v4:
  - Add missing rcu_read_lock/unlock to cpuset_cancel_fork() in patch 3.
  - Add patch 5 to reduce performance impact for the
    non-CLONE_INTO_CGROUP case.

 v3:
  - Update patches 2 & 3 to put task_cs() call under rcu_read_lock().

 v2:
  - Drop v1 patch 3
  - Add a new patch to fix an issue in cpuset_cancel_attach() and
    another patch to add cpuset_can_fork() and cpuset_cacnel_fork()
    methods.

The first patch in this series fixes a problem in
cpuset_cancel_attach(). Patches 2 and 3 fixes the CLONE_INTO_CGROUP
problem in cpuset. Patch 4 is a minor fix. The last patch is a
performance optimization patch for the non-CLONE_INTO_CGROUP case.

Waiman Long (5):
  cgroup/cpuset: Wake up cpuset_attach_wq tasks in
    cpuset_cancel_attach()
  cgroup/cpuset: Make cpuset_fork() handle CLONE_INTO_CGROUP properly
  cgroup/cpuset: Add cpuset_can_fork() and cpuset_cancel_fork() methods
  cgroup/cpuset: Make cpuset_attach_task() skip subpartitions CPUs for
    top_cpuset
  cgroup/cpuset: Optimize out unneeded
    cpuset_can_fork/cpuset_cancel_fork calls

 include/linux/cgroup-defs.h |   6 ++
 kernel/cgroup/cgroup.c      |  23 +++--
 kernel/cgroup/cpuset.c      | 167 +++++++++++++++++++++++++++++-------
 3 files changed, 159 insertions(+), 37 deletions(-)

-- 
2.31.1

On Tue, Apr 11, 2023 at 09:35:56AM -0400, Waiman Long wrote:
>  v4:
>   - Add missing rcu_read_lock/unlock to cpuset_cancel_fork() in patch 3.
>   - Add patch 5 to reduce performance impact for the
>     non-CLONE_INTO_CGROUP case.
> 
>  v3:
>   - Update patches 2 & 3 to put task_cs() call under rcu_read_lock().
> 
>  v2:
>   - Drop v1 patch 3
>   - Add a new patch to fix an issue in cpuset_cancel_attach() and
>     another patch to add cpuset_can_fork() and cpuset_cacnel_fork()
>     methods.
> 
> The first patch in this series fixes a problem in
> cpuset_cancel_attach(). Patches 2 and 3 fixes the CLONE_INTO_CGROUP
> problem in cpuset. Patch 4 is a minor fix. The last patch is a
> performance optimization patch for the non-CLONE_INTO_CGROUP case.

Applied 1-4 to cgroup/for-6.3-fixes w/ stable cc'd. Given that the fixes are
a bit involved, the breakages have been there for quite a while and the
cpuset code has changed quite a bit, backporting might not be trivial tho.
Let's see how that goes.

Thanks.

-- 
tejun

On 4/12/23 14:26, Tejun Heo wrote:
> On Tue, Apr 11, 2023 at 09:35:56AM -0400, Waiman Long wrote:
>>   v4:
>>    - Add missing rcu_read_lock/unlock to cpuset_cancel_fork() in patch 3.
>>    - Add patch 5 to reduce performance impact for the
>>      non-CLONE_INTO_CGROUP case.
>>
>>   v3:
>>    - Update patches 2 & 3 to put task_cs() call under rcu_read_lock().
>>
>>   v2:
>>    - Drop v1 patch 3
>>    - Add a new patch to fix an issue in cpuset_cancel_attach() and
>>      another patch to add cpuset_can_fork() and cpuset_cacnel_fork()
>>      methods.
>>
>> The first patch in this series fixes a problem in
>> cpuset_cancel_attach(). Patches 2 and 3 fixes the CLONE_INTO_CGROUP
>> problem in cpuset. Patch 4 is a minor fix. The last patch is a
>> performance optimization patch for the non-CLONE_INTO_CGROUP case.
> Applied 1-4 to cgroup/for-6.3-fixes w/ stable cc'd. Given that the fixes are
> a bit involved, the breakages have been there for quite a while and the
> cpuset code has changed quite a bit, backporting might not be trivial tho.
> Let's see how that goes.

I know stable backport won't be straight forward. I am planning to help 
in the backporting effort. Thanks for taking these into your cgroup tree.

Cheers,
Longman