Re: [PATCH 2/5] xen/sched: rtds: enforce admission control in xl sched-rtds

[Furkan Çalışkan <frn1furkan10@xxxxxxxxx>]

On 9/14/26 12:16, Juergen Gross wrote:
> On 26.08.26 06:57, Furkan Caliskan wrote:
>> Now that we have introduced admission control for new and
>> removed units. Extend it to XEN_DOMCTL_SCHEDOP_putinfo and
>> putvcpuinfo, so growing an existing reservation via
>> xl sched-rtds is checked too.
>>
>> putinfo sets the same (period, budget) for every unit of a
>> domain at once, so it tests the whole domain's utilization
>> delta atomically, rather than unit-by-unit, which could
>> spuriously reject an overall-acceptable change depending on
>> iteration order.
>>
>> putvcpuinfo changes one unit at a time, so it just calls
>> rt_admission_test() directly.
>>
>> Signed-off-by: Furkan Caliskan <frn1furkan10@xxxxxxxxx>
>> ---
>>   xen/common/sched/rt.c | 42 ++++++++++++++++++++++++++++++++++++++++++
>>   1 file changed, 42 insertions(+)
>>
>> diff --git a/xen/common/sched/rt.c b/xen/common/sched/rt.c
>> index 9126320801..9643a277fe 100644
>> --- a/xen/common/sched/rt.c
>> +++ b/xen/common/sched/rt.c
>> @@ -1527,11 +1527,43 @@ rt_dom_cntl(
>>           op->u.rtds.budget = RTDS_DEFAULT_BUDGET / MICROSECS(1);
>>           break;
>>       case XEN_DOMCTL_SCHEDOP_putinfo:
>> +    {
>> +        uint64_t dom_old_util = 0, new_util, new_total;
>> +        unsigned int nr_units = 0;
>> +
>>           rc = rt_validate_params(&op->u.rtds, &period, &budget);
>>           if ( rc )
>>               break;
>>   +        new_util = rt_unit_utilization(period, budget);
>> +
>>           spin_lock_irqsave(&prv->lock, flags);
>> +
>> +        /*
>> +         * Same (period, budget) for every unit of d: test and commit
>> +         * the domain's whole utilization delta atomically, rather
>> +         * than unit-by-unit, which could spuriously reject an
>> +         * overall-acceptable change depending on iteration order.
>> +         */
>> +        for_each_sched_unit ( d, unit )
>> +        {
>> +            svc = rt_unit(unit);
>> +            dom_old_util += rt_unit_utilization(svc->period, svc->budget);
>> +            nr_units++;
>> +        }
>> +
>> +        new_total = prv->utilization - dom_old_util +
>> +                    (uint64_t)nr_units * new_util;
>> +
>> +        if ( new_total > prv->utilization && new_total > 
>> rt_utilization_cap(d) )
>> +        {
>> +            rc = -EINVAL;
> 
> Is EINVAL a good choice here? I think a different error code might be wanted.

Would EBUSY be okay here? Or what else would you suggest?

> 
>> +            spin_unlock_irqrestore(&prv->lock, flags);
>> +            break;
>> +        }
>> +
>> +        prv->utilization = new_total;
>> +
>>           for_each_sched_unit ( d, unit )
>>           {
>>               svc = rt_unit(unit);
>> @@ -1540,6 +1572,7 @@ rt_dom_cntl(
>>           }
>>           spin_unlock_irqrestore(&prv->lock, flags);
>>           break;
>> +    }
>>       case XEN_DOMCTL_SCHEDOP_getvcpuinfo:
>>       case XEN_DOMCTL_SCHEDOP_putvcpuinfo:
>>           while ( index < op->u.v.nr_vcpus )
>> @@ -1584,6 +1617,15 @@ rt_dom_cntl(
>>                     spin_lock_irqsave(&prv->lock, flags);
>>                   svc = rt_unit(d->vcpu[local_sched.vcpuid]->sched_unit);
>> +
>> +                rc = rt_admission_test(prv, d, svc->period, svc->budget,
>> +                                        period, budget);
>> +                if ( rc )
>> +                {
>> +                    spin_unlock_irqrestore(&prv->lock, flags);
>> +                    break;
>> +                }
>> +
>>                   svc->period = period;
>>                   svc->budget = budget;
>>                   if ( local_sched.u.rtds.flags & XEN_DOMCTL_SCHEDRT_extra )
> 
> 
> Juergen

Furkan