On 10/04/2011 06:55 PM, Grant Likely :

On Tue, Oct 04, 2011 at 08:08:27AM -0500, Rob Herring wrote:

quoted

Nicolas,

On 10/04/2011 03:18 AM, Nicolas Ferre wrote:

quoted

On 10/04/2011 03:58 AM, Rob Herring :

quoted

On 10/03/2011 04:51 AM, Nicolas Ferre wrote:

quoted

+Optional properties:
+- atmel,use-dma-rx: use of PDC or DMA for receiving data
+- atmel,use-dma-tx: use of PDC or DMA for transmitting data

Is this an internal DMA or separate DMA controller? If the latter, these
should be phandles to a DMA channel/request.

Well, for now it relies on PDC (looks like an internal DMA). There is no
notion of channel/request so I think it is appropriate to keep it like this.

Okay. Although, this is a bit of Linux creeping into DT. DT should
describe the h/w. You should really be describing whether the h/w
supports DMA or not rather than whether you want to use DMA or not.
Since DMA support is really tied to the SOC rather than board, you could
just use the compatible string to distinguish platforms that support DMA
or not. So what is determining this setting? Is it purely user choice?

It is okay for a binding to still requires config properties like this
to state if the DMA can be used, even if compatible is enough to
distinquish.  However, you're right to bring it up.  It is a design
choice that should be made thoughtfully.

Grant, Rob,

Thanks for your inputs.

I would like to keep this binding to allow the user to choose the way he
likes to use integrated DMA.

I plan to resend a patch series about this serial driver once I will be
able to propose a rs485 binding.

Cheers,
-- 
Nicolas Ferre