LTSP - Linux Terminal Server Project

LTSP - Linux Terminal Server Project - v4.1 James McQuillan

jam@LTSP.org

4.1.3-en 2004-06-20 jam 2004 James A. McQuillan GNU/Linux makes a great platform for deploying diskless thin clients. The primary purpose of this document is to show you how to deploy thin clients using LTSP. But, this document also covers many issues with diskless workstations in general. Introduction The LTSP provides a simple way to utilize low cost workstations as either graphical or character based terminals on a GNU/Linux server. In a traditional office setting, there are relatively high powered Intel based PC's spread around at every desk. Each with several gigabytes of hard disk space. Users store their own data on the local hard drives and backups are rarely (if ever) performed. Does it really make sense to have a full computer at each desk? We say no. Fortunately, there is another way. Utilizing the LTSP, you can take very low-end PCs, remove the hard drive, floppy and CDRom, and add a bootable network card. Many network cards have bootrom sockets, just waiting for a bootrom to be inserted. During the boot phase, the diskless workstation obtains its IP info and a kernel from the server, then mounts the root filesystem from the server via NFS. The workstation can be configured in one of 3 modes: Graphical X Window System interface Using X Windows, the workstation can be used to access any applications on the server, or on other servers within the network. Character based Telnet sessions The workstation can invoke multiple telnet sessions to the server. Each Telnet session will be on a separate virtual screen. Pressing Alt-F1 through Alt-F9 will switch between the telnet sessions. Shell prompt The workstation can be configured to drop you right into a bash shell at the console. This is very useful when debugging problems with X Windows or NFS. The really neat thing is that you can have lots of workstations served by a single GNU/Linux server. How many workstations? Well, that depends on the size of the server, and the applications that will be used. It's not unusual to have 50 workstations, all running Mozilla and OpenOffice from a Dual P4-2.4 with 4GB of ram. We know this works. In fact, the load-average is rarely above 1.0! Disclaimer Neither the author nor the distributors, or any other contributor of this document are in any way responsible for physical, financial, moral or any other type of damage incurred by following the suggestions in this text. Copyright and License This document is copyright 2004 by James McQuillan, and is released under the terms of the GNU Free Documentation License, which is hereby incorporated by reference. The GNU Free Documentation License can be found at http://www.gnu.org/licenses/fdl.html. Theory of operation Booting a diskless workstation involves several steps. Understanding what is happening along the way will make it much easier to solve problems, should they arise. There are four basic services required to boot an LTSP workstation. They are: DHCP TFTP NFS XDMCP LTSP is very flexible. Each of the above services can be served from the same server, or from several different servers. For our example, we'll describe a simple setup which consists of a single server, handling all of the above services. The steps that the workstation will go through Load the linux kernel into the memory of the workstation. This can be done several different ways, including: Bootrom (Etherboot,PXE,MBA,Netboot) Floppy Harddisk CD-ROM USB Memory device Each of the above booting methods will be explained later in this chapter. Once the kernel has been loaded into memory, it will begin executing. The kernel will initialize the entire system and all of the peripherals that it recognizes. This is where the fun really begins. During the kernel loading process, a ramdisk image will also be loaded into memory. A kernel command line argument of root=/dev/ram0 tells the kernel to mount the image as the root directory. Normally, when the kernel is finished booting, it will launch the init program. But, in this case, we've instructed the kernel to load a small shell script instead. We do this by passing init=/linuxrc on the kernel command line. The /linuxrc script begins by scanning the PCI bus, looking for a network card. For each PCI device it finds, it does a lookup in the /etc/niclist file, to see if it finds a match. Once a match is found, the name of the NIC driver module is returned, and that kernel module is loaded. For ISA cards, the driver module MUST be specified on the kernel command line, along with any IRQ or address parameters that may be required. A small DHCP client called dhclient will then be run, to make another query from the DHCP server. We need to do this separate user-space query, because we need more information than the bootrom retrieved with the first dhcp query. When dhclient gets a reply from the server, it will run the /etc/dhclient-script file, which will take the information retrieved, and configure the eth0 interface. Upto this point, the root filesystem has been a ram disk. Now, the /linuxrc script will mount a new root filesystem via NFS. The directory that is exported from the server is typlically /opt/ltsp/i386. It can't just mount the new filesystem as /. It must first mount it as /mnt. Then, it will do a pivot_root. pivot_root will swap the current root filesystem for a new filesystem. When it completes, the NFS filesystem will be mounted on /, and the old root filesystem will be mounted on /oldroot. Once the mounting and pivoting of the new root filesystem is complete, we are done with the /linuxrc shell script and we need to invoke the real /sbin/init program. Init will read the /etc/inittab file and begin setting up the workstation environment. One of the first items in the inittab file is the rc.sysinit command that will be run while the workstation is in the 'sysinit' state. The rc.sysinit script will create a 1mb ramdisk to contain all of the things that need to be written to or modified in any way. The ramdisk will be mounted as the /tmp directory. Any files that need to be written will actually exist in the /tmp directory, and there are symbolic links pointing to these files. The /proc filesystem is mounted. The lts.conf file will be parsed, and all of the parameters in that file that pertain to this workstation will be set as environment variables for the rc.sysinit script to use. If the workstation is configured to swap over NFS, the /var/opt/ltsp/swapfiles directory will be mounted as /tmp/swapfiles. Then, if there isn't a swapfile for this workstation yet, it will be created automatically. The size of the swapfile is configured in the lts.conf file. The swapfile will then be enabled, using the swapon command. The loopback network interface is configured. This is the networking interface that has 127.0.0.1 as its IP address. If Local apps is enabled, then the /home directory will be mounted, so that the apps can access the users home directories. Several directories are created in the /tmp filesystem for holding some of the transient files that are needed while the system is running. Directories such as: /tmp/compiled /tmp/var /tmp/var/run /tmp/var/log /tmp/var/lock /tmp/var/lock/subsys will all be created. The /tmp/syslog.conf file will be created. This file will contain information telling the syslogd daemon which host on the network to send the logging information to. The syslog host is specified in the lts.conf file. There is a symbolic link called /etc/syslog.conf that points to the /tmp/syslog.conf file. The syslogd daemon is started, using the config file created in the previous step. Once the rc.sysinit script is finished, control returns back to the /sbin/init program, which will change the runlevel from sysinit to 5. This will cause any of the entries in the /etc/inittab file to be executed. By default, there are entries in inittab to run the /etc/screen_session script on tty1, tty2 and tty3. That means that you can run 3 sessions at a time, and the type of session is controlled by the SCREEN_01, SCREEN_02 and SCREEN_03 entries in lts.conf. More entries can be setup in inittab for more sessions, if desired. If SCREEN_01 is set to a value of startx, then the /etc/screen.d/startx script will be executed, which will launch the X Windows System, giving you a graphical user interface. In the lts.conf file, there is a parameter called XSERVER. If this parameter is missing, or set to "auto", then an automatic detection of the video card will be attempted. If the card is PCI or AGP, then it will get the PCI Vendor and Device id, and do a lookup in the /etc/vidlist file. If the card is supported by Xorg 6.7, the pci_scan routine will return the name of the driver module. If it is only supported by XFree86 3.3.6, pci_scan will return the name of the X server to use. The startx script can tell the difference because the older 3.3.6 server names start with 'XF86_', whereas the newer Xorg X sesrver modules are typically lowercase names, like ati or trident. If Xorg is used, then the /etc/build_x4_cfg script will be called to build an XF86Config file. If XFree86 3.3.6 is used, then /etc/build_x3_cfg will be called to build the XF86Config file. These files are placed in the /tmp directory. Which, if you'll remember, is a ramdisk, seen only by the workstation. The XF86Config file will be built, based on entries in the /etc/lts.conf file. Once the XF86Config file has been built, then the startx script will launch the X server with that new config file. The X server will send an XDMCP query to the LTSP server, which will offer a login dialog. At this point, the user can log in. They'll get a session on the server. This confuses alot of people at first. They are sitting at a workstation, but they are running a session on the server. All commands they run, will be run on the server, but the output will be displayed on the workstation. Loading the kernel into memory Getting the Linux kernel into the workstations memory can be accomplished in a variety of ways. Boot ROM Local media Boot ROM Etherboot Etherboot is a very popular open-source bootrom project. It contains drivers for many common network cards, and works very well with LTSP. Linux kernels must be tagged with the mknbi-linux, which will prepare the kernel for network booting, by prefixing the kernel with some additional code, and appending the initrd to the end of the kernel. The kernels that are supplied with LTSP are already tagged, and ready to boot with Etherboot. Etherboot can also be written to a floppy, which works great for testing. PXE Part of the 'Wired for Management' specification from the late 1990's included a specification for a bootrom technology known as the Pre-boot Execution Environment commonly abreviated as PXE. A PXE bootrom can load at most a 32 kilo-byte file. A Linux kernel is quite a bit larger than that. Therefore, we setup PXE to load a 2nd stage boot-loader called pxelinux. pxelinux is small enough to be loaded, and it knows how to load much larger files, such as a Linux kernel. MBA Managed Boot Agent (MBA) is a bootrom from a company named emBoot. emBoot used to be the Lanworks division of 3Com. MBA is really four bootroms in one. It will handle PXE, TCP/IP, RPL and Netware. The MBA implementation of PXE works very well. You can use it with pxelinux to boot a Linux kernel. The TCP/IP method can be used, but the kernel must first be prepared with a utility called imggen. Netboot Netboot, like Etherboot, is a free software project that provides boot ROM images. The difference is that Netboot is a wrapper around the NDIS driver or packet drivers that the manufacturer ships with the card. Local media Floppy disk There are 2 ways to boot a LTSP workstation with a floppy. One way is to load Etherboot in the boot sector of the floppy. Then, it will act just like a bootrom. The boot code will be executed, the network card will be initialized, and the kernel will be loaded from the network server. You can also write the kernel and initrd to the floppy, and boot that way. However, it is actually faster to load the kernel over the network. Hard disk The hard disk can be used with LILO or GRUB, to load the Linux kernel and initrd. OR, you can load the Etherboot bootrom image from the harddisk, and it will act like a bootrom. CD-ROM A bootable CD-ROM can be loaded either with a Linux kernel, or an Etherboot image. USB Memory device Just like a CD-ROM, Floppy disk and Hard disk, you can use a USB Memory device to boot either an Etherboot module, or a complete Linux kernel and initrd image. Installing LTSP on the server LTSP is best thought of as a complete distribution of Linux. It's a distribution that sits on top of a host distribution. the Host distro can be any Linux distro that you want. In fact, there's no real requirement that the host be running Linux. The only requirement is that the host system needs to be able to serve NFS (Network File System). Most any Unix system can handle that. In fact, even some versions of Microsoft Windows can be configured to work as a LTSP server. There are three phases to building an LTSP server. Installing the LTSP utilities Installing the LTSP client packages Configuring the services needed by LTSP Installing the LTSP utilities Starting with version 4.1, LTSP has a package of utilities for installing and managing the LTSP client packages (The stuff that is executed on the thin clients), and for configuring the services on the LTSP server. The administration utility is called ltspadmin and the configuration tool is called ltspcfg. Both of these tools are part of the ltsp-utils package. The ltsp-utils package is available in both RPM and TGZ formats. Choose which format you prefer to install, and follow the appropriate instructions. Installing the RPM package Download the latest release of the ltsp-utils RPM package, and install it using the following command: rpm -ivh ltsp-utils-0.1-0.noarch.rpm The above commands will install the utilities on the server. Installing the TGZ packages Download the latest release of the ltsp-utils TGZ package and install it using the following commands: tar xzf ltsp-utils-0.1-0.noarch.tgz cd ltsp_utils ./install.sh cd .. The above commands will install the utilities on the server. This is useful for non-RPM based systems. Installing the LTSP client packages Once the installation of the ltsp-utils package is complete, you can run the ltspadmin command. This utility is used to manage the LTSP Client packages. It will query the LTSP download repository, and get the list of currently available packages. Run the ltspadmin command and you'll see a screen like the following:

LTSP installer - Main Screen From this screen, you can choose "Install/Update", and if this is your first time running the utility, it will display the installer configuration screen.

LTSP installer - Configuration Screen In the configuration screen, you can set several values that the installer will use, for downloading and installing the LTSP packages. Those values are: Where to retrieve packages from This is a URL, pointing to the the package repository. Typically, this would be http://www.ltsp.org/ltsp-4.1, but if you want to install the packages from a local filesystem, you can use the file: instead. For instance, if the packages are on a CD-ROM, and you've mounted the CD-ROM on /mnt/cdrom, then the value for the package repository would be: file:///mnt/cdrom. (Notice the 3 slashes). In which directory would you like to place the LTSP client tree This is the directory on the server, where you'd like to put the LTSP client tree. Typically, this would be: /opt/ltsp. The directory will be created, if it does not already exist. Within this directory, the root directories for each architecture will be created. Currently, only x86 workstations are officially supported by LTSP, but there are several people working on ports to other architectures, such as PPC and Sparc. HTTP Proxy If the server is behind a firewall, and all web access must go through a proxy, you can configure the installer to use the proxy here. The value should contain the URL to the proxy, including the protocol and the port. An example for this setting is: http://firewall.yourdomain.com:3128. If you don't need a proxy, you should set this to "none". FTP Proxy For packages located on an FTP server, if you need to go through an FTP proxy, you can enter it here. The syntax is similar to the HTTP Proxy option above. If you don't need a proxy, you should set this to "none". Once you get past the configuration screen, the installer will query the package repository, and obtain the list of currently available components.

LTSP installer - Component list Select each component that you want to install. To select it, move the high-lighted line to that component, and press 'I' to select the individual component. You can also press 'A' to select ALL of the components. Most of the time, this will be what you want. That way, you can support the broadest range of thin client hardware. There are several keys that can be used to navigate around this screen. You can get help on those keys by pressing the 'H' key.

LTSP installer - Help window If you want to see the list of packages that are in a particular component, you can press 'S', and the list of packages will be displayed. It will show the version currently installed, and also the latest available version.

LTSP installer - Package list Once the desired components are selected, you can exit the component selection screen. The installer will prompt you, to see if you really want to install/update the selected packages. If you answer 'Y', then it will download and install the selected packages. Configuring the services needed by LTSP There are four basic services required to support the booting of an LTSP workstation. They are: DHCP TFTP NFS XDMCP The ltspcfg can be used to configure all of those services, plus alot of other LTSP related things. You can access ltspcfg from the ltspadmin, or you can run ltspcfg by typing it at the command line. When you run the ltspcfg utility, it will scan the server, to assess what is currently installed and running. You'll see a screen like:

ltspcfg - Initial screen This shows all of the things that the utility is looking for. To configure all of the things that need to be setup, choose 'C', and the configuration menu will be displayed. From the configuration menu, you'll need to go through each item, to make sure it is configured properly for serving LTSP workstations.

ltspcfg - Initial screen 1 - Runlevel The Runlevel is variable used by the init program. With Linux and Unix systems, at any given time, the system is said to be in a specific "Runlevel". Runlevel 2 or 3 is typically used when the server is in text mode. Runlevel 5 typically indicates the system is in graphical mode on a network. For an LTSP server, traditionally Runlevel 5 is used. Most systems are already configured to serve NFS and XDMCP when in Runlevel 5. For those systems that aren't already configured for that, this utility will take care of them. 2 - Interface selection For systems that have multiple network interfaces, you'll need to specify which interface the Thin clients are connected to. By selecting the interface, the configuration tool will be able to properly create other configuration files, such as the dhcpd.conf and the /etc/exports files. 3 - DHCP configuration DHCP needs to be configured to supply the required fields to the workstation. Among those fields are fixed-address, filename, subnet-mask, broadcast-address and root-path. By selecting this menu item, you'll be able to create the dhcpd.conf configuration file, and enable dhcpd to run at startup time. 4 - TFTP configuration TFTP is used by the thin client to download the Linux kernel. The tftpd service needs to be enabled on the server, to serve up the kernel. 5 - Portmapper configuration The Portmapper is used by RPC services. Each RPC service, such as NFS. 6 - NFS configuration NFS is the service that allows local directory trees to be mounted by remote machines. This is required for LTSP, because the workstations mount their root filesystems from the server. This menu item will take care of configuring NFS to start at bootup time. The configuration file is /etc/exports and its creating is described later in this section. 7 - XDMCP configuration XDMCP is the "X Display Manager Control Protocol". The X server sends an XDMCP query to the Display manager on the server, to get a login prompt. Common display managers in use are XDM, GDM and KDM. This menu item will display which display managers are found, and which one is configured to run. For security purposes, the Display manager is configured by default to NOT allow remote workstations to connect. This is usually the reason for the infamous Gray screen with large X cursor. ltspcfg can usually configure the display manager to allow remote workstations to get a connection. 8 - Create /etc/hosts entries Many services, like NFS and the Display manager need to be able to map the IP address of the workstation to a hostname. You could setup the Berkeley Intenet Naming Daemon (BIND) to do that, but you'd have to make sure you get the reverses setup properly. Ultimately, using bind is probably the best way to do it, but configuration of bind is beyond the scope of this document and the ltspcfg utility. A much simpler approach for configuring the mapping of IP addresses and hostnames is the /etc/hosts file. 9 - Create /etc/hosts.allow entries Some services use a layer of security known as tcpwrappers. This is configured through the /etc/hosts.allow file. This menu item will setup that file for you. 10 - Create the /etc/exports file This is the file that NFS uses, to determine which directories are allowed to be mounted by remote machines. This menu item will create that file. 11 - Create the lts.conf file The configuration of each workstation is directed by entries in the lts.conf file. For fairly modern workstations with a PCI bus, it shouldn't require any additional entries in the lts.conf file. But, the file still needs to exist. This menu item will create the default lts.conf file for you. Workstation specific configuration Now, it's time to tell the LTSP server about your specific workstation. There are three files that contain information about the workstation. /etc/dhcpd.conf /etc/hosts /opt/ltsp/i386/etc/lts.conf /etc/dhcpd.conf The workstation needs an IP address and other information. It will get the following from the DHCP server: IP address hostname Server IP address Default gateway Pathname of kernel to load Server and directory path to be mounted as the root filesystem For our example LTSP environment, we have chosen DHCP for handling IP address assignment to the workstations. During the ltsp_initialize script, a sample dhcpd.conf file is installed. It is called /etc/dhcpd.conf.example you can copy that file to /etc/dhcpd.conf to use it as a basis for your dhcp configuration. You will need to modify the parts of this file that pertain to your specific workstation and server environment.

/etc/dhcpd.conf default-lease-time 21600; max-lease-time 21600; option subnet-mask 255.255.255.0; option broadcast-address 192.168.0.255; option routers 192.168.0.254; option domain-name-servers 192.168.0.254; option domain-name "ltsp.org"; option root-path "192.168.0.254:/opt/ltsp/i386"; shared-network WORKSTATIONS { subnet 192.168.0.0 netmask 255.255.255.0 { } } group { use-host-decl-names on; option log-servers 192.168.0.254; host ws001 { hardware ethernet 00:E0:18:E0:04:82; fixed-address 192.168.0.1; filename "/lts/vmlinuz.ltsp"; } } As of LTSP, version 2.09pre2, you no longer have to specify a particular kernel to load. The standard kernel package supports all of the network cards that Linux supports. There are two kernel files included in the LTSP kernel package. One kernel has the Linux Progress Patch (LPP) applied, and the other kernel does not. The names for the kernels are: vmlinuz-2.4.9-ltsp-5 vmlinuz-2.4.9-ltsp-lpp-5 You may have noticed that the kernel is sitting in the /tftpboot/lts directory, but in the "filename" entry in the /etc/dhcpd.conf file is missing the leading /tftpboot component of the pathname. That is because on Redhat versions 7.1 and above, TFTP is run with the '-s' option. That causes the tftpd daemon to run in secure mode. That is, it does a chroot to the /tftpboot directory when it starts. Therefore, all files that are available to the tftpd daemon are relative to the /tftpboot directory. Other Linux distributions may not have the '-s' option set for tftpd, so you will need to add the /tftpboot prefix to the kernel pathname. /etc/hosts IP address to hostname mapping Computers generally communicate just fine with IP addresses. Then, us humans come along and need to put names on the computers, because we can't remember the numbers. That's where DNS or the /etc/hosts file come into play. This IP address to hostname mapping generally isn't required, except in an LTSP environment. That's because without it, NFS will give you permissions errors when the workstation attempts to mount the root filesystem. In addition to NFS problems, if the workstation is not listed in the /etc/hosts file, you may also have problems with the GDM or KDM display managers. /opt/ltsp/i386/etc/lts.conf There are a number of configuration entries that can be specified in the lts.conf file. The lts.conf file has a simple syntax, that consists of multiple sections. There is a default section called [default] and there can be sections for individual workstations. The workstations can be identified by hostname, IP address or MAC address. A typical lts.conf file looks like this: lts.conf file # # Config file for the Linux Terminal Server Project (www.ltsp.org) # [Default] SERVER = 192.168.0.254 XSERVER = auto X_MOUSE_PROTOCOL = "PS/2" X_MOUSE_DEVICE = "/dev/psaux" X_MOUSE_RESOLUTION = 400 X_MOUSE_BUTTONS = 3 USE_XFS = N LOCAL_APPS = N RUNLEVEL = 5 [ws001] USE_NFS_SWAP = Y SWAPFILE_SIZE = 48m RUNLEVEL = 5 [ws002] XSERVER = XF86_SVGA LOCAL_APPS = N USE_NFS_SWAP = Y SWAPFILE_SIZE = 64m RUNLEVEL = 3 The following is a list of some of those entries: XSERVER If your video card is a PCI card, and if it is supported by X.org 6.7.0, then you just need the lts_x_core package. That contains all of the driver modules for X4. There are several XFree86 3.3.6 packages available for LTSP. This is in case your card isn't supported by X.org 6.7.0. You can make entries in the lts.conf for each individual workstation, or you can make default entries that are shared by all workstations. Our workstation has an Intel i810 video chipset, and it can correctly be auto detected, so we don't need any XSERVER entry in the lts.conf file. The XSERVER entry can be specified, if you want, or it can be set to 'auto' to show that it is going to be auto detected. RUNLEVEL We want to run the workstation in graphical mode, so we need to set the runlevel to '5'. This is done by another entry in the lts.conf file. Displaying the current configuration With ltspcfg, you can get a look at the current status of all of the services required by LTSP. From the ltspcfg main menu, press 'S', and you will see the current status.

ltspcfg - Current Status Setting up the workstation Once the server is setup, it is time to focus on setting up the workstation. The LTSP project is all about what happens after the kernel is in memory. There are several ways to get the kernel into memory, including Etherboot, Netboot, PXE and floppy disk. Booting with PXE If your network card or PC has PXE built into it, then you can use that to load the Linux kernel. PXE is a bootrom technology, similar to Etherboot or Netboot. You may need to enable the PXE bootrom on your NIC. You may also need to change the boot device order in your BIOS, to make "Boot from LAN" the first choice, rather than booting from the floppy or the hard disk. PXE has a limitation of only being able to load files that are 32kb or smaller. The Linux kernel is quite a bit larger than that, so you can't load the Linux kernel directly with PXE. You need to load something known as a 'Network Bootstrap Program' or NBP. There is an NBP available for loading Linux kernels called pxelinux.0. This is part of the syslinux package from kernel developer H. Peter Anvin. The LTSP kernel package includes the pxelinux.0 NBP and the configuration file needed to load the Linux kernel and an initial ramdisk image. The way it works, is this: The PXE bootrom initialializes the network card and sends a DHCP request. The DHCP server replies with an IP address, and the name of the NBP to load. The PXE bootrom downloads the NBP, places it in memory, and starts executing it. The NBP uses tftp to download the configuration file from the server. The configuration file contains the name of the kernel, the name of the initial ramdisk file, and options to pass to the kernel, once it is loaded. Here is an example of the pxelinux configuration file: prompt=0 label linux kernel bzImage-2.4.24-ltsp-4 append init=/linuxrc rw root=/dev/ram0 initrd=initrd-2.4.24-ltsp-4.gz The NBP then uses tftp to download the Linux kernel, and the initial ramdisk (initrd). Control is then passed to the Linux kernel, it boots up, mounts the initrd, and continues on with the startup of the Thin client. Booting with Etherboot

Ken Yap Etherboot is a software package for creating ROM images that can download code over an Ethernet network to be executed on an x86 computer. Many network adapters have a socket where a ROM chip can be installed. Etherboot is code that can be put in such a ROM.

Etherboot is also Open Source, protected under the GNU General Public License, Version 2 (GPL2). To use Etherboot, if you have a network card with an Etherboot bootrom, may need to change your BIOS configuration to tell it to "Boot from LAN" before booting from floppy or hard disk. If you don't yet have an Etherboot bootrom, you can either make a bootrom, or you can make a floppy disk with an Etherboot image on its boot sector. Etherboot supports a very large number of network cards. Over 200 models, with more being added all the time. Whether you choose to make a floppy or a burn the code to an Eprom, you'll need to determine which model network card you have. Choosing an Etherboot driver for an ISA network card For older ISA based network cards, it isn't so important that you determine the exact type. First of all, most of them are either ne2000 or 3Com 3c509 cards. You just need to pick the right Etherboot driver, the one that selects the correct media type on cards that both 10 base-2 (Coax) and 10 base-T (Twisted pair). Choosing an Etherboot driver for a PCI network card For PCI cards, it is important that you pick the Etherboot driver that matches the PCI Vendor and Device ID of the network card. Sometimes, you'll get lucky. You'll know exactly which model card you have, because the model number is printed on the card, and it exactly matches the description of one of the Etherboot modules. But, in many cases, you will need to find the PCI ID numbers. If your workstation has a floppy drive, you can boot a tomsrtbt (Tom's Root Boot) floppy. Or, if your workstation has a CD-ROM drive, you can boot a Knoppix CD. If you can't load linux on your workstation, then your only hope may be to move the network card to a machine that can boot Linux. Once you have Linux booted, you can use the lspci command. with the '-n' option. [root@jamlap root]# lspci -n 0000:00:00.0 Class 0600: 8086:7190 (rev 03) 0000:00:01.0 Class 0604: 8086:7191 (rev 03) 0000:00:03.0 Class 0607: 104c:ac1c (rev 01) 0000:00:03.1 Class 0607: 104c:ac1c (rev 01) 0000:00:07.0 Class 0680: 8086:7110 (rev 02) 0000:00:07.1 Class 0101: 8086:7111 (rev 01) 0000:00:07.2 Class 0c03: 8086:7112 (rev 01) 0000:00:07.3 Class 0680: 8086:7113 (rev 03) 0000:00:08.0 Class 0401: 125d:1978 (rev 10) 0000:01:00.0 Class 0300: 1002:4c4d (rev 64) 0000:06:00.0 Class 0200: 8086:1229 (rev 09) In the example above, you can see an entry for each of the PCI cards in the system. You only need to look at the Class 0200 devices. So, running the command again, looking only at the Ethernet interfaces, the list becomes much more manageable. [root@jamlap root]# lspci -n | grep "Class 0200" 0000:06:00.0 Class 0200: 8086:1229 (rev 09) The PCI ID numbers are: 8086:1229. The first field, 8086 is the PCI Vendor ID. In this example, the vendor is the Intel Corporation. The second field, 1229 is the PCI Device ID. That tells us which model of the network card. In this case, it's an EtherExpress 100 card. Creating a boot floppy You can download the Etherboot package and configure it for the type of bootrom that you need. Then, you can compile the source to produce a bootrom image that can be written to an EPROM, or written to a floppy disk. A much simpler approach is to go to Marty Connor's www.Rom-O-Matic.net website. Marty has done an excellent job of putting a web based front-end on the configuration and compilation process of making bootrom images with Etherboot. On his site, you select what type of network card you have, and what kind of image you want. Then, you have an opportunity to modify many Etherboot configuration options. Then, you can press the 'Get ROM' button and a custom bootrom image will be generated while you wait. For the ROM output format, choose 'Floppy Bootable ROM Image'. This will cause it to include a 512 byte header that is a boot loader for loading the etherboot image into ram where it can be executed. Press the 'Get ROM' button. The bootrom image will be generated while you wait. It only takes a few seconds, and when it completes, your browser will pop-up a "Save As" window where you can designate where you want the bootrom image to be saved on your computer. Once you've saved the image to your hard drive, you need to write it out to a floppy disk. Insert a floppy diskette into the drive and run the following command to write the floppy: dd if=Etherboot_Image of=/dev/fd0 Creating a bootrom Writing the Etherboot image to an EPROM requires an EPROM programmer. This is a piece of equipment that ranges in price from few hundred dollars to several thousand dollars, depending on the features. The process of creating a bootrom is entirely dependent on the EPROM programmer. This is beyond the scope of this document. Running the workstation Assuming that the server and workstation are configured properly, it should just be a matter of inserting the boot floppy in the floppy drive and turning on the workstation. The Etherboot code will be read from the floppy into memory, the network card will be found and initialized, the dhcp request will be sent on the network and a reply will be sent from the server and the kernel will be downloaded to the workstation. Once the kernel has initialized the workstation hardware, X windows will start and a login box should appear on the workstation, similar to the example below.

Login screen At this point, you can log in. An important thing to keep in mind is that you are logging into the server. All of the commands that you run are actually running on the server, and displaying their output on the workstation. That's the power of X windows. You can run any program that is supported by the server. Printing Aside from the workstation being a fully functioning GUI or character mode terminal, it can also act as a print server, allowing upto 3 printers to be attached to the parallel and serial ports. This is all transparent to the user of the workstation. They won't even notice the small amount of traffic that is going through the workstation to the printers. Client side setup LTSP uses the lp_server program on the workstation, to redirect print jobs from the server to the printer attached to one of the ports on the workstation. To enable the printer on the workstation, there are a set of configuration entries in the lts.conf file. [ws001] PRINTER_0_DEVICE = /dev/lp0 PRINTER_0_TYPE = P The above entry will cause the lp_server program to run as a daemon, listening on TCP/IP port 9100 for a print stream from the server. The print data will then be redirected to the printer attached to the /dev/lp0 parallel port. There are many more options available. Check the lts.conf section later in this document for more information on printer configuration entries. Server side setup Setting up the printer on the server is a matter of defining a print queue, using the printer configuration tool on the server. On Redhat 7.2, there is both a GUI and a Text based printer configuration tool. The GUI tool is called printconf-gui, and the text based tool is called printconf-tui. Older versions of Redhat have a program called printtool. Printtool also exists on Redhat 7.2, but it will call printconf-gui. Other Linux distributions have their own printer configuration tool.

Printconf-gui Adding new printer Once you launch the printer configuration tool, you need to add a new printer. The lp_server program allows the workstation to emulate an HP JetDirect print server. You just need to create a JetDirect printer. You need to give the printer a Queue name. The name can be anything, but make it a meaningful name, and the name can contain only the following characters: "a-z" lower-case letters "A-Z" upper-case letters "0-9" numeric digits "-" hyphen "_" underscore The name chosen in the example above is ws001_lp. This name makes it easy to see that the printer is associated with ws001.

Printconf-gui Detail info There are two fields required to communicate with the printer: IP Address or hostname of the workstation that the printer is associated with. The TCP port that the lp_server daemon is listening on. The first printer you connect to a workstation will be on TCP/IP port 9100. The second printer will be on port 9101, and the third printer will be on port 9102. Screen Scripts A feature of LTSP that was added with version 4.0 is something called Screen Scripts. These are scripts for starting various types of sessions. You can specify multiple screen scripts for a workstation. Doing so, will all you to have multiple sessions. They can be different types of sessions, or they can be the same type of session. For example, you could specify the following: SCREEN_01 = startx SCREEN_02 = shell This would start an X server on the first screen, and a shell prompt on the second screen. You can switch between the screens by pressing Ctrl-Alt-F1 to get to the first screen, and Ctrl-Alt-F2 to get to the second screen. You can specify upto 12 screen scripts for a workstation, but most people would just have a single one. Available types of screen scripts: startx This script will launch the Xserver with a -query option, to send an XDMCP request to a display manager to get a login dialog box to display on the screen. shell This script will launch a shell on the terminal. This is really intended to be used for troubleshooting problems with the workstation. Because it gives you a session on the thin client, rather than the server, it isn't very useful for running applications. telnet This script will launch a telnet session that will connect to the server. This will give you a character based session on the server. By default, telnet will connect to the LTSP server. If you want to specify a different server, you can pass it on the same line as the screen script. For example: SCREEN_01 = telnet server2.mydomain.com You can also specify any options that Telnet understands, But, if you do specify any options, then you must also specify the server to connect to. rdesktop This script will launch the rdesktop program, which will connect to a Microsoft Window server. You can specify any rdesktop options you want on the line, directly after the name of the screen script name. For example, if you want to specify the server to connect to, you can do it like this: SCREEN_01 = rdesktop -f w2k.mydomain.com The above example will launch rdesktop in full screen mode. The user will see a windows login dialog, and they will only have to log in once. This is very useful when you just want to get a windows login, without having a Linux login or window manager. The user won't even know they are running Linux. The screen scripts reside in the /opt/ltsp/i386/etc/screen.d directory. You can create your own screen scripts and place them in this directory. It's best to follow one of the existing scripts as an example. Troubleshooting If, after following through the previous chapters, your workstation doesn't boot, then you've got to start the process of troubleshooting the installation. The first thing to do is figure out how far through the bootup the workstation has gotten. Troubleshooting Etherboot floppy image When you boot from the floppy, you should see something similar to this: loaded ROM segment 0x0800 length 0x4000 reloc 0x9400 Etherboot 5.0.1 (GPL) Tagged ELF for [LANCE/PCI] Found AMD Lance/PCI at 0x1000, ROM address 0x0000 Probing...[LANCE/PCI] PCnet/PCI-II 79C970A base 0x1000, addr 00:50:56:81:00:01 Searching for server (DHCP)... <sleep> The above example shows what you can expect to see on the screen when booting from a floppy. If you don't see those messages, indicating that Etherboot has started, then you may have a bad floppy disk, or you didn't write the image to it properly. If, you see a message like the following, then it probably indicates that the Etherboot image you have generated is not the correct image for your network card. ROM segment 0x0800 length 0x8000 reloc 0x9400 Etherboot 5.0.2 (GPL) Tagged ELF for [Tulip] Probing...[Tulip]No adapter found <sleep> <abort> If it does get to the point where it detects the network card and displays the proper MAC address, then the floppy is probably good. Troubleshooting DHCP Once the network card is initialized, it will send out a DHCP broadcast on the local network, looking for a DHCP server. If the workstation gets a valid reply from the DHCP server, then it will configure the network card. You can tell if it worked properly if the IP address information is displayed on the screen. Here's an example of what it should look like: ROM segment 0x0800 length 0x4000 reloc 0x9400 Etherboot 5.0.1 (GPL) Tagged ELF for [LANCE/PCI] Found AMD Lance/PCI at 0x1000, ROM address 0x0000 Probing...[LANCE/PCI] PCnet/PCI-II 79C970A base 0x1000, addr 00:50:56:81:00:01 Searching for server (DHCP)... <sleep> Me: 192.168.0.1, Server: 192.168.0.254, Gateway 192.168.0.254 If you see the line that starts with 'Me:', following by an IP address, then you know that DHCP is working properly. You can move on to checking to see if TFTP is working. If instead, you see the following message on the workstation, followed by lots of <sleep> messages, then something is wrong. Although, it is common to see one or two <sleep> messages, after which the dhcp server replies. Searching for server (DHCP)... Figuring out what is wrong can sometimes be difficult, but here are some things to look for. Check the connections Is the workstation physically connected to the same network that the server is connected to? With the workstation turned on, make sure that the link lights are lit at all of the connections. If you are connecting directly between the workstation and the server (no hub or switch), make sure that you are using a cross-over cable. If you are using a hub or switch, then make sure you are using a normal straight-thru cable, both between the workstation and hub, and also between the hub and server. Is dhcpd running? You need to determine whether dhcpd is running on the server. We can find the answer a couple of ways. dhcpd normally sits in the background, listening on udp port 67. Try running the netstat command to see if anything is listening on that port: netstat -an | grep ":67 " You should see output similar to the following: udp 0 0 0.0.0.0:67 0.0.0.0:* The 4th column contains the IP address and port, separated by a colon (':'). An address of all zeroes ('0.0.0.0') indicates that it is listening on all interfaces. That is, you may have an eth0 and an eth1 interface, and dhcpd is listening on both interfaces. Just because netstat shows that something is listening on udp port 67, it doesn't mean that it is definately dhcpd that is listening. It could be bootpd, but that is unlikely, because bootp is no longer included on most distributions of Linux. To make sure that it is the dhcpd that is running, try running the ps command. ps aux | grep dhcpd You should see something like the following: root 23814 0.0 0.3 1676 820 ? S 15:13 0:00 /usr/sbin/dhcpd root 23834 0.0 0.2 1552 600 pts/0 S 15:52 0:00 grep dhcp The first line shows that dhcpd is running. The second line is just our grep command. If you don't see any lines showing that dhcpd is running, then you need to check that the server is configured for runlevel 5, and that dhcpd is configured to start in runlevel 5. On Redhat based systems, you can run the ntsysv and scroll down to make sure dhcpd is configured to start. You can try starting dhcpd with this command: service dhcpd start Pay attention to the output, it may show errors. Double-check the dhcpd configuration Does the /etc/dhcpd.conf file have an entry for our workstation? You should double-check the 'fixed-address' setting in the config file, to make sure it exactly matches the card in the workstation. Is ipchains or iptables blocking the request? Checking for ipchains Run the following command to see what it says: ipchains -L -v If you see something like this: Chain input (policy ACCEPT: 229714 packets, 115477216 bytes): Chain forward (policy ACCEPT: 10 packets, 1794 bytes): Chain output (policy ACCEPT: 188978 packets, 66087385 bytes): Then it isn't ipchains that is getting in the way. Checking for iptables Run the following command to see what it says: iptables -L -v If you see something like this: Chain INPUT (policy ACCEPT 18148 packets, 2623K bytes) pkts bytes target prot opt in out source destination Chain FORWARD (policy ACCEPT 0 packets, 0 bytes) pkts bytes target prot opt in out source destination Chain OUTPUT (policy ACCEPT 17721 packets, 2732K bytes) pkts bytes target prot opt in out source destination Then it is not iptables getting in the way. Is the workstation sending the request? Try watching the /var/log/messages file while the workstation is booting. You can do that with the following command: tail -f /var/log/messages This will 'follow' the log file as new records are written to it. server dhcpd: DHCPDISCOVER from 00:50:56:81:00:01 via eth0 server dhcpd: no free leases on subnet WORKSTATIONS server dhcpd: DHCPDISCOVER from 00:50:56:81:00:01 via eth0 server dhcpd: no free leases on subnet WORKSTATIONS If you see messages like those above, saying 'no free leases', that indicates that dhcpd is running, but it doesn't know anything about the workstation that is requesting an IP address. Troubleshooting TFTP Etherboot uses TFTP to retrieve a Linux kernel from the server. This is a fairly simple protocol, but sometimes there are problems trying to get it to work. If you see a message similar to this: Loading 192.168.0.254:/lts/vmlinuz-2.4.24-ltsp-4......... with the dots filling the screen rather quickly, that normally indicates that TFTP is working properly, and the kernel is downloading. If, instead, you don't see the dots, there is a problem. Possible problems include: <command>tftpd</command> is not running If tftpd isn't configured to run, then it certainly won't be able to answer the request from the workstation. You can see if it is running, you can use the netstat command, like this: [root@bigdog]# netstat -anp | grep ":69 " udp 0 0 0.0.0.0:69 0.0.0.0:* 453/inetd If you don't see any output from that command, then tftpd is likely not running. There are two common methods for invoking tftpd, They are inetd and the newer xinetd inetd uses a configuration file called /etc/inetd.conf. In that file, make sure that the line for starting tftpd is NOT commented out. This is what the line should look like: tftp dgram udp wait nobody /usr/sbin/tcpd /usr/sbin/in.tftpd -s /tftpboot xinetd uses a directory of individual configuration files. One file for each service. If your server is using xinetd, then the configuration file for tftpd is called /etc/xinetd.d/tftp. Below is an example: service tftp { disable = no socket_type = dgram protocol = udp wait = yes user = root server = /usr/sbin/in.tftpd server_args = -s /tftpboot } Make sure that the disable line doesn't say yes. Kernel not where tftpd expects to find it The kernel needs to be in a location that the tftpd daemon can access it. If the '-s' option is specified for the tftpd daemon, then whatever the workstation is asking for must be relative to the /tftpboot. So, if the filename entry in the /etc/dhcpd.conf file is set to /lts/vmlinuz-2.4.24-ltsp-4, then the kernel actually needs to be /tftpboot/lts/vmlinuz-2.4.24-ltsp-4 Troubleshooting NFS root filesystem There are several things that can prevent a root filesystem from being mounted. Including the following: No init found If you get the following error: Kernel panic: No init found. Try passing init= option to kernel. Then it is is most likely that either you are mounting the wrong directory as the root filesystem, or the /opt/ltsp/i386 directory is empty. Server returned error -13 If you get the following error: Root-NFS: Server returned error -13 while mounting /opt/ltsp/i386 This indicates that either the /opt/ltsp/i386 directory isn't listed in the /etc/exports file. Take a look in the /var/log/messages file to see if there are any clues. An entry like this: Jul 20 00:28:39 bigdog rpc.mountd: refused mount request from ws004 for /opt/ltsp/i386 (/): no export entry Then it confirms our suspicion that the entry in /etc/exports isn't correct. NFS Daemon problems (portmap, nfsd & mountd) NFS can be a complex and difficult service to trouble-shoot, but understanding what should be setup and what tools are available to diagnose the problems will surely help to make it easier. There are three daemons that need to be running on the server for NFS to work properly. portmap, nfsd and mountd. The Portmapper (portmap) If you get the following messages: Looking up port of RPC 100003/2 on 192.168.0.254 portmap: server 192.168.0.254 not responding, timed out Root-NFS: Unable to get nfsd port number from server, using default Looking up port of RPC 100005/2 on 192.168.0.254 portmap: server 192.168.0.254 not responding, timed out Root-NFS: Unable to get mountd port number from server, using default mount: server 192.168.0.254 not responding, timed out Root-NFS: Server returned error -5 while mounting /opt/ltsp/i386 VFS: unable to mount root fs via NFS, trying floppy. VFS: Cannot open root device "nfs" or 02:00 Please append a correct "root=" boot option Kernel panic: VFS: Unable to mount root fs on 02:00 This most likely is caused by the portmap daemon not running. You can confirm whether or not the portmapper is running by using the ps command: ps -e | grep portmap If the portmapper is running, you should see output like this: 30455 ? 00:00:00 portmap Another test is to use the netstat. The portmapper uses TCP and UDP ports 111. Try running this: netstat -an | grep ":111 " You should see the following output: tcp 0 0 0.0.0.0:111 0.0.0.0:* LISTEN udp 0 0 0.0.0.0:111 0.0.0.0:* If you don't see similar output, then the portmapper isn't running. You start the portmapper by running: /etc/rc.d/init.d/portmap start Then, you should make sure that the portmapper is setup to start when the server boots. Run ntsysv to make sure it is selected to run. The NFS and MOUNT daemons (nfsd & mountd) NFS has 2 daemons that need to be running. nfsd and mountd. They are both started by the /etc/rc.d/init.d/nfs script. You can run the ps command to make sure that they are running. ps -e | grep nfs ps -e | grep mountd If it shows that one or both of the daemons are not running, then you will need to start them. Normally, you should be able to run the startup script with the restart argument to cause them both to startup, but for some reason, the /etc/rc.d/init.d/nfs script doesn't restart nfsd that way. It only restarts mountd (bug?). So, you should instead run the following sequence of commands: /etc/rc.d/init.d/nfs stop /etc/rc.d/init.d/nfs start You may get errors on the stop command, but that is Ok. The start command should show OK as the status. If the daemons are running, but NFS is still not working, you can verify that they have registered themselves with the portmapper by using the rpcinfo command. rpcinfo -p localhost You should see results similar to the following: program vers proto port 100000 2 tcp 111 portmapper 100000 2 udp 111 portmapper 100003 2 udp 2049 nfs 100003 3 udp 2049 nfs 100021 1 udp 32771 nlockmgr 100021 3 udp 32771 nlockmgr 100021 4 udp 32771 nlockmgr 100005 1 udp 648 mountd 100005 1 tcp 651 mountd 100005 2 udp 648 mountd 100005 2 tcp 651 mountd 100005 3 udp 648 mountd 100005 3 tcp 651 mountd 100024 1 udp 750 status 100024 1 tcp 753 status This indicates that nfs (nfsd) and mountd are both running and have registered themselves with the portmapper. Troubleshooting the Xserver Oh boy!, Probably the single most difficult part of setting up an LTSP workstation is getting the X server configured properly. If you are using a fairly new video card, and it is supported by the Xorg Xservers, and you have a fairly new monitor that can handle a large range of frequencies and resolutions, then it is fairly straight forward. Usually, in that case, if it doesn't work, it is most likely the wrong X server for that card. When an X server doesn't work with your card, it is usually pretty obvious. Either the X server won't start, or the display will be incorrect. When the workstation is ready to start the X server, it calls the startx script, which starts the X server on the local workstation, with a -query option pointing to a server, where a display manager, such as XDM, GDM or KDM is running. Because the X server is started by the startx script, which is itself started by the init program, when it fails, init will attempt to run it again. init will continue this loop of trying to run the X server 10 times, then give up, because it thinks that it is respawning too quickly. After it finally gives up, the error message from the X server should be left on the screen. Waiting for the X server to fail 10 times can be rather irritating, so a simple way to avoid the repeated failures is to start the workstation in runlevel 3, so that the X server is NOT started automatically. Instead, when you boot the workstation, you will get a bash prompt. From the bash prompt, you can start the X server manually with the following command: sh /tmp/start_ws The X server will attempt to start, then when it fails, it will return back to the bash prompt, so you can see what the reason for the failure is. Troubleshooting the Display manager The display manager is the daemon that runs on the server, waiting for an X server to make contact with it. Once contact has been made, it will display a login dialog box on the screen, offering the user a chance to log into the server. The three most common display managers are: XDM - It's been around forever. It is included with the standard X windows system. GDM - The 'Gnome Display Manager'. This is part of the Gnome package. KDM - The 'KDE Display Manager'. This is part of the K Desktop system. Most recent GNU/Linux distributions include all three display managers. Grey screen with large X cursor This indicates that the X server is running, but it has not been able to make contact with a display manager. Some possible reasons for that are: The display manager may not be running On recent versions of Redhat (7.0 and above), the display manager is started from init. In the /etc/inittab file, there is a line that looks like this: x:5:respawn:/etc/X11/prefdm -nodaemon The prefdm script will make the determination of which display manager to run. The default display manager depends on which packages have been installed. If Gnome is installed, then GDM is the default display manager. If Gnome is not installed, then the prefdm script will check to see if KDE is installed. If it is, then KDM will be the default display manager. If KDE also is not installed, then XDM will be the default display manager. Using the netstat command, you should be able to see if there is a display manager running. On the server, run the following command: netstat -ap | grep xdmcp You should see results showing that there is a process listening on the xdmcp port (177). udp 0 0 *:xdmcp *:* 1493/gdm This shows clearly that gdm is running with a PID of 1493, and it is listening on the xdmcp port. If you see a line like the one shown above, indicating that there is definately a display manager listening, then you need to make sure that the workstation is sending the XDMCP query to the correct server. In the lts.conf file, you can have an entry which specifies the IP address of the server that is running the display manager. the entry is optional, but if present, should look like this: XDM_SERVER = 192.168.0.254 of course, the IP address for your network may be different than the example above. If the 'XDM_SERVER' entry is not present, it will then use the value of the 'SERVER' entry, if present. If that is not present, then it will use 192.168.0.254. Which ever way it is specified, you just need to make sure that the IP address is actually the correct address of the server running the display manager. The display manager may be configured to ignore requests from remote hosts. If you've determined that the display manager is running, then it is possible that it has been configured to ignore XDMCP requests from remote hosts. You will need to check the configuration files of the particular running display manager, to determine if it is configured properly. XDM The default configuration for Redhat is to disable the ability for workstations to get login access from XDM. The ltsp_initialize script will take care of enabling this for you, but if it's not working, you should check the /etc/X11/xdm/xdm-config file. Look for an entry that looks like this: DisplayManager.requestPort: 0 This entry MUST be commented out in order for XDM to listen on port 177 for remote requests. Another configuration file is also important for XDM to serve up remote login requests. There is a file called /etc/X11/xdm/Xaccess that MUST have a line that starts with an asterisk '*'. the line is normally included in the file, but Redhat leaves the line commented out. the ltsp_initialize script will fix the line for you, but if XDM doesn't seem to be working, you should check this file. A valid line should look like this: * #any host can get a login window KDM Newer versions of KDM have a file called kdmrc. Different Linux distributions store that file in different locations. For Redhat 7.2, it is /etc/kde/kdm/kdmrc. For the other distros, you should run the locate command to find out where it is stored. The entry that controls whether remote workstations can get a login is in the [Xdmcp] section. Make sure that the Enable entry is set to true. Older versions of KDM use the XDM configuration files, located in /etc/X11/xdm. GDM GDM uses a different set of configuration files. They are located in the /etc/X11/gdm directory. The main file to look at is the gdm.conf file. Look for the [xdmcp] section. you should see an entry within that section called 'Enable'. It must be set to '1' or 'true', depending on the version of GDM. Here is an example: [xdmcp] Enable=true HonorIndirect=0 MaxPending=4 MaxPendingIndirect=4 MaxSessions=16 MaxWait=30 MaxWaitIndirect=30 Port=177 Notice the 'Enable=true' line. Older versions of GDM use '0' and '1' to signify whether to Disable or Enable the remote XDMCP. Newer versions use 'false' and 'true'. If the Display manager is definately running, and it is listening for requests from remote workstations, it may be a simple problem that the display manager is unable to map the IP address to a hostname. The workstation either needs to be listed in the /etc/hosts file, or it needs to be correctly setup in the DNS tables. Kernels There are a few decisions that need to be made about the kernel that is run on the workstation. You need to decide whether you want to run one of the standard kernels available for download, or build your own. And, you need to decide if you want to display the graphic screen, complete with the progress bar which, which is made possible by the Linux Progress Patch (LPP). Standard LTSP supplied kernels The kernel package that is available with LTSP actually includes two kernels. One has the Linux Progress Patch already applied and configured, and the other does not have the patch applied. Both of the kernels already have the NFS Swap patch applied. Build your own kernel There are two ways to configure a kernel for LTSP. The default method is to use something called an 'Initial Ram Disk', or initrd for short. The initrd image is a small filesystem that is appended to the kernel. The initrd filesystem image is loaded into memory, and once the kernel is booted, it will mount the ramdisk as its root filesystem. There are a couple of advantages of using an initrd image. First, we can compile the network drivers as modules and load the correct module during bootup. This allows a single kernel which will support virtually all network cards. The other advantage is that we can run the DHCP client as a "user-land" program rather than in kernel-space. Running the client in user-land provides better control over the options requested and received from the server. Also, it makes the kernel slightly smaller. The other way to configure the kernel is without the initrd. Building a kernel without an initrd requires that the specific network card driver be statically linked into the kernel, and it also requires that IP-Autoconfig and "Root filesystem on NFS" are set when building the kernel. The advantage of not using an initrd is that the kernel is slightly smaller, and it will boot slightly faster. once the workstation is up and running, there is virtually no difference in how the workstation functions. The standard kernel for LTSP includes an Initial Ramdisk (initrd) that takes care of detecting the network card, and making a user-space DHCP request. A major goal for the image was to make it as small as possible. So, we chose the uClinux libc replacement library, and busybox for the utilities that we need during the boot. If you want to build your own kernels, you should download the ltsp_initrd_kit package. It contains the root filesystem hierarchy, and a script for building the image. Obtaining the source for the kernel When building a custom kernel, it's usually a good idea to start with fresh kernel sources direct from the ftp.kernel.org site. The reason for this is that the distributions, like Redhat, apply many patches to their kernel sources, leaving you with a set of source code that really doesn't match that of the official kernel. Download the kernel source package of your choice, and save it in the /usr/src directory. The kernels are located in the /pub/linux/kernel directory of the ftp.kernel.org ftp server. You will need to grab a recent 2.4.x series kernel, because you need to include devfs support. Also, if you want to include support for swapping over NFS or the Linux Progress Patch (LPP), you will need to make sure that you get the patches and kernel sources that are all the same. At the time of this writing, the 2.4.9 kernel is the latest to support those features. For our example, we will use the 2.4.9 kernel. The entire path is ftp://ftp.kernel.org/pub/linux/kernel/v2.4/linux-2.4.9.tar.bz2 Unpack the kernel sources in the /usr/src directory. You need to be careful, because when you un-tar the package, it will be in a directory called linux. You may already have a directory called linux with a different set of source code, and you don't want to clobber it. So, check for an existing directory, and if it is there, rename it to something else before unpacking the sources. The source package we downloaded had been compressed with the bzip2 compression utility. So, we need to uncompress it before we feed it to the tar program. You can use the following command to unpack it: bunzip2 <linux-2.4.9.tar.bz2 | tar xf - When the unpacking completes, you will have a directory called linux containing the entire source tree. At this point, I usually like to rename the directory to something more meaningful. mv linux linux-2.4.9 Once the directory has been renamed, then change into the new directory: cd linux-2.4.9 I usually like to modify the Makefile before I start configuring the new kernel. Near the top of the file is a variable named EXTRAVERSION. I set this to 'ltsp-1', so that the actual version number of the kernel will be '2.4.9-ltsp-1', which makes it easier to identify the kernel later. The top of the Makefile should look like this when you are done: VERSION = 2 PATCHLEVEL = 4 SUBLEVEL = 9 EXTRAVERSION = -ltsp-1 KERNELRELEASE=$(VERSION).$(PATCHLEVEL).$(SUBLEVEL)$(EXTRAVERSION) Kernel Patches After unpacking the kernel, you may have various patches you want to apply. For example, the NFS Swap patch or the Linux Progress Patch. These patches MUST be applied before configuring the kernel. NFS Swap patch The NFS Swap patch will allow the workstation kernel to use a swapfile located on an NFS server. While it usually recommended to have enough memory in the workstation to not require swapping, it can sometimes be difficult to add more memory, especially on older computers. So, the ability to swap over NFS can make an otherwise unuseable computer actually useable. If the current directory is /usr/src/linux-2.4.9, and the patch is in /usr/src, you can do the following to test the patch: patch -p1 --dry-run <../linux-2.4.9-nfs-swap.diff This will test the patch, to make sure it can be applied cleanly. If it finishes with no errors, then you can apply the patch without the --dry-run option. patch -p1 <../linux-2.4.9-nfs-swap.diff Linux Progress Patch (LPP) The Linux Progress Patch (LPP) will allow you to configure a graphic logo to display during the boot process. The normal kernel boot messages are redirected to another tty screen, and special instructions are added to the boot scripts to cause the Progress bar to reflect how far along the boot process is. Like the NFS Swap patch, you can test the LPP patch by issuing this command: patch -p1 --dry-run <../lpp-2.4.9 If the test completes successfully, then you can apply the patch with: patch -p1 <../lpp-2.4.9 Configuring kernel options You can now run the kernel configuration program of your choice. Available choices are: make xconfig This will invoke the X Windows version of the kernel configuration utility. make menuconfig This will invoke the curses based version of the kernel configuration utility. make config This will invoke the simple line-at-a-time version of the kernel configuration utility.