Skip to contentSkip to conversation

ASoC: Fix dapm_is_shared_kcontrol so everything isn't shared

PATCH In review Stephen Warren 2 participants 6 messages May 24, 2011 – May 26, 2011

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 ASoC: Fix dapm_is_shared_kcontrol so everything isn't shared

Commit af46800 ("ASoC: Implement mux control sharing") introduced function dapm_is_shared_kcontrol.

When this function returns true, the naming of DAPM controls is derived from the kcontrol_new. Otherwise, the name comes from the widget (and possibly a widget's naming prefix).

A bug in the implementation of dapm_is_shared_kcontrol made it return 1 in all cases. Hence, that commit caused a change in control naming for all controls instead of just shared controls.

Specifically, a control is always considered shared because it is always compared against itself. Solve this by never comparing against the widget containing the control being created.

I tested that with the Tegra WM8903 driver:
* Shared is now mostly 0 as expected, and sometimes 1.
* The expected controls are still generated after this change.

Howwever, I don't have any systems that have a widget/control naming prefix, so I can't test that aspect.

---
 sound/soc/soc-dapm.c |    5 ++++-
 1 files changed, 4 insertions(+), 1 deletions(-)
sound/soc/soc-dapm.c +4 -1
index 456617e..5397699 100644
@@ -325,6 +325,7 @@ static int dapm_connect_mixer(struct snd_soc_dapm_context *dapm,
 }
 
 static int dapm_is_shared_kcontrol(struct snd_soc_dapm_context *dapm,
+	struct snd_soc_dapm_widget *kcontrolw,
 	const struct snd_kcontrol_new *kcontrol_new,
 	struct snd_kcontrol **kcontrol)
 {
@@ -334,6 +335,8 @@ static int dapm_is_shared_kcontrol(struct snd_soc_dapm_context *dapm,
 	*kcontrol = NULL;
 
 	list_for_each_entry(w, &dapm->card->widgets, list) {
+		if (w == kcontrolw)
+			continue;
 		for (i = 0; i < w->num_kcontrols; i++) {
 			if (&w->kcontrol_news[i] == kcontrol_new) {
 				if (w->kcontrols)
@@ -468,7 +471,7 @@ static int dapm_new_mux(struct snd_soc_dapm_context *dapm,
 		return -EINVAL;
 	}
 
-	shared = dapm_is_shared_kcontrol(dapm, &w->kcontrol_news[0],
+	shared = dapm_is_shared_kcontrol(dapm, w, &w->kcontrol_news[0],
 					 &kcontrol);
 	if (kcontrol) {
 		wlist = kcontrol->private_data;
Signature · 2 lines
1.7.0.4
nvpublic
#1 lore raw
Quoted 42 lines · On Tue, 24 May 2011 17:12:01 -0600 Stephen Warren <[email protected]> wrote:

Commit af46800 ("ASoC: Implement mux control sharing") introduced function dapm_is_shared_kcontrol.

When this function returns true, the naming of DAPM controls is derived from the kcontrol_new. Otherwise, the name comes from the widget (and possibly a widget's naming prefix).

A bug in the implementation of dapm_is_shared_kcontrol made it return 1 in all cases. Hence, that commit caused a change in control naming for all controls instead of just shared controls.

Specifically, a control is always considered shared because it is always compared against itself. Solve this by never comparing against the widget containing the control being created.

I tested that with the Tegra WM8903 driver:
* Shared is now mostly 0 as expected, and sometimes 1.
* The expected controls are still generated after this change.

Howwever, I don't have any systems that have a widget/control naming prefix, so I can't test that aspect.

---
 sound/soc/soc-dapm.c |    5 ++++-
 1 files changed, 4 insertions(+), 1 deletions(-)
sound/soc/soc-dapm.c +1 -0
index 456617e..5397699 100644
@@ -325,6 +325,7 @@ static int dapm_connect_mixer(struct snd_soc_dapm_context *dapm,
 }
 
 static int dapm_is_shared_kcontrol(struct snd_soc_dapm_context *dapm,
+	struct snd_soc_dapm_widget *kcontrolw,
 	const struct snd_kcontrol_new *kcontrol_new,
 	struct snd_kcontrol **kcontrol)
 {

sound/soc/soc-dapm.c static int dapm_is_shared_kcontrol(struct snd_soc_dapm_context *dapm, · line 339

Quoted 5 lines
sound/soc/soc-dapm.c +2 -0
@@ -334,6 +335,8 @@ static int dapm_is_shared_kcontrol(struct snd_soc_dapm_context *dapm,
 	*kcontrol = NULL;
 
 	list_for_each_entry(w, &dapm->card->widgets, list) {
+		if (w == kcontrolw)
+			continue;

Hmm.. it seems this test is not enough with tlv320aic3x.c. It still classifies a lot of kcontrols to be shared. Will keep hunting.

Signature · 1 line
Jarkko
#2 lore raw

Hi Stephen

Quoted 22 lines · On Wed, 25 May 2011 10:01:55 +0300 Jarkko Nikula <[email protected]> wrote:
Quoted 19 lines · On Tue, 24 May 2011 17:12:01 -0600 Stephen Warren <[email protected]> wrote:

Commit af46800 ("ASoC: Implement mux control sharing") introduced function dapm_is_shared_kcontrol.

When this function returns true, the naming of DAPM controls is derived from the kcontrol_new. Otherwise, the name comes from the widget (and possibly a widget's naming prefix).

A bug in the implementation of dapm_is_shared_kcontrol made it return 1 in all cases. Hence, that commit caused a change in control naming for all controls instead of just shared controls.

Specifically, a control is always considered shared because it is always compared against itself. Solve this by never comparing against the widget containing the control being created.

I tested that with the Tegra WM8903 driver:
* Shared is now mostly 0 as expected, and sometimes 1.
* The expected controls are still generated after this change.

Quoted 5 lines
Quoted 5 lines

Howwever, I don't have any systems that have a widget/control naming prefix, so I can't test that aspect.

I would rather see here Reported-by/Tested-by as I don't feel that I broke here something :-)

Quoted 3 lines

Hmm.. it seems this test is not enough with tlv320aic3x.c. It still classifies a lot of kcontrols to be shared. Will keep hunting.

I managed to find why this didn't work with tlv320aic3x.c on rx51.c.

This triggered out that "[Left | Right] Line1[L | R]" Muxes in tlv320aic3x were pointing to same mux controls which is wrong since there are separate registers for them. I prepare a fix for this hopefully tomorrow.

Then we must skip test for widgets that are in different dapm context, i.e. we don't want to share mux controls if they are in different device instance. Does it work for you if you add the dapm test below to your patch?

@@ -334,6 +335,8 @@ static int dapm_is_shared_kcontrol(struct snd_soc_dapm_context *dapm, *kcontrol = NULL;

list_for_each_entry(w, &dapm->card->widgets, list) {
+ if (w == kcontrolw || w->dapm != kcontrolw->dapm)
+ continue;

Signature · 1 line
Jarkko
#3 lore raw

Jarkko Nikula wrote Wednesday, May 25, 2011 8:52 AM:

Quoted 24 lines · On Wed, 25 May 2011 10:01:55 +0300 Jarkko Nikula <[email protected]> wrote:
Quoted 23 lines
Quoted 19 lines · On Tue, 24 May 2011 17:12:01 -0600 Stephen Warren <[email protected]> wrote:

Commit af46800 ("ASoC: Implement mux control sharing") introduced function dapm_is_shared_kcontrol.

When this function returns true, the naming of DAPM controls is derived from the kcontrol_new. Otherwise, the name comes from the widget (and possibly a widget's naming prefix).

A bug in the implementation of dapm_is_shared_kcontrol made it return 1 in all cases. Hence, that commit caused a change in control naming for all controls instead of just shared controls.

Specifically, a control is always considered shared because it is always compared against itself. Solve this by never comparing against the widget containing the control being created.

I tested that with the Tegra WM8903 driver:
* Shared is now mostly 0 as expected, and sometimes 1.
* The expected controls are still generated after this change.

Quoted 8 lines
Quoted 5 lines
Quoted 5 lines

Howwever, I don't have any systems that have a widget/control naming prefix, so I can't test that aspect.

I would rather see here Reported-by/Tested-by as I don't feel that I broke here something :-)

OK, I can change it to Tested-by, but what I meant by that (admittedly non-standard) tag was simply that you'd pointed out what the basic problem was, i.e. done the work to determine the root-cause, not that you were the root cause!

Thanks for testing.

Quoted 10 lines
Quoted 2 lines

Hmm.. it seems this test is not enough with tlv320aic3x.c. It still classifies a lot of kcontrols to be shared. Will keep hunting.

I managed to find why this didn't work with tlv320aic3x.c on rx51.c.

This triggered out that "[Left | Right] Line1[L | R]" Muxes in tlv320aic3x were pointing to same mux controls which is wrong since there are separate registers for them. I prepare a fix for this hopefully tomorrow.

Quoted 4 lines

Then we must skip test for widgets that are in different dapm context, i.e. we don't want to share mux controls if they are in different device instance. Does it work for you if you add the dapm test below to your patch?

I suppose that's technically correct, and I'm fine adding this to the patch. However, I'm surprised that change is needed; do we actually have a situation where two widgets in different DAPM contexts point at the same control definition? It seems like that'd be a bug; why would there be two controls pointing at the same register field, yet in different DAPM contexts?

Quoted 6 lines
sound/soc/soc-dapm.c +2 -0
@@ -334,6 +335,8 @@ static int dapm_is_shared_kcontrol(struct
 snd_soc_dapm_context *dapm, *kcontrol = NULL;
 
 	list_for_each_entry(w, &dapm->card->widgets, list) {
+		if (w == kcontrolw || w->dapm != kcontrolw->dapm)
+			continue;
Signature · 1 line
nvpublic
#4 lore raw
Quoted 5 lines · On Wed, 25 May 2011 08:24:59 -0700 Stephen Warren <[email protected]> wrote:
Quoted 4 lines

Then we must skip test for widgets that are in different dapm context, i.e. we don't want to share mux controls if they are in different device instance. Does it work for you if you add the dapm test below to your patch?

Quoted 7 lines

I suppose that's technically correct, and I'm fine adding this to the patch. However, I'm surprised that change is needed; do we actually have a situation where two widgets in different DAPM contexts point at the same control definition? It seems like that'd be a bug; why would there be two controls pointing at the same register field, yet in different DAPM contexts?

Yes, I noticed this can happen when you have multiple same codecs in a system. Let's assume we have multiple wm8903 codecs and we have these in the driver:

static const struct snd_kcontrol_new adcinput_mux =

	SOC_DAPM_ENUM("ADC Input", adcinput_enum);

static const struct snd_soc_dapm_widget wm8903_dapm_widgets[] = {
SND_SOC_DAPM_MUX("Left ADC Input", SND_SOC_NOPM, 0, 0, &adcinput_mux),
};

Even the core duplicates widgets for each codec instance in this example, they still point to same adcinput_mux and thus dapm_is_shared_kcontrol() miss detects them as shared if we don't skip different dapm context.

Signature · 1 line
Jarkko
#5 lore raw

Jarkko Nikula wrote at Thursday, May 26, 2011 12:58 AM:

Quoted 13 lines · On Wed, 25 May 2011 08:24:59 -0700 Stephen Warren <[email protected]> wrote:
Quoted 11 lines
Quoted 4 lines

Then we must skip test for widgets that are in different dapm context, i.e. we don't want to share mux controls if they are in different device instance. Does it work for you if you add the dapm test below to your patch?

I suppose that's technically correct, and I'm fine adding this to the patch. However, I'm surprised that change is needed; do we actually have a situation where two widgets in different DAPM contexts point at the same control definition? It seems like that'd be a bug; why would there be two controls pointing at the same register field, yet in different DAPM contexts?

Quoted 3 lines

Yes, I noticed this can happen when you have multiple same codecs in a system. Let's assume we have multiple wm8903 codecs and we have these in the driver:

Yes, of course, that makes perfect sense. I'll rev the patch as you suggested.

Quoted 11 lines

static const struct snd_kcontrol_new adcinput_mux =

	SOC_DAPM_ENUM("ADC Input", adcinput_enum);

static const struct snd_soc_dapm_widget wm8903_dapm_widgets[] = {
SND_SOC_DAPM_MUX("Left ADC Input", SND_SOC_NOPM, 0, 0, &adcinput_mux),
};

Even the core duplicates widgets for each codec instance in this example, they still point to same adcinput_mux and thus dapm_is_shared_kcontrol() miss detects them as shared if we don't skip different dapm context.

Signature · 1 line
nvpublic