On 02/09/16 13:54, Rob Herring wrote:
> On Tue, Aug 30, 2016 at 04:53:01PM +0100, Paul Burton wrote:
>> On 26/08/16 18:44, Stephen Boyd wrote:
>>> On 08/26, Paul Burton wrote:
>>>> diff --git a/Documentation/devicetree/bindings/clock/img,boston-clock.txt b/Documentation/devicetree/bindings/clock/img,boston-clock.txt
>>>> new file mode 100644
>>>> index 0000000..c01ea60
>>>> --- /dev/null
>>>> +++ b/Documentation/devicetree/bindings/clock/img,boston-clock.txt
>>>> @@ -0,0 +1,27 @@
>>>> +Binding for Imagination Technologies MIPS Boston clock sources.
>>>> +
>>>> +This binding uses the common clock binding[1].
>>>> +
>>>> +[1] Documentation/devicetree/bindings/clock/clock-bindings.txt
>>>> +
>>>> +Required properties:
>>>> +- compatible : Should be "img,boston-clock".
>>>> +- #clock-cells : Should be set to 1.
>>>> + Values available for clock consumers can be found in the header file:
>>>> + <dt-bindings/clock/boston-clock.h>
>>>> +- regmap : Phandle to the Boston platform register system controller.
>>>> + This should contain a phandle to the system controller node covering the
>>>> + platform registers provided by the Boston board.
>>>> +
>>>> +Example:
>>>> +
>>>> + clk_boston: clock {
>>>> + compatible = "img,boston-clock";
>>>> + #clock-cells = <1>;
>>>> + regmap = <&plat_regs>;
>>>
>>> Isn't syscon more standard than regmap as the property name? Is
>>> there a binding for the plat_regs device? Is there any reason the
>>> clks can't be populated in that syscon driver?
>>
>> Hi Stephen,
>>
>> The plat_regs device doesn't have a custom driver, it simply makes use
>> of the generic "syscon" driver which can provide a regmap.
>>
>> It would be possible to register the clocks from a register for the
>> plat_regs device, but I don't think it would make much sense. The
>> platform registers in question are essentially just a convenient place
>> where various bits of information about the system are exposed,
>> including the clock frequencies but also other bits & pieces like
>> connectivity of PCIe controllers or I/O coherence units, the RTL
>> revision of the CPU or the wrapper RTL that runs on this FPGA-based
>> board, a register that allows for resetting the board, etc. It's not a
>> single piece of hardware, more a dumping ground for miscellanea. So in
>> my opinion using the syscon approach works best here, and drivers for
>> well defined pieces of hardware or functionality can reference that
>> syscon to retrieve the regmap.
>
> That is all quite common for any SoC. Whether it's 2 nodes or 2 drivers
> are independent questions. You can easily have 1 node and 2 drivers. The
> decision factor is really how many registers we're dealing with. We
> don't want to end up with a node per register or register field. That's
> too fine grained.
Absolutely, I don't think we disagree there.
>> As for whether "syscon" is a more standard property name than "regmap",
>> both seem to be used based on a grep of
>> Documentation/devicetree/bindings/. I believe I picked up use of
>> "regmap" from the generic syscon-poweroff & syscon-reboot drivers, which
>> both use "regmap" as a property name.
>
> syscon is much more common.
>
> Avoid the phandle altogether and make this a child node.
I could do that, but it would feel rather odd to describe the clock
hardware as a child of a bunch of miscellaneous registers that happen to
expose some information about those clocks. Is that really what you
prefer? I think as-is the DT is a better description of the hardware.
Thanks,
Paul
>
> Rob
>
