Question about soc_bind_aux_dev()

Question about soc_bind_aux_dev()

[Kuninori Morimoto]

Hi ALSA ML

Do you know why snd_soc_aux_dev has both name and codec_name ??

	struct snd_soc_aux_dev {
=>		const char *name;		/* Codec name */
		...
=>		const char *codec_name;
		...
	};

For example, samsung's speyside.c has both name and codec_name

	static struct snd_soc_aux_dev speyside_aux_dev[] = {
		{
=>			.name = "wm9081",
=>			.codec_name = "wm9081.1-006c",
			...
		},
	};

In my research, there is no driver which is using .name only.
All drivers are having codec_of_node, or .codec_name

Now, in the soc_bind_aux_dev(), it checks .codec_name or codec_of_node first.
This means, there is zero chance to use .name ?

	static int soc_bind_aux_dev(struct snd_soc_card *card, int num)
	{
		...
=>		if (aux_dev->codec_of_node || aux_dev->codec_name) {
			...
			component = soc_find_component(...);
			...
=>		} else if (aux_dev->name) {
			...
			component = soc_find_component(...);
			...
		}
		...
	}

Why .name is needed ??

Thank you for your help !!
Best regards
---
Kuninori Morimoto

Re: Question about soc_bind_aux_dev()

[Takashi Iwai]

On Wed, 22 May 2019 10:30:51 +0200,
Kuninori Morimoto wrote:
> 
> 
> Hi ALSA ML
> 
> Do you know why snd_soc_aux_dev has both name and codec_name ??
> 
> 	struct snd_soc_aux_dev {
> =>		const char *name;		/* Codec name */
> 		...
> =>		const char *codec_name;
> 		...
> 	};
> 
> For example, samsung's speyside.c has both name and codec_name
> 
> 	static struct snd_soc_aux_dev speyside_aux_dev[] = {
> 		{
> =>			.name = "wm9081",
> =>			.codec_name = "wm9081.1-006c",
> 			...
> 		},
> 	};
> 
> In my research, there is no driver which is using .name only.
> All drivers are having codec_of_node, or .codec_name
> 
> Now, in the soc_bind_aux_dev(), it checks .codec_name or codec_of_node first.
> This means, there is zero chance to use .name ?
> 
> 	static int soc_bind_aux_dev(struct snd_soc_card *card, int num)
> 	{
> 		...
> =>		if (aux_dev->codec_of_node || aux_dev->codec_name) {
> 			...
> 			component = soc_find_component(...);
> 			...
> =>		} else if (aux_dev->name) {
> 			...
> 			component = soc_find_component(...);
> 			...
> 		}
> 		...
> 	}
> 
> Why .name is needed ??

AFAIK, the binding with aux_dev->name was introduced later as a
fallback in case of no codec.  Through a quick glance at git log, the
commit f2ed6b07645e brought the behavior.  Other than that, the field
was used merely for some label, e.g. passed to
snd_soc_post_component_init().


Takashi

Re: Question about soc_bind_aux_dev()

[Kuninori Morimoto]

Hi Takashi

Thank you for your feedback

> > Why .name is needed ??
>
> AFAIK, the binding with aux_dev->name was introduced later as a
> fallback in case of no codec.  Through a quick glance at git log, the
> commit f2ed6b07645e brought the behavior.  Other than that, the field
> was used merely for some label, e.g. passed to
> snd_soc_post_component_init().

Ahh, OK. for non codec...
Then, the reason was "main" is for search for codec,
and it is using "codec_xxx" variable name.
Then, using .name and .codec_name/.codec_of_node in the same time is strange.

Anyway, the purpose of these are for finding component.
I think this complicated code can be simple if we could replace it to
snd_soc_dai_link_component which exist for this purpose ?

	struct snd_soc_aux_dev {
-		const char *name;		/* Codec name */

		/*
		 * specify multi-codec either by device name, or by
		 * DT/OF node, but not both.
		 */
-		const char *codec_name;
-		struct device_node *codec_of_node;
+		struct snd_soc_dai_link_component dlc;
+
		/* codec/machine specific init - e.g. add machine controls */
		int (*init)(struct snd_soc_component *component);
	};

Re: Question about soc_bind_aux_dev()

[Kuninori Morimoto]

Hi Takashi, again

> AFAIK, the binding with aux_dev->name was introduced later as a
> fallback in case of no codec.  Through a quick glance at git log, the
> commit f2ed6b07645e brought the behavior.  Other than that, the field
> was used merely for some label, e.g. passed to
> snd_soc_post_component_init().

soc_post_component_init() is using dai_link->name, not aux_dev->name.
But, am I misunderstanding ?

Thank you for your help !!
Best regards
---
Kuninori Morimoto

Re: Question about soc_bind_aux_dev()

[Takashi Iwai]

On Thu, 23 May 2019 03:01:10 +0200,
Kuninori Morimoto wrote:
> 
> 
> Hi Takashi, again
> 
> > AFAIK, the binding with aux_dev->name was introduced later as a
> > fallback in case of no codec.  Through a quick glance at git log, the
> > commit f2ed6b07645e brought the behavior.  Other than that, the field
> > was used merely for some label, e.g. passed to
> > snd_soc_post_component_init().
> 
> soc_post_component_init() is using dai_link->name, not aux_dev->name.
> But, am I misunderstanding ?

It used to be called, but was removed by the commit above.

The author of the patch, Mengdong, should be able to give a better
answer about the use case of aux_dev->name.


thanks,

Takashi