UBUNTU-CVE-2026-46223
In the Linux kernel, the following vulnerability has been resolved: cgroup: Defer css percpu_ref kill on rmdir until cgroup is depopulated A chain of commits going back to v7.0 reworked rmdir to satisfy the controller invariant that a subsystem's ->css_offline() must not run while tasks are still doing kernel-side work in the cgroup. [1] d245698d727a ("cgroup: Defer task cgroup unlink until after the task is done switching out") [2] a72f73c4dd9b ("cgroup: Don't expose dead tasks in cgroup") [3] 1b164b876c36 ("cgroup: Wait for dying tasks to leave on rmdir") [4] 4c56a8ac6869 ("cgroup: Fix cgroup_drain_dying() testing the wrong condition") [5] 13e786b64bd3 ("cgroup: Increment nr_dying_subsys_* from rmdir context") [1] moved task cset unlink from do_exit() to finish_task_switch() so a task's cset link drops only after the task has fully stopped scheduling. That made tasks past exit_signals() linger on cset->tasks until their final context switch, which led to a series of problems as what userspace expected to see after rmdir diverged from what the kernel needs to wait for. [2]-[5] tried to bridge that divergence: [2] filtered the exiting tasks from cgroup.procs; [3] had rmdir(2) sleep in TASK_UNINTERRUPTIBLE for them; [4] fixed the wait's condition; [5] made nr_dying_subsys_* visible synchronously. The cgroup_drain_dying() wait in [3] turned out to be a dead end. When the rmdir caller is also the reaper of a zombie that pins a pidns teardown (e.g. host PID 1 systemd reaping orphan pids that were re-parented to it during the same teardown), rmdir blocks in TASK_UNINTERRUPTIBLE waiting for those pids to free, the pids can't free because PID 1 is the reaper and it's stuck in rmdir, and the system A-A deadlocks. No internal lock ordering breaks this; the wait itself is the bug. The css killing side that drove the original reorder, however, can be made cleanly asynchronous: ->css_offline() is already async, run from css_killed_work_fn() driven by percpu_ref_kill_and_confirm(). The fix is to make that chain start only after all tasks have left the cgroup. rmdir's user-visible side then returns as soon as cgroup.procs and friends are empty, while ->css_offline() still runs only after the cgroup is fully drained. Verified by the original reproducer (pidns teardown + zombie reaper, runs under vng) which hangs vanilla and succeeds here, and by per-commit deterministic repros for [2], [3], [4], [5] with a boot parameter that widens the post-exit_signals() window so each state is reliably reachable. Some stress tests on top of that. cgroup_apply_control_disable() has the same shape of pre-existing race: when a controller is disabled via subtree_control, kill_css() ran synchronously while tasks past exit_signals() could still be linked to the cgroup's csets, and ->css_offline() could fire before they drained. This patch preserves the existing synchronous behavior at that call site (kill_css_sync() + kill_css_finish() back-to-back) and a follow-up patch will defer kill_css_finish() there using a per-css trigger. This seems like the right approach and I don't see problems with it. The changes are somewhat invasive but not excessively so, so backporting to -stable should be okay. If something does turn out to be wrong, the fallback is to revert the entire chain ([1]-[5]) and rework in the development branch instead. v2: Pin cgrp across the deferred destroy work with explicit cgroup_get()/cgroup_put() around queue_work() and the work_fn. v1 wasn't actually broken (ordered cgroup_offline_wq + queue_work order in cgroup_task_dead() saved it) but the explicit ref removes the dependency on those non-obvious invariants. Also note the pre-existing cgroup_apply_control_disable() race in the description; a follow-up will defer kill_css_finish() there.
02 / AFFECTED SOFTWARE
Affected packages
11 explicit affected versions
4 explicit affected versions
50 explicit affected versions
6 explicit affected versions
12 explicit affected versions
19 explicit affected versions
10 explicit affected versions
6 explicit affected versions
24 explicit affected versions
2 explicit affected versions
10 explicit affected versions
26 explicit affected versions
7 explicit affected versions
13 explicit affected versions
35 explicit affected versions
1 explicit affected versions
18 explicit affected versions
1 explicit affected versions
13 explicit affected versions
4 explicit affected versions
23 explicit affected versions
1 explicit affected versions
21 explicit affected versions
16 explicit affected versions
10 explicit affected versions
7 explicit affected versions
8 explicit affected versions
7 explicit affected versions
12 explicit affected versions
11 explicit affected versions
7 explicit affected versions
12 explicit affected versions
14 explicit affected versions
1 explicit affected versions
7 explicit affected versions
12 explicit affected versions
12 explicit affected versions
9 explicit affected versions
9 explicit affected versions
16 explicit affected versions
6 explicit affected versions
10 explicit affected versions
37 explicit affected versions
13 explicit affected versions
10 explicit affected versions
11 explicit affected versions
26 explicit affected versions
13 explicit affected versions
29 explicit affected versions
13 explicit affected versions
12 explicit affected versions
33 explicit affected versions
23 explicit affected versions
10 explicit affected versions
11 explicit affected versions
12 explicit affected versions
7 explicit affected versions
12 explicit affected versions
17 explicit affected versions
10 explicit affected versions
26 explicit affected versions
11 explicit affected versions
13 explicit affected versions
16 explicit affected versions
10 explicit affected versions
3 explicit affected versions
7 explicit affected versions
51 explicit affected versions
18 explicit affected versions
12 explicit affected versions
27 explicit affected versions
3 explicit affected versions
8 explicit affected versions
8 explicit affected versions
3 explicit affected versions
5 explicit affected versions
8 explicit affected versions
6 explicit affected versions
10 explicit affected versions
7 explicit affected versions
13 explicit affected versions
13 explicit affected versions
7 explicit affected versions
8 explicit affected versions
10 explicit affected versions
12 explicit affected versions
9 explicit affected versions
44 explicit affected versions
16 explicit affected versions
3 explicit affected versions
13 explicit affected versions
26 explicit affected versions
45 explicit affected versions
37 explicit affected versions
14 explicit affected versions
38 explicit affected versions
9 explicit affected versions
11 explicit affected versions
7 explicit affected versions
4 explicit affected versions
1 explicit affected versions
7 explicit affected versions
13 explicit affected versions
43 explicit affected versions
10 explicit affected versions
8 explicit affected versions
14 explicit affected versions
4 explicit affected versions
78 explicit affected versions
11 explicit affected versions
1 explicit affected versions
6 explicit affected versions
03 / CONNECTIONS
Connected vulnerabilities
04 / EVIDENCE
Source records
In the Linux kernel, the following vulnerability has been resolved: cgroup: Defer css percpu_ref kill on rmdir until cgroup is depopulated A chain of commits going back to v7.0 reworked rmdir to satisfy the controller invariant that a subsystem's ->css_offline() must not run while tasks are still doing kernel-side work in the cgroup. [1] d245698d727a ("cgroup: Defer task cgroup unlink until after the task is done switching out") [2] a72f73c4dd9b ("cgroup: Don't expose dead tasks in cgroup") [3] 1b164b876c36 ("cgroup: Wait for dying tasks to leave on rmdir") [4] 4c56a8ac6869 ("cgroup: Fix cgroup_drain_dying() testing the wrong condition") [5] 13e786b64bd3 ("cgroup: Increment nr_dying_subsys_* from rmdir context") [1] moved task cset unlink from do_exit() to finish_task_switch() so a task's cset link drops only after the task has fully stopped scheduling. That made tasks past exit_signals() linger on cset->tasks until their final context switch, which led to a series of problems as what userspace expected to see after rmdir diverged from what the kernel needs to wait for. [2]-[5] tried to bridge that divergence: [2] filtered the exiting tasks from cgroup.procs; [3] had rmdir(2) sleep in TASK_UNINTERRUPTIBLE for them; [4] fixed the wait's condition; [5] made nr_dying_subsys_* visible synchronously. The cgroup_drain_dying() wait in [3] turned out to be a dead end. When the rmdir caller is also the reaper of a zombie that pins a pidns teardown (e.g. host PID 1 systemd reaping orphan pids that were re-parented to it during the same teardown), rmdir blocks in TASK_UNINTERRUPTIBLE waiting for those pids to free, the pids can't free because PID 1 is the reaper and it's stuck in rmdir, and the system A-A deadlocks. No internal lock ordering breaks this; the wait itself is the bug. The css killing side that drove the original reorder, however, can be made cleanly asynchronous: ->css_offline() is already async, run from css_killed_work_fn() driven by percpu_ref_kill_and_confirm(). The fix is to make that chain start only after all tasks have left the cgroup. rmdir's user-visible side then returns as soon as cgroup.procs and friends are empty, while ->css_offline() still runs only after the cgroup is fully drained. Verified by the original reproducer (pidns teardown + zombie reaper, runs under vng) which hangs vanilla and succeeds here, and by per-commit deterministic repros for [2], [3], [4], [5] with a boot parameter that widens the post-exit_signals() window so each state is reliably reachable. Some stress tests on top of that. cgroup_apply_control_disable() has the same shape of pre-existing race: when a controller is disabled via subtree_control, kill_css() ran synchronously while tasks past exit_signals() could still be linked to the cgroup's csets, and ->css_offline() could fire before they drained. This patch preserves the existing synchronous behavior at that call site (kill_css_sync() + kill_css_finish() back-to-back) and a follow-up patch will defer kill_css_finish() there using a per-css trigger. This seems like the right approach and I don't see problems with it. The changes are somewhat invasive but not excessively so, so backporting to -stable should be okay. If something does turn out to be wrong, the fallback is to revert the entire chain ([1]-[5]) and rework in the development branch instead. v2: Pin cgrp across the deferred destroy work with explicit cgroup_get()/cgroup_put() around queue_work() and the work_fn. v1 wasn't actually broken (ordered cgroup_offline_wq + queue_work order in cgroup_task_dead() saved it) but the explicit ref removes the dependency on those non-obvious invariants. Also note the pre-existing cgroup_apply_control_disable() race in the description; a follow-up will defer kill_css_finish() there.
05 / REFERENCES
Further evidence
- https://git.kernel.org/linus/93618edf753838a727dbff63c7c291dee22d656b
- https://git.kernel.org/stable/c/33fa2e6b1507a0a377a151a8826438bedad1d0b0
- https://git.kernel.org/stable/c/93618edf753838a727dbff63c7c291dee22d656b
- https://ubuntu.com/security/CVE-2026-46223
- https://ubuntu.com/security/notices/USN-8566-1
- https://ubuntu.com/security/notices/USN-8568-1
- https://ubuntu.com/security/notices/USN-8569-1
- https://ubuntu.com/security/notices/USN-8593-1
- https://ubuntu.com/security/notices/USN-8603-1
- https://ubuntu.com/security/notices/USN-8618-1
- https://www.cve.org/CVERecord?id=CVE-2026-46223