Re: [PATCH v2 2/5] cpuidle: psci: Mark as PREEMPT_RT safe

[Ulf Hansson <ulf.hansson@linaro.org>]

On Mon, 19 Dec 2022 at 16:15, Krzysztof Kozlowski
<krzysztof.kozlowski@linaro.org> wrote:
>
> The PSCI cpuidle power domain in power_off callback uses
> __this_cpu_write() so it is PREEMPT_RT safe.  This allows to use it in
> Realtime kernels and solves errors like:
>
>   BUG: scheduling while atomic: swapper/2/0/0x00000002
>   Hardware name: Qualcomm Technologies, Inc. Robotics RB5 (DT)
>   Call trace:
>    dump_backtrace.part.0+0xe0/0xf0
>    show_stack+0x18/0x40
>    dump_stack_lvl+0x68/0x84
>    dump_stack+0x18/0x34
>    __schedule_bug+0x60/0x80
>    __schedule+0x628/0x800
>    schedule_rtlock+0x28/0x5c
>    rtlock_slowlock_locked+0x360/0xd30
>    rt_spin_lock+0x88/0xb0
>    genpd_lock_nested_spin+0x1c/0x30
>    genpd_power_off.part.0.isra.0+0x20c/0x2a0
>    genpd_runtime_suspend+0x150/0x2bc
>    __rpm_callback+0x48/0x170
>    rpm_callback+0x6c/0x7c
>    rpm_suspend+0x108/0x660
>    __pm_runtime_suspend+0x4c/0x8c
>    __psci_enter_domain_idle_state.constprop.0+0x54/0xe0
>    psci_enter_domain_idle_state+0x18/0x2c
>    cpuidle_enter_state+0x8c/0x4e0
>    cpuidle_enter+0x38/0x50
>    do_idle+0x248/0x2f0
>    cpu_startup_entry+0x24/0x30
>    secondary_start_kernel+0x130/0x154
>    __secondary_switched+0xb0/0xb4
>
> Cc: Adrien Thierry <athierry@redhat.com>
> Cc: Brian Masney <bmasney@redhat.com>
> Cc: linux-rt-users@vger.kernel.org
> Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
> ---
>  drivers/cpuidle/cpuidle-psci-domain.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/cpuidle/cpuidle-psci-domain.c b/drivers/cpuidle/cpuidle-psci-domain.c
> index c80cf9ddabd8..d15a91fb7048 100644
> --- a/drivers/cpuidle/cpuidle-psci-domain.c
> +++ b/drivers/cpuidle/cpuidle-psci-domain.c
> @@ -62,7 +62,8 @@ static int psci_pd_init(struct device_node *np, bool use_osi)
>         if (!pd_provider)
>                 goto free_pd;
>
> -       pd->flags |= GENPD_FLAG_IRQ_SAFE | GENPD_FLAG_CPU_DOMAIN;
> +       pd->flags |= GENPD_FLAG_IRQ_SAFE | GENPD_FLAG_RT_SAFE | \
> +                    GENPD_FLAG_CPU_DOMAIN;

My main concern with this, is that it will affect the parent domains
too. Whether those would be able to use the GENPD_FLAG_RT_SAFE or not,
is a different story.

In one way or the other, I think it would be better to limit the
GENPD_FLAG_RT_SAFE to be used only for PREEMPT_RT kernels.

>
>         /* Allow power off when OSI has been successfully enabled. */
>         if (use_osi)
> --
> 2.34.1
>

Kind regards
Uffe