Computer Crashes

[bashrules]

Hi all,

My desktop (i3-2120 from 2012) runs NetBSD 9 for the last 11 months. It runs 24/7 and crashes once a month. This is very bad as I rely on it being remotely accessible.

Before, it was not my computer and I don't know of any known hardware problems.

After the crash:

• the power LED is still on and the fan continues to spin. This makes me believe it's not a hardware problem.
• the monitor does not detect a signal from the graphics card. Here I'm not sure what that means. If the kernel crashes, what is supposed to be on the screen? Who is driving the graphics card/monitor?
• pings are not answered.
• there was no power outage - all other computers in the household are still running.
• BIOS is set up to "keep the power-state" after a power-failure, e.g. when it was previously on, it will be turned on. As NetBSD did not boot again, the power was never really lost.
• it was not a graceful shutdown. When NetBSD came up after a manual power-cycle, filesystem checks took place (well, the journal is replayed)

I follow NetBSD cvs netbsd-9 branch. I believe that is the stable branch. The NetBSD config is the default.

NetBSD XXX 9.1_STABLE NetBSD 9.1_STABLE (GENERIC) #2: Sun Jan 3 11:19:52 PST 2021 root@XXX:/usr/obj/sys/arch/amd64/compile/GENERIC amd64

The last crash must have happened early morning of Feb 24 (I power-cyceled the computer at 9:00am. Last cron job activity at 3:00am.) Log files are "blank" in the early morning. /var/log/messages:

Feb 21 13:16:04 XXX /netbsd: [ 3620751.3847045] kern error: [drm:(/usr/src/sys/external/bsd/drm2/dist/drm/i915/intel_fifo_underrun.c:230)cpt_set_fifo_underrun_reporting] *ERROR* uncleared pch fifo underrun on pch transcoder A
Feb 21 13:16:04 XXX /netbsd: [ 3620751.3847045] kern error: [drm:(/usr/src/sys/external/bsd/drm2/dist/drm/i915/intel_fifo_underrun.c:381)intel_pch_fifo_underrun_irq_handler] *ERROR* PCH transcoder A FIFO underrun
Feb 21 14:33:25 XXX /netbsd: [ 3625392.1028982] kern error: [drm:(/usr/src/sys/external/bsd/drm2/dist/drm/i915/intel_fifo_underrun.c:230)cpt_set_fifo_underrun_reporting] *ERROR* uncleared pch fifo underrun on pch transcoder A
Feb 21 14:33:25 XXX /netbsd: [ 3625392.1028982] kern error: [drm:(/usr/src/sys/external/bsd/drm2/dist/drm/i915/intel_fifo_underrun.c:381)intel_pch_fifo_underrun_irq_handler] *ERROR* PCH transcoder A FIFO underrun
Feb 24 09:03:22 XXX syslogd[184]: restart
Feb 24 09:03:22 XXX /netbsd: [   1.0000000] Copyright (c) 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003, 2004, 2005,

No other log files have timestamps around that time (well, besides cron.log).

What do you suspect, hardware problem or kernel crash?

Is there a knob to keep the kernel in a debugger upon a crash?

What can I do to root-cause the crash?

[ohmpr]

i don't know anything about netbsd, but i would suspect maybe kernel panic due to mem leak or mem going bad. As for how to diagnose i would say use whatever methods to test via process of elimination with those suspected causes in mind. IDK, just a guess.

[gogh20100]

2 suggestions:

- A core dump should have been created, you can load it with gdb to see where it crashed.

- /var/log/messages reports an error in the video i915 kernel module. You may want to configure X to use the wsfb driver (instead of the automatically intel driver) to confirm this hypothesis (your machine should no longer crash).

If this is the correct diagnosis, you should open a PR including the error messages in /var/log/messages, the output of uname -a, your full dmesg and the information obtained from the core dump.

[ohmpr]

- /var/log/messages reports an error in the video i915 kernel module. You may want to configure X to use the wsfb driver (instead of the automatically intel driver) to confirm this hypothesis (your machine should no longer crash).

yes, i ignored this because i assumed it had already been ruled out as the source of the issue since it was right there in the log posted and not mentioned. Although i probably shouldn't have since it says "kern error" and that sounds like something that could cause a kernel panic.

[bashrules]

Thank you for getting back to me.

Originally Posted by ohmpr

i don't know anything about netbsd, but i would suspect maybe kernel panic due to mem leak or mem going bad. As for how to diagnose i would say use whatever methods to test via process of elimination with those suspected causes in mind. IDK, just a guess.

I was running memtest over night. Memory is good.

A kernel panic should create a core-file. There is none.

Originally Posted by gogh20100

2 suggestions:

- A core dump should have been created, you can load it with gdb to see where it crashed.

- /var/log/messages reports an error in the video i915 kernel module. You may want to configure X to use the wsfb driver (instead of the automatically intel driver) to confirm this hypothesis (your machine should no longer crash).

If this is the correct diagnosis, you should open a PR including the error messages in /var/log/messages, the output of uname -a, your full dmesg and the information obtained from the core dump.

There is no core-file. Do you happen to know how core files are created, e.g. under which circumstance are they / are they not created?

I doubt that these i915 kernel logs are responsible for this crash 3 days later.

[Sehnsucht94]

Originally Posted by bashrules

There is no core-file. Do you happen to know how core files are created, e.g. under which circumstance are they / are they not created?

I doubt that these i915 kernel logs are responsible for this crash 3 days later.

It's covered in the wiki