Skip site navigation (1)Skip section navigation (2)
Date:      Sun, 4 Jan 1998 12:39:08 +0100 (MET)
From:      j@uriah.heep.sax.de (J Wunsch)
To:        freebsd-hackers@FreeBSD.ORG
Subject:   Re: Why dump to /var??
Message-ID:  <199801041139.MAA18280@uriah.heep.sax.de>
References:  <Pine.BSF.3.96.980103200655.299A-100000@zerium.newmedia.no> <19980104102759.11459@lemis.com>

next in thread | previous in thread | raw e-mail | index | archive | help
Greg Lehey <grog@lemis.com> wrote:

> Well, one reason not to do this change is the possible symbolic link,
> which makes it possible to put /var/crash on some other file system.

You could even mount a separate /var/crash. ;-)

> The other thing is that savecore almost invariably saves to
> /var/crash, so it doesn't add much value.

Why?  /var/crash always has to be provided as an argument to savecore,
there's no reason it could not point to something else.  HOWEVER, if
you move it somewhere else, make pretty sure you install a `minfree'
file.  The default installation does this in the /var/crash case (and
savecore now finally obeyes it correctly).

>> Anyone? Any why not send this error msg to syslogd? 
> 
> It's not running at this point.

It should be running at this point, i'm going to commit this change.
The only case where it's perhaps more useful the other way 'round is a
4 MB only machine (since starting syslogd might clobber the coredump
in the swap partition then), but these presumably require a lot of
personal tweaking anyway these days, and it's not longer useful to
make the default behaviour dependant on a low-mem setup.

> Yes, there is some merit in the idea of starting syslogd earlier and
> logging *all* the startup messages.

There's nothing more except savecore attempting to log anything
earlier.  Also remember, you need a writable /var before starting
syslogd, so syslogging something like the fsck messages is virtually
impossible (with the current scheme).

I think Poul-Henning has once been playing with a true memory
filesystem, this could provide a workaround for this early stage (by
offering some pseudo-disk storage that can be reclaimed later).

As always: context diffs welcome. ;-)  You can argue a lot about what
might be better or not, the devil's always in the detail, and playing
with this part of the system will quickly turn out to be more than a
weekend-only project if you do it right.  So it's not whether changes
in this area are acceptable or not, but whether someone can be found
actually doing these (as opposed to _only_ talking about them).

-- 
cheers, J"org

joerg_wunsch@uriah.heep.sax.de -- http://www.sax.de/~joerg/ -- NIC: JW11-RIPE
Never trust an operating system you don't have sources for. ;-)



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199801041139.MAA18280>