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

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/btrfs-unstable:
  Btrfs: make sure reserve_metadata_bytes doesn't leak out strange errors
  Btrfs: use the commit_root for reading free_space_inode crcs
  Btrfs: reduce extent_state lock contention for metadata
  Btrfs: remove lockdep magic from btrfs_next_leaf
  Btrfs: make a lockdep class for each root
  Btrfs: switch the btrfs tree locks to reader/writer
  Btrfs: fix deadlock when throttling transactions
  Btrfs: stop using highmem for extent_buffers
  Btrfs: fix BUG_ON() caused by ENOSPC when relocating space
  Btrfs: tag pages for writeback in sync
  Btrfs: fix enospc problems with delalloc
  Btrfs: don't flush delalloc arbitrarily
  Btrfs: use find_or_create_page instead of grab_cache_page
  Btrfs: use a worker thread to do caching
  Btrfs: fix how we merge extent states and deal with cached states
  Btrfs: use the normal checksumming infrastructure for free space cache
  Btrfs: serialize flushers in reserve_metadata_bytes
  Btrfs: do transaction space reservation before joining the transaction
  Btrfs: try to only do one btrfs_search_slot in do_setxattr

fs/btrfs/btrfs_inode.h

@@ -34,6 +34,9 @@ struct btrfs_inode {
 	 */
 	struct btrfs_key location;
 
+	/* Lock for counters */
+	spinlock_t lock;
+
 	/* the extent_tree has caches of all the extent mappings to disk */
 	struct extent_map_tree extent_tree;
 
@@ -134,8 +137,8 @@ struct btrfs_inode {
 	 * items we think we'll end up using, and reserved_extents is the number
 	 * of extent items we've reserved metadata for.
 	 */
-	atomic_t outstanding_extents;
-	atomic_t reserved_extents;
+	unsigned outstanding_extents;
+	unsigned reserved_extents;
 
 	/*
 	 * ordered_data_close is set by truncate when a file that used
@@ -184,4 +187,13 @@ static inline void btrfs_i_size_write(struct inode *inode, u64 size)
 	BTRFS_I(inode)->disk_i_size = size;
 }
 
+static inline bool btrfs_is_free_space_inode(struct btrfs_root *root,
+				       struct inode *inode)
+{
+	if (root == root->fs_info->tree_root ||
+	    BTRFS_I(inode)->location.objectid == BTRFS_FREE_INO_OBJECTID)
+		return true;
+	return false;
+}
+
 #endif

fs/btrfs/ctree.h

@@ -755,6 +755,8 @@ struct btrfs_space_info {
 				   chunks for this space */
 	unsigned int chunk_alloc:1;	/* set if we are allocating a chunk */
 
+	unsigned int flush:1;		/* set if we are trying to make space */
+
 	unsigned int force_alloc;	/* set if we need to force a chunk
 					   alloc for this space */
 
@@ -764,7 +766,7 @@ struct btrfs_space_info {
 	struct list_head block_groups[BTRFS_NR_RAID_TYPES];
 	spinlock_t lock;
 	struct rw_semaphore groups_sem;
-	atomic_t caching_threads;
+	wait_queue_head_t wait;
 };
 
 struct btrfs_block_rsv {
@@ -824,6 +826,7 @@ struct btrfs_caching_control {
 	struct list_head list;
 	struct mutex mutex;
 	wait_queue_head_t wait;
+	struct btrfs_work work;
 	struct btrfs_block_group_cache *block_group;
 	u64 progress;
 	atomic_t count;
@@ -1032,6 +1035,8 @@ struct btrfs_fs_info {
 	struct btrfs_workers endio_write_workers;
 	struct btrfs_workers endio_freespace_worker;
 	struct btrfs_workers submit_workers;
+	struct btrfs_workers caching_workers;
+
 	/*
 	 * fixup workers take dirty pages that didn't properly go through
 	 * the cow mechanism and make them safe to write.  It happens
@@ -2128,7 +2133,7 @@ static inline bool btrfs_mixed_space_info(struct btrfs_space_info *space_info)
 
 /* extent-tree.c */
 static inline u64 btrfs_calc_trans_metadata_size(struct btrfs_root *root,
-						 int num_items)
+						 unsigned num_items)
 {
 	return (root->leafsize + root->nodesize * (BTRFS_MAX_LEVEL - 1)) *
 		3 * num_items;
@@ -2222,9 +2227,6 @@ void btrfs_set_inode_space_info(struct btrfs_root *root, struct inode *ionde);
 void btrfs_clear_space_info_full(struct btrfs_fs_info *info);
 int btrfs_check_data_free_space(struct inode *inode, u64 bytes);
 void btrfs_free_reserved_data_space(struct inode *inode, u64 bytes);
-int btrfs_trans_reserve_metadata(struct btrfs_trans_handle *trans,
-				struct btrfs_root *root,
-				int num_items);
 void btrfs_trans_release_metadata(struct btrfs_trans_handle *trans,
 				struct btrfs_root *root);
 int btrfs_orphan_reserve_metadata(struct btrfs_trans_handle *trans,
@@ -2330,7 +2332,7 @@ struct btrfs_path *btrfs_alloc_path(void);
 void btrfs_free_path(struct btrfs_path *p);
 void btrfs_set_path_blocking(struct btrfs_path *p);
 void btrfs_clear_path_blocking(struct btrfs_path *p,
-			       struct extent_buffer *held);
+			       struct extent_buffer *held, int held_rw);
 void btrfs_unlock_up_safe(struct btrfs_path *p, int level);
 
 int btrfs_del_items(struct btrfs_trans_handle *trans, struct btrfs_root *root,

fs/btrfs/delayed-inode.c

@@ -735,7 +735,7 @@ static int btrfs_batch_insert_items(struct btrfs_trans_handle *trans,
 	}
 
 	/* reset all the locked nodes in the patch to spinning locks. */
-	btrfs_clear_path_blocking(path, NULL);
+	btrfs_clear_path_blocking(path, NULL, 0);
 
 	/* insert the keys of the items */
 	ret = setup_items_for_insert(trans, root, path, keys, data_size,

fs/btrfs/dir-item.c

@@ -89,13 +89,8 @@ int btrfs_insert_xattr_item(struct btrfs_trans_handle *trans,
 	data_size = sizeof(*dir_item) + name_len + data_len;
 	dir_item = insert_with_overflow(trans, root, path, &key, data_size,
 					name, name_len);
-	/*
-	 * FIXME: at some point we should handle xattr's that are larger than
-	 * what we can fit in our leaf.  We set location to NULL b/c we arent
-	 * pointing at anything else, that will change if we store the xattr
-	 * data in a separate inode.
-	 */
-	BUG_ON(IS_ERR(dir_item));
+	if (IS_ERR(dir_item))
+		return PTR_ERR(dir_item);
 	memset(&location, 0, sizeof(location));
 
 	leaf = path->nodes[0];

fs/btrfs/disk-io.c

@@ -100,38 +100,83 @@ struct async_submit_bio {
 	struct btrfs_work work;
 };
 
-/* These are used to set the lockdep class on the extent buffer locks.
- * The class is set by the readpage_end_io_hook after the buffer has
- * passed csum validation but before the pages are unlocked.
+/*
+ * Lockdep class keys for extent_buffer->lock's in this root.  For a given
+ * eb, the lockdep key is determined by the btrfs_root it belongs to and
+ * the level the eb occupies in the tree.
+ *
+ * Different roots are used for different purposes and may nest inside each
+ * other and they require separate keysets.  As lockdep keys should be
+ * static, assign keysets according to the purpose of the root as indicated
+ * by btrfs_root->objectid.  This ensures that all special purpose roots
+ * have separate keysets.
  *
- * The lockdep class is also set by btrfs_init_new_buffer on freshly
- * allocated blocks.
+ * Lock-nesting across peer nodes is always done with the immediate parent
+ * node locked thus preventing deadlock.  As lockdep doesn't know this, use
+ * subclass to avoid triggering lockdep warning in such cases.
  *
- * The class is based on the level in the tree block, which allows lockdep
- * to know that lower nodes nest inside the locks of higher nodes.
+ * The key is set by the readpage_end_io_hook after the buffer has passed
+ * csum validation but before the pages are unlocked.  It is also set by
+ * btrfs_init_new_buffer on freshly allocated blocks.
  *
- * We also add a check to make sure the highest level of the tree is
- * the same as our lockdep setup here.  If BTRFS_MAX_LEVEL changes, this
- * code needs update as well.
+ * We also add a check to make sure the highest level of the tree is the
+ * same as our lockdep setup here.  If BTRFS_MAX_LEVEL changes, this code
+ * needs update as well.
  */
 #ifdef CONFIG_DEBUG_LOCK_ALLOC
 # if BTRFS_MAX_LEVEL != 8
 #  error
 # endif
-static struct lock_class_key btrfs_eb_class[BTRFS_MAX_LEVEL + 1];
-static const char *btrfs_eb_name[BTRFS_MAX_LEVEL + 1] = {
-	/* leaf */
-	"btrfs-extent-00",
-	"btrfs-extent-01",
-	"btrfs-extent-02",
-	"btrfs-extent-03",
-	"btrfs-extent-04",
-	"btrfs-extent-05",
-	"btrfs-extent-06",
-	"btrfs-extent-07",
-	/* highest possible level */
-	"btrfs-extent-08",
+
+static struct btrfs_lockdep_keyset {
+	u64			id;		/* root objectid */
+	const char		*name_stem;	/* lock name stem */
+	char			names[BTRFS_MAX_LEVEL + 1][20];
+	struct lock_class_key	keys[BTRFS_MAX_LEVEL + 1];
+} btrfs_lockdep_keysets[] = {
+	{ .id = BTRFS_ROOT_TREE_OBJECTID,	.name_stem = "root"	},
+	{ .id = BTRFS_EXTENT_TREE_OBJECTID,	.name_stem = "extent"	},
+	{ .id = BTRFS_CHUNK_TREE_OBJECTID,	.name_stem = "chunk"	},
+	{ .id = BTRFS_DEV_TREE_OBJECTID,	.name_stem = "dev"	},
+	{ .id = BTRFS_FS_TREE_OBJECTID,		.name_stem = "fs"	},
+	{ .id = BTRFS_CSUM_TREE_OBJECTID,	.name_stem = "csum"	},
+	{ .id = BTRFS_ORPHAN_OBJECTID,		.name_stem = "orphan"	},
+	{ .id = BTRFS_TREE_LOG_OBJECTID,	.name_stem = "log"	},
+	{ .id = BTRFS_TREE_RELOC_OBJECTID,	.name_stem = "treloc"	},
+	{ .id = BTRFS_DATA_RELOC_TREE_OBJECTID,	.name_stem = "dreloc"	},
+	{ .id = 0,				.name_stem = "tree"	},
 };
+
+void __init btrfs_init_lockdep(void)
+{
+	int i, j;
+
+	/* initialize lockdep class names */
+	for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) {
+		struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i];
+
+		for (j = 0; j < ARRAY_SIZE(ks->names); j++)
+			snprintf(ks->names[j], sizeof(ks->names[j]),
+				 "btrfs-%s-%02d", ks->name_stem, j);
+	}
+}
+
+void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb,
+				    int level)
+{
+	struct btrfs_lockdep_keyset *ks;
+
+	BUG_ON(level >= ARRAY_SIZE(ks->keys));
+
+	/* find the matching keyset, id 0 is the default entry */
+	for (ks = btrfs_lockdep_keysets; ks->id; ks++)
+		if (ks->id == objectid)
+			break;
+
+	lockdep_set_class_and_name(&eb->lock,
+				   &ks->keys[level], ks->names[level]);
+}
+
 #endif
 
 /*
@@ -217,7 +262,6 @@ static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
 	unsigned long len;
 	unsigned long cur_len;
 	unsigned long offset = BTRFS_CSUM_SIZE;
-	char *map_token = NULL;
 	char *kaddr;
 	unsigned long map_start;
 	unsigned long map_len;
@@ -228,8 +272,7 @@ static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
 	len = buf->len - offset;
 	while (len > 0) {
 		err = map_private_extent_buffer(buf, offset, 32,
-					&map_token, &kaddr,
-					&map_start, &map_len, KM_USER0);
+					&kaddr, &map_start, &map_len);
 		if (err)
 			return 1;
 		cur_len = min(len, map_len - (offset - map_start));
@@ -237,7 +280,6 @@ static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
 				      crc, cur_len);
 		len -= cur_len;
 		offset += cur_len;
-		unmap_extent_buffer(buf, map_token, KM_USER0);
 	}
 	if (csum_size > sizeof(inline_result)) {
 		result = kzalloc(csum_size * sizeof(char), GFP_NOFS);
@@ -494,15 +536,6 @@ static noinline int check_leaf(struct btrfs_root *root,
 	return 0;
 }
 
-#ifdef CONFIG_DEBUG_LOCK_ALLOC
-void btrfs_set_buffer_lockdep_class(struct extent_buffer *eb, int level)
-{
-	lockdep_set_class_and_name(&eb->lock,
-			   &btrfs_eb_class[level],
-			   btrfs_eb_name[level]);
-}
-#endif
-
 static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end,
 			       struct extent_state *state)
 {
@@ -553,7 +586,8 @@ static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end,
 	}
 	found_level = btrfs_header_level(eb);
 
-	btrfs_set_buffer_lockdep_class(eb, found_level);
+	btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb),
+				       eb, found_level);
 
 	ret = csum_tree_block(root, eb, 1);
 	if (ret) {
@@ -1598,7 +1632,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 		goto fail_bdi;
 	}
 
-	fs_info->btree_inode->i_mapping->flags &= ~__GFP_FS;
+	mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
 
 	INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
 	INIT_LIST_HEAD(&fs_info->trans_list);
@@ -1802,6 +1836,9 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 			   fs_info->thread_pool_size),
 			   &fs_info->generic_worker);
 
+	btrfs_init_workers(&fs_info->caching_workers, "cache",
+			   2, &fs_info->generic_worker);
+
 	/* a higher idle thresh on the submit workers makes it much more
 	 * likely that bios will be send down in a sane order to the
 	 * devices
@@ -1855,6 +1892,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	btrfs_start_workers(&fs_info->endio_write_workers, 1);
 	btrfs_start_workers(&fs_info->endio_freespace_worker, 1);
 	btrfs_start_workers(&fs_info->delayed_workers, 1);
+	btrfs_start_workers(&fs_info->caching_workers, 1);
 
 	fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
 	fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
@@ -2112,6 +2150,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	btrfs_stop_workers(&fs_info->endio_freespace_worker);
 	btrfs_stop_workers(&fs_info->submit_workers);
 	btrfs_stop_workers(&fs_info->delayed_workers);
+	btrfs_stop_workers(&fs_info->caching_workers);
 fail_alloc:
 	kfree(fs_info->delayed_root);
 fail_iput:
@@ -2577,6 +2616,7 @@ int close_ctree(struct btrfs_root *root)
 	btrfs_stop_workers(&fs_info->endio_freespace_worker);
 	btrfs_stop_workers(&fs_info->submit_workers);
 	btrfs_stop_workers(&fs_info->delayed_workers);
+	btrfs_stop_workers(&fs_info->caching_workers);
 
 	btrfs_close_devices(fs_info->fs_devices);
 	btrfs_mapping_tree_free(&fs_info->mapping_tree);

fs/btrfs/disk-io.h

@@ -87,10 +87,14 @@ int btree_lock_page_hook(struct page *page);
 
 
 #ifdef CONFIG_DEBUG_LOCK_ALLOC
-void btrfs_set_buffer_lockdep_class(struct extent_buffer *eb, int level);
+void btrfs_init_lockdep(void);
+void btrfs_set_buffer_lockdep_class(u64 objectid,
+			            struct extent_buffer *eb, int level);
 #else
-static inline void btrfs_set_buffer_lockdep_class(struct extent_buffer *eb,
-						 int level)
+static inline void btrfs_init_lockdep(void)
+{ }
+static inline void btrfs_set_buffer_lockdep_class(u64 objectid,
+					struct extent_buffer *eb, int level)
 {
 }
 #endif

fs/btrfs/extent_io.h

@@ -120,8 +120,6 @@ struct extent_state {
 struct extent_buffer {
 	u64 start;
 	unsigned long len;
-	char *map_token;
-	char *kaddr;
 	unsigned long map_start;
 	unsigned long map_len;
 	struct page *first_page;
@@ -130,14 +128,26 @@ struct extent_buffer {
 	struct rcu_head rcu_head;
 	atomic_t refs;
 
-	/* the spinlock is used to protect most operations */
-	spinlock_t lock;
+	/* count of read lock holders on the extent buffer */
+	atomic_t write_locks;
+	atomic_t read_locks;
+	atomic_t blocking_writers;
+	atomic_t blocking_readers;
+	atomic_t spinning_readers;
+	atomic_t spinning_writers;
+
+	/* protects write locks */
+	rwlock_t lock;
 
-	/*
-	 * when we keep the lock held while blocking, waiters go onto
-	 * the wq
+	/* readers use lock_wq while they wait for the write
+	 * lock holders to unlock
 	 */
-	wait_queue_head_t lock_wq;
+	wait_queue_head_t write_lock_wq;
+
+	/* writers use read_lock_wq while they wait for readers
+	 * to unlock
+	 */
+	wait_queue_head_t read_lock_wq;
 };
 
 static inline void extent_set_compress_type(unsigned long *bio_flags,
@@ -279,15 +289,10 @@ int clear_extent_buffer_uptodate(struct extent_io_tree *tree,
 int extent_buffer_uptodate(struct extent_io_tree *tree,
 			   struct extent_buffer *eb,
 			   struct extent_state *cached_state);
-int map_extent_buffer(struct extent_buffer *eb, unsigned long offset,
-		      unsigned long min_len, char **token, char **map,
-		      unsigned long *map_start,
-		      unsigned long *map_len, int km);
 int map_private_extent_buffer(struct extent_buffer *eb, unsigned long offset,
-		      unsigned long min_len, char **token, char **map,
+		      unsigned long min_len, char **map,
 		      unsigned long *map_start,
-		      unsigned long *map_len, int km);
-void unmap_extent_buffer(struct extent_buffer *eb, char *token, int km);
+		      unsigned long *map_len);
 int extent_range_uptodate(struct extent_io_tree *tree,
 			  u64 start, u64 end);
 int extent_clear_unlock_delalloc(struct inode *inode,

fs/btrfs/file-item.c

@@ -177,6 +177,15 @@ static int __btrfs_lookup_bio_sums(struct btrfs_root *root,
 
 	WARN_ON(bio->bi_vcnt <= 0);
 
+	/*
+	 * the free space stuff is only read when it hasn't been
+	 * updated in the current transaction.  So, we can safely
+	 * read from the commit root and sidestep a nasty deadlock
+	 * between reading the free space cache and updating the csum tree.
+	 */
+	if (btrfs_is_free_space_inode(root, inode))
+		path->search_commit_root = 1;
+
 	disk_bytenr = (u64)bio->bi_sector << 9;
 	if (dio)
 		offset = logical_offset;
@@ -664,10 +673,6 @@ int btrfs_csum_file_blocks(struct btrfs_trans_handle *trans,
 	struct btrfs_sector_sum *sector_sum;
 	u32 nritems;
 	u32 ins_size;
-	char *eb_map;
-	char *eb_token;
-	unsigned long map_len;
-	unsigned long map_start;
 	u16 csum_size =
 		btrfs_super_csum_size(&root->fs_info->super_copy);
 
@@ -814,30 +819,9 @@ int btrfs_csum_file_blocks(struct btrfs_trans_handle *trans,
 	item_end = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_csum_item);
 	item_end = (struct btrfs_csum_item *)((unsigned char *)item_end +
 				      btrfs_item_size_nr(leaf, path->slots[0]));
-	eb_token = NULL;
 next_sector:
 
-	if (!eb_token ||
-	   (unsigned long)item + csum_size >= map_start + map_len) {
-		int err;
-
-		if (eb_token)
-			unmap_extent_buffer(leaf, eb_token, KM_USER1);
-		eb_token = NULL;
-		err = map_private_extent_buffer(leaf, (unsigned long)item,
-						csum_size,
-						&eb_token, &eb_map,
-						&map_start, &map_len, KM_USER1);
-		if (err)
-			eb_token = NULL;
-	}
-	if (eb_token) {
-		memcpy(eb_token + ((unsigned long)item & (PAGE_CACHE_SIZE - 1)),
-		       &sector_sum->sum, csum_size);
-	} else {
-		write_extent_buffer(leaf, &sector_sum->sum,
-				    (unsigned long)item, csum_size);
-	}
+	write_extent_buffer(leaf, &sector_sum->sum, (unsigned long)item, csum_size);
 
 	total_bytes += root->sectorsize;
 	sector_sum++;
@@ -850,10 +834,7 @@ int btrfs_csum_file_blocks(struct btrfs_trans_handle *trans,
 			goto next_sector;
 		}
 	}
-	if (eb_token) {
-		unmap_extent_buffer(leaf, eb_token, KM_USER1);
-		eb_token = NULL;
-	}
+
 	btrfs_mark_buffer_dirty(path->nodes[0]);
 	if (total_bytes < sums->len) {
 		btrfs_release_path(path);

fs/btrfs/file.c

@@ -1081,7 +1081,8 @@ static noinline int prepare_pages(struct btrfs_root *root, struct file *file,
 
 again:
 	for (i = 0; i < num_pages; i++) {
-		pages[i] = grab_cache_page(inode->i_mapping, index + i);
+		pages[i] = find_or_create_page(inode->i_mapping, index + i,
+					       GFP_NOFS);
 		if (!pages[i]) {
 			faili = i - 1;
 			err = -ENOMEM;
@@ -1238,9 +1239,11 @@ static noinline ssize_t __btrfs_buffered_write(struct file *file,
 		 * managed to copy.
 		 */
 		if (num_pages > dirty_pages) {
-			if (copied > 0)
-				atomic_inc(
-					&BTRFS_I(inode)->outstanding_extents);
+			if (copied > 0) {
+				spin_lock(&BTRFS_I(inode)->lock);
+				BTRFS_I(inode)->outstanding_extents++;
+				spin_unlock(&BTRFS_I(inode)->lock);
+			}
 			btrfs_delalloc_release_space(inode,
 					(num_pages - dirty_pages) <<
 					PAGE_CACHE_SHIFT);

fs/btrfs/inode.c

@@ -750,15 +750,6 @@ static u64 get_extent_allocation_hint(struct inode *inode, u64 start,
 	return alloc_hint;
 }
 
-static inline bool is_free_space_inode(struct btrfs_root *root,
-				       struct inode *inode)
-{
-	if (root == root->fs_info->tree_root ||
-	    BTRFS_I(inode)->location.objectid == BTRFS_FREE_INO_OBJECTID)
-		return true;
-	return false;
-}
-
 /*
  * when extent_io.c finds a delayed allocation range in the file,
  * the call backs end up in this code.  The basic idea is to
@@ -791,7 +782,7 @@ static noinline int cow_file_range(struct inode *inode,
 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
 	int ret = 0;
 
-	BUG_ON(is_free_space_inode(root, inode));
+	BUG_ON(btrfs_is_free_space_inode(root, inode));
 	trans = btrfs_join_transaction(root);
 	BUG_ON(IS_ERR(trans));
 	trans->block_rsv = &root->fs_info->delalloc_block_rsv;
@@ -1072,7 +1063,7 @@ static noinline int run_delalloc_nocow(struct inode *inode,
 	path = btrfs_alloc_path();
 	BUG_ON(!path);
 
-	nolock = is_free_space_inode(root, inode);
+	nolock = btrfs_is_free_space_inode(root, inode);
 
 	if (nolock)
 		trans = btrfs_join_transaction_nolock(root);
@@ -1298,7 +1289,9 @@ static int btrfs_split_extent_hook(struct inode *inode,
 	if (!(orig->state & EXTENT_DELALLOC))
 		return 0;
 
-	atomic_inc(&BTRFS_I(inode)->outstanding_extents);
+	spin_lock(&BTRFS_I(inode)->lock);
+	BTRFS_I(inode)->outstanding_extents++;
+	spin_unlock(&BTRFS_I(inode)->lock);
 	return 0;
 }
 
@@ -1316,7 +1309,9 @@ static int btrfs_merge_extent_hook(struct inode *inode,
 	if (!(other->state & EXTENT_DELALLOC))
 		return 0;
 
-	atomic_dec(&BTRFS_I(inode)->outstanding_extents);
+	spin_lock(&BTRFS_I(inode)->lock);
+	BTRFS_I(inode)->outstanding_extents--;
+	spin_unlock(&BTRFS_I(inode)->lock);
 	return 0;
 }
 
@@ -1337,12 +1332,15 @@ static int btrfs_set_bit_hook(struct inode *inode,
 	if (!(state->state & EXTENT_DELALLOC) && (*bits & EXTENT_DELALLOC)) {
 		struct btrfs_root *root = BTRFS_I(inode)->root;
 		u64 len = state->end + 1 - state->start;
-		bool do_list = !is_free_space_inode(root, inode);
+		bool do_list = !btrfs_is_free_space_inode(root, inode);
 
-		if (*bits & EXTENT_FIRST_DELALLOC)
+		if (*bits & EXTENT_FIRST_DELALLOC) {
 			*bits &= ~EXTENT_FIRST_DELALLOC;
-		else
-			atomic_inc(&BTRFS_I(inode)->outstanding_extents);
+		} else {
+			spin_lock(&BTRFS_I(inode)->lock);
+			BTRFS_I(inode)->outstanding_extents++;
+			spin_unlock(&BTRFS_I(inode)->lock);
+		}
 
 		spin_lock(&root->fs_info->delalloc_lock);
 		BTRFS_I(inode)->delalloc_bytes += len;
@@ -1370,12 +1368,15 @@ static int btrfs_clear_bit_hook(struct inode *inode,
 	if ((state->state & EXTENT_DELALLOC) && (*bits & EXTENT_DELALLOC)) {
 		struct btrfs_root *root = BTRFS_I(inode)->root;
 		u64 len = state->end + 1 - state->start;
-		bool do_list = !is_free_space_inode(root, inode);
+		bool do_list = !btrfs_is_free_space_inode(root, inode);
 
-		if (*bits & EXTENT_FIRST_DELALLOC)
+		if (*bits & EXTENT_FIRST_DELALLOC) {
 			*bits &= ~EXTENT_FIRST_DELALLOC;
-		else if (!(*bits & EXTENT_DO_ACCOUNTING))
-			atomic_dec(&BTRFS_I(inode)->outstanding_extents);
+		} else if (!(*bits & EXTENT_DO_ACCOUNTING)) {
+			spin_lock(&BTRFS_I(inode)->lock);
+			BTRFS_I(inode)->outstanding_extents--;
+			spin_unlock(&BTRFS_I(inode)->lock);
+		}
 
 		if (*bits & EXTENT_DO_ACCOUNTING)
 			btrfs_delalloc_release_metadata(inode, len);
@@ -1477,7 +1478,7 @@ static int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
 
 	skip_sum = BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM;
 
-	if (is_free_space_inode(root, inode))
+	if (btrfs_is_free_space_inode(root, inode))
 		ret = btrfs_bio_wq_end_io(root->fs_info, bio, 2);
 	else
 		ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0);
@@ -1726,7 +1727,7 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end)
 		return 0;
 	BUG_ON(!ordered_extent);
 
-	nolock = is_free_space_inode(root, inode);
+	nolock = btrfs_is_free_space_inode(root, inode);
 
 	if (test_bit(BTRFS_ORDERED_NOCOW, &ordered_extent->flags)) {
 		BUG_ON(!list_empty(&ordered_extent->list));
@@ -2531,13 +2532,6 @@ static void btrfs_read_locked_inode(struct inode *inode)
 
 	inode_item = btrfs_item_ptr(leaf, path->slots[0],
 				    struct btrfs_inode_item);
-	if (!leaf->map_token)
-		map_private_extent_buffer(leaf, (unsigned long)inode_item,
-					  sizeof(struct btrfs_inode_item),
-					  &leaf->map_token, &leaf->kaddr,
-					  &leaf->map_start, &leaf->map_len,
-					  KM_USER1);
-
 	inode->i_mode = btrfs_inode_mode(leaf, inode_item);
 	inode->i_nlink = btrfs_inode_nlink(leaf, inode_item);
 	inode->i_uid = btrfs_inode_uid(leaf, inode_item);
@@ -2575,11 +2569,6 @@ static void btrfs_read_locked_inode(struct inode *inode)
 	if (!maybe_acls)
 		cache_no_acl(inode);
 
-	if (leaf->map_token) {
-		unmap_extent_buffer(leaf, leaf->map_token, KM_USER1);
-		leaf->map_token = NULL;
-	}
-
 	btrfs_free_path(path);
 
 	switch (inode->i_mode & S_IFMT) {
@@ -2624,13 +2613,6 @@ static void fill_inode_item(struct btrfs_trans_handle *trans,
 			    struct btrfs_inode_item *item,
 			    struct inode *inode)
 {
-	if (!leaf->map_token)
-		map_private_extent_buffer(leaf, (unsigned long)item,
-					  sizeof(struct btrfs_inode_item),
-					  &leaf->map_token, &leaf->kaddr,
-					  &leaf->map_start, &leaf->map_len,
-					  KM_USER1);
-
 	btrfs_set_inode_uid(leaf, item, inode->i_uid);
 	btrfs_set_inode_gid(leaf, item, inode->i_gid);
 	btrfs_set_inode_size(leaf, item, BTRFS_I(inode)->disk_i_size);
@@ -2659,11 +2641,6 @@ static void fill_inode_item(struct btrfs_trans_handle *trans,
 	btrfs_set_inode_rdev(leaf, item, inode->i_rdev);
 	btrfs_set_inode_flags(leaf, item, BTRFS_I(inode)->flags);
 	btrfs_set_inode_block_group(leaf, item, 0);
-
-	if (leaf->map_token) {
-		unmap_extent_buffer(leaf, leaf->map_token, KM_USER1);
-		leaf->map_token = NULL;
-	}
 }
 
 /*
@@ -2684,7 +2661,7 @@ noinline int btrfs_update_inode(struct btrfs_trans_handle *trans,
 	 * The data relocation inode should also be directly updated
 	 * without delay
 	 */
-	if (!is_free_space_inode(root, inode)
+	if (!btrfs_is_free_space_inode(root, inode)
 	    && root->root_key.objectid != BTRFS_DATA_RELOC_TREE_OBJECTID) {
 		ret = btrfs_delayed_update_inode(trans, root, inode);
 		if (!ret)
@@ -3398,7 +3375,7 @@ static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
 
 	ret = -ENOMEM;
 again:
-	page = grab_cache_page(mapping, index);
+	page = find_or_create_page(mapping, index, GFP_NOFS);
 	if (!page) {
 		btrfs_delalloc_release_space(inode, PAGE_CACHE_SIZE);
 		goto out;
@@ -3634,7 +3611,7 @@ void btrfs_evict_inode(struct inode *inode)
 
 	truncate_inode_pages(&inode->i_data, 0);
 	if (inode->i_nlink && (btrfs_root_refs(&root->root_item) != 0 ||
-			       is_free_space_inode(root, inode)))
+			       btrfs_is_free_space_inode(root, inode)))
 		goto no_delete;
 
 	if (is_bad_inode(inode)) {
@@ -4271,7 +4248,7 @@ int btrfs_write_inode(struct inode *inode, struct writeback_control *wbc)
 	if (BTRFS_I(inode)->dummy_inode)
 		return 0;
 
-	if (btrfs_fs_closing(root->fs_info) && is_free_space_inode(root, inode))
+	if (btrfs_fs_closing(root->fs_info) && btrfs_is_free_space_inode(root, inode))
 		nolock = true;
 
 	if (wbc->sync_mode == WB_SYNC_ALL) {
@@ -6728,8 +6705,9 @@ struct inode *btrfs_alloc_inode(struct super_block *sb)
 	ei->index_cnt = (u64)-1;
 	ei->last_unlink_trans = 0;
 
-	atomic_set(&ei->outstanding_extents, 0);
-	atomic_set(&ei->reserved_extents, 0);
+	spin_lock_init(&ei->lock);
+	ei->outstanding_extents = 0;
+	ei->reserved_extents = 0;
 
 	ei->ordered_data_close = 0;
 	ei->orphan_meta_reserved = 0;
@@ -6767,8 +6745,8 @@ void btrfs_destroy_inode(struct inode *inode)
 
 	WARN_ON(!list_empty(&inode->i_dentry));
 	WARN_ON(inode->i_data.nrpages);
-	WARN_ON(atomic_read(&BTRFS_I(inode)->outstanding_extents));
-	WARN_ON(atomic_read(&BTRFS_I(inode)->reserved_extents));
+	WARN_ON(BTRFS_I(inode)->outstanding_extents);
+	WARN_ON(BTRFS_I(inode)->reserved_extents);
 
 	/*
 	 * This can happen where we create an inode, but somebody else also
@@ -6823,7 +6801,7 @@ int btrfs_drop_inode(struct inode *inode)
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 
 	if (btrfs_root_refs(&root->root_item) == 0 &&
-	    !is_free_space_inode(root, inode))
+	    !btrfs_is_free_space_inode(root, inode))
 		return 1;
 	else
 		return generic_drop_inode(inode);

fs/btrfs/ioctl.c

@@ -859,8 +859,8 @@ static int cluster_pages_for_defrag(struct inode *inode,
 	/* step one, lock all the pages */
 	for (i = 0; i < num_pages; i++) {
 		struct page *page;
-		page = grab_cache_page(inode->i_mapping,
-					    start_index + i);
+		page = find_or_create_page(inode->i_mapping,
+					    start_index + i, GFP_NOFS);
 		if (!page)
 			break;
 
@@ -930,7 +930,9 @@ static int cluster_pages_for_defrag(struct inode *inode,
 			  GFP_NOFS);
 
 	if (i_done != num_pages) {
-		atomic_inc(&BTRFS_I(inode)->outstanding_extents);
+		spin_lock(&BTRFS_I(inode)->lock);
+		BTRFS_I(inode)->outstanding_extents++;
+		spin_unlock(&BTRFS_I(inode)->lock);
 		btrfs_delalloc_release_space(inode,
 				     (num_pages - i_done) << PAGE_CACHE_SHIFT);
 	}

fs/btrfs/locking.c

@@ -24,185 +24,197 @@
 #include "extent_io.h"
 #include "locking.h"
 
-static inline void spin_nested(struct extent_buffer *eb)
-{
-	spin_lock(&eb->lock);
-}
+void btrfs_assert_tree_read_locked(struct extent_buffer *eb);
 
 /*
- * Setting a lock to blocking will drop the spinlock and set the
- * flag that forces other procs who want the lock to wait.  After
- * this you can safely schedule with the lock held.
+ * if we currently have a spinning reader or writer lock
+ * (indicated by the rw flag) this will bump the count
+ * of blocking holders and drop the spinlock.
  */
-void btrfs_set_lock_blocking(struct extent_buffer *eb)
+void btrfs_set_lock_blocking_rw(struct extent_buffer *eb, int rw)
 {
-	if (!test_bit(EXTENT_BUFFER_BLOCKING, &eb->bflags)) {
-		set_bit(EXTENT_BUFFER_BLOCKING, &eb->bflags);
-		spin_unlock(&eb->lock);
+	if (rw == BTRFS_WRITE_LOCK) {
+		if (atomic_read(&eb->blocking_writers) == 0) {
+			WARN_ON(atomic_read(&eb->spinning_writers) != 1);
+			atomic_dec(&eb->spinning_writers);
+			btrfs_assert_tree_locked(eb);
+			atomic_inc(&eb->blocking_writers);
+			write_unlock(&eb->lock);
+		}
+	} else if (rw == BTRFS_READ_LOCK) {
+		btrfs_assert_tree_read_locked(eb);
+		atomic_inc(&eb->blocking_readers);
+		WARN_ON(atomic_read(&eb->spinning_readers) == 0);
+		atomic_dec(&eb->spinning_readers);
+		read_unlock(&eb->lock);
 	}
-	/* exit with the spin lock released and the bit set */
+	return;
 }
 
 /*
- * clearing the blocking flag will take the spinlock again.
- * After this you can't safely schedule
+ * if we currently have a blocking lock, take the spinlock
+ * and drop our blocking count
  */
-void btrfs_clear_lock_blocking(struct extent_buffer *eb)
+void btrfs_clear_lock_blocking_rw(struct extent_buffer *eb, int rw)
 {
-	if (test_bit(EXTENT_BUFFER_BLOCKING, &eb->bflags)) {
-		spin_nested(eb);
-		clear_bit(EXTENT_BUFFER_BLOCKING, &eb->bflags);
-		smp_mb__after_clear_bit();
+	if (rw == BTRFS_WRITE_LOCK_BLOCKING) {
+		BUG_ON(atomic_read(&eb->blocking_writers) != 1);
+		write_lock(&eb->lock);
+		WARN_ON(atomic_read(&eb->spinning_writers));
+		atomic_inc(&eb->spinning_writers);
+		if (atomic_dec_and_test(&eb->blocking_writers))
+			wake_up(&eb->write_lock_wq);
+	} else if (rw == BTRFS_READ_LOCK_BLOCKING) {
+		BUG_ON(atomic_read(&eb->blocking_readers) == 0);
+		read_lock(&eb->lock);
+		atomic_inc(&eb->spinning_readers);
+		if (atomic_dec_and_test(&eb->blocking_readers))
+			wake_up(&eb->read_lock_wq);
 	}
-	/* exit with the spin lock held */
+	return;
 }
 
 /*
- * unfortunately, many of the places that currently set a lock to blocking
- * don't end up blocking for very long, and often they don't block
- * at all.  For a dbench 50 run, if we don't spin on the blocking bit
- * at all, the context switch rate can jump up to 400,000/sec or more.
- *
- * So, we're still stuck with this crummy spin on the blocking bit,
- * at least until the most common causes of the short blocks
- * can be dealt with.
+ * take a spinning read lock.  This will wait for any blocking
+ * writers
  */
-static int btrfs_spin_on_block(struct extent_buffer *eb)
+void btrfs_tree_read_lock(struct extent_buffer *eb)
 {
-	int i;
-
-	for (i = 0; i < 512; i++) {
-		if (!test_bit(EXTENT_BUFFER_BLOCKING, &eb->bflags))
-			return 1;
-		if (need_resched())
-			break;
-		cpu_relax();
+again:
+	wait_event(eb->write_lock_wq, atomic_read(&eb->blocking_writers) == 0);
+	read_lock(&eb->lock);
+	if (atomic_read(&eb->blocking_writers)) {
+		read_unlock(&eb->lock);
+		wait_event(eb->write_lock_wq,
+			   atomic_read(&eb->blocking_writers) == 0);
+		goto again;
 	}
-	return 0;
+	atomic_inc(&eb->read_locks);
+	atomic_inc(&eb->spinning_readers);
 }
 
 /*
- * This is somewhat different from trylock.  It will take the
- * spinlock but if it finds the lock is set to blocking, it will
- * return without the lock held.
- *
- * returns 1 if it was able to take the lock and zero otherwise
- *
- * After this call, scheduling is not safe without first calling
- * btrfs_set_lock_blocking()
+ * returns 1 if we get the read lock and 0 if we don't
+ * this won't wait for blocking writers
  */
-int btrfs_try_spin_lock(struct extent_buffer *eb)
+int btrfs_try_tree_read_lock(struct extent_buffer *eb)
 {
-	int i;
+	if (atomic_read(&eb->blocking_writers))
+		return 0;
 
-	if (btrfs_spin_on_block(eb)) {
-		spin_nested(eb);
-		if (!test_bit(EXTENT_BUFFER_BLOCKING, &eb->bflags))
-			return 1;
-		spin_unlock(&eb->lock);
+	read_lock(&eb->lock);
+	if (atomic_read(&eb->blocking_writers)) {
+		read_unlock(&eb->lock);
+		return 0;
 	}
-	/* spin for a bit on the BLOCKING flag */
-	for (i = 0; i < 2; i++) {
-		cpu_relax();
-		if (!btrfs_spin_on_block(eb))
-			break;
-
-		spin_nested(eb);
-		if (!test_bit(EXTENT_BUFFER_BLOCKING, &eb->bflags))
-			return 1;
-		spin_unlock(&eb->lock);
-	}
-	return 0;
+	atomic_inc(&eb->read_locks);
+	atomic_inc(&eb->spinning_readers);
+	return 1;
 }
 
 /*
- * the autoremove wake function will return 0 if it tried to wake up
- * a process that was already awake, which means that process won't
- * count as an exclusive wakeup.  The waitq code will continue waking
- * procs until it finds one that was actually sleeping.
- *
- * For btrfs, this isn't quite what we want.  We want a single proc
- * to be notified that the lock is ready for taking.  If that proc
- * already happen to be awake, great, it will loop around and try for
- * the lock.
- *
- * So, btrfs_wake_function always returns 1, even when the proc that we
- * tried to wake up was already awake.
+ * returns 1 if we get the read lock and 0 if we don't
+ * this won't wait for blocking writers or readers
  */
-static int btrfs_wake_function(wait_queue_t *wait, unsigned mode,
-			       int sync, void *key)
+int btrfs_try_tree_write_lock(struct extent_buffer *eb)
 {
-	autoremove_wake_function(wait, mode, sync, key);
+	if (atomic_read(&eb->blocking_writers) ||
+	    atomic_read(&eb->blocking_readers))
+		return 0;
+	write_lock(&eb->lock);
+	if (atomic_read(&eb->blocking_writers) ||
+	    atomic_read(&eb->blocking_readers)) {
+		write_unlock(&eb->lock);
+		return 0;
+	}
+	atomic_inc(&eb->write_locks);
+	atomic_inc(&eb->spinning_writers);
 	return 1;
 }
 
 /*
- * returns with the extent buffer spinlocked.
- *
- * This will spin and/or wait as required to take the lock, and then
- * return with the spinlock held.
- *
- * After this call, scheduling is not safe without first calling
- * btrfs_set_lock_blocking()
+ * drop a spinning read lock
+ */
+void btrfs_tree_read_unlock(struct extent_buffer *eb)
+{
+	btrfs_assert_tree_read_locked(eb);
+	WARN_ON(atomic_read(&eb->spinning_readers) == 0);
+	atomic_dec(&eb->spinning_readers);
+	atomic_dec(&eb->read_locks);
+	read_unlock(&eb->lock);
+}
+
+/*
+ * drop a blocking read lock
+ */
+void btrfs_tree_read_unlock_blocking(struct extent_buffer *eb)
+{
+	btrfs_assert_tree_read_locked(eb);
+	WARN_ON(atomic_read(&eb->blocking_readers) == 0);
+	if (atomic_dec_and_test(&eb->blocking_readers))
+		wake_up(&eb->read_lock_wq);
+	atomic_dec(&eb->read_locks);
+}
+
+/*
+ * take a spinning write lock.  This will wait for both
+ * blocking readers or writers
  */
 int btrfs_tree_lock(struct extent_buffer *eb)
 {
-	DEFINE_WAIT(wait);
-	wait.func = btrfs_wake_function;
-
-	if (!btrfs_spin_on_block(eb))
-		goto sleep;
-
-	while(1) {
-		spin_nested(eb);
-
-		/* nobody is blocking, exit with the spinlock held */
-		if (!test_bit(EXTENT_BUFFER_BLOCKING, &eb->bflags))
-			return 0;
-
-		/*
-		 * we have the spinlock, but the real owner is blocking.
-		 * wait for them
-		 */
-		spin_unlock(&eb->lock);
-
-		/*
-		 * spin for a bit, and if the blocking flag goes away,
-		 * loop around
-		 */
-		cpu_relax();
-		if (btrfs_spin_on_block(eb))
-			continue;
-sleep:
-		prepare_to_wait_exclusive(&eb->lock_wq, &wait,
-					  TASK_UNINTERRUPTIBLE);
-
-		if (test_bit(EXTENT_BUFFER_BLOCKING, &eb->bflags))
-			schedule();
-
-		finish_wait(&eb->lock_wq, &wait);
+again:
+	wait_event(eb->read_lock_wq, atomic_read(&eb->blocking_readers) == 0);
+	wait_event(eb->write_lock_wq, atomic_read(&eb->blocking_writers) == 0);
+	write_lock(&eb->lock);
+	if (atomic_read(&eb->blocking_readers)) {
+		write_unlock(&eb->lock);
+		wait_event(eb->read_lock_wq,
+			   atomic_read(&eb->blocking_readers) == 0);
+		goto again;
 	}
+	if (atomic_read(&eb->blocking_writers)) {
+		write_unlock(&eb->lock);
+		wait_event(eb->write_lock_wq,
+			   atomic_read(&eb->blocking_writers) == 0);
+		goto again;
+	}
+	WARN_ON(atomic_read(&eb->spinning_writers));
+	atomic_inc(&eb->spinning_writers);
+	atomic_inc(&eb->write_locks);
 	return 0;
 }
 
+/*
+ * drop a spinning or a blocking write lock.
+ */
 int btrfs_tree_unlock(struct extent_buffer *eb)
 {
-	/*
-	 * if we were a blocking owner, we don't have the spinlock held
-	 * just clear the bit and look for waiters
-	 */
-	if (test_and_clear_bit(EXTENT_BUFFER_BLOCKING, &eb->bflags))
-		smp_mb__after_clear_bit();
-	else
-		spin_unlock(&eb->lock);
-
-	if (waitqueue_active(&eb->lock_wq))
-		wake_up(&eb->lock_wq);
+	int blockers = atomic_read(&eb->blocking_writers);
+
+	BUG_ON(blockers > 1);
+
+	btrfs_assert_tree_locked(eb);
+	atomic_dec(&eb->write_locks);
+
+	if (blockers) {
+		WARN_ON(atomic_read(&eb->spinning_writers));
+		atomic_dec(&eb->blocking_writers);
+		smp_wmb();
+		wake_up(&eb->write_lock_wq);
+	} else {
+		WARN_ON(atomic_read(&eb->spinning_writers) != 1);
+		atomic_dec(&eb->spinning_writers);
+		write_unlock(&eb->lock);
+	}
 	return 0;
 }
 
 void btrfs_assert_tree_locked(struct extent_buffer *eb)
 {
-	if (!test_bit(EXTENT_BUFFER_BLOCKING, &eb->bflags))
-		assert_spin_locked(&eb->lock);
+	BUG_ON(!atomic_read(&eb->write_locks));
+}
+
+void btrfs_assert_tree_read_locked(struct extent_buffer *eb)
+{
+	BUG_ON(!atomic_read(&eb->read_locks));
 }

fs/btrfs/locking.h

@@ -19,11 +19,43 @@
 #ifndef __BTRFS_LOCKING_
 #define __BTRFS_LOCKING_
 
+#define BTRFS_WRITE_LOCK 1
+#define BTRFS_READ_LOCK 2
+#define BTRFS_WRITE_LOCK_BLOCKING 3
+#define BTRFS_READ_LOCK_BLOCKING 4
+
 int btrfs_tree_lock(struct extent_buffer *eb);
 int btrfs_tree_unlock(struct extent_buffer *eb);
 int btrfs_try_spin_lock(struct extent_buffer *eb);
 
-void btrfs_set_lock_blocking(struct extent_buffer *eb);
-void btrfs_clear_lock_blocking(struct extent_buffer *eb);
+void btrfs_tree_read_lock(struct extent_buffer *eb);
+void btrfs_tree_read_unlock(struct extent_buffer *eb);
+void btrfs_tree_read_unlock_blocking(struct extent_buffer *eb);
+void btrfs_set_lock_blocking_rw(struct extent_buffer *eb, int rw);
+void btrfs_clear_lock_blocking_rw(struct extent_buffer *eb, int rw);
 void btrfs_assert_tree_locked(struct extent_buffer *eb);
+int btrfs_try_tree_read_lock(struct extent_buffer *eb);
+int btrfs_try_tree_write_lock(struct extent_buffer *eb);
+
+static inline void btrfs_tree_unlock_rw(struct extent_buffer *eb, int rw)
+{
+	if (rw == BTRFS_WRITE_LOCK || rw == BTRFS_WRITE_LOCK_BLOCKING)
+		btrfs_tree_unlock(eb);
+	else if (rw == BTRFS_READ_LOCK_BLOCKING)
+		btrfs_tree_read_unlock_blocking(eb);
+	else if (rw == BTRFS_READ_LOCK)
+		btrfs_tree_read_unlock(eb);
+	else
+		BUG();
+}
+
+static inline void btrfs_set_lock_blocking(struct extent_buffer *eb)
+{
+	btrfs_set_lock_blocking_rw(eb, BTRFS_WRITE_LOCK);
+}
+
+static inline void btrfs_clear_lock_blocking(struct extent_buffer *eb)
+{
+	btrfs_clear_lock_blocking_rw(eb, BTRFS_WRITE_LOCK_BLOCKING);
+}
 #endif

fs/btrfs/relocation.c

@@ -2955,7 +2955,8 @@ static int relocate_file_extent_cluster(struct inode *inode,
 			page_cache_sync_readahead(inode->i_mapping,
 						  ra, NULL, index,
 						  last_index + 1 - index);
-			page = grab_cache_page(inode->i_mapping, index);
+			page = find_or_create_page(inode->i_mapping, index,
+						   GFP_NOFS);
 			if (!page) {
 				btrfs_delalloc_release_metadata(inode,
 							PAGE_CACHE_SIZE);

fs/btrfs/struct-funcs.c

@@ -50,36 +50,22 @@ u##bits btrfs_##name(struct extent_buffer *eb,				\
 	unsigned long part_offset = (unsigned long)s;			\
 	unsigned long offset = part_offset + offsetof(type, member);	\
 	type *p;							\
-	/* ugly, but we want the fast path here */			\
-	if (eb->map_token && offset >= eb->map_start &&			\
-	    offset + sizeof(((type *)0)->member) <= eb->map_start +	\
-	    eb->map_len) {						\
-		p = (type *)(eb->kaddr + part_offset - eb->map_start);	\
-		return le##bits##_to_cpu(p->member);			\
-	}								\
-	{								\
-		int err;						\
-		char *map_token;					\
-		char *kaddr;						\
-		int unmap_on_exit = (eb->map_token == NULL);		\
-		unsigned long map_start;				\
-		unsigned long map_len;					\
-		u##bits res;						\
-		err = map_extent_buffer(eb, offset,			\
-				sizeof(((type *)0)->member),		\
-				&map_token, &kaddr,			\
-				&map_start, &map_len, KM_USER1);	\
-		if (err) {						\
-			__le##bits leres;				\
-			read_eb_member(eb, s, type, member, &leres);	\
-			return le##bits##_to_cpu(leres);		\
-		}							\
-		p = (type *)(kaddr + part_offset - map_start);		\
-		res = le##bits##_to_cpu(p->member);			\
-		if (unmap_on_exit)					\
-			unmap_extent_buffer(eb, map_token, KM_USER1);	\
-		return res;						\
-	}								\
+	int err;						\
+	char *kaddr;						\
+	unsigned long map_start;				\
+	unsigned long map_len;					\
+	u##bits res;						\
+	err = map_private_extent_buffer(eb, offset,		\
+			sizeof(((type *)0)->member),		\
+			&kaddr, &map_start, &map_len);		\
+	if (err) {						\
+		__le##bits leres;				\
+		read_eb_member(eb, s, type, member, &leres);	\
+		return le##bits##_to_cpu(leres);		\
+	}							\
+	p = (type *)(kaddr + part_offset - map_start);		\
+	res = le##bits##_to_cpu(p->member);			\
+	return res;						\
 }									\
 void btrfs_set_##name(struct extent_buffer *eb,				\
 				    type *s, u##bits val)		\
@@ -87,36 +73,21 @@ void btrfs_set_##name(struct extent_buffer *eb,				\
 	unsigned long part_offset = (unsigned long)s;			\
 	unsigned long offset = part_offset + offsetof(type, member);	\
 	type *p;							\
-	/* ugly, but we want the fast path here */			\
-	if (eb->map_token && offset >= eb->map_start &&			\
-	    offset + sizeof(((type *)0)->member) <= eb->map_start +	\
-	    eb->map_len) {						\
-		p = (type *)(eb->kaddr + part_offset - eb->map_start);	\
-		p->member = cpu_to_le##bits(val);			\
-		return;							\
-	}								\
-	{								\
-		int err;						\
-		char *map_token;					\
-		char *kaddr;						\
-		int unmap_on_exit = (eb->map_token == NULL);		\
-		unsigned long map_start;				\
-		unsigned long map_len;					\
-		err = map_extent_buffer(eb, offset,			\
-				sizeof(((type *)0)->member),		\
-				&map_token, &kaddr,			\
-				&map_start, &map_len, KM_USER1);	\
-		if (err) {						\
-			__le##bits val2;				\
-			val2 = cpu_to_le##bits(val);			\
-			write_eb_member(eb, s, type, member, &val2);	\
-			return;						\
-		}							\
-		p = (type *)(kaddr + part_offset - map_start);		\
-		p->member = cpu_to_le##bits(val);			\
-		if (unmap_on_exit)					\
-			unmap_extent_buffer(eb, map_token, KM_USER1);	\
-	}								\
+	int err;						\
+	char *kaddr;						\
+	unsigned long map_start;				\
+	unsigned long map_len;					\
+	err = map_private_extent_buffer(eb, offset,		\
+			sizeof(((type *)0)->member),		\
+			&kaddr, &map_start, &map_len);		\
+	if (err) {						\
+		__le##bits val2;				\
+		val2 = cpu_to_le##bits(val);			\
+		write_eb_member(eb, s, type, member, &val2);	\
+		return;						\
+	}							\
+	p = (type *)(kaddr + part_offset - map_start);		\
+	p->member = cpu_to_le##bits(val);			\
 }
 
 #include "ctree.h"
@@ -125,15 +96,6 @@ void btrfs_node_key(struct extent_buffer *eb,
 		    struct btrfs_disk_key *disk_key, int nr)
 {
 	unsigned long ptr = btrfs_node_key_ptr_offset(nr);
-	if (eb->map_token && ptr >= eb->map_start &&
-	    ptr + sizeof(*disk_key) <= eb->map_start + eb->map_len) {
-		memcpy(disk_key, eb->kaddr + ptr - eb->map_start,
-			sizeof(*disk_key));
-		return;
-	} else if (eb->map_token) {
-		unmap_extent_buffer(eb, eb->map_token, KM_USER1);
-		eb->map_token = NULL;
-	}
 	read_eb_member(eb, (struct btrfs_key_ptr *)ptr,
 		       struct btrfs_key_ptr, key, disk_key);
 }

fs/btrfs/transaction.c

@@ -260,7 +260,7 @@ static struct btrfs_trans_handle *start_transaction(struct btrfs_root *root,
 {
 	struct btrfs_trans_handle *h;
 	struct btrfs_transaction *cur_trans;
-	int retries = 0;
+	u64 num_bytes = 0;
 	int ret;
 
 	if (root->fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR)
@@ -274,6 +274,19 @@ static struct btrfs_trans_handle *start_transaction(struct btrfs_root *root,
 		h->block_rsv = NULL;
 		goto got_it;
 	}
+
+	/*
+	 * Do the reservation before we join the transaction so we can do all
+	 * the appropriate flushing if need be.
+	 */
+	if (num_items > 0 && root != root->fs_info->chunk_root) {
+		num_bytes = btrfs_calc_trans_metadata_size(root, num_items);
+		ret = btrfs_block_rsv_add(NULL, root,
+					  &root->fs_info->trans_block_rsv,
+					  num_bytes);
+		if (ret)
+			return ERR_PTR(ret);
+	}
 again:
 	h = kmem_cache_alloc(btrfs_trans_handle_cachep, GFP_NOFS);
 	if (!h)
@@ -310,24 +323,9 @@ static struct btrfs_trans_handle *start_transaction(struct btrfs_root *root,
 		goto again;
 	}
 
-	if (num_items > 0) {
-		ret = btrfs_trans_reserve_metadata(h, root, num_items);
-		if (ret == -EAGAIN && !retries) {
-			retries++;
-			btrfs_commit_transaction(h, root);
-			goto again;
-		} else if (ret == -EAGAIN) {
-			/*
-			 * We have already retried and got EAGAIN, so really we
-			 * don't have space, so set ret to -ENOSPC.
-			 */
-			ret = -ENOSPC;
-		}
-
-		if (ret < 0) {
-			btrfs_end_transaction(h, root);
-			return ERR_PTR(ret);
-		}
+	if (num_bytes) {
+		h->block_rsv = &root->fs_info->trans_block_rsv;
+		h->bytes_reserved = num_bytes;
 	}
 
 got_it:
@@ -499,10 +497,17 @@ static int __btrfs_end_transaction(struct btrfs_trans_handle *trans,
 	}
 
 	if (lock && cur_trans->blocked && !cur_trans->in_commit) {
-		if (throttle)
+		if (throttle) {
+			/*
+			 * We may race with somebody else here so end up having
+			 * to call end_transaction on ourselves again, so inc
+			 * our use_count.
+			 */
+			trans->use_count++;
 			return btrfs_commit_transaction(trans, root);
-		else
+		} else {
 			wake_up_process(info->transaction_kthread);
+		}
 	}
 
 	WARN_ON(cur_trans != info->running_transaction);

fs/btrfs/tree-log.c

@@ -1730,8 +1730,8 @@ static noinline int walk_down_log_tree(struct btrfs_trans_handle *trans,
 				btrfs_read_buffer(next, ptr_gen);
 
 				btrfs_tree_lock(next);
-				clean_tree_block(trans, root, next);
 				btrfs_set_lock_blocking(next);
+				clean_tree_block(trans, root, next);
 				btrfs_wait_tree_block_writeback(next);
 				btrfs_tree_unlock(next);
 
@@ -1796,8 +1796,8 @@ static noinline int walk_up_log_tree(struct btrfs_trans_handle *trans,
 				next = path->nodes[*level];
 
 				btrfs_tree_lock(next);
-				clean_tree_block(trans, root, next);
 				btrfs_set_lock_blocking(next);
+				clean_tree_block(trans, root, next);
 				btrfs_wait_tree_block_writeback(next);
 				btrfs_tree_unlock(next);
 
@@ -1864,8 +1864,8 @@ static int walk_log_tree(struct btrfs_trans_handle *trans,
 			next = path->nodes[orig_level];
 
 			btrfs_tree_lock(next);
-			clean_tree_block(trans, log, next);
 			btrfs_set_lock_blocking(next);
+			clean_tree_block(trans, log, next);
 			btrfs_wait_tree_block_writeback(next);
 			btrfs_tree_unlock(next);
 

fs/btrfs/volumes.c

@@ -3595,7 +3595,7 @@ int btrfs_read_sys_array(struct btrfs_root *root)
 	if (!sb)
 		return -ENOMEM;
 	btrfs_set_buffer_uptodate(sb);
-	btrfs_set_buffer_lockdep_class(sb, 0);
+	btrfs_set_buffer_lockdep_class(root->root_key.objectid, sb, 0);
 
 	write_extent_buffer(sb, super_copy, 0, BTRFS_SUPER_INFO_SIZE);
 	array_size = btrfs_super_sys_array_size(super_copy);

fs/btrfs/xattr.c

@@ -102,43 +102,57 @@ static int do_setxattr(struct btrfs_trans_handle *trans,
 	if (!path)
 		return -ENOMEM;
 
-	/* first lets see if we already have this xattr */
-	di = btrfs_lookup_xattr(trans, root, path, btrfs_ino(inode), name,
-				strlen(name), -1);
-	if (IS_ERR(di)) {
-		ret = PTR_ERR(di);
-		goto out;
-	}
-
-	/* ok we already have this xattr, lets remove it */
-	if (di) {
-		/* if we want create only exit */
-		if (flags & XATTR_CREATE) {
-			ret = -EEXIST;
+	if (flags & XATTR_REPLACE) {
+		di = btrfs_lookup_xattr(trans, root, path, btrfs_ino(inode), name,
+					name_len, -1);
+		if (IS_ERR(di)) {
+			ret = PTR_ERR(di);
+			goto out;
+		} else if (!di) {
+			ret = -ENODATA;
 			goto out;
 		}
-
 		ret = btrfs_delete_one_dir_name(trans, root, path, di);
-		BUG_ON(ret);
+		if (ret)
+			goto out;
 		btrfs_release_path(path);
+	}
 
-		/* if we don't have a value then we are removing the xattr */
-		if (!value)
+again:
+	ret = btrfs_insert_xattr_item(trans, root, path, btrfs_ino(inode),
+				      name, name_len, value, size);
+	if (ret == -EEXIST) {
+		if (flags & XATTR_CREATE)
 			goto out;
-	} else {
+		/*
+		 * We can't use the path we already have since we won't have the
+		 * proper locking for a delete, so release the path and
+		 * re-lookup to delete the thing.
+		 */
 		btrfs_release_path(path);
+		di = btrfs_lookup_xattr(trans, root, path, btrfs_ino(inode),
+					name, name_len, -1);
+		if (IS_ERR(di)) {
+			ret = PTR_ERR(di);
+			goto out;
+		} else if (!di) {
+			/* Shouldn't happen but just in case... */
+			btrfs_release_path(path);
+			goto again;
+		}
 
-		if (flags & XATTR_REPLACE) {
-			/* we couldn't find the attr to replace */
-			ret = -ENODATA;
+		ret = btrfs_delete_one_dir_name(trans, root, path, di);
+		if (ret)
 			goto out;
+
+		/*
+		 * We have a value to set, so go back and try to insert it now.
+		 */
+		if (value) {
+			btrfs_release_path(path);
+			goto again;
 		}
 	}
-
-	/* ok we have to create a completely new xattr */
-	ret = btrfs_insert_xattr_item(trans, root, path, btrfs_ino(inode),
-				      name, name_len, value, size);
-	BUG_ON(ret);
 out:
 	btrfs_free_path(path);
 	return ret;