Excerpts from netbsd: 31-Jul-98 Re: Disk partitioning (agai.. Zdenek
Salvet@ics.muni.c (2109*)

> > I _think_ what happened is that I lost track of the original
> > translated geometry, and I partitioned the disk with the wrong translated
> > geometry, and the C/H/S partition labels don't point to the right
> > spot

> Maybe, I think this is not critical for NetBSD boot blocks.

That's not completely correct. The MBR code will load
the NetBSD bootblock using the CHS information in the
Partition table. If this doesn't comply to the geometry
used by the BIOS at this point, it will load a random block
which of course does not have the 55aa signature, causing
the "no operating system" message.
The NetBSD bootblocks are not better in this respect, and
I thought of fixing this several times... but without fixed MBR
it would be worthless.

> > - Is there a way to determine the "correct" translated geometry
> >   associated with an IDE drive?

The DOS "pfdisk" is most reliable.

> > - As I understand it, the crucial limitation is that the BIOS only
> >   has enough bits for 1023 cylinders, so the other parameters have
> >   to be made larger so the cylinders can be made smaller.

Either so, or make sure your 'a' partition is within the
first 1023 cylinders.

best regards
Matthias