GFS2: Fix unlock of fcntl locks during withdrawn state
authorSteven Whitehouse <swhiteho@redhat.com>
Thu, 14 Mar 2013 15:49:59 +0000 (15:49 +0000)
committerSteven Whitehouse <swhiteho@redhat.com>
Thu, 4 Apr 2013 08:53:46 +0000 (09:53 +0100)
When withdraw occurs, we need to continue to allow unlocks of fcntl
locks to occur, however these will only be local, since the node has
withdrawn from the cluster. This prevents triggering a VFS level
bug trap due to locks remaining when a file is closed.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
fs/gfs2/file.c

index 019f45e450971341f9cfba1c50a0116295b2dfd3..d79c2dadc536662370c3c084b100f1cecd303855 100644 (file)
@@ -923,8 +923,11 @@ static int gfs2_lock(struct file *file, int cmd, struct file_lock *fl)
                cmd = F_SETLK;
                fl->fl_type = F_UNLCK;
        }
-       if (unlikely(test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
+       if (unlikely(test_bit(SDF_SHUTDOWN, &sdp->sd_flags))) {
+               if (fl->fl_type == F_UNLCK)
+                       posix_lock_file_wait(file, fl);
                return -EIO;
+       }
        if (IS_GETLK(cmd))
                return dlm_posix_get(ls->ls_dlm, ip->i_no_addr, file, fl);
        else if (fl->fl_type == F_UNLCK)