Re: [RFC PATCH v4 8/8] acpi_idle: Disallow play_dead with FFH cstate on AMD platforms

[Peter Zijlstra <peterz@infradead.org>]

On Mon, Nov 25, 2024 at 02:20:28PM +0100, Patryk Wlazlyn wrote:
> Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@linux.intel.com>
> Suggested-by: Gautham R. Shenoy <gautham.shenoy@amd.com>
> ---
>  drivers/acpi/processor_idle.c | 6 +++++-
>  1 file changed, 5 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/acpi/processor_idle.c b/drivers/acpi/processor_idle.c
> index 586cc7d1d8aa..4b4ac8d55b55 100644
> --- a/drivers/acpi/processor_idle.c
> +++ b/drivers/acpi/processor_idle.c
> @@ -803,7 +803,11 @@ static int acpi_processor_setup_cstates(struct acpi_processor *pr)
>  
>  		state->flags = 0;
>  
> -		state->enter_dead = acpi_idle_play_dead;
> +		/* AMD doesn't want to use mwait for play dead. */
> +		bool amd_or_hygon = boot_cpu_data.x86_vendor == X86_VENDOR_AMD ||
> +				    boot_cpu_data.x86_vendor == X86_VENDOR_HYGON;
> +		if (!(cx->entry_method == ACPI_CSTATE_FFH && amd_or_hygon))
> +			state->enter_dead = acpi_idle_play_dead;

So I don't like this. Less exceptions is better.

This *SHOULD* never trigger on AMD anyway, because they recommend IO
port C[23]. But if their partner BIOS engineer does a wobbly and they
end up in MWAIT anyway, it *should* all work regardless.