Thanks for confirming this bug and for the fast turnaround.

Update on the HWE 7.0 kernel test on srv05: it has now been running
kernel 7.0.0-29-generic for **over 33 hours (uptime since 2026-08-18
16:20:03)** under normal production load, with all 20 guest VMs running
continuously and no kernel panic.

```
ubuntu@srv05:~$ uptime -s
2026-08-18 16:20:03
ubuntu@srv05:~$ uptime
 09:05:26 up 1 day, 16:45,  1 user,  load average: 4.20, 4.39, 4.44
```

This comfortably exceeds the 1-24 hour window in which panics previously
occurred on 6.8.0-137-generic across all 5 affected hosts, so I'm fairly
confident the HWE 7.0 kernel resolves the issue on this host.

I took a closer look at the commit you linked
(9709eb0f845b713ba163f2c461537d8add3e4e04, "sched/numa: fix task swap by
skipping kernel threads"). It modifies `task_numa_compare()` in
`kernel/sched/fair.c` to also exclude kernel threads (`PF_KTHREAD`) and
userspace tasks without an `mm` from being selected as NUMA balancing
swap targets - previously only `PF_EXITING` and the idle task were
excluded. Given that our panics consistently trace through
`multi_cpu_stop`/`cpu_stopper_thread`, and that NUMA task-swap
interacting with the stopper machinery matches a known historical
failure mode (a 2018 LKML thread on cpu stopper threads and load
balancing deadlocking, which I referenced in the original report), this
looks like a strong candidate for the actual root cause rather than a
coincidental fix.

I'll continue monitoring srv05 and plan to roll the same kernel out to
the remaining affected hosts (srv04, srv06, srv07, srv08) if stability
holds.

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2163642

Title:
  Kernel panic "Attempted to kill the idle task" on AMD Opteron multi-
  node NUMA under KVM (6.8.0-137)

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2163642/+subscriptions


-- 
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Reply via email to