[SOLVED] Post-upgrade (F25) Nvidia Not Loading unless on old (F24) Kernel

DJOtaku · 02-19-2017, 07:13 PM

Today I upgraded to F25 to get access to the latest version of Kodi. After rebooting sddm would not load. I went to another terminal and ran startx. It complained there was no OpenGL. So I went to Negativo's page and saw the following 2 comments:

Quote:

Joshua
February 16, 2017 at 9:29 pm

Help! Installed nvidia updates today Thu Feb 16 and no GL — KDE plasma fails, sddm fails. Xorg starts from console with no DE but can run firefox to send this help request. I can not figure out what broke. I’m running kernel-4.7.10 but everything else is up to date.

Joshua
February 17, 2017 at 1:07 am

I’m not sure what happened but I think my initrd.img had the previous driver version in it. Luckily I had a backup initrd.img. I’m really tempted to just disable updates once my system is running the way I like. Seems like something breaks with every other update on fedora 25 — it’s the buggiest release in a while.
Anyway, I need to figure out how to configure dracut so it doesn’t happen again.

So I rebooted and loaded into my F24 kernel and everything works. Is there a way to confirm this is the issue and if it is, to redo my initrd.img so that I can boot into my F25 kernel?

Thanks!

business_kid · 02-21-2017, 04:25 AM

The nvidia module has to get installed in the new module tree. For yourself, try this:

Code:

cd /lib/modules
find -name 'nvidia*'

The module that shows in your F24 module tree should show in the F25 one.You need the kernel source to reinstall, but look it all up on Fedora's site and that takes some of the pain away.

DJOtaku · 02-21-2017, 07:05 PM

looks like it's already there:

Code:

*find -name "nvidia*"
./4.9.9-200.fc25.x86_64/extra/nvidia
./4.9.9-200.fc25.x86_64/extra/nvidia/nvidia-drm.ko.xz
./4.9.9-200.fc25.x86_64/extra/nvidia/nvidia-uvm.ko.xz
./4.9.9-200.fc25.x86_64/extra/nvidia/nvidia-modeset.ko.xz
./4.9.9-200.fc25.x86_64/extra/nvidia/nvidia.ko.xz
./4.9.9-200.fc25.x86_64/kernel/drivers/net/ethernet/nvidia
./4.9.9-100.fc24.x86_64/extra/nvidia
./4.9.9-100.fc24.x86_64/extra/nvidia/nvidia-drm.ko.xz
./4.9.9-100.fc24.x86_64/extra/nvidia/nvidia-uvm.ko.xz
./4.9.9-100.fc24.x86_64/extra/nvidia/nvidia-modeset.ko.xz
./4.9.9-100.fc24.x86_64/extra/nvidia/nvidia.ko.xz
./4.9.9-100.fc24.x86_64/kernel/drivers/net/ethernet/nvidia
./4.9.5-100.fc24.x86_64/extra/nvidia
./4.9.5-100.fc24.x86_64/extra/nvidia/nvidia-drm.ko.xz
./4.9.5-100.fc24.x86_64/extra/nvidia/nvidia-uvm.ko.xz
./4.9.5-100.fc24.x86_64/extra/nvidia/nvidia-modeset.ko.xz                                                                                                                                                           
./4.9.5-100.fc24.x86_64/extra/nvidia/nvidia.ko.xz
./4.9.5-100.fc24.x86_64/kernel/drivers/net/ethernet/nvidia

business_kid · 02-22-2017, 04:13 AM

Hmmm. No expert here on the nvidia binary - it's been a while. What does

Code:

lsmod |grep nvidia

sho is being loaded? If it's there, you can eliminate that.

DJOtaku · 03-16-2017, 07:29 AM

Sorry it's taken me so long to come back and reply. I've been busy and didn't want to disturb my computer because I needed to complete some tasks. Although it's not satisfying for anyone who would find this thread while googling, I can report that with the latest kernel update everything works. So I'm not sure if something didn't get triggered correctly on upgrade for getting nvidia working or if it went along with some selinux or other such kmod issues I read about in the last month, but it's all working now.

Frankly, every time this happens I consider switching to AMD since they're mostly free (I think the latest cards do require a proprietary blob if you want full functionality for video games), but so far their performance just hasn't matched nvidia consistently. Perhaps on their new chipset coming out this year they'll do well enough to change on my next hardware upgrade.