Installation of qemu
As the root user, add any users
that might use the KVM device to that group:
usermod -a -G kvm <username>
Install qemu by running the
following commands:
![[Note]](../images/note.png)
Note
Qemu is capable of running many targets. The build process is
also capable of building multiple targets at one time in a comma
delimited list assigned to --target-list. Run ./configure --help to get a
complete list of available targets.
if [ $(uname -m) = i686 ]; then
QEMU_ARCH=i386-softmmu
else
QEMU_ARCH=x86_64-softmmu
fi
mkdir -vp build &&
cd build &&
../configure --prefix=/usr \
--sysconfdir=/etc \
--localstatedir=/var \
--target-list=$QEMU_ARCH \
--audio-drv-list=alsa \
--disable-pa \
--docdir=/usr/share/doc/qemu-7.2.0 &&
unset QEMU_ARCH &&
make
qemu uses ninja as a subprocess when
building. To run the tests, issue: ninja test.
Now, as the root user:
make install
You will also need to add an Udev rule so that the KVM device gets
correct permissions. As the root
user, issue:
cat > /lib/udev/rules.d/65-kvm.rules << "EOF"
KERNEL=="kvm", GROUP="kvm", MODE="0660"
EOF
Change the permissions and ownership of a helper script, which is
needed when using the “bridge” network device (see below).
Again as the root user, issue:
chgrp kvm /usr/libexec/qemu-bridge-helper &&
chmod 4750 /usr/libexec/qemu-bridge-helper
Note
For convenience you may want to create a symbolic link to run the
installed program. For instance (as the root user):
ln -sv qemu-system-`uname -m` /usr/bin/qemu
Using Qemu
Since using qemu means using a virtual computer, the steps to set
up the virtual machine are in close analogy with those to set up a
real computer. You'll need to decide about CPU, memory, disk, USB
devices, network card(s), screen size, etc. Once the “hardware” is
decided, you'll have for example to choose how to connect the
machine to internet, and/or to install an OS. In the following, we
show basic ways of performing those steps. But qemu is much more
than this, and it is strongly advised to read the qemu
documentation in /usr/share/doc/qemu-7.2.0/qemu-doc.html.
Note
It is standard practice to name the computer running qemu
“host” and the emulated machine
running under qemu the “guest”. We'll use those notations in
the following.
Note
The following instructions assume the optional symbolic link,
qemu, has been created.
Additionally, qemu
should be run in a graphical environment. But it is possible to
use qemu “headless” or through SSH. See the
documentation for the various possibilities.
Disk
A virtual disk may be set up in the following way:
VDISK_SIZE=50G
VDISK_FILENAME=vdisk.img
qemu-img create -f qcow2 $VDISK_FILENAME $VDISK_SIZE
The virtual disk size and filename should be adjusted as desired.
The actual size of the file will be less than specified, but will
expand as needed, so it is safe to put a high value.
Operating System
To install an operating system, download an iso image from your
preferred Linux distribution. For the purposes of this example,
we'll use Fedora-16-x86_64-Live-LXDE.iso in the current
directory. Run the following:
qemu -enable-kvm \
-drive file=$VDISK_FILENAME \
-cdrom Fedora-16-x86_64-Live-LXDE.iso \
-boot d \
-m 1G
Follow the normal installation procedures for the chosen
distribution. The -boot
option specifies the boot order of drives as a string of drive
letters. Valid drive letters are: a, b (floppy 1 and 2), c (first
hard disk), d (first CD-ROM). The -m option is the amount of memory to
use for the virtual machine. The choice depends on the load of the
host. Modern distributions should be comfortable with 1GB. The
-enable-kvm option allows
hardware acceleration. Without this switch, the emulation is much
slower.
Defining the virtual hardware
The virtual machine hardware is defined by the qemu command line.
An example command is given below:
qemu -enable-kvm \
-smp 4 \
-cpu host \
-m 1G \
-drive file=$VDISK_FILENAME \
-cdrom grub-img.iso \
-boot order=c,once=d,menu=on \
-net nic,netdev=net0 \
-netdev user,id=net0 \
-device ac97 \
-vga std \
-serial mon:stdio \
-name "fedora-16"
Meaning of the command line options
-enable-kvm: enable full
KVM virtualization support. On some hardware, it may be necessary
to add the undocumented -machine
smm=off option in order to enable KVM.
-smp <N>: enable
symmetric multiprocessing with <N> CPUs.
-cpu <model>:
simulate CPU <model>. the list of supported models can be
obtained with -cpu help.
-drive
file=<filename>: defines a virtual disk whose
image is stored in <filename>.
-cdrom grub-img.iso:
defines an iso formatted file to use as a cdrom. Here we use a grub
rescue disk, which may turn handy when something goes wrong at boot
time.
-boot
order=c,once=d,menu=on: defines the boot order for the
virtual BIOS.
-net
nic,netdev=<netid>: defines a network card
connected to the network device with id <netid>.
-netdev
user,id=<netid>: defines the network “user”
device. This is a virtual local network with addresses 10.0.2.0/24,
where the host has address 10.0.2.2 and acts as a gateway to
internet, and with a name server at address 10.0.2.3, and an smb
server at address 10.0.2.4. A builtin DHCP server can allocate
addresses between 10.0.2.15 and 10.0.2.31.
-soundhw <model>:
defines the soundcard model. The list may be obtained with
-soundhw help.
-vga <type>: defines
the type of VGA card to emulate. For -vga std, if you are building a Linux
kernel for the guest, it's recommended to enable CONFIG_DRM_BOCHS (as a part of the kernel or a
kernel module) to drive all the features of the emulated VGA card,
and CONFIG_FB to display the Linux
console on it. The other <type>
values are not tested by the editors and may require additional
dependencies.
-serial mon:stdio: sends
the serial port of the guest (/dev/ttyS0 on linux guests), multiplexed with the
qemu monitor, to the standard input and output of the qemu process.
-name <name>: sets
the name of the guest. This name is displayed in the guest window
caption. It may be useful if you run several guests at the same
time.
-pflash
/usr/share/qemu/edk2-x86_64-code.fd: Load a pre-built EDK2
UEFI firmware, instead of the default PC BIOS. Use this option if
you want to boot the guest OS with UEFI.
-drive file=<filename>,if=virtio:
Provide Virtio interface to the guest kernel for accessing the disk
image, instead of simulating a real disk hardware. This can improve
disk I/O performance, but it requires a Virtio driver in guest
kernel. Use it instead of a plain -drive if the guest kernel supports Virtio. To
build a Linux kernel with Virtio support for the guest, use
make defconfig && make
kvm_guest.config to create an initial kernel
configuration with the Virtio drives enabled, then make your
customization. And, if the guest kernel is Linux, the virtual disks
using Virtio interface will be named vdx in the devtmpfs, instead of sdx.
-net
nic,netdev=net0,model=virtio-net-pci: Provide Virtio
interface to the guest kernel for accessing the network interface,
instead of simulating a real network interface card. This can
improve network I/O performance, but it requires a Virtio driver in
guest kernel. Use it instead of a plain -net if the guest kernel supports Virtio.
Controlling the Emulated Display
It may happen that the guest window displayed by qemu does not
correspond to the full capability of the emulated VGA card. For
example, the vmware card is 1600x900 capable, but only 1024x768 is
displayed by default. A suitable Xorg configuration on the guest
allows to use the full size (Note that the Xorg video driver to use
is Xorg VMware Driver-13.4.0):
cat > /usr/share/X11/xorg.conf.d/20-vmware.conf << "EOF"
Section "Monitor"
Identifier "Monitor0"
# cvt 1600 900
# 1600x900 59.95 Hz (CVT 1.44M9) hsync: 55.99 kHz; pclk: 118.25 MHz
Modeline "1600x900" 118.25 1600 1696 1856 2112 900 903 908 934 -hsync +vsync
Option "PreferredMode" "1600x900"
HorizSync 1-200
VertRefresh 1-200
EndSection
Section "Device"
Identifier "VMware SVGA II Adapter"
Option "Monitor" "default"
Driver "vmware"
EndSection
Section "Screen"
Identifier "Default Screen"
Device "VMware SVGA II Adapter"
Monitor "Monitor0"
SubSection "Display"
Depth 24
Modes "1600x900" "1440x900" "1366x768" "1280x720" "800x480"
EndSubSection
EndSection
EOF
New sizes will be available besides the native ones. You need to
restart X in order to have the new sizes available.
Networking
The above solution for networking allows the guest to access the
local network through the host (and possibly to access internet
through the local routers), but the converse is not true. Not even
the host can access the guest, unless port forwarding is enabled.
And in the case several guests are running, they cannot communicate
with each other. Other network devices can be used for this
purpose. For example, there is the “socket”
device, which allows several guests to share a common virtual
network. In the following, we describe in more details how to set
up the “bridge” device, which allows the guests
to appear as if connected to the local network. All the commands
below should be run as the root
user.
Allow the host to forward IP packets:
sysctl -w net.ipv4.ip_forward=1
To make this permanent, add the command to /etc/sysctl.d/60-net-forward.conf:
cat >> /etc/sysctl.d/60-net-forward.conf << EOF
net.ipv4.ip_forward=1
EOF
Set up a required configuration file:
install -vdm 755 /etc/qemu &&
echo allow br0 > /etc/qemu/bridge.conf
In the qemu command line above, replace the switch -netdev user,... with -netdev bridge,....