xfs: do not log/recover swapext extent owner changes for deleted inodes
authorEric Sandeen <sandeen@sandeen.net>
Thu, 29 Mar 2018 00:48:08 +0000 (17:48 -0700)
committerDarrick J. Wong <darrick.wong@oracle.com>
Thu, 29 Mar 2018 17:19:15 +0000 (10:19 -0700)
Today if we run xfs_fsr and crash[1], log replay can fail because
the recovery code tries to instantiate the donor inode from
disk to replay the swapext, but it's been deleted and we get
verifier failures when we try to read the inode off disk with
i_mode == 0.

This fixes both sides: We don't log the swapext change if the
inode has been deleted, and we don't try to recover it either.

[1] or if systemd doesn't cleanly unmount root, as it is wont
    to do ...

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
fs/xfs/xfs_inode.c
fs/xfs/xfs_log_recover.c

index 09ba970..3e3aab3 100644 (file)
@@ -2475,6 +2475,10 @@ xfs_ifree(
        ip->i_d.di_forkoff = 0;         /* mark the attr fork not in use */
        ip->i_d.di_format = XFS_DINODE_FMT_EXTENTS;
        ip->i_d.di_aformat = XFS_DINODE_FMT_EXTENTS;
+
+       /* Don't attempt to replay owner changes for a deleted inode */
+       ip->i_itemp->ili_fields &= ~(XFS_ILOG_AOWNER|XFS_ILOG_DOWNER);
+
        /*
         * Bump the generation count so no one will be confused
         * by reincarnations of this inode.
index 59134f6..2b2383f 100644 (file)
@@ -3245,7 +3245,9 @@ xlog_recover_inode_pass2(
        }
 
 out_owner_change:
-       if (in_f->ilf_fields & (XFS_ILOG_DOWNER|XFS_ILOG_AOWNER))
+       /* Recover the swapext owner change unless inode has been deleted */
+       if ((in_f->ilf_fields & (XFS_ILOG_DOWNER|XFS_ILOG_AOWNER)) &&
+           (dip->di_mode != 0))
                error = xfs_recover_inode_owner_change(mp, dip, in_f,
                                                       buffer_list);
        /* re-generate the checksum. */