Re: [PATCH 25/25] KVM: x86: Add CPUID bits missing from KVM_GET_SUPPORTED_CPUID

[Sean Christopherson <seanjc@google.com>]

On Wed, Aug 14, 2024, Rick P Edgecombe wrote:
> On Wed, 2024-08-14 at 06:35 -0700, Sean Christopherson wrote:
> > > One scenario where "fixed-1" bits can help is: we discover a security issue
> > > and
> > > release a microcode update to expose a feature indicating which CPUs are
> > > vulnerable. if the TDX module allows the VMM to configure the feature as 0
> > > (i.e., not vulnerable) on vulnerable CPUs, a TD might incorrectly assume
> > > it's
> > > not vulnerable, creating a security issue.
> > > 
> > > I think in above case, the TDX module has to add a "fixed-1" bit. An example
> > > of
> > > such a feature is RRSBA in the IA32_ARCH_CAPABILITIES MSR.
> > 
> > That would be fine, I would classify that as reasonable.  However, that
> > scenario
> > doesn't really work in practice, at least not the way Intel probably hopes it
> > plays out.  For the new fixed-1 bit to provide value, it would require a guest
> > reboot and likely a guets kernel upgrade.
> 
> If we allow "reasonable" fixed bits, we need to decide how to handle any that
> KVM sees but doesn't know about. Not filtering them is simpler to implement.
> Filtering them seems a little more controlled to me.
> 
> It might depend on how reasonable, "reasonable" turns out. Maybe we give not
> filtering a try and see how it goes. If we run into a problem, we can filter new
> bits from that point, and add a quirk for whatever the issue is. I'm still on
> the fence.

As I see it, it's ultimately unlikely to be KVM's problem.  If Intel ships a
TDX-Module that does bad things, and someone's setup breaks when they upgrade to
that TDX-Module, then their gripe is with Intel.  KVM can't do anything to remedy
the problem.

If the upgrade breaks a setup because it confuses _KVM_, then I'll care, but I
suspect/hope that won't happen in practice, purely because KVM has so little
visiblity into the guest, i.e. doesn't care what is/isn't advertised to the guest.

FWIW, AMD has effectively gone the "fixed-1" route for a few things[*], e.g. KVM
can't intercept XCR0 or XSS writes.  And while I detest the behavior, I haven't
refused to merge support for SEV-ES+.  I just grumble every time it comes up :-)

[*] https://lore.kernel.org/all/ZUQvNIE9iU5TqJfw@google.com