Re: [RFC][PATCH] cpufreq: intel_pstate: Delete intel_pstate sysfs if failed to register the driver

[Rafael J. Wysocki]

On Fri, Oct 9, 2020 at 5:29 AM Chen Yu <yu.c.chen@xxxxxxxxx> wrote:
>
> There is a corner case that if the intel_pstate driver failed to be
> registered(might be due to invalid MSR access) and with the acpi_cpufreq
> loaded, the intel_pstate sysfs might still be created, which makes the
> user confusing(turbostat for example):
>
> grep . /sys/devices/system/cpu/cpu0/cpufreq/scaling_driver
> acpi-cpufreq
>
> grep . /sys/devices/system/cpu/intel_pstate/*
> /sys/devices/system/cpu/intel_pstate/max_perf_pct:0
> /sys/devices/system/cpu/intel_pstate/min_perf_pct:0
> grep: /sys/devices/system/cpu/intel_pstate/no_turbo: Resource temporarily unavailable
> grep: /sys/devices/system/cpu/intel_pstate/num_pstates: Resource temporarily unavailable
> /sys/devices/system/cpu/intel_pstate/status:off
> grep: /sys/devices/system/cpu/intel_pstate/turbo_pct: Resource temporarily unavailable
>
> The existing of intel_pstate sysfs does not mean that the intel_pstate driver
> has been successfully loaded(for example, echo off to status), but the
> intel_pstate sysfs should not co-exist when acpi-cpufreq is also present.
> Fix this issue by deleting the intel_pstate sysfs if the driver failed
> to be loaded during bootup.
>
> Reported-by: Wendy Wang <wendy.wang@xxxxxxxxx>
> Suggested-by: Zhang Rui <rui.zhang@xxxxxxxxx>
> Signed-off-by: Chen Yu <yu.c.chen@xxxxxxxxx>
> ---
>  drivers/cpufreq/intel_pstate.c | 24 +++++++++++++++++++++++-
>  1 file changed, 23 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/cpufreq/intel_pstate.c b/drivers/cpufreq/intel_pstate.c
> index 9a515c460a00..8c5f9680de83 100644
> --- a/drivers/cpufreq/intel_pstate.c
> +++ b/drivers/cpufreq/intel_pstate.c
> @@ -1420,6 +1420,26 @@ static void __init intel_pstate_sysfs_expose_params(void)
>         }
>  }
>
> +static void __init intel_pstate_sysfs_clean(void)
> +{
> +       if (!intel_pstate_kobject)
> +               return;
> +
> +       sysfs_remove_group(intel_pstate_kobject, &intel_pstate_attr_group);
> +
> +       if (per_cpu_limits)
> +               goto release_kobj;
> +
> +       sysfs_remove_file(intel_pstate_kobject, &max_perf_pct.attr);
> +       sysfs_remove_file(intel_pstate_kobject, &min_perf_pct.attr);
> +
> +       if (x86_match_cpu(intel_pstate_cpu_ee_disable_ids))
> +               sysfs_remove_file(intel_pstate_kobject, &energy_efficiency.attr);
> +
> +release_kobj:
> +       kobject_put(intel_pstate_kobject);
> +}
> +
>  static void intel_pstate_sysfs_expose_hwp_dynamic_boost(void)
>  {
>         int rc;
> @@ -3063,8 +3083,10 @@ static int __init intel_pstate_init(void)
>         mutex_lock(&intel_pstate_driver_lock);
>         rc = intel_pstate_register_driver(default_driver);
>         mutex_unlock(&intel_pstate_driver_lock);
> -       if (rc)
> +       if (rc) {
> +               intel_pstate_sysfs_clean();
>                 return rc;
> +       }
>
>         if (hwp_active) {
>                 const struct x86_cpu_id *id;
> --

Applied as 5.10-rc material with some minor changes and the Srinivas'
ACK, thanks!