Gnome apps segfault in Nvidia (535) Wayland sessions on Noble, but 550 works

Bug #2063827 reported by Claudiu Constantin Cojocaru
72
This bug affects 12 people
Affects Status Importance Assigned to Milestone
nvidia-graphics-drivers-535 (Ubuntu)
Status tracked in Oracular
Noble
Fix Committed
High
Alessandro Astone
Oracular
Fix Committed
High
Alessandro Astone

Bug Description

[ Impact ]

Graphical programs in the GNOME Wayland session crash in a Ubuntu 24.04 system with NVIDIA as the sole GPU and the NVIDIA 535 driver.

[ Test Plan ]

 * Log into the Wayland session from a machine with an NVIDIA GPU and no other GPU (such as integrated graphics, or iGPU).

 * Run a graphical application such as `gnome-control-center` or `nautilus`

 * Verify that the application opens without crashing.

[ Where problems could occur ]

The fix should be distributed as a simple new library dependency, with low risk of side effects.

[ Original description ]

Hello, I just installed Ubuntu 24.04 LTS and I realized that Gnome Control Center does not open on Wayland, but does on Xorg, attempting to open it will simply open the crash dialog after some time.

When running from the terminal:

$ gnome-control-center
libEGL warning: egl: failed to create dri2 screen
Segmentation fault (core dumped)

Version installed:

gnome-control-center:
  Installed: 1:46.0.1-1ubuntu7
  Candidate: 1:46.0.1-1ubuntu7

My current installation is very lightweight, and I haven't removed any packages or changed any system files.

This occurs on a Nvidia GPU, specifically the NVIDA GeForce 1060 3GB with the driver version being 535.171.04

If more information is needed, please let me know, however I might be slow to respond.

Thank you.

Revision history for this message
Sebastien Bacher (seb128) wrote :

Thank you for taking the time to report this bug and helping to make Ubuntu better. Please try to obtain a backtrace following the instructions at http://wiki.ubuntu.com/DebuggingProgramCrash and upload the backtrace (as an attachment) to the bug report. This will greatly help us in tracking down your problem.

Changed in gnome-control-center (Ubuntu):
status: New → Incomplete
Revision history for this message
Claudiu Constantin Cojocaru (rexase) wrote :

Interesting, after getting the backtrace on Wayland I tried to attach the file, however this bug also seems to affect Nautilus as well, the browser will try to open Nautilus and fail. I had to switch back to Xorg to attach the report. Trying to open Nautilus from the terminal gives exactly the same error as trying to open Gnome Control Center.

I assume that the backtrace for Gnome Control Center is enough, but if it's necessary I will also provide the backtrace for Nautilus if requested.

Thank you for your time.

Revision history for this message
Sebastien Bacher (seb128) wrote :

Thanks, the backtrace indicates a bug in the nvidia driver

#5 0x00007fffcc201e0d in ProducerCleanup ()
   from /lib/x86_64-linux-gnu/libnvidia-vulkan-producer.so
No symbol table info available.
#6 0x00007fffcc2020db in ProducerInit ()
   from /lib/x86_64-linux-gnu/libnvidia-vulkan-producer.so
No symbol table info available.
#7 0x00007fffb5e1a0e2 in ?? ()
   from /lib/x86_64-linux-gnu/libnvidia-glcore.so.535.171.04
No symbol table info available.
#8 0x00007fffb5e4442f in ?? ()
   from /lib/x86_64-linux-gnu/libnvidia-glcore.so.535.171.04
No symbol table info available.
#9 0x00007fffb5e6844d in ?? ()
   from /lib/x86_64-linux-gnu/libnvidia-glcore.so.535.171.04
No symbol table info available.
#10 0x00007fffec4a9b50 in ?? () from /lib/x86_64-linux-gnu/libGLX_nvidia.so.0

Changed in gnome-control-center (Ubuntu):
status: Incomplete → New
affects: gnome-control-center (Ubuntu) → nvidia-graphics-drivers-535 (Ubuntu)
Revision history for this message
Sebastien Bacher (seb128) wrote :

Could you try if the 550 drivers behaves any better and also provide details on your videocard?

Revision history for this message
Claudiu Constantin Cojocaru (rexase) wrote :

The 550 drivers seem to solve this issue completely, currently using version 550.67.

If it's still relevant, here's the output of hwinfo --gfxcard:

15: PCI 100.0: 0300 VGA compatible controller (VGA)
  [Created at pci.386]
  Unique ID: VCu0.AhXiurDHit1
  Parent ID: vSkL.Hv91_qPnDS1
  SysFS ID: /devices/pci0000:00/0000:00:01.0/0000:01:00.0
  SysFS BusID: 0000:01:00.0
  Hardware Class: graphics card
  Model: "nVidia GP106 [GeForce GTX 1060 3GB]"
  Vendor: pci 0x10de "nVidia Corporation"
  Device: pci 0x1c02 "GP106 [GeForce GTX 1060 3GB]"
  SubVendor: pci 0x174b "PC Partner Limited / Sapphire Technology"
  SubDevice: pci 0x2438
  Revision: 0xa1
  Driver: "nvidia"
  Driver Modules: "nvidia"
  Memory Range: 0xde000000-0xdeffffff (rw,non-prefetchable)
  Memory Range: 0xc0000000-0xcfffffff (ro,non-prefetchable)
  Memory Range: 0xd0000000-0xd1ffffff (ro,non-prefetchable)
  I/O Ports: 0xe000-0xe07f (rw)
  Memory Range: 0x000c0000-0x000dffff (rw,non-prefetchable,disabled)
  IRQ: 137 (36587 events)
  Module Alias: "pci:v000010DEd00001C02sv0000174Bsd00002438bc03sc00i00"
  Driver Info #0:
    Driver Status: nvidiafb is not active
    Driver Activation Cmd: "modprobe nvidiafb"
  Driver Info #1:
    Driver Status: nouveau is not active
    Driver Activation Cmd: "modprobe nouveau"
  Driver Info #2:
    Driver Status: nvidia_drm is active
    Driver Activation Cmd: "modprobe nvidia_drm"
  Driver Info #3:
    Driver Status: nvidia is active
    Driver Activation Cmd: "modprobe nvidia"
  Config Status: cfg=new, avail=yes, need=no, active=unknown
  Attached to: #12 (PCI bridge)

Thank you very much for your time.

Revision history for this message
Launchpad Janitor (janitor) wrote :

Status changed to 'Confirmed' because the bug affects multiple users.

Changed in nvidia-graphics-drivers-535 (Ubuntu):
status: New → Confirmed
Revision history for this message
John Doe (imakemalware) wrote (last edit ):

I want to add: I faced a similar issue. Though updating to the 550 drivers through apt seemingly fixed the issue, there was actually more issues at play.

First and foremost, on the Wayland session, running "glxinfo | egrep "OpenGL vendor|OpenGL renderer" returns:
"OpenGL vendor string: Mesa
OpenGL renderer string: llvmpipe (LLVM 17.0.6, 256 bits)"

Upon some further research, I learned that the renderer being "llvmpipe" means that the nvidia driver isn't properly working. I'll note, also, that I could only install the driver using apt, as the ubuntu-drivers utility didn't show any drivers newer than 535. I can't tell if the driver installation broke, but nvidia-smi does return the expected output. And on XOrg, glxinfo mentions the nvidia drivers and my graphics card (RTX 4080).

I don't know if this implies that the nvidia drivers are currently bugged on wayland.

Again, I had a similar issue as OP upon a fresh install. All the gnome apps crashed (settings, the first-installation greeter) and going through the errors, it listed a segmentation fault (I'm sorry, but I can't reproduce it now).

Revision history for this message
Sebastien Bacher (seb128) wrote :

> I don't know if this implies that the nvidia drivers are currently bugged on wayland.

Unsure if it would be right to call them buggy, but we still default to xorg instead of wayland for nvidia because there are known issues with nvidia on wayland

Revision history for this message
Nir Nachmani Major (mazemeshane) wrote :

I have the same issue with NVIDA GeForce 4080 and driver 535.171.04 on Ubuntu 24.04 with Wayland. Problem is that if I update to a newer driver (which fixes the issue) I get another bug with letters and icons disappearing after return from suspend (described here: https://askubuntu.com/questions/1512797/most-characters-disappearing-in-gnome-after-suspend/1516115#1516115)

Revision history for this message
Daniel van Vugt (vanvugt) wrote :

That last issue is bug 1876632.

tags: added: noble nvidia nvidia-wayland
Revision history for this message
Daniel van Vugt (vanvugt) wrote :

See also bug 2067872.

Revision history for this message
Daniel van Vugt (vanvugt) wrote :

Sorry I didn't see this bug earlier. For now I'm going to group it with bug 2067872 because that one is more generic and has more relevant links.

Revision history for this message
Daniel van Vugt (vanvugt) wrote :

Actually there is more relevant detail here so let's keep this one.

summary: - Gnome Control Center fails to open on Wayland
+ Gnome apps segfault in Nvidia Wayland sessions on Noble
Changed in egl-wayland (Ubuntu):
status: New → Confirmed
Revision history for this message
Jonathan Cave (jocave) wrote : Re: Gnome apps segfault in Nvidia Wayland sessions on Noble

I have now tried nvidia driver 550 (which isn't currently suggested by Additional Drivers). I found that gnome apps no longer crash, but I did notice the following messages are printed when calling from commandline:

❯ gnome-control-center
MESA: error: ZINK: failed to choose pdev
libEGL warning: egl: failed to create dri2 screen

I have also experienced a problem on a few boots where no desktop manager is displayed and I have to reset from power button and make another boot attempt to bring up the login screen.

Revision history for this message
Daniel van Vugt (vanvugt) wrote :

Looks like there's one root cause in mesa/egl-wayland/GNOME causing that Zink error and the original crash here. Because I don't think(?) Nvidia should be trying to use Zink, and if it didn't then the original crash also would not happen.

Revision history for this message
Daniel van Vugt (vanvugt) wrote :

Just tested again. Switching back to Nvidia 535.171.04-0ubuntu2 on Noble I don't get any crashes in GNOME apps. Definitely using Wayland. They all just work.

I wonder why it works for some people but not others. Maybe the driver installation method? I ended up installing my driver using 'sudo apt install nvidia-driver-535' on the command line because the Additional Drivers GUI was so slow it felt broken. I suspect I was just too impatient. But then I did hit bug 2061738 and had to figure out the workaround for that.

Revision history for this message
Daniel van Vugt (vanvugt) wrote :

Looks like this bug has only ever been observed with AMD CPUs, which might be a factor.

Also I wonder if the Nvidia driver misconfiguration might be caused by third party software like Steam?

Revision history for this message
Jonathan Cave (jocave) wrote :

In my specific case I've never had Steam installed, can't really think of any third apps that should impact the driver on my install.

Revision history for this message
Nir Nachmani Major (mazemeshane) wrote :

I have Intel CPU and have never installed Steam. I switched to Xorg for now because between this bug and bug 1876632, Wayland is not usable.

Revision history for this message
Daniel van Vugt (vanvugt) wrote :

If you add GSK_RENDERER=gl to /etc/environment and reboot, does it improve the situation?

Revision history for this message
Marat Fatkullin (marat123) wrote :

> If you add GSK_RENDERER=gl to /etc/environment and reboot, does it improve the situation?

no.

also just update on 535.183.01 - there the same - anyway gnome apps crushes under wayland session.

Revision history for this message
Alessandro Astone (aleasto) wrote :

I hit this bug when installing 24.04 by selecting the option in the installer to install proprietary drivers.

`libnvidia-egl-wayland1` is not installed, which causes apps that try to use GPU acceleration to crash on startup. Installing that library solves the issue.

We should make sure that this gets always installed -- regardless of installation method -- such as by making it a required dependency of `libnvidia-gl-535` (and every other version)

Revision history for this message
Daniel van Vugt (vanvugt) wrote :

It would make sense that we might forget to install that in the live session, but not forget in the 'Additional Drivers' app.

tags: added: udeng-3440
Revision history for this message
Alessandro Astone (aleasto) wrote :

BTW, the reason this actually crashes is because since mesa 23.3 the Zink driver is the first fallback in mesa. After the nvidia driver fails to load due to missing egl-wayland, the next EGL driver to try is Zink and unfortunately that crashes with nvidia 535. With nvidia 550 the Zink crash seems to be resolved, so applications work again - but through Zink+nvidia-vulkan which is not the optimal path today.

Prior to mesa 23.3 this would fall back to llvmpipe instead.

Revision history for this message
Daniel van Vugt (vanvugt) wrote (last edit ):

Oh it wouldn't be the installer's fault. More likely my dev machine always worked around the issue because in order to build Mutter I have always:

  sudo apt-get build-dep mutter

So I was getting libnvidia-egl-wayland-dev and its dependencies installed.

Everyone else should just use:

  sudo apt install libnvidia-egl-wayland1

to solve the bug.

Changed in egl-wayland (Ubuntu):
status: Confirmed → Invalid
no longer affects: egl-wayland (Ubuntu)
summary: - Gnome apps segfault in Nvidia Wayland sessions on Noble
+ Gnome apps segfault in Nvidia (535) Wayland sessions on Noble, but 550
+ works
tags: added: performance
Revision history for this message
Daniel van Vugt (vanvugt) wrote (last edit ):

> With nvidia 550 the Zink crash seems to be resolved, so applications work again -
> but through Zink+nvidia-vulkan which is not the optimal path today.

I think we need a new bug to track that in both 550 and 535, which would be resolved by the same fix as this except that the 550 driver also needs patching.

Changed in nvidia-graphics-drivers-535 (Ubuntu):
assignee: nobody → Alessandro Astone (aleasto)
no longer affects: ubuntu-drivers-common (Ubuntu Noble)
no longer affects: ubuntu-drivers-common (Ubuntu Oracular)
no longer affects: ubuntu-drivers-common (Ubuntu)
Changed in nvidia-graphics-drivers-535 (Ubuntu Oracular):
importance: Undecided → High
Changed in nvidia-graphics-drivers-535 (Ubuntu Noble):
importance: Undecided → High
assignee: nobody → Alessandro Astone (aleasto)
status: New → Triaged
Changed in nvidia-graphics-drivers-535 (Ubuntu Oracular):
status: Confirmed → Triaged
milestone: none → ubuntu-24.10
Changed in nvidia-graphics-drivers-535 (Ubuntu Noble):
milestone: none → ubuntu-24.04.1
Revision history for this message
Daniel van Vugt (vanvugt) wrote :

The general case also showing that 550 needs fixing is being tracked in bug 2062082.

Revision history for this message
Daniel van Vugt (vanvugt) wrote :
Changed in nvidia-graphics-drivers-535 (Ubuntu Oracular):
status: Triaged → In Progress
Revision history for this message
Daniel van Vugt (vanvugt) wrote :

Bug 2062082 tracks the fix for this, so just copy that status.

Changed in nvidia-graphics-drivers-535 (Ubuntu Oracular):
status: In Progress → Fix Committed
Changed in nvidia-graphics-drivers-535 (Ubuntu Noble):
status: Triaged → Fix Committed
description: updated
description: updated
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.