From: Paul Bickford <bickford@jimmy.harvard.edu>
> >* Get yourself the file NetBSD/arch/pmax/32M-diskimage.gz
> 
> >* dd that onto the disk you want to boot your 3100 first. It is very much 
> >  easier if you have two disks, but let's boot the machine first :)
> >  - Connect the boot harddisk to your i386 and find out its device.

This *should* be a simple method for installing, but it doesn't work
(as I just found out).  Details below...

> It is known as "sd1". It is very likely that the disklabel is bogus (see
> previous posts).

> So, I re-labeled the disk so that the "a" partition spans the entire disk,
> removed all the other partitions, and lastly "newfs /dev/sd1a".
> 
> Then, I:
> 	dd if=32M-diskimage of=/dev/sd1a
> 
> and get no errors.

The operation is "take the diskimage and write it to the disk".
On NetBSD/i386, the whole-disk partition should be /dev/rsd1d,
on Ultrix or NetBSD/pmax it's /dev/rrz1c,
on the other NetBSD ports it should be /dev/rsd1c.

Unfortunately, "dd if=32M-diskimage of=/dev/rrz1c" fails (at
least on my NetBSD/pmax machine) with "Read-only file system".
While using /dev/rz1c *seems* to work, it seems like it didn't
actually write the partition table/label/bootblocks to the disk.

The read-only file system error is because of disklabel protection.
I would guess that's also the reason for the silent failure when
using the /dev/rz1c device as well.  To actually get the bootblocks
*written* to disk I had to write a C program using the DIOCWLABEL
ioctl, and /dev/rrz1c.

> So, I then take the sd1 drive and install it in the DS3100 (after changing
> its SCSI ID to zero).

> boot -f rz(0,0,0)netbsd
> 
> which returns the error:
> ?417 btblk fmt
> ?470 bt err: rz(0,0,0)netbsd

yeah - I guess this is the same: you didn't actually manage
to get the bootblocks written to the disk.

This is bad.  The silent failure to actually write stuff to the
disk when using /dev/rz1c is *very* bad.  That it's so hard to
write a diskimage from NetBSD generally isn't very good either.

(Note that since I've only tested on NetBSD/pmax I may be on
the wrong track, but it *sounds* like just the same type
of problem that I had).

  -  Arne H. J.