ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/IO-AIO/AIO.pm

Comparing IO-AIO/AIO.pm (file contents):
Revision 1.2 by root, Sun Jul 10 18:16:49 2005 UTC vs.
Revision 1.253 by root, Thu Jun 25 13:34:28 2015 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", IO::AIO::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
9	=head1 DESCRIPTION	31	=head1 DESCRIPTION
10		32
11	This module implements asynchronous I/O using whatever means your	33	This module implements asynchronous I/O using whatever means your
12	operating system supports.	34	operating system supports. It is implemented as an interface to C<libeio>
		35	(L<http://software.schmorp.de/pkg/libeio.html>).
13		36
		37	Asynchronous means that operations that can normally block your program
		38	(e.g. reading from disk) will be done asynchronously: the operation
		39	will still block, but you can do something else in the meantime. This
		40	is extremely useful for programs that need to stay interactive even
		41	when doing heavy I/O (GUI programs, high performance network servers
		42	etc.), but can also be used to easily do operations in parallel that are
		43	normally done sequentially, e.g. stat'ing many files, which is much faster
		44	on a RAID volume or over NFS when you do a number of stat operations
		45	concurrently.
		46
		47	While most of this works on all types of file descriptors (for
		48	example sockets), using these functions on file descriptors that
		49	support nonblocking operation (again, sockets, pipes etc.) is
		50	very inefficient. Use an event loop for that (such as the L<EV>
		51	module): IO::AIO will naturally fit into such an event loop itself.
		52
14	Currently, a number of threads are started that execute your read/writes	53	In this version, a number of threads are started that execute your
15	and signal their completion. You don't need thread support in your libc or	54	requests and signal their completion. You don't need thread support
16	perl, and the threads created by this module will not be visible to the	55	in perl, and the threads created by this module will not be visible
17	pthreads library. In the future, this module might make use of the native	56	to perl. In the future, this module might make use of the native aio
18	aio functions available on many operating systems. However, they are often	57	functions available on many operating systems. However, they are often
19	not well-supported (Linux doesn't allow them on normal files currently,	58	not well-supported or restricted (GNU/Linux doesn't allow them on normal
20	for example), and they would only support aio_read and aio_write, so the	59	files currently, for example), and they would only support aio_read and
21	remaining functionality would have to be implemented using threads anyway.	60	aio_write, so the remaining functionality would have to be implemented
		61	using threads anyway.
22		62
23	Although the module will work with in the presence of other threads, it is	63	Although the module will work in the presence of other (Perl-) threads,
24	currently not reentrant, so use appropriate locking yourself.	64	it is currently not reentrant in any way, so use appropriate locking
		65	yourself, always call C<poll_cb> from within the same thread, or never
		66	call C<poll_cb> (or other C<aio_> functions) recursively.
		67
		68	=head2 EXAMPLE
		69
		70	This is a simple example that uses the EV module and loads
		71	F</etc/passwd> asynchronously:
		72
		73	use EV;
		74	use IO::AIO;
		75
		76	# register the IO::AIO callback with EV
		77	my $aio_w = EV::io IO::AIO::poll_fileno, EV::READ, \&IO::AIO::poll_cb;
		78
		79	# queue the request to open /etc/passwd
		80	aio_open "/etc/passwd", IO::AIO::O_RDONLY, 0, sub {
		81	my $fh = shift
		82	or die "error while opening: $!";
		83
		84	# stat'ing filehandles is generally non-blocking
		85	my $size = -s $fh;
		86
		87	# queue a request to read the file
		88	my $contents;
		89	aio_read $fh, 0, $size, $contents, 0, sub {
		90	$_[0] == $size
		91	or die "short read: $!";
		92
		93	close $fh;
		94
		95	# file contents now in $contents
		96	print $contents;
		97
		98	# exit event loop and program
		99	EV::unloop;
		100	};
		101	};
		102
		103	# possibly queue up other requests, or open GUI windows,
		104	# check for sockets etc. etc.
		105
		106	# process events as long as there are some:
		107	EV::loop;
		108
		109	=head1 REQUEST ANATOMY AND LIFETIME
		110
		111	Every C<aio_*> function creates a request. which is a C data structure not
		112	directly visible to Perl.
		113
		114	If called in non-void context, every request function returns a Perl
		115	object representing the request. In void context, nothing is returned,
		116	which saves a bit of memory.
		117
		118	The perl object is a fairly standard ref-to-hash object. The hash contents
		119	are not used by IO::AIO so you are free to store anything you like in it.
		120
		121	During their existance, aio requests travel through the following states,
		122	in order:
		123
		124	=over 4
		125
		126	=item ready
		127
		128	Immediately after a request is created it is put into the ready state,
		129	waiting for a thread to execute it.
		130
		131	=item execute
		132
		133	A thread has accepted the request for processing and is currently
		134	executing it (e.g. blocking in read).
		135
		136	=item pending
		137
		138	The request has been executed and is waiting for result processing.
		139
		140	While request submission and execution is fully asynchronous, result
		141	processing is not and relies on the perl interpreter calling C<poll_cb>
		142	(or another function with the same effect).
		143
		144	=item result
		145
		146	The request results are processed synchronously by C<poll_cb>.
		147
		148	The C<poll_cb> function will process all outstanding aio requests by
		149	calling their callbacks, freeing memory associated with them and managing
		150	any groups they are contained in.
		151
		152	=item done
		153
		154	Request has reached the end of its lifetime and holds no resources anymore
		155	(except possibly for the Perl object, but its connection to the actual
		156	aio request is severed and calling its methods will either do nothing or
		157	result in a runtime error).
		158
		159	=back
		160
		161	=cut
		162
		163	package IO::AIO;
		164
		165	use Carp ();
		166
		167	use common::sense;
		168
		169	use base 'Exporter';
		170
		171	BEGIN {
		172	our $VERSION = 4.32;
		173
		174	our @AIO_REQ = qw(aio_sendfile aio_seek aio_read aio_write aio_open aio_close
		175	aio_stat aio_lstat aio_unlink aio_rmdir aio_readdir aio_readdirx
		176	aio_scandir aio_symlink aio_readlink aio_realpath aio_sync
		177	aio_fsync aio_syncfs aio_fdatasync aio_sync_file_range aio_allocate
		178	aio_pathsync aio_readahead aio_fiemap
		179	aio_rename aio_link aio_move aio_copy aio_group
		180	aio_nop aio_mknod aio_load aio_rmtree aio_mkdir aio_chown
		181	aio_chmod aio_utime aio_truncate
		182	aio_msync aio_mtouch aio_mlock aio_mlockall
		183	aio_statvfs
		184	aio_wd);
		185
		186	our @EXPORT = (@AIO_REQ, qw(aioreq_pri aioreq_nice));
		187	our @EXPORT_OK = qw(poll_fileno poll_cb poll_wait flush
		188	min_parallel max_parallel max_idle idle_timeout
		189	nreqs nready npending nthreads
		190	max_poll_time max_poll_reqs
		191	sendfile fadvise madvise
		192	mmap munmap munlock munlockall);
		193
		194	push @AIO_REQ, qw(aio_busy); # not exported
		195
		196	@IO::AIO::GRP::ISA = 'IO::AIO::REQ';
		197
		198	require XSLoader;
		199	XSLoader::load ("IO::AIO", $VERSION);
		200	}
		201
		202	=head1 FUNCTIONS
		203
		204	=head2 QUICK OVERVIEW
		205
		206	This section simply lists the prototypes most of the functions for
		207	quick reference. See the following sections for function-by-function
		208	documentation.
		209
		210	aio_wd $pathname, $callback->($wd)
		211	aio_open $pathname, $flags, $mode, $callback->($fh)
		212	aio_close $fh, $callback->($status)
		213	aio_seek $fh,$offset,$whence, $callback->($offs)
		214	aio_read $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
		215	aio_write $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
		216	aio_sendfile $out_fh, $in_fh, $in_offset, $length, $callback->($retval)
		217	aio_readahead $fh,$offset,$length, $callback->($retval)
		218	aio_stat $fh_or_path, $callback->($status)
		219	aio_lstat $fh, $callback->($status)
		220	aio_statvfs $fh_or_path, $callback->($statvfs)
		221	aio_utime $fh_or_path, $atime, $mtime, $callback->($status)
		222	aio_chown $fh_or_path, $uid, $gid, $callback->($status)
		223	aio_chmod $fh_or_path, $mode, $callback->($status)
		224	aio_truncate $fh_or_path, $offset, $callback->($status)
		225	aio_allocate $fh, $mode, $offset, $len, $callback->($status)
		226	aio_fiemap $fh, $start, $length, $flags, $count, $cb->(\@extents)
		227	aio_unlink $pathname, $callback->($status)
		228	aio_mknod $pathname, $mode, $dev, $callback->($status)
		229	aio_link $srcpath, $dstpath, $callback->($status)
		230	aio_symlink $srcpath, $dstpath, $callback->($status)
		231	aio_readlink $pathname, $callback->($link)
		232	aio_realpath $pathname, $callback->($path)
		233	aio_rename $srcpath, $dstpath, $callback->($status)
		234	aio_mkdir $pathname, $mode, $callback->($status)
		235	aio_rmdir $pathname, $callback->($status)
		236	aio_readdir $pathname, $callback->($entries)
		237	aio_readdirx $pathname, $flags, $callback->($entries, $flags)
		238	IO::AIO::READDIR_DENTS IO::AIO::READDIR_DIRS_FIRST
		239	IO::AIO::READDIR_STAT_ORDER IO::AIO::READDIR_FOUND_UNKNOWN
		240	aio_scandir $pathname, $maxreq, $callback->($dirs, $nondirs)
		241	aio_load $pathname, $data, $callback->($status)
		242	aio_copy $srcpath, $dstpath, $callback->($status)
		243	aio_move $srcpath, $dstpath, $callback->($status)
		244	aio_rmtree $pathname, $callback->($status)
		245	aio_sync $callback->($status)
		246	aio_syncfs $fh, $callback->($status)
		247	aio_fsync $fh, $callback->($status)
		248	aio_fdatasync $fh, $callback->($status)
		249	aio_sync_file_range $fh, $offset, $nbytes, $flags, $callback->($status)
		250	aio_pathsync $pathname, $callback->($status)
		251	aio_msync $scalar, $offset = 0, $length = undef, flags = 0, $callback->($status)
		252	aio_mtouch $scalar, $offset = 0, $length = undef, flags = 0, $callback->($status)
		253	aio_mlock $scalar, $offset = 0, $length = undef, $callback->($status)
		254	aio_mlockall $flags, $callback->($status)
		255	aio_group $callback->(...)
		256	aio_nop $callback->()
		257
		258	$prev_pri = aioreq_pri [$pri]
		259	aioreq_nice $pri_adjust
		260
		261	IO::AIO::poll_wait
		262	IO::AIO::poll_cb
		263	IO::AIO::poll
		264	IO::AIO::flush
		265	IO::AIO::max_poll_reqs $nreqs
		266	IO::AIO::max_poll_time $seconds
		267	IO::AIO::min_parallel $nthreads
		268	IO::AIO::max_parallel $nthreads
		269	IO::AIO::max_idle $nthreads
		270	IO::AIO::idle_timeout $seconds
		271	IO::AIO::max_outstanding $maxreqs
		272	IO::AIO::nreqs
		273	IO::AIO::nready
		274	IO::AIO::npending
		275
		276	IO::AIO::sendfile $ofh, $ifh, $offset, $count
		277	IO::AIO::fadvise $fh, $offset, $len, $advice
		278	IO::AIO::mmap $scalar, $length, $prot, $flags[, $fh[, $offset]]
		279	IO::AIO::munmap $scalar
		280	IO::AIO::madvise $scalar, $offset, $length, $advice
		281	IO::AIO::mprotect $scalar, $offset, $length, $protect
		282	IO::AIO::munlock $scalar, $offset = 0, $length = undef
		283	IO::AIO::munlockall
25		284
26	=head2 API NOTES	285	=head2 API NOTES
27		286
28	All the C<aio_*> calls are more or less thin wrappers around the syscall	287	All the C<aio_*> calls are more or less thin wrappers around the syscall
29	with the same name (sans C<aio_>). The arguments are similar or identical,	288	with the same name (sans C<aio_>). The arguments are similar or identical,
30	and they all accept an additional C<$callback> argument which must be	289	and they all accept an additional (and optional) C<$callback> argument
31	a code reference. This code reference will get called with the syscall	290	which must be a code reference. This code reference will be called after
		291	the syscall has been executed in an asynchronous fashion. The results
		292	of the request will be passed as arguments to the callback (and, if an
		293	error occured, in C<$!>) - for most requests the syscall return code (e.g.
32	return code (e.g. most syscalls return C<-1> on error, unlike perl, which	294	most syscalls return C<-1> on error, unlike perl, which usually delivers
33	usually delivers "false") as it's sole argument when the given syscall has	295	"false").
34	been executed asynchronously.
35		296
36	All functions that expect a filehandle will also accept a file descriptor.	297	Some requests (such as C<aio_readdir>) pass the actual results and
		298	communicate failures by passing C<undef>.
37		299
		300	All functions expecting a filehandle keep a copy of the filehandle
		301	internally until the request has finished.
		302
		303	All functions return request objects of type L<IO::AIO::REQ> that allow
		304	further manipulation of those requests while they are in-flight.
		305
38	The filenames you pass to these routines I<must> be absolute. The reason	306	The pathnames you pass to these routines I<should> be absolute. The
39	is that at the time the request is being executed, the current working	307	reason for this is that at the time the request is being executed, the
40	directory could have changed. Alternatively, you can make sure that you	308	current working directory could have changed. Alternatively, you can
41	never change the current working directory.	309	make sure that you never change the current working directory anywhere
		310	in the program and then use relative paths. You can also take advantage
		311	of IO::AIOs working directory abstraction, that lets you specify paths
		312	relative to some previously-opened "working directory object" - see the
		313	description of the C<IO::AIO::WD> class later in this document.
		314
		315	To encode pathnames as octets, either make sure you either: a) always pass
		316	in filenames you got from outside (command line, readdir etc.) without
		317	tinkering, b) are in your native filesystem encoding, c) use the Encode
		318	module and encode your pathnames to the locale (or other) encoding in
		319	effect in the user environment, d) use Glib::filename_from_unicode on
		320	unicode filenames or e) use something else to ensure your scalar has the
		321	correct contents.
		322
		323	This works, btw. independent of the internal UTF-8 bit, which IO::AIO
		324	handles correctly whether it is set or not.
		325
		326	=head2 AIO REQUEST FUNCTIONS
42		327
43	=over 4	328	=over 4
44		329
45	=cut	330	=item $prev_pri = aioreq_pri [$pri]
46		331
47	package IO::AIO;	332	Returns the priority value that would be used for the next request and, if
		333	C<$pri> is given, sets the priority for the next aio request.
48		334
49	use base 'Exporter';	335	The default priority is C<0>, the minimum and maximum priorities are C<-4>
		336	and C<4>, respectively. Requests with higher priority will be serviced
		337	first.
50		338
51	use Fcntl ();	339	The priority will be reset to C<0> after each call to one of the C<aio_*>
		340	functions.
52		341
53	BEGIN {	342	Example: open a file with low priority, then read something from it with
54	$VERSION = 0.1;	343	higher priority so the read request is serviced before other low priority
		344	open requests (potentially spamming the cache):
55		345
56	@EXPORT = qw(aio_read aio_write aio_open aio_close aio_stat aio_lstat aio_unlink	346	aioreq_pri -3;
57	aio_fsync aio_fdatasync aio_readahead);	347	aio_open ..., sub {
58	@EXPORT_OK = qw(poll_fileno poll_cb min_parallel max_parallel nreqs);	348	return unless $_[0];
59		349
60	require XSLoader;	350	aioreq_pri -2;
61	XSLoader::load IO::AIO, $VERSION;	351	aio_read $_[0], ..., sub {
62	}	352	...
		353	};
		354	};
63		355
64	=item IO::AIO::min_parallel $nthreads
65		356
66	Set the minimum number of AIO threads to C<$nthreads>. The default is	357	=item aioreq_nice $pri_adjust
67	C<1>, which means a single asynchronous operation can be done at one time
68	(the number of outstanding operations, however, is unlimited).
69		358
70	It is recommended to keep the number of threads low, as some linux	359	Similar to C<aioreq_pri>, but subtracts the given value from the current
71	kernel versions will scale negatively with the number of threads (higher	360	priority, so the effect is cumulative.
72	parallelity => MUCH higher latency).
73		361
74	Under normal circumstances you don't need to call this function, as this
75	module automatically starts a single async thread.
76		362
77	=item IO::AIO::max_parallel $nthreads
78
79	Sets the maximum number of AIO threads to C<$nthreads>. If more than
80	the specified number of threads are currently running, kill them. This
81	function blocks until the limit is reached.
82
83	This module automatically runs C<max_parallel 0> at program end, to ensure
84	that all threads are killed and that there are no outstanding requests.
85
86	Under normal circumstances you don't need to call this function.
87
88	=item $fileno = IO::AIO::poll_fileno
89
90	Return the I<request result pipe filehandle>. This filehandle must be
91	polled for reading by some mechanism outside this module (e.g. Event
92	or select, see below). If the pipe becomes readable you have to call
93	C<poll_cb> to check the results.
94
95	See C<poll_cb> for an example.
96
97	=item IO::AIO::poll_cb
98
99	Process all outstanding events on the result pipe. You have to call this
100	regularly. Returns the number of events processed. Returns immediately
101	when no events are outstanding.
102
103	You can use Event to multiplex, e.g.:
104
105	Event->io (fd => IO::AIO::poll_fileno,
106	poll => 'r', async => 1,
107	cb => \&IO::AIO::poll_cb);
108
109	=item IO::AIO::poll_wait
110
111	Wait till the result filehandle becomes ready for reading (simply does a
112	select on the filehandle. This is useful if you want to synchronously wait
113	for some requests to finish).
114
115	See C<nreqs> for an example.
116
117	=item IO::AIO::nreqs
118
119	Returns the number of requests currently outstanding.
120
121	Example: wait till there are no outstanding requests anymore:
122
123	IO::AIO::poll_wait, IO::AIO::poll_cb
124	while IO::AIO::nreqs;
125
126	=item aio_open $pathname, $flags, $mode, $callback	363	=item aio_open $pathname, $flags, $mode, $callback->($fh)
127		364
128	Asynchronously open or create a file and call the callback with a newly	365	Asynchronously open or create a file and call the callback with a newly
129	created filehandle for the file.	366	created filehandle for the file (or C<undef> in case of an error).
130		367
131	The pathname passed to C<aio_open> must be absolute. See API NOTES, above,	368	The pathname passed to C<aio_open> must be absolute. See API NOTES, above,
132	for an explanation.	369	for an explanation.
133		370
134	The C<$mode> argument is a bitmask. See the C<Fcntl> module for a	371	The C<$flags> argument is a bitmask. See the C<Fcntl> module for a
135	list. They are the same as used in C<sysopen>.	372	list. They are the same as used by C<sysopen>.
		373
		374	Likewise, C<$mode> specifies the mode of the newly created file, if it
		375	didn't exist and C<O_CREAT> has been given, just like perl's C<sysopen>,
		376	except that it is mandatory (i.e. use C<0> if you don't create new files,
		377	and C<0666> or C<0777> if you do). Note that the C<$mode> will be modified
		378	by the umask in effect then the request is being executed, so better never
		379	change the umask.
136		380
137	Example:	381	Example:
138		382
139	aio_open "/etc/passwd", O_RDONLY, 0, sub {	383	aio_open "/etc/passwd", IO::AIO::O_RDONLY, 0, sub {
140	if ($_[0]) {	384	if ($_[0]) {
141	print "open successful, fh is $_[0]\n";	385	print "open successful, fh is $_[0]\n";
142	...	386	...
143	} else {	387	} else {
144	die "open failed: $!\n";	388	die "open failed: $!\n";
145	}	389	}
146	};	390	};
147		391
		392	In addition to all the common open modes/flags (C<O_RDONLY>, C<O_WRONLY>,
		393	C<O_RDWR>, C<O_CREAT>, C<O_TRUNC>, C<O_EXCL> and C<O_APPEND>), the
		394	following POSIX and non-POSIX constants are available (missing ones on
		395	your system are, as usual, C<0>):
		396
		397	C<O_ASYNC>, C<O_DIRECT>, C<O_NOATIME>, C<O_CLOEXEC>, C<O_NOCTTY>, C<O_NOFOLLOW>,
		398	C<O_NONBLOCK>, C<O_EXEC>, C<O_SEARCH>, C<O_DIRECTORY>, C<O_DSYNC>,
		399	C<O_RSYNC>, C<O_SYNC> and C<O_TTY_INIT>.
		400
		401
148	=item aio_close $fh, $callback	402	=item aio_close $fh, $callback->($status)
149		403
150	Asynchronously close a file and call the callback with the result	404	Asynchronously close a file and call the callback with the result
151	code. I<WARNING:> although accepted, you should not pass in a perl	405	code.
152	filehandle here, as perl will likely close the file descriptor itself when
153	the filehandle is destroyed. Normally, you can safely call perls C<close>
154	or just let filehandles go out of scope.
155		406
		407	Unfortunately, you can't do this to perl. Perl I<insists> very strongly on
		408	closing the file descriptor associated with the filehandle itself.
		409
		410	Therefore, C<aio_close> will not close the filehandle - instead it will
		411	use dup2 to overwrite the file descriptor with the write-end of a pipe
		412	(the pipe fd will be created on demand and will be cached).
		413
		414	Or in other words: the file descriptor will be closed, but it will not be
		415	free for reuse until the perl filehandle is closed.
		416
		417	=cut
		418
		419	=item aio_seek $fh, $offset, $whence, $callback->($offs)
		420
		421	Seeks the filehandle to the new C<$offset>, similarly to perl's
		422	C<sysseek>. The C<$whence> can use the traditional values (C<0> for
		423	C<IO::AIO::SEEK_SET>, C<1> for C<IO::AIO::SEEK_CUR> or C<2> for
		424	C<IO::AIO::SEEK_END>).
		425
		426	The resulting absolute offset will be passed to the callback, or C<-1> in
		427	case of an error.
		428
		429	In theory, the C<$whence> constants could be different than the
		430	corresponding values from L<Fcntl>, but perl guarantees they are the same,
		431	so don't panic.
		432
		433	As a GNU/Linux (and maybe Solaris) extension, also the constants
		434	C<IO::AIO::SEEK_DATA> and C<IO::AIO::SEEK_HOLE> are available, if they
		435	could be found. No guarantees about suitability for use in C<aio_seek> or
		436	Perl's C<sysseek> can be made though, although I would naively assume they
		437	"just work".
		438
156	=item aio_read $fh,$offset,$length, $data,$dataoffset,$callback	439	=item aio_read $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
157		440
158	=item aio_write $fh,$offset,$length, $data,$dataoffset,$callback	441	=item aio_write $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
159		442
160	Reads or writes C<length> bytes from the specified C<fh> and C<offset>	443	Reads or writes C<$length> bytes from or to the specified C<$fh> and
161	into the scalar given by C<data> and offset C<dataoffset> and calls the	444	C<$offset> into the scalar given by C<$data> and offset C<$dataoffset>
162	callback without the actual number of bytes read (or -1 on error, just	445	and calls the callback without the actual number of bytes read (or -1 on
163	like the syscall).	446	error, just like the syscall).
164		447
		448	C<aio_read> will, like C<sysread>, shrink or grow the C<$data> scalar to
		449	offset plus the actual number of bytes read.
		450
		451	If C<$offset> is undefined, then the current file descriptor offset will
		452	be used (and updated), otherwise the file descriptor offset will not be
		453	changed by these calls.
		454
		455	If C<$length> is undefined in C<aio_write>, use the remaining length of
		456	C<$data>.
		457
		458	If C<$dataoffset> is less than zero, it will be counted from the end of
		459	C<$data>.
		460
		461	The C<$data> scalar I<MUST NOT> be modified in any way while the request
		462	is outstanding. Modifying it can result in segfaults or World War III (if
		463	the necessary/optional hardware is installed).
		464
165	Example: Read 15 bytes at offset 7 into scalar C<$buffer>, strating at	465	Example: Read 15 bytes at offset 7 into scalar C<$buffer>, starting at
166	offset C<0> within the scalar:	466	offset C<0> within the scalar:
167		467
168	aio_read $fh, 7, 15, $buffer, 0, sub {	468	aio_read $fh, 7, 15, $buffer, 0, sub {
169	$_[0] >= 0 or die "read error: $!";	469	$_[0] > 0 or die "read error: $!";
170	print "read <$buffer>\n";	470	print "read $_[0] bytes: <$buffer>\n";
171	};	471	};
172		472
		473
		474	=item aio_sendfile $out_fh, $in_fh, $in_offset, $length, $callback->($retval)
		475
		476	Tries to copy C<$length> bytes from C<$in_fh> to C<$out_fh>. It starts
		477	reading at byte offset C<$in_offset>, and starts writing at the current
		478	file offset of C<$out_fh>. Because of that, it is not safe to issue more
		479	than one C<aio_sendfile> per C<$out_fh>, as they will interfere with each
		480	other. The same C<$in_fh> works fine though, as this function does not
		481	move or use the file offset of C<$in_fh>.
		482
		483	Please note that C<aio_sendfile> can read more bytes from C<$in_fh> than
		484	are written, and there is no way to find out how many more bytes have been
		485	read from C<aio_sendfile> alone, as C<aio_sendfile> only provides the
		486	number of bytes written to C<$out_fh>. Only if the result value equals
		487	C<$length> one can assume that C<$length> bytes have been read.
		488
		489	Unlike with other C<aio_> functions, it makes a lot of sense to use
		490	C<aio_sendfile> on non-blocking sockets, as long as one end (typically
		491	the C<$in_fh>) is a file - the file I/O will then be asynchronous, while
		492	the socket I/O will be non-blocking. Note, however, that you can run
		493	into a trap where C<aio_sendfile> reads some data with readahead, then
		494	fails to write all data, and when the socket is ready the next time, the
		495	data in the cache is already lost, forcing C<aio_sendfile> to again hit
		496	the disk. Explicit C<aio_read> + C<aio_write> let's you better control
		497	resource usage.
		498
		499	This call tries to make use of a native C<sendfile>-like syscall to
		500	provide zero-copy operation. For this to work, C<$out_fh> should refer to
		501	a socket, and C<$in_fh> should refer to an mmap'able file.
		502
		503	If a native sendfile cannot be found or it fails with C<ENOSYS>,
		504	C<EINVAL>, C<ENOTSUP>, C<EOPNOTSUPP>, C<EAFNOSUPPORT>, C<EPROTOTYPE> or
		505	C<ENOTSOCK>, it will be emulated, so you can call C<aio_sendfile> on any
		506	type of filehandle regardless of the limitations of the operating system.
		507
		508	As native sendfile syscalls (as practically any non-POSIX interface hacked
		509	together in a hurry to improve benchmark numbers) tend to be rather buggy
		510	on many systems, this implementation tries to work around some known bugs
		511	in Linux and FreeBSD kernels (probably others, too), but that might fail,
		512	so you really really should check the return value of C<aio_sendfile> -
		513	fewre bytes than expected might have been transferred.
		514
		515
173	=item aio_readahead $fh,$offset,$length, $callback	516	=item aio_readahead $fh,$offset,$length, $callback->($retval)
174		517
175	Asynchronously reads the specified byte range into the page cache, using
176	the C<readahead> syscall. If that syscall doesn't exist the status will be
177	C<-1> and C<$!> is set to ENOSYS.
178
179	readahead() populates the page cache with data from a file so that	518	C<aio_readahead> populates the page cache with data from a file so that
180	subsequent reads from that file will not block on disk I/O. The C<$offset>	519	subsequent reads from that file will not block on disk I/O. The C<$offset>
181	argument specifies the starting point from which data is to be read and	520	argument specifies the starting point from which data is to be read and
182	C<$length> specifies the number of bytes to be read. I/O is performed in	521	C<$length> specifies the number of bytes to be read. I/O is performed in
183	whole pages, so that offset is effectively rounded down to a page boundary	522	whole pages, so that offset is effectively rounded down to a page boundary
184	and bytes are read up to the next page boundary greater than or equal to	523	and bytes are read up to the next page boundary greater than or equal to
185	(off-set+length). aio_readahead() does not read beyond the end of the	524	(off-set+length). C<aio_readahead> does not read beyond the end of the
186	file. The current file offset of the file is left unchanged.	525	file. The current file offset of the file is left unchanged.
187		526
		527	If that syscall doesn't exist (likely if your OS isn't Linux) it will be
		528	emulated by simply reading the data, which would have a similar effect.
		529
		530
188	=item aio_stat $fh_or_path, $callback	531	=item aio_stat $fh_or_path, $callback->($status)
189		532
190	=item aio_lstat $fh, $callback	533	=item aio_lstat $fh, $callback->($status)
191		534
192	Works like perl's C<stat> or C<lstat> in void context. The callback will	535	Works like perl's C<stat> or C<lstat> in void context. The callback will
193	be called after the stat and the results will be available using C<stat _>	536	be called after the stat and the results will be available using C<stat _>
194	or C<-s _> etc...	537	or C<-s _> etc...
195		538
…		…
197	for an explanation.	540	for an explanation.
198		541
199	Currently, the stats are always 64-bit-stats, i.e. instead of returning an	542	Currently, the stats are always 64-bit-stats, i.e. instead of returning an
200	error when stat'ing a large file, the results will be silently truncated	543	error when stat'ing a large file, the results will be silently truncated
201	unless perl itself is compiled with large file support.	544	unless perl itself is compiled with large file support.
		545
		546	To help interpret the mode and dev/rdev stat values, IO::AIO offers the
		547	following constants and functions (if not implemented, the constants will
		548	be C<0> and the functions will either C<croak> or fall back on traditional
		549	behaviour).
		550
		551	C<S_IFMT>, C<S_IFIFO>, C<S_IFCHR>, C<S_IFBLK>, C<S_IFLNK>, C<S_IFREG>,
		552	C<S_IFDIR>, C<S_IFWHT>, C<S_IFSOCK>, C<IO::AIO::major $dev_t>,
		553	C<IO::AIO::minor $dev_t>, C<IO::AIO::makedev $major, $minor>.
202		554
203	Example: Print the length of F</etc/passwd>:	555	Example: Print the length of F</etc/passwd>:
204		556
205	aio_stat "/etc/passwd", sub {	557	aio_stat "/etc/passwd", sub {
206	$_[0] and die "stat failed: $!";	558	$_[0] and die "stat failed: $!";
207	print "size is ", -s _, "\n";	559	print "size is ", -s _, "\n";
208	};	560	};
209		561
		562
		563	=item aio_statvfs $fh_or_path, $callback->($statvfs)
		564
		565	Works like the POSIX C<statvfs> or C<fstatvfs> syscalls, depending on
		566	whether a file handle or path was passed.
		567
		568	On success, the callback is passed a hash reference with the following
		569	members: C<bsize>, C<frsize>, C<blocks>, C<bfree>, C<bavail>, C<files>,
		570	C<ffree>, C<favail>, C<fsid>, C<flag> and C<namemax>. On failure, C<undef>
		571	is passed.
		572
		573	The following POSIX IO::AIO::ST_* constants are defined: C<ST_RDONLY> and
		574	C<ST_NOSUID>.
		575
		576	The following non-POSIX IO::AIO::ST_* flag masks are defined to
		577	their correct value when available, or to C<0> on systems that do
		578	not support them: C<ST_NODEV>, C<ST_NOEXEC>, C<ST_SYNCHRONOUS>,
		579	C<ST_MANDLOCK>, C<ST_WRITE>, C<ST_APPEND>, C<ST_IMMUTABLE>, C<ST_NOATIME>,
		580	C<ST_NODIRATIME> and C<ST_RELATIME>.
		581
		582	Example: stat C</wd> and dump out the data if successful.
		583
		584	aio_statvfs "/wd", sub {
		585	my $f = $_[0]
		586	or die "statvfs: $!";
		587
		588	use Data::Dumper;
		589	say Dumper $f;
		590	};
		591
		592	# result:
		593	{
		594	bsize => 1024,
		595	bfree => 4333064312,
		596	blocks => 10253828096,
		597	files => 2050765568,
		598	flag => 4096,
		599	favail => 2042092649,
		600	bavail => 4333064312,
		601	ffree => 2042092649,
		602	namemax => 255,
		603	frsize => 1024,
		604	fsid => 1810
		605	}
		606
		607	Here is a (likely partial - send me updates!) list of fsid values used by
		608	Linux - it is safe to hardcode these when C<$^O> is C<linux>:
		609
		610	0x0000adf5 adfs
		611	0x0000adff affs
		612	0x5346414f afs
		613	0x09041934 anon-inode filesystem
		614	0x00000187 autofs
		615	0x42465331 befs
		616	0x1badface bfs
		617	0x42494e4d binfmt_misc
		618	0x9123683e btrfs
		619	0x0027e0eb cgroupfs
		620	0xff534d42 cifs
		621	0x73757245 coda
		622	0x012ff7b7 coh
		623	0x28cd3d45 cramfs
		624	0x453dcd28 cramfs-wend (wrong endianness)
		625	0x64626720 debugfs
		626	0x00001373 devfs
		627	0x00001cd1 devpts
		628	0x0000f15f ecryptfs
		629	0x00414a53 efs
		630	0x0000137d ext
		631	0x0000ef53 ext2/ext3
		632	0x0000ef51 ext2
		633	0x00004006 fat
		634	0x65735546 fuseblk
		635	0x65735543 fusectl
		636	0x0bad1dea futexfs
		637	0x01161970 gfs2
		638	0x47504653 gpfs
		639	0x00004244 hfs
		640	0xf995e849 hpfs
		641	0x958458f6 hugetlbfs
		642	0x2bad1dea inotifyfs
		643	0x00009660 isofs
		644	0x000072b6 jffs2
		645	0x3153464a jfs
		646	0x6b414653 k-afs
		647	0x0bd00bd0 lustre
		648	0x0000137f minix
		649	0x0000138f minix 30 char names
		650	0x00002468 minix v2
		651	0x00002478 minix v2 30 char names
		652	0x00004d5a minix v3
		653	0x19800202 mqueue
		654	0x00004d44 msdos
		655	0x0000564c novell
		656	0x00006969 nfs
		657	0x6e667364 nfsd
		658	0x00003434 nilfs
		659	0x5346544e ntfs
		660	0x00009fa1 openprom
		661	0x7461636F ocfs2
		662	0x00009fa0 proc
		663	0x6165676c pstorefs
		664	0x0000002f qnx4
		665	0x858458f6 ramfs
		666	0x52654973 reiserfs
		667	0x00007275 romfs
		668	0x67596969 rpc_pipefs
		669	0x73636673 securityfs
		670	0xf97cff8c selinux
		671	0x0000517b smb
		672	0x534f434b sockfs
		673	0x73717368 squashfs
		674	0x62656572 sysfs
		675	0x012ff7b6 sysv2
		676	0x012ff7b5 sysv4
		677	0x01021994 tmpfs
		678	0x15013346 udf
		679	0x00011954 ufs
		680	0x54190100 ufs byteswapped
		681	0x00009fa2 usbdevfs
		682	0x01021997 v9fs
		683	0xa501fcf5 vxfs
		684	0xabba1974 xenfs
		685	0x012ff7b4 xenix
		686	0x58465342 xfs
		687	0x012fd16d xia
		688
		689	=item aio_utime $fh_or_path, $atime, $mtime, $callback->($status)
		690
		691	Works like perl's C<utime> function (including the special case of $atime
		692	and $mtime being undef). Fractional times are supported if the underlying
		693	syscalls support them.
		694
		695	When called with a pathname, uses utimes(2) if available, otherwise
		696	utime(2). If called on a file descriptor, uses futimes(2) if available,
		697	otherwise returns ENOSYS, so this is not portable.
		698
		699	Examples:
		700
		701	# set atime and mtime to current time (basically touch(1)):
		702	aio_utime "path", undef, undef;
		703	# set atime to current time and mtime to beginning of the epoch:
		704	aio_utime "path", time, undef; # undef==0
		705
		706
		707	=item aio_chown $fh_or_path, $uid, $gid, $callback->($status)
		708
		709	Works like perl's C<chown> function, except that C<undef> for either $uid
		710	or $gid is being interpreted as "do not change" (but -1 can also be used).
		711
		712	Examples:
		713
		714	# same as "chown root path" in the shell:
		715	aio_chown "path", 0, -1;
		716	# same as above:
		717	aio_chown "path", 0, undef;
		718
		719
		720	=item aio_truncate $fh_or_path, $offset, $callback->($status)
		721
		722	Works like truncate(2) or ftruncate(2).
		723
		724
		725	=item aio_allocate $fh, $mode, $offset, $len, $callback->($status)
		726
		727	Allocates or frees disk space according to the C<$mode> argument. See the
		728	linux C<fallocate> documentation for details.
		729
		730	C<$mode> is usually C<0> or C<IO::AIO::FALLOC_FL_KEEP_SIZE> to allocate
		731	space, or C<IO::AIO::FALLOC_FL_PUNCH_HOLE | IO::AIO::FALLOC_FL_KEEP_SIZE>,
		732	to deallocate a file range.
		733
		734	IO::AIO also supports C<FALLOC_FL_COLLAPSE_RANGE>, to remove a range
		735	(without leaving a hole) and C<FALLOC_FL_ZERO_RANGE>, to zero a range (see
		736	your L<fallocate(2)> manpage).
		737
		738	The file system block size used by C<fallocate> is presumably the
		739	C<f_bsize> returned by C<statvfs>.
		740
		741	If C<fallocate> isn't available or cannot be emulated (currently no
		742	emulation will be attempted), passes C<-1> and sets C<$!> to C<ENOSYS>.
		743
		744
		745	=item aio_chmod $fh_or_path, $mode, $callback->($status)
		746
		747	Works like perl's C<chmod> function.
		748
		749
210	=item aio_unlink $pathname, $callback	750	=item aio_unlink $pathname, $callback->($status)
211		751
212	Asynchronously unlink (delete) a file and call the callback with the	752	Asynchronously unlink (delete) a file and call the callback with the
213	result code.	753	result code.
214		754
		755
		756	=item aio_mknod $pathname, $mode, $dev, $callback->($status)
		757
		758	[EXPERIMENTAL]
		759
		760	Asynchronously create a device node (or fifo). See mknod(2).
		761
		762	The only (POSIX-) portable way of calling this function is:
		763
		764	aio_mknod $pathname, IO::AIO::S_IFIFO | $mode, 0, sub { ...
		765
		766	See C<aio_stat> for info about some potentially helpful extra constants
		767	and functions.
		768
		769	=item aio_link $srcpath, $dstpath, $callback->($status)
		770
		771	Asynchronously create a new link to the existing object at C<$srcpath> at
		772	the path C<$dstpath> and call the callback with the result code.
		773
		774
		775	=item aio_symlink $srcpath, $dstpath, $callback->($status)
		776
		777	Asynchronously create a new symbolic link to the existing object at C<$srcpath> at
		778	the path C<$dstpath> and call the callback with the result code.
		779
		780
		781	=item aio_readlink $pathname, $callback->($link)
		782
		783	Asynchronously read the symlink specified by C<$path> and pass it to
		784	the callback. If an error occurs, nothing or undef gets passed to the
		785	callback.
		786
		787
		788	=item aio_realpath $pathname, $callback->($path)
		789
		790	Asynchronously make the path absolute and resolve any symlinks in
		791	C<$path>. The resulting path only consists of directories (same as
		792	L<Cwd::realpath>).
		793
		794	This request can be used to get the absolute path of the current working
		795	directory by passing it a path of F<.> (a single dot).
		796
		797
		798	=item aio_rename $srcpath, $dstpath, $callback->($status)
		799
		800	Asynchronously rename the object at C<$srcpath> to C<$dstpath>, just as
		801	rename(2) and call the callback with the result code.
		802
		803	On systems that support the AIO::WD working directory abstraction
		804	natively, the case C<[$wd, "."]> as C<$srcpath> is specialcased - instead
		805	of failing, C<rename> is called on the absolute path of C<$wd>.
		806
		807
		808	=item aio_mkdir $pathname, $mode, $callback->($status)
		809
		810	Asynchronously mkdir (create) a directory and call the callback with
		811	the result code. C<$mode> will be modified by the umask at the time the
		812	request is executed, so do not change your umask.
		813
		814
		815	=item aio_rmdir $pathname, $callback->($status)
		816
		817	Asynchronously rmdir (delete) a directory and call the callback with the
		818	result code.
		819
		820	On systems that support the AIO::WD working directory abstraction
		821	natively, the case C<[$wd, "."]> is specialcased - instead of failing,
		822	C<rmdir> is called on the absolute path of C<$wd>.
		823
		824
		825	=item aio_readdir $pathname, $callback->($entries)
		826
		827	Unlike the POSIX call of the same name, C<aio_readdir> reads an entire
		828	directory (i.e. opendir + readdir + closedir). The entries will not be
		829	sorted, and will B<NOT> include the C<.> and C<..> entries.
		830
		831	The callback is passed a single argument which is either C<undef> or an
		832	array-ref with the filenames.
		833
		834
		835	=item aio_readdirx $pathname, $flags, $callback->($entries, $flags)
		836
		837	Quite similar to C<aio_readdir>, but the C<$flags> argument allows one to
		838	tune behaviour and output format. In case of an error, C<$entries> will be
		839	C<undef>.
		840
		841	The flags are a combination of the following constants, ORed together (the
		842	flags will also be passed to the callback, possibly modified):
		843
		844	=over 4
		845
		846	=item IO::AIO::READDIR_DENTS
		847
		848	When this flag is off, then the callback gets an arrayref consisting of
		849	names only (as with C<aio_readdir>), otherwise it gets an arrayref with
		850	C<[$name, $type, $inode]> arrayrefs, each describing a single directory
		851	entry in more detail.
		852
		853	C<$name> is the name of the entry.
		854
		855	C<$type> is one of the C<IO::AIO::DT_xxx> constants:
		856
		857	C<IO::AIO::DT_UNKNOWN>, C<IO::AIO::DT_FIFO>, C<IO::AIO::DT_CHR>, C<IO::AIO::DT_DIR>,
		858	C<IO::AIO::DT_BLK>, C<IO::AIO::DT_REG>, C<IO::AIO::DT_LNK>, C<IO::AIO::DT_SOCK>,
		859	C<IO::AIO::DT_WHT>.
		860
		861	C<IO::AIO::DT_UNKNOWN> means just that: readdir does not know. If you need to
		862	know, you have to run stat yourself. Also, for speed reasons, the C<$type>
		863	scalars are read-only: you can not modify them.
		864
		865	C<$inode> is the inode number (which might not be exact on systems with 64
		866	bit inode numbers and 32 bit perls). This field has unspecified content on
		867	systems that do not deliver the inode information.
		868
		869	=item IO::AIO::READDIR_DIRS_FIRST
		870
		871	When this flag is set, then the names will be returned in an order where
		872	likely directories come first, in optimal stat order. This is useful when
		873	you need to quickly find directories, or you want to find all directories
		874	while avoiding to stat() each entry.
		875
		876	If the system returns type information in readdir, then this is used
		877	to find directories directly. Otherwise, likely directories are names
		878	beginning with ".", or otherwise names with no dots, of which names with
		879	short names are tried first.
		880
		881	=item IO::AIO::READDIR_STAT_ORDER
		882
		883	When this flag is set, then the names will be returned in an order
		884	suitable for stat()'ing each one. That is, when you plan to stat()
		885	all files in the given directory, then the returned order will likely
		886	be fastest.
		887
		888	If both this flag and C<IO::AIO::READDIR_DIRS_FIRST> are specified, then
		889	the likely dirs come first, resulting in a less optimal stat order.
		890
		891	=item IO::AIO::READDIR_FOUND_UNKNOWN
		892
		893	This flag should not be set when calling C<aio_readdirx>. Instead, it
		894	is being set by C<aio_readdirx>, when any of the C<$type>'s found were
		895	C<IO::AIO::DT_UNKNOWN>. The absence of this flag therefore indicates that all
		896	C<$type>'s are known, which can be used to speed up some algorithms.
		897
		898	=back
		899
		900
		901	=item aio_load $pathname, $data, $callback->($status)
		902
		903	This is a composite request that tries to fully load the given file into
		904	memory. Status is the same as with aio_read.
		905
		906	=cut
		907
		908	sub aio_load($$;$) {
		909	my ($path, undef, $cb) = @_;
		910	my $data = \$_[1];
		911
		912	my $pri = aioreq_pri;
		913	my $grp = aio_group $cb;
		914
		915	aioreq_pri $pri;
		916	add $grp aio_open $path, O_RDONLY, 0, sub {
		917	my $fh = shift
		918	or return $grp->result (-1);
		919
		920	aioreq_pri $pri;
		921	add $grp aio_read $fh, 0, (-s $fh), $$data, 0, sub {
		922	$grp->result ($_[0]);
		923	};
		924	};
		925
		926	$grp
		927	}
		928
		929	=item aio_copy $srcpath, $dstpath, $callback->($status)
		930
		931	Try to copy the I<file> (directories not supported as either source or
		932	destination) from C<$srcpath> to C<$dstpath> and call the callback with
		933	a status of C<0> (ok) or C<-1> (error, see C<$!>).
		934
		935	This is a composite request that creates the destination file with
		936	mode 0200 and copies the contents of the source file into it using
		937	C<aio_sendfile>, followed by restoring atime, mtime, access mode and
		938	uid/gid, in that order.
		939
		940	If an error occurs, the partial destination file will be unlinked, if
		941	possible, except when setting atime, mtime, access mode and uid/gid, where
		942	errors are being ignored.
		943
		944	=cut
		945
		946	sub aio_copy($$;$) {
		947	my ($src, $dst, $cb) = @_;
		948
		949	my $pri = aioreq_pri;
		950	my $grp = aio_group $cb;
		951
		952	aioreq_pri $pri;
		953	add $grp aio_open $src, O_RDONLY, 0, sub {
		954	if (my $src_fh = $_[0]) {
		955	my @stat = stat $src_fh; # hmm, might block over nfs?
		956
		957	aioreq_pri $pri;
		958	add $grp aio_open $dst, O_CREAT | O_WRONLY | O_TRUNC, 0200, sub {
		959	if (my $dst_fh = $_[0]) {
		960	aioreq_pri $pri;
		961	add $grp aio_sendfile $dst_fh, $src_fh, 0, $stat[7], sub {
		962	if ($_[0] == $stat[7]) {
		963	$grp->result (0);
		964	close $src_fh;
		965
		966	my $ch = sub {
		967	aioreq_pri $pri;
		968	add $grp aio_chmod $dst_fh, $stat[2] & 07777, sub {
		969	aioreq_pri $pri;
		970	add $grp aio_chown $dst_fh, $stat[4], $stat[5], sub {
		971	aioreq_pri $pri;
		972	add $grp aio_close $dst_fh;
		973	}
		974	};
		975	};
		976
		977	aioreq_pri $pri;
		978	add $grp aio_utime $dst_fh, $stat[8], $stat[9], sub {
		979	if ($_[0] < 0 && $! == ENOSYS) {
		980	aioreq_pri $pri;
		981	add $grp aio_utime $dst, $stat[8], $stat[9], $ch;
		982	} else {
		983	$ch->();
		984	}
		985	};
		986	} else {
		987	$grp->result (-1);
		988	close $src_fh;
		989	close $dst_fh;
		990
		991	aioreq $pri;
		992	add $grp aio_unlink $dst;
		993	}
		994	};
		995	} else {
		996	$grp->result (-1);
		997	}
		998	},
		999
		1000	} else {
		1001	$grp->result (-1);
		1002	}
		1003	};
		1004
		1005	$grp
		1006	}
		1007
		1008	=item aio_move $srcpath, $dstpath, $callback->($status)
		1009
		1010	Try to move the I<file> (directories not supported as either source or
		1011	destination) from C<$srcpath> to C<$dstpath> and call the callback with
		1012	a status of C<0> (ok) or C<-1> (error, see C<$!>).
		1013
		1014	This is a composite request that tries to rename(2) the file first; if
		1015	rename fails with C<EXDEV>, it copies the file with C<aio_copy> and, if
		1016	that is successful, unlinks the C<$srcpath>.
		1017
		1018	=cut
		1019
		1020	sub aio_move($$;$) {
		1021	my ($src, $dst, $cb) = @_;
		1022
		1023	my $pri = aioreq_pri;
		1024	my $grp = aio_group $cb;
		1025
		1026	aioreq_pri $pri;
		1027	add $grp aio_rename $src, $dst, sub {
		1028	if ($_[0] && $! == EXDEV) {
		1029	aioreq_pri $pri;
		1030	add $grp aio_copy $src, $dst, sub {
		1031	$grp->result ($_[0]);
		1032
		1033	unless ($_[0]) {
		1034	aioreq_pri $pri;
		1035	add $grp aio_unlink $src;
		1036	}
		1037	};
		1038	} else {
		1039	$grp->result ($_[0]);
		1040	}
		1041	};
		1042
		1043	$grp
		1044	}
		1045
		1046	=item aio_scandir $pathname, $maxreq, $callback->($dirs, $nondirs)
		1047
		1048	Scans a directory (similar to C<aio_readdir>) but additionally tries to
		1049	efficiently separate the entries of directory C<$path> into two sets of
		1050	names, directories you can recurse into (directories), and ones you cannot
		1051	recurse into (everything else, including symlinks to directories).
		1052
		1053	C<aio_scandir> is a composite request that creates of many sub requests_
		1054	C<$maxreq> specifies the maximum number of outstanding aio requests that
		1055	this function generates. If it is C<< <= 0 >>, then a suitable default
		1056	will be chosen (currently 4).
		1057
		1058	On error, the callback is called without arguments, otherwise it receives
		1059	two array-refs with path-relative entry names.
		1060
		1061	Example:
		1062
		1063	aio_scandir $dir, 0, sub {
		1064	my ($dirs, $nondirs) = @_;
		1065	print "real directories: @$dirs\n";
		1066	print "everything else: @$nondirs\n";
		1067	};
		1068
		1069	Implementation notes.
		1070
		1071	The C<aio_readdir> cannot be avoided, but C<stat()>'ing every entry can.
		1072
		1073	If readdir returns file type information, then this is used directly to
		1074	find directories.
		1075
		1076	Otherwise, after reading the directory, the modification time, size etc.
		1077	of the directory before and after the readdir is checked, and if they
		1078	match (and isn't the current time), the link count will be used to decide
		1079	how many entries are directories (if >= 2). Otherwise, no knowledge of the
		1080	number of subdirectories will be assumed.
		1081
		1082	Then entries will be sorted into likely directories a non-initial dot
		1083	currently) and likely non-directories (see C<aio_readdirx>). Then every
		1084	entry plus an appended C</.> will be C<stat>'ed, likely directories first,
		1085	in order of their inode numbers. If that succeeds, it assumes that the
		1086	entry is a directory or a symlink to directory (which will be checked
		1087	separately). This is often faster than stat'ing the entry itself because
		1088	filesystems might detect the type of the entry without reading the inode
		1089	data (e.g. ext2fs filetype feature), even on systems that cannot return
		1090	the filetype information on readdir.
		1091
		1092	If the known number of directories (link count - 2) has been reached, the
		1093	rest of the entries is assumed to be non-directories.
		1094
		1095	This only works with certainty on POSIX (= UNIX) filesystems, which
		1096	fortunately are the vast majority of filesystems around.
		1097
		1098	It will also likely work on non-POSIX filesystems with reduced efficiency
		1099	as those tend to return 0 or 1 as link counts, which disables the
		1100	directory counting heuristic.
		1101
		1102	=cut
		1103
		1104	sub aio_scandir($$;$) {
		1105	my ($path, $maxreq, $cb) = @_;
		1106
		1107	my $pri = aioreq_pri;
		1108
		1109	my $grp = aio_group $cb;
		1110
		1111	$maxreq = 4 if $maxreq <= 0;
		1112
		1113	# get a wd object
		1114	aioreq_pri $pri;
		1115	add $grp aio_wd $path, sub {
		1116	$_[0]
		1117	or return $grp->result ();
		1118
		1119	my $wd = [shift, "."];
		1120
		1121	# stat once
		1122	aioreq_pri $pri;
		1123	add $grp aio_stat $wd, sub {
		1124	return $grp->result () if $_[0];
		1125	my $now = time;
		1126	my $hash1 = join ":", (stat _)[0,1,3,7,9];
		1127
		1128	# read the directory entries
		1129	aioreq_pri $pri;
		1130	add $grp aio_readdirx $wd, READDIR_DIRS_FIRST, sub {
		1131	my $entries = shift
		1132	or return $grp->result ();
		1133
		1134	# stat the dir another time
		1135	aioreq_pri $pri;
		1136	add $grp aio_stat $wd, sub {
		1137	my $hash2 = join ":", (stat _)[0,1,3,7,9];
		1138
		1139	my $ndirs;
		1140
		1141	# take the slow route if anything looks fishy
		1142	if ($hash1 ne $hash2 or (stat _)[9] == $now) {
		1143	$ndirs = -1;
		1144	} else {
		1145	# if nlink == 2, we are finished
		1146	# for non-posix-fs's, we rely on nlink < 2
		1147	$ndirs = (stat _)[3] - 2
		1148	or return $grp->result ([], $entries);
		1149	}
		1150
		1151	my (@dirs, @nondirs);
		1152
		1153	my $statgrp = add $grp aio_group sub {
		1154	$grp->result (\@dirs, \@nondirs);
		1155	};
		1156
		1157	limit $statgrp $maxreq;
		1158	feed $statgrp sub {
		1159	return unless @$entries;
		1160	my $entry = shift @$entries;
		1161
		1162	aioreq_pri $pri;
		1163	$wd->[1] = "$entry/.";
		1164	add $statgrp aio_stat $wd, sub {
		1165	if ($_[0] < 0) {
		1166	push @nondirs, $entry;
		1167	} else {
		1168	# need to check for real directory
		1169	aioreq_pri $pri;
		1170	$wd->[1] = $entry;
		1171	add $statgrp aio_lstat $wd, sub {
		1172	if (-d _) {
		1173	push @dirs, $entry;
		1174
		1175	unless (--$ndirs) {
		1176	push @nondirs, @$entries;
		1177	feed $statgrp;
		1178	}
		1179	} else {
		1180	push @nondirs, $entry;
		1181	}
		1182	}
		1183	}
		1184	};
		1185	};
		1186	};
		1187	};
		1188	};
		1189	};
		1190
		1191	$grp
		1192	}
		1193
		1194	=item aio_rmtree $pathname, $callback->($status)
		1195
		1196	Delete a directory tree starting (and including) C<$path>, return the
		1197	status of the final C<rmdir> only. This is a composite request that
		1198	uses C<aio_scandir> to recurse into and rmdir directories, and unlink
		1199	everything else.
		1200
		1201	=cut
		1202
		1203	sub aio_rmtree;
		1204	sub aio_rmtree($;$) {
		1205	my ($path, $cb) = @_;
		1206
		1207	my $pri = aioreq_pri;
		1208	my $grp = aio_group $cb;
		1209
		1210	aioreq_pri $pri;
		1211	add $grp aio_scandir $path, 0, sub {
		1212	my ($dirs, $nondirs) = @_;
		1213
		1214	my $dirgrp = aio_group sub {
		1215	add $grp aio_rmdir $path, sub {
		1216	$grp->result ($_[0]);
		1217	};
		1218	};
		1219
		1220	(aioreq_pri $pri), add $dirgrp aio_rmtree "$path/$_" for @$dirs;
		1221	(aioreq_pri $pri), add $dirgrp aio_unlink "$path/$_" for @$nondirs;
		1222
		1223	add $grp $dirgrp;
		1224	};
		1225
		1226	$grp
		1227	}
		1228
		1229	=item aio_sync $callback->($status)
		1230
		1231	Asynchronously call sync and call the callback when finished.
		1232
215	=item aio_fsync $fh, $callback	1233	=item aio_fsync $fh, $callback->($status)
216		1234
217	Asynchronously call fsync on the given filehandle and call the callback	1235	Asynchronously call fsync on the given filehandle and call the callback
218	with the fsync result code.	1236	with the fsync result code.
219		1237
220	=item aio_fdatasync $fh, $callback	1238	=item aio_fdatasync $fh, $callback->($status)
221		1239
222	Asynchronously call fdatasync on the given filehandle and call the	1240	Asynchronously call fdatasync on the given filehandle and call the
223	callback with the fdatasync result code.	1241	callback with the fdatasync result code.
224		1242
		1243	If this call isn't available because your OS lacks it or it couldn't be
		1244	detected, it will be emulated by calling C<fsync> instead.
		1245
		1246	=item aio_syncfs $fh, $callback->($status)
		1247
		1248	Asynchronously call the syncfs syscall to sync the filesystem associated
		1249	to the given filehandle and call the callback with the syncfs result
		1250	code. If syncfs is not available, calls sync(), but returns C<-1> and sets
		1251	errno to C<ENOSYS> nevertheless.
		1252
		1253	=item aio_sync_file_range $fh, $offset, $nbytes, $flags, $callback->($status)
		1254
		1255	Sync the data portion of the file specified by C<$offset> and C<$length>
		1256	to disk (but NOT the metadata), by calling the Linux-specific
		1257	sync_file_range call. If sync_file_range is not available or it returns
		1258	ENOSYS, then fdatasync or fsync is being substituted.
		1259
		1260	C<$flags> can be a combination of C<IO::AIO::SYNC_FILE_RANGE_WAIT_BEFORE>,
		1261	C<IO::AIO::SYNC_FILE_RANGE_WRITE> and
		1262	C<IO::AIO::SYNC_FILE_RANGE_WAIT_AFTER>: refer to the sync_file_range
		1263	manpage for details.
		1264
		1265	=item aio_pathsync $pathname, $callback->($status)
		1266
		1267	This request tries to open, fsync and close the given path. This is a
		1268	composite request intended to sync directories after directory operations
		1269	(E.g. rename). This might not work on all operating systems or have any
		1270	specific effect, but usually it makes sure that directory changes get
		1271	written to disc. It works for anything that can be opened for read-only,
		1272	not just directories.
		1273
		1274	Future versions of this function might fall back to other methods when
		1275	C<fsync> on the directory fails (such as calling C<sync>).
		1276
		1277	Passes C<0> when everything went ok, and C<-1> on error.
		1278
225	=cut	1279	=cut
226		1280
227	# support function to convert a fd into a perl filehandle	1281	sub aio_pathsync($;$) {
228	sub _fd2fh {	1282	my ($path, $cb) = @_;
229	return undef if $_[0] < 0;
230		1283
231	# try to be perl5.6-compatible	1284	my $pri = aioreq_pri;
232	local *AIO_FH;	1285	my $grp = aio_group $cb;
233	open AIO_FH, "+<&=$_[0]"
234	or return undef;
235		1286
236	*AIO_FH	1287	aioreq_pri $pri;
		1288	add $grp aio_open $path, O_RDONLY, 0, sub {
		1289	my ($fh) = @_;
		1290	if ($fh) {
		1291	aioreq_pri $pri;
		1292	add $grp aio_fsync $fh, sub {
		1293	$grp->result ($_[0]);
		1294
		1295	aioreq_pri $pri;
		1296	add $grp aio_close $fh;
		1297	};
		1298	} else {
		1299	$grp->result (-1);
		1300	}
		1301	};
		1302
		1303	$grp
237	}	1304	}
238		1305
		1306	=item aio_msync $scalar, $offset = 0, $length = undef, flags = 0, $callback->($status)
		1307
		1308	This is a rather advanced IO::AIO call, which only works on mmap(2)ed
		1309	scalars (see the C<IO::AIO::mmap> function, although it also works on data
		1310	scalars managed by the L<Sys::Mmap> or L<Mmap> modules, note that the
		1311	scalar must only be modified in-place while an aio operation is pending on
		1312	it).
		1313
		1314	It calls the C<msync> function of your OS, if available, with the memory
		1315	area starting at C<$offset> in the string and ending C<$length> bytes
		1316	later. If C<$length> is negative, counts from the end, and if C<$length>
		1317	is C<undef>, then it goes till the end of the string. The flags can be
		1318	a combination of C<IO::AIO::MS_ASYNC>, C<IO::AIO::MS_INVALIDATE> and
		1319	C<IO::AIO::MS_SYNC>.
		1320
		1321	=item aio_mtouch $scalar, $offset = 0, $length = undef, flags = 0, $callback->($status)
		1322
		1323	This is a rather advanced IO::AIO call, which works best on mmap(2)ed
		1324	scalars.
		1325
		1326	It touches (reads or writes) all memory pages in the specified
		1327	range inside the scalar. All caveats and parameters are the same
		1328	as for C<aio_msync>, above, except for flags, which must be either
		1329	C<0> (which reads all pages and ensures they are instantiated) or
		1330	C<IO::AIO::MT_MODIFY>, which modifies the memory pages (by reading and
		1331	writing an octet from it, which dirties the page).
		1332
		1333	=item aio_mlock $scalar, $offset = 0, $length = undef, $callback->($status)
		1334
		1335	This is a rather advanced IO::AIO call, which works best on mmap(2)ed
		1336	scalars.
		1337
		1338	It reads in all the pages of the underlying storage into memory (if any)
		1339	and locks them, so they are not getting swapped/paged out or removed.
		1340
		1341	If C<$length> is undefined, then the scalar will be locked till the end.
		1342
		1343	On systems that do not implement C<mlock>, this function returns C<-1>
		1344	and sets errno to C<ENOSYS>.
		1345
		1346	Note that the corresponding C<munlock> is synchronous and is
		1347	documented under L<MISCELLANEOUS FUNCTIONS>.
		1348
		1349	Example: open a file, mmap and mlock it - both will be undone when
		1350	C<$data> gets destroyed.
		1351
		1352	open my $fh, "<", $path or die "$path: $!";
		1353	my $data;
		1354	IO::AIO::mmap $data, -s $fh, IO::AIO::PROT_READ, IO::AIO::MAP_SHARED, $fh;
		1355	aio_mlock $data; # mlock in background
		1356
		1357	=item aio_mlockall $flags, $callback->($status)
		1358
		1359	Calls the C<mlockall> function with the given C<$flags> (a combination of
		1360	C<IO::AIO::MCL_CURRENT> and C<IO::AIO::MCL_FUTURE>).
		1361
		1362	On systems that do not implement C<mlockall>, this function returns C<-1>
		1363	and sets errno to C<ENOSYS>.
		1364
		1365	Note that the corresponding C<munlockall> is synchronous and is
		1366	documented under L<MISCELLANEOUS FUNCTIONS>.
		1367
		1368	Example: asynchronously lock all current and future pages into memory.
		1369
		1370	aio_mlockall IO::AIO::MCL_FUTURE;
		1371
		1372	=item aio_fiemap $fh, $start, $length, $flags, $count, $cb->(\@extents)
		1373
		1374	Queries the extents of the given file (by calling the Linux C<FIEMAP>
		1375	ioctl, see L<http://cvs.schmorp.de/IO-AIO/doc/fiemap.txt> for details). If
		1376	the ioctl is not available on your OS, then this request will fail with
		1377	C<ENOSYS>.
		1378
		1379	C<$start> is the starting offset to query extents for, C<$length> is the
		1380	size of the range to query - if it is C<undef>, then the whole file will
		1381	be queried.
		1382
		1383	C<$flags> is a combination of flags (C<IO::AIO::FIEMAP_FLAG_SYNC> or
		1384	C<IO::AIO::FIEMAP_FLAG_XATTR> - C<IO::AIO::FIEMAP_FLAGS_COMPAT> is also
		1385	exported), and is normally C<0> or C<IO::AIO::FIEMAP_FLAG_SYNC> to query
		1386	the data portion.
		1387
		1388	C<$count> is the maximum number of extent records to return. If it is
		1389	C<undef>, then IO::AIO queries all extents of the range. As a very special
		1390	case, if it is C<0>, then the callback receives the number of extents
		1391	instead of the extents themselves (which is unreliable, see below).
		1392
		1393	If an error occurs, the callback receives no arguments. The special
		1394	C<errno> value C<IO::AIO::EBADR> is available to test for flag errors.
		1395
		1396	Otherwise, the callback receives an array reference with extent
		1397	structures. Each extent structure is an array reference itself, with the
		1398	following members:
		1399
		1400	[$logical, $physical, $length, $flags]
		1401
		1402	Flags is any combination of the following flag values (typically either C<0>
		1403	or C<IO::AIO::FIEMAP_EXTENT_LAST> (1)):
		1404
		1405	C<IO::AIO::FIEMAP_EXTENT_LAST>, C<IO::AIO::FIEMAP_EXTENT_UNKNOWN>,
		1406	C<IO::AIO::FIEMAP_EXTENT_DELALLOC>, C<IO::AIO::FIEMAP_EXTENT_ENCODED>,
		1407	C<IO::AIO::FIEMAP_EXTENT_DATA_ENCRYPTED>, C<IO::AIO::FIEMAP_EXTENT_NOT_ALIGNED>,
		1408	C<IO::AIO::FIEMAP_EXTENT_DATA_INLINE>, C<IO::AIO::FIEMAP_EXTENT_DATA_TAIL>,
		1409	C<IO::AIO::FIEMAP_EXTENT_UNWRITTEN>, C<IO::AIO::FIEMAP_EXTENT_MERGED> or
		1410	C<IO::AIO::FIEMAP_EXTENT_SHARED>.
		1411
		1412	At the time of this writing (Linux 3.2), this requets is unreliable unless
		1413	C<$count> is C<undef>, as the kernel has all sorts of bugs preventing
		1414	it to return all extents of a range for files with large number of
		1415	extents. The code works around all these issues if C<$count> is undef.
		1416
		1417	=item aio_group $callback->(...)
		1418
		1419	This is a very special aio request: Instead of doing something, it is a
		1420	container for other aio requests, which is useful if you want to bundle
		1421	many requests into a single, composite, request with a definite callback
		1422	and the ability to cancel the whole request with its subrequests.
		1423
		1424	Returns an object of class L<IO::AIO::GRP>. See its documentation below
		1425	for more info.
		1426
		1427	Example:
		1428
		1429	my $grp = aio_group sub {
		1430	print "all stats done\n";
		1431	};
		1432
		1433	add $grp
		1434	(aio_stat ...),
		1435	(aio_stat ...),
		1436	...;
		1437
		1438	=item aio_nop $callback->()
		1439
		1440	This is a special request - it does nothing in itself and is only used for
		1441	side effects, such as when you want to add a dummy request to a group so
		1442	that finishing the requests in the group depends on executing the given
		1443	code.
		1444
		1445	While this request does nothing, it still goes through the execution
		1446	phase and still requires a worker thread. Thus, the callback will not
		1447	be executed immediately but only after other requests in the queue have
		1448	entered their execution phase. This can be used to measure request
		1449	latency.
		1450
		1451	=item IO::AIO::aio_busy $fractional_seconds, $callback->() *NOT EXPORTED*
		1452
		1453	Mainly used for debugging and benchmarking, this aio request puts one of
		1454	the request workers to sleep for the given time.
		1455
		1456	While it is theoretically handy to have simple I/O scheduling requests
		1457	like sleep and file handle readable/writable, the overhead this creates is
		1458	immense (it blocks a thread for a long time) so do not use this function
		1459	except to put your application under artificial I/O pressure.
		1460
		1461	=back
		1462
		1463
		1464	=head2 IO::AIO::WD - multiple working directories
		1465
		1466	Your process only has one current working directory, which is used by all
		1467	threads. This makes it hard to use relative paths (some other component
		1468	could call C<chdir> at any time, and it is hard to control when the path
		1469	will be used by IO::AIO).
		1470
		1471	One solution for this is to always use absolute paths. This usually works,
		1472	but can be quite slow (the kernel has to walk the whole path on every
		1473	access), and can also be a hassle to implement.
		1474
		1475	Newer POSIX systems have a number of functions (openat, fdopendir,
		1476	futimensat and so on) that make it possible to specify working directories
		1477	per operation.
		1478
		1479	For portability, and because the clowns who "designed", or shall I write,
		1480	perpetrated this new interface were obviously half-drunk, this abstraction
		1481	cannot be perfect, though.
		1482
		1483	IO::AIO allows you to convert directory paths into a so-called IO::AIO::WD
		1484	object. This object stores the canonicalised, absolute version of the
		1485	path, and on systems that allow it, also a directory file descriptor.
		1486
		1487	Everywhere where a pathname is accepted by IO::AIO (e.g. in C<aio_stat>
		1488	or C<aio_unlink>), one can specify an array reference with an IO::AIO::WD
		1489	object and a pathname instead (or the IO::AIO::WD object alone, which
		1490	gets interpreted as C<[$wd, "."]>). If the pathname is absolute, the
		1491	IO::AIO::WD object is ignored, otherwise the pathname is resolved relative
		1492	to that IO::AIO::WD object.
		1493
		1494	For example, to get a wd object for F</etc> and then stat F<passwd>
		1495	inside, you would write:
		1496
		1497	aio_wd "/etc", sub {
		1498	my $etcdir = shift;
		1499
		1500	# although $etcdir can be undef on error, there is generally no reason
		1501	# to check for errors here, as aio_stat will fail with ENOENT
		1502	# when $etcdir is undef.
		1503
		1504	aio_stat [$etcdir, "passwd"], sub {
		1505	# yay
		1506	};
		1507	};
		1508
		1509	The fact that C<aio_wd> is a request and not a normal function shows that
		1510	creating an IO::AIO::WD object is itself a potentially blocking operation,
		1511	which is why it is done asynchronously.
		1512
		1513	To stat the directory obtained with C<aio_wd> above, one could write
		1514	either of the following three request calls:
		1515
		1516	aio_lstat "/etc" , sub { ... # pathname as normal string
		1517	aio_lstat [$wd, "."], sub { ... # "." relative to $wd (i.e. $wd itself)
		1518	aio_lstat $wd , sub { ... # shorthand for the previous
		1519
		1520	As with normal pathnames, IO::AIO keeps a copy of the working directory
		1521	object and the pathname string, so you could write the following without
		1522	causing any issues due to C<$path> getting reused:
		1523
		1524	my $path = [$wd, undef];
		1525
		1526	for my $name (qw(abc def ghi)) {
		1527	$path->[1] = $name;
		1528	aio_stat $path, sub {
		1529	# ...
		1530	};
		1531	}
		1532
		1533	There are some caveats: when directories get renamed (or deleted), the
		1534	pathname string doesn't change, so will point to the new directory (or
		1535	nowhere at all), while the directory fd, if available on the system,
		1536	will still point to the original directory. Most functions accepting a
		1537	pathname will use the directory fd on newer systems, and the string on
		1538	older systems. Some functions (such as realpath) will always rely on the
		1539	string form of the pathname.
		1540
		1541	So this functionality is mainly useful to get some protection against
		1542	C<chdir>, to easily get an absolute path out of a relative path for future
		1543	reference, and to speed up doing many operations in the same directory
		1544	(e.g. when stat'ing all files in a directory).
		1545
		1546	The following functions implement this working directory abstraction:
		1547
		1548	=over 4
		1549
		1550	=item aio_wd $pathname, $callback->($wd)
		1551
		1552	Asynchonously canonicalise the given pathname and convert it to an
		1553	IO::AIO::WD object representing it. If possible and supported on the
		1554	system, also open a directory fd to speed up pathname resolution relative
		1555	to this working directory.
		1556
		1557	If something goes wrong, then C<undef> is passwd to the callback instead
		1558	of a working directory object and C<$!> is set appropriately. Since
		1559	passing C<undef> as working directory component of a pathname fails the
		1560	request with C<ENOENT>, there is often no need for error checking in the
		1561	C<aio_wd> callback, as future requests using the value will fail in the
		1562	expected way.
		1563
		1564	=item IO::AIO::CWD
		1565
		1566	This is a compiletime constant (object) that represents the process
		1567	current working directory.
		1568
		1569	Specifying this object as working directory object for a pathname is as if
		1570	the pathname would be specified directly, without a directory object. For
		1571	example, these calls are functionally identical:
		1572
		1573	aio_stat "somefile", sub { ... };
		1574	aio_stat [IO::AIO::CWD, "somefile"], sub { ... };
		1575
		1576	=back
		1577
		1578	To recover the path associated with an IO::AIO::WD object, you can use
		1579	C<aio_realpath>:
		1580
		1581	aio_realpath $wd, sub {
		1582	warn "path is $_[0]\n";
		1583	};
		1584
		1585	Currently, C<aio_statvfs> always, and C<aio_rename> and C<aio_rmdir>
		1586	sometimes, fall back to using an absolue path.
		1587
		1588	=head2 IO::AIO::REQ CLASS
		1589
		1590	All non-aggregate C<aio_*> functions return an object of this class when
		1591	called in non-void context.
		1592
		1593	=over 4
		1594
		1595	=item cancel $req
		1596
		1597	Cancels the request, if possible. Has the effect of skipping execution
		1598	when entering the B<execute> state and skipping calling the callback when
		1599	entering the the B<result> state, but will leave the request otherwise
		1600	untouched (with the exception of readdir). That means that requests that
		1601	currently execute will not be stopped and resources held by the request
		1602	will not be freed prematurely.
		1603
		1604	=item cb $req $callback->(...)
		1605
		1606	Replace (or simply set) the callback registered to the request.
		1607
		1608	=back
		1609
		1610	=head2 IO::AIO::GRP CLASS
		1611
		1612	This class is a subclass of L<IO::AIO::REQ>, so all its methods apply to
		1613	objects of this class, too.
		1614
		1615	A IO::AIO::GRP object is a special request that can contain multiple other
		1616	aio requests.
		1617
		1618	You create one by calling the C<aio_group> constructing function with a
		1619	callback that will be called when all contained requests have entered the
		1620	C<done> state:
		1621
		1622	my $grp = aio_group sub {
		1623	print "all requests are done\n";
		1624	};
		1625
		1626	You add requests by calling the C<add> method with one or more
		1627	C<IO::AIO::REQ> objects:
		1628
		1629	$grp->add (aio_unlink "...");
		1630
		1631	add $grp aio_stat "...", sub {
		1632	$_[0] or return $grp->result ("error");
		1633
		1634	# add another request dynamically, if first succeeded
		1635	add $grp aio_open "...", sub {
		1636	$grp->result ("ok");
		1637	};
		1638	};
		1639
		1640	This makes it very easy to create composite requests (see the source of
		1641	C<aio_move> for an application) that work and feel like simple requests.
		1642
		1643	=over 4
		1644
		1645	=item * The IO::AIO::GRP objects will be cleaned up during calls to
		1646	C<IO::AIO::poll_cb>, just like any other request.
		1647
		1648	=item * They can be canceled like any other request. Canceling will cancel not
		1649	only the request itself, but also all requests it contains.
		1650
		1651	=item * They can also can also be added to other IO::AIO::GRP objects.
		1652
		1653	=item * You must not add requests to a group from within the group callback (or
		1654	any later time).
		1655
		1656	=back
		1657
		1658	Their lifetime, simplified, looks like this: when they are empty, they
		1659	will finish very quickly. If they contain only requests that are in the
		1660	C<done> state, they will also finish. Otherwise they will continue to
		1661	exist.
		1662
		1663	That means after creating a group you have some time to add requests
		1664	(precisely before the callback has been invoked, which is only done within
		1665	the C<poll_cb>). And in the callbacks of those requests, you can add
		1666	further requests to the group. And only when all those requests have
		1667	finished will the the group itself finish.
		1668
		1669	=over 4
		1670
		1671	=item add $grp ...
		1672
		1673	=item $grp->add (...)
		1674
		1675	Add one or more requests to the group. Any type of L<IO::AIO::REQ> can
		1676	be added, including other groups, as long as you do not create circular
		1677	dependencies.
		1678
		1679	Returns all its arguments.
		1680
		1681	=item $grp->cancel_subs
		1682
		1683	Cancel all subrequests and clears any feeder, but not the group request
		1684	itself. Useful when you queued a lot of events but got a result early.
		1685
		1686	The group request will finish normally (you cannot add requests to the
		1687	group).
		1688
		1689	=item $grp->result (...)
		1690
		1691	Set the result value(s) that will be passed to the group callback when all
		1692	subrequests have finished and set the groups errno to the current value
		1693	of errno (just like calling C<errno> without an error number). By default,
		1694	no argument will be passed and errno is zero.
		1695
		1696	=item $grp->errno ([$errno])
		1697
		1698	Sets the group errno value to C<$errno>, or the current value of errno
		1699	when the argument is missing.
		1700
		1701	Every aio request has an associated errno value that is restored when
		1702	the callback is invoked. This method lets you change this value from its
		1703	default (0).
		1704
		1705	Calling C<result> will also set errno, so make sure you either set C<$!>
		1706	before the call to C<result>, or call c<errno> after it.
		1707
		1708	=item feed $grp $callback->($grp)
		1709
		1710	Sets a feeder/generator on this group: every group can have an attached
		1711	generator that generates requests if idle. The idea behind this is that,
		1712	although you could just queue as many requests as you want in a group,
		1713	this might starve other requests for a potentially long time. For example,
		1714	C<aio_scandir> might generate hundreds of thousands of C<aio_stat>
		1715	requests, delaying any later requests for a long time.
		1716
		1717	To avoid this, and allow incremental generation of requests, you can
		1718	instead a group and set a feeder on it that generates those requests. The
		1719	feed callback will be called whenever there are few enough (see C<limit>,
		1720	below) requests active in the group itself and is expected to queue more
		1721	requests.
		1722
		1723	The feed callback can queue as many requests as it likes (i.e. C<add> does
		1724	not impose any limits).
		1725
		1726	If the feed does not queue more requests when called, it will be
		1727	automatically removed from the group.
		1728
		1729	If the feed limit is C<0> when this method is called, it will be set to
		1730	C<2> automatically.
		1731
		1732	Example:
		1733
		1734	# stat all files in @files, but only ever use four aio requests concurrently:
		1735
		1736	my $grp = aio_group sub { print "finished\n" };
		1737	limit $grp 4;
		1738	feed $grp sub {
		1739	my $file = pop @files
		1740	or return;
		1741
		1742	add $grp aio_stat $file, sub { ... };
		1743	};
		1744
		1745	=item limit $grp $num
		1746
		1747	Sets the feeder limit for the group: The feeder will be called whenever
		1748	the group contains less than this many requests.
		1749
		1750	Setting the limit to C<0> will pause the feeding process.
		1751
		1752	The default value for the limit is C<0>, but note that setting a feeder
		1753	automatically bumps it up to C<2>.
		1754
		1755	=back
		1756
		1757	=head2 SUPPORT FUNCTIONS
		1758
		1759	=head3 EVENT PROCESSING AND EVENT LOOP INTEGRATION
		1760
		1761	=over 4
		1762
		1763	=item $fileno = IO::AIO::poll_fileno
		1764
		1765	Return the I<request result pipe file descriptor>. This filehandle must be
		1766	polled for reading by some mechanism outside this module (e.g. EV, Glib,
		1767	select and so on, see below or the SYNOPSIS). If the pipe becomes readable
		1768	you have to call C<poll_cb> to check the results.
		1769
		1770	See C<poll_cb> for an example.
		1771
		1772	=item IO::AIO::poll_cb
		1773
		1774	Process some requests that have reached the result phase (i.e. they have
		1775	been executed but the results are not yet reported). You have to call
		1776	this "regularly" to finish outstanding requests.
		1777
		1778	Returns C<0> if all events could be processed (or there were no
		1779	events to process), or C<-1> if it returned earlier for whatever
		1780	reason. Returns immediately when no events are outstanding. The amount
		1781	of events processed depends on the settings of C<IO::AIO::max_poll_req>,
		1782	C<IO::AIO::max_poll_time> and C<IO::AIO::max_outstanding>.
		1783
		1784	If not all requests were processed for whatever reason, the poll file
		1785	descriptor will still be ready when C<poll_cb> returns, so normally you
		1786	don't have to do anything special to have it called later.
		1787
		1788	Apart from calling C<IO::AIO::poll_cb> when the event filehandle becomes
		1789	ready, it can be beneficial to call this function from loops which submit
		1790	a lot of requests, to make sure the results get processed when they become
		1791	available and not just when the loop is finished and the event loop takes
		1792	over again. This function returns very fast when there are no outstanding
		1793	requests.
		1794
		1795	Example: Install an Event watcher that automatically calls
		1796	IO::AIO::poll_cb with high priority (more examples can be found in the
		1797	SYNOPSIS section, at the top of this document):
		1798
		1799	Event->io (fd => IO::AIO::poll_fileno,
		1800	poll => 'r', async => 1,
		1801	cb => \&IO::AIO::poll_cb);
		1802
		1803	=item IO::AIO::poll_wait
		1804
		1805	Wait until either at least one request is in the result phase or no
		1806	requests are outstanding anymore.
		1807
		1808	This is useful if you want to synchronously wait for some requests to
		1809	become ready, without actually handling them.
		1810
		1811	See C<nreqs> for an example.
		1812
		1813	=item IO::AIO::poll
		1814
		1815	Waits until some requests have been handled.
		1816
		1817	Returns the number of requests processed, but is otherwise strictly
		1818	equivalent to:
		1819
		1820	IO::AIO::poll_wait, IO::AIO::poll_cb
		1821
		1822	=item IO::AIO::flush
		1823
		1824	Wait till all outstanding AIO requests have been handled.
		1825
		1826	Strictly equivalent to:
		1827
		1828	IO::AIO::poll_wait, IO::AIO::poll_cb
		1829	while IO::AIO::nreqs;
		1830
		1831	=item IO::AIO::max_poll_reqs $nreqs
		1832
		1833	=item IO::AIO::max_poll_time $seconds
		1834
		1835	These set the maximum number of requests (default C<0>, meaning infinity)
		1836	that are being processed by C<IO::AIO::poll_cb> in one call, respectively
		1837	the maximum amount of time (default C<0>, meaning infinity) spent in
		1838	C<IO::AIO::poll_cb> to process requests (more correctly the mininum amount
		1839	of time C<poll_cb> is allowed to use).
		1840
		1841	Setting C<max_poll_time> to a non-zero value creates an overhead of one
		1842	syscall per request processed, which is not normally a problem unless your
		1843	callbacks are really really fast or your OS is really really slow (I am
		1844	not mentioning Solaris here). Using C<max_poll_reqs> incurs no overhead.
		1845
		1846	Setting these is useful if you want to ensure some level of
		1847	interactiveness when perl is not fast enough to process all requests in
		1848	time.
		1849
		1850	For interactive programs, values such as C<0.01> to C<0.1> should be fine.
		1851
		1852	Example: Install an Event watcher that automatically calls
		1853	IO::AIO::poll_cb with low priority, to ensure that other parts of the
		1854	program get the CPU sometimes even under high AIO load.
		1855
		1856	# try not to spend much more than 0.1s in poll_cb
		1857	IO::AIO::max_poll_time 0.1;
		1858
		1859	# use a low priority so other tasks have priority
		1860	Event->io (fd => IO::AIO::poll_fileno,
		1861	poll => 'r', nice => 1,
		1862	cb => &IO::AIO::poll_cb);
		1863
		1864	=back
		1865
		1866	=head3 CONTROLLING THE NUMBER OF THREADS
		1867
		1868	=over
		1869
		1870	=item IO::AIO::min_parallel $nthreads
		1871
		1872	Set the minimum number of AIO threads to C<$nthreads>. The current
		1873	default is C<8>, which means eight asynchronous operations can execute
		1874	concurrently at any one time (the number of outstanding requests,
		1875	however, is unlimited).
		1876
		1877	IO::AIO starts threads only on demand, when an AIO request is queued and
		1878	no free thread exists. Please note that queueing up a hundred requests can
		1879	create demand for a hundred threads, even if it turns out that everything
		1880	is in the cache and could have been processed faster by a single thread.
		1881
		1882	It is recommended to keep the number of threads relatively low, as some
		1883	Linux kernel versions will scale negatively with the number of threads
		1884	(higher parallelity => MUCH higher latency). With current Linux 2.6
		1885	versions, 4-32 threads should be fine.
		1886
		1887	Under most circumstances you don't need to call this function, as the
		1888	module selects a default that is suitable for low to moderate load.
		1889
		1890	=item IO::AIO::max_parallel $nthreads
		1891
		1892	Sets the maximum number of AIO threads to C<$nthreads>. If more than the
		1893	specified number of threads are currently running, this function kills
		1894	them. This function blocks until the limit is reached.
		1895
		1896	While C<$nthreads> are zero, aio requests get queued but not executed
		1897	until the number of threads has been increased again.
		1898
		1899	This module automatically runs C<max_parallel 0> at program end, to ensure
		1900	that all threads are killed and that there are no outstanding requests.
		1901
		1902	Under normal circumstances you don't need to call this function.
		1903
		1904	=item IO::AIO::max_idle $nthreads
		1905
		1906	Limit the number of threads (default: 4) that are allowed to idle
		1907	(i.e., threads that did not get a request to process within the idle
		1908	timeout (default: 10 seconds). That means if a thread becomes idle while
		1909	C<$nthreads> other threads are also idle, it will free its resources and
		1910	exit.
		1911
		1912	This is useful when you allow a large number of threads (e.g. 100 or 1000)
		1913	to allow for extremely high load situations, but want to free resources
		1914	under normal circumstances (1000 threads can easily consume 30MB of RAM).
		1915
		1916	The default is probably ok in most situations, especially if thread
		1917	creation is fast. If thread creation is very slow on your system you might
		1918	want to use larger values.
		1919
		1920	=item IO::AIO::idle_timeout $seconds
		1921
		1922	Sets the minimum idle timeout (default 10) after which worker threads are
		1923	allowed to exit. SEe C<IO::AIO::max_idle>.
		1924
		1925	=item IO::AIO::max_outstanding $maxreqs
		1926
		1927	Sets the maximum number of outstanding requests to C<$nreqs>. If
		1928	you do queue up more than this number of requests, the next call to
		1929	C<IO::AIO::poll_cb> (and other functions calling C<poll_cb>, such as
		1930	C<IO::AIO::flush> or C<IO::AIO::poll>) will block until the limit is no
		1931	longer exceeded.
		1932
		1933	In other words, this setting does not enforce a queue limit, but can be
		1934	used to make poll functions block if the limit is exceeded.
		1935
		1936	This is a very bad function to use in interactive programs because it
		1937	blocks, and a bad way to reduce concurrency because it is inexact: Better
		1938	use an C<aio_group> together with a feed callback.
		1939
		1940	Its main use is in scripts without an event loop - when you want to stat
		1941	a lot of files, you can write somehting like this:
		1942
		1943	IO::AIO::max_outstanding 32;
		1944
		1945	for my $path (...) {
		1946	aio_stat $path , ...;
		1947	IO::AIO::poll_cb;
		1948	}
		1949
		1950	IO::AIO::flush;
		1951
		1952	The call to C<poll_cb> inside the loop will normally return instantly, but
		1953	as soon as more thna C<32> reqeusts are in-flight, it will block until
		1954	some requests have been handled. This keeps the loop from pushing a large
		1955	number of C<aio_stat> requests onto the queue.
		1956
		1957	The default value for C<max_outstanding> is very large, so there is no
		1958	practical limit on the number of outstanding requests.
		1959
		1960	=back
		1961
		1962	=head3 STATISTICAL INFORMATION
		1963
		1964	=over
		1965
		1966	=item IO::AIO::nreqs
		1967
		1968	Returns the number of requests currently in the ready, execute or pending
		1969	states (i.e. for which their callback has not been invoked yet).
		1970
		1971	Example: wait till there are no outstanding requests anymore:
		1972
		1973	IO::AIO::poll_wait, IO::AIO::poll_cb
		1974	while IO::AIO::nreqs;
		1975
		1976	=item IO::AIO::nready
		1977
		1978	Returns the number of requests currently in the ready state (not yet
		1979	executed).
		1980
		1981	=item IO::AIO::npending
		1982
		1983	Returns the number of requests currently in the pending state (executed,
		1984	but not yet processed by poll_cb).
		1985
		1986	=back
		1987
		1988	=head3 MISCELLANEOUS FUNCTIONS
		1989
		1990	IO::AIO implements some functions that are useful when you want to use
		1991	some "Advanced I/O" function not available to in Perl, without going the
		1992	"Asynchronous I/O" route. Many of these have an asynchronous C<aio_*>
		1993	counterpart.
		1994
		1995	=over 4
		1996
		1997	=item IO::AIO::sendfile $ofh, $ifh, $offset, $count
		1998
		1999	Calls the C<eio_sendfile_sync> function, which is like C<aio_sendfile>,
		2000	but is blocking (this makes most sense if you know the input data is
		2001	likely cached already and the output filehandle is set to non-blocking
		2002	operations).
		2003
		2004	Returns the number of bytes copied, or C<-1> on error.
		2005
		2006	=item IO::AIO::fadvise $fh, $offset, $len, $advice
		2007
		2008	Simply calls the C<posix_fadvise> function (see its
		2009	manpage for details). The following advice constants are
		2010	available: C<IO::AIO::FADV_NORMAL>, C<IO::AIO::FADV_SEQUENTIAL>,
		2011	C<IO::AIO::FADV_RANDOM>, C<IO::AIO::FADV_NOREUSE>,
		2012	C<IO::AIO::FADV_WILLNEED>, C<IO::AIO::FADV_DONTNEED>.
		2013
		2014	On systems that do not implement C<posix_fadvise>, this function returns
		2015	ENOSYS, otherwise the return value of C<posix_fadvise>.
		2016
		2017	=item IO::AIO::madvise $scalar, $offset, $len, $advice
		2018
		2019	Simply calls the C<posix_madvise> function (see its
		2020	manpage for details). The following advice constants are
		2021	available: C<IO::AIO::MADV_NORMAL>, C<IO::AIO::MADV_SEQUENTIAL>,
		2022	C<IO::AIO::MADV_RANDOM>, C<IO::AIO::MADV_WILLNEED>, C<IO::AIO::MADV_DONTNEED>.
		2023
		2024	On systems that do not implement C<posix_madvise>, this function returns
		2025	ENOSYS, otherwise the return value of C<posix_madvise>.
		2026
		2027	=item IO::AIO::mprotect $scalar, $offset, $len, $protect
		2028
		2029	Simply calls the C<mprotect> function on the preferably AIO::mmap'ed
		2030	$scalar (see its manpage for details). The following protect
		2031	constants are available: C<IO::AIO::PROT_NONE>, C<IO::AIO::PROT_READ>,
		2032	C<IO::AIO::PROT_WRITE>, C<IO::AIO::PROT_EXEC>.
		2033
		2034	On systems that do not implement C<mprotect>, this function returns
		2035	ENOSYS, otherwise the return value of C<mprotect>.
		2036
		2037	=item IO::AIO::mmap $scalar, $length, $prot, $flags, $fh[, $offset]
		2038
		2039	Memory-maps a file (or anonymous memory range) and attaches it to the
		2040	given C<$scalar>, which will act like a string scalar. Returns true on
		2041	success, and false otherwise.
		2042
		2043	The only operations allowed on the scalar are C<substr>/C<vec> that don't
		2044	change the string length, and most read-only operations such as copying it
		2045	or searching it with regexes and so on.
		2046
		2047	Anything else is unsafe and will, at best, result in memory leaks.
		2048
		2049	The memory map associated with the C<$scalar> is automatically removed
		2050	when the C<$scalar> is destroyed, or when the C<IO::AIO::mmap> or
		2051	C<IO::AIO::munmap> functions are called.
		2052
		2053	This calls the C<mmap>(2) function internally. See your system's manual
		2054	page for details on the C<$length>, C<$prot> and C<$flags> parameters.
		2055
		2056	The C<$length> must be larger than zero and smaller than the actual
		2057	filesize.
		2058
		2059	C<$prot> is a combination of C<IO::AIO::PROT_NONE>, C<IO::AIO::PROT_EXEC>,
		2060	C<IO::AIO::PROT_READ> and/or C<IO::AIO::PROT_WRITE>,
		2061
		2062	C<$flags> can be a combination of C<IO::AIO::MAP_SHARED> or
		2063	C<IO::AIO::MAP_PRIVATE>, or a number of system-specific flags (when
		2064	not available, the are defined as 0): C<IO::AIO::MAP_ANONYMOUS>
		2065	(which is set to C<MAP_ANON> if your system only provides this
		2066	constant), C<IO::AIO::MAP_HUGETLB>, C<IO::AIO::MAP_LOCKED>,
		2067	C<IO::AIO::MAP_NORESERVE>, C<IO::AIO::MAP_POPULATE> or
		2068	C<IO::AIO::MAP_NONBLOCK>
		2069
		2070	If C<$fh> is C<undef>, then a file descriptor of C<-1> is passed.
		2071
		2072	C<$offset> is the offset from the start of the file - it generally must be
		2073	a multiple of C<IO::AIO::PAGESIZE> and defaults to C<0>.
		2074
		2075	Example:
		2076
		2077	use Digest::MD5;
		2078	use IO::AIO;
		2079
		2080	open my $fh, "<verybigfile"
		2081	or die "$!";
		2082
		2083	IO::AIO::mmap my $data, -s $fh, IO::AIO::PROT_READ, IO::AIO::MAP_SHARED, $fh
		2084	or die "verybigfile: $!";
		2085
		2086	my $fast_md5 = md5 $data;
		2087
		2088	=item IO::AIO::munmap $scalar
		2089
		2090	Removes a previous mmap and undefines the C<$scalar>.
		2091
		2092	=item IO::AIO::munlock $scalar, $offset = 0, $length = undef
		2093
		2094	Calls the C<munlock> function, undoing the effects of a previous
		2095	C<aio_mlock> call (see its description for details).
		2096
		2097	=item IO::AIO::munlockall
		2098
		2099	Calls the C<munlockall> function.
		2100
		2101	On systems that do not implement C<munlockall>, this function returns
		2102	ENOSYS, otherwise the return value of C<munlockall>.
		2103
		2104	=item IO::AIO::splice $r_fh, $r_off, $w_fh, $w_off, $length, $flags
		2105
		2106	Calls the GNU/Linux C<splice(2)> syscall, if available. If C<$r_off> or
		2107	C<$w_off> are C<undef>, then C<NULL> is passed for these, otherwise they
		2108	should be the file offset.
		2109
		2110	C<$r_fh> and C<$w_fh> should not refer to the same file, as splice might
		2111	silently corrupt the data in this case.
		2112
		2113	The following symbol flag values are available: C<IO::AIO::SPLICE_F_MOVE>,
		2114	C<IO::AIO::SPLICE_F_NONBLOCK>, C<IO::AIO::SPLICE_F_MORE> and
		2115	C<IO::AIO::SPLICE_F_GIFT>.
		2116
		2117	See the C<splice(2)> manpage for details.
		2118
		2119	=item IO::AIO::tee $r_fh, $w_fh, $length, $flags
		2120
		2121	Calls the GNU/Linux C<tee(2)> syscall, see its manpage and the
		2122	description for C<IO::AIO::splice> above for details.
		2123
		2124	=item $actual_size = IO::AIO::pipesize $r_fh[, $new_size]
		2125
		2126	Attempts to query or change the pipe buffer size. Obviously works only
		2127	on pipes, and currently works only on GNU/Linux systems, and fails with
		2128	C<-1>/C<ENOSYS> everywhere else. If anybody knows how to influence pipe buffer
		2129	size on other systems, drop me a note.
		2130
		2131	=item ($rfh, $wfh) = IO::AIO::pipe2 [$flags]
		2132
		2133	This is a direct interface to the Linux L<pipe2(2)> system call. If
		2134	C<$flags> is missing or C<0>, then this should be the same as a call to
		2135	perl's built-in C<pipe> function and create a new pipe, and should work
		2136	everywhere.
		2137
		2138	If C<$flags> is non-zero, it tries to invoke the pipe2 system call with
		2139	the given flags (Linux 2.6.27, glibc 2.9).
		2140
		2141	On success, the read and write file handles are returned.
		2142
		2143	On error, nothing will be returned. If the pipe2 syscall is missing and
		2144	C<$flags> is non-zero, fails with C<ENOSYS>.
		2145
		2146	Please refer to L<pipe2(2)> for more info on the C<$flags>, but at the
		2147	time of this writing, C<IO::AIO::O_CLOEXEC>, C<IO::AIO::O_NONBLOCK> and
		2148	C<IO::AIO::O_DIRECT> (Linux 3.4, for packet-based pipes) were supported.
		2149
		2150	=back
		2151
		2152	=cut
		2153
239	min_parallel 4;	2154	min_parallel 8;
240		2155
241	END {	2156	END { flush }
242	max_parallel 0;
243	}
244		2157
245	1;	2158	1;
246		2159
		2160	=head1 EVENT LOOP INTEGRATION
		2161
		2162	It is recommended to use L<AnyEvent::AIO> to integrate IO::AIO
		2163	automatically into many event loops:
		2164
		2165	# AnyEvent integration (EV, Event, Glib, Tk, POE, urxvt, pureperl...)
		2166	use AnyEvent::AIO;
		2167
		2168	You can also integrate IO::AIO manually into many event loops, here are
		2169	some examples of how to do this:
		2170
		2171	# EV integration
		2172	my $aio_w = EV::io IO::AIO::poll_fileno, EV::READ, \&IO::AIO::poll_cb;
		2173
		2174	# Event integration
		2175	Event->io (fd => IO::AIO::poll_fileno,
		2176	poll => 'r',
		2177	cb => \&IO::AIO::poll_cb);
		2178
		2179	# Glib/Gtk2 integration
		2180	add_watch Glib::IO IO::AIO::poll_fileno,
		2181	in => sub { IO::AIO::poll_cb; 1 };
		2182
		2183	# Tk integration
		2184	Tk::Event::IO->fileevent (IO::AIO::poll_fileno, "",
		2185	readable => \&IO::AIO::poll_cb);
		2186
		2187	# Danga::Socket integration
		2188	Danga::Socket->AddOtherFds (IO::AIO::poll_fileno =>
		2189	\&IO::AIO::poll_cb);
		2190
		2191	=head2 FORK BEHAVIOUR
		2192
		2193	Usage of pthreads in a program changes the semantics of fork
		2194	considerably. Specifically, only async-safe functions can be called after
		2195	fork. Perl doesn't know about this, so in general, you cannot call fork
		2196	with defined behaviour in perl if pthreads are involved. IO::AIO uses
		2197	pthreads, so this applies, but many other extensions and (for inexplicable
		2198	reasons) perl itself often is linked against pthreads, so this limitation
		2199	applies to quite a lot of perls.
		2200
		2201	This module no longer tries to fight your OS, or POSIX. That means IO::AIO
		2202	only works in the process that loaded it. Forking is fully supported, but
		2203	using IO::AIO in the child is not.
		2204
		2205	You might get around by not I<using> IO::AIO before (or after)
		2206	forking. You could also try to call the L<IO::AIO::reinit> function in the
		2207	child:
		2208
		2209	=over 4
		2210
		2211	=item IO::AIO::reinit
		2212
		2213	Abandons all current requests and I/O threads and simply reinitialises all
		2214	data structures. This is not an operation supported by any standards, but
		2215	happens to work on GNU/Linux and some newer BSD systems.
		2216
		2217	The only reasonable use for this function is to call it after forking, if
		2218	C<IO::AIO> was used in the parent. Calling it while IO::AIO is active in
		2219	the process will result in undefined behaviour. Calling it at any time
		2220	will also result in any undefined (by POSIX) behaviour.
		2221
247	=back	2222	=back
248		2223
		2224	=head2 MEMORY USAGE
		2225
		2226	Per-request usage:
		2227
		2228	Each aio request uses - depending on your architecture - around 100-200
		2229	bytes of memory. In addition, stat requests need a stat buffer (possibly
		2230	a few hundred bytes), readdir requires a result buffer and so on. Perl
		2231	scalars and other data passed into aio requests will also be locked and
		2232	will consume memory till the request has entered the done state.
		2233
		2234	This is not awfully much, so queuing lots of requests is not usually a
		2235	problem.
		2236
		2237	Per-thread usage:
		2238
		2239	In the execution phase, some aio requests require more memory for
		2240	temporary buffers, and each thread requires a stack and other data
		2241	structures (usually around 16k-128k, depending on the OS).
		2242
249	=head1 BUGS	2243	=head1 KNOWN BUGS
250		2244
251	- could be optimized to use more semaphores instead of filehandles.	2245	Known bugs will be fixed in the next release.
252		2246
253	=head1 SEE ALSO	2247	=head1 SEE ALSO
254		2248
255	L<Coro>, L<Linux::AIO>.	2249	L<AnyEvent::AIO> for easy integration into event loops, L<Coro::AIO> for a
		2250	more natural syntax.
256		2251
257	=head1 AUTHOR	2252	=head1 AUTHOR
258		2253
259	Marc Lehmann <schmorp@schmorp.de>	2254	Marc Lehmann <schmorp@schmorp.de>
260	http://home.schmorp.de/	2255	http://home.schmorp.de/

Diff Legend

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