Re: [Qemu-devel] [PATCH v3 2/2] i386/kvm: lower requirements for Hyper-V frequency MSRs exposure

From: Roman Kagan <rkagan@virtuozzo.com>
To: Paolo Bonzini <pbonzini@redhat.com>
Cc: Vitaly Kuznetsov <vkuznets@redhat.com>,
	Richard Henderson <rth@twiddle.net>,
	Eduardo Habkost <ehabkost@redhat.com>,
	Marcelo Tosatti <mtosatti@redhat.com>,
	qemu-devel@nongnu.org
Subject: Re: [Qemu-devel] [PATCH v3 2/2] i386/kvm: lower requirements for Hyper-V frequency MSRs exposure
Date: Wed, 21 Mar 2018 19:47:08 +0300	[thread overview]
Message-ID: <20180321164707.GG14983@rkaganb.sw.ru> (raw)
In-Reply-To: <46c08ea1-27f7-e581-f93f-cf635950062f@redhat.com>

On Wed, Mar 21, 2018 at 04:33:33PM +0100, Paolo Bonzini wrote:
> On 20/03/2018 18:35, Vitaly Kuznetsov wrote:
> > +        if (has_msr_hv_frequencies && env->tsc_khz) {
> >              env->features[FEAT_HYPERV_EAX] |= HV_ACCESS_FREQUENCY_MSRS;
> >              env->features[FEAT_HYPERV_EDX] |= HV_FREQUENCY_MSRS_AVAILABLE;
> >          }
> 
> Since you have added cpu->hyperv_reenlightenment, I'd rather change this
> so that we don't make the "license to change guest ABI across migration"
> apply more.  We can exploit the fact that Windows doesn't even use the
> MSRs unless either invtsc or re-enlightenment is present.

That's not a given.  It reportedly doesn't use these MSRs *now*, but I
see no reason for it not to start using it at some point, say, to avoid
TSC calibration.  E.g. Linux already does so with its hv_get_tsc_khz().

And the bad thing is that it'll kill your guest with #GP when you
migrate from a new KVM to an old one.

Roman.