On Wed, Oct 19, 2011 at 01:33:55PM +0800, Shawn Guo wrote:

On Tue, Oct 18, 2011 at 05:00:46PM +0100, Mark Brown wrote:

quoted

It's not just Linux-specific stuff, some of this is even specific to
what current Linux drivers can do - updating the kernel could mean a
different set of constraints.

Well, from what I see, the 'struct regulation_constraints' is defined
in machine.h and meant to be the regulator machine/board interface.

...which will depend on the system integrator's understanding of what
their system is capable of right now.

With the example I'm looking at, mc13892, the regulation_constraints
configuration is fully passed from machine/board file.  If there is
something specific to what drivers can do, it probably should be encoded
in regulator driver rather than staying in regulation_constraints.

I don't think you're quite understanding the issue - it's an integration
problem with three different variables.  It's a combination of what the
chips can do, what the drivers can do and if the board design affects
any of this stuff.  Only the board can come to a final decision.