[3/3] PM: s2idle: Fully block the system from entering s2idle when cpuidle isn't supported

s2idle isn't supported on platforms that don't support cpuidle as of
31a3409065d1 ("cpuidle / sleep: Do sanity checks in cpuidle_enter_freeze()
too").

There is a check in the cpuidle subsystem which would prevent the
system from entering s2idle. However, there is nothing in the suspend
framework which prevents this, which can cause the suspend subsystem to
think that the machine is entering s2idle while the cpuidle subsystem is
not, which can completely break the system.

Block the machine from entering s2idle when cpuidle isn't supported in
the suspend subsystem as well.

Link: https://lore.kernel.org/all/20230204152747.drte4uitljzngdt6@kazuki-mac
Fixes: 31a3409065d1 ("cpuidle / sleep: Do sanity checks in cpuidle_enter_freeze() too")
Signed-off-by: Kazuki H <kazukih0205@gmail.com>
---
 kernel/power/main.c    | 12 +++++++++---
 kernel/power/suspend.c |  5 +++++
 2 files changed, 14 insertions(+), 3 deletions(-)

Hi, apologies for the late response.

On Mon, Mar 27, 2023 at 07:43:02PM +0200, Rafael J. Wysocki wrote:
> On Thu, Mar 16, 2023 at 8:49 AM Kazuki H <kazukih0205@gmail.com> wrote:
> >
> > s2idle isn't supported on platforms that don't support cpuidle as of
> > 31a3409065d1 ("cpuidle / sleep: Do sanity checks in cpuidle_enter_freeze()
> > too").
> 
> This simply isn't correct.
> 
> As of the above commit, s2idle was still supported without cpuidle as
> long as arch_cpu_idle() worked.
Sorry I linked the wrong commit, ef2b22ac540c (cpuidle / sleep: Use
broadcast timer for states that stop local timer) is the correct one.

This adds this block of code

if (cpuidle_not_available(drv, dev))
	goto use_default;

which prevents this code, critical for s2idle, from executing

if (idle_should_freeze()) {
	entered_state = cpuidle_enter_freeze(drv, dev);
	if (entered_state >= 0) {
		local_irq_enable();
		goto exit_idle;
	}
	reflect = false;
	next_state = cpuidle_find_deepest_state(drv, dev);
} else {
	reflect = true;
	/*
	 * Ask the cpuidle framework to choose a convenient idle state.
	 */
	next_state = cpuidle_select(drv, dev);
}
> > There is a check in the cpuidle subsystem which would prevent the
> > system from entering s2idle. However, there is nothing in the suspend
> > framework which prevents this, which can cause the suspend subsystem to
> > think that the machine is entering s2idle while the cpuidle subsystem is
> > not, which can completely break the system.
> >
> > Block the machine from entering s2idle when cpuidle isn't supported in
> > the suspend subsystem as well.
> 
> First, please explain why the cpuidle_not_available() check in
> cpuidle_idle_call() is not sufficient to avoid the breakage.
cpuidle_idle_call() doesn't notify the PM subsystem that the host
doesn't support s2idle. We need a seperate check here, otherwise the PM
subsystem will happily continue entering s2idle without the other
subsystem's knowledge.
> 
> > Link: https://lore.kernel.org/all/20230204152747.drte4uitljzngdt6@kazuki-mac
> > Fixes: 31a3409065d1 ("cpuidle / sleep: Do sanity checks in cpuidle_enter_freeze() too")
> > Signed-off-by: Kazuki H <kazukih0205@gmail.com>
> > ---
> >  kernel/power/main.c    | 12 +++++++++---
> >  kernel/power/suspend.c |  5 +++++
> >  2 files changed, 14 insertions(+), 3 deletions(-)
> >
> > diff --git a/kernel/power/main.c b/kernel/power/main.c
> > index 31ec4a9b9d70..b14765447989 100644
> > --- a/kernel/power/main.c
> > +++ b/kernel/power/main.c
> > @@ -133,6 +133,8 @@ static ssize_t mem_sleep_show(struct kobject *kobj, struct kobj_attribute *attr,
> >         for (i = PM_SUSPEND_MIN; i < PM_SUSPEND_MAX; i++) {
> >                 if (i >= PM_SUSPEND_MEM && cxl_mem_active())
> >                         continue;
> > +               if (i == PM_SUSPEND_TO_IDLE && cpuidle_not_available())
> > +                       continue;
> 
> Not really.
> 
> >                 if (mem_sleep_states[i]) {
> >                         const char *label = mem_sleep_states[i];
> >
> > @@ -185,11 +187,15 @@ static ssize_t mem_sleep_store(struct kobject *kobj, struct kobj_attribute *attr
> >         }
> >
> >         state = decode_suspend_state(buf, n);
> > -       if (state < PM_SUSPEND_MAX && state > PM_SUSPEND_ON)
> > +       if (state == PM_SUSPEND_TO_IDLE && cpuidle_not_available())
> 
> And same here.
> 
> > +               goto einval;
> > +       if (state < PM_SUSPEND_MAX && state > PM_SUSPEND_ON) {
> >                 mem_sleep_current = state;
> > -       else
> > -               error = -EINVAL;
> > +               goto out;
> > +       }
> >
> > + einval:
> > +       error = -EINVAL;
> >   out:
> >         pm_autosleep_unlock();
> >         return error ? error : n;
> > diff --git a/kernel/power/suspend.c b/kernel/power/suspend.c
> > index 3f436282547c..55ddf525aaaf 100644
> > --- a/kernel/power/suspend.c
> > +++ b/kernel/power/suspend.c
> > @@ -556,6 +556,11 @@ static int enter_state(suspend_state_t state)
> >
> >         trace_suspend_resume(TPS("suspend_enter"), state, true);
> >         if (state == PM_SUSPEND_TO_IDLE) {
> > +               if (cpuidle_not_available()) {
> > +                       pr_warn("s2idle is unsupported when cpuidle is unavailable");
> > +                       return -EINVAL;
> > +               }
> > +
> >  #ifdef CONFIG_PM_DEBUG
> >                 if (pm_test_level != TEST_NONE && pm_test_level <= TEST_CPUS) {
> >                         pr_warn("Unsupported test mode for suspend to idle, please choose none/freezer/devices/platform.\n");
> > --
> 
> Again, I need to understand what the problem is in the first place.
Thanks,
Kazuki