Skip to contentSkip to conversation

simple-card: Improve probe's error messages

RFC In review Fabio Forni 3 participants 11 messages Jan 12 – Jan 26

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

cover simple-card: Improve probe's error messages

Hello,

this patch tries to improve errors messages that are printed by the probe function of simple-card. Some messages were missing and I think that made debugging difficult, others were printed even though the underlying error is unrelated (e.g. when devm_snd_soc_register_card is called).

This series is tagged as RFC because it's my first contribution. Plus, patch number 3 touches many lines. I'm not sure I'm allowed to touch so many lines in one run, also I need more than a pair of eyes because even though patch #3 is just a refactor, I'm not 100% familiar with ASoC.

Fabio Forni (3):
ASoC: simple-card: Do not print "parse error" twice
ASoC: simple-card: Replace "parse error" with a more descriptive msg
ASoC: simple-card: Split alloc and init logics in probe function

 sound/soc/generic/simple-card-utils.c |   9 ++
 sound/soc/generic/simple-card.c       | 141 +++++++++++++-------------
 2 files changed, 82 insertions(+), 68 deletions(-)
Signature · 1 line
2.52.0
#1 lore raw

RFC 1/3 ASoC: simple-card: Do not print "parse error" twice

If an error occurred on line 747, we'd print "parse error" and then go to the err label, where we print the same message.
Let's remove the duplicated print.

---
 sound/soc/generic/simple-card.c | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)
sound/soc/generic/simple-card.c +1 -3
index 5af6d1b308f2..8d36cbbfca1d 100644
@@ -745,10 +745,8 @@ static int simple_probe(struct platform_device *pdev)
 	if (np && of_device_is_available(np)) {
 
 		ret = simple_parse_of(priv, li);
-		if (ret < 0) {
-			dev_err_probe(dev, ret, "parse error\n");
+		if (ret < 0)
 			goto err;
-		}
 
 	} else {
 		struct simple_util_info *cinfo;
Signature · 1 line
2.52.0
#2 lore raw

Re: [RFC 1/3] ASoC: simple-card: Do not print "parse error" twice

Hi Fabio

Thank you for your patch

Quoted 6 lines

If an error occurred on line 747, we'd print "parse error" and then go to the err label, where we print the same message.
Let's remove the duplicated print.

---

I think this patch effect will be removed/freshed by [3/3] patch anyway ? Same comment to [2/3] patch

Thank you for your help !!

Best regards
---
Kuninori Morimoto

author
#3 lore raw

Re: [RFC 1/3] ASoC: simple-card: Do not print "parse error" twice

Thank you for your fast review, Kuninori!

Quoted 2 lines

I think this patch effect will be removed/freshed by [3/3] patch anyway ? Same comment to [2/3] patch

Yes, definitely. I kept my commit history to document my flow of thought and to address specific problems. At this point I'll just squash #1 and #2 into one commit in v2.

#4 lore raw

RFC 2/3 ASoC: simple-card: Replace "parse error" with a more descriptive msg

In reply to Fabio Forni

"parse error" is misleading: it made me think the device tree contains syntax errors, or phandles of the wrong type were passed to the simple-audio-card node. Instead in my case, the sound-dai subnode of simple-audio-card,cpu was referring to an uninitialized device, because its kernel module wasn't loaded yet.

As far as I understood, simple_probe fails when components composing the sound card are missing or simply not ready.

---
 sound/soc/generic/simple-card.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
sound/soc/generic/simple-card.c +1 -1
index 8d36cbbfca1d..298f8cc98130 100644
@@ -806,7 +806,7 @@ static int simple_probe(struct platform_device *pdev)
 err:
 	simple_util_clean_reference(card);
 end:
-	return dev_err_probe(dev, ret, "parse error\n");
+	return dev_err_probe(dev, ret, "components missing or uninitialized\n");
 }
 
 static const struct of_device_id simple_of_match[] = {
Signature · 1 line
2.52.0
#5 lore raw

RFC 3/3 ASoC: simple-card: Split alloc and init logics in probe function

In reply to Fabio Forni

It wasn't clear to me when different simple_probe's gotos were to be used, plus some allocation errors and snd_soc* errors were printed as if they were a parsing error. This commit:

1. Splits the allocation logic and the initialization logic, respectively in simple_probe and simple_probe_merge_components
2. Adds more error messages to improve the debugging experience
3. Reduces the cognitive load of gotos, as there is now only one label (I concede that's subjective).

This commit also documents simple_util_init_priv.

---
 sound/soc/generic/simple-card-utils.c |   9 ++
 sound/soc/generic/simple-card.c       | 137 ++++++++++++++------------
 2 files changed, 81 insertions(+), 65 deletions(-)
sound/soc/generic/simple-card-utils.c +9 -0
index 355f7ec8943c..46a8ee04c6ef 100644
@@ -888,6 +888,15 @@ static struct simple_util_dai dummy_util_dais = {
 	.name = "dummy_util_dais",
 };
 
+/**
+ * simple_util_init_priv - Initializes private data of a simple audio card.
+ * @priv: Private data to initialize. Must be pre-allocated.
+ * @li: Links of the card. Cannot be NULL.
+ *
+ * Returns:
+ * * %0       - OK.
+ * * %-ENOMEM - Could not allocate memory.
+ */
 int simple_util_init_priv(struct simple_util_priv *priv,
 			  struct link_info *li)
 {
sound/soc/generic/simple-card.c +72 -65

176 lines not on this page · raw message

Signature · 1 line
2.52.0
#6 lore raw

Re: [RFC 3/3] ASoC: simple-card: Split alloc and init logics in probe function

Hi Fabio

Thank you for your patch

Quoted 14 lines

It wasn't clear to me when different simple_probe's gotos were to be used, plus some allocation errors and snd_soc* errors were printed as if they were a parsing error. This commit:

1. Splits the allocation logic and the initialization logic, respectively in simple_probe and simple_probe_merge_components
2. Adds more error messages to improve the debugging experience
3. Reduces the cognitive load of gotos, as there is now only one label (I concede that's subjective).

This commit also documents simple_util_init_priv.

---

(snip)

Quoted 9 lines

+/**
+ * simple_util_init_priv - Initializes private data of a simple audio card.
+ * @priv: Private data to initialize. Must be pre-allocated.
+ * @li: Links of the card. Cannot be NULL.
+ *
+ * Returns:
+ * * %0 - OK.
+ * * %-ENOMEM - Could not allocate memory.
+ */

This should be separate patch.

Quoted 35 lines
sound/soc/generic/simple-card.c +5 -19
index 298f8cc98130..b8004aa8f452 100644
@@ -706,47 +706,26 @@ static int simple_soc_probe(struct snd_soc_card *card)
 	return simple_ret(priv, ret);
 }
 
-static int simple_probe(struct platform_device *pdev)
+static int simple_probe_merge_components(const struct device *dev,
+					 struct simple_util_priv *priv,
+					 struct link_info *li)
 {
-	struct simple_util_priv *priv;
-	struct device *dev = &pdev->dev;
-	struct device_node *np = dev->of_node;
-	struct snd_soc_card *card;
+	const struct device_node *np = dev->of_node;
+	struct snd_soc_card *card = simple_priv_to_card(priv);
 	int ret;
 
-	/* Allocate the private data and the DAI link array */
-	priv = devm_kzalloc(dev, sizeof(*priv), GFP_KERNEL);
-	if (!priv)
-		return -ENOMEM;
-
-	card = simple_priv_to_card(priv);
-	card->owner		= THIS_MODULE;
-	card->dev		= dev;
-	card->probe		= simple_soc_probe;
-	card->driver_name       = "simple-card";
-
-	ret = -ENOMEM;
-	struct link_info *li __free(kfree) = kzalloc(sizeof(*li), GFP_KERNEL);
-	if (!li)

sound/soc/generic/simple-card.c static int simple_soc_probe(struct snd_soc_card *card) · line 718

Quoted 10 lines
sound/soc/generic/simple-card.c +1 -8
-		goto end;
-
-	ret = simple_get_dais_count(priv, li);
-	if (ret < 0)
-		goto end;
-
-	ret = -EINVAL;
 	if (!li->link)
-		goto end;
+		return -EINVAL;

no dev_err_probe() ?

sound/soc/generic/simple-card.c static int simple_soc_probe(struct snd_soc_card *card) · line 722

Quoted 6 lines
sound/soc/generic/simple-card.c +2 -2
-	ret = simple_util_init_priv(priv, li);
+	ret = simple_get_dais_count(priv, li);
 	if (ret < 0)
-		goto end;
+		dev_err_probe(dev, ret, "failed to count DAIs");

Here is missing return ?

sound/soc/generic/simple-card.c static int simple_soc_probe(struct snd_soc_card *card) · line 728

Quoted 7 lines
sound/soc/generic/simple-card.c +2 -2
 	if (np && of_device_is_available(np)) {
-
 		ret = simple_parse_of(priv, li);
 		if (ret < 0)
-			goto err;
+			dev_err_probe(dev, ret,
+				      "components missing or uninitialized");

Here is missing return ?

Is it clear error message ? It failed parse DT or something ?

Quoted 7 lines
sound/soc/generic/simple-card.c +0 -2
@@ -756,57 +735,85 @@ static int simple_probe(struct platform_device *pdev)
 		struct snd_soc_dai_link *dai_link = priv->dai_link;
 		struct simple_dai_props *dai_props = priv->dai_props;
 
-		ret = -EINVAL;
-
 		cinfo = dev->platform_data;

sound/soc/generic/simple-card.c static int simple_probe(struct platform_device *pdev) · line 744

Quoted 10 lines
sound/soc/generic/simple-card.c +6 -4
-		if (!cinfo) {
-			dev_err(dev, "no info for asoc-simple-card\n");
-			goto err;
-		}
+		if (!cinfo)
+			return dev_err_probe(dev, -EINVAL,
+					     "no info for asoc-simple-card\n");
+
+		if (!cinfo->name || !cinfo->codec_dai.name || !cinfo->codec ||
+		    !cinfo->platform || !cinfo->cpu_dai.name)

I like aligned condition, same as original code.

sound/soc/generic/simple-card.c static int simple_probe(struct platform_device *pdev) · line 747

Quoted 3 lines
sound/soc/generic/simple-card.c +3 -0
+			return dev_err_probe(
+				dev, -EINVAL,
+				"insufficient simple_util_info settings\n");

This can be 2 or 1 line ?

Quoted 5 lines
sound/soc/generic/simple-card.c +5 -0
+		cpus = dai_link->cpus;
+		cpus->dai_name = cinfo->cpu_dai.name;
+
+		codecs = dai_link->codecs;
+		codecs->name = cinfo->codec;

sound/soc/generic/simple-card.c static int simple_probe(struct platform_device *pdev) · line 763

Quoted 10 lines
sound/soc/generic/simple-card.c +10 -0
+		codecs->dai_name = cinfo->codec_dai.name;
+
+		platform = dai_link->platforms;
+		platform->name = cinfo->platform;
+
+		card->name = (cinfo->card) ? cinfo->card : cinfo->name;
+		dai_link->name = cinfo->name;
+		dai_link->stream_name = cinfo->name;
+		dai_link->dai_fmt = cinfo->daifmt;
+		dai_link->init = simple_util_dai_init;

I like tag aligned code, same as original

sound/soc/generic/simple-card.c static int simple_probe(struct platform_device *pdev) · line 767

Quoted 4 lines
sound/soc/generic/simple-card.c +4 -0
+		memcpy(dai_props->cpu_dai, &cinfo->cpu_dai,
+		       sizeof(*dai_props->cpu_dai));
+		memcpy(dai_props->codec_dai, &cinfo->codec_dai,
+		       sizeof(*dai_props->codec_dai));

These can be 1 line (with aligned. I like aligned code ;) ?

	memcpy(dai_props->cpu_dai,   &cinfo->cpu_dai,   sizeof(*dai_props->cpu_dai));
	memcpy(dai_props->codec_dai, &cinfo->codec_dai, sizeof(*dai_props->codec_dai));
Quoted 12 lines
sound/soc/generic/simple-card.c +8 -3
+static int simple_probe(struct platform_device *pdev)
+{
+	struct simple_util_priv *priv;
+	struct device *dev = &pdev->dev;
+	struct snd_soc_card *card;
+	int ret;
 
-		codecs			= dai_link->codecs;
-		codecs->name		= cinfo->codec;
-		codecs->dai_name	= cinfo->codec_dai.name;
+	/* Allocate the private data and the DAI link array */
+	priv = devm_kzalloc(dev, sizeof(*priv), GFP_KERNEL);

sound/soc/generic/simple-card.c static int simple_probe(struct platform_device *pdev) · line 789

Quoted 10 lines
sound/soc/generic/simple-card.c +7 -2
+	if (!priv)
+		return -ENOMEM;
 
-		platform		= dai_link->platforms;
-		platform->name		= cinfo->platform;
+	card = simple_priv_to_card(priv);
+	card->owner = THIS_MODULE;
+	card->dev = dev;
+	card->probe = simple_soc_probe;
+	card->driver_name = "simple-card";

I like to have tab align here, same as original code.

sound/soc/generic/simple-card.c static int simple_probe(struct platform_device *pdev) · line 793

Quoted 13 lines
sound/soc/generic/simple-card.c +3 -10
-		card->name		= (cinfo->card) ? cinfo->card : cinfo->name;
-		dai_link->name		= cinfo->name;
-		dai_link->stream_name	= cinfo->name;
-		dai_link->dai_fmt	= cinfo->daifmt;
-		dai_link->init		= simple_util_dai_init;
-		memcpy(dai_props->cpu_dai, &cinfo->cpu_dai,
-					sizeof(*dai_props->cpu_dai));
-		memcpy(dai_props->codec_dai, &cinfo->codec_dai,
-					sizeof(*dai_props->codec_dai));
-	}
+	struct link_info *li __free(kfree) = kzalloc(sizeof(*li), GFP_KERNEL);
+	if (!li)
+		return -ENOMEM;

no dev_err_probe() ?

sound/soc/generic/simple-card.c static int simple_probe(struct platform_device *pdev) · line 797

Quoted 3 lines
sound/soc/generic/simple-card.c +3 -0
+	ret = simple_util_init_priv(priv, li);
+	if (ret < 0)
+		return ret;

no dev_err_probe() ?

It is calling simple_util_init_priv() (A) without calling simple_get_dais_count() (B).
((B) simple_get_dais_count() fill li->num[x].xxx count, and
(A) simple_util_init_priv() will use it)

I think it doesn't work correctly ?

Thank you for your help !!

Best regards
---
Kuninori Morimoto

#7 lore raw

Re: [RFC 3/3] ASoC: simple-card: Split alloc and init logics in probe function

I don't think dev_err_probe() should be called when -ENOMEM is thrown as I think it's obvious what failed. It also seems to do anything at all per commit d6137f25b191. Do you agree with that commit?

Quoted 8 lines
Quoted 7 lines
  	if (np && of_device_is_available(np)) {
sound/soc/generic/simple-card.c +2 -2
-
  		ret = simple_parse_of(priv, li);
  		if (ret < 0)
-			goto err;
+			dev_err_probe(dev, ret,
+				      "components missing or uninitialized");

[snip]

Quoted 2 lines

Is it clear error message ? It failed parse DT or something ?

That's exactly the part that made me provide this patch. There's so much going on in simple_parse_of(). I think it may fail when a phandle of the wrong type is passed in any of the simple-card node.
But I'm certain it can also fail when the DT is correct but kernel modules of the simple-card's components aren't loaded yet, since this very case happened to me. "components missing or uninitialized" basically sums it up, without saying "parse error" that suggests a purely syntactic error. If you don't agree with that message I can dive into the rabbit hole and learn what's going on inside simple_parse_of().

Quoted 1 line

with aligned. I like aligned code ;) ?

Blame clang-format style! I have autoformatting enabled. I'll manually fix the alignment :)

---

I'll fix the other errors in v2. And sorry for those missing returns, I shouldn't code in the night...

Quoted 214 lines · Kuninori Morimoto

Hi Fabio

Thank you for your patch

Quoted 14 lines

It wasn't clear to me when different simple_probe's gotos were to be used, plus some allocation errors and snd_soc* errors were printed as if they were a parsing error. This commit:

1. Splits the allocation logic and the initialization logic, respectively in simple_probe and simple_probe_merge_components
2. Adds more error messages to improve the debugging experience
3. Reduces the cognitive load of gotos, as there is now only one label (I concede that's subjective).

This commit also documents simple_util_init_priv.

---

(snip)

Quoted 9 lines

+/**
+ * simple_util_init_priv - Initializes private data of a simple audio card.
+ * @priv: Private data to initialize. Must be pre-allocated.
+ * @li: Links of the card. Cannot be NULL.
+ *
+ * Returns:
+ * * %0 - OK.
+ * * %-ENOMEM - Could not allocate memory.
+ */

This should be separate patch.

Quoted 35 lines
sound/soc/generic/simple-card.c +5 -19
index 298f8cc98130..b8004aa8f452 100644
@@ -706,47 +706,26 @@ static int simple_soc_probe(struct snd_soc_card *card)
  	return simple_ret(priv, ret);
  }
 
-static int simple_probe(struct platform_device *pdev)
+static int simple_probe_merge_components(const struct device *dev,
+					 struct simple_util_priv *priv,
+					 struct link_info *li)
  {
-	struct simple_util_priv *priv;
-	struct device *dev = &pdev->dev;
-	struct device_node *np = dev->of_node;
-	struct snd_soc_card *card;
+	const struct device_node *np = dev->of_node;
+	struct snd_soc_card *card = simple_priv_to_card(priv);
  	int ret;
 
-	/* Allocate the private data and the DAI link array */
-	priv = devm_kzalloc(dev, sizeof(*priv), GFP_KERNEL);
-	if (!priv)
-		return -ENOMEM;
-
-	card = simple_priv_to_card(priv);
-	card->owner		= THIS_MODULE;
-	card->dev		= dev;
-	card->probe		= simple_soc_probe;
-	card->driver_name       = "simple-card";
-
-	ret = -ENOMEM;
-	struct link_info *li __free(kfree) = kzalloc(sizeof(*li), GFP_KERNEL);
-	if (!li)

sound/soc/generic/simple-card.c static int simple_soc_probe(struct snd_soc_card *card) · line 718

Quoted 10 lines
sound/soc/generic/simple-card.c +1 -8
-		goto end;
-
-	ret = simple_get_dais_count(priv, li);
-	if (ret < 0)
-		goto end;
-
-	ret = -EINVAL;
  	if (!li->link)
-		goto end;
+		return -EINVAL;

no dev_err_probe() ?

sound/soc/generic/simple-card.c static int simple_soc_probe(struct snd_soc_card *card) · line 722

Quoted 6 lines
sound/soc/generic/simple-card.c +2 -2
-	ret = simple_util_init_priv(priv, li);
+	ret = simple_get_dais_count(priv, li);
  	if (ret < 0)
-		goto end;
+		dev_err_probe(dev, ret, "failed to count DAIs");

Here is missing return ?

sound/soc/generic/simple-card.c static int simple_soc_probe(struct snd_soc_card *card) · line 728

Quoted 7 lines
sound/soc/generic/simple-card.c +2 -2
  	if (np && of_device_is_available(np)) {
-
  		ret = simple_parse_of(priv, li);
  		if (ret < 0)
-			goto err;
+			dev_err_probe(dev, ret,
+				      "components missing or uninitialized");

Here is missing return ?

Is it clear error message ? It failed parse DT or something ?

Quoted 7 lines
sound/soc/generic/simple-card.c +0 -2
@@ -756,57 +735,85 @@ static int simple_probe(struct platform_device *pdev)
  		struct snd_soc_dai_link *dai_link = priv->dai_link;
  		struct simple_dai_props *dai_props = priv->dai_props;
 
-		ret = -EINVAL;
-
  		cinfo = dev->platform_data;

sound/soc/generic/simple-card.c static int simple_probe(struct platform_device *pdev) · line 744

Quoted 10 lines
sound/soc/generic/simple-card.c +6 -4
-		if (!cinfo) {
-			dev_err(dev, "no info for asoc-simple-card\n");
-			goto err;
-		}
+		if (!cinfo)
+			return dev_err_probe(dev, -EINVAL,
+					     "no info for asoc-simple-card\n");
+
+		if (!cinfo->name || !cinfo->codec_dai.name || !cinfo->codec ||
+		    !cinfo->platform || !cinfo->cpu_dai.name)

I like aligned condition, same as original code.

sound/soc/generic/simple-card.c static int simple_probe(struct platform_device *pdev) · line 747

Quoted 3 lines
sound/soc/generic/simple-card.c +3 -0
+			return dev_err_probe(
+				dev, -EINVAL,
+				"insufficient simple_util_info settings\n");

This can be 2 or 1 line ?

Quoted 5 lines
sound/soc/generic/simple-card.c +5 -0
+		cpus = dai_link->cpus;
+		cpus->dai_name = cinfo->cpu_dai.name;
+
+		codecs = dai_link->codecs;
+		codecs->name = cinfo->codec;

sound/soc/generic/simple-card.c static int simple_probe(struct platform_device *pdev) · line 763

Quoted 10 lines
sound/soc/generic/simple-card.c +10 -0
+		codecs->dai_name = cinfo->codec_dai.name;
+
+		platform = dai_link->platforms;
+		platform->name = cinfo->platform;
+
+		card->name = (cinfo->card) ? cinfo->card : cinfo->name;
+		dai_link->name = cinfo->name;
+		dai_link->stream_name = cinfo->name;
+		dai_link->dai_fmt = cinfo->daifmt;
+		dai_link->init = simple_util_dai_init;

I like tag aligned code, same as original

sound/soc/generic/simple-card.c static int simple_probe(struct platform_device *pdev) · line 767

Quoted 4 lines
sound/soc/generic/simple-card.c +4 -0
+		memcpy(dai_props->cpu_dai, &cinfo->cpu_dai,
+		       sizeof(*dai_props->cpu_dai));
+		memcpy(dai_props->codec_dai, &cinfo->codec_dai,
+		       sizeof(*dai_props->codec_dai));

These can be 1 line (with aligned. I like aligned code ;) ?

	memcpy(dai_props->cpu_dai,   &cinfo->cpu_dai,   sizeof(*dai_props->cpu_dai));
	memcpy(dai_props->codec_dai, &cinfo->codec_dai, sizeof(*dai_props->codec_dai));
Quoted 12 lines
sound/soc/generic/simple-card.c +8 -3
+static int simple_probe(struct platform_device *pdev)
+{
+	struct simple_util_priv *priv;
+	struct device *dev = &pdev->dev;
+	struct snd_soc_card *card;
+	int ret;
 
-		codecs			= dai_link->codecs;
-		codecs->name		= cinfo->codec;
-		codecs->dai_name	= cinfo->codec_dai.name;
+	/* Allocate the private data and the DAI link array */
+	priv = devm_kzalloc(dev, sizeof(*priv), GFP_KERNEL);

sound/soc/generic/simple-card.c static int simple_probe(struct platform_device *pdev) · line 789

Quoted 10 lines
sound/soc/generic/simple-card.c +7 -2
+	if (!priv)
+		return -ENOMEM;
 
-		platform		= dai_link->platforms;
-		platform->name		= cinfo->platform;
+	card = simple_priv_to_card(priv);
+	card->owner = THIS_MODULE;
+	card->dev = dev;
+	card->probe = simple_soc_probe;
+	card->driver_name = "simple-card";

I like to have tab align here, same as original code.

sound/soc/generic/simple-card.c static int simple_probe(struct platform_device *pdev) · line 793

Quoted 13 lines
sound/soc/generic/simple-card.c +3 -10
-		card->name		= (cinfo->card) ? cinfo->card : cinfo->name;
-		dai_link->name		= cinfo->name;
-		dai_link->stream_name	= cinfo->name;
-		dai_link->dai_fmt	= cinfo->daifmt;
-		dai_link->init		= simple_util_dai_init;
-		memcpy(dai_props->cpu_dai, &cinfo->cpu_dai,
-					sizeof(*dai_props->cpu_dai));
-		memcpy(dai_props->codec_dai, &cinfo->codec_dai,
-					sizeof(*dai_props->codec_dai));
-	}
+	struct link_info *li __free(kfree) = kzalloc(sizeof(*li), GFP_KERNEL);
+	if (!li)
+		return -ENOMEM;

no dev_err_probe() ?

sound/soc/generic/simple-card.c static int simple_probe(struct platform_device *pdev) · line 797

Quoted 3 lines
sound/soc/generic/simple-card.c +3 -0
+	ret = simple_util_init_priv(priv, li);
+	if (ret < 0)
+		return ret;

no dev_err_probe() ?

It is calling simple_util_init_priv() (A) without calling simple_get_dais_count() (B).
((B) simple_get_dais_count() fill li->num[x].xxx count, and
(A) simple_util_init_priv() will use it)

I think it doesn't work correctly ?

Thank you for your help !!

Best regards
---
Kuninori Morimoto

#8 lore raw

Re: [RFC 3/3] ASoC: simple-card: Split alloc and init logics in probe function

Hi Fabio

Sorry for my late response

Quoted 12 lines
Quoted 8 lines
Quoted 7 lines
  	if (np && of_device_is_available(np)) {
sound/soc/generic/simple-card.c +2 -2
-
  		ret = simple_parse_of(priv, li);
  		if (ret < 0)
-			goto err;
+			dev_err_probe(dev, ret,
+				      "components missing or uninitialized");

[snip]

Quoted 2 lines

Is it clear error message ? It failed parse DT or something ?

Quoted 8 lines

That's exactly the part that made me provide this patch. There's so much going on in simple_parse_of(). I think it may fail when a phandle of the wrong type is passed in any of the simple-card node.
But I'm certain it can also fail when the DT is correct but kernel modules of the simple-card's components aren't loaded yet, since this very case happened to me. "components missing or uninitialized" basically sums it up, without saying "parse error" that suggests a purely syntactic error. If you don't agree with that message I can dive into the rabbit hole and learn what's going on inside simple_parse_of().

To be honest, I don't super mind about error message :P

But I have thought that the style like below can easy for people ?
(A) part checks something and return error, then indicate error reasons.
(B) part returns error because the called function returns error.

	func () {
		...
(A)		if (...)
			error("indicate error reason1")
(A)		if (...)
			error("indicate error reason2")
		...
	}


	main() {
		...
		ret = func();
(B)		if (ret < 0)
			error("func() error")
		...
	}

Thank you for your help !!

Best regards
---
Kuninori Morimoto

#9 lore raw

Re: [RFC 3/3] ASoC: simple-card: Split alloc and init logics in probe function

Hi, here I am.

Quoted 14 lines

But I have thought that the style like below can easy for people ?
(A) part checks something and return error, then indicate error reasons.
(B) part returns error because the called function returns error.

	func () {
		...
(A)		if (...)
			error("indicate error reason1")
(A)		if (...)
			error("indicate error reason2")
		...
	}
Quoted 7 lines
	main() {
		...
		ret = func();
(B)		if (ret < 0)
			error("func() error")
		...
	}

Hm, I think you're being too abstract for my comprehension, sorry about that. Let me assume main() is the probe function, and func() is whatever function that probe() calls. If my assumptions are correct, then:

- In (A), are we letting child functions print the reason of their own error? If so, I like the idea! In C we only return errors codes without a message, so the actual reason of, say, EINVAL is only known where the error was thrown. Outside of the context of the function an EINVAL just means "something was invalid". So yeah, I'd choose option (A).
- In (B), I think it's the approach already being used for simple_parse_of(), isn't it? You cannot know the reason of the error for sure, again because of error codes without a message being carried together. The best you can do here is thinking of a generic error message like I did.

Option (A) requires more work, of course.

#10 lore raw

Re: [RFC 3/3] ASoC: simple-card: Split alloc and init logics in probe function

Hi Fabio

Quoted 18 lines
Quoted 10 lines
	func () {
		...
(A)		if (...)
			error("indicate error reason1")
(A)		if (...)
			error("indicate error reason2")
		...
	}
Quoted 7 lines
	main() {
		...
		ret = func();
(B)		if (ret < 0)
			error("func() error")
		...
	}
Quoted 3 lines

Hm, I think you're being too abstract for my comprehension, sorry about that. Let me assume main() is the probe function, and func() is whatever function that probe() calls. If my assumptions are correct, then:

Sorry for my not good explanation, indeed it is too abstract. Above (A) and (B) exists in the same time, I meant.

Quoted 1 line

- In (A), are we letting child functions print the reason of their own error?

Yes. indicate why it become error in each function. Because size is too small, memory alloc failed, etc, etc...

Quoted 2 lines

- In (B), I think it's the approach already being used for simple_parse_of(), isn't it?

(B) indicates func() returns error. I think this can indicate error path, like below

	xxx size failed (error reason)		// (A)
	funcB() error				// (B)
	funcA() error				// (B)

Developer can understand that funcA() calls funcB() and funcB() returns error from this message.
But this is not mandatory, just an idea.

I think simple_ret() can help you, but I'm not sure whether you can use it on probe function.

Thank you for your help !!

Best regards
---
Kuninori Morimoto