Prev: perf-probe updates - data-structure support improvements, etc.
Next: [PATCH 2/2] ipc semaphores: order wakeups based on waiter CPU
From: Divyesh Shah on 12 Apr 2010 14:40
On Mon, Apr 12, 2010 at 7:30 AM, Vivek Goyal <vgoyal(a)redhat.com> wrote:
> On Fri, Apr 09, 2010 at 07:35:35PM -0700, Divyesh Shah wrote:
>> review comments.
>>
>> Changelog: (in response to Vivek Goyal's comments)
>> o group blkiocg_update_blkio_group_dequeue_stats() with other DEBUG functions
>> o rename blkiocg_update_set_active_queue_stats() to
>> � blkiocg_update_avg_queue_size_stats()
>> o s/request/io/ in blkiocg_update_request_add_stats() and
>> � blkiocg_update_request_remove_stats()
>> o Call cfq_del_timer() at request dispatch() instead of
>> � blkiocg_update_idle_time_stats()
>>
>> Signed-off-by: Divyesh Shah<dpshah(a)google.com>
>> ---
>>
>
> [..]
>> @@ -2395,11 +2395,7 @@ static int cfq_dispatch_requests(struct request_queue *q, int force)
>> � � � }
>>
>> � � � cfq_log_cfqq(cfqd, cfqq, "dispatched a request");
>> - � � /*
>> - � � �* This is needed since we don't exactly match the mod_timer() and
>> - � � �* del_timer() calls in CFQ.
>> - � � �*/
>> - � � blkiocg_update_idle_time_stats(&cfqq->cfqg->blkg);
>> + � � cfq_del_timer(cfqd, cfqq);
>
> I am trying to think that why do we need this call here. What is that path
> which does not delete the timer upon receiving request and lets fix that.
>
> One of the possible path when you got the request but still leave the
> timer on is following.
>
> cfq_rq_enqueued()
> � � � �if(request_not_big_enough)
> � � � � � � � �leave_timer_on
> � � � � � � � �cfq_mark_cfqq_must_dispatch(cfqq);
>
> I guess we need to leave timer on so that if an unplug does not come in, a
> timer expirty will dispatch the request from the queue.
>
> In this case we can probably call here blkiocg_update_idle_time_stats()
> directly. Anyway, that's the right thing to do otherwise our idle time
> stats are little wrong as we got a request from the application and idle
> timer is over. Now this is additional wait time enforced by cfq so that
> lots of small requests are not dispatched to disk.

Good point. Will send an updated version of this patch.

>
> Thanks
> Vivek
>
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
 | 
Pages: 1
Prev: perf-probe updates - data-structure support improvements, etc.
Next: [PATCH 2/2] ipc semaphores: order wakeups based on waiter CPU