Hello Jouni, Rajendra,
On Wed, 13 Aug 2008, Högander Jouni wrote:
> "ext Rajendra Nayak" <rnayak@xxxxxx> writes:
> > Its needed so that a power domain is turned ON while the first clock node in it is
> > requested, and turned OFF while the last clock node is disabled.
>
> Isn't this already taken care by clock framework? If not I think it
> should be. I mean directly requesting domain to wakeup rather than
> requesting resource which causes it to wakeup.
>
> Paul, Any comment on this?
On the clock enable path, the clock code will wake up the clockdomain (or
add a wakeup dependency from MPU), which should cause the powerdomain to
automatically wake up if it is currently in retention or off.
Similarly, on the clock disable path, the clock code puts the clockdomain
to sleep (removes the MPU/IVA2 wakeup dependencies), which should cause
the powerdomain to enter whatever state was previously configured in
PM_PWSTCTRL_x.POWERSTATE. POWERSTATE would have been set earlier by
device drivers calling into the OMAP PM code ("set_max_wakeup_lat" etc.)
If all of that is correct, then as you wrote, Jouni, there shouldn't be a
need for omap_pm_pwrdm_{in,}active() in Rajendra's current patch set. In
fact the current code might actually prevent the powerdomain from dropping
below ON state.
Rajendra, can you confirm this? Or am I missing something in the above?
- Paul
