On Tue Feb 10, 2026 at 9:19 AM CET, Roger Pau Monné wrote:
> On Fri, Feb 06, 2026 at 05:15:23PM +0100, Alejandro Vallejo wrote:
>> While in principle it's possible to have a vendor virtualising another,
>> this is fairly tricky in practice. Not doing so enables certain
>> optimisations with regards to vendor checks in later patches.
>> 
>> Signed-off-by: Alejandro Vallejo <[email protected]>
>> ---
>> Patch 1 from the cross-vendor series. Do not merge here. It's simply for
>> consistency.
>> ---
>>  xen/lib/x86/policy.c | 3 ++-
>>  1 file changed, 2 insertions(+), 1 deletion(-)
>> 
>> diff --git a/xen/lib/x86/policy.c b/xen/lib/x86/policy.c
>> index f033d22785..079c42a29b 100644
>> --- a/xen/lib/x86/policy.c
>> +++ b/xen/lib/x86/policy.c
>> @@ -15,7 +15,8 @@ int x86_cpu_policies_are_compatible(const struct 
>> cpu_policy *host,
>>  #define FAIL_MSR(m) \
>>      do { e.msr = (m); goto out; } while ( 0 )
>>  
>> -    if ( guest->basic.max_leaf > host->basic.max_leaf )
>> +    if ( (guest->x86_vendor     != host->x86_vendor) ||
>> +         (guest->basic.max_leaf >  host->basic.max_leaf) )
>
> You possibly want to expand test-cpu-policy.c to add a small test to
> ensure this works as expected?  Not that it shouldn't, but it's
> trivial to expand test_is_compatible_{success,failure}() to add a
> small test for the vendor checking.
>
> Thanks, Roger.

I didn't consider it. I guess I could. What I'm thinking also is that this
vendor check should probably check the encoded vendor (ebcx/ecx/edx) rather
than the decoded one.

Cheers,
Alejandro

Reply via email to