[alsa-devel] [PATCH] ASoC: SAMSUNG: Add sound card driver for Snow board

[Tushar Behera]

On 23 April 2014 00:03, Mark Brown <broonie at kernel.org> wrote:
> On Tue, Apr 22, 2014 at 07:17:54PM +0530, Tushar Behera wrote:
>> On 22 April 2014 16:14, Mark Brown <broonie at kernel.org> wrote:
>
>> > In general this isn't up to modern standards, please do try to check
>> > that new code is following best practices.  Did the support for setting
>> > the clocking up in the device tree get merged already?
>
>> I didn't get this point. Would you please elaborate?
>
> The out of tree driver for these boards has a bunch of code in it which
> reprograms the clock tree that parents the I2S block so that the I2S
> block has inputs at suitable rates to allow it to generate useful
> outputs.
>

Currently the clocks are configured through clk-exynos-audss.c. The
reparenting of I2S bus clock is not done anymore, but is left to use
the default value set in bootloader. With default setup (XXTI as the
parent), I have tested wave files with various sample rates on Snow.

The missing part right now is setting XCLKOUT (mclk for the audio
codec) properly. In internal tree, I have directly programmed the
XCLKOUT register (we don't have a clock for it as this register is not
part of clock domain). I am now working on preparing a clock provider
for XCLKOUT.

>> > Please do also pay attention to the CC lists when posting patches, this
>> > seems to have been sent to a fairly random selection of people and
>> > lists.
>
>> Okay, I will update the CC list as per get_maintainer script during
>> next revision.
>
> Please think about the results when doing that - get_maintainers is very
> useful but it does generate false positives and miss people.
>

I will do my best to add relevant people in the discussion.

>> >> +     ret = snd_soc_dai_set_clkdiv(cpu_dai, SAMSUNG_I2S_DIV_BCLK, bfs);
>> >> +     if (ret < 0)
>> >> +             return ret;
>
>> > Set this stuff up on probe.  I'm surprised that you need to set BCLK at
>> > all...
>> >
>
>> Should I create a late_probe call for this (in line with tobermory.c)?
>
> Yes.

Ok.

Thanks.
-- 
Tushar Behera