Re: [PATCH] cgroup/cpuset: Disable cpuset_cpumask_can_shrink() test if not load balancing

[Waiman Long <llong@redhat.com>]

On 11/18/24 4:39 AM, Juri Lelli wrote:
> On 15/11/24 12:55, Waiman Long wrote:
>> On 11/15/24 5:54 AM, Juri Lelli wrote:
>>> Hi Waiman,
>>>
>>> On 14/11/24 13:19, Waiman Long wrote:
>>>> With some recent proposed changes [1] in the deadline server code,
>>>> it has caused a test failure in test_cpuset_prs.sh when a change
>>>> is being made to an isolated partition. This is due to failing
>>>> the cpuset_cpumask_can_shrink() check for SCHED_DEADLINE tasks at
>>>> validate_change().
>>> What sort of change is being made to that isolated partition? Which test
>>> is failing from the test_cpuset_prs.sh collection? Asking because I now
>>> see "All tests PASSED" running that locally (with all my 3 patches on
>>> top of cgroup/for-6.13 w/o this last patch from you).
>> The failing test isn't an isolated partition. The actual test failure is
>>
>> Test TEST_MATRIX[62] failed result check!
>> C0-4:X2-4:S+ C1-4:X2-4:S+:P2 C2-4:X4:P1 . . X5 . . 0 A1:0-4,A2:1-4,A3:2-4
>> A1:P0,A2:P-2,A3:P-1
>>
>> In this particular case, cgroup A3 has the following setting before the X5
>> operation.
>>
>> A1/A2/A3/cpuset.cpus: 2-4
>> A1/A2/A3/cpuset.cpus.exclusive: 4
>> A1/A2/A3/cpuset.cpus.effective: 4
>> A1/A2/A3/cpuset.cpus.exclusive.effective: 4
>> A1/A2/A3/cpuset.cpus.partition: root
> Right, and is this problematic already?
We allow nested partition setup. So there can be a child partition 
underneath a parent partition. So this is OK.
>
> Then the test, I believe, does
>
> # echo 5 >cgroup/A1/A2/cpuset.cpus.exclusive
>
> and that goes through and makes the setup invalid - root domain reconf
> and the following
>
> # cat cgroup/A1/cpuset.cpus.partition
> member
> # cat cgroup/A1/A2/cpuset.cpus.partition
> isolated invalid (Parent is not a partition root)
> # cat cgroup/A1/A2/A3/cpuset.cpus.partition
> root invalid (Parent is an invalid partition root)
>
> Is this what shouldn't happen?
>
A3 should become invalid because none of the CPUs in 
cpuset.cpus.exclusive can be granted. However A2 should remain a valid 
partition. I will look further into that. Thank for spotting this 
inconsistency.

Cheers,
Longman