Re: [PATCH 1/2] dt-bindings: gpio: add binding doc for siflower,sf19a2890-gpio

[Chuanhong Guo]

Hello!

On Sat, Dec 28, 2024 at 12:45 AM Linus Walleij <linus.walleij@xxxxxxxxxx> wrote:
>
> Hi Chuanhong,
>
> thanks for your patch!
>
> On Wed, Dec 25, 2024 at 4:59 AM Chuanhong Guo <gch981213@xxxxxxxxx> wrote:
>
> > Add dt binding doc for the GPIO controller found on Siflower SF19A2890
> > and various other Siflower MIPS and RISC-V SoCs.
> >
> > Signed-off-by: Chuanhong Guo <gch981213@xxxxxxxxx>
> (...)
> > +  interrupts:
> > +    description:
> > +      Interrupt mapping, one interrupt per 16 GPIOs.
>
> So from the driver it is very clear that this is lumping together several
> GPIO blocks with 16 GPIOs in each into a bigger GPIO controller, despite
> the instances are identical. They even each have an individual IRQ.
>
> > +  ngpios:
> > +    description:
> > +      The number of GPIOs available on the controller implementation.
> > +    minimum: 1
>
> I would say minimum: 1 maximum: 16 default: 16.
>
> One instance per block/bank.
>
> > +examples:
> > +  - |
> > +    #include <dt-bindings/interrupt-controller/irq.h>
> > +    #include <dt-bindings/interrupt-controller/mips-gic.h>
> > +    gpio@19d00000 {
> > +      compatible = "siflower,sf19a2890-gpio";
> > +      reg = <0x19d00000 0x100000>;
>
> Just use 4 instances. Since (looking at the driver) it seems there
> is an IRQ register that is "off the bulk" I would do something like:
>
> instance 0:
>
> reg = <0x19d00000 0x40>, <0x19d04000 4>;
>
> instance: 1:
>
> reg = <0x19d00040 0x40>, <0x19d04004 4>;
>
> (...etc...)

Actually, this weird GPIO controller uses one 0x40 block per pin (The
49 pin controller uses 0x40 * 49 or about 3K memory space just for IO
registers!), and they share a single reset signal from the reset
controller.
I don't know how one reset could be shared across multiple platform
devices so I don't think this kind of split is possible.

--
Regards,
Chuanhong Guo