Bug: X server hangs on Linux with NVIDIA Quadro P620

So on Linux (I am using Arch, but it’s also happened on other distributions on the same machine), after a certain amount of time using the machine with the Nvidia card output, the X server just hangs. I can still perform inputs on whatever the last window I clicked was, and I can still switch to a TTY with a hotkey, but I cannot switch windows, perform mouse actions, or do just about anything else. the windows are still present on my screen and my mouse can still move, Linux just completely ceases processing inputs. This is still the case if I leave the computer on and come back to it hours later. It still just does not take any input.

Despite this bug being both VERY disruptive and VERY common for me to run into, I haven’t completely isolated its cause. My best guess is that it has something to do with the amount of work that the Nvidia card is asked to do; it seems to happen more often when I’m watching a video or doing work on multiple screens, for example.

It’s worth mentioning that for whatever reason as well, this bug seems to appear with the open-source Nouveau drivers as well as the official Nvidia ones. I don’t know if that means that it’s a hardware issue or just the two drivers happening to run into the same issue.

Log is attached.

nvidia-bug-report.log (674.4 KB)

Unfortunately, the log was created after reboot, so there’s little left from the crash. Seems nvidia_glxserver.so is crashing, this might be a problem with system memory. Please check if removing memory modules/running some rounds of memtest86 brings something up.

I found some time to work on this again and I was able to get a new crash log while the X server was still alive. It’s below.

nvidia-bug-report-2.log (822.7 KB)

Haven’t done any memtests yet because I figured that this would be more useful. I also have 32GB of RAM on this machine, so if that’s the issue then something must be going really, really, really wrong…

There are a lot of crashdumps included. Seems it’s always triggered by Discord so it’s most likely no memory defect. Still not really clear to me why Discord is able to crash the driver. Can you check if really only Discord is trigggering this by running for some time without it?