On 06/17/2014 07:16 AM, Sean Cross wrote:
A couple of small bits inline.
[...]
--- /dev/null +++ b/sound/soc/codecs/es8328.c @@ -0,0 +1,711 @@
[...]
+#include <linux/platform_device.h>
You don't use platform devices in here
[...]
+uint8_t sample_ratios[] = {
static const, also a prefix like es8328_ wont hurt
- [ES8328_RATE_8019] = 0x9,
- [ES8328_RATE_11025] = 0x7,
- [ES8328_RATE_22050] = 0x4,
- [ES8328_RATE_44100] = 0x2,
+};
+#define ES8328_RATES (SNDRV_PCM_RATE_44100 | \
SNDRV_PCM_RATE_22050 | \SNDRV_PCM_RATE_11025)+#define ES8328_FORMATS (SNDRV_PCM_FMTBIT_S16_LE)
+/* codec private data */ +struct es8328_priv {
- struct regmap *regmap;
- int sysclk;
+};
+/*
- ES8328 Controls
- */
+static const char * const deemph_txt[] = {"None", "32Khz", "44.1Khz", "48Khz"}; +static SOC_ENUM_SINGLE_DECL(deemph,
ES8328_DACCONTROL6, 6, deemph_txt);
deemph should just be a single boolean switch (On or Off) and automatically select the correct setting based on the configured sample rate.
[...]
+static const char * const es8328_line_texts[] = {
- "Line 1", "Line 2", "PGA", "Differential"};
+static const unsigned int es8328_line_values[] = {
- 0, 1, 2, 3};
You don't need to use a value enum for this.
[...]
+static const char * const es8328_pga_sel[] = {
- "Line 1", "Line 2", "Line 3", "Differential"};
+static const unsigned int es8328_pga_val[] = { 0, 1, 2, 3 };
same here
[...]
+static int es8328_hw_params(struct snd_pcm_substream *substream,
- struct snd_pcm_hw_params *params,
- struct snd_soc_dai *dai)
+{
- struct snd_soc_codec *codec = dai->codec;
- if (substream->stream == SNDRV_PCM_STREAM_PLAYBACK) {
u8 dac = snd_soc_read(codec, ES8328_DACCONTROL2);dac &= ~ES8328_DACCONTROL2_RATEMASK;switch (params_rate(params)) {case 8000:dac |= sample_ratios[ES8328_RATE_8019];break;case 11025:dac |= sample_ratios[ES8328_RATE_11025];break;case 22050:dac |= sample_ratios[ES8328_RATE_22050];break;case 44100:dac |= sample_ratios[ES8328_RATE_44100];break;default:dev_err(codec->dev, "%s: unknown rate %d\n",__func__, params_rate(params));return -EINVAL;}snd_soc_write(codec, ES8328_DACCONTROL2, dac);- }
- if (substream->stream == SNDRV_PCM_STREAM_CAPTURE) {
u8 adc = snd_soc_read(codec, ES8328_ADCCONTROL5);adc &= ~ES8328_ADCCONTROL5_RATEMASK;switch (params_rate(params)) {case 8000:adc |= sample_ratios[ES8328_RATE_8019];break;case 11025:adc |= sample_ratios[ES8328_RATE_11025];break;case 22050:adc |= sample_ratios[ES8328_RATE_22050];break;case 44100:adc |= sample_ratios[ES8328_RATE_44100];break;default:dev_err(codec->dev, "%s: unknown rate %d\n",__func__, params_rate(params));return -EINVAL;}snd_soc_write(codec, ES8328_ADCCONTROL5, adc);- }
The capture and the playback case look pretty much the same. This could probably be handled without duplicating the code. Also use snd_soc_update_bits() for updating the register.
- return 0;
+}
[...][
+static int es8328_probe(struct snd_soc_codec *codec) +{
- /* power on device */
- return es8328_set_bias_level(codec, SND_SOC_BIAS_STANDBY);
+}
You don't need this one, the core will automatically bring the dapm context to idle bias after the device has been probed.
+static int es8328_remove(struct snd_soc_codec *codec) +{
- return 0;
+}
No, need to have empty callbacks, just leave .remove of the codec driver struct NULL.
+static struct snd_soc_codec_driver soc_codec_dev_es8328 = {
const
the soc_codec_dev_foobar naming pattern comes from a time when this actually used to be a device rather than a driver. Something like es8328_codec_driver is a more appropriate name these days.
- .probe = es8328_probe,
- .remove = es8328_remove,
- .suspend = es8328_suspend,
- .resume = es8328_resume,
- .set_bias_level = es8328_set_bias_level,
- .controls = es8328_snd_controls,
- .num_controls = ARRAY_SIZE(es8328_snd_controls),
- .dapm_widgets = es8328_dapm_widgets,
- .num_dapm_widgets = ARRAY_SIZE(es8328_dapm_widgets),
- .dapm_routes = es8328_dapm_routes,
- .num_dapm_routes = ARRAY_SIZE(es8328_dapm_routes),
+};
[...]
+#if defined(CONFIG_SPI_MASTER)
For CODECs with more than one bus interface we started splitting the bus interface support code into separate modules (e.g. one module for the core logic of the driver, one module for I2C and one module for SPI). This was done to avoid issue when one bus is built-in and the the other is built as a module. New drivers should follow this scheme. E.g. see http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/sound/...
+static int es8328_spi_probe(struct spi_device *spi) +{
- struct es8328_priv *es8328;
- int ret;
- es8328 = devm_kzalloc(&spi->dev, sizeof(struct es8328_priv),
sizeof(*es8328) is the preferred style
GFP_KERNEL);- if (es8328 == NULL)
return -ENOMEM;- es8328->regmap = devm_regmap_init_spi(spi, &es8328_regmap);
- if (IS_ERR(es8328->regmap))
return PTR_ERR(es8328->regmap);- spi_set_drvdata(spi, es8328);
- ret = snd_soc_register_codec(&spi->dev,
&soc_codec_dev_es8328, &es8328_dai, 1);- if (ret < 0)
dev_err(&spi->dev, "unable to register codec: %d\n", ret);- return ret;
+}
+static int es8328_spi_remove(struct spi_device *spi) +{
- snd_soc_unregister_codec(&spi->dev);
- return 0;
[...]
+static const struct i2c_device_id es8328_i2c_id[] = {
- { "es8328", 0x11 },
The second field is the drvdata, this is not the I2C address of the device. If you don't use drvdata just leave it 0.
- { }
+}; +MODULE_DEVICE_TABLE(i2c, es8328_i2c_id);
[...]