Creating sockets

Socket Programming

How do we build Internet applications? In this lecture, we will discuss the socket API and support for TCP communications between end hosts. Socket programing is the key API for programming distributed applications on the Internet. Note, we do not cover the UDP API in the course. If interested take CS60 Computer Networks.

Socket program is a key skill needed for the robotics project for exerting control - in this case the controller running on your laptop will connect to the server running on the bot.

Goals

We plan to learn the following from these lectures:

• What is a socket?
• The client-server model
• Byte order
• TCP socket API
• Concurrent server design
• Example of echo client and iterative server
• Example of echo client and concurrent server

The basics

Program. A program is an executable file residing on a disk in a directory. A program is read into memory and is executed by the kernel as a result of an exec() function. The exec() has six variants, but we only consider the simplest one (exec()) in this course.

Process. An executing instance of a program is called a process. Sometimes, task is used instead of process with the same meaning. UNIX guarantees that every process has a unique identifier called the process ID. The process ID is always a non-negative integer.

File descriptors. File descriptors are normally small non-negative integers that the kernel uses to identify the files being accessed by a particular process. Whenever it opens an existing file or creates a new file, the kernel returns a file descriptor that is used to read or write the file. As we will see in this course, sockets are based on a very similar mechanism (socket descriptors).

The client-server model

The client-server model is one of the most used communication paradigms in networked systems. Clients normally communicates with one server at a time. From a server’s perspective, at any point in time, it is not unusual for a server to be communicating with multiple clients. Client need to know of the existence of and the address of the server, but the server does not need to know the address of (or even the existence of) the client prior to the connection being established

Client and servers communicate by means of multiple layers of network protocols. In this course we will focus on the TCP/IP protocol suite.

The scenario of the client and the server on the same local network (usually called LAN, Local Area Network) is shown in Figure 1

Figure 1: Client and server on the same Ethernet communicating using TCP/IP.

The client and the server may be in different LANs, with both LANs connected to a Wide Area Network (WAN) by means of routers. The largest WAN is the Internet, but companies may have their own WANs. This scenario is depicted in Figure 2.

Figure 2: Client and server on different LANs connected through WAN/Internet.

The flow of information between the client and the server goes down the protocol stack on one side, then across the network and then up the protocol stack on the other side.

Transmission Control Protocol (TCP)

TCP provides a connection oriented service, since it is based on connections between clients and servers.

TCP provides reliability. When a TCP client send data to the server, it requires an acknowledgement in return. If an acknowledgement is not received, TCP automatically retransmit the data and waits for a longer period of time.

TCP is instead a byte-stream protocol, without any boundaries at all.

TCP is described in RFC 793, RFC 1323, RFC 2581 and RFC 3390.

Socket addresses

IPv4 socket address structure is named sockaddr_in and is defined by including the <netinet/in.h> header.

The POSIX definition is the following:

struct in_addr{
in_addr_t s_addr; /*32 bit IPv4 network byte ordered address*/
};


struct sockaddr_in {
   uint8_t sin_len; /* length of structure (16)*/
   sa_family_t sin_family; /* AF_INET*/
   in_port_t sin_port; /* 16 bit TCP or UDP port number */
   struct in_addr sin_addr; /* 32 bit IPv4 address*/
   char sin_zero[8]; /* not used but always set to zero */
};

The uint8_t datatype is unsigned 8-bit integer.

Generic Socket Address Structure

A socket address structure is always passed by reference as an argument to any socket functions. But any socket function that takes one of these pointers as an argument must deal with socket address structures from any of the supported protocol families.

A problem arises in declaring the type of pointer that is passed. With ANSI C, the solution is to use void * (the generic pointer type). But the socket functions predate the definition of ANSI C and the solution chosen was to define a generic socket address as follows:

struct sockaddr {
   uint8_t sa_len;
   sa_family_t sa_family; /* address family: AD_xxx value */
   char sa_data[14];
};

Host Byte Order to Network Byte Order Conversion

There are two ways to store two bytes in memory: with the lower-order byte at the starting address (little-endian byte order) or with the high-order byte at the starting address (big-endian byte order). We call them collectively host byte order. For example, an Intel processor stores the 32-bit integer as four consecutives bytes in memory in the order 1-2-3-4, where 1 is the most significant byte. IBM PowerPC processors would store the integer in the byte order 4-3-2-1.

Networking protocols such as TCP are based on a specific network byte order. The Internet protocols use big-endian byte ordering.

The htons(), htonl(), ntohs(), and ntohl() Functions

The follwowing functions are used for the conversion:

#include <netinet/in.h>

uint16_t htons(uint16_t host16bitvalue);

uint32_t htonl(uint32_t host32bitvalue);

uint16_t ntohs(uint16_t net16bitvalue);

uint32_t ntohl(uint32_t net32bitvalue);

The first two return the value in network byte order (16 and 32 bit, respectively). The latter return the value in host byte order (16 and 32 bit, respectively).

TCP Socket API

The sequence of function calls for the client and a server participating in a TCP connection is presented in Figure 3.

Figure 3: TCP client-server.

As shown in the figure, the steps for establishing a TCP socket on the client side are the following:

• Create a socket using the socket() function;
• Connect the socket to the address of the server using the connect() function;
• Send and receive data by means of the read() and write() functions.

The steps involved in establishing a TCP socket on the server side are as follows:

• Create a socket with the socket() function;
• Bind the socket to an address using the bind() function;
• Listen for connections with the listen() function;
• Accept a connection with the accept() function system call. This call typically blocks until a client connects with the server.
• Send and receive data by means of send() and receive().

The socket() Function

The first step is to call the socket function, specifying the type of communication protocol (TCP based on IPv4, TCP based on IPv6, UDP).

The function is defined as follows:

#include <sys/socket.h>

int socket (int family, int type, int protocol);

where family specifies the protocol family (AF_INET for the IPv4 protocols), type is a constant described the type of socket (SOCK_STREAM for stream sockets and SOCK_DGRAM for datagram sockets.

The function returns a non-negative integer number, similar to a file descriptor, that we define socket descriptor or -1 on error.

The connect() Function

The connect() function is used by a TCP client to establish a connection with a TCP server/

The function is defined as follows:

#include <sys/socket.h>

int connect (int sockfd, const struct sockaddr *servaddr, socklen_t addrlen);

where sockfd is the socket descriptor returned by the socket function.

The function returns 0 if the it succeeds in establishing a connection (i.e., successful TCP three-way handshake, -1 otherwise.

The client does not have to call bind() in Section before calling this function: the kernel will choose both an ephemeral port and the source IP if necessary.

The bind() Function

The bind() assigns a local protocol address to a socket. With the Internet protocols, the address is the combination of an IPv4 or IPv6 address (32-bit or 128-bit) address along with a 16 bit TCP port number.

The function is defined as follows:

#include <sys/socket.h>

int bind(int sockfd, const struct sockaddr *servaddr, socklen_t addrlen);

where sockfd is the socket descriptor, myaddr is a pointer to a protocol-specific address and addrlen is the size of the address structure.

bind() returns 0 if it succeeds, -1 on error.

This use of the generic socket address sockaddr requires that any calls to these functions must cast the pointer to the protocol-specific address structure. For example for and IPv4 socket structure:

struct sockaddr_in serv; /* IPv4 socket address structure */

bind(sockfd, (struct sockaddr*) &serv, sizeof(serv))

A process can bind a specific IP address to its socket: for a TCP client, this assigns the source IP address that will be used for IP datagrams sent on the sockets. For a TCP server, this restricts the socket to receive incoming client connections destined only to that IP address.

Normally, a TCP client does not bind an IP address to its socket. The kernel chooses the source IP socket is connected, based on the outgoing interface that is used. If a TCP server does not bind an IP address to its socket, the kernel uses the destination IP address of the incoming packets as the server’s source address.

bind() allows to specify the IP address, the port, both or neither.

The table below summarizes the combinations for IPv4.

The listen() Function

The listen() function converts an unconnected socket into a passive socket, indicating that the kernel should accept incoming connection requests directed to this socket. It is defined as follows:

#include <sys/socket.h>

int listen(int sockfd, int backlog);

where sockfd is the socket descriptor and backlog is the maximum number of connections the kernel should queue for this socket. The backlog argument provides an hint to the system of the number of outstanding connect requests that is should enqueue in behalf of the process. Once the queue is full, the system will reject additional connection requests. The backlog value must be chosen based on the expected load of the server.

The function listen() return 0 if it succeeds, -1 on error.

The accept() Function

The accept() is used to retrieve a connect request and convert that into a request. It is defined as follows:

#include <sys/socket.h>

int accept(int sockfd, struct sockaddr *cliaddr,
socklen_t *addrlen);

where sockfd is a new file descriptor that is connected to the client that called the connect(). The cliaddr and addrlen arguments are used to return the protocol address of the client. The new socket descriptor has the same socket type and address family of the original socket. The original socket passed to accept() is not associated with the connection, but instead remains available to receive additional connect requests. The kernel creates one connected socket for each client connection that is accepted.

If we don’t care about the client’s identity, we can set the cliaddr and addrlen to NULL. Otherwise, before calling the accept function, the cliaddr parameter has to be set to a buffer large enough to hold the address and set the interger pointed by addrlen to the size of the buffer.

The send() Function

Since a socket endpoint is represented as a file descriptor, we can use read and write to communicate with a socket as long as it is connected. However, if we want to specify options we need another set of functions.

For example, send() is similar to write() but allows to specify some options. send() is defined as follows:

#include <sys/socket.h>
ssize_t send(int sockfd, const void *buf, size_t nbytes, int flags);

where buf and nbytes have the same meaning as they have with write. The additional argument flags is used to specify how we want the data to be transmitted. We will not consider the possible options in this course. We will assume it equal to 0.

The function returns the number of bytes if it succeeds, -1 on error.

The receive() Function

The recv() function is similar to read(), but allows to specify some options to control how the data are received. We will not consider the possible options in this course. We will assume it is equal to 0.

receive is defined as follows:

#include <sys/socket.h>
ssize_t recv(int sockfd, void *buf, size_t nbytes, int flags);

The function returns the length of the message in bytes, 0 if no messages are available and peer had done an orderly shutdown, or -1 on error.

The close() Function

The normal close() function is used to close a socket and terminate a TCP socket. It returns 0 if it succeeds, -1 on error. It is defined as follows:

#include <unistd.h>

int close(int sockfd);