On Feb 28, 2006, at 9:54 PM, Jonathan A. Kollasch wrote:

>  As there are now shipping EFI-based i386 machines, we should consider
> how the partitioning will work on all the EFI machines.  Since EFI  
> itself
> may often need to read (and write) into at least one partition, a  
> simple
> disklabel-only scheme won't work.  Are wedges over the GPT[1]  
> partitions
> the way to go?  Or do we do a i386-like GPT+disklabel thing?  I  
> believe
> FreeBSD/ia64 is using a GPT-only scheme.

Wedges will be the solution for GPT.  There are still some wedge  
issues to work out yet.  I have plans to work on this some more in  
the near future.

> For starters the GPT partitioning program from FreeBSD should be  
> ported
> and imported.  It uses some FreeBSD-specific ioctls for determining
> size of the disk, and I couldn't find an equivalent in NetBSD.  I'd be
> willing to try to get it ported if I got some pointers in this area.

I'm not so sure the GPT program from FreeBSD should be imported as-is.

> Also, for the GPT+disklabel possibility a gptlabel(8) program  
> similar to
> mbrlabel(8) may need to be written.  Or is a wedges-based system  
> with only
> the GPT on disk the way of the future?

Wedges only, please.  The disklabel in-core format can't express what  
GPT can do.

>
> 	Jonathan Kollasch
>
> [1] http://en.wikipedia.org/wiki/GUID_Partition_Table

-- thorpej