Re: [PATCH 07/11] irqchip: stm32: add stm32mp1 support with hierarchy domain

[Ludovic BARRE]

On 05/04/2018 10:38 PM, Rob Herring wrote:
> On Thu, May 3, 2018 at 4:55 AM, Ludovic BARRE <ludovic.barre@xxxxxx> wrote:
> > On 05/02/2018 07:45 PM, Rob Herring wrote:
> > > On Wed, May 2, 2018 at 11:03 AM, Ludovic BARRE <ludovic.barre@xxxxxx>
> > > wrote:
> > > > Hi Rob
> > > >
> > > >
> > > >
> > > > On 05/01/2018 04:56 PM, Rob Herring wrote:
> > > > > On Thu, Apr 26, 2018 at 06:18:30PM +0200, Ludovic Barre wrote:
> > > > > > From: Ludovic Barre <ludovic.barre@xxxxxx>
> > > > > >
> > > > > > Exti controller has been differently integrated on stm32mp1 SoC.
> > > > > > A parent irq has only one external interrupt. A hierachy domain could
> > > > > > be used. Handlers are call by parent, each parent interrupt could be
> > > > > > masked and unmasked according to the needs.
> > > > > >
> > > > > > Signed-off-by: Ludovic Barre <ludovic.barre@xxxxxx>
> > > > > > ---
> > > > > >     .../interrupt-controller/st,stm32-exti.txt         |   3 +
> > > > > >     drivers/irqchip/irq-stm32-exti.c                   | 322
> > > > > > +++++++++++++++++++++
> > > > > >     2 files changed, 325 insertions(+)
> > > > > >
> > > > > > diff --git
> > > > > >
> > > > > > a/Documentation/devicetree/bindings/interrupt-controller/st,stm32-exti.txt
> > > > > >
> > > > > > b/Documentation/devicetree/bindings/interrupt-controller/st,stm32-exti.txt
> > > > > > index edf03f0..136bd61 100644
> > > > > > ---
> > > > > >
> > > > > > a/Documentation/devicetree/bindings/interrupt-controller/st,stm32-exti.txt
> > > > > > +++
> > > > > >
> > > > > > b/Documentation/devicetree/bindings/interrupt-controller/st,stm32-exti.txt
> > > > > > @@ -5,11 +5,14 @@ Required properties:
> > > > > >     - compatible: Should be:
> > > > > >         "st,stm32-exti"
> > > > > >         "st,stm32h7-exti"
> > > > > > +    "st,stm32mp1-exti"
> > > > > >     - reg: Specifies base physical address and size of the registers
> > > > > >     - interrupt-controller: Indentifies the node as an interrupt
> > > > > > controller
> > > > > >     - #interrupt-cells: Specifies the number of cells to encode an
> > > > > > interrupt
> > > > > >       specifier, shall be 2
> > > > > >     - interrupts: interrupts references to primary interrupt controller
> > > > > > +  (only needed for exti controller with multiple exti under
> > > > > > +  same parent interrupt: st,stm32-exti and st,stm32h7-exti")
> > > > > >       Example:
> > > > > >     diff --git a/drivers/irqchip/irq-stm32-exti.c
> > > > > > b/drivers/irqchip/irq-stm32-exti.c
> > > > > > index b38c655..ebf7146 100644
> > > > > > --- a/drivers/irqchip/irq-stm32-exti.c
> > > > > > +++ b/drivers/irqchip/irq-stm32-exti.c
> > > > >
> > > > >
> > > > >
> > > > > [...]
> > > > >
> > > > > > +static const struct stm32_desc_irq stm32mp1_desc_irq[] = {
> > > > > > +       { .exti = 1, .irq_parent = 7 },
> > > > > > +       { .exti = 2, .irq_parent = 8 },
> > > > > > +       { .exti = 3, .irq_parent = 9 },
> > > > > > +       { .exti = 4, .irq_parent = 10 },
> > > > > > +       { .exti = 5, .irq_parent = 23 },
> > > > > > +       { .exti = 6, .irq_parent = 64 },
> > > > > > +       { .exti = 7, .irq_parent = 65 },
> > > > > > +       { .exti = 8, .irq_parent = 66 },
> > > > > > +       { .exti = 9, .irq_parent = 67 },
> > > > > > +       { .exti = 10, .irq_parent = 40 },
> > > > > > +       { .exti = 11, .irq_parent = 42 },
> > > > > > +       { .exti = 12, .irq_parent = 76 },
> > > > > > +       { .exti = 13, .irq_parent = 77 },
> > > > > > +       { .exti = 14, .irq_parent = 121 },
> > > > > > +       { .exti = 15, .irq_parent = 127 },
> > > > > > +       { .exti = 16, .irq_parent = 1 },
> > > > > > +       { .exti = 65, .irq_parent = 144 },
> > > > > > +       { .exti = 68, .irq_parent = 143 },
> > > > > > +       { .exti = 73, .irq_parent = 129 },
> > > > > > +};
> > > > >
> > > > >
> > > > >
> > > > > You can use an interrupt-map property rather than put this into the
> > > > > driver.
> > > >
> > > >
> > > >
> > > > interrupt-map seemed interesting and promising like used in pci host.
> > > > At first sight this property can't be used into node with
> > > > "interrupt-controller" property (see in drivers/of/irq.c function:
> > > > of_irq_parse_raw) because "of_irq_parse_raw" checks if node got
> > > > it first, and after lookup the interrupt-map.
> > > >
> > > > Rob, Thomas, Jason, Marc what do you prefers or the right ways...?
> > >
> > >
> > > I believe the correct thing to do is simply drop "interrupt-controller".
>
> Actually, that's not right.
>
> > if I drop "interrupt-controller" of my node, my driver will not be
> > initialized by "of_irq_init"
> > (start_kernel->init_IRQ->irqchip_init->of_irq_init).
> > Probably, we could replace "IRQCHIP_DECLARE" by a
> > "module_platform_driver" or "postcore_initcall" but I'm not a big fan of
> > this solution. what do you think ?
>
> You'd have to parse 'interrupt-map' yourself to extract the data for
> this to work because interrupt-map currently only works when the
> translation is transparent (i.e. doesn't need s/w handling). So I
> guess leave this in the driver. Sorry for the noise.

Sorry, I don't know if I've correctly understand the answer.
I hesitate between:
- keep the code like this, with stm32mp1_desc_irq tab.
- parse interrupt-map property into stm32 exti driver.

Just to be sure, and avoid misunderstand.

BR
Ludo

> Rob