man7/ip.7

   1 '\" t
   2 .\" This man page is Copyright (C) 1999 Andi Kleen <ak@muc.de>.
   3 .\"
   4 .\" %%%LICENSE_START(VERBATIM_ONE_PARA)
   5 .\" Permission is granted to distribute possibly modified copies
   6 .\" of this page provided the header is included verbatim,
   7 .\" and in case of nontrivial modification author and date
   8 .\" of the modification is added to the header.
   9 .\" %%%LICENSE_END
  10 .\"
  11 .\" $Id: ip.7,v 1.19 2000/12/20 18:10:31 ak Exp $
  12 .\"
  13 .\" FIXME The following socket options are yet to be documented
  14 .\"
  15 .\"     IP_XFRM_POLICY (2.5.48)
  16 .\"         Needs CAP_NET_ADMIN
  17 .\"
  18 .\"     IP_IPSEC_POLICY (2.5.47)
  19 .\"         Needs CAP_NET_ADMIN
  20 .\"
  21 .\"     IP_PASSSEC (2.6.17)
  22 .\"         Boolean
  23 .\"         commit 2c7946a7bf45ae86736ab3b43d0085e43947945c
  24 .\"         Author: Catherine Zhang <cxzhang@watson.ibm.com>
  25 .\"
  26 .\"     IP_MINTTL (2.6.34)
  27 .\"         commit d218d11133d888f9745802146a50255a4781d37a
  28 .\"         Author: Stephen Hemminger <shemminger@vyatta.com>
  29 .\"
  30 .\"     MCAST_JOIN_GROUP (2.4.22 / 2.6)
  31 .\"
  32 .\"     MCAST_BLOCK_SOURCE (2.4.22 / 2.6)
  33 .\"
  34 .\"     MCAST_UNBLOCK_SOURCE (2.4.22 / 2.6)
  35 .\"
  36 .\"     MCAST_LEAVE_GROUP (2.4.22 / 2.6)
  37 .\"
  38 .\"     MCAST_JOIN_SOURCE_GROUP (2.4.22 / 2.6)
  39 .\"
  40 .\"     MCAST_LEAVE_SOURCE_GROUP (2.4.22 / 2.6)
  41 .\"
  42 .\"     MCAST_MSFILTER (2.4.22 / 2.6)
  43 .\"
  44 .\"     IP_UNICAST_IF (3.4)
  45 .\"         commit 76e21053b5bf33a07c76f99d27a74238310e3c71
  46 .\"         Author: Erich E. Hoover <ehoover@mines.edu>
  47 .\"
  48 .TH IP  7 2017-03-13 "Linux" "Linux Programmer's Manual"
  49 .SH NAME
  50 ip \- Linux IPv4 protocol implementation
  51 .SH SYNOPSIS
  52 .B #include <sys/socket.h>
  53 .br
  54 .\" .B #include <net/netinet.h> -- does not exist anymore
  55 .\" .B #include <linux/errqueue.h> -- never include <linux/foo.h>
  56 .B #include <netinet/in.h>
  57 .br
  58 .B #include <netinet/ip.h>        \fR/* superset of previous */
  59 .PP
  60 .IB tcp_socket " = socket(AF_INET, SOCK_STREAM, 0);"
  61 .br
  62 .IB udp_socket " = socket(AF_INET, SOCK_DGRAM, 0);"
  63 .br
  64 .IB raw_socket " = socket(AF_INET, SOCK_RAW, " protocol ");"
  65 .SH DESCRIPTION
  66 Linux implements the Internet Protocol, version 4,
  67 described in RFC\ 791 and RFC\ 1122.
  68 .B ip
  69 contains a level 2 multicasting implementation conforming to RFC\ 1112.
  70 It also contains an IP router including a packet filter.
  71 .PP
  72 The programming interface is BSD-sockets compatible.
  73 For more information on sockets, see
  74 .BR socket (7).
  75 .PP
  76 An IP socket is created using
  77 .BR socket (2):
  78 .PP
  79     socket(AF_INET, socket_type, protocol);
  80 .PP
  81 Valid socket types are
  82 .B SOCK_STREAM
  83 to open a
  84 .BR tcp (7)
  85 socket,
  86 .B SOCK_DGRAM
  87 to open a
  88 .BR udp (7)
  89 socket, or
  90 .B SOCK_RAW
  91 to open a
  92 .BR raw (7)
  93 socket to access the IP protocol directly.
  94 .I protocol
  95 is the IP protocol in the IP header to be received or sent.
  96 The only valid values for
  97 .I protocol
  98 are 0 and
  99 .B IPPROTO_TCP
 100 for TCP sockets, and 0 and
 101 .B IPPROTO_UDP
 102 for UDP sockets.
 103 For
 104 .B SOCK_RAW
 105 you may specify a valid IANA IP protocol defined in
 106 RFC\ 1700 assigned numbers.
 107 .PP
 108 When a process wants to receive new incoming packets or connections, it
 109 should bind a socket to a local interface address using
 110 .BR bind (2).
 111 In this case, only one IP socket may be bound to any given local
 112 (address, port) pair.
 113 When
 114 .B INADDR_ANY
 115 is specified in the bind call, the socket will be bound to
 116 .I all
 117 local interfaces.
 118 When
 119 .BR listen (2)
 120 is called on an unbound socket, the socket is automatically bound
 121 to a random free port with the local address set to
 122 .BR INADDR_ANY .
 123 When
 124 .BR connect (2)
 125 is called on an unbound socket, the socket is automatically bound
 126 to a random free port or to a usable shared port with the local address
 127 set to
 128 .BR INADDR_ANY .
 129 .PP
 130 A TCP local socket address that has been bound is unavailable for
 131 some time after closing, unless the
 132 .B SO_REUSEADDR
 133 flag has been set.
 134 Care should be taken when using this flag as it makes TCP less reliable.
 135 .SS Address format
 136 An IP socket address is defined as a combination of an IP interface
 137 address and a 16-bit port number.
 138 The basic IP protocol does not supply port numbers, they
 139 are implemented by higher level protocols like
 140 .BR udp (7)
 141 and
 142 .BR tcp (7).
 143 On raw sockets
 144 .I sin_port
 145 is set to the IP protocol.
 146 .PP
 147 .in +4n
 148 .EX
 149 struct sockaddr_in {
 150     sa_family_t    sin_family; /* address family: AF_INET */
 151     in_port_t      sin_port;   /* port in network byte order */
 152     struct in_addr sin_addr;   /* internet address */
 153 };
 154
 155 /* Internet address. */
 156 struct in_addr {
 157     uint32_t       s_addr;     /* address in network byte order */
 158 };
 159 .EE
 160 .in
 161 .PP
 162 .I sin_family
 163 is always set to
 164 .BR AF_INET .
 165 This is required; in Linux 2.2 most networking functions return
 166 .B EINVAL
 167 when this setting is missing.
 168 .I sin_port
 169 contains the port in network byte order.
 170 The port numbers below 1024 are called
 171 .IR "privileged ports"
 172 (or sometimes:
 173 .IR "reserved ports" ).
 174 Only a privileged process
 175 (on Linux: a process that has the
 176 .B CAP_NET_BIND_SERVICE
 177 capability in the user namespace governing its network namespace) may
 178 .BR bind (2)
 179 to these sockets.
 180 Note that the raw IPv4 protocol as such has no concept of a
 181 port, they are implemented only by higher protocols like
 182 .BR tcp (7)
 183 and
 184 .BR udp (7).
 185 .PP
 186 .I sin_addr
 187 is the IP host address.
 188 The
 189 .I s_addr
 190 member of
 191 .I struct in_addr
 192 contains the host interface address in network byte order.
 193 .I in_addr
 194 should be assigned one of the
 195 .BR INADDR_*
 196 values (e.g.,
 197 .BR INADDR_ANY )
 198 or set using the
 199 .BR inet_aton (3),
 200 .BR inet_addr (3),
 201 .BR inet_makeaddr (3)
 202 library functions or directly with the name resolver (see
 203 .BR gethostbyname (3)).
 204 .PP
 205 IPv4 addresses are divided into unicast, broadcast,
 206 and multicast addresses.
 207 Unicast addresses specify a single interface of a host,
 208 broadcast addresses specify all hosts on a network, and multicast
 209 addresses address all hosts in a multicast group.
 210 Datagrams to broadcast addresses can be sent or received only when the
 211 .B SO_BROADCAST
 212 socket flag is set.
 213 In the current implementation, connection-oriented sockets are allowed
 214 to use only unicast addresses.
 215 .\" Leave a loophole for XTP @)
 216 .PP
 217 Note that the address and the port are always stored in
 218 network byte order.
 219 In particular, this means that you need to call
 220 .BR htons (3)
 221 on the number that is assigned to a port.
 222 All address/port manipulation
 223 functions in the standard library work in network byte order.
 224 .PP
 225 There are several special addresses:
 226 .B INADDR_LOOPBACK
 227 (127.0.0.1)
 228 always refers to the local host via the loopback device;
 229 .B INADDR_ANY
 230 (0.0.0.0)
 231 means any address for binding;
 232 .B INADDR_BROADCAST
 233 (255.255.255.255)
 234 means any host and has the same effect on bind as
 235 .B INADDR_ANY
 236 for historical reasons.
 237 .SS Socket options
 238 IP supports some protocol-specific socket options that can be set with
 239 .BR setsockopt (2)
 240 and read with
 241 .BR getsockopt (2).
 242 The socket option level for IP is
 243 .BR IPPROTO_IP .
 244 .\" or SOL_IP on Linux
 245 A boolean integer flag is zero when it is false, otherwise true.
 246 .PP
 247 When an invalid socket option is specified,
 248 .BR getsockopt (2)
 249 and
 250 .BR setsockopt (2)
 251 fail with the error
 252 .BR ENOPROTOOPT .
 253 .TP
 254 .BR IP_ADD_MEMBERSHIP " (since Linux 1.2)"
 255 Join a multicast group.
 256 Argument is an
 257 .I ip_mreqn
 258 structure.
 259 .PP
 260 .in +4n
 261 .EX
 262 struct ip_mreqn {
 263     struct in_addr imr_multiaddr; /* IP multicast group
 264                                      address */
 265     struct in_addr imr_address;   /* IP address of local
 266                                      interface */
 267     int            imr_ifindex;   /* interface index */
 268 };
 269 .EE
 270 .in
 271 .PP
 272 .I imr_multiaddr
 273 contains the address of the multicast group the application
 274 wants to join or leave.
 275 It must be a valid multicast address
 276 .\" (i.e., within the 224.0.0.0-239.255.255.255 range)
 277 (or
 278 .BR setsockopt (2)
 279 fails with the error
 280 .BR EINVAL ).
 281 .I imr_address
 282 is the address of the local interface with which the system
 283 should join the multicast group; if it is equal to
 284 .BR INADDR_ANY ,
 285 an appropriate interface is chosen by the system.
 286 .I imr_ifindex
 287 is the interface index of the interface that should join/leave the
 288 .I imr_multiaddr
 289 group, or 0 to indicate any interface.
 290 .IP
 291 The
 292 .I ip_mreqn
 293 structure is available only since Linux 2.2.
 294 For compatibility, the old
 295 .I ip_mreq
 296 structure (present since Linux 1.2) is still supported;
 297 it differs from
 298 .I ip_mreqn
 299 only by not including the
 300 .I imr_ifindex
 301 field.
 302 (The kernel determines which structure is being passed based
 303 on the size passed in
 304 .IR optlen .)
 305 .IP
 306 .B IP_ADD_MEMBERSHIP
 307 is valid only for
 308 .BR setsockopt (2).
 309 .\"
 310 .TP
 311 .BR IP_ADD_SOURCE_MEMBERSHIP " (since Linux 2.4.22 / 2.5.68)"
 312 Join a multicast group and allow receiving data only
 313 from a specified source.
 314 Argument is an
 315 .I ip_mreq_source
 316 structure.
 317 .PP
 318 .in +4n
 319 .EX
 320 struct ip_mreq_source {
 321     struct in_addr imr_multiaddr;  /* IP multicast group
 322                                       address */
 323     struct in_addr imr_interface;  /* IP address of local
 324                                       interface */
 325     struct in_addr imr_sourceaddr; /* IP address of
 326                                       multicast source */
 327 };
 328 .EE
 329 .in
 330 .PP
 331 The
 332 .I ip_mreq_source
 333 structure is similar to
 334 .I ip_mreqn
 335 described under
 336 .BR IP_ADD_MEMBERSIP .
 337 The
 338 .I imr_multiaddr
 339 field contains the address of the multicast group the application
 340 wants to join or leave.
 341 The
 342 .I imr_interface
 343 field is the address of the local interface with which
 344 the system should join the multicast group.
 345 Finally, the
 346 .I imr_sourceaddr
 347 field contains the address of the source the
 348 application wants to receive data from.
 349 .IP
 350 This option can be used multiple times to allow
 351 receiving data from more than one source.
 352 .TP
 353 .BR IP_BIND_ADDRESS_NO_PORT " (since Linux 4.2)"
 354 .\" commit 90c337da1524863838658078ec34241f45d8394d
 355 Inform the kernel to not reserve an ephemeral port when using
 356 .BR bind (2)
 357 with a port number of 0.
 358 The port will later be automatically chosen at
 359 .BR connect (2)
 360 time,
 361 in a way that allows sharing a source port as long as the 4-tuple is unique.
 362 .TP
 363 .BR IP_BLOCK_SOURCE " (since Linux 2.4.22 / 2.5.68)"
 364 Stop receiving multicast data from a specific source in a given group.
 365 This is valid only after the application has subscribed
 366 to the multicast group using either
 367 .BR IP_ADD_MEMBERSHIP
 368 or
 369 .BR IP_ADD_SOURCE_MEMBERSHIP .
 370 .IP
 371 Argument is an
 372 .I ip_mreq_source
 373 structure as described under
 374 .BR IP_ADD_SOURCE_MEMBERSHIP .
 375 .TP
 376 .BR IP_DROP_MEMBERSHIP " (since Linux 1.2)"
 377 Leave a multicast group.
 378 Argument is an
 379 .I ip_mreqn
 380 or
 381 .I ip_mreq
 382 structure similar to
 383 .BR IP_ADD_MEMBERSHIP .
 384 .TP
 385 .BR IP_DROP_SOURCE_MEMBERSHIP " (since Linux 2.4.22 / 2.5.68)"
 386 Leave a source-specific group\(emthat is, stop receiving data from
 387 a given multicast group that come from a given source.
 388 If the application has subscribed to multiple sources within
 389 the same group, data from the remaining sources will still be delivered.
 390 To stop receiving data from all sources at once, use
 391 .BR IP_DROP_MEMBERSHIP .
 392 .IP
 393 Argument is an
 394 .I ip_mreq_source
 395 structure as described under
 396 .BR IP_ADD_SOURCE_MEMBERSHIP .
 397 .TP
 398 .BR IP_FREEBIND " (since Linux 2.4)"
 399 .\" Precisely: 2.4.0-test10
 400 If enabled, this boolean option allows binding to an IP address
 401 that is nonlocal or does not (yet) exist.
 402 This permits listening on a socket,
 403 without requiring the underlying network interface or the
 404 specified dynamic IP address to be up at the time that
 405 the application is trying to bind to it.
 406 This option is the per-socket equivalent of the
 407 .IR ip_nonlocal_bind
 408 .I /proc
 409 interface described below.
 410 .TP
 411 .BR IP_HDRINCL " (since Linux 2.0)"
 412 If enabled,
 413 the user supplies an IP header in front of the user data.
 414 Valid only for
 415 .B SOCK_RAW
 416 sockets; see
 417 .BR raw (7)
 418 for more information.
 419 When this flag is enabled, the values set by
 420 .BR IP_OPTIONS ,
 421 .BR IP_TTL ,
 422 and
 423 .B IP_TOS
 424 are ignored.
 425 .TP
 426 .BR IP_MSFILTER " (since Linux 2.4.22 / 2.5.68)"
 427 This option provides access to the advanced full-state filtering API.
 428 Argument is an
 429 .I ip_msfilter
 430 structure.
 431 .PP
 432 .in +4n
 433 .EX
 434 struct ip_msfilter {
 435     struct in_addr imsf_multiaddr; /* IP multicast group
 436                                       address */
 437     struct in_addr imsf_interface; /* IP address of local
 438                                       interface */
 439     uint32_t       imsf_fmode;     /* Filter-mode */
 440
 441     uint32_t       imsf_numsrc;    /* Number of sources in
 442                                       the following array */
 443     struct in_addr imsf_slist[1];  /* Array of source
 444                                       addresses */
 445 };
 446 .EE
 447 .in
 448 .PP
 449 There are two macros,
 450 .BR MCAST_INCLUDE
 451 and
 452 .BR MCAST_EXCLUDE ,
 453 which can be used to specify the filtering mode.
 454 Additionally, the
 455 .BR IP_MSFILTER_SIZE (n)
 456 macro exists to determine how much memory is needed to store
 457 .I ip_msfilter
 458 structure with
 459 .I n
 460 sources in the source list.
 461 .IP
 462 For the full description of multicast source filtering
 463 refer to RFC 3376.
 464 .TP
 465 .BR IP_MTU " (since Linux 2.2)"
 466 .\" Precisely: 2.1.124
 467 Retrieve the current known path MTU of the current socket.
 468 Returns an integer.
 469 .IP
 470 .B IP_MTU
 471 is valid only for
 472 .BR getsockopt (2)
 473 and can be employed only when the socket has been connected.
 474 .TP
 475 .BR IP_MTU_DISCOVER " (since Linux 2.2)"
 476 .\" Precisely: 2.1.124
 477 Set or receive the Path MTU Discovery setting for a socket.
 478 When enabled, Linux will perform Path MTU Discovery
 479 as defined in RFC\ 1191 on
 480 .B SOCK_STREAM
 481 sockets.
 482 For
 483 .RB non- SOCK_STREAM
 484 sockets,
 485 .B IP_PMTUDISC_DO
 486 forces the don't-fragment flag to be set on all outgoing packets.
 487 It is the user's responsibility to packetize the data
 488 in MTU-sized chunks and to do the retransmits if necessary.
 489 The kernel will reject (with
 490 .BR EMSGSIZE )
 491 datagrams that are bigger than the known path MTU.
 492 .B IP_PMTUDISC_WANT
 493 will fragment a datagram if needed according to the path MTU,
 494 or will set the don't-fragment flag otherwise.
 495 .IP
 496 The system-wide default can be toggled between
 497 .B IP_PMTUDISC_WANT
 498 and
 499 .B IP_PMTUDISC_DONT
 500 by writing (respectively, zero and nonzero values) to the
 501 .I /proc/sys/net/ipv4/ip_no_pmtu_disc
 502 file.
 503 .TS
 504 tab(:);
 505 c l
 506 l l.
 507 Path MTU discovery value:Meaning
 508 IP_PMTUDISC_WANT:Use per-route settings.
 509 IP_PMTUDISC_DONT:Never do Path MTU Discovery.
 510 IP_PMTUDISC_DO:Always do Path MTU Discovery.
 511 IP_PMTUDISC_PROBE:Set DF but ignore Path MTU.
 512 .TE
 513 .sp 1
 514 When PMTU discovery is enabled, the kernel automatically keeps track of
 515 the path MTU per destination host.
 516 When it is connected to a specific peer with
 517 .BR connect (2),
 518 the currently known path MTU can be retrieved conveniently using the
 519 .B IP_MTU
 520 socket option (e.g., after an
 521 .B EMSGSIZE
 522 error occurred).
 523 The path MTU may change over time.
 524 For connectionless sockets with many destinations,
 525 the new MTU for a given destination can also be accessed using the
 526 error queue (see
 527 .BR IP_RECVERR ).
 528 A new error will be queued for every incoming MTU update.
 529 .IP
 530 While MTU discovery is in progress, initial packets from datagram sockets
 531 may be dropped.
 532 Applications using UDP should be aware of this and not
 533 take it into account for their packet retransmit strategy.
 534 .IP
 535 To bootstrap the path MTU discovery process on unconnected sockets, it
 536 is possible to start with a big datagram size
 537 (headers up to 64 kilobytes long) and let it shrink by updates of the path MTU.
 538 .IP
 539 To get an initial estimate of the
 540 path MTU, connect a datagram socket to the destination address using
 541 .BR connect (2)
 542 and retrieve the MTU by calling
 543 .BR getsockopt (2)
 544 with the
 545 .B IP_MTU
 546 option.
 547 .IP
 548 It is possible to implement RFC 4821 MTU probing with
 549 .B SOCK_DGRAM
 550 or
 551 .B SOCK_RAW
 552 sockets by setting a value of
 553 .BR IP_PMTUDISC_PROBE
 554 (available since Linux 2.6.22).
 555 This is also particularly useful for diagnostic tools such as
 556 .BR tracepath (8)
 557 that wish to deliberately send probe packets larger than
 558 the observed Path MTU.
 559 .TP
 560 .BR IP_MULTICAST_ALL " (since Linux 2.6.31)"
 561 This option can be used to modify the delivery policy of multicast messages
 562 to sockets bound to the wildcard
 563 .B INADDR_ANY
 564 address.
 565 The argument is a boolean integer (defaults to 1).
 566 If set to 1,
 567 the socket will receive messages from all the groups that have been joined
 568 globally on the whole system.
 569 Otherwise, it will deliver messages only from
 570 the groups that have been explicitly joined (for example via the
 571 .B IP_ADD_MEMBERSHIP
 572 option) on this particular socket.
 573 .TP
 574 .BR IP_MULTICAST_IF " (since Linux 1.2)"
 575 Set the local device for a multicast socket.
 576 The argument for
 577 .BR setsockopt (2)
 578 is an
 579 .I ip_mreqn
 580 or
 581 .\" net: IP_MULTICAST_IF setsockopt now recognizes struct mreq
 582 .\" Commit: 3a084ddb4bf299a6e898a9a07c89f3917f0713f7
 583 (since Linux 3.5)
 584 .I ip_mreq
 585 structure similar to
 586 .BR IP_ADD_MEMBERSHIP ,
 587 or an
 588 .I in_addr
 589 structure.
 590 (The kernel determines which structure is being passed based
 591 on the size passed in
 592 .IR optlen .)
 593 For
 594 .BR getsockopt (2),
 595 the argument is an
 596 .I in_addr
 597 structure.
 598 .TP
 599 .BR IP_MULTICAST_LOOP " (since Linux 1.2)"
 600 Set or read a boolean integer argument that determines whether
 601 sent multicast packets should be looped back to the local sockets.
 602 .TP
 603 .BR IP_MULTICAST_TTL " (since Linux 1.2)"
 604 Set or read the time-to-live value of outgoing multicast packets for this
 605 socket.
 606 It is very important for multicast packets to set the smallest TTL possible.
 607 The default is 1 which means that multicast packets don't leave the local
 608 network unless the user program explicitly requests it.
 609 Argument is an integer.
 610 .TP
 611 .BR IP_NODEFRAG " (since Linux 2.6.36)"
 612 If enabled (argument is nonzero),
 613 the reassembly of outgoing packets is disabled in the netfilter layer.
 614 The argument is an integer.
 615 .IP
 616 This option is valid only for
 617 .B SOCK_RAW
 618 sockets.
 619 .TP
 620 .BR IP_OPTIONS " (since Linux 2.0)"
 621 .\" Precisely: 1.3.30
 622 Set or get the IP options to be sent with every packet from this socket.
 623 The arguments are a pointer to a memory buffer containing the options
 624 and the option length.
 625 The
 626 .BR setsockopt (2)
 627 call sets the IP options associated with a socket.
 628 The maximum option size for IPv4 is 40 bytes.
 629 See RFC\ 791 for the allowed options.
 630 When the initial connection request packet for a
 631 .B SOCK_STREAM
 632 socket contains IP options, the IP options will be set automatically
 633 to the options from the initial packet with routing headers reversed.
 634 Incoming packets are not allowed to change options after the connection
 635 is established.
 636 The processing of all incoming source routing options
 637 is disabled by default and can be enabled by using the
 638 .I accept_source_route
 639 .I /proc
 640 interface.
 641 Other options like timestamps are still handled.
 642 For datagram sockets, IP options can be only set by the local user.
 643 Calling
 644 .BR getsockopt (2)
 645 with
 646 .B IP_OPTIONS
 647 puts the current IP options used for sending into the supplied buffer.
 648 .TP
 649 .BR IP_PKTINFO " (since Linux 2.2)"
 650 .\" Precisely: 2.1.68
 651 Pass an
 652 .B IP_PKTINFO
 653 ancillary message that contains a
 654 .I pktinfo
 655 structure that supplies some information about the incoming packet.
 656 This only works for datagram oriented sockets.
 657 The argument is a flag that tells the socket whether the
 658 .B IP_PKTINFO
 659 message should be passed or not.
 660 The message itself can only be sent/retrieved
 661 as control message with a packet using
 662 .BR recvmsg (2)
 663 or
 664 .BR sendmsg (2).
 665 .IP
 666 .in +4n
 667 .EX
 668 struct in_pktinfo {
 669     unsigned int   ipi_ifindex;  /* Interface index */
 670     struct in_addr ipi_spec_dst; /* Local address */
 671     struct in_addr ipi_addr;     /* Header Destination
 672                                     address */
 673 };
 674 .EE
 675 .in
 676 .IP
 677 .I ipi_ifindex
 678 is the unique index of the interface the packet was received on.
 679 .I ipi_spec_dst
 680 is the local address of the packet and
 681 .I ipi_addr
 682 is the destination address in the packet header.
 683 If
 684 .B IP_PKTINFO
 685 is passed to
 686 .BR sendmsg (2)
 687 and
 688 .\" This field is grossly misnamed
 689 .I ipi_spec_dst
 690 is not zero, then it is used as the local source address for the routing
 691 table lookup and for setting up IP source route options.
 692 When
 693 .I ipi_ifindex
 694 is not zero, the primary local address of the interface specified by the
 695 index overwrites
 696 .I ipi_spec_dst
 697 for the routing table lookup.
 698 .TP
 699 .BR IP_RECVERR " (since Linux 2.2)"
 700 .\" Precisely: 2.1.15
 701 Enable extended reliable error message passing.
 702 When enabled on a datagram socket, all
 703 generated errors will be queued in a per-socket error queue.
 704 When the user receives an error from a socket operation,
 705 the errors can be received by calling
 706 .BR recvmsg (2)
 707 with the
 708 .B MSG_ERRQUEUE
 709 flag set.
 710 The
 711 .I sock_extended_err
 712 structure describing the error will be passed in an ancillary message with
 713 the type
 714 .B IP_RECVERR
 715 and the level
 716 .BR IPPROTO_IP .
 717 .\" or SOL_IP on Linux
 718 This is useful for reliable error handling on unconnected sockets.
 719 The received data portion of the error queue contains the error packet.
 720 .IP
 721 The
 722 .B IP_RECVERR
 723 control message contains a
 724 .I sock_extended_err
 725 structure:
 726 .IP
 727 .in +4n
 728 .EX
 729 #define SO_EE_ORIGIN_NONE    0
 730 #define SO_EE_ORIGIN_LOCAL   1
 731 #define SO_EE_ORIGIN_ICMP    2
 732 #define SO_EE_ORIGIN_ICMP6   3
 733
 734 struct sock_extended_err {
 735     uint32_t ee_errno;   /* error number */
 736     uint8_t  ee_origin;  /* where the error originated */
 737     uint8_t  ee_type;    /* type */
 738     uint8_t  ee_code;    /* code */
 739     uint8_t  ee_pad;
 740     uint32_t ee_info;    /* additional information */
 741     uint32_t ee_data;    /* other data */
 742     /* More data may follow */
 743 };
 744
 745 struct sockaddr *SO_EE_OFFENDER(struct sock_extended_err *);
 746 .EE
 747 .in
 748 .IP
 749 .I ee_errno
 750 contains the
 751 .I errno
 752 number of the queued error.
 753 .I ee_origin
 754 is the origin code of where the error originated.
 755 The other fields are protocol-specific.
 756 The macro
 757 .B SO_EE_OFFENDER
 758 returns a pointer to the address of the network object
 759 where the error originated from given a pointer to the ancillary message.
 760 If this address is not known, the
 761 .I sa_family
 762 member of the
 763 .I sockaddr
 764 contains
 765 .B AF_UNSPEC
 766 and the other fields of the
 767 .I sockaddr
 768 are undefined.
 769 .IP
 770 IP uses the
 771 .I sock_extended_err
 772 structure as follows:
 773 .I ee_origin
 774 is set to
 775 .B SO_EE_ORIGIN_ICMP
 776 for errors received as an ICMP packet, or
 777 .B SO_EE_ORIGIN_LOCAL
 778 for locally generated errors.
 779 Unknown values should be ignored.
 780 .I ee_type
 781 and
 782 .I ee_code
 783 are set from the type and code fields of the ICMP header.
 784 .I ee_info
 785 contains the discovered MTU for
 786 .B EMSGSIZE
 787 errors.
 788 The message also contains the
 789 .I sockaddr_in of the node
 790 caused the error, which can be accessed with the
 791 .B SO_EE_OFFENDER
 792 macro.
 793 The
 794 .I sin_family
 795 field of the
 796 .B SO_EE_OFFENDER
 797 address is
 798 .B AF_UNSPEC
 799 when the source was unknown.
 800 When the error originated from the network, all IP options
 801 .RB ( IP_OPTIONS ", " IP_TTL ", "
 802 etc.) enabled on the socket and contained in the
 803 error packet are passed as control messages.
 804 The payload of the packet causing the error is returned as normal payload.
 805 .\" FIXME . Is it a good idea to document that? It is a dubious feature.
 806 .\" On
 807 .\" .B SOCK_STREAM
 808 .\" sockets,
 809 .\" .B IP_RECVERR
 810 .\" has slightly different semantics. Instead of
 811 .\" saving the errors for the next timeout, it passes all incoming
 812 .\" errors immediately to the user.
 813 .\" This might be useful for very short-lived TCP connections which
 814 .\" need fast error handling. Use this option with care:
 815 .\" it makes TCP unreliable
 816 .\" by not allowing it to recover properly from routing
 817 .\" shifts and other normal
 818 .\" conditions and breaks the protocol specification.
 819 Note that TCP has no error queue;
 820 .B MSG_ERRQUEUE
 821 is not permitted on
 822 .B SOCK_STREAM
 823 sockets.
 824 .B IP_RECVERR
 825 is valid for TCP, but all errors are returned by socket function return or
 826 .B SO_ERROR
 827 only.
 828 .IP
 829 For raw sockets,
 830 .B IP_RECVERR
 831 enables passing of all received ICMP errors to the
 832 application, otherwise errors are only reported on connected sockets
 833 .IP
 834 It sets or retrieves an integer boolean flag.
 835 .B IP_RECVERR
 836 defaults to off.
 837 .TP
 838 .BR IP_RECVOPTS " (since Linux 2.2)"
 839 .\" Precisely: 2.1.15
 840 Pass all incoming IP options to the user in a
 841 .B IP_OPTIONS
 842 control message.
 843 The routing header and other options are already filled in
 844 for the local host.
 845 Not supported for
 846 .B SOCK_STREAM
 847 sockets.
 848 .TP
 849 .BR IP_RECVORIGDSTADDR " (since Linux 2.6.29)"
 850 .\" commit e8b2dfe9b4501ed0047459b2756ba26e5a940a69
 851 This boolean option enables the
 852 .B IP_ORIGDSTADDR
 853 ancillary message in
 854 .BR recvmsg (2),
 855 in which the kernel returns the original destination address
 856 of the datagram being received.
 857 The ancillary message contains a
 858 .IR "struct sockaddr_in" .
 859 .TP
 860 .BR IP_RECVTOS " (since Linux 2.2)"
 861 .\" Precisely: 2.1.68
 862 If enabled, the
 863 .B IP_TOS
 864 ancillary message is passed with incoming packets.
 865 It contains a byte which specifies the Type of Service/Precedence
 866 field of the packet header.
 867 Expects a boolean integer flag.
 868 .TP
 869 .BR IP_RECVTTL " (since Linux 2.2)"
 870 .\" Precisely: 2.1.68
 871 When this flag is set, pass a
 872 .B IP_TTL
 873 control message with the time-to-live
 874 field of the received packet as a byte.
 875 Not supported for
 876 .B SOCK_STREAM
 877 sockets.
 878 .TP
 879 .BR IP_RETOPTS " (since Linux 2.2)"
 880 .\" Precisely: 2.1.15
 881 Identical to
 882 .BR IP_RECVOPTS ,
 883 but returns raw unprocessed options with timestamp and route record
 884 options not filled in for this hop.
 885 .TP
 886 .BR IP_ROUTER_ALERT " (since Linux 2.2)"
 887 .\" Precisely: 2.1.68
 888 Pass all to-be forwarded packets with the
 889 IP Router Alert option set to this socket.
 890 Valid only for raw sockets.
 891 This is useful, for instance, for user-space RSVP daemons.
 892 The tapped packets are not forwarded by the kernel; it is
 893 the user's responsibility to send them out again.
 894 Socket binding is ignored,
 895 such packets are only filtered by protocol.
 896 Expects an integer flag.
 897 .TP
 898 .BR IP_TOS " (since Linux 1.0)"
 899 Set or receive the Type-Of-Service (TOS) field that is sent
 900 with every IP packet originating from this socket.
 901 It is used to prioritize packets on the network.
 902 TOS is a byte.
 903 There are some standard TOS flags defined:
 904 .B IPTOS_LOWDELAY
 905 to minimize delays for interactive traffic,
 906 .B IPTOS_THROUGHPUT
 907 to optimize throughput,
 908 .B IPTOS_RELIABILITY
 909 to optimize for reliability,
 910 .B IPTOS_MINCOST
 911 should be used for "filler data" where slow transmission doesn't matter.
 912 At most one of these TOS values can be specified.
 913 Other bits are invalid and shall be cleared.
 914 Linux sends
 915 .B IPTOS_LOWDELAY
 916 datagrams first by default,
 917 but the exact behavior depends on the configured queueing discipline.
 918 .\" FIXME elaborate on this
 919 Some high-priority levels may require superuser privileges (the
 920 .B CAP_NET_ADMIN
 921 capability).
 922 .\" The priority can also be set in a protocol-independent way by the
 923 .\" .RB ( SOL_SOCKET ", " SO_PRIORITY )
 924 .\" socket option (see
 925 .\" .BR socket (7)).
 926 .TP
 927 .BR IP_TRANSPARENT " (since Linux 2.6.24)"
 928 .\" commit f5715aea4564f233767ea1d944b2637a5fd7cd2e
 929 .\"     This patch introduces the IP_TRANSPARENT socket option: enabling that
 930 .\"     will make the IPv4 routing omit the non-local source address check on
 931 .\"     output. Setting IP_TRANSPARENT requires NET_ADMIN capability.
 932 .\" http://lwn.net/Articles/252545/
 933 Setting this boolean option enables transparent proxying on this socket.
 934 This socket option allows
 935 the calling application to bind to a nonlocal IP address and operate
 936 both as a client and a server with the foreign address as the local endpoint.
 937 NOTE: this requires that routing be set up in a way that
 938 packets going to the foreign address are routed through the TProxy box
 939 (i.e., the system hosting the application that employs the
 940 .B IP_TRANSPARENT
 941 socket option).
 942 Enabling this socket option requires superuser privileges
 943 (the
 944 .BR CAP_NET_ADMIN
 945 capability).
 946 .IP
 947 TProxy redirection with the iptables TPROXY target also requires that
 948 this option be set on the redirected socket.
 949 .TP
 950 .BR IP_TTL " (since Linux 1.0)"
 951 Set or retrieve the current time-to-live field that is used in every packet
 952 sent from this socket.
 953 .TP
 954 .BR IP_UNBLOCK_SOURCE " (since Linux 2.4.22 / 2.5.68)"
 955 Unblock previously blocked multicast source.
 956 Returns
 957 .BR EADDRNOTAVAIL
 958 when given source is not being blocked.
 959 .IP
 960 Argument is an
 961 .I ip_mreq_source
 962 structure as described under
 963 .BR IP_ADD_SOURCE_MEMBERSHIP .
 964 .SS /proc interfaces
 965 The IP protocol
 966 supports a set of
 967 .I /proc
 968 interfaces to configure some global parameters.
 969 The parameters can be accessed by reading or writing files in the directory
 970 .IR /proc/sys/net/ipv4/ .
 971 .\" FIXME As at 2.6.12, 14 Jun 2005, the following are undocumented:
 972 .\"  ip_queue_maxlen
 973 .\"  ip_conntrack_max
 974 Interfaces described as
 975 .I Boolean
 976 take an integer value, with a nonzero value ("true") meaning that
 977 the corresponding option is enabled, and a zero value ("false")
 978 meaning that the option is disabled.
 979 .\"
 980 .TP
 981 .IR ip_always_defrag " (Boolean; since Linux 2.2.13)"
 982 [New with kernel 2.2.13; in earlier kernel versions this feature
 983 was controlled at compile time by the
 984 .B CONFIG_IP_ALWAYS_DEFRAG
 985 option; this option is not present in 2.4.x and later]
 986 .IP
 987 When this boolean flag is enabled (not equal 0), incoming fragments
 988 (parts of IP packets
 989 that arose when some host between origin and destination decided
 990 that the packets were too large and cut them into pieces) will be
 991 reassembled (defragmented) before being processed, even if they are
 992 about to be forwarded.
 993 .IP
 994 Enable only if running either a firewall that is the sole link
 995 to your network or a transparent proxy; never ever use it for a
 996 normal router or host.
 997 Otherwise, fragmented communication can be disturbed
 998 if the fragments travel over different links.
 999 Defragmentation also has a large memory and CPU time cost.
1000 .IP
1001 This is automagically turned on when masquerading or transparent
1002 proxying are configured.
1003 .\"
1004 .TP
1005 .IR ip_autoconfig " (since Linux 2.2 to 2.6.17)"
1006 .\" Precisely: since 2.1.68
1007 .\" FIXME document ip_autoconfig
1008 Not documented.
1009 .\"
1010 .TP
1011 .IR ip_default_ttl " (integer; default: 64; since Linux 2.2)"
1012 .\" Precisely: 2.1.15
1013 Set the default time-to-live value of outgoing packets.
1014 This can be changed per socket with the
1015 .B IP_TTL
1016 option.
1017 .\"
1018 .TP
1019 .IR ip_dynaddr " (Boolean; default: disabled; since Linux 2.0.31)"
1020 Enable dynamic socket address and masquerading entry rewriting on interface
1021 address change.
1022 This is useful for dialup interface with changing IP addresses.
1023 0 means no rewriting, 1 turns it on and 2 enables verbose mode.
1024 .\"
1025 .TP
1026 .IR ip_forward " (Boolean; default: disabled; since Linux 1.2)"
1027 Enable IP forwarding with a boolean flag.
1028 IP forwarding can be also set on a per-interface basis.
1029 .\"
1030 .TP
1031 .IR ip_local_port_range " (since Linux 2.2)"
1032 .\" Precisely: since 2.1.68
1033 This file contains two integers that define the default local port range
1034 allocated to sockets that are not explicitly bound to a port number\(emthat
1035 is, the range used for
1036 .IR "ephemeral ports" .
1037 An ephemeral port is allocated to a socket in the following circumstances:
1038 .RS
1039 .IP * 3
1040 the port number in a socket address is specified as 0 when calling
1041 .BR bind (2);
1042 .IP *
1043 .BR listen (2)
1044 is called on a stream socket that was not previously bound;
1045 .IP *
1046 .BR connect (2)
1047 was called on a socket that was not previously bound;
1048 .IP *
1049 .BR sendto (2)
1050 is called on a datagram socket that was not previously bound.
1051 .RE
1052 .IP
1053 Allocation of ephemeral ports starts with the first number in
1054 .IR ip_local_port_range
1055 and ends with the second number.
1056 If the range of ephemeral ports is exhausted,
1057 then the relevant system call returns an error (but see BUGS).
1058 .IP
1059 Note that the port range in
1060 .IR ip_local_port_range
1061 should not conflict with the ports used by masquerading
1062 (although the case is handled).
1063 Also, arbitrary choices may cause problems with some firewall packet
1064 filters that make assumptions about the local ports in use.
1065 The first number should be at least greater than 1024,
1066 or better, greater than 4096, to avoid clashes
1067 with well known ports and to minimize firewall problems.
1068 .\"
1069 .TP
1070 .IR ip_no_pmtu_disc " (Boolean; default: disabled; since Linux 2.2)"
1071 .\" Precisely: 2.1.15
1072 If enabled, don't do Path MTU Discovery for TCP sockets by default.
1073 Path MTU discovery may fail if misconfigured firewalls (that drop
1074 all ICMP packets) or misconfigured interfaces (e.g., a point-to-point
1075 link where the both ends don't agree on the MTU) are on the path.
1076 It is better to fix the broken routers on the path than to turn off
1077 Path MTU Discovery globally, because not doing it incurs a high cost
1078 to the network.
1079 .\"
1080 .\" The following is from 2.6.12: Documentation/networking/ip-sysctl.txt
1081 .TP
1082 .IR ip_nonlocal_bind " (Boolean; default: disabled; since Linux 2.4)"
1083 .\" Precisely: patch-2.4.0-test10
1084 If set, allows processes to
1085 .BR bind (2)
1086 to nonlocal IP addresses,
1087 which can be quite useful, but may break some applications.
1088 .\"
1089 .\" The following is from 2.6.12: Documentation/networking/ip-sysctl.txt
1090 .TP
1091 .IR ip6frag_time " (integer; default: 30)"
1092 Time in seconds to keep an IPv6 fragment in memory.
1093 .\"
1094 .\" The following is from 2.6.12: Documentation/networking/ip-sysctl.txt
1095 .TP
1096 .IR ip6frag_secret_interval " (integer; default: 600)"
1097 Regeneration interval (in seconds) of the hash secret (or lifetime
1098 for the hash secret) for IPv6 fragments.
1099 .TP
1100 .IR ipfrag_high_thresh " (integer), " ipfrag_low_thresh " (integer)"
1101 If the amount of queued IP fragments reaches
1102 .IR ipfrag_high_thresh ,
1103 the queue is pruned down to
1104 .IR ipfrag_low_thresh .
1105 Contains an integer with the number of bytes.
1106 .TP
1107 .I neigh/*
1108 See
1109 .BR arp (7).
1110 .\" FIXME Document the conf/*/* interfaces
1111 .\"
1112 .\" FIXME Document the route/* interfaces
1113 .SS Ioctls
1114 All ioctls described in
1115 .BR socket (7)
1116 apply to
1117 .BR ip .
1118 .\" 2006-04-02, mtk
1119 .\" commented out the following because ipchains is obsolete
1120 .\" .PP
1121 .\" The ioctls to configure firewalling are documented in
1122 .\" .BR ipfw (4)
1123 .\" from the
1124 .\" .B ipchains
1125 .\" package.
1126 .PP
1127 Ioctls to configure generic device parameters are described in
1128 .BR netdevice (7).
1129 .\" FIXME Add a discussion of multicasting
1130 .SH ERRORS
1131 .\" FIXME document all errors.
1132 .\"     We should really fix the kernels to give more uniform
1133 .\"     error returns (ENOMEM vs ENOBUFS, EPERM vs EACCES etc.)
1134 .TP
1135 .B EACCES
1136 The user tried to execute an operation without the necessary permissions.
1137 These include:
1138 sending a packet to a broadcast address without having the
1139 .B SO_BROADCAST
1140 flag set;
1141 sending a packet via a
1142 .I prohibit
1143 route;
1144 modifying firewall settings without superuser privileges (the
1145 .B CAP_NET_ADMIN
1146 capability);
1147 binding to a privileged port without superuser privileges (the
1148 .B CAP_NET_BIND_SERVICE
1149 capability).
1150 .TP
1151 .B EADDRINUSE
1152 Tried to bind to an address already in use.
1153 .TP
1154 .B EADDRNOTAVAIL
1155 A nonexistent interface was requested or the requested source
1156 address was not local.
1157 .TP
1158 .B EAGAIN
1159 Operation on a nonblocking socket would block.
1160 .TP
1161 .B EALREADY
1162 A connection operation on a nonblocking socket is already in progress.
1163 .TP
1164 .B ECONNABORTED
1165 A connection was closed during an
1166 .BR accept (2).
1167 .TP
1168 .B EHOSTUNREACH
1169 No valid routing table entry matches the destination address.
1170 This error can be caused by an ICMP message from a remote router or
1171 for the local routing table.
1172 .TP
1173 .B EINVAL
1174 Invalid argument passed.
1175 For send operations this can be caused by sending to a
1176 .I blackhole
1177 route.
1178 .TP
1179 .B EISCONN
1180 .BR connect (2)
1181 was called on an already connected socket.
1182 .TP
1183 .B EMSGSIZE
1184 Datagram is bigger than an MTU on the path and it cannot be fragmented.
1185 .TP
1186 .BR ENOBUFS ", " ENOMEM
1187 Not enough free memory.
1188 This often means that the memory allocation is limited by the socket
1189 buffer limits, not by the system memory, but this is not 100% consistent.
1190 .TP
1191 .B ENOENT
1192 .B SIOCGSTAMP
1193 was called on a socket where no packet arrived.
1194 .TP
1195 .B ENOPKG
1196 A kernel subsystem was not configured.
1197 .TP
1198 .BR ENOPROTOOPT " and " EOPNOTSUPP
1199 Invalid socket option passed.
1200 .TP
1201 .B ENOTCONN
1202 The operation is defined only on a connected socket, but the socket wasn't
1203 connected.
1204 .TP
1205 .B EPERM
1206 User doesn't have permission to set high priority, change configuration,
1207 or send signals to the requested process or group.
1208 .TP
1209 .B EPIPE
1210 The connection was unexpectedly closed or shut down by the other end.
1211 .TP
1212 .B ESOCKTNOSUPPORT
1213 The socket is not configured or an unknown socket type was requested.
1214 .PP
1215 Other errors may be generated by the overlaying protocols; see
1216 .BR tcp (7),
1217 .BR raw (7),
1218 .BR udp (7),
1219 and
1220 .BR socket (7).
1221 .SH NOTES
1222 .BR IP_FREEBIND ,
1223 .BR IP_MSFILTER ,
1224 .BR IP_MTU ,
1225 .BR IP_MTU_DISCOVER ,
1226 .BR IP_RECVORIGDSTADDR ,
1227 .BR IP_PKTINFO ,
1228 .BR IP_RECVERR ,
1229 .BR IP_ROUTER_ALERT ,
1230 and
1231 .BR IP_TRANSPARENT
1232 are Linux-specific.
1233 .\" IP_PASSSEC is Linux-specific
1234 .\" IP_XFRM_POLICY is Linux-specific
1235 .\" IP_IPSEC_POLICY is a nonstandard extension, also present on some BSDs
1236 .PP
1237 Be very careful with the
1238 .B SO_BROADCAST
1239 option \- it is not privileged in Linux.
1240 It is easy to overload the network
1241 with careless broadcasts.
1242 For new application protocols
1243 it is better to use a multicast group instead of broadcasting.
1244 Broadcasting is discouraged.
1245 .PP
1246 Some other BSD sockets implementations provide
1247 .B IP_RCVDSTADDR
1248 and
1249 .B IP_RECVIF
1250 socket options to get the destination address and the interface of
1251 received datagrams.
1252 Linux has the more general
1253 .B IP_PKTINFO
1254 for the same task.
1255 .PP
1256 Some BSD sockets implementations also provide an
1257 .B IP_RECVTTL
1258 option, but an ancillary message with type
1259 .B IP_RECVTTL
1260 is passed with the incoming packet.
1261 This is different from the
1262 .B IP_TTL
1263 option used in Linux.
1264 .PP
1265 Using the
1266 .B SOL_IP
1267 socket options level isn't portable; BSD-based stacks use the
1268 .B IPPROTO_IP
1269 level.
1270 .SS Compatibility
1271 For compatibility with Linux 2.0, the obsolete
1272 .BI "socket(AF_INET, SOCK_PACKET, " protocol )
1273 syntax is still supported to open a
1274 .BR packet (7)
1275 socket.
1276 This is deprecated and should be replaced by
1277 .BI "socket(AF_PACKET, SOCK_RAW, " protocol )
1278 instead.
1279 The main difference is the new
1280 .I sockaddr_ll
1281 address structure for generic link layer information instead of the old
1282 .BR sockaddr_pkt .
1283 .SH BUGS
1284 There are too many inconsistent error values.
1285 .PP
1286 The error used to diagnose exhaustion of the ephemeral port range differs
1287 across the various system calls
1288 .RB ( connect (2),
1289 .BR bind (2),
1290 .BR listen (2),
1291 .BR sendto (2))
1292 that can assign ephemeral ports.
1293 .PP
1294 The ioctls to configure IP-specific interface options and ARP tables are
1295 not described.
1296 .\" .PP
1297 .\" Some versions of glibc forget to declare
1298 .\" .IR in_pktinfo .
1299 .\" Workaround currently is to copy it into your program from this man page.
1300 .PP
1301 Receiving the original destination address with
1302 .B MSG_ERRQUEUE
1303 in
1304 .I msg_name
1305 by
1306 .BR recvmsg (2)
1307 does not work in some 2.2 kernels.
1308 .\" .SH AUTHORS
1309 .\" This man page was written by Andi Kleen.
1310 .SH SEE ALSO
1311 .BR recvmsg (2),
1312 .BR sendmsg (2),
1313 .BR byteorder (3),
1314 .BR ipfw (4),
1315 .BR capabilities (7),
1316 .BR icmp (7),
1317 .BR ipv6 (7),
1318 .BR netlink (7),
1319 .BR raw (7),
1320 .BR socket (7),
1321 .BR tcp (7),
1322 .BR udp (7),
1323 .BR ip (8)
1324 .PP
1325 RFC\ 791 for the original IP specification.
1326 RFC\ 1122 for the IPv4 host requirements.
1327 RFC\ 1812 for the IPv4 router requirements.