xfs: relax is_reflink_inode assert in xfs_reflink_find_cow_mapping

We don't hold the ilock through the entire sequence of xfs_writepage_map
-> xfs_map_cow -> xfs_reflink_find_cow_mapping.  This means that we can
race with another thread that is trying to clear the inode reflink flag,
with the result that the flag is set for the xfs_map_cow check but
cleared before we get to the assert in find_cow_mapping.  When this
happens, we blow the assert even though everything is fine.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>

diff --git a/fs/xfs/xfs_reflink.c b/fs/xfs/xfs_reflink.c
index e13f5ad57a03eb9727059b18341e5b8472a24265..99c5852f9fe785b3bb2522f770f0345ec8d635e6 100644 (file)
--- a/fs/xfs/xfs_reflink.c
+++ b/fs/xfs/xfs_reflink.c
@@ -490,8 +490,9 @@ xfs_reflink_find_cow_mapping(
        struct xfs_iext_cursor          icur;
 
        ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL | XFS_ILOCK_SHARED));
-       ASSERT(xfs_is_reflink_inode(ip));
 
+       if (!xfs_is_reflink_inode(ip))
+               return false;
        offset_fsb = XFS_B_TO_FSBT(ip->i_mount, offset);
        if (!xfs_iext_lookup_extent(ip, ifp, offset_fsb, &icur, &got))
                return false;