On Thu, 28 Oct 1999, NetBSD Bob wrote:

> Is there some combination of any of the earlier relases or bits from
> here and there that will allow booting and installing from a TK50?
> I am willing to try to roll a few shots in the dark, if I can do
> it from Ultrix.

Maybe... with some modifications.  On further consideration, it's most
likely that something is just hung up with the bootloader.  We got the
1.4.1 bootloader to work on non-MSCP SCSI and MFM disks by forcing it to
use the builtin rom() routines.  I don't know that we did the same for the
straight SCSI tape devices.


> Pardon me VAX ignorance.... what is the TK50Z that I have, relative to
> TMSCP?  

Basically, it's not a TMSCP device.  A TK50 drive hooked up to one of the
Q-bus TMSCP controllers will appear as a TMSCP device on the system.  A
TK50Z is a TK50 drive which also has a logic board to allow the TK50 to
communicate as a SCSI tape drive.  And we're still working out the bugs
with the whole SCSI thing.

> I was thinking that was one of the few scsi tapes that was supported,
> although when I last tried tapebooting a VAX, it was on a 3500 with a
> TK70, if memory serves me correctly (about 6 months back).

That would've been a TK70 connected to a Q-bus TMSCP controller board.
They do work.

> Well, for the sake of discussion, since the beast has 5 hard drives up
> under Ultrix, what would it take to build it up from Ultrix?  I could
> dd boot.fs to the c partition on one of the 5 drives that I could
> unmount from Ultrix, then boot that.  Would that require 2 drives, one
> to dd to and boot from, and a second to install onto, or can that be
> done from one drive? Is everything there in the boot.fs to complete
> the install?

In theory you can do this with just one drive, but in practice it's
been sort of tricky.  Basically, you dd boot.fs to the c partition on the
first SCSI drive, and you install the system using sysinst onto the second
drive like you've guessed above.  The following notes are applicable to
the 1.4.1 release.

The boot.fs is a filesystem image which contains /boot and the /netbsd.ram
image in some form, I forget whether it's gziped or not.  Once NetBSD
boots up and sysinst is running, you should be able to perform an install
over the SCSI drive you just booted from.  However, when you try to
re-disklabel the drive, the disklabel program only ever sees the 2000
blocks from the boot.fs image you dumped to the drive.  Only having a
1000KB partion on you're nifty 1GB SCSI drive is sort of disappointing.

But just a few days ago I found that using "dd if=/dev/zero of=/dev/rsd0c
count=2000 bs=512" would effectively nuke the stale partition info.  I've
not actually tried a single disk boot-and-build like this, but I think
it'll work. Or at least it should as long as dd is present on the ramdisk
inside in the netbsd.ram kernel.  I don't recall if it is or not.  
Another option might to be "cat /dev/zero > /dev/rsd0c" and then crtl-C it
after a bit.  Once you're booted into sysinst you would exit out of it,
nuke the boot.fs based partition on your boot disk, and then rerun
sysinst.

The down side of the one disk approach is that you'd better get the
install right on the first try.  If not then you'll be starting from the
beginning, having to boot Ultrix and dd'ing boot.fs onto the disk again.

As for the questions about seeing Ultrix partitions from NetBSD or NetBSD
partitions from Ultrix -- I've no idea.  Since they've got the same BSD
lineage, it's plausible.  I've never tried though, and have to means to
investigate it.

-brian.
--- Brian Chase | bdc@world.std.com | http://world.std.com/~bdc/ -----
I've said it before, and I'll say it again.  You can diagnose people 
based on where they sit on the bus.  -- K.