Hi Krystian, Important update - srv06 panicked a **second time**, this time even with Viktor's boot parameters (`processor.max_cstate=1 idle=nomwait`) applied, on the same v6.14.0 kernel:
``` Kernel panic - not syncing: Attempted to kill the idle task! RIP: 0010:rt_mutex_adjust_prio_chain+0x5e8/0x9c0 ``` This is a **different code path** than the first crash on this host (which was in `migrate_swap`/`multi_cpu_stop`) - this one is in `rt_mutex_adjust_prio_chain` (priority inheritance/RT mutex machinery), with `AMD-Vi: IO_PAGE_FAULT` events logged just before the panic. So on srv06 alone we've now seen two distinct crash signatures, and Viktor's boot parameters did not prevent this one. Given srv06 has now panicked twice on v6.14.0 (once without the boot params, once with), while srv07 on the identical kernel/commit configuration (v6.15.6, commit 1 only, no boot params) is still stable past 10 days, I think srv06 may simply be more prone to triggering this on this hardware - possibly consistent with it being a timing/hardware- sensitive race rather than something cleanly fixed by either the boot params or presence/absence of a single commit. Since srv06 has crashed twice, I've taken it out of the active bisection test and moved it onto `linux-generic-hwe-24.04` (kernel 7.0.0-31-generic, no extra boot params) to restore a stable production state on this host, given it hosts 16 production VMs. Updated status: - srv04: v6.8.12 (neither commit) + boot params - 12d+, stable - srv08: v6.13.0 (neither commit) - 11d+, stable - srv07: v6.15.6 (commit 1 only) - 10d+, stable - srv05: v6.15.7 (both commits) - 10d+, stable - srv06: moved to HWE 7.0.0-31 (stabilized after 2 panics on v6.14.0) Let me know if you'd like me to move srv06 to a different kernel for further bisection once it's had a chance to recover, or if this is enough data to work with for now. -- 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
