Parent


Attendees

  • David Rusling (Linaro)
  • Kiko (Linaro)
  • Sree Kumar (Samsung)
  • Roger Teague (ARM)
  • Paul McKenney (IBM)

  • Andrea Gallo (ST-Ericsson)
  • Yves Vandervennet (Freescale)

Absent

  • Vijay Pasam (TI)

Information

Guest Participants

Agenda

  • Review Kernel WG requirements
  • Previous actions
  • AOB

Minutes

  • Actions from previous meetings
  • hardware baseline is ARMv7A Thumb2, Neon and SMP; all Cortex cores (A5, A8, A9, A15)
  • Priorities
    • Kernel consolidation
    • embedded Linux kernel enhancements
  • Linus - "why don't you guys get together and organise this!"
    • A key part of consolidation
  • Embedded kernel features
    • Reliability, Availability, Servicability (RAS)
  • Priority - general TSC agreement, plus a need to see detail in upstreaming process
    • Sree (Samsung) (agree with the list)
    • Roger (ARM) (agree with the list, but keen to make sure that the process is effective (getting stuff into the mainline kernel))
    • Andrea (ST Ericsson) (apologies have
    • Yves (agree with the list, agree that K1.1 is the most important)
  • Linaro next
    • Confirmed that the tree only contains patches on their way upstream
    • Device Tree is a tolerated exception
  • 2.6.39 is it, we have no control over the kernel release cycle, so may be able to support 2.6.40 if it comes out within the 6 months cycle
  • Priorities
    • Priorities are in terms of how priorities will be fed into the kernel working group
    • Have to be aggressive about leveraging the community (manufacturers and open source)
    • This is the list for the kernel team
    • Low priority means will most probably not happen
    • Medium is 'at risk'
    • Some work items sized, others not (are impossible to size)
  • K1. Kernel consolidation
    • Still discussing K1.1
    • Discuss at ELC and Linaro@UDS
    • Need buy in from Russell, Linus etc
  • Roger, UEFI work looks 'at risk'
    • Punting boot architecture to OCTO
    • K3.2 needed as a priority
    • Bracket K3.1 and K3.2 together as medium?
      • Use Elba for this?
      • Elba availability - 2 boards, 1 in the May time frame
      • In limited numbers currently, will eventually be available commercially
    • ARM request going in to Linaro is that UEFI makes progress
    • ACTION: Figure out how to execute K5.3 (Dave and Kiko).
  • K2.2 - need to demonstrate DT end to end, all the way up the stack
  • K4.2 - is this a single binary (ACTION: Paul McKenney to investigate)

  • K5.2 A15 support
    • ARM work has been pushed upstream
    • LAVA can support models
    • ARM interested in A15 being validated by Linaro
      • Linaro prefers silicon, FPGA and simulation tends to be less efficient (because of the learning curve)
    • ARM wants to show the ARM partners that Linaro has incorporated good A15 support
    • There will be a validated (fast) model, silicon becomes available very shortly after
      • ACTION: How do we integrate the validated model in Linaro's validation flow (LAVA)?
  • UEFI
    • is one of Samsung's priorities, but Samsung is happy with Paul's essentially priorities (K1, K2, K4, K8)
  • What is our position with respect to xfat? Do we rely on third parties for xfat?
    • K7.8, SDXC controller support, standard format for SDXC is xfat.
    • Prototype patches posted in 2009, don't know the upstream status
  • K8.3
    • this is SMP only
    • ACTION: David Rusling investigate how this could be done elsewhere, might be a server topic
  • Resourcing discussion
  • AOB
    • Upstreaming of patches
      • All patches from the landing teams and working groups go both into Linaro next and upstream
      • How far behind upstream is Linaro next?
        • 3 to 6 months (1 to 2 releases), variable
      • Patches get queued in a maintainer tree waiting for a merge window
        • Nico is a maintainer of several subsystems (co-maintainer in most cases)
          • Nico pulls submitted patches into his tree and Linaro next
          • Linus pulls from Nico's maintainer tree, Russell's tree
      • Nobody upstream is pulling patches directly from Linaro next (which is a consolidation tree not a maintainer tree)
      • Patches get submitted to the appropriate maintainer tree (TI tree, Samsung tree etc)
      • K1.1 needs a separate maintainer tree
      • Do the monthly releases change this process?
        • No, but we have to produce trees that have the 'right' patches in them
      • This is the most optimal way to get upstream
        • The process is submission, discussion, conflict resolution

Actions

Number

Action

Owner

State

2011-04-06-1

Figure out how to execute K5.3

Dave and Kiko

2011-04-06-1

How do we integrate the validated model in Linaro's validation flow (LAVA)?

Paul Larson

2011-04-06-1

K8.3 investigate how this could be done elsewhere, might be a server topic

David Rusling

Next Meeting

TSC/2011-04-06 (last modified 2011-04-15 18:08:04)