Re: [PATCH v2 03/16] ARM: dts: provide DMA config to pxamci

[Lars-Peter Clausen <lars@metafoo.de>]

On 12/16/2013 12:47 PM, Sergei Ianovich wrote:
> On Mon, 2013-12-16 at 10:58 +0100, Daniel Mack wrote:
>> On 12/14/2013 08:34 PM, Sergei Ianovich wrote:
>>> On Sat, 2013-12-14 at 20:06 +0100, Arnd Bergmann wrote:
>>
>>>> The patch looks ok in case we are merging your patches for 3.14
>>>> and Daniel's patches later than that. If they end up in the
>>>> same merge window however, we'd have to be care to resolve
>>>> the obvious conflict in a proper way.
>>>
>>> The most recently published Daniel's patch (Aug 2013) wraps
>>> IORESOURCE_DMA handling on DT presence in a similar way,
>>
>> Erm, no it doesn't. My patch uses dma_request_slave_channel_compat() in
>> DT case, and that works fine with the current version of pdma, and
>> there's no need to read the "dmas" properties directly.
>>
>> If you want to provide a way to simply denote the dma channel numbers,
>> without looking at the actual phandle, then yes, we could merge this
>> patch first, but it would be effectively reverted a proper implementation.
> 
> Daniel is right. His patch doesn't need to read "dmas" directly. So my
> patch won't need to change drivers/mmc/host/pxamci.c at all, if it is
> applied after Daniel's one.
> 

Btw. any driver that parses the dmas property manually should be considered
broken. This is a classical layering violation. The layout of the dma
specifier is DMA controller specific and should be completely transparent to
the device driver.

- Lars