Date: Thu, 21 Mar 1996 11:09:47 -0700 (MST) From: Terry Lambert <terry@lambert.org> To: jgreco@brasil.moneng.mei.com (Joe Greco) Cc: davidg@root.com, phk@critter.tfs.com, terry@lambert.org, julian@ref.tfs.com, scrappy@ki.net, current@FreeBSD.org Subject: Re: DEVFS vs "regular /dev" Message-ID: <199603211809.LAA00435@phaeton.artisoft.com> In-Reply-To: <199603211618.KAA02588@brasil.moneng.mei.com> from "Joe Greco" at Mar 21, 96 10:17:54 am
next in thread | previous in thread | raw e-mail | index | archive | help
> > If we do the mount in /sbin/init and it fails for some reason (like the > > mount point doesn't exist), then we'll have no way to inform the operator > > (there isn't a /dev/console to write to). If we do it in the kernel, we can > > emit a message saying "/dev: not found" or something. > > They just got through arguing this in NetBSD's port-sun3 list (folks tend to > forget to do the MAKEDEV every once in a while). It would seem to me that a > little "special handling" for init would not be so bad, i.e. pass init file > descriptors 0, 1, and 2 already pointing at the console, so that the console > does not have to be accessible via the filesystem immediately. > > I know that there is some "ugh" associated with this, but it would seem the > most robust method to provide USEFUL error trapping and reporting. OR Provide a file system to mount as "/" that has a "/dev" directory in it as part of devfs, and mount it before mounting root (and mount root over top of the "/" portion of it using a translucent mount). Then you don't need a mount point and you don't need to special case all this crap for init... This would be an easy change for devfs itself, and the root mount changes I've already submitted allow the overlay. Terry Lambert terry@lambert.org --- Any opinions in this posting are my own and not those of my present or previous employers.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199603211809.LAA00435>