Bug 6981

Summary: akmod-nvidia-550.90.07 fails in nv_drm_atomic_commit with kernel-6.9.5-100
Product: Fedora Reporter: Fritz Elfert <fritz>
Component: nvidia-kmodAssignee: Nicolas Chauvet <kwizart>
Status: RESOLVED FIXED    
Severity: enhancement CC: fritz, leigh123linux, leigh123linux
Priority: P1    
Version: f39   
Hardware: x86_64   
OS: GNU/Linux   
namespace:
Attachments: nvidia related kernel messages from booting kernel 6.9.5
nvidia related kernel messages from booting kernel 6.9.4 (as a reference)

Description Fritz Elfert 2024-06-26 00:34:35 CEST
Created attachment 2560 [details]
nvidia related kernel messages from booting kernel 6.9.5

I have akmod-nvidia-550.90.07 installed.

Since the update from kernel-6.9.4-100 to kernel 6.9.5-100 the nvidia_drm
module produces many errors in nv_drm_atomic_commit. The result is a *very*
slow boot. GUI login appears eventually, but it takes approx 2 minutes.

The same nvidia driver works fine when selecting kernel 6.9.4-100 in grub.

Attached is a log excerpt of the boot process, which was generated
by running journalctl -b -t kernel | grep -i nvidia
Comment 1 Fritz Elfert 2024-06-26 00:35:48 CEST
Created attachment 2561 [details]
nvidia related kernel messages from booting kernel 6.9.4 (as a reference)
Comment 2 leigh scott 2024-06-26 08:38:25 CEST
Try downgrading Plymouth.


sudo dnf downgrade *\Plymouth\*
Comment 3 leigh scott 2024-06-26 08:40:05 CEST
(In reply to leigh scott from comment #2)
> Try downgrading Plymouth.
> 
> 
> sudo dnf downgrade *\Plymouth\*

Scratch that, I didn't notice your running F39
Comment 4 leigh scott 2024-06-26 08:41:37 CEST
File the issue at nvidia https://forums.developer.nvidia.com/c/gpu-graphics/linux/148
Comment 5 Nicolas Chauvet 2024-06-26 09:21:29 CEST
Also you might try to reproduce with 555 beta
https://rpmfusion.org/Howto/NVIDIA#Rawhide
Comment 7 Fritz Elfert 2024-06-26 22:41:23 CEST
(In reply to Nicolas Chauvet from comment #5)
> Also you might try to reproduce with 555 beta
> https://rpmfusion.org/Howto/NVIDIA#Rawhide

This helped.
With 555 beta, the errors shown in the kernel log are gone and boot is fast again.

I did however struggle with a strange behavior (which probably is specific to my graphics card - an older EVGA GTX 980 Ti):

This card has 1xDVI 1xHDMI and 3xDP. Until now, my primary display was connectetd to the card's DVI port and the second display to the HDMI port.
Now with 555 beta, I was unable to get a signal on the DVI port. Luckily,
connecting the primary display to one of the DP ports did remedy the situation.

FTR:
I did upgrade to 555 beta in a slightly different way than described here:
https://rpmfusion.org/Howto/NVIDIA#Latest.2FBeta_driver
The last of the 5 cmdlines mentioned there needed to be modified:

sudo dnf --enablerepo=rpmfusion-nonfree-rawhide upgrade akmod-nvidia

Thanks a lot for the hint.