man7/socket.7

   1 '\" t
   2 .\" This man page is Copyright (C) 1999 Andi Kleen <ak@muc.de>.
   3 .\" and copyright (c) 1999 Matthew Wilcox.
   4 .\"
   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.
  10 .\" %%%LICENSE_END
  11 .\"
  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.
  20 .\" FIXME
  21 .\" The following are not yet documented:
  22 .\"     SO_PEERNAME (2.4?)
  23 .\"             get only
  24 .\"             Seems to do something similar to getpeername(), 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_WIFI_STATUS (3.3)
  35 .\"             commit 6e3e939f3b1bf8534b32ad09ff199d88800835a0
  36 .\"             Author: Johannes Berg <johannes.berg@intel.com>
  37 .\"             Also: SCM_WIFI_STATUS
  38 .\"     SO_NOFCS (3.4)
  39 .\"             commit 3bdc0eba0b8b47797f4a76e377dd8360f317450f
  40 .\"             Author: Ben Greear <greearb@candelatech.com>
  41 .\"     SO_GET_FILTER (3.8)
  42 .\"             commit a8fc92778080c845eaadc369a0ecf5699a03bef0
  43 .\"             Author: Pavel Emelyanov <xemul@parallels.com>
  44 .\"     SO_LOCK_FILTER (3.9)
  45 .\"             commit d59577b6ffd313d0ab3be39cb1ab47e29bdc9182
  46 .\"             Author: Vincent Bernat <bernat@luffy.cx>
  47 .\"     SO_SELECT_ERR_QUEUE (3.10)
  48 .\"             commit 7d4c04fc170087119727119074e72445f2bb192b
  49 .\"             Author: Keller, Jacob E <jacob.e.keller@intel.com>
  50 .\"     SO_MAX_PACING_RATE (3.13)
  51 .\"             commit 62748f32d501f5d3712a7c372bbb92abc7c62bc7
  52 .\"     SO_BPF_EXTENSIONS (3.14)
  53 .\"             commit ea02f9411d9faa3553ed09ce0ec9f00ceae9885e
  54 .\"
  55 .TH SOCKET 7 2015-05-07 Linux "Linux Programmer's Manual"
  56 .SH NAME
  57 socket \- Linux socket interface
  58 .SH SYNOPSIS
  59 .B #include <sys/socket.h>
  60 .sp
  61 .IB sockfd " = socket(int " socket_family ", int " socket_type ", int " protocol );
  62 .SH DESCRIPTION
  63 This manual page describes the Linux networking socket layer user
  64 interface.
  65 The BSD compatible sockets
  66 are the uniform interface
  67 between the user process and the network protocol stacks in the kernel.
  68 The protocol modules are grouped into
  69 .I protocol families
  70 such as
  71 .BR AF_INET ", " AF_IPX ", and " AF_PACKET ,
  72 and
  73 .I socket types
  74 such as
  75 .B SOCK_STREAM
  76 or
  77 .BR SOCK_DGRAM .
  78 See
  79 .BR socket (2)
  80 for more information on families and types.
  81 .SS Socket-layer functions
  82 These functions are used by the user process to send or receive packets
  83 and to do other socket operations.
  84 For more information see their respective manual pages.
  85
  86 .BR socket (2)
  87 creates a socket,
  88 .BR connect (2)
  89 connects a socket to a remote socket address,
  90 the
  91 .BR bind (2)
  92 function binds a socket to a local socket address,
  93 .BR listen (2)
  94 tells the socket that new connections shall be accepted, and
  95 .BR accept (2)
  96 is used to get a new socket with a new incoming connection.
  97 .BR socketpair (2)
  98 returns two connected anonymous sockets (implemented only for a few
  99 local families like
 100 .BR AF_UNIX )
 101 .PP
 102 .BR send (2),
 103 .BR sendto (2),
 104 and
 105 .BR sendmsg (2)
 106 send data over a socket, and
 107 .BR recv (2),
 108 .BR recvfrom (2),
 109 .BR recvmsg (2)
 110 receive data from a socket.
 111 .BR poll (2)
 112 and
 113 .BR select (2)
 114 wait for arriving data or a readiness to send data.
 115 In addition, the standard I/O operations like
 116 .BR write (2),
 117 .BR writev (2),
 118 .BR sendfile (2),
 119 .BR read (2),
 120 and
 121 .BR readv (2)
 122 can be used to read and write data.
 123 .PP
 124 .BR getsockname (2)
 125 returns the local socket address and
 126 .BR getpeername (2)
 127 returns the remote socket address.
 128 .BR getsockopt (2)
 129 and
 130 .BR setsockopt (2)
 131 are used to set or get socket layer or protocol options.
 132 .BR ioctl (2)
 133 can be used to set or read some other options.
 134 .PP
 135 .BR close (2)
 136 is used to close a socket.
 137 .BR shutdown (2)
 138 closes parts of a full-duplex socket connection.
 139 .PP
 140 Seeking, or calling
 141 .BR pread (2)
 142 or
 143 .BR pwrite (2)
 144 with a nonzero position is not supported on sockets.
 145 .PP
 146 It is possible to do nonblocking I/O on sockets by setting the
 147 .B O_NONBLOCK
 148 flag on a socket file descriptor using
 149 .BR fcntl (2).
 150 Then all operations that would block will (usually)
 151 return with
 152 .B EAGAIN
 153 (operation should be retried later);
 154 .BR connect (2)
 155 will return
 156 .B EINPROGRESS
 157 error.
 158 The user can then wait for various events via
 159 .BR poll (2)
 160 or
 161 .BR select (2).
 162 .TS
 163 tab(:) allbox;
 164 c s s
 165 l l l.
 166 I/O events
 167 Event:Poll flag:Occurrence
 168 Read:POLLIN:T{
 169 New data arrived.
 170 T}
 171 Read:POLLIN:T{
 172 A connection setup has been completed
 173 (for connection-oriented sockets)
 174 T}
 175 Read:POLLHUP:T{
 176 A disconnection request has been initiated by the other end.
 177 T}
 178 Read:POLLHUP:T{
 179 A connection is broken (only for connection-oriented protocols).
 180 When the socket is written
 181 .B SIGPIPE
 182 is also sent.
 183 T}
 184 Write:POLLOUT:T{
 185 Socket has enough send buffer space for writing new data.
 186 T}
 187 Read/Write:T{
 188 POLLIN|
 189 .br
 190 POLLOUT
 191 T}:T{
 192 An outgoing
 193 .BR connect (2)
 194 finished.
 195 T}
 196 Read/Write:POLLERR:An asynchronous error occurred.
 197 Read/Write:POLLHUP:The other end has shut down one direction.
 198 Exception:POLLPRI:T{
 199 Urgent data arrived.
 200 .B SIGURG
 201 is sent then.
 202 T}
 203 .\" FIXME . The following is not true currently:
 204 .\" It is no I/O event when the connection
 205 .\" is broken from the local end using
 206 .\" .BR shutdown (2)
 207 .\" or
 208 .\" .BR close (2).
 209 .TE
 210 .PP
 211 An alternative to
 212 .BR poll (2)
 213 and
 214 .BR select (2)
 215 is to let the kernel inform the application about events
 216 via a
 217 .B SIGIO
 218 signal.
 219 For that the
 220 .B O_ASYNC
 221 flag must be set on a socket file descriptor via
 222 .BR fcntl (2)
 223 and a valid signal handler for
 224 .B SIGIO
 225 must be installed via
 226 .BR sigaction (2).
 227 See the
 228 .I Signals
 229 discussion below.
 230 .SS Socket address structures
 231 Each socket domain has its own format for socket addresses,
 232 with a domain-specific address structure.
 233 Each of these structures begins with an
 234 integer "family" field (typed as
 235 .IR sa_family_t )
 236 that indicates the type of the address structure.
 237 This allows
 238 the various system calls (e.g.,
 239 .BR connect (2),
 240 .BR bind (2),
 241 .BR accept (2),
 242 .BR getsockname (2),
 243 .BR getpeername (2)),
 244 which are generic to all socket domains,
 245 to determine the domain of a particular socket address.
 246
 247 To allow any type of socket address to be passed to
 248 interfaces in the sockets API,
 249 the type
 250 .IR "struct sockaddr"
 251 is defined.
 252 The purpose of this type is purely to allow casting of
 253 domain-specific socket address types to a "generic" type,
 254 so as to avoid compiler warnings about type mismatches in
 255 calls to the sockets API.
 256
 257 In addition, the sockets API provides the data type
 258 .IR "struct sockaddr_storage".
 259 This type
 260 is suitable to accommodate all supported domain-specific socket
 261 address structures; it is large enough and is aligned properly.
 262 (In particular, it is large enough to hold
 263 IPv6 socket addresses.)
 264 The structure includes the following field, which can be used to identify
 265 the type of socket address actually stored in the structure:
 266
 267 .in +4n
 268 .nf
 269     sa_family_t ss_family;
 270 .fi
 271 .in
 272
 273 The
 274 .I sockaddr_storage
 275 structure is useful in programs that must handle socket addresses
 276 in a generic way
 277 (e.g., programs that must deal with both IPv4 and IPv6 socket addresses).
 278 .SS Socket options
 279 The socket options listed below can be set by using
 280 .BR setsockopt (2)
 281 and read with
 282 .BR getsockopt (2)
 283 with the socket level set to
 284 .B SOL_SOCKET
 285 for all sockets.
 286 Unless otherwise noted,
 287 .I optval
 288 is a pointer to an
 289 .IR int .
 290 .\" FIXME .
 291 .\" In the list below, the text used to describe argument types
 292 .\" for each socket option should be more consistent
 293 .\"
 294 .\" SO_ACCEPTCONN is in POSIX.1-2001, and its origin is explained in
 295 .\" W R Stevens, UNPv1
 296 .TP
 297 .B SO_ACCEPTCONN
 298 Returns a value indicating whether or not this socket has been marked
 299 to accept connections with
 300 .BR listen (2).
 301 The value 0 indicates that this is not a listening socket,
 302 the value 1 indicates that this is a listening socket.
 303 This socket option is read-only.
 304 .TP
 305 .B SO_BINDTODEVICE
 306 Bind this socket to a particular device like \(lqeth0\(rq,
 307 as specified in the passed interface name.
 308 If the
 309 name is an empty string or the option length is zero, the socket device
 310 binding is removed.
 311 The passed option is a variable-length null-terminated
 312 interface name string with the maximum size of
 313 .BR IFNAMSIZ .
 314 If a socket is bound to an interface,
 315 only packets received from that particular interface are processed by the
 316 socket.
 317 Note that this works only for some socket types, particularly
 318 .B AF_INET
 319 sockets.
 320 It is not supported for packet sockets (use normal
 321 .BR bind (2)
 322 there).
 323
 324 Before Linux 3.8,
 325 this socket option could be set, but could not retrieved with
 326 .BR getsockopt (2).
 327 Since Linux 3.8, it is readable.
 328 The
 329 .I optlen
 330 argument should contain the buffer size available
 331 to receive the device name and is recommended to be
 332 .BR IFNAMSZ
 333 bytes.
 334 The real device name length is reported back in the
 335 .I optlen
 336 argument.
 337 .TP
 338 .B SO_BROADCAST
 339 Set or get the broadcast flag.
 340 When enabled, datagram sockets are allowed to send
 341 packets to a broadcast address.
 342 This option has no effect on stream-oriented sockets.
 343 .TP
 344 .B SO_BSDCOMPAT
 345 Enable BSD bug-to-bug compatibility.
 346 This is used by the UDP protocol module in Linux 2.0 and 2.2.
 347 If enabled, ICMP errors received for a UDP socket will not be passed
 348 to the user program.
 349 In later kernel versions, support for this option has been phased out:
 350 Linux 2.4 silently ignores it, and Linux 2.6 generates a kernel warning
 351 (printk()) if a program uses this option.
 352 Linux 2.0 also enabled BSD bug-to-bug compatibility
 353 options (random header changing, skipping of the broadcast flag) for raw
 354 sockets with this option, but that was removed in Linux 2.2.
 355 .TP
 356 .B SO_DEBUG
 357 Enable socket debugging.
 358 Only allowed for processes with the
 359 .B CAP_NET_ADMIN
 360 capability or an effective user ID of 0.
 361 .TP
 362 .BR SO_DOMAIN " (since Linux 2.6.32)"
 363 Retrieves the socket domain as an integer, returning a value such as
 364 .BR AF_INET6 .
 365 See
 366 .BR socket (2)
 367 for details.
 368 This socket option is read-only.
 369 .TP
 370 .B SO_ERROR
 371 Get and clear the pending socket error.
 372 This socket option is read-only.
 373 Expects an integer.
 374 .TP
 375 .B SO_DONTROUTE
 376 Don't send via a gateway, send only to directly connected hosts.
 377 The same effect can be achieved by setting the
 378 .B MSG_DONTROUTE
 379 flag on a socket
 380 .BR send (2)
 381 operation.
 382 Expects an integer boolean flag.
 383 .TP
 384 .B SO_KEEPALIVE
 385 Enable sending of keep-alive messages on connection-oriented sockets.
 386 Expects an integer boolean flag.
 387 .TP
 388 .B SO_LINGER
 389 Sets or gets the
 390 .B SO_LINGER
 391 option.
 392 The argument is a
 393 .I linger
 394 structure.
 395 .sp
 396 .in +4n
 397 .nf
 398 struct linger {
 399     int l_onoff;    /* linger active */
 400     int l_linger;   /* how many seconds to linger for */
 401 };
 402 .fi
 403 .in
 404 .IP
 405 When enabled, a
 406 .BR close (2)
 407 or
 408 .BR shutdown (2)
 409 will not return until all queued messages for the socket have been
 410 successfully sent or the linger timeout has been reached.
 411 Otherwise,
 412 the call returns immediately and the closing is done in the background.
 413 When the socket is closed as part of
 414 .BR exit (2),
 415 it always lingers in the background.
 416 .TP
 417 .BR SO_MARK " (since Linux 2.6.25)"
 418 .\" commit 4a19ec5800fc3bb64e2d87c4d9fdd9e636086fe0
 419 .\" and    914a9ab386a288d0f22252fc268ecbc048cdcbd5
 420 Set the mark for each packet sent through this socket
 421 (similar to the netfilter MARK target but socket-based).
 422 Changing the mark can be used for mark-based
 423 routing without netfilter or for packet filtering.
 424 Setting this option requires the
 425 .B CAP_NET_ADMIN
 426 capability.
 427 .TP
 428 .B SO_OOBINLINE
 429 If this option is enabled,
 430 out-of-band data is directly placed into the receive data stream.
 431 Otherwise, out-of-band data is passed only when the
 432 .B MSG_OOB
 433 flag is set during receiving.
 434 .\" don't document it because it can do too much harm.
 435 .\".B SO_NO_CHECK
 436 .\"     The kernel has support for the SO_NO_CHECK socket
 437 .\"     option (boolean: 0 == default, calculate checksum on xmit,
 438 .\"     1 == do not calculate checksum on xmit).
 439 .\" Additional note from Andi Kleen on SO_NO_CHECK (2010-08-30)
 440 .\"     On Linux UDP checksums are essentially free and there's no reason
 441 .\"     to turn them off and it would disable another safety line.
 442 .\"     That is why I didn't document the option.
 443 .TP
 444 .B SO_PASSCRED
 445 Enable or disable the receiving of the
 446 .B SCM_CREDENTIALS
 447 control message.
 448 For more information see
 449 .BR unix (7).
 450 .\" FIXME Document SO_PASSSEC, added in 2.6.18; there is some info
 451 .\" in the 2.6.18 ChangeLog
 452 .TP
 453 .BR SO_PEEK_OFF " (since Linux 3.4)"
 454 .\" commit ef64a54f6e558155b4f149bb10666b9e914b6c54
 455 This option, which is currently supported only for
 456 .BR unix (7)
 457 sockets, sets the value of the "peek offset" for the
 458 .BR recv (2)
 459 system call when used with
 460 .BR MSG_PEEK
 461 flag.
 462
 463 When this option is set to a negative value
 464 (it is set to \-1 for all new sockets),
 465 traditional behavior is provided:
 466 .BR recv (2)
 467 with the
 468 .BR MSG_PEEK
 469 flag will peek data from the front of the queue.
 470
 471 When the option is set to a value greater than or equal to zero,
 472 then the next peek at data queued in the socket will occur at
 473 the byte offset specified by the option value.
 474 At the same time, the "peek offset" will be
 475 incremented by the number of bytes that were peeked from the queue,
 476 so that a subsequent peek will return the next data in the queue.
 477
 478 If data is removed from the front of the queue via a call to
 479 .BR recv (2)
 480 (or similar) without the
 481 .BR MSG_PEEK
 482 flag, the "peek offset" will be decreased by the number of bytes removed.
 483 In other words, receiving data without the
 484 .B MSG_PEEK
 485 flag will cause the "peek offset" to be adjusted to maintain
 486 the correct relative position in the queued data,
 487 so that a subsequent peek will retrieve the data that would have been
 488 retrieved had the data not been removed.
 489
 490 For datagram sockets, if the "peek offset" points to the middle of a packet,
 491 the data returned will be marked with the
 492 .BR MSG_TRUNC
 493 flag.
 494
 495 The following example serves to illustrate the use of
 496 .BR SO_PEEK_OFF .
 497 Suppose a stream socket has the following queued input data:
 498
 499     aabbccddeeff
 500
 501 .IP
 502 The following sequence of
 503 .BR recv (2)
 504 calls would have the effect noted in the comments:
 505
 506 .in +4n
 507 .nf
 508 int ov = 4;                  // Set peek offset to 4
 509 setsockopt(fd, SOL_SOCKET, SO_PEEK_OFF, &ov, sizeof(ov));
 510
 511 recv(fd, buf, 2, MSG_PEEK);  // Peeks "cc"; offset set to 6
 512 recv(fd, buf, 2, MSG_PEEK);  // Peeks "dd"; offset set to 8
 513 recv(fd, buf, 2, 0);         // Reads "aa"; offset set to 6
 514 recv(fd, buf, 2, MSG_PEEK);  // Peeks "ee"; offset set to 8
 515 .fi
 516 .in
 517 .TP
 518 .B SO_PEERCRED
 519 Return the credentials of the foreign process connected to this socket.
 520 This is possible only for connected
 521 .B AF_UNIX
 522 stream sockets and
 523 .B AF_UNIX
 524 stream and datagram socket pairs created using
 525 .BR socketpair (2);
 526 see
 527 .BR unix (7).
 528 The returned credentials are those that were in effect at the time
 529 of the call to
 530 .BR connect (2)
 531 or
 532 .BR socketpair (2).
 533 The argument is a
 534 .I ucred
 535 structure; define the
 536 .B _GNU_SOURCE
 537 feature test macro to obtain the definition of that structure from
 538 .IR <sys/socket.h> .
 539 This socket option is read-only.
 540 .TP
 541 .B SO_PRIORITY
 542 Set the protocol-defined priority for all packets to be sent on
 543 this socket.
 544 Linux uses this value to order the networking queues:
 545 packets with a higher priority may be processed first depending
 546 on the selected device queueing discipline.
 547 .\" For
 548 .\" .BR ip (7),
 549 .\" this also sets the IP type-of-service (TOS) field for outgoing packets.
 550 Setting a priority outside the range 0 to 6 requires the
 551 .B CAP_NET_ADMIN
 552 capability.
 553 .TP
 554 .BR SO_PROTOCOL " (since Linux 2.6.32)"
 555 Retrieves the socket protocol as an integer, returning a value such as
 556 .BR IPPROTO_SCTP .
 557 See
 558 .BR socket (2)
 559 for details.
 560 This socket option is read-only.
 561 .TP
 562 .B SO_RCVBUF
 563 Sets or gets the maximum socket receive buffer in bytes.
 564 The kernel doubles this value (to allow space for bookkeeping overhead)
 565 when it is set using
 566 .\" Most (all?) other implementations do not do this -- MTK, Dec 05
 567 .BR setsockopt (2),
 568 and this doubled value is returned by
 569 .BR getsockopt (2).
 570 .\" The following thread on LMKL is quite informative:
 571 .\" getsockopt/setsockopt with SO_RCVBUF and SO_SNDBUF "non-standard" behavior
 572 .\" 17 July 2012
 573 .\" http://thread.gmane.org/gmane.linux.kernel/1328935
 574 The default value is set by the
 575 .I /proc/sys/net/core/rmem_default
 576 file, and the maximum allowed value is set by the
 577 .I /proc/sys/net/core/rmem_max
 578 file.
 579 The minimum (doubled) value for this option is 256.
 580 .TP
 581 .BR SO_RCVBUFFORCE " (since Linux 2.6.14)"
 582 Using this socket option, a privileged
 583 .RB ( CAP_NET_ADMIN )
 584 process can perform the same task as
 585 .BR SO_RCVBUF ,
 586 but the
 587 .I rmem_max
 588 limit can be overridden.
 589 .TP
 590 .BR SO_RCVLOWAT " and " SO_SNDLOWAT
 591 Specify the minimum number of bytes in the buffer until the socket layer
 592 will pass the data to the protocol
 593 .RB ( SO_SNDLOWAT )
 594 or the user on receiving
 595 .RB ( SO_RCVLOWAT ).
 596 These two values are initialized to 1.
 597 .B SO_SNDLOWAT
 598 is not changeable on Linux
 599 .RB ( setsockopt (2)
 600 fails with the error
 601 .BR ENOPROTOOPT ).
 602 .B SO_RCVLOWAT
 603 is changeable
 604 only since Linux 2.4.
 605 The
 606 .BR select (2)
 607 and
 608 .BR poll (2)
 609 system calls currently do not respect the
 610 .B SO_RCVLOWAT
 611 setting on Linux,
 612 and mark a socket readable when even a single byte of data is available.
 613 A subsequent read from the socket will block until
 614 .B SO_RCVLOWAT
 615 bytes are available.
 616 .\" See http://marc.theaimsgroup.com/?l=linux-kernel&m=111049368106984&w=2
 617 .\" Tested on kernel 2.6.14 -- mtk, 30 Nov 05
 618 .TP
 619 .BR SO_RCVTIMEO " and " SO_SNDTIMEO
 620 .\" Not implemented in 2.0.
 621 .\" Implemented in 2.1.11 for getsockopt: always return a zero struct.
 622 .\" Implemented in 2.3.41 for setsockopt, and actually used.
 623 Specify the receiving or sending timeouts until reporting an error.
 624 The argument is a
 625 .IR "struct timeval" .
 626 If an input or output function blocks for this period of time, and
 627 data has been sent or received, the return value of that function
 628 will be the amount of data transferred; if no data has been transferred
 629 and the timeout has been reached, then \-1 is returned with
 630 .I errno
 631 set to
 632 .BR EAGAIN
 633 or
 634 .BR EWOULDBLOCK ,
 635 .\" in fact to EAGAIN
 636 or
 637 .B EINPROGRESS
 638 (for
 639 .BR connect (2))
 640 just as if the socket was specified to be nonblocking.
 641 If the timeout is set to zero (the default),
 642 then the operation will never timeout.
 643 Timeouts only have effect for system calls that perform socket I/O (e.g.,
 644 .BR read (2),
 645 .BR recvmsg (2),
 646 .BR send (2),
 647 .BR sendmsg (2));
 648 timeouts have no effect for
 649 .BR select (2),
 650 .BR poll (2),
 651 .BR epoll_wait (2),
 652 and so on.
 653 .TP
 654 .B SO_REUSEADDR
 655 .\"     commit c617f398edd4db2b8567a28e899a88f8f574798d
 656 .\"     https://lwn.net/Articles/542629/
 657 Indicates that the rules used in validating addresses supplied in a
 658 .BR bind (2)
 659 call should allow reuse of local addresses.
 660 For
 661 .B AF_INET
 662 sockets this
 663 means that a socket may bind, except when there
 664 is an active listening socket bound to the address.
 665 When the listening socket is bound to
 666 .B INADDR_ANY
 667 with a specific port then it is not possible
 668 to bind to this port for any local address.
 669 Argument is an integer boolean flag.
 670 .TP
 671 .BR SO_REUSEPORT " (since Linux 3.9)"
 672 Permits multiple
 673 .B AF_INET
 674 or
 675 .B AF_INET6
 676 sockets to be bound to an identical socket address.
 677 This option must be set on each socket (including the first socket)
 678 prior to calling
 679 .BR bind (2)
 680 on the socket.
 681 To prevent port hijacking,
 682 all of the processes binding to the same address must have the same
 683 effective UID.
 684 This option can be employed with both TCP and UDP sockets.
 685
 686 For TCP sockets, this option allows
 687 .BR accept (2)
 688 load distribution in a multi-threaded server to be improved by
 689 using a distinct listener socket for each thread.
 690 This provides improved load distribution as compared
 691 to traditional techniques such using a single
 692 .BR accept (2)ing
 693 thread that distributes connections,
 694 or having multiple threads that compete to
 695 .BR accept (2)
 696 from the same socket.
 697
 698 For UDP sockets,
 699 the use of this option can provide better distribution
 700 of incoming datagrams to multiple processes (or threads) as compared
 701 to the traditional technique of having multiple processes
 702 compete to receive datagrams on the same socket.
 703 .TP
 704 .BR SO_RXQ_OVFL " (since Linux 2.6.33)"
 705 .\" commit 3b885787ea4112eaa80945999ea0901bf742707f
 706 Indicates that an unsigned 32-bit value ancillary message (cmsg)
 707 should be attached to received skbs indicating
 708 the number of packets dropped by the socket between
 709 the last received packet and this received packet.
 710 .TP
 711 .B SO_SNDBUF
 712 Sets or gets the maximum socket send buffer in bytes.
 713 The kernel doubles this value (to allow space for bookkeeping overhead)
 714 when it is set using
 715 .\" Most (all?) other implementations do not do this -- MTK, Dec 05
 716 .\" See also the comment to SO_RCVBUF (17 Jul 2012 LKML mail)
 717 .BR setsockopt (2),
 718 and this doubled value is returned by
 719 .BR getsockopt (2).
 720 The default value is set by the
 721 .I /proc/sys/net/core/wmem_default
 722 file and the maximum allowed value is set by the
 723 .I /proc/sys/net/core/wmem_max
 724 file.
 725 The minimum (doubled) value for this option is 2048.
 726 .TP
 727 .BR SO_SNDBUFFORCE " (since Linux 2.6.14)"
 728 Using this socket option, a privileged
 729 .RB ( CAP_NET_ADMIN )
 730 process can perform the same task as
 731 .BR SO_SNDBUF ,
 732 but the
 733 .I wmem_max
 734 limit can be overridden.
 735 .TP
 736 .B SO_TIMESTAMP
 737 Enable or disable the receiving of the
 738 .B SO_TIMESTAMP
 739 control message.
 740 The timestamp control message is sent with level
 741 .B SOL_SOCKET
 742 and the
 743 .I cmsg_data
 744 field is a
 745 .I "struct timeval"
 746 indicating the
 747 reception time of the last packet passed to the user in this call.
 748 See
 749 .BR cmsg (3)
 750 for details on control messages.
 751 .TP
 752 .B SO_TYPE
 753 Gets the socket type as an integer (e.g.,
 754 .BR SOCK_STREAM ).
 755 This socket option is read-only.
 756 .TP
 757 .BR SO_BUSY_POLL " (since Linux 3.11)"
 758 Sets the approximate time in microseconds to busy poll on a blocking receive
 759 when there is no data.
 760 Increasing this value requires
 761 .BR CAP_NET_ADMIN .
 762 The default for this option is controlled by the
 763 .I /proc/sys/net/core/busy_read
 764 file.
 765
 766 The value in the
 767 .I /proc/sys/net/core/busy_poll
 768 file determines how long
 769 .BR select (2)
 770 and
 771 .BR poll (2)
 772 will busy poll when they operate on sockets with
 773 .BR SO_BUSY_POLL
 774 set and no events to report are found.
 775
 776 In both cases,
 777 busy polling will only be done when the socket last received data
 778 from a network device that supports this option.
 779
 780 While busy polling may improve latency of some applications,
 781 care must be taken when using it since this will increase
 782 both CPU utilization and power usage.
 783 .SS Signals
 784 When writing onto a connection-oriented socket that has been shut down
 785 (by the local or the remote end)
 786 .B SIGPIPE
 787 is sent to the writing process and
 788 .B EPIPE
 789 is returned.
 790 The signal is not sent when the write call
 791 specified the
 792 .B MSG_NOSIGNAL
 793 flag.
 794 .PP
 795 When requested with the
 796 .B FIOSETOWN
 797 .BR fcntl (2)
 798 or
 799 .B SIOCSPGRP
 800 .BR ioctl (2),
 801 .B SIGIO
 802 is sent when an I/O event occurs.
 803 It is possible to use
 804 .BR poll (2)
 805 or
 806 .BR select (2)
 807 in the signal handler to find out which socket the event occurred on.
 808 An alternative (in Linux 2.2) is to set a real-time signal using the
 809 .B F_SETSIG
 810 .BR fcntl (2);
 811 the handler of the real time signal will be called with
 812 the file descriptor in the
 813 .I si_fd
 814 field of its
 815 .IR siginfo_t .
 816 See
 817 .BR fcntl (2)
 818 for more information.
 819 .PP
 820 Under some circumstances (e.g., multiple processes accessing a
 821 single socket), the condition that caused the
 822 .B SIGIO
 823 may have already disappeared when the process reacts to the signal.
 824 If this happens, the process should wait again because Linux
 825 will resend the signal later.
 826 .\" .SS Ancillary messages
 827 .SS /proc interfaces
 828 The core socket networking parameters can be accessed
 829 via files in the directory
 830 .IR /proc/sys/net/core/ .
 831 .TP
 832 .I rmem_default
 833 contains the default setting in bytes of the socket receive buffer.
 834 .TP
 835 .I rmem_max
 836 contains the maximum socket receive buffer size in bytes which a user may
 837 set by using the
 838 .B SO_RCVBUF
 839 socket option.
 840 .TP
 841 .I wmem_default
 842 contains the default setting in bytes of the socket send buffer.
 843 .TP
 844 .I wmem_max
 845 contains the maximum socket send buffer size in bytes which a user may
 846 set by using the
 847 .B SO_SNDBUF
 848 socket option.
 849 .TP
 850 .IR message_cost " and " message_burst
 851 configure the token bucket filter used to load limit warning messages
 852 caused by external network events.
 853 .TP
 854 .I netdev_max_backlog
 855 Maximum number of packets in the global input queue.
 856 .TP
 857 .I optmem_max
 858 Maximum length of ancillary data and user control data like the iovecs
 859 per socket.
 860 .\" netdev_fastroute is not documented because it is experimental
 861 .SS Ioctls
 862 These operations can be accessed using
 863 .BR ioctl (2):
 864
 865 .in +4n
 866 .nf
 867 .IB error " = ioctl(" ip_socket ", " ioctl_type ", " &value_result ");"
 868 .fi
 869 .in
 870 .TP
 871 .B SIOCGSTAMP
 872 Return a
 873 .I struct timeval
 874 with the receive timestamp of the last packet passed to the user.
 875 This is useful for accurate round trip time measurements.
 876 See
 877 .BR setitimer (2)
 878 for a description of
 879 .IR "struct timeval" .
 880 .\"
 881 This ioctl should be used only if the socket option
 882 .B SO_TIMESTAMP
 883 is not set on the socket.
 884 Otherwise, it returns the timestamp of the
 885 last packet that was received while
 886 .B SO_TIMESTAMP
 887 was not set, or it fails if no such packet has been received,
 888 (i.e.,
 889 .BR ioctl (2)
 890 returns \-1 with
 891 .I errno
 892 set to
 893 .BR ENOENT ).
 894 .TP
 895 .B SIOCSPGRP
 896 Set the process or process group to send
 897 .B SIGIO
 898 or
 899 .B SIGURG
 900 signals
 901 to when an
 902 asynchronous I/O operation has finished or urgent data is available.
 903 The argument is a pointer to a
 904 .IR pid_t .
 905 If the argument is positive, send the signals to that process.
 906 If the
 907 argument is negative, send the signals to the process group with the ID
 908 of the absolute value of the argument.
 909 The process may only choose itself or its own process group to receive
 910 signals unless it has the
 911 .B CAP_KILL
 912 capability or an effective UID of 0.
 913 .TP
 914 .B FIOASYNC
 915 Change the
 916 .B O_ASYNC
 917 flag to enable or disable asynchronous I/O mode of the socket.
 918 Asynchronous I/O mode means that the
 919 .B SIGIO
 920 signal or the signal set with
 921 .B F_SETSIG
 922 is raised when a new I/O event occurs.
 923 .IP
 924 Argument is an integer boolean flag.
 925 (This operation is synonymous with the use of
 926 .BR fcntl (2)
 927 to set the
 928 .B O_ASYNC
 929 flag.)
 930 .\"
 931 .TP
 932 .B SIOCGPGRP
 933 Get the current process or process group that receives
 934 .B SIGIO
 935 or
 936 .B SIGURG
 937 signals,
 938 or 0
 939 when none is set.
 940 .PP
 941 Valid
 942 .BR fcntl (2)
 943 operations:
 944 .TP
 945 .B FIOGETOWN
 946 The same as the
 947 .B SIOCGPGRP
 948 .BR ioctl (2).
 949 .TP
 950 .B FIOSETOWN
 951 The same as the
 952 .B SIOCSPGRP
 953 .BR ioctl (2).
 954 .SH VERSIONS
 955 .B SO_BINDTODEVICE
 956 was introduced in Linux 2.0.30.
 957 .B SO_PASSCRED
 958 is new in Linux 2.2.
 959 The
 960 .I /proc
 961 interfaces were introduced in Linux 2.2.
 962 .B SO_RCVTIMEO
 963 and
 964 .B SO_SNDTIMEO
 965 are supported since Linux 2.3.41.
 966 Earlier, timeouts were fixed to
 967 a protocol-specific setting, and could not be read or written.
 968 .SH NOTES
 969 Linux assumes that half of the send/receive buffer is used for internal
 970 kernel structures; thus the values in the corresponding
 971 .I /proc
 972 files are twice what can be observed on the wire.
 973
 974 Linux will allow port reuse only with the
 975 .B SO_REUSEADDR
 976 option
 977 when this option was set both in the previous program that performed a
 978 .BR bind (2)
 979 to the port and in the program that wants to reuse the port.
 980 This differs from some implementations (e.g., FreeBSD)
 981 where only the later program needs to set the
 982 .B SO_REUSEADDR
 983 option.
 984 Typically this difference is invisible, since, for example, a server
 985 program is designed to always set this option.
 986 .SH BUGS
 987 The
 988 .B CONFIG_FILTER
 989 socket options
 990 .B SO_ATTACH_FILTER
 991 and
 992 .B SO_DETACH_FILTER
 993 .\" FIXME Document SO_ATTACH_FILTER and SO_DETACH_FILTER
 994 are not documented.
 995 The suggested interface to use them is via the libpcap
 996 library.
 997 .\" .SH AUTHORS
 998 .\" This man page was written by Andi Kleen.
 999 .SH SEE ALSO
1000 .BR connect (2),
1001 .BR getsockopt (2),
1002 .BR setsockopt (2),
1003 .BR socket (2),
1004 .BR capabilities (7),
1005 .BR ddp (7),
1006 .BR ip (7),
1007 .BR packet (7),
1008 .BR tcp (7),
1009 .BR udp (7),
1010 .BR unix (7)