On Wed, Aug 16, 2023, Kai Huang wrote:
> On Tue, 2023-08-15 at 13:36 -0700, Sean Christopherson wrote:
> > Track "VMX exposed to L1" via a governed feature flag instead of using a
> > dedicated helper to provide the same functionality. The main goal is to
> > drive convergence between VMX and SVM with respect to querying features
> > that are controllable via module param (SVM likes to cache nested
> > features), avoiding the guest CPUID lookups at runtime is just a bonus
> > and unlikely to provide any meaningful performance benefits.
> >
> > No functional change intended.
> >
> >
> [...]
>
> >
> > -/*
> > - * nested_vmx_allowed() checks whether a guest should be allowed to use VMX
> > - * instructions and MSRs (i.e., nested VMX). Nested VMX is disabled for
> > - * all guests if the "nested" module option is off, and can also be disabled
> > - * for a single guest by disabling its VMX cpuid bit.
> > - */
> > -bool nested_vmx_allowed(struct kvm_vcpu *vcpu)
> > -{
> > - return nested && guest_cpuid_has(vcpu, X86_FEATURE_VMX);
> > -}
> > -
> >
>
> [...]
>
> > @@ -7750,13 +7739,15 @@ static void vmx_vcpu_after_set_cpuid(struct kvm_vcpu *vcpu)
> > guest_cpuid_has(vcpu, X86_FEATURE_XSAVE))
> > kvm_governed_feature_check_and_set(vcpu, X86_FEATURE_XSAVES);
> >
> > + kvm_governed_feature_check_and_set(vcpu, X86_FEATURE_VMX);
> > +
> >
>
> Nit:
>
> nested_vmx_allowed() also checks 'nested' global variable. However
> kvm_governed_feature_check_and_set() is called unconditionally. Although IIUC
> it should never actually set the VMX governed bit when 'nested=0', it's not that
> obvious in _this_ patch.
Yeah, 100% agree after re-reading this patch without context. I had to go search
the code to remember where "nested" is checked. :-)
> Should we explicitly call this out in the changelog so git blamers can
> understand this more easily in the future?
Yep, I'll add a blurb to point out the dependency in vmx_set_cpu_caps().
Thanks!
