On Wed, 9 Oct 2002, Alexander Viro wrote:
>
> That's what I'm asking about - do we want to have objects on the _partition_
> side of things that would require per-partition directory? Not on the
> filesystem/swap/whatnot side...
You're going about this the wrong way.
It doesn't _matter_ if you have associations on the partition side. The
_only_ thing that matters is that somebody else has associations _to_ a
partition. That is already enough to require a "node" to associate to.
How would you otherwise describe that association in filesystem terms?
Look, let me give you an example of an existing association.
/devices/bus/scsi/devices:
lrwxrwxrwx 1 root root 40 Oct 7 17:17 1:0:0:0 -> ../../../root/pci0/00:0b.0/scsi1/1:0:0:0
Here the association is from the "list of scsi devices" to the "hardware
device node that is associated with that device".
The thing I want to point out is that you need a "target node" in order to
be able to _have_ this association.
In other words, if you think that it is reasonable to have an assocation
from the MD device to the list of partitions that are part of that MD
device, then you _need_ to have a "partition node", because otherwise you
cannot have the pointer to it. See?
So your "partition side" vs "filesystem side" thing doesn't matter. It
doesn't matter which side the associations are, you need to have a node to
associate _with_.
Noth sides need a node. Even if the relationship is only going in one
direction.
Linus
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Tue Oct 15 2002 - 22:00:32 EST