Bootstrap Idea

Bootstrap Idea

[A.F.R. Bain]

On Fri, 12 Sep 1997, Warren Toomey wrote:

> All,
> 	I had an idea about bootstrapping images into PDP-11s, please shoot
> it down! Ok, I don't know much about the -11 hardware, how hard would it
> be to bootstrap as follows:
> 
> 	hand-toggle in a small bit of code, which
> 	sucks in a bigger bootstrap over a serial line, which
> 	then can pull in a disk image over the serial line & write to disk

Well I've started to use this approach allready and have written most of
the code.  My scenario is that I have an 11/34 with one RK05, and no other
operating systems.  I have access to almost unlimited number of PDP-11
serial cards, and other unix boxes.

The PDP 11 is configured with two serial ports, one the standard serial
console, the other a 9600 baud serial line.  These go to the two serial
ports on my PC running (which incendtally runs linux).  First I send a
program (binary) as console emulator instructions to the PDP on the
console line.  This program is basically a hacked version of 
/mdec/tboot.s (TU10 boot) which provides getc() and putc() and other bits
and bobs.  It reads in from console a number (rather than a file name)
which is the length of the program to load.  This is read in from the
second serial port and loaded into memory.  I then jump to this and start 
going.  

So the procedure is (1) Load tboot.s via console (2) load RKF to format
RK-05 (this works fine) (2) load the copy program whose name escapes me at
the moment, and transfer the tape image to the RK-05.  (3) Boot the RK-05.

This seemed to work, and progress was only interupted by the need to move
the PDP-11 from Cambridge to home!  I have copies of all the programs
which need a little finishing off, but I could let people have copies if
they are interested.  Incidentally I assembled the on Supnik's emulator
and then punched them out to the virtual PTP.

In fact I also have a longish document which goes through the entire
process with all the code (and some new comments) to try and explain how
it all works.

All this was done using V5 unix code btw.

>
 > Flaws:	need different bootstraps for different disks 
> 	need different bootstraps for different serial hardware
> 	how to deal with bad blocks?
> 	very slooow
> 
> Other problems: 5th, 6th Edition came as RK05 images. We could probably
> 	build images for different drives.
> 	7th Edition did a mkfs during installation, but I don't know if
> 	bad blocks were ever dealt with.
> 
> Anyway, this solution would allow a simple program + disk images to be
> put on your nearby PC running Linux/whatever, so no tapes or tape drives
> would be required.

It is slow -- but that doesn't matter if you can go somewhere else while
it happens (as long as your RK-05 doesn't catch fire)

Alan


Received: (from major at localhost)
	by minnie.cs.adfa.oz.au (8.8.5/8.8.5) id NAA04307
	for pups-liszt; Sat, 13 Sep 1997 13:07:30 +1000 (EST)
X-Authentication-Warning: minnie.cs.adfa.oz.au: major set sender to owner-pups at minnie.cs.adfa.oz.au using -f
Received: from henry.cs.adfa.oz.au (henry.cs.adfa.oz.au [131.236.21.158])
	by minnie.cs.adfa.oz.au (8.8.5/8.8.5) with ESMTP id NAA04302
	for <pups at minnie.cs.adfa.oz.au>; Sat, 13 Sep 1997 13:07:25 +1000 (EST)
Received: (from wkt at localhost) by henry.cs.adfa.oz.au (8.7.5/8.7.3) id NAA24922 for pups at minnie.cs.adfa.oz.au; Sat, 13 Sep 1997 13:07:43 +1000 (EST)
Received: from csadfa.cs.adfa.oz.au (csadfa.cs.adfa.oz.au [131.236.20.6]) by henry.cs.adfa.oz.au (8.7.5/8.7.3) with ESMTP id TAA13378 for <wkt at henry.cs.adfa.oz.au>; Fri, 12 Sep 1997 19:19:00 +1000 (EST)
Received: from oxmail4.ox.ac.uk (oxmail4.ox.ac.uk [163.1.2.33])
	by csadfa.cs.adfa.oz.au (8.8.5/8.8.5) with SMTP id TAA18733
	for <wkt at cs.adfa.oz.au>; Fri, 12 Sep 1997 19:16:26 +1000 (EST)
Received: from europa.ox.ac.uk (actually host max18.public.ox.ac.uk) 
          by oxmail4 with SMTP (PP); Fri, 12 Sep 1997 10:18:32 +0100
Message-Id: <m0x9RhP-0003d7C at europa.ox.ac.uk>
From: rjm@swift.eng.ox.ac.uk (Bob)
Subject: Re: Bootstrap Idea
To: wkt at cs.adfa.oz.au
Date: Fri, 12 Sep 1997 10:07:35 +0100 (BST)
In-Reply-To: <199709112331.JAA09209 at henry.cs.adfa.oz.au> from "Warren Toomey" at Sep 12, 97 09:31:22 am
X-Mailer: ELM [version 2.4 PL23]
Content-Type: text
Sender: owner-pups at minnie.cs.adfa.oz.au
Precedence: bulk

>All,
>	I had an idea about bootstrapping images into PDP-11s, please shoot
>it down! Ok, I don't know much about the -11 hardware, how hard would it
>be to bootstrap as follows:
>
>	hand-toggle in a small bit of code, which
>	sucks in a bigger bootstrap over a serial line, which
>	then can pull in a disk image over the serial line & write to disk


Well, for what it's worth here's my contribution. I have been able to
set up working v5 and v6 systems on my 11/34c (which uses an RK05
drive) uisng the distributed images (thanks Warren). 

The key here I think is to make use of the excellent work that has
been already done on the various PDP-11 emulators. The emulator I use
is that of Bob Supnic, which runs wonderfully on my Linux box.

v6 is available as an image of a tape. The docs helpfully say:

This is a copy of the Sixth Edition distribution tape which was sent
to me by Ken Wellsch. The file v6.tape.gz is the tape image, with the
first 100 512-byte tape blocks containing tape bootstrap stuff. Blocks
100 - 4099 are the RK05 root image, blocks 4100 - 8099 are the /usr
RK05 image, and the blocks 8100 - 12099 are the /doc RK05 image.

It is trivial with the UNIX command dd to split the tape image into
its constituent parts:

	1) Tape bootstap (useless for me)
	2) RK05 root image
	3) RK05 /usr image
	4) RK05 /doc image

I can then start the Supnic emulator, attach the three RK05 images to
drives 0, 1 and 2 and boot. My system has only a single RK05 drive. As
a result I had to mess around in the emulator to create a fresh RK05
image containing a useful subset of the root and /usr images. In
practice it's not too hard to get v6 onto a single RK05 pack -- I
think it's only necessary to lose stuff like the spell disctionary and
so on.

I also wanted support for a second DL11 serial line, so I used the
emulator (and some extra emulated disk space) to rebuild the kernel.

Once I was happy with the disk image I'd created, it was time to
transfer it to real hardware. My 11/34c runs RT11 and I was able to
get hold of a small stripped down Kermit server-only program from John
Wilson. I forget the name of this utility, but it's available from
ftp.dbit.com. The stardard RT11 Kermit that I have is unable to
transfer entire disk images, the Wilson implementation can do so.

Booting the 11/34c from RX01 floppy (one RK05 drive only, remember) I
ran the Kermit server. From the PC end it was straightforward to
transfer the image: "PUT imagefile.img RK01:". Aside from a
spectacular RK05 head crash that put my machine out of action for a
while, all was well.

For v7, I started with a single RL02 image. That would boot on the
simulator. I added a couple of emulated RK05s to the emulator set up
and proceded to build a single RK05 which would (just!) hold a
bootable v7 image. I warn you that v7 on a 2.5Mb disk is tight, but it
can be done. Having built the image, I rebuilt the kernel to enable
RK05 support (and a second serial line, again) and disable the other
disk drivers. That kernel went onto my new image. Again, the same
Kermit trick enabled the image to be transferred to the 11/34a.

If anyone wants these RK05 bootable image files, please let me know
(rjm at swift.eng.ox.ac.uk - not this address)!

A few points though:

1) RK05s packs _can not have_ bad blocks. There is no bad block
revectoring on these drives. A bad block on a pack suggests the pack
is ready for the dumpster. This makes life a little easier. I'm not
sure how you'd deal with bad block revectoring on an RL01/RL02 for
example. I guess it's not too hard, but I shan't speculate there.

2) My 11/73, on the other hand, uses an RD53 (last time I looked). I'm
not sure if the emulator can deal with these kind of drives. In any
case, support for MSCP drives only came with 2BSD, which I've not
played with. Maybe the same emulator tricks can be employed to get
RDxx images, modulo the bad blocks problem.

3) If you don't have a PDP-11 operating system running, the Kermit
approach won't be much use. There may be a stand-alone file transfer
program that can write to raw devices. No idea -- perhaps someone
needs to write this (both for the PDP and the PC ends).

Hope this helps.

Cheers,
Bob

-- 
------------------------------------------------------------------------------
Bob Manners                       (My REAL address is: rjm at swift.eng.ox.ac.uk)
BOB'S COMPUTER MUSEUM:               http://swift.eng.ox.ac.uk/rjm/museum.html
------------------------------------------------------------------------------

Bootstrap Idea

[Allison J Parent]

Hi,

<Allison, you got a copy? I emailed Dan Ts'o to get a copy, but haven't hear
<back from him yet.

I got it off the net in one of the archive sites. I have the sources (in C)
for it.

<The early Unixes didn't have a concept of bad blocks. This will always
<be a pain.

Not true.  They all did as bad blocks have been a fact of life for all 
computers since day one.  Some of the ealy unixes used crude methods
from a perfomance standoint but, the bad block replacement was there.

Do read LIONS commentary.  I was able to get a copy from the local library
here in eastern MA (USA) along with several books on BSD design.  Unix was 
really ahead of the pack on many things. 

Allison