2009-10-26 09:16:22

by Shaohua Li

Subject: [RFC]cfq-iosched: improve hw_tag detection

If active queue hasn't enough requests and idle window opens, cfq will not
dispatch sufficient requests to hardware. In such situation, current code
will zero hw_tag. But this is because cfq doesn't dispatch enough requests
instead of hardware queue doesn't work. Don't zero hw_tag in such case.

Signed-off-by: Shaohua Li <[email protected]>
diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c
index 069a610..0467049 100644
--- a/block/cfq-iosched.c
+++ b/block/cfq-iosched.c
@@ -2166,6 +2166,8 @@ static void cfq_insert_request(struct request_queue *q, struct request *rq)
*/
static void cfq_update_hw_tag(struct cfq_data *cfqd)
{
+ struct cfq_queue *cfqq = cfqd->active_queue;
+
if (rq_in_driver(cfqd) > cfqd->rq_in_driver_peak)
cfqd->rq_in_driver_peak = rq_in_driver(cfqd);

@@ -2173,6 +2175,16 @@ static void cfq_update_hw_tag(struct cfq_data *cfqd)
rq_in_driver(cfqd) <= CFQ_HW_QUEUE_MIN)
return;

+ /*
+ * If active queue hasn't enough requests and can idle, cfq might not
+ * dispatch sufficient requests to hardware. Don't zero hw_tag in this
+ * case
+ */
+ if (cfqq && cfq_cfqq_idle_window(cfqq) &&
+ cfqq->dispatched + cfqq->queued[0] + cfqq->queued[1] <
+ CFQ_HW_QUEUE_MIN && rq_in_driver(cfqd) < CFQ_HW_QUEUE_MIN)
+ return;
+
if (cfqd->hw_tag_samples++ < 50)
return;

2009-10-26 11:35:52

by Jens Axboe

Subject: Re: [RFC]cfq-iosched: improve hw_tag detection

On Mon, Oct 26 2009, Shaohua Li wrote:
> If active queue hasn't enough requests and idle window opens, cfq will not
> dispatch sufficient requests to hardware. In such situation, current code
> will zero hw_tag. But this is because cfq doesn't dispatch enough requests
> instead of hardware queue doesn't work. Don't zero hw_tag in such case.

Thanks, that looks like a good fix. Curious, is this fixing some
regression at your end?

--
Jens Axboe

2009-10-27 00:50:37

by Shaohua Li

Subject: Re: [RFC]cfq-iosched: improve hw_tag detection

On Mon, Oct 26, 2009 at 07:35:55PM +0800, Jens Axboe wrote:
> On Mon, Oct 26 2009, Shaohua Li wrote:
> > If active queue hasn't enough requests and idle window opens, cfq will not
> > dispatch sufficient requests to hardware. In such situation, current code
> > will zero hw_tag. But this is because cfq doesn't dispatch enough requests
> > instead of hardware queue doesn't work. Don't zero hw_tag in such case.
>
> Thanks, that looks like a good fix. Curious, is this fixing some
> regression at your end?
No, just observed such issue when I play with cfq.

Thanks,
Shaohua

2009-10-27 07:45:40

by Jens Axboe

Subject: Re: [RFC]cfq-iosched: improve hw_tag detection

On Tue, Oct 27 2009, Shaohua Li wrote:
> On Mon, Oct 26, 2009 at 07:35:55PM +0800, Jens Axboe wrote:
> > On Mon, Oct 26 2009, Shaohua Li wrote:
> > > If active queue hasn't enough requests and idle window opens, cfq will not
> > > dispatch sufficient requests to hardware. In such situation, current code
> > > will zero hw_tag. But this is because cfq doesn't dispatch enough requests
> > > instead of hardware queue doesn't work. Don't zero hw_tag in such case.
> >
> > Thanks, that looks like a good fix. Curious, is this fixing some
> > regression at your end?
> No, just observed such issue when I play with cfq.

OK thanks, just curious. I will apply it for 2.6.33.

--
Jens Axboe