Re: [PATCH v6 07/11] arm64: dts: qcom: sm6115: Add Crypto Engine support

[Stephan Gerhold]

Hi Bhupesh,

Not sure if this is the latest version of this series since it's pretty 
old but I didn't find a new one. Just came here because you mentioned 
RB1/RB2 [1] in my bam_dma patch and they don't have any BAM defined
upstream yet.

[1]: https://lore.kernel.org/linux-arm-msm/CAH=2Ntw0BZH=RGp14mYLhX7D6jV5O5eDKRQbby=uCy85xMDU_g@xxxxxxxxxxxxxx/

On Wed, Apr 05, 2023 at 12:58:32PM +0530, Bhupesh Sharma wrote:
> Add crypto engine (CE) and CE BAM related nodes and definitions to
> 'sm6115.dtsi'.
> 
> Signed-off-by: Bhupesh Sharma <bhupesh.sharma@xxxxxxxxxx>
> ---
>  arch/arm64/boot/dts/qcom/sm6115.dtsi | 22 ++++++++++++++++++++++
>  1 file changed, 22 insertions(+)
> 
> diff --git a/arch/arm64/boot/dts/qcom/sm6115.dtsi b/arch/arm64/boot/dts/qcom/sm6115.dtsi
> index 2a51c938bbcb..ebac026b4cc7 100644
> --- a/arch/arm64/boot/dts/qcom/sm6115.dtsi
> +++ b/arch/arm64/boot/dts/qcom/sm6115.dtsi
> @@ -650,6 +650,28 @@ usb_hsphy: phy@1613000 {
>  			status = "disabled";
>  		};
>  
> +		cryptobam: dma-controller@1b04000 {
> +			compatible = "qcom,bam-v1.7.4", "qcom,bam-v1.7.0";
> +			reg = <0x0 0x01b04000 0x0 0x24000>;
> +			interrupts = <GIC_SPI 247 IRQ_TYPE_LEVEL_HIGH>;
> +			#dma-cells = <1>;
> +			qcom,ee = <0>;
> +			qcom,controlled-remotely;
> +			num-channels = <8>;
> +			qcom,num-ees = <2>;
> +			iommus = <&apps_smmu 0x94 0x11>,
> +				 <&apps_smmu 0x96 0x11>;
> +		};
> +
> +		crypto: crypto@1b3a000 {
> +			compatible = "qcom,sm6115-qce", "qcom,sm8150-qce", "qcom,qce";
> +			reg = <0x0 0x01b3a000 0x0 0x6000>;
> +			dmas = <&cryptobam 6>, <&cryptobam 7>;
> +			dma-names = "rx", "tx";
> +			iommus = <&apps_smmu 0x94 0x11>,
> +				 <&apps_smmu 0x96 0x11>;

Shouldn't you have clocks = <&rpmcc RPM_SMD_CE1_CLK> here to make sure
the clock for the crypto engine is on? Your binding patch (PATCH 06/11)
says "Crypto Engine block on Qualcomm SoCs SM6115 and QCM2290 do not
require clocks strictly" but doesn't say why.

Make sure you don't rely on having rpmcc keep unused clocks on
permanently. This is the case at the moment, but we would like to change
this [2]. Adding new users that rely on this broken behavior would just
make this effort even more complicated.

If you also add the clock to the cryptobam then you should be able to
see the advantage of my bam_dma patch [3]. It allows you to drop
"num-channels" and "qcom,num-ees" from the cryptobam in your changes
above because it can then be read directly from the BAM registers.

Thanks,
Stephan

[2]: https://lore.kernel.org/linux-arm-msm/20230303-topic-rpmcc_sleep-v2-0-ae80a325fe94@xxxxxxxxxx/
[3]: https://lore.kernel.org/linux-arm-msm/20230518-bamclk-dt-v1-1-82f738c897d9@xxxxxxxxxxx/