On Sun, 18 Jan 2004, der Mouse wrote:

> > " disklabel wd1, and ensure that wd1a starts at least 100 blocks from the
> > start of the disk .. wd1c should be identical to wd1d .."
>
> > 16 partitions:
> > #        size    offset     fstype  [fsize bsize cpg/sgs]
> >   a:   4142880      1008       RAID                      # (Cyl.    1 -
> > 4110)
> >   c:   4144707         0     unused      0     0         # (Cyl.    0 -
> > 4111*)
> >   d:   4144707         0     unused      0     0         # (Cyl.    0 -
> > 4111*)
>
> > NetBSD slice at 63, partition C at 0
> > Write outside MBR partition? [n]:
>
> I suspect the instructions are the part that's broken here.  d should
> always begin at offset 0 and be the whole disk (that's what ends up
> getting used anyway, since it's RAW_PART), but c is only the NetBSD
> portion of the disk, and should agree with the NetBSD partition's
> boundaries in the MBR ("fdisk") partitioning.  I suspect this is what
> the "slice at 63, c at 0" complaint is about, and that the instructions
> are broken for telling you to make c and d identical.
>
> As for whether 63 is the right number, I can't say.  I'd have to ask
> someone who knows the RAID stuff.

	Putting both c and d to cover the entire disk with a offset
	100 blocks or so allows you to write the bootblocks
	contiguously at the start of the disk. They then pick up
	the kernel from the RAID partition a. All of this is unnecessary
	in -current as the first stage boot blocks understand RAID1, but
	I have many i386 1.6.x boxes setup according to the above.

	None of them have any non NetBSD partitions on the disk, so the
	'mangled' fdisk does not affect anything.

-- 
		David/absolute          -- www.netbsd.org: No hype required --