Merge branch 'for-chris-4.7' of git://git.kernel.org/pub/scm/linux/kernel/git/fdmanana/linux into for-linus-4.7

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

fs/btrfs/btrfs_inode.h

@@ -196,6 +196,16 @@ struct btrfs_inode {
 	struct list_head delayed_iput;
 	long delayed_iput_count;
 
+	/*
+	 * To avoid races between lockless (i_mutex not held) direct IO writes
+	 * and concurrent fsync requests. Direct IO writes must acquire read
+	 * access on this semaphore for creating an extent map and its
+	 * corresponding ordered extent. The fast fsync path must acquire write
+	 * access on this semaphore before it collects ordered extents and
+	 * extent maps.
+	 */
+	struct rw_semaphore dio_sem;
+
 	struct inode vfs_inode;
 };
 

fs/btrfs/ctree.h

@@ -618,6 +618,27 @@ struct btrfs_block_group_cache {
 
 	struct btrfs_io_ctl io_ctl;
 
+	/*
+	 * Incremented when doing extent allocations and holding a read lock
+	 * on the space_info's groups_sem semaphore.
+	 * Decremented when an ordered extent that represents an IO against this
+	 * block group's range is created (after it's added to its inode's
+	 * root's list of ordered extents) or immediately after the allocation
+	 * if it's a metadata extent or fallocate extent (for these cases we
+	 * don't create ordered extents).
+	 */
+	atomic_t reservations;
+
+	/*
+	 * Incremented while holding the spinlock *lock* by a task checking if
+	 * it can perform a nocow write (incremented if the value for the *ro*
+	 * field is 0). Decremented by such tasks once they create an ordered
+	 * extent or before that if some error happens before reaching that step.
+	 * This is to prevent races between block group relocation and nocow
+	 * writes through direct IO.
+	 */
+	atomic_t nocow_writers;
+
 	/* Lock for free space tree operations. */
 	struct mutex free_space_lock;
 
@@ -2487,6 +2508,12 @@ int btrfs_should_throttle_delayed_refs(struct btrfs_trans_handle *trans,
 				       struct btrfs_root *root);
 int btrfs_check_space_for_delayed_refs(struct btrfs_trans_handle *trans,
 				       struct btrfs_root *root);
+void btrfs_dec_block_group_reservations(struct btrfs_fs_info *fs_info,
+					 const u64 start);
+void btrfs_wait_block_group_reservations(struct btrfs_block_group_cache *bg);
+bool btrfs_inc_nocow_writers(struct btrfs_fs_info *fs_info, u64 bytenr);
+void btrfs_dec_nocow_writers(struct btrfs_fs_info *fs_info, u64 bytenr);
+void btrfs_wait_nocow_writers(struct btrfs_block_group_cache *bg);
 void btrfs_put_block_group(struct btrfs_block_group_cache *cache);
 int btrfs_run_delayed_refs(struct btrfs_trans_handle *trans,
 			   struct btrfs_root *root, unsigned long count);

fs/btrfs/dev-replace.c

@@ -385,7 +385,7 @@ int btrfs_dev_replace_start(struct btrfs_root *root, char *tgtdev_name,
 	if (ret)
 		btrfs_err(fs_info, "kobj add dev failed %d\n", ret);
 
-	btrfs_wait_ordered_roots(fs_info, -1);
+	btrfs_wait_ordered_roots(root->fs_info, -1, 0, (u64)-1);
 
 	/* force writing the updated state information to disk */
 	trans = btrfs_start_transaction(root, 0);
@@ -504,7 +504,7 @@ static int btrfs_dev_replace_finishing(struct btrfs_fs_info *fs_info,
 		mutex_unlock(&dev_replace->lock_finishing_cancel_unmount);
 		return ret;
 	}
-	btrfs_wait_ordered_roots(root->fs_info, -1);
+	btrfs_wait_ordered_roots(root->fs_info, -1, 0, (u64)-1);
 
 	trans = btrfs_start_transaction(root, 0);
 	if (IS_ERR(trans)) {

fs/btrfs/extent-tree.c

@@ -3824,6 +3824,59 @@ int btrfs_extent_readonly(struct btrfs_root *root, u64 bytenr)
 	return readonly;
 }
 
+bool btrfs_inc_nocow_writers(struct btrfs_fs_info *fs_info, u64 bytenr)
+{
+	struct btrfs_block_group_cache *bg;
+	bool ret = true;
+
+	bg = btrfs_lookup_block_group(fs_info, bytenr);
+	if (!bg)
+		return false;
+
+	spin_lock(&bg->lock);
+	if (bg->ro)
+		ret = false;
+	else
+		atomic_inc(&bg->nocow_writers);
+	spin_unlock(&bg->lock);
+
+	/* no put on block group, done by btrfs_dec_nocow_writers */
+	if (!ret)
+		btrfs_put_block_group(bg);
+
+	return ret;
+
+}
+
+void btrfs_dec_nocow_writers(struct btrfs_fs_info *fs_info, u64 bytenr)
+{
+	struct btrfs_block_group_cache *bg;
+
+	bg = btrfs_lookup_block_group(fs_info, bytenr);
+	ASSERT(bg);
+	if (atomic_dec_and_test(&bg->nocow_writers))
+		wake_up_atomic_t(&bg->nocow_writers);
+	/*
+	 * Once for our lookup and once for the lookup done by a previous call
+	 * to btrfs_inc_nocow_writers()
+	 */
+	btrfs_put_block_group(bg);
+	btrfs_put_block_group(bg);
+}
+
+static int btrfs_wait_nocow_writers_atomic_t(atomic_t *a)
+{
+	schedule();
+	return 0;
+}
+
+void btrfs_wait_nocow_writers(struct btrfs_block_group_cache *bg)
+{
+	wait_on_atomic_t(&bg->nocow_writers,
+			 btrfs_wait_nocow_writers_atomic_t,
+			 TASK_UNINTERRUPTIBLE);
+}
+
 static const char *alloc_name(u64 flags)
 {
 	switch (flags) {
@@ -4141,7 +4194,7 @@ commit_trans:
 
 			if (need_commit > 0) {
 				btrfs_start_delalloc_roots(fs_info, 0, -1);
-				btrfs_wait_ordered_roots(fs_info, -1);
+				btrfs_wait_ordered_roots(fs_info, -1, 0, (u64)-1);
 			}
 
 			trans = btrfs_join_transaction(root);
@@ -4583,7 +4636,8 @@ static void btrfs_writeback_inodes_sb_nr(struct btrfs_root *root,
 		 */
 		btrfs_start_delalloc_roots(root->fs_info, 0, nr_items);
 		if (!current->journal_info)
-			btrfs_wait_ordered_roots(root->fs_info, nr_items);
+			btrfs_wait_ordered_roots(root->fs_info, nr_items,
+						 0, (u64)-1);
 	}
 }
 
@@ -4632,7 +4686,8 @@ static void shrink_delalloc(struct btrfs_root *root, u64 to_reclaim, u64 orig,
 		if (trans)
 			return;
 		if (wait_ordered)
-			btrfs_wait_ordered_roots(root->fs_info, items);
+			btrfs_wait_ordered_roots(root->fs_info, items,
+						 0, (u64)-1);
 		return;
 	}
 
@@ -4671,7 +4726,8 @@ skip_async:
 
 		loops++;
 		if (wait_ordered && !trans) {
-			btrfs_wait_ordered_roots(root->fs_info, items);
+			btrfs_wait_ordered_roots(root->fs_info, items,
+						 0, (u64)-1);
 		} else {
 			time_left = schedule_timeout_killable(1);
 			if (time_left)
@@ -6172,6 +6228,57 @@ int btrfs_exclude_logged_extents(struct btrfs_root *log,
 	return 0;
 }
 
+static void
+btrfs_inc_block_group_reservations(struct btrfs_block_group_cache *bg)
+{
+	atomic_inc(&bg->reservations);
+}
+
+void btrfs_dec_block_group_reservations(struct btrfs_fs_info *fs_info,
+					const u64 start)
+{
+	struct btrfs_block_group_cache *bg;
+
+	bg = btrfs_lookup_block_group(fs_info, start);
+	ASSERT(bg);
+	if (atomic_dec_and_test(&bg->reservations))
+		wake_up_atomic_t(&bg->reservations);
+	btrfs_put_block_group(bg);
+}
+
+static int btrfs_wait_bg_reservations_atomic_t(atomic_t *a)
+{
+	schedule();
+	return 0;
+}
+
+void btrfs_wait_block_group_reservations(struct btrfs_block_group_cache *bg)
+{
+	struct btrfs_space_info *space_info = bg->space_info;
+
+	ASSERT(bg->ro);
+
+	if (!(bg->flags & BTRFS_BLOCK_GROUP_DATA))
+		return;
+
+	/*
+	 * Our block group is read only but before we set it to read only,
+	 * some task might have had allocated an extent from it already, but it
+	 * has not yet created a respective ordered extent (and added it to a
+	 * root's list of ordered extents).
+	 * Therefore wait for any task currently allocating extents, since the
+	 * block group's reservations counter is incremented while a read lock
+	 * on the groups' semaphore is held and decremented after releasing
+	 * the read access on that semaphore and creating the ordered extent.
+	 */
+	down_write(&space_info->groups_sem);
+	up_write(&space_info->groups_sem);
+
+	wait_on_atomic_t(&bg->reservations,
+			 btrfs_wait_bg_reservations_atomic_t,
+			 TASK_UNINTERRUPTIBLE);
+}
+
 /**
  * btrfs_update_reserved_bytes - update the block_group and space info counters
  * @cache:	The cache we are manipulating
@@ -7430,6 +7537,7 @@ checks:
 			btrfs_add_free_space(block_group, offset, num_bytes);
 			goto loop;
 		}
+		btrfs_inc_block_group_reservations(block_group);
 
 		/* we are all good, lets return */
 		ins->objectid = search_start;
@@ -7611,8 +7719,10 @@ again:
 	WARN_ON(num_bytes < root->sectorsize);
 	ret = find_free_extent(root, num_bytes, empty_size, hint_byte, ins,
 			       flags, delalloc);
-
-	if (ret == -ENOSPC) {
+	if (!ret && !is_data) {
+		btrfs_dec_block_group_reservations(root->fs_info,
+						   ins->objectid);
+	} else if (ret == -ENOSPC) {
 		if (!final_tried && ins->offset) {
 			num_bytes = min(num_bytes >> 1, ins->offset);
 			num_bytes = round_down(num_bytes, root->sectorsize);

fs/btrfs/inode.c

@@ -824,6 +824,7 @@ retry:
 						async_extent->ram_size - 1, 0);
 			goto out_free_reserve;
 		}
+		btrfs_dec_block_group_reservations(root->fs_info, ins.objectid);
 
 		/*
 		 * clear dirty, set writeback and unlock the pages.
@@ -861,6 +862,7 @@ retry:
 	}
 	return;
 out_free_reserve:
+	btrfs_dec_block_group_reservations(root->fs_info, ins.objectid);
 	btrfs_free_reserved_extent(root, ins.objectid, ins.offset, 1);
 out_free:
 	extent_clear_unlock_delalloc(inode, async_extent->start,
@@ -1038,6 +1040,8 @@ static noinline int cow_file_range(struct inode *inode,
 				goto out_drop_extent_cache;
 		}
 
+		btrfs_dec_block_group_reservations(root->fs_info, ins.objectid);
+
 		if (disk_num_bytes < cur_alloc_size)
 			break;
 
@@ -1066,6 +1070,7 @@ out:
 out_drop_extent_cache:
 	btrfs_drop_extent_cache(inode, start, start + ram_size - 1, 0);
 out_reserve:
+	btrfs_dec_block_group_reservations(root->fs_info, ins.objectid);
 	btrfs_free_reserved_extent(root, ins.objectid, ins.offset, 1);
 out_unlock:
 	extent_clear_unlock_delalloc(inode, start, end, locked_page,
@@ -1377,6 +1382,9 @@ next_slot:
 			 */
 			if (csum_exist_in_range(root, disk_bytenr, num_bytes))
 				goto out_check;
+			if (!btrfs_inc_nocow_writers(root->fs_info,
+						     disk_bytenr))
+				goto out_check;
 			nocow = 1;
 		} else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
 			extent_end = found_key.offset +
@@ -1391,6 +1399,9 @@ out_check:
 			path->slots[0]++;
 			if (!nolock && nocow)
 				btrfs_end_write_no_snapshoting(root);
+			if (nocow)
+				btrfs_dec_nocow_writers(root->fs_info,
+							disk_bytenr);
 			goto next_slot;
 		}
 		if (!nocow) {
@@ -1411,6 +1422,9 @@ out_check:
 			if (ret) {
 				if (!nolock && nocow)
 					btrfs_end_write_no_snapshoting(root);
+				if (nocow)
+					btrfs_dec_nocow_writers(root->fs_info,
+								disk_bytenr);
 				goto error;
 			}
 			cow_start = (u64)-1;
@@ -1453,6 +1467,8 @@ out_check:
 
 		ret = btrfs_add_ordered_extent(inode, cur_offset, disk_bytenr,
 					       num_bytes, num_bytes, type);
+		if (nocow)
+			btrfs_dec_nocow_writers(root->fs_info, disk_bytenr);
 		BUG_ON(ret); /* -ENOMEM */
 
 		if (root->root_key.objectid ==
@@ -7129,6 +7145,43 @@ out:
 	return em;
 }
 
+static struct extent_map *btrfs_create_dio_extent(struct inode *inode,
+						  const u64 start,
+						  const u64 len,
+						  const u64 orig_start,
+						  const u64 block_start,
+						  const u64 block_len,
+						  const u64 orig_block_len,
+						  const u64 ram_bytes,
+						  const int type)
+{
+	struct extent_map *em = NULL;
+	int ret;
+
+	down_read(&BTRFS_I(inode)->dio_sem);
+	if (type != BTRFS_ORDERED_NOCOW) {
+		em = create_pinned_em(inode, start, len, orig_start,
+				      block_start, block_len, orig_block_len,
+				      ram_bytes, type);
+		if (IS_ERR(em))
+			goto out;
+	}
+	ret = btrfs_add_ordered_extent_dio(inode, start, block_start,
+					   len, block_len, type);
+	if (ret) {
+		if (em) {
+			free_extent_map(em);
+			btrfs_drop_extent_cache(inode, start,
+						start + len - 1, 0);
+		}
+		em = ERR_PTR(ret);
+	}
+ out:
+	up_read(&BTRFS_I(inode)->dio_sem);
+
+	return em;
+}
+
 static struct extent_map *btrfs_new_extent_direct(struct inode *inode,
 						  u64 start, u64 len)
 {
@@ -7144,41 +7197,13 @@ static struct extent_map *btrfs_new_extent_direct(struct inode *inode,
 	if (ret)
 		return ERR_PTR(ret);
 
-	/*
-	 * Create the ordered extent before the extent map. This is to avoid
-	 * races with the fast fsync path that would lead to it logging file
-	 * extent items that point to disk extents that were not yet written to.
-	 * The fast fsync path collects ordered extents into a local list and
-	 * then collects all the new extent maps, so we must create the ordered
-	 * extent first and make sure the fast fsync path collects any new
-	 * ordered extents after collecting new extent maps as well.
-	 * The fsync path simply can not rely on inode_dio_wait() because it
-	 * causes deadlock with AIO.
-	 */
-	ret = btrfs_add_ordered_extent_dio(inode, start, ins.objectid,
-					   ins.offset, ins.offset, 0);
-	if (ret) {
+	em = btrfs_create_dio_extent(inode, start, ins.offset, start,
+				     ins.objectid, ins.offset, ins.offset,
+				     ins.offset, 0);
+	btrfs_dec_block_group_reservations(root->fs_info, ins.objectid);
+	if (IS_ERR(em))
 		btrfs_free_reserved_extent(root, ins.objectid, ins.offset, 1);
-		return ERR_PTR(ret);
-	}
 
-	em = create_pinned_em(inode, start, ins.offset, start, ins.objectid,
-			      ins.offset, ins.offset, ins.offset, 0);
-	if (IS_ERR(em)) {
-		struct btrfs_ordered_extent *oe;
-
-		btrfs_free_reserved_extent(root, ins.objectid, ins.offset, 1);
-		oe = btrfs_lookup_ordered_extent(inode, start);
-		ASSERT(oe);
-		if (WARN_ON(!oe))
-			return em;
-		set_bit(BTRFS_ORDERED_IOERR, &oe->flags);
-		set_bit(BTRFS_ORDERED_IO_DONE, &oe->flags);
-		btrfs_remove_ordered_extent(inode, oe);
-		/* Once for our lookup and once for the ordered extents tree. */
-		btrfs_put_ordered_extent(oe);
-		btrfs_put_ordered_extent(oe);
-	}
 	return em;
 }
 
@@ -7650,24 +7675,21 @@ static int btrfs_get_blocks_direct(struct inode *inode, sector_t iblock,
 		block_start = em->block_start + (start - em->start);
 
 		if (can_nocow_extent(inode, start, &len, &orig_start,
-				     &orig_block_len, &ram_bytes) == 1) {
+				     &orig_block_len, &ram_bytes) == 1 &&
+		    btrfs_inc_nocow_writers(root->fs_info, block_start)) {
+			struct extent_map *em2;
+
+			em2 = btrfs_create_dio_extent(inode, start, len,
+						      orig_start, block_start,
+						      len, orig_block_len,
+						      ram_bytes, type);
+			btrfs_dec_nocow_writers(root->fs_info, block_start);
 			if (type == BTRFS_ORDERED_PREALLOC) {
 				free_extent_map(em);
-				em = create_pinned_em(inode, start, len,
-						       orig_start,
-						       block_start, len,
-						       orig_block_len,
-						       ram_bytes, type);
-				if (IS_ERR(em)) {
-					ret = PTR_ERR(em);
-					goto unlock_err;
-				}
+				em = em2;
 			}
-
-			ret = btrfs_add_ordered_extent_dio(inode, start,
-					   block_start, len, len, type);
-			if (ret) {
-				free_extent_map(em);
+			if (em2 && IS_ERR(em2)) {
+				ret = PTR_ERR(em2);
 				goto unlock_err;
 			}
 			goto unlock;
@@ -9230,6 +9252,7 @@ struct inode *btrfs_alloc_inode(struct super_block *sb)
 	INIT_LIST_HEAD(&ei->delalloc_inodes);
 	INIT_LIST_HEAD(&ei->delayed_iput);
 	RB_CLEAR_NODE(&ei->rb_node);
+	init_rwsem(&ei->dio_sem);
 
 	return inode;
 }
@@ -9387,18 +9410,290 @@ static int btrfs_getattr(struct vfsmount *mnt,
 	return 0;
 }
 
-static int btrfs_rename(struct inode *old_dir, struct dentry *old_dentry,
-			   struct inode *new_dir, struct dentry *new_dentry)
+static int btrfs_rename_exchange(struct inode *old_dir,
+			      struct dentry *old_dentry,
+			      struct inode *new_dir,
+			      struct dentry *new_dentry)
 {
 	struct btrfs_trans_handle *trans;
 	struct btrfs_root *root = BTRFS_I(old_dir)->root;
 	struct btrfs_root *dest = BTRFS_I(new_dir)->root;
+	struct inode *new_inode = new_dentry->d_inode;
+	struct inode *old_inode = old_dentry->d_inode;
+	struct timespec ctime = CURRENT_TIME;
+	struct dentry *parent;
+	u64 old_ino = btrfs_ino(old_inode);
+	u64 new_ino = btrfs_ino(new_inode);
+	u64 old_idx = 0;
+	u64 new_idx = 0;
+	u64 root_objectid;
+	int ret;
+	bool root_log_pinned = false;
+	bool dest_log_pinned = false;
+
+	/* we only allow rename subvolume link between subvolumes */
+	if (old_ino != BTRFS_FIRST_FREE_OBJECTID && root != dest)
+		return -EXDEV;
+
+	/* close the race window with snapshot create/destroy ioctl */
+	if (old_ino == BTRFS_FIRST_FREE_OBJECTID)
+		down_read(&root->fs_info->subvol_sem);
+	if (new_ino == BTRFS_FIRST_FREE_OBJECTID)
+		down_read(&dest->fs_info->subvol_sem);
+
+	/*
+	 * We want to reserve the absolute worst case amount of items.  So if
+	 * both inodes are subvols and we need to unlink them then that would
+	 * require 4 item modifications, but if they are both normal inodes it
+	 * would require 5 item modifications, so we'll assume their normal
+	 * inodes.  So 5 * 2 is 10, plus 2 for the new links, so 12 total items
+	 * should cover the worst case number of items we'll modify.
+	 */
+	trans = btrfs_start_transaction(root, 12);
+	if (IS_ERR(trans)) {
+		ret = PTR_ERR(trans);
+		goto out_notrans;
+	}
+
+	/*
+	 * We need to find a free sequence number both in the source and
+	 * in the destination directory for the exchange.
+	 */
+	ret = btrfs_set_inode_index(new_dir, &old_idx);
+	if (ret)
+		goto out_fail;
+	ret = btrfs_set_inode_index(old_dir, &new_idx);
+	if (ret)
+		goto out_fail;
+
+	BTRFS_I(old_inode)->dir_index = 0ULL;
+	BTRFS_I(new_inode)->dir_index = 0ULL;
+
+	/* Reference for the source. */
+	if (old_ino == BTRFS_FIRST_FREE_OBJECTID) {
+		/* force full log commit if subvolume involved. */
+		btrfs_set_log_full_commit(root->fs_info, trans);
+	} else {
+		btrfs_pin_log_trans(root);
+		root_log_pinned = true;
+		ret = btrfs_insert_inode_ref(trans, dest,
+					     new_dentry->d_name.name,
+					     new_dentry->d_name.len,
+					     old_ino,
+					     btrfs_ino(new_dir), old_idx);
+		if (ret)
+			goto out_fail;
+	}
+
+	/* And now for the dest. */
+	if (new_ino == BTRFS_FIRST_FREE_OBJECTID) {
+		/* force full log commit if subvolume involved. */
+		btrfs_set_log_full_commit(dest->fs_info, trans);
+	} else {
+		btrfs_pin_log_trans(dest);
+		dest_log_pinned = true;
+		ret = btrfs_insert_inode_ref(trans, root,
+					     old_dentry->d_name.name,
+					     old_dentry->d_name.len,
+					     new_ino,
+					     btrfs_ino(old_dir), new_idx);
+		if (ret)
+			goto out_fail;
+	}
+
+	/* Update inode version and ctime/mtime. */
+	inode_inc_iversion(old_dir);
+	inode_inc_iversion(new_dir);
+	inode_inc_iversion(old_inode);
+	inode_inc_iversion(new_inode);
+	old_dir->i_ctime = old_dir->i_mtime = ctime;
+	new_dir->i_ctime = new_dir->i_mtime = ctime;
+	old_inode->i_ctime = ctime;
+	new_inode->i_ctime = ctime;
+
+	if (old_dentry->d_parent != new_dentry->d_parent) {
+		btrfs_record_unlink_dir(trans, old_dir, old_inode, 1);
+		btrfs_record_unlink_dir(trans, new_dir, new_inode, 1);
+	}
+
+	/* src is a subvolume */
+	if (old_ino == BTRFS_FIRST_FREE_OBJECTID) {
+		root_objectid = BTRFS_I(old_inode)->root->root_key.objectid;
+		ret = btrfs_unlink_subvol(trans, root, old_dir,
+					  root_objectid,
+					  old_dentry->d_name.name,
+					  old_dentry->d_name.len);
+	} else { /* src is an inode */
+		ret = __btrfs_unlink_inode(trans, root, old_dir,
+					   old_dentry->d_inode,
+					   old_dentry->d_name.name,
+					   old_dentry->d_name.len);
+		if (!ret)
+			ret = btrfs_update_inode(trans, root, old_inode);
+	}
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
+		goto out_fail;
+	}
+
+	/* dest is a subvolume */
+	if (new_ino == BTRFS_FIRST_FREE_OBJECTID) {
+		root_objectid = BTRFS_I(new_inode)->root->root_key.objectid;
+		ret = btrfs_unlink_subvol(trans, dest, new_dir,
+					  root_objectid,
+					  new_dentry->d_name.name,
+					  new_dentry->d_name.len);
+	} else { /* dest is an inode */
+		ret = __btrfs_unlink_inode(trans, dest, new_dir,
+					   new_dentry->d_inode,
+					   new_dentry->d_name.name,
+					   new_dentry->d_name.len);
+		if (!ret)
+			ret = btrfs_update_inode(trans, dest, new_inode);
+	}
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
+		goto out_fail;
+	}
+
+	ret = btrfs_add_link(trans, new_dir, old_inode,
+			     new_dentry->d_name.name,
+			     new_dentry->d_name.len, 0, old_idx);
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
+		goto out_fail;
+	}
+
+	ret = btrfs_add_link(trans, old_dir, new_inode,
+			     old_dentry->d_name.name,
+			     old_dentry->d_name.len, 0, new_idx);
+	if (ret) {
+		btrfs_abort_transaction(trans, root, ret);
+		goto out_fail;
+	}
+
+	if (old_inode->i_nlink == 1)
+		BTRFS_I(old_inode)->dir_index = old_idx;
+	if (new_inode->i_nlink == 1)
+		BTRFS_I(new_inode)->dir_index = new_idx;
+
+	if (root_log_pinned) {
+		parent = new_dentry->d_parent;
+		btrfs_log_new_name(trans, old_inode, old_dir, parent);
+		btrfs_end_log_trans(root);
+		root_log_pinned = false;
+	}
+	if (dest_log_pinned) {
+		parent = old_dentry->d_parent;
+		btrfs_log_new_name(trans, new_inode, new_dir, parent);
+		btrfs_end_log_trans(dest);
+		dest_log_pinned = false;
+	}
+out_fail:
+	/*
+	 * If we have pinned a log and an error happened, we unpin tasks
+	 * trying to sync the log and force them to fallback to a transaction
+	 * commit if the log currently contains any of the inodes involved in
+	 * this rename operation (to ensure we do not persist a log with an
+	 * inconsistent state for any of these inodes or leading to any
+	 * inconsistencies when replayed). If the transaction was aborted, the
+	 * abortion reason is propagated to userspace when attempting to commit
+	 * the transaction. If the log does not contain any of these inodes, we
+	 * allow the tasks to sync it.
+	 */
+	if (ret && (root_log_pinned || dest_log_pinned)) {
+		if (btrfs_inode_in_log(old_dir, root->fs_info->generation) ||
+		    btrfs_inode_in_log(new_dir, root->fs_info->generation) ||
+		    btrfs_inode_in_log(old_inode, root->fs_info->generation) ||
+		    (new_inode &&
+		     btrfs_inode_in_log(new_inode, root->fs_info->generation)))
+		    btrfs_set_log_full_commit(root->fs_info, trans);
+
+		if (root_log_pinned) {
+			btrfs_end_log_trans(root);
+			root_log_pinned = false;
+		}
+		if (dest_log_pinned) {
+			btrfs_end_log_trans(dest);
+			dest_log_pinned = false;
+		}
+	}
+	ret = btrfs_end_transaction(trans, root);
+out_notrans:
+	if (new_ino == BTRFS_FIRST_FREE_OBJECTID)
+		up_read(&dest->fs_info->subvol_sem);
+	if (old_ino == BTRFS_FIRST_FREE_OBJECTID)
+		up_read(&root->fs_info->subvol_sem);
+
+	return ret;
+}
+
+static int btrfs_whiteout_for_rename(struct btrfs_trans_handle *trans,
+				     struct btrfs_root *root,
+				     struct inode *dir,
+				     struct dentry *dentry)
+{
+	int ret;
+	struct inode *inode;
+	u64 objectid;
+	u64 index;
+
+	ret = btrfs_find_free_ino(root, &objectid);
+	if (ret)
+		return ret;
+
+	inode = btrfs_new_inode(trans, root, dir,
+				dentry->d_name.name,
+				dentry->d_name.len,
+				btrfs_ino(dir),
+				objectid,
+				S_IFCHR | WHITEOUT_MODE,
+				&index);
+
+	if (IS_ERR(inode)) {
+		ret = PTR_ERR(inode);
+		return ret;
+	}
+
+	inode->i_op = &btrfs_special_inode_operations;
+	init_special_inode(inode, inode->i_mode,
+		WHITEOUT_DEV);
+
+	ret = btrfs_init_inode_security(trans, inode, dir,
+				&dentry->d_name);
+	if (ret)
+		goto out;
+
+	ret = btrfs_add_nondir(trans, dir, dentry,
+				inode, 0, index);
+	if (ret)
+		goto out;
+
+	ret = btrfs_update_inode(trans, root, inode);
+out:
+	unlock_new_inode(inode);
+	if (ret)
+		inode_dec_link_count(inode);
+	iput(inode);
+
+	return ret;
+}
+
+static int btrfs_rename(struct inode *old_dir, struct dentry *old_dentry,
+			   struct inode *new_dir, struct dentry *new_dentry,
+			   unsigned int flags)
+{
+	struct btrfs_trans_handle *trans;
+	unsigned int trans_num_items;
+	struct btrfs_root *root = BTRFS_I(old_dir)->root;
+	struct btrfs_root *dest = BTRFS_I(new_dir)->root;
 	struct inode *new_inode = d_inode(new_dentry);
 	struct inode *old_inode = d_inode(old_dentry);
 	u64 index = 0;
 	u64 root_objectid;
 	int ret;
 	u64 old_ino = btrfs_ino(old_inode);
+	bool log_pinned = false;
 
 	if (btrfs_ino(new_dir) == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID)
 		return -EPERM;
@@ -9449,15 +9744,21 @@ static int btrfs_rename(struct inode *old_dir, struct dentry *old_dentry,
 	 * We want to reserve the absolute worst case amount of items.  So if
 	 * both inodes are subvols and we need to unlink them then that would
 	 * require 4 item modifications, but if they are both normal inodes it
-	 * would require 5 item modifications, so we'll assume their normal
+	 * would require 5 item modifications, so we'll assume they are normal
 	 * inodes.  So 5 * 2 is 10, plus 1 for the new link, so 11 total items
 	 * should cover the worst case number of items we'll modify.
+	 * If our rename has the whiteout flag, we need more 5 units for the
+	 * new inode (1 inode item, 1 inode ref, 2 dir items and 1 xattr item
+	 * when selinux is enabled).
 	 */
-	trans = btrfs_start_transaction(root, 11);
+	trans_num_items = 11;
+	if (flags & RENAME_WHITEOUT)
+		trans_num_items += 5;
+	trans = btrfs_start_transaction(root, trans_num_items);
 	if (IS_ERR(trans)) {
-                ret = PTR_ERR(trans);
-                goto out_notrans;
-        }
+		ret = PTR_ERR(trans);
+		goto out_notrans;
+	}
 
 	if (dest != root)
 		btrfs_record_root_in_trans(trans, dest);
@@ -9471,6 +9772,8 @@ static int btrfs_rename(struct inode *old_dir, struct dentry *old_dentry,
 		/* force full log commit if subvolume involved. */
 		btrfs_set_log_full_commit(root->fs_info, trans);
 	} else {
+		btrfs_pin_log_trans(root);
+		log_pinned = true;
 		ret = btrfs_insert_inode_ref(trans, dest,
 					     new_dentry->d_name.name,
 					     new_dentry->d_name.len,
@@ -9478,14 +9781,6 @@ static int btrfs_rename(struct inode *old_dir, struct dentry *old_dentry,
 					     btrfs_ino(new_dir), index);
 		if (ret)
 			goto out_fail;
-		/*
-		 * this is an ugly little race, but the rename is required
-		 * to make sure that if we crash, the inode is either at the
-		 * old name or the new one.  pinning the log transaction lets
-		 * us make sure we don't allow a log commit to come in after
-		 * we unlink the name but before we add the new name back in.
-		 */
-		btrfs_pin_log_trans(root);
 	}
 
 	inode_inc_iversion(old_dir);
@@ -9552,12 +9847,46 @@ static int btrfs_rename(struct inode *old_dir, struct dentry *old_dentry,
 	if (old_inode->i_nlink == 1)
 		BTRFS_I(old_inode)->dir_index = index;
 
-	if (old_ino != BTRFS_FIRST_FREE_OBJECTID) {
+	if (log_pinned) {
 		struct dentry *parent = new_dentry->d_parent;
+
 		btrfs_log_new_name(trans, old_inode, old_dir, parent);
 		btrfs_end_log_trans(root);
+		log_pinned = false;
+	}
+
+	if (flags & RENAME_WHITEOUT) {
+		ret = btrfs_whiteout_for_rename(trans, root, old_dir,
+						old_dentry);
+
+		if (ret) {
+			btrfs_abort_transaction(trans, root, ret);
+			goto out_fail;
+		}
 	}
 out_fail:
+	/*
+	 * If we have pinned the log and an error happened, we unpin tasks
+	 * trying to sync the log and force them to fallback to a transaction
+	 * commit if the log currently contains any of the inodes involved in
+	 * this rename operation (to ensure we do not persist a log with an
+	 * inconsistent state for any of these inodes or leading to any
+	 * inconsistencies when replayed). If the transaction was aborted, the
+	 * abortion reason is propagated to userspace when attempting to commit
+	 * the transaction. If the log does not contain any of these inodes, we
+	 * allow the tasks to sync it.
+	 */
+	if (ret && log_pinned) {
+		if (btrfs_inode_in_log(old_dir, root->fs_info->generation) ||
+		    btrfs_inode_in_log(new_dir, root->fs_info->generation) ||
+		    btrfs_inode_in_log(old_inode, root->fs_info->generation) ||
+		    (new_inode &&
+		     btrfs_inode_in_log(new_inode, root->fs_info->generation)))
+		    btrfs_set_log_full_commit(root->fs_info, trans);
+
+		btrfs_end_log_trans(root);
+		log_pinned = false;
+	}
 	btrfs_end_transaction(trans, root);
 out_notrans:
 	if (old_ino == BTRFS_FIRST_FREE_OBJECTID)
@@ -9570,10 +9899,14 @@ static int btrfs_rename2(struct inode *old_dir, struct dentry *old_dentry,
 			 struct inode *new_dir, struct dentry *new_dentry,
 			 unsigned int flags)
 {
-	if (flags & ~RENAME_NOREPLACE)
+	if (flags & ~(RENAME_NOREPLACE | RENAME_EXCHANGE | RENAME_WHITEOUT))
 		return -EINVAL;
 
-	return btrfs_rename(old_dir, old_dentry, new_dir, new_dentry);
+	if (flags & RENAME_EXCHANGE)
+		return btrfs_rename_exchange(old_dir, old_dentry, new_dir,
+					  new_dentry);
+
+	return btrfs_rename(old_dir, old_dentry, new_dir, new_dentry, flags);
 }
 
 static void btrfs_run_delalloc_work(struct btrfs_work *work)
@@ -9942,6 +10275,7 @@ static int __btrfs_prealloc_file_range(struct inode *inode, int mode,
 				btrfs_end_transaction(trans, root);
 			break;
 		}
+		btrfs_dec_block_group_reservations(root->fs_info, ins.objectid);
 
 		last_alloc = ins.offset;
 		ret = insert_reserved_file_extent(trans, inode,

fs/btrfs/ioctl.c

@@ -690,7 +690,7 @@ static int create_snapshot(struct btrfs_root *root, struct inode *dir,
 	if (ret)
 		goto dec_and_free;
 
-	btrfs_wait_ordered_extents(root, -1);
+	btrfs_wait_ordered_extents(root, -1, 0, (u64)-1);
 
 	btrfs_init_block_rsv(&pending_snapshot->block_rsv,
 			     BTRFS_BLOCK_RSV_TEMP);

fs/btrfs/ordered-data.c

@@ -661,14 +661,15 @@ static void btrfs_run_ordered_extent_work(struct btrfs_work *work)
  * wait for all the ordered extents in a root.  This is done when balancing
  * space between drives.
  */
-int btrfs_wait_ordered_extents(struct btrfs_root *root, int nr)
+int btrfs_wait_ordered_extents(struct btrfs_root *root, int nr,
+			       const u64 range_start, const u64 range_len)
 {
-	struct list_head splice, works;
+	LIST_HEAD(splice);
+	LIST_HEAD(skipped);
+	LIST_HEAD(works);
 	struct btrfs_ordered_extent *ordered, *next;
 	int count = 0;
-
-	INIT_LIST_HEAD(&splice);
-	INIT_LIST_HEAD(&works);
+	const u64 range_end = range_start + range_len;
 
 	mutex_lock(&root->ordered_extent_mutex);
 	spin_lock(&root->ordered_extent_lock);
@@ -676,6 +677,14 @@ int btrfs_wait_ordered_extents(struct btrfs_root *root, int nr)
 	while (!list_empty(&splice) && nr) {
 		ordered = list_first_entry(&splice, struct btrfs_ordered_extent,
 					   root_extent_list);
+
+		if (range_end <= ordered->start ||
+		    ordered->start + ordered->disk_len <= range_start) {
+			list_move_tail(&ordered->root_extent_list, &skipped);
+			cond_resched_lock(&root->ordered_extent_lock);
+			continue;
+		}
+
 		list_move_tail(&ordered->root_extent_list,
 			       &root->ordered_extents);
 		atomic_inc(&ordered->refs);
@@ -694,6 +703,7 @@ int btrfs_wait_ordered_extents(struct btrfs_root *root, int nr)
 			nr--;
 		count++;
 	}
+	list_splice_tail(&skipped, &root->ordered_extents);
 	list_splice_tail(&splice, &root->ordered_extents);
 	spin_unlock(&root->ordered_extent_lock);
 
@@ -708,7 +718,8 @@ int btrfs_wait_ordered_extents(struct btrfs_root *root, int nr)
 	return count;
 }
 
-void btrfs_wait_ordered_roots(struct btrfs_fs_info *fs_info, int nr)
+void btrfs_wait_ordered_roots(struct btrfs_fs_info *fs_info, int nr,
+			      const u64 range_start, const u64 range_len)
 {
 	struct btrfs_root *root;
 	struct list_head splice;
@@ -728,7 +739,8 @@ void btrfs_wait_ordered_roots(struct btrfs_fs_info *fs_info, int nr)
 			       &fs_info->ordered_roots);
 		spin_unlock(&fs_info->ordered_root_lock);
 
-		done = btrfs_wait_ordered_extents(root, nr);
+		done = btrfs_wait_ordered_extents(root, nr,
+						  range_start, range_len);
 		btrfs_put_fs_root(root);
 
 		spin_lock(&fs_info->ordered_root_lock);

fs/btrfs/ordered-data.h

@@ -197,8 +197,10 @@ int btrfs_ordered_update_i_size(struct inode *inode, u64 offset,
 				struct btrfs_ordered_extent *ordered);
 int btrfs_find_ordered_sum(struct inode *inode, u64 offset, u64 disk_bytenr,
 			   u32 *sum, int len);
-int btrfs_wait_ordered_extents(struct btrfs_root *root, int nr);
-void btrfs_wait_ordered_roots(struct btrfs_fs_info *fs_info, int nr);
+int btrfs_wait_ordered_extents(struct btrfs_root *root, int nr,
+			       const u64 range_start, const u64 range_len);
+void btrfs_wait_ordered_roots(struct btrfs_fs_info *fs_info, int nr,
+			      const u64 range_start, const u64 range_len);
 void btrfs_get_logged_extents(struct inode *inode,
 			      struct list_head *logged_list,
 			      const loff_t start,

fs/btrfs/relocation.c

@@ -4254,12 +4254,11 @@ int btrfs_relocate_block_group(struct btrfs_root *extent_root, u64 group_start)
 	btrfs_info(extent_root->fs_info, "relocating block group %llu flags %llu",
 	       rc->block_group->key.objectid, rc->block_group->flags);
 
-	ret = btrfs_start_delalloc_roots(fs_info, 0, -1);
-	if (ret < 0) {
-		err = ret;
-		goto out;
-	}
-	btrfs_wait_ordered_roots(fs_info, -1);
+	btrfs_wait_block_group_reservations(rc->block_group);
+	btrfs_wait_nocow_writers(rc->block_group);
+	btrfs_wait_ordered_roots(fs_info, -1,
+				 rc->block_group->key.objectid,
+				 rc->block_group->key.offset);
 
 	while (1) {
 		mutex_lock(&fs_info->cleaner_mutex);

fs/btrfs/super.c

@@ -1156,7 +1156,7 @@ int btrfs_sync_fs(struct super_block *sb, int wait)
 		return 0;
 	}
 
-	btrfs_wait_ordered_roots(fs_info, -1);
+	btrfs_wait_ordered_roots(fs_info, -1, 0, (u64)-1);
 
 	trans = btrfs_attach_transaction_barrier(root);
 	if (IS_ERR(trans)) {

fs/btrfs/transaction.c

@@ -1907,7 +1907,7 @@ static inline int btrfs_start_delalloc_flush(struct btrfs_fs_info *fs_info)
 static inline void btrfs_wait_delalloc_flush(struct btrfs_fs_info *fs_info)
 {
 	if (btrfs_test_opt(fs_info->tree_root, FLUSHONCOMMIT))
-		btrfs_wait_ordered_roots(fs_info, -1);
+		btrfs_wait_ordered_roots(fs_info, -1, 0, (u64)-1);
 }
 
 static inline void

fs/btrfs/tree-log.c

@@ -4141,6 +4141,7 @@ static int btrfs_log_changed_extents(struct btrfs_trans_handle *trans,
 
 	INIT_LIST_HEAD(&extents);
 
+	down_write(&BTRFS_I(inode)->dio_sem);
 	write_lock(&tree->lock);
 	test_gen = root->fs_info->last_trans_committed;
 
@@ -4169,13 +4170,20 @@ static int btrfs_log_changed_extents(struct btrfs_trans_handle *trans,
 	}
 
 	list_sort(NULL, &extents, extent_cmp);
-	/*
-	 * Collect any new ordered extents within the range. This is to
-	 * prevent logging file extent items without waiting for the disk
-	 * location they point to being written. We do this only to deal
-	 * with races against concurrent lockless direct IO writes.
-	 */
 	btrfs_get_logged_extents(inode, logged_list, start, end);
+	/*
+	 * Some ordered extents started by fsync might have completed
+	 * before we could collect them into the list logged_list, which
+	 * means they're gone, not in our logged_list nor in the inode's
+	 * ordered tree. We want the application/user space to know an
+	 * error happened while attempting to persist file data so that
+	 * it can take proper action. If such error happened, we leave
+	 * without writing to the log tree and the fsync must report the
+	 * file data write error and not commit the current transaction.
+	 */
+	ret = btrfs_inode_check_errors(inode);
+	if (ret)
+		ctx->io_err = ret;
 process:
 	while (!list_empty(&extents)) {
 		em = list_entry(extents.next, struct extent_map, list);
@@ -4202,6 +4210,7 @@ process:
 	}
 	WARN_ON(!list_empty(&extents));
 	write_unlock(&tree->lock);
+	up_write(&BTRFS_I(inode)->dio_sem);
 
 	btrfs_release_path(path);
 	return ret;
@@ -4622,23 +4631,6 @@ static int btrfs_log_inode(struct btrfs_trans_handle *trans,
 
 	mutex_lock(&BTRFS_I(inode)->log_mutex);
 
-	/*
-	 * Collect ordered extents only if we are logging data. This is to
-	 * ensure a subsequent request to log this inode in LOG_INODE_ALL mode
-	 * will process the ordered extents if they still exists at the time,
-	 * because when we collect them we test and set for the flag
-	 * BTRFS_ORDERED_LOGGED to prevent multiple log requests to process the
-	 * same ordered extents. The consequence for the LOG_INODE_ALL log mode
-	 * not processing the ordered extents is that we end up logging the
-	 * corresponding file extent items, based on the extent maps in the
-	 * inode's extent_map_tree's modified_list, without logging the
-	 * respective checksums (since the may still be only attached to the
-	 * ordered extents and have not been inserted in the csum tree by
-	 * btrfs_finish_ordered_io() yet).
-	 */
-	if (inode_only == LOG_INODE_ALL)
-		btrfs_get_logged_extents(inode, &logged_list, start, end);
-
 	/*
 	 * a brute force approach to making sure we get the most uptodate
 	 * copies of everything.
@@ -4846,21 +4838,6 @@ log_extents:
 			goto out_unlock;
 	}
 	if (fast_search) {
-		/*
-		 * Some ordered extents started by fsync might have completed
-		 * before we collected the ordered extents in logged_list, which
-		 * means they're gone, not in our logged_list nor in the inode's
-		 * ordered tree. We want the application/user space to know an
-		 * error happened while attempting to persist file data so that
-		 * it can take proper action. If such error happened, we leave
-		 * without writing to the log tree and the fsync must report the
-		 * file data write error and not commit the current transaction.
-		 */
-		err = btrfs_inode_check_errors(inode);
-		if (err) {
-			ctx->io_err = err;
-			goto out_unlock;
-		}
 		ret = btrfs_log_changed_extents(trans, root, inode, dst_path,
 						&logged_list, ctx, start, end);
 		if (ret) {
@@ -5158,7 +5135,7 @@ process_leaf:
 			}
 
 			ctx->log_new_dentries = false;
-			if (type == BTRFS_FT_DIR)
+			if (type == BTRFS_FT_DIR || type == BTRFS_FT_SYMLINK)
 				log_mode = LOG_INODE_ALL;
 			btrfs_release_path(path);
 			ret = btrfs_log_inode(trans, root, di_inode,
@@ -5278,11 +5255,16 @@ static int btrfs_log_all_parents(struct btrfs_trans_handle *trans,
 			if (IS_ERR(dir_inode))
 				continue;
 
+			if (ctx)
+				ctx->log_new_dentries = false;
 			ret = btrfs_log_inode(trans, root, dir_inode,
 					      LOG_INODE_ALL, 0, LLONG_MAX, ctx);
 			if (!ret &&
 			    btrfs_must_commit_transaction(trans, dir_inode))
 				ret = 1;
+			if (!ret && ctx && ctx->log_new_dentries)
+				ret = log_new_dir_dentries(trans, root,
+							   dir_inode, ctx);
 			iput(dir_inode);
 			if (ret)
 				goto out;
@@ -5652,11 +5634,9 @@ void btrfs_record_unlink_dir(struct btrfs_trans_handle *trans,
 	 * into the file.  When the file is logged we check it and
 	 * don't log the parents if the file is fully on disk.
 	 */
-	if (S_ISREG(inode->i_mode)) {
-		mutex_lock(&BTRFS_I(inode)->log_mutex);
-		BTRFS_I(inode)->last_unlink_trans = trans->transid;
-		mutex_unlock(&BTRFS_I(inode)->log_mutex);
-	}
+	mutex_lock(&BTRFS_I(inode)->log_mutex);
+	BTRFS_I(inode)->last_unlink_trans = trans->transid;
+	mutex_unlock(&BTRFS_I(inode)->log_mutex);
 
 	/*
 	 * if this directory was already logged any new