Brad Salai wrote:
> 
> Last night, I moved an external drive into my SE/30 mail server. I ran into
> some problems that were mostly related to lack of sleep, but some might
> bear looking into.

yeah, lack of sleep will definitely get you into trouble when making major
changes :-)

> I didn't change /etc/fstab before trying to boot, and since I had changed
> the scsi id from 5 to 0 the booter failed. It actually quits rather than
> helping, but if you watch it, it tells you what the problem is before it
> quits. You just have to be vigilant. this is 1.9.5 I think, newer booters
> may be more helpful.

Hmmmm...I assume that you forgot to change the root SCSI ID in the Booter
dialog box?  Every booter I've ever used and screwed this one up gives me
a message like "can't find kernel netbsd".  This is usually the clue that
I've specified the kernel location incorrectly.  The Booter doesn't read
/etc/fstab at all to the best of my knowledge.

> When I tried to modify /etc/fstab I realized that I had to mount the volume
> r/w, duh... Anyway, if you havent manually mounted anything in a while, and
> mistakenly try
> mount -uw /dev/sd0a

Where did you do this?  The Installer?  If so, that's probably easy to
fix, we just need to give a usage message on failure.

> it fails with a very non helpful message about the device being unknown. It
> would be _much_ more helpful if it mentioned that I had forgotten to
> include a mount point.

I assume that this was from within NetBSD, then.  Kind of a strange error
message I'll admit.  Perhaps you can send in a PR asking for a usage
message as well as the error message?

> I finally got it, changed the fstab, and guess what, still no good.
> 
> fsck -? /dev/sd1a is hard wired into /etc/rc.  Is there a reason for this?

The only reason that I can think of is that _you_ put it there in order to
get around onetime problems with fsck in the rc scripts.  I'm pretty sure
that NetBSD doesn't ship this way; if it did, it would probably break on
most people's setups.  The correct way is to have an 'fsck -p', I believe,
although it might still be broken and require an 'fsck -p /' (although I
doubt it since mine doesn't require this and I'm running fairly recent
-current).

> Once I fixed this, the only remaining thing is that even though fstab says
> that /dev/sd0a is a ufs volume, dmesg says it is ffs. I don't know if this
> is or will cause problems, but I'd be more comfortable if it were
> consistant.

You need to change /etc/fstab to use 'ffs' instead of 'ufs'.  The 'ufs'
name is the old one and its use is somewhat deprecated.  I think that
recent Installers create a proper fstab, but I'm not too sure.

> All in all this is much more fun than in MacOS where you just hook em up
> and they work. Who would want that ;)

Ain't it, tho?  ;-)

I hope this helps.

Later.

-- 
Colin Wood                                 cwood@ichips.intel.com
Component Design Engineer - MD6                 Intel Corporation
-----------------------------------------------------------------
I speak only on my own behalf, not for my employer.