Re: [PATCH] of: support an enumerated-bus compatible value

From: Stephen Warren <swarren-3lzwWm7+Weoh9ZMKESR00Q@public.gmane.org>
To: Segher Boessenkool
	<segher-XVmvHMARGAS8U2dJNN8I7kB+6BGkLq7r@public.gmane.org>
Cc: devicetree-discuss-uLR06cmDAlY/bJ5BZ2RsiQ@public.gmane.org
Subject: Re: [PATCH] of: support an enumerated-bus compatible value
Date: Mon, 02 Jul 2012 16:26:57 -0600	[thread overview]
Message-ID: <4FF22031.3060206@wwwdotorg.org> (raw)
In-Reply-To: <6BC22F77-77D7-45DF-821A-6CA2DBADEA59-XVmvHMARGAS8U2dJNN8I7kB+6BGkLq7r@public.gmane.org>

On 07/02/2012 03:44 PM, Segher Boessenkool wrote:
>> But don't get too hung up on regulators; there are plenty of other
>> devices that can exist in DT that aren't memory-mapped; GPIO-keys,
>> aggregate sound complexes, perhaps WiFi/rfkill nodes, etc. All are
>> affected by the same DT representation issue.
> 
> So what is the issue?  You have some node (the "bus node") under which
> you have the "regulator@0", "frobnicator@1" etc. nodes.  You refer to
> these device nodes by phandle always, you cannot programmatically
> control those devices (otherwise, they should be child node of a *real*
> bus node).  The bus node does not represent any device at all, and no
> device driver can ever do anything with it; it should not have a
> "compatible" property either.

Without the "bus" node having a compatible property, nothing will
recurse into it to instantiate the devices that the child nodes
represent; the fact that they're reference by phandle doesn't
instantiate the device, just references it.