Re: Crash followed by green db> prompt

[Martin Husemann <martin%duskware.de@localhost>]

On Tue, Aug 03, 2010 at 08:37:37AM +0000, Thomas Mueller wrote:
> Running NetBSD 5.1-RC3 on i386, editing a file, the kernel configuration, 
> with vi, I had a sudden crash followed by green 
> db>

The interesting part should have been written out before the green db>
prompt - there is probably a panic: massages (midway in your vi screen).

You are now at the kernel debugger ddb(4). Most intersting things to do
are: bt (for backtrace), which should show a stack trace.

An interesting question, however, is how you ended up there at all - on a
new installation of NetBSD you probably would get a kernel crash dump and
automatic reboot, followed by savecore writing the crash dump to /var/crash
for later analyzis.

> Maybe there is some known instability in NetBSD 5.1_RC3, judging from the long
> time elapsed since 5.1_RC3 with no release or RC4?

Not that I'm aware of.

>  But my main question is
> what I can do at the post-crash db> prompt, which looks like something that
> would not be just for -current or release candidates.

You could check the output of sysctl ddb.onpanic - if this is 0 you should
get the (intended) crash dump, otherwise (id ddb is available) you will end
in the interactive debugger. This is normally done by an entry like

ddb.onpanic?=0

in /etc/sysctl.conf.

Martin