You are not logged in.
For two weeks I've been encountering a system freeze. It only occurs when X server is running (I've had the machine running for two days straight console-only, it didn't occur in those two days). At first I've attributed this to the GPU failing as it's quite old, but switching to the integrated GPU didn't help. The freeze doesn't leave any traces in the logs displayed by journalctl, unfortunately. The freeze itself looks like this: the screen stops updating, the last bit of sound plays several times and then the sound stops, neither mouse nor keyboard respond. My biggest problem is that I don't know how to localize the issue. How do I do that? What tools do I have?
> Don't get impatient, you cannot solve problems by coding, you can only solve them by thinking.
Offline
How are you determining that the journal doesn't contain any traces? Often messages do tip you off despite not looking like a error, or the relevant messages happen way earlier than the freeze. You might want to post the journal, maybe someone else can infer information from it. Right now, there's a prevalent issue with the enablement of iommu in the 4.13 kernel, so ruling that out might be a good step currently.
For example:
https://bbs.archlinux.org/viewtopic.php?id=230362
https://bbs.archlinux.org/viewtopic.php?id=230115
If you see similar messages that might be connected.
Offline
I'll leave the machine running with X server on overnight. I expect it to freeze by the time I get back home from work. That'll give some fresh logs, which I'll post. iommu is set to soft, for quite some time already. I can't remember why I set it this way, I vaguely recall it fixing some warning during boot.
> Don't get impatient, you cannot solve problems by coding, you can only solve them by thinking.
Offline
When it freezes, can you switch to a different console using CTRL-ALT-F2 (or F3 - F6) ?
Can you ssh into the box?
Can you ping the box?
Are the keyboard LEDs (Caps lock, Num lock) blinking?
Nothing is too wonderful to be true, if it be consistent with the laws of nature -- Michael Faraday
The shortest way to ruin a country is to give power to demagogues.— Dionysius of Halicarnassus
---
How to Ask Questions the Smart Way
Offline
No, I can't switch consoles. I haven't tried ssh'ing into the box, I'll try it next time. Same goes for pinging it. LEDs are locked in the state they were prior to a freeze, pressing Num/Caps lock does nothing.
> Don't get impatient, you cannot solve problems by coding, you can only solve them by thinking.
Offline
Check the log of the previous boot: "journalctl -b-1"
Next check xorg logs, share hardware specs and of course try https://fedoraproject.org/wiki/QA/Sysrq
Also details on the Xorg session and whether it happens with a "pure" X11 server (+xterm) as well could be interesting.
Do you see any pattern of related actions (video playback)?
Offline
One other thing, Ensure you have installed and configured your processor's microcode updates.
Nothing is too wonderful to be true, if it be consistent with the laws of nature -- Michael Faraday
The shortest way to ruin a country is to give power to demagogues.— Dionysius of Halicarnassus
---
How to Ask Questions the Smart Way
Offline
To my surprise, 20 hours of running just X server didn't cause a freeze. I'll leave it running overnight again, but this time with some slightly higher than usual load (mpd and several instances of vlc).
Hardware used (short):
CPU: AMD A10-6790K APU with Radeon(tm) HD Graphics
RAM: 10 GB, I think.
Hardware used (long): here
I don't really see a pattern in these. It happened once when I was editing something in Vim in console (X was running in another console, a browser was open, mpd was running), one more time when I was watching a video in vlc, and - if memory serves - when I was watching something on Youtube. We'll see if higher load on the system can cause a freeze to happen again.
> Don't get impatient, you cannot solve problems by coding, you can only solve them by thinking.
Offline
The problem doesn't seem to be reproducible anymore. I'm not sure if it's fixed by something I did, or if it's hiding. In case it resurfaces again, what should I do? Post here with the info gained, or start a new thread, and close this one for good?
> Don't get impatient, you cannot solve problems by coding, you can only solve them by thinking.
Offline