Nvidia driver being unloaded on boot Dell 9520 laptop

Ask Fedora
, , ,
1

Opening this in response to request by ankursinha.

(Edited to reformat lines copied from terminal)

(New as of December 2022) Dell 9520 laptop: i7-12700, RTX 3050, 32G RAM, 1TB SSD, dual-boot Fedora 37 && Windows 11 Home (that’s what the system came with; I left Windows on there, for now, because the 3050 works under Windows).

Secure Boot is disabled in UEFI.

The RPM Fusion Nvidia drivers are installed - 525.60.11

On boot the Nvidia driver is loaded but then kicked out again because some config failure occurs - there’s no info that I can find as to what.

Here are the salient lines from journalctl after the most recent boot containing “nvidia”:

Jan 08 06:51:19 fedora kernel: Command line: BOOT_IMAGE=(hd0,gpt9)/vmlinuz-6.0.17-300.fc37.x86_64 root=UUID=121a08d1-2561-4fa7-89cc-fef89a3dcdff ro rootflags=subvol=root rd.driver.blacklist=nouveau modprobe.blacklist=nouveau nvidia-drm.modeset=1 initcall_blacklist=simpledrm_platform_driver_init rhgb quiet rd.driver.blacklist=nouveau modprobe.blacklist=nouveau nvidia-drm.modeset=1 initcall_blacklist=simpledrm_platform_driver_init             
Jan 08 06:51:19 fedora kernel: Kernel command line: BOOT_IMAGE=(hd0,gpt9)/vmlinuz-6.0.17-300.fc37.x86_64 root=UUID=121a08d1-2561-4fa7-89cc-fef89a3dcdff ro rootflags=subvol=root rd.driver.blacklist=nouveau modprobe.blacklist=nouveau nvidia-drm.modeset=1 initcall_blacklist=simpledrm_platform_driver_init rhgb quiet rd.driver.blacklist=nouveau modprobe.blacklist=nouveau nvidia-drm.modeset=1 initcall_blacklist=simpledrm_platform_driver_init      
Jan 08 06:51:19 fedora dracut-cmdline[387]: Using kernel command line parameters:  rd.driver.pre=btrfs   BOOT_IMAGE=(hd0,gpt9)/vmlinuz-6.0.17-300.fc37.x86_64 root=UUID=121a08d1-2561-4fa7-89cc-fef89a3dcdff ro rootflags=subvol=root rd.driver.blacklist=nouveau modprobe.blacklist=nouveau nvidia-drm.modeset=1 initcall_blacklist=simpledrm_platform_driver_init rhgb quiet rd.driver.blacklist=nouveau modprobe.blacklist=nouveau nvidia-drm.modeset=1 initcall_blacklist=simpledrm_platform_driver_init                                                                                                     
Jan 08 14:51:24 fedora systemd[1]: Starting nvidia-powerd.service - nvidia-powerd service...                                                         
Jan 08 14:51:24 fedora /usr/bin/nvidia-powerd[994]: nvidia-powerd version:1.0(build 1)                                                               
Jan 08 14:51:24 fedora kernel: nvidia: module license 'NVIDIA' taints kernel.                                                                        
Jan 08 14:51:25 fedora kernel: nvidia-nvlink: Nvlink Core is being initialized, major device number 508                                              
Jan 08 14:51:25 fedora kernel: NVRM: loading NVIDIA UNIX x86_64 Kernel Module  525.60.11  Wed Nov 23 23:04:03 UTC 2022                               
Jan 08 14:51:25 fedora kernel: nvidia_uvm: module uses symbols nvUvmInterfaceDisableAccessCntr from proprietary module nvidia, inheriting taint.     
Jan 08 14:51:25 fedora kernel: nvidia-uvm: Loaded the UVM driver, major device number 506.                                                           
Jan 08 14:51:25 fedora kernel: nvidia-modeset: Loading NVIDIA Kernel Mode Setting Driver for UNIX platforms  525.60.11  Wed Nov 23 22:49:17 UTC 2022 
Jan 08 14:51:25 fedora kernel: [drm] [nvidia-drm] [GPU ID 0x00000100] Loading driver                                                                 
Jan 08 14:51:26 fedora systemd-udevd[791]: nvidia: Process '/usr/bin/bash -c '/usr/bin/mknod -Z -m 666 /dev/nvidiactl c 195 255'' failed with exit code 1.                                                                                                                                               
Jan 08 14:51:26 fedora systemd-udevd[749]: nvidia: Process '/usr/bin/bash -c '/usr/bin/mknod -Z -m 666 /dev/nvidiactl c 195 255'' failed with exit code 1.                                                                                                                                               
Jan 08 14:51:26 fedora kernel: [drm] Initialized nvidia-drm 0.0.0 20160202 for 0000:01:00.0 on minor 1                                               
Jan 08 14:51:26 fedora systemd-udevd[791]: nvidia: Process '/usr/bin/bash -c 'for i in $(cat /proc/driver/nvidia/gpus/*/information | grep Minor | cut -d \  -f 4); do /usr/bin/mknod -Z -m 666 /dev/nvidia${i} c 195 ${i}; done'' failed with exit code 1.                                              
Jan 08 14:51:26 fedora systemd-udevd[749]: nvidia: Process '/usr/bin/bash -c 'for i in $(cat /proc/driver/nvidia/gpus/*/information | grep Minor | cut -d \  -f 4); do /usr/bin/mknod -Z -m 666 /dev/nvidia${i} c 195 ${i}; done'' failed with exit code 1.                                              
Jan 08 14:51:26 fedora /usr/bin/nvidia-powerd[994]: Dbus Connection is established                                                                   
Jan 08 14:51:26 fedora systemd[1]: Started nvidia-powerd.service - nvidia-powerd service.                                                            
Jan 08 14:51:26 fedora audit[1]: SERVICE_START pid=1 uid=0 auid=4294967295 ses=4294967295 subj=system_u:system_r:init_t:s0 msg='unit=nvidia-powerd comm="systemd" exe="/usr/lib/systemd/systemd" hostname=? addr=? terminal=? res=success'                                                               
Jan 08 14:51:26 fedora systemd[1]: nvidia-fallback.service - Fallback to nouveau as nvidia did not load was skipped because of a failed condition check (ConditionPathExists=!/sys/module/nvidia).

I have a desktop i9-12900 system with a discrete RTX 3050. I have Fedora 37 running on that too and it successfully loads the same RPM fusion driver(s). It also has Secure Boot disabled, and the drivers are not signed (by me) on either platform. The lines where it seems errors have occurred in the journalctl output (“failed with exit code 1”) are the same on both laptop and desktop platforms.

There’s something peculiar about this laptop that is stopping the RPM Fusion Nvidia driver being loaded, but I don’t know what it is.

The laptop is working well with the iGPU under Fedora 37, but I’d like the 3050 to work too (so I can ditch Windows entirely!)

1 Like
2

That log snippet shows the nvidia driver properly loading.
What is the output of lsmod | grep nvidia

3 
nvidia_drm             73728  3
nvidia_modeset       1208320  2 nvidia_drm
nvidia_uvm           2904064  0
nvidia              56176640  391 nvidia_uvm,nvidia_modeset
4 
Jan 08 14:51:26 fedora systemd[1]: nvidia-fallback.service - Fallback to nouveau *as nvidia did not load* was skipped because of a failed condition check (ConditionPathExists=!/sys/module/nvidia).

Emphasis added. As additional proof, Blender can’t find a CUDA-capable GPU.

5

  1. The output of lsmod shows the nvidia drivers are actually loaded and active.

  2. That line shows the fallback to nouveau failed because it was not needed.

  3. Blender can’t find the GPU probably because the cuda driver is not installed.

The cuda driver can be installed from the rpmfusion repo with dnf install xorg-x11-drv-nvidia-cuda

6

I believe CUDA drivers are loaded already:

$ dnf list installed *nvidia*
Installed Packages
akmod-nvidia.x86_64                                                      3:525.60.11-1.fc37                                @rpmfusion-nonfree-updates
kmod-nvidia-6.0.15-300.fc37.x86_64.x86_64                                3:525.60.11-1.fc37                                @@commandline             
kmod-nvidia-6.0.16-300.fc37.x86_64.x86_64                                3:525.60.11-1.fc37                                @@commandline             
kmod-nvidia-6.0.17-300.fc37.x86_64.x86_64                                3:525.60.11-1.fc37                                @@commandline             
nvidia-gpu-firmware.noarch                                               20221214-145.fc37                                 @updates                  
nvidia-persistenced.x86_64                                               3:525.60.11-1.fc37                                @rpmfusion-nonfree-updates
nvidia-settings.x86_64                                                   3:525.60.11-1.fc37                                @rpmfusion-nonfree-updates
xorg-x11-drv-nvidia.x86_64                                               3:525.60.11-1.fc37                                @rpmfusion-nonfree-updates
xorg-x11-drv-nvidia-cuda.x86_64                                          3:525.60.11-1.fc37                                @rpmfusion-nonfree-updates
xorg-x11-drv-nvidia-cuda-libs.x86_64                                     3:525.60.11-1.fc37                                @rpmfusion-nonfree-updates
xorg-x11-drv-nvidia-kmodsrc.x86_64                                       3:525.60.11-1.fc37                                @rpmfusion-nonfree-updates
xorg-x11-drv-nvidia-libs.x86_64                                          3:525.60.11-1.fc37                                @rpmfusion-nonfree-updates
xorg-x11-drv-nvidia-power.x86_64                                         3:525.60.11-1.fc37                                @rpmfusion-nonfree-updates

If true then that has to be the most confusing and unintuitive error message ever.
What part of the message does the failed condition check refer to? The nvidia driver not being loaded or the fallback not happening? Fallback can’t happen in any case because the nouveau driver is blacklisted. Why would nvidia-fallback-service be involved at all if the the driver is successfully loaded? What is there to fallback to (or attempt it) in that instance?

7

The fallback not happening. The condition needed to trigger the fallback was not met.
I agree it is not clearly worded for a cursory read.

In the situation where the nvidia driver is installed but remains unsigned with secure boot active then the driver cannot load. The fallback forces the nouveau driver to be loaded instead (I think that happens even with nouveau blacklisted).
I suspect it would happen similarly if the nvidia driver were corrupt and could not properly load.

If you still cannot get Blender to run then that is a separate topic entirely since the nvidia drivers are loaded and cuda is installed.

Blender runs properly for me.