Outstanding actions from last meeting
amitk to meet with keybuk on automated boot tests.
Didn’t really talk to keybuk about that in particular. But I now have the info necessary to start my own PXE booting automated installs.
Lucid Release Status: Bugs
Lucid Release Status: Milestoned Features
Lucid Release Status: Bugs with patches attached
Arsenal scripts are currently running in dryrun mode daily. I am finishing up adding the -r option in each script to enable the actual run. Reviewed the X debugging pages. I have started working up an outline of what we need to put in place and how we can use most of what we already have to do it. I’m also making up a listing of what information needs to be updated on our wiki.
Nothing to report this week.
Testing for any boot-speed regression with the PATA and SATA disk controllers removed is ongoing. We have also cleaned up those patches which were under discusssion, dropping a number of redundant patches.
Question: Did we decide anything w.r.t. having builtin drivers vs. modules?
Answer: That is part of what the first item there is about. I am planning to test that, and use that to start an email discussion on kernel-team.
Preliminary testing with nouveau has been pretty good. An upload of the new Lucid LBM is pending in the New queue currently.
There is some discussion starting that KMS on radeon may be not as good as it could be and we need to figure out what if anything we need to do there also
I (manjo) am looking at backporting some patches for radeon that might fix some of the memory issues that are being reported.
I have 2 patches:
- Prints a lot of info driver:device:resumetime
- Is based on tracepoints.
I have tested patch #1, and I am in the process of testing patch #2. I plan on tweaking patch #1 and carrying it as a sauce in Lucid.
Removed an unused feature that is potential security vulnerability (other option was complete the feature and do heavy code auditing). The feature could allow privilege escalation when combined with change_profile. No CVE as there is no way to exploit it with out it being explicitly used in policy, which has never happened except in experimentation. finished up policy optimization, it provided a 50% reduction in size and 4x speedup, also laid the ground work for further improvements (table sharing, and faster table compression) in the future. pam_apparmor update should be done later in the week. Fixing locking deadlock. The freeing of profiles has a race that can potentially deadlock. This can occur when free_profile is triggered from an rcu callback that is occurring at the same time as a profile replacement/removal.
Boot is looking better, plymouth has been pulled back out of initramfs and we are below budget again. We have one outstanding issue with khubd which seems to be hanging about, investigation is ongoing. We have also pulled some of the boot speed improvements back to the ARM branches for lucid, they are under testing now. We look to be at about 1.8s to rootfs right now
Other Release Tasks: Lucid Audio Support
Nothing new here, wiki and bug work needing to be done
Other Release Tasks: Lucid Better Power Mgt
I’ve been playing with CPU wakeups on a stock lucid isntall got it down to 5-6 wake ups after using some laptop-mode-tools tricks, blacklisting modules, etc. 5 wakeups/s now trying to figure out what parts of these hacks can be bundled with lucid I hope to have atleast some scripts for pm-utils.
Other Release Tasks: EC2 Lucid Kernel Status
Updated to latest set of patches, and updated configs to more closely match -server (with exception of physical device drivers, which diverged even more). Need to look at linux-meta-ec2 package as it has been reported by smoser that it isn’t updating to latest version of kernel.
We have refactored the debian abstraction to reduce the separation and thereby increase commonality of the rules and scripts. The kernel is currently being pulled up to v18.104.22.168 and should be uploaded today. Overall we are about half way through alpha-3 and we have completed about 58% of the milestoned tasks. Much of what remains undone are investigation items. We are making some progress to returning below the burn-down line: Canonical Kernel Team, Burndown Chart
Security & Bugfix Kernels
|2.6.31-20.57||(proposed)||4/18 verifications done|
Security release is done. This time it included updates to fsl-imx51, mvl-dove and ec2 packages. The ports-meta will be refreshed as well. confirmation. A new upload to Hardy proposed is waiting for approval and will also be done for netbook-lpia. The Karmic proposed upload includes a few more patches than before and also fixes a regression through 22.214.171.124 that was catched before.
Incoming Bugs: Regressions
Current regression stats (broken down by release):
regression-potential (up 14)
regression-update (no change)
regression-release (no change)
regression-proposed (no change)
Incoming Bugs: Bug day report
There was no bug day, the platform team was sprinting.
Open Discussion or Questions
- amitk wants to revisit our decision (dtchen’s really) to turn off HDA power save mode. He is looking for numbers on how many users were affected by the sound glitches. An email will be sent to the kernel-team mailing list starting a discussion and bradf will work with amitk on discussions with upstream.
- apw wants to remind everyone that we are frezing the kernel on March the 11th, which means we have to have everything ready and in by March 8th.
- smb points out that Karmic SRU for the kernel are about to end as well.