Glad to hear it’s working. No need for a pic of the error. What’s probably happening is the new kmod triggers dracut to regenerate the initramfs for all the kernels installed. The ones that aren’t kABI compatible will complain. It’s expected behavior.
Not sure what you meant by the comment about the 570.190 driver in the main repo. I keep that updated too…at least until it looks like 580.xx is stable enough.
But until it is updated, everyone who updates the kernel will now get a black screen.
It is updated already.
rpm -qRp kmod-nvidia-570.190-1.1.el9_6.elrepo.x86_64.rpm | grep "kernel "
kernel >= 5.14.0-570.42.2.el9_6
Hi! Not trying to hijack this thread but I’m attempting to solve a similar issue that popped up with kernel updates AFTER 5.14.0-570.39.1.el9_6.x86_64. I’m stuck on .39 since the best I can do with Xorg is a framebuffer driver with the subsequent kernel updates. I can move this post to a new thread if need be.
What I am running into is the nvidia-drm.ko and nvidia-modeset.ko bits don’t seem to be getting built for kernels newer than .39. I tried removing (PITA with dnf so used rpm) the .52 kernel and reinstalling. That got me nvidia-modeset.ko but still no drm module and Xorg still comes up with the framebuffer driver although nvidia-smi at least runs and tells me the the driver isn’t loaded.
Looking for suggestions like just reinstalling kmod-nvidia-5.14.0-570.el9_6-570.172.08-1.el9.x86_64 (going to try that next) or moving to a more recent kmod package. Suggestions welcome and can provide more info if needed. Video card is:
84:00.0 VGA compatible controller: NVIDIA Corporation TU106 [GeForce GTX 1650] (rev a1) (prog-if 00 [VGA controller])
Subsystem: PNY Device 1400
Thanks!
Update: The reinstall kmod-nvidia gets me lots of the following messages in /var/log/messages during boot:
Oct 21 21:35:45 bend kernel: nvidia: loading out-of-tree module taints kernel.
Oct 21 21:35:45 bend kernel: nvidia: module license ‘NVIDIA’ taints kernel.
Oct 21 21:35:45 bend kernel: Disabling lock debugging due to kernel taint
Oct 21 21:35:45 bend kernel: nvidia: module verification failed: signature and/or required key missing - tainting kernel
Oct 21 21:35:45 bend kernel: nvidia: module license taints kernel.
Oct 21 21:35:45 bend kernel: nvidia-nvlink: Nvlink Core is being initialized, major device number 237
Oct 21 21:35:45 bend kernel:
Oct 21 21:35:45 bend kernel: nvidia 0000:84:00.0: vgaarb: VGA decodes changed: olddecodes=io+mem,decodes=none:owns=io+mem
Oct 21 21:35:45 bend kernel: NVRM: loading NVIDIA UNIX x86_64 Kernel Module 580.95.05 Tue Sep 23 10:11:16 UTC 2025
Oct 21 21:35:46 bend kernel: nvidia-modeset: Loading NVIDIA Kernel Mode Setting Driver for UNIX platforms 580.95.05 Tue Sep 23 09:41:17 UTC 2025
Oct 21 21:35:46 bend kernel: nvidia_drm: disagrees about version of symbol drm_framebuffer_cleanup
Oct 21 21:35:46 bend kernel: nvidia_drm: Unknown symbol drm_framebuffer_cleanup (err -22)
Oct 21 21:35:46 bend kernel: nvidia_drm: disagrees about version of symbol drm_framebuffer_init
Oct 21 21:35:46 bend kernel: nvidia_drm: Unknown symbol drm_framebuffer_init (err -22)
Oct 21 21:35:46 bend kernel: nvidia_drm: disagrees about version of symbol drm_property_create
Oct 21 21:35:46 bend kernel: nvidia_drm: Unknown symbol drm_property_create (err -22)
Oct 21 21:35:46 bend kernel: nvidia_drm: disagrees about version of symbol drm_atomic_add_affected_planes
Oct 21 21:35:46 bend kernel: nvidia_drm: Unknown symbol drm_atomic_add_affected_planes (err -22)
Oct 21 21:35:46 bend kernel: nvidia_drm: disagrees about version of symbol drm_atomic_add_affected_connectors
Oct 21 21:35:46 bend kernel: nvidia_drm: Unknown symbol drm_atomic_add_affected_connectors (err -22)
Oct 21 21:35:46 bend kernel: nvidia_drm: disagrees about version of symbol drm_atomic_helper_disable_plane
Oct 21 21:35:46 bend kernel: nvidia_drm: Unknown symbol drm_atomic_helper_disable_plane (err -22)
Oct 21 21:35:46 bend kernel: nvidia_drm: disagrees about version of symbol drm_atomic_helper_shutdown
Oct 21 21:35:46 bend kernel: nvidia_drm: Unknown symbol drm_atomic_helper_shutdown (err -22)
Oct 21 21:35:46 bend kernel: nvidia_drm: disagrees about version of symbol drm_atomic_state_default_clear
Oct 21 21:35:46 bend kernel: nvidia_drm: Unknown symbol drm_atomic_state_default_clear (err -22)
…
There are a little over 200 lines of similar messages.
Going by memory, it was around kernel-5.14.0-570.42.2.el9_6.x86_64 that the kABI was broken and it required a kmod-nvidia rebuild for the drivers from ELRepo.
Your choices are:
Thanks! The kABI problem was mentioned earlier in the thread and .39 worked but .42 didn’t which fits with what I experienced. Would this possibly be fixed in the kmod-nvidia-580 packages or is that just going to get me different headaches?
Sorry, I don’t build modules, so I just use ready-made packages from the elrepo repository. Everything works fine.
5.14.0-570.52.1.el9_6.x86_64
kmod-nvidia-580.95.05-1.1.el9_6.elrepo.x86_64
01:00.0 VGA compatible controller: NVIDIA Corporation GP106GL [Quadro P2000] (rev a1)
It’s not the NVIDIA version that is the problem. It’s a matter of which kernel the driver is built against and if that binary is backwards & forwards compatible.
I’ll give kmod-nvidia-580.95.05-1.1.el9_6.elrepo.x86_64 a shot tomorrow with the .52 kernel. Just hit midnight here so time to call it a day. Thanks again!
Just make sure you uninstall all of the RPMFusion NVIDIA packages before trying another repositories packages.
FYI, currently the ELRepo NVIDIA packages are:
When 580 gets to be a little more stable (across EL 8/9/10), we’ll promote the 580.x to the main repo.
Everything working but with the official Nvidia drivers. Ended up getting stuck attempting to install the elrepo drivers with dnf wanting to do modules. Found documentation for the official Nvidia drivers that included the module to use. Realized after I started their install that you probably used the same module nomenclature for elrepo.
Up and running on the .52 kernel but still seeing some artifacting on full motion video. May be back since it sounds like you’re trying to get it working for gamers.
Thanks again for the help!
Yes, it’s best to not mix the 3rd party repos that have packages that overlap. Glad you’re up and running.
FWIW, we’re not doing anything special for gamers. I’ve notice a few problems with the 580 drivers myself:
I have el9 systems, with NVidia’s packages and limited to X11. The lag has been noticeable.
I had 580 series, but see it also on 570 series. One machine had 575 series before, without lag,
but there were also other package upgrades at the time.
Yes, I use Gnome and Wayland on RL9. Everything works fast and well, but in some apps, a black screen sometimes appears, sometimes in parts of the window. In Pidgin, etc., you just move the window a little and the black screen (it’s only inside the application window) disappears, but this is rare.
Is this (in any way) related to this “solved” post?
https://forums.rockylinux.org/t/crushingly-slow-display-since-upgrade-to-9-6/19147/18
I’ll give messing with the power state a try. Thanks again.
This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.