Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <670ce789f6139143f781bdd5ebfead79d5a4fadb.camel@redhat.com>
Subject: Re: [PATCH 5/7] KVM: SVM: Add VNMI support in inject_nmi
From:   Maxim Levitsky <mlevitsk@redhat.com>
To:     "Shukla, Santosh" <santosh.shukla@amd.com>,
        Paolo Bonzini <pbonzini@redhat.com>
Cc:     Sean Christopherson <seanjc@google.com>,
        Vitaly Kuznetsov <vkuznets@redhat.com>,
        Jim Mattson <jmattson@google.com>,
        Joerg Roedel <joro@8bytes.org>,
        Tom Lendacky <thomas.lendacky@amd.com>, kvm@vger.kernel.org,
        linux-kernel@vger.kernel.org
Date:   Sun, 10 Jul 2022 19:07:23 +0300
In-Reply-To: <0981fd1c-b4b4-84ad-27e9-babcfa2524db@amd.com>
References: <20220602142620.3196-1-santosh.shukla@amd.com>
         <20220602142620.3196-6-santosh.shukla@amd.com>
         <dfc2ce68a10181b1ac6c07ca3927d474e13ca973.camel@redhat.com>
         <0981fd1c-b4b4-84ad-27e9-babcfa2524db@amd.com>
Content-Type: text/plain; charset="UTF-8"
User-Agent: Evolution 3.36.5 (3.36.5-2.fc32) 
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Precedence: bulk

On Fri, 2022-06-17 at 20:35 +0530, Shukla, Santosh wrote:
> 
> On 6/7/2022 6:44 PM, Maxim Levitsky wrote:
> > On Thu, 2022-06-02 at 19:56 +0530, Santosh Shukla wrote:
> > > Inject the NMI by setting V_NMI in the VMCB interrupt control. processor
> > > will clear V_NMI to acknowledge processing has started and will keep the
> > > V_NMI_MASK set until the processor is done with processing the NMI event.
> > > 
> > > Signed-off-by: Santosh Shukla <santosh.shukla@amd.com>
> > > ---
> > >  arch/x86/kvm/svm/svm.c | 7 ++++++-
> > >  1 file changed, 6 insertions(+), 1 deletion(-)
> > > 
> > > diff --git a/arch/x86/kvm/svm/svm.c b/arch/x86/kvm/svm/svm.c
> > > index a405e414cae4..200f979169e0 100644
> > > --- a/arch/x86/kvm/svm/svm.c
> > > +++ b/arch/x86/kvm/svm/svm.c
> > > @@ -3385,11 +3385,16 @@ static void svm_inject_nmi(struct kvm_vcpu *vcpu)
> > >  {
> > >         struct vcpu_svm *svm = to_svm(vcpu);
> > >  
> > > +       ++vcpu->stat.nmi_injections;
> > > +       if (is_vnmi_enabled(svm->vmcb)) {
> > > +               svm->vmcb->control.int_ctl |= V_NMI_PENDING;
> > > +               return;
> > > +       }
> > Here I would advice to have a warning to check if vNMI is already pending.
> > 
> Yes, in v2.
> 
> > Also we need to check what happens if we make vNMI pending and get #SMI,
> > while in #NMI handler, or if #NMI is blocked due to interrupt window.
> > 
> 
> V_NMI_MASK should be saved as 1 in the save area and
> hypervisor will resume the NMI handler after handling the SMI.


Answering my own question, because I did some homework in this area while
working on that SMI int shadow bug:

Actually what will happen (now I checked) is that we have a special KVM host state flag
(called X86EMUL_SMM_MASK, and HF_SMM_INSIDE_NMI_MASK), and what we do is that if we
receive SMI while in NMI handler, we don't mask the NMI again, on RSM we don't unmask NMI.

Also, as I found out the hard way recently, the NMI is not blocked by the interrupt shadow.

I don't think that anything is saved to the SMRAM in this case.

Best regards,
	Maxim Levitsky

> 
> Thanks,
> Santosh
>  
> > Best regards,
> > 	Maxim Levitsky
> > 
> > 
> > > +
> > >         svm->vmcb->control.event_inj = SVM_EVTINJ_VALID | SVM_EVTINJ_TYPE_NMI;
> > >         vcpu->arch.hflags |= HF_NMI_MASK;
> > >         if (!sev_es_guest(vcpu->kvm))
> > >                 svm_set_intercept(svm, INTERCEPT_IRET);
> > > -       ++vcpu->stat.nmi_injections;
> > >  }
> > >  
> > >  static void svm_inject_irq(struct kvm_vcpu *vcpu)