Nvidia Driver - How to proceed with post-install and configuration?

User Support
1

Hello,

OS: Qubes 4.2.3
Kernel: [user@untrusted ~]$ uname -a
Linux untrusted 6.6.54-1.qubes.fc37.x86_64 #1 SMP PREEMPT_DYNAMIC Mon Oct 7 20:17:52 GMT 2024 x86_64 GNU/Linux
VGA: Nvidia GTX 1060 6GB
Language: english (of course) isn’t my native language (as you can see, easy to note)

Question
How to configure the nvidia tools (nvidia-smi, window config etc, CUDA support etc) after install the driver manually? How to proceed after post-install properly and completely? Is it possible that the driver works just on templates or appvms?

Obs.: that’s my only card on the machine (i7 8700k, mb gigabyte z370 aorus gaming 7, GTX 1060 6GB);
I can’t afford a new computer with the designated approved HCL for maximum hardware compatibility.
I haven’t another computer.

I think that’s a good and easy way to complete install the tools using: sudo dnf install nvidia-smi.

Solutions tried:

  • Add RPMFusion repos in dom0 but no way to make it work.

Proceed with config the repos:

sudo dnf repo list
sudo dnf config-manager setopt rpmfusion-free.enabled=1
sudo dnf config-manager setopt rpmfusion-nonfree.enabled=1
sudo dnf config-manager setopt rpmfusion-free-updates.enabled=1
sudo dnf config-manager setopt rpmfusion-nonfree-updates.enabled=1

_____ Don’t work at dom0.
_____ Work in fedora-41-xfce template and work qube
(I try the first, the second and both)

As result the system freezes and I lose all work.

sudo dnf install https://mirrors.rpmfusion.org/free/fedora/rpmfusion-free-release-$(rpm -E %fedora).noarch.rpm https://mirrors.rpmfusion.org/nonfree/fedora/rpmfusion-nonfree-release-$(rpm -E %fedora).noarch.rpm

and

sudo dnf config-manager --enable fedora-cisco-openh264

and

sudo dnf config-manager --set-enabled rpmfusion-free
sudo dnf config-manager --set-enabled rpmfusion-free-updates
sudo dnf config-manager --set-enabled rpmfusion-nonfree
sudo dnf config-manager --set-enabled rpmfusion-nonfree-updates
sudo dnf upgrade --refresh

______ nothing work.

It’s my 6th Qubes OS new installation in 3 days. I’m spending more than 18 hours day reading, learning, testing.

  • I check in Qubes Docs for nvidia troubleshooting but the page return error 404 (don’t exist anymore). I found nvidia-troubleshooting-guide but… doesn’t help much.

  • I’m trying to find a way to config and enable rpmfusion to install nvidia tools and support in Qubes, like nvidia-smi and nvidia gui manager, and of course the support for X11/Xorg.

I’ve been working on this for a few days now.

  • I managed to install the Nvidia driver manually after using 3 different guides on the forum. The based one is: Nvidia install.
    I do some modifications and reinstall the entire system and test it.

_____ Apparently works.
(Driver version used: 535.216.01)

As result I have:

[root@dom0 Downloads]# lsmod |grep nvidia
nvidia              62857216  0

But when I check in any qube, template or appvm, the nvidia isn’t loaded.

Summary

[user@untrusted ~]$ lsmod |grep nvidia
[user@untrusted ~]$
[user@untrusted ~]$ lsmod
Module Size Used by
vfat 24576 0
fat 106496 1 vfat
snd_seq_dummy 12288 0
snd_hrtimer 12288 1
snd_seq 126976 7 snd_seq_dummy
snd_seq_device 16384 1 snd_seq
snd_timer 53248 2 snd_seq,snd_hrtimer
snd 147456 5 snd_seq,snd_seq_device,snd_timer
soundcore 16384 1 snd
nft_reject_ipv6 12288 1
nf_reject_ipv6 20480 1 nft_reject_ipv6
nft_reject_ipv4 12288 1
nf_reject_ipv4 12288 1 nft_reject_ipv4
nft_reject 12288 2 nft_reject_ipv6,nft_reject_ipv4
nft_ct 24576 8
nft_masq 12288 2
nft_chain_nat 12288 2
nf_nat 61440 2 nft_masq,nft_chain_nat
nf_conntrack 184320 3 nf_nat,nft_ct,nft_masq
nf_defrag_ipv6 24576 1 nf_conntrack
nf_defrag_ipv4 12288 1 nf_conntrack
intel_rapl_msr 20480 0
intel_rapl_common 40960 1 intel_rapl_msr
intel_pmc_core 86016 0
crct10dif_pclmul 12288 1
crc32_pclmul 12288 0
nf_tables 376832 133 nft_reject_ipv6,nft_reject_ipv4,nft_ct,nft_masq,nft_chain_nat,nft_reject
crc32c_intel 16384 0
polyval_clmulni 12288 0
polyval_generic 12288 1 polyval_clmulni
ghash_clmulni_intel 16384 0
sha512_ssse3 49152 0
sha256_ssse3 32768 0
sha1_ssse3 32768 0
xen_netfront 57344 1
pcspkr 12288 0
xen_scsiback 49152 0
target_core_mod 581632 1 xen_scsiback
xen_netback 86016 0
xen_privcmd 28672 48
xen_gntdev 45056 1
xen_gntalloc 20480 6
xen_blkback 61440 0
xen_evtchn 12288 6
loop 45056 0
fuse 221184 3
nfnetlink 20480 2 nf_tables
overlay 233472 1
xen_blkfront 57344 9
[user@untrusted ~]$

Just to inform, first time I put the nvidia.ko (compiled) on /lib/modules/$(uname -r)/extra as indicated in this tutorial - user:dhn.
Reading other tutorials and info I found a indication of a new directory: /lib/modules/$(uname -r)/kernel/drivers/video/ in my case: /lib/modules/6.6.54-1.qubes.fc37.x86_64/kernel/drivers/video/

Thank you.

2

Are you trying to do this because the install with Nvidia card isn’t working or because you want to install Nvidia drivers for another reason ?

3

Answer in next post.

4

Hey.

I install the driver, manually at dom0.

But where is the config?
Why nvidia-smi doens’t work?
Can I have the nvidia panel/manager working?

It’s about that.

After manually install, the only thing that is working (apparently) is the module nvidia.ko that is loaded. But I can’t check the gpu status, frequency etc. All tutorials using the rpmfusion doesn’t works. Many of than are old and the commands are depredicated.

And I’m using 6.6.54-1.qubes.fc37.x86_64 kernel.

Ps.: I edit the original post to a better markdown and reading.

5

Ok, sorry, I don’t know about that. Some nvidia cards are having issues with your kernel so I thought it might be that but it sounds the kernel works fine.

Perhaps @apparatus can suggest a fix ?

1 Like
6

Solved:

Just copy nvidia-smi or any executable that do you want from installation folder to /usr/bin and the necessary libs to /usr/lib64/. Done.