Difference between revisions of "TCT Meeting/January 2014 Minutes"

From Xen
m
 
(2 intermediate revisions by the same user not shown)
Line 9: Line 9:
 
| content1 = [[Xen Technical Coordination Team (TCT) Meeting|Home]]
 
| content1 = [[Xen Technical Coordination Team (TCT) Meeting|Home]]
 
}}
 
}}
  +
   
 
= Attendees =
 
= Attendees =
   
  +
* Andres Lagar-Cavilla
  +
* Andrew Cooper
  +
* Anthony Liguori
  +
* Daniel Kiper
  +
* Don Dugger
 
* Ian Campbell
 
* Ian Campbell
* Don Dugger
+
* Jan Beuluch
  +
* Jun Nakajima
  +
* Kelly Zytaruk
 
* Konrad Rzeszutek Wilk
 
* Konrad Rzeszutek Wilk
  +
* Olaf Herring
* Kelly Zytaruk
 
* Sherry Hurwitz
+
* Sheery Hurwitz
  +
* Suravee Suthikulpanit
* James Bulpin
 
* Ian Pratt
 
* Jun Nakajima
 
* Olaf Hering
 
* Boris Ostrovsky
 
* Daniel De Graaf
 
* Daniel Kiper
 
 
= TCT Mailing List =
 
 
List for administrivia, agenda, minutes etc. No objections
 
 
= Review Action items =
 
 
None
 
 
= Current technical challenges =
 
 
== Perf subsystem ==
 
 
Sherry: Interested in Boris' work on the perf subsystem. How solid/complete is it.
 
 
Boris: Very. Xen side being reviewed on the list, Linux side not posted yet, can send out on demand.
 
 
= Coordination on current and future work =
 
 
== GPU passthrough ==
 
 
Kelly: Is anyone working on GPU passthrough as a primary device?
 
 
Ian C: Not sure
 
 
Konrad: Intel has patches?
 
 
Ian P: Traditionally Intel can only be used as a primary, while AMD and NVIDIA can only be used as a secondary
 
 
Konrad: Was thinking of the XenGT.
 
 
Jun: This is virtualising the GPU, not assignment.
 
 
Jun: Passthrough depends on the GFX card being used, depends on (V)BIOS etc
 
 
Kelley: A lot of internal architecture doesn't support primary passthrough, sent Konrad a list of deficiencies. Has it working on a test system but not ready for primetime. If no one else is working on it then may pick it, willing to work with other interested parties. Looking into the qemu side.
 
 
Konrad: Three versions of qemu, upstream, version in Xen and the old one
 
 
IanC: The first two are the same thing.
 
 
Jun: QEMU/KVM guys are using VFIO for gfx passthrough.
 
 
IanC: Can Xen use VFIO? Thought it was very KVM centric.
 
 
Jun: Might be areas where we can share.
 
 
Konrad: Bus reset and slot reset code in pciback is buggy and needs work.
 
 
IanC: thought we did that via sysfs these days.
 
 
Konrad: For devices yes, but not for slot or bus etc. Generic code will refuse to reset if there are multiple functions etc. There's a bunch of code in Linux which VFIO uses to work out whether to do a slot or bus reset. Code for pciback to use this same infrastructure has been posted but needs some rework.
 
 
Jan: Changes to memory type changes are a prerequisite for GPU passthrough to work, at least whenever shared RAM is used for communications with the GPU. Currently we have an arbitrary mix of UC and WB, which leaves us at risk of not doing what the programmer wanted. Some patches went in, the remainder depend on vmx maintainers to comment on certain aspects of EPT and why IOMMU related things are done the way they are.
 
 
{{Action|Don & Jun}} Don and Jun will chase this up inside Intel.
 
 
== Xen nested on Xen performance ==
 
 
Ian P: have been doing some work on this usecase. VMCS shadow on Haswell help, but can do better if detect you are nested and modify your behaviour, can do even better than VMCS shadowing. No need to change the root hypervisor. Effectively a PV hypervisor with no changes to the host/root hypervisor. Writing to VMCS directly avoiding VMREAD/VMWRITE -- non-architectural but goes an awful lot faster. Would like to propose a version identifier in the virtual VMCS for nested hypervisor to use, and access VMCS directly on match.
 
 
We use a virtual VMCS which does not match the physical VMCS, when VMCS shadow is disabled. Layout is arbitrary.
 
 
Code in Xen looks for changes in the virtual-VMCS and syncs to the physical-VMCS. A kind of lazy update, update everyting and then trigger a resync with VMLOAD. Relies on Xen resyncing the entire VMCS. Effectively batches up.
 
 
The nested VM needs to know something about how the root Xen behaves so need some way to determine when it does/doesn't know about this.
 
 
No changes to the root Xen, but lots to the nested Xen.
 
 
IanC: Need to see both the nested and root side things, in order to test and guarentee.
 
   
  +
= Review ACTION Items =
Don: Hard to justify without something to take advantage of it.
 
   
  +
* Konrad to arrange for some PVH instructions for test days. Blog post in progress.
Konrad: Adding an ABI to the root hypervisor.
 
   
  +
= Current Technical Challenges =
Don: Need to be careful to bump the version appropriately. Someone needs to think about it for every change. Need to test etc.
 
   
  +
* Konrad: Reports PCI passthrough Regresssion in 4.4 qemu-xen for PCI passthrough.
Ian P: Could use hypervisor version number.
 
  +
** 4.3 qemu-xen did work.
  +
** Anthony didn't have hardware, has been directed to where to find some.
  +
** Konrad is seeing it with a normal NIC (Intel 82576), passing through the PF and not the VF.
  +
** Don: Intel have tested passthrough with 4.4 and qemu-xen and it works for them.
  +
** Andrew: Could also be the SeaBIOS update
  +
** Konrad: Tested version from June/July.
  +
** {{Action|Konrad}} Konrad to contact Don to see if we can resolve
  +
* Andrew Cooper: 4.4 has been through a XenRT run with XenServer. But didn't show PCI passthrough because that isn't tested.
  +
** Seeing similar Windows issues to those seen with osstest
  +
** Some PV guests didn't reboot, still to diagnose
   
  +
* Sherry: Is the community asking for testing of various PCI devices
Ian C: This can break on backport.
 
  +
** Ian: Yes.
   
  +
* Suravee: Regression on 4.3/4.4 for AMD IOMMU. Fix was committed yesterday, should probably be back ported.
Ian P: Currently the version field is the hardware version, which is very wrong. A hardcoded number would be more correct.
 
   
Don: Intel are doing ongoing work on nested.
+
* Boris: What are the plans for 4.3.2?
  +
** Jan: was planning 4.2.4. Waiting on Ian Jackson for tools backports.
  +
** For 4.3.2, at least a month before the regular schedule would call for a new 4.3.x release. There don't seem to be many big changes.
  +
** May align 4.3.2 and 4.2.4 releases at end of Jan early Feb.
   
  +
= Coordinaton of future work =
Ian P: Will propose patches for comments.
 
   
  +
* Konrad is continuing to work with testing PCI passthrough.
== Roadmap for 4.5 ==
 
  +
** Concentrating on reports for now, once there is some firm idea of what needs to be done will start looking at patches (but rather tied up with PVH at the moment)
  +
** Sherry: Kelly (on the call) can answer questions about graphics passthrough to radeon cards etc.
  +
** Kelly: Is working with 4.2 and has identified some issues with GFX passthrough, with qemu and seabios in particular. Has some fixes but needs to rebase to 4.4. Issues are mainly with passthrough of primary.
  +
** Konrad: Having issues with secondary passthrough, issues with rombios fixup.
  +
** ROMBIOS needs patches from David Techer in some circumstances
  +
** Kelly: Hardware has two BARs, which are not remapped correctly. video bios does not go through pci config space to get the bars so the accesses are not trapped and qemu doesn't update the mapping.
  +
** Andy Cooper: One of David Tetcher's patches looks relevant.
  +
** Kelly: Problems with multiple graphics cards in the host. Takes BIOS from 0xc000 (i.e. out of memory instead of out of ROM). Which means the boot card's BIOS gets used, which fails with passing through a non-boot graphics card. Fixed by reading the image from the rombar.
  +
** {{Action|Kelly}} Kelly to mail Konrad and/or xen-devel with the current set of issues.
  +
** Focus should be on qemu-xen+seabios and not qemu-trad+rombios
   
  +
* Andrew Cooper: There is an issue with high MMIO regions, PV pvops kernels cannot handle these well. Which breaks passthrough of these cards. Affects dom0
Sherry: Have we started a roadmap, wiki page etc
 
  +
** Konrad: Known and in progress.
   
  +
* Release: Looking like mid Feb.
Ian C: People are still catching their breath. Once the dust has settle and the Release Manager has been decided can revisit. Next time perhaps?
 
   
  +
* Konrad: We have the option of sponsoring an internship via the GNOME Outreach Program for Women.
= Community news, activities =
 
  +
** Soliciting ideas for interesting projects on Xen or Linux. Wants people to think about it
  +
*** Andres: Recent Xen paging discussions.
  +
** Will be a formal request as the process ramps up.
   
  +
* Sherry: Next Xen test day.
None
 
  +
** Ian: I should know but I don't. Plan was maybe to have one alongside each RC? Should know but doesn't
   
 
[[Category:Technical Coordination Team]]
 
[[Category:Technical Coordination Team]]

Latest revision as of 12:32, 31 March 2014


Attendees

  • Andres Lagar-Cavilla
  • Andrew Cooper
  • Anthony Liguori
  • Daniel Kiper
  • Don Dugger
  • Ian Campbell
  • Jan Beuluch
  • Jun Nakajima
  • Kelly Zytaruk
  • Konrad Rzeszutek Wilk
  • Olaf Herring
  • Sheery Hurwitz
  • Suravee Suthikulpanit

Review ACTION Items

  • Konrad to arrange for some PVH instructions for test days. Blog post in progress.

Current Technical Challenges

  • Konrad: Reports PCI passthrough Regresssion in 4.4 qemu-xen for PCI passthrough.
    • 4.3 qemu-xen did work.
    • Anthony didn't have hardware, has been directed to where to find some.
    • Konrad is seeing it with a normal NIC (Intel 82576), passing through the PF and not the VF.
    • Don: Intel have tested passthrough with 4.4 and qemu-xen and it works for them.
    • Andrew: Could also be the SeaBIOS update
    • Konrad: Tested version from June/July.

    • Pictogram voting comment 15px.png Action Konrad: Konrad to contact Don to see if we can resolve
  • Andrew Cooper: 4.4 has been through a XenRT run with XenServer. But didn't show PCI passthrough because that isn't tested.
    • Seeing similar Windows issues to those seen with osstest
    • Some PV guests didn't reboot, still to diagnose
  • Sherry: Is the community asking for testing of various PCI devices
    • Ian: Yes.
  • Suravee: Regression on 4.3/4.4 for AMD IOMMU. Fix was committed yesterday, should probably be back ported.
  • Boris: What are the plans for 4.3.2?
    • Jan: was planning 4.2.4. Waiting on Ian Jackson for tools backports.
    • For 4.3.2, at least a month before the regular schedule would call for a new 4.3.x release. There don't seem to be many big changes.
    • May align 4.3.2 and 4.2.4 releases at end of Jan early Feb.

Coordinaton of future work

  • Konrad is continuing to work with testing PCI passthrough.
    • Concentrating on reports for now, once there is some firm idea of what needs to be done will start looking at patches (but rather tied up with PVH at the moment)
    • Sherry: Kelly (on the call) can answer questions about graphics passthrough to radeon cards etc.
    • Kelly: Is working with 4.2 and has identified some issues with GFX passthrough, with qemu and seabios in particular. Has some fixes but needs to rebase to 4.4. Issues are mainly with passthrough of primary.
    • Konrad: Having issues with secondary passthrough, issues with rombios fixup.
    • ROMBIOS needs patches from David Techer in some circumstances
    • Kelly: Hardware has two BARs, which are not remapped correctly. video bios does not go through pci config space to get the bars so the accesses are not trapped and qemu doesn't update the mapping.
    • Andy Cooper: One of David Tetcher's patches looks relevant.
    • Kelly: Problems with multiple graphics cards in the host. Takes BIOS from 0xc000 (i.e. out of memory instead of out of ROM). Which means the boot card's BIOS gets used, which fails with passing through a non-boot graphics card. Fixed by reading the image from the rombar.

    • Pictogram voting comment 15px.png Action Kelly: Kelly to mail Konrad and/or xen-devel with the current set of issues.
    • Focus should be on qemu-xen+seabios and not qemu-trad+rombios
  • Andrew Cooper: There is an issue with high MMIO regions, PV pvops kernels cannot handle these well. Which breaks passthrough of these cards. Affects dom0
    • Konrad: Known and in progress.
  • Release: Looking like mid Feb.
  • Konrad: We have the option of sponsoring an internship via the GNOME Outreach Program for Women.
    • Soliciting ideas for interesting projects on Xen or Linux. Wants people to think about it
      • Andres: Recent Xen paging discussions.
    • Will be a formal request as the process ramps up.
  • Sherry: Next Xen test day.
    • Ian: I should know but I don't. Plan was maybe to have one alongside each RC? Should know but doesn't