Ubuntu Manpages

ip

Linux IPv4 protocol implementation

#include <sys/socket.h>
#include <netinet/in.h>
#include <netinet/ip.h>  /* superset of previous */
tcp_socket = socket(AF_INET, SOCK_STREAM, 0);
udp_socket = socket(AF_INET, SOCK_DGRAM, 0);
raw_socket = socket(AF_INET, SOCK_RAW, protocol);

Linux implements the Internet Protocol, version 4, described in RFC 791 and RFC 1122. ip contains a level 2 multicasting implementation conforming to RFC 1112. It also contains an IP router including a packet filter.

The programming interface is BSD-sockets compatible. For more information on sockets, see socket(7).

An IP socket is created using socket(2):


socket(AF_INET, socket_type, protocol);

Valid socket types include SOCK_STREAM to open a stream socket, SOCK_DGRAM to open a datagram socket, and SOCK_RAW to open a raw(7) socket to access the IP protocol directly.

protocol is the IP protocol in the IP header to be received or sent. Valid values for protocol include:

  • 0 and IPPROTO_TCP for tcp(7) stream sockets;
  • 0 and IPPROTO_UDP for udp(7) datagram sockets;
  • IPPROTO_SCTP for sctp(7) stream sockets; and
  • IPPROTO_UDPLITE for udplite(7) datagram sockets.

For SOCK_RAW you may specify a valid IANA IP protocol defined in RFC 1700 assigned numbers.

When a process wants to receive new incoming packets or connections, it should bind a socket to a local interface address using bind(2). In this case, only one IP socket may be bound to any given local (address, port) pair. When INADDR_ANY is specified in the bind call, the socket will be bound to all local interfaces. When listen(2) is called on an unbound socket, the socket is automatically bound to a random free port with the local address set to INADDR_ANY. When connect(2) is called on an unbound socket, the socket is automatically bound to a random free port or to a usable shared port with the local address set to INADDR_ANY.

A TCP local socket address that has been bound is unavailable for some time after closing, unless the SO_REUSEADDR flag has been set. Care should be taken when using this flag as it makes TCP less reliable.

An IP socket address is defined as a combination of an IP interface address and a 16-bit port number. The basic IP protocol does not supply port numbers, they are implemented by higher level protocols like udp(7) and tcp(7). On raw sockets .sin_port is set to the IP protocol.

See sockaddr_in(3type).

.sin_family is always set to AF_INET. This is required; in Linux 2.2 most networking functions return EINVAL when this setting is missing. .sin_port contains the port in network byte order. The port numbers below 1024 are called privileged ports (or sometimes: reserved ports). Only a privileged process (on Linux: a process that has the CAP_NET_BIND_SERVICE capability in the user namespace governing its network namespace) may bind(2) to these sockets. Note that the raw IPv4 protocol as such has no concept of a port, they are implemented only by higher protocols like tcp(7) and udp(7).

.sin_addr is the IP host address. The .s_addr member of the in_addr(3type) structure contains the host interface address in network byte order. in_addr(3type) should be assigned one of the INADDR_* values (e.g., INADDR_LOOPBACK) using htonl(3) or set using the inet_aton(3), inet_addr(3), inet_makeaddr(3) library functions or directly with the name resolver (see gethostbyname(3)).

IPv4 addresses are divided into unicast, broadcast, and multicast addresses. Unicast addresses specify a single interface of a host, broadcast addresses specify all hosts on a network, and multicast addresses address all hosts in a multicast group. Datagrams to broadcast addresses can be sent or received only when the SO_BROADCAST socket flag is set. In the current implementation, connection-oriented sockets are allowed to use only unicast addresses.

Note that the address and the port are always stored in network byte order. In particular, this means that you need to call htons(3) on the number that is assigned to a port. All address/port manipulation functions in the standard library work in network byte order.

There are several special addresses:

always refers to the local host via the loopback device;
means any address for socket binding;
A packet addressed to INADDR_BROADCAST through a socket which has SO_BROADCAST set will be broadcast to all hosts on the local network segment, as long as the link is broadcast-capable.
On any locally-attached non-point-to-point IP subnet with a link type that supports broadcasts, the highest-numbered address (e.g., the .255 address on a subnet with netmask 255.255.255.0) is designated as a broadcast address. It cannot usefully be assigned to an individual interface, and can only be addressed with a socket on which the SO_BROADCAST option has been set. Internet standards have historically also reserved the lowest-numbered address (e.g., the .0 address on a subnet with netmask 255.255.255.0) for broadcast, though they call it "obsolete" for this purpose. (Some sources also refer to this as the "network address.") Since Linux 5.14, it is treated as an ordinary unicast address and can be assigned to an interface.

Internet standards have traditionally also reserved various addresses for particular uses, though Linux no longer treats some of these specially.

[0.0.0.1, 0.255.255.255]
[240.0.0.0, 255.255.255.254]
Addresses in these ranges (0/8 and 240/4) are reserved globally. Since Linux 5.3 and Linux 2.6.25, respectively, the 0/8 and 240/4 addresses, other than INADDR_ANY and INADDR_BROADCAST, are treated as ordinary unicast addresses. Systems that follow the traditional behaviors may not interoperate with these historically reserved addresses.
[127.0.0.1, 127.255.255.254]
Addresses in this range (127/8) are treated as loopback addresses akin to the standardized local loopback address INADDR_LOOPBACK (127.0.0.1);
[224.0.0.0, 239.255.255.255]
Addresses in this range (224/4) are dedicated to multicast use.

See IPPROTO_IP(2const).

See proc_sys_net_ipv4(5).

All ioctls described in socket(7) apply to ip.

Ioctls to configure generic device parameters are described in netdevice(7).

The user tried to execute an operation without the necessary permissions. These include: sending a packet to a broadcast address without having the SO_BROADCAST flag set; sending a packet via a prohibit route; modifying firewall settings without superuser privileges (the CAP_NET_ADMIN capability); binding to a privileged port without superuser privileges (the CAP_NET_BIND_SERVICE capability).
Tried to bind to an address already in use.
A nonexistent interface was requested or the requested source address was not local.
Operation on a nonblocking socket would block.
A connection operation on a nonblocking socket is already in progress.
A connection was closed during an accept(2).
No valid routing table entry matches the destination address. This error can be caused by an ICMP message from a remote router or for the local routing table.
Invalid argument passed. For send operations this can be caused by sending to a blackhole route.
connect(2) was called on an already connected socket.
Datagram is bigger than an MTU on the path and it cannot be fragmented.
Not enough free memory. This often means that the memory allocation is limited by the socket buffer limits, not by the system memory, but this is not 100% consistent.
SIOCGSTAMP was called on a socket where no packet arrived.
A kernel subsystem was not configured.
Invalid socket option passed.
The operation is defined only on a connected socket, but the socket wasn't connected.
User doesn't have permission to set high priority, change configuration, or send signals to the requested process or group.
The connection was unexpectedly closed or shut down by the other end.
The socket is not configured or an unknown socket type was requested.

Other errors may be generated by the overlaying protocols; see tcp(7), raw(7), udp(7), and socket(7).

Be very careful with the SO_BROADCAST option - it is not privileged in Linux. It is easy to overload the network with careless broadcasts. For new application protocols it is better to use a multicast group instead of broadcasting. Broadcasting is discouraged. See RFC 6762 for an example of a protocol (mDNS) using the more modern multicast approach to communicating with an open-ended group of hosts on the local network.

Using the SOL_IP socket options level isn't portable; BSD-based stacks use the IPPROTO_IP level.

INADDR_ANY (0.0.0.0) and INADDR_BROADCAST (255.255.255.255) are byte-order-neutral. This means htonl(3) has no effect on them.

For compatibility with Linux 2.0, the obsolete socket(AF_INET, SOCK_PACKET, protocol) syntax is still supported to open a packet(7) socket. This is deprecated and should be replaced by socket(AF_PACKET, SOCK_RAW, protocol) instead. The main difference is the new sockaddr_ll address structure for generic link layer information instead of the old sockaddr_pkt.

There are too many inconsistent error values.

The error used to diagnose exhaustion of the ephemeral port range differs across the various system calls (connect(2), bind(2), listen(2), sendto(2)) that can assign ephemeral ports.

The ioctls to configure IP-specific interface options and ARP tables are not described.

Receiving the original destination address with MSG_ERRQUEUE in msg_name by recvmsg(2) does not work in some Linux 2.2 kernels.

IPPROTO_IP(2const), recvmsg(2), sendmsg(2), byteorder(3), capabilities(7), icmp(7), ipv6(7), netdevice(7), netlink(7), raw(7), socket(7), tcp(7), udp(7), ip(8)

The kernel source file Documentation/networking/ip-sysctl.rst.

RFC 791 for the original IP specification. RFC 1122 for the IPv4 host requirements. RFC 1812 for the IPv4 router requirements.