man2/select.2

   1 .\" Hey Emacs! This file is -*- nroff -*- source.
   2 .\"
   3 .\" This manpage is copyright (C) 1992 Drew Eckhardt,
   4 .\"                 copyright (C) 1995 Michael Shields.
   5 .\"
   6 .\" Permission is granted to make and distribute verbatim copies of this
   7 .\" manual provided the copyright notice and this permission notice are
   8 .\" preserved on all copies.
   9 .\"
  10 .\" Permission is granted to copy and distribute modified versions of this
  11 .\" manual under the conditions for verbatim copying, provided that the
  12 .\" entire resulting derived work is distributed under the terms of a
  13 .\" permission notice identical to this one.
  14 .\"
  15 .\" Since the Linux kernel and libraries are constantly changing, this
  16 .\" manual page may be incorrect or out-of-date.  The author(s) assume no
  17 .\" responsibility for errors or omissions, or for damages resulting from
  18 .\" the use of the information contained herein.  The author(s) may not
  19 .\" have taken the same level of care in the production of this manual,
  20 .\" which is licensed free of charge, as they might when working
  21 .\" professionally.
  22 .\"
  23 .\" Formatted or processed versions of this manual, if unaccompanied by
  24 .\" the source, must acknowledge the copyright and authors of this work.
  25 .\"
  26 .\" Modified 1993-07-24 by Rik Faith <faith@cs.unc.edu>
  27 .\" Modified 1995-05-18 by Jim Van Zandt <jrv@vanzandt.mv.com>
  28 .\" Sun Feb 11 14:07:00 MET 1996  Martin Schulze  <joey@linux.de>
  29 .\"     * layout slightly modified
  30 .\"
  31 .\" Modified Mon Oct 21 23:05:29 EDT 1996 by Eric S. Raymond <esr@thyrsus.com>
  32 .\" Modified Thu Feb 24 01:41:09 CET 2000 by aeb
  33 .\" Modified Thu Feb  9 22:32:09 CET 2001 by bert hubert <ahu@ds9a.nl>, aeb
  34 .\" Modified Mon Nov 11 14:35:00 PST 2002 by Ben Woodard <ben@zork.net>
  35 .\" 2005-03-11, mtk, modified pselect() text (it is now a system
  36 .\"     call in 2.6.16.
  37 .\"
  38 .TH SELECT 2 2007-07-26 "Linux" "Linux Programmer's Manual"
  39 .SH NAME
  40 select, pselect, FD_CLR, FD_ISSET, FD_SET, FD_ZERO \-
  41 synchronous I/O multiplexing
  42 .SH SYNOPSIS
  43 .nf
  44 /* According to POSIX.1-2001 */
  45 .br
  46 .B #include <sys/select.h>
  47 .sp
  48 /* According to earlier standards */
  49 .br
  50 .B #include <sys/time.h>
  51 .br
  52 .B #include <sys/types.h>
  53 .br
  54 .B #include <unistd.h>
  55 .sp
  56 .BI "int select(int " nfds ", fd_set *" readfds ", fd_set *" writefds ,
  57 .BI "           fd_set *" exceptfds ", struct timeval *" timeout );
  58 .sp
  59 .BI "void FD_CLR(int " fd ", fd_set *" set );
  60 .br
  61 .BI "int  FD_ISSET(int " fd ", fd_set *" set );
  62 .br
  63 .BI "void FD_SET(int " fd ", fd_set *" set );
  64 .br
  65 .BI "void FD_ZERO(fd_set *" set );
  66 .sp
  67 .B #include <sys/select.h>
  68 .sp
  69 .BI "int pselect(int " nfds ", fd_set *" readfds ", fd_set *" writefds ,
  70 .BI "            fd_set *" exceptfds ", const struct timespec *" timeout ,
  71 .BI "            const sigset_t *" sigmask );
  72 .fi
  73 .sp
  74 .in -4n
  75 Feature Test Macro Requirements for glibc (see
  76 .BR feature_test_macros (7)):
  77 .in
  78 .sp
  79 .BR pselect ():
  80 _POSIX_C_SOURCE\ >=\ 200112L || _XOPEN_SOURCE\ >=\ 600
  81 .SH DESCRIPTION
  82 .BR select ()
  83 and
  84 .BR pselect ()
  85 allow a program to monitor multiple file descriptors,
  86 waiting until one or more of the file descriptors become "ready"
  87 for some class of I/O operation (e.g., input possible).
  88 A file descriptor is considered ready if it is possible to
  89 perform the corresponding I/O operation (e.g.,
  90 .BR read (2))
  91 without blocking.
  92 .PP
  93 The operation of
  94 .BR select ()
  95 and
  96 .BR pselect ()
  97 is identical, with three differences:
  98 .TP
  99 (i)
 100 .BR select ()
 101 uses a timeout that is a
 102 .I struct timeval
 103 (with seconds and microseconds), while
 104 .BR pselect ()
 105 uses a
 106 .I struct timespec
 107 (with seconds and nanoseconds).
 108 .TP
 109 (ii)
 110 .BR select ()
 111 may update the
 112 .I timeout
 113 argument to indicate how much time was left.
 114 .BR pselect ()
 115 does not change this argument.
 116 .TP
 117 (iii)
 118 .BR select ()
 119 has no
 120 .I sigmask
 121 argument, and behaves as
 122 .BR pselect ()
 123 called with NULL
 124 .IR sigmask .
 125 .PP
 126 Three independent sets of file descriptors are watched.
 127 Those listed in
 128 .I readfds
 129 will be watched to see if characters become
 130 available for reading (more precisely, to see if a read will not
 131 block; in particular, a file descriptor is also ready on end-of-file),
 132 those in
 133 .I writefds
 134 will be watched to see if a write will not block, and
 135 those in
 136 .I exceptfds
 137 will be watched for exceptions.
 138 On exit, the sets are modified in place
 139 to indicate which file descriptors actually changed status.
 140 Each of the three file descriptor sets may be specified as NULL
 141 if no file descriptors are to be watched for the corresponding class
 142 of events.
 143 .PP
 144 Four macros are provided to manipulate the sets.
 145 .BR FD_ZERO ()
 146 clears a set.
 147 .BR FD_SET ()
 148 and
 149 .BR FD_CLR ()
 150 respectively add and remove a given file descriptor from a set.
 151 .BR FD_ISSET ()
 152 tests to see if a file descriptor is part of the set;
 153 this is useful after
 154 .BR select ()
 155 returns.
 156 .PP
 157 .I nfds
 158 is the highest-numbered file descriptor in any of the three sets, plus 1.
 159 .PP
 160 .I timeout
 161 is an upper bound on the amount of time elapsed before
 162 .BR select ()
 163 returns.
 164 It may be zero, causing
 165 .BR select ()
 166 to return immediately.
 167 (This is useful for polling.)
 168 If
 169 .I timeout
 170 is NULL (no timeout),
 171 .BR select ()
 172 can block indefinitely.
 173 .PP
 174 .I sigmask
 175 is a pointer to a signal mask (see
 176 .BR sigprocmask (2));
 177 if it is not NULL, then
 178 .BR pselect ()
 179 first replaces the current signal mask by the one pointed to by
 180 .IR sigmask ,
 181 then does the "select" function, and then restores the original
 182 signal mask.
 183 .PP
 184 Other than the difference in the precision of the
 185 .I timeout
 186 argument, the following
 187 .BR pselect ()
 188 call:
 189 .nf
 190
 191     ready = pselect(nfds, &readfds, &writefds, &exceptfds,
 192                     timeout, &sigmask);
 193
 194 .fi
 195 is equivalent to
 196 .I atomically
 197 executing the following calls:
 198 .nf
 199
 200     sigset_t origmask;
 201
 202     sigprocmask(SIG_SETMASK, &sigmask, &origmask);
 203     ready = select(nfds, &readfds, &writefds, &exceptfds, timeout);
 204     sigprocmask(SIG_SETMASK, &origmask, NULL);
 205 .fi
 206 .PP
 207 The reason that
 208 .BR pselect ()
 209 is needed is that if one wants to wait for either a signal
 210 or for a file descriptor to become ready, then
 211 an atomic test is needed to prevent race conditions.
 212 (Suppose the signal handler sets a global flag and
 213 returns.
 214 Then a test of this global flag followed by a call of
 215 .BR select ()
 216 could hang indefinitely if the signal arrived just after the test
 217 but just before the call.
 218 By contrast,
 219 .BR pselect ()
 220 allows one to first block signals, handle the signals that have come in,
 221 then call
 222 .BR pselect ()
 223 with the desired
 224 .IR sigmask ,
 225 avoiding the race.)
 226 .SS "The timeout"
 227 The time structures involved are defined in
 228 .I <sys/time.h>
 229 and look like
 230
 231 .in +4n
 232 .nf
 233 struct timeval {
 234     long    tv_sec;         /* seconds */
 235     long    tv_usec;        /* microseconds */
 236 };
 237 .fi
 238 .in
 239
 240 and
 241
 242 .in +4n
 243 .nf
 244 struct timespec {
 245     long    tv_sec;         /* seconds */
 246     long    tv_nsec;        /* nanoseconds */
 247 };
 248 .fi
 249 .in
 250
 251 (However, see below on the POSIX.1-2001 versions.)
 252 .PP
 253 Some code calls
 254 .BR select ()
 255 with all three sets empty,
 256 .I nfds
 257 zero, and a non-NULL
 258 .I timeout
 259 as a fairly portable way to sleep with subsecond precision.
 260 .PP
 261 On Linux,
 262 .BR select ()
 263 modifies
 264 .I timeout
 265 to reflect the amount of time not slept; most other implementations
 266 do not do this.
 267 (POSIX.1-2001 permits either behavior.)
 268 This causes problems both when Linux code which reads
 269 .I timeout
 270 is ported to other operating systems, and when code is ported to Linux
 271 that reuses a \fIstruct timeval\fP for multiple
 272 .BR select ()s
 273 in a loop without reinitializing it.
 274 Consider
 275 .I timeout
 276 to be undefined after
 277 .BR select ()
 278 returns.
 279 .\" .PP - it is rumored that:
 280 .\" On BSD, when a timeout occurs, the file descriptor bits are not changed.
 281 .\" - it is certainly true that:
 282 .\" Linux follows SUSv2 and sets the bit masks to zero upon a timeout.
 283 .SH "RETURN VALUE"
 284 On success,
 285 .BR select ()
 286 and
 287 .BR pselect ()
 288 return the number of file descriptors contained in the three returned
 289 descriptor sets (that is, the total number of bits that are set in
 290 .IR readfds ,
 291 .IR writefds ,
 292 .IR exceptfds )
 293 which may be zero if the timeout expires before anything interesting happens.
 294 On error, \-1 is returned, and
 295 .I errno
 296 is set appropriately; the sets and
 297 .I timeout
 298 become undefined, so do not
 299 rely on their contents after an error.
 300 .SH ERRORS
 301 .TP
 302 .B EBADF
 303 An invalid file descriptor was given in one of the sets.
 304 (Perhaps a file descriptor that was already closed,
 305 or one on which an error has occurred.)
 306 .TP
 307 .B EINTR
 308 A signal was caught.
 309 .TP
 310 .B EINVAL
 311 .I nfds
 312 is negative or the value contained within
 313 .I timeout
 314 is invalid.
 315 .TP
 316 .B ENOMEM
 317 unable to allocate memory for internal tables.
 318 .SH VERSIONS
 319 .BR pselect ()
 320 was added to Linux in kernel 2.6.16.
 321 Prior to this,
 322 .BR pselect ()
 323 was emulated in glibc (but see BUGS).
 324 .SH "CONFORMING TO"
 325 .BR select ()
 326 conforms to POSIX.1-2001 and
 327 4.4BSD
 328 .RB ( select ()
 329 first appeared in 4.2BSD).
 330 Generally portable to/from
 331 non-BSD systems supporting clones of the BSD socket layer (including
 332 System V variants).
 333 However, note that the System V variant typically
 334 sets the timeout variable before exit, but the BSD variant does not.
 335 .PP
 336 .BR pselect ()
 337 is defined in POSIX.1g, and in
 338 POSIX.1-2001.
 339 .SH NOTES
 340 An
 341 .I fd_set
 342 is a fixed size buffer.
 343 Executing
 344 .BR FD_CLR ()
 345 or
 346 .BR FD_SET ()
 347 with a value of
 348 .I fd
 349 that is negative or is equal to or larger than
 350 .B FD_SETSIZE
 351 will result
 352 in undefined behavior.
 353 Moreover, POSIX requires
 354 .I fd
 355 to be a valid file descriptor.
 356
 357 Concerning the types involved, the classical situation is that
 358 the two fields of a
 359 .I timeval
 360 structure are typed as
 361 .I long
 362 (as shown above), and the structure is defined in
 363 .IR <sys/time.h> .
 364 The POSIX.1-2001 situation is
 365
 366 .in +4n
 367 .nf
 368 struct timeval {
 369     time_t         tv_sec;     /* seconds */
 370     suseconds_t    tv_usec;    /* microseconds */
 371 };
 372 .fi
 373 .in
 374
 375 where the structure is defined in
 376 .I <sys/select.h>
 377 and the data types
 378 .I time_t
 379 and
 380 .I suseconds_t
 381 are defined in
 382 .IR <sys/types.h> .
 383 .LP
 384 Concerning prototypes, the classical situation is that one should
 385 include
 386 .I <time.h>
 387 for
 388 .BR select ().
 389 The POSIX.1-2001 situation is that one should include
 390 .I <sys/select.h>
 391 for
 392 .BR select ()
 393 and
 394 .BR pselect ().
 395 Libc4 and libc5 do not have a
 396 .I <sys/select.h>
 397 header; under glibc 2.0 and later this header exists.
 398 Under glibc 2.0 it unconditionally gives the wrong prototype for
 399 .BR pselect (),
 400 under glibc 2.1-2.2.1 it gives
 401 .BR pselect ()
 402 when
 403 .B _GNU_SOURCE
 404 is defined, under glibc 2.2.2-2.2.4 it gives it when
 405 .B _XOPEN_SOURCE
 406 is defined and has a value of 600 or larger.
 407 No doubt, since POSIX.1-2001, it should give the prototype by default.
 408 .SS "Linux Notes"
 409 The Linux
 410 .BR pselect ()
 411 system call modifies its
 412 .I timeout
 413 argument.
 414 However, the glibc wrapper function hides this behavior
 415 by using a local variable for the timeout argument that
 416 is passed to the system call.
 417 Thus, the glibc
 418 .BR pselect ()
 419 function does not modify its timeout argument;
 420 this is the behavior required by POSIX.1-2001.
 421 .SH BUGS
 422 Glibc 2.0 provided a version of
 423 .BR pselect ()
 424 that did not take a
 425 .I sigmask
 426 argument.
 427
 428 Since version 2.1, glibc has provided an emulation of
 429 .BR pselect ()
 430 that is implemented using
 431 .BR sigprocmask (2)
 432 and
 433 .BR select ().
 434 This implementation remains vulnerable to the very race condition that
 435 .BR pselect ()
 436 was designed to prevent.
 437 On systems that lack
 438 .BR pselect ()
 439 reliable (and more portable) signal trapping can be achieved
 440 using the self-pipe trick
 441 (where a signal handler writes a byte to a pipe whose other end
 442 is monitored by
 443 .BR select ()
 444 in the main program.)
 445
 446 Under Linux,
 447 .BR select ()
 448 may report a socket file descriptor as "ready for reading", while
 449 nevertheless a subsequent read blocks.
 450 This could for example
 451 happen when data has arrived but upon examination has wrong
 452 checksum and is discarded.
 453 There may be other circumstances
 454 in which a file descriptor is spuriously reported as ready.
 455 .\" Stevens discusses a case where accept can block after select
 456 .\" returns successfully because of an intervening RST from the client.
 457 Thus it may be safer to use
 458 .B O_NONBLOCK
 459 on sockets that should not block.
 460 .\" Maybe the kernel should have returned EIO in such a situation?
 461
 462 On Linux,
 463 .BR select ()
 464 also modifies
 465 .I timeout
 466 if the call is interrupted by a signal handler (i.e., the
 467 .B EINTR
 468 error return).
 469 This is not permitted by POSIX.1-2001.
 470 The Linux
 471 .BR pselect ()
 472 system call has the same behavior,
 473 but the glibc wrapper hides this behavior by internally copying the
 474 .I timeout
 475 to a local variable and passing that variable to the system call.
 476 .SH EXAMPLE
 477 .nf
 478 #include <stdio.h>
 479 #include <stdlib.h>
 480 #include <sys/time.h>
 481 #include <sys/types.h>
 482 #include <unistd.h>
 483
 484 int
 485 main(void)
 486 {
 487     fd_set rfds;
 488     struct timeval tv;
 489     int retval;
 490
 491     /* Watch stdin (fd 0) to see when it has input. */
 492     FD_ZERO(&rfds);
 493     FD_SET(0, &rfds);
 494
 495     /* Wait up to five seconds. */
 496     tv.tv_sec = 5;
 497     tv.tv_usec = 0;
 498
 499     retval = select(1, &rfds, NULL, NULL, &tv);
 500     /* Don't rely on the value of tv now! */
 501
 502     if (retval == \-1)
 503         perror("select()");
 504     else if (retval)
 505         printf("Data is available now.\\n");
 506         /* FD_ISSET(0, &rfds) will be true. */
 507     else
 508         printf("No data within five seconds.\\n");
 509
 510     exit(EXIT_SUCCESS);
 511 }
 512 .fi
 513 .SH "SEE ALSO"
 514 For a tutorial with discussion and examples, see
 515 .BR select_tut (2).
 516 .LP
 517 For vaguely related stuff, see
 518 .BR accept (2),
 519 .BR connect (2),
 520 .BR poll (2),
 521 .BR read (2),
 522 .BR recv (2),
 523 .BR send (2),
 524 .BR sigprocmask (2),
 525 .BR write (2),
 526 .BR epoll (7),
 527 .BR time (7)