[RFC PATCH] irq: handle private interrupt registration

[felipe.balbi@nokia.com (Felipe Balbi)]

Hi,

On Wed, Jun 02, 2010 at 01:14:32AM +0200, ext Thomas Gleixner wrote:
>1) Use PER_CPU offsets for the irq numbers. The generic irq code does
>   not care whether the interrupt number is matching any physical
>   numbering scheme in the hardware, as long as the arch specific chip
>   implementation knows how to deal with it, which is not rocket
>   science to do.

FWIW, I think (1) is a better approach as the problem will vanish 
altogether and to me it sounds like the simpler approach as well. Archs 
which have more than one IRQ chip (like OMAP with the twl4030 family) 
will already use sequencial numbering anyway, so using the same approach 
for N cpus, to me, sounds like a good deal.

While at that, a question from my side: do we have a generic way of 
fetching the last IRQ number so we can easily use that to calculate the 
physical number of the IRQ line on the chip ?

On OMAP, we have been passing that number down to twl4030 via 
platform_data, but it would be better to ask genirq to tell us which was 
the last irq number "claimed".

-- 
balbi

DefectiveByDesign.org