Re: [PATCH v2 1/2] irqchip: stm32: Optimizes and cleans up stm32-exti irq_domain

[Ludovic BARRE]

On 02/23/2018 04:46 PM, Ludovic BARRE wrote:
> hi Radoslaw
>
> The gpio-keys tests it's now functional on H7, great.
> However the gpio-key test only the bank1 (like stm32f429).
> Like the H7 introduce the multi-bank management,
> we must perform complementary test.
>
> comment below about ack in handler
>
> On 02/23/2018 09:31 AM, Radoslaw Pietrzyk wrote:
> > - discards setting handle_simple_irq handler for hierarchy interrupts
> > - removes acking in chained irq handler as this is done by
> > irq_chip itself inside handle_edge_irq
> > - removes unneeded irq_domain_ops.xlate callback
> >
> > Signed-off-by: Radoslaw Pietrzyk <radoslaw.pietrzyk@xxxxxxxxx>
> > ---
> > Â drivers/irqchip/irq-stm32-exti.c | 13 -------------
> > Â 1 file changed, 13 deletions(-)
> >
> > diff --git a/drivers/irqchip/irq-stm32-exti.c 
> > b/drivers/irqchip/irq-stm32-exti.c
> > index 36f0fbe..8013a87 100644
> > --- a/drivers/irqchip/irq-stm32-exti.c
> > +++ b/drivers/irqchip/irq-stm32-exti.c
> > @@ -79,13 +79,6 @@ static unsigned long stm32_exti_pending(struct 
> > irq_chip_generic *gc)
> > ÂÂÂÂÂ return irq_reg_readl(gc, stm32_bank->pr_ofst);
> > Â }
> > -static void stm32_exti_irq_ack(struct irq_chip_generic *gc, u32 mask)
> > -{
> > -ÂÂÂ const struct stm32_exti_bank *stm32_bank = gc->private;
> > -
> > -ÂÂÂ irq_reg_writel(gc, mask, stm32_bank->pr_ofst);
> > -}
> > -
> > Â static void stm32_irq_handler(struct irq_desc *desc)
> > Â {
> > ÂÂÂÂÂ struct irq_domain *domain = irq_desc_get_handler_data(desc);
> > @@ -106,7 +99,6 @@ static void stm32_irq_handler(struct irq_desc *desc)
> > ÂÂÂÂÂÂÂÂÂÂÂÂÂ for_each_set_bit(n, &pending, IRQS_PER_BANK) {
> > ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ virq = irq_find_mapping(domain, irq_base + n);
> > ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ generic_handle_irq(virq);
> > -ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ stm32_exti_irq_ack(gc, BIT(n));
>
> the while loop " while ((pending = " has been introduce while original
> upstream of this driver in V2 to V3
> https://patchwork.ozlabs.org/patch/604190/
> https://patchwork.ozlabs.org/patch/665242/
>
> the "ack" is needed to acknowledge the irq which are handled and which 
> is not the original irq for which we enter.
>
> > ÂÂÂÂÂÂÂÂÂÂÂÂÂ }
> > ÂÂÂÂÂÂÂÂÂ }
> > ÂÂÂÂÂ }
> > @@ -176,16 +168,12 @@ static int stm32_irq_set_wake(struct irq_data 
> > *data, unsigned int on)
> > Â static int stm32_exti_alloc(struct irq_domain *d, unsigned int virq,
> > ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ unsigned int nr_irqs, void *data)
> > Â {
> > -ÂÂÂ struct irq_chip_generic *gc;
> > ÂÂÂÂÂ struct irq_fwspec *fwspec = data;
> > ÂÂÂÂÂ irq_hw_number_t hwirq;
> > ÂÂÂÂÂ hwirq = fwspec->param[0];
> > -ÂÂÂ gc = irq_get_domain_generic_chip(d, hwirq);
> > ÂÂÂÂÂ irq_map_generic_chip(d, virq, hwirq);
> > -ÂÂÂ irq_domain_set_info(d, virq, hwirq, &gc->chip_types->chip, gc,
> > -ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ handle_simple_irq, NULL, NULL);
>
> I see if it is not disturb the multi-bank.

you are right, normally irq_domain_set_info is already set in
irq_map_generic_chip.

> > ÂÂÂÂÂ return 0;
> > Â }
> > @@ -200,7 +188,6 @@ static void stm32_exti_free(struct irq_domain *d, 
> > unsigned int virq,
> > Â struct irq_domain_ops irq_exti_domain_ops = {
> > ÂÂÂÂÂ .mapÂÂÂ = irq_map_generic_chip,
> > -ÂÂÂ .xlateÂÂÂ = irq_domain_xlate_onetwocell,
> >  .alloc = stm32_exti_alloc,
> > ÂÂÂÂÂ .freeÂÂÂ = stm32_exti_free,
> > Â };
>
> BR
> Ludo