Re: [PATCH V2 RESEND 6/6] arm64: dts: qcom: sm8650: Add video and camera clock controllers

[Dmitry Baryshkov <dmitry.baryshkov@linaro.org>]

On Fri, 5 Apr 2024 at 09:01, Jagadeesh Kona <quic_jkona@quicinc.com> wrote:
>
>
>
> On 4/4/2024 9:35 PM, Dmitry Baryshkov wrote:
> > On Thu, 4 Apr 2024 at 13:06, Jagadeesh Kona <quic_jkona@quicinc.com> wrote:
> >>
> >>
> >>
> >> On 4/4/2024 11:00 AM, Dmitry Baryshkov wrote:
> >>> On Thu, 4 Apr 2024 at 08:13, Jagadeesh Kona <quic_jkona@quicinc.com> wrote:
> >>>>
> >>>>
> >>>>
> >>>> On 4/3/2024 9:24 PM, Dmitry Baryshkov wrote:
> >>>>> On Wed, 3 Apr 2024 at 10:16, Jagadeesh Kona <quic_jkona@quicinc.com> wrote:
> >>>>>>
> >>>>>>
> >>>>>>
> >>>>>> On 3/25/2024 11:38 AM, Jagadeesh Kona wrote:
> >>>>>>>
> >>>>>>>
> >>>>>>> On 3/21/2024 6:43 PM, Dmitry Baryshkov wrote:
> >>>>>>>> On Thu, 21 Mar 2024 at 11:27, Jagadeesh Kona <quic_jkona@quicinc.com>
> >>>>>>>> wrote:
> >>>>>>>>>
> >>>>>>>>> Add device nodes for video and camera clock controllers on Qualcomm
> >>>>>>>>> SM8650 platform.
> >>>>>>>>>
> >>>>>>>>> Signed-off-by: Jagadeesh Kona <quic_jkona@quicinc.com>
> >>>>>>>>> ---
> >>>>>>>>>      arch/arm64/boot/dts/qcom/sm8650.dtsi | 28 ++++++++++++++++++++++++++++
> >>>>>>>>>      1 file changed, 28 insertions(+)
> >>>>>>>>>
> >>>>>>>>> diff --git a/arch/arm64/boot/dts/qcom/sm8650.dtsi
> >>>>>>>>> b/arch/arm64/boot/dts/qcom/sm8650.dtsi
> >>>>>>>>> index 32c0a7b9aded..d862aa6be824 100644
> >>>>>>>>> --- a/arch/arm64/boot/dts/qcom/sm8650.dtsi
> >>>>>>>>> +++ b/arch/arm64/boot/dts/qcom/sm8650.dtsi
> >>>>>>>>> @@ -4,6 +4,8 @@
> >>>>>>>>>       */
> >>>>>>>>>
> >>>>>>>>>      #include <dt-bindings/clock/qcom,rpmh.h>
> >>>>>>>>> +#include <dt-bindings/clock/qcom,sm8450-videocc.h>
> >>>>>>>>> +#include <dt-bindings/clock/qcom,sm8650-camcc.h>
> >>>>>>>>>      #include <dt-bindings/clock/qcom,sm8650-dispcc.h>
> >>>>>>>>>      #include <dt-bindings/clock/qcom,sm8650-gcc.h>
> >>>>>>>>>      #include <dt-bindings/clock/qcom,sm8650-gpucc.h>
> >>>>>>>>> @@ -3110,6 +3112,32 @@ opp-202000000 {
> >>>>>>>>>                             };
> >>>>>>>>>                     };
> >>>>>>>>>
> >>>>>>>>> +               videocc: clock-controller@aaf0000 {
> >>>>>>>>> +                       compatible = "qcom,sm8650-videocc";
> >>>>>>>>> +                       reg = <0 0x0aaf0000 0 0x10000>;
> >>>>>>>>> +                       clocks = <&bi_tcxo_div2>,
> >>>>>>>>> +                                <&gcc GCC_VIDEO_AHB_CLK>;
> >>>>>>>>> +                       power-domains = <&rpmhpd RPMHPD_MMCX>;
> >>>>>>>>> +                       required-opps = <&rpmhpd_opp_low_svs>;
> >>>>>>>>
> >>>>>>>> The required-opps should no longer be necessary.
> >>>>>>>>
> >>>>>>>
> >>>>>>> Sure, will check and remove this if not required.
> >>>>>>
> >>>>>>
> >>>>>> I checked further on this and without required-opps, if there is no vote
> >>>>>> on the power-domain & its peer from any other consumers, when runtime
> >>>>>> get is called on device, it enables the power domain just at the minimum
> >>>>>> non-zero level. But in some cases, the minimum non-zero level of
> >>>>>> power-domain could be just retention and is not sufficient for clock
> >>>>>> controller to operate, hence required-opps property is needed to specify
> >>>>>> the minimum level required on power-domain for this clock controller.
> >>>>>
> >>>>> In which cases? If it ends up with the retention vote, it is a bug
> >>>>> which must be fixed.
> >>>>>
> >>>>
> >>>> The minimum non-zero level(configured from bootloaders) of MMCX is
> >>>> retention on few chipsets but it can vary across the chipsets. Hence to
> >>>> be on safer side from our end, it is good to have required-opps in DT to
> >>>> specify the minimum level required for this clock controller.
> >>>
> >>> We are discussing sm8650, not some abstract chipset. Does it list
> >>> retention or low_svs as a minimal level for MMCX?
> >>>
> >>
> >> Actually, the minimum level for MMCX is external to the clock
> >> controllers.
> >
> > Yes, it comes from cmd-db
> >
> >>   But the clock controller requires MMCX to be atleast at
> >> lowsvs for it to be functional.
> >
> > Correct
> >
> >> Hence we need to keep required-opps to
> >> ensure the same without relying on the actual minimum level for MMCX.
> >
> > And this is not correct. There is no need for the DT to be redundant.
> > I plan to send patches removing the existing required-opps when they
> > are not required.
> >
> I agree this is not required if cmd-db minimum level is already at
> lowsvs. But since MMCX running at lowsvs is a mandatory requirement for
> clock controller to operate, I believe it is good to have required-opps
> to ensure we meet this requirement in all cases, rather than relying on
> the cmd-db minimum level which we have no control over.

IIf we follow this logic, we should throw cmd-db away and hardcode all
those values in the RPMh drivers.

We have cmd-db. If it is correct, there is no need to duplicate it. If
it is incorrect, it is a bug that should be fixed or worked around.

>
> Thanks,
> Jagadeesh
>


-- 
With best wishes
Dmitry