NVidia GeForce GT610 driver fails after upgrade to kernel 6.6

Support Graphics & Display
1

I just did a pacman sYu which upgraded to kernel 6.6.126. On first boot I got a blank screen with a still cursor. I could still ssh in, and from there I discovered that the ALPM script that does the dkms upgrade failed. I don’t really know what that means though…

TLDR:
pacman.log has

[2026-02-26T09:28:29+1100] [ALPM-SCRIPTLET] ==> WARNING: `dkms install --no-depmod nvidia/390.157 -k 6.6.126-1-MANJARO' exited 6

System journal has lots of

Feb 27 11:05:24 radagast kwin_wayland[19254]: gbm_bo_get_fd_for_plane() failed: Invalid argument

nvidia-driver assistant tells me to run sudo pacman -S linux66-nvidia-390xx but that just reinstalls the same driver I had before and doesn’t change anything.

I’ve read lots of posts with driver/package conflicts, but i don’t seem to have any of those. I’ll post more of the pacman logs and journalctl, happy to add any other info (I rarely post to the forums - it’s rare that i don’t find my question already answered, but this is one of those cases)

https://pastebin.com/ahWKMKhg (that’ll only last a week)

2

It means it didn’t build the 390xx driver, which will prevent your graphics to load.

You may consider removing the nvidia-390xx-dkms package and instead sync the linux66-nvidia-390xx

3 Likes
3

Thanks…
I uninstalled that driver and reinstalled the nvidia one

sudo pacman -R nvidia-390xx-dkms
sudo pacman -S linux66-nvidia-390xx

No change. It’s loading the module

Mar 01 12:09:39 radagast kernel: NVRM: loading NVIDIA UNIX x86_64 Kernel Module  390.157  Wed Oct 12 09:19:07 UTC 2022 (using threaded interrupts)
Mar 01 12:09:39 radagast kernel: tsc: Refined TSC clocksource calibration: 3700.001 MHz
Mar 01 12:09:39 radagast kernel: clocksource: tsc: mask: 0xffffffffffffffff max_cycles: 0x6aaaabc71c7, max_idle_ns: 881590412124 ns
Mar 01 12:09:39 radagast kernel: clocksource: Switched to clocksource tsc
Mar 01 12:09:39 radagast kernel: nvidia-modeset: Loading NVIDIA Kernel Mode Setting Driver for UNIX platforms  390.157  Wed Oct 12 09:15:25 UTC 2022
Mar 01 12:09:39 radagast kernel: nvidia-uvm: Loaded the UVM driver in 8 mode, major device number 236
Mar 01 12:09:39 radagast kernel: nvidia_drm: unknown parameter 'fbdev' ignored
Mar 01 12:09:39 radagast kernel: [drm] [nvidia-drm] [GPU ID 0x00002d00] Loading driver
Mar 01 12:09:39 radagast kernel: resource: resource sanity check: requesting [mem 0x00000000000c0000-0x00000000000fffff], which spans more than PCI Bus 0000:00 [mem 0x000a0000-0x000dffff window]
Mar 01 12:09:39 radagast kernel: caller _nv001015rm+0x1bf/0x1f0 [nvidia] mapping multiple BARs
Mar 01 12:09:39 radagast kernel: [drm] Initialized nvidia-drm 0.0.0 20160202 for 0000:2d:00.0 on minor 1

but failing with the two errors

Mar 01 12:10:18 radagast kernel: [drm:nv_drm_gem_prime_export [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00002d00] Gem object 0x000000008cc56958 is not suitable to export
Mar 01 12:10:18 radagast kwin_wayland[966]: gbm_bo_get_fd_for_plane() failed: Invalid argument

I’d love to be able to live without drm support… is it possible to try? doesn’t half the internet have drm? or would I not miss it?

4

I just removed nvidia_drm from the modules list in /etc/mkinitcpio.conf
Reinstalled the driver sudo pacman -S linux66-nvidia-390xx and rebuilt the modules sudo mkinitcpio -P (i’m sure you don’t have to do both, but just to be sure). The boot log says it’s still loading the module, and has the same errors?

Mar 01 12:31:22 radagast kernel: nvidia_drm: unknown parameter 'fbdev' ignored
Mar 01 12:31:22 radagast kernel: [drm] [nvidia-drm] [GPU ID 0x00002d00] Loading driver
Mar 01 12:31:22 radagast systemd[1]: Finished Monitoring of LVM2 mirrors, snapshots etc. using dmeventd or progress polling.
Mar 01 12:31:22 radagast systemd[1]: Finished Load/Save OS Random Seed.
Mar 01 12:31:22 radagast kernel: resource: resource sanity check: requesting [mem 0x00000000000c0000-0x00000000000fffff], which spans more than PCI Bus 0000:00 [mem 0x000a0000-0x000dffff window]
Mar 01 12:31:22 radagast kernel: caller _nv001015rm+0x1bf/0x1f0 [nvidia] mapping multiple BARs
Mar 01 12:31:22 radagast systemd[1]: Finished Create Static Device Nodes in /dev gracefully.
Mar 01 12:31:22 radagast systemd[1]: Starting Create System Users...
Mar 01 12:31:22 radagast kernel: [drm] Initialized nvidia-drm 0.0.0 20160202 for 0000:2d:00.0 on minor 1

PS I should have said in my first post, falling back to the previous kernel doesn’t fix it. So it’s not an issue with the new kernel, it’s something else that has changed that affects this driver / all kernels.

5

DRM can be two things

In the context of your GPU it is the latter which makes sense when reading [drm] and nvidia-drm from the system journal.

I am thinking - with a GPU that old GT610 - it may be worth trying the opensource nouveau driver.

A search for Linux GPU driver DRM - sx.nix.dk

4 Likes
7

Thank you again.
I followed the guide at Configure NVIDIA (non-free) settings and load them on Startup - Manjaro - very simply, removed all the nvidia stuff from my mkinitcpio.conf file and replace it with just MODULES="nouveau", then rebuilt with sudo mkinitcpio -P and also sudo mhwd -a pci free 0300 (which I think should not do anything different?).
The journal still showed the kernel loading the nvidia module, and still with the drm which should have been gone many build/boots ago.
So i removed the non-free module completely with sudo mhwd -r pci video-nvidia-390xx, rebuilt again with mkinitcpio -P, and it’s working! It’s forgotten all my desktop settings, but I have a gui and a menu and my mouse works.

Thanks

8

This topic was automatically closed 3 days after the last reply. New replies are no longer allowed.