diff mbox

[04/11] block: Protect queue flag changes with the queue lock

Message ID 20180228192823.5191-5-bart.vanassche@wdc.com (mailing list archive)
State New, archived
Headers show

Commit Message

Bart Van Assche Feb. 28, 2018, 7:28 p.m. UTC
Since the queue flags may be changed concurrently from multiple
contexts after a queue becomes visible in sysfs, make these changes
safe by protecting these with the queue lock.

Signed-off-by: Bart Van Assche <bart.vanassche@wdc.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Hannes Reinecke <hare@suse.de>
Cc: Johannes Thumshirn <jthumshirn@suse.de>
Cc: Ming Lei <ming.lei@redhat.com>
---
 block/blk-mq.c   | 3 +--
 block/blk-stat.c | 6 +++---
 2 files changed, 4 insertions(+), 5 deletions(-)

Comments

Johannes Thumshirn March 1, 2018, 8:51 a.m. UTC | #1
On Wed, Feb 28, 2018 at 11:28:16AM -0800, Bart Van Assche wrote:
>  static bool blk_poll_stats_enable(struct request_queue *q)
>  {
> -	if (test_bit(QUEUE_FLAG_POLL_STATS, &q->queue_flags) ||
> -	    test_and_set_bit(QUEUE_FLAG_POLL_STATS, &q->queue_flags))
> +	if (blk_queue_flag_test_and_set(QUEUE_FLAG_POLL_STATS, q))

Is this one really needed or just for symmetry? Even if something
would change the queue_flags after the first test_bit() call, the
test_and_set_bit() would still do the right thing, wouldn't it?
Bart Van Assche March 1, 2018, 3:19 p.m. UTC | #2
On Thu, 2018-03-01 at 09:51 +0100, Johannes Thumshirn wrote:
> On Wed, Feb 28, 2018 at 11:28:16AM -0800, Bart Van Assche wrote:

> >  static bool blk_poll_stats_enable(struct request_queue *q)

> >  {

> > -	if (test_bit(QUEUE_FLAG_POLL_STATS, &q->queue_flags) ||

> > -	    test_and_set_bit(QUEUE_FLAG_POLL_STATS, &q->queue_flags))

> > +	if (blk_queue_flag_test_and_set(QUEUE_FLAG_POLL_STATS, q))

> 

> Is this one really needed or just for symmetry? Even if something

> would change the queue_flags after the first test_bit() call, the

> test_and_set_bit() would still do the right thing, wouldn't it?


Hello Johannes,

Since blk_poll_stats_enable() is called from the hot path (polling code) I
think we need the optimization of calling test_bit() before calling
test_and_set_bit(). I will restore the test_bit() call.

Bart.
Johannes Thumshirn March 1, 2018, 3:22 p.m. UTC | #3
On Thu, Mar 01, 2018 at 03:19:08PM +0000, Bart Van Assche wrote:
> Hello Johannes,
> 
> Since blk_poll_stats_enable() is called from the hot path (polling code) I
> think we need the optimization of calling test_bit() before calling
> test_and_set_bit(). I will restore the test_bit() call.

Thanks,
	Johannes
Martin K. Petersen March 2, 2018, 11:12 p.m. UTC | #4
Bart,

> Since the queue flags may be changed concurrently from multiple
> contexts after a queue becomes visible in sysfs, make these changes
> safe by protecting these with the queue lock.

Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
diff mbox

Patch

diff --git a/block/blk-mq.c b/block/blk-mq.c
index 67026494119b..899f357962e8 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -3022,8 +3022,7 @@  EXPORT_SYMBOL_GPL(blk_mq_update_nr_hw_queues);
 /* Enable polling stats and return whether they were already enabled. */
 static bool blk_poll_stats_enable(struct request_queue *q)
 {
-	if (test_bit(QUEUE_FLAG_POLL_STATS, &q->queue_flags) ||
-	    test_and_set_bit(QUEUE_FLAG_POLL_STATS, &q->queue_flags))
+	if (blk_queue_flag_test_and_set(QUEUE_FLAG_POLL_STATS, q))
 		return true;
 	blk_stat_add_callback(q, q->poll_cb);
 	return false;
diff --git a/block/blk-stat.c b/block/blk-stat.c
index b664aa6df725..bd365a95fcf8 100644
--- a/block/blk-stat.c
+++ b/block/blk-stat.c
@@ -152,7 +152,7 @@  void blk_stat_add_callback(struct request_queue *q,
 
 	spin_lock(&q->stats->lock);
 	list_add_tail_rcu(&cb->list, &q->stats->callbacks);
-	queue_flag_set(QUEUE_FLAG_STATS, q);
+	blk_queue_flag_set(QUEUE_FLAG_STATS, q);
 	spin_unlock(&q->stats->lock);
 }
 EXPORT_SYMBOL_GPL(blk_stat_add_callback);
@@ -163,7 +163,7 @@  void blk_stat_remove_callback(struct request_queue *q,
 	spin_lock(&q->stats->lock);
 	list_del_rcu(&cb->list);
 	if (list_empty(&q->stats->callbacks) && !q->stats->enable_accounting)
-		queue_flag_clear(QUEUE_FLAG_STATS, q);
+		blk_queue_flag_clear(QUEUE_FLAG_STATS, q);
 	spin_unlock(&q->stats->lock);
 
 	del_timer_sync(&cb->timer);
@@ -191,7 +191,7 @@  void blk_stat_enable_accounting(struct request_queue *q)
 {
 	spin_lock(&q->stats->lock);
 	q->stats->enable_accounting = true;
-	queue_flag_set(QUEUE_FLAG_STATS, q);
+	blk_queue_flag_set(QUEUE_FLAG_STATS, q);
 	spin_unlock(&q->stats->lock);
 }