Re: [PATCH] arm: dts: broadcom: Add missing required fields

[Dave Stevenson]

Hi Stefan and Florian

On Mon, 30 Sept 2024 at 19:36, Stefan Wahren <wahrenst@xxxxxxx> wrote:
>
> Hi Florian,
>
> Am 25.09.24 um 22:38 schrieb Florian Fainelli:
> > On 9/25/24 09:39, Stefan Wahren wrote:
> >> Hi Karan,
> >>
> >> Am 25.09.24 um 18:14 schrieb Karan Sanghavi:
> >>> Added below mentioned required fields
> >>>    1. interrupt-controller
> >>>    2. #interrupt-cells
> >>> in the bcm2711.dtsi file for the
> >>> interrupt-controller@40000000 block as defined in the
> >>> bindings/interrupt-controller/brcm,bcm2836-l1-intc.yaml.
> >>> This issue was noticed while compiling the dtb file
> >>> for broadcom/bcm2711-rpi-4-b.dts file.
> >>> After including the above fields in the dtsi file
> >>> interrupt-conntroller error was resolved.
> >> looks like you made the same mistake like me [1]. This change breaks
> >> boot of Raspberry Pi 4 [2].
> >>
> >> There are a lot of DT schema warnings to fix, but this doesn't belong to
> >> the trivial ones.
> >
> > Including the #interrupt-cells would not have a functional impact
> > however, and we ought to be able to do that.
> >
> > The 'interrupt-controller' property presence means that the controller
> > will be picked up by of_irq_init() and that is was causes problems for
> > people testing this. Stefan, do you know if the VPU firmware
> > removes/inserts that property to tell Linux which interrupt controller
> > (bcm2836-l1-intc or ARM GIC) to use or does it make use of the
> > "status" property which would be the canonical way about doing that?
> There is a config.txt parameter for this, which is called "enable_gic".
> But if i use this i couldn't see any difference to /proc/device-tree.
> Also i couldn't see any modifications by the firmware to the node in
> general:
>
> interrupt-controller@40000000 {
>          compatible = "brcm,bcm2836-l1-intc";
>          reg = <0x40000000 0x100>;
>          phandle = <0x8e>;
> };
>
> Except of this i don't have any clue about the VPU firmware.

cc Phil so he can correct me if I get this wrong.

The firmware looks at the DTB and automatically sets the enable_gic
property if DT /interrupt-parent points at a node with compatible
string "arm,gic-400". It doesn't modify DT around the interrupt
controller nodes.

Manually setting enable_gic should only be necessary on a system which
isn't using DT where they wish to control whether to use the GIC or
bcm2836-l1-intc.

  Dave