Re: [PATCH 1/4] ASoC: Add ADAV80x codec driver

[Mark Brown <broonie@opensource.wolfsonmicro.com>]

On Thu, Jun 23, 2011 at 03:36:39AM +0200, Lars-Peter Clausen wrote:
> On 06/23/2011 03:21 AM, Mark Brown wrote:

> >> +	if (adav80x->deemph) {
> >> +		switch (adav80x->rate) {
> >> +		case 0:
> >> +			val = ADAV80X_DAC_CTRL2_DEEMPH_NONE;
> >> +			break;
> >> +		case 32000:
> >> +			val = ADAV80X_DAC_CTRL2_DEEMPH_32;
> >> +			break;
> >> +		case 44100:
> >> +			val = ADAV80X_DAC_CTRL2_DEEMPH_44;
> >> +			break;
> >> +		case 48000:
> >> +		default:
> >> +			val = ADAV80X_DAC_CTRL2_DEEMPH_48;
> >> +			break;

> > Really?  I'd have expected a check for the closest matching rate (which
> > would get 32k for most low rates) or a requirement for an exact match.

> Since the codec supports 32k, 44.1k, 48k, 64k, 48.2k and 96k this will select
> the closest match.

It'd be better to write the code to say that.

> >> +	} else	{
> >> +		if (adav80x->clk_src == new_src)
> >> +			return 0;
> >> +
> >> +		adav80x->clk_src = new_src;
> >> +
> >> +		if (new_src == ADAV80X_CLK_XIN) {
> >> +			/* DAC, ADC, ICLK clock source - XIN */
> >> +			snd_soc_write(codec, ADAV80X_ICLK_CTRL1, 0x00);
> >> +			snd_soc_write(codec, ADAV80X_ICLK_CTRL2, 0x00);
> >> +		} else {
> >> +			/* DAC, ADC, ICLK clock source - MCLKI */
> >> +			snd_soc_write(codec, ADAV80X_ICLK_CTRL1, 0x25);
> >> +			snd_soc_write(codec, ADAV80X_ICLK_CTRL2, 0x01);
> >> +		}
> >> +
> >> +		pll_ctrl1 |= ADAV80X_PLL_CTRL1_PLL1PD;
> >> +		snd_soc_write(codec, ADAV80X_PLL_CTRL1, pll_ctrl1);

> > What's this doing?  Setting the PLL output to zero means stop the PLL.

> That's exactly what's it doing. Switching to an external clock and powering the
> PLL down. Or what do you mean?

It really doesn't look like that - if it were just stopping the PLL it'd
not need to look at the clock source selection.  Perhaps some of this
should be in set_sysclk()?

> >> +static int adav80x_resume(struct snd_soc_codec *codec)
> >> +{
> >> +	return adav80x_set_bias_level(codec, SND_SOC_BIAS_STANDBY);
> >> +}

> > This doesn't appear to restore the register cache, nor does
> > set_bias_level().

> The register contents is not lost unless we'd cut external power, but I could
> add restoring for completeness.

Cutting external power is the sort of thing that one would expect to
happen over suspend on most systems, especially power sensitive ones.