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15 .\" manual page may be incorrect or out-of-date.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 2017-05-03 "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 );
57 .BI "int FD_ISSET(int " fd ", fd_set *" set );
59 .BI "void FD_SET(int " fd ", fd_set *" set );
61 .BI "void FD_ZERO(fd_set *" set );
63 .B #include <sys/select.h>
65 .BI "int pselect(int " nfds ", fd_set *" readfds ", fd_set *" writefds ,
66 .BI " fd_set *" exceptfds ", const struct timespec *" timeout ,
67 .BI " const sigset_t *" sigmask );
71 Feature Test Macro Requirements for glibc (see
72 .BR feature_test_macros (7)):
76 _POSIX_C_SOURCE\ >=\ 200112L
81 allow a program to monitor multiple file descriptors,
82 waiting until one or more of the file descriptors become "ready"
83 for some class of I/O operation (e.g., input possible).
84 A file descriptor is considered ready if it is possible to
85 perform a corresponding I/O operation (e.g.,
87 without blocking, or a sufficiently small
91 can monitor only file descriptors numbers that are less than
94 does not have this limitation.
101 is identical, other than these three differences:
105 uses a timeout that is a
107 (with seconds and microseconds), while
111 (with seconds and nanoseconds).
117 argument to indicate how much time was left.
119 does not change this argument.
125 argument, and behaves as
130 Three independent sets of file descriptors are watched.
131 The file descriptors listed in
133 will be watched to see if characters become
134 available for reading (more precisely, to see if a read will not
135 block; in particular, a file descriptor is also ready on end-of-file).
136 The file descriptors in
138 will be watched to see if space is available for write (though a large
139 write may still block).
140 The file descriptors in
142 will be watched for exceptional conditions.
143 (For examples of some exceptional conditions, see the discussion of
148 On exit, each of the file descriptor sets is modified in place
149 to indicate which file descriptors actually changed status.
152 within a loop, the sets must be reinitialized before each call.)
154 Each of the three file descriptor sets may be specified as NULL
155 if no file descriptors are to be watched for the corresponding class
158 Four macros are provided to manipulate the sets.
164 respectively add and remove a given file descriptor from a set.
166 tests to see if a file descriptor is part of the set;
172 should be set to the highest-numbered file descriptor in any
173 of the three sets, plus 1.
174 The indicated file descriptors in each set are checked, up to this limit
179 argument specifies the interval that
181 should block waiting for a file descriptor to become ready.
182 The call will block until either:
184 a file descriptor becomes ready;
186 the call is interrupted by a signal handler; or
192 interval will be rounded up to the system clock granularity,
193 and kernel scheduling delays mean that the blocking interval
194 may overrun by a small amount.
195 If both fields of the
197 structure are zero, then
200 (This is useful for polling.)
203 is NULL (no timeout),
205 can block indefinitely.
208 is a pointer to a signal mask (see
209 .BR sigprocmask (2));
210 if it is not NULL, then
212 first replaces the current signal mask by the one pointed to by
214 then does the "select" function, and then restores the original
217 Other than the difference in the precision of the
219 argument, the following
224 ready = pselect(nfds, &readfds, &writefds, &exceptfds,
230 executing the following calls:
235 pthread_sigmask(SIG_SETMASK, &sigmask, &origmask);
236 ready = select(nfds, &readfds, &writefds, &exceptfds, timeout);
237 pthread_sigmask(SIG_SETMASK, &origmask, NULL);
242 is needed is that if one wants to wait for either a signal
243 or for a file descriptor to become ready, then
244 an atomic test is needed to prevent race conditions.
245 (Suppose the signal handler sets a global flag and
247 Then a test of this global flag followed by a call of
249 could hang indefinitely if the signal arrived just after the test
250 but just before the call.
253 allows one to first block signals, handle the signals that have come in,
260 The time structures involved are defined in
267 long tv_sec; /* seconds */
268 long tv_usec; /* microseconds */
278 long tv_sec; /* seconds */
279 long tv_nsec; /* nanoseconds */
284 (However, see below on the POSIX.1 versions.)
288 with all three sets empty,
292 as a fairly portable way to sleep with subsecond precision.
298 to reflect the amount of time not slept; most other implementations
300 (POSIX.1 permits either behavior.)
301 This causes problems both when Linux code which reads
303 is ported to other operating systems, and when code is ported to Linux
304 that reuses a \fIstruct timeval\fP for multiple
306 in a loop without reinitializing it.
309 to be undefined after
312 .\" .PP - it is rumored that:
313 .\" On BSD, when a timeout occurs, the file descriptor bits are not changed.
314 .\" - it is certainly true that:
315 .\" Linux follows SUSv2 and sets the bit masks to zero upon a timeout.
321 return the number of file descriptors contained in the three returned
322 descriptor sets (that is, the total number of bits that are set in
326 which may be zero if the timeout expires before anything interesting happens.
327 On error, \-1 is returned, and
329 is set to indicate the error;
330 the file descriptor sets are unmodified,
337 An invalid file descriptor was given in one of the sets.
338 (Perhaps a file descriptor that was already closed,
339 or one on which an error has occurred.)
343 A signal was caught; see
348 is negative or exceeds the
354 The value contained within
359 Unable to allocate memory for internal tables.
362 was added to Linux in kernel 2.6.16.
365 was emulated in glibc (but see BUGS).
368 conforms to POSIX.1-2001, POSIX.1-2008, and
371 first appeared in 4.2BSD).
372 Generally portable to/from
373 non-BSD systems supporting clones of the BSD socket layer (including
375 However, note that the System\ V variant typically
376 sets the timeout variable before exit, but the BSD variant does not.
379 is defined in POSIX.1g, and in
380 POSIX.1-2001 and POSIX.1-2008.
384 is a fixed size buffer.
391 that is negative or is equal to or larger than
394 in undefined behavior.
395 Moreover, POSIX requires
397 to be a valid file descriptor.
399 On some other UNIX systems,
400 .\" Darwin, according to a report by Jeremy Sequoia, relayed by Josh Triplett
402 can fail with the error
404 if the system fails to allocate kernel-internal resources, rather than
407 POSIX specifies this error for
411 Portable programs may wish to check for
413 and loop, just as with
418 reliable (and more portable) signal trapping can be achieved
419 using the self-pipe trick.
421 a signal handler writes a byte to a pipe whose other end
425 (To avoid possibly blocking when writing to a pipe that may be full
426 or reading from a pipe that may be empty,
427 nonblocking I/O is used when reading from and writing to the pipe.)
429 Concerning the types involved, the classical situation is that
432 structure are typed as
434 (as shown above), and the structure is defined in
436 The POSIX.1 situation is
441 time_t tv_sec; /* seconds */
442 suseconds_t tv_usec; /* microseconds */
447 where the structure is defined in
456 Concerning prototypes, the classical situation is that one should
461 The POSIX.1 situation is that one should include
470 gives the wrong prototype for
472 Under glibc 2.1 to 2.2.1, it gives
477 Since glibc 2.2.2, the requirements are as shown in the SYNOPSIS.
479 .SS Correspondence between select() and poll() notifications
480 Within the Linux kernel source,
482 we find the following definitions which show the correspondence
483 between the readable, writable, and exceptional condition notifications of
485 and the event notifications provided by
492 #define POLLIN_SET (POLLRDNORM | POLLRDBAND | POLLIN | POLLHUP |
494 /* Ready for reading */
495 #define POLLOUT_SET (POLLWRBAND | POLLWRNORM | POLLOUT | POLLERR)
496 /* Ready for writing */
497 #define POLLEX_SET (POLLPRI)
498 /* Exceptional condition */
502 .SS Multithreaded applications
503 If a file descriptor being monitored by
505 is closed in another thread, the result is unspecified.
506 On some UNIX systems,
508 unblocks and returns, with an indication that the file descriptor is ready
509 (a subsequent I/O operation will likely fail with an error,
510 unless another the file descriptor reopened between the time
512 returned and the I/O operations was performed).
513 On Linux (and some other systems),
514 closing the file descriptor in another thread has no effect on
516 In summary, any application that relies on a particular behavior
517 in this scenario must be considered buggy.
519 .SS C library/kernel differences
520 The Linux kernel allows file descriptor sets of arbitrary size,
521 determining the length of the sets to be checked from the value of
523 However, in the glibc implementation, the
525 type is fixed in size.
530 interface described in this page is implemented by glibc.
531 The underlying Linux system call is named
533 This system call has somewhat different behavior from the glibc
538 system call modifies its
541 However, the glibc wrapper function hides this behavior
542 by using a local variable for the timeout argument that
543 is passed to the system call.
546 function does not modify its
549 this is the behavior required by POSIX.1-2001.
551 The final argument of the
555 pointer, but is instead a structure of the form:
560 const kernel_sigset_t *ss; /* Pointer to signal set */
561 size_t ss_len; /* Size (in bytes) of object
562 pointed to by 'ss' */
567 This allows the system call to obtain both
568 a pointer to the signal set and its size,
569 while allowing for the fact that most architectures
570 support a maximum of 6 arguments to a system call.
573 for a discussion of the difference between the kernel and libc
574 notion of the signal set.
576 POSIX allows an implementation to define an upper limit,
577 advertised via the constant
579 on the range of file descriptors that can be specified
580 in a file descriptor set.
581 The Linux kernel imposes no fixed limit, but the glibc implementation makes
583 a fixed-size type, with
585 defined as 1024, and the
587 macros operating according to that limit.
588 To monitor file descriptors greater than 1023, use
594 should check all specified file descriptors in the three file descriptor sets,
597 However, the current implementation ignores any file descriptor in
598 these sets that is greater than the maximum file descriptor number
599 that the process currently has open.
600 According to POSIX, any such file descriptor that is specified in one
601 of the sets should result in the error
604 Glibc 2.0 provided a version of
610 Starting with version 2.1, glibc provided an emulation of
612 that was implemented using
616 This implementation remained vulnerable to the very race condition that
618 was designed to prevent.
619 Modern versions of glibc use the (race-free)
621 system call on kernels where it is provided.
625 may report a socket file descriptor as "ready for reading", while
626 nevertheless a subsequent read blocks.
627 This could for example
628 happen when data has arrived but upon examination has wrong
629 checksum and is discarded.
630 There may be other circumstances
631 in which a file descriptor is spuriously reported as ready.
632 .\" Stevens discusses a case where accept can block after select
633 .\" returns successfully because of an intervening RST from the client.
634 Thus it may be safer to use
636 on sockets that should not block.
637 .\" Maybe the kernel should have returned EIO in such a situation?
643 if the call is interrupted by a signal handler (i.e., the
646 This is not permitted by POSIX.1.
649 system call has the same behavior,
650 but the glibc wrapper hides this behavior by internally copying the
652 to a local variable and passing that variable to the system call.
657 #include <sys/time.h>
658 #include <sys/types.h>
668 /* Watch stdin (fd 0) to see when it has input. */
673 /* Wait up to five seconds. */
678 retval = select(1, &rfds, NULL, NULL, &tv);
679 /* Don't rely on the value of tv now! */
684 printf("Data is available now.\\n");
685 /* FD_ISSET(0, &rfds) will be true. */
687 printf("No data within five seconds.\\n");
698 .BR restart_syscall (2),
705 For a tutorial with discussion and examples, see