2 .\" This man page is Copyright (C) 1999 Andi Kleen <ak@muc.de>.
3 .\" and copyright (c) 1999 Matthew Wilcox.
5 .\" %%%LICENSE_START(VERBATIM_ONE_PARA)
6 .\" Permission is granted to distribute possibly modified copies
7 .\" of this page provided the header is included verbatim,
8 .\" and in case of nontrivial modification author and date
9 .\" of the modification is added to the header.
12 .\" 2002-10-30, Michael Kerrisk, <mtk.manpages@gmail.com>
13 .\" Added description of SO_ACCEPTCONN
14 .\" 2004-05-20, aeb, added SO_RCVTIMEO/SO_SNDTIMEO text.
15 .\" Modified, 27 May 2004, Michael Kerrisk <mtk.manpages@gmail.com>
16 .\" Added notes on capability requirements
17 .\" A few small grammar fixes
18 .\" 2010-06-13 Jan Engelhardt <jengelh@medozas.de>
19 .\" Documented SO_DOMAIN and SO_PROTOCOL.
21 .\" The following are not yet documented:
22 .\" SO_PEERNAME (2.4?)
24 .\" Seems to do something similar to getpeernam(), but then
25 .\" why is it necessary / how does it differ?
26 .\" SO_TIMESTAMPNS (2.6.22)
27 .\" Documentation/networking/timestamping.txt
28 .\" commit 92f37fd2ee805aa77925c1e64fd56088b46094fc
29 .\" Author: Eric Dumazet <dada1@cosmosbay.com>
30 .\" SO_TIMESTAMPING (2.6.30)
31 .\" Documentation/networking/timestamping.txt
32 .\" commit cb9eff097831007afb30d64373f29d99825d0068
33 .\" Author: Patrick Ohly <patrick.ohly@intel.com>
34 .\" SO_RXQ_OVFL (2.6.33)
35 .\" commit 3b885787ea4112eaa80945999ea0901bf742707f
36 .\" Author: Neil Horman <nhorman@tuxdriver.com>
37 .\" SO_WIFI_STATUS (3.3)
38 .\" commit 6e3e939f3b1bf8534b32ad09ff199d88800835a0
39 .\" Author: Johannes Berg <johannes.berg@intel.com>
40 .\" Also: SCM_WIFI_STATUS
42 .\" commit 3bdc0eba0b8b47797f4a76e377dd8360f317450f
43 .\" Author: Ben Greear <greearb@candelatech.com>
45 .TH SOCKET 7 2013-04-17 Linux "Linux Programmer's Manual"
47 socket \- Linux socket interface
49 .B #include <sys/socket.h>
51 .IB sockfd " = socket(int " socket_family ", int " socket_type ", int " protocol );
53 This manual page describes the Linux networking socket layer user
55 The BSD compatible sockets
56 are the uniform interface
57 between the user process and the network protocol stacks in the kernel.
58 The protocol modules are grouped into
61 .BR AF_INET ", " AF_IPX ", " AF_PACKET
70 for more information on families and types.
71 .SS Socket-layer functions
72 These functions are used by the user process to send or receive packets
73 and to do other socket operations.
74 For more information see their respective manual pages.
79 connects a socket to a remote socket address,
82 function binds a socket to a local socket address,
84 tells the socket that new connections shall be accepted, and
86 is used to get a new socket with a new incoming connection.
88 returns two connected anonymous sockets (implemented only for a few
96 send data over a socket, and
100 receive data from a socket.
104 wait for arriving data or a readiness to send data.
105 In addition, the standard I/O operations like
112 can be used to read and write data.
115 returns the local socket address and
117 returns the remote socket address.
121 are used to set or get socket layer or protocol options.
123 can be used to set or read some other options.
126 is used to close a socket.
128 closes parts of a full-duplex socket connection.
134 with a nonzero position is not supported on sockets.
136 It is possible to do nonblocking I/O on sockets by setting the
138 flag on a socket file descriptor using
140 Then all operations that would block will (usually)
143 (operation should be retried later);
148 The user can then wait for various events via
157 Event:Poll flag:Occurrence
162 A connection setup has been completed
163 (for connection-oriented sockets)
166 A disconnection request has been initiated by the other end.
169 A connection is broken (only for connection-oriented protocols).
170 When the socket is written
175 Socket has enough send buffer space for writing new data.
186 Read/Write:POLLERR:An asynchronous error occurred.
187 Read/Write:POLLHUP:The other end has shut down one direction.
193 .\" FIXME . The following is not true currently:
194 .\" It is no I/O event when the connection
195 .\" is broken from the local end using
205 is to let the kernel inform the application about events
211 flag must be set on a socket file descriptor via
213 and a valid signal handler for
215 must be installed via
220 .SS Socket address structures
221 Each socket domain has its own format for socket addresses,
222 with a domain-specific address structure.
223 Each of these structures begins with an
224 integer "family" field (typed as
226 that indicates the type of the address structure.
228 the various system calls (e.g.,
233 .BR getpeername (2)),
234 which are generic to all socket domains,
235 to determine the domain of a particular socket address.
237 To allow any type of socket address to be passed to
238 interfaces in the sockets API,
240 .IR "struct sockaddr"
242 The purpose of this type is purely to allow casting of
243 domain-specific socket address types to a "generic" type,
244 so as to avoid compiler warnings about type mismatches in
245 calls to the sockets API.
247 In addition, the sockets API provides the data type
248 .IR "struct sockaddr_storage".
250 is suitable to accommodate all supported domain-specific socket
251 address structures; it is large enough and is aligned properly.
252 (In particular, it is large enough to hold
253 IPv6 socket addresses.)
254 The structure includes the following field, which can be used to identify
255 the type of socket address actually stored in the structure:
259 sa_family_t ss_family;
265 structure is useful in programs that must handle socket addresses
267 (e.g., programs that must deal with both IPv4 and IPv6 socket addresses).
269 These socket options can be set by using
273 with the socket level set to
277 .\" In the list below, the text used to describe argument types
278 .\" for each socket option should be more consistent
280 .\" SO_ACCEPTCONN is in POSIX.1-2001, and its origin is explained in
281 .\" W R Stevens, UNPv1
284 Returns a value indicating whether or not this socket has been marked
285 to accept connections with
287 The value 0 indicates that this is not a listening socket,
288 the value 1 indicates that this is a listening socket.
289 This socket option is read-only.
292 Bind this socket to a particular device like \(lqeth0\(rq,
293 as specified in the passed interface name.
295 name is an empty string or the option length is zero, the socket device
297 The passed option is a variable-length null-terminated
298 interface name string with the maximum size of
300 If a socket is bound to an interface,
301 only packets received from that particular interface are processed by the
303 Note that this works only for some socket types, particularly
306 It is not supported for packet sockets (use normal
311 this socket option could be set, but could not retrieved with
313 Since Linux 3.8, it is readable.
316 argument should contain the buffer size available
317 to receive the device name and is recommended to be
320 The real device name length is reported back in the
325 Set or get the broadcast flag.
326 When enabled, datagram sockets are allowed to send
327 packets to a broadcast address.
328 This option has no effect on stream-oriented sockets.
331 Enable BSD bug-to-bug compatibility.
332 This is used by the UDP protocol module in Linux 2.0 and 2.2.
333 If enabled ICMP errors received for a UDP socket will not be passed
335 In later kernel versions, support for this option has been phased out:
336 Linux 2.4 silently ignores it, and Linux 2.6 generates a kernel warning
337 (printk()) if a program uses this option.
338 Linux 2.0 also enabled BSD bug-to-bug compatibility
339 options (random header changing, skipping of the broadcast flag) for raw
340 sockets with this option, but that was removed in Linux 2.2.
343 Enable socket debugging.
344 Only allowed for processes with the
346 capability or an effective user ID of 0.
348 .BR SO_DOMAIN " (since Linux 2.6.32)"
349 Retrieves the socket domain as an integer, returning a value such as
354 This socket option is read-only.
357 Get and clear the pending socket error.
358 This socket option is read-only.
362 Don't send via a gateway, send only to directly connected hosts.
363 The same effect can be achieved by setting the
368 Expects an integer boolean flag.
371 Enable sending of keep-alive messages on connection-oriented sockets.
372 Expects an integer boolean flag.
385 int l_onoff; /* linger active */
386 int l_linger; /* how many seconds to linger for */
395 will not return until all queued messages for the socket have been
396 successfully sent or the linger timeout has been reached.
398 the call returns immediately and the closing is done in the background.
399 When the socket is closed as part of
401 it always lingers in the background.
403 .BR SO_MARK " (since Linux 2.6.25)"
404 .\" commit 4a19ec5800fc3bb64e2d87c4d9fdd9e636086fe0
405 .\" and 914a9ab386a288d0f22252fc268ecbc048cdcbd5
406 Set the mark for each packet sent through this socket
407 (similar to the netfilter MARK target but socket-based).
408 Changing the mark can be used for mark-based
409 routing without netfilter or for packet filtering.
410 Setting this option requires the
415 If this option is enabled,
416 out-of-band data is directly placed into the receive data stream.
417 Otherwise out-of-band data is passed only when the
419 flag is set during receiving.
420 .\" don't document it because it can do too much harm.
424 Enable or disable the receiving of the
427 For more information see
429 .\" FIXME Document SO_PASSSEC, added in 2.6.18; there is some info
430 .\" in the 2.6.18 ChangeLog
432 .BR SO_PEEK_OFF " (since Linux 3.4)"
433 .\" commit ef64a54f6e558155b4f149bb10666b9e914b6c54
434 This option, which is currently supported only for
436 sockets, sets the value of the "peek offset" for the
438 system call when used with
442 When this option is set to a negative value
443 (it is set to \-1 for all new sockets),
444 traditional behavior is provided:
448 flag will peek data from the front of the queue.
450 When the option is set to a value greater than or equal to zero,
451 then the next peek at data queued in the socket will occur at
452 the byte offset specified by the option value.
453 At the same time, the "peek offset" will be
454 incremented by the number of bytes that were peeked from the queue,
455 so that a subsequent peek will return the next data in the queue.i
457 If data is removed from the front of the queue via a call to
459 (or similar) without the
461 flag, the "peek offset" will be decreased by the number of bytes removed.
462 In other words, receiving data without the
464 flag will cause the "peek offset" to be adjusted to maintain
465 the correct relative position in the queued data,
466 so that a subsequent peek will retrieve the data that would have been
467 retrieved had the data not been removed.
469 For datagram sockets, if the "peek offset" points to the middle of a packet,
470 the data returned will be marked with the
474 The following example serves to illustrate the use of
476 Suppose a stream socket has the following queued input data:
481 The following sequence of
483 calls would have the effect noted in the comments:
487 int ov = 4; // Set peek offset to 4
488 setsockopt(fd, SOL_SOCKET, SO_PEEK_OFF, &ov, sizeof(ov));
490 recv(fd, buf, 2, MSG_PEEK); // Peeks "cc"; offset set to 6
491 recv(fd, buf, 2, MSG_PEEK); // Peeks "dd"; offset set to 8
492 recv(fd, buf, 2, 0); // Reads "aa"; offset set to 6
493 recv(fd, buf, 2, MSG_PEEK); // Peeks "ee"; offset set to 8
498 Return the credentials of the foreign process connected to this socket.
499 This is possible only for connected
503 stream and datagram socket pairs created using
507 The returned credentials are those that were in effect at the time
514 structure; define the
516 feature test macro to obtain the definition of that structure from
518 This socket option is read-only.
521 Set the protocol-defined priority for all packets to be sent on
523 Linux uses this value to order the networking queues:
524 packets with a higher priority may be processed first depending
525 on the selected device queueing discipline.
528 this also sets the IP type-of-service (TOS) field for outgoing packets.
529 Setting a priority outside the range 0 to 6 requires the
533 .BR SO_PROTOCOL " (since Linux 2.6.32)"
534 Retrieves the socket protocol as an integer, returning a value such as
539 This socket option is read-only.
542 Sets or gets the maximum socket receive buffer in bytes.
543 The kernel doubles this value (to allow space for bookkeeping overhead)
545 .\" Most (all?) other implementations do not do this -- MTK, Dec 05
547 and this doubled value is returned by
549 .\" The following thread on LMKL is quite informative:
550 .\" getsockopt/setsockopt with SO_RCVBUF and SO_SNDBUF "non-standard" behaviour
552 .\" http://thread.gmane.org/gmane.linux.kernel/1328935
553 The default value is set by the
554 .I /proc/sys/net/core/rmem_default
555 file, and the maximum allowed value is set by the
556 .I /proc/sys/net/core/rmem_max
558 The minimum (doubled) value for this option is 256.
560 .BR SO_RCVBUFFORCE " (since Linux 2.6.14)"
561 Using this socket option, a privileged
562 .RB ( CAP_NET_ADMIN )
563 process can perform the same task as
567 limit can be overridden.
569 .BR SO_RCVLOWAT " and " SO_SNDLOWAT
570 Specify the minimum number of bytes in the buffer until the socket layer
571 will pass the data to the protocol
573 or the user on receiving
575 These two values are initialized to 1.
577 is not changeable on Linux
583 only since Linux 2.4.
588 system calls currently do not respect the
591 and mark a socket readable when even a single byte of data is available.
592 A subsequent read from the socket will block until
595 .\" See http://marc.theaimsgroup.com/?l=linux-kernel&m=111049368106984&w=2
596 .\" Tested on kernel 2.6.14 -- mtk, 30 Nov 05
598 .BR SO_RCVTIMEO " and " SO_SNDTIMEO
599 .\" Not implemented in 2.0.
600 .\" Implemented in 2.1.11 for getsockopt: always return a zero struct.
601 .\" Implemented in 2.3.41 for setsockopt, and actually used.
602 Specify the receiving or sending timeouts until reporting an error.
604 .IR "struct timeval" .
605 If an input or output function blocks for this period of time, and
606 data has been sent or received, the return value of that function
607 will be the amount of data transferred; if no data has been transferred
608 and the timeout has been reached then \-1 is returned with
614 .\" in fact to EAGAIN
619 just as if the socket was specified to be nonblocking.
620 If the timeout is set to zero (the default)
621 then the operation will never timeout.
622 Timeouts only have effect for system calls that perform socket I/O (e.g.,
627 timeouts have no effect for
634 Indicates that the rules used in validating addresses supplied in a
636 call should allow reuse of local addresses.
640 means that a socket may bind, except when there
641 is an active listening socket bound to the address.
642 When the listening socket is bound to
644 with a specific port then it is not possible
645 to bind to this port for any local address.
646 Argument is an integer boolean flag.
649 Sets or gets the maximum socket send buffer in bytes.
650 The kernel doubles this value (to allow space for bookkeeping overhead)
652 .\" Most (all?) other implementations do not do this -- MTK, Dec 05
653 .\" See also the comment to SO_RCVBUF (17 Jul 2012 LKML mail)
655 and this doubled value is returned by
657 The default value is set by the
658 .I /proc/sys/net/core/wmem_default
659 file and the maximum allowed value is set by the
660 .I /proc/sys/net/core/wmem_max
662 The minimum (doubled) value for this option is 2048.
664 .BR SO_SNDBUFFORCE " (since Linux 2.6.14)"
665 Using this socket option, a privileged
666 .RB ( CAP_NET_ADMIN )
667 process can perform the same task as
671 limit can be overridden.
674 Enable or disable the receiving of the
677 The timestamp control message is sent with level
684 reception time of the last packet passed to the user in this call.
687 for details on control messages.
690 Gets the socket type as an integer (e.g.,
692 This socket option is read-only.
694 When writing onto a connection-oriented socket that has been shut down
695 (by the local or the remote end)
697 is sent to the writing process and
700 The signal is not sent when the write call
705 When requested with the
712 is sent when an I/O event occurs.
713 It is possible to use
717 in the signal handler to find out which socket the event occurred on.
718 An alternative (in Linux 2.2) is to set a real-time signal using the
721 the handler of the real time signal will be called with
722 the file descriptor in the
728 for more information.
730 Under some circumstances (e.g., multiple processes accessing a
731 single socket), the condition that caused the
733 may have already disappeared when the process reacts to the signal.
734 If this happens, the process should wait again because Linux
735 will resend the signal later.
736 .\" .SS Ancillary messages
738 The core socket networking parameters can be accessed
739 via files in the directory
740 .IR /proc/sys/net/core/ .
743 contains the default setting in bytes of the socket receive buffer.
746 contains the maximum socket receive buffer size in bytes which a user may
752 contains the default setting in bytes of the socket send buffer.
755 contains the maximum socket send buffer size in bytes which a user may
760 .IR message_cost " and " message_burst
761 configure the token bucket filter used to load limit warning messages
762 caused by external network events.
764 .I netdev_max_backlog
765 Maximum number of packets in the global input queue.
768 Maximum length of ancillary data and user control data like the iovecs
770 .\" netdev_fastroute is not documented because it is experimental
772 These operations can be accessed using
777 .IB error " = ioctl(" ip_socket ", " ioctl_type ", " &value_result ");"
784 with the receive timestamp of the last packet passed to the user.
785 This is useful for accurate round trip time measurements.
789 .IR "struct timeval" .
791 This ioctl should be used only if the socket option
793 is not set on the socket.
794 Otherwise, it returns the timestamp of the
795 last packet that was received while
797 was not set, or it fails if no such packet has been received,
806 Set the process or process group to send
812 asynchronous I/O operation has finished or urgent data is available.
813 The argument is a pointer to a
815 If the argument is positive, send the signals to that process.
817 argument is negative, send the signals to the process group with the ID
818 of the absolute value of the argument.
819 The process may only choose itself or its own process group to receive
820 signals unless it has the
822 capability or an effective UID of 0.
827 flag to enable or disable asynchronous I/O mode of the socket.
828 Asynchronous I/O mode means that the
830 signal or the signal set with
832 is raised when a new I/O event occurs.
834 Argument is an integer boolean flag.
835 (This operation is synonymous with the use of
843 Get the current process or process group that receives
866 was introduced in Linux 2.0.30.
871 interfaces was introduced in Linux 2.2.
875 are supported since Linux 2.3.41.
876 Earlier, timeouts were fixed to
877 a protocol-specific setting, and could not be read or written.
879 Linux assumes that half of the send/receive buffer is used for internal
880 kernel structures; thus the values in the corresponding
882 files are twice what can be observed on the wire.
884 Linux will only allow port reuse with the
887 when this option was set both in the previous program that performed a
889 to the port and in the program that wants to reuse the port.
890 This differs from some implementations (e.g., FreeBSD)
891 where only the later program needs to set the
894 Typically this difference is invisible, since, for example, a server
895 program is designed to always set this option.
903 .\" FIXME Document SO_ATTACH_FILTER and SO_DETACH_FILTER
905 The suggested interface to use them is via the libpcap
908 .\" This man page was written by Andi Kleen.
914 .BR capabilities (7),