Re: [PATCH 3/3] clockevents: Switch state to ONESHOT_STOPPED for unused clockevent devices

[Preeti U Murthy]

On 03/27/2015 10:44 PM, Viresh Kumar wrote:
> Clockevent device can now be stopped by switching to ONESHOT_STOPPED state, to
> avoid getting spurious interrupts on a tickless CPU.
> 
> Switch state to ONESHOT_STOPPED at following places:
> 
> 1.) NOHZ_MODE_LOWRES Mode
> 
> Timers & hrtimers are dependent on tick for their working in this mode and the
> only place from where clockevent device is programmed is the tick-code.
> 
> In LOWRES mode we skip reprogramming the clockevent device in
> tick_nohz_stop_sched_tick() if expires == KTIME_MAX. In addition to that we must
> also switch the clockevent device to ONESHOT_STOPPED state to avoid all spurious
> interrupts that may follow.
> 
> 2.) NOHZ_MODE_HIGHRES Mode
> 
> Tick & timers are dependent on hrtimers for their working in this mode and the
> only place from where clockevent device is programmed is the hrtimer-code.
> 
> There are two places here from which we reprogram the clockevent device or skip
> reprogramming it on expires == KTIME_MAX.
> 
> Signed-off-by: Viresh Kumar <viresh.kumar@xxxxxxxxxx>
> ---
>  kernel/time/hrtimer.c    | 46 ++++++++++++++++++++++++++++++++++++++++++----
>  kernel/time/tick-sched.c |  3 +++
>  2 files changed, 45 insertions(+), 4 deletions(-)
> 
> diff --git a/kernel/time/hrtimer.c b/kernel/time/hrtimer.c
> index 045ba7e2be6c..89d4b593dfc0 100644
> --- a/kernel/time/hrtimer.c
> +++ b/kernel/time/hrtimer.c
> @@ -543,8 +543,19 @@ hrtimer_force_reprogram(struct hrtimer_cpu_base *cpu_base, int skip_equal)
>  	if (cpu_base->hang_detected)
>  		return;
> 
> -	if (cpu_base->expires_next.tv64 != KTIME_MAX)
> +	if (cpu_base->expires_next.tv64 != KTIME_MAX) {
>  		tick_program_event(cpu_base->expires_next, 1);
> +	} else {
> +		struct clock_event_device *dev =
> +			__this_cpu_read(tick_cpu_device.evtdev);
> +		/*
> +		 * Don't need clockevent device anymore, stop it.
> +		 *
> +		 * We reach here only for NOHZ_MODE_HIGHRES mode and we are
> +		 * guaranteed that no timers/hrtimers are enqueued on this cpu.
> +		 */
> +		clockevents_set_state(dev, CLOCK_EVT_STATE_ONESHOT_STOPPED);
> +	}
>  }
> 
>  /*
> @@ -1312,9 +1323,36 @@ void hrtimer_interrupt(struct clock_event_device *dev)
>  	cpu_base->in_hrtirq = 0;
>  	raw_spin_unlock(&cpu_base->lock);
> 
> -	/* Reprogramming necessary ? */
> -	if (expires_next.tv64 == KTIME_MAX ||
> -	    !tick_program_event(expires_next, 0)) {
> +	if (expires_next.tv64 == KTIME_MAX) {
> +		struct clock_event_device *dev =
> +			__this_cpu_read(tick_cpu_device.evtdev);
> +
> +		cpu_base->hang_detected = 0;
> +
> +		/*
> +		 * Don't need clockevent device anymore, stop it.
> +		 *
> +		 * We reach here only for NOHZ_MODE_HIGHRES mode and we are
> +		 * guaranteed that no timers/hrtimers are enqueued on this cpu.
> +		 *
> +		 * Most of the scenarios will be covered by similar code
> +		 * present in hrtimer_force_reprogram(), as we always try to
> +		 * evaluate tick requirement on idle/irq exit and cancel
> +		 * tick-sched hrtimer when tick isn't required anymore.
> +		 *
> +		 * It is required here as well as a special case.
> +		 *
> +		 * Last hrtimer fires on a tickless CPU and doesn't rearm
> +		 * itself. tick_nohz_irq_exit() reevaluates next event and it
> +		 * gets expires == KTIME_MAX. Because tick was already stopped,
> +		 * and last expires == new_expires, we return early. And the
> +		 * clockevent device is never stopped.
> +		 */
> +		clockevents_set_state(dev, CLOCK_EVT_STATE_ONESHOT_STOPPED);
> +		return;
> +	}
> +
> +	if (!tick_program_event(expires_next, 0)) {
>  		cpu_base->hang_detected = 0;
>  		return;
>  	}
> diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
> index 47c04edd07df..ff271a26fa4d 100644
> --- a/kernel/time/tick-sched.c
> +++ b/kernel/time/tick-sched.c
> @@ -685,6 +685,9 @@ static ktime_t tick_nohz_stop_sched_tick(struct tick_sched *ts,
>  		 if (unlikely(expires.tv64 == KTIME_MAX)) {
>  			if (ts->nohz_mode == NOHZ_MODE_HIGHRES)
>  				hrtimer_cancel(&ts->sched_timer);
> +			else
> +				/* stop clock event device */
> +				clockevents_set_state(dev, CLOCK_EVT_STATE_ONESHOT_STOPPED);
>  			goto out;
>  		}
> 
Reviewed-by: Preeti U. Murthy <preeti@xxxxxxxxxxxxxxxxxx>

--
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/