Subject: Re: NetBSD/pmax installation
To: Gregory McGarry <gmcgarry@lux.sprc.qut.edu.au>
From: Jonathan Stone <jonathan@DSG.Stanford.EDU>
List: port-pmax
Date: 07/12/1997 04:37:23
Gregory McGarry <gmcgarry@lux.sprc.qut.edu.au> writes:


>Well, my NetBSD/pmax installation onto the DS5000/25 continues.
>Last weekend Jonathan solved my tftpboot problems, and thanks to
>David Conran I have myself a supported graphics and monitor setup.

>There were some messages on the list earlier in the week about getting
>the tftpboot working with the KN03-AA V5.1b prom?  Was this sorted out
>or is there still numerous machines with problems?  I ask this because
>this problem is probably more important than mine!

To me, this looks like a PROM problem.  The PROM is dying with a bad
access even beofre it gets to ask for its adress via BOOTP.  There is
no way I can fix that.  (Well, I could, by disassembling the PROM and
fixing it, but I am not going to try. Life's too short.) 

So the only way around this problem is to upgrade your PROM, or to use
MOP for diskless booting.  Sorry for not making that plainer.

Digital UNIX and Ultrix support MOP boot of DECstations.  We have a
mopd in the tree, but it only supports the standard PDP-11-like MOP
boot formats, and sending a.out files to vaxes.  (i.e,. the vax a.out
support does *NOT* work for sending a.out-format kernels to pmaxes).

If someone wants to take the existing mopd code and add ELF or ecoff
support , that would be great.  I won't have time to do it myself
before 1.3.

>Jonathan expressed interest in improving the NetBSD/pmax installation
>procedure.  I would tend to agree, given that the first contact many
>have with the OS is when installing it.

Yes, given the recent discussion on comp.sys.ultrix, I think so too.

But there is already a diskimage and tarball of the diskimage.  This
is functionally identical to the OpenBSD `simpleroot'; a 32 Mbyte
filesystem image with the normal contents of a root partition.  DD
that onto a raw disk, boot it, mount the installation tarballs from
somewhere, and go.

Writing that up on the installation-notes Webpage would be a good
start.  

I'd also like to fix a to miniroot that you can drop onto a hard disk
and then run out of RAM, so people who really need to can tailor the
size of their root partition. But for now, either netbooting a kernel
(other than using the LK501 on a Personal DECstation as a console!),
or clobbering an existing root partition and part of swap with the
diskimage is the best we have.


>Tohru Nishimura wrote:

>> I learned MAXINE can not complete netboot procedure with video
>> console.  (Jonathan already has some guess of the cause).  So, you
>> have to use serial console hooked up MAXINE serial port for TFTPboot.

>I can verify this.  I don't have a serial console either.  Is this the
>next logical problem in the NetBSD/pmax installation procedure?

Er... 

Netbooting and asking for a root filesystem has worked fine for me on
a 5000/150, a 3100, and a 5000/200.  This is only a problem for
Personal DECstations.  AFAIK it's a problem which is specific to the
``desktop bus'' keyboard.

It sounds like a simple interrupt or autoconfiguration code.  I have
found figuring those out from desk-checking code is very
time-consuming, especially wiht hardware like the Desktop Bus [sic]
for which I have no documentation other than the driver source code.

I have no idea if I will get time to figure it out before 1.3 is out.
I doubt it. It would be much easier if I had physical access to a
Personal DECstation.  I don't....


>> Try "boot 3/tftp", and see what will happen on the serial console.
>> you will eventually get;
>> 
>> boot device: <unknown>
>> root device:
>> 
>> on display, and your serial input is available on the root device
>> selection prompt.

>I'm looking for a serial connection to the DS, however, I'd prefer a
>solution to the video console problem.

But,  AFAIK, once you get the system installed there *is* no problem.

I think Toru-san said as much in an earlier message also.  Once you've
installed and can boot a kernel that doesn't have to ask where the
root filesystem is, the framebuffer console code should Just Work.



>Is Simon Burge going to be the official Webpage maintainer then?

That's fine with me, but it will take some time to set up Simon with
access to the NetBSD Foundation's server where the pages are.
Meanwhile I'm happy to to paste in Simon's list, if it's finished, or
to commit updated HTML source code.