Re: [PATCH v2 4/4] arm64: dts: qcom: sm6350: Add video clock controller

[Konrad Dybcio]

On 4/15/25 6:05 AM, Taniya Das wrote:
> 
> 
> On 4/12/2025 12:56 AM, Konrad Dybcio wrote:
>> On 4/11/25 1:37 PM, Jagadeesh Kona wrote:
>>>
>>>
>>> On 4/11/2025 2:42 PM, Konrad Dybcio wrote:
>>>> On 4/11/25 9:15 AM, Jagadeesh Kona wrote:
>>>>>
>>>>>
>>>>> On 4/1/2025 10:03 PM, Konrad Dybcio wrote:
>>>>>> On 3/24/25 9:41 AM, Luca Weiss wrote:
>>>>>>> Add a node for the videocc found on the SM6350 SoC.
>>>>>>>
>>>>>>> Signed-off-by: Luca Weiss <luca.weiss@xxxxxxxxxxxxx>
>>>>>>> ---
>>>>>>>  arch/arm64/boot/dts/qcom/sm6350.dtsi | 14 ++++++++++++++
>>>>>>>  1 file changed, 14 insertions(+)
>>>>>>>
>>>>>>> diff --git a/arch/arm64/boot/dts/qcom/sm6350.dtsi b/arch/arm64/boot/dts/qcom/sm6350.dtsi
>>>>>>> index 42f9d16c2fa6da66a8bb524a33c2687a1e4b40e0..4498d6dfd61a7e30a050a8654d54dae2d06c220c 100644
>>>>>>> --- a/arch/arm64/boot/dts/qcom/sm6350.dtsi
>>>>>>> +++ b/arch/arm64/boot/dts/qcom/sm6350.dtsi
>>>>>>> @@ -1952,6 +1952,20 @@ usb_1_dwc3_ss_out: endpoint {
>>>>>>>  			};
>>>>>>>  		};
>>>>>>>  
>>>>>>> +		videocc: clock-controller@aaf0000 {
>>>>>>> +			compatible = "qcom,sm6350-videocc";
>>>>>>> +			reg = <0x0 0x0aaf0000 0x0 0x10000>;
>>>>>>> +			clocks = <&gcc GCC_VIDEO_AHB_CLK>,
>>>>>>> +				 <&rpmhcc RPMH_CXO_CLK>,
>>>>>>> +				 <&sleep_clk>;
>>>>>>> +			clock-names = "iface",
>>>>>>> +				      "bi_tcxo",
>>>>>>> +				      "sleep_clk";
>>>>>>> +			#clock-cells = <1>;
>>>>>>> +			#reset-cells = <1>;
>>>>>>> +			#power-domain-cells = <1>;
>>>>>>> +		};
>>>>>>
>>>>>> You'll probably want to hook up some additional power domains here, see
>>>>>>
>>>>>> https://lore.kernel.org/linux-arm-msm/20250327-videocc-pll-multi-pd-voting-v3-0-895fafd62627@xxxxxxxxxxx/
>>>>>>
>>>>>
>>>>> On SM6350, videocc doesn't need multiple power domains at HW level, it is only on CX rail which would be ON
>>>>> when system is active, hence power-domains are not mandatory here.
>>>>
>>>> 6350 doesn't have either MMCX nor a split MX - shouldn't both normal
>>>> CX and MX be in there?
>>>>
>>>
>>> All clocks & GDSC's of SM6350 videocc are only on CX rail, so it requires only CX power domain. But when HLOS
>>> is active, CX rail will be ON and operate at a level above retention, which is sufficient for videocc to operate.
>>> Hence clock driver don't need to explicitly vote on CX rail.
>>>
>>> The same is not true for other rails like MMCX and Split MX(MXC), hence clock drivers had to explicitly vote on
>>> those rails.
>>
>> I'm worried about MX being undervolted for higher OPPs
>>
> 
> From a videocc PoV there is no requirement of Mx on SM6350. The CX
> levels would be taken care by Video SW driver from their defined OPP. Mx
> at system level would be catered via the BW votes.

So I'm specifically thinking about the videocc (and other) PLLs, which
have defined vdd levels downstream - currently we're relying on random
luck rather than ensuring each one of them has its requirements fulfilled

Konrad