Subject: Re: MAKEDEV init
To: None <tech-userlevel@NetBSD.org>
From: Johnny C. Lam <jlam@pkgsrc.org>
List: tech-userlevel
Date: 06/01/2007 18:45:27
On Fri, Jun 01, 2007 at 07:08:25PM +0200, Joerg Sonnenberger wrote:
> On Fri, Jun 01, 2007 at 03:50:18PM +0000, Johnny C. Lam wrote:
> > I ask because I'd like to make it a easier to make a read-only-root
> > installation on my Xen boxes.
> 
> Unless I'm proven wrong, I'd say the following two things are needed for
> read-only root:
> - dhclient has to be changed to create resolv.conf somewhere else. It
> should also be fixed to not loop around if it can't create resolv.conf,
> but that's a separate question.
> - mount ptyfs and drop /dev/tty* (not all, but the ptys). This works for
> most software, would be nice to know what breaks to fix that to use
> openpt and friends.

I didn't mean for this thread to become something on read-only root
setups.  I have a special setup that I use[1], but I was mostly
wondering about "MAKEDEV init".  I think the changes to MAKEDEV and
init(8) so that the init process doesn't create the MFS mount, but
rather lets MAKEDEV do it basically solved the problem of the MFS
mount not being big enough.  I'm wondering why we didn't do the next
logical step, which is to nuke the custom "init" targets on that
small handful of platforms' MAKEDEV scripts.

	Cheers,

	-- Johnny Lam <jlam@pkgsrc.org>

[1] http://www.netbsd.org/~jlam/cgd_laptop.html