Re: [PATCH v2 1/5] dt-bindings: interrupt-controller: Update STM32 EXTI interrupt controller

[Marc Zyngier <maz@kernel.org>]

On Fri, 17 Dec 2021 13:39:11 +0000,
Alexandre TORGUE <alexandre.torgue@foss.st.com> wrote:
> 
> On 12/16/21 9:15 PM, Rob Herring wrote:
> > On Wed, Dec 15, 2021 at 11:58:43AM +0100, Alexandre Torgue wrote:
> >> Document new entry "st,exti-mapping" which links EXTI lines with GIC
> >> interrupt lines and add an include file to define EXTI interrupt type.
> >> 
> >> Signed-off-by: Alexandre Torgue <alexandre.torgue@foss.st.com>
> >> 
> >> diff --git a/Documentation/devicetree/bindings/interrupt-controller/st,stm32-exti.yaml b/Documentation/devicetree/bindings/interrupt-controller/st,stm32-exti.yaml
> >> index d19c881b4abc..e08bb51e97a8 100644
> >> --- a/Documentation/devicetree/bindings/interrupt-controller/st,stm32-exti.yaml
> >> +++ b/Documentation/devicetree/bindings/interrupt-controller/st,stm32-exti.yaml
> >> @@ -41,6 +41,17 @@ properties:
> >>       description:
> >>         Interrupts references to primary interrupt controller
> >>   +  st,exti-mapping:
> >> +    $ref: "/schemas/types.yaml#/definitions/uint32-matrix"
> >> +    description: |
> >> +            Define mapping between EXTI lines and GIC irq lines. Should be:
> >> +            st,exti-mapping = <EXTI_LINE GIC_IRQ EXTI_TYPE>, ...;
> >> +            With:
> >> +            - EXTI_LINE: EXTI line number.
> >> +            - GIC_IRQ: GIC IRQ associated to the EXTI line.
> >> +            - EXTI_TYPE: STM32_EXTI_TYPE_CONFIGURABLE or STM32_EXTI_TYPE_DIRECT.
> >> +              Defined in include/dt-bindings/interrupt-controller/stm32-exti.h
> > 
> > No custom properties for this. See[1][2][3].
> > 
> 
> Thanks for inputs. In my case the mapping consists to map an EXTI line
> with a GIC irq line which could be done using interrupt-map (avoiding
> to parse it in my driver).

The problem is that 'interrupt-map' defines an interrupt mapping
between an input and an output, and that mentioning the GIC in such a
table will only result in your EXTI to be bypassed.

'interrupt-map' really is a dispatch table for targeting an interrupt
controller (or multiple controllers, even), but really isn't the
correct tool to carry configuration informations to an interrupt
controller driver.

> But for each EXTI/GIC association I would
> like also to describe the EXTI_TYPE (which actually describe the well
> irqchip to use inside my exti driver) . This property is not generic
> and so I assume I can't use a generic binding such "interrupt-map".
> 
> If the solution consists to use a common binding (i.e. interrupt-map)
> plus a conversion table in exti driver to affect the well irq_chip to
> the well EXTI line then we could envisage to keep the whole mapping
> inside the driver (even if it's not the best solution).

A possible solution would be to have:

- A set of standard 'interrupts' properties describing the output
signals

- A set of properties describing the input to output mapping (if
relevant) and additional configuration information that for the
interrupt controller driver.

	M.

-- 
Without deviation from the norm, progress is not possible.