Re: [PATCH] cpufreq: Honour transition_latency over transition_delay_us

[Rafael J. Wysocki]

On Wed, Feb 28, 2024 at 12:35 AM Qais Yousef <qyousef@xxxxxxxxxxx> wrote:
>
> Some platforms like Arm's Juno can have a high transition latency that
> can be larger than the 2ms cap introduced. If a driver report
> a transition_latency that is higher than the cap, then use it as-is.
>
> Update comment s/10/2/ to reflect the new cap of 2ms.
>
> Reported-by: Pierre Gondois <pierre.gondois@xxxxxxx>
> Signed-off-by: Qais Yousef <qyousef@xxxxxxxxxxx>
> ---
>  drivers/cpufreq/cpufreq.c | 13 +++++++++++--
>  1 file changed, 11 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c
> index 66cef33c4ec7..926a51cb7e52 100644
> --- a/drivers/cpufreq/cpufreq.c
> +++ b/drivers/cpufreq/cpufreq.c
> @@ -576,8 +576,17 @@ unsigned int cpufreq_policy_transition_delay_us(struct cpufreq_policy *policy)
>
>         latency = policy->cpuinfo.transition_latency / NSEC_PER_USEC;
>         if (latency) {
> +               unsigned int max_delay_us = 2 * MSEC_PER_SEC;
> +
> +               /*
> +                * If the platform already has high transition_latency, use it
> +                * as-is.
> +                */
> +               if (latency > max_delay_us)
> +                       return latency;
> +
>                 /*
> -                * For platforms that can change the frequency very fast (< 10
> +                * For platforms that can change the frequency very fast (< 2
>                  * us), the above formula gives a decent transition delay But
>                  * for platforms where transition_latency is in milliseconds, it
>                  * ends up giving unrealistic values.
> @@ -586,7 +595,7 @@ unsigned int cpufreq_policy_transition_delay_us(struct cpufreq_policy *policy)
>                  * a reasonable amount of time after which we should reevaluate
>                  * the frequency.
>                  */
> -               return min(latency * LATENCY_MULTIPLIER, (unsigned int)(2 * MSEC_PER_SEC));
> +               return min(latency * LATENCY_MULTIPLIER, max_delay_us);
>         }
>
>         return LATENCY_MULTIPLIER;
> --

Applied as 6.9 material, thanks!