Re: sh: replace init_IRQ_ipr() and init_IRQ_intc2() with plat_irq_setup()

[Magnus D. <mag...@gm...>]

On 7/18/07, Paul Mundt <le...@li...> wrote:
> On Wed, Jul 18, 2007 at 10:23:05AM +0900, Magnus Damm wrote:
> > --- 0001/arch/sh/kernel/irq.c
> > +++ work/arch/sh/kernel/irq.c 2007-07-17 17:01:58.000000000 +0900
> > @@ -254,12 +254,9 @@ void __init init_IRQ(void)
> >       init_IRQ_pint();
> >  #endif
> >
> > -#ifdef CONFIG_CPU_HAS_INTC2_IRQ
> > -     init_IRQ_intc2();
> > -#endif
> > -
> > -#ifdef CONFIG_CPU_HAS_IPR_IRQ
> > -     init_IRQ_ipr();
> > +#if defined(CONFIG_CPU_HAS_INTC2_IRQ) || defined(CONFIG_CPU_HAS_INTC2_IRQ) || \
> > +    defined(CONFIG_CPU_HAS_INTC_IRQ)
> > +     plat_irq_setup();
> >  #endif
> >
>
> Eek. How about a weak dummy symbol instead? You've also double-cased
> CPU_HAS_INTC2_IRQ, whereas I assume you meant to make one of them
> CPU_HAS_IPR_IRQ. On the other hand, is it really a valid state for any
> CPU not to have at least one of INTC2/INTC/IPR? In which case, make the
> reference to the symbol explicit, and let the linker whine.

Yeah, both ugly and incorrect - great. Sorry about that. I was
thinking along the lines of a weak dummy symbol, but then I hesitated
because I didn't know if it was needed at all. And then I looked at
for instance setup-sh7300.c which doesn't seem to have any interrupt
controller setup code at all...

I guess the best solution is to require plat_irq_setup() to be present
and fix the broken platforms.

Thanks.

/ magnus