Subject: Re: new kernel woes
To: Nathan Parrish <nparrish@maxine.cc.vt.edu>
From: Jonathan Stone <jonathan@DSG.Stanford.EDU>
List: port-pmax
Date: 01/29/1996 20:39:20
[[reports of problems with a 5000/20 or /25]]

Nathan,

There are several things going on here.

First, the NetBSD/pamx kernel may not be able to recognize the
devices you have on units 1 and 2 as being supported disks.
I'm not sure how to fix that; I'd need more info from
the SCSI driver.  After a few seconds perusing rz.c, I'd
guess that your rz1 and rz1 are SCSI-2 drives, but that
for some reason the rz driver isn't getting the vendor and
product ID strings from the drive.  Some SCSI-2 drives
reportedly return more data to an inquriy command. Can you
try the following patch and let me know if it helps?

*** sys/arch/pmax/dev/scsi.h	Fri Oct 13 19:16:09 1995
--- sys/arch/pmax/dev/scsi.h.dsg	Tue Oct 24 18:06:14 1995
***************
*** 380,385 ****
--- 380,386 ----
  	u_char	revLevel[4];	/* Revision level (ASCII) */
  	u_char	revData[8];	/* Revision data (ASCII) */
  #endif
+ 	u_char	pading[1024];	/* newer SCSI II drives give additional data */
  } ScsiInquiryData;
  
  /*



Second, because the kernel's root partition is on /dev/rz2, and
the rz driver doesn't recognize rz2, the kernel is almost
certainly failing to read /sbin/init.  Without /sbin/init,
there's no way the kernel will prompt for input.

Thus, the first thing you should try is, if at all possible,
to put a minimal root filesystem and bootbloks on your rz0 disk.
(just a kernel, /sbin/init and /bin/sh would be a useful experiment.)

Third, it looks like you've typed in the kernel's console messages
from booting.  That log has the wrong interrupt priority/slot number
for the desktop bus, which is where the keyboard is attached on
a 5k/25.   It would be really helpful if you could try to boot
a kernel from /dev/rz0a, with an /sbin/init in /dev/rz0a,
and see if  *that* setup at least gets a prompt for a shell.

Fourth, the clock bug is fixed (for the next two years) in more
current source. I would advise you to *not* fetch newer source,
though, until you can at least a kernel that gives you get a /bin/sh
prompt in single-user mode.

Lastly, the config.old configurations are ancient history. They just
don't work any more, and should have been removed from the tree ages
ago.  I'll do so now.