[DT] API to determine if interrupt is PPI or SPI ?

[DT] API to determine if interrupt is PPI or SPI ?

[rvaswani]

Hello,

We can specify the interrupt information in the device tree using the arm
gic convention.
Is there an API to read directly the 1st cell and determine if an
interrupt is a PPI or a SPI ? This information is directly relevant to the
driver because it needs to call request_irq or request_percpu_irq based on
this data.

How should this be done ideally ?

Thanks,
Rohit

-- 
Sent by an employee of the Qualcomm Innovation Center, Inc.
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.
--

[DT] API to determine if interrupt is PPI or SPI ?

[rvaswani at codeaurora.org]

Hello,

We can specify the interrupt information in the device tree using the arm
gic convention.
Is there an API to read directly the 1st cell and determine if an
interrupt is a PPI or a SPI ? This information is directly relevant to the
driver because it needs to call request_irq or request_percpu_irq based on
this data.

How should this be done ideally ?

Thanks,
Rohit

-- 
Sent by an employee of the Qualcomm Innovation Center, Inc.
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.
--

Re: [DT] API to determine if interrupt is PPI or SPI ?

[Rob Herring]

On 07/26/2012 02:18 PM, rvaswani-sgV2jX0FEOL9JmXXK+q4OQ@public.gmane.org wrote:
> Hello,
> 
> We can specify the interrupt information in the device tree using the arm
> gic convention.
> Is there an API to read directly the 1st cell and determine if an
> interrupt is a PPI or a SPI ? This information is directly relevant to the
> driver because it needs to call request_irq or request_percpu_irq based on
> this data.
> 
> How should this be done ideally ?

Seems a bit strange to me that you would have a h/w block that can be
connected either way...

Perhaps the irq resource should have a flag for percpu irq.

Rob

[DT] API to determine if interrupt is PPI or SPI ?

[Rob Herring]

On 07/26/2012 02:18 PM, rvaswani at codeaurora.org wrote:
> Hello,
> 
> We can specify the interrupt information in the device tree using the arm
> gic convention.
> Is there an API to read directly the 1st cell and determine if an
> interrupt is a PPI or a SPI ? This information is directly relevant to the
> driver because it needs to call request_irq or request_percpu_irq based on
> this data.
> 
> How should this be done ideally ?

Seems a bit strange to me that you would have a h/w block that can be
connected either way...

Perhaps the irq resource should have a flag for percpu irq.

Rob

Re: [DT] API to determine if interrupt is PPI or SPI ?

[Will Deacon]

On Fri, Jul 27, 2012 at 01:33:36PM +0100, Rob Herring wrote:
> On 07/26/2012 02:18 PM, rvaswani-sgV2jX0FEOL9JmXXK+q4OQ@public.gmane.org wrote:
> > Hello,
> > 
> > We can specify the interrupt information in the device tree using the arm
> > gic convention.
> > Is there an API to read directly the 1st cell and determine if an
> > interrupt is a PPI or a SPI ? This information is directly relevant to the
> > driver because it needs to call request_irq or request_percpu_irq based on
> > this data.
> > 
> > How should this be done ideally ?
> 
> Seems a bit strange to me that you would have a h/w block that can be
> connected either way...

The CPU PMU can be wired as either SPI or PPI. The latter makes more sense to
me but usually (although not exclusively) it's wired using the former.

Will

[DT] API to determine if interrupt is PPI or SPI ?

[Will Deacon]

On Fri, Jul 27, 2012 at 01:33:36PM +0100, Rob Herring wrote:
> On 07/26/2012 02:18 PM, rvaswani at codeaurora.org wrote:
> > Hello,
> > 
> > We can specify the interrupt information in the device tree using the arm
> > gic convention.
> > Is there an API to read directly the 1st cell and determine if an
> > interrupt is a PPI or a SPI ? This information is directly relevant to the
> > driver because it needs to call request_irq or request_percpu_irq based on
> > this data.
> > 
> > How should this be done ideally ?
> 
> Seems a bit strange to me that you would have a h/w block that can be
> connected either way...

The CPU PMU can be wired as either SPI or PPI. The latter makes more sense to
me but usually (although not exclusively) it's wired using the former.

Will

Re: [DT] API to determine if interrupt is PPI or SPI ?

[rvaswani]

> On Fri, Jul 27, 2012 at 01:33:36PM +0100, Rob Herring wrote:
>> On 07/26/2012 02:18 PM, rvaswani@codeaurora.org wrote:
>> > Hello,
>> >
>> > We can specify the interrupt information in the device tree using the
>> arm
>> > gic convention.
>> > Is there an API to read directly the 1st cell and determine if an
>> > interrupt is a PPI or a SPI ? This information is directly relevant to
>> the
>> > driver because it needs to call request_irq or request_percpu_irq
>> based on
>> > this data.
>> >
>> > How should this be done ideally ?
>>
>> Seems a bit strange to me that you would have a h/w block that can be
>> connected either way...
>
> The CPU PMU can be wired as either SPI or PPI. The latter makes more sense
> to
> me but usually (although not exclusively) it's wired using the former.
>
> Will
>

We have the same driver for a hardware for different chipsets and
unfortunately the hardware team decided to connect this GIC interrupt
differently across these chipsets. This leaves us with a software choice
to make.
We could add another property to the device tree (like irq-is-percpu,
irq-is-not-percpu), but that seems redundant.
We were hoping to have an API to tell us the SPI/PPI difference since this
information is already encoded in the interrupt-cells in the device tree
for this device.
Would that be the right way to go or there is something better that can be
done ?

Thanks,
Rohit

-- 
Sent by a consultant of the Qualcomm Innovation Center, Inc.
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.

[DT] API to determine if interrupt is PPI or SPI ?

[rvaswani at codeaurora.org]

> On Fri, Jul 27, 2012 at 01:33:36PM +0100, Rob Herring wrote:
>> On 07/26/2012 02:18 PM, rvaswani at codeaurora.org wrote:
>> > Hello,
>> >
>> > We can specify the interrupt information in the device tree using the
>> arm
>> > gic convention.
>> > Is there an API to read directly the 1st cell and determine if an
>> > interrupt is a PPI or a SPI ? This information is directly relevant to
>> the
>> > driver because it needs to call request_irq or request_percpu_irq
>> based on
>> > this data.
>> >
>> > How should this be done ideally ?
>>
>> Seems a bit strange to me that you would have a h/w block that can be
>> connected either way...
>
> The CPU PMU can be wired as either SPI or PPI. The latter makes more sense
> to
> me but usually (although not exclusively) it's wired using the former.
>
> Will
>

We have the same driver for a hardware for different chipsets and
unfortunately the hardware team decided to connect this GIC interrupt
differently across these chipsets. This leaves us with a software choice
to make.
We could add another property to the device tree (like irq-is-percpu,
irq-is-not-percpu), but that seems redundant.
We were hoping to have an API to tell us the SPI/PPI difference since this
information is already encoded in the interrupt-cells in the device tree
for this device.
Would that be the right way to go or there is something better that can be
done ?

Thanks,
Rohit

-- 
Sent by a consultant of the Qualcomm Innovation Center, Inc.
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.