Skip to contentSkip to conversation

Problems with simple-card

DISCUSSION Jean-Francois Moine 4 participants 5 messages Jan 15, 2014 – Jan 16, 2014

lore
Share & tools
lore
#0 lore raw

Xiubo,

I did a mistake in the v1 of my 'ASoC: simple-card: simplify code': I did not initialize the pointer to the asoc_simple_card_dai_init() function when DT. Then, I fixed that, and the simple card does not work for me.

First, without any 'format' in the DT, I get a fmt for each CPU / CODEC
DAI: SND_SOC_DAIFMT_CBS_CFS is always set. Well, some code is executed for nothing, but this is not critical.

The main problem is about sysclk: when there is no clock in the DT, you get the clock of the CPU or CODEC DAI. In my system, two clocks are declared in the controller CPU DAI, but there is no .set_sysclk pointer. So, snd_soc_dai_set_sysclk() returns -EINVAL and the card is not created.

As I don't know why you need these fmt and sysclk, may you have a look at these problems?

Signature · 2 lines
Ken ar c'hentañ	|	      ** Breizh ha Linux atav! **
Jef		|		http://moinejf.free.fr/
#1 lore raw
Quoted 1 line · Jean-Francois Moine

Xiubo,

Adding Morimoto-san.

Quoted 22 lines

I did a mistake in the v1 of my 'ASoC: simple-card: simplify code': I did not initialize the pointer to the asoc_simple_card_dai_init() function when DT. Then, I fixed that, and the simple card does not work for me.

First, without any 'format' in the DT, I get a fmt for each CPU / CODEC
DAI: SND_SOC_DAIFMT_CBS_CFS is always set. Well, some code is executed for nothing, but this is not critical.

The main problem is about sysclk: when there is no clock in the DT, you get the clock of the CPU or CODEC DAI. In my system, two clocks are declared in the controller CPU DAI, but there is no .set_sysclk pointer. So, snd_soc_dai_set_sysclk() returns -EINVAL and the card is not created.

As I don't know why you need these fmt and sysclk, may you have a look at these problems?

Signature · 2 lines
Ken ar c'hentañ	|	      ** Breizh ha Linux atav! **
Jef		|		http://moinejf.free.fr/
#2 lore raw
In reply to Jean-Francois Moine
Quoted 5 lines · Jean-Francois Moine

The main problem is about sysclk: when there is no clock in the DT, you get the clock of the CPU or CODEC DAI. In my system, two clocks are declared in the controller CPU DAI, but there is no .set_sysclk pointer. So, snd_soc_dai_set_sysclk() returns -EINVAL and the card is not created.

set_sysclk() should be returning -ENOTSUPP in this case as with set_fmt() and the generic code then ignoring that error.

Quoted 2 lines

As I don't know why you need these fmt and sysclk, may you have a look at these problems?

The overwhelming majority of combinations of devices need a format specifying (usually one will be imposed by the system design even if both devices could be master).

#3 lore raw

Hi

Quoted 9 lines
Quoted 5 lines

The main problem is about sysclk: when there is no clock in the DT, you get the clock of the CPU or CODEC DAI. In my system, two clocks are declared in the controller CPU DAI, but there is no .set_sysclk pointer. So, snd_soc_dai_set_sysclk() returns -EINVAL and the card is not created.

set_sysclk() should be returning -ENOTSUPP in this case as with set_fmt() and the generic code then ignoring that error.

Quoted 6 lines
Quoted 2 lines

As I don't know why you need these fmt and sysclk, may you have a look at these problems?

The overwhelming majority of combinations of devices need a format specifying (usually one will be imposed by the system design even if both devices could be master).

If my understanding is correct, this patch is the reason ?

71467e46414d3bab220de77d3d085be0c0aa03e1
(ASoC: simple-card: Add device's module clock selection)

It seems break non .set_sysclk drivers.

#4 lore raw
In reply to Jean-Francois Moine

Hi,

Quoted 9 lines

I did a mistake in the v1 of my 'ASoC: simple-card: simplify code': I did not initialize the pointer to the asoc_simple_card_dai_init() function when DT. Then, I fixed that, and the simple card does not work for me.

First, without any 'format' in the DT, I get a fmt for each CPU / CODEC
DAI: SND_SOC_DAIFMT_CBS_CFS is always set. Well, some code is executed for nothing, but this is not critical.

This is the old issue that I have raised before, which is one limitation of snd_soc_of_parse_daifmt(). As descript above, if without any fmt in the DT node, and just using snd_soc_of_parse_daifmt() for each CPU/CODEC DAI fmt the SND_SOC_DAIFMT_CBS_CFS (none zero) is always set.

@Mark and Morimoto-san, there has two ways to deal with this in my mind:
Fist, in the snd_soc_of_parse_daifmt(), modifying the
boolen : "[prefix]bitclock-master"
boolen : "[prefix]frame-master"

	==> 

[prefix]master = "XXX" and "XXX" is :
"cbm-cfm" --> SND_SOC_DAIFMT_CBM_CFM
"cbs-cfm" --> SND_SOC_DAIFMT_CBS_CFM
"cbm-cfs" --> SND_SOC_DAIFMT_CBM_CFS
"cbs-cfs" --> SND_SOC_DAIFMT_CBS_CFS

And the default value will be zero...

If this method is applied, and there is no need any masks for snd_soc_of_parse_daifmt(), like in simple card driver:

/* get CPU/CODEC common format via simple-audio-card,format */ info->daifmt = snd_soc_of_parse_daifmt(node, "simple-audio-card,") &

           (SND_SOC_DAIFMT_FORMAT_MASK | SND_SOC_DAIFMT_INV_MASK);

Second, if there hasn't any DAI fmts in DT node and at the same time the CPU and CODEC DAI devices don't need to care about of them, and set_dai() pointer could be set to be NULL, and then snd_soc_dai_set_fmt() just returned -ENOTSUPP.

Thanks,