xfs: check kthread_should_stop() after the setting of task state

A umount hang is possible when a race occurs between the umount
process and the xfsaild kthread. The following sequences outline
the race:

    xfsaild: kthread_should_stop()
	     => return false, so xfsaild continue

    umount: set_bit(KTHREAD_SHOULD_STOP, &kthread->flags)
	    => by kthread_stop()
    umount: wake_up_process()
	    => because xfsaild is still running, so 0 is returned

    xfsaild: __set_current_state(TASK_INTERRUPTIBLE)
    xfsaild: schedule()
	    => now, xfsaild will wait indefinitely

    umount: wait_for_completion()
	    => and umount will hang

To fix that, we need to check kthread_should_stop() after we set
the task state, so the xfsaild will either see the stop bit and
exit or the task state is reset to runnable by wake_up_process()
such that it isn't scheduled out indefinitely and detects the stop
bit at the next iteration.

Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Hou Tao <houtao1@huawei.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>

diff --git a/fs/xfs/xfs_trans_ail.c b/fs/xfs/xfs_trans_ail.c
index 354368a906e593c9e1e3a268c38919bbb5a20ab7..4b1669f9d2b21634e11b8ce756248fef209e0000 100644 (file)
--- a/fs/xfs/xfs_trans_ail.c
+++ b/fs/xfs/xfs_trans_ail.c
@@ -514,11 +514,26 @@ xfsaild(
        current->flags |= PF_MEMALLOC;
        set_freezable();
 
-       while (!kthread_should_stop()) {
+       while (1) {
                if (tout && tout <= 20)
-                       __set_current_state(TASK_KILLABLE);
+                       set_current_state(TASK_KILLABLE);
                else
-                       __set_current_state(TASK_INTERRUPTIBLE);
+                       set_current_state(TASK_INTERRUPTIBLE);
+
+               /*
+                * Check kthread_should_stop() after we set the task state
+                * to guarantee that we either see the stop bit and exit or
+                * the task state is reset to runnable such that it's not
+                * scheduled out indefinitely and detects the stop bit at
+                * next iteration.
+                *
+                * A memory barrier is included in above task state set to
+                * serialize again kthread_stop().
+                */
+               if (kthread_should_stop()) {
+                       __set_current_state(TASK_RUNNING);
+                       break;
+               }
 
                spin_lock(&ailp->xa_lock);