On 02/06, Florian Fainelli wrote:

On 02/06/2017 02:59 PM, Stephen Boyd wrote:

quoted

On 02/03, Florian Fainelli wrote:

quoted

We already have another piece of drive code that manipulates registers
in the Bus Interface Unit located in drivers/soc/bcm/brcmstb/biuctrl.c
and which has little to nothing to do with the CPU's clock ratio. And
actually another one being submitted that deals with the CPU's
read-ahead cache. I would very much prefer we keep all of them separate
and dealing with just the register offset they need to do, but that does
not mean the Device Tree binding has to look that way though.

The binding for the BIUCTRL register made it here:

Documentation/devicetree/bindings/arm/bcm/brcm,brcmstb.txt

so we should re-use that, and have a small piece of clock provided that
just uses the relevant register range within that larger register space
and provide the CLOCK_RATIO. Does that work?

Ok. That's fine. The existing binding will be updated to include
this new subnode then for the clock component?

Humm, I suppose we could do that yes, my original thought was to just
have this CPU clock provider remap the entire BIUCTRL register range
(of_iomap() etc.) and manipulate just the relevant register range of
interest (CPU_CLOCK_CONFIG_REG), but if you want a sub-node to appear,
we could probably do that as well.

One node is fine as well. I thought the plan was many subnodes
based on the binding document you mentioned earlier.

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