Re: [PATCH] ASoC: meson: Use managed DMA buffer allocation

[Lars-Peter Clausen <lars@xxxxxxxxxx>]

On 12/18/20 5:28 PM, Jerome Brunet wrote:
> On Fri 18 Dec 2020 at 16:45, Lars-Peter Clausen <lars@xxxxxxxxxx> wrote:
>
> > Using a managed buffer will pre-allocate the buffer using
> > snd_pcm_lib_preallocate_pages() and automatically free it when the PCM is
> > destroyed.
> >
> > In addition it will call snd_pcm_lib_malloc_pages() before the driver's
> > hw_params() callback and snd_pcm_lib_free_pages() after the driver's
> > hw_free() callback.
> >
> > This slightly reduces the boilerplate code of the driver.
> >
> > Signed-off-by: Lars-Peter Clausen <lars@xxxxxxxxxx>
> > ---
> >   sound/soc/meson/aiu-fifo-i2s.c   |  1 -
> >   sound/soc/meson/aiu-fifo-spdif.c |  1 -
> >   sound/soc/meson/aiu-fifo.c       | 18 ++----------------
> >   3 files changed, 2 insertions(+), 18 deletions(-)
> >
> > diff --git a/sound/soc/meson/aiu-fifo-i2s.c b/sound/soc/meson/aiu-fifo-i2s.c
> > index d91b0d874342..2388a2d0b3a6 100644
> > --- a/sound/soc/meson/aiu-fifo-i2s.c
> > +++ b/sound/soc/meson/aiu-fifo-i2s.c
> > @@ -124,7 +124,6 @@ const struct snd_soc_dai_ops aiu_fifo_i2s_dai_ops = {
> >   	.trigger	= aiu_fifo_i2s_trigger,
> >   	.prepare	= aiu_fifo_i2s_prepare,
> >   	.hw_params	= aiu_fifo_i2s_hw_params,
> > -	.hw_free	= aiu_fifo_hw_free,
> >   	.startup	= aiu_fifo_startup,
> >   	.shutdown	= aiu_fifo_shutdown,
> >   };
> > diff --git a/sound/soc/meson/aiu-fifo-spdif.c b/sound/soc/meson/aiu-fifo-spdif.c
> > index 44eb6faacf44..2fb30f89bf7a 100644
> > --- a/sound/soc/meson/aiu-fifo-spdif.c
> > +++ b/sound/soc/meson/aiu-fifo-spdif.c
> > @@ -158,7 +158,6 @@ const struct snd_soc_dai_ops aiu_fifo_spdif_dai_ops = {
> >   	.trigger	= fifo_spdif_trigger,
> >   	.prepare	= fifo_spdif_prepare,
> >   	.hw_params	= fifo_spdif_hw_params,
> > -	.hw_free	= aiu_fifo_hw_free,
> >   	.startup	= aiu_fifo_startup,
> >   	.shutdown	= aiu_fifo_shutdown,
> >   };
> > diff --git a/sound/soc/meson/aiu-fifo.c b/sound/soc/meson/aiu-fifo.c
> > index aa88aae8e517..4ad23267cace 100644
> > --- a/sound/soc/meson/aiu-fifo.c
> > +++ b/sound/soc/meson/aiu-fifo.c
> > @@ -99,11 +99,6 @@ int aiu_fifo_hw_params(struct snd_pcm_substream *substream,
> >   	struct snd_soc_component *component = dai->component;
> >   	struct aiu_fifo *fifo = dai->playback_dma_data;
> >   	dma_addr_t end;
> > -	int ret;
> > -
> > -	ret = snd_pcm_lib_malloc_pages(substream, params_buffer_bytes(params));
> > -	if (ret < 0)
> > -		return ret;
> >   
> >   	/* Setup the fifo boundaries */
> >   	end = runtime->dma_addr + runtime->dma_bytes - fifo->fifo_block;
> > @@ -124,12 +119,6 @@ int aiu_fifo_hw_params(struct snd_pcm_substream *substream,
> >   	return 0;
> >   }
> >   
> > -int aiu_fifo_hw_free(struct snd_pcm_substream *substream,
> > -		     struct snd_soc_dai *dai)
> > -{
> > -	return snd_pcm_lib_free_pages(substream);
> > -}
> > -
> >   static irqreturn_t aiu_fifo_isr(int irq, void *dev_id)
> >   {
> >   	struct snd_pcm_substream *playback = dev_id;
> > @@ -187,15 +176,12 @@ void aiu_fifo_shutdown(struct snd_pcm_substream *substream,
> >   int aiu_fifo_pcm_new(struct snd_soc_pcm_runtime *rtd,
> >   		     struct snd_soc_dai *dai)
> >   {
> > -	struct snd_pcm_substream *substream =
> > -		rtd->pcm->streams[SNDRV_PCM_STREAM_PLAYBACK].substream;
> >   	struct snd_card *card = rtd->card->snd_card;
> >   	struct aiu_fifo *fifo = dai->playback_dma_data;
> >   	size_t size = fifo->pcm->buffer_bytes_max;
> >   
> > -	snd_pcm_lib_preallocate_pages(substream,
> > -				      SNDRV_DMA_TYPE_DEV,
> > -				      card->dev, size, size);
> > +	snd_pcm_set_managed_buffer_all(rtd->pcm, SNDRV_DMA_TYPE_DEV,
> > +				       card->dev, size, size);
> Hi Lars-Peter,
>
> These FIFOs only do playback so to avoid wasting memory
> s/snd_pcm_set_managed_buffer_all/snd_pcm_set_managed_buffer ?

snd_pcm_set_managed_buffer_all() will skip substreams that do not exist. E.g. if the there is not capture support it wont allocate memory for it.

To keep things simple I prefer snd_pcm_set_managed_buffer_all(). snd_pcm_set_managed_buffer() only makes sense if you have a different DMA device for capture and playback or you want different buffer sizes.