Kevin,
[...]
>
> > set_gpio_trigger(bank, gpio, trigger);
> > } else if (bank->regs->irqctrl) {
> > reg += bank->regs->irqctrl;
> > @@ -295,13 +296,7 @@ static int _set_gpio_triggering(struct gpio_bank
> *bank, int gpio, int trigger)
> > if (trigger & IRQ_TYPE_EDGE_FALLING)
> > l |= 1 << (gpio << 1);
> >
> > - if (trigger)
> > - /* Enable wake-up during idle for dynamic tick */
> > - __raw_writel(1 << gpio, bank->base
> > - + bank->regs->wkup_set);
> > - else
> > - __raw_writel(1 << gpio, bank->base
> > - + bank->regs->wkup_clear);
> > + MOD_REG_BIT(bank->regs->wkup_status, 1 << gpio, trigger != 0);
>
> This isn't right, so I'm not sure how this is working. At this point:
>
> base = bank->base;
> reg = bank->base + bank->regs->edgectrl[12];
Right so far...
>
> but if you look at MOD_REG_BIT(), it does register access using 'base +
> reg', but here that will be 'bank->base + bank->base + reg offset',
> which will be doing a read/modify/write to who-knows-where.
My understanding was MOD_REG_BIT(bank->regs->wkup_status, ...) translate to:
bank->base + bank->regs->wkup_status, for the following reason:
In the following macro, reg is equivalent to bank->regs->wkup_status.
But this is not the case with base, which remains as bank->base.
Please correct me.
#define MOD_REG_BIT(reg, bit_mask, set) \
do { \
int l = __raw_readl(base + reg); \
if (set) l |= bit_mask; \
else l &= ~bit_mask; \
__raw_writel(l, base + reg); \
} while(0)
>
> > __raw_writel(l, reg);
>
> Oh, now I see why it works: because you have an additional write here,
> which isn't right either because MOD_REG_BIT() already does the read
> *and* write.
This should be writing to following still.
reg = bank->base + bank->regs->edgectrl[12];
--
Tarun
>
[...]
--
To unsubscribe from this list: send the line "unsubscribe linux-omap" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
