While shutting down loop controller, we first quiesce the admin/IO queue,
delete the admin/IO tag-set and then at last destroy the admin/IO queue.
However it's quite possible that during the window between quiescing and
destroying of the admin/IO queue, some admin/IO request might sneak in
and if that happens then we could potentially encounter a hung task
because shutdown operation can't forward progress until any pending I/O
is flushed off.
This commit helps ensure that before destroying the admin/IO queue, we
unquiesce the admin/IO queue so that any outstanding requests, which are
added after the admin/IO queue is quiesced, are now flushed to its
completion.
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Nilay Shroff <nilay@linux.ibm.com>
Signed-off-by: Keith Busch <kbusch@kernel.org>
 {
        if (!test_and_clear_bit(NVME_LOOP_Q_LIVE, &ctrl->queues[0].flags))
                return;
+       /*
+        * It's possible that some requests might have been added
+        * after admin queue is stopped/quiesced. So now start the
+        * queue to flush these requests to the completion.
+        */
+       nvme_unquiesce_admin_queue(&ctrl->ctrl);
+
        nvmet_sq_destroy(&ctrl->queues[0].nvme_sq);
        nvme_remove_admin_tag_set(&ctrl->ctrl);
 }
                nvmet_sq_destroy(&ctrl->queues[i].nvme_sq);
        }
        ctrl->ctrl.queue_count = 1;
+       /*
+        * It's possible that some requests might have been added
+        * after io queue is stopped/quiesced. So now start the
+        * queue to flush these requests to the completion.
+        */
+       nvme_unquiesce_io_queues(&ctrl->ctrl);
 }
 
 static int nvme_loop_init_io_queues(struct nvme_loop_ctrl *ctrl)