WebRTC-Related Software Triggering System Reboot

Spíosra

Spíosra

So it seems that things like Discord and Google Meet for whatever reason trigger an instant system reboot on my FreeBSD 14.0-RELEASE. And by "instant" I mean like an unprompted power cycle. Can't seem to find any interesting crash logs of the sort. It seems to happen quite predictably within 1-5 minutes of joining a call. Tried on both Firefox and Chromium. The interesting thing is that there are absolutely no issues on Linux where I not only use Discord regularly but also play very demanding games. Other things on FreeBSD such as games, Blender, and compilation work like a charm. Playing videos presents no issues. But it seems to be fairly isolate to FreeBSD. For this reason I don't suspect hardware issues.

I am at a loss as to what it could be nonetheless, especially since I can barely find any useful traces. dmesg does not seem to present any interesting info. For some reason I suspect on a whim it might be related to the GPU. Have tried different variations and versions of drm_kmod to no avail.

Curious if any might have an idea? I realize there are similar issues popping up around drm_kmod in particular, but no solution seems to have worked for me at least.

For what it's worth, my system is an AMD Ryzen 5 7600X and an AMD RX 6800 XT.
 
Alexander88207 said:
You have maybe the current drm-5.15 panic problem as many others PR 274770.

Using the suggested branch from the bug report fixes the panics for me.
Click to expand...
Maybe, but compiling and installing drm_kmod from the 5.15-lts-focal branch doesn't seem to fix it for me. Neither does downgrading to 5.10.

It is very annoying. I'm quite intent on driving FreeBSD daily. Curiously, Telegram calls do not seem to present any issue.
 
After thinking a bit an instant reboot without a trace sounds for me hardware issue.

Even if it would be not, like you have said, what if FreeBSD triggers something what others do not?
 
Alexander88207 said:
After thinking a bit an instant reboot without a trace sounds for me hardware issue.

Even if it would be not, like you have said, what if FreeBSD triggers something what others do not?
Click to expand...
I would believe it more if it wasn't as predictable as it seems to be. But even if it was seemingly at random, I find it difficult to believe that Linux would perpetually leave it untouched.

That said, I do have a spare RX 580 around. One thing I could try doing is plugging it in and see if I experience the same thing. At least then it might isolate the issue more.
 
You would only find crashdumps if you turn them on and have enough space for them (usually swapspace).

I had cases of hardware errors only appearing under one OS but not another. But you are right, if you reliably crash within 5 minutes it is more likely a driver problem. Unless it is heat. FreeBSD might produce more heat faster.
 
After a bit more tinkering with voltages and timings I was able to achieve stability thus far, though I haven't stress-tested the system with prolonged daily use yet. Wondering if it turns out to be a memory issue after all or related to RAM configuration.

Will keep this posted if it turns out to be resolved as such.
 
So after looking into it more I think it may be related to the network driver. I happen to have a Realtek 8125 and silly me had not actually enabled crash dumps before. I can see now that there are page faults around re0 and taskq processes similar to those in a few other issue reports pertaining to this device and the realtek-re-kmod driver.

I suppose it could still be a memory issue in general, but I think this can almost be ruled out as pertaining to anything GPU-related.
 
You must log in or register to reply here.
Back
Top