| One of the most important tasks of the network subsystem of an operating system is to process data packets according to the protocols used. In the designing of such a system, the multitude and flexibility of available methods play an important role, in addition to the performance and correctness of these protocols. Many network protocols differ a lot externally, but, when you implement them within an operating system, you can see quickly that the algorithms and operations on data packets are similar, and most of them can be reused. This chapter uses a Linux system as an example to show how data packets can be realized and what general methods are available to manipulate them. One main reason for the flexibility and efficiency of the Linux network implementation is the architecture of the buffers that manage network packets the so-called socket buffers, or skb for short. This central structure of the network implementation represents a packet during its entire processing lifetime in the kernel, representing one of the two basic elements of this network implementation, in addition to network devices. This means that a socket buffer corresponds to a sending or received packet. This chapter introduces buffer management (i.e., the structure of socket buffers) and the operations used to manage or manipulate them. Beginning with an introduction to the sk_buff structure, we will use an example to show how an IP packet is represented in this structure and how it changes along its way across different protocols and layers. In addition, this chapter introduces functions used to manage and change the structure. |
