Re: [RFCv6 PATCH 07/10] sched/fair: jump to max OPP when crossing UP threshold

[Steve Muckle]

Hi Juri,

On 12/11/2015 03:12 AM, Juri Lelli wrote:
>> @@ -2895,6 +2934,8 @@ void scheduler_tick(void)
>> >  	trigger_load_balance(rq);
>> >  #endif
>> >  	rq_last_tick_reset(rq);
>> > +
>> > +	sched_freq_tick(cpu);
> We are not holding rq->lock anymore at this points, and this collides
> with comment in update_cpu_capacity_request(). Can't you just move this
> up before raw_spin_unlock(&rq->lock)?

My thinking in putting it last was to have it after the possible
periodic load balance, so that we don't initiate a frequency change only
to have to modify the frequency again immediately afterwards.

Thinking more about it, the way we currently have the policy defined
there's no concern with having it earlier since sched_freq_tick only
causes the frequency to go to fmax (or do nothing). If we modify the
policy so that sched_freq_tick can cause arbitrary frequency changes
then I think this may need more thought.

thanks,
Steve
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/