Re: [RFC][PATCH 07/10 v2] ASoC: hisilicon: Add hi6210 i2s audio driver

[John Stultz <john.stultz@linaro.org>]

On Tue, Jul 19, 2016 at 5:28 PM, Mark Brown <broonie@kernel.org> wrote:
> On Tue, Jul 19, 2016 at 04:22:43PM -0700, John Stultz wrote:
>
>>  sound/soc/Kconfig                |   1 +
>>  sound/soc/Makefile               |   1 +
>>  sound/soc/hisilicon/Kconfig      |   5 +
>>  sound/soc/hisilicon/Makefile     |   1 +
>>  sound/soc/hisilicon/hi6210-i2s.c | 678 +++++++++++++++++++++++++++++++++++++++
>>  sound/soc/hisilicon/hi6210-i2s.h | 276 ++++++++++++++++
>>  6 files changed, 962 insertions(+)
>
> This is adding a new binding without documenting it and still looks like
> it's far more than an I2S controller.

So I added some binding documentation in the patch that proceeded this...

Oh Crud. I forgot to add the cc list to those files. Sorry for that.

Here they are on the list:
https://lkml.org/lkml/2016/7/19/820
https://lkml.org/lkml/2016/7/19/815


As for being more then an i2s controller, I don't have access to the
docs this was written with, so I'm not really sure what I can do to
properly extend this driver beyond acting as an i2s driver.  I'll take
a look, but again, my lack of familiarity with ASoC means I may need
some extra guidance. So apologies up front.

>> +     switch (params_rate(params)) {
>> +     default:
>> +             dev_err(cpu_dai->dev, "Bad rate\n");
>> +             return -EINVAL;
>
> We should tell the user what rate.
>
>> +     if (bits == HII2S_BITS_24) {
>> +             i2s->bits = 32;
>> +             dma_data->addr_width = 3;
>> +     } else {
>> +             i2s->bits = 16;
>> +             dma_data->addr_width = 2;
>> +     }
>
> This looks like it should be a switch statement, there's some similar
> stuff for the channels.

So yea, the switch selection above this does the validation and then
I'm just applying the change here.

I can duplicate the switch in both cases, but this seems a little more terse.

>> +     _hi6210_i2s_set_fmt(i2s, substream);
>
> Why is this not in line given that this is the only user?

I think it breaks up the function some, but I can move it inline.


>> +     res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
>> +     if (!res) {
>> +             ret = -ENODEV;
>> +             goto err2;
>> +     }
>> +
>> +     i2s->base = devm_ioremap_resource(dev, res);
>
> devm_ioremap_resource() will check the error.
>
>> +err3:
>> +     while (--i2s->clocks)
>> +             clk_put(i2s->clk[i2s->clocks]);
>> +
>> +err2:
>> +     kfree(i2s);
>
> You switched to using devm_ but left the error handling.

Ok, will address.

Thanks for the review and feedback!
-john