Re: [PATCH] irqdomain: Initialize number of IRQs for simple domains

[Rob Herring <robherring2-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>]

On 01/06/2012 10:07 AM, David Brown wrote:
> On Fri, Jan 06, 2012 at 03:28:25PM +0100, Thierry Reding wrote:
>> diff --git a/arch/arm/mach-msm/board-msm8x60.c b/arch/arm/mach-msm/board-msm8x60.c
>> index 0a11342..a50c7e2 100644
>> --- a/arch/arm/mach-msm/board-msm8x60.c
>> +++ b/arch/arm/mach-msm/board-msm8x60.c
>> @@ -84,8 +84,11 @@ static void __init msm8x60_dt_init(void)
>>  
>>  	node = of_find_matching_node_by_address(NULL, msm_dt_gic_match,
>>  			MSM8X60_QGIC_DIST_PHYS);
>> -	if (node)
>> -		irq_domain_add_simple(node, GIC_SPI_START);
>> +	if (node) {
>> +		struct irq_domain *domain = irq_domain_add_simple(node,
>> +				GIC_SPI_START, NR_MSM_IRQS);
>> +		WARN_ON(IS_ERR(domain));
>> +	}
>>  
>>  	if (of_machine_is_compatible("qcom,msm8660-surf")) {
>>  		printk(KERN_INFO "Init surf UART registers\n");
> 
> This is probably a consequence of MSM not really being "simple", but
> just using that.  However, NR_MSM_IRQS is only the number of IRQs on
> the MSM core.  There are also GPIO irqs, and potentially board IRQs
> (the board has an I2C-based chip with a bunch of IRQ lines on it).
> 
> The only define that captures this now is 'NR_IRQS', even though we're
> trying to get rid of that.
> 
> Ultimately, the correct answer will be to get the various interrupt
> controllers using their own domains, but for now, this needs to be a
> larger value to avoid missing a bunch of the interrupts.

No. This should only be the number of interrupts for a controller as the
interrupt numbers in the device tree should be relative to a controller
and not the Linux virq number. The numbering in the dts needs to be
correct. You don't need a domain until you start getting the interrupts
from the dts.

Rob