1. INTRODUCTION, CREDITS, COPYRIGHT AND SUMMARY

1.1 Introduction and Credits

In this HOWTO I will try to explain my experience in the configuration of a system that allows a diskless PC (one without floppy, CD-ROM or Hard-Disk) to access a remote server, boot up from it and then work as if it were a fully operational PC. The reader is supposed to know the standard tools of a Linux system as there are for example: Text editor, shell, kernel configuration, some tool to analyze net traffic as iptraf or tcpdump...

I want to thank Salo alias Gonzalo Perez de Olaguer ([email protected]) for his tips and corrections that where very important.

1.2 License

Copyright Lorenz von Krosigk. Permission to distribute and modify this document is granted under the GNU General Public License. An on-line copy is available at http:\\www.gnu.org\copyleft\gpl.html and alwqays has to be aplicated the last version of the GPL.

1.3 Summary

As a simple definition of a Diskless Client we take a PC without any local space for permanent data storage. That means specially that there is no way to keep configuration information or system files in the local computer.

1. When the client's bios boots up it invokes the Ethernet's card eprom to make a bootp or dhcp broadcast call to address 255.255.255.255. This general broadcast direction is used because the client has no idea about the network it forms part of nor the network mask in this moment. The only known fact is it's card MAC number. The client expects to get it's basic IP specifications as an answer to this call that goes to to every computer connected physically in the same local network. That means that the server that is supposed to answer to this call has to be inside of this local (sub)-network.
2. The bootp or dhcp server gets the information associated with that MAC address from the file /etc/bootptab or /etc/dhcpd.conf and sends back to the calling client at least the following data:
   • IP address
   • Network mask
3. With this information the client starts up a tftp session to download a kernel from the server by making another broadcast call to the local network using Network Mask and it's IP address
4. The client runs the kernel, which makes a new bootp/dhcp call to obtain again its network configuration.
5. The server paciently answers to the client with at least the following data:
   • IP address
   • Network mask
6. The next step is a call from the client to the server to mount its root filesystem, usually from the server's directory /tftpboot/<client's IP address>.
7. This call gets answered from the server's rpc. mountd daemon after having confirmed in /etc/exports that the client has permissions to remotely access those files.
8. At the end of the startup process the client mounts his root filesystem from the server's /tftpboot/<client's IP address> using the NFS protocol that is provided by the rpc.nfsd daemon in the server.