[RFC v2 4/9] of: add clock providers

[Rob Herring]

On 12/15/2011 07:51 AM, Shawn Guo wrote:
> On Tue, Dec 13, 2011 at 10:54:58AM -0700, Grant Likely wrote:
>> On Mon, Dec 12, 2011 at 4:29 PM, Jamie Iles <jamie at jamieiles.com> wrote:
>>> Hi Grant,
>>>
>>> I'm still going through these and trying to digest them but a couple of
>>> quick questions/comments.
>>>
>>> Jamie
>>>
>>> On Mon, Dec 12, 2011 at 03:02:04PM -0700, Grant Likely wrote:
>>>> diff --git a/Documentation/devicetree/bindings/clock/clock-bindings.txt b/Documentation/devicetree/bindings/clock/clock-bindings.txt
>>>> new file mode 100644
>>>> index 0000000..e40c436
>>>> --- /dev/null
>>>> +++ b/Documentation/devicetree/bindings/clock/clock-bindings.txt
>>>> @@ -0,0 +1,114 @@
>>>> +This binding is a work-in-progress, and are based on some experimental
>>>> +work by benh[1].
>>>> +
>>>> +Sources of clock signal can be represented by any node in the device
>>>> +tree.  Those nodes are designated as clock providers.  Clock consumer
>>>> +nodes use a phandle and clock specifier pair to connect clock provider
>>>> +outputs to clock inputs.  Similar to the gpio specifiers, a clock
>>>> +specifier is an array of one more more cells identifying the clock
>>>> +output on a device.  The length of a clock specifier is defined by the
>>>> +value of a #clock-cells property in the clock provider node.
>>>> +
>>>> +[1] http://patchwork.ozlabs.org/patch/31551/
>>>> +
>>>> +==Clock providers==
>>>> +
>>>> +Required properties:
>>>> +#clock-cells:           Number of cells in a clock specifier; typically will be
>>>> +                set to 1
>>>
>>> I'm not sure I fully understand what the extra cells actually mean for
>>> clocks.  I think the first integer is the clock output to use but some
>>> of the versatile and highbank ones only have a phandle or is it more
>>> implementation defined?  The clock-output-names description hints at
>>> recommended, so I find this a little confusing, but that could just be
>>> me!
>>
>> I'm following convention here that has been established with
>> interrupts, gpios, and others.  Sometimes more information is needed
>> that just the clock number.  Using #clock-cells gives a clock provider
>> the option of having additional fields for clock flags or other data.
>> This is very much implementation defined.  Simple clock providers that
>> only have a single clock output can easily use #clock-cells = <0>.
>> Providers with multiple outputs will need to use 1 or more cells.
>>
> It totally destroys my understanding on #clock-cells :)
> 
> I thought it's introduced to reduce the clock nodes in dts.  That said,
> #clock-cells stands for the number of clks we describe in the node.
> When #clock-cells > 1 for a node, the node becomes a clk blob which
> actually contains multiple clks.  I migrated the imx6 clock to your
> first post with this approach [1], using 70 nodes to describe 110
> clocks (~35% nodes reduced).
> 
> Now, you are saying it's not designed for this purpose.  I'm pretty
> confused, because the only reasonable use of #clock-cells to me is
> just that way, and I fail to see why we need #clock-cells if we do
> not use it that way.
> 
> http://comments.gmane.org/gmane.linux.drivers.devicetree/9631

Think of clock-cells as the size of array elements (minus 1), not the
number of array elements. If you have more than 1 clock per node, then
you need more information on the clock input side than just a phandle to
describe which clock you are using. Typically this would just be the
phandle plus index of the clock you want. This is just like gpio where
you have controller phandle plus gpio line number.

Rob