Re: [PATCH v1 1/4] ACPI: processor: Reorder acpi_processor_driver_init()

From: Rafael J. Wysocki
Date: Mon Mar 13 2023 - 11:00:46 EST


On Mon, Mar 13, 2023 at 3:54 PM Zhang, Rui <rui.zhang@xxxxxxxxx> wrote:
>
> On Mon, 2023-03-13 at 14:48 +0100, Rafael J. Wysocki wrote:
> > On Sun, Mar 12, 2023 at 5:09 PM Zhang, Rui <rui.zhang@xxxxxxxxx>
> > wrote:
> > > On Fri, 2023-03-03 at 20:19 +0100, Rafael J. Wysocki wrote:
> > > > From: Rafael J. Wysocki <rafael.j.wysocki@xxxxxxxxx>
> > > >
> > > > The cpufreq policy notifier in the ACPI processor driver may as
> > > > well be registered before the driver itself, which causes
> > > > acpi_processor_cpufreq_init to be true (unless the notifier
> > > > registration fails, which is unlikely at that point) when the
> > > > ACPI CPU thermal cooling devices are registered, so the
> > > > processor_get_max_state() result does not change while
> > > > acpi_processor_driver_init() is running.
> > > >
> > > > Change the ordering in acpi_processor_driver_init() accordingly
> > > > to prevent the max_state value from remaining 0 permanently for
> > > > all
> > > > ACPI CPU cooling devices.
> > > >
> > > > Fixes: a365105c685c("thermal: sysfs: Reuse cdev->max_state")
> > > > Reported-by: Wang, Quanxian <quanxian.wang@xxxxxxxxx>
> > > > Link:
> > > > https://lore.kernel.org/linux-pm/53ec1f06f61c984100868926f282647e57ecfb2d.camel@xxxxxxxxx/
> > > > Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@xxxxxxxxx>
> > >
> > > The full patch series fixes the problem but this one does not.
> >
> > That is a correct observation, but the $subject patch fixes part of
> > the problem (which is not addressed by the rest of the series AFAICS)
> > and so it deserves a Fixes tag of its own IMO.
>
> Am I understanding this correctly that this patch helps in below case?
>
> cpufreq driver like intel_pstate is registered before we register the
> notifier callback in processor_driver. In this case, we are not able to
> catch the CPUFREQ_CREATE_POLICY notification and cpufreq should be
> counted as part of cooling states when registering the ACPI CPU cooling
> device. (acpi_processor_cpufreq_init must be set at this time)

Yes.

> Or else, in normal case, the ACPI CPU cdev->max_state always return 0
> (when t-state not available) until we receive the CPUFREQ_CREATE_POLICY
> notification and call thermal_cooling_device_update(), both with and
> without this patch.
>
> Patch 2,3,4 works on my test platform, without applying patch 1/4.

OK

> > I guess I should clarify that in the changelog.
> >
> > > This is because,
> > >
> > > static int cpu_has_cpufreq(unsigned int cpu)
> > > {
> > > struct
> > > cpufreq_policy *policy;
> > >
> > > if (!acpi_processor_cpufreq_init)
> > > return 0;
> > >
> > > policy = cpufreq_cpu_get(cpu);
> > > if (policy) {
> > > cpufreq_cpu_put(policy);
> > > return 1;
> > > }
> > > return 0;
> > > }
> > >
> > > Although acpi_processor_cpufreq_init is set to true with patch 1/4,
> > > but
> > > we don't have cpufreq driver registered, thus cpufreq_cpu_get()
> > > return
> > > NULL.
> > > so acpi_processor_cpufreq_init is not the only dependency here. :(
> >
> > Right. That's why the other patches in the series are needed too.
> >
> > > > ---
> > > > drivers/acpi/processor_driver.c | 12 ++++++------
> > > > 1 file changed, 6 insertions(+), 6 deletions(-)
> > > >
> > > > Index: linux-pm/drivers/acpi/processor_driver.c
> > > > =================================================================
> > > > ==
> > > > --- linux-pm.orig/drivers/acpi/processor_driver.c
> > > > +++ linux-pm/drivers/acpi/processor_driver.c
> > > > @@ -263,6 +263,12 @@ static int __init acpi_processor_driver_
> > > > if (acpi_disabled)
> > > > return 0;
> > > >
> > > > + if
> > > > (!cpufreq_register_notifier(&acpi_processor_notifier_block,
> > > > + CPUFREQ_POLICY_NOTIFIER)) {
> > > > + acpi_processor_cpufreq_init = true;
> > > > + acpi_processor_ignore_ppc_init();
> > > > + }
> > > > +
> > > > result = driver_register(&acpi_processor_driver);
> > > > if (result < 0)
> > > > return result;
> > > > @@ -276,12 +282,6 @@ static int __init acpi_processor_driver_
> > > > cpuhp_setup_state_nocalls(CPUHP_ACPI_CPUDRV_DEAD,
> > > > "acpi/cpu-
> > > > drv:dead",
> > > > NULL, acpi_soft_cpu_dead);
> > > >
> > > > - if
> > > > (!cpufreq_register_notifier(&acpi_processor_notifier_block,
> > > > - CPUFREQ_POLICY_NOTIFIER)) {
> > > > - acpi_processor_cpufreq_init = true;
> > > > - acpi_processor_ignore_ppc_init();
> > > > - }
> > > > -
> > > > acpi_processor_throttling_init();
> > > > return 0;
> > > > err:
> > > >
> > > >
> > > >