1PERLFORK(1) Perl Programmers Reference Guide PERLFORK(1)
2
6 perlfork - Perl's fork() emulation
7
9 NOTE: As of the 5.8.0 release, fork() emulation has considerably
10 matured. However, there are still a few known bugs and differences
11 from real fork() that might affect you. See the "BUGS" and
12 "CAVEATS AND LIMITATIONS" sections below.
13 Perl provides a fork() keyword that corresponds to the Unix system call
15 of the same name. On most Unix-like platforms where the fork() system
16 call is available, Perl's fork() simply calls it.
17 On some platforms such as Windows where the fork() system call is not
19 available, Perl can be built to emulate fork() at the interpreter
20 level. While the emulation is designed to be as compatible as possible
21 with the real fork() at the level of the Perl program, there are cer‐
22 tain important differences that stem from the fact that all the pseudo
23 child "processes" created this way live in the same real process as far
24 as the operating system is concerned.
25 This document provides a general overview of the capabilities and limi‐
27 tations of the fork() emulation. Note that the issues discussed here
28 are not applicable to platforms where a real fork() is available and
29 Perl has been configured to use it.
30
32 The fork() emulation is implemented at the level of the Perl inter‐
33 preter. What this means in general is that running fork() will actu‐
34 ally clone the running interpreter and all its state, and run the
35 cloned interpreter in a separate thread, beginning execution in the new
36 thread just after the point where the fork() was called in the parent.
37 We will refer to the thread that implements this child "process" as the
38 pseudo-process.
39 To the Perl program that called fork(), all this is designed to be
41 transparent. The parent returns from the fork() with a pseudo-process
42 ID that can be subsequently used in any process manipulation functions;
43 the child returns from the fork() with a value of 0 to signify that it
44 is the child pseudo-process.
45 Behavior of other Perl features in forked pseudo-processes
47 Most Perl features behave in a natural way within pseudo-processes.
49 $$ or $PROCESS_ID
51 This special variable is correctly set to the pseudo-process
52 ID. It can be used to identify pseudo-processes within a par‐
53 ticular session. Note that this value is subject to recycling
54 if any pseudo-processes are launched after others have been
55 wait()-ed on.
56 %ENV Each pseudo-process maintains its own virtual environment.
58 Modifications to %ENV affect the virtual environment, and are
59 only visible within that pseudo-process, and in any processes
60 (or pseudo-processes) launched from it.
61 chdir() and all other builtins that accept filenames
63 Each pseudo-process maintains its own virtual idea of the cur‐
64 rent directory. Modifications to the current directory using
65 chdir() are only visible within that pseudo-process, and in any
66 processes (or pseudo-processes) launched from it. All file and
67 directory accesses from the pseudo-process will correctly map
68 the virtual working directory to the real working directory
69 appropriately.
70 wait() and waitpid()
72 wait() and waitpid() can be passed a pseudo-process ID returned
73 by fork(). These calls will properly wait for the termination
74 of the pseudo-process and return its status.
75 kill() kill() can be used to terminate a pseudo-process by passing it
77 the ID returned by fork(). This should not be used except
78 under dire circumstances, because the operating system may not
79 guarantee integrity of the process resources when a running
80 thread is terminated. Note that using kill() on a
81 pseudo-process() may typically cause memory leaks, because the
82 thread that implements the pseudo-process does not get a chance
83 to clean up its resources.
84 exec() Calling exec() within a pseudo-process actually spawns the
86 requested executable in a separate process and waits for it to
87 complete before exiting with the same exit status as that
88 process. This means that the process ID reported within the
89 running executable will be different from what the earlier Perl
90 fork() might have returned. Similarly, any process manipula‐
91 tion functions applied to the ID returned by fork() will affect
92 the waiting pseudo-process that called exec(), not the real
93 process it is waiting for after the exec().
94 exit() exit() always exits just the executing pseudo-process, after
96 automatically wait()-ing for any outstanding child pseudo-pro‐
97 cesses. Note that this means that the process as a whole will
98 not exit unless all running pseudo-processes have exited.
99 Open handles to files, directories and network sockets
101 All open handles are dup()-ed in pseudo-processes, so that
102 closing any handles in one process does not affect the others.
103 See below for some limitations.
104 Resource limits
106 In the eyes of the operating system, pseudo-processes created via the
108 fork() emulation are simply threads in the same process. This means
109 that any process-level limits imposed by the operating system apply to
110 all pseudo-processes taken together. This includes any limits imposed
111 by the operating system on the number of open file, directory and
112 socket handles, limits on disk space usage, limits on memory size, lim‐
113 its on CPU utilization etc.
114 Killing the parent process
116 If the parent process is killed (either using Perl's kill() builtin, or
118 using some external means) all the pseudo-processes are killed as well,
119 and the whole process exits.
120 Lifetime of the parent process and pseudo-processes
122 During the normal course of events, the parent process and every
124 pseudo-process started by it will wait for their respective pseudo-
125 children to complete before they exit. This means that the parent and
126 every pseudo-child created by it that is also a pseudo-parent will only
127 exit after their pseudo-children have exited.
128 A way to mark a pseudo-processes as running detached from their parent
130 (so that the parent would not have to wait() for them if it doesn't
131 want to) will be provided in future.
132 CAVEATS AND LIMITATIONS
134 BEGIN blocks
136 The fork() emulation will not work entirely correctly when
137 called from within a BEGIN block. The forked copy will run the
138 contents of the BEGIN block, but will not continue parsing the
139 source stream after the BEGIN block. For example, consider the
140 following code:
141 BEGIN {
143 fork and exit; # fork child and exit the parent
144 print "inner\n";
145 }
146 print "outer\n";
147 This will print:
149 inner
151 rather than the expected:
153 inner
155 outer
156 This limitation arises from fundamental technical difficulties
158 in cloning and restarting the stacks used by the Perl parser in
159 the middle of a parse.
160 Open filehandles
162 Any filehandles open at the time of the fork() will be
163 dup()-ed. Thus, the files can be closed independently in the
164 parent and child, but beware that the dup()-ed handles will
165 still share the same seek pointer. Changing the seek position
166 in the parent will change it in the child and vice-versa. One
167 can avoid this by opening files that need distinct seek point‐
168 ers separately in the child.
169 Forking pipe open() not yet implemented
171 The "open(FOO, "⎪-")" and "open(BAR, "-⎪")" constructs are not
172 yet implemented. This limitation can be easily worked around
173 in new code by creating a pipe explicitly. The following exam‐
174 ple shows how to write to a forked child:
175 # simulate open(FOO, "⎪-")
177 sub pipe_to_fork ($) {
178 my $parent = shift;
179 pipe my $child, $parent or die;
180 my $pid = fork();
181 die "fork() failed: $!" unless defined $pid;
182 if ($pid) {
183 close $child;
184 }
185 else {
186 close $parent;
187 open(STDIN, "<&=" . fileno($child)) or die;
188 }
189 $pid;
190 }
191 if (pipe_to_fork('FOO')) {
193 # parent
194 print FOO "pipe_to_fork\n";
195 close FOO;
196 }
197 else {
198 # child
199 while (<STDIN>) { print; }
200 exit(0);
201 }
202 And this one reads from the child:
204 # simulate open(FOO, "-⎪")
206 sub pipe_from_fork ($) {
207 my $parent = shift;
208 pipe $parent, my $child or die;
209 my $pid = fork();
210 die "fork() failed: $!" unless defined $pid;
211 if ($pid) {
212 close $child;
213 }
214 else {
215 close $parent;
216 open(STDOUT, ">&=" . fileno($child)) or die;
217 }
218 $pid;
219 }
220 if (pipe_from_fork('BAR')) {
222 # parent
223 while (<BAR>) { print; }
224 close BAR;
225 }
226 else {
227 # child
228 print "pipe_from_fork\n";
229 exit(0);
230 }
231 Forking pipe open() constructs will be supported in future.
233 Global state maintained by XSUBs
235 External subroutines (XSUBs) that maintain their own global
236 state may not work correctly. Such XSUBs will either need to
237 maintain locks to protect simultaneous access to global data
238 from different pseudo-processes, or maintain all their state on
239 the Perl symbol table, which is copied naturally when fork() is
240 called. A callback mechanism that provides extensions an
241 opportunity to clone their state will be provided in the near
242 future.
243 Interpreter embedded in larger application
245 The fork() emulation may not behave as expected when it is exe‐
246 cuted in an application which embeds a Perl interpreter and
247 calls Perl APIs that can evaluate bits of Perl code. This
248 stems from the fact that the emulation only has knowledge about
249 the Perl interpreter's own data structures and knows nothing
250 about the containing application's state. For example, any
251 state carried on the application's own call stack is out of
252 reach.
253 Thread-safety of extensions
255 Since the fork() emulation runs code in multiple threads,
256 extensions calling into non-thread-safe libraries may not work
257 reliably when calling fork(). As Perl's threading support
258 gradually becomes more widely adopted even on platforms with a
259 native fork(), such extensions are expected to be fixed for
260 thread-safety.
261
263 · Having pseudo-process IDs be negative integers breaks down for
264 the integer "-1" because the wait() and waitpid() functions
265 treat this number as being special. The tacit assumption in
266 the current implementation is that the system never allocates a
267 thread ID of 1 for user threads. A better representation for
268 pseudo-process IDs will be implemented in future.
269 · In certain cases, the OS-level handles created by the pipe(),
271 socket(), and accept() operators are apparently not duplicated
272 accurately in pseudo-processes. This only happens in some sit‐
273 uations, but where it does happen, it may result in deadlocks
274 between the read and write ends of pipe handles, or inability
275 to send or receive data across socket handles.
276 · This document may be incomplete in some respects.
278
280 Support for concurrent interpreters and the fork() emulation was imple‐
281 mented by ActiveState, with funding from Microsoft Corporation.
282 This document is authored and maintained by Gurusamy Sarathy
284 <gsar@activestate.com>.
285
287 "fork" in perlfunc, perlipc
288perl v5.8.8 2006-01-07 PERLFORK(1)