On Fri, Mar 28, 2014 at 12:40:09PM +0100, Daniel Mack wrote:
> On 03/28/2014 12:35 PM, Mark Brown wrote:

> > It'd be better style to enable the regulators during probe to read the
> > ID register rather than deferring.

> I had it that way, but that meant I had to introduce device-level pm
> functions in parallel to the soc-level ones, in order to really bring
> down the supply in suspend, even in setups where the codec is not
> actually used.

> Nevermind, I can change it back. Will resend the patches for tas5086 and
> ak4104.

The ususal thing is to power on and off during the probe (though this is
mainly for devices which actively manage the regualtor at runtime rather
than just leaving the regulator on while the sound card is active.