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 |
|
|
=head1 WHAT THIS MODULE IS NOT |
33 |
|
|
|
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 |
|
|
If you need some form of RPC, you can either implement it yourself |
40 |
|
|
in whatever way you like, use some message-passing module such |
41 |
|
|
as L<AnyEvent::MP>, some pipe such as L<AnyEvent::ZeroMQ>, use |
42 |
|
|
L<AnyEvent::Handle> on both sides to send e.g. JSON or Storable messages, |
43 |
|
|
and so on. |
44 |
|
|
|
45 |
root |
1.25 |
=head1 PROBLEM STATEMENT |
46 |
|
|
|
47 |
|
|
There are two traditional ways to implement parallel processing on UNIX |
48 |
|
|
like operating systems - fork and process, and fork+exec and process. They |
49 |
|
|
have different advantages and disadvantages that I describe below, |
50 |
|
|
together with how this module tries to mitigate the disadvantages. |
51 |
|
|
|
52 |
|
|
=over 4 |
53 |
|
|
|
54 |
|
|
=item Forking from a big process can be very slow. |
55 |
|
|
|
56 |
|
|
A 5GB process needs 0.05s to fork on my 3.6GHz amd64 GNU/Linux box. This |
57 |
|
|
overhead is often shared with exec (because you have to fork first), but |
58 |
|
|
in some circumstances (e.g. when vfork is used), fork+exec can be much |
59 |
|
|
faster. |
60 |
|
|
|
61 |
|
|
This module can help here by telling a small(er) helper process to fork, |
62 |
|
|
which is faster then forking the main process, and also uses vfork where |
63 |
|
|
possible. This gives the speed of vfork, with the flexibility of fork. |
64 |
|
|
|
65 |
|
|
=item Forking usually creates a copy-on-write copy of the parent |
66 |
|
|
process. |
67 |
|
|
|
68 |
|
|
For example, modules or data files that are loaded will not use additional |
69 |
|
|
memory after a fork. When exec'ing a new process, modules and data files |
70 |
|
|
might need to be loaded again, at extra CPU and memory cost. But when |
71 |
|
|
forking, literally all data structures are copied - if the program frees |
72 |
|
|
them and replaces them by new data, the child processes will retain the |
73 |
|
|
old version even if it isn't used, which can suddenly and unexpectedly |
74 |
|
|
increase memory usage when freeing memory. |
75 |
|
|
|
76 |
|
|
The trade-off is between more sharing with fork (which can be good or |
77 |
|
|
bad), and no sharing with exec. |
78 |
|
|
|
79 |
|
|
This module allows the main program to do a controlled fork, and allows |
80 |
|
|
modules to exec processes safely at any time. When creating a custom |
81 |
|
|
process pool you can take advantage of data sharing via fork without |
82 |
|
|
risking to share large dynamic data structures that will blow up child |
83 |
|
|
memory usage. |
84 |
|
|
|
85 |
|
|
In other words, this module puts you into control over what is being |
86 |
|
|
shared and what isn't, at all times. |
87 |
|
|
|
88 |
|
|
=item Exec'ing a new perl process might be difficult. |
89 |
|
|
|
90 |
|
|
For example, it is not easy to find the correct path to the perl |
91 |
|
|
interpreter - C<$^X> might not be a perl interpreter at all. |
92 |
|
|
|
93 |
|
|
This module tries hard to identify the correct path to the perl |
94 |
|
|
interpreter. With a cooperative main program, exec'ing the interpreter |
95 |
|
|
might not even be necessary, but even without help from the main program, |
96 |
|
|
it will still work when used from a module. |
97 |
|
|
|
98 |
|
|
=item Exec'ing a new perl process might be slow, as all necessary modules |
99 |
|
|
have to be loaded from disk again, with no guarantees of success. |
100 |
|
|
|
101 |
|
|
Long running processes might run into problems when perl is upgraded |
102 |
|
|
and modules are no longer loadable because they refer to a different |
103 |
|
|
perl version, or parts of a distribution are newer than the ones already |
104 |
|
|
loaded. |
105 |
|
|
|
106 |
|
|
This module supports creating pre-initialised perl processes to be used as |
107 |
|
|
a template for new processes. |
108 |
|
|
|
109 |
|
|
=item Forking might be impossible when a program is running. |
110 |
|
|
|
111 |
|
|
For example, POSIX makes it almost impossible to fork from a |
112 |
|
|
multi-threaded program while doing anything useful in the child - in |
113 |
|
|
fact, if your perl program uses POSIX threads (even indirectly via |
114 |
|
|
e.g. L<IO::AIO> or L<threads>), you cannot call fork on the perl level |
115 |
|
|
anymore without risking corruption issues on a number of operating |
116 |
|
|
systems. |
117 |
|
|
|
118 |
|
|
This module can safely fork helper processes at any time, by calling |
119 |
|
|
fork+exec in C, in a POSIX-compatible way (via L<Proc::FastSpawn>). |
120 |
|
|
|
121 |
|
|
=item Parallel processing with fork might be inconvenient or difficult |
122 |
|
|
to implement. Modules might not work in both parent and child. |
123 |
|
|
|
124 |
|
|
For example, when a program uses an event loop and creates watchers it |
125 |
|
|
becomes very hard to use the event loop from a child program, as the |
126 |
|
|
watchers already exist but are only meaningful in the parent. Worse, a |
127 |
|
|
module might want to use such a module, not knowing whether another module |
128 |
|
|
or the main program also does, leading to problems. |
129 |
|
|
|
130 |
root |
1.26 |
Apart from event loops, graphical toolkits also commonly fall into the |
131 |
|
|
"unsafe module" category, or just about anything that communicates with |
132 |
|
|
the external world, such as network libraries and file I/O modules, which |
133 |
|
|
usually don't like being copied and then allowed to continue in two |
134 |
|
|
processes. |
135 |
|
|
|
136 |
root |
1.25 |
With this module only the main program is allowed to create new processes |
137 |
|
|
by forking (because only the main program can know when it is still safe |
138 |
|
|
to do so) - all other processes are created via fork+exec, which makes it |
139 |
|
|
possible to use modules such as event loops or window interfaces safely. |
140 |
|
|
|
141 |
|
|
=back |
142 |
|
|
|
143 |
root |
1.24 |
=head1 EXAMPLES |
144 |
|
|
|
145 |
|
|
=head2 Create a single new process, tell it to run your worker function. |
146 |
root |
1.9 |
|
147 |
|
|
AnyEvent::Fork |
148 |
|
|
->new |
149 |
|
|
->require ("MyModule") |
150 |
|
|
->run ("MyModule::worker, sub { |
151 |
|
|
my ($master_filehandle) = @_; |
152 |
|
|
|
153 |
|
|
# now $master_filehandle is connected to the |
154 |
|
|
# $slave_filehandle in the new process. |
155 |
|
|
}); |
156 |
|
|
|
157 |
|
|
# MyModule::worker might look like this |
158 |
|
|
sub MyModule::worker { |
159 |
|
|
my ($slave_filehandle) = @_; |
160 |
|
|
|
161 |
|
|
# now $slave_filehandle is connected to the $master_filehandle |
162 |
|
|
# in the original prorcess. have fun! |
163 |
|
|
} |
164 |
|
|
|
165 |
root |
1.24 |
=head2 Create a pool of server processes all accepting on the same socket. |
166 |
root |
1.9 |
|
167 |
|
|
# create listener socket |
168 |
|
|
my $listener = ...; |
169 |
|
|
|
170 |
|
|
# create a pool template, initialise it and give it the socket |
171 |
|
|
my $pool = AnyEvent::Fork |
172 |
|
|
->new |
173 |
|
|
->require ("Some::Stuff", "My::Server") |
174 |
|
|
->send_fh ($listener); |
175 |
|
|
|
176 |
|
|
# now create 10 identical workers |
177 |
|
|
for my $id (1..10) { |
178 |
|
|
$pool |
179 |
|
|
->fork |
180 |
|
|
->send_arg ($id) |
181 |
|
|
->run ("My::Server::run"); |
182 |
|
|
} |
183 |
|
|
|
184 |
|
|
# now do other things - maybe use the filehandle provided by run |
185 |
|
|
# to wait for the processes to die. or whatever. |
186 |
|
|
|
187 |
|
|
# My::Server::run might look like this |
188 |
|
|
sub My::Server::run { |
189 |
|
|
my ($slave, $listener, $id) = @_; |
190 |
|
|
|
191 |
|
|
close $slave; # we do not use the socket, so close it to save resources |
192 |
|
|
|
193 |
|
|
# we could go ballistic and use e.g. AnyEvent here, or IO::AIO, |
194 |
|
|
# or anything we usually couldn't do in a process forked normally. |
195 |
|
|
while (my $socket = $listener->accept) { |
196 |
|
|
# do sth. with new socket |
197 |
|
|
} |
198 |
|
|
} |
199 |
|
|
|
200 |
root |
1.24 |
=head2 use AnyEvent::Fork as a faster fork+exec |
201 |
root |
1.23 |
|
202 |
root |
1.24 |
This runs /bin/echo hi, with stdout redirected to /tmp/log and stderr to |
203 |
|
|
the communications socket. It is usually faster than fork+exec, but still |
204 |
|
|
let's you prepare the environment. |
205 |
root |
1.23 |
|
206 |
|
|
open my $output, ">/tmp/log" or die "$!"; |
207 |
|
|
|
208 |
|
|
AnyEvent::Fork |
209 |
|
|
->new |
210 |
|
|
->eval (' |
211 |
|
|
sub run { |
212 |
|
|
my ($fh, $output, @cmd) = @_; |
213 |
|
|
|
214 |
|
|
# perl will clear close-on-exec on STDOUT/STDERR |
215 |
|
|
open STDOUT, ">&", $output or die; |
216 |
|
|
open STDERR, ">&", $fh or die; |
217 |
|
|
|
218 |
|
|
exec @cmd; |
219 |
|
|
} |
220 |
|
|
') |
221 |
|
|
->send_fh ($output) |
222 |
|
|
->send_arg ("/bin/echo", "hi") |
223 |
|
|
->run ("run", my $cv = AE::cv); |
224 |
|
|
|
225 |
|
|
my $stderr = $cv->recv; |
226 |
|
|
|
227 |
root |
1.3 |
=head1 CONCEPTS |
228 |
|
|
|
229 |
|
|
This module can create new processes either by executing a new perl |
230 |
|
|
process, or by forking from an existing "template" process. |
231 |
|
|
|
232 |
|
|
Each such process comes with its own file handle that can be used to |
233 |
|
|
communicate with it (it's actually a socket - one end in the new process, |
234 |
|
|
one end in the main process), and among the things you can do in it are |
235 |
|
|
load modules, fork new processes, send file handles to it, and execute |
236 |
|
|
functions. |
237 |
|
|
|
238 |
|
|
There are multiple ways to create additional processes to execute some |
239 |
|
|
jobs: |
240 |
|
|
|
241 |
|
|
=over 4 |
242 |
|
|
|
243 |
|
|
=item fork a new process from the "default" template process, load code, |
244 |
|
|
run it |
245 |
|
|
|
246 |
|
|
This module has a "default" template process which it executes when it is |
247 |
|
|
needed the first time. Forking from this process shares the memory used |
248 |
|
|
for the perl interpreter with the new process, but loading modules takes |
249 |
|
|
time, and the memory is not shared with anything else. |
250 |
|
|
|
251 |
|
|
This is ideal for when you only need one extra process of a kind, with the |
252 |
root |
1.17 |
option of starting and stopping it on demand. |
253 |
root |
1.3 |
|
254 |
root |
1.9 |
Example: |
255 |
|
|
|
256 |
|
|
AnyEvent::Fork |
257 |
|
|
->new |
258 |
|
|
->require ("Some::Module") |
259 |
|
|
->run ("Some::Module::run", sub { |
260 |
|
|
my ($fork_fh) = @_; |
261 |
|
|
}); |
262 |
|
|
|
263 |
root |
1.3 |
=item fork a new template process, load code, then fork processes off of |
264 |
|
|
it and run the code |
265 |
|
|
|
266 |
|
|
When you need to have a bunch of processes that all execute the same (or |
267 |
|
|
very similar) tasks, then a good way is to create a new template process |
268 |
|
|
for them, loading all the modules you need, and then create your worker |
269 |
|
|
processes from this new template process. |
270 |
|
|
|
271 |
|
|
This way, all code (and data structures) that can be shared (e.g. the |
272 |
|
|
modules you loaded) is shared between the processes, and each new process |
273 |
|
|
consumes relatively little memory of its own. |
274 |
|
|
|
275 |
|
|
The disadvantage of this approach is that you need to create a template |
276 |
|
|
process for the sole purpose of forking new processes from it, but if you |
277 |
root |
1.17 |
only need a fixed number of processes you can create them, and then destroy |
278 |
root |
1.3 |
the template process. |
279 |
|
|
|
280 |
root |
1.9 |
Example: |
281 |
|
|
|
282 |
|
|
my $template = AnyEvent::Fork->new->require ("Some::Module"); |
283 |
|
|
|
284 |
|
|
for (1..10) { |
285 |
|
|
$template->fork->run ("Some::Module::run", sub { |
286 |
|
|
my ($fork_fh) = @_; |
287 |
|
|
}); |
288 |
|
|
} |
289 |
|
|
|
290 |
|
|
# at this point, you can keep $template around to fork new processes |
291 |
|
|
# later, or you can destroy it, which causes it to vanish. |
292 |
|
|
|
293 |
root |
1.3 |
=item execute a new perl interpreter, load some code, run it |
294 |
|
|
|
295 |
|
|
This is relatively slow, and doesn't allow you to share memory between |
296 |
|
|
multiple processes. |
297 |
|
|
|
298 |
|
|
The only advantage is that you don't have to have a template process |
299 |
|
|
hanging around all the time to fork off some new processes, which might be |
300 |
|
|
an advantage when there are long time spans where no extra processes are |
301 |
|
|
needed. |
302 |
|
|
|
303 |
root |
1.9 |
Example: |
304 |
|
|
|
305 |
|
|
AnyEvent::Fork |
306 |
|
|
->new_exec |
307 |
|
|
->require ("Some::Module") |
308 |
|
|
->run ("Some::Module::run", sub { |
309 |
|
|
my ($fork_fh) = @_; |
310 |
|
|
}); |
311 |
|
|
|
312 |
root |
1.3 |
=back |
313 |
|
|
|
314 |
|
|
=head1 FUNCTIONS |
315 |
|
|
|
316 |
root |
1.1 |
=over 4 |
317 |
|
|
|
318 |
|
|
=cut |
319 |
|
|
|
320 |
root |
1.4 |
package AnyEvent::Fork; |
321 |
root |
1.1 |
|
322 |
|
|
use common::sense; |
323 |
|
|
|
324 |
root |
1.18 |
use Errno (); |
325 |
root |
1.1 |
|
326 |
|
|
use AnyEvent; |
327 |
|
|
use AnyEvent::Util (); |
328 |
|
|
|
329 |
root |
1.15 |
use IO::FDPass; |
330 |
|
|
|
331 |
root |
1.21 |
our $VERSION = 0.5; |
332 |
root |
1.12 |
|
333 |
root |
1.4 |
our $PERL; # the path to the perl interpreter, deduces with various forms of magic |
334 |
root |
1.1 |
|
335 |
root |
1.4 |
=item my $pool = new AnyEvent::Fork key => value... |
336 |
root |
1.1 |
|
337 |
|
|
Create a new process pool. The following named parameters are supported: |
338 |
|
|
|
339 |
|
|
=over 4 |
340 |
|
|
|
341 |
|
|
=back |
342 |
|
|
|
343 |
|
|
=cut |
344 |
|
|
|
345 |
root |
1.5 |
# the early fork template process |
346 |
|
|
our $EARLY; |
347 |
|
|
|
348 |
root |
1.4 |
# the empty template process |
349 |
|
|
our $TEMPLATE; |
350 |
|
|
|
351 |
|
|
sub _cmd { |
352 |
|
|
my $self = shift; |
353 |
|
|
|
354 |
root |
1.18 |
# ideally, we would want to use "a (w/a)*" as format string, but perl |
355 |
|
|
# versions from at least 5.8.9 to 5.16.3 are all buggy and can't unpack |
356 |
|
|
# it. |
357 |
root |
1.19 |
push @{ $self->[2] }, pack "a L/a*", $_[0], $_[1]; |
358 |
root |
1.4 |
|
359 |
root |
1.19 |
$self->[3] ||= AE::io $self->[1], 1, sub { |
360 |
|
|
do { |
361 |
|
|
# send the next "thing" in the queue - either a reference to an fh, |
362 |
|
|
# or a plain string. |
363 |
|
|
|
364 |
|
|
if (ref $self->[2][0]) { |
365 |
|
|
# send fh |
366 |
|
|
unless (IO::FDPass::send fileno $self->[1], fileno ${ $self->[2][0] }) { |
367 |
|
|
return if $! == Errno::EAGAIN || $! == Errno::EWOULDBLOCK; |
368 |
|
|
undef $self->[3]; |
369 |
|
|
die "AnyEvent::Fork: file descriptor send failure: $!"; |
370 |
root |
1.18 |
} |
371 |
root |
1.4 |
|
372 |
root |
1.19 |
shift @{ $self->[2] }; |
373 |
root |
1.18 |
|
374 |
root |
1.19 |
} else { |
375 |
|
|
# send string |
376 |
|
|
my $len = syswrite $self->[1], $self->[2][0]; |
377 |
|
|
|
378 |
|
|
unless ($len) { |
379 |
|
|
return if $! == Errno::EAGAIN || $! == Errno::EWOULDBLOCK; |
380 |
|
|
undef $self->[3]; |
381 |
|
|
die "AnyEvent::Fork: command write failure: $!"; |
382 |
|
|
} |
383 |
root |
1.18 |
|
384 |
root |
1.19 |
substr $self->[2][0], 0, $len, ""; |
385 |
|
|
shift @{ $self->[2] } unless length $self->[2][0]; |
386 |
|
|
} |
387 |
|
|
} while @{ $self->[2] }; |
388 |
|
|
|
389 |
|
|
# everything written |
390 |
|
|
undef $self->[3]; |
391 |
|
|
|
392 |
|
|
# invoke run callback, if any |
393 |
root |
1.20 |
$self->[4]->($self->[1]) if $self->[4]; |
394 |
root |
1.19 |
}; |
395 |
root |
1.14 |
|
396 |
|
|
() # make sure we don't leak the watcher |
397 |
root |
1.4 |
} |
398 |
root |
1.1 |
|
399 |
root |
1.4 |
sub _new { |
400 |
root |
1.19 |
my ($self, $fh, $pid) = @_; |
401 |
root |
1.1 |
|
402 |
root |
1.6 |
AnyEvent::Util::fh_nonblocking $fh, 1; |
403 |
|
|
|
404 |
root |
1.4 |
$self = bless [ |
405 |
root |
1.20 |
$pid, |
406 |
root |
1.1 |
$fh, |
407 |
root |
1.4 |
[], # write queue - strings or fd's |
408 |
|
|
undef, # AE watcher |
409 |
|
|
], $self; |
410 |
|
|
|
411 |
|
|
$self |
412 |
root |
1.1 |
} |
413 |
|
|
|
414 |
root |
1.6 |
# fork template from current process, used by AnyEvent::Fork::Early/Template |
415 |
|
|
sub _new_fork { |
416 |
|
|
my ($fh, $slave) = AnyEvent::Util::portable_socketpair; |
417 |
root |
1.7 |
my $parent = $$; |
418 |
|
|
|
419 |
root |
1.6 |
my $pid = fork; |
420 |
|
|
|
421 |
|
|
if ($pid eq 0) { |
422 |
|
|
require AnyEvent::Fork::Serve; |
423 |
root |
1.7 |
$AnyEvent::Fork::Serve::OWNER = $parent; |
424 |
root |
1.6 |
close $fh; |
425 |
root |
1.7 |
$0 = "$_[1] of $parent"; |
426 |
root |
1.16 |
$SIG{CHLD} = 'IGNORE'; |
427 |
root |
1.6 |
AnyEvent::Fork::Serve::serve ($slave); |
428 |
root |
1.15 |
exit 0; |
429 |
root |
1.6 |
} elsif (!$pid) { |
430 |
|
|
die "AnyEvent::Fork::Early/Template: unable to fork template process: $!"; |
431 |
|
|
} |
432 |
|
|
|
433 |
root |
1.19 |
AnyEvent::Fork->_new ($fh, $pid) |
434 |
root |
1.6 |
} |
435 |
|
|
|
436 |
root |
1.4 |
=item my $proc = new AnyEvent::Fork |
437 |
root |
1.1 |
|
438 |
root |
1.4 |
Create a new "empty" perl interpreter process and returns its process |
439 |
|
|
object for further manipulation. |
440 |
root |
1.1 |
|
441 |
root |
1.4 |
The new process is forked from a template process that is kept around |
442 |
|
|
for this purpose. When it doesn't exist yet, it is created by a call to |
443 |
|
|
C<new_exec> and kept around for future calls. |
444 |
|
|
|
445 |
root |
1.9 |
When the process object is destroyed, it will release the file handle |
446 |
|
|
that connects it with the new process. When the new process has not yet |
447 |
|
|
called C<run>, then the process will exit. Otherwise, what happens depends |
448 |
|
|
entirely on the code that is executed. |
449 |
|
|
|
450 |
root |
1.4 |
=cut |
451 |
|
|
|
452 |
|
|
sub new { |
453 |
|
|
my $class = shift; |
454 |
root |
1.1 |
|
455 |
root |
1.4 |
$TEMPLATE ||= $class->new_exec; |
456 |
|
|
$TEMPLATE->fork |
457 |
root |
1.1 |
} |
458 |
|
|
|
459 |
root |
1.4 |
=item $new_proc = $proc->fork |
460 |
|
|
|
461 |
|
|
Forks C<$proc>, creating a new process, and returns the process object |
462 |
|
|
of the new process. |
463 |
|
|
|
464 |
|
|
If any of the C<send_> functions have been called before fork, then they |
465 |
|
|
will be cloned in the child. For example, in a pre-forked server, you |
466 |
|
|
might C<send_fh> the listening socket into the template process, and then |
467 |
|
|
keep calling C<fork> and C<run>. |
468 |
|
|
|
469 |
|
|
=cut |
470 |
|
|
|
471 |
|
|
sub fork { |
472 |
|
|
my ($self) = @_; |
473 |
root |
1.1 |
|
474 |
|
|
my ($fh, $slave) = AnyEvent::Util::portable_socketpair; |
475 |
root |
1.4 |
|
476 |
|
|
$self->send_fh ($slave); |
477 |
|
|
$self->_cmd ("f"); |
478 |
|
|
|
479 |
|
|
AnyEvent::Fork->_new ($fh) |
480 |
|
|
} |
481 |
|
|
|
482 |
|
|
=item my $proc = new_exec AnyEvent::Fork |
483 |
|
|
|
484 |
|
|
Create a new "empty" perl interpreter process and returns its process |
485 |
|
|
object for further manipulation. |
486 |
|
|
|
487 |
|
|
Unlike the C<new> method, this method I<always> spawns a new perl process |
488 |
|
|
(except in some cases, see L<AnyEvent::Fork::Early> for details). This |
489 |
|
|
reduces the amount of memory sharing that is possible, and is also slower. |
490 |
|
|
|
491 |
|
|
You should use C<new> whenever possible, except when having a template |
492 |
|
|
process around is unacceptable. |
493 |
|
|
|
494 |
root |
1.17 |
The path to the perl interpreter is divined using various methods - first |
495 |
root |
1.4 |
C<$^X> is investigated to see if the path ends with something that sounds |
496 |
|
|
as if it were the perl interpreter. Failing this, the module falls back to |
497 |
|
|
using C<$Config::Config{perlpath}>. |
498 |
|
|
|
499 |
|
|
=cut |
500 |
|
|
|
501 |
|
|
sub new_exec { |
502 |
|
|
my ($self) = @_; |
503 |
|
|
|
504 |
root |
1.5 |
return $EARLY->fork |
505 |
|
|
if $EARLY; |
506 |
|
|
|
507 |
root |
1.4 |
# first find path of perl |
508 |
|
|
my $perl = $; |
509 |
|
|
|
510 |
|
|
# first we try $^X, but the path must be absolute (always on win32), and end in sth. |
511 |
|
|
# that looks like perl. this obviously only works for posix and win32 |
512 |
|
|
unless ( |
513 |
root |
1.15 |
($^O eq "MSWin32" || $perl =~ m%^/%) |
514 |
root |
1.4 |
&& $perl =~ m%[/\\]perl(?:[0-9]+(\.[0-9]+)+)?(\.exe)?$%i |
515 |
|
|
) { |
516 |
|
|
# if it doesn't look perlish enough, try Config |
517 |
|
|
require Config; |
518 |
|
|
$perl = $Config::Config{perlpath}; |
519 |
|
|
$perl =~ s/(?:\Q$Config::Config{_exe}\E)?$/$Config::Config{_exe}/; |
520 |
|
|
} |
521 |
|
|
|
522 |
|
|
require Proc::FastSpawn; |
523 |
|
|
|
524 |
|
|
my ($fh, $slave) = AnyEvent::Util::portable_socketpair; |
525 |
|
|
Proc::FastSpawn::fd_inherit (fileno $slave); |
526 |
|
|
|
527 |
root |
1.10 |
# new fh's should always be set cloexec (due to $^F), |
528 |
|
|
# but hey, not on win32, so we always clear the inherit flag. |
529 |
|
|
Proc::FastSpawn::fd_inherit (fileno $fh, 0); |
530 |
|
|
|
531 |
root |
1.4 |
# quick. also doesn't work in win32. of course. what did you expect |
532 |
|
|
#local $ENV{PERL5LIB} = join ":", grep !ref, @INC; |
533 |
root |
1.1 |
my %env = %ENV; |
534 |
root |
1.15 |
$env{PERL5LIB} = join +($^O eq "MSWin32" ? ";" : ":"), grep !ref, @INC; |
535 |
root |
1.1 |
|
536 |
root |
1.19 |
my $pid = Proc::FastSpawn::spawn ( |
537 |
root |
1.4 |
$perl, |
538 |
root |
1.7 |
["perl", "-MAnyEvent::Fork::Serve", "-e", "AnyEvent::Fork::Serve::me", fileno $slave, $$], |
539 |
root |
1.4 |
[map "$_=$env{$_}", keys %env], |
540 |
|
|
) or die "unable to spawn AnyEvent::Fork server: $!"; |
541 |
|
|
|
542 |
root |
1.19 |
$self->_new ($fh, $pid) |
543 |
root |
1.4 |
} |
544 |
|
|
|
545 |
root |
1.20 |
=item $pid = $proc->pid |
546 |
|
|
|
547 |
|
|
Returns the process id of the process I<iff it is a direct child of the |
548 |
|
|
process> running AnyEvent::Fork, and C<undef> otherwise. |
549 |
|
|
|
550 |
|
|
Normally, only processes created via C<< AnyEvent::Fork->new_exec >> and |
551 |
|
|
L<AnyEvent::Fork::Template> are direct children, and you are responsible |
552 |
|
|
to clean up their zombies when they die. |
553 |
|
|
|
554 |
|
|
All other processes are not direct children, and will be cleaned up by |
555 |
|
|
AnyEvent::Fork. |
556 |
|
|
|
557 |
|
|
=cut |
558 |
|
|
|
559 |
|
|
sub pid { |
560 |
|
|
$_[0][0] |
561 |
|
|
} |
562 |
|
|
|
563 |
root |
1.9 |
=item $proc = $proc->eval ($perlcode, @args) |
564 |
|
|
|
565 |
|
|
Evaluates the given C<$perlcode> as ... perl code, while setting C<@_> to |
566 |
root |
1.23 |
the strings specified by C<@args>, in the "main" package. |
567 |
root |
1.9 |
|
568 |
|
|
This call is meant to do any custom initialisation that might be required |
569 |
|
|
(for example, the C<require> method uses it). It's not supposed to be used |
570 |
|
|
to completely take over the process, use C<run> for that. |
571 |
|
|
|
572 |
|
|
The code will usually be executed after this call returns, and there is no |
573 |
|
|
way to pass anything back to the calling process. Any evaluation errors |
574 |
|
|
will be reported to stderr and cause the process to exit. |
575 |
|
|
|
576 |
root |
1.23 |
If you want to execute some code to take over the process (see the |
577 |
|
|
"fork+exec" example in the SYNOPSIS), you should compile a function via |
578 |
|
|
C<eval> first, and then call it via C<run>. This also gives you access to |
579 |
|
|
any arguments passed via the C<send_xxx> methods, such as file handles. |
580 |
|
|
|
581 |
root |
1.9 |
Returns the process object for easy chaining of method calls. |
582 |
|
|
|
583 |
|
|
=cut |
584 |
|
|
|
585 |
|
|
sub eval { |
586 |
|
|
my ($self, $code, @args) = @_; |
587 |
|
|
|
588 |
root |
1.19 |
$self->_cmd (e => pack "(w/a*)*", $code, @args); |
589 |
root |
1.9 |
|
590 |
|
|
$self |
591 |
|
|
} |
592 |
|
|
|
593 |
root |
1.4 |
=item $proc = $proc->require ($module, ...) |
594 |
root |
1.1 |
|
595 |
root |
1.9 |
Tries to load the given module(s) into the process |
596 |
root |
1.1 |
|
597 |
root |
1.4 |
Returns the process object for easy chaining of method calls. |
598 |
root |
1.1 |
|
599 |
root |
1.9 |
=cut |
600 |
|
|
|
601 |
|
|
sub require { |
602 |
|
|
my ($self, @modules) = @_; |
603 |
|
|
|
604 |
|
|
s%::%/%g for @modules; |
605 |
|
|
$self->eval ('require "$_.pm" for @_', @modules); |
606 |
|
|
|
607 |
|
|
$self |
608 |
|
|
} |
609 |
|
|
|
610 |
root |
1.4 |
=item $proc = $proc->send_fh ($handle, ...) |
611 |
root |
1.1 |
|
612 |
root |
1.4 |
Send one or more file handles (I<not> file descriptors) to the process, |
613 |
|
|
to prepare a call to C<run>. |
614 |
root |
1.1 |
|
615 |
root |
1.4 |
The process object keeps a reference to the handles until this is done, |
616 |
|
|
so you must not explicitly close the handles. This is most easily |
617 |
|
|
accomplished by simply not storing the file handles anywhere after passing |
618 |
|
|
them to this method. |
619 |
|
|
|
620 |
|
|
Returns the process object for easy chaining of method calls. |
621 |
|
|
|
622 |
root |
1.17 |
Example: pass a file handle to a process, and release it without |
623 |
|
|
closing. It will be closed automatically when it is no longer used. |
624 |
root |
1.9 |
|
625 |
|
|
$proc->send_fh ($my_fh); |
626 |
|
|
undef $my_fh; # free the reference if you want, but DO NOT CLOSE IT |
627 |
|
|
|
628 |
root |
1.4 |
=cut |
629 |
|
|
|
630 |
|
|
sub send_fh { |
631 |
|
|
my ($self, @fh) = @_; |
632 |
|
|
|
633 |
|
|
for my $fh (@fh) { |
634 |
|
|
$self->_cmd ("h"); |
635 |
|
|
push @{ $self->[2] }, \$fh; |
636 |
|
|
} |
637 |
|
|
|
638 |
|
|
$self |
639 |
root |
1.1 |
} |
640 |
|
|
|
641 |
root |
1.4 |
=item $proc = $proc->send_arg ($string, ...) |
642 |
|
|
|
643 |
|
|
Send one or more argument strings to the process, to prepare a call to |
644 |
|
|
C<run>. The strings can be any octet string. |
645 |
|
|
|
646 |
root |
1.18 |
The protocol is optimised to pass a moderate number of relatively short |
647 |
|
|
strings - while you can pass up to 4GB of data in one go, this is more |
648 |
|
|
meant to pass some ID information or other startup info, not big chunks of |
649 |
|
|
data. |
650 |
|
|
|
651 |
root |
1.17 |
Returns the process object for easy chaining of method calls. |
652 |
root |
1.4 |
|
653 |
|
|
=cut |
654 |
root |
1.1 |
|
655 |
root |
1.4 |
sub send_arg { |
656 |
|
|
my ($self, @arg) = @_; |
657 |
root |
1.1 |
|
658 |
root |
1.19 |
$self->_cmd (a => pack "(w/a*)*", @arg); |
659 |
root |
1.1 |
|
660 |
|
|
$self |
661 |
|
|
} |
662 |
|
|
|
663 |
root |
1.4 |
=item $proc->run ($func, $cb->($fh)) |
664 |
|
|
|
665 |
root |
1.23 |
Enter the function specified by the function name in C<$func> in the |
666 |
|
|
process. The function is called with the communication socket as first |
667 |
root |
1.4 |
argument, followed by all file handles and string arguments sent earlier |
668 |
|
|
via C<send_fh> and C<send_arg> methods, in the order they were called. |
669 |
|
|
|
670 |
root |
1.23 |
The function name should be fully qualified, but if it isn't, it will be |
671 |
|
|
looked up in the main package. |
672 |
root |
1.4 |
|
673 |
root |
1.23 |
If the called function returns, doesn't exist, or any error occurs, the |
674 |
|
|
process exits. |
675 |
root |
1.4 |
|
676 |
root |
1.23 |
Preparing the process is done in the background - when all commands have |
677 |
|
|
been sent, the callback is invoked with the local communications socket |
678 |
|
|
as argument. At this point you can start using the socket in any way you |
679 |
|
|
like. |
680 |
|
|
|
681 |
|
|
The process object becomes unusable on return from this function - any |
682 |
|
|
further method calls result in undefined behaviour. |
683 |
root |
1.4 |
|
684 |
|
|
If the communication socket isn't used, it should be closed on both sides, |
685 |
|
|
to save on kernel memory. |
686 |
|
|
|
687 |
|
|
The socket is non-blocking in the parent, and blocking in the newly |
688 |
root |
1.23 |
created process. The close-on-exec flag is set in both. |
689 |
|
|
|
690 |
|
|
Even if not used otherwise, the socket can be a good indicator for the |
691 |
|
|
existence of the process - if the other process exits, you get a readable |
692 |
|
|
event on it, because exiting the process closes the socket (if it didn't |
693 |
|
|
create any children using fork). |
694 |
root |
1.4 |
|
695 |
root |
1.9 |
Example: create a template for a process pool, pass a few strings, some |
696 |
|
|
file handles, then fork, pass one more string, and run some code. |
697 |
|
|
|
698 |
|
|
my $pool = AnyEvent::Fork |
699 |
|
|
->new |
700 |
|
|
->send_arg ("str1", "str2") |
701 |
|
|
->send_fh ($fh1, $fh2); |
702 |
|
|
|
703 |
|
|
for (1..2) { |
704 |
|
|
$pool |
705 |
|
|
->fork |
706 |
|
|
->send_arg ("str3") |
707 |
|
|
->run ("Some::function", sub { |
708 |
|
|
my ($fh) = @_; |
709 |
|
|
|
710 |
|
|
# fh is nonblocking, but we trust that the OS can accept these |
711 |
root |
1.22 |
# few octets anyway. |
712 |
root |
1.9 |
syswrite $fh, "hi #$_\n"; |
713 |
|
|
|
714 |
|
|
# $fh is being closed here, as we don't store it anywhere |
715 |
|
|
}); |
716 |
|
|
} |
717 |
|
|
|
718 |
|
|
# Some::function might look like this - all parameters passed before fork |
719 |
|
|
# and after will be passed, in order, after the communications socket. |
720 |
|
|
sub Some::function { |
721 |
|
|
my ($fh, $str1, $str2, $fh1, $fh2, $str3) = @_; |
722 |
|
|
|
723 |
root |
1.22 |
print scalar <$fh>; # prints "hi #1\n" and "hi #2\n" in any order |
724 |
root |
1.9 |
} |
725 |
|
|
|
726 |
root |
1.4 |
=cut |
727 |
|
|
|
728 |
|
|
sub run { |
729 |
|
|
my ($self, $func, $cb) = @_; |
730 |
|
|
|
731 |
root |
1.20 |
$self->[4] = $cb; |
732 |
root |
1.9 |
$self->_cmd (r => $func); |
733 |
root |
1.4 |
} |
734 |
|
|
|
735 |
root |
1.1 |
=back |
736 |
|
|
|
737 |
root |
1.16 |
=head1 PERFORMANCE |
738 |
|
|
|
739 |
|
|
Now for some unscientific benchmark numbers (all done on an amd64 |
740 |
|
|
GNU/Linux box). These are intended to give you an idea of the relative |
741 |
root |
1.18 |
performance you can expect, they are not meant to be absolute performance |
742 |
|
|
numbers. |
743 |
root |
1.16 |
|
744 |
root |
1.17 |
OK, so, I ran a simple benchmark that creates a socket pair, forks, calls |
745 |
root |
1.16 |
exit in the child and waits for the socket to close in the parent. I did |
746 |
root |
1.18 |
load AnyEvent, EV and AnyEvent::Fork, for a total process size of 5100kB. |
747 |
root |
1.16 |
|
748 |
root |
1.18 |
2079 new processes per second, using manual socketpair + fork |
749 |
root |
1.16 |
|
750 |
|
|
Then I did the same thing, but instead of calling fork, I called |
751 |
|
|
AnyEvent::Fork->new->run ("CORE::exit") and then again waited for the |
752 |
|
|
socket form the child to close on exit. This does the same thing as manual |
753 |
root |
1.17 |
socket pair + fork, except that what is forked is the template process |
754 |
root |
1.16 |
(2440kB), and the socket needs to be passed to the server at the other end |
755 |
|
|
of the socket first. |
756 |
|
|
|
757 |
|
|
2307 new processes per second, using AnyEvent::Fork->new |
758 |
|
|
|
759 |
|
|
And finally, using C<new_exec> instead C<new>, using vforks+execs to exec |
760 |
|
|
a new perl interpreter and compile the small server each time, I get: |
761 |
|
|
|
762 |
|
|
479 vfork+execs per second, using AnyEvent::Fork->new_exec |
763 |
|
|
|
764 |
root |
1.17 |
So how can C<< AnyEvent->new >> be faster than a standard fork, even |
765 |
|
|
though it uses the same operations, but adds a lot of overhead? |
766 |
root |
1.16 |
|
767 |
|
|
The difference is simply the process size: forking the 6MB process takes |
768 |
|
|
so much longer than forking the 2.5MB template process that the overhead |
769 |
|
|
introduced is canceled out. |
770 |
|
|
|
771 |
|
|
If the benchmark process grows, the normal fork becomes even slower: |
772 |
|
|
|
773 |
|
|
1340 new processes, manual fork in a 20MB process |
774 |
|
|
731 new processes, manual fork in a 200MB process |
775 |
|
|
235 new processes, manual fork in a 2000MB process |
776 |
|
|
|
777 |
root |
1.17 |
What that means (to me) is that I can use this module without having a |
778 |
|
|
very bad conscience because of the extra overhead required to start new |
779 |
root |
1.16 |
processes. |
780 |
|
|
|
781 |
root |
1.15 |
=head1 TYPICAL PROBLEMS |
782 |
|
|
|
783 |
|
|
This section lists typical problems that remain. I hope by recognising |
784 |
|
|
them, most can be avoided. |
785 |
|
|
|
786 |
|
|
=over 4 |
787 |
|
|
|
788 |
|
|
=item "leaked" file descriptors for exec'ed processes |
789 |
|
|
|
790 |
|
|
POSIX systems inherit file descriptors by default when exec'ing a new |
791 |
|
|
process. While perl itself laudably sets the close-on-exec flags on new |
792 |
|
|
file handles, most C libraries don't care, and even if all cared, it's |
793 |
|
|
often not possible to set the flag in a race-free manner. |
794 |
|
|
|
795 |
|
|
That means some file descriptors can leak through. And since it isn't |
796 |
root |
1.17 |
possible to know which file descriptors are "good" and "necessary" (or |
797 |
|
|
even to know which file descriptors are open), there is no good way to |
798 |
root |
1.15 |
close the ones that might harm. |
799 |
|
|
|
800 |
|
|
As an example of what "harm" can be done consider a web server that |
801 |
|
|
accepts connections and afterwards some module uses AnyEvent::Fork for the |
802 |
|
|
first time, causing it to fork and exec a new process, which might inherit |
803 |
|
|
the network socket. When the server closes the socket, it is still open |
804 |
|
|
in the child (which doesn't even know that) and the client might conclude |
805 |
|
|
that the connection is still fine. |
806 |
|
|
|
807 |
|
|
For the main program, there are multiple remedies available - |
808 |
|
|
L<AnyEvent::Fork::Early> is one, creating a process early and not using |
809 |
|
|
C<new_exec> is another, as in both cases, the first process can be exec'ed |
810 |
|
|
well before many random file descriptors are open. |
811 |
|
|
|
812 |
|
|
In general, the solution for these kind of problems is to fix the |
813 |
|
|
libraries or the code that leaks those file descriptors. |
814 |
|
|
|
815 |
root |
1.17 |
Fortunately, most of these leaked descriptors do no harm, other than |
816 |
root |
1.15 |
sitting on some resources. |
817 |
|
|
|
818 |
|
|
=item "leaked" file descriptors for fork'ed processes |
819 |
|
|
|
820 |
|
|
Normally, L<AnyEvent::Fork> does start new processes by exec'ing them, |
821 |
|
|
which closes file descriptors not marked for being inherited. |
822 |
|
|
|
823 |
|
|
However, L<AnyEvent::Fork::Early> and L<AnyEvent::Fork::Template> offer |
824 |
|
|
a way to create these processes by forking, and this leaks more file |
825 |
|
|
descriptors than exec'ing them, as there is no way to mark descriptors as |
826 |
|
|
"close on fork". |
827 |
|
|
|
828 |
|
|
An example would be modules like L<EV>, L<IO::AIO> or L<Gtk2>. Both create |
829 |
|
|
pipes for internal uses, and L<Gtk2> might open a connection to the X |
830 |
|
|
server. L<EV> and L<IO::AIO> can deal with fork, but Gtk2 might have |
831 |
|
|
trouble with a fork. |
832 |
|
|
|
833 |
|
|
The solution is to either not load these modules before use'ing |
834 |
|
|
L<AnyEvent::Fork::Early> or L<AnyEvent::Fork::Template>, or to delay |
835 |
|
|
initialising them, for example, by calling C<init Gtk2> manually. |
836 |
|
|
|
837 |
root |
1.19 |
=item exit runs destructors |
838 |
|
|
|
839 |
|
|
This only applies to users of Lc<AnyEvent::Fork:Early> and |
840 |
|
|
L<AnyEvent::Fork::Template>. |
841 |
|
|
|
842 |
|
|
When a process created by AnyEvent::Fork exits, it might do so by calling |
843 |
|
|
exit, or simply letting perl reach the end of the program. At which point |
844 |
|
|
Perl runs all destructors. |
845 |
|
|
|
846 |
|
|
Not all destructors are fork-safe - for example, an object that represents |
847 |
|
|
the connection to an X display might tell the X server to free resources, |
848 |
|
|
which is inconvenient when the "real" object in the parent still needs to |
849 |
|
|
use them. |
850 |
|
|
|
851 |
|
|
This is obviously not a problem for L<AnyEvent::Fork::Early>, as you used |
852 |
|
|
it as the very first thing, right? |
853 |
|
|
|
854 |
|
|
It is a problem for L<AnyEvent::Fork::Template> though - and the solution |
855 |
|
|
is to not create objects with nontrivial destructors that might have an |
856 |
|
|
effect outside of Perl. |
857 |
|
|
|
858 |
root |
1.15 |
=back |
859 |
|
|
|
860 |
root |
1.8 |
=head1 PORTABILITY NOTES |
861 |
|
|
|
862 |
root |
1.10 |
Native win32 perls are somewhat supported (AnyEvent::Fork::Early is a nop, |
863 |
|
|
and ::Template is not going to work), and it cost a lot of blood and sweat |
864 |
|
|
to make it so, mostly due to the bloody broken perl that nobody seems to |
865 |
|
|
care about. The fork emulation is a bad joke - I have yet to see something |
866 |
root |
1.17 |
useful that you can do with it without running into memory corruption |
867 |
root |
1.10 |
issues or other braindamage. Hrrrr. |
868 |
|
|
|
869 |
|
|
Cygwin perl is not supported at the moment, as it should implement fd |
870 |
|
|
passing, but doesn't, and rolling my own is hard, as cygwin doesn't |
871 |
|
|
support enough functionality to do it. |
872 |
root |
1.8 |
|
873 |
root |
1.13 |
=head1 SEE ALSO |
874 |
|
|
|
875 |
|
|
L<AnyEvent::Fork::Early> (to avoid executing a perl interpreter), |
876 |
|
|
L<AnyEvent::Fork::Template> (to create a process by forking the main |
877 |
|
|
program at a convenient time). |
878 |
|
|
|
879 |
root |
1.1 |
=head1 AUTHOR |
880 |
|
|
|
881 |
|
|
Marc Lehmann <schmorp@schmorp.de> |
882 |
|
|
http://home.schmorp.de/ |
883 |
|
|
|
884 |
|
|
=cut |
885 |
|
|
|
886 |
|
|
1 |
887 |
|
|
|