1 |
root |
1.1 |
=head1 NAME |
2 |
|
|
|
3 |
root |
1.4 |
AnyEvent::Fork - everything you wanted to use fork() for, but couldn't |
4 |
root |
1.1 |
|
5 |
|
|
=head1 SYNOPSIS |
6 |
|
|
|
7 |
root |
1.4 |
use AnyEvent::Fork; |
8 |
root |
1.1 |
|
9 |
root |
1.24 |
AnyEvent::Fork |
10 |
|
|
->new |
11 |
|
|
->require ("MyModule") |
12 |
|
|
->run ("MyModule::server", my $cv = AE::cv); |
13 |
|
|
|
14 |
|
|
my $fh = $cv->recv; |
15 |
|
|
|
16 |
|
|
=head1 DESCRIPTION |
17 |
|
|
|
18 |
|
|
This module allows you to create new processes, without actually forking |
19 |
|
|
them from your current process (avoiding the problems of forking), but |
20 |
|
|
preserving most of the advantages of fork. |
21 |
|
|
|
22 |
|
|
It can be used to create new worker processes or new independent |
23 |
|
|
subprocesses for short- and long-running jobs, process pools (e.g. for use |
24 |
|
|
in pre-forked servers) but also to spawn new external processes (such as |
25 |
|
|
CGI scripts from a web server), which can be faster (and more well behaved) |
26 |
|
|
than using fork+exec in big processes. |
27 |
|
|
|
28 |
|
|
Special care has been taken to make this module useful from other modules, |
29 |
|
|
while still supporting specialised environments such as L<App::Staticperl> |
30 |
|
|
or L<PAR::Packer>. |
31 |
|
|
|
32 |
root |
1.39 |
=head2 WHAT THIS MODULE IS NOT |
33 |
root |
1.24 |
|
34 |
|
|
This module only creates processes and lets you pass file handles and |
35 |
|
|
strings to it, and run perl code. It does not implement any kind of RPC - |
36 |
|
|
there is no back channel from the process back to you, and there is no RPC |
37 |
|
|
or message passing going on. |
38 |
|
|
|
39 |
root |
1.43 |
If you need some form of RPC, you could use the L<AnyEvent::Fork::RPC> |
40 |
|
|
companion module, which adds simple RPC/job queueing to a process created |
41 |
|
|
by this module. |
42 |
|
|
|
43 |
root |
1.51 |
And if you need some automatic process pool management on top of |
44 |
|
|
L<AnyEvent::Fork::RPC>, you can look at the L<AnyEvent::Fork::Pool> |
45 |
|
|
companion module. |
46 |
|
|
|
47 |
|
|
Or you can implement it yourself in whatever way you like: use some |
48 |
root |
1.43 |
message-passing module such as L<AnyEvent::MP>, some pipe such as |
49 |
|
|
L<AnyEvent::ZeroMQ>, use L<AnyEvent::Handle> on both sides to send |
50 |
|
|
e.g. JSON or Storable messages, and so on. |
51 |
root |
1.24 |
|
52 |
root |
1.39 |
=head2 COMPARISON TO OTHER MODULES |
53 |
|
|
|
54 |
|
|
There is an abundance of modules on CPAN that do "something fork", such as |
55 |
|
|
L<Parallel::ForkManager>, L<AnyEvent::ForkManager>, L<AnyEvent::Worker> |
56 |
|
|
or L<AnyEvent::Subprocess>. There are modules that implement their own |
57 |
|
|
process management, such as L<AnyEvent::DBI>. |
58 |
|
|
|
59 |
|
|
The problems that all these modules try to solve are real, however, none |
60 |
|
|
of them (from what I have seen) tackle the very real problems of unwanted |
61 |
root |
1.63 |
memory sharing, efficiency or not being able to use event processing, GUI |
62 |
|
|
toolkits or similar modules in the processes they create. |
63 |
root |
1.39 |
|
64 |
|
|
This module doesn't try to replace any of them - instead it tries to solve |
65 |
|
|
the problem of creating processes with a minimum of fuss and overhead (and |
66 |
|
|
also luxury). Ideally, most of these would use AnyEvent::Fork internally, |
67 |
|
|
except they were written before AnyEvent:Fork was available, so obviously |
68 |
|
|
had to roll their own. |
69 |
|
|
|
70 |
|
|
=head2 PROBLEM STATEMENT |
71 |
root |
1.25 |
|
72 |
|
|
There are two traditional ways to implement parallel processing on UNIX |
73 |
|
|
like operating systems - fork and process, and fork+exec and process. They |
74 |
|
|
have different advantages and disadvantages that I describe below, |
75 |
|
|
together with how this module tries to mitigate the disadvantages. |
76 |
|
|
|
77 |
|
|
=over 4 |
78 |
|
|
|
79 |
|
|
=item Forking from a big process can be very slow. |
80 |
|
|
|
81 |
|
|
A 5GB process needs 0.05s to fork on my 3.6GHz amd64 GNU/Linux box. This |
82 |
|
|
overhead is often shared with exec (because you have to fork first), but |
83 |
|
|
in some circumstances (e.g. when vfork is used), fork+exec can be much |
84 |
|
|
faster. |
85 |
|
|
|
86 |
|
|
This module can help here by telling a small(er) helper process to fork, |
87 |
|
|
which is faster then forking the main process, and also uses vfork where |
88 |
|
|
possible. This gives the speed of vfork, with the flexibility of fork. |
89 |
|
|
|
90 |
|
|
=item Forking usually creates a copy-on-write copy of the parent |
91 |
|
|
process. |
92 |
|
|
|
93 |
|
|
For example, modules or data files that are loaded will not use additional |
94 |
root |
1.63 |
memory after a fork. Exec'ing a new process, in contrast, means modules |
95 |
|
|
and data files might need to be loaded again, at extra CPU and memory |
96 |
|
|
cost. |
97 |
|
|
|
98 |
|
|
But when forking, you still create a copy of your data structures - if |
99 |
|
|
the program frees them and replaces them by new data, the child processes |
100 |
|
|
will retain the old version even if it isn't used, which can suddenly and |
101 |
|
|
unexpectedly increase memory usage when freeing memory. |
102 |
|
|
|
103 |
|
|
For example, L<Gtk2::CV> is an image viewer optimised for large |
104 |
|
|
directories (millions of pictures). It also forks subprocesses for |
105 |
|
|
thumbnail generation, which inherit the data structure that stores all |
106 |
|
|
file information. If the user changes the directory, it gets freed in |
107 |
|
|
the main process, leaving a copy in the thumbnailer processes. This can |
108 |
|
|
lead to many times the memory usage that would actually be required. The |
109 |
|
|
solution is to fork early (and being unable to dynamically generate more |
110 |
|
|
subprocesses or do this from a module)... or to use L<AnyEvent:Fork>. |
111 |
root |
1.25 |
|
112 |
root |
1.63 |
There is a trade-off between more sharing with fork (which can be good or |
113 |
root |
1.25 |
bad), and no sharing with exec. |
114 |
|
|
|
115 |
|
|
This module allows the main program to do a controlled fork, and allows |
116 |
|
|
modules to exec processes safely at any time. When creating a custom |
117 |
|
|
process pool you can take advantage of data sharing via fork without |
118 |
|
|
risking to share large dynamic data structures that will blow up child |
119 |
|
|
memory usage. |
120 |
|
|
|
121 |
|
|
In other words, this module puts you into control over what is being |
122 |
|
|
shared and what isn't, at all times. |
123 |
|
|
|
124 |
|
|
=item Exec'ing a new perl process might be difficult. |
125 |
|
|
|
126 |
|
|
For example, it is not easy to find the correct path to the perl |
127 |
root |
1.63 |
interpreter - C<$^X> might not be a perl interpreter at all. Worse, there |
128 |
|
|
might not even be a perl binary installed on the system. |
129 |
root |
1.25 |
|
130 |
|
|
This module tries hard to identify the correct path to the perl |
131 |
|
|
interpreter. With a cooperative main program, exec'ing the interpreter |
132 |
|
|
might not even be necessary, but even without help from the main program, |
133 |
|
|
it will still work when used from a module. |
134 |
|
|
|
135 |
|
|
=item Exec'ing a new perl process might be slow, as all necessary modules |
136 |
|
|
have to be loaded from disk again, with no guarantees of success. |
137 |
|
|
|
138 |
|
|
Long running processes might run into problems when perl is upgraded |
139 |
|
|
and modules are no longer loadable because they refer to a different |
140 |
|
|
perl version, or parts of a distribution are newer than the ones already |
141 |
|
|
loaded. |
142 |
|
|
|
143 |
|
|
This module supports creating pre-initialised perl processes to be used as |
144 |
root |
1.63 |
a template for new processes at a later time, e.g. for use in a process |
145 |
|
|
pool. |
146 |
root |
1.25 |
|
147 |
|
|
=item Forking might be impossible when a program is running. |
148 |
|
|
|
149 |
|
|
For example, POSIX makes it almost impossible to fork from a |
150 |
|
|
multi-threaded program while doing anything useful in the child - in |
151 |
|
|
fact, if your perl program uses POSIX threads (even indirectly via |
152 |
|
|
e.g. L<IO::AIO> or L<threads>), you cannot call fork on the perl level |
153 |
root |
1.63 |
anymore without risking memory corruption or worse on a number of |
154 |
|
|
operating systems. |
155 |
root |
1.25 |
|
156 |
|
|
This module can safely fork helper processes at any time, by calling |
157 |
|
|
fork+exec in C, in a POSIX-compatible way (via L<Proc::FastSpawn>). |
158 |
|
|
|
159 |
|
|
=item Parallel processing with fork might be inconvenient or difficult |
160 |
|
|
to implement. Modules might not work in both parent and child. |
161 |
|
|
|
162 |
|
|
For example, when a program uses an event loop and creates watchers it |
163 |
|
|
becomes very hard to use the event loop from a child program, as the |
164 |
|
|
watchers already exist but are only meaningful in the parent. Worse, a |
165 |
|
|
module might want to use such a module, not knowing whether another module |
166 |
|
|
or the main program also does, leading to problems. |
167 |
|
|
|
168 |
root |
1.26 |
Apart from event loops, graphical toolkits also commonly fall into the |
169 |
|
|
"unsafe module" category, or just about anything that communicates with |
170 |
|
|
the external world, such as network libraries and file I/O modules, which |
171 |
|
|
usually don't like being copied and then allowed to continue in two |
172 |
|
|
processes. |
173 |
|
|
|
174 |
root |
1.25 |
With this module only the main program is allowed to create new processes |
175 |
|
|
by forking (because only the main program can know when it is still safe |
176 |
|
|
to do so) - all other processes are created via fork+exec, which makes it |
177 |
|
|
possible to use modules such as event loops or window interfaces safely. |
178 |
|
|
|
179 |
|
|
=back |
180 |
|
|
|
181 |
root |
1.24 |
=head1 EXAMPLES |
182 |
|
|
|
183 |
root |
1.64 |
This is where the wall of text ends and code speaks. |
184 |
|
|
|
185 |
root |
1.24 |
=head2 Create a single new process, tell it to run your worker function. |
186 |
root |
1.9 |
|
187 |
|
|
AnyEvent::Fork |
188 |
|
|
->new |
189 |
|
|
->require ("MyModule") |
190 |
|
|
->run ("MyModule::worker, sub { |
191 |
|
|
my ($master_filehandle) = @_; |
192 |
|
|
|
193 |
|
|
# now $master_filehandle is connected to the |
194 |
|
|
# $slave_filehandle in the new process. |
195 |
|
|
}); |
196 |
|
|
|
197 |
root |
1.32 |
C<MyModule> might look like this: |
198 |
root |
1.30 |
|
199 |
root |
1.31 |
package MyModule; |
200 |
|
|
|
201 |
|
|
sub worker { |
202 |
root |
1.9 |
my ($slave_filehandle) = @_; |
203 |
|
|
|
204 |
|
|
# now $slave_filehandle is connected to the $master_filehandle |
205 |
root |
1.65 |
# in the original process. have fun! |
206 |
root |
1.9 |
} |
207 |
|
|
|
208 |
root |
1.24 |
=head2 Create a pool of server processes all accepting on the same socket. |
209 |
root |
1.9 |
|
210 |
|
|
# create listener socket |
211 |
|
|
my $listener = ...; |
212 |
|
|
|
213 |
|
|
# create a pool template, initialise it and give it the socket |
214 |
|
|
my $pool = AnyEvent::Fork |
215 |
|
|
->new |
216 |
|
|
->require ("Some::Stuff", "My::Server") |
217 |
|
|
->send_fh ($listener); |
218 |
|
|
|
219 |
|
|
# now create 10 identical workers |
220 |
|
|
for my $id (1..10) { |
221 |
|
|
$pool |
222 |
|
|
->fork |
223 |
|
|
->send_arg ($id) |
224 |
|
|
->run ("My::Server::run"); |
225 |
|
|
} |
226 |
|
|
|
227 |
|
|
# now do other things - maybe use the filehandle provided by run |
228 |
|
|
# to wait for the processes to die. or whatever. |
229 |
|
|
|
230 |
root |
1.32 |
C<My::Server> might look like this: |
231 |
root |
1.31 |
|
232 |
|
|
package My::Server; |
233 |
root |
1.30 |
|
234 |
root |
1.31 |
sub run { |
235 |
root |
1.9 |
my ($slave, $listener, $id) = @_; |
236 |
|
|
|
237 |
|
|
close $slave; # we do not use the socket, so close it to save resources |
238 |
|
|
|
239 |
|
|
# we could go ballistic and use e.g. AnyEvent here, or IO::AIO, |
240 |
|
|
# or anything we usually couldn't do in a process forked normally. |
241 |
|
|
while (my $socket = $listener->accept) { |
242 |
|
|
# do sth. with new socket |
243 |
|
|
} |
244 |
|
|
} |
245 |
|
|
|
246 |
root |
1.24 |
=head2 use AnyEvent::Fork as a faster fork+exec |
247 |
root |
1.23 |
|
248 |
root |
1.44 |
This runs C</bin/echo hi>, with standard output redirected to F</tmp/log> |
249 |
root |
1.32 |
and standard error redirected to the communications socket. It is usually |
250 |
|
|
faster than fork+exec, but still lets you prepare the environment. |
251 |
root |
1.23 |
|
252 |
|
|
open my $output, ">/tmp/log" or die "$!"; |
253 |
|
|
|
254 |
|
|
AnyEvent::Fork |
255 |
|
|
->new |
256 |
|
|
->eval (' |
257 |
root |
1.40 |
# compile a helper function for later use |
258 |
root |
1.23 |
sub run { |
259 |
|
|
my ($fh, $output, @cmd) = @_; |
260 |
|
|
|
261 |
|
|
# perl will clear close-on-exec on STDOUT/STDERR |
262 |
|
|
open STDOUT, ">&", $output or die; |
263 |
|
|
open STDERR, ">&", $fh or die; |
264 |
|
|
|
265 |
|
|
exec @cmd; |
266 |
|
|
} |
267 |
|
|
') |
268 |
|
|
->send_fh ($output) |
269 |
|
|
->send_arg ("/bin/echo", "hi") |
270 |
|
|
->run ("run", my $cv = AE::cv); |
271 |
|
|
|
272 |
|
|
my $stderr = $cv->recv; |
273 |
|
|
|
274 |
root |
1.51 |
=head2 For stingy users: put the worker code into a C<DATA> section. |
275 |
|
|
|
276 |
root |
1.61 |
When you want to be stingy with files, you can put your code into the |
277 |
root |
1.51 |
C<DATA> section of your module (or program): |
278 |
|
|
|
279 |
|
|
use AnyEvent::Fork; |
280 |
|
|
|
281 |
|
|
AnyEvent::Fork |
282 |
|
|
->new |
283 |
|
|
->eval (do { local $/; <DATA> }) |
284 |
|
|
->run ("doit", sub { ... }); |
285 |
|
|
|
286 |
|
|
__DATA__ |
287 |
|
|
|
288 |
|
|
sub doit { |
289 |
|
|
... do something! |
290 |
|
|
} |
291 |
|
|
|
292 |
|
|
=head2 For stingy standalone programs: do not rely on external files at |
293 |
|
|
all. |
294 |
|
|
|
295 |
|
|
For single-file scripts it can be inconvenient to rely on external |
296 |
root |
1.61 |
files - even when using a C<DATA> section, you still need to C<exec> an |
297 |
|
|
external perl interpreter, which might not be available when using |
298 |
root |
1.51 |
L<App::Staticperl>, L<Urlader> or L<PAR::Packer> for example. |
299 |
|
|
|
300 |
|
|
Two modules help here - L<AnyEvent::Fork::Early> forks a template process |
301 |
|
|
for all further calls to C<new_exec>, and L<AnyEvent::Fork::Template> |
302 |
|
|
forks the main program as a template process. |
303 |
|
|
|
304 |
|
|
Here is how your main program should look like: |
305 |
|
|
|
306 |
|
|
#! perl |
307 |
|
|
|
308 |
|
|
# optional, as the very first thing. |
309 |
|
|
# in case modules want to create their own processes. |
310 |
|
|
use AnyEvent::Fork::Early; |
311 |
|
|
|
312 |
|
|
# next, load all modules you need in your template process |
313 |
|
|
use Example::My::Module |
314 |
|
|
use Example::Whatever; |
315 |
|
|
|
316 |
|
|
# next, put your run function definition and anything else you |
317 |
|
|
# need, but do not use code outside of BEGIN blocks. |
318 |
|
|
sub worker_run { |
319 |
|
|
my ($fh, @args) = @_; |
320 |
|
|
... |
321 |
|
|
} |
322 |
|
|
|
323 |
|
|
# now preserve everything so far as AnyEvent::Fork object |
324 |
root |
1.62 |
# in $TEMPLATE. |
325 |
root |
1.51 |
use AnyEvent::Fork::Template; |
326 |
|
|
|
327 |
|
|
# do not put code outside of BEGIN blocks until here |
328 |
|
|
|
329 |
|
|
# now use the $TEMPLATE process in any way you like |
330 |
|
|
|
331 |
|
|
# for example: create 10 worker processes |
332 |
|
|
my @worker; |
333 |
|
|
my $cv = AE::cv; |
334 |
|
|
for (1..10) { |
335 |
|
|
$cv->begin; |
336 |
|
|
$TEMPLATE->fork->send_arg ($_)->run ("worker_run", sub { |
337 |
|
|
push @worker, shift; |
338 |
|
|
$cv->end; |
339 |
|
|
}); |
340 |
|
|
} |
341 |
|
|
$cv->recv; |
342 |
|
|
|
343 |
root |
1.52 |
=head1 CONCEPTS |
344 |
root |
1.3 |
|
345 |
|
|
This module can create new processes either by executing a new perl |
346 |
|
|
process, or by forking from an existing "template" process. |
347 |
|
|
|
348 |
root |
1.45 |
All these processes are called "child processes" (whether they are direct |
349 |
|
|
children or not), while the process that manages them is called the |
350 |
|
|
"parent process". |
351 |
|
|
|
352 |
root |
1.3 |
Each such process comes with its own file handle that can be used to |
353 |
|
|
communicate with it (it's actually a socket - one end in the new process, |
354 |
|
|
one end in the main process), and among the things you can do in it are |
355 |
|
|
load modules, fork new processes, send file handles to it, and execute |
356 |
|
|
functions. |
357 |
|
|
|
358 |
|
|
There are multiple ways to create additional processes to execute some |
359 |
|
|
jobs: |
360 |
|
|
|
361 |
|
|
=over 4 |
362 |
|
|
|
363 |
|
|
=item fork a new process from the "default" template process, load code, |
364 |
|
|
run it |
365 |
|
|
|
366 |
|
|
This module has a "default" template process which it executes when it is |
367 |
|
|
needed the first time. Forking from this process shares the memory used |
368 |
|
|
for the perl interpreter with the new process, but loading modules takes |
369 |
|
|
time, and the memory is not shared with anything else. |
370 |
|
|
|
371 |
|
|
This is ideal for when you only need one extra process of a kind, with the |
372 |
root |
1.17 |
option of starting and stopping it on demand. |
373 |
root |
1.3 |
|
374 |
root |
1.9 |
Example: |
375 |
|
|
|
376 |
|
|
AnyEvent::Fork |
377 |
|
|
->new |
378 |
|
|
->require ("Some::Module") |
379 |
|
|
->run ("Some::Module::run", sub { |
380 |
|
|
my ($fork_fh) = @_; |
381 |
|
|
}); |
382 |
|
|
|
383 |
root |
1.3 |
=item fork a new template process, load code, then fork processes off of |
384 |
|
|
it and run the code |
385 |
|
|
|
386 |
|
|
When you need to have a bunch of processes that all execute the same (or |
387 |
|
|
very similar) tasks, then a good way is to create a new template process |
388 |
|
|
for them, loading all the modules you need, and then create your worker |
389 |
|
|
processes from this new template process. |
390 |
|
|
|
391 |
|
|
This way, all code (and data structures) that can be shared (e.g. the |
392 |
|
|
modules you loaded) is shared between the processes, and each new process |
393 |
|
|
consumes relatively little memory of its own. |
394 |
|
|
|
395 |
|
|
The disadvantage of this approach is that you need to create a template |
396 |
|
|
process for the sole purpose of forking new processes from it, but if you |
397 |
root |
1.17 |
only need a fixed number of processes you can create them, and then destroy |
398 |
root |
1.3 |
the template process. |
399 |
|
|
|
400 |
root |
1.9 |
Example: |
401 |
|
|
|
402 |
|
|
my $template = AnyEvent::Fork->new->require ("Some::Module"); |
403 |
|
|
|
404 |
|
|
for (1..10) { |
405 |
|
|
$template->fork->run ("Some::Module::run", sub { |
406 |
|
|
my ($fork_fh) = @_; |
407 |
|
|
}); |
408 |
|
|
} |
409 |
|
|
|
410 |
|
|
# at this point, you can keep $template around to fork new processes |
411 |
|
|
# later, or you can destroy it, which causes it to vanish. |
412 |
|
|
|
413 |
root |
1.3 |
=item execute a new perl interpreter, load some code, run it |
414 |
|
|
|
415 |
|
|
This is relatively slow, and doesn't allow you to share memory between |
416 |
|
|
multiple processes. |
417 |
|
|
|
418 |
|
|
The only advantage is that you don't have to have a template process |
419 |
|
|
hanging around all the time to fork off some new processes, which might be |
420 |
|
|
an advantage when there are long time spans where no extra processes are |
421 |
|
|
needed. |
422 |
|
|
|
423 |
root |
1.9 |
Example: |
424 |
|
|
|
425 |
|
|
AnyEvent::Fork |
426 |
|
|
->new_exec |
427 |
|
|
->require ("Some::Module") |
428 |
|
|
->run ("Some::Module::run", sub { |
429 |
|
|
my ($fork_fh) = @_; |
430 |
|
|
}); |
431 |
|
|
|
432 |
root |
1.3 |
=back |
433 |
|
|
|
434 |
root |
1.27 |
=head1 THE C<AnyEvent::Fork> CLASS |
435 |
|
|
|
436 |
|
|
This module exports nothing, and only implements a single class - |
437 |
|
|
C<AnyEvent::Fork>. |
438 |
|
|
|
439 |
root |
1.28 |
There are two class constructors that both create new processes - C<new> |
440 |
|
|
and C<new_exec>. The C<fork> method creates a new process by forking an |
441 |
root |
1.27 |
existing one and could be considered a third constructor. |
442 |
|
|
|
443 |
|
|
Most of the remaining methods deal with preparing the new process, by |
444 |
|
|
loading code, evaluating code and sending data to the new process. They |
445 |
|
|
usually return the process object, so you can chain method calls. |
446 |
|
|
|
447 |
|
|
If a process object is destroyed before calling its C<run> method, then |
448 |
|
|
the process simply exits. After C<run> is called, all responsibility is |
449 |
|
|
passed to the specified function. |
450 |
root |
1.3 |
|
451 |
root |
1.29 |
As long as there is any outstanding work to be done, process objects |
452 |
|
|
resist being destroyed, so there is no reason to store them unless you |
453 |
|
|
need them later - configure and forget works just fine. |
454 |
|
|
|
455 |
root |
1.1 |
=over 4 |
456 |
|
|
|
457 |
|
|
=cut |
458 |
|
|
|
459 |
root |
1.4 |
package AnyEvent::Fork; |
460 |
root |
1.1 |
|
461 |
|
|
use common::sense; |
462 |
|
|
|
463 |
root |
1.18 |
use Errno (); |
464 |
root |
1.1 |
|
465 |
|
|
use AnyEvent; |
466 |
|
|
use AnyEvent::Util (); |
467 |
|
|
|
468 |
root |
1.15 |
use IO::FDPass; |
469 |
|
|
|
470 |
root |
1.73 |
our $VERSION = 1.32; |
471 |
root |
1.12 |
|
472 |
root |
1.5 |
# the early fork template process |
473 |
|
|
our $EARLY; |
474 |
|
|
|
475 |
root |
1.4 |
# the empty template process |
476 |
|
|
our $TEMPLATE; |
477 |
|
|
|
478 |
root |
1.42 |
sub QUEUE() { 0 } |
479 |
|
|
sub FH() { 1 } |
480 |
|
|
sub WW() { 2 } |
481 |
|
|
sub PID() { 3 } |
482 |
|
|
sub CB() { 4 } |
483 |
|
|
|
484 |
|
|
sub _new { |
485 |
|
|
my ($self, $fh, $pid) = @_; |
486 |
|
|
|
487 |
|
|
AnyEvent::Util::fh_nonblocking $fh, 1; |
488 |
|
|
|
489 |
|
|
$self = bless [ |
490 |
|
|
[], # write queue - strings or fd's |
491 |
|
|
$fh, |
492 |
|
|
undef, # AE watcher |
493 |
|
|
$pid, |
494 |
|
|
], $self; |
495 |
|
|
|
496 |
|
|
$self |
497 |
|
|
} |
498 |
|
|
|
499 |
root |
1.4 |
sub _cmd { |
500 |
|
|
my $self = shift; |
501 |
|
|
|
502 |
root |
1.18 |
# ideally, we would want to use "a (w/a)*" as format string, but perl |
503 |
|
|
# versions from at least 5.8.9 to 5.16.3 are all buggy and can't unpack |
504 |
|
|
# it. |
505 |
root |
1.55 |
push @{ $self->[QUEUE] }, pack "a L/a*", $_[0], $_[1]; |
506 |
root |
1.4 |
|
507 |
root |
1.42 |
$self->[WW] ||= AE::io $self->[FH], 1, sub { |
508 |
root |
1.19 |
do { |
509 |
|
|
# send the next "thing" in the queue - either a reference to an fh, |
510 |
|
|
# or a plain string. |
511 |
|
|
|
512 |
root |
1.42 |
if (ref $self->[QUEUE][0]) { |
513 |
root |
1.19 |
# send fh |
514 |
root |
1.42 |
unless (IO::FDPass::send fileno $self->[FH], fileno ${ $self->[QUEUE][0] }) { |
515 |
root |
1.19 |
return if $! == Errno::EAGAIN || $! == Errno::EWOULDBLOCK; |
516 |
root |
1.42 |
undef $self->[WW]; |
517 |
root |
1.19 |
die "AnyEvent::Fork: file descriptor send failure: $!"; |
518 |
root |
1.18 |
} |
519 |
root |
1.4 |
|
520 |
root |
1.42 |
shift @{ $self->[QUEUE] }; |
521 |
root |
1.18 |
|
522 |
root |
1.19 |
} else { |
523 |
|
|
# send string |
524 |
root |
1.42 |
my $len = syswrite $self->[FH], $self->[QUEUE][0]; |
525 |
root |
1.19 |
|
526 |
|
|
unless ($len) { |
527 |
|
|
return if $! == Errno::EAGAIN || $! == Errno::EWOULDBLOCK; |
528 |
root |
1.50 |
undef $self->[WW]; |
529 |
root |
1.19 |
die "AnyEvent::Fork: command write failure: $!"; |
530 |
|
|
} |
531 |
root |
1.18 |
|
532 |
root |
1.42 |
substr $self->[QUEUE][0], 0, $len, ""; |
533 |
|
|
shift @{ $self->[QUEUE] } unless length $self->[QUEUE][0]; |
534 |
root |
1.19 |
} |
535 |
root |
1.42 |
} while @{ $self->[QUEUE] }; |
536 |
root |
1.19 |
|
537 |
|
|
# everything written |
538 |
root |
1.42 |
undef $self->[WW]; |
539 |
root |
1.19 |
|
540 |
|
|
# invoke run callback, if any |
541 |
root |
1.48 |
if ($self->[CB]) { |
542 |
|
|
$self->[CB]->($self->[FH]); |
543 |
|
|
@$self = (); |
544 |
|
|
} |
545 |
root |
1.19 |
}; |
546 |
root |
1.14 |
|
547 |
|
|
() # make sure we don't leak the watcher |
548 |
root |
1.4 |
} |
549 |
root |
1.1 |
|
550 |
root |
1.6 |
# fork template from current process, used by AnyEvent::Fork::Early/Template |
551 |
|
|
sub _new_fork { |
552 |
|
|
my ($fh, $slave) = AnyEvent::Util::portable_socketpair; |
553 |
root |
1.7 |
my $parent = $$; |
554 |
|
|
|
555 |
root |
1.6 |
my $pid = fork; |
556 |
|
|
|
557 |
|
|
if ($pid eq 0) { |
558 |
|
|
require AnyEvent::Fork::Serve; |
559 |
root |
1.7 |
$AnyEvent::Fork::Serve::OWNER = $parent; |
560 |
root |
1.6 |
close $fh; |
561 |
root |
1.69 |
$0 = "$parent AnyEvent::Fork/exec"; |
562 |
root |
1.6 |
AnyEvent::Fork::Serve::serve ($slave); |
563 |
root |
1.15 |
exit 0; |
564 |
root |
1.6 |
} elsif (!$pid) { |
565 |
|
|
die "AnyEvent::Fork::Early/Template: unable to fork template process: $!"; |
566 |
|
|
} |
567 |
|
|
|
568 |
root |
1.19 |
AnyEvent::Fork->_new ($fh, $pid) |
569 |
root |
1.6 |
} |
570 |
|
|
|
571 |
root |
1.4 |
=item my $proc = new AnyEvent::Fork |
572 |
root |
1.1 |
|
573 |
root |
1.4 |
Create a new "empty" perl interpreter process and returns its process |
574 |
|
|
object for further manipulation. |
575 |
root |
1.1 |
|
576 |
root |
1.4 |
The new process is forked from a template process that is kept around |
577 |
|
|
for this purpose. When it doesn't exist yet, it is created by a call to |
578 |
root |
1.29 |
C<new_exec> first and then stays around for future calls. |
579 |
root |
1.9 |
|
580 |
root |
1.4 |
=cut |
581 |
|
|
|
582 |
|
|
sub new { |
583 |
|
|
my $class = shift; |
584 |
root |
1.1 |
|
585 |
root |
1.4 |
$TEMPLATE ||= $class->new_exec; |
586 |
|
|
$TEMPLATE->fork |
587 |
root |
1.1 |
} |
588 |
|
|
|
589 |
root |
1.4 |
=item $new_proc = $proc->fork |
590 |
|
|
|
591 |
|
|
Forks C<$proc>, creating a new process, and returns the process object |
592 |
|
|
of the new process. |
593 |
|
|
|
594 |
|
|
If any of the C<send_> functions have been called before fork, then they |
595 |
|
|
will be cloned in the child. For example, in a pre-forked server, you |
596 |
|
|
might C<send_fh> the listening socket into the template process, and then |
597 |
|
|
keep calling C<fork> and C<run>. |
598 |
|
|
|
599 |
|
|
=cut |
600 |
|
|
|
601 |
|
|
sub fork { |
602 |
|
|
my ($self) = @_; |
603 |
root |
1.1 |
|
604 |
|
|
my ($fh, $slave) = AnyEvent::Util::portable_socketpair; |
605 |
root |
1.4 |
|
606 |
|
|
$self->send_fh ($slave); |
607 |
|
|
$self->_cmd ("f"); |
608 |
|
|
|
609 |
|
|
AnyEvent::Fork->_new ($fh) |
610 |
|
|
} |
611 |
|
|
|
612 |
|
|
=item my $proc = new_exec AnyEvent::Fork |
613 |
|
|
|
614 |
|
|
Create a new "empty" perl interpreter process and returns its process |
615 |
|
|
object for further manipulation. |
616 |
|
|
|
617 |
|
|
Unlike the C<new> method, this method I<always> spawns a new perl process |
618 |
|
|
(except in some cases, see L<AnyEvent::Fork::Early> for details). This |
619 |
|
|
reduces the amount of memory sharing that is possible, and is also slower. |
620 |
|
|
|
621 |
|
|
You should use C<new> whenever possible, except when having a template |
622 |
|
|
process around is unacceptable. |
623 |
|
|
|
624 |
root |
1.17 |
The path to the perl interpreter is divined using various methods - first |
625 |
root |
1.57 |
C<$^X> is investigated to see if the path ends with something that looks |
626 |
root |
1.4 |
as if it were the perl interpreter. Failing this, the module falls back to |
627 |
|
|
using C<$Config::Config{perlpath}>. |
628 |
|
|
|
629 |
root |
1.71 |
The path to perl can also be overridden by setting the global variable |
630 |
root |
1.56 |
C<$AnyEvent::Fork::PERL> - it's value will be used for all subsequent |
631 |
|
|
invocations. |
632 |
|
|
|
633 |
root |
1.4 |
=cut |
634 |
|
|
|
635 |
root |
1.56 |
our $PERL; |
636 |
|
|
|
637 |
root |
1.4 |
sub new_exec { |
638 |
|
|
my ($self) = @_; |
639 |
|
|
|
640 |
root |
1.5 |
return $EARLY->fork |
641 |
|
|
if $EARLY; |
642 |
|
|
|
643 |
root |
1.56 |
unless (defined $PERL) { |
644 |
|
|
# first find path of perl |
645 |
root |
1.62 |
my $perl = $^X; |
646 |
root |
1.56 |
|
647 |
|
|
# first we try $^X, but the path must be absolute (always on win32), and end in sth. |
648 |
|
|
# that looks like perl. this obviously only works for posix and win32 |
649 |
|
|
unless ( |
650 |
|
|
($^O eq "MSWin32" || $perl =~ m%^/%) |
651 |
|
|
&& $perl =~ m%[/\\]perl(?:[0-9]+(\.[0-9]+)+)?(\.exe)?$%i |
652 |
|
|
) { |
653 |
|
|
# if it doesn't look perlish enough, try Config |
654 |
|
|
require Config; |
655 |
|
|
$perl = $Config::Config{perlpath}; |
656 |
|
|
$perl =~ s/(?:\Q$Config::Config{_exe}\E)?$/$Config::Config{_exe}/; |
657 |
|
|
} |
658 |
root |
1.4 |
|
659 |
root |
1.56 |
$PERL = $perl; |
660 |
root |
1.4 |
} |
661 |
|
|
|
662 |
|
|
require Proc::FastSpawn; |
663 |
|
|
|
664 |
|
|
my ($fh, $slave) = AnyEvent::Util::portable_socketpair; |
665 |
|
|
Proc::FastSpawn::fd_inherit (fileno $slave); |
666 |
|
|
|
667 |
root |
1.10 |
# new fh's should always be set cloexec (due to $^F), |
668 |
|
|
# but hey, not on win32, so we always clear the inherit flag. |
669 |
|
|
Proc::FastSpawn::fd_inherit (fileno $fh, 0); |
670 |
|
|
|
671 |
root |
1.4 |
# quick. also doesn't work in win32. of course. what did you expect |
672 |
|
|
#local $ENV{PERL5LIB} = join ":", grep !ref, @INC; |
673 |
root |
1.1 |
my %env = %ENV; |
674 |
root |
1.15 |
$env{PERL5LIB} = join +($^O eq "MSWin32" ? ";" : ":"), grep !ref, @INC; |
675 |
root |
1.1 |
|
676 |
root |
1.19 |
my $pid = Proc::FastSpawn::spawn ( |
677 |
root |
1.56 |
$PERL, |
678 |
root |
1.66 |
[$PERL, "-MAnyEvent::Fork::Serve", "-e", "AnyEvent::Fork::Serve::me", fileno $slave, $$], |
679 |
root |
1.4 |
[map "$_=$env{$_}", keys %env], |
680 |
|
|
) or die "unable to spawn AnyEvent::Fork server: $!"; |
681 |
|
|
|
682 |
root |
1.19 |
$self->_new ($fh, $pid) |
683 |
root |
1.4 |
} |
684 |
|
|
|
685 |
root |
1.20 |
=item $pid = $proc->pid |
686 |
|
|
|
687 |
|
|
Returns the process id of the process I<iff it is a direct child of the |
688 |
root |
1.59 |
process running AnyEvent::Fork>, and C<undef> otherwise. As a general |
689 |
|
|
rule (that you cannot rely upon), processes created via C<new_exec>, |
690 |
|
|
L<AnyEvent::Fork::Early> or L<AnyEvent::Fork::Template> are direct |
691 |
|
|
children, while all other processes are not. |
692 |
|
|
|
693 |
|
|
Or in other words, you do not normally have to take care of zombies for |
694 |
|
|
processes created via C<new>, but when in doubt, or zombies are a problem, |
695 |
|
|
you need to check whether a process is a diretc child by calling this |
696 |
|
|
method, and possibly creating a child watcher or reap it manually. |
697 |
root |
1.20 |
|
698 |
|
|
=cut |
699 |
|
|
|
700 |
|
|
sub pid { |
701 |
root |
1.42 |
$_[0][PID] |
702 |
root |
1.20 |
} |
703 |
|
|
|
704 |
root |
1.9 |
=item $proc = $proc->eval ($perlcode, @args) |
705 |
|
|
|
706 |
root |
1.72 |
Evaluates the given C<$perlcode> as ... Perl code, while setting C<@_> |
707 |
|
|
to the strings specified by C<@args>, in the "main" package (so you can |
708 |
root |
1.74 |
access the args using C<$_[0]> and so on, but not using implicit C<shift> |
709 |
root |
1.72 |
as the latter works on C<@ARGV>). |
710 |
root |
1.9 |
|
711 |
|
|
This call is meant to do any custom initialisation that might be required |
712 |
|
|
(for example, the C<require> method uses it). It's not supposed to be used |
713 |
|
|
to completely take over the process, use C<run> for that. |
714 |
|
|
|
715 |
|
|
The code will usually be executed after this call returns, and there is no |
716 |
|
|
way to pass anything back to the calling process. Any evaluation errors |
717 |
|
|
will be reported to stderr and cause the process to exit. |
718 |
|
|
|
719 |
root |
1.33 |
If you want to execute some code (that isn't in a module) to take over the |
720 |
|
|
process, you should compile a function via C<eval> first, and then call |
721 |
|
|
it via C<run>. This also gives you access to any arguments passed via the |
722 |
|
|
C<send_xxx> methods, such as file handles. See the L<use AnyEvent::Fork as |
723 |
root |
1.34 |
a faster fork+exec> example to see it in action. |
724 |
root |
1.23 |
|
725 |
root |
1.9 |
Returns the process object for easy chaining of method calls. |
726 |
|
|
|
727 |
root |
1.65 |
It's common to want to call an iniitalisation function with some |
728 |
|
|
arguments. Make sure you actually pass C<@_> to that function (for example |
729 |
|
|
by using C<&name> syntax), and do not just specify a function name: |
730 |
|
|
|
731 |
|
|
$proc->eval ('&MyModule::init', $string1, $string2); |
732 |
|
|
|
733 |
root |
1.9 |
=cut |
734 |
|
|
|
735 |
|
|
sub eval { |
736 |
|
|
my ($self, $code, @args) = @_; |
737 |
|
|
|
738 |
root |
1.19 |
$self->_cmd (e => pack "(w/a*)*", $code, @args); |
739 |
root |
1.9 |
|
740 |
|
|
$self |
741 |
|
|
} |
742 |
|
|
|
743 |
root |
1.4 |
=item $proc = $proc->require ($module, ...) |
744 |
root |
1.1 |
|
745 |
root |
1.9 |
Tries to load the given module(s) into the process |
746 |
root |
1.1 |
|
747 |
root |
1.4 |
Returns the process object for easy chaining of method calls. |
748 |
root |
1.1 |
|
749 |
root |
1.9 |
=cut |
750 |
|
|
|
751 |
|
|
sub require { |
752 |
|
|
my ($self, @modules) = @_; |
753 |
|
|
|
754 |
|
|
s%::%/%g for @modules; |
755 |
|
|
$self->eval ('require "$_.pm" for @_', @modules); |
756 |
|
|
|
757 |
|
|
$self |
758 |
|
|
} |
759 |
|
|
|
760 |
root |
1.4 |
=item $proc = $proc->send_fh ($handle, ...) |
761 |
root |
1.1 |
|
762 |
root |
1.4 |
Send one or more file handles (I<not> file descriptors) to the process, |
763 |
|
|
to prepare a call to C<run>. |
764 |
root |
1.1 |
|
765 |
root |
1.35 |
The process object keeps a reference to the handles until they have |
766 |
|
|
been passed over to the process, so you must not explicitly close the |
767 |
|
|
handles. This is most easily accomplished by simply not storing the file |
768 |
|
|
handles anywhere after passing them to this method - when AnyEvent::Fork |
769 |
|
|
is finished using them, perl will automatically close them. |
770 |
root |
1.4 |
|
771 |
|
|
Returns the process object for easy chaining of method calls. |
772 |
|
|
|
773 |
root |
1.17 |
Example: pass a file handle to a process, and release it without |
774 |
|
|
closing. It will be closed automatically when it is no longer used. |
775 |
root |
1.9 |
|
776 |
|
|
$proc->send_fh ($my_fh); |
777 |
|
|
undef $my_fh; # free the reference if you want, but DO NOT CLOSE IT |
778 |
|
|
|
779 |
root |
1.4 |
=cut |
780 |
|
|
|
781 |
|
|
sub send_fh { |
782 |
|
|
my ($self, @fh) = @_; |
783 |
|
|
|
784 |
|
|
for my $fh (@fh) { |
785 |
|
|
$self->_cmd ("h"); |
786 |
root |
1.42 |
push @{ $self->[QUEUE] }, \$fh; |
787 |
root |
1.4 |
} |
788 |
|
|
|
789 |
|
|
$self |
790 |
root |
1.1 |
} |
791 |
|
|
|
792 |
root |
1.4 |
=item $proc = $proc->send_arg ($string, ...) |
793 |
|
|
|
794 |
|
|
Send one or more argument strings to the process, to prepare a call to |
795 |
root |
1.35 |
C<run>. The strings can be any octet strings. |
796 |
root |
1.4 |
|
797 |
root |
1.18 |
The protocol is optimised to pass a moderate number of relatively short |
798 |
|
|
strings - while you can pass up to 4GB of data in one go, this is more |
799 |
|
|
meant to pass some ID information or other startup info, not big chunks of |
800 |
|
|
data. |
801 |
|
|
|
802 |
root |
1.17 |
Returns the process object for easy chaining of method calls. |
803 |
root |
1.4 |
|
804 |
|
|
=cut |
805 |
root |
1.1 |
|
806 |
root |
1.4 |
sub send_arg { |
807 |
|
|
my ($self, @arg) = @_; |
808 |
root |
1.1 |
|
809 |
root |
1.19 |
$self->_cmd (a => pack "(w/a*)*", @arg); |
810 |
root |
1.1 |
|
811 |
|
|
$self |
812 |
|
|
} |
813 |
|
|
|
814 |
root |
1.4 |
=item $proc->run ($func, $cb->($fh)) |
815 |
|
|
|
816 |
root |
1.23 |
Enter the function specified by the function name in C<$func> in the |
817 |
|
|
process. The function is called with the communication socket as first |
818 |
root |
1.4 |
argument, followed by all file handles and string arguments sent earlier |
819 |
|
|
via C<send_fh> and C<send_arg> methods, in the order they were called. |
820 |
|
|
|
821 |
root |
1.35 |
The process object becomes unusable on return from this function - any |
822 |
|
|
further method calls result in undefined behaviour. |
823 |
|
|
|
824 |
root |
1.23 |
The function name should be fully qualified, but if it isn't, it will be |
825 |
root |
1.35 |
looked up in the C<main> package. |
826 |
root |
1.4 |
|
827 |
root |
1.23 |
If the called function returns, doesn't exist, or any error occurs, the |
828 |
|
|
process exits. |
829 |
root |
1.4 |
|
830 |
root |
1.23 |
Preparing the process is done in the background - when all commands have |
831 |
|
|
been sent, the callback is invoked with the local communications socket |
832 |
|
|
as argument. At this point you can start using the socket in any way you |
833 |
|
|
like. |
834 |
|
|
|
835 |
root |
1.4 |
If the communication socket isn't used, it should be closed on both sides, |
836 |
|
|
to save on kernel memory. |
837 |
|
|
|
838 |
|
|
The socket is non-blocking in the parent, and blocking in the newly |
839 |
root |
1.23 |
created process. The close-on-exec flag is set in both. |
840 |
|
|
|
841 |
|
|
Even if not used otherwise, the socket can be a good indicator for the |
842 |
|
|
existence of the process - if the other process exits, you get a readable |
843 |
|
|
event on it, because exiting the process closes the socket (if it didn't |
844 |
|
|
create any children using fork). |
845 |
root |
1.4 |
|
846 |
root |
1.58 |
=over 4 |
847 |
|
|
|
848 |
|
|
=item Compatibility to L<AnyEvent::Fork::Remote> |
849 |
|
|
|
850 |
|
|
If you want to write code that works with both this module and |
851 |
|
|
L<AnyEvent::Fork::Remote>, you need to write your code so that it assumes |
852 |
|
|
there are two file handles for communications, which might not be unix |
853 |
|
|
domain sockets. The C<run> function should start like this: |
854 |
|
|
|
855 |
|
|
sub run { |
856 |
|
|
my ($rfh, @args) = @_; # @args is your normal arguments |
857 |
|
|
my $wfh = fileno $rfh ? $rfh : *STDOUT; |
858 |
|
|
|
859 |
|
|
# now use $rfh for reading and $wfh for writing |
860 |
|
|
} |
861 |
|
|
|
862 |
|
|
This checks whether the passed file handle is, in fact, the process |
863 |
|
|
C<STDIN> handle. If it is, then the function was invoked visa |
864 |
|
|
L<AnyEvent::Fork::Remote>, so STDIN should be used for reading and |
865 |
|
|
C<STDOUT> should be used for writing. |
866 |
|
|
|
867 |
|
|
In all other cases, the function was called via this module, and there is |
868 |
|
|
only one file handle that should be sued for reading and writing. |
869 |
|
|
|
870 |
|
|
=back |
871 |
|
|
|
872 |
root |
1.9 |
Example: create a template for a process pool, pass a few strings, some |
873 |
|
|
file handles, then fork, pass one more string, and run some code. |
874 |
|
|
|
875 |
|
|
my $pool = AnyEvent::Fork |
876 |
|
|
->new |
877 |
|
|
->send_arg ("str1", "str2") |
878 |
|
|
->send_fh ($fh1, $fh2); |
879 |
|
|
|
880 |
|
|
for (1..2) { |
881 |
|
|
$pool |
882 |
|
|
->fork |
883 |
|
|
->send_arg ("str3") |
884 |
|
|
->run ("Some::function", sub { |
885 |
|
|
my ($fh) = @_; |
886 |
|
|
|
887 |
|
|
# fh is nonblocking, but we trust that the OS can accept these |
888 |
root |
1.22 |
# few octets anyway. |
889 |
root |
1.9 |
syswrite $fh, "hi #$_\n"; |
890 |
|
|
|
891 |
|
|
# $fh is being closed here, as we don't store it anywhere |
892 |
|
|
}); |
893 |
|
|
} |
894 |
|
|
|
895 |
|
|
# Some::function might look like this - all parameters passed before fork |
896 |
|
|
# and after will be passed, in order, after the communications socket. |
897 |
|
|
sub Some::function { |
898 |
|
|
my ($fh, $str1, $str2, $fh1, $fh2, $str3) = @_; |
899 |
|
|
|
900 |
root |
1.22 |
print scalar <$fh>; # prints "hi #1\n" and "hi #2\n" in any order |
901 |
root |
1.9 |
} |
902 |
|
|
|
903 |
root |
1.4 |
=cut |
904 |
|
|
|
905 |
|
|
sub run { |
906 |
|
|
my ($self, $func, $cb) = @_; |
907 |
|
|
|
908 |
root |
1.42 |
$self->[CB] = $cb; |
909 |
root |
1.9 |
$self->_cmd (r => $func); |
910 |
root |
1.4 |
} |
911 |
|
|
|
912 |
root |
1.53 |
=back |
913 |
|
|
|
914 |
root |
1.69 |
|
915 |
|
|
=head2 CHILD PROCESS INTERFACE |
916 |
|
|
|
917 |
|
|
This module has a limited API for use in child processes. |
918 |
|
|
|
919 |
|
|
=over 4 |
920 |
|
|
|
921 |
|
|
=item @args = AnyEvent::Fork::Serve::run_args |
922 |
|
|
|
923 |
|
|
This function, which only exists before the C<run> method is called, |
924 |
|
|
returns the arguments that would be passed to the run function, and clears |
925 |
|
|
them. |
926 |
|
|
|
927 |
|
|
This is mainly useful to get any file handles passed via C<send_fh>, but |
928 |
|
|
works for any arguments passed via C<< send_I<xxx> >> methods. |
929 |
|
|
|
930 |
|
|
=back |
931 |
|
|
|
932 |
|
|
|
933 |
root |
1.53 |
=head2 EXPERIMENTAL METHODS |
934 |
|
|
|
935 |
root |
1.55 |
These methods might go away completely or change behaviour, at any time. |
936 |
root |
1.53 |
|
937 |
|
|
=over 4 |
938 |
|
|
|
939 |
root |
1.50 |
=item $proc->to_fh ($cb->($fh)) # EXPERIMENTAL, MIGHT BE REMOVED |
940 |
root |
1.48 |
|
941 |
|
|
Flushes all commands out to the process and then calls the callback with |
942 |
|
|
the communications socket. |
943 |
|
|
|
944 |
|
|
The process object becomes unusable on return from this function - any |
945 |
|
|
further method calls result in undefined behaviour. |
946 |
|
|
|
947 |
root |
1.58 |
The point of this method is to give you a file handle that you can pass |
948 |
root |
1.48 |
to another process. In that other process, you can call C<new_from_fh |
949 |
root |
1.58 |
AnyEvent::Fork $fh> to create a new C<AnyEvent::Fork> object from it, |
950 |
|
|
thereby effectively passing a fork object to another process. |
951 |
root |
1.48 |
|
952 |
|
|
=cut |
953 |
|
|
|
954 |
|
|
sub to_fh { |
955 |
|
|
my ($self, $cb) = @_; |
956 |
|
|
|
957 |
|
|
$self->[CB] = $cb; |
958 |
|
|
|
959 |
|
|
unless ($self->[WW]) { |
960 |
|
|
$self->[CB]->($self->[FH]); |
961 |
|
|
@$self = (); |
962 |
|
|
} |
963 |
|
|
} |
964 |
|
|
|
965 |
root |
1.50 |
=item new_from_fh AnyEvent::Fork $fh # EXPERIMENTAL, MIGHT BE REMOVED |
966 |
root |
1.48 |
|
967 |
|
|
Takes a file handle originally rceeived by the C<to_fh> method and creates |
968 |
|
|
a new C<AnyEvent:Fork> object. The child process itself will not change in |
969 |
|
|
any way, i.e. it will keep all the modifications done to it before calling |
970 |
|
|
C<to_fh>. |
971 |
|
|
|
972 |
|
|
The new object is very much like the original object, except that the |
973 |
|
|
C<pid> method will return C<undef> even if the process is a direct child. |
974 |
|
|
|
975 |
|
|
=cut |
976 |
|
|
|
977 |
|
|
sub new_from_fh { |
978 |
|
|
my ($class, $fh) = @_; |
979 |
|
|
|
980 |
|
|
$class->_new ($fh) |
981 |
|
|
} |
982 |
|
|
|
983 |
root |
1.1 |
=back |
984 |
|
|
|
985 |
root |
1.16 |
=head1 PERFORMANCE |
986 |
|
|
|
987 |
|
|
Now for some unscientific benchmark numbers (all done on an amd64 |
988 |
|
|
GNU/Linux box). These are intended to give you an idea of the relative |
989 |
root |
1.18 |
performance you can expect, they are not meant to be absolute performance |
990 |
|
|
numbers. |
991 |
root |
1.16 |
|
992 |
root |
1.17 |
OK, so, I ran a simple benchmark that creates a socket pair, forks, calls |
993 |
root |
1.16 |
exit in the child and waits for the socket to close in the parent. I did |
994 |
root |
1.18 |
load AnyEvent, EV and AnyEvent::Fork, for a total process size of 5100kB. |
995 |
root |
1.16 |
|
996 |
root |
1.18 |
2079 new processes per second, using manual socketpair + fork |
997 |
root |
1.16 |
|
998 |
|
|
Then I did the same thing, but instead of calling fork, I called |
999 |
|
|
AnyEvent::Fork->new->run ("CORE::exit") and then again waited for the |
1000 |
root |
1.48 |
socket from the child to close on exit. This does the same thing as manual |
1001 |
root |
1.17 |
socket pair + fork, except that what is forked is the template process |
1002 |
root |
1.16 |
(2440kB), and the socket needs to be passed to the server at the other end |
1003 |
|
|
of the socket first. |
1004 |
|
|
|
1005 |
|
|
2307 new processes per second, using AnyEvent::Fork->new |
1006 |
|
|
|
1007 |
|
|
And finally, using C<new_exec> instead C<new>, using vforks+execs to exec |
1008 |
|
|
a new perl interpreter and compile the small server each time, I get: |
1009 |
|
|
|
1010 |
|
|
479 vfork+execs per second, using AnyEvent::Fork->new_exec |
1011 |
|
|
|
1012 |
root |
1.17 |
So how can C<< AnyEvent->new >> be faster than a standard fork, even |
1013 |
|
|
though it uses the same operations, but adds a lot of overhead? |
1014 |
root |
1.16 |
|
1015 |
root |
1.36 |
The difference is simply the process size: forking the 5MB process takes |
1016 |
|
|
so much longer than forking the 2.5MB template process that the extra |
1017 |
root |
1.43 |
overhead is canceled out. |
1018 |
root |
1.16 |
|
1019 |
|
|
If the benchmark process grows, the normal fork becomes even slower: |
1020 |
|
|
|
1021 |
root |
1.36 |
1340 new processes, manual fork of a 20MB process |
1022 |
|
|
731 new processes, manual fork of a 200MB process |
1023 |
|
|
235 new processes, manual fork of a 2000MB process |
1024 |
|
|
|
1025 |
|
|
What that means (to me) is that I can use this module without having a bad |
1026 |
|
|
conscience because of the extra overhead required to start new processes. |
1027 |
root |
1.16 |
|
1028 |
root |
1.15 |
=head1 TYPICAL PROBLEMS |
1029 |
|
|
|
1030 |
|
|
This section lists typical problems that remain. I hope by recognising |
1031 |
|
|
them, most can be avoided. |
1032 |
|
|
|
1033 |
|
|
=over 4 |
1034 |
|
|
|
1035 |
root |
1.36 |
=item leaked file descriptors for exec'ed processes |
1036 |
root |
1.15 |
|
1037 |
|
|
POSIX systems inherit file descriptors by default when exec'ing a new |
1038 |
|
|
process. While perl itself laudably sets the close-on-exec flags on new |
1039 |
|
|
file handles, most C libraries don't care, and even if all cared, it's |
1040 |
|
|
often not possible to set the flag in a race-free manner. |
1041 |
|
|
|
1042 |
|
|
That means some file descriptors can leak through. And since it isn't |
1043 |
root |
1.17 |
possible to know which file descriptors are "good" and "necessary" (or |
1044 |
|
|
even to know which file descriptors are open), there is no good way to |
1045 |
root |
1.15 |
close the ones that might harm. |
1046 |
|
|
|
1047 |
|
|
As an example of what "harm" can be done consider a web server that |
1048 |
|
|
accepts connections and afterwards some module uses AnyEvent::Fork for the |
1049 |
|
|
first time, causing it to fork and exec a new process, which might inherit |
1050 |
|
|
the network socket. When the server closes the socket, it is still open |
1051 |
|
|
in the child (which doesn't even know that) and the client might conclude |
1052 |
|
|
that the connection is still fine. |
1053 |
|
|
|
1054 |
|
|
For the main program, there are multiple remedies available - |
1055 |
|
|
L<AnyEvent::Fork::Early> is one, creating a process early and not using |
1056 |
|
|
C<new_exec> is another, as in both cases, the first process can be exec'ed |
1057 |
|
|
well before many random file descriptors are open. |
1058 |
|
|
|
1059 |
|
|
In general, the solution for these kind of problems is to fix the |
1060 |
|
|
libraries or the code that leaks those file descriptors. |
1061 |
|
|
|
1062 |
root |
1.17 |
Fortunately, most of these leaked descriptors do no harm, other than |
1063 |
root |
1.15 |
sitting on some resources. |
1064 |
|
|
|
1065 |
root |
1.36 |
=item leaked file descriptors for fork'ed processes |
1066 |
root |
1.15 |
|
1067 |
|
|
Normally, L<AnyEvent::Fork> does start new processes by exec'ing them, |
1068 |
|
|
which closes file descriptors not marked for being inherited. |
1069 |
|
|
|
1070 |
|
|
However, L<AnyEvent::Fork::Early> and L<AnyEvent::Fork::Template> offer |
1071 |
|
|
a way to create these processes by forking, and this leaks more file |
1072 |
|
|
descriptors than exec'ing them, as there is no way to mark descriptors as |
1073 |
|
|
"close on fork". |
1074 |
|
|
|
1075 |
|
|
An example would be modules like L<EV>, L<IO::AIO> or L<Gtk2>. Both create |
1076 |
|
|
pipes for internal uses, and L<Gtk2> might open a connection to the X |
1077 |
|
|
server. L<EV> and L<IO::AIO> can deal with fork, but Gtk2 might have |
1078 |
|
|
trouble with a fork. |
1079 |
|
|
|
1080 |
|
|
The solution is to either not load these modules before use'ing |
1081 |
|
|
L<AnyEvent::Fork::Early> or L<AnyEvent::Fork::Template>, or to delay |
1082 |
|
|
initialising them, for example, by calling C<init Gtk2> manually. |
1083 |
|
|
|
1084 |
root |
1.37 |
=item exiting calls object destructors |
1085 |
root |
1.19 |
|
1086 |
root |
1.38 |
This only applies to users of L<AnyEvent::Fork:Early> and |
1087 |
root |
1.44 |
L<AnyEvent::Fork::Template>, or when initialising code creates objects |
1088 |
root |
1.38 |
that reference external resources. |
1089 |
root |
1.19 |
|
1090 |
|
|
When a process created by AnyEvent::Fork exits, it might do so by calling |
1091 |
|
|
exit, or simply letting perl reach the end of the program. At which point |
1092 |
|
|
Perl runs all destructors. |
1093 |
|
|
|
1094 |
|
|
Not all destructors are fork-safe - for example, an object that represents |
1095 |
|
|
the connection to an X display might tell the X server to free resources, |
1096 |
|
|
which is inconvenient when the "real" object in the parent still needs to |
1097 |
|
|
use them. |
1098 |
|
|
|
1099 |
|
|
This is obviously not a problem for L<AnyEvent::Fork::Early>, as you used |
1100 |
|
|
it as the very first thing, right? |
1101 |
|
|
|
1102 |
|
|
It is a problem for L<AnyEvent::Fork::Template> though - and the solution |
1103 |
|
|
is to not create objects with nontrivial destructors that might have an |
1104 |
|
|
effect outside of Perl. |
1105 |
|
|
|
1106 |
root |
1.15 |
=back |
1107 |
|
|
|
1108 |
root |
1.8 |
=head1 PORTABILITY NOTES |
1109 |
|
|
|
1110 |
root |
1.10 |
Native win32 perls are somewhat supported (AnyEvent::Fork::Early is a nop, |
1111 |
|
|
and ::Template is not going to work), and it cost a lot of blood and sweat |
1112 |
|
|
to make it so, mostly due to the bloody broken perl that nobody seems to |
1113 |
|
|
care about. The fork emulation is a bad joke - I have yet to see something |
1114 |
root |
1.17 |
useful that you can do with it without running into memory corruption |
1115 |
root |
1.10 |
issues or other braindamage. Hrrrr. |
1116 |
|
|
|
1117 |
root |
1.49 |
Since fork is endlessly broken on win32 perls (it doesn't even remotely |
1118 |
|
|
work within it's documented limits) and quite obviously it's not getting |
1119 |
|
|
improved any time soon, the best way to proceed on windows would be to |
1120 |
|
|
always use C<new_exec> and thus never rely on perl's fork "emulation". |
1121 |
|
|
|
1122 |
root |
1.36 |
Cygwin perl is not supported at the moment due to some hilarious |
1123 |
root |
1.49 |
shortcomings of its API - see L<IO::FDPoll> for more details. If you never |
1124 |
|
|
use C<send_fh> and always use C<new_exec> to create processes, it should |
1125 |
|
|
work though. |
1126 |
root |
1.8 |
|
1127 |
root |
1.63 |
=head1 USING AnyEvent::Fork IN SUBPROCESSES |
1128 |
|
|
|
1129 |
|
|
AnyEvent::Fork itself cannot generally be used in subprocesses. As long as |
1130 |
|
|
only one process ever forks new processes, sharing the template processes |
1131 |
|
|
is possible (you could use a pipe as a lock by writing a byte into it to |
1132 |
|
|
unlock, and reading the byte to lock for example) |
1133 |
|
|
|
1134 |
|
|
To make concurrent calls possible after fork, you should get rid of the |
1135 |
|
|
template and early fork processes. AnyEvent::Fork will create a new |
1136 |
|
|
template process as needed. |
1137 |
|
|
|
1138 |
|
|
undef $AnyEvent::Fork::EARLY; |
1139 |
|
|
undef $AnyEvent::Fork::TEMPLATE; |
1140 |
|
|
|
1141 |
|
|
It doesn't matter whether you get rid of them in the parent or child after |
1142 |
|
|
a fork. |
1143 |
|
|
|
1144 |
root |
1.13 |
=head1 SEE ALSO |
1145 |
|
|
|
1146 |
root |
1.46 |
L<AnyEvent::Fork::Early>, to avoid executing a perl interpreter at all |
1147 |
|
|
(part of this distribution). |
1148 |
|
|
|
1149 |
|
|
L<AnyEvent::Fork::Template>, to create a process by forking the main |
1150 |
|
|
program at a convenient time (part of this distribution). |
1151 |
|
|
|
1152 |
root |
1.57 |
L<AnyEvent::Fork::Remote>, for another way to create processes that is |
1153 |
|
|
mostly compatible to this module and modules building on top of it, but |
1154 |
|
|
works better with remote processes. |
1155 |
|
|
|
1156 |
root |
1.46 |
L<AnyEvent::Fork::RPC>, for simple RPC to child processes (on CPAN). |
1157 |
root |
1.13 |
|
1158 |
root |
1.51 |
L<AnyEvent::Fork::Pool>, for simple worker process pool (on CPAN). |
1159 |
|
|
|
1160 |
root |
1.43 |
=head1 AUTHOR AND CONTACT INFORMATION |
1161 |
root |
1.1 |
|
1162 |
|
|
Marc Lehmann <schmorp@schmorp.de> |
1163 |
root |
1.43 |
http://software.schmorp.de/pkg/AnyEvent-Fork |
1164 |
root |
1.1 |
|
1165 |
|
|
=cut |
1166 |
|
|
|
1167 |
|
|
1 |
1168 |
|
|
|