Re: [PATCH v6 2/5] dt-bindings: media: Add qcom,x1e80100-camss

[Dmitry Baryshkov]

On Thu, Apr 24, 2025 at 11:17:13AM +0100, Bryan O'Donoghue wrote:
> On 24/04/2025 11:07, Krzysztof Kozlowski wrote:
> > On 24/04/2025 11:34, Bryan O'Donoghue wrote:
> > > On 24/04/2025 07:40, Krzysztof Kozlowski wrote:
> > > > > +  vdd-csiphy-0p8-supply:
> > > > Same comment as other series on the lists - this is wrong name. There
> > > > are no pins named like this and all existing bindings use different name.
> > > 
> > > The existing bindings are unfortunately not granular enough.
> > > 
> > > I'll post s series to capture pin-names per the SoC pinout shortly.
> > How are the pins/supplies actually called?
> > 
> > Best regards,
> > Krzysztof
> 
> I don't think strictly algning to pin-names is what we want.
> 
> Here are the input pins
> 
> VDD_A_CSI_0_1_1P2
> VDD_A_CSI_2_4_1P2
> VDD_A_CSI_0_1_0P9
> VDD_A_CSI_2_4_0P9
> 
> I think the right way to represent this
> 
> yaml:
> csiphy0-1p2-supply
> csiphy1-1p2-supply
> csiphy2-1p2-supply
> csiphy3-1p2-supply
> 
> dts:
> 
> vdd-csiphy0-0p9-supply = <&vreg_l2c_0p8>;
> vdd-csiphy1-0p9-supply = <&vreg_l2c_0p8>;
> 
> etc
> 
> vdda-csiphy0-1p2-supply = <&vreg_l1c_1p2>;
> 
> because that captures the fact we could have separate lines for each phy,
> names it generically and then leaves it up to the dts implementation to
> represent what actually happened on the PCB.
> 
> That would also work for qcm2290 and sm8650.
> 
> https://lore.kernel.org/linux-arm-msm/20250423221954.1926453-2-vladimir.zapolskiy@xxxxxxxxxx/
> 
> So for sm8650 instead of
> 
> +  vdda-csi01-0p9-supply:
> +
> +  vdda-csi24-0p9-supply:
> +
> +  vdda-csi35-0p9-supply:
> +
> +  vdda-csi01-1p2-supply:
> +
> +  vdda-csi24-1p2-supply:
> +
> +  vdda-csi35-1p2-supply:
> 
> you would have
> 
> +  vdda-csiphy0-0p9-supply:
> +
> +  vdda-csiphy1-0p9-supply:
> 
> +  vdda-csiphy0-1p2-supply:
> +
> +  vdda-csiphy1-1p2-supply:
> 
> Which would then be consistent across SoCs for as long as 0p9 and 1p2 are
> the power-domains used by these PHYs.

This won't be consistent with other cases where we have a shared power
pin. For example, for PMICs we provide supply names which match pin
names rather than one-supply-per-LDO.

-- 
With best wishes
Dmitry