> This doc dosn't suggest to put disklabel into MBR. If you devote the whole 
> disk to NetBSD or to set of systems which understand disklabel IMO it is much 
> better to get rid of ancient MBR and use only disklabel so this would look 
> like:
> 
> CHS 0:0:1, 1 sector - primary boot loader which understand disklabel.
> CHS 0:0:2, sectors for disklabel which actually "partitions" the disk (I don't

If you do that, then some system will not boot the disk.
For the sake if 63 sectors, you should really, really put an mbr on the disk.


> I'm not sure if exactly boot code does understand several disklabels wrapped 
> into MBR. Of course to be sure that's enough to look into the code.
> 
> > 4) It appears as though that assumption, "the first netbsd" slice,
> >    implies that while there's room for a second bsd disklabel on a
> >    second netbsd slice, it could never be read.  Is this correct?
> 
> Technically, you can create up to four disklabels into one MBR so there can be 
> several disklabels on one hard disk. Even more if you put disklabels into 
> extended partitions.

Nope - you can only have 1 disklabel.
You can (now) have more than 1 netbsd partition, and they can all be
bootable - but they all share the same disklabel.

> Last time i checked mbrlabel code (i386/1.6.1) i found that it cares about 
> disklabels inside extended partitions. So in general you may have more then 
> one disklabel on hdd with one "primary" inside master MBR partition and 
> several "slave" inside one or more extended partitions.

mbrlabel is not relevant to this discussion.


	David

-- 
David Laight: david@l8s.co.uk