On Wed, Mar 24, 2021 at 03:21:24PM +0100, Christoph Hellwig wrote:

The i_cowextsize field is only used for v3 inodes, and the i_flushiter
field is only used for v1/v2 inodes.  Use a union to pack the inode a
littler better after adding a few missing guards around their usage.

Signed-off-by: Christoph Hellwig <hch@lst.de>

Hmm, so this patch caused a regression on V4 filesystems xfs/051.  It
looks like the flush iter gets set to zero and then log recovery forgets
to replay the inode(?)

The following patch fixes it for me, FWIW...

--D

xfs: fix regression in xfs/051 with this patch

Signed-off-by: Darrick J. Wong <djwong@kernel.org>

diff --git a/fs/xfs/xfs_ioctl.c b/fs/xfs/xfs_ioctl.c
index 205ee7da8fa5..795c23e5c655 100644
--- a/fs/xfs/xfs_ioctl.c
+++ b/fs/xfs/xfs_ioctl.c

@@ -1525,11 +1525,12 @@ xfs_ioctl_setattr(
 		ip->i_extsize = fa->fsx_extsize >> mp->m_sb.sb_blocklog;
 	else
 		ip->i_extsize = 0;
-	if (xfs_sb_version_has_v3inode(&mp->m_sb) &&
-	    (ip->i_d.di_flags2 & XFS_DIFLAG2_COWEXTSIZE))
-		ip->i_cowextsize = fa->fsx_cowextsize >> mp->m_sb.sb_blocklog;
-	else
-		ip->i_cowextsize = 0;
+	if (xfs_sb_version_has_v3inode(&mp->m_sb)) {
+		if (ip->i_d.di_flags2 & XFS_DIFLAG2_COWEXTSIZE)
+			ip->i_cowextsize = XFS_B_TO_FSBT(mp, fa->fsx_cowextsize);
+		else
+			ip->i_cowextsize = 0;
+	}
 
 	error = xfs_trans_commit(tp);