KVM in 14.1

fdonkers · 01-23-2014, 12:39 PM

I had the same error when I tried to install windows server. It seems that the new 0.10.0 version of virt-manager doesn't create a USB host controller in the virtual machine. Also, the default pointer device seems to be a graphics USB tablet instead of a mouse. Without a USB bus, the graphics tablet cannot be handled correctly.

I got around that by copying an existing machine using virsh. The "virsh dumpxml" command will output the XML definition of an existing machine. I copied it to a file and edited the definition to match my new machine (name, hd path). Be sure to remove the unique uuid string. Then I used "virsh define" to define the new machine.

Note: when i create a new slackware vm, it works without errors. This new vm has a PS/2 mouse and no USB bus.

Gerardo Zamudio · 01-23-2014, 01:03 PM

Do you recommend downgrading to the previous version of virt-manager, then? I don't have any machines created at the moment as I've just set up this box. Would creating a new Slackware VM then dumping and editing its xml to create the Windows VM work?

fdonkers · 01-23-2014, 01:40 PM

When I create a new Slackware vm, it adds a PCI, SATA and IDE controller, but not USB. So copying from that machine won't work.

Perhapt your best bet is to downgrade virt-manager. Make sure you check the required packages, because those have changed in the latest version.

ponce · 01-23-2014, 02:05 PM

maybe I have understood the culprit: virt-manager is fine, you got to upgrade libvirt/libvirt-python to 1.2.1.

the patches for them are in git and should go in the next upgrade

http://slackbuilds.org/cgit/slackbui...535b95ae48dea5
http://slackbuilds.org/cgit/slackbui...1db134b9ac936f

I just tried with libvirt-1.2.0 and got your same error, after upgrading to 1.2.1 all went fine.

Richard Cranium · 01-23-2014, 02:27 PM

ponce saves the day again!

Dude, what would we do without you?

Gerardo Zamudio · 01-23-2014, 09:56 PM

Quote:

Originally Posted by ponce

maybe I have understood the culprit: virt-manager is fine, you got to upgrade libvirt/libvirt-python to 1.2.1.

the patches for them are in git and should go in the next upgrade

http://slackbuilds.org/cgit/slackbui...535b95ae48dea5
http://slackbuilds.org/cgit/slackbui...1db134b9ac936f

I just tried with libvirt-1.2.0 and got your same error, after upgrading to 1.2.1 all went fine.

Thanks for this, it worked! Now I have the other issue which is virt-manager appears to not be picking up my QtCurve/icon theme in KDE. My other GTK programs like Firefox and Pidgin recognize it just fine after installing the QtCurve-Gtk2 package and changing the etc/gtk-2.0/gtkrc symlink. It looks like an issue similar to this one. I suspected spice-gtk and gtk-vnc were using GTK3, so I rebuilt those packages explicitly using only their GTK2 version, but still no go. I get the following error:

Code:

Error starting Virtual Machine Manager: Requiring namespace 'Gtk' version '2.0', but '3.0' is already loaded

Traceback (most recent call last):
  File "/usr/share/virt-manager/virt-manager", line 303, in <module>
    main()
  File "/usr/share/virt-manager/virt-manager", line 267, in main
    from virtManager.engine import vmmEngine
  File "/usr/share/virt-manager/virtManager/engine.py", line 45, in <module>
    from virtManager.details import vmmDetails
  File "/usr/share/virt-manager/virtManager/details.py", line 37, in <module>
    from virtManager.console import vmmConsolePages
  File "/usr/share/virt-manager/virtManager/console.py", line 27, in <module>
    from gi.repository import GtkVnc
  File "/usr/lib64/python2.7/site-packages/gi/importer.py", line 68, in load_module
    dynamic_module._load()
  File "/usr/lib64/python2.7/site-packages/gi/module.py", line 289, in _load
    self._introspection_module = get_introspection_module(self._namespace)
  File "/usr/lib64/python2.7/site-packages/gi/module.py", line 268, in get_introspection_module
    module = IntrospectionModule(namespace, version)
  File "/usr/lib64/python2.7/site-packages/gi/module.py", line 117, in __init__
    repository.require(namespace, version)
RepositoryError: Requiring namespace 'Gtk' version '2.0', but '3.0' is already loaded

Any thoughts?

ponce · 01-23-2014, 11:39 PM

virt-manager 0.10 uses gtk+3 so you *have* to build gtk-vnc and spice-gtk with gtk+3 support and use a gtk+3 theme.

Gerardo Zamudio · 01-24-2014, 12:31 PM

Thanks for the reply, yesterday I fiddled around with it and did discover from the virt-manager changelog that as of version 0.10.0, it has been ported over to GTK3.

fdonkers · 01-25-2014, 08:52 AM

You can install "kde-gtk-config" from slackbuilds.org. It shows up in KDE System Settings, Application appearance, GTK configuration. There you can set the themes for GTK2 and GTK3. It fixes the icon visibility issue in virt-manager.

TobiSGD · 02-19-2014, 10:19 PM

I just had the same problem with this error-message

Quote:

Unable to complete install: 'Controller 'devices' is not wanted, but 'cpu' is co-mounted: Invalid argument'

with libvirt 1.2.1 and libvirt-python 1.2.1.
I found this blog post with the solution, after putting that script into my rc.local (before the start of libvirtd) it works without a problem, though I am using Qemu and not LXC.
https://strangeworlds.co.uk/blog/201...and-slackware/

ponce · 02-19-2014, 11:51 PM

Hi TobiSGD,

if you're using 14.1 mounting cgroups in subfolder is already done by the /etc/rc.d/rc.S script, as it contains this block that should take care of it

Code:

# Mount Control Groups filesystem interface:
if grep -wq cgroup /proc/filesystems ; then
  if [ -d /sys/fs/cgroup ]; then
    # See linux-*/Documentation/cgroups/cgroups.txt (section 1.6)
    # Check if we have some tools to autodetect the available cgroup controllers
    if [ -x /usr/bin/lssubsys -a -x /usr/bin/tr -a -x /usr/bin/sed ]; then
      # Mount a tmpfs as the cgroup filesystem root
      mount -t tmpfs -o mode=0755 cgroup_root /sys/fs/cgroup
      # Autodetect available controllers and mount them in subfolders
      controllers="$(lssubsys -a 2>/dev/null | tr '\n' ' ' | sed s/.$//)"
      for i in $controllers; do
        mkdir /sys/fs/cgroup/$i
        mount -t cgroup -o $i $i /sys/fs/cgroup/$i
      done
      unset i controllers
    else
      # We can't use autodetection so fall back mounting them all together
      mount -t cgroup cgroup /sys/fs/cgroup
    fi
  else
    mkdir -p /dev/cgroup
    mount -t cgroup cgroup /dev/cgroup
  fi
fi

I'm guessing, maybe for some reason autodetecting doesn't work?

have a look also at this topic.

TobiSGD · 02-20-2014, 05:19 AM

OK, the problem was that in the rc.S on my laptop that I used for testing the mounting of cgroups looked totally different. This may be a remainder from previous experiments of from version upgrades, I replaced it with a stock rc.S and now it works as it should. Thanks for that link.