Messages in this thread | | | Date | Mon, 14 Nov 2016 16:11:05 +0100 | From | Peter Zijlstra <> | Subject | Re: [PATCH 3/3] thermal/powerclamp: use PF_IDLE in injection kthread |
| |
On Wed, Nov 09, 2016 at 11:05:12AM -0800, Jacob Pan wrote: > +static enum hrtimer_restart idle_inject_timer_fn(struct hrtimer *hrtimer) > +{ > + set_tsk_need_resched(current);
So drivers really should not use this, which is why I had that in play_idle().
> + return HRTIMER_NORESTART; > +} > + > static void clamp_idle_injection_func(struct kthread_work *work) > { > struct powerclamp_worker_data *w_data; > - unsigned long target_jiffies; > + unsigned long end_time; > + unsigned int duration_ms; > > w_data = container_of(work, struct powerclamp_worker_data, > idle_injection_work.work); > > + hrtimer_init(&w_data->timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL_PINNED); > + w_data->timer.function = idle_inject_timer_fn; > + > /* > * only elected controlling cpu can collect stats and update > * control parameters. > @@ -453,31 +459,17 @@ static void clamp_idle_injection_func(struct kthread_work *work) > if (should_skip) > goto balance; > > + end_time = jiffies + w_data->duration_jiffies; > + duration_ms = jiffies_to_msecs(w_data->duration_jiffies); > + hrtimer_start(&w_data->timer, ms_to_ktime(duration_ms), > + HRTIMER_MODE_REL_PINNED); > + > + cpuidle_use_deepest_state(true); > + while (time_after(end_time, jiffies)) > + play_idle();
Why that loop?
> + cpuidle_use_deepest_state(false);
> + > + hrtimer_cancel(&w_data->timer);
| |