| Message ID | 20221213-sm6350-cci-v2-2-15c2c14c34bb@fairphone.com |
|---|---|
| State | Not Applicable |
| Headers | show |
| Series | Add CCI bus support for SM6350 | expand |
On 20/01/2023 13:13, Luca Weiss wrote: > + camcc: clock-controller@ad00000 { > + compatible = "qcom,sm6350-camcc"; > + reg = <0 0x0ad00000 0 0x16000>; > + clocks = <&rpmhcc RPMH_CXO_CLK>; > + #clock-cells = <1>; > + #reset-cells = <1>; > + #power-domain-cells = <1>; > + }; Should you include required-opps = <&rpmhpd_opp_low_svs>; ? --- bod
On Fri Jan 20, 2023 at 5:49 PM CET, Bryan O'Donoghue wrote: > On 20/01/2023 13:13, Luca Weiss wrote: > > + camcc: clock-controller@ad00000 { > > + compatible = "qcom,sm6350-camcc"; > > + reg = <0 0x0ad00000 0 0x16000>; > > + clocks = <&rpmhcc RPMH_CXO_CLK>; > > + #clock-cells = <1>; > > + #reset-cells = <1>; > > + #power-domain-cells = <1>; > > + }; > > Should you include > > required-opps = <&rpmhpd_opp_low_svs>; > > ? I don't know, it works without. But doesn't this property not just affect power-domains? I haven't passed any here. > > --- > bod
On 24/01/2023 14:48, Luca Weiss wrote: > On Fri Jan 20, 2023 at 5:49 PM CET, Bryan O'Donoghue wrote: >> On 20/01/2023 13:13, Luca Weiss wrote: >>> + camcc: clock-controller@ad00000 { >>> + compatible = "qcom,sm6350-camcc"; >>> + reg = <0 0x0ad00000 0 0x16000>; >>> + clocks = <&rpmhcc RPMH_CXO_CLK>; >>> + #clock-cells = <1>; >>> + #reset-cells = <1>; >>> + #power-domain-cells = <1>; >>> + }; >> >> Should you include >> >> required-opps = <&rpmhpd_opp_low_svs>; >> >> ? > > I don't know, it works without. But doesn't this property not just > affect power-domains? I haven't passed any here. > Should you have a TITAN_TOP pd though ? --- bod
On Tue Jan 24, 2023 at 4:25 PM CET, Bryan O'Donoghue wrote: > On 24/01/2023 14:48, Luca Weiss wrote: > > On Fri Jan 20, 2023 at 5:49 PM CET, Bryan O'Donoghue wrote: > >> On 20/01/2023 13:13, Luca Weiss wrote: > >>> + camcc: clock-controller@ad00000 { > >>> + compatible = "qcom,sm6350-camcc"; > >>> + reg = <0 0x0ad00000 0 0x16000>; > >>> + clocks = <&rpmhcc RPMH_CXO_CLK>; > >>> + #clock-cells = <1>; > >>> + #reset-cells = <1>; > >>> + #power-domain-cells = <1>; > >>> + }; > >> > >> Should you include > >> > >> required-opps = <&rpmhpd_opp_low_svs>; > >> > >> ? > > > > I don't know, it works without. But doesn't this property not just > > affect power-domains? I haven't passed any here. > > > > Should you have a TITAN_TOP pd though ? Can I reference <&camcc TITAN_TOP_GDSC> from itself? I know that having it on is required to turn on at least some clocks (maybe all clocks). But from what I understand how power domains are normally handled, the driver core enables them before the driver is probed, so self referencing wouldn't work. And at least no other SoC upstream references TITAN_TOP_GDSC in camcc. Regards Luca > > --- > bod
On 27/01/2023 12:45, Luca Weiss wrote: > Can I reference <&camcc TITAN_TOP_GDSC> from itself? I know that having > it on is required to turn on at least some clocks (maybe all clocks). > But from what I understand how power domains are normally handled, the > driver core enables them before the driver is probed, so self > referencing wouldn't work. > > And at least no other SoC upstream references TITAN_TOP_GDSC in camcc. > > Regards > Luca Doh I meant to say a power-domain to an mmcx a la power-domains = <&rpmhpd SM8250_MMCX>; required-opps = <&rpmhpd_opp_low_svs>; TITAN_TOP should be in your cci and camss dt nodes. --- bod
On Fri Jan 27, 2023 at 1:49 PM CET, Bryan O'Donoghue wrote: > On 27/01/2023 12:45, Luca Weiss wrote: > > Can I reference <&camcc TITAN_TOP_GDSC> from itself? I know that having > > it on is required to turn on at least some clocks (maybe all clocks). > > But from what I understand how power domains are normally handled, the > > driver core enables them before the driver is probed, so self > > referencing wouldn't work. > > > > And at least no other SoC upstream references TITAN_TOP_GDSC in camcc. > > > > Regards > > Luca > > Doh I meant to say a power-domain to an mmcx a la > > power-domains = <&rpmhpd SM8250_MMCX>; > required-opps = <&rpmhpd_opp_low_svs>; > > TITAN_TOP should be in your cci and camss dt nodes. Okay, that makes more sense. What I don't quite understand is why sm8250 only has MMCX listed there since downstream has both vdd_mx-supply = <&VDD_MX_LEVEL> and vdd_mm-supply = <&VDD_MMCX_LEVEL> and both "supplies" are used for different clocks using .vdd_class But back to sm6350, downstream has vdd_mx-supply = <&VDD_MX_LEVEL> and vdd_cx-supply = <&VDD_CX_LEVEL> and like sm8250 uses cx and mx for different clocks. Not sure if I should add both, and I guess mainline also currently doesn't use higher ops for the power domain when higher clock rate is needed, from what I understand? > > --- > bod
On 27.01.2023 14:11, Luca Weiss wrote: > On Fri Jan 27, 2023 at 1:49 PM CET, Bryan O'Donoghue wrote: >> On 27/01/2023 12:45, Luca Weiss wrote: >>> Can I reference <&camcc TITAN_TOP_GDSC> from itself? I know that having >>> it on is required to turn on at least some clocks (maybe all clocks). >>> But from what I understand how power domains are normally handled, the >>> driver core enables them before the driver is probed, so self >>> referencing wouldn't work. >>> >>> And at least no other SoC upstream references TITAN_TOP_GDSC in camcc. >>> >>> Regards >>> Luca >> >> Doh I meant to say a power-domain to an mmcx a la >> >> power-domains = <&rpmhpd SM8250_MMCX>; >> required-opps = <&rpmhpd_opp_low_svs>; >> >> TITAN_TOP should be in your cci and camss dt nodes. > > Okay, that makes more sense. > > What I don't quite understand is why sm8250 only has MMCX listed there > since downstream has both vdd_mx-supply = <&VDD_MX_LEVEL> and > vdd_mm-supply = <&VDD_MMCX_LEVEL> and both "supplies" are used for > different clocks using .vdd_class > > But back to sm6350, downstream has vdd_mx-supply = <&VDD_MX_LEVEL> and > vdd_cx-supply = <&VDD_CX_LEVEL> and like sm8250 uses cx and mx for > different clocks. > Not sure if I should add both, and I guess mainline also currently > doesn't use higher ops for the power domain when higher clock rate is > needed, from what I understand? Basically if you don't need to power any of these power rails to have access to the clock controller, you don't need any of them. What you will need to do however, is make sure that they are scaled with child devices then.. but that's no bueno since they all need TITAN_GDSC. That's why Bryan suggests leaving a vote on a power rail in the clock controller, so that if no other votes are present (as improbable as that may be), you will still be able to get the clocks going. That OTOH will require you to add power management support (PM ops) to the clock controller, as otherwise you can say goodbye to battery life.. Konrad > >> >> --- >> bod >
On 27/01/2023 13:11, Luca Weiss wrote: >> Doh I meant to say a power-domain to an mmcx a la >> >> power-domains = <&rpmhpd SM8250_MMCX>; >> required-opps = <&rpmhpd_opp_low_svs>; >> >> TITAN_TOP should be in your cci and camss dt nodes. > Okay, that makes more sense. > > What I don't quite understand is why sm8250 only has MMCX listed there > since downstream has both vdd_mx-supply = <&VDD_MX_LEVEL> and > vdd_mm-supply = <&VDD_MMCX_LEVEL> and both "supplies" are used for > different clocks using .vdd_class power-domains = <&rpmhpd SM8250_MMCX>; == MMCX_LEVEL required for camcc power-domains = <&camcc TITAN_TOP_GDSC>; required for cci/camss now that you ask the question about MX_LEVEL you're making me doubt we have a 100% complete representation upstream TB perfectly honest, warrants a deep dive.. I just remember that on 8250 we tripped over MMCX not being switched on when - display I think was switched off. --- bod
On 27.01.2023 14:54, Bryan O'Donoghue wrote: > On 27/01/2023 13:11, Luca Weiss wrote: >>> Doh I meant to say a power-domain to an mmcx a la >>> >>> power-domains = <&rpmhpd SM8250_MMCX>; >>> required-opps = <&rpmhpd_opp_low_svs>; >>> >>> TITAN_TOP should be in your cci and camss dt nodes. >> Okay, that makes more sense. >> >> What I don't quite understand is why sm8250 only has MMCX listed there >> since downstream has both vdd_mx-supply = <&VDD_MX_LEVEL> and >> vdd_mm-supply = <&VDD_MMCX_LEVEL> and both "supplies" are used for >> different clocks using .vdd_class > > power-domains = <&rpmhpd SM8250_MMCX>; == MMCX_LEVEL required for camcc > power-domains = <&camcc TITAN_TOP_GDSC>; required for cci/camss > > now that you ask the question about MX_LEVEL you're making me doubt we have a 100% complete representation upstream TB perfectly honest, warrants a deep dive.. > > I just remember that on 8250 we tripped over MMCX not being switched on when - display I think was switched off. There's no MMCX on 6350 and MX is a parent of CX, so if we just stick CX here and add the lowest level to required-opps and add corresponding PM ops to the clk driver, it'll all be taken care of! Konrad > > --- > bod >
diff --git a/arch/arm64/boot/dts/qcom/sm6350.dtsi b/arch/arm64/boot/dts/qcom/sm6350.dtsi index 8224adb99948..300ced5cda57 100644 --- a/arch/arm64/boot/dts/qcom/sm6350.dtsi +++ b/arch/arm64/boot/dts/qcom/sm6350.dtsi @@ -1435,6 +1435,15 @@ usb_1_dwc3: usb@a600000 { }; }; + camcc: clock-controller@ad00000 { + compatible = "qcom,sm6350-camcc"; + reg = <0 0x0ad00000 0 0x16000>; + clocks = <&rpmhcc RPMH_CXO_CLK>; + #clock-cells = <1>; + #reset-cells = <1>; + #power-domain-cells = <1>; + }; + pdc: interrupt-controller@b220000 { compatible = "qcom,sm6350-pdc", "qcom,pdc"; reg = <0 0x0b220000 0 0x30000>, <0 0x17c000f0 0 0x64>;