ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/IO-AIO/AIO.pm

Comparing IO-AIO/AIO.pm (file contents):
Revision 1.39 by root, Sun Aug 28 11:05:50 2005 UTC vs.
Revision 1.299 by root, Fri Dec 28 12:09:50 2018 UTC

1	=head1 NAME	1	=head1 NAME
2		2
3	IO::AIO - Asynchronous Input/Output	3	IO::AIO - Asynchronous/Advanced Input/Output
4		4
5	=head1 SYNOPSIS	5	=head1 SYNOPSIS
6		6
7	use IO::AIO;	7	use IO::AIO;
8		8
9	aio_open "/etc/passwd", O_RDONLY, 0, sub {	9	aio_open "/etc/passwd", IO::AIO::O_RDONLY, 0, sub {
10	my ($fh) = @_;	10	my $fh = shift
		11	or die "/etc/passwd: $!";
11	...	12	...
12	};	13	};
13		14
14	aio_unlink "/tmp/file", sub { };	15	aio_unlink "/tmp/file", sub { };
15		16
16	aio_read $fh, 30000, 1024, $buffer, 0, sub {	17	aio_read $fh, 30000, 1024, $buffer, 0, sub {
17	$_[0] > 0 or die "read error: $!";	18	$_[0] > 0 or die "read error: $!";
18	};	19	};
19		20
20	# Event	21	# version 2+ has request and group objects
21	Event->io (fd => IO::AIO::poll_fileno,	22	use IO::AIO 2;
22	poll => 'r',
23	cb => \&IO::AIO::poll_cb);
24		23
25	# Glib/Gtk2	24	aioreq_pri 4; # give next request a very high priority
26	add_watch Glib::IO IO::AIO::poll_fileno,	25	my $req = aio_unlink "/tmp/file", sub { };
27	in => sub { IO::AIO::poll_cb; 1 };	26	$req->cancel; # cancel request if still in queue
28		27
29	# Tk	28	my $grp = aio_group sub { print "all stats done\n" };
30	Tk::Event::IO->fileevent (IO::AIO::poll_fileno, "",	29	add $grp aio_stat "..." for ...;
31	readable => \&IO::AIO::poll_cb);
32
33	# Danga::Socket
34	Danga::Socket->AddOtherFds (IO::AIO::poll_fileno =>
35	\&IO::AIO::poll_cb);
36
37		30
38	=head1 DESCRIPTION	31	=head1 DESCRIPTION
39		32
40	This module implements asynchronous I/O using whatever means your	33	This module implements asynchronous I/O using whatever means your
41	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>).
42		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
43	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
44	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
45	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
46	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
47	aio functions available on many operating systems. However, they are often	57	functions available on many operating systems. However, they are often
48	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
49	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
50	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.
51		62
		63	In addition to asynchronous I/O, this module also exports some rather
		64	arcane interfaces, such as C<madvise> or linux's C<splice> system call,
		65	which is why the C<A> in C<AIO> can also mean I<advanced>.
		66
52	Although the module will work with in the presence of other threads, it is	67	Although the module will work in the presence of other (Perl-) threads,
53	currently not reentrant, so use appropriate locking yourself, always call	68	it is currently not reentrant in any way, so use appropriate locking
54	C<poll_cb> from within the same thread, or never call C<poll_cb> (or other	69	yourself, always call C<poll_cb> from within the same thread, or never
55	C<aio_> functions) recursively.	70	call C<poll_cb> (or other C<aio_> functions) recursively.
		71
		72	=head2 EXAMPLE
		73
		74	This is a simple example that uses the EV module and loads
		75	F</etc/passwd> asynchronously:
		76
		77	use EV;
		78	use IO::AIO;
		79
		80	# register the IO::AIO callback with EV
		81	my $aio_w = EV::io IO::AIO::poll_fileno, EV::READ, \&IO::AIO::poll_cb;
		82
		83	# queue the request to open /etc/passwd
		84	aio_open "/etc/passwd", IO::AIO::O_RDONLY, 0, sub {
		85	my $fh = shift
		86	or die "error while opening: $!";
		87
		88	# stat'ing filehandles is generally non-blocking
		89	my $size = -s $fh;
		90
		91	# queue a request to read the file
		92	my $contents;
		93	aio_read $fh, 0, $size, $contents, 0, sub {
		94	$_[0] == $size
		95	or die "short read: $!";
		96
		97	close $fh;
		98
		99	# file contents now in $contents
		100	print $contents;
		101
		102	# exit event loop and program
		103	EV::break;
		104	};
		105	};
		106
		107	# possibly queue up other requests, or open GUI windows,
		108	# check for sockets etc. etc.
		109
		110	# process events as long as there are some:
		111	EV::run;
		112
		113	=head1 REQUEST ANATOMY AND LIFETIME
		114
		115	Every C<aio_*> function creates a request. which is a C data structure not
		116	directly visible to Perl.
		117
		118	If called in non-void context, every request function returns a Perl
		119	object representing the request. In void context, nothing is returned,
		120	which saves a bit of memory.
		121
		122	The perl object is a fairly standard ref-to-hash object. The hash contents
		123	are not used by IO::AIO so you are free to store anything you like in it.
		124
		125	During their existance, aio requests travel through the following states,
		126	in order:
		127
		128	=over 4
		129
		130	=item ready
		131
		132	Immediately after a request is created it is put into the ready state,
		133	waiting for a thread to execute it.
		134
		135	=item execute
		136
		137	A thread has accepted the request for processing and is currently
		138	executing it (e.g. blocking in read).
		139
		140	=item pending
		141
		142	The request has been executed and is waiting for result processing.
		143
		144	While request submission and execution is fully asynchronous, result
		145	processing is not and relies on the perl interpreter calling C<poll_cb>
		146	(or another function with the same effect).
		147
		148	=item result
		149
		150	The request results are processed synchronously by C<poll_cb>.
		151
		152	The C<poll_cb> function will process all outstanding aio requests by
		153	calling their callbacks, freeing memory associated with them and managing
		154	any groups they are contained in.
		155
		156	=item done
		157
		158	Request has reached the end of its lifetime and holds no resources anymore
		159	(except possibly for the Perl object, but its connection to the actual
		160	aio request is severed and calling its methods will either do nothing or
		161	result in a runtime error).
		162
		163	=back
56		164
57	=cut	165	=cut
58		166
59	package IO::AIO;	167	package IO::AIO;
60		168
61	no warnings;	169	use Carp ();
		170
		171	use common::sense;
62		172
63	use base 'Exporter';	173	use base 'Exporter';
64		174
65	use Fcntl ();
66
67	BEGIN {	175	BEGIN {
68	$VERSION = 1.6;	176	our $VERSION = 4.7;
69		177
70	@EXPORT = qw(aio_sendfile aio_read aio_write aio_open aio_close aio_stat	178	our @AIO_REQ = qw(aio_sendfile aio_seek aio_read aio_write aio_open aio_close
71	aio_lstat aio_unlink aio_rmdir aio_readdir aio_symlink	179	aio_stat aio_lstat aio_unlink aio_rmdir aio_readdir aio_readdirx
72	aio_fsync aio_fdatasync aio_readahead);	180	aio_scandir aio_symlink aio_readlink aio_realpath aio_fcntl aio_ioctl
73	@EXPORT_OK = qw(poll_fileno poll_cb min_parallel max_parallel	181	aio_sync aio_fsync aio_syncfs aio_fdatasync aio_sync_file_range
74	max_outstanding nreqs);	182	aio_pathsync aio_readahead aio_fiemap aio_allocate
		183	aio_rename aio_rename2 aio_link aio_move aio_copy aio_group
		184	aio_nop aio_mknod aio_load aio_rmtree aio_mkdir aio_chown
		185	aio_chmod aio_utime aio_truncate
		186	aio_msync aio_mtouch aio_mlock aio_mlockall
		187	aio_statvfs
		188	aio_slurp
		189	aio_wd);
		190
		191	our @EXPORT = (@AIO_REQ, qw(aioreq_pri aioreq_nice));
		192	our @EXPORT_OK = qw(poll_fileno poll_cb poll_wait flush
		193	min_parallel max_parallel max_idle idle_timeout
		194	nreqs nready npending nthreads
		195	max_poll_time max_poll_reqs
		196	sendfile fadvise madvise
		197	mmap munmap mremap munlock munlockall);
		198
		199	push @AIO_REQ, qw(aio_busy); # not exported
		200
		201	@IO::AIO::GRP::ISA = 'IO::AIO::REQ';
75		202
76	require XSLoader;	203	require XSLoader;
77	XSLoader::load IO::AIO, $VERSION;	204	XSLoader::load ("IO::AIO", $VERSION);
78	}	205	}
79		206
80	=head1 FUNCTIONS	207	=head1 FUNCTIONS
81		208
82	=head2 AIO FUNCTIONS	209	=head2 QUICK OVERVIEW
		210
		211	This section simply lists the prototypes most of the functions for
		212	quick reference. See the following sections for function-by-function
		213	documentation.
		214
		215	aio_wd $pathname, $callback->($wd)
		216	aio_open $pathname, $flags, $mode, $callback->($fh)
		217	aio_close $fh, $callback->($status)
		218	aio_seek $fh,$offset,$whence, $callback->($offs)
		219	aio_read $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
		220	aio_write $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
		221	aio_sendfile $out_fh, $in_fh, $in_offset, $length, $callback->($retval)
		222	aio_readahead $fh,$offset,$length, $callback->($retval)
		223	aio_stat $fh_or_path, $callback->($status)
		224	aio_lstat $fh, $callback->($status)
		225	aio_statvfs $fh_or_path, $callback->($statvfs)
		226	aio_utime $fh_or_path, $atime, $mtime, $callback->($status)
		227	aio_chown $fh_or_path, $uid, $gid, $callback->($status)
		228	aio_chmod $fh_or_path, $mode, $callback->($status)
		229	aio_truncate $fh_or_path, $offset, $callback->($status)
		230	aio_allocate $fh, $mode, $offset, $len, $callback->($status)
		231	aio_fiemap $fh, $start, $length, $flags, $count, $cb->(\@extents)
		232	aio_unlink $pathname, $callback->($status)
		233	aio_mknod $pathname, $mode, $dev, $callback->($status)
		234	aio_link $srcpath, $dstpath, $callback->($status)
		235	aio_symlink $srcpath, $dstpath, $callback->($status)
		236	aio_readlink $pathname, $callback->($link)
		237	aio_realpath $pathname, $callback->($path)
		238	aio_rename $srcpath, $dstpath, $callback->($status)
		239	aio_rename2 $srcpath, $dstpath, $flags, $callback->($status)
		240	aio_mkdir $pathname, $mode, $callback->($status)
		241	aio_rmdir $pathname, $callback->($status)
		242	aio_readdir $pathname, $callback->($entries)
		243	aio_readdirx $pathname, $flags, $callback->($entries, $flags)
		244	IO::AIO::READDIR_DENTS IO::AIO::READDIR_DIRS_FIRST
		245	IO::AIO::READDIR_STAT_ORDER IO::AIO::READDIR_FOUND_UNKNOWN
		246	aio_scandir $pathname, $maxreq, $callback->($dirs, $nondirs)
		247	aio_load $pathname, $data, $callback->($status)
		248	aio_copy $srcpath, $dstpath, $callback->($status)
		249	aio_move $srcpath, $dstpath, $callback->($status)
		250	aio_rmtree $pathname, $callback->($status)
		251	aio_fcntl $fh, $cmd, $arg, $callback->($status)
		252	aio_ioctl $fh, $request, $buf, $callback->($status)
		253	aio_sync $callback->($status)
		254	aio_syncfs $fh, $callback->($status)
		255	aio_fsync $fh, $callback->($status)
		256	aio_fdatasync $fh, $callback->($status)
		257	aio_sync_file_range $fh, $offset, $nbytes, $flags, $callback->($status)
		258	aio_pathsync $pathname, $callback->($status)
		259	aio_msync $scalar, $offset = 0, $length = undef, flags = MS_SYNC, $callback->($status)
		260	aio_mtouch $scalar, $offset = 0, $length = undef, flags = 0, $callback->($status)
		261	aio_mlock $scalar, $offset = 0, $length = undef, $callback->($status)
		262	aio_mlockall $flags, $callback->($status)
		263	aio_group $callback->(...)
		264	aio_nop $callback->()
		265
		266	$prev_pri = aioreq_pri [$pri]
		267	aioreq_nice $pri_adjust
		268
		269	IO::AIO::poll_wait
		270	IO::AIO::poll_cb
		271	IO::AIO::poll
		272	IO::AIO::flush
		273	IO::AIO::max_poll_reqs $nreqs
		274	IO::AIO::max_poll_time $seconds
		275	IO::AIO::min_parallel $nthreads
		276	IO::AIO::max_parallel $nthreads
		277	IO::AIO::max_idle $nthreads
		278	IO::AIO::idle_timeout $seconds
		279	IO::AIO::max_outstanding $maxreqs
		280	IO::AIO::nreqs
		281	IO::AIO::nready
		282	IO::AIO::npending
		283	$nfd = IO::AIO::get_fdlimit [EXPERIMENTAL]
		284	IO::AIO::min_fdlimit $nfd [EXPERIMENTAL]
		285
		286	IO::AIO::sendfile $ofh, $ifh, $offset, $count
		287	IO::AIO::fadvise $fh, $offset, $len, $advice
		288	IO::AIO::mmap $scalar, $length, $prot, $flags[, $fh[, $offset]]
		289	IO::AIO::munmap $scalar
		290	IO::AIO::mremap $scalar, $new_length, $flags[, $new_address]
		291	IO::AIO::madvise $scalar, $offset, $length, $advice
		292	IO::AIO::mprotect $scalar, $offset, $length, $protect
		293	IO::AIO::munlock $scalar, $offset = 0, $length = undef
		294	IO::AIO::munlockall
		295
		296	=head2 API NOTES
83		297
84	All the C<aio_*> calls are more or less thin wrappers around the syscall	298	All the C<aio_*> calls are more or less thin wrappers around the syscall
85	with the same name (sans C<aio_>). The arguments are similar or identical,	299	with the same name (sans C<aio_>). The arguments are similar or identical,
86	and they all accept an additional (and optional) C<$callback> argument	300	and they all accept an additional (and optional) C<$callback> argument
87	which must be a code reference. This code reference will get called with	301	which must be a code reference. This code reference will be called after
88	the syscall return code (e.g. most syscalls return C<-1> on error, unlike	302	the syscall has been executed in an asynchronous fashion. The results
89	perl, which usually delivers "false") as it's sole argument when the given	303	of the request will be passed as arguments to the callback (and, if an
90	syscall has been executed asynchronously.	304	error occured, in C<$!>) - for most requests the syscall return code (e.g.
		305	most syscalls return C<-1> on error, unlike perl, which usually delivers
		306	"false").
		307
		308	Some requests (such as C<aio_readdir>) pass the actual results and
		309	communicate failures by passing C<undef>.
91		310
92	All functions expecting a filehandle keep a copy of the filehandle	311	All functions expecting a filehandle keep a copy of the filehandle
93	internally until the request has finished.	312	internally until the request has finished.
94		313
		314	All functions return request objects of type L<IO::AIO::REQ> that allow
		315	further manipulation of those requests while they are in-flight.
		316
95	The pathnames you pass to these routines I<must> be absolute and	317	The pathnames you pass to these routines I<should> be absolute. The
96	encoded in byte form. The reason for the former is that at the time the	318	reason for this is that at the time the request is being executed, the
97	request is being executed, the current working directory could have	319	current working directory could have changed. Alternatively, you can
98	changed. Alternatively, you can make sure that you never change the	320	make sure that you never change the current working directory anywhere
99	current working directory.	321	in the program and then use relative paths. You can also take advantage
		322	of IO::AIOs working directory abstraction, that lets you specify paths
		323	relative to some previously-opened "working directory object" - see the
		324	description of the C<IO::AIO::WD> class later in this document.
100		325
101	To encode pathnames to byte form, either make sure you either: a)	326	To encode pathnames as octets, either make sure you either: a) always pass
102	always pass in filenames you got from outside (command line, readdir	327	in filenames you got from outside (command line, readdir etc.) without
103	etc.), b) are ASCII or ISO 8859-1, c) use the Encode module and encode	328	tinkering, b) are in your native filesystem encoding, c) use the Encode
104	your pathnames to the locale (or other) encoding in effect in the user	329	module and encode your pathnames to the locale (or other) encoding in
105	environment, d) use Glib::filename_from_unicode on unicode filenames or e)	330	effect in the user environment, d) use Glib::filename_from_unicode on
106	use something else.	331	unicode filenames or e) use something else to ensure your scalar has the
		332	correct contents.
		333
		334	This works, btw. independent of the internal UTF-8 bit, which IO::AIO
		335	handles correctly whether it is set or not.
		336
		337	=head2 AIO REQUEST FUNCTIONS
107		338
108	=over 4	339	=over 4
109		340
		341	=item $prev_pri = aioreq_pri [$pri]
		342
		343	Returns the priority value that would be used for the next request and, if
		344	C<$pri> is given, sets the priority for the next aio request.
		345
		346	The default priority is C<0>, the minimum and maximum priorities are C<-4>
		347	and C<4>, respectively. Requests with higher priority will be serviced
		348	first.
		349
		350	The priority will be reset to C<0> after each call to one of the C<aio_*>
		351	functions.
		352
		353	Example: open a file with low priority, then read something from it with
		354	higher priority so the read request is serviced before other low priority
		355	open requests (potentially spamming the cache):
		356
		357	aioreq_pri -3;
		358	aio_open ..., sub {
		359	return unless $_[0];
		360
		361	aioreq_pri -2;
		362	aio_read $_[0], ..., sub {
		363	...
		364	};
		365	};
		366
		367
		368	=item aioreq_nice $pri_adjust
		369
		370	Similar to C<aioreq_pri>, but subtracts the given value from the current
		371	priority, so the effect is cumulative.
		372
		373
110	=item aio_open $pathname, $flags, $mode, $callback	374	=item aio_open $pathname, $flags, $mode, $callback->($fh)
111		375
112	Asynchronously open or create a file and call the callback with a newly	376	Asynchronously open or create a file and call the callback with a newly
113	created filehandle for the file.	377	created filehandle for the file (or C<undef> in case of an error).
114		378
115	The pathname passed to C<aio_open> must be absolute. See API NOTES, above,	379	The pathname passed to C<aio_open> must be absolute. See API NOTES, above,
116	for an explanation.	380	for an explanation.
117		381
118	The C<$flags> argument is a bitmask. See the C<Fcntl> module for a	382	The C<$flags> argument is a bitmask. See the C<Fcntl> module for a
119	list. They are the same as used by C<sysopen>.	383	list. They are the same as used by C<sysopen>.
120		384
121	Likewise, C<$mode> specifies the mode of the newly created file, if it	385	Likewise, C<$mode> specifies the mode of the newly created file, if it
122	didn't exist and C<O_CREAT> has been given, just like perl's C<sysopen>,	386	didn't exist and C<O_CREAT> has been given, just like perl's C<sysopen>,
123	except that it is mandatory (i.e. use C<0> if you don't create new files,	387	except that it is mandatory (i.e. use C<0> if you don't create new files,
124	and C<0666> or C<0777> if you do).	388	and C<0666> or C<0777> if you do). Note that the C<$mode> will be modified
		389	by the umask in effect then the request is being executed, so better never
		390	change the umask.
125		391
126	Example:	392	Example:
127		393
128	aio_open "/etc/passwd", O_RDONLY, 0, sub {	394	aio_open "/etc/passwd", IO::AIO::O_RDONLY, 0, sub {
129	if ($_[0]) {	395	if ($_[0]) {
130	print "open successful, fh is $_[0]\n";	396	print "open successful, fh is $_[0]\n";
131	...	397	...
132	} else {	398	} else {
133	die "open failed: $!\n";	399	die "open failed: $!\n";
134	}	400	}
135	};	401	};
136		402
		403	In addition to all the common open modes/flags (C<O_RDONLY>, C<O_WRONLY>,
		404	C<O_RDWR>, C<O_CREAT>, C<O_TRUNC>, C<O_EXCL> and C<O_APPEND>), the
		405	following POSIX and non-POSIX constants are available (missing ones on
		406	your system are, as usual, C<0>):
		407
		408	C<O_ASYNC>, C<O_DIRECT>, C<O_NOATIME>, C<O_CLOEXEC>, C<O_NOCTTY>, C<O_NOFOLLOW>,
		409	C<O_NONBLOCK>, C<O_EXEC>, C<O_SEARCH>, C<O_DIRECTORY>, C<O_DSYNC>,
		410	C<O_RSYNC>, C<O_SYNC>, C<O_PATH>, C<O_TMPFILE>, C<O_TTY_INIT> and C<O_ACCMODE>.
		411
		412
137	=item aio_close $fh, $callback	413	=item aio_close $fh, $callback->($status)
138		414
139	Asynchronously close a file and call the callback with the result	415	Asynchronously close a file and call the callback with the result
140	code. I<WARNING:> although accepted, you should not pass in a perl	416	code.
141	filehandle here, as perl will likely close the file descriptor another
142	time when the filehandle is destroyed. Normally, you can safely call perls
143	C<close> or just let filehandles go out of scope.
144		417
145	This is supposed to be a bug in the API, so that might change. It's	418	Unfortunately, you can't do this to perl. Perl I<insists> very strongly on
146	therefore best to avoid this function.	419	closing the file descriptor associated with the filehandle itself.
147		420
		421	Therefore, C<aio_close> will not close the filehandle - instead it will
		422	use dup2 to overwrite the file descriptor with the write-end of a pipe
		423	(the pipe fd will be created on demand and will be cached).
		424
		425	Or in other words: the file descriptor will be closed, but it will not be
		426	free for reuse until the perl filehandle is closed.
		427
		428	=cut
		429
		430	=item aio_seek $fh, $offset, $whence, $callback->($offs)
		431
		432	Seeks the filehandle to the new C<$offset>, similarly to perl's
		433	C<sysseek>. The C<$whence> can use the traditional values (C<0> for
		434	C<IO::AIO::SEEK_SET>, C<1> for C<IO::AIO::SEEK_CUR> or C<2> for
		435	C<IO::AIO::SEEK_END>).
		436
		437	The resulting absolute offset will be passed to the callback, or C<-1> in
		438	case of an error.
		439
		440	In theory, the C<$whence> constants could be different than the
		441	corresponding values from L<Fcntl>, but perl guarantees they are the same,
		442	so don't panic.
		443
		444	As a GNU/Linux (and maybe Solaris) extension, also the constants
		445	C<IO::AIO::SEEK_DATA> and C<IO::AIO::SEEK_HOLE> are available, if they
		446	could be found. No guarantees about suitability for use in C<aio_seek> or
		447	Perl's C<sysseek> can be made though, although I would naively assume they
		448	"just work".
		449
148	=item aio_read $fh,$offset,$length, $data,$dataoffset,$callback	450	=item aio_read $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
149		451
150	=item aio_write $fh,$offset,$length, $data,$dataoffset,$callback	452	=item aio_write $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
151		453
152	Reads or writes C<length> bytes from the specified C<fh> and C<offset>	454	Reads or writes C<$length> bytes from or to the specified C<$fh> and
153	into the scalar given by C<data> and offset C<dataoffset> and calls the	455	C<$offset> into the scalar given by C<$data> and offset C<$dataoffset> and
154	callback without the actual number of bytes read (or -1 on error, just	456	calls the callback with the actual number of bytes transferred (or -1 on
155	like the syscall).	457	error, just like the syscall).
		458
		459	C<aio_read> will, like C<sysread>, shrink or grow the C<$data> scalar to
		460	offset plus the actual number of bytes read.
		461
		462	If C<$offset> is undefined, then the current file descriptor offset will
		463	be used (and updated), otherwise the file descriptor offset will not be
		464	changed by these calls.
		465
		466	If C<$length> is undefined in C<aio_write>, use the remaining length of
		467	C<$data>.
		468
		469	If C<$dataoffset> is less than zero, it will be counted from the end of
		470	C<$data>.
156		471
157	The C<$data> scalar I<MUST NOT> be modified in any way while the request	472	The C<$data> scalar I<MUST NOT> be modified in any way while the request
158	is outstanding. Modifying it can result in segfaults or WW3 (if the	473	is outstanding. Modifying it can result in segfaults or World War III (if
159	necessary/optional hardware is installed).	474	the necessary/optional hardware is installed).
160		475
161	Example: Read 15 bytes at offset 7 into scalar C<$buffer>, starting at	476	Example: Read 15 bytes at offset 7 into scalar C<$buffer>, starting at
162	offset C<0> within the scalar:	477	offset C<0> within the scalar:
163		478
164	aio_read $fh, 7, 15, $buffer, 0, sub {	479	aio_read $fh, 7, 15, $buffer, 0, sub {
165	$_[0] > 0 or die "read error: $!";	480	$_[0] > 0 or die "read error: $!";
166	print "read $_[0] bytes: <$buffer>\n";	481	print "read $_[0] bytes: <$buffer>\n";
167	};	482	};
168		483
		484
169	=item aio_sendfile $out_fh, $in_fh, $in_offset, $length, $callback	485	=item aio_sendfile $out_fh, $in_fh, $in_offset, $length, $callback->($retval)
170		486
171	Tries to copy C<$length> bytes from C<$in_fh> to C<$out_fh>. It starts	487	Tries to copy C<$length> bytes from C<$in_fh> to C<$out_fh>. It starts
172	reading at byte offset C<$in_offset>, and starts writing at the current	488	reading at byte offset C<$in_offset>, and starts writing at the current
173	file offset of C<$out_fh>. Because of that, it is not safe to issue more	489	file offset of C<$out_fh>. Because of that, it is not safe to issue more
174	than one C<aio_sendfile> per C<$out_fh>, as they will interfere with each	490	than one C<aio_sendfile> per C<$out_fh>, as they will interfere with each
175	other.	491	other. The same C<$in_fh> works fine though, as this function does not
		492	move or use the file offset of C<$in_fh>.
176		493
		494	Please note that C<aio_sendfile> can read more bytes from C<$in_fh> than
		495	are written, and there is no way to find out how many more bytes have been
		496	read from C<aio_sendfile> alone, as C<aio_sendfile> only provides the
		497	number of bytes written to C<$out_fh>. Only if the result value equals
		498	C<$length> one can assume that C<$length> bytes have been read.
		499
		500	Unlike with other C<aio_> functions, it makes a lot of sense to use
		501	C<aio_sendfile> on non-blocking sockets, as long as one end (typically
		502	the C<$in_fh>) is a file - the file I/O will then be asynchronous, while
		503	the socket I/O will be non-blocking. Note, however, that you can run
		504	into a trap where C<aio_sendfile> reads some data with readahead, then
		505	fails to write all data, and when the socket is ready the next time, the
		506	data in the cache is already lost, forcing C<aio_sendfile> to again hit
		507	the disk. Explicit C<aio_read> + C<aio_write> let's you better control
		508	resource usage.
		509
177	This call tries to make use of a native C<sendfile> syscall to provide	510	This call tries to make use of a native C<sendfile>-like syscall to
178	zero-copy operation. For this to work, C<$out_fh> should refer to a	511	provide zero-copy operation. For this to work, C<$out_fh> should refer to
179	socket, and C<$in_fh> should refer to mmap'able file.	512	a socket, and C<$in_fh> should refer to an mmap'able file.
180		513
181	If the native sendfile call fails or is not implemented, it will be	514	If a native sendfile cannot be found or it fails with C<ENOSYS>,
182	emulated, so you can call C<aio_sendfile> on any type of filehandle	515	C<EINVAL>, C<ENOTSUP>, C<EOPNOTSUPP>, C<EAFNOSUPPORT>, C<EPROTOTYPE> or
		516	C<ENOTSOCK>, it will be emulated, so you can call C<aio_sendfile> on any
183	regardless of the limitations of the operating system.	517	type of filehandle regardless of the limitations of the operating system.
184		518
185	Please note, however, that C<aio_sendfile> can read more bytes from	519	As native sendfile syscalls (as practically any non-POSIX interface hacked
186	C<$in_fh> than are written, and there is no way to find out how many	520	together in a hurry to improve benchmark numbers) tend to be rather buggy
187	bytes have been read from C<aio_sendfile> alone, as C<aio_sendfile> only	521	on many systems, this implementation tries to work around some known bugs
188	provides the number of bytes written to C<$out_fh>. Only if the result	522	in Linux and FreeBSD kernels (probably others, too), but that might fail,
189	value equals C<$length> one can assume that C<$length> bytes have been	523	so you really really should check the return value of C<aio_sendfile> -
190	read.	524	fewer bytes than expected might have been transferred.
191		525
		526
192	=item aio_readahead $fh,$offset,$length, $callback	527	=item aio_readahead $fh,$offset,$length, $callback->($retval)
193		528
194	C<aio_readahead> populates the page cache with data from a file so that	529	C<aio_readahead> populates the page cache with data from a file so that
195	subsequent reads from that file will not block on disk I/O. The C<$offset>	530	subsequent reads from that file will not block on disk I/O. The C<$offset>
196	argument specifies the starting point from which data is to be read and	531	argument specifies the starting point from which data is to be read and
197	C<$length> specifies the number of bytes to be read. I/O is performed in	532	C<$length> specifies the number of bytes to be read. I/O is performed in
198	whole pages, so that offset is effectively rounded down to a page boundary	533	whole pages, so that offset is effectively rounded down to a page boundary
199	and bytes are read up to the next page boundary greater than or equal to	534	and bytes are read up to the next page boundary greater than or equal to
200	(off-set+length). C<aio_readahead> does not read beyond the end of the	535	(off-set+length). C<aio_readahead> does not read beyond the end of the
201	file. The current file offset of the file is left unchanged.	536	file. The current file offset of the file is left unchanged.
202		537
203	If that syscall doesn't exist (likely if your OS isn't Linux) it will be	538	If that syscall doesn't exist (likely if your kernel isn't Linux) it will
204	emulated by simply reading the data, which would have a similar effect.	539	be emulated by simply reading the data, which would have a similar effect.
205		540
		541
206	=item aio_stat $fh_or_path, $callback	542	=item aio_stat $fh_or_path, $callback->($status)
207		543
208	=item aio_lstat $fh, $callback	544	=item aio_lstat $fh, $callback->($status)
209		545
210	Works like perl's C<stat> or C<lstat> in void context. The callback will	546	Works almost exactly like perl's C<stat> or C<lstat> in void context. The
211	be called after the stat and the results will be available using C<stat _>	547	callback will be called after the stat and the results will be available
212	or C<-s _> etc...	548	using C<stat _> or C<-s _> and other tests (with the exception of C<-B>
		549	and C<-T>).
213		550
214	The pathname passed to C<aio_stat> must be absolute. See API NOTES, above,	551	The pathname passed to C<aio_stat> must be absolute. See API NOTES, above,
215	for an explanation.	552	for an explanation.
216		553
217	Currently, the stats are always 64-bit-stats, i.e. instead of returning an	554	Currently, the stats are always 64-bit-stats, i.e. instead of returning an
218	error when stat'ing a large file, the results will be silently truncated	555	error when stat'ing a large file, the results will be silently truncated
219	unless perl itself is compiled with large file support.	556	unless perl itself is compiled with large file support.
		557
		558	To help interpret the mode and dev/rdev stat values, IO::AIO offers the
		559	following constants and functions (if not implemented, the constants will
		560	be C<0> and the functions will either C<croak> or fall back on traditional
		561	behaviour).
		562
		563	C<S_IFMT>, C<S_IFIFO>, C<S_IFCHR>, C<S_IFBLK>, C<S_IFLNK>, C<S_IFREG>,
		564	C<S_IFDIR>, C<S_IFWHT>, C<S_IFSOCK>, C<IO::AIO::major $dev_t>,
		565	C<IO::AIO::minor $dev_t>, C<IO::AIO::makedev $major, $minor>.
		566
		567	To access higher resolution stat timestamps, see L<SUBSECOND STAT TIME
		568	ACCESS>.
220		569
221	Example: Print the length of F</etc/passwd>:	570	Example: Print the length of F</etc/passwd>:
222		571
223	aio_stat "/etc/passwd", sub {	572	aio_stat "/etc/passwd", sub {
224	$_[0] and die "stat failed: $!";	573	$_[0] and die "stat failed: $!";
225	print "size is ", -s _, "\n";	574	print "size is ", -s _, "\n";
226	};	575	};
227		576
		577
		578	=item aio_statvfs $fh_or_path, $callback->($statvfs)
		579
		580	Works like the POSIX C<statvfs> or C<fstatvfs> syscalls, depending on
		581	whether a file handle or path was passed.
		582
		583	On success, the callback is passed a hash reference with the following
		584	members: C<bsize>, C<frsize>, C<blocks>, C<bfree>, C<bavail>, C<files>,
		585	C<ffree>, C<favail>, C<fsid>, C<flag> and C<namemax>. On failure, C<undef>
		586	is passed.
		587
		588	The following POSIX IO::AIO::ST_* constants are defined: C<ST_RDONLY> and
		589	C<ST_NOSUID>.
		590
		591	The following non-POSIX IO::AIO::ST_* flag masks are defined to
		592	their correct value when available, or to C<0> on systems that do
		593	not support them: C<ST_NODEV>, C<ST_NOEXEC>, C<ST_SYNCHRONOUS>,
		594	C<ST_MANDLOCK>, C<ST_WRITE>, C<ST_APPEND>, C<ST_IMMUTABLE>, C<ST_NOATIME>,
		595	C<ST_NODIRATIME> and C<ST_RELATIME>.
		596
		597	Example: stat C</wd> and dump out the data if successful.
		598
		599	aio_statvfs "/wd", sub {
		600	my $f = $_[0]
		601	or die "statvfs: $!";
		602
		603	use Data::Dumper;
		604	say Dumper $f;
		605	};
		606
		607	# result:
		608	{
		609	bsize => 1024,
		610	bfree => 4333064312,
		611	blocks => 10253828096,
		612	files => 2050765568,
		613	flag => 4096,
		614	favail => 2042092649,
		615	bavail => 4333064312,
		616	ffree => 2042092649,
		617	namemax => 255,
		618	frsize => 1024,
		619	fsid => 1810
		620	}
		621
		622	=item aio_utime $fh_or_path, $atime, $mtime, $callback->($status)
		623
		624	Works like perl's C<utime> function (including the special case of $atime
		625	and $mtime being undef). Fractional times are supported if the underlying
		626	syscalls support them.
		627
		628	When called with a pathname, uses utimensat(2) or utimes(2) if available,
		629	otherwise utime(2). If called on a file descriptor, uses futimens(2)
		630	or futimes(2) if available, otherwise returns ENOSYS, so this is not
		631	portable.
		632
		633	Examples:
		634
		635	# set atime and mtime to current time (basically touch(1)):
		636	aio_utime "path", undef, undef;
		637	# set atime to current time and mtime to beginning of the epoch:
		638	aio_utime "path", time, undef; # undef==0
		639
		640
		641	=item aio_chown $fh_or_path, $uid, $gid, $callback->($status)
		642
		643	Works like perl's C<chown> function, except that C<undef> for either $uid
		644	or $gid is being interpreted as "do not change" (but -1 can also be used).
		645
		646	Examples:
		647
		648	# same as "chown root path" in the shell:
		649	aio_chown "path", 0, -1;
		650	# same as above:
		651	aio_chown "path", 0, undef;
		652
		653
		654	=item aio_truncate $fh_or_path, $offset, $callback->($status)
		655
		656	Works like truncate(2) or ftruncate(2).
		657
		658
		659	=item aio_allocate $fh, $mode, $offset, $len, $callback->($status)
		660
		661	Allocates or frees disk space according to the C<$mode> argument. See the
		662	linux C<fallocate> documentation for details.
		663
		664	C<$mode> is usually C<0> or C<IO::AIO::FALLOC_FL_KEEP_SIZE> to allocate
		665	space, or C<IO::AIO::FALLOC_FL_PUNCH_HOLE | IO::AIO::FALLOC_FL_KEEP_SIZE>,
		666	to deallocate a file range.
		667
		668	IO::AIO also supports C<FALLOC_FL_COLLAPSE_RANGE>, to remove a range
		669	(without leaving a hole), C<FALLOC_FL_ZERO_RANGE>, to zero a range,
		670	C<FALLOC_FL_INSERT_RANGE> to insert a range and C<FALLOC_FL_UNSHARE_RANGE>
		671	to unshare shared blocks (see your L<fallocate(2)> manpage).
		672
		673	The file system block size used by C<fallocate> is presumably the
		674	C<f_bsize> returned by C<statvfs>, but different filesystems and filetypes
		675	can dictate other limitations.
		676
		677	If C<fallocate> isn't available or cannot be emulated (currently no
		678	emulation will be attempted), passes C<-1> and sets C<$!> to C<ENOSYS>.
		679
		680
		681	=item aio_chmod $fh_or_path, $mode, $callback->($status)
		682
		683	Works like perl's C<chmod> function.
		684
		685
228	=item aio_unlink $pathname, $callback	686	=item aio_unlink $pathname, $callback->($status)
229		687
230	Asynchronously unlink (delete) a file and call the callback with the	688	Asynchronously unlink (delete) a file and call the callback with the
231	result code.	689	result code.
232		690
		691
		692	=item aio_mknod $pathname, $mode, $dev, $callback->($status)
		693
		694	[EXPERIMENTAL]
		695
		696	Asynchronously create a device node (or fifo). See mknod(2).
		697
		698	The only (POSIX-) portable way of calling this function is:
		699
		700	aio_mknod $pathname, IO::AIO::S_IFIFO | $mode, 0, sub { ...
		701
		702	See C<aio_stat> for info about some potentially helpful extra constants
		703	and functions.
		704
		705	=item aio_link $srcpath, $dstpath, $callback->($status)
		706
		707	Asynchronously create a new link to the existing object at C<$srcpath> at
		708	the path C<$dstpath> and call the callback with the result code.
		709
		710
		711	=item aio_symlink $srcpath, $dstpath, $callback->($status)
		712
		713	Asynchronously create a new symbolic link to the existing object at C<$srcpath> at
		714	the path C<$dstpath> and call the callback with the result code.
		715
		716
		717	=item aio_readlink $pathname, $callback->($link)
		718
		719	Asynchronously read the symlink specified by C<$path> and pass it to
		720	the callback. If an error occurs, nothing or undef gets passed to the
		721	callback.
		722
		723
		724	=item aio_realpath $pathname, $callback->($path)
		725
		726	Asynchronously make the path absolute and resolve any symlinks in
		727	C<$path>. The resulting path only consists of directories (same as
		728	L<Cwd::realpath>).
		729
		730	This request can be used to get the absolute path of the current working
		731	directory by passing it a path of F<.> (a single dot).
		732
		733
		734	=item aio_rename $srcpath, $dstpath, $callback->($status)
		735
		736	Asynchronously rename the object at C<$srcpath> to C<$dstpath>, just as
		737	rename(2) and call the callback with the result code.
		738
		739	On systems that support the AIO::WD working directory abstraction
		740	natively, the case C<[$wd, "."]> as C<$srcpath> is specialcased - instead
		741	of failing, C<rename> is called on the absolute path of C<$wd>.
		742
		743
		744	=item aio_rename2 $srcpath, $dstpath, $flags, $callback->($status)
		745
		746	Basically a version of C<aio_rename> with an additional C<$flags>
		747	argument. Calling this with C<$flags=0> is the same as calling
		748	C<aio_rename>.
		749
		750	Non-zero flags are currently only supported on GNU/Linux systems that
		751	support renameat2. Other systems fail with C<ENOSYS> in this case.
		752
		753	The following constants are available (missing ones are, as usual C<0>),
		754	see renameat2(2) for details:
		755
		756	C<IO::AIO::RENAME_NOREPLACE>, C<IO::AIO::RENAME_EXCHANGE>
		757	and C<IO::AIO::RENAME_WHITEOUT>.
		758
		759
		760	=item aio_mkdir $pathname, $mode, $callback->($status)
		761
		762	Asynchronously mkdir (create) a directory and call the callback with
		763	the result code. C<$mode> will be modified by the umask at the time the
		764	request is executed, so do not change your umask.
		765
		766
233	=item aio_rmdir $pathname, $callback	767	=item aio_rmdir $pathname, $callback->($status)
234		768
235	Asynchronously rmdir (delete) a directory and call the callback with the	769	Asynchronously rmdir (delete) a directory and call the callback with the
236	result code.	770	result code.
237		771
		772	On systems that support the AIO::WD working directory abstraction
		773	natively, the case C<[$wd, "."]> is specialcased - instead of failing,
		774	C<rmdir> is called on the absolute path of C<$wd>.
		775
		776
238	=item aio_readdir $pathname $callback	777	=item aio_readdir $pathname, $callback->($entries)
239		778
240	Unlike the POSIX call of the same name, C<aio_readdir> reads an entire	779	Unlike the POSIX call of the same name, C<aio_readdir> reads an entire
241	directory (i.e. opendir + readdir + closedir). The entries will not be	780	directory (i.e. opendir + readdir + closedir). The entries will not be
242	sorted, and will B<NOT> include the C<.> and C<..> entries.	781	sorted, and will B<NOT> include the C<.> and C<..> entries.
243		782
244	The callback a single argument which is either C<undef> or an array-ref	783	The callback is passed a single argument which is either C<undef> or an
245	with the filenames.	784	array-ref with the filenames.
246		785
		786
		787	=item aio_readdirx $pathname, $flags, $callback->($entries, $flags)
		788
		789	Quite similar to C<aio_readdir>, but the C<$flags> argument allows one to
		790	tune behaviour and output format. In case of an error, C<$entries> will be
		791	C<undef>.
		792
		793	The flags are a combination of the following constants, ORed together (the
		794	flags will also be passed to the callback, possibly modified):
		795
		796	=over 4
		797
		798	=item IO::AIO::READDIR_DENTS
		799
		800	Normally the callback gets an arrayref consisting of names only (as
		801	with C<aio_readdir>). If this flag is set, then the callback gets an
		802	arrayref with C<[$name, $type, $inode]> arrayrefs, each describing a
		803	single directory entry in more detail:
		804
		805	C<$name> is the name of the entry.
		806
		807	C<$type> is one of the C<IO::AIO::DT_xxx> constants:
		808
		809	C<IO::AIO::DT_UNKNOWN>, C<IO::AIO::DT_FIFO>, C<IO::AIO::DT_CHR>, C<IO::AIO::DT_DIR>,
		810	C<IO::AIO::DT_BLK>, C<IO::AIO::DT_REG>, C<IO::AIO::DT_LNK>, C<IO::AIO::DT_SOCK>,
		811	C<IO::AIO::DT_WHT>.
		812
		813	C<IO::AIO::DT_UNKNOWN> means just that: readdir does not know. If you need
		814	to know, you have to run stat yourself. Also, for speed/memory reasons,
		815	the C<$type> scalars are read-only: you must not modify them.
		816
		817	C<$inode> is the inode number (which might not be exact on systems with 64
		818	bit inode numbers and 32 bit perls). This field has unspecified content on
		819	systems that do not deliver the inode information.
		820
		821	=item IO::AIO::READDIR_DIRS_FIRST
		822
		823	When this flag is set, then the names will be returned in an order where
		824	likely directories come first, in optimal stat order. This is useful when
		825	you need to quickly find directories, or you want to find all directories
		826	while avoiding to stat() each entry.
		827
		828	If the system returns type information in readdir, then this is used
		829	to find directories directly. Otherwise, likely directories are names
		830	beginning with ".", or otherwise names with no dots, of which names with
		831	short names are tried first.
		832
		833	=item IO::AIO::READDIR_STAT_ORDER
		834
		835	When this flag is set, then the names will be returned in an order
		836	suitable for stat()'ing each one. That is, when you plan to stat() most or
		837	all files in the given directory, then the returned order will likely be
		838	faster.
		839
		840	If both this flag and C<IO::AIO::READDIR_DIRS_FIRST> are specified,
		841	then the likely dirs come first, resulting in a less optimal stat order
		842	for stat'ing all entries, but likely a more optimal order for finding
		843	subdirectories.
		844
		845	=item IO::AIO::READDIR_FOUND_UNKNOWN
		846
		847	This flag should not be set when calling C<aio_readdirx>. Instead, it
		848	is being set by C<aio_readdirx>, when any of the C<$type>'s found were
		849	C<IO::AIO::DT_UNKNOWN>. The absence of this flag therefore indicates that all
		850	C<$type>'s are known, which can be used to speed up some algorithms.
		851
		852	=back
		853
		854
		855	=item aio_slurp $pathname, $offset, $length, $data, $callback->($status)
		856
		857	Opens, reads and closes the given file. The data is put into C<$data>,
		858	which is resized as required.
		859
		860	If C<$offset> is negative, then it is counted from the end of the file.
		861
		862	If C<$length> is zero, then the remaining length of the file is
		863	used. Also, in this case, the same limitations to modifying C<$data> apply
		864	as when IO::AIO::mmap is used, i.e. it must only be modified in-place
		865	with C<substr>. If the size of the file is known, specifying a non-zero
		866	C<$length> results in a performance advantage.
		867
		868	This request is similar to the older C<aio_load> request, but since it is
		869	a single request, it might be more efficient to use.
		870
		871	Example: load F</etc/passwd> into C<$passwd>.
		872
		873	my $passwd;
		874	aio_slurp "/etc/passwd", 0, 0, $passwd, sub {
		875	$_[0] >= 0
		876	or die "/etc/passwd: $!\n";
		877
		878	printf "/etc/passwd is %d bytes long, and contains:\n", length $passwd;
		879	print $passwd;
		880	};
		881	IO::AIO::flush;
		882
		883
		884	=item aio_load $pathname, $data, $callback->($status)
		885
		886	This is a composite request that tries to fully load the given file into
		887	memory. Status is the same as with aio_read.
		888
		889	Using C<aio_slurp> might be more efficient, as it is a single request.
		890
		891	=cut
		892
		893	sub aio_load($$;$) {
		894	my ($path, undef, $cb) = @_;
		895	my $data = \$_[1];
		896
		897	my $pri = aioreq_pri;
		898	my $grp = aio_group $cb;
		899
		900	aioreq_pri $pri;
		901	add $grp aio_open $path, O_RDONLY, 0, sub {
		902	my $fh = shift
		903	or return $grp->result (-1);
		904
		905	aioreq_pri $pri;
		906	add $grp aio_read $fh, 0, (-s $fh), $$data, 0, sub {
		907	$grp->result ($_[0]);
		908	};
		909	};
		910
		911	$grp
		912	}
		913
		914	=item aio_copy $srcpath, $dstpath, $callback->($status)
		915
		916	Try to copy the I<file> (directories not supported as either source or
		917	destination) from C<$srcpath> to C<$dstpath> and call the callback with
		918	a status of C<0> (ok) or C<-1> (error, see C<$!>).
		919
		920	Existing destination files will be truncated.
		921
		922	This is a composite request that creates the destination file with
		923	mode 0200 and copies the contents of the source file into it using
		924	C<aio_sendfile>, followed by restoring atime, mtime, access mode and
		925	uid/gid, in that order.
		926
		927	If an error occurs, the partial destination file will be unlinked, if
		928	possible, except when setting atime, mtime, access mode and uid/gid, where
		929	errors are being ignored.
		930
		931	=cut
		932
		933	sub aio_copy($$;$) {
		934	my ($src, $dst, $cb) = @_;
		935
		936	my $pri = aioreq_pri;
		937	my $grp = aio_group $cb;
		938
		939	aioreq_pri $pri;
		940	add $grp aio_open $src, O_RDONLY, 0, sub {
		941	if (my $src_fh = $_[0]) {
		942	my @stat = stat $src_fh; # hmm, might block over nfs?
		943
		944	aioreq_pri $pri;
		945	add $grp aio_open $dst, O_CREAT | O_WRONLY | O_TRUNC, 0200, sub {
		946	if (my $dst_fh = $_[0]) {
		947	aioreq_pri $pri;
		948	add $grp aio_sendfile $dst_fh, $src_fh, 0, $stat[7], sub {
		949	if ($_[0] == $stat[7]) {
		950	$grp->result (0);
		951	close $src_fh;
		952
		953	my $ch = sub {
		954	aioreq_pri $pri;
		955	add $grp aio_chmod $dst_fh, $stat[2] & 07777, sub {
		956	aioreq_pri $pri;
		957	add $grp aio_chown $dst_fh, $stat[4], $stat[5], sub {
		958	aioreq_pri $pri;
		959	add $grp aio_close $dst_fh;
		960	}
		961	};
		962	};
		963
		964	aioreq_pri $pri;
		965	add $grp aio_utime $dst_fh, $stat[8], $stat[9], sub {
		966	if ($_[0] < 0 && $! == ENOSYS) {
		967	aioreq_pri $pri;
		968	add $grp aio_utime $dst, $stat[8], $stat[9], $ch;
		969	} else {
		970	$ch->();
		971	}
		972	};
		973	} else {
		974	$grp->result (-1);
		975	close $src_fh;
		976	close $dst_fh;
		977
		978	aioreq $pri;
		979	add $grp aio_unlink $dst;
		980	}
		981	};
		982	} else {
		983	$grp->result (-1);
		984	}
		985	},
		986
		987	} else {
		988	$grp->result (-1);
		989	}
		990	};
		991
		992	$grp
		993	}
		994
		995	=item aio_move $srcpath, $dstpath, $callback->($status)
		996
		997	Try to move the I<file> (directories not supported as either source or
		998	destination) from C<$srcpath> to C<$dstpath> and call the callback with
		999	a status of C<0> (ok) or C<-1> (error, see C<$!>).
		1000
		1001	This is a composite request that tries to rename(2) the file first; if
		1002	rename fails with C<EXDEV>, it copies the file with C<aio_copy> and, if
		1003	that is successful, unlinks the C<$srcpath>.
		1004
		1005	=cut
		1006
		1007	sub aio_move($$;$) {
		1008	my ($src, $dst, $cb) = @_;
		1009
		1010	my $pri = aioreq_pri;
		1011	my $grp = aio_group $cb;
		1012
		1013	aioreq_pri $pri;
		1014	add $grp aio_rename $src, $dst, sub {
		1015	if ($_[0] && $! == EXDEV) {
		1016	aioreq_pri $pri;
		1017	add $grp aio_copy $src, $dst, sub {
		1018	$grp->result ($_[0]);
		1019
		1020	unless ($_[0]) {
		1021	aioreq_pri $pri;
		1022	add $grp aio_unlink $src;
		1023	}
		1024	};
		1025	} else {
		1026	$grp->result ($_[0]);
		1027	}
		1028	};
		1029
		1030	$grp
		1031	}
		1032
		1033	=item aio_scandir $pathname, $maxreq, $callback->($dirs, $nondirs)
		1034
		1035	Scans a directory (similar to C<aio_readdir>) but additionally tries to
		1036	efficiently separate the entries of directory C<$path> into two sets of
		1037	names, directories you can recurse into (directories), and ones you cannot
		1038	recurse into (everything else, including symlinks to directories).
		1039
		1040	C<aio_scandir> is a composite request that generates many sub requests.
		1041	C<$maxreq> specifies the maximum number of outstanding aio requests that
		1042	this function generates. If it is C<< <= 0 >>, then a suitable default
		1043	will be chosen (currently 4).
		1044
		1045	On error, the callback is called without arguments, otherwise it receives
		1046	two array-refs with path-relative entry names.
		1047
		1048	Example:
		1049
		1050	aio_scandir $dir, 0, sub {
		1051	my ($dirs, $nondirs) = @_;
		1052	print "real directories: @$dirs\n";
		1053	print "everything else: @$nondirs\n";
		1054	};
		1055
		1056	Implementation notes.
		1057
		1058	The C<aio_readdir> cannot be avoided, but C<stat()>'ing every entry can.
		1059
		1060	If readdir returns file type information, then this is used directly to
		1061	find directories.
		1062
		1063	Otherwise, after reading the directory, the modification time, size etc.
		1064	of the directory before and after the readdir is checked, and if they
		1065	match (and isn't the current time), the link count will be used to decide
		1066	how many entries are directories (if >= 2). Otherwise, no knowledge of the
		1067	number of subdirectories will be assumed.
		1068
		1069	Then entries will be sorted into likely directories a non-initial dot
		1070	currently) and likely non-directories (see C<aio_readdirx>). Then every
		1071	entry plus an appended C</.> will be C<stat>'ed, likely directories first,
		1072	in order of their inode numbers. If that succeeds, it assumes that the
		1073	entry is a directory or a symlink to directory (which will be checked
		1074	separately). This is often faster than stat'ing the entry itself because
		1075	filesystems might detect the type of the entry without reading the inode
		1076	data (e.g. ext2fs filetype feature), even on systems that cannot return
		1077	the filetype information on readdir.
		1078
		1079	If the known number of directories (link count - 2) has been reached, the
		1080	rest of the entries is assumed to be non-directories.
		1081
		1082	This only works with certainty on POSIX (= UNIX) filesystems, which
		1083	fortunately are the vast majority of filesystems around.
		1084
		1085	It will also likely work on non-POSIX filesystems with reduced efficiency
		1086	as those tend to return 0 or 1 as link counts, which disables the
		1087	directory counting heuristic.
		1088
		1089	=cut
		1090
		1091	sub aio_scandir($$;$) {
		1092	my ($path, $maxreq, $cb) = @_;
		1093
		1094	my $pri = aioreq_pri;
		1095
		1096	my $grp = aio_group $cb;
		1097
		1098	$maxreq = 4 if $maxreq <= 0;
		1099
		1100	# get a wd object
		1101	aioreq_pri $pri;
		1102	add $grp aio_wd $path, sub {
		1103	$_[0]
		1104	or return $grp->result ();
		1105
		1106	my $wd = [shift, "."];
		1107
		1108	# stat once
		1109	aioreq_pri $pri;
		1110	add $grp aio_stat $wd, sub {
		1111	return $grp->result () if $_[0];
		1112	my $now = time;
		1113	my $hash1 = join ":", (stat _)[0,1,3,7,9];
		1114	my $rdxflags = READDIR_DIRS_FIRST;
		1115
		1116	if ((stat _)[3] < 2) {
		1117	# at least one non-POSIX filesystem exists
		1118	# that returns useful DT_type values: btrfs,
		1119	# so optimise for this here by requesting dents
		1120	$rdxflags |= READDIR_DENTS;
		1121	}
		1122
		1123	# read the directory entries
		1124	aioreq_pri $pri;
		1125	add $grp aio_readdirx $wd, $rdxflags, sub {
		1126	my ($entries, $flags) = @_
		1127	or return $grp->result ();
		1128
		1129	if ($rdxflags & READDIR_DENTS) {
		1130	# if we requested type values, see if we can use them directly.
		1131
		1132	# if there were any DT_UNKNOWN entries then we assume we
		1133	# don't know. alternatively, we could assume that if we get
		1134	# one DT_DIR, then all directories are indeed marked with
		1135	# DT_DIR, but this seems not required for btrfs, and this
		1136	# is basically the "btrfs can't get it's act together" code
		1137	# branch.
		1138	unless ($flags & READDIR_FOUND_UNKNOWN) {
		1139	# now we have valid DT_ information for all entries,
		1140	# so use it as an optimisation without further stat's.
		1141	# they must also all be at the beginning of @$entries
		1142	# by now.
		1143
		1144	my $dirs;
		1145
		1146	if (@$entries) {
		1147	for (0 .. $#$entries) {
		1148	if ($entries->[$_][1] != DT_DIR) {
		1149	# splice out directories
		1150	$dirs = [splice @$entries, 0, $_];
		1151	last;
		1152	}
		1153	}
		1154
		1155	# if we didn't find any non-dir, then all entries are dirs
		1156	unless ($dirs) {
		1157	($dirs, $entries) = ($entries, []);
		1158	}
		1159	} else {
		1160	# directory is empty, so there are no sbdirs
		1161	$dirs = [];
		1162	}
		1163
		1164	# either splice'd the directories out or the dir was empty.
		1165	# convert dents to filenames
		1166	$_ = $_->[0] for @$dirs;
		1167	$_ = $_->[0] for @$entries;
		1168
		1169	return $grp->result ($dirs, $entries);
		1170	}
		1171
		1172	# cannot use, so return to our old ways
		1173	# by pretending we only scanned for names.
		1174	$_ = $_->[0] for @$entries;
		1175	}
		1176
		1177	# stat the dir another time
		1178	aioreq_pri $pri;
		1179	add $grp aio_stat $wd, sub {
		1180	my $hash2 = join ":", (stat _)[0,1,3,7,9];
		1181
		1182	my $ndirs;
		1183
		1184	# take the slow route if anything looks fishy
		1185	if ($hash1 ne $hash2 or (stat _)[9] == $now) {
		1186	$ndirs = -1;
		1187	} else {
		1188	# if nlink == 2, we are finished
		1189	# for non-posix-fs's, we rely on nlink < 2
		1190	$ndirs = (stat _)[3] - 2
		1191	or return $grp->result ([], $entries);
		1192	}
		1193
		1194	my (@dirs, @nondirs);
		1195
		1196	my $statgrp = add $grp aio_group sub {
		1197	$grp->result (\@dirs, \@nondirs);
		1198	};
		1199
		1200	limit $statgrp $maxreq;
		1201	feed $statgrp sub {
		1202	return unless @$entries;
		1203	my $entry = shift @$entries;
		1204
		1205	aioreq_pri $pri;
		1206	$wd->[1] = "$entry/.";
		1207	add $statgrp aio_stat $wd, sub {
		1208	if ($_[0] < 0) {
		1209	push @nondirs, $entry;
		1210	} else {
		1211	# need to check for real directory
		1212	aioreq_pri $pri;
		1213	$wd->[1] = $entry;
		1214	add $statgrp aio_lstat $wd, sub {
		1215	if (-d _) {
		1216	push @dirs, $entry;
		1217
		1218	unless (--$ndirs) {
		1219	push @nondirs, @$entries;
		1220	feed $statgrp;
		1221	}
		1222	} else {
		1223	push @nondirs, $entry;
		1224	}
		1225	}
		1226	}
		1227	};
		1228	};
		1229	};
		1230	};
		1231	};
		1232	};
		1233
		1234	$grp
		1235	}
		1236
		1237	=item aio_rmtree $pathname, $callback->($status)
		1238
		1239	Delete a directory tree starting (and including) C<$path>, return the
		1240	status of the final C<rmdir> only. This is a composite request that
		1241	uses C<aio_scandir> to recurse into and rmdir directories, and unlink
		1242	everything else.
		1243
		1244	=cut
		1245
		1246	sub aio_rmtree;
		1247	sub aio_rmtree($;$) {
		1248	my ($path, $cb) = @_;
		1249
		1250	my $pri = aioreq_pri;
		1251	my $grp = aio_group $cb;
		1252
		1253	aioreq_pri $pri;
		1254	add $grp aio_scandir $path, 0, sub {
		1255	my ($dirs, $nondirs) = @_;
		1256
		1257	my $dirgrp = aio_group sub {
		1258	add $grp aio_rmdir $path, sub {
		1259	$grp->result ($_[0]);
		1260	};
		1261	};
		1262
		1263	(aioreq_pri $pri), add $dirgrp aio_rmtree "$path/$_" for @$dirs;
		1264	(aioreq_pri $pri), add $dirgrp aio_unlink "$path/$_" for @$nondirs;
		1265
		1266	add $grp $dirgrp;
		1267	};
		1268
		1269	$grp
		1270	}
		1271
		1272	=item aio_fcntl $fh, $cmd, $arg, $callback->($status)
		1273
		1274	=item aio_ioctl $fh, $request, $buf, $callback->($status)
		1275
		1276	These work just like the C<fcntl> and C<ioctl> built-in functions, except
		1277	they execute asynchronously and pass the return value to the callback.
		1278
		1279	Both calls can be used for a lot of things, some of which make more sense
		1280	to run asynchronously in their own thread, while some others make less
		1281	sense. For example, calls that block waiting for external events, such
		1282	as locking, will also lock down an I/O thread while it is waiting, which
		1283	can deadlock the whole I/O system. At the same time, there might be no
		1284	alternative to using a thread to wait.
		1285
		1286	So in general, you should only use these calls for things that do
		1287	(filesystem) I/O, not for things that wait for other events (network,
		1288	other processes), although if you are careful and know what you are doing,
		1289	you still can.
		1290
		1291	The following constants are available (missing ones are, as usual C<0>):
		1292
		1293	C<F_DUPFD_CLOEXEC>,
		1294
		1295	C<F_OFD_GETLK>, C<F_OFD_SETLK>, C<F_OFD_GETLKW>,
		1296
		1297	C<FIFREEZE>, C<FITHAW>, C<FITRIM>, C<FICLONE>, C<FICLONERANGE>, C<FIDEDUPERANGE>.
		1298
		1299	C<FS_IOC_GETFLAGS>, C<FS_IOC_SETFLAGS>, C<FS_IOC_GETVERSION>, C<FS_IOC_SETVERSION>,
		1300	C<FS_IOC_FIEMAP>.
		1301
		1302	C<FS_IOC_FSGETXATTR>, C<FS_IOC_FSSETXATTR>, C<FS_IOC_SET_ENCRYPTION_POLICY>,
		1303	C<FS_IOC_GET_ENCRYPTION_PWSALT>, C<FS_IOC_GET_ENCRYPTION_POLICY>, C<FS_KEY_DESCRIPTOR_SIZE>.
		1304
		1305	C<FS_SECRM_FL>, C<FS_UNRM_FL>, C<FS_COMPR_FL>, C<FS_SYNC_FL>, C<FS_IMMUTABLE_FL>,
		1306	C<FS_APPEND_FL>, C<FS_NODUMP_FL>, C<FS_NOATIME_FL>, C<FS_DIRTY_FL>,
		1307	C<FS_COMPRBLK_FL>, C<FS_NOCOMP_FL>, C<FS_ENCRYPT_FL>, C<FS_BTREE_FL>,
		1308	C<FS_INDEX_FL>, C<FS_JOURNAL_DATA_FL>, C<FS_NOTAIL_FL>, C<FS_DIRSYNC_FL>, C<FS_TOPDIR_FL>,
		1309	C<FS_FL_USER_MODIFIABLE>.
		1310
		1311	C<FS_XFLAG_REALTIME>, C<FS_XFLAG_PREALLOC>, C<FS_XFLAG_IMMUTABLE>, C<FS_XFLAG_APPEND>,
		1312	C<FS_XFLAG_SYNC>, C<FS_XFLAG_NOATIME>, C<FS_XFLAG_NODUMP>, C<FS_XFLAG_RTINHERIT>,
		1313	C<FS_XFLAG_PROJINHERIT>, C<FS_XFLAG_NOSYMLINKS>, C<FS_XFLAG_EXTSIZE>, C<FS_XFLAG_EXTSZINHERIT>,
		1314	C<FS_XFLAG_NODEFRAG>, C<FS_XFLAG_FILESTREAM>, C<FS_XFLAG_DAX>, C<FS_XFLAG_HASATTR>,
		1315
		1316	=item aio_sync $callback->($status)
		1317
		1318	Asynchronously call sync and call the callback when finished.
		1319
247	=item aio_fsync $fh, $callback	1320	=item aio_fsync $fh, $callback->($status)
248		1321
249	Asynchronously call fsync on the given filehandle and call the callback	1322	Asynchronously call fsync on the given filehandle and call the callback
250	with the fsync result code.	1323	with the fsync result code.
251		1324
252	=item aio_fdatasync $fh, $callback	1325	=item aio_fdatasync $fh, $callback->($status)
253		1326
254	Asynchronously call fdatasync on the given filehandle and call the	1327	Asynchronously call fdatasync on the given filehandle and call the
255	callback with the fdatasync result code.	1328	callback with the fdatasync result code.
256		1329
257	If this call isn't available because your OS lacks it or it couldn't be	1330	If this call isn't available because your OS lacks it or it couldn't be
258	detected, it will be emulated by calling C<fsync> instead.	1331	detected, it will be emulated by calling C<fsync> instead.
259		1332
		1333	=item aio_syncfs $fh, $callback->($status)
		1334
		1335	Asynchronously call the syncfs syscall to sync the filesystem associated
		1336	to the given filehandle and call the callback with the syncfs result
		1337	code. If syncfs is not available, calls sync(), but returns C<-1> and sets
		1338	errno to C<ENOSYS> nevertheless.
		1339
		1340	=item aio_sync_file_range $fh, $offset, $nbytes, $flags, $callback->($status)
		1341
		1342	Sync the data portion of the file specified by C<$offset> and C<$length>
		1343	to disk (but NOT the metadata), by calling the Linux-specific
		1344	sync_file_range call. If sync_file_range is not available or it returns
		1345	ENOSYS, then fdatasync or fsync is being substituted.
		1346
		1347	C<$flags> can be a combination of C<IO::AIO::SYNC_FILE_RANGE_WAIT_BEFORE>,
		1348	C<IO::AIO::SYNC_FILE_RANGE_WRITE> and
		1349	C<IO::AIO::SYNC_FILE_RANGE_WAIT_AFTER>: refer to the sync_file_range
		1350	manpage for details.
		1351
		1352	=item aio_pathsync $pathname, $callback->($status)
		1353
		1354	This request tries to open, fsync and close the given path. This is a
		1355	composite request intended to sync directories after directory operations
		1356	(E.g. rename). This might not work on all operating systems or have any
		1357	specific effect, but usually it makes sure that directory changes get
		1358	written to disc. It works for anything that can be opened for read-only,
		1359	not just directories.
		1360
		1361	Future versions of this function might fall back to other methods when
		1362	C<fsync> on the directory fails (such as calling C<sync>).
		1363
		1364	Passes C<0> when everything went ok, and C<-1> on error.
		1365
		1366	=cut
		1367
		1368	sub aio_pathsync($;$) {
		1369	my ($path, $cb) = @_;
		1370
		1371	my $pri = aioreq_pri;
		1372	my $grp = aio_group $cb;
		1373
		1374	aioreq_pri $pri;
		1375	add $grp aio_open $path, O_RDONLY, 0, sub {
		1376	my ($fh) = @_;
		1377	if ($fh) {
		1378	aioreq_pri $pri;
		1379	add $grp aio_fsync $fh, sub {
		1380	$grp->result ($_[0]);
		1381
		1382	aioreq_pri $pri;
		1383	add $grp aio_close $fh;
		1384	};
		1385	} else {
		1386	$grp->result (-1);
		1387	}
		1388	};
		1389
		1390	$grp
		1391	}
		1392
		1393	=item aio_msync $scalar, $offset = 0, $length = undef, flags = MS_SYNC, $callback->($status)
		1394
		1395	This is a rather advanced IO::AIO call, which only works on mmap(2)ed
		1396	scalars (see the C<IO::AIO::mmap> function, although it also works on data
		1397	scalars managed by the L<Sys::Mmap> or L<Mmap> modules, note that the
		1398	scalar must only be modified in-place while an aio operation is pending on
		1399	it).
		1400
		1401	It calls the C<msync> function of your OS, if available, with the memory
		1402	area starting at C<$offset> in the string and ending C<$length> bytes
		1403	later. If C<$length> is negative, counts from the end, and if C<$length>
		1404	is C<undef>, then it goes till the end of the string. The flags can be
		1405	either C<IO::AIO::MS_ASYNC> or C<IO::AIO::MS_SYNC>, plus an optional
		1406	C<IO::AIO::MS_INVALIDATE>.
		1407
		1408	=item aio_mtouch $scalar, $offset = 0, $length = undef, flags = 0, $callback->($status)
		1409
		1410	This is a rather advanced IO::AIO call, which works best on mmap(2)ed
		1411	scalars.
		1412
		1413	It touches (reads or writes) all memory pages in the specified
		1414	range inside the scalar. All caveats and parameters are the same
		1415	as for C<aio_msync>, above, except for flags, which must be either
		1416	C<0> (which reads all pages and ensures they are instantiated) or
		1417	C<IO::AIO::MT_MODIFY>, which modifies the memory pages (by reading and
		1418	writing an octet from it, which dirties the page).
		1419
		1420	=item aio_mlock $scalar, $offset = 0, $length = undef, $callback->($status)
		1421
		1422	This is a rather advanced IO::AIO call, which works best on mmap(2)ed
		1423	scalars.
		1424
		1425	It reads in all the pages of the underlying storage into memory (if any)
		1426	and locks them, so they are not getting swapped/paged out or removed.
		1427
		1428	If C<$length> is undefined, then the scalar will be locked till the end.
		1429
		1430	On systems that do not implement C<mlock>, this function returns C<-1>
		1431	and sets errno to C<ENOSYS>.
		1432
		1433	Note that the corresponding C<munlock> is synchronous and is
		1434	documented under L<MISCELLANEOUS FUNCTIONS>.
		1435
		1436	Example: open a file, mmap and mlock it - both will be undone when
		1437	C<$data> gets destroyed.
		1438
		1439	open my $fh, "<", $path or die "$path: $!";
		1440	my $data;
		1441	IO::AIO::mmap $data, -s $fh, IO::AIO::PROT_READ, IO::AIO::MAP_SHARED, $fh;
		1442	aio_mlock $data; # mlock in background
		1443
		1444	=item aio_mlockall $flags, $callback->($status)
		1445
		1446	Calls the C<mlockall> function with the given C<$flags> (a
		1447	combination of C<IO::AIO::MCL_CURRENT>, C<IO::AIO::MCL_FUTURE> and
		1448	C<IO::AIO::MCL_ONFAULT>).
		1449
		1450	On systems that do not implement C<mlockall>, this function returns C<-1>
		1451	and sets errno to C<ENOSYS>. Similarly, flag combinations not supported
		1452	by the system result in a return value of C<-1> with errno being set to
		1453	C<EINVAL>.
		1454
		1455	Note that the corresponding C<munlockall> is synchronous and is
		1456	documented under L<MISCELLANEOUS FUNCTIONS>.
		1457
		1458	Example: asynchronously lock all current and future pages into memory.
		1459
		1460	aio_mlockall IO::AIO::MCL_FUTURE;
		1461
		1462	=item aio_fiemap $fh, $start, $length, $flags, $count, $cb->(\@extents)
		1463
		1464	Queries the extents of the given file (by calling the Linux C<FIEMAP>
		1465	ioctl, see L<http://cvs.schmorp.de/IO-AIO/doc/fiemap.txt> for details). If
		1466	the ioctl is not available on your OS, then this request will fail with
		1467	C<ENOSYS>.
		1468
		1469	C<$start> is the starting offset to query extents for, C<$length> is the
		1470	size of the range to query - if it is C<undef>, then the whole file will
		1471	be queried.
		1472
		1473	C<$flags> is a combination of flags (C<IO::AIO::FIEMAP_FLAG_SYNC> or
		1474	C<IO::AIO::FIEMAP_FLAG_XATTR> - C<IO::AIO::FIEMAP_FLAGS_COMPAT> is also
		1475	exported), and is normally C<0> or C<IO::AIO::FIEMAP_FLAG_SYNC> to query
		1476	the data portion.
		1477
		1478	C<$count> is the maximum number of extent records to return. If it is
		1479	C<undef>, then IO::AIO queries all extents of the range. As a very special
		1480	case, if it is C<0>, then the callback receives the number of extents
		1481	instead of the extents themselves (which is unreliable, see below).
		1482
		1483	If an error occurs, the callback receives no arguments. The special
		1484	C<errno> value C<IO::AIO::EBADR> is available to test for flag errors.
		1485
		1486	Otherwise, the callback receives an array reference with extent
		1487	structures. Each extent structure is an array reference itself, with the
		1488	following members:
		1489
		1490	[$logical, $physical, $length, $flags]
		1491
		1492	Flags is any combination of the following flag values (typically either C<0>
		1493	or C<IO::AIO::FIEMAP_EXTENT_LAST> (1)):
		1494
		1495	C<IO::AIO::FIEMAP_EXTENT_LAST>, C<IO::AIO::FIEMAP_EXTENT_UNKNOWN>,
		1496	C<IO::AIO::FIEMAP_EXTENT_DELALLOC>, C<IO::AIO::FIEMAP_EXTENT_ENCODED>,
		1497	C<IO::AIO::FIEMAP_EXTENT_DATA_ENCRYPTED>, C<IO::AIO::FIEMAP_EXTENT_NOT_ALIGNED>,
		1498	C<IO::AIO::FIEMAP_EXTENT_DATA_INLINE>, C<IO::AIO::FIEMAP_EXTENT_DATA_TAIL>,
		1499	C<IO::AIO::FIEMAP_EXTENT_UNWRITTEN>, C<IO::AIO::FIEMAP_EXTENT_MERGED> or
		1500	C<IO::AIO::FIEMAP_EXTENT_SHARED>.
		1501
		1502	At the time of this writing (Linux 3.2), this request is unreliable unless
		1503	C<$count> is C<undef>, as the kernel has all sorts of bugs preventing
		1504	it to return all extents of a range for files with a large number of
		1505	extents. The code (only) works around all these issues if C<$count> is
		1506	C<undef>.
		1507
		1508	=item aio_group $callback->(...)
		1509
		1510	This is a very special aio request: Instead of doing something, it is a
		1511	container for other aio requests, which is useful if you want to bundle
		1512	many requests into a single, composite, request with a definite callback
		1513	and the ability to cancel the whole request with its subrequests.
		1514
		1515	Returns an object of class L<IO::AIO::GRP>. See its documentation below
		1516	for more info.
		1517
		1518	Example:
		1519
		1520	my $grp = aio_group sub {
		1521	print "all stats done\n";
		1522	};
		1523
		1524	add $grp
		1525	(aio_stat ...),
		1526	(aio_stat ...),
		1527	...;
		1528
		1529	=item aio_nop $callback->()
		1530
		1531	This is a special request - it does nothing in itself and is only used for
		1532	side effects, such as when you want to add a dummy request to a group so
		1533	that finishing the requests in the group depends on executing the given
		1534	code.
		1535
		1536	While this request does nothing, it still goes through the execution
		1537	phase and still requires a worker thread. Thus, the callback will not
		1538	be executed immediately but only after other requests in the queue have
		1539	entered their execution phase. This can be used to measure request
		1540	latency.
		1541
		1542	=item IO::AIO::aio_busy $fractional_seconds, $callback->() *NOT EXPORTED*
		1543
		1544	Mainly used for debugging and benchmarking, this aio request puts one of
		1545	the request workers to sleep for the given time.
		1546
		1547	While it is theoretically handy to have simple I/O scheduling requests
		1548	like sleep and file handle readable/writable, the overhead this creates is
		1549	immense (it blocks a thread for a long time) so do not use this function
		1550	except to put your application under artificial I/O pressure.
		1551
260	=back	1552	=back
261		1553
		1554
		1555	=head2 IO::AIO::WD - multiple working directories
		1556
		1557	Your process only has one current working directory, which is used by all
		1558	threads. This makes it hard to use relative paths (some other component
		1559	could call C<chdir> at any time, and it is hard to control when the path
		1560	will be used by IO::AIO).
		1561
		1562	One solution for this is to always use absolute paths. This usually works,
		1563	but can be quite slow (the kernel has to walk the whole path on every
		1564	access), and can also be a hassle to implement.
		1565
		1566	Newer POSIX systems have a number of functions (openat, fdopendir,
		1567	futimensat and so on) that make it possible to specify working directories
		1568	per operation.
		1569
		1570	For portability, and because the clowns who "designed", or shall I write,
		1571	perpetrated this new interface were obviously half-drunk, this abstraction
		1572	cannot be perfect, though.
		1573
		1574	IO::AIO allows you to convert directory paths into a so-called IO::AIO::WD
		1575	object. This object stores the canonicalised, absolute version of the
		1576	path, and on systems that allow it, also a directory file descriptor.
		1577
		1578	Everywhere where a pathname is accepted by IO::AIO (e.g. in C<aio_stat>
		1579	or C<aio_unlink>), one can specify an array reference with an IO::AIO::WD
		1580	object and a pathname instead (or the IO::AIO::WD object alone, which
		1581	gets interpreted as C<[$wd, "."]>). If the pathname is absolute, the
		1582	IO::AIO::WD object is ignored, otherwise the pathname is resolved relative
		1583	to that IO::AIO::WD object.
		1584
		1585	For example, to get a wd object for F</etc> and then stat F<passwd>
		1586	inside, you would write:
		1587
		1588	aio_wd "/etc", sub {
		1589	my $etcdir = shift;
		1590
		1591	# although $etcdir can be undef on error, there is generally no reason
		1592	# to check for errors here, as aio_stat will fail with ENOENT
		1593	# when $etcdir is undef.
		1594
		1595	aio_stat [$etcdir, "passwd"], sub {
		1596	# yay
		1597	};
		1598	};
		1599
		1600	The fact that C<aio_wd> is a request and not a normal function shows that
		1601	creating an IO::AIO::WD object is itself a potentially blocking operation,
		1602	which is why it is done asynchronously.
		1603
		1604	To stat the directory obtained with C<aio_wd> above, one could write
		1605	either of the following three request calls:
		1606
		1607	aio_lstat "/etc" , sub { ... # pathname as normal string
		1608	aio_lstat [$wd, "."], sub { ... # "." relative to $wd (i.e. $wd itself)
		1609	aio_lstat $wd , sub { ... # shorthand for the previous
		1610
		1611	As with normal pathnames, IO::AIO keeps a copy of the working directory
		1612	object and the pathname string, so you could write the following without
		1613	causing any issues due to C<$path> getting reused:
		1614
		1615	my $path = [$wd, undef];
		1616
		1617	for my $name (qw(abc def ghi)) {
		1618	$path->[1] = $name;
		1619	aio_stat $path, sub {
		1620	# ...
		1621	};
		1622	}
		1623
		1624	There are some caveats: when directories get renamed (or deleted), the
		1625	pathname string doesn't change, so will point to the new directory (or
		1626	nowhere at all), while the directory fd, if available on the system,
		1627	will still point to the original directory. Most functions accepting a
		1628	pathname will use the directory fd on newer systems, and the string on
		1629	older systems. Some functions (such as C<aio_realpath>) will always rely on
		1630	the string form of the pathname.
		1631
		1632	So this functionality is mainly useful to get some protection against
		1633	C<chdir>, to easily get an absolute path out of a relative path for future
		1634	reference, and to speed up doing many operations in the same directory
		1635	(e.g. when stat'ing all files in a directory).
		1636
		1637	The following functions implement this working directory abstraction:
		1638
		1639	=over 4
		1640
		1641	=item aio_wd $pathname, $callback->($wd)
		1642
		1643	Asynchonously canonicalise the given pathname and convert it to an
		1644	IO::AIO::WD object representing it. If possible and supported on the
		1645	system, also open a directory fd to speed up pathname resolution relative
		1646	to this working directory.
		1647
		1648	If something goes wrong, then C<undef> is passwd to the callback instead
		1649	of a working directory object and C<$!> is set appropriately. Since
		1650	passing C<undef> as working directory component of a pathname fails the
		1651	request with C<ENOENT>, there is often no need for error checking in the
		1652	C<aio_wd> callback, as future requests using the value will fail in the
		1653	expected way.
		1654
		1655	=item IO::AIO::CWD
		1656
		1657	This is a compiletime constant (object) that represents the process
		1658	current working directory.
		1659
		1660	Specifying this object as working directory object for a pathname is as if
		1661	the pathname would be specified directly, without a directory object. For
		1662	example, these calls are functionally identical:
		1663
		1664	aio_stat "somefile", sub { ... };
		1665	aio_stat [IO::AIO::CWD, "somefile"], sub { ... };
		1666
		1667	=back
		1668
		1669	To recover the path associated with an IO::AIO::WD object, you can use
		1670	C<aio_realpath>:
		1671
		1672	aio_realpath $wd, sub {
		1673	warn "path is $_[0]\n";
		1674	};
		1675
		1676	Currently, C<aio_statvfs> always, and C<aio_rename> and C<aio_rmdir>
		1677	sometimes, fall back to using an absolue path.
		1678
		1679	=head2 IO::AIO::REQ CLASS
		1680
		1681	All non-aggregate C<aio_*> functions return an object of this class when
		1682	called in non-void context.
		1683
		1684	=over 4
		1685
		1686	=item cancel $req
		1687
		1688	Cancels the request, if possible. Has the effect of skipping execution
		1689	when entering the B<execute> state and skipping calling the callback when
		1690	entering the the B<result> state, but will leave the request otherwise
		1691	untouched (with the exception of readdir). That means that requests that
		1692	currently execute will not be stopped and resources held by the request
		1693	will not be freed prematurely.
		1694
		1695	=item cb $req $callback->(...)
		1696
		1697	Replace (or simply set) the callback registered to the request.
		1698
		1699	=back
		1700
		1701	=head2 IO::AIO::GRP CLASS
		1702
		1703	This class is a subclass of L<IO::AIO::REQ>, so all its methods apply to
		1704	objects of this class, too.
		1705
		1706	A IO::AIO::GRP object is a special request that can contain multiple other
		1707	aio requests.
		1708
		1709	You create one by calling the C<aio_group> constructing function with a
		1710	callback that will be called when all contained requests have entered the
		1711	C<done> state:
		1712
		1713	my $grp = aio_group sub {
		1714	print "all requests are done\n";
		1715	};
		1716
		1717	You add requests by calling the C<add> method with one or more
		1718	C<IO::AIO::REQ> objects:
		1719
		1720	$grp->add (aio_unlink "...");
		1721
		1722	add $grp aio_stat "...", sub {
		1723	$_[0] or return $grp->result ("error");
		1724
		1725	# add another request dynamically, if first succeeded
		1726	add $grp aio_open "...", sub {
		1727	$grp->result ("ok");
		1728	};
		1729	};
		1730
		1731	This makes it very easy to create composite requests (see the source of
		1732	C<aio_move> for an application) that work and feel like simple requests.
		1733
		1734	=over 4
		1735
		1736	=item * The IO::AIO::GRP objects will be cleaned up during calls to
		1737	C<IO::AIO::poll_cb>, just like any other request.
		1738
		1739	=item * They can be canceled like any other request. Canceling will cancel not
		1740	only the request itself, but also all requests it contains.
		1741
		1742	=item * They can also can also be added to other IO::AIO::GRP objects.
		1743
		1744	=item * You must not add requests to a group from within the group callback (or
		1745	any later time).
		1746
		1747	=back
		1748
		1749	Their lifetime, simplified, looks like this: when they are empty, they
		1750	will finish very quickly. If they contain only requests that are in the
		1751	C<done> state, they will also finish. Otherwise they will continue to
		1752	exist.
		1753
		1754	That means after creating a group you have some time to add requests
		1755	(precisely before the callback has been invoked, which is only done within
		1756	the C<poll_cb>). And in the callbacks of those requests, you can add
		1757	further requests to the group. And only when all those requests have
		1758	finished will the the group itself finish.
		1759
		1760	=over 4
		1761
		1762	=item add $grp ...
		1763
		1764	=item $grp->add (...)
		1765
		1766	Add one or more requests to the group. Any type of L<IO::AIO::REQ> can
		1767	be added, including other groups, as long as you do not create circular
		1768	dependencies.
		1769
		1770	Returns all its arguments.
		1771
		1772	=item $grp->cancel_subs
		1773
		1774	Cancel all subrequests and clears any feeder, but not the group request
		1775	itself. Useful when you queued a lot of events but got a result early.
		1776
		1777	The group request will finish normally (you cannot add requests to the
		1778	group).
		1779
		1780	=item $grp->result (...)
		1781
		1782	Set the result value(s) that will be passed to the group callback when all
		1783	subrequests have finished and set the groups errno to the current value
		1784	of errno (just like calling C<errno> without an error number). By default,
		1785	no argument will be passed and errno is zero.
		1786
		1787	=item $grp->errno ([$errno])
		1788
		1789	Sets the group errno value to C<$errno>, or the current value of errno
		1790	when the argument is missing.
		1791
		1792	Every aio request has an associated errno value that is restored when
		1793	the callback is invoked. This method lets you change this value from its
		1794	default (0).
		1795
		1796	Calling C<result> will also set errno, so make sure you either set C<$!>
		1797	before the call to C<result>, or call c<errno> after it.
		1798
		1799	=item feed $grp $callback->($grp)
		1800
		1801	Sets a feeder/generator on this group: every group can have an attached
		1802	generator that generates requests if idle. The idea behind this is that,
		1803	although you could just queue as many requests as you want in a group,
		1804	this might starve other requests for a potentially long time. For example,
		1805	C<aio_scandir> might generate hundreds of thousands of C<aio_stat>
		1806	requests, delaying any later requests for a long time.
		1807
		1808	To avoid this, and allow incremental generation of requests, you can
		1809	instead a group and set a feeder on it that generates those requests. The
		1810	feed callback will be called whenever there are few enough (see C<limit>,
		1811	below) requests active in the group itself and is expected to queue more
		1812	requests.
		1813
		1814	The feed callback can queue as many requests as it likes (i.e. C<add> does
		1815	not impose any limits).
		1816
		1817	If the feed does not queue more requests when called, it will be
		1818	automatically removed from the group.
		1819
		1820	If the feed limit is C<0> when this method is called, it will be set to
		1821	C<2> automatically.
		1822
		1823	Example:
		1824
		1825	# stat all files in @files, but only ever use four aio requests concurrently:
		1826
		1827	my $grp = aio_group sub { print "finished\n" };
		1828	limit $grp 4;
		1829	feed $grp sub {
		1830	my $file = pop @files
		1831	or return;
		1832
		1833	add $grp aio_stat $file, sub { ... };
		1834	};
		1835
		1836	=item limit $grp $num
		1837
		1838	Sets the feeder limit for the group: The feeder will be called whenever
		1839	the group contains less than this many requests.
		1840
		1841	Setting the limit to C<0> will pause the feeding process.
		1842
		1843	The default value for the limit is C<0>, but note that setting a feeder
		1844	automatically bumps it up to C<2>.
		1845
		1846	=back
		1847
		1848
262	=head2 SUPPORT FUNCTIONS	1849	=head2 SUPPORT FUNCTIONS
263		1850
		1851	=head3 EVENT PROCESSING AND EVENT LOOP INTEGRATION
		1852
264	=over 4	1853	=over 4
265		1854
266	=item $fileno = IO::AIO::poll_fileno	1855	=item $fileno = IO::AIO::poll_fileno
267		1856
268	Return the I<request result pipe file descriptor>. This filehandle must be	1857	Return the I<request result pipe file descriptor>. This filehandle must be
269	polled for reading by some mechanism outside this module (e.g. Event or	1858	polled for reading by some mechanism outside this module (e.g. EV, Glib,
270	select, see below or the SYNOPSIS). If the pipe becomes readable you have	1859	select and so on, see below or the SYNOPSIS). If the pipe becomes readable
271	to call C<poll_cb> to check the results.	1860	you have to call C<poll_cb> to check the results.
272		1861
273	See C<poll_cb> for an example.	1862	See C<poll_cb> for an example.
274		1863
275	=item IO::AIO::poll_cb	1864	=item IO::AIO::poll_cb
276		1865
277	Process all outstanding events on the result pipe. You have to call this	1866	Process some requests that have reached the result phase (i.e. they have
278	regularly. Returns the number of events processed. Returns immediately	1867	been executed but the results are not yet reported). You have to call
279	when no events are outstanding.	1868	this "regularly" to finish outstanding requests.
		1869
		1870	Returns C<0> if all events could be processed (or there were no
		1871	events to process), or C<-1> if it returned earlier for whatever
		1872	reason. Returns immediately when no events are outstanding. The amount
		1873	of events processed depends on the settings of C<IO::AIO::max_poll_req>,
		1874	C<IO::AIO::max_poll_time> and C<IO::AIO::max_outstanding>.
		1875
		1876	If not all requests were processed for whatever reason, the poll file
		1877	descriptor will still be ready when C<poll_cb> returns, so normally you
		1878	don't have to do anything special to have it called later.
		1879
		1880	Apart from calling C<IO::AIO::poll_cb> when the event filehandle becomes
		1881	ready, it can be beneficial to call this function from loops which submit
		1882	a lot of requests, to make sure the results get processed when they become
		1883	available and not just when the loop is finished and the event loop takes
		1884	over again. This function returns very fast when there are no outstanding
		1885	requests.
280		1886
281	Example: Install an Event watcher that automatically calls	1887	Example: Install an Event watcher that automatically calls
282	IO::AIO::poll_cb with high priority:	1888	IO::AIO::poll_cb with high priority (more examples can be found in the
		1889	SYNOPSIS section, at the top of this document):
283		1890
284	Event->io (fd => IO::AIO::poll_fileno,	1891	Event->io (fd => IO::AIO::poll_fileno,
285	poll => 'r', async => 1,	1892	poll => 'r', async => 1,
286	cb => \&IO::AIO::poll_cb);	1893	cb => \&IO::AIO::poll_cb);
287		1894
288	=item IO::AIO::poll_wait	1895	=item IO::AIO::poll_wait
289		1896
290	Wait till the result filehandle becomes ready for reading (simply does a	1897	Wait until either at least one request is in the result phase or no
291	C<select> on the filehandle. This is useful if you want to synchronously wait	1898	requests are outstanding anymore.
292	for some requests to finish).	1899
		1900	This is useful if you want to synchronously wait for some requests to
		1901	become ready, without actually handling them.
293		1902
294	See C<nreqs> for an example.	1903	See C<nreqs> for an example.
295		1904
		1905	=item IO::AIO::poll
		1906
		1907	Waits until some requests have been handled.
		1908
		1909	Returns the number of requests processed, but is otherwise strictly
		1910	equivalent to:
		1911
		1912	IO::AIO::poll_wait, IO::AIO::poll_cb
		1913
296	=item IO::AIO::nreqs	1914	=item IO::AIO::flush
297		1915
298	Returns the number of requests currently outstanding (i.e. for which their	1916	Wait till all outstanding AIO requests have been handled.
299	callback has not been invoked yet).
300		1917
301	Example: wait till there are no outstanding requests anymore:	1918	Strictly equivalent to:
302		1919
303	IO::AIO::poll_wait, IO::AIO::poll_cb	1920	IO::AIO::poll_wait, IO::AIO::poll_cb
304	while IO::AIO::nreqs;	1921	while IO::AIO::nreqs;
305		1922
		1923	This function can be useful at program aborts, to make sure outstanding
		1924	I/O has been done (C<IO::AIO> uses an C<END> block which already calls
		1925	this function on normal exits), or when you are merely using C<IO::AIO>
		1926	for its more advanced functions, rather than for async I/O, e.g.:
		1927
		1928	my ($dirs, $nondirs);
		1929	IO::AIO::aio_scandir "/tmp", 0, sub { ($dirs, $nondirs) = @_ };
306	=item IO::AIO::flush	1930	IO::AIO::flush;
		1931	# $dirs, $nondirs are now set
307		1932
308	Wait till all outstanding AIO requests have been handled.	1933	=item IO::AIO::max_poll_reqs $nreqs
309		1934
310	Strictly equivalent to:	1935	=item IO::AIO::max_poll_time $seconds
311		1936
312	IO::AIO::poll_wait, IO::AIO::poll_cb	1937	These set the maximum number of requests (default C<0>, meaning infinity)
313	while IO::AIO::nreqs;	1938	that are being processed by C<IO::AIO::poll_cb> in one call, respectively
		1939	the maximum amount of time (default C<0>, meaning infinity) spent in
		1940	C<IO::AIO::poll_cb> to process requests (more correctly the mininum amount
		1941	of time C<poll_cb> is allowed to use).
314		1942
315	=item IO::AIO::poll	1943	Setting C<max_poll_time> to a non-zero value creates an overhead of one
		1944	syscall per request processed, which is not normally a problem unless your
		1945	callbacks are really really fast or your OS is really really slow (I am
		1946	not mentioning Solaris here). Using C<max_poll_reqs> incurs no overhead.
316		1947
317	Waits until some requests have been handled.	1948	Setting these is useful if you want to ensure some level of
		1949	interactiveness when perl is not fast enough to process all requests in
		1950	time.
318		1951
319	Strictly equivalent to:	1952	For interactive programs, values such as C<0.01> to C<0.1> should be fine.
320		1953
321	IO::AIO::poll_wait, IO::AIO::poll_cb	1954	Example: Install an Event watcher that automatically calls
322	if IO::AIO::nreqs;	1955	IO::AIO::poll_cb with low priority, to ensure that other parts of the
		1956	program get the CPU sometimes even under high AIO load.
		1957
		1958	# try not to spend much more than 0.1s in poll_cb
		1959	IO::AIO::max_poll_time 0.1;
		1960
		1961	# use a low priority so other tasks have priority
		1962	Event->io (fd => IO::AIO::poll_fileno,
		1963	poll => 'r', nice => 1,
		1964	cb => &IO::AIO::poll_cb);
		1965
		1966	=back
		1967
		1968
		1969	=head3 CONTROLLING THE NUMBER OF THREADS
		1970
		1971	=over
323		1972
324	=item IO::AIO::min_parallel $nthreads	1973	=item IO::AIO::min_parallel $nthreads
325		1974
326	Set the minimum number of AIO threads to C<$nthreads>. The current default	1975	Set the minimum number of AIO threads to C<$nthreads>. The current
327	is C<4>, which means four asynchronous operations can be done at one time	1976	default is C<8>, which means eight asynchronous operations can execute
328	(the number of outstanding operations, however, is unlimited).	1977	concurrently at any one time (the number of outstanding requests,
		1978	however, is unlimited).
329		1979
330	IO::AIO starts threads only on demand, when an AIO request is queued and	1980	IO::AIO starts threads only on demand, when an AIO request is queued and
331	no free thread exists.	1981	no free thread exists. Please note that queueing up a hundred requests can
		1982	create demand for a hundred threads, even if it turns out that everything
		1983	is in the cache and could have been processed faster by a single thread.
332		1984
333	It is recommended to keep the number of threads low, as some Linux	1985	It is recommended to keep the number of threads relatively low, as some
334	kernel versions will scale negatively with the number of threads (higher	1986	Linux kernel versions will scale negatively with the number of threads
335	parallelity => MUCH higher latency). With current Linux 2.6 versions, 4-32	1987	(higher parallelity => MUCH higher latency). With current Linux 2.6
336	threads should be fine.	1988	versions, 4-32 threads should be fine.
337		1989
338	Under most circumstances you don't need to call this function, as the	1990	Under most circumstances you don't need to call this function, as the
339	module selects a default that is suitable for low to moderate load.	1991	module selects a default that is suitable for low to moderate load.
340		1992
341	=item IO::AIO::max_parallel $nthreads	1993	=item IO::AIO::max_parallel $nthreads
…		…
350	This module automatically runs C<max_parallel 0> at program end, to ensure	2002	This module automatically runs C<max_parallel 0> at program end, to ensure
351	that all threads are killed and that there are no outstanding requests.	2003	that all threads are killed and that there are no outstanding requests.
352		2004
353	Under normal circumstances you don't need to call this function.	2005	Under normal circumstances you don't need to call this function.
354		2006
		2007	=item IO::AIO::max_idle $nthreads
		2008
		2009	Limit the number of threads (default: 4) that are allowed to idle
		2010	(i.e., threads that did not get a request to process within the idle
		2011	timeout (default: 10 seconds). That means if a thread becomes idle while
		2012	C<$nthreads> other threads are also idle, it will free its resources and
		2013	exit.
		2014
		2015	This is useful when you allow a large number of threads (e.g. 100 or 1000)
		2016	to allow for extremely high load situations, but want to free resources
		2017	under normal circumstances (1000 threads can easily consume 30MB of RAM).
		2018
		2019	The default is probably ok in most situations, especially if thread
		2020	creation is fast. If thread creation is very slow on your system you might
		2021	want to use larger values.
		2022
		2023	=item IO::AIO::idle_timeout $seconds
		2024
		2025	Sets the minimum idle timeout (default 10) after which worker threads are
		2026	allowed to exit. SEe C<IO::AIO::max_idle>.
		2027
355	=item $oldnreqs = IO::AIO::max_outstanding $nreqs	2028	=item IO::AIO::max_outstanding $maxreqs
356		2029
357	Sets the maximum number of outstanding requests to C<$nreqs>. If you	2030	Sets the maximum number of outstanding requests to C<$nreqs>. If
358	try to queue up more than this number of requests, the caller will block until	2031	you do queue up more than this number of requests, the next call to
359	some requests have been handled.	2032	C<IO::AIO::poll_cb> (and other functions calling C<poll_cb>, such as
		2033	C<IO::AIO::flush> or C<IO::AIO::poll>) will block until the limit is no
		2034	longer exceeded.
360		2035
361	The default is very large, so normally there is no practical limit. If you	2036	In other words, this setting does not enforce a queue limit, but can be
362	queue up many requests in a loop it often improves speed if you set	2037	used to make poll functions block if the limit is exceeded.
363	this to a relatively low number, such as C<100>.
364		2038
365	Under normal circumstances you don't need to call this function.	2039	This is a very bad function to use in interactive programs because it
		2040	blocks, and a bad way to reduce concurrency because it is inexact: Better
		2041	use an C<aio_group> together with a feed callback.
		2042
		2043	Its main use is in scripts without an event loop - when you want to stat
		2044	a lot of files, you can write something like this:
		2045
		2046	IO::AIO::max_outstanding 32;
		2047
		2048	for my $path (...) {
		2049	aio_stat $path , ...;
		2050	IO::AIO::poll_cb;
		2051	}
		2052
		2053	IO::AIO::flush;
		2054
		2055	The call to C<poll_cb> inside the loop will normally return instantly, but
		2056	as soon as more thna C<32> reqeusts are in-flight, it will block until
		2057	some requests have been handled. This keeps the loop from pushing a large
		2058	number of C<aio_stat> requests onto the queue.
		2059
		2060	The default value for C<max_outstanding> is very large, so there is no
		2061	practical limit on the number of outstanding requests.
366		2062
367	=back	2063	=back
368		2064
		2065
		2066	=head3 STATISTICAL INFORMATION
		2067
		2068	=over
		2069
		2070	=item IO::AIO::nreqs
		2071
		2072	Returns the number of requests currently in the ready, execute or pending
		2073	states (i.e. for which their callback has not been invoked yet).
		2074
		2075	Example: wait till there are no outstanding requests anymore:
		2076
		2077	IO::AIO::poll_wait, IO::AIO::poll_cb
		2078	while IO::AIO::nreqs;
		2079
		2080	=item IO::AIO::nready
		2081
		2082	Returns the number of requests currently in the ready state (not yet
		2083	executed).
		2084
		2085	=item IO::AIO::npending
		2086
		2087	Returns the number of requests currently in the pending state (executed,
		2088	but not yet processed by poll_cb).
		2089
		2090	=back
		2091
		2092
		2093	=head3 SUBSECOND STAT TIME ACCESS
		2094
		2095	Both C<aio_stat>/C<aio_lstat> and perl's C<stat>/C<lstat> functions can
		2096	generally find access/modification and change times with subsecond time
		2097	accuracy of the system supports it, but perl's built-in functions only
		2098	return the integer part.
		2099
		2100	The following functions return the timestamps of the most recent
		2101	stat with subsecond precision on most systems and work both after
		2102	C<aio_stat>/C<aio_lstat> and perl's C<stat>/C<lstat> calls. Their return
		2103	value is only meaningful after a successful C<stat>/C<lstat> call, or
		2104	during/after a successful C<aio_stat>/C<aio_lstat> callback.
		2105
		2106	This is similar to the L<Time::HiRes> C<stat> functions, but can return
		2107	full resolution without rounding and work with standard perl C<stat>,
		2108	alleviating the need to call the special C<Time::HiRes> functions, which
		2109	do not act like their perl counterparts.
		2110
		2111	On operating systems or file systems where subsecond time resolution is
		2112	not supported or could not be detected, a fractional part of C<0> is
		2113	returned, so it is always safe to call these functions.
		2114
		2115	=over 4
		2116
		2117	=item $seconds = IO::AIO::st_atime, IO::AIO::st_mtime, IO::AIO::st_ctime, IO::AIO::st_btime
		2118
		2119	Return the access, modication, change or birth time, respectively,
		2120	including fractional part. Due to the limited precision of floating point,
		2121	the accuracy on most platforms is only a bit better than milliseconds
		2122	for times around now - see the I<nsec> function family, below, for full
		2123	accuracy.
		2124
		2125	File birth time is only available when the OS and perl support it (on
		2126	FreeBSD and NetBSD at the time of this writing, although support is
		2127	adaptive, so if your OS/perl gains support, IO::AIO can take avdantage of
		2128	it). On systems where it isn't available, C<0> is currently returned, but
		2129	this might change to C<undef> in a future version.
		2130
		2131	=item ($atime, $mtime, $ctime, $btime, ...) = IO::AIO::st_xtime
		2132
		2133	Returns access, modification, change and birth time all in one go, and
		2134	maybe more times in the future version.
		2135
		2136	=item $nanoseconds = IO::AIO::st_atimensec, IO::AIO::st_mtimensec, IO::AIO::st_ctimensec, IO::AIO::st_btimensec
		2137
		2138	Return the fractional access, modifcation, change or birth time, in nanoseconds,
		2139	as an integer in the range C<0> to C<999999999>.
		2140
		2141	Note that no accessors are provided for access, modification and
		2142	change times - you need to get those from C<stat _> if required (C<int
		2143	IO::AIO::st_atime> and so on will I<not> generally give you the correct
		2144	value).
		2145
		2146	=item $seconds = IO::AIO::st_btimesec
		2147
		2148	The (integral) seconds part of the file birth time, if available.
		2149
		2150	=item ($atime, $mtime, $ctime, $btime, ...) = IO::AIO::st_xtimensec
		2151
		2152	Like the functions above, but returns all four times in one go (and maybe
		2153	more in future versions).
		2154
		2155	=item $counter = IO::AIO::st_gen
		2156
		2157	Returns the generation counter (in practice this is just a random number)
		2158	of the file. This is only available on platforms which have this member in
		2159	their C<struct stat> (most BSDs at the time of this writing) and generally
		2160	only to the root usert. If unsupported, C<0> is returned, but this might
		2161	change to C<undef> in a future version.
		2162
		2163	=back
		2164
		2165	Example: print the high resolution modification time of F</etc>, using
		2166	C<stat>, and C<IO::AIO::aio_stat>.
		2167
		2168	if (stat "/etc") {
		2169	printf "stat(/etc) mtime: %f\n", IO::AIO::st_mtime;
		2170	}
		2171
		2172	IO::AIO::aio_stat "/etc", sub {
		2173	$_[0]
		2174	and return;
		2175
		2176	printf "aio_stat(/etc) mtime: %d.%09d\n", (stat _)[9], IO::AIO::st_mtimensec;
		2177	};
		2178
		2179	IO::AIO::flush;
		2180
		2181	Output of the awbove on my system, showing reduced and full accuracy:
		2182
		2183	stat(/etc) mtime: 1534043702.020808
		2184	aio_stat(/etc) mtime: 1534043702.020807792
		2185
		2186
		2187	=head3 MISCELLANEOUS FUNCTIONS
		2188
		2189	IO::AIO implements some functions that are useful when you want to use
		2190	some "Advanced I/O" function not available to in Perl, without going the
		2191	"Asynchronous I/O" route. Many of these have an asynchronous C<aio_*>
		2192	counterpart.
		2193
		2194	=over 4
		2195
		2196	=item $numfd = IO::AIO::get_fdlimit
		2197
		2198	This function is I<EXPERIMENTAL> and subject to change.
		2199
		2200	Tries to find the current file descriptor limit and returns it, or
		2201	C<undef> and sets C<$!> in case of an error. The limit is one larger than
		2202	the highest valid file descriptor number.
		2203
		2204	=item IO::AIO::min_fdlimit [$numfd]
		2205
		2206	This function is I<EXPERIMENTAL> and subject to change.
		2207
		2208	Try to increase the current file descriptor limit(s) to at least C<$numfd>
		2209	by changing the soft or hard file descriptor resource limit. If C<$numfd>
		2210	is missing, it will try to set a very high limit, although this is not
		2211	recommended when you know the actual minimum that you require.
		2212
		2213	If the limit cannot be raised enough, the function makes a best-effort
		2214	attempt to increase the limit as much as possible, using various
		2215	tricks, while still failing. You can query the resulting limit using
		2216	C<IO::AIO::get_fdlimit>.
		2217
		2218	If an error occurs, returns C<undef> and sets C<$!>, otherwise returns
		2219	true.
		2220
		2221	=item IO::AIO::sendfile $ofh, $ifh, $offset, $count
		2222
		2223	Calls the C<eio_sendfile_sync> function, which is like C<aio_sendfile>,
		2224	but is blocking (this makes most sense if you know the input data is
		2225	likely cached already and the output filehandle is set to non-blocking
		2226	operations).
		2227
		2228	Returns the number of bytes copied, or C<-1> on error.
		2229
		2230	=item IO::AIO::fadvise $fh, $offset, $len, $advice
		2231
		2232	Simply calls the C<posix_fadvise> function (see its
		2233	manpage for details). The following advice constants are
		2234	available: C<IO::AIO::FADV_NORMAL>, C<IO::AIO::FADV_SEQUENTIAL>,
		2235	C<IO::AIO::FADV_RANDOM>, C<IO::AIO::FADV_NOREUSE>,
		2236	C<IO::AIO::FADV_WILLNEED>, C<IO::AIO::FADV_DONTNEED>.
		2237
		2238	On systems that do not implement C<posix_fadvise>, this function returns
		2239	ENOSYS, otherwise the return value of C<posix_fadvise>.
		2240
		2241	=item IO::AIO::madvise $scalar, $offset, $len, $advice
		2242
		2243	Simply calls the C<posix_madvise> function (see its
		2244	manpage for details). The following advice constants are
		2245	available: C<IO::AIO::MADV_NORMAL>, C<IO::AIO::MADV_SEQUENTIAL>,
		2246	C<IO::AIO::MADV_RANDOM>, C<IO::AIO::MADV_WILLNEED>,
		2247	C<IO::AIO::MADV_DONTNEED>.
		2248
		2249	If C<$offset> is negative, counts from the end. If C<$length> is negative,
		2250	the remaining length of the C<$scalar> is used. If possible, C<$length>
		2251	will be reduced to fit into the C<$scalar>.
		2252
		2253	On systems that do not implement C<posix_madvise>, this function returns
		2254	ENOSYS, otherwise the return value of C<posix_madvise>.
		2255
		2256	=item IO::AIO::mprotect $scalar, $offset, $len, $protect
		2257
		2258	Simply calls the C<mprotect> function on the preferably AIO::mmap'ed
		2259	$scalar (see its manpage for details). The following protect
		2260	constants are available: C<IO::AIO::PROT_NONE>, C<IO::AIO::PROT_READ>,
		2261	C<IO::AIO::PROT_WRITE>, C<IO::AIO::PROT_EXEC>.
		2262
		2263	If C<$offset> is negative, counts from the end. If C<$length> is negative,
		2264	the remaining length of the C<$scalar> is used. If possible, C<$length>
		2265	will be reduced to fit into the C<$scalar>.
		2266
		2267	On systems that do not implement C<mprotect>, this function returns
		2268	ENOSYS, otherwise the return value of C<mprotect>.
		2269
		2270	=item IO::AIO::mmap $scalar, $length, $prot, $flags, $fh[, $offset]
		2271
		2272	Memory-maps a file (or anonymous memory range) and attaches it to the
		2273	given C<$scalar>, which will act like a string scalar. Returns true on
		2274	success, and false otherwise.
		2275
		2276	The scalar must exist, but its contents do not matter - this means you
		2277	cannot use a nonexistant array or hash element. When in doubt, C<undef>
		2278	the scalar first.
		2279
		2280	The only operations allowed on the mmapped scalar are C<substr>/C<vec>,
		2281	which don't change the string length, and most read-only operations such
		2282	as copying it or searching it with regexes and so on.
		2283
		2284	Anything else is unsafe and will, at best, result in memory leaks.
		2285
		2286	The memory map associated with the C<$scalar> is automatically removed
		2287	when the C<$scalar> is undef'd or destroyed, or when the C<IO::AIO::mmap>
		2288	or C<IO::AIO::munmap> functions are called on it.
		2289
		2290	This calls the C<mmap>(2) function internally. See your system's manual
		2291	page for details on the C<$length>, C<$prot> and C<$flags> parameters.
		2292
		2293	The C<$length> must be larger than zero and smaller than the actual
		2294	filesize.
		2295
		2296	C<$prot> is a combination of C<IO::AIO::PROT_NONE>, C<IO::AIO::PROT_EXEC>,
		2297	C<IO::AIO::PROT_READ> and/or C<IO::AIO::PROT_WRITE>,
		2298
		2299	C<$flags> can be a combination of
		2300	C<IO::AIO::MAP_SHARED> or
		2301	C<IO::AIO::MAP_PRIVATE>,
		2302	or a number of system-specific flags (when not available, the are C<0>):
		2303	C<IO::AIO::MAP_ANONYMOUS> (which is set to C<MAP_ANON> if your system only provides this constant),
		2304	C<IO::AIO::MAP_LOCKED>,
		2305	C<IO::AIO::MAP_NORESERVE>,
		2306	C<IO::AIO::MAP_POPULATE>,
		2307	C<IO::AIO::MAP_NONBLOCK>,
		2308	C<IO::AIO::MAP_FIXED>,
		2309	C<IO::AIO::MAP_GROWSDOWN>,
		2310	C<IO::AIO::MAP_32BIT>,
		2311	C<IO::AIO::MAP_HUGETLB> or
		2312	C<IO::AIO::MAP_STACK>.
		2313
		2314	If C<$fh> is C<undef>, then a file descriptor of C<-1> is passed.
		2315
		2316	C<$offset> is the offset from the start of the file - it generally must be
		2317	a multiple of C<IO::AIO::PAGESIZE> and defaults to C<0>.
		2318
		2319	Example:
		2320
		2321	use Digest::MD5;
		2322	use IO::AIO;
		2323
		2324	open my $fh, "<verybigfile"
		2325	or die "$!";
		2326
		2327	IO::AIO::mmap my $data, -s $fh, IO::AIO::PROT_READ, IO::AIO::MAP_SHARED, $fh
		2328	or die "verybigfile: $!";
		2329
		2330	my $fast_md5 = md5 $data;
		2331
		2332	=item IO::AIO::munmap $scalar
		2333
		2334	Removes a previous mmap and undefines the C<$scalar>.
		2335
		2336	=item IO::AIO::mremap $scalar, $new_length, $flags = MREMAP_MAYMOVE[, $new_address = 0]
		2337
		2338	Calls the Linux-specific mremap(2) system call. The C<$scalar> must have
		2339	been mapped by C<IO::AIO::mmap>, and C<$flags> must currently either be
		2340	C<0> or C<IO::AIO::MREMAP_MAYMOVE>.
		2341
		2342	Returns true if successful, and false otherwise. If the underlying mmapped
		2343	region has changed address, then the true value has the numerical value
		2344	C<1>, otherwise it has the numerical value C<0>:
		2345
		2346	my $success = IO::AIO::mremap $mmapped, 8192, IO::AIO::MREMAP_MAYMOVE
		2347	or die "mremap: $!";
		2348
		2349	if ($success*1) {
		2350	warn "scalar has chanegd address in memory\n";
		2351	}
		2352
		2353	C<IO::AIO::MREMAP_FIXED> and the C<$new_address> argument are currently
		2354	implemented, but not supported and might go away in a future version.
		2355
		2356	On systems where this call is not supported or is not emulated, this call
		2357	returns falls and sets C<$!> to C<ENOSYS>.
		2358
		2359	=item IO::AIO::mlockall $flags
		2360
		2361	Calls the C<eio_mlockall_sync> function, which is like C<aio_mlockall>,
		2362	but is blocking.
		2363
		2364	=item IO::AIO::munlock $scalar, $offset = 0, $length = undef
		2365
		2366	Calls the C<munlock> function, undoing the effects of a previous
		2367	C<aio_mlock> call (see its description for details).
		2368
		2369	=item IO::AIO::munlockall
		2370
		2371	Calls the C<munlockall> function.
		2372
		2373	On systems that do not implement C<munlockall>, this function returns
		2374	ENOSYS, otherwise the return value of C<munlockall>.
		2375
		2376	=item IO::AIO::splice $r_fh, $r_off, $w_fh, $w_off, $length, $flags
		2377
		2378	Calls the GNU/Linux C<splice(2)> syscall, if available. If C<$r_off> or
		2379	C<$w_off> are C<undef>, then C<NULL> is passed for these, otherwise they
		2380	should be the file offset.
		2381
		2382	C<$r_fh> and C<$w_fh> should not refer to the same file, as splice might
		2383	silently corrupt the data in this case.
		2384
		2385	The following symbol flag values are available: C<IO::AIO::SPLICE_F_MOVE>,
		2386	C<IO::AIO::SPLICE_F_NONBLOCK>, C<IO::AIO::SPLICE_F_MORE> and
		2387	C<IO::AIO::SPLICE_F_GIFT>.
		2388
		2389	See the C<splice(2)> manpage for details.
		2390
		2391	=item IO::AIO::tee $r_fh, $w_fh, $length, $flags
		2392
		2393	Calls the GNU/Linux C<tee(2)> syscall, see its manpage and the
		2394	description for C<IO::AIO::splice> above for details.
		2395
		2396	=item $actual_size = IO::AIO::pipesize $r_fh[, $new_size]
		2397
		2398	Attempts to query or change the pipe buffer size. Obviously works only
		2399	on pipes, and currently works only on GNU/Linux systems, and fails with
		2400	C<-1>/C<ENOSYS> everywhere else. If anybody knows how to influence pipe buffer
		2401	size on other systems, drop me a note.
		2402
		2403	=item ($rfh, $wfh) = IO::AIO::pipe2 [$flags]
		2404
		2405	This is a direct interface to the Linux L<pipe2(2)> system call. If
		2406	C<$flags> is missing or C<0>, then this should be the same as a call to
		2407	perl's built-in C<pipe> function and create a new pipe, and works on
		2408	systems that lack the pipe2 syscall. On win32, this case invokes C<_pipe
		2409	(..., 4096, O_BINARY)>.
		2410
		2411	If C<$flags> is non-zero, it tries to invoke the pipe2 system call with
		2412	the given flags (Linux 2.6.27, glibc 2.9).
		2413
		2414	On success, the read and write file handles are returned.
		2415
		2416	On error, nothing will be returned. If the pipe2 syscall is missing and
		2417	C<$flags> is non-zero, fails with C<ENOSYS>.
		2418
		2419	Please refer to L<pipe2(2)> for more info on the C<$flags>, but at the
		2420	time of this writing, C<IO::AIO::O_CLOEXEC>, C<IO::AIO::O_NONBLOCK> and
		2421	C<IO::AIO::O_DIRECT> (Linux 3.4, for packet-based pipes) were supported.
		2422
		2423	Example: create a pipe race-free w.r.t. threads and fork:
		2424
		2425	my ($rfh, $wfh) = IO::AIO::pipe2 IO::AIO::O_CLOEXEC
		2426	or die "pipe2: $!\n";
		2427
		2428	=item $fh = IO::AIO::eventfd [$initval, [$flags]]
		2429
		2430	This is a direct interface to the Linux L<eventfd(2)> system call. The
		2431	(unhelpful) defaults for C<$initval> and C<$flags> are C<0> for both.
		2432
		2433	On success, the new eventfd filehandle is returned, otherwise returns
		2434	C<undef>. If the eventfd syscall is missing, fails with C<ENOSYS>.
		2435
		2436	Please refer to L<eventfd(2)> for more info on this call.
		2437
		2438	The following symbol flag values are available: C<IO::AIO::EFD_CLOEXEC>,
		2439	C<IO::AIO::EFD_NONBLOCK> and C<IO::AIO::EFD_SEMAPHORE> (Linux 2.6.30).
		2440
		2441	Example: create a new eventfd filehandle:
		2442
		2443	$fh = IO::AIO::eventfd 0, IO::AIO::O_CLOEXEC
		2444	or die "eventfd: $!\n";
		2445
		2446	=item $fh = IO::AIO::timerfd_create $clockid[, $flags]
		2447
		2448	This is a direct interface to the Linux L<timerfd_create(2)> system call. The
		2449	(unhelpful) default for C<$flags> is C<0>.
		2450
		2451	On success, the new timerfd filehandle is returned, otherwise returns
		2452	C<undef>. If the eventfd syscall is missing, fails with C<ENOSYS>.
		2453
		2454	Please refer to L<timerfd_create(2)> for more info on this call.
		2455
		2456	The following C<$clockid> values are
		2457	available: C<IO::AIO::CLOCK_REALTIME>, C<IO::AIO::CLOCK_MONOTONIC>
		2458	C<IO::AIO::CLOCK_CLOCK_BOOTTIME> (Linux 3.15)
		2459	C<IO::AIO::CLOCK_CLOCK_REALTIME_ALARM> (Linux 3.11) and
		2460	C<IO::AIO::CLOCK_CLOCK_BOOTTIME_ALARM> (Linux 3.11).
		2461
		2462	The following C<$flags> values are available (Linux
		2463	2.6.27): C<IO::AIO::TFD_NONBLOCK> and C<IO::AIO::TFD_CLOEXEC>.
		2464
		2465	Example: create a new timerfd and set it to one-second repeated alarms,
		2466	then wait for two alarms:
		2467
		2468	my $fh = IO::AIO::timerfd_create IO::AIO::CLOCK_BOOTTIME, IO::AIO::TFD_CLOEXEC
		2469	or die "timerfd_create: $!\n";
		2470
		2471	defined IO::AIO::timerfd_settime $fh, 0, 1, 1
		2472	or die "timerfd_settime: $!\n";
		2473
		2474	for (1..2) {
		2475	8 == sysread $fh, my $buf, 8
		2476	or die "timerfd read failure\n";
		2477
		2478	printf "number of expirations (likely 1): %d\n",
		2479	unpack "Q", $buf;
		2480	}
		2481
		2482	=item ($cur_interval, $cur_value) = IO::AIO::timerfd_settime $fh, $flags, $new_interval, $nbw_value
		2483
		2484	This is a direct interface to the Linux L<timerfd_settime(2)> system
		2485	call. Please refer to its manpage for more info on this call.
		2486
		2487	The new itimerspec is specified using two (possibly fractional) second
		2488	values, C<$new_interval> and C<$new_value>).
		2489
		2490	On success, the current interval and value are returned (as per
		2491	C<timerfd_gettime>). On failure, the empty list is returned.
		2492
		2493	The following C<$flags> values are
		2494	available: C<IO::AIO::TFD_TIMER_ABSTIME> and
		2495	C<IO::AIO::TFD_TIMER_CANCEL_ON_SET>.
		2496
		2497	See C<IO::AIO::timerfd_create> for a full example.
		2498
		2499	=item ($cur_interval, $cur_value) = IO::AIO::timerfd_gettime $fh
		2500
		2501	This is a direct interface to the Linux L<timerfd_gettime(2)> system
		2502	call. Please refer to its manpage for more info on this call.
		2503
		2504	On success, returns the current values of interval and value for the given
		2505	timerfd (as potentially fractional second values). On failure, the empty
		2506	list is returned.
		2507
		2508	=back
		2509
369	=cut	2510	=cut
370		2511
371	# support function to convert a fd into a perl filehandle
372	sub _fd2fh {
373	return undef if $_[0] < 0;
374
375	# try to generate nice filehandles
376	my $sym = "IO::AIO::fd#$_[0]";
377	local *$sym;
378
379	open *$sym, "+<&=$_[0]" # usually works under any unix
380	or open *$sym, "<&=$_[0]" # cygwin needs this
381	or open *$sym, ">&=$_[0]" # or this
382	or return undef;
383
384	*$sym
385	}
386
387	min_parallel 4;	2512	min_parallel 8;
388		2513
389	END {	2514	END { flush }
390	max_parallel 0;
391	}
392		2515
393	1;	2516	1;
394		2517
		2518	=head1 EVENT LOOP INTEGRATION
		2519
		2520	It is recommended to use L<AnyEvent::AIO> to integrate IO::AIO
		2521	automatically into many event loops:
		2522
		2523	# AnyEvent integration (EV, Event, Glib, Tk, POE, urxvt, pureperl...)
		2524	use AnyEvent::AIO;
		2525
		2526	You can also integrate IO::AIO manually into many event loops, here are
		2527	some examples of how to do this:
		2528
		2529	# EV integration
		2530	my $aio_w = EV::io IO::AIO::poll_fileno, EV::READ, \&IO::AIO::poll_cb;
		2531
		2532	# Event integration
		2533	Event->io (fd => IO::AIO::poll_fileno,
		2534	poll => 'r',
		2535	cb => \&IO::AIO::poll_cb);
		2536
		2537	# Glib/Gtk2 integration
		2538	add_watch Glib::IO IO::AIO::poll_fileno,
		2539	in => sub { IO::AIO::poll_cb; 1 };
		2540
		2541	# Tk integration
		2542	Tk::Event::IO->fileevent (IO::AIO::poll_fileno, "",
		2543	readable => \&IO::AIO::poll_cb);
		2544
		2545	# Danga::Socket integration
		2546	Danga::Socket->AddOtherFds (IO::AIO::poll_fileno =>
		2547	\&IO::AIO::poll_cb);
		2548
395	=head2 FORK BEHAVIOUR	2549	=head2 FORK BEHAVIOUR
396		2550
397	Before the fork, IO::AIO enters a quiescent state where no requests	2551	Usage of pthreads in a program changes the semantics of fork
398	can be added in other threads and no results will be processed. After	2552	considerably. Specifically, only async-safe functions can be called after
399	the fork the parent simply leaves the quiescent state and continues	2553	fork. Perl doesn't know about this, so in general, you cannot call fork
400	request/result processing, while the child clears the request/result	2554	with defined behaviour in perl if pthreads are involved. IO::AIO uses
401	queue (so the requests started before the fork will only be handled in	2555	pthreads, so this applies, but many other extensions and (for inexplicable
402	the parent). Threats will be started on demand until the limit ste in the	2556	reasons) perl itself often is linked against pthreads, so this limitation
403	parent process has been reached again.	2557	applies to quite a lot of perls.
		2558
		2559	This module no longer tries to fight your OS, or POSIX. That means IO::AIO
		2560	only works in the process that loaded it. Forking is fully supported, but
		2561	using IO::AIO in the child is not.
		2562
		2563	You might get around by not I<using> IO::AIO before (or after)
		2564	forking. You could also try to call the L<IO::AIO::reinit> function in the
		2565	child:
		2566
		2567	=over 4
		2568
		2569	=item IO::AIO::reinit
		2570
		2571	Abandons all current requests and I/O threads and simply reinitialises all
		2572	data structures. This is not an operation supported by any standards, but
		2573	happens to work on GNU/Linux and some newer BSD systems.
		2574
		2575	The only reasonable use for this function is to call it after forking, if
		2576	C<IO::AIO> was used in the parent. Calling it while IO::AIO is active in
		2577	the process will result in undefined behaviour. Calling it at any time
		2578	will also result in any undefined (by POSIX) behaviour.
		2579
		2580	=back
		2581
		2582	=head2 LINUX-SPECIFIC CALLS
		2583
		2584	When a call is documented as "linux-specific" then this means it
		2585	originated on GNU/Linux. C<IO::AIO> will usually try to autodetect the
		2586	availability and compatibility of such calls regardless of the platform
		2587	it is compiled on, so platforms such as FreeBSD which often implement
		2588	these calls will work. When in doubt, call them and see if they fail wth
		2589	C<ENOSYS>.
		2590
		2591	=head2 MEMORY USAGE
		2592
		2593	Per-request usage:
		2594
		2595	Each aio request uses - depending on your architecture - around 100-200
		2596	bytes of memory. In addition, stat requests need a stat buffer (possibly
		2597	a few hundred bytes), readdir requires a result buffer and so on. Perl
		2598	scalars and other data passed into aio requests will also be locked and
		2599	will consume memory till the request has entered the done state.
		2600
		2601	This is not awfully much, so queuing lots of requests is not usually a
		2602	problem.
		2603
		2604	Per-thread usage:
		2605
		2606	In the execution phase, some aio requests require more memory for
		2607	temporary buffers, and each thread requires a stack and other data
		2608	structures (usually around 16k-128k, depending on the OS).
		2609
		2610	=head1 KNOWN BUGS
		2611
		2612	Known bugs will be fixed in the next release :)
		2613
		2614	=head1 KNOWN ISSUES
		2615
		2616	Calls that try to "import" foreign memory areas (such as C<IO::AIO::mmap>
		2617	or C<IO::AIO::aio_slurp>) do not work with generic lvalues, such as
		2618	non-created hash slots or other scalars I didn't think of. It's best to
		2619	avoid such and either use scalar variables or making sure that the scalar
		2620	exists (e.g. by storing C<undef>) and isn't "funny" (e.g. tied).
		2621
		2622	I am not sure anything can be done about this, so this is considered a
		2623	known issue, rather than a bug.
404		2624
405	=head1 SEE ALSO	2625	=head1 SEE ALSO
406		2626
407	L<Coro>, L<Linux::AIO>.	2627	L<AnyEvent::AIO> for easy integration into event loops, L<Coro::AIO> for a
		2628	more natural syntax.
408		2629
409	=head1 AUTHOR	2630	=head1 AUTHOR
410		2631
411	Marc Lehmann <schmorp@schmorp.de>	2632	Marc Lehmann <schmorp@schmorp.de>
412	http://home.schmorp.de/	2633	http://home.schmorp.de/

Diff Legend

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