linux-kernel - Re: [PATCH v2 2/2] KVM: x86: add MSR_KVM_MIGRATION

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]

Date:   Wed, 28 Apr 2021 20:06:39 +0000
From:   Sean Christopherson <seanjc@...gle.com>
To:     Steve Rutherford <srutherford@...gle.com>
Cc:     Paolo Bonzini <pbonzini@...hat.com>,
        LKML <linux-kernel@...r.kernel.org>,
        KVM list <kvm@...r.kernel.org>, Joerg Roedel <joro@...tes.org>,
        Brijesh Singh <brijesh.singh@....com>,
        Tom Lendacky <thomas.lendacky@....com>,
        Venu Busireddy <venu.busireddy@...cle.com>
Subject: Re: [PATCH v2 2/2] KVM: x86: add MSR_KVM_MIGRATION_CONTROL

On Wed, Apr 28, 2021, Steve Rutherford wrote:
> On Tue, Apr 27, 2021 at 3:14 PM Sean Christopherson <seanjc@...gle.com> wrote:
> >
> > On Wed, Apr 21, 2021, Paolo Bonzini wrote:
> > > Add a new MSR that can be used to communicate whether the page
> > > encryption status bitmap is up to date and therefore whether live
> > > migration of an encrypted guest is possible.
> > >
> > > The MSR should be processed by userspace if it is going to live
> > > migrate the guest; the default implementation does nothing.
> > >
> > > Signed-off-by: Paolo Bonzini <pbonzini@...hat.com>
> > > ---
> >
> > ...
> >
> > > @@ -91,6 +93,8 @@ struct kvm_clock_pairing {
> > >  /* MSR_KVM_ASYNC_PF_INT */
> > >  #define KVM_ASYNC_PF_VEC_MASK                        GENMASK(7, 0)
> > >
> > > +/* MSR_KVM_MIGRATION_CONTROL */
> > > +#define KVM_PAGE_ENC_STATUS_UPTODATE         (1 << 0)
> >
> > Why explicitly tie this to encryption status?  AFAICT, doing so serves no real
> > purpose and can only hurt us in the long run.  E.g. if a new use case for
> > "disabling" migration comes along and it has nothing to do with encryption, then
> > it has the choice of either using a different bit or bastardizing the existing
> > control.
> >
> > I've no idea if such a use case is remotely likely to pop up, but allowing for
> > such a possibility costs us nothing.
>
> Using a different bit sounds fine to me. It would allow us to avoid
> stuffing multiple meanings into a single bit, which would still happen
> even if we had a better name.

But there's only multiple meanings if we define the bit to be specific to
page encryption.  E.g. if the bit is KVM_READY_FOR_MIGRATION, then its meaning
(when cleared) is simply "please don't migrate me, I will die".  KVM doesn't
care _why_ the guest is telling userspace that it's not ready for migration, nor
does KVM care if userspace honors the indicator.