Re: [PATCH V2 1/4] i2c: Add irq_gpio field to struct i2c_client.

[Arnd Bergmann <arnd-r2nGTMty4D4@public.gmane.org>]

On Friday 02 September 2011, Stephen Warren wrote:
> The idea was specifically to replace the need to call irq_to_gpio(i2c->irq).
> If we did just rename it plain "gpio" and allow it to be used for anything,
> then that does indeed start looking more like device-specific platform data.
> 
> I guess it sounds like consensus is to go that way. It does seem like that
> will end up creating a bunch more device-specific platform-data files though.
> I wonder if adding IORESOURCE_GPIO would make sense so this could be handled
> in a generic way without custom platform data types?

Interesting point. That's probably best for Grant to comment on, because
it depends on the long-term direction he wants to take with this.

I suppose that an IORESOURCE_GPIO makes a lot of sense if we expect
to keep having a flat system-wide gpio number space in the long run,
similar to irq numbers.

It would not fit well if we expect gpio numbers to be local to
a gpio controller, with no unique global identifier for them, similar
to how dma channels in the dma-engine subsystem are handled.

	Arnd