Re: [PATCH] KVM: x86/cpuid: Clear XFD for component i if the base feature is missing

[Paolo Bonzini <pbonzini@redhat.com>]

On 1/18/22 07:43, Like Xu wrote:
> On 18/1/2022 1:31 am, Paolo Bonzini wrote:
>> On 1/17/22 08:45, Like Xu wrote:
>>> From: Like Xu <likexu@tencent.com>
>>>
>>> According to Intel extended feature disable (XFD) spec, the 
>>> sub-function i
>>> (i > 1) of CPUID function 0DH enumerates "details for state component i.
>>> ECX[2] enumerates support for XFD support for this state component."
>>>
>>> If KVM does not report F(XFD) feature (e.g. due to CONFIG_X86_64),
>>> then the corresponding XFD support for any state component i
>>> should also be removed. Translate this dependency into KVM terms.
>>>
>>> Fixes: 690a757d610e ("kvm: x86: Add CPUID support for Intel AMX")
>>> Signed-off-by: Like Xu <likexu@tencent.com>
>>> ---
>>>   arch/x86/kvm/cpuid.c | 3 +++
>>>   1 file changed, 3 insertions(+)
>>>
>>> diff --git a/arch/x86/kvm/cpuid.c b/arch/x86/kvm/cpuid.c
>>> index c55e57b30e81..e96efef4f048 100644
>>> --- a/arch/x86/kvm/cpuid.c
>>> +++ b/arch/x86/kvm/cpuid.c
>>> @@ -886,6 +886,9 @@ static inline int __do_cpuid_func(struct 
>>> kvm_cpuid_array *array, u32 function)
>>>                   --array->nent;
>>>                   continue;
>>>               }
>>> +
>>> +            if (!kvm_cpu_cap_has(X86_FEATURE_XFD))
>>> +                entry->ecx &= ~BIT_ULL(2);
>>>               entry->edx = 0;
>>>           }
>>>           break;
>>
>> Generally this is something that is left to userspace.  Apart from the 
>> usecase of "call KVM_GET_SUPPORTED_CPUID and pass it to 
>> KVM_SET_CPUID2", userspace should know what any changed bits mean.
>>
>> Paolo
>>
> 
> I totally agree that setting the appropriate CPUID bits for a feature is 
> a user space thing.
> 
> But this patch is more focused on fixing a different type of problem, 
> which is
> that the capabilities exposed by KVM should not *contradict each other* :
> 
>      a user space may be confused with the current code base :
>      - why KVM does not have F(XFD) feature (MSR_IA32_XFD and XFD_ERR 
> non-exit),
>      - but KVM reports XFD support for state component i individually;

Got it.  Yeah, the patch makes sense for the sake of CONFIG_X86_64.

Paolo

> This is like KVM reporting PEBS when no PMU capacity is reported (due to 
> module param).