[SOLVED] Many GNOME applications (+ using the portal) fail to start

lama_noob · 2026-05-12 08:50:34

Hello,

The situation is probably related to mesa/nouveau, but since it seems to only apply to GNOME-related environments/applications, I'm posting in this forum. Feel free to move the topic (or to tell me to repost on an appropriate subforum) if necessary.

I run a laptop with an NVIDIA GPU, ArchLinux, using nouveau. My desktop environment is GNOME. Everything is up-to-date.

Here is info on the GPU:

lspci -k | grep Ada
01:00.0 VGA compatible controller: NVIDIA Corporation AD107GLM [RTX 1000 Ada Generation Laptop GPU] (rev a1)

When booting the computer, everything runs smoothly, with no bugs.
After some time (around 2 hours today), everything GNOME-related fails, except the graphical environment itself. This concerns:
- GNOME-related applications fail to launch. This concerns utility apps like the calculator, but also the image viewer (loupe), the file viewer (nautilus) or the console. Other applications (terminator, firefox, thunderbird, texstudio...) launch correctly.
- Applications that seem to use a GNOME-related "file browser" fail to load it. So, for recent versions of Firefox/Thunderbird, when trying to download or upload a file on my computer, the interface does not load. So I am unable to download/upload, attach files to mails. The application freezes until I press "Escape" enough to cancel. Applications using other interfaces do not exhibit this problem (for example, the open/save file for my LaTeX editor TeXstudio still works).
- The display itself has no problem: I can still move windows between screens, etc.

I tried launching loupe from my terminal, and got the following error message (same for nautilus):

(loupe:50104): Gdk-WARNING **: 10:24:10.708: Vulkan: ../mesa-26.0.6/src/nouveau/vulkan/nvkmd/nouveau/nvkmd_nouveau_ctx.c:149: DRM_NOUVEAU_EXEC failed: Aucun périphérique de ce type (VK_ERROR_DEVICE_LOST)

(loupe:50104): Gsk-WARNING **: 10:24:10.708: vkQueueSubmit(): The logical or physical device has been lost. (VK_ERROR_DEVICE_LOST) (-4)

(loupe:50104): Gdk-WARNING **: 10:24:10.708: Vulkan: ../mesa-26.0.6/src/nouveau/vulkan/nvk_queue.c:306: Submit failed (VK_ERROR_DEVICE_LOST)

(loupe:50104): Gdk-WARNING **: 10:24:10.708: vkQueuePresentKHR(): The logical or physical device has been lost. (VK_ERROR_DEVICE_LOST) (-4)

The terminal then hangs until I Ctrl-C. The same error happens for trying to launch nautilus.

A few notes:
- As stated above, this happens after my computer has been running for some time, but not very long. I am not doing much in the meantime: using firefox, thunderbird and texstudio. The laptop does not go to sleep or anything. It just... happens after some time.
- The problem happens on my work computer, which has a GPU. On my home computer (same setup but no GPU), no problem arises.
- This has been ongoing for a few... weeks? months? But I never took time to investigate or ask for help...

I can provide any additional detail if required.
Thanks in advance.

EDIT: Seems to be solved!
The problem arose from gnome not using the correct driver. And nouveau seems to die ~2 hours into the session (not sure why).
Installing vulkan-intel seems to have fixed the problem: gnome uses the correct driver, and everything seems to work fine.
Read the whole thread for full context/discussion.
Thanks seth!

Last edited by lama_noob (2026-05-13 13:30:13)

seth · 2026-05-12 09:59:23

vulkaninfo --summary
gtk4-demo
GSK_RENDERER=gl gtk4-demo
GSK_RENDERER=cairo gtk4-demo

lama_noob · 2026-05-12 11:49:05

Hi,

vulkaninfo --summary
ERROR at /usr/src/debug/vulkan-tools/Vulkan-Tools/vulkaninfo/./vulkaninfo.h:249:vkGetPhysicalDeviceDisplayPlanePropertiesKHR failed with ERROR_OUT_OF_HOST_MEMORY

and the program ends correctly (meaning that there is no further error, the terminal does not freeze/hang, etc).

gtk4-demo runs into this and then keeps running but no further output appears, until I kill it.

gtk4-demo

(gtk4-demo:79476): Gdk-WARNING **: 12:06:17.326: Vulkan: ../mesa-26.0.6/src/nouveau/vulkan/nvkmd/nouveau/nvkmd_nouveau_ctx.c:149: DRM_NOUVEAU_EXEC failed: Aucun périphérique de ce type (VK_ERROR_DEVICE_LOST)

(gtk4-demo:79476): Gsk-WARNING **: 12:06:17.326: vkQueueSubmit(): The logical or physical device has been lost. (VK_ERROR_DEVICE_LOST) (-4)

(gtk4-demo:79476): Gdk-WARNING **: 12:06:17.326: Vulkan: ../mesa-26.0.6/src/nouveau/vulkan/nvk_queue.c:306: Submit failed (VK_ERROR_DEVICE_LOST)

(gtk4-demo:79476): Gdk-WARNING **: 12:06:17.326: vkQueuePresentKHR(): The logical or physical device has been lost. (VK_ERROR_DEVICE_LOST) (-4)

^C

For the other two

GSK_RENDERER=gl gtk4-demo
GSK_RENDERER=cairo gtk4-demo

the window appears (I don't know if I should run the tests or if the goal is to have the tests appear).

Last edited by lama_noob (2026-05-12 11:49:33)

seth · 2026-05-12 12:08:05

the goal is to have the tests appear

This.

GSK defaults to vulkan and you your problem is

failed with ERROR_OUT_OF_HOST_MEMORY

Not sure how reliable "VK_ERROR_DEVICE_LOST" is.

eglinfo -B
ls -l /usr/share/vulkan/icd.d/

Is it sufficient to restart gnome to get out of this situation?

Last edited by seth (2026-05-12 12:08:24)

lama_noob · 2026-05-12 12:36:39

seth wrote:

GSK defaults to vulkan and you your problem is
failed with ERROR_OUT_OF_HOST_MEMORY
Not sure how reliable "VK_ERROR_DEVICE_LOST" is.

I see! Thanks!

eglinfo -B displays this (and already takes a few seconds to output past "GBM platform"), and then nothing happens until I kill it...

GBM platform:
EGL API version: 1.5
EGL vendor string: Mesa Project
EGL version string: 1.5
EGL client APIs: OpenGL OpenGL_ES 
OpenGL core profile vendor: Mesa
OpenGL core profile renderer: zink Vulkan 1.4(NVIDIA RTX 1000 Ada Generation Laptop GPU (NVK AD107) (MESA_NVK))
OpenGL core profile version: 4.6 (Core Profile) Mesa 26.0.6-arch1.1
OpenGL core profile shading language version: 4.60
MESA: error: ZINK: vkQueueSubmit failed (VK_ERROR_DEVICE_LOST)
OpenGL compatibility profile vendor: Mesa
OpenGL compatibility profile renderer: zink Vulkan 1.4(NVIDIA RTX 1000 Ada Generation Laptop GPU (NVK AD107) (MESA_NVK))
OpenGL compatibility profile version: 4.6 (Compatibility Profile) Mesa 26.0.6-arch1.1
OpenGL compatibility profile shading language version: 4.60

^C

As for

ls -l /usr/share/vulkan/icd.d/

total 4
-rw-r--r-- 1 root root 139 30 avril 03:06 nouveau_icd.json

seth wrote:

Is it sufficient to restart gnome to get out of this situation?

Unfortunately, I am using Wayland, and from what I know I have no way of restarting gnome without rebooting (which is a temporary fix, since the problem only arises after a few hours).

seth · 2026-05-12 12:42:39

Unfortunately, I am using Wayland

I have no way of restarting gnome without rebooting

Logging out and back in will get you a new gnome session and restarting the graphical.target will also restart GDM, this has nothing to do with wayland.

nothing happens until I kill it

MESA: error: ZINK: vkQueueSubmit failed (VK_ERROR_DEVICE_LOST)

Please post your complete system journal for the boot:

sudo journalctl -b | curl -s -H "Accept: application/json, */*" --upload-file - 'https://paste.c-net.org/'

Sanity check: using the nvidia driver is deliberately not an option?

lama_noob · 2026-05-12 13:10:33

seth wrote:

Logging out and back in will get you a new gnome session and restarting the graphical.target will also restart GDM, this has nothing to do with wayland.

I was mistaken, good to know. Unfortunately, this led to a full black screen where the only existing thing was my mouse pointer (without any lag/freeze, I must say!). I had to force restart (partition clean at reboot). So now I won't have the issue for a bit of time. ^^

seth wrote:

Please post your complete system journal for the boot:
sudo journalctl -b | curl -s -H "Accept: application/json, */*" --upload-file - 'https://paste.c-net.org/'

I was able to get it before killing: https://paste.c-net.org/IncurRebound
EDIT: Removed after resolution.

seth wrote:

Sanity check: using the nvidia driver is deliberately not an option?

So, philosophically I would rather not use the nvidia (or nvidia-open, since it is the supported one if I followed correctly?) drivers. Practically, I am a bit afraid of switching, in particular since the issue is annoying but not destructive (it is a work computer, so I would like to avoid the risk of having to manually repair everything should there be an issue). And also a bit annoyed in advance when nvidia will stop supporting my GPU and I would again have to switch...
However, if most advice is to switch to the nvidia driver, I will give up on nouveau.

Last edited by lama_noob (2026-05-13 14:02:31)

tekstryder · 2026-05-12 13:19:11

lama_noob wrote:

when nvidia will stop supporting my GPU(

(Aside from thread topic): 580 series driver will be officially supported until June 2028.

• https://docs.nvidia.com/datacenter/tesl … sions.html

Given the history of patches to older, unsupported Nvidia hardware/drivers still existing today, I'd expect functionality for many years after official support has ended.

There will likely be Arch AUR packages with patches for future kernels, etc.

EDIT: Actually, your GPU is also supported by the current Nvidia driver.

Last edited by tekstryder (2026-05-12 13:24:07)

seth · 2026-05-12 13:31:01

mai 12 08:21:02 23-1471A24082-P kernel: nouveau 0000:01:00.0: [drm] Cannot find any crtc or sizes

Hybrid graphics, there're no outputs attached to the nvidia driver

mai 12 08:21:07 23-1471A24082-P gnome-shell[1028]: Added device '/dev/dri/card1' (i915) using atomic mode setting.
mai 12 08:21:14 23-1471A24082-P gnome-shell[4082]: GPU /dev/dri/card1 selected primary from builtin panel presence

and gnome runs on the IGP, install https://archlinux.org/packages/extra/x8 … kan-intel/

mai 12 10:18:31 23-1471A24082-P kernel: nouveau 0000:01:00.0: loupe[47687]: job timeout, channel 4 killed!
mai 12 10:18:38 23-1471A24082-P kernel: nouveau 0000:01:00.0: gsp: rc engn:00000001 chid:4 gfid:0 level:2 type:38 scope:1 part:233 fault_addr:0000000000000000 fault_type:00000000
mai 12 10:18:38 23-1471A24082-P kernel: nouveau 0000:01:00.0: fifo:c00000:0004:0004:[loupe[47687]] errored - disabling channel
mai 12 10:18:46 23-1471A24082-P gnome-shell[4082]: meta_window_set_stack_position_no_sync: assertion 'window->stack_position >= 0' failed
mai 12 10:21:10 23-1471A24082-P NetworkManager[829]: <info>  [1778574070.3672] dhcp4 (wlp0s20f3): state changed new lease, address=10.160.17.216
mai 12 10:21:10 23-1471A24082-P systemd[1]: Starting Network Manager Script Dispatcher Service...
mai 12 10:21:10 23-1471A24082-P systemd[1]: Started Network Manager Script Dispatcher Service.
mai 12 10:21:20 23-1471A24082-P systemd[1]: NetworkManager-dispatcher.service: Deactivated successfully.
mai 12 10:24:10 23-1471A24082-P kernel: nouveau 0000:01:00.0: loupe[50104]: job timeout, channel 4 killed!
mai 12 10:24:16 23-1471A24082-P kernel: nouveau 0000:01:00.0: gsp: rc engn:00000001 chid:4 gfid:0 level:2 type:38 scope:1 part:233 fault_addr:0000000000000000 fault_type:00000000
mai 12 10:24:16 23-1471A24082-P kernel: nouveau 0000:01:00.0: fifo:c00000:0004:0004:[loupe[50104]] errored - disabling channel

nouveau dies ~2h into the session, might be the GSP.

However, if most advice is to switch to the nvidia driver

If you were to rely on the chip, maybe. Probably.
But for now keeping things on the IGP might easily suffice.

lama_noob · 2026-05-12 13:45:34

tekstryder wrote:

(Aside from thread topic): 580 series driver will be officially supported until June 2028.
• https://docs.nvidia.com/datacenter/tesl … sions.html
Given the history of patches to older, unsupported Nvidia hardware/drivers still existing today, I'd expect functionality for many years after official support has ended.
There will likely be Arch AUR packages with patches for future kernels, etc.
EDIT: Actually, your GPU is also supported by the current Nvidia driver.

Oh, I know, this is the least important reason. Really, this is more about avoiding nvidia proprietary parts (which, from what I gathered, are in the drivers) + not having to manage a full switch. Thanks!

seth wrote:

and gnome runs on the IGP, install https://archlinux.org/packages/extra/x8 … kan-intel/

Alright, do you advise also installing vulkan-mesa-layers?

seth wrote:

nouveau dies ~2h into the session, might be the GSP.

This matches my experience. Maybe disabling the GSP would help then?

seth wrote:

If you were to rely on the chip, maybe. Probably.
But for now keeping things on the IGP might easily suffice.

Hopefully!

Thanks a lot, I will report later/tomorrow (or next week, we have a long week-end here in France ).

seth · 2026-05-12 13:53:04

Alright, do you advise also installing vulkan-mesa-layers?

No, just make sure that the gtk clients now fully run on the intel GPU (nvtop/inte_gpu_top)

Maybe disabling the GSP would help then?

This is only an option w/ the closed-source 580xx nvidia drivers, nvidia-open and nouveau rely on the GSP

we have a long week-end here in France

Vatertag is upon us all

lama_noob · 2026-05-12 14:00:50

seth wrote:

No, just make sure that the gtk clients now fully run on the intel GPU (nvtop/inte_gpu_top)

Alright! Just to make sure, this is again a journalctl -b where the lines:

mai 12 08:21:07 23-1471A24082-P gnome-shell[1028]: Added device '/dev/dri/card1' (i915) using atomic mode setting.
mai 12 08:21:14 23-1471A24082-P gnome-shell[4082]: GPU /dev/dri/card1 selected primary from builtin panel presence

would appear with the other device? (And probably at next boot!)

seth wrote:

This is only an option w/ the closed-source 580xx nvidia drivers, nvidia-open and nouveau rely on the GSP

Alright, thanks! I should have insisted for a work computer without a GPU... (esp. an nvidia one)

seth wrote:

Vatertag is upon us all

L'Ascension and its famed "bridge". (same date, different meaning XD)

Last edited by lama_noob (2026-05-12 14:07:16)

seth · 2026-05-12 17:57:46

The card order is not 100% stable, you want to make sure that gnome uses the one for the i915 module.

The holiday doubles as Fête des Pères in Germany - the weather forecast looks bleak, but usually it's an execution for all men (who is to say that you don't have a child somewhere…) to stroll around, pulling a keg of beer in a hand-wagon and get mindlessly drunk in the sun

lama_noob · 2026-05-13 07:03:36

seth wrote:

The card order is not 100% stable, you want to make sure that gnome uses the one for the i915 module.

From this morning boot, I get this:

mai 13 08:53:46 23-1471A24082-P gnome-shell[1034]: Thread 'KMS thread' will be using high priority scheduling
mai 13 08:53:46 23-1471A24082-P gnome-shell[1034]: Device '/dev/dri/card1' prefers shadow buffer
mai 13 08:53:46 23-1471A24082-P gnome-shell[1034]: Added device '/dev/dri/card1' (nouveau) using non-atomic mode setting.
mai 13 08:53:46 23-1471A24082-P gnome-shell[1034]: Device '/dev/dri/card2' prefers shadow buffer
mai 13 08:53:46 23-1471A24082-P gnome-shell[1034]: Added device '/dev/dri/card2' (i915) using atomic mode setting.
mai 13 08:53:46 23-1471A24082-P gnome-shell[1034]: Created gbm renderer for '/dev/dri/card1'
mai 13 08:53:46 23-1471A24082-P gnome-shell[1034]: Created gbm renderer for '/dev/dri/card2'
mai 13 08:53:46 23-1471A24082-P gnome-shell[1034]: GPU /dev/dri/card2 selected primary from builtin panel presence
mai 13 08:53:46 23-1471A24082-P gnome-shell[1034]: Obtained a high priority EGL context

Funnily it appears again 10 seconds later...

mai 13 08:53:56 23-1471A24082-P gnome-shell[4014]: Thread 'KMS thread' will be using high priority scheduling
mai 13 08:53:56 23-1471A24082-P gnome-shell[4014]: Device '/dev/dri/card1' prefers shadow buffer
mai 13 08:53:56 23-1471A24082-P gnome-shell[4014]: Added device '/dev/dri/card1' (nouveau) using non-atomic mode setting.
mai 13 08:53:56 23-1471A24082-P gnome-shell[4014]: Device '/dev/dri/card2' prefers shadow buffer
mai 13 08:53:56 23-1471A24082-P gnome-shell[4014]: Added device '/dev/dri/card2' (i915) using atomic mode setting.
mai 13 08:53:56 23-1471A24082-P gnome-shell[4014]: Created gbm renderer for '/dev/dri/card1'
mai 13 08:53:56 23-1471A24082-P gnome-shell[4014]: Created gbm renderer for '/dev/dri/card2'
mai 13 08:53:56 23-1471A24082-P gnome-shell[4014]: GPU /dev/dri/card2 selected primary from builtin panel presence
mai 13 08:53:56 23-1471A24082-P gnome-shell[4014]: Obtained a high priority EGL context

There is no error inbetween. Maybe the first one is the displayed login page, and the second one is when I logged in? Who knows!
I will report with whether the problem appears again. Thanks again!

seth wrote:

The holiday doubles as Fête des Pères in Germany - the weather forecast looks bleak, but usually it's an execution for all men (who is to say that you don't have a child somewhere…) to stroll around, pulling a keg of beer in a hand-wagon and get mindlessly drunk in the sun

Good to know! Here it is mostly a way to see family, since it is always a Thursday and many people can also take Friday as a holiday.
I hope you'll have good weather, for us it looks bad!

seth · 2026-05-13 07:15:24

Funnily it appears again 10 seconds later...

The first gnome shell is GDM and the second one your session.

I hope you'll have good weather, for us it looks bad!

Cloudy, rainy, possible thunderstorms, 11°C - everyone out tomorrow to get drunk has probably have a serious alcohol problem.

lama_noob · 2026-05-13 13:23:43

seth wrote:

The first gnome shell is GDM and the second one your session.

Yes, this is what I thought.
Well, the problem seems to be solved since it has been more than two hours since boot and the problem has not arisen again, so it seems solved. I will modify the title and indicate the fix. Hopefully it stays right!

Thanks a lot again.

seth wrote:

Cloudy, rainy, possible thunderstorms, 11°C - everyone out tomorrow to get drunk has probably have a serious alcohol problem.

Slightly worse than us, then. Good to know we're not alone.

Arch Linux

#1 2026-05-12 08:50:34

[SOLVED] Many GNOME applications (+ using the portal) fail to start

#2 2026-05-12 09:59:23

Re: [SOLVED] Many GNOME applications (+ using the portal) fail to start

#3 2026-05-12 11:49:05

Re: [SOLVED] Many GNOME applications (+ using the portal) fail to start

#4 2026-05-12 12:08:05

Re: [SOLVED] Many GNOME applications (+ using the portal) fail to start

#5 2026-05-12 12:36:39

Re: [SOLVED] Many GNOME applications (+ using the portal) fail to start

#6 2026-05-12 12:42:39

Re: [SOLVED] Many GNOME applications (+ using the portal) fail to start

#7 2026-05-12 13:10:33

Re: [SOLVED] Many GNOME applications (+ using the portal) fail to start

#8 2026-05-12 13:19:11

Re: [SOLVED] Many GNOME applications (+ using the portal) fail to start

#9 2026-05-12 13:31:01

Re: [SOLVED] Many GNOME applications (+ using the portal) fail to start

#10 2026-05-12 13:45:34

Re: [SOLVED] Many GNOME applications (+ using the portal) fail to start

#11 2026-05-12 13:53:04

Re: [SOLVED] Many GNOME applications (+ using the portal) fail to start

#12 2026-05-12 14:00:50

Re: [SOLVED] Many GNOME applications (+ using the portal) fail to start

#13 2026-05-12 17:57:46

Re: [SOLVED] Many GNOME applications (+ using the portal) fail to start

#14 2026-05-13 07:03:36

Re: [SOLVED] Many GNOME applications (+ using the portal) fail to start

#15 2026-05-13 07:15:24

Re: [SOLVED] Many GNOME applications (+ using the portal) fail to start

#16 2026-05-13 13:23:43

Re: [SOLVED] Many GNOME applications (+ using the portal) fail to start

Board footer