On Fri, Nov 4, 2011 at 2:57 PM, Mark Brown
[off-list ref] wrote:

On Fri, Nov 04, 2011 at 02:47:05PM -0700, Olof Johansson wrote:

quoted

On Fri, Nov 4, 2011 at 2:25 PM, Mark Brown

quoted

quoted

I don't see how you can usefully do that, the task of plumbing a
regulator into a board is largely orthogonal to the specific feature set
of a given regulator. ?The specific bindings for a fixed voltage
regulator would be useful or unhelpful for most regultors controlled via
I2C.

quoted

I meant more that the fixed regulators should reuse as much as
possible from the generic regulator bindings, instead of completely
forking them.

That appears to be what's going on? ?The fixed voltage regulator
includes by reference the core regulator binding, all of the properties
it defines with the possible exception of the supply name are not
covered in the core binding.

The fixed-regulator binding makes no reference to the generic binding,
and the example includes no properties that are defined in that one.
So I definitely did not make that conclusion based on what I saw.

quoted

Then, depending on how they are controlled, there will be more
specific bindings. So the case of a gpio-controlled fixed regulator
would have a binding where the format of the properties to find the
gpio, etc, would be described. But things like voltage (without a
range, obviously) would be using the same bindings as the other
regulators.

The only overlap I'm seeing is the voltage?

The intended semantic for the voltage is rather different. ?The core
binding for the voltage specifies the range of voltages it is possible
to set a regulator to on a given board and is used to give permission to
the system to reconfigure the regulator. ?The binding here tells the
system what voltage a fixed voltage regulator is running at. ?We could
have the fixed voltage regulator read the same binding - though there's
some risk of mild confusion it shouldn't be too bad.

Yeah, voltage is the obvious one where fixed would have a max and min
that is the same when you have a fixed regulator.

But even things like allowing (optional) attributes such as
startup-delay on non-fixed regulators could make sense. Keep in mind
that the device tree should focus on describing the hardware, not just
what the linux driver needs from it. So maybe instead of
startup-delay, specifying ramp-up speed instead of time needed until
power is good could be the way to go there.


-Olof