On Tue, 23 Jan 2024 at 15:43, Ritesh Kumar <quic_riteshk@xxxxxxxxxxx> wrote:
>
>
> On 1/16/2024 6:27 PM, Dmitry Baryshkov wrote:
>
> > On Tue, 16 Jan 2024 at 14:06, Konrad Dybcio <konrad.dybcio@xxxxxxxxxx> wrote:
> >>
> >>
> >> On 1/16/24 10:49, Ritesh Kumar wrote:
> >>> Enable Display Subsystem with Novatek NT36672E Panel
> >>> on qcm6490 idp platform.
> >>>
> >>> Signed-off-by: Ritesh Kumar <quic_riteshk@xxxxxxxxxxx>
> >>> ---
> >>> arch/arm64/boot/dts/qcom/qcm6490-idp.dts | 100 +++++++++++++++++++++++
> >>> 1 file changed, 100 insertions(+)
> >>>
> >>> diff --git a/arch/arm64/boot/dts/qcom/qcm6490-idp.dts b/arch/arm64/boot/dts/qcom/qcm6490-idp.dts
> >>> index 2a6e4907c5ee..efa5252130a1 100644
> >>> --- a/arch/arm64/boot/dts/qcom/qcm6490-idp.dts
> >>> +++ b/arch/arm64/boot/dts/qcom/qcm6490-idp.dts
> >>> @@ -9,6 +9,7 @@
> >>> #define PM7250B_SID 8
> >>> #define PM7250B_SID1 9
> >>>
> >>> +#include <dt-bindings/pinctrl/qcom,pmic-gpio.h>
> >>> #include <dt-bindings/regulator/qcom,rpmh-regulator.h>
> >>> #include "sc7280.dtsi"
> >>> #include "pm7250b.dtsi"
> >>> @@ -38,6 +39,25 @@
> >>> stdout-path = "serial0:115200n8";
> >>> };
> >>>
> >>> + lcd_disp_bias: lcd-disp-bias-regulator {
> >>> + compatible = "regulator-fixed";
> >>> + regulator-name = "lcd_disp_bias";
> >>> + regulator-min-microvolt = <5500000>;
> >>> + regulator-max-microvolt = <5500000>;
> >>> + gpio = <&pm7250b_gpios 2 GPIO_ACTIVE_HIGH>;
> >>> + enable-active-high;
> >>> + pinctrl-names = "default";
> >>> + pinctrl-0 = <&lcd_disp_bias_en>;
> >> property-n
> >> property-names
> >>
> >> all throughout the patch
>
> Thanks, I will update in the new version.
>
> >>> +&gpu {
> >>> + status = "disabled";
> >>> +};
> >> Hm.. generally we disable the GPU in the SoC DT, but that doesn't
> >> seem to have happened here..
> >>
> >> Thinking about it more, is disabling it here necessary? Does it
> >> not fail gracefully?
> > Missed this.
> >
> > I'd say, I don't see a reason to disable it at all. The GPU should be
> > working on sc7280 / qcm4290.
>
> With GPU device node enabled, adreno_bind failure is seen as the
> "speed_bin" was not populated on QCM6490 target which leads to display
> bind failure.
Excuse me please. The GPU node for sc7280 already has speed_bin, which
points to qfprom + 0x1e9, bits 5 to 9.
Do you mean that qcm6490 uses different speed bin location? Or
different values for the speed bins?
> Spoke with GPU team and on QCM6490 board, only CPU rendering is
> supported for now and there is no plan to enable GPU rendering in near
> future.
This sounds like having the feature disabled for no particular reason.
Both the kernel and Mesa have supported the Adreno 635 for quite a
while.
> In this regard, what do you suggest
>
> 1) Disable GPU in QCM6490 DT (as per the current patch)
> 2) Disable GPU in the SoC DT, but enable it in other platform DTs. (This
> will prompt change in all the dt's and we don't have all the devices to
> test)
The second option definitely follows what is present on other platforms.
> Please let me know your views on it.
Please enable the GPU instead.
--
With best wishes
Dmitry
