Hi Scott!

In message <8523.199610141839@mortimer.cc.ic.ac.uk>
          Scott Stevens <s.k.stevens@ic.ac.uk> wrote:

> Hmmm. Use the one bb_riscbsd gives you as this is the one (AFAIK) the
> the RiscBSD disklabel code will use.

OK, I have updated my Cumana installation notes accordingly. You've probably
already seen them on the documentation mailinglist.

Btw, I did use the bb_riscbsd values for installation according to my
disklabel, ie. my RiscBSD part starts on cylinder 2241.
So this shouldn't be the cause for my disklabel losses.

Where is the disklabel situated anyway, ie. between which parts on the disk?
Before the RiscBSD root partition?

> This is not the fault of bb_riscbsd. as far as it knows there are 2
> distinct disks in the machine.

Isn't there any way to find out? This might puzzle people. I tried to explain
this in the installation guide. It could probably still be a bit clearer though.

> I assume you have 2 riscos partitions on one disk if so then you only
> need to run bb_riscbsd once on the first partition on the disk.

Well I only need to run it once to change the filecore bootblock, but twice to
get the size of each partition. I have to sum them up to get the size of the
RiscOS part of the disk.

> bb_riscbsd obtains this figure from the cylinder offset in the
> bootblock, this address is initially calculated anf placed there by
> !SCSImgr.

OK, so bb_riscbsd should be right unless there's a bug in it.
Good to know.

> I think the heart of the problem is !SCSImgr, in that it give you a
> size in sectors along with an offset but due to the way the offset is
> stored must convert to a cylinder offset. It must round up to the
> nearest boundary but looks like it's not reporting the rounded value.

We should tell Cumana about this.
Do you want to do this, or should I do it because I have a concrete example at
hand? I will try to be polite to them despite the NDA. ;-)
Hmm, a copy of my message to this list mailed to them should almost be enough.
Just tell me if you want and I'll drop it to them.

So long, Markus