• Stephen Doel (Linaro)
  • David Rusling (Linaro)
  • Kiko (Linaro)
  • Roger Teague (ARM)
  • Paul McKenney (IBM)

  • Vijay Pasam (TI)
  • Andrea Gallo (ST-Ericsson)
  • Sree Kumar (Samsung)
  • Yves Vandervennet (Freescale)


Guest Participants

  • David Mandala (Canonical TSC Advisor)
  • Mark Orvek (MontaVista TSC Advisor)

TSC Face to Face Dallas 11th/12th January 2011


  • Set the scene for the work in this cycle
  • Review previous cycle
  • Preview this cycle

Two Day Agenda

  • Tuesday 11th January 2011
    • PM: Advisor's slot
    • Evening: TSC Dinner
  • Wednesday 12th January 2011
    • All day, face to face TSC meeting


Day 1: Tuesday Afternoon


Chair / Presenter


2.30 - 3.00


Welcome and introductions

3.00 - 4.00

Mark Orvek

MontaVista TSC Advisor

4.00 - 4.10



4.10 - 5.10

David Mandala

Canonical TSC Advisor

5.10 - 5.20



5.20 - ~6.00


Agree Wednesday agenda

7.00 -


Evening meal


Montavista: Challenges and Opportinities (Mark Orvek, Jim Lewis)

  • Linaro - Challenges and Opportunities.pdf

  • Challenges
    • Competition from Yocto?
      • Primarily Windriver/Intel backed
      • Completely tied to Poky/OE
      • Consumed linaro-toolchain
      • It's mostly a build system
      • Kiko: very useful for the roll-your-own market, less so for others
      • Also working on developer tools
    • ACTION: Probably worth issuing a blog post and a joint statement with LF/Yocto about Linaro; maybe wiki documentation
      • Make it clear that Linaro is here to fix the problems that nobody else is fixing
      • ACTION: Yocto may be interested in using more Linaro components
    • Developers need to contribute to both and Linaro?
      • Paul: we push things into both areas
    • ACTION: check on LWN statistics on Kernel, how do we ensure Linaro is mentioned?
    • Working with OSS distributions
    • ACTION: Figure out what Linaro wants to do about support
      • Open ad-hoc support on mailing lists
      • We will offer escalation support if a partner provides support
    • Balacing needs and requirements from primary members
      • Kiko: Haven't been significant conflicts up to now
      • David: Good input from members up to now
    • Evolution and growth
      • History is that consortiums do fail
      • Having our own engineering resources helps
      • CELF challenges
        • Fastboot filesystem
        • CELF 1.0
        • Building specifications and code
        • Code was dropped
      • David: growth challenges
        • Size of TSC
        • Broaden technical scope
          • ... when we are beyond phase I?
          • Automotive, server, etc
        • Growing influence in OSS and ARM
    • Continued investment by primary members
  • Opportunities
    • Position ARM to be more competitive with other arches
    • Interaction with customers resulting in more design wins
      • When customers come to Montavista they have already decided what platform they will use
    • Expand ecosystem in ARM Linux
      • Citrix browser plug-in example
      • Flash/Skype/Java
      • Binary Drivers/Codecs
      • Montavista has good expertise in the legal area
      • ACTION: question to the TSC: is it worth doing something as a group?
    • Help ARM expand into other markets
    • Use other open source projects without requiring official endorsement
    • More focus on A-Z embedded
      • Not focused on cross-compiling at the moment
      • True gap is in userland recipes, which are missing for lots of components
        • Kiko: Interested in more specific use cases of cross-compilation stories
        • ACTION: publish the list of packages that can be cross-compiled from our Ubuntu-based images
    • Working Group suggestions
      • Multicore/Virtualization: KVM on ARM?
        • Enterprise workloads
        • AMP; assymetric multicore on cores of different ISAs
      • Real-time
        • Real-time patches; "one day they will be in the kernel"
        • Important technology, and it works on Linux; why are people are still using RTOS?

Canonical: Challenges of Developing & Deploying an ARM Linux Distribution (David Mandala)

  • Challenges from a distribution perspective
    • No single ARM source kernel tree
      • Conflicting patches make it impossible to coalesce BSPs into a single source tree
    • Vendors have little interest in upstreaming code
      • Distributions end up testing
      • Mark: vendors have gotten better over the years
      • Kiko: Linaro Landing Team work has this as its primary goal; not there yet, but should solve both problems at once
    • Userland
      • NEON optional; requires code to be smart
      • Applications hacked up applications to run optimally on device X, but not on others
      • Thumb2 not well tested and porting still necessary
        • Mark: backtrace supported in thumb2?
        • Mark: Thumb2 quite used (4 out of 10 projects?)
        • Mark: hard-float seen for the first time
      • Kiko: do note the ARM porting project being worked out
    • Binary Drivers
      • Licenses restrict redistribution
      • Require specific kernels; not really ready to be built out-of-tree
        • Kiko: would it be enough for you if it was ready to be built out of tree?
        • nVidia model does work, and we could use it
        • The stub needs to be rebuildable
      • Missing X and OpenGLES drivers
        • We need something that actually boots up to graphics
        • Framebuffer is the bare minimum
      • Power Management different per-platform
        • Suspend/Resume/Hibernate
        • Kiko: Talk to Amit K about specific plans; we're already invested in this
      • Audio Drivers
      • Jim: Every customer wants the ability to do field-upgrading of devices
  • From an OEM perspective
    • Conflict between working on trunk and stable
      • BSP uses kernel with a different version than distribution kernel
        • Mismatch with distribution userspace
    • Why do Ubuntu ARM projects take longer than x86
      • SoC and board unreliability, custom OEM/ODM bootloader, image installation and upgrade is custom
        • Boot speed requirements; in particular for automotive
        • Display needs to be active before the kernel boots
        • Mostly happens because vendors historically used another bootloader
        • None of TI's customers use uboot
        • Montavista has uboot used in most customers
        • Security is probably the biggest issue
      • Factory process isn't well-suited to ARM devices
        • NAND flash issues; how are bad blocks handled?
      • Kernel
      • Graphics and Codecs
      • Target market demands for the product
        • Unrealistic expectations from OEM and ODMs
      • Codecs, Flash, Skype and other ISV software not available
        • Mark reinforces an Ecosystem WG
        • Linaro could make a big difference for Adobe
    • Suggestions
      • Legal education
        • Looks like a Binary Driver Conformance Program for Joe Bates
      • Push a specific bootloader
      • Unified kernel (device tree single source tree, single binary long-term)
    • If it's attached to the Internet, field upgradability is important








Kiko to check with Jesse and Alexander on his knowledge of any plans for graphics change in Honeycomb



bring to the meeting considerations on the last cycle and things we'd like to see changed




Benchmarking information from each member: what's used internally, what sort of measurement is done, sharing policy

Obsoleted, will hold a discussion at an upcoming TSC teleconference



All attendees to sprint need to be listed



Kiko, David

Consider a blog post and / or a joint statement with LF/Yocto about Linaro


Kiko, David

What Linaro components are appropriate for Yocto?


Rob Coombs

Check on LWN statistics on Kernel, how do we ensure Linaro is mentioned?



Do we want to do more with binary drivers (flash etc)? Could be an action for Joe.



publish the list of packages that can be cross-compiled from our Ubuntu-based images

Next Meeting

Wednesday, 12th January 2011

TSC/2011-01-11 (last modified 2011-01-17 10:14:06)