Date:   Tue, 7 Dec 2021 16:34:43 +0000
From:   Sean Christopherson <seanjc@google.com>
To:     Tom Lendacky <thomas.lendacky@amd.com>
Cc:     Marc Orr <marcorr@google.com>, pbonzini@redhat.com,
        vkuznets@redhat.com, wanpengli@tencent.com, jmattson@google.com,
        joro@8bytes.org, tglx@linutronix.de, mingo@redhat.com,
        bp@alien8.de, dave.hansen@linux.intel.com, x86@kernel.org,
        hpa@zytor.com, kvm@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH] KVM: x86: Always set kvm_run->if_flag
Message-ID: <Ya+NIxO5pIkB8057@google.com>
References: <20211207043100.3357474-1-marcorr@google.com>
 <c8889028-9c4e-cade-31b6-ea92a32e4f66@amd.com>
 <CAA03e5E7-ns7w9B9Tu7pSWzCo0Nh7Ba5jwQXcn_XYPf_reRq9Q@mail.gmail.com>
 <5e69c0ca-389c-3ace-7559-edd901a0ab3c@amd.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <5e69c0ca-389c-3ace-7559-edd901a0ab3c@amd.com>
Precedence: bulk

On Tue, Dec 07, 2021, Tom Lendacky wrote:
> On 12/7/21 9:14 AM, Marc Orr wrote:
> > On Tue, Dec 7, 2021 at 6:43 AM Tom Lendacky <thomas.lendacky@amd.com> wrote:
> > > > +static bool svm_get_if_flag(struct kvm_vcpu *vcpu)
> > > > +{
> > > > +     struct vmcb *vmcb = to_svm(vcpu)->vmcb;
> > > > +
> > > > +     return !!(vmcb->control.int_state & SVM_GUEST_INTERRUPT_MASK);
> > > 
> > > I'm not sure if this is always valid to use for non SEV-ES guests. Maybe
> > > the better thing would be:
> > > 
> > >          return sev_es_guest(vcpu->kvm) ? vmcb->control.int_state & SVM_GUEST_INTERRUPT_MASK
> > >                                         : kvm_get_rflags(vcpu) & X86_EFLAGS_IF;
> > > 
> > > (Since this function returns a bool, I don't think you need the !!)
> > 
> > I had the same reservations when writing the patch. (Why fix what's
> > not broken.) The reason I wrote the patch this way is based on what I
> > read in APM vol2: Appendix B Layout of VMCB: "GUEST_INTERRUPT_MASK -
> > Value of the RFLAGS.IF bit for the guest."
> 
> I just verified with the hardware team that this flag is indeed only set for
> a guest with protected state (SEV-ES / SEV-SNP). An update to the APM will
> be made.

svm_interrupt_blocked() should be modified to use the new svm_get_if_flag()
helper so that the SEV-{ES,SN} behavior is contained in a single location, e.g.

diff --git a/arch/x86/kvm/svm/svm.c b/arch/x86/kvm/svm/svm.c
index 208566f63bce..fef04e9fa9c9 100644
--- a/arch/x86/kvm/svm/svm.c
+++ b/arch/x86/kvm/svm/svm.c
@@ -3583,14 +3583,10 @@ bool svm_interrupt_blocked(struct kvm_vcpu *vcpu)
        if (!gif_set(svm))
                return true;

-       if (sev_es_guest(vcpu->kvm)) {
-               /*
-                * SEV-ES guests to not expose RFLAGS. Use the VMCB interrupt mask
-                * bit to determine the state of the IF flag.
-                */
-               if (!(vmcb->control.int_state & SVM_GUEST_INTERRUPT_MASK))
+       if (!is_guest_mode(vcpu)) {
+               if (!svm_get_if_flag(vcpu))
                        return true;
-       } else if (is_guest_mode(vcpu)) {
+       } else {
                /* As long as interrupts are being delivered...  */
                if ((svm->nested.ctl.int_ctl & V_INTR_MASKING_MASK)
                    ? !(svm->vmcb01.ptr->save.rflags & X86_EFLAGS_IF)
@@ -3600,9 +3596,6 @@ bool svm_interrupt_blocked(struct kvm_vcpu *vcpu)
                /* ... vmexits aren't blocked by the interrupt shadow  */
                if (nested_exit_on_intr(svm))
                        return false;
-       } else {
-               if (!(kvm_get_rflags(vcpu) & X86_EFLAGS_IF))
-                       return true;
        }

        return (vmcb->control.int_state & SVM_INTERRUPT_SHADOW_MASK);