README for OpenBSD users
------------------------
==> Quick Start
NOTE: amd64 cannot use '-net user' without SEGV'ing, to work around
this, either run qemu as root by replacing
qemu -m 32 ..
in this quick start section with
sudo env ETHER=em0 qemu -net nic,model=rtl8139 -net tap -m 32 ..
or by reading the 'tap mode' section below.
1. Get a bootable CDROM image
$ ftp ftp://ftp.openbsd.org/pub/OpenBSD/snapshots/i386/cd47.iso
2. Create a virtual disk image:
$ qemu-img create -f qcow2 virtual.img 10G
3. Install the os:
$ qemu -m 32 -monitor stdio -no-fd-bootchk -hda virtual.img \
-cdrom cd47.iso -boot d
NOTE: start this inside an xterm or equivalent
NOTE: be sure to choose serial console during install
NOTE: -no-fd-bootchk permits booting faster when no floppy is in use
4. Compress the virtual disk:
$ qemu-img convert -c -O qcow2 virtual.img v.tmp && mv v.tmp virtual.img
NOTE: do not do this while qemu is running / using this virtual disk
5. Boot normally from the virtual disk:
$ qemu -m 32 -nographic -no-fd-bootchk -hda virtual.img
==> Networking
1. Default Settings
By default, qemu sets up the equivalent of the following networking:
-net nic,vlan=0,model=rtl8139,macaddr=52:54:00:12:34:56
-net user,vlan=0
Also, inside this virtual usermode network, it uses the 10.0.2.0/24 and
serves dhcp from inside this virtual network. Static address can be used
if one cannot or does not want to do dhcp in the guest os:
Guest OS IP : 10.0.2.15
Default Gateway : 10.0.2.2
Nameserver : 10.0.2.3
It is sufficient for most operations, qemu itself performs NAT and then
makes userland network calls for tcp/udp operations. icmp and other things
are not possible in this mode.
NOTE: If you use one '-net' cmdline argument, qemu assumes you know what you
want and clears defaults for the rest of the -net defaults.
NOTE: The guest mode networking does not currently support IPv6, and
qemu will complain that it cannot find a dns server if
/etc/resolv.conf contains only IPv6 dns servers.
2. tap mode
Sometimes it is desirable to configure qemu to access a network via layer2
directly. One way of doing this without having to run qemu as root is to
let root open /dev/tunN and pass the file descriptor to qemu. The tun(4)
interface should preferrably be configured before starting qemu:
$ sudo ifconfig tun0 192.168.0.254 link0
The interface can also be configured as part of a bridge(4), in which case
the ip address can be omitted:
$ sudo ifconfig tun0 link0
$ sudo ifconfig bridge0 create
$ sudo ifconfig bridge0 add tun0 add em0 up
The tunnel and bridge interfaces can also be configured at system startup by
editing /etc/hostname.tunN and /etc/bridgename.bridgeN,
respectively (see hostname.if(5)).
After configuring the virtual network we can use sudo to let root open the
tunnel device and then use sudo again to drop privileges and start qemu:
$ sudo sh -c "sudo -u $USER qemu -nographic -net nic -net tap,fd=3 \
-no-fd-bootchk -hda virtual.img 3<>/dev/tun0"
NOTE: sudo calls closefrom(2). In order to have more than one fd passed
tap interface, a line to sudoers akin to:
Defaults closefrom_override
then calling sudo via 'sudo -C 5 -u $USER qemu ..' is required.
See sudoers(5) and sudo(8) for details.
An alternative to the procedure described above is to have qemu set up the
network via ${SYSCONFDIR}/qemu-ifup. This is not recommended however, since
you would have to run qemu as root, and there is no way to drop from root
privileges at this point.
${SYSCONFDIR}/qemu-ifup contains some default settings that permit one to do
the following:
# qemu -net nic -net tap -no-fd-bootchk -hda virtual.img
It presumes you wish the tun(4) interface to talk to the interface
holding the default IPv4 route (falling back to trunk0 if no route is
found), and that you want 'bridge0' to be used to bridge the two.
If you wish to over-ride these settings, setting the environment variables
ETHER and BRIDGE will over-ride these settings, respectively.
When starting qemu, the script attempts to output useful information, but
there are also error messages that occur as well. On my laptop, I want to
route / nat natively using pf and also have layer2 access to the qemu
networks. I thus have this as /etc/hostname.trunk101:
inet6 fe80::1c 64 lladdr 00:03:25:0d:7a:2c
inet 10.7.255.1 255.255.255.0
inet6 alias 2001:240:58a:45::1c
I have dhcpd configured to run on trunk101, and also run rtadvd. For qemu,
the startup looks like this:
# export ETHER=trunk101
# export BRIDGE=bridge101
# qemu -net nic,vlan=0,model=rtl8139,macaddr=52:54:00:12:35:00 \
-net tap,vlan=0 -vnc :0 -localtime -usb -usbdevice tablet \
-m 256 -no-fd-bootchk -hda virtual.img -monitor stdio
{tun0 (bridge101 <-> trunk101)ifconfig: bridge101: No such process
ifconfig: bridge101: No such process
}
(qemu)
The errors are normal and should be ignored. One can verify the networking
is properly configured by verifying the bridge interface:
$ ifconfig bridge101
bridge101: flags=41<UP,RUNNING>
priority 32768 hellotime 2 fwddelay 15 maxage 20 holdcnt 6 proto rstp
designated: id 00:00:00:00:00:00 priority 0
tun0 flags=3<LEARNING,DISCOVER>
port 16 ifpriority 0 ifcost 0
trunk101 flags=3<LEARNING,DISCOVER>
port 6 ifpriority 0 ifcost 0
Addresses (max cache: 100, timeout: 240):
NOTE: When running multiple qemu sessions simultaneously on the same bridge,
care must be taken because the network mac address defaults to
52:54:00:12:34:56 for every qemu instance. To change this, observe
the macaddr= syntax in the above example and choose a unique lladdr
per qemu nic.
==> Mice
NOTE: Certain OS's work much better with the tablet usb device than
the normal ps2 mouse handling. See the above example for usage.
==> Serial Console
1. Installing OpenBSD via serial console is sometimes desirable. X may not
be available, and so on. There are two ways to accomplish this, both in
effect the same solution:
a. qemu -vnc :0 -serial stdio .. virtual.img -cdrom install43.iso -boot d
- this option permits you to use vnc from some system to connect to
the qemu instance and 'set tty com0' at the 'boot>' prompt.
- you may then disconnect vnc and use the terminal from which you
started qemu to do the install
b. qemu -nographic .. virtual.img -fda floppy47.fs -boot a
- this maps both the serial port and the (qemu) monitor prompt to
the terminal qemu was started on
- to flip between them, Ctrl-a c; see the qemu man page for other
commands that work in -nographic mode.
- preparation of the floppy image to force serial console mode is
straightforward:
vnconfig svnd0 floppy47.fs
mount /dev/svnd0c /mnt
mkdir /mnt/etc
echo set tty com0 > /mnt/etc/boot.conf
umount /mnt
vnconfig -u svnd0
.. be sure to choose 'yes' for setting com0 to be the serial console.
NOTE: OpenBSD poweroff does work with qemu, which actually causes qemu
itself to exit. This is a good thing, as it is currently not
possible to set what block device is booted from at runtime from
qemu. So if you start an installation booting from a cdrom, you
will always boot off a cdrom every time you reboot that qemu session
until you exit and start qemu again booting off the virtual hard
drive.
==> daemonized qemu
Sometimes you want qemu to start as part of a system script.
Adding to some of the above, the -daemonize option comes in handy,
as well as the telnet: designator for -serial and monitor. This
is a complete example that may be cut-and-pasted into rc.local:
hddir=/var/vm
USER=qemu
if [ -x ${TRUEPREFIX}/bin/qemu ]; then
echo -n 'Qemu: vmi386'
(
ifconfig tun0 link0
ifconfig bridge101 create
ifconfig bridge101 add trunk101 add tun0 up
sh -c "sudo -u $USER \
${TRUEPREFIX}/bin/qemu \
-daemonize \
-nographic \
-net nic,vlan=0,model=rtl8139,macaddr=52:54:00:4e:62:8f \
-net tap,vlan=0,fd=3 \
-m 128 \
-hda $hddir/virtual.img \
-serial telnet:127.0.0.1:1080,server,nowait \
-monitor telnet:127.0.0.1:1081,server,nowait \
-no-fd-bootchk 3<>/dev/tun0"
)
echo "."
fi
NOTE: this presumes the user `qemu' exists, create it or set USER= to an
existing user to use this example.
