2 .\" Don't change the first line, it tells man that we need tbl.
3 .\" This man page is Copyright (C) 1999 Andi Kleen <ak@muc.de>.
4 .\" and copyright (c) 1999 Matthew Wilcox.
5 .\" Permission is granted to distribute possibly modified copies
6 .\" of this page provided the header is included verbatim,
7 .\" and in case of nontrivial modification author and date
8 .\" of the modification is added to the header.
10 .\" 2002-10-30, Michael Kerrisk, <mtk.manpages@gmail.com>
11 .\" Added description of SO_ACCEPTCONN
12 .\" 2004-05-20, aeb, added SO_RCVTIMEO/SO_SNDTIMEO text.
13 .\" Modified, 27 May 2004, Michael Kerrisk <mtk.manpages@gmail.com>
14 .\" Added notes on capability requirements
15 .\" A few small grammar fixes
16 .\" 2010-06-13 Jan Engelhardt <jengelh@medozas.de>
17 .\" Documented SO_DOMAIN and SO_PROTOCOL.
19 .\" The following are not yet documented:
22 .\" SO_TIMESTAMPING (2.6.30)
23 .\" SO_RXQ_OVFL (2.6.33)
25 .TH SOCKET 7 2012-07-07 Linux "Linux Programmer's Manual"
27 socket \- Linux socket interface
29 .B #include <sys/socket.h>
31 .IB sockfd " = socket(int " socket_family ", int " socket_type ", int " protocol );
33 This manual page describes the Linux networking socket layer user
35 The BSD compatible sockets
36 are the uniform interface
37 between the user process and the network protocol stacks in the kernel.
38 The protocol modules are grouped into
41 .BR AF_INET ", " AF_IPX ", " AF_PACKET
50 for more information on families and types.
51 .SS Socket Layer Functions
52 These functions are used by the user process to send or receive packets
53 and to do other socket operations.
54 For more information see their respective manual pages.
59 connects a socket to a remote socket address,
62 function binds a socket to a local socket address,
64 tells the socket that new connections shall be accepted, and
66 is used to get a new socket with a new incoming connection.
68 returns two connected anonymous sockets (only implemented for a few
76 send data over a socket, and
80 receive data from a socket.
84 wait for arriving data or a readiness to send data.
85 In addition, the standard I/O operations like
92 can be used to read and write data.
95 returns the local socket address and
97 returns the remote socket address.
101 are used to set or get socket layer or protocol options.
103 can be used to set or read some other options.
106 is used to close a socket.
108 closes parts of a full-duplex socket connection.
114 with a nonzero position is not supported on sockets.
116 It is possible to do nonblocking I/O on sockets by setting the
118 flag on a socket file descriptor using
120 Then all operations that would block will (usually)
123 (operation should be retried later);
128 The user can then wait for various events via
137 Event:Poll flag:Occurrence
142 A connection setup has been completed
143 (for connection-oriented sockets)
146 A disconnection request has been initiated by the other end.
149 A connection is broken (only for connection-oriented protocols).
150 When the socket is written
155 Socket has enough send buffer space for writing new data.
166 Read/Write:POLLERR:An asynchronous error occurred.
167 Read/Write:POLLHUP:The other end has shut down one direction.
173 .\" FIXME . The following is not true currently:
174 .\" It is no I/O event when the connection
175 .\" is broken from the local end using
186 is to let the kernel inform the application about events
192 flag must be set on a socket file descriptor via
194 and a valid signal handler for
196 must be installed via
202 These socket options can be set by using
206 with the socket level set to
210 .\" In the list below, the text used to describe argument types
211 .\" for each socket option should be more consistent
213 .\" SO_ACCEPTCONN is in POSIX.1-2001, and its origin is explained in
214 .\" W R Stevens, UNPv1
217 Returns a value indicating whether or not this socket has been marked
218 to accept connections with
220 The value 0 indicates that this is not a listening socket,
221 the value 1 indicates that this is a listening socket.
222 This socket option is read-only.
225 Bind this socket to a particular device like \(lqeth0\(rq,
226 as specified in the passed interface name.
228 name is an empty string or the option length is zero, the socket device
230 The passed option is a variable-length null-terminated
231 interface name string with the maximum size of
233 If a socket is bound to an interface,
234 only packets received from that particular interface are processed by the
236 Note that this only works for some socket types, particularly
239 It is not supported for packet sockets (use normal
244 Set or get the broadcast flag.
245 When enabled, datagram sockets are allowed to send
246 packets to a broadcast address.
247 This option has no effect on stream-oriented sockets.
250 Enable BSD bug-to-bug compatibility.
251 This is used by the UDP protocol module in Linux 2.0 and 2.2.
252 If enabled ICMP errors received for a UDP socket will not be passed
254 In later kernel versions, support for this option has been phased out:
255 Linux 2.4 silently ignores it, and Linux 2.6 generates a kernel warning
256 (printk()) if a program uses this option.
257 Linux 2.0 also enabled BSD bug-to-bug compatibility
258 options (random header changing, skipping of the broadcast flag) for raw
259 sockets with this option, but that was removed in Linux 2.2.
262 Enable socket debugging.
263 Only allowed for processes with the
265 capability or an effective user ID of 0.
267 .BR SO_DOMAIN " (since Linux 2.6.32)"
268 Retrieves the socket domain as an integer, returning a value such as
273 This socket option is read-only.
276 Get and clear the pending socket error.
277 This socket option is read-only.
281 Don't send via a gateway, only send to directly connected hosts.
282 The same effect can be achieved by setting the
287 Expects an integer boolean flag.
290 Enable sending of keep-alive messages on connection-oriented sockets.
291 Expects an integer boolean flag.
304 int l_onoff; /* linger active */
305 int l_linger; /* how many seconds to linger for */
314 will not return until all queued messages for the socket have been
315 successfully sent or the linger timeout has been reached.
317 the call returns immediately and the closing is done in the background.
318 When the socket is closed as part of
320 it always lingers in the background.
322 .BR SO_MARK " (since Linux 2.6.25)"
323 .\" commit 4a19ec5800fc3bb64e2d87c4d9fdd9e636086fe0
324 .\" and 914a9ab386a288d0f22252fc268ecbc048cdcbd5
325 Set the mark for each packet sent through this socket
326 (similar to the netfilter MARK target but socket-based).
327 Changing the mark can be used for mark-based
328 routing without netfilter or for packet filtering.
329 Setting this option requires the
334 If this option is enabled,
335 out-of-band data is directly placed into the receive data stream.
336 Otherwise out-of-band data is only passed when the
338 flag is set during receiving.
339 .\" don't document it because it can do too much harm.
343 Enable or disable the receiving of the
346 For more information see
348 .\" FIXME Document SO_PASSSEC, added in 2.6.18; there is some info
349 .\" in the 2.6.18 ChangeLog
352 Return the credentials of the foreign process connected to this socket.
353 This is only possible for connected
357 stream and datagram socket pairs created using
361 The returned credentials are those that were in effect at the time
369 This socket option is read-only.
372 Set the protocol-defined priority for all packets to be sent on
374 Linux uses this value to order the networking queues:
375 packets with a higher priority may be processed first depending
376 on the selected device queueing discipline.
379 this also sets the IP type-of-service (TOS) field for outgoing packets.
380 Setting a priority outside the range 0 to 6 requires the
384 .BR SO_PROTOCOL " (since Linux 2.6.32)"
385 Retrieves the socket protocol as an integer, returning a value such as
390 This socket option is read-only.
393 Sets or gets the maximum socket receive buffer in bytes.
394 The kernel doubles this value (to allow space for bookkeeping overhead)
396 .\" Most (all?) other implementations do not do this -- MTK, Dec 05
398 and this doubled value is returned by
400 .\" The following thread on LMKL is quite informative:
401 .\" getsockopt/setsockopt with SO_RCVBUF and SO_SNDBUF "non-standard" behaviour
403 .\" http://thread.gmane.org/gmane.linux.kernel/1328935
404 The default value is set by the
405 .I /proc/sys/net/core/rmem_default
406 file, and the maximum allowed value is set by the
407 .I /proc/sys/net/core/rmem_max
409 The minimum (doubled) value for this option is 256.
411 .BR SO_RCVBUFFORCE " (since Linux 2.6.14)"
412 Using this socket option, a privileged
413 .RB ( CAP_NET_ADMIN )
414 process can perform the same task as
418 limit can be overridden.
420 .BR SO_RCVLOWAT " and " SO_SNDLOWAT
421 Specify the minimum number of bytes in the buffer until the socket layer
422 will pass the data to the protocol
424 or the user on receiving
426 These two values are initialized to 1.
428 is not changeable on Linux
434 only since Linux 2.4.
439 system calls currently do not respect the
442 and mark a socket readable when even a single byte of data is available.
443 A subsequent read from the socket will block until
446 .\" See http://marc.theaimsgroup.com/?l=linux-kernel&m=111049368106984&w=2
447 .\" Tested on kernel 2.6.14 -- mtk, 30 Nov 05
449 .BR SO_RCVTIMEO " and " SO_SNDTIMEO
450 .\" Not implemented in 2.0.
451 .\" Implemented in 2.1.11 for getsockopt: always return a zero struct.
452 .\" Implemented in 2.3.41 for setsockopt, and actually used.
453 Specify the receiving or sending timeouts until reporting an error.
455 .IR "struct timeval" .
456 If an input or output function blocks for this period of time, and
457 data has been sent or received, the return value of that function
458 will be the amount of data transferred; if no data has been transferred
459 and the timeout has been reached then \-1 is returned with
465 .\" in fact to EAGAIN
470 just as if the socket was specified to be nonblocking.
471 If the timeout is set to zero (the default)
472 then the operation will never timeout.
473 Timeouts only have effect for system calls that perform socket I/O (e.g.,
478 timeouts have no effect for
485 Indicates that the rules used in validating addresses supplied in a
487 call should allow reuse of local addresses.
491 means that a socket may bind, except when there
492 is an active listening socket bound to the address.
493 When the listening socket is bound to
495 with a specific port then it is not possible
496 to bind to this port for any local address.
497 Argument is an integer boolean flag.
500 Sets or gets the maximum socket send buffer in bytes.
501 The kernel doubles this value (to allow space for bookkeeping overhead)
503 .\" Most (all?) other implementations do not do this -- MTK, Dec 05
504 .\" See also the comment to SO_RCVBUF (17 Jul 2012 LKML mail)
506 and this doubled value is returned by
508 The default value is set by the
509 .I /proc/sys/net/core/wmem_default
510 file and the maximum allowed value is set by the
511 .I /proc/sys/net/core/wmem_max
513 The minimum (doubled) value for this option is 2048.
515 .BR SO_SNDBUFFORCE " (since Linux 2.6.14)"
516 Using this socket option, a privileged
517 .RB ( CAP_NET_ADMIN )
518 process can perform the same task as
522 limit can be overridden.
525 Enable or disable the receiving of the
528 The timestamp control message is sent with level
535 reception time of the last packet passed to the user in this call.
538 for details on control messages.
541 Gets the socket type as an integer (e.g.,
543 This socket option is read-only.
545 When writing onto a connection-oriented socket that has been shut down
546 (by the local or the remote end)
548 is sent to the writing process and
551 The signal is not sent when the write call
556 When requested with the
563 is sent when an I/O event occurs.
564 It is possible to use
568 in the signal handler to find out which socket the event occurred on.
569 An alternative (in Linux 2.2) is to set a real-time signal using the
572 the handler of the real time signal will be called with
573 the file descriptor in the
579 for more information.
581 Under some circumstances (e.g., multiple processes accessing a
582 single socket), the condition that caused the
584 may have already disappeared when the process reacts to the signal.
585 If this happens, the process should wait again because Linux
586 will resend the signal later.
587 .\" .SS Ancillary Messages
589 The core socket networking parameters can be accessed
590 via files in the directory
591 .IR /proc/sys/net/core/ .
594 contains the default setting in bytes of the socket receive buffer.
597 contains the maximum socket receive buffer size in bytes which a user may
603 contains the default setting in bytes of the socket send buffer.
606 contains the maximum socket send buffer size in bytes which a user may
611 .IR message_cost " and " message_burst
612 configure the token bucket filter used to load limit warning messages
613 caused by external network events.
615 .I netdev_max_backlog
616 Maximum number of packets in the global input queue.
619 Maximum length of ancillary data and user control data like the iovecs
621 .\" netdev_fastroute is not documented because it is experimental
623 These operations can be accessed using
628 .IB error " = ioctl(" ip_socket ", " ioctl_type ", " &value_result ");"
635 with the receive timestamp of the last packet passed to the user.
636 This is useful for accurate round trip time measurements.
640 .IR "struct timeval" .
642 This ioctl should only be used if the socket option
644 is not set on the socket.
645 Otherwise, it returns the timestamp of the
646 last packet that was received while
648 was not set, or it fails if no such packet has been received,
657 Set the process or process group to send
663 asynchronous I/O operation has finished or urgent data is available.
664 The argument is a pointer to a
666 If the argument is positive, send the signals to that process.
668 argument is negative, send the signals to the process group with the ID
669 of the absolute value of the argument.
670 The process may only choose itself or its own process group to receive
671 signals unless it has the
673 capability or an effective UID of 0.
678 flag to enable or disable asynchronous I/O mode of the socket.
679 Asynchronous I/O mode means that the
681 signal or the signal set with
683 is raised when a new I/O event occurs.
685 Argument is an integer boolean flag.
686 (This operation is synonymous with the use of
694 Get the current process or process group that receives
717 was introduced in Linux 2.0.30.
722 interfaces was introduced in Linux 2.2.
726 are supported since Linux 2.3.41.
727 Earlier, timeouts were fixed to
728 a protocol-specific setting, and could not be read or written.
730 Linux assumes that half of the send/receive buffer is used for internal
731 kernel structures; thus the values in the corresponding
733 files are twice what can be observed on the wire.
735 Linux will only allow port reuse with the
738 when this option was set both in the previous program that performed a
740 to the port and in the program that wants to reuse the port.
741 This differs from some implementations (e.g., FreeBSD)
742 where only the later program needs to set the
745 Typically this difference is invisible, since, for example, a server
746 program is designed to always set this option.
754 .\" FIXME Document SO_ATTACH_FILTER and SO_DETACH_FILTER
756 The suggested interface to use them is via the libpcap
759 .\" This man page was written by Andi Kleen.
765 .BR capabilities (7),