====== Getting iSCSI booting to work with Ubuntu including logon and ibft support! ====== **Tested on the following releases:** * Lucid Lynx 10.04 * Maverick Meerkat 10.10 == For this example, I will use these values: == * iSCSI Qualified Name: iqn.2010-09.org.test:diskimg * Server ip: 192.168.1.1 * Client ip: 192.168.1.2 * iSCSI username: user * iSCSI password: password * Client disk image: /mnt/disk-images/disk.img ===== Client side setup (initiator): ===== On the client side, install a standard Ubuntu Desktop on a physical or virtual machine. If you use a virtual machine, make the sure disk image is in 'raw' format, meaning no headers or such on the image. If it is a valid raw image, you should be able to see the partition table with, for example “fdisk -l disk.img” or even the “disktype” command, like so: $ disktype disk.img --- disk.img Regular file, size 6 GiB (6442450944 bytes) DOS/MBR partition map Partition 1: 5.858 GiB (6290407424 bytes, 12285952 sectors from 2048, bootable) Type 0x83 (Linux) Ext3 file system UUID 745D45C4-38FA-4B7C-85A3-71D37AE23AF3 (DCE, v4) Volume size 5.858 GiB (6290407424 bytes, 1535744 blocks of 4 KiB) Partition 2: 144 MiB (150994944 bytes, 294912 sectors from 12288000) Type 0x82 (Linux swap / Solaris) Linux swap, version 2, subversion 1, 4 KiB pages, little-endian Swap size 144.0 MiB (150986752 bytes, 36862 pages of 4 KiB) ==== Use the following steps to make the Ubuntu client iSCSI capable: ==== == Paste the following code into a new file called /etc/initramfs-tools/hooks/iscsi: == #!/bin/sh PREREQ="" prereqs() { echo "$PREREQ" } case $1 in prereqs) prereqs exit 0 ;; esac # Begin real processing below this line if [ ! -x /sbin/iscsistart ]; then exit 0 fi . /usr/share/initramfs-tools/hook-functions copy_exec /sbin/iscsistart /sbin # Disable stuff that can affect the network interface if [ -x /usr/lib/pm-utils/power.d/disable_wol ]; then chmod -x /usr/lib/pm-utils/power.d/disable_wol # affects usb network adapters fi if [ -x /sbin/iscsid ]; then chmod -x /sbin/iscsid # iscsid will interfere by re-discovering fi if [ -e /lib/udev/rules.d/75-persistent-net-generator.rules ]; then rm /etc/udev/rules.d/70-persistent-net.rules # prevent renaming of the network interfaces rm /lib/udev/rules.d/75-persistent-net-generator.rules # prevent re-creating the above rule fi # pcmcia/cardbus networking copy_exec /lib/udev/pcmcia-socket-startup copy_exec /lib/udev/pcmcia-check-broken-cis copy_exec /lib/udev/rules.d/85-pcmcia.rules copy_exec /etc/pcmcia/config.opts copy_modules_dir kernel/drivers/pcmcia copy_modules_dir kernel/drivers/net/pcmcia for x in pcmcia_core yenta_socket pcmcia; do force_load ${x} done # usb networking copy_modules_dir kernel/drivers/net/usb # iscsi module dependencies for x in scsi_transport_iscsi libiscsi libiscsi_tcp iscsi_tcp \ crc32c iscsi_ibft; do force_load ${x} done == Paste the following code into a new file called /etc/initramfs-tools/scripts/local-top/iscsi: == #!/bin/sh # Check if are in the initramfs environment if [ "$BOOT" = "local" ]; then # Get any iscsi command line parameters - cmdline parsing template taken from init script in initramfs for x in $(cat /proc/cmdline); do case $x in iscsi_ip=*) iscsi_ip=${x#iscsi_ip=} ;; iscsi_iqn=*) iscsi_iqn=${x#iscsi_iqn=} ;; iscsi_user=*) iscsi_user=${x#iscsi_user=} ;; iscsi_pw=*) iscsi_pw=${x#iscsi_pw=} ;; esac done # end of cmdline parsing if [ -n "${iscsi_ip}" ]; then iscsi_name='iqn.2000-09.org.local:UNKNOWN' fi # Check if an iscsi environment exists and start up the networking environment if [ -e /sys/firmware/ibft/target0/target-name -o -n "${iscsi_ip}" ]; then x=5 # Wait up to x seconds for a network interface to pop up echo -n "Searching for network cards..." until ifconfig -a|grep -q Ethernet; do sleep 1 echo -n . x=$(($x - 1)) if [ $x = 0 ]; then echo "No network cards detected!" exit 0 fi done # end of until echo "Searching for an dhcp server on all network interfaces..." CNT=5 until ifconfig ${netdev}|grep -q "inet addr"; do for netdev in `ifconfig -a|grep Ethernet|cut -d' ' -f1`; do # Do a round-robin search for dhcp servers #ip link set ${netdev} mtu 2500 # Optional: increase mtu for performance ip link set ${netdev} up # try to bring up the interface ipconfig -t 2 -c dhcp -d ${netdev} # Get an IP if ifconfig -a|grep -q "inet addr"; then break 2; fi # If we have got an address, stop searching. done # end of netdev probing CNT=$((${CNT} - 1)) if [ ${CNT} = 0 ]; then echo "No dhcp servers found!" exit 0 fi echo "Tries left: ${CNT}" done # end of until if [ -e /sys/firmware/ibft/target0/target-name ]; then #iscsistart -b # This should get the ibft parameters but it does not yet. # Lets do the following instead iscsistart \ -i `cat /sys/firmware/ibft/initiator/initiator-name` \ -t `cat /sys/firmware/ibft/target0/target-name` \ -a `cat /sys/firmware/ibft/target0/ip-addr` \ -u `cat /sys/firmware/ibft/target0/chap-name` \ -w `cat /sys/firmware/ibft/target0/chap-secret` \ -g 1 else iscsistart -i ${iscsi_name} -t ${iscsi_iqn} -a ${iscsi_ip} -u ${iscsi_user} -w ${iscsi_pw} -g 1 fi # end of iscsi firmware check fi # end of iscsi env check fi # end of initramfs check Make sure that eth0 is set to manual in /etc/network/interfaces so that Network Manager does not try to reconfigure the interface: # This file describes the network interfaces available on your system # and how to activate them. For more information, see interfaces(5). # The loopback network interface auto lo iface lo inet loopback # The primary network interface #auto eth0 iface eth0 inet manual iface eth1 inet manual iface eth2 inet manual Then do these commands at the terminal: sudo -s # Become root, if it asks for password, use the users password chmod +x /etc/initramfs-tools/hooks/iscsi # Make your hook script executable chmod +x /etc/initramfs-tools/scripts/local-top/iscsi # Make your script executable apt-get install open-iscsi open-iscsi-utils # Make sure you have the iscsi tools installed update-initramfs -u # Update initramfs with your changes, # the changes will also take effect even with upgrades You are now finished with editing the client. Shutdown the client now. ==== Last step: Transfer the image to the server ==== You could just copy the image as is, but it is even smarter to make a sparse file. That way only actual data is copied and not the empty sectors. You can do this under linux with the cp command: cp --sparse=always disk.img /media/server-data/disk-images/disk.img If you installed to a physical machine, boot up the machine with a live linux cd, become root and do this command: cp --sparse=always /dev/sda /media/server-data/disk-images/disk.img ===== Server side setup (target): ===== I am assuming that you are using either Ubuntu or Debian linux on the server. Make sure you have a dhcp and a tftp server installed and that you have followed the directions for setting up gpxe on the dhcp server from etherboot.org. Install the "iscsitarget" package using this command: sudo apt-get install iscsitarget Assuming /mnt/disk-images/disk.img is your client image, create the /etc/ietf.conf file with the following contents: Target iqn.2010-09.org.test:diskimg IncomingUser user password OutgoingUser Lun 0 Path=/mnt/disk-images/disk.img,IOMode=wb,Type=fileio Make sure iscsitarget can start at boot by using this value in /etc/default/iscsitarget ISCSITARGET_ENABLE=true Make the gpxe script /tftpboot/iscsi.gpxe with these contents: dhcp net0 set username user set password password sanboot iscsi:192.168.1.1::::iqn.2010-09.org.test:diskimg Edit your dhcpd.conf file and make sure the following is in your gpxe section of the config file: filename=iscsi.gpxe That's it! Start your server by running "sudo /etc/init.d/iscsitarget restart" and boot your pxe client and see what happens. ===== BTRFS workaround ===== In order to work around the kernel oops when iscsi image files are residing on a btrfs file system, change the /etc/ietf.conf file as follows: Target iqn.2010-09.org.test:diskimg IncomingUser user password OutgoingUser Lun 0 Path=/dev/loop0,IOMode=wb,Type=blockio ...and create a loop device that points to the image: losetup /dev/loop0 /mnt/disk-images/disk.img Remember to restart the iscsitarget service. ===== Random Notes ===== * If you don't install any proprietary video drivers, then what you got here is an image that will boot from either iSCSI, USB, or any internal/external disk on any machine. * Since Ubuntu is Debian based, these steps should work with Debian too but I haven't tried it myself * If you can't use the ibft interface or need to load the kernel and initrd from another source than iscsi, then you can pass the following kernel command line arguments to get it to connect to iscsi: iscsi_iqn= The iscsi qualifier name iscsi_ip= The target server's ip address iscsi_user= The username to login to the target server iscsi_pw= The password to login to the target server * pcmcia support is now complete. All linux supported pcmcia and cardbus network cards should work * You cannot create sparse images if you copy to a windows or samba network share, use nfs/ssh/nc (netcat) instead. Here is an example with nc (netcat): Client side: cat /dev/sda | nc -l -p 1234 Server side: nc 1234 | cp --sparse=always /dev/stdin disk.img --- //Quinn Plattel 2010/10/14// * This method disables iscsid. I have detected (probably the same) problem with it under Maverick (2.6.35 kernel), but not under Lucid (2.6.32 kernel). iscsid fails to reestablish the iscsi session and fails with response status code 0201. The problem is: I need iscsid to attach more iscsi shares... any idea how to solve that issue? --- cinquero --- //Trip-G 2011/1/19// * Quinn - Simply add another iscsistart to the file /etc/initramfs-tools/scripts/local-top/iscsi * if your second target is on the same server all you need is the iqn * like this: MYEXTRANAME="iqn.numbers.com.whatever:name:something.videostorage.mac" iscsistart \ -i `cat /sys/firmware/ibft/initiator/initiator-name` \ -t ${MYEXTRANAME} \ -a `cat /sys/firmware/ibft/target0/ip-addr` \ -g 1 if you have relevant username password stuff you can add that too. I added this after if [ -e /sys/firmware/ibft/target0/target-name ]; then and before "else" but you should be able to tag it on the end if you want. --- Trip-G