Linux Kernel Networking: Implementation and Theory (2014)

Posted 2021-09-09 rtoax

https://apprize.best/linux/kernel/6.html

CHAPTER 5. The IPv4 Routing Subsystem

Chapter 4 discussed the IPv4 subsystem. In this chapter and the next I discuss one of the most important Linux subsystems, the routing subsystem, and its implementation in Linux. The Linux routing subsystem is used in a wide range of routers—from home and small office routers, to enterprise routers (which connect organizations or ISPs) and core high speed routers on the Internet backbone. It is impossible to imagine the modern world without these devices. The discussion in these two chapters is limited to the IPv4 routing subsystem, which is very similar to the IPv6 implementation. This chapter is mainly an introduction and presents the main data structures that are used by the IPv4 routing subsystem, like the routing tables, the Forwarding Information Base (FIB) info and the FIB alias, the FIB TRIE and more. (TRIE is not an acronym, by the way, but it is derived from the word retrieval). The TRIE is a data structure, a special tree that replaced the FIB hash table. You will learn how a lookup in the routing subsystem is performed, how and when ICMP Redirect messages are generated, and about the removal of the routing cache code. Note that the discussion and the code examples in this chapter relate to kernel 3.9, except for two sections where a different kernel version is explicitly mentioned.

Forwarding and the FIB

One of the important goals of the Linux Networking stack is to forward traffic. This is relevant especially when discussing core routers, which operate in the Internet backbone. The Linux IP stack layer, responsible for forwarding packets and maintaining the forwarding database, is called the routing subsystem. For small networks, management of the FIB can be done by a system administrator, because most of the network topology is static. When discussing core routers, the situation is a bit different, as the topology is dynamic and there is a vast amount of ever-changing information. In this case, management of the FIB is done usually by userspace routing daemons, sometimes in conjunction with special hardware enhancements. These userspace daemons usually maintain routing tables of their own, which sometimes interact with the kernel routing tables.

Let’s start with the basics: what is routing? Take a look at a very simple forwarding example: you have two Ethernet Local Area Networks, LAN1 and LAN2. On LAN1 you have a subnet of 192.168.1.0/24, and on LAN2 you have a subnet of 192.168.2.0/24. There is a machine between these two LANs, which will be called a “forwarding router.” There are two Ethernet network interface cards (NICs) in the forwarding router. The network interface connected to LAN1 is eth0 and has an IP address of 192.168.1.200, and the network interface connected to LAN2 is eth1and has an IP address of 192.168.2.200, as you can see in Figure 5-1. For the sake of simplicity, let’s assume that no firewall daemon runs on the forwarding router. You start sending traffic from LAN1, which is destined to LAN2. The process of forwarding incoming packets, which are sent from LAN1 and which are destined to LAN2 (or vice versa), according to data structures that are called routing tables, is called routing. I discuss this process and the routing table data structures in this chapter and in the next as well.

更多信息查看原文：https://apprize.best/linux/kernel/6.html