Thanks for the review! I'm answering here, but would like to thank
Lars-Peter for the review as well.
Mark Brown wrote:
> On Thu, Nov 06, 2014 at 01:54:00PM +0100, Peter Rosin wrote:
>
> > +#define TFA9879_REG(codec, reg, field, value) \
> > + snd_soc_update_bits(codec, TFA9879_ ## reg, \
> > + TFA9879_ ## field ## _MASK, \
> > + (value) << TFA9879_ ## field ## _SHIFT)
>
> Please don't do stuff like this, it just makes the code look more obscure than
> it should be for people who work with multiple devices - just use
> update_bits() directly.
Ok, I'll zap it. Pity though, I really liked the actual code even if the macro
was a
bit hard to digest...
> > +static int tfa9879_prepare(struct snd_pcm_substream *substream,
> > + struct snd_soc_dai *dai)
> > +{
> > + struct snd_soc_codec *codec = dai->codec;
> > +
> > + TFA9879_REG(codec, DEVICE_CONTROL, POWERUP, 1);
>
> Use DAPM to manage power, you're probably looking for a supply widget.
I'll try to find out how to hook that up. BTW, there is very limited info about
the supply widget at http://www.alsa-project.org/main/index.php/DAPM
> > + switch (params_format(params)) {
> > + case SNDRV_PCM_FORMAT_S16_LE:
>
> Use params_width() rather than a specific memory layout.
Ok.
> > + if (tfa9879->lsb_justified)
> > + TFA9879_REG(codec, SERIAL_INTERFACE_1, I2S_SET,
> i2s_set);
>
> Why does this need to be reset every time, shouldn't we just be setting the
> register in set_fmt().?
Yes, I'd sure like to do that, but how do I get to the width in set_fmt()?
> > +static int tfa9879_probe(struct snd_soc_codec *codec) {
> > + struct tfa9879_priv *tfa9879 = snd_soc_codec_get_drvdata(codec);
> > +
> > + codec->control_data = tfa9879->regmap;
>
> This is redundant, just remove the entire function - ASoC will get the regmap
> automatically.
Oops, sorry. Forward-porting in progress...
> > + { TFA9879_MISC_STATUS, 0x0000 }, /* 0x15, read-only */
> > +};
> > +
> > +static bool tfa9879_volatile_register(struct device *dev, unsigned
> > +int reg) {
> > + return reg == TFA9879_MISC_STATUS;
>
> If the register is volatile it shouldn't have a default value provided.
Then I misunderstood what volatile was meant to do. I'll just nuke the
function. It works fine anyway...
I'll send a v2 later, with the other bits from Lars-Peter taken care of, and
with 2/2 squashed. Need to test first though...
Cheers,
Peter
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
