Help with a crash dump

nik.taylor

It's definitely nut. I uninstalled and there were no crashes. I re-installed and it's started crashing. Is this an integration issue you can look at or should I contact the nut team?

Thanks.

stephenw10

Yet it still crashed with Nut installed but disabled previously?

Were you able to replicate that? It seems hard to imagine that could happen if it really was disabled.

If it's a problem with the nut binaries in FreeBSD that would need to be reported upstream but there must be be a lot of people running that in FreeBSD.

Do you have any additional crash reports? Anything showing the NUT package specifically?

Steve

nik.taylor

I'm pretty sure it did. I'm going to disable it again and see if I get a crash dump with nut installed but disabled.

Here is the latest crash:

0_1551543086047_nut dump.txt

stephenw10

Mmm, well identical crash then. Implies probably software at least.

nik.taylor

Bumping this thread back up. I've continued to have this problem. I disabled nut for a few months and it didn't go away. I'm seeing crashes about once or twice a week still. Any next debugging steps?

Latest crash dump attached.

Thanks in advance.

crash_dump.txt

stephenw10

Hmm, well that is three almost identical crashes:

hardclock_cnt() at hardclock_cnt+0x131/frame 0xfffffe010e4d44e0
handleevents() at handleevents+0xc9/frame 0xfffffe010e4d4530
timercb() at timercb+0xad/frame 0xfffffe010e4d4580
lapic_handle_timer() at lapic_handle_timer+0xa2/frame 0xfffffe010e4d45c0
Xtimerint() at Xtimerint+0xa8/frame 0xfffffe010e4d45c0

I got to think it's some issue with the system clock being used on that system.

I see it's loading the speedstep driver (est), is powerd enabled? You might disabling it if so. It's been a while since I've seen one but some systems has issues with varying the cpu clock that would throw errors.

You could usually work past that by selevting a non variable system timer instead.
For example:

[2.5.0-DEVELOPMENT][admin@apu.stevew.lan]/root: sysctl kern.timecounter.choice
kern.timecounter.choice: ACPI-fast(900) HPET(950) i8254(0) TSC(800) dummy(-1000000)
[2.5.0-DEVELOPMENT][admin@apu.stevew.lan]/root: sysctl kern.timecounter.hardware
kern.timecounter.hardware: HPET

Steve

nik.taylor

@stephenw10 said in Help with a crash dump:

sysctl kern.timecounter.hardware

Thanks very much. powerd is not running.

I changed to HPET and will see what happens. I have to be honest, I know next to nothing about system timers so this is a stab in the dark for me. Will report back if anything happens.

nik.taylor

HPET didnt work. My system froze with 're1 watchdog timeout' within about 5 mins. Re-booted, reset HPET and same thing happened.

stephenw10

@nik-taylor said in Help with a crash dump:

watchdog timeout

You saw that error only using the HPET timecounter? What is the default timecounter there?

Did it actually 'freeze' or just stop responding to the network? That error is typical when using Realtek NICs but the system usually still responds at the console for example. It's only the NICs that fail.
You might try the alternative driver if so. That has been shown to help if the default driver is triggering watchdog timeouts.
https://forum.netgate.com/topic/135850/official-realtek-driver-binary-1-95-for-2-4-4-release

Steve

nik.taylor

I have seen that error once or twice in the past but it's pretty consistent when i changed to HPET.

It just stopped responding to the network. I could still use a keyboard directly attached to the box.

Default timecounter is TSC.

I'll try the updated driver.

Thanks for helping me keep on top of this.

stephenw10

No problem.
Yeah if the console was still active but the NIC/driver crashed out definitely try the alternative driver.

Ultimately there's not too much you can do. The Realtek NICs are budget items and not that well supported in FreeBSD.

Steve

nik.taylor

@stephenw10 - 5 days and no crash. I think the NIC driver patch fixed it. Thanks for all your help. I think I learned my lesson, Intel NIC's from here on out.