Rob Herring wrote at Thursday, August 04, 2011 7:55 PM:

On 08/04/2011 06:00 PM, Stephen Warren wrote:

...

Many IRQs are associated with GPIO pins. When the pin is used as an IRQ, it can't be used as anything else; it should be requested. Enhance the core interrupt code to call gpio_request() and gpio_direction_input() for any IRQ that is also a GPIO. This prevents duplication of these calls in each driver that uses an IRQ.

...

...

diff --git a/kernel/irq/manage.c b/kernel/irq/manage.c

...

...

@@ -978,6 +979,16 @@ __setup_irq(unsigned int irq, struct irq_desc *desc, struct irqaction *new) old = *old_ptr; } while (old); shared = 1;

• } else {

• gpio = irq_to_gpio(irq);
  

If you read the documentation for gpio, it is not recommended to use irq_to_gpio. There's only a handful of users. Part of the problem is it is platform specific and the gpio core cannot convert irq to gpio number

It seems like that's a soluble problem though?

I was thinking about adding a to_gpio function to struct irq_chip, as the inverse of struct gpio_chip's to_irq. Then, presumably any platform would be able to convert back from IRQ to GPIO, provided the platform called a new __irq_to_gpio from the platform-specific gpio_to_irq, just like most gpio_to_irq implementations defer to __gpio_to_irq.

I ended up not doing that in this patchset, since Tegra's gpio_to_irq function already works for the IRQs/GPIOs that were relevant for my testing, and I wanted to post a simple patch first to driver discussion.

Here is the relevant section:

Non-error values returned from irq_to_gpio() would most commonly be used with gpio_get_value(), for example to initialize or update driver state when the IRQ is edge-triggered. Note that some platforms don't support this reverse mapping, so you should avoid using it.