2020-06-12 07:51:14

by Shengjiu Wang

Subject: [RFC PATCH v3 4/4] ASoC: fsl_asrc_dma: Fix data copying speed issue with EDMA

With EDMA, there is two dma channels can be used for dev_to_dev,
one is from ASRC, one is from another peripheral (ESAI or SAI).

If we select the dma channel of ASRC, there is an issue for ideal
ratio case, the speed of copy data is faster than sample
frequency, because ASRC output data is very fast in ideal ratio
mode.

So it is reasonable to use the dma channel of Back-End peripheral.
then copying speed of DMA is controlled by data consumption
speed in the peripheral FIFO,

Signed-off-by: Shengjiu Wang <[email protected]>
---
sound/soc/fsl/fsl_asrc_common.h | 2 ++
sound/soc/fsl/fsl_asrc_dma.c | 26 +++++++++++++++-----------
2 files changed, 17 insertions(+), 11 deletions(-)

diff --git a/sound/soc/fsl/fsl_asrc_common.h b/sound/soc/fsl/fsl_asrc_common.h
index 77665b15c8db..7e1c13ca37f1 100644
--- a/sound/soc/fsl/fsl_asrc_common.h
+++ b/sound/soc/fsl/fsl_asrc_common.h
@@ -32,6 +32,7 @@ enum asrc_pair_index {
* @dma_chan: inputer and output DMA channels
* @dma_data: private dma data
* @pos: hardware pointer position
+ * @req_dma_chan: flag to release dev_to_dev chan
* @private: pair private area
*/
struct fsl_asrc_pair {
@@ -45,6 +46,7 @@ struct fsl_asrc_pair {
struct dma_chan *dma_chan[2];
struct imx_dma_data dma_data;
unsigned int pos;
+ bool req_dma_chan;

void *private;
};
diff --git a/sound/soc/fsl/fsl_asrc_dma.c b/sound/soc/fsl/fsl_asrc_dma.c
index d88e6343e0a2..5f01a58f422a 100644
--- a/sound/soc/fsl/fsl_asrc_dma.c
+++ b/sound/soc/fsl/fsl_asrc_dma.c
@@ -233,11 +233,11 @@ static int fsl_asrc_dma_hw_params(struct snd_soc_component *component,

pair->dma_chan[dir] =
dma_request_channel(mask, filter, &pair->dma_data);
+ pair->req_dma_chan = true;
} else {
- if (!be_chan)
- dma_release_channel(tmp_chan);
- pair->dma_chan[dir] =
- asrc->get_dma_channel(pair, dir);
+ pair->dma_chan[dir] = tmp_chan;
+ /* Do not flag to release if we are reusing the Back-End one */
+ pair->req_dma_chan = !be_chan;
}

if (!pair->dma_chan[dir]) {
@@ -276,7 +276,8 @@ static int fsl_asrc_dma_hw_params(struct snd_soc_component *component,
ret = dmaengine_slave_config(pair->dma_chan[dir], &config_be);
if (ret) {
dev_err(dev, "failed to config DMA channel for Back-End\n");
- dma_release_channel(pair->dma_chan[dir]);
+ if (pair->req_dma_chan)
+ dma_release_channel(pair->dma_chan[dir]);
return ret;
}

@@ -288,19 +289,22 @@ static int fsl_asrc_dma_hw_params(struct snd_soc_component *component,
static int fsl_asrc_dma_hw_free(struct snd_soc_component *component,
struct snd_pcm_substream *substream)
{
+ bool tx = substream->stream == SNDRV_PCM_STREAM_PLAYBACK;
struct snd_pcm_runtime *runtime = substream->runtime;
struct fsl_asrc_pair *pair = runtime->private_data;
+ u8 dir = tx ? OUT : IN;

snd_pcm_set_runtime_buffer(substream, NULL);

- if (pair->dma_chan[IN])
- dma_release_channel(pair->dma_chan[IN]);
+ if (pair->dma_chan[!dir])
+ dma_release_channel(pair->dma_chan[!dir]);

- if (pair->dma_chan[OUT])
- dma_release_channel(pair->dma_chan[OUT]);
+ /* release dev_to_dev chan if we aren't reusing the Back-End one */
+ if (pair->dma_chan[dir] && pair->req_dma_chan)
+ dma_release_channel(pair->dma_chan[dir]);

- pair->dma_chan[IN] = NULL;
- pair->dma_chan[OUT] = NULL;
+ pair->dma_chan[!dir] = NULL;
+ pair->dma_chan[dir] = NULL;

return 0;
}
--
2.21.0

2020-06-12 08:14:38

by Nicolin Chen

Subject: Re: [RFC PATCH v3 4/4] ASoC: fsl_asrc_dma: Fix data copying speed issue with EDMA

On Fri, Jun 12, 2020 at 03:37:51PM +0800, Shengjiu Wang wrote:
> With EDMA, there is two dma channels can be used for dev_to_dev,
> one is from ASRC, one is from another peripheral (ESAI or SAI).
>
> If we select the dma channel of ASRC, there is an issue for ideal
> ratio case, the speed of copy data is faster than sample
> frequency, because ASRC output data is very fast in ideal ratio
> mode.
>
> So it is reasonable to use the dma channel of Back-End peripheral.
> then copying speed of DMA is controlled by data consumption
> speed in the peripheral FIFO,
>
> Signed-off-by: Shengjiu Wang <[email protected]>

Reviewed-by: Nicolin Chen <[email protected]>