8.11 fcntl -- The fcntl() and ioctl() system calls

Python PEP

8.11 fcntl -- The fcntl() and ioctl() system calls

Availability: Unix.

This module performs file control and I/O control on file descriptors. It is an interface to the fcntl() and ioctl() Unix routines.

All functions in this module take a file descriptor fd as their first argument. This can be an integer file descriptor, such as returned by sys.stdin.fileno(), or a file object, such as sys.stdin itself, which provides a fileno() which returns a genuine file descriptor.

The module defines the following functions:

Perform the requested operation on file descriptor fd (file objects providing a fileno() method are accepted as well). The operation is defined by op and is operating system dependent. These codes are also found in the fcntl module. The argument arg is optional, and defaults to the integer value 0. When present, it can either be an integer value, or a string. With the argument missing or an integer value, the return value of this function is the integer return value of the C fcntl() call. When the argument is a string it represents a binary structure, e.g. created by struct.pack(). The binary data is copied to a buffer whose address is passed to the C fcntl() call. The return value after a successful call is the contents of the buffer, converted to a string object. The length of the returned string will be the same as the length of the arg argument. This is limited to 1024 bytes. If the information returned in the buffer by the operating system is larger than 1024 bytes, this is most likely to result in a segmentation violation or a more subtle data corruption.

If the fcntl() fails, an IOError is raised.

This function is identical to the fcntl() function, except that the operations are typically defined in the library module termios and the argument handling is even more complicated.

The parameter arg can be one of an integer, absent (treated identically to the integer 0), an object supporting the read-only buffer interface (most likely a plain Python string) or an object supporting the read-write buffer interface.

In all but the last case, behaviour is as for the fcntl() function.

If a mutable buffer is passed, then the behaviour is determined by the value of the mutate_flag parameter.

If it is false, the buffer's mutability is ignored and behaviour is as for a read-only buffer, except that the 1024 byte limit mentioned above is avoided - so long as the buffer you pass is longer than what the operating system wants to put there, things should work.

If mutate_flag is true, then the buffer is (in effect) passed to the underlying ioctl() system call, the latter's return code is passed back to the calling Python, and the buffer's new contents reflect the action of the ioctl(). This is a slight simplification, because if the supplied buffer is less than 1024 bytes long it is first copied into a static buffer 1024 bytes long which is then passed to ioctl() and copied back into the supplied buffer.

If mutate_flag is not supplied, then in 2.3 it defaults to false. This is planned to change over the next few Python versions: in 2.4 failing to supply mutate_flag will get a warning but the same behavior and in versions later than 2.5 it will default to true.

An example:

>>> import array, fcntl, struct, termios, os
>>> os.getpgrp()
13341
>>> struct.unpack('h', fcntl.ioctl(0, termios.TIOCGPGRP, "  "))[0]
13341
>>> buf = array.array('h', [0])
>>> fcntl.ioctl(0, termios.TIOCGPGRP, buf, 1)
0
>>> buf
array('h', [13341])

Perform the lock operation op on file descriptor fd (file objects providing a fileno() method are accepted as well). See the Unix manual flock(3) for details. (On some systems, this function is emulated using fcntl().)

This is essentially a wrapper around the fcntl() locking calls. fd is the file descriptor of the file to lock or unlock, and operation is one of the following values:

  • LOCK_UN - unlock
  • LOCK_SH - acquire a shared lock
  • LOCK_EX - acquire an exclusive lock

When operation is LOCK_SH or LOCK_EX, it can also be bit-wise OR'd with LOCK_NB to avoid blocking on lock acquisition. If LOCK_NB is used and the lock cannot be acquired, an IOError will be raised and the exception will have an errno attribute set to EACCES or EAGAIN (depending on the operating system; for portability, check for both values). On at least some systems, LOCK_EX can only be used if the file descriptor refers to a file opened for writing.

length is the number of bytes to lock, start is the byte offset at which the lock starts, relative to whence, and whence is as with fileobj.seek(), specifically:

  • 0 - relative to the start of the file (SEEK_SET)
  • 1 - relative to the current buffer position (SEEK_CUR)
  • 2 - relative to the end of the file (SEEK_END)

The default for start is 0, which means to start at the beginning of the file. The default for length is 0 which means to lock to the end of the file. The default for whence is also 0.

Examples (all on a SVR4 compliant system):

import struct, fcntl, os

f = open(...)
rv = fcntl.fcntl(f, fcntl.F_SETFL, os.O_NDELAY)

lockdata = struct.pack('hhllhh', fcntl.F_WRLCK, 0, 0, 0, 0, 0)
rv = fcntl.fcntl(f, fcntl.F_SETLKW, lockdata)

Note that in the first example the return value variable rv will hold an integer value; in the second example it will hold a string value. The structure lay-out for the lockdata variable is system dependent -- therefore using the flock() call may be better.

See Also:

The os.open() function supports locking flags and is available on a wider variety of platforms than the lockf() and flock() functions, providing a more platform-independent file locking facility.
See About this document... for information on suggesting changes.