Merge branch 'freespace-tree' into for-linus-4.5

Signed-off-by: Chris Mason <clm@fb.com>

15 files changed, 2932 insertions, 112 deletions

diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile
index 6d1d0b93b1aa..128ce17a80b0 100644
--- a/fs/btrfs/Makefile
+++ b/fs/btrfs/Makefile
@@ -9,11 +9,12 @@ btrfs-y += super.o ctree.o extent-tree.o print-tree.o root-tree.o dir-item.o \
 	   export.o tree-log.o free-space-cache.o zlib.o lzo.o \
 	   compression.o delayed-ref.o relocation.o delayed-inode.o scrub.o \
 	   reada.o backref.o ulist.o qgroup.o send.o dev-replace.o raid56.o \
-	   uuid-tree.o props.o hash.o
+	   uuid-tree.o props.o hash.o free-space-tree.o
 
 btrfs-$(CONFIG_BTRFS_FS_POSIX_ACL) += acl.o
 btrfs-$(CONFIG_BTRFS_FS_CHECK_INTEGRITY) += check-integrity.o
 
 btrfs-$(CONFIG_BTRFS_FS_RUN_SANITY_TESTS) += tests/free-space-tests.o \
 	tests/extent-buffer-tests.o tests/btrfs-tests.o \
-	tests/extent-io-tests.o tests/inode-tests.o tests/qgroup-tests.o
+	tests/extent-io-tests.o tests/inode-tests.o tests/qgroup-tests.o \
+	tests/free-space-tree-tests.o
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 35489e7129a7..cf87979a153e 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -96,6 +96,9 @@ struct btrfs_ordered_sum;
 /* for storing items that use the BTRFS_UUID_KEY* types */
 #define BTRFS_UUID_TREE_OBJECTID 9ULL
 
+/* tracks free space in block groups. */
+#define BTRFS_FREE_SPACE_TREE_OBJECTID 10ULL
+
 /* for storing balance parameters in the root tree */
 #define BTRFS_BALANCE_OBJECTID -4ULL
 
@@ -500,6 +503,8 @@ struct btrfs_super_block {
  * Compat flags that we support.  If any incompat flags are set other than the
  * ones specified below then we will fail to mount
  */
+#define BTRFS_FEATURE_COMPAT_RO_FREE_SPACE_TREE	(1ULL << 0)
+
 #define BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF	(1ULL << 0)
 #define BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL	(1ULL << 1)
 #define BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS	(1ULL << 2)
@@ -526,7 +531,10 @@ struct btrfs_super_block {
 #define BTRFS_FEATURE_COMPAT_SUPP		0ULL
 #define BTRFS_FEATURE_COMPAT_SAFE_SET		0ULL
 #define BTRFS_FEATURE_COMPAT_SAFE_CLEAR		0ULL
-#define BTRFS_FEATURE_COMPAT_RO_SUPP		0ULL
+
+#define BTRFS_FEATURE_COMPAT_RO_SUPP			\
+	(BTRFS_FEATURE_COMPAT_RO_FREE_SPACE_TREE)
+
 #define BTRFS_FEATURE_COMPAT_RO_SAFE_SET	0ULL
 #define BTRFS_FEATURE_COMPAT_RO_SAFE_CLEAR	0ULL
 
@@ -1088,6 +1096,13 @@ struct btrfs_block_group_item {
 	__le64 flags;
 } __attribute__ ((__packed__));
 
+struct btrfs_free_space_info {
+	__le32 extent_count;
+	__le32 flags;
+} __attribute__ ((__packed__));
+
+#define BTRFS_FREE_SPACE_USING_BITMAPS (1ULL << 0)
+
 #define BTRFS_QGROUP_LEVEL_SHIFT		48
 static inline u64 btrfs_qgroup_level(u64 qgroupid)
 {
@@ -1296,6 +1311,9 @@ struct btrfs_caching_control {
 	atomic_t count;
 };
 
+/* Once caching_thread() finds this much free space, it will wake up waiters. */
+#define CACHING_CTL_WAKE_UP (1024 * 1024 * 2)
+
 struct btrfs_io_ctl {
 	void *cur, *orig;
 	struct page *page;
@@ -1321,8 +1339,20 @@ struct btrfs_block_group_cache {
 	u64 delalloc_bytes;
 	u64 bytes_super;
 	u64 flags;
-	u64 sectorsize;
 	u64 cache_generation;
+	u32 sectorsize;
+
+	/*
+	 * If the free space extent count exceeds this number, convert the block
+	 * group to bitmaps.
+	 */
+	u32 bitmap_high_thresh;
+
+	/*
+	 * If the free space extent count drops below this number, convert the
+	 * block group back to extents.
+	 */
+	u32 bitmap_low_thresh;
 
 	/*
 	 * It is just used for the delayed data space allocation because
@@ -1378,6 +1408,15 @@ struct btrfs_block_group_cache {
 	struct list_head io_list;
 
 	struct btrfs_io_ctl io_ctl;
+
+	/* Lock for free space tree operations. */
+	struct mutex free_space_lock;
+
+	/*
+	 * Does the block group need to be added to the free space tree?
+	 * Protected by free_space_lock.
+	 */
+	int needs_free_space;
 };
 
 /* delayed seq elem */
@@ -1429,6 +1468,7 @@ struct btrfs_fs_info {
 	struct btrfs_root *csum_root;
 	struct btrfs_root *quota_root;
 	struct btrfs_root *uuid_root;
+	struct btrfs_root *free_space_root;
 
 	/* the log root tree is a directory of all the other log roots */
 	struct btrfs_root *log_root_tree;
@@ -2092,6 +2132,27 @@ struct btrfs_ioctl_defrag_range_args {
  */
 #define BTRFS_BLOCK_GROUP_ITEM_KEY 192
 
+/*
+ * Every block group is represented in the free space tree by a free space info
+ * item, which stores some accounting information. It is keyed on
+ * (block_group_start, FREE_SPACE_INFO, block_group_length).
+ */
+#define BTRFS_FREE_SPACE_INFO_KEY 198
+
+/*
+ * A free space extent tracks an extent of space that is free in a block group.
+ * It is keyed on (start, FREE_SPACE_EXTENT, length).
+ */
+#define BTRFS_FREE_SPACE_EXTENT_KEY 199
+
+/*
+ * When a block group becomes very fragmented, we convert it to use bitmaps
+ * instead of extents. A free space bitmap is keyed on
+ * (start, FREE_SPACE_BITMAP, length); the corresponding item is a bitmap with
+ * (length / sectorsize) bits.
+ */
+#define BTRFS_FREE_SPACE_BITMAP_KEY 200
+
 #define BTRFS_DEV_EXTENT_KEY	204
 #define BTRFS_DEV_ITEM_KEY	216
 #define BTRFS_CHUNK_ITEM_KEY	228
@@ -2184,6 +2245,7 @@ struct btrfs_ioctl_defrag_range_args {
 #define BTRFS_MOUNT_RESCAN_UUID_TREE	(1 << 23)
 #define BTRFS_MOUNT_FRAGMENT_DATA	(1 << 24)
 #define BTRFS_MOUNT_FRAGMENT_METADATA	(1 << 25)
+#define BTRFS_MOUNT_FREE_SPACE_TREE	(1 << 26)
 
 #define BTRFS_DEFAULT_COMMIT_INTERVAL	(30)
 #define BTRFS_DEFAULT_MAX_INLINE	(8192)
@@ -2506,6 +2568,11 @@ BTRFS_SETGET_FUNCS(disk_block_group_flags,
 BTRFS_SETGET_STACK_FUNCS(block_group_flags,
 			struct btrfs_block_group_item, flags, 64);
 
+/* struct btrfs_free_space_info */
+BTRFS_SETGET_FUNCS(free_space_extent_count, struct btrfs_free_space_info,
+		   extent_count, 32);
+BTRFS_SETGET_FUNCS(free_space_flags, struct btrfs_free_space_info, flags, 32);
+
 /* struct btrfs_inode_ref */
 BTRFS_SETGET_FUNCS(inode_ref_name_len, struct btrfs_inode_ref, name_len, 16);
 BTRFS_SETGET_FUNCS(inode_ref_index, struct btrfs_inode_ref, index, 64);
@@ -3573,6 +3640,9 @@ void btrfs_end_write_no_snapshoting(struct btrfs_root *root);
 void check_system_chunk(struct btrfs_trans_handle *trans,
 			struct btrfs_root *root,
 			const u64 type);
+u64 add_new_free_space(struct btrfs_block_group_cache *block_group,
+		       struct btrfs_fs_info *info, u64 start, u64 end);
+
 /* ctree.c */
 int btrfs_bin_search(struct extent_buffer *eb, struct btrfs_key *key,
 		     int level, int *slot);
@@ -3737,6 +3807,7 @@ static inline void free_fs_info(struct btrfs_fs_info *fs_info)
 	kfree(fs_info->csum_root);
 	kfree(fs_info->quota_root);
 	kfree(fs_info->uuid_root);
+	kfree(fs_info->free_space_root);
 	kfree(fs_info->super_copy);
 	kfree(fs_info->super_for_commit);
 	security_free_mnt_opts(&fs_info->security_opts);
@@ -4247,6 +4318,30 @@ static inline void __btrfs_set_fs_incompat(struct btrfs_fs_info *fs_info,
 	}
 }
 
+#define btrfs_clear_fs_incompat(__fs_info, opt) \
+	__btrfs_clear_fs_incompat((__fs_info), BTRFS_FEATURE_INCOMPAT_##opt)
+
+static inline void __btrfs_clear_fs_incompat(struct btrfs_fs_info *fs_info,
+					     u64 flag)
+{
+	struct btrfs_super_block *disk_super;
+	u64 features;
+
+	disk_super = fs_info->super_copy;
+	features = btrfs_super_incompat_flags(disk_super);
+	if (features & flag) {
+		spin_lock(&fs_info->super_lock);
+		features = btrfs_super_incompat_flags(disk_super);
+		if (features & flag) {
+			features &= ~flag;
+			btrfs_set_super_incompat_flags(disk_super, features);
+			btrfs_info(fs_info, "clearing %llu feature flag",
+					 flag);
+		}
+		spin_unlock(&fs_info->super_lock);
+	}
+}
+
 #define btrfs_fs_incompat(fs_info, opt) \
 	__btrfs_fs_incompat((fs_info), BTRFS_FEATURE_INCOMPAT_##opt)
 
@@ -4257,6 +4352,64 @@ static inline int __btrfs_fs_incompat(struct btrfs_fs_info *fs_info, u64 flag)
 	return !!(btrfs_super_incompat_flags(disk_super) & flag);
 }
 
+#define btrfs_set_fs_compat_ro(__fs_info, opt) \
+	__btrfs_set_fs_compat_ro((__fs_info), BTRFS_FEATURE_COMPAT_RO_##opt)
+
+static inline void __btrfs_set_fs_compat_ro(struct btrfs_fs_info *fs_info,
+					    u64 flag)
+{
+	struct btrfs_super_block *disk_super;
+	u64 features;
+
+	disk_super = fs_info->super_copy;
+	features = btrfs_super_compat_ro_flags(disk_super);
+	if (!(features & flag)) {
+		spin_lock(&fs_info->super_lock);
+		features = btrfs_super_compat_ro_flags(disk_super);
+		if (!(features & flag)) {
+			features |= flag;
+			btrfs_set_super_compat_ro_flags(disk_super, features);
+			btrfs_info(fs_info, "setting %llu ro feature flag",
+				   flag);
+		}
+		spin_unlock(&fs_info->super_lock);
+	}
+}
+
+#define btrfs_clear_fs_compat_ro(__fs_info, opt) \
+	__btrfs_clear_fs_compat_ro((__fs_info), BTRFS_FEATURE_COMPAT_RO_##opt)
+
+static inline void __btrfs_clear_fs_compat_ro(struct btrfs_fs_info *fs_info,
+					      u64 flag)
+{
+	struct btrfs_super_block *disk_super;
+	u64 features;
+
+	disk_super = fs_info->super_copy;
+	features = btrfs_super_compat_ro_flags(disk_super);
+	if (features & flag) {
+		spin_lock(&fs_info->super_lock);
+		features = btrfs_super_compat_ro_flags(disk_super);
+		if (features & flag) {
+			features &= ~flag;
+			btrfs_set_super_compat_ro_flags(disk_super, features);
+			btrfs_info(fs_info, "clearing %llu ro feature flag",
+				   flag);
+		}
+		spin_unlock(&fs_info->super_lock);
+	}
+}
+
+#define btrfs_fs_compat_ro(fs_info, opt) \
+	__btrfs_fs_compat_ro((fs_info), BTRFS_FEATURE_COMPAT_RO_##opt)
+
+static inline int __btrfs_fs_compat_ro(struct btrfs_fs_info *fs_info, u64 flag)
+{
+	struct btrfs_super_block *disk_super;
+	disk_super = fs_info->super_copy;
+	return !!(btrfs_super_compat_ro_flags(disk_super) & flag);
+}
+
 /*
  * Call btrfs_abort_transaction as early as possible when an error condition is
  * detected, that way the exact line number is reported.
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 974be09e7556..52e623f59848 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -42,6 +42,7 @@
 #include "locking.h"
 #include "tree-log.h"
 #include "free-space-cache.h"
+#include "free-space-tree.h"
 #include "inode-map.h"
 #include "check-integrity.h"
 #include "rcu-string.h"
@@ -1650,6 +1651,9 @@ struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info,
 	if (location->objectid == BTRFS_UUID_TREE_OBJECTID)
 		return fs_info->uuid_root ? fs_info->uuid_root :
 					    ERR_PTR(-ENOENT);
+	if (location->objectid == BTRFS_FREE_SPACE_TREE_OBJECTID)
+		return fs_info->free_space_root ? fs_info->free_space_root :
+						  ERR_PTR(-ENOENT);
 again:
 	root = btrfs_lookup_fs_root(fs_info, location->objectid);
 	if (root) {
@@ -2148,6 +2152,7 @@ static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root)
 	free_root_extent_buffers(info->uuid_root);
 	if (chunk_root)
 		free_root_extent_buffers(info->chunk_root);
+	free_root_extent_buffers(info->free_space_root);
 }
 
 void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info)
@@ -2448,6 +2453,15 @@ static int btrfs_read_roots(struct btrfs_fs_info *fs_info,
 		fs_info->uuid_root = root;
 	}
 
+	if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
+		location.objectid = BTRFS_FREE_SPACE_TREE_OBJECTID;
+		root = btrfs_read_tree_root(tree_root, &location);
+		if (IS_ERR(root))
+			return PTR_ERR(root);
+		set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
+		fs_info->free_space_root = root;
+	}
+
 	return 0;
 }
 
@@ -3076,6 +3090,30 @@ retry_root_backup:
 
 	btrfs_qgroup_rescan_resume(fs_info);
 
+	if (btrfs_test_opt(tree_root, CLEAR_CACHE) &&
+	    btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
+		pr_info("BTRFS: clearing free space tree\n");
+		ret = btrfs_clear_free_space_tree(fs_info);
+		if (ret) {
+			pr_warn("BTRFS: failed to clear free space tree %d\n",
+				ret);
+			close_ctree(tree_root);
+			return ret;
+		}
+	}
+
+	if (btrfs_test_opt(tree_root, FREE_SPACE_TREE) &&
+	    !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
+		pr_info("BTRFS: creating free space tree\n");
+		ret = btrfs_create_free_space_tree(fs_info);
+		if (ret) {
+			pr_warn("BTRFS: failed to create free space tree %d\n",
+				ret);
+			close_ctree(tree_root);
+			return ret;
+		}
+	}
+
 	if (!fs_info->uuid_root) {
 		pr_info("BTRFS: creating UUID tree\n");
 		ret = btrfs_create_uuid_tree(fs_info);
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 4b89680a1923..8abb344e3dcb 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -33,6 +33,7 @@
 #include "raid56.h"
 #include "locking.h"
 #include "free-space-cache.h"
+#include "free-space-tree.h"
 #include "math.h"
 #include "sysfs.h"
 #include "qgroup.h"
@@ -357,8 +358,8 @@ static void fragment_free_space(struct btrfs_root *root,
  * we need to check the pinned_extents for any extents that can't be used yet
  * since their free space will be released as soon as the transaction commits.
  */
-static u64 add_new_free_space(struct btrfs_block_group_cache *block_group,
-			      struct btrfs_fs_info *info, u64 start, u64 end)
+u64 add_new_free_space(struct btrfs_block_group_cache *block_group,
+		       struct btrfs_fs_info *info, u64 start, u64 end)
 {
 	u64 extent_start, extent_end, size, total_added = 0;
 	int ret;
@@ -395,11 +396,10 @@ static u64 add_new_free_space(struct btrfs_block_group_cache *block_group,
 	return total_added;
 }
 
-static noinline void caching_thread(struct btrfs_work *work)
+static int load_extent_tree_free(struct btrfs_caching_control *caching_ctl)
 {
 	struct btrfs_block_group_cache *block_group;
 	struct btrfs_fs_info *fs_info;
-	struct btrfs_caching_control *caching_ctl;
 	struct btrfs_root *extent_root;
 	struct btrfs_path *path;
 	struct extent_buffer *leaf;
@@ -407,17 +407,16 @@ static noinline void caching_thread(struct btrfs_work *work)
 	u64 total_found = 0;
 	u64 last = 0;
 	u32 nritems;
-	int ret = -ENOMEM;
+	int ret;
 	bool wakeup = true;
 
-	caching_ctl = container_of(work, struct btrfs_caching_control, work);
 	block_group = caching_ctl->block_group;
 	fs_info = block_group->fs_info;
 	extent_root = fs_info->extent_root;
 
 	path = btrfs_alloc_path();
 	if (!path)
-		goto out;
+		return -ENOMEM;
 
 	last = max_t(u64, block_group->key.objectid, BTRFS_SUPER_INFO_OFFSET);
 
@@ -443,15 +442,11 @@ static noinline void caching_thread(struct btrfs_work *work)
 	key.objectid = last;
 	key.offset = 0;
 	key.type = BTRFS_EXTENT_ITEM_KEY;
-again:
-	mutex_lock(&caching_ctl->mutex);
-	/* need to make sure the commit_root doesn't disappear */
-	down_read(&fs_info->commit_root_sem);
 
 next:
 	ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0);
 	if (ret < 0)
-		goto err;
+		goto out;
 
 	leaf = path->nodes[0];
 	nritems = btrfs_header_nritems(leaf);
@@ -477,12 +472,14 @@ next:
 				up_read(&fs_info->commit_root_sem);
 				mutex_unlock(&caching_ctl->mutex);
 				cond_resched();
-				goto again;
+				mutex_lock(&caching_ctl->mutex);
+				down_read(&fs_info->commit_root_sem);
+				goto next;
 			}
 
 			ret = btrfs_next_leaf(extent_root, path);
 			if (ret < 0)
-				goto err;
+				goto out;
 			if (ret)
 				break;
 			leaf = path->nodes[0];
@@ -521,7 +518,7 @@ next:
 			else
 				last = key.objectid + key.offset;
 
-			if (total_found > (1024 * 1024 * 2)) {
+			if (total_found > CACHING_CTL_WAKE_UP) {
 				total_found = 0;
 				if (wakeup)
 					wake_up(&caching_ctl->wait);
@@ -534,9 +531,35 @@ next:
 	total_found += add_new_free_space(block_group, fs_info, last,
 					  block_group->key.objectid +
 					  block_group->key.offset);
+	caching_ctl->progress = (u64)-1;
+
+out:
+	btrfs_free_path(path);
+	return ret;
+}
+
+static noinline void caching_thread(struct btrfs_work *work)
+{
+	struct btrfs_block_group_cache *block_group;
+	struct btrfs_fs_info *fs_info;
+	struct btrfs_caching_control *caching_ctl;
+	int ret;
+
+	caching_ctl = container_of(work, struct btrfs_caching_control, work);
+	block_group = caching_ctl->block_group;
+	fs_info = block_group->fs_info;
+
+	mutex_lock(&caching_ctl->mutex);
+	down_read(&fs_info->commit_root_sem);
+
+	if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE))
+		ret = load_free_space_tree(caching_ctl);
+	else
+		ret = load_extent_tree_free(caching_ctl);
+
 	spin_lock(&block_group->lock);
 	block_group->caching_ctl = NULL;
-	block_group->cached = BTRFS_CACHE_FINISHED;
+	block_group->cached = ret ? BTRFS_CACHE_ERROR : BTRFS_CACHE_FINISHED;
 	spin_unlock(&block_group->lock);
 
 #ifdef CONFIG_BTRFS_DEBUG
@@ -555,20 +578,11 @@ next:
 #endif
 
 	caching_ctl->progress = (u64)-1;
-err:
-	btrfs_free_path(path);
-	up_read(&fs_info->commit_root_sem);
-
-	free_excluded_extents(extent_root, block_group);
 
+	up_read(&fs_info->commit_root_sem);
+	free_excluded_extents(fs_info->extent_root, block_group);
 	mutex_unlock(&caching_ctl->mutex);
-out:
-	if (ret) {
-		spin_lock(&block_group->lock);
-		block_group->caching_ctl = NULL;
-		block_group->cached = BTRFS_CACHE_ERROR;
-		spin_unlock(&block_group->lock);
-	}
+
 	wake_up(&caching_ctl->wait);
 
 	put_caching_control(caching_ctl);
@@ -680,8 +694,8 @@ static int cache_block_group(struct btrfs_block_group_cache *cache,
 		}
 	} else {
 		/*
-		 * We are not going to do the fast caching, set cached to the
-		 * appropriate value and wakeup any waiters.
+		 * We're either using the free space tree or no caching at all.
+		 * Set cached to the appropriate value and wakeup any waiters.
 		 */
 		spin_lock(&cache->lock);
 		if (load_cache_only) {
@@ -6661,6 +6675,13 @@ static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
 			}
 		}
 
+		ret = add_to_free_space_tree(trans, root->fs_info, bytenr,
+					     num_bytes);
+		if (ret) {
+			btrfs_abort_transaction(trans, extent_root, ret);
+			goto out;
+		}
+
 		ret = update_block_group(trans, root, bytenr, num_bytes, 0);
 		if (ret) {
 			btrfs_abort_transaction(trans, extent_root, ret);
@@ -7672,6 +7693,11 @@ static int alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
 	btrfs_mark_buffer_dirty(path->nodes[0]);
 	btrfs_free_path(path);
 
+	ret = remove_from_free_space_tree(trans, fs_info, ins->objectid,
+					  ins->offset);
+	if (ret)
+		return ret;
+
 	ret = update_block_group(trans, root, ins->objectid, ins->offset, 1);
 	if (ret) { /* -ENOENT, logic error */
 		btrfs_err(fs_info, "update block group failed for %llu %llu",
@@ -7752,6 +7778,11 @@ static int alloc_reserved_tree_block(struct btrfs_trans_handle *trans,
 	btrfs_mark_buffer_dirty(leaf);
 	btrfs_free_path(path);
 
+	ret = remove_from_free_space_tree(trans, fs_info, ins->objectid,
+					  num_bytes);
+	if (ret)
+		return ret;
+
 	ret = update_block_group(trans, root, ins->objectid, root->nodesize,
 				 1);
 	if (ret) { /* -ENOENT, logic error */
@@ -9656,6 +9687,8 @@ btrfs_create_block_group_cache(struct btrfs_root *root, u64 start, u64 size)
 	cache->full_stripe_len = btrfs_full_stripe_len(root,
 					       &root->fs_info->mapping_tree,
 					       start);
+	set_free_space_tree_thresholds(cache);
+
 	atomic_set(&cache->count, 1);
 	spin_lock_init(&cache->lock);
 	init_rwsem(&cache->data_rwsem);
@@ -9667,6 +9700,7 @@ btrfs_create_block_group_cache(struct btrfs_root *root, u64 start, u64 size)
 	INIT_LIST_HEAD(&cache->io_list);
 	btrfs_init_free_space_ctl(cache);
 	atomic_set(&cache->trimming, 0);
+	mutex_init(&cache->free_space_lock);
 
 	return cache;
 }
@@ -9877,6 +9911,8 @@ void btrfs_create_pending_block_groups(struct btrfs_trans_handle *trans,
 					       key.objectid, key.offset);
 		if (ret)
 			btrfs_abort_transaction(trans, extent_root, ret);
+		add_block_group_free_space(trans, root->fs_info, block_group);
+		/* already aborted the transaction if it failed. */
 next:
 		list_del_init(&block_group->bg_list);
 	}
@@ -9907,6 +9943,7 @@ int btrfs_make_block_group(struct btrfs_trans_handle *trans,
 	cache->flags = type;
 	cache->last_byte_to_unpin = (u64)-1;
 	cache->cached = BTRFS_CACHE_FINISHED;
+	cache->needs_free_space = 1;
 	ret = exclude_super_stripes(root, cache);
 	if (ret) {
 		/*
@@ -10277,6 +10314,10 @@ int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
 
 	unlock_chunks(root);
 
+	ret = remove_block_group_free_space(trans, root->fs_info, block_group);
+	if (ret)
+		goto out;
+
 	btrfs_put_block_group(block_group);
 	btrfs_put_block_group(block_group);
 
diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
index 9abe18763a7f..2b3f26326565 100644
--- a/fs/btrfs/extent_io.c
+++ b/fs/btrfs/extent_io.c
@@ -4797,24 +4797,14 @@ struct extent_buffer *btrfs_clone_extent_buffer(struct extent_buffer *src)
 	return new;
 }
 
-struct extent_buffer *alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
-						u64 start)
+struct extent_buffer *__alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
+						  u64 start, unsigned long len)
 {
 	struct extent_buffer *eb;
-	unsigned long len;
 	unsigned long num_pages;
 	unsigned long i;
 
-	if (!fs_info) {
-		/*
-		 * Called only from tests that don't always have a fs_info
-		 * available, but we know that nodesize is 4096
-		 */
-		len = 4096;
-	} else {
-		len = fs_info->tree_root->nodesize;
-	}
-	num_pages = num_extent_pages(0, len);
+	num_pages = num_extent_pages(start, len);
 
 	eb = __alloc_extent_buffer(fs_info, start, len);
 	if (!eb)
@@ -4837,6 +4827,24 @@ err:
 	return NULL;
 }
 
+struct extent_buffer *alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
+						u64 start)
+{
+	unsigned long len;
+
+	if (!fs_info) {
+		/*
+		 * Called only from tests that don't always have a fs_info
+		 * available, but we know that nodesize is 4096
+		 */
+		len = 4096;
+	} else {
+		len = fs_info->tree_root->nodesize;
+	}
+
+	return __alloc_dummy_extent_buffer(fs_info, start, len);
+}
+
 static void check_buffer_tree_ref(struct extent_buffer *eb)
 {
 	int refs;
@@ -5594,6 +5602,155 @@ void copy_extent_buffer(struct extent_buffer *dst, struct extent_buffer *src,
 	}
 }
 
+/*
+ * The extent buffer bitmap operations are done with byte granularity because
+ * bitmap items are not guaranteed to be aligned to a word and therefore a
+ * single word in a bitmap may straddle two pages in the extent buffer.
+ */
+#define BIT_BYTE(nr) ((nr) / BITS_PER_BYTE)
+#define BYTE_MASK ((1 << BITS_PER_BYTE) - 1)
+#define BITMAP_FIRST_BYTE_MASK(start) \
+	((BYTE_MASK << ((start) & (BITS_PER_BYTE - 1))) & BYTE_MASK)
+#define BITMAP_LAST_BYTE_MASK(nbits) \
+	(BYTE_MASK >> (-(nbits) & (BITS_PER_BYTE - 1)))
+
+/*
+ * eb_bitmap_offset() - calculate the page and offset of the byte containing the
+ * given bit number
+ * @eb: the extent buffer
+ * @start: offset of the bitmap item in the extent buffer
+ * @nr: bit number
+ * @page_index: return index of the page in the extent buffer that contains the
+ * given bit number
+ * @page_offset: return offset into the page given by page_index
+ *
+ * This helper hides the ugliness of finding the byte in an extent buffer which
+ * contains a given bit.
+ */
+static inline void eb_bitmap_offset(struct extent_buffer *eb,
+				    unsigned long start, unsigned long nr,
+				    unsigned long *page_index,
+				    size_t *page_offset)
+{
+	size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
+	size_t byte_offset = BIT_BYTE(nr);
+	size_t offset;
+
+	/*
+	 * The byte we want is the offset of the extent buffer + the offset of
+	 * the bitmap item in the extent buffer + the offset of the byte in the
+	 * bitmap item.
+	 */
+	offset = start_offset + start + byte_offset;
+
+	*page_index = offset >> PAGE_CACHE_SHIFT;
+	*page_offset = offset & (PAGE_CACHE_SIZE - 1);
+}
+
+/**
+ * extent_buffer_test_bit - determine whether a bit in a bitmap item is set
+ * @eb: the extent buffer
+ * @start: offset of the bitmap item in the extent buffer
+ * @nr: bit number to test
+ */
+int extent_buffer_test_bit(struct extent_buffer *eb, unsigned long start,
+			   unsigned long nr)
+{
+	char *kaddr;
+	struct page *page;
+	unsigned long i;
+	size_t offset;
+
+	eb_bitmap_offset(eb, start, nr, &i, &offset);
+	page = eb->pages[i];
+	WARN_ON(!PageUptodate(page));
+	kaddr = page_address(page);
+	return 1U & (kaddr[offset] >> (nr & (BITS_PER_BYTE - 1)));
+}
+
+/**
+ * extent_buffer_bitmap_set - set an area of a bitmap
+ * @eb: the extent buffer
+ * @start: offset of the bitmap item in the extent buffer
+ * @pos: bit number of the first bit
+ * @len: number of bits to set
+ */
+void extent_buffer_bitmap_set(struct extent_buffer *eb, unsigned long start,
+			      unsigned long pos, unsigned long len)
+{
+	char *kaddr;
+	struct page *page;
+	unsigned long i;
+	size_t offset;
+	const unsigned int size = pos + len;
+	int bits_to_set = BITS_PER_BYTE - (pos % BITS_PER_BYTE);
+	unsigned int mask_to_set = BITMAP_FIRST_BYTE_MASK(pos);
+
+	eb_bitmap_offset(eb, start, pos, &i, &offset);
+	page = eb->pages[i];
+	WARN_ON(!PageUptodate(page));
+	kaddr = page_address(page);
+
+	while (len >= bits_to_set) {
+		kaddr[offset] |= mask_to_set;
+		len -= bits_to_set;
+		bits_to_set = BITS_PER_BYTE;
+		mask_to_set = ~0U;
+		if (++offset >= PAGE_CACHE_SIZE && len > 0) {
+			offset = 0;
+			page = eb->pages[++i];
+			WARN_ON(!PageUptodate(page));
+			kaddr = page_address(page);
+		}
+	}
+	if (len) {
+		mask_to_set &= BITMAP_LAST_BYTE_MASK(size);
+		kaddr[offset] |= mask_to_set;
+	}
+}
+
+
+/**
+ * extent_buffer_bitmap_clear - clear an area of a bitmap
+ * @eb: the extent buffer
+ * @start: offset of the bitmap item in the extent buffer
+ * @pos: bit number of the first bit
+ * @len: number of bits to clear
+ */
+void extent_buffer_bitmap_clear(struct extent_buffer *eb, unsigned long start,
+				unsigned long pos, unsigned long len)
+{
+	char *kaddr;
+	struct page *page;
+	unsigned long i;
+	size_t offset;
+	const unsigned int size = pos + len;
+	int bits_to_clear = BITS_PER_BYTE - (pos % BITS_PER_BYTE);
+	unsigned int mask_to_clear = BITMAP_FIRST_BYTE_MASK(pos);
+
+	eb_bitmap_offset(eb, start, pos, &i, &offset);
+	page = eb->pages[i];
+	WARN_ON(!PageUptodate(page));
+	kaddr = page_address(page);
+
+	while (len >= bits_to_clear) {
+		kaddr[offset] &= ~mask_to_clear;
+		len -= bits_to_clear;
+		bits_to_clear = BITS_PER_BYTE;
+		mask_to_clear = ~0U;
+		if (++offset >= PAGE_CACHE_SIZE && len > 0) {
+			offset = 0;
+			page = eb->pages[++i];
+			WARN_ON(!PageUptodate(page));
+			kaddr = page_address(page);
+		}
+	}
+	if (len) {
+		mask_to_clear &= BITMAP_LAST_BYTE_MASK(size);
+		kaddr[offset] &= ~mask_to_clear;
+	}
+}
+
 static inline bool areas_overlap(unsigned long src, unsigned long dst, unsigned long len)
 {
 	unsigned long distance = (src > dst) ? src - dst : dst - src;
diff --git a/fs/btrfs/extent_io.h b/fs/btrfs/extent_io.h
index f4c1ae11855f..350c8b0a8582 100644
--- a/fs/btrfs/extent_io.h
+++ b/fs/btrfs/extent_io.h
@@ -282,8 +282,10 @@ void set_page_extent_mapped(struct page *page);
 
 struct extent_buffer *alloc_extent_buffer(struct btrfs_fs_info *fs_info,
 					  u64 start);
+struct extent_buffer *__alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
+						  u64 start, unsigned long len);
 struct extent_buffer *alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
-		u64 start);
+						u64 start);
 struct extent_buffer *btrfs_clone_extent_buffer(struct extent_buffer *src);
 struct extent_buffer *find_extent_buffer(struct btrfs_fs_info *fs_info,
 					 u64 start);
@@ -328,6 +330,12 @@ void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
 			   unsigned long src_offset, unsigned long len);
 void memset_extent_buffer(struct extent_buffer *eb, char c,
 			  unsigned long start, unsigned long len);
+int extent_buffer_test_bit(struct extent_buffer *eb, unsigned long start,
+			   unsigned long pos);
+void extent_buffer_bitmap_set(struct extent_buffer *eb, unsigned long start,
+			      unsigned long pos, unsigned long len);
+void extent_buffer_bitmap_clear(struct extent_buffer *eb, unsigned long start,
+				unsigned long pos, unsigned long len);
 void clear_extent_buffer_dirty(struct extent_buffer *eb);
 int set_extent_buffer_dirty(struct extent_buffer *eb);
 int set_extent_buffer_uptodate(struct extent_buffer *eb);
diff --git a/fs/btrfs/free-space-tree.c b/fs/btrfs/free-space-tree.c
new file mode 100644
index 000000000000..cbe36dd0d97b
--- /dev/null
+++ b/fs/btrfs/free-space-tree.c
@@ -0,0 +1,1584 @@
+/*
+ * Copyright (C) 2015 Facebook.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/kernel.h>
+#include <linux/vmalloc.h>
+#include "ctree.h"
+#include "disk-io.h"
+#include "locking.h"
+#include "free-space-tree.h"
+#include "transaction.h"
+
+static int __add_block_group_free_space(struct btrfs_trans_handle *trans,
+					struct btrfs_fs_info *fs_info,
+					struct btrfs_block_group_cache *block_group,
+					struct btrfs_path *path);
+
+void set_free_space_tree_thresholds(struct btrfs_block_group_cache *cache)
+{
+	u32 bitmap_range;
+	size_t bitmap_size;
+	u64 num_bitmaps, total_bitmap_size;
+
+	/*
+	 * We convert to bitmaps when the disk space required for using extents
+	 * exceeds that required for using bitmaps.
+	 */
+	bitmap_range = cache->sectorsize * BTRFS_FREE_SPACE_BITMAP_BITS;
+	num_bitmaps = div_u64(cache->key.offset + bitmap_range - 1,
+			      bitmap_range);
+	bitmap_size = sizeof(struct btrfs_item) + BTRFS_FREE_SPACE_BITMAP_SIZE;
+	total_bitmap_size = num_bitmaps * bitmap_size;
+	cache->bitmap_high_thresh = div_u64(total_bitmap_size,
+					    sizeof(struct btrfs_item));
+
+	/*
+	 * We allow for a small buffer between the high threshold and low
+	 * threshold to avoid thrashing back and forth between the two formats.
+	 */
+	if (cache->bitmap_high_thresh > 100)
+		cache->bitmap_low_thresh = cache->bitmap_high_thresh - 100;
+	else
+		cache->bitmap_low_thresh = 0;
+}
+
+static int add_new_free_space_info(struct btrfs_trans_handle *trans,
+				   struct btrfs_fs_info *fs_info,
+				   struct btrfs_block_group_cache *block_group,
+				   struct btrfs_path *path)
+{
+	struct btrfs_root *root = fs_info->free_space_root;
+	struct btrfs_free_space_info *info;
+	struct btrfs_key key;
+	struct extent_buffer *leaf;
+	int ret;
+
+	key.objectid = block_group->key.objectid;
+	key.type = BTRFS_FREE_SPACE_INFO_KEY;
+	key.offset = block_group->key.offset;
+
+	ret = btrfs_insert_empty_item(trans, root, path, &key, sizeof(*info));
+	if (ret)
+		goto out;
+
+	leaf = path->nodes[0];
+	info = btrfs_item_ptr(leaf, path->slots[0],
+			      struct btrfs_free_space_info);
+	btrfs_set_free_space_extent_count(leaf, info, 0);