ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/IO-AIO/AIO.pm

Comparing IO-AIO/AIO.pm (file contents):
Revision 1.1 by root, Sun Jul 10 17:07:44 2005 UTC vs.
Revision 1.118 by root, Thu Nov 29 20:57:02 2007 UTC

…		…
4		4
5	=head1 SYNOPSIS	5	=head1 SYNOPSIS
6		6
7	use IO::AIO;	7	use IO::AIO;
8		8
		9	aio_open "/etc/passwd", O_RDONLY, 0, sub {
		10	my $fh = shift
		11	or die "/etc/passwd: $!";
		12	...
		13	};
		14
		15	aio_unlink "/tmp/file", sub { };
		16
		17	aio_read $fh, 30000, 1024, $buffer, 0, sub {
		18	$_[0] > 0 or die "read error: $!";
		19	};
		20
		21	# version 2+ has request and group objects
		22	use IO::AIO 2;
		23
		24	aioreq_pri 4; # give next request a very high priority
		25	my $req = aio_unlink "/tmp/file", sub { };
		26	$req->cancel; # cancel request if still in queue
		27
		28	my $grp = aio_group sub { print "all stats done\n" };
		29	add $grp aio_stat "..." for ...;
		30
		31	# AnyEvent integration (EV, Event, Glib, Tk, urxvt, pureperl...)
		32	open my $fh, "<&=" . IO::AIO::poll_fileno or die "$!";
		33	my $w = AnyEvent->io (fh => $fh, poll => 'r', cb => sub { IO::AIO::poll_cb });
		34
		35	# EV integration
		36	my $w = EV::io IO::AIO::poll_fileno, EV::READ, \&IO::AIO::poll_cb;
		37
		38	# Event integration
		39	Event->io (fd => IO::AIO::poll_fileno,
		40	poll => 'r',
		41	cb => \&IO::AIO::poll_cb);
		42
		43	# Glib/Gtk2 integration
		44	add_watch Glib::IO IO::AIO::poll_fileno,
		45	in => sub { IO::AIO::poll_cb; 1 };
		46
		47	# Tk integration
		48	Tk::Event::IO->fileevent (IO::AIO::poll_fileno, "",
		49	readable => \&IO::AIO::poll_cb);
		50
		51	# Danga::Socket integration
		52	Danga::Socket->AddOtherFds (IO::AIO::poll_fileno =>
		53	\&IO::AIO::poll_cb);
		54
9	=head1 DESCRIPTION	55	=head1 DESCRIPTION
10		56
11	This module implements asynchronous I/O using whatever means your	57	This module implements asynchronous I/O using whatever means your
12	operating system supports. Currently, it falls back to Linux::AIO if that	58	operating system supports.
13	module is available, or uses pthreads to emulato aio functionality.
14		59
		60	Asynchronous means that operations that can normally block your program
		61	(e.g. reading from disk) will be done asynchronously: the operation
		62	will still block, but you can do something else in the meantime. This
		63	is extremely useful for programs that need to stay interactive even
		64	when doing heavy I/O (GUI programs, high performance network servers
		65	etc.), but can also be used to easily do operations in parallel that are
		66	normally done sequentially, e.g. stat'ing many files, which is much faster
		67	on a RAID volume or over NFS when you do a number of stat operations
		68	concurrently.
		69
		70	While most of this works on all types of file descriptors (for
		71	example sockets), using these functions on file descriptors that
		72	support nonblocking operation (again, sockets, pipes etc.) is very
		73	inefficient. Use an event loop for that (such as the L<Event|Event>
		74	module): IO::AIO will naturally fit into such an event loop itself.
		75
15	Currently, in this module a number of threads are started that execute	76	In this version, a number of threads are started that execute your
16	your read/writes and signal their completion. You don't need thread	77	requests and signal their completion. You don't need thread support
17	support in your libc or perl, and the threads created by this module will	78	in perl, and the threads created by this module will not be visible
18	not be visible to the pthreads library.	79	to perl. In the future, this module might make use of the native aio
		80	functions available on many operating systems. However, they are often
		81	not well-supported or restricted (GNU/Linux doesn't allow them on normal
		82	files currently, for example), and they would only support aio_read and
		83	aio_write, so the remaining functionality would have to be implemented
		84	using threads anyway.
19		85
20	Although the module will work with in the presence of other threads, it is	86	Although the module will work in the presence of other (Perl-) threads,
21	not reentrant, so use appropriate locking yourself.	87	it is currently not reentrant in any way, so use appropriate locking
		88	yourself, always call C<poll_cb> from within the same thread, or never
		89	call C<poll_cb> (or other C<aio_> functions) recursively.
22		90
23	=head2 API NOTES	91	=head2 EXAMPLE
		92
		93	This is a simple example that uses the Event module and loads
		94	F</etc/passwd> asynchronously:
		95
		96	use Fcntl;
		97	use Event;
		98	use IO::AIO;
		99
		100	# register the IO::AIO callback with Event
		101	Event->io (fd => IO::AIO::poll_fileno,
		102	poll => 'r',
		103	cb => \&IO::AIO::poll_cb);
		104
		105	# queue the request to open /etc/passwd
		106	aio_open "/etc/passwd", O_RDONLY, 0, sub {
		107	my $fh = shift
		108	or die "error while opening: $!";
		109
		110	# stat'ing filehandles is generally non-blocking
		111	my $size = -s $fh;
		112
		113	# queue a request to read the file
		114	my $contents;
		115	aio_read $fh, 0, $size, $contents, 0, sub {
		116	$_[0] == $size
		117	or die "short read: $!";
		118
		119	close $fh;
		120
		121	# file contents now in $contents
		122	print $contents;
		123
		124	# exit event loop and program
		125	Event::unloop;
		126	};
		127	};
		128
		129	# possibly queue up other requests, or open GUI windows,
		130	# check for sockets etc. etc.
		131
		132	# process events as long as there are some:
		133	Event::loop;
		134
		135	=head1 REQUEST ANATOMY AND LIFETIME
		136
		137	Every C<aio_*> function creates a request. which is a C data structure not
		138	directly visible to Perl.
		139
		140	If called in non-void context, every request function returns a Perl
		141	object representing the request. In void context, nothing is returned,
		142	which saves a bit of memory.
		143
		144	The perl object is a fairly standard ref-to-hash object. The hash contents
		145	are not used by IO::AIO so you are free to store anything you like in it.
		146
		147	During their existance, aio requests travel through the following states,
		148	in order:
		149
		150	=over 4
		151
		152	=item ready
		153
		154	Immediately after a request is created it is put into the ready state,
		155	waiting for a thread to execute it.
		156
		157	=item execute
		158
		159	A thread has accepted the request for processing and is currently
		160	executing it (e.g. blocking in read).
		161
		162	=item pending
		163
		164	The request has been executed and is waiting for result processing.
		165
		166	While request submission and execution is fully asynchronous, result
		167	processing is not and relies on the perl interpreter calling C<poll_cb>
		168	(or another function with the same effect).
		169
		170	=item result
		171
		172	The request results are processed synchronously by C<poll_cb>.
		173
		174	The C<poll_cb> function will process all outstanding aio requests by
		175	calling their callbacks, freeing memory associated with them and managing
		176	any groups they are contained in.
		177
		178	=item done
		179
		180	Request has reached the end of its lifetime and holds no resources anymore
		181	(except possibly for the Perl object, but its connection to the actual
		182	aio request is severed and calling its methods will either do nothing or
		183	result in a runtime error).
		184
		185	=back
		186
		187	=cut
		188
		189	package IO::AIO;
		190
		191	use Carp ();
		192
		193	no warnings;
		194	use strict 'vars';
		195
		196	use base 'Exporter';
		197
		198	BEGIN {
		199	our $VERSION = '2.51';
		200
		201	our @AIO_REQ = qw(aio_sendfile aio_read aio_write aio_open aio_close aio_stat
		202	aio_lstat aio_unlink aio_rmdir aio_readdir aio_scandir aio_symlink
		203	aio_readlink aio_fsync aio_fdatasync aio_readahead aio_rename aio_link
		204	aio_move aio_copy aio_group aio_nop aio_mknod aio_load aio_rmtree aio_mkdir
		205	aio_chown aio_chmod aio_utime aio_truncate);
		206	our @EXPORT = (@AIO_REQ, qw(aioreq_pri aioreq_nice aio_block));
		207	our @EXPORT_OK = qw(poll_fileno poll_cb poll_wait flush
		208	min_parallel max_parallel max_idle
		209	nreqs nready npending nthreads
		210	max_poll_time max_poll_reqs);
		211
		212	@IO::AIO::GRP::ISA = 'IO::AIO::REQ';
		213
		214	require XSLoader;
		215	XSLoader::load ("IO::AIO", $VERSION);
		216	}
		217
		218	=head1 FUNCTIONS
		219
		220	=head2 AIO REQUEST FUNCTIONS
24		221
25	All the C<aio_*> calls are more or less thin wrappers around the syscall	222	All the C<aio_*> calls are more or less thin wrappers around the syscall
26	with the same name (sans C<aio_>). The arguments are similar or identical,	223	with the same name (sans C<aio_>). The arguments are similar or identical,
27	and they all accept an additional C<$callback> argument which must be	224	and they all accept an additional (and optional) C<$callback> argument
28	a code reference. This code reference will get called with the syscall	225	which must be a code reference. This code reference will get called with
29	return code (e.g. most syscalls return C<-1> on error, unlike perl, which	226	the syscall return code (e.g. most syscalls return C<-1> on error, unlike
30	usually delivers "false") as it's sole argument when the given syscall has	227	perl, which usually delivers "false") as it's sole argument when the given
31	been executed asynchronously.	228	syscall has been executed asynchronously.
32		229
33	All functions that expect a filehandle will also accept a file descriptor.	230	All functions expecting a filehandle keep a copy of the filehandle
		231	internally until the request has finished.
34		232
		233	All functions return request objects of type L<IO::AIO::REQ> that allow
		234	further manipulation of those requests while they are in-flight.
		235
35	The filenames you pass to these routines I<must> be absolute. The reason	236	The pathnames you pass to these routines I<must> be absolute and
36	is that at the time the request is being executed, the current working	237	encoded as octets. The reason for the former is that at the time the
37	directory could have changed. Alternatively, you can make sure that you	238	request is being executed, the current working directory could have
38	never change the current working directory.	239	changed. Alternatively, you can make sure that you never change the
		240	current working directory anywhere in the program and then use relative
		241	paths.
		242
		243	To encode pathnames as octets, either make sure you either: a) always pass
		244	in filenames you got from outside (command line, readdir etc.) without
		245	tinkering, b) are ASCII or ISO 8859-1, c) use the Encode module and encode
		246	your pathnames to the locale (or other) encoding in effect in the user
		247	environment, d) use Glib::filename_from_unicode on unicode filenames or e)
		248	use something else to ensure your scalar has the correct contents.
		249
		250	This works, btw. independent of the internal UTF-8 bit, which IO::AIO
		251	handles correctly wether it is set or not.
39		252
40	=over 4	253	=over 4
41		254
42	=cut	255	=item $prev_pri = aioreq_pri [$pri]
43		256
44	package IO::AIO;	257	Returns the priority value that would be used for the next request and, if
		258	C<$pri> is given, sets the priority for the next aio request.
45		259
46	use base 'Exporter';	260	The default priority is C<0>, the minimum and maximum priorities are C<-4>
		261	and C<4>, respectively. Requests with higher priority will be serviced
		262	first.
47		263
48	BEGIN {	264	The priority will be reset to C<0> after each call to one of the C<aio_*>
49	$VERSION = 0.1;	265	functions.
50		266
51	@EXPORT = qw(aio_read aio_write aio_open aio_close aio_stat aio_lstat aio_unlink	267	Example: open a file with low priority, then read something from it with
52	aio_fsync aio_fdatasync aio_readahead);	268	higher priority so the read request is serviced before other low priority
53	@EXPORT_OK = qw(poll_fileno poll_cb min_parallel max_parallel nreqs);	269	open requests (potentially spamming the cache):
54		270
55	require XSLoader;	271	aioreq_pri -3;
56	XSLoader::load IO::AIO, $VERSION;	272	aio_open ..., sub {
57	}	273	return unless $_[0];
58		274
59	=item IO::AIO::min_parallel $nthreads	275	aioreq_pri -2;
		276	aio_read $_[0], ..., sub {
		277	...
		278	};
		279	};
60		280
61	Set the minimum number of AIO threads to C<$nthreads>. The default is
62	C<1>, which means a single asynchronous operation can be done at one time
63	(the number of outstanding operations, however, is unlimited).
64		281
65	It is recommended to keep the number of threads low, as some linux	282	=item aioreq_nice $pri_adjust
66	kernel versions will scale negatively with the number of threads (higher
67	parallelity => MUCH higher latency).
68		283
69	Under normal circumstances you don't need to call this function, as this	284	Similar to C<aioreq_pri>, but subtracts the given value from the current
70	module automatically starts a single async thread.	285	priority, so the effect is cumulative.
71		286
72	=item IO::AIO::max_parallel $nthreads
73		287
74	Sets the maximum number of AIO threads to C<$nthreads>. If more than
75	the specified number of threads are currently running, kill them. This
76	function blocks until the limit is reached.
77
78	This module automatically runs C<max_parallel 0> at program end, to ensure
79	that all threads are killed and that there are no outstanding requests.
80
81	Under normal circumstances you don't need to call this function.
82
83	=item $fileno = IO::AIO::poll_fileno
84
85	Return the I<request result pipe filehandle>. This filehandle must be
86	polled for reading by some mechanism outside this module (e.g. Event
87	or select, see below). If the pipe becomes readable you have to call
88	C<poll_cb> to check the results.
89
90	See C<poll_cb> for an example.
91
92	=item IO::AIO::poll_cb
93
94	Process all outstanding events on the result pipe. You have to call this
95	regularly. Returns the number of events processed. Returns immediately
96	when no events are outstanding.
97
98	You can use Event to multiplex, e.g.:
99
100	Event->io (fd => IO::AIO::poll_fileno,
101	poll => 'r', async => 1,
102	cb => \&IO::AIO::poll_cb);
103
104	=item IO::AIO::poll_wait
105
106	Wait till the result filehandle becomes ready for reading (simply does a
107	select on the filehandle. This is useful if you want to synchronously wait
108	for some requests to finish).
109
110	See C<nreqs> for an example.
111
112	=item IO::AIO::nreqs
113
114	Returns the number of requests currently outstanding.
115
116	Example: wait till there are no outstanding requests anymore:
117
118	IO::AIO::poll_wait, IO::AIO::poll_cb
119	while IO::AIO::nreqs;
120
121	=item aio_open $pathname, $flags, $mode, $callback	288	=item aio_open $pathname, $flags, $mode, $callback->($fh)
122		289
123	Asynchronously open or create a file and call the callback with the	290	Asynchronously open or create a file and call the callback with a newly
124	filedescriptor (NOT a perl filehandle, sorry for that, but watch out, this	291	created filehandle for the file.
125	might change in the future).
126		292
127	The pathname passed to C<aio_open> must be absolute. See API NOTES, above,	293	The pathname passed to C<aio_open> must be absolute. See API NOTES, above,
128	for an explanation.	294	for an explanation.
129		295
130	The C<$mode> argument is a bitmask. See the C<Fcntl> module for a	296	The C<$flags> argument is a bitmask. See the C<Fcntl> module for a
131	list. They are the same as used in C<sysopen>.	297	list. They are the same as used by C<sysopen>.
		298
		299	Likewise, C<$mode> specifies the mode of the newly created file, if it
		300	didn't exist and C<O_CREAT> has been given, just like perl's C<sysopen>,
		301	except that it is mandatory (i.e. use C<0> if you don't create new files,
		302	and C<0666> or C<0777> if you do). Note that the C<$mode> will be modified
		303	by the umask in effect then the request is being executed, so better never
		304	change the umask.
132		305
133	Example:	306	Example:
134		307
135	aio_open "/etc/passwd", O_RDONLY, 0, sub {	308	aio_open "/etc/passwd", O_RDONLY, 0, sub {
136	if ($_[0] >= 0) {	309	if ($_[0]) {
137	open my $fh, "<&$_[0]"; # create a copy for perl
138	aio_close $_[0], sub { }; # close the aio handle
139	print "open successful, fh is $fh\n";	310	print "open successful, fh is $_[0]\n";
140	...	311	...
141	} else {	312	} else {
142	die "open failed: $!\n";	313	die "open failed: $!\n";
143	}	314	}
144	};	315	};
145		316
		317
146	=item aio_close $fh, $callback	318	=item aio_close $fh, $callback->($status)
147		319
148	Asynchronously close a file and call the callback with the result code.	320	Asynchronously close a file and call the callback with the result
		321	code.
149		322
		323	Unfortunately, you can't do this to perl. Perl I<insists> very strongly on
		324	closing the file descriptor associated with the filehandle itself. Here is
		325	what aio_close will try:
		326
		327	1. dup()licate the fd
		328	2. asynchronously close() the duplicated fd
		329	3. dup()licate the fd once more
		330	4. let perl close() the filehandle
		331	5. asynchronously close the duplicated fd
		332
		333	The idea is that the first close() flushes stuff to disk that closing an
		334	fd will flush, so when perl closes the fd, nothing much will need to be
		335	flushed. The second async. close() will then flush stuff to disk that
		336	closing the last fd to the file will flush.
		337
		338	Just FYI, SuSv3 has this to say on close:
		339
		340	All outstanding record locks owned by the process on the file
		341	associated with the file descriptor shall be removed.
		342
		343	If fildes refers to a socket, close() shall cause the socket to be
		344	destroyed. ... close() shall block for up to the current linger
		345	interval until all data is transmitted.
		346	[this actually sounds like a specification bug, but who knows]
		347
		348	And at least Linux additionally actually flushes stuff on every close,
		349	even when the file itself is still open.
		350
		351	Sounds enourmously inefficient and complicated? Yes... please show me how
		352	to nuke perl's fd out of existence...
		353
		354	=cut
		355
		356	sub aio_close($;$) {
		357	aio_block {
		358	my ($fh, $cb) = @_;
		359
		360	my $pri = aioreq_pri;
		361	my $grp = aio_group $cb;
		362
		363	my $fd = fileno $fh;
		364
		365	defined $fd or Carp::croak "aio_close called with fd-less filehandle";
		366
		367	# if the dups fail we will simply get EBADF
		368	my $fd2 = _dup $fd;
		369	aioreq_pri $pri;
		370	add $grp _aio_close $fd2, sub {
		371	my $fd2 = _dup $fd;
		372	close $fh;
		373	aioreq_pri $pri;
		374	add $grp _aio_close $fd2, sub {
		375	$grp->result ($_[0]);
		376	};
		377	};
		378
		379	$grp
		380	}
		381	}
		382
		383
150	=item aio_read $fh,$offset,$length, $data,$dataoffset,$callback	384	=item aio_read $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
151		385
152	=item aio_write $fh,$offset,$length, $data,$dataoffset,$callback	386	=item aio_write $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
153		387
154	Reads or writes C<length> bytes from the specified C<fh> and C<offset>	388	Reads or writes C<$length> bytes from the specified C<$fh> and C<$offset>
155	into the scalar given by C<data> and offset C<dataoffset> and calls the	389	into the scalar given by C<$data> and offset C<$dataoffset> and calls the
156	callback without the actual number of bytes read (or -1 on error, just	390	callback without the actual number of bytes read (or -1 on error, just
157	like the syscall).	391	like the syscall).
158		392
		393	If C<$offset> is undefined, then the current file descriptor offset will
		394	be used (and updated), otherwise the file descriptor offset will not be
		395	changed by these calls.
		396
		397	If C<$length> is undefined in C<aio_write>, use the remaining length of C<$data>.
		398
		399	If C<$dataoffset> is less than zero, it will be counted from the end of
		400	C<$data>.
		401
		402	The C<$data> scalar I<MUST NOT> be modified in any way while the request
		403	is outstanding. Modifying it can result in segfaults or World War III (if
		404	the necessary/optional hardware is installed).
		405
159	Example: Read 15 bytes at offset 7 into scalar C<$buffer>, strating at	406	Example: Read 15 bytes at offset 7 into scalar C<$buffer>, starting at
160	offset C<0> within the scalar:	407	offset C<0> within the scalar:
161		408
162	aio_read $fh, 7, 15, $buffer, 0, sub {	409	aio_read $fh, 7, 15, $buffer, 0, sub {
163	$_[0] >= 0 or die "read error: $!";	410	$_[0] > 0 or die "read error: $!";
164	print "read <$buffer>\n";	411	print "read $_[0] bytes: <$buffer>\n";
165	};	412	};
166		413
		414
		415	=item aio_sendfile $out_fh, $in_fh, $in_offset, $length, $callback->($retval)
		416
		417	Tries to copy C<$length> bytes from C<$in_fh> to C<$out_fh>. It starts
		418	reading at byte offset C<$in_offset>, and starts writing at the current
		419	file offset of C<$out_fh>. Because of that, it is not safe to issue more
		420	than one C<aio_sendfile> per C<$out_fh>, as they will interfere with each
		421	other.
		422
		423	This call tries to make use of a native C<sendfile> syscall to provide
		424	zero-copy operation. For this to work, C<$out_fh> should refer to a
		425	socket, and C<$in_fh> should refer to mmap'able file.
		426
		427	If the native sendfile call fails or is not implemented, it will be
		428	emulated, so you can call C<aio_sendfile> on any type of filehandle
		429	regardless of the limitations of the operating system.
		430
		431	Please note, however, that C<aio_sendfile> can read more bytes from
		432	C<$in_fh> than are written, and there is no way to find out how many
		433	bytes have been read from C<aio_sendfile> alone, as C<aio_sendfile> only
		434	provides the number of bytes written to C<$out_fh>. Only if the result
		435	value equals C<$length> one can assume that C<$length> bytes have been
		436	read.
		437
		438
167	=item aio_readahead $fh,$offset,$length, $callback	439	=item aio_readahead $fh,$offset,$length, $callback->($retval)
168		440
169	Asynchronously reads the specified byte range into the page cache, using
170	the C<readahead> syscall. If that syscall doesn't exist the status will be
171	C<-1> and C<$!> is set to ENOSYS.
172
173	readahead() populates the page cache with data from a file so that	441	C<aio_readahead> populates the page cache with data from a file so that
174	subsequent reads from that file will not block on disk I/O. The C<$offset>	442	subsequent reads from that file will not block on disk I/O. The C<$offset>
175	argument specifies the starting point from which data is to be read and	443	argument specifies the starting point from which data is to be read and
176	C<$length> specifies the number of bytes to be read. I/O is performed in	444	C<$length> specifies the number of bytes to be read. I/O is performed in
177	whole pages, so that offset is effectively rounded down to a page boundary	445	whole pages, so that offset is effectively rounded down to a page boundary
178	and bytes are read up to the next page boundary greater than or equal to	446	and bytes are read up to the next page boundary greater than or equal to
179	(off-set+length). aio_readahead() does not read beyond the end of the	447	(off-set+length). C<aio_readahead> does not read beyond the end of the
180	file. The current file offset of the file is left unchanged.	448	file. The current file offset of the file is left unchanged.
181		449
		450	If that syscall doesn't exist (likely if your OS isn't Linux) it will be
		451	emulated by simply reading the data, which would have a similar effect.
		452
		453
182	=item aio_stat $fh_or_path, $callback	454	=item aio_stat $fh_or_path, $callback->($status)
183		455
184	=item aio_lstat $fh, $callback	456	=item aio_lstat $fh, $callback->($status)
185		457
186	Works like perl's C<stat> or C<lstat> in void context. The callback will	458	Works like perl's C<stat> or C<lstat> in void context. The callback will
187	be called after the stat and the results will be available using C<stat _>	459	be called after the stat and the results will be available using C<stat _>
188	or C<-s _> etc...	460	or C<-s _> etc...
189		461
…		…
199	aio_stat "/etc/passwd", sub {	471	aio_stat "/etc/passwd", sub {
200	$_[0] and die "stat failed: $!";	472	$_[0] and die "stat failed: $!";
201	print "size is ", -s _, "\n";	473	print "size is ", -s _, "\n";
202	};	474	};
203		475
		476
		477	=item aio_utime $fh_or_path, $atime, $mtime, $callback->($status)
		478
		479	Works like perl's C<utime> function (including the special case of $atime
		480	and $mtime being undef). Fractional times are supported if the underlying
		481	syscalls support them.
		482
		483	When called with a pathname, uses utimes(2) if available, otherwise
		484	utime(2). If called on a file descriptor, uses futimes(2) if available,
		485	otherwise returns ENOSYS, so this is not portable.
		486
		487	Examples:
		488
		489	# set atime and mtime to current time (basically touch(1)):
		490	aio_utime "path", undef, undef;
		491	# set atime to current time and mtime to beginning of the epoch:
		492	aio_utime "path", time, undef; # undef==0
		493
		494
		495	=item aio_chown $fh_or_path, $uid, $gid, $callback->($status)
		496
		497	Works like perl's C<chown> function, except that C<undef> for either $uid
		498	or $gid is being interpreted as "do not change" (but -1 can also be used).
		499
		500	Examples:
		501
		502	# same as "chown root path" in the shell:
		503	aio_chown "path", 0, -1;
		504	# same as above:
		505	aio_chown "path", 0, undef;
		506
		507
		508	=item aio_truncate $fh_or_path, $offset, $callback->($status)
		509
		510	Works like truncate(2) or ftruncate(2).
		511
		512
		513	=item aio_chmod $fh_or_path, $mode, $callback->($status)
		514
		515	Works like perl's C<chmod> function.
		516
		517
204	=item aio_unlink $pathname, $callback	518	=item aio_unlink $pathname, $callback->($status)
205		519
206	Asynchronously unlink (delete) a file and call the callback with the	520	Asynchronously unlink (delete) a file and call the callback with the
207	result code.	521	result code.
208		522
		523
		524	=item aio_mknod $path, $mode, $dev, $callback->($status)
		525
		526	[EXPERIMENTAL]
		527
		528	Asynchronously create a device node (or fifo). See mknod(2).
		529
		530	The only (POSIX-) portable way of calling this function is:
		531
		532	aio_mknod $path, IO::AIO::S_IFIFO | $mode, 0, sub { ...
		533
		534
		535	=item aio_link $srcpath, $dstpath, $callback->($status)
		536
		537	Asynchronously create a new link to the existing object at C<$srcpath> at
		538	the path C<$dstpath> and call the callback with the result code.
		539
		540
		541	=item aio_symlink $srcpath, $dstpath, $callback->($status)
		542
		543	Asynchronously create a new symbolic link to the existing object at C<$srcpath> at
		544	the path C<$dstpath> and call the callback with the result code.
		545
		546
		547	=item aio_readlink $path, $callback->($link)
		548
		549	Asynchronously read the symlink specified by C<$path> and pass it to
		550	the callback. If an error occurs, nothing or undef gets passed to the
		551	callback.
		552
		553
		554	=item aio_rename $srcpath, $dstpath, $callback->($status)
		555
		556	Asynchronously rename the object at C<$srcpath> to C<$dstpath>, just as
		557	rename(2) and call the callback with the result code.
		558
		559
		560	=item aio_mkdir $pathname, $mode, $callback->($status)
		561
		562	Asynchronously mkdir (create) a directory and call the callback with
		563	the result code. C<$mode> will be modified by the umask at the time the
		564	request is executed, so do not change your umask.
		565
		566
		567	=item aio_rmdir $pathname, $callback->($status)
		568
		569	Asynchronously rmdir (delete) a directory and call the callback with the
		570	result code.
		571
		572
		573	=item aio_readdir $pathname, $callback->($entries)
		574
		575	Unlike the POSIX call of the same name, C<aio_readdir> reads an entire
		576	directory (i.e. opendir + readdir + closedir). The entries will not be
		577	sorted, and will B<NOT> include the C<.> and C<..> entries.
		578
		579	The callback a single argument which is either C<undef> or an array-ref
		580	with the filenames.
		581
		582
		583	=item aio_load $path, $data, $callback->($status)
		584
		585	This is a composite request that tries to fully load the given file into
		586	memory. Status is the same as with aio_read.
		587
		588	=cut
		589
		590	sub aio_load($$;$) {
		591	aio_block {
		592	my ($path, undef, $cb) = @_;
		593	my $data = \$_[1];
		594
		595	my $pri = aioreq_pri;
		596	my $grp = aio_group $cb;
		597
		598	aioreq_pri $pri;
		599	add $grp aio_open $path, O_RDONLY, 0, sub {
		600	my $fh = shift
		601	or return $grp->result (-1);
		602
		603	aioreq_pri $pri;
		604	add $grp aio_read $fh, 0, (-s $fh), $$data, 0, sub {
		605	$grp->result ($_[0]);
		606	};
		607	};
		608
		609	$grp
		610	}
		611	}
		612
		613	=item aio_copy $srcpath, $dstpath, $callback->($status)
		614
		615	Try to copy the I<file> (directories not supported as either source or
		616	destination) from C<$srcpath> to C<$dstpath> and call the callback with
		617	the C<0> (error) or C<-1> ok.
		618
		619	This is a composite request that it creates the destination file with
		620	mode 0200 and copies the contents of the source file into it using
		621	C<aio_sendfile>, followed by restoring atime, mtime, access mode and
		622	uid/gid, in that order.
		623
		624	If an error occurs, the partial destination file will be unlinked, if
		625	possible, except when setting atime, mtime, access mode and uid/gid, where
		626	errors are being ignored.
		627
		628	=cut
		629
		630	sub aio_copy($$;$) {
		631	aio_block {
		632	my ($src, $dst, $cb) = @_;
		633
		634	my $pri = aioreq_pri;
		635	my $grp = aio_group $cb;
		636
		637	aioreq_pri $pri;
		638	add $grp aio_open $src, O_RDONLY, 0, sub {
		639	if (my $src_fh = $_[0]) {
		640	my @stat = stat $src_fh;
		641
		642	aioreq_pri $pri;
		643	add $grp aio_open $dst, O_CREAT | O_WRONLY | O_TRUNC, 0200, sub {
		644	if (my $dst_fh = $_[0]) {
		645	aioreq_pri $pri;
		646	add $grp aio_sendfile $dst_fh, $src_fh, 0, $stat[7], sub {
		647	if ($_[0] == $stat[7]) {
		648	$grp->result (0);
		649	close $src_fh;
		650
		651	# those should not normally block. should. should.
		652	utime $stat[8], $stat[9], $dst;
		653	chmod $stat[2] & 07777, $dst_fh;
		654	chown $stat[4], $stat[5], $dst_fh;
		655	close $dst_fh;
		656	} else {
		657	$grp->result (-1);
		658	close $src_fh;
		659	close $dst_fh;
		660
		661	aioreq $pri;
		662	add $grp aio_unlink $dst;
		663	}
		664	};
		665	} else {
		666	$grp->result (-1);
		667	}
		668	},
		669
		670	} else {
		671	$grp->result (-1);
		672	}
		673	};
		674
		675	$grp
		676	}
		677	}
		678
		679	=item aio_move $srcpath, $dstpath, $callback->($status)
		680
		681	Try to move the I<file> (directories not supported as either source or
		682	destination) from C<$srcpath> to C<$dstpath> and call the callback with
		683	the C<0> (error) or C<-1> ok.
		684
		685	This is a composite request that tries to rename(2) the file first. If
		686	rename files with C<EXDEV>, it copies the file with C<aio_copy> and, if
		687	that is successful, unlinking the C<$srcpath>.
		688
		689	=cut
		690
		691	sub aio_move($$;$) {
		692	aio_block {
		693	my ($src, $dst, $cb) = @_;
		694
		695	my $pri = aioreq_pri;
		696	my $grp = aio_group $cb;
		697
		698	aioreq_pri $pri;
		699	add $grp aio_rename $src, $dst, sub {
		700	if ($_[0] && $! == EXDEV) {
		701	aioreq_pri $pri;
		702	add $grp aio_copy $src, $dst, sub {
		703	$grp->result ($_[0]);
		704
		705	if (!$_[0]) {
		706	aioreq_pri $pri;
		707	add $grp aio_unlink $src;
		708	}
		709	};
		710	} else {
		711	$grp->result ($_[0]);
		712	}
		713	};
		714
		715	$grp
		716	}
		717	}
		718
		719	=item aio_scandir $path, $maxreq, $callback->($dirs, $nondirs)
		720
		721	Scans a directory (similar to C<aio_readdir>) but additionally tries to
		722	efficiently separate the entries of directory C<$path> into two sets of
		723	names, directories you can recurse into (directories), and ones you cannot
		724	recurse into (everything else, including symlinks to directories).
		725
		726	C<aio_scandir> is a composite request that creates of many sub requests_
		727	C<$maxreq> specifies the maximum number of outstanding aio requests that
		728	this function generates. If it is C<< <= 0 >>, then a suitable default
		729	will be chosen (currently 4).
		730
		731	On error, the callback is called without arguments, otherwise it receives
		732	two array-refs with path-relative entry names.
		733
		734	Example:
		735
		736	aio_scandir $dir, 0, sub {
		737	my ($dirs, $nondirs) = @_;
		738	print "real directories: @$dirs\n";
		739	print "everything else: @$nondirs\n";
		740	};
		741
		742	Implementation notes.
		743
		744	The C<aio_readdir> cannot be avoided, but C<stat()>'ing every entry can.
		745
		746	After reading the directory, the modification time, size etc. of the
		747	directory before and after the readdir is checked, and if they match (and
		748	isn't the current time), the link count will be used to decide how many
		749	entries are directories (if >= 2). Otherwise, no knowledge of the number
		750	of subdirectories will be assumed.
		751
		752	Then entries will be sorted into likely directories (everything without
		753	a non-initial dot currently) and likely non-directories (everything
		754	else). Then every entry plus an appended C</.> will be C<stat>'ed,
		755	likely directories first. If that succeeds, it assumes that the entry
		756	is a directory or a symlink to directory (which will be checked
		757	seperately). This is often faster than stat'ing the entry itself because
		758	filesystems might detect the type of the entry without reading the inode
		759	data (e.g. ext2fs filetype feature).
		760
		761	If the known number of directories (link count - 2) has been reached, the
		762	rest of the entries is assumed to be non-directories.
		763
		764	This only works with certainty on POSIX (= UNIX) filesystems, which
		765	fortunately are the vast majority of filesystems around.
		766
		767	It will also likely work on non-POSIX filesystems with reduced efficiency
		768	as those tend to return 0 or 1 as link counts, which disables the
		769	directory counting heuristic.
		770
		771	=cut
		772
		773	sub aio_scandir($$;$) {
		774	aio_block {
		775	my ($path, $maxreq, $cb) = @_;
		776
		777	my $pri = aioreq_pri;
		778
		779	my $grp = aio_group $cb;
		780
		781	$maxreq = 4 if $maxreq <= 0;
		782
		783	# stat once
		784	aioreq_pri $pri;
		785	add $grp aio_stat $path, sub {
		786	return $grp->result () if $_[0];
		787	my $now = time;
		788	my $hash1 = join ":", (stat _)[0,1,3,7,9];
		789
		790	# read the directory entries
		791	aioreq_pri $pri;
		792	add $grp aio_readdir $path, sub {
		793	my $entries = shift
		794	or return $grp->result ();
		795
		796	# stat the dir another time
		797	aioreq_pri $pri;
		798	add $grp aio_stat $path, sub {
		799	my $hash2 = join ":", (stat _)[0,1,3,7,9];
		800
		801	my $ndirs;
		802
		803	# take the slow route if anything looks fishy
		804	if ($hash1 ne $hash2 or (stat _)[9] == $now) {
		805	$ndirs = -1;
		806	} else {
		807	# if nlink == 2, we are finished
		808	# on non-posix-fs's, we rely on nlink < 2
		809	$ndirs = (stat _)[3] - 2
		810	or return $grp->result ([], $entries);
		811	}
		812
		813	# sort into likely dirs and likely nondirs
		814	# dirs == files without ".", short entries first
		815	$entries = [map $_->[0],
		816	sort { $b->[1] cmp $a->[1] }
		817	map [$_, sprintf "%s%04d", (/.\./ ? "1" : "0"), length],
		818	@$entries];
		819
		820	my (@dirs, @nondirs);
		821
		822	my $statgrp = add $grp aio_group sub {
		823	$grp->result (\@dirs, \@nondirs);
		824	};
		825
		826	limit $statgrp $maxreq;
		827	feed $statgrp sub {
		828	return unless @$entries;
		829	my $entry = pop @$entries;
		830
		831	aioreq_pri $pri;
		832	add $statgrp aio_stat "$path/$entry/.", sub {
		833	if ($_[0] < 0) {
		834	push @nondirs, $entry;
		835	} else {
		836	# need to check for real directory
		837	aioreq_pri $pri;
		838	add $statgrp aio_lstat "$path/$entry", sub {
		839	if (-d _) {
		840	push @dirs, $entry;
		841
		842	unless (--$ndirs) {
		843	push @nondirs, @$entries;
		844	feed $statgrp;
		845	}
		846	} else {
		847	push @nondirs, $entry;
		848	}
		849	}
		850	}
		851	};
		852	};
		853	};
		854	};
		855	};
		856
		857	$grp
		858	}
		859	}
		860
		861	=item aio_rmtree $path, $callback->($status)
		862
		863	Delete a directory tree starting (and including) C<$path>, return the
		864	status of the final C<rmdir> only. This is a composite request that
		865	uses C<aio_scandir> to recurse into and rmdir directories, and unlink
		866	everything else.
		867
		868	=cut
		869
		870	sub aio_rmtree;
		871	sub aio_rmtree($;$) {
		872	aio_block {
		873	my ($path, $cb) = @_;
		874
		875	my $pri = aioreq_pri;
		876	my $grp = aio_group $cb;
		877
		878	aioreq_pri $pri;
		879	add $grp aio_scandir $path, 0, sub {
		880	my ($dirs, $nondirs) = @_;
		881
		882	my $dirgrp = aio_group sub {
		883	add $grp aio_rmdir $path, sub {
		884	$grp->result ($_[0]);
		885	};
		886	};
		887
		888	(aioreq_pri $pri), add $dirgrp aio_rmtree "$path/$_" for @$dirs;
		889	(aioreq_pri $pri), add $dirgrp aio_unlink "$path/$_" for @$nondirs;
		890
		891	add $grp $dirgrp;
		892	};
		893
		894	$grp
		895	}
		896	}
		897
209	=item aio_fsync $fh, $callback	898	=item aio_fsync $fh, $callback->($status)
210		899
211	Asynchronously call fsync on the given filehandle and call the callback	900	Asynchronously call fsync on the given filehandle and call the callback
212	with the fsync result code.	901	with the fsync result code.
213		902
214	=item aio_fdatasync $fh, $callback	903	=item aio_fdatasync $fh, $callback->($status)
215		904
216	Asynchronously call fdatasync on the given filehandle and call the	905	Asynchronously call fdatasync on the given filehandle and call the
217	callback with the fdatasync result code.	906	callback with the fdatasync result code.
218		907
		908	If this call isn't available because your OS lacks it or it couldn't be
		909	detected, it will be emulated by calling C<fsync> instead.
		910
		911	=item aio_group $callback->(...)
		912
		913	This is a very special aio request: Instead of doing something, it is a
		914	container for other aio requests, which is useful if you want to bundle
		915	many requests into a single, composite, request with a definite callback
		916	and the ability to cancel the whole request with its subrequests.
		917
		918	Returns an object of class L<IO::AIO::GRP>. See its documentation below
		919	for more info.
		920
		921	Example:
		922
		923	my $grp = aio_group sub {
		924	print "all stats done\n";
		925	};
		926
		927	add $grp
		928	(aio_stat ...),
		929	(aio_stat ...),
		930	...;
		931
		932	=item aio_nop $callback->()
		933
		934	This is a special request - it does nothing in itself and is only used for
		935	side effects, such as when you want to add a dummy request to a group so
		936	that finishing the requests in the group depends on executing the given
		937	code.
		938
		939	While this request does nothing, it still goes through the execution
		940	phase and still requires a worker thread. Thus, the callback will not
		941	be executed immediately but only after other requests in the queue have
		942	entered their execution phase. This can be used to measure request
		943	latency.
		944
		945	=item IO::AIO::aio_busy $fractional_seconds, $callback->() *NOT EXPORTED*
		946
		947	Mainly used for debugging and benchmarking, this aio request puts one of
		948	the request workers to sleep for the given time.
		949
		950	While it is theoretically handy to have simple I/O scheduling requests
		951	like sleep and file handle readable/writable, the overhead this creates is
		952	immense (it blocks a thread for a long time) so do not use this function
		953	except to put your application under artificial I/O pressure.
		954
		955	=back
		956
		957	=head2 IO::AIO::REQ CLASS
		958
		959	All non-aggregate C<aio_*> functions return an object of this class when
		960	called in non-void context.
		961
		962	=over 4
		963
		964	=item cancel $req
		965
		966	Cancels the request, if possible. Has the effect of skipping execution
		967	when entering the B<execute> state and skipping calling the callback when
		968	entering the the B<result> state, but will leave the request otherwise
		969	untouched. That means that requests that currently execute will not be
		970	stopped and resources held by the request will not be freed prematurely.
		971
		972	=item cb $req $callback->(...)
		973
		974	Replace (or simply set) the callback registered to the request.
		975
		976	=back
		977
		978	=head2 IO::AIO::GRP CLASS
		979
		980	This class is a subclass of L<IO::AIO::REQ>, so all its methods apply to
		981	objects of this class, too.
		982
		983	A IO::AIO::GRP object is a special request that can contain multiple other
		984	aio requests.
		985
		986	You create one by calling the C<aio_group> constructing function with a
		987	callback that will be called when all contained requests have entered the
		988	C<done> state:
		989
		990	my $grp = aio_group sub {
		991	print "all requests are done\n";
		992	};
		993
		994	You add requests by calling the C<add> method with one or more
		995	C<IO::AIO::REQ> objects:
		996
		997	$grp->add (aio_unlink "...");
		998
		999	add $grp aio_stat "...", sub {
		1000	$_[0] or return $grp->result ("error");
		1001
		1002	# add another request dynamically, if first succeeded
		1003	add $grp aio_open "...", sub {
		1004	$grp->result ("ok");
		1005	};
		1006	};
		1007
		1008	This makes it very easy to create composite requests (see the source of
		1009	C<aio_move> for an application) that work and feel like simple requests.
		1010
		1011	=over 4
		1012
		1013	=item * The IO::AIO::GRP objects will be cleaned up during calls to
		1014	C<IO::AIO::poll_cb>, just like any other request.
		1015
		1016	=item * They can be canceled like any other request. Canceling will cancel not
		1017	only the request itself, but also all requests it contains.
		1018
		1019	=item * They can also can also be added to other IO::AIO::GRP objects.
		1020
		1021	=item * You must not add requests to a group from within the group callback (or
		1022	any later time).
		1023
		1024	=back
		1025
		1026	Their lifetime, simplified, looks like this: when they are empty, they
		1027	will finish very quickly. If they contain only requests that are in the
		1028	C<done> state, they will also finish. Otherwise they will continue to
		1029	exist.
		1030
		1031	That means after creating a group you have some time to add requests. And
		1032	in the callbacks of those requests, you can add further requests to the
		1033	group. And only when all those requests have finished will the the group
		1034	itself finish.
		1035
		1036	=over 4
		1037
		1038	=item add $grp ...
		1039
		1040	=item $grp->add (...)
		1041
		1042	Add one or more requests to the group. Any type of L<IO::AIO::REQ> can
		1043	be added, including other groups, as long as you do not create circular
		1044	dependencies.
		1045
		1046	Returns all its arguments.
		1047
		1048	=item $grp->cancel_subs
		1049
		1050	Cancel all subrequests and clears any feeder, but not the group request
		1051	itself. Useful when you queued a lot of events but got a result early.
		1052
		1053	=item $grp->result (...)
		1054
		1055	Set the result value(s) that will be passed to the group callback when all
		1056	subrequests have finished and set thre groups errno to the current value
		1057	of errno (just like calling C<errno> without an error number). By default,
		1058	no argument will be passed and errno is zero.
		1059
		1060	=item $grp->errno ([$errno])
		1061
		1062	Sets the group errno value to C<$errno>, or the current value of errno
		1063	when the argument is missing.
		1064
		1065	Every aio request has an associated errno value that is restored when
		1066	the callback is invoked. This method lets you change this value from its
		1067	default (0).
		1068
		1069	Calling C<result> will also set errno, so make sure you either set C<$!>
		1070	before the call to C<result>, or call c<errno> after it.
		1071
		1072	=item feed $grp $callback->($grp)
		1073
		1074	Sets a feeder/generator on this group: every group can have an attached
		1075	generator that generates requests if idle. The idea behind this is that,
		1076	although you could just queue as many requests as you want in a group,
		1077	this might starve other requests for a potentially long time. For
		1078	example, C<aio_scandir> might generate hundreds of thousands C<aio_stat>
		1079	requests, delaying any later requests for a long time.
		1080
		1081	To avoid this, and allow incremental generation of requests, you can
		1082	instead a group and set a feeder on it that generates those requests. The
		1083	feed callback will be called whenever there are few enough (see C<limit>,
		1084	below) requests active in the group itself and is expected to queue more
		1085	requests.
		1086
		1087	The feed callback can queue as many requests as it likes (i.e. C<add> does
		1088	not impose any limits).
		1089
		1090	If the feed does not queue more requests when called, it will be
		1091	automatically removed from the group.
		1092
		1093	If the feed limit is C<0>, it will be set to C<2> automatically.
		1094
		1095	Example:
		1096
		1097	# stat all files in @files, but only ever use four aio requests concurrently:
		1098
		1099	my $grp = aio_group sub { print "finished\n" };
		1100	limit $grp 4;
		1101	feed $grp sub {
		1102	my $file = pop @files
		1103	or return;
		1104
		1105	add $grp aio_stat $file, sub { ... };
		1106	};
		1107
		1108	=item limit $grp $num
		1109
		1110	Sets the feeder limit for the group: The feeder will be called whenever
		1111	the group contains less than this many requests.
		1112
		1113	Setting the limit to C<0> will pause the feeding process.
		1114
		1115	=back
		1116
		1117	=head2 SUPPORT FUNCTIONS
		1118
		1119	=head3 EVENT PROCESSING AND EVENT LOOP INTEGRATION
		1120
		1121	=over 4
		1122
		1123	=item $fileno = IO::AIO::poll_fileno
		1124
		1125	Return the I<request result pipe file descriptor>. This filehandle must be
		1126	polled for reading by some mechanism outside this module (e.g. Event or
		1127	select, see below or the SYNOPSIS). If the pipe becomes readable you have
		1128	to call C<poll_cb> to check the results.
		1129
		1130	See C<poll_cb> for an example.
		1131
		1132	=item IO::AIO::poll_cb
		1133
		1134	Process some outstanding events on the result pipe. You have to call this
		1135	regularly. Returns the number of events processed. Returns immediately
		1136	when no events are outstanding. The amount of events processed depends on
		1137	the settings of C<IO::AIO::max_poll_req> and C<IO::AIO::max_poll_time>.
		1138
		1139	If not all requests were processed for whatever reason, the filehandle
		1140	will still be ready when C<poll_cb> returns.
		1141
		1142	Example: Install an Event watcher that automatically calls
		1143	IO::AIO::poll_cb with high priority:
		1144
		1145	Event->io (fd => IO::AIO::poll_fileno,
		1146	poll => 'r', async => 1,
		1147	cb => \&IO::AIO::poll_cb);
		1148
		1149	=item IO::AIO::max_poll_reqs $nreqs
		1150
		1151	=item IO::AIO::max_poll_time $seconds
		1152
		1153	These set the maximum number of requests (default C<0>, meaning infinity)
		1154	that are being processed by C<IO::AIO::poll_cb> in one call, respectively
		1155	the maximum amount of time (default C<0>, meaning infinity) spent in
		1156	C<IO::AIO::poll_cb> to process requests (more correctly the mininum amount
		1157	of time C<poll_cb> is allowed to use).
		1158
		1159	Setting C<max_poll_time> to a non-zero value creates an overhead of one
		1160	syscall per request processed, which is not normally a problem unless your
		1161	callbacks are really really fast or your OS is really really slow (I am
		1162	not mentioning Solaris here). Using C<max_poll_reqs> incurs no overhead.
		1163
		1164	Setting these is useful if you want to ensure some level of
		1165	interactiveness when perl is not fast enough to process all requests in
		1166	time.
		1167
		1168	For interactive programs, values such as C<0.01> to C<0.1> should be fine.
		1169
		1170	Example: Install an Event watcher that automatically calls
		1171	IO::AIO::poll_cb with low priority, to ensure that other parts of the
		1172	program get the CPU sometimes even under high AIO load.
		1173
		1174	# try not to spend much more than 0.1s in poll_cb
		1175	IO::AIO::max_poll_time 0.1;
		1176
		1177	# use a low priority so other tasks have priority
		1178	Event->io (fd => IO::AIO::poll_fileno,
		1179	poll => 'r', nice => 1,
		1180	cb => &IO::AIO::poll_cb);
		1181
		1182	=item IO::AIO::poll_wait
		1183
		1184	If there are any outstanding requests and none of them in the result
		1185	phase, wait till the result filehandle becomes ready for reading (simply
		1186	does a C<select> on the filehandle. This is useful if you want to
		1187	synchronously wait for some requests to finish).
		1188
		1189	See C<nreqs> for an example.
		1190
		1191	=item IO::AIO::poll
		1192
		1193	Waits until some requests have been handled.
		1194
		1195	Returns the number of requests processed, but is otherwise strictly
		1196	equivalent to:
		1197
		1198	IO::AIO::poll_wait, IO::AIO::poll_cb
		1199
		1200	=item IO::AIO::flush
		1201
		1202	Wait till all outstanding AIO requests have been handled.
		1203
		1204	Strictly equivalent to:
		1205
		1206	IO::AIO::poll_wait, IO::AIO::poll_cb
		1207	while IO::AIO::nreqs;
		1208
		1209	=back
		1210
		1211	=head3 CONTROLLING THE NUMBER OF THREADS
		1212
		1213	=over
		1214
		1215	=item IO::AIO::min_parallel $nthreads
		1216
		1217	Set the minimum number of AIO threads to C<$nthreads>. The current
		1218	default is C<8>, which means eight asynchronous operations can execute
		1219	concurrently at any one time (the number of outstanding requests,
		1220	however, is unlimited).
		1221
		1222	IO::AIO starts threads only on demand, when an AIO request is queued and
		1223	no free thread exists. Please note that queueing up a hundred requests can
		1224	create demand for a hundred threads, even if it turns out that everything
		1225	is in the cache and could have been processed faster by a single thread.
		1226
		1227	It is recommended to keep the number of threads relatively low, as some
		1228	Linux kernel versions will scale negatively with the number of threads
		1229	(higher parallelity => MUCH higher latency). With current Linux 2.6
		1230	versions, 4-32 threads should be fine.
		1231
		1232	Under most circumstances you don't need to call this function, as the
		1233	module selects a default that is suitable for low to moderate load.
		1234
		1235	=item IO::AIO::max_parallel $nthreads
		1236
		1237	Sets the maximum number of AIO threads to C<$nthreads>. If more than the
		1238	specified number of threads are currently running, this function kills
		1239	them. This function blocks until the limit is reached.
		1240
		1241	While C<$nthreads> are zero, aio requests get queued but not executed
		1242	until the number of threads has been increased again.
		1243
		1244	This module automatically runs C<max_parallel 0> at program end, to ensure
		1245	that all threads are killed and that there are no outstanding requests.
		1246
		1247	Under normal circumstances you don't need to call this function.
		1248
		1249	=item IO::AIO::max_idle $nthreads
		1250
		1251	Limit the number of threads (default: 4) that are allowed to idle (i.e.,
		1252	threads that did not get a request to process within 10 seconds). That
		1253	means if a thread becomes idle while C<$nthreads> other threads are also
		1254	idle, it will free its resources and exit.
		1255
		1256	This is useful when you allow a large number of threads (e.g. 100 or 1000)
		1257	to allow for extremely high load situations, but want to free resources
		1258	under normal circumstances (1000 threads can easily consume 30MB of RAM).
		1259
		1260	The default is probably ok in most situations, especially if thread
		1261	creation is fast. If thread creation is very slow on your system you might
		1262	want to use larger values.
		1263
		1264	=item $oldmaxreqs = IO::AIO::max_outstanding $maxreqs
		1265
		1266	This is a very bad function to use in interactive programs because it
		1267	blocks, and a bad way to reduce concurrency because it is inexact: Better
		1268	use an C<aio_group> together with a feed callback.
		1269
		1270	Sets the maximum number of outstanding requests to C<$nreqs>. If you
		1271	do queue up more than this number of requests, the next call to the
		1272	C<poll_cb> (and C<poll_some> and other functions calling C<poll_cb>)
		1273	function will block until the limit is no longer exceeded.
		1274
		1275	The default value is very large, so there is no practical limit on the
		1276	number of outstanding requests.
		1277
		1278	You can still queue as many requests as you want. Therefore,
		1279	C<max_oustsanding> is mainly useful in simple scripts (with low values) or
		1280	as a stop gap to shield against fatal memory overflow (with large values).
		1281
		1282	=back
		1283
		1284	=head3 STATISTICAL INFORMATION
		1285
		1286	=over
		1287
		1288	=item IO::AIO::nreqs
		1289
		1290	Returns the number of requests currently in the ready, execute or pending
		1291	states (i.e. for which their callback has not been invoked yet).
		1292
		1293	Example: wait till there are no outstanding requests anymore:
		1294
		1295	IO::AIO::poll_wait, IO::AIO::poll_cb
		1296	while IO::AIO::nreqs;
		1297
		1298	=item IO::AIO::nready
		1299
		1300	Returns the number of requests currently in the ready state (not yet
		1301	executed).
		1302
		1303	=item IO::AIO::npending
		1304
		1305	Returns the number of requests currently in the pending state (executed,
		1306	but not yet processed by poll_cb).
		1307
		1308	=back
		1309
219	=cut	1310	=cut
220		1311
221	min_parallel 4;	1312	min_parallel 8;
222		1313
223	END {	1314	END { flush }
224	max_parallel 0;
225	}
226		1315
227	1;	1316	1;
228		1317
229	=back	1318	=head2 FORK BEHAVIOUR
230		1319
		1320	This module should do "the right thing" when the process using it forks:
		1321
		1322	Before the fork, IO::AIO enters a quiescent state where no requests
		1323	can be added in other threads and no results will be processed. After
		1324	the fork the parent simply leaves the quiescent state and continues
		1325	request/result processing, while the child frees the request/result queue
		1326	(so that the requests started before the fork will only be handled in the
		1327	parent). Threads will be started on demand until the limit set in the
		1328	parent process has been reached again.
		1329
		1330	In short: the parent will, after a short pause, continue as if fork had
		1331	not been called, while the child will act as if IO::AIO has not been used
		1332	yet.
		1333
		1334	=head2 MEMORY USAGE
		1335
		1336	Per-request usage:
		1337
		1338	Each aio request uses - depending on your architecture - around 100-200
		1339	bytes of memory. In addition, stat requests need a stat buffer (possibly
		1340	a few hundred bytes), readdir requires a result buffer and so on. Perl
		1341	scalars and other data passed into aio requests will also be locked and
		1342	will consume memory till the request has entered the done state.
		1343
		1344	This is not awfully much, so queuing lots of requests is not usually a
		1345	problem.
		1346
		1347	Per-thread usage:
		1348
		1349	In the execution phase, some aio requests require more memory for
		1350	temporary buffers, and each thread requires a stack and other data
		1351	structures (usually around 16k-128k, depending on the OS).
		1352
231	=head1 BUGS	1353	=head1 KNOWN BUGS
232		1354
233	- aio_open gives a fd, but all other functions expect a perl filehandle.	1355	Known bugs will be fixed in the next release.
234		1356
235	=head1 SEE ALSO	1357	=head1 SEE ALSO
236		1358
237	L<Coro>, L<Linux::AIO>.	1359	L<Coro::AIO>.
238		1360
239	=head1 AUTHOR	1361	=head1 AUTHOR
240		1362
241	Marc Lehmann <schmorp@schmorp.de>	1363	Marc Lehmann <schmorp@schmorp.de>
242	http://home.schmorp.de/	1364	http://home.schmorp.de/

Diff Legend

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