Re: Ping²: [PATCH] x86/PV: conditionally avoid raising #GP for early guest MSR accesses

[Jürgen Groß <jgross@xxxxxxxx>]

On 05.02.21 14:00, Jan Beulich wrote:
> On 05.02.2021 11:56, Jürgen Groß wrote:
> > As we need to consider backports of processor bug mitigations
> > in old guests, too, I think we need to have a "catch-all"
> > fallback.
> >
> > Not being able to run an old updated guest until we add handling
> > for a new MSR isn't a viable option IMO.
> I'm not sure I follow you here: Such backports should still make
> use of the respective CPUID bits, and hence shouldn't contain
> "blind" MSR accesses. And if there's really something needing to
> probe an MSR, then I'd expect such a backport to make sure the
> probing actually works in a prereq (presumably) backport.
We know that Linux partially relies on blind MSR accesses to be
tolerated on bare metal. With this and older Xen versions having
worked by _not_ faulting when the guest accesses illegal MSRs
it is not that unlikely that some guests will suddenly fail to
boot.

Especially when backporting processor bug mitigations I could
envision not all distros will thoroughly test the kernel to still
run as PV guest on newest Xen.


Juergen

Attachment: OpenPGP_0xB0DE9DD628BF132F.asc
Description: application/pgp-keys

Attachment: OpenPGP_signature
Description: OpenPGP digital signature