On 163, 06 12, 2003 at 09:34:06PM +1000, Anton Blanchard wrote:
>
> > I believe Andrey's IRQ consolidation provides a single flat IRQ
> > structure. Unfortunately, this doesn't reflect the reality that we
> > have on many ARM platforms - it remains the case that we need to
> > decode IRQs on a multi-level basis.
>
> Yes its still a flat structure. On ppc32/64 we offset the interrupts on
> the main controller to provide a space for ISA interrupts to go. Not
> great but it works for us.
May be I missed the point, but it isn't flat.
You can define HAVE_ARCH_IRQ_DESC and provide your own irq_desc(irq)
function which will translate irq number to the corresponding
irq_desc_t structure. You are free to implement any irq mappings
behind the irq_desc(). NR_IRQS is used only as maximal irq number.
So what is the problem ?
> One thing Paul suggested was to have a flag to mark an interrupt as a
> cascade in the irq descriptor. If its set then we also provide a
> get_irq() method (perhaps stashed away in the ->action field). That gives
> us nested interrupt handling in generic code. (assuming you can
> partition your irq numbers somehow)
>
> Anton
-- Andrey Panin | Linux and UNIX system administrator pazke@donpac.ru | PGP key: wwwkeys.pgp.net
This archive was generated by hypermail 2b29 : Mon Jun 23 2003 - 22:00:18 EST