Arch Linux

Stimmar

Member

Registered: 2021-12-15

Posts: 3

Random system hang that can only be resolved by hard reset

Seemingly randomly I get a total hang where audio stops working, the screen freezes, a black square appears on it, fans start running wild, and I can't even ssh into the system.
I tried letting it run for a while in that state, but it seems that the only way to make it usable is a hard reset (pressing the power button for a few seconds).

This seems to be random, as it happened recently with only a few hours of uptime, and didn't happen at all for the last session of over two weeks of uptime. It doesn't seem related to suspending since I usually notice it while actively using the system.

I suspect it has to do with the new GPU, since I first noticed it happen after I installed it a couple months back and the errors in the journalctl log mention it. I first thought it might be due to the proprietary driver I installed because of Blender so I removed that, but it kept happening.
I've tried searching if others had the same issue, but haven't managed to find anything I thought was relevant.
The GPU is AMD RX6600 XT, and the old one was AMD RX580.

Here's some kernel-space error I noticed in the output of journalctl -b-1, and here's the last 10 minutes of the session where that error occured.

Possibly related but otherwise minor, lightdm.service often fails on boot, but once I manually restart it everything's fine. I think this is the relevant Xorg log of that failure, and here's the output of journalctl -b -u lightdm.service. I think I messed around with it at some point as PulseEffects (now EasyEffects) wouldn't start on boot at the time, so that might be at fault.

I run i3 with picom as a compositor and have two monitors, one 1440p and the other 1080p in case that matters.

Please let me know if any other logs or info could be useful, and thank you for your time.

Edit: I finally looked into the lightdm issue and it seems like setting logind-check-graphical=true should solve it, but again that's a minor issue.

Last edited by Stimmar (2021-12-15 02:07:40)

V1del

Forum Moderator

Registered: 2012-10-16

Posts: 25,296

Re: Random system hang that can only be resolved by hard reset

The kernel errors in the first excerpt don't mean much that's pretty normal for CD drives with a CD in them. The second excerpt doesn't really contain anything which can be normal if you actually get a proper crash that prevents the journal from flushing. SMU errors come up with power issues some time so maybe trying to enable a high power profile might help: https://wiki.archlinux.org/title/AMDGPU … cy_problem

Other than that I'd say test a newer kernel like linux-mainline or linux-amd-staging-drm-next-git if you want to go full experimental.

FWIW the lightdm issue should indeed be solvable with that, otherwise for the more generic method, see https://wiki.archlinux.org/title/Kernel … _KMS_start

tonitch

Member

Registered: 2020-08-09

Posts: 11

Re: Random system hang that can only be resolved by hard reset

Fwiw, I have had the same issue with my last setup, when I was running out of memory... I did my first install without swap fs ☺️

I noticed it because I placed the memory usage in my taskbar and on my desktop with conky...

Stimmar

Member

Registered: 2021-12-15

Posts: 3

Re: Random system hang that can only be resolved by hard reset

The kernel errors in the first excerpt don't mean much that's pretty normal for CD drives with a CD in them. The second excerpt doesn't really contain anything which can be normal if you actually get a proper crash that prevents the journal from flushing. SMU errors come up with power issues some time so maybe trying to enable a high power profile might help: https://wiki.archlinux.org/title/AMDGPU … cy_problem

Other than that I'd say test a newer kernel like linux-mainline or linux-amd-staging-drm-next-git if you want to go full experimental.

FWIW the lightdm issue should indeed be solvable with that, otherwise for the more generic method, see https://wiki.archlinux.org/title/Kernel … _KMS_start

I don't have a CD inside, but oh well, if it's not causing problems I don't care.
I'll try out changing the power profile, thanks for the tip. This won't make it run at max power all the time or anything silly like that, it'll just raise power consumption, right?

Fwiw, I have had the same issue with my last setup, when I was running out of memory... I did my first install without swap fs ☺️

I noticed it because I placed the memory usage in my taskbar and on my desktop with conky...

Firefox does end up eating a lot of RAM after being run for as long and with as many tabs as I sometimes run it, but as I have 16GB of RAM and the frankly wasteful 8GB of swap I doubt it'd happen on a relatively new session without me noticing. Swapping usually means everything's very slow, so it's pretty noticeable.

Stimmar

Member

Registered: 2021-12-15

Posts: 3

Re: Random system hang that can only be resolved by hard reset

Do you think something like this would make it so the power saving profile (2) doesn't get used, but all other ones are switched between automatically? I assume nothing should break if I try it, but I just wanted to ask while I'm here.

echo "0 1 3 4 5 6" > /sys/class/drm/card0/device/pp_power_profile_mode