socat

socat

Feb 2008

CONTENTS

NAME

SYNOPSIS

DESCRIPTION

Socat is a command line based utility that establishes two bidirectional byte streams and transfers data between them. Because the streams can be constructed from a large set of different types of data sinks and sources (see address types), and because lots of address options may be applied to the streams, socat can be used for many different purposes. It might be one of the tools that one `has already needed'.

Filan is a utility that prints information about its active file descriptors to stdout. It has been written for debugging socat, but might be useful for other purposes too. Use the -h option to find more infos.

Procan is a utility that prints information about process parameters to stdout. It has been written to better understand some UNIX process properties and for debugging socat, but might be useful for other purposes too.

The life cycle of a socat instance typically consists of four phases.

In the init phase, the command line options are parsed and logging is initialized.

During the open phase, socat opens the first address and afterwards the second address. These steps are usually blocking; thus, especially for complex address types like socks, connection requests or authentication dialogs must be completed before the next step is started.

In the transfer phase, socat watches both streams' read and write file descriptors via select(), and, when data is available on one side and can be written to the other side, socat reads it, performs newline character conversions if required, and writes the data to the write file descriptor of the other stream, then continues waiting for more data in both directions.

When one of the streams effectively reaches EOF, the closing phase begins. Socat transfers the EOF condition to the other stream, i.e. tries to shutdown only its write stream, giving it a chance to terminate gracefully. For a defined time socat continues to transfer data in the other direction, but then closes all remaining channels and terminates.

OPTIONS

Socat provides some command line options that modify the behaviour of the program. They have nothing to do with so called address options that are used as parts of address specifications.

-V

Print version and available feature information to stdout, and exit.

-h | -?

Print a help text to stdout describing command line options and available address types, and exit.

-hh | -??

Like -h, plus a list of the short names of all available address options. Some options are platform dependend, so this output is helpful for checking the particular implementation.

-hhh | -???

Like -hh, plus a list of all available address option names.

-d

Without this option, only fatal and error messages are generated; applying this option also prints warning messages. See DIAGNOSTICS for more information.

-d -d

Prints fatal, error, warning, and notice messages.

-d -d -d

Prints fatal, error, warning, notice, and info messages.

-d -d -d -d

Prints fatal, error, warning, notice, info, and debug messages.

-D

Logs information about file descriptors before starting the transfer phase.

-ly[<facility>]

Writes messages to syslog instead of stderr; severity as defined with -d option. With optional <facility>, the syslog type can be selected, default is "daemon".

-lf <logfile>

Writes messages to <logfile> [filename] instead of stderr.

-ls

Writes messages to stderr (this is the default).

-lp<progname>

Overrides the program name printed in error messages.

-lu

Extends the timestamp of error messages to microsecond resolution. Does not work when logging to syslog.

-lm[<facility>]

Mixed log mode. During startup messages are printed to stderr; when socat starts the transfer phase loop or daemon mode (i.e. after opening all streams and before starting data transfer, or, with listening sockets with fork option, before the first accept call), it switches logging to syslog. With optional <facility>, the syslog type can be selected, default is "daemon".

-lh

Adds hostname to log messages. Uses the value from environment variable HOSTNAME or the value retrieved with uname() if HOSTNAME is not set.

-v

Writes the transferred data not only to their target streams, but also to stderr. The output format is text with some conversions for readability, and prefixed with "> " or "< " indicating flow directions.

-x

Writes the transferred data not only to their target streams, but also to stderr. The output format is hexadecimal, prefixed with "> " or "< " indicating flow directions. Can be combined with -v.

-b<size>

Sets the data transfer block <size> [size_t]. At most <size> bytes are transferred per step. Default is 8192 bytes.

-s

By default, socat terminates when an error occurred to prevent the process from running when some option could not be applied. With this option, socat is sloppy with errors and tries to continue. Even with this option, socat will exit on fatals, and will abort connection attempts when security checks failed.

-t<timeout>

When one channel has reached EOF, the write part of the other channel is shut down. Then, socat waits <timeout> [timeval] seconds before terminating. Default is 0.5 seconds. This timeout only applies to addresses where write and read part can be closed independently. When during the timeout interval the read part gives EOF, socat terminates without awaiting the timeout.

-T<timeout>

Total inactivity timeout: when socat is already in the transfer loop and nothing has happened for <timeout> [timeval] seconds (no data arrived, no interrupt occurred...) then it terminates. Useful with protocols like UDP that cannot transfer EOF.

-u

Uses unidirectional mode. The first address is only used for reading, and the second address is only used for writing (example).

-U

Uses unidirectional mode in reverse direction. The first address is only used for writing, and the second address is only used for reading.

-g

During address option parsing, don't check if the option is considered useful in the given address environment. Use it if you want to force, e.g., appliance of a socket option to a serial device.

-L<lockfile>

If lockfile exists, exits with error. If lockfile does not exist, creates it and continues, unlinks lockfile on exit.

-W<lockfile>

If lockfile exists, waits until it disappears. When lockfile does not exist, creates it and continues, unlinks lockfile on exit.

-4

Use IP version 4 in case that the addresses do not implicitly or explicitly specify a version; this is the default.

-6

Use IP version 6 in case that the addresses do not implicitly or explicitly specify a version.

ADDRESS SPECIFICATIONS

With the address command line arguments, the user gives socat instructions and the necessary information for establishing the byte streams.

An address specification usually consists of an address type keyword, zero or more required address parameters separated by ':' from the keyword and from each other, and zero or more address options separated by ','.

The keyword specifies the address type (e.g., TCP4, OPEN, EXEC). For some keywords there exist synonyms ('-' for STDIO, TCP for TCP4). Keywords are case insensitive. For a few special address types, the keyword may be omitted: Address specifications starting with a number are assumed to be FD (raw file descriptor) addresses; if a '/' is found before the first ':' or ',', GOPEN (generic file open) is assumed.

The required number and type of address parameters depend on the address type. E.g., TCP4 requires a server specification (name or address), and a port specification (number or service name).

Zero or more address options may be given with each address. They influence the address in some ways. Options consist of an option keyword or an option keyword and a value, separated by '='. Option keywords are case insensitive. For filtering the options that are useful with an address type, each option is member of one option group. For each address type there is a set of option groups allowed. Only options belonging to one of these address groups may be used (except with option -g).

Address specifications following the above schema are also called single address specifications. Two single addresses can be combined with "!!" to form a dual type address for one channel. Here, the first address is used by socat for reading data, and the second address for writing data. There is no way to specify an option only once for being applied to both single addresses.

Usually, addresses are opened in read/write mode. When an address is part of a dual address specification, or when option -u or -U is used, an address might be used only for reading or for writing. Considering this is important with some address types.

With socat version 1.5.0 and higher, the lexical analysis tries to handle quotes and parenthesis meaningfully and allows escaping of special characters. If one of the characters ( { [ ' is found, the corresponding closing character - ) } ] ' - is looked for; they may also be nested. Within these constructs, socats special characters and strings : , !! are not handled specially. All those characters and strings can be escaped with \ or within ""

ADDRESS TYPES

This section describes the available address types with their keywords, parameters, and semantics.

CREATE:<filename>

Opens <filename> with creat() and uses the file descriptor for writing. This address type requires write-only context, because a file opened with creat cannot be read from. <filename> must be a valid existing or not existing path. If <filename> is a named pipe, creat() might block; if <filename> refers to a socket, this is an error.
Option groups: FD,REG,NAMED
Useful options: mode, user, group, unlink-early, unlink-late, append
See also: OPEN, GOPEN

EXEC:<command-line>

Forks a sub process that establishes communication with its parent process and invokes the specified program with execvp(). <command-line> is a simple command with arguments separated by single spaces. If the program name contains a '/', the part after the last '/' is taken as ARGV[0]. If the program name is a relative path, the execvp() semantics for finding the program via $PATH apply. After successful program start, socat writes data to stdin of the process and reads from its stdout using a UNIX domain socket generated by socketpair() per default. (example)
Option groups: FD,SOCKET,EXEC,FORK,TERMIOS
Useful options: path, fdin, fdout, chroot, su, su-d, nofork, pty, stderr, ctty, setsid, pipes, login, sigint, sigquit
See also: SYSTEM

FD:<fdnum>

Uses the file descriptor <fdnum>. It must already exist as valid UN*X file descriptor.
Option groups: FD (TERMIOS,REG,SOCKET)
See also: STDIO, STDIN, STDOUT, STDERR

GOPEN:<filename>

(Generic open) This address type tries to handle any file system entry except directories usefully. <filename> may be a relative or absolute path. If it already exists, its type is checked. In case of a UNIX domain socket, socat connects; if connecting fails, socat assumes a datagram socket and uses sendto() calls. If the entry is not a socket, socat opens it applying the O_APPEND flag. If it does not exist, it is opened with flag O_CREAT as a regular file (example).
Option groups: FD,REG,SOCKET,NAMED,OPEN
See also: OPEN, CREATE, UNIX-CONNECT

IP-SENDTO:<host>:<protocol>

Opens a raw IP socket. Depending on host specification or option pf, IP procotol version 4 or 6 is used. It uses <protocol> to send packets to <host> [IP address] and receives packets from host, ignores packets from other hosts. Protocol 255 uses the raw socket with the IP header being part of the data.
Option groups: FD,SOCKET,IP4,IP6
Useful options: pf, ttl See also: IP4-SENDTO, IP6-SENDTO, IP-RECVFROM, IP-RECV, UDP-SENDTO UNIX-SENDTO

IP4-SENDTO:<host>:<protocol>

Like IP-SENDTO, but always uses IPv4.
Option groups: FD,SOCKET,IP4

IP6-SENDTO:<host>:<protocol>

Like IP-SENDTO, but always uses IPv6.
Option groups: FD,SOCKET,IP6

IP-DATAGRAM:<address>:<protocol>

Sends outgoing data to the specified address which may in particular be a broadcast or multicast address. Packets arriving on the local socket are checked if their source addresses match eventual RANGE or TCPWRAP options. This address type can for example be used for implementing symmetric or asymmetric broadcast or multicast communications.
Option groups: FD, SOCKET, IP4, IP6, RANGE
Useful options: range, tcpwrap, broadcast, ip-multicast-loop, ip-multicast-ttl, ip-multicast-if, ip-add-membership, ttl, tos, bind, pf
See also: IP4-DATAGRAM, IP6-DATAGRAM, IP-SENDTO, IP-RECVFROM, IP-RECV, UDP-DATAGRAM

IP4-DATAGRAM:<host>:<protocol>

Like IP-DATAGRAM, but always uses IPv4. (example)
Option groups: FD, SOCKET, IP4, RANGE

IP6-DATAGRAM:<host>:<protocol>

Like IP-DATAGRAM, but always uses IPv6. Please note that IPv6 does not know broadcasts.
Option groups: FD, SOCKET, IP6, RANGE

IP-RECVFROM:<protocol>

Opens a raw IP socket of <protocol>. Depending on option pf, IP procotol version 4 or 6 is used. It receives one packet from an unspecified peer and may send one or more answer packets to that peer. This mode is particularly useful with fork option where each arriving packet - from arbitrary peers - is handled by its own sub process. This allows a behaviour similar to typical UDP based servers like ntpd or named. This address works well with IP-SENDTO address peers (see above). Protocol 255 uses the raw socket with the IP header being part of the data.
Option groups: FD,SOCKET,IP4,IP6,CHILD,RANGE
Useful options: pf, fork, range, ttl, broadcast
See also: IP4-RECVFROM, IP6-RECVFROM, IP-SENDTO, IP-RECV, UDP-RECVFROM, UNIX-RECVFROM

IP4-RECVFROM:<protocol>

Like IP-RECVFROM, but always uses IPv4.
Option groups: FD,SOCKET,IP4,CHILD,RANGE

IP6-RECVFROM:<protocol>

Like IP-RECVFROM, but always uses IPv6.
Option groups: FD,SOCKET,IP6,CHILD,RANGE

IP-RECV:<protocol>

Opens a raw IP socket of <protocol>. Depending on option pf, IP procotol version 4 or 6 is used. It receives packets from multiple unspecified peers and merges the data. No replies are possible. It can be, e.g., addressed by socat IP-SENDTO address peers. Protocol 255 uses the raw socket with the IP header being part of the data.
Option groups: FD,SOCKET,IP4,IP6,RANGE
Useful options: pf, range
See also: IP4-RECV, IP6-RECV, IP-SENDTO, IP-RECVFROM, UDP-RECV, UNIX-RECV

IP4-RECV:<protocol>

Like IP-RECV, but always uses IPv4.
Option groups: FD,SOCKET,IP4,RANGE

IP6-RECV:<protocol>

Like IP-RECV, but always uses IPv6.
Option groups: FD,SOCKET,IP6,RANGE

OPEN:<filename>

Opens <filename> using the open() system call (example). This operation fails on UNIX domain sockets.
Note: This address type is rarly useful in bidirectional mode.
Option groups: FD,REG,NAMED,OPEN
Useful options: creat, excl, noatime, nofollow, append, rdonly, wronly, lock, readbytes, ignoreeof
See also: CREATE, GOPEN, UNIX-CONNECT

OPENSSL:<host>:<port>

Tries to establish a SSL connection to <port> [TCP service] on <host> [IP address] using TCP/IP version 4 or 6 depending on address specification, name resolution, or option pf.
NOTE: The server certificate is only checked for validity against cafile or capath, but not for match with the server's name or its IP address!
Option groups: FD,SOCKET,IP4,IP6,TCP,OPENSSL,RETRY
Useful options: cipher, method, verify, cafile, capath, certificate, bind, pf, connect-timeout, sourceport, retry
See also: OPENSSL-LISTEN, TCP

OPENSSL-LISTEN:<port>

Listens on tcp <port> [TCP service]. The IP version is 4 or the one specified with pf. When a connection is accepted, this address behaves as SSL server.
Note: You probably want to use the certificate option with this address.
NOTE: The client certificate is only checked for validity against cafile or capath, but not for match with the client's name or its IP address!
Option groups: FD,SOCKET,IP4,IP6,TCP,LISTEN,OPENSSL,CHILD,RANGE,RETRY
Useful options: pf, cipher, method, verify, cafile, capath, certificate, fork, bind, range, tcpwrap, su, reuseaddr, retry
See also: OPENSSL, TCP

PIPE:<filename>

If <filename> already exists, it is opened. If is does not exist, a named pipe is created and opened. Beginning with socat version 1.4.3, the named pipe is removed when the address is closed (but see option unlink-close
Note: When a pipe is used for both reading and writing, it works as echo service.
Note: When a pipe is used for both reading and writing, and socat tries to write more bytes than the pipe can buffer (Linux 2.4: 2048 bytes), socat might block. Consider using socat option, e.g., -b 2048
Option groups: FD,NAMED,OPEN
Useful options: rdonly, nonblock, group, user, mode, unlink-early
See also: unnamed pipe

PIPE

Creates an unnamed pipe and uses it for reading and writing. It works as an echo, because everything written to it appeares immediately as read data.
Note: When socat tries to write more bytes than the pipe can queue (Linux 2.4: 2048 bytes), socat might block. Consider, e.g., using option -b 2048
Option groups: FD
See also: named pipe

PROXY:<proxy>:<hostname>:<port>

Connects to an HTTP proxy server on port 8080 using TCP/IP version 4 or 6 depending on address specification, name resolution, or option pf, and sends a CONNECT request for hostname:port. If the proxy grants access and succeeds to connect to the target, data transfer between socat and the target can start. Note that the traffic need not be HTTP but can be an arbitrary protocol.
Option groups: FD,SOCKET,IP4,IP6,TCP,HTTP,RETRY
Useful options: proxyport, ignorecr, proxyauth, resolve, crnl, bind, connect-timeout, mss, sourceport, retry
See also: SOCKS, TCP

PTY

Generates a pseudo terminal (pty) and uses its master side. Another process may open the pty's slave side using it like a serial line or terminal. (example). If both the ptmx and the openpty mechanisms are available, ptmx is used (POSIX).
Option groups: FD,NAMED,PTY,TERMIOS
Useful options: link, openpty, wait-slave, mode, user, group
See also: UNIX-LISTEN, PIPE, EXEC, SYSTEM

READLINE

Uses GNU readline and history on stdio to allow editing and reusing input lines (example). This requires the GNU readline and history libraries. Note that stdio should be a (pseudo) terminal device, otherwise readline does not seem to work.
Option groups: FD,READLINE,TERMIOS
Useful options: history, noecho
See also: STDIO

SOCKS4:<socks-server>:<host>:<port>

Connects via <socks-server> [IP address] to <host> [IPv4 address] on <port> [TCP service], using socks version 4 protocol over IP version 4 or 6 depending on address specification, name resolution, or option pf (example).
Option groups: FD,SOCKET,IP4,IP6,TCP,SOCKS4,RETRY
Useful options: socksuser, socksport, sourceport, pf, retry
See also: SOCKS4A, PROXY, TCP

SOCKS4A:<socks-server>:<host>:<port>

like SOCKS4, but uses socks protocol version 4a, thus leaving host name resolution to the socks server.
Option groups: FD,SOCKET,IP4,IP6,TCP,SOCKS4,RETRY

STDERR

Uses file descriptor 2.
Option groups: FD (TERMIOS,REG,SOCKET)
See also: FD

STDIN

Uses file descriptor 0.
Option groups: FD (TERMIOS,REG,SOCKET)
Useful options: readbytes
See also: FD

STDIO

Uses file descriptor 0 for reading, and 1 for writing.
Option groups: FD (TERMIOS,REG,SOCKET)
Useful options: readbytes
See also: FD

STDOUT

Uses file descriptor 1.
Option groups: FD (TERMIOS,REG,SOCKET)
See also: FD

SYSTEM:<shell-command>

Forks a sub process that establishes communication with its parent process and invokes the specified program with system(). Please note that <shell-command> [string] must not contain ',' or "!!", and that shell meta characters may have to be protected. After successful program start, socat writes data to stdin of the process and reads from its stdout.
Option groups: FD,SOCKET,EXEC,FORK,TERMIOS
Useful options: path, fdin, fdout, chroot, su, su-d, nofork, pty, stderr, ctty, setsid, pipes, sigint, sigquit
See also: EXEC

TCP:<host>:<port>

Connects to <port> [TCP service] on <host> [IP address] using TCP/IP version 4 or 6 depending on address specification, name resolution, or option pf.
Option groups: FD,SOCKET,IP4,IP6,TCP,RETRY
Useful options: crnl, bind, pf, connect-timeout, tos, mtudiscover, mss, nodelay, nonblock, sourceport, retry, readbytes
See also: TCP4, TCP6, TCP-LISTEN, UDP, UNIX-CONNECT

TCP4:<host>:<port>

Like TCP, but only supports IPv4 protocol (example).
Option groups: FD,SOCKET,IP4,TCP,RETRY

TCP6:<host>:<port>

Like TCP, but only supports IPv6 protocol.
Option groups: FD,SOCKET,IP6,TCP,RETRY

TCP-LISTEN:<port>

Listens on <port> [TCP service] and accepts a TCP/IP connection. The IP version is 4 or the one specified with pf. Note that opening this address usually blocks until a client connects.
Option groups: FD,SOCKET,LISTEN,CHILD,RANGE,IP4,IP6,TCP,RETRY
Useful options: crnl, fork, bind, range, tcpwrap, pf, backlog, mss, su, reuseaddr, retry, retry
See also: TCP4-LISTEN, TCP6-LISTEN, UDP-LISTEN, UNIX-LISTEN, OPENSSL-LISTEN

TCP4-LISTEN:<port>

Like TCP-LISTEN, but only supports IPv4 protocol (example).
Option groups: FD,SOCKET,LISTEN,CHILD,RANGE,IP4,TCP,RETRY

TCP6-LISTEN:<port>

Like TCP-LISTEN, but only supports IPv6 protocol.
Additional useful option: ipv6only
Option groups: FD,SOCKET,LISTEN,CHILD,RANGE,IP6,TCP,RETRY

TUN:<if-addr>/<bits>

Creates a Linux TUN/TAP device and assignes to it the address and netmask defined by the parameters. The resulting network interface is ready for use by other processes; socat serves its "wire side". This address requires read and write access to the tunnel cloning device, usually /dev/net/tun.
Option groups: FD,NAMED,OPEN,TUN
Useful options: iff-up, tun-device, tun-name, tun-type, iff-no-pi
See also: ip-recv

UDP:<host>:<port>

Connects to <port> [UDP service] on <host> [IP address] using UDP/IP version 4 or 6 depending on address specification, name resolution, or option pf.
Please note that, due to UDP protocol properties, no real connection is established; data has to be sent for `connecting' to the server, and no end-of-file condition can be transported.
Option groups: FD,SOCKET,IP4,IP6
Useful options: ttl, tos, bind, sourceport, pf
See also: UDP4, UDP6, UDP-LISTEN, TCP, IP

UDP4:<host>:<port>

Like UDP, but only supports IPv4 protocol.
Option groups: FD,SOCKET,IP4

UDP6:<host>:<port>

Like UDP, but only supports IPv6 protocol.
Option groups: FD,SOCKET,IP6

UDP-DATAGRAM:<address>:<port>

Sends outgoing data to the specified address which may in particular be a broadcast or multicast address. Packets arriving on the local socket are checked for the correct remote port and if their source addresses match eventual RANGE or TCPWRAP options. This address type can for example be used for implementing symmetric or asymmetric broadcast or multicast communications.
Option groups: FD,SOCKET,IP4,IP6,RANGE
Useful options: range, tcpwrap, broadcast, ip-multicast-loop, ip-multicast-ttl, ip-multicast-if, ip-add-membership, ttl, tos, bind, sourceport, pf
See also: UDP4-DATAGRAM, UDP6-DATAGRAM, UDP-SENDTO, UDP-RECVFROM, UDP-RECV, UDP-CONNECT, UDP-LISTEN, IP-DATAGRAM

UDP4-DATAGRAM:<address>:<port>

Like UDP-DATAGRAM, but only supports IPv4 protocol (example1, example2).
Option groups: FD, SOCKET, IP4, RANGE

UDP6-DATAGRAM:<address>:<port>

Like UDP-DATAGRAM, but only supports IPv6 protocol.
Option groups: FD,SOCKET, IP6,RANGE

UDP-LISTEN:<port>

Waits for a UDP/IP packet arriving on <port> [UDP service] and `connects' back to sender. The accepted IP version is 4 or the one specified with option pf. Please note that, due to UDP protocol properties, no real connection is established; data has to arrive from the peer first, and no end-of-file condition can be transported. Note that opening this address usually blocks until a client connects.
Option groups: FD,SOCKET,LISTEN,CHILD,RANGE,IP4,IP6
Useful options: fork, bind, range, pf
See also: UDP, UDP4-LISTEN, UDP6-LISTEN, TCP-LISTEN

UDP4-LISTEN:<port>

Like UDP-LISTEN, but only support IPv4 protocol.
Option groups: FD,SOCKET,LISTEN,CHILD,RANGE,IP4

UDP6-LISTEN:<port>

Like UDP-LISTEN, but only support IPv6 protocol.
Option groups: FD,SOCKET,LISTEN,CHILD,RANGE,IP6

UDP-SENDTO:<host>:<port>

Communicates with the specified peer socket, defined by <port> [UDP service] on <host> [IP address], using UDP/IP version 4 or 6 depending on address specification, name resolution, or option pf. It sends packets to and receives packets from that peer socket only. This address effectively implements a datagram client. It works well with socat UDP-RECVFROM and UDP-RECV address peers.
Option groups: FD,SOCKET,IP4,IP6
Useful options: ttl, tos, bind, sourceport, pf
See also: UDP4-SENDTO, UDP6-SENDTO, UDP-RECVFROM, UDP-RECV, UDP-CONNECT, UDP-LISTEN, IP-SENDTO

UDP4-SENDTO:<host>:<port>

Like UDP-SENDTO, but only supports IPv4 protocol.
Option groups: FD,SOCKET,IP4

UDP6-SENDTO:<host>:<port>

Like UDP-SENDTO, but only supports IPv6 protocol.
Option groups: FD,SOCKET,IP6

UDP-RECVFROM:<port>

Creates a UDP socket on <port> [UDP service] using UDP/IP version 4 or 6 depending on option pf. It receives one packet from an unspecified peer and may send one or more answer packets to that peer. This mode is particularly useful with fork option where each arriving packet - from arbitrary peers - is handled by its own sub process. This allows a behaviour similar to typical UDP based servers like ntpd or named. This address works well with socat SENDTO address peers.
Option groups: FD,SOCKET,IP4,IP6,CHILD,RANGE
Useful options: fork, ttl, tos, bind, sourceport, pf
See also: UDP4-RECVFROM, UDP6-RECVFROM, UDP-SENDTO, UDP-RECV, UDP-CONNECT, UDP-LISTEN, IP-RECVFROM, UNIX-RECVFROM

UDP4-RECVFROM:<port>

Like UDP-RECVFROM, but only supports IPv4 protocol.
Option groups: FD,SOCKET,IP4,CHILD,RANGE

UDP6-RECVFROM:<port>

Like UDP-RECVFROM, but only supports IPv6 protocol.
Option groups: FD,SOCKET,IP6,CHILD,RANGE

UDP-RECV:<port>

Creates a UDP socket on <port> [UDP service] using UDP/IP version 4 or 6 depending on option pf. It receives packets from multiple unspecified peers and merges the data. No replies are possible. It works well with, e.g., socat UDP-SENDTO address peers; it behaves similar to a syslog server.
Option groups: FD,SOCKET,IP4,IP6,RANGE
Useful options: fork, pf, bind, sourceport, ttl, tos
See also: UDP4-RECV, UDP6-RECV, UDP-SENDTO, UDP-RECVFROM, UDP-CONNECT, UDP-LISTEN, IP-RECV, UNIX-RECV

UDP4-RECV:<port>

Like UDP-RECV, but only supports IPv4 protocol.
Option groups: FD,SOCKET,IP4,RANGE

UDP6-RECV:<port>

Like UDP-RECV, but only supports IPv6 protocol.
Option groups: FD,SOCKET,IP6,RANGE

UNIX-CONNECT:<filename>

Connects to <filename> assuming it is a UNIX domain socket. If <filename> does not exist, this is an error; if <filename> is not a UNIX domain socket, this is an error; if <filename> is a UNIX domain socket, but no process is listening, this is an error.
Option groups: FD,SOCKET, NAMED,RETRY, UNIX
) Useful options: bind
See also: UNIX-LISTEN, UNIX-SENDTO, TCP

UNIX-LISTEN:<filename>

Listens on <filename> using a UNIX domain stream socket and accepts a connection. If <filename> exists and is not a socket, this is an error. If <filename> exists and is a UNIX domain socket, binding to the address fails (use option unlink-early!). Note that opening this address usually blocks until a client connects. Beginning with socat version 1.4.3, the file system entry is removed when this address is closed (but see option unlink-close) (example).
Option groups: FD,SOCKET, NAMED,LISTEN, CHILD,RETRY, UNIX
Useful options: fork, umask, mode, user, group, unlink-early
See also: UNIX-CONNECT, UNIX-RECVFROM, UNIX-RECV, TCP-LISTEN

UNIX-SENDTO:<filename>

Communicates with the specified peer socket, defined by [<filename>] assuming it is a UNIX domain datagram socket. It sends packets to and receives packets from that peer socket only. It works well with socat UNIX-RECVFROM and UNIX-RECV address peers.
Option groups: FD,SOCKET, NAMED,UNIX
Useful options: bind
See also: UNIX-RECVFROM, UNIX-RECV, UNIX-CONNECT, UDP-SENDTO, IP-SENDTO

UNIX-RECVFROM:<filename>

Creates a UNIX domain datagram socket [<filename>]. Receives one packet and may send one or more answer packets to that peer. This mode is particularly useful with fork option where each arriving packet - from arbitrary peers - is handled by its own sub process. This address works well with socat UNIX-SENDTO address peers.
Option groups: FD,SOCKET, NAMED,CHILD, UNIX
Useful options: fork
See also: UNIX-SENDTO, UNIX-RECV, UNIX-LISTEN, UDP-RECVFROM, IP-RECVFROM

UNIX-RECV:<filename>

Creates a UNIX domain datagram socket [<filename>]. Receives packets from multiple unspecified peers and merges the data. No replies are possible. It can be, e.g., addressed by socat UNIX-SENDTO address peers. It behaves similar to a syslog server. Option groups: FD,SOCKET, NAMED,UNIX
See also: UNIX-SENDTO, UNIX-RECVFROM, UNIX-LISTEN, UDP-RECV, IP-RECV

UNIX-CLIENT:<filename>

Communicates with the specified peer socket, defined by [<filename>] assuming it is a UNIX domain socket. It first tries to connect and, if that fails, assumes it is a datagram socket, thus supporting both types.
Option groups: FD,SOCKET, NAMED,UNIX
Useful options: bind
See also: UNIX-CONNECT, UNIX-SENDTO, GOPEN

ABSTRACT-CONNECT:<string>

ABSTRACT-LISTEN:<string>

ABSTRACT-SENDTO:<string>

ABSTRACT-RECVFROM:<string>

ABSTRACT-RECV:<string>

ABSTRACT-CLIENT:<string>

The ABSTRACT addresses are almost identical to the related UNIX addresses except that they do not address file system based sockets but an alternate UNIX domain address space. To archieve this the socket address strings are prefixed with "\0" internally. This feature is available (only?) on Linux. Option groups are the same as with the related UNIX addresses, except that the ABSTRACT addresses are not member of the NAMED group.

ADDRESS OPTIONS

Address options can be applied to address specifications to influence the process of opening the addresses and the properties of the resulting data channels.

For technical reasons not every option can be applied to every address type; e.g., applying a socket option to a regular file will fail. To catch most useless combinations as early as in the open phase, the concept of option groups was introduced. Each option belongs to one or more option groups. Options can be used only with address types that support at least one of their option groups (but see option -g).

Address options have data types that their values must conform to. Every address option consists of just a keyword or a keyword followed by "=value", where value must conform to the options type. Some address options manipulate parameters of system calls; e.g., option sync sets the O_SYNC flag with the open() call. Other options cause a system or library call; e.g., with option `ttl=value' the setsockopt(fd, SOL_IP, IP_TTL, value, sizeof(int)) call is applied. Other options set internal socat variables that are used during data transfer; e.g., `crnl' causes explicit character conversions. A few options have more complex implementations; e.g., su-d (substuser-delayed) inquires some user and group infos, stores them, and applies them later after a possible chroot() call.

If multiple options are given to an address, their sequence in the address specification has (almost) no effect on the sequence of their execution/application. Instead, socat has built in an option phase model that tries to bring the options in a useful order. Some options exist in different forms (e.g., unlink, unlink-early, unlink-late) to control the time of their execution.

If the same option is specified more than once within one address specification, with equal or different values, the effect depends on the kind of option. Options resulting in function calls like setsockopt() cause multiple invocations. With options that set parameters for a required call like open() or set internal flags, the value of the last option occurrence is effective.

The existence or semantics of many options are system dependent. Socat usually does NOT try to emulate missing libc or kernel features, it just provides an interface to the underlying system. So, if an operating system lacks a feature, the related option is simply not available on this platform.

The following paragraphs introduce just the more common address options. For a more comprehensive reference and to find information about canonical option names, alias names, option phases, and platforms see file xio.help.

FD option group

This option group contains options that are applied to a UN*X style file descriptor, no matter how it was generated. Because all current socat address types are file descriptor based, these options may be applied to any address.
Note: Some of these options are also member of another option group, that provides an other, non-fd based mechanism. For these options, it depends on the actual address type and its option groups which mechanism is used. The second, non-fd based mechanism is prioritized.

cloexec=<bool>

Sets the FD_CLOEXEC flag with the fcntl() system call to value <bool>. If set, the file descriptor is closed on exec() family function calls. Socat internally handles this flag for the fds it controls, so in most cases there will be no need to apply this option.

setlk

Tries to set a discretionary write lock to the whole file using the fcntl(fd, F_SETLK, ...) system call. If the file is already locked, this call results in an error. On Linux, when the file permissions for group are "S" (g-x,g+s), and the file system is locally mounted with the "mand" option, the lock is mandatory, i.e. prevents other processes from opening the file.

setlkw

Tries to set a discretionary waiting write lock to the whole file using the fcntl(fd, F_SETLKW, ...) system call. If the file is already locked, this call blocks. See option setlk for information about making this lock mandatory.

setlk-rd

Tries to set a discretionary read lock to the whole file using the fcntl(fd, F_SETLK, ...) system call. If the file is already write locked, this call results in an error. See option setlk for information about making this lock mandatory.

setlkw-rd

Tries to set a discretionary waiting read lock to the whole file using the fcntl(fd, F_SETLKW, ...) system call. If the file is already write locked, this call blocks. See option setlk for information about making this lock mandatory.

flock-ex

Tries to set a blocking exclusive advisory lock to the file using the flock(fd, LOCK_EX) system call. Socat hangs in this call if the file is locked by another process.

flock-ex-nb

Tries to set a nonblocking exclusive advisory lock to the file using the flock(fd, LOCK_EX|LOCK_NB) system call. If the file is already locked, this option results in an error.

flock-sh

Tries to set a blocking shared advisory lock to the file using the flock(fd, LOCK_SH) system call. Socat hangs in this call if the file is locked by another process.

flock-sh-nb

Tries to set a nonblocking shared advisory lock to the file using the flock(fd, LOCK_SH|LOCK_NB) system call. If the file is already locked, this option results in an error.

lock

Sets a blocking lock on the file. Uses the setlk or flock mechanism depending on availability on the particular platform. If both are available, the POSIX variant (setlkw) is used.

user=<user>

Sets the <user> (owner) of the stream. If the address is member of the NAMED option group, socat uses the chown() system call after opening the file or binding to the UNIX domain socket (race condition!). Without filesystem entry, socat sets the user of the stream using the fchown() system call. These calls might require root privilege.

user-late=<user>

Sets the owner of the fd to <user> with the fchown() system call after opening or connecting the channel. This is useful only on file system entries.

group=<group>

Sets the <group> of the stream. If the address is member of the NAMED option group, socat uses the chown() system call after opening the file or binding to the UNIX domain socket (race condition!). Without filesystem entry, socat sets the group of the stream with the fchown() system call. These calls might require group membership or root privilege.

group-late=<group>

Sets the group of the fd to <group> with the fchown() system call after opening or connecting the channel. This is useful only on file system entries.

mode=<mode>

Sets the <mode> [mode_t] (permissions) of the stream. If the address is member of the NAMED option group and uses the open() or creat() call, the mode is applied with these. If the address is member of the NAMED option group without using these system calls, socat uses the chmod() system call after opening the filesystem entry or binding to the UNIX domain socket (race condition!). Otherwise, socat sets the mode of the stream using fchmod(). These calls might require ownership or root privilege.

perm-late=<mode>

Sets the permissions of the fd to value <mode> [mode_t] using the fchmod() system call after opening or connecting the channel. This is useful only on file system entries.

append=<bool>

Always writes data to the actual end of file. If the address is member of the OPEN option group, socat uses the O_APPEND flag with the open() system call (example). Otherwise, socat applies the fcntl(fd, F_SETFL, O_APPEND) call.

nonblock=<bool>

Tries to open or use file in nonblocking mode. Its only effects are that the connect() call of TCP addresses does not block, and that opening a named pipe for reading does not block. If the address is member of the OPEN option group, socat uses the O_NONBLOCK flag with the open() system call. Otherwise, socat applies the fcntl(fd, F_SETFL, O_NONBLOCK) call.

binary

Opens the file in binary mode to avoid implicit line terminator conversions (Cygwin).

text

Opens the file in text mode to force implicit line terminator conversions (Cygwin).

noinherit

Does not keep this file open in a spawned process (Cygwin).

cool-write

Takes it easy when write fails with EPIPE or ECONNRESET and logs the message with notice level instead of error. This prevents the log file from being filled with useless error messages when socat is used as a high volume server or proxy where clients often abort the connection.
This option is experimental.

end-close

Changes the (address dependent) method of ending a connection to just close the file descriptors. This is useful when the connection is to be reused by or shared with other processes (example).
Normally, socket connections will be ended with shutdown(2) which terminates the socket even if it is shared by multiple processes. close(2) "unlinks" the socket from the process but keeps it active as long as there are still links from other processes.
Similarly, when an address of type EXEC or SYSTEM is ended, socat usually will explicitely kill the sub process. With this option, it will just close the file descriptors.

NAMED option group

These options work on file system entries.
See also options user, group, and mode.

user-early=<user>

Changes the <user> (owner) of the file system entry before accessing it, using the chown() system call. This call might require root privilege.

group-early=<group>

Changes the <group> of the file system entry before accessing it, using the chown() system call. This call might require group membership or root privilege.

perm-early=<mode>

Changes the <mode> [mode_t] of the file system entry before accessing it, using the chmod() system call. This call might require ownership or root privilege.

umask=<mode>

Sets the umask of the process to <mode> [mode_t] before accessing the file system entry (useful with UNIX domain sockets!). This call might affect all further operations of the socat process!

unlink-early

Unlinks (removes) the file before opening it and even before applying user-early etc.

unlink

Unlinks (removes) the file before accessing it, but after user-early etc.

unlink-late

Unlinks (removes) the file after opening it to make it inaccessible for other processes after a short race condition.

unlink-close

Removes the addresses file system entry when closing the address. For named pipes, listening unix domain sockets, and the symbolic links of pty addresses, the default is 1; for created files, opened files, generic opened files, and client unix domain sockets the default is 0.

OPEN option group

The OPEN group options allow to set flags with the open() system call. E.g., option `creat' sets the O_CREAT flag.
See also options append and nonblock.

creat=<bool>

Creates the file if it does not exist (example).

dsync=<bool>

Blocks write() calls until metainfo is physically written to media.

excl=<bool>

With option creat, if file exists this is an error.

largefile=<bool>

On 32 bit systems, allows a file larger than 2^31 bytes.

noatime

Sets the O_NOATIME options, so reads do not change the access timestamp.

noctty=<bool>

Does not make this file the controlling terminal.

nofollow=<bool>

Does not follow symbolic links.

nshare=<bool>

Does not allow to share this file with other processes.

rshare=<bool>

Does not allow other processes to open this file for writing.

rsync=<bool>

Blocks write() until metainfo is physically written to media.

sync=<bool>

Blocks write() until data is physically written to media.

rdonly=<bool>

Opens the file for reading only.

wronly=<bool>

Opens the file for writing only.

trunc

Truncates the file to size 0 during opening it.

REG and BLK option group

These options are usually applied to a UN*X file descriptor, but their semantics make sense only on a file supporting random access.

seek=<offset>

Applies the lseek(fd, <offset>, SEEK_SET) (or lseek64) system call, thus positioning the file pointer absolutely to <offset> [off_t or off64_t].

seek-cur=<offset>

Applies the lseek(fd, <offset>, SEEK_CUR) (or lseek64) system call, thus positioning the file pointer <offset> [off_t or off64_t] bytes relatively to its current position (which is usually 0).

seek-end=<offset>

Applies the lseek(fd, <offset>, SEEK_END) (or lseek64) system call, thus positioning the file pointer <offset> [off_t or off64_t] bytes relatively to the files current end.

ftruncate=<offset>

Applies the ftruncate(fd, <offset>) (or ftruncate64 if available) system call, thus truncating the file at the position <offset> [off_t or off64_t].

secrm=<bool>

unrm=<bool>

compr=<bool>

ext2-sync=<bool>

immutable=<bool>

ext2-append=<bool>

nodump=<bool>

ext2-noatime=<bool>

journal-data=<bool>

notail=<bool>

dirsync=<bool>

These options change non standard file attributes on operating systems and file systems that support these features, like Linux with ext2fs, ext3fs, or reiserfs. See man 1 chattr for information on these options. Please note that there might be a race condition between creating the file and applying these options.

PROCESS option group

Options of this group change the process properties instead of just affecting one data channel. For EXEC and SYSTEM addresses and for LISTEN and CONNECT type addresses with option FORK, these options apply to the child processes instead of the main socat process.

chroot=<directory>

Performs a chroot() operation to <directory> after processing the address (example). This call might require root privilege.

chroot-early=<directory>

Performs a chroot() operation to <directory> before opening the address. This call might require root privilege.

setgid=<group>

Changes the primary <group> of the process after processing the address. This call might require root privilege.

setgid-early=<group>

Changes the primary <group> of the process before opening the address. This call might require root privilege.

setuid=<user>

Changes the <user> (owner) of the process after processing the address. This call might require root privilege.

setuid-early=<user>

Changes the <user> (owner) of the process before opening the address. This call might require root privilege.

su=<user>

Changes the <user> (owner) and groups of the process after processing the address (example). This call might require root privilege.

su-d=<user>

Short name for substuser-delayed. Changes the <user> (owner) and groups of the process after processing the address (example). The user and his groups are retrieved before a possible chroot(). This call might require root privilege.

setpgid=<pid_t>

Makes the process a member of the specified process group <pid_t>. If no value is given, or if the value is 0 or 1, the process becomes leader of a new process group.

setsid

Makes the process the leader of a new session (example).

READLINE option group

These options apply to the readline address type.

history=<filename>

Reads and writes history from/to <filename> (example).

noprompt

Since version 1.4.0, socat per default tries to determine a prompt - that is then passed to the readline call - by remembering the last incomplete line of the output. With this option, socat does not pass a prompt to readline, so it begins line editing in the first column of the terminal.

noecho=<pattern>

Specifies a regular pattern for a prompt that prevents the following input line from being displayed on the screen and from being added to the history. The prompt is defined as the text that was output to the readline address after the lastest newline character and before an input character was typed. The pattern is a regular expression, e.g. "^[Pp]assword:.*$" or "([Uu]ser:|[Pp]assword:)". See regex(7) for details. (example)

prompt=<string>

Passes the string as prompt to the readline function. readline prints this prompt when stepping through the history. If this string matches a constant prompt issued by an interactive program on the other socat address, consistent look and feel can be archieved.

APPLICATION option group

This group contains options that work at data level. Note that these options only apply to the "raw" data transferred by socat, but not to protocol data used by addresses like PROXY.

cr

Converts the default line termination character NL ('\n', 0x0a) to/from CR ('\r', 0x0d) when writing/reading on this channel.

crnl

Converts the default line termination character NL ('\n', 0x0a) to/from CRNL ("\r\n", 0x0d0a) when writing/reading on this channel (example). Note: socat simply strips all CR characters.

ignoreeof

When EOF occurs on this channel, socat ignores it and tries to read more data (like "tail -f") (example).

readbytes=<bytes>

socat reads only so many bytes from this address (the address provides only so many bytes for transfer and pretends to be at EOF afterwards). Must be greater than 0.

lockfile=<filename>

If lockfile exists, exits with error. If lockfile does not exist, creates it and continues, unlinks lockfile on exit.

waitlock=<filename>

If lockfile exists, waits until it disappears. When lockfile does not exist, creates it and continues, unlinks lockfile on exit.

SOCKET option group

These options are intended for all kinds of sockets, e.g. IP or UNIX domain. Most are applied with a setsockopt() call.

bind=<sockname>

Binds the socket to the given socket address using the bind() system call. The form of <sockname> is socket domain dependent: IP4 and IP6 allow the form [hostname|hostaddress][:(service|port)] (example), UNIX domain sockets require <filename>.

connect-timeout=<seconds>

Abort the connection attempt after <seconds> [timeval] with error status.

interface=<interface>

Binds the socket to the given <interface>. This option might require root privilege.

broadcast

For datagram sockets, allows sending to broadcast addresses and receiving packets addressed to broadcast addresses.

bsdcompat

Emulates some (old?) bugs of the BSD socket implementation.

debug

Enables socket debugging.

dontroute

Only communicates with directly connected peers, does not use routers.

keepalive

Enables sending keepalives on the socket.

linger=<seconds>

Blocks shutdown() or close() until data transfers have finished or the given timeout [int] expired.

oobinline

Places out-of-band data in the input data stream.

priority=<priority>

Sets the protocol defined <priority> [<int>] for outgoing packets.

rcvbuf=<bytes>

Sets the size of the receive buffer after the socket() call to <bytes> [int]. With TCP sockets, this value corresponds to the socket's maximal window size.

rcvbuf-late=<bytes>

Sets the size of the receive buffer when the socket is already connected to <bytes> [int]. With TCP sockets, this value corresponds to the socket's maximal window size.

rcvlowat=<bytes>

Specifies the minimum number of received bytes [int] until the socket layer will pass the buffered data to socat.

rcvtimeo=<seconds>

Sets the receive timeout [timeval].

reuseaddr

Allows other sockets to bind to an address even if parts of it (e.g. the local port) are already in use by socat (example).

sndbuf=<bytes>

Sets the size of the send buffer after the socket() call to <bytes> [int].

sndbuf-late=<bytes>

Sets the size of the send buffer when the socket is connected to <bytes> [int].

sndlowat=<bytes>

Specifies the minimum number of bytes in the send buffer until the socket layer will send the data to <bytes> [int].

sndtimeo=<seconds>

Sets the send timeout to seconds [timeval].

type=<type>

Sets the type of the socket, usually as argument to the socket() or socketpair() call, to <type> [int]. Under Linux, 1 means stream oriented socket, 2 means datagram socket, and 3 means raw socket.

pf=<string>

Forces the use of the specified IP version. <string> can be something like "ip4" or "ip6".

UNIX option group

These options apply to UNIX domain based addresses.

unix-tightsocklen=[0|1]

On socket operations, pass a socket address length that does not include the whole struct sockaddr_un record but (besides other components) only the relevant part of the filename or abstract string. Default is 1.

IP4 and IP6 option groups

These options can be used with IPv4 and IPv6 based sockets.

tos=<tos>

Sets the TOS (type of service) field of outgoing packets to <tos> [byte] (see RFC 791).

ttl=<ttl>

Sets the TTL (time to live) field of outgoing packets to <ttl> [byte].

ipoptions=<data>

Sets IP options like source routing. Must be given in binary form, recommended format is a leading "x" followed by an even number of hex digits. This option may be used multiple times, data are appended. E.g., to connect to host 10.0.0.1 via some gateway using a loose source route, use the gateway as address parameter and set a loose source route using the option ipoptions=x8307040a000001.
IP options are defined in RFC 791.

mtudiscover=<0|1|2>

Takes 0, 1, 2 to never, want, or always use path MTU discover on this socket.

ip-add-membership=<multicast-address:interface-address>

ip-add-membership=<multicast-address:interface-name>

ip-add-membership=<multicast-address:interface-index>

ip-add-membership=<multicast-address:interface-address:interface-name>

ip-add-membership=<multicast-address:interface-address:interface-index>

Makes the socket member of the specified multicast group. This is currently only implemented for IPv4. The option takes the IP address of the multicast group and info about the desired network interface. The most common syntax is the first one, while the others are only available on systems that provide struct mreqn (Linux).
The indices of active network interfaces can be shown using the utility procan.

ip-multicast-if=<hostname>

Specifies hostname or address of the network interface to be used for multicast traffic.

ip-multicast-loop=<bool>

Specifies if outgoing multicast traffic should loop back to the interface.

ip-multicast-ttl=<byte>

Sets the TTL used for outgoing multicast traffic. Default is 1.

res-debug

res-aaonly

res-usevc

res-primary

res-igntc

res-recurse

res-defnames

res-stayopen

res-dnsrch

These options set the corresponding resolver (name resolution) option flags. Append "=0" to clear a default option. See man resolver(5) for more information on these options. Note: these options are valid only for the address they are applied to.

IP6 option group

These options can only be used on IPv6 based sockets. See IP options for options that can be applied to both IPv4 and IPv6 sockets.

ipv6only=<bool>

Sets the IPV6_V6ONLY socket option. If 0, the TCP stack will also accept connections using IPv4 protocol on the same port. The default is system dependent.

TCP option group

These options may be applied to TCP sockets. They work by invoking setsockopt() with the appropriate parameters.

cork

Doesn't send packets smaller than MSS (maximal segment size).

defer-accept

While listening, accepts connections only when data from the peer arrived.

keepcnt=<count>

Sets the number of keepalives before shutting down the socket to <count> [int].

keepidle=<seconds>

Sets the idle time before sending the first keepalive to <seconds> [int].

keepintvl=<seconds>

Sets the interval between two keepalives to <seconds> [int].

linger2=<seconds>

Sets the time to keep the socket in FIN-WAIT-2 state to <seconds> [int].

mss=<bytes>

Sets the MSS (maximum segment size) after the socket() call to <bytes> [int]. This value is then proposed to the peer with the SYN or SYN/ACK packet (example).

mss-late=<bytes>

Sets the MSS of the socket after connection has been established to <bytes> [int].

nodelay

Turns off the Nagle algorithm for measuring the RTT (round trip time).

rfc1323

Enables RFC1323 TCP options: TCP window scale, round-trip time measurement (RTTM), and protect against wrapped sequence numbers (PAWS) (AIX).

stdurg

Enables RFC1122 compliant urgent pointer handling (AIX).

syncnt=<count>

Sets the maximal number of SYN retransmits during connect to <count> [int].

md5sig

Enables generation of MD5 digests on the packets (FreeBSD).

noopt

Disables use of TCP options (FreeBSD, MacOSX).

nopush

sets the TCP_NOPUSH socket option (FreeBSD, MacOSX).

sack-disable

Disables use the selective acknowledge feature (OpenBSD).

signature-enable

Enables generation of MD5 digests on the packets (OpenBSD).

abort-threshold=<milliseconds>

Sets the time to wait for an answer of the peer on an established connection (HP-UX).

conn-abort-threshold=<milliseconds>

Sets the time to wait for an answer of the server during the initial connect (HP-UX).

keepinit

Sets the time to wait for an answer of the server during connect() before giving up. Value in half seconds, default is 150 (75s) (Tru64).

paws

Enables the "protect against wrapped sequence numbers" feature (Tru64).

sackena

Enables selective acknowledge (Tru64).

tsoptena

Enables the time stamp option that allows RTT recalculation on existing connections (Tru64).

UDP and TCP option groups

Here we find options that are related to the network port mechanism and that thus can be used with UDP and TCP, client and server addresses.

sourceport=<port>

For outgoing (client) TCP and UDP connections, it sets the source <port> using an extra bind() call. With TCP or UDP listen addresses, socat immediately shuts down the connection if the client does not use this sourceport (example).

lowport

Outgoing (client) TCP and UDP connections with this option use an unused random source port between 640 and 1023 incl. On UNIX class operating systems, this requires root privilege, and thus indicates that the client process is authorized by local root. TCP and UDP listen addresses with this option immediately shut down the connection if the client does not use a sourceport <= 1023. This mechanism can provide limited authorization under some circumstances.

SOCKS option group

When using SOCKS type addresses, some socks specific options can be set.

socksport=<tcp service>

Overrides the default "socks" service or port 1080 for the socks server port with <TCP service>.

socksuser=<user>

Sends the <user> [string] in the username field to the socks server. Default is the actual user name ($LOGNAME or $USER) (example).

HTTP option group

Options that can be provided with HTTP type addresses. The only HTTP address currently implemented is proxy-connect.

proxyport=<TCP service>

Overrides the default HTTP proxy port 8080 with <TCP service>.

ignorecr

The HTTP protocol requires the use of CR+NL as line terminator. When a proxy server violates this standard, socat might not understand its answer. This option directs socat to interprete NL as line terminator and to ignore CR in the answer. Nevertheless, socat sends CR+NL to the proxy.

proxyauth=<username>:<password>

Provide "basic" authentication to the proxy server. The argument to the option is used with a "Proxy-Authorization: Base" header in base64 encoded form.
Note: username and password are visible for every user on the local machine in the process list; username and password are transferred to the proxy server unencrypted (base64 encoded) and might be sniffed.

resolve

Per default, socat sends to the proxy a CONNECT request containing the target hostname. With this option, socat resolves the hostname locally and sends the IP address. Please note that, according to RFC 2396, only name resolution to IPv4 addresses is implemented.

RANGE option group

These options check if a connecting client should be granted access. They can be applied to listening and receiving network sockets. tcp-wrappers options fall into this group.

range=<address-range>

After accepting a connection, tests if the peer is within range. For IPv4 addresses, address-range takes the form address/bits, e.g. 10.0.0.0/8, or address:mask, e.g. 10.0.0.0:255.0.0.0 (example); for IPv6, it is [ip6-address/bits], e.g. [::1/128]. If the client address does not match, socat issues a warning and keeps listening/receiving.

tcpwrap[=<name>]

Uses Wietse Venema's libwrap (tcpd) library to determine if the client is allowed to connect. The configuration files are /etc/hosts.allow and /etc/hosts.deny per default, see "man 5 hosts_access" for more information. The optional <name> (type string) is passed to the wrapper functions as daemon process name (example). If omitted, the basename of socats invocation (argv[0]) is passed. If both tcpwrap and range options are applied to an address, both conditions must be fulfilled to allow the connection.

allow-table=<filename>

Takes the specified file instead of /etc/hosts.allow.

deny-table=<filename>

Takes the specified file instead of /etc/hosts.deny.

tcpwrap-etc=<directoryname>

Looks for hosts.allow and hosts.deny in the specified directory. Is overridden by options hosts-allow and hosts-deny.

LISTEN option group

Options specific to listening sockets.

backlog=<count>

Sets the backlog value passed with the listen() system call to <count> [int]. Default is 5.

CHILD option group

Options for addresses with multiple connections via child processes.

fork

After establishing a connection, handles its channel in a child process and keeps the parent process attempting to produce more connections, either by listening or by connecting in a loop (example).
SSL-CONNECT and SSL-LISTEN differ in when they actually fork off the child: SSL-LISTEN forks before the SSL handshake, while SSL-CONNECT forks afterwards. RETRY and FOREVER options are not inherited by the child process.

EXEC option group

Options for addresses that invoke a program.

path=<string>

Overrides the PATH environment variable for searching the program with <string>. This $PATH value is effective in the child process too.

login

Prefixes argv[0] for the execvp() call with '-', thus making a shell behave as login shell.

FORK option group

EXEC or SYSTEM addresses invoke a program using a child process and transfer data between socat and the program. The interprocess communication mechanism can be influenced with the following options. Per default, a socketpair() is created and assigned to stdin and stdout of the child process, while stderr is inherited from the socat process, and the child process uses file descriptors 0 and 1 for communicating with the main socat process.

nofork

Does not fork a subprocess for executing the program, instead calls execvp() or system() directly from the actual socat instance. This avoids the overhead of another process between the program and its peer, but introduces a lot of restrictions:

this option can only be applied to the second socat address.

it cannot be applied to a part of a dual address.

the first socat address cannot be OPENSSL or READLINE

socat options -b, -t, -D, -l, -v, -x become useless

for both addresses, options ignoreeof, cr, and crnl become useless

for the second address (the one with option nofork), options append, cloexec, flock, user, group, mode, nonblock, perm-late, setlk, and setpgid cannot be applied. Some of these could be used on the first address though.

pipes

Creates a pair of unnamed pipes for interprocess communication instead of a socket pair.

openpty

Establishes communication with the sub process using a pseudo terminal created with openpty() instead of the default (socketpair or ptmx).

ptmx

Establishes communication with the sub process using a pseudo terminal created by opening /dev/ptmx or /dev/ptc instead of the default (socketpair).

pty

Establishes communicat