Re: [Patch -mm 2/5] driver core: Introduce device_move(): move a device to a new parent.

["Dmitry Torokhov" <dmitry.torokhov@gmail.com>]

On 11/15/06, Cornelia Huck <cornelia.huck@de.ibm.com> wrote:
> On Wed, 15 Nov 2006 10:31:59 +0100,
> Kay Sievers <kay.sievers@vrfy.org> wrote:
>
> > We need the old DEVPATH in the environment (or something similar),
> > otherwise we can't connect the event with the new device location to the
> > current device. :)
>
> Duh. I've attached another completely untested patch below.
>
> > > Wouldn't we need something similar for kobject_rename()
> > > as well?
> >
> > Maybe kobject_rename() can go, if we have a move function which can be
> > used. In any case, the events should look identical to userspace, yes.
>
> I think kobject_move() and kobject_rename() are two different beasts.
> kobject_move() changes the topology, kobject_rename() changes an
> identifier. Shouldn't they be reported in two different ways to
> userspace?
>

Why do we need to have them at all? Devices should not "move" in the
trees - it it moves we should just treat them as old devices going
away and new devices appearing... Renames - they are only used to
rename net devices, don't they? I wonder if we could have just a
separate "alias" or "name" sysfs attribute for them and get away with
renaming of devices.

-- 
Dmitry