Xen

From ArchWiki
Jump to: navigation, search

From Xen Overview:

Xen is an open-source type-1 or baremetal hypervisor, which makes it possible to run many instances of an operating system or indeed different operating systems in parallel on a single machine (or host). Xen is the only type-1 hypervisor that is available as open source. Xen is used as the basis for a number of different commercial and open source applications, such as: server virtualization, Infrastructure as a Service (IaaS), desktop virtualization, security applications, embedded and hardware appliances.
Warning: Do not run other virtualization software such as VirtualBox when running Xen hypervisor, it might hang your system. See this bug report (wontfix).

Introduction

The Xen hypervisor is a thin layer of software which emulates a computer architecture allowing multiple operating systems to run simultaneously. The hypervisor is started by the boot loader of the computer it is installed on. Once the hypervisor is loaded, it starts the dom0 (short for "domain 0", sometimes called the host or privileged domain) which in our case runs Arch Linux. Once the dom0 has started, one or more domU (short for user domains, sometimes called VMs or guests) can be started and controlled from the dom0. Xen supports both paravirtualized (PV) and hardware virtualized (HVM) domU. See Xen.org for a full overview.

System requirements

The Xen hypervisor requires kernel level support which is included in recent Linux kernels and is built into the linux and linux-lts Arch kernel packages. To run HVM domU, the physical hardware must have either Intel VT-x or AMD-V (SVM) virtualization support. In order to verify this, run the following command when the Xen hypervisor is not running:

$ grep -E "(vmx|svm)" --color=always /proc/cpuinfo

If the above command does not produce output, then hardware virtualization support is unavailable and your hardware is unable to run HVM domU (or you are already running the Xen hypervisor). If you believe the CPU supports one of these features you should access the host system's BIOS configuration menu during the boot process and look if options related to virtualization support have been disabled. If such an option exists and is disabled, then enable it, boot the system and repeat the above command. The Xen hypervisor also supports PCI passthrough where PCI devices can be passed directly to the domU even in the absence of dom0 support for the device. In order to use PCI passthrough, the CPU must support IOMMU/VT-d.

Configuring dom0

The Xen hypervisor relies on a full install of the base operating system. Before attempting to install the Xen hypervisor, the host machine should have a fully operational and up-to-date install of Arch Linux. This installation can be a minimal install with only the base package and does not require a Desktop environment or even Xorg. If you are building a new host from scratch, see the Installation guide for instructions on installing Arch Linux. The following configuration steps are required to convert a standard installation into a working dom0 running on top of the Xen hypervisor:

  1. Installation of the Xen hypervisor
  2. Modification of the bootloader to boot the Xen hypervisor
  3. Creation of a network bridge
  4. Installation of Xen systemd services

Installation of the Xen hypervisor

To install the Xen hypervisor install either the current stable xenAUR or the bleeding edge unstable xen-gitAUR[broken link: archived in aur-mirror] packages available in the Arch User Repository. Both packages provide the Xen hypervisor, current xl interface and all configuration and support files, including systemd services. The multilib repository needs to be enabled and the multilib-devel package group installed to compile Xen. Install the xen-docsAUR package from the Arch User Repository for the man pages and documentation.

With UEFI support

It's possible to boot the Xen hypervisor though the bare UEFI system on a modern computer but requires you to first recompile binutils to add support for x86_64-pep emulation. Using the archway of doing things you would use the Arch Build System and add --enable-targets=x86_64-pep to the build options of the binutils PKGBUILD file:

--disable-werror --enable-targets=x86_64-pep
Note:

Tango-inaccurate.pngThe factual accuracy of this article or section is disputed.Tango-inaccurate.png

Reason: This Note is not very meaningful without a link to a bug report. (Discuss in Talk:Xen#)

This will not work on the newest version of binutils you will need to downgrade to an older version from the svn:

$ svn checkout --depth empty svn://svn.archlinux.org/packages
$ cd packages
$ svn update -r 215066 binutils
Then compile and install. See [1] for details of the procedure.

The next time binutils gets updated on your system it will be overwritten with the official version again. However, you only need this change to (re-)compile the UEFI aware Xen hypervisor, it is not needed at either boot or run time.

Now when you compile Xen with your x86_64-pep aware binutils a UEFI kernel will be built and installed by default. It is located at /usr/lib/efi/xen-?.?.?.efi where "?" represent the version digits. The other files you find that also begin with "xen" are simply symlinks back to the real file and can be ignored. However, the efi-binary needs to be manually copied to /boot, e.g.:

# cp /usr/lib/efi/xen-4.4.0.efi /boot

Modification of the bootloader

Tango-view-fullscreen.pngThis article or section needs expansion.Tango-view-fullscreen.png

Reason: Lots of other boot loaders could/should be covered, at least the most common like systemd-boot. (Discuss in Talk:Xen#)
Warning: Never assume your system will boot after changes to the boot system. This might be the most common error new as well as old users do. Make sure you have a alternative way to boot your system like a USB stick or other livemedia BEFORE you make changes to your boot system.

The boot loader must be modified to load a special Xen kernel (xen.gz or in the case of UEFI xen.efi) which is then used to boot the normal kernel. To do this a new bootloader entry is needed.

UEFI

There are several ways UEFI can be involved in booting Xen but this section will cover the most simple way to get Xen to boot with help of EFI-stub.

Make sure that you have compiled Xen with UEFI support enabled accoring to #With UEFI support.

It is possible to boot a kernel from UEFI just by placing it on the EFI partition, but since Xen at least needs to know what kernel should be booted as dom0, a minimum configuration file is required. Create or edit a /boot/xen.cfg file according to system requirements, for example:

/boot/xen.cfg
[global]
default=xen

[xen]
options=console=vga loglvl=all noreboot
kernel=vmlinuz-linux root=/dev/sda2 rw ignore_loglevel #earlyprintk=xen
ramdisk=initramfs-linux.img

It might be necessary to use efibootmgr to set boot order and other parameters. If booting fails, drop to the build-in UEFI shell and try to launch manually. For example:

Shell> fs0:
FS0:\> xen-4.4.0.efi

GRUB

For GRUB users, the Xen package provides the /etc/grub.d/09_xen generator file. The file /etc/xen/grub.conf can be edited to customize the Xen boot commands. For example, to allocate 512 MiB of RAM to dom0 at boot, modify /etc/xen/grub.conf by replacing the line:

#XEN_HYPERVISOR_CMDLINE="xsave=1"

with

XEN_HYPERVISOR_CMDLINE="dom0_mem=512M xsave=1"

After customizing the options, update the bootloader configuration with the following command:

# grub-mkconfig -o /boot/grub/grub.cfg

More information on using the GRUB bootloader is available at GRUB.

Syslinux

For syslinux users, add a stanza like this to your /boot/syslinux/syslinux.cfg:

LABEL xen
    MENU LABEL Xen
    KERNEL mboot.c32
    APPEND ../xen-X.Y.Z.gz --- ../vmlinuz-linux console=tty0 root=/dev/sdaX ro --- ../initramfs-linux.img

where X.Y.Z is your xen version and /dev/sdaX is your root partition.

This also requires mboot.c32 to be in the same directory as syslinux.cfg. If you do not have mboot.c32 in /boot/syslinux, copy it from:

# cp /usr/lib/syslinux/bios/mboot.c32 /boot/syslinux

Creation of a network bridge

Xen requires that network communications between domU and the dom0 (and beyond) be set up manually. The use of both DHCP and static addressing is possible, and the choice should be determined by the network topology. Complex setups are possible, see the Networking article on the Xen wiki for details and /etc/xen/scripts for scripts for various networking configurations. A basic bridged network, in which a virtual switch is created in dom0 that every domU is attached to, can be set up by creating a network bridge with the expected name xenbr0.

See Network bridge#Creating a bridge for details.

Creating bridge with Network Manager

Merge-arrows-2.pngThis article or section is a candidate for merging with Network_bridge#With_NetworkManager.Merge-arrows-2.png

Notes: Duplicates the main page. (Discuss in Talk:Xen#)

Gnome's Network Manager can sometime be troublesome. If following the bridge creation section outlined in the bridges section of the wiki are unclear or do not work, then the following steps may work.

Open the Network Settings and disable the interface you wish to use in your bridge (ex enp5s0). Edit the setting to off and uncheck "connect automatically."

Create a new bridge connection profile by clicking on the "+" symbol in the bottom left of the network settings. Optionally, run:

# nm-connection-editor

to bring up the window immediately. Once the window opens, select Bridge.

Click "Add" next to the "Bridged Connections" and select the interface you wished to use in your bridge (ex. Ethernet). Select the device mac address that corresponds to the interface you intend to use and save the settings

If your bridge is going to receive an IP address via DHCP, leave the IPv4/IPv6 sections as they are. If DHCP is not running for this particular connection, make sure to give your bridge an IP address. Needless to say, all connections will fail if an IP address is not assigned to the bridge. If you forget to add the IP address when you first create the bridge, it can always be edited later.

Now, as root, run:

# nmcli con show

You should see a connection that matches the name of the bridge you just created. Highlight and copy the UUID on that connection, and then run (again as root):

# nmcli con up <UUID OF CONNECTION>

A new connection should appear under the network settings. It may take 30 seconds to a minute. To confirm that it is up and running, run:

# brctl show

to show a list of active bridges.

Reboot. If everything works properly after a reboot (ie. bridge starts automatically), then you are all set.

<optional> In your network settings, remove the connection profile on your bridge interface that does NOT connect to the bridge. This just keeps things from being confusing later on.

Installation of Xen systemd services

The Xen dom0 requires the xenstored, xenconsoled, xendomains and xen-init-dom0 services to be started and possibly enabled.

Confirming successful installation

Reboot your dom0 host and ensure that the Xen kernel boots correctly and that all settings survive a reboot. A properly set up dom0 should report the following when you run xl list as root:

# xl list
Name                                        ID   Mem VCPUs	State	Time(s)
Domain-0                                     0   511     2     r-----   41652.9

Of course, the Mem, VCPUs and Time columns will be different depending on machine configuration and uptime. The important thing is that dom0 is listed.

In addition to the required steps above, see best practices for running Xen which includes information on allocating a fixed amount of memory and how to dedicate (pin) a CPU core for dom0 use. It also may be beneficial to create a xenfs filesystem mount point by including in /etc/fstab

none /proc/xen xenfs defaults 0 0

Using Xen

Xen supports both paravirtualized (PV) and hardware virtualized (HVM) domU. In the following sections the steps for creating HVM and PV domU running Arch Linux are described. In general, the steps for creating an HVM domU are independent of the domU OS and HVM domU support a wide range of operating systems including Microsoft Windows. To use HVM domU the dom0 hardware must have virtualization support. Paravirtualized domU do not require virtualization support, but instead require modifications to the guest operating system making the installation procedure different for each operating system (see the Guest Install page of the Xen wiki for links to instructions). Some operating systems (e.g., Microsoft Windows) cannot be installed as a PV domU. In general, HVM domU often run slower than PV domU since HVMs run on emulated hardware. While there are some common steps involved in setting up PV and HVM domU, the processes are substantially different. In both cases, for each domU, a "hard disk" will need to be created and a configuration file needs to be written. Additionally, for installation each domU will need access to a copy of the installation ISO stored on the dom0 (see the Download Page to obtain the Arch Linux ISO).

Create a domU "hard disk"

Xen supports a number of different types of "hard disks" including Logical Volumes, raw partitions, and image files. To create a sparse file, that will grow to a maximum of 10GiB, called domU.img, use:

$ truncate -s 10G domU.img

If file IO speed is of greater importance than domain portability, using Logical Volumes or raw partitions may be a better choice.

Xen may present any partition / disk available to the host machine to a domain as either a partition or disk. This means that, for example, an LVM partition on the host can appear as a hard drive (and hold multiple partitions) to a domain. Note that making sub-partitons on a partition will make accessing those partitions on the host machine more difficult. See the kpartx man page for information on how to map out partitions within a partition.

Create a domU configuration

Each domU requires a separate configuration file that is used to create the virtual machine. Full details about the configuration files can be found at the Xen Wiki or the xl.cfg man page. Both HVM and PV domU share some components of the configuration file. These include

name = "domU"
memory = 256
disk = [ "file:/path/to/ISO,sdb,r", "phy:/path/to/partition,sda1,w" ]
vif = [ 'mac=00:16:3e:XX:XX:XX,bridge=xenbr0' ]

The name= is the name by which the xl tools manage the domU and needs to be unique across all domU. The disk= includes information about both the the installation media (file:) and the partition created for the domU phy. If an image file is being used instead of a physical partition, the phy: needs to be changed to file:. The vif= defines a network controller. The 00:16:3e MAC block is reserved for Xen domains, so the last three digits of the mac= must be randomly filled in (hex values 0-9 and a-f only).

Managing a domU

If a domU should be started on boot, create a symlink to the configuration file in /etc/xen/auto and ensure the xendomains service is set up correctly. Some useful commands for managing domU are:

# xl top
# xl list
# xl console domUname
# xl shutdown domUname
# xl destroy domUname

Configuring a hardware virtualized (HVM) Arch domU

In order to use HVM domU install the mesa-libgl and bluez-libs packages.

A minimal configuration file for a HVM Arch domU is:

name = 'HVM_domU'
builder = 'hvm'
memory = 256
vcpus = 2
disk = [ 'phy:/dev/mapper/vg0-hvm_arch,xvda,w', 'file:/path/to/ISO,hdc:cdrom,r' ]
vif = [ 'mac=00:16:3e:00:00:00,bridge=xenbr0' ]
vnc = 1
vnclisten = '0.0.0.0'
vncdisplay = 1

Since HVM machines do not have a console, they can only be connected to via a vncviewer. The configuration file allows for unauthenticated remote access of the domU vncserver and is not suitable for unsecured networks. The vncserver will be available on port 590X, where X is the value of vncdisplay, of the dom0. The domU can be created with:

# xl create /path/to/config/file

and its status can be checked with

# xl list

Once the domU is created, connect to it via the vncserver and install Arch Linux as described in the Installation guide.

Configuring a paravirtualized (PV) Arch domU

A minimal configuration file for a PV Arch domU is:

name = "PV_domU"
kernel = "/mnt/arch/boot/x86_64/vmlinuz"
ramdisk = "/mnt/arch/boot/x86_64/archiso.img"
extra = "archisobasedir=arch archisolabel=ARCH_201301"
memory = 256
disk = [ "phy:/path/to/partition,sda1,w", "file:/path/to/ISO,sdb,r" ]
vif = [ 'mac=00:16:3e:XX:XX:XX,bridge=xenbr0' ]

This file needs to tweaked for your specific use. Most importantly, the archisolabel=ARCH_201301 line must be edited to use the release year/month of the ISO being used. If you want to install 32-bit Arch, change the kernel and ramdisk paths from x86_64 to i686.

Before creating the domU, the installation ISO must be loop-mounted. To do this, ensure the directory /mnt exists and is empty, then run the following command (being sure to fill in the correct ISO path):

# mount -o loop /path/to/iso /mnt

Once the ISO is mounted, the domU can be created with:

# xl create -c /path/to/config/file

The "-c" option will enter the domU's console when successfully created. Then you can install Arch Linux as described in the Installation guide, but with the following deviations. The block devices listed in the disks line of the cfg file will show up as /dev/xvd*. Use these devices when partitioning the domU. After installation and before the domU is rebooted, the xen-blkfront, xen-fbfront, xen-netfront, xen-kbdfront modules must be added to Mkinitcpio. Without these modules, the domU will not boot correctly. For booting, it is not necessary to install Grub. Xen has a Python-based grub emulator, so all that is needed to boot is a grub.cfg file: (It may be necessary to create the /boot/grub directory)

/boot/grub/grub.cfg
menuentry 'Arch GNU/Linux, with Linux core repo kernel' --class arch --class gnu-linux --class gnu --class os $menuentry_id_option 'gnulinux-core repo kernel-true-__UUID__' {
        insmod gzio
        insmod part_msdos
        insmod ext2
        set root='hd0,msdos1'
        if [ x$feature_platform_search_hint = xy ]; then
          search --no-floppy --fs-uuid --set=root --hint-bios=hd0,msdos1 --hint-efi=hd0,msdos1 --hint-baremetal=ahci0,msdos1  __UUID__
        else
          search --no-floppy --fs-uuid --set=root __UUID__
        fi
        echo    'Loading Linux core repo kernel ...'
        linux   /boot/vmlinuz-linux root=UUID=__UUID__ ro
        echo    'Loading initial ramdisk ...'
        initrd  /boot/initramfs-linux.img
}

This file must be edited to match the UUID of the root partition. From within the domU, run the following command:

# blkid

Replace all instances of __UUID__ with the real UUID of the root partition (the one that mounts as /).:

# sed -i 's/__UUID__/12345678-1234-1234-1234-123456789abcd/g' /boot/grub/grub.cfg

Shutdown the domU with the poweroff command. The console will be returned to the hypervisor when the domain is fully shut down, and the domain will no longer appear in the xl domains list. Now the ISO file may be unmounted:

# umount /mnt

The domU cfg file should now be edited. Delete the kernel =, ramdisk =, and extra = lines and replace them with the following line:

bootloader = "pygrub"

Also remove the ISO disk from the disk = line.

The Arch domU is now set up. It may be started with the same line as before:

# xl create -c /etc/xen/archdomu.cfg

Common Errors

"xl list" complains about libxl

Either you have not booted into the Xen system, or xen modules listed in xencommons script are not installed.

"xl create" fails

Check the guest's kernel is located correctly, check the pv-xxx.cfg file for spelling mistakes (like using initrd instead of ramdisk).

Arch Linux guest hangs with a ctrl-d message

Press ctrl-d until you get back to a prompt, rebuild its initramfs described

Error message "failed to execute '/usr/lib/udev/socket:/org/xen/xend/udev_event' 'socket:/org/xen/xend/udev_event': No such file or directory"

This is caused by /etc/udev/rules.d/xend.rules. Xend is deprecated and not used, so it is safe to remove that file.

Resources