[PATCH v3] ARM: i.MX27 pcm970: Add camera support

[broonie@opensource.wolfsonmicro.com (Mark Brown)]

On Wed, Aug 18, 2010 at 10:05:25PM +0200, Guennadi Liakhovetski wrote:

> 1. unbalanced regulator enable. IMHO, this is no exception to the general 
> enable / disable, alloc / free, lock / unlock etc. call balancing rule. If 
> your regulator has to stay enabled for the whole system run-time, you 
> don't have to (even try to) enable it multiple times. Please, either make 
> enabling / disabling symmetric, or leave only one enable in the 
> initialisation call.

If the regulator needs to be on all the time on a given board then
forcing this with always_on in the constraints is much more idiomatic
and straightforward.

> > +static int pcm970_camera_power_bus(struct device *dev, int toggle)
> > +{
> > +	struct regulator *regulator;
> > +
> > +	regulator = regulator_get(NULL, "imx_cam_vcc");
> > +	if (IS_ERR(regulator)) {
> > +		pr_err("unable to get regulator: %ld\n", PTR_ERR(regulator));
> > +		return -ENODEV;
> > +	} else {
> > +		if (!regulator_is_enabled(regulator) && toggle)
> > +			regulator_enable(regulator);
> > +	}
> > +	return 0;
> > +}

This is all very suspicious - your regulator_get() should be using the
struct device it was passed to look up the regulator by supply name
(probably "cam_vcc").  The regulator_is_enabled() call is going to break
if the supply is shared, too.

It all looks like something that I'd expect the camera driver to be
taking care of for itself, there's nothing board specific in here.

> > +	regulator = regulator_get(NULL, "imx_cam_vcc");
> > +	if (IS_ERR(regulator)) {
> > +		pr_err("unable to get regulator: %ld\n", PTR_ERR(regulator));
> > +		return -ENODEV;
> > +	} else if (!regulator_is_enabled(regulator)) {
> > +		regulator_enable(regulator);
> > +	}

Again, this looks like you need to add regulator support to the camera
driver.