Re: [PATCH 3/3] x86: Use CpuidUserDis if an AMD HVM guest toggles CPUID faulting

[Alejandro Vallejo <alejandro.vallejo@cloud.com>]

On Tue, May 09, 2023 at 04:41:49PM +0200, Jan Beulich wrote:
> > I asked Alejandro to do it like this.
> > 
> > Advertising this to guests requires plumbing another MSR into the
> > infrastructure which isn't quite set up properly let, and is in flux
> > from my work.
> > 
> > For now, this just lets Xen enforce the policy over PV guests, which is
> > an improvement in and of itself.
> 
> But as per the title this patch is about HVM guests (aiui the PV aspect
> is taken care of already without the patch here). In any event - if the
> omissions are intentional (for the time being), then I think that wants
> mentioning in the description.
> 
> Jan

HVM guests are always exposed the Intel interface (emulated if not natively
available). The HVM max policy forces it on, and I don't see anything in
the default policy overriding it. My attempt here was to let AMD guests use
the emulated Intel MSR and trigger levellling that would itself rely on
AMD's CpuidUserDis without guest intervention. That said, several cans of
worms exist in mantaining this internal routing. I'll get rid of that last
patch and leave HVM guests alone for the time being. They are functionally 
correct (albeit their CPUID take 2 faults, whereas 1 would suffice).

Alejandro