Skip to contentSkip to conversation

ASoC: meson: aiu: Fix resource leak

PATCH In review Ethan Tidmore 4 participants 6 messages Jul 7 – Jul 10

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 ASoC: meson: aiu: Fix resource leak

In snd_soc_register_component() error path, the aiu-formatter is not being freed. Instead of returning now goto the err tag so the aiu-formatter can properly be freed.

Fixes 83b83024cdbfd ("ASoC: meson: aiu: use aiu-formatter-i2s to format I2S output data")
---
 sound/soc/meson/aiu.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
sound/soc/meson/aiu.c +1 -1
index 64ace4d25d92..d6b644a53f7e 100644
@@ -310,7 +310,7 @@ static int aiu_probe(struct platform_device *pdev)
 					 ARRAY_SIZE(aiu_cpu_dai_drv));
 	if (ret) {
 		dev_err(dev, "Failed to register cpu component\n");
-		return ret;
+		goto err;
 	}
 
 	/* Register the hdmi codec control component */
Signature · 3 lines
Thanks,
ET
https://github.com/sponsors/ethantidmore
#1 lore raw
Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider: - [Medium] The patch modifies the error path to purportedly free the `aiu-formatter` resource by jumping to…

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] The patch modifies the error path to purportedly free the aiu-formatter resource by jumping to the err label, referencing commit 83b83024cdbfd. However, the aiu-formatter allocation and cleanup code (specifically gx_formatter_create and gx_formatter_free) does not exist in the codebase at this revision. The patch is a "ghost fix" attempting to resolve a resource leak for a feature that is missing in this tree. --

--- Patch [1]: [PATCH] ASoC: meson: aiu: Fix resource leak --- commit e1f43f4a59c62657488472af87aa2e5d80576b03
Author: Ethan Tidmore <[email protected]>

ASoC: meson: aiu: Fix resource leak

This commit changes an error path in aiu_probe() to use a goto label instead of returning directly. The commit message indicates this is to ensure the aiu-formatter is properly freed when component registration fails.

Quoted 4 lines
sound/soc/meson/aiu.c +0 -0
index f2890111c1d2c..1cfa165fb9cfc 100644

[ ... ]

sound/soc/meson/aiu.c static int aiu_probe(struct platform_device *pdev) · line 291

Quoted 6 lines
sound/soc/meson/aiu.c +1 -1
@@ -288,7 +288,7 @@ static int aiu_probe(struct platform_device *pdev)
 					 ARRAY_SIZE(aiu_cpu_dai_drv));
 	if (ret) {
 		dev_err(dev, "Failed to register cpu component\n");
-		return ret;
+		goto err;

[Severity: Medium]
Is there a missing dependency for this patch, or is it an incomplete backport?

The commit message mentions that the aiu-formatter is not being freed. However, looking at aiu_probe() in this tree, there is no aiu-formatter allocation or cleanup code present anywhere in the function.

Quoted 3 lines
sound/soc/meson/aiu.c +0 -0
 	}
 
 	/* Register the hdmi codec control component */

[ ... ]

sound/soc/meson/aiu.c

Quoted 4 lines

err:

sound/soc/meson/aiu.c +0 -0
 	snd_soc_unregister_component(dev);
 	return ret;
 }

[Severity: Medium]
At this error label, only snd_soc_unregister_component() is called, and no formatter cleanup logic exists. Does this fix apply to a different tree where the formatter feature is actually present?

Signature · 6 lines
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1

_______________________________________________
linux-amlogic mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-amlogic
#2 lore raw
Quoted 3 lines · sashiko-bot

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] The patch modifies the error path to purportedly free the aiu-formatter resource by jumping to the err label, referencing commit 83b83024cdbfd. However, the aiu-formatter allocation and cleanup code (specifically gx_formatter_create and gx_formatter_free) does not exist in the codebase at this revision. The patch is a "ghost fix" attempting to resolve a resource leak for a feature that is missing in this tree. --

This bug exists in linux-next. The AI is using the mainline.

--
Thanks,
ET
https://github.com/sponsors/ethantidmore

#3 lore raw
In reply to Ethan Tidmore
Quoted 15 lines · Ethan Tidmore

In snd_soc_register_component() error path, the aiu-formatter is not being freed. Instead of returning now goto the err tag so the aiu-formatter can properly be freed.

Fixes 83b83024cdbfd ("ASoC: meson: aiu: use aiu-formatter-i2s to format I2S output data")
---
  sound/soc/meson/aiu.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
sound/soc/meson/aiu.c +0 -0
index 64ace4d25d92..d6b644a53f7e 100644

sound/soc/meson/aiu.c static int aiu_probe(struct platform_device *pdev) · line 316

Quoted 8 lines
sound/soc/meson/aiu.c +1 -1
@@ -310,7 +310,7 @@ static int aiu_probe(struct platform_device *pdev)
  					 ARRAY_SIZE(aiu_cpu_dai_drv));
  	if (ret) {
  		dev_err(dev, "Failed to register cpu component\n");
-		return ret;
+		goto err;
  	}
 
  	/* Register the hdmi codec control component */

Technically the fix is OK, but I don't think the commit title and message are correct. aiu-formatter data is allocated in 'gx_formatter_create' using 'devm_kzalloc' so, unless I'm missing something, the data should be automatically freed when the device is removed in case of error. What 'gx_formatter_free' does is simply clearing widget's private pointer (i.e. 'w->priv'). I think that not clearing it should not have any bad consequence since the pointer would be overwritten on the next probe. However I agree that for consistency reasons it's a good thing to clear to do when 'aiu_probe' fails.

So perhaps the title should be rewritten to something like "ASoC: meson: aiu: reset widget private data on probe error path"?

Signature · 1 line
Valerio
#4 lore raw

…

Quoted 9 lines
Quoted 9 lines

+++ b/sound/soc/meson/aiu.c

sound/soc/meson/aiu.c +1 -1
@@ -310,7 +310,7 @@ static int aiu_probe(struct platform_device *pdev)
  					 ARRAY_SIZE(aiu_cpu_dai_drv));
  	if (ret) {
  		dev_err(dev, "Failed to register cpu component\n");
-		return ret;
+		goto err;
  	}
 
Quoted 3 lines

sound/soc/meson/aiu.c static int aiu_probe(struct platform_device *pdev) · line 316

Quoted 1 line
sound/soc/meson/aiu.c +0 -0
  	/* Register the hdmi codec control component */

Technically the fix is OK, …

I doubt it.

How can a snd_soc_unregister_component(dev) call ever fit to a failed snd_soc_register_component() call (according to the presented suggestion)?
https://elixir.bootlin.com/linux/v7.2-rc1/source/sound/soc/meson/aiu.c#L240-L315

Regards,
Markus

#5 lore raw
Quoted 14 lines · Markus Elfring

…

Quoted 9 lines
Quoted 9 lines

+++ b/sound/soc/meson/aiu.c

sound/soc/meson/aiu.c +1 -1
@@ -310,7 +310,7 @@ static int aiu_probe(struct platform_device *pdev)
   					 ARRAY_SIZE(aiu_cpu_dai_drv));
   	if (ret) {
   		dev_err(dev, "Failed to register cpu component\n");
-		return ret;
+		goto err;
   	}
 
Quoted 3 lines

sound/soc/meson/aiu.c static int aiu_probe(struct platform_device *pdev) · line 316

Quoted 1 line
sound/soc/meson/aiu.c +0 -0
   	/* Register the hdmi codec control component */

Technically the fix is OK, …

Quoted 5 lines

I doubt it.

How can a snd_soc_unregister_component(dev) call ever fit to a failed snd_soc_register_component() call (according to the presented suggestion)?
https://elixir.bootlin.com/linux/v7.2-rc1/source/sound/soc/meson/aiu.c#L240-L315

The reply was based on the fact that looking at 'snd_soc_unregister_component' that's a macro for 'snd_soc_unregister_component_by_driver'. The latter cycles through all registered components and if nothing is found related to the given dev, then it results in a no-op. So calling 'snd_soc_unregister_component' on a failed 'snd_soc_register_component' looked harmless to me.

What did I miss?

Signature · 1 line
Valerio