Hi Thomas,

On 09/14/2016 11:19 AM, Thomas Gleixner wrote:

On Tue, 13 Sep 2016, Alexandre Torgue wrote:

quoted

On 09/13/2016 05:21 PM, Thomas Gleixner wrote:

quoted

On Fri, 9 Sep 2016, Alexandre TORGUE wrote:

quoted

+static void stm32_exti_free(struct irq_domain *d, unsigned int virq,
+			    unsigned int nr_irqs)
+{
+	struct irq_data *data = irq_get_irq_data(virq);
+
+	irq_gc_mask_clr_bit(data->parent_data);

I have a hard time to understand this. The irq domain is not hierarchical.

Actually, I wanted to test ".free" callback function of gpio_irq_domain in
STM32 pinctrl driver. To do that I modified gpio driver: just after getting
virq through gpio_to_irq, I called "irq_dispose_mapping(virq)".
I know it is dirty but I thought it was the only way to test.

Doing that, I see that ".free" callback of gpio domain is called but as it is
hirerchical ".free" callback for parent domain (exti one) is also called. I
observed that virq was well unmapped, but not masked at exti level. It is for
this reason than I added "irq_gc_mask_clr_bit(data->parent_data);" which mask
interrupt at exti level.

Aargh. I really misread the patch, but this is entirely non obvious and you
should do:

       struct irq_data *data = irq_domain_get_irq_data(d, virq);

       irq_gc_mask_clr_bit(d);

Then it is entirely clear that you mask the interrupt of _this_ (the exti)
domain.

Ok, it's easier to understand like that.

Now what really bugs me is that you do that at all. An interrupt which is
freed must be masked already. Why is it unmasked in the first place?

Honestly I don't know. When "devm_free_irq" is called to release virq, 
there is no issue and interrupt is well masked. But, when I tried to use
"irq_dispose_mapping(virq)" I observed that .free is called (child and 
parent domain) but interrupt is not masked.

Regards

Alex

Thanks,

	tglx