Re: SKL BOOT FAILURE unless idle=nomwait (was Re: PROBLEM: Cpufreq constantly keeps frequency at maximum on 4.5-rc4)

[Rik van Riel <riel@redhat.com>]

[-- Attachment #1: Type: text/plain, Size: 1358 bytes --]

On Wed, 2016-03-16 at 01:32 +0100, Rafael J. Wysocki wrote:
> 
>  		 * We want to default to C1 (hlt), not to busy
> polling
> -		 * unless the timer is happening really really soon.
> +		 * unless the timer is happening really really
> soon.  Still, if
> +		 * the exit latency of C1 is too high, we need to
> poll anyway.
>  		 */
> -		if (interactivity_req > 20 &&
> +		if (data->next_timer_us > 20 &&
> +		    drv-
> >states[CPUIDLE_DRIVER_STATE_START].exit_latency <= 2 &&
>  		    !drv-

I wonder if replacing these two tests with
           data->next_timer_us
> states[CPUIDLE_DRIVER_STATE_START].exit_latency
would do the right thing?

That way any time we do not know for sure we are waking
up before the HLT exit latency has gone, we will poll.

Any time we suspect we will wake up sooner (data->predicted_us
is smaller than the HLT exit latency) we will still pick HLT,
anyway.

This makes the behaviour consistent between CPUs.

> >states[CPUIDLE_DRIVER_STATE_START].disabled &&
> -			dev-
> >states_usage[CPUIDLE_DRIVER_STATE_START].disable == 0)
> +		    !dev-
> >states_usage[CPUIDLE_DRIVER_STATE_START].disable)
>  			data->last_state_idx =
> CPUIDLE_DRIVER_STATE_START;
>  	} else {
>  		data->last_state_idx = CPUIDLE_DRIVER_STATE_START;
-- 
All Rights Reversed.


[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 473 bytes --]