[PATCH 3/3] ARM: gic: add OF based initialization

[Rob Herring]

On 09/26/2011 04:11 PM, Jamie Iles wrote:
> On Mon, Sep 26, 2011 at 03:49:11PM -0500, Rob Herring wrote:
>> On 09/26/2011 02:57 PM, Jamie Iles wrote:
>>> Hi Rob,
>>>
>>> On Tue, Sep 20, 2011 at 03:24:04PM -0500, Rob Herring wrote:
>>> [...]
>>>> +int __init gic_of_init(struct device_node *node, struct device_node *parent)
>>>> +{
>>>> +	void __iomem *cpu_base;
>>>> +	void __iomem *dist_base;
>>>> +	int irq;
>>>> +	struct irq_domain *domain = &gic_data[gic_cnt].domain;
>>>> +
>>>> +	if (WARN_ON(!node))
>>>> +		return -ENODEV;
>>>> +
>>>> +	dist_base = of_iomap(node, 0);
>>>> +	WARN(!dist_base, "unable to map gic dist registers\n");
>>>> +
>>>> +	cpu_base = of_iomap(node, 1);
>>>> +	WARN(!cpu_base, "unable to map gic cpu registers\n");
>>>> +
>>>> +	domain->nr_irq = gic_irq_count(dist_base);
>>>> +	/* subtract off SGIs. Also subtract off PPIs for secondary GICs */
>>>> +	if (parent)
>>>> +		domain->nr_irq -= 32;
>>>> +	else
>>>> +		domain->nr_irq -= 16;
>>>> +
>>>> +	domain->irq_base = irq_alloc_descs(-1, 16, domain->nr_irq, numa_node_id());
>>>
>>> The way I understand irq_alloc_descs() (probably not very well) is that 
>>> having the irq parameter < 0 and the from parameter 16 means that it 
>>> needs to find domain->nr_irq descs starting from at least 16.  But if 
>>> the base is greater than 16, does this still work with the gic entry 
>>> macros as they are?
>>
>> No, but that would only happen if a platform calls irq_alloc_descs prior
>> to this code. The root controller must be initialized first (for other
>> reasons as well). There are no calls to irq_alloc_descs in arch/arm.
>>
>> With the MULTI_IRQ GIC support Marc Z is working on, we could make the
>> GIC irq mapping be completely dynamic. Although, there's probably not
>> much reason to do so for the root controller.
> 
> OK, that makes sense.  I think that if you were to do 
> irq_alloc_descs(16, 16, domain->nr_irqs, num_node_id()) then that will 
> guarantee the descriptors start from 16 (if they are available) which is 
> probably nicer.
> 
That would break secondary GICs though as it would always fail. For
secondary GIC, we skip SGIs and PPIs and allocate the first available
block of irq_desc's.

Rob