When testing security updates, it is important to test the update in a full Ubuntu environment for the release being tested. Put simply, an update for an Ubuntu 16.04 LTS package should be tested in a full install of Xenial. The Ubuntu Security team has created some scripts and put them into ubuntu-qa-tools. These tools use kvm and libvirt, the preferred virtualization technology in Ubuntu. KVM requires the virtualization extensions to be available and enabled in your BIOS. You can test to see if you have these by using the kvm-ok command. QEMU is an alternative and can be used with libvirt and uvt, but it is slow. If you cannot use kvm, then it is worth looking at another virtualization technology such as virtualbox.

uvt is essentially a wrapper script for virsh and virt-install for both making VM creation repeatable and to help batch commands to multiple VMs. Using uvt should not be considered mandatory as all this can be achieved via other means (though use of uvt new is encouraged and should save you time).

Setting up uvt on Precise

Much of this (and more) can be found in the vm-tools/README file.

  1. Install the necessary software:

    $ sudo apt-get install ubuntu-virt-mgmt ubuntu-virt-server genisoimage xorriso
  2. Add your user to the libvirtd group:

    $ sudo adduser <username> libvirtd
  3. Download the ubuntu-qa-tools branch:

    $ git clone lp:ubuntu-qa-tools
      Note: If you have issue trying the above you can try directly on one of these: git://git.launchpad.net/ubuntu-qa-tools, git+ssh://git.launchpad.net/ubuntu-qa-tools, https://git.launchpad.net/ubuntu-qa-tools
  4. Add the UQT_VM_TOOLS environment variable to your startup scripts (eg ~/.bashrc) and have it point to the ubuntu-qa-tools branch:

    export UQT_VM_TOOLS="$HOME/git-pulls/ubuntu-qa-tools/vm-tools"
  5. update your PATH to include the vm-tools directory (eg via ~/.bashrc):

    export PATH="$PATH:$UQT_VM_TOOLS"
  6. Add bash-completion support for uvt by adding the following to your ~/.bash_completion files:

    if which uvt 1>/dev/null; then
      source $(dirname $(realpath $(which uvt)))/uvt-completion.bash
  7. [OPTIONAL] While uvt will use reasonable defaults without a config file, you may also use uvt config to create $HOME/.uvt.conf. It can then be customized to have something like:

    # Configuration file for uvt tool
    # vm_dir_iso: sets the location where .iso images are to be found
    #             and downloaded to. This defaults to $HOME/iso if unset.
    # vm_dir_iso_cache: sets the location where preseeded iso images will be
    #                   cached. This defaults to $HOME/iso/cache if unset.
    # vm_path: sets the location where the libvirt virtual machines will be
    #          stored. This defaults to $HOME/machines if unset
    # vm_locale: This sets the default locale in the virtual machines. Default
    #            setting is inherited from current user.
    # vm_setkeyboard: If this is set to "true", keyboard configuration will
    #                 be setup in the virtual machines using the four following
    #                 options. Default is "true", with settings inherited from
    #                 host.
    # vm_image_size: Default size for virtual machine images (hard disk size).
    #                Default is 8GB
    # vm_memory: Default size for virtual machine RAM. A minimum of 384MB is
    #            needed for desktops, 256MB for servers. Default is 512MB.
    # vm_username: Username of the initial user that is set up in the virtual
    #              machines. Default is current user.
    # vm_password: Password of the initial user that is set up in the virtual
    #              machines. Default is hardcoded to "ubuntu".
    # vm_timezone: Timezone that is set up in the virtual machines. Default
    #              is inherited from the host.
    # vm_ssh_key: SSH public key to copy over to the virtual machines. This
    #             sets up SSH public key authentication to the VMs by default.
    #             Default is $HOME/.ssh/id_rsa.pub
    # vm_aptproxy: If set, this sets up an apt proxy in the virtual machine.
    #              No default setting. Can be used to set up apt-cacher-ng, for
    #              example.
    # vm_mirror: This is used as the default mirror in the sources.list file.
    #            Default is main archive.
    # vm_security_mirror: This is used as the default security mirror in the
    #                     sources.list file. Default is main security archive.
    # vm_src_mirror: This is used to override the archive used for source
    #                packages. By default, source packages are obtained from
    #                the archives set by vm_mirror and vm_security_mirror.
    # vm_mirror_host: FIXME: used with mini iso
    # vm_mirror_dir: FIXME: used with mini iso
    # vm_repo_url: This is used by the 'uvt repo' command to add or remove
    #              a local software repository to a VM. Defaults to a repo
    #              on the default libvirt host address:
    # vm_latecmd: This allows specifying an additional latecommand that is
    #             executed by the installer.
    # vm_extra_packages: A list of extra packages can be specified here to be
    #                    installed in the VM, for example, "screen". Empty by
    #                    default.
    vm_extra_packages="screen vim"
  8. Download the desktop CD images for each release and put them in the directory specified in the vm_dir_iso configuration option.

Virtual machines for testing

The security team should have at least one virtual machine per release and one for the development release. The recommended method is to use kvm with libvirt, which is what is documented here.

kvm on 64-bit will allow 32-bit OS guests, so if running a 64-bit host OS, then can have both i386 and amd64 installs. It's often useful to spend some time getting a pristine image exactly how you want it, then you clone/snapshot off of it for development work and/or testing.

WARNING: in this configuration you need to take special care to isolate the virtual machines from hostile network traffic, since anyone could login as root or your user to a VM. One way to do this is by using NAT with libvirt (the default).

Networking with libvirt

In order to perform name lookups for virtual machines, you must do one of the following changes:

Use the libvirt NSS module (Option 1 - preferred)

  1. Install the NSS module:

    $ sudo apt install libnss-libvirt
  2. Enable the NSS module by editing the /etc/nsswitch.conf and adding libvirt to the hosts line:

    hosts:          files mdns4_minimal [NOTFOUND=return] libvirt dns myhostname

Disable dnsmasq (Option 2)

  1. Put a line into /etc/dhcp/dhclient.conf like so:

    prepend domain-name-servers;
  2. Disable the system dnsmasq to prevent it from looping with libvirt's dnsmasq by modifying /etc/NetworkManager/NetworkManager.conf to comment out the following line:


Forward non-qualified lookups (Option 3)

  1. Configure the Network Manager dnsmasq to send non-qualified lookups to the libvirt dnsmasq by creating a /etc/NetworkManager/dnsmasq.d/libvirt file containing:

  2. Configure the libvirt dnsmasq to not forward non-qualified lookups by modifying /etc/libvirt/qemu/networks/default.xml to add the following to the <network> section:

      <dns forwardPlainNames="no">

Tell systemd-resolved to use libvirt's dnsmasq for VMs only (17.04+) (Option 4)

Ubuntu 16.10 uses systemd-resolved by default and 17.04 enables the stub resolver. With this configuration you can no longer put libvirt's dnsmasq (eg, 'nameserver') in /etc/resolv.conf or otherwise pass the server in as a nameserver because it creates a loop with the systemd-resolve stub resolver. Instead:

  1. adjust the libvirt network xml to put all VMs into their own domain (eg, 'vm') and to not forward to host:

    $ virsh net-edit default
      <domain name='vm' localOnly='yes'/>
    $ virsh net-destroy default
    $ virsh net-start default
  2. adjust systemd-resolved to use for queries on virbr0 with the 'vm' search domain:

    $ systemd-resolve --status virbr0
    Link 11 (virbr0)
          Current Scopes: DNS LLMNR/IPv4
           LLMNR setting: yes
    MulticastDNS setting: no
          DNSSEC setting: allow-downgrade
        DNSSEC supported: no
    # set virbr0 (systemd-resolved link '11') DNS to use
    $ sudo gdbus call --system --dest=org.freedesktop.resolve1 --object-path=/org/freedesktop/resolve1 --method=org.freedesktop.resolve1.Manager.SetLinkDNS 11 '[(2, [byte 0xc0, 0xa8, 0x7a, 0x01])]'
    # set virbr0 (systemd-resolved link '11') search domains to 'vm'
    $ sudo gdbus call --system --dest=org.freedesktop.resolve1 --object-path=/org/freedesktop/resolve1 --method=org.freedesktop.resolve1.Manager.SetLinkDomains 11 '[("vm", true)]'
    $ systemd-resolve --status virbr0
    Link 11 (virbr0)
          Current Scopes: DNS LLMNR/IPv4
           LLMNR setting: yes
    MulticastDNS setting: no
          DNSSEC setting: allow-downgrade
        DNSSEC supported: no
             DNS Servers:
              DNS Domain: ~vm

    Unfortunately the above is not preserved on reboot. Ideally the above would happen on boot or whenever the 'default' libvirt network is brought up. TODO: figure out the best place to put this. For now, this can be added to your ~/.bashrc so you can simply run update_resolved_for_libvirt as needed:

    update_resolved_for_libvirt() {
        echo "Was:"
        systemd-resolve --status "$libvirt_iface"
        iface_idx=`systemd-resolve --status "$libvirt_iface" | grep '(virbr0)' | cut -f 2 -d ' '`
        if [ -z "$iface_idx" ]; then
            echo "Could not find interface with 'systemd-resolve --status'"
        # set virbr0 DNS to use
        sudo gdbus call --system --dest=org.freedesktop.resolve1 --object-path=/org/freedesktop/resolve1 --method=org.freedesktop.resolve1.Manager.SetLinkDNS "$iface_idx" '[(2, [byte 0xc0, 0xa8, 0x7a, 0x01])]'
        # set virbr0 search domains to 'vm'
        sudo gdbus call --system --dest=org.freedesktop.resolve1 --object-path=/org/freedesktop/resolve1 --method=org.freedesktop.resolve1.Manager.SetLinkDomains "$iface_idx" '[("vm", true)]'
        # show the configuration
        echo "Now:"
        systemd-resolve --status "$libvirt_iface"
        echo "(requires '<domain name=\"vm\" localOnly=\"yes\"/>' in libvirt net xml"
        echo "and removed from /etc/resolv.conf)"

Skip DNS resolution entirely (Option 5)

Skipping DNS resolution entirely is also possible by adjusting ~/.ssh/config to have:

Host sec-*-amd64 sec-*-i386
#    StrictHostKeyChecking no
#    UserKnownHostsFile /dev/null
    ProxyCommand nc $(virsh domifaddr $(echo %h | sed "s/\.vm//g") | awk -F'[ /]+' '{if (NR>2 && $5) print $5}'|tail -1) %p
    # terminate the connection after the VM is shutdown

After one of the above options is chosen, you then will be able to ssh into the machines with:

  • 16.10 and lower or if adjusting ~/.ssh/config:

    $ ssh sec-lucid-amd64


    $ ssh sec-lucid-amd64.
    Notice the '.' at the end of the second command. This may be needed due to a bug in dnsmasq when using NAT with some versions of Ubuntu.
  • 17.04 and higher:

    $ ssh sec-lucid-amd64.vm

If avahi is installed in the guest, you can also use:

$ ssh sec-lucid-amd64.local

IMPORTANT: When using the tools, keep in mind that you may connect to the same VM with different hostnames. Eg, you could connect to the sec-lucid-amd64 VM as 'sec-lucid-amd64.', 'sec-lucid-amd64.local' or 'sec-lucid-amd64'. uvt tests if the VM is up by testing hostnames in this order and it is possible for the first to fail and the second to succeed. Therefore, you should login via ssh to at least the first two (if not all three), so that you have the host keys for the host. You may be having this problem if you see something like:

----- sec-maverick-amd64 -----
Starting sec-maverick-amd64 (snapshotted) ...
Waiting for 'sec-maverick-amd64' to come up host is up
Command: ssh -t -o BatchMode=yes -l root sec-maverick-amd64.local "ls /tmp"
Host key verification failed.

In this case, the ssh host key was in ~/.ssh/known_hosts for 'sec-maverick-amd64.', but not for 'sec-maverick-amd64.local'.

mDNS (avahi) (Option 6)

vm-tools will by default try '.local' mDNS (avahi) addresses when connecting to VMs. This may not work in all environments and can be disabled by using the following in $HOME/.uqt-vm-tools.conf:

# Set to 'no' to disable '.local' mDNS (avahi) lookups for VMs

You'll know you are encountering this problem when host name lookups fail for <machine>.local. This can also be caused be restrictive firewalls. If you have ufw enabled, then instead of disabling avahi lookups you can do something similar to:

$ sudo ufw allow in on virbr0

Host upgrades to 19.04+ (osxsave dropped from qemu)

Old virtual machines built with uvt prior to October 2016 or any VM that was built using --cpu=host or specific CPU models may end up with the following in the domain XML:

<cpu mode='custom' match='exact' check='partial'>
  <feature policy='require' name='osxsave'/>

qemu in 19.04 dropped the osxsave feature so VMs with this cpu definition fail to start:

$ virsh start sec-trusty-i386
error: Failed to start domain sec-trusty-i386
error: internal error: process exited while connecting to monitor: 2019-04-17T14:10:37.222226Z qemu-system-x86_64: can't apply global Broadwell-noTSX-x86_64-cpu.osxsave=on: Property '.osxsave' not found

Simply use virsh edit <vm name> to remove <feature policy='require' name='osxsave'/>. If using snapshots, revert to a pristine snapshot, edit the xml, then snapshot the vm.

Snapshotted virtual machines

With this method, we have VMs which are initially created with a pristine libvirt snapshot. Eg, might have the following virtual machines:

  • sec-trusty-i386
  • sec-trusty-amd64
  • sec-xenial-i386
  • sec-xenial-amd64
  • sec-bionic-i386
  • sec-bionic-amd64
  • sec-disco-i386
  • sec-disco-amd64
  • sec-eoan-i386
  • sec-eoan-amd64

Create your VMs (best to do this sequentially, even though it takes a while):

. $HOME/.uqt-vm-tools.conf
for i in $vm_release_list ; do
    uvt new $i i386 sec
    uvt new $i amd64 sec

The basic idea is as follows:

  1. uvt new after installing the OS will create the pristine snapshot

  2. using uvt start <vm> starts the image normally, and uvt stop <vm> shuts it down cleanly. Changes to the VM are preserved across reboots

  3. using uvt start -r <vm> reverts all changes made to the VM since the last snapshot, then starts the VM in the pristine state. Note that if you know you are going to revert to the previous snapshot, you can use uvt stop -f <vm> which does an unclean shutdown akin to pulling the plug.

Typical uses:

  • revert to pristine snapshot and discard:

    $ uvt start -r <vm>
    ... do your stuff ...
    $ uvt stop -f <vm>
  • snapshot with persistence across stops:

    $ uvt start -r <vm>            # revert all changes and start with a clean slate
    ... do your stuff ...
    $ uvt stop <vm>                # no '-f' so a clean shutdown is performed
    $ uvt start <vm>               # notice no '-r', so changes are not reverted
    ... do more stuff ...
    $ uvt stop <vm>
    ... do even more stuff ...
    $ uvt stop -f <vm>             # done with work, so pull the plug for a quick shutdown (assumes -r on next start)

IMPORTANT: Changes made in a snapshot will be lost if you use '-r' with uvt start or otherwise remove the snapshots.

To update the pristine image and make a new snapshot:

$ uvt start -r <vm>            # revert all changes and start with a clean slate
... make changes to the VM ...
$ uvt stop <vm>                # cleanly shut it down
$ uvt snapshot <vm>            # update the pristine snapshot

As a convenience, you can perform package upgrades using:

$ uvt update --autoremove <vm> # reverts to pristine snapshot, starts the VM, dist-upgrades, cleans up, then updates the pristine snapshot

NOTE: as of 2017/09/20 if adjusting ~/.ssh/config to skip DNS resolution, invoke uvt with UVT_USE_DOMIFADDR=1 uvt ...

Batch commands

If using dnsmasq as above, you can also use uvt cmd to do batch commands for the virtual machines, like so:

$ uvt cmd -p sec 'uname -a'
$ uvt cmd -r -p sec "apt-get update && apt-get -y upgrade"

uvt cmd uses 'release_list' in $HOME/.uqt-vm-tools.conf and will ssh in to all running sec-*-* machines and perform the specified command. Specifying -r to uvt cmd will login to the machine and run the command as root, otherwise it runs as non-root (ie your username in the guest).

Other useful commands:

  • uvt start: start a single VM of a group of VMs. Eg:

    $ uvt start -r sec-xenial-amd64           # start a single VM, reverting to the last snapshot
    $ uvt start -p sec -a i386               # start all i386 VMs starting with 'sec'
    $ uvt start -v -p sec -a i386            # start all i386 VMs starting with 'sec' without virt-viewer
  • uvt stop: stop a single running VM or a group of running VMs. Eg:

    $ uvt stop sec-xenial-amd64               # stop a single VM via ACPI
    $ uvt stop -r sec-xenial-amd64            # stop a single VM via ACPI, and revert to pristine snapshot
    $ uvt stop -f -p sec                     # hard stop all VMs starting with 'sec'
  • uvt snapshot: snapshot a single running VM or a group of running VMs. Eg:

    $ uvt snapshot sec-xenial-amd64           # snapshot a single VM
    $ uvt snapshot -p sec                    # snapshot all VMs starting with 'sec'
  • uvt revert: reverts a single running VM or a group of running VMs to pristine snapshot. Eg:

    $ uvt revert sec-xenial-amd64             # snapshot a single VM
    $ uvt revert -p sec                      # snapshot all VMs starting with 'sec'
  • uvt update: update and snapshot a single running VM or a group of running VMs. Eg:

    $ uvt update sec-xenial-amd64            # dist-upgrade and snapshot a single VM
    $ uvt update -p sec                     # dist-upgrade and snapshot all VMs starting with 'sec'
  • uvt remove: remove a single VM or a group of VMs. Eg:

    $ uvt remove sec-xenial-amd64             # delete a single VM
    $ uvt remove -p sec                      # delete all VMs starting with 'sec'
  • uvt repo: toggle the local repo (eg where 'umt repo' puts stuff) for a single running VM or a group of running VMs. Eg:

    $ uvt repo -e sec-xenial-amd64            # enable the local repo for a single VM
    $ uvt repo -d sec-xenial-amd64            # disable the local repo for a single VM
    $ uvt repo -e -p sec                     # enable the local repo for all running VMs starting with 'sec'
  • uvt view: connect to the VNC console of a single VM or a group of VMs using virt-viewer. Eg:

    $ uvt view sec-xenial-amd64
    $ uvt view -p sec

Setting up an ESM VM

Extra steps are required for setting up a VM for Ubuntu 12.04 ESM testing.

  1. Create the VM as described above. ESM is targetted for amd64 server environments so you can simply create an amd64 server VM:

    $ uvt new -t server precise amd64 sec
  2. Start the VM and log in:

    $ uvt start -v sec-precise-amd64
    $ ssh sec-precise-amd64
  3. Install the apt-transport-https package that's required to communicate with a private Launchpad PPA:

    $ sudo apt-get install apt-transport-https
  4. Take note of the required information for accessing the PPA
    1. Adjust https://launchpad.net/~<LP_USER>/+archivesubscriptions with your Launchpad username and go to the page

    2. Find the Extended Security Maintenance (ppa:ubuntu-esm/esm) row and click the "view" link to the right

    3. You'll need the apt sources lines as well as the key ID of the PPA archive key
  5. Restrict file access to the ESM apt sources list file prior to writing your credentials to the file
    • $ sudo touch /etc/apt/sources.list.d/esm-ppa.list
      $ sudo chown root:sudo /etc/apt/sources.list.d/esm-ppa.list
      $ sudo chmod 640 /etc/apt/sources.list.d/esm-ppa.list
  6. Add the apt sources lines to a new sources file (replace the echo'ed strings with your personalized apt sources lines):

    $ echo "deb ..." | sudo tee -a /etc/apt/sources.list.d/esm-ppa.list
  7. Import the PPA key:

    $ sudo apt-key adv --recv-key --keyserver keyserver.ubuntu.com <PPA_ARCHIVE_KEY_ID>
  8. Synchronize the package index files and upgrade any outdated packages:

    $ sudo apt-get update
    $ sudo apt-get dist-upgrade
  9. Shut down the VM:

    $ sudo shutdown -h now
  10. In the host environment, snapshot the VM:

    $ uvt snapshot sec-precise-amd64

You can now use the precise-esm-amd64 chroot to prepare source packages and perform local test builds using UMT, as documented below.



If you use uvt repo you may notice that while your apt archive is signed, the VM doesn't know about your key. This can be solved with:

$ gpg --armor --export <your key id> | ssh root@<vm> sudo apt-key add - # for deb
$ gpg --armor --export <your key id> | ssh <vm> gpg --no-default-keyring --keyring ~/.gnupg/trustedkeys.gpg --import - # for deb-src


Many VM operations can be quite slow on spinning-metal hard drives; be sure your ssh key is added to your ssh-agent for a long enough time to handle all operations. Consider adding the key without a timeout:

$ ssh-add

Migrate existing VMs into uvt

You can migrate VMs to use uvt in the following manner:

  1. copy the disk image into vm_path (by default, $HOME/machines) as <domain>.qcow2. Eg, for a machine named 'my-vm':

    $ cp <path to>/my-vm/disk0.qcow2 $HOME/machines/my-vm.qcow2
    • uvt only supports qcow2 images (due to its use of snapshots), so if the VM uses a raw image, do something like:

      $ qemu-img convert -f raw <path to>/my-vm/disk0.img -O qcow2 $HOME/machines/my-vm.qcow2
  2. update the libvirt XML to use the new path to the disk in vm_path (eg, virsh edit my-vm). Eg:

        <disk type='file' device='disk'>
          <driver name='qemu' type='qcow2'/>
          <source file='<vm_path>/my-vm.qcow2'/>
  3. create the pristine snapshot:

    $ uvt snapshot my-vm
    • You can verify that the snapshot was created with virsh snapshot-list <domain>. Eg:

      $ virsh snapshot-list my-vm
       Name                 Creation Time             State
       pristine             2012-10-09 11:09:27 -0500 shutoff

Importing qcow2 images

Similar to the above, you can import cloud images (eg, those used by autopkgtest) like so:

$ autopkgtest-buildvm-ubuntu-cloud -r bionic
$ mv ./autopkgtest-bionic-amd64.img test-bionic-amd64.qcow2 # cloud images are qcow2
$ virsh dumpxml <some existing vm, eg, sec-bionic-amd64> > ./xml
... edit xml to remove `<uuid>` and `<mac address='...'>`, to update `<name>` (eg, 'test-bionic-amd64') and the path to the disk (eg, test-bionic-amd64.qcow2) ...
$ virsh define ./xml
$ uvt start <name>
... assuming it boots, login, then shutdown ...
$ uvt snapshot <name>
  • on systems with netplan (eg, 18.04+), be sure to edit /etc/netplan/50-cloud-init.yaml so that it has the expected interface name and MAC addresss (both seen with ip addr). Eg:

    $ cat /etc/netplan/50-cloud-init.yaml 
                dhcp4: true
                    macaddress: '52:54:00:5A:57:24'
                set-name: ens3
        version: 2
    $ sudo netplan generate
    $ sudo netplan apply
    • adjust /etc/hostname and anything else to taste, then shutdown and rerun uvt snapshot <name>

Reclaim qcow2 space

You can compact the qcow2 images periodically to save hard drive space; on a dozen VMs using 89 gigabytes of disk space, the following was able to shrink them 45% to only 61 gigabytes of disk space:

for f in sec-{lucid,precise,quantal,saucy,trusty}-{amd64,i386} ; do echo $f ; qemu-img convert -s pristine -p -f qcow2 -O qcow2 $f.qcow2 reclaimed.qcow2 ; mv reclaimed.qcow2 $f.qcow2 ; virsh snapshot-delete $f --snapshotname pristine ; uvt snapshot $f ; done

SecurityTeam/TestingEnvironment (last edited 2021-10-19 19:59:16 by sbeattie)