Re: [patch 2.6.28-rc3] regulator: add REGULATOR_MODE_OFF

[David Brownell <david-b@pacbell.net>]

On Sunday 16 November 2008, Mark Brown wrote:
> > Are you arguing that there should be some new regulator_ops call to
> > expose the actual regulator state?  If so, then what should happen to
> 
> Yes, exactly - possibly multiple calls.

OK, so I finally got back to this issue.  I'll post a patch
with such a separate call in a moment ... just a single status
value, sufficient for the need I observe, and only exposed
through sysfs since it's more useful just now as a kind of
introspection.  Example:  troubleshooting, as we previously
discussed; or, I can see some regulators that are wrongly
enabled, primarily because of bootloader goofage.

That raises an issue:  how can Linux get such regulators to
turn off?  Clock frameworks have the same issue, and they
tend to resolve this with a SoC-specific Kconfig option to
disable unused clocks (in a late_initcall, after everthing
has had a chance to start up).  That conserves power.

I'm thinking it'd be worth having a similar Kconfig option
for regulators too.  It could kick in regulator core code to
call regulator_ops.disable() whenever regulator_dev.use_count
is zero, possibly warning if !regulator_ops.is_disabled().
(Handling constraints.always_on too.)

Comments?

- Dave