Re: [bk/patch] driver model update: device_unregister()

[Linus Torvalds]

On Wed, 9 Oct 2002, Alexander Viro wrote:
> 
> It makes sense, but that should be done for gendisk.  I.e. we should have
> (name, base, range) - not a node for each partition.

Actually, we _do_ want to have a node for each partition, if we want to
show the things that are associated with one particular partition. And we
do have those things - mounts and (onc eit's working again) LVM
relationships etc.

It's a perfectly valid question to ask "what partitions are part of this 
extended disk?" or "which partition is the backing store for this 
filesystem?".  Which implies that a partition is a real first-class 
entity, not just "one of a range".

		Linus