[nvidia] External display sometimes frozen, after logout/login has only black screen.

[Andriy Khomych]

After some system updates my external screen is not working properly. After logout/log in it keeps black or with a background image and is frozen. Moreover, it happens from time to time even when you were logged in without any explanation.

ProblemType: Bug
DistroRelease: Ubuntu 22.04
Package: nvidia-driver-530 530.41.03-0ubuntu0.22.04.2
ProcVersionSignature: Ubuntu 5.19.0-45.46~22.04.1-generic 5.19.17
Uname: Linux 5.19.0-45-generic x86_64
NonfreeKernelModules: nvidia_modeset nvidia
ApportVersion: 2.20.11-0ubuntu82.5
Architecture: amd64
CasperMD5CheckResult: pass
CurrentDesktop: ubuntu:GNOME
Date: Thu Jun 29 09:06:10 2023
InstallationDate: Installed on 2023-02-19 (129 days ago)
InstallationMedia: Ubuntu 22.04.1 LTS "Jammy Jellyfish" - Release amd64 (20220809.1)
RebootRequiredPkgs: Error: path contained symlinks.
SourcePackage: nvidia-graphics-drivers-530
UpgradeStatus: No upgrade log present (probably fresh install)

[Daniel van Vugt (vanvugt)]

Thanks for the bug report. Next time the problem happens, please run:

  lspci -k > lspci.txt
  grep . /sys/class/drm/*/status > drmstatus.txt
  journalctl -b0 > journal.txt

or if you had to reboot then:

  journalctl -b-1 > prevboot.txt

Please then attach the resulting text files here.

[Andriy Khomych (andriy-khomych)]

I didn't make a reboot but attached it and Nvidia bugs report (sudo nvidia-bug-report.sh). All is archived.

[Daniel van Vugt (vanvugt)]

Thanks. I can see the external display is connected by HDMI to the NVIDIA GPU. And you seem to be using Xorg exclusively so this is a bug between Xorg and the NVIDIA driver.

I think it's likely the issue correlates with these error messages:

[drm:nv_drm_master_set [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00000100] Failed to grab modeset ownership

Because that means the NVIDIA driver has lost control of its displays (the external monitor).

As a possible workaround you might like to try adding a kernel parameter:

  nvidia_drm.modeset=0

which should at least disable the code path causing the above error message. But to use Wayland you will need to keep the default (nvidia_drm.modeset=1).

[Andriy Khomych (andriy-khomych)]

Thanks, I cannot remember why I used X11, switched to Wayland and it helped.
Thanks a lot!

[Daniel van Vugt (vanvugt)]

Beware: Wayland is slower with external monitors connected to NVIDIA (bug 1970291)

[Andriy Khomych (andriy-khomych)]

Thanks, I'll take this into account :)