LEG engineering meeting 15 Nov 2012

Attendance

In the EU/ASIA TZ meeting

  • Andrea Gallo
  • Anmar Oueija
  • Ilias Biris
  • Leif Lindholm
  • Ryan Harkin
  • Steve Capper
  • Steve McIntyre

In the US TZ meeting

  • Al Stone
  • Andrea Gallo
  • Anmar Oueija
  • Ilias Biris

Actions from last meeting

  • ACTION: Al to look into getting the list of Fedora packages which are dependent on inline assembly and would need to be reworked for ARM servers.
    • INPROGRESS: Al has started looking at the files meanwhile Jon Masters has provided his script to scan packages for inline assembly dependencies. Al is looking at cleaning that up and getting the list of scanned files.
  • ACTION: Leif should update the backlog wiki page with the items which are now done for GRUB work

    • DONE
  • ACTION: Rony to send the LAVA scripts for UEFI testing on Origen to Dave Pigott, and review the corresponding card with Ilias so that it can be closed
    • TODO: Rony needs to clean the scripts up a bit more and send them to Dave
  • ACTION Andrea to ping Grant Likely, Olivier Martin in order to organise the UEFI Cambridge sprint
    • DONE: The minutes have more details.

  • ACTION: Andrea will discuss this with Loic to clear what content should be placed under the external LEG wiki
  • ACTION: Ilias to initiate and followup the discussion with the necessary Linaro folks related to the FSF copyright assignment process for the upstreaming of GRUB significant updates
    • DONE: needs to be discussed at LEG-SC level and a recommendation should be sent to the TSC. Ilias has requested this from Andrea.

Meeting minutes

  • News
    • A discussion over UEFI was held on 13.Nov, covering the UEFI mini-sprint in December (Cambridge) and going over some of the details related to the UEFI SCT. The minutes have more details.

      • ACTION Anmar will take care of sending the invitation letter to Rony for the sprint as he needs to prepare his visa.
    • Project management activities: Anmar and Ilias will be the project managers looking after work in LEG. Anmar will take care of activities in Kernel, and bootloader sides. Ilias will be looking after work for benchmarking and optimisation as well as toolchain. Both will be looking after validation tasks.
    • News from the LEG-SC:
      • Last week there was first 1-1 with the representative from Facebook discussing the possible assignees
      • LEG-SC to discuss the issue of the representative for the TSC from LEG. This was proposed to be postponed but should not wait too long.
      • Review of the roadmap spreadsheet that Andrea has put together: we are in the process of aligning the spreadsheet targets with the Linaro WGs and TLs and getting the planned items in the spreadsheet into JIRA cards.
        • Toolchain discussion with Michael Hope and Matt Gretton-Dann on the Toolchain related requirements took place on the 15.Nov, their comments were incorporated directly on the requirements spreadsheet directly.
      • Reminder: we will be doing all the planning of the work using JIRA, rather than using JIRA for cards and Launchpad for Blueprints (bugs will still be handled by Launchpad). Ilias has the action point to set the guidelines on how this is supposed to work in JIRA. Anmar and Ilias are now scheduling discussions with the teams to go through their supposed deliverables, starting to look from the epic cards (CARD-210 and CARD-211).

        • The main interest why we are raising the use of JIRA is to remove the need of manual sync between cards and blueprints (for example looking at the picture starting from the cards it may often seem that there is nothing being done, whereas the blueprints say a different story - there is then some unnecessary manual synchronisation of status at the card level based on what is going on at blueprint level). A second reason is that we need to have detailed and almost real-time visibility on the status - so the more structure we create to cover status updates the less individual probing we are going to have when more and more assignees are coming in.
        • Andrea wants to have a special field per card to provide a succint status update by the engineers working on the card. This is possible using the "Engineering status update" field in each card, and can be collected automatically to populate status reports. ACTION Ilias to create a proposal related to the way we are going to use JIRA for planning and status updates.
  • Status
    • Al Stone:
      • Trying to get the list of inline assembly dependent packages released - the original list was actually against RH7 not against Fedora, so there is a bit more checking involved on the Redhat side to release that list. Al will follow up on that.

      • Al will talk to doko on how the list of packages was generated for the Ubuntu side.
      • ACPI: how does Al see the activities on starting the ACPI work

        • 1st step would be to review the specification
        • 2nd: get the acpica source code to understand how it builds/works on ARMv7 - Al should receive an Arndale board. Starting with Arndale (as opposed to v8) will help to think about potential solutions in automatically converting the device tree into acpi.
        • Starting to add the first static table once we know what are the possible issues. Should also add checking the support for the new ACPI v5 flag for the non-hw compliance. The current support in Linux kernel for acpica does not support non-hw compliance - so adding this support will be needed when sending patches upstream.

        • ACTION: Al to start breaking down the ACPI for ARMv7 to have something usable and demonstrable by Connect @ Hong Kong. Al should start breaking down into tasks to look then at how to start working with the cards with Ilias.
      • Related to Al's participation into the mini-sprint, it is not really necessary for him to be there as it will be a UEFI sprint. Perhaps we will need to have Al in a subsequent sprint when we start bringing UEFI and ACPI pieces together. We could nevertheless organise a hangout to review the summary (perhaps a daily quick review at the end of each sprint day).
    • Leif: GRUB2 on uboot patches have been rewritten and to be sent upstream this week for a second round of review (Leif was off-sick).

      • Ilias has requested from Andrea to discuss the issue of copyright assignment for GRUB2 with the LEG-SC (email discussion as there is no LEG-SC on wk 47).

    • Rony:
      • Has been a short week due to holiday season in India.
      • UEFI LAVA testing - still pending, need to push the scripts to LAVA, scripts were changed lately by the Rony's colleague who works on LAVA and he is on vacation so it will be closed when he is back.

      • A15 UEFI discussion with Ryan.
      • Arndale UEFI: A15 support for the SoC there is good support on the EDK2. To port UEFI on Arndale we need to enable a number of devices: sdcard, SATA, network in order. Then ACPI would need to be worked on. On the run up to the December mini sprint we should have sdcard support for UEFI on Arndale.

        • The UEFI code for the Arndale will be done on the same model used for Origen until we decide on a better way to handle the code in relation to Olivier's code.
          • Regarding how UEFI code will appear in the various trees, the Tianocore tree will have only core Tianocore code and ARM CPU support. Then platform specific code will reside in the "common tree" which will be maintained by the LEG UEFI maintainer (Ryan for the moment). All board support is then expected to move out of Olivier's tree and into Ryan's tree.
          • We want then to stage getting Olivier's patches pushed into the "common tree" maintained at Linaro, and then we push that to Tianocore, in order to get Olivier's changes and the board support changes in a common point before Olivier's changes land into the mainline tree. This will avoid breakages in the common tree maintained by Linaro. In this way mainline will always be more stable for us. This is also why it is important to get our changes also validated in LAVA so that we are taking care to have good quality tested code on the common tree.
          • For LAVA : When UEFI for Arndale will be available in order to get testing going on LAVA? Currently Rony is able to get the command prompt for UEFI. There is some problem with MMC driver - Rony is looking at the uboot MMC code to port into EFI. Since A15 support is proper Rony does not foresee major issues ahead. Also keep in mind that on Arndale we need the sdmux support for more robust testing of UEFI on LAVA. Without sdmux we can only test by calling UEFI from uboot, which allows testing of UEFI booting only. We need to also add SATA support in LAVA as well as the linaro media create tools. We need to contact Andy Doan and discuss of these extensions to LAVA. Anmar has already discussed this briefly - sdmux will work for Arndale but will not work for Versatile Express: Ryan will be writing some scripts to allow bootloader testing on versatile express and which Andy will integrate on LAVA. For the sprint we should get Dave Pigott to be present at least partly so that the LAVA actions are aligned and some solutions are reached.

    • Ryan: working on scripts to work locally on his board allowing reflashing the bootloader and boot a kernel with it, in order to get all the steps needed to put a new version of UEFI on the board, reboot it and prove that it works. The LAVA team will be able to take this in and make sure that they get the best-case scenario in booting cleared, as well as identify exceptions and possible error cases.
    • Steve Capper:
      • NUMA optimisations - getting familiar with the Linux memory model options. Has a barebones NUMA skeleton system almost working.

        • Will try to transfer some of the topology information regarding clusters as separate NUMA nodes. Will do some stability testing to confirm if there are any flat memory model assumptions hiding in the ARM specific bits of the code. Will also set some userspace hooks to test userspace libraries.
    • Steve McIntyre: finished off strace work upstream (work for armv8).

      • List of Ubuntu packages which are dependent on inline assembly is now available. Need to go through that and prioritise what to focus on. As soon as the Fedora package list is available we will be able to look into it as well in order to get a priority list of packages to focus on benchmarking and stress-testing. The list of ubuntu packages shows some low-level packages which a lot of other packages depend upon (eg libatomic) - there are however a lot of packages which have dependency on assembly which are not really important.

  • Other business
    • We now have a wiki for the LEG Engineering related items: http://wiki.linaro.org/LEG/Engineering. ACTION Ilias should move related wiki pages from the previous Internal wiki location to this new area in the wiki.

    • ACTION Anmar to take over running the US LEG engineering meeting.

Actions from this meeting

  • ACTION: Al to start breaking down the ACPI for ARMv7 to have something usable and demonstrable by Connect @ Hong Kong. Al should start breaking down into tasks to look then at how to start working with the cards with Ilias.
  • CARRIEDOVER: Al to look into getting the list of Fedora packages which are dependent on inline assembly and would need to be reworked for ARM servers.
  • ACTION Anmar will take care of sending the invitation letter to Rony for the sprint as he needs to prepare his visa.
  • ACTION Anmar to take over running the US LEG engineering meeting.
  • ACTION Ilias to create a proposal related to the way we are going to use JIRA for planning and status updates.
  • ACTION Ilias should move related wiki pages from the previous Internal wiki location to this new area in the wiki.
  • CARRIEDOVER: Rony to send the LAVA scripts for UEFI testing on Origen to Dave Pigott, and review the corresponding card with Ilias so that it can be closed
  • ACTION: Steve McIntyre should document his effort with the inline assembly dependent packages under the LEG Engineering wiki pages.

LEG/Engineering/Meetings/2012-11-15 (last modified 2012-11-17 07:38:48)