[REGRESSION] [495.29.05] NVIDIA drivers dead on resume / software suspend fails intermittently

After suspending my PC for long periods of time, when it resumes, the screen is dead and the NVIDIA drivers are seemingly also dead as I cannot do anything with them including reading any files under /proc/driver/nvidia

This has never happened before.

I’ve lost the last nvidia-bug-report log, will upload it later when it happens again.

PC Specs:

Linux 5.14.14 (vanilla)
Fedora 35
NVDIDIA drivers: 495.29.05
GPU: GTX 1660 Ti
CPU: Ryzen 7 5800X
Some x570 based motherboard.

Upgraded to 5.14.17, the issue persists.

I’ve no idea what to do because on resume Xorg is completely dead and unresponsive (doesn’t update its log file), NVIDIA drivers fail to return anything - all their interfaces are dead - I cannot read any files under /proc/driver/nvidia and nvidia* utilities freeze. Oh, I cannot rmmod -f nvidia* either. NVIDIA drivers essentially die on resume.

Strangely my laptop with a different kernel and no NVIDIA drivers fails to resume occasionally as well. I’m thinking there’s a regression in 5.14.x which is hard to spot and which affects very few devices. It’s doubly weird considering the laptop and the PC have completely different CPUs (Intel vs AMD), different chipsets and are separated by 5 years.

cat /sys/power/mem_sleep
s2idle [deep]

Having the same problem, no idea how to debug. No “resume” action shows up in the logs at all. Power comes back, no visual output. Have to hard reset.

Can’t reproduce it any longer after updating to Linux 5.15.4 and NVIDIA drivers 495.44

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.