Re: [PATCH] ARM: SPEAr600: Add device-tree support to SPEAr600 boards

On 03/14/2012 08:27 AM, Arnd Bergmann wrote:

On Wednesday 14 March 2012, Stefan Roese wrote:

quoted

On Wednesday 14 March 2012 10:48:44 Arnd Bergmann wrote:

quoted

quoted

quoted

I would suggest you convert these to DT next so you can remove the
amba_devs list. Which devices are these? If they are pl061 and
pl010/pl011, the binding should be really easy to do.

Yes. I really wanted to do that. But from a quick look at the pl011
driver (drivers/tty/serial/amba-pl011.c), this driver doesn't support DT
probing. I might have missed something here though. And ideas?

amba primecell devices don't actually need to register to a "compatible"
property, they are probed using the primecell ID, and the device tree
is just used to tell the system about memory and IRQ resources.

They do need to have "arm,primecell" for things to work as the type of
h/w is probeable, but the location of it is not. There should also be a
compatible property for the specific module (i.e. arm,pl011), but Linux
doesn't look at it. There's support for overriding the hw id in the DT
as well which some spear platforms need IIRC.

I thought I had added documentation for some of the primecells and Dave
Martin was going to do the rest, but seems it got dropped on the
floor... Off to dig in my git tree.

quoted

Ahh, I see. Let me see, if I can get this working...

I just saw that there is a patch series for pl061 that Rob Herring did at  

git://sources.calxeda.com/kernel/linux.git irqdomain-for-grant

Please have a look at that branch first.

This is just interrupt support. The pl061 driver already supports DT.
And so do pl011, pl03x, pl022, pl330 (memory to memory only).

Rob

quoted

quoted

Don't worry about the the gpio and uart devices if they are not in the
first initial version. I do think that they should be done fairly soon
though, before we get into most of the other devices.

It's ok if you put a lot of the other devices in the dts file though,
like the ethernet device, it gives a better overview of what is actually
there, even if the driver does not actually use it yet.

My initial idea was to push the devices once their DT support is accepted. 
Otherwise the bindings are still not settled.

Right, that makes sense. However, you can add a property containing

	status = "disabled;

so that the device is visible in the source but no platform device gets
generated. That would have no consequences at run time but can be helpful
for review.

	Arnd