Re: [PATCH V3 2/6] sched: idle: cpuidle: Check the latency req before idle

[Peter Zijlstra <peterz@infradead.org>]

On Fri, Nov 07, 2014 at 03:31:23PM +0100, Daniel Lezcano wrote:
> @@ -216,19 +219,26 @@ static void cpu_idle_loop(void)
>  			local_irq_disable();
>  			arch_cpu_idle_enter();
>  
> +			latency_req = pm_qos_request(PM_QOS_CPU_DMA_LATENCY);
> +
>  			/*
>  			 * In poll mode we reenable interrupts and spin.
>  			 *
> +			 * If the latency req is zero, we don't want to
> +			 * enter any idle state and we jump to the poll
> +			 * function directly
> +			 *
>  			 * Also if we detected in the wakeup from idle
>  			 * path that the tick broadcast device expired
>  			 * for us, we don't want to go deep idle as we
>  			 * know that the IPI is going to arrive right
>  			 * away
>  			 */
> -			if (cpu_idle_force_poll || tick_check_broadcast_expired())
> +			if (!latency_req || cpu_idle_force_poll ||
> +			    tick_check_broadcast_expired())
>  				cpu_idle_poll();

Is this why you wanted that weak poll function?

Should we not instead allow an arch to deal with !latency_req and only
fall back to this polling if there is no actual way for it to implement
this better?