Re: [PATCH v2 2/3] x86/vpmu: Disable VPMU if guest's CPUID indicates no PMU support

[Boris Ostrovsky <boris.ostrovsky@oracle.com>]

On 02/22/2017 09:38 AM, Jan Beulich wrote:
>>>> On 22.02.17 at 15:15, <boris.ostrovsky@oracle.com> wrote:
>> On 02/22/2017 04:55 AM, Jan Beulich wrote:
>>>>>> On 17.02.17 at 18:40, <boris.ostrovsky@oracle.com> wrote:
>>>> --- a/xen/arch/x86/cpu/vpmu_intel.c
>>>> +++ b/xen/arch/x86/cpu/vpmu_intel.c
>>>> @@ -884,6 +884,10 @@ int vmx_vpmu_initialise(struct vcpu *v)
>>>>      if ( vpmu_mode == XENPMU_MODE_OFF )
>>>>          return 0;
>>>>  
>>>> +    if ( MASK_EXTR(v->domain->arch.cpuid->basic.raw[0xa].a, 
>>>> +                   PMU_VERSION_MASK) == 0 )
>>>> +        return -EINVAL;
>>> How about other unsupported (too large) values?
>> Yes, we can check here for version >=5 as well.
>>
>> (I don't think we should make this additional test in
>> update_domain_cpuid_info())
> ... because of ... ? After all it's the purpose of this patch to not
> expose the vPMU in such cases, which imo ought to be done
> consistently in both places.

Because I felt that having zero as a version is an indication of
explicit admin's desire to disable VPMU. Too high a version is more
likely misconfiguration and can be taken care of during VPMU initialization.

-boris


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel