send(2)

NAME

send(), sendmsg(), sendto() — send a message from a socket

SYNOPSIS

#include <sys/socket.h> 

int send(int s, const void *msg, int len, int flags); 

int sendto( 
    int         s, 
    const void *msg, 
    int         len, 
    int         flags, 
    const void *to, 
    int         tolen 
); 

int sendmsg(int s, const struct msghdr msg[], int flags); 

UNIX 03 Only (X/Open Sockets)

ssize_t send(int s, const void *msg, size_t len, int flags); 

ssize_t sendto( 
        int                    s, 
        const void            *msg, 
        size_t                 len, 
        int                    flags, 
        const struct sockaddr *to, 
        socklen_t              tolen 
); 

ssize_t sendmsg(int s, const struct msghdr *msg, int flags); 

Obsolescent UNIX 95 Only (X/Open Sockets)

ssize_t sendto( 
        int                    s, 
        const void            *msg, 
        size_t                 len, 
        int                    flags, 
        const struct sockaddr *to, 
        size_t                 tolen 
); 

DESCRIPTION

The send(), sendmsg(), and sendto() system calls transmit a message to another socket. send() can be used only when the socket is in a connected state, whereas sendmsg() and sendto() can be used at any time. sendmsg() allows the send data to be gathered from several buffers specified in the msghdr structure.

s is a socket descriptor that specifies the socket on which the message will be sent.

msg points to the buffer containing the message.

If the socket uses connection-based communications, such as a SOCK_STREAM socket, these calls can only be used after the connection has been established (see connect(2)). In this case, any destination specified by to is ignored. For connectionless sockets, such as SOCK_DGRAM, sendto() must be used unless the destination address has already been specified by connect(). If the destination address has been specified and sendto() is used, an error results if any address is specified by to.

The address of the target socket is contained in a socket address structure pointed to by to, with tolen specifying the size of the structure.

If a sendto() is attempted on a SOCK_DGRAM socket before any local address has been bound to it, the system automatically selects a local address to be used for the message. In this case, there is no guarantee that the same local address will be used for successive sendto() requests on the same socket.

The length of the message is given by len in bytes. The length of data actually sent is returned. If the message is too long to pass atomically through the underlying protocol, the message is not transmitted, -1 is returned, and errno is set to EMSGSIZE. For SOCK_DGRAM sockets, this size is fixed by the implementation (see the DEPENDENCIES section). Otherwise there is no size limit.

When send() or sendto() returns a positive value, it only indicates this number of bytes have been sent to the local transport provider. It does not mean this number of bytes have been delivered to the peer socket application. A SOCK_DGRAM socket does not guarantee end-to-end delivery. A SOCK_STREAM socket guarantees eventual end-to-end delivery, however its underlying transport provider may later detect an irrecoverable error and returns a value of -1 at another socket function call.

When send() or sendto() returns a value of -1 , it indicates a locally detected error. errno is set to indicate the error.

sendmsg() performs the same action as send(), but it gathers the output data from the buffers specified in the msghdr structure (see X/Open Sockets Only below). This structure is defined in <sys/socket.h> and has the following form (HP-UX BSD Sockets Only):

struct msghdr { 
    caddr_t   msg_name;            /* optional address */ 
    int       msg_namelen;         /* size of address  */ 
    struct    iovec *msg_iov;      /* scatter array for data */ 
    int       msg_iovlen;          /* # of elements in msg_iov */ 
    caddr_t   msg_accrights;       /* access rights */ 
    int       msg_accrightslen;    /* size of msg_accrights */ 
} 

msg_name points to a sockaddr structure in which the address of the destination socket should be stored, if the socket is connectionless; msg_name may be a null pointer if no name is specified. msg_iov specifies the locations of the character arrays for storing the outbound data. msg_accrights specifies a buffer that contains any access rights to be sent along with the message. Access rights are limited to file descriptors of size int. If access rights are not being transferred, set the msg_accrights field to NULL. Access rights are supported only for AF_UNIX.

If no buffer space is available to hold the data to be transmitted, send() blocks unless nonblocking mode is enabled. The three ways to enable nonblocking mode are:

with the FIOSNBIO ioctl() request,
with the O_NONBLOCK flag, and
with the O_NDELAY fcntl() flag.

If nonblocking I/O is enabled using FIOSNBIO or the equivalent FIONBIO request (defined in <sys/ioctl.h> and explained in ioctl(2), ioctl(5), and socket(7)), although the use of FIONBIO is not recommended, the send() request completes in one of three ways:

If there is enough space available in the system to buffer all of the data, send() completes successfully, having written out all of the data, and returns the number of bytes written.
If there is not enough space in the buffer to write out the entire request, send() completes successfully, having written as much data as possible, and returns the number of bytes it was able to write.
If there is no space in the system to buffer any of the data, send() fails, having written no data, and errno is set to EWOULDBLOCK.

If nonblocking I/O is disabled using FIOSNBIO, send() always executes completely (blocking as necessary) and returns the number of bytes written.

If the O_NONBLOCK flag is set using fcntl() (defined in <sys/fcntl.h> and explained in fcntl(2) and fcntl(5)), POSIX-style nonblocking I/O is enabled. In this case, the send() request completes in one of three ways:

If there is enough space available in the system to buffer all of the data, send() completes successfully, having written out all of the data, and returns the number of bytes written.
If there is not enough space in the buffer to write out the entire request, send() completes successfully, having written as much data as possible, and returns the number of bytes it was able to write.
If there is no space in the system to buffer any of the data, send() completes, having written no data, and returns -1, with errno set to EAGAIN.

If the O_NDELAY flag is set using fcntl() (defined in <sys/fcntl.h> and explained in fcntl(2) and fcntl(5)), nonblocking I/O is enabled. In this case, the send() request completes in one of three ways:

If there is enough space available in the system to buffer all of the data, send() completes successfully, having written out all of the data, and returns the number of bytes written.
If there is not enough space in the buffer to write out the entire request, send() completes successfully, having written as much data as possible, and returns the number of bytes it was able to write.
If there is no space in the system to buffer any of the data, send() completes successfully, having written no data, and returns 0.

If the O_NDELAY flag is cleared using fcntl(), nonblocking I/O is disabled. In this case, the send() always executes completely (blocking as necessary) and returns the number of bytes written.

Since the fcntl() O_NONBLOCK and O_NDELAY flags and ioctl() FIOSNBIO requests are supported, the following clarifies on how these features interact. If the O_NONBLOCK or O_NDELAY flag has been set, send() requests behave accordingly, regardless of any FIOSNBIO requests. If neither the O_NONBLOCK flag nor the O_NDELAY flag has been set, FIOSNBIO requests control the behavior of send().

By default nonblocking I/O is disabled.

The supported values for flags are zero or MSG_OOB (to send out-of-band data). A write() call made to a socket behaves in exactly the same way as send() with flags set to zero. MSG_OOB is not supported for AF_UNIX or AF_VME_LINK sockets.

select() can be used to determine when it is possible to send more data.

AF_CCITT Only

Sockets of the address family AF_CCITT operate in message mode. Although they are specified as connection-based (SOCK_STREAM) sockets, the X.25 subsystem communicates via messages. They require that a connection be established with the connect() or accept() calls.

The O_NDELAY flag is not supported. Use FIOSNBIO requests to control nonblocking I/O. If the available buffer space is not large enough for the entire message and the socket is in nonblocking mode, errno is set to EWOULDBLOCK. If the amount of data in the send() exceeds the maximum outbound message size, errno is set to EMSGSIZE.

The sendto() call is not supported.

Each call sends either a complete or a partial X.25 message. This is controlled by the setting of the More-Data-To-Follow (MDTF) bit. If the user wants to send a partial message, MDTF should be set to 1 before the send() call. The MDTF bit should be cleared to 0 before sending the final message fragment.

Message fragment length may range from 0 bytes up to the size of the socket's send buffer. The MDTF bit and multiple send() calls can be combined to transmit complete X.25 packet sequences (zero or more DATA packets in which the More Data bit is set, followed by one DATA packet in which the More Data bit is clear) of arbitrary length. Note that a 0-byte message is not actually sent, but may be necessary to flush a complete X.25 message if the user is controlling the MDTF bit.

Sockets of the AF_CCITT address family can send 1 byte of out-of-band data (known as an INTERRUPT data packet in X.25 terminology), or up to 32 bytes if the X.25 interface is configured for 1984 CCITT X.25 recommendations. INTERRUPT data packets sent in blocking mode cause the process to block until confirmation is received. INTERRUPT data packets sent with the socket in nonblocking mode do not cause the process to block; instead, an out-of-band message is queued to the socket when the INTERRUPT confirmation packet is received (see recv(2)).

X/Open Sockets Only

For X/Open Sockets, the msghdr structure has the following form:

(UNIX 03) 
  
struct msghdr { 
    void       *msg_name;           /* optional address */ 
    socklen_t   msg_namelen;        /* size of address  */ 
    struct      iovec *msg_iov;     /* scatter array for data */ 
    int         msg_iovlen;         /* # of elements in msg_iov */ 
    void       *msg_control;        /* ancillary data, see below */ 
    socklen_t   msg_controllen;     /* ancillary data buffer len */ 
    int         msg_flags;          /* flags on received message */ 
} 
  
Obsolescent (UNIX 95) 
  
struct msghdr { 
    void       *msg_name;           /* optional address */ 
    size_t      msg_namelen;        /* size of address  */ 
    struct      iovec *msg_iov;     /* scatter array for data */ 
    int         msg_iovlen;         /* # of elements in msg_iov */ 
    void       *msg_control;        /* ancillary data, see below */ 
    size_t      msg_controllen;     /* ancillary data buffer len */ 
    int         msg_flags;          /* flags on received message */ 
} 

msg_control specifies a buffer of ancillary data to send along with the message. Ancillary data consists of a sequence of pairs, each consisting of a cmsghdr structure followed by a data array. The data array contains the ancillary data message, and the cmsghdr structure contains descriptive information that allows an application to correctly parse the data. cmsghdr has the following structure:

(UNIX 03) 
  
struct cmsghdr { 
    socklen_t  cmsg_len;      /* data byte count, including hdr*/ 
    int        cmsg_level;    /* originating protocol */ 
    int        cmsg_type;     /* protocol-specific type */ 
} 
  
Obsolescent (UNIX 95) 
  
struct cmsghdr { 
    size_t     cmsg_len;      /* data byte count, including hdr*/ 
    int        cmsg_level;    /* originating protocol */ 
    int        cmsg_type;     /* protocol-specific type */ 
} 

If the cmsg_level is SOL_SOCKET, and cmsg_type is SCM_RIGHTS, then it indicates that the data array contains the access rights to be sent. Access rights are supported only for AF_UNIX. Access rights are limited to file descriptors of size int.

If the cmsg_level is IPPROTO_IPV6, then cmsg_type must be one of the supported types: IPV6_PKTINFO, IPV6_HOPLIMIT, IPV6_NEXTHOP, IPV6_RTHDR, IPV6_HOPOPTS, IPV6_DSTOPTS or IPV6_RTHDRDSTOPTS. (See description in IPv6(7P)).

If ancillary data are not being transferred, set the msg_control field to NULL, and set the msg_controllen field to 0.

The msg_flags member is ignored.

X/Open Sockets Compilation Environment

See xopen_networking(7).

RETURN VALUE

send(), sendmsg(), and sendto() return the following values:

n: Successful completion. n is the number of bytes sent.
-1: Failure. errno is set to indicate the error.

ERRORS

If send(), sendmsg(), or sendto() fails, errno is set to one of the following values.

EACCES

Process doing a send() of a broadcast packet does not have broadcast capability enabled for the socket. Use setsockopt() to enable broadcast capability.

EAFNOSUPPORT

The specified address is not a valid address for the address family of this socket.

EAGAIN

Nonblocking I/O is enabled using the O_NONBLOCK flag with fcntl(), and the requested operation would block, or the socket has an error that was set asynchronously. An asynchronous error can be caused by a gateway failing to forward a datagram from this socket because the datagram exceeds the MTU of the next-hop network and the "Don't Fragment" (DF) bit in the datagram is set. (See SO_PMTU in getsockopt(2)).

EBADF

s is not a valid file descriptor.

ECONNRESET

A connection was forcibly closed by a peer.

EDESTADDRREQ

The to parameter needs to specify a destination address for the message. This is also given if the specified address contains unspecified fields (see inet(7F)).

EFAULT

An invalid pointer was specified in the msg or to parameter, or in the msghdr structure.

EINTR

The operation was interrupted by a signal before any data was sent. (If some data was sent, send() returns the number of bytes sent before the signal, and EINTR is not set).

EINVAL

The len or tolen parameter, or a length in the msghdr structure is invalid. A sendto() system call was issued on an X.25 socket, or the connection is in its reset sequence and cannot accept data.

EIO

A timeout occurred.

EISCONN

An address was specified by to for a SOCK_DGRAM socket which is already connected.

EMSGSIZE

A length in the msghdr structure is invalid. The socket requires that messages be sent atomically, and the size of the message to be sent made this impossible.

SOCK_DGRAM/AF_INET or SOCK_STREAM/AF_CCITT: The message size exceeded the outbound buffer size.

ENETDOWN

The interface used for the specified address is "down" (see ifconfig(1M)), no interface for the specified address can be found (SO_DONTROUTE socket option in use), or the X.25 Level 2 is down.

EHOSTUNREACH

The destination host is not reachable.

ENETUNREACH

The destination network is not reachable. Some of the possible causes for this error are:

(LAN) All encapsulations (for example, Ethernet, IEEE) have been turned off (see also ifconfig(1M)).

(X.25) The X.25 Level 2 is down. The X.25 link layer is not working (wires might be broken, connections are loose on the interface hoods at the modem, the modem failed, the packet switch at the remote end lost power or failed for some reason, or electrical noise interfered with the line for an extremely long period of time).

ENOBUFS

No buffer space is available in the system to perform the operation.

ENOMEM

No memory is available in the system to perform the operation.

ENOPROTOOPT

The remote system or an intermediate system in the communications path does not support a protocol option sent by the local system. This option may have been set using a getsockopt() or setsockopt() call, or set as a system parameter.

ENOTCONN

A send() on a socket that is not connected, or a send() on a socket that has not completed the connect sequence with its peer, or is no longer connected to its peer.

ENOTSOCK

s is a valid file descriptor, but it is not a socket.

EOPNOTSUPP

The MSG_OOB flag was specified; it is not supported for AF_UNIX or AF_VME_LINK sockets.

[EPIPE] and SIGPIPE signal

An attempt was made to send on a socket that was connected, but the connection has been shut down either by the remote peer or by this side of the connection. Note that the default action for SIGPIPE, unless the process has established a signal handler for this signal, is to terminate the process.

EWOULDBLOCK

Nonblocking I/O is enabled using ioctl() FIOSNBIO request and the requested operation would block.

DEPENDENCIES

UDP messages are fragmented at the IP level into Maximum Transmission Unit (MTU) sized pieces; MTU varies for different link types. These pieces, called IP fragments, can be transmitted, but IP does not guarantee delivery. Sending large messages may cause too many fragments and overrun a receiver's ability to receive them. If this happens the complete message cannot be reassembled. This affects the apparent reliability and throughput of the network as viewed by the end user.

The default and maximum buffer sizes are protocol-specific. Refer to the appropriate entries in Sections 7F and 7P for details. The buffer size can be set by calling setsockopt() with SO_SNDBUF.

OBSOLESCENCE

Currently, the socklen_t and size_t types are the same size. This is compatible with the UNIX 95 and UNIX 03 profiles. However, in a future release, socklen_t might be a different size. In that case, the size of the msghdr and cmsghdr structures and the relative position of their members will be different, which might affect application behavior. Applications that use socklen_t now, where appropriate, will avoid such migration problems. On the other hand, applications that need to be portable to the UNIX 95 profile should follow the X/Open specification (see xopen_networking(7)).

WARNINGS

Linking binary objects compiled to X/Open Sockets specification and binary objects compiled to HP-UX BSD Sockets specification to the same executable may result in unexpected behavior, including application abnormal termination and unexpected socket errors. See xopen_networking(7) for details and remedy.

FUTURE DIRECTION

Currently, the default behavior is the HP-UX BSD Sockets; however, it might be changed to X/Open Sockets in a future release. At that time, any HP-UX BSD Sockets behavior that is incompatible with X/Open Sockets might be obsoleted. Applications that conform to the X/Open specification now will avoid migration problems (see xopen_networking(7)).

AUTHOR

send(), sendmsg(), and sendto() were developed by HP and the University of California, Berkeley.

STANDARDS CONFORMANCE

send(): XPG4, UNIX 95, UNIX 03