Track lead: Matthew Garrett mjg@redhat.com

We're seeing increasing interest in development of aggressive runtime power management. This throws up a couple of problems:

• We need to be able to facilitate that aggressive runtime PM in the first place
• We need to understand how userspace can communicate its requirements to the kernel

Providing support for runtime PM

Rafael's PM framework provides most of the support required for this, and we're seeing it applied to various bus types. However, we don't currently have a solution for more general system-wide power states. OMAP uses cpuidle to enter deep power states when idle, but this doesn't scale to multi-core systems which are starting to appear in the embedded market. The main continuing point of disagreement is over how resilient the PM implementation should be to applications which don't idle when unused.

Communicating userspace requirements

pm_qos is now a reasonably well-established interface for expressing certain kinds of userspace requirements for PM transitions. Is it suitable for all purposes?

There's a certain amount of crossover between this and the PM minisummit in August. We'll need to try to work out which topics are most relevant to each.

Key people

Rafael Wysocki (runtime PM framework) Kevin Hilman (would be good for OMAP side of things) Mark Gross (pm_QoS) Someone from Android