You are not logged in.
Today my system has been randomly freezing and crashing to the point of me having to kill power to it to reboot.
None of what I tried really did anything, changing WM, DE or even kernel.
Dmesg and journalctl didn't give me anything solid.
So I decided to boot into a live session of Fedora through a usb stick just to see if a hardware or a software issue is causing this freeze. It did not freeze once.
I thought to myself that I must've borked my install somehow so I did a brand new fresh install of arch (I used archinstall just to point it out), and it is still freezing, the frequency of the freezes is much lower (from once every 5-10 minutes to hours now)
I am at a loss at this point, if I need to post any of my crash/error logs I will.
Thank you,
Offline
After reboot, have you tried to use
jounalctl -b -1to see if anything is wrong? As dmesg is only work for the current boot.
When system is freezed, can the capslock led on keyboard response to clicking the capslock key? Can the system repond for ping? Can you ssh to it?
Have you checked the /var/lib/systemd/coredump, if kernel crashes, there should be some file.
Could you try to use multi-user.target to see if it's OK? Or if there is any gpu driver, try to disable them?
Offline
After reboot, have you tried to use
jounalctl -b -1to see if anything is wrong? As dmesg is only work for the current boot.
When system is freezed, can the capslock led on keyboard response to clicking the capslock key? Can the system repond for ping? Can you ssh to it?
Have you checked the /var/lib/systemd/coredump, if kernel crashes, there should be some file.
Could you try to use multi-user.target to see if it's OK? Or if there is any gpu driver, try to disable them?
https://pastebin.com/4FMSigK1 results of the journalctl, when I search for this kind of error I get to add to my boot settings the dcdebugmask thing (which I already have added).
Capslock LED does not change when crashed. It also disconnects from network (tested by being in a discord call with friends).
There are three entirely different files in the coredump directory "wineboot, wine64-preloade, steamwebhelper" but those appear only if a crash occurs if steam is open/a wine game is running.
Crashes happen even if literally nothing is running, at which point nothing is added to coredump.
Could you elaborate on the "multi-user.target" thing ,how do I do that? I also have a single dedicated GPU, I am running mesa, radeon-vulkan and amdgpu (amdgpu is loaded as a driver, checked with lspci -k).
Offline
It seems to be the problem of amd gpu.
Please post the full log. The log you post is that amd driver is doing error analysis. There should be some detailed info in the following and before.
When I search for this kind of error I get to add to my boot settings the dcdebugmask thing (which I already have added).
Do you mean after configuring dcdebugmask, the problem still exist? Or you are now trying and to see if that helps? ( sorry for my poor English)
Could you elaborate on the "multi-user.target" thing ,how do I do that?
you could add
systemd.unit=multi-user.targetin kernel parameter or use
systemctl get-default # write down the original target
systemctl set-default multi-user.targetto check if it's indeed the gpu thing that leads to system freeze.
Offline
After multiple testing, reboots and reinstalls. I narrowed down the freeze to using linux-lts. I still don't know why it causes the freeze, but that freeze does not happen on the standard linux kernel.
Nothing else I tried seemed to fix it except switching kernels.
Offline