select.2: Show correspondence between select() and poll() readiness notifications

[thirdparty/man-pages.git] / man2 / select.2

.\" This manpage is copyright (C) 1992 Drew Eckhardt,
.\"                 copyright (C) 1995 Michael Shields.
.\"
.\" %%%LICENSE_START(VERBATIM)
.\" Permission is granted to make and distribute verbatim copies of this
.\" manual provided the copyright notice and this permission notice are
.\" preserved on all copies.
.\"
.\" Permission is granted to copy and distribute modified versions of this
.\" manual under the conditions for verbatim copying, provided that the
.\" entire resulting derived work is distributed under the terms of a
.\" permission notice identical to this one.
.\"
.\" Since the Linux kernel and libraries are constantly changing, this
.\" manual page may be incorrect or out-of-date.  The author(s) assume no
.\" responsibility for errors or omissions, or for damages resulting from
.\" the use of the information contained herein.  The author(s) may not
.\" have taken the same level of care in the production of this manual,
.\" which is licensed free of charge, as they might when working
.\" professionally.
.\"
.\" Formatted or processed versions of this manual, if unaccompanied by
.\" the source, must acknowledge the copyright and authors of this work.
.\" %%%LICENSE_END
.\"
.\" Modified 1993-07-24 by Rik Faith <faith@cs.unc.edu>
.\" Modified 1995-05-18 by Jim Van Zandt <jrv@vanzandt.mv.com>
.\" Sun Feb 11 14:07:00 MET 1996  Martin Schulze  <joey@linux.de>
.\"     * layout slightly modified
.\"
.\" Modified Mon Oct 21 23:05:29 EDT 1996 by Eric S. Raymond <esr@thyrsus.com>
.\" Modified Thu Feb 24 01:41:09 CET 2000 by aeb
.\" Modified Thu Feb  9 22:32:09 CET 2001 by bert hubert <ahu@ds9a.nl>, aeb
.\" Modified Mon Nov 11 14:35:00 PST 2002 by Ben Woodard <ben@zork.net>
.\" 2005-03-11, mtk, modified pselect() text (it is now a system
.\"     call in 2.6.16.
.\"
.TH SELECT 2 2017-03-13 "Linux" "Linux Programmer's Manual"
.SH NAME
select, pselect, FD_CLR, FD_ISSET, FD_SET, FD_ZERO \-
synchronous I/O multiplexing
.SH SYNOPSIS
.nf
/* According to POSIX.1-2001, POSIX.1-2008 */
.br
.B #include <sys/select.h>
.sp
/* According to earlier standards */
.br
.B #include <sys/time.h>
.br
.B #include <sys/types.h>
.br
.B #include <unistd.h>
.sp
.BI "int select(int " nfds ", fd_set *" readfds ", fd_set *" writefds ,
.BI "           fd_set *" exceptfds ", struct timeval *" timeout );
.sp
.BI "void FD_CLR(int " fd ", fd_set *" set );
.br
.BI "int  FD_ISSET(int " fd ", fd_set *" set );
.br
.BI "void FD_SET(int " fd ", fd_set *" set );
.br
.BI "void FD_ZERO(fd_set *" set );
.sp
.B #include <sys/select.h>
.sp
.BI "int pselect(int " nfds ", fd_set *" readfds ", fd_set *" writefds ,
.BI "            fd_set *" exceptfds ", const struct timespec *" timeout ,
.BI "            const sigset_t *" sigmask );
.fi
.sp
.in -4n
Feature Test Macro Requirements for glibc (see
.BR feature_test_macros (7)):
.in
.sp
.BR pselect ():
_POSIX_C_SOURCE\ >=\ 200112L
.SH DESCRIPTION
.BR select ()
and
.BR pselect ()
allow a program to monitor multiple file descriptors,
waiting until one or more of the file descriptors become "ready"
for some class of I/O operation (e.g., input possible).
A file descriptor is considered ready if it is possible to
perform a corresponding I/O operation (e.g.,
.BR read (2)
without blocking, or a sufficiently small
.BR write (2)).
.PP
.BR select ()
can monitor only file descriptors numbers that are less than
.BR FD_SETSIZE ;
.BR poll (2)
does not have this limitation.
See BUGS.
.PP
The operation of
.BR select ()
and
.BR pselect ()
is identical, other than these three differences:
.TP
(i)
.BR select ()
uses a timeout that is a
.I struct timeval
(with seconds and microseconds), while
.BR pselect ()
uses a
.I struct timespec
(with seconds and nanoseconds).
.TP
(ii)
.BR select ()
may update the
.I timeout
argument to indicate how much time was left.
.BR pselect ()
does not change this argument.
.TP
(iii)
.BR select ()
has no
.I sigmask
argument, and behaves as
.BR pselect ()
called with NULL
.IR sigmask .
.PP
Three independent sets of file descriptors are watched.
Those listed in
.I readfds
will be watched to see if characters become
available for reading (more precisely, to see if a read will not
block; in particular, a file descriptor is also ready on end-of-file),
those in
.I writefds
will be watched to see if space is available for write (though a large
write may still block), and those in
.I exceptfds
will be watched for exceptions.
On exit, the sets are modified in place
to indicate which file descriptors actually changed status.
Each of the three file descriptor sets may be specified as NULL
if no file descriptors are to be watched for the corresponding class
of events.
.PP
Four macros are provided to manipulate the sets.
.BR FD_ZERO ()
clears a set.
.BR FD_SET ()
and
.BR FD_CLR ()
respectively add and remove a given file descriptor from a set.
.BR FD_ISSET ()
tests to see if a file descriptor is part of the set;
this is useful after
.BR select ()
returns.
.PP
.I nfds
should be set to the highest-numbered file descriptor in any
of the three sets, plus 1.
The indicated file descriptors in each set are checked, up to this limit
(but see BUGS).
.PP
The
.I timeout
argument specifies the interval that
.BR select ()
should block waiting for a file descriptor to become ready.
The call will block until either:
.IP * 3
a file descriptor becomes ready;
.IP *
the call is interrupted by a signal handler; or
.IP *
the timeout expires.
.PP
Note that the
.I timeout
interval will be rounded up to the system clock granularity,
and kernel scheduling delays mean that the blocking interval
may overrun by a small amount.
If both fields of the
.I timeval
structure are zero, then
.BR select ()
returns immediately.
(This is useful for polling.)
If
.I timeout
is NULL (no timeout),
.BR select ()
can block indefinitely.
.PP
.I sigmask
is a pointer to a signal mask (see
.BR sigprocmask (2));
if it is not NULL, then
.BR pselect ()
first replaces the current signal mask by the one pointed to by
.IR sigmask ,
then does the "select" function, and then restores the original
signal mask.
.PP
Other than the difference in the precision of the
.I timeout
argument, the following
.BR pselect ()
call:
.nf

    ready = pselect(nfds, &readfds, &writefds, &exceptfds,
                    timeout, &sigmask);

.fi
is equivalent to
.I atomically
executing the following calls:
.nf

    sigset_t origmask;

    pthread_sigmask(SIG_SETMASK, &sigmask, &origmask);
    ready = select(nfds, &readfds, &writefds, &exceptfds, timeout);
    pthread_sigmask(SIG_SETMASK, &origmask, NULL);
.fi
.PP
The reason that
.BR pselect ()
is needed is that if one wants to wait for either a signal
or for a file descriptor to become ready, then
an atomic test is needed to prevent race conditions.
(Suppose the signal handler sets a global flag and
returns.
Then a test of this global flag followed by a call of
.BR select ()
could hang indefinitely if the signal arrived just after the test
but just before the call.
By contrast,
.BR pselect ()
allows one to first block signals, handle the signals that have come in,
then call
.BR pselect ()
with the desired
.IR sigmask ,
avoiding the race.)
.SS The timeout
The time structures involved are defined in
.I <sys/time.h>
and look like

.in +4n
.nf
struct timeval {
    long    tv_sec;         /* seconds */
    long    tv_usec;        /* microseconds */
};
.fi
.in

and

.in +4n
.nf
struct timespec {
    long    tv_sec;         /* seconds */
    long    tv_nsec;        /* nanoseconds */
};
.fi
.in

(However, see below on the POSIX.1 versions.)
.PP
Some code calls
.BR select ()
with all three sets empty,
.I nfds
zero, and a non-NULL
.I timeout
as a fairly portable way to sleep with subsecond precision.
.PP
On Linux,
.BR select ()
modifies
.I timeout
to reflect the amount of time not slept; most other implementations
do not do this.
(POSIX.1 permits either behavior.)
This causes problems both when Linux code which reads
.I timeout
is ported to other operating systems, and when code is ported to Linux
that reuses a \fIstruct timeval\fP for multiple
.BR select ()s
in a loop without reinitializing it.
Consider
.I timeout
to be undefined after
.BR select ()
returns.
.\" .PP - it is rumored that:
.\" On BSD, when a timeout occurs, the file descriptor bits are not changed.
.\" - it is certainly true that:
.\" Linux follows SUSv2 and sets the bit masks to zero upon a timeout.
.SH RETURN VALUE
On success,
.BR select ()
and
.BR pselect ()
return the number of file descriptors contained in the three returned
descriptor sets (that is, the total number of bits that are set in
.IR readfds ,
.IR writefds ,
.IR exceptfds )
which may be zero if the timeout expires before anything interesting happens.
On error, \-1 is returned, and
.I errno
is set to indicate the error;
the file descriptor sets are unmodified,
and
.I timeout
becomes undefined.
.SH ERRORS
.TP
.B EBADF
An invalid file descriptor was given in one of the sets.
(Perhaps a file descriptor that was already closed,
or one on which an error has occurred.)
However, see BUGS.
.TP
.B EINTR
A signal was caught; see
.BR signal (7).
.TP
.B EINVAL
.I nfds
is negative or exceeds the
.BR RLIMIT_NOFILE
resource limit (see
.BR getrlimit (2)).
.TP
.B EINVAL
The value contained within
.I timeout
is invalid.
.TP
.B ENOMEM
Unable to allocate memory for internal tables.
.SH VERSIONS
.BR pselect ()
was added to Linux in kernel 2.6.16.
Prior to this,
.BR pselect ()
was emulated in glibc (but see BUGS).
.SH CONFORMING TO
.BR select ()
conforms to POSIX.1-2001, POSIX.1-2008, and
4.4BSD
.RB ( select ()
first appeared in 4.2BSD).
Generally portable to/from
non-BSD systems supporting clones of the BSD socket layer (including
System\ V variants).
However, note that the System\ V variant typically
sets the timeout variable before exit, but the BSD variant does not.
.PP
.BR pselect ()
is defined in POSIX.1g, and in
POSIX.1-2001 and POSIX.1-2008.
.SH NOTES
An
.I fd_set
is a fixed size buffer.
Executing
.BR FD_CLR ()
or
.BR FD_SET ()
with a value of
.I fd
that is negative or is equal to or larger than
.B FD_SETSIZE
will result
in undefined behavior.
Moreover, POSIX requires
.I fd
to be a valid file descriptor.

On some other UNIX systems,
.\" Darwin, according to a report by Jeremy Sequoia, relayed by Josh Triplett
.BR select ()
can fail with the error
.B EAGAIN
if the system fails to allocate kernel-internal resources, rather than
.B ENOMEM
as Linux does.
POSIX specifies this error for
.BR poll (2),
but not for
.BR select ().
Portable programs may wish to check for
.B EAGAIN
and loop, just as with
.BR EINTR .

On systems that lack
.BR pselect (),
reliable (and more portable) signal trapping can be achieved
using the self-pipe trick.
In this technique,
a signal handler writes a byte to a pipe whose other end
is monitored by
.BR select ()
in the main program.
(To avoid possibly blocking when writing to a pipe that may be full
or reading from a pipe that may be empty,
nonblocking I/O is used when reading from and writing to the pipe.)

Concerning the types involved, the classical situation is that
the two fields of a
.I timeval
structure are typed as
.I long
(as shown above), and the structure is defined in
.IR <sys/time.h> .
The POSIX.1 situation is

.in +4n
.nf
struct timeval {
    time_t         tv_sec;     /* seconds */
    suseconds_t    tv_usec;    /* microseconds */
};
.fi
.in

where the structure is defined in
.I <sys/select.h>
and the data types
.I time_t
and
.I suseconds_t
are defined in
.IR <sys/types.h> .
.LP
Concerning prototypes, the classical situation is that one should
include
.I <time.h>
for
.BR select ().
The POSIX.1 situation is that one should include
.I <sys/select.h>
for
.BR select ()
and
.BR pselect ().

Under glibc 2.0,
.I <sys/select.h>
gives the wrong prototype for
.BR pselect ().
Under glibc 2.1 to 2.2.1, it gives
.BR pselect ()
when
.B _GNU_SOURCE
is defined.
Since glibc 2.2.2, the requirements are as shown in the SYNOPSIS.
.\"
.SS Correspondence between select() and poll() notifications
Within the Linux kernel source,
.\" fs/select.c
we find the following definitions which show the correspondence
between the readable, writable, and exceptional condition notifications of
.BR select ()
and the event notifications provided by
.BR poll (2)
(and
.BR epoll (7)):

.nf
.in +4n
#define POLLIN_SET (POLLRDNORM | POLLRDBAND | POLLIN | POLLHUP |
                    POLLERR)
                   /* Ready for reading */
#define POLLOUT_SET (POLLWRBAND | POLLWRNORM | POLLOUT | POLLERR)
                   /* Ready for writing */
#define POLLEX_SET (POLLPRI)
                   /* Exceptional condition */
.in
.fi
.\"
.SS Multithreaded applications
If a file descriptor being monitored by
.BR select ()
is closed in another thread, the result is unspecified.
On some UNIX systems,
.BR select ()
unblocks and returns, with an indication that the file descriptor is ready
(a subsequent I/O operation will likely fail with an error,
unless another the file descriptor reopened between the time
.BR select ()
returned and the I/O operations was performed).
On Linux (and some other systems),
closing the file descriptor in another thread has no effect on
.BR select ().
In summary, any application that relies on a particular behavior
in this scenario must be considered buggy.
.\"
.SS C library/kernel differences
The Linux kernel allows file descriptor sets of arbitrary size,
determining the length of the sets to be checked from the value of
.IR nfds .
However, in the glibc implementation, the
.IR fd_set
type is fixed in size.
See also BUGS.

The
.BR pselect ()
interface described in this page is implemented by glibc.
The underlying Linux system call is named
.BR pselect6 ().
This system call has somewhat different behavior from the glibc
wrapper function.

The Linux
.BR pselect6 ()
system call modifies its
.I timeout
argument.
However, the glibc wrapper function hides this behavior
by using a local variable for the timeout argument that
is passed to the system call.
Thus, the glibc
.BR pselect ()
function does not modify its
.I timeout
argument;
this is the behavior required by POSIX.1-2001.

The final argument of the
.BR pselect6 ()
system call is not a
.I "sigset_t\ *"
pointer, but is instead a structure of the form:
.in +4
.nf

struct {
    const kernel_sigset_t *ss;   /* Pointer to signal set */
    size_t ss_len;               /* Size (in bytes) of object
                                    pointed to by 'ss' */
};

.fi
.in
This allows the system call to obtain both
a pointer to the signal set and its size,
while allowing for the fact that most architectures
support a maximum of 6 arguments to a system call.
See
.BR sigprocmask (2)
for a discussion of the difference between the kernel and libc
notion of the signal set.
.SH BUGS
POSIX allows an implementation to define an upper limit,
advertised via the constant
.BR FD_SETSIZE ,
on the range of file descriptors that can be specified
in a file descriptor set.
The Linux kernel imposes no fixed limit, but the glibc implementation makes
.IR fd_set
a fixed-size type, with
.BR FD_SETSIZE
defined as 1024, and the
.BR FD_* ()
macros operating according to that limit.
To monitor file descriptors greater than 1023, use
.BR poll (2)
instead.

According to POSIX,
.BR select ()
should check all specified file descriptors in the three file descriptor sets,
up to the limit
.IR nfds\-1 .
However, the current implementation ignores any file descriptor in
these sets that is greater than the maximum file descriptor number
that the process currently has open.
According to POSIX, any such file descriptor that is specified in one
of the sets should result in the error
.BR EBADF .

Glibc 2.0 provided a version of
.BR pselect ()
that did not take a
.I sigmask
argument.

Starting with version 2.1, glibc provided an emulation of
.BR pselect ()
that was implemented using
.BR sigprocmask (2)
and
.BR select ().
This implementation remained vulnerable to the very race condition that
.BR pselect ()
was designed to prevent.
Modern versions of glibc use the (race-free)
.BR pselect ()
system call on kernels where it is provided.

Under Linux,
.BR select ()
may report a socket file descriptor as "ready for reading", while
nevertheless a subsequent read blocks.
This could for example
happen when data has arrived but upon examination has wrong
checksum and is discarded.
There may be other circumstances
in which a file descriptor is spuriously reported as ready.
.\" Stevens discusses a case where accept can block after select
.\" returns successfully because of an intervening RST from the client.
Thus it may be safer to use
.B O_NONBLOCK
on sockets that should not block.
.\" Maybe the kernel should have returned EIO in such a situation?

On Linux,
.BR select ()
also modifies
.I timeout
if the call is interrupted by a signal handler (i.e., the
.B EINTR
error return).
This is not permitted by POSIX.1.
The Linux
.BR pselect ()
system call has the same behavior,
but the glibc wrapper hides this behavior by internally copying the
.I timeout
to a local variable and passing that variable to the system call.
.SH EXAMPLE
.nf
#include <stdio.h>
#include <stdlib.h>
#include <sys/time.h>
#include <sys/types.h>
#include <unistd.h>

int
main(void)
{
    fd_set rfds;
    struct timeval tv;
    int retval;

    /* Watch stdin (fd 0) to see when it has input. */

    FD_ZERO(&rfds);
    FD_SET(0, &rfds);

    /* Wait up to five seconds. */

    tv.tv_sec = 5;
    tv.tv_usec = 0;

    retval = select(1, &rfds, NULL, NULL, &tv);
    /* Don't rely on the value of tv now! */

    if (retval == \-1)
        perror("select()");
    else if (retval)
        printf("Data is available now.\\n");
        /* FD_ISSET(0, &rfds) will be true. */
    else
        printf("No data within five seconds.\\n");

    exit(EXIT_SUCCESS);
}
.fi
.SH SEE ALSO
.BR accept (2),
.BR connect (2),
.BR poll (2),
.BR read (2),
.BR recv (2),
.BR restart_syscall (2),
.BR send (2),
.BR sigprocmask (2),
.BR write (2),
.BR epoll (7),
.BR time (7)

For a tutorial with discussion and examples, see
.BR select_tut (2).