MMP2 Interrupt Controller forcing IRQ

[Andrei E. Warkentin]

2011/9/27 Mitch Bradley <wmb at firmworks.com>:
> On 9/26/2011 4:40 PM, Andrei E. Warkentin wrote:
>>
>> 2011/9/26 Andrei E. Warkentin<andrey.warkentin at gmail.com>:
>>>
>>> 2011/9/26 Mitch Bradley<wmb at firmworks.com>:
>>>>
>>>> If you write 0x400 to APB_VIRT_BASE + 0x1d008, it will assert interrupt
>>>> 56 .
>>>>
>>>> You can then clear it by writing 0x400 to APB_VIRT_BASE + 0x1d40c.
>>>>
>>>> This uses an inter-processor communication unit.  Presumably the FIG
>>>> handler
>>>> would assert the interrupt and the IRQ 56 handler would clear it.
>>>>
>>>
>>> Awesome! Thanks a lot!!!
>>>
>>>
>>
>> Hmm. I was hoping that this was really GIC SGI functionality
>>
>> (http://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.ihi0048a/index.html),
>> but I guess PJ4 has something completely different. Is there an easy
>> formula for the values written into those two regs and the IRQ value?
>> Or is this specific
>> functionality that only triggers IRQ 56? (which I note is absent from
>> mach/irqs.h).
>
>
> It is documented in the Marvell datasheet, which we can't release.
>
> As far as I know it is specific to this SoC, not a standard ARM thing.
>
> arch/arm/mach-mmp/include/mach/irqs.h in the Android gingerbread-gold kernel
> has:
>
> #define IRQ_MMP2_IPC2                   56
>
> "IPC" refers to the "InterProcessor Communication" unit.
>
> There are different bits that you can write to the "set" register to get
> "different" interrupts, but they are all or-ed together at the input to the
> SoC's interrupt controller, so they all appear as INT56.  I didn't think it
> was worthwhile to give you all the details of the different IPC unit options
> - nor to I really have time to do so.  I hope the recipe I gave you will
> suffice.  I presume that the Linux interrupt handling infrastructure will
> take care of the mechanics of registering and dispatching interrupts on this
> hardware.
>

Oh, absolutely! I was just wondering if there was anything other than 56, but it
definitely works for what I'm doing (in concept, in practice - as soon
as I test it).
Many thanks again for helping me out!

A