Laptop troubles...

[Adam K Kirchhoff]

Nate Lawson wrote:

> Adam K Kirchhoff wrote:
>
>> Nate Lawson wrote:
>>
>>> What happens if you boot single-user?
>>
>>
>> So far, it's been fine.  I've been in single user long enough to run 
>> fsck on my filesystems...  I launched sshd manually once (without any 
>> problems) since that's where it appeared to lockup most frequently.  
>> When I then continued into multi-user mode, it locked up around the 
>> time it loaded the linux compat module.
>
>
> It's important to isolate this more.  Try running some of the rc.d 
> scripts to see if you can trigger it.  I think there's an rc.d debug 
> mode that prints everything before it does it also.  Then just enable 
> that feature and send me the last few lines it prints before hanging.
>
>>>
>>> You can't break to the debugger with ctrl-alt-backspace? 
>>
>>
>> Nope. Didn't realize that was usually a possibility, but when I tried 
>> it just now, it didn't break into the debugger.
>
>
> Well, you have to have options BREAK_TO_DEBUGGER in the kernel for 
> this to work.  If you have it but this doesn't work, then it's a 
> harder hang.  If not, we can get debug info as to what's hung.
>
>>> Try a boot -v -s with the bad kernel and see how long it takes to 
>>> hang while sitting there idle.
>>>
>> Well, so far so good.  I've booted up with -s and -v...  fsck'ed the 
>> filesystems, and I'm now sitting at the single user prompt.  I 
>> ifconfiged my wireless card, and all seems well.  I'm about to leave 
>> work for the day.  I'll leave it like this during my drive home and 
>> see if it stays running the entire time.
>>
>> When I'm home, I'll reboot with -v -s and dump the kernel output to 
>> the serial port, and then post it here.
>
>
> The -v is just to get more info from right before the hang.  Try doing 
> things like sysctl -a, kldload linux, or whatever to see if you can 
> isolate what's triggering this.
>

Woohoo...  It's /etc/rc.d/devd:
# ./cron start
Starting cron.
# ./devd start
Starting devd.
hw.acpi.cpu.cx_lowest: C1 -> C3
hw.acpi.cpu.throttle_state: 8 -> 8

And then, immediately, the lockup.  Want me to try adding the 
BREAK_TO_DEBUGGER option in the kernel and see if I can get a backtrace?

Adam