Merge git://git.kernel.org/pub/scm/linux/kernel/git/mason/btrfs-unstable

* git://git.kernel.org/pub/scm/linux/kernel/git/mason/btrfs-unstable: (23 commits)
  Btrfs: fix extent_buffer leak during tree log replay
  Btrfs: fix oops when btrfs_inherit_iflags called with a NULL dir
  Btrfs: fix -o nodatasum printk spelling
  Btrfs: check duplicate backrefs for both data and metadata
  Btrfs: init worker struct fields before kthread-run
  Btrfs: pin buffers during write_dev_supers
  Btrfs: avoid races between super writeout and device list updates
  Fix btrfs when ACLs are configured out
  Btrfs: fdatasync should skip metadata writeout
  Btrfs: remove crc32c.h and use libcrc32c directly.
  Btrfs: implement FS_IOC_GETFLAGS/SETFLAGS/GETVERSION
  Btrfs: autodetect SSD devices
  Btrfs: add mount -o ssd_spread to spread allocations out
  Btrfs: avoid allocation clusters that are too spread out
  Btrfs: Add mount -o nossd
  Btrfs: avoid IO stalls behind congested devices in a multi-device FS
  Btrfs: don't allow WRITE_SYNC bios to starve out regular writes
  Btrfs: fix metadata dirty throttling limits
  Btrfs: reduce mount -o ssd CPU usage
  Btrfs: balance btree more often
  ...

fs/btrfs/Makefile

@@ -6,5 +6,5 @@ btrfs-y += super.o ctree.o extent-tree.o print-tree.o root-tree.o dir-item.o \
 	   transaction.o inode.o file.o tree-defrag.o \
 	   extent_map.o sysfs.o struct-funcs.o xattr.o ordered-data.o \
 	   extent_io.o volumes.o async-thread.o ioctl.o locking.o orphan.o \
-	   ref-cache.o export.o tree-log.o acl.o free-space-cache.o zlib.o \
-	   compression.o delayed-ref.o
+	   export.o tree-log.o acl.o free-space-cache.o zlib.o \
+	   compression.o delayed-ref.o relocation.o

fs/btrfs/acl.c

@@ -351,9 +351,4 @@ int btrfs_init_acl(struct inode *inode, struct inode *dir)
 	return 0;
 }
 
-int btrfs_check_acl(struct inode *inode, int mask)
-{
-	return 0;
-}
-
 #endif /* CONFIG_FS_POSIX_ACL */

fs/btrfs/async-thread.c

@@ -294,10 +294,10 @@ int btrfs_start_workers(struct btrfs_workers *workers, int num_workers)
 		INIT_LIST_HEAD(&worker->worker_list);
 		spin_lock_init(&worker->lock);
 		atomic_set(&worker->num_pending, 0);
+		worker->workers = workers;
 		worker->task = kthread_run(worker_loop, worker,
 					   "btrfs-%s-%d", workers->name,
 					   workers->num_workers + i);
-		worker->workers = workers;
 		if (IS_ERR(worker->task)) {
 			kfree(worker);
 			ret = PTR_ERR(worker->task);

fs/btrfs/btrfs_inode.h

@@ -72,6 +72,9 @@ struct btrfs_inode {
 	 */
 	struct list_head ordered_operations;
 
+	/* node for the red-black tree that links inodes in subvolume root */
+	struct rb_node rb_node;
+
 	/* the space_info for where this inode's data allocations are done */
 	struct btrfs_space_info *space_info;
 
@@ -154,5 +157,4 @@ static inline void btrfs_i_size_write(struct inode *inode, u64 size)
 	BTRFS_I(inode)->disk_i_size = size;
 }
 
-
 #endif

fs/btrfs/compression.c

@@ -123,7 +123,7 @@ static int check_compressed_csum(struct inode *inode,
 	u32 csum;
 	u32 *cb_sum = &cb->sums;
 
-	if (btrfs_test_flag(inode, NODATASUM))
+	if (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)
 		return 0;
 
 	for (i = 0; i < cb->nr_pages; i++) {
@@ -670,7 +670,7 @@ int btrfs_submit_compressed_read(struct inode *inode, struct bio *bio,
 			 */
 			atomic_inc(&cb->pending_bios);
 
-			if (!btrfs_test_flag(inode, NODATASUM)) {
+			if (!(BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)) {
 				btrfs_lookup_bio_sums(root, inode, comp_bio,
 						      sums);
 			}
@@ -697,7 +697,7 @@ int btrfs_submit_compressed_read(struct inode *inode, struct bio *bio,
 	ret = btrfs_bio_wq_end_io(root->fs_info, comp_bio, 0);
 	BUG_ON(ret);
 
-	if (!btrfs_test_flag(inode, NODATASUM))
+	if (!(BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
 		btrfs_lookup_bio_sums(root, inode, comp_bio, sums);
 
 	ret = btrfs_map_bio(root, READ, comp_bio, mirror_num, 0);

fs/btrfs/crc32c.h

@@ -1,29 +0,0 @@
-/*
- * Copyright (C) 2008 Oracle.  All rights reserved.
- *
- * This program is free software; you can redistribute it and/or
- * modify it under the terms of the GNU General Public
- * License v2 as published by the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful,
- * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * General Public License for more details.
- *
- * You should have received a copy of the GNU General Public
- * License along with this program; if not, write to the
- * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
- * Boston, MA 021110-1307, USA.
- */
-
-#ifndef __BTRFS_CRC32C__
-#define __BTRFS_CRC32C__
-#include <linux/crc32c.h>
-
-/*
- * this file used to do more for selecting the HW version of crc32c,
- * perhaps it will one day again soon.
- */
-#define btrfs_crc32c(seed, data, length) crc32c(seed, data, length)
-#endif
-

fs/btrfs/ctree.h

@@ -45,6 +45,8 @@ struct btrfs_ordered_sum;
 
 #define BTRFS_MAX_LEVEL 8
 
+#define BTRFS_COMPAT_EXTENT_TREE_V0
+
 /*
  * files bigger than this get some pre-flushing when they are added
  * to the ordered operations list.  That way we limit the total
@@ -267,7 +269,18 @@ static inline unsigned long btrfs_chunk_item_size(int num_stripes)
 }
 
 #define BTRFS_FSID_SIZE 16
-#define BTRFS_HEADER_FLAG_WRITTEN (1 << 0)
+#define BTRFS_HEADER_FLAG_WRITTEN	(1ULL << 0)
+#define BTRFS_HEADER_FLAG_RELOC		(1ULL << 1)
+#define BTRFS_SUPER_FLAG_SEEDING	(1ULL << 32)
+#define BTRFS_SUPER_FLAG_METADUMP	(1ULL << 33)
+
+#define BTRFS_BACKREF_REV_MAX		256
+#define BTRFS_BACKREF_REV_SHIFT		56
+#define BTRFS_BACKREF_REV_MASK		(((u64)BTRFS_BACKREF_REV_MAX - 1) << \
+					 BTRFS_BACKREF_REV_SHIFT)
+
+#define BTRFS_OLD_BACKREF_REV		0
+#define BTRFS_MIXED_BACKREF_REV		1
 
 /*
  * every tree block (leaf or node) starts with this header.
@@ -296,7 +309,6 @@ struct btrfs_header {
 					sizeof(struct btrfs_item) - \
 					sizeof(struct btrfs_file_extent_item))
 
-#define BTRFS_SUPER_FLAG_SEEDING (1ULL << 32)
 
 /*
  * this is a very generous portion of the super block, giving us
@@ -355,9 +367,12 @@ 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_SUPP	0x0
-#define BTRFS_FEATURE_COMPAT_RO_SUPP	0x0
-#define BTRFS_FEATURE_INCOMPAT_SUPP	0x0
+#define BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF	(1ULL << 0)
+
+#define BTRFS_FEATURE_COMPAT_SUPP		0ULL
+#define BTRFS_FEATURE_COMPAT_RO_SUPP		0ULL
+#define BTRFS_FEATURE_INCOMPAT_SUPP		\
+	BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF
 
 /*
  * A leaf is full of items. offset and size tell us where to find
@@ -421,23 +436,65 @@ struct btrfs_path {
 	unsigned int keep_locks:1;
 	unsigned int skip_locking:1;
 	unsigned int leave_spinning:1;
+	unsigned int search_commit_root:1;
 };
 
 /*
  * items in the extent btree are used to record the objectid of the
  * owner of the block and the number of references
  */
+
 struct btrfs_extent_item {
+	__le64 refs;
+	__le64 generation;
+	__le64 flags;
+} __attribute__ ((__packed__));
+
+struct btrfs_extent_item_v0 {
 	__le32 refs;
 } __attribute__ ((__packed__));
 
-struct btrfs_extent_ref {
+#define BTRFS_MAX_EXTENT_ITEM_SIZE(r) ((BTRFS_LEAF_DATA_SIZE(r) >> 4) - \
+					sizeof(struct btrfs_item))
+
+#define BTRFS_EXTENT_FLAG_DATA		(1ULL << 0)
+#define BTRFS_EXTENT_FLAG_TREE_BLOCK	(1ULL << 1)
+
+/* following flags only apply to tree blocks */
+
+/* use full backrefs for extent pointers in the block */
+#define BTRFS_BLOCK_FLAG_FULL_BACKREF	(1ULL << 8)
+
+struct btrfs_tree_block_info {
+	struct btrfs_disk_key key;
+	u8 level;
+} __attribute__ ((__packed__));
+
+struct btrfs_extent_data_ref {
+	__le64 root;
+	__le64 objectid;
+	__le64 offset;
+	__le32 count;
+} __attribute__ ((__packed__));
+
+struct btrfs_shared_data_ref {
+	__le32 count;
+} __attribute__ ((__packed__));
+
+struct btrfs_extent_inline_ref {
+	u8 type;
+	u64 offset;
+} __attribute__ ((__packed__));
+
+/* old style backrefs item */
+struct btrfs_extent_ref_v0 {
 	__le64 root;
 	__le64 generation;
 	__le64 objectid;
-	__le32 num_refs;
+	__le32 count;
 } __attribute__ ((__packed__));
 
+
 /* dev extents record free space on individual devices.  The owner
  * field points back to the chunk allocation mapping tree that allocated
  * the extent.  The chunk tree uuid field is a way to double check the owner
@@ -695,12 +752,7 @@ struct btrfs_block_group_cache {
 	struct list_head cluster_list;
 };
 
-struct btrfs_leaf_ref_tree {
-	struct rb_root root;
-	struct list_head list;
-	spinlock_t lock;
-};
-
+struct reloc_control;
 struct btrfs_device;
 struct btrfs_fs_devices;
 struct btrfs_fs_info {
@@ -831,18 +883,11 @@ struct btrfs_fs_info {
 	struct task_struct *cleaner_kthread;
 	int thread_pool_size;
 
-	/* tree relocation relocated fields */
-	struct list_head dead_reloc_roots;
-	struct btrfs_leaf_ref_tree reloc_ref_tree;
-	struct btrfs_leaf_ref_tree shared_ref_tree;
-
 	struct kobject super_kobj;
 	struct completion kobj_unregister;
 	int do_barriers;
 	int closing;
 	int log_root_recovering;
-	atomic_t throttles;
-	atomic_t throttle_gen;
 
 	u64 total_pinned;
 
@@ -861,6 +906,8 @@ struct btrfs_fs_info {
 	 */
 	struct list_head space_info;
 
+	struct reloc_control *reloc_ctl;
+
 	spinlock_t delalloc_lock;
 	spinlock_t new_trans_lock;
 	u64 delalloc_bytes;
@@ -891,7 +938,6 @@ struct btrfs_fs_info {
  * in ram representation of the tree.  extent_root is used for all allocations
  * and for the extent tree extent_root root.
  */
-struct btrfs_dirty_root;
 struct btrfs_root {
 	struct extent_buffer *node;
 
@@ -899,9 +945,6 @@ struct btrfs_root {
 	spinlock_t node_lock;
 
 	struct extent_buffer *commit_root;
-	struct btrfs_leaf_ref_tree *ref_tree;
-	struct btrfs_leaf_ref_tree ref_tree_struct;
-	struct btrfs_dirty_root *dirty_root;
 	struct btrfs_root *log_root;
 	struct btrfs_root *reloc_root;
 
@@ -952,10 +995,15 @@ struct btrfs_root {
 	/* the dirty list is only used by non-reference counted roots */
 	struct list_head dirty_list;
 
+	struct list_head root_list;
+
 	spinlock_t list_lock;
-	struct list_head dead_list;
 	struct list_head orphan_list;
 
+	spinlock_t inode_lock;
+	/* red-black tree that keeps track of in-memory inodes */
+	struct rb_root inode_tree;
+
 	/*
 	 * right now this just gets used so that a root has its own devid
 	 * for stat.  It may be used for more later
@@ -1017,7 +1065,16 @@ struct btrfs_root {
  * are used, and how many references there are to each block
  */
 #define BTRFS_EXTENT_ITEM_KEY	168
-#define BTRFS_EXTENT_REF_KEY	180
+
+#define BTRFS_TREE_BLOCK_REF_KEY	176
+
+#define BTRFS_EXTENT_DATA_REF_KEY	178
+
+#define BTRFS_EXTENT_REF_V0_KEY		180
+
+#define BTRFS_SHARED_BLOCK_REF_KEY	182
+
+#define BTRFS_SHARED_DATA_REF_KEY	184
 
 /*
  * block groups give us hints into the extent allocation trees.  Which
@@ -1043,6 +1100,8 @@ struct btrfs_root {
 #define BTRFS_MOUNT_COMPRESS		(1 << 5)
 #define BTRFS_MOUNT_NOTREELOG           (1 << 6)
 #define BTRFS_MOUNT_FLUSHONCOMMIT       (1 << 7)
+#define BTRFS_MOUNT_SSD_SPREAD		(1 << 8)
+#define BTRFS_MOUNT_NOSSD		(1 << 9)
 
 #define btrfs_clear_opt(o, opt)		((o) &= ~BTRFS_MOUNT_##opt)
 #define btrfs_set_opt(o, opt)		((o) |= BTRFS_MOUNT_##opt)
@@ -1056,12 +1115,14 @@ struct btrfs_root {
 #define BTRFS_INODE_READONLY		(1 << 2)
 #define BTRFS_INODE_NOCOMPRESS		(1 << 3)
 #define BTRFS_INODE_PREALLOC		(1 << 4)
-#define btrfs_clear_flag(inode, flag)	(BTRFS_I(inode)->flags &= \
-					 ~BTRFS_INODE_##flag)
-#define btrfs_set_flag(inode, flag)	(BTRFS_I(inode)->flags |= \
-					 BTRFS_INODE_##flag)
-#define btrfs_test_flag(inode, flag)	(BTRFS_I(inode)->flags & \
-					 BTRFS_INODE_##flag)
+#define BTRFS_INODE_SYNC		(1 << 5)
+#define BTRFS_INODE_IMMUTABLE		(1 << 6)
+#define BTRFS_INODE_APPEND		(1 << 7)
+#define BTRFS_INODE_NODUMP		(1 << 8)
+#define BTRFS_INODE_NOATIME		(1 << 9)
+#define BTRFS_INODE_DIRSYNC		(1 << 10)
+
+
 /* some macros to generate set/get funcs for the struct fields.  This
  * assumes there is a lefoo_to_cpu for every type, so lets make a simple
  * one for u8:
@@ -1317,24 +1378,67 @@ static inline u8 *btrfs_dev_extent_chunk_tree_uuid(struct btrfs_dev_extent *dev)
 	return (u8 *)((unsigned long)dev + ptr);
 }
 
-/* struct btrfs_extent_ref */
-BTRFS_SETGET_FUNCS(ref_root, struct btrfs_extent_ref, root, 64);
-BTRFS_SETGET_FUNCS(ref_generation, struct btrfs_extent_ref, generation, 64);
-BTRFS_SETGET_FUNCS(ref_objectid, struct btrfs_extent_ref, objectid, 64);
-BTRFS_SETGET_FUNCS(ref_num_refs, struct btrfs_extent_ref, num_refs, 32);
+BTRFS_SETGET_FUNCS(extent_refs, struct btrfs_extent_item, refs, 64);
+BTRFS_SETGET_FUNCS(extent_generation, struct btrfs_extent_item,
+		   generation, 64);
+BTRFS_SETGET_FUNCS(extent_flags, struct btrfs_extent_item, flags, 64);
 
-BTRFS_SETGET_STACK_FUNCS(stack_ref_root, struct btrfs_extent_ref, root, 64);
-BTRFS_SETGET_STACK_FUNCS(stack_ref_generation, struct btrfs_extent_ref,
-			 generation, 64);
-BTRFS_SETGET_STACK_FUNCS(stack_ref_objectid, struct btrfs_extent_ref,
-			 objectid, 64);
-BTRFS_SETGET_STACK_FUNCS(stack_ref_num_refs, struct btrfs_extent_ref,
-			 num_refs, 32);
+BTRFS_SETGET_FUNCS(extent_refs_v0, struct btrfs_extent_item_v0, refs, 32);
+
+
+BTRFS_SETGET_FUNCS(tree_block_level, struct btrfs_tree_block_info, level, 8);
+
+static inline void btrfs_tree_block_key(struct extent_buffer *eb,
+					struct btrfs_tree_block_info *item,
+					struct btrfs_disk_key *key)
+{
+	read_eb_member(eb, item, struct btrfs_tree_block_info, key, key);
+}
+
+static inline void btrfs_set_tree_block_key(struct extent_buffer *eb,
+					    struct btrfs_tree_block_info *item,
+					    struct btrfs_disk_key *key)
+{
+	write_eb_member(eb, item, struct btrfs_tree_block_info, key, key);
+}
+
+BTRFS_SETGET_FUNCS(extent_data_ref_root, struct btrfs_extent_data_ref,
+		   root, 64);
+BTRFS_SETGET_FUNCS(extent_data_ref_objectid, struct btrfs_extent_data_ref,
+		   objectid, 64);
+BTRFS_SETGET_FUNCS(extent_data_ref_offset, struct btrfs_extent_data_ref,
+		   offset, 64);
+BTRFS_SETGET_FUNCS(extent_data_ref_count, struct btrfs_extent_data_ref,
+		   count, 32);
+
+BTRFS_SETGET_FUNCS(shared_data_ref_count, struct btrfs_shared_data_ref,
+		   count, 32);
 
-/* struct btrfs_extent_item */
-BTRFS_SETGET_FUNCS(extent_refs, struct btrfs_extent_item, refs, 32);
-BTRFS_SETGET_STACK_FUNCS(stack_extent_refs, struct btrfs_extent_item,
-			 refs, 32);
+BTRFS_SETGET_FUNCS(extent_inline_ref_type, struct btrfs_extent_inline_ref,
+		   type, 8);
+BTRFS_SETGET_FUNCS(extent_inline_ref_offset, struct btrfs_extent_inline_ref,
+		   offset, 64);
+
+static inline u32 btrfs_extent_inline_ref_size(int type)
+{
+	if (type == BTRFS_TREE_BLOCK_REF_KEY ||
+	    type == BTRFS_SHARED_BLOCK_REF_KEY)
+		return sizeof(struct btrfs_extent_inline_ref);
+	if (type == BTRFS_SHARED_DATA_REF_KEY)
+		return sizeof(struct btrfs_shared_data_ref) +
+		       sizeof(struct btrfs_extent_inline_ref);
+	if (type == BTRFS_EXTENT_DATA_REF_KEY)
+		return sizeof(struct btrfs_extent_data_ref) +
+		       offsetof(struct btrfs_extent_inline_ref, offset);
+	BUG();
+	return 0;
+}
+
+BTRFS_SETGET_FUNCS(ref_root_v0, struct btrfs_extent_ref_v0, root, 64);
+BTRFS_SETGET_FUNCS(ref_generation_v0, struct btrfs_extent_ref_v0,
+		   generation, 64);
+BTRFS_SETGET_FUNCS(ref_objectid_v0, struct btrfs_extent_ref_v0, objectid, 64);
+BTRFS_SETGET_FUNCS(ref_count_v0, struct btrfs_extent_ref_v0, count, 32);
 
 /* struct btrfs_node */
 BTRFS_SETGET_FUNCS(key_blockptr, struct btrfs_key_ptr, blockptr, 64);
@@ -1558,6 +1662,21 @@ static inline int btrfs_clear_header_flag(struct extent_buffer *eb, u64 flag)
 	return (flags & flag) == flag;
 }
 
+static inline int btrfs_header_backref_rev(struct extent_buffer *eb)
+{
+	u64 flags = btrfs_header_flags(eb);
+	return flags >> BTRFS_BACKREF_REV_SHIFT;
+}
+
+static inline void btrfs_set_header_backref_rev(struct extent_buffer *eb,
+						int rev)
+{
+	u64 flags = btrfs_header_flags(eb);
+	flags &= ~BTRFS_BACKREF_REV_MASK;
+	flags |= (u64)rev << BTRFS_BACKREF_REV_SHIFT;
+	btrfs_set_header_flags(eb, flags);
+}
+
 static inline u8 *btrfs_header_fsid(struct extent_buffer *eb)
 {
 	unsigned long ptr = offsetof(struct btrfs_header, fsid);
@@ -1790,39 +1909,32 @@ int btrfs_update_pinned_extents(struct btrfs_root *root,
 int btrfs_drop_leaf_ref(struct btrfs_trans_handle *trans,
 			struct btrfs_root *root, struct extent_buffer *leaf);
 int btrfs_cross_ref_exist(struct btrfs_trans_handle *trans,
-			  struct btrfs_root *root, u64 objectid, u64 bytenr);
+			  struct btrfs_root *root,
+			  u64 objectid, u64 offset, u64 bytenr);
 int btrfs_copy_pinned(struct btrfs_root *root, struct extent_io_tree *copy);
 struct btrfs_block_group_cache *btrfs_lookup_block_group(
 						 struct btrfs_fs_info *info,
 						 u64 bytenr);
+void btrfs_put_block_group(struct btrfs_block_group_cache *cache);
 u64 btrfs_find_block_group(struct btrfs_root *root,
 			   u64 search_start, u64 search_hint, int owner);
 struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
-					     struct btrfs_root *root,
-					     u32 blocksize, u64 parent,
-					     u64 root_objectid,
-					     u64 ref_generation,
-					     int level,
-					     u64 hint,
-					     u64 empty_size);
+					struct btrfs_root *root, u32 blocksize,
+					u64 parent, u64 root_objectid,
+					struct btrfs_disk_key *key, int level,
+					u64 hint, u64 empty_size);
 struct extent_buffer *btrfs_init_new_buffer(struct btrfs_trans_handle *trans,
 					    struct btrfs_root *root,
 					    u64 bytenr, u32 blocksize,
 					    int level);
-int btrfs_alloc_extent(struct btrfs_trans_handle *trans,
-		       struct btrfs_root *root,
-		       u64 num_bytes, u64 parent, u64 min_bytes,
-		       u64 root_objectid, u64 ref_generation,
-		       u64 owner, u64 empty_size, u64 hint_byte,
-		       u64 search_end, struct btrfs_key *ins, u64 data);
-int btrfs_alloc_reserved_extent(struct btrfs_trans_handle *trans,
-				struct btrfs_root *root, u64 parent,
-				u64 root_objectid, u64 ref_generation,
-				u64 owner, struct btrfs_key *ins);
-int btrfs_alloc_logged_extent(struct btrfs_trans_handle *trans,
-				struct btrfs_root *root, u64 parent,
-				u64 root_objectid, u64 ref_generation,
-				u64 owner, struct btrfs_key *ins);
+int btrfs_alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
+				     struct btrfs_root *root,
+				     u64 root_objectid, u64 owner,
+				     u64 offset, struct btrfs_key *ins);
+int btrfs_alloc_logged_file_extent(struct btrfs_trans_handle *trans,
+				   struct btrfs_root *root,
+				   u64 root_objectid, u64 owner, u64 offset,
+				   struct btrfs_key *ins);
 int btrfs_reserve_extent(struct btrfs_trans_handle *trans,
 				  struct btrfs_root *root,
 				  u64 num_bytes, u64 min_alloc_size,
@@ -1830,18 +1942,18 @@ int btrfs_reserve_extent(struct btrfs_trans_handle *trans,
 				  u64 search_end, struct btrfs_key *ins,
 				  u64 data);
 int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
-		  struct extent_buffer *orig_buf, struct extent_buffer *buf,
-		  u32 *nr_extents);
-int btrfs_cache_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
-		    struct extent_buffer *buf, u32 nr_extents);
-int btrfs_update_ref(struct btrfs_trans_handle *trans,
-		     struct btrfs_root *root, struct extent_buffer *orig_buf,
-		     struct extent_buffer *buf, int start_slot, int nr);
+		  struct extent_buffer *buf, int full_backref);
+int btrfs_dec_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+		  struct extent_buffer *buf, int full_backref);
+int btrfs_set_disk_extent_flags(struct btrfs_trans_handle *trans,
+				struct btrfs_root *root,
+				u64 bytenr, u64 num_bytes, u64 flags,
+				int is_data);
 int btrfs_free_extent(struct btrfs_trans_handle *trans,
 		      struct btrfs_root *root,
 		      u64 bytenr, u64 num_bytes, u64 parent,
-		      u64 root_objectid, u64 ref_generation,
-		      u64 owner_objectid, int pin);
+		      u64 root_objectid, u64 owner, u64 offset);
+
 int btrfs_free_reserved_extent(struct btrfs_root *root, u64 start, u64 len);
 int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans,
 			       struct btrfs_root *root,
@@ -1849,13 +1961,8 @@ int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans,
 int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
 			 struct btrfs_root *root,
 			 u64 bytenr, u64 num_bytes, u64 parent,
-			 u64 root_objectid, u64 ref_generation,
-			 u64 owner_objectid);
-int btrfs_update_extent_ref(struct btrfs_trans_handle *trans,
-			    struct btrfs_root *root, u64 bytenr, u64 num_bytes,
-			    u64 orig_parent, u64 parent,
-			    u64 root_objectid, u64 ref_generation,
-			    u64 owner_objectid);
+			 u64 root_objectid, u64 owner, u64 offset);
+
 int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans,
 				    struct btrfs_root *root);
 int btrfs_extent_readonly(struct btrfs_root *root, u64 bytenr);
@@ -1867,16 +1974,9 @@ int btrfs_make_block_group(struct btrfs_trans_handle *trans,
 			   u64 size);
 int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
 			     struct btrfs_root *root, u64 group_start);
-int btrfs_relocate_block_group(struct btrfs_root *root, u64 group_start);
-int btrfs_free_reloc_root(struct btrfs_trans_handle *trans,
-			  struct btrfs_root *root);
-int btrfs_drop_dead_reloc_roots(struct btrfs_root *root);
-int btrfs_reloc_tree_cache_ref(struct btrfs_trans_handle *trans,
-			       struct btrfs_root *root,
-			       struct extent_buffer *buf, u64 orig_start);
-int btrfs_add_dead_reloc_root(struct btrfs_root *root);
-int btrfs_cleanup_reloc_trees(struct btrfs_root *root);
-int btrfs_reloc_clone_csums(struct inode *inode, u64 file_pos, u64 len);
+int btrfs_prepare_block_group_relocation(struct btrfs_root *root,
+				struct btrfs_block_group_cache *group);
+
 u64 btrfs_reduce_alloc_profile(struct btrfs_root *root, u64 flags);
 void btrfs_set_inode_space_info(struct btrfs_root *root, struct inode *ionde);
 void btrfs_clear_space_info_full(struct btrfs_fs_info *info);
@@ -1891,13 +1991,12 @@ void btrfs_delalloc_reserve_space(struct btrfs_root *root, struct inode *inode,
 void btrfs_delalloc_free_space(struct btrfs_root *root, struct inode *inode,
 			      u64 bytes);
 /* ctree.c */
+int btrfs_bin_search(struct extent_buffer *eb, struct btrfs_key *key,
+		     int level, int *slot);
+int btrfs_comp_cpu_keys(struct btrfs_key *k1, struct btrfs_key *k2);
 int btrfs_previous_item(struct btrfs_root *root,
 			struct btrfs_path *path, u64 min_objectid,
 			int type);
-int btrfs_merge_path(struct btrfs_trans_handle *trans,
-		     struct btrfs_root *root,
-		     struct btrfs_key *node_keys,
-		     u64 *nodes, int lowest_level);
 int btrfs_set_item_key_safe(struct btrfs_trans_handle *trans,
 			    struct btrfs_root *root, struct btrfs_path *path,
 			    struct btrfs_key *new_key);
@@ -1918,6 +2017,8 @@ int btrfs_copy_root(struct btrfs_trans_handle *trans,
 		      struct btrfs_root *root,
 		      struct extent_buffer *buf,
 		      struct extent_buffer **cow_ret, u64 new_root_objectid);
+int btrfs_block_can_be_shared(struct btrfs_root *root,
+			      struct extent_buffer *buf);
 int btrfs_extend_item(struct btrfs_trans_handle *trans, struct btrfs_root
 		      *root, struct btrfs_path *path, u32 data_size);
 int btrfs_truncate_item(struct btrfs_trans_handle *trans,
@@ -1944,9 +2045,6 @@ void btrfs_unlock_up_safe(struct btrfs_path *p, int level);
 
 int btrfs_del_items(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 		   struct btrfs_path *path, int slot, int nr);
-int btrfs_del_leaf(struct btrfs_trans_handle *trans,
-			    struct btrfs_root *root,
-			    struct btrfs_path *path, u64 bytenr);
 static inline int btrfs_del_item(struct btrfs_trans_handle *trans,
 				 struct btrfs_root *root,
 				 struct btrfs_path *path)
@@ -2005,8 +2103,9 @@ int btrfs_find_last_root(struct btrfs_root *root, u64 objectid, struct
 			 btrfs_root_item *item, struct btrfs_key *key);
 int btrfs_search_root(struct btrfs_root *root, u64 search_start,
 		      u64 *found_objectid);
-int btrfs_find_dead_roots(struct btrfs_root *root, u64 objectid,
-			  struct btrfs_root *latest_root);
+int btrfs_find_dead_roots(struct btrfs_root *root, u64 objectid);
+int btrfs_set_root_node(struct btrfs_root_item *item,
+			struct extent_buffer *node);
 /* dir-item.c */
 int btrfs_insert_dir_item(struct btrfs_trans_handle *trans,
 			  struct btrfs_root *root, const char *name,
@@ -2139,7 +2238,6 @@ int btrfs_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
 int btrfs_readpage(struct file *file, struct page *page);
 void btrfs_delete_inode(struct inode *inode);
 void btrfs_put_inode(struct inode *inode);
-void btrfs_read_locked_inode(struct inode *inode);
 int btrfs_write_inode(struct inode *inode, int wait);
 void btrfs_dirty_inode(struct inode *inode);
 struct inode *btrfs_alloc_inode(struct super_block *sb);
@@ -2147,12 +2245,8 @@ void btrfs_destroy_inode(struct inode *inode);
 int btrfs_init_cachep(void);
 void btrfs_destroy_cachep(void);
 long btrfs_ioctl_trans_end(struct file *file);
-struct inode *btrfs_ilookup(struct super_block *s, u64 objectid,
-			    struct btrfs_root *root, int wait);
-struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid,
-				struct btrfs_root *root);
 struct inode *btrfs_iget(struct super_block *s, struct btrfs_key *location,
-			 struct btrfs_root *root, int *is_new);
+			 struct btrfs_root *root);
 int btrfs_commit_write(struct file *file, struct page *page,
 		       unsigned from, unsigned to);
 struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
@@ -2168,6 +2262,8 @@ int btrfs_cont_expand(struct inode *inode, loff_t size);
 
 /* ioctl.c */
 long btrfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
+void btrfs_update_iflags(struct inode *inode);
+void btrfs_inherit_iflags(struct inode *inode, struct inode *dir);
 
 /* file.c */
 int btrfs_sync_file(struct file *file, struct dentry *dentry, int datasync);
@@ -2205,8 +2301,20 @@ int btrfs_parse_options(struct btrfs_root *root, char *options);
 int btrfs_sync_fs(struct super_block *sb, int wait);
 
 /* acl.c */
+#ifdef CONFIG_FS_POSIX_ACL
 int btrfs_check_acl(struct inode *inode, int mask);
+#else
+#define btrfs_check_acl NULL
+#endif
 int btrfs_init_acl(struct inode *inode, struct inode *dir);
 int btrfs_acl_chmod(struct inode *inode);
 
+/* relocation.c */
+int btrfs_relocate_block_group(struct btrfs_root *root, u64 group_start);
+int btrfs_init_reloc_root(struct btrfs_trans_handle *trans,
+			  struct btrfs_root *root);
+int btrfs_update_reloc_root(struct btrfs_trans_handle *trans,
+			    struct btrfs_root *root);
+int btrfs_recover_relocation(struct btrfs_root *root);
+int btrfs_reloc_clone_csums(struct inode *inode, u64 file_pos, u64 len);
 #endif

fs/btrfs/delayed-ref.c

@@ -29,27 +29,87 @@
  * add extents in the middle of btrfs_search_slot, and it allows
  * us to buffer up frequently modified backrefs in an rb tree instead
  * of hammering updates on the extent allocation tree.
- *
- * Right now this code is only used for reference counted trees, but
- * the long term goal is to get rid of the similar code for delayed
- * extent tree modifications.
  */
 
 /*
- * entries in the rb tree are ordered by the byte number of the extent
- * and by the byte number of the parent block.
+ * compare two delayed tree backrefs with same bytenr and type
+ */
+static int comp_tree_refs(struct btrfs_delayed_tree_ref *ref2,
+			  struct btrfs_delayed_tree_ref *ref1)
+{
+	if (ref1->node.type == BTRFS_TREE_BLOCK_REF_KEY) {
+		if (ref1->root < ref2->root)
+			return -1;
+		if (ref1->root > ref2->root)
+			return 1;
+	} else {
+		if (ref1->parent < ref2->parent)
+			return -1;
+		if (ref1->parent > ref2->parent)
+			return 1;
+	}
+	return 0;
+}
+
+/*
+ * compare two delayed data backrefs with same bytenr and type
  */
-static int comp_entry(struct btrfs_delayed_ref_node *ref,
-		      u64 bytenr, u64 parent)
+static int comp_data_refs(struct btrfs_delayed_data_ref *ref2,
+			  struct btrfs_delayed_data_ref *ref1)
 {
-	if (bytenr < ref->bytenr)
+	if (ref1->node.type == BTRFS_EXTENT_DATA_REF_KEY) {
+		if (ref1->root < ref2->root)
+			return -1;
+		if (ref1->root > ref2->root)
+			return 1;
+		if (ref1->objectid < ref2->objectid)
+			return -1;
+		if (ref1->objectid > ref2->objectid)
+			return 1;
+		if (ref1->offset < ref2->offset)
+			return -1;
+		if (ref1->offset > ref2->offset)
+			return 1;
+	} else {
+		if (ref1->parent < ref2->parent)
+			return -1;
+		if (ref1->parent > ref2->parent)
+			return 1;
+	}
+	return 0;
+}
+
+/*
+ * entries in the rb tree are ordered by the byte number of the extent,
+ * type of the delayed backrefs and content of delayed backrefs.
+ */
+static int comp_entry(struct btrfs_delayed_ref_node *ref2,
+		      struct btrfs_delayed_ref_node *ref1)
+{
+	if (ref1->bytenr < ref2->bytenr)
 		return -1;
-	if (bytenr > ref->bytenr)
+	if (ref1->bytenr > ref2->bytenr)
 		return 1;
-	if (parent < ref->parent)
+	if (ref1->is_head && ref2->is_head)
+		return 0;
+	if (ref2->is_head)
 		return -1;
-	if (parent > ref->parent)
+	if (ref1->is_head)
 		return 1;
+	if (ref1->type < ref2->type)
+		return -1;
+	if (ref1->type > ref2->type)
+		return 1;
+	if (ref1->type == BTRFS_TREE_BLOCK_REF_KEY ||
+	    ref1->type == BTRFS_SHARED_BLOCK_REF_KEY) {
+		return comp_tree_refs(btrfs_delayed_node_to_tree_ref(ref2),
+				      btrfs_delayed_node_to_tree_ref(ref1));
+	} else if (ref1->type == BTRFS_EXTENT_DATA_REF_KEY ||
+		   ref1->type == BTRFS_SHARED_DATA_REF_KEY) {
+		return comp_data_refs(btrfs_delayed_node_to_data_ref(ref2),
+				      btrfs_delayed_node_to_data_ref(ref1));
+	}
+	BUG();
 	return 0;
 }
 
@@ -59,20 +119,21 @@ static int comp_entry(struct btrfs_delayed_ref_node *ref,
  * inserted.
  */
 static struct btrfs_delayed_ref_node *tree_insert(struct rb_root *root,
-						  u64 bytenr, u64 parent,
 						  struct rb_node *node)
 {
 	struct rb_node **p = &root->rb_node;
 	struct rb_node *parent_node = NULL;
 	struct btrfs_delayed_ref_node *entry;
+	struct btrfs_delayed_ref_node *ins;
 	int cmp;
 
+	ins = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);
 	while (*p) {
 		parent_node = *p;
 		entry = rb_entry(parent_node, struct btrfs_delayed_ref_node,
 				 rb_node);
 
-		cmp = comp_entry(entry, bytenr, parent);
+		cmp = comp_entry(entry, ins);
 		if (cmp < 0)
 			p = &(*p)->rb_left;
 		else if (cmp > 0)
@@ -81,18 +142,17 @@ static struct btrfs_delayed_ref_node *tree_insert(struct rb_root *root,
 			return entry;
 	}
 
-	entry = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);
 	rb_link_node(node, parent_node, p);
 	rb_insert_color(node, root);
 	return NULL;
 }
 
 /*
- * find an entry based on (bytenr,parent).  This returns the delayed
- * ref if it was able to find one, or NULL if nothing was in that spot
+ * find an head entry based on bytenr. This returns the delayed ref
+ * head if it was able to find one, or NULL if nothing was in that spot
  */
-static struct btrfs_delayed_ref_node *tree_search(struct rb_root *root,
-				  u64 bytenr, u64 parent,
+static struct btrfs_delayed_ref_node *find_ref_head(struct rb_root *root,
+				  u64 bytenr,
 				  struct btrfs_delayed_ref_node **last)
 {
 	struct rb_node *n = root->rb_node;
@@ -105,7 +165,15 @@ static struct btrfs_delayed_ref_node *tree_search(struct rb_root *root,
 		if (last)
 			*last = entry;
 
-		cmp = comp_entry(entry, bytenr, parent);
+		if (bytenr < entry->bytenr)
+			cmp = -1;
+		else if (bytenr > entry->bytenr)
+			cmp = 1;
+		else if (!btrfs_delayed_ref_is_head(entry))
+			cmp = 1;
+		else
+			cmp = 0;
+
 		if (cmp < 0)
 			n = n->rb_left;
 		else if (cmp > 0)
@@ -154,7 +222,7 @@ int btrfs_find_ref_cluster(struct btrfs_trans_handle *trans,
 		node = rb_first(&delayed_refs->root);
 	} else {
 		ref = NULL;
-		tree_search(&delayed_refs->root, start, (u64)-1, &ref);
+		find_ref_head(&delayed_refs->root, start, &ref);
 		if (ref) {
 			struct btrfs_delayed_ref_node *tmp;
 
@@ -234,7 +302,7 @@ int btrfs_delayed_ref_pending(struct btrfs_trans_handle *trans, u64 bytenr)
 	delayed_refs = &trans->transaction->delayed_refs;
 	spin_lock(&delayed_refs->lock);
 
-	ref = tree_search(&delayed_refs->root, bytenr, (u64)-1, NULL);
+	ref = find_ref_head(&delayed_refs->root, bytenr, NULL);
 	if (ref) {
 		prev_node = rb_prev(&ref->rb_node);
 		if (!prev_node)
@@ -250,25 +318,28 @@ out:
 }
 
 /*
- * helper function to lookup reference count
+ * helper function to lookup reference count and flags of extent.
  *
  * the head node for delayed ref is used to store the sum of all the
- * reference count modifications queued up in the rbtree.  This way you
- * can check to see what the reference count would be if all of the
- * delayed refs are processed.
+ * reference count modifications queued up in the rbtree. the head
+ * node may also store the extent flags to set. This way you can check
+ * to see what the reference count and extent flags would be if all of
+ * the delayed refs are not processed.
  */
-int btrfs_lookup_extent_ref(struct btrfs_trans_handle *trans,
-			    struct btrfs_root *root, u64 bytenr,
-			    u64 num_bytes, u32 *refs)
+int btrfs_lookup_extent_info(struct btrfs_trans_handle *trans,
+			     struct btrfs_root *root, u64 bytenr,
+			     u64 num_bytes, u64 *refs, u64 *flags)
 {
 	struct btrfs_delayed_ref_node *ref;
 	struct btrfs_delayed_ref_head *head;
 	struct btrfs_delayed_ref_root *delayed_refs;
 	struct btrfs_path *path;
-	struct extent_buffer *leaf;
 	struct btrfs_extent_item *ei;
+	struct extent_buffer *leaf;
 	struct btrfs_key key;
-	u32 num_refs;
+	u32 item_size;
+	u64 num_refs;
+	u64 extent_flags;
 	int ret;
 
 	path = btrfs_alloc_path();
@@ -287,37 +358,60 @@ again:
 
 	if (ret == 0) {
 		leaf = path->nodes[0];
-		ei = btrfs_item_ptr(leaf, path->slots[0],
-				    struct btrfs_extent_item);
-		num_refs = btrfs_extent_refs(leaf, ei);
+		item_size = btrfs_item_size_nr(leaf, path->slots[0]);
+		if (item_size >= sizeof(*ei)) {
+			ei = btrfs_item_ptr(leaf, path->slots[0],
+					    struct btrfs_extent_item);
+			num_refs = btrfs_extent_refs(leaf, ei);
+			extent_flags = btrfs_extent_flags(leaf, ei);
+		} else {
+#ifdef BTRFS_COMPAT_EXTENT_TREE_V0
+			struct btrfs_extent_item_v0 *ei0;
+			BUG_ON(item_size != sizeof(*ei0));
+			ei0 = btrfs_item_ptr(leaf, path->slots[0],
+					     struct btrfs_extent_item_v0);
+			num_refs = btrfs_extent_refs_v0(leaf, ei0);
+			/* FIXME: this isn't correct for data */
+			extent_flags = BTRFS_BLOCK_FLAG_FULL_BACKREF;
+#else
+			BUG();
+#endif
+		}
+		BUG_ON(num_refs == 0);
 	} else {
 		num_refs = 0;
+		extent_flags = 0;
 		ret = 0;
 	}
 
 	spin_lock(&delayed_refs->lock);
-	ref = tree_search(&delayed_refs->root, bytenr, (u64)-1, NULL);
+	ref = find_ref_head(&delayed_refs->root, bytenr, NULL);
 	if (ref) {
 		head = btrfs_delayed_node_to_head(ref);
-		if (mutex_trylock(&head->mutex)) {
-			num_refs += ref->ref_mod;
-			mutex_unlock(&head->mutex);
-			*refs = num_refs;
-			goto out;
-		}
+		if (!mutex_trylock(&head->mutex)) {
+			atomic_inc(&ref->refs);
+			spin_unlock(&delayed_refs->lock);
 
-		atomic_inc(&ref->refs);
-		spin_unlock(&delayed_refs->lock);
+			btrfs_release_path(root->fs_info->extent_root, path);
 
-		btrfs_release_path(root->fs_info->extent_root, path);
+			mutex_lock(&head->mutex);
+			mutex_unlock(&head->mutex);
+			btrfs_put_delayed_ref(ref);
+			goto again;
+		}
+		if (head->extent_op && head->extent_op->update_flags)
+			extent_flags |= head->extent_op->flags_to_set;
+		else
+			BUG_ON(num_refs == 0);
 
-		mutex_lock(&head->mutex);
+		num_refs += ref->ref_mod;
 		mutex_unlock(&head->mutex);
-		btrfs_put_delayed_ref(ref);
-		goto again;
-	} else {
-		*refs = num_refs;
 	}
+	WARN_ON(num_refs == 0);
+	if (refs)
+		*refs = num_refs;
+	if (flags)
+		*flags = extent_flags;
 out:
 	spin_unlock(&delayed_refs->lock);
 	btrfs_free_path(path);
@@ -338,16 +432,7 @@ update_existing_ref(struct btrfs_trans_handle *trans,
 		    struct btrfs_delayed_ref_node *existing,
 		    struct btrfs_delayed_ref_node *update)
 {
-	struct btrfs_delayed_ref *existing_ref;
-	struct btrfs_delayed_ref *ref;
-
-	existing_ref = btrfs_delayed_node_to_ref(existing);
-	ref = btrfs_delayed_node_to_ref(update);
-
-	if (ref->pin)
-		existing_ref->pin = 1;
-
-	if (ref->action != existing_ref->action) {
+	if (update->action != existing->action) {
 		/*
 		 * this is effectively undoing either an add or a
 		 * drop.  We decrement the ref_mod, and if it goes
@@ -363,20 +448,13 @@ update_existing_ref(struct btrfs_trans_handle *trans,
 			delayed_refs->num_entries--;
 			if (trans->delayed_ref_updates)
 				trans->delayed_ref_updates--;
+		} else {
+			WARN_ON(existing->type == BTRFS_TREE_BLOCK_REF_KEY ||
+				existing->type == BTRFS_SHARED_BLOCK_REF_KEY);
 		}
 	} else {
-		if (existing_ref->action == BTRFS_ADD_DELAYED_REF) {
-			/* if we're adding refs, make sure all the
-			 * details match up.  The extent could
-			 * have been totally freed and reallocated
-			 * by a different owner before the delayed
-			 * ref entries were removed.
-			 */
-			existing_ref->owner_objectid = ref->owner_objectid;
-			existing_ref->generation = ref->generation;
-			existing_ref->root = ref->root;
-			existing->num_bytes = update->num_bytes;
-		}
+		WARN_ON(existing->type == BTRFS_TREE_BLOCK_REF_KEY ||
+			existing->type == BTRFS_SHARED_BLOCK_REF_KEY);
 		/*
 		 * the action on the existing ref matches
 		 * the action on the ref we're trying to add.
@@ -401,6 +479,7 @@ update_existing_head_ref(struct btrfs_delayed_ref_node *existing,
 
 	existing_ref = btrfs_delayed_node_to_head(existing);
 	ref = btrfs_delayed_node_to_head(update);
+	BUG_ON(existing_ref->is_data != ref->is_data);
 
 	if (ref->must_insert_reserved) {
 		/* if the extent was freed and then
@@ -420,6 +499,24 @@ update_existing_head_ref(struct btrfs_delayed_ref_node *existing,
 
 	}
 
+	if (ref->extent_op) {
+		if (!existing_ref->extent_op) {
+			existing_ref->extent_op = ref->extent_op;
+		} else {
+			if (ref->extent_op->update_key) {
+				memcpy(&existing_ref->extent_op->key,
+				       &ref->extent_op->key,
+				       sizeof(ref->extent_op->key));
+				existing_ref->extent_op->update_key = 1;
+			}
+			if (ref->extent_op->update_flags) {
+				existing_ref->extent_op->flags_to_set |=
+					ref->extent_op->flags_to_set;
+				existing_ref->extent_op->update_flags = 1;
+			}
+			kfree(ref->extent_op);
+		}
+	}
 	/*
 	 * update the reference mod on the head to reflect this new operation
 	 */
@@ -427,19 +524,16 @@ update_existing_head_ref(struct btrfs_delayed_ref_node *existing,
 }
 
 /*
- * helper function to actually insert a delayed ref into the rbtree.
+ * helper function to actually insert a head node into the rbtree.
  * this does all the dirty work in terms of maintaining the correct
- * overall modification count in the head node and properly dealing
- * with updating existing nodes as new modifications are queued.
+ * overall modification count.
  */
-static noinline int __btrfs_add_delayed_ref(struct btrfs_trans_handle *trans,
-			  struct btrfs_delayed_ref_node *ref,
-			  u64 bytenr, u64 num_bytes, u64 parent, u64 ref_root,
-			  u64 ref_generation, u64 owner_objectid, int action,
-			  int pin)
+static noinline int add_delayed_ref_head(struct btrfs_trans_handle *trans,
+					struct btrfs_delayed_ref_node *ref,
+					u64 bytenr, u64 num_bytes,
+					int action, int is_data)
 {
 	struct btrfs_delayed_ref_node *existing;
-	struct btrfs_delayed_ref *full_ref;
 	struct btrfs_delayed_ref_head *head_ref = NULL;
 	struct btrfs_delayed_ref_root *delayed_refs;
 	int count_mod = 1;
@@ -449,12 +543,10 @@ static noinline int __btrfs_add_delayed_ref(struct btrfs_trans_handle *trans,
 	 * the head node stores the sum of all the mods, so dropping a ref
 	 * should drop the sum in the head node by one.
 	 */
-	if (parent == (u64)-1) {
-		if (action == BTRFS_DROP_DELAYED_REF)
-			count_mod = -1;
-		else if (action == BTRFS_UPDATE_DELAYED_HEAD)
-			count_mod = 0;
-	}
+	if (action == BTRFS_UPDATE_DELAYED_HEAD)
+		count_mod = 0;
+	else if (action == BTRFS_DROP_DELAYED_REF)
+		count_mod = -1;
 
 	/*
 	 * BTRFS_ADD_DELAYED_EXTENT means that we need to update
@@ -467,57 +559,148 @@ static noinline int __btrfs_add_delayed_ref(struct btrfs_trans_handle *trans,
 	 * Once we record must_insert_reserved, switch the action to
 	 * BTRFS_ADD_DELAYED_REF because other special casing is not required.
 	 */
-	if (action == BTRFS_ADD_DELAYED_EXTENT) {
+	if (action == BTRFS_ADD_DELAYED_EXTENT)
 		must_insert_reserved = 1;
-		action = BTRFS_ADD_DELAYED_REF;
-	} else {
+	else
 		must_insert_reserved = 0;
-	}
-
 
 	delayed_refs = &trans->transaction->delayed_refs;
 
 	/* first set the basic ref node struct up */
 	atomic_set(&ref->refs, 1);
 	ref->bytenr = bytenr;
-	ref->parent = parent;
+	ref->num_bytes = num_bytes;
 	ref->ref_mod = count_mod;
+	ref->type  = 0;
+	ref->action  = 0;
+	ref->is_head = 1;
 	ref->in_tree = 1;
+
+	head_ref = btrfs_delayed_node_to_head(ref);
+	head_ref->must_insert_reserved = must_insert_reserved;
+	head_ref->is_data = is_data;
+
+	INIT_LIST_HEAD(&head_ref->cluster);
+	mutex_init(&head_ref->mutex);
+
+	existing = tree_insert(&delayed_refs->root, &ref->rb_node);
+
+	if (existing) {
+		update_existing_head_ref(existing, ref);
+		/*
+		 * we've updated the existing ref, free the newly
+		 * allocated ref
+		 */
+		kfree(ref);
+	} else {
+		delayed_refs->num_heads++;
+		delayed_refs->num_heads_ready++;
+		delayed_refs->num_entries++;
+		trans->delayed_ref_updates++;
+	}
+	return 0;
+}
+
+/*
+ * helper to insert a delayed tree ref into the rbtree.
+ */
+static noinline int add_delayed_tree_ref(struct btrfs_trans_handle *trans,
+					 struct btrfs_delayed_ref_node *ref,
+					 u64 bytenr, u64 num_bytes, u64 parent,
+					 u64 ref_root, int level, int action)
+{
+	struct btrfs_delayed_ref_node *existing;
+	struct btrfs_delayed_tree_ref *full_ref;
+	struct btrfs_delayed_ref_root *delayed_refs;
+
+	if (action == BTRFS_ADD_DELAYED_EXTENT)
+		action = BTRFS_ADD_DELAYED_REF;
+
+	delayed_refs = &trans->transaction->delayed_refs;
+
+	/* first set the basic ref node struct up */
+	atomic_set(&ref->refs, 1);
+	ref->bytenr = bytenr;
 	ref->num_bytes = num_bytes;
+	ref->ref_mod = 1;
+	ref->action = action;
+	ref->is_head = 0;
+	ref->in_tree = 1;
 
-	if (btrfs_delayed_ref_is_head(ref)) {
-		head_ref = btrfs_delayed_node_to_head(ref);
-		head_ref->must_insert_reserved = must_insert_reserved;
-		INIT_LIST_HEAD(&head_ref->cluster);
-		mutex_init(&head_ref->mutex);
+	full_ref = btrfs_delayed_node_to_tree_ref(ref);
+	if (parent) {
+		full_ref->parent = parent;
+		ref->type = BTRFS_SHARED_BLOCK_REF_KEY;
 	} else {
-		full_ref = btrfs_delayed_node_to_ref(ref);
 		full_ref->root = ref_root;
-		full_ref->generation = ref_generation;
-		full_ref->owner_objectid = owner_objectid;
-		full_ref->pin = pin;
-		full_ref->action = action;
+		ref->type = BTRFS_TREE_BLOCK_REF_KEY;
 	}
+	full_ref->level = level;
 
-	existing = tree_insert(&delayed_refs->root, bytenr,
-			       parent, &ref->rb_node);
+	existing = tree_insert(&delayed_refs->root, &ref->rb_node);
 
 	if (existing) {
-		if (btrfs_delayed_ref_is_head(ref))
-			update_existing_head_ref(existing, ref);
-		else
-			update_existing_ref(trans, delayed_refs, existing, ref);
+		update_existing_ref(trans, delayed_refs, existing, ref);
+		/*
+		 * we've updated the existing ref, free the newly
+		 * allocated ref
+		 */
+		kfree(ref);
+	} else {
+		delayed_refs->num_entries++;
+		trans->delayed_ref_updates++;
+	}
+	return 0;
+}
+
+/*
+ * helper to insert a delayed data ref into the rbtree.
+ */
+static noinline int add_delayed_data_ref(struct btrfs_trans_handle *trans,
+					 struct btrfs_delayed_ref_node *ref,
+					 u64 bytenr, u64 num_bytes, u64 parent,
+					 u64 ref_root, u64 owner, u64 offset,
+					 int action)
+{
+	struct btrfs_delayed_ref_node *existing;
+	struct btrfs_delayed_data_ref *full_ref;
+	struct btrfs_delayed_ref_root *delayed_refs;
+
+	if (action == BTRFS_ADD_DELAYED_EXTENT)
+		action = BTRFS_ADD_DELAYED_REF;
+
+	delayed_refs = &trans->transaction->delayed_refs;
+
+	/* first set the basic ref node struct up */
+	atomic_set(&ref->refs, 1);
+	ref->bytenr = bytenr;
+	ref->num_bytes = num_bytes;
+	ref->ref_mod = 1;
+	ref->action = action;
+	ref->is_head = 0;
+	ref->in_tree = 1;
+
+	full_ref = btrfs_delayed_node_to_data_ref(ref);
+	if (parent) {
+		full_ref->parent = parent;
+		ref->type = BTRFS_SHARED_DATA_REF_KEY;
+	} else {
+		full_ref->root = ref_root;
+		ref->type = BTRFS_EXTENT_DATA_REF_KEY;
+	}
+	full_ref->objectid = owner;
+	full_ref->offset = offset;
 
+	existing = tree_insert(&delayed_refs->root, &ref->rb_node);
+
+	if (existing) {
+		update_existing_ref(trans, delayed_refs, existing, ref);
 		/*
 		 * we've updated the existing ref, free the newly
 		 * allocated ref
 		 */
 		kfree(ref);
 	} else {
-		if (btrfs_delayed_ref_is_head(ref)) {
-			delayed_refs->num_heads++;
-			delayed_refs->num_heads_ready++;
-		}
 		delayed_refs->num_entries++;
 		trans->delayed_ref_updates++;
 	}
@@ -525,37 +708,78 @@ static noinline int __btrfs_add_delayed_ref(struct btrfs_trans_handle *trans,
 }
 
 /*
- * add a delayed ref to the tree.  This does all of the accounting required
+ * add a delayed tree ref.  This does all of the accounting required
  * to make sure the delayed ref is eventually processed before this
  * transaction commits.
  */
-int btrfs_add_delayed_ref(struct btrfs_trans_handle *trans,
-			  u64 bytenr, u64 num_bytes, u64 parent, u64 ref_root,
-			  u64 ref_generation, u64 owner_objectid, int action,
-			  int pin)
+int btrfs_add_delayed_tree_ref(struct btrfs_trans_handle *trans,
+			       u64 bytenr, u64 num_bytes, u64 parent,
+			       u64 ref_root,  int level, int action,
+			       struct btrfs_delayed_extent_op *extent_op)
 {
-	struct btrfs_delayed_ref *ref;
+	struct btrfs_delayed_tree_ref *ref;
 	struct btrfs_delayed_ref_head *head_ref;
 	struct btrfs_delayed_ref_root *delayed_refs;
 	int ret;
 
+	BUG_ON(extent_op && extent_op->is_data);
 	ref = kmalloc(sizeof(*ref), GFP_NOFS);
 	if (!ref)
 		return -ENOMEM;
 
+	head_ref = kmalloc(sizeof(*head_ref), GFP_NOFS);
+	if (!head_ref) {
+		kfree(ref);
+		return -ENOMEM;
+	}
+
+	head_ref->extent_op = extent_op;
+
+	delayed_refs = &trans->transaction->delayed_refs;
+	spin_lock(&delayed_refs->lock);
+
 	/*
-	 * the parent = 0 case comes from cases where we don't actually
-	 * know the parent yet.  It will get updated later via a add/drop
-	 * pair.
+	 * insert both the head node and the new ref without dropping
+	 * the spin lock
 	 */
-	if (parent == 0)
-		parent = bytenr;
+	ret = add_delayed_ref_head(trans, &head_ref->node, bytenr, num_bytes,
+				   action, 0);
+	BUG_ON(ret);
+
+	ret = add_delayed_tree_ref(trans, &ref->node, bytenr, num_bytes,
+				   parent, ref_root, level, action);
+	BUG_ON(ret);
+	spin_unlock(&delayed_refs->lock);
+	return 0;
+}
+
+/*
+ * add a delayed data ref. it's similar to btrfs_add_delayed_tree_ref.
+ */
+int btrfs_add_delayed_data_ref(struct btrfs_trans_handle *trans,
+			       u64 bytenr, u64 num_bytes,
+			       u64 parent, u64 ref_root,
+			       u64 owner, u64 offset, int action,
+			       struct btrfs_delayed_extent_op *extent_op)
+{
+	struct btrfs_delayed_data_ref *ref;
+	struct btrfs_delayed_ref_head *head_ref;
+	struct btrfs_delayed_ref_root *delayed_refs;
+	int ret;
+
+	BUG_ON(extent_op && !extent_op->is_data);
+	ref = kmalloc(sizeof(*ref), GFP_NOFS);
+	if (!ref)
+		return -ENOMEM;
 
 	head_ref = kmalloc(sizeof(*head_ref), GFP_NOFS);
 	if (!head_ref) {
 		kfree(ref);
 		return -ENOMEM;
 	}
+
+	head_ref->extent_op = extent_op;
+
 	delayed_refs = &trans->transaction->delayed_refs;
 	spin_lock(&delayed_refs->lock);
 
@@ -563,14 +787,39 @@ int btrfs_add_delayed_ref(struct btrfs_trans_handle *trans,
 	 * insert both the head node and the new ref without dropping
 	 * the spin lock
 	 */
-	ret = __btrfs_add_delayed_ref(trans, &head_ref->node, bytenr, num_bytes,
-				      (u64)-1, 0, 0, 0, action, pin);
+	ret = add_delayed_ref_head(trans, &head_ref->node, bytenr, num_bytes,
+				   action, 1);
 	BUG_ON(ret);
 
-	ret = __btrfs_add_delayed_ref(trans, &ref->node, bytenr, num_bytes,
-				      parent, ref_root, ref_generation,
-				      owner_objectid, action, pin);
+	ret = add_delayed_data_ref(trans, &ref->node, bytenr, num_bytes,
+				   parent, ref_root, owner, offset, action);
+	BUG_ON(ret);
+	spin_unlock(&delayed_refs->lock);
+	return 0;
+}
+
+int btrfs_add_delayed_extent_op(struct btrfs_trans_handle *trans,
+				u64 bytenr, u64 num_bytes,
+				struct btrfs_delayed_extent_op *extent_op)
+{
+	struct btrfs_delayed_ref_head *head_ref;
+	struct btrfs_delayed_ref_root *delayed_refs;
+	int ret;
+
+	head_ref = kmalloc(sizeof(*head_ref), GFP_NOFS);
+	if (!head_ref)
+		return -ENOMEM;
+
+	head_ref->extent_op = extent_op;
+
+	delayed_refs = &trans->transaction->delayed_refs;
+	spin_lock(&delayed_refs->lock);
+
+	ret = add_delayed_ref_head(trans, &head_ref->node, bytenr,
+				   num_bytes, BTRFS_UPDATE_DELAYED_HEAD,
+				   extent_op->is_data);
 	BUG_ON(ret);
+
 	spin_unlock(&delayed_refs->lock);
 	return 0;
 }
@@ -587,7 +836,7 @@ btrfs_find_delayed_ref_head(struct btrfs_trans_handle *trans, u64 bytenr)
 	struct btrfs_delayed_ref_root *delayed_refs;
 
 	delayed_refs = &trans->transaction->delayed_refs;
-	ref = tree_search(&delayed_refs->root, bytenr, (u64)-1, NULL);
+	ref = find_ref_head(&delayed_refs->root, bytenr, NULL);
 	if (ref)
 		return btrfs_delayed_node_to_head(ref);
 	return NULL;
@@ -603,6 +852,7 @@ btrfs_find_delayed_ref_head(struct btrfs_trans_handle *trans, u64 bytenr)
  *
  * It is the same as doing a ref add and delete in two separate calls.
  */
+#if 0
 int btrfs_update_delayed_ref(struct btrfs_trans_handle *trans,
 			  u64 bytenr, u64 num_bytes, u64 orig_parent,
 			  u64 parent, u64 orig_ref_root, u64 ref_root,
@@ -666,3 +916,4 @@ int btrfs_update_delayed_ref(struct btrfs_trans_handle *trans,
 	spin_unlock(&delayed_refs->lock);
 	return 0;
 }
+#endif

fs/btrfs/delayed-ref.h

@@ -30,9 +30,6 @@ struct btrfs_delayed_ref_node {
 	/* the starting bytenr of the extent */
 	u64 bytenr;
 
-	/* the parent our backref will point to */
-	u64 parent;
-
 	/* the size of the extent */
 	u64 num_bytes;
 
@@ -50,10 +47,21 @@ struct btrfs_delayed_ref_node {
 	 */
 	int ref_mod;
 
+	unsigned int action:8;
+	unsigned int type:8;
 	/* is this node still in the rbtree? */
+	unsigned int is_head:1;
 	unsigned int in_tree:1;
 };
 
+struct btrfs_delayed_extent_op {
+	struct btrfs_disk_key key;
+	u64 flags_to_set;
+	unsigned int update_key:1;
+	unsigned int update_flags:1;
+	unsigned int is_data:1;
+};
+
 /*
  * the head refs are used to hold a lock on a given extent, which allows us
  * to make sure that only one process is running the delayed refs
@@ -71,6 +79,7 @@ struct btrfs_delayed_ref_head {
 
 	struct list_head cluster;
 
+	struct btrfs_delayed_extent_op *extent_op;
 	/*
 	 * when a new extent is allocated, it is just reserved in memory
 	 * The actual extent isn't inserted into the extent allocation tree
@@ -84,27 +93,26 @@ struct btrfs_delayed_ref_head {
 	 * the free has happened.
 	 */
 	unsigned int must_insert_reserved:1;
+	unsigned int is_data:1;
 };
 
-struct btrfs_delayed_ref {
+struct btrfs_delayed_tree_ref {
 	struct btrfs_delayed_ref_node node;
+	union {
+		u64 root;
+		u64 parent;
+	};
+	int level;
+};
 
-	/* the root objectid our ref will point to */
-	u64 root;
-
-	/* the generation for the backref */
-	u64 generation;
-
-	/* owner_objectid of the backref  */
-	u64 owner_objectid;
-
-	/* operation done by this entry in the rbtree */
-	u8 action;
-
-	/* if pin == 1, when the extent is freed it will be pinned until
-	 * transaction commit
-	 */
-	unsigned int pin:1;
+struct btrfs_delayed_data_ref {
+	struct btrfs_delayed_ref_node node;
+	union {
+		u64 root;
+		u64 parent;
+	};
+	u64 objectid;
+	u64 offset;
 };
 
 struct btrfs_delayed_ref_root {
@@ -143,17 +151,25 @@ static inline void btrfs_put_delayed_ref(struct btrfs_delayed_ref_node *ref)
 	}
 }
 
-int btrfs_add_delayed_ref(struct btrfs_trans_handle *trans,
-			  u64 bytenr, u64 num_bytes, u64 parent, u64 ref_root,
-			  u64 ref_generation, u64 owner_objectid, int action,
-			  int pin);
+int btrfs_add_delayed_tree_ref(struct btrfs_trans_handle *trans,
+			       u64 bytenr, u64 num_bytes, u64 parent,
+			       u64 ref_root, int level, int action,
+			       struct btrfs_delayed_extent_op *extent_op);
+int btrfs_add_delayed_data_ref(struct btrfs_trans_handle *trans,
+			       u64 bytenr, u64 num_bytes,
+			       u64 parent, u64 ref_root,
+			       u64 owner, u64 offset, int action,
+			       struct btrfs_delayed_extent_op *extent_op);
+int btrfs_add_delayed_extent_op(struct btrfs_trans_handle *trans,
+				u64 bytenr, u64 num_bytes,
+				struct btrfs_delayed_extent_op *extent_op);
 
 struct btrfs_delayed_ref_head *
 btrfs_find_delayed_ref_head(struct btrfs_trans_handle *trans, u64 bytenr);
 int btrfs_delayed_ref_pending(struct btrfs_trans_handle *trans, u64 bytenr);
-int btrfs_lookup_extent_ref(struct btrfs_trans_handle *trans,
-			    struct btrfs_root *root, u64 bytenr,
-			    u64 num_bytes, u32 *refs);
+int btrfs_lookup_extent_info(struct btrfs_trans_handle *trans,
+			     struct btrfs_root *root, u64 bytenr,
+			     u64 num_bytes, u64 *refs, u64 *flags);
 int btrfs_update_delayed_ref(struct btrfs_trans_handle *trans,
 			  u64 bytenr, u64 num_bytes, u64 orig_parent,
 			  u64 parent, u64 orig_ref_root, u64 ref_root,
@@ -169,18 +185,24 @@ int btrfs_find_ref_cluster(struct btrfs_trans_handle *trans,
  */
 static int btrfs_delayed_ref_is_head(struct btrfs_delayed_ref_node *node)
 {
-	return node->parent == (u64)-1;
+	return node->is_head;
 }
 
 /*
  * helper functions to cast a node into its container
  */
-static inline struct btrfs_delayed_ref *
-btrfs_delayed_node_to_ref(struct btrfs_delayed_ref_node *node)
+static inline struct btrfs_delayed_tree_ref *
+btrfs_delayed_node_to_tree_ref(struct btrfs_delayed_ref_node *node)
 {
 	WARN_ON(btrfs_delayed_ref_is_head(node));
-	return container_of(node, struct btrfs_delayed_ref, node);
+	return container_of(node, struct btrfs_delayed_tree_ref, node);
+}
 
+static inline struct btrfs_delayed_data_ref *
+btrfs_delayed_node_to_data_ref(struct btrfs_delayed_ref_node *node)
+{
+	WARN_ON(btrfs_delayed_ref_is_head(node));
+	return container_of(node, struct btrfs_delayed_data_ref, node);
 }
 
 static inline struct btrfs_delayed_ref_head *
@@ -188,6 +210,5 @@ btrfs_delayed_node_to_head(struct btrfs_delayed_ref_node *node)
 {
 	WARN_ON(!btrfs_delayed_ref_is_head(node));
 	return container_of(node, struct btrfs_delayed_ref_head, node);
-
 }
 #endif

fs/btrfs/disk-io.c

@@ -26,8 +26,8 @@
 #include <linux/workqueue.h>
 #include <linux/kthread.h>
 #include <linux/freezer.h>
+#include <linux/crc32c.h>
 #include "compat.h"
-#include "crc32c.h"
 #include "ctree.h"
 #include "disk-io.h"
 #include "transaction.h"
@@ -36,7 +36,6 @@
 #include "print-tree.h"
 #include "async-thread.h"
 #include "locking.h"
-#include "ref-cache.h"
 #include "tree-log.h"
 #include "free-space-cache.h"
 
@@ -172,7 +171,7 @@ out:
 
 u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len)
 {
-	return btrfs_crc32c(seed, data, len);
+	return crc32c(seed, data, len);
 }
 
 void btrfs_csum_final(u32 crc, char *result)
@@ -884,7 +883,6 @@ static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
 {
 	root->node = NULL;
 	root->commit_root = NULL;
-	root->ref_tree = NULL;
 	root->sectorsize = sectorsize;
 	root->nodesize = nodesize;
 	root->leafsize = leafsize;
@@ -899,12 +897,14 @@ static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
 	root->last_inode_alloc = 0;
 	root->name = NULL;
 	root->in_sysfs = 0;
+	root->inode_tree.rb_node = NULL;
 
 	INIT_LIST_HEAD(&root->dirty_list);
 	INIT_LIST_HEAD(&root->orphan_list);
-	INIT_LIST_HEAD(&root->dead_list);
+	INIT_LIST_HEAD(&root->root_list);
 	spin_lock_init(&root->node_lock);
 	spin_lock_init(&root->list_lock);
+	spin_lock_init(&root->inode_lock);
 	mutex_init(&root->objectid_mutex);
 	mutex_init(&root->log_mutex);
 	init_waitqueue_head(&root->log_writer_wait);
@@ -918,9 +918,6 @@ static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
 	extent_io_tree_init(&root->dirty_log_pages,
 			     fs_info->btree_inode->i_mapping, GFP_NOFS);
 
-	btrfs_leaf_ref_tree_init(&root->ref_tree_struct);
-	root->ref_tree = &root->ref_tree_struct;
-
 	memset(&root->root_key, 0, sizeof(root->root_key));
 	memset(&root->root_item, 0, sizeof(root->root_item));
 	memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
@@ -959,6 +956,7 @@ static int find_and_setup_root(struct btrfs_root *tree_root,
 	blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
 	root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
 				     blocksize, generation);
+	root->commit_root = btrfs_root_node(root);
 	BUG_ON(!root->node);
 	return 0;
 }
@@ -1025,20 +1023,19 @@ static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
 	 */
 	root->ref_cows = 0;
 
-	leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
-				      0, BTRFS_TREE_LOG_OBJECTID,
-				      trans->transid, 0, 0, 0);
+	leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0,
+				      BTRFS_TREE_LOG_OBJECTID, NULL, 0, 0, 0);
 	if (IS_ERR(leaf)) {
 		kfree(root);
 		return ERR_CAST(leaf);
 	}
 
+	memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
+	btrfs_set_header_bytenr(leaf, leaf->start);
+	btrfs_set_header_generation(leaf, trans->transid);
+	btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
+	btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
 	root->node = leaf;
-	btrfs_set_header_nritems(root->node, 0);
-	btrfs_set_header_level(root->node, 0);
-	btrfs_set_header_bytenr(root->node, root->node->start);
-	btrfs_set_header_generation(root->node, trans->transid);
-	btrfs_set_header_owner(root->node, BTRFS_TREE_LOG_OBJECTID);
 
 	write_extent_buffer(root->node, root->fs_info->fsid,
 			    (unsigned long)btrfs_header_fsid(root->node),
@@ -1081,8 +1078,7 @@ int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
 	inode_item->nbytes = cpu_to_le64(root->leafsize);
 	inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
 
-	btrfs_set_root_bytenr(&log_root->root_item, log_root->node->start);
-	btrfs_set_root_generation(&log_root->root_item, trans->transid);
+	btrfs_set_root_node(&log_root->root_item, log_root->node);
 
 	WARN_ON(root->log_root);
 	root->log_root = log_root;
@@ -1144,6 +1140,7 @@ out:
 	blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
 	root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
 				     blocksize, generation);
+	root->commit_root = btrfs_root_node(root);
 	BUG_ON(!root->node);
 insert:
 	if (location->objectid != BTRFS_TREE_LOG_OBJECTID) {
@@ -1210,7 +1207,7 @@ struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info,
 	}
 	if (!(fs_info->sb->s_flags & MS_RDONLY)) {
 		ret = btrfs_find_dead_roots(fs_info->tree_root,
-					    root->root_key.objectid, root);
+					    root->root_key.objectid);
 		BUG_ON(ret);
 		btrfs_orphan_cleanup(root);
 	}
@@ -1569,8 +1566,6 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	atomic_set(&fs_info->async_delalloc_pages, 0);
 	atomic_set(&fs_info->async_submit_draining, 0);
 	atomic_set(&fs_info->nr_async_bios, 0);
-	atomic_set(&fs_info->throttles, 0);
-	atomic_set(&fs_info->throttle_gen, 0);
 	fs_info->sb = sb;
 	fs_info->max_extent = (u64)-1;
 	fs_info->max_inline = 8192 * 1024;
@@ -1598,6 +1593,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
 	fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;
 
+	RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node);
 	extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
 			     fs_info->btree_inode->i_mapping,
 			     GFP_NOFS);
@@ -1613,10 +1609,6 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 			     fs_info->btree_inode->i_mapping, GFP_NOFS);
 	fs_info->do_barriers = 1;
 
-	INIT_LIST_HEAD(&fs_info->dead_reloc_roots);
-	btrfs_leaf_ref_tree_init(&fs_info->reloc_ref_tree);
-	btrfs_leaf_ref_tree_init(&fs_info->shared_ref_tree);
-
 	BTRFS_I(fs_info->btree_inode)->root = tree_root;
 	memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
 	       sizeof(struct btrfs_key));
@@ -1674,6 +1666,12 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 		goto fail_iput;
 	}
 
+	features = btrfs_super_incompat_flags(disk_super);
+	if (!(features & BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF)) {
+		features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
+		btrfs_set_super_incompat_flags(disk_super, features);
+	}
+
 	features = btrfs_super_compat_ro_flags(disk_super) &
 		~BTRFS_FEATURE_COMPAT_RO_SUPP;
 	if (!(sb->s_flags & MS_RDONLY) && features) {
@@ -1771,7 +1769,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	if (ret) {
 		printk(KERN_WARNING "btrfs: failed to read the system "
 		       "array on %s\n", sb->s_id);
-		goto fail_sys_array;
+		goto fail_sb_buffer;
 	}
 
 	blocksize = btrfs_level_size(tree_root,
@@ -1785,6 +1783,8 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 					   btrfs_super_chunk_root(disk_super),
 					   blocksize, generation);
 	BUG_ON(!chunk_root->node);
+	btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
+	chunk_root->commit_root = btrfs_root_node(chunk_root);
 
 	read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
 	   (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node),
@@ -1810,7 +1810,8 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 					  blocksize, generation);
 	if (!tree_root->node)
 		goto fail_chunk_root;
-
+	btrfs_set_root_node(&tree_root->root_item, tree_root->node);
+	tree_root->commit_root = btrfs_root_node(tree_root);
 
 	ret = find_and_setup_root(tree_root, fs_info,
 				  BTRFS_EXTENT_TREE_OBJECTID, extent_root);
@@ -1820,14 +1821,14 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 
 	ret = find_and_setup_root(tree_root, fs_info,
 				  BTRFS_DEV_TREE_OBJECTID, dev_root);
-	dev_root->track_dirty = 1;
 	if (ret)
 		goto fail_extent_root;
+	dev_root->track_dirty = 1;
 
 	ret = find_and_setup_root(tree_root, fs_info,
 				  BTRFS_CSUM_TREE_OBJECTID, csum_root);
 	if (ret)
-		goto fail_extent_root;
+		goto fail_dev_root;
 
 	csum_root->track_dirty = 1;
 
@@ -1849,6 +1850,14 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	if (IS_ERR(fs_info->transaction_kthread))
 		goto fail_cleaner;
 
+	if (!btrfs_test_opt(tree_root, SSD) &&
+	    !btrfs_test_opt(tree_root, NOSSD) &&
+	    !fs_info->fs_devices->rotating) {
+		printk(KERN_INFO "Btrfs detected SSD devices, enabling SSD "
+		       "mode\n");
+		btrfs_set_opt(fs_info->mount_opt, SSD);
+	}
+
 	if (btrfs_super_log_root(disk_super) != 0) {
 		u64 bytenr = btrfs_super_log_root(disk_super);
 
@@ -1881,7 +1890,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	}
 
 	if (!(sb->s_flags & MS_RDONLY)) {
-		ret = btrfs_cleanup_reloc_trees(tree_root);
+		ret = btrfs_recover_relocation(tree_root);
 		BUG_ON(ret);
 	}
 
@@ -1892,6 +1901,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
 	if (!fs_info->fs_root)
 		goto fail_trans_kthread;
+
 	return tree_root;
 
 fail_trans_kthread:
@@ -1908,14 +1918,19 @@ fail_cleaner:
 
 fail_csum_root:
 	free_extent_buffer(csum_root->node);
+	free_extent_buffer(csum_root->commit_root);
+fail_dev_root:
+	free_extent_buffer(dev_root->node);
+	free_extent_buffer(dev_root->commit_root);
 fail_extent_root:
 	free_extent_buffer(extent_root->node);
+	free_extent_buffer(extent_root->commit_root);
 fail_tree_root:
 	free_extent_buffer(tree_root->node);
+	free_extent_buffer(tree_root->commit_root);
 fail_chunk_root:
 	free_extent_buffer(chunk_root->node);
-fail_sys_array:
-	free_extent_buffer(dev_root->node);
+	free_extent_buffer(chunk_root->commit_root);
 fail_sb_buffer:
 	btrfs_stop_workers(&fs_info->fixup_workers);
 	btrfs_stop_workers(&fs_info->delalloc_workers);
@@ -2005,6 +2020,17 @@ struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
 	return latest;
 }
 
+/*
+ * this should be called twice, once with wait == 0 and
+ * once with wait == 1.  When wait == 0 is done, all the buffer heads
+ * we write are pinned.
+ *
+ * They are released when wait == 1 is done.
+ * max_mirrors must be the same for both runs, and it indicates how
+ * many supers on this one device should be written.
+ *
+ * max_mirrors == 0 means to write them all.
+ */
 static int write_dev_supers(struct btrfs_device *device,
 			    struct btrfs_super_block *sb,
 			    int do_barriers, int wait, int max_mirrors)
@@ -2040,12 +2066,16 @@ static int write_dev_supers(struct btrfs_device *device,
 			bh = __find_get_block(device->bdev, bytenr / 4096,
 					      BTRFS_SUPER_INFO_SIZE);
 			BUG_ON(!bh);
-			brelse(bh);
 			wait_on_buffer(bh);
-			if (buffer_uptodate(bh)) {
-				brelse(bh);
-				continue;
-			}
+			if (!buffer_uptodate(bh))
+				errors++;
+
+			/* drop our reference */
+			brelse(bh);
+
+			/* drop the reference from the wait == 0 run */
+			brelse(bh);
+			continue;
 		} else {
 			btrfs_set_super_bytenr(sb, bytenr);
 
@@ -2056,12 +2086,18 @@ static int write_dev_supers(struct btrfs_device *device,
 					      BTRFS_CSUM_SIZE);
 			btrfs_csum_final(crc, sb->csum);
 
+			/*
+			 * one reference for us, and we leave it for the
+			 * caller
+			 */
 			bh = __getblk(device->bdev, bytenr / 4096,
 				      BTRFS_SUPER_INFO_SIZE);
 			memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
 
-			set_buffer_uptodate(bh);
+			/* one reference for submit_bh */
 			get_bh(bh);
+
+			set_buffer_uptodate(bh);
 			lock_buffer(bh);
 			bh->b_end_io = btrfs_end_buffer_write_sync;
 		}
@@ -2073,6 +2109,7 @@ static int write_dev_supers(struct btrfs_device *device,
 				       device->name);
 				set_buffer_uptodate(bh);
 				device->barriers = 0;
+				/* one reference for submit_bh */
 				get_bh(bh);
 				lock_buffer(bh);
 				ret = submit_bh(WRITE_SYNC, bh);
@@ -2081,22 +2118,15 @@ static int write_dev_supers(struct btrfs_device *device,
 			ret = submit_bh(WRITE_SYNC, bh);
 		}
 
-		if (!ret && wait) {
-			wait_on_buffer(bh);
-			if (!buffer_uptodate(bh))
-				errors++;
-		} else if (ret) {
+		if (ret)
 			errors++;
-		}
-		if (wait)
-			brelse(bh);
 	}
 	return errors < i ? 0 : -1;
 }
 
 int write_all_supers(struct btrfs_root *root, int max_mirrors)
 {
-	struct list_head *head = &root->fs_info->fs_devices->devices;
+	struct list_head *head;
 	struct btrfs_device *dev;
 	struct btrfs_super_block *sb;
 	struct btrfs_dev_item *dev_item;
@@ -2111,6 +2141,9 @@ int write_all_supers(struct btrfs_root *root, int max_mirrors)
 
 	sb = &root->fs_info->super_for_commit;
 	dev_item = &sb->dev_item;
+
+	mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
+	head = &root->fs_info->fs_devices->devices;
 	list_for_each_entry(dev, head, dev_list) {
 		if (!dev->bdev) {
 			total_errors++;
@@ -2154,6 +2187,7 @@ int write_all_supers(struct btrfs_root *root, int max_mirrors)
 		if (ret)
 			total_errors++;
 	}
+	mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
 	if (total_errors > max_errors) {
 		printk(KERN_ERR "btrfs: %d errors while writing supers\n",
 		       total_errors);
@@ -2173,6 +2207,7 @@ int write_ctree_super(struct btrfs_trans_handle *trans,
 
 int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
 {
+	WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
 	radix_tree_delete(&fs_info->fs_roots_radix,
 			  (unsigned long)root->root_key.objectid);
 	if (root->anon_super.s_dev) {
@@ -2219,10 +2254,12 @@ int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
 					     ARRAY_SIZE(gang));
 		if (!ret)
 			break;
+
+		root_objectid = gang[ret - 1]->root_key.objectid + 1;
 		for (i = 0; i < ret; i++) {
 			root_objectid = gang[i]->root_key.objectid;
 			ret = btrfs_find_dead_roots(fs_info->tree_root,
-						    root_objectid, gang[i]);
+						    root_objectid);
 			BUG_ON(ret);
 			btrfs_orphan_cleanup(gang[i]);
 		}
@@ -2278,20 +2315,16 @@ int close_ctree(struct btrfs_root *root)
 		       (unsigned long long)fs_info->total_ref_cache_size);
 	}
 
-	if (fs_info->extent_root->node)
-		free_extent_buffer(fs_info->extent_root->node);
-
-	if (fs_info->tree_root->node)
-		free_extent_buffer(fs_info->tree_root->node);
-
-	if (root->fs_info->chunk_root->node)
-		free_extent_buffer(root->fs_info->chunk_root->node);
-
-	if (root->fs_info->dev_root->node)
-		free_extent_buffer(root->fs_info->dev_root->node);
-
-	if (root->fs_info->csum_root->node)
-		free_extent_buffer(root->fs_info->csum_root->node);
+	free_extent_buffer(fs_info->extent_root->node);
+	free_extent_buffer(fs_info->extent_root->commit_root);
+	free_extent_buffer(fs_info->tree_root->node);
+	free_extent_buffer(fs_info->tree_root->commit_root);
+	free_extent_buffer(root->fs_info->chunk_root->node);
+	free_extent_buffer(root->fs_info->chunk_root->commit_root);
+	free_extent_buffer(root->fs_info->dev_root->node);
+	free_extent_buffer(root->fs_info->dev_root->commit_root);
+	free_extent_buffer(root->fs_info->csum_root->node);
+	free_extent_buffer(root->fs_info->csum_root->commit_root);
 
 	btrfs_free_block_groups(root->fs_info);
 
@@ -2373,17 +2406,14 @@ void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
 	 * looks as though older kernels can get into trouble with
 	 * this code, they end up stuck in balance_dirty_pages forever
 	 */
-	struct extent_io_tree *tree;
 	u64 num_dirty;
-	u64 start = 0;
 	unsigned long thresh = 32 * 1024 * 1024;
-	tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
 
 	if (current->flags & PF_MEMALLOC)
 		return;
 
-	num_dirty = count_range_bits(tree, &start, (u64)-1,
-				     thresh, EXTENT_DIRTY);
+	num_dirty = root->fs_info->dirty_metadata_bytes;
+
 	if (num_dirty > thresh) {
 		balance_dirty_pages_ratelimited_nr(
 				   root->fs_info->btree_inode->i_mapping, 1);

fs/btrfs/export.c

@@ -78,7 +78,7 @@ static struct dentry *btrfs_get_dentry(struct super_block *sb, u64 objectid,
 	btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
 	key.offset = 0;
 
-	inode = btrfs_iget(sb, &key, root, NULL);
+	inode = btrfs_iget(sb, &key, root);
 	if (IS_ERR(inode))
 		return (void *)inode;
 
@@ -192,7 +192,7 @@ static struct dentry *btrfs_get_parent(struct dentry *child)
 	btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
 	key.offset = 0;
 
-	return d_obtain_alias(btrfs_iget(root->fs_info->sb, &key, root, NULL));
+	return d_obtain_alias(btrfs_iget(root->fs_info->sb, &key, root));
 }
 
 const struct export_operations btrfs_export_ops = {

fs/btrfs/extent_io.c

@@ -476,6 +476,7 @@ int clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
 	struct extent_state *state;
 	struct extent_state *prealloc = NULL;
 	struct rb_node *node;
+	u64 last_end;
 	int err;
 	int set = 0;
 
@@ -498,6 +499,7 @@ again:
 	if (state->start > end)
 		goto out;
 	WARN_ON(state->end < start);
+	last_end = state->end;
 
 	/*
 	 *     | ---- desired range ---- |
@@ -524,9 +526,11 @@ again:
 		if (err)
 			goto out;
 		if (state->end <= end) {
-			start = state->end + 1;
 			set |= clear_state_bit(tree, state, bits,
 					wake, delete);
+			if (last_end == (u64)-1)
+				goto out;
+			start = last_end + 1;
 		} else {
 			start = state->start;
 		}
@@ -552,8 +556,10 @@ again:
 		goto out;
 	}
 
-	start = state->end + 1;
 	set |= clear_state_bit(tree, state, bits, wake, delete);
+	if (last_end == (u64)-1)
+		goto out;
+	start = last_end + 1;
 	goto search_again;
 
 out:
@@ -707,8 +713,10 @@ again:
 			goto out;
 		}
 		set_state_bits(tree, state, bits);
-		start = state->end + 1;
 		merge_state(tree, state);
+		if (last_end == (u64)-1)
+			goto out;
+		start = last_end + 1;
 		goto search_again;
 	}
 
@@ -742,8 +750,10 @@ again:
 			goto out;
 		if (state->end <= end) {
 			set_state_bits(tree, state, bits);
-			start = state->end + 1;
 			merge_state(tree, state);
+			if (last_end == (u64)-1)
+				goto out;
+			start = last_end + 1;
 		} else {
 			start = state->start;
 		}

fs/btrfs/file.c

@@ -291,16 +291,12 @@ noinline int btrfs_drop_extents(struct btrfs_trans_handle *trans,
 {
 	u64 extent_end = 0;
 	u64 search_start = start;
-	u64 leaf_start;
 	u64 ram_bytes = 0;
-	u64 orig_parent = 0;
 	u64 disk_bytenr = 0;
 	u64 orig_locked_end = locked_end;
 	u8 compression;
 	u8 encryption;
 	u16 other_encoding = 0;
-	u64 root_gen;
-	u64 root_owner;
 	struct extent_buffer *leaf;
 	struct btrfs_file_extent_item *extent;
 	struct btrfs_path *path;
@@ -340,9 +336,6 @@ next_slot:
 		bookend = 0;
 		found_extent = 0;
 		found_inline = 0;
-		leaf_start = 0;
-		root_gen = 0;
-		root_owner = 0;
 		compression = 0;
 		encryption = 0;
 		extent = NULL;
@@ -417,9 +410,6 @@ next_slot:
 		if (found_extent) {
 			read_extent_buffer(leaf, &old, (unsigned long)extent,
 					   sizeof(old));
-			root_gen = btrfs_header_generation(leaf);
-			root_owner = btrfs_header_owner(leaf);
-			leaf_start = leaf->start;
 		}
 
 		if (end < extent_end && end >= key.offset) {
@@ -443,14 +433,14 @@ next_slot:
 				}
 				locked_end = extent_end;
 			}
-			orig_parent = path->nodes[0]->start;
 			disk_bytenr = le64_to_cpu(old.disk_bytenr);
 			if (disk_bytenr != 0) {
 				ret = btrfs_inc_extent_ref(trans, root,
 					   disk_bytenr,
-					   le64_to_cpu(old.disk_num_bytes),
-					   orig_parent, root->root_key.objectid,
-					   trans->transid, inode->i_ino);
+					   le64_to_cpu(old.disk_num_bytes), 0,
+					   root->root_key.objectid,
+					   key.objectid, key.offset -
+					   le64_to_cpu(old.offset));
 				BUG_ON(ret);
 			}
 		}
@@ -568,17 +558,6 @@ next_slot:
 			btrfs_mark_buffer_dirty(path->nodes[0]);
 			btrfs_set_lock_blocking(path->nodes[0]);
 
-			if (disk_bytenr != 0) {
-				ret = btrfs_update_extent_ref(trans, root,
-						disk_bytenr,
-						le64_to_cpu(old.disk_num_bytes),
-						orig_parent,
-						leaf->start,
-						root->root_key.objectid,
-						trans->transid, ins.objectid);
-
-				BUG_ON(ret);
-			}
 			path->leave_spinning = 0;
 			btrfs_release_path(root, path);
 			if (disk_bytenr != 0)
@@ -594,8 +573,9 @@ next_slot:
 				ret = btrfs_free_extent(trans, root,
 						old_disk_bytenr,
 						le64_to_cpu(old.disk_num_bytes),
-						leaf_start, root_owner,
-						root_gen, key.objectid, 0);
+						0, root->root_key.objectid,
+						key.objectid, key.offset -
+						le64_to_cpu(old.offset));
 				BUG_ON(ret);
 				*hint_byte = old_disk_bytenr;
 			}
@@ -664,12 +644,11 @@ int btrfs_mark_extent_written(struct btrfs_trans_handle *trans,
 	u64 bytenr;
 	u64 num_bytes;
 	u64 extent_end;
-	u64 extent_offset;
+	u64 orig_offset;
 	u64 other_start;
 	u64 other_end;
 	u64 split = start;
 	u64 locked_end = end;
-	u64 orig_parent;
 	int extent_type;
 	int split_end = 1;
 	int ret;
@@ -703,7 +682,7 @@ again:
 
 	bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
 	num_bytes = btrfs_file_extent_disk_num_bytes(leaf, fi);
-	extent_offset = btrfs_file_extent_offset(leaf, fi);
+	orig_offset = key.offset - btrfs_file_extent_offset(leaf, fi);
 
 	if (key.offset == start)
 		split = end;
@@ -711,8 +690,6 @@ again:
 	if (key.offset == start && extent_end == end) {
 		int del_nr = 0;
 		int del_slot = 0;
-		u64 leaf_owner = btrfs_header_owner(leaf);
-		u64 leaf_gen = btrfs_header_generation(leaf);
 		other_start = end;
 		other_end = 0;
 		if (extent_mergeable(leaf, path->slots[0] + 1, inode->i_ino,
@@ -721,8 +698,8 @@ again:
 			del_slot = path->slots[0] + 1;
 			del_nr++;
 			ret = btrfs_free_extent(trans, root, bytenr, num_bytes,
-						leaf->start, leaf_owner,
-						leaf_gen, inode->i_ino, 0);
+						0, root->root_key.objectid,
+						inode->i_ino, orig_offset);
 			BUG_ON(ret);
 		}
 		other_start = 0;
@@ -733,8 +710,8 @@ again:
 			del_slot = path->slots[0];
 			del_nr++;
 			ret = btrfs_free_extent(trans, root, bytenr, num_bytes,
-						leaf->start, leaf_owner,
-						leaf_gen, inode->i_ino, 0);
+						0, root->root_key.objectid,
+						inode->i_ino, orig_offset);
 			BUG_ON(ret);
 		}
 		split_end = 0;
@@ -768,13 +745,12 @@ again:
 			locked_end = extent_end;
 		}
 		btrfs_set_file_extent_num_bytes(leaf, fi, split - key.offset);
-		extent_offset += split - key.offset;
 	} else  {
 		BUG_ON(key.offset != start);
-		btrfs_set_file_extent_offset(leaf, fi, extent_offset +
-					     split - key.offset);
-		btrfs_set_file_extent_num_bytes(leaf, fi, extent_end - split);
 		key.offset = split;
+		btrfs_set_file_extent_offset(leaf, fi, key.offset -
+					     orig_offset);
+		btrfs_set_file_extent_num_bytes(leaf, fi, extent_end - split);
 		btrfs_set_item_key_safe(trans, root, path, &key);
 		extent_end = split;
 	}
@@ -793,7 +769,8 @@ again:
 					    struct btrfs_file_extent_item);
 			key.offset = split;
 			btrfs_set_item_key_safe(trans, root, path, &key);
-			btrfs_set_file_extent_offset(leaf, fi, extent_offset);
+			btrfs_set_file_extent_offset(leaf, fi, key.offset -
+						     orig_offset);
 			btrfs_set_file_extent_num_bytes(leaf, fi,
 							other_end - split);
 			goto done;
@@ -815,10 +792,9 @@ again:
 
 	btrfs_mark_buffer_dirty(leaf);
 
-	orig_parent = leaf->start;
-	ret = btrfs_inc_extent_ref(trans, root, bytenr, num_bytes,
-				   orig_parent, root->root_key.objectid,
-				   trans->transid, inode->i_ino);
+	ret = btrfs_inc_extent_ref(trans, root, bytenr, num_bytes, 0,
+				   root->root_key.objectid,
+				   inode->i_ino, orig_offset);
 	BUG_ON(ret);
 	btrfs_release_path(root, path);
 
@@ -833,20 +809,12 @@ again:
 	btrfs_set_file_extent_type(leaf, fi, extent_type);
 	btrfs_set_file_extent_disk_bytenr(leaf, fi, bytenr);
 	btrfs_set_file_extent_disk_num_bytes(leaf, fi, num_bytes);
-	btrfs_set_file_extent_offset(leaf, fi, extent_offset);
+	btrfs_set_file_extent_offset(leaf, fi, key.offset - orig_offset);
 	btrfs_set_file_extent_num_bytes(leaf, fi, extent_end - key.offset);
 	btrfs_set_file_extent_ram_bytes(leaf, fi, num_bytes);
 	btrfs_set_file_extent_compression(leaf, fi, 0);
 	btrfs_set_file_extent_encryption(leaf, fi, 0);
 	btrfs_set_file_extent_other_encoding(leaf, fi, 0);
-
-	if (orig_parent != leaf->start) {
-		ret = btrfs_update_extent_ref(trans, root, bytenr, num_bytes,
-					      orig_parent, leaf->start,
-					      root->root_key.objectid,
-					      trans->transid, inode->i_ino);
-		BUG_ON(ret);
-	}
 done:
 	btrfs_mark_buffer_dirty(leaf);
 
@@ -1189,6 +1157,8 @@ int btrfs_sync_file(struct file *file, struct dentry *dentry, int datasync)
 	btrfs_wait_ordered_range(inode, 0, (u64)-1);
 	root->log_batch++;
 
+	if (datasync && !(inode->i_state & I_DIRTY_PAGES))
+		goto out;
 	/*
 	 * ok we haven't committed the transaction yet, lets do a commit
 	 */

fs/btrfs/free-space-cache.c

@@ -579,6 +579,7 @@ out:
  * it returns -enospc
  */
 int btrfs_find_space_cluster(struct btrfs_trans_handle *trans,
+			     struct btrfs_root *root,
 			     struct btrfs_block_group_cache *block_group,
 			     struct btrfs_free_cluster *cluster,
 			     u64 offset, u64 bytes, u64 empty_size)
@@ -595,7 +596,9 @@ int btrfs_find_space_cluster(struct btrfs_trans_handle *trans,
 	int ret;
 
 	/* for metadata, allow allocates with more holes */
-	if (block_group->flags & BTRFS_BLOCK_GROUP_METADATA) {
+	if (btrfs_test_opt(root, SSD_SPREAD)) {
+		min_bytes = bytes + empty_size;
+	} else if (block_group->flags & BTRFS_BLOCK_GROUP_METADATA) {
 		/*
 		 * we want to do larger allocations when we are
 		 * flushing out the delayed refs, it helps prevent
@@ -645,14 +648,15 @@ again:
 		 * we haven't filled the empty size and the window is
 		 * very large.  reset and try again
 		 */
-		if (next->offset - window_start > (bytes + empty_size) * 2) {
+		if (next->offset - (last->offset + last->bytes) > 128 * 1024 ||
+		    next->offset - window_start > (bytes + empty_size) * 2) {
 			entry = next;
 			window_start = entry->offset;
 			window_free = entry->bytes;
 			last = entry;
 			max_extent = 0;
 			total_retries++;
-			if (total_retries % 256 == 0) {
+			if (total_retries % 64 == 0) {
 				if (min_bytes >= (bytes + empty_size)) {
 					ret = -ENOSPC;
 					goto out;

fs/btrfs/free-space-cache.h

@@ -31,6 +31,7 @@ void btrfs_dump_free_space(struct btrfs_block_group_cache *block_group,
 			   u64 bytes);
 u64 btrfs_block_group_free_space(struct btrfs_block_group_cache *block_group);
 int btrfs_find_space_cluster(struct btrfs_trans_handle *trans,
+			     struct btrfs_root *root,
 			     struct btrfs_block_group_cache *block_group,
 			     struct btrfs_free_cluster *cluster,
 			     u64 offset, u64 bytes, u64 empty_size);

fs/btrfs/hash.h

@@ -19,9 +19,9 @@
 #ifndef __HASH__
 #define __HASH__
 
-#include "crc32c.h"
+#include <linux/crc32c.h>
 static inline u64 btrfs_name_hash(const char *name, int len)
 {
-	return btrfs_crc32c((u32)~1, name, len);
+	return crc32c((u32)~1, name, len);
 }
 #endif

fs/btrfs/inode.c

@@ -48,7 +48,6 @@
 #include "ordered-data.h"
 #include "xattr.h"
 #include "tree-log.h"
-#include "ref-cache.h"
 #include "compression.h"
 #include "locking.h"
 
@@ -369,7 +368,7 @@ again:
 	 * inode has not been flagged as nocompress.  This flag can
 	 * change at any time if we discover bad compression ratios.
 	 */
-	if (!btrfs_test_flag(inode, NOCOMPRESS) &&
+	if (!(BTRFS_I(inode)->flags & BTRFS_INODE_NOCOMPRESS) &&
 	    btrfs_test_opt(root, COMPRESS)) {
 		WARN_ON(pages);
 		pages = kzalloc(sizeof(struct page *) * nr_pages, GFP_NOFS);
@@ -470,7 +469,7 @@ again:
 		nr_pages_ret = 0;
 
 		/* flag the file so we don't compress in the future */
-		btrfs_set_flag(inode, NOCOMPRESS);
+		BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
 	}
 	if (will_compress) {
 		*num_added += 1;
@@ -863,7 +862,7 @@ static int cow_file_range_async(struct inode *inode, struct page *locked_page,
 		async_cow->locked_page = locked_page;
 		async_cow->start = start;
 
-		if (btrfs_test_flag(inode, NOCOMPRESS))
+		if (BTRFS_I(inode)->flags & BTRFS_INODE_NOCOMPRESS)
 			cur_end = end;
 		else
 			cur_end = min(end, start + 512 * 1024 - 1);
@@ -944,6 +943,7 @@ static noinline int run_delalloc_nocow(struct inode *inode,
 	u64 cow_start;
 	u64 cur_offset;
 	u64 extent_end;
+	u64 extent_offset;
 	u64 disk_bytenr;
 	u64 num_bytes;
 	int extent_type;
@@ -1005,6 +1005,7 @@ next_slot:
 		if (extent_type == BTRFS_FILE_EXTENT_REG ||
 		    extent_type == BTRFS_FILE_EXTENT_PREALLOC) {
 			disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
+			extent_offset = btrfs_file_extent_offset(leaf, fi);
 			extent_end = found_key.offset +
 				btrfs_file_extent_num_bytes(leaf, fi);
 			if (extent_end <= start) {
@@ -1022,9 +1023,10 @@ next_slot:
 			if (btrfs_extent_readonly(root, disk_bytenr))
 				goto out_check;
 			if (btrfs_cross_ref_exist(trans, root, inode->i_ino,
-						  disk_bytenr))
+						  found_key.offset -
+						  extent_offset, disk_bytenr))
 				goto out_check;
-			disk_bytenr += btrfs_file_extent_offset(leaf, fi);
+			disk_bytenr += extent_offset;
 			disk_bytenr += cur_offset - found_key.offset;
 			num_bytes = min(end + 1, extent_end) - cur_offset;
 			/*
@@ -1131,10 +1133,10 @@ static int run_delalloc_range(struct inode *inode, struct page *locked_page,
 	int ret;
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 
-	if (btrfs_test_flag(inode, NODATACOW))
+	if (BTRFS_I(inode)->flags & BTRFS_INODE_NODATACOW)
 		ret = run_delalloc_nocow(inode, locked_page, start, end,
 					 page_started, 1, nr_written);
-	else if (btrfs_test_flag(inode, PREALLOC))
+	else if (BTRFS_I(inode)->flags & BTRFS_INODE_PREALLOC)
 		ret = run_delalloc_nocow(inode, locked_page, start, end,
 					 page_started, 0, nr_written);
 	else if (!btrfs_test_opt(root, COMPRESS))
@@ -1288,7 +1290,7 @@ static int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
 	int ret = 0;
 	int skip_sum;
 
-	skip_sum = btrfs_test_flag(inode, NODATASUM);
+	skip_sum = BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM;
 
 	ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0);
 	BUG_ON(ret);
@@ -1489,9 +1491,9 @@ static int insert_reserved_file_extent(struct btrfs_trans_handle *trans,
 	ins.objectid = disk_bytenr;
 	ins.offset = disk_num_bytes;
 	ins.type = BTRFS_EXTENT_ITEM_KEY;
-	ret = btrfs_alloc_reserved_extent(trans, root, leaf->start,
-					  root->root_key.objectid,
-					  trans->transid, inode->i_ino, &ins);
+	ret = btrfs_alloc_reserved_file_extent(trans, root,
+					root->root_key.objectid,
+					inode->i_ino, file_pos, &ins);
 	BUG_ON(ret);
 	btrfs_free_path(path);
 
@@ -1788,7 +1790,8 @@ static int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end,
 		ClearPageChecked(page);
 		goto good;
 	}
-	if (btrfs_test_flag(inode, NODATASUM))
+
+	if (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)
 		return 0;
 
 	if (root->root_key.objectid == BTRFS_DATA_RELOC_TREE_OBJECTID &&
@@ -1956,23 +1959,13 @@ void btrfs_orphan_cleanup(struct btrfs_root *root)
 		 * crossing root thing.  we store the inode number in the
 		 * offset of the orphan item.
 		 */
-		inode = btrfs_iget_locked(root->fs_info->sb,
-					  found_key.offset, root);
-		if (!inode)
+		found_key.objectid = found_key.offset;
+		found_key.type = BTRFS_INODE_ITEM_KEY;
+		found_key.offset = 0;
+		inode = btrfs_iget(root->fs_info->sb, &found_key, root);
+		if (IS_ERR(inode))
 			break;
 
-		if (inode->i_state & I_NEW) {
-			BTRFS_I(inode)->root = root;
-
-			/* have to set the location manually */
-			BTRFS_I(inode)->location.objectid = inode->i_ino;
-			BTRFS_I(inode)->location.type = BTRFS_INODE_ITEM_KEY;
-			BTRFS_I(inode)->location.offset = 0;
-
-			btrfs_read_locked_inode(inode);
-			unlock_new_inode(inode);
-		}
-
 		/*
 		 * add this inode to the orphan list so btrfs_orphan_del does
 		 * the proper thing when we hit it
@@ -2069,7 +2062,7 @@ static noinline int acls_after_inode_item(struct extent_buffer *leaf,
 /*
  * read an inode from the btree into the in-memory inode
  */
-void btrfs_read_locked_inode(struct inode *inode)
+static void btrfs_read_locked_inode(struct inode *inode)
 {
 	struct btrfs_path *path;
 	struct extent_buffer *leaf;
@@ -2164,6 +2157,8 @@ void btrfs_read_locked_inode(struct inode *inode)
 		init_special_inode(inode, inode->i_mode, rdev);
 		break;
 	}
+
+	btrfs_update_iflags(inode);
 	return;
 
 make_bad:
@@ -2599,9 +2594,8 @@ noinline int btrfs_truncate_inode_items(struct btrfs_trans_handle *trans,
 	struct btrfs_file_extent_item *fi;
 	u64 extent_start = 0;
 	u64 extent_num_bytes = 0;
+	u64 extent_offset = 0;
 	u64 item_end = 0;
-	u64 root_gen = 0;
-	u64 root_owner = 0;
 	int found_extent;
 	int del_item;
 	int pending_del_nr = 0;
@@ -2716,6 +2710,9 @@ search_again:
 				extent_num_bytes =
 					btrfs_file_extent_disk_num_bytes(leaf,
 									 fi);
+				extent_offset = found_key.offset -
+					btrfs_file_extent_offset(leaf, fi);
+
 				/* FIXME blocksize != 4096 */
 				num_dec = btrfs_file_extent_num_bytes(leaf, fi);
 				if (extent_start != 0) {
@@ -2723,8 +2720,6 @@ search_again:
 					if (root->ref_cows)
 						inode_sub_bytes(inode, num_dec);
 				}
-				root_gen = btrfs_header_generation(leaf);
-				root_owner = btrfs_header_owner(leaf);
 			}
 		} else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
 			/*
@@ -2768,12 +2763,12 @@ delete:
 		} else {
 			break;
 		}
-		if (found_extent) {
+		if (found_extent && root->ref_cows) {
 			btrfs_set_path_blocking(path);
 			ret = btrfs_free_extent(trans, root, extent_start,
-						extent_num_bytes,
-						leaf->start, root_owner,
-						root_gen, inode->i_ino, 0);
+						extent_num_bytes, 0,
+						btrfs_header_owner(leaf),
+						inode->i_ino, extent_offset);
 			BUG_ON(ret);
 		}
 next:
@@ -3105,6 +3100,45 @@ static int fixup_tree_root_location(struct btrfs_root *root,
 	return 0;
 }
 
+static void inode_tree_add(struct inode *inode)
+{
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct btrfs_inode *entry;
+	struct rb_node **p = &root->inode_tree.rb_node;
+	struct rb_node *parent = NULL;
+
+	spin_lock(&root->inode_lock);
+	while (*p) {
+		parent = *p;
+		entry = rb_entry(parent, struct btrfs_inode, rb_node);
+
+		if (inode->i_ino < entry->vfs_inode.i_ino)
+			p = &(*p)->rb_left;
+		else if (inode->i_ino > entry->vfs_inode.i_ino)
+			p = &(*p)->rb_right;
+		else {
+			WARN_ON(!(entry->vfs_inode.i_state &
+				  (I_WILL_FREE | I_FREEING | I_CLEAR)));
+			break;
+		}
+	}
+	rb_link_node(&BTRFS_I(inode)->rb_node, parent, p);
+	rb_insert_color(&BTRFS_I(inode)->rb_node, &root->inode_tree);
+	spin_unlock(&root->inode_lock);
+}
+
+static void inode_tree_del(struct inode *inode)
+{
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+
+	if (!RB_EMPTY_NODE(&BTRFS_I(inode)->rb_node)) {
+		spin_lock(&root->inode_lock);
+		rb_erase(&BTRFS_I(inode)->rb_node, &root->inode_tree);
+		spin_unlock(&root->inode_lock);
+		RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node);
+	}
+}
+
 static noinline void init_btrfs_i(struct inode *inode)
 {
 	struct btrfs_inode *bi = BTRFS_I(inode);
@@ -3130,6 +3164,7 @@ static noinline void init_btrfs_i(struct inode *inode)
 			     inode->i_mapping, GFP_NOFS);
 	INIT_LIST_HEAD(&BTRFS_I(inode)->delalloc_inodes);
 	INIT_LIST_HEAD(&BTRFS_I(inode)->ordered_operations);
+	RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node);
 	btrfs_ordered_inode_tree_init(&BTRFS_I(inode)->ordered_tree);
 	mutex_init(&BTRFS_I(inode)->extent_mutex);
 	mutex_init(&BTRFS_I(inode)->log_mutex);
@@ -3152,26 +3187,9 @@ static int btrfs_find_actor(struct inode *inode, void *opaque)
 		args->root == BTRFS_I(inode)->root;
 }
 
-struct inode *btrfs_ilookup(struct super_block *s, u64 objectid,
-			    struct btrfs_root *root, int wait)
-{
-	struct inode *inode;
-	struct btrfs_iget_args args;
-	args.ino = objectid;
-	args.root = root;
-
-	if (wait) {
-		inode = ilookup5(s, objectid, btrfs_find_actor,
-				 (void *)&args);
-	} else {
-		inode = ilookup5_nowait(s, objectid, btrfs_find_actor,
-					(void *)&args);
-	}
-	return inode;
-}
-
-struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid,
-				struct btrfs_root *root)
+static struct inode *btrfs_iget_locked(struct super_block *s,
+				       u64 objectid,
+				       struct btrfs_root *root)
 {
 	struct inode *inode;
 	struct btrfs_iget_args args;
@@ -3188,24 +3206,21 @@ struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid,
  * Returns in *is_new if the inode was read from disk
  */
 struct inode *btrfs_iget(struct super_block *s, struct btrfs_key *location,
-			 struct btrfs_root *root, int *is_new)
+			 struct btrfs_root *root)
 {
 	struct inode *inode;
 
 	inode = btrfs_iget_locked(s, location->objectid, root);
 	if (!inode)
-		return ERR_PTR(-EACCES);
+		return ERR_PTR(-ENOMEM);
 
 	if (inode->i_state & I_NEW) {
 		BTRFS_I(inode)->root = root;
 		memcpy(&BTRFS_I(inode)->location, location, sizeof(*location));
 		btrfs_read_locked_inode(inode);
+
+		inode_tree_add(inode);
 		unlock_new_inode(inode);
-		if (is_new)
-			*is_new = 1;
-	} else {
-		if (is_new)
-			*is_new = 0;
 	}
 
 	return inode;
@@ -3218,7 +3233,7 @@ struct inode *btrfs_lookup_dentry(struct inode *dir, struct dentry *dentry)
 	struct btrfs_root *root = bi->root;
 	struct btrfs_root *sub_root = root;
 	struct btrfs_key location;
-	int ret, new;
+	int ret;
 
 	if (dentry->d_name.len > BTRFS_NAME_LEN)
 		return ERR_PTR(-ENAMETOOLONG);
@@ -3236,7 +3251,7 @@ struct inode *btrfs_lookup_dentry(struct inode *dir, struct dentry *dentry)
 			return ERR_PTR(ret);
 		if (ret > 0)
 			return ERR_PTR(-ENOENT);
-		inode = btrfs_iget(dir->i_sb, &location, sub_root, &new);
+		inode = btrfs_iget(dir->i_sb, &location, sub_root);
 		if (IS_ERR(inode))
 			return ERR_CAST(inode);
 	}
@@ -3574,9 +3589,9 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
 			btrfs_find_block_group(root, 0, alloc_hint, owner);
 	if ((mode & S_IFREG)) {
 		if (btrfs_test_opt(root, NODATASUM))
-			btrfs_set_flag(inode, NODATASUM);
+			BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
 		if (btrfs_test_opt(root, NODATACOW))
-			btrfs_set_flag(inode, NODATACOW);
+			BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
 	}
 
 	key[0].objectid = objectid;
@@ -3630,7 +3645,10 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
 	location->offset = 0;
 	btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
 
+	btrfs_inherit_iflags(inode, dir);
+
 	insert_inode_hash(inode);
+	inode_tree_add(inode);
 	return inode;
 fail:
 	if (dir)
@@ -4683,6 +4701,7 @@ void btrfs_destroy_inode(struct inode *inode)
 			btrfs_put_ordered_extent(ordered);
 		}
 	}
+	inode_tree_del(inode);
 	btrfs_drop_extent_cache(inode, 0, (u64)-1, 0);
 	kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
 }
@@ -5061,7 +5080,7 @@ static int prealloc_file_range(struct btrfs_trans_handle *trans,
 out:
 	if (cur_offset > start) {
 		inode->i_ctime = CURRENT_TIME;
-		btrfs_set_flag(inode, PREALLOC);
+		BTRFS_I(inode)->flags |= BTRFS_INODE_PREALLOC;
 		if (!(mode & FALLOC_FL_KEEP_SIZE) &&
 		    cur_offset > i_size_read(inode))
 			btrfs_i_size_write(inode, cur_offset);
@@ -5182,7 +5201,7 @@ static int btrfs_set_page_dirty(struct page *page)
 
 static int btrfs_permission(struct inode *inode, int mask)
 {
-	if (btrfs_test_flag(inode, READONLY) && (mask & MAY_WRITE))
+	if ((BTRFS_I(inode)->flags & BTRFS_INODE_READONLY) && (mask & MAY_WRITE))
 		return -EACCES;
 	return generic_permission(inode, mask, btrfs_check_acl);
 }

fs/btrfs/ioctl.c

@@ -50,7 +50,177 @@
 #include "volumes.h"
 #include "locking.h"
 
+/* Mask out flags that are inappropriate for the given type of inode. */
+static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
+{
+	if (S_ISDIR(mode))
+		return flags;
+	else if (S_ISREG(mode))
+		return flags & ~FS_DIRSYNC_FL;
+	else
+		return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
+}
+
+/*
+ * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
+ */
+static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
+{
+	unsigned int iflags = 0;
+
+	if (flags & BTRFS_INODE_SYNC)
+		iflags |= FS_SYNC_FL;
+	if (flags & BTRFS_INODE_IMMUTABLE)
+		iflags |= FS_IMMUTABLE_FL;
+	if (flags & BTRFS_INODE_APPEND)
+		iflags |= FS_APPEND_FL;
+	if (flags & BTRFS_INODE_NODUMP)
+		iflags |= FS_NODUMP_FL;
+	if (flags & BTRFS_INODE_NOATIME)
+		iflags |= FS_NOATIME_FL;
+	if (flags & BTRFS_INODE_DIRSYNC)
+		iflags |= FS_DIRSYNC_FL;
+
+	return iflags;
+}
+
+/*
+ * Update inode->i_flags based on the btrfs internal flags.
+ */
+void btrfs_update_iflags(struct inode *inode)
+{
+	struct btrfs_inode *ip = BTRFS_I(inode);
+
+	inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
+
+	if (ip->flags & BTRFS_INODE_SYNC)
+		inode->i_flags |= S_SYNC;
+	if (ip->flags & BTRFS_INODE_IMMUTABLE)
+		inode->i_flags |= S_IMMUTABLE;
+	if (ip->flags & BTRFS_INODE_APPEND)
+		inode->i_flags |= S_APPEND;
+	if (ip->flags & BTRFS_INODE_NOATIME)
+		inode->i_flags |= S_NOATIME;
+	if (ip->flags & BTRFS_INODE_DIRSYNC)
+		inode->i_flags |= S_DIRSYNC;
+}
+
+/*
+ * Inherit flags from the parent inode.
+ *
+ * Unlike extN we don't have any flags we don't want to inherit currently.
+ */
+void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
+{
+	unsigned int flags;
+
+	if (!dir)
+		return;
+
+	flags = BTRFS_I(dir)->flags;
+
+	if (S_ISREG(inode->i_mode))
+		flags &= ~BTRFS_INODE_DIRSYNC;
+	else if (!S_ISDIR(inode->i_mode))
+		flags &= (BTRFS_INODE_NODUMP | BTRFS_INODE_NOATIME);
+
+	BTRFS_I(inode)->flags = flags;
+	btrfs_update_iflags(inode);
+}
+
+static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
+{
+	struct btrfs_inode *ip = BTRFS_I(file->f_path.dentry->d_inode);
+	unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
+
+	if (copy_to_user(arg, &flags, sizeof(flags)))
+		return -EFAULT;
+	return 0;
+}
+
+static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
+{
+	struct inode *inode = file->f_path.dentry->d_inode;
+	struct btrfs_inode *ip = BTRFS_I(inode);
+	struct btrfs_root *root = ip->root;
+	struct btrfs_trans_handle *trans;
+	unsigned int flags, oldflags;
+	int ret;
+
+	if (copy_from_user(&flags, arg, sizeof(flags)))
+		return -EFAULT;
+
+	if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
+		      FS_NOATIME_FL | FS_NODUMP_FL | \
+		      FS_SYNC_FL | FS_DIRSYNC_FL))
+		return -EOPNOTSUPP;
 
+	if (!is_owner_or_cap(inode))
+		return -EACCES;
+
+	mutex_lock(&inode->i_mutex);
+
+	flags = btrfs_mask_flags(inode->i_mode, flags);
+	oldflags = btrfs_flags_to_ioctl(ip->flags);
+	if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
+		if (!capable(CAP_LINUX_IMMUTABLE)) {
+			ret = -EPERM;
+			goto out_unlock;
+		}
+	}
+
+	ret = mnt_want_write(file->f_path.mnt);
+	if (ret)
+		goto out_unlock;
+
+	if (flags & FS_SYNC_FL)
+		ip->flags |= BTRFS_INODE_SYNC;
+	else
+		ip->flags &= ~BTRFS_INODE_SYNC;
+	if (flags & FS_IMMUTABLE_FL)
+		ip->flags |= BTRFS_INODE_IMMUTABLE;
+	else
+		ip->flags &= ~BTRFS_INODE_IMMUTABLE;
+	if (flags & FS_APPEND_FL)
+		ip->flags |= BTRFS_INODE_APPEND;
+	else
+		ip->flags &= ~BTRFS_INODE_APPEND;
+	if (flags & FS_NODUMP_FL)
+		ip->flags |= BTRFS_INODE_NODUMP;
+	else
+		ip->flags &= ~BTRFS_INODE_NODUMP;
+	if (flags & FS_NOATIME_FL)
+		ip->flags |= BTRFS_INODE_NOATIME;
+	else
+		ip->flags &= ~BTRFS_INODE_NOATIME;
+	if (flags & FS_DIRSYNC_FL)
+		ip->flags |= BTRFS_INODE_DIRSYNC;
+	else
+		ip->flags &= ~BTRFS_INODE_DIRSYNC;
+
+
+	trans = btrfs_join_transaction(root, 1);
+	BUG_ON(!trans);
+
+	ret = btrfs_update_inode(trans, root, inode);
+	BUG_ON(ret);
+
+	btrfs_update_iflags(inode);
+	inode->i_ctime = CURRENT_TIME;
+	btrfs_end_transaction(trans, root);
+
+	mnt_drop_write(file->f_path.mnt);
+ out_unlock:
+	mutex_unlock(&inode->i_mutex);
+	return 0;
+}
+
+static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
+{
+	struct inode *inode = file->f_path.dentry->d_inode;
+
+	return put_user(inode->i_generation, arg);
+}
 
 static noinline int create_subvol(struct btrfs_root *root,
 				  struct dentry *dentry,
@@ -82,22 +252,25 @@ static noinline int create_subvol(struct btrfs_root *root,
 	if (ret)
 		goto fail;
 
-	leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0,
-				      objectid, trans->transid, 0, 0, 0);
+	leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
+				      0, objectid, NULL, 0, 0, 0);
 	if (IS_ERR(leaf)) {
 		ret = PTR_ERR(leaf);
 		goto fail;
 	}
 
-	btrfs_set_header_nritems(leaf, 0);
-	btrfs_set_header_level(leaf, 0);
+	memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
 	btrfs_set_header_bytenr(leaf, leaf->start);
 	btrfs_set_header_generation(leaf, trans->transid);
+	btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
 	btrfs_set_header_owner(leaf, objectid);
 
 	write_extent_buffer(leaf, root->fs_info->fsid,
 			    (unsigned long)btrfs_header_fsid(leaf),
 			    BTRFS_FSID_SIZE);
+	write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
+			    (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
+			    BTRFS_UUID_SIZE);
 	btrfs_mark_buffer_dirty(leaf);
 
 	inode_item = &root_item.inode;
@@ -125,7 +298,7 @@ static noinline int create_subvol(struct btrfs_root *root,
 	btrfs_set_root_dirid(&root_item, new_dirid);
 
 	key.objectid = objectid;
-	key.offset = 1;
+	key.offset = 0;
 	btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
 	ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
 				&root_item);
@@ -911,10 +1084,10 @@ static long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
 				if (disko) {
 					inode_add_bytes(inode, datal);
 					ret = btrfs_inc_extent_ref(trans, root,
-						   disko, diskl, leaf->start,
-						   root->root_key.objectid,
-						   trans->transid,
-						   inode->i_ino);
+							disko, diskl, 0,
+							root->root_key.objectid,
+							inode->i_ino,
+							new_key.offset - datao);
 					BUG_ON(ret);
 				}
 			} else if (type == BTRFS_FILE_EXTENT_INLINE) {
@@ -1074,6 +1247,12 @@ long btrfs_ioctl(struct file *file, unsigned int
 	void __user *argp = (void __user *)arg;
 
 	switch (cmd) {
+	case FS_IOC_GETFLAGS:
+		return btrfs_ioctl_getflags(file, argp);
+	case FS_IOC_SETFLAGS:
+		return btrfs_ioctl_setflags(file, argp);
+	case FS_IOC_GETVERSION:
+		return btrfs_ioctl_getversion(file, argp);
 	case BTRFS_IOC_SNAP_CREATE:
 		return btrfs_ioctl_snap_create(file, argp, 0);
 	case BTRFS_IOC_SUBVOL_CREATE:

fs/btrfs/print-tree.c

@@ -45,22 +45,132 @@ static void print_dev_item(struct extent_buffer *eb,
 	       (unsigned long long)btrfs_device_total_bytes(eb, dev_item),
 	       (unsigned long long)btrfs_device_bytes_used(eb, dev_item));
 }
+static void print_extent_data_ref(struct extent_buffer *eb,
+				  struct btrfs_extent_data_ref *ref)
+{
+	printk(KERN_INFO "\t\textent data backref root %llu "
+	       "objectid %llu offset %llu count %u\n",
+	       (unsigned long long)btrfs_extent_data_ref_root(eb, ref),
+	       (unsigned long long)btrfs_extent_data_ref_objectid(eb, ref),
+	       (unsigned long long)btrfs_extent_data_ref_offset(eb, ref),
+	       btrfs_extent_data_ref_count(eb, ref));
+}
+
+static void print_extent_item(struct extent_buffer *eb, int slot)
+{
+	struct btrfs_extent_item *ei;
+	struct btrfs_extent_inline_ref *iref;
+	struct btrfs_extent_data_ref *dref;
+	struct btrfs_shared_data_ref *sref;
+	struct btrfs_disk_key key;
+	unsigned long end;
+	unsigned long ptr;
+	int type;
+	u32 item_size = btrfs_item_size_nr(eb, slot);
+	u64 flags;
+	u64 offset;
+
+	if (item_size < sizeof(*ei)) {
+#ifdef BTRFS_COMPAT_EXTENT_TREE_V0
+		struct btrfs_extent_item_v0 *ei0;
+		BUG_ON(item_size != sizeof(*ei0));
+		ei0 = btrfs_item_ptr(eb, slot, struct btrfs_extent_item_v0);
+		printk(KERN_INFO "\t\textent refs %u\n",
+		       btrfs_extent_refs_v0(eb, ei0));
+		return;
+#else
+		BUG();
+#endif
+	}
+
+	ei = btrfs_item_ptr(eb, slot, struct btrfs_extent_item);
+	flags = btrfs_extent_flags(eb, ei);
+
+	printk(KERN_INFO "\t\textent refs %llu gen %llu flags %llu\n",
+	       (unsigned long long)btrfs_extent_refs(eb, ei),
+	       (unsigned long long)btrfs_extent_generation(eb, ei),
+	       (unsigned long long)flags);
+
+	if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
+		struct btrfs_tree_block_info *info;
+		info = (struct btrfs_tree_block_info *)(ei + 1);
+		btrfs_tree_block_key(eb, info, &key);
+		printk(KERN_INFO "\t\ttree block key (%llu %x %llu) "
+		       "level %d\n",
+		       (unsigned long long)btrfs_disk_key_objectid(&key),
+		       key.type,
+		       (unsigned long long)btrfs_disk_key_offset(&key),
+		       btrfs_tree_block_level(eb, info));
+		iref = (struct btrfs_extent_inline_ref *)(info + 1);
+	} else {
+		iref = (struct btrfs_extent_inline_ref *)(ei + 1);
+	}
+
+	ptr = (unsigned long)iref;
+	end = (unsigned long)ei + item_size;
+	while (ptr < end) {
+		iref = (struct btrfs_extent_inline_ref *)ptr;
+		type = btrfs_extent_inline_ref_type(eb, iref);
+		offset = btrfs_extent_inline_ref_offset(eb, iref);
+		switch (type) {
+		case BTRFS_TREE_BLOCK_REF_KEY:
+			printk(KERN_INFO "\t\ttree block backref "
+				"root %llu\n", (unsigned long long)offset);
+			break;
+		case BTRFS_SHARED_BLOCK_REF_KEY:
+			printk(KERN_INFO "\t\tshared block backref "
+				"parent %llu\n", (unsigned long long)offset);
+			break;
+		case BTRFS_EXTENT_DATA_REF_KEY:
+			dref = (struct btrfs_extent_data_ref *)(&iref->offset);
+			print_extent_data_ref(eb, dref);
+			break;
+		case BTRFS_SHARED_DATA_REF_KEY:
+			sref = (struct btrfs_shared_data_ref *)(iref + 1);
+			printk(KERN_INFO "\t\tshared data backref "
+			       "parent %llu count %u\n",
+			       (unsigned long long)offset,
+			       btrfs_shared_data_ref_count(eb, sref));
+			break;
+		default:
+			BUG();
+		}
+		ptr += btrfs_extent_inline_ref_size(type);
+	}
+	WARN_ON(ptr > end);
+}
+
+#ifdef BTRFS_COMPAT_EXTENT_TREE_V0
+static void print_extent_ref_v0(struct extent_buffer *eb, int slot)
+{
+	struct btrfs_extent_ref_v0 *ref0;
+
+	ref0 = btrfs_item_ptr(eb, slot, struct btrfs_extent_ref_v0);
+	printk("\t\textent back ref root %llu gen %llu "
+		"owner %llu num_refs %lu\n",
+		(unsigned long long)btrfs_ref_root_v0(eb, ref0),
+		(unsigned long long)btrfs_ref_generation_v0(eb, ref0),
+		(unsigned long long)btrfs_ref_objectid_v0(eb, ref0),
+		(unsigned long)btrfs_ref_count_v0(eb, ref0));
+}
+#endif
+
 void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l)
 {
 	int i;
+	u32 type;
 	u32 nr = btrfs_header_nritems(l);
 	struct btrfs_item *item;
-	struct btrfs_extent_item *ei;
 	struct btrfs_root_item *ri;
 	struct btrfs_dir_item *di;
 	struct btrfs_inode_item *ii;
 	struct btrfs_block_group_item *bi;
 	struct btrfs_file_extent_item *fi;
+	struct btrfs_extent_data_ref *dref;
+	struct btrfs_shared_data_ref *sref;
+	struct btrfs_dev_extent *dev_extent;
 	struct btrfs_key key;
 	struct btrfs_key found_key;
-	struct btrfs_extent_ref *ref;
-	struct btrfs_dev_extent *dev_extent;
-	u32 type;
 
 	printk(KERN_INFO "leaf %llu total ptrs %d free space %d\n",
 		(unsigned long long)btrfs_header_bytenr(l), nr,
@@ -100,20 +210,25 @@ void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l)
 				btrfs_disk_root_refs(l, ri));
 			break;
 		case BTRFS_EXTENT_ITEM_KEY:
-			ei = btrfs_item_ptr(l, i, struct btrfs_extent_item);
-			printk(KERN_INFO "\t\textent data refs %u\n",
-				btrfs_extent_refs(l, ei));
-			break;
-		case BTRFS_EXTENT_REF_KEY:
-			ref = btrfs_item_ptr(l, i, struct btrfs_extent_ref);
-			printk(KERN_INFO "\t\textent back ref root %llu "
-			       "gen %llu owner %llu num_refs %lu\n",
-			       (unsigned long long)btrfs_ref_root(l, ref),
-			       (unsigned long long)btrfs_ref_generation(l, ref),
-			       (unsigned long long)btrfs_ref_objectid(l, ref),
-			       (unsigned long)btrfs_ref_num_refs(l, ref));
+			print_extent_item(l, i);
+			break;
+		case BTRFS_TREE_BLOCK_REF_KEY:
+			printk(KERN_INFO "\t\ttree block backref\n");
+			break;
+		case BTRFS_SHARED_BLOCK_REF_KEY:
+			printk(KERN_INFO "\t\tshared block backref\n");
+			break;
+		case BTRFS_EXTENT_DATA_REF_KEY:
+			dref = btrfs_item_ptr(l, i,
+					      struct btrfs_extent_data_ref);
+			print_extent_data_ref(l, dref);
+			break;
+		case BTRFS_SHARED_DATA_REF_KEY:
+			sref = btrfs_item_ptr(l, i,
+					      struct btrfs_shared_data_ref);
+			printk(KERN_INFO "\t\tshared data backref count %u\n",
+			       btrfs_shared_data_ref_count(l, sref));
 			break;
-
 		case BTRFS_EXTENT_DATA_KEY:
 			fi = btrfs_item_ptr(l, i,
 					    struct btrfs_file_extent_item);
@@ -139,6 +254,12 @@ void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l)
 			       (unsigned long long)
 			       btrfs_file_extent_ram_bytes(l, fi));
 			break;
+		case BTRFS_EXTENT_REF_V0_KEY:
+#ifdef BTRFS_COMPAT_EXTENT_TREE_V0
+			print_extent_ref_v0(l, i);
+#else
+			BUG();
+#endif
 		case BTRFS_BLOCK_GROUP_ITEM_KEY:
 			bi = btrfs_item_ptr(l, i,
 					    struct btrfs_block_group_item);

fs/btrfs/relocation.c

@@ -0,0 +1,3711 @@
+/*
+ * Copyright (C) 2009 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/sched.h>
+#include <linux/pagemap.h>
+#include <linux/writeback.h>
+#include <linux/blkdev.h>
+#include <linux/rbtree.h>
+#include "ctree.h"
+#include "disk-io.h"
+#include "transaction.h"
+#include "volumes.h"
+#include "locking.h"
+#include "btrfs_inode.h"
+#include "async-thread.h"
+
+/*
+ * backref_node, mapping_node and tree_block start with this
+ */
+struct tree_entry {
+	struct rb_node rb_node;
+	u64 bytenr;
+};
+
+/*
+ * present a tree block in the backref cache
+ */
+struct backref_node {
+	struct rb_node rb_node;
+	u64 bytenr;
+	/* objectid tree block owner */
+	u64 owner;
+	/* list of upper level blocks reference this block */
+	struct list_head upper;
+	/* list of child blocks in the cache */
+	struct list_head lower;
+	/* NULL if this node is not tree root */
+	struct btrfs_root *root;
+	/* extent buffer got by COW the block */
+	struct extent_buffer *eb;
+	/* level of tree block */
+	unsigned int level:8;
+	/* 1 if the block is root of old snapshot */
+	unsigned int old_root:1;
+	/* 1 if no child blocks in the cache */
+	unsigned int lowest:1;
+	/* is the extent buffer locked */
+	unsigned int locked:1;
+	/* has the block been processed */
+	unsigned int processed:1;
+	/* have backrefs of this block been checked */
+	unsigned int checked:1;
+};
+
+/*
+ * present a block pointer in the backref cache
+ */
+struct backref_edge {
+	struct list_head list[2];
+	struct backref_node *node[2];
+	u64 blockptr;
+};
+
+#define LOWER	0
+#define UPPER	1
+
+struct backref_cache {
+	/* red black tree of all backref nodes in the cache */
+	struct rb_root rb_root;
+	/* list of backref nodes with no child block in the cache */
+	struct list_head pending[BTRFS_MAX_LEVEL];
+	spinlock_t lock;
+};
+
+/*
+ * map address of tree root to tree
+ */
+struct mapping_node {
+	struct rb_node rb_node;
+	u64 bytenr;
+	void *data;
+};
+
+struct mapping_tree {
+	struct rb_root rb_root;
+	spinlock_t lock;
+};
+
+/*
+ * present a tree block to process
+ */
+struct tree_block {
+	struct rb_node rb_node;
+	u64 bytenr;
+	struct btrfs_key key;
+	unsigned int level:8;
+	unsigned int key_ready:1;
+};
+
+/* inode vector */
+#define INODEVEC_SIZE 16
+
+struct inodevec {
+	struct list_head list;
+	struct inode *inode[INODEVEC_SIZE];
+	int nr;
+};
+
+struct reloc_control {
+	/* block group to relocate */
+	struct btrfs_block_group_cache *block_group;
+	/* extent tree */
+	struct btrfs_root *extent_root;
+	/* inode for moving data */
+	struct inode *data_inode;
+	struct btrfs_workers workers;
+	/* tree blocks have been processed */
+	struct extent_io_tree processed_blocks;
+	/* map start of tree root to corresponding reloc tree */
+	struct mapping_tree reloc_root_tree;
+	/* list of reloc trees */
+	struct list_head reloc_roots;
+	u64 search_start;
+	u64 extents_found;
+	u64 extents_skipped;
+	int stage;
+	int create_reloc_root;
+	unsigned int found_file_extent:1;
+	unsigned int found_old_snapshot:1;
+};
+
+/* stages of data relocation */
+#define MOVE_DATA_EXTENTS	0
+#define UPDATE_DATA_PTRS	1
+
+/*
+ * merge reloc tree to corresponding fs tree in worker threads
+ */
+struct async_merge {
+	struct btrfs_work work;
+	struct reloc_control *rc;
+	struct btrfs_root *root;
+	struct completion *done;
+	atomic_t *num_pending;
+};
+
+static void mapping_tree_init(struct mapping_tree *tree)
+{
+	tree->rb_root.rb_node = NULL;
+	spin_lock_init(&tree->lock);
+}
+
+static void backref_cache_init(struct backref_cache *cache)
+{
+	int i;
+	cache->rb_root.rb_node = NULL;
+	for (i = 0; i < BTRFS_MAX_LEVEL; i++)
+		INIT_LIST_HEAD(&cache->pending[i]);
+	spin_lock_init(&cache->lock);
+}
+
+static void backref_node_init(struct backref_node *node)
+{
+	memset(node, 0, sizeof(*node));
+	INIT_LIST_HEAD(&node->upper);
+	INIT_LIST_HEAD(&node->lower);
+	RB_CLEAR_NODE(&node->rb_node);
+}
+
+static struct rb_node *tree_insert(struct rb_root *root, u64 bytenr,
+				   struct rb_node *node)
+{
+	struct rb_node **p = &root->rb_node;
+	struct rb_node *parent = NULL;
+	struct tree_entry *entry;
+
+	while (*p) {
+		parent = *p;
+		entry = rb_entry(parent, struct tree_entry, rb_node);
+
+		if (bytenr < entry->bytenr)
+			p = &(*p)->rb_left;
+		else if (bytenr > entry->bytenr)
+			p = &(*p)->rb_right;
+		else
+			return parent;
+	}
+
+	rb_link_node(node, parent, p);
+	rb_insert_color(node, root);
+	return NULL;
+}
+
+static struct rb_node *tree_search(struct rb_root *root, u64 bytenr)
+{
+	struct rb_node *n = root->rb_node;
+	struct tree_entry *entry;
+
+	while (n) {
+		entry = rb_entry(n, struct tree_entry, rb_node);
+
+		if (bytenr < entry->bytenr)
+			n = n->rb_left;
+		else if (bytenr > entry->bytenr)
+			n = n->rb_right;
+		else
+			return n;
+	}
+	return NULL;
+}
+
+/*
+ * walk up backref nodes until reach node presents tree root
+ */
+static struct backref_node *walk_up_backref(struct backref_node *node,
+					    struct backref_edge *edges[],
+					    int *index)
+{
+	struct backref_edge *edge;
+	int idx = *index;
+
+	while (!list_empty(&node->upper)) {
+		edge = list_entry(node->upper.next,
+				  struct backref_edge, list[LOWER]);
+		edges[idx++] = edge;
+		node = edge->node[UPPER];
+	}
+	*index = idx;
+	return node;
+}
+
+/*
+ * walk down backref nodes to find start of next reference path
+ */
+static struct backref_node *walk_down_backref(struct backref_edge *edges[],
+					      int *index)
+{
+	struct backref_edge *edge;
+	struct backref_node *lower;
+	int idx = *index;
+
+	while (idx > 0) {
+		edge = edges[idx - 1];
+		lower = edge->node[LOWER];
+		if (list_is_last(&edge->list[LOWER], &lower->upper)) {
+			idx--;
+			continue;
+		}
+		edge = list_entry(edge->list[LOWER].next,
+				  struct backref_edge, list[LOWER]);
+		edges[idx - 1] = edge;
+		*index = idx;
+		return edge->node[UPPER];
+	}
+	*index = 0;
+	return NULL;
+}
+
+static void drop_node_buffer(struct backref_node *node)
+{
+	if (node->eb) {
+		if (node->locked) {
+			btrfs_tree_unlock(node->eb);
+			node->locked = 0;
+		}
+		free_extent_buffer(node->eb);
+		node->eb = NULL;
+	}
+}
+
+static void drop_backref_node(struct backref_cache *tree,
+			      struct backref_node *node)
+{
+	BUG_ON(!node->lowest);
+	BUG_ON(!list_empty(&node->upper));
+
+	drop_node_buffer(node);
+	list_del(&node->lower);
+
+	rb_erase(&node->rb_node, &tree->rb_root);
+	kfree(node);
+}
+
+/*
+ * remove a backref node from the backref cache
+ */
+static void remove_backref_node(struct backref_cache *cache,
+				struct backref_node *node)
+{
+	struct backref_node *upper;
+	struct backref_edge *edge;
+
+	if (!node)
+		return;
+
+	BUG_ON(!node->lowest);
+	while (!list_empty(&node->upper)) {
+		edge = list_entry(node->upper.next, struct backref_edge,
+				  list[LOWER]);
+		upper = edge->node[UPPER];
+		list_del(&edge->list[LOWER]);
+		list_del(&edge->list[UPPER]);
+		kfree(edge);
+		/*
+		 * add the node to pending list if no other
+		 * child block cached.
+		 */
+		if (list_empty(&upper->lower)) {
+			list_add_tail(&upper->lower,
+				      &cache->pending[upper->level]);
+			upper->lowest = 1;
+		}
+	}
+	drop_backref_node(cache, node);
+}
+
+/*
+ * find reloc tree by address of tree root
+ */
+static struct btrfs_root *find_reloc_root(struct reloc_control *rc,
+					  u64 bytenr)
+{
+	struct rb_node *rb_node;
+	struct mapping_node *node;
+	struct btrfs_root *root = NULL;
+
+	spin_lock(&rc->reloc_root_tree.lock);
+	rb_node = tree_search(&rc->reloc_root_tree.rb_root, bytenr);
+	if (rb_node) {
+		node = rb_entry(rb_node, struct mapping_node, rb_node);
+		root = (struct btrfs_root *)node->data;
+	}
+	spin_unlock(&rc->reloc_root_tree.lock);
+	return root;
+}
+
+static int is_cowonly_root(u64 root_objectid)
+{
+	if (root_objectid == BTRFS_ROOT_TREE_OBJECTID ||
+	    root_objectid == BTRFS_EXTENT_TREE_OBJECTID ||
+	    root_objectid == BTRFS_CHUNK_TREE_OBJECTID ||
+	    root_objectid == BTRFS_DEV_TREE_OBJECTID ||
+	    root_objectid == BTRFS_TREE_LOG_OBJECTID ||
+	    root_objectid == BTRFS_CSUM_TREE_OBJECTID)
+		return 1;
+	return 0;
+}
+
+static struct btrfs_root *read_fs_root(struct btrfs_fs_info *fs_info,
+					u64 root_objectid)
+{
+	struct btrfs_key key;
+
+	key.objectid = root_objectid;
+	key.type = BTRFS_ROOT_ITEM_KEY;
+	if (is_cowonly_root(root_objectid))
+		key.offset = 0;
+	else
+		key.offset = (u64)-1;
+
+	return btrfs_read_fs_root_no_name(fs_info, &key);
+}
+
+#ifdef BTRFS_COMPAT_EXTENT_TREE_V0
+static noinline_for_stack
+struct btrfs_root *find_tree_root(struct reloc_control *rc,
+				  struct extent_buffer *leaf,
+				  struct btrfs_extent_ref_v0 *ref0)
+{
+	struct btrfs_root *root;
+	u64 root_objectid = btrfs_ref_root_v0(leaf, ref0);
+	u64 generation = btrfs_ref_generation_v0(leaf, ref0);
+
+	BUG_ON(root_objectid == BTRFS_TREE_RELOC_OBJECTID);
+
+	root = read_fs_root(rc->extent_root->fs_info, root_objectid);
+	BUG_ON(IS_ERR(root));
+
+	if (root->ref_cows &&
+	    generation != btrfs_root_generation(&root->root_item))
+		return NULL;
+
+	return root;
+}
+#endif
+
+static noinline_for_stack
+int find_inline_backref(struct extent_buffer *leaf, int slot,
+			unsigned long *ptr, unsigned long *end)
+{
+	struct btrfs_extent_item *ei;
+	struct btrfs_tree_block_info *bi;
+	u32 item_size;
+
+	item_size = btrfs_item_size_nr(leaf, slot);
+#ifdef BTRFS_COMPAT_EXTENT_TREE_V0
+	if (item_size < sizeof(*ei)) {
+		WARN_ON(item_size != sizeof(struct btrfs_extent_item_v0));
+		return 1;
+	}
+#endif
+	ei = btrfs_item_ptr(leaf, slot, struct btrfs_extent_item);
+	WARN_ON(!(btrfs_extent_flags(leaf, ei) &
+		  BTRFS_EXTENT_FLAG_TREE_BLOCK));
+
+	if (item_size <= sizeof(*ei) + sizeof(*bi)) {
+		WARN_ON(item_size < sizeof(*ei) + sizeof(*bi));
+		return 1;
+	}
+
+	bi = (struct btrfs_tree_block_info *)(ei + 1);
+	*ptr = (unsigned long)(bi + 1);
+	*end = (unsigned long)ei + item_size;
+	return 0;
+}
+
+/*
+ * build backref tree for a given tree block. root of the backref tree
+ * corresponds the tree block, leaves of the backref tree correspond
+ * roots of b-trees that reference the tree block.
+ *
+ * the basic idea of this function is check backrefs of a given block
+ * to find upper level blocks that refernece the block, and then check
+ * bakcrefs of these upper level blocks recursively. the recursion stop
+ * when tree root is reached or backrefs for the block is cached.
+ *
+ * NOTE: if we find backrefs for a block are cached, we know backrefs
+ * for all upper level blocks that directly/indirectly reference the
+ * block are also cached.
+ */
+static struct backref_node *build_backref_tree(struct reloc_control *rc,
+					       struct backref_cache *cache,
+					       struct btrfs_key *node_key,
+					       int level, u64 bytenr)
+{
+	struct btrfs_path *path1;
+	struct btrfs_path *path2;
+	struct extent_buffer *eb;
+	struct btrfs_root *root;
+	struct backref_node *cur;
+	struct backref_node *upper;
+	struct backref_node *lower;
+	struct backref_node *node = NULL;
+	struct backref_node *exist = NULL;
+	struct backref_edge *edge;
+	struct rb_node *rb_node;
+	struct btrfs_key key;
+	unsigned long end;
+	unsigned long ptr;
+	LIST_HEAD(list);
+	int ret;
+	int err = 0;
+
+	path1 = btrfs_alloc_path();
+	path2 = btrfs_alloc_path();
+	if (!path1 || !path2) {
+		err = -ENOMEM;
+		goto out;
+	}
+
+	node = kmalloc(sizeof(*node), GFP_NOFS);
+	if (!node) {
+		err = -ENOMEM;
+		goto out;
+	}
+
+	backref_node_init(node);
+	node->bytenr = bytenr;
+	node->owner = 0;
+	node->level = level;
+	node->lowest = 1;
+	cur = node;
+again:
+	end = 0;
+	ptr = 0;
+	key.objectid = cur->bytenr;
+	key.type = BTRFS_EXTENT_ITEM_KEY;
+	key.offset = (u64)-1;
+
+	path1->search_commit_root = 1;
+	path1->skip_locking = 1;
+	ret = btrfs_search_slot(NULL, rc->extent_root, &key, path1,
+				0, 0);
+	if (ret < 0) {
+		err = ret;
+		goto out;
+	}
+	BUG_ON(!ret || !path1->slots[0]);
+
+	path1->slots[0]--;
+
+	WARN_ON(cur->checked);
+	if (!list_empty(&cur->upper)) {
+		/*
+		 * the backref was added previously when processsing
+		 * backref of type BTRFS_TREE_BLOCK_REF_KEY
+		 */
+		BUG_ON(!list_is_singular(&cur->upper));
+		edge = list_entry(cur->upper.next, struct backref_edge,
+				  list[LOWER]);
+		BUG_ON(!list_empty(&edge->list[UPPER]));
+		exist = edge->node[UPPER];
+		/*
+		 * add the upper level block to pending list if we need
+		 * check its backrefs
+		 */
+		if (!exist->checked)
+			list_add_tail(&edge->list[UPPER], &list);
+	} else {
+		exist = NULL;
+	}
+
+	while (1) {
+		cond_resched();
+		eb = path1->nodes[0];
+
+		if (ptr >= end) {
+			if (path1->slots[0] >= btrfs_header_nritems(eb)) {
+				ret = btrfs_next_leaf(rc->extent_root, path1);
+				if (ret < 0) {
+					err = ret;
+					goto out;
+				}
+				if (ret > 0)
+					break;
+				eb = path1->nodes[0];
+			}
+
+			btrfs_item_key_to_cpu(eb, &key, path1->slots[0]);
+			if (key.objectid != cur->bytenr) {
+				WARN_ON(exist);
+				break;
+			}
+
+			if (key.type == BTRFS_EXTENT_ITEM_KEY) {
+				ret = find_inline_backref(eb, path1->slots[0],
+							  &ptr, &end);
+				if (ret)
+					goto next;
+			}
+		}
+
+		if (ptr < end) {
+			/* update key for inline back ref */
+			struct btrfs_extent_inline_ref *iref;
+			iref = (struct btrfs_extent_inline_ref *)ptr;
+			key.type = btrfs_extent_inline_ref_type(eb, iref);
+			key.offset = btrfs_extent_inline_ref_offset(eb, iref);
+			WARN_ON(key.type != BTRFS_TREE_BLOCK_REF_KEY &&
+				key.type != BTRFS_SHARED_BLOCK_REF_KEY);
+		}
+
+		if (exist &&
+		    ((key.type == BTRFS_TREE_BLOCK_REF_KEY &&
+		      exist->owner == key.offset) ||
+		     (key.type == BTRFS_SHARED_BLOCK_REF_KEY &&
+		      exist->bytenr == key.offset))) {
+			exist = NULL;
+			goto next;
+		}
+
+#ifdef BTRFS_COMPAT_EXTENT_TREE_V0
+		if (key.type == BTRFS_SHARED_BLOCK_REF_KEY ||
+		    key.type == BTRFS_EXTENT_REF_V0_KEY) {
+			if (key.objectid == key.offset &&
+			    key.type == BTRFS_EXTENT_REF_V0_KEY) {
+				struct btrfs_extent_ref_v0 *ref0;
+				ref0 = btrfs_item_ptr(eb, path1->slots[0],
+						struct btrfs_extent_ref_v0);
+				root = find_tree_root(rc, eb, ref0);
+				if (root)
+					cur->root = root;
+				else
+					cur->old_root = 1;
+				break;
+			}
+#else
+		BUG_ON(key.type == BTRFS_EXTENT_REF_V0_KEY);
+		if (key.type == BTRFS_SHARED_BLOCK_REF_KEY) {
+#endif
+			if (key.objectid == key.offset) {
+				/*
+				 * only root blocks of reloc trees use
+				 * backref of this type.
+				 */
+				root = find_reloc_root(rc, cur->bytenr);
+				BUG_ON(!root);
+				cur->root = root;
+				break;
+			}
+
+			edge = kzalloc(sizeof(*edge), GFP_NOFS);
+			if (!edge) {
+				err = -ENOMEM;
+				goto out;
+			}
+			rb_node = tree_search(&cache->rb_root, key.offset);
+			if (!rb_node) {
+				upper = kmalloc(sizeof(*upper), GFP_NOFS);
+				if (!upper) {
+					kfree(edge);
+					err = -ENOMEM;
+					goto out;
+				}
+				backref_node_init(upper);
+				upper->bytenr = key.offset;
+				upper->owner = 0;
+				upper->level = cur->level + 1;
+				/*
+				 *  backrefs for the upper level block isn't
+				 *  cached, add the block to pending list
+				 */
+				list_add_tail(&edge->list[UPPER], &list);
+			} else {
+				upper = rb_entry(rb_node, struct backref_node,
+						 rb_node);
+				INIT_LIST_HEAD(&edge->list[UPPER]);
+			}
+			list_add(&edge->list[LOWER], &cur->upper);
+			edge->node[UPPER] = upper;
+			edge->node[LOWER] = cur;
+
+			goto next;
+		} else if (key.type != BTRFS_TREE_BLOCK_REF_KEY) {
+			goto next;
+		}
+
+		/* key.type == BTRFS_TREE_BLOCK_REF_KEY */
+		root = read_fs_root(rc->extent_root->fs_info, key.offset);
+		if (IS_ERR(root)) {
+			err = PTR_ERR(root);
+			goto out;
+		}
+
+		if (btrfs_root_level(&root->root_item) == cur->level) {
+			/* tree root */
+			BUG_ON(btrfs_root_bytenr(&root->root_item) !=
+			       cur->bytenr);
+			cur->root = root;
+			break;
+		}
+
+		level = cur->level + 1;
+
+		/*
+		 * searching the tree to find upper level blocks
+		 * reference the block.
+		 */
+		path2->search_commit_root = 1;
+		path2->skip_locking = 1;
+		path2->lowest_level = level;
+		ret = btrfs_search_slot(NULL, root, node_key, path2, 0, 0);
+		path2->lowest_level = 0;
+		if (ret < 0) {
+			err = ret;
+			goto out;
+		}
+
+		eb = path2->nodes[level];
+		WARN_ON(btrfs_node_blockptr(eb, path2->slots[level]) !=
+			cur->bytenr);
+
+		lower = cur;
+		for (; level < BTRFS_MAX_LEVEL; level++) {
+			if (!path2->nodes[level]) {
+				BUG_ON(btrfs_root_bytenr(&root->root_item) !=
+				       lower->bytenr);
+				lower->root = root;
+				break;
+			}
+
+			edge = kzalloc(sizeof(*edge), GFP_NOFS);
+			if (!edge) {
+				err = -ENOMEM;
+				goto out;
+			}
+
+			eb = path2->nodes[level];
+			rb_node = tree_search(&cache->rb_root, eb->start);
+			if (!rb_node) {
+				upper = kmalloc(sizeof(*upper), GFP_NOFS);
+				if (!upper) {
+					kfree(edge);
+					err = -ENOMEM;
+					goto out;
+				}
+				backref_node_init(upper);
+				upper->bytenr = eb->start;
+				upper->owner = btrfs_header_owner(eb);
+				upper->level = lower->level + 1;
+
+				/*
+				 * if we know the block isn't shared
+				 * we can void checking its backrefs.
+				 */
+				if (btrfs_block_can_be_shared(root, eb))
+					upper->checked = 0;
+				else
+					upper->checked = 1;
+
+				/*
+				 * add the block to pending list if we
+				 * need check its backrefs. only block
+				 * at 'cur->level + 1' is added to the
+				 * tail of pending list. this guarantees
+				 * we check backrefs from lower level
+				 * blocks to upper level blocks.
+				 */
+				if (!upper->checked &&
+				    level == cur->level + 1) {
+					list_add_tail(&edge->list[UPPER],
+						      &list);
+				} else
+					INIT_LIST_HEAD(&edge->list[UPPER]);
+			} else {
+				upper = rb_entry(rb_node, struct backref_node,
+						 rb_node);
+				BUG_ON(!upper->checked);
+				INIT_LIST_HEAD(&edge->list[UPPER]);
+			}
+			list_add_tail(&edge->list[LOWER], &lower->upper);
+			edge->node[UPPER] = upper;
+			edge->node[LOWER] = lower;
+
+			if (rb_node)
+				break;
+			lower = upper;
+			upper = NULL;
+		}
+		btrfs_release_path(root, path2);
+next:
+		if (ptr < end) {
+			ptr += btrfs_extent_inline_ref_size(key.type);
+			if (ptr >= end) {
+				WARN_ON(ptr > end);
+				ptr = 0;
+				end = 0;
+			}
+		}
+		if (ptr >= end)
+			path1->slots[0]++;
+	}
+	btrfs_release_path(rc->extent_root, path1);
+
+	cur->checked = 1;
+	WARN_ON(exist);
+
+	/* the pending list isn't empty, take the first block to process */
+	if (!list_empty(&list)) {
+		edge = list_entry(list.next, struct backref_edge, list[UPPER]);
+		list_del_init(&edge->list[UPPER]);
+		cur = edge->node[UPPER];
+		goto again;
+	}
+
+	/*
+	 * everything goes well, connect backref nodes and insert backref nodes
+	 * into the cache.
+	 */
+	BUG_ON(!node->checked);
+	rb_node = tree_insert(&cache->rb_root, node->bytenr, &node->rb_node);
+	BUG_ON(rb_node);
+
+	list_for_each_entry(edge, &node->upper, list[LOWER])
+		list_add_tail(&edge->list[UPPER], &list);
+
+	while (!list_empty(&list)) {
+		edge = list_entry(list.next, struct backref_edge, list[UPPER]);
+		list_del_init(&edge->list[UPPER]);
+		upper = edge->node[UPPER];
+
+		if (!RB_EMPTY_NODE(&upper->rb_node)) {
+			if (upper->lowest) {
+				list_del_init(&upper->lower);
+				upper->lowest = 0;
+			}
+
+			list_add_tail(&edge->list[UPPER], &upper->lower);
+			continue;
+		}
+
+		BUG_ON(!upper->checked);
+		rb_node = tree_insert(&cache->rb_root, upper->bytenr,
+				      &upper->rb_node);
+		BUG_ON(rb_node);
+
+		list_add_tail(&edge->list[UPPER], &upper->lower);
+
+		list_for_each_entry(edge, &upper->upper, list[LOWER])
+			list_add_tail(&edge->list[UPPER], &list);
+	}
+out:
+	btrfs_free_path(path1);
+	btrfs_free_path(path2);
+	if (err) {
+		INIT_LIST_HEAD(&list);
+		upper = node;
+		while (upper) {
+			if (RB_EMPTY_NODE(&upper->rb_node)) {
+				list_splice_tail(&upper->upper, &list);
+				kfree(upper);
+			}
+
+			if (list_empty(&list))
+				break;
+
+			edge = list_entry(list.next, struct backref_edge,
+					  list[LOWER]);
+			upper = edge->node[UPPER];
+			kfree(edge);
+		}
+		return ERR_PTR(err);
+	}
+	return node;
+}
+
+/*
+ * helper to add 'address of tree root -> reloc tree' mapping
+ */
+static int __add_reloc_root(struct btrfs_root *root)
+{
+	struct rb_node *rb_node;
+	struct mapping_node *node;
+	struct reloc_control *rc = root->fs_info->reloc_ctl;
+
+	node = kmalloc(sizeof(*node), GFP_NOFS);
+	BUG_ON(!node);
+
+	node->bytenr = root->node->start;
+	node->data = root;
+
+	spin_lock(&rc->reloc_root_tree.lock);
+	rb_node = tree_insert(&rc->reloc_root_tree.rb_root,
+			      node->bytenr, &node->rb_node);
+	spin_unlock(&rc->reloc_root_tree.lock);
+	BUG_ON(rb_node);
+
+	list_add_tail(&root->root_list, &rc->reloc_roots);
+	return 0;
+}
+
+/*
+ * helper to update/delete the 'address of tree root -> reloc tree'
+ * mapping
+ */
+static int __update_reloc_root(struct btrfs_root *root, int del)
+{
+	struct rb_node *rb_node;
+	struct mapping_node *node = NULL;
+	struct reloc_control *rc = root->fs_info->reloc_ctl;
+
+	spin_lock(&rc->reloc_root_tree.lock);
+	rb_node = tree_search(&rc->reloc_root_tree.rb_root,
+			      root->commit_root->start);
+	if (rb_node) {
+		node = rb_entry(rb_node, struct mapping_node, rb_node);
+		rb_erase(&node->rb_node, &rc->reloc_root_tree.rb_root);
+	}
+	spin_unlock(&rc->reloc_root_tree.lock);
+
+	BUG_ON((struct btrfs_root *)node->data != root);
+
+	if (!del) {
+		spin_lock(&rc->reloc_root_tree.lock);
+		node->bytenr = root->node->start;
+		rb_node = tree_insert(&rc->reloc_root_tree.rb_root,
+				      node->bytenr, &node->rb_node);
+		spin_unlock(&rc->reloc_root_tree.lock);
+		BUG_ON(rb_node);
+	} else {
+		list_del_init(&root->root_list);
+		kfree(node);
+	}
+	return 0;
+}
+
+/*
+ * create reloc tree for a given fs tree. reloc tree is just a
+ * snapshot of the fs tree with special root objectid.
+ */
+int btrfs_init_reloc_root(struct btrfs_trans_handle *trans,
+			  struct btrfs_root *root)
+{
+	struct btrfs_root *reloc_root;
+	struct extent_buffer *eb;
+	struct btrfs_root_item *root_item;
+	struct btrfs_key root_key;
+	int ret;
+
+	if (root->reloc_root) {
+		reloc_root = root->reloc_root;
+		reloc_root->last_trans = trans->transid;
+		return 0;
+	}
+
+	if (!root->fs_info->reloc_ctl ||
+	    !root->fs_info->reloc_ctl->create_reloc_root ||
+	    root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID)
+		return 0;
+
+	root_item = kmalloc(sizeof(*root_item), GFP_NOFS);
+	BUG_ON(!root_item);
+
+	root_key.objectid = BTRFS_TREE_RELOC_OBJECTID;
+	root_key.type = BTRFS_ROOT_ITEM_KEY;
+	root_key.offset = root->root_key.objectid;
+
+	ret = btrfs_copy_root(trans, root, root->commit_root, &eb,
+			      BTRFS_TREE_RELOC_OBJECTID);
+	BUG_ON(ret);
+
+	btrfs_set_root_last_snapshot(&root->root_item, trans->transid - 1);
+	memcpy(root_item, &root->root_item, sizeof(*root_item));
+	btrfs_set_root_refs(root_item, 1);
+	btrfs_set_root_bytenr(root_item, eb->start);
+	btrfs_set_root_level(root_item, btrfs_header_level(eb));
+	btrfs_set_root_generation(root_item, trans->transid);
+	memset(&root_item->drop_progress, 0, sizeof(struct btrfs_disk_key));
+	root_item->drop_level = 0;
+
+	btrfs_tree_unlock(eb);
+	free_extent_buffer(eb);
+
+	ret = btrfs_insert_root(trans, root->fs_info->tree_root,
+				&root_key, root_item);
+	BUG_ON(ret);
+	kfree(root_item);
+
+	reloc_root = btrfs_read_fs_root_no_radix(root->fs_info->tree_root,
+						 &root_key);
+	BUG_ON(IS_ERR(reloc_root));
+	reloc_root->last_trans = trans->transid;
+
+	__add_reloc_root(reloc_root);
+	root->reloc_root = reloc_root;
+	return 0;
+}
+
+/*
+ * update root item of reloc tree
+ */
+int btrfs_update_reloc_root(struct btrfs_trans_handle *trans,
+			    struct btrfs_root *root)
+{
+	struct btrfs_root *reloc_root;
+	struct btrfs_root_item *root_item;
+	int del = 0;
+	int ret;
+
+	if (!root->reloc_root)
+		return 0;
+
+	reloc_root = root->reloc_root;
+	root_item = &reloc_root->root_item;
+
+	if (btrfs_root_refs(root_item) == 0) {
+		root->reloc_root = NULL;
+		del = 1;
+	}
+
+	__update_reloc_root(reloc_root, del);
+
+	if (reloc_root->commit_root != reloc_root->node) {
+		btrfs_set_root_node(root_item, reloc_root->node);
+		free_extent_buffer(reloc_root->commit_root);
+		reloc_root->commit_root = btrfs_root_node(reloc_root);
+	}
+
+	ret = btrfs_update_root(trans, root->fs_info->tree_root,
+				&reloc_root->root_key, root_item);
+	BUG_ON(ret);
+	return 0;
+}
+
+/*
+ * helper to find first cached inode with inode number >= objectid
+ * in a subvolume
+ */
+static struct inode *find_next_inode(struct btrfs_root *root, u64 objectid)
+{
+	struct rb_node *node;
+	struct rb_node *prev;
+	struct btrfs_inode *entry;
+	struct inode *inode;
+
+	spin_lock(&root->inode_lock);
+again:
+	node = root->inode_tree.rb_node;
+	prev = NULL;
+	while (node) {
+		prev = node;
+		entry = rb_entry(node, struct btrfs_inode, rb_node);
+
+		if (objectid < entry->vfs_inode.i_ino)
+			node = node->rb_left;
+		else if (objectid > entry->vfs_inode.i_ino)
+			node = node->rb_right;
+		else
+			break;
+	}
+	if (!node) {
+		while (prev) {
+			entry = rb_entry(prev, struct btrfs_inode, rb_node);
+			if (objectid <= entry->vfs_inode.i_ino) {
+				node = prev;
+				break;
+			}
+			prev = rb_next(prev);
+		}
+	}
+	while (node) {
+		entry = rb_entry(node, struct btrfs_inode, rb_node);
+		inode = igrab(&entry->vfs_inode);
+		if (inode) {
+			spin_unlock(&root->inode_lock);
+			return inode;
+		}
+
+		objectid = entry->vfs_inode.i_ino + 1;
+		if (cond_resched_lock(&root->inode_lock))
+			goto again;
+
+		node = rb_next(node);
+	}
+	spin_unlock(&root->inode_lock);
+	return NULL;
+}
+
+static int in_block_group(u64 bytenr,
+			  struct btrfs_block_group_cache *block_group)
+{
+	if (bytenr >= block_group->key.objectid &&
+	    bytenr < block_group->key.objectid + block_group->key.offset)
+		return 1;
+	return 0;
+}
+
+/*
+ * get new location of data
+ */
+static int get_new_location(struct inode *reloc_inode, u64 *new_bytenr,
+			    u64 bytenr, u64 num_bytes)
+{
+	struct btrfs_root *root = BTRFS_I(reloc_inode)->root;
+	struct btrfs_path *path;
+	struct btrfs_file_extent_item *fi;
+	struct extent_buffer *leaf;
+	int ret;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	bytenr -= BTRFS_I(reloc_inode)->index_cnt;
+	ret = btrfs_lookup_file_extent(NULL, root, path, reloc_inode->i_ino,
+				       bytenr, 0);
+	if (ret < 0)
+		goto out;
+	if (ret > 0) {
+		ret = -ENOENT;
+		goto out;
+	}
+
+	leaf = path->nodes[0];
+	fi = btrfs_item_ptr(leaf, path->slots[0],
+			    struct btrfs_file_extent_item);
+
+	BUG_ON(btrfs_file_extent_offset(leaf, fi) ||
+	       btrfs_file_extent_compression(leaf, fi) ||
+	       btrfs_file_extent_encryption(leaf, fi) ||
+	       btrfs_file_extent_other_encoding(leaf, fi));
+
+	if (num_bytes != btrfs_file_extent_disk_num_bytes(leaf, fi)) {
+		ret = 1;
+		goto out;
+	}
+
+	if (new_bytenr)
+		*new_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
+	ret = 0;
+out:
+	btrfs_free_path(path);
+	return ret;
+}
+
+/*
+ * update file extent items in the tree leaf to point to
+ * the new locations.
+ */
+static int replace_file_extents(struct btrfs_trans_handle *trans,
+				struct reloc_control *rc,
+				struct btrfs_root *root,
+				struct extent_buffer *leaf,
+				struct list_head *inode_list)
+{
+	struct btrfs_key key;
+	struct btrfs_file_extent_item *fi;
+	struct inode *inode = NULL;
+	struct inodevec *ivec = NULL;
+	u64 parent;
+	u64 bytenr;
+	u64 new_bytenr;
+	u64 num_bytes;
+	u64 end;
+	u32 nritems;
+	u32 i;
+	int ret;
+	int first = 1;
+	int dirty = 0;
+
+	if (rc->stage != UPDATE_DATA_PTRS)
+		return 0;
+
+	/* reloc trees always use full backref */
+	if (root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID)
+		parent = leaf->start;
+	else
+		parent = 0;
+
+	nritems = btrfs_header_nritems(leaf);
+	for (i = 0; i < nritems; i++) {
+		cond_resched();
+		btrfs_item_key_to_cpu(leaf, &key, i);
+		if (key.type != BTRFS_EXTENT_DATA_KEY)
+			continue;
+		fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
+		if (btrfs_file_extent_type(leaf, fi) ==
+		    BTRFS_FILE_EXTENT_INLINE)
+			continue;
+		bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
+		num_bytes = btrfs_file_extent_disk_num_bytes(leaf, fi);
+		if (bytenr == 0)
+			continue;
+		if (!in_block_group(bytenr, rc->block_group))
+			continue;
+
+		/*
+		 * if we are modifying block in fs tree, wait for readpage
+		 * to complete and drop the extent cache
+		 */
+		if (root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID) {
+			if (!ivec || ivec->nr == INODEVEC_SIZE) {
+				ivec = kmalloc(sizeof(*ivec), GFP_NOFS);
+				BUG_ON(!ivec);
+				ivec->nr = 0;
+				list_add_tail(&ivec->list, inode_list);
+			}
+			if (first) {
+				inode = find_next_inode(root, key.objectid);
+				if (inode)
+					ivec->inode[ivec->nr++] = inode;
+				first = 0;
+			} else if (inode && inode->i_ino < key.objectid) {
+				inode = find_next_inode(root, key.objectid);
+				if (inode)
+					ivec->inode[ivec->nr++] = inode;
+			}
+			if (inode && inode->i_ino == key.objectid) {
+				end = key.offset +
+				      btrfs_file_extent_num_bytes(leaf, fi);
+				WARN_ON(!IS_ALIGNED(key.offset,
+						    root->sectorsize));
+				WARN_ON(!IS_ALIGNED(end, root->sectorsize));
+				end--;
+				ret = try_lock_extent(&BTRFS_I(inode)->io_tree,
+						      key.offset, end,
+						      GFP_NOFS);
+				if (!ret)
+					continue;
+
+				btrfs_drop_extent_cache(inode, key.offset, end,
+							1);
+				unlock_extent(&BTRFS_I(inode)->io_tree,
+					      key.offset, end, GFP_NOFS);
+			}
+		}
+
+		ret = get_new_location(rc->data_inode, &new_bytenr,
+				       bytenr, num_bytes);
+		if (ret > 0)
+			continue;
+		BUG_ON(ret < 0);
+
+		btrfs_set_file_extent_disk_bytenr(leaf, fi, new_bytenr);
+		dirty = 1;
+
+		key.offset -= btrfs_file_extent_offset(leaf, fi);
+		ret = btrfs_inc_extent_ref(trans, root, new_bytenr,
+					   num_bytes, parent,
+					   btrfs_header_owner(leaf),
+					   key.objectid, key.offset);
+		BUG_ON(ret);
+
+		ret = btrfs_free_extent(trans, root, bytenr, num_bytes,
+					parent, btrfs_header_owner(leaf),
+					key.objectid, key.offset);
+		BUG_ON(ret);
+	}
+	if (dirty)
+		btrfs_mark_buffer_dirty(leaf);
+	return 0;
+}
+
+static noinline_for_stack
+int memcmp_node_keys(struct extent_buffer *eb, int slot,
+		     struct btrfs_path *path, int level)
+{
+	struct btrfs_disk_key key1;
+	struct btrfs_disk_key key2;
+	btrfs_node_key(eb, &key1, slot);
+	btrfs_node_key(path->nodes[level], &key2, path->slots[level]);
+	return memcmp(&key1, &key2, sizeof(key1));
+}
+
+/*
+ * try to replace tree blocks in fs tree with the new blocks
+ * in reloc tree. tree blocks haven't been modified since the
+ * reloc tree was create can be replaced.
+ *
+ * if a block was replaced, level of the block + 1 is returned.
+ * if no block got replaced, 0 is returned. if there are other
+ * errors, a negative error number is returned.
+ */
+static int replace_path(struct btrfs_trans_handle *trans,
+			struct btrfs_root *dest, struct btrfs_root *src,
+			struct btrfs_path *path, struct btrfs_key *next_key,
+			struct extent_buffer **leaf,
+			int lowest_level, int max_level)
+{
+	struct extent_buffer *eb;
+	struct extent_buffer *parent;
+	struct btrfs_key key;
+	u64 old_bytenr;
+	u64 new_bytenr;
+	u64 old_ptr_gen;
+	u64 new_ptr_gen;
+	u64 last_snapshot;
+	u32 blocksize;
+	int level;
+	int ret;
+	int slot;
+
+	BUG_ON(src->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID);
+	BUG_ON(dest->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID);
+	BUG_ON(lowest_level > 1 && leaf);
+
+	last_snapshot = btrfs_root_last_snapshot(&src->root_item);
+
+	slot = path->slots[lowest_level];
+	btrfs_node_key_to_cpu(path->nodes[lowest_level], &key, slot);
+
+	eb = btrfs_lock_root_node(dest);
+	btrfs_set_lock_blocking(eb);
+	level = btrfs_header_level(eb);
+
+	if (level < lowest_level) {
+		btrfs_tree_unlock(eb);
+		free_extent_buffer(eb);
+		return 0;
+	}
+
+	ret = btrfs_cow_block(trans, dest, eb, NULL, 0, &eb);
+	BUG_ON(ret);
+	btrfs_set_lock_blocking(eb);
+
+	if (next_key) {
+		next_key->objectid = (u64)-1;
+		next_key->type = (u8)-1;
+		next_key->offset = (u64)-1;
+	}
+
+	parent = eb;
+	while (1) {
+		level = btrfs_header_level(parent);
+		BUG_ON(level < lowest_level);
+
+		ret = btrfs_bin_search(parent, &key, level, &slot);
+		if (ret && slot > 0)
+			slot--;
+
+		if (next_key && slot + 1 < btrfs_header_nritems(parent))
+			btrfs_node_key_to_cpu(parent, next_key, slot + 1);
+
+		old_bytenr = btrfs_node_blockptr(parent, slot);
+		blocksize = btrfs_level_size(dest, level - 1);
+		old_ptr_gen = btrfs_node_ptr_generation(parent, slot);
+
+		if (level <= max_level) {
+			eb = path->nodes[level];
+			new_bytenr = btrfs_node_blockptr(eb,
+							path->slots[level]);
+			new_ptr_gen = btrfs_node_ptr_generation(eb,
+							path->slots[level]);
+		} else {
+			new_bytenr = 0;
+			new_ptr_gen = 0;
+		}
+
+		if (new_bytenr > 0 && new_bytenr == old_bytenr) {
+			WARN_ON(1);
+			ret = level;
+			break;
+		}
+
+		if (new_bytenr == 0 || old_ptr_gen > last_snapshot ||
+		    memcmp_node_keys(parent, slot, path, level)) {
+			if (level <= lowest_level && !leaf) {
+				ret = 0;
+				break;
+			}
+
+			eb = read_tree_block(dest, old_bytenr, blocksize,
+					     old_ptr_gen);
+			btrfs_tree_lock(eb);
+			ret = btrfs_cow_block(trans, dest, eb, parent,
+					      slot, &eb);
+			BUG_ON(ret);
+			btrfs_set_lock_blocking(eb);
+
+			if (level <= lowest_level) {
+				*leaf = eb;
+				ret = 0;
+				break;
+			}
+
+			btrfs_tree_unlock(parent);
+			free_extent_buffer(parent);
+
+			parent = eb;
+			continue;
+		}
+
+		btrfs_node_key_to_cpu(path->nodes[level], &key,
+				      path->slots[level]);
+		btrfs_release_path(src, path);
+
+		path->lowest_level = level;
+		ret = btrfs_search_slot(trans, src, &key, path, 0, 1);
+		path->lowest_level = 0;
+		BUG_ON(ret);
+
+		/*
+		 * swap blocks in fs tree and reloc tree.
+		 */
+		btrfs_set_node_blockptr(parent, slot, new_bytenr);
+		btrfs_set_node_ptr_generation(parent, slot, new_ptr_gen);
+		btrfs_mark_buffer_dirty(parent);
+
+		btrfs_set_node_blockptr(path->nodes[level],
+					path->slots[level], old_bytenr);
+		btrfs_set_node_ptr_generation(path->nodes[level],
+					      path->slots[level], old_ptr_gen);
+		btrfs_mark_buffer_dirty(path->nodes[level]);
+
+		ret = btrfs_inc_extent_ref(trans, src, old_bytenr, blocksize,
+					path->nodes[level]->start,
+					src->root_key.objectid, level - 1, 0);
+		BUG_ON(ret);
+		ret = btrfs_inc_extent_ref(trans, dest, new_bytenr, blocksize,
+					0, dest->root_key.objectid, level - 1,
+					0);
+		BUG_ON(ret);
+
+		ret = btrfs_free_extent(trans, src, new_bytenr, blocksize,
+					path->nodes[level]->start,
+					src->root_key.objectid, level - 1, 0);
+		BUG_ON(ret);
+
+		ret = btrfs_free_extent(trans, dest, old_bytenr, blocksize,
+					0, dest->root_key.objectid, level - 1,
+					0);
+		BUG_ON(ret);
+
+		btrfs_unlock_up_safe(path, 0);
+
+		ret = level;
+		break;
+	}
+	btrfs_tree_unlock(parent);
+	free_extent_buffer(parent);
+	return ret;
+}
+
+/*
+ * helper to find next relocated block in reloc tree
+ */
+static noinline_for_stack
+int walk_up_reloc_tree(struct btrfs_root *root, struct btrfs_path *path,
+		       int *level)
+{
+	struct extent_buffer *eb;
+	int i;
+	u64 last_snapshot;
+	u32 nritems;
+
+	last_snapshot = btrfs_root_last_snapshot(&root->root_item);
+
+	for (i = 0; i < *level; i++) {
+		free_extent_buffer(path->nodes[i]);
+		path->nodes[i] = NULL;
+	}
+
+	for (i = *level; i < BTRFS_MAX_LEVEL && path->nodes[i]; i++) {
+		eb = path->nodes[i];
+		nritems = btrfs_header_nritems(eb);
+		while (path->slots[i] + 1 < nritems) {
+			path->slots[i]++;
+			if (btrfs_node_ptr_generation(eb, path->slots[i]) <=
+			    last_snapshot)
+				continue;
+
+			*level = i;
+			return 0;
+		}
+		free_extent_buffer(path->nodes[i]);
+		path->nodes[i] = NULL;
+	}
+	return 1;
+}
+
+/*
+ * walk down reloc tree to find relocated block of lowest level
+ */
+static noinline_for_stack
+int walk_down_reloc_tree(struct btrfs_root *root, struct btrfs_path *path,
+			 int *level)
+{
+	struct extent_buffer *eb = NULL;
+	int i;
+	u64 bytenr;
+	u64 ptr_gen = 0;
+	u64 last_snapshot;
+	u32 blocksize;
+	u32 nritems;
+
+	last_snapshot = btrfs_root_last_snapshot(&root->root_item);
+
+	for (i = *level; i > 0; i--) {
+		eb = path->nodes[i];
+		nritems = btrfs_header_nritems(eb);
+		while (path->slots[i] < nritems) {
+			ptr_gen = btrfs_node_ptr_generation(eb, path->slots[i]);
+			if (ptr_gen > last_snapshot)
+				break;
+			path->slots[i]++;
+		}
+		if (path->slots[i] >= nritems) {
+			if (i == *level)
+				break;
+			*level = i + 1;
+			return 0;
+		}
+		if (i == 1) {
+			*level = i;
+			return 0;
+		}
+
+		bytenr = btrfs_node_blockptr(eb, path->slots[i]);
+		blocksize = btrfs_level_size(root, i - 1);
+		eb = read_tree_block(root, bytenr, blocksize, ptr_gen);
+		BUG_ON(btrfs_header_level(eb) != i - 1);
+		path->nodes[i - 1] = eb;
+		path->slots[i - 1] = 0;
+	}
+	return 1;
+}
+
+/*
+ * invalidate extent cache for file extents whose key in range of
+ * [min_key, max_key)
+ */
+static int invalidate_extent_cache(struct btrfs_root *root,
+				   struct btrfs_key *min_key,
+				   struct btrfs_key *max_key)
+{
+	struct inode *inode = NULL;
+	u64 objectid;
+	u64 start, end;
+
+	objectid = min_key->objectid;
+	while (1) {
+		cond_resched();
+		iput(inode);
+
+		if (objectid > max_key->objectid)
+			break;
+
+		inode = find_next_inode(root, objectid);
+		if (!inode)
+			break;
+
+		if (inode->i_ino > max_key->objectid) {
+			iput(inode);
+			break;
+		}
+
+		objectid = inode->i_ino + 1;
+		if (!S_ISREG(inode->i_mode))
+			continue;
+
+		if (unlikely(min_key->objectid == inode->i_ino)) {
+			if (min_key->type > BTRFS_EXTENT_DATA_KEY)
+				continue;
+			if (min_key->type < BTRFS_EXTENT_DATA_KEY)
+				start = 0;
+			else {
+				start = min_key->offset;
+				WARN_ON(!IS_ALIGNED(start, root->sectorsize));
+			}
+		} else {
+			start = 0;
+		}
+
+		if (unlikely(max_key->objectid == inode->i_ino)) {
+			if (max_key->type < BTRFS_EXTENT_DATA_KEY)
+				continue;
+			if (max_key->type > BTRFS_EXTENT_DATA_KEY) {
+				end = (u64)-1;
+			} else {
+				if (max_key->offset == 0)
+					continue;
+				end = max_key->offset;
+				WARN_ON(!IS_ALIGNED(end, root->sectorsize));
+				end--;
+			}
+		} else {
+			end = (u64)-1;
+		}
+
+		/* the lock_extent waits for readpage to complete */
+		lock_extent(&BTRFS_I(inode)->io_tree, start, end, GFP_NOFS);
+		btrfs_drop_extent_cache(inode, start, end, 1);
+		unlock_extent(&BTRFS_I(inode)->io_tree, start, end, GFP_NOFS);
+	}
+	return 0;
+}
+
+static int find_next_key(struct btrfs_path *path, int level,
+			 struct btrfs_key *key)
+
+{
+	while (level < BTRFS_MAX_LEVEL) {
+		if (!path->nodes[level])
+			break;
+		if (path->slots[level] + 1 <
+		    btrfs_header_nritems(path->nodes[level])) {
+			btrfs_node_key_to_cpu(path->nodes[level], key,
+					      path->slots[level] + 1);
+			return 0;
+		}
+		level++;
+	}
+	return 1;
+}
+
+/*
+ * merge the relocated tree blocks in reloc tree with corresponding
+ * fs tree.
+ */
+static noinline_for_stack int merge_reloc_root(struct reloc_control *rc,
+					       struct btrfs_root *root)
+{
+	LIST_HEAD(inode_list);
+	struct btrfs_key key;
+	struct btrfs_key next_key;
+	struct btrfs_trans_handle *trans;
+	struct btrfs_root *reloc_root;
+	struct btrfs_root_item *root_item;
+	struct btrfs_path *path;
+	struct extent_buffer *leaf = NULL;
+	unsigned long nr;
+	int level;
+	int max_level;
+	int replaced = 0;
+	int ret;
+	int err = 0;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	reloc_root = root->reloc_root;
+	root_item = &reloc_root->root_item;
+
+	if (btrfs_disk_key_objectid(&root_item->drop_progress) == 0) {
+		level = btrfs_root_level(root_item);
+		extent_buffer_get(reloc_root->node);
+		path->nodes[level] = reloc_root->node;
+		path->slots[level] = 0;
+	} else {
+		btrfs_disk_key_to_cpu(&key, &root_item->drop_progress);
+
+		level = root_item->drop_level;
+		BUG_ON(level == 0);
+		path->lowest_level = level;
+		ret = btrfs_search_slot(NULL, reloc_root, &key, path, 0, 0);
+		if (ret < 0) {
+			btrfs_free_path(path);
+			return ret;
+		}
+
+		btrfs_node_key_to_cpu(path->nodes[level], &next_key,
+				      path->slots[level]);
+		WARN_ON(memcmp(&key, &next_key, sizeof(key)));
+
+		btrfs_unlock_up_safe(path, 0);
+	}
+
+	if (level == 0 && rc->stage == UPDATE_DATA_PTRS) {
+		trans = btrfs_start_transaction(root, 1);
+
+		leaf = path->nodes[0];
+		btrfs_item_key_to_cpu(leaf, &key, 0);
+		btrfs_release_path(reloc_root, path);
+
+		ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
+		if (ret < 0) {
+			err = ret;
+			goto out;
+		}
+
+		leaf = path->nodes[0];
+		btrfs_unlock_up_safe(path, 1);
+		ret = replace_file_extents(trans, rc, root, leaf,
+					   &inode_list);
+		if (ret < 0)
+			err = ret;
+		goto out;
+	}
+
+	memset(&next_key, 0, sizeof(next_key));
+
+	while (1) {
+		leaf = NULL;
+		replaced = 0;
+		trans = btrfs_start_transaction(root, 1);
+		max_level = level;
+
+		ret = walk_down_reloc_tree(reloc_root, path, &level);
+		if (ret < 0) {
+			err = ret;
+			goto out;
+		}
+		if (ret > 0)
+			break;
+
+		if (!find_next_key(path, level, &key) &&
+		    btrfs_comp_cpu_keys(&next_key, &key) >= 0) {
+			ret = 0;
+		} else if (level == 1 && rc->stage == UPDATE_DATA_PTRS) {
+			ret = replace_path(trans, root, reloc_root,
+					   path, &next_key, &leaf,
+					   level, max_level);
+		} else {
+			ret = replace_path(trans, root, reloc_root,
+					   path, &next_key, NULL,
+					   level, max_level);
+		}
+		if (ret < 0) {
+			err = ret;
+			goto out;
+		}
+
+		if (ret > 0) {
+			level = ret;
+			btrfs_node_key_to_cpu(path->nodes[level], &key,
+					      path->slots[level]);
+			replaced = 1;
+		} else if (leaf) {
+			/*
+			 * no block got replaced, try replacing file extents
+			 */
+			btrfs_item_key_to_cpu(leaf, &key, 0);
+			ret = replace_file_extents(trans, rc, root, leaf,
+						   &inode_list);
+			btrfs_tree_unlock(leaf);
+			free_extent_buffer(leaf);
+			BUG_ON(ret < 0);
+		}
+
+		ret = walk_up_reloc_tree(reloc_root, path, &level);
+		if (ret > 0)
+			break;
+
+		BUG_ON(level == 0);
+		/*
+		 * save the merging progress in the drop_progress.
+		 * this is OK since root refs == 1 in this case.
+		 */
+		btrfs_node_key(path->nodes[level], &root_item->drop_progress,
+			       path->slots[level]);
+		root_item->drop_level = level;
+
+		nr = trans->blocks_used;
+		btrfs_end_transaction(trans, root);
+
+		btrfs_btree_balance_dirty(root, nr);
+
+		if (replaced && rc->stage == UPDATE_DATA_PTRS)
+			invalidate_extent_cache(root, &key, &next_key);
+	}
+
+	/*
+	 * handle the case only one block in the fs tree need to be
+	 * relocated and the block is tree root.
+	 */
+	leaf = btrfs_lock_root_node(root);
+	ret = btrfs_cow_block(trans, root, leaf, NULL, 0, &leaf);
+	btrfs_tree_unlock(leaf);
+	free_extent_buffer(leaf);
+	if (ret < 0)
+		err = ret;
+out:
+	btrfs_free_path(path);
+
+	if (err == 0) {
+		memset(&root_item->drop_progress, 0,
+		       sizeof(root_item->drop_progress));
+		root_item->drop_level = 0;
+		btrfs_set_root_refs(root_item, 0);
+	}
+
+	nr = trans->blocks_used;
+	btrfs_end_transaction(trans, root);
+
+	btrfs_btree_balance_dirty(root, nr);
+
+	/*
+	 * put inodes while we aren't holding the tree locks
+	 */
+	while (!list_empty(&inode_list)) {
+		struct inodevec *ivec;
+		ivec = list_entry(inode_list.next, struct inodevec, list);
+		list_del(&ivec->list);
+		while (ivec->nr > 0) {
+			ivec->nr--;
+			iput(ivec->inode[ivec->nr]);
+		}
+		kfree(ivec);
+	}
+
+	if (replaced && rc->stage == UPDATE_DATA_PTRS)
+		invalidate_extent_cache(root, &key, &next_key);
+
+	return err;
+}
+
+/*
+ * callback for the work threads.
+ * this function merges reloc tree with corresponding fs tree,
+ * and then drops the reloc tree.
+ */
+static void merge_func(struct btrfs_work *work)
+{
+	struct btrfs_trans_handle *trans;
+	struct btrfs_root *root;
+	struct btrfs_root *reloc_root;
+	struct async_merge *async;
+
+	async = container_of(work, struct async_merge, work);
+	reloc_root = async->root;
+
+	if (btrfs_root_refs(&reloc_root->root_item) > 0) {
+		root = read_fs_root(reloc_root->fs_info,
+				    reloc_root->root_key.offset);
+		BUG_ON(IS_ERR(root));
+		BUG_ON(root->reloc_root != reloc_root);
+
+		merge_reloc_root(async->rc, root);
+
+		trans = btrfs_start_transaction(root, 1);
+		btrfs_update_reloc_root(trans, root);
+		btrfs_end_transaction(trans, root);
+	}
+
+	btrfs_drop_dead_root(reloc_root);
+
+	if (atomic_dec_and_test(async->num_pending))
+		complete(async->done);
+
+	kfree(async);
+}
+
+static int merge_reloc_roots(struct reloc_control *rc)
+{
+	struct async_merge *async;
+	struct btrfs_root *root;
+	struct completion done;
+	atomic_t num_pending;
+
+	init_completion(&done);
+	atomic_set(&num_pending, 1);
+
+	while (!list_empty(&rc->reloc_roots)) {
+		root = list_entry(rc->reloc_roots.next,
+				  struct btrfs_root, root_list);
+		list_del_init(&root->root_list);
+
+		async = kmalloc(sizeof(*async), GFP_NOFS);
+		BUG_ON(!async);
+		async->work.func = merge_func;
+		async->work.flags = 0;
+		async->rc = rc;
+		async->root = root;
+		async->done = &done;
+		async->num_pending = &num_pending;
+		atomic_inc(&num_pending);
+		btrfs_queue_worker(&rc->workers, &async->work);
+	}
+
+	if (!atomic_dec_and_test(&num_pending))
+		wait_for_completion(&done);
+
+	BUG_ON(!RB_EMPTY_ROOT(&rc->reloc_root_tree.rb_root));
+	return 0;
+}
+
+static void free_block_list(struct rb_root *blocks)
+{
+	struct tree_block *block;
+	struct rb_node *rb_node;
+	while ((rb_node = rb_first(blocks))) {
+		block = rb_entry(rb_node, struct tree_block, rb_node);
+		rb_erase(rb_node, blocks);
+		kfree(block);
+	}
+}
+
+static int record_reloc_root_in_trans(struct btrfs_trans_handle *trans,
+				      struct btrfs_root *reloc_root)
+{
+	struct btrfs_root *root;
+
+	if (reloc_root->last_trans == trans->transid)
+		return 0;
+
+	root = read_fs_root(reloc_root->fs_info, reloc_root->root_key.offset);
+	BUG_ON(IS_ERR(root));
+	BUG_ON(root->reloc_root != reloc_root);
+
+	return btrfs_record_root_in_trans(trans, root);
+}
+
+/*
+ * select one tree from trees that references the block.
+ * for blocks in refernce counted trees, we preper reloc tree.
+ * if no reloc tree found and reloc_only is true, NULL is returned.
+ */
+static struct btrfs_root *__select_one_root(struct btrfs_trans_handle *trans,
+					    struct backref_node *node,
+					    struct backref_edge *edges[],
+					    int *nr, int reloc_only)
+{
+	struct backref_node *next;
+	struct btrfs_root *root;
+	int index;
+	int loop = 0;
+again:
+	index = 0;
+	next = node;
+	while (1) {
+		cond_resched();
+		next = walk_up_backref(next, edges, &index);
+		root = next->root;
+		if (!root) {
+			BUG_ON(!node->old_root);
+			goto skip;
+		}
+
+		/* no other choice for non-refernce counted tree */
+		if (!root->ref_cows) {
+			BUG_ON(reloc_only);
+			break;
+		}
+
+		if (root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID) {
+			record_reloc_root_in_trans(trans, root);
+			break;
+		}
+
+		if (loop) {
+			btrfs_record_root_in_trans(trans, root);
+			break;
+		}
+
+		if (reloc_only || next != node) {
+			if (!root->reloc_root)
+				btrfs_record_root_in_trans(trans, root);
+			root = root->reloc_root;
+			/*
+			 * if the reloc tree was created in current
+			 * transation, there is no node in backref tree
+			 * corresponds to the root of the reloc tree.
+			 */
+			if (btrfs_root_last_snapshot(&root->root_item) ==
+			    trans->transid - 1)
+				break;
+		}
+skip:
+		root = NULL;
+		next = walk_down_backref(edges, &index);
+		if (!next || next->level <= node->level)
+			break;
+	}
+
+	if (!root && !loop && !reloc_only) {
+		loop = 1;
+		goto again;
+	}
+
+	if (root)
+		*nr = index;
+	else
+		*nr = 0;
+
+	return root;
+}
+
+static noinline_for_stack
+struct btrfs_root *select_one_root(struct btrfs_trans_handle *trans,
+				   struct backref_node *node)
+{
+	struct backref_edge *edges[BTRFS_MAX_LEVEL - 1];
+	int nr;
+	return __select_one_root(trans, node, edges, &nr, 0);
+}
+
+static noinline_for_stack
+struct btrfs_root *select_reloc_root(struct btrfs_trans_handle *trans,
+				     struct backref_node *node,
+				     struct backref_edge *edges[], int *nr)
+{
+	return __select_one_root(trans, node, edges, nr, 1);
+}
+
+static void grab_path_buffers(struct btrfs_path *path,
+			      struct backref_node *node,
+			      struct backref_edge *edges[], int nr)
+{
+	int i = 0;
+	while (1) {
+		drop_node_buffer(node);
+		node->eb = path->nodes[node->level];
+		BUG_ON(!node->eb);
+		if (path->locks[node->level])
+			node->locked = 1;
+		path->nodes[node->level] = NULL;
+		path->locks[node->level] = 0;
+
+		if (i >= nr)
+			break;
+
+		edges[i]->blockptr = node->eb->start;
+		node = edges[i]->node[UPPER];
+		i++;
+	}
+}
+
+/*
+ * relocate a block tree, and then update pointers in upper level
+ * blocks that reference the block to point to the new location.
+ *
+ * if called by link_to_upper, the block has already been relocated.
+ * in that case this function just updates pointers.
+ */
+static int do_relocation(struct btrfs_trans_handle *trans,
+			 struct backref_node *node,
+			 struct btrfs_key *key,
+			 struct btrfs_path *path, int lowest)
+{
+	struct backref_node *upper;
+	struct backref_edge *edge;
+	struct backref_edge *edges[BTRFS_MAX_LEVEL - 1];
+	struct btrfs_root *root;
+	struct extent_buffer *eb;
+	u32 blocksize;
+	u64 bytenr;
+	u64 generation;
+	int nr;
+	int slot;
+	int ret;
+	int err = 0;
+
+	BUG_ON(lowest && node->eb);
+
+	path->lowest_level = node->level + 1;
+	list_for_each_entry(edge, &node->upper, list[LOWER]) {
+		cond_resched();
+		if (node->eb && node->eb->start == edge->blockptr)
+			continue;
+
+		upper = edge->node[UPPER];
+		root = select_reloc_root(trans, upper, edges, &nr);
+		if (!root)
+			continue;
+
+		if (upper->eb && !upper->locked)
+			drop_node_buffer(upper);
+
+		if (!upper->eb) {
+			ret = btrfs_search_slot(trans, root, key, path, 0, 1);
+			if (ret < 0) {
+				err = ret;
+				break;
+			}
+			BUG_ON(ret > 0);
+
+			slot = path->slots[upper->level];
+
+			btrfs_unlock_up_safe(path, upper->level + 1);
+			grab_path_buffers(path, upper, edges, nr);
+
+			btrfs_release_path(NULL, path);
+		} else {
+			ret = btrfs_bin_search(upper->eb, key, upper->level,
+					       &slot);
+			BUG_ON(ret);
+		}
+
+		bytenr = btrfs_node_blockptr(upper->eb, slot);
+		if (!lowest) {
+			if (node->eb->start == bytenr) {
+				btrfs_tree_unlock(upper->eb);
+				upper->locked = 0;
+				continue;
+			}
+		} else {
+			BUG_ON(node->bytenr != bytenr);
+		}
+
+		blocksize = btrfs_level_size(root, node->level);
+		generation = btrfs_node_ptr_generation(upper->eb, slot);
+		eb = read_tree_block(root, bytenr, blocksize, generation);
+		btrfs_tree_lock(eb);
+		btrfs_set_lock_blocking(eb);
+
+		if (!node->eb) {
+			ret = btrfs_cow_block(trans, root, eb, upper->eb,
+					      slot, &eb);
+			if (ret < 0) {
+				err = ret;
+				break;
+			}
+			btrfs_set_lock_blocking(eb);
+			node->eb = eb;
+			node->locked = 1;
+		} else {
+			btrfs_set_node_blockptr(upper->eb, slot,
+						node->eb->start);
+			btrfs_set_node_ptr_generation(upper->eb, slot,
+						      trans->transid);
+			btrfs_mark_buffer_dirty(upper->eb);
+
+			ret = btrfs_inc_extent_ref(trans, root,
+						node->eb->start, blocksize,
+						upper->eb->start,
+						btrfs_header_owner(upper->eb),
+						node->level, 0);
+			BUG_ON(ret);
+
+			ret = btrfs_drop_subtree(trans, root, eb, upper->eb);
+			BUG_ON(ret);
+
+			btrfs_tree_unlock(eb);
+			free_extent_buffer(eb);
+		}
+		if (!lowest) {
+			btrfs_tree_unlock(upper->eb);
+			upper->locked = 0;
+		}
+	}
+	path->lowest_level = 0;
+	return err;
+}
+
+static int link_to_upper(struct btrfs_trans_handle *trans,
+			 struct backref_node *node,
+			 struct btrfs_path *path)
+{
+	struct btrfs_key key;
+	if (!node->eb || list_empty(&node->upper))
+		return 0;
+
+	btrfs_node_key_to_cpu(node->eb, &key, 0);
+	return do_relocation(trans, node, &key, path, 0);
+}
+
+static int finish_pending_nodes(struct btrfs_trans_handle *trans,
+				struct backref_cache *cache,
+				struct btrfs_path *path)
+{
+	struct backref_node *node;
+	int level;
+	int ret;
+	int err = 0;
+
+	for (level = 0; level < BTRFS_MAX_LEVEL; level++) {
+		while (!list_empty(&cache->pending[level])) {
+			node = list_entry(cache->pending[level].next,
+					  struct backref_node, lower);
+			BUG_ON(node->level != level);
+
+			ret = link_to_upper(trans, node, path);
+			if (ret < 0)
+				err = ret;
+			/*
+			 * this remove the node from the pending list and
+			 * may add some other nodes to the level + 1
+			 * pending list
+			 */
+			remove_backref_node(cache, node);
+		}
+	}
+	BUG_ON(!RB_EMPTY_ROOT(&cache->rb_root));
+	return err;
+}
+
+static void mark_block_processed(struct reloc_control *rc,
+				 struct backref_node *node)
+{
+	u32 blocksize;
+	if (node->level == 0 ||
+	    in_block_group(node->bytenr, rc->block_group)) {
+		blocksize = btrfs_level_size(rc->extent_root, node->level);
+		set_extent_bits(&rc->processed_blocks, node->bytenr,
+				node->bytenr + blocksize - 1, EXTENT_DIRTY,
+				GFP_NOFS);
+	}
+	node->processed = 1;
+}
+
+/*
+ * mark a block and all blocks directly/indirectly reference the block
+ * as processed.
+ */
+static void update_processed_blocks(struct reloc_control *rc,
+				    struct backref_node *node)
+{
+	struct backref_node *next = node;
+	struct backref_edge *edge;
+	struct backref_edge *edges[BTRFS_MAX_LEVEL - 1];
+	int index = 0;
+
+	while (next) {
+		cond_resched();
+		while (1) {
+			if (next->processed)
+				break;
+
+			mark_block_processed(rc, next);
+
+			if (list_empty(&next->upper))
+				break;
+
+			edge = list_entry(next->upper.next,
+					  struct backref_edge, list[LOWER]);
+			edges[index++] = edge;
+			next = edge->node[UPPER];
+		}
+		next = walk_down_backref(edges, &index);
+	}
+}
+
+static int tree_block_processed(u64 bytenr, u32 blocksize,
+				struct reloc_control *rc)
+{
+	if (test_range_bit(&rc->processed_blocks, bytenr,
+			   bytenr + blocksize - 1, EXTENT_DIRTY, 1))
+		return 1;
+	return 0;
+}
+
+/*
+ * check if there are any file extent pointers in the leaf point to
+ * data require processing
+ */
+static int check_file_extents(struct reloc_control *rc,
+			      u64 bytenr, u32 blocksize, u64 ptr_gen)
+{
+	struct btrfs_key found_key;
+	struct btrfs_file_extent_item *fi;
+	struct extent_buffer *leaf;
+	u32 nritems;
+	int i;
+	int ret = 0;
+
+	leaf = read_tree_block(rc->extent_root, bytenr, blocksize, ptr_gen);
+
+	nritems = btrfs_header_nritems(leaf);
+	for (i = 0; i < nritems; i++) {
+		cond_resched();
+		btrfs_item_key_to_cpu(leaf, &found_key, i);
+		if (found_key.type != BTRFS_EXTENT_DATA_KEY)
+			continue;
+		fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
+		if (btrfs_file_extent_type(leaf, fi) ==
+		    BTRFS_FILE_EXTENT_INLINE)
+			continue;
+		bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
+		if (bytenr == 0)
+			continue;
+		if (in_block_group(bytenr, rc->block_group)) {
+			ret = 1;
+			break;
+		}
+	}
+	free_extent_buffer(leaf);
+	return ret;
+}
+
+/*
+ * scan child blocks of a given block to find blocks require processing
+ */
+static int add_child_blocks(struct btrfs_trans_handle *trans,
+			    struct reloc_control *rc,
+			    struct backref_node *node,
+			    struct rb_root *blocks)
+{
+	struct tree_block *block;
+	struct rb_node *rb_node;
+	u64 bytenr;
+	u64 ptr_gen;
+	u32 blocksize;
+	u32 nritems;
+	int i;
+	int err = 0;
+
+	nritems = btrfs_header_nritems(node->eb);
+	blocksize = btrfs_level_size(rc->extent_root, node->level - 1);
+	for (i = 0; i < nritems; i++) {
+		cond_resched();
+		bytenr = btrfs_node_blockptr(node->eb, i);
+		ptr_gen = btrfs_node_ptr_generation(node->eb, i);
+		if (ptr_gen == trans->transid)
+			continue;
+		if (!in_block_group(bytenr, rc->block_group) &&
+		    (node->level > 1 || rc->stage == MOVE_DATA_EXTENTS))
+			continue;
+		if (tree_block_processed(bytenr, blocksize, rc))
+			continue;
+
+		readahead_tree_block(rc->extent_root,
+				     bytenr, blocksize, ptr_gen);
+	}
+
+	for (i = 0; i < nritems; i++) {
+		cond_resched();
+		bytenr = btrfs_node_blockptr(node->eb, i);
+		ptr_gen = btrfs_node_ptr_generation(node->eb, i);
+		if (ptr_gen == trans->transid)
+			continue;
+		if (!in_block_group(bytenr, rc->block_group) &&
+		    (node->level > 1 || rc->stage == MOVE_DATA_EXTENTS))
+			continue;
+		if (tree_block_processed(bytenr, blocksize, rc))
+			continue;
+		if (!in_block_group(bytenr, rc->block_group) &&
+		    !check_file_extents(rc, bytenr, blocksize, ptr_gen))
+			continue;
+
+		block = kmalloc(sizeof(*block), GFP_NOFS);
+		if (!block) {
+			err = -ENOMEM;
+			break;
+		}
+		block->bytenr = bytenr;
+		btrfs_node_key_to_cpu(node->eb, &block->key, i);
+		block->level = node->level - 1;
+		block->key_ready = 1;
+		rb_node = tree_insert(blocks, block->bytenr, &block->rb_node);
+		BUG_ON(rb_node);
+	}
+	if (err)
+		free_block_list(blocks);
+	return err;
+}
+
+/*
+ * find adjacent blocks require processing
+ */
+static noinline_for_stack
+int add_adjacent_blocks(struct btrfs_trans_handle *trans,
+			struct reloc_control *rc,
+			struct backref_cache *cache,
+			struct rb_root *blocks, int level,
+			struct backref_node **upper)
+{
+	struct backref_node *node;
+	int ret = 0;
+
+	WARN_ON(!list_empty(&cache->pending[level]));
+
+	if (list_empty(&cache->pending[level + 1]))
+		return 1;
+
+	node = list_entry(cache->pending[level + 1].next,
+			  struct backref_node, lower);
+	if (node->eb)
+		ret = add_child_blocks(trans, rc, node, blocks);
+
+	*upper = node;
+	return ret;
+}
+
+static int get_tree_block_key(struct reloc_control *rc,
+			      struct tree_block *block)
+{
+	struct extent_buffer *eb;
+
+	BUG_ON(block->key_ready);
+	eb = read_tree_block(rc->extent_root, block->bytenr,
+			     block->key.objectid, block->key.offset);
+	WARN_ON(btrfs_header_level(eb) != block->level);
+	if (block->level == 0)
+		btrfs_item_key_to_cpu(eb, &block->key, 0);
+	else
+		btrfs_node_key_to_cpu(eb, &block->key, 0);
+	free_extent_buffer(eb);
+	block->key_ready = 1;
+	return 0;
+}
+
+static int reada_tree_block(struct reloc_control *rc,
+			    struct tree_block *block)
+{
+	BUG_ON(block->key_ready);
+	readahead_tree_block(rc->extent_root, block->bytenr,
+			     block->key.objectid, block->key.offset);
+	return 0;
+}
+
+/*
+ * helper function to relocate a tree block
+ */
+static int relocate_tree_block(struct btrfs_trans_handle *trans,
+				struct reloc_control *rc,
+				struct backref_node *node,
+				struct btrfs_key *key,
+				struct btrfs_path *path)
+{
+	struct btrfs_root *root;
+	int ret;
+
+	root = select_one_root(trans, node);
+	if (unlikely(!root)) {
+		rc->found_old_snapshot = 1;
+		update_processed_blocks(rc, node);
+		return 0;
+	}
+
+	if (root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID) {
+		ret = do_relocation(trans, node, key, path, 1);
+		if (ret < 0)
+			goto out;
+		if (node->level == 0 && rc->stage == UPDATE_DATA_PTRS) {
+			ret = replace_file_extents(trans, rc, root,
+						   node->eb, NULL);
+			if (ret < 0)
+				goto out;
+		}
+		drop_node_buffer(node);
+	} else if (!root->ref_cows) {
+		path->lowest_level = node->level;
+		ret = btrfs_search_slot(trans, root, key, path, 0, 1);
+		btrfs_release_path(root, path);
+		if (ret < 0)
+			goto out;
+	} else if (root != node->root) {
+		WARN_ON(node->level > 0 || rc->stage != UPDATE_DATA_PTRS);
+	}
+
+	update_processed_blocks(rc, node);
+	ret = 0;
+out:
+	drop_node_buffer(node);
+	return ret;
+}
+
+/*
+ * relocate a list of blocks
+ */
+static noinline_for_stack
+int relocate_tree_blocks(struct btrfs_trans_handle *trans,
+			 struct reloc_control *rc, struct rb_root *blocks)
+{
+	struct backref_cache *cache;
+	struct backref_node *node;
+	struct btrfs_path *path;
+	struct tree_block *block;
+	struct rb_node *rb_node;
+	int level = -1;
+	int ret;
+	int err = 0;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	cache = kmalloc(sizeof(*cache), GFP_NOFS);
+	if (!cache) {
+		btrfs_free_path(path);
+		return -ENOMEM;
+	}
+
+	backref_cache_init(cache);
+
+	rb_node = rb_first(blocks);
+	while (rb_node) {
+		block = rb_entry(rb_node, struct tree_block, rb_node);
+		if (level == -1)
+			level = block->level;
+		else
+			BUG_ON(level != block->level);
+		if (!block->key_ready)
+			reada_tree_block(rc, block);
+		rb_node = rb_next(rb_node);
+	}
+
+	rb_node = rb_first(blocks);
+	while (rb_node) {
+		block = rb_entry(rb_node, struct tree_block, rb_node);
+		if (!block->key_ready)
+			get_tree_block_key(rc, block);
+		rb_node = rb_next(rb_node);
+	}
+
+	rb_node = rb_first(blocks);
+	while (rb_node) {
+		block = rb_entry(rb_node, struct tree_block, rb_node);
+
+		node = build_backref_tree(rc, cache, &block->key,
+					  block->level, block->bytenr);
+		if (IS_ERR(node)) {
+			err = PTR_ERR(node);
+			goto out;
+		}
+
+		ret = relocate_tree_block(trans, rc, node, &block->key,
+					  path);
+		if (ret < 0) {
+			err = ret;
+			goto out;
+		}
+		remove_backref_node(cache, node);
+		rb_node = rb_next(rb_node);
+	}
+
+	if (level > 0)
+		goto out;
+
+	free_block_list(blocks);
+
+	/*
+	 * now backrefs of some upper level tree blocks have been cached,
+	 * try relocating blocks referenced by these upper level blocks.
+	 */
+	while (1) {
+		struct backref_node *upper = NULL;
+		if (trans->transaction->in_commit ||
+		    trans->transaction->delayed_refs.flushing)
+			break;
+
+		ret = add_adjacent_blocks(trans, rc, cache, blocks, level,
+					  &upper);
+		if (ret < 0)
+			err = ret;
+		if (ret != 0)
+			break;
+
+		rb_node = rb_first(blocks);
+		while (rb_node) {
+			block = rb_entry(rb_node, struct tree_block, rb_node);
+			if (trans->transaction->in_commit ||
+			    trans->transaction->delayed_refs.flushing)
+				goto out;
+			BUG_ON(!block->key_ready);
+			node = build_backref_tree(rc, cache, &block->key,
+						  level, block->bytenr);
+			if (IS_ERR(node)) {
+				err = PTR_ERR(node);
+				goto out;
+			}
+
+			ret = relocate_tree_block(trans, rc, node,
+						  &block->key, path);
+			if (ret < 0) {
+				err = ret;
+				goto out;
+			}
+			remove_backref_node(cache, node);
+			rb_node = rb_next(rb_node);
+		}
+		free_block_list(blocks);
+
+		if (upper) {
+			ret = link_to_upper(trans, upper, path);
+			if (ret < 0) {
+				err = ret;
+				break;
+			}
+			remove_backref_node(cache, upper);
+		}
+	}
+out:
+	free_block_list(blocks);
+
+	ret = finish_pending_nodes(trans, cache, path);
+	if (ret < 0)
+		err = ret;
+
+	kfree(cache);
+	btrfs_free_path(path);
+	return err;
+}
+
+static noinline_for_stack
+int relocate_inode_pages(struct inode *inode, u64 start, u64 len)
+{
+	u64 page_start;
+	u64 page_end;
+	unsigned long i;
+	unsigned long first_index;
+	unsigned long last_index;
+	unsigned int total_read = 0;
+	unsigned int total_dirty = 0;
+	struct page *page;
+	struct file_ra_state *ra;
+	struct btrfs_ordered_extent *ordered;
+	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
+	int ret = 0;
+
+	ra = kzalloc(sizeof(*ra), GFP_NOFS);
+	if (!ra)
+		return -ENOMEM;
+
+	mutex_lock(&inode->i_mutex);
+	first_index = start >> PAGE_CACHE_SHIFT;
+	last_index = (start + len - 1) >> PAGE_CACHE_SHIFT;
+
+	/* make sure the dirty trick played by the caller work */
+	ret = invalidate_inode_pages2_range(inode->i_mapping,
+					    first_index, last_index);
+	if (ret)
+		goto out_unlock;
+
+	file_ra_state_init(ra, inode->i_mapping);
+
+	for (i = first_index ; i <= last_index; i++) {
+		if (total_read % ra->ra_pages == 0) {
+			btrfs_force_ra(inode->i_mapping, ra, NULL, i,
+				min(last_index, ra->ra_pages + i - 1));
+		}
+		total_read++;
+again:
+		if (((u64)i << PAGE_CACHE_SHIFT) > i_size_read(inode))
+			BUG_ON(1);
+		page = grab_cache_page(inode->i_mapping, i);
+		if (!page) {
+			ret = -ENOMEM;
+			goto out_unlock;
+		}
+		if (!PageUptodate(page)) {
+			btrfs_readpage(NULL, page);
+			lock_page(page);
+			if (!PageUptodate(page)) {
+				unlock_page(page);
+				page_cache_release(page);
+				ret = -EIO;
+				goto out_unlock;
+			}
+		}
+		wait_on_page_writeback(page);
+
+		page_start = (u64)page->index << PAGE_CACHE_SHIFT;
+		page_end = page_start + PAGE_CACHE_SIZE - 1;
+		lock_extent(io_tree, page_start, page_end, GFP_NOFS);
+
+		ordered = btrfs_lookup_ordered_extent(inode, page_start);
+		if (ordered) {
+			unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
+			unlock_page(page);
+			page_cache_release(page);
+			btrfs_start_ordered_extent(inode, ordered, 1);
+			btrfs_put_ordered_extent(ordered);
+			goto again;
+		}
+		set_page_extent_mapped(page);
+
+		if (i == first_index)
+			set_extent_bits(io_tree, page_start, page_end,
+					EXTENT_BOUNDARY, GFP_NOFS);
+		btrfs_set_extent_delalloc(inode, page_start, page_end);
+
+		set_page_dirty(page);
+		total_dirty++;
+
+		unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
+		unlock_page(page);
+		page_cache_release(page);
+	}
+out_unlock:
+	mutex_unlock(&inode->i_mutex);
+	kfree(ra);
+	balance_dirty_pages_ratelimited_nr(inode->i_mapping, total_dirty);
+	return ret;
+}
+
+static noinline_for_stack
+int relocate_data_extent(struct inode *inode, struct btrfs_key *extent_key)
+{
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
+	struct extent_map *em;
+	u64 start = extent_key->objectid - BTRFS_I(inode)->index_cnt;
+	u64 end = start + extent_key->offset - 1;
+
+	em = alloc_extent_map(GFP_NOFS);
+	em->start = start;
+	em->len = extent_key->offset;
+	em->block_len = extent_key->offset;
+	em->block_start = extent_key->objectid;
+	em->bdev = root->fs_info->fs_devices->latest_bdev;
+	set_bit(EXTENT_FLAG_PINNED, &em->flags);
+
+	/* setup extent map to cheat btrfs_readpage */
+	lock_extent(&BTRFS_I(inode)->io_tree, start, end, GFP_NOFS);
+	while (1) {
+		int ret;
+		spin_lock(&em_tree->lock);
+		ret = add_extent_mapping(em_tree, em);
+		spin_unlock(&em_tree->lock);
+		if (ret != -EEXIST) {
+			free_extent_map(em);
+			break;
+		}
+		btrfs_drop_extent_cache(inode, start, end, 0);
+	}
+	unlock_extent(&BTRFS_I(inode)->io_tree, start, end, GFP_NOFS);
+
+	return relocate_inode_pages(inode, start, extent_key->offset);
+}
+
+#ifdef BTRFS_COMPAT_EXTENT_TREE_V0
+static int get_ref_objectid_v0(struct reloc_control *rc,
+			       struct btrfs_path *path,
+			       struct btrfs_key *extent_key,
+			       u64 *ref_objectid, int *path_change)
+{
+	struct btrfs_key key;
+	struct extent_buffer *leaf;
+	struct btrfs_extent_ref_v0 *ref0;
+	int ret;
+	int slot;
+
+	leaf = path->nodes[0];
+	slot = path->slots[0];
+	while (1) {
+		if (slot >= btrfs_header_nritems(leaf)) {
+			ret = btrfs_next_leaf(rc->extent_root, path);
+			if (ret < 0)
+				return ret;
+			BUG_ON(ret > 0);
+			leaf = path->nodes[0];
+			slot = path->slots[0];
+			if (path_change)
+				*path_change = 1;
+		}
+		btrfs_item_key_to_cpu(leaf, &key, slot);
+		if (key.objectid != extent_key->objectid)
+			return -ENOENT;
+
+		if (key.type != BTRFS_EXTENT_REF_V0_KEY) {
+			slot++;
+			continue;
+		}
+		ref0 = btrfs_item_ptr(leaf, slot,
+				struct btrfs_extent_ref_v0);
+		*ref_objectid = btrfs_ref_objectid_v0(leaf, ref0);
+		break;
+	}
+	return 0;
+}
+#endif
+
+/*
+ * helper to add a tree block to the list.
+ * the major work is getting the generation and level of the block
+ */
+static int add_tree_block(struct reloc_control *rc,
+			  struct btrfs_key *extent_key,
+			  struct btrfs_path *path,
+			  struct rb_root *blocks)
+{
+	struct extent_buffer *eb;
+	struct btrfs_extent_item *ei;
+	struct btrfs_tree_block_info *bi;
+	struct tree_block *block;
+	struct rb_node *rb_node;
+	u32 item_size;
+	int level = -1;
+	int generation;
+
+	eb =  path->nodes[0];
+	item_size = btrfs_item_size_nr(eb, path->slots[0]);
+
+	if (item_size >= sizeof(*ei) + sizeof(*bi)) {
+		ei = btrfs_item_ptr(eb, path->slots[0],
+				struct btrfs_extent_item);
+		bi = (struct btrfs_tree_block_info *)(ei + 1);
+		generation = btrfs_extent_generation(eb, ei);
+		level = btrfs_tree_block_level(eb, bi);
+	} else {
+#ifdef BTRFS_COMPAT_EXTENT_TREE_V0
+		u64 ref_owner;
+		int ret;
+
+		BUG_ON(item_size != sizeof(struct btrfs_extent_item_v0));
+		ret = get_ref_objectid_v0(rc, path, extent_key,
+					  &ref_owner, NULL);
+		BUG_ON(ref_owner >= BTRFS_MAX_LEVEL);
+		level = (int)ref_owner;
+		/* FIXME: get real generation */
+		generation = 0;
+#else
+		BUG();
+#endif
+	}
+
+	btrfs_release_path(rc->extent_root, path);
+
+	BUG_ON(level == -1);
+
+	block = kmalloc(sizeof(*block), GFP_NOFS);
+	if (!block)
+		return -ENOMEM;
+
+	block->bytenr = extent_key->objectid;
+	block->key.objectid = extent_key->offset;
+	block->key.offset = generation;
+	block->level = level;
+	block->key_ready = 0;
+
+	rb_node = tree_insert(blocks, block->bytenr, &block->rb_node);
+	BUG_ON(rb_node);
+
+	return 0;
+}
+
+/*
+ * helper to add tree blocks for backref of type BTRFS_SHARED_DATA_REF_KEY
+ */
+static int __add_tree_block(struct reloc_control *rc,
+			    u64 bytenr, u32 blocksize,
+			    struct rb_root *blocks)
+{
+	struct btrfs_path *path;
+	struct btrfs_key key;
+	int ret;
+
+	if (tree_block_processed(bytenr, blocksize, rc))
+		return 0;
+
+	if (tree_search(blocks, bytenr))
+		return 0;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	key.objectid = bytenr;
+	key.type = BTRFS_EXTENT_ITEM_KEY;
+	key.offset = blocksize;
+
+	path->search_commit_root = 1;
+	path->skip_locking = 1;
+	ret = btrfs_search_slot(NULL, rc->extent_root, &key, path, 0, 0);
+	if (ret < 0)
+		goto out;
+	BUG_ON(ret);
+
+	btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
+	ret = add_tree_block(rc, &key, path, blocks);
+out:
+	btrfs_free_path(path);
+	return ret;
+}
+
+/*
+ * helper to check if the block use full backrefs for pointers in it
+ */
+static int block_use_full_backref(struct reloc_control *rc,
+				  struct extent_buffer *eb)
+{
+	struct btrfs_path *path;
+	struct btrfs_extent_item *ei;
+	struct btrfs_key key;
+	u64 flags;
+	int ret;
+
+	if (btrfs_header_flag(eb, BTRFS_HEADER_FLAG_RELOC) ||
+	    btrfs_header_backref_rev(eb) < BTRFS_MIXED_BACKREF_REV)
+		return 1;
+
+	path = btrfs_alloc_path();
+	BUG_ON(!path);
+
+	key.objectid = eb->start;
+	key.type = BTRFS_EXTENT_ITEM_KEY;
+	key.offset = eb->len;
+
+	path->search_commit_root = 1;
+	path->skip_locking = 1;
+	ret = btrfs_search_slot(NULL, rc->extent_root,
+				&key, path, 0, 0);
+	BUG_ON(ret);
+
+	ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
+			    struct btrfs_extent_item);
+	flags = btrfs_extent_flags(path->nodes[0], ei);
+	BUG_ON(!(flags & BTRFS_EXTENT_FLAG_TREE_BLOCK));
+	if (flags & BTRFS_BLOCK_FLAG_FULL_BACKREF)
+		ret = 1;
+	else
+		ret = 0;
+	btrfs_free_path(path);
+	return ret;
+}
+
+/*
+ * 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
+ */
+static int find_data_references(struct reloc_control *rc,
+				struct btrfs_key *extent_key,
+				struct extent_buffer *leaf,
+				struct btrfs_extent_data_ref *ref,
+				struct rb_root *blocks)
+{
+	struct btrfs_path *path;
+	struct tree_block *block;
+	struct btrfs_root *root;
+	struct btrfs_file_extent_item *fi;
+	struct rb_node *rb_node;
+	struct btrfs_key key;
+	u64 ref_root;
+	u64 ref_objectid;
+	u64 ref_offset;
+	u32 ref_count;
+	u32 nritems;
+	int err = 0;
+	int added = 0;
+	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);
+
+	root = read_fs_root(rc->extent_root->fs_info, ref_root);
+	if (IS_ERR(root)) {
+		err = PTR_ERR(root);
+		goto out;
+	}
+
+	key.objectid = ref_objectid;
+	key.offset = ref_offset;
+	key.type = BTRFS_EXTENT_DATA_KEY;
+
+	path->search_commit_root = 1;
+	path->skip_locking = 1;
+	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+	if (ret < 0) {
+		err = ret;
+		goto out;
+	}
+
+	leaf = path->nodes[0];
+	nritems = btrfs_header_nritems(leaf);
+	/*
+	 * the references in tree blocks that use full backrefs
+	 * are not counted in
+	 */
+	if (block_use_full_backref(rc, leaf))
+		counted = 0;
+	else
+		counted = 1;
+	rb_node = tree_search(blocks, leaf->start);
+	if (rb_node) {
+		if (counted)
+			added = 1;
+		else
+			path->slots[0] = nritems;
+	}
+
+	while (ref_count > 0) {
+		while (path->slots[0] >= nritems) {
+			ret = btrfs_next_leaf(root, path);
+			if (ret < 0) {
+				err = ret;
+				goto out;
+			}
+			if (ret > 0) {
+				WARN_ON(1);
+				goto out;
+			}
+
+			leaf = path->nodes[0];
+			nritems = btrfs_header_nritems(leaf);
+			added = 0;
+
+			if (block_use_full_backref(rc, leaf))
+				counted = 0;
+			else
+				counted = 1;
+			rb_node = tree_search(blocks, leaf->start);
+			if (rb_node) {
+				if (counted)
+					added = 1;
+				else
+					path->slots[0] = nritems;
+			}
+		}
+
+		btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
+		if (key.objectid != ref_objectid ||
+		    key.type != BTRFS_EXTENT_DATA_KEY) {
+			WARN_ON(1);
+			break;
+		}
+
+		fi = btrfs_item_ptr(leaf, path->slots[0],
+				    struct btrfs_file_extent_item);
+
+		if (btrfs_file_extent_type(leaf, fi) ==
+		    BTRFS_FILE_EXTENT_INLINE)
+			goto next;
+
+		if (btrfs_file_extent_disk_bytenr(leaf, fi) !=
+		    extent_key->objectid)
+			goto next;
+
+		key.offset -= btrfs_file_extent_offset(leaf, fi);
+		if (key.offset != ref_offset)
+			goto next;
+
+		if (counted)
+			ref_count--;
+		if (added)
+			goto next;
+
+		if (!tree_block_processed(leaf->start, leaf->len, rc)) {
+			block = kmalloc(sizeof(*block), GFP_NOFS);
+			if (!block) {
+				err = -ENOMEM;
+				break;
+			}
+			block->bytenr = leaf->start;
+			btrfs_item_key_to_cpu(leaf, &block->key, 0);
+			block->level = 0;
+			block->key_ready = 1;
+			rb_node = tree_insert(blocks, block->bytenr,
+					      &block->rb_node);
+			BUG_ON(rb_node);
+		}
+		if (counted)
+			added = 1;
+		else
+			path->slots[0] = nritems;
+next:
+		path->slots[0]++;
+
+	}
+out:
+	btrfs_free_path(path);
+	return err;
+}
+
+/*
+ * hepler to find all tree blocks that reference a given data extent
+ */
+static noinline_for_stack
+int add_data_references(struct reloc_control *rc,
+			struct btrfs_key *extent_key,
+			struct btrfs_path *path,
+			struct rb_root *blocks)
+{
+	struct btrfs_key key;
+	struct extent_buffer *eb;
+	struct btrfs_extent_data_ref *dref;
+	struct btrfs_extent_inline_ref *iref;
+	unsigned long ptr;
+	unsigned long end;
+	u32 blocksize;
+	int ret;
+	int err = 0;
+
+	ret = get_new_location(rc->data_inode, NULL, extent_key->objectid,
+			       extent_key->offset);
+	BUG_ON(ret < 0);
+	if (ret > 0) {
+		/* the relocated data is fragmented */
+		rc->extents_skipped++;
+		btrfs_release_path(rc->extent_root, path);
+		return 0;
+	}
+
+	blocksize = btrfs_level_size(rc->extent_root, 0);
+
+	eb = path->nodes[0];
+	ptr = btrfs_item_ptr_offset(eb, path->slots[0]);
+	end = ptr + btrfs_item_size_nr(eb, path->slots[0]);
+#ifdef BTRFS_COMPAT_EXTENT_TREE_V0
+	if (ptr + sizeof(struct btrfs_extent_item_v0) == end)
+		ptr = end;
+	else
+#endif
+		ptr += sizeof(struct btrfs_extent_item);
+
+	while (ptr < end) {
+		iref = (struct btrfs_extent_inline_ref *)ptr;
+		key.type = btrfs_extent_inline_ref_type(eb, iref);
+		if (key.type == BTRFS_SHARED_DATA_REF_KEY) {
+			key.offset = btrfs_extent_inline_ref_offset(eb, iref);
+			ret = __add_tree_block(rc, key.offset, blocksize,
+					       blocks);
+		} else if (key.type == BTRFS_EXTENT_DATA_REF_KEY) {
+			dref = (struct btrfs_extent_data_ref *)(&iref->offset);
+			ret = find_data_references(rc, extent_key,
+						   eb, dref, blocks);
+		} else {
+			BUG();
+		}
+		ptr += btrfs_extent_inline_ref_size(key.type);
+	}
+	WARN_ON(ptr > end);
+
+	while (1) {
+		cond_resched();
+		eb = path->nodes[0];
+		if (path->slots[0] >= btrfs_header_nritems(eb)) {
+			ret = btrfs_next_leaf(rc->extent_root, path);
+			if (ret < 0) {
+				err = ret;
+				break;
+			}
+			if (ret > 0)
+				break;
+			eb = path->nodes[0];
+		}
+
+		btrfs_item_key_to_cpu(eb, &key, path->slots[0]);
+		if (key.objectid != extent_key->objectid)
+			break;
+
+#ifdef BTRFS_COMPAT_EXTENT_TREE_V0
+		if (key.type == BTRFS_SHARED_DATA_REF_KEY ||
+		    key.type == BTRFS_EXTENT_REF_V0_KEY) {
+#else
+		BUG_ON(key.type == BTRFS_EXTENT_REF_V0_KEY);
+		if (key.type == BTRFS_SHARED_DATA_REF_KEY) {
+#endif
+			ret = __add_tree_block(rc, key.offset, blocksize,
+					       blocks);
+		} else if (key.type == BTRFS_EXTENT_DATA_REF_KEY) {
+			dref = btrfs_item_ptr(eb, path->slots[0],
+					      struct btrfs_extent_data_ref);
+			ret = find_data_references(rc, extent_key,
+						   eb, dref, blocks);
+		} else {
+			ret = 0;
+		}
+		if (ret) {
+			err = ret;
+			break;
+		}
+		path->slots[0]++;
+	}
+	btrfs_release_path(rc->extent_root, path);
+	if (err)
+		free_block_list(blocks);
+	return err;
+}
+
+/*
+ * hepler to find next unprocessed extent
+ */
+static noinline_for_stack
+int find_next_extent(struct btrfs_trans_handle *trans,
+		     struct reloc_control *rc, struct btrfs_path *path)
+{
+	struct btrfs_key key;
+	struct extent_buffer *leaf;
+	u64 start, end, last;
+	int ret;
+
+	last = rc->block_group->key.objectid + rc->block_group->key.offset;
+	while (1) {
+		cond_resched();
+		if (rc->search_start >= last) {
+			ret = 1;
+			break;
+		}
+
+		key.objectid = rc->search_start;
+		key.type = BTRFS_EXTENT_ITEM_KEY;
+		key.offset = 0;
+
+		path->search_commit_root = 1;
+		path->skip_locking = 1;
+		ret = btrfs_search_slot(NULL, rc->extent_root, &key, path,
+					0, 0);
+		if (ret < 0)
+			break;
+next:
+		leaf = path->nodes[0];
+		if (path->slots[0] >= btrfs_header_nritems(leaf)) {
+			ret = btrfs_next_leaf(rc->extent_root, path);
+			if (ret != 0)
+				break;
+			leaf = path->nodes[0];
+		}
+
+		btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
+		if (key.objectid >= last) {
+			ret = 1;
+			break;
+		}
+
+		if (key.type != BTRFS_EXTENT_ITEM_KEY ||
+		    key.objectid + key.offset <= rc->search_start) {
+			path->slots[0]++;
+			goto next;
+		}
+
+		ret = find_first_extent_bit(&rc->processed_blocks,
+					    key.objectid, &start, &end,
+					    EXTENT_DIRTY);
+
+		if (ret == 0 && start <= key.objectid) {
+			btrfs_release_path(rc->extent_root, path);
+			rc->search_start = end + 1;
+		} else {
+			rc->search_start = key.objectid + key.offset;
+			return 0;
+		}
+	}
+	btrfs_release_path(rc->extent_root, path);
+	return ret;
+}
+
+static void set_reloc_control(struct reloc_control *rc)
+{
+	struct btrfs_fs_info *fs_info = rc->extent_root->fs_info;
+	mutex_lock(&fs_info->trans_mutex);
+	fs_info->reloc_ctl = rc;
+	mutex_unlock(&fs_info->trans_mutex);
+}
+
+static void unset_reloc_control(struct reloc_control *rc)
+{
+	struct btrfs_fs_info *fs_info = rc->extent_root->fs_info;
+	mutex_lock(&fs_info->trans_mutex);
+	fs_info->reloc_ctl = NULL;
+	mutex_unlock(&fs_info->trans_mutex);
+}
+
+static int check_extent_flags(u64 flags)
+{
+	if ((flags & BTRFS_EXTENT_FLAG_DATA) &&
+	    (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK))
+		return 1;
+	if (!(flags & BTRFS_EXTENT_FLAG_DATA) &&
+	    !(flags & BTRFS_EXTENT_FLAG_TREE_BLOCK))
+		return 1;
+	if ((flags & BTRFS_EXTENT_FLAG_DATA) &&
+	    (flags & BTRFS_BLOCK_FLAG_FULL_BACKREF))
+		return 1;
+	return 0;
+}
+
+static noinline_for_stack int relocate_block_group(struct reloc_control *rc)
+{
+	struct rb_root blocks = RB_ROOT;
+	struct btrfs_key key;
+	struct btrfs_trans_handle *trans = NULL;
+	struct btrfs_path *path;
+	struct btrfs_extent_item *ei;
+	unsigned long nr;
+	u64 flags;
+	u32 item_size;
+	int ret;
+	int err = 0;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	rc->search_start = rc->block_group->key.objectid;
+	clear_extent_bits(&rc->processed_blocks, 0, (u64)-1, EXTENT_DIRTY,
+			  GFP_NOFS);
+
+	rc->create_reloc_root = 1;
+	set_reloc_control(rc);
+
+	trans = btrfs_start_transaction(rc->extent_root, 1);
+	btrfs_commit_transaction(trans, rc->extent_root);
+
+	while (1) {
+		trans = btrfs_start_transaction(rc->extent_root, 1);
+
+		ret = find_next_extent(trans, rc, path);
+		if (ret < 0)
+			err = ret;
+		if (ret != 0)
+			break;
+
+		rc->extents_found++;
+
+		ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
+				    struct btrfs_extent_item);
+		btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
+		item_size = btrfs_item_size_nr(path->nodes[0],
+					       path->slots[0]);
+		if (item_size >= sizeof(*ei)) {
+			flags = btrfs_extent_flags(path->nodes[0], ei);
+			ret = check_extent_flags(flags);
+			BUG_ON(ret);
+
+		} else {
+#ifdef BTRFS_COMPAT_EXTENT_TREE_V0
+			u64 ref_owner;
+			int path_change = 0;
+
+			BUG_ON(item_size !=
+			       sizeof(struct btrfs_extent_item_v0));
+			ret = get_ref_objectid_v0(rc, path, &key, &ref_owner,
+						  &path_change);
+			if (ref_owner < BTRFS_FIRST_FREE_OBJECTID)
+				flags = BTRFS_EXTENT_FLAG_TREE_BLOCK;
+			else
+				flags = BTRFS_EXTENT_FLAG_DATA;
+
+			if (path_change) {
+				btrfs_release_path(rc->extent_root, path);
+
+				path->search_commit_root = 1;
+				path->skip_locking = 1;
+				ret = btrfs_search_slot(NULL, rc->extent_root,
+							&key, path, 0, 0);
+				if (ret < 0) {
+					err = ret;
+					break;
+				}
+				BUG_ON(ret > 0);
+			}
+#else
+			BUG();
+#endif
+		}
+
+		if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
+			ret = add_tree_block(rc, &key, path, &blocks);
+		} else if (rc->stage == UPDATE_DATA_PTRS &&
+			 (flags & BTRFS_EXTENT_FLAG_DATA)) {
+			ret = add_data_references(rc, &key, path, &blocks);
+		} else {
+			btrfs_release_path(rc->extent_root, path);
+			ret = 0;
+		}
+		if (ret < 0) {
+			err = 0;
+			break;
+		}
+
+		if (!RB_EMPTY_ROOT(&blocks)) {
+			ret = relocate_tree_blocks(trans, rc, &blocks);
+			if (ret < 0) {
+				err = ret;
+				break;
+			}
+		}
+
+		nr = trans->blocks_used;
+		btrfs_end_transaction_throttle(trans, rc->extent_root);
+		trans = NULL;
+		btrfs_btree_balance_dirty(rc->extent_root, nr);
+
+		if (rc->stage == MOVE_DATA_EXTENTS &&
+		    (flags & BTRFS_EXTENT_FLAG_DATA)) {
+			rc->found_file_extent = 1;
+			ret = relocate_data_extent(rc->data_inode, &key);
+			if (ret < 0) {
+				err = ret;
+				break;
+			}
+		}
+	}
+	btrfs_free_path(path);
+
+	if (trans) {
+		nr = trans->blocks_used;
+		btrfs_end_transaction(trans, rc->extent_root);
+		btrfs_btree_balance_dirty(rc->extent_root, nr);
+	}
+
+	rc->create_reloc_root = 0;
+	smp_mb();
+
+	if (rc->extents_found > 0) {
+		trans = btrfs_start_transaction(rc->extent_root, 1);
+		btrfs_commit_transaction(trans, rc->extent_root);
+	}
+
+	merge_reloc_roots(rc);
+
+	unset_reloc_control(rc);
+
+	/* get rid of pinned extents */
+	trans = btrfs_start_transaction(rc->extent_root, 1);
+	btrfs_commit_transaction(trans, rc->extent_root);
+
+	return err;
+}
+
+static int __insert_orphan_inode(struct btrfs_trans_handle *trans,
+				 struct btrfs_root *root,
+				 u64 objectid, u64 size)
+{
+	struct btrfs_path *path;
+	struct btrfs_inode_item *item;
+	struct extent_buffer *leaf;
+	int ret;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	ret = btrfs_insert_empty_inode(trans, root, path, objectid);
+	if (ret)
+		goto out;
+
+	leaf = path->nodes[0];
+	item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_inode_item);
+	memset_extent_buffer(leaf, 0, (unsigned long)item, sizeof(*item));
+	btrfs_set_inode_generation(leaf, item, 1);
+	btrfs_set_inode_size(leaf, item, size);
+	btrfs_set_inode_mode(leaf, item, S_IFREG | 0600);
+	btrfs_set_inode_flags(leaf, item, BTRFS_INODE_NOCOMPRESS);
+	btrfs_mark_buffer_dirty(leaf);
+	btrfs_release_path(root, path);
+out:
+	btrfs_free_path(path);
+	return ret;
+}
+
+/*
+ * helper to create inode for data relocation.
+ * the inode is in data relocation tree and its link count is 0
+ */
+static struct inode *create_reloc_inode(struct btrfs_fs_info *fs_info,
+					struct btrfs_block_group_cache *group)
+{
+	struct inode *inode = NULL;
+	struct btrfs_trans_handle *trans;
+	struct btrfs_root *root;
+	struct btrfs_key key;
+	unsigned long nr;
+	u64 objectid = BTRFS_FIRST_FREE_OBJECTID;
+	int err = 0;
+
+	root = read_fs_root(fs_info, BTRFS_DATA_RELOC_TREE_OBJECTID);
+	if (IS_ERR(root))
+		return ERR_CAST(root);
+
+	trans = btrfs_start_transaction(root, 1);
+	BUG_ON(!trans);
+
+	err = btrfs_find_free_objectid(trans, root, objectid, &objectid);
+	if (err)
+		goto out;
+
+	err = __insert_orphan_inode(trans, root, objectid, group->key.offset);
+	BUG_ON(err);
+
+	err = btrfs_insert_file_extent(trans, root, objectid, 0, 0, 0,
+				       group->key.offset, 0, group->key.offset,
+				       0, 0, 0);
+	BUG_ON(err);
+
+	key.objectid = objectid;
+	key.type = BTRFS_INODE_ITEM_KEY;
+	key.offset = 0;
+	inode = btrfs_iget(root->fs_info->sb, &key, root);
+	BUG_ON(IS_ERR(inode) || is_bad_inode(inode));
+	BTRFS_I(inode)->index_cnt = group->key.objectid;
+
+	err = btrfs_orphan_add(trans, inode);
+out:
+	nr = trans->blocks_used;
+	btrfs_end_transaction(trans, root);
+
+	btrfs_btree_balance_dirty(root, nr);
+	if (err) {
+		if (inode)
+			iput(inode);
+		inode = ERR_PTR(err);
+	}
+	return inode;
+}
+
+/*
+ * function to relocate all extents in a block group.
+ */
+int btrfs_relocate_block_group(struct btrfs_root *extent_root, u64 group_start)
+{
+	struct btrfs_fs_info *fs_info = extent_root->fs_info;
+	struct reloc_control *rc;
+	int ret;
+	int err = 0;
+
+	rc = kzalloc(sizeof(*rc), GFP_NOFS);
+	if (!rc)
+		return -ENOMEM;
+
+	mapping_tree_init(&rc->reloc_root_tree);
+	extent_io_tree_init(&rc->processed_blocks, NULL, GFP_NOFS);
+	INIT_LIST_HEAD(&rc->reloc_roots);
+
+	rc->block_group = btrfs_lookup_block_group(fs_info, group_start);
+	BUG_ON(!rc->block_group);
+
+	btrfs_init_workers(&rc->workers, "relocate",
+			   fs_info->thread_pool_size);
+
+	rc->extent_root = extent_root;
+	btrfs_prepare_block_group_relocation(extent_root, rc->block_group);
+
+	rc->data_inode = create_reloc_inode(fs_info, rc->block_group);
+	if (IS_ERR(rc->data_inode)) {
+		err = PTR_ERR(rc->data_inode);
+		rc->data_inode = NULL;
+		goto out;
+	}
+
+	printk(KERN_INFO "btrfs: relocating block group %llu flags %llu\n",
+	       (unsigned long long)rc->block_group->key.objectid,
+	       (unsigned long long)rc->block_group->flags);
+
+	btrfs_start_delalloc_inodes(fs_info->tree_root);
+	btrfs_wait_ordered_extents(fs_info->tree_root, 0);
+
+	while (1) {
+		mutex_lock(&fs_info->cleaner_mutex);
+		btrfs_clean_old_snapshots(fs_info->tree_root);
+		mutex_unlock(&fs_info->cleaner_mutex);
+
+		rc->extents_found = 0;
+		rc->extents_skipped = 0;
+
+		ret = relocate_block_group(rc);
+		if (ret < 0) {
+			err = ret;
+			break;
+		}
+
+		if (rc->extents_found == 0)
+			break;
+
+		printk(KERN_INFO "btrfs: found %llu extents\n",
+			(unsigned long long)rc->extents_found);
+
+		if (rc->stage == MOVE_DATA_EXTENTS && rc->found_file_extent) {
+			btrfs_wait_ordered_range(rc->data_inode, 0, (u64)-1);
+			invalidate_mapping_pages(rc->data_inode->i_mapping,
+						 0, -1);
+			rc->stage = UPDATE_DATA_PTRS;
+		} else if (rc->stage == UPDATE_DATA_PTRS &&
+			   rc->extents_skipped >= rc->extents_found) {
+			iput(rc->data_inode);
+			rc->data_inode = create_reloc_inode(fs_info,
+							    rc->block_group);
+			if (IS_ERR(rc->data_inode)) {
+				err = PTR_ERR(rc->data_inode);
+				rc->data_inode = NULL;
+				break;
+			}
+			rc->stage = MOVE_DATA_EXTENTS;
+			rc->found_file_extent = 0;
+		}
+	}
+
+	filemap_fdatawrite_range(fs_info->btree_inode->i_mapping,
+				 rc->block_group->key.objectid,
+				 rc->block_group->key.objectid +
+				 rc->block_group->key.offset - 1);
+
+	WARN_ON(rc->block_group->pinned > 0);
+	WARN_ON(rc->block_group->reserved > 0);
+	WARN_ON(btrfs_block_group_used(&rc->block_group->item) > 0);
+out:
+	iput(rc->data_inode);
+	btrfs_stop_workers(&rc->workers);
+	btrfs_put_block_group(rc->block_group);
+	kfree(rc);
+	return err;
+}
+
+/*
+ * recover relocation interrupted by system crash.
+ *
+ * this function resumes merging reloc trees with corresponding fs trees.
+ * this is important for keeping the sharing of tree blocks
+ */
+int btrfs_recover_relocation(struct btrfs_root *root)
+{
+	LIST_HEAD(reloc_roots);
+	struct btrfs_key key;
+	struct btrfs_root *fs_root;
+	struct btrfs_root *reloc_root;
+	struct btrfs_path *path;
+	struct extent_buffer *leaf;
+	struct reloc_control *rc = NULL;
+	struct btrfs_trans_handle *trans;
+	int ret;
+	int err = 0;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	key.objectid = BTRFS_TREE_RELOC_OBJECTID;
+	key.type = BTRFS_ROOT_ITEM_KEY;
+	key.offset = (u64)-1;
+
+	while (1) {
+		ret = btrfs_search_slot(NULL, root->fs_info->tree_root, &key,
+					path, 0, 0);
+		if (ret < 0) {
+			err = ret;
+			goto out;
+		}
+		if (ret > 0) {
+			if (path->slots[0] == 0)
+				break;
+			path->slots[0]--;
+		}
+		leaf = path->nodes[0];
+		btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
+		btrfs_release_path(root->fs_info->tree_root, path);
+
+		if (key.objectid != BTRFS_TREE_RELOC_OBJECTID ||
+		    key.type != BTRFS_ROOT_ITEM_KEY)
+			break;
+
+		reloc_root = btrfs_read_fs_root_no_radix(root, &key);
+		if (IS_ERR(reloc_root)) {
+			err = PTR_ERR(reloc_root);
+			goto out;
+		}
+
+		list_add(&reloc_root->root_list, &reloc_roots);
+
+		if (btrfs_root_refs(&reloc_root->root_item) > 0) {
+			fs_root = read_fs_root(root->fs_info,
+					       reloc_root->root_key.offset);
+			if (IS_ERR(fs_root)) {
+				err = PTR_ERR(fs_root);
+				goto out;
+			}
+		}
+
+		if (key.offset == 0)
+			break;
+
+		key.offset--;
+	}
+	btrfs_release_path(root->fs_info->tree_root, path);
+
+	if (list_empty(&reloc_roots))
+		goto out;
+
+	rc = kzalloc(sizeof(*rc), GFP_NOFS);
+	if (!rc) {
+		err = -ENOMEM;
+		goto out;
+	}
+
+	mapping_tree_init(&rc->reloc_root_tree);
+	INIT_LIST_HEAD(&rc->reloc_roots);
+	btrfs_init_workers(&rc->workers, "relocate",
+			   root->fs_info->thread_pool_size);
+	rc->extent_root = root->fs_info->extent_root;
+
+	set_reloc_control(rc);
+
+	while (!list_empty(&reloc_roots)) {
+		reloc_root = list_entry(reloc_roots.next,
+					struct btrfs_root, root_list);
+		list_del(&reloc_root->root_list);
+
+		if (btrfs_root_refs(&reloc_root->root_item) == 0) {
+			list_add_tail(&reloc_root->root_list,
+				      &rc->reloc_roots);
+			continue;
+		}
+
+		fs_root = read_fs_root(root->fs_info,
+				       reloc_root->root_key.offset);
+		BUG_ON(IS_ERR(fs_root));
+
+		__add_reloc_root(reloc_root);
+		fs_root->reloc_root = reloc_root;
+	}
+
+	trans = btrfs_start_transaction(rc->extent_root, 1);
+	btrfs_commit_transaction(trans, rc->extent_root);
+
+	merge_reloc_roots(rc);
+
+	unset_reloc_control(rc);
+
+	trans = btrfs_start_transaction(rc->extent_root, 1);
+	btrfs_commit_transaction(trans, rc->extent_root);
+out:
+	if (rc) {
+		btrfs_stop_workers(&rc->workers);
+		kfree(rc);
+	}
+	while (!list_empty(&reloc_roots)) {
+		reloc_root = list_entry(reloc_roots.next,
+					struct btrfs_root, root_list);
+		list_del(&reloc_root->root_list);
+		free_extent_buffer(reloc_root->node);
+		free_extent_buffer(reloc_root->commit_root);
+		kfree(reloc_root);
+	}
+	btrfs_free_path(path);
+
+	if (err == 0) {
+		/* cleanup orphan inode in data relocation tree */
+		fs_root = read_fs_root(root->fs_info,
+				       BTRFS_DATA_RELOC_TREE_OBJECTID);
+		if (IS_ERR(fs_root))
+			err = PTR_ERR(fs_root);
+	}
+	return err;
+}
+
+/*
+ * helper to add ordered checksum for data relocation.
+ *
+ * cloning checksum properly handles the nodatasum extents.
+ * it also saves CPU time to re-calculate the checksum.
+ */
+int btrfs_reloc_clone_csums(struct inode *inode, u64 file_pos, u64 len)
+{
+	struct btrfs_ordered_sum *sums;
+	struct btrfs_sector_sum *sector_sum;
+	struct btrfs_ordered_extent *ordered;
+	struct btrfs_root *root = BTRFS_I(inode)->root;
+	size_t offset;
+	int ret;
+	u64 disk_bytenr;
+	LIST_HEAD(list);
+
+	ordered = btrfs_lookup_ordered_extent(inode, file_pos);
+	BUG_ON(ordered->file_offset != file_pos || ordered->len != len);
+
+	disk_bytenr = file_pos + BTRFS_I(inode)->index_cnt;
+	ret = btrfs_lookup_csums_range(root->fs_info->csum_root, disk_bytenr,
+				       disk_bytenr + len - 1, &list);
+
+	while (!list_empty(&list)) {
+		sums = list_entry(list.next, struct btrfs_ordered_sum, list);
+		list_del_init(&sums->list);
+
+		sector_sum = sums->sums;
+		sums->bytenr = ordered->start;
+
+		offset = 0;
+		while (offset < sums->len) {
+			sector_sum->bytenr += ordered->start - disk_bytenr;
+			sector_sum++;
+			offset += root->sectorsize;
+		}
+
+		btrfs_add_ordered_sum(inode, ordered, sums);
+	}
+	btrfs_put_ordered_extent(ordered);
+	return 0;
+}

fs/btrfs/root-tree.c

@@ -111,6 +111,15 @@ out:
 	return ret;
 }
 
+int btrfs_set_root_node(struct btrfs_root_item *item,
+			struct extent_buffer *node)
+{
+	btrfs_set_root_bytenr(item, node->start);
+	btrfs_set_root_level(item, btrfs_header_level(node));
+	btrfs_set_root_generation(item, btrfs_header_generation(node));
+	return 0;
+}
+
 /*
  * copy the data in 'item' into the btree
  */
@@ -164,8 +173,7 @@ int btrfs_insert_root(struct btrfs_trans_handle *trans, struct btrfs_root
  * offset lower than the latest root.  They need to be queued for deletion to
  * finish what was happening when we crashed.
  */
-int btrfs_find_dead_roots(struct btrfs_root *root, u64 objectid,
-			  struct btrfs_root *latest)
+int btrfs_find_dead_roots(struct btrfs_root *root, u64 objectid)
 {
 	struct btrfs_root *dead_root;
 	struct btrfs_item *item;
@@ -227,10 +235,7 @@ again:
 			goto err;
 		}
 
-		if (objectid == BTRFS_TREE_RELOC_OBJECTID)
-			ret = btrfs_add_dead_reloc_root(dead_root);
-		else
-			ret = btrfs_add_dead_root(dead_root, latest);
+		ret = btrfs_add_dead_root(dead_root);
 		if (ret)
 			goto err;
 		goto again;

fs/btrfs/super.c

@@ -52,7 +52,6 @@
 #include "export.h"
 #include "compression.h"
 
-
 static struct super_operations btrfs_super_ops;
 
 static void btrfs_put_super(struct super_block *sb)
@@ -67,8 +66,8 @@ static void btrfs_put_super(struct super_block *sb)
 enum {
 	Opt_degraded, Opt_subvol, Opt_device, Opt_nodatasum, Opt_nodatacow,
 	Opt_max_extent, Opt_max_inline, Opt_alloc_start, Opt_nobarrier,
-	Opt_ssd, Opt_thread_pool, Opt_noacl,  Opt_compress, Opt_notreelog,
-	Opt_ratio, Opt_flushoncommit, Opt_err,
+	Opt_ssd, Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl,
+	Opt_compress, Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_err,
 };
 
 static match_table_t tokens = {
@@ -84,6 +83,8 @@ static match_table_t tokens = {
 	{Opt_thread_pool, "thread_pool=%d"},
 	{Opt_compress, "compress"},
 	{Opt_ssd, "ssd"},
+	{Opt_ssd_spread, "ssd_spread"},
+	{Opt_nossd, "nossd"},
 	{Opt_noacl, "noacl"},
 	{Opt_notreelog, "notreelog"},
 	{Opt_flushoncommit, "flushoncommit"},
@@ -158,7 +159,7 @@ int btrfs_parse_options(struct btrfs_root *root, char *options)
 			 */
 			break;
 		case Opt_nodatasum:
-			printk(KERN_INFO "btrfs: setting nodatacsum\n");
+			printk(KERN_INFO "btrfs: setting nodatasum\n");
 			btrfs_set_opt(info->mount_opt, NODATASUM);
 			break;
 		case Opt_nodatacow:
@@ -174,6 +175,19 @@ int btrfs_parse_options(struct btrfs_root *root, char *options)
 			printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
 			btrfs_set_opt(info->mount_opt, SSD);
 			break;
+		case Opt_ssd_spread:
+			printk(KERN_INFO "btrfs: use spread ssd "
+			       "allocation scheme\n");
+			btrfs_set_opt(info->mount_opt, SSD);
+			btrfs_set_opt(info->mount_opt, SSD_SPREAD);
+			break;
+		case Opt_nossd:
+			printk(KERN_INFO "btrfs: not using ssd allocation "
+			       "scheme\n");
+			btrfs_set_opt(info->mount_opt, NOSSD);
+			btrfs_clear_opt(info->mount_opt, SSD);
+			btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
+			break;
 		case Opt_nobarrier:
 			printk(KERN_INFO "btrfs: turning off barriers\n");
 			btrfs_set_opt(info->mount_opt, NOBARRIER);
@@ -322,7 +336,7 @@ static int btrfs_fill_super(struct super_block *sb,
 	struct dentry *root_dentry;
 	struct btrfs_super_block *disk_super;
 	struct btrfs_root *tree_root;
-	struct btrfs_inode *bi;
+	struct btrfs_key key;
 	int err;
 
 	sb->s_maxbytes = MAX_LFS_FILESIZE;
@@ -341,23 +355,15 @@ static int btrfs_fill_super(struct super_block *sb,
 	}
 	sb->s_fs_info = tree_root;
 	disk_super = &tree_root->fs_info->super_copy;
-	inode = btrfs_iget_locked(sb, BTRFS_FIRST_FREE_OBJECTID,
-				  tree_root->fs_info->fs_root);
-	bi = BTRFS_I(inode);
-	bi->location.objectid = inode->i_ino;
-	bi->location.offset = 0;
-	bi->root = tree_root->fs_info->fs_root;
 
-	btrfs_set_key_type(&bi->location, BTRFS_INODE_ITEM_KEY);
-
-	if (!inode) {
-		err = -ENOMEM;
+	key.objectid = BTRFS_FIRST_FREE_OBJECTID;
+	key.type = BTRFS_INODE_ITEM_KEY;
+	key.offset = 0;
+	inode = btrfs_iget(sb, &key, tree_root->fs_info->fs_root);
+	if (IS_ERR(inode)) {
+		err = PTR_ERR(inode);
 		goto fail_close;
 	}
-	if (inode->i_state & I_NEW) {
-		btrfs_read_locked_inode(inode);
-		unlock_new_inode(inode);
-	}
 
 	root_dentry = d_alloc_root(inode);
 	if (!root_dentry) {
@@ -433,7 +439,11 @@ static int btrfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
 		seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
 	if (btrfs_test_opt(root, COMPRESS))
 		seq_puts(seq, ",compress");
-	if (btrfs_test_opt(root, SSD))
+	if (btrfs_test_opt(root, NOSSD))
+		seq_puts(seq, ",nossd");
+	if (btrfs_test_opt(root, SSD_SPREAD))
+		seq_puts(seq, ",ssd_spread");
+	else if (btrfs_test_opt(root, SSD))
 		seq_puts(seq, ",ssd");
 	if (btrfs_test_opt(root, NOTREELOG))
 		seq_puts(seq, ",notreelog");
@@ -584,7 +594,8 @@ static int btrfs_remount(struct super_block *sb, int *flags, char *data)
 		if (btrfs_super_log_root(&root->fs_info->super_copy) != 0)
 			return -EINVAL;
 
-		ret = btrfs_cleanup_reloc_trees(root);
+		/* recover relocation */
+		ret = btrfs_recover_relocation(root);
 		WARN_ON(ret);
 
 		ret = btrfs_cleanup_fs_roots(root->fs_info);

fs/btrfs/transaction.c

@@ -25,7 +25,6 @@
 #include "disk-io.h"
 #include "transaction.h"
 #include "locking.h"
-#include "ref-cache.h"
 #include "tree-log.h"
 
 #define BTRFS_ROOT_TRANS_TAG 0
@@ -94,45 +93,37 @@ static noinline int join_transaction(struct btrfs_root *root)
  * to make sure the old root from before we joined the transaction is deleted
  * when the transaction commits
  */
-noinline int btrfs_record_root_in_trans(struct btrfs_root *root)
+static noinline int record_root_in_trans(struct btrfs_trans_handle *trans,
+					 struct btrfs_root *root)
 {
-	struct btrfs_dirty_root *dirty;
-	u64 running_trans_id = root->fs_info->running_transaction->transid;
-	if (root->ref_cows && root->last_trans < running_trans_id) {
+	if (root->ref_cows && root->last_trans < trans->transid) {
 		WARN_ON(root == root->fs_info->extent_root);
-		if (root->root_item.refs != 0) {
-			radix_tree_tag_set(&root->fs_info->fs_roots_radix,
-				   (unsigned long)root->root_key.objectid,
-				   BTRFS_ROOT_TRANS_TAG);
-
-			dirty = kmalloc(sizeof(*dirty), GFP_NOFS);
-			BUG_ON(!dirty);
-			dirty->root = kmalloc(sizeof(*dirty->root), GFP_NOFS);
-			BUG_ON(!dirty->root);
-			dirty->latest_root = root;
-			INIT_LIST_HEAD(&dirty->list);
-
-			root->commit_root = btrfs_root_node(root);
-
-			memcpy(dirty->root, root, sizeof(*root));
-			spin_lock_init(&dirty->root->node_lock);
-			spin_lock_init(&dirty->root->list_lock);
-			mutex_init(&dirty->root->objectid_mutex);
-			mutex_init(&dirty->root->log_mutex);
-			INIT_LIST_HEAD(&dirty->root->dead_list);
-			dirty->root->node = root->commit_root;
-			dirty->root->commit_root = NULL;
+		WARN_ON(root->root_item.refs == 0);
+		WARN_ON(root->commit_root != root->node);
+
+		radix_tree_tag_set(&root->fs_info->fs_roots_radix,
+			   (unsigned long)root->root_key.objectid,
+			   BTRFS_ROOT_TRANS_TAG);
+		root->last_trans = trans->transid;
+		btrfs_init_reloc_root(trans, root);
+	}
+	return 0;
+}
 
-			spin_lock(&root->list_lock);
-			list_add(&dirty->root->dead_list, &root->dead_list);
-			spin_unlock(&root->list_lock);
+int btrfs_record_root_in_trans(struct btrfs_trans_handle *trans,
+			       struct btrfs_root *root)
+{
+	if (!root->ref_cows)
+		return 0;
 
-			root->dirty_root = dirty;
-		} else {
-			WARN_ON(1);
-		}
-		root->last_trans = running_trans_id;
+	mutex_lock(&root->fs_info->trans_mutex);
+	if (root->last_trans == trans->transid) {
+		mutex_unlock(&root->fs_info->trans_mutex);
+		return 0;
 	}
+
+	record_root_in_trans(trans, root);
+	mutex_unlock(&root->fs_info->trans_mutex);
 	return 0;
 }
 
@@ -181,7 +172,6 @@ static struct btrfs_trans_handle *start_transaction(struct btrfs_root *root,
 	ret = join_transaction(root);
 	BUG_ON(ret);
 
-	btrfs_record_root_in_trans(root);
 	h->transid = root->fs_info->running_transaction->transid;
 	h->transaction = root->fs_info->running_transaction;
 	h->blocks_reserved = num_blocks;
@@ -192,6 +182,7 @@ static struct btrfs_trans_handle *start_transaction(struct btrfs_root *root,
 	h->delayed_ref_updates = 0;
 
 	root->fs_info->running_transaction->use_count++;
+	record_root_in_trans(h, root);
 	mutex_unlock(&root->fs_info->trans_mutex);
 	return h;
 }
@@ -233,6 +224,7 @@ static noinline int wait_for_commit(struct btrfs_root *root,
 	return 0;
 }
 
+#if 0
 /*
  * rate limit against the drop_snapshot code.  This helps to slow down new
  * operations if the drop_snapshot code isn't able to keep up.
@@ -273,6 +265,7 @@ harder:
 			goto harder;
 	}
 }
+#endif
 
 void btrfs_throttle(struct btrfs_root *root)
 {
@@ -280,7 +273,6 @@ void btrfs_throttle(struct btrfs_root *root)
 	if (!root->fs_info->open_ioctl_trans)
 		wait_current_trans(root);
 	mutex_unlock(&root->fs_info->trans_mutex);
-	throttle_on_drops(root);
 }
 
 static int __btrfs_end_transaction(struct btrfs_trans_handle *trans,
@@ -323,9 +315,6 @@ static int __btrfs_end_transaction(struct btrfs_trans_handle *trans,
 	memset(trans, 0, sizeof(*trans));
 	kmem_cache_free(btrfs_trans_handle_cachep, trans);
 
-	if (throttle)
-		throttle_on_drops(root);
-
 	return 0;
 }
 
@@ -462,12 +451,8 @@ static int update_cowonly_root(struct btrfs_trans_handle *trans,
 		old_root_bytenr = btrfs_root_bytenr(&root->root_item);
 		if (old_root_bytenr == root->node->start)
 			break;
-		btrfs_set_root_bytenr(&root->root_item,
-				       root->node->start);
-		btrfs_set_root_level(&root->root_item,
-				     btrfs_header_level(root->node));
-		btrfs_set_root_generation(&root->root_item, trans->transid);
 
+		btrfs_set_root_node(&root->root_item, root->node);
 		ret = btrfs_update_root(trans, tree_root,
 					&root->root_key,
 					&root->root_item);
@@ -477,14 +462,16 @@ static int update_cowonly_root(struct btrfs_trans_handle *trans,
 		ret = btrfs_run_delayed_refs(trans, root, (unsigned long)-1);
 		BUG_ON(ret);
 	}
+	free_extent_buffer(root->commit_root);
+	root->commit_root = btrfs_root_node(root);
 	return 0;
 }
 
 /*
  * update all the cowonly tree roots on disk
  */
-int btrfs_commit_tree_roots(struct btrfs_trans_handle *trans,
-			    struct btrfs_root *root)
+static noinline int commit_cowonly_roots(struct btrfs_trans_handle *trans,
+					 struct btrfs_root *root)
 {
 	struct btrfs_fs_info *fs_info = root->fs_info;
 	struct list_head *next;
@@ -520,118 +507,54 @@ int btrfs_commit_tree_roots(struct btrfs_trans_handle *trans,
  * a dirty root struct and adds it into the list of dead roots that need to
  * be deleted
  */
-int btrfs_add_dead_root(struct btrfs_root *root, struct btrfs_root *latest)
+int btrfs_add_dead_root(struct btrfs_root *root)
 {
-	struct btrfs_dirty_root *dirty;
-
-	dirty = kmalloc(sizeof(*dirty), GFP_NOFS);
-	if (!dirty)
-		return -ENOMEM;
-	dirty->root = root;
-	dirty->latest_root = latest;
-
 	mutex_lock(&root->fs_info->trans_mutex);
-	list_add(&dirty->list, &latest->fs_info->dead_roots);
+	list_add(&root->root_list, &root->fs_info->dead_roots);
 	mutex_unlock(&root->fs_info->trans_mutex);
 	return 0;
 }
 
 /*
- * at transaction commit time we need to schedule the old roots for
- * deletion via btrfs_drop_snapshot.  This runs through all the
- * reference counted roots that were modified in the current
- * transaction and puts them into the drop list
+ * update all the cowonly tree roots on disk
  */
-static noinline int add_dirty_roots(struct btrfs_trans_handle *trans,
-				    struct radix_tree_root *radix,
-				    struct list_head *list)
+static noinline int commit_fs_roots(struct btrfs_trans_handle *trans,
+				    struct btrfs_root *root)
 {
-	struct btrfs_dirty_root *dirty;
 	struct btrfs_root *gang[8];
-	struct btrfs_root *root;
+	struct btrfs_fs_info *fs_info = root->fs_info;
 	int i;
 	int ret;
 	int err = 0;
-	u32 refs;
 
 	while (1) {
-		ret = radix_tree_gang_lookup_tag(radix, (void **)gang, 0,
+		ret = radix_tree_gang_lookup_tag(&fs_info->fs_roots_radix,
+						 (void **)gang, 0,
 						 ARRAY_SIZE(gang),
 						 BTRFS_ROOT_TRANS_TAG);
 		if (ret == 0)
 			break;
 		for (i = 0; i < ret; i++) {
 			root = gang[i];
-			radix_tree_tag_clear(radix,
-				     (unsigned long)root->root_key.objectid,
-				     BTRFS_ROOT_TRANS_TAG);
-
-			BUG_ON(!root->ref_tree);
-			dirty = root->dirty_root;
+			radix_tree_tag_clear(&fs_info->fs_roots_radix,
+					(unsigned long)root->root_key.objectid,
+					BTRFS_ROOT_TRANS_TAG);
 
 			btrfs_free_log(trans, root);
-			btrfs_free_reloc_root(trans, root);
-
-			if (root->commit_root == root->node) {
-				WARN_ON(root->node->start !=
-					btrfs_root_bytenr(&root->root_item));
-
-				free_extent_buffer(root->commit_root);
-				root->commit_root = NULL;
-				root->dirty_root = NULL;
-
-				spin_lock(&root->list_lock);
-				list_del_init(&dirty->root->dead_list);
-				spin_unlock(&root->list_lock);
+			btrfs_update_reloc_root(trans, root);
 
-				kfree(dirty->root);
-				kfree(dirty);
-
-				/* make sure to update the root on disk
-				 * so we get any updates to the block used
-				 * counts
-				 */
-				err = btrfs_update_root(trans,
-						root->fs_info->tree_root,
-						&root->root_key,
-						&root->root_item);
+			if (root->commit_root == root->node)
 				continue;
-			}
 
-			memset(&root->root_item.drop_progress, 0,
-			       sizeof(struct btrfs_disk_key));
-			root->root_item.drop_level = 0;
-			root->commit_root = NULL;
-			root->dirty_root = NULL;
-			root->root_key.offset = root->fs_info->generation;
-			btrfs_set_root_bytenr(&root->root_item,
-					      root->node->start);
-			btrfs_set_root_level(&root->root_item,
-					     btrfs_header_level(root->node));
-			btrfs_set_root_generation(&root->root_item,
-						  root->root_key.offset);
-
-			err = btrfs_insert_root(trans, root->fs_info->tree_root,
+			free_extent_buffer(root->commit_root);
+			root->commit_root = btrfs_root_node(root);
+
+			btrfs_set_root_node(&root->root_item, root->node);
+			err = btrfs_update_root(trans, fs_info->tree_root,
 						&root->root_key,
 						&root->root_item);
 			if (err)
 				break;
-
-			refs = btrfs_root_refs(&dirty->root->root_item);
-			btrfs_set_root_refs(&dirty->root->root_item, refs - 1);
-			err = btrfs_update_root(trans, root->fs_info->tree_root,
-						&dirty->root->root_key,
-						&dirty->root->root_item);
-
-			BUG_ON(err);
-			if (refs == 1) {
-				list_add(&dirty->list, list);
-			} else {
-				WARN_ON(1);
-				free_extent_buffer(dirty->root->node);
-				kfree(dirty->root);
-				kfree(dirty);
-			}
 		}
 	}
 	return err;
@@ -688,12 +611,8 @@ static noinline int wait_transaction_pre_flush(struct btrfs_fs_info *info)
 				TASK_UNINTERRUPTIBLE);
 		mutex_unlock(&info->trans_mutex);
 
-		atomic_dec(&info->throttles);
-		wake_up(&info->transaction_throttle);
-
 		schedule();
 
-		atomic_inc(&info->throttles);
 		mutex_lock(&info->trans_mutex);
 		finish_wait(&info->transaction_wait, &wait);
 	}
@@ -705,111 +624,61 @@ static noinline int wait_transaction_pre_flush(struct btrfs_fs_info *info)
  * Given a list of roots that need to be deleted, call btrfs_drop_snapshot on
  * all of them
  */
-static noinline int drop_dirty_roots(struct btrfs_root *tree_root,
-				     struct list_head *list)
+int btrfs_drop_dead_root(struct btrfs_root *root)
 {
-	struct btrfs_dirty_root *dirty;
 	struct btrfs_trans_handle *trans;
+	struct btrfs_root *tree_root = root->fs_info->tree_root;
 	unsigned long nr;
-	u64 num_bytes;
-	u64 bytes_used;
-	u64 max_useless;
-	int ret = 0;
-	int err;
-
-	while (!list_empty(list)) {
-		struct btrfs_root *root;
-
-		dirty = list_entry(list->prev, struct btrfs_dirty_root, list);
-		list_del_init(&dirty->list);
-
-		num_bytes = btrfs_root_used(&dirty->root->root_item);
-		root = dirty->latest_root;
-		atomic_inc(&root->fs_info->throttles);
-
-		while (1) {
-			/*
-			 * we don't want to jump in and create a bunch of
-			 * delayed refs if the transaction is starting to close
-			 */
-			wait_transaction_pre_flush(tree_root->fs_info);
-			trans = btrfs_start_transaction(tree_root, 1);
-
-			/*
-			 * we've joined a transaction, make sure it isn't
-			 * closing right now
-			 */
-			if (trans->transaction->delayed_refs.flushing) {
-				btrfs_end_transaction(trans, tree_root);
-				continue;
-			}
-
-			mutex_lock(&root->fs_info->drop_mutex);
-			ret = btrfs_drop_snapshot(trans, dirty->root);
-			if (ret != -EAGAIN)
-				break;
-			mutex_unlock(&root->fs_info->drop_mutex);
+	int ret;
 
-			err = btrfs_update_root(trans,
-					tree_root,
-					&dirty->root->root_key,
-					&dirty->root->root_item);
-			if (err)
-				ret = err;
-			nr = trans->blocks_used;
-			ret = btrfs_end_transaction(trans, tree_root);
-			BUG_ON(ret);
+	while (1) {
+		/*
+		 * we don't want to jump in and create a bunch of
+		 * delayed refs if the transaction is starting to close
+		 */
+		wait_transaction_pre_flush(tree_root->fs_info);
+		trans = btrfs_start_transaction(tree_root, 1);
 
-			btrfs_btree_balance_dirty(tree_root, nr);
-			cond_resched();
+		/*
+		 * we've joined a transaction, make sure it isn't
+		 * closing right now
+		 */
+		if (trans->transaction->delayed_refs.flushing) {
+			btrfs_end_transaction(trans, tree_root);
+			continue;
 		}
-		BUG_ON(ret);
-		atomic_dec(&root->fs_info->throttles);
-		wake_up(&root->fs_info->transaction_throttle);
 
-		num_bytes -= btrfs_root_used(&dirty->root->root_item);
-		bytes_used = btrfs_root_used(&root->root_item);
-		if (num_bytes) {
-			mutex_lock(&root->fs_info->trans_mutex);
-			btrfs_record_root_in_trans(root);
-			mutex_unlock(&root->fs_info->trans_mutex);
-			btrfs_set_root_used(&root->root_item,
-					    bytes_used - num_bytes);
-		}
+		ret = btrfs_drop_snapshot(trans, root);
+		if (ret != -EAGAIN)
+			break;
 
-		ret = btrfs_del_root(trans, tree_root, &dirty->root->root_key);
-		if (ret) {
-			BUG();
+		ret = btrfs_update_root(trans, tree_root,
+					&root->root_key,
+					&root->root_item);
+		if (ret)
 			break;
-		}
-		mutex_unlock(&root->fs_info->drop_mutex);
-
-		spin_lock(&root->list_lock);
-		list_del_init(&dirty->root->dead_list);
-		if (!list_empty(&root->dead_list)) {
-			struct btrfs_root *oldest;
-			oldest = list_entry(root->dead_list.prev,
-					    struct btrfs_root, dead_list);
-			max_useless = oldest->root_key.offset - 1;
-		} else {
-			max_useless = root->root_key.offset - 1;
-		}
-		spin_unlock(&root->list_lock);
 
 		nr = trans->blocks_used;
 		ret = btrfs_end_transaction(trans, tree_root);
 		BUG_ON(ret);
 
-		ret = btrfs_remove_leaf_refs(root, max_useless, 0);
-		BUG_ON(ret);
-
-		free_extent_buffer(dirty->root->node);
-		kfree(dirty->root);
-		kfree(dirty);
-
 		btrfs_btree_balance_dirty(tree_root, nr);
 		cond_resched();
 	}
+	BUG_ON(ret);
+
+	ret = btrfs_del_root(trans, tree_root, &root->root_key);
+	BUG_ON(ret);
+
+	nr = trans->blocks_used;
+	ret = btrfs_end_transaction(trans, tree_root);
+	BUG_ON(ret);
+
+	free_extent_buffer(root->node);
+	free_extent_buffer(root->commit_root);
+	kfree(root);
+
+	btrfs_btree_balance_dirty(tree_root, nr);
 	return ret;
 }
 
@@ -839,24 +708,23 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans,
 	if (ret)
 		goto fail;
 
-	btrfs_record_root_in_trans(root);
+	record_root_in_trans(trans, root);
 	btrfs_set_root_last_snapshot(&root->root_item, trans->transid);
 	memcpy(new_root_item, &root->root_item, sizeof(*new_root_item));
 
 	key.objectid = objectid;
-	key.offset = trans->transid;
+	key.offset = 0;
 	btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
 
 	old = btrfs_lock_root_node(root);
 	btrfs_cow_block(trans, root, old, NULL, 0, &old);
+	btrfs_set_lock_blocking(old);
 
 	btrfs_copy_root(trans, root, old, &tmp, objectid);
 	btrfs_tree_unlock(old);
 	free_extent_buffer(old);
 
-	btrfs_set_root_bytenr(new_root_item, tmp->start);
-	btrfs_set_root_level(new_root_item, btrfs_header_level(tmp));
-	btrfs_set_root_generation(new_root_item, trans->transid);
+	btrfs_set_root_node(new_root_item, tmp);
 	ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
 				new_root_item);
 	btrfs_tree_unlock(tmp);
@@ -964,6 +832,24 @@ static noinline int finish_pending_snapshots(struct btrfs_trans_handle *trans,
 	return 0;
 }
 
+static void update_super_roots(struct btrfs_root *root)
+{
+	struct btrfs_root_item *root_item;
+	struct btrfs_super_block *super;
+
+	super = &root->fs_info->super_copy;
+
+	root_item = &root->fs_info->chunk_root->root_item;
+	super->chunk_root = root_item->bytenr;
+	super->chunk_root_generation = root_item->generation;
+	super->chunk_root_level = root_item->level;
+
+	root_item = &root->fs_info->tree_root->root_item;
+	super->root = root_item->bytenr;
+	super->generation = root_item->generation;
+	super->root_level = root_item->level;
+}
+
 int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 			     struct btrfs_root *root)
 {
@@ -971,8 +857,6 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 	unsigned long timeout = 1;
 	struct btrfs_transaction *cur_trans;
 	struct btrfs_transaction *prev_trans = NULL;
-	struct btrfs_root *chunk_root = root->fs_info->chunk_root;
-	struct list_head dirty_fs_roots;
 	struct extent_io_tree *pinned_copy;
 	DEFINE_WAIT(wait);
 	int ret;
@@ -999,7 +883,6 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 	BUG_ON(ret);
 
 	mutex_lock(&root->fs_info->trans_mutex);
-	INIT_LIST_HEAD(&dirty_fs_roots);
 	if (cur_trans->in_commit) {
 		cur_trans->use_count++;
 		mutex_unlock(&root->fs_info->trans_mutex);
@@ -1105,41 +988,36 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 	 * with the tree-log code.
 	 */
 	mutex_lock(&root->fs_info->tree_log_mutex);
-	/*
-	 * keep tree reloc code from adding new reloc trees
-	 */
-	mutex_lock(&root->fs_info->tree_reloc_mutex);
-
 
-	ret = add_dirty_roots(trans, &root->fs_info->fs_roots_radix,
-			      &dirty_fs_roots);
+	ret = commit_fs_roots(trans, root);
 	BUG_ON(ret);
 
-	/* add_dirty_roots gets rid of all the tree log roots, it is now
+	/* commit_fs_roots gets rid of all the tree log roots, it is now
 	 * safe to free the root of tree log roots
 	 */
 	btrfs_free_log_root_tree(trans, root->fs_info);
 
-	ret = btrfs_commit_tree_roots(trans, root);
+	ret = commit_cowonly_roots(trans, root);
 	BUG_ON(ret);
 
 	cur_trans = root->fs_info->running_transaction;
 	spin_lock(&root->fs_info->new_trans_lock);
 	root->fs_info->running_transaction = NULL;
 	spin_unlock(&root->fs_info->new_trans_lock);
-	btrfs_set_super_generation(&root->fs_info->super_copy,
-				   cur_trans->transid);
-	btrfs_set_super_root(&root->fs_info->super_copy,
-			     root->fs_info->tree_root->node->start);
-	btrfs_set_super_root_level(&root->fs_info->super_copy,
-			   btrfs_header_level(root->fs_info->tree_root->node));
-
-	btrfs_set_super_chunk_root(&root->fs_info->super_copy,
-				   chunk_root->node->start);
-	btrfs_set_super_chunk_root_level(&root->fs_info->super_copy,
-					 btrfs_header_level(chunk_root->node));
-	btrfs_set_super_chunk_root_generation(&root->fs_info->super_copy,
-				btrfs_header_generation(chunk_root->node));
+
+	btrfs_set_root_node(&root->fs_info->tree_root->root_item,
+			    root->fs_info->tree_root->node);
+	free_extent_buffer(root->fs_info->tree_root->commit_root);
+	root->fs_info->tree_root->commit_root =
+				btrfs_root_node(root->fs_info->tree_root);
+
+	btrfs_set_root_node(&root->fs_info->chunk_root->root_item,
+			    root->fs_info->chunk_root->node);
+	free_extent_buffer(root->fs_info->chunk_root->commit_root);
+	root->fs_info->chunk_root->commit_root =
+				btrfs_root_node(root->fs_info->chunk_root);
+
+	update_super_roots(root);
 
 	if (!root->fs_info->log_root_recovering) {
 		btrfs_set_super_log_root(&root->fs_info->super_copy, 0);
@@ -1153,7 +1031,6 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 
 	trans->transaction->blocked = 0;
 
-	wake_up(&root->fs_info->transaction_throttle);
 	wake_up(&root->fs_info->transaction_wait);
 
 	mutex_unlock(&root->fs_info->trans_mutex);
@@ -1170,9 +1047,6 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 	btrfs_finish_extent_commit(trans, root, pinned_copy);
 	kfree(pinned_copy);
 
-	btrfs_drop_dead_reloc_roots(root);
-	mutex_unlock(&root->fs_info->tree_reloc_mutex);
-
 	/* do the directory inserts of any pending snapshot creations */
 	finish_pending_snapshots(trans, root->fs_info);
 
@@ -1186,16 +1060,9 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 	put_transaction(cur_trans);
 	put_transaction(cur_trans);
 
-	list_splice_init(&dirty_fs_roots, &root->fs_info->dead_roots);
-	if (root->fs_info->closing)
-		list_splice_init(&root->fs_info->dead_roots, &dirty_fs_roots);
-
 	mutex_unlock(&root->fs_info->trans_mutex);
 
 	kmem_cache_free(btrfs_trans_handle_cachep, trans);
-
-	if (root->fs_info->closing)
-		drop_dirty_roots(root->fs_info->tree_root, &dirty_fs_roots);
 	return ret;
 }
 
@@ -1204,16 +1071,17 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
  */
 int btrfs_clean_old_snapshots(struct btrfs_root *root)
 {
-	struct list_head dirty_roots;
-	INIT_LIST_HEAD(&dirty_roots);
-again:
-	mutex_lock(&root->fs_info->trans_mutex);
-	list_splice_init(&root->fs_info->dead_roots, &dirty_roots);
-	mutex_unlock(&root->fs_info->trans_mutex);
+	LIST_HEAD(list);
+	struct btrfs_fs_info *fs_info = root->fs_info;
+
+	mutex_lock(&fs_info->trans_mutex);
+	list_splice_init(&fs_info->dead_roots, &list);
+	mutex_unlock(&fs_info->trans_mutex);
 
-	if (!list_empty(&dirty_roots)) {
-		drop_dirty_roots(root, &dirty_roots);
-		goto again;
+	while (!list_empty(&list)) {
+		root = list_entry(list.next, struct btrfs_root, root_list);
+		list_del_init(&root->root_list);
+		btrfs_drop_dead_root(root);
 	}
 	return 0;
 }

fs/btrfs/transaction.h

@@ -62,12 +62,6 @@ struct btrfs_pending_snapshot {
 	struct list_head list;
 };
 
-struct btrfs_dirty_root {
-	struct list_head list;
-	struct btrfs_root *root;
-	struct btrfs_root *latest_root;
-};
-
 static inline void btrfs_set_trans_block_group(struct btrfs_trans_handle *trans,
 					       struct inode *inode)
 {
@@ -100,7 +94,8 @@ int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans,
 int btrfs_commit_tree_roots(struct btrfs_trans_handle *trans,
 			    struct btrfs_root *root);
 
-int btrfs_add_dead_root(struct btrfs_root *root, struct btrfs_root *latest);
+int btrfs_add_dead_root(struct btrfs_root *root);
+int btrfs_drop_dead_root(struct btrfs_root *root);
 int btrfs_defrag_root(struct btrfs_root *root, int cacheonly);
 int btrfs_clean_old_snapshots(struct btrfs_root *root);
 int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
@@ -108,7 +103,8 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 int btrfs_end_transaction_throttle(struct btrfs_trans_handle *trans,
 				   struct btrfs_root *root);
 void btrfs_throttle(struct btrfs_root *root);
-int btrfs_record_root_in_trans(struct btrfs_root *root);
+int btrfs_record_root_in_trans(struct btrfs_trans_handle *trans,
+				struct btrfs_root *root);
 int btrfs_write_and_wait_marked_extents(struct btrfs_root *root,
 					struct extent_io_tree *dirty_pages);
 #endif

fs/btrfs/tree-log.c

@@ -430,18 +430,16 @@ no_copy:
 static noinline struct inode *read_one_inode(struct btrfs_root *root,
 					     u64 objectid)
 {
+	struct btrfs_key key;
 	struct inode *inode;
-	inode = btrfs_iget_locked(root->fs_info->sb, objectid, root);
-	if (inode->i_state & I_NEW) {
-		BTRFS_I(inode)->root = root;
-		BTRFS_I(inode)->location.objectid = objectid;
-		BTRFS_I(inode)->location.type = BTRFS_INODE_ITEM_KEY;
-		BTRFS_I(inode)->location.offset = 0;
-		btrfs_read_locked_inode(inode);
-		unlock_new_inode(inode);
 
-	}
-	if (is_bad_inode(inode)) {
+	key.objectid = objectid;
+	key.type = BTRFS_INODE_ITEM_KEY;
+	key.offset = 0;
+	inode = btrfs_iget(root->fs_info->sb, &key, root);
+	if (IS_ERR(inode)) {
+		inode = NULL;
+	} else if (is_bad_inode(inode)) {
 		iput(inode);
 		inode = NULL;
 	}
@@ -541,6 +539,7 @@ static noinline int replay_one_extent(struct btrfs_trans_handle *trans,
 
 	if (found_type == BTRFS_FILE_EXTENT_REG ||
 	    found_type == BTRFS_FILE_EXTENT_PREALLOC) {
+		u64 offset;
 		unsigned long dest_offset;
 		struct btrfs_key ins;
 
@@ -555,6 +554,7 @@ static noinline int replay_one_extent(struct btrfs_trans_handle *trans,
 		ins.objectid = btrfs_file_extent_disk_bytenr(eb, item);
 		ins.offset = btrfs_file_extent_disk_num_bytes(eb, item);
 		ins.type = BTRFS_EXTENT_ITEM_KEY;
+		offset = key->offset - btrfs_file_extent_offset(eb, item);
 
 		if (ins.objectid > 0) {
 			u64 csum_start;
@@ -569,19 +569,16 @@ static noinline int replay_one_extent(struct btrfs_trans_handle *trans,
 			if (ret == 0) {
 				ret = btrfs_inc_extent_ref(trans, root,
 						ins.objectid, ins.offset,
-						path->nodes[0]->start,
-						root->root_key.objectid,
-						trans->transid, key->objectid);
+						0, root->root_key.objectid,
+						key->objectid, offset);
 			} else {
 				/*
 				 * insert the extent pointer in the extent
 				 * allocation tree
 				 */
-				ret = btrfs_alloc_logged_extent(trans, root,
-						path->nodes[0]->start,
-						root->root_key.objectid,
-						trans->transid, key->objectid,
-						&ins);
+				ret = btrfs_alloc_logged_file_extent(trans,
+						root, root->root_key.objectid,
+						key->objectid, offset, &ins);
 				BUG_ON(ret);
 			}
 			btrfs_release_path(root, path);
@@ -1706,9 +1703,6 @@ static noinline int walk_down_log_tree(struct btrfs_trans_handle *trans,
 				btrfs_wait_tree_block_writeback(next);
 				btrfs_tree_unlock(next);
 
-				ret = btrfs_drop_leaf_ref(trans, root, next);
-				BUG_ON(ret);
-
 				WARN_ON(root_owner !=
 					BTRFS_TREE_LOG_OBJECTID);
 				ret = btrfs_free_reserved_extent(root,
@@ -1753,10 +1747,6 @@ static noinline int walk_down_log_tree(struct btrfs_trans_handle *trans,
 		btrfs_wait_tree_block_writeback(next);
 		btrfs_tree_unlock(next);
 
-		if (*level == 0) {
-			ret = btrfs_drop_leaf_ref(trans, root, next);
-			BUG_ON(ret);
-		}
 		WARN_ON(root_owner != BTRFS_TREE_LOG_OBJECTID);
 		ret = btrfs_free_reserved_extent(root, bytenr, blocksize);
 		BUG_ON(ret);
@@ -1811,12 +1801,6 @@ static noinline int walk_up_log_tree(struct btrfs_trans_handle *trans,
 				btrfs_wait_tree_block_writeback(next);
 				btrfs_tree_unlock(next);
 
-				if (*level == 0) {
-					ret = btrfs_drop_leaf_ref(trans, root,
-								  next);
-					BUG_ON(ret);
-				}
-
 				WARN_ON(root_owner != BTRFS_TREE_LOG_OBJECTID);
 				ret = btrfs_free_reserved_extent(root,
 						path->nodes[*level]->start,
@@ -1884,11 +1868,6 @@ static int walk_log_tree(struct btrfs_trans_handle *trans,
 			btrfs_wait_tree_block_writeback(next);
 			btrfs_tree_unlock(next);
 
-			if (orig_level == 0) {
-				ret = btrfs_drop_leaf_ref(trans, log,
-							  next);
-				BUG_ON(ret);
-			}
 			WARN_ON(log->root_key.objectid !=
 				BTRFS_TREE_LOG_OBJECTID);
 			ret = btrfs_free_reserved_extent(log, next->start,
@@ -2027,9 +2006,7 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
 	ret = btrfs_write_and_wait_marked_extents(log, &log->dirty_log_pages);
 	BUG_ON(ret);
 
-	btrfs_set_root_bytenr(&log->root_item, log->node->start);
-	btrfs_set_root_generation(&log->root_item, trans->transid);
-	btrfs_set_root_level(&log->root_item, btrfs_header_level(log->node));
+	btrfs_set_root_node(&log->root_item, log->node);
 
 	root->log_batch = 0;
 	root->log_transid++;
@@ -2581,7 +2558,7 @@ static noinline int copy_items(struct btrfs_trans_handle *trans,
 				       ins_keys, ins_sizes, nr);
 	BUG_ON(ret);
 
-	for (i = 0; i < nr; i++) {
+	for (i = 0; i < nr; i++, dst_path->slots[0]++) {
 		dst_offset = btrfs_item_ptr_offset(dst_path->nodes[0],
 						   dst_path->slots[0]);
 
@@ -2617,36 +2594,31 @@ static noinline int copy_items(struct btrfs_trans_handle *trans,
 			found_type = btrfs_file_extent_type(src, extent);
 			if (found_type == BTRFS_FILE_EXTENT_REG ||
 			    found_type == BTRFS_FILE_EXTENT_PREALLOC) {
-				u64 ds = btrfs_file_extent_disk_bytenr(src,
-								   extent);
-				u64 dl = btrfs_file_extent_disk_num_bytes(src,
-								      extent);
-				u64 cs = btrfs_file_extent_offset(src, extent);
-				u64 cl = btrfs_file_extent_num_bytes(src,
-								     extent);;
+				u64 ds, dl, cs, cl;
+				ds = btrfs_file_extent_disk_bytenr(src,
+								extent);
+				/* ds == 0 is a hole */
+				if (ds == 0)
+					continue;
+
+				dl = btrfs_file_extent_disk_num_bytes(src,
+								extent);
+				cs = btrfs_file_extent_offset(src, extent);
+				cl = btrfs_file_extent_num_bytes(src,
+								extent);;
 				if (btrfs_file_extent_compression(src,
 								  extent)) {
 					cs = 0;
 					cl = dl;
 				}
-				/* ds == 0 is a hole */
-				if (ds != 0) {
-					ret = btrfs_inc_extent_ref(trans, log,
-						   ds, dl,
-						   dst_path->nodes[0]->start,
-						   BTRFS_TREE_LOG_OBJECTID,
-						   trans->transid,
-						   ins_keys[i].objectid);
-					BUG_ON(ret);
-					ret = btrfs_lookup_csums_range(
-						   log->fs_info->csum_root,
-						   ds + cs, ds + cs + cl - 1,
-						   &ordered_sums);
-					BUG_ON(ret);
-				}
+
+				ret = btrfs_lookup_csums_range(
+						log->fs_info->csum_root,
+						ds + cs, ds + cs + cl - 1,
+						&ordered_sums);
+				BUG_ON(ret);
 			}
 		}
-		dst_path->slots[0]++;
 	}
 
 	btrfs_mark_buffer_dirty(dst_path->nodes[0]);
@@ -3029,9 +3001,7 @@ again:
 		BUG_ON(!wc.replay_dest);
 
 		wc.replay_dest->log_root = log;
-		mutex_lock(&fs_info->trans_mutex);
-		btrfs_record_root_in_trans(wc.replay_dest);
-		mutex_unlock(&fs_info->trans_mutex);
+		btrfs_record_root_in_trans(trans, wc.replay_dest);
 		ret = walk_log_tree(trans, log, &wc);
 		BUG_ON(ret);
 
@@ -3049,6 +3019,7 @@ again:
 		key.offset = found_key.offset - 1;
 		wc.replay_dest->log_root = NULL;
 		free_extent_buffer(log->node);
+		free_extent_buffer(log->commit_root);
 		kfree(log);
 
 		if (found_key.offset == 0)

fs/btrfs/volumes.c

@@ -161,8 +161,10 @@ static noinline int run_scheduled_bios(struct btrfs_device *device)
 	int again = 0;
 	unsigned long num_run;
 	unsigned long num_sync_run;
+	unsigned long batch_run = 0;
 	unsigned long limit;
 	unsigned long last_waited = 0;
+	int force_reg = 0;
 
 	bdi = blk_get_backing_dev_info(device->bdev);
 	fs_info = device->dev_root->fs_info;
@@ -176,19 +178,22 @@ static noinline int run_scheduled_bios(struct btrfs_device *device)
 
 loop:
 	spin_lock(&device->io_lock);
-	num_run = 0;
 
 loop_lock:
+	num_run = 0;
 
 	/* take all the bios off the list at once and process them
 	 * later on (without the lock held).  But, remember the
 	 * tail and other pointers so the bios can be properly reinserted
 	 * into the list if we hit congestion
 	 */
-	if (device->pending_sync_bios.head)
+	if (!force_reg && device->pending_sync_bios.head) {
 		pending_bios = &device->pending_sync_bios;
-	else
+		force_reg = 1;
+	} else {
 		pending_bios = &device->pending_bios;
+		force_reg = 0;
+	}
 
 	pending = pending_bios->head;
 	tail = pending_bios->tail;
@@ -228,10 +233,14 @@ loop_lock:
 	while (pending) {
 
 		rmb();
-		if (pending_bios != &device->pending_sync_bios &&
-		    device->pending_sync_bios.head &&
-		    num_run > 16) {
-			cond_resched();
+		/* we want to work on both lists, but do more bios on the
+		 * sync list than the regular list
+		 */
+		if ((num_run > 32 &&
+		    pending_bios != &device->pending_sync_bios &&
+		    device->pending_sync_bios.head) ||
+		   (num_run > 64 && pending_bios == &device->pending_sync_bios &&
+		    device->pending_bios.head)) {
 			spin_lock(&device->io_lock);
 			requeue_list(pending_bios, pending, tail);
 			goto loop_lock;
@@ -249,6 +258,8 @@ loop_lock:
 		BUG_ON(atomic_read(&cur->bi_cnt) == 0);
 		submit_bio(cur->bi_rw, cur);
 		num_run++;
+		batch_run++;
+
 		if (bio_sync(cur))
 			num_sync_run++;
 
@@ -265,7 +276,7 @@ loop_lock:
 		 * is now congested.  Back off and let other work structs
 		 * run instead
 		 */
-		if (pending && bdi_write_congested(bdi) && num_run > 16 &&
+		if (pending && bdi_write_congested(bdi) && batch_run > 32 &&
 		    fs_info->fs_devices->open_devices > 1) {
 			struct io_context *ioc;
 
@@ -366,6 +377,7 @@ static noinline int device_list_add(const char *path,
 		memcpy(fs_devices->fsid, disk_super->fsid, BTRFS_FSID_SIZE);
 		fs_devices->latest_devid = devid;
 		fs_devices->latest_trans = found_transid;
+		mutex_init(&fs_devices->device_list_mutex);
 		device = NULL;
 	} else {
 		device = __find_device(&fs_devices->devices, devid,
@@ -392,7 +404,11 @@ static noinline int device_list_add(const char *path,
 			return -ENOMEM;
 		}
 		INIT_LIST_HEAD(&device->dev_alloc_list);
+
+		mutex_lock(&fs_devices->device_list_mutex);
 		list_add(&device->dev_list, &fs_devices->devices);
+		mutex_unlock(&fs_devices->device_list_mutex);
+
 		device->fs_devices = fs_devices;
 		fs_devices->num_devices++;
 	}
@@ -418,10 +434,12 @@ static struct btrfs_fs_devices *clone_fs_devices(struct btrfs_fs_devices *orig)
 	INIT_LIST_HEAD(&fs_devices->devices);
 	INIT_LIST_HEAD(&fs_devices->alloc_list);
 	INIT_LIST_HEAD(&fs_devices->list);
+	mutex_init(&fs_devices->device_list_mutex);
 	fs_devices->latest_devid = orig->latest_devid;
 	fs_devices->latest_trans = orig->latest_trans;
 	memcpy(fs_devices->fsid, orig->fsid, sizeof(fs_devices->fsid));
 
+	mutex_lock(&orig->device_list_mutex);
 	list_for_each_entry(orig_dev, &orig->devices, dev_list) {
 		device = kzalloc(sizeof(*device), GFP_NOFS);
 		if (!device)
@@ -443,8 +461,10 @@ static struct btrfs_fs_devices *clone_fs_devices(struct btrfs_fs_devices *orig)
 		device->fs_devices = fs_devices;
 		fs_devices->num_devices++;
 	}
+	mutex_unlock(&orig->device_list_mutex);
 	return fs_devices;
 error:
+	mutex_unlock(&orig->device_list_mutex);
 	free_fs_devices(fs_devices);
 	return ERR_PTR(-ENOMEM);
 }
@@ -455,6 +475,7 @@ int btrfs_close_extra_devices(struct btrfs_fs_devices *fs_devices)
 
 	mutex_lock(&uuid_mutex);
 again:
+	mutex_lock(&fs_devices->device_list_mutex);
 	list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
 		if (device->in_fs_metadata)
 			continue;
@@ -474,6 +495,7 @@ again:
 		kfree(device->name);
 		kfree(device);
 	}
+	mutex_unlock(&fs_devices->device_list_mutex);
 
 	if (fs_devices->seed) {
 		fs_devices = fs_devices->seed;
@@ -594,6 +616,9 @@ static int __btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
 		device->in_fs_metadata = 0;
 		device->mode = flags;
 
+		if (!blk_queue_nonrot(bdev_get_queue(bdev)))
+			fs_devices->rotating = 1;
+
 		fs_devices->open_devices++;
 		if (device->writeable) {
 			fs_devices->rw_devices++;
@@ -1121,12 +1146,14 @@ int btrfs_rm_device(struct btrfs_root *root, char *device_path)
 
 		device = NULL;
 		devices = &root->fs_info->fs_devices->devices;
+		mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
 		list_for_each_entry(tmp, devices, dev_list) {
 			if (tmp->in_fs_metadata && !tmp->bdev) {
 				device = tmp;
 				break;
 			}
 		}
+		mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
 		bdev = NULL;
 		bh = NULL;
 		disk_super = NULL;
@@ -1181,7 +1208,16 @@ int btrfs_rm_device(struct btrfs_root *root, char *device_path)
 		goto error_brelse;
 
 	device->in_fs_metadata = 0;
+
+	/*
+	 * the device list mutex makes sure that we don't change
+	 * the device list while someone else is writing out all
+	 * the device supers.
+	 */
+	mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
 	list_del_init(&device->dev_list);
+	mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
+
 	device->fs_devices->num_devices--;
 
 	next_device = list_entry(root->fs_info->fs_devices->devices.next,
@@ -1275,6 +1311,7 @@ static int btrfs_prepare_sprout(struct btrfs_trans_handle *trans,
 	seed_devices->opened = 1;
 	INIT_LIST_HEAD(&seed_devices->devices);
 	INIT_LIST_HEAD(&seed_devices->alloc_list);
+	mutex_init(&seed_devices->device_list_mutex);
 	list_splice_init(&fs_devices->devices, &seed_devices->devices);
 	list_splice_init(&fs_devices->alloc_list, &seed_devices->alloc_list);
 	list_for_each_entry(device, &seed_devices->devices, dev_list) {
@@ -1400,6 +1437,10 @@ int btrfs_init_new_device(struct btrfs_root *root, char *device_path)
 	mutex_lock(&root->fs_info->volume_mutex);
 
 	devices = &root->fs_info->fs_devices->devices;
+	/*
+	 * we have the volume lock, so we don't need the extra
+	 * device list mutex while reading the list here.
+	 */
 	list_for_each_entry(device, devices, dev_list) {
 		if (device->bdev == bdev) {
 			ret = -EEXIST;
@@ -1454,6 +1495,12 @@ int btrfs_init_new_device(struct btrfs_root *root, char *device_path)
 	}
 
 	device->fs_devices = root->fs_info->fs_devices;
+
+	/*
+	 * we don't want write_supers to jump in here with our device
+	 * half setup
+	 */
+	mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
 	list_add(&device->dev_list, &root->fs_info->fs_devices->devices);
 	list_add(&device->dev_alloc_list,
 		 &root->fs_info->fs_devices->alloc_list);
@@ -1462,6 +1509,9 @@ int btrfs_init_new_device(struct btrfs_root *root, char *device_path)
 	root->fs_info->fs_devices->rw_devices++;
 	root->fs_info->fs_devices->total_rw_bytes += device->total_bytes;
 
+	if (!blk_queue_nonrot(bdev_get_queue(bdev)))
+		root->fs_info->fs_devices->rotating = 1;
+
 	total_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy);
 	btrfs_set_super_total_bytes(&root->fs_info->super_copy,
 				    total_bytes + device->total_bytes);
@@ -1469,6 +1519,7 @@ int btrfs_init_new_device(struct btrfs_root *root, char *device_path)
 	total_bytes = btrfs_super_num_devices(&root->fs_info->super_copy);
 	btrfs_set_super_num_devices(&root->fs_info->super_copy,
 				    total_bytes + 1);
+	mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
 
 	if (seeding_dev) {
 		ret = init_first_rw_device(trans, root, device);
@@ -1671,8 +1722,6 @@ static int btrfs_relocate_chunk(struct btrfs_root *root,
 	int ret;
 	int i;
 
-	printk(KERN_INFO "btrfs relocating chunk %llu\n",
-	       (unsigned long long)chunk_offset);
 	root = root->fs_info->chunk_root;
 	extent_root = root->fs_info->extent_root;
 	em_tree = &root->fs_info->mapping_tree.map_tree;

fs/btrfs/volumes.h

@@ -96,7 +96,12 @@ struct btrfs_fs_devices {
 	u64 rw_devices;
 	u64 total_rw_bytes;
 	struct block_device *latest_bdev;
-	/* all of the devices in the FS */
+
+	/* all of the devices in the FS, protected by a mutex
+	 * so we can safely walk it to write out the supers without
+	 * worrying about add/remove by the multi-device code
+	 */
+	struct mutex device_list_mutex;
 	struct list_head devices;
 
 	/* devices not currently being allocated */
@@ -107,6 +112,11 @@ struct btrfs_fs_devices {
 	int seeding;
 
 	int opened;
+
+	/* set when we find or add a device that doesn't have the
+	 * nonrot flag set
+	 */
+	int rotating;
 };
 
 struct btrfs_bio_stripe {