On Tue, Aug 19, 2025, Xin Li wrote:
> On 8/19/2025 9:11 AM, Sean Christopherson wrote:
> > On Mon, Aug 11, 2025, Chao Gao wrote:
> > > From: Yang Weijiang <weijiang.yang@xxxxxxxxx>
> > >
> > > Enable/disable CET MSRs interception per associated feature configuration.
> > >
> > > Shadow Stack feature requires all CET MSRs passed through to guest to make
> > > it supported in user and supervisor mode
> >
> > I doubt that SS _requires_ CET MSRs to be passed through. IIRC, the actual
> > reason for passing through most of the MSRs is that they are managed via XSAVE,
> > i.e. _can't_ be intercepted without also intercepting XRSTOR.
> >
> > > while IBT feature only depends on
> > > MSR_IA32_{U,S}_CETS_CET to enable user and supervisor IBT.
> > >
> > > Note, this MSR design introduced an architectural limitation of SHSTK and
> > > IBT control for guest, i.e., when SHSTK is exposed, IBT is also available
> > > to guest from architectural perspective since IBT relies on subset of SHSTK
> > > relevant MSRs.
> > >
> > > Suggested-by: Sean Christopherson <seanjc@xxxxxxxxxx>
> > > Signed-off-by: Yang Weijiang <weijiang.yang@xxxxxxxxx>
> > > Tested-by: Mathias Krause <minipli@xxxxxxxxxxxxxx>
> > > Tested-by: John Allen <john.allen@xxxxxxx>
> > > Signed-off-by: Chao Gao <chao.gao@xxxxxxxxx>
> > > ---
> > > arch/x86/kvm/vmx/vmx.c | 20 ++++++++++++++++++++
> > > 1 file changed, 20 insertions(+)
> > >
> > > diff --git a/arch/x86/kvm/vmx/vmx.c b/arch/x86/kvm/vmx/vmx.c
> > > index bd572c8c7bc3..130ffbe7dc1a 100644
> > > --- a/arch/x86/kvm/vmx/vmx.c
> > > +++ b/arch/x86/kvm/vmx/vmx.c
> > > @@ -4088,6 +4088,8 @@ void pt_update_intercept_for_msr(struct kvm_vcpu *vcpu)
> > > void vmx_recalc_msr_intercepts(struct kvm_vcpu *vcpu)
> > > {
> > > + bool set;
> >
> > s/set/intercept
> >
>
> Maybe because you asked me to change "flag" to "set" when reviewing FRED
> patches, however "intercept" does sound better, and I just changed it :)
Ah crud. I had a feeling I was flip-flopping. I obviously don't have a strong
preference.
> > > +
> > > if (!cpu_has_vmx_msr_bitmap())
> > > return;
> > > @@ -4133,6 +4135,24 @@ void vmx_recalc_msr_intercepts(struct kvm_vcpu *vcpu)
> > > vmx_set_intercept_for_msr(vcpu, MSR_IA32_FLUSH_CMD, MSR_TYPE_W,
> > > !guest_cpu_cap_has(vcpu, X86_FEATURE_FLUSH_L1D));
> > > + if (kvm_cpu_cap_has(X86_FEATURE_SHSTK)) {
> > > + set = !guest_cpu_cap_has(vcpu, X86_FEATURE_SHSTK);
> > > +
> > > + vmx_set_intercept_for_msr(vcpu, MSR_IA32_PL0_SSP, MSR_TYPE_RW, set);
> > > + vmx_set_intercept_for_msr(vcpu, MSR_IA32_PL1_SSP, MSR_TYPE_RW, set);
> > > + vmx_set_intercept_for_msr(vcpu, MSR_IA32_PL2_SSP, MSR_TYPE_RW, set);
> > > + vmx_set_intercept_for_msr(vcpu, MSR_IA32_PL3_SSP, MSR_TYPE_RW, set);
> > > + vmx_set_intercept_for_msr(vcpu, MSR_IA32_INT_SSP_TAB, MSR_TYPE_RW, set);
> >
> > MSR_IA32_INT_SSP_TAB isn't managed via XSAVE, so why is it being passed through?
>
> It's managed in VMCS host and guest areas, i.e. HOST_INTR_SSP_TABLE and
> GUEST_INTR_SSP_TABLE, if the "load CET" bits are set in both VM entry
> and exit controls.
Ah, "because it's essentially free". Unless there's a true need to pass it through,
I think it makes sense to intercept. Merging KVM's bitmap with vmcs12's bitmap
isn't completely free (though it's quite cheap). More importantly, this is technically
wrong due to MSR_IA32_INT_SSP_TAB not existing if the vCPU doesn't have LM. That's
obviously easy to solve, I just don't see the point.
