Subject: Re: /kern/kernel
To: NetBSD-current Discussion List <current-users@netbsd.org>
From: Gandhi woulda smacked you <greywolf@starwolf.com>
List: current-users
Date: 09/16/1998 00:52:24
On Mon, 14 Sep 1998, Greg A. Woods wrote:

 * [ On Mon, September 14, 1998 at 08:23:55 (-0400), der Mouse wrote: ]
 * > Subject: Re: /kern/kernel
 * >
 * > The problem is, / is mounted read-only.  To remount it read-write,
 * > userland does "mount -o update /dev/whatever /".  This doesn't work,
 * > because when the kernel looks up / to get the mounted-on vnode, it gets
 * > a different vnode from the one the root device is currently mounted on.
 * ...
 * 
 * I believe this is why SCO Xenix (and possibly other implementations)
 * have a /dev/root and /dev/rroot.  These are virtual devices that always
 * point at the "real" one -- i.e. the one set at boot time either because
 * that's where the kernel was loaded from, or that's where the user
 * specified for the equivalent of a "boot -a".  That way "mount -o
 * update,rw /dev/root /" is always correct.  Although if I remember right
 * finding out what the real root disk is from user-land is left as an
 * exercise to the user.....

/dev/root and /dev/rroot were originally hard-links to the "real"
devices under SVRx (0 <= x < 4).  How they got there, I'm not sure,
but ls -l showed them to have the exact same major and minor numbers
as the disk and partition (if possible) from whence they were booted.

SCO's implementation must be much more bloa\aolb/slick than that.

 * 
 * 

				--*greywolf;
--
Friends don't let friends use Microsoft.