Email log

From:   John Stultz <john.stultz@linaro.org>
Subject:        Android-Kernel-Subteam weekly email meeting - 1/18/12

[Administrative stuff]========================
* Nothing much from me here.

* I went through my blueprints and tried to specify 12.1 vs 12.2+ work
items. Mounir, Deepak,  anything I need to tweak/adjust here?


[Lowmemory killer]===============================
https://blueprints.launchpad.net/linux-linaro/+spec/android-lowmem-upstreaming

* Anton, Still need a little more granularity on work items.

* Anton, Also please update blueprint with related work done, like
sending the missing Android lowmemory killer patch to Greg, as well as
other planned work like file cache issues, and tasklist_lock changes.


[ashmem]==========================================
https://blueprints.launchpad.net/linux-linaro/+spec/android-ashmem-upstreaming

* At Andrew Morton's prodding, spending a good chunk of my time this
week working on a more efficient tree based storage method for ranges as
opposed to a simple list. Ends up its a non-trivial design. I've managed
to get simple tree based insert and deletion, but the tree also needs to
be self-balancing, so I need to add that. Considering using splay-like
functionality.

* In discussions with Andrew, he's also really strongly pushing for
whatever storage structure ends up being used, we have to be able to use
it for posix/bsd file locks, which have similar range properties. While
they are conceptually similar, I'm not 100% sure the ashmem ranges and
file locks are totally equivalent semantic wise, so I'm reading up more
to better understand the details. I suspect reworking the file locking
code to use whatever efficient structure I come up with is going to end
up being a prerequisite for merging ashmem work, which is frustrating. 



[monotonic evdev]=================================
https://blueprints.launchpad.net/linux-linaro/+spec/android-evdev-upstreaming

* Patch has been merged in maintainer queue for 3.4. So this can likely
be backgrounded until 3.4 opens and Android moves over to using 3.4
(prob 4-6 months out), when we'll need to convert Android userland to
using the new ioctl.


[alarmtimer mending]==============================
https://blueprints.launchpad.net/linux-linaro/+spec/android-alarmtimer-mending

* I haven't had time to get to this, this week.


[Misc]===============================================

* My config_merge.sh script was merged into 3.3-rc, so I'll probably
have some additional work to help move the Linaro kernel and build
systems over to using it for 12.2 or whatever release picks up 3.3-rc
kernels.

* Andrey is overloaded at the moment, so I'm going to help him out and
release the 12.1 Linaro Android kernel much as I did for 11.12. This
will chew up a day or so this week.

* Anton: You going to be attending Linaro Connect?

* Anton: Any issues on your side? 

From:   john stultz <johnstul@us.ibm.com>
Subject:        Re: Android-Kernel-Subteam weekly email meeting - 1/18/12

On Tue, 2012-01-17 at 18:18 -0800, John Stultz wrote:
> [Administrative stuff]========================
> * Nothing much from me here.
> 
> * I went through my blueprints and tried to specify 12.1 vs 12.2+ work
> items. Mounir, Deepak,  anything I need to tweak/adjust here?
> 
> 
> [Lowmemory killer]===============================
> https://blueprints.launchpad.net/linux-linaro/+spec/android-lowmem-upstreaming
> 
> * Anton, Still need a little more granularity on work items.
> 
> * Anton, Also please update blueprint with related work done, like
> sending the missing Android lowmemory killer patch to Greg, as well as
> other planned work like file cache issues, and tasklist_lock changes.

Anton, also just in case you missed it, there's a patch on lkml today
for /dev/low_mem_notify that you might want to check out and comment on.

http://lkml.indiana.edu/hypermail/linux/kernel/1201.2/00569.html

thanks
-john

From:   Anton Vorontsov <anton.vorontsov@linaro.org>
Subject:        Re: Android-Kernel-Subteam weekly email meeting - 1/18/12

On Tue, Jan 17, 2012 at 06:56:22PM -0800, john stultz wrote:
> On Tue, 2012-01-17 at 18:18 -0800, John Stultz wrote:
> > [Administrative stuff]========================
> > * Nothing much from me here.
> > 
> > * I went through my blueprints and tried to specify 12.1 vs 12.2+ work
> > items. Mounir, Deepak,  anything I need to tweak/adjust here?
> > 
> > 
> > [Lowmemory killer]===============================
> > https://blueprints.launchpad.net/linux-linaro/+spec/android-lowmem-upstreaming
> > 
> > * Anton, Still need a little more granularity on work items.
> > 
> > * Anton, Also please update blueprint with related work done, like
> > sending the missing Android lowmemory killer patch to Greg, as well as
> > other planned work like file cache issues, and tasklist_lock changes.
> 
> Anton, also just in case you missed it, there's a patch on lkml today
> for /dev/low_mem_notify that you might want to check out and comment on.
> 
> http://lkml.indiana.edu/hypermail/linux/kernel/1201.2/00569.html

The good thing I did not start implementing this myself, and
instead focused on the Android's ActivityManager to wrap my head over
needed requirements and to try to implement a working "Userland
Lowmemory Killer" proof of concept (using existing cgroups memory
notifiers, but keeping in mind that we might use other notifier API
some time later).

The bad thing is that the current driver is overly simple, but I
see some discussion on extending it.

So, all is good: these efforts are not going to interfere w/ us
(for now), but I surely will say some words wrt this driver,
at least to show that we're interested in this feature.

From:   john stultz <johnstul@us.ibm.com>
Subject:        Re: Android-Kernel-Subteam weekly email meeting - 1/18/12

On Wed, 2012-01-18 at 21:50 +0400, Anton Vorontsov wrote:
> On Tue, Jan 17, 2012 at 06:56:22PM -0800, john stultz wrote:
> > On Tue, 2012-01-17 at 18:18 -0800, John Stultz wrote:
> > > [Administrative stuff]========================
> > > * Nothing much from me here.
> > > 
> > > * I went through my blueprints and tried to specify 12.1 vs 12.2+ work
> > > items. Mounir, Deepak,  anything I need to tweak/adjust here?
> > > 
> > > 
> > > [Lowmemory killer]===============================
> > > https://blueprints.launchpad.net/linux-linaro/+spec/android-lowmem-upstreaming
> > > 
> > > * Anton, Still need a little more granularity on work items.
> > > 
> > > * Anton, Also please update blueprint with related work done, like
> > > sending the missing Android lowmemory killer patch to Greg, as well as
> > > other planned work like file cache issues, and tasklist_lock changes.
> > 
> > Anton, also just in case you missed it, there's a patch on lkml today
> > for /dev/low_mem_notify that you might want to check out and comment on.
> > 
> > http://lkml.indiana.edu/hypermail/linux/kernel/1201.2/00569.html
> 
> The good thing I did not start implementing this myself, and
> instead focused on the Android's ActivityManager to wrap my head over
> needed requirements and to try to implement a working "Userland
> Lowmemory Killer" proof of concept (using existing cgroups memory
> notifiers, but keeping in mind that we might use other notifier API
> some time later).
> 
> The bad thing is that the current driver is overly simple, but I
> see some discussion on extending it.
> 
> So, all is good: these efforts are not going to interfere w/ us
> (for now), but I surely will say some words wrt this driver,
> at least to show that we're interested in this feature.

Also saw this related item today:
https://lwn.net/Articles/475793/

thanks
-john

WorkingGroups/KernelArchived/AndroidUpstreamMeetings/2012-01-18 (last modified 2013-01-14 19:34:09)