[fedora-virt] Fedora virt status
Mark McLoughlin
markmc at redhat.com
Fri Jul 3 16:30:07 UTC 2009
F-12
====
F-12 is still coming at us fast:
http://fedoraproject.org/wiki/Releases/12/Schedule
2009-07-28 Feature Freeze (26 days)
2009-09-22 Beta (Final Development) Free (82 days)
2009-10-20 Final Release Compose (110 days)
libguestfs
==========
Much libguestfs activity has continued on fedora-virt at redhat.com
lately and Rich Jones has posted two related F-12 feature pages:
https://fedoraproject.org/wiki/Features/libguestfs
https://fedoraproject.org/wiki/Features/Xen_to_KVM_migration
FWN
===
Yet another weekly installment from Dale Bewley:
https://fedoraproject.org/wiki/FWN/Issue182#Virtualization
libvirt-0.6.5
=============
The libvirt folks have released another version of libvirt:
http://www.redhat.com/archives/libvir-list/2009-July/msg00060.html
This is mostly a bug fix release, though it includes some serious
improvements for storage/NPIV and on the OpenNebula driver:
* New features
- create storage columes on disk backend (Henrik Persson)
- drop of capabilities based on libcap-ng when possible (Daniel Berrange)
qemu-fedora.git
===============
git is now being used to manage Fedora's qemu patches:
http://www.redhat.com/archives/fedora-virt/2009-June/msg00157.html
Glauber and I are going to start using git to manage the patches for
the qemu package. The idea is that we'll use git's magical powers to
make it easier for us to fix problems in patches, cherry-pick patches
from upstream, re-base our patches to a newer upstream, etc.
http://git.et.redhat.com/?p=qemu-fedora.git
Already, this has proved useful re-basing to the latest
qemu-kvm-0.10.50 snapshot (kvm-87).
F-11 virtio-blk Performance
===========================
Charles Duffy, Michael Tokarev, Rich Jones and others recently
stumbled across the apparent cause of a dramatic slowdown in
virtio-blk performance in F-11 guests:
https://bugzilla.redhat.com/509383
Thanks for Michael Tokarev and Charles Duffy for the
following suggestion which has mostly fixed this problem:
for f in /sys/block/vd*/queue/rotational; do echo 1 > $f; done
It seems the kernel folks deemed it prudent to turn KVM and Xen
paravirt disks into "non-rotational" disks in 2.6.29, thereby
preventing guests from attempting to optimize block I/O and, instead,
leaving it up to the host:
http://lkml.org/lkml/2008/10/27/84
No doubt there'll be more on this as people look deeper into the
issue.
KVM USB Passthrough
===================
A discussion at FUDCon lead to some tips on USB passthrough being
posted to the fedora-virt list:
http://www.redhat.com/archives/fedora-virt/2009-June/msg00182.html
"KVM, enabling iPhone-touting geeks everywhere to use iTunes" :-)
Wiki
====
Dan Berrange has pulled together a couple of new wiki pages which
people might find interesting.
The first outlines the history of virt in Fedora, with details of what
changes were made in each release:
https://fedoraproject.org/wiki/Virtualization/History
The second pulls together links to all the virt status reports and FWN
virt sections:
https://fedoraproject.org/wiki/Virtualization/News
Xen
===
Michael Young posted more Dom0 kernel builds here:
http://www.redhat.com/archives/fedora-virt/2009-June/msg00205.html
Michael also pointed out Jeremy Fitzhardinge's latest prognosis of
getting the dom0 patches merged upstream:
http://lists.xensource.com/archives/html/xen-devel/2009-06/msg01193.html
For now I've backed off putting anything much into the current merge
window, or even the next, in favour of getting the tree into good
shape for Xen users and working out what needs to happen to get a
solid set of upstreamable patches.
Bugs
====
DOOM-O-METER: 242 open bugs a couple of weeks ago, 250 now
= New =
== kvm ==
https://bugzilla.redhat.com/508801
kvm "emulation failed (pagetable)" during OpenBSD 4.6-beta guest
boot (OOS shadow bug?)
It looks like this OpenBSD guest boot failure on F-11 might be due
to an OOS shadow bug but, then again, that should have been fixed
in 2.6.29.3.
https://bugzilla.redhat.com/509383
mkfs.ext2 much slower with virtio-backed disks than with
IDE-backed disks (NOT sparseness issue)
The problem here seems to be that virtio_blk is now defaulting to
non-rotational mode since 2.6.29. This may also be the cause of
bug #505695 and bug #504376.
== qemu ==
https://bugzilla.redhat.com/507899
qemu's console-via-VNC generates invalid rect position(s)
qemu's VNC server sends bogus rectangle sizes causing vncviewer to
abort.
https://bugzilla.redhat.com/507834
Clock drift in RHEL5.3 KVM Guests on Heavily Loaded Servers
A report of x86_64 RHEL5 guests time drifting under KVM with the
jiffies clocksource. Suggestion to boot with "notsc
divider=10". Macelo also suggests "clock=tsc" or "clock=jiffies"
with -no-kvm-pit-reinjection.
https://bugzilla.redhat.com/508258
qemu init script does nothing useful by default
The qemu initscript for registering binary handlers for the
usermode emulators does nothing unless you obtain the runtime
environment for the emulator from somewhere else. This is
lame.
https://bugzilla.redhat.com/509532
Move /usr/bin/qemu-kvm into the qemu-kvm package
A minor packaging change we should make for F-12.
== libvirt ==
https://bugzilla.redhat.com/508645
libvirt should use qemu's autoconnect mode for USB device
passthrough
qemu has a nice autoconnect mode for USB passthrough which libvirt
should probably use by default. Dan suggests that it should only
be used with managed='yes'.
https://bugzilla.redhat.com/508317
Allow sound devices to be used with svirt - tunnel sound over VNC
Covers the fact that with svirt we now disable sound cards until
we can tunnel sound over VNC.
https://bugzilla.redhat.com/509458
libvirtd default virtual network should not be enabled when
libvirt installed in a guest
This issue has been known for a while now - if you install
libvirtd in a guest, the default network is enabled and since it's
using the same IP range as the default network in the host, things
get screwed up. Suggestions on how to avoid having libvirtd or the
default network enabled by default in guest installs welcome :-)
== virt-manager=
https://bugzilla.redhat.com/508665
virt-manager (running as unprivileged user) doesn't let me install
from NFS locations
Because virt-manager no longer runs as root, NFS installs don't
work. Suggestion is that we'll use PolicyKit to allow virt-manager
to do NFS mounts.
== virt-top ==
https://bugzilla.redhat.com/508197
virt-top segfaults if run in Japanese
Now there's a bug that's right up Rich's alley!
= Resolved =
== qemu ==
https://bugzilla.redhat.com/508326
Force use of /sys/bus/usb for all USB host devices
qemu is currently prefering to use usbfs instead of sysfs for USB
passthrough. Fix sent upstream and built in rawhide.
https://bugzilla.redhat.com/506170
removing virtualization packages leaves qemu-system-x86
This is NOTABUG - if 'yum groupinstall' pulls in packages via
dependencies, then 'yum groupremove' doesn't remove them again.
== libvirt ==
https://bugzilla.redhat.com/506590
libvirt should ignore NUMA cells with missing topology
The libvirt fix for this was committed upstream but a recent F-11
numactl update is enough to fix the original issue for the
reported, so this has been marked as fixed upstream.
https://bugzilla.redhat.com/507555
libvirt starting a guest with ISO on NFS mount fails when
unnecessarily setting SELinux file context
Patch from Tim Waugh sent upstream and included in
libvirt-0.6.2-13.fc11 F-11 update.
https://bugzilla.redhat.com/505625
libvirtd segfault during virsh dominfo on security model
This libvirtd segfault which occurs when you do "virsh
capabilities" followed by "virsh dominfo" is fix in the
libvirt-0.6.2-13.fc11 F-11 update.
https://bugzilla.redhat.com/493692
Creating new VM; SELinux prevents opening iso image of install
media
https://bugzilla.redhat.com/499569
Guest with source-less cdrom fails to start :: Failed to set
security label
Two patches which we included in F-11, but neglected to send
upstream and were then dropped accidentally when we rebased to
0.6.4. All sorted out again now for F-12.
== virt-viewer ==
https://bugzilla.redhat.com/508825
virt-viewer doesnt have "full screen" hot key
It seems sensible that we would hook up F11 for fullscreen and
have virt-manager handle it, rather than passing F11 through to
the guest. Dan thinks this is a dumb idea, so the bug is closed as
WONTFIX.
More information about the virt
mailing list