Public bug reported:
# NVIDIA kernel Oops and complete freeze after display power-
off/reactivation
Suggested Ubuntu source package: nvidia-graphics-drivers-595
## Environment
- Ubuntu 24.04.5 LTS, amd64.
- MSI Sword 16 HX laptop, Intel Core i9-14900HX, 32 GiB RAM, NVIDIA GeForce RTX
4060 Laptop GPU (8 GiB).
- Three displays: internal 1920x1200 at 144 Hz, external DisplayPort and HDMI
each 1920x1080 at 60 Hz.
- NVIDIA proprietary driver: 595.99.02; Ubuntu package
595.99.02-0ubuntu0.24.04.1.
- Failing kernel: 7.0.0-42-generic, package 7.0.0-42.42~24.04.1.
- Comparison kernel: 7.0.0-38-generic, package 7.0.0-38.38~24.04.4.
- GNOME Shell 46.0-0ubuntu6~24.04.17; libmutter 46.2-1ubuntu0.24.04.18; Xorg
core 2:21.1.12-1ubuntu1.9.
- Secure Boot enabled. NVIDIA PreserveVideoMemoryAllocations=1; NVIDIA
suspend/resume services enabled.
## Observed behaviour
1. Under Wayland, display reactivation sometimes loses the DisplayPort monitor.
Kernel connector becomes disconnected and has no EDID, while the compositor
remains running. Reconnecting the video cable restored detection; cycling
monitor power did not.
2. Separate incidents under Wayland included GNOME Shell SIGSEGV after EGL
errors, terminating the user session and applications. Logged messages included
`EGL failed to allocate resources` and `drmModeAtomicCommit: Invalid argument`.
These symptoms may have related or distinct causes.
3. Switching to X11 did not eliminate the problem. On kernel 7.0.0-42, a
controlled display-off/on test initially appeared successful, but a complete
unresponsive freeze occurred minutes later. Recovery required holding the power
button.
4. EFI pstore recovered after that freeze contains two Oops in the proprietary
NVIDIA kernel module. These are stronger evidence than the earlier compositor
errors.
## Sanitized kernel evidence
Selected messages manually transcribed; complete logs, memory dumps and
register values are intentionally not attached.
```text
[1004.590869] Oops: general protection fault [#1] SMP NOPTI
Kernel: 7.0.0-42-generic #42~24.04.1-Ubuntu
Comm: kworker/0:0; taint: proprietary/out-of-tree module
Workqueue: kacpi_notify acpi_os_execute_deferred
RIP: _nv015620rm+0x604/0xa40 [nvidia]
Selected call-trace entries (not a complete stack):
rm_acpi_notify+0x48/0x27a [nvidia]
rm_acpi_notify+0x53/0x27a [nvidia]
nv_acpi_notify_event+0x1a/0x30 [nvidia]
[1009.906451] Oops: general protection fault [#2] SMP NOPTI
Comm: nv_queue; taint: proprietary/out-of-tree module, DIE
RIP: _nv013730rm+0x3c/0x310 [nvidia]
```
Xid 31/MMU faults involving Xorg were logged on the following boot; they
are not direct evidence of the preceding freeze.
## Reproduction context
- Three displays attached; automatic system suspend disabled. No actual
suspend/resume transition was found in the reviewed journal windows.
- Lock GNOME session, request Mutter DisplayConfig PowerSaveMode=3, then
request PowerSaveMode=0 after approximately 10 seconds.
- Observe the desktop for several minutes, not just immediately after power-on.
- This procedure has already caused a complete freeze and should only be
attempted with all work saved. Reproduction is intermittent; no repeatability
rate is established.
Expected: all displays return, the desktop stays responsive, and applications
remain open.
Actual: loss of DisplayPort, compositor crashes, or complete freeze with NVIDIA
kernel Oops as described above.
## Comparison and temporary mitigation
Booted kernel 7.0.0-38 with the same NVIDIA driver and X11. One brief
display-off/reactivation test passed, with all three monitors and the same
Xorg/GNOME processes retained. The user reactivated with the keyboard, so a
full uninterrupted 10-second physical off interval is not established.
Unlocking succeeded. No new NVIDIA/Oops/Xid errors were found during the short
observation window.
However, the 38 boot had already logged:
`nvidia-modeset: ERROR: GPU:0: Idling display engine timed out:
0x0000c67e:4:0:1169`
Kernel 38 is temporarily selected as default, keeping both kernels
installed. This is not a confirmed fix or proof that kernel 42 alone is
responsible. A preceding update changed multiple graphics components and
the kernel; the causal regression is not isolated. Long-term stability
is unverified.
## Privacy
This report contains no username, hostname, email, home-directory path, IP/MAC
address, filesystem UUID, device serial number, EDID, window title, account
identifier, or raw crash dump. Hardware models and software versions are
retained for diagnosis. No original journal, pstore file or Apport crash
attachment is included.
** Affects: nvidia-graphics-drivers-595 (Ubuntu)
Importance: Undecided
Status: New
** Tags: multi-monitor noble nvidia
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2173502
Title:
NVIDIA kernel Oops and complete freeze after display power-
off/reactivation
To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/nvidia-graphics-drivers-595/+bug/2173502/+subscriptions
--
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
