Hi Mark, thank for comment.
On Tue, 2016-06-07 at 16:54 +0100, Mark Brown wrote:
On Fri, Jun 03, 2016 at 12:56:17PM +0800, Garlic Tseng wrote:
- /*enable agent*/
Lots of comments with missing spaces in them, there's quite a few examples of that in here.
I'll check all the code and fix them.
+EXPORT_SYMBOL(mtk_afe_fe_startup);
All the ASoC and regmap APIs are _GPL(), you really shouldn't export new interfaces based on them without it - the point with the _GPL() is that non-GPL code shouldn't be able to use the APIs.
Thanks for comment, I'll fix them.
+EXPORT_SYMBOL(mtk_afe_fe_hw_params);
Do you need to export the individual ops rather than just the ops structure?
Yes, in 2701 driver we modify some ops.
+/* FE DAIs */ +static const struct snd_soc_dai_ops mt2701_single_memif_dai_ops = { + .startup = mt2701_simple_fe_startup, + .shutdown = mtk_afe_fe_shutdown, + .hw_params = mt2701_simple_fe_hw_params, + .hw_free = mtk_afe_fe_hw_free, + .prepare = mtk_afe_fe_prepare, + .trigger = mtk_afe_fe_trigger, + +};
And here is one of them. Here MT2701 need one more reg control.
+static int mt2701_simple_fe_hw_params(struct snd_pcm_substream *substream, + struct snd_pcm_hw_params *params, + struct snd_soc_dai *dai) +{ + struct snd_soc_pcm_runtime *rtd = substream->private_data; + struct mtk_base_afe *afe = snd_soc_platform_get_drvdata(rtd->platform); + int stream_dir = substream->stream; + + /*single DL use PAIR_INTERLEAVE*/ + if (stream_dir == SNDRV_PCM_STREAM_PLAYBACK) { + regmap_update_bits(afe->regmap, + AFE_MEMIF_PBUF_SIZE, + AFE_MEMIF_PBUF_SIZE_DLM_MASK, + AFE_MEMIF_PBUF_SIZE_PAIR_INTERLEAVE); + } + return mtk_afe_fe_hw_params(substream, params, dai); +}
MT2701 need some specific reg controls. If the control is platform-specific I tend not to put them in the common ops structure.
+int mtk_afe_fe_trigger(struct snd_pcm_substream *substream, int cmd,
struct snd_soc_dai *dai)+{
- struct snd_soc_pcm_runtime *rtd = substream->private_data;
- struct snd_pcm_runtime * const runtime = substream->runtime;
- struct mtk_base_afe *afe = snd_soc_platform_get_drvdata(rtd->platform);
- struct mtk_base_afe_memif *memif = &afe->memif[rtd->cpu_dai->id];
- struct mtk_base_afe_irq *irqs = &afe->irqs[memif->irq_usage];
- const struct mtk_base_irq_data *irq_data = irqs->irq_data;
- unsigned int counter = runtime->period_size;
- int fs;
- dev_info(afe->dev, "%s %s cmd=%d\n", __func__, memif->data->name, cmd);
That's way too noisy, dev_dbg() at most.
OK. I'll fix it.
+static DEFINE_MUTEX(irqs_lock); +int mtk_dynamic_irq_acquire(struct mtk_base_afe *afe) +{
- int i;
- mutex_lock(&irqs_lock);
- for (i = 0; i < afe->irqs_size; ++i) {
Why is the lock global and not part of the AFE struct?
I'll put it in AFE struct. Thanks.
+void mtk_simple_isr(struct mtk_base_afe *afe, struct mtk_base_afe_memif *memif) +{
- snd_pcm_period_elapsed(memif->substream);
+} +EXPORT_SYMBOL(mtk_simple_isr);
Is this really worth it over just calling _period_elapsed() directly?
I'll remove the function and just calling period_elapsed() directly.