On Fri, Aug 05, 2011 at 09:06:20AM +0100, Ben Dooks wrote:

I would say that setting the interrupt should deal with all the necessary
configuration to get that interrupt working. In the case of pretty much all
of the SoCs I have been involved with have always set the GPIO's function
to allow the interrupt to pass through.

That's what Stephen is trying to do, essentially.  It's looking like
it's more sensible to fix it in the Tegra interrupt drivers, though.