Re: more on proff crash building native

[Brad Parker <brad%heeltoe.com@localhost>]

Johnny Billquist wrote:
>Brad Parker wrote:
>> "David Brownlee" wrote:
>>> Whats next on the list, bootblocks? :)
>> 
>> sort of :-)  full install from tape.  seems very broken right now for 11/7xx
>.
>
>Considering the fact that the 11/7xx machines never supported booting 
>from tape in the first place, that's no surprise. :-)

well ok, I didn't mean booting per se.   I ment installing from tape.
sorry for the confusion.

>That is, VMB can't even boot from tape. There is no way to get the 
>machine to even try booting from a tape.

Yes, I understand.   I think I've previously talked about "boot.bin"
being on the 730 boot tape (tu58) or on the 780 boot floppy.  Or,
I assume the 750 console tape (also tu58).

I'm assuming you have console media with 'boot' and 'copy' on it.  and
I'm just focusing on 11/7xx's right now.

>But I didn't know we had problems with the bootblocks. What kind of 
>problems are we talking about here?

Not a problem with boot blocks.  A coordination problem getting
the miniroot in place.

I can't find a recipe for installing netbsd on an 11/7xx from a tape.
Nor can I find a way to make a suitable tape image.  I don't think you
can do it anymore.  I'll describe one solution before I describe the
problem.

My current thinking is that I will do the following to rectify:

- write a simple program which can be loaded from a tape by 'boot'
which will create a reasonable disk label.

- create a tape image which has the following (separated by tape marks)

 simple-partition-program (elf)
 ffs-miniroot-filesystem
 tar of component to install
 tar of component to install
 ...

Why?

Everything in the old days assumed you could copy the miniroot to the
swap partition.  And to make that true the "copy" program had a
predefined "default" paritition map.  By agreement, all of the "disktab"
partitions and the default partition map in "copy" have the swap
partition (partition B) starting at the same place.  That way running
the miniroot won't get trashed when you repartition.

In the old days, no matter what disk you had, you could use "copy" to
copy the miniroot to swap and then repartition the disk and copy to it
safely.

But the old partition maps and the default map on in "copy" are now much
too small.  So I think we need some way to make a default partition map
which has an "a" partition big enough for an install.

Since I'm only worried about mscp disks, it's a simple job.  I'll just
creating a map with a small swap at the end of the disk and big "a"
partition.  You could argue about that, but at least it will work.  I'll
include a way to override that.

With that, I can first create the map, then copy the miniroot, then
run the normal install program.

As an aside, I originally found the generic kernel would not run on a
780.  Right now, after fixes, I'm seeing the kernel will run from the
miniroot but it does not see the 'ra' disk during the unibus probe (I'm
debugging that - it may be that 'boot' leaves ra in a some odd state or
who knows, something else).  I'm currently debugging on netbsd-3.

I've also had problems getting some of the sa utilties to run on a 780.
bitrot I think. no doubt they work fine on a 4000.

-brad