On Fri, May 01, 2020 at 03:53:09PM +0200, Alexandre Belloni wrote:
> On 01/05/2020 08:00:11-0500, Rob Herring wrote:
> > > I don't think this is true because in the case of a discrete RTC, its
> > > interrupt pin can be connected directly to a PMIC to power up a board
> > > instead of being connected to the SoC. In that case we don't have an
> > > interrupt property but the RTC is still a wakeup source. This is the
> > > usual use case for wakeup-source in the RTC subsystem. Else, if there is
> > > an interrupt, then we assume the RTC is a wakeup source and there is no
> > > need to have the wakeup-source property.
> > 
> > Yes, that would be an example of "unless the wakeup mechanism is
> > somehow not an interrupt". I guess I should add not an interrupt from
> > the perspective of the OS.
> > 
> > So if the wakeup is self contained within the PMIC, why do we need a
> > DT property? The capability is always there and enabling/disabling
> > wakeup from it is userspace policy.
> > 
> 
> Yes, for this particular case, I'm not sure wakeup-source is actually
> necessary. If the interrupt line is used to wakeup the SoC, then the
> presence of the interrupts property is enough to enable wakeup.

So yes, the wakeup-source property isn't necessary. The goal of patches
1 and 2 was to allow the RTC to be actually disabled as a wakeup-source
in case it didn't work as intended. But since the RTC is enabled as a
wakeup source on these PMICs by default, the idea was to add a new sub-
node for the RTC and required the wakeup-source in that subnode if that
subnode was present.

That said, patch 3 actually does make the RTC work as a wakeup source
on the particular board that I tested this, so patches 1 and 2 are no
longer really required from my point of view.

Do you want me to send patch 3/3 again separately or can you pick it up
from this series?

Thanks,
Thierry