RE: [PATCH v12 01/37] x86/cpufeatures: Add the cpu feature bit for WRMSRNS

[Li, Xin3]

> Then, further down in the patchset, it says:
> 
> +	if (cpu_feature_enabled(X86_FEATURE_FRED)) {
> +		/* WRMSRNS is a baseline feature for FRED. */
> 
> but WRMSRNS is not mentioned in the FRED spec "Document Number:
> 346446-005US, Revision: 5.0" which, according to
> 
> https://www.intel.com/content/www/us/en/content-details/780121/flexible-
> return-and-event-delivery-fred-specification.html
> 
> is the latest.
> 
> Am I looking at the wrong one?

No.  tglx asked for it:
https://lkml.kernel.org/kvm/87y1h81ht4.ffs@tglx/

> 
> And now I'm wondering: when you're adding a separate CPUID bit, then the
> above should be
> 
> +       if (cpu_feature_enabled(X86_FEATURE_WRMSRNS)) {
> +               /* WRMSRNS is a baseline feature for FRED. */

Because we are doing 
		wrmsrns(MSR_IA32_FRED_RSP0, ...)
here, and X86_FEATURE_WRMSRNS doesn't guarantee MSR_IA32_FRED_RSP0 exists.

Or I missed something?

> 
> I see that you're adding a dependency:
> 
> +	{ X86_FEATURE_FRED,			X86_FEATURE_WRMSRNS   },
> 
> which then means you don't need the X86_FEATURE_WRMSRNS definition at all
> and can use X86_FEATURE_FRED only.
> 
> So, what's up?

FRED just gets the honor to introduce WRMSRNS and its first usage:
https://lkml.kernel.org/kvm/b05e3092-8ba3-f4e1-b5a3-2125944936fd@xxxxxxxxx/

Another patch set should replace WRMSR with WRMSRNS, with SERIALIZE added
when needed.

Sorry for the late response, it was a long weekend in the US.

Thanks!
    Xin