[RFC_iii/iv 2/2] ASoC: Allow register dailess codecs in machine driver

[RFC_iii/iv 2/2] ASoC: Allow register dailess codecs in machine driver

[Jarkko Nikula]

This makes possible to register dailess codecs in a machine driver. Term
dailess is used here for amplifiers and codecs without DAI or DAI being
unused.

Dailess codecs are registered by letting cpu_dai_name, codec_dai_name and
platform_name to be NULL in struct snd_soc_dev_map. Other fields remain the
same so machine driver can setup paths and widgets for dailess codecs as
usual. Only difference is that the PCM is not created.

Signed-off-by: Jarkko Nikula <jhnikula@gmail.com>
---
 include/sound/soc.h  |    1 +
 sound/soc/soc-core.c |   32 ++++++++++++++++++++++----------
 2 files changed, 23 insertions(+), 10 deletions(-)

diff --git a/include/sound/soc.h b/include/sound/soc.h
index 49e48f0..17f985c 100644
--- a/include/sound/soc.h
+++ b/include/sound/soc.h
@@ -603,6 +603,7 @@ struct snd_soc_pcm_runtime  {
 
 	unsigned int complete:1;
 	unsigned int dev_registered:1;
+	unsigned int dailess:1;
 
 	/* Symmetry data - only valid if symmetry is being enforced */
 	unsigned int rate;
diff --git a/sound/soc/soc-core.c b/sound/soc/soc-core.c
index fa05fa8..b904a20 100644
--- a/sound/soc/soc-core.c
+++ b/sound/soc/soc-core.c
@@ -1217,7 +1217,7 @@ static int soc_bind_dev_map(struct snd_soc_card *card, int num)
 	dev_dbg(card->dev, "binding %s at idx %d\n", dev_map->name, num);
 
 	/* do we already have the CPU DAI for this link ? */
-	if (rtd->cpu_dai) {
+	if (rtd->cpu_dai  || !dev_map->cpu_dai_name) {
 		goto find_codec;
 	}
 	/* no, then find CPU DAI from registered DAIs*/
@@ -1248,6 +1248,9 @@ find_codec:
 			if (!try_module_get(codec->dev->driver->owner))
 				return -ENODEV;
 
+			if (!dev_map->codec_dai_name)
+				goto find_platform;
+
 			/* CODEC found, so find CODEC DAI from registered DAIs from this CODEC*/
 			list_for_each_entry(codec_dai, &dai_list, list) {
 				if (codec->dev == codec_dai->dev &&
@@ -1267,7 +1270,7 @@ find_codec:
 
 find_platform:
 	/* do we already have the CODEC DAI for this link ? */
-	if (rtd->platform) {
+	if (rtd->platform || !dev_map->platform_name) {
 		goto out;
 	}
 	/* no, then find CPU DAI from registered DAIs*/
@@ -1287,8 +1290,12 @@ find_platform:
 	return 0;
 
 out:
+	rtd->dailess = (!dev_map->platform_name && !dev_map->cpu_dai_name &&
+			!dev_map->codec_dai_name);
 	/* mark rtd as complete if we found all 4 of our client devices */
-	if (rtd->codec && rtd->codec_dai && rtd->platform && rtd->cpu_dai) {
+	if (rtd->codec &&
+	    ((rtd->codec_dai && rtd->platform && rtd->cpu_dai) ||
+	     rtd->dailess)) {
 		rtd->complete = 1;
 		card->num_rtd++;
 	}
@@ -1378,19 +1385,21 @@ static int soc_probe_dev_map(struct snd_soc_card *card, int num)
 	dev_dbg(card->dev, "probe %s dai link %d\n", card->name, num);
 
 	/* config components */
-	codec_dai->codec = codec;
 	codec->card = card;
-	cpu_dai->platform = platform;
 	rtd->card = card;
 	rtd->dev.parent = card->dev;
-	codec_dai->card = card;
-	cpu_dai->card = card;
+	if (!rtd->dailess) {
+		codec_dai->codec = codec;
+		codec_dai->card = card;
+		cpu_dai->platform = platform;
+		cpu_dai->card = card;
+	}
 
 	/* set default power off timeout */
 	rtd->pmdown_time = pmdown_time;
 
 	/* probe the cpu_dai */
-	if (!cpu_dai->probed) {
+	if (cpu_dai && !cpu_dai->probed) {
 		if (cpu_dai->driver->probe) {
 			ret = cpu_dai->driver->probe(cpu_dai);
 			if (ret < 0) {
@@ -1425,7 +1434,7 @@ static int soc_probe_dev_map(struct snd_soc_card *card, int num)
 	}
 
 	/* probe the platform */
-	if (!platform->probed) {
+	if (platform && !platform->probed) {
 		if (platform->driver->probe) {
 			ret = platform->driver->probe(platform);
 			if (ret < 0) {
@@ -1440,7 +1449,7 @@ static int soc_probe_dev_map(struct snd_soc_card *card, int num)
 	}
 
 	/* probe the CODEC DAI */
-	if (!codec_dai->probed) {
+	if (codec_dai && !codec_dai->probed) {
 		if (codec_dai->driver->probe) {
 			ret = codec_dai->driver->probe(codec_dai);
 			if (ret < 0) {
@@ -1495,6 +1504,9 @@ static int soc_probe_dev_map(struct snd_soc_card *card, int num)
 	if (ret < 0)
 		printk(KERN_WARNING "asoc: failed to add codec sysfs files\n");
 
+	if (rtd->dailess)
+		return 0;
+
 	/* create the pcm */
 	ret = soc_new_pcm(rtd, num);
 	if (ret < 0) {
-- 
1.7.2.3

Re: [RFC_iii/iv 1/2] ASoC: Rename dai_link as dev_map

[Mark Brown]

On Fri, Oct 29, 2010 at 03:02:59PM +0300, Jarkko Nikula wrote:
> Thus far the struct snd_soc_dai_link is used to tie together platform, codec
> and cpu dai drivers in a machine driver and all of these devices has been
> required for a link.

My main thought with this now that I see the change is that it'd be
easier and probably clearer to add an explicit list of devices rather
than to repurpose the DAI links.  This will probably be simpler since we
can avoid things like having to work out if we've seen and initialised a
device before when it has multiple DAIs involved.

Re: [RFC_iii/iv 1/2] ASoC: Rename dai_link as dev_map

[Jarkko Nikula]

On Fri, 29 Oct 2010 14:22:49 -0700
Mark Brown <broonie@opensource.wolfsonmicro.com> wrote:

> On Fri, Oct 29, 2010 at 03:02:59PM +0300, Jarkko Nikula wrote:
> > Thus far the struct snd_soc_dai_link is used to tie together platform, codec
> > and cpu dai drivers in a machine driver and all of these devices has been
> > required for a link.
> 
> My main thought with this now that I see the change is that it'd be
> easier and probably clearer to add an explicit list of devices rather
> than to repurpose the DAI links.  This will probably be simpler since we
> can avoid things like having to work out if we've seen and initialised a
> device before when it has multiple DAIs involved.

My concern was mostly that we end up having somewhat similar struct
than snd_soc_dai_link and adding pointer to such table in snd_soc_card.
But that's not necessary reduncandy at all if it makes cleanier move to
other cases like codec-codec etc. I'll experiment that in next version.


-- 
Jarkko

Re: [RFC_iii/iv 1/2] ASoC: Rename dai_link as dev_map

[Mark Brown]

On Sun, Oct 31, 2010 at 08:12:15PM +0200, Jarkko Nikula wrote:
> Mark Brown <broonie@opensource.wolfsonmicro.com> wrote:

> > My main thought with this now that I see the change is that it'd be
> > easier and probably clearer to add an explicit list of devices rather
> > than to repurpose the DAI links.  This will probably be simpler since we
> > can avoid things like having to work out if we've seen and initialised a
> > device before when it has multiple DAIs involved.

> My concern was mostly that we end up having somewhat similar struct
> than snd_soc_dai_link and adding pointer to such table in snd_soc_card.

We probably want to be pulling stuff out of the DAI link if anything
(for example, the CODEC is generally going to be implied by the CODEC
DAI).

> But that's not necessary reduncandy at all if it makes cleanier move to
> other cases like codec-codec etc. I'll experiment that in next version.

Please do - I'd rather have a series of small, clear descriptive lists
than a single big list since I think that this will be much clearer and
easier to work with.