linux-kernel - Re: [PATCH 3/7] KVM: SVM: Add VNMI support in get/set_nmi

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <a0970f50d5de92917dee44c6d23dce87ad49d862.camel@redhat.com>
Date:   Sun, 10 Jul 2022 19:09:08 +0300
From:   Maxim Levitsky <mlevitsk@...hat.com>
To:     "Shukla, Santosh" <santosh.shukla@....com>,
        Paolo Bonzini <pbonzini@...hat.com>
Cc:     Sean Christopherson <seanjc@...gle.com>,
        Vitaly Kuznetsov <vkuznets@...hat.com>,
        Jim Mattson <jmattson@...gle.com>,
        Joerg Roedel <joro@...tes.org>,
        Tom Lendacky <thomas.lendacky@....com>, kvm@...r.kernel.org,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH 3/7] KVM: SVM: Add VNMI support in get/set_nmi_mask

On Fri, 2022-06-17 at 20:18 +0530, Shukla, Santosh wrote:
> 
> On 6/17/2022 8:15 PM, Shukla, Santosh wrote:
> > 
> > On 6/7/2022 6:37 PM, Maxim Levitsky wrote:
> > > On Thu, 2022-06-02 at 19:56 +0530, Santosh Shukla wrote:
> > > > VMCB intr_ctrl bit12 (V_NMI_MASK) is set by the processor when handling
> > > > NMI in guest and is cleared after the NMI is handled. Treat V_NMI_MASK as
> > > > read-only in the hypervisor and do not populate set accessors.
> > > > 
> > > > Signed-off-by: Santosh Shukla <santosh.shukla@....com>
> > > > ---
> > > >  arch/x86/kvm/svm/svm.c | 20 +++++++++++++++++++-
> > > >  1 file changed, 19 insertions(+), 1 deletion(-)
> > > > 
> > > > diff --git a/arch/x86/kvm/svm/svm.c b/arch/x86/kvm/svm/svm.c
> > > > index 860f28c668bd..d67a54517d95 100644
> > > > --- a/arch/x86/kvm/svm/svm.c
> > > > +++ b/arch/x86/kvm/svm/svm.c
> > > > @@ -323,6 +323,16 @@ static int is_external_interrupt(u32 info)
> > > >         return info == (SVM_EVTINJ_VALID | SVM_EVTINJ_TYPE_INTR);
> > > >  }
> > > >  
> > > > +static bool is_vnmi_enabled(struct vmcb *vmcb)
> > > > +{
> > > > +       return vnmi && (vmcb->control.int_ctl & V_NMI_ENABLE);
> > > > +}
> > > 
> > > Following Paolo's suggestion I recently removed vgif_enabled(),
> > > based on the logic that vgif_enabled == vgif, because
> > > we always enable vGIF for L1 as long as 'vgif' module param is set,
> > > which is set unless either hardware or user cleared it.
> > > 
> > Yes. In v2, Thanks!.
> > 
> > > Note that here vmcb is the current vmcb, which can be vmcb02,
> > > and it might be wrong
> > > 
> > > > +
> > > > +static bool is_vnmi_mask_set(struct vmcb *vmcb)
> > > > +{
> > > > +       return !!(vmcb->control.int_ctl & V_NMI_MASK);
> > > > +}
> > > > +
> > > >  static u32 svm_get_interrupt_shadow(struct kvm_vcpu *vcpu)
> > > >  {
> > > >         struct vcpu_svm *svm = to_svm(vcpu);
> > > > @@ -3502,13 +3512,21 @@ static int svm_nmi_allowed(struct kvm_vcpu *vcpu, bool for_injection)
> > > >  
> > > >  static bool svm_get_nmi_mask(struct kvm_vcpu *vcpu)
> > > >  {
> > > > -       return !!(vcpu->arch.hflags & HF_NMI_MASK);
> > > > +       struct vcpu_svm *svm = to_svm(vcpu);
> > > > +
> > > > +       if (is_vnmi_enabled(svm->vmcb))
> > > > +               return is_vnmi_mask_set(svm->vmcb);
> > > > +       else
> > > > +               return !!(vcpu->arch.hflags & HF_NMI_MASK);
> > > >  }
> > > >  
> > > >  static void svm_set_nmi_mask(struct kvm_vcpu *vcpu, bool masked)
> > > >  {
> > > >         struct vcpu_svm *svm = to_svm(vcpu);
> > > >  
> > > > +       if (is_vnmi_enabled(svm->vmcb))
> > > > +               return;
> > > 
> > > What if the KVM wants to mask NMI, shoudn't we update the 
> > > V_NMI_MASK value in int_ctl instead of doing nothing?
> > > 
> 
> V_NMI_MASK is cpu controlled meaning HW sets the mask while processing
> event and clears right after processing, so in away its Read-only for hypervisor.

And yet, svm_set_nmi_mask is called when KVM wants to explicitly mask NMI
without injecting a NMI, it does this when entering (emulated) SMI.

So the KVM has to set V_NMI_MASK here, becaue no real NMI is injected,
and thus the CPU will not set this bit itself.

Best regards,
	Maxim Levitsky
> 
> > > Best regards,
> > > 	Maxim Levitsky
> > > 
> > > 
> > > > +
> > > >         if (masked) {
> > > >                 vcpu->arch.hflags |= HF_NMI_MASK;
> > > >                 if (!sev_es_guest(vcpu->kvm))