Re: [PATCH V2 05/15] genirq: Generic chip: Add big endian I/O accessors

From: Thomas Gleixner
Date: Thu Oct 30 2014 - 08:30:20 EST


On Thu, 30 Oct 2014, Arnd Bergmann wrote:
> On Wednesday 29 October 2014 19:17:58 Kevin Cernekee wrote:
> > static LIST_HEAD(gc_list);
> > static DEFINE_RAW_SPINLOCK(gc_lock);
> >
> > +static int is_big_endian(struct irq_chip_generic *gc)
> > +{
> > + return !!(gc->domain->gc->gc_flags & IRQ_GC_BE_IO);
> > +}
> > +
> > static void irq_reg_writel(struct irq_chip_generic *gc,
> > u32 val, int reg_offset)
> > {
> > - writel(val, gc->reg_base + reg_offset);
> > + if (is_big_endian(gc))
> > + iowrite32be(val, gc->reg_base + reg_offset);
> > + else
> > + writel(val, gc->reg_base + reg_offset);
> > }
> >
>
> What I had in mind was to use indirect function calls instead, like
>
> #ifndef irq_reg_writel
> static inline void irq_reg_writel_le(u32 val, void __iomem *addr)
> {
> return writel(val, addr);
> }
> #endif
>
> #ifndef irq_reg_writel_be
> static inline void irq_reg_writel_be(u32 val, void __iomem *addr)
> {
> return iowrite32_be(val, addr);
> }
> #endif
>
>
> static inline void irq_reg_writel(struct irq_chip_generic *gc, u32 val, int reg_offset)
> {
> if (IS_ENABLED(CONFIG_GENERIC_IRQ_CHIP) &&

That's inside of the generic irq chip, so CONFIG_GENERIC_IRQ_CHIP is
always set when this is compiled.

> !IS_ENABLED(CONFIG_GENERIC_IRQ_CHIP_BE))
> return irq_reg_writel_le(val, gc->reg_base + reg_offset);
>
> if (IS_ENABLED(CONFIG_GENERIC_IRQ_CHIP) &&
> !IS_ENABLED(CONFIG_GENERIC_IRQ_CHIP_BE))

s/!// ?

> return irq_reg_writel_be(val, gc->reg_base + reg_offset);

I don't think the above will cover all combinations.

..._CHIP_BE ...CHIP_LE
N N ; Default behaviour: readl/writel
Y N ; ioread/write32be
N Y ; Default behaviour: readl/writel
Y Y ; Runtime selected

> return gc->writel(val, gc->reg_base + reg_offset);
> }
>
> This would take the condition out of the callers.

So you trade a conditional for an indirect call. Not sure what's more
expensive. The indirect call is definitely a smaller text footprint,
so we should opt for this.

Thanks,

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