Re: [RFC PATCH 3/6] PM / Voltagedomain: introduce voltage domain driver support

[Nishanth Menon <nm@ti.com>]

On 03/10/2014 12:22 PM, Mark Brown wrote:
> On Mon, Mar 10, 2014 at 12:11:44PM -0500, Nishanth Menon wrote:
>> On 03/02/2014 09:54 PM, Mark Brown wrote:
>>> On Mon, Feb 24, 2014 at 08:38:07AM -0600, Nishanth Menon wrote:
> 
>>>> Intent here is to allow drivers such as cpufreq-cpu0 to be reused on
>>>> platforms such as TI's OMAP derivatives, and other SoCs which differ
>>>> only by the sequence involved in voltage scale operations. So, this
>>>> patch provides a framework for registering the underlying
>>>> implementation of the SoC specific voltage change methodology.
> 
>>> That bit is clear, what's very opaque from the code is how this is going
>>> to be accomplished.
> 
>> The SoC specific voltage domain drivers register with
>> devm_voltdm_register. the fops provide the abstraction needed for the
>> SoC (example in patch #5 - which introduces OMAP specific voltage
>> domain which handles ABB and VDD regulators).
> 
>> What would you suggest that we do to clarify the usage here?
> 
> Probably saying something about this in the commit message would be
> enough - mentioning how the registration occurs and that things are
> triggered by clock frequency changes.

OK. will do, thanks for suggesting the same.

>>> So the first question I have here is what happens if multiple clocks
>>> need to be updated in lock step - if we're only triggering off clock
>>> notifiers that seems tricky.  The other thing here is that the fact that
> 
>> Yes, that is true, however, there are ways to implement them, for
>> example: We could implement an higher level clock that takes care of
>> the multiple clock node control to handle this kind of scenario.
> 
> That seems concerning given the fact that people seem to like describing
> their entire clock trees in DT, we shouldn't be putting implementation
> stuff there.

The only other options are:
a) Abstract it at a higher level at "user drivers", since they are
aware of the sequencing needs - but this partially defeats the
purpose, unless ofcourse, we do a tricky implementation such as:
clk a, b, c -> prenotifiers in a, postnotifiers in c (which as you
mentioned is a little trickier to get right).
b) introduce a higher level generic dvfs function[1] which does not
seem very attractive either.


Any other suggestions other than limiting the usage(and documenting it
so) and hoping for a future evolution to take this into consideration?

[1] http://marc.info/?t=139275581900004&r=1&w=2

-- 
Regards,
Nishanth Menon