Gnome/GDM glitches after suspending/sleep

dingaling · September 20, 2024, 3:09am

What is the expected outcome:
Typically if you click “suspend” on gnome to sleep your computer, if you turn it back on it will with no visual problems.

What output did you get
Extreme visual glitches on GDM and and DE. Only seems to be fixable with a full restart of the computer.

Meta information
OS: NixOS 24.11.20240906.574d1ea (Vicuna) x86_64
DE: GNOME 46.4
WM: Mutter (Wayland)
Kernel: Linux 6.11.0-cachyos
Nvidia driver version: 560.35.03

My nvidia.nix:

{ config, pkgs, ... }:

{
  hardware = {
    graphics.enable = true;
    nvidia = {
    modesetting.enable = true;
    powerManagement = {
      enable = false; # Disable power management to avoid potential issues
      finegrained = false;
    };
    open = false; # Use proprietary NVIDIA drivers
    nvidiaSettings = true; # Enable NVIDIA settings panel
    package = config.boot.kernelPackages.nvidiaPackages.beta; # Use beta drivers
    };
  };
}

My gdm-display-manager-config.nix:

{
  services.xserver.displayManager.gdm = {
  enable = true;
  wayland = true;
  autoSuspend = false;
  };

  systemd.tmpfiles.rules = [
    # Configuration for GDM monitor setup
    ''f+ /run/gdm/.config/monitors.xml - gdm gdm - <monitors version="2"><configuration><logicalmonitor><x>3840</x><y>0</y><scale>1</scale><monitor><monitorspec><connector>DP-1</connector><vendor>GSM</vendor><product>LG IPS FULLHD</product><serial>0x01010101</serial></monitorspec><mode><width>1920</width><height>1080</height><rate>60.000</rate></mode></monitor></logicalmonitor><logicalmonitor><x>1920</x><y>0</y><scale>1</scale><primary>yes</primary><monitor><monitorspec><connector>DP-2</connector><vendor>AUS</vendor><product>XG248Q</product><serial>KCLMRS019540</serial></monitorspec><mode><width>1920</width><height>1080</height><rate>239.964</rate></mode></monitor></logicalmonitor><logicalmonitor><x>0</x><y>0</y><scale>1</scale><monitor><monitorspec><connector>HDMI-1</connector><vendor>JXC</vendor><product>LED MONITOR</product><serial>0x00000000</serial></monitorspec><mode><width>1920</width><height>1080</height><rate>74.973</rate></mode></monitor></logicalmonitor></configuration></monitors>''
  ];
}

My gnome-desktop-config.nix:

{pkgs, ...}:
{
  services.xserver.desktopManager.gnome = {
  enable = true;
  extraGSettingsOverridePackages = [ pkgs.mutter ];
  extraGSettingsOverrides = ''
    [org.gnome.mutter.wayland]
    xwayland-allow-grabs=true
    xwayland-grab-access-rules=['parsecd']
    [org.gnome.mutter]
    experimental-features=['variable-refresh-rate']
  '';
  };
  
  environment.gnome.excludePackages = [ pkgs.epiphany ]; # Remove unwanted GNOME packages

}

Please let me know if there is any other information I can provide.

eljamm · September 20, 2024, 8:23am

According to the Wiki, you might need to enable hardware.nvidia.powerManagement or hardware.nvidia.powerManagement.finegrained to remove the graphical corruption.

dingaling · September 20, 2024, 12:16pm

I see, thank you. Both do not appear to be ideal however. The `hardware.nvidia.powerManagement does not seem to fix issues with suspending the desktop, instead the desktop just does not show a display. For the latter, is there a known decrease in performance using hardware.nvidia.powerManagement.finegrained ? I will opt into just avoiding suspending my desktop completely if it does have an impact on performance (e.g. for gaming).

eljamm · September 20, 2024, 3:09pm

Although I haven’t had corruption issues before, I have both options enabled and I haven’t noticed any decrease in performance. Did the finegrained option fix the corruption issue for you at the cost of performance?

dingaling · September 21, 2024, 3:03am

I am unable to enable the finegrained feature it seems. First I had an error that I needed to enable PRIME, so I did, but when I tried to build it said I needed to put my PCIE ID to enable it even though I have:

{ config, pkgs, ... }:

{
  hardware = {
    graphics.enable = true;
    nvidia = {
      modesetting.enable = true;
      prime = {
        offload.enable = true;
        nvidiaBusId = "PCI:26:0:0";
      };
      powerManagement = {
        enable = true; # Disable power management to avoid potential issues
        finegrained = true;
        };
      open = false; # Use proprietary NVIDIA drivers  
      nvidiaSettings = true; # Enable NVIDIA settings panel
      package = config.boot.kernelPackages.nvidiaPackages.beta; # Use beta drivers
    };
  };
}

I’m also worried that PRIME might be inappropriate to enable on my machine since it’s a desktop not a laptop.

eljamm · September 21, 2024, 9:17am

Ah, reading the Nvidia page on finegrained PM, turns out this feature is only supported for laptops. Could you enable hardware.nvidia.open and try again with and without powerManagement.enable?

dingaling · September 22, 2024, 3:13am

I have now tested both and they have identical results (visual corruption).

eljamm · September 22, 2024, 9:48am

Are you sure that the nvidia-{suspend,hibernate,resume} services are running when enabling powerManagement?

doronbehar · September 22, 2024, 6:35pm

I’m not using this option, but I too experienced issues in the past with Nvidia drivers and Gnome. So you might want to try linuxPackages.nvidia_x11_legacy535, that was added due to this issue reported in the ArchLinux forums.

dingaling · September 23, 2024, 9:55am

It seems to be running

~ ❱ systemctl status nvidia-{suspend,resume,hibernate}.service
○ nvidia-suspend.service - NVIDIA system suspend actions
     Loaded: loaded (/etc/systemd/system/nvidia-suspend.service; enabled; preset: enabled)
     Active: inactive (dead)

○ nvidia-resume.service - NVIDIA system resume actions
     Loaded: loaded (/etc/systemd/system/nvidia-resume.service; enabled; preset: enabled)
     Active: inactive (dead)

○ nvidia-hibernate.service - NVIDIA system hibernate actions
     Loaded: loaded (/etc/systemd/system/nvidia-hibernate.service; enabled; preset: enabled)
     Active: inactive (dead)
~ 1.1s | 3 ❱ journalctl -u nvidia-suspend.service -n 20
             journalctl -u nvidia-resume.service -n 20
             journalctl -u nvidia-hibernate.service -n 20
Sep 20 22:22:16 sengoku systemd[1]: Starting NVIDIA system suspend actions...
Sep 20 22:22:18 sengoku systemd[1]: nvidia-suspend.service: Deactivated successfully.
Sep 20 22:22:18 sengoku systemd[1]: Finished NVIDIA system suspend actions.
Sep 20 22:22:18 sengoku systemd[1]: nvidia-suspend.service: Consumed 1.403s CPU time, 815.6M memory peak.
-- Boot a8fd3111656f4ebdaba96504879c62aa --
Sep 22 13:00:45 sengoku systemd[1]: Starting NVIDIA system suspend actions...
Sep 22 13:00:48 sengoku systemd[1]: nvidia-suspend.service: Deactivated successfully.
Sep 22 13:00:48 sengoku systemd[1]: Finished NVIDIA system suspend actions.
Sep 22 13:00:48 sengoku systemd[1]: nvidia-suspend.service: Consumed 1.487s CPU time, 667.1M memory peak.
-- Boot 634d4dd4f5ca4013981661028e7b37d8 --
Sep 23 19:50:59 sengoku systemd[1]: Starting NVIDIA system suspend actions...
Sep 23 19:51:01 sengoku systemd[1]: nvidia-suspend.service: Deactivated successfully.
Sep 23 19:51:01 sengoku systemd[1]: Finished NVIDIA system suspend actions.
Sep 23 19:51:01 sengoku systemd[1]: nvidia-suspend.service: Consumed 1.389s CPU time, 785.2M memory peak.
Sep 20 22:24:17 sengoku systemd[1]: Starting NVIDIA system resume actions...
Sep 20 22:24:20 sengoku systemd[1]: nvidia-resume.service: Deactivated successfully.
Sep 20 22:24:20 sengoku systemd[1]: Finished NVIDIA system resume actions.
Sep 20 22:24:20 sengoku systemd[1]: nvidia-resume.service: Consumed 3.280s CPU time, 1.5M memory peak.
-- Boot a8fd3111656f4ebdaba96504879c62aa --
Sep 22 13:02:10 sengoku systemd[1]: Starting NVIDIA system resume actions...
Sep 22 13:02:13 sengoku systemd[1]: nvidia-resume.service: Deactivated successfully.
Sep 22 13:02:13 sengoku systemd[1]: Finished NVIDIA system resume actions.
Sep 22 13:02:13 sengoku systemd[1]: nvidia-resume.service: Consumed 2.371s CPU time, 1.5M memory peak.
-- Boot 634d4dd4f5ca4013981661028e7b37d8 --
Sep 23 19:52:44 sengoku systemd[1]: Starting NVIDIA system resume actions...
Sep 23 19:52:46 sengoku systemd[1]: nvidia-resume.service: Deactivated successfully.
Sep 23 19:52:46 sengoku systemd[1]: Finished NVIDIA system resume actions.
Sep 23 19:52:46 sengoku systemd[1]: nvidia-resume.service: Consumed 2.405s CPU time, 1.5M memory peak.
-- No entries --
~ 1.4s ❱

dingaling · September 23, 2024, 9:56am

Would this work if I am using Wayland? Just from the name it seems to be for x11 only.

doronbehar · September 23, 2024, 10:27am

Yes the choice of GPU drivers doesn’t even depend on actually showing graphical stuff - you can even use it with a pure command line setup and access it for computations only.

dingaling · September 30, 2024, 1:15pm

For now I will avoid suspending my computer since I need latest drivers to play games. Hopefully there is a fix in the near future.

eljamm · October 23, 2024, 7:41pm

@dingaling if you’re still having this issue, it might be worth it to try the beta drivers:

dingaling · October 26, 2024, 3:25am

Thank you for this. Would I need to do what this person has done? I am already using the beta drivers but I’m still on driver version 560.35.03.

eljamm · October 26, 2024, 8:32am

Yes, you can get the hashes as described in that thread and then set:

hardware.nvidia.package = config.boot.kernelPackages.nvidiaPackages.mkDriver {
  version = "565.57.01"; # beta
  sha256_64bit = "sha256-buvpTlheOF6IBPWnQVLfQUiHv4GcwhvZW3Ks0PsYLHo=";
  sha256_aarch64 = "sha256-aDVc3sNTG4O3y+vKW87mw+i9AqXCY29GVqEIUlsvYfE=";
  openSha256 = "sha256-/tM3n9huz1MTE6KKtTCBglBMBGGL/GOHi5ZSUag4zXA=";
  settingsSha256 = "sha256-H7uEe34LdmUFcMcS6bz7sbpYhg9zPCb/5AmZZFTx1QA=";
  persistencedSha256 = "sha256-hdszsACWNqkCh8G4VBNitDT85gk9gJe1BlQ8LdrYIkg=";
};

I think you meant 560.31.02 and not 560.35.03, as the latter is the latest stable, not beta.

dingaling · October 26, 2024, 11:33am

I think you meant 560.31.02 and not 560.35.03, as the latter is the latest stable, not beta.

Huh, this is strange. I did not notice the versions being wrong. I seem to have the stable versions even though I’ve declared the beta version of the drivers.

{ config, pkgs, ... }:

{
  hardware = {
    graphics = {
      enable = true;
      enable32Bit = true;
      extraPackages = with pkgs; [ libva vaapiVdpau ];
    };
    nvidia = {
    modesetting.enable = true;
    powerManagement = {
      enable = false; # Disable power management to avoid potential issues
      finegrained = false;
    };
    open = false; # Use proprietary NVIDIA drivers
    nvidiaSettings = true; # Enable NVIDIA settings panel
    package = config.boot.kernelPackages.nvidiaPackages.beta; # Use beta drivers
    };
  };
}

Yes, you can get the hashes as described in that thread

Ok I will try this and let you know in this thread

MichaelOwenDyer · February 23, 2025, 8:02pm

Hi @dingaling, did you ever manage to get to the bottom of this? I am struggling with the same issue.

The relevant section of my configuration is here.

And this is the driver version I currently have installed:

> cat /proc/driver/nvidia/version
NVRM version: NVIDIA UNIX Open Kernel Module for x86_64  565.77  Release Build  (nixbld@)  
GCC version:  gcc version 14.2.1 20241116 (GCC)

I previously was using config.boot.kernelPackages.nvidiaPackages.stable, which I believe was driver version 550.142. Before the switch I was seeing major graphical glitches after suspend, like you describe. On the newer drivers I only see a blank screen after suspend and I am forced to cut power to the machine to restart it.

dingaling · March 14, 2025, 10:56pm

No, this is still unsolved for me as well sadly

fareycircles · April 2, 2025, 7:52am

Joining this thread because I’m having a similar issue! It would be interesting to know your GPU model, @dingaling, if you’re willing to share. Mine is a GeForce GTX 1060 3GT OC.