Re: Support for trees of clocks served by one clock manager

[Jared McNeill <jmcneill%invisible.ca@localhost>]

On Wed, 15 Aug 2018, Aymeric Vincent wrote:

> Sure, the clock tree is in
> sys/external/gpl2/dts/dist/arch/arm/boot/dts/socfpga.dtsi
> and I include below a dump of the exact dtb I use for completeness.

Thanks, that helps a lot.

> In order to avoid the loop, I could register all clocks as clock
> managers but that's _very_ ugly (it works).

I think this is fine. Consider the binding docs:

https://www.kernel.org/doc/Documentation/devicetree/bindings/clock/clock-bindings.txt
https://www.kernel.org/doc/Documentation/devicetree/bindings/clock/altr_socfpga.txt

In the generic binding docs, there is no mention of searching for parent 
clock controllers. The cross reference from clock consumer contains the 
phandle of the clock controller and optional specifier.

For SoCFPGA platform, it looks like each node under clock is meant to be a 
separate clock device, and in socfpga.dtsi there are multiple 
"fixed-clock" nodes for which we already have a driver 
(dev/fdt/fixedclock.c).

Walking the tree to find the clock controller might work in this case but 
it is contrary to the binding docs and could cause issues in a 
scenario where we have nested clock controllers and are missing a driver 
for the child one.

Cheers,
Jared