Re: [PATCH] x86/tdx: Fix zero-extension for CPUID emulation

[Edgecombe, Rick P]

On Tue, 2026-05-12 at 15:14 -0700, Dave Hansen wrote:
> The end result is that a TDX guest can use CPUID and end up having bits
> set in rax/rbx/rcx/rdx that are architecturally impossible. This patch
> is effectively fixing up the VMM naughtiness before the guest CPUID
> instance can see it.

A naughty VMM could mess with the guest in a number of ways though. For example
setting impossible bits in specific leafs in the lower 32 bits. This patch is a
relatively simple sanity check compared to a complete check of CPUID arch
matching (or MSR, etc) of course.

> 
> Does anybody disagree with any of that?
> 
> Do we *want* to fix this up silently? If we catch a malicious VMM trying
> to stuff garbage into the guest, shouldn't we be a bit more upset than
> silently papering over it?

I agree a warning would be appropriate. This should probably trigger a bug fix
in the VMM. For example, BIOS might hit it too. So I kind of wonder, how
valuable is catching this specific bug in the guest? Do we need to worry about
the specific issue for some reason?

On the other hand, the #VE handler is supposed to do the emulation of the
instruction, with the help of the TDVMCALL, so maybe the correctness should be
in the guest... Hmm...