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15 .\" manual page may be incorrect or out-of-date.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 2019-03-06 "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, POSIX.1-2008 */
45 .B #include <sys/select.h>
47 /* According to earlier standards */
48 .B #include <sys/time.h>
49 .B #include <sys/types.h>
50 .B #include <unistd.h>
52 .BI "int select(int " nfds ", fd_set *" readfds ", fd_set *" writefds ,
53 .BI " fd_set *" exceptfds ", struct timeval *" timeout );
55 .BI "void FD_CLR(int " fd ", fd_set *" set );
56 .BI "int FD_ISSET(int " fd ", fd_set *" set );
57 .BI "void FD_SET(int " fd ", fd_set *" set );
58 .BI "void FD_ZERO(fd_set *" set );
60 .B #include <sys/select.h>
62 .BI "int pselect(int " nfds ", fd_set *" readfds ", fd_set *" writefds ,
63 .BI " fd_set *" exceptfds ", const struct timespec *" timeout ,
64 .BI " const sigset_t *" sigmask );
68 Feature Test Macro Requirements for glibc (see
69 .BR feature_test_macros (7)):
73 _POSIX_C_SOURCE\ >=\ 200112L
78 allow a program to monitor multiple file descriptors,
79 waiting until one or more of the file descriptors become "ready"
80 for some class of I/O operation (e.g., input possible).
81 A file descriptor is considered ready if it is possible to
82 perform a corresponding I/O operation (e.g.,
84 or a sufficiently small
89 can monitor only file descriptors numbers that are less than
92 does not have this limitation.
99 is identical, other than these three differences:
103 uses a timeout that is a
105 (with seconds and microseconds), while
109 (with seconds and nanoseconds).
115 argument to indicate how much time was left.
117 does not change this argument.
123 argument, and behaves as
128 Three independent sets of file descriptors are watched.
129 The file descriptors listed in
131 will be watched to see if characters become
132 available for reading (more precisely, to see if a read will not
133 block; in particular, a file descriptor is also ready on end-of-file).
134 The file descriptors in
136 will be watched to see if space is available for write (though a large
137 write may still block).
138 The file descriptors in
140 will be watched for exceptional conditions.
141 (For examples of some exceptional conditions, see the discussion of
146 On exit, each of the file descriptor sets is modified in place
147 to indicate which file descriptors actually changed status.
150 within a loop, the sets must be reinitialized before each call.)
152 Each of the three file descriptor sets may be specified as NULL
153 if no file descriptors are to be watched for the corresponding class
156 Four macros are provided to manipulate the sets.
162 add and remove a given file descriptor from a set.
164 tests to see if a file descriptor is part of the set;
170 should be set to the highest-numbered file descriptor in any
171 of the three sets, plus 1.
172 The indicated file descriptors in each set are checked, up to this limit
177 argument specifies the interval that
179 should block waiting for a file descriptor to become ready.
180 The call will block until either:
182 a file descriptor becomes ready;
184 the call is interrupted by a signal handler; or
190 interval will be rounded up to the system clock granularity,
191 and kernel scheduling delays mean that the blocking interval
192 may overrun by a small amount.
193 If both fields of the
195 structure are zero, then
198 (This is useful for polling.)
201 is NULL (no timeout),
203 can block indefinitely.
206 is a pointer to a signal mask (see
207 .BR sigprocmask (2));
208 if it is not NULL, then
210 first replaces the current signal mask by the one pointed to by
212 then does the "select" function, and then restores the original
215 Other than the difference in the precision of the
217 argument, the following
223 ready = pselect(nfds, &readfds, &writefds, &exceptfds,
230 executing the following calls:
236 pthread_sigmask(SIG_SETMASK, &sigmask, &origmask);
237 ready = select(nfds, &readfds, &writefds, &exceptfds, timeout);
238 pthread_sigmask(SIG_SETMASK, &origmask, NULL);
245 is needed is that if one wants to wait for either a signal
246 or for a file descriptor to become ready, then
247 an atomic test is needed to prevent race conditions.
248 (Suppose the signal handler sets a global flag and
250 Then a test of this global flag followed by a call of
252 could hang indefinitely if the signal arrived just after the test
253 but just before the call.
256 allows one to first block signals, handle the signals that have come in,
263 The time structures involved are defined in
270 long tv_sec; /* seconds */
271 long tv_usec; /* microseconds */
281 long tv_sec; /* seconds */
282 long tv_nsec; /* nanoseconds */
287 (However, see below on the POSIX.1 versions.)
291 with all three sets empty,
295 as a fairly portable way to sleep with subsecond precision.
301 to reflect the amount of time not slept; most other implementations
303 (POSIX.1 permits either behavior.)
304 This causes problems both when Linux code which reads
306 is ported to other operating systems, and when code is ported to Linux
307 that reuses a \fIstruct timeval\fP for multiple
309 in a loop without reinitializing it.
312 to be undefined after
315 .\" .PP - it is rumored that:
316 .\" On BSD, when a timeout occurs, the file descriptor bits are not changed.
317 .\" - it is certainly true that:
318 .\" Linux follows SUSv2 and sets the bit masks to zero upon a timeout.
324 return the number of file descriptors contained in the three returned
325 descriptor sets (that is, the total number of bits that are set in
329 which may be zero if the timeout expires before anything interesting happens.
330 On error, \-1 is returned, and
332 is set to indicate the error;
333 the file descriptor sets are unmodified,
340 An invalid file descriptor was given in one of the sets.
341 (Perhaps a file descriptor that was already closed,
342 or one on which an error has occurred.)
346 A signal was caught; see
351 is negative or exceeds the
357 The value contained within
362 Unable to allocate memory for internal tables.
365 was added to Linux in kernel 2.6.16.
368 was emulated in glibc (but see BUGS).
371 conforms to POSIX.1-2001, POSIX.1-2008, and
374 first appeared in 4.2BSD).
375 Generally portable to/from
376 non-BSD systems supporting clones of the BSD socket layer (including
378 However, note that the System\ V variant typically
379 sets the timeout variable before exit, but the BSD variant does not.
382 is defined in POSIX.1g, and in
383 POSIX.1-2001 and POSIX.1-2008.
387 is a fixed size buffer.
394 that is negative or is equal to or larger than
397 in undefined behavior.
398 Moreover, POSIX requires
400 to be a valid file descriptor.
406 is not affected by the
410 On some other UNIX systems,
411 .\" Darwin, according to a report by Jeremy Sequoia, relayed by Josh Triplett
413 can fail with the error
415 if the system fails to allocate kernel-internal resources, rather than
418 POSIX specifies this error for
422 Portable programs may wish to check for
424 and loop, just as with
429 reliable (and more portable) signal trapping can be achieved
430 using the self-pipe trick.
432 a signal handler writes a byte to a pipe whose other end
436 (To avoid possibly blocking when writing to a pipe that may be full
437 or reading from a pipe that may be empty,
438 nonblocking I/O is used when reading from and writing to the pipe.)
440 Concerning the types involved, the classical situation is that
443 structure are typed as
445 (as shown above), and the structure is defined in
447 The POSIX.1 situation is
452 time_t tv_sec; /* seconds */
453 suseconds_t tv_usec; /* microseconds */
458 where the structure is defined in
467 Concerning prototypes, the classical situation is that one should
472 The POSIX.1 situation is that one should include
481 gives the wrong prototype for
483 Under glibc 2.1 to 2.2.1, it gives
488 Since glibc 2.2.2, the requirements are as shown in the SYNOPSIS.
490 .SS Correspondence between select() and poll() notifications
491 Within the Linux kernel source,
493 we find the following definitions which show the correspondence
494 between the readable, writable, and exceptional condition notifications of
496 and the event notifications provided by
503 #define POLLIN_SET (POLLRDNORM | POLLRDBAND | POLLIN | POLLHUP |
505 /* Ready for reading */
506 #define POLLOUT_SET (POLLWRBAND | POLLWRNORM | POLLOUT | POLLERR)
507 /* Ready for writing */
508 #define POLLEX_SET (POLLPRI)
509 /* Exceptional condition */
513 .SS Multithreaded applications
514 If a file descriptor being monitored by
516 is closed in another thread, the result is unspecified.
517 On some UNIX systems,
519 unblocks and returns, with an indication that the file descriptor is ready
520 (a subsequent I/O operation will likely fail with an error,
521 unless another process reopens file descriptor between the time
523 returned and the I/O operation is performed).
524 On Linux (and some other systems),
525 closing the file descriptor in another thread has no effect on
527 In summary, any application that relies on a particular behavior
528 in this scenario must be considered buggy.
530 .SS C library/kernel differences
531 The Linux kernel allows file descriptor sets of arbitrary size,
532 determining the length of the sets to be checked from the value of
534 However, in the glibc implementation, the
536 type is fixed in size.
541 interface described in this page is implemented by glibc.
542 The underlying Linux system call is named
544 This system call has somewhat different behavior from the glibc
549 system call modifies its
552 However, the glibc wrapper function hides this behavior
553 by using a local variable for the timeout argument that
554 is passed to the system call.
557 function does not modify its
560 this is the behavior required by POSIX.1-2001.
562 The final argument of the
566 pointer, but is instead a structure of the form:
571 const kernel_sigset_t *ss; /* Pointer to signal set */
572 size_t ss_len; /* Size (in bytes) of object
573 pointed to by 'ss' */
578 This allows the system call to obtain both
579 a pointer to the signal set and its size,
580 while allowing for the fact that most architectures
581 support a maximum of 6 arguments to a system call.
584 for a discussion of the difference between the kernel and libc
585 notion of the signal set.
587 POSIX allows an implementation to define an upper limit,
588 advertised via the constant
590 on the range of file descriptors that can be specified
591 in a file descriptor set.
592 The Linux kernel imposes no fixed limit, but the glibc implementation makes
594 a fixed-size type, with
596 defined as 1024, and the
598 macros operating according to that limit.
599 To monitor file descriptors greater than 1023, use
603 The implementation of the
605 arguments as value-result arguments means that they must be
606 reinitialized on each call to
608 This design error is avoided by
610 which uses separate structure fields for the input and output of the call.
614 should check all specified file descriptors in the three file descriptor sets,
617 However, the current implementation ignores any file descriptor in
618 these sets that is greater than the maximum file descriptor number
619 that the process currently has open.
620 According to POSIX, any such file descriptor that is specified in one
621 of the sets should result in the error
624 Glibc 2.0 provided a version of
630 Starting with version 2.1, glibc provided an emulation of
632 that was implemented using
636 This implementation remained vulnerable to the very race condition that
638 was designed to prevent.
639 Modern versions of glibc use the (race-free)
641 system call on kernels where it is provided.
645 may report a socket file descriptor as "ready for reading", while
646 nevertheless a subsequent read blocks.
647 This could for example
648 happen when data has arrived but upon examination has wrong
649 checksum and is discarded.
650 There may be other circumstances
651 in which a file descriptor is spuriously reported as ready.
652 .\" Stevens discusses a case where accept can block after select
653 .\" returns successfully because of an intervening RST from the client.
654 Thus it may be safer to use
656 on sockets that should not block.
657 .\" Maybe the kernel should have returned EIO in such a situation?
663 if the call is interrupted by a signal handler (i.e., the
666 This is not permitted by POSIX.1.
669 system call has the same behavior,
670 but the glibc wrapper hides this behavior by internally copying the
672 to a local variable and passing that variable to the system call.
677 #include <sys/time.h>
678 #include <sys/types.h>
688 /* Watch stdin (fd 0) to see when it has input. */
693 /* Wait up to five seconds. */
698 retval = select(1, &rfds, NULL, NULL, &tv);
699 /* Don't rely on the value of tv now! */
704 printf("Data is available now.\en");
705 /* FD_ISSET(0, &rfds) will be true. */
707 printf("No data within five seconds.\en");
718 .BR restart_syscall (2),
725 For a tutorial with discussion and examples, see