LCU13: Android upstreaming - status and future
2 likes3,721 views
The document outlines the progress and future plans for Android upstreaming as of October 2013, detailing accomplishments like merged features and ongoing discussions among community developers. Key topics include improvements to the Android sync driver, configfs gadget integration, and netfilter extensions, with next steps focusing on collaboration and avoiding fragmentation of frameworks. It emphasizes the need for additional resources and resolving conflicts in ongoing projects, particularly in ARM changes and the Android gadget driver.
LCU13: Android upstreaming - status and future
- 1. Android Upstreaming Status and Future John Stultz, Oct 29th 2013 john.stultz@linaro.org
- 2. www.linaro.org ● Accomplishments so far this year ● Discussions from Linux Plumbers ● What is still to be done ● Next steps / Discussion Overview
- 3. www.linaro.org ● Mempressure events merged in 3.10 (AntonV) – Android developers planning to move from low-memory-killer in a future release. ● Android Sync driver in staging in 3.10 (John) ● FAT_GET_VOLUME_ID ioctl merged in 3.11 (Bintian) ● Suspend time logging merged in 3.11 (Zoran) ● Keyreset functionality merged into sysrq in 3.12 (Mathieu) ● compat_ioctl enablement for ashmem, alarmdev, sync, FAT_GET_VOLUME_ID and others merged during 3.10-3.12 (DmitryP, John, Serban, Bintian) – Binder support hopefully soon ● pstore suspend debugging queued for 3.13 (Zoran) ● wakeup_source enablement for RTC and power_supply drivers queued for 3.13 (Zoran) – Hopefully MMC as well (Ulf) Linaro Accomplishments so far in 2013
- 4. www.linaro.org ● ConfigFS gadget merged in 3.10 (Sebastian/Andrzej) – Will likely replace Android Gadget eventually ● Alarmtimer support added to timerfd merged in 3.11 (Todd) – Allowed android-battery meta-driver to move to user land – Likely to replace /dev/alarm ● Freezer optimizations merged in 3.11 (Colin) – Merged upstream in parallel with AOSP! Community Accomplishments so far in 2013
- 5. www.linaro.org ● Android Sync ● Atomic Display Framework ● ION ● Binder ● Netfilter ● Android Gadget Discussion from Linux Plumbers
- 6. www.linaro.org ● Dmabuf-fences is the upstream preferred solution ● Implicit vs explicit sync contracts ● Seemingly agreement on allowing for optional explicit sync arguments in the APIs – Likely trying to implement Android's Sync api on top of kernel-internal dmabuf-fences ● Next steps: – Maarten started looking at this, but may need assistance Android Sync
- 7. www.linaro.org ● Simplified KMS-like framework. ● Sort of an experiment to see what an ideal framework for Android would look like ● Similar to upstream atomic mode-setting / nuclear pageflip work ● Integrates Android sync points in the API ● Presented as fodder for how KMS could be extended ● Next steps: – Continue pushing community and Android developers to collaborate on this. – Hopefully avoid forking display frameworks! Atomic Display Framework
- 8. www.linaro.org ● Discussed ION interface issues ● Upstream preferred solution is post-attach delayed allocation – Android developers have issues with this approach ● Likely to continue having separate interfaces, but hopefully will share underlying logic. ● Next steps: – Push ION into staging – Try to integrate ION's “heap” allocators in the dmabuf post-attach delayed allocation approach. ION
- 9. www.linaro.org ● Greg KH earlier claimed he would get libbinder running over KDBUS, allowing in-kernel binder driver to be dropped ● This is apparently quite hard, as KDBUS has very different semantics then binder. ● Not looking very likely a way forward, but Greg wants us to check back in a year. ● Next steps: – None, at the moment Binder
- 10. www.linaro.org ● Got background on Android netfilter extensions: – xt_quota2, xt_idletimer and xt_qtaguid ● Upstream developers suggest using NFQUEUE instead of xt_qtaguid ● Next steps: – xt_quota2 likely to be done w/ nfacct – xt_idletimer changes look possible to merge – NFQUEUE viability research on-going Netfilter
- 11. www.linaro.org ● ConfigFS driver merged in 3.10 likely to eventually replace Android Gadget ● ConfigFS still missing some functionality: – Needs FunctionFS support – Also upstream needs MTP, Android Accessory and Android Audio function support ● Next steps: – Find hands to work on this Android Gadget
- 12. www.linaro.org ● MMC improvements (Ulf) ● ION moving to staging & making it arch neutral (Jesse, Sumit, John) ● Netfilter changes (Mathieu) ● Volatile Ranges work to upstream part of ashmem (John) ● Binder compat_ioctl (Serban) Current work in progress
- 13. www.linaro.org ● Lots of graphics work – ION, Sync, ADF – Want to avoid a forked graphics infrastructure ● ETM/ETB ● FIQ Debugger functionality in KDB (stalled) ● ARM improvements (ro kernel text, console flush, etc) ● Android Gadget driver – Enable FunctionFS in ConfigFS gadget – Support MTP, Android Accessory, Android Audio gadget functions ● Interactive cpufreq gov – Hopefully replaced by power-aware scheduling ● Bluetooth, wireless, other networking changes ● Moving items out of staging (binder, ashmem, logger, sync, etc) What is left?
- 15. net/netfilter/xt_qtaguid.c | 3012 ++++++++++++++++++++++++++ drivers/usb/gadget/android.c | 1517 +++++++++++++ drivers/gpu/ion/ion.c | 1495 ++++++++++++ arch/arm/common/fiq_debugger.c | 1376 +++++++++++ drivers/usb/gadget/f_mtp.c | 1274 ++++++++++ drivers/cpufreq/cpufreq_interactive.c | 1247 ++++++++++ drivers/usb/gadget/f_accessory.c | 1180 ++++++++++ drivers/usb/gadget/f_audio_source.c | 828 +++++++ arch/arm/kernel/etm.c | 680 ++++- net/netfilter/xt_qtaguid_print.c | 566 ++++ drivers/gpu/ion/ion_system_heap.c | 450 +++ drivers/net/ppp/pppolac.c | 449 +++ drivers/input/misc/gpio_matrix.c | 441 +++ drivers/net/ppp/pppopns.c | 428 +++ drivers/input/misc/keychord.c | 391 +++ drivers/input/misc/gpio_input.c | 390 +++ net/netfilter/xt_quota2.c | 385 +++ include/linux/ion.h | 376 +++ drivers/gpu/ion/ion_priv.h | 359 +++ net/netfilter/xt_qtaguid_internal.h | 352 +++ drivers/char/dcc_tty.c | 326 ++ drivers/gpu/ion/ion_heap.c | 320 ++ drivers/input/keyreset.c | 239 ++ drivers/input/misc/gpio_event.c | 228 + drivers/gpu/ion/ion_chunk_heap.c | 213 + drivers/gpu/ion/ion_cma_heap.c | 205 + drivers/gpu/ion/ion_page_pool.c | 200 + drivers/input/misc/gpio_axis.c | 192 + drivers/gpu/ion/ion_carveout_heap.c | 187 + kernel/trace/trace_output.c | 182 + drivers/switch/switch_class.c | 174 + drivers/usb/phy/otg-wakelock.c | 173 + drivers/switch/switch_gpio.c | 172 + include/linux/freezer.h | 171 + include/linux/gpio_event.h | 170 + arch/arm/mm/rodata.c | 159 + ... Key: Upstreaming in progress Upstream since 3.10 diffstat of android-3.10 sorted by delta
- 16. www.linaro.org ● Resource issues – Plenty of work, not too many full time hands – Need to leverage the graphics team ● Next areas of focus? – ARM changes ● Tend to collide and cause frequent pain ● Would be good to resolve – Android Gadget (low hanging fruit) – FIQ Debugger (currently stalled) – ???? Discussion
- 17. More about Linaro: http://www.linaro.org/about/ More about Linaro engineering: http://www.linaro.org/engineering/ How to join: http://www.linaro.org/about/how-to-join Linaro members: www.linaro.org/members connect.linaro.org