Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 14 Jun 2001 13:00:22 -0700 (PDT)
From:      Gordon Tetlow <gordont@bluemtn.net>
To:        Andrew Hesford <ajh3@usrlib.org>
Cc:        Warner Losh <imp@harmony.village.org>, Doug Barton <DougB@DougBarton.net>, Cyrille Lefevre <clefevre@redirect.to>, Jon Parise <jon@csh.rit.edu>, Matt Dillon <dillon@earth.backplane.com>, Will Andrews <will@physics.purdue.edu>, Mark Santcroos <marks@ripe.net>, <bsddiy@163.net>, <freebsd-hackers@FreeBSD.ORG>, <eivind@FreeBSD.ORG>
Subject:   Boot Order (was Re: import NetBSD rc system)
Message-ID:  <Pine.BSF.4.33.0106141243460.94127-100000@sdmail0.sd.bmarts.com>
In-Reply-To: <20010614144245.A7853@core.usrlib.org>

next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, 14 Jun 2001, Andrew Hesford wrote:

> On Thu, Jun 14, 2001 at 11:57:18AM -0700, Gordon Tetlow wrote:
> >
> > I've also made some (slight) changes to the boot order. Mostly just
> > formalizing the lock step nature of things. I did make one change that I
> > think will be helpful, I moved pccard initialization to before the
> > mounting of /usr (but after mounting /var). This way people can use pccard
> > for mounting filesystems from /etc/fstab which I don't think you can do
> > now. Anyone have any objections to this move? it would require moving
> > pccardd and pccardc to /sbin, but that should be it. Other than that, I
> > made the (rash) assumption that /var is local storage, which simplifies
> > things greatly. That makes it diskless un-friendly, but we've always been
> > unfriendly to diskless anyway....
>
> Wow, this is the thread that won't die.
>
> I take objection to the moving of pccard before mounting /usr. It works
> just fine for those who place /var on the same partition as /, but for
> guys like me, it means you are mounting /var at a different time than
> other filesystems. I think that, for cohesiveness and sanity, all
> filesystems should be mounted at the same time, with the mount script
> simply calling `mount -a`. Otherwise, we have to wory about mount-var.sh
> and mount-rest.sh.

That's fine, it was just an idea. So... should it go:
mount root
mount local (which will have to include /var)
network init (things like dhclient which need /var)
remote filesystems

A little bit of a chicken and an egg problem. I'm trying to design these
scripts with the notion of being as versatile as possible, I don't want
the diskless folks to have to jump through lots of hops to get their stuff
working, and if one set of people are going to be using dhclient, it's
probably going to be them.

> If we want to mount pccard stuff from fstab, we should do something like
> the smbfs script does. Then we can have a mount-local.sh and
> mount-pccard.sh script, which is a bit more sane.

Not really as it's not a particular filesystem type that can't be mounted
(like smbfs) but a block device that's just missing. I suppose with DEVFS,
you could do some magic in mount, but if a filesystem is missing (and
mount doesn't know why), you are going to get an error on boot and drop to
single user. I could see a new field in /etc/fstab (like noauto) called
discon which will mount a drive if it is there and not complain (maybe
just a warning "/zip/drive isn't connected") if it isn't.

Any other thoughts?

-gordon


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-hackers" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.4.33.0106141243460.94127-100000>