Re: [RFC][PATCH 3/9] cpuidle: Move IRQ state validation

[Frederic Weisbecker <frederic@kernel.org>]

On Thu, May 19, 2022 at 11:27:53PM +0200, Peter Zijlstra wrote:
> Make cpuidle_enter_state() consistent with the s2idle variant and
> verify ->enter() always returns with interrupts disabled.
> 
> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
> ---
>  drivers/cpuidle/cpuidle.c |   10 +++++-----
>  1 file changed, 5 insertions(+), 5 deletions(-)
> 
> --- a/drivers/cpuidle/cpuidle.c
> +++ b/drivers/cpuidle/cpuidle.c
> @@ -234,7 +234,11 @@ int cpuidle_enter_state(struct cpuidle_d
>  	stop_critical_timings();
>  	if (!(target_state->flags & CPUIDLE_FLAG_RCU_IDLE))
>  		rcu_idle_enter();
> +
>  	entered_state = target_state->enter(dev, drv, index);
> +	if (WARN_ON_ONCE(!irqs_disabled()))
> +		raw_local_irq_disable();

So it means that idle functions are supposed to return with IRQs disabled
without tracing, right? I can see that at least acpi_safe_halt() is using
the non-raw local_irq_disable().

> +
>  	if (!(target_state->flags & CPUIDLE_FLAG_RCU_IDLE))
>  		rcu_idle_exit();
>  	start_critical_timings();
> @@ -246,12 +250,8 @@ int cpuidle_enter_state(struct cpuidle_d
>  	/* The cpu is no longer idle or about to enter idle. */
>  	sched_idle_set_state(NULL);
>  
> -	if (broadcast) {
> -		if (WARN_ON_ONCE(!irqs_disabled()))
> -			local_irq_disable();
> -
> +	if (broadcast)
>  		tick_broadcast_exit();
> -	}
>  
>  	if (!cpuidle_state_is_coupled(drv, index))
>  		local_irq_enable();
> 
>