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;
MIME-Version: 1.0
References: <20220617195141.2866706-1-pgonda@google.com> <Yqz+CZlGCoQo7lMQ@google.com>
In-Reply-To: <Yqz+CZlGCoQo7lMQ@google.com>
From:   Peter Gonda <pgonda@google.com>
Date:   Wed, 22 Jun 2022 13:33:32 -0600
Message-ID: <CAMkAt6p6U6G0N0UP-Dnatr1cLHnaU1DY-1gxvmYhvxjckhbx4g@mail.gmail.com>
Subject: Re: [PATCH] KVM: SEV: Init target VMCBs in sev_migrate_from
To:     Sean Christopherson <seanjc@google.com>
Cc:     kvm list <kvm@vger.kernel.org>, Marc Orr <marcorr@google.com>,
        Paolo Bonzini <pbonzini@redhat.com>,
        Tom Lendacky <thomas.lendacky@amd.com>,
        LKML <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

> >
> > +void sev_init_vmcb(struct vcpu_svm *svm)
> > +{
> > +     svm->vmcb->control.nested_ctl |= SVM_NESTED_CTL_SEV_ENABLE;
> > +     clr_exception_intercept(svm, UD_VECTOR);
>
> I don't love separating SEV and SEV-ES VMCB initialization, especially since they're
> both doing RMW operations and not straight writes.  E.g. migration ends up reversing
> the order between the two relatively to init_vmcb().  That's just asking for a subtle
> bug to be introduced that affects only due to the ordering difference.
>
> What about using common top-level flows for SEV and SEV-ES so that the sequencing
> between SEV and SEV-ES is more rigid?  The resulting sev_migrate_from() is a little
> gross, but IMO it's worth having a fixed sequence, and the flip side to the ugliness
> it that it documents some of the differences between SEV and SEV-ES migration.

Thanks for the suggestion Sean! I like your suggestion here. I'll test
it out, clean it up and send it out as V2. I think the distinction
between SEV and SEV-ES migration was largely due to how I split up the
set of patches that enabled this feature.