Re: [PATCH v2 8/9] ASoC: qdsp6-dai: add gapless support

[Srinivas Kandagatla]

Thanks Pierre for quick review!

On 21/07/2020 20:53, Pierre-Louis Bossart wrote:
> > ÂÂÂÂÂ case ASM_CLIENT_EVENT_CMD_EOS_DONE:
> > -ÂÂÂÂÂÂÂ prtd->state = Q6ASM_STREAM_STOPPED;
> > +ÂÂÂÂÂÂÂ spin_lock_irqsave(&prtd->lock, flags);
> > +ÂÂÂÂÂÂÂ if (prtd->notify_on_drain) {
> > +ÂÂÂÂÂÂÂÂÂÂÂ if (substream->partial_drain) {
> > +ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ /**
>
> why the kernel-doc style comment?
Nothing intentional, will fix this!

> [...]

> > -static int q6asm_dai_compr_set_params(struct snd_soc_component 
> > *component,
> > -ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ struct snd_compr_stream *stream,
> > -ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ struct snd_compr_params *params)
> > +static int __q6asm_dai_compr_set_codec_params(struct 
> > snd_soc_component *component,
> > +ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ struct snd_compr_stream *stream,
> > +ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ struct snd_codec *codec,
> > +ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ int stream_id)
>
> not sure I get why you added the __ prefix, does it have any semantic 
> meaning?
Nope, just to mark them it as internal function, as the function name is 
very much similar to q6asm_dai_compr_set_params() callback!
I will try to come up with better naming and also move the indirection 
changes to a separate patch!

> > Â {
> > ÂÂÂÂÂ struct snd_compr_runtime *runtime = stream->runtime;
> > ÂÂÂÂÂ struct q6asm_dai_rtd *prtd = runtime->private_data;
> > -ÂÂÂ struct snd_soc_pcm_runtime *rtd = stream->private_data;
> > -ÂÂÂ int dir = stream->direction;
> > -ÂÂÂ struct q6asm_dai_data *pdata;
> > ÂÂÂÂÂ struct q6asm_flac_cfg flac_cfg;
> > ÂÂÂÂÂ struct q6asm_wma_cfg wma_cfg;
> > ÂÂÂÂÂ struct q6asm_alac_cfg alac_cfg;
> > @@ -667,53 +718,18 @@ static int q6asm_dai_compr_set_params(struct 
> > snd_soc_component *component,
> > ÂÂÂÂÂ struct snd_dec_alac *alac;
> > ÂÂÂÂÂ struct snd_dec_ape *ape;
> > -ÂÂÂ codec_options = &(prtd->codec_param.codec.options);
> > -
> > -
> > -ÂÂÂ memcpy(&prtd->codec_param, params, sizeof(*params));
> > -
> > -ÂÂÂ pdata = snd_soc_component_get_drvdata(component);
> > -ÂÂÂ if (!pdata)
> > -ÂÂÂÂÂÂÂ return -EINVAL;
> > -
> > -ÂÂÂ if (!prtd || !prtd->audio_client) {
> > -ÂÂÂÂÂÂÂ dev_err(dev, "private data null or audio client freed\n");
> > -ÂÂÂÂÂÂÂ return -EINVAL;
> > -ÂÂÂ }
> > -
> > -ÂÂÂ prtd->periods = runtime->fragments;
> > -ÂÂÂ prtd->pcm_count = runtime->fragment_size;
> > -ÂÂÂ prtd->pcm_size = runtime->fragments * runtime->fragment_size;
> > -ÂÂÂ prtd->bits_per_sample = 16;
> > -ÂÂÂ if (dir == SND_COMPRESS_PLAYBACK) {
> > -ÂÂÂÂÂÂÂ ret = q6asm_open_write(prtd->audio_client, prtd->stream_id,
> > -ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ params->codec.id, params->codec.profile,
> > -ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ prtd->bits_per_sample, true);
> > -
> > -ÂÂÂÂÂÂÂ if (ret < 0) {
> > -ÂÂÂÂÂÂÂÂÂÂÂ dev_err(dev, "q6asm_open_write failed\n");
> > -ÂÂÂÂÂÂÂÂÂÂÂ q6asm_audio_client_free(prtd->audio_client);
> > -ÂÂÂÂÂÂÂÂÂÂÂ prtd->audio_client = NULL;
> > -ÂÂÂÂÂÂÂÂÂÂÂ return ret;
> > -ÂÂÂÂÂÂÂ }
> > -ÂÂÂ }
> > +ÂÂÂ codec_options = &(prtd->codec.options);
> > -ÂÂÂ prtd->session_id = q6asm_get_session_id(prtd->audio_client);
> > -ÂÂÂ ret = q6routing_stream_open(rtd->dai_link->id, LEGACY_PCM_MODE,
> > -ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ prtd->session_id, dir);
> > -ÂÂÂ if (ret) {
> > -ÂÂÂÂÂÂÂ dev_err(dev, "Stream reg failed ret:%d\n", ret);
> > -ÂÂÂÂÂÂÂ return ret;
> > -ÂÂÂ }
> > +ÂÂÂ memcpy(&prtd->codec, codec, sizeof(*codec));
> > -ÂÂÂ switch (params->codec.id) {
> > +ÂÂÂ switch (codec->id) {
> > ÂÂÂÂÂ case SND_AUDIOCODEC_FLAC:
> > ÂÂÂÂÂÂÂÂÂ memset(&flac_cfg, 0x0, sizeof(struct q6asm_flac_cfg));
> > ÂÂÂÂÂÂÂÂÂ flac = &codec_options->flac_d;
> > -ÂÂÂÂÂÂÂ flac_cfg.ch_cfg = params->codec.ch_in;
> > -ÂÂÂÂÂÂÂ flac_cfg.sample_rate =Â params->codec.sample_rate;
>
> all these indirection changes could have gone in a earlier patch, this 
> adds a lot of changes that make this patch long to review. Same comment 
> for all formats
>
> > +ÂÂÂÂÂÂÂ flac_cfg.ch_cfg = codec->ch_in;
> > +ÂÂÂÂÂÂÂ flac_cfg.sample_rate = codec->sample_rate;
> > ÂÂÂÂÂÂÂÂÂ flac_cfg.stream_info_present = 1;
> > ÂÂÂÂÂÂÂÂÂ flac_cfg.sample_size = flac->sample_size;
> > ÂÂÂÂÂÂÂÂÂ flac_cfg.min_blk_size = flac->min_blk_size;
>
> [...]
...

> > @@ -917,6 +1018,14 @@ static int q6asm_dai_compr_trigger(struct 
> > snd_soc_component *component,
> > ÂÂÂÂÂÂÂÂÂ ret = q6asm_cmd_nowait(prtd->audio_client, prtd->stream_id,
> > ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ CMD_PAUSE);
> > ÂÂÂÂÂÂÂÂÂ break;
> > +ÂÂÂ case SND_COMPR_TRIGGER_NEXT_TRACK:
> > +ÂÂÂÂÂÂÂ prtd->next_track = true;
> > +ÂÂÂÂÂÂÂ prtd->next_track_stream_id = (prtd->stream_id == 1 ? 2 : 1);
>
> it's rather odd, the initialization above uses next_track_stream_id = 0?

Vaild stream ids start from 1, So we are toggling this between 1 and 2.
So when we set next_track_stream_id to 0, that means we have opened the 
new next stream id and is set to prtd->stream_id. This logic is to 
ensure that we are not going to open next stream id twice!

> > +ÂÂÂÂÂÂÂ break;
> > +ÂÂÂ case SND_COMPR_TRIGGER_DRAIN:
> > +ÂÂÂ case SND_COMPR_TRIGGER_PARTIAL_DRAIN:
> > +ÂÂÂÂÂÂÂ prtd->notify_on_drain = true;
> > +ÂÂÂÂÂÂÂ break;
> > ÂÂÂÂÂ default:
> > ÂÂÂÂÂÂÂÂÂ ret = -EINVAL;
> > ÂÂÂÂÂÂÂÂÂ break;
> > diff --git a/sound/soc/qcom/qdsp6/q6asm.h b/sound/soc/qcom/qdsp6/q6asm.h
> > index 69513ac1fa55..a8dddfeb28da 100644
> > --- a/sound/soc/qcom/qdsp6/q6asm.h
> > +++ b/sound/soc/qcom/qdsp6/q6asm.h
> > @@ -34,6 +34,7 @@ enum {
> > Â #define MAX_SESSIONSÂÂÂ 8
> > Â #define NO_TIMESTAMPÂÂÂ 0xFF00
> > Â #define FORMAT_LINEAR_PCMÂÂ 0x0000
> > +#define ASM_LAST_BUFFER_FLAGÂÂÂÂÂÂÂÂÂÂ BIT(30)
> > Â struct q6asm_flac_cfg {
> > ÂÂÂÂÂÂÂÂÂ u32 sample_rate;