Re: [PATCH V3 2/3] cpufreq: intel_pstate: Migrate away from ->stop_cpu() callback

["Rafael J. Wysocki" <rafael@kernel.org>]

On Mon, Jun 21, 2021 at 5:09 AM Viresh Kumar <viresh.kumar@linaro.org> wrote:
>
> On 18-06-21, 14:00, Rafael J. Wysocki wrote:
> > On Fri, Jun 18, 2021 at 5:22 AM Viresh Kumar <viresh.kumar@linaro.org> wrote:
> > >
> > > commit 367dc4aa932b ("cpufreq: Add stop CPU callback to cpufreq_driver
> > > interface") added the stop_cpu() callback to allow the drivers to do
> > > clean up before the CPU is completely down and its state can't be
> > > modified.
> > >
> > > At that time the CPU hotplug framework used to call the cpufreq core's
> > > registered notifier for different events like CPU_DOWN_PREPARE and
> > > CPU_POST_DEAD. The stop_cpu() callback was called during the
> > > CPU_DOWN_PREPARE event.
> > >
> > > This is no longer the case, cpuhp_cpufreq_offline() is called only once
> > > by the CPU hotplug core now and we don't really need to separately
> > > call stop_cpu() for cpufreq drivers.
> > >
> > > Migrate to using the exit() and offline() callbacks instead of
> > > stop_cpu().
> > >
> > > We need to clear util hook from both the callbacks, exit() and
> > > offline(), since it is possible that only exit() gets called sometimes
> > > (specially on errors) or both get called at other times.
> > > intel_pstate_clear_update_util_hook() anyway have enough protection in
> > > place if it gets called a second time and will return early then.
> > >
> > > Cc: Dirk Brandewie <dirk.brandewie@gmail.com>
> > > Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
> > > ---
> > > V2->V3:
> > > - Update intel_pstate_cpu_offline() as well.
> > > - Improved commit log.
> > >
> > >  drivers/cpufreq/intel_pstate.c | 12 ++++--------
> > >  1 file changed, 4 insertions(+), 8 deletions(-)
> > >
> > > diff --git a/drivers/cpufreq/intel_pstate.c b/drivers/cpufreq/intel_pstate.c
> > > index 0e69dffd5a76..8f8a2d9d7daa 100644
> > > --- a/drivers/cpufreq/intel_pstate.c
> > > +++ b/drivers/cpufreq/intel_pstate.c
> > > @@ -2335,6 +2335,8 @@ static int intel_pstate_cpu_offline(struct cpufreq_policy *policy)
> > >
> > >         pr_debug("CPU %d going offline\n", cpu->cpu);
> > >
> > > +       intel_pstate_clear_update_util_hook(policy->cpu);
> > > +
> > >         if (cpu->suspended)
> > >                 return 0;
> > >
> > > @@ -2374,17 +2376,12 @@ static int intel_pstate_cpu_online(struct cpufreq_policy *policy)
> > >         return 0;
> > >  }
> > >
> > > -static void intel_pstate_stop_cpu(struct cpufreq_policy *policy)
> > > -{
> > > -       pr_debug("CPU %d stopping\n", policy->cpu);
> > > -
> > > -       intel_pstate_clear_update_util_hook(policy->cpu);
> > > -}
> > > -
> > >  static int intel_pstate_cpu_exit(struct cpufreq_policy *policy)
> > >  {
> > >         pr_debug("CPU %d exiting\n", policy->cpu);
> > >
> > > +       intel_pstate_clear_update_util_hook(policy->cpu);
> >
> > This change is not needed now, because ->offline always runs before
> > ->exit if present.
>
> Not necessarily, we don't call ->offline() for many error paths in
> cpufreq_online().

I guess you mean the error paths in cpufreq_offline()?

IMO this is confusing/broken, because ->offline should always be
called after ->online has returned success.

> offline() only comes into play after driver is registered properly once.

The relevant intel_pstate case is a ->setpolicy driver where
->setpolicy or ->online, if successful, need to be followed by
->offline.

If ->setpolicy is successful in the cpufreq_online() path, the entire
cpufreq_online() is successful and the error paths in question are not
executed.

So the change I was talking about is not needed AFAICS.