Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 07 Feb 1997 12:22:24 -0600
From:      Gabor Kincses <gabor@acm.org>
To:        Joerg Wunsch <joerg_wunsch@uriah.heep.sax.de>
Cc:        freebsd-hackers@FreeBSD.ORG
Subject:   Re: Panic in probe, but no dump
Message-ID:  <32FB72E0.2781E494@acm.org>
References:  <32FA29A1.41C67EA6@acm.org> <Mutt.19970207005345.j@uriah.heep.sax.de>

next in thread | previous in thread | raw e-mail | index | archive | help
J Wunsch wrote:
> 
> (Mailing lists trimmed.  Whenever you feel like posting to more than
> one list, you probably feel wrong. ;-)

Desperation speeds up the fingers and slows down the gray stuff.

> As Gabor Kincses wrote:
> 
> > The problem I have is to get a dump out of the sound driver probe.  I
> > pretty much followed the handbook on this.  I have set config root on
> > wd2 dumps on wd2 in my config file, but no dump occurred (I checked .  I
> 
> Well, the normal sequence is to use `dumpon'.  This is also supported
> by /etc/sysconfig.

AFAIK that is not an option here, since we are not executing /etc/rc,
yet.

> > 1. Is there a way to load the kernel.debug in gdb and correlate
> > addresses with lines of code?  (Like on HP-UX w/ xdb: one can use 'td')
> 
> Basically, though a little hard.  Please, read the section about
> kernel debugging in the handbook (gdb -k aka. kgdb).

I did.  Before I started all this.  It seems that once I can single step
or have a dump it will print the addresses.  In xdb I can 'td' (toggle
disassembly) and voila I see the machine instructions with line numbers
next to the ones that start a line.  I'm looking for a similar feature.

> > 2. Why am I not getting a core dump?
> 
> No idea.  What did the kernel say?  Did it say ``dumping to ...''?

No, it did not.  I guess there might be a bug in handling the panic.  I
did have a dump device configured.  I will triple-check this one.  No
need to accuse the designer in case of pilot error.

> > save the kernel core, or is it already on the swap device?  Ie. what
> > macroscopic events would tell me that I got a dump?  (Possibly: "core
> > dumped" msg on the console :-)
> 
> Almost.  Dumping to dev ...., followed by the number of outstanding
> megabytes to be dumped.

That is good info, I didn't see anything like that.  The panic message
printed the current EIP, then "waiting 15 seconds to reboot" or
something like that.

> > 3. I have turned on savecore in /etc/sysconfig, but I have noticed that
> > the swapon gets executed first in /etc/rc before the savecore.  Wouldn't
> > this wipe out a core in the swap device?
> 
> No.  The dump is done to the end of the swap partition.
> 
> > 5. Is my only option left is ddb?
> 
> Your best one.  Note that you can also force a dump from within DDB by
> saying `panic' (followed by `c').

20.3. call diediedie().  I'll try that.  ddb looks fairly similar to adb
on HP-UX.  Probably share roots, too.

> > 6. Is a page fault essentially like a segmentation violation in user
> > mode?  AFAIK on i386 you have to install a page fault handler, which is
> > probably what the VM is all about.  Is this true?
> 
> Well, basically.  A page fault normally tries to see if it has to load
> a page from secondary storage.  If there's no page to load, it will be
> reported.  By a SIGSEGV to a user process, by a panic (or a jump into
> DDB) in the kernel.
> 
> > 7. Why is the "dumps on" feature being deprecated?  How else can we get
> > a dump if the system crashes before dumpon can be executed?  BTW, My
> 

Thanks for your help.  DDB here I come!

-- 
Gabor Kincses
(gabor@acm.org)



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?32FB72E0.2781E494>