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

[Shawn Guo]

On Fri, Jan 13, 2012 at 08:30:30PM -0800, Turquette, Mike wrote:
...
> I had envisioned fixed clocks as being clocks whose rates could never
> change; obviously this is mostly useful for root clocks like
> oscillators and whatnot.
> 
> There is nothing wrong with using fixed clock for sgtl5000-sys-mclk,
> but it's rate *can* change if it's parent rate ever changes.  This may
> be very unlikely on your platform, in which case again it is OK to use
> fixed clock here if you want to.
> 
> However... I'm inclined to say that sgtl5000-sys-mclk is good
> candidate for a dummy clock: it follows it's parent rate, doesn't
> gate, doesn't divide it's parent rate, only has one input.  There
> isn't a common dummy clock in the v4 patches, but there is in v5.  The
> key difference between the fixed rate clock and the dummy clock is
> that the dummy clock looks at clk->parent->rate in it's .get_rate,
> whereas a fixed rate clock will have it's rate cached in struct
> clk_fixed.
> 
> Thoughts?
> 
I would say this modeling looks all sane to me, except both the
fixed-clock soc-26m-clk and dummy-clock sgtl5000-sys-mclk in this
case have a gate which is controlled by gpio.

-- 
Regards,
Shawn