《Linux networking Kernel》.pdf

Linux Networking Kernel Version 0.1 February, 12, 2003 1 – Introduction This report tries to describe the Networking part of the linux networking kernel. We try to describe the path the the packets follow in the forwarding path. We restrict this work to the IP code leaving other less used protocols like X.25 for other works. In the lower layers we also concentrate in the Ethernet protocol. In the IP code we describe only the IPv4 code although the IPv6 code does not have many differences (bigger addressing, no fragmentation, etc). Networking in the linux kernel In Figure \ref{fig:tree} we can see where the relevant code is in the linux kernel. Figure 1 - Networking in the linux tree Report Structure The report follows a bottom-up approach when describing the kernel. A brief introduction to the most relevant data structures is presented in chapter ... In chapter ... the sub-ip layer is described. In chapter ... we describe the IP layer. TCP and UDP are described in chapters ... The networking part of the kernel is populated with netfilter hooks where users can hang their code and analyse or change the packets. These are marked as HOOK in the kernel maps. NOTE: We need References, specially RFCs (TCP, PAWS, SACKs, etc) and netfilter. NOTE: We might need a final chapter on Socket layer and interation with the user space Acknowledgments We would like to thank 2 – Data Structures The networking part of the kernel uses mainly two data structures. One to keep the state of a connection called sock and other to keep the status of each packet (incoming and/or outgoing) called sk_buff. Here we proved a brief description of both. We also include a brief description of tcp_opt, a structure part of the sock structure which is used to maintain the tcp connection state. The details of tcp are presented in chapter \ref{chapter:tcp}. sk_buff The sk_buff structure is d