Btrfs: Fix race in btrfs_mark_extent_written (c36047d7) · Commits · 戴 / test

When extent needs to be split, btrfs_mark_extent_written truncates the extent first, then inserts a new extent and increases the reference count. The race happens if someone else deletes the old extent before the new extent is inserted. The fix here is increase the reference count in advance. This race is similar to the race in btrfs_drop_extents that was recently fixed. Signed-off-by:

Yan Zheng <zheng.yan@oracle.com>

fs/btrfs/file.c

+14 −4

Original line number	Diff line number	Diff line
		@@ -746,6 +746,7 @@ int btrfs_mark_extent_written(struct btrfs_trans_handle *trans,
		u64 other_end;
		u64 split = start;
		u64 locked_end = end;
		u64 orig_parent;
		int extent_type;
		int split_end = 1;
		int ret;
		@@ -890,6 +891,12 @@ 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);
		BUG_ON(ret);
		btrfs_release_path(root, path);

		key.offset = start;
		@@ -910,10 +917,13 @@ again:
		btrfs_set_file_extent_encryption(leaf, fi, 0);
		btrfs_set_file_extent_other_encoding(leaf, fi, 0);

		ret = btrfs_inc_extent_ref(trans, root, bytenr, num_bytes,
		leaf->start, root->root_key.objectid,
		if (orig_parent != leaf->start) {
		ret = btrfs_update_extent_ref(trans, root, bytenr,
		orig_parent, leaf->start,
		root->root_key.objectid,
		trans->transid, inode->i_ino);
		BUG_ON(ret);
		}
		done:
		btrfs_mark_buffer_dirty(leaf);
		btrfs_release_path(root, path);

Admin message