* Sven Peter [off-list ref] [210530 11:09]:

Hi,

On Wed, May 26, 2021, at 09:18, Tony Lindgren wrote:

quoted

Hi,

* Rob Herring [off-list ref] [210525 18:09]:

quoted

I would do a single node per mmio region with the register offset (or
offset / 4) being the clock id. This can still support new SoCs easily
if you have a fallback compatible. If you want/need to get all the
clocks, just walk the DT 'clocks' properties and extract all the IDs.

I mostly agree.. Except I'd also leave out the artificial clock ID and
just use real register offsets from the clock controller base instead.

Sure, I'll do that.

quoted

So a single clock controller node for each MMIO range, then set
#clock=cells = <1>. Then the binding follows what we have for the
interrupts-extended binding for example.

If the clock controller optionally needs some data in the dts,
that can be added to the clock controller node. Or it can be driver
internal built-in data. If the data for dts can be described in a
generic way, even better :)

Now the big question is *how* to describe this additional data in the
dts. Essentially I need to specify that e.g. to enable clock 0x270
I first need to enable the (internal) clocks 0x1c0 and then 0x220.
Are you aware of any generic way to describe this? I'm not even sure
how a sane non-generic way would look like when I just have a single
clock controller node.

To me it seems you might be able to recycle the assigned-clocks and
assigned-clock-parents etc properties in the clock controller node.

Sure the assigned-clocks property will point to clocks in the
clock controller itself, and will have tens of entries, but should
work :)

And sounds like you can generate that list with some script from the
Apple dtb.

Regards,

Tony

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel