Re: [PATCH v2 1/5] dt-bindings: interrupt-controller: Add Renesas RZ/A1 Interrupt Controller

[Geert Uytterhoeven]

Hi Rob,

On Tue, Apr 30, 2019 at 10:26 PM Rob Herring <robh+dt@xxxxxxxxxx> wrote:
> On Tue, Apr 30, 2019 at 10:34 AM Marc Zyngier <marc.zyngier@xxxxxxx> wrote:
> > On 30/04/2019 16:02, Rob Herring wrote:
> > > On Tue, Apr 30, 2019 at 7:13 AM Geert Uytterhoeven
> > > <geert+renesas@xxxxxxxxx> wrote:
> > >>
> > >> Add DT bindings for the Renesas RZ/A1 Interrupt Controller.
> > >>
> > >> Signed-off-by: Geert Uytterhoeven <geert+renesas@xxxxxxxxx>
> > >> ---
> > >> v2:
> > >>   - Add "renesas,gic-spi-base",
> > >>   - Document RZ/A2M.
> > >> ---
> > >>  .../renesas,rza1-irqc.txt                     | 30 +++++++++++++++++++
> > >>  1 file changed, 30 insertions(+)
> > >>  create mode 100644 Documentation/devicetree/bindings/interrupt-controller/renesas,rza1-irqc.txt
> > >>
> > >> diff --git a/Documentation/devicetree/bindings/interrupt-controller/renesas,rza1-irqc.txt b/Documentation/devicetree/bindings/interrupt-controller/renesas,rza1-irqc.txt
> > >> new file mode 100644
> > >> index 0000000000000000..ea8ddb6955338ccd
> > >> --- /dev/null
> > >> +++ b/Documentation/devicetree/bindings/interrupt-controller/renesas,rza1-irqc.txt
> > >> @@ -0,0 +1,30 @@
> > >> +DT bindings for the Renesas RZ/A1 Interrupt Controller
> > >> +
> > >> +The RZ/A1 Interrupt Controller is a front-end for the GIC found on Renesas
> > >> +RZ/A1 and RZ/A2 SoCs:
> > >> +  - IRQ sense select for 8 external interrupts, 1:1-mapped to 8 GIC SPI
> > >> +    interrupts,
> > >> +  - NMI edge select.
> > >> +
> > >> +Required properties:
> > >> +  - compatible: Must be "renesas,<soctype>-irqc", and "renesas,rza1-irqc" as
> > >> +               fallback.
> > >> +               Examples with soctypes are:
> > >> +                 - "renesas,r7s72100-irqc" (RZ/A1H)
> > >> +                 - "renesas,r7s9210-irqc" (RZ/A2M)
> > >> +  - #interrupt-cells: Must be 2 (an interrupt index and flags, as defined
> > >> +                                in interrupts.txt in this directory)
> > >> +  - interrupt-controller: Marks the device as an interrupt controller
> > >> +  - reg: Base address and length of the memory resource used by the interrupt
> > >> +         controller
> > >> +  - renesas,gic-spi-base: Lowest GIC SPI interrupt number this block maps to.
> > >
> > > Why isn't this just an 'interrupts' property?
> >
> > That's likely because of kernel limitations. The DT code does an
> > of_populate() on any device that it finds, parse the "interrupts"
> > propertiy, resulting in the irq_descs being populated.
> >
> > That creates havoc, as these interrupts are not for this device, but for
> > something that is connected to it. This is merely a bridge of some sort.
>
> 'interrupt-map' would avoid that problem I think.

"interrupt-map" seems to be meant for translation on a bus?
What to do with the child and parent unit addresses fields?
The parent unit address size depends on the #address-cells of the parent
interrupt-controller (i.e. GIC, so it's zero).
But the child unit address size depends on the #address-cells of the bus node
on which the child is located, so that's a (non-zero) bus #address-cells
(from the root node), not an interrupt-controller #address-cells.

Each line in an interrupt-map also contains a child interrupt specifier.
As the RZ/A1 IRQC supports 8 interrupt inputs with 4 sense types,
that would mean 32 lines? Or should I just ignore the senses here,
and specify 0?

i.e. interrupt-map = <0 0 0 &gic GIC_SPI 0 IRQ_TYPE_LEVEL_HIGH,
                      0 1 0 &gic GIC_SPI 1 IRQ_TYPE_LEVEL_HIGH,
                      0 2 0 &gic GIC_SPI 2 IRQ_TYPE_LEVEL_HIGH,
                      0 3 0 &gic GIC_SPI 3 IRQ_TYPE_LEVEL_HIGH,
                      0 4 0 &gic GIC_SPI 4 IRQ_TYPE_LEVEL_HIGH,
                      0 5 0 &gic GIC_SPI 5 IRQ_TYPE_LEVEL_HIGH,
                      0 6 0 &gic GIC_SPI 6 IRQ_TYPE_LEVEL_HIGH,
                      0 7 0 &gic GIC_SPI 7 IRQ_TYPE_LEVEL_HIGH>;

(using zero for the child unit addresses, too)?

> > Furthermore, this is a rather long established practice: gic-v2m,
> > gic-v3-mbi, mediatek,sysirq, mediatek,cirq... All the bits of glue that
> > for one reason or another plug onto the GIC use the same method.
>
> All handling the mapping to the parent in their own way...
>
> > > Plus, without 'interrupts' walking the hierarchy is broken.
> >
> > Erm... Which hierarchy?
>
> of_irq_init() expects that an interrupt-controller without an
> interrupt-parent is the root controller. So you're right. We only need

That applies to IRQCHIP_DECLARE() drivers only, not platform device
drivers, right?

> to have an 'interrupt-parent', but not 'interrupts'.

This is implied by "interrupt-parent = <&gic>;" on the soc node.

Thanks!

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds