Re: [PATCH 2/4] KVM: VMX: Unconditionally clear CPUID.INVPCID if !CPUID.PCID

[Paolo Bonzini <pbonzini@redhat.com>]

On 23/09/20 19:15, Jim Mattson wrote:
> On Wed, Sep 23, 2020 at 9:51 AM Sean Christopherson
> <sean.j.christopherson@intel.com> wrote:
>>
>> If PCID is not exposed to the guest, clear INVPCID in the guest's CPUID
>> even if the VMCS INVPCID enable is not supported.  This will allow
>> consolidating the secondary execution control adjustment code without
>> having to special case INVPCID.
>>
>> Technically, this fixes a bug where !CPUID.PCID && CPUID.INVCPID would
>> result in unexpected guest behavior (#UD instead of #GP/#PF), but KVM
>> doesn't support exposing INVPCID if it's not supported in the VMCS, i.e.
>> such a config is broken/bogus no matter what.
>>
>> Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
>> ---
>>  arch/x86/kvm/vmx/vmx.c | 16 +++++++++++-----
>>  1 file changed, 11 insertions(+), 5 deletions(-)
>>
>> diff --git a/arch/x86/kvm/vmx/vmx.c b/arch/x86/kvm/vmx/vmx.c
>> index cfed29329e4f..57e48c5a1e91 100644
>> --- a/arch/x86/kvm/vmx/vmx.c
>> +++ b/arch/x86/kvm/vmx/vmx.c
>> @@ -4149,16 +4149,22 @@ static void vmx_compute_secondary_exec_control(struct vcpu_vmx *vmx)
>>                 }
>>         }
>>
>> +       /*
>> +        * Expose INVPCID if and only if PCID is also exposed to the guest.
>> +        * INVPCID takes a #UD when it's disabled in the VMCS, but a #GP or #PF
>> +        * if CR4.PCIDE=0.  Enumerating CPUID.INVPCID=1 would lead to incorrect
>> +        * behavior from the guest perspective (it would expect #GP or #PF).
>> +        */
>> +       if (!guest_cpuid_has(vcpu, X86_FEATURE_PCID))
>> +               guest_cpuid_clear(vcpu, X86_FEATURE_INVPCID);
>> +
> 
> I thought the general rule for userspace provided CPUID bits was that
> kvm only made any adjustments necessary to prevent bad things from
> happening at the host level. Proper guest semantics are entirely the
> responsibility of userspace. Or did I misunderstand?
> 

Yes, that's generally the idea.  INVPCID has always been a bit special
due to the secondary execution control being of the "enable" kind; this
led the original author to try and disable the instruction (which is by
itself something we do not always do, and sometimes cannot always do).

So I agree that Sean's patch is of marginal utility by itself; however
it lets him use the new macros in patch 4 and it is a good idea to
separate the small functional change into its own commit.

Paolo