I think I remember something now....

I added svscan.sh to /etc (stupid, I think for more than one reason) which =
is symlinked to /usr which is in critical_filesystems but not in before_net=
.  If svscan.sh got rcordered before the network started (and before the =
critical_filesystems got mounted, wouldn't the fact that it existed in an =
umounted filesystem cause this kind of disaster?

I am going to change it, but thought I would throw this out.  It's an =
awfully easy way to bork a system, if it's true.

Ian Harding
Programmer/Analyst II
Tacoma-Pierce County Health Department
iharding@tpchd.org
(253) 798-3549

>>> Frederick Bruckman <fredb@immanent.net> 01/11/03 10:09AM >>>
On Sat, 11 Jan 2003, Ian Harding wrote:

> I installed a second hard drive into a perfectly good NetBSD 1.5.2
> system.  I thought I would be clever and install 1.6 on the second
> drive, leaving the first drive intact.
>
> To make a long story short, I screwed up because now I can boot
> sd0a:netbsd, but it doesn't mount /usr so it drops into single user
> mode since I have symlinked some of my config files from /usr/... to
> /etc and it can't find them.  Booting sd1a:netbsd works fine.

Try adding

    critical_filesystems_local=3D"/usr"

to /etc/rc.conf, or look in "/etc/defaults/rc.conf" for something like
that. (This is the 1.5.2 system that doesn't work now, right? I forgot
if that was the same in 1.5.2.)

> The /etc/fstab is correct. but something funny is happening in the
> boot sequence.
>
> I think the mistake I made was in the sysinst question "Do you want
> to install ... bootblocks?"

Don't think so... The bootblocks are compatible with 1.5.2, else you
wouldn't even get to single user mode.

> How can I fix this?

So what error do you get if you try, from single user mode,

    fsck -p
    mount -a

? If none, you should be able to exit the single-user shell into
multi-user.

Frederick