Mark,
I sended the v2 patches and fixed the follows,
On 2015年10月20日 03:09, Mark Brown wrote:
> On Tue, Oct 13, 2015 at 09:01:06AM +0800, Shunqian Zheng wrote:
>
>> +config SND_SOC_INNO_RK3036
>> + tristate "Inno codec driver for RK3036 SoC"
>> + depends on ARCH_ROCKCHIP
>> +
> There doesn't appear to be any build dependency here so can we have an
> || COMPILE_TEST please and also add to SND_SOC_ALL_CODECS.
Delete the dependency and add to SND_SOC_ALL_CODECS.
>
>> +#define DEBUG
> Please remove this for upstream.
Removed.
>
>> +static int rk3036_codec_dai_set_sysclk(struct snd_soc_dai *dai, int clk_id,
>> + unsigned int fre, int dir)
>> +{
>> + /* Nothing to be done here */
>> + return 0;
>> +}
> Remove empty functions, if they can be empty they can just be omitted.
Removed.
>
>> +static int rk3036_codec_dai_digital_mute(struct snd_soc_dai *dai, int mute)
>> +{
>> + struct snd_soc_codec *codec = dai->codec;
>> + unsigned int val = 0;
>> +
>> + dev_dbg(codec->dev, "rk3036_codec dai mute : %d\n", mute);
>> + if (mute)
>> + val |= INNO_REG_09_HP_OUTR_MUTE_YES |
>> + INNO_REG_09_HP_OUTL_MUTE_YES;
>> + else
>> + val |= INNO_REG_09_HP_OUTR_MUTE_NO |
>> + INNO_REG_09_HP_OUTL_MUTE_NO;
> Does the device actually need muting to avoid audio issues? If not it
> looks like this is something that might be better made a user visible
> control instead.
Make mute as a control.
>
>> +static void rk3036_codec_power_off(struct snd_soc_codec *codec)
>> +{
>> + struct inno_reg_val *reg_val;
>> + int i;
>> +static void rk3036_codec_power_on(struct snd_soc_codec *codec)
>> +{
>> + struct inno_reg_val *reg_val;
>> + int i;
> I'd expect to see these be part of the set_bias_level() operation.
Move the part of setting precharge current to set_bias_level()
>
>> + dev_dbg(codec->dev, "rk3036_codec power on\n");
>> + /* set a big current for capacitor discharging. */
>> + snd_soc_write(codec, INNO_REG_10, INNO_REG_10_MAX_CUR);
>> + mdelay(10);
>> + /* start precharge */
>> + snd_soc_write(codec, INNO_REG_06, INNO_REG_06_DAC_PRECHARGE);
>> + mdelay(100);
>> +
>> + for (i = 0; i < ARRAY_SIZE(inno_codec_open_path); i++) {
>> + reg_val = &inno_codec_open_path[i];
>> + snd_soc_write(codec, reg_val->reg, reg_val->val);
>> + mdelay(5);
>> + }
> This looks suspicious... why isn't the power up sequence managed via
> DAPM?
Use DAPM to manage the power up path.
>
>> +static struct regmap *rk3036_codec_get_regmap(struct device *dev)
>> +{
>> + struct rk3036_codec_priv *priv = dev_get_drvdata(dev);
>> +
>> + dev_dbg(dev, "rk3036_codec get regmap\n");
>> + return priv->regmap;
>> +}
> There's already dev_get_regmap().
Delete this functions.
>
>> +static int codec_reg_read(void *context, unsigned int reg, unsigned int *val)
>> +{
>> + struct rk3036_codec_priv *priv = context;
>> + void __iomem *base = priv->base;
>> +
>> + *val = readl(base + reg);
>> + return 0;
>> +}
>> +static struct regmap_bus codec_regmap_bus = {
>> + .reg_read = codec_reg_read,
>> + .reg_write = codec_reg_write,
> This looks like it's just regmap_mmio, why not just use that?
Use the devm_regmap_init_mmio().
>
>> + ret = regmap_write(grf, 0x00140, BIT(16 + 10) | BIT(10));
>> + if (ret != 0) {
> Can we have some defines for this rather than all the magic numbers?
Define some macros for grf select.
>
>> +static void rk3036_codec_platform_shutdown(struct platform_device *pdev)
>> +{
>> + /*TODO:*/
>> +}
> If you use DAPM you shouldn't need to do anything here.
Delete it.
Thank you very much,
Shunqian