Re: [PATCH] ASoC: TWL4030: PM fix for output amplifiers

[Mark Brown <broonie@opensource.wolfsonmicro.com>]

On Mon, Mar 22, 2010 at 04:46:08PM +0200, Peter Ujfalusi wrote:
> On Monday 22 March 2010 16:15:44 ext Mark Brown wrote:

> > I don't see any fundamental problem here - mostly this just maps on to
> > "if the widget is powered on write this value, otherwise write the
> > value configured by the control.".

> Hmm, I suppose it could be possible to pass the name of the DAPM widget, a 
> bitmask to a control...
> Than in the handler in the core we could walk through the DAPM widgets, and find 
> the one, check the status, and if it is on, we allow the write, otherwise we 
> could use the mask to write only the things, that it is needed.

Something like that, yes - the walk could happen at startup time to
avoid having to do it repeatedly.

> But, in TWL case I'd like to filter out also the writes which is done by the 
> DAPM widgets (which is visible from the user space, the mixers).
> As it is now, if user changes the mixer, which is associated with DAPM widget 
> (and it is different than the PGA, they are DAPM_MIXER), than that write goes 
> directly to the register, and this write takes the cached value, makes the 
> changes and than writes it to register. This also enables the gain (which 
> enables the amp, which takes power).

Remember, the register cache is below the controls and transparent to
them - if the controls haven't written a value to the chip then it will
not appear in the register cache so other controls will not be affected.

> > Well, I was rather hoping I could convince you to make this more
> > generic.  But if that's not possible then yes, please do make the commit
> > message clearer.

> I need to study other codes as well, before I could start doing something more 
> generic, which I totally agree would be beneficial for others as well.
> But for now, I would like to update the commit message, and come back later, and 
> revisit the possibility of similar, and more generic ways of doing this.

I guess.