From a networking point of view, we use BPF programs for two main use cases: packet capturing and filtering.
This means that a user-space program can attach a filter to any socket and extract information about packets flowing through it and allow/disallow/redirect certain kinds of packets as they are seen at that level.
The goal of this chapter is to explain how BPF programs can interact with the Socket Buffer structure at different stages of the network data path in the Linux kernel network stack. We are identifying, as common use cases two types of programs:
Program types related to sockets
Programs written for the BPF-based classifier for Traffic Control
The Socket Buffer structure, also called SKB or
sk_buff, is the one in the kernel that is created and used for every packet sent or received. By reading the SKB you can pass or drop packets and populate BPF maps to create statistics and flow metrics about the current traffic.
In addition some BPF programs allow you to manipulate the SKB and, by extension, transform the final packets in order to redirect them or change their fundamental structure. For example, on an IPv6-only system, you might write a program that converts all the received packets from IPv4 to IPv6, which can be accomplished by mangling with the packets’ SKB.
Understanding the differences between the different kinds of programs we can write and how different programs lead to the same goal is the key to understanding ...