1PERLIPC(1)             Perl Programmers Reference Guide             PERLIPC(1)
2
3
4

NAME

6       perlipc - Perl interprocess communication (signals, fifos, pipes, safe
7       subprocesses, sockets, and semaphores)
8

DESCRIPTION

10       The basic IPC facilities of Perl are built out of the good old Unix
11       signals, named pipes, pipe opens, the Berkeley socket routines, and
12       SysV IPC calls.  Each is used in slightly different situations.
13

Signals

15       Perl uses a simple signal handling model: the %SIG hash contains names
16       or references of user-installed signal handlers.  These handlers will
17       be called with an argument which is the name of the signal that
18       triggered it.  A signal may be generated intentionally from a
19       particular keyboard sequence like control-C or control-Z, sent to you
20       from another process, or triggered automatically by the kernel when
21       special events transpire, like a child process exiting, your own
22       process running out of stack space, or hitting a process file-size
23       limit.
24
25       For example, to trap an interrupt signal, set up a handler like this:
26
27           our $shucks;
28
29           sub catch_zap {
30               my $signame = shift;
31               $shucks++;
32               die "Somebody sent me a SIG$signame";
33           }
34           $SIG{INT} = __PACKAGE__ . "::catch_zap";
35           $SIG{INT} = \&catch_zap;  # best strategy
36
37       Prior to Perl 5.8.0 it was necessary to do as little as you possibly
38       could in your handler; notice how all we do is set a global variable
39       and then raise an exception.  That's because on most systems, libraries
40       are not re-entrant; particularly, memory allocation and I/O routines
41       are not.  That meant that doing nearly anything in your handler could
42       in theory trigger a memory fault and subsequent core dump - see
43       "Deferred Signals (Safe Signals)" below.
44
45       The names of the signals are the ones listed out by "kill -l" on your
46       system, or you can retrieve them using the CPAN module IPC::Signal.
47
48       You may also choose to assign the strings "IGNORE" or "DEFAULT" as the
49       handler, in which case Perl will try to discard the signal or do the
50       default thing.
51
52       On most Unix platforms, the "CHLD" (sometimes also known as "CLD")
53       signal has special behavior with respect to a value of "IGNORE".
54       Setting $SIG{CHLD} to "IGNORE" on such a platform has the effect of not
55       creating zombie processes when the parent process fails to "wait()" on
56       its child processes (i.e., child processes are automatically reaped).
57       Calling "wait()" with $SIG{CHLD} set to "IGNORE" usually returns "-1"
58       on such platforms.
59
60       Some signals can be neither trapped nor ignored, such as the KILL and
61       STOP (but not the TSTP) signals. Note that ignoring signals makes them
62       disappear.  If you only want them blocked temporarily without them
63       getting lost you'll have to use the "POSIX" module's sigprocmask.
64
65       Sending a signal to a negative process ID means that you send the
66       signal to the entire Unix process group.  This code sends a hang-up
67       signal to all processes in the current process group, and also sets
68       $SIG{HUP} to "IGNORE" so it doesn't kill itself:
69
70           # block scope for local
71           {
72               local $SIG{HUP} = "IGNORE";
73               kill HUP => -getpgrp();
74               # snazzy writing of: kill("HUP", -getpgrp())
75           }
76
77       Another interesting signal to send is signal number zero.  This doesn't
78       actually affect a child process, but instead checks whether it's alive
79       or has changed its UIDs.
80
81           unless (kill 0 => $kid_pid) {
82               warn "something wicked happened to $kid_pid";
83           }
84
85       Signal number zero may fail because you lack permission to send the
86       signal when directed at a process whose real or saved UID is not
87       identical to the real or effective UID of the sending process, even
88       though the process is alive.  You may be able to determine the cause of
89       failure using $! or "%!".
90
91           unless (kill(0 => $pid) || $!{EPERM}) {
92               warn "$pid looks dead";
93           }
94
95       You might also want to employ anonymous functions for simple signal
96       handlers:
97
98           $SIG{INT} = sub { die "\nOutta here!\n" };
99
100       SIGCHLD handlers require some special care.  If a second child dies
101       while in the signal handler caused by the first death, we won't get
102       another signal. So must loop here else we will leave the unreaped child
103       as a zombie. And the next time two children die we get another zombie.
104       And so on.
105
106           use POSIX ":sys_wait_h";
107           $SIG{CHLD} = sub {
108               while ((my $child = waitpid(-1, WNOHANG)) > 0) {
109                   $Kid_Status{$child} = $?;
110               }
111           };
112           # do something that forks...
113
114       Be careful: qx(), system(), and some modules for calling external
115       commands do a fork(), then wait() for the result. Thus, your signal
116       handler will be called. Because wait() was already called by system()
117       or qx(), the wait() in the signal handler will see no more zombies and
118       will therefore block.
119
120       The best way to prevent this issue is to use waitpid(), as in the
121       following example:
122
123           use POSIX ":sys_wait_h"; # for nonblocking read
124
125           my %children;
126
127           $SIG{CHLD} = sub {
128               # don't change $! and $? outside handler
129               local ($!, $?);
130               while ( (my $pid = waitpid(-1, WNOHANG)) > 0 ) {
131                   delete $children{$pid};
132                   cleanup_child($pid, $?);
133               }
134           };
135
136           while (1) {
137               my $pid = fork();
138               die "cannot fork" unless defined $pid;
139               if ($pid == 0) {
140                   # ...
141                   exit 0;
142               } else {
143                   $children{$pid}=1;
144                   # ...
145                   system($command);
146                   # ...
147              }
148           }
149
150       Signal handling is also used for timeouts in Unix.  While safely
151       protected within an "eval{}" block, you set a signal handler to trap
152       alarm signals and then schedule to have one delivered to you in some
153       number of seconds.  Then try your blocking operation, clearing the
154       alarm when it's done but not before you've exited your "eval{}" block.
155       If it goes off, you'll use die() to jump out of the block.
156
157       Here's an example:
158
159           my $ALARM_EXCEPTION = "alarm clock restart";
160           eval {
161               local $SIG{ALRM} = sub { die $ALARM_EXCEPTION };
162               alarm 10;
163               flock($fh, 2)    # blocking write lock
164                               || die "cannot flock: $!";
165               alarm 0;
166           };
167           if ($@ && $@ !~ quotemeta($ALARM_EXCEPTION)) { die }
168
169       If the operation being timed out is system() or qx(), this technique is
170       liable to generate zombies.    If this matters to you, you'll need to
171       do your own fork() and exec(), and kill the errant child process.
172
173       For more complex signal handling, you might see the standard POSIX
174       module.  Lamentably, this is almost entirely undocumented, but the
175       ext/POSIX/t/sigaction.t file from the Perl source distribution has some
176       examples in it.
177
178   Handling the SIGHUP Signal in Daemons
179       A process that usually starts when the system boots and shuts down when
180       the system is shut down is called a daemon (Disk And Execution
181       MONitor). If a daemon process has a configuration file which is
182       modified after the process has been started, there should be a way to
183       tell that process to reread its configuration file without stopping the
184       process. Many daemons provide this mechanism using a "SIGHUP" signal
185       handler. When you want to tell the daemon to reread the file, simply
186       send it the "SIGHUP" signal.
187
188       The following example implements a simple daemon, which restarts itself
189       every time the "SIGHUP" signal is received. The actual code is located
190       in the subroutine "code()", which just prints some debugging info to
191       show that it works; it should be replaced with the real code.
192
193         #!/usr/bin/perl
194
195         use strict;
196         use warnings;
197
198         use POSIX ();
199         use FindBin ();
200         use File::Basename ();
201         use File::Spec::Functions qw(catfile);
202
203         $| = 1;
204
205         # make the daemon cross-platform, so exec always calls the script
206         # itself with the right path, no matter how the script was invoked.
207         my $script = File::Basename::basename($0);
208         my $SELF  = catfile($FindBin::Bin, $script);
209
210         # POSIX unmasks the sigprocmask properly
211         $SIG{HUP} = sub {
212             print "got SIGHUP\n";
213             exec($SELF, @ARGV)        || die "$0: couldn't restart: $!";
214         };
215
216         code();
217
218         sub code {
219             print "PID: $$\n";
220             print "ARGV: @ARGV\n";
221             my $count = 0;
222             while (1) {
223                 sleep 2;
224                 print ++$count, "\n";
225             }
226         }
227
228   Deferred Signals (Safe Signals)
229       Before Perl 5.8.0, installing Perl code to deal with signals exposed
230       you to danger from two things.  First, few system library functions are
231       re-entrant.  If the signal interrupts while Perl is executing one
232       function (like malloc(3) or printf(3)), and your signal handler then
233       calls the same function again, you could get unpredictable
234       behavior--often, a core dump.  Second, Perl isn't itself re-entrant at
235       the lowest levels.  If the signal interrupts Perl while Perl is
236       changing its own internal data structures, similarly unpredictable
237       behavior may result.
238
239       There were two things you could do, knowing this: be paranoid or be
240       pragmatic.  The paranoid approach was to do as little as possible in
241       your signal handler.  Set an existing integer variable that already has
242       a value, and return.  This doesn't help you if you're in a slow system
243       call, which will just restart.  That means you have to "die" to
244       longjmp(3) out of the handler.  Even this is a little cavalier for the
245       true paranoiac, who avoids "die" in a handler because the system is out
246       to get you.  The pragmatic approach was to say "I know the risks, but
247       prefer the convenience", and to do anything you wanted in your signal
248       handler, and be prepared to clean up core dumps now and again.
249
250       Perl 5.8.0 and later avoid these problems by "deferring" signals.  That
251       is, when the signal is delivered to the process by the system (to the C
252       code that implements Perl) a flag is set, and the handler returns
253       immediately.  Then at strategic "safe" points in the Perl interpreter
254       (e.g. when it is about to execute a new opcode) the flags are checked
255       and the Perl level handler from %SIG is executed. The "deferred" scheme
256       allows much more flexibility in the coding of signal handlers as we
257       know the Perl interpreter is in a safe state, and that we are not in a
258       system library function when the handler is called.  However the
259       implementation does differ from previous Perls in the following ways:
260
261       Long-running opcodes
262           As the Perl interpreter looks at signal flags only when it is about
263           to execute a new opcode, a signal that arrives during a long-
264           running opcode (e.g. a regular expression operation on a very large
265           string) will not be seen until the current opcode completes.
266
267           If a signal of any given type fires multiple times during an opcode
268           (such as from a fine-grained timer), the handler for that signal
269           will be called only once, after the opcode completes; all other
270           instances will be discarded.  Furthermore, if your system's signal
271           queue gets flooded to the point that there are signals that have
272           been raised but not yet caught (and thus not deferred) at the time
273           an opcode completes, those signals may well be caught and deferred
274           during subsequent opcodes, with sometimes surprising results.  For
275           example, you may see alarms delivered even after calling alarm(0)
276           as the latter stops the raising of alarms but does not cancel the
277           delivery of alarms raised but not yet caught.  Do not depend on the
278           behaviors described in this paragraph as they are side effects of
279           the current implementation and may change in future versions of
280           Perl.
281
282       Interrupting IO
283           When a signal is delivered (e.g., SIGINT from a control-C) the
284           operating system breaks into IO operations like read(2), which is
285           used to implement Perl's readline() function, the "<>" operator. On
286           older Perls the handler was called immediately (and as "read" is
287           not "unsafe", this worked well). With the "deferred" scheme the
288           handler is not called immediately, and if Perl is using the
289           system's "stdio" library that library may restart the "read"
290           without returning to Perl to give it a chance to call the %SIG
291           handler. If this happens on your system the solution is to use the
292           ":perlio" layer to do IO--at least on those handles that you want
293           to be able to break into with signals. (The ":perlio" layer checks
294           the signal flags and calls %SIG handlers before resuming IO
295           operation.)
296
297           The default in Perl 5.8.0 and later is to automatically use the
298           ":perlio" layer.
299
300           Note that it is not advisable to access a file handle within a
301           signal handler where that signal has interrupted an I/O operation
302           on that same handle. While perl will at least try hard not to
303           crash, there are no guarantees of data integrity; for example, some
304           data might get dropped or written twice.
305
306           Some networking library functions like gethostbyname() are known to
307           have their own implementations of timeouts which may conflict with
308           your timeouts.  If you have problems with such functions, try using
309           the POSIX sigaction() function, which bypasses Perl safe signals.
310           Be warned that this does subject you to possible memory corruption,
311           as described above.
312
313           Instead of setting $SIG{ALRM}:
314
315              local $SIG{ALRM} = sub { die "alarm" };
316
317           try something like the following:
318
319            use POSIX qw(SIGALRM);
320            POSIX::sigaction(SIGALRM,
321                             POSIX::SigAction->new(sub { die "alarm" }))
322                     || die "Error setting SIGALRM handler: $!\n";
323
324           Another way to disable the safe signal behavior locally is to use
325           the "Perl::Unsafe::Signals" module from CPAN, which affects all
326           signals.
327
328       Restartable system calls
329           On systems that supported it, older versions of Perl used the
330           SA_RESTART flag when installing %SIG handlers.  This meant that
331           restartable system calls would continue rather than returning when
332           a signal arrived.  In order to deliver deferred signals promptly,
333           Perl 5.8.0 and later do not use SA_RESTART.  Consequently,
334           restartable system calls can fail (with $! set to "EINTR") in
335           places where they previously would have succeeded.
336
337           The default ":perlio" layer retries "read", "write" and "close" as
338           described above; interrupted "wait" and "waitpid" calls will always
339           be retried.
340
341       Signals as "faults"
342           Certain signals like SEGV, ILL, and BUS are generated by virtual
343           memory addressing errors and similar "faults". These are normally
344           fatal: there is little a Perl-level handler can do with them.  So
345           Perl delivers them immediately rather than attempting to defer
346           them.
347
348       Signals triggered by operating system state
349           On some operating systems certain signal handlers are supposed to
350           "do something" before returning. One example can be CHLD or CLD,
351           which indicates a child process has completed. On some operating
352           systems the signal handler is expected to "wait" for the completed
353           child process. On such systems the deferred signal scheme will not
354           work for those signals: it does not do the "wait". Again the
355           failure will look like a loop as the operating system will reissue
356           the signal because there are completed child processes that have
357           not yet been "wait"ed for.
358
359       If you want the old signal behavior back despite possible memory
360       corruption, set the environment variable "PERL_SIGNALS" to "unsafe".
361       This feature first appeared in Perl 5.8.1.
362

Named Pipes

364       A named pipe (often referred to as a FIFO) is an old Unix IPC mechanism
365       for processes communicating on the same machine.  It works just like
366       regular anonymous pipes, except that the processes rendezvous using a
367       filename and need not be related.
368
369       To create a named pipe, use the "POSIX::mkfifo()" function.
370
371           use POSIX qw(mkfifo);
372           mkfifo($path, 0700)     ||  die "mkfifo $path failed: $!";
373
374       You can also use the Unix command mknod(1), or on some systems,
375       mkfifo(1).  These may not be in your normal path, though.
376
377           # system return val is backwards, so && not ||
378           #
379           $ENV{PATH} .= ":/etc:/usr/etc";
380           if  (      system("mknod",  $path, "p")
381                   && system("mkfifo", $path) )
382           {
383               die "mk{nod,fifo} $path failed";
384           }
385
386       A fifo is convenient when you want to connect a process to an unrelated
387       one.  When you open a fifo, the program will block until there's
388       something on the other end.
389
390       For example, let's say you'd like to have your .signature file be a
391       named pipe that has a Perl program on the other end.  Now every time
392       any program (like a mailer, news reader, finger program, etc.) tries to
393       read from that file, the reading program will read the new signature
394       from your program.  We'll use the pipe-checking file-test operator, -p,
395       to find out whether anyone (or anything) has accidentally removed our
396       fifo.
397
398           chdir();    # go home
399           my $FIFO = ".signature";
400
401           while (1) {
402               unless (-p $FIFO) {
403                   unlink $FIFO;   # discard any failure, will catch later
404                   require POSIX;  # delayed loading of heavy module
405                   POSIX::mkfifo($FIFO, 0700)
406                                         || die "can't mkfifo $FIFO: $!";
407               }
408
409               # next line blocks till there's a reader
410               open (my $fh, ">", $FIFO) || die "can't open $FIFO: $!";
411               print $fh "John Smith (smith\@host.org)\n", `fortune -s`;
412               close($fh)                || die "can't close $FIFO: $!";
413               sleep 2;                # to avoid dup signals
414           }
415

Using open() for IPC

417       Perl's basic open() statement can also be used for unidirectional
418       interprocess communication by specifying the open mode as "|-" or "-|".
419       Here's how to start something up in a child process you intend to write
420       to:
421
422           open(my $spooler, "|-", "cat -v | lpr -h 2>/dev/null")
423                               || die "can't fork: $!";
424           local $SIG{PIPE} = sub { die "spooler pipe broke" };
425           print $spooler "stuff\n";
426           close $spooler      || die "bad spool: $! $?";
427
428       And here's how to start up a child process you intend to read from:
429
430           open(my $status, "-|", "netstat -an 2>&1")
431                               || die "can't fork: $!";
432           while (<$status>) {
433               next if /^(tcp|udp)/;
434               print;
435           }
436           close $status       || die "bad netstat: $! $?";
437
438       Be aware that these operations are full Unix forks, which means they
439       may not be correctly implemented on all alien systems.  See "open" in
440       perlport for portability details.
441
442       In the two-argument form of open(), a pipe open can be achieved by
443       either appending or prepending a pipe symbol to the second argument:
444
445           open(my $spooler, "| cat -v | lpr -h 2>/dev/null")
446                               || die "can't fork: $!";
447           open(my $status, "netstat -an 2>&1 |")
448                               || die "can't fork: $!";
449
450       This can be used even on systems that do not support forking, but this
451       possibly allows code intended to read files to unexpectedly execute
452       programs.  If one can be sure that a particular program is a Perl
453       script expecting filenames in @ARGV using the two-argument form of
454       open() or the "<>" operator, the clever programmer can write something
455       like this:
456
457           % program f1 "cmd1|" - f2 "cmd2|" f3 < tmpfile
458
459       and no matter which sort of shell it's called from, the Perl program
460       will read from the file f1, the process cmd1, standard input (tmpfile
461       in this case), the f2 file, the cmd2 command, and finally the f3 file.
462       Pretty nifty, eh?
463
464       You might notice that you could use backticks for much the same effect
465       as opening a pipe for reading:
466
467           print grep { !/^(tcp|udp)/ } `netstat -an 2>&1`;
468           die "bad netstatus ($?)" if $?;
469
470       While this is true on the surface, it's much more efficient to process
471       the file one line or record at a time because then you don't have to
472       read the whole thing into memory at once.  It also gives you finer
473       control of the whole process, letting you kill off the child process
474       early if you'd like.
475
476       Be careful to check the return values from both open() and close().  If
477       you're writing to a pipe, you should also trap SIGPIPE.  Otherwise,
478       think of what happens when you start up a pipe to a command that
479       doesn't exist: the open() will in all likelihood succeed (it only
480       reflects the fork()'s success), but then your output will
481       fail--spectacularly.  Perl can't know whether the command worked,
482       because your command is actually running in a separate process whose
483       exec() might have failed.  Therefore, while readers of bogus commands
484       return just a quick EOF, writers to bogus commands will get hit with a
485       signal, which they'd best be prepared to handle.  Consider:
486
487           open(my $fh, "|-", "bogus") || die "can't fork: $!";
488           print $fh "bang\n";         #  neither necessary nor sufficient
489                                       #  to check print retval!
490           close($fh)                  || die "can't close: $!";
491
492       The reason for not checking the return value from print() is because of
493       pipe buffering; physical writes are delayed.  That won't blow up until
494       the close, and it will blow up with a SIGPIPE.  To catch it, you could
495       use this:
496
497           $SIG{PIPE} = "IGNORE";
498           open(my $fh, "|-", "bogus") || die "can't fork: $!";
499           print $fh "bang\n";
500           close($fh)                  || die "can't close: status=$?";
501
502   Filehandles
503       Both the main process and any child processes it forks share the same
504       STDIN, STDOUT, and STDERR filehandles.  If both processes try to access
505       them at once, strange things can happen.  You may also want to close or
506       reopen the filehandles for the child.  You can get around this by
507       opening your pipe with open(), but on some systems this means that the
508       child process cannot outlive the parent.
509
510   Background Processes
511       You can run a command in the background with:
512
513           system("cmd &");
514
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 details.
518
519   Complete Dissociation of Child from Parent
520       In some cases (starting server processes, for instance) you'll want to
521       completely dissociate the child process from the parent.  This is often
522       called daemonization.  A well-behaved daemon will also chdir() to the
523       root directory so it doesn't prevent unmounting the filesystem
524       containing the directory from which it was launched, and redirect its
525       standard file descriptors from and to /dev/null so that random output
526       doesn't wind up on the user's terminal.
527
528        use POSIX "setsid";
529
530        sub daemonize {
531            chdir("/")                     || die "can't chdir to /: $!";
532            open(STDIN,  "<", "/dev/null") || die "can't read /dev/null: $!";
533            open(STDOUT, ">", "/dev/null") || die "can't write /dev/null: $!";
534            defined(my $pid = fork())      || die "can't fork: $!";
535            exit if $pid;              # non-zero now means I am the parent
536            (setsid() != -1)           || die "Can't start a new session: $!";
537            open(STDERR, ">&", STDOUT) || die "can't dup stdout: $!";
538        }
539
540       The fork() has to come before the setsid() to ensure you aren't a
541       process group leader; the setsid() will fail if you are.  If your
542       system doesn't have the setsid() function, open /dev/tty and use the
543       "TIOCNOTTY" ioctl() on it instead.  See tty(4) for details.
544
545       Non-Unix users should check their "Your_OS::Process" module for other
546       possible solutions.
547
548   Safe Pipe Opens
549       Another interesting approach to IPC is making your single program go
550       multiprocess and communicate between--or even amongst--yourselves.  The
551       two-argument form of the open() function will accept a file argument of
552       either "-|" or "|-" to do a very interesting thing: it forks a child
553       connected to the filehandle you've opened.  The child is running the
554       same program as the parent.  This is useful for safely opening a file
555       when running under an assumed UID or GID, for example.  If you open a
556       pipe to minus, you can write to the filehandle you opened and your kid
557       will find it in his STDIN.  If you open a pipe from minus, you can read
558       from the filehandle you opened whatever your kid writes to his STDOUT.
559
560           my $PRECIOUS = "/path/to/some/safe/file";
561           my $sleep_count;
562           my $pid;
563           my $kid_to_write;
564
565           do {
566               $pid = open($kid_to_write, "|-");
567               unless (defined $pid) {
568                   warn "cannot fork: $!";
569                   die "bailing out" if $sleep_count++ > 6;
570                   sleep 10;
571               }
572           } until defined $pid;
573
574           if ($pid) {                 # I am the parent
575               print $kid_to_write @some_data;
576               close($kid_to_write)    || warn "kid exited $?";
577           } else {                    # I am the child
578               # drop permissions in setuid and/or setgid programs:
579               ($>, $)) = ($<, $();
580               open (my $outfile, ">", $PRECIOUS)
581                                       || die "can't open $PRECIOUS: $!";
582               while (<STDIN>) {
583                   print $outfile;     # child STDIN is parent $kid_to_write
584               }
585               close($outfile)         || die "can't close $PRECIOUS: $!";
586               exit(0);                # don't forget this!!
587           }
588
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
593       on your arguments.   Instead, use lower-level control to call exec()
594       directly.
595
596       Here's a safe backtick or pipe open for read:
597
598           my $pid = open(my $kid_to_read, "-|");
599           defined($pid)            || die "can't fork: $!";
600
601           if ($pid) {             # parent
602               while (<$kid_to_read>) {
603                                   # do something interesting
604               }
605               close($kid_to_read)  || warn "kid exited $?";
606
607           } else {                # child
608               ($>, $)) = ($<, $(); # suid only
609               exec($program, @options, @args)
610                                    || die "can't exec program: $!";
611               # NOTREACHED
612           }
613
614       And here's a safe pipe open for writing:
615
616           my $pid = open(my $kid_to_write, "|-");
617           defined($pid)            || die "can't fork: $!";
618
619           $SIG{PIPE} = sub { die "whoops, $program pipe broke" };
620
621           if ($pid) {             # parent
622               print $kid_to_write @data;
623               close($kid_to_write) || warn "kid exited $?";
624
625           } else {                # child
626               ($>, $)) = ($<, $();
627               exec($program, @options, @args)
628                                    || die "can't exec program: $!";
629               # NOTREACHED
630           }
631
632       It is very easy to dead-lock a process using this form of open(), or
633       indeed with any use of pipe() with multiple subprocesses.  The example
634       above is "safe" because it is simple and calls exec().  See "Avoiding
635       Pipe Deadlocks" for general safety principles, but there are extra
636       gotchas with Safe Pipe Opens.
637
638       In particular, if you opened the pipe using "open $fh, "|-"", then you
639       cannot simply use close() in the parent process to close an unwanted
640       writer.  Consider this code:
641
642           my $pid = open(my $writer, "|-");        # fork open a kid
643           defined($pid)               || die "first fork failed: $!";
644           if ($pid) {
645               if (my $sub_pid = fork()) {
646                   defined($sub_pid)   || die "second fork failed: $!";
647                   close($writer)      || die "couldn't close writer: $!";
648                   # now do something else...
649               }
650               else {
651                   # first write to $writer
652                   # ...
653                   # then when finished
654                   close($writer)      || die "couldn't close writer: $!";
655                   exit(0);
656               }
657           }
658           else {
659               # first do something with STDIN, then
660               exit(0);
661           }
662
663       In the example above, the true parent does not want to write to the
664       $writer filehandle, so it closes it.  However, because $writer was
665       opened using "open $fh, "|-"", it has a special behavior: closing it
666       calls waitpid() (see "waitpid" in perlfunc), which waits for the
667       subprocess to exit.  If the child process ends up waiting for something
668       happening in the section marked "do something else", you have deadlock.
669
670       This can also be a problem with intermediate subprocesses in more
671       complicated code, which will call waitpid() on all open filehandles
672       during global destruction--in no predictable order.
673
674       To solve this, you must manually use pipe(), fork(), and the form of
675       open() which sets one file descriptor to another, as shown below:
676
677           pipe(my $reader, my $writer)   || die "pipe failed: $!";
678           my $pid = fork();
679           defined($pid)                  || die "first fork failed: $!";
680           if ($pid) {
681               close $reader;
682               if (my $sub_pid = fork()) {
683                   defined($sub_pid)      || die "first fork failed: $!";
684                   close($writer)         || die "can't close writer: $!";
685               }
686               else {
687                   # write to $writer...
688                   # ...
689                   # then  when finished
690                   close($writer)         || die "can't close writer: $!";
691                   exit(0);
692               }
693               # write to $writer...
694           }
695           else {
696               open(STDIN, "<&", $reader) || die "can't reopen STDIN: $!";
697               close($writer)             || die "can't close writer: $!";
698               # do something...
699               exit(0);
700           }
701
702       Since Perl 5.8.0, you can also use the list form of "open" for pipes.
703       This is preferred when you wish to avoid having the shell interpret
704       metacharacters that may be in your command string.
705
706       So for example, instead of using:
707
708           open(my $ps_pipe, "-|", "ps aux") || die "can't open ps pipe: $!";
709
710       One would use either of these:
711
712           open(my $ps_pipe, "-|", "ps", "aux")
713                                             || die "can't open ps pipe: $!";
714
715           my @ps_args = qw[ ps aux ];
716           open(my $ps_pipe, "-|", @ps_args)
717                                             || die "can't open @ps_args|: $!";
718
719       Because there are more than three arguments to open(), it forks the
720       ps(1) command without spawning a shell, and reads its standard output
721       via the $ps_pipe filehandle.  The corresponding syntax to write to
722       command pipes is to use "|-" in place of "-|".
723
724       This was admittedly a rather silly example, because you're using string
725       literals whose content is perfectly safe.  There is therefore no cause
726       to resort to the harder-to-read, multi-argument form of pipe open().
727       However, whenever you cannot be assured that the program arguments are
728       free of shell metacharacters, the fancier form of open() should be
729       used.  For example:
730
731           my @grep_args = ("egrep", "-i", $some_pattern, @many_files);
732           open(my $grep_pipe, "-|", @grep_args)
733                               || die "can't open @grep_args|: $!";
734
735       Here the multi-argument form of pipe open() is preferred because the
736       pattern and indeed even the filenames themselves might hold
737       metacharacters.
738
739   Avoiding Pipe Deadlocks
740       Whenever you have more than one subprocess, you must be careful that
741       each closes whichever half of any pipes created for interprocess
742       communication it is not using.  This is because any child process
743       reading from the pipe and expecting an EOF will never receive it, and
744       therefore never exit. A single process closing a pipe is not enough to
745       close it; the last process with the pipe open must close it for it to
746       read EOF.
747
748       Certain built-in Unix features help prevent this most of the time.  For
749       instance, filehandles have a "close on exec" flag, which is set en
750       masse under control of the $^F variable.  This is so any filehandles
751       you didn't explicitly route to the STDIN, STDOUT or STDERR of a child
752       program will be automatically closed.
753
754       Always explicitly and immediately call close() on the writable end of
755       any pipe, unless that process is actually writing to it.  Even if you
756       don't explicitly call close(), Perl will still close() all filehandles
757       during global destruction.  As previously discussed, if those
758       filehandles have been opened with Safe Pipe Open, this will result in
759       calling waitpid(), which may again deadlock.
760
761   Bidirectional Communication with Another Process
762       While this works reasonably well for unidirectional communication, what
763       about bidirectional communication?  The most obvious approach doesn't
764       work:
765
766           # THIS DOES NOT WORK!!
767           open(my $prog_for_reading_and_writing, "| some program |")
768
769       If you forget to "use warnings", you'll miss out entirely on the
770       helpful diagnostic message:
771
772           Can't do bidirectional pipe at -e line 1.
773
774       If you really want to, you can use the standard open2() from the
775       IPC::Open2 module to catch both ends.  There's also an open3() in
776       IPC::Open3 for tridirectional I/O so you can also catch your child's
777       STDERR, but doing so would then require an awkward select() loop and
778       wouldn't allow you to use normal Perl input operations.
779
780       If you look at its source, you'll see that open2() uses low-level
781       primitives like the pipe() and exec() syscalls to create all the
782       connections.  Although it might have been more efficient by using
783       socketpair(), this would have been even less portable than it already
784       is. The open2() and open3() functions are unlikely to work anywhere
785       except on a Unix system, or at least one purporting POSIX compliance.
786
787       Here's an example of using open2():
788
789           use IPC::Open2;
790           my $pid = open2(my $reader, my $writer, "cat -un");
791           print $writer "stuff\n";
792           my $got = <$reader>;
793           waitpid $pid, 0;
794
795       The problem with this is that buffering is really going to ruin your
796       day.  Even though your $writer filehandle is auto-flushed so the
797       process on the other end gets your data in a timely manner, you can't
798       usually do anything to force that process to give its data to you in a
799       similarly quick fashion.  In this special case, we could actually so,
800       because we gave cat a -u flag to make it unbuffered.  But very few
801       commands are designed to operate over pipes, so this seldom works
802       unless you yourself wrote the program on the other end of the double-
803       ended pipe.
804
805       A solution to this is to use a library which uses pseudottys to make
806       your program behave more reasonably.  This way you don't have to have
807       control over the source code of the program you're using.  The "Expect"
808       module from CPAN also addresses this kind of thing.  This module
809       requires two other modules from CPAN, "IO::Pty" and "IO::Stty".  It
810       sets up a pseudo terminal to interact with programs that insist on
811       talking to the terminal device driver.  If your system is supported,
812       this may be your best bet.
813
814   Bidirectional Communication with Yourself
815       If you want, you may make low-level pipe() and fork() syscalls to
816       stitch this together by hand.  This example only talks to itself, but
817       you could reopen the appropriate handles to STDIN and STDOUT and call
818       other processes.  (The following example lacks proper error checking.)
819
820        #!/usr/bin/perl
821        # pipe1 - bidirectional communication using two pipe pairs
822        #         designed for the socketpair-challenged
823        use strict;
824        use warnings;
825        use IO::Handle;  # enable autoflush method before Perl 5.14
826        pipe(my $parent_rdr, my $child_wtr);  # XXX: check failure?
827        pipe(my $child_rdr,  my $parent_wtr); # XXX: check failure?
828        $child_wtr->autoflush(1);
829        $parent_wtr->autoflush(1);
830
831        if ($pid = fork()) {
832            close $parent_rdr;
833            close $parent_wtr;
834            print $child_wtr "Parent Pid $$ is sending this\n";
835            chomp(my $line = <$child_rdr>);
836            print "Parent Pid $$ just read this: '$line'\n";
837            close $child_rdr; close $child_wtr;
838            waitpid($pid, 0);
839        } else {
840            die "cannot fork: $!" unless defined $pid;
841            close $child_rdr;
842            close $child_wtr;
843            chomp(my $line = <$parent_rdr>);
844            print "Child Pid $$ just read this: '$line'\n";
845            print $parent_wtr "Child Pid $$ is sending this\n";
846            close $parent_rdr;
847            close $parent_wtr;
848            exit(0);
849        }
850
851       But you don't actually have to make two pipe calls.  If you have the
852       socketpair() system call, it will do this all for you.
853
854        #!/usr/bin/perl
855        # pipe2 - bidirectional communication using socketpair
856        #   "the best ones always go both ways"
857
858        use strict;
859        use warnings;
860        use Socket;
861        use IO::Handle;  # enable autoflush method before Perl 5.14
862
863        # We say AF_UNIX because although *_LOCAL is the
864        # POSIX 1003.1g form of the constant, many machines
865        # still don't have it.
866        socketpair(my $child, my $parent, AF_UNIX, SOCK_STREAM, PF_UNSPEC)
867                                    ||  die "socketpair: $!";
868
869        $child->autoflush(1);
870        $parent->autoflush(1);
871
872        if ($pid = fork()) {
873            close $parent;
874            print $child "Parent Pid $$ is sending this\n";
875            chomp(my $line = <$child>);
876            print "Parent Pid $$ just read this: '$line'\n";
877            close $child;
878            waitpid($pid, 0);
879        } else {
880            die "cannot fork: $!" unless defined $pid;
881            close $child;
882            chomp(my $line = <$parent>);
883            print "Child Pid $$ just read this: '$line'\n";
884            print $parent "Child Pid $$ is sending this\n";
885            close $parent;
886            exit(0);
887        }
888

Sockets: Client/Server Communication

890       While not entirely limited to Unix-derived operating systems (e.g.,
891       WinSock on PCs provides socket support, as do some VMS libraries), you
892       might not have sockets on your system, in which case this section
893       probably isn't going to do you much good.  With sockets, you can do
894       both virtual circuits like TCP streams and datagrams like UDP packets.
895       You may be able to do even more depending on your system.
896
897       The Perl functions for dealing with sockets have the same names as the
898       corresponding system calls in C, but their arguments tend to differ for
899       two reasons.  First, Perl filehandles work differently than C file
900       descriptors.  Second, Perl already knows the length of its strings, so
901       you don't need to pass that information.
902
903       One of the major problems with ancient, antemillennial socket code in
904       Perl was that it used hard-coded values for some of the constants,
905       which severely hurt portability.  If you ever see code that does
906       anything like explicitly setting "$AF_INET = 2", you know you're in for
907       big trouble.  An immeasurably superior approach is to use the Socket
908       module, which more reliably grants access to the various constants and
909       functions you'll need.
910
911       If you're not writing a server/client for an existing protocol like
912       NNTP or SMTP, you should give some thought to how your server will know
913       when the client has finished talking, and vice-versa.  Most protocols
914       are based on one-line messages and responses (so one party knows the
915       other has finished when a "\n" is received) or multi-line messages and
916       responses that end with a period on an empty line ("\n.\n" terminates a
917       message/response).
918
919   Internet Line Terminators
920       The Internet line terminator is "\015\012".  Under ASCII variants of
921       Unix, that could usually be written as "\r\n", but under other systems,
922       "\r\n" might at times be "\015\015\012", "\012\012\015", or something
923       completely different.  The standards specify writing "\015\012" to be
924       conformant (be strict in what you provide), but they also recommend
925       accepting a lone "\012" on input (be lenient in what you require).  We
926       haven't always been very good about that in the code in this manpage,
927       but unless you're on a Mac from way back in its pre-Unix dark ages,
928       you'll probably be ok.
929
930   Internet TCP Clients and Servers
931       Use Internet-domain sockets when you want to do client-server
932       communication that might extend to machines outside of your own system.
933
934       Here's a sample TCP client using Internet-domain sockets:
935
936           #!/usr/bin/perl
937           use strict;
938           use warnings;
939           use Socket;
940
941           my $remote  = shift || "localhost";
942           my $port    = shift || 2345;  # random port
943           if ($port =~ /\D/) { $port = getservbyname($port, "tcp") }
944           die "No port" unless $port;
945           my $iaddr   = inet_aton($remote)       || die "no host: $remote";
946           my $paddr   = sockaddr_in($port, $iaddr);
947
948           my $proto   = getprotobyname("tcp");
949           socket(my $sock, PF_INET, SOCK_STREAM, $proto)  || die "socket: $!";
950           connect($sock, $paddr)              || die "connect: $!";
951           while (my $line = <$sock>) {
952               print $line;
953           }
954
955           close ($sock)                        || die "close: $!";
956           exit(0);
957
958       And here's a corresponding server to go along with it.  We'll leave the
959       address as "INADDR_ANY" so that the kernel can choose the appropriate
960       interface on multihomed hosts.  If you want sit on a particular
961       interface (like the external side of a gateway or firewall machine),
962       fill this in with your real address instead.
963
964        #!/usr/bin/perl -T
965        use strict;
966        use warnings;
967        BEGIN { $ENV{PATH} = "/usr/bin:/bin" }
968        use Socket;
969        use Carp;
970        my $EOL = "\015\012";
971
972        sub logmsg { print "$0 $$: @_ at ", scalar localtime(), "\n" }
973
974        my $port  = shift || 2345;
975        die "invalid port" unless $port =~ /^ \d+ $/x;
976
977        my $proto = getprotobyname("tcp");
978
979        socket(my $server, PF_INET, SOCK_STREAM, $proto) || die "socket: $!";
980        setsockopt($server, SOL_SOCKET, SO_REUSEADDR, pack("l", 1))
981                                                      || die "setsockopt: $!";
982        bind($server, sockaddr_in($port, INADDR_ANY)) || die "bind: $!";
983        listen($server, SOMAXCONN)                    || die "listen: $!";
984
985        logmsg "server started on port $port";
986
987        for (my $paddr; $paddr = accept(my $client, $server); close $client) {
988            my($port, $iaddr) = sockaddr_in($paddr);
989            my $name = gethostbyaddr($iaddr, AF_INET);
990
991            logmsg "connection from $name [",
992                    inet_ntoa($iaddr), "]
993                    at port $port";
994
995            print $client "Hello there, $name, it's now ",
996                            scalar localtime(), $EOL;
997        }
998
999       And here's a multitasking version.  It's multitasked in that like most
1000       typical servers, it spawns (fork()s) a slave server to handle the
1001       client request so that the master server can quickly go back to service
1002       a new client.
1003
1004        #!/usr/bin/perl -T
1005        use strict;
1006        use warnings;
1007        BEGIN { $ENV{PATH} = "/usr/bin:/bin" }
1008        use Socket;
1009        use Carp;
1010        my $EOL = "\015\012";
1011
1012        sub spawn;  # forward declaration
1013        sub logmsg { print "$0 $$: @_ at ", scalar localtime(), "\n" }
1014
1015        my $port  = shift || 2345;
1016        die "invalid port" unless $port =~ /^ \d+ $/x;
1017
1018        my $proto = getprotobyname("tcp");
1019
1020        socket(my $server, PF_INET, SOCK_STREAM, $proto) || die "socket: $!";
1021        setsockopt($server, SOL_SOCKET, SO_REUSEADDR, pack("l", 1))
1022                                                      || die "setsockopt: $!";
1023        bind($server, sockaddr_in($port, INADDR_ANY)) || die "bind: $!";
1024        listen($server, SOMAXCONN)                    || die "listen: $!";
1025
1026        logmsg "server started on port $port";
1027
1028        my $waitedpid = 0;
1029
1030        use POSIX ":sys_wait_h";
1031        use Errno;
1032
1033        sub REAPER {
1034            local $!;   # don't let waitpid() overwrite current error
1035            while ((my $pid = waitpid(-1, WNOHANG)) > 0 && WIFEXITED($?)) {
1036                logmsg "reaped $waitedpid" . ($? ? " with exit $?" : "");
1037            }
1038            $SIG{CHLD} = \&REAPER;  # loathe SysV
1039        }
1040
1041        $SIG{CHLD} = \&REAPER;
1042
1043        while (1) {
1044            my $paddr = accept(my $client, $server) || do {
1045                # try again if accept() returned because got a signal
1046                next if $!{EINTR};
1047                die "accept: $!";
1048            };
1049            my ($port, $iaddr) = sockaddr_in($paddr);
1050            my $name = gethostbyaddr($iaddr, AF_INET);
1051
1052            logmsg "connection from $name [",
1053                   inet_ntoa($iaddr),
1054                   "] at port $port";
1055
1056            spawn $client, sub {
1057                $| = 1;
1058                print "Hello there, $name, it's now ",
1059                      scalar localtime(),
1060                      $EOL;
1061                exec "/usr/games/fortune"       # XXX: "wrong" line terminators
1062                    or confess "can't exec fortune: $!";
1063            };
1064            close $client;
1065        }
1066
1067        sub spawn {
1068            my $client = shift;
1069            my $coderef = shift;
1070
1071            unless (@_ == 0 && $coderef && ref($coderef) eq "CODE") {
1072                confess "usage: spawn CLIENT CODEREF";
1073            }
1074
1075            my $pid;
1076            unless (defined($pid = fork())) {
1077                logmsg "cannot fork: $!";
1078                return;
1079            }
1080            elsif ($pid) {
1081                logmsg "begat $pid";
1082                return; # I'm the parent
1083            }
1084            # else I'm the child -- go spawn
1085
1086            open(STDIN,  "<&", $client)   || die "can't dup client to stdin";
1087            open(STDOUT, ">&", $client)   || die "can't dup client to stdout";
1088            ## open(STDERR, ">&", STDOUT) || die "can't dup stdout to stderr";
1089            exit($coderef->());
1090        }
1091
1092       This server takes the trouble to clone off a child version via fork()
1093       for each incoming request.  That way it can handle many requests at
1094       once, which you might not always want.  Even if you don't fork(), the
1095       listen() will allow that many pending connections.  Forking servers
1096       have to be particularly careful about cleaning up their dead children
1097       (called "zombies" in Unix parlance), because otherwise you'll quickly
1098       fill up your process table.  The REAPER subroutine is used here to call
1099       waitpid() for any child processes that have finished, thereby ensuring
1100       that they terminate cleanly and don't join the ranks of the living
1101       dead.
1102
1103       Within the while loop we call accept() and check to see if it returns a
1104       false value.  This would normally indicate a system error needs to be
1105       reported.  However, the introduction of safe signals (see "Deferred
1106       Signals (Safe Signals)" above) in Perl 5.8.0 means that accept() might
1107       also be interrupted when the process receives a signal.  This typically
1108       happens when one of the forked subprocesses exits and notifies the
1109       parent process with a CHLD signal.
1110
1111       If accept() is interrupted by a signal, $! will be set to EINTR.  If
1112       this happens, we can safely continue to the next iteration of the loop
1113       and another call to accept().  It is important that your signal
1114       handling code not modify the value of $!, or else this test will likely
1115       fail.  In the REAPER subroutine we create a local version of $! before
1116       calling waitpid().  When waitpid() sets $! to ECHILD as it inevitably
1117       does when it has no more children waiting, it updates the local copy
1118       and leaves the original unchanged.
1119
1120       You should use the -T flag to enable taint checking (see perlsec) even
1121       if we aren't running setuid or setgid.  This is always a good idea for
1122       servers or any program run on behalf of someone else (like CGI
1123       scripts), because it lessens the chances that people from the outside
1124       will be able to compromise your system.
1125
1126       Let's look at another TCP client.  This one connects to the TCP "time"
1127       service on a number of different machines and shows how far their
1128       clocks differ from the system on which it's being run:
1129
1130           #!/usr/bin/perl
1131           use strict;
1132           use warnings;
1133           use Socket;
1134
1135           my $SECS_OF_70_YEARS = 2208988800;
1136           sub ctime { scalar localtime(shift() || time()) }
1137
1138           my $iaddr = gethostbyname("localhost");
1139           my $proto = getprotobyname("tcp");
1140           my $port = getservbyname("time", "tcp");
1141           my $paddr = sockaddr_in(0, $iaddr);
1142
1143           $| = 1;
1144           printf "%-24s %8s %s\n", "localhost", 0, ctime();
1145
1146           foreach my $host (@ARGV) {
1147               printf "%-24s ", $host;
1148               my $hisiaddr = inet_aton($host)     || die "unknown host";
1149               my $hispaddr = sockaddr_in($port, $hisiaddr);
1150               socket(my $socket, PF_INET, SOCK_STREAM, $proto)
1151                                                   || die "socket: $!";
1152               connect($socket, $hispaddr)         || die "connect: $!";
1153               my $rtime = pack("C4", ());
1154               read($socket, $rtime, 4);
1155               close($socket);
1156               my $histime = unpack("N", $rtime) - $SECS_OF_70_YEARS;
1157               printf "%8d %s\n", $histime - time(), ctime($histime);
1158           }
1159
1160   Unix-Domain TCP Clients and Servers
1161       That's fine for Internet-domain clients and servers, but what about
1162       local communications?  While you can use the same setup, sometimes you
1163       don't want to.  Unix-domain sockets are local to the current host, and
1164       are often used internally to implement pipes.  Unlike Internet domain
1165       sockets, Unix domain sockets can show up in the file system with an
1166       ls(1) listing.
1167
1168           % ls -l /dev/log
1169           srw-rw-rw-  1 root            0 Oct 31 07:23 /dev/log
1170
1171       You can test for these with Perl's -S file test:
1172
1173           unless (-S "/dev/log") {
1174               die "something's wicked with the log system";
1175           }
1176
1177       Here's a sample Unix-domain client:
1178
1179           #!/usr/bin/perl
1180           use Socket;
1181           use strict;
1182           use warnings;
1183
1184           my $rendezvous = shift || "catsock";
1185           socket(my $sock, PF_UNIX, SOCK_STREAM, 0) || die "socket: $!";
1186           connect($sock, sockaddr_un($rendezvous))  || die "connect: $!";
1187           while (defined(my $line = <$sock>)) {
1188               print $line;
1189           }
1190           exit(0);
1191
1192       And here's a corresponding server.  You don't have to worry about silly
1193       network terminators here because Unix domain sockets are guaranteed to
1194       be on the localhost, and thus everything works right.
1195
1196           #!/usr/bin/perl -T
1197           use strict;
1198           use warnings;
1199           use Socket;
1200           use Carp;
1201
1202           BEGIN { $ENV{PATH} = "/usr/bin:/bin" }
1203           sub spawn;  # forward declaration
1204           sub logmsg { print "$0 $$: @_ at ", scalar localtime(), "\n" }
1205
1206           my $NAME = "catsock";
1207           my $uaddr = sockaddr_un($NAME);
1208           my $proto = getprotobyname("tcp");
1209
1210           socket(my $server, PF_UNIX, SOCK_STREAM, 0) || die "socket: $!";
1211           unlink($NAME);
1212           bind  ($server, $uaddr)                     || die "bind: $!";
1213           listen($server, SOMAXCONN)                  || die "listen: $!";
1214
1215           logmsg "server started on $NAME";
1216
1217           my $waitedpid;
1218
1219           use POSIX ":sys_wait_h";
1220           sub REAPER {
1221               my $child;
1222               while (($waitedpid = waitpid(-1, WNOHANG)) > 0) {
1223                   logmsg "reaped $waitedpid" . ($? ? " with exit $?" : "");
1224               }
1225               $SIG{CHLD} = \&REAPER;  # loathe SysV
1226           }
1227
1228           $SIG{CHLD} = \&REAPER;
1229
1230
1231           for ( $waitedpid = 0;
1232                 accept(my $client, $server) || $waitedpid;
1233                 $waitedpid = 0, close $client)
1234           {
1235               next if $waitedpid;
1236               logmsg "connection on $NAME";
1237               spawn $client, sub {
1238                   print "Hello there, it's now ", scalar localtime(), "\n";
1239                   exec("/usr/games/fortune")  || die "can't exec fortune: $!";
1240               };
1241           }
1242
1243           sub spawn {
1244               my $client = shift();
1245               my $coderef = shift();
1246
1247               unless (@_ == 0 && $coderef && ref($coderef) eq "CODE") {
1248                   confess "usage: spawn CLIENT CODEREF";
1249               }
1250
1251               my $pid;
1252               unless (defined($pid = fork())) {
1253                   logmsg "cannot fork: $!";
1254                   return;
1255               }
1256               elsif ($pid) {
1257                   logmsg "begat $pid";
1258                   return; # I'm the parent
1259               }
1260               else {
1261                   # I'm the child -- go spawn
1262               }
1263
1264               open(STDIN,  "<&", $client)
1265                   || die "can't dup client to stdin";
1266               open(STDOUT, ">&", $client)
1267                   || die "can't dup client to stdout";
1268               ## open(STDERR, ">&", STDOUT)
1269               ##  || die "can't dup stdout to stderr";
1270               exit($coderef->());
1271           }
1272
1273       As you see, it's remarkably similar to the Internet domain TCP server,
1274       so much so, in fact, that we've omitted several duplicate
1275       functions--spawn(), logmsg(), ctime(), and REAPER()--which are the same
1276       as in the other server.
1277
1278       So why would you ever want to use a Unix domain socket instead of a
1279       simpler named pipe?  Because a named pipe doesn't give you sessions.
1280       You can't tell one process's data from another's.  With socket
1281       programming, you get a separate session for each client; that's why
1282       accept() takes two arguments.
1283
1284       For example, let's say that you have a long-running database server
1285       daemon that you want folks to be able to access from the Web, but only
1286       if they go through a CGI interface.  You'd have a small, simple CGI
1287       program that does whatever checks and logging you feel like, and then
1288       acts as a Unix-domain client and connects to your private server.
1289

TCP Clients with IO::Socket

1291       For those preferring a higher-level interface to socket programming,
1292       the IO::Socket module provides an object-oriented approach.  If for
1293       some reason you lack this module, you can just fetch IO::Socket from
1294       CPAN, where you'll also find modules providing easy interfaces to the
1295       following systems: DNS, FTP, Ident (RFC 931), NIS and NISPlus, NNTP,
1296       Ping, POP3, SMTP, SNMP, SSLeay, Telnet, and Time--to name just a few.
1297
1298   A Simple Client
1299       Here's a client that creates a TCP connection to the "daytime" service
1300       at port 13 of the host name "localhost" and prints out everything that
1301       the server there cares to provide.
1302
1303           #!/usr/bin/perl
1304           use strict;
1305           use warnings;
1306           use IO::Socket;
1307           my $remote = IO::Socket::INET->new(
1308                               Proto    => "tcp",
1309                               PeerAddr => "localhost",
1310                               PeerPort => "daytime(13)",
1311                           )
1312                        || die "can't connect to daytime service on localhost";
1313           while (<$remote>) { print }
1314
1315       When you run this program, you should get something back that looks
1316       like this:
1317
1318           Wed May 14 08:40:46 MDT 1997
1319
1320       Here are what those parameters to the new() constructor mean:
1321
1322       "Proto"
1323           This is which protocol to use.  In this case, the socket handle
1324           returned will be connected to a TCP socket, because we want a
1325           stream-oriented connection, that is, one that acts pretty much like
1326           a plain old file.  Not all sockets are this of this type.  For
1327           example, the UDP protocol can be used to make a datagram socket,
1328           used for message-passing.
1329
1330       "PeerAddr"
1331           This is the name or Internet address of the remote host the server
1332           is running on.  We could have specified a longer name like
1333           "www.perl.com", or an address like "207.171.7.72".  For
1334           demonstration purposes, we've used the special hostname
1335           "localhost", which should always mean the current machine you're
1336           running on.  The corresponding Internet address for localhost is
1337           "127.0.0.1", if you'd rather use that.
1338
1339       "PeerPort"
1340           This is the service name or port number we'd like to connect to.
1341           We could have gotten away with using just "daytime" on systems with
1342           a well-configured system services file,[FOOTNOTE: The system
1343           services file is found in /etc/services under Unixy systems.] but
1344           here we've specified the port number (13) in parentheses.  Using
1345           just the number would have also worked, but numeric literals make
1346           careful programmers nervous.
1347
1348   A Webget Client
1349       Here's a simple client that takes a remote host to fetch a document
1350       from, and then a list of files to get from that host.  This is a more
1351       interesting client than the previous one because it first sends
1352       something to the server before fetching the server's response.
1353
1354           #!/usr/bin/perl
1355           use strict;
1356           use warnings;
1357           use IO::Socket;
1358           unless (@ARGV > 1) { die "usage: $0 host url ..." }
1359           my $host = shift(@ARGV);
1360           my $EOL = "\015\012";
1361           my $BLANK = $EOL x 2;
1362           for my $document (@ARGV) {
1363               my $remote = IO::Socket::INET->new( Proto     => "tcp",
1364                                                   PeerAddr  => $host,
1365                                                   PeerPort  => "http(80)",
1366                         )     || die "cannot connect to httpd on $host";
1367               $remote->autoflush(1);
1368               print $remote "GET $document HTTP/1.0" . $BLANK;
1369               while ( <$remote> ) { print }
1370               close $remote;
1371           }
1372
1373       The web server handling the HTTP service is assumed to be at its
1374       standard port, number 80.  If the server you're trying to connect to is
1375       at a different port, like 1080 or 8080, you should specify it as the
1376       named-parameter pair, "PeerPort => 8080".  The "autoflush" method is
1377       used on the socket because otherwise the system would buffer up the
1378       output we sent it.  (If you're on a prehistoric Mac, you'll also need
1379       to change every "\n" in your code that sends data over the network to
1380       be a "\015\012" instead.)
1381
1382       Connecting to the server is only the first part of the process: once
1383       you have the connection, you have to use the server's language.  Each
1384       server on the network has its own little command language that it
1385       expects as input.  The string that we send to the server starting with
1386       "GET" is in HTTP syntax.  In this case, we simply request each
1387       specified document.  Yes, we really are making a new connection for
1388       each document, even though it's the same host.  That's the way you
1389       always used to have to speak HTTP.  Recent versions of web browsers may
1390       request that the remote server leave the connection open a little
1391       while, but the server doesn't have to honor such a request.
1392
1393       Here's an example of running that program, which we'll call webget:
1394
1395           % webget www.perl.com /guanaco.html
1396           HTTP/1.1 404 File Not Found
1397           Date: Thu, 08 May 1997 18:02:32 GMT
1398           Server: Apache/1.2b6
1399           Connection: close
1400           Content-type: text/html
1401
1402           <HEAD><TITLE>404 File Not Found</TITLE></HEAD>
1403           <BODY><H1>File Not Found</H1>
1404           The requested URL /guanaco.html was not found on this server.<P>
1405           </BODY>
1406
1407       Ok, so that's not very interesting, because it didn't find that
1408       particular document.  But a long response wouldn't have fit on this
1409       page.
1410
1411       For a more featureful version of this program, you should look to the
1412       lwp-request program included with the LWP modules from CPAN.
1413
1414   Interactive Client with IO::Socket
1415       Well, that's all fine if you want to send one command and get one
1416       answer, but what about setting up something fully interactive, somewhat
1417       like the way telnet works?  That way you can type a line, get the
1418       answer, type a line, get the answer, etc.
1419
1420       This client is more complicated than the two we've done so far, but if
1421       you're on a system that supports the powerful "fork" call, the solution
1422       isn't that rough.  Once you've made the connection to whatever service
1423       you'd like to chat with, call "fork" to clone your process.  Each of
1424       these two identical process has a very simple job to do: the parent
1425       copies everything from the socket to standard output, while the child
1426       simultaneously copies everything from standard input to the socket.  To
1427       accomplish the same thing using just one process would be much harder,
1428       because it's easier to code two processes to do one thing than it is to
1429       code one process to do two things.  (This keep-it-simple principle a
1430       cornerstones of the Unix philosophy, and good software engineering as
1431       well, which is probably why it's spread to other systems.)
1432
1433       Here's the code:
1434
1435           #!/usr/bin/perl
1436           use strict;
1437           use warnings;
1438           use IO::Socket;
1439
1440           unless (@ARGV == 2) { die "usage: $0 host port" }
1441           my ($host, $port) = @ARGV;
1442
1443           # create a tcp connection to the specified host and port
1444           my $handle = IO::Socket::INET->new(Proto     => "tcp",
1445                                              PeerAddr  => $host,
1446                                              PeerPort  => $port)
1447                      || die "can't connect to port $port on $host: $!";
1448
1449           $handle->autoflush(1);       # so output gets there right away
1450           print STDERR "[Connected to $host:$port]\n";
1451
1452           # split the program into two processes, identical twins
1453           die "can't fork: $!" unless defined(my $kidpid = fork());
1454
1455           # the if{} block runs only in the parent process
1456           if ($kidpid) {
1457               # copy the socket to standard output
1458               while (defined (my $line = <$handle>)) {
1459                   print STDOUT $line;
1460               }
1461               kill("TERM", $kidpid);   # send SIGTERM to child
1462           }
1463           # the else{} block runs only in the child process
1464           else {
1465               # copy standard input to the socket
1466               while (defined (my $line = <STDIN>)) {
1467                   print $handle $line;
1468               }
1469               exit(0);                # just in case
1470           }
1471
1472       The "kill" function in the parent's "if" block is there to send a
1473       signal to our child process, currently running in the "else" block, as
1474       soon as the remote server has closed its end of the connection.
1475
1476       If the remote server sends data a byte at time, and you need that data
1477       immediately without waiting for a newline (which might not happen), you
1478       may wish to replace the "while" loop in the parent with the following:
1479
1480           my $byte;
1481           while (sysread($handle, $byte, 1) == 1) {
1482               print STDOUT $byte;
1483           }
1484
1485       Making a system call for each byte you want to read is not very
1486       efficient (to put it mildly) but is the simplest to explain and works
1487       reasonably well.
1488

TCP Servers with IO::Socket

1490       As always, setting up a server is little bit more involved than running
1491       a client.  The model is that the server creates a special kind of
1492       socket that does nothing but listen on a particular port for incoming
1493       connections.  It does this by calling the "IO::Socket::INET->new()"
1494       method with slightly different arguments than the client did.
1495
1496       Proto
1497           This is which protocol to use.  Like our clients, we'll still
1498           specify "tcp" here.
1499
1500       LocalPort
1501           We specify a local port in the "LocalPort" argument, which we
1502           didn't do for the client.  This is service name or port number for
1503           which you want to be the server. (Under Unix, ports under 1024 are
1504           restricted to the superuser.)  In our sample, we'll use port 9000,
1505           but you can use any port that's not currently in use on your
1506           system.  If you try to use one already in used, you'll get an
1507           "Address already in use" message.  Under Unix, the "netstat -a"
1508           command will show which services current have servers.
1509
1510       Listen
1511           The "Listen" parameter is set to the maximum number of pending
1512           connections we can accept until we turn away incoming clients.
1513           Think of it as a call-waiting queue for your telephone.  The low-
1514           level Socket module has a special symbol for the system maximum,
1515           which is SOMAXCONN.
1516
1517       Reuse
1518           The "Reuse" parameter is needed so that we restart our server
1519           manually without waiting a few minutes to allow system buffers to
1520           clear out.
1521
1522       Once the generic server socket has been created using the parameters
1523       listed above, the server then waits for a new client to connect to it.
1524       The server blocks in the "accept" method, which eventually accepts a
1525       bidirectional connection from the remote client.  (Make sure to
1526       autoflush this handle to circumvent buffering.)
1527
1528       To add to user-friendliness, our server prompts the user for commands.
1529       Most servers don't do this.  Because of the prompt without a newline,
1530       you'll have to use the "sysread" variant of the interactive client
1531       above.
1532
1533       This server accepts one of five different commands, sending output back
1534       to the client.  Unlike most network servers, this one handles only one
1535       incoming client at a time.  Multitasking servers are covered in Chapter
1536       16 of the Camel.
1537
1538       Here's the code.
1539
1540        #!/usr/bin/perl
1541        use strict;
1542        use warnings;
1543        use IO::Socket;
1544        use Net::hostent;      # for OOish version of gethostbyaddr
1545
1546        my $PORT = 9000;       # pick something not in use
1547
1548        my $server = IO::Socket::INET->new( Proto     => "tcp",
1549                                            LocalPort => $PORT,
1550                                            Listen    => SOMAXCONN,
1551                                            Reuse     => 1);
1552
1553        die "can't setup server" unless $server;
1554        print "[Server $0 accepting clients]\n";
1555
1556        while (my $client = $server->accept()) {
1557          $client->autoflush(1);
1558          print $client "Welcome to $0; type help for command list.\n";
1559          my $hostinfo = gethostbyaddr($client->peeraddr);
1560          printf "[Connect from %s]\n",
1561                 $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`       }
1570            else {
1571              print $client "Commands: quit date who cookie motd\n";
1572            }
1573          } continue {
1574             print $client "Command? ";
1575          }
1576          close $client;
1577        }
1578

UDP: Message Passing

1580       Another kind of client-server setup is one that uses not connections,
1581       but messages.  UDP communications involve much lower overhead but also
1582       provide less reliability, as there are no promises that messages will
1583       arrive at all, let alone in order and unmangled.  Still, UDP offers
1584       some advantages over TCP, including being able to "broadcast" or
1585       "multicast" to a whole bunch of destination hosts at once (usually on
1586       your local subnet).  If you find yourself overly concerned about
1587       reliability and start building checks into your message system, then
1588       you probably should use just TCP to start with.
1589
1590       UDP datagrams are not a bytestream and should not be treated as such.
1591       This makes using I/O mechanisms with internal buffering like stdio
1592       (i.e.  print() and friends) especially cumbersome. Use syswrite(), or
1593       better send(), like in the example below.
1594
1595       Here's a UDP program similar to the sample Internet TCP client given
1596       earlier.  However, instead of checking one host at a time, the UDP
1597       version will check many of them asynchronously by simulating a
1598       multicast and then using select() to do a timed-out wait for I/O.  To
1599       do something similar with TCP, you'd have to use a different socket
1600       handle for each host.
1601
1602        #!/usr/bin/perl
1603        use strict;
1604        use warnings;
1605        use Socket;
1606        use Sys::Hostname;
1607
1608        my $SECS_OF_70_YEARS = 2_208_988_800;
1609
1610        my $iaddr = gethostbyname(hostname());
1611        my $proto = getprotobyname("udp");
1612        my $port = getservbyname("time", "udp");
1613        my $paddr = sockaddr_in(0, $iaddr); # 0 means let kernel pick
1614
1615        socket(my $socket, PF_INET, SOCK_DGRAM, $proto) || die "socket: $!";
1616        bind($socket, $paddr)                           || die "bind: $!";
1617
1618        $| = 1;
1619        printf "%-12s %8s %s\n",  "localhost", 0, scalar localtime();
1620        my $count = 0;
1621        for my $host (@ARGV) {
1622            $count++;
1623            my $hisiaddr = inet_aton($host)         || die "unknown host";
1624            my $hispaddr = sockaddr_in($port, $hisiaddr);
1625            defined(send($socket, 0, 0, $hispaddr)) || die "send $host: $!";
1626        }
1627
1628        my $rout = my $rin = "";
1629        vec($rin, fileno($socket), 1) = 1;
1630
1631        # timeout after 10.0 seconds
1632        while ($count && select($rout = $rin, undef, undef, 10.0)) {
1633            my $rtime = "";
1634            my $hispaddr = recv($socket, $rtime, 4, 0) || die "recv: $!";
1635            my ($port, $hisiaddr) = sockaddr_in($hispaddr);
1636            my $host = gethostbyaddr($hisiaddr, AF_INET);
1637            my $histime = unpack("N", $rtime) - $SECS_OF_70_YEARS;
1638            printf "%-12s ", $host;
1639            printf "%8d %s\n", $histime - time(), scalar localtime($histime);
1640            $count--;
1641        }
1642
1643       This example does not include any retries and may consequently fail to
1644       contact a reachable host. The most prominent reason for this is
1645       congestion of the queues on the sending host if the number of hosts to
1646       contact is sufficiently large.
1647

SysV IPC

1649       While System V IPC isn't so widely used as sockets, it still has some
1650       interesting uses.  However, you cannot use SysV IPC or Berkeley mmap()
1651       to have a variable shared amongst several processes.  That's because
1652       Perl would reallocate your string when you weren't wanting it to.  You
1653       might look into the "IPC::Shareable" or "threads::shared" modules for
1654       that.
1655
1656       Here's a small example showing shared memory usage.
1657
1658           use IPC::SysV qw(IPC_PRIVATE IPC_RMID S_IRUSR S_IWUSR);
1659
1660           my $size = 2000;
1661           my $id = shmget(IPC_PRIVATE, $size, S_IRUSR | S_IWUSR);
1662           defined($id)                    || die "shmget: $!";
1663           print "shm key $id\n";
1664
1665           my $message = "Message #1";
1666           shmwrite($id, $message, 0, 60)  || die "shmwrite: $!";
1667           print "wrote: '$message'\n";
1668           shmread($id, my $buff, 0, 60)      || die "shmread: $!";
1669           print "read : '$buff'\n";
1670
1671           # the buffer of shmread is zero-character end-padded.
1672           substr($buff, index($buff, "\0")) = "";
1673           print "un" unless $buff eq $message;
1674           print "swell\n";
1675
1676           print "deleting shm $id\n";
1677           shmctl($id, IPC_RMID, 0)        || die "shmctl: $!";
1678
1679       Here's an example of a semaphore:
1680
1681           use IPC::SysV qw(IPC_CREAT);
1682
1683           my $IPC_KEY = 1234;
1684           my $id = semget($IPC_KEY, 10, 0666 | IPC_CREAT);
1685           defined($id)                    || die "semget: $!";
1686           print "sem id $id\n";
1687
1688       Put this code in a separate file to be run in more than one process.
1689       Call the file take:
1690
1691           # create a semaphore
1692
1693           my $IPC_KEY = 1234;
1694           my $id = semget($IPC_KEY, 0, 0);
1695           defined($id)                    || die "semget: $!";
1696
1697           my $semnum  = 0;
1698           my $semflag = 0;
1699
1700           # "take" semaphore
1701           # wait for semaphore to be zero
1702           my $semop = 0;
1703           my $opstring1 = pack("s!s!s!", $semnum, $semop, $semflag);
1704
1705           # Increment the semaphore count
1706           $semop = 1;
1707           my $opstring2 = pack("s!s!s!", $semnum, $semop,  $semflag);
1708           my $opstring  = $opstring1 . $opstring2;
1709
1710           semop($id, $opstring)   || die "semop: $!";
1711
1712       Put this code in a separate file to be run in more than one process.
1713       Call this file give:
1714
1715           # "give" the semaphore
1716           # run this in the original process and you will see
1717           # that the second process continues
1718
1719           my $IPC_KEY = 1234;
1720           my $id = semget($IPC_KEY, 0, 0);
1721           die unless defined($id);
1722
1723           my $semnum  = 0;
1724           my $semflag = 0;
1725
1726           # Decrement the semaphore count
1727           my $semop = -1;
1728           my $opstring = pack("s!s!s!", $semnum, $semop, $semflag);
1729
1730           semop($id, $opstring)   || die "semop: $!";
1731
1732       The SysV IPC code above was written long ago, and it's definitely
1733       clunky looking.  For a more modern look, see the IPC::SysV module.
1734
1735       A small example demonstrating SysV message queues:
1736
1737           use IPC::SysV qw(IPC_PRIVATE IPC_RMID IPC_CREAT S_IRUSR S_IWUSR);
1738
1739           my $id = msgget(IPC_PRIVATE, IPC_CREAT | S_IRUSR | S_IWUSR);
1740           defined($id)                || die "msgget failed: $!";
1741
1742           my $sent      = "message";
1743           my $type_sent = 1234;
1744
1745           msgsnd($id, pack("l! a*", $type_sent, $sent), 0)
1746                                       || die "msgsnd failed: $!";
1747
1748           msgrcv($id, my $rcvd_buf, 60, 0, 0)
1749                                       || die "msgrcv failed: $!";
1750
1751           my($type_rcvd, $rcvd) = unpack("l! a*", $rcvd_buf);
1752
1753           if ($rcvd eq $sent) {
1754               print "okay\n";
1755           } else {
1756               print "not okay\n";
1757           }
1758
1759           msgctl($id, IPC_RMID, 0)    || die "msgctl failed: $!\n";
1760

NOTES

1762       Most of these routines quietly but politely return "undef" when they
1763       fail instead of causing your program to die right then and there due to
1764       an uncaught exception.  (Actually, some of the new Socket conversion
1765       functions do croak() on bad arguments.)  It is therefore essential to
1766       check return values from these functions.  Always begin your socket
1767       programs this way for optimal success, and don't forget to add the -T
1768       taint-checking flag to the "#!" line for servers:
1769
1770           #!/usr/bin/perl -T
1771           use strict;
1772           use warnings;
1773           use sigtrap;
1774           use Socket;
1775

BUGS

1777       These routines all create system-specific portability problems.  As
1778       noted elsewhere, Perl is at the mercy of your C libraries for much of
1779       its system behavior.  It's probably safest to assume broken SysV
1780       semantics for signals and to stick with simple TCP and UDP socket
1781       operations; e.g., don't try to pass open file descriptors over a local
1782       UDP datagram socket if you want your code to stand a chance of being
1783       portable.
1784

AUTHOR

1786       Tom Christiansen, with occasional vestiges of Larry Wall's original
1787       version and suggestions from the Perl Porters.
1788

SEE ALSO

1790       There's a lot more to networking than this, but this should get you
1791       started.
1792
1793       For intrepid programmers, the indispensable textbook is Unix Network
1794       Programming, 2nd Edition, Volume 1 by W. Richard Stevens (published by
1795       Prentice-Hall).  Most books on networking address the subject from the
1796       perspective of a C programmer; translation to Perl is left as an
1797       exercise for the reader.
1798
1799       The IO::Socket(3) manpage describes the object library, and the
1800       Socket(3) manpage describes the low-level interface to sockets.
1801       Besides the obvious functions in perlfunc, you should also check out
1802       the modules file at your nearest CPAN site, especially
1803       <http://www.cpan.org/modules/00modlist.long.html#ID5_Networking_>.  See
1804       perlmodlib or best yet, the Perl FAQ for a description of what CPAN is
1805       and where to get it if the previous link doesn't work for you.
1806
1807       Section 5 of CPAN's modules file is devoted to "Networking, Device
1808       Control (modems), and Interprocess Communication", and contains
1809       numerous unbundled modules numerous networking modules, Chat and Expect
1810       operations, CGI programming, DCE, FTP, IPC, NNTP, Proxy, Ptty, RPC,
1811       SNMP, SMTP, Telnet, Threads, and ToolTalk--to name just a few.
1812
1813
1814
1815perl v5.32.1                      2021-05-31                        PERLIPC(1)
Impressum