Hi Everyone,
There are two issues resulting from asynchronous suspend of thermal zones.
One of them is platform-specific and related to some firmware issue (I think)
causing battery readings to become invalid after a system resume due to
interference between ACPI battery resume and a thermal zone update. This
can be addressed by running the thermal PM notifier after all of the other
PM notifiers (including the ACPI battery one) which is done by patch [2/2].
The other one is mostly theoretical, but I couldn't convince myself that it
cannot happen. Namely, a leftover thermal zone resume running during the
next system suspend (if it is carried out back-to-back with the previous
suspend-resume cycle) can accidentally reset tz->suspended set for a thermal
zone by the thermal pre-suspend PM notifier. This is addressed by patch
[1/2] (which goes before the second one because the latter increases the
likelihood of the issue slightly).
Thanks!