Re: [PATCH] thermal/intel_powerclamp: Don't report an error for AMD CPUs

[Thomas Gleixner <tglx@linutronix.de>]

Alexander Koskovich <zvnexus@gmail.com> writes:

> Resolves dmesg error "intel_powerclamp: CPU does not support MWAIT".
>
> The error that is outputted in dmesg prior to this patch
> is innacurate, AMD Ryzen CPUs do support MWAIT. We could
> also add the AMD vendor to the MWAIT check, but even though
> AMD CPUs do support MWAIT, they fail the C-state package
> check so it's better just to bail out in the beginning.
>
> Signed-off-by: Alexander Koskovich <zvnexus@outlook.com>
> ---
>  drivers/thermal/intel/intel_powerclamp.c | 10 ++++++++++
>  1 file changed, 10 insertions(+)
>
> diff --git a/drivers/thermal/intel/intel_powerclamp.c b/drivers/thermal/intel/intel_powerclamp.c
> index 53216dcbe173..3c5b25bfa596 100644
> --- a/drivers/thermal/intel/intel_powerclamp.c
> +++ b/drivers/thermal/intel/intel_powerclamp.c
> @@ -650,6 +650,11 @@ static struct thermal_cooling_device_ops powerclamp_cooling_ops = {
>  	.set_cur_state = powerclamp_set_cur_state,
>  };
>  
> +static const struct x86_cpu_id amd_cpu[] = {
> +	{ X86_VENDOR_AMD },
> +	{},
> +};
> +
>  static const struct x86_cpu_id __initconst intel_powerclamp_ids[] = {
>  	{ X86_VENDOR_INTEL, X86_FAMILY_ANY, X86_MODEL_ANY, X86_FEATURE_MWAIT },
>  	{}
> @@ -659,6 +664,11 @@ MODULE_DEVICE_TABLE(x86cpu, intel_powerclamp_ids);
>  static int __init powerclamp_probe(void)
>  {
>  
> +	if (x86_match_cpu(amd_cpu)) {
> +		pr_info("Intel PowerClamp does not support AMD CPUs\n");
> +		return -ENODEV;

This is still running into the same problem on all other non Intel
vendors, e.g. HYGON, VIA ....

> +	}
> +
>  	if (!x86_match_cpu(intel_powerclamp_ids)) {
>  		pr_err("CPU does not support MWAIT\n");
>  		return -ENODEV;

The right thing to do is to remove this silly pr_err(). It's not an
error when a CPU does not support MWAIT. It's a fact and even older
Intel CPUs do not have MWAIT.

We do not print "Machine does not have $FEATURE" in device drivers and
whatever code which depends on runtime detection either. dmesg would be
full of this.

Thanks,

        tglx