Mainline Android kernel functionality
- Mainline Android kernel functionality
- Tim Bird
Android has some unique functionality that is not currently mainlined, e.g. ashmem, logger, binder and so forth.
Mainlining these aspects of Android would reduce the overall maintenance and integration cost for working with Android.
The problem of wakelocks (/suspend blockers, etc) have been discussed and debated extensively. Other specific features that Android adds to the kernel are less controversial, and should be added incrementally over time.
A big issue is whether the Android user-space code would be modified to match the kernel code as modified to be suitable for mainlining.
- Android Kernel Features
- Linaro kernel-android-upstreaming project
- Linaro seems to have a project in-progress in this area, with some involvement by Paul McKenney
- https://lkml.org/lkml/2010/11/6/154 - e-mail by Ted Ts'o about Android out-of-tree patches
- https://lkml.org/lkml/2010/11/6/177 - response by Greg KH
- https://lkml.org/lkml/2010/11/10/433 - note by Rafael Wysocki that wakeup sources may solve some problems
- Presentation by Greg KH on Android patches
I would recommend starting with something simple (like logger or ashmem) and working from there. The amount of time to mainline stuff is notoriously difficult to estimate, but possibly 4 weeks per function, for simple things.