[alsa-devel] ASoC: TLV320AIC3x: Adding additional functionality for 3106 with [Patch] for discuss

[Timur Karaldin]

16.03.2016 18:23, Peter Ujfalusi пишет:
> On 03/16/2016 12:26 PM, Timur Karaldin wrote:
>> Hi Peter!
>>
>> 16.03.2016 11:57, Peter Ujfalusi пишет:
>>> Hi Timur,
>>>
>>> please avoid top posting.
>>>
>>> On 03/15/16 18:13, Timur Karaldin wrote:
>>>> Ok, I choose the way #1 without changing kcontrol names.
>>>> Another thing I don't understand is the way for caching gain volume if I have
>>>> switches muted. Is there any proper way for storing cached values between gain
>>>> and switch handlers includes soft reset or I need manually check kcontrol name
>>>> in handler and cache values in extended private structure? The same question
>>>> for switch handler, because I need to share cached values between these
>>>> handlers.
>>> Since the problematic registers are: 17, 18, .. 24. I would have an array
>>> inside struct aic3x_priv to store the configured gain. You could store the
>>> mute as well, but since regmap is caching the content I would simply read back
>>> the gain value and if it is 0xf, it is muted, so update only the local shadow,
>>> if it is not muted, write the gain.
>>> In the mute/unmute controls I would then save the gain to shadow when muting
>>> the path and write the cached gain to the register when unmuting it. You
>>> should also return correct 0/1 for the unmute state and not the actual
>>> register value.
> your mail client does not seem to wrap the lines correctly, can you check that.
I have no idea how these lines should looks, so it's very hard for me to 
see what's wrong. Could you point me how it looks in original?
>
>> Actually I try to avoid complex handler, with switch-case which depends on reg
>> number or kcontrol name inside of it.
>> I have found similar discussion with you about one year ago here
>> http://mailman.alsa-project.org/pipermail/alsa-devel/2015-February/087430.html. There
>> is a proposal to use TLV_DB_RANGE_HEAD() and TLV_DB_SCALE_ITEM() because of
>> invert with SOC_DAPM_SINGLE_TLV.
> As I recall there were some issues with that back then, but can not recall. At
> least I don't see followup patches regarding to that.
>
>> As I understand we could make tlv with using
>> 0 as mute, skip values 1-6, use values 7-15 (7 = -12dB and 15=0dB) then
>> declare SOC_DAPM_SINGLE_TLV as "Line1L Volume" for example. Is it still
>> actual?
> Worth a try ;)
> The naming needs to be different, take a look at twl4030 codec's
> twl4030_dapm_dbypass_tlv and how it is used.
>
>> Few things which I didn't understand:
>> - should I use "Line1L Switch" together with "Line1L Volume" in this case. I
>> think no, because DAPM_SINGLE_TLV should replace switch.
> Since we will have the gain and mute with volume control we will have only one
> control.
>
>> - i don't understand how declare tlv with skipping values 1-6
> Yeah, I'm not sure how leaving 'hole' between the TLV_DB_SCALE_ITEMs will be
> handled if it works.
>
>> Cheers, Tim
Ok, now it's much more clear for me.
Another question is register behaviour during soft reset. There is 
"aic3x_set_power" handle. In this handle kernel makes SOFT_RESET, markes 
cache as dirty, then power down the codec for handle power down request.
But as I could see main volumes like "HP DAC" and "PCM" stores values 
between close and open in mixer and I could not see any code to handle 
it. On the other hand my controls do not save states, as you mentioned 
because of SOFT_RESET, could you explain such different behaviour?

Cheers,
Tim