1 .\" This manpage is copyright (C) 1992 Drew Eckhardt,
2 .\" copyright (C) 1995 Michael Shields.
4 .\" %%%LICENSE_START(VERBATIM)
5 .\" Permission is granted to make and distribute verbatim copies of this
6 .\" manual provided the copyright notice and this permission notice are
7 .\" preserved on all copies.
9 .\" Permission is granted to copy and distribute modified versions of this
10 .\" manual under the conditions for verbatim copying, provided that the
11 .\" entire resulting derived work is distributed under the terms of a
12 .\" permission notice identical to this one.
14 .\" Since the Linux kernel and libraries are constantly changing, this
15 .\" manual page may be incorrect or out-of-date. The author(s) assume no
16 .\" responsibility for errors or omissions, or for damages resulting from
17 .\" the use of the information contained herein. The author(s) may not
18 .\" have taken the same level of care in the production of this manual,
19 .\" which is licensed free of charge, as they might when working
22 .\" Formatted or processed versions of this manual, if unaccompanied by
23 .\" the source, must acknowledge the copyright and authors of this work.
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
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
38 .TH SELECT 2 2014-12-31 "Linux" "Linux Programmer's Manual"
40 select, pselect, FD_CLR, FD_ISSET, FD_SET, FD_ZERO \-
41 synchronous I/O multiplexing
44 /* According to POSIX.1-2001 */
46 .B #include <sys/select.h>
48 /* According to earlier standards */
50 .B #include <sys/time.h>
52 .B #include <sys/types.h>
54 .B #include <unistd.h>
56 .BI "int select(int " nfds ", fd_set *" readfds ", fd_set *" writefds ,
57 .BI " fd_set *" exceptfds ", struct timeval *" timeout );
59 .BI "void FD_CLR(int " fd ", fd_set *" set );
61 .BI "int FD_ISSET(int " fd ", fd_set *" set );
63 .BI "void FD_SET(int " fd ", fd_set *" set );
65 .BI "void FD_ZERO(fd_set *" set );
67 .B #include <sys/select.h>
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 );
75 Feature Test Macro Requirements for glibc (see
76 .BR feature_test_macros (7)):
80 _POSIX_C_SOURCE\ >=\ 200112L || _XOPEN_SOURCE\ >=\ 600
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 a corresponding I/O operation (e.g.,
91 without blocking, or a sufficiently small
98 is identical, other than these three differences:
102 uses a timeout that is a
104 (with seconds and microseconds), while
108 (with seconds and nanoseconds).
114 argument to indicate how much time was left.
116 does not change this argument.
122 argument, and behaves as
127 Three independent sets of file descriptors are watched.
130 will be watched to see if characters become
131 available for reading (more precisely, to see if a read will not
132 block; in particular, a file descriptor is also ready on end-of-file),
135 will be watched to see if space is available for write (though a large
136 write may still block), and those in
138 will be watched for exceptions.
139 On exit, the sets are modified in place
140 to indicate which file descriptors actually changed status.
141 Each of the three file descriptor sets may be specified as NULL
142 if no file descriptors are to be watched for the corresponding class
145 Four macros are provided to manipulate the sets.
151 respectively add and remove a given file descriptor from a set.
153 tests to see if a file descriptor is part of the set;
159 is the highest-numbered file descriptor in any of the three sets, plus 1.
163 argument specifies the interval that
165 should block waiting for a file descriptor to become ready.
166 The call will block until either:
168 a file descriptor becomes ready;
170 the call is interrupted by a signal handler; or
176 interval will be rounded up to the system clock granularity,
177 and kernel scheduling delays mean that the blocking interval
178 may overrun by a small amount.
179 If both fields of the
181 structure are zero, then
184 (This is useful for polling.)
187 is NULL (no timeout),
189 can block indefinitely.
192 is a pointer to a signal mask (see
193 .BR sigprocmask (2));
194 if it is not NULL, then
196 first replaces the current signal mask by the one pointed to by
198 then does the "select" function, and then restores the original
201 Other than the difference in the precision of the
203 argument, the following
208 ready = pselect(nfds, &readfds, &writefds, &exceptfds,
214 executing the following calls:
219 pthread_sigmask(SIG_SETMASK, &sigmask, &origmask);
220 ready = select(nfds, &readfds, &writefds, &exceptfds, timeout);
221 pthread_sigmask(SIG_SETMASK, &origmask, NULL);
226 is needed is that if one wants to wait for either a signal
227 or for a file descriptor to become ready, then
228 an atomic test is needed to prevent race conditions.
229 (Suppose the signal handler sets a global flag and
231 Then a test of this global flag followed by a call of
233 could hang indefinitely if the signal arrived just after the test
234 but just before the call.
237 allows one to first block signals, handle the signals that have come in,
244 The time structures involved are defined in
251 long tv_sec; /* seconds */
252 long tv_usec; /* microseconds */
262 long tv_sec; /* seconds */
263 long tv_nsec; /* nanoseconds */
268 (However, see below on the POSIX.1-2001 versions.)
272 with all three sets empty,
276 as a fairly portable way to sleep with subsecond precision.
282 to reflect the amount of time not slept; most other implementations
284 (POSIX.1-2001 permits either behavior.)
285 This causes problems both when Linux code which reads
287 is ported to other operating systems, and when code is ported to Linux
288 that reuses a \fIstruct timeval\fP for multiple
290 in a loop without reinitializing it.
293 to be undefined after
296 .\" .PP - it is rumored that:
297 .\" On BSD, when a timeout occurs, the file descriptor bits are not changed.
298 .\" - it is certainly true that:
299 .\" Linux follows SUSv2 and sets the bit masks to zero upon a timeout.
305 return the number of file descriptors contained in the three returned
306 descriptor sets (that is, the total number of bits that are set in
310 which may be zero if the timeout expires before anything interesting happens.
311 On error, \-1 is returned, and
313 is set to indicate the error;
314 the file descriptor sets are unmodified,
321 An invalid file descriptor was given in one of the sets.
322 (Perhaps a file descriptor that was already closed,
323 or one on which an error has occurred.)
326 A signal was caught; see
331 is negative or the value contained within
336 unable to allocate memory for internal tables.
339 was added to Linux in kernel 2.6.16.
342 was emulated in glibc (but see BUGS).
345 conforms to POSIX.1-2001 and
348 first appeared in 4.2BSD).
349 Generally portable to/from
350 non-BSD systems supporting clones of the BSD socket layer (including
352 However, note that the System\ V variant typically
353 sets the timeout variable before exit, but the BSD variant does not.
356 is defined in POSIX.1g, and in
361 is a fixed size buffer.
368 that is negative or is equal to or larger than
371 in undefined behavior.
372 Moreover, POSIX requires
374 to be a valid file descriptor.
376 Concerning the types involved, the classical situation is that
379 structure are typed as
381 (as shown above), and the structure is defined in
383 The POSIX.1-2001 situation is
388 time_t tv_sec; /* seconds */
389 suseconds_t tv_usec; /* microseconds */
394 where the structure is defined in
403 Concerning prototypes, the classical situation is that one should
408 The POSIX.1-2001 situation is that one should include
417 gives the wrong prototype for
419 Under glibc 2.1 to 2.2.1, it gives
424 Since glibc 2.2.2, the requirements are as shown in the SYNOPSIS.
425 .SS Multithreaded applications
426 If a file descriptor being monitored by
428 is closed in another thread, the result is unspecified.
429 On some UNIX systems,
431 unblocks and returns, with an indication that the file descriptor is ready
432 (a subsequent I/O operation will likely fail with an error,
433 unless another the file descriptor reopened between the time
435 returned and the I/O operations was performed).
436 On Linux (and some other systems),
437 closing the file descriptor in another thread has no effect on
439 In summary, any application that relies on a particular behavior
440 in this scenario must be considered buggy.
442 .SS C library/kernel ABI differences
445 interface described in this page is implemented by glibc.
446 The underlying Linux system call is named
448 This system call has somewhat different behavior from the glibc
453 system call modifies its
456 However, the glibc wrapper function hides this behavior
457 by using a local variable for the timeout argument that
458 is passed to the system call.
461 function does not modify its
464 this is the behavior required by POSIX.1-2001.
466 The final argument of the
470 pointer, but is instead a structure of the form:
475 const sigset_t *ss; /* Pointer to signal set */
476 size_t ss_len; /* Size (in bytes) of object pointed
482 This allows the system call to obtain both
483 a pointer to the signal set and its size,
484 while allowing for the fact that most architectures
485 support a maximum of 6 arguments to a system call.
487 Glibc 2.0 provided a version of
493 Starting with version 2.1, glibc provided an emulation of
495 that was implemented using
499 This implementation remained vulnerable to the very race condition that
501 was designed to prevent.
502 Modern versions of glibc use the (race-free)
504 system call on kernels where it is provided.
508 reliable (and more portable) signal trapping can be achieved
509 using the self-pipe trick.
511 a signal handler writes a byte to a pipe whose other end
515 (To avoid possibly blocking when writing to a pipe that may be full
516 or reading from a pipe that may be empty,
517 nonblocking I/O is used when reading from and writing to the pipe.)
521 may report a socket file descriptor as "ready for reading", while
522 nevertheless a subsequent read blocks.
523 This could for example
524 happen when data has arrived but upon examination has wrong
525 checksum and is discarded.
526 There may be other circumstances
527 in which a file descriptor is spuriously reported as ready.
528 .\" Stevens discusses a case where accept can block after select
529 .\" returns successfully because of an intervening RST from the client.
530 Thus it may be safer to use
532 on sockets that should not block.
533 .\" Maybe the kernel should have returned EIO in such a situation?
539 if the call is interrupted by a signal handler (i.e., the
542 This is not permitted by POSIX.1-2001.
545 system call has the same behavior,
546 but the glibc wrapper hides this behavior by internally copying the
548 to a local variable and passing that variable to the system call.
553 #include <sys/time.h>
554 #include <sys/types.h>
564 /* Watch stdin (fd 0) to see when it has input. */
568 /* Wait up to five seconds. */
572 retval = select(1, &rfds, NULL, NULL, &tv);
573 /* Don't rely on the value of tv now! */
578 printf("Data is available now.\\n");
579 /* FD_ISSET(0, &rfds) will be true. */
581 printf("No data within five seconds.\\n");
592 .BR restart_syscall (2),
599 For a tutorial with discussion and examples, see