[ViewVC] Diff of: cvs/AnyEvent-Fork/Fork.pm

Comparing AnyEvent-Fork/Fork.pm (file contents):
Revision 1.2 by root, Sun Mar 31 03:26:50 2013 UTC vs.
Revision 1.30 by root, Sat Apr 6 09:28:45 2013 UTC

1	=head1 NAME	1	=head1 NAME
2		2
3	AnyEvent::ProcessPool - manage pools of perl worker processes, exec'ed or fork'ed	3	AnyEvent::Fork - everything you wanted to use fork() for, but couldn't
4		4
5	=head1 SYNOPSIS	5	=head1 SYNOPSIS
6		6
7	use AnyEvent::ProcessPool;	7	use AnyEvent::Fork;
		8
		9	AnyEvent::Fork
		10	->new
		11	->require ("MyModule")
		12	->run ("MyModule::server", my $cv = AE::cv);
		13
		14	my $fh = $cv->recv;
8		15
9	=head1 DESCRIPTION	16	=head1 DESCRIPTION
10		17
11	This module allows you to create single worker processes but also worker	18	This module allows you to create new processes, without actually forking
12	pool that share memory, by forking from the main program, or exec'ing new	19	them from your current process (avoiding the problems of forking), but
13	perl interpreters from a module.	20	preserving most of the advantages of fork.
14		21
15	You create a new processes in a pool by specifying a function to call	22	It can be used to create new worker processes or new independent
16	with any combination of string values and file handles.	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.
17		27
18	A pool can have initialisation code which is executed before forking. The	28	Special care has been taken to make this module useful from other modules,
19	initialisation code is only executed once and the resulting process is	29	while still supporting specialised environments such as L<App::Staticperl>
20	cached, to be used as a template.	30	or L<PAR::Packer>.
21		31
22	Pools without such initialisation code don't cache an extra process.	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.
23		44
24	=head1 PROBLEM STATEMENT	45	=head1 PROBLEM STATEMENT
25		46
26	There are two ways to implement parallel processing on UNIX like operating	47	There are two traditional ways to implement parallel processing on UNIX
27	systems - fork and process, and fork+exec and process. They have different	48	like operating systems - fork and process, and fork+exec and process. They
28	advantages and disadvantages that I describe below, together with how this	49	have different advantages and disadvantages that I describe below,
29	module tries to mitigate the disadvantages.	50	together with how this module tries to mitigate the disadvantages.
30		51
31	=over 4	52	=over 4
32		53
33	=item Forking from a big process can be very slow (a 5GB process needs	54	=item Forking from a big process can be very slow.
34	0.05s to fork on my 3.6GHz amd64 GNU/Linux box for example). This overhead	55
		56	A 5GB process needs 0.05s to fork on my 3.6GHz amd64 GNU/Linux box. This
35	is often shared with exec (because you have to fork first), but in some	57	overhead is often shared with exec (because you have to fork first), but
36	circumstances (e.g. when vfork is used), fork+exec can be much faster.	58	in some circumstances (e.g. when vfork is used), fork+exec can be much
		59	faster.
37		60
38	This module can help here by telling a small(er) helper process to fork,	61	This module can help here by telling a small(er) helper process to fork,
39	or fork+exec instead.	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.
40		64
41	=item Forking usually creates a copy-on-write copy of the parent	65	=item Forking usually creates a copy-on-write copy of the parent
42	process. Memory (for example, modules or data files that have been	66	process.
43	will not take additional memory). When exec'ing a new process, modules	67
44	and data files might need to be loaded again, at extra cpu and memory	68	For example, modules or data files that are loaded will not use additional
45	cost. Likewise when forking, all data structures are copied as well - if	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
46	the program frees them and replaces them by new data, the child processes	72	them and replaces them by new data, the child processes will retain the
47	will retain the memory even if it isn't used.	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.
48		78
49	This module allows the main program to do a controlled fork, and allows	79	This module allows the main program to do a controlled fork, and allows
50	modules to exec processes safely at any time. When creating a custom	80	modules to exec processes safely at any time. When creating a custom
51	process pool you can take advantage of data sharing via fork without	81	process pool you can take advantage of data sharing via fork without
52	risking to share large dynamic data structures that will blow up child	82	risking to share large dynamic data structures that will blow up child
53	memory usage.	83	memory usage.
54		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
55	=item Exec'ing a new perl process might be difficult and slow. For	88	=item Exec'ing a new perl process might be difficult.
		89
56	example, it is not easy to find the correct path to the perl interpreter,	90	For example, it is not easy to find the correct path to the perl
57	and all modules have to be loaded from disk again. Long running processes	91	interpreter - C<$^X> might not be a perl interpreter at all.
58	might run into problems when perl is upgraded for example.
59		92
60	This module supports creating pre-initialised perl processes to be used
61	as template, and also tries hard to identify the correct path to the perl	93	This module tries hard to identify the correct path to the perl
62	interpreter. With a cooperative main program, exec'ing the interpreter	94	interpreter. With a cooperative main program, exec'ing the interpreter
63	might not even be necessary.	95	might not even be necessary, but even without help from the main program,
		96	it will still work when used from a module.
64		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
65	=item Forking might be impossible when a program is running. For example,	109	=item Forking might be impossible when a program is running.
66	POSIX makes it almost impossible to fork from a multithreaded program and
67	do anything useful in the child - strictly speaking, if your perl program
68	uses posix threads (even indirectly via e.g. L<IO::AIO> or L<threads>),
69	you cannot call fork on the perl level anymore, at all.
70		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
71	This module can safely fork helper processes at any time, by caling	118	This module can safely fork helper processes at any time, by calling
72	fork+exec in C, in a POSIX-compatible way.	119	fork+exec in C, in a POSIX-compatible way (via L<Proc::FastSpawn>).
73		120
74	=item Parallel processing with fork might be inconvenient or difficult	121	=item Parallel processing with fork might be inconvenient or difficult
		122	to implement. Modules might not work in both parent and child.
		123
75	to implement. For example, when a program uses an event loop and creates	124	For example, when a program uses an event loop and creates watchers it
76	watchers it becomes very hard to use the event loop from a child	125	becomes very hard to use the event loop from a child program, as the
77	program, as the watchers already exist but are only meaningful in the	126	watchers already exist but are only meaningful in the parent. Worse, a
78	parent. Worse, a module might want to use such a system, not knowing	127	module might want to use such a module, not knowing whether another module
79	whether another module or the main program also does, leading to problems.	128	or the main program also does, leading to problems.
80		129
81	This module only lets the main program create pools by forking (because	130	Apart from event loops, graphical toolkits also commonly fall into the
82	only the main program can know when it is still safe to do so) - all other	131	"unsafe module" category, or just about anything that communicates with
83	pools are created by fork+exec, after which such modules can again be	132	the external world, such as network libraries and file I/O modules, which
84	loaded.	133	usually don't like being copied and then allowed to continue in two
		134	processes.
		135
		136	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.
85		140
86	=back	141	=back
87		142
		143	=head1 EXAMPLES
		144
		145	=head2 Create a single new process, tell it to run your worker function.
		146
		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	# $slave_filehandle in the new process.
		155	});
		156
		157	MyModule::worker might look like this:
		158
		159	sub MyModule::worker {
		160	my ($slave_filehandle) = @_;
		161
		162	# now $slave_filehandle is connected to the $master_filehandle
		163	# in the original prorcess. have fun!
		164	}
		165
		166	=head2 Create a pool of server processes all accepting on the same socket.
		167
		168	# create listener socket
		169	my $listener = ...;
		170
		171	# create a pool template, initialise it and give it the socket
		172	my $pool = AnyEvent::Fork
		173	->new
		174	->require ("Some::Stuff", "My::Server")
		175	->send_fh ($listener);
		176
		177	# now create 10 identical workers
		178	for my $id (1..10) {
		179	$pool
		180	->fork
		181	->send_arg ($id)
		182	->run ("My::Server::run");
		183	}
		184
		185	# now do other things - maybe use the filehandle provided by run
		186	# to wait for the processes to die. or whatever.
		187
		188	My::Server::run might look like this:
		189
		190	sub My::Server::run {
		191	my ($slave, $listener, $id) = @_;
		192
		193	close $slave; # we do not use the socket, so close it to save resources
		194
		195	# we could go ballistic and use e.g. AnyEvent here, or IO::AIO,
		196	# or anything we usually couldn't do in a process forked normally.
		197	while (my $socket = $listener->accept) {
		198	# do sth. with new socket
		199	}
		200	}
		201
		202	=head2 use AnyEvent::Fork as a faster fork+exec
		203
		204	This runs /bin/echo hi, with stdout redirected to /tmp/log and stderr to
		205	the communications socket. It is usually faster than fork+exec, but still
		206	let's you prepare the environment.
		207
		208	open my $output, ">/tmp/log" or die "$!";
		209
		210	AnyEvent::Fork
		211	->new
		212	->eval ('
		213	sub run {
		214	my ($fh, $output, @cmd) = @_;
		215
		216	# perl will clear close-on-exec on STDOUT/STDERR
		217	open STDOUT, ">&", $output or die;
		218	open STDERR, ">&", $fh or die;
		219
		220	exec @cmd;
		221	}
		222	')
		223	->send_fh ($output)
		224	->send_arg ("/bin/echo", "hi")
		225	->run ("run", my $cv = AE::cv);
		226
		227	my $stderr = $cv->recv;
		228
		229	=head1 CONCEPTS
		230
		231	This module can create new processes either by executing a new perl
		232	process, or by forking from an existing "template" process.
		233
		234	Each such process comes with its own file handle that can be used to
		235	communicate with it (it's actually a socket - one end in the new process,
		236	one end in the main process), and among the things you can do in it are
		237	load modules, fork new processes, send file handles to it, and execute
		238	functions.
		239
		240	There are multiple ways to create additional processes to execute some
		241	jobs:
		242
88	=over 4	243	=over 4
89		244
90	=cut	245	=item fork a new process from the "default" template process, load code,
		246	run it
91		247
		248	This module has a "default" template process which it executes when it is
		249	needed the first time. Forking from this process shares the memory used
		250	for the perl interpreter with the new process, but loading modules takes
		251	time, and the memory is not shared with anything else.
		252
		253	This is ideal for when you only need one extra process of a kind, with the
		254	option of starting and stopping it on demand.
		255
		256	Example:
		257
		258	AnyEvent::Fork
		259	->new
		260	->require ("Some::Module")
		261	->run ("Some::Module::run", sub {
		262	my ($fork_fh) = @_;
		263	});
		264
		265	=item fork a new template process, load code, then fork processes off of
		266	it and run the code
		267
		268	When you need to have a bunch of processes that all execute the same (or
		269	very similar) tasks, then a good way is to create a new template process
		270	for them, loading all the modules you need, and then create your worker
		271	processes from this new template process.
		272
		273	This way, all code (and data structures) that can be shared (e.g. the
		274	modules you loaded) is shared between the processes, and each new process
		275	consumes relatively little memory of its own.
		276
		277	The disadvantage of this approach is that you need to create a template
		278	process for the sole purpose of forking new processes from it, but if you
		279	only need a fixed number of processes you can create them, and then destroy
		280	the template process.
		281
		282	Example:
		283
		284	my $template = AnyEvent::Fork->new->require ("Some::Module");
		285
		286	for (1..10) {
		287	$template->fork->run ("Some::Module::run", sub {
		288	my ($fork_fh) = @_;
		289	});
		290	}
		291
		292	# at this point, you can keep $template around to fork new processes
		293	# later, or you can destroy it, which causes it to vanish.
		294
		295	=item execute a new perl interpreter, load some code, run it
		296
		297	This is relatively slow, and doesn't allow you to share memory between
		298	multiple processes.
		299
		300	The only advantage is that you don't have to have a template process
		301	hanging around all the time to fork off some new processes, which might be
		302	an advantage when there are long time spans where no extra processes are
		303	needed.
		304
		305	Example:
		306
		307	AnyEvent::Fork
		308	->new_exec
		309	->require ("Some::Module")
		310	->run ("Some::Module::run", sub {
		311	my ($fork_fh) = @_;
		312	});
		313
		314	=back
		315
		316	=head1 THE C<AnyEvent::Fork> CLASS
		317
		318	This module exports nothing, and only implements a single class -
		319	C<AnyEvent::Fork>.
		320
		321	There are two class constructors that both create new processes - C<new>
		322	and C<new_exec>. The C<fork> method creates a new process by forking an
		323	existing one and could be considered a third constructor.
		324
		325	Most of the remaining methods deal with preparing the new process, by
		326	loading code, evaluating code and sending data to the new process. They
		327	usually return the process object, so you can chain method calls.
		328
		329	If a process object is destroyed before calling its C<run> method, then
		330	the process simply exits. After C<run> is called, all responsibility is
		331	passed to the specified function.
		332
		333	As long as there is any outstanding work to be done, process objects
		334	resist being destroyed, so there is no reason to store them unless you
		335	need them later - configure and forget works just fine.
		336
		337	=over 4
		338
		339	=cut
		340
92	package AnyEvent::ProcessPool;	341	package AnyEvent::Fork;
93		342
94	use common::sense;	343	use common::sense;
95		344
96	use Socket ();	345	use Errno ();
97		346
98	use Proc::FastSpawn;
99	use AnyEvent;	347	use AnyEvent;
100	use AnyEvent::ProcessPool::Util;
101	use AnyEvent::Util ();	348	use AnyEvent::Util ();
102		349
103	BEGIN {	350	use IO::FDPass;
104	# require Exporter;
105	}
106		351
107	=item my $pool = new AnyEvent::ProcessPool key => value...	352	our $VERSION = 0.5;
108		353
109	Create a new process pool. The following named parameters are supported:	354	our $PERL; # the path to the perl interpreter, deduces with various forms of magic
110		355
111	=over 4	356	=over 4
112		357
113	=back	358	=back
114		359
115	=cut	360	=cut
116		361
		362	# the early fork template process
		363	our $EARLY;
		364
117	# the template process	365	# the empty template process
118	our $template;	366	our $TEMPLATE;
119		367
120	sub _queue {	368	sub _cmd {
121	my ($pid, $fh) = @_;	369	my $self = shift;
122		370
123	[	371	# ideally, we would want to use "a (w/a)*" as format string, but perl
		372	# versions from at least 5.8.9 to 5.16.3 are all buggy and can't unpack
		373	# it.
		374	push @{ $self->[2] }, pack "a L/a*", $_[0], $_[1];
		375
		376	$self->[3] \|\|= AE::io $self->[1], 1, sub {
		377	do {
		378	# send the next "thing" in the queue - either a reference to an fh,
		379	# or a plain string.
		380
		381	if (ref $self->[2][0]) {
		382	# send fh
		383	unless (IO::FDPass::send fileno $self->[1], fileno ${ $self->[2][0] }) {
		384	return if $! == Errno::EAGAIN \|\| $! == Errno::EWOULDBLOCK;
		385	undef $self->[3];
		386	die "AnyEvent::Fork: file descriptor send failure: $!";
		387	}
		388
		389	shift @{ $self->[2] };
		390
		391	} else {
		392	# send string
		393	my $len = syswrite $self->[1], $self->[2][0];
		394
		395	unless ($len) {
		396	return if $! == Errno::EAGAIN \|\| $! == Errno::EWOULDBLOCK;
		397	undef $self->[3];
		398	die "AnyEvent::Fork: command write failure: $!";
		399	}
		400
		401	substr $self->[2][0], 0, $len, "";
		402	shift @{ $self->[2] } unless length $self->[2][0];
		403	}
		404	} while @{ $self->[2] };
		405
		406	# everything written
		407	undef $self->[3];
		408
		409	# invoke run callback, if any
		410	$self->[4]->($self->[1]) if $self->[4];
		411	};
		412
		413	() # make sure we don't leak the watcher
		414	}
		415
		416	sub _new {
		417	my ($self, $fh, $pid) = @_;
		418
		419	AnyEvent::Util::fh_nonblocking $fh, 1;
		420
		421	$self = bless [
124	$pid,	422	$pid,
125	$fh,	423	$fh,
126	[],	424	[], # write queue - strings or fd's
127	undef	425	undef, # AE watcher
128	]	426	], $self;
129	}
130		427
131	sub queue_cmd {	428	$self
132	my ($queue, $cmd) = @_;
133
134	push @{ $queue->[2] }, pack "N/a", $cmd;
135
136	$queue->[3] \|\|= AE::io $queue->[1], 1, sub {
137	if (ref $queue->[2][0]) {
138	AnyEvent::ProcessPool::Util::fd_send fileno $queue->[1], fileno ${ $queue->[2][0] }
139	and shift @{ $queue->[2] };
140	} else {
141	my $len = syswrite $queue->[1], $queue->[2][0]
142	or do { undef $queue->[3]; die "AnyEvent::ProcessPool::queue write failure: $!" };
143	substr $queue->[2][0], 0, $len, "";
144	shift @{ $queue->[2] } unless length $queue->[2][0];
145	}
146
147	undef $queue->[3] unless @{ $queue->[2] };
148	};
149	}	429	}
150		430
151	sub run_template {	431	# fork template from current process, used by AnyEvent::Fork::Early/Template
152	return if $template;	432	sub _new_fork {
153
154	my ($fh, $slave) = AnyEvent::Util::portable_socketpair;	433	my ($fh, $slave) = AnyEvent::Util::portable_socketpair;
155	AnyEvent::Util::fh_nonblocking $fh, 1;	434	my $parent = $$;
156	fd_inherit fileno $slave;
157		435
158	my %env = %ENV;	436	my $pid = fork;
159	$env{PERL5LIB} = join ":", grep !ref, @INC;
160		437
161	my $pid = spawn	438	if ($pid eq 0) {
162	$^X,	439	require AnyEvent::Fork::Serve;
163	["perl", "-MAnyEvent::ProcessPool::Serve", "-e", "AnyEvent::ProcessPool::Serve::me", fileno $slave],	440	$AnyEvent::Fork::Serve::OWNER = $parent;
164	[map "$_=$env{$_}", keys %env],	441	close $fh;
165	or die "unable to spawn AnyEvent::ProcessPool server: $!";	442	$0 = "$_[1] of $parent";
		443	$SIG{CHLD} = 'IGNORE';
		444	AnyEvent::Fork::Serve::serve ($slave);
		445	exit 0;
		446	} elsif (!$pid) {
		447	die "AnyEvent::Fork::Early/Template: unable to fork template process: $!";
		448	}
166		449
167	close $slave;	450	AnyEvent::Fork->_new ($fh, $pid)
168
169	$template = _queue $pid, $fh;
170
171	my ($a, $b) = AnyEvent::Util::portable_socketpair;
172
173	queue_cmd $template, "Iabc";
174	push @{ $template->[2] }, \$b;
175
176	use Coro::AnyEvent; Coro::AnyEvent::sleep 1;
177	undef $b;
178	die "x" . <$a>;
179	}	451	}
		452
		453	=item my $proc = new AnyEvent::Fork
		454
		455	Create a new "empty" perl interpreter process and returns its process
		456	object for further manipulation.
		457
		458	The new process is forked from a template process that is kept around
		459	for this purpose. When it doesn't exist yet, it is created by a call to
		460	C<new_exec> first and then stays around for future calls.
		461
		462	=cut
180		463
181	sub new {	464	sub new {
182	my $class = shift;	465	my $class = shift;
183		466
184	my $self = bless {	467	$TEMPLATE \|\|= $class->new_exec;
185	@_	468	$TEMPLATE->fork
186	}, $class;	469	}
187		470
188	run_template;	471	=item $new_proc = $proc->fork
		472
		473	Forks C<$proc>, creating a new process, and returns the process object
		474	of the new process.
		475
		476	If any of the C<send_> functions have been called before fork, then they
		477	will be cloned in the child. For example, in a pre-forked server, you
		478	might C<send_fh> the listening socket into the template process, and then
		479	keep calling C<fork> and C<run>.
		480
		481	=cut
		482
		483	sub fork {
		484	my ($self) = @_;
		485
		486	my ($fh, $slave) = AnyEvent::Util::portable_socketpair;
		487
		488	$self->send_fh ($slave);
		489	$self->_cmd ("f");
		490
		491	AnyEvent::Fork->_new ($fh)
		492	}
		493
		494	=item my $proc = new_exec AnyEvent::Fork
		495
		496	Create a new "empty" perl interpreter process and returns its process
		497	object for further manipulation.
		498
		499	Unlike the C<new> method, this method I<always> spawns a new perl process
		500	(except in some cases, see L<AnyEvent::Fork::Early> for details). This
		501	reduces the amount of memory sharing that is possible, and is also slower.
		502
		503	You should use C<new> whenever possible, except when having a template
		504	process around is unacceptable.
		505
		506	The path to the perl interpreter is divined using various methods - first
		507	C<$^X> is investigated to see if the path ends with something that sounds
		508	as if it were the perl interpreter. Failing this, the module falls back to
		509	using C<$Config::Config{perlpath}>.
		510
		511	=cut
		512
		513	sub new_exec {
		514	my ($self) = @_;
		515
		516	return $EARLY->fork
		517	if $EARLY;
		518
		519	# first find path of perl
		520	my $perl = $;
		521
		522	# first we try $^X, but the path must be absolute (always on win32), and end in sth.
		523	# that looks like perl. this obviously only works for posix and win32
		524	unless (
		525	($^O eq "MSWin32" \|\| $perl =~ m%^/%)
		526	&& $perl =~ m%[/\\]perl(?:[0-9]+(\.[0-9]+)+)?(\.exe)?$%i
		527	) {
		528	# if it doesn't look perlish enough, try Config
		529	require Config;
		530	$perl = $Config::Config{perlpath};
		531	$perl =~ s/(?:\Q$Config::Config{_exe}\E)?$/$Config::Config{_exe}/;
		532	}
		533
		534	require Proc::FastSpawn;
		535
		536	my ($fh, $slave) = AnyEvent::Util::portable_socketpair;
		537	Proc::FastSpawn::fd_inherit (fileno $slave);
		538
		539	# new fh's should always be set cloexec (due to $^F),
		540	# but hey, not on win32, so we always clear the inherit flag.
		541	Proc::FastSpawn::fd_inherit (fileno $fh, 0);
		542
		543	# quick. also doesn't work in win32. of course. what did you expect
		544	#local $ENV{PERL5LIB} = join ":", grep !ref, @INC;
		545	my %env = %ENV;
		546	$env{PERL5LIB} = join +($^O eq "MSWin32" ? ";" : ":"), grep !ref, @INC;
		547
		548	my $pid = Proc::FastSpawn::spawn (
		549	$perl,
		550	["perl", "-MAnyEvent::Fork::Serve", "-e", "AnyEvent::Fork::Serve::me", fileno $slave, $$],
		551	[map "$_=$env{$_}", keys %env],
		552	) or die "unable to spawn AnyEvent::Fork server: $!";
		553
		554	$self->_new ($fh, $pid)
		555	}
		556
		557	=item $pid = $proc->pid
		558
		559	Returns the process id of the process I<iff it is a direct child of the
		560	process> running AnyEvent::Fork, and C<undef> otherwise.
		561
		562	Normally, only processes created via C<< AnyEvent::Fork->new_exec >> and
		563	L<AnyEvent::Fork::Template> are direct children, and you are responsible
		564	to clean up their zombies when they die.
		565
		566	All other processes are not direct children, and will be cleaned up by
		567	AnyEvent::Fork itself.
		568
		569	=cut
		570
		571	sub pid {
		572	$_[0][0]
		573	}
		574
		575	=item $proc = $proc->eval ($perlcode, @args)
		576
		577	Evaluates the given C<$perlcode> as ... perl code, while setting C<@_> to
		578	the strings specified by C<@args>, in the "main" package.
		579
		580	This call is meant to do any custom initialisation that might be required
		581	(for example, the C<require> method uses it). It's not supposed to be used
		582	to completely take over the process, use C<run> for that.
		583
		584	The code will usually be executed after this call returns, and there is no
		585	way to pass anything back to the calling process. Any evaluation errors
		586	will be reported to stderr and cause the process to exit.
		587
		588	If you want to execute some code to take over the process (see the
		589	"fork+exec" example in the SYNOPSIS), you should compile a function via
		590	C<eval> first, and then call it via C<run>. This also gives you access to
		591	any arguments passed via the C<send_xxx> methods, such as file handles.
		592
		593	Returns the process object for easy chaining of method calls.
		594
		595	=cut
		596
		597	sub eval {
		598	my ($self, $code, @args) = @_;
		599
		600	$self->_cmd (e => pack "(w/a)", $code, @args);
189		601
190	$self	602	$self
191	}	603	}
192		604
		605	=item $proc = $proc->require ($module, ...)
		606
		607	Tries to load the given module(s) into the process
		608
		609	Returns the process object for easy chaining of method calls.
		610
		611	=cut
		612
		613	sub require {
		614	my ($self, @modules) = @_;
		615
		616	s%::%/%g for @modules;
		617	$self->eval ('require "$_.pm" for @_', @modules);
		618
		619	$self
		620	}
		621
		622	=item $proc = $proc->send_fh ($handle, ...)
		623
		624	Send one or more file handles (I<not> file descriptors) to the process,
		625	to prepare a call to C<run>.
		626
		627	The process object keeps a reference to the handles until this is done,
		628	so you must not explicitly close the handles. This is most easily
		629	accomplished by simply not storing the file handles anywhere after passing
		630	them to this method.
		631
		632	Returns the process object for easy chaining of method calls.
		633
		634	Example: pass a file handle to a process, and release it without
		635	closing. It will be closed automatically when it is no longer used.
		636
		637	$proc->send_fh ($my_fh);
		638	undef $my_fh; # free the reference if you want, but DO NOT CLOSE IT
		639
		640	=cut
		641
		642	sub send_fh {
		643	my ($self, @fh) = @_;
		644
		645	for my $fh (@fh) {
		646	$self->_cmd ("h");
		647	push @{ $self->[2] }, \$fh;
		648	}
		649
		650	$self
		651	}
		652
		653	=item $proc = $proc->send_arg ($string, ...)
		654
		655	Send one or more argument strings to the process, to prepare a call to
		656	C<run>. The strings can be any octet string.
		657
		658	The protocol is optimised to pass a moderate number of relatively short
		659	strings - while you can pass up to 4GB of data in one go, this is more
		660	meant to pass some ID information or other startup info, not big chunks of
		661	data.
		662
		663	Returns the process object for easy chaining of method calls.
		664
		665	=cut
		666
		667	sub send_arg {
		668	my ($self, @arg) = @_;
		669
		670	$self->_cmd (a => pack "(w/a)", @arg);
		671
		672	$self
		673	}
		674
		675	=item $proc->run ($func, $cb->($fh))
		676
		677	Enter the function specified by the function name in C<$func> in the
		678	process. The function is called with the communication socket as first
		679	argument, followed by all file handles and string arguments sent earlier
		680	via C<send_fh> and C<send_arg> methods, in the order they were called.
		681
		682	The function name should be fully qualified, but if it isn't, it will be
		683	looked up in the main package.
		684
		685	If the called function returns, doesn't exist, or any error occurs, the
		686	process exits.
		687
		688	Preparing the process is done in the background - when all commands have
		689	been sent, the callback is invoked with the local communications socket
		690	as argument. At this point you can start using the socket in any way you
		691	like.
		692
		693	The process object becomes unusable on return from this function - any
		694	further method calls result in undefined behaviour.
		695
		696	If the communication socket isn't used, it should be closed on both sides,
		697	to save on kernel memory.
		698
		699	The socket is non-blocking in the parent, and blocking in the newly
		700	created process. The close-on-exec flag is set in both.
		701
		702	Even if not used otherwise, the socket can be a good indicator for the
		703	existence of the process - if the other process exits, you get a readable
		704	event on it, because exiting the process closes the socket (if it didn't
		705	create any children using fork).
		706
		707	Example: create a template for a process pool, pass a few strings, some
		708	file handles, then fork, pass one more string, and run some code.
		709
		710	my $pool = AnyEvent::Fork
		711	->new
		712	->send_arg ("str1", "str2")
		713	->send_fh ($fh1, $fh2);
		714
		715	for (1..2) {
		716	$pool
		717	->fork
		718	->send_arg ("str3")
		719	->run ("Some::function", sub {
		720	my ($fh) = @_;
		721
		722	# fh is nonblocking, but we trust that the OS can accept these
		723	# few octets anyway.
		724	syswrite $fh, "hi #$_\n";
		725
		726	# $fh is being closed here, as we don't store it anywhere
		727	});
		728	}
		729
		730	# Some::function might look like this - all parameters passed before fork
		731	# and after will be passed, in order, after the communications socket.
		732	sub Some::function {
		733	my ($fh, $str1, $str2, $fh1, $fh2, $str3) = @_;
		734
		735	print scalar <$fh>; # prints "hi #1\n" and "hi #2\n" in any order
		736	}
		737
		738	=cut
		739
		740	sub run {
		741	my ($self, $func, $cb) = @_;
		742
		743	$self->[4] = $cb;
		744	$self->_cmd (r => $func);
		745	}
		746
193	=back	747	=back
		748
		749	=head1 PERFORMANCE
		750
		751	Now for some unscientific benchmark numbers (all done on an amd64
		752	GNU/Linux box). These are intended to give you an idea of the relative
		753	performance you can expect, they are not meant to be absolute performance
		754	numbers.
		755
		756	OK, so, I ran a simple benchmark that creates a socket pair, forks, calls
		757	exit in the child and waits for the socket to close in the parent. I did
		758	load AnyEvent, EV and AnyEvent::Fork, for a total process size of 5100kB.
		759
		760	2079 new processes per second, using manual socketpair + fork
		761
		762	Then I did the same thing, but instead of calling fork, I called
		763	AnyEvent::Fork->new->run ("CORE::exit") and then again waited for the
		764	socket form the child to close on exit. This does the same thing as manual
		765	socket pair + fork, except that what is forked is the template process
		766	(2440kB), and the socket needs to be passed to the server at the other end
		767	of the socket first.
		768
		769	2307 new processes per second, using AnyEvent::Fork->new
		770
		771	And finally, using C<new_exec> instead C<new>, using vforks+execs to exec
		772	a new perl interpreter and compile the small server each time, I get:
		773
		774	479 vfork+execs per second, using AnyEvent::Fork->new_exec
		775
		776	So how can C<< AnyEvent->new >> be faster than a standard fork, even
		777	though it uses the same operations, but adds a lot of overhead?
		778
		779	The difference is simply the process size: forking the 6MB process takes
		780	so much longer than forking the 2.5MB template process that the overhead
		781	introduced is canceled out.
		782
		783	If the benchmark process grows, the normal fork becomes even slower:
		784
		785	1340 new processes, manual fork in a 20MB process
		786	731 new processes, manual fork in a 200MB process
		787	235 new processes, manual fork in a 2000MB process
		788
		789	What that means (to me) is that I can use this module without having a
		790	very bad conscience because of the extra overhead required to start new
		791	processes.
		792
		793	=head1 TYPICAL PROBLEMS
		794
		795	This section lists typical problems that remain. I hope by recognising
		796	them, most can be avoided.
		797
		798	=over 4
		799
		800	=item "leaked" file descriptors for exec'ed processes
		801
		802	POSIX systems inherit file descriptors by default when exec'ing a new
		803	process. While perl itself laudably sets the close-on-exec flags on new
		804	file handles, most C libraries don't care, and even if all cared, it's
		805	often not possible to set the flag in a race-free manner.
		806
		807	That means some file descriptors can leak through. And since it isn't
		808	possible to know which file descriptors are "good" and "necessary" (or
		809	even to know which file descriptors are open), there is no good way to
		810	close the ones that might harm.
		811
		812	As an example of what "harm" can be done consider a web server that
		813	accepts connections and afterwards some module uses AnyEvent::Fork for the
		814	first time, causing it to fork and exec a new process, which might inherit
		815	the network socket. When the server closes the socket, it is still open
		816	in the child (which doesn't even know that) and the client might conclude
		817	that the connection is still fine.
		818
		819	For the main program, there are multiple remedies available -
		820	L<AnyEvent::Fork::Early> is one, creating a process early and not using
		821	C<new_exec> is another, as in both cases, the first process can be exec'ed
		822	well before many random file descriptors are open.
		823
		824	In general, the solution for these kind of problems is to fix the
		825	libraries or the code that leaks those file descriptors.
		826
		827	Fortunately, most of these leaked descriptors do no harm, other than
		828	sitting on some resources.
		829
		830	=item "leaked" file descriptors for fork'ed processes
		831
		832	Normally, L<AnyEvent::Fork> does start new processes by exec'ing them,
		833	which closes file descriptors not marked for being inherited.
		834
		835	However, L<AnyEvent::Fork::Early> and L<AnyEvent::Fork::Template> offer
		836	a way to create these processes by forking, and this leaks more file
		837	descriptors than exec'ing them, as there is no way to mark descriptors as
		838	"close on fork".
		839
		840	An example would be modules like L<EV>, L<IO::AIO> or L<Gtk2>. Both create
		841	pipes for internal uses, and L<Gtk2> might open a connection to the X
		842	server. L<EV> and L<IO::AIO> can deal with fork, but Gtk2 might have
		843	trouble with a fork.
		844
		845	The solution is to either not load these modules before use'ing
		846	L<AnyEvent::Fork::Early> or L<AnyEvent::Fork::Template>, or to delay
		847	initialising them, for example, by calling C<init Gtk2> manually.
		848
		849	=item exit runs destructors
		850
		851	This only applies to users of Lc<AnyEvent::Fork:Early> and
		852	L<AnyEvent::Fork::Template>.
		853
		854	When a process created by AnyEvent::Fork exits, it might do so by calling
		855	exit, or simply letting perl reach the end of the program. At which point
		856	Perl runs all destructors.
		857
		858	Not all destructors are fork-safe - for example, an object that represents
		859	the connection to an X display might tell the X server to free resources,
		860	which is inconvenient when the "real" object in the parent still needs to
		861	use them.
		862
		863	This is obviously not a problem for L<AnyEvent::Fork::Early>, as you used
		864	it as the very first thing, right?
		865
		866	It is a problem for L<AnyEvent::Fork::Template> though - and the solution
		867	is to not create objects with nontrivial destructors that might have an
		868	effect outside of Perl.
		869
		870	=back
		871
		872	=head1 PORTABILITY NOTES
		873
		874	Native win32 perls are somewhat supported (AnyEvent::Fork::Early is a nop,
		875	and ::Template is not going to work), and it cost a lot of blood and sweat
		876	to make it so, mostly due to the bloody broken perl that nobody seems to
		877	care about. The fork emulation is a bad joke - I have yet to see something
		878	useful that you can do with it without running into memory corruption
		879	issues or other braindamage. Hrrrr.
		880
		881	Cygwin perl is not supported at the moment, as it should implement fd
		882	passing, but doesn't, and rolling my own is hard, as cygwin doesn't
		883	support enough functionality to do it.
		884
		885	=head1 SEE ALSO
		886
		887	L<AnyEvent::Fork::Early> (to avoid executing a perl interpreter),
		888	L<AnyEvent::Fork::Template> (to create a process by forking the main
		889	program at a convenient time).
194		890
195	=head1 AUTHOR	891	=head1 AUTHOR
196		892
197	Marc Lehmann <schmorp@schmorp.de>	893	Marc Lehmann <schmorp@schmorp.de>
198	http://home.schmorp.de/	894	http://home.schmorp.de/

Diff Legend

-–
+Removed lines
-+
+Added lines
-<
+Changed lines
->
+Changed lines

Comparing AnyEvent-Fork/Fork.pm (file contents): Revision 1.2 by root, Sun Mar 31 03:26:50 2013 UTC vs. Revision 1.30 by root, Sat Apr 6 09:28:45 2013 UTC

Diff Legend

Comparing AnyEvent-Fork/Fork.pm (file contents):
Revision 1.2 by root, Sun Mar 31 03:26:50 2013 UTC vs.
Revision 1.30 by root, Sat Apr 6 09:28:45 2013 UTC