On Wed, Apr 03, 2002 at 07:50:34AM -0500, Thomas Mueller wrote:
> from Manuel Bouyer:
> 
> > See the mbrlabel program. There has been discussions about this in tech-kern,
> > and if I remember rigth the conclusion was that we don't want to put that
> > much knowledge in the kernel ...
> 
> You mean it's better to render a hard disk nonbootable and render a DOS 
> partition nonbootable, and risk overwriting less-easily-rescuable data when the
> hard disk is shared with DOS or Windows, and Linux?  I suppose that would not be
> a concern for BSD servers, where BSD is installed to the whole hard disk and
> there is no need to partition to share with any other OS.  But I am led to
> believe NetBSD is not really compatible with Linux and DOS on the same hard
> disk.

If netbsd wants to be compatible with other x86 PC operating systems it
really ought to be able to parse the mbr partition table and extended
partition list (I suspect I know the format - it isn't the 4-way
branching tree it appears to be at first sight).

OTOH if netbsd x86 is the x86 instantiation of a generic OS then
it doesn't care about other formats :-)

The amount of code required to parse the extended partition chain
is trivial compared to the size of the current default kernel and
the memory size of modern systems.  The slight difficulty is
picking names for the devices and fitting the minor device format...

I suspect there is space to actually boot netbsd from an extended
partition using the mbr_bootsel code.  It might have to assume
that LBA reads are possible - anyone know how many systems have
BIOS that don't support LBA?


	David

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