> To: port-hp300@NetBSD.ORG
> Cc: thorpej@nas.nasa.gov
> Subject: SUCCESS!!! (Was Re: Still no luck booting on HP-340...)
> Date: Tue, 25 Jul 1995 13:03:08 -0600
> From: Greg Oster <oster@cs.usask.ca>
> Sender: owner-port-hp300@NetBSD.ORG
> 
> ...
> The solution: We added "-DSLOWSCSI" to the Makefile in 
> /usr/src/sys/arch/hp300/stand
> ...

See if your hp300/stand/scsivar.h has a 7 byte sc_msg field.  If so, bummer!
I had hoped that the SLOWSCSI hack was no longer needed.  Here is from our
RCS log:
----
revision 1.2
date: 1992/08/12 06:45:29;  author: mike;  state: Exp;  lines: +1 -1
increase MESG_IN/OUT field to 7 bytes.  Most drives only return 1 bytes
but some (MO) return more.  If the MESG_IN phase code fails to read all
the pending bytes, it throws the controller out of sync leading to a
variety of odd behaviors:
        boot program failures on warm reboot (ixfer_start failures)
        read requests that silently do nothing (VJ's old printf in sd.c)
this may also be the cause of the occasional failures of the boot
program on 68040s though I haven't tested it yet.