Re: [PATCH v2] x86/speculation/srbds: do not try to turn mitigation off when not supported

[Pawan Gupta]

On Thu, Mar 31, 2022 at 05:25:27PM +0200, Borislav Petkov wrote:
> On Thu, Mar 31, 2022 at 03:18:42PM +0200, Ricardo Cañuelo wrote:
> > Just to clarify, this changes the behavior wrt the hypervisor case:
> > currently it just bails out of update_srbds_msr(), with your patch it'd
> > clear RNGDS_MITG_DIS from MSR_IA32_MCU_OPT_CTRL. Is that what you
> > intended?

Yes, this should be fine as update_srbds_msr() would also check for
X86_FEATURE_SRBDS_CTRL. If a hypervisor doesn't want a guest to write to
the MSR it should not export X86_FEATURE_SRBDS_CTRL.

> Just do the simple thing I pasted earlier - no need to rewrite the whole
> function for no good reason.

I was trying to address earlier comment on split logic in two functions.
I am okay with keeping it as is (and just adding X86_FEATURE_SRBDS_CTRL
check in update_srbds_msr()).

Thanks,
Pawan