OSDN Git Service

xfs: fix iclog release error check race with shutdown
authorBrian Foster <bfoster@redhat.com>
Fri, 21 Feb 2020 15:36:24 +0000 (07:36 -0800)
committerDarrick J. Wong <darrick.wong@oracle.com>
Tue, 3 Mar 2020 04:55:50 +0000 (20:55 -0800)
Prior to commit df732b29c8 ("xfs: call xlog_state_release_iclog with
l_icloglock held"), xlog_state_release_iclog() always performed a
locked check of the iclog error state before proceeding into the
sync state processing code. As of this commit, part of
xlog_state_release_iclog() was open-coded into
xfs_log_release_iclog() and as a result the locked error state check
was lost.

The lockless check still exists, but this doesn't account for the
possibility of a race with a shutdown being performed by another
task causing the iclog state to change while the original task waits
on ->l_icloglock. This has reproduced very rarely via generic/475
and manifests as an assert failure in __xlog_state_release_iclog()
due to an unexpected iclog state.

Restore the locked error state check in xlog_state_release_iclog()
to ensure that an iclog state update via shutdown doesn't race with
the iclog release state processing code.

Fixes: df732b29c807 ("xfs: call xlog_state_release_iclog with l_icloglock held")
Reported-by: Zorro Lang <zlang@redhat.com>
Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
fs/xfs/xfs_log.c

index f6006d9..796ff37 100644 (file)
@@ -605,18 +605,23 @@ xfs_log_release_iclog(
        struct xlog             *log = mp->m_log;
        bool                    sync;
 
-       if (iclog->ic_state == XLOG_STATE_IOERROR) {
-               xfs_force_shutdown(mp, SHUTDOWN_LOG_IO_ERROR);
-               return -EIO;
-       }
+       if (iclog->ic_state == XLOG_STATE_IOERROR)
+               goto error;
 
        if (atomic_dec_and_lock(&iclog->ic_refcnt, &log->l_icloglock)) {
+               if (iclog->ic_state == XLOG_STATE_IOERROR) {
+                       spin_unlock(&log->l_icloglock);
+                       goto error;
+               }
                sync = __xlog_state_release_iclog(log, iclog);
                spin_unlock(&log->l_icloglock);
                if (sync)
                        xlog_sync(log, iclog);
        }
        return 0;
+error:
+       xfs_force_shutdown(mp, SHUTDOWN_LOG_IO_ERROR);
+       return -EIO;
 }
 
 /*