Re: [PATCH v4.1] cpufreq: Fail driver register if it has adjust_perf without fast_switch

From: Rafael J. Wysocki
Date: Thu Jun 15 2023 - 12:16:36 EST


On Mon, May 29, 2023 at 4:26 PM Wyes Karny <wyes.karny@xxxxxxx> wrote:
>
> If fast_switch_possible flag is set by the scaling driver, the governor
> is free to select fast_switch function even if adjust_perf is set. Some
> scaling drivers which use adjust_perf don't set fast_switch thinking
> that the governor would never fall back to fast_switch. But the governor
> can fall back to fast_switch even in runtime if frequency invariance is
> disabled due to some reason. This could crash the kernel if the driver
> didn't set the fast_switch function pointer.
>
> Therefore, fail driver registration if it has adjust_perf without
> fast_switch.
>
> Suggested-by: Rafael J. Wysocki <rafael@xxxxxxxxxx>
> Suggested-by: Viresh Kumar <viresh.kumar@xxxxxxxxxx>
> Signed-off-by: Wyes Karny <wyes.karny@xxxxxxx>
> ---
> drivers/cpufreq/cpufreq.c | 3 ++-
> include/linux/cpufreq.h | 5 ++++-
> 2 files changed, 6 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c
> index 6b52ebe5a890..50bbc969ffe5 100644
> --- a/drivers/cpufreq/cpufreq.c
> +++ b/drivers/cpufreq/cpufreq.c
> @@ -2828,7 +2828,8 @@ int cpufreq_register_driver(struct cpufreq_driver *driver_data)
> (driver_data->setpolicy && (driver_data->target_index ||
> driver_data->target)) ||
> (!driver_data->get_intermediate != !driver_data->target_intermediate) ||
> - (!driver_data->online != !driver_data->offline))
> + (!driver_data->online != !driver_data->offline) ||
> + (driver_data->adjust_perf && !driver_data->fast_switch))
> return -EINVAL;
>
> pr_debug("trying to register driver %s\n", driver_data->name);
> diff --git a/include/linux/cpufreq.h b/include/linux/cpufreq.h
> index 26e2eb399484..172ff51c1b2a 100644
> --- a/include/linux/cpufreq.h
> +++ b/include/linux/cpufreq.h
> @@ -340,7 +340,10 @@ struct cpufreq_driver {
> /*
> * ->fast_switch() replacement for drivers that use an internal
> * representation of performance levels and can pass hints other than
> - * the target performance level to the hardware.
> + * the target performance level to the hardware. This can only be set
> + * if ->fast_switch is set too, because in those cases (under specific
> + * conditions) scale invariance can be disabled, which causes the
> + * schedutil governor to fall back to the latter.
> */
> void (*adjust_perf)(unsigned int cpu,
> unsigned long min_perf,
> --

Applied as 6.5 material, thanks!