Skip to contentSkip to conversation

ASoC: Fix passing platform_data to ac97 bus users and fix a leak.

PATCH In review Graham Gower 3 participants 6 messages Mar 24, 2010 – Mar 25, 2010

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

PATCH 1/1 ASoC: Fix passing platform_data to ac97 bus users and fix a leak.

snd_soc_new_ac97_codec() allocates for codec->ac97, snd_soc_new_pcms() then sets the platform_data. However, snd_ac97_mixer() overwrites codec->ac97 with its own allocated struct snd_ac97.

---
 sound/soc/codecs/ac97.c |   12 ++++++++++++
 1 files changed, 12 insertions(+), 0 deletions(-)
sound/soc/codecs/ac97.c +12 -0
index a1bbe16..64ec797 100644
@@ -80,9 +80,11 @@ static int ac97_write(struct snd_soc_codec *codec, unsigned int reg,
 static int ac97_soc_probe(struct platform_device *pdev)
 {
 	struct snd_soc_device *socdev = platform_get_drvdata(pdev);
+	struct snd_soc_card *card = socdev->card;
 	struct snd_soc_codec *codec;
 	struct snd_ac97_bus *ac97_bus;
 	struct snd_ac97_template ac97_template;
+	int i;
 	int ret = 0;
 
 	printk(KERN_INFO "AC97 SoC Audio Codec %s\n", AC97_VERSION);
@@ -118,11 +120,21 @@ static int ac97_soc_probe(struct platform_device *pdev)
 	if (ret < 0)
 		goto bus_err;
 
+	/* free the ac97 here so that we don't leak it in snd_ac97_mixer */
+	snd_soc_free_ac97_codec(codec);
+
 	memset(&ac97_template, 0, sizeof(struct snd_ac97_template));
 	ret = snd_ac97_mixer(ac97_bus, &ac97_template, &codec->ac97);
 	if (ret < 0)
 		goto bus_err;
 
+	for (i = 0; i < card->num_links; i++) {
+		if (card->dai_link[i].codec_dai->ac97_control) {
+			snd_ac97_dev_add_pdata(codec->ac97,
+				card->dai_link[i].cpu_dai->ac97_pdata);
+		}
+	}
+
 	return 0;
 
 bus_err:
Signature · 1 line
1.6.4
#1 lore raw
Quoted 26 lines · Graham Gower

snd_soc_new_ac97_codec() allocates for codec->ac97, snd_soc_new_pcms() then sets the platform_data. However, snd_ac97_mixer() overwrites codec->ac97 with its own allocated struct snd_ac97.

---
 sound/soc/codecs/ac97.c |   12 ++++++++++++
 1 files changed, 12 insertions(+), 0 deletions(-)
sound/soc/codecs/ac97.c +2 -0
index a1bbe16..64ec797 100644
@@ -80,9 +80,11 @@ static int ac97_write(struct snd_soc_codec *codec, unsigned int reg,
 static int ac97_soc_probe(struct platform_device *pdev)
 {
 	struct snd_soc_device *socdev = platform_get_drvdata(pdev);
+	struct snd_soc_card *card = socdev->card;
 	struct snd_soc_codec *codec;
 	struct snd_ac97_bus *ac97_bus;
 	struct snd_ac97_template ac97_template;
+	int i;
 	int ret = 0;
 
 	printk(KERN_INFO "AC97 SoC Audio Codec %s\n", AC97_VERSION);

sound/soc/codecs/ac97.c static int ac97_soc_probe(struct platform_device *pdev) · line 125

Quoted 6 lines
sound/soc/codecs/ac97.c +3 -0
@@ -118,11 +120,21 @@ static int ac97_soc_probe(struct platform_device *pdev)
 	if (ret < 0)
 		goto bus_err;
 
+	/* free the ac97 here so that we don't leak it in snd_ac97_mixer */
+	snd_soc_free_ac97_codec(codec);
+

Can you try removing the call to snd_soc_new_ac97_codec() instead. I've just had a quick look and the codec->ac97 is not used by any calls upto this point except for snd_soc_new_pcms() adding the AC97 pdata (please also add a check for valid codec->ac97 in snd_soc_new_pcms() before setting pdata).

Sorry, I don't have any working AC97 hardware to test this myself today.

Thanks

Liam

Signature · 3 lines
Freelance Developer, SlimLogic Ltd
ASoC and Voltage Regulator Maintainer.
http://www.slimlogic.co.uk
#2 lore raw
In reply to Graham Gower
Quoted 3 lines · Graham Gower

snd_soc_new_ac97_codec() allocates for codec->ac97, snd_soc_new_pcms() then sets the platform_data. However, snd_ac97_mixer() overwrites codec->ac97 with its own allocated struct snd_ac97.

Please remember to CC maintainers on patches, this helps avoid patches getting lost in mailing lists - the procedure in SubmittingPatches should generally be followed.

sound/soc/codecs/ac97.c static int ac97_soc_probe(struct platform_device *pdev) · line 125

Quoted 7 lines
sound/soc/codecs/ac97.c +3 -0
@@ -118,11 +120,21 @@ static int ac97_soc_probe(struct platform_device *pdev)
 	if (ret < 0)
 		goto bus_err;
 
+	/* free the ac97 here so that we don't leak it in snd_ac97_mixer */
+	snd_soc_free_ac97_codec(codec);
+

Like Liam says this seems wrong - why would we have allocated this early on in the probe() only to free it later?

Note also that this code needs to work with both ASoC-native AC97 drivers and the ac97.c bodge.

#3 lore raw

Mark, Liam,

As requested, this patch removes the initial allocation of codec->ac97.

---
 sound/soc/codecs/ac97.c |   15 +++++++++------
 sound/soc/soc-core.c    |    2 +-
 2 files changed, 10 insertions(+), 7 deletions(-)
sound/soc/codecs/ac97.c +9 -6
index a1bbe16..bcfa532 100644
@@ -80,9 +80,11 @@ static int ac97_write(struct snd_soc_codec *codec, unsigned int reg,
 static int ac97_soc_probe(struct platform_device *pdev)
 {
 	struct snd_soc_device *socdev = platform_get_drvdata(pdev);
+	struct snd_soc_card *card = socdev->card;
 	struct snd_soc_codec *codec;
 	struct snd_ac97_bus *ac97_bus;
 	struct snd_ac97_template ac97_template;
+	int i;
 	int ret = 0;
 
 	printk(KERN_INFO "AC97 SoC Audio Codec %s\n", AC97_VERSION);
@@ -102,12 +104,6 @@ static int ac97_soc_probe(struct platform_device *pdev)
 	INIT_LIST_HEAD(&codec->dapm_widgets);
 	INIT_LIST_HEAD(&codec->dapm_paths);
 
-	ret = snd_soc_new_ac97_codec(codec, &soc_ac97_ops, 0);
-	if (ret < 0) {
-		printk(KERN_ERR "ASoC: failed to init gen ac97 glue\n");
-		goto err;
-	}
-
 	/* register pcms */
 	ret = snd_soc_new_pcms(socdev, SNDRV_DEFAULT_IDX1, SNDRV_DEFAULT_STR1);
 	if (ret < 0)
@@ -123,6 +119,13 @@ static int ac97_soc_probe(struct platform_device *pdev)
 	if (ret < 0)
 		goto bus_err;
 
+	for (i = 0; i < card->num_links; i++) {
+		if (card->dai_link[i].codec_dai->ac97_control) {
+			snd_ac97_dev_add_pdata(codec->ac97,
+				card->dai_link[i].cpu_dai->ac97_pdata);
+		}
+	}
+
 	return 0;
 
 bus_err:
sound/soc/soc-core.c +1 -1
index c8b0556..5704923 100644
@@ -1548,7 +1548,7 @@ int snd_soc_new_pcms(struct snd_soc_device *socdev, int idx, const char *xid)
 			mutex_unlock(&codec->mutex);
 			return ret;
 		}
-		if (card->dai_link[i].codec_dai->ac97_control) {
+		if (card->dai_link[i].codec_dai->ac97_control && codec->ac97) {
 			snd_ac97_dev_add_pdata(codec->ac97,
 				card->dai_link[i].cpu_dai->ac97_pdata);
 		}
Signature · 1 line
1.6.4
#5 lore raw
In reply to Graham Gower
Quoted 3 lines · Graham Gower

Mark, Liam,

As requested, this patch removes the initial allocation of codec->ac97.

Please recheck SubmittingPatches - this'd end up in the changelog (and now your changelog is empty so it's not clear what the actual problem is any more). I've applied this with a changelog added and...

sound/soc/soc-core.c int snd_soc_new_pcms(struct snd_soc_device *socdev, int idx, const char *xid) · line 1553

Quoted 5 lines
sound/soc/soc-core.c +1 -1
 		}
-		if (card->dai_link[i].codec_dai->ac97_control) {
+		if (card->dai_link[i].codec_dai->ac97_control && codec->ac97) {
 			snd_ac97_dev_add_pdata(codec->ac97,
 				card->dai_link[i].cpu_dai->ac97_pdata);

...a comment here since otherwise it's not obvious why we're not complaining about the missing codec->ac97 otherwise.