Re: [PATCH 2/9] irq-remove: arch non-trivial

[Jeff Garzik <jeff@garzik.org>]

Jeremy Fitzhardinge wrote:
> Jeff Garzik wrote:
> > commit 8d45690dd90b18daaa21b981ab20caf393220bf0
> > Author: Jeff Garzik <jeff@garzik.org>
> > Date:   Fri Oct 19 00:46:23 2007 -0400
> >
> >     [IRQ ARG REMOVAL] various non-trivial arch updates
> >
> >  arch/x86/kernel/vm86_32.c     |    3 ++-
> >  include/asm-x86/irq_regs_32.h |   25 +++++++++++++++++++++++++
> >  2 files changed, 27 insertions(+), 1 deletion(-)
> >
> > 8d45690dd90b18daaa21b981ab20caf393220bf0
> > diff --git a/arch/x86/kernel/vm86_32.c b/arch/x86/kernel/vm86_32.c
> > index 157e4be..18aae9e 100644
> > --- a/arch/x86/kernel/vm86_32.c
> > +++ b/arch/x86/kernel/vm86_32.c
> > @@ -739,10 +739,11 @@ static int irqbits;
> >  	| (1 << SIGUSR1) | (1 << SIGUSR2) | (1 << SIGIO)  | (1 << SIGURG) \
> >  	| (1 << SIGUNUSED) )
> >  	
> > -static irqreturn_t irq_handler(int intno, void *dev_id)
> > +static irqreturn_t irq_handler(void *dev_id)
> >  {
> >  	int irq_bit;
> >  	unsigned long flags;
> > +	unsigned int intno = get_irqfunc_irq();
> >  
> >  	spin_lock_irqsave(&irqbits_lock, flags);	
> >  	irq_bit = 1 << intno;
> > diff --git a/include/asm-x86/irq_regs_32.h b/include/asm-x86/irq_regs_32.h
> > index 3368b20..68a531d 100644
> > --- a/include/asm-x86/irq_regs_32.h
> > +++ b/include/asm-x86/irq_regs_32.h
> > @@ -26,4 +26,29 @@ static inline struct pt_regs *set_irq_regs(struct pt_regs *new_regs)
> >  	return old_regs;
> >  }
> >  
> > +DECLARE_PER_CPU(unsigned int, __irqfunc_irqs);
> > +
> > +static inline unsigned int get_irqfunc_irq(void)
> > +{
> > +	return __get_cpu_var(__irqfunc_irqs);
> > +}
> > +
> > +#if 0
> > +static inline unsigned int set_irqfunc_irq(unsigned int new_irq)
> > +{
> > +	unsigned int old_irq, *pirq = &__get_cpu_var(__irqfunc_irqs);
> > +
> > +	old_irq = *pirq;
> > +	*pirq = new_irq;
> > +	return old_irq;
> > +}
> > +#else
> > +static inline void set_irqfunc_irq(unsigned int new_irq)
> > +{
> > +	int *pirq = &__get_cpu_var(__irqfunc_irqs);
> > +
> > +	*pirq = new_irq;
> >   
>
> x86_write_percpu(__irqfunc_irqs, new_irq) would be slightly more
> efficient here.  Any why the pointer anyway?

Answering the latter question -- that's the way 
include/asm-generic/irq_regs.h was written, and I simply followed their 
lead.

One attribute of this method is to avoid preemption, which is probably 
necessary in the bowels of irq handling.

	Jeff



-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/