ocfs2: fix slot overwritten if storage link down during mount
authorjiangyiwen <jiangyiwen@huawei.com>
Thu, 14 Jan 2016 23:17:33 +0000 (15:17 -0800)
committerLinus Torvalds <torvalds@linux-foundation.org>
Fri, 15 Jan 2016 00:00:49 +0000 (16:00 -0800)
The following case will lead to slot overwritten.

N1                               N2
mount ocfs2 volume, find and
allocate slot 0, then set
osb->slot_num to 0, begin to
write slot info to disk
                                 mount ocfs2 volume, wait for super lock
write block fail because of
storage link down, unlock
super lock
                                 got super lock and also allocate slot 0
                                 then unlock super lock

mount fail and then dismount,
since osb->slot_num is 0, try to
put invalid slot to disk. And it
will succeed if storage link
restores.
                                 N2 slot info is now overwritten

Once another node say N3 mount, it will find and allocate slot 0 again,
which will lead to mount hung because journal has already been locked by
N2.  so when write slot info failed, invalidate slot in advance to avoid
overwrite slot.

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Yiwen Jiang <jiangyiwen@huawei.com>
Reviewed-by: Joseph Qi <joseph.qi@huawei.com>
Cc: Mark Fasheh <mfasheh@suse.de>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
fs/ocfs2/slot_map.c

index 47bba1be2a51f6d89ee95fdd91213d40f4c9192e..1e09592148ad2823c8b073ebe9c03b66c99e3366 100644 (file)
@@ -502,8 +502,17 @@ int ocfs2_find_slot(struct ocfs2_super *osb)
        trace_ocfs2_find_slot(osb->slot_num);
 
        status = ocfs2_update_disk_slot(osb, si, osb->slot_num);
-       if (status < 0)
+       if (status < 0) {
                mlog_errno(status);
+               /*
+                * if write block failed, invalidate slot to avoid overwrite
+                * slot during dismount in case another node rightly has mounted
+                */
+               spin_lock(&osb->osb_lock);
+               ocfs2_invalidate_slot(si, osb->slot_num);
+               osb->slot_num = OCFS2_INVALID_SLOT;
+               spin_unlock(&osb->osb_lock);
+       }
 
 bail:
        return status;