[PATCH 07/27] irqchip: Declare cortex-a7's irqchip to initialize gic from dt

[Marc Zyngier]

On Thu, Apr 10 2014 at 11:42:56 am BST, armdev <armdev.ftm at gmail.com> wrote:
> On 10-Apr-2014, at 4:11 pm, Marc Zyngier <marc.zyngier at arm.com> wrote:
>
>> On Thu, Apr 10 2014 at 11:30:41 am BST, armdev <armdev.ftm at gmail.com> wrote:
>>> On 10-Apr-2014, at 3:51 pm, Marc Zyngier <marc.zyngier at arm.com> wrote:
>>> 
>>>> On Thu, Apr 10 2014 at 11:09:02 am BST, armdev <armdev.ftm at gmail.com> wrote:
>>>>> On 10-Apr-2014, at 3:34 pm, Marc Zyngier <marc.zyngier at arm.com> wrote:
>>>>> 
>>>>>> On Thu, Apr 10 2014 at 10:28:24 am BST, Chanwoo Choi <cw00.choi at samsung.com> wrote:
>>>>>>> This patch declare coretex-a7's irqchip to initialze gic from dt
>>>>>>> with "arm,cortex-a7-gic" data.
>>>>>>> 
>>>>>>> Cc: Thomas Gleixner <tglx at linutronix.de>
>>>>>>> Signed-off-by: Chanwoo Choi <cw00.choi at samsung.com>
>>>>>>> Signed-off-by: Kyungmin Park <kyungmin.park at samsung.com>
>>>>>>> ---
>>>>>>> drivers/irqchip/irq-gic.c | 1 +
>>>>>>> 1 file changed, 1 insertion(+)
>>>>>>> 
>>>>>>> diff --git a/drivers/irqchip/irq-gic.c b/drivers/irqchip/irq-gic.c
>>>>>>> index 4300b66..8e906e4 100644
>>>>>>> --- a/drivers/irqchip/irq-gic.c
>>>>>>> +++ b/drivers/irqchip/irq-gic.c
>>>>>>> @@ -1069,6 +1069,7 @@ gic_of_init(struct device_node *node, struct device_node *parent)
>>>>>>> }
>>>>>>> IRQCHIP_DECLARE(cortex_a15_gic, "arm,cortex-a15-gic", gic_of_init);
>>>>>>> IRQCHIP_DECLARE(cortex_a9_gic, "arm,cortex-a9-gic", gic_of_init);
>>>>>>> +IRQCHIP_DECLARE(cortex_a7_gic, "arm,cortex-a7-gic", gic_of_init);
>>>>>>> IRQCHIP_DECLARE(msm_8660_qgic, "qcom,msm-8660-qgic", gic_of_init);
>>>>>>> IRQCHIP_DECLARE(msm_qgic2, "qcom,msm-qgic2", gic_of_init);
>>>>>> 
>>>>>> Frankly, this patch adds no value. Are we going to add
>>>>>> "arm,cortex-a12-gic", "arm,cortex-a17-gic", "arm,cortex-a53-gic",
>>>>>> "arm,cortex-a57-gic"? And that's just to mention the ARM Ltd cores...
>>>>>> 
>>>>>> Instead, how about defining a generic "arm,gic" property, and mandate
>>>>>> that new DT files are using that? We can always use a more precise
>>>>>> compatible for quirks.
>>>>>> 
>>>>> 
>>>>> How about keeping it simple and tied to arm gic versions
>>>>> arm,gicv1, arm,gicv2, arm,gicv2ve
>>>> 
>>>> That's a variation on the same theme. As for GICv2, we don't need to
>>>> distinguish between having the Virtualization Extentions, the binding
>>>> already allows you to tell one from the other.
>>>> 
>>> So if there be just 2 types of gic, it would be simple.
>> 
>> Not exactly. We just happen to support two revisions of the GIC
>> architecture with the same binding. GICv3 has an entierely separate
>> binding.
>> 
>>> gicv1 - 2 address sets (gicc and gicd)
>> 
>> Yes.
>> 
>>> gicv2 - 4 sets (gicc gicd gicv gich) and 1 maintenance interrupt. Right?
>> 
>> No.
>> 
>> The presence of the GICV, GICH and maintenance interrupt are indicative
>> of the support for the Virtualization Extentions. GICv2 itself can
>> perfectly be built without it.
>
> then does gicv2-ve makes sense ?

Read what I just wrote. You find the GICV region, you have the VE
extensions. You don't find them, they are not present. No need for an
overloaded compatible string, they both conform to the same
*Architecture Spec*.

	M.
-- 
Jazz is not dead. It just smells funny.