Btrfs: fill_holes: Fix slot number passed to hole_mergeable() call.

For a non-existent key, btrfs_search_slot() sets path->slots[0] to the slot
where the key could have been present, which in this case would be the slot
containing the extent item which would be the next neighbor of the file range
being punched. The current code passes an incremented path->slots[0] and we
skip to the wrong file extent item. This would mean that we would fail to
merge the "yet to be created" hole with the next neighboring hole (if one
exists). Fix this.

Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com>
Reviewed-by: Wang Shilong <wangsl.fnst@cn.fujitsu.com>
Signed-off-by: Chris Mason <clm@fb.com>

diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index d3afac292d677e0ed492255ca7cb3890cde717b1..77e33534c7d6d05a94cd77eabd113e3a733466e8 100644 (file)
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -2088,10 +2088,9 @@ static int fill_holes(struct btrfs_trans_handle *trans, struct inode *inode,
                goto out;
        }
 
-       if (hole_mergeable(inode, leaf, path->slots[0]+1, offset, end)) {
+       if (hole_mergeable(inode, leaf, path->slots[0], offset, end)) {
                u64 num_bytes;
 
-               path->slots[0]++;
                key.offset = offset;
                btrfs_set_item_key_safe(root, path, &key);
                fi = btrfs_item_ptr(leaf, path->slots[0],