2 .\" Don't change the line above. it tells man that tbl is needed.
3 .\" This man page is Copyright (C) 1999 Andi Kleen <ak@muc.de>.
4 .\" Permission is granted to distribute possibly modified copies
5 .\" of this page provided the header is included verbatim,
6 .\" and in case of nontrivial modification author and date
7 .\" of the modification is added to the header.
8 .\" $Id: ip.7,v 1.19 2000/12/20 18:10:31 ak Exp $
10 .\" FIXME: Document IP_ORIGDSTADDR+IP_RECVORIGDSTADDR, added in Linux 2.6.29
11 .\" FIXME: Document IP_MINTTL, added in Linux 2.6.34
13 .TH IP 7 2010-09-11 "Linux" "Linux Programmer's Manual"
15 ip \- Linux IPv4 protocol implementation
17 .B #include <sys/socket.h>
19 .\" .B #include <net/netinet.h> -- does not exist anymore
20 .\" .B #include <linux/errqueue.h> -- never include <linux/foo.h>
21 .B #include <netinet/in.h>
23 .B #include <netinet/ip.h> \fR/* superset of previous */
25 .IB tcp_socket " = socket(AF_INET, SOCK_STREAM, 0);"
27 .IB udp_socket " = socket(AF_INET, SOCK_DGRAM, 0);"
29 .IB raw_socket " = socket(AF_INET, SOCK_RAW, " protocol ");"
31 Linux implements the Internet Protocol, version 4,
32 described in RFC\ 791 and RFC\ 1122.
34 contains a level 2 multicasting implementation conforming to RFC\ 1112.
35 It also contains an IP router including a packet filter.
36 .\" FIXME has someone verified that 2.1 is really 1812 compliant?
38 The programming interface is BSD-sockets compatible.
39 For more information on sockets, see
42 An IP socket is created by calling the
45 .BR "socket(AF_INET, socket_type, protocol)" .
46 Valid socket types are
58 socket to access the IP protocol directly.
60 is the IP protocol in the IP header to be received or sent.
61 The only valid values for
65 for TCP sockets, and 0 and
70 you may specify a valid IANA IP protocol defined in
71 RFC\ 1700 assigned numbers.
73 .\" FIXME ip current does an autobind in listen, but I'm not sure
74 .\" if that should be documented.
75 When a process wants to receive new incoming packets or connections, it
76 should bind a socket to a local interface address using
78 Only one IP socket may be bound to any given local (address, port) pair.
81 is specified in the bind call, the socket will be bound to
88 are called on an unbound socket, it is automatically bound to a
89 random free port with the local address set to
92 A TCP local socket address that has been bound is unavailable for
93 some time after closing, unless the
96 Care should be taken when using this flag as it makes TCP less reliable.
98 An IP socket address is defined as a combination of an IP interface
99 address and a 16-bit port number.
100 The basic IP protocol does not supply port numbers, they
101 are implemented by higher level protocols like
107 is set to the IP protocol.
112 sa_family_t sin_family; /* address family: AF_INET */
113 in_port_t sin_port; /* port in network byte order */
114 struct in_addr sin_addr; /* internet address */
117 /* Internet address. */
119 uint32_t s_addr; /* address in network byte order */
127 This is required; in Linux 2.2 most networking functions return
129 when this setting is missing.
131 contains the port in network byte order.
132 The port numbers below 1024 are called
133 .IR "privileged ports"
135 .IR "reserved ports" ).
136 Only privileged processes (i.e., those having the
137 .B CAP_NET_BIND_SERVICE
141 Note that the raw IPv4 protocol as such has no concept of a
142 port, they are only implemented by higher protocols like
148 is the IP host address.
153 contains the host interface address in network byte order.
155 should be assigned one of the INADDR_* values (e.g.,
160 .BR inet_makeaddr (3)
161 library functions or directly with the name resolver (see
162 .BR gethostbyname (3)).
164 IPv4 addresses are divided into unicast, broadcast
165 and multicast addresses.
166 Unicast addresses specify a single interface of a host,
167 broadcast addresses specify all hosts on a network and multicast
168 addresses address all hosts in a multicast group.
169 Datagrams to broadcast addresses can be only sent or received when the
172 In the current implementation, connection-oriented sockets are only allowed
173 to use unicast addresses.
174 .\" Leave a loophole for XTP @)
176 Note that the address and the port are always stored in
178 In particular, this means that you need to call
180 on the number that is assigned to a port.
181 All address/port manipulation
182 functions in the standard library work in network byte order.
184 There are several special addresses:
187 always refers to the local host via the loopback device;
190 means any address for binding;
193 means any host and has the same effect on bind as
195 for historical reasons.
197 IP supports some protocol-specific socket options that can be set with
201 The socket option level for IP is
203 .\" or SOL_IP on Linux
204 A boolean integer flag is zero when it is false, otherwise true.
206 .BR IP_ADD_MEMBERSHIP " (since Linux 1.2)"
207 Join a multicast group.
215 struct in_addr imr_multiaddr; /* IP multicast group
217 struct in_addr imr_address; /* IP address of local
219 int imr_ifindex; /* interface index */
225 contains the address of the multicast group the application
226 wants to join or leave.
227 It must be a valid multicast address
228 .\" (i.e., within the 224.0.0.0-239.255.255.255 range)
234 is the address of the local interface with which the system
235 should join the multicast group; if it is equal to
237 an appropriate interface is chosen by the system.
239 is the interface index of the interface that should join/leave the
241 group, or 0 to indicate any interface.
245 structure is available only since Linux 2.2.
246 For compatibility, the old
248 structure (present since Linux 1.2) is still supported;
251 only by not including the
258 .BR IP_DROP_MEMBERSHIP " (since Linux 1.2)"
259 Leave a multicast group.
265 .BR IP_ADD_MEMBERSHIP .
267 .BR IP_HDRINCL " (since Linux 2.0)"
269 the user supplies an IP header in front of the user data.
275 for more information.
276 When this flag is enabled the values set by
283 .BR IP_FREEBIND " (since Linux 2.4)"
284 .\" Precisely: 2.4.0-test10
285 If enabled, this boolean option allows binding to an IP address
286 that is nonlocal or does not (yet) exist.
287 This permits listening on a socket,
288 without requiring the underlying network interface or the
289 specified dynamic IP address to be up at the time that
290 the application is trying to bind to it.
291 This option is the per-socket equivalent of the
294 interface described below.
296 .\" FIXME Document IP_IPSEC_POLICY
297 .\" Since Linux 2.5.47
298 .\" Needs CAP_NET_ADMIN
300 .BR IP_MTU " (since Linux 2.2)"
301 .\" Precisely: 2.1.124
302 Retrieve the current known path MTU of the current socket.
303 Only valid when the socket has been connected.
308 .BR IP_MTU_DISCOVER " (since Linux 2.2)"
309 .\" Precisely: 2.1.124
310 Set or receive the Path MTU Discovery setting for a socket.
311 When enabled, Linux will perform Path MTU Discovery
312 as defined in RFC\ 1191
314 The don't-fragment flag is set on all outgoing datagrams.
315 The system-wide default is controlled by the
316 .I /proc/sys/net/ipv4/ip_no_pmtu_disc
319 sockets, and disabled on all others.
322 sockets, it is the user's responsibility to packetize the data
323 in MTU sized chunks and to do the retransmits if necessary.
324 The kernel will reject packets that are bigger than the known
325 path MTU if this flag is set (with
332 Path MTU discovery flags:Meaning
333 IP_PMTUDISC_WANT:Use per-route settings.
334 IP_PMTUDISC_DONT:Never do Path MTU Discovery.
335 IP_PMTUDISC_DO:Always do Path MTU Discovery.
336 IP_PMTUDISC_PROBE:Set DF but ignore Path MTU.
339 When PMTU discovery is enabled, the kernel automatically keeps track of
340 the path MTU per destination host.
341 When it is connected to a specific peer with
343 the currently known path MTU can be retrieved conveniently using the
345 socket option (e.g., after a
348 It may change over time.
349 For connectionless sockets with many destinations,
350 the new MTU for a given destination can also be accessed using the
353 A new error will be queued for every incoming MTU update.
355 While MTU discovery is in progress, initial packets from datagram sockets
357 Applications using UDP should be aware of this and not
358 take it into account for their packet retransmit strategy.
360 To bootstrap the path MTU discovery process on unconnected sockets, it
361 is possible to start with a big datagram size
362 (up to 64K-headers bytes long) and let it shrink by updates of the path MTU.
363 .\" FIXME this is an ugly hack
365 To get an initial estimate of the
366 path MTU, connect a datagram socket to the destination address using
368 and retrieve the MTU by calling
374 It is possible to implement RFC 4821 MTU probing with
378 sockets by setting a value of
379 .BR IP_PMTUDISC_PROBE
380 (available since Linux 2.6.22).
381 This is also particularly useful for diagnostic tools such as
383 that wish to deliberately send probe packets larger than
384 the observed Path MTU.
386 .BR IP_MULTICAST_IF " (since Linux 1.2)"
387 Set the local device for a multicast socket.
393 .BR IP_ADD_MEMBERSHIP .
395 When an invalid socket option is passed,
399 .BR IP_MULTICAST_LOOP " (since Linux 1.2)"
400 Set or read a boolean integer argument that determines whether
401 sent multicast packets should be looped back to the local sockets.
403 .BR IP_MULTICAST_TTL " (since Linux 1.2)"
404 Set or read the time-to-live value of outgoing multicast packets for this
406 It is very important for multicast packets to set the smallest TTL possible.
407 The default is 1 which means that multicast packets don't leave the local
408 network unless the user program explicitly requests it.
409 Argument is an integer.
411 .BR IP_NODEFRAG " (since Linux 2.6.36)"
412 If enabled (argument is nonzero),
413 the reassembly of outgoing packets is disabled in the netfilter layer.
414 This option is only valid for
417 The argument is an integer.
419 .BR IP_OPTIONS " (since Linux 2.0)"
420 .\" Precisely: 1.3.30
421 Set or get the IP options to be sent with every packet from this socket.
422 The arguments are a pointer to a memory buffer containing the options
423 and the option length.
426 call sets the IP options associated with a socket.
427 The maximum option size for IPv4 is 40 bytes.
428 See RFC\ 791 for the allowed options.
429 When the initial connection request packet for a
431 socket contains IP options, the IP options will be set automatically
432 to the options from the initial packet with routing headers reversed.
433 Incoming packets are not allowed to change options after the connection
435 The processing of all incoming source routing options
436 is disabled by default and can be enabled by using the
437 .I accept_source_route
440 Other options like timestamps are still handled.
441 For datagram sockets, IP options can be only set by the local user.
446 puts the current IP options used for sending into the supplied buffer.
447 .\" FIXME Document IP_PASSSEC
449 .\" Since Linux 2.6.17
450 .\" commit 2c7946a7bf45ae86736ab3b43d0085e43947945c
451 .\" Author: Catherine Zhang <cxzhang@watson.ibm.com>
453 .BR IP_PKTINFO " (since Linux 2.2)"
454 .\" Precisely: 2.1.68
457 ancillary message that contains a
459 structure that supplies some information about the incoming packet.
460 This only works for datagram oriented sockets.
461 The argument is a flag that tells the socket whether the
463 message should be passed or not.
464 The message itself can only be sent/retrieved
465 as control message with a packet using
473 unsigned int ipi_ifindex; /* Interface index */
474 struct in_addr ipi_spec_dst; /* Local address */
475 struct in_addr ipi_addr; /* Header Destination
481 .\" FIXME elaborate on that.
483 is the unique index of the interface the packet was received on.
485 is the local address of the packet and
487 is the destination address in the packet header.
493 .\" This field is grossly misnamed
495 is not zero, then it is used as the local source address for the routing
496 table lookup and for setting up IP source route options.
499 is not zero, the primary local address of the interface specified by the
502 for the routing table lookup.
504 .BR IP_RECVERR " (since Linux 2.2)"
505 .\" Precisely: 2.1.15
506 Enable extended reliable error message passing.
507 When enabled on a datagram socket, all
508 generated errors will be queued in a per-socket error queue.
509 When the user receives an error from a socket operation,
510 the errors can be received by calling
517 structure describing the error will be passed in an ancillary message with
522 .\" or SOL_IP on Linux
523 This is useful for reliable error handling on unconnected sockets.
524 The received data portion of the error queue contains the error packet.
528 control message contains a
535 #define SO_EE_ORIGIN_NONE 0
536 #define SO_EE_ORIGIN_LOCAL 1
537 #define SO_EE_ORIGIN_ICMP 2
538 #define SO_EE_ORIGIN_ICMP6 3
540 struct sock_extended_err {
541 uint32_t ee_errno; /* error number */
542 uint8_t ee_origin; /* where the error originated */
543 uint8_t ee_type; /* type */
544 uint8_t ee_code; /* code */
546 uint32_t ee_info; /* additional information */
547 uint32_t ee_data; /* other data */
548 /* More data may follow */
551 struct sockaddr *SO_EE_OFFENDER(struct sock_extended_err *);
558 number of the queued error.
560 is the origin code of where the error originated.
561 The other fields are protocol-specific.
564 returns a pointer to the address of the network object
565 where the error originated from given a pointer to the ancillary message.
566 If this address is not known, the
572 and the other fields of the
578 structure as follows:
582 for errors received as an ICMP packet, or
583 .B SO_EE_ORIGIN_LOCAL
584 for locally generated errors.
585 Unknown values should be ignored.
589 are set from the type and code fields of the ICMP header.
591 contains the discovered MTU for
594 The message also contains the
595 .I sockaddr_in of the node
596 caused the error, which can be accessed with the
601 field of the SO_EE_OFFENDER address is
603 when the source was unknown.
604 When the error originated from the network, all IP options
605 .RI ( IP_OPTIONS ", " IP_TTL ", "
606 etc.) enabled on the socket and contained in the
607 error packet are passed as control messages.
608 The payload of the packet causing the error is returned as normal payload.
609 .\" FIXME . Is it a good idea to document that? It is a dubious feature.
614 .\" has slightly different semantics. Instead of
615 .\" saving the errors for the next timeout, it passes all incoming
616 .\" errors immediately to the user.
617 .\" This might be useful for very short-lived TCP connections which
618 .\" need fast error handling. Use this option with care:
619 .\" it makes TCP unreliable
620 .\" by not allowing it to recover properly from routing
621 .\" shifts and other normal
622 .\" conditions and breaks the protocol specification.
623 Note that TCP has no error queue;
629 is valid for TCP, but all errors are returned by socket function return or
635 enables passing of all received ICMP errors to the
636 application, otherwise errors are only reported on connected sockets
638 It sets or retrieves an integer boolean flag.
642 .BR IP_RECVOPTS " (since Linux 2.2)"
643 .\" Precisely: 2.1.15
644 Pass all incoming IP options to the user in a
647 The routing header and other options are already filled in
653 .BR IP_RECVTOS " (since Linux 2.2)"
654 .\" Precisely: 2.1.68
657 ancillary message is passed with incoming packets.
658 It contains a byte which specifies the Type of Service/Precedence
659 field of the packet header.
660 Expects a boolean integer flag.
662 .BR IP_RECVTTL " (since Linux 2.2)"
663 .\" Precisely: 2.1.68
664 When this flag is set, pass a
666 control message with the time to live
667 field of the received packet as a byte.
672 .BR IP_RETOPTS " (since Linux 2.2)"
673 .\" Precisely: 2.1.15
676 but returns raw unprocessed options with timestamp and route record
677 options not filled in for this hop.
679 .BR IP_ROUTER_ALERT " (since Linux 2.2)"
680 .\" Precisely: 2.1.68
681 Pass all to-be forwarded packets with the
682 IP Router Alert option set to this socket.
683 Only valid for raw sockets.
684 This is useful, for instance, for user-space RSVP daemons.
685 The tapped packets are not forwarded by the kernel; it is
686 the user's responsibility to send them out again.
687 Socket binding is ignored,
688 such packets are only filtered by protocol.
689 Expects an integer flag.
691 .BR IP_TOS " (since Linux 1.0)"
692 Set or receive the Type-Of-Service (TOS) field that is sent
693 with every IP packet originating from this socket.
694 It is used to prioritize packets on the network.
696 There are some standard TOS flags defined:
698 to minimize delays for interactive traffic,
700 to optimize throughput,
702 to optimize for reliability,
704 should be used for "filler data" where slow transmission doesn't matter.
705 At most one of these TOS values can be specified.
706 Other bits are invalid and shall be cleared.
709 datagrams first by default,
710 but the exact behavior depends on the configured queueing discipline.
711 .\" FIXME elaborate on this
712 Some high priority levels may require superuser privileges (the
715 The priority can also be set in a protocol independent way by the
716 .RB ( SOL_SOCKET ", " SO_PRIORITY )
719 .\" FIXME Document IP_TRANSPARENT
720 .\" Needs CAP_NET_ADMIN
722 .\" Since Linux 2.6.27
723 .\" commit f5715aea4564f233767ea1d944b2637a5fd7cd2e
724 .\" Author: KOVACS Krisztian <hidden@sch.bme.hu>
726 .BR IP_TTL " (since Linux 1.0)"
727 Set or retrieve the current time-to-live field that is used in every packet
728 sent from this socket.
729 .\" FIXME Document IP_XFRM_POLICY
730 .\" Since Linux 2.5.48
731 .\" Needs CAP_NET_ADMIN
736 interfaces to configure some global parameters.
737 The parameters can be accessed by reading or writing files in the directory
738 .IR /proc/sys/net/ipv4/ .
739 .\" FIXME As at 2.6.12, 14 Jun 2005, the following are undocumented:
742 Interfaces described as
744 take an integer value, with a nonzero value ("true") meaning that
745 the corresponding option is enabled, and a zero value ("false")
746 meaning that the option is disabled.
749 .IR ip_always_defrag " (Boolean; since Linux 2.2.13)"
750 [New with kernel 2.2.13; in earlier kernel versions this feature
751 was controlled at compile time by the
752 .B CONFIG_IP_ALWAYS_DEFRAG
753 option; this option is not present in 2.4.x and later]
755 When this boolean flag is enabled (not equal 0), incoming fragments
757 that arose when some host between origin and destination decided
758 that the packets were too large and cut them into pieces) will be
759 reassembled (defragmented) before being processed, even if they are
760 about to be forwarded.
762 Only enable if running either a firewall that is the sole link
763 to your network or a transparent proxy; never ever use it for a
764 normal router or host.
765 Otherwise fragmented communication can be disturbed
766 if the fragments travel over different links.
767 Defragmentation also has a large memory and CPU time cost.
769 This is automagically turned on when masquerading or transparent
770 proxying are configured.
773 .IR ip_autoconfig " (since Linux 2.2 to 2.6.17)"
774 .\" Precisely: since 2.1.68
775 .\" FIXME document ip_autoconfig
779 .IR ip_default_ttl " (integer; default: 64; since Linux 2.2)"
780 .\" Precisely: 2.1.15
781 Set the default time-to-live value of outgoing packets.
782 This can be changed per socket with the
787 .IR ip_dynaddr " (Boolean; default: disabled; since Linux 2.0.31)"
788 Enable dynamic socket address and masquerading entry rewriting on interface
790 This is useful for dialup interface with changing IP addresses.
791 0 means no rewriting, 1 turns it on and 2 enables verbose mode.
794 .IR ip_forward " (Boolean; default: disabled; since Linux 1.2)"
795 Enable IP forwarding with a boolean flag.
796 IP forwarding can be also set on a per-interface basis.
799 .IR ip_local_port_range " (since Linux 2.2)"
800 .\" Precisely: since 2.1.68
801 Contains two integers that define the default local port range
802 allocated to sockets.
803 Allocation starts with the first number and ends with the second number.
804 Note that these should not conflict with the ports used by masquerading
805 (although the case is handled).
806 Also arbitrary choices may cause problems with some firewall packet
807 filters that make assumptions about the local ports in use.
808 First number should be at least greater than 1024,
809 or better, greater than 4096, to avoid clashes
810 with well known ports and to minimize firewall problems.
813 .IR ip_no_pmtu_disc " (Boolean; default: disabled; since Linux 2.2)"
814 .\" Precisely: 2.1.15
815 If enabled, don't do Path MTU Discovery for TCP sockets by default.
816 Path MTU discovery may fail if misconfigured firewalls (that drop
817 all ICMP packets) or misconfigured interfaces (e.g., a point-to-point
818 link where the both ends don't agree on the MTU) are on the path.
819 It is better to fix the broken routers on the path than to turn off
820 Path MTU Discovery globally, because not doing it incurs a high cost
823 .\" The following is from 2.6.12: Documentation/networking/ip-sysctl.txt
825 .IR ip_nonlocal_bind " (Boolean; default: disabled; since Linux 2.4)"
826 .\" Precisely: patch-2.4.0-test10
827 If set, allows processes to
829 to nonlocal IP addresses,
830 which can be quite useful, but may break some applications.
832 .\" The following is from 2.6.12: Documentation/networking/ip-sysctl.txt
834 .IR ip6frag_time " (integer; default: 30)"
835 Time in seconds to keep an IPv6 fragment in memory.
837 .\" The following is from 2.6.12: Documentation/networking/ip-sysctl.txt
839 .IR ip6frag_secret_interval " (integer; default: 600)"
840 Regeneration interval (in seconds) of the hash secret (or lifetime
841 for the hash secret) for IPv6 fragments.
843 .IR ipfrag_high_thresh " (integer), " ipfrag_low_thresh " (integer)"
844 If the amount of queued IP fragments reaches
845 .IR ipfrag_high_thresh ,
846 the queue is pruned down to
847 .IR ipfrag_low_thresh .
848 Contains an integer with the number of bytes.
853 .\" FIXME Document the conf/*/* interfaces
854 .\" FIXME Document the route/* interfaces
855 .\" FIXME document them all
857 All ioctls described in
862 .\" commented out the following because ipchains is obsolete
864 .\" The ioctls to configure firewalling are documented in
870 Ioctls to configure generic device parameters are described in
872 .\" FIXME Add a discussion of multicasting
874 .\" FIXME document all errors.
875 .\" We should really fix the kernels to give more uniform
876 .\" error returns (ENOMEM vs ENOBUFS, EPERM vs EACCES etc.)
879 The user tried to execute an operation without the necessary permissions.
881 sending a packet to a broadcast address without having the
884 sending a packet via a
887 modifying firewall settings without superuser privileges (the
890 binding to a privileged port without superuser privileges (the
891 .B CAP_NET_BIND_SERVICE
895 Tried to bind to an address already in use.
898 A nonexistent interface was requested or the requested source
899 address was not local.
902 Operation on a nonblocking socket would block.
905 An connection operation on a nonblocking socket is already in progress.
908 A connection was closed during an
912 No valid routing table entry matches the destination address.
913 This error can be caused by a ICMP message from a remote router or
914 for the local routing table.
917 Invalid argument passed.
918 For send operations this can be caused by sending to a
924 was called on an already connected socket.
927 Datagram is bigger than an MTU on the path and it cannot be fragmented.
929 .BR ENOBUFS ", " ENOMEM
930 Not enough free memory.
931 This often means that the memory allocation is limited by the socket
932 buffer limits, not by the system memory, but this is not 100% consistent.
936 was called on a socket where no packet arrived.
939 A kernel subsystem was not configured.
941 .BR ENOPROTOOPT " and " EOPNOTSUPP
942 Invalid socket option passed.
945 The operation is only defined on a connected socket, but the socket wasn't
949 User doesn't have permission to set high priority, change configuration,
950 or send signals to the requested process or group.
953 The connection was unexpectedly closed or shut down by the other end.
956 The socket is not configured or an unknown socket type was requested.
958 Other errors may be generated by the overlaying protocols; see
967 .BR IP_MTU_DISCOVER ,
973 .\" IP_PASSSEC is Linux-specific
974 .\" IP_XFRM_POLICY is Linux-specific
975 .\" IP_IPSEC_POLICY is a nonstandard extension, also present on some BSDs
976 Be very careful with the
978 option \- it is not privileged in Linux.
979 It is easy to overload the network
980 with careless broadcasts.
981 For new application protocols
982 it is better to use a multicast group instead of broadcasting.
983 Broadcasting is discouraged.
985 Some other BSD sockets implementations provide
989 socket options to get the destination address and the interface of
991 Linux has the more general
995 Some BSD sockets implementations also provide an
997 option, but an ancillary message with type
999 is passed with the incoming packet.
1000 This is different from the
1002 option used in Linux.
1006 socket options level isn't portable, BSD-based stacks use
1010 For compatibility with Linux 2.0, the obsolete
1011 .BI "socket(AF_INET, SOCK_PACKET, " protocol )
1012 syntax is still supported to open a
1015 This is deprecated and should be replaced by
1016 .BI "socket(AF_PACKET, SOCK_RAW, " protocol )
1018 The main difference is the new
1020 address structure for generic link layer information instead of the old
1023 There are too many inconsistent error values.
1025 The ioctls to configure IP-specific interface options and ARP tables are
1028 Some versions of glibc forget to declare
1030 Workaround currently is to copy it into your program from this man page.
1032 Receiving the original destination address with
1038 does not work in some 2.2 kernels.
1040 .\" This man page was written by Andi Kleen.
1046 .BR capabilities (7),
1053 RFC\ 791 for the original IP specification.
1055 RFC\ 1122 for the IPv4 host requirements.
1057 RFC\ 1812 for the IPv4 router requirements.
1058 .\" FIXME autobind INADDR REUSEADDR