On Mon, Feb 07, 2022 at 06:34:19PM +0000, Sean Christopherson wrote:
> On Mon, Feb 07, 2022, Oliver Upton wrote:
> > Until recently, this all sort of 'worked'. Since we called
> > kvm_update_cpuid() all the time it was possible for KVM to overwrite the
> > bits after the MSR write, just not immediately so. After the whole CPUID
> > rework, we only update the VMX control MSRs immediately after a
> > KVM_SET_CPUID2, meaning we've missed the case of MSR write after CPUID.
>
> That needs to be explained in the changelog (ditto for patch 02), and arguably
> the Fixes tag is wrong too, or at least incomplete. The commit that truly broke
> things was
>
> aedbaf4f6afd ("KVM: x86: Extract kvm_update_cpuid_runtime() from kvm_update_cpuid()")
>
> I'm guessing this is why Paolo is also confused. Without understanding that KVM
> used too (eventually) enforce its overrides, it looks like you're proposing an
> arbitrary, unnecessary ABI change.
Gah, sorry, I really didn't provide the full context on this. I chose to
blame the original commits for these since it was still possible to
write the MSR and avoid a KVM update (just looking for paths where
kvm_update_cpuid() is not called), but agree that full breakage came
from the above commit.
I'll add some language discussing how commit aedbaf4f6afd ("KVM: x86: Extract
kvm_update_cpuid_runtime() from kvm_update_cpuid()") fully broke this.
--
Thanks,
Oliver
