Re: [PATCH] ASoC: DAPM: Add support for multi register mux

[Lars-Peter Clausen <lars@metafoo.de>]

On 04/01/2014 08:26 PM, Arun Shamanna Lakshmi wrote:
[...]
>>> diff --git a/sound/soc/soc-dapm.c b/sound/soc/soc-dapm.c index
>>> c8a780d..4d2b35c 100644
>>> --- a/sound/soc/soc-dapm.c
>>> +++ b/sound/soc/soc-dapm.c
>>> @@ -514,9 +514,9 @@ static int dapm_connect_mux(struct
>> snd_soc_dapm_context *dapm,
>>>    	unsigned int val, item;
>>>    	int i;
>>>
>>> -	if (e->reg != SND_SOC_NOPM) {
>>> -		soc_widget_read(dest, e->reg, &val);
>>> -		val = (val >> e->shift_l) & e->mask;
>>> +	if (e->reg[0] != SND_SOC_NOPM) {
>>> +		soc_widget_read(dest, e->reg[0], &val);
>>> +		val = (val >> e->shift_l) & e->mask[0];
>>>    		item = snd_soc_enum_val_to_item(e, val);
>>
>> This probably should handle the new enum type as well. You'll probably
>> need some kind of flag in the struct to distinguish between the two
>> enum types.
>
> Any suggestion on the flag name ?
>

How about 'onehot'?

[...]
>>> +		reg_val = BIT(bit_pos);
>>> +	}
>>> +
>>> +	for (i = 0; i < e->num_regs; i++) {
>>> +		if (i == reg_idx) {
>>> +			change = snd_soc_test_bits(codec, e->reg[i],
>>> +							e->mask[i],
>> reg_val);
>>> +
>>> +		} else {
>>> +			/* accumulate the change to update the DAPM
>> path
>>> +			    when none is selected */
>>> +			change += snd_soc_test_bits(codec, e->reg[i],
>>> +							e->mask[i], 0);
>>
>> change |=
>>
>>> +
>>> +			/* clear the register when not selected */
>>> +			snd_soc_write(codec, e->reg[i], 0);
>>
>> I think this should happen as part of the DAPM update sequence like
>> you had earlier. Some special care should probably be take to make
>> sure that you de-select the previous mux input before selecting the
>> new one if the new one is in a different register than the previous one.
>
> I am not sure I follow this part. We are clearing the 'not selected'
> registers before we set the one we want. Do you want us to loop the
> logic of soc_dapm_mux_update_power for each register ? or do you
> want to change the dapm_update structure so that it takes all the regs,
> masks, and values together ?

The idea with the dapm_update struct is that the register updates are done 
in the middle of the power-down and power-up sequence. So yes, change the 
dapm_update struct to be able to hold all register updates and do all 
register updates in dapm_widget_update. I think an earlier version of your 
patch already had this.