Hi Rob,
On Mon, 13 May 2024 09:53:58 -0500
Rob Herring <robh@xxxxxxxxxx> wrote:
> On Mon, May 13, 2024 at 02:37:20PM +0200, Herve Codina wrote:
> > Hi Rob,
> >
> > On Tue, 7 May 2024 10:28:06 -0500
> > Rob Herring <robh@xxxxxxxxxx> wrote:
> >
> > ...
> > > > +examples:
> > > > + - |
> > > > + interrupt-controller@e00c0120 {
> > > > + compatible = "microchip,lan966x-oic";
> > > > + reg = <0xe00c0120 0x190>;
> > >
> > > Looks like this is part of some larger block?
> > >
> >
> > According to the registers information document:
> > https://microchip-ung.github.io/lan9662_reginfo/reginfo_LAN9662.html?select=cpu,intr
> >
> > The interrupt controller is mapped at offset 0x48 (offset in number of
> > 32bit words).
> > -> Address offset: 0x48 * 4 = 0x120
> > -> size: (0x63 + 1) * 4 = 0x190
> >
> > IMHO, the reg property value looks correct.
>
> What I mean is h/w blocks don't just start at some address with small
> alignment. That wouldn't work from a physical design standpoint. The
> larger block here is "CPU System Regs". The block as a whole should be
> documented, but maybe that ship already sailed.
The clock controller, also part of the "CPU System Regs" is already defined
and used without the larger block
Documentation/devicetree/bindings/clock/microchip,lan966x-gck.yaml
IMHO, the binding related to the interrupt controller should be consistent
with the one related to the clock controller.
>
> Also, here you call it the OIC, but the link above calls it the VCore
> interrupt controller.
Yes, I call it OIC (Outband Interrupt Controller) as it is its name in the
datasheet explaining how it works.
The datasheet I have is not publicly available and so, I can point only to
the register map (url provided).
I think it would be better to keep "Outband Interrupt Controller" as
mentioned in the datasheet.
Best regards,
Hervé
--
Hervé Codina, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com
