RE: [PATCH] cpuidle, ACPI: Evaluate LPI arch_flags for broadcast timer

[Pawandeep Oza (QUIC)]

Hi Sudeep,

Please find my comments inline.

Regards,
Oza.

-----Original Message-----
From: Sudeep Holla <sudeep.holla@xxxxxxx> 
Sent: Tuesday, July 18, 2023 6:32 AM
To: Pawandeep Oza (QUIC) <quic_poza@xxxxxxxxxxx>
Cc: Catalin Marinas <catalin.marinas@xxxxxxx>; Will Deacon <will@xxxxxxxxxx>; Sudeep Holla <sudeep.holla@xxxxxxx>; Rafael J . Wysocki <rafael@xxxxxxxxxx>; Len Brown <lenb@xxxxxxxxxx>; linux-arm-kernel@xxxxxxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx; linux-acpi@xxxxxxxxxxxxxxx; Jamie Iles <jiles@xxxxxxxxxxxxxxxx>
Subject: Re: [PATCH] cpuidle, ACPI: Evaluate LPI arch_flags for broadcast timer

On Wed, Jul 12, 2023 at 10:24:58AM -0700, Oza Pawandeep wrote:
> Arm(r) Functional Fixed Hardware Specification defines LPI states, which 
> provides an architectural context loss flags field that can be used to 
> describe the context that might be lost when an LPI state is entered.
> 
> - Core context Lost
> 	- General purpose registers.
> 	- Floating point and SIMD registers.
> 	- System registers, include the System register based
> 	- generic timer for the core.
> 	- Debug register in the core power domain.
> 	- PMU registers in the core power domain.
> 	- Trace register in the core power domain.
> - Trace context loss
> - GICR
> - GICD
> 
> Qualcomm's custom CPUs preserves the architectural state, including 
> keeping the power domain for local timers active.
> when core is power gated, the local timers are sufficient to wake the 
> core up without needing broadcast timer.
> 
> The patch fixes the evaluation of cpuidle arch_flags, and moves only 
> to broadcast timer if core context lost is defined in ACPI LPI.
> 
> Signed-off-by: Oza Pawandeep <quic_poza@xxxxxxxxxxx>
> 
> diff --git a/arch/arm64/include/asm/acpi.h 
> b/arch/arm64/include/asm/acpi.h index bd68e1b7f29f..9c335968316c 
> 100644
> --- a/arch/arm64/include/asm/acpi.h
> +++ b/arch/arm64/include/asm/acpi.h
> @@ -42,6 +42,24 @@
>  #define ACPI_MADT_GICC_SPE  (offsetof(struct acpi_madt_generic_interrupt, \
>  	spe_interrupt) + sizeof(u16))
>  
> +/*
> + * Arm(r) Functional Fixed Hardware Specification Version 1.2.
> + * Table 2: Arm Architecture context loss flags  */
> +#define CPUIDLE_CORE_CTXT		BIT(0) /* Core context Lost */
> +
> +#ifndef arch_acpi_lpi_timer_stopped
> +static __always_inline bool arch_acpi_lpi_timer_stopped(u32 
> +arch_flags)

As mentioned by you above, the core context is not just timer context, so calling this function so is misleading.

 Oza: Will take care of you last 2 comments, probably this should be taken care as well with that.

> +{
> +  return arch_flags & CPUIDLE_CORE_CTXT; } #define 
> +arch_acpi_lpi_timer_stopped arch_acpi_lpi_timer_stopped #endif
> +
> +#define CPUIDLE_TRACE_CTXT		BIT(1) /* Trace context loss */
> +#define CPUIDLE_GICR_CTXT		BIT(2) /* GICR */
> +#define CPUIDLE_GICD_CTXT		BIT(3) /* GICD */
> +

Do we really need to define these unused bitfields ? DO you have plans to use them ?

Oza: this is basically defined in ARM FFH - I think no harm is having them here for sake of completeness.

>  /* Basic configuration for ACPI */
>  #ifdef	CONFIG_ACPI
>  pgprot_t __acpi_get_mem_attribute(phys_addr_t addr); diff --git 
> a/drivers/acpi/processor_idle.c b/drivers/acpi/processor_idle.c index 
> 9718d07cc2a2..8ea1f2b3bf96 100644
> --- a/drivers/acpi/processor_idle.c
> +++ b/drivers/acpi/processor_idle.c
> @@ -1221,7 +1221,7 @@ static int acpi_processor_setup_lpi_states(struct acpi_processor *pr)
>  		strscpy(state->desc, lpi->desc, CPUIDLE_DESC_LEN);
>  		state->exit_latency = lpi->wake_latency;
>  		state->target_residency = lpi->min_residency;
> -		if (lpi->arch_flags)
> +		if (arch_acpi_lpi_timer_stopped(lpi->arch_flags))

While setting CPUIDLE_FLAG_TIMER_STOP if any flags set is already questionable, checking for arch specific flag in the generic code is even more questionable now. I wonder if it makes more sense to have a arch specific helper to update the state->flags based on how arch specific interpretation of lpi->arch_flags ?

>  			state->flags |= CPUIDLE_FLAG_TIMER_STOP;
>  		if (i != 0 && lpi->entry_method == ACPI_CSTATE_FFH)
>  			state->flags |= CPUIDLE_FLAG_RCU_IDLE; diff --git 
> a/include/linux/acpi.h b/include/linux/acpi.h index 
> d584f94409e1..b24f1cd1cebb 100644
> --- a/include/linux/acpi.h
> +++ b/include/linux/acpi.h
> @@ -1471,6 +1471,14 @@ static inline int 
> lpit_read_residency_count_address(u64 *address)  }  #endif
>  
> +#ifndef arch_acpi_lpi_timer_stopped
> +static __always_inline bool arch_acpi_lpi_timer_stopped(u32 
> +arch_flags) {
> +  return (arch_flags != 0);
> +}
> +#define arch_acpi_lpi_timer_stopped arch_acpi_lpi_timer_stopped 
> +#endif
> +

This looks ugly and main reason for my above comment. I am thinking of arch_update_idle_state_flags(lpi->arch_flags, &state->flags) and make it do nothing on non arm platforms. I don't think we will be breaking anything(i.e. no need to check arch_flags != 0. It is incorrect strictly speaking but there are no non-arm users ATM, but that doesn't mean we can trickle the arch specific LPI FFH details into the generic code.

Oza: Let me work on above 2 comments and post the new patch.

--
Regards,
Sudeep