[v2,4/4] block: block new I/O just after queue is set as dying
diff mbox

Message ID 20170324123621.5227-5-tom.leiming@gmail.com
State New
Headers show

Commit Message

Ming Lei March 24, 2017, 12:36 p.m. UTC
Before commit 780db2071a(blk-mq: decouble blk-mq freezing
from generic bypassing), the dying flag is checked before
entering queue, and Tejun converts the checking into .mq_freeze_depth,
and assumes the counter is increased just after dying flag
is set. Unfortunately we doesn't do that in blk_set_queue_dying().

This patch calls blk_freeze_queue_start() in blk_set_queue_dying(),
so that we can block new I/O coming once the queue is set as dying.

Given blk_set_queue_dying() is always called in remove path
of block device, and queue will be cleaned up later, we don't
need to worry about undoing the counter.

Cc: Bart Van Assche <bart.vanassche@sandisk.com>
Cc: Tejun Heo <tj@kernel.org>
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Ming Lei <tom.leiming@gmail.com>
---
 block/blk-core.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

Comments

Bart Van Assche March 24, 2017, 5:45 p.m. UTC | #1
On Fri, 2017-03-24 at 20:36 +0800, Ming Lei wrote: 
> +	/* block new I/O coming */
> +	blk_freeze_queue_start(q);

As I have already mentioned two times, the comment above
blk_freeze_queue_start() should be made more clear. It should mention that
without that call blk_queue_enter() won't check the "dying" flag after it
has been set. If that is not mentioned in a comment the next person who
reads the blk_set_queue_dying() function will wonder why the
blk_freeze_queue_start() call is really needed and whether it can be removed.

>  		/*
>  		 * read pair of barrier in blk_freeze_queue_start(),
>  		 * we need to order reading DEAD flag of .q_usage_counter
> -		 * and reading .mq_freeze_depth, otherwise the following
> -		 * wait may never return if the two read are reordered.
> +		 * and reading .mq_freeze_depth or dying flag, otherwise
> +		 * the following wait may never return if the two read
> +		 * are reordered.
>  		 */
>  		smp_rmb();

Please fix the spelling in the above comment ("two read").

Thanks,

Bart.

Patch
diff mbox

diff --git a/block/blk-core.c b/block/blk-core.c
index 5901133d105f..f0dd9b0054ed 100644
--- a/block/blk-core.c
+++ b/block/blk-core.c
@@ -500,6 +500,9 @@  void blk_set_queue_dying(struct request_queue *q)
 	queue_flag_set(QUEUE_FLAG_DYING, q);
 	spin_unlock_irq(q->queue_lock);
 
+	/* block new I/O coming */
+	blk_freeze_queue_start(q);
+
 	if (q->mq_ops)
 		blk_mq_wake_waiters(q);
 	else {
@@ -672,8 +675,9 @@  int blk_queue_enter(struct request_queue *q, bool nowait)
 		/*
 		 * read pair of barrier in blk_freeze_queue_start(),
 		 * we need to order reading DEAD flag of .q_usage_counter
-		 * and reading .mq_freeze_depth, otherwise the following
-		 * wait may never return if the two read are reordered.
+		 * and reading .mq_freeze_depth or dying flag, otherwise
+		 * the following wait may never return if the two read
+		 * are reordered.
 		 */
 		smp_rmb();