Subject: Re: savecore busted with /var/crash a symlink off of {/var,/usr}
To: None <perry@piermont.com, veego@NetBSD.ORG, cjs@NetBSD.ORG, drochner@NetBSD.ORG,>
From: Matthias Drochner <drochner@zelux6.zel.kfa-juelich.de>
List: current-users
Date: 09/19/1997 11:00:17
Excerpts from netbsd: 18-Sep-97 Re: savecore busted with /v.. John
Kohl@kolvir.arlingt (678)

> >> The current /etc/rc doesn't mount local file systems before running
> >> savecore or checking if /var/crash resolves to a directory.

"/var" is mounted before, so it is not completely broken, at least
for standard setups.
The idea was to run "savecore" early, before the swap space
might have been overwritten by eg BIND/YP, which can in turn
be necessary to mount local file systems.

> this stuff is hairy enough and other folks have been trying to
> sort it out

Oh yes, it is guaranteed to create 3 new problems if you solve one.
A linear /etc/rc is quite inflexible, and it can only take into account
more or less conventional setups.
(I'm not an /etc/rc.d/Snn* fan, using arbitrary numbers to choose
order of actions is quite unnatural for me. Something inspired by
"lorder | tsort" or "make" would be nicer... But there have been
many proposals and discussions before.)

To get the problem solved: Should we really add a new rc.conf
variable like "savecore_after_localmount", or is it acceptable to
run "savecore" from /etc/rc.local in such cases?
(next potential problem: if you want to have "syslog" log into
another filesystem...)

best regards
Matthias