On 02.08.2022 18:08, Dario Faggioli wrote:
> On Tue, 2022-08-02 at 16:56 +0200, Jan Beulich wrote:
>> On 02.08.2022 15:51, Dario Faggioli wrote:
>>> If dom0_vcpus_pin is used, make sure the pinning is only done for
>>> dom0 vcpus, instead of for the hardware domain (which might not be
>>> dom0 at all!).
>>
>> Hmm, but the control domain may not be either, as it's derived from
>> d->is_privileged. I think ...
>>
> Mmm... Right.
> 
>>> --- a/xen/common/sched/core.c
>>> +++ b/xen/common/sched/core.c
>>> @@ -575,7 +575,7 @@ int sched_init_vcpu(struct vcpu *v)
>>>       * Initialize affinity settings. The idler, and potentially
>>>       * domain-0 VCPUs, are pinned onto their respective physical
>>> CPUs.
>>>       */
>>> -    if ( is_idle_domain(d) || (is_hardware_domain(d) &&
>>> opt_dom0_vcpus_pin) )
>>> +    if ( is_idle_domain(d) || (is_control_domain(d) &&
>>> opt_dom0_vcpus_pin) )
>>
>> ... for it to be strictly only Dom0, you want to check d->domain_id
>> here.
>>
> Ok, I'll send an update that does that.

Well - if you agree, I'd be happy to make the change while committing
and then adding
Reviewed-by: Jan Beulich <[email protected]>

Jan

Reply via email to