Re: [RFC PATCH] drivers/amba: probe via device tree reworked

On Wed, May 18, 2011 at 02:39:10PM -0500, Rob Herring wrote:

On 05/18/2011 01:47 PM, Grant Likely wrote:

quoted

On Wed, May 18, 2011 at 01:14:47PM -0500, Rob Herring wrote:

quoted

From: Rob Herring<rob.herring-bsGFqQB8/DxBDgjK7y7TUQ@public.gmane.org>

This reworks the original amba bus device tree probing to be more inline with
how platform bus probing via device tree is done using a match table.

You should also mention that this patch builds on the code currently
in devicetree/test on git://git.secretlab.ca/git/linux-2.6.  Otherwise
people will be very confused about what this applies to.

Yeah. I was ultimately expecting to squash this one into the first
amba bus patch assuming you agree with the general direction. Would
you rather me combine it and send that out instead of an incremental
patch?

yes.

quoted

quoted

The amba bus code doesn't support creation of parent devices, so a flat tree
is created even if the device tree contains a heirarchy of devices.

Hmmm, why do you say this?  The code doesn't currently set the parent
pointer, but nothing prevents a caller of amba_device_register() from
setting the parent pointer before it is registered.  I also don't see
anything in the amba bus code that assume it must all be flat in the
driver model.  What bit am I missing?

The problem is not setting a parent pointer. That can be passed in.
The problem I had was creating an amba_device for the bus, but your
comment below to use a platform device clarifies that issue.

okay, good.

g.