Proof Of Concept results of Wake Lock replacement experiments for Android 2.2
This proposal has been rejected.
One Line Summary
A report out of experiences from hacking the Android PM model to use PM wakeup event's instead of kernel wake locks.
Ever since the collaboration summit in April 2010 there has been a lot of traffic on the linux-pm list about wake locks, suspend blockers, wake notifications, pm-qos based wake locks (which resulted in plist-based pm-qos and is a nice enhancement to the pm-qos implementation) … its never completely ended.
The current patches in response to the wake locks that have potential to solve some of the wake lock crisis is Rafael’s pm_wakeup_event patch. http://lkml.org/lkml/2010/6/19/200 However; this solution requires a tweak to user mode. Someone needs to give it a good try within a credible google-Android implementation to see where the gaps are and compare how it well it works WRT the original Android PM design.
So, as a side project I’m hacking together an android stack and kernel to use the pm wakeup event mechanism to find out where it falls over or is incomplete. (and perhaps fix things up a bit.)
My day job includes helping to make standard Android PM work as well as possible on UMG hardware. I hope to be able to do a good job at comparing this POC against the standard android PM.
This is a work in progress but I can commit to having results and something to say about it by late October.
(draft – will update later)
Mark Gross works in the Intel Ultra Mobile Group at Intel and is focused on enabling Android on Intel based handset and tablet platforms. (Moorestown at the moment)
Mark is also active in power linux kernel power management, mostly through PM-QoS, and has been participating in the Linux kernel for around 10 years.