Re: [PATCH v8 3/4] arm64: dts: qcom: sm6115: Add EUD dt node and dwc3 connector

[Bhupesh Sharma]

On Tue, 18 Jul 2023 at 01:54, Konrad Dybcio <konrad.dybcio@xxxxxxxxxx> wrote:
>
> On 17.07.2023 22:22, Bhupesh Sharma wrote:
> > On Tue, 18 Jul 2023 at 01:49, Konrad Dybcio <konrad.dybcio@xxxxxxxxxx> wrote:
> >>
> >> On 17.07.2023 22:09, Bhupesh Sharma wrote:
> >>> On Mon, 17 Jul 2023 at 23:58, Stephan Gerhold <stephan@xxxxxxxxxxx> wrote:
> >>>>
> >>>> On Mon, Jul 17, 2023 at 11:33:40PM +0530, Bhupesh Sharma wrote:
> >>>>> On Mon, 17 Jul 2023 at 16:15, Stephan Gerhold <stephan@xxxxxxxxxxx> wrote:
> >>>>>>
> >>>>>> On Mon, Jul 17, 2023 at 04:02:35PM +0530, Bhupesh Sharma wrote:
> >>>>>>> Add the Embedded USB Debugger(EUD) device tree node for
> >>>>>>> SM6115 / SM4250 SoC.
> >>>>>>>
> >>>>>>> The node contains EUD base register region, EUD mode manager
> >>>>>>> register region and TCSR Base register region along with the
> >>>>>>> interrupt entry.
> >>>>>>>
> >>>>>>> [...]
> >>>>>>>
> >>>>>>> Reviewed-by: Konrad Dybcio <konrad.dybcio@xxxxxxxxxx>
> >>>>>>> Signed-off-by: Bhupesh Sharma <bhupesh.sharma@xxxxxxxxxx>
> >>>>>>> ---
> >>>>>>>  arch/arm64/boot/dts/qcom/sm6115.dtsi | 50 ++++++++++++++++++++++++++++
> >>>>>>>  1 file changed, 50 insertions(+)
> >>>>>>>
> >>>>>>> diff --git a/arch/arm64/boot/dts/qcom/sm6115.dtsi b/arch/arm64/boot/dts/qcom/sm6115.dtsi
> >>>>>>> index 839c603512403..db45337c1082c 100644
> >>>>>>> --- a/arch/arm64/boot/dts/qcom/sm6115.dtsi
> >>>>>>> +++ b/arch/arm64/boot/dts/qcom/sm6115.dtsi
> >>>>>>> [...]
> >>>>>>> @@ -789,6 +801,37 @@ gcc: clock-controller@1400000 {
> >>>>>>>                       #power-domain-cells = <1>;
> >>>>>>>               };
> >>>>>>>
> >>>>>>> +             eud: eud@1610000 {
> >>>>>>> +                     compatible = "qcom,sm6115-eud", "qcom,eud";
> >>>>>>> +                     reg = <0x0 0x01610000 0x0 0x2000>,
> >>>>>>> +                           <0x0 0x01612000 0x0 0x1000>,
> >>>>>>> +                           <0x0 0x003c0000 0x0 0x40000>;
> >>>>>>> +                     reg-names = "eud-base", "eud-mode-mgr", "tcsr-base";
> >>>>>>
> >>>>>> TCSR is a separate hardware block unrelated to the EUD. IMHO it
> >>>>>> shouldn't be listed as "reg" here.
> >>>>>>
> >>>>>> Typically we describe it as syscon and then reference it from other
> >>>>>> nodes. See e.g. sm8450.dtsi "tcsr: syscon@1fc0000" referenced in &scm
> >>>>>> "qcom,dload-mode = <&tcsr 0x13000>". This is pretty much exactly the
> >>>>>> same use case as you have. It also uses this to write something with
> >>>>>> qcom_scm_io_writel() at the end.
> >>>>>
> >>>>> That was discussed a bit during v1 patchset review. Basically, if we
> >>>>> use a tcsr syscon approach here, we will need to define a 'qcom,xx'
> >>>>> vendor specific dt-property and use something like this in the eud
> >>>>> node:
> >>>>>
> >>>>> qcom,eud-sec-reg = <&tcsr_reg yyyy>
> >>>>>
> >>>>> which would be then used by the eud driver (via
> >>>>> syscon_regmap_lookup_by_phandle()).
> >>>>>
> >>>>> But for sm6115 / qcm2290 this would be an over complicated solution as
> >>>>> normally the eud driver (say sc7280) doesn't need tcsr based secure
> >>>>> mode manager access. So defining a new soc / vendor specific
> >>>>> dt-property might be an overkill.
> >>>>>
> >>>>
> >>>> IMO a vendor-specific DT property is still better than messing up the
> >>>> device separation in the device tree. The same "tcsr-base" reg would
> >>>> also appear on the actual tcsr syscon device tree node. Having two
> >>>> device tree nodes with the same reg region is generally not valid.
> >>>>
> >>>> Something like qcom,eud-sec-reg = <&tcsr_reg yyyy> would at least make
> >>>> clear that this points into a region that is shared between multiple
> >>>> different devices, while adding it as reg suggests that TCSR belongs
> >>>> exclusively to EUD.
> >>>
> >>> I understand your point but since for sm6115 / qcm2290 devices TCSR is
> >>> not used for any other purpose than EUD, I still think introducing a
> >>> new soc / vendor specific dt-property might be an overkill for this
> >>> changeset.
> >> Untrue, there's some mumblings around the PHY properties and PSHOLD.
> >> I think Stephan may be correct here.
> >
> > Can you share the links to those discussions?
> It just seemed off to me that TCSR was not used by anything else (even
> from Linux, it would obviously be used by something else higher up in
> the boot chain as it contains various configuration registers), so I
> took a glance at the downstream device tree and I noticed there are
> more users.

Ok, let me recheck the downstream code and come back.

Thanks,
Bhupesh