On Mon, May 14, 2012 at 07:06:50PM +0900, Chanwoo Choi wrote:
> On 05/14/2012 06:57 PM, Mark Brown wrote:
> > On Mon, May 14, 2012 at 02:20:35PM +0900, Chanwoo Choi wrote:

> >> +	if (!max77693->irq) {
> >> +		dev_warn(max77693->dev, "No interrupt specified.\n");
> >> +		return 0;
> >> +	}

> > I'd suggest just setting up the IRQ domain anyway - it makes life
> > simpler in the other drivers for the chip as they don't need to worry if
> > interrupts are registered unless it's very important to them that they
> > go off which is often not the case.  Just skip actually claiming the
> > IRQ.

> This patch add support irq domain for MAX77693 in the
> max77693_irq_init() function.
> You can check on patchset. Then, The 'max77693->irq' is gpio pin for
> IRQ. max77693->irq
> is meant to indicate to the application processor that the status of
> MAX77693 has changed.

Sure.  What I'm saying is that it's probably going to be easier all
round if you run through all the init except actually calling
request_threaded_irq() (and unmasking the external interrupt if you have
control of that) - that way the interrupt numbers for the interrupts on
the PMIC can always be requested, even if they can't actually occur.
Only child drivers that *really* need the interrupts to happen will need
to have conditional code.