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 or
14 references of user-installed signal handlers. These handlers will be called
15 with an argument which is the name of the signal that triggered it. A
16 signal may be generated intentionally from a particular keyboard sequence like
17 control-C or control-Z, sent to you from another process, or
18 triggered automatically by the kernel when special events transpire, like
19 a child process exiting, your process running out of stack space, or
20 hitting file size limit.
22 For example, to trap an interrupt signal, set up a handler like this.
23 Do as little as you possibly can in your handler; notice how all we do is
24 set a global variable and then raise an exception. That's because on most
25 systems, libraries are not re-entrant; particularly, memory allocation and
26 I/O routines are not. That means that doing nearly I<anything> in your
27 handler could in theory trigger a memory fault and subsequent core dump.
32 die "Somebody sent me a SIG$signame";
34 $SIG{INT} = 'catch_zap'; # could fail in modules
35 $SIG{INT} = \&catch_zap; # best strategy
37 The names of the signals are the ones listed out by C<kill -l> on your
38 system, or you can retrieve them from the Config module. Set up an
39 @signame list indexed by number to get the name and a %signo table
40 indexed by name to get the number:
43 defined $Config{sig_name} || die "No sigs?";
44 foreach $name (split(' ', $Config{sig_name})) {
50 So to check whether signal 17 and SIGALRM were the same, do just this:
52 print "signal #17 = $signame[17]\n";
54 print "SIGALRM is $signo{ALRM}\n";
57 You may also choose to assign the strings C<'IGNORE'> or C<'DEFAULT'> as
58 the handler, in which case Perl will try to discard the signal or do the
59 default thing. Some signals can be neither trapped nor ignored, such as
60 the KILL and STOP (but not the TSTP) signals. One strategy for
61 temporarily ignoring signals is to use a local() statement, which will be
62 automatically restored once your block is exited. (Remember that local()
63 values are "inherited" by functions called from within that block.)
66 local $SIG{INT} = 'IGNORE';
70 # interrupts still ignored, for now...
73 Sending a signal to a negative process ID means that you send the signal
74 to the entire Unix process-group. This code sends a hang-up signal to all
75 processes in the current process group (and sets $SIG{HUP} to IGNORE so
76 it doesn't kill itself):
79 local $SIG{HUP} = 'IGNORE';
81 # snazzy writing of: kill('HUP', -$$)
84 Another interesting signal to send is signal number zero. This doesn't
85 actually affect another process, but instead checks whether it's alive
86 or has changed its UID.
88 unless (kill 0 => $kid_pid) {
89 warn "something wicked happened to $kid_pid";
92 You might also want to employ anonymous functions for simple signal
95 $SIG{INT} = sub { die "\nOutta here!\n" };
97 But that will be problematic for the more complicated handlers that need
98 to reinstall themselves. Because Perl's signal mechanism is currently
99 based on the signal(3) function from the C library, you may sometimes be so
100 misfortunate as to run on systems where that function is "broken", that
101 is, it behaves in the old unreliable SysV way rather than the newer, more
102 reasonable BSD and POSIX fashion. So you'll see defensive people writing
103 signal handlers like this:
107 # loathe sysV: it makes us not only reinstate
108 # the handler, but place it after the wait
109 $SIG{CHLD} = \&REAPER;
111 $SIG{CHLD} = \&REAPER;
112 # now do something that forks...
114 or even the more elaborate:
116 use POSIX ":sys_wait_h";
119 while ($child = waitpid(-1,WNOHANG)) {
120 $Kid_Status{$child} = $?;
122 $SIG{CHLD} = \&REAPER; # still loathe sysV
124 $SIG{CHLD} = \&REAPER;
125 # do something that forks...
127 Signal handling is also used for timeouts in Unix, While safely
128 protected within an C<eval{}> block, you set a signal handler to trap
129 alarm signals and then schedule to have one delivered to you in some
130 number of seconds. Then try your blocking operation, clearing the alarm
131 when it's done but not before you've exited your C<eval{}> block. If it
132 goes off, you'll use die() to jump out of the block, much as you might
133 using longjmp() or throw() in other languages.
138 local $SIG{ALRM} = sub { die "alarm clock restart" };
140 flock(FH, 2); # blocking write lock
143 if ($@ and $@ !~ /alarm clock restart/) { die }
145 For more complex signal handling, you might see the standard POSIX
146 module. Lamentably, this is almost entirely undocumented, but
147 the F<t/lib/posix.t> file from the Perl source distribution has some
152 A named pipe (often referred to as a FIFO) is an old Unix IPC
153 mechanism for processes communicating on the same machine. It works
154 just like a regular, connected anonymous pipes, except that the
155 processes rendezvous using a filename and don't have to be related.
157 To create a named pipe, use the Unix command mknod(1) or on some
158 systems, mkfifo(1). These may not be in your normal path.
160 # system return val is backwards, so && not ||
162 $ENV{PATH} .= ":/etc:/usr/etc";
163 if ( system('mknod', $path, 'p')
164 && system('mkfifo', $path) )
166 die "mk{nod,fifo} $path failed";
170 A fifo is convenient when you want to connect a process to an unrelated
171 one. When you open a fifo, the program will block until there's something
174 For example, let's say you'd like to have your F<.signature> file be a
175 named pipe that has a Perl program on the other end. Now every time any
176 program (like a mailer, news reader, finger program, etc.) tries to read
177 from that file, the reading program will block and your program will
178 supply the new signature. We'll use the pipe-checking file test B<-p>
179 to find out whether anyone (or anything) has accidentally removed our fifo.
182 $FIFO = '.signature';
183 $ENV{PATH} .= ":/etc:/usr/games";
188 system('mknod', $FIFO, 'p')
189 && die "can't mknod $FIFO: $!";
192 # next line blocks until there's a reader
193 open (FIFO, "> $FIFO") || die "can't write $FIFO: $!";
194 print FIFO "John Smith (smith\@host.org)\n", `fortune -s`;
196 sleep 2; # to avoid dup signals
201 By installing Perl code to deal with signals, you're exposing yourself
202 to danger from two things. First, few system library functions are
203 re-entrant. If the signal interrupts while Perl is executing one function
204 (like malloc(3) or printf(3)), and your signal handler then calls the
205 same function again, you could get unpredictable behavior--often, a
206 core dump. Second, Perl isn't itself re-entrant at the lowest levels.
207 If the signal interrupts Perl while Perl is changing its own internal
208 data structures, similarly unpredictable behaviour may result.
210 There are two things you can do, knowing this: be paranoid or be
211 pragmatic. The paranoid approach is to do as little as possible in your
212 signal handler. Set an existing integer variable that already has a
213 value, and return. This doesn't help you if you're in a slow system call,
214 which will just restart. That means you have to C<die> to longjump(3) out
215 of the handler. Even this is a little cavalier for the true paranoiac,
216 who avoids C<die> in a handler because the system I<is> out to get you.
217 The pragmatic approach is to say ``I know the risks, but prefer the
218 convenience'', and to do anything you want in your signal handler,
219 prepared to clean up core dumps now and again.
221 To forbid signal handlers altogether would bars you from
222 many interesting programs, including virtually everything in this manpage,
223 since you could no longer even write SIGCHLD handlers. Their dodginess
224 is expected to be addresses in the 5.005 release.
227 =head1 Using open() for IPC
229 Perl's basic open() statement can also be used for unidirectional interprocess
230 communication by either appending or prepending a pipe symbol to the second
231 argument to open(). Here's how to start something up in a child process you
234 open(SPOOLER, "| cat -v | lpr -h 2>/dev/null")
235 || die "can't fork: $!";
236 local $SIG{PIPE} = sub { die "spooler pipe broke" };
237 print SPOOLER "stuff\n";
238 close SPOOLER || die "bad spool: $! $?";
240 And here's how to start up a child process you intend to read from:
242 open(STATUS, "netstat -an 2>&1 |")
243 || die "can't fork: $!";
245 next if /^(tcp|udp)/;
248 close STATUS || die "bad netstat: $! $?";
250 If one can be sure that a particular program is a Perl script that is
251 expecting filenames in @ARGV, the clever programmer can write something
254 % program f1 "cmd1|" - f2 "cmd2|" f3 < tmpfile
256 and irrespective of which shell it's called from, the Perl program will
257 read from the file F<f1>, the process F<cmd1>, standard input (F<tmpfile>
258 in this case), the F<f2> file, the F<cmd2> command, and finally the F<f3>
259 file. Pretty nifty, eh?
261 You might notice that you could use backticks for much the
262 same effect as opening a pipe for reading:
264 print grep { !/^(tcp|udp)/ } `netstat -an 2>&1`;
265 die "bad netstat" if $?;
267 While this is true on the surface, it's much more efficient to process the
268 file one line or record at a time because then you don't have to read the
269 whole thing into memory at once. It also gives you finer control of the
270 whole process, letting you to kill off the child process early if you'd
273 Be careful to check both the open() and the close() return values. If
274 you're I<writing> to a pipe, you should also trap SIGPIPE. Otherwise,
275 think of what happens when you start up a pipe to a command that doesn't
276 exist: the open() will in all likelihood succeed (it only reflects the
277 fork()'s success), but then your output will fail--spectacularly. Perl
278 can't know whether the command worked because your command is actually
279 running in a separate process whose exec() might have failed. Therefore,
280 while readers of bogus commands return just a quick end of file, writers
281 to bogus command will trigger a signal they'd better be prepared to
284 open(FH, "|bogus") or die "can't fork: $!";
285 print FH "bang\n" or die "can't write: $!";
286 close FH or die "can't close: $!";
288 That won't blow up until the close, and it will blow up with a SIGPIPE.
289 To catch it, you could use this:
291 $SIG{PIPE} = 'IGNORE';
292 open(FH, "|bogus") or die "can't fork: $!";
293 print FH "bang\n" or die "can't write: $!";
294 close FH or die "can't close: status=$?";
298 Both the main process and any child processes it forks share the same
299 STDIN, STDOUT, and STDERR filehandles. If both processes try to access
300 them at once, strange things can happen. You'll certainly want to any
301 stdio flush output buffers before forking. You may also want to close
302 or reopen the filehandles for the child. You can get around this by
303 opening your pipe with open(), but on some systems this means that the
304 child process cannot outlive the parent.
306 =head2 Background Processes
308 You can run a command in the background with:
312 The command's STDOUT and STDERR (and possibly STDIN, depending on your
313 shell) will be the same as the parent's. You won't need to catch
314 SIGCHLD because of the double-fork taking place (see below for more
317 =head2 Complete Dissociation of Child from Parent
319 In some cases (starting server processes, for instance) you'll want to
320 complete dissociate the child process from the parent. The easiest
323 use POSIX qw(setsid);
324 setsid() or die "Can't start a new session: $!";
326 However, you may not be on POSIX. The following process is reported
327 to work on most Unixish systems. Non-Unix users should check their
328 Your_OS::Process module for other solutions.
334 Open /dev/tty and use the TIOCNOTTY ioctl on it. See L<tty(4)>
339 Change directory to /
343 Reopen STDIN, STDOUT, and STDERR so they're not connected to the old
348 Background yourself like this:
354 Ignore hangup signals in case you're running on a shell that doesn't
355 automatically no-hup you:
357 $SIG{HUP} = 'IGNORE'; # or whatever you'd like
361 =head2 Safe Pipe Opens
363 Another interesting approach to IPC is making your single program go
364 multiprocess and communicate between (or even amongst) yourselves. The
365 open() function will accept a file argument of either C<"-|"> or C<"|-">
366 to do a very interesting thing: it forks a child connected to the
367 filehandle you've opened. The child is running the same program as the
368 parent. This is useful for safely opening a file when running under an
369 assumed UID or GID, for example. If you open a pipe I<to> minus, you can
370 write to the filehandle you opened and your kid will find it in his
371 STDIN. If you open a pipe I<from> minus, you can read from the filehandle
372 you opened whatever your kid writes to his STDOUT.
378 $pid = open(KID_TO_WRITE, "|-");
379 unless (defined $pid) {
380 warn "cannot fork: $!";
381 die "bailing out" if $sleep_count++ > 6;
384 } until defined $pid;
387 print KID_TO_WRITE @some_data;
388 close(KID_TO_WRITE) || warn "kid exited $?";
390 ($EUID, $EGID) = ($UID, $GID); # suid progs only
391 open (FILE, "> /safe/file")
392 || die "can't open /safe/file: $!";
394 print FILE; # child's STDIN is parent's KID
396 exit; # don't forget this
399 Another common use for this construct is when you need to execute
400 something without the shell's interference. With system(), it's
401 straightforward, but you can't use a pipe open or backticks safely.
402 That's because there's no way to stop the shell from getting its hands on
403 your arguments. Instead, use lower-level control to call exec() directly.
405 Here's a safe backtick or pipe open for read:
407 # add error processing as above
408 $pid = open(KID_TO_READ, "-|");
411 while (<KID_TO_READ>) {
412 # do something interesting
414 close(KID_TO_READ) || warn "kid exited $?";
417 ($EUID, $EGID) = ($UID, $GID); # suid only
418 exec($program, @options, @args)
419 || die "can't exec program: $!";
424 And here's a safe pipe open for writing:
426 # add error processing as above
427 $pid = open(KID_TO_WRITE, "|-");
428 $SIG{ALRM} = sub { die "whoops, $program pipe broke" };
434 close(KID_TO_WRITE) || warn "kid exited $?";
437 ($EUID, $EGID) = ($UID, $GID);
438 exec($program, @options, @args)
439 || die "can't exec program: $!";
443 Note that these operations are full Unix forks, which means they may not be
444 correctly implemented on alien systems. Additionally, these are not true
445 multithreading. If you'd like to learn more about threading, see the
446 F<modules> file mentioned below in the SEE ALSO section.
448 =head2 Bidirectional Communication with Another Process
450 While this works reasonably well for unidirectional communication, what
451 about bidirectional communication? The obvious thing you'd like to do
452 doesn't actually work:
454 open(PROG_FOR_READING_AND_WRITING, "| some program |")
456 and if you forget to use the B<-w> flag, then you'll miss out
457 entirely on the diagnostic message:
459 Can't do bidirectional pipe at -e line 1.
461 If you really want to, you can use the standard open2() library function
462 to catch both ends. There's also an open3() for tridirectional I/O so you
463 can also catch your child's STDERR, but doing so would then require an
464 awkward select() loop and wouldn't allow you to use normal Perl input
467 If you look at its source, you'll see that open2() uses low-level
468 primitives like Unix pipe() and exec() calls to create all the connections.
469 While it might have been slightly more efficient by using socketpair(), it
470 would have then been even less portable than it already is. The open2()
471 and open3() functions are unlikely to work anywhere except on a Unix
472 system or some other one purporting to be POSIX compliant.
474 Here's an example of using open2():
478 $pid = open2(*Reader, *Writer, "cat -u -n" );
479 Writer->autoflush(); # default here, actually
480 print Writer "stuff\n";
483 The problem with this is that Unix buffering is really going to
484 ruin your day. Even though your C<Writer> filehandle is auto-flushed,
485 and the process on the other end will get your data in a timely manner,
486 you can't usually do anything to force it to give it back to you
487 in a similarly quick fashion. In this case, we could, because we
488 gave I<cat> a B<-u> flag to make it unbuffered. But very few Unix
489 commands are designed to operate over pipes, so this seldom works
490 unless you yourself wrote the program on the other end of the
493 A solution to this is the nonstandard F<Comm.pl> library. It uses
494 pseudo-ttys to make your program behave more reasonably:
497 $ph = open_proc('cat -n');
499 print $ph "a line\n";
500 print "got back ", scalar <$ph>;
503 This way you don't have to have control over the source code of the
504 program you're using. The F<Comm> library also has expect()
505 and interact() functions. Find the library (and we hope its
506 successor F<IPC::Chat>) at your nearest CPAN archive as detailed
507 in the SEE ALSO section below.
509 =head2 Bidirectional Communication with Yourself
511 If you want, you may make low-level pipe() and fork()
512 to stitch this together by hand. This example only
513 talks to itself, but you could reopen the appropriate
514 handles to STDIN and STDOUT and call other processes.
517 # pipe1 - bidirectional communication using two pipe pairs
518 # designed for the socketpair-challenged
519 use IO::Handle; # thousands of lines just for autoflush :-(
520 pipe(PARENT_RDR, CHILD_WTR); # XXX: failure?
521 pipe(CHILD_RDR, PARENT_WTR); # XXX: failure?
522 CHILD_WTR->autoflush(1);
523 PARENT_WTR->autoflush(1);
526 close PARENT_RDR; close PARENT_WTR;
527 print CHILD_WTR "Parent Pid $$ is sending this\n";
528 chomp($line = <CHILD_RDR>);
529 print "Parent Pid $$ just read this: `$line'\n";
530 close CHILD_RDR; close CHILD_WTR;
533 die "cannot fork: $!" unless defined $pid;
534 close CHILD_RDR; close CHILD_WTR;
535 chomp($line = <PARENT_RDR>);
536 print "Child Pid $$ just read this: `$line'\n";
537 print PARENT_WTR "Child Pid $$ is sending this\n";
538 close PARENT_RDR; close PARENT_WTR;
542 But you don't actually have to make two pipe calls. If you
543 have the socketpair() system call, it will do this all for you.
546 # pipe2 - bidirectional communication using socketpair
547 # "the best ones always go both ways"
550 use IO::Handle; # thousands of lines just for autoflush :-(
551 # We say AF_UNIX because although *_LOCAL is the
552 # POSIX 1003.1g form of the constant, many machines
553 # still don't have it.
554 socketpair(CHILD, PARENT, AF_UNIX, SOCK_STREAM, PF_UNSPEC)
555 or die "socketpair: $!";
558 PARENT->autoflush(1);
562 print CHILD "Parent Pid $$ is sending this\n";
563 chomp($line = <CHILD>);
564 print "Parent Pid $$ just read this: `$line'\n";
568 die "cannot fork: $!" unless defined $pid;
570 chomp($line = <PARENT>);
571 print "Child Pid $$ just read this: `$line'\n";
572 print PARENT "Child Pid $$ is sending this\n";
577 =head1 Sockets: Client/Server Communication
579 While not limited to Unix-derived operating systems (e.g., WinSock on PCs
580 provides socket support, as do some VMS libraries), you may not have
581 sockets on your system, in which case this section probably isn't going to do
582 you much good. With sockets, you can do both virtual circuits (i.e., TCP
583 streams) and datagrams (i.e., UDP packets). You may be able to do even more
584 depending on your system.
586 The Perl function calls for dealing with sockets have the same names as
587 the corresponding system calls in C, but their arguments tend to differ
588 for two reasons: first, Perl filehandles work differently than C file
589 descriptors. Second, Perl already knows the length of its strings, so you
590 don't need to pass that information.
592 One of the major problems with old socket code in Perl was that it used
593 hard-coded values for some of the constants, which severely hurt
594 portability. If you ever see code that does anything like explicitly
595 setting C<$AF_INET = 2>, you know you're in for big trouble: An
596 immeasurably superior approach is to use the C<Socket> module, which more
597 reliably grants access to various constants and functions you'll need.
599 If you're not writing a server/client for an existing protocol like
600 NNTP or SMTP, you should give some thought to how your server will
601 know when the client has finished talking, and vice-versa. Most
602 protocols are based on one-line messages and responses (so one party
603 knows the other has finished when a "\n" is received) or multi-line
604 messages and responses that end with a period on an empty line
605 ("\n.\n" terminates a message/response).
607 =head2 Internet Line Terminators
609 The Internet line terminator is "\015\012". Under ASCII variants of
610 Unix, that could usually be written as "\r\n", but under other systems,
611 "\r\n" might at times be "\015\015\012", "\012\012\015", or something
612 completely different. The standards specify writing "\015\012" to be
613 conformant (be strict in what you provide), but they also recommend
614 accepting a lone "\012" on input (but be lenient in what you require).
615 We haven't always been very good about that in the code in this manpage,
616 but unless you're on a Mac, you'll probably be ok.
618 =head2 Internet TCP Clients and Servers
620 Use Internet-domain sockets when you want to do client-server
621 communication that might extend to machines outside of your own system.
623 Here's a sample TCP client using Internet-domain sockets:
628 my ($remote,$port, $iaddr, $paddr, $proto, $line);
630 $remote = shift || 'localhost';
631 $port = shift || 2345; # random port
632 if ($port =~ /\D/) { $port = getservbyname($port, 'tcp') }
633 die "No port" unless $port;
634 $iaddr = inet_aton($remote) || die "no host: $remote";
635 $paddr = sockaddr_in($port, $iaddr);
637 $proto = getprotobyname('tcp');
638 socket(SOCK, PF_INET, SOCK_STREAM, $proto) || die "socket: $!";
639 connect(SOCK, $paddr) || die "connect: $!";
640 while (defined($line = <SOCK>)) {
644 close (SOCK) || die "close: $!";
647 And here's a corresponding server to go along with it. We'll
648 leave the address as INADDR_ANY so that the kernel can choose
649 the appropriate interface on multihomed hosts. If you want sit
650 on a particular interface (like the external side of a gateway
651 or firewall machine), you should fill this in with your real address
656 BEGIN { $ENV{PATH} = '/usr/ucb:/bin' }
661 sub logmsg { print "$0 $$: @_ at ", scalar localtime, "\n" }
663 my $port = shift || 2345;
664 my $proto = getprotobyname('tcp');
665 $port = $1 if $port =~ /(\d+)/; # untaint port number
667 socket(Server, PF_INET, SOCK_STREAM, $proto) || die "socket: $!";
668 setsockopt(Server, SOL_SOCKET, SO_REUSEADDR,
669 pack("l", 1)) || die "setsockopt: $!";
670 bind(Server, sockaddr_in($port, INADDR_ANY)) || die "bind: $!";
671 listen(Server,SOMAXCONN) || die "listen: $!";
673 logmsg "server started on port $port";
677 $SIG{CHLD} = \&REAPER;
679 for ( ; $paddr = accept(Client,Server); close Client) {
680 my($port,$iaddr) = sockaddr_in($paddr);
681 my $name = gethostbyaddr($iaddr,AF_INET);
683 logmsg "connection from $name [",
684 inet_ntoa($iaddr), "]
687 print Client "Hello there, $name, it's now ",
688 scalar localtime, $EOL;
691 And here's a multithreaded version. It's multithreaded in that
692 like most typical servers, it spawns (forks) a slave server to
693 handle the client request so that the master server can quickly
694 go back to service a new client.
698 BEGIN { $ENV{PATH} = '/usr/ucb:/bin' }
703 sub spawn; # forward declaration
704 sub logmsg { print "$0 $$: @_ at ", scalar localtime, "\n" }
706 my $port = shift || 2345;
707 my $proto = getprotobyname('tcp');
708 $port = $1 if $port =~ /(\d+)/; # untaint port number
710 socket(Server, PF_INET, SOCK_STREAM, $proto) || die "socket: $!";
711 setsockopt(Server, SOL_SOCKET, SO_REUSEADDR,
712 pack("l", 1)) || die "setsockopt: $!";
713 bind(Server, sockaddr_in($port, INADDR_ANY)) || die "bind: $!";
714 listen(Server,SOMAXCONN) || die "listen: $!";
716 logmsg "server started on port $port";
723 $SIG{CHLD} = \&REAPER; # loathe sysV
724 logmsg "reaped $waitedpid" . ($? ? " with exit $?" : '');
727 $SIG{CHLD} = \&REAPER;
729 for ( $waitedpid = 0;
730 ($paddr = accept(Client,Server)) || $waitedpid;
731 $waitedpid = 0, close Client)
733 next if $waitedpid and not $paddr;
734 my($port,$iaddr) = sockaddr_in($paddr);
735 my $name = gethostbyaddr($iaddr,AF_INET);
737 logmsg "connection from $name [",
738 inet_ntoa($iaddr), "]
742 print "Hello there, $name, it's now ", scalar localtime, $EOL;
743 exec '/usr/games/fortune' # XXX: `wrong' line terminators
744 or confess "can't exec fortune: $!";
752 unless (@_ == 0 && $coderef && ref($coderef) eq 'CODE') {
753 confess "usage: spawn CODEREF";
757 if (!defined($pid = fork)) {
758 logmsg "cannot fork: $!";
762 return; # I'm the parent
764 # else I'm the child -- go spawn
766 open(STDIN, "<&Client") || die "can't dup client to stdin";
767 open(STDOUT, ">&Client") || die "can't dup client to stdout";
768 ## open(STDERR, ">&STDOUT") || die "can't dup stdout to stderr";
772 This server takes the trouble to clone off a child version via fork() for
773 each incoming request. That way it can handle many requests at once,
774 which you might not always want. Even if you don't fork(), the listen()
775 will allow that many pending connections. Forking servers have to be
776 particularly careful about cleaning up their dead children (called
777 "zombies" in Unix parlance), because otherwise you'll quickly fill up your
780 We suggest that you use the B<-T> flag to use taint checking (see L<perlsec>)
781 even if we aren't running setuid or setgid. This is always a good idea
782 for servers and other programs run on behalf of someone else (like CGI
783 scripts), because it lessens the chances that people from the outside will
784 be able to compromise your system.
786 Let's look at another TCP client. This one connects to the TCP "time"
787 service on a number of different machines and shows how far their clocks
788 differ from the system on which it's being run:
794 my $SECS_of_70_YEARS = 2208988800;
795 sub ctime { scalar localtime(shift) }
797 my $iaddr = gethostbyname('localhost');
798 my $proto = getprotobyname('tcp');
799 my $port = getservbyname('time', 'tcp');
800 my $paddr = sockaddr_in(0, $iaddr);
804 printf "%-24s %8s %s\n", "localhost", 0, ctime(time());
806 foreach $host (@ARGV) {
807 printf "%-24s ", $host;
808 my $hisiaddr = inet_aton($host) || die "unknown host";
809 my $hispaddr = sockaddr_in($port, $hisiaddr);
810 socket(SOCKET, PF_INET, SOCK_STREAM, $proto) || die "socket: $!";
811 connect(SOCKET, $hispaddr) || die "bind: $!";
813 read(SOCKET, $rtime, 4);
815 my $histime = unpack("N", $rtime) - $SECS_of_70_YEARS ;
816 printf "%8d %s\n", $histime - time, ctime($histime);
819 =head2 Unix-Domain TCP Clients and Servers
821 That's fine for Internet-domain clients and servers, but what about local
822 communications? While you can use the same setup, sometimes you don't
823 want to. Unix-domain sockets are local to the current host, and are often
824 used internally to implement pipes. Unlike Internet domain sockets, Unix
825 domain sockets can show up in the file system with an ls(1) listing.
828 srw-rw-rw- 1 root 0 Oct 31 07:23 /dev/log
830 You can test for these with Perl's B<-S> file test:
832 unless ( -S '/dev/log' ) {
833 die "something's wicked with the print system";
836 Here's a sample Unix-domain client:
841 my ($rendezvous, $line);
843 $rendezvous = shift || '/tmp/catsock';
844 socket(SOCK, PF_UNIX, SOCK_STREAM, 0) || die "socket: $!";
845 connect(SOCK, sockaddr_un($rendezvous)) || die "connect: $!";
846 while (defined($line = <SOCK>)) {
851 And here's a corresponding server. You don't have to worry about silly
852 network terminators here because Unix domain sockets are guaranteed
853 to be on the localhost, and thus everything works right.
860 BEGIN { $ENV{PATH} = '/usr/ucb:/bin' }
861 sub logmsg { print "$0 $$: @_ at ", scalar localtime, "\n" }
863 my $NAME = '/tmp/catsock';
864 my $uaddr = sockaddr_un($NAME);
865 my $proto = getprotobyname('tcp');
867 socket(Server,PF_UNIX,SOCK_STREAM,0) || die "socket: $!";
869 bind (Server, $uaddr) || die "bind: $!";
870 listen(Server,SOMAXCONN) || die "listen: $!";
872 logmsg "server started on $NAME";
878 $SIG{CHLD} = \&REAPER; # loathe sysV
879 logmsg "reaped $waitedpid" . ($? ? " with exit $?" : '');
882 $SIG{CHLD} = \&REAPER;
885 for ( $waitedpid = 0;
886 accept(Client,Server) || $waitedpid;
887 $waitedpid = 0, close Client)
890 logmsg "connection on $NAME";
892 print "Hello there, it's now ", scalar localtime, "\n";
893 exec '/usr/games/fortune' or die "can't exec fortune: $!";
897 As you see, it's remarkably similar to the Internet domain TCP server, so
898 much so, in fact, that we've omitted several duplicate functions--spawn(),
899 logmsg(), ctime(), and REAPER()--which are exactly the same as in the
902 So why would you ever want to use a Unix domain socket instead of a
903 simpler named pipe? Because a named pipe doesn't give you sessions. You
904 can't tell one process's data from another's. With socket programming,
905 you get a separate session for each client: that's why accept() takes two
908 For example, let's say that you have a long running database server daemon
909 that you want folks from the World Wide Web to be able to access, but only
910 if they go through a CGI interface. You'd have a small, simple CGI
911 program that does whatever checks and logging you feel like, and then acts
912 as a Unix-domain client and connects to your private server.
914 =head1 TCP Clients with IO::Socket
916 For those preferring a higher-level interface to socket programming, the
917 IO::Socket module provides an object-oriented approach. IO::Socket is
918 included as part of the standard Perl distribution as of the 5.004
919 release. If you're running an earlier version of Perl, just fetch
920 IO::Socket from CPAN, where you'll also find find modules providing easy
921 interfaces to the following systems: DNS, FTP, Ident (RFC 931), NIS and
922 NISPlus, NNTP, Ping, POP3, SMTP, SNMP, SSLeay, Telnet, and Time--just
925 =head2 A Simple Client
927 Here's a client that creates a TCP connection to the "daytime"
928 service at port 13 of the host name "localhost" and prints out everything
929 that the server there cares to provide.
933 $remote = IO::Socket::INET->new(
935 PeerAddr => "localhost",
936 PeerPort => "daytime(13)",
938 or die "cannot connect to daytime port at localhost";
939 while ( <$remote> ) { print }
941 When you run this program, you should get something back that
944 Wed May 14 08:40:46 MDT 1997
946 Here are what those parameters to the C<new> constructor mean:
952 This is which protocol to use. In this case, the socket handle returned
953 will be connected to a TCP socket, because we want a stream-oriented
954 connection, that is, one that acts pretty much like a plain old file.
955 Not all sockets are this of this type. For example, the UDP protocol
956 can be used to make a datagram socket, used for message-passing.
960 This is the name or Internet address of the remote host the server is
961 running on. We could have specified a longer name like C<"www.perl.com">,
962 or an address like C<"204.148.40.9">. For demonstration purposes, we've
963 used the special hostname C<"localhost">, which should always mean the
964 current machine you're running on. The corresponding Internet address
965 for localhost is C<"127.1">, if you'd rather use that.
969 This is the service name or port number we'd like to connect to.
970 We could have gotten away with using just C<"daytime"> on systems with a
971 well-configured system services file,[FOOTNOTE: The system services file
972 is in I</etc/services> under Unix] but just in case, we've specified the
973 port number (13) in parentheses. Using just the number would also have
974 worked, but constant numbers make careful programmers nervous.
978 Notice how the return value from the C<new> constructor is used as
979 a filehandle in the C<while> loop? That's what's called an indirect
980 filehandle, a scalar variable containing a filehandle. You can use
981 it the same way you would a normal filehandle. For example, you
982 can read one line from it this way:
986 all remaining lines from is this way:
990 and send a line of data to it this way:
992 print $handle "some data\n";
994 =head2 A Webget Client
996 Here's a simple client that takes a remote host to fetch a document
997 from, and then a list of documents to get from that host. This is a
998 more interesting client than the previous one because it first sends
999 something to the server before fetching the server's response.
1003 unless (@ARGV > 1) { die "usage: $0 host document ..." }
1004 $host = shift(@ARGV);
1007 foreach $document ( @ARGV ) {
1008 $remote = IO::Socket::INET->new( Proto => "tcp",
1010 PeerPort => "http(80)",
1012 unless ($remote) { die "cannot connect to http daemon on $host" }
1013 $remote->autoflush(1);
1014 print $remote "GET $document HTTP/1.0" . $BLANK;
1015 while ( <$remote> ) { print }
1019 The web server handing the "http" service, which is assumed to be at
1020 its standard port, number 80. If your the web server you're trying to
1021 connect to is at a different port (like 1080 or 8080), you should specify
1022 as the named-parameter pair, C<PeerPort =E<gt> 8080>. The C<autoflush>
1023 method is used on the socket because otherwise the system would buffer
1024 up the output we sent it. (If you're on a Mac, you'll also need to
1025 change every C<"\n"> in your code that sends data over the network to
1026 be a C<"\015\012"> instead.)
1028 Connecting to the server is only the first part of the process: once you
1029 have the connection, you have to use the server's language. Each server
1030 on the network has its own little command language that it expects as
1031 input. The string that we send to the server starting with "GET" is in
1032 HTTP syntax. In this case, we simply request each specified document.
1033 Yes, we really are making a new connection for each document, even though
1034 it's the same host. That's the way you always used to have to speak HTTP.
1035 Recent versions of web browsers may request that the remote server leave
1036 the connection open a little while, but the server doesn't have to honor
1039 Here's an example of running that program, which we'll call I<webget>:
1041 % webget www.perl.com /guanaco.html
1042 HTTP/1.1 404 File Not Found
1043 Date: Thu, 08 May 1997 18:02:32 GMT
1044 Server: Apache/1.2b6
1046 Content-type: text/html
1048 <HEAD><TITLE>404 File Not Found</TITLE></HEAD>
1049 <BODY><H1>File Not Found</H1>
1050 The requested URL /guanaco.html was not found on this server.<P>
1053 Ok, so that's not very interesting, because it didn't find that
1054 particular document. But a long response wouldn't have fit on this page.
1056 For a more fully-featured version of this program, you should look to
1057 the I<lwp-request> program included with the LWP modules from CPAN.
1059 =head2 Interactive Client with IO::Socket
1061 Well, that's all fine if you want to send one command and get one answer,
1062 but what about setting up something fully interactive, somewhat like
1063 the way I<telnet> works? That way you can type a line, get the answer,
1064 type a line, get the answer, etc.
1066 This client is more complicated than the two we've done so far, but if
1067 you're on a system that supports the powerful C<fork> call, the solution
1068 isn't that rough. Once you've made the connection to whatever service
1069 you'd like to chat with, call C<fork> to clone your process. Each of
1070 these two identical process has a very simple job to do: the parent
1071 copies everything from the socket to standard output, while the child
1072 simultaneously copies everything from standard input to the socket.
1073 To accomplish the same thing using just one process would be I<much>
1074 harder, because it's easier to code two processes to do one thing than it
1075 is to code one process to do two things. (This keep-it-simple principle
1076 a cornerstones of the Unix philosophy, and good software engineering as
1077 well, which is probably why it's spread to other systems.)
1084 my ($host, $port, $kidpid, $handle, $line);
1086 unless (@ARGV == 2) { die "usage: $0 host port" }
1087 ($host, $port) = @ARGV;
1089 # create a tcp connection to the specified host and port
1090 $handle = IO::Socket::INET->new(Proto => "tcp",
1093 or die "can't connect to port $port on $host: $!";
1095 $handle->autoflush(1); # so output gets there right away
1096 print STDERR "[Connected to $host:$port]\n";
1098 # split the program into two processes, identical twins
1099 die "can't fork: $!" unless defined($kidpid = fork());
1101 # the if{} block runs only in the parent process
1103 # copy the socket to standard output
1104 while (defined ($line = <$handle>)) {
1107 kill("TERM", $kidpid); # send SIGTERM to child
1109 # the else{} block runs only in the child process
1111 # copy standard input to the socket
1112 while (defined ($line = <STDIN>)) {
1113 print $handle $line;
1117 The C<kill> function in the parent's C<if> block is there to send a
1118 signal to our child process (current running in the C<else> block)
1119 as soon as the remote server has closed its end of the connection.
1121 If the remote server sends data a byte at time, and you need that
1122 data immediately without waiting for a newline (which might not happen),
1123 you may wish to replace the C<while> loop in the parent with the
1127 while (sysread($handle, $byte, 1) == 1) {
1131 Making a system call for each byte you want to read is not very efficient
1132 (to put it mildly) but is the simplest to explain and works reasonably
1135 =head1 TCP Servers with IO::Socket
1137 As always, setting up a server is little bit more involved than running a client.
1138 The model is that the server creates a special kind of socket that
1139 does nothing but listen on a particular port for incoming connections.
1140 It does this by calling the C<IO::Socket::INET-E<gt>new()> method with
1141 slightly different arguments than the client did.
1147 This is which protocol to use. Like our clients, we'll
1148 still specify C<"tcp"> here.
1153 port in the C<LocalPort> argument, which we didn't do for the client.
1154 This is service name or port number for which you want to be the
1155 server. (Under Unix, ports under 1024 are restricted to the
1156 superuser.) In our sample, we'll use port 9000, but you can use
1157 any port that's not currently in use on your system. If you try
1158 to use one already in used, you'll get an "Address already in use"
1159 message. Under Unix, the C<netstat -a> command will show
1160 which services current have servers.
1164 The C<Listen> parameter is set to the maximum number of
1165 pending connections we can accept until we turn away incoming clients.
1166 Think of it as a call-waiting queue for your telephone.
1167 The low-level Socket module has a special symbol for the system maximum, which
1172 The C<Reuse> parameter is needed so that we restart our server
1173 manually without waiting a few minutes to allow system buffers to
1178 Once the generic server socket has been created using the parameters
1179 listed above, the server then waits for a new client to connect
1180 to it. The server blocks in the C<accept> method, which eventually an
1181 bidirectional connection to the remote client. (Make sure to autoflush
1182 this handle to circumvent buffering.)
1184 To add to user-friendliness, our server prompts the user for commands.
1185 Most servers don't do this. Because of the prompt without a newline,
1186 you'll have to use the C<sysread> variant of the interactive client above.
1188 This server accepts one of five different commands, sending output
1189 back to the client. Note that unlike most network servers, this one
1190 only handles one incoming client at a time. Multithreaded servers are
1191 covered in Chapter 6 of the Camel as well as later in this manpage.
1193 Here's the code. We'll
1197 use Net::hostent; # for OO version of gethostbyaddr
1199 $PORT = 9000; # pick something not in use
1201 $server = IO::Socket::INET->new( Proto => 'tcp',
1203 Listen => SOMAXCONN,
1206 die "can't setup server" unless $server;
1207 print "[Server $0 accepting clients]\n";
1209 while ($client = $server->accept()) {
1210 $client->autoflush(1);
1211 print $client "Welcome to $0; type help for command list.\n";
1212 $hostinfo = gethostbyaddr($client->peeraddr);
1213 printf "[Connect from %s]\n", $hostinfo->name || $client->peerhost;
1214 print $client "Command? ";
1215 while ( <$client>) {
1216 next unless /\S/; # blank line
1217 if (/quit|exit/i) { last; }
1218 elsif (/date|time/i) { printf $client "%s\n", scalar localtime; }
1219 elsif (/who/i ) { print $client `who 2>&1`; }
1220 elsif (/cookie/i ) { print $client `/usr/games/fortune 2>&1`; }
1221 elsif (/motd/i ) { print $client `cat /etc/motd 2>&1`; }
1223 print $client "Commands: quit date who cookie motd\n";
1226 print $client "Command? ";
1231 =head1 UDP: Message Passing
1233 Another kind of client-server setup is one that uses not connections, but
1234 messages. UDP communications involve much lower overhead but also provide
1235 less reliability, as there are no promises that messages will arrive at
1236 all, let alone in order and unmangled. Still, UDP offers some advantages
1237 over TCP, including being able to "broadcast" or "multicast" to a whole
1238 bunch of destination hosts at once (usually on your local subnet). If you
1239 find yourself overly concerned about reliability and start building checks
1240 into your message system, then you probably should use just TCP to start
1243 Here's a UDP program similar to the sample Internet TCP client given
1244 earlier. However, instead of checking one host at a time, the UDP version
1245 will check many of them asynchronously by simulating a multicast and then
1246 using select() to do a timed-out wait for I/O. To do something similar
1247 with TCP, you'd have to use a different socket handle for each host.
1254 my ( $count, $hisiaddr, $hispaddr, $histime,
1255 $host, $iaddr, $paddr, $port, $proto,
1256 $rin, $rout, $rtime, $SECS_of_70_YEARS);
1258 $SECS_of_70_YEARS = 2208988800;
1260 $iaddr = gethostbyname(hostname());
1261 $proto = getprotobyname('udp');
1262 $port = getservbyname('time', 'udp');
1263 $paddr = sockaddr_in(0, $iaddr); # 0 means let kernel pick
1265 socket(SOCKET, PF_INET, SOCK_DGRAM, $proto) || die "socket: $!";
1266 bind(SOCKET, $paddr) || die "bind: $!";
1269 printf "%-12s %8s %s\n", "localhost", 0, scalar localtime time;
1273 $hisiaddr = inet_aton($host) || die "unknown host";
1274 $hispaddr = sockaddr_in($port, $hisiaddr);
1275 defined(send(SOCKET, 0, 0, $hispaddr)) || die "send $host: $!";
1279 vec($rin, fileno(SOCKET), 1) = 1;
1281 # timeout after 10.0 seconds
1282 while ($count && select($rout = $rin, undef, undef, 10.0)) {
1284 ($hispaddr = recv(SOCKET, $rtime, 4, 0)) || die "recv: $!";
1285 ($port, $hisiaddr) = sockaddr_in($hispaddr);
1286 $host = gethostbyaddr($hisiaddr, AF_INET);
1287 $histime = unpack("N", $rtime) - $SECS_of_70_YEARS ;
1288 printf "%-12s ", $host;
1289 printf "%8d %s\n", $histime - time, scalar localtime($histime);
1295 While System V IPC isn't so widely used as sockets, it still has some
1296 interesting uses. You can't, however, effectively use SysV IPC or
1297 Berkeley mmap() to have shared memory so as to share a variable amongst
1298 several processes. That's because Perl would reallocate your string when
1299 you weren't wanting it to.
1301 Here's a small example showing shared memory usage.
1306 $key = shmget($IPC_PRIVATE, $size , 0777 );
1307 die unless defined $key;
1309 $message = "Message #1";
1310 shmwrite($key, $message, 0, 60 ) || die "$!";
1311 shmread($key,$buff,0,60) || die "$!";
1315 print "deleting $key\n";
1316 shmctl($key ,$IPC_RMID, 0) || die "$!";
1318 Here's an example of a semaphore:
1322 $IPC_CREATE = 0001000;
1323 $key = semget($IPC_KEY, $nsems , 0666 | $IPC_CREATE );
1324 die if !defined($key);
1327 Put this code in a separate file to be run in more than one process.
1328 Call the file F<take>:
1330 # create a semaphore
1333 $key = semget($IPC_KEY, 0 , 0 );
1334 die if !defined($key);
1340 # wait for semaphore to be zero
1342 $opstring1 = pack("sss", $semnum, $semop, $semflag);
1344 # Increment the semaphore count
1346 $opstring2 = pack("sss", $semnum, $semop, $semflag);
1347 $opstring = $opstring1 . $opstring2;
1349 semop($key,$opstring) || die "$!";
1351 Put this code in a separate file to be run in more than one process.
1352 Call this file F<give>:
1354 # 'give' the semaphore
1355 # run this in the original process and you will see
1356 # that the second process continues
1359 $key = semget($IPC_KEY, 0, 0);
1360 die if !defined($key);
1365 # Decrement the semaphore count
1367 $opstring = pack("sss", $semnum, $semop, $semflag);
1369 semop($key,$opstring) || die "$!";
1371 The SysV IPC code above was written long ago, and it's definitely
1372 clunky looking. It should at the very least be made to C<use strict>
1373 and C<require "sys/ipc.ph">. Better yet, check out the IPC::SysV modules
1378 Most of these routines quietly but politely return C<undef> when they
1379 fail instead of causing your program to die right then and there due to
1380 an uncaught exception. (Actually, some of the new I<Socket> conversion
1381 functions croak() on bad arguments.) It is therefore essential to
1382 check return values from these functions. Always begin your socket
1383 programs this way for optimal success, and don't forget to add B<-T>
1384 taint checking flag to the #! line for servers:
1393 All these routines create system-specific portability problems. As noted
1394 elsewhere, Perl is at the mercy of your C libraries for much of its system
1395 behaviour. It's probably safest to assume broken SysV semantics for
1396 signals and to stick with simple TCP and UDP socket operations; e.g., don't
1397 try to pass open file descriptors over a local UDP datagram socket if you
1398 want your code to stand a chance of being portable.
1400 As mentioned in the signals section, because few vendors provide C
1401 libraries that are safely re-entrant, the prudent programmer will do
1402 little else within a handler beyond setting a numeric variable that
1403 already exists; or, if locked into a slow (restarting) system call,
1404 using die() to raise an exception and longjmp(3) out. In fact, even
1405 these may in some cases cause a core dump. It's probably best to avoid
1406 signals except where they are absolutely inevitable. This
1407 will be addressed in a future release of Perl.
1411 Tom Christiansen, with occasional vestiges of Larry Wall's original
1412 version and suggestions from the Perl Porters.
1416 There's a lot more to networking than this, but this should get you
1419 For intrepid programmers, the indispensable textbook is I<Unix Network
1420 Programming> by W. Richard Stevens (published by Addison-Wesley). Note
1421 that most books on networking address networking from the perspective of
1422 a C programmer; translation to Perl is left as an exercise for the reader.
1424 The IO::Socket(3) manpage describes the object library, and the Socket(3)
1425 manpage describes the low-level interface to sockets. Besides the obvious
1426 functions in L<perlfunc>, you should also check out the F<modules> file
1427 at your nearest CPAN site. (See L<perlmodlib> or best yet, the F<Perl
1428 FAQ> for a description of what CPAN is and where to get it.)
1430 Section 5 of the F<modules> file is devoted to "Networking, Device Control
1431 (modems), and Interprocess Communication", and contains numerous unbundled
1432 modules numerous networking modules, Chat and Expect operations, CGI
1433 programming, DCE, FTP, IPC, NNTP, Proxy, Ptty, RPC, SNMP, SMTP, Telnet,
1434 Threads, and ToolTalk--just to name a few.