Dnia czwartek, 29 lipca 2010 o 13:57:21 Amit Arora napisał(a):
> I have tested this only on my desktop. I am yet to receive an ARM based
> board, and hence couldn't test it there. So, any kind of testing and/or
> feedback are most welcome! Thanks!
I fetched 9f4efd1c7cc0bc9a9078fa6b83a98763a7637e2f revision, applied both
patches and built it for my desktop. Segfaulted on first run.
Desktop is Core2Quad q6600, powertop without your changes works. C-states are
not available, frequencies are listed.
Regards,
--
JID: hrw(a)jabber.org
Website: http://marcin.juszkiewicz.com.pl/
LinkedIn: http://www.linkedin.com/in/marcinjuszkiewicz
Tim,
Based on your suggestions last week I have prepared a new linux-linaro
tree. The source package name is now linux-linaro. The debian.linaro
changelog only has high level changes not all the git commits. I
removed all the cruft from the debian.linaro subtree. The url is:
git://git.linaro.org/jcrigby/linux-linaro.git
The linaro and master branches are currently identical. The latest
tag is Linaro-2.6.35-1000.2. A source package based on this is in a
ppa building at this very moment.
Thanks
John
The Agenda and the Minutes for this meeting can also be found at
https://wiki.linaro.org/WorkingGroups/PowerManagement/Meetings/2010-07-28
Attendees : Amit Arora, Amit Kucheria, Bobby Bhatacharia, Srinivas
Kalaga, Vishwanath Sripathy
Minutes:
* Review action items from last meeting
o ACTION: Bobby to investigate android contraints (e.g. display keep-alive)
Will target it for the next week.
o ACTION: Srinivas to provide details of where he believes
userspace - kernel interaction is required.
pm_qos is low prio .. so this may take a few weeks for this action
This may take a few weeks for this action. Currently at low
priority (since pm_qos work is of low prio)
o ACTION: Srinivas to add ftrace item to task list and send patches to linaro
Srinivas has been working on this in last few days. We should see
something soon on this.
o ACTION: Amit K. to inform ARM about multi-core availability
We are not getting any multi-core boards. Vishwa should have one.
Now, the question is - will one board be enough ?
Bobby to see possibility of getting more boards for Linaro. Not
easy, but will consider the request open.
o ACTION: ARM to discuss giving out powertop tool
ARM doesn't have a PowerTOP tool, so to speak, to give out.
Though they have an Eclipse based tool, which is very much an ongoing
work. Can work on making it available. But, no ETA as of now.
o ACTION: TI to elaborate on requirement of cpuidle instrumentation task
Vishwa : The currently "known" c-states sleep time and latency
information on OMAP boards are NOT optimal. There is a requirement for
some kind of instrumentation to measure the s/w and the h/w latencies.
Later is difficult part. And once we can get the optimal latencies, we
will have to work on optimizing them.
NEWACTION : Vishwa to create a blue print for this. It should
mention the codepaths to be measured (eg. wakeup latencies, idle loop
to idle loop latencies for various c-states) and other details.
o ACTION: ARM to share internal instrumentation flow
(BAB: we might also align with Linaro on workload discussions)
Bobby to make this available. But, it might take a couple of months.
* HW availability
o Amit Arora is following up to expedite the process to acquire the
development boards. May take some more time.
o Vishwa has access to couple of OMAP3/4 SDP boards. Amit Kucheria
mentioned that it may be a good idea to get the publically available
boards, like the Beagleboards or the PANDA blades. Vishwa to find out
about this.
o Amit Kucheria has a Freescale i.mx51 babbage board and a beagleboard
* Git trees on git.linaro.org
o We will have to host a couple of git trees in git.linaro.org
o To start with, we may need two trees for the tools - for powertop
and powerdebug (a new tool).
o Need access to the host system and some process needs to be followed.
o Once we start hosting the trees, we will have to send a mail to
linaro-dev mailing list, requesting everyone to try out the code.
o When we have patches for kernel, we may have to host a separate
kernel git tree for Power Management, which should get merged with the
linaro kernel git tree.
* gobby / ubuntu
o Amit Kucheria recommended everyone to start using this tool,
which is a cross-platform collaboration tool and will make things like
taking minutes much easier (everyone will be able to contribute to it)
* Blueprint status
o Tools blueprint : Amit Arora gave the update on the work he has
been doing on powertop and powerdebug tools. He has a few patches
ready, which he plans to host in a git tree. For the time being, we
plan to host this code as tarballs at some location, which will help
others to test it.
o Master blueprint : Amit Kucheria had split the master blue print
into "infrastructure" and "tools" blueprints. More such blueprints
will need to be created as we go forward.
o There are some known issues with Ondemand governor, which Vishwa
and AmitK will take it to the the community (LKML/linux-pm and make
sure to copy right people). Will need to share the constraints and ask
for feedback on how its best to implement some of the things in the
current framework.
* Others
o Linaro's first released will be based on 2.6.35. Hence, all our
work also should be inline to this release.
ACTION ITEMS:
Carried Forward from past :
-------------------------
* CLOSED :
o ACTION: TI to elaborate on requirement of cpuidle instrumentation task
This has given birth to a new ACTION item, discussed below.
* ACTIVE :
o ACTION: Bobby to investigate android contraints (e.g. display keep-alive)
o ACTION: Srinivas to provide details of where he believes
userspace - kernel interaction is required.
o ACTION: Srinivas to add ftrace item to task list and send patches to linaro
o ACTION: Amit K. to inform ARM about multi-core availability
o ACTION: ARM to discuss giving out powertop tool
* DORMANT :
o ACTION: ARM to share internal instrumentation flow
(BAB: we might also align with Linaro on workload discussions)
Might take couple of months.
New Items:
---------
o ACTION: Vishwa to create a new blue print for the instrumentation
of cpuidle related work
o ACTION: Amit Arora to create tarballs of his current work and
share it internally for testing
o ACTION: Amit Kucheria and Vishwa to get inputs from community on
the issues related to Ondemand governor.
Thanks!
Amit Arora
Hi,
Starting tomorrow (28.07.2010, Wednesday), the PM WG will start weekly
calls to discuss status of the work items.
Meeting minutes will be posted at
https://wiki.linaro.org/WorkingGroups/PowerManagement
Regards,
Amit
The weekly report for the Linaro Infrastructure team may be found at:-
https://wiki.linaro.org/Releases/InfrastructureStatus
Summary:
Most of the team were in Prague for the Sprint.
There is a concern that the changes to LexBuilder are behind schedule.
The first licence request was submitted to the TSC.
Progress:
A lot of tasks remain outstanding for the 10.10 beta. This is being
investigated to establish of this is realistic. An update will be
provided in the next report.
We're a bit lacking in hardware at the moment. I've gone through my
setup, given it a bit of a polish, and made the following available to
the outside world:
* pavo1 - a BeagelBoard clone with a hard disk, suitable for running long jobs
* pavo2 - a BragleBoard clone
* orion - an Atom N450 based netbook
* scorpius - an AMD Phenom based machine, useful as a distcc host
Please contact me if you'd like access. Both pavo1 and orion are
suitable for running long jobs on. More details are available at
https://wiki.linaro.org/WorkingGroups/ToolChain/Hardware.
These will shift into the data centre in the future but it's nice to
have them available now.
-- Michael
Hey
Please find below the minutes from yesterday's Kernel Consolidation
Working Group meeting and the activity reports from previous week.
You can find the full meeting notes at
https://wiki.linaro.org/WorkingGroups/KernelConsolidation/Meetings/2010-07-…
== Agenda ==
* Action items from last meeting.
* mwaddel to code review the arm_next.git tree and check that all
patches are present.
* [DONE] mwaddel will try the kernel on the versatile board.
* jk to report on progress of talking to Grant WRT upstreaming his patches.
* [CARRY OVER] no response from Grant yet.
* mwaddel to talk to ARM about device-tree work.
* [DONE]
* Jeremy to chat with Nicolas about documenting boot interface.
* [DONE] some patches already produced.
* npitre to find a way to expose the list of trees merged in
arm_next.git via a wiki page.
* [DONE] documented in the tag instead.
* jcrigby to talk to slangasek about the list of kernel flavours to
produce.
* [CARRY OVER] jcrigby to talk face-to-face at the sprint.
* amitk to report back the status on linux-omap's patches.
* [CARRY OVER] amitk absent.
* jcrigby to determine if versatile is still needed.
* [CARRY OVER] jcrigby to talk face-to-face at the sprint.
* lool to check that -fPIE support is currently enabled in the Linaro
toolchain.
* [DONE]
http://lists.linaro.org/pipermail/linaro-dev/2010-July/000150.html
* npitre and arnd to review ericm's single zImage patches.
* [DONE] ongoing and doesn't need an action item.
* ericm to send single zImage patches to russelking after they have
been reviewed.
* [DONE] patches are going to the public mailing lists.
* jcrigby and lool to arrange a uboot specific call for interested
team members.
* [CARRY OVER] Arrange at the sprint -
http://lists.linaro.org/pipermail/linaro-dev/2010-July/000149.html
* Blueprint status.
arm-m-using-device-tree-on-arm
arm-m-kernel-version-alignment
arm-m-missing-security-features
arm-m-debugging-with-oprofile
spec/kernel-maverick-arm-single-zimage
arm-m-versatile-express-a9-enablement
* Gathering ideas for face to face discussions / work items for the sprint
* Personal work plan for the sprint
* Plenaries
* Meeting canceled for week of sprint
=== Action Points ===
* jk to report back on Grant Likely’s response to upstreaming patches
* ericm to discuss with lool how to get more hardware.
=== Minutes ===
* lool on vacation today, short status call.
daverusling asked about the status of the Ubuntu kernel patches in
the Linaro tree after recent discussion on the mailing list. nico
explained that they were not to be merged as the benefits of merging
were small and individual patches could be merged at a later date.
* mwaddel code reviewed arm_next.git and checked that required patches
were present. He is checking versatile express with it now.
* jk emailed Grant Likely to discuss the upstreaming on his patches, no
response yet.
* [ACTION] jk to report on progress of talking to Grant WRT
upstreaming his patches
* Matt talked to Lorenzo at ARM about device tree status. Based on jk’s
work patches are not quite ready yet but are close.
* jk pushed some patches on the boot interface. More work can be done
face-to-face with nico at the sprint. nico mentioned that
documentation is already there. Both considered this action done.
* There was some discussion on where the list of trees merged in to the
arm_next.git tree should be documented. nico included this
information directly into a tag in the tree. The question was asked
if a separate wiki page needed to also convey this information, the
consensus was not due to the constantly updating nature and nicely
seeing the documents history.
* jcrigby has been producing omap and versatile kernel flavours, no
others have been identified yet. jcrigby will talk to slangasek to
determine if versatile is still needed and if other flavours should
be added. This will be done face-to-face at the sprint.
* eric forwarded some patches to upstream public mailing list for the
single zImage blueprint. Consensus was that this work spans many
weeks and isn’t a specific action item but should be monitored.
=== Blueprints ===
* jk had a bunch of patches around for debugging mapping for device
tree’s and has been working on them this week. Good progress.
Discussions with Grant about mainlining device-tree’s proved
insightful.
* jk also mentioned that he will be working on imx51 in the near
future.
* ericm said that there may be a need for pbuild hacks to allow
multiple platforms to be built at once to help on the single zImage
blueprint. Work is ongoing.
* A lack of hardware is hampering progress at the moment as ericm only
has one platform.
* [ACTION] ericm to discuss with lool how to get more hardware.
* dmart mentioned that the missing security features spec was waiting
on kees but most items were either nearly complete or completed.
* dmart went on to inform the group that work on oProfile by willdeacon
was on track.
* Work on the versatile express board by mwaddel is ongoing and lots
have work items have been defined but mwaddel noted that once a
booting kernel is available most work items will be quick to
complete.
* JamieBennett asked the group to fill in their personal goals for the
sprint at
https://wiki.linaro.org/Events/2010-07-PlatformSprint/KernelConsolidationWG
and to suggest plenary topics.
* There will be no meeting next week due to the face-to-face sprint.
Activity reports from previous week:
= Arnd Bergmann =
== Getting started ==
* Built a qemu chroot for cross compiling
* Research about ARM CPU families
* Tried to set up Gumstix with uweigand
* Looked at the kernel for my WM8505 tablet
= Nicolas Pitre =
* Reviewed some of Eric Miao's patches on the linux-arm-kernel list.
* Reviewed Will Deacon's second round of patches to fix asm constraints
for atomic ops on ARM.
* Reviewed more of Uwe Kleine-König's Makefile cleanup patches and
asked him to produce a pull request for RMK. I also intend to
pull his patches into the arm-next tree.
* Had another look at the DT patch series. Somehow I still am not as
pleased with them as I would. Maybe I should take more time to study
the generic DT docs and possibly come up with more concrete
suggestions.
* Discussed with Kees Cook about the issue I was having with the
stack protector test module. Seems that it works only on array of
chars only and not of all types.
* Tryed to improve design in the OMAP hsmmc driver by discussing
alternative solutions on the mailing lists where some new development
were posted for review. The people involved seem to have agreed to
change their code.
* The usual mailing list monitoring and replying.
* Prepared a Git pull request for the Orion/Kirkwood/Dove stuff I'm
still maintaining for the next merge window.
= Jeremy Kerr =
== ARM ==
* device tree: rebased to gcl's new test-devicetree branch and rmk's
updated master branch
* device tree: implemented DT-pointer-in-r2 boot interface in ARM
kernel boot code (r2-direct branch)
* device tree: implemented DT-pointer-in-r2 boot interface in qemu
(r2-direct branch)
* device tree: maintenance on versatile branch
* chased down versatile timer issues; problem has since been fixed in
qemu
* built a simple debug initramfs
= Jamie Bennett =
== General ==
* Spent much of Monday morning catching up with emails, wiki's and
various other sources after my vacation. Furious wiki editing and
tidy-up. Decided to unsubscribe from a few less read mailing lists to
try to avoid the '' '3000-emails-sitting-in-my-inbox-after-vacation'
''syndrome.
* Took on TLE (Technical Liaison Engineer) role to be ARM's technical
point of contact. Extra workload at this stage should be manageable
but will be reassessed at a later date.
* Usual meetings, kernel consolidation minutes and associated wiki
pages done.
== Sprint Planning ==
* Lots of coordinating the Linaro sprint, answering peoples questions,
arranging hardware to be available, making sure we have enough room
and working with Anmar to take some tasks from him as he is on
vacation next week.
* Arranged with Stephen Doel to supply hardware for the sprint,
recommended specific items which should arrive by Friday next week.
== Blueprints and Releases ==
* Started discussions about how we can improve the blueprint/work item
tracking. For next cycle I will propose some changes to make progress
on work items much more transparent.
* Prodded team members individually to ensure they are on top of their
work items. There is a worrying amount of work still to do this cycle
and keeping on top of this is essential. Some blueprints need more
love than others so talked to the managers to ensure everything is
still on track.
* The Power Management and Toolchain work groups now have blueprints
associated with them. These are not tracked with the main release
burn-downs. Set up launchpad projects for each working group and
worked with pitti to get the work-item tracker recognising the new
projects.
== Development ==
* Spent some time looking at the launchpad-work-item tracker project to
try to understand how it can be hacked for our purposes. There are a few
features I will be working on before next cycle to make the Linaro
project easier to track.
* Pushed 'members as groups' support to the work item tracker. Now
projects with members who are actually groups can be tracked allowing
'meta-projects' to be created to track disparate projects all
together.
http://people.canonical.com/~pitti/workitems/maverick/arm-ubuntu.html
is an example of one such meta-project which is set up to track the 3
Linaro groups progress.
= Loïc Minier =
* Reviewed Amit's power management WG plans
* Tried out UBIFS on my sheevaplug, but couldn't get it to attach to
mtd2 on boot, would only attach manually after boot
* Built a maverick kernel for my beagleboard with the workaround for
LP #591941
* Looked at ARM's Android patches for RealView; lack hardware to test
them out, but I could try in QEMU
Bye,
--
Loïc Minier