Hi Matthew, Just report another PID 1 hang variant on an AWS ec2 machine (c8a.metal-48xl, kernel 7.0.0-1004-aws). It seems not related to cpuset_migrate_mm.
What the kernel's hung-task reports show: - PID 39263 (systemd-user-runtime-dir), from a user login at 10:20:57, was moving itself into its cgroup. Path: cgroup_procs_write_start → percpu_down_write → rcu_sync_enter → synchronize_rcu (→ __wait_rcu_gp / call_rcu_hurry). It never returned and held cgroup_mutex the whole time. - PID 1 blocked in cgroup_mkdir → cgroup_kn_lock_live → mutex_lock. Kernel message: "task systemd:1 is blocked on a mutex likely owned by task (systemd-user-):39263". It was reported blocked for more than 614 s by 10:31:48. We tried to offline a CPU to recover the host, but that got stuck in synchronize_rcu the same way. We suspect an RCU callback thread (rcuos) running on an isolated core was starved by the busy application threads there, so the waiter's wake- up callback never ran. Have you seen similar cases on your side, and what should we capture if it happens again? Thanks Cindy -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/2165410 Title: Ubuntu LTS 26.04 linux-aws: systemd enters D state and blocks SSH during cpuset migration on nohz_full CPUs To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/linux-aws/+bug/2165410/+subscriptions -- ubuntu-bugs mailing list [email protected] https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
