man2/select_tut.2

   1 .\" This manpage is copyright (C) 2001 Paul Sheer.
   2 .\"
   3 .\" SPDX-License-Identifier: Linux-man-pages-copyleft
   4 .\"
   5 .\" very minor changes, aeb
   6 .\"
   7 .\" Modified 5 June 2002, Michael Kerrisk <mtk.manpages@gmail.com>
   8 .\" 2006-05-13, mtk, removed much material that is redundant with select.2
   9 .\"             various other changes
  10 .\" 2008-01-26, mtk, substantial changes and rewrites
  11 .\"
  12 .TH SELECT_TUT 2 (date) "Linux man-pages (unreleased)"
  13 .SH NAME
  14 select, pselect \- synchronous I/O multiplexing
  15 .SH LIBRARY
  16 Standard C library
  17 .RI ( libc ", " \-lc )
  18 .SH SYNOPSIS
  19 See
  20 .BR select (2)
  21 .SH DESCRIPTION
  22 The
  23 .BR select ()
  24 and
  25 .BR pselect ()
  26 system calls are used to efficiently monitor multiple file descriptors,
  27 to see if any of them is, or becomes, "ready";
  28 that is, to see whether I/O becomes possible,
  29 or an "exceptional condition" has occurred on any of the file descriptors.
  30 .P
  31 This page provides background and tutorial information
  32 on the use of these system calls.
  33 For details of the arguments and semantics of
  34 .BR select ()
  35 and
  36 .BR pselect (),
  37 see
  38 .BR select (2).
  39 .\"
  40 .SS Combining signal and data events
  41 .BR pselect ()
  42 is useful if you are waiting for a signal as well as
  43 for file descriptor(s) to become ready for I/O.
  44 Programs that receive signals
  45 normally use the signal handler only to raise a global flag.
  46 The global flag will indicate that the event must be processed
  47 in the main loop of the program.
  48 A signal will cause the
  49 .BR select ()
  50 (or
  51 .BR pselect ())
  52 call to return with \fIerrno\fP set to \fBEINTR\fP.
  53 This behavior is essential so that signals can be processed
  54 in the main loop of the program, otherwise
  55 .BR select ()
  56 would block indefinitely.
  57 .P
  58 Now, somewhere
  59 in the main loop will be a conditional to check the global flag.
  60 So we must ask:
  61 what if a signal arrives after the conditional, but before the
  62 .BR select ()
  63 call?
  64 The answer is that
  65 .BR select ()
  66 would block indefinitely, even though an event is actually pending.
  67 This race condition is solved by the
  68 .BR pselect ()
  69 call.
  70 This call can be used to set the signal mask to a set of signals
  71 that are to be received only within the
  72 .BR pselect ()
  73 call.
  74 For instance, let us say that the event in question
  75 was the exit of a child process.
  76 Before the start of the main loop, we
  77 would block \fBSIGCHLD\fP using
  78 .BR sigprocmask (2).
  79 Our
  80 .BR pselect ()
  81 call would enable
  82 .B SIGCHLD
  83 by using an empty signal mask.
  84 Our program would look like:
  85 .P
  86 .EX
  87 static volatile sig_atomic_t got_SIGCHLD = 0;
  88 \&
  89 static void
  90 child_sig_handler(int sig)
  91 {
  92     got_SIGCHLD = 1;
  93 }
  94 \&
  95 int
  96 main(int argc, char *argv[])
  97 {
  98     sigset_t sigmask, empty_mask;
  99     struct sigaction sa;
 100     fd_set readfds, writefds, exceptfds;
 101     int r;
 102 \&
 103     sigemptyset(&sigmask);
 104     sigaddset(&sigmask, SIGCHLD);
 105     if (sigprocmask(SIG_BLOCK, &sigmask, NULL) == \-1) {
 106         perror("sigprocmask");
 107         exit(EXIT_FAILURE);
 108     }
 109 \&
 110     sa.sa_flags = 0;
 111     sa.sa_handler = child_sig_handler;
 112     sigemptyset(&sa.sa_mask);
 113     if (sigaction(SIGCHLD, &sa, NULL) == \-1) {
 114         perror("sigaction");
 115         exit(EXIT_FAILURE);
 116     }
 117 \&
 118     sigemptyset(&empty_mask);
 119 \&
 120     for (;;) {          /* main loop */
 121         /* Initialize readfds, writefds, and exceptfds
 122            before the pselect() call. (Code omitted.) */
 123 \&
 124         r = pselect(nfds, &readfds, &writefds, &exceptfds,
 125                     NULL, &empty_mask);
 126         if (r == \-1 && errno != EINTR) {
 127             /* Handle error */
 128         }
 129 \&
 130         if (got_SIGCHLD) {
 131             got_SIGCHLD = 0;
 132 \&
 133             /* Handle signalled event here; e.g., wait() for all
 134                terminated children. (Code omitted.) */
 135         }
 136 \&
 137         /* main body of program */
 138     }
 139 }
 140 .EE
 141 .SS Practical
 142 So what is the point of
 143 .BR select ()?
 144 Can't I just read and write to my file descriptors whenever I want?
 145 The point of
 146 .BR select ()
 147 is that it watches
 148 multiple descriptors at the same time and properly puts the process to
 149 sleep if there is no activity.
 150 UNIX programmers often find
 151 themselves in a position where they have to handle I/O from more than one
 152 file descriptor where the data flow may be intermittent.
 153 If you were to merely create a sequence of
 154 .BR read (2)
 155 and
 156 .BR write (2)
 157 calls, you would
 158 find that one of your calls may block waiting for data from/to a file
 159 descriptor, while another file descriptor is unused though ready for I/O.
 160 .BR select ()
 161 efficiently copes with this situation.
 162 .SS Select law
 163 Many people who try to use
 164 .BR select ()
 165 come across behavior that is
 166 difficult to understand and produces nonportable or borderline results.
 167 For instance, the above program is carefully written not to
 168 block at any point, even though it does not set its file descriptors to
 169 nonblocking mode.
 170 It is easy to introduce
 171 subtle errors that will remove the advantage of using
 172 .BR select (),
 173 so here is a list of essentials to watch for when using
 174 .BR select ().
 175 .TP 4
 176 1.
 177 You should always try to use
 178 .BR select ()
 179 without a timeout.
 180 Your program
 181 should have nothing to do if there is no data available.
 182 Code that
 183 depends on timeouts is not usually portable and is difficult to debug.
 184 .TP
 185 2.
 186 The value \fInfds\fP must be properly calculated for efficiency as
 187 explained above.
 188 .TP
 189 3.
 190 No file descriptor must be added to any set if you do not intend
 191 to check its result after the
 192 .BR select ()
 193 call, and respond appropriately.
 194 See next rule.
 195 .TP
 196 4.
 197 After
 198 .BR select ()
 199 returns, all file descriptors in all sets
 200 should be checked to see if they are ready.
 201 .TP
 202 5.
 203 The functions
 204 .BR read (2),
 205 .BR recv (2),
 206 .BR write (2),
 207 and
 208 .BR send (2)
 209 do \fInot\fP necessarily read/write the full amount of data
 210 that you have requested.
 211 If they do read/write the full amount, it's
 212 because you have a low traffic load and a fast stream.
 213 This is not always going to be the case.
 214 You should cope with the case of your
 215 functions managing to send or receive only a single byte.
 216 .TP
 217 6.
 218 Never read/write only in single bytes at a time unless you are really
 219 sure that you have a small amount of data to process.
 220 It is extremely
 221 inefficient not to read/write as much data as you can buffer each time.
 222 The buffers in the example below are 1024 bytes although they could
 223 easily be made larger.
 224 .TP
 225 7.
 226 Calls to
 227 .BR read (2),
 228 .BR recv (2),
 229 .BR write (2),
 230 .BR send (2),
 231 and
 232 .BR select ()
 233 can fail with the error
 234 \fBEINTR\fP,
 235 and calls to
 236 .BR read (2),
 237 .BR recv (2),
 238 .BR write (2),
 239 and
 240 .BR send (2)
 241 can fail with
 242 .I errno
 243 set to \fBEAGAIN\fP (\fBEWOULDBLOCK\fP).
 244 These results must be properly managed (not done properly above).
 245 If your program is not going to receive any signals, then
 246 it is unlikely you will get \fBEINTR\fP.
 247 If your program does not set nonblocking I/O,
 248 you will not get \fBEAGAIN\fP.
 249 .\" Nonetheless, you should still cope with these errors for completeness.
 250 .TP
 251 8.
 252 Never call
 253 .BR read (2),
 254 .BR recv (2),
 255 .BR write (2),
 256 or
 257 .BR send (2)
 258 with a buffer length of zero.
 259 .TP
 260 9.
 261 If the functions
 262 .BR read (2),
 263 .BR recv (2),
 264 .BR write (2),
 265 and
 266 .BR send (2)
 267 fail with errors other than those listed in \fB7.\fP,
 268 or one of the input functions returns 0, indicating end of file,
 269 then you should \fInot\fP pass that file descriptor to
 270 .BR select ()
 271 again.
 272 In the example below,
 273 I close the file descriptor immediately, and then set it to \-1
 274 to prevent it being included in a set.
 275 .TP
 276 10.
 277 The timeout value must be initialized with each new call to
 278 .BR select (),
 279 since some operating systems modify the structure.
 280 .BR pselect ()
 281 however does not modify its timeout structure.
 282 .TP
 283 11.
 284 Since
 285 .BR select ()
 286 modifies its file descriptor sets,
 287 if the call is being used in a loop,
 288 then the sets must be reinitialized before each call.
 289 .\" "I have heard" does not fill me with confidence, and doesn't
 290 .\" belong in a man page, so I've commented this point out.
 291 .\" .TP
 292 .\" 11.
 293 .\" I have heard that the Windows socket layer does not cope with OOB data
 294 .\" properly.
 295 .\" It also does not cope with
 296 .\" .BR select ()
 297 .\" calls when no file descriptors are set at all.
 298 .\" Having no file descriptors set is a useful
 299 .\" way to sleep the process with subsecond precision by using the timeout.
 300 .\" (See further on.)
 301 .SH RETURN VALUE
 302 See
 303 .BR select (2).
 304 .SH NOTES
 305 Generally speaking,
 306 all operating systems that support sockets also support
 307 .BR select ().
 308 .BR select ()
 309 can be used to solve
 310 many problems in a portable and efficient way that naive programmers try
 311 to solve in a more complicated manner using
 312 threads, forking, IPCs, signals, memory sharing, and so on.
 313 .P
 314 The
 315 .BR poll (2)
 316 system call has the same functionality as
 317 .BR select (),
 318 and is somewhat more efficient when monitoring sparse
 319 file descriptor sets.
 320 It is nowadays widely available, but historically was less portable than
 321 .BR select ().
 322 .P
 323 The Linux-specific
 324 .BR epoll (7)
 325 API provides an interface that is more efficient than
 326 .BR select (2)
 327 and
 328 .BR poll (2)
 329 when monitoring large numbers of file descriptors.
 330 .SH EXAMPLES
 331 Here is an example that better demonstrates the true utility of
 332 .BR select ().
 333 The listing below is a TCP forwarding program that forwards
 334 from one TCP port to another.
 335 .P
 336 .\" SRC BEGIN (select.c)
 337 .EX
 338 #include <arpa/inet.h>
 339 #include <errno.h>
 340 #include <netinet/in.h>
 341 #include <signal.h>
 342 #include <stdio.h>
 343 #include <stdlib.h>
 344 #include <string.h>
 345 #include <sys/select.h>
 346 #include <sys/socket.h>
 347 #include <unistd.h>
 348 \&
 349 static int forward_port;
 350 \&
 351 #undef max
 352 #define max(x, y) ((x) > (y) ? (x) : (y))
 353 \&
 354 static int
 355 listen_socket(int listen_port)
 356 {
 357     int                 lfd;
 358     int                 yes;
 359     struct sockaddr_in  addr;
 360 \&
 361     lfd = socket(AF_INET, SOCK_STREAM, 0);
 362     if (lfd == \-1) {
 363         perror("socket");
 364         return \-1;
 365     }
 366 \&
 367     yes = 1;
 368     if (setsockopt(lfd, SOL_SOCKET, SO_REUSEADDR,
 369                    &yes, sizeof(yes)) == \-1)
 370     {
 371         perror("setsockopt");
 372         close(lfd);
 373         return \-1;
 374     }
 375 \&
 376     memset(&addr, 0, sizeof(addr));
 377     addr.sin_port = htons(listen_port);
 378     addr.sin_family = AF_INET;
 379     if (bind(lfd, (struct sockaddr *) &addr, sizeof(addr)) == \-1) {
 380         perror("bind");
 381         close(lfd);
 382         return \-1;
 383     }
 384 \&
 385     printf("accepting connections on port %d\en", listen_port);
 386     listen(lfd, 10);
 387     return lfd;
 388 }
 389 \&
 390 static int
 391 connect_socket(int connect_port, char *address)
 392 {
 393     int                 cfd;
 394     struct sockaddr_in  addr;
 395 \&
 396     cfd = socket(AF_INET, SOCK_STREAM, 0);
 397     if (cfd == \-1) {
 398         perror("socket");
 399         return \-1;
 400     }
 401 \&
 402     memset(&addr, 0, sizeof(addr));
 403     addr.sin_port = htons(connect_port);
 404     addr.sin_family = AF_INET;
 405 \&
 406     if (!inet_aton(address, (struct in_addr *) &addr.sin_addr.s_addr)) {
 407         fprintf(stderr, "inet_aton(): bad IP address format\en");
 408         close(cfd);
 409         return \-1;
 410     }
 411 \&
 412     if (connect(cfd, (struct sockaddr *) &addr, sizeof(addr)) == \-1) {
 413         perror("connect()");
 414         shutdown(cfd, SHUT_RDWR);
 415         close(cfd);
 416         return \-1;
 417     }
 418     return cfd;
 419 }
 420 \&
 421 #define SHUT_FD1 do {                                \e
 422                      if (fd1 >= 0) {                 \e
 423                          shutdown(fd1, SHUT_RDWR);   \e
 424                          close(fd1);                 \e
 425                          fd1 = \-1;                   \e
 426                      }                               \e
 427                  } while (0)
 428 \&
 429 #define SHUT_FD2 do {                                \e
 430                      if (fd2 >= 0) {                 \e
 431                          shutdown(fd2, SHUT_RDWR);   \e
 432                          close(fd2);                 \e
 433                          fd2 = \-1;                   \e
 434                      }                               \e
 435                  } while (0)
 436 \&
 437 #define BUF_SIZE 1024
 438 \&
 439 int
 440 main(int argc, char *argv[])
 441 {
 442     int      h;
 443     int      ready, nfds;
 444     int      fd1 = \-1, fd2 = \-1;
 445     int      buf1_avail = 0, buf1_written = 0;
 446     int      buf2_avail = 0, buf2_written = 0;
 447     char     buf1[BUF_SIZE], buf2[BUF_SIZE];
 448     fd_set   readfds, writefds, exceptfds;
 449     ssize_t  nbytes;
 450 \&
 451     if (argc != 4) {
 452         fprintf(stderr, "Usage\en\etfwd <listen\-port> "
 453                 "<forward\-to\-port> <forward\-to\-ip\-address>\en");
 454         exit(EXIT_FAILURE);
 455     }
 456 \&
 457     signal(SIGPIPE, SIG_IGN);
 458 \&
 459     forward_port = atoi(argv[2]);
 460 \&
 461     h = listen_socket(atoi(argv[1]));
 462     if (h == \-1)
 463         exit(EXIT_FAILURE);
 464 \&
 465     for (;;) {
 466         nfds = 0;
 467 \&
 468         FD_ZERO(&readfds);
 469         FD_ZERO(&writefds);
 470         FD_ZERO(&exceptfds);
 471         FD_SET(h, &readfds);
 472         nfds = max(nfds, h);
 473 \&
 474         if (fd1 > 0 && buf1_avail < BUF_SIZE)
 475             FD_SET(fd1, &readfds);
 476             /* Note: nfds is updated below, when fd1 is added to
 477                exceptfds. */
 478         if (fd2 > 0 && buf2_avail < BUF_SIZE)
 479             FD_SET(fd2, &readfds);
 480 \&
 481         if (fd1 > 0 && buf2_avail \- buf2_written > 0)
 482             FD_SET(fd1, &writefds);
 483         if (fd2 > 0 && buf1_avail \- buf1_written > 0)
 484             FD_SET(fd2, &writefds);
 485 \&
 486         if (fd1 > 0) {
 487             FD_SET(fd1, &exceptfds);
 488             nfds = max(nfds, fd1);
 489         }
 490         if (fd2 > 0) {
 491             FD_SET(fd2, &exceptfds);
 492             nfds = max(nfds, fd2);
 493         }
 494 \&
 495         ready = select(nfds + 1, &readfds, &writefds, &exceptfds, NULL);
 496 \&
 497         if (ready == \-1 && errno == EINTR)
 498             continue;
 499 \&
 500         if (ready == \-1) {
 501             perror("select()");
 502             exit(EXIT_FAILURE);
 503         }
 504 \&
 505         if (FD_ISSET(h, &readfds)) {
 506             socklen_t addrlen;
 507             struct sockaddr_in client_addr;
 508             int fd;
 509 \&
 510             addrlen = sizeof(client_addr);
 511             memset(&client_addr, 0, addrlen);
 512             fd = accept(h, (struct sockaddr *) &client_addr, &addrlen);
 513             if (fd == \-1) {
 514                 perror("accept()");
 515             } else {
 516                 SHUT_FD1;
 517                 SHUT_FD2;
 518                 buf1_avail = buf1_written = 0;
 519                 buf2_avail = buf2_written = 0;
 520                 fd1 = fd;
 521                 fd2 = connect_socket(forward_port, argv[3]);
 522                 if (fd2 == \-1)
 523                     SHUT_FD1;
 524                 else
 525                     printf("connect from %s\en",
 526                            inet_ntoa(client_addr.sin_addr));
 527 \&
 528                 /* Skip any events on the old, closed file
 529                    descriptors. */
 530 \&
 531                 continue;
 532             }
 533         }
 534 \&
 535         /* NB: read OOB data before normal reads. */
 536 \&
 537         if (fd1 > 0 && FD_ISSET(fd1, &exceptfds)) {
 538             char c;
 539 \&
 540             nbytes = recv(fd1, &c, 1, MSG_OOB);
 541             if (nbytes < 1)
 542                 SHUT_FD1;
 543             else
 544                 send(fd2, &c, 1, MSG_OOB);
 545         }
 546         if (fd2 > 0 && FD_ISSET(fd2, &exceptfds)) {
 547             char c;
 548 \&
 549             nbytes = recv(fd2, &c, 1, MSG_OOB);
 550             if (nbytes < 1)
 551                 SHUT_FD2;
 552             else
 553                 send(fd1, &c, 1, MSG_OOB);
 554         }
 555         if (fd1 > 0 && FD_ISSET(fd1, &readfds)) {
 556             nbytes = read(fd1, buf1 + buf1_avail,
 557                           BUF_SIZE \- buf1_avail);
 558             if (nbytes < 1)
 559                 SHUT_FD1;
 560             else
 561                 buf1_avail += nbytes;
 562         }
 563         if (fd2 > 0 && FD_ISSET(fd2, &readfds)) {
 564             nbytes = read(fd2, buf2 + buf2_avail,
 565                           BUF_SIZE \- buf2_avail);
 566             if (nbytes < 1)
 567                 SHUT_FD2;
 568             else
 569                 buf2_avail += nbytes;
 570         }
 571         if (fd1 > 0 && FD_ISSET(fd1, &writefds) && buf2_avail > 0) {
 572             nbytes = write(fd1, buf2 + buf2_written,
 573                            buf2_avail \- buf2_written);
 574             if (nbytes < 1)
 575                 SHUT_FD1;
 576             else
 577                 buf2_written += nbytes;
 578         }
 579         if (fd2 > 0 && FD_ISSET(fd2, &writefds) && buf1_avail > 0) {
 580             nbytes = write(fd2, buf1 + buf1_written,
 581                            buf1_avail \- buf1_written);
 582             if (nbytes < 1)
 583                 SHUT_FD2;
 584             else
 585                 buf1_written += nbytes;
 586         }
 587 \&
 588         /* Check if write data has caught read data. */
 589 \&
 590         if (buf1_written == buf1_avail)
 591             buf1_written = buf1_avail = 0;
 592         if (buf2_written == buf2_avail)
 593             buf2_written = buf2_avail = 0;
 594 \&
 595         /* One side has closed the connection, keep
 596            writing to the other side until empty. */
 597 \&
 598         if (fd1 < 0 && buf1_avail \- buf1_written == 0)
 599             SHUT_FD2;
 600         if (fd2 < 0 && buf2_avail \- buf2_written == 0)
 601             SHUT_FD1;
 602     }
 603     exit(EXIT_SUCCESS);
 604 }
 605 .EE
 606 .\" SRC END
 607 .P
 608 The above program properly forwards most kinds of TCP connections
 609 including OOB signal data transmitted by \fBtelnet\fP servers.
 610 It handles the tricky problem of having data flow in both directions
 611 simultaneously.
 612 You might think it more efficient to use a
 613 .BR fork (2)
 614 call and devote a thread to each stream.
 615 This becomes more tricky than you might suspect.
 616 Another idea is to set nonblocking I/O using
 617 .BR fcntl (2).
 618 This also has its problems because you end up using
 619 inefficient timeouts.
 620 .P
 621 The program does not handle more than one simultaneous connection at a
 622 time, although it could easily be extended to do this with a linked list
 623 of buffers\[em]one for each connection.
 624 At the moment, new
 625 connections cause the current connection to be dropped.
 626 .SH SEE ALSO
 627 .BR accept (2),
 628 .BR connect (2),
 629 .BR poll (2),
 630 .BR read (2),
 631 .BR recv (2),
 632 .BR select (2),
 633 .BR send (2),
 634 .BR sigprocmask (2),
 635 .BR write (2),
 636 .BR epoll (7)
 637 .\" .SH AUTHORS
 638 .\" This man page was written by Paul Sheer.