[PATCH 0/2] ASoC: atmel-pcm: use generic dmaengine framework

[PATCH 0/2] ASoC: atmel-pcm: use generic dmaengine framework

[Bo Shen]

this patch series is based on next-20130701

Bo Shen (2):
  ASoC: atmel_ssc_dai: move set dma data to startup callback
  ASoC: atmel-pcm: use generic dmaengine framework

 sound/soc/atmel/atmel-pcm-dma.c |  118 ++++++---------------------------------
 sound/soc/atmel/atmel_ssc_dai.c |   33 ++++++-----
 2 files changed, 32 insertions(+), 119 deletions(-)

-- 
1.7.9.5

[PATCH 1/2] ASoC: atmel_ssc_dai: move set dma data to startup callback

[Bo Shen]

move set dma data to startup callback function, if the set dma
data exist in hw_params callback, so the dma data only usable when
call hw_params, if want use it before hw_params callback, it will
cause NULL pointer access oops

Signed-off-by: Bo Shen <voice.shen@atmel.com>
---
 sound/soc/atmel/atmel_ssc_dai.c |   33 ++++++++++++++++-----------------
 1 file changed, 16 insertions(+), 17 deletions(-)

diff --git a/sound/soc/atmel/atmel_ssc_dai.c b/sound/soc/atmel/atmel_ssc_dai.c
index f3fdfa0..6cf9cf1 100644
--- a/sound/soc/atmel/atmel_ssc_dai.c
+++ b/sound/soc/atmel/atmel_ssc_dai.c
@@ -196,15 +196,27 @@ static int atmel_ssc_startup(struct snd_pcm_substream *substream,
 			     struct snd_soc_dai *dai)
 {
 	struct atmel_ssc_info *ssc_p = &ssc_info[dai->id];
-	int dir_mask;
+	struct atmel_pcm_dma_params *dma_params;
+	int dir, dir_mask;
 
 	pr_debug("atmel_ssc_startup: SSC_SR=0x%u\n",
 		ssc_readl(ssc_p->ssc->regs, SR));
 
-	if (substream->stream == SNDRV_PCM_STREAM_PLAYBACK)
+	if (substream->stream == SNDRV_PCM_STREAM_PLAYBACK) {
+		dir = 0;
 		dir_mask = SSC_DIR_MASK_PLAYBACK;
-	else
+	} else {
+		dir = 1;
 		dir_mask = SSC_DIR_MASK_CAPTURE;
+	}
+
+	dma_params = &ssc_dma_params[dai->id][dir];
+	dma_params->ssc = ssc_p->ssc;
+	dma_params->substream = substream;
+
+	ssc_p->dma_params[dir] = dma_params;
+
+	snd_soc_dai_set_dma_data(dai, substream, dma_params);
 
 	spin_lock_irq(&ssc_p->lock);
 	if (ssc_p->dir_mask & dir_mask) {
@@ -325,7 +337,6 @@ static int atmel_ssc_hw_params(struct snd_pcm_substream *substream,
 	struct snd_pcm_hw_params *params,
 	struct snd_soc_dai *dai)
 {
-	struct snd_soc_pcm_runtime *rtd = snd_pcm_substream_chip(substream);
 	int id = dai->id;
 	struct atmel_ssc_info *ssc_p = &ssc_info[id];
 	struct atmel_pcm_dma_params *dma_params;
@@ -344,19 +355,7 @@ static int atmel_ssc_hw_params(struct snd_pcm_substream *substream,
 	else
 		dir = 1;
 
-	dma_params = &ssc_dma_params[id][dir];
-	dma_params->ssc = ssc_p->ssc;
-	dma_params->substream = substream;
-
-	ssc_p->dma_params[dir] = dma_params;
-
-	/*
-	 * The snd_soc_pcm_stream->dma_data field is only used to communicate
-	 * the appropriate DMA parameters to the pcm driver hw_params()
-	 * function.  It should not be used for other purposes
-	 * as it is common to all substreams.
-	 */
-	snd_soc_dai_set_dma_data(rtd->cpu_dai, substream, dma_params);
+	dma_params = ssc_p->dma_params[dir];
 
 	channels = params_channels(params);
 
-- 
1.7.9.5

[alsa-devel] [PATCH 1/2] ASoC: atmel_ssc_dai: move set dma data to startup callback

[Lars-Peter Clausen]

On 07/02/2013 11:04 AM, Bo Shen wrote:
[...]
> +	dma_params->substream = substream;

Unrelated to this patch, but having a substream field in the dma_params seems
to be redundant, since you need to know the substream to get the dma_params in
the first place. Looking at the code it also seems to be unused other than this
assignment. Maybe you can send a follow up patch to remove it.

[...]

[alsa-devel] [PATCH 1/2] ASoC: atmel_ssc_dai: move set dma data to startup callback

[Bo Shen]

Hi Lars-Peter,

On 7/2/2013 23:25, Lars-Peter Clausen wrote:
> On 07/02/2013 11:04 AM, Bo Shen wrote:
> [...]
>> +	dma_params->substream = substream;
>
> Unrelated to this patch, but having a substream field in the dma_params seems
> to be redundant, since you need to know the substream to get the dma_params in
> the first place. Looking at the code it also seems to be unused other than this
> assignment. Maybe you can send a follow up patch to remove it.
>
> [...]
>

Yes, this is redundant. I will remove it in next version patches.

Best Regards,
Bo Shen

[alsa-devel] [PATCH 1/2] ASoC: atmel_ssc_dai: move set dma data to startup callback

[Bo Shen]

Hi Lars,

On 7/3/2013 09:20, Bo Shen wrote:
> Hi Lars-Peter,
>
> On 7/2/2013 23:25, Lars-Peter Clausen wrote:
>> On 07/02/2013 11:04 AM, Bo Shen wrote:
>> [...]
>>> +    dma_params->substream = substream;
>>
>> Unrelated to this patch, but having a substream field in the
>> dma_params seems
>> to be redundant, since you need to know the substream to get the
>> dma_params in
>> the first place. Looking at the code it also seems to be unused other
>> than this
>> assignment. Maybe you can send a follow up patch to remove it.
>>
>> [...]
>>
>
> Yes, this is redundant. I will remove it in next version patches.

Sorry for mistake, this substream parameter need by the interrupt for 
PDC transfer, so, I will keep it.

Best Regards,
Bo Shen

[PATCH 2/2] ASoC: atmel-pcm: use generic dmaengine framework

[Bo Shen]

Align atmel pcm to use ASoC generic dmaengine framework

Signed-off-by: Bo Shen <voice.shen@atmel.com>

---
 sound/soc/atmel/atmel-pcm-dma.c |  118 ++++++---------------------------------
 1 file changed, 16 insertions(+), 102 deletions(-)

diff --git a/sound/soc/atmel/atmel-pcm-dma.c b/sound/soc/atmel/atmel-pcm-dma.c
index 1d38fd0..0f6c701 100644
--- a/sound/soc/atmel/atmel-pcm-dma.c
+++ b/sound/soc/atmel/atmel-pcm-dma.c
@@ -89,138 +89,52 @@ static void atmel_pcm_dma_irq(u32 ssc_sr,
 	}
 }
 
-/*--------------------------------------------------------------------------*\
- * DMAENGINE operations
-\*--------------------------------------------------------------------------*/
-static bool filter(struct dma_chan *chan, void *slave)
-{
-	struct at_dma_slave *sl = slave;
-
-	if (sl->dma_dev == chan->device->dev) {
-		chan->private = sl;
-		return true;
-	} else {
-		return false;
-	}
-}
-
 static int atmel_pcm_configure_dma(struct snd_pcm_substream *substream,
-	struct snd_pcm_hw_params *params, struct atmel_pcm_dma_params *prtd)
+	struct snd_pcm_hw_params *params, struct dma_slave_config *slave_config)
 {
+	struct snd_soc_pcm_runtime *rtd = substream->private_data;
+	struct atmel_pcm_dma_params *prtd;
 	struct ssc_device *ssc;
-	struct dma_chan *dma_chan;
-	struct dma_slave_config slave_config;
 	int ret;
 
+	prtd = snd_soc_dai_get_dma_data(rtd->cpu_dai, substream);
 	ssc = prtd->ssc;
 
-	ret = snd_hwparams_to_dma_slave_config(substream, params,
-			&slave_config);
+	ret = snd_hwparams_to_dma_slave_config(substream, params, slave_config);
 	if (ret) {
 		pr_err("atmel-pcm: hwparams to dma slave configure failed\n");
 		return ret;
 	}
 
 	if (substream->stream == SNDRV_PCM_STREAM_PLAYBACK) {
-		slave_config.dst_addr = (dma_addr_t)ssc->phybase + SSC_THR;
-		slave_config.dst_maxburst = 1;
+		slave_config->dst_addr = (dma_addr_t)ssc->phybase + SSC_THR;
+		slave_config->dst_maxburst = 1;
 	} else {
-		slave_config.src_addr = (dma_addr_t)ssc->phybase + SSC_RHR;
-		slave_config.src_maxburst = 1;
-	}
-
-	dma_chan = snd_dmaengine_pcm_get_chan(substream);
-	if (dmaengine_slave_config(dma_chan, &slave_config)) {
-		pr_err("atmel-pcm: failed to configure dma channel\n");
-		ret = -EBUSY;
-		return ret;
-	}
-
-	return 0;
-}
-
-static int atmel_pcm_hw_params(struct snd_pcm_substream *substream,
-	struct snd_pcm_hw_params *params)
-{
-	struct snd_soc_pcm_runtime *rtd = substream->private_data;
-	struct atmel_pcm_dma_params *prtd;
-	struct ssc_device *ssc;
-	struct at_dma_slave *sdata = NULL;
-	int ret;
-
-	snd_pcm_set_runtime_buffer(substream, &substream->dma_buffer);
-
-	prtd = snd_soc_dai_get_dma_data(rtd->cpu_dai, substream);
-	ssc = prtd->ssc;
-	if (ssc->pdev)
-		sdata = ssc->pdev->dev.platform_data;
-
-	ret = snd_dmaengine_pcm_open_request_chan(substream, filter, sdata);
-	if (ret) {
-		pr_err("atmel-pcm: dmaengine pcm open failed\n");
-		return -EINVAL;
-	}
-
-	ret = atmel_pcm_configure_dma(substream, params, prtd);
-	if (ret) {
-		pr_err("atmel-pcm: failed to configure dmai\n");
-		goto err;
+		slave_config->src_addr = (dma_addr_t)ssc->phybase + SSC_RHR;
+		slave_config->src_maxburst = 1;
 	}
 
 	prtd->dma_intr_handler = atmel_pcm_dma_irq;
 
 	return 0;
-err:
-	snd_dmaengine_pcm_close_release_chan(substream);
-	return ret;
 }
 
-static int atmel_pcm_dma_prepare(struct snd_pcm_substream *substream)
-{
-	struct snd_soc_pcm_runtime *rtd = substream->private_data;
-	struct atmel_pcm_dma_params *prtd;
-
-	prtd = snd_soc_dai_get_dma_data(rtd->cpu_dai, substream);
-
-	ssc_writex(prtd->ssc->regs, SSC_IER, prtd->mask->ssc_error);
-	ssc_writex(prtd->ssc->regs, SSC_CR, prtd->mask->ssc_enable);
-
-	return 0;
-}
-
-static int atmel_pcm_open(struct snd_pcm_substream *substream)
-{
-	snd_soc_set_runtime_hwparams(substream, &atmel_pcm_dma_hardware);
-
-	return 0;
-}
-
-static struct snd_pcm_ops atmel_pcm_ops = {
-	.open		= atmel_pcm_open,
-	.close		= snd_dmaengine_pcm_close_release_chan,
-	.ioctl		= snd_pcm_lib_ioctl,
-	.hw_params	= atmel_pcm_hw_params,
-	.prepare	= atmel_pcm_dma_prepare,
-	.trigger	= snd_dmaengine_pcm_trigger,
-	.pointer	= snd_dmaengine_pcm_pointer_no_residue,
-	.mmap		= atmel_pcm_mmap,
-};
-
-static struct snd_soc_platform_driver atmel_soc_platform = {
-	.ops		= &atmel_pcm_ops,
-	.pcm_new	= atmel_pcm_new,
-	.pcm_free	= atmel_pcm_free,
+static const struct snd_dmaengine_pcm_config atmel_dmaengine_pcm_config = {
+	.prepare_slave_config = atmel_pcm_configure_dma,
+	.pcm_hardware = &atmel_pcm_dma_hardware,
+	.prealloc_buffer_size = ATMEL_SSC_DMABUF_SIZE,
 };
 
 int atmel_pcm_dma_platform_register(struct device *dev)
 {
-	return snd_soc_register_platform(dev, &atmel_soc_platform);
+	return snd_dmaengine_pcm_register(dev, &atmel_dmaengine_pcm_config,
+			SND_DMAENGINE_PCM_FLAG_NO_RESIDUE);
 }
 EXPORT_SYMBOL(atmel_pcm_dma_platform_register);
 
 void atmel_pcm_dma_platform_unregister(struct device *dev)
 {
-	snd_soc_unregister_platform(dev);
+	snd_dmaengine_pcm_unregister(dev);
 }
 EXPORT_SYMBOL(atmel_pcm_dma_platform_unregister);
 
-- 
1.7.9.5

[alsa-devel] [PATCH 2/2] ASoC: atmel-pcm: use generic dmaengine framework

[Lars-Peter Clausen]

On 07/02/2013 11:04 AM, Bo Shen wrote:
[...]
> -/*--------------------------------------------------------------------------*\
> - * DMAENGINE operations
> -\*--------------------------------------------------------------------------*/
> -static bool filter(struct dma_chan *chan, void *slave)
> -{
> -	struct at_dma_slave *sl = slave;
> -
> -	if (sl->dma_dev == chan->device->dev) {
> -		chan->private = sl;
> -		return true;
> -	} else {
> -		return false;
> -	}
> -}

I'm wondering why is filtering no longer required?

[...]
> +		slave_config->dst_addr = (dma_addr_t)ssc->phybase + SSC_THR;
[...]
> +		slave_config->src_addr = (dma_addr_t)ssc->phybase + SSC_RHR;

Change the type of phybase to dma_addr_t

[...]
>  }
>  
> -static int atmel_pcm_dma_prepare(struct snd_pcm_substream *substream)
> -{
> -	struct snd_soc_pcm_runtime *rtd = substream->private_data;
> -	struct atmel_pcm_dma_params *prtd;
> -
> -	prtd = snd_soc_dai_get_dma_data(rtd->cpu_dai, substream);
> -
> -	ssc_writex(prtd->ssc->regs, SSC_IER, prtd->mask->ssc_error);
> -	ssc_writex(prtd->ssc->regs, SSC_CR, prtd->mask->ssc_enable);

These writes are also completely lost

> -
> -	return 0;
> -}

I think it would also be good to get rid of any references to the ssc struct
from the DMA driver and move them over to the DAI driver.

- Lars

[alsa-devel] [PATCH 2/2] ASoC: atmel-pcm: use generic dmaengine framework

[Bo Shen]

Hi Lars-Peter,

On 7/2/2013 23:34, Lars-Peter Clausen wrote:
> On 07/02/2013 11:04 AM, Bo Shen wrote:
> [...]
>> -/*--------------------------------------------------------------------------*\
>> - * DMAENGINE operations
>> -\*--------------------------------------------------------------------------*/
>> -static bool filter(struct dma_chan *chan, void *slave)
>> -{
>> -	struct at_dma_slave *sl = slave;
>> -
>> -	if (sl->dma_dev == chan->device->dev) {
>> -		chan->private = sl;
>> -		return true;
>> -	} else {
>> -		return false;
>> -	}
>> -}
>
> I'm wondering why is filtering no longer required?

Now this is handled by at_dma_filter in at_hdmac driver. More 
information, you can reference: <driver/dma/at_hdmac.c> file

> [...]
>> +		slave_config->dst_addr = (dma_addr_t)ssc->phybase + SSC_THR;
> [...]
>> +		slave_config->src_addr = (dma_addr_t)ssc->phybase + SSC_RHR;
>
> Change the type of phybase to dma_addr_t

It should be:
slave_config->dst_addr = (dma_addr_t)(ssc->phybase + SSC_THR);
slave_config->src_addr = (dma_addr_t)(ssc->phybase + SSC_RHR);

> [...]
>>   }
>>
>> -static int atmel_pcm_dma_prepare(struct snd_pcm_substream *substream)
>> -{
>> -	struct snd_soc_pcm_runtime *rtd = substream->private_data;
>> -	struct atmel_pcm_dma_params *prtd;
>> -
>> -	prtd = snd_soc_dai_get_dma_data(rtd->cpu_dai, substream);
>> -
>> -	ssc_writex(prtd->ssc->regs, SSC_IER, prtd->mask->ssc_error);
>> -	ssc_writex(prtd->ssc->regs, SSC_CR, prtd->mask->ssc_enable);
>
> These writes are also completely lost

This will be moved to DAI driver startup callback function.

>> -
>> -	return 0;
>> -}
>
> I think it would also be good to get rid of any references to the ssc struct
> from the DMA driver and move them over to the DAI driver.

I will try to implement this in next version patch.

Thanks.

> - Lars
>

Best Regards,
Bo Shen

[alsa-devel] [PATCH 2/2] ASoC: atmel-pcm: use generic dmaengine framework

[Lars-Peter Clausen]

On 07/03/2013 03:33 AM, Bo Shen wrote:
> Hi Lars-Peter,
> 
> On 7/2/2013 23:34, Lars-Peter Clausen wrote:
>> On 07/02/2013 11:04 AM, Bo Shen wrote:
>> [...]
>>> -/*--------------------------------------------------------------------------*\
>>> - * DMAENGINE operations
>>> -\*--------------------------------------------------------------------------*/
>>> -static bool filter(struct dma_chan *chan, void *slave)
>>> -{
>>> -    struct at_dma_slave *sl = slave;
>>> -
>>> -    if (sl->dma_dev == chan->device->dev) {
>>> -        chan->private = sl;
>>> -        return true;
>>> -    } else {
>>> -        return false;
>>> -    }
>>> -}
>>
>> I'm wondering why is filtering no longer required?
> 
> Now this is handled by at_dma_filter in at_hdmac driver. More information, you
> can reference: <driver/dma/at_hdmac.c> file

Ok, so the DMA channel is now supplied via devicetree. This is really a change
that should be listed in the changelog.

> 
>> [...]
>>> +        slave_config->dst_addr = (dma_addr_t)ssc->phybase + SSC_THR;
>> [...]
>>> +        slave_config->src_addr = (dma_addr_t)ssc->phybase + SSC_RHR;
>>
>> Change the type of phybase to dma_addr_t
> 
> It should be:
> slave_config->dst_addr = (dma_addr_t)(ssc->phybase + SSC_THR);
> slave_config->src_addr = (dma_addr_t)(ssc->phybase + SSC_RHR);

The type of phybase in the ssc_device struct should be changed from
resource_size_t to dma_addr_t.

[alsa-devel] [PATCH 2/2] ASoC: atmel-pcm: use generic dmaengine framework

[Bo Shen]

Hi Lars,

On 7/3/2013 15:36, Lars-Peter Clausen wrote:
>>>> +        slave_config->dst_addr = (dma_addr_t)ssc->phybase + SSC_THR;
>>> >>[...]
>>>> >>>+        slave_config->src_addr = (dma_addr_t)ssc->phybase + SSC_RHR;
>>> >>
>>> >>Change the type of phybase to dma_addr_t
>> >
>> >It should be:
>> >slave_config->dst_addr = (dma_addr_t)(ssc->phybase + SSC_THR);
>> >slave_config->src_addr = (dma_addr_t)(ssc->phybase + SSC_RHR);
> The type of phybase in the ssc_device struct should be changed from
> resource_size_t to dma_addr_t.

Actually, the phybase in ssc_device structure is resource_size_t which 
used to access the ssc device register.
Here, we use ssc register as dma source address or destination address, 
so we convert it to dma_addr_t type. So, not need to change, am I right?

Best Regards,
Bo Shen

[alsa-devel] [PATCH 2/2] ASoC: atmel-pcm: use generic dmaengine framework

[Lars-Peter Clausen]

On 07/03/2013 09:57 AM, Bo Shen wrote:
> Hi Lars,
> 
> On 7/3/2013 15:36, Lars-Peter Clausen wrote:
>>>>> +        slave_config->dst_addr = (dma_addr_t)ssc->phybase + SSC_THR;
>>>> >>[...]
>>>>> >>>+        slave_config->src_addr = (dma_addr_t)ssc->phybase + SSC_RHR;
>>>> >>
>>>> >>Change the type of phybase to dma_addr_t
>>> >
>>> >It should be:
>>> >slave_config->dst_addr = (dma_addr_t)(ssc->phybase + SSC_THR);
>>> >slave_config->src_addr = (dma_addr_t)(ssc->phybase + SSC_RHR);
>> The type of phybase in the ssc_device struct should be changed from
>> resource_size_t to dma_addr_t.
> 
> Actually, the phybase in ssc_device structure is resource_size_t which used to
> access the ssc device register.
> Here, we use ssc register as dma source address or destination address, so we
> convert it to dma_addr_t type. So, not need to change, am I right?

Isn't 'regs' used to access the registers? As far as I can see phybase is only
used in the audio DMA driver.

- Lars

[alsa-devel] [PATCH 2/2] ASoC: atmel-pcm: use generic dmaengine framework

[Bo Shen]

Hi Lars,

On 7/3/2013 16:08, Lars-Peter Clausen wrote:
> On 07/03/2013 09:57 AM, Bo Shen wrote:
>> Hi Lars,
>>
>> On 7/3/2013 15:36, Lars-Peter Clausen wrote:
>>>>>> +        slave_config->dst_addr = (dma_addr_t)ssc->phybase + SSC_THR;
>>>>>>> [...]
>>>>>>>>> +        slave_config->src_addr = (dma_addr_t)ssc->phybase + SSC_RHR;
>>>>>>>
>>>>>>> Change the type of phybase to dma_addr_t
>>>>>
>>>>> It should be:
>>>>> slave_config->dst_addr = (dma_addr_t)(ssc->phybase + SSC_THR);
>>>>> slave_config->src_addr = (dma_addr_t)(ssc->phybase + SSC_RHR);
>>> The type of phybase in the ssc_device struct should be changed from
>>> resource_size_t to dma_addr_t.
>>
>> Actually, the phybase in ssc_device structure is resource_size_t which used to
>> access the ssc device register.
>> Here, we use ssc register as dma source address or destination address, so we
>> convert it to dma_addr_t type. So, not need to change, am I right?
>
> Isn't 'regs' used to access the registers? As far as I can see phybase is only
> used in the audio DMA driver.

Thanks for point out this. I will change the phybase type from 
resource_size_t to dma_addr_t.

> - Lars
>

Best Regards,
Bo Shen