Chapter 11. Application-Layer Elements Ports

Transcription

1 Chapter 11 Application-Layer Elements 11.1 Ports Sockets Socket Domains, Types and Protocols Operations on Sockets Network Programming Internet Addresses Checking Open Ports Socket Information Input and Output Daytime Protocol Client Socket Options Glossary Network applications do not have to concern themselves with the details of network communication. Operating systems and libraries provide abstractions and APIs that make network programming relatively easy. This chapter introduces the socket abstraction that is used in most network programming, and shows how sockets (and some related abstractions) can be used to write simple network programs in Java Ports Computers make connections for many purposes, and many different services can be provided by a single device, so a computer needs more than just an 93

2 94 CHAPTER 11. APPLICATION-LAYER ELEMENTS service port conventional name number purpose echo 7 echo received data back to the sender daytime 13 local time (as an ASCII string) FTP 21 file transfer SSH 22 secure remote login and copy Telnet 23 insecure remote login SMTP 25 sending time 37 local time (32-bit binary, seconds since ) whois 43 ownership of a domain name finger 79 system and user status HTTP 80 Word Wide Web pages POP 110 Post-Office Protocol: reading IMAP 143 Internet Message Access Protocol: reading Table 11.1: Commonly-used IP port numbers. On Unix-like systems (including Darwin/MacOS X and Linux) a comprehensive list of TCP and UDP well-known and registered port numbers can be found in the file /etc/services. Internet address to communicate with a service. Ports are numbers that identify a communication endpoint. Many services operate at a specific port number. Port numbers are usually in written in decimal. TCP and UDP ports are distinct, and so for each of the two protocols in the IP transport layer (TCP and UDP) there are 65,535 ports available. Ports 1 to 1023 are reserved for critical or well-known public services such as (smtp), web sites (HTTP), secure shell, etc. (Table?? lists some of the more commonly-used ones.) Some other numbers, between 1024 and 49151, are also registered with the Internet Association for Assigned Numbers (IANA) or are used unofficially (but consistently) by convention. Ports above can be used for any purpose, or are assigned dynamically to communication endpoints (when the endpoint is created) and released automatically (when the endpoint is destroyed).

3 11.2. SOCKETS Sockets A socket is a software structure that represents a communication endpoint associated with a specific port number. Applications can use sockets to send and receive datagrams (via UDP) or to create and communicate over connection-oriented virtual circuits (using TCP). The basic operations that can be performed on a socket are: create a socket and bind it to a specified or randomly-assigned port set and get socket options send and receive apacketofdata(udpclientsandservers) connect to remote machine (TCP clients) listen for and accept incoming connection requests (TCP severs) read and write to a data stream (TCP clients and servers) shutdown and close (destroy) a socket Each socket is assigned a transport protocol and a port number when it is created. A socket is also implicitly associated with an address (for example, the IP address of the machine on which the socket exists). Once sockets have been created, bi-directional communication is possible between pairs of sockets residing on two different machines, or sometimes on the same machine. Most operating systems provide an Application Programming Interface (API) corresponding to the basic operations that can be performed on a socket. Because the incorrect use of sockets can disrupt communications on the local machine, or even on the network, the socket API is privileged; its functions are operating system calls. The specific API provided depends on the operating system but in general Unix-based machines (including MacOS X and Linux) provide the Berkeley socket API (BSD Sockets) and Widows machines provide the Windows Sockets API (WSA, or Winsock) which is based very closely on a subset of BSD sockets (with a few Windows-specific extensions) Socket Domains, Types and Protocols Sockets are created with a particular socket type and protocol within a particular address domain. These parameters determine how the socket will behave. The domain determines what kind of addressing the socket will use, the socket type determines whether the socket deals with packets or with con-

4 96 CHAPTER 11. APPLICATION-LAYER ELEMENTS tinuous streams, and the protocol determines which network protocol will be used to send the data. Socket Domains The three most common socket domains are called INET, INET6 and LOCAL. INET and INET6 domain sockets use IPv4 or IPv6 addresses, respectively. Some operating systems provide local inter-process communication (IPC) using sockets within the LOCAL domain; since the filesystem is a simple, shared namespace accessible to all processes on the local machine, LOCAL domain sockets typically use filenames for their addresses. Socket Types and Protocols Socket types include DGRAM, STREAM and RAW. Protocols include UDP, TCP and SCTP. DGRAM (datagram) sockets provide unreliable, connectionless packet-based communication. They typically only support UDP as their transport protocol. STREAM sockets provide reliable, connection-oriented, full-duplex, streambased communication. Applications usually use TCP with a stream socket, although other protocols do exist. For example, the stream control transmission protocol (SCTP) implements reliable, message-oriented packet delivery in which each packet belongs to a particular stream. A single SCTP socket can support multiple independent streams of packets, allowing the separation of (for example) control packets and data packets within a single virtual circuit. RAW sockets are a third type that applications can use to bypass the TCP/IP transport layer. They behave almost like datagram sockets, except that the network layer headers are not removed from packets when they are delivered to the receiving application. (In principle this allows applications to implement their own IP-based transport protocols, but raw sockets are not often used because there is no support for them in Windows.) Operations on Sockets The following functions are provided by the BSD socket API. (Identical or equivalent functions are provided by the Windows Socket API.)

5 11.2. SOCKETS 97 socket(domain, type, protocol) creates a new socket. Initially the socket is not bound to any address. bind(socket, address) associates an address with a socket. The contents of the address depend on the domain in which the socket was created. Local domain sockets will use a structure containing a path name. Internet domain sockets use a structure containing an IP address and aportnumber. connect(socket, address) on the client side, assigns an available port number to a socket and then tries to open a connection to the given address. listen(socket) on the server side, listens for incoming connection requests to a connection-oriented socket. accept(socket, *address) on the server side, accepts an incoming connection request on a listening socket and creates a new socket on which client-server communication will take place. The address parameter is filled-in with the address of the client. (The original listening socket remains unconnected and available for further connections. The new socket is assigned an available port number for the duration of the connection.) sendto(socket, address, data) sends a packet containing the given data to the indicated address. This operation is typically used with UDP sockets. recv(socket, *buffer) recvfrom(socket, *buffer, *address) receives a packet and stores the payload data in the given buffer. In the second form, the sender s IP address and port number are stored in the given address structure. write(socket, data) read(socket, *data) send and receive a buffer of data over a streamoriented connection. poll(sockets[], timeout) checks on the state of the given sockets and informs the caller whether those sockets are readable, writable or have encountered some error condition. shutdown(socket, how) closes down all or part of a connection, disallowing further reads, writes or both. This is useful to signal the end of the connection to the peer, without actually closing the socket. (Closing

6 98 CHAPTER 11. APPLICATION-LAYER ELEMENTS the socket would would prevent any pending data from the peer being delivered.) close(socket) shuts down any connection associated with the socket and releases the socket s resources. No further communication is possible on the socket. gethostbyname(name) gethostbyaddr(address) finds the name(s) and address(es) associated with the given host name or host address. (More than one name and/or address can be associated with a given host.) 11.3 Network Programming In the rest of this chapter we will show some examples of network programming using Java. Java is useful for this because it is an object-oriented programming (OOP) language. OOP languages often provide libraries of classes that make it very convenient to write programs that are short, easy to read, and easy to understand. In the following sections, Java class names (and primitive types) that appear in the text are written using a typewriter font Internet Addresses The InetAddress class represents a network address. It includes a 32-bit int (the IP address) and a String (the host name). New instances are created through the static methods getbyname() and getbyaddress(). They attempt to resolve host names and addresses using DNS and will throw an UnknownHostException if no match is found. A SecurityException can also be thrown if, for example, the program does not have permission to access network resources. The local machine can be known by several names and addresses, including those assigned by a DHCP server, by DNS records, or by local configuration files (such as /etc/hosts on Unix-based machines). Any of these names or addresses can be used to find information about the local host. InetAddress local = InetAddress.getByName("localhost"); InetAddress local = InetAddress.getByAddress(0x7F000001);

7 11.3. NETWORK PROGRAMMING 99 A try-catch construct should always be placed around any network operation that might fail, so the preceding examples would be much better written: try { InetAddress local = InetAddress.getByName("localhost"); catch (UnknownHostException e) { println(e); Checking Open Ports Figure 11.1 shows how to use the method testport() to test for services on all TCP ports from 1 to The program can discover whether the local machine is, for example, a web server (HTTP uses port 80). On my machine the program gives the following result: Server on port 22 of localhost Server on port 80 of localhost Server on port 631 of localhost (Port 22 is the Secure Shell Protocol, port 80 is HTTP, and 631 is the Internet Printing Protocol, IPP.) Socket Information The above example creates a Socket to check whether or not a server is accepting connections on a particular port. The client port number does not matter so it is assigned randomly by the operating system, each time aconnectionismade,fromoutsidetherangeofportnumbersthathave official uses. The server s port number does matter, since it uniquely identifies the server to any clients that want to connect. After establishing the connection, however, the server will be communicating using a new socket, associated with a different port that is assigned randomly (leaving the socket associated with the well-known port available to accept more incoming client connections).

8 100 CHAPTER 11. APPLICATION-LAYER ELEMENTS // save this in a file called ProbePorts.java and then // compile with: javac ProbePorts.java // run with: java ProbePorts import java.io.*; import java.net.*; public class ProbePorts { static boolean testport(inetaddress addr, int p) { try { Socket s = new Socket(addr, p); System.out.println("Server on port " + p + " of " + addr.gethostname()); s.close(); catch (IOException e) { return false; return true; // host found public static void main(string[] args) { try { InetAddress local = InetAddress.getByName("localhost"); int i; for (i = 1; i < 1024; ++i) { testport(local, i); catch (UnknownHostException e) { System.out.println(e); Figure 11.1: Probing for servers on the local machine. The testport() method tries to open a connection to a specific port on a specific host. The main() method uses testport() to attempt a connection to each port in the range 1 to 1023 on the local host. Be warned that running this kind of port scan on any machine other than your own could be considered an attempt to break into that machine, with potentially serious consequences.

9 11.3. NETWORK PROGRAMMING 101 Sockets are therefore associated with two endpoints, one on the local machine and one on the remote machine where the server is running. Clientserver communication happens between these two endpoints. The Socket object can provide information about both of these addresses. The two methods InetAddress asocket.getlocaladdress() InetAddress asocket.getlocalport() return information about the local address associated with the socket, and the methods InetAddress asocket.getinetaddress() InetAddress asocket.getport() return information about the remote address Input and Output In Java, input/output can be done using objects called Streams and Buffers. A Stream is an object to which a stream of bytes can be written and/or from which a stream of bytes can be read. A Buffer is used to store those bytes as they are being transferred to or from a Stream. Buffers can be used to accumulate data that arrives in several pieces, which is an important capability in network programming where fragmentation can (and does) happen. Specific kinds of Stream and Buffer are used for different purposes. StringBuffers store sequences of characters and InputStreams support reading characters from devices such as the keyboard or a network communication endpoint. To create an InputStream for a network endpoint we can use Socket s method called getinputstream(). The example in the next section shows how this is used to create a client for the daytime protocol Daytime Protocol Client Figure 11.2 shows an example TCP client program. It uses the daytime protocol to retrieve a remote machine s local time as an ASCII string. The daytime protocol is very easy to use and gives a good demonstration of a simple client in action. We will use a known daytime server in the US called time.nist.gov.

10 102 CHAPTER 11. APPLICATION-LAYER ELEMENTS // save this in a file called DayTime.java and then // compile with: javac DayTime.java // run with: java DayTime import java.io.*; import java.net.*; public class DayTime { public static void main(string[] args) { try { Socket s = new Socket ("time.nist.gov", 13); InputStream stream = s.getinputstream(); StringBuffer time = new StringBuffer(); int c = 0; while ((c = stream.read())!= -1) time.append((char)c); String timestr = time.tostring().trim(); System.out.println("It is now " + timestr + " at nist.gov"); catch (Exception e) { System.out.println(e); Figure 11.2: A simple daytime protocol client. A connection is made to the daytime service running on port 13 of the machine time.nist.gov and a client socket created for the local communication endpoint. The server sends us its local time as an ASCII string. We can read that string from the socket as a sequence of ASCII characters and then print them. When run on my machine, this program prints: It is now :48: UTC(NIST) * at nist.gov As you can see from this very small client, much of the work in network programming is dealing with input/output. The BSD socket abstraction and Java s corresponding Socket object lets the system take care of most of the details of network programming.

11 11.4. GLOSSARY Socket Options Sockets provide many options to tune how network communication behaves. Connection-oriented Sockets have a NoDelay option which makes sure that data is sent as soon as possible. The default behaviour is to combine smaller pieces of data into a single larger piece before sending a packet, to make better use of network bandwidth. In some situations this additional latency, caused by waiting for additional data before sending a packet, is undesirable. The Linger option tells a socket that it should wait for a short while when closed beforedestroyingitselfifthereisanydatathathasnotyet been sent. This gives the network time to transmit that data. Timeout options are provided to make sure a Socket does not block for too long during long read() or write() operations. Finally, the transport layer buffer size can be adjusted to allow larger or smaller buffers according to network connection performance and the application s ability to deal with data in a timely manner Glossary accept establish a connection with a client that has requested a connection to a listening TCP socket. The result is a new socket, with a randomlyassigned port number, that provides the server with an endpoint for exchanging application data with the client. Application Programming Interface (API) Asetoffunctionsand data types associated with a library or operating system service. bi-directional capableofworkingintwodirections. Ofcommunication channels, capable of transmitting data in either direction between the two endpoints. bind associate an IP address and port number with a new socket. close release all resources associated with a socket and terminate any connection it is part of.

12 104 CHAPTER 11. APPLICATION-LAYER ELEMENTS connect send a request to a server socket listening on a well-known port asking it to establish two-way communication with the requester. If the server accepts the connection request then a new socket will be created on the server, connected to the client s socket, on which data exchange can occur. datagram aself-containedpacketofapplication-leveldata. domain an address space determined by the type of address that will be used within it. IPv4 and IPv6 have their own domains that use Internet addresses to identify devices connected to the Internet. A local domain can use filenames as addresses to identify endpoints belonging to processes running on the local machine. endpoint theendofacommunicationchannel,residingonanetwork device. Data transmission occurs from the device to the network via an endpoint, and data reception from the network to a device occur via another endpoint. Endpoints can be permanently associated with a single channel (e.g., when using TCP) or capable of sending/receiving datagrams to/from anywhere (e.g., when using UDP). full-duplex abi-directionalchannelcapableofsimultaneous,overlapping transmission and reception. inter-process communication (IPC) communicationbetweentwoprocesses. Remote processes must use network services for IPC. Local processes (running on the same machine) typically use either shared memory or network services with local addresses (such as filenames) to identify their endpoints. Internet Association for Assigned Numbers (IANA) thecentral registry for protocol names and numbers used in many Internet protocols. listen towaitforincomingconnectionrequestsonaserversocket. packet-based communication that happens one packet at a time, with each packet being a complete and indivisible unit of communication. (The opposite of stream-based communication.) port number thenumericidentifier,between0and65535,ofaparticular port. port alogicalstructurerepresentingthetransmitterorintendedrecipient of network communication. read receivedatafromastream-basedsocket.

13 11.4. GLOSSARY 105 receive receiveasinglepacketofdatafromadatagramsocket. send sendasinglepacketofdataonadatagramsocket. service anapplicationrunningonanetworkdevice,communicatingona specific port number, that provides some capability (storage, computation, etc.) to remote clients. shutdown terminaltransmissionand/orreceptionofdataonasocket without actually destroying the socket or releasing its resources. For example, shutting down a client socket will inform the remote server that the connection is being terminated, while allowing any data in-transit to still be received by the client. socket option a flag or value associated with a socket that can be set or cleared to change the behaviour of communication that uses the socket. Examples include switching off delayed transmission in TCP, and adjusting the sizes of the low-level transmit or receive buffers. socket type the characteristic of a socket that determines whether it is connection-oriented or connectionless. The semantics of communication and the operations permitted are fundamentally different between different types of socket. socket asoftwareabstractionforacommunicationendpointidentifiedby some address (such as a port number and IP address for Internet domain sockets, or a filesystem entry for local domain sockets). stream control transmission protocol (SCTP) aniptransportprotocol that provides UDP-like packet delivery with TCP-like reliability and sequencing. stream-based communication that happens one byte at a time, with no correspondence between the way data is divided during transmission and the way it will be presented during reception. A write operation can provide any number of bytes to the stream in a single operation, and a read operation can request any number of bytes from the stream in a single operation. (The opposite of packet-based communication.) system calls servicesprovidedbytheoperatingsystemtoanapplication, typically presented like library functions, that require elevated privilege. Examples include opening a file, forking a new process, and creating a socket.

14 106 CHAPTER 11. APPLICATION-LAYER ELEMENTS virtual circuits the computer networking equivalent of a switched telephone circuit. A virtual circuit is established between two peers for the purpose of an extended exchange of data between them. Establishing a virtual circuit is more expensive than sending a single datagram, but subsequent exchange of data over a virtual circuit is more efficient than it would be if the data were sent one datagram at a time. write senddatatoastream-basedsocket. Copyright 2014 E.W. Cooper and I.K. Piumarta All rights reserved. Permission is granted for you to download, save and then view or print one copy of this document for personal study purposes. No other form of publishing, duplication or redistribution is permitted without explicit prior consent.