Hard freezes with opensource driver about open-gpu-kernel-modules HOT 12 OPEN

nvidia moment

from open-gpu-kernel-modules.

There were actually conflict between nouveau and nvidia opensource flavor. I will let you know when i boot again with the opensource nvidia flavor to collect some data.

Regards

from open-gpu-kernel-modules.

nvidia moment

LOL ya

from open-gpu-kernel-modules.

I've managed to make it "work" but it freezes regularly, and only hard reset helps.

from open-gpu-kernel-modules.

@mtijanic Do you have any idea about this, or do someone suffer it?

from open-gpu-kernel-modules.

Hi, thanks for the report!

From the attached logs, I see:

NVRM: Open nvidia.ko is only ready for use on Data Center GPUs.
NVRM: To force use of Open nvidia.ko on other GPUs, see the
NVRM: 'OpenRmEnableUnsupportedGpus' kernel module parameter described
NVRM: in the README.

Are you sure you are running the open source flavor of the driver? You can verify with modinfo nvidia | grep license

The logs also previously show failure to initialize nvidia.ko because another driver (probably nouveau) is already active.

Let's first make sure we're running the open source driver, and if so we can try to up the logging level and see what went wrong.

from open-gpu-kernel-modules.

Hello again, well there's the correct bugreport.

nvidia-bug-report.log.gz

Don't have freeze at the moment (and i cannot provide any report while it freeze)

from open-gpu-kernel-modules.

@sandikata, here are few possible experiments:

(1) Do you see similar freezes with the 515.43.04 binary kernel modules, as well as the open kernel modules?

(2) Did you see similar freezes with previous NVIDIA driver releases?

(3) When the system freezes, is the system still accessible over the network? Or, is the only option at that point to reboot?

from open-gpu-kernel-modules.

Hello,

1. Yes only on OpenSource 515.43.04, proprietary from same version works.
2. Here can say no
3. Mostly on heavy load (gaming, encoding and etc), not accessible in any way the only is reboot.

from open-gpu-kernel-modules.

I am having the same issue here:
Kernel: 5.17.7-arch1-1
Driver: nvidia-open 515.43.04

Seems like something in the kernel module is locking up whenever I run an OpenGL game (i.e. Minecraft), as there is a lot of reports of timeouts. Strangely running a Vulkan game (GTA 5 through Proton and DXVK) works perfectly.

This does not happen with the proprietary kernel modules; both OpenGL and Vulkan games work perfectly.

Please note I have tried adding the nvidia-drm.modeset=1 kernel parameter which changed nothing.

Attached is a snippet of a log that I think is of interest.
nvidia.log

Hopefully this helps

from open-gpu-kernel-modules.

I should add that if I am quick enough after the freeze occurs, I can switch to a tty and have it be fully functional.

from open-gpu-kernel-modules.

I'm not sure if that was a reply to my comment, but I attempted to blacklist the nouveau driver, which didn't solve my issues. Upon further inspection, on my system nouveau was already blacklisted by the nvidia-utils package in /usr/lib/modprobe.d.

from open-gpu-kernel-modules.