Spontaneous reboot: GPU hang on 14.5-RELEASE

I was reading some news articles, running waterfox, when my machine (thinkpad X220) spontaneously rebooted itself. After the reboot, I can see the following was the last thing to happen in /var/log/messages

Nov 5 20:06:38 box kernel: drmn0: [drm] GPU HANG: ecode 6:1:85fffffc, in Renderer [200136]
Nov 5 20:06:38 box kernel: drmn0: [drm] Resetting chip for stopped heartbeat on rcs0
Nov 5 20:06:38 box kernel: drmn0: [drm] Renderer[200136] context reset due to GPU hang
Nov 5 20:45:50 box syslogd: kernel boot file is /boot/kernel/kernel
Nov 5 20:45:50 box kernel: ---<<BOOT>>---
Nov 5 20:45:50 box kernel: Copyright (c) 1992-2023 The FreeBSD Project.

There's a big gap in timestamps between the gpu hang at 20:06 and the reboot at 20:45. Maybe the event that caused the reboot doesn't have any log entries, but I wondered if there is a connection. I can't find any coredumps on the box.

$ freebsd-version -ukr
14.5-RELEASE-p2
14.5-RELEASE-p2
14.5-RELEASE-p3

It's pretty strange, I don't think I've ever seen freebsd do that before on this hardware.
Has anyone else seen this? Is it a known bug? Any ideas?
 
I've seen similar (in my case it was a GPU hang, had to power off / on the machine to recover), but probably different hardware as mine was a gfx card in my desktop workstation.
What GPU is in that ThinkPad of yours?
 
Lenovo ThinkPad X220 (4290) Specifications, Tests and Review Screen Size 12.5" Resolution 1366 x 768 (HD) CPU Intel Core i5 (2nd Gen) 2520M / 2.5 GHz GPU Intel HD Graphics 3000 RAM 2 GB (1 x 2 GB) Core Qty Dual-Core. It's a sandy bridge cpu. But given that the reboot happened appox 39 minutes after the gpu hang, it's probably unrelated. It's the only error I could see in messages. I suspect a hardware glitch. It's running i915kms.

There are no corefiles on the system. There was no kernel panic message. It's like someone pulled the reset line on the cpu. Perhaps the heatsink compound needs renewing. I have dumpdev set to AUTO in rc.conf and 4GB of swap space configured, but 'service savecore onestart' reports that no crash dumps are available, so there is no kernel dump to debug. The reboot was instantaneous, there was no delay as would be expected it was writing a corefile or dump to disk.

I just hope I haven't acquired a rootkit from somewhere. Seems unlikely. Maybe it's time for a fresh install on another box and move to 15.x.
 
I checked microcode update, I have the message

CPU microcode: updated from 0x29 to 0x2f

in dmesg. So it appears to be at the latest level.

I am thinking that the most likely explanation for this event is a hardware fault. I might try renewing the heatsink paste. It's pretty unusual, this box has been totally stable up to now, for over a year. I guess it's quite an old laptop. I will watch to see if it happens again.
 
drm-61-kmod-6.1.128.1405000_11 is already installed; but it was worth checking.

I was playing a yt video with mpv when the reboot occurred, using the 'gpu' driver; I'm going to try switching to 'gpu-next' which is their latest stuff. The 'gpu-next' driver is their new updated driver, while 'gpu' is old and receiving few updates. Although they recommend 'gpu' for older hardware, it won't hurt to try gpu-next, since I definitely had the gpu hang error occur while running the gpu driver, so perhaps there is a bug in it.
 
Oh well, that didn't work; I get a large number of ioctl errors like this with the first video I tried.

ioctl(I915_GEM_USERPTR) failed. Try running as root but expect poor stability.
ioctl(I915_GEM_USERPTR) failed. Try running as root but expect poor stability.
... repeats ad infinitum


The final post says it's broken in freebsd, although that was in 2023.

Back to 'gpu' for the time being, gpu-next looks like a being red herring.
 
Back
Top