NVIDIA Driver Linux Kernel 6.15

Graphics / Linux Linux
1

Problems with 575.51.02 or 570.153.02 Driver can’t install on Linux Kernel 6.15
OS: openSUSE Tumbleweed
nvidia-installer.log (8.5 MB)
exist a useable Solution ?
Regards

2

Same OS, perfectly working out of RPM (570.153.02)

3

570.153.02 should work fine, 575 is not supported without patches.

4

570.153.02 works with the open source driver.
570.153.02 closed source driver needs patching as it uses gpl-only symbols which prevents it from loading.

5

The just released 575.57.08 should do the trick for you when using the MIT/GPL modules (=default option).

6

No does not work

7

how you manage to fail it working?

8

nvidia
nvidia1534×789 155 KB

Just showing to proof. NVidia’s repo. Closed driver. All work like a charm. No single finger moved to make it work. How you guys manage to fail at all???

9

You haven’t considered the possibility that Suse reverted the kernel commit that made the symbols gpl-only.
It’s definitely a real issue , loading the module fails due to gpl symbols on kernel-6.15.
I am using the cachyos patch to workarround the issue for the rpmfusion driver.

1 Like
10

Also your screenshot doesn’t prove your point, it show that nvidia is working but doesn’t show which module is used!

modinfo nvidia |grep license:
1 Like
11

How much more proof needed?
license

12

repos
repos1344×752 175 KB

official NVidia repo install source from RPM, closed driver. All works out the box. Didn’t touch anything at all. Just updated kernel and it works. Got to know there are any issues only here in topic.

I have legacy GPU without GSP, so it doesn’t support open driver at all, only proprietary.

As an option try this repo for kernel: https://cdn.opensuse.org/repositories/Kernel:/stable/standard/ . I use this one, OpenSUSE is slow rolling out kernels and often end up with bugs, so I ignore shipped with distro kernels. And it worked as a charm, didn’t know issue exists at all. And not willing to try distro default one.

And just checked, official OpenSUSE Tumbleweed repo doesn’t ship 6.15 at all yet. So my question remains the same: how did you manage to fail? Where is kernel from and how it manages to fail?

13

It’s possible to patch the .run file

wget https://raw.githubusercontent.com/leigh123linux/nvidia_patch/refs/heads/main/Workaround-nv_vm_flags_-calling-GPL-only-code.patch
sh NVIDIA-Linux-x86_64-575.57.08.run --apply-patch  Workaround-nv_vm_flags_-calling-GPL-only-code.patch

Then install the custom .run file

sh NVIDIA-Linux-x86_64-575.57.08-custom.run
3 Likes
14

I use the fedora kernel from

$ dnf list kernel
Updating and loading repositories:
 Fedora 42 - x86_64 - Updates                                          100% |  87.8 KiB/s |   9.0 KiB |  00m00s
 Fedora 42 - x86_64 - Updates                                          100% |   3.1 MiB/s |   1.8 MiB |  00m01s
Repositories loaded.
Installed packages
kernel.x86_64 6.14.6-300.fc42      updates
kernel.x86_64 6.15.0-0.rc6.49.fc43 fedora-rawhide-nodebug
kernel.x86_64 6.15.0-61.fc43       fedora-rawhide-nodebug
15

Hmm… then they probably broke kernel build somehow. Likely for me no any sorts of patches needed and all works out the box. It’s anyway stupid how kernel insists on gpl nonsense at times. but in given case seems kernel vendor’s fault

16

Why do you think it’s everyone’s kernel that is broken and not your kernel (my guess is your ego is bigger than your brain :-) )?
multiple distros are reporting the same issue.

Have you considered your kernel is just some early git snapshot from before the kernel changes that broke nvidia.

17

that patch just worked for me, thank you.