QemuAsDevelTarget
Image(wiki:[[QemuAsDevelTarget:sae-qemu-devel.png)]]
apt-get install qemu
dd if=/dev/zero of=rootfs.img bs=1024 count=1048576
losetup /dev/loop0 rootfs.img
mkfs -t ext3 -m 1 -v /dev/loop0
mkdir /mnt/qemu-rootfs
mount -t ext3 /dev/loop0 /mnt/qemu-rootfs
Don't worry if you don't know what size to choose. If the suggested 1GB is not sufficient you can always ResizeYourQemuImage.
sudo debootstrap sid /mnt/qemu-rootfs http://ftp.debian.org/debian
3.1 Install useful packages
chroot /mnt/qemu-rootfs /bin/bash
apt-get install iw vim sudo make gcc sshfs libnl2 tcpdump
4.1 autologin as root via serial console (i.e. the terminal console from which you invoke qemu):
Remove password from root account:
/etc/shadow
- root:x:14483:0:99999:7:::
+ root::14483:0:99999:7:::
Enable serial console with auto login
/etc/inittab
- #T0:23:respawn:/sbin/getty -L ttyS0 9600 vt100
+ T0:23:respawn:/sbin/getty -n -l /bin/autologin.sh -L ttyS0 38400 linux
Create /bin/autologin.sh as:
#!/bin/sh
exec /bin/login -f root
And make it executable
chmod 700 /bin/autologin.sh
git clone \
git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless-testing
cd wireless-testing
make defconfig
# For mesh development you want to edit your .config and enable
# CONFIG_MAC80211_MESH=y
# CONFIG_MAC80211_HWSIM=m
#
# and maybe some debugging
# CONFIG_CFG80211_DEBUGFS=y
# CONFIG_MAC80211_DEBUGFS=y
# CONFIG_MAC80211_DEBUG_MENU=y
# CONFIG_MAC80211_VERBOSE_MPL_DEBUG=y
# CONFIG_MAC80211_VERBOSE_MHWMP_DEBUG=y
#
# for sshfs support, enable
# CONFIG_FUSE_FS=y
make -j 2
INSTALL_MOD_PATH=/mnt/qemu-rootfs make modules_install
From time to time you'll have to install new modules, etc. This is a useful convenience script for that:
LOOPDEV=/dev/loop0
MNTPOINT=/mnt/qemu-rootfs
sudo losetup $LOOPDEV || sudo losetup /dev/loop0 /path/to/your/rootfs.img
sudo killall qemu
sudo mount -t ext3 $LOOPDEV $MNTPOINT
sudo INSTALL_MOD_PATH=$MNTPOINT make modules_install
sudo umount $MNTPOINT
umount /mnt/qemu-rootfs
qemu -kernel arch/i386/boot/bzImage \
-hda rootfs.img \
-append "root=/dev/sda console=ttyS0" -nographic
You should be greeted by a login prompt:
Debian GNU/Linux 6.0 phooey ttyS0
phooey login:
"is a bit convoluted but avoids running qemu as root. Also, throughout the rest of the doc, the host's tap interface address will be assumed to be 192.168.55.1 and the guest address 192.168.55.2":/cozybit/open80211s/wiki/This
On your host edit the following:
/etc/qemu-ifup:
#!/bin/sh
/sbin/ifconfig $1 192.168.55.1 up
/etc/qemu-ifdown:
# leave this blank. We'll do the cleanup in our wrapper script
And create this wrapper script to invoke qemu
#! /bin/bash
IFNAME=`sudo tunctl -u javier -b`
sudo /etc/qemu-ifup $IFNAME
qemu -kernel ~/dev/wireless-testing/arch/i386/boot/bzImage \
-hda ~/qemu/rootfs.img \
-append "root=/dev/sda combined_mode=ide console=ttyS0" \
-nographic -net nic,model=e1000 \
-net tap,ifname=$IFNAME,script=no
tunctl -d $IFNAME
To get tunctl, install the user mode linux utilities package:
sudo apt-get install uml-utilities
On your guest edit the following:
/etc/network/interfaces
auto eth0
iface eth0 inet static
address 192.168.55.2
netmask 255.255.255.0
remove udev net rules, we'll be using one interface per guest:
rm /etc/udev/rules.d/70-persistent-net.rules
Now boot and you should be able to ping from guest to host and vice versa.
Generate a passphrase-less private key on guest and copy public key to host to automate login. Could be pretier, but adding the line below to /etc/rc.local on guest works.
sshfs -o allow_other -o idmap=user [email protected]:/path/to/remote/dir \
/local/mount/point
It is also possible to mount external directories from within qemu using virtio. The full instructions can be found here, but here is a brief overview.
The distro11s qemu config already has the correct kernel config settings. You will probably have to build qemu with virtio support which is automatically enabled if you run configure with libattr1-dev installed. USB pass-through is a nice feature to have, and you might as well enable it now. The newest libusb can be found at http://libusbx.org/.
NOTE: you can also try installing your distribution's qemu source package. The rebuild then shouldn't require a new libusb-1.0-dev.
- uninstall any existing qemu package
git clone http://git.qemu.org/git/qemu.gitcd qemu && /configure '--enable-libusb' '--target-list=x86_64-softmmu' '--enable-debug' '--enable-kvm'- make && sudo make install
This could be useful for launching multiple qemu instances using the same rootfs, and there is no need to rebuild the rootfs image after every change. However, in a benchmark involving boot and running 10 hwsim tests, the current ext3fs img approach was over 2x faster in both cases, so this is still the default.
# virtfs rootfs
-append "root=root rw rootflags=rw,trans=virtio,version=9p2000.L rootfstype=9p combined_mode=ide console=ttyS0" \
-fsdev local,id=root,path=${STAGING},security_model=none \
-device virtio-9p-pci,fsdev=root,mount_tag=/dev/root \
Bob Copeland suggested a compromise, where the rootfs would still be contained in a filesystem image, but the module and hwsim test paths would be virtfs mounts. We can then do away with the SSHFS_AUTOMOUNT complexity currently needed for accessing the hwsim tests.
# selective virtfs mounts
-fsdev local,id=modules,path=${STAGING}/lib/modules,security_model=mapped-xattr \
-fsdev local,id=home,path=/home,security_model=mapped-xattr \
-device virtio-9p-pci,fsdev=modules,mount_tag=modules \
-device virtio-9p-pci,fsdev=home,mount_tag=home \
Then in the rootfs image, you have to enter the right fstab parameters:
home /home 9p trans=virtio,version=9p2000.L 0 0
modules /lib/modules 9p trans=virtio,version=9p2000.L 0 0
This is now the default.
We could take this further and have a separate external /cozy_usr as the installation target for any custom packages such as iw, authsae, etc. and then mount this to /usr inside qemu.
-
losetup does not exist. just create the filesystem and double-click to mount it
-
sudo port install e2fsprogsto get mkfs
TODO: Finish this section
Symptom: In dmesg, kvm: disabled by bios
Solution: enable VT and VT-d in CPU page of bios, then power cycle your machine.
Symptom:
[ 37.498636] ata1.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x0
Solution: Make sure rootfs.img has the appropriate readwrite permissions, duh
/etc/init.d/udev: 1: cannot create /run/udev//root-link-rule: Read-only file system
That also causes other failures like no serial console.
Deleting /run on your guest filesystem did the trick.
we don't need them anyway, edit /lib/udev/net.agent like so:
- ppp*|ippp*|isdn*|plip*|lo|irda*|ipsec*)
+ ppp*|ippp*|isdn*|plip*|lo|irda*|ipsec*|mesh*)
to disable spawning of net.agents for mesh interfaces. If the net.agents still persist after that edit, add 'exit 1' near the beginning of /lib/udev/net.agent:
./lib/udev/hotplug.functions
+ exit 1
if [ -z "$INTERFACE" ]; then
http://www.cryptosystem.org/archives/2005/03/kernel-development-with-qemu http://www.coreboot.org/QEMU_Build_Tutorial#Creating_a_hard_disk_image http://www.walkernews.net/2007/07/01/create-linux-loopback-file-system-on-disk-file/