Public bug reported:

Hardware: Lenovo IdeaPad 3 15ADA05 (81W1), AMD Ryzen 5 3500U with Radeon
Vega Mobile Gfx, 8 GB DDR4 (4 GB soldered + 4 GB SODIMM), BIOS E8CN34WW
(04/28/2022), WD SN530 NVMe, Qualcomm QCA6174 Wi-Fi (ath10k).

Regression:
- The machine ran stable for five years, including kernels 7.0.0-28, -30 and 
-31 (installed 2026-07-18, 2026-08-22, 2026-09-05, on Ubuntu 24.04 HWE).
- 7.0.0-34 was installed on 2026-09-27. The first crash occurred on 2026-10-06. 
Since then the system crashes repeatedly, also with 7.0.0-42 and 7.0.0-43 (now 
on Ubuntu 26.04.1 LTS).

Symptoms:
- Hard crash / kernel panic ("Fatal exception in interrupt"), several times per 
week.
- Triggers: typically a few minutes after resume from suspend (S3/deep), but 
also shortly after a cold boot (7 seconds and about 2 minutes after boot in two 
cases). Suspend has now been disabled as a workaround; crashes after cold boot 
remain possible.

Crash signatures (from pstore, files attached):
- 2026-10-06, 7.0.0-34: "BUG: kernel NULL pointer dereference, address: 
0000000000000121", RIP task_h_load+0x99, call trace wake_affine -> 
select_task_rq_fair -> select_task_rq -> try_to_wake_up (Comm: gnome-shell); 
followed by page faults in update_load_avg / enqueue_entity on other CPUs. 
(This October dump is no longer available; signature taken from my notes.)
- 2026-11-09, 7.0.0-42: page faults, RIP desc_read_finalized_seq, 
kernel_fpu_begin_mask / irq_fpu_usable (Comm: apt-cache), "Fatal exception in 
interrupt".
- 2026-11-13, 7.0.0-43: oops 7 s after boot; the preserved part of the call 
trace runs through drm_ioctl (only part of the dump was saved).
- 2026-11-18, 7.0.0-43: "unable to handle page fault", RIP 
__sched_balance_update_blocked_averages+0x531/0x690 and +0x31b/0x690, then 
"kernel NULL pointer dereference, address: 0000000000000121" (Comm: swapper/1, 
cpuidle path), ~110 s after boot.

The NULL dereference at offset 0x121 in task_h_load / the load-tracking
path resembles the h_load_next use-after-free that was addressed
upstream by "sched/fair: Clear ->h_load_next when unregistering a
cgroup" (tip sched/urgent, December 2025). It may be worth checking
whether this fix (or a related one) is missing or was regressed between
7.0.0-31 and 7.0.0-34.

Hardware checks:
- Memtest86+ 8.00: 15 full passes (~8 h), 0 errors. The system did not crash 
during this 8 h full-load test.
- NVMe SMART: healthy (3 % used, 0 media errors).
- Secure Boot disabled.

I can test kernels or provide further logs if needed.

** Affects: linux (Ubuntu)
     Importance: Undecided
         Status: New

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2175492

Title:
  Kernel panic / NULL pointer dereference in CFS scheduler (task_h_load,
  update_load_avg, sched_balance_update_blocked_averages) since
  7.0.0-34; 7.0.0-31 was stable

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2175492/+subscriptions


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

Reply via email to