Revision 54 as of 2008-01-17 11:31:38

Clear message

MobileTeam Reporting Page


[:BuildingCommunity/TeamReporting:Back to the Team Reporting page...]

Welcome to the MobileAndEmbedded reporting page. Every month this page reveals a snapshot of the work that has been going on in the Ubuntu Mobile and Embedded Community. If you have have been working hard on the project and want to let others know what you have been up to, try to summarize it here. Community contributions are actively encouraged. A few guidelines for how to write good points are:

  • Keep it simple and short - single sentences are best. Smile :)

  • Try to summarise your work and not be over-detailed - we don't need to know specifics, just a summary of the kind
  • Where possible, include relevant links.

You should get all your reporting included here by the 22nd of every month.

The Report

Ubuntu Mobile [ MootBot Logs]

Status report for week 51

Moblin Kernel and Driver Status

  • 2.6.24 Kernel Support Alek continues to port the patches from 2.6.22 to 2.6.24. He is using/referencing the work previously done by Amit.
    • PSB / SLT HW Support
    Jacob still owes me a patch to work around the MWAIT/local APIC timer issue. We decided to go with the plan to mark the MWAIT instruction broken in the kernel taking out the CPU_FEATURE_MWAIT after cpuid is read for the effected processor. SDIO

    Feng has released a kernel patch on containing Marvell's GPL release of the 8688 WLAN + BT drivers. In the process we've removed the older & outdated MSS SDIO kernel patch from Pierre Ossman's SDIO stack is the only choice for SDIO support on moblin. Dabney Thermal patches No changes this week. USB Client Drivers Alek continues to add support for the USB Client solution that was described previously. I don't have an update on his status this week.

    PSB Gfx & Video Drivers Inuka has been investigating a bug with the PSB Gfx Beta3 RC driver where powertop is reporting that X is generating 120+ interrupts/second when the system is idle. This prevents the CPU from going into a

    lower power C state and blows battery life. The issue is believed to be in the implementation/design on the interface between closed & open pieces of the driver. After the holidays we'll be talking directly with the 3rd party developer to get this corrected. Power and Performance Testing/Optimization Jay has numbers that compares performance between the mobile browser compiled with gcc and browser compiled with Intel C/C++ compiler. As expected there are some significant improvement, for example browser startup appears to be about 25% faster with the version compiled with the Intel compiler. Before we get too excited, we need spend some time scrubbing the data and make sure the numbers we've collected are accurate.

Status report for week 02

Moblin Kernel and Driver Status

  • 2.6.24 & 2.6.22 Kernel Support Alek set up a 2.6.24-rc5 kernel tree with Moblin patches on In order to support both of 2.6.22 and 2.6.24 kernels we made relevant changes to have more consistent kernel package numbering:

    1. Original kernel-mid package was renamed to kernel-mid-2.6.22 and the binary packages it produced got renamed to linux-menlow-default-2.6.22 and linux-menlow-developer-2.6.22
    2. A new kernel-mid-2.6.24 package (now based on rc5) was added to Moblin already, and the binary package names were changed to linux-menlow-default-2.6.24 and linux-menlow-developer-2.6.24. The
    3. psb-kmd package was updated to build Gfx modules for both of these two kernels. The binary packages are psb-kmd-default-2.6.22, psb-kmd-developer-2.6.22 psb-kmd-default-2.6.24 and
      • psb-kmd-developer-2.6.24.
    4. The fset files for Moblin Image Creator got updated too.
    PSB / SLT HW Support The broken MWAIT/local APIC timer issue is assumed to be fixed in a new BIOS release through microcode patch to SLT. Once the fix is in a released BIOS and before we can enable-HT, we still need to take out the kernel boot option nolapic_timer in Moblin image-creator and then verify there is no boot hang and the extra interrupts issue goes away. This is the work our team needs to do. SDIO Feng is submitting a SDIO multi-function fix patch to Pierre Ossman's SDIO stack. The patch is still under review. Dabney Thermal patches No changes this week. USB Client Drivers No changes this week. The USBC solution as described before has been done and checked in already.

    PSB Gfx & Video Drivers Inuka is checking if the interrupts PSB Gfx Beta3 RC driver generating issue (120+ interrupts/second) is gone with the latest driver release. Inuka is also investigating a screen resolution issue which only happens on 2.6.24 kernel. With 2.6.24-rc5 and psb-kmd 2.6.24 module, the default X screen resolution is set to 1280x1024 instead of 1024x768. Power and Performance Testing/Optimization No changes this week. Grub fast boot/resume Martin enabled boot splash and resuming progress bar feature to grub package. Martin fixed the grub fast resume bug for 2.6.24 kernel and would check into the soon. Martin also discussed with EFI team to find how to take advantage of fast EFI bios booting. The fast EFI bios would only cost 2~3 second to hand over control right to boot loader after power up. In this case, Martin need port the fast boot patch from grub to elilo or grub2 which support EFI boot.

Midbrowser status

  • Right before the Holiday break, Asac helped Jimmy resolved some of the FF3.0 and Midbrowser integration issues. Asac also created a debian package for Pat McGowan so that Pat can demo it at CES. A gold star for Asac. We are planning to move Midbdrowser development to FF3.0. FF3.0 beta2 based MIdbrowser will be used for Moblin Beta3 release. QA from PRC has experienced browser performance issues on Crownbeach. Someone has been assigned to look into these performance bugs. Continued to work on bug fixes.

Status report for week 03

Moblin Kernel and Driver Status

  • Kernel Support

    We are talking with Ubuntu kernel team about how to better sync up kernel patches between us. The initial plan is: we will have a dedicated engineer -- Chetty, Jay<> in our team to take all of our Moblin patches and port, support and push them into Ubuntu's new branch of the Ubuntu kernel. Other Moblin kernel members will continue to commit changes to Moblin kernel with notification to Jay. And Jay may need assistance from the whole team when porting the changes to Ubuntu's kernel branch. Hope to finalize the plan during Intel/Ubuntu sprint.

    File-system corruption on hard shutdown ( issue got fixed after fixing the /dev/sda1 ext2 partition mounted as read-write bug. Thanks Michael Frey for his patch. PSB / SLT HW Support The broken MWAIT/local APIC timer issue got fixed after applied BIOS 66 to Crown-beach boards. Thus the kernel cmdline option nolapic_timer is not necessary any more. And when removed nolapic_timer options, the HT got enabled without 200+ interrupts per second issue.

    Jacob investigated many "APIC error on xx(xx)" kernel messages during booting issue (see Unclear APIC error msg reported when enable HT in CRB BIOS and found:

    1. BIOS does not set DTS (Digital thermal sensor) thresholds correctly.
    2. BIOS configures LAPIC thermal interrupts to be delivered via SMI.
    SDIO Feng committed a Marvell 8688 combo-card workaround (a force 100ms delay after power-on) to Moblin kernel. Dabney Thermal patches No changes this week. USB Client Drivers No changes this week..

    PSB Gfx & Video Drivers Inuka found an issue where PSB Gfx will set default resolution to 1280x1024 when drivers/acpi/video module was not inserted. But still need to rootcause this issue. Inuka has been working with Bryce to try to integrate PSB Gfx Beta3 driver to Hardy. Inuka has fixed the patch applying issue, but the built binary package does not function correctly. Inuka will continue to support on this issue. Power and Performance Testing/Optimization Jay completed a complete ACPI verification for Crown-beach board and fixed the issue where procfs did not get correct temperature. Jay also fired a bug “Closing the virtual lid does not turn the display off”

    ( and need to rootcause it. Jay also successfully built the midbrowser with Intel compiler (ICC) and did a performance comparison between the GCC built midbrowser and found a very good increase in performance (25%+ for browser launching, and 20%+ page loading from QA team) Grub fast boot/resume Martin is working on fixing the issue where fast resume gets hung when HT is enabled.