;
+	u64 transid;
+
+	trans = btrfs_start_transaction(root, 0);
+	transid = trans->transid;
+	btrfs_commit_transaction_async(trans, root, 0);
+
+	if (argp)
+		if (copy_to_user(argp, &transid, sizeof(transid)))
+			return -EFAULT;
+	return 0;
+}
+
+static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp)
+{
+	struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
+	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);
+}
+
 long btrfs_ioctl(struct file *file, unsigned int
 		cmd, unsigned long arg)
 {
···
 	case FS_IOC_GETVERSION:
 		return btrfs_ioctl_getversion(file, argp);
 	case BTRFS_IOC_SNAP_CREATE:
-		return btrfs_ioctl_snap_create(file, argp, 0);
+		return btrfs_ioctl_snap_create(file, argp, 0, 0);
+	case BTRFS_IOC_SNAP_CREATE_ASYNC:
+		return btrfs_ioctl_snap_create(file, argp, 0, 1);
 	case BTRFS_IOC_SUBVOL_CREATE:
-		return btrfs_ioctl_snap_create(file, argp, 1);
+		return btrfs_ioctl_snap_create(file, argp, 1, 0);
 	case BTRFS_IOC_SNAP_DESTROY:
 		return btrfs_ioctl_snap_destroy(file, argp);
 	case BTRFS_IOC_DEFAULT_SUBVOL:
···
 	case BTRFS_IOC_SYNC:
 		btrfs_sync_fs(file->f_dentry->d_sb, 1);
 		return 0;
+	case BTRFS_IOC_START_SYNC:
+		return btrfs_ioctl_start_sync(file, argp);
+	case BTRFS_IOC_WAIT_SYNC:
+		return btrfs_ioctl_wait_sync(file, argp);
 	}
 
 	return -ENOTTY;

fs/btrfs/ioctl.h

···
 
 #define BTRFS_IOCTL_MAGIC 0x94
 #define BTRFS_VOL_NAME_MAX 255
-#define BTRFS_PATH_NAME_MAX 4087
 
 /* this should be 4k */
+#define BTRFS_PATH_NAME_MAX 4087
 struct btrfs_ioctl_vol_args {
 	__s64 fd;
 	char name[BTRFS_PATH_NAME_MAX + 1];
+};
+
+#define BTRFS_SNAPSHOT_NAME_MAX 4079
+struct btrfs_ioctl_async_vol_args {
+	__s64 fd;
+	__u64 transid;
+	char name[BTRFS_SNAPSHOT_NAME_MAX + 1];
 };
 
 #define BTRFS_INO_LOOKUP_PATH_MAX 4080
···
 #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)
+#define BTRFS_IOC_START_SYNC _IOR(BTRFS_IOCTL_MAGIC, 24, __u64)
+#define BTRFS_IOC_WAIT_SYNC  _IOW(BTRFS_IOCTL_MAGIC, 22, __u64)
+#define BTRFS_IOC_SNAP_CREATE_ASYNC _IOW(BTRFS_IOCTL_MAGIC, 23, \
+				   struct btrfs_ioctl_async_vol_args)
 #endif

fs/btrfs/ordered-data.c

···
 {
 	u64 end;
 	u64 orig_end;
-	u64 wait_end;
 	struct btrfs_ordered_extent *ordered;
 	int found;
 
···
 		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

fs/btrfs/relocation.c

···
 #include "locking.h"
 #include "btrfs_inode.h"
 #include "async-thread.h"
+#include "free-space-cache.h"
 
 /*
  * backref_node, mapping_node and tree_block start with this
···
 
 	u64 search_start;
 	u64 extents_found;
-
-	int block_rsv_retries;
 
 	unsigned int stage:8;
 	unsigned int create_reloc_tree:1;
···
 	LIST_HEAD(reloc_roots);
 	u64 num_bytes = 0;
 	int ret;
-	int retries = 0;
 
 	mutex_lock(&root->fs_info->trans_mutex);
 	rc->merging_rsv_size += root->nodesize * (BTRFS_MAX_LEVEL - 1) * 2;
···
 	if (!err) {
 		num_bytes = rc->merging_rsv_size;
 		ret = btrfs_block_rsv_add(NULL, root, rc->block_rsv,
-					  num_bytes, &retries);
+					  num_bytes);
 		if (ret)
 			err = ret;
 	}
···
 			btrfs_end_transaction(trans, rc->extent_root);
 			btrfs_block_rsv_release(rc->extent_root,
 						rc->block_rsv, num_bytes);
-			retries = 0;
 			goto again;
 		}
 	}
···
 	num_bytes = calcu_metadata_size(rc, node, 1) * 2;
 
 	trans->block_rsv = rc->block_rsv;
-	ret = btrfs_block_rsv_add(trans, root, rc->block_rsv, num_bytes,
-				  &rc->block_rsv_retries);
+	ret = btrfs_block_rsv_add(trans, root, rc->block_rsv, num_bytes);
 	if (ret) {
 		if (ret == -EAGAIN)
 			rc->commit_transaction = 1;
 		return ret;
 	}
 
-	rc->block_rsv_retries = 0;
 	return 0;
 }
 
···
 		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 */
···
 	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_path *path;
+	struct btrfs_root *root = fs_info->tree_root;
+	struct btrfs_trans_handle *trans;
+	unsigned long nr;
+	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 (!inode || IS_ERR(inode) || is_bad_inode(inode)) {
+		if (inode && !IS_ERR(inode))
+			iput(inode);
+		return -ENOENT;
+	}
+
+truncate:
+	path = btrfs_alloc_path();
+	if (!path) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	trans = btrfs_join_transaction(root, 0);
+	if (IS_ERR(trans)) {
+		btrfs_free_path(path);
+		goto out;
+	}
+
+	ret = btrfs_truncate_free_space_cache(root, trans, path, inode);
+
+	btrfs_free_path(path);
+	nr = trans->blocks_used;
+	btrfs_end_transaction(trans, root);
+	btrfs_btree_balance_dirty(root, nr);
+out:
+	iput(inode);
+	return ret;
+}
+
 /*
  * helper to add tree blocks for backref of type BTRFS_EXTENT_DATA_REF_KEY
  * this function scans fs tree to find blocks reference the data extent
···
 	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;
 
 	root = read_fs_root(rc->extent_root->fs_info, ref_root);
 	if (IS_ERR(root)) {
···
 	 * is no reservation in transaction handle.
 	 */
 	ret = btrfs_block_rsv_add(NULL, rc->extent_root, rc->block_rsv,
-				  rc->extent_root->nodesize * 256,
-				  &rc->block_rsv_retries);
+				  rc->extent_root->nodesize * 256);
 	if (ret)
 		return ret;
 
···
 	rc->extents_found = 0;
 	rc->nodes_relocated = 0;
 	rc->merging_rsv_size = 0;
-	rc->block_rsv_retries = 0;
 
 	rc->create_reloc_tree = 1;
 	set_reloc_control(rc);
···
 {
 	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;
···
 			goto out;
 		}
 		rw = 1;
+	}
+
+	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);
···
 		btrfs_add_ordered_sum(inode, ordered, sums);
 	}
 	btrfs_put_ordered_extent(ordered);
-	return 0;
+	return ret;
 }
 
 void btrfs_reloc_cow_block(struct btrfs_trans_handle *trans,

fs/btrfs/root-tree.c

···
 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;
···
 			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;

fs/btrfs/super.c

···
 
 	ret = close_ctree(root);
 	sb->s_fs_info = NULL;
+
+	(void)ret; /* FIXME: need to fix VFS to return error? */
 }
 
 enum {
···
 	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_discard, Opt_space_cache, Opt_clear_cache, Opt_err,
+	Opt_user_subvol_rm_allowed,
 };
 
 static match_table_t tokens = {
···
 	{Opt_flushoncommit, "flushoncommit"},
 	{Opt_ratio, "metadata_ratio=%d"},
 	{Opt_discard, "discard"},
+	{Opt_space_cache, "space_cache"},
+	{Opt_clear_cache, "clear_cache"},
+	{Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
 	{Opt_err, NULL},
 };
 
···
 			break;
 		case Opt_discard:
 			btrfs_set_opt(info->mount_opt, DISCARD);
+			break;
+		case Opt_space_cache:
+			printk(KERN_INFO "btrfs: enabling disk space caching\n");
+			btrfs_set_opt(info->mount_opt, SPACE_CACHE);
+		case Opt_clear_cache:
+			printk(KERN_INFO "btrfs: force clearing of disk cache\n");
+			btrfs_set_opt(info->mount_opt, CLEAR_CACHE);
+			break;
+		case Opt_user_subvol_rm_allowed:
+			btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
 			break;
 		case Opt_err:
 			printk(KERN_INFO "btrfs: unrecognized mount option "
···
 find_root:
 	new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
 	if (IS_ERR(new_root))
-		return ERR_PTR(PTR_ERR(new_root));
+		return ERR_CAST(new_root);
 
 	if (btrfs_root_refs(&new_root->root_item) == 0)
 		return ERR_PTR(-ENOENT);
···
 {
 	struct inode *inode;
 	struct dentry *root_dentry;
-	struct btrfs_super_block *disk_super;
 	struct btrfs_root *tree_root;
 	struct btrfs_key key;
 	int err;
···
 		return PTR_ERR(tree_root);
 	}
 	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;
···
 	char *subvol_name = NULL;
 	u64 subvol_objectid = 0;
 	int error = 0;
-	int found = 0;
 
 	if (!(flags & MS_RDONLY))
 		mode |= FMODE_WRITE;
···
 			goto error_close_devices;
 		}
 
-		found = 1;
 		btrfs_close_devices(fs_devices);
 	} else {
 		char b[BDEVNAME_SIZE];
···
 	if (IS_ERR(root)) {
 		error = PTR_ERR(root);
 		deactivate_locked_super(s);
-		goto error;
+		goto error_free_subvol_name;
 	}
 	/* if they gave us a subvolume name bind mount into that */
 	if (strcmp(subvol_name, ".")) {
···
 			deactivate_locked_super(s);
 			error = PTR_ERR(new_root);
 			dput(root);
-			goto error_close_devices;
+			goto error_free_subvol_name;
 		}
 		if (!new_root->d_inode) {
 			dput(root);
 			dput(new_root);
 			deactivate_locked_super(s);
 			error = -ENXIO;
-			goto error_close_devices;
+			goto error_free_subvol_name;
 		}
 		dput(root);
 		root = new_root;
···
 	btrfs_close_devices(fs_devices);
 error_free_subvol_name:
 	kfree(subvol_name);
-error:
 	return ERR_PTR(error);
 }
 
···
 	struct list_head *head = &root->fs_info->space_info;
 	struct btrfs_space_info *found;
 	u64 total_used = 0;
+	u64 total_used_data = 0;
 	int bits = dentry->d_sb->s_blocksize_bits;
 	__be32 *fsid = (__be32 *)root->fs_info->fsid;
 
 	rcu_read_lock();
-	list_for_each_entry_rcu(found, head, list)
+	list_for_each_entry_rcu(found, head, list) {
+		if (found->flags & (BTRFS_BLOCK_GROUP_METADATA |
+				    BTRFS_BLOCK_GROUP_SYSTEM))
+			total_used_data += found->disk_total;
+		else
+			total_used_data += found->disk_used;
 		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_bfree;
+	buf->f_bavail = buf->f_blocks - (total_used_data >> bits);
 	buf->f_bsize = dentry->d_sb->s_blocksize;
 	buf->f_type = BTRFS_SUPER_MAGIC;
 

fs/btrfs/transaction.c

···
 	TRANS_START,
 	TRANS_JOIN,
 	TRANS_USERSPACE,
+	TRANS_JOIN_NOLOCK,
 };
 
 static int may_wait_transaction(struct btrfs_root *root, int type)
···
 {
 	struct btrfs_trans_handle *h;
 	struct btrfs_transaction *cur_trans;
-	int retries = 0;
 	int ret;
 again:
 	h = kmem_cache_alloc(btrfs_trans_handle_cachep, GFP_NOFS);
 	if (!h)
 		return ERR_PTR(-ENOMEM);
 
-	mutex_lock(&root->fs_info->trans_mutex);
+	if (type != TRANS_JOIN_NOLOCK)
+		mutex_lock(&root->fs_info->trans_mutex);
 	if (may_wait_transaction(root, type))
 		wait_current_trans(root);
 
···
 
 	cur_trans = root->fs_info->running_transaction;
 	cur_trans->use_count++;
-	mutex_unlock(&root->fs_info->trans_mutex);
+	if (type != TRANS_JOIN_NOLOCK)
+		mutex_unlock(&root->fs_info->trans_mutex);
 
 	h->transid = cur_trans->transid;
 	h->transaction = cur_trans;
···
 	}
 
 	if (num_items > 0) {
-		ret = btrfs_trans_reserve_metadata(h, root, num_items,
-						   &retries);
+		ret = btrfs_trans_reserve_metadata(h, root, num_items);
 		if (ret == -EAGAIN) {
 			btrfs_commit_transaction(h, root);
 			goto again;
···
 		}
 	}
 
-	mutex_lock(&root->fs_info->trans_mutex);
+	if (type != TRANS_JOIN_NOLOCK)
+		mutex_lock(&root->fs_info->trans_mutex);
 	record_root_in_trans(h, root);
-	mutex_unlock(&root->fs_info->trans_mutex);
+	if (type != TRANS_JOIN_NOLOCK)
+		mutex_unlock(&root->fs_info->trans_mutex);
 
 	if (!current->journal_info && type != TRANS_USERSPACE)
 		current->journal_info = h;
···
 						   int num_blocks)
 {
 	return start_transaction(root, 0, TRANS_JOIN);
+}
+
+struct btrfs_trans_handle *btrfs_join_transaction_nolock(struct btrfs_root *root,
+							  int num_blocks)
+{
+	return start_transaction(root, 0, TRANS_JOIN_NOLOCK);
 }
 
 struct btrfs_trans_handle *btrfs_start_ioctl_transaction(struct btrfs_root *r,
···
 	mutex_unlock(&root->fs_info->trans_mutex);
 	finish_wait(&commit->commit_wait, &wait);
 	return 0;
+}
+
+int btrfs_wait_for_commit(struct btrfs_root *root, u64 transid)
+{
+	struct btrfs_transaction *cur_trans = NULL, *t;
+	int ret;
+
+	mutex_lock(&root->fs_info->trans_mutex);
+
+	ret = 0;
+	if (transid) {
+		if (transid <= root->fs_info->last_trans_committed)
+			goto out_unlock;
+
+		/* find specified transaction */
+		list_for_each_entry(t, &root->fs_info->trans_list, list) {
+			if (t->transid == transid) {
+				cur_trans = t;
+				break;
+			}
+			if (t->transid > transid)
+				break;
+		}
+		ret = -EINVAL;
+		if (!cur_trans)
+			goto out_unlock;  /* bad transid */
+	} else {
+		/* find newest transaction that is committing | committed */
+		list_for_each_entry_reverse(t, &root->fs_info->trans_list,
+					    list) {
+			if (t->in_commit) {
+				if (t->commit_done)
+					goto out_unlock;
+				cur_trans = t;
+				break;
+			}
+		}
+		if (!cur_trans)
+			goto out_unlock;  /* nothing committing|committed */
+	}
+
+	cur_trans->use_count++;
+	mutex_unlock(&root->fs_info->trans_mutex);
+
+	wait_for_commit(root, cur_trans);
+
+	mutex_lock(&root->fs_info->trans_mutex);
+	put_transaction(cur_trans);
+	ret = 0;
+out_unlock:
+	mutex_unlock(&root->fs_info->trans_mutex);
+	return ret;
 }
 
 #if 0
···
 }
 
 static int __btrfs_end_transaction(struct btrfs_trans_handle *trans,
-			  struct btrfs_root *root, int throttle)
+			  struct btrfs_root *root, int throttle, int lock)
 {
 	struct btrfs_transaction *cur_trans = trans->transaction;
 	struct btrfs_fs_info *info = root->fs_info;
···
 
 	btrfs_trans_release_metadata(trans, root);
 
-	if (!root->fs_info->open_ioctl_trans &&
+	if (lock && !root->fs_info->open_ioctl_trans &&
 	    should_end_transaction(trans, root))
 		trans->transaction->blocked = 1;
 
-	if (cur_trans->blocked && !cur_trans->in_commit) {
+	if (lock && cur_trans->blocked && !cur_trans->in_commit) {
 		if (throttle)
 			return btrfs_commit_transaction(trans, root);
 		else
 			wake_up_process(info->transaction_kthread);
 	}
 
-	mutex_lock(&info->trans_mutex);
+	if (lock)
+		mutex_lock(&info->trans_mutex);
 	WARN_ON(cur_trans != info->running_transaction);
 	WARN_ON(cur_trans->num_writers < 1);
 	cur_trans->num_writers--;
 
+	smp_mb();
 	if (waitqueue_active(&cur_trans->writer_wait))
 		wake_up(&cur_trans->writer_wait);
 	put_transaction(cur_trans);
-	mutex_unlock(&info->trans_mutex);
+	if (lock)
+		mutex_unlock(&info->trans_mutex);
 
 	if (current->journal_info == trans)
 		current->journal_info = NULL;
···
 int btrfs_end_transaction(struct btrfs_trans_handle *trans,
 			  struct btrfs_root *root)
 {
-	return __btrfs_end_transaction(trans, root, 0);
+	return __btrfs_end_transaction(trans, root, 0, 1);
 }
 
 int btrfs_end_transaction_throttle(struct btrfs_trans_handle *trans,
 				   struct btrfs_root *root)
 {
-	return __btrfs_end_transaction(trans, root, 1);
+	return __btrfs_end_transaction(trans, root, 1, 1);
+}
+
+int btrfs_end_transaction_nolock(struct btrfs_trans_handle *trans,
+				 struct btrfs_root *root)
+{
+	return __btrfs_end_transaction(trans, root, 0, 0);
 }
 
 /*
···
 	struct extent_buffer *tmp;
 	struct extent_buffer *old;
 	int ret;
-	int retries = 0;
 	u64 to_reserve = 0;
 	u64 index = 0;
 	u64 objectid;
···
 
 	if (to_reserve > 0) {
 		ret = btrfs_block_rsv_add(trans, root, &pending->block_rsv,
-					  to_reserve, &retries);
+					  to_reserve);
 		if (ret) {
 			pending->error = ret;
 			goto fail;
···
 	super->root = root_item->bytenr;
 	super->generation = root_item->generation;
 	super->root_level = root_item->level;
+	if (super->cache_generation != 0 || btrfs_test_opt(root, SPACE_CACHE))
+		super->cache_generation = root_item->generation;
 }
 
 int btrfs_transaction_in_commit(struct btrfs_fs_info *info)
···
 	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)
+{
+	DEFINE_WAIT(wait);
+
+	if (trans->in_commit)
+		return;
+
+	while (1) {
+		prepare_to_wait(&root->fs_info->transaction_blocked_wait, &wait,
+				TASK_UNINTERRUPTIBLE);
+		if (trans->in_commit) {
+			finish_wait(&root->fs_info->transaction_blocked_wait,
+				    &wait);
+			break;
+		}
+		mutex_unlock(&root->fs_info->trans_mutex);
+		schedule();
+		mutex_lock(&root->fs_info->trans_mutex);
+		finish_wait(&root->fs_info->transaction_blocked_wait, &wait);
+	}
+}
+
+/*
+ * 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)
+{
+	DEFINE_WAIT(wait);
+
+	if (trans->commit_done || (trans->in_commit && !trans->blocked))
+		return;
+
+	while (1) {
+		prepare_to_wait(&root->fs_info->transaction_wait, &wait,
+				TASK_UNINTERRUPTIBLE);
+		if (trans->commit_done ||
+		    (trans->in_commit && !trans->blocked)) {
+			finish_wait(&root->fs_info->transaction_wait,
+				    &wait);
+			break;
+		}
+		mutex_unlock(&root->fs_info->trans_mutex);
+		schedule();
+		mutex_lock(&root->fs_info->trans_mutex);
+		finish_wait(&root->fs_info->transaction_wait,
+			    &wait);
+	}
+}
+
+/*
+ * 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 delayed_work work;
+};
+
+static void do_async_commit(struct work_struct *work)
+{
+	struct btrfs_async_commit *ac =
+		container_of(work, struct btrfs_async_commit, work.work);
+
+	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);
+	BUG_ON(!ac);
+
+	INIT_DELAYED_WORK(&ac->work, do_async_commit);
+	ac->root = root;
+	ac->newtrans = btrfs_join_transaction(root, 0);
+
+	/* take transaction reference */
+	mutex_lock(&root->fs_info->trans_mutex);
+	cur_trans = trans->transaction;
+	cur_trans->use_count++;
+	mutex_unlock(&root->fs_info->trans_mutex);
+
+	btrfs_end_transaction(trans, root);
+	schedule_delayed_work(&ac->work, 0);
+
+	/* wait for transaction to start and unblock */
+	mutex_lock(&root->fs_info->trans_mutex);
+	if (wait_for_unblock)
+		wait_current_trans_commit_start_and_unblock(root, cur_trans);
+	else
+		wait_current_trans_commit_start(root, cur_trans);
+	put_transaction(cur_trans);
+	mutex_unlock(&root->fs_info->trans_mutex);
+
+	return 0;
+}
+
+/*
+ * btrfs_transaction state sequence:
+ *    in_commit = 0, blocked = 0  (initial)
+ *    in_commit = 1, blocked = 1
+ *    blocked = 0
+ *    commit_done = 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 *prev_trans = NULL;
 	DEFINE_WAIT(wait);
···
 
 	trans->transaction->in_commit = 1;
 	trans->transaction->blocked = 1;
+	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);
···
 			snap_pending = 1;
 
 		WARN_ON(cur_trans != trans->transaction);
-		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) {
···
 				TASK_UNINTERRUPTIBLE);
 
 		smp_mb();
-		if (cur_trans->num_writers > 1 || should_grow)
-			schedule_timeout(timeout);
+		if (cur_trans->num_writers > 1)
+			schedule_timeout(MAX_SCHEDULE_TIMEOUT);
+		else if (should_grow)
+			schedule_timeout(1);
 
 		mutex_lock(&root->fs_info->trans_mutex);
 		finish_wait(&cur_trans->writer_wait, &wait);

fs/btrfs/transaction.h

···
 
 int btrfs_end_transaction(struct btrfs_trans_handle *trans,
 			  struct btrfs_root *root);
+int btrfs_end_transaction_nolock(struct btrfs_trans_handle *trans,
+				 struct btrfs_root *root);
 struct btrfs_trans_handle *btrfs_start_transaction(struct btrfs_root *root,
 						   int num_items);
 struct btrfs_trans_handle *btrfs_join_transaction(struct btrfs_root *root,
 						  int num_blocks);
+struct btrfs_trans_handle *btrfs_join_transaction_nolock(struct btrfs_root *root,
+							  int num_blocks);
 struct btrfs_trans_handle *btrfs_start_ioctl_transaction(struct btrfs_root *r,
 							 int num_blocks);
+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,
···
 int btrfs_clean_old_snapshots(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,

fs/btrfs/tree-defrag.c

···
 	int ret = 0;
 	int wret;
 	int level;
-	int orig_level;
 	int is_extent = 0;
 	int next_key_ret = 0;
 	u64 last_ret = 0;
···
 		return -ENOMEM;
 
 	level = btrfs_header_level(root->node);
-	orig_level = level;
 
 	if (level == 0)
 		goto out;

fs/btrfs/tree-log.c

···
 {
 	struct inode *dir;
 	int ret;
-	struct btrfs_key location;
 	struct btrfs_inode_ref *ref;
 	struct btrfs_dir_item *di;
 	struct inode *inode;
···
 	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
···
 	struct btrfs_path *path;
 	struct btrfs_root *root = wc->replay_dest;
 	struct btrfs_key key;
-	u32 item_size;
 	int level;
 	int i;
 	int ret;
···
 	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 &&
···
 				   struct walk_control *wc)
 {
 	u64 root_owner;
-	u64 root_gen;
 	u64 bytenr;
 	u64 ptr_gen;
 	struct extent_buffer *next;
···
 
 		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);
 
···
 				 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 + 1 < btrfs_header_nritems(path->nodes[i])) {
-			struct extent_buffer *node;
-			node = path->nodes[i];
 			path->slots[i]++;
 			*level = i;
 			WARN_ON(*level == 0);
···
 				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,
 				 btrfs_header_generation(path->nodes[*level]));
 			if (wc->free) {
···
 	}
 	btrfs_end_log_trans(root);
 
-	return 0;
+	return err;
 }
 
 /* see comments for btrfs_del_dir_entries_in_log */
···
 	struct btrfs_key max_key;
 	struct btrfs_root *log = root->log_root;
 	struct extent_buffer *src = NULL;
-	u32 size;
 	int err = 0;
 	int ret;
 	int nritems;
···
 			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;

fs/btrfs/volumes.c

···
 	u64 size_to_free;
 	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 (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;
···
 		}
 		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) {
+		if (dev && dev->bdev && (rw != WRITE || dev->writeable)) {
 			bio->bi_bdev = dev->bdev;
 			if (async_submit)
 				schedule_bio(root, dev, rw, bio);

fs/btrfs/xattr.c

···
 	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;
···
 		}
 		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 */

fs/btrfs/zlib.c

···
 	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;
···
 	workspace->def_strm.next_out = cpage_out;
 	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);

fs/fs-writeback.c

···
 }
 
 /**
- * writeback_inodes_sb	-	writeback dirty inodes from given super_block
+ * writeback_inodes_sb_nr -	writeback dirty inodes from given super_block
  * @sb: the superblock
+ * @nr: the number of pages to write
  *
  * Start writeback on some inodes on this super_block. No guarantees are made
  * on how many (if any) will be written, and this function does not wait
- * for IO completion of submitted IO. The number of pages submitted is
- * returned.
+ * for IO completion of submitted IO.
  */
-void writeback_inodes_sb(struct super_block *sb)
+void writeback_inodes_sb_nr(struct super_block *sb, unsigned long nr)
 {
 	DECLARE_COMPLETION_ONSTACK(done);
 	struct wb_writeback_work work = {
 		.sb		= sb,
 		.sync_mode	= WB_SYNC_NONE,
 		.done		= &done,
+		.nr_pages	= nr,
 	};
 
 	WARN_ON(!rwsem_is_locked(&sb->s_umount));
-
-	work.nr_pages = get_nr_dirty_pages();
-
 	bdi_queue_work(sb->s_bdi, &work);
 	wait_for_completion(&done);
+}
+EXPORT_SYMBOL(writeback_inodes_sb_nr);
+
+/**
+ * writeback_inodes_sb	-	writeback dirty inodes from given super_block
+ * @sb: the superblock
+ *
+ * Start writeback on some inodes on this super_block. No guarantees are made
+ * on how many (if any) will be written, and this function does not wait
+ * for IO completion of submitted IO.
+ */
+void writeback_inodes_sb(struct super_block *sb)
+{
+	return writeback_inodes_sb_nr(sb, get_nr_dirty_pages());
 }
 EXPORT_SYMBOL(writeback_inodes_sb);
 
···
 		return 0;
 }
 EXPORT_SYMBOL(writeback_inodes_sb_if_idle);
+
+/**
+ * writeback_inodes_sb_if_idle	-	start writeback if none underway
+ * @sb: the superblock
+ * @nr: the number of pages to write
+ *
+ * Invoke writeback_inodes_sb if no writeback is currently underway.
+ * Returns 1 if writeback was started, 0 if not.
+ */
+int writeback_inodes_sb_nr_if_idle(struct super_block *sb,
+				   unsigned long nr)
+{
+	if (!writeback_in_progress(sb->s_bdi)) {
+		down_read(&sb->s_umount);
+		writeback_inodes_sb_nr(sb, nr);
+		up_read(&sb->s_umount);
+		return 1;
+	} else
+		return 0;
+}
+EXPORT_SYMBOL(writeback_inodes_sb_nr_if_idle);
 
 /**
  * sync_inodes_sb	-	sync sb inode pages

include/linux/writeback.h

···
 struct bdi_writeback;
 int inode_wait(void *);
 void writeback_inodes_sb(struct super_block *);
+void writeback_inodes_sb_nr(struct super_block *, unsigned long nr);
 int writeback_inodes_sb_if_idle(struct super_block *);
+int writeback_inodes_sb_nr_if_idle(struct super_block *, unsigned long nr);
 void sync_inodes_sb(struct super_block *);
 void writeback_inodes_wb(struct bdi_writeback *wb,
 		struct writeback_control *wbc);