Ubuntu 23.04 System crashes when Display Port is connected

Hi all,

I have a RTX2060 connected through an eGPU via TB3 to my C940 Laptop. When I attach an external Monitor through Display Port the System instantly crashes (no response anymore) and I get a white screen with the following error: “Oh no! Something has gone wrong. A problem has occurred and the system can’t recover. Please contact a system administrator”. I can only force-shutdown the system via the power-button.
When the Monitor is connected during boot, everything works fine until the login-screen would display. At that point it crashes and displays the same error message

When I attach the Monitor through HDMI, everything works fine. It also works fine with Windows and it did work with Ubuntu 22.10. So I can rule out HW problems.

OS: Ubuntu 23.04, Kernel 6.2, Xorg, Standard Ubuntu-Gnome desktop
Nvidia Driver: 525.105.17 (remark: Issue also present with version 470 and 515)
GPU: RTX2060
eGPU: Lenovo Booststation.
Laptop: Lenovo C940
Monitor: Acer Predator XB273K GP

As the system crashes, I cannot run the nvidia-bugreport.
Would be great, if anyone can help here! Let me know, if I there is other data needed.

Thanks!

/etc/X11/xorg.conf:

Section “Module”
Load “modesetting”
EndSection

Section “Device”
Identifier “Device0”
Driver “nvidia”
VendorName “NVIDIA Corporation”
Option “AllowEmptyInitialConfiguration”
Option “AllowExternalGpus” “True”
EndSection

prime profile set to performance mode

Anyone from Nvidia could have a look at this? I upgraded to Driver Version 530, but Problem still persists. Thanks!

I meanwhile figured out a (bit painful) workaround, that avoids the crash:

  1. unplug the DP cable from the GPU
  2. Boot and login
  3. Plugin the DP cable

a bit disappointing, not to receive a Feedback.
the problem persists also with a RTX4070 card, as well as on Ubuntu 23.10

Without any nvidia-bug-report.log there can’t be any response. The “oh no!” message is just gnome-shell crashing, so the system should be still accessible, either by ssh or on a VT (ctrl+alt+f6).

Thanks for your reply. Highly Appreciated!
VT (ctrl+alt+F6) did not work. Screen turns black and stays black. No prompt.
SSH did work. File is attached. Thanks for your help.
nvidia-bug-report.log.gz (449.7 KB)

Xserver and nvidia gpu seem to be working fine, just gnome-shell is crashing:

[   48.569946] traps: gnome-shell[3308] trap int3 ip:7fdcbbd46867 sp:7ffee4a2a780 error:0 in libglib-2.0.so.0.7800.0[7fdcbbd03000+9a000]
[   48.942309] traps: gnome-shell[3338] trap int3 ip:7f6fa4346867 sp:7ffe802a2b10 error:0 in libglib-2.0.so.0.7800.0[7f6fa4303000+9a000]

You might check what gnome shell is logging
journalctl -b0 |grep gnome-shell
or after reboot
journalctl -b-1 |grep gnome-shell

Here is the output. Sadly I did not find any solution to these errors online. Can you help also here? Thanks!

Nov 16 19:24:38 XYZ gnome-shell[3376]: Running GNOME Shell (using mutter 45.0) as a X11 window and compositing manager
Nov 16 19:24:39 XYZ /usr/libexec/gdm-x-session[3366]: dbus-daemon[3366]: Activating service name=‘org.a11y.atspi.Registry’ requested by ‘:1.1’ (uid=125 pid=3376 comm="/usr/bin/gnome>
Nov 16 19:24:39 XYZ gnome-shell[3376]: Received an X Window System error.
Nov 16 19:24:39 XYZ kernel: traps: gnome-shell[3376] trap int3 ip:7f8a36946867 sp:7fffa4a90ac0 error:0 in libglib-2.0.so.0.7800.0[7f8a36903000+9a000]
Nov 16 19:24:39 XYZ gnome-shell[3407]: Running GNOME Shell (using mutter 45.0) as a X11 window and compositing manager
Nov 16 19:24:39 XYZ gnome-shell[3407]: Received an X Window System error.
Nov 16 19:24:39 XYZ kernel: traps: gnome-shell[3407] trap int3 ip:7fa605709867 sp:7ffcd7167670 error:0 in libglib-2.0.so.0.7800.0[7fa6056c6000+9a000]

Checking the log again, it seems that gnome-shell of gdm tries to set a weird mode when the monitor is connected to DP-0: Setting mode "DP-0: 1920x2160 @1920x2160 +3840+0 {ViewPortIn=1920x2160, ViewPortOut=1920x2160+0+0}" and then crashes. While the monitor is connected to HDMI, it doesn’t set any mode.
Does /var/lib/gdm/.config/monitors.xml (or /var/lib/gdm3/.config/monitors.xml) exist and does it work if you delete it?
If not, please create /etc/X11/xorg.conf.d/nvidia-modedebug.conf

Section "OutputClass"
    Identifier "nvidia-modedebug"
    MatchDriver "nvidia-drm"
    Driver "nvidia"
    Option "ModeDebug" "true"
EndSection

and create a new nvidia-bug-report.log after crash.

The file does not exist. I did have a /home/username/.config/monitors.xml, but deleting that is not changing anything.
Bug Report is attached. Thanks!
nvidia-bug-report.log.gz (487.7 KB)

You might try doing it the other way round: use gnome control center to arrange your monitors so ~/.config/monitors.xml is recreated and then copy it to /var/lib/gdm3/.config/ so gdm has a valid monitor config.

Sadly, that did not work. It keeps crashing.
The workaround I described (unplug DP cable, boot & login, plugin cable again) stopped working since the Ubuntu 23.10 upgrade. Even if I use the same config files that I had in 23.04.
Two more observations:

  • I manged to grab a Lenovo Monitor for testing. Connecting this Screen through DP does work. The Problem seems thus related to the combination with my Acer Predator Screen
  • Changing display manager to LightDM does display a Login screen. It then crashes after logging in.

Any further ideas, what to test out? Thanks!

Actually, no. I guess it’s clearly a gnome-shell bug which got worse by updating it (alongside ubuntu) so you can only report a bug with ubuntu.

Will do. Thanks a lot for your help so far!

You might want to attach your edid when opening a bug report so the devs don’t have to buy the same monitor for repro, it’s this:
edidcrash.txt (1.1 KB)