On Mon, Mar 01, 2021 at 05:16:34PM +0000, Ian Jackson wrote:
> Roger Pau Monne writes ("[PATCH RFC for-4.15] x86/msr: introduce an option
> for legacy MSR behavior selection"):
> > Introduce an option to allow selecting the legacy behavior for
> > accesses to MSRs not explicitly handled. Since commit
> > 84e848fd7a162f669 and 322ec7c89f6640e accesses to MSRs not explicitly
> > handled by Xen result in the injection of a #GP to the guest. This is
> > a behavior change since previously a #GP was only injected if
> > accessing the MSR on the real hardware will also trigger a #GP.
> >
> > This seems to be problematic for some guests, so introduce an option
> > to fallback to this legacy behavior. The main difference between what
> > was previously done is that the hardware MSR value is not leaked to
> > the guests on reads.
> >
> > Signed-off-by: Roger Pau Monné <roger....@citrix.com>
> > ---
> > Cc: Boris Ostrovsky <boris.ostrov...@oracle.com>
> > ---
> > Note that this option is not made available to dom0. I'm not sure
> > whether it makes sense to do so, since anyone updating Xen to such
> > newer version will also likely pair it with a newish kernel that
> > doesn't require such workarounds.
> >
> > RFC because there's still some debate as to how we should solve the
> > MSR issue, this is one possible way, but IMO we need to make a
> > decision soon-ish because of the release timeline.
> >
> > Boris, could you please test with Solaris to see if this fixes the
> > issue?
>
> So AIUI this patch is to make it possible for Xen 4.15 to behave like
> Xen 4.14, thus avoiding a regression for these troublesome guests.
Yes, sorry I haven't provided a release executive summary, as I wasn't
sure this would be acceptable in it's current form. Can do if there's
consensus this is an acceptable fix.
> Have we diffed the result of this against 4.14 and if not would it be
> a sensible thing to do ?
I think there will likely bee too much noise, we have changed the MSR
handling a bit from 4.14, so it's likely a diff to 4.14 is not going
to be helpful as the context will have too many changes (albeit I
haven't tried the exercise myself).
Thanks, Roger.
