[PATCH] cgroup: Don't expose dead tasks in cgroup

[Sebastian Andrzej Siewior <bigeasy@linutronix.de>]

Once a task exits it has its state set to TASK_DEAD and then it is
removed the cgroup it belonged to. The last step happens on the task
gets out of its last schedule() invocation and is delayed on PREEMPT_RT
due to locking constrains.

As a result it is possible to receive a pid via waitpid() of a task
which is still listed in cgroup.procs for the cgroup it belonged
to. This is something that systemd does not expect and as a result it
waits for its exit until a time out occurs.

This can be avoided by skipping tasks which are in the DEAD state. There
is no need to verify both task states under task_struct::pi_lock because
once the task is exiting after is its last schedule, there will not be
any sleeping locks.

Reported-by: Bert Karwatzki <spasswolf@web.de>
Closes: https://lore.kernel.org/all/20260219164648.3014-1-spasswolf@web.de/
Tested-by: Bert Karwatzki <spasswolf@web.de>
Fixes: 9311e6c29b348 ("cgroup: Fix sleeping from invalid context warning on PREEMPT_RT")
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
---

Tejun, with this change, would it be okay to
- replace the irq-work with kworker? With this change it should address
  your concern regarding "run in definite time" as mentioned in [0]. So
  it might be significantly delayed but it shouldn't be visible.
  This would lift the restriction that a irq-work needs to run on this
  CPU and the kworker could run on any CPU. 

- would it be okay to treat RT and !RT equally here (and do this delayed
  cgroup_task_dead() in both cases)

[0] https://lore.kernel.org/all/aQzg9kcnCsdRQiB4@slm.duckdns.org/

 kernel/cgroup/cgroup.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/kernel/cgroup/cgroup.c b/kernel/cgroup/cgroup.c
index c22cda7766d84..a8254229d62d3 100644
--- a/kernel/cgroup/cgroup.c
+++ b/kernel/cgroup/cgroup.c
@@ -5283,6 +5283,11 @@ static void *cgroup_procs_start(struct seq_file *s, loff_t *pos)
 
 static int cgroup_procs_show(struct seq_file *s, void *v)
 {
+	struct task_struct *tsk = v;
+
+	if (READ_ONCE(tsk->__state) & TASK_DEAD)
+		return 0;
+
 	seq_printf(s, "%d\n", task_pid_vnr(v));
 	return 0;
 }
-- 
2.51.0