[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: [fedora-virt] Fedora virt status



Xen Dom0
========

Jeremy Fitzhardinge's latest submission of the Xen Dom0 patches for
2.6.31 has caused the kernel community to get themselves into a bit of
a flap. LWN has a nice article on the 'discussion':

   http://lwn.net/Articles/335812/

   But the stronger voice looks to be the one saying that the problems
   need to be fixed first. The deciding factors seem to be (1) the
   user-space ABI, and (2) the intrusion into the core x86 code; those
   issues make Xen different from yet another driver or
   filesystem. That, in turn, suggests that the Dom0 code is not
   destined for the mainline anytime soon. Instead, the Xen developers
   will be expected to go back and fix a list of problems - a lot of
   work with an uncertain result at the end.

Some more details:  The controversial bits are:

  (1) MTRR support.  Not that a big deal, can be skipped for the
      initial merge.  Might be PAT obsoletes this anyway ...
  (2) swiotlb hooks.  Not *that* problematic I think.
  (3) lapic + ioapic.  That is the big hot topic.  Current code lacks
      sensible interfaces to the different apic types out there in the
      wild.  The current dom0 patches hook just into that mess instead
      of cleaning it up.  Right now it looks like Jeremy has to sort
      that mess to get the xen bits in, using the to-be-created apic
      interfaces.

So it looks like Jeremy will again end up doing x86 arch code cleanups to get the xen bits in, like it happened before. Jeremy already did alot of x86 cleanup and unification work as part of the xen domU merge.

Will we have Dom0 support in the F-12 kernel? It's looking
increasingly unlikely, isn't it?

No way the apic stuff will make it into the 2.6.31 merge window opening RSN, the work on that barely started. Thus we can expect functional dom0 support upstream in 2.6.32 earliest. And even that isn't a safe bet, the apic stuff has to go quite smooth for that, on a road full of quirks for bugs-in-silicon ...

Oh yes - Gerd Hoffman has pushed a xen-3.4.0 update to F-11:

Pushed to rawhide. F-11 will follow, but most likely after the 3.4.1 bugfix release.


Also noteworthy: There is a discussion on xen-devel regarding the future of the xen linux trees. Which one should be used & maintained and so on. The options / kernel trees are:

 (1) The old 2.6.18 tree.
 (2) The 2.6.27 tree (patches forward-ported by suse, i.e. jan beulich).
 (3) The 2.6.29 gentoo patches (the 2.6.27 suse patches forward-ported).
 (4) jeremys kernel.org tree (aka pv_ops/dom0).

Looks like Keir seriously considers switching xen-unstable to the pv_ops kernel by default. YES! FINALLY! Well, it is probably to early to uncork the champagne, but I think we can at least put a bottle into the fridge ;)

cheers,
  Gerd


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]