[PATCH 1/3] ASoC: add 88pm860x codec driver

[Mark Brown]

On Tue, Aug 17, 2010 at 06:44:44PM +0800, Haojian Zhuang wrote:
> On Tue, Aug 17, 2010 at 5:58 PM, Mark Brown

> >> +     /* unmute DAC */
> >> +     snd_soc_update_bits(codec, PM860X_DAC_OFFSET, DAC_MUTE, 0);

> > Can you explain what's going on with this mute handling please?

> Em. Actually there should be automute variable. I shouldn't delete
> that variable.  In order to anti-pop, mute DAC before enabling DAC.
> Unmute it after enabling DAC. It's required by silicon.

As requested in the previous review comments you need to provide better
documentation of this in the driver.  I'd also expect to see a custom
mute control somewhere which implements this by keeping the DAC muted
unless it has been enabled rather than some sort of automute handling.

> >> +     switch (event) {
> >> +     case SND_SOC_DAPM_PRE_PMU:
> >> +             snd_soc_update_bits(codec, PM860X_ADC_EN_1, en1, en1);
> >> +             snd_soc_update_bits(codec, PM860X_ADC_EN_2, en2, en2);

> > I still don't follow why you need a custom event for this.

> Enabling both bit 0 of ADC_EN_1 and bit 5 of ADC_EN_2 can enable left
> ADC. Enabling both bit 1 of ADC_EN_1 and bit 4 of ADC_EN_2 can enable
> right ADC. I can't find any DAPM API can handle this. So I implement
> the custom event.

As I said previously I would expect you to be using a DAPM supply widget
for this.

> >> +static int pm860x_mic1_event(struct snd_soc_dapm_widget *w,
> >> +                          struct snd_kcontrol *kcontrol, int event)
> >> +{
> >> +     struct snd_soc_codec *codec = w->codec;
> >> +
> >> +     switch (event) {
> >> +     case SND_SOC_DAPM_POST_PMU:
> >> +             /* Enable Mic1 Bias & MICDET, HSDET */
> >> +             snd_soc_update_bits(codec, PM860X_ADC_ANA_1, MIC1BIAS_MASK,
> >> +                                 MIC1BIAS_MASK);

> > As I said last time you should handle this via DAPM.

> I registered it as DAPM widget. Why do you say it's not handled via DAPM?

I say this because this code is manually going in and enabling the
microphone bias here.  If the microphone bias is being handled using
DAPM you shouldn't be going around explicitly writing to the power
control bit like this.

> >> +             pm860x_set_bits(codec->control_data, REG_MIC_DET,
> >> +                             MICDET_MASK, MICDET_MASK);
> >> +             pm860x_set_bits(codec->control_data, REG_HS_DET,
> >> +                             EN_HS_DET, EN_HS_DET);

> > This should be associated with enabling microphone detection.

> Yes, but you said that it could be controlled by enable_pin(). I
> forced to enable microphone pins in machine driver.

I don't follow what you're saying at all.  This code will unconditonally
enable headset and microphone detection while the microphone is enabled,
regardless of the user deciding if they want to use that feature or not.
The use of enable_pin() was for the microphone bias which is a separate
thing.

> >> +     /* set master/slave audio interface */
> >> +     switch (fmt & SND_SOC_DAIFMT_MASTER_MASK) {
> >> +     case SND_SOC_DAIFMT_CBM_CFM:
> >> +     case SND_SOC_DAIFMT_CBM_CFS:
> >> +             if (pm860x->dir == PM860X_CLK_DIR_OUT)
> >> +                     *inf |= PCM_INF2_MASTER;
> >> +             else
> >> +                     return -EINVAL;
> >> +             break;

> > You're setting the same register configuration for two different DAI
> > clock master configurations here.  Presumably one of the settings is
> > inaccurate?

> No, they're different registers. But offsets are same. So I just return pointer.

I can't associate your comment there with the code at all.  The code
does nothing different for the two case statements and there's no other
code I can see.

> >> +static int pm860x_set_dai_sysclk(struct snd_soc_dai *codec_dai,
> >> +                              int clk_id, unsigned int freq, int dir)
> >> +{
> >> +     struct snd_soc_codec *codec = codec_dai->codec;
> >> +     struct pm860x_priv *pm860x = snd_soc_codec_get_drvdata(codec);
> >> +
> >> +     if (dir == PM860X_CLK_DIR_OUT)
> >> +             pm860x->dir = PM860X_CLK_DIR_OUT;
> >> +     else
> >> +             pm860x->dir = PM860X_CLK_DIR_IN;
> >> +
> >> +     return 0;
> >> +}

> > What is this actually setting - which clock is being configured here?

> While codec is master, the clock is fixed. I needn't set detail clock.
> While codec is slave, it's not supported in this patch yet.

What is "the clock", and if slave mode is not supported surely an error
should be returned?

> >> +     if (shrt & (SHORT_LO1 | SHORT_LO2))
> >> +             report |= PM860X_SHORT_LINEOUT;
> >> +     if (shrt & (SHORT_HS1 | SHORT_HS2))
> >> +             report |= PM860X_SHORT_HEADSET;
> >> +     dev_dbg(pm860x->codec->dev, "report:0x%x\n", report);
> >> +     return IRQ_HANDLED;

> > It would seem better to just remove the interrupt handling support
> > entirely if you're not going to implement jack detection.  Right now all
> > the curernt code will do is waste power by enabling the feature but
> > ignoring the result.

> I need a document on illustrating jack on alsa. Could you share one?

There's a number of in tree examples - seach for snd_soc_jack.