Re: [PATCH 05/11] dt-bindings: document the Qualcomm TEE Shared Memory Bridge

[Bartosz Golaszewski]

On Wed, 30 Aug 2023 at 16:58, Krzysztof Kozlowski
<krzysztof.kozlowski@xxxxxxxxxx> wrote:
>
> On 30/08/2023 16:39, Bartosz Golaszewski wrote:
> > On Wed, 30 Aug 2023 at 16:31, Krzysztof Kozlowski
> > <krzysztof.kozlowski@xxxxxxxxxx> wrote:
> >>
> >> On 30/08/2023 15:48, Bartosz Golaszewski wrote:
> >>> On Tue, 29 Aug 2023 at 11:30, Konrad Dybcio <konrad.dybcio@xxxxxxxxxx> wrote:
> >>>>
> >>>> On 29.08.2023 10:02, Krzysztof Kozlowski wrote:
> >>>>> On 28/08/2023 21:25, Bartosz Golaszewski wrote:
> >>>>>> Add Device Tree bindings for Qualcomm TEE Shared Memory Brige - a
> >>>>>> mechanism that allows sharing memory buffers between trustzone and the
> >>>>>> kernel.
> >>>>>
> >>>>> Subject prefix:
> >>>>> dt-bindings: firmware:
> >>>>>
> >>>>>
> >>>>>
> >>>>>>
> >>>>>> Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@xxxxxxxxxx>
> >>>>>> ---
> >>>>>>  .../bindings/firmware/qcom,shm-bridge.yaml    | 36 +++++++++++++++++++
> >>>>>>  1 file changed, 36 insertions(+)
> >>>>>>  create mode 100644 Documentation/devicetree/bindings/firmware/qcom,shm-bridge.yaml
> >>>>>>
> >>>>>> diff --git a/Documentation/devicetree/bindings/firmware/qcom,shm-bridge.yaml b/Documentation/devicetree/bindings/firmware/qcom,shm-bridge.yaml
> >>>>>> new file mode 100644
> >>>>>> index 000000000000..f660962b7b86
> >>>>>> --- /dev/null
> >>>>>> +++ b/Documentation/devicetree/bindings/firmware/qcom,shm-bridge.yaml
> >>>>>> @@ -0,0 +1,36 @@
> >>>>>> +# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
> >>>>>> +%YAML 1.2
> >>>>>> +---
> >>>>>> +$id: http://devicetree.org/schemas/firmware/qcom,shm-bridge.yaml#
> >>>>>> +$schema: http://devicetree.org/meta-schemas/core.yaml#
> >>>>>> +
> >>>>>> +title: QCOM Shared Memory Bridge
> >>>>>> +
> >>>>>> +description: |
> >>>>>
> >>>>> Do not need '|' unless you need to preserve formatting.
> >>>>>
> >>>>>> +  Qualcomm TEE Shared Memory Bridge allows sharing limited areas of kernel's
> >>>>>> +  virtual memory with the trustzone in order to avoid mapping the entire RAM.
> >>>>>> +
> >>>>>> +maintainers:
> >>>>>> +  - Bjorn Andersson <andersson@xxxxxxxxxx>
> >>>>>> +  - Konrad Dybcio <konrad.dybcio@xxxxxxxxxx>
> >>>>>> +  - Bartosz Golaszewski <bartosz.golaszewski@xxxxxxxxxx>
> >>>>>> +
> >>>>>> +properties:
> >>>>>> +  compatible:
> >>>>>> +    items:
> >>>>>> +      - enum:
> >>>>>> +          - qcom,shm-bridge-sa8775p
> >>>>>> +          - qcom,shm-bridge-sm8150
> >>>>>> +          - qcom,shm-bridge-sm8450
> >>>>>> +      - const: qcom,shm-bridge
> >>>>>> +
> >>>>>
> >>>>> Looks quite empty... Why this cannot be part of qcom,scm? IOW, why do
> >>>>> you need new binding if you do not have any resources here and the block
> >>>>> is essentially feature of qcom,scm firmware?
> >>>> Since it's "discoverable" (via retval of an scm call), I'd second the
> >>>> idea of probing this from within the SCM driver.
> >>>>
> >>>> Konrad
> >>>
> >>> Downstream has a bunch of DT switches that we don't support for now
> >>> upstream. I disagree about shoehorning this into the SCM driver. It
> >>> really is a layer on top of SCM but also SCM is a user of this
> >>> interface.
> >>
> >> Sure, for the driver makes sense, but it does not really explain why DT
> >> node is needed. It is not separate hardware. I doubt it is even separate
> >> firmware, but part of SCM.
> >>
> >> Best regards,
> >> Krzysztof
> >>
> >
> > Because not all platforms support it and it's the simplest way of
>
> Platforms like SoCs or boards?
>
> > marking the ones that do. Both SHM and SCM nodes sit on the firmware
> > node anyway. What do you recommend? A property of the SCM node? Like
> > 'qcom,shm-bridge` or something?
>
> If the first - you talk about SoCs - then you have everything needed
> already: SCM compatibles. This defines it fully.
>
> If it varies by boards with one SoC, would be different case, but I
> really doubt it.
>

Ok, makes sense. Thanks.

Bartosz