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 .\"
  21 .\" FIXME
  22 .\" The following are not yet documented:
  23 .\"
  24 .\" SO_PEERNAME (2.4?)
  25 .\"     get only
  26 .\"     Seems to do something similar to getpeername(), but then
  27 .\"     why is it necessary / how does it differ?
  28 .\"
  29 .\" SO_TIMESTAMPNS (2.6.22)
  30 .\"     Documentation/networking/timestamping.txt
  31 .\"     commit 92f37fd2ee805aa77925c1e64fd56088b46094fc
  32 .\"     Author: Eric Dumazet <dada1@cosmosbay.com>
  33 .\"
  34 .\" SO_TIMESTAMPING (2.6.30)
  35 .\"     Documentation/networking/timestamping.txt
  36 .\"     commit cb9eff097831007afb30d64373f29d99825d0068
  37 .\"     Author: Patrick Ohly <patrick.ohly@intel.com>
  38 .\"
  39 .\"  SO_WIFI_STATUS (3.3)
  40 .\"     commit 6e3e939f3b1bf8534b32ad09ff199d88800835a0
  41 .\"     Author: Johannes Berg <johannes.berg@intel.com>
  42 .\"     Also: SCM_WIFI_STATUS
  43 .\"
  44 .\" SO_NOFCS (3.4)
  45 .\"     commit 3bdc0eba0b8b47797f4a76e377dd8360f317450f
  46 .\"     Author: Ben Greear <greearb@candelatech.com>
  47 .\"
  48 .\"  SO_GET_FILTER (3.8)
  49 .\"     commit a8fc92778080c845eaadc369a0ecf5699a03bef0
  50 .\"     Author: Pavel Emelyanov <xemul@parallels.com>
  51 .\"
  52 .\" SO_MAX_PACING_RATE (3.13)
  53 .\"     commit 62748f32d501f5d3712a7c372bbb92abc7c62bc7
  54 .\"     Author: Eric Dumazet <edumazet@google.com>
  55 .\"
  56 .\" SO_BPF_EXTENSIONS (3.14)
  57 .\"     commit ea02f9411d9faa3553ed09ce0ec9f00ceae9885e
  58 .\"     Author: Michal Sekletar <msekleta@redhat.com>
  59 .\"
  60 .TH SOCKET 7 2019-08-02 Linux "Linux Programmer's Manual"
  61 .SH NAME
  62 socket \- Linux socket interface
  63 .SH SYNOPSIS
  64 .B #include <sys/socket.h>
  65 .PP
  66 .IB sockfd " = socket(int " socket_family ", int " socket_type ", int " protocol );
  67 .SH DESCRIPTION
  68 This manual page describes the Linux networking socket layer user
  69 interface.
  70 The BSD compatible sockets
  71 are the uniform interface
  72 between the user process and the network protocol stacks in the kernel.
  73 The protocol modules are grouped into
  74 .I protocol families
  75 such as
  76 .BR AF_INET ", " AF_IPX ", and " AF_PACKET ,
  77 and
  78 .I socket types
  79 such as
  80 .B SOCK_STREAM
  81 or
  82 .BR SOCK_DGRAM .
  83 See
  84 .BR socket (2)
  85 for more information on families and types.
  86 .SS Socket-layer functions
  87 These functions are used by the user process to send or receive packets
  88 and to do other socket operations.
  89 For more information see their respective manual pages.
  90 .PP
  91 .BR socket (2)
  92 creates a socket,
  93 .BR connect (2)
  94 connects a socket to a remote socket address,
  95 the
  96 .BR bind (2)
  97 function binds a socket to a local socket address,
  98 .BR listen (2)
  99 tells the socket that new connections shall be accepted, and
 100 .BR accept (2)
 101 is used to get a new socket with a new incoming connection.
 102 .BR socketpair (2)
 103 returns two connected anonymous sockets (implemented only for a few
 104 local families like
 105 .BR AF_UNIX )
 106 .PP
 107 .BR send (2),
 108 .BR sendto (2),
 109 and
 110 .BR sendmsg (2)
 111 send data over a socket, and
 112 .BR recv (2),
 113 .BR recvfrom (2),
 114 .BR recvmsg (2)
 115 receive data from a socket.
 116 .BR poll (2)
 117 and
 118 .BR select (2)
 119 wait for arriving data or a readiness to send data.
 120 In addition, the standard I/O operations like
 121 .BR write (2),
 122 .BR writev (2),
 123 .BR sendfile (2),
 124 .BR read (2),
 125 and
 126 .BR readv (2)
 127 can be used to read and write data.
 128 .PP
 129 .BR getsockname (2)
 130 returns the local socket address and
 131 .BR getpeername (2)
 132 returns the remote socket address.
 133 .BR getsockopt (2)
 134 and
 135 .BR setsockopt (2)
 136 are used to set or get socket layer or protocol options.
 137 .BR ioctl (2)
 138 can be used to set or read some other options.
 139 .PP
 140 .BR close (2)
 141 is used to close a socket.
 142 .BR shutdown (2)
 143 closes parts of a full-duplex socket connection.
 144 .PP
 145 Seeking, or calling
 146 .BR pread (2)
 147 or
 148 .BR pwrite (2)
 149 with a nonzero position is not supported on sockets.
 150 .PP
 151 It is possible to do nonblocking I/O on sockets by setting the
 152 .B O_NONBLOCK
 153 flag on a socket file descriptor using
 154 .BR fcntl (2).
 155 Then all operations that would block will (usually)
 156 return with
 157 .B EAGAIN
 158 (operation should be retried later);
 159 .BR connect (2)
 160 will return
 161 .B EINPROGRESS
 162 error.
 163 The user can then wait for various events via
 164 .BR poll (2)
 165 or
 166 .BR select (2).
 167 .TS
 168 tab(:) allbox;
 169 c s s
 170 l l l.
 171 I/O events
 172 Event:Poll flag:Occurrence
 173 Read:POLLIN:T{
 174 New data arrived.
 175 T}
 176 Read:POLLIN:T{
 177 A connection setup has been completed
 178 (for connection-oriented sockets)
 179 T}
 180 Read:POLLHUP:T{
 181 A disconnection request has been initiated by the other end.
 182 T}
 183 Read:POLLHUP:T{
 184 A connection is broken (only for connection-oriented protocols).
 185 When the socket is written
 186 .B SIGPIPE
 187 is also sent.
 188 T}
 189 Write:POLLOUT:T{
 190 Socket has enough send buffer space for writing new data.
 191 T}
 192 Read/Write:T{
 193 POLLIN |
 194 .br
 195 POLLOUT
 196 T}:T{
 197 An outgoing
 198 .BR connect (2)
 199 finished.
 200 T}
 201 Read/Write:POLLERR:An asynchronous error occurred.
 202 Read/Write:POLLHUP:The other end has shut down one direction.
 203 Exception:POLLPRI:T{
 204 Urgent data arrived.
 205 .B SIGURG
 206 is sent then.
 207 T}
 208 .\" FIXME . The following is not true currently:
 209 .\" It is no I/O event when the connection
 210 .\" is broken from the local end using
 211 .\" .BR shutdown (2)
 212 .\" or
 213 .\" .BR close (2).
 214 .TE
 215 .PP
 216 An alternative to
 217 .BR poll (2)
 218 and
 219 .BR select (2)
 220 is to let the kernel inform the application about events
 221 via a
 222 .B SIGIO
 223 signal.
 224 For that the
 225 .B O_ASYNC
 226 flag must be set on a socket file descriptor via
 227 .BR fcntl (2)
 228 and a valid signal handler for
 229 .B SIGIO
 230 must be installed via
 231 .BR sigaction (2).
 232 See the
 233 .I Signals
 234 discussion below.
 235 .SS Socket address structures
 236 Each socket domain has its own format for socket addresses,
 237 with a domain-specific address structure.
 238 Each of these structures begins with an
 239 integer "family" field (typed as
 240 .IR sa_family_t )
 241 that indicates the type of the address structure.
 242 This allows
 243 the various system calls (e.g.,
 244 .BR connect (2),
 245 .BR bind (2),
 246 .BR accept (2),
 247 .BR getsockname (2),
 248 .BR getpeername (2)),
 249 which are generic to all socket domains,
 250 to determine the domain of a particular socket address.
 251 .PP
 252 To allow any type of socket address to be passed to
 253 interfaces in the sockets API,
 254 the type
 255 .IR "struct sockaddr"
 256 is defined.
 257 The purpose of this type is purely to allow casting of
 258 domain-specific socket address types to a "generic" type,
 259 so as to avoid compiler warnings about type mismatches in
 260 calls to the sockets API.
 261 .PP
 262 In addition, the sockets API provides the data type
 263 .IR "struct sockaddr_storage".
 264 This type
 265 is suitable to accommodate all supported domain-specific socket
 266 address structures; it is large enough and is aligned properly.
 267 (In particular, it is large enough to hold
 268 IPv6 socket addresses.)
 269 The structure includes the following field, which can be used to identify
 270 the type of socket address actually stored in the structure:
 271 .PP
 272 .in +4n
 273 .EX
 274     sa_family_t ss_family;
 275 .EE
 276 .in
 277 .PP
 278 The
 279 .I sockaddr_storage
 280 structure is useful in programs that must handle socket addresses
 281 in a generic way
 282 (e.g., programs that must deal with both IPv4 and IPv6 socket addresses).
 283 .SS Socket options
 284 The socket options listed below can be set by using
 285 .BR setsockopt (2)
 286 and read with
 287 .BR getsockopt (2)
 288 with the socket level set to
 289 .B SOL_SOCKET
 290 for all sockets.
 291 Unless otherwise noted,
 292 .I optval
 293 is a pointer to an
 294 .IR int .
 295 .\" FIXME .
 296 .\" In the list below, the text used to describe argument types
 297 .\" for each socket option should be more consistent
 298 .\"
 299 .\" SO_ACCEPTCONN is in POSIX.1-2001, and its origin is explained in
 300 .\" W R Stevens, UNPv1
 301 .TP
 302 .B SO_ACCEPTCONN
 303 Returns a value indicating whether or not this socket has been marked
 304 to accept connections with
 305 .BR listen (2).
 306 The value 0 indicates that this is not a listening socket,
 307 the value 1 indicates that this is a listening socket.
 308 This socket option is read-only.
 309 .TP
 310 .BR SO_ATTACH_FILTER " (since Linux 2.2), " SO_ATTACH_BPF " (since Linux 3.19)"
 311 Attach a classic BPF
 312 .RB ( SO_ATTACH_FILTER )
 313 or an extended BPF
 314 .RB ( SO_ATTACH_BPF )
 315 program to the socket for use as a filter of incoming packets.
 316 A packet will be dropped if the filter program returns zero.
 317 If the filter program returns a
 318 nonzero value which is less than the packet's data length,
 319 the packet will be truncated to the length returned.
 320 If the value returned by the filter is greater than or equal to the
 321 packet's data length, the packet is allowed to proceed unmodified.
 322 .IP
 323 The argument for
 324 .BR SO_ATTACH_FILTER
 325 is a
 326 .I sock_fprog
 327 structure, defined in
 328 .IR <linux/filter.h> :
 329 .IP
 330 .in +4n
 331 .EX
 332 struct sock_fprog {
 333     unsigned short      len;
 334     struct sock_filter *filter;
 335 };
 336 .EE
 337 .in
 338 .IP
 339 The argument for
 340 .BR SO_ATTACH_BPF
 341 is a file descriptor returned by the
 342 .BR bpf (2)
 343 system call and must refer to a program of type
 344 .BR BPF_PROG_TYPE_SOCKET_FILTER .
 345 .IP
 346 These options may be set multiple times for a given socket,
 347 each time replacing the previous filter program.
 348 The classic and extended versions may be called on the same socket,
 349 but the previous filter will always be replaced such that a socket
 350 never has more than one filter defined.
 351 .IP
 352 Both classic and extended BPF are explained in the kernel source file
 353 .I Documentation/networking/filter.txt
 354 .TP
 355 .BR SO_ATTACH_REUSEPORT_CBPF ", " SO_ATTACH_REUSEPORT_EBPF
 356 For use with the
 357 .BR SO_REUSEPORT
 358 option, these options allow the user to set a classic BPF
 359 .RB ( SO_ATTACH_REUSEPORT_CBPF )
 360 or an extended BPF
 361 .RB ( SO_ATTACH_REUSEPORT_EBPF )
 362 program which defines how packets are assigned to
 363 the sockets in the reuseport group (that is, all sockets which have
 364 .BR SO_REUSEPORT
 365 set and are using the same local address to receive packets).
 366 .IP
 367 The BPF program must return an index between 0 and N\-1 representing
 368 the socket which should receive the packet
 369 (where N is the number of sockets in the group).
 370 If the BPF program returns an invalid index,
 371 socket selection will fall back to the plain
 372 .BR SO_REUSEPORT
 373 mechanism.
 374 .IP
 375 Sockets are numbered in the order in which they are added to the group
 376 (that is, the order of
 377 .BR bind (2)
 378 calls for UDP sockets or the order of
 379 .BR listen (2)
 380 calls for TCP sockets).
 381 New sockets added to a reuseport group will inherit the BPF program.
 382 When a socket is removed from a reuseport group (via
 383 .BR close (2)),
 384 the last socket in the group will be moved into the closed socket's
 385 position.
 386 .IP
 387 These options may be set repeatedly at any time on any socket in the group
 388 to replace the current BPF program used by all sockets in the group.
 389 .IP
 390 .BR SO_ATTACH_REUSEPORT_CBPF
 391 takes the same argument type as
 392 .BR SO_ATTACH_FILTER
 393 and
 394 .BR SO_ATTACH_REUSEPORT_EBPF
 395 takes the same argument type as
 396 .BR SO_ATTACH_BPF .
 397 .IP
 398 UDP support for this feature is available since Linux 4.5;
 399 TCP support is available since Linux 4.6.
 400 .TP
 401 .B SO_BINDTODEVICE
 402 Bind this socket to a particular device like \(lqeth0\(rq,
 403 as specified in the passed interface name.
 404 If the
 405 name is an empty string or the option length is zero, the socket device
 406 binding is removed.
 407 The passed option is a variable-length null-terminated
 408 interface name string with the maximum size of
 409 .BR IFNAMSIZ .
 410 If a socket is bound to an interface,
 411 only packets received from that particular interface are processed by the
 412 socket.
 413 Note that this works only for some socket types, particularly
 414 .B AF_INET
 415 sockets.
 416 It is not supported for packet sockets (use normal
 417 .BR bind (2)
 418 there).
 419 .IP
 420 Before Linux 3.8,
 421 this socket option could be set, but could not retrieved with
 422 .BR getsockopt (2).
 423 Since Linux 3.8, it is readable.
 424 The
 425 .I optlen
 426 argument should contain the buffer size available
 427 to receive the device name and is recommended to be
 428 .BR IFNAMSIZ
 429 bytes.
 430 The real device name length is reported back in the
 431 .I optlen
 432 argument.
 433 .TP
 434 .B SO_BROADCAST
 435 Set or get the broadcast flag.
 436 When enabled, datagram sockets are allowed to send
 437 packets to a broadcast address.
 438 This option has no effect on stream-oriented sockets.
 439 .TP
 440 .B SO_BSDCOMPAT
 441 Enable BSD bug-to-bug compatibility.
 442 This is used by the UDP protocol module in Linux 2.0 and 2.2.
 443 If enabled, ICMP errors received for a UDP socket will not be passed
 444 to the user program.
 445 In later kernel versions, support for this option has been phased out:
 446 Linux 2.4 silently ignores it, and Linux 2.6 generates a kernel warning
 447 (printk()) if a program uses this option.
 448 Linux 2.0 also enabled BSD bug-to-bug compatibility
 449 options (random header changing, skipping of the broadcast flag) for raw
 450 sockets with this option, but that was removed in Linux 2.2.
 451 .TP
 452 .B SO_DEBUG
 453 Enable socket debugging.
 454 Allowed only for processes with the
 455 .B CAP_NET_ADMIN
 456 capability or an effective user ID of 0.
 457 .TP
 458 .BR SO_DETACH_FILTER " (since Linux 2.2), " SO_DETACH_BPF " (since Linux 3.19)"
 459 These two options, which are synonyms,
 460 may be used to remove the classic or extended BPF
 461 program attached to a socket with either
 462 .BR SO_ATTACH_FILTER
 463 or
 464 .BR SO_ATTACH_BPF .
 465 The option value is ignored.
 466 .TP
 467 .BR SO_DOMAIN " (since Linux 2.6.32)"
 468 Retrieves the socket domain as an integer, returning a value such as
 469 .BR AF_INET6 .
 470 See
 471 .BR socket (2)
 472 for details.
 473 This socket option is read-only.
 474 .TP
 475 .B SO_ERROR
 476 Get and clear the pending socket error.
 477 This socket option is read-only.
 478 Expects an integer.
 479 .TP
 480 .B SO_DONTROUTE
 481 Don't send via a gateway, send only to directly connected hosts.
 482 The same effect can be achieved by setting the
 483 .B MSG_DONTROUTE
 484 flag on a socket
 485 .BR send (2)
 486 operation.
 487 Expects an integer boolean flag.
 488 .TP
 489 .BR SO_INCOMING_CPU " (gettable since Linux 3.19, settable since Linux 4.4)"
 490 .\" getsockopt 2c8c56e15df3d4c2af3d656e44feb18789f75837
 491 .\" setsockopt 70da268b569d32a9fddeea85dc18043de9d89f89
 492 Sets or gets the CPU affinity of a socket.
 493 Expects an integer flag.
 494 .IP
 495 .in +4n
 496 .EX
 497 int cpu = 1;
 498 setsockopt(fd, SOL_SOCKET, SO_INCOMING_CPU, &cpu, sizeof(cpu));
 499 .EE
 500 .in
 501 .IP
 502 Because all of the packets for a single stream
 503 (i.e., all packets for the same 4-tuple)
 504 arrive on the single RX queue that is associated with a particular CPU,
 505 the typical use case is to employ one listening process per RX queue,
 506 with the incoming flow being handled by a listener
 507 on the same CPU that is handling the RX queue.
 508 This provides optimal NUMA behavior and keeps CPU caches hot.
 509 .\"
 510 .\" From an email conversation with Eric Dumazet:
 511 .\" >> Note that setting the option is not supported if SO_REUSEPORT is used.
 512 .\" >
 513 .\" > Please define "not supported". Does this yield an API diagnostic?
 514 .\" > If so, what is it?
 515 .\" >
 516 .\" >> Socket will be selected from an array, either by a hash or BPF program
 517 .\" >> that has no access to this information.
 518 .\" >
 519 .\" > Sorry -- I'm lost here. How does this comment relate to the proposed
 520 .\" > man page text above?
 521 .\"
 522 .\" Simply that :
 523 .\"
 524 .\" If an application uses both SO_INCOMING_CPU and SO_REUSEPORT, then
 525 .\" SO_REUSEPORT logic, selecting the socket to receive the packet, ignores
 526 .\" SO_INCOMING_CPU setting.
 527 .TP
 528 .B SO_KEEPALIVE
 529 Enable sending of keep-alive messages on connection-oriented sockets.
 530 Expects an integer boolean flag.
 531 .TP
 532 .B SO_LINGER
 533 Sets or gets the
 534 .B SO_LINGER
 535 option.
 536 The argument is a
 537 .I linger
 538 structure.
 539 .IP
 540 .in +4n
 541 .EX
 542 struct linger {
 543     int l_onoff;    /* linger active */
 544     int l_linger;   /* how many seconds to linger for */
 545 };
 546 .EE
 547 .in
 548 .IP
 549 When enabled, a
 550 .BR close (2)
 551 or
 552 .BR shutdown (2)
 553 will not return until all queued messages for the socket have been
 554 successfully sent or the linger timeout has been reached.
 555 Otherwise,
 556 the call returns immediately and the closing is done in the background.
 557 When the socket is closed as part of
 558 .BR exit (2),
 559 it always lingers in the background.
 560 .TP
 561 .B SO_LOCK_FILTER
 562 .\" commit d59577b6ffd313d0ab3be39cb1ab47e29bdc9182
 563 When set, this option will prevent
 564 changing the filters associated with the socket.
 565 These filters include any set using the socket options
 566 .BR SO_ATTACH_FILTER ,
 567 .BR SO_ATTACH_BPF ,
 568 .BR SO_ATTACH_REUSEPORT_CBPF ,
 569 and
 570 .BR SO_ATTACH_REUSEPORT_EBPF .
 571 .IP
 572 The typical use case is for a privileged process to set up a raw socket
 573 (an operation that requires the
 574 .BR CAP_NET_RAW
 575 capability), apply a restrictive filter, set the
 576 .BR SO_LOCK_FILTER
 577 option,
 578 and then either drop its privileges or pass the socket file descriptor
 579 to an unprivileged process via a UNIX domain socket.
 580 .IP
 581 Once the
 582 .BR SO_LOCK_FILTER
 583 option has been enabled, attempts to change or remove the filter
 584 attached to a socket, or to disable the
 585 .BR SO_LOCK_FILTER
 586 option will fail with the error
 587 .BR EPERM .
 588 .TP
 589 .BR SO_MARK " (since Linux 2.6.25)"
 590 .\" commit 4a19ec5800fc3bb64e2d87c4d9fdd9e636086fe0
 591 .\" and    914a9ab386a288d0f22252fc268ecbc048cdcbd5
 592 Set the mark for each packet sent through this socket
 593 (similar to the netfilter MARK target but socket-based).
 594 Changing the mark can be used for mark-based
 595 routing without netfilter or for packet filtering.
 596 Setting this option requires the
 597 .B CAP_NET_ADMIN
 598 capability.
 599 .TP
 600 .B SO_OOBINLINE
 601 If this option is enabled,
 602 out-of-band data is directly placed into the receive data stream.
 603 Otherwise, out-of-band data is passed only when the
 604 .B MSG_OOB
 605 flag is set during receiving.
 606 .\" don't document it because it can do too much harm.
 607 .\".B SO_NO_CHECK
 608 .\"     The kernel has support for the SO_NO_CHECK socket
 609 .\"     option (boolean: 0 == default, calculate checksum on xmit,
 610 .\"     1 == do not calculate checksum on xmit).
 611 .\" Additional note from Andi Kleen on SO_NO_CHECK (2010-08-30)
 612 .\"     On Linux UDP checksums are essentially free and there's no reason
 613 .\"     to turn them off and it would disable another safety line.
 614 .\"     That is why I didn't document the option.
 615 .TP
 616 .B SO_PASSCRED
 617 Enable or disable the receiving of the
 618 .B SCM_CREDENTIALS
 619 control message.
 620 For more information see
 621 .BR unix (7).
 622 .TP
 623 .B SO_PASSSEC
 624 Enable or disable the receiving of the
 625 .B SCM_SECURITY
 626 control message.
 627 For more information see
 628 .BR unix (7).
 629 .TP
 630 .BR SO_PEEK_OFF " (since Linux 3.4)"
 631 .\" commit ef64a54f6e558155b4f149bb10666b9e914b6c54
 632 This option, which is currently supported only for
 633 .BR unix (7)
 634 sockets, sets the value of the "peek offset" for the
 635 .BR recv (2)
 636 system call when used with
 637 .BR MSG_PEEK
 638 flag.
 639 .IP
 640 When this option is set to a negative value
 641 (it is set to \-1 for all new sockets),
 642 traditional behavior is provided:
 643 .BR recv (2)
 644 with the
 645 .BR MSG_PEEK
 646 flag will peek data from the front of the queue.
 647 .IP
 648 When the option is set to a value greater than or equal to zero,
 649 then the next peek at data queued in the socket will occur at
 650 the byte offset specified by the option value.
 651 At the same time, the "peek offset" will be
 652 incremented by the number of bytes that were peeked from the queue,
 653 so that a subsequent peek will return the next data in the queue.
 654 .IP
 655 If data is removed from the front of the queue via a call to
 656 .BR recv (2)
 657 (or similar) without the
 658 .BR MSG_PEEK
 659 flag, the "peek offset" will be decreased by the number of bytes removed.
 660 In other words, receiving data without the
 661 .B MSG_PEEK
 662 flag will cause the "peek offset" to be adjusted to maintain
 663 the correct relative position in the queued data,
 664 so that a subsequent peek will retrieve the data that would have been
 665 retrieved had the data not been removed.
 666 .IP
 667 For datagram sockets, if the "peek offset" points to the middle of a packet,
 668 the data returned will be marked with the
 669 .BR MSG_TRUNC
 670 flag.
 671 .IP
 672 The following example serves to illustrate the use of
 673 .BR SO_PEEK_OFF .
 674 Suppose a stream socket has the following queued input data:
 675 .IP
 676     aabbccddeeff
 677 .IP
 678 The following sequence of
 679 .BR recv (2)
 680 calls would have the effect noted in the comments:
 681 .IP
 682 .in +4n
 683 .EX
 684 int ov = 4;                  // Set peek offset to 4
 685 setsockopt(fd, SOL_SOCKET, SO_PEEK_OFF, &ov, sizeof(ov));
 686
 687 recv(fd, buf, 2, MSG_PEEK);  // Peeks "cc"; offset set to 6
 688 recv(fd, buf, 2, MSG_PEEK);  // Peeks "dd"; offset set to 8
 689 recv(fd, buf, 2, 0);         // Reads "aa"; offset set to 6
 690 recv(fd, buf, 2, MSG_PEEK);  // Peeks "ee"; offset set to 8
 691 .EE
 692 .in
 693 .TP
 694 .B SO_PEERCRED
 695 Return the credentials of the peer process connected to this socket.
 696 For further details, see
 697 .BR unix (7).
 698 .TP
 699 .B SO_PRIORITY
 700 Set the protocol-defined priority for all packets to be sent on
 701 this socket.
 702 Linux uses this value to order the networking queues:
 703 packets with a higher priority may be processed first depending
 704 on the selected device queueing discipline.
 705 .\" For
 706 .\" .BR ip (7),
 707 .\" this also sets the IP type-of-service (TOS) field for outgoing packets.
 708 Setting a priority outside the range 0 to 6 requires the
 709 .B CAP_NET_ADMIN
 710 capability.
 711 .TP
 712 .BR SO_PROTOCOL " (since Linux 2.6.32)"
 713 Retrieves the socket protocol as an integer, returning a value such as
 714 .BR IPPROTO_SCTP .
 715 See
 716 .BR socket (2)
 717 for details.
 718 This socket option is read-only.
 719 .TP
 720 .B SO_RCVBUF
 721 Sets or gets the maximum socket receive buffer in bytes.
 722 The kernel doubles this value (to allow space for bookkeeping overhead)
 723 when it is set using
 724 .\" Most (all?) other implementations do not do this -- MTK, Dec 05
 725 .BR setsockopt (2),
 726 and this doubled value is returned by
 727 .BR getsockopt (2).
 728 .\" The following thread on LMKL is quite informative:
 729 .\" getsockopt/setsockopt with SO_RCVBUF and SO_SNDBUF "non-standard" behavior
 730 .\" 17 July 2012
 731 .\" http://thread.gmane.org/gmane.linux.kernel/1328935
 732 The default value is set by the
 733 .I /proc/sys/net/core/rmem_default
 734 file, and the maximum allowed value is set by the
 735 .I /proc/sys/net/core/rmem_max
 736 file.
 737 The minimum (doubled) value for this option is 256.
 738 .TP
 739 .BR SO_RCVBUFFORCE " (since Linux 2.6.14)"
 740 Using this socket option, a privileged
 741 .RB ( CAP_NET_ADMIN )
 742 process can perform the same task as
 743 .BR SO_RCVBUF ,
 744 but the
 745 .I rmem_max
 746 limit can be overridden.
 747 .TP
 748 .BR SO_RCVLOWAT " and " SO_SNDLOWAT
 749 Specify the minimum number of bytes in the buffer until the socket layer
 750 will pass the data to the protocol
 751 .RB ( SO_SNDLOWAT )
 752 or the user on receiving
 753 .RB ( SO_RCVLOWAT ).
 754 These two values are initialized to 1.
 755 .B SO_SNDLOWAT
 756 is not changeable on Linux
 757 .RB ( setsockopt (2)
 758 fails with the error
 759 .BR ENOPROTOOPT ).
 760 .B SO_RCVLOWAT
 761 is changeable
 762 only since Linux 2.4.
 763 .IP
 764 Before Linux 2.6.28
 765 .\" commit c7004482e8dcb7c3c72666395cfa98a216a4fb70
 766 .BR select (2),
 767 .BR poll (2),
 768 and
 769 .BR epoll (7)
 770 did not respect the
 771 .B SO_RCVLOWAT
 772 setting on Linux,
 773 and indicated a socket as readable when even a single byte of data
 774 was available.
 775 A subsequent read from the socket would then block until
 776 .B SO_RCVLOWAT
 777 bytes are available.
 778 .\" See http://marc.theaimsgroup.com/?l=linux-kernel&m=111049368106984&w=2
 779 .\" Tested on kernel 2.6.14 -- mtk, 30 Nov 05
 780 .TP
 781 .BR SO_RCVTIMEO " and " SO_SNDTIMEO
 782 .\" Not implemented in 2.0.
 783 .\" Implemented in 2.1.11 for getsockopt: always return a zero struct.
 784 .\" Implemented in 2.3.41 for setsockopt, and actually used.
 785 Specify the receiving or sending timeouts until reporting an error.
 786 The argument is a
 787 .IR "struct timeval" .
 788 If an input or output function blocks for this period of time, and
 789 data has been sent or received, the return value of that function
 790 will be the amount of data transferred; if no data has been transferred
 791 and the timeout has been reached, then \-1 is returned with
 792 .I errno
 793 set to
 794 .BR EAGAIN
 795 or
 796 .BR EWOULDBLOCK ,
 797 .\" in fact to EAGAIN
 798 or
 799 .B EINPROGRESS
 800 (for
 801 .BR connect (2))
 802 just as if the socket was specified to be nonblocking.
 803 If the timeout is set to zero (the default),
 804 then the operation will never timeout.
 805 Timeouts only have effect for system calls that perform socket I/O (e.g.,
 806 .BR read (2),
 807 .BR recvmsg (2),
 808 .BR send (2),
 809 .BR sendmsg (2));
 810 timeouts have no effect for
 811 .BR select (2),
 812 .BR poll (2),
 813 .BR epoll_wait (2),
 814 and so on.
 815 .TP
 816 .B SO_REUSEADDR
 817 .\"     commit c617f398edd4db2b8567a28e899a88f8f574798d
 818 .\"     https://lwn.net/Articles/542629/
 819 Indicates that the rules used in validating addresses supplied in a
 820 .BR bind (2)
 821 call should allow reuse of local addresses.
 822 For
 823 .B AF_INET
 824 sockets this
 825 means that a socket may bind, except when there
 826 is an active listening socket bound to the address.
 827 When the listening socket is bound to
 828 .B INADDR_ANY
 829 with a specific port then it is not possible
 830 to bind to this port for any local address.
 831 Argument is an integer boolean flag.
 832 .TP
 833 .BR SO_REUSEPORT " (since Linux 3.9)"
 834 Permits multiple
 835 .B AF_INET
 836 or
 837 .B AF_INET6
 838 sockets to be bound to an identical socket address.
 839 This option must be set on each socket (including the first socket)
 840 prior to calling
 841 .BR bind (2)
 842 on the socket.
 843 To prevent port hijacking,
 844 all of the processes binding to the same address must have the same
 845 effective UID.
 846 This option can be employed with both TCP and UDP sockets.
 847 .IP
 848 For TCP sockets, this option allows
 849 .BR accept (2)
 850 load distribution in a multi-threaded server to be improved by
 851 using a distinct listener socket for each thread.
 852 This provides improved load distribution as compared
 853 to traditional techniques such using a single
 854 .BR accept (2)ing
 855 thread that distributes connections,
 856 or having multiple threads that compete to
 857 .BR accept (2)
 858 from the same socket.
 859 .IP
 860 For UDP sockets,
 861 the use of this option can provide better distribution
 862 of incoming datagrams to multiple processes (or threads) as compared
 863 to the traditional technique of having multiple processes
 864 compete to receive datagrams on the same socket.
 865 .TP
 866 .BR SO_RXQ_OVFL " (since Linux 2.6.33)"
 867 .\" commit 3b885787ea4112eaa80945999ea0901bf742707f
 868 Indicates that an unsigned 32-bit value ancillary message (cmsg)
 869 should be attached to received skbs indicating
 870 the number of packets dropped by the socket since its creation.
 871 .TP
 872 .BR SO_SELECT_ERR_QUEUE " (since Linux 3.10)"
 873 .\"     commit 7d4c04fc170087119727119074e72445f2bb192b
 874 .\"     Author: Keller, Jacob E <jacob.e.keller@intel.com>
 875 Makes poll adding
 876 .B POLLPRI
 877 when
 878 .B POLLERR
 879 event is returned. It does not affect wake up.
 880 .IP
 881 Background: The flag was added when waking up on
 882 .B POLLERR
 883 required requesting
 884 .B POLLIN
 885 or
 886 .B POLLPRI.
 887 After the commit 6e5d58fdc9bedd0255a8 ("skbuff: Fix not
 888 waking applications when errors are enqueued"), introduced
 889 in Linux 4.16, waking up on
 890 .B POLLERR
 891 does not require requesting other events. The flag is kept
 892 only for backwards compatibility.
 893 .TP
 894 .B SO_SNDBUF
 895 Sets or gets the maximum socket send buffer in bytes.
 896 The kernel doubles this value (to allow space for bookkeeping overhead)
 897 when it is set using
 898 .\" Most (all?) other implementations do not do this -- MTK, Dec 05
 899 .\" See also the comment to SO_RCVBUF (17 Jul 2012 LKML mail)
 900 .BR setsockopt (2),
 901 and this doubled value is returned by
 902 .BR getsockopt (2).
 903 The default value is set by the
 904 .I /proc/sys/net/core/wmem_default
 905 file and the maximum allowed value is set by the
 906 .I /proc/sys/net/core/wmem_max
 907 file.
 908 The minimum (doubled) value for this option is 2048.
 909 .TP
 910 .BR SO_SNDBUFFORCE " (since Linux 2.6.14)"
 911 Using this socket option, a privileged
 912 .RB ( CAP_NET_ADMIN )
 913 process can perform the same task as
 914 .BR SO_SNDBUF ,
 915 but the
 916 .I wmem_max
 917 limit can be overridden.
 918 .TP
 919 .B SO_TIMESTAMP
 920 Enable or disable the receiving of the
 921 .B SO_TIMESTAMP
 922 control message.
 923 The timestamp control message is sent with level
 924 .B SOL_SOCKET
 925 and the
 926 .I cmsg_data
 927 field is a
 928 .I "struct timeval"
 929 indicating the
 930 reception time of the last packet passed to the user in this call.
 931 See
 932 .BR cmsg (3)
 933 for details on control messages.
 934 .TP
 935 .B SO_TYPE
 936 Gets the socket type as an integer (e.g.,
 937 .BR SOCK_STREAM ).
 938 This socket option is read-only.
 939 .TP
 940 .BR SO_BUSY_POLL " (since Linux 3.11)"
 941 Sets the approximate time in microseconds to busy poll on a blocking receive
 942 when there is no data.
 943 Increasing this value requires
 944 .BR CAP_NET_ADMIN .
 945 The default for this option is controlled by the
 946 .I /proc/sys/net/core/busy_read
 947 file.
 948 .IP
 949 The value in the
 950 .I /proc/sys/net/core/busy_poll
 951 file determines how long
 952 .BR select (2)
 953 and
 954 .BR poll (2)
 955 will busy poll when they operate on sockets with
 956 .BR SO_BUSY_POLL
 957 set and no events to report are found.
 958 .IP
 959 In both cases,
 960 busy polling will only be done when the socket last received data
 961 from a network device that supports this option.
 962 .IP
 963 While busy polling may improve latency of some applications,
 964 care must be taken when using it since this will increase
 965 both CPU utilization and power usage.
 966 .SS Signals
 967 When writing onto a connection-oriented socket that has been shut down
 968 (by the local or the remote end)
 969 .B SIGPIPE
 970 is sent to the writing process and
 971 .B EPIPE
 972 is returned.
 973 The signal is not sent when the write call
 974 specified the
 975 .B MSG_NOSIGNAL
 976 flag.
 977 .PP
 978 When requested with the
 979 .B FIOSETOWN
 980 .BR fcntl (2)
 981 or
 982 .B SIOCSPGRP
 983 .BR ioctl (2),
 984 .B SIGIO
 985 is sent when an I/O event occurs.
 986 It is possible to use
 987 .BR poll (2)
 988 or
 989 .BR select (2)
 990 in the signal handler to find out which socket the event occurred on.
 991 An alternative (in Linux 2.2) is to set a real-time signal using the
 992 .B F_SETSIG
 993 .BR fcntl (2);
 994 the handler of the real time signal will be called with
 995 the file descriptor in the
 996 .I si_fd
 997 field of its
 998 .IR siginfo_t .
 999 See
1000 .BR fcntl (2)
1001 for more information.
1002 .PP
1003 Under some circumstances (e.g., multiple processes accessing a
1004 single socket), the condition that caused the
1005 .B SIGIO
1006 may have already disappeared when the process reacts to the signal.
1007 If this happens, the process should wait again because Linux
1008 will resend the signal later.
1009 .\" .SS Ancillary messages
1010 .SS /proc interfaces
1011 The core socket networking parameters can be accessed
1012 via files in the directory
1013 .IR /proc/sys/net/core/ .
1014 .TP
1015 .I rmem_default
1016 contains the default setting in bytes of the socket receive buffer.
1017 .TP
1018 .I rmem_max
1019 contains the maximum socket receive buffer size in bytes which a user may
1020 set by using the
1021 .B SO_RCVBUF
1022 socket option.
1023 .TP
1024 .I wmem_default
1025 contains the default setting in bytes of the socket send buffer.
1026 .TP
1027 .I wmem_max
1028 contains the maximum socket send buffer size in bytes which a user may
1029 set by using the
1030 .B SO_SNDBUF
1031 socket option.
1032 .TP
1033 .IR message_cost " and " message_burst
1034 configure the token bucket filter used to load limit warning messages
1035 caused by external network events.
1036 .TP
1037 .I netdev_max_backlog
1038 Maximum number of packets in the global input queue.
1039 .TP
1040 .I optmem_max
1041 Maximum length of ancillary data and user control data like the iovecs
1042 per socket.
1043 .\" netdev_fastroute is not documented because it is experimental
1044 .SS Ioctls
1045 These operations can be accessed using
1046 .BR ioctl (2):
1047 .PP
1048 .in +4n
1049 .EX
1050 .IB error " = ioctl(" ip_socket ", " ioctl_type ", " &value_result ");"
1051 .EE
1052 .in
1053 .TP
1054 .B SIOCGSTAMP
1055 Return a
1056 .I struct timeval
1057 with the receive timestamp of the last packet passed to the user.
1058 This is useful for accurate round trip time measurements.
1059 See
1060 .BR setitimer (2)
1061 for a description of
1062 .IR "struct timeval" .
1063 .\"
1064 This ioctl should be used only if the socket option
1065 .B SO_TIMESTAMP
1066 is not set on the socket.
1067 Otherwise, it returns the timestamp of the
1068 last packet that was received while
1069 .B SO_TIMESTAMP
1070 was not set, or it fails if no such packet has been received,
1071 (i.e.,
1072 .BR ioctl (2)
1073 returns \-1 with
1074 .I errno
1075 set to
1076 .BR ENOENT ).
1077 .TP
1078 .B SIOCSPGRP
1079 Set the process or process group that is to receive
1080 .B SIGIO
1081 or
1082 .B SIGURG
1083 signals when I/O becomes possible or urgent data is available.
1084 The argument is a pointer to a
1085 .IR pid_t .
1086 For further details, see the description of
1087 .BR F_SETOWN
1088 in
1089 .BR fcntl (2).
1090 .TP
1091 .B FIOASYNC
1092 Change the
1093 .B O_ASYNC
1094 flag to enable or disable asynchronous I/O mode of the socket.
1095 Asynchronous I/O mode means that the
1096 .B SIGIO
1097 signal or the signal set with
1098 .B F_SETSIG
1099 is raised when a new I/O event occurs.
1100 .IP
1101 Argument is an integer boolean flag.
1102 (This operation is synonymous with the use of
1103 .BR fcntl (2)
1104 to set the
1105 .B O_ASYNC
1106 flag.)
1107 .\"
1108 .TP
1109 .B SIOCGPGRP
1110 Get the current process or process group that receives
1111 .B SIGIO
1112 or
1113 .B SIGURG
1114 signals,
1115 or 0
1116 when none is set.
1117 .PP
1118 Valid
1119 .BR fcntl (2)
1120 operations:
1121 .TP
1122 .B FIOGETOWN
1123 The same as the
1124 .B SIOCGPGRP
1125 .BR ioctl (2).
1126 .TP
1127 .B FIOSETOWN
1128 The same as the
1129 .B SIOCSPGRP
1130 .BR ioctl (2).
1131 .SH VERSIONS
1132 .B SO_BINDTODEVICE
1133 was introduced in Linux 2.0.30.
1134 .B SO_PASSCRED
1135 is new in Linux 2.2.
1136 The
1137 .I /proc
1138 interfaces were introduced in Linux 2.2.
1139 .B SO_RCVTIMEO
1140 and
1141 .B SO_SNDTIMEO
1142 are supported since Linux 2.3.41.
1143 Earlier, timeouts were fixed to
1144 a protocol-specific setting, and could not be read or written.
1145 .SH NOTES
1146 Linux assumes that half of the send/receive buffer is used for internal
1147 kernel structures; thus the values in the corresponding
1148 .I /proc
1149 files are twice what can be observed on the wire.
1150 .PP
1151 Linux will allow port reuse only with the
1152 .B SO_REUSEADDR
1153 option
1154 when this option was set both in the previous program that performed a
1155 .BR bind (2)
1156 to the port and in the program that wants to reuse the port.
1157 This differs from some implementations (e.g., FreeBSD)
1158 where only the later program needs to set the
1159 .B SO_REUSEADDR
1160 option.
1161 Typically this difference is invisible, since, for example, a server
1162 program is designed to always set this option.
1163 .\" .SH AUTHORS
1164 .\" This man page was written by Andi Kleen.
1165 .SH SEE ALSO
1166 .BR wireshark (1),
1167 .BR bpf (2),
1168 .BR connect (2),
1169 .BR getsockopt (2),
1170 .BR setsockopt (2),
1171 .BR socket (2),
1172 .BR pcap (3),
1173 .BR address_families (7),
1174 .BR capabilities (7),
1175 .BR ddp (7),
1176 .BR ip (7),
1177 .BR packet (7),
1178 .BR tcp (7),
1179 .BR udp (7),
1180 .BR unix (7),
1181 .BR tcpdump (8)