runc update error when there is a nested cpuset cgroup

[Ace-Tang]

I try this command runc --root /run/containerd/runc/default/ update --cpuset-cpus "0,1" docker2, and get error

failed to write 0,1 to cpuset.cpus: write /sys/fs/cgroup/cpuset/default/docker2/cpuset.cpus: device or resource busy

this is because a nested cpuset cgroup existed, cause child cpusets must contain a subset of the parents CPU and Memory resources, so we cannot simply change the data in parent cgroup file.

this case occurs in a real business scene. here is my test environment, docker2 is a container cpuset cgroup, t1, t2, l1, l2, c3 is a nested cpuset cgroup.

root@www:/sys/fs/cgroup/cpuset/default/docker2# tree
...
├── cpuset.sched_load_balance
├── cpuset.sched_relax_domain_level
├── notify_on_release
├── t1
│   ...
│   ├── cpuset.memory_migrate
│   ├── cpuset.memory_pressure
│   ├── cpuset.memory_spread_page
│   ├── cpuset.memory_spread_slab
│   ├── cpuset.mems
│   ├── cpuset.sched_load_balance
│   ├── cpuset.sched_relax_domain_level
│   ├── l1
│   │   ├── cgroup.clone_children
│   │   ├── cgroup.procs
│   │   ├── cpuset.cpu_exclusive
│   │   ├── cpuset.cpus
│   │   ├── cpuset.effective_cpus
│   │   ├── cpuset.effective_mems
│   │   ├── cpuset.mem_exclusive
│   │   ├── cpuset.mem_hardwall
│   │   ├── cpuset.memory_migrate
│   │   ├── cpuset.memory_pressure
│   │   ├── cpuset.memory_spread_page
│   │   ├── cpuset.memory_spread_slab
│   │   ├── cpuset.mems
│   │   ├── cpuset.sched_load_balance
│   │   ├── cpuset.sched_relax_domain_level
│   │   ├── notify_on_release
│   │   └── tasks
│   ├── notify_on_release
│   └── tasks
├── t2
│    ...
│   ├── cpuset.memory_spread_slab
│   ├── cpuset.mems
│   ├── cpuset.sched_load_balance
│   ├── cpuset.sched_relax_domain_level
│   ├── l2
│   │   ├── c3
│   │   │   ├── cgroup.clone_children
│   │   │   ├── cgroup.procs
│   │   │   ├── cpuset.cpu_exclusive
│   │   │   ├── cpuset.cpus
│   │   │   ├── cpuset.effective_cpus
│   │   │   ├── cpuset.effective_mems
│   │   │   ├── cpuset.mem_exclusive
│   │   │   ├── cpuset.mem_hardwall
│   │   │   ├── cpuset.memory_migrate
│   │   │   ├── cpuset.memory_pressure
│   │   │   ├── cpuset.memory_spread_page
│   │   │   ├── cpuset.memory_spread_slab
│   │   │   ├── cpuset.mems
│   │   │   ├── cpuset.sched_load_balance
│   │   │   ├── cpuset.sched_relax_domain_level
│   │   │   ├── notify_on_release
│   │   │   └── tasks
│   │   ├── cgroup.clone_children
│   │   ├── cgroup.procs
│   │   ├── cpuset.cpu_exclusive
│   │   ├── cpuset.cpus
│   │   ├── cpuset.effective_cpus
│   │   ├── cpuset.effective_mems
│   │   ├── cpuset.mem_exclusive
│   │   ├── cpuset.mem_hardwall
│   │   ├── cpuset.memory_migrate
│   │   ├── cpuset.memory_pressure
│   │   ├── cpuset.memory_spread_page
│   │   ├── cpuset.memory_spread_slab
│   │   ├── cpuset.mems
│   │   ├── cpuset.sched_load_balance
│   │   ├── cpuset.sched_relax_domain_level
│   │   ├── notify_on_release
│   │   └── tasks
│   ├── notify_on_release
│   └── tasks
└── tasks

[P1ng-W1n]

Is there any update about this issue?
It was opened quite some time ago, but nobody seems to checked it yet.

[vladimir-kozyrev]

This issue affects quite a lot of people. Will there be a fix any time soon?
#66614 #54967

[omriperi]

Hey, any update on this issue?