Linux TCP/IP Stack详解：嵌入式系统网络入门指南

需积分: 13 116 浏览量更新于2024-08-02 收藏 2.54MB PDF 举报

《Linux TCP/IP Stack：嵌入式系统网络技术详解》是一本由Thomas F. Herbert编著的深入指南，由Charles River Media出版于2004年，共600页。该书针对嵌入式系统程序员、工程师以及网络专业人员，旨在全面剖析在嵌入式系统项目中实现和利用Linux TCP/IP栈的过程。本书内容涵盖了丰富的主题，从TCP/IP基础知识到高级应用，适合那些希望深入了解Linux网络协议栈的读者。第1章是引言，概述了本书的目的和结构，让读者对整个Linux TCP/IP堆栈有一个总体认识。接下来的章节逐步展开：第2章讨论了宽带网络协议的历史与现状，对比了过去和现在的网络技术，帮助读者理解TCP/IP协议在不断变化的网络环境中的角色。第3章专门讲解TCP/IP在嵌入式系统中的应用，包括其原理、适用场景和设计挑战。这部分内容对于理解嵌入式系统如何与外部网络交互至关重要。第4章深入解析Linux网络接口和设备驱动，探讨如何配置和管理网络连接，确保硬件设备与TCP/IP协议的无缝对接。第5章详述Linux套接字（Sockets），这是一种编程接口，用于在应用程序和底层网络层之间进行通信，是构建网络应用程序的基础。第6章至第11章逐层深入，分别介绍了套接字缓冲区、内存分配、数据发送（通过UDP和TCP）、网络层（IP协议）以及传输层（UDP和TCP接收数据）的功能和实现细节。IPv6，下一代互联网协议，也在第11章有所涉及，展示了其与IPv4的区别和未来趋势。附录A列出相关的RFC（Request for Comments，互联网工程任务组文档），这些都是网络协议设计和标准的基石，有助于读者查阅最新的技术规范。附录B提供了关于本书光盘内容的详细信息，可能包含示例代码、实验数据或其他辅助学习资料。最后，书后还提供了参考文献和索引，便于读者查找特定主题或技术细节，以及一系列图表和表格来辅助理解复杂的概念。《Linux TCP/IP Stack：嵌入式系统网络技术详解》不仅是一本技术手册，也是学习者和专业人员在嵌入式系统开发过程中不可或缺的参考资料，能够帮助读者掌握Linux网络协议栈的实施和应用技巧。

among machines that attempt simultaneous access to the network. The first and still the most

popular LAN is Ethernet. From the standpoint of TCP/IP and some other data communication

stacks, the link layer is far simpler in Ethernet and other LANs than it is for WANs. This is

because there is no requirement for connection-oriented service. However, if the particular

network configuration includes packet filtering, bridging or switching, or address translation,

these capabilities will add quite a bit of complexity to the data link layer. Almost always, the

data link layer for Ethernet consists only of a framing layer and the MAC header. All of

Ethernet’s complexity is hidden in the physical layer (PHY). It is at the physical layer where

error detection and collision avoidance occur.

Generally, most LAN reception and transmission is handled at the device driver level. Incoming

packets are written directly in a circular buffer by the DMA capability of the Ethernet interface.

Outgoing packets are queued at the Ethernet chip for transmission. There are more complex

LANs such as WIFI 802.11. Although these protocols are more complex than Ethernet, the

details are hidden in the physical layer, and the data link layer provides a fairly simple Ethernet-

compatible interface. Of course, authentication and authorization complicates the picture, but

once a user is validated, the interface presents a simple LAN type interface where each machine

on the immediate LAN can see all the other machines.

1.6.2 Wide Area Networks

In this book, we look at things from the viewpoint of TCP/IP. Essentially, we consider protocols

other than TCP/IP from two points of view. We will see how some historical protocols

contributed essential technology that later became in corporated in TCP/IP. Other protocols are

still commonly used for WAN data transmission and often are interfaced to TCP/IP.

WANs are concerned with using capacity on the public carrier networks. There are many

methods to interface computers to WANs, from simple modems, DSL, and cable modems, up to

high-speed optical interfaces such as OC192. Essentially, however, there are two types of WAN

interfaces. The first is a direct point-to-point connection through a dedicated or private link. An

example would be a dedicated physical twisted pair or a leased line. These will be interfaced as a

network interface driver with hardware support for the interface. The other more complex type of

WAN interface involves a complete separate protocol suite with elements of the protocol

provided to manage negotiated bandwidth and data traffic classes. These protocols are for use in

a public network where data traffic is merged with other rate-payer’s data or where data packets

are submodulated on a shared carrier along with other data or voice traffic. An example of this

protocol would be Frame Relay or ATM. Interfacing protocols such as these to TCP/IP involves

one or more network interfaces without low-level hardware support. Examples of these

interfaces might be IP over Frame Relay, or Classical IP over ATM (CLIP). Some examples of

WAN data communication protocols are discussed in more detail in Chapter 2.

1.7 Packets and Frames

When discussing computer networking, there is confusion between when to call a chunk of data

a packet and when to call it a frame. Often, when IP-based networking is discussed over Ethernet,

the difference between these two terms is blurred even though there is a difference between them.

When discussing data transmission at the data link layer or the physical layer, the term frame is

used to describe a unit of data. A frame is thought to consist of a beginning sequence of

characters or bits defining the start of the frame. This is followed by a sequence of characters or

bits containing the payload data and followed by a sequence of characters or bits defining the end

of the frame. When discussing data transmission at the network layer or above, we consider the

unit of data transmission a packet. A packet is a data chunk following a header, and we think of

the header as encapsulating the data. The header typically includes a length field because packets

are generally considered variable length. In contrast to a packet, a frame is a data unit delineated

by a starting sequence of bits or flags and terminated by a stream of specific bits or flags.

Another important difference is that a packet is intended to be processed by higher-level software,

and a frame is intended to be processed by lower-level hardware.

1.8 The Digital Data Rate Hierarchy in The Public Network

Most of the protocols included in this chapter are for transmitting data across the public network.

As discussed earlier, the public network evolved from early hardwired voice circuits and evolved

over many years to be able to provide a variety of services and different rates. Originally, a

separate twisted pair of wires was used for each voice channel. In the mid-twentieth century, the

demand for voice circuits rose rapidly and telephone companies looked for a way to combine

voice channels in trunk lines without having to use huge cables consisting of bundles of twisted

pairs. Voice channels were multiplexed using analog methods across a single modulated

broadband carrier using a method called frequency domain multiplexing (FDM). With FDM,

each voice channel is modulated on a specific frequency band within the broadband carrier using

frequency modulation (FM) techniques. This technique continued to be used for about 20 years

until it became apparent that it was not the most efficient use of broadband channel capacity.

Moreover, FDM required expensive analog switching equipment in the telephone company’s

switching stations and therefore was not scalable to ever-increasing demands for increased

circuit carrying capacity [EDWAR00a].

Later, in the 1960s and 1970s, the public network was converted from analog to digital. Voice

phone calls were transmitted over individual channels modulated using pulse code modulation

(PCM) over a base band carrier. The voice information is modulated and divided into frames

where the local analog loop terminates at the phone company’s central office (CO). Each voice

channel can be transmitted using time division modulation (TDM) techniques. It was known that

to maintain minimum fidelity for voice transmission, each channel requires 3 kHz of bandwidth.

It was determined that 64 K bits per second was required to transmit a single voice line.

Therefore, the fundamental unit of bandwidth for the public network is 64 KB, and all digital

data rates are calculated in units of 64 KB of capacity. Table 1.3 shows the hierarchy of digital

channel capacity carried across electrical networks in North America. Table 1.4 shows the

equivalent hierarchy in use in Europe. Table 1.5 shows the Sonet and SDH Optical Hierarchy

[MCDYSO99].

Table 1.3: TDM Digital Hierarchy in North America

Name Multiplexing Level Number of Voice

Channels

Rate in Bits per

Second

Chapter 2: Broadband Networking Protocols

of Yesterday and Today

In Chapter 1, “Introduction,” we discussed the basics of data communication and presented the

OSI seven-layer model. We illustrated the difference between broadband and LAN protocols and

covered networking standards. Now, to fully appreciate the TCP/IP protocol suite and how it

came to be so widely used, it is helpful to discuss other networking protocols that influenced the

development of TCP/IP. Some protocols in this chapter are included because they are often used

to interface to TCP/IP when we want to send IP datagrams across a broadband or public network.

Others are included because they are significant in a historical context and have concepts that

were later borrowed by TCP/IP. Still others are included because they extend the reach of IP

packets into broadband realm and allow IP packets to be forwarded over the public networks.

2.1 Introduction

In this chapter, we extend our discussion on basic data communication begun in Chapter 1. We

will discuss the difference between asynchronous and synchronous data communication

protocols. From a historical context, we will introduce different types of synchronous protocols

and show an example of how they are used in the X.25 stack. High Level Data Link Protocol

(HDLC) will be covered, too, because it is the basis of both early and modern protocols using

electrical transmission methods. We will see how concepts used as part of TCP connection

management evolved from X.25 and HDLC. Frame Relay is important, too, and we will look at

it to see how it differs from X.25. The last but far from the least important protocol is

Asynchronous Transmission Mode (ATM). It is the most recent of the broadband protocols and

is important because it forms the basic infrastructure in the public telephony network that is the

basis for most voice and data communication. In addition, ATM provides a common point of

interface for sending TCP/IP traffic to the world outside our LANs.

The material in this chapter will include some important concepts that should help in the detailed

examination of Linux TCP/IP in later chapters. It is important to note that most of the broadband

protocols we cover are primarily concerned with providing reliable or connection-oriented

service. As we will see, when we go into more detail in the chapters on TCP/IP, IP traffic was

intended to be carried over connectionless networks. TCP, which runs over IP, is depended on to

provide a connection-oriented transport for most of the common application layer protocols.

However, we will see by examining the lower layer protocols in this chapter that the connection-

oriented protocols here are analogous to TCP’s implementation of connection management.

2.2 Connection-Oriented and Connectionless Protocols

WAN or broadband protocols are typically connection oriented because they provide sequenced

reliable delivery service. As we saw earlier, WANs were intended to use excess capacity in the

Public Switched Telephone Network (PSTN). The PSTN evolved from the earlier discretely

wired telephone system that was in use well into the last century. The virtual circuits have to

maintain the audio signal quality, and they work the same way as a physically connected pair of

telephones does. The same public network, originally designed to carry voice traffic, evolved to

a mixed traffic network where it is asked to carry data simultaneously as it carries voice data in

connections implemented as virtual circuits. This orientation to voice traffic naturally led to the

development of connection-oriented protocols, which are also circuit oriented. The simpler of the

connection-oriented or WAN protocols use the network in the way for which it was designed: to

carry voice channels over the public network.

In a similar fashion to circuit-based voice traffic, connection-oriented pro tocols work over

virtual circuits (VCs). The part of the protocols that establish the circuits is implemented at Layer

2 or higher in the protocol stack. (Refer to Chapter 1 for a description of the OSI seven layer

model.) The VCs are either configured at startup time or are dynamically set up. When they are

established at startup time, they are called provisioned virtual circuits (PVCs). When they are set

up dynamically by a signaling protocol, they are called switched virtual circuits (SVCs). The

signaling methods are actually implemented as separate protocols that negotiate the connection

between pairs of switches. Once they are set up, the VCs are actually full-duplex paths between

pairs of nodes.

LANs can be contrasted to WANs by the way in which the neighboring machines are connected.

Machines on WANs don’t really have neighbors; every connection between two nodes is either

pre-provisioned or separately negotiated. However, machines on LANs physically near each

other don’t need to run special software to negotiate a connection with other nodes before they

can talk to each other. On a LAN, every machine can see all its neighbors. The machines on a

LAN are all peers; every machine has the equal right to transmit to another machine. The

physical layer arbitrates in some fashion between the transmitting machines based on their link

layer addresses, and the upper layers are responsible for filtering out unwanted packets.

2.3 Broadband Networking Versus Local Area Networking

In this chapter and elsewhere in the book, we use the terms broadband and WAN as synonyms.

Although not precisely correct, this is the terminology in common use at the time of publication.

WAN network interfaces faster than analog modems are called “broadband” connections. As

discussed in earlier sections, the simpler broadband or WAN protocols with some exceptions

establish point-to-point connections. They rely on higher layer protocols to provide multicast or

broadcast capability. In contrast, LANs are democratic. All the machines on a LAN share the

same chunk of bandwidth. They can all talk at once or at least try to talk at once. They can all see

each other’s transmissions. Therefore, LANs provide a mechanism at the physical layer to

negotiate access to the physical network to make sure that there is only one speaker at a time. In

general, LANs have a fairly complicated physical layer but very simple data link layers. LANs

will checksum packets to make sure that bit-level communications are okay. Although packet

integrity is usually guaranteed, errors will result in dropped packets.

WANs usually have relatively simple direct serial interfaces at the physical layer. Higher layers,

usually the data link layer, have to do all the work of arbitrating bandwidth and negotiating the

connections. Therefore, when compared to LANs, WANs have simpler physical layers but more

complicated data link layers. The data link layers will compensate for transmission problems at

the physical layer provided sufficient bandwidth is available.

剩余480页未读，继续阅读

xaodoudou

粉丝: 0
资源: 3

Linux TCP/IP Stack详解：嵌入式系统网络入门指南

R10循环机芯 UI使用手册

The Linux TCP IP Stack Networking for Embedded Systems.rar

The Linux TCP IP Stack Networking for Embedded Systems

FNET - Embedded TCP/IP Stack-开源

The Linux TCP-IP Stack.rar

Building-AngularJS-and-Node.js-Apps-with-the-MEAN-Stack:[MongoDB, Express.js, Angular.js, Node.js] Building-AngularJS-and-Node.js-Apps-with-the-MEAN-Stack

Validate-the-ns-3-implementation-of-BIC-TCP-using-Network-Stack-Tester-NeST

Reading-and-comprehense-linux-Kernel-network-protocol-stack_y123456yz.tar.gz

TCP-IP.rar_Stack_tcp/ip stack

Using the NicheStack TCP_IP Stack - Nios II Edition Tutorial.pdf

最新资源