[RFC PATCH 0/6] Those patches is used for dw_hdmi audio.

[ykk@rock-chips.com (Kuankuan.Yang)]

Hi Russell:

I got an idea that we can split the pcm dma part code out, after that we 
can chose the buffer transmit way (AUD_DMA or I2S).
In that way i will make another i2s driver to transmit those buffer, but 
in the mainline kernel already lanched an rockchip i2s driver
(rockchip_i2s.c), so seams it maybe not an good way.

what's your opinion, russell?

Best Regards.

? 2014?12?15? 20:00, Russell King - ARM Linux ??:
> On Mon, Dec 15, 2014 at 07:52:06PM +0800, Kuankuan.Yang wrote:
>>   Hi Russell:
>>
>>       thks for your replay, actually you also have send me those
>> dw-hdmi-audio.c patches, and I also agree it's an beautiful way to make
>> hdmi-audio works. Beside,
>>   I try to reuse it into our platform, and actually the system have created
>> the DW_HDMI sound card successfully, but i cannot play any sound  with this
>> sound card.
>>       After dump the registers, I found the part of "Audio DMA Registers"
>> cannot write and always read with 0x00.  So I searching the document
>> "Designware Core
>>   HDMI Transmitter Controller Databook", and found that  "Audio DMA
>> Registers" only present when the hardware configuration parameter AUDIO_IF
>> is set to
>>   AHBAUDDMA. Than I communicate with our IC colleagues, they told me that our
>> cpu rk3288 only support two way to  transmit audio data( I2S & SPDIF ), in
>> that
>> way we do not support AHB_DMA, it's very sad, and this it why i give up this
>> way, also it's my bad that i should replay to u first in the before mail.
> Okay, that means there is some work to be done to figure out how to
> support this correctly so that both the iMX and Rockchip code can
> co-exist together in the mainline kernel - that means we _both_ need
> to work together on this problem _before_ this code gets merged, so
> that we have a common approach between the two code bases.
>
> I really don't want to end up in another cocked up situation like
> what happened with the Dove audio, where it became politically
> impossible for the SolidRun platform to be properly supported by
> mainline kernels.
>