Re: [PATCHv2] cpufreq/schedutil: Only bind threads if needed

[Rafael J. Wysocki]

On Fri, Sep 27, 2024 at 10:59 AM Christian Loehle
<christian.loehle@xxxxxxx> wrote:
>
> Remove the unconditional binding of sugov kthreads to the affected CPUs
> if the cpufreq driver indicates that updates can happen from any CPU.
> This allows userspace to set affinities to either save power (waking up
> bigger CPUs on HMP can be expensive) or increasing performance (by
> letting the utilized CPUs run without preemption of the sugov kthread).
>
> Signed-off-by: Christian Loehle <christian.loehle@xxxxxxx>

Acked-by: Rafael J. Wysocki <rafael@xxxxxxxxxx>

and I'm assuming that this will go in via tip.

> ---
> - v2: Add comment for the dl_task_check_affinity return (Juri)
> v1: https://lore.kernel.org/lkml/480f2140-ea59-4e1d-a68d-18cbcec10941@xxxxxxx/
>
>  kernel/sched/cpufreq_schedutil.c | 6 +++++-
>  kernel/sched/syscalls.c          | 7 +++++++
>  2 files changed, 12 insertions(+), 1 deletion(-)
>
> diff --git a/kernel/sched/cpufreq_schedutil.c b/kernel/sched/cpufreq_schedutil.c
> index c6ba15388ea7..10faab849a3e 100644
> --- a/kernel/sched/cpufreq_schedutil.c
> +++ b/kernel/sched/cpufreq_schedutil.c
> @@ -691,7 +691,11 @@ static int sugov_kthread_create(struct sugov_policy *sg_policy)
>         }
>
>         sg_policy->thread = thread;
> -       kthread_bind_mask(thread, policy->related_cpus);
> +       if (policy->dvfs_possible_from_any_cpu)
> +               set_cpus_allowed_ptr(thread, policy->related_cpus);
> +       else
> +               kthread_bind_mask(thread, policy->related_cpus);
> +
>         init_irq_work(&sg_policy->irq_work, sugov_irq_work);
>         mutex_init(&sg_policy->work_lock);
>
> diff --git a/kernel/sched/syscalls.c b/kernel/sched/syscalls.c
> index aa70beee9895..2ef1cb8626bc 100644
> --- a/kernel/sched/syscalls.c
> +++ b/kernel/sched/syscalls.c
> @@ -1172,6 +1172,13 @@ int dl_task_check_affinity(struct task_struct *p, const struct cpumask *mask)
>         if (!task_has_dl_policy(p) || !dl_bandwidth_enabled())
>                 return 0;
>
> +       /*
> +        * The special/sugov task isn't part of regular bandwidth/admission
> +        * control so let userspace change affinities.
> +        */
> +       if (dl_entity_is_special(&p->dl))
> +               return 0;
> +
>         /*
>          * Since bandwidth control happens on root_domain basis,
>          * if admission test is enabled, we only admit -deadline
> --
> 2.34.1