Re: [PATCH v2 3/3] cgroup/cpuset: Release isolated CPUs when the parent is invalid

[Guopeng Zhang <guopeng.zhang@linux.dev>]

在 2026/11/6 12:35, Ridong Chen 写道:
> 
> 
> On 11/5/2026 9:03 PM, Guopeng Zhang wrote:
>> From: Guopeng Zhang <zhangguopeng@kylinos.cn>
>>
>> An isolated partition can be invalidated after its parent has already
>> become invalid. update_parent_effective_cpumask() then skips returning
>> CPUs to the parent, but also skips removing them from isolated_cpus.
>> The invalid child continues to isolate CPUs that no longer belong to
>> any valid isolated partition.
>>
>> For example, on a 32-CPU system with all CPUs online and without
>> isolcpus= or nohz_full= boot parameters:
>>
>>    cd /sys/fs/cgroup
>>    echo +cpuset > cgroup.subtree_control
>>    mkdir A
>>    echo 0,2-31 > A/cpuset.cpus
>>    echo root > A/cpuset.cpus.partition
>>    echo +cpuset > A/cgroup.subtree_control
>>    mkdir A/B
>>    echo 2-3 > A/B/cpuset.cpus
>>    echo isolated > A/B/cpuset.cpus.partition
>>    cat cpuset.cpus.isolated
>>    2-3
>>    echo 0 > /sys/devices/system/cpu/cpu1/online
>>    cat A/B/cpuset.cpus.partition
>>    isolated invalid
>>    cat cpuset.cpus.isolated
>>    2-3
>>
>> Release the invalidated child's isolated CPUs when its parent is no
>> longer a valid partition. Use isolated_cpus_release() to preserve CPUs
>> still covered by a valid isolated ancestor and by boot isolation.
>>
>> After this fix, cpuset.cpus.isolated is empty for this reproducer.
>>
>> Fixes: 11e5f407b64a ("cgroup/cpuset: Keep track of CPUs in isolated partitions")
>> Signed-off-by: Guopeng Zhang <zhangguopeng@kylinos.cn>
>> ---
>>   kernel/cgroup/cpuset.c | 7 ++++++-
>>   1 file changed, 6 insertions(+), 1 deletion(-)
>>
>> diff --git a/kernel/cgroup/cpuset.c b/kernel/cgroup/cpuset.c
>> index fc5454c2740f..f3b808cea51c 100644
>> --- a/kernel/cgroup/cpuset.c
>> +++ b/kernel/cgroup/cpuset.c
>> @@ -1976,10 +1976,15 @@ static int update_parent_effective_cpumask(struct cpuset *cs, int cmd,
>>           /*
>>            * Make the current partition invalid.
>>            */
>> -        if (is_partition_valid(parent))
>> +        if (is_partition_valid(parent)) {
>>               adding = cpumask_and(tmp->addmask,
>>                            cs->effective_xcpus,
>>                            parent->effective_xcpus);
>> +        } else if (old_prs == PRS_ISOLATED) {
>> +            /* No valid parent will release these isolated CPUs. */
>> +            cpumask_copy(tmp->addmask, cs->effective_xcpus);
>> +            isolated_cpus_release(cs, tmp->addmask);
>> +        }
>>           if (old_prs > 0)
>>               new_prs = -old_prs;
>>   
> 
> Hi Guopeng,
> 
> Thanks for the fix. Could you try this patch to see if it fixes the
> issue?
> 
> https://lore.kernel.org/cgroups/20261105125215.2205892-4-ridong.chen@linux.dev/T/#u
> 

Yes, I took another look at the code and tested your patch. It fixes the issue, so I'll drop mine in v3.

> I'm concerned about whether the related issue has been fixed in all
> the places that need fixing. In your reproducer, CPUs 2-3 are already
> wrong when A becomes invalid: partition_xcpus_del() only updates
> isolated_cpus when the old state differs from the parent's state, so
> the CPUs that A gave back but had distributed to the isolated child B
> are left in isolated_cpus. This patch only cleans them up later, when
> B is invalidated with an invalid parent. Other paths that give back
> CPUs of an isolated descendant may have the same problem.
> 

Haha, good point! I missed that again.

Thanks,
Guopeng
> My patch makes partition_xcpus_del() always set the isolation state of
> the given back CPUs, including those of the descendants, to the state
> of the parent that receives them. With it, isolated_cpus_release()
> should no longer be needed.
> 
> It would also be good to add your reproducer to test_cpuset_prs.sh.
> It can be expressed with O1=0 in the test matrix.
>