On 03/11/2015 09:26 PM, Tony Lindgren wrote:

* Tero Kristo [off-list ref] [150311 12:09]:

quoted

On 03/11/2015 07:17 PM, Tony Lindgren wrote:

quoted

Hi Tero,

* Tero Kristo [off-list ref] [150225 11:09]:

quoted

Add node for system control module, and move all the existing system
control IO space users under this new node as its children. A new node
for scm_conf area is also added.

...

quoted

--- a/arch/arm/boot/dts/dra7.dtsi
+++ b/arch/arm/boot/dts/dra7.dtsi

@@ -203,26 +203,47 @@
  			};
  		};

+		scm: scm at 4a002000 {
+			compatible = "ti,dra7-ctrl", "simple-bus";
+			reg = <0x4a002000 0x1400>,
+			      <0x4a003400 0x600>,
+			      <0x4ae0c000 0x600>;
+			#address-cells = <2>;
+			#size-cells = <1>;
+			ranges = <0 0 0x4a002000 0x1400>,
+				 <1 0 0x4a003400 0x600>,
+				 <2 0 0x4ae0c000 0x600>;
+
+			scm_conf: tisyscon at 0,0 {
+				compatible = "syscon";
+				reg = <0 0x0 0x1400>;
+				#address-cells = <1>;
+				#size-cells = <1>;
+			};
+
+			dra7_pmx_core: pinmux at 1,0 {
+				compatible = "ti,dra7-padconf",
+					     "pinctrl-single";
+				reg = <1 0x0 0x0464>;
+				#address-cells = <1>;
+				#size-cells = <0>;
+				#interrupt-cells = <1>;
+				interrupt-controller;
+				pinctrl-single,register-width = <32>;
+				pinctrl-single,function-mask = <0x3fffffff>;
+			};
+		};

Wouldn't it make more sense to have separate device_scm, core_scm and
wkup_scm instead of stuffing multiple ranges here?

Or are there other reasons for the multiple ranges?

Yea that was the alternative I was thinking about, I ended up with this for
some reason. I think personally I liked having them all under the same SCM
part, because they are nicely grouped then, and well, its the same system
control part in the chip. We can split it up easily of course. Should we
have a higher level scm part and then have core_scm and wkup_scm under this
followed by the sub-functions, or just drop the top level scm part
completely?

Well I'd model it after the hardware so we can have one or more scm driver
instances managing the clock for those blocks. If we squash them together,
we won't have a chance to pass interrupts and clocks device tree property
to the right driver instance. And for example 5432 TRM has them as separate
devices in "Figure 18-1. Control Module Overview".

I don't think we need the top level scm to group them under, these are all
connected seprately to the interconnect, right?

Yea, can't really think of any real need for the top-level node.

quoted

This same question applies to omap4 + omap5 also. In some part for omap3
also, as it also has pmx_core + pmx_wkup separately, even if they are part
of the same register space.

Anyway, just a political decision from your side, I am fine either way. :)

OK thanks for confirming that, to me it makes sense to set them up as
separate instances then.

All right, you got fair points there, I'll rework this for next revision 
of the set. Had a quick look at OMAP3 TRM and it is also basically 
listing these as separate instances also, so I'll change all OMAP3+.

-Tero