Subject: Re: Savecore Complexity Questions
To: Manuel Bouyer <bouyer@antioche.eu.org>
From: Thor Lancelot Simon <tls@rek.tjls.com>
List: tech-kern
Date: 12/10/2007 13:25:08
On Mon, Dec 10, 2007 at 07:09:29PM +0100, Manuel Bouyer wrote:
>
> savecore tries to extract the panic string to syslog it. I think it's
> valuable to keep this functionality.

The thing is, this is trivial to do with a "dense" on-disk core format
because you could, practically, just write the core out with dd and you'd
get the same thing -- kvm can operate just as it would on a kernel in
memory or a core saved to the filesystem.

With any core format that isn't what will be saved for analysis with
libkvm tools or gdb, this means teaching libkvm how to directly look up
symbols in the on-disk sparse core format, and making savecore even more
complex so it can use the new libkvm functionality.  I for one don't think
it's worth it.

How about multiple savecore backends, as we have for fsck and newfs, where
those for sparse formats might simply not return a panic string even if
asked?

Thor