TCT Meeting/June 2014 Minutes

From Xen
Jump to navigationJump to search

Attendees

  • Don Dugger
  • Jun Nakajima
  • Ian Campbell
  • Daniel De Graaf
  • Olaf Herring
  • Jim Fehlig
  • Konrad Rzeszutek Wilk
  • Sherry Hurwitz
  • Aravind Gopalakrishnan
  • Daniel Kiper
  • James Bulpin

Review ACTION items.

  • Konrad to investigate qemu agent. (Was: DEFER)
    • DEFER AGAIN
  • Konrad to investigate KVP spec WRT compatibility etc. (Was: DEFER)
    • DEFER AGAIN
  • Jun to get status on sharing EPT and VTd page tables question
    • Jun: Right now we don't have issues from sharing. Jan had some concerns but not ???(I missed what was said here). Jun thinks we are OK today. At the same time might need asa feature, e.g. large pages for VT-d. Low priority.
    • Konrad: Discussion at hackathon, issues when doing GFX passthrough and trying to do DMA to the cards framebuffer. Print a warning and provide a workaround? Will mail people regarding hackathon discussion

Current technical challenges.

  • Jun: What is the status of PVH for dom0?
    • Konrad: Mukesh is working on it on the Xen side. Linux 3.16 has all the patches for the dom0 parts. Would appreciate more help from others.
  • Don: How to deal with conflict between maintainers. In particular co-maintainers of the same subsystem who disagree on a patch.
    • Konrad: Governance model provide backstop for when consensus has failed.
    • Don: Prefer consensus but try and resolve on a case by basis
    • Ian: Need to highlight to the co-maintainers that the disagreement is blocking something and they need to decide.

Coordination on current and future work.

(e.g. planned development activities for next release(s), roadmap)

  • Jim: Bamvor is looking into working on snapshotting via libxl and posting proposals but things have gone quiet.
    • Ian: hasn't been following very closely but isn't really sure what is going on.
    • Jim will advise him to gain consensus before working on the code much more.
    • Ian: Would be useful to get a highlevel overview without getting into specifics of implementation details (e.g qdisk mechanisms)