Creating a daemon the Python way

                """Disk And Execution MONitor (Daemon)
Configurable daemon behaviors:
   1.) The current working directory set to the "/" directory.
   2.) The current file creation mode mask set to 0.
   3.) Close all open files (1024).
   4.) Redirect standard I/O streams to "/dev/null".
A failed call to fork() now raises an exception.
References:
   1) Advanced Programming in the Unix Environment: W. Richard Stevens
   2) Unix Programming Frequently Asked Questions:
         http://www.erlenstar.demon.co.uk/unix/faq_toc.html
"""
__author__ = "Chad J. Schroeder"
__copyright__ = "Copyright (C) 2005 Chad J. Schroeder"
__revision__ = "$Id$"
__version__ = "0.2"
# Standard Python modules.
import os               # Miscellaneous OS interfaces.
import sys              # System-specific parameters and functions.
# Default daemon parameters.
# File mode creation mask of the daemon.
UMASK = 0
# Default working directory for the daemon.
WORKDIR = "/"
# Default maximum for the number of available file descriptors.
MAXFD = 1024
# The standard I/O file descriptors are redirected to /dev/null by default.
if (hasattr(os, "devnull")):
   REDIRECT_TO = os.devnull
else:
   REDIRECT_TO = "/dev/null"
def createDaemon():
   """Detach a process from the controlling terminal and run it in the
   background as a daemon.
   """
   try:
      # Fork a child process so the parent can exit.  This returns control to
      # the command-line or shell.  It also guarantees that the child will not
      # be a process group leader, since the child receives a new process ID
      # and inherits the parent's process group ID.  This step is required
      # to insure that the next call to os.setsid is successful.
      pid = os.fork()
   except OSError, e:
      raise Exception, "%s [%d]" % (e.strerror, e.errno)
   if (pid == 0):        # The first child.
      # To become the session leader of this new session and the process group
      # leader of the new process group, we call os.setsid().  The process is
      # also guaranteed not to have a controlling terminal.
      os.setsid()
      # Is ignoring SIGHUP necessary?
      #
      # It's often suggested that the SIGHUP signal should be ignored before
      # the second fork to avoid premature termination of the process.  The
      # reason is that when the first child terminates, all processes, e.g.
      # the second child, in the orphaned group will be sent a SIGHUP.
      #
      # "However, as part of the session management system, there are exactly
      # two cases where SIGHUP is sent on the death of a process:
      #
      #   1) When the process that dies is the session leader of a session that
      #      is attached to a terminal device, SIGHUP is sent to all processes
      #      in the foreground process group of that terminal device.
      #   2) When the death of a process causes a process group to become
      #      orphaned, and one or more processes in the orphaned group are
      #      stopped, then SIGHUP and SIGCONT are sent to all members of the
      #      orphaned group." [2]
      #
      # The first case can be ignored since the child is guaranteed not to have
      # a controlling terminal.  The second case isn't so easy to dismiss.
      # The process group is orphaned when the first child terminates and
      # POSIX.1 requires that every STOPPED process in an orphaned process
      # group be sent a SIGHUP signal followed by a SIGCONT signal.  Since the
      # second child is not STOPPED though, we can safely forego ignoring the
      # SIGHUP signal.  In any case, there are no ill-effects if it is ignored.
      #
      # import signal           # Set handlers for asynchronous events.
      # signal.signal(signal.SIGHUP, signal.SIG_IGN)
      try:
         # Fork a second child and exit immediately to prevent zombies.  This
         # causes the second child process to be orphaned, making the init
         # process responsible for its cleanup.  And, since the first child is
         # a session leader without a controlling terminal, it's possible for
         # it to acquire one by opening a terminal in the future (System V-
         # based systems).  This second fork guarantees that the child is no
         # longer a session leader, preventing the daemon from ever acquiring
         # a controlling terminal.
         pid = os.fork()        # Fork a second child.
      except OSError, e:
         raise Exception, "%s [%d]" % (e.strerror, e.errno)
      if (pid == 0):        # The second child.
         # Since the current working directory may be a mounted filesystem, we
         # avoid the issue of not being able to unmount the filesystem at
         # shutdown time by changing it to the root directory.
         os.chdir(WORKDIR)
         # We probably don't want the file mode creation mask inherited from
         # the parent, so we give the child complete control over permissions.
         os.umask(UMASK)
      else:
         # exit() or _exit()?  See below.
         os._exit(0)        # Exit parent (the first child) of the second child.
   else:
      # exit() or _exit()?
      # _exit is like exit(), but it doesn't call any functions registered
      # with atexit (and on_exit) or any registered signal handlers.  It also
      # closes any open file descriptors.  Using exit() may cause all stdio
      # streams to be flushed twice and any temporary files may be unexpectedly
      # removed.  It's therefore recommended that child branches of a fork()
      # and the parent branch(es) of a daemon use _exit().
      os._exit(0)        # Exit parent of the first child.
   # Close all open file descriptors.  This prevents the child from keeping
   # open any file descriptors inherited from the parent.  There is a variety
   # of methods to accomplish this task.  Three are listed below.
   #
   # Try the system configuration variable, SC_OPEN_MAX, to obtain the maximum
   # number of open file descriptors to close.  If it doesn't exists, use
   # the default value (configurable).
   #
   # try:
   #    maxfd = os.sysconf("SC_OPEN_MAX")
   # except (AttributeError, ValueError):
   #    maxfd = MAXFD
   #
   # OR
   #
   # if (os.sysconf_names.has_key("SC_OPEN_MAX")):
   #    maxfd = os.sysconf("SC_OPEN_MAX")
   # else:
   #    maxfd = MAXFD
   #
   # OR
   #
   # Use the getrlimit method to retrieve the maximum file descriptor number
   # that can be opened by this process.  If there is not limit on the
   # resource, use the default value.
   #
   import resource                # Resource usage information.
   maxfd = resource.getrlimit(resource.RLIMIT_NOFILE)[1]
   if (maxfd == resource.RLIM_INFINITY):
      maxfd = MAXFD
  
   # Iterate through and close all file descriptors.
   for fd in range(0, maxfd):
      try:
         os.close(fd)
      except OSError:        # ERROR, fd wasn't open to begin with (ignored)
         pass
   # Redirect the standard I/O file descriptors to the specified file.  Since
   # the daemon has no controlling terminal, most daemons redirect stdin,
   # stdout, and stderr to /dev/null.  This is done to prevent side-effects
   # from reads and writes to the standard I/O file descriptors.
   # This call to open is guaranteed to return the lowest file descriptor,
   # which will be 0 (stdin), since it was closed above.
   os.open(REDIRECT_TO, os.O_RDWR)        # standard input (0)
   # Duplicate standard input to standard output and standard error.
   os.dup2(0, 1)                        # standard output (1)
   os.dup2(0, 2)                        # standard error (2)
   return(0)
if __name__ == "__main__":
   retCode = createDaemon()
   # The code, as is, will create a new file in the root directory, when
   # executed with superuser privileges.  The file will contain the following
   # daemon related process parameters: return code, process ID, parent
   # process group ID, session ID, user ID, effective user ID, real group ID,
   # and the effective group ID.  Notice the relationship between the daemon's
   # process ID, process group ID, and its parent's process ID.
   procParams = """
   return code = %s
   process ID = %s
   parent process ID = %s
   process group ID = %s
   session ID = %s
   user ID = %s
   effective user ID = %s
   real group ID = %s
   effective group ID = %s
   """ % (retCode, os.getpid(), os.getppid(), os.getpgrp(), os.getsid(0),
   os.getuid(), os.geteuid(), os.getgid(), os.getegid())
   open("createDaemon.log", "w").write(procParams + "\n")
   sys.exit(retCode)