Kernel crash before dumpon
Lowell Gilbert
freebsd-questions-local at be-well.ilk.org
Tue Nov 25 12:02:32 PST 2008
Frank Solensky <frank at solensky.org> writes:
Wow. Frank Solensky. Long time no see.
> I'm trying to get a dump off a machine with a 7.1-beta2 kernel that's
> been crashing during the boot process, following the instructions on
> http://www.freebsd.org/doc/en/books/developers-handbook/kerneldebug.html#KERNELDEBUG-OBTAIN
> I've recompiled the kernel with "-g" but no vmcore file appears so I'm
> assuming that the crash occurs before dumpon is executed.
>
> The page includes the following suggestion on how I might be able to
> proceed:
> Alternatively, the dump device can be hard-coded via the
> dump clause in the config(5) line of a kernel configuration
> file. This approach is deprecated and should be used only
> if a kernel is crashing before dumpon(8) can be executed.
>
> I tried adding
> config dump "/dev/ad4s3b"
> to the configuration file but that option appears to be no longer
> supported: the config command gives an error message of:
> root/dump/swap specification obsolete
>
> Is the paragraph above obsolete? If so, what's the preferred way to
> collect the dump?
Yep, it looks like the config(8) code to handle that has been gone for a
while. At a quick glance, I can't figure out where the dump device is
chosen, but it's supposedly iterating through devices looking for
something that would work. Sticking in a device closer to the top of
the search order might help. I suppose it's possible that just sticking
in an appropriately formatted USB disk might help.
Enabling minidumps would let you get away with a smaller space for
storing the dump, which would be useful if you have some space to throw
at it. Alternatively, in the same spot I would be tempted to build a
separate disk just for debugging this particular problem, making sure to
leave space for a swap partition close to the front.
Using DDB might be an option, but I suspect that if you're having
trouble getting a dump, you'll have problems dropping to a live debugger
as well.
All of my above advice is a bit shot-in-the-dark; if no one else
suggests anything better, you may want to go to the freebsd-hackers
list, or look at the cvs logs for whoever's modified the dump code in
the last year or so.
Good luck.
--
Lowell Gilbert, embedded/networking software engineer, Boston area
http://be-well.ilk.org/~lowell/
More information about the freebsd-questions
mailing list