Re: [PATCH] PM: s2idle: Don't allow s2idle when cpuidle isn't supported

[Hector Martin]

On 15/02/2023 06.50, Kazuki Hashimoto 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"), so update the suspend framework to reflect this in order to avoid
> breakages.
> 
> 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 Hashimoto <kazukih0205@xxxxxxxxx>
> ---
>  kernel/power/suspend.c | 6 ++++++
>  1 file changed, 6 insertions(+)
> 
> diff --git a/kernel/power/suspend.c b/kernel/power/suspend.c
> index 3f436282547c..27507ae7c9c9 100644
> --- a/kernel/power/suspend.c
> +++ b/kernel/power/suspend.c
> @@ -556,6 +556,12 @@ static int enter_state(suspend_state_t state)
>  
>  	trace_suspend_resume(TPS("suspend_enter"), state, true);
>  	if (state == PM_SUSPEND_TO_IDLE) {
> +		struct cpuidle_device *dev = cpuidle_get_device();
> +		struct cpuidle_driver *drv = cpuidle_get_cpu_driver(dev);
> +
> +		if (cpuidle_not_available(drv, dev))
> +			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");

Well... this turns s2idle from "slighly broken" to "fully broken". I'm
not sure that's a good idea.

If this (or something equivalent) goes in, we'll have to carry a revert
in the Asahi tree until the PSCI-replacement story goes in, because lots
of people are using the current somewhat broken s2idle. It's a lot
better than having no sleep modes at all.

Also, if you intend to disable s2idle on these platforms, you have to
actually stop advertising it. Advertising it and then refusing to enter
s2idle is not acceptable, it means people will get "sleep" buttons and
expect sleep to work and then it won't. But then that could also
introduce race conditions with userspace checking for sleep support
before the cpuidle driver loads. So there is definitely a can of worms here.

- Hector