Re: [PATCH] clk: stm32h7: Add stm32h743 clock driver

[Stephen Boyd <sboyd@codeaurora.org>]

On 04/06, Gabriel Fernandez wrote:
> On 04/06/2017 12:32 AM, Stephen Boyd wrote:
> >On 03/15, gabriel.fernandez@st.com wrote:
> >>diff --git a/Documentation/devicetree/bindings/clock/st,stm32h7-rcc.txt b/Documentation/devicetree/bindings/clock/st,stm32h7-rcc.txt
> >>new file mode 100644
> >>index 0000000..9d4b587
> >>--- /dev/null
> >>+++ b/Documentation/devicetree/bindings/clock/st,stm32h7-rcc.txt
> >>@@ -0,0 +1,152 @@
> >>+
> >>+	rcc: rcc@58024400 {
> >>+		#reset-cells = <1>;
> >>+		#clock-cells = <2>
> >>+		compatible = "st,stm32h743-rcc", "st,stm32-rcc";
> >>+		reg = <0x58024400 0x400>;
> >>+		clocks = <&clk_hse>, <&clk_lse>, <&clk_i2s_ckin>;
> >>+
> >>+		st,syscfg = <&pwrcfg>;
> >>+
> >>+		#address-cells = <1>;
> >>+		#size-cells = <0>;
> >>+
> >>+		vco1@58024430 {
> >>+			#clock-cells = <0>;
> >>+			compatible = "stm32,pll";
> >>+			reg = <0>;
> >reg is super confusing and doesn't match unit address.
> ok i fixed it in the v2
> 
> >
> >>+		};
> >Why? Shouldn't we know this from the compatible string how many
> >PLLs there are and where they're located? Export the PLLs through
> >rcc node's clock-cells?
> >
> Because i need to offer the possibility to change the PLL VCO
> frequencies at the start-up of this driver clock.
> The VCO algorithm needs a division factor, a multiplication factor
> and a fractional factor.
> Lot's of solution are possible for one frequency and it's nightmare
> to satisfy the 3 output dividers of the PLL.

Sure, but do we need to configure that on a per-board basis or a
per-SoC basis? If it's just some configuration, I wonder why we
don't put that into the driver and base it off some compatible
string that includes the SoC the device is for. 

-- 
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project