View Issue Details

IDProjectCategoryView StatusLast Update
0000590AlmaLinux-9kmodpublic2025-12-03 22:01
Reportersboresch Assigned Toalukoshko  
PrioritynormalSeveritymajorReproducibilityalways
Status closedResolutionfixed 
Platformx86_64OSAlmalinuxOS Version9.7
Summary0000590: No nvidia kernel modules for kernel-5.14.0-611.9.1.el9_7
DescriptionI just set up two boxes with nvidia cards using the 9.6 ISO, followed by dnf upgrade to bring them to 9.7. Both are installed in text mode only, but I need nvidia drivers for cuda.

Box 1 is working great, installing nvidia drivers etc as described in the recent blog post worked without a hitch. [This was about 2 weeks ago (around Nov 17), last explicit 'dnf upgrade' about one week later]

Box 2 was set up only yesterday; installation of the nvidia packages reported no error, but even after a reboot, no nvidia module is loaded. Manual attempts produce:

modprobe: FATAL: Module nvidia-drm not found in directory /lib/modules/5.14.0-611.9.1.el9_7.x86_64

I was puzzled, as I thought that the two boxes were for all practical purposes identical, but:

On the working box, kernel and modules are kernel-5.14.0-611.5.1.el9_7; the (relevant?) nvidia package is kmod-nvidia-open-580.105.08-2.el9
/lib/modules/5.14.0-611.5.1.el9_7.x86_64 contains the required nvidia modules and all works well.

Box 2, I installed only yesterday, followed by dnf update/upgrade, and this installed a newer kernel: kernel-5.14.0-611.9.1.el9_7 [NOTE the 9 instead of the 5, it took me a while to see the difference!!], but the nivida package is still: kmod-nvidia-open-580.105.08-2.el9.

On box 2, I have /lib/modules/5.14.0-611.5.1.el9_7.x86_64 (containing the nvidia modules), but /lib/modules/5.14.0-611.9.1.el9_7.x86_64 is the one used by the kernel, and 611.9 does NOT contain any nvidia modules. Thus, this machines is 'severely broken'; fortunately, it's text mode only (as by now nouveau has been blacklisted).

The older 611.5 kernel is 'gone', so I cannot force grub to boot into the older kernel. (I still see the kernel from Almalinux 9.6, not sure whether this should have happened or not).

So, it seems to me that the most recent kernel upgrade has no matching nvidia modules, and as no dkms seems involved, it's not as if I could rebuild these modules myself.

Steps To ReproduceI guess: do 'dnf upgrade' that pulls in the new kernel on box 1. Since the working box is in semi-production, I can't do/try this at the moment.
Additional InformationI am really unexperienced with RHEL derived Linux systems (I do mostly Debian/Ubuntu). Hence it may well be that I am doing / did something stupid or overlooked something trivial
TagsNo tags attached.

Activities

sboresch

2025-12-03 18:59

reporter   ~0001196

The just released kmod-nvidia-open-580.105.08-3.el9 rpm(s) fixed this. See thread at https://chat.almalinux.org/almalinux/pl/wojxosmzhpg6xyj5k76jcqkhio
So, as far as I am concerned, please close.

Issue History

Date Modified Username Field Change
2025-12-03 14:04 sboresch New Issue
2025-12-03 18:59 sboresch Note Added: 0001196
2025-12-03 22:01 alukoshko Assigned To => alukoshko
2025-12-03 22:01 alukoshko Status new => closed
2025-12-03 22:01 alukoshko Resolution open => fixed