18.1.1. Socket families

Depending on the system and the build options, various socket families are supported by this module.

The address format required by a particular socket object is automatically selected based on the address family specified when the socket object was created. Socket addresses are represented as follows:

• The address of an AF_UNIX socket bound to a file system node is represented as a string, using the file system encoding and the 'surrogateescape' error handler (see PEP 383). An address in Linux’s abstract namespace is returned as a bytes object with an initial null byte; note that sockets in this namespace can communicate with normal file system sockets, so programs intended to run on Linux may need to deal with both types of address. A string or bytes object can be used for either type of address when passing it as an argument.

  Changed in version 3.3: Previously, AF_UNIX socket paths were assumed to use UTF-8 encoding.

• A pair (host, port) is used for the AF_INET address family, where host is a string representing either a hostname in Internet domain notation like 'daring.cwi.nl' or an IPv4 address like '100.50.200.5', and port is an integer.

• For AF_INET6 address family, a four-tuple (host, port, flowinfo, scopeid) is used, where flowinfo and scopeid represent the sin6_flowinfo and sin6_scope_id members in struct sockaddr_in6 in C. For socket module methods, flowinfo and scopeid can be omitted just for backward compatibility. Note, however, omission of scopeid can cause problems in manipulating scoped IPv6 addresses.

• AF_NETLINK sockets are represented as pairs (pid, groups).

• Linux-only support for TIPC is available using the AF_TIPC address family. TIPC is an open, non-IP based networked protocol designed for use in clustered computer environments. Addresses are represented by a tuple, and the fields depend on the address type. The general tuple form is (addr_type, v1, v2, v3 [, scope]), where:

  • addr_type is one of TIPC_ADDR_NAMESEQ, TIPC_ADDR_NAME, or TIPC_ADDR_ID.

  • scope is one of TIPC_ZONE_SCOPE, TIPC_CLUSTER_SCOPE, and TIPC_NODE_SCOPE.

  • If addr_type is TIPC_ADDR_NAME, then v1 is the server type, v2 is the port identifier, and v3 should be 0.

    If addr_type is TIPC_ADDR_NAMESEQ, then v1 is the server type, v2 is the lower port number, and v3 is the upper port number.

    If addr_type is TIPC_ADDR_ID, then v1 is the node, v2 is the reference, and v3 should be set to 0.

    If addr_type is TIPC_ADDR_ID, then v1 is the node, v2 is the reference, and v3 should be set to 0.

• A tuple (interface, ) is used for the AF_CAN address family, where interface is a string representing a network interface name like 'can0'. The network interface name '' can be used to receive packets from all network interfaces of this family.

• A string or a tuple (id, unit) is used for the SYSPROTO_CONTROL protocol of the PF_SYSTEM family. The string is the name of a kernel control using a dynamically-assigned ID. The tuple can be used if ID and unit number of the kernel control are known or if a registered ID is used.

  New in version 3.3.

• Certain other address families (AF_BLUETOOTH, AF_PACKET, AF_CAN) support specific representations.

For IPv4 addresses, two special forms are accepted instead of a host address: the empty string represents INADDR_ANY, and the string '<broadcast>' represents INADDR_BROADCAST. This behavior is not compatible with IPv6, therefore, you may want to avoid these if you intend to support IPv6 with your Python programs.

If you use a hostname in the host portion of IPv4/v6 socket address, the program may show a nondeterministic behavior, as Python uses the first address returned from the DNS resolution. The socket address will be resolved differently into an actual IPv4/v6 address, depending on the results from DNS resolution and/or the host configuration. For deterministic behavior use a numeric address in host portion.

All errors raise exceptions. The normal exceptions for invalid argument types and out-of-memory conditions can be raised; starting from Python 3.3, errors related to socket or address semantics raise OSError or one of its subclasses (they used to raise socket.error).

Non-blocking mode is supported through setblocking(). A generalization of this based on timeouts is supported through settimeout().

18.1.2. Module contents

The module socket exports the following elements.

18.1.3. Socket Objects

Socket objects have the following methods. Except for makefile(), these correspond to Unix system calls applicable to sockets.

socket.accept()

Accept a connection. The socket must be bound to an address and listening for connections. The return value is a pair (conn, address) where conn is a new socket object usable to send and receive data on the connection, and address is the address bound to the socket on the other end of the connection.

The newly created socket is non-inheritable.

Changed in version 3.4: The socket is now non-inheritable.

socket.bind(address)

Bind the socket to address. The socket must not already be bound. (The format of address depends on the address family — see above.)

socket.close()

Mark the socket closed. The underlying system resource (e.g. a file descriptor) is also closed when all file objects from makefile() are closed. Once that happens, all future operations on the socket object will fail. The remote end will receive no more data (after queued data is flushed).

Sockets are automatically closed when they are garbage-collected, but it is recommended to close() them explicitly, or to use a with statement around them.

Note

close() releases the resource associated with a connection but does not necessarily close the connection immediately. If you want to close the connection in a timely fashion, call shutdown() before close().

socket.connect(address)

Connect to a remote socket at address. (The format of address depends on the address family — see above.)

socket.connect_ex(address)

Like connect(address), but return an error indicator instead of raising an exception for errors returned by the C-level connect() call (other problems, such as “host not found,” can still raise exceptions). The error indicator is 0 if the operation succeeded, otherwise the value of the errno variable. This is useful to support, for example, asynchronous connects.

socket.detach()

Put the socket object into closed state without actually closing the underlying file descriptor. The file descriptor is returned, and can be reused for other purposes.

New in version 3.2.

socket.dup()

Duplicate the socket.

The newly created socket is non-inheritable.

Changed in version 3.4: The socket is now non-inheritable.

socket.fileno()

Return the socket’s file descriptor (a small integer). This is useful with select.select().

Under Windows the small integer returned by this method cannot be used where a file descriptor can be used (such as os.fdopen()). Unix does not have this limitation.

socket.get_inheritable()

Get the inheritable flag of the socket’s file descriptor or socket’s handle: True if the socket can be inherited in child processes, False if it cannot.

New in version 3.4.

socket.getpeername()

Return the remote address to which the socket is connected. This is useful to find out the port number of a remote IPv4/v6 socket, for instance. (The format of the address returned depends on the address family — see above.) On some systems this function is not supported.

socket.getsockname()

Return the socket’s own address. This is useful to find out the port number of an IPv4/v6 socket, for instance. (The format of the address returned depends on the address family — see above.)

socket.getsockopt(level, optname[, buflen])

Return the value of the given socket option (see the Unix man page getsockopt(2)). The needed symbolic constants (SO_* etc.) are defined in this module. If buflen is absent, an integer option is assumed and its integer value is returned by the function. If buflen is present, it specifies the maximum length of the buffer used to receive the option in, and this buffer is returned as a bytes object. It is up to the caller to decode the contents of the buffer (see the optional built-in module struct for a way to decode C structures encoded as byte strings).

socket.gettimeout()

Return the timeout in seconds (float) associated with socket operations, or None if no timeout is set. This reflects the last call to setblocking() or settimeout().

socket.ioctl(control, option)

Platform:	Windows

The ioctl() method is a limited interface to the WSAIoctl system interface. Please refer to the Win32 documentation for more information.

On other platforms, the generic fcntl.fcntl() and fcntl.ioctl() functions may be used; they accept a socket object as their first argument.

socket.listen(backlog)

Listen for connections made to the socket. The backlog argument specifies the maximum number of queued connections and should be at least 0; the maximum value is system-dependent (usually 5), the minimum value is forced to 0.

socket.makefile(mode='r', buffering=None, *, encoding=None, errors=None, newline=None)

Return a file object associated with the socket. The exact returned type depends on the arguments given to makefile(). These arguments are interpreted the same way as by the built-in open() function.

The socket must be in blocking mode; it can have a timeout, but the file object’s internal buffer may end up in a inconsistent state if a timeout occurs.

Closing the file object returned by makefile() won’t close the original socket unless all other file objects have been closed and socket.close() has been called on the socket object.

Note

On Windows, the file-like object created by makefile() cannot be used where a file object with a file descriptor is expected, such as the stream arguments of subprocess.Popen().

socket.recv(bufsize[, flags])

Receive data from the socket. The return value is a bytes object representing the data received. The maximum amount of data to be received at once is specified by bufsize. See the Unix manual page recv(2) for the meaning of the optional argument flags; it defaults to zero.

Note

For best match with hardware and network realities, the value of bufsize should be a relatively small power of 2, for example, 4096.

socket.recvfrom(bufsize[, flags])

Receive data from the socket. The return value is a pair (bytes, address) where bytes is a bytes object representing the data received and address is the address of the socket sending the data. See the Unix manual page recv(2) for the meaning of the optional argument flags; it defaults to zero. (The format of address depends on the address family — see above.)

socket.recvmsg(bufsize[, ancbufsize[, flags]])

Receive normal data (up to bufsize bytes) and ancillary data from the socket. The ancbufsize argument sets the size in bytes of the internal buffer used to receive the ancillary data; it defaults to 0, meaning that no ancillary data will be received. Appropriate buffer sizes for ancillary data can be calculated using CMSG_SPACE() or CMSG_LEN(), and items which do not fit into the buffer might be truncated or discarded. The flags argument defaults to 0 and has the same meaning as for recv().

The return value is a 4-tuple: (data, ancdata, msg_flags, address). The data item is a bytes object holding the non-ancillary data received. The ancdata item is a list of zero or more tuples (cmsg_level, cmsg_type, cmsg_data) representing the ancillary data (control messages) received: cmsg_level and cmsg_type are integers specifying the protocol level and protocol-specific type respectively, and cmsg_data is a bytes object holding the associated data. The msg_flags item is the bitwise OR of various flags indicating conditions on the received message; see your system documentation for details. If the receiving socket is unconnected, address is the address of the sending socket, if available; otherwise, its value is unspecified.

On some systems, sendmsg() and recvmsg() can be used to pass file descriptors between processes over an AF_UNIX socket. When this facility is used (it is often restricted to SOCK_STREAM sockets), recvmsg() will return, in its ancillary data, items of the form (socket.SOL_SOCKET, socket.SCM_RIGHTS, fds), where fds is a bytes object representing the new file descriptors as a binary array of the native C int type. If recvmsg() raises an exception after the system call returns, it will first attempt to close any file descriptors received via this mechanism.

Some systems do not indicate the truncated length of ancillary data items which have been only partially received. If an item appears to extend beyond the end of the buffer, recvmsg() will issue a RuntimeWarning, and will return the part of it which is inside the buffer provided it has not been truncated before the start of its associated data.

On systems which support the SCM_RIGHTS mechanism, the following function will receive up to maxfds file descriptors, returning the message data and a list containing the descriptors (while ignoring unexpected conditions such as unrelated control messages being received). See also sendmsg().

import socket, array

def recv_fds(sock, msglen, maxfds):
    fds = array.array("i")   # Array of ints
    msg, ancdata, flags, addr = sock.recvmsg(msglen, socket.CMSG_LEN(maxfds * fds.itemsize))
    for cmsg_level, cmsg_type, cmsg_data in ancdata:
        if (cmsg_level == socket.SOL_SOCKET and cmsg_type == socket.SCM_RIGHTS):
            # Append data, ignoring any truncated integers at the end.
            fds.fromstring(cmsg_data[:len(cmsg_data) - (len(cmsg_data) % fds.itemsize)])
    return msg, list(fds)

Availability: most Unix platforms, possibly others.

New in version 3.3.

socket.recvmsg_into(buffers[, ancbufsize[, flags]])

Receive normal data and ancillary data from the socket, behaving as recvmsg() would, but scatter the non-ancillary data into a series of buffers instead of returning a new bytes object. The buffers argument must be an iterable of objects that export writable buffers (e.g. bytearray objects); these will be filled with successive chunks of the non-ancillary data until it has all been written or there are no more buffers. The operating system may set a limit (sysconf() value SC_IOV_MAX) on the number of buffers that can be used. The ancbufsize and flags arguments have the same meaning as for recvmsg().

The return value is a 4-tuple: (nbytes, ancdata, msg_flags, address), where nbytes is the total number of bytes of non-ancillary data written into the buffers, and ancdata, msg_flags and address are the same as for recvmsg().

Example:

>>> import socket
>>> s1, s2 = socket.socketpair()
>>> b1 = bytearray(b'----')
>>> b2 = bytearray(b'0123456789')
>>> b3 = bytearray(b'--------------')
>>> s1.send(b'Mary had a little lamb')
22
>>> s2.recvmsg_into([b1, memoryview(b2)[2:9], b3])
(22, [], 0, None)
>>> [b1, b2, b3]
[bytearray(b'Mary'), bytearray(b'01 had a 9'), bytearray(b'little lamb---')]

Availability: most Unix platforms, possibly others.

New in version 3.3.

socket.recvfrom_into(buffer[, nbytes[, flags]])

Receive data from the socket, writing it into buffer instead of creating a new bytestring. The return value is a pair (nbytes, address) where nbytes is the number of bytes received and address is the address of the socket sending the data. See the Unix manual page recv(2) for the meaning of the optional argument flags; it defaults to zero. (The format of address depends on the address family — see above.)

socket.recv_into(buffer[, nbytes[, flags]])

Receive up to nbytes bytes from the socket, storing the data into a buffer rather than creating a new bytestring. If nbytes is not specified (or 0), receive up to the size available in the given buffer. Returns the number of bytes received. See the Unix manual page recv(2) for the meaning of the optional argument flags; it defaults to zero.

socket.send(bytes[, flags])

Send data to the socket. The socket must be connected to a remote socket. The optional flags argument has the same meaning as for recv() above. Returns the number of bytes sent. Applications are responsible for checking that all data has been sent; if only some of the data was transmitted, the application needs to attempt delivery of the remaining data. For further information on this topic, consult the Socket Programming HOWTO.

socket.sendall(bytes[, flags])

Send data to the socket. The socket must be connected to a remote socket. The optional flags argument has the same meaning as for recv() above. Unlike send(), this method continues to send data from bytes until either all data has been sent or an error occurs. None is returned on success. On error, an exception is raised, and there is no way to determine how much data, if any, was successfully sent.

socket.sendto(bytes, address)

socket.sendto(bytes, flags, address)

Send data to the socket. The socket should not be connected to a remote socket, since the destination socket is specified by address. The optional flags argument has the same meaning as for recv() above. Return the number of bytes sent. (The format of address depends on the address family — see above.)

socket.sendmsg(buffers[, ancdata[, flags[, address]]])

Send normal and ancillary data to the socket, gathering the non-ancillary data from a series of buffers and concatenating it into a single message. The buffers argument specifies the non-ancillary data as an iterable of buffer-compatible objects (e.g. bytes objects); the operating system may set a limit (sysconf() value SC_IOV_MAX) on the number of buffers that can be used. The ancdata argument specifies the ancillary data (control messages) as an iterable of zero or more tuples (cmsg_level, cmsg_type, cmsg_data), where cmsg_level and cmsg_type are integers specifying the protocol level and protocol-specific type respectively, and cmsg_data is a buffer-compatible object holding the associated data. Note that some systems (in particular, systems without CMSG_SPACE()) might support sending only one control message per call. The flags argument defaults to 0 and has the same meaning as for send(). If address is supplied and not None, it sets a destination address for the message. The return value is the number of bytes of non-ancillary data sent.

The following function sends the list of file descriptors fds over an AF_UNIX socket, on systems which support the SCM_RIGHTS mechanism. See also recvmsg().

import socket, array

def send_fds(sock, msg, fds):
    return sock.sendmsg([msg], [(socket.SOL_SOCKET, socket.SCM_RIGHTS, array.array("i", fds))])

Availability: most Unix platforms, possibly others.

New in version 3.3.

socket.set_inheritable(inheritable)

Set the inheritable flag of the socket’s file descriptor or socket’s handle.

New in version 3.4.

socket.setblocking(flag)

Set blocking or non-blocking mode of the socket: if flag is false, the socket is set to non-blocking, else to blocking mode.

This method is a shorthand for certain settimeout() calls:

• sock.setblocking(True) is equivalent to sock.settimeout(None)
• sock.setblocking(False) is equivalent to sock.settimeout(0.0)

socket.settimeout(value)

Set a timeout on blocking socket operations. The value argument can be a nonnegative floating point number expressing seconds, or None. If a non-zero value is given, subsequent socket operations will raise a timeout exception if the timeout period value has elapsed before the operation has completed. If zero is given, the socket is put in non-blocking mode. If None is given, the socket is put in blocking mode.

For further information, please consult the notes on socket timeouts.

socket.setsockopt(level, optname, value)

Set the value of the given socket option (see the Unix manual page setsockopt(2)). The needed symbolic constants are defined in the socket module (SO_* etc.). The value can be an integer or a bytes object representing a buffer. In the latter case it is up to the caller to ensure that the bytestring contains the proper bits (see the optional built-in module struct for a way to encode C structures as bytestrings).

socket.shutdown(how)

Shut down one or both halves of the connection. If how is SHUT_RD, further receives are disallowed. If how is SHUT_WR, further sends are disallowed. If how is SHUT_RDWR, further sends and receives are disallowed.

socket.share(process_id)

Duplicate a socket and prepare it for sharing with a target process. The target process must be provided with process_id. The resulting bytes object can then be passed to the target process using some form of interprocess communication and the socket can be recreated there using fromshare(). Once this method has been called, it is safe to close the socket since the operating system has already duplicated it for the target process.

Availability: Windows.

New in version 3.3.

Note that there are no methods read() or write(); use recv() and send() without flags argument instead.

Socket objects also have these (read-only) attributes that correspond to the values given to the socket constructor.

socket.family

The socket family.

socket.type

The socket type.

socket.proto

The socket protocol.

18.1.4. Notes on socket timeouts

A socket object can be in one of three modes: blocking, non-blocking, or timeout. Sockets are by default always created in blocking mode, but this can be changed by calling setdefaulttimeout().

• In blocking mode, operations block until complete or the system returns an error (such as connection timed out).
• In non-blocking mode, operations fail (with an error that is unfortunately system-dependent) if they cannot be completed immediately: functions from the select can be used to know when and whether a socket is available for reading or writing.
• In timeout mode, operations fail if they cannot be completed within the timeout specified for the socket (they raise a timeout exception) or if the system returns an error.

18.1.5. Example

Here are four minimal example programs using the TCP/IP protocol: a server that echoes all data that it receives back (servicing only one client), and a client using it. Note that a server must perform the sequence socket(), bind(), listen(), accept() (possibly repeating the accept() to service more than one client), while a client only needs the sequence socket(), connect(). Also note that the server does not sendall()/recv() on the socket it is listening on but on the new socket returned by accept().

The first two examples support IPv4 only.

# Echo server program
import socket

HOST = ''                 # Symbolic name meaning all available interfaces
PORT = 50007              # Arbitrary non-privileged port
s = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
s.bind((HOST, PORT))
s.listen(1)
conn, addr = s.accept()
print('Connected by', addr)
while True:
    data = conn.recv(1024)
    if not data: break
    conn.sendall(data)
conn.close()

# Echo client program
import socket

HOST = 'daring.cwi.nl'    # The remote host
PORT = 50007              # The same port as used by the server
s = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
s.connect((HOST, PORT))
s.sendall(b'Hello, world')
data = s.recv(1024)
s.close()
print('Received', repr(data))

The next two examples are identical to the above two, but support both IPv4 and IPv6. The server side will listen to the first address family available (it should listen to both instead). On most of IPv6-ready systems, IPv6 will take precedence and the server may not accept IPv4 traffic. The client side will try to connect to the all addresses returned as a result of the name resolution, and sends traffic to the first one connected successfully.

# Echo server program
import socket
import sys

HOST = None               # Symbolic name meaning all available interfaces
PORT = 50007              # Arbitrary non-privileged port
s = None
for res in socket.getaddrinfo(HOST, PORT, socket.AF_UNSPEC,
                              socket.SOCK_STREAM, 0, socket.AI_PASSIVE):
    af, socktype, proto, canonname, sa = res
    try:
        s = socket.socket(af, socktype, proto)
    except OSError as msg:
        s = None
        continue
    try:
        s.bind(sa)
        s.listen(1)
    except OSError as msg:
        s.close()
        s = None
        continue
    break
if s is None:
    print('could not open socket')
    sys.exit(1)
conn, addr = s.accept()
print('Connected by', addr)
while True:
    data = conn.recv(1024)
    if not data: break
    conn.send(data)
conn.close()

# Echo client program
import socket
import sys

HOST = 'daring.cwi.nl'    # The remote host
PORT = 50007              # The same port as used by the server
s = None
for res in socket.getaddrinfo(HOST, PORT, socket.AF_UNSPEC, socket.SOCK_STREAM):
    af, socktype, proto, canonname, sa = res
    try:
        s = socket.socket(af, socktype, proto)
    except OSError as msg:
        s = None
        continue
    try:
        s.connect(sa)
    except OSError as msg:
        s.close()
        s = None
        continue
    break
if s is None:
    print('could not open socket')
    sys.exit(1)
s.sendall(b'Hello, world')
data = s.recv(1024)
s.close()
print('Received', repr(data))

The next example shows how to write a very simple network sniffer with raw sockets on Windows. The example requires administrator privileges to modify the interface:

import socket

# the public network interface
HOST = socket.gethostbyname(socket.gethostname())

# create a raw socket and bind it to the public interface
s = socket.socket(socket.AF_INET, socket.SOCK_RAW, socket.IPPROTO_IP)
s.bind((HOST, 0))

# Include IP headers
s.setsockopt(socket.IPPROTO_IP, socket.IP_HDRINCL, 1)

# receive all packages
s.ioctl(socket.SIO_RCVALL, socket.RCVALL_ON)

# receive a package
print(s.recvfrom(65565))

# disabled promiscuous mode
s.ioctl(socket.SIO_RCVALL, socket.RCVALL_OFF)

The last example shows how to use the socket interface to communicate to a CAN network using the raw socket protocol. To use CAN with the broadcast manager protocol instead, open a socket with:

socket.socket(socket.AF_CAN, socket.SOCK_DGRAM, socket.CAN_BCM)

After binding (CAN_RAW) or connecting (CAN_BCM) the socket, you can use the socket.send(), and the socket.recv() operations (and their counterparts) on the socket object as usual.

This example might require special priviledge:

import socket
import struct


# CAN frame packing/unpacking (see 'struct can_frame' in <linux/can.h>)

can_frame_fmt = "=IB3x8s"
can_frame_size = struct.calcsize(can_frame_fmt)

def build_can_frame(can_id, data):
    can_dlc = len(data)
    data = data.ljust(8, b'\x00')
    return struct.pack(can_frame_fmt, can_id, can_dlc, data)

def dissect_can_frame(frame):
    can_id, can_dlc, data = struct.unpack(can_frame_fmt, frame)
    return (can_id, can_dlc, data[:can_dlc])


# create a raw socket and bind it to the 'vcan0' interface
s = socket.socket(socket.AF_CAN, socket.SOCK_RAW, socket.CAN_RAW)
s.bind(('vcan0',))

while True:
    cf, addr = s.recvfrom(can_frame_size)

    print('Received: can_id=%x, can_dlc=%x, data=%s' % dissect_can_frame(cf))

    try:
        s.send(cf)
    except OSError:
        print('Error sending CAN frame')

    try:
        s.send(build_can_frame(0x01, b'\x01\x02\x03'))
    except OSError:
        print('Error sending CAN frame')

Running an example several times with too small delay between executions, could lead to this error:

OSError: [Errno 98] Address already in use

This is because the previous execution has left the socket in a TIME_WAIT state, and can’t be immediately reused.

There is a socket flag to set, in order to prevent this, socket.SO_REUSEADDR:

s = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
s.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
s.bind((HOST, PORT))

the SO_REUSEADDR flag tells the kernel to reuse a local socket in TIME_WAIT state, without waiting for its natural timeout to expire.