Home
小杰的博客 Prev Page Prev Page
?
Main Page
Table of content
Copyright
Addison-Wesley Professional Computing Series
Foreword
Preface
Introduction
Changes from the Second Edition
Using This Book
Source Code and Errata Availability
Acknowledgments
Part 1: Introduction and TCP/IP
Chapter 1. Introduction
1.1 Introduction
1.2 A Simple Daytime Client
1.3 Protocol Independence
1.4 Error Handling: Wrapper Functions
1.5 A Simple Daytime Server
1.6 Roadmap to Client/Server Examples in the Text
1.7 OSI Model
1.8 BSD Networking History
1.9 Test Networks and Hosts
1.10 Unix Standards
1.11 64-Bit Architectures
1.12 Summary
Exercises
Chapter 2. The Transport Layer: TCP, UDP, and SCTP
2.1 Introduction
2.2 The Big Picture
2.3 User Datagram Protocol (UDP)
2.4 Transmission Control Protocol (TCP)
2.5 Stream Control Transmission Protocol (SCTP)
2.6 TCP Connection Establishment and Termination
2.7 TIME_WAIT State
2.8 SCTP Association Establishment and Termination
2.9 Port Numbers
2.10 TCP Port Numbers and Concurrent Servers
2.11 Buffer Sizes and Limitations
2.12 Standard Internet Services
2.13 Protocol Usage by Common Internet Applications
2.14 Summary
Exercises
Part 2: Elementary Sockets
Chapter 3. Sockets Introduction
3.1 Introduction
3.2 Socket Address Structures
3.3 Value-Result Arguments
3.4 Byte Ordering Functions
3.5 Byte Manipulation Functions
3.6 'inet_aton', 'inet_addr', and 'inet_ntoa' Functions
3.7 'inet_pton' and 'inet_ntop' Functions
3.8 'sock_ntop' and Related Functions
3.9 'readn', 'writen', and 'readline' Functions
3.10 Summary
Exercises
Chapter 4. Elementary TCP Sockets
4.1 Introduction
4.2 'socket' Function
4.3 'connect' Function
4.4 'bind' Function
4.5 'listen' Function
4.6 'accept' Function
4.7 'fork' and 'exec' Functions
4.8 Concurrent Servers
4.9 'close' Function
4.10 'getsockname' and 'getpeername' Functions
4.11 Summary
Exercises
Chapter 5. TCP Client/Server Example
5.1 Introduction
5.2 TCP Echo Server: 'main' Function
5.3 TCP Echo Server: 'str_echo' Function
5.4 TCP Echo Client: 'main' Function
5.5 TCP Echo Client: 'str_cli' Function
5.6 Normal Startup
5.7 Normal Termination
5.8 POSIX Signal Handling
5.9 Handling 'SIGCHLD' Signals
5.10 'wait' and 'waitpid' Functions
5.11 Connection Abort before 'accept' Returns
5.12 Termination of Server Process
5.13 'SIGPIPE' Signal
5.14 Crashing of Server Host
5.15 Crashing and Rebooting of Server Host
5.16 Shutdown of Server Host
5.17 Summary of TCP Example
5.18 Data Format
5.19 Summary
Exercises
Chapter 6. I/O Multiplexing: The 'select' and 'poll' Functions
6.1 Introduction
6.2 I/O Models
6.3 'select' Function
6.4 'str_cli' Function (Revisited)
6.5 Batch Input and Buffering
6.6 'shutdown' Function
6.7 'str_cli' Function (Revisited Again)
6.8 TCP Echo Server (Revisited)
6.9 'pselect' Function
6.10 'poll' Function
6.11 TCP Echo Server (Revisited Again)
6.12 Summary
Exercises
Chapter 7. Socket Options
7.1 Introduction
7.2 'getsockopt' and 'setsockopt' Functions
7.3 Checking if an Option Is Supported and Obtaining the Default
7.4 Socket States
7.5 Generic Socket Options
7.6 IPv4 Socket Options
7.7 ICMPv6 Socket Option
7.8 IPv6 Socket Options
7.9 TCP Socket Options
7.10 SCTP Socket Options
7.11 'fcntl' Function
7.12 Summary
Exercises
Chapter 8. Elementary UDP Sockets
8.1 Introduction
8.2 'recvfrom' and 'sendto' Functions
8.3 UDP Echo Server: 'main' Function
8.4 UDP Echo Server: 'dg_echo' Function
8.5 UDP Echo Client: 'main' Function
8.6 UDP Echo Client: 'dg_cli' Function
8.7 Lost Datagrams
8.8 Verifying Received Response
8.9 Server Not Running
8.10 Summary of UDP Example
8.11 'connect' Function with UDP
8.12 'dg_cli' Function (Revisited)
8.13 Lack of Flow Control with UDP
8.14 Determining Outgoing Interface with UDP
8.15 TCP and UDP Echo Server Using 'select'
8.16 Summary
Exercises
Chapter 9. Elementary SCTP Sockets
9.1 Introduction
9.2 Interface Models
9.3 'sctp_bindx' Function
9.4 'sctp_connectx' Function
9.5 'sctp_getpaddrs' Function
9.6 'sctp_freepaddrs' Function
9.7 'sctp_getladdrs' Function
9.8 'sctp_freeladdrs' Function
9.9 'sctp_sendmsg' Function
9.10 'sctp_recvmsg' Function
9.11 'sctp_opt_info' Function
9.12 'sctp_peeloff' Function
9.13 'shutdown' Function
9.14 Notifications
9.15 Summary
Exercises
Chapter 10. SCTP Client/Server Example
10.1 Introduction
10.2 SCTP One-to-Many-Style Streaming Echo Server: 'main' Function
10.3 SCTP One-to-Many-Style Streaming Echo Client: 'main' Function
10.4 SCTP Streaming Echo Client: 'str_cli' Function
10.5 Exploring Head-of-Line Blocking
10.6 Controlling the Number of Streams
10.7 Controlling Termination
10.8 Summary
Exercises
Chapter 11. Name and Address Conversions
11.1 Introduction
11.2 Domain Name System (DNS)
11.3 'gethostbyname' Function
11.4 'gethostbyaddr' Function
11.5 'getservbyname' and 'getservbyport' Functions
11.6 'getaddrinfo' Function
11.7 'gai_strerror' Function
11.8 'freeaddrinfo' Function
11.9 'getaddrinfo' Function: IPv6
11.10 'getaddrinfo' Function: Examples
11.11 'host_serv' Function
11.12 'tcp_connect' Function
11.13 'tcp_listen' Function
11.14 'udp_client' Function
11.15 'udp_connect' Function
11.16 'udp_server' Function
11.17 'getnameinfo' Function
11.18 Re-entrant Functions
11.19 'gethostbyname_r' and 'gethostbyaddr_r' Functions
11.20 Obsolete IPv6 Address Lookup Functions
11.21 Other Networking Information
11.22 Summary
Exercises
Part 3: Advanced Sockets
Chapter 12. IPv4 and IPv6 Interoperability
12.1 Introduction
12.2 IPv4 Client, IPv6 Server
12.3 IPv6 Client, IPv4 Server
12.4 IPv6 Address-Testing Macros
12.5 Source Code Portability
12.6 Summary
Exercises
Chapter 13. Daemon Processes and the 'inetd' Superserver
13.1 Introduction
13.2 'syslogd' Daemon
13.3 'syslog' Function
13.4 'daemon_init' Function
13.5 'inetd' Daemon
13.6 'daemon_inetd' Function
13.7 Summary
Exercises
Chapter 14. Advanced I/O Functions
14.1 Introduction
14.2 Socket Timeouts
14.3 'recv' and 'send' Functions
14.4 'readv' and 'writev' Functions
14.5 'recvmsg' and 'sendmsg' Functions
14.6 Ancillary Data
14.7 How Much Data Is Queued?
14.8 Sockets and Standard I/O
14.9 Advanced Polling
14.10 Summary
Exercises
Chapter 15. Unix Domain Protocols
15.1 Introduction
15.2 Unix Domain Socket Address Structure
15.3 'socketpair' Function
15.4 Socket Functions
15.5 Unix Domain Stream Client/Server
15.6 Unix Domain Datagram Client/Server
15.7 Passing Descriptors
15.8 Receiving Sender Credentials
15.9 Summary
Exercises
Chapter 16. Nonblocking I/O
16.1 Introduction
16.2 Nonblocking Reads and Writes: 'str_cli' Function (Revisited)
16.3 Nonblocking 'connect'
16.4 Nonblocking 'connect:' Daytime Client
16.5 Nonblocking 'connect:' Web Client
16.6 Nonblocking 'accept'
16.7 Summary
Exercises
Chapter 17. 'ioctl' Operations
17.1 Introduction
17.2 'ioctl' Function
17.3 Socket Operations
17.4 File Operations
17.5 Interface Configuration
17.6 'get_ifi_info' Function
17.7 Interface Operations
17.8 ARP Cache Operations
17.9 Routing Table Operations
17.10 Summary
Exercises
Chapter 18. Routing Sockets
18.1 Introduction
18.2 Datalink Socket Address Structure
18.3 Reading and Writing
18.4 'sysctl' Operations
18.5 'get_ifi_info' Function (Revisited)
18.6 Interface Name and Index Functions
18.7 Summary
Exercises
Chapter 19. Key Management Sockets
19.1 Introduction
19.2 Reading and Writing
19.3 Dumping the Security Association Database (SADB)
19.4 Creating a Static Security Association (SA)
19.5 Dynamically Maintaining SAs
19.6 Summary
Exercises
Chapter 20. Broadcasting
20.1 Introduction
20.2 Broadcast Addresses
20.3 Unicast versus Broadcast
20.4 'dg_cli' Function Using Broadcasting
20.5 Race Conditions
20.6 Summary
Exercises
Chapter 21. Multicasting
21.1 Introduction
21.2 Multicast Addresses
21.3 Multicasting versus Broadcasting on a LAN
21.4 Multicasting on a WAN
21.5 Source-Specific Multicast
21.6 Multicast Socket Options
21.7 'mcast_join' and Related Functions
21.8 'dg_cli' Function Using Multicasting
21.9 Receiving IP Multicast Infrastructure Session Announcements
21.10 Sending and Receiving
21.11 Simple Network Time Protocol (SNTP)
21.12 Summary
Exercises
Chapter 22. Advanced UDP Sockets
22.1 Introduction
22.2 Receiving Flags, Destination IP Address, and Interface Index
22.3 Datagram Truncation
22.4 When to Use UDP Instead of TCP
22.5 Adding Reliability to a UDP Application
22.6 Binding Interface Addresses
22.7 Concurrent UDP Servers
22.8 IPv6 Packet Information
22.9 IPv6 Path MTU Control
22.10 Summary
Exercises
Chapter 23. Advanced SCTP Sockets
23.1 Introduction
23.2 An Autoclosing One-to-Many-Style Server
23.3 Partial Delivery
23.4 Notifications
23.5 Unordered Data
23.6 Binding a Subset of Addresses
23.7 Determining Peer and Local Address Information
23.8 Finding an Association ID Given an IP Address
23.9 Heartbeating and Address Failure
23.10 Peeling Off an Association
23.11 Controlling Timing
23.12 When to Use SCTP Instead of TCP
23.13 Summary
Exercises
Chapter 24. Out-of-Band Data
24.1 Introduction
24.2 TCP Out-of-Band Data
24.3 'sockatmark' Function
24.4 TCP Out-of-Band Data Recap
24.5 Summary
Exercises
Chapter 25. Signal-Driven I/O
25.1 Introduction
25.2 Signal-Driven I/O for Sockets
25.3 UDP Echo Server Using 'SIGIO'
25.4 Summary
Exercises
Chapter 26. Threads
26.1 Introduction
26.2 Basic Thread Functions: Creation and Termination
26.3 'str_cli' Function Using Threads
26.4 TCP Echo Server Using Threads
26.5 Thread-Specific Data
26.6 Web Client and Simultaneous Connections (Continued)
26.7 Mutexes: Mutual Exclusion
26.8 Condition Variables
26.9 Web Client and Simultaneous Connections (Continued)
26.10 Summary
Exercises
Chapter 27. IP Options
27.1 Introduction
27.2 IPv4 Options
27.3 IPv4 Source Route Options
27.4 IPv6 Extension Headers
27.5 IPv6 Hop-by-Hop Options and Destination Options
27.6 IPv6 Routing Header
27.7 IPv6 Sticky Options
27.8 Historical IPv6 Advanced API
27.9 Summary
Exercises
Chapter 28. Raw Sockets
28.1 Introduction
28.2 Raw Socket Creation
28.3 Raw Socket Output
28.4 Raw Socket Input
28.5 'ping' Program
28.6 'traceroute' Program
28.7 An ICMP Message Daemon
28.8 Summary
Exercises
Chapter 29. Datalink Access
29.1 Introduction
29.2 BSD Packet Filter (BPF)
29.3 Datalink Provider Interface (DLPI)
29.4 Linux: 'SOCK_PACKET' and 'PF_PACKET'
29.5 'libpcap': Packet Capture Library
29.6 'libnet': Packet Creation and Injection Library
29.7 Examining the UDP Checksum Field
29.8 Summary
Exercises
Chapter 30. Client/Server Design Alternatives
30.1 Introduction
30.2 TCP Client Alternatives
30.3 TCP Test Client
30.4 TCP Iterative Server
30.5 TCP Concurrent Server, One Child per Client
30.6 TCP Preforked Server, No Locking Around 'accept'
30.7 TCP Preforked Server, File Locking Around 'accept'
30.8 TCP Preforked Server, Thread Locking Around 'accept'
30.9 TCP Preforked Server, Descriptor Passing
30.10 TCP Concurrent Server, One Thread per Client
30.11 TCP Prethreaded Server, per-Thread 'accept'
30.12 TCP Prethreaded Server, Main Thread 'accept'
30.13 Summary
Exercises
Chapter 31. Streams
31.1 Introduction
31.2 Overview
31.3 'getmsg' and 'putmsg' Functions
31.4 'getpmsg' and 'putpmsg' Functions
31.5 'ioctl' Function
31.6 Transport Provider Interface (TPI)
31.7 Summary
Exercises
Appendix A. IPv4, IPv6, ICMPv4, and ICMPv6
A.1 Introduction
A.2 IPv4 Header
A.3 IPv6 Header
A.4 IPv4 Addresses
A.5 IPv6 Addresses
A.6 Internet Control Message Protocols (ICMPv4 and ICMPv6)
Appendix B. Virtual Networks
B.1 Introduction
B.2 The MBone
B.3 The 6bone
B.4 IPv6 Transition: 6to4
Appendix C. Debugging Techniques
C.1 System Call Tracing
C.2 Standard Internet Services
C.3 'sock' Program
C.4 Small Test Programs
C.5 'tcpdump' Program
C.6 'netstat' Program
C.7 'lsof' Program
Appendix D. Miscellaneous Source Code
D.1 'unp.h' Header
D.2 'config.h' Header
D.3 Standard Error Functions
Appendix E. Solutions to Selected Exercises
Chapter 1
Chapter 2
Chapter 3
Chapter 4
Chapter 5
Chapter 6
Chapter 7
Chapter 8
Chapter 9
Chapter 10
Chapter 11
Chapter 12
Chapter 13
Chapter 14
Chapter 15
Chapter 16
Chapter 17
Chapter 18
Chapter 20
Chapter 21
Chapter 22
Chapter 24
Chapter 25
Chapter 26
Chapter 27
Chapter 28
Chapter 29
Chapter 30
Chapter 31
Bibliography
?
[ Team LiB ] Previous Section Next Section

24.2 TCP Out-of-Band Data

TCP does not have true out-of-band data. Instead, TCP provides an urgent mode. Assume a process has written N bytes of data to a TCP socket and that data is queued by TCP in the socket send buffer, waiting to be sent to the peer. We show this in Figure 24.1 and have labeled the data bytes 1 through N.

Figure 24.1. Socket send buffer containing data to send.

graphics/24fig01.gif

The process now writes a single byte of out-of-band data, containing the ASCII character a, using the send function and the MSG_OOB flag.


send(fd, "a", 1, MSG_OOB);

TCP places the data in the next available position in the socket send buffer and sets its urgent pointer for this connection to be the next available location. We show this in Figure 24.2 and have labeled the out-of-band byte "OOB."

Figure 24.2. Socket send buffer after one byte of out-of-band data is written by application.

graphics/24fig02.gif

TCP's urgent pointer has a sequence number one greater than the byte of data that is written with the MSG_OOB flag. As discussed on pp. 292鈥?96 of TCPv1, this is an historical artifact that is now emulated by all implementations. As long as the sending TCP and the receiving TCP agree on the interpretation of TCP's urgent pointer, all is fine.

Given the state of the TCP socket send buffer shown in Figure 24.2, the next segment sent by TCP will have its URG flag set in the TCP header and the urgent offset field in the TCP header will point to the byte following the out-of-band byte. But this segment may or may not contain the byte that we have labeled as OOB. Whether the OOB byte is sent depends on the number of bytes ahead of it in the socket send buffer, the segment size that TCP is sending to the peer, and the current window advertised by the peer.

We have used the terms urgent pointer and urgent offset. At the TCP level, the two are different. The 16-bit value in the TCP header is called the urgent offset and it must be added to the sequence number field in the header to obtain the 32-bit urgent pointer. TCP looks at the urgent offset only if another bit in the header is set, and this bit is called the URG flag. From a programming perspective, we need not worry about this detail and just refer to TCP's urgent pointer.

This is an important characteristic of TCP's urgent mode: The TCP header indicates that the sender has entered urgent mode (i.e., the URG flag is set along with the urgent offset), but the actual byte of data referred to by the urgent pointer need not be sent. Indeed, if the sending TCP is stopped by flow control (the receiver's socket receive buffer is full, so its TCP has advertised a window of 0 to the sending TCP), the urgent notification is sent without any data (pp. 1016鈥?017 of TCPv2), as we will show in Figures 24.10 and 24.11. This is one reason why applications use TCP's urgent mode (i.e., out-of-band data): The urgent notification is always sent to the peer TCP, even if the flow of data is stopped by TCP's flow control.

What happens if we send multiple bytes of out-of-band data, as in


send(fd, "abc", 3, MSG_OOB);

In this example, TCP's urgent pointer points one beyond the final byte; that is, the final byte (the c) is considered the out-of-band byte.

Now that we have covered the sending of out-of-band data, let's look at it from the receiver's side:

  1. When TCP receives a segment with the URG flag set, the urgent pointer is examined to see whether it refers to new out-of-band data, that is, whether this is the first time TCP's urgent mode has referenced this particular byte in the stream of data from the sender to the receiver. It is common for the sending TCP to send multiple segments (typically over a short period of time) containing the URG flag, but with the urgent pointer pointing to the same byte of data. Only the first of these segments causes the receiving process to be notified that new out-of-band data has arrived.

  2. The receiving process is notified when a new urgent pointer arrives. First the SIGURG signal is sent to the owner of the socket, assuming either fcntl or ioctl has been called to establish an owner for the socket (Figure 7.20). Second, if the process is blocked in a call to select waiting for this socket descriptor to have an exception condition, select returns.

    These two potential notifications to the receiving process take place when a new urgent pointer arrives, regardless of whether the actual byte of data pointed to by the urgent pointer has arrived at the receiving TCP.

    There is only one OOB mark; if a new OOB byte arrives before the old OOB byte was read, the old byte is discarded.

  3. When the actual byte of data pointed to by the urgent pointer arrives at the receiving TCP, the data byte can be pulled out-of-band or left inline. By default, the SO_OOBINLINE socket option is not set for a socket, so the single byte of data is not placed into the socket receive buffer. Instead, the data byte is placed into a separate one-byte out-of-band buffer for this connection (pp. 986鈥?88 of TCPv2). The only way for the process to read from this special one-byte buffer is to call recv, recvfrom, or recvmsg and specify the MSG_OOB flag. If a new OOB byte arrives before the old byte is read, the previous value in this buffer is discarded.

    If, however, the process sets the SO_OOBINLINE socket option, then the single byte of data referred to by TCP's urgent pointer is left in the normal socket receive buffer. The process cannot specify the MSG_OOB flag to read the data byte in this case. The process will know when it reaches this byte of data by checking the out-of-band mark for this connection, as we will describe in Section 24.3.

Some of the following errors are possible:

  1. If the process asks for out-of-band data (e.g., specifying the MSG_OOB flag), but the peer has not sent any, EINVAL is returned.

  2. If the process has been notified that the peer has sent an out-of-band byte (e.g., by SIGURG or select), and the process tries to read it but that byte has not yet arrived, EWOULDBLOCK is returned. All the process can do at this point is read from the socket receive buffer (possibly discarding the data if it has no room to store the data), to make space in the buffer so that the peer TCP can send the out-of-band byte.

  3. If the process tries to read the same out-of-band byte multiple times, EINVAL is returned.

  4. If the process has set the SO_OOBINLINE socket option and then tries to read the out-of-band data by specifying MSG_OOB, EINVAL is returned.

Simple Example Using SIGURG

We now show a trivial example of sending and receiving out-of-band data. Figure 24.3 shows the sending program.

Nine bytes are sent, with a one-second sleep between each output operation. The purpose of the pause is to let each write or send be transmitted as a single TCP segment and received as such by the other end. We'll talk later about some of the timing considerations with out-of-band data. When we run this program, we see the expected output.


macosx % tcpsend01 freebsd4 9999
wrote 3 bytes of normal data
wrote 1 byte of OOB data
wrote 2 bytes of normal data
wrote 1 byte of OOB data
wrote 2 bytes of normal data
Figure 24.3 Simple out-of-band sending program.

oob/tcpsend01.c

 1 #include    "unp.h"

 2 int
 3 main(int argc, char **argv)
 4 {
 5     int     sockfd;

 6     if (argc != 3)
 7         err_quit("usage: tcpsend01 <host> <port#>");

 8     sockfd = Tcp_connect(argv[1], argv[2]);

 9     Write(sockfd, "123", 3);
10     printf("wrote 3 bytes of normal data\n");
11     sleep(1);

12     Send(sockfd, "4", 1, MSG_OOB);
13     printf("wrote 1 byte of OOB data\n");
14     sleep(1);

15     Write(sockfd, "56", 2);
16     printf("wrote 2 bytes of normal data\n");
17     sleep(1);

18     Send(sockfd, "7", 1, MSG_OOB);
19     printf("wrote 1 byte of OOB data\n");
20     sleep(1);

21     Write(sockfd, "89", 2);
22     printf("wrote 2 bytes of normal data\n");
23     sleep(1);

24     exit(0);
25 }

Figure 24.4 is the receiving program.

Establish signal handler and socket owner

16鈥?7 The signal handler for SIGURG is established, and fcntl sets the owner of the connected socket.

Notice that we do not establish the signal handler until accept returns. There is a small probability that out-of-band data can arrive after our TCP completes the three-way handshake, but before accept returns, which we would miss. But if we established the signal handler before calling accept and also set the owner of the listening socket (which carries over to the connected socket), then if out-of-band data arrives before accept returns, our signal handler won't yet have a value for connfd. If this scenario is important for the application, it should initialize connfd to 鈥?, check for this value in the signal handler, and if true, just set a flag for the main loop to check after accept returns. Alternately, it could block the signal around the call to accept, but this is subject to all the signal race conditions we discussed in Section 20.5.

Figure 24.4 Simple out-of-band receiving program.

oob/tcprecv01.c

 1 #include    "unp.h"

 2 int     listenfd, connfd;

 3 void    sig_urg(int);

 4 int
 5 main(int argc, char **argv)
 6 {
 7     int     n;
 8     char    buff[100];

 9     if (argc == 2)
10         listenfd = Tcp_listen(NULL, argv[1], NULL);
11     else if (argc == 3)
12         listenfd = Tcp_listen(argv[1], argv[2], NULL);
13     else
14         err_quit("usage: tcprecv01 [ <host> ] <port#>");

15     connfd = Accept(listenfd, NULL, NULL);

16     Signal(SIGURG, sig_urg);
17     Fcntl(connfd, F_SETOWN, getpid());

18     for ( ; ; ) {
19         if ( (n = Read(connfd, buff, sizeof(buff) - 1)) == 0) {
20             printf("received EOF\n");
21             exit(0);
22         }
23         buff[n] = 0;            /* null terminate */
24         printf("read %d bytes: %s\n", n, buff);
25     }
26 }

27 void
28 sig_urg(int signo)
29 {
30     int     n;
31     char    buff[100];

32     printf("SIGURG received\n");
33     n = Recv(connfd, buff, sizeof(buff) - 1, MSG_OOB);
34     buff[n] = 0;                /* null terminate */
35     printf("read %d OOB byte: %s\n", n, buff);
36 }

18鈥?5 The process reads from the socket, printing each string that is returned by read. When the sender terminates the connection, the receiver then terminates.

SIGURG handler

27鈥?6 Our signal handler calls printf, reads the out-of-band byte by specifying the MSG_OOB flag, and then prints the returned data. Notice that we ask for up to 100 bytes in the call to recv, but as we will see shortly, only 1 byte is ever returned as out-of-band data.

As stated earlier, calling the unsafe printf function from a signal handler is not recommended. We do it just to see what's happening with our programs.

Here is the output when we run the receiving program, and then run the sending program from Figure 24.3:


freebsd4 % tcprecv01 9999
read 3 bytes: 123
SIGURG received
read 1 OOB byte: 4
read 2 bytes: 56
SIGURG received
read 1 OOB byte: 7
read 2 bytes: 89
received EOF

The results are as we expect. Each sending of out-of-band data by the sender generates SIGURG for the receiver, which then reads the single out-of-band byte.

Simple Example Using select

We now redo our out-of-band receiver to use select instead of the SIGURG signal. Figure 24.5 is the receiving program.

15鈥?0 The process calls select while waiting for either normal data (the read set, rset) or out-of-band data (the exception set, xset). In each case, the received data is printed.

When we run this program and then run the same sending program as earlier (Figure 24.3), we encounter the following error:


freebsd4 % tcprecv02 9999
read 3 bytes: 123
read 1 OOB byte: 4
recv error: Invalid argument

The problem is that select indicates an exception condition until the process reads beyond the out-of-band data (pp. 530鈥?31 of TCPv2). We cannot read the out-of-band data more than once because after we read it the first time, the kernel clears the one-byte out-of-band buffer. When we call recv specifying the MSG_OOB flag the second time, it returns EINVAL.

Figure 24.5 Receiving program that (incorrectly) uses select to be notified of out-of-band data.

oob/tcprecv02.c

 1 #include    "unp.h"

 2 int
 3 main(int argc, char **argv)
 4 {
 5     int     listenfd, connfd, n;
 6     char    buff[100];
 7     fd_set  rset, xset;

 8     if (argc == 2)
 9         listenfd = Tcp_listen(NULL, argv[1], NULL);
10     else if (argc == 3)
11         listenfd = Tcp_listen(argv[1], argv[2], NULL);
12     else
13         err_quit("usage: tcprecv02 [ <host> ] <port#>");

14     connfd = Accept(listenfd, NULL, NULL);

15     FD_ZERO(&rset);
16     FD_ZERO(&xset);
17     for ( ; ; ) {
18         FD_SET(connfd, &rset);
19         FD_SET(connfd, &xset);

20         Select(connfd + 1, &rset, NULL, &xset, NULL);

21         if (FD_ISSET(connfd, &xset)) {
22             n = Recv(connfd, buff, sizeof(buff) - 1, MSG_OOB);
23             buff[n] = 0;        /* null terminate */
24             printf("read %d OOB byte: %s\n", n, buff);
25         }

26         if (FD_ISSET(connfd, &rset)) {
27             if ( (n = Read(connfd, buff, sizeof(buff) - 1)) == 0) {
28                 printf("received EOF\n");
29                 exit(0);
30             }
31             buff[n] = 0;        /* null terminate */
32             printf("read %d bytes: %s\n", n, buff);
33         }
34     }
35 }

The solution is to select for an exception condition only after reading normal data. Figure 24.6 is a modification of Figure 24.5 that handles this scenario correctly.

5 We declare a new variable named justreadoob, which indicates whether we just read out-of-band data or not. This flag determines whether or not to select for an exception condition.

26鈥?7 When we set the justreadoob flag, we must also clear the bit for this descriptor in the exception set.

The program now works as expected.

Figure 24.6 Modification of Figure 24.5 to select for an exception condition correctly.

oob/tcprecv03.c

 1 #include    "unp.h"

 2 int
 3 main(int argc, char **argv)
 4 {
 5     int     listenfd, connfd, n, justreadoob = 0;
 6     char    buff[100];
 7     fd_set  rset, xset;

 8     if (argc == 2)
 9         listenfd = Tcp_listen(NULL, argv[1], NULL);
10     else if (argc == 3)
11         listenfd = Tcp_listen(argv[1], argv[2], NULL);
12     else
13         err_quit("usage: tcprecv03 [ <host> ] <port#>");

14     connfd = Accept(listenfd, NULL, NULL);

15     FD_ZERO(&rset);
16     FD_ZERO(&xset);
17     for ( ; ; ) {
18         FD_SET(connfd, &rset);
19         if (justreadoob == 0)
20             FD_SET(connfd, &xset);

21         Select(connfd + 1, &rset, NULL, &xset, NULL);

22         if (FD_ISSET(connfd, &xset)) {
23             n = Recv(connfd, buff, sizeof(buff) - 1, MSG_OOB);
24             buff[n] = 0;        /* null terminate */
25             printf("read %d OOB byte: %s\n", n, buff);
26             justreadoob = 1;
27             FD_CLR(connfd, &xset);
28         }

29         if (FD_ISSET(connfd, &rset)) {
30             if ( (n = Read(connfd, buff, sizeof(buff) - 1) ) == 0) {
31                 printf("received EOF\n");
32                 exit(0);
33             }
34             buff[n] = 0;        /* null terminate */
35             printf("read %d bytes: %s\n", n, buff);
36             justreadoob = 0;
37         }
38     }
39 }
[ Team LiB ] Previous Section Next Section
Converted from CHM to HTML with chm2web Pro 2.85 (unicode)