[PATCH 1/9] dt-bindings: dma: brcm,bcm2835-dma: Add BCM2711 and BCM2712

[Peter Robinson]

On Mon, 2 Nov 2026 at 20:38, Florian Fainelli <f.fainelli at gmail.com> wrote:
>
> On 11/2/26 12:18, Peter Robinson wrote:
> > On Mon, 2 Nov 2026 at 19:39, Conor Dooley <conor at kernel.org> wrote:
> >>
> >> On Mon, Nov 02, 2026 at 09:06:51AM +0000, Peter Robinson wrote:
> >>> Add compatibles for the BCM2711 40-bit DMA4 channels, which are
> >>> described by a separate node covering channels 11-14, and for the
> >>> BCM2712 DMA controllers. The register address of a node identifies its
> >>> first channel, allowing a controller to be split across nodes.
> >>>
> >>> Signed-off-by: Peter Robinson <pbrobinson at gmail.com>
> >>> Cc: Rob Herring <robh at kernel.org>
> >>> Cc: Krzysztof Kozlowski <krzk+dt at kernel.org>
> >>> Cc: Conor Dooley <conor+dt at kernel.org>
> >>>
> >>> ---
> >>>   .../bindings/dma/brcm,bcm2835-dma.yaml        | 31 +++++++++++++++++--
> >>>   1 file changed, 29 insertions(+), 2 deletions(-)
> >>>
> >>> diff --git a/Documentation/devicetree/bindings/dma/brcm,bcm2835-dma.yaml b/Documentation/devicetree/bindings/dma/brcm,bcm2835-dma.yaml
> >>> index c9b9a5490826c..31d97a10a0ea9 100644
> >>> --- a/Documentation/devicetree/bindings/dma/brcm,bcm2835-dma.yaml
> >>> +++ b/Documentation/devicetree/bindings/dma/brcm,bcm2835-dma.yaml
> >>> @@ -9,18 +9,27 @@ title: BCM2835 DMA controller
> >>>   maintainers:
> >>>     - Nicolas Saenz Julienne <nsaenz at kernel.org>
> >>>
> >>> -description:
> >>> +description: |
> >>>     The BCM2835 DMA controller has 16 channels in total. Only the lower
> >>>     13 channels have an associated IRQ. Some arbitrary channels are used by the
> >>>     VideoCore firmware (1,3,6,7 in the current firmware version). The channels
> >>>     0, 2 and 3 have special functionality and should not be used by the driver.
> >>>
> >>> +  BCM2711 channels 11 to 14 are 40-bit capable DMA4 channels, described by a
> >>> +  separate brcm,bcm2711-dma node. BCM2712 has 40-bit addressing on all
> >>> +  channels, with channels 6 to 11 being DMA4 channels. A controller may be
> >>> +  split across several nodes, the register address identifies the first
> >>> +  channel of a node.
> >>
> >> Uh no, a device should not be split across different nodes. This should
> >> not be needed, why can't your driver handle the upper channels having
> >> different capabilities?
> >
> > This is the way it's done in the downstream kernel, unfortunately
> > Broadcom doesn't publish the HW details so either they'd need to
> > contribute the support (which they've not done since the RPi4 was
> > available) or I'm not sure what the other options would be. Open to
> > suggestions.
>
> Neither here nor there, IMHO, yes it would be nice if documentation was
> published (and a customer like Raspberry Pi is probably one of the few
> that can get the needle to move there).
>
> You have the bcm2711 and bcm2712 compatible strings which should be
> enough for you to know which channels are 40-bit capable and which are
> not, yet keep the same Device Tree node.

>From reading the downstream commits though it reads like the different
channels have different capabilities? Or am I reading that wrong and
it's the different generations of the HW that have different
capabilities?

Peter