[dm-devel] dm-mq and end_clone_request()

[Mike Snitzer]

On Mon, Jul 25 2016 at  6:00P -0400,
Bart Van Assche <bart.vanassche at sandisk.com> wrote:

> On 07/25/2016 02:23 PM, Mike Snitzer wrote:
> >So I'd be curious to know if your debugging has enabled you to identify
> >exactly where in the dm-mapth.c code the -EIO return is being
> >established.  do_end_io() is the likely candidate -- but again the
> >__must_push_back() check should prevent it and DM_ENDIO_REQUEUE should
> >be returned.
> 
> Hello Mike,
> 
> Thanks for looking further into this. The pr_info() statement that I had
> added in the following code block in __multipath_map() fired what told me
> that the following code block triggered the -EIO return:
> 
> 	if (!pgpath) {
> 		if (!must_push_back(m))
> 			r = -EIO;	/* Failed */
> 		pr_info("%s(): (a) returning %d\n", __func__, r);
> 		return r;
> 	}
> 
> From the system log:
> 
> kernel: mpath 254:0: queue_if_no_path 1 -> 0
> kernel: __multipath_map(): (a) returning -5
> 
> The code that I had added in queue_if_no_path() is as follows:
> 
> 	old = test_bit(MPATHF_QUEUE_IF_NO_PATH, &m->flags);
>         [ ... ]
> 	pr_info("mpath %s: queue_if_no_path %d -> %d\n",
> 		dm_device_name(dm_table_get_md(m->ti->table)), old,
> 		queue_if_no_path);

Hi Bart,

Please try this patch to see if it fixes your issue, thanks.

diff --git a/drivers/md/dm-mpath.c b/drivers/md/dm-mpath.c
index 52baf8a..287caa7 100644
--- a/drivers/md/dm-mpath.c
+++ b/drivers/md/dm-mpath.c
@@ -433,10 +433,17 @@ failed:
  */
 static int must_push_back(struct multipath *m)
 {
-	return (test_bit(MPATHF_QUEUE_IF_NO_PATH, &m->flags) ||
-		((test_bit(MPATHF_QUEUE_IF_NO_PATH, &m->flags) !=
-		  test_bit(MPATHF_SAVED_QUEUE_IF_NO_PATH, &m->flags)) &&
-		 dm_noflush_suspending(m->ti)));
+	bool r;
+	unsigned long flags;
+
+	spin_lock_irqsave(&m->lock, flags);
+	r = (test_bit(MPATHF_QUEUE_IF_NO_PATH, &m->flags) ||
+	     ((test_bit(MPATHF_QUEUE_IF_NO_PATH, &m->flags) !=
+	       test_bit(MPATHF_SAVED_QUEUE_IF_NO_PATH, &m->flags)) &&
+	      dm_noflush_suspending(m->ti)));
+	spin_unlock_irqrestore(&m->lock, flags);
+
+	return r;
 }
 
 /*