[Software] Issues with the nvidia proprietary drivers installation

[Puffer]

Greetings everyone!

I have a fresh Debian install in my laptop, the following information might be relevant:

• dGPU: NVIDIA GeForce RTX 3060 Mobile / Max-Q
• iGPU: Intel Alder Lake-P
• CPU: 12th Gen Intel i7-12700H (20) @ 4.600GHz
• Kernel: 6.1.0-13-amd64
• OS: Debian GNU/Linux 12 (bookworm) x86_64

Everything was smooth until I attempted to install the nvidia drivers; I followed this page: https://wiki.debian.org/NvidiaGraphicsD ... ary_Driver. The driver itself seems to work fine (couldn't test it), but my external display wouldn't be identified. If I uninstall the drivers, the display works again. After uninstalling the drivers the KDE animations are all over the place, some work at 20Hz others are completely gone. I'm on KDE Plasma, wayland; my laptop has 2 graphics cards as mentioned above. I'd like to have the iGPU work most of the time and the dGPU summoned once necessary (for a render or playing a game per say). Currently I purged the nvidia stuff since I rely on the external display for work, but I'm willing to try again once possible. I've been all over Google so the most "obvious" stuff seems to be out of the plate, but you never know, maybe I misread something. If any additional information is required please leave a reply and I'll answer ASAP.

Thanks in advance!


11/02/2023 Edit: All Kde Plasma animations are gone, wayland takes up to, somehow, 600% of the cpu. System is barely usable. When using some applications such as Firefox, it feels as if the display is in 60Hz instead of 160.

[Puffer]

So I installed the drivers from the official Nvidia website, now they work and when necessary they appear (like magic). I would like to know how to summon them though, so I'd be able to use it only when necessary. I also added "KWIN_DRM_USE_MODIFIERS=0" to my /etc/environment which solved the lag. The only issue now is really how to control the GPU since in wayland I can't use nvidia-settings.

[CynicalDebian]

So I installed the drivers from the official Nvidia website,

I think the wiki makes clear why this is a bad idea. Enjoy.

but my external display wouldn't be identified

User-space utilities for managing external displays are wayland compositor dependent. On wlr-roots we use wlr-randr. I have no idea what you mean by "can't be identified." You should manage your display through your desktop environments display settings.

I'm on KDE Plasma, wayland; my laptop has 2 graphics cards as mentioned above. I'd like to have the iGPU work most of the time and the dGPU summoned once necessary (for a render or playing a game per say).

On Nvidia Propietary Drivers this technology is known as Nvidia OptimusTM.
https://wiki.debian.org/NVIDIA%20Optimus

now they work and when necessary they appear (like magic). I would like to know how to summon them though,

They as in the NVIDIA driver? This is the only thing I think you could be referring to, and I have no idea what "summoning" means here.

The only issue now is really how to control the GPU since in wayland I can't use nvidia-settings.

No clue what you mean. The management frontend in NVIDIA is mainly just a duplication of efforts to control the X-server that already exist in X userspace utilities (Xrandr), the idea is that configuration should be GPU-Agnostic, a GPU "settings" panel is largely a Windowsism. You should manage your display settings using your Desktop Environments display manager.

Your experience with Wayland with the NVIDIA propietary driver is just going to be more rabbit-holes and other pitfalls. My recommendation is to use the much more mature system for managing your display. The X11 Display Server.
https://wiki.debian.org/Xorg

If you want to use Wayland for all of its benefits, you should use the Nouveau display driver. But this will have poor support with your RTX 3060.

Welcome to proprietary hell.

Hope that helps.

[Puffer]

I think the wiki makes clear why this is a bad idea. Enjoy.

Might be a bad idea but at least they work. The drivers available in the standard repository won't allow me to use the GPU no matter what I throw at it.

...but my external display wouldn't be identified...

Now it's my main display. It simply stays black, here's what journalctl says:

Code: Select all

...
sddm[1022]: Failed to read display number from pipe
sddm[1022]: Could not start Display server on vt 7
kernel: watchdog: watchdog0: watchdog did not stop!
...

They as in the NVIDIA driver? This is the only thing I think you could be referring to, and I have no idea what "summoning" means here.

I meant that the GPU works. With the drivers from the official source/repository my GPU won't start no matter what.

By summoning the GPU I meant: use the iGPU for all the minor tasks and left the heavy duty for the dGPU. Like nvidia-prime on Arch does, once you run something with "prime-run" attached to it, the dGPU renders the software.

And again, I followed the wiki's page and that led to nothing. I'll try the recommended drivers again but I'm sure the result won't change; I'll report once I have any news.

[CynicalDebian]

I meant that the GPU works. With the drivers from the official source/repository my GPU won't start no matter what.

I'm highly skeptical. It's not a car, it's a GPU. Is this a Waylandism or is the actual nvidia-driver install not working? If it's not it points to a more serious problem. Now that you have downloaded using the NVIDIA installer script your system is left in a strange state.
https://wiki.debian.org/DontBreakDebian ... ll_scripts

Like nvidia-prime on Arch does

Arch doesn't have access to some secret nvidia-prime technology Debian doesn't, the technology is known as Nvidia OptimusTM and the Wiki page I sent describes setting the environment variables to launch programs with the dGPU.

My guess is with Wayland you need to force enable drm-modeset with the NVIDIA drivers from Debian. (and that your external output is wired directly the the dGPU, as is a possible situation... explained in the Nvidia OptimusTM wiki page.
https://forums.developer.nvidia.com/t/u ... ing/204068

So...
1. Install the nvidia-driver from Debian repositories
2. Enable nvidia-drm in /etc/modprobe.d
3. Verify the driver is loaded using a program like inxi or just modprobe.
4. Try launching a program from the dGPU using the environment variables explained in the NVIDIA OptimusTM wiki page

[Puffer]

I'm highly skeptical. It's not a car, it's a GPU. Is this a Waylandism or is the actual nvidia-driver install not working? If it's not it points to a more serious problem..

I don't know what to tell ya. When I install from NVIDIA's website website it works, when I do it from the repositories it doesn't. Beats me.

Arch doesn't have access to some secret nvidia-prime technology

It was just an example.

When I tried to run a game with the parameters as instructed it simply froze. Same game, drivers from the Nvidia's website it works.

Now my main monitor came back to life, the external is now without signal. I'm confused.

Thanks for the quick response.

[CynicalDebian]

I don't know what to tell ya. When I install from NVIDIA's website website it works, when I do it from the repositories it doesn't. Beats me.

I think my confusion is, does it work on X? I'm saying you shouldn't write off the entire Debian driver just because it appears broken on Wayland.

When I tried to run a game with the parameters as instructed it simply froze. Same game, drivers from the Nvidia's website it works.

The driver will work till kernel upgrade time, if it's working then perhaps your issue is solved?

[Puffer]

think my confusion is, does it work on X?

Same behavior on X.

Tried re installing the driver from Nvidia’s website, my external screen now goes black, no signal. Either version seems to break something somewhere now. Maybe X conf? But could it affect monitor detection even under Wayland? I’m out of ideas now.

[Puffer]

I've been doing some tests and it seems I can't come back to the driver that was working (for some godforsaken reason). I mean, I can, it just won't work. It makes me think that maybe a system file/configuration might have gotten in the way of the other driver or something like that. I'll leave some info I found in journalctl -b -1 | grep nvidia, might be useful:

Code: Select all

[sudo] password for puffer: 
nov 04 22:13:37 puffer kernel: Command line: BOOT_IMAGE=/boot/vmlinuz-6.1.0-13-amd64 root=UUID=e2e9de85-e32a-4755-9105-b2fe3b2e2277 ro nvidia-drm.modeset=1
nov 04 22:13:37 puffer kernel: Kernel command line: BOOT_IMAGE=/boot/vmlinuz-6.1.0-13-amd64 root=UUID=e2e9de85-e32a-4755-9105-b2fe3b2e2277 ro nvidia-drm.modeset=1
nov 04 22:13:37 puffer systemd-modules-load[432]: Failed to find module 'nvidia-drm'
nov 04 22:13:37 puffer audit[633]: AVC apparmor="STATUS" operation="profile_load" profile="unconfined" name="nvidia_modprobe" pid=633 comm="apparmor_parser"
nov 04 22:13:37 puffer audit[633]: AVC apparmor="STATUS" operation="profile_load" profile="unconfined" name="nvidia_modprobe//kmod" pid=633 comm="apparmor_parser"
nov 04 22:13:37 puffer kernel: audit: type=1400 audit(1699150417.340:6): apparmor="STATUS" operation="profile_load" profile="unconfined" name="nvidia_modprobe" pid=633 comm="apparmor_parser"
nov 04 22:13:37 puffer kernel: audit: type=1400 audit(1699150417.340:7): apparmor="STATUS" operation="profile_load" profile="unconfined" name="nvidia_modprobe//kmod" pid=633 comm="apparmor_parser"
nov 04 22:13:37 puffer kernel: nvidia: loading out-of-tree module taints kernel.
nov 04 22:13:37 puffer kernel: nvidia: module license 'NVIDIA' taints kernel.
nov 04 22:13:37 puffer kernel: nvidia: module verification failed: signature and/or required key missing - tainting kernel
nov 04 22:13:37 puffer kernel: nvidia-nvlink: Nvlink Core is being initialized, major device number 236
nov 04 22:13:37 puffer kernel: nvidia 0000:01:00.0: enabling device (0000 -> 0003)
nov 04 22:13:37 puffer kernel: nvidia 0000:01:00.0: vgaarb: changed VGA decodes: olddecodes=io+mem,decodes=none:owns=none
nov 04 22:13:37 puffer kernel: nvidia-modeset: Loading NVIDIA Kernel Mode Setting Driver for UNIX platforms  525.125.06  Tue May 30 04:58:48 UTC 2023
nov 04 22:13:37 puffer kernel: [drm] [nvidia-drm] [GPU ID 0x00000100] Loading driver
nov 04 22:13:37 puffer kernel: [drm] Initialized nvidia-drm 0.0.0 20160202 for 0000:01:00.0 on minor 1
nov 04 22:13:38 puffer systemd[1]: Starting nvidia-persistenced.service - NVIDIA Persistence Daemon...
nov 04 22:13:38 puffer nvidia-persistenced[873]: Started (873)
nov 04 22:13:39 puffer systemd[1]: Started nvidia-persistenced.service - NVIDIA Persistence Daemon.
nov 04 22:14:35 puffer plasmashell[1992]: /bin/sh\0-c\0__NV_PRIME_RENDER_OFFLOAD=1 __GLX_VENDOR_LIBRARY_NAME=nvidia /home/puffer/.steam/debian-installation/ubuntu12_32/reaper SteamLaunch AppId=782330 -- /home/puffer/.steam/debian-installation/ubuntu12_32/steam-launch-wrapper -- '/data/SteamLibrary/steamapps/common/SteamLinuxRuntime_sniper'/_v2-entry-point --verb=waitforexitandrun -- '/data/SteamLibrary/steamapps/common/Proton 8.0'/proton waitforexitandrun  '/data/SteamLibrary/steamapps/common/DOOMEternal/idTechLauncher.exe' +com_skipIntroVideo 1\0
nov 04 22:15:50 puffer plasmashell[1992]: /bin/sh\0-c\0__NV_PRIME_RENDER_OFFLOAD=1 __GLX_VENDOR_LIBRARY_NAME=nvidia /home/puffer/.steam/debian-installation/ubuntu12_32/reaper SteamLaunch AppId=782330 -- /home/puffer/.steam/debian-installation/ubuntu12_32/steam-launch-wrapper -- '/data/SteamLibrary/steamapps/common/SteamLinuxRuntime_sniper'/_v2-entry-point --verb=waitforexitandrun -- '/data/SteamLibrary/steamapps/common/Proton 8.0'/proton waitforexitandrun  '/data/SteamLibrary/steamapps/common/DOOMEternal/idTechLauncher.exe' +com_skipIntroVideo 1\0
nov 04 22:22:27 puffer systemd[1]: Stopping nvidia-persistenced.service - NVIDIA Persistence Daemon...
nov 04 22:22:27 puffer nvidia-persistenced[873]: Shutdown (873)
nov 04 22:22:27 puffer systemd[1]: nvidia-persistenced.service: Deactivated successfully.
nov 04 22:22:27 puffer systemd[1]: Stopped nvidia-persistenced.service - NVIDIA Persistence Daemon.
nov 04 22:22:27 puffer systemd[1]: nvidia-persistenced.service: Consumed 1.006s CPU time.

Once again if extra info is required, just hit me up.

[sunrat]

Please use code tags, not quote tags, for terminal text output.

[Puffer]

Please use code tags, not quote tags, for terminal text output.

My bad, typo.

[@ttila]

Hi,
i am fighting with my black screen too under Wayland. If you need to configure your secondary display under KDE plasma, i read exist an app called kscreen-doctor or from a command line wlr-randr.

[Puffer]

Hi,
i am fighting with my black screen too under Wayland. If you need to configure your secondary display under KDE plasma, i read exist an app called kscreen-doctor or from a command line wlr-randr.

Hi, yes I do have such software. When I try to list the outputs, it only shows the main screen unfortunately.

[Puffer]

So I spent about some hours already and no luck whatsoever... Kscreen-doctor only shows one output. I installed the drivers from the Nvidia's website and they worked again (I swear I don't understand anything anymore). In the installation process it says I should be fine even with kernel updates, is it true? If so, honestly I don't see why I shouldn't use them. Is there any other Nvidia driver version available that I don't know of? Maybe a more recent one? From Nvidia's website I get the 535... and from the official sources I have the 525...

[CynicalDebian]

Do you have drm modesetting force enabled with modprobe? It is probably necessary to run Wayland, my guess is that newer NVIDIA driver versions enable this by default now, which would explain the difference between versions.

The other issue with external displays is how they are wired to the dGPU/iGPU in the laptop. Depending on the configuration sometimes workarounds are necessary, which is I why suggest sticking with X11 where this stuff is well tested if your going to be using Hybrid graphics. It's the price you pay for proprietary.

This thread may be helpful...
viewtopic.php?p=769511

it says I should be fine even with kernel updates, is it true?

Probably not

[Puffer]

Do you have drm modesetting force enabled with modprobe? It is probably necessary to run Wayland, my guess is that newer NVIDIA driver versions enable this by default now, which would explain the difference between versions.

I have modesetting enabled, what do you mean by force?