Suspend failures have returned even on nvidia-driver-580-open, but under
a newer kernel than the earlier working test.

The previous working 580-open result was on kernel 7.0.0-10-generic. I
am now on:

- kernel 7.0.0-12-generic
- nvidia-driver-580-open 580.126.09-0ubuntu9
- nvidia-dkms-580-open 580.126.09-0ubuntu9
- loaded module path /lib/modules/7.0.0-12-generic/updates/dkms/nvidia.ko.zst
- loaded module version 580.126.09

Fresh Apr 8 reproductions on this 580-open plus 7.0.0-12-generic setup:

Boot (2026-04-07 23:47:26 -> 2026-04-08 02:02:21):
  - nvidia-suspend.service completed
  - systemd-sleep: Performing sleep operation 'suspend'...
  - kernel: PM: suspend entry (deep)
  - no matching PM: suspend exit or wake lines
- The immediately following boot -4 is a kdump dump boot:
  - systemd.unit=kdump-tools-dump.service
  - Starting kdump-tools-dump.service - Kernel crash dump capture service...

Boot (2026-04-08 08:13:36 -> 2026-04-08 08:14:21) shows the same pattern:
  - nvidia-suspend.service completed
  - systemd-sleep: Performing sleep operation 'suspend'...
  - kernel: PM: suspend entry (deep)
  - no matching PM: suspend exit or wake lines
- The immediately following boot -1 is again a kdump dump boot.

I was able to extract dmesg from both Apr 8 kdump files, and both show
the same crash signature as the earlier 595 reproductions:

- PM: suspend entry (deep)
- jump_label: Fatal kernel bug, unexpected op at 
nvkms_kthread_q_callback+0x8e/0x1a0 [nvidia_modeset]
- kernel BUG at arch/x86/kernel/jump_label.c:73!
- Oops: invalid opcode: 0000 [#1]
- call trace through freeze_processes() -> enter_state() -> pm_suspend()

Additional clue from the first normal boot after the latest failure
(08:15:09):

- x86/amd: Previous system reset reason [0x00080800]: software wrote 0x6 to 
reset control register 0xCF9
- PM:   Magic number: 6:49:268
- Unpatched return thunk in use. This should not happen!
- WARNING: arch/x86/kernel/cpu/bugs.c:3736 at __warn_thunk+0x10/0x20, CPU#8: 
(udev-worker)/405
- trace includes nvidia_init_module+0x29/0x710 [nvidia]
- then the 580.126.09 open module loads normally

So the earlier conclusion needs tightening: 580-open was a working
control on 7.0.0-10-generic, but it is not a stable workaround after
moving to 7.0.0-12-generic. The readable Apr 8 dumps show that the exact
same nvidia_modeset suspend crash path now reproduces on 580-open as
well.


2026-04-08 08:14 crash: /var/crash/202604080814/dump.202604080814

[   47.416910] [   T9515] PM: suspend entry (deep)
[   47.755882] [    T330] Filesystems sync: 0.338 seconds
[   47.817637] [   T9515] jump_label: Fatal kernel bug, unexpected op at 
nvkms_kthread_q_callback+0x8e/0x1a0 [nvidia_modeset] [00000000d2e552db] (e9 9b 
00 00 00 != 0f 1f 44 00 00)) size:5 type:1
[   47.817670] [   T9515] kernel BUG at arch/x86/kernel/jump_label.c:73!
[   47.817676] [   T9515] Oops: invalid opcode: 0000 [#1] SMP NOPTI
[   47.817682] [   T9515] CPU: 12 UID: 0 PID: 9515 Comm: systemd-sleep Kdump: 
loaded Tainted: G        W  O        7.0.0-12-generic #12-Ubuntu PREEMPT(lazy)
[   47.817693] [   T9515] Hardware name: Gigabyte Technology Co., Ltd. X570 
GAMING X/X570 GAMING X, BIOS F40 10/28/2025
[   47.817756] [   T9515]  arch_jump_label_transform_queue+0x37/0x90
[   47.817762] [   T9515]  __jump_label_update+0x47/0x100
[   47.817768] [   T9515]  jump_label_update+0x5c/0x110
[   47.817781] [   T9515]  freeze_processes+0xd1/0xe0
[   47.817786] [   T9515]  enter_state+0xe0/0x610
[   47.817791] [   T9515]  pm_suspend+0x49/0x90

2026-04-08 02:02 crash: /var/crash/202604080202/dump.202604080202

[ 8096.644804] [  T35051] PM: suspend entry (deep)
[ 8097.022635] [  T33548] Filesystems sync: 0.377 seconds
[ 8097.082439] [  T35051] jump_label: Fatal kernel bug, unexpected op at 
nvkms_kthread_q_callback+0x8e/0x1a0 [nvidia_modeset] [000000008231907c] (e9 9b 
00 00 00 != 0f 1f 44 00 00)) size:5 type:1
[ 8097.082479] [  T35051] kernel BUG at arch/x86/kernel/jump_label.c:73!
[ 8097.082487] [  T35051] Oops: invalid opcode: 0000 [#1] SMP NOPTI
[ 8097.082494] [  T35051] CPU: 10 UID: 0 PID: 35051 Comm: systemd-sleep Kdump: 
loaded Tainted: G        W  O        7.0.0-12-generic #12-Ubuntu PREEMPT(lazy)
[ 8097.082507] [  T35051] Hardware name: Gigabyte Technology Co., Ltd. X570 
GAMING X/X570 GAMING X, BIOS F40 10/28/2025
[ 8097.082585] [  T35051]  arch_jump_label_transform_queue+0x37/0x90
[ 8097.082592] [  T35051]  __jump_label_update+0x47/0x100
[ 8097.082600] [  T35051]  jump_label_update+0x5c/0x110
[ 8097.082615] [  T35051]  freeze_processes+0xd1/0xe0
[ 8097.082621] [  T35051]  enter_state+0xe0/0x610
[ 8097.082627] [  T35051]  pm_suspend+0x49/0x90

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2147391

Title:
  Suspend crashes in nvidia_modeset on RTX 3080 Ti with NVIDIA 595 on
  Ubuntu 26.04 beta kernel 7.0.0-10-generic; 580-open works on the same
  system

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/nvidia-graphics-drivers-595/+bug/2147391/+subscriptions


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

Reply via email to