Re: [PATCH 8/9] x86/cpuid: Introduce missing feature adjustment in calculate_pv_def_policy()

[Andrew Cooper <andrew.cooper3@xxxxxxxxxx>]

On 16/06/2020 10:40, Jan Beulich wrote:
> On 15.06.2020 16:15, Andrew Cooper wrote:
>> This was an accidental asymmetry with the HVM side.
>>
>> No change in behaviour at this point.
>>
>> Fixes: 83b387382 ("x86/cpuid: Introduce and use default CPUID policies")
>> Signed-off-by: Andrew Cooper <andrew.cooper3@xxxxxxxxxx>
> Reviewed-by: Jan Beulich <jbeulich@xxxxxxxx>

Thanks.

> with a remark:
>
>> --- a/xen/arch/x86/cpuid.c
>> +++ b/xen/arch/x86/cpuid.c
>> @@ -402,6 +402,8 @@ static void __init calculate_pv_def_policy(void)
>>      for ( i = 0; i < ARRAY_SIZE(pv_featureset); ++i )
>>          pv_featureset[i] &= pv_def_featuremask[i];
>>  
>> +    guest_common_feature_adjustments(pv_featureset);
>> +
>>      sanitise_featureset(pv_featureset);
>>      cpuid_featureset_to_policy(pv_featureset, p);
>>      recalculate_xstate(p);
> These four calls are common to all three callers of the function.
> Perhaps them going out of sync would be less likely if all four
> called the same helper to carry out these four steps?

I'm not sure how many of them are going to survive the transformation to
a fully libx86 based world.

I expect it not to look exactly like this.

~Andrew