Re: [PATCH v2 2/7] dt-bindings: interrupt-controller: apple,aic2: New binding for AICv2

From: Marc Zyngier
Date: Mon Mar 07 2022 - 06:35:34 EST


On Fri, 25 Feb 2022 21:58:34 +0000,
Hector Martin <marcan@xxxxxxxxx> wrote:
>
> On 26/02/2022 05.19, Rob Herring wrote:
> >> +properties:
> >> + compatible:
> >> + items:
> >> + - const: apple,t6000-aic
> >> + - const: apple,aic2
> >
> > I feel I was sold on Apple doesn't change h/w and we're the 2nd chip in
> > and the h/w changed. Just my musings, but aic3 will be rejected. :(
>
> Well yes, after not changing hardware for N phone/tablet generations,
> they figured out they *finally* had to make some changes for real
> desktop chips... (t8103 was a tablet chip they shoehorned into laptops;
> t6000 is the first real laptop/desktop chip). This isn't the 2nd chip
> in, this is the 26th chip in or so, and yet it's called AIC2 (by Apple
> even)... We aren't starting from chip #1, just the first chip they
> decided to *let* us put Linux on.
>
> It's pretty clear that the t6000 changes were made with future-proofing
> in mind. I guess we'll find out in a couple weeks, since the rumor mill
> says M2 is coming. If I'm right and we end up needing *zero* kernel
> changes to boot on M2, will you be happy? ;-)
>
> >> + apple,event-reg:
> >> + $ref: /schemas/types.yaml#/definitions/uint32
> >> + description:
> >> + Specifies the offset of the event register, which lies after all the
> >> + implemented die register sets, page aligned. This is not computable from
> >> + capability register values, so we have to specify it explicitly.
> >
> > If this is last, then couldn't it be a 2nd 'reg' entry?
> >
> > 'page aligned' is ambiguous. I assume that means 16K since that's what
> > Apple uses, but I might assume 4K not knowing that.
>
> 16K, and yeah, it could be a 2nd reg entry if you think that works
> better. Makes sense.

Do you plan to respin this? If I'm going to that this series for 5.18,
it needs to be this week.

Thanks,

M.

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