3 perlipc - Perl interprocess communication (signals, fifos, pipes, safe subprocesses, sockets, and semaphores)
7 The basic IPC facilities of Perl are built out of the good old Unix
8 signals, named pipes, pipe opens, the Berkeley socket routines, and SysV
9 IPC calls. Each is used in slightly different situations.
13 Perl uses a simple signal handling model: the %SIG hash contains names
14 or references of user-installed signal handlers. These handlers will
15 be called with an argument which is the name of the signal that
16 triggered it. A signal may be generated intentionally from a
17 particular keyboard sequence like control-C or control-Z, sent to you
18 from another process, or triggered automatically by the kernel when
19 special events transpire, like a child process exiting, your process
20 running out of stack space, or hitting file size limit.
22 For example, to trap an interrupt signal, set up a handler like this:
27 die "Somebody sent me a SIG$signame";
29 $SIG{INT} = 'catch_zap'; # could fail in modules
30 $SIG{INT} = \&catch_zap; # best strategy
32 Prior to Perl 5.7.3 it was necessary to do as little as you possibly
33 could in your handler; notice how all we do is set a global variable
34 and then raise an exception. That's because on most systems,
35 libraries are not re-entrant; particularly, memory allocation and I/O
36 routines are not. That meant that doing nearly I<anything> in your
37 handler could in theory trigger a memory fault and subsequent core
38 dump - see L</Deferred Signals (Safe Signals)> below.
40 The names of the signals are the ones listed out by C<kill -l> on your
41 system, or you can retrieve them from the Config module. Set up an
42 @signame list indexed by number to get the name and a %signo table
43 indexed by name to get the number:
46 defined $Config{sig_name} || die "No sigs?";
47 foreach $name (split(' ', $Config{sig_name})) {
53 So to check whether signal 17 and SIGALRM were the same, do just this:
55 print "signal #17 = $signame[17]\n";
57 print "SIGALRM is $signo{ALRM}\n";
60 You may also choose to assign the strings C<'IGNORE'> or C<'DEFAULT'> as
61 the handler, in which case Perl will try to discard the signal or do the
64 On most Unix platforms, the C<CHLD> (sometimes also known as C<CLD>) signal
65 has special behavior with respect to a value of C<'IGNORE'>.
66 Setting C<$SIG{CHLD}> to C<'IGNORE'> on such a platform has the effect of
67 not creating zombie processes when the parent process fails to C<wait()>
68 on its child processes (i.e. child processes are automatically reaped).
69 Calling C<wait()> with C<$SIG{CHLD}> set to C<'IGNORE'> usually returns
70 C<-1> on such platforms.
72 Some signals can be neither trapped nor ignored, such as
73 the KILL and STOP (but not the TSTP) signals. One strategy for
74 temporarily ignoring signals is to use a local() statement, which will be
75 automatically restored once your block is exited. (Remember that local()
76 values are "inherited" by functions called from within that block.)
79 local $SIG{INT} = 'IGNORE';
83 # interrupts still ignored, for now...
86 Sending a signal to a negative process ID means that you send the signal
87 to the entire Unix process-group. This code sends a hang-up signal to all
88 processes in the current process group (and sets $SIG{HUP} to IGNORE so
89 it doesn't kill itself):
92 local $SIG{HUP} = 'IGNORE';
94 # snazzy writing of: kill('HUP', -$$)
97 Another interesting signal to send is signal number zero. This doesn't
98 actually affect a child process, but instead checks whether it's alive
99 or has changed its UID.
101 unless (kill 0 => $kid_pid) {
102 warn "something wicked happened to $kid_pid";
105 When directed at a process whose UID is not identical to that
106 of the sending process, signal number zero may fail because
107 you lack permission to send the signal, even though the process is alive.
108 You may be able to determine the cause of failure using C<%!>.
110 unless (kill 0 => $pid or $!{EPERM}) {
111 warn "$pid looks dead";
114 You might also want to employ anonymous functions for simple signal
117 $SIG{INT} = sub { die "\nOutta here!\n" };
119 But that will be problematic for the more complicated handlers that need
120 to reinstall themselves. Because Perl's signal mechanism is currently
121 based on the signal(3) function from the C library, you may sometimes be so
122 unfortunate as to run on systems where that function is "broken", that
123 is, it behaves in the old unreliable SysV way rather than the newer, more
124 reasonable BSD and POSIX fashion. So you'll see defensive people writing
125 signal handlers like this:
129 # loathe SysV: it makes us not only reinstate
130 # the handler, but place it after the wait
131 $SIG{CHLD} = \&REAPER;
133 $SIG{CHLD} = \&REAPER;
134 # now do something that forks...
138 use POSIX ":sys_wait_h";
141 # If a second child dies while in the signal handler caused by the
142 # first death, we won't get another signal. So must loop here else
143 # we will leave the unreaped child as a zombie. And the next time
144 # two children die we get another zombie. And so on.
145 while (($child = waitpid(-1,WNOHANG)) > 0) {
146 $Kid_Status{$child} = $?;
148 $SIG{CHLD} = \&REAPER; # still loathe SysV
150 $SIG{CHLD} = \&REAPER;
151 # do something that forks...
153 Signal handling is also used for timeouts in Unix, While safely
154 protected within an C<eval{}> block, you set a signal handler to trap
155 alarm signals and then schedule to have one delivered to you in some
156 number of seconds. Then try your blocking operation, clearing the alarm
157 when it's done but not before you've exited your C<eval{}> block. If it
158 goes off, you'll use die() to jump out of the block, much as you might
159 using longjmp() or throw() in other languages.
164 local $SIG{ALRM} = sub { die "alarm clock restart" };
166 flock(FH, 2); # blocking write lock
169 if ($@ and $@ !~ /alarm clock restart/) { die }
171 If the operation being timed out is system() or qx(), this technique
172 is liable to generate zombies. If this matters to you, you'll
173 need to do your own fork() and exec(), and kill the errant child process.
175 For more complex signal handling, you might see the standard POSIX
176 module. Lamentably, this is almost entirely undocumented, but
177 the F<t/lib/posix.t> file from the Perl source distribution has some
180 =head2 Handling the SIGHUP Signal in Daemons
182 A process that usually starts when the system boots and shuts down
183 when the system is shut down is called a daemon (Disk And Execution
184 MONitor). If a daemon process has a configuration file which is
185 modified after the process has been started, there should be a way to
186 tell that process to re-read its configuration file, without stopping
187 the process. Many daemons provide this mechanism using the C<SIGHUP>
188 signal handler. When you want to tell the daemon to re-read the file
189 you simply send it the C<SIGHUP> signal.
191 Not all platforms automatically reinstall their (native) signal
192 handlers after a signal delivery. This means that the handler works
193 only the first time the signal is sent. The solution to this problem
194 is to use C<POSIX> signal handlers if available, their behaviour
197 The following example implements a simple daemon, which restarts
198 itself every time the C<SIGHUP> signal is received. The actual code is
199 located in the subroutine C<code()>, which simply prints some debug
200 info to show that it works and should be replaced with the real code.
206 use File::Basename ();
207 use File::Spec::Functions;
211 # make the daemon cross-platform, so exec always calls the script
212 # itself with the right path, no matter how the script was invoked.
213 my $script = File::Basename::basename($0);
214 my $SELF = catfile $FindBin::Bin, $script;
216 # POSIX unmasks the sigprocmask properly
217 my $sigset = POSIX::SigSet->new();
218 my $action = POSIX::SigAction->new('sigHUP_handler',
221 POSIX::sigaction(&POSIX::SIGHUP, $action);
224 print "got SIGHUP\n";
225 exec($SELF, @ARGV) or die "Couldn't restart: $!\n";
232 print "ARGV: @ARGV\n";
244 A named pipe (often referred to as a FIFO) is an old Unix IPC
245 mechanism for processes communicating on the same machine. It works
246 just like a regular, connected anonymous pipes, except that the
247 processes rendezvous using a filename and don't have to be related.
249 To create a named pipe, use the C<POSIX::mkfifo()> function.
251 use POSIX qw(mkfifo);
252 mkfifo($path, 0700) or die "mkfifo $path failed: $!";
254 You can also use the Unix command mknod(1) or on some
255 systems, mkfifo(1). These may not be in your normal path.
257 # system return val is backwards, so && not ||
259 $ENV{PATH} .= ":/etc:/usr/etc";
260 if ( system('mknod', $path, 'p')
261 && system('mkfifo', $path) )
263 die "mk{nod,fifo} $path failed";
267 A fifo is convenient when you want to connect a process to an unrelated
268 one. When you open a fifo, the program will block until there's something
271 For example, let's say you'd like to have your F<.signature> file be a
272 named pipe that has a Perl program on the other end. Now every time any
273 program (like a mailer, news reader, finger program, etc.) tries to read
274 from that file, the reading program will block and your program will
275 supply the new signature. We'll use the pipe-checking file test B<-p>
276 to find out whether anyone (or anything) has accidentally removed our fifo.
279 $FIFO = '.signature';
285 POSIX::mkfifo($FIFO, 0700)
286 or die "can't mkfifo $FIFO: $!";
289 # next line blocks until there's a reader
290 open (FIFO, "> $FIFO") || die "can't write $FIFO: $!";
291 print FIFO "John Smith (smith\@host.org)\n", `fortune -s`;
293 sleep 2; # to avoid dup signals
296 =head2 Deferred Signals (Safe Signals)
298 In Perls before Perl 5.7.3 by installing Perl code to deal with
299 signals, you were exposing yourself to danger from two things. First,
300 few system library functions are re-entrant. If the signal interrupts
301 while Perl is executing one function (like malloc(3) or printf(3)),
302 and your signal handler then calls the same function again, you could
303 get unpredictable behavior--often, a core dump. Second, Perl isn't
304 itself re-entrant at the lowest levels. If the signal interrupts Perl
305 while Perl is changing its own internal data structures, similarly
306 unpredictable behaviour may result.
308 There were two things you could do, knowing this: be paranoid or be
309 pragmatic. The paranoid approach was to do as little as possible in your
310 signal handler. Set an existing integer variable that already has a
311 value, and return. This doesn't help you if you're in a slow system call,
312 which will just restart. That means you have to C<die> to longjmp(3) out
313 of the handler. Even this is a little cavalier for the true paranoiac,
314 who avoids C<die> in a handler because the system I<is> out to get you.
315 The pragmatic approach was to say "I know the risks, but prefer the
316 convenience", and to do anything you wanted in your signal handler,
317 and be prepared to clean up core dumps now and again.
319 In Perl 5.7.3 and later to avoid these problems signals are
320 "deferred"-- that is when the signal is delivered to the process by
321 the system (to the C code that implements Perl) a flag is set, and the
322 handler returns immediately. Then at strategic "safe" points in the
323 Perl interpreter (e.g. when it is about to execute a new opcode) the
324 flags are checked and the Perl level handler from %SIG is
325 executed. The "deferred" scheme allows much more flexibility in the
326 coding of signal handler as we know Perl interpreter is in a safe
327 state, and that we are not in a system library function when the
328 handler is called. However the implementation does differ from
329 previous Perls in the following ways:
333 =item Long-running opcodes
335 As the Perl interpreter only looks at the signal flags when it is about
336 to execute a new opcode, a signal that arrives during a long-running
337 opcode (e.g. a regular expression operation on a very large string) will
338 not be seen until the current opcode completes.
340 N.B. If a signal of any given type fires multiple times during an opcode
341 (such as from a fine-grained timer), the handler for that signal will
342 only be called once after the opcode completes, and all the other
343 instances will be discarded. Furthermore, if your system's signal queue
344 gets flooded to the point that there are signals that have been raised
345 but not yet caught (and thus not deferred) at the time an opcode
346 completes, those signals may well be caught and deferred during
347 subsequent opcodes, with sometimes surprising results. For example, you
348 may see alarms delivered even after calling C<alarm(0)> as the latter
349 stops the raising of alarms but does not cancel the delivery of alarms
350 raised but not yet caught. Do not depend on the behaviors described in
351 this paragraph as they are side effects of the current implementation and
352 may change in future versions of Perl.
355 =item Interrupting IO
357 When a signal is delivered (e.g. INT control-C) the operating system
358 breaks into IO operations like C<read> (used to implement Perls
359 E<lt>E<gt> operator). On older Perls the handler was called
360 immediately (and as C<read> is not "unsafe" this worked well). With
361 the "deferred" scheme the handler is not called immediately, and if
362 Perl is using system's C<stdio> library that library may re-start the
363 C<read> without returning to Perl and giving it a chance to call the
364 %SIG handler. If this happens on your system the solution is to use
365 C<:perlio> layer to do IO - at least on those handles which you want
366 to be able to break into with signals. (The C<:perlio> layer checks
367 the signal flags and calls %SIG handlers before resuming IO operation.)
369 Note that the default in Perl 5.7.3 and later is to automatically use
370 the C<:perlio> layer.
372 Note that some networking library functions like gethostbyname() are
373 known to have their own implementations of timeouts which may conflict
374 with your timeouts. If you are having problems with such functions,
375 you can try using the POSIX sigaction() function, which bypasses the
376 Perl safe signals (note that this means subjecting yourself to
377 possible memory corruption, as described above). Instead of setting
380 local $SIG{ALRM} = sub { die "alarm" };
382 try something like the following:
384 use POSIX qw(SIGALRM);
385 POSIX::sigaction(SIGALRM,
386 POSIX::SigAction->new(sub { die "alarm" }))
387 or die "Error setting SIGALRM handler: $!\n";
389 Another way to disable the safe signal behavior locally is to use
390 the C<Perl::Unsafe::Signals> module from CPAN (which will affect
393 =item Restartable system calls
395 On systems that supported it, older versions of Perl used the
396 SA_RESTART flag when installing %SIG handlers. This meant that
397 restartable system calls would continue rather than returning when
398 a signal arrived. In order to deliver deferred signals promptly,
399 Perl 5.7.3 and later do I<not> use SA_RESTART. Consequently,
400 restartable system calls can fail (with $! set to C<EINTR>) in places
401 where they previously would have succeeded.
403 Note that the default C<:perlio> layer will retry C<read>, C<write>
404 and C<close> as described above and that interrupted C<wait> and
405 C<waitpid> calls will always be retried.
407 =item Signals as "faults"
409 Certain signals, e.g. SEGV, ILL, and BUS, are generated as a result of
410 virtual memory or other "faults". These are normally fatal and there is
411 little a Perl-level handler can do with them, so Perl now delivers them
412 immediately rather than attempting to defer them.
414 =item Signals triggered by operating system state
416 On some operating systems certain signal handlers are supposed to "do
417 something" before returning. One example can be CHLD or CLD which
418 indicates a child process has completed. On some operating systems the
419 signal handler is expected to C<wait> for the completed child
420 process. On such systems the deferred signal scheme will not work for
421 those signals (it does not do the C<wait>). Again the failure will
422 look like a loop as the operating system will re-issue the signal as
423 there are un-waited-for completed child processes.
427 If you want the old signal behaviour back regardless of possible
428 memory corruption, set the environment variable C<PERL_SIGNALS> to
429 C<"unsafe"> (a new feature since Perl 5.8.1).
431 =head1 Using open() for IPC
433 Perl's basic open() statement can also be used for unidirectional
434 interprocess communication by either appending or prepending a pipe
435 symbol to the second argument to open(). Here's how to start
436 something up in a child process you intend to write to:
438 open(SPOOLER, "| cat -v | lpr -h 2>/dev/null")
439 || die "can't fork: $!";
440 local $SIG{PIPE} = sub { die "spooler pipe broke" };
441 print SPOOLER "stuff\n";
442 close SPOOLER || die "bad spool: $! $?";
444 And here's how to start up a child process you intend to read from:
446 open(STATUS, "netstat -an 2>&1 |")
447 || die "can't fork: $!";
449 next if /^(tcp|udp)/;
452 close STATUS || die "bad netstat: $! $?";
454 If one can be sure that a particular program is a Perl script that is
455 expecting filenames in @ARGV, the clever programmer can write something
458 % program f1 "cmd1|" - f2 "cmd2|" f3 < tmpfile
460 and irrespective of which shell it's called from, the Perl program will
461 read from the file F<f1>, the process F<cmd1>, standard input (F<tmpfile>
462 in this case), the F<f2> file, the F<cmd2> command, and finally the F<f3>
463 file. Pretty nifty, eh?
465 You might notice that you could use backticks for much the
466 same effect as opening a pipe for reading:
468 print grep { !/^(tcp|udp)/ } `netstat -an 2>&1`;
469 die "bad netstat" if $?;
471 While this is true on the surface, it's much more efficient to process the
472 file one line or record at a time because then you don't have to read the
473 whole thing into memory at once. It also gives you finer control of the
474 whole process, letting you to kill off the child process early if you'd
477 Be careful to check both the open() and the close() return values. If
478 you're I<writing> to a pipe, you should also trap SIGPIPE. Otherwise,
479 think of what happens when you start up a pipe to a command that doesn't
480 exist: the open() will in all likelihood succeed (it only reflects the
481 fork()'s success), but then your output will fail--spectacularly. Perl
482 can't know whether the command worked because your command is actually
483 running in a separate process whose exec() might have failed. Therefore,
484 while readers of bogus commands return just a quick end of file, writers
485 to bogus command will trigger a signal they'd better be prepared to
488 open(FH, "|bogus") or die "can't fork: $!";
489 print FH "bang\n" or die "can't write: $!";
490 close FH or die "can't close: $!";
492 That won't blow up until the close, and it will blow up with a SIGPIPE.
493 To catch it, you could use this:
495 $SIG{PIPE} = 'IGNORE';
496 open(FH, "|bogus") or die "can't fork: $!";
497 print FH "bang\n" or die "can't write: $!";
498 close FH or die "can't close: status=$?";
502 Both the main process and any child processes it forks share the same
503 STDIN, STDOUT, and STDERR filehandles. If both processes try to access
504 them at once, strange things can happen. You may also want to close
505 or reopen the filehandles for the child. You can get around this by
506 opening your pipe with open(), but on some systems this means that the
507 child process cannot outlive the parent.
509 =head2 Background Processes
511 You can run a command in the background with:
515 The command's STDOUT and STDERR (and possibly STDIN, depending on your
516 shell) will be the same as the parent's. You won't need to catch
517 SIGCHLD because of the double-fork taking place (see below for more
520 =head2 Complete Dissociation of Child from Parent
522 In some cases (starting server processes, for instance) you'll want to
523 completely dissociate the child process from the parent. This is
524 often called daemonization. A well behaved daemon will also chdir()
525 to the root directory (so it doesn't prevent unmounting the filesystem
526 containing the directory from which it was launched) and redirect its
527 standard file descriptors from and to F</dev/null> (so that random
528 output doesn't wind up on the user's terminal).
533 chdir '/' or die "Can't chdir to /: $!";
534 open STDIN, '/dev/null' or die "Can't read /dev/null: $!";
535 open STDOUT, '>/dev/null'
536 or die "Can't write to /dev/null: $!";
537 defined(my $pid = fork) or die "Can't fork: $!";
539 die "Can't start a new session: $!" if setsid == -1;
540 open STDERR, '>&STDOUT' or die "Can't dup stdout: $!";
543 The fork() has to come before the setsid() to ensure that you aren't a
544 process group leader (the setsid() will fail if you are). If your
545 system doesn't have the setsid() function, open F</dev/tty> and use the
546 C<TIOCNOTTY> ioctl() on it instead. See tty(4) for details.
548 Non-Unix users should check their Your_OS::Process module for other
551 =head2 Safe Pipe Opens
553 Another interesting approach to IPC is making your single program go
554 multiprocess and communicate between (or even amongst) yourselves. The
555 open() function will accept a file argument of either C<"-|"> or C<"|-">
556 to do a very interesting thing: it forks a child connected to the
557 filehandle you've opened. The child is running the same program as the
558 parent. This is useful for safely opening a file when running under an
559 assumed UID or GID, for example. If you open a pipe I<to> minus, you can
560 write to the filehandle you opened and your kid will find it in his
561 STDIN. If you open a pipe I<from> minus, you can read from the filehandle
562 you opened whatever your kid writes to his STDOUT.
564 use English '-no_match_vars';
568 $pid = open(KID_TO_WRITE, "|-");
569 unless (defined $pid) {
570 warn "cannot fork: $!";
571 die "bailing out" if $sleep_count++ > 6;
574 } until defined $pid;
577 print KID_TO_WRITE @some_data;
578 close(KID_TO_WRITE) || warn "kid exited $?";
580 ($EUID, $EGID) = ($UID, $GID); # suid progs only
581 open (FILE, "> /safe/file")
582 || die "can't open /safe/file: $!";
584 print FILE; # child's STDIN is parent's KID_TO_WRITE
586 exit; # don't forget this
589 Another common use for this construct is when you need to execute
590 something without the shell's interference. With system(), it's
591 straightforward, but you can't use a pipe open or backticks safely.
592 That's because there's no way to stop the shell from getting its hands on
593 your arguments. Instead, use lower-level control to call exec() directly.
595 Here's a safe backtick or pipe open for read:
597 # add error processing as above
598 $pid = open(KID_TO_READ, "-|");
601 while (<KID_TO_READ>) {
602 # do something interesting
604 close(KID_TO_READ) || warn "kid exited $?";
607 ($EUID, $EGID) = ($UID, $GID); # suid only
608 exec($program, @options, @args)
609 || die "can't exec program: $!";
614 And here's a safe pipe open for writing:
616 # add error processing as above
617 $pid = open(KID_TO_WRITE, "|-");
618 $SIG{PIPE} = sub { die "whoops, $program pipe broke" };
624 close(KID_TO_WRITE) || warn "kid exited $?";
627 ($EUID, $EGID) = ($UID, $GID);
628 exec($program, @options, @args)
629 || die "can't exec program: $!";
633 It is very easy to dead-lock a process using this form of open(), or
634 indeed any use of pipe() and multiple sub-processes. The above
635 example is 'safe' because it is simple and calls exec(). See
636 L</"Avoiding Pipe Deadlocks"> for general safety principles, but there
637 are extra gotchas with Safe Pipe Opens.
639 In particular, if you opened the pipe using C<open FH, "|-">, then you
640 cannot simply use close() in the parent process to close an unwanted
641 writer. Consider this code:
643 $pid = open WRITER, "|-";
644 defined $pid or die "fork failed; $!";
646 if (my $sub_pid = fork()) {
648 # do something else...
656 # do something with STDIN...
660 In the above, the true parent does not want to write to the WRITER
661 filehandle, so it closes it. However, because WRITER was opened using
662 C<open FH, "|-">, it has a special behaviour: closing it will call
663 waitpid() (see L<perlfunc/waitpid>), which waits for the sub-process
664 to exit. If the child process ends up waiting for something happening
665 in the section marked "do something else", then you have a deadlock.
667 This can also be a problem with intermediate sub-processes in more
668 complicated code, which will call waitpid() on all open filehandles
669 during global destruction; in no predictable order.
671 To solve this, you must manually use pipe(), fork(), and the form of
672 open() which sets one file descriptor to another, as below:
674 pipe(READER, WRITER);
676 defined $pid or die "fork failed; $!";
679 if (my $sub_pid = fork()) {
689 open STDIN, "<&READER";
695 Since Perl 5.8.0, you can also use the list form of C<open> for pipes :
698 open KID_PS, "-|", "ps", "aux" or die $!;
700 forks the ps(1) command (without spawning a shell, as there are more than
701 three arguments to open()), and reads its standard output via the
702 C<KID_PS> filehandle. The corresponding syntax to write to command
703 pipes (with C<"|-"> in place of C<"-|">) is also implemented.
705 Note that these operations are full Unix forks, which means they may not be
706 correctly implemented on alien systems. Additionally, these are not true
707 multithreading. If you'd like to learn more about threading, see the
708 F<modules> file mentioned below in the SEE ALSO section.
710 =head2 Avoiding Pipe Deadlocks
712 In general, if you have more than one sub-process, you need to be very
713 careful that any process which does not need the writer half of any
714 pipe you create for inter-process communication does not have it open.
716 The reason for this is that any child process which is reading from
717 the pipe and expecting an EOF will never receive it, and therefore
718 never exit. A single process closing a pipe is not enough to close it;
719 the last process with the pipe open must close it for it to read EOF.
721 There are some features built-in to unix to help prevent this most of
722 the time. For instance, filehandles have a 'close on exec' flag (set
723 I<en masse> with Perl using the C<$^F> L<perlvar>), so that any
724 filehandles which you didn't explicitly route to the STDIN, STDOUT or
725 STDERR of a child I<program> will automatically be closed for you.
727 So, always explicitly and immediately call close() on the writable end
728 of any pipe, unless that process is actually writing to it. If you
729 don't explicitly call close() then be warned Perl will still close()
730 all the filehandles during global destruction. As warned above, if
731 those filehandles were opened with Safe Pipe Open, they will also call
732 waitpid() and you might again deadlock.
734 =head2 Bidirectional Communication with Another Process
736 While this works reasonably well for unidirectional communication, what
737 about bidirectional communication? The obvious thing you'd like to do
738 doesn't actually work:
740 open(PROG_FOR_READING_AND_WRITING, "| some program |")
742 and if you forget to use the C<use warnings> pragma or the B<-w> flag,
743 then you'll miss out entirely on the diagnostic message:
745 Can't do bidirectional pipe at -e line 1.
747 If you really want to, you can use the standard open2() library function
748 to catch both ends. There's also an open3() for tridirectional I/O so you
749 can also catch your child's STDERR, but doing so would then require an
750 awkward select() loop and wouldn't allow you to use normal Perl input
753 If you look at its source, you'll see that open2() uses low-level
754 primitives like Unix pipe() and exec() calls to create all the connections.
755 While it might have been slightly more efficient by using socketpair(), it
756 would have then been even less portable than it already is. The open2()
757 and open3() functions are unlikely to work anywhere except on a Unix
758 system or some other one purporting to be POSIX compliant.
760 Here's an example of using open2():
764 $pid = open2(*Reader, *Writer, "cat -u -n" );
765 print Writer "stuff\n";
768 The problem with this is that Unix buffering is really going to
769 ruin your day. Even though your C<Writer> filehandle is auto-flushed,
770 and the process on the other end will get your data in a timely manner,
771 you can't usually do anything to force it to give it back to you
772 in a similarly quick fashion. In this case, we could, because we
773 gave I<cat> a B<-u> flag to make it unbuffered. But very few Unix
774 commands are designed to operate over pipes, so this seldom works
775 unless you yourself wrote the program on the other end of the
778 A solution to this is the nonstandard F<Comm.pl> library. It uses
779 pseudo-ttys to make your program behave more reasonably:
782 $ph = open_proc('cat -n');
784 print $ph "a line\n";
785 print "got back ", scalar <$ph>;
788 This way you don't have to have control over the source code of the
789 program you're using. The F<Comm> library also has expect()
790 and interact() functions. Find the library (and we hope its
791 successor F<IPC::Chat>) at your nearest CPAN archive as detailed
792 in the SEE ALSO section below.
794 The newer Expect.pm module from CPAN also addresses this kind of thing.
795 This module requires two other modules from CPAN: IO::Pty and IO::Stty.
796 It sets up a pseudo-terminal to interact with programs that insist on
797 using talking to the terminal device driver. If your system is
798 amongst those supported, this may be your best bet.
800 =head2 Bidirectional Communication with Yourself
802 If you want, you may make low-level pipe() and fork()
803 to stitch this together by hand. This example only
804 talks to itself, but you could reopen the appropriate
805 handles to STDIN and STDOUT and call other processes.
808 # pipe1 - bidirectional communication using two pipe pairs
809 # designed for the socketpair-challenged
810 use IO::Handle; # thousands of lines just for autoflush :-(
811 pipe(PARENT_RDR, CHILD_WTR); # XXX: failure?
812 pipe(CHILD_RDR, PARENT_WTR); # XXX: failure?
813 CHILD_WTR->autoflush(1);
814 PARENT_WTR->autoflush(1);
817 close PARENT_RDR; close PARENT_WTR;
818 print CHILD_WTR "Parent Pid $$ is sending this\n";
819 chomp($line = <CHILD_RDR>);
820 print "Parent Pid $$ just read this: `$line'\n";
821 close CHILD_RDR; close CHILD_WTR;
824 die "cannot fork: $!" unless defined $pid;
825 close CHILD_RDR; close CHILD_WTR;
826 chomp($line = <PARENT_RDR>);
827 print "Child Pid $$ just read this: `$line'\n";
828 print PARENT_WTR "Child Pid $$ is sending this\n";
829 close PARENT_RDR; close PARENT_WTR;
833 But you don't actually have to make two pipe calls. If you
834 have the socketpair() system call, it will do this all for you.
837 # pipe2 - bidirectional communication using socketpair
838 # "the best ones always go both ways"
841 use IO::Handle; # thousands of lines just for autoflush :-(
842 # We say AF_UNIX because although *_LOCAL is the
843 # POSIX 1003.1g form of the constant, many machines
844 # still don't have it.
845 socketpair(CHILD, PARENT, AF_UNIX, SOCK_STREAM, PF_UNSPEC)
846 or die "socketpair: $!";
849 PARENT->autoflush(1);
853 print CHILD "Parent Pid $$ is sending this\n";
854 chomp($line = <CHILD>);
855 print "Parent Pid $$ just read this: `$line'\n";
859 die "cannot fork: $!" unless defined $pid;
861 chomp($line = <PARENT>);
862 print "Child Pid $$ just read this: `$line'\n";
863 print PARENT "Child Pid $$ is sending this\n";
868 =head1 Sockets: Client/Server Communication
870 While not limited to Unix-derived operating systems (e.g., WinSock on PCs
871 provides socket support, as do some VMS libraries), you may not have
872 sockets on your system, in which case this section probably isn't going to do
873 you much good. With sockets, you can do both virtual circuits (i.e., TCP
874 streams) and datagrams (i.e., UDP packets). You may be able to do even more
875 depending on your system.
877 The Perl function calls for dealing with sockets have the same names as
878 the corresponding system calls in C, but their arguments tend to differ
879 for two reasons: first, Perl filehandles work differently than C file
880 descriptors. Second, Perl already knows the length of its strings, so you
881 don't need to pass that information.
883 One of the major problems with old socket code in Perl was that it used
884 hard-coded values for some of the constants, which severely hurt
885 portability. If you ever see code that does anything like explicitly
886 setting C<$AF_INET = 2>, you know you're in for big trouble: An
887 immeasurably superior approach is to use the C<Socket> module, which more
888 reliably grants access to various constants and functions you'll need.
890 If you're not writing a server/client for an existing protocol like
891 NNTP or SMTP, you should give some thought to how your server will
892 know when the client has finished talking, and vice-versa. Most
893 protocols are based on one-line messages and responses (so one party
894 knows the other has finished when a "\n" is received) or multi-line
895 messages and responses that end with a period on an empty line
896 ("\n.\n" terminates a message/response).
898 =head2 Internet Line Terminators
900 The Internet line terminator is "\015\012". Under ASCII variants of
901 Unix, that could usually be written as "\r\n", but under other systems,
902 "\r\n" might at times be "\015\015\012", "\012\012\015", or something
903 completely different. The standards specify writing "\015\012" to be
904 conformant (be strict in what you provide), but they also recommend
905 accepting a lone "\012" on input (but be lenient in what you require).
906 We haven't always been very good about that in the code in this manpage,
907 but unless you're on a Mac, you'll probably be ok.
909 =head2 Internet TCP Clients and Servers
911 Use Internet-domain sockets when you want to do client-server
912 communication that might extend to machines outside of your own system.
914 Here's a sample TCP client using Internet-domain sockets:
919 my ($remote,$port, $iaddr, $paddr, $proto, $line);
921 $remote = shift || 'localhost';
922 $port = shift || 2345; # random port
923 if ($port =~ /\D/) { $port = getservbyname($port, 'tcp') }
924 die "No port" unless $port;
925 $iaddr = inet_aton($remote) || die "no host: $remote";
926 $paddr = sockaddr_in($port, $iaddr);
928 $proto = getprotobyname('tcp');
929 socket(SOCK, PF_INET, SOCK_STREAM, $proto) || die "socket: $!";
930 connect(SOCK, $paddr) || die "connect: $!";
931 while (defined($line = <SOCK>)) {
935 close (SOCK) || die "close: $!";
938 And here's a corresponding server to go along with it. We'll
939 leave the address as INADDR_ANY so that the kernel can choose
940 the appropriate interface on multihomed hosts. If you want sit
941 on a particular interface (like the external side of a gateway
942 or firewall machine), you should fill this in with your real address
947 BEGIN { $ENV{PATH} = '/usr/ucb:/bin' }
950 my $EOL = "\015\012";
952 sub logmsg { print "$0 $$: @_ at ", scalar localtime, "\n" }
954 my $port = shift || 2345;
955 my $proto = getprotobyname('tcp');
957 ($port) = $port =~ /^(\d+)$/ or die "invalid port";
959 socket(Server, PF_INET, SOCK_STREAM, $proto) || die "socket: $!";
960 setsockopt(Server, SOL_SOCKET, SO_REUSEADDR,
961 pack("l", 1)) || die "setsockopt: $!";
962 bind(Server, sockaddr_in($port, INADDR_ANY)) || die "bind: $!";
963 listen(Server,SOMAXCONN) || die "listen: $!";
965 logmsg "server started on port $port";
969 $SIG{CHLD} = \&REAPER;
971 for ( ; $paddr = accept(Client,Server); close Client) {
972 my($port,$iaddr) = sockaddr_in($paddr);
973 my $name = gethostbyaddr($iaddr,AF_INET);
975 logmsg "connection from $name [",
976 inet_ntoa($iaddr), "]
979 print Client "Hello there, $name, it's now ",
980 scalar localtime, $EOL;
983 And here's a multithreaded version. It's multithreaded in that
984 like most typical servers, it spawns (forks) a slave server to
985 handle the client request so that the master server can quickly
986 go back to service a new client.
990 BEGIN { $ENV{PATH} = '/usr/ucb:/bin' }
993 my $EOL = "\015\012";
995 sub spawn; # forward declaration
996 sub logmsg { print "$0 $$: @_ at ", scalar localtime, "\n" }
998 my $port = shift || 2345;
999 my $proto = getprotobyname('tcp');
1001 ($port) = $port =~ /^(\d+)$/ or die "invalid port";
1003 socket(Server, PF_INET, SOCK_STREAM, $proto) || die "socket: $!";
1004 setsockopt(Server, SOL_SOCKET, SO_REUSEADDR,
1005 pack("l", 1)) || die "setsockopt: $!";
1006 bind(Server, sockaddr_in($port, INADDR_ANY)) || die "bind: $!";
1007 listen(Server,SOMAXCONN) || die "listen: $!";
1009 logmsg "server started on port $port";
1014 use POSIX ":sys_wait_h";
1018 local $!; # don't let waitpid() overwrite current error
1019 while ((my $pid = waitpid(-1,WNOHANG)) > 0 && WIFEXITED($?)) {
1020 logmsg "reaped $waitedpid" . ($? ? " with exit $?" : '');
1022 $SIG{CHLD} = \&REAPER; # loathe SysV
1025 $SIG{CHLD} = \&REAPER;
1028 $paddr = accept(Client, Server) || do {
1029 # try again if accept() returned because a signal was received
1033 my ($port, $iaddr) = sockaddr_in($paddr);
1034 my $name = gethostbyaddr($iaddr, AF_INET);
1036 logmsg "connection from $name [",
1042 print "Hello there, $name, it's now ", scalar localtime, $EOL;
1043 exec '/usr/games/fortune' # XXX: `wrong' line terminators
1044 or confess "can't exec fortune: $!";
1050 my $coderef = shift;
1052 unless (@_ == 0 && $coderef && ref($coderef) eq 'CODE') {
1053 confess "usage: spawn CODEREF";
1057 if (! defined($pid = fork)) {
1058 logmsg "cannot fork: $!";
1062 logmsg "begat $pid";
1063 return; # I'm the parent
1065 # else I'm the child -- go spawn
1067 open(STDIN, "<&Client") || die "can't dup client to stdin";
1068 open(STDOUT, ">&Client") || die "can't dup client to stdout";
1069 ## open(STDERR, ">&STDOUT") || die "can't dup stdout to stderr";
1073 This server takes the trouble to clone off a child version via fork()
1074 for each incoming request. That way it can handle many requests at
1075 once, which you might not always want. Even if you don't fork(), the
1076 listen() will allow that many pending connections. Forking servers
1077 have to be particularly careful about cleaning up their dead children
1078 (called "zombies" in Unix parlance), because otherwise you'll quickly
1079 fill up your process table. The REAPER subroutine is used here to
1080 call waitpid() for any child processes that have finished, thereby
1081 ensuring that they terminate cleanly and don't join the ranks of the
1084 Within the while loop we call accept() and check to see if it returns
1085 a false value. This would normally indicate a system error that needs
1086 to be reported. However the introduction of safe signals (see
1087 L</Deferred Signals (Safe Signals)> above) in Perl 5.7.3 means that
1088 accept() may also be interrupted when the process receives a signal.
1089 This typically happens when one of the forked sub-processes exits and
1090 notifies the parent process with a CHLD signal.
1092 If accept() is interrupted by a signal then $! will be set to EINTR.
1093 If this happens then we can safely continue to the next iteration of
1094 the loop and another call to accept(). It is important that your
1095 signal handling code doesn't modify the value of $! or this test will
1096 most likely fail. In the REAPER subroutine we create a local version
1097 of $! before calling waitpid(). When waitpid() sets $! to ECHILD (as
1098 it inevitably does when it has no more children waiting), it will
1099 update the local copy leaving the original unchanged.
1101 We suggest that you use the B<-T> flag to use taint checking (see L<perlsec>)
1102 even if we aren't running setuid or setgid. This is always a good idea
1103 for servers and other programs run on behalf of someone else (like CGI
1104 scripts), because it lessens the chances that people from the outside will
1105 be able to compromise your system.
1107 Let's look at another TCP client. This one connects to the TCP "time"
1108 service on a number of different machines and shows how far their clocks
1109 differ from the system on which it's being run:
1115 my $SECS_of_70_YEARS = 2208988800;
1116 sub ctime { scalar localtime(shift) }
1118 my $iaddr = gethostbyname('localhost');
1119 my $proto = getprotobyname('tcp');
1120 my $port = getservbyname('time', 'tcp');
1121 my $paddr = sockaddr_in(0, $iaddr);
1125 printf "%-24s %8s %s\n", "localhost", 0, ctime(time());
1127 foreach $host (@ARGV) {
1128 printf "%-24s ", $host;
1129 my $hisiaddr = inet_aton($host) || die "unknown host";
1130 my $hispaddr = sockaddr_in($port, $hisiaddr);
1131 socket(SOCKET, PF_INET, SOCK_STREAM, $proto) || die "socket: $!";
1132 connect(SOCKET, $hispaddr) || die "connect: $!";
1134 read(SOCKET, $rtime, 4);
1136 my $histime = unpack("N", $rtime) - $SECS_of_70_YEARS;
1137 printf "%8d %s\n", $histime - time, ctime($histime);
1140 =head2 Unix-Domain TCP Clients and Servers
1142 That's fine for Internet-domain clients and servers, but what about local
1143 communications? While you can use the same setup, sometimes you don't
1144 want to. Unix-domain sockets are local to the current host, and are often
1145 used internally to implement pipes. Unlike Internet domain sockets, Unix
1146 domain sockets can show up in the file system with an ls(1) listing.
1149 srw-rw-rw- 1 root 0 Oct 31 07:23 /dev/log
1151 You can test for these with Perl's B<-S> file test:
1153 unless ( -S '/dev/log' ) {
1154 die "something's wicked with the log system";
1157 Here's a sample Unix-domain client:
1162 my ($rendezvous, $line);
1164 $rendezvous = shift || 'catsock';
1165 socket(SOCK, PF_UNIX, SOCK_STREAM, 0) || die "socket: $!";
1166 connect(SOCK, sockaddr_un($rendezvous)) || die "connect: $!";
1167 while (defined($line = <SOCK>)) {
1172 And here's a corresponding server. You don't have to worry about silly
1173 network terminators here because Unix domain sockets are guaranteed
1174 to be on the localhost, and thus everything works right.
1181 BEGIN { $ENV{PATH} = '/usr/ucb:/bin' }
1182 sub spawn; # forward declaration
1183 sub logmsg { print "$0 $$: @_ at ", scalar localtime, "\n" }
1185 my $NAME = 'catsock';
1186 my $uaddr = sockaddr_un($NAME);
1187 my $proto = getprotobyname('tcp');
1189 socket(Server,PF_UNIX,SOCK_STREAM,0) || die "socket: $!";
1191 bind (Server, $uaddr) || die "bind: $!";
1192 listen(Server,SOMAXCONN) || die "listen: $!";
1194 logmsg "server started on $NAME";
1198 use POSIX ":sys_wait_h";
1201 while (($waitedpid = waitpid(-1,WNOHANG)) > 0) {
1202 logmsg "reaped $waitedpid" . ($? ? " with exit $?" : '');
1204 $SIG{CHLD} = \&REAPER; # loathe SysV
1207 $SIG{CHLD} = \&REAPER;
1210 for ( $waitedpid = 0;
1211 accept(Client,Server) || $waitedpid;
1212 $waitedpid = 0, close Client)
1215 logmsg "connection on $NAME";
1217 print "Hello there, it's now ", scalar localtime, "\n";
1218 exec '/usr/games/fortune' or die "can't exec fortune: $!";
1223 my $coderef = shift;
1225 unless (@_ == 0 && $coderef && ref($coderef) eq 'CODE') {
1226 confess "usage: spawn CODEREF";
1230 if (!defined($pid = fork)) {
1231 logmsg "cannot fork: $!";
1234 logmsg "begat $pid";
1235 return; # I'm the parent
1237 # else I'm the child -- go spawn
1239 open(STDIN, "<&Client") || die "can't dup client to stdin";
1240 open(STDOUT, ">&Client") || die "can't dup client to stdout";
1241 ## open(STDERR, ">&STDOUT") || die "can't dup stdout to stderr";
1245 As you see, it's remarkably similar to the Internet domain TCP server, so
1246 much so, in fact, that we've omitted several duplicate functions--spawn(),
1247 logmsg(), ctime(), and REAPER()--which are exactly the same as in the
1250 So why would you ever want to use a Unix domain socket instead of a
1251 simpler named pipe? Because a named pipe doesn't give you sessions. You
1252 can't tell one process's data from another's. With socket programming,
1253 you get a separate session for each client: that's why accept() takes two
1256 For example, let's say that you have a long running database server daemon
1257 that you want folks from the World Wide Web to be able to access, but only
1258 if they go through a CGI interface. You'd have a small, simple CGI
1259 program that does whatever checks and logging you feel like, and then acts
1260 as a Unix-domain client and connects to your private server.
1262 =head1 TCP Clients with IO::Socket
1264 For those preferring a higher-level interface to socket programming, the
1265 IO::Socket module provides an object-oriented approach. IO::Socket is
1266 included as part of the standard Perl distribution as of the 5.004
1267 release. If you're running an earlier version of Perl, just fetch
1268 IO::Socket from CPAN, where you'll also find modules providing easy
1269 interfaces to the following systems: DNS, FTP, Ident (RFC 931), NIS and
1270 NISPlus, NNTP, Ping, POP3, SMTP, SNMP, SSLeay, Telnet, and Time--just
1273 =head2 A Simple Client
1275 Here's a client that creates a TCP connection to the "daytime"
1276 service at port 13 of the host name "localhost" and prints out everything
1277 that the server there cares to provide.
1281 $remote = IO::Socket::INET->new(
1283 PeerAddr => "localhost",
1284 PeerPort => "daytime(13)",
1286 or die "cannot connect to daytime port at localhost";
1287 while ( <$remote> ) { print }
1289 When you run this program, you should get something back that
1292 Wed May 14 08:40:46 MDT 1997
1294 Here are what those parameters to the C<new> constructor mean:
1300 This is which protocol to use. In this case, the socket handle returned
1301 will be connected to a TCP socket, because we want a stream-oriented
1302 connection, that is, one that acts pretty much like a plain old file.
1303 Not all sockets are this of this type. For example, the UDP protocol
1304 can be used to make a datagram socket, used for message-passing.
1308 This is the name or Internet address of the remote host the server is
1309 running on. We could have specified a longer name like C<"www.perl.com">,
1310 or an address like C<"204.148.40.9">. For demonstration purposes, we've
1311 used the special hostname C<"localhost">, which should always mean the
1312 current machine you're running on. The corresponding Internet address
1313 for localhost is C<"127.1">, if you'd rather use that.
1317 This is the service name or port number we'd like to connect to.
1318 We could have gotten away with using just C<"daytime"> on systems with a
1319 well-configured system services file,[FOOTNOTE: The system services file
1320 is in I</etc/services> under Unix] but just in case, we've specified the
1321 port number (13) in parentheses. Using just the number would also have
1322 worked, but constant numbers make careful programmers nervous.
1326 Notice how the return value from the C<new> constructor is used as
1327 a filehandle in the C<while> loop? That's what's called an indirect
1328 filehandle, a scalar variable containing a filehandle. You can use
1329 it the same way you would a normal filehandle. For example, you
1330 can read one line from it this way:
1334 all remaining lines from is this way:
1338 and send a line of data to it this way:
1340 print $handle "some data\n";
1342 =head2 A Webget Client
1344 Here's a simple client that takes a remote host to fetch a document
1345 from, and then a list of documents to get from that host. This is a
1346 more interesting client than the previous one because it first sends
1347 something to the server before fetching the server's response.
1351 unless (@ARGV > 1) { die "usage: $0 host document ..." }
1352 $host = shift(@ARGV);
1355 foreach $document ( @ARGV ) {
1356 $remote = IO::Socket::INET->new( Proto => "tcp",
1358 PeerPort => "http(80)",
1360 unless ($remote) { die "cannot connect to http daemon on $host" }
1361 $remote->autoflush(1);
1362 print $remote "GET $document HTTP/1.0" . $BLANK;
1363 while ( <$remote> ) { print }
1367 The web server handing the "http" service, which is assumed to be at
1368 its standard port, number 80. If the web server you're trying to
1369 connect to is at a different port (like 1080 or 8080), you should specify
1370 as the named-parameter pair, C<< PeerPort => 8080 >>. The C<autoflush>
1371 method is used on the socket because otherwise the system would buffer
1372 up the output we sent it. (If you're on a Mac, you'll also need to
1373 change every C<"\n"> in your code that sends data over the network to
1374 be a C<"\015\012"> instead.)
1376 Connecting to the server is only the first part of the process: once you
1377 have the connection, you have to use the server's language. Each server
1378 on the network has its own little command language that it expects as
1379 input. The string that we send to the server starting with "GET" is in
1380 HTTP syntax. In this case, we simply request each specified document.
1381 Yes, we really are making a new connection for each document, even though
1382 it's the same host. That's the way you always used to have to speak HTTP.
1383 Recent versions of web browsers may request that the remote server leave
1384 the connection open a little while, but the server doesn't have to honor
1387 Here's an example of running that program, which we'll call I<webget>:
1389 % webget www.perl.com /guanaco.html
1390 HTTP/1.1 404 File Not Found
1391 Date: Thu, 08 May 1997 18:02:32 GMT
1392 Server: Apache/1.2b6
1394 Content-type: text/html
1396 <HEAD><TITLE>404 File Not Found</TITLE></HEAD>
1397 <BODY><H1>File Not Found</H1>
1398 The requested URL /guanaco.html was not found on this server.<P>
1401 Ok, so that's not very interesting, because it didn't find that
1402 particular document. But a long response wouldn't have fit on this page.
1404 For a more fully-featured version of this program, you should look to
1405 the I<lwp-request> program included with the LWP modules from CPAN.
1407 =head2 Interactive Client with IO::Socket
1409 Well, that's all fine if you want to send one command and get one answer,
1410 but what about setting up something fully interactive, somewhat like
1411 the way I<telnet> works? That way you can type a line, get the answer,
1412 type a line, get the answer, etc.
1414 This client is more complicated than the two we've done so far, but if
1415 you're on a system that supports the powerful C<fork> call, the solution
1416 isn't that rough. Once you've made the connection to whatever service
1417 you'd like to chat with, call C<fork> to clone your process. Each of
1418 these two identical process has a very simple job to do: the parent
1419 copies everything from the socket to standard output, while the child
1420 simultaneously copies everything from standard input to the socket.
1421 To accomplish the same thing using just one process would be I<much>
1422 harder, because it's easier to code two processes to do one thing than it
1423 is to code one process to do two things. (This keep-it-simple principle
1424 a cornerstones of the Unix philosophy, and good software engineering as
1425 well, which is probably why it's spread to other systems.)
1432 my ($host, $port, $kidpid, $handle, $line);
1434 unless (@ARGV == 2) { die "usage: $0 host port" }
1435 ($host, $port) = @ARGV;
1437 # create a tcp connection to the specified host and port
1438 $handle = IO::Socket::INET->new(Proto => "tcp",
1441 or die "can't connect to port $port on $host: $!";
1443 $handle->autoflush(1); # so output gets there right away
1444 print STDERR "[Connected to $host:$port]\n";
1446 # split the program into two processes, identical twins
1447 die "can't fork: $!" unless defined($kidpid = fork());
1449 # the if{} block runs only in the parent process
1451 # copy the socket to standard output
1452 while (defined ($line = <$handle>)) {
1455 kill("TERM", $kidpid); # send SIGTERM to child
1457 # the else{} block runs only in the child process
1459 # copy standard input to the socket
1460 while (defined ($line = <STDIN>)) {
1461 print $handle $line;
1465 The C<kill> function in the parent's C<if> block is there to send a
1466 signal to our child process (current running in the C<else> block)
1467 as soon as the remote server has closed its end of the connection.
1469 If the remote server sends data a byte at time, and you need that
1470 data immediately without waiting for a newline (which might not happen),
1471 you may wish to replace the C<while> loop in the parent with the
1475 while (sysread($handle, $byte, 1) == 1) {
1479 Making a system call for each byte you want to read is not very efficient
1480 (to put it mildly) but is the simplest to explain and works reasonably
1483 =head1 TCP Servers with IO::Socket
1485 As always, setting up a server is little bit more involved than running a client.
1486 The model is that the server creates a special kind of socket that
1487 does nothing but listen on a particular port for incoming connections.
1488 It does this by calling the C<< IO::Socket::INET->new() >> method with
1489 slightly different arguments than the client did.
1495 This is which protocol to use. Like our clients, we'll
1496 still specify C<"tcp"> here.
1501 port in the C<LocalPort> argument, which we didn't do for the client.
1502 This is service name or port number for which you want to be the
1503 server. (Under Unix, ports under 1024 are restricted to the
1504 superuser.) In our sample, we'll use port 9000, but you can use
1505 any port that's not currently in use on your system. If you try
1506 to use one already in used, you'll get an "Address already in use"
1507 message. Under Unix, the C<netstat -a> command will show
1508 which services current have servers.
1512 The C<Listen> parameter is set to the maximum number of
1513 pending connections we can accept until we turn away incoming clients.
1514 Think of it as a call-waiting queue for your telephone.
1515 The low-level Socket module has a special symbol for the system maximum, which
1520 The C<Reuse> parameter is needed so that we restart our server
1521 manually without waiting a few minutes to allow system buffers to
1526 Once the generic server socket has been created using the parameters
1527 listed above, the server then waits for a new client to connect
1528 to it. The server blocks in the C<accept> method, which eventually accepts a
1529 bidirectional connection from the remote client. (Make sure to autoflush
1530 this handle to circumvent buffering.)
1532 To add to user-friendliness, our server prompts the user for commands.
1533 Most servers don't do this. Because of the prompt without a newline,
1534 you'll have to use the C<sysread> variant of the interactive client above.
1536 This server accepts one of five different commands, sending output
1537 back to the client. Note that unlike most network servers, this one
1538 only handles one incoming client at a time. Multithreaded servers are
1539 covered in Chapter 6 of the Camel.
1541 Here's the code. We'll
1545 use Net::hostent; # for OO version of gethostbyaddr
1547 $PORT = 9000; # pick something not in use
1549 $server = IO::Socket::INET->new( Proto => 'tcp',
1551 Listen => SOMAXCONN,
1554 die "can't setup server" unless $server;
1555 print "[Server $0 accepting clients]\n";
1557 while ($client = $server->accept()) {
1558 $client->autoflush(1);
1559 print $client "Welcome to $0; type help for command list.\n";
1560 $hostinfo = gethostbyaddr($client->peeraddr);
1561 printf "[Connect from %s]\n", $hostinfo ? $hostinfo->name : $client->peerhost;
1562 print $client "Command? ";
1563 while ( <$client>) {
1564 next unless /\S/; # blank line
1565 if (/quit|exit/i) { last; }
1566 elsif (/date|time/i) { printf $client "%s\n", scalar localtime; }
1567 elsif (/who/i ) { print $client `who 2>&1`; }
1568 elsif (/cookie/i ) { print $client `/usr/games/fortune 2>&1`; }
1569 elsif (/motd/i ) { print $client `cat /etc/motd 2>&1`; }
1571 print $client "Commands: quit date who cookie motd\n";
1574 print $client "Command? ";
1579 =head1 UDP: Message Passing
1581 Another kind of client-server setup is one that uses not connections, but
1582 messages. UDP communications involve much lower overhead but also provide
1583 less reliability, as there are no promises that messages will arrive at
1584 all, let alone in order and unmangled. Still, UDP offers some advantages
1585 over TCP, including being able to "broadcast" or "multicast" to a whole
1586 bunch of destination hosts at once (usually on your local subnet). If you
1587 find yourself overly concerned about reliability and start building checks
1588 into your message system, then you probably should use just TCP to start
1591 Note that UDP datagrams are I<not> a bytestream and should not be treated
1592 as such. This makes using I/O mechanisms with internal buffering
1593 like stdio (i.e. print() and friends) especially cumbersome. Use syswrite(),
1594 or better send(), like in the example below.
1596 Here's a UDP program similar to the sample Internet TCP client given
1597 earlier. However, instead of checking one host at a time, the UDP version
1598 will check many of them asynchronously by simulating a multicast and then
1599 using select() to do a timed-out wait for I/O. To do something similar
1600 with TCP, you'd have to use a different socket handle for each host.
1607 my ( $count, $hisiaddr, $hispaddr, $histime,
1608 $host, $iaddr, $paddr, $port, $proto,
1609 $rin, $rout, $rtime, $SECS_of_70_YEARS);
1611 $SECS_of_70_YEARS = 2208988800;
1613 $iaddr = gethostbyname(hostname());
1614 $proto = getprotobyname('udp');
1615 $port = getservbyname('time', 'udp');
1616 $paddr = sockaddr_in(0, $iaddr); # 0 means let kernel pick
1618 socket(SOCKET, PF_INET, SOCK_DGRAM, $proto) || die "socket: $!";
1619 bind(SOCKET, $paddr) || die "bind: $!";
1622 printf "%-12s %8s %s\n", "localhost", 0, scalar localtime time;
1626 $hisiaddr = inet_aton($host) || die "unknown host";
1627 $hispaddr = sockaddr_in($port, $hisiaddr);
1628 defined(send(SOCKET, 0, 0, $hispaddr)) || die "send $host: $!";
1632 vec($rin, fileno(SOCKET), 1) = 1;
1634 # timeout after 10.0 seconds
1635 while ($count && select($rout = $rin, undef, undef, 10.0)) {
1637 ($hispaddr = recv(SOCKET, $rtime, 4, 0)) || die "recv: $!";
1638 ($port, $hisiaddr) = sockaddr_in($hispaddr);
1639 $host = gethostbyaddr($hisiaddr, AF_INET);
1640 $histime = unpack("N", $rtime) - $SECS_of_70_YEARS;
1641 printf "%-12s ", $host;
1642 printf "%8d %s\n", $histime - time, scalar localtime($histime);
1646 Note that this example does not include any retries and may consequently
1647 fail to contact a reachable host. The most prominent reason for this
1648 is congestion of the queues on the sending host if the number of
1649 list of hosts to contact is sufficiently large.
1653 While System V IPC isn't so widely used as sockets, it still has some
1654 interesting uses. You can't, however, effectively use SysV IPC or
1655 Berkeley mmap() to have shared memory so as to share a variable amongst
1656 several processes. That's because Perl would reallocate your string when
1657 you weren't wanting it to.
1659 Here's a small example showing shared memory usage.
1661 use IPC::SysV qw(IPC_PRIVATE IPC_RMID S_IRUSR S_IWUSR);
1664 $id = shmget(IPC_PRIVATE, $size, S_IRUSR|S_IWUSR) || die "$!";
1665 print "shm key $id\n";
1667 $message = "Message #1";
1668 shmwrite($id, $message, 0, 60) || die "$!";
1669 print "wrote: '$message'\n";
1670 shmread($id, $buff, 0, 60) || die "$!";
1671 print "read : '$buff'\n";
1673 # the buffer of shmread is zero-character end-padded.
1674 substr($buff, index($buff, "\0")) = '';
1675 print "un" unless $buff eq $message;
1678 print "deleting shm $id\n";
1679 shmctl($id, IPC_RMID, 0) || die "$!";
1681 Here's an example of a semaphore:
1683 use IPC::SysV qw(IPC_CREAT);
1686 $id = semget($IPC_KEY, 10, 0666 | IPC_CREAT ) || die "$!";
1687 print "shm key $id\n";
1689 Put this code in a separate file to be run in more than one process.
1690 Call the file F<take>:
1692 # create a semaphore
1695 $id = semget($IPC_KEY, 0 , 0 );
1696 die if !defined($id);
1702 # wait for semaphore to be zero
1704 $opstring1 = pack("s!s!s!", $semnum, $semop, $semflag);
1706 # Increment the semaphore count
1708 $opstring2 = pack("s!s!s!", $semnum, $semop, $semflag);
1709 $opstring = $opstring1 . $opstring2;
1711 semop($id,$opstring) || die "$!";
1713 Put this code in a separate file to be run in more than one process.
1714 Call this file F<give>:
1716 # 'give' the semaphore
1717 # run this in the original process and you will see
1718 # that the second process continues
1721 $id = semget($IPC_KEY, 0, 0);
1722 die if !defined($id);
1727 # Decrement the semaphore count
1729 $opstring = pack("s!s!s!", $semnum, $semop, $semflag);
1731 semop($id,$opstring) || die "$!";
1733 The SysV IPC code above was written long ago, and it's definitely
1734 clunky looking. For a more modern look, see the IPC::SysV module
1735 which is included with Perl starting from Perl 5.005.
1737 A small example demonstrating SysV message queues:
1739 use IPC::SysV qw(IPC_PRIVATE IPC_RMID IPC_CREAT S_IRUSR S_IWUSR);
1741 my $id = msgget(IPC_PRIVATE, IPC_CREAT | S_IRUSR | S_IWUSR);
1743 my $sent = "message";
1744 my $type_sent = 1234;
1749 if (msgsnd($id, pack("l! a*", $type_sent, $sent), 0)) {
1750 if (msgrcv($id, $rcvd, 60, 0, 0)) {
1751 ($type_rcvd, $rcvd) = unpack("l! a*", $rcvd);
1752 if ($rcvd eq $sent) {
1758 die "# msgrcv failed\n";
1761 die "# msgsnd failed\n";
1763 msgctl($id, IPC_RMID, 0) || die "# msgctl failed: $!\n";
1765 die "# msgget failed\n";
1770 Most of these routines quietly but politely return C<undef> when they
1771 fail instead of causing your program to die right then and there due to
1772 an uncaught exception. (Actually, some of the new I<Socket> conversion
1773 functions croak() on bad arguments.) It is therefore essential to
1774 check return values from these functions. Always begin your socket
1775 programs this way for optimal success, and don't forget to add B<-T>
1776 taint checking flag to the #! line for servers:
1785 All these routines create system-specific portability problems. As noted
1786 elsewhere, Perl is at the mercy of your C libraries for much of its system
1787 behaviour. It's probably safest to assume broken SysV semantics for
1788 signals and to stick with simple TCP and UDP socket operations; e.g., don't
1789 try to pass open file descriptors over a local UDP datagram socket if you
1790 want your code to stand a chance of being portable.
1794 Tom Christiansen, with occasional vestiges of Larry Wall's original
1795 version and suggestions from the Perl Porters.
1799 There's a lot more to networking than this, but this should get you
1802 For intrepid programmers, the indispensable textbook is I<Unix
1803 Network Programming, 2nd Edition, Volume 1> by W. Richard Stevens
1804 (published by Prentice-Hall). Note that most books on networking
1805 address the subject from the perspective of a C programmer; translation
1806 to Perl is left as an exercise for the reader.
1808 The IO::Socket(3) manpage describes the object library, and the Socket(3)
1809 manpage describes the low-level interface to sockets. Besides the obvious
1810 functions in L<perlfunc>, you should also check out the F<modules> file
1811 at your nearest CPAN site. (See L<perlmodlib> or best yet, the F<Perl
1812 FAQ> for a description of what CPAN is and where to get it.)
1814 Section 5 of the F<modules> file is devoted to "Networking, Device Control
1815 (modems), and Interprocess Communication", and contains numerous unbundled
1816 modules numerous networking modules, Chat and Expect operations, CGI
1817 programming, DCE, FTP, IPC, NNTP, Proxy, Ptty, RPC, SNMP, SMTP, Telnet,
1818 Threads, and ToolTalk--just to name a few.