Re: [PATCH v4 1/6] ARM: at91/aic: add irq domain and device tree support

[Nicolas Ferre <nicolas.ferre-AIFe0yeh4nAAvxtiuMwx3w@public.gmane.org>]

Hi Grant,

Thanks for your review.

Some answer following...

On 01/04/2012 08:40 PM, Grant Likely :

[..]

>> --- a/arch/arm/mach-at91/irq.c
>> +++ b/arch/arm/mach-at91/irq.c
>> @@ -24,6 +24,10 @@
>>  #include <linux/module.h>
>>  #include <linux/mm.h>
>>  #include <linux/types.h>
>> +#include <linux/of_address.h>
>> +#include <linux/of_irq.h>
>> +#include <linux/irqdomain.h>
>> +#include <linux/err.h>
>>  
>>  #include <mach/hardware.h>
>>  #include <asm/irq.h>
>> @@ -34,22 +38,28 @@
>>  #include <asm/mach/map.h>
>>  
>>  void __iomem *at91_aic_base;
>> +static struct irq_domain at91_aic_domain;
>> +
>> +static inline unsigned int aic_irq(struct irq_data *d)
>> +{
>> +	return d->hwirq;
>> +}
> 
> This is a useless macro.  Drop it.

Ok. That was overkill ;-)

[..]

>> @@ -90,13 +100,13 @@ static u32 backups;
>>  
>>  static int at91_aic_set_wake(struct irq_data *d, unsigned value)
>>  {
>> -	if (unlikely(d->irq >= 32))
>> +	if (unlikely(aic_irq(d) >= at91_aic_domain.nr_irq))
> I don't think you need to derefernce the irq_domain.nr_irq here.  Just
> keep the test for >= 32.

Well I would like to keep it referring to this value as we may increase
the number of irq handled by this controller in the future...

[..]

>> +#if defined(CONFIG_OF)
>> +static int __init __at91_aic_of_init(struct device_node *node,
>> +				     struct device_node *parent)
>> +{
>> +	at91_aic_base = of_iomap(node, 0);
>> +	at91_aic_domain.of_node = of_node_get(node);
>> +
>> +	return 0;
>> +}
>> +
>> +static const struct of_device_id aic_ids[] __initconst = {
>> +	{ .compatible = "atmel,at91rm9200-aic", .data = __at91_aic_of_init },
>> +	{ /*sentinel*/ }
>> +};
>> +
>> +static void __init at91_aic_of_init(void)
>> +{
>> +	of_irq_init(aic_ids);
>> +}
>> +#else
>> +static void __init at91_aic_of_init(void) {}
>> +#endif
>> +
>>  /*
>>   * Initialize the AIC interrupt controller.
>>   */
>>  void __init at91_aic_init(unsigned int priority[NR_AIC_IRQS])
>>  {
>>  	unsigned int i;
>> +	int hwirq, irq;
>>  
>> -	at91_aic_base = ioremap(AT91_AIC, 512);
>> +	if(of_have_populated_dt())
>> +		at91_aic_of_init();
>> +	else
>> +		at91_aic_base = ioremap(AT91_AIC, 512);
>>  
>>  	if (!at91_aic_base)
>> -		panic("Impossible to ioremap AT91_AIC\n");
>> +		panic("Unable to ioremap AIC registers\n");
>> +
>> +	/* Add irq domain for AIC */
>> +	at91_aic_domain.nr_irq = NR_AIC_IRQS;
>> +	at91_aic_domain.irq_base = irq_alloc_descs(-1, AIC_BASE_IRQ,
>> +						at91_aic_domain.nr_irq, 0);
>> +	if (IS_ERR_VALUE(at91_aic_domain.irq_base)) {
>> +		WARN(1, "Cannot allocate irq_descs, assuming pre-allocated\n");
>> +		at91_aic_domain.irq_base = AIC_BASE_IRQ;
>> +	}
>> +	at91_aic_domain.ops = &irq_domain_simple_ops;
>> +	irq_domain_add(&at91_aic_domain);
> 
> Nothing special is being done with the irq domain.  Please use
> irq_domain_add_simple().  It will make things easier for the next
> round of irq_domain patches that I'm getting ready to post.

Fair enough, but I have one question: how do I specify "nr_irq" in the
domain created by irq_domain_add_simple()? Maybe it is not needed?


>>  	/*
>>  	 * The IVR is used by macro get_irqnr_and_base to read and verify.
>>  	 * The irq number is NR_AIC_IRQS when a spurious interrupt has occurred.
>>  	 */
>> -	for (i = 0; i < NR_AIC_IRQS; i++) {
>> +	irq_domain_for_each_irq((&at91_aic_domain), hwirq, irq) {
>> +
> 
> Same here; the driver is written for a fixed number of irqs
> (NR_AIC_IRQs).  You don't need to reference the irq_domain for that.

Yes, Ditto: this fixed number may change in the future...

> Otherwise;
> 
> Acked-by: Grant Likely <grant.likely-s3s/WqlpOiPyB63q8FvJNQ@public.gmane.org>

Ok, I will rework a new revision...

Thanks, best regards,
-- 
Nicolas Ferre