On Wed, Feb 19, 2014 at 03:19:00PM +0100, Krzysztof Kozlowski wrote: > As I understand the suspend mode (correct me if I'm wrong), the > regulator core during suspend to mem: > 1. Calls suspend_set_state(). > 2. rstate->disabled is true so the ops->set_suspend_disable() is called. > 3. The ops->set_suspend_disable() function (implemented by the driver) > disables the regulator (e.g. through I2C commands /regmap/). > 4. During resume the regulator is enabled normal way (ops->enable(), I2C > again). No, set_suspend_disable() should *not* disable the regulator, it should configure what the regulator will do when the hardware enters suspend mode. If we were just disabling the regulator there would be no point in having a special operation, we could just use the normal operation. > Possible problems: > A. What happens if some driver using this regulator resumes earlier then > regulator_suspend_finish()? Nothing, if the system is not in suspend mode then configuring suspend mode will have no impact. > B. What happens if resuming regulator requires some other driver to be > resumed earlier (e.g. I2C bus)? If regulator resumes before I2C bus then > calling ops->enable() would fail. It is expected that as part of exiting suspend mode the hardware will revert to normal operational mode.
Attachment:
signature.asc
Description: Digital signature
