ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/IO-AIO/AIO.pm

Comparing IO-AIO/AIO.pm (file contents):
Revision 1.14 by root, Mon Jul 11 02:53:59 2005 UTC vs.
Revision 1.303 by root, Wed Apr 3 03:09:04 2019 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	\&IO::AIO::poll_cb;	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.	68	it is currently not reentrant in any way, so use appropriate locking
		69	yourself, always call C<poll_cb> from within the same thread, or never
		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
54		164
55	=cut	165	=cut
56		166
57	package IO::AIO;	167	package IO::AIO;
58		168
		169	use Carp ();
		170
		171	use common::sense;
		172
59	use base 'Exporter';	173	use base 'Exporter';
60		174
61	use Fcntl ();
62
63	BEGIN {	175	BEGIN {
64	$VERSION = 0.3;	176	our $VERSION = 4.72;
65		177
66	@EXPORT = qw(aio_read aio_write aio_open aio_close aio_stat aio_lstat aio_unlink	178	our @AIO_REQ = qw(aio_sendfile aio_seek aio_read aio_write aio_open aio_close
67	aio_fsync aio_fdatasync aio_readahead);	179	aio_stat aio_lstat aio_unlink aio_rmdir aio_readdir aio_readdirx
68	@EXPORT_OK = qw(poll_fileno poll_cb min_parallel max_parallel max_outstanding nreqs);	180	aio_scandir aio_symlink aio_readlink aio_realpath aio_fcntl aio_ioctl
		181	aio_sync aio_fsync aio_syncfs aio_fdatasync aio_sync_file_range
		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';
69		202
70	require XSLoader;	203	require XSLoader;
71	XSLoader::load IO::AIO, $VERSION;	204	XSLoader::load ("IO::AIO", $VERSION);
72	}	205	}
73		206
74	=head1 FUNCTIONS	207	=head1 FUNCTIONS
75		208
76	=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	IO::AIO::reinit
		284
		285	$nfd = IO::AIO::get_fdlimit [EXPERIMENTAL]
		286	IO::AIO::min_fdlimit $nfd [EXPERIMENTAL]
		287
		288	IO::AIO::sendfile $ofh, $ifh, $offset, $count
		289	IO::AIO::fadvise $fh, $offset, $len, $advice
		290
		291	IO::AIO::mmap $scalar, $length, $prot, $flags[, $fh[, $offset]]
		292	IO::AIO::munmap $scalar
		293	IO::AIO::mremap $scalar, $new_length, $flags[, $new_address]
		294	IO::AIO::madvise $scalar, $offset, $length, $advice
		295	IO::AIO::mprotect $scalar, $offset, $length, $protect
		296	IO::AIO::munlock $scalar, $offset = 0, $length = undef
		297	IO::AIO::munlockall
		298
		299	# stat extensions
		300	$counter = IO::AIO::st_gen
		301	$seconds = IO::AIO::st_atime, IO::AIO::st_mtime, IO::AIO::st_ctime, IO::AIO::st_btime
		302	($atime, $mtime, $ctime, $btime, ...) = IO::AIO::st_xtime
		303	$nanoseconds = IO::AIO::st_atimensec, IO::AIO::st_mtimensec, IO::AIO::st_ctimensec, IO::AIO::st_btimensec
		304	$seconds = IO::AIO::st_btimesec
		305	($atime, $mtime, $ctime, $btime, ...) = IO::AIO::st_xtimensec
		306
		307	# very much unportable syscalls
		308	IO::AIO::splice $r_fh, $r_off, $w_fh, $w_off, $length, $flags
		309	IO::AIO::tee $r_fh, $w_fh, $length, $flags
		310	$actual_size = IO::AIO::pipesize $r_fh[, $new_size]
		311	($rfh, $wfh) = IO::AIO::pipe2 [$flags]
		312	$fh = IO::AIO::memfd_create $pathname[, $flags]
		313	$fh = IO::AIO::eventfd [$initval, [$flags]]
		314	$fh = IO::AIO::timerfd_create $clockid[, $flags]
		315	($cur_interval, $cur_value) = IO::AIO::timerfd_settime $fh, $flags, $new_interval, $nbw_value
		316	($cur_interval, $cur_value) = IO::AIO::timerfd_gettime $fh
		317
		318	=head2 API NOTES
77		319
78	All the C<aio_*> calls are more or less thin wrappers around the syscall	320	All the C<aio_*> calls are more or less thin wrappers around the syscall
79	with the same name (sans C<aio_>). The arguments are similar or identical,	321	with the same name (sans C<aio_>). The arguments are similar or identical,
80	and they all accept an additional (and optional) C<$callback> argument	322	and they all accept an additional (and optional) C<$callback> argument
81	which must be a code reference. This code reference will get called with	323	which must be a code reference. This code reference will be called after
82	the syscall return code (e.g. most syscalls return C<-1> on error, unlike	324	the syscall has been executed in an asynchronous fashion. The results
83	perl, which usually delivers "false") as it's sole argument when the given	325	of the request will be passed as arguments to the callback (and, if an
84	syscall has been executed asynchronously.	326	error occured, in C<$!>) - for most requests the syscall return code (e.g.
		327	most syscalls return C<-1> on error, unlike perl, which usually delivers
		328	"false").
85		329
86	All functions that expect a filehandle will also accept a file descriptor.	330	Some requests (such as C<aio_readdir>) pass the actual results and
		331	communicate failures by passing C<undef>.
87		332
		333	All functions expecting a filehandle keep a copy of the filehandle
		334	internally until the request has finished.
		335
		336	All functions return request objects of type L<IO::AIO::REQ> that allow
		337	further manipulation of those requests while they are in-flight.
		338
88	The filenames you pass to these routines I<must> be absolute. The reason	339	The pathnames you pass to these routines I<should> be absolute. The
89	is that at the time the request is being executed, the current working	340	reason for this is that at the time the request is being executed, the
90	directory could have changed. Alternatively, you can make sure that you	341	current working directory could have changed. Alternatively, you can
91	never change the current working directory.	342	make sure that you never change the current working directory anywhere
		343	in the program and then use relative paths. You can also take advantage
		344	of IO::AIOs working directory abstraction, that lets you specify paths
		345	relative to some previously-opened "working directory object" - see the
		346	description of the C<IO::AIO::WD> class later in this document.
		347
		348	To encode pathnames as octets, either make sure you either: a) always pass
		349	in filenames you got from outside (command line, readdir etc.) without
		350	tinkering, b) are in your native filesystem encoding, c) use the Encode
		351	module and encode your pathnames to the locale (or other) encoding in
		352	effect in the user environment, d) use Glib::filename_from_unicode on
		353	unicode filenames or e) use something else to ensure your scalar has the
		354	correct contents.
		355
		356	This works, btw. independent of the internal UTF-8 bit, which IO::AIO
		357	handles correctly whether it is set or not.
		358
		359	=head2 AIO REQUEST FUNCTIONS
92		360
93	=over 4	361	=over 4
94		362
		363	=item $prev_pri = aioreq_pri [$pri]
		364
		365	Returns the priority value that would be used for the next request and, if
		366	C<$pri> is given, sets the priority for the next aio request.
		367
		368	The default priority is C<0>, the minimum and maximum priorities are C<-4>
		369	and C<4>, respectively. Requests with higher priority will be serviced
		370	first.
		371
		372	The priority will be reset to C<0> after each call to one of the C<aio_*>
		373	functions.
		374
		375	Example: open a file with low priority, then read something from it with
		376	higher priority so the read request is serviced before other low priority
		377	open requests (potentially spamming the cache):
		378
		379	aioreq_pri -3;
		380	aio_open ..., sub {
		381	return unless $_[0];
		382
		383	aioreq_pri -2;
		384	aio_read $_[0], ..., sub {
		385	...
		386	};
		387	};
		388
		389
		390	=item aioreq_nice $pri_adjust
		391
		392	Similar to C<aioreq_pri>, but subtracts the given value from the current
		393	priority, so the effect is cumulative.
		394
		395
95	=item aio_open $pathname, $flags, $mode, $callback	396	=item aio_open $pathname, $flags, $mode, $callback->($fh)
96		397
97	Asynchronously open or create a file and call the callback with a newly	398	Asynchronously open or create a file and call the callback with a newly
98	created filehandle for the file.	399	created filehandle for the file (or C<undef> in case of an error).
99		400
100	The pathname passed to C<aio_open> must be absolute. See API NOTES, above,	401	The pathname passed to C<aio_open> must be absolute. See API NOTES, above,
101	for an explanation.	402	for an explanation.
102		403
103	The C<$mode> argument is a bitmask. See the C<Fcntl> module for a	404	The C<$flags> argument is a bitmask. See the C<Fcntl> module for a
104	list. They are the same as used in C<sysopen>.	405	list. They are the same as used by C<sysopen>.
		406
		407	Likewise, C<$mode> specifies the mode of the newly created file, if it
		408	didn't exist and C<O_CREAT> has been given, just like perl's C<sysopen>,
		409	except that it is mandatory (i.e. use C<0> if you don't create new files,
		410	and C<0666> or C<0777> if you do). Note that the C<$mode> will be modified
		411	by the umask in effect then the request is being executed, so better never
		412	change the umask.
105		413
106	Example:	414	Example:
107		415
108	aio_open "/etc/passwd", O_RDONLY, 0, sub {	416	aio_open "/etc/passwd", IO::AIO::O_RDONLY, 0, sub {
109	if ($_[0]) {	417	if ($_[0]) {
110	print "open successful, fh is $_[0]\n";	418	print "open successful, fh is $_[0]\n";
111	...	419	...
112	} else {	420	} else {
113	die "open failed: $!\n";	421	die "open failed: $!\n";
114	}	422	}
115	};	423	};
116		424
		425	In addition to all the common open modes/flags (C<O_RDONLY>, C<O_WRONLY>,
		426	C<O_RDWR>, C<O_CREAT>, C<O_TRUNC>, C<O_EXCL> and C<O_APPEND>), the
		427	following POSIX and non-POSIX constants are available (missing ones on
		428	your system are, as usual, C<0>):
		429
		430	C<O_ASYNC>, C<O_DIRECT>, C<O_NOATIME>, C<O_CLOEXEC>, C<O_NOCTTY>, C<O_NOFOLLOW>,
		431	C<O_NONBLOCK>, C<O_EXEC>, C<O_SEARCH>, C<O_DIRECTORY>, C<O_DSYNC>,
		432	C<O_RSYNC>, C<O_SYNC>, C<O_PATH>, C<O_TMPFILE>, C<O_TTY_INIT> and C<O_ACCMODE>.
		433
		434
117	=item aio_close $fh, $callback	435	=item aio_close $fh, $callback->($status)
118		436
119	Asynchronously close a file and call the callback with the result	437	Asynchronously close a file and call the callback with the result
120	code. I<WARNING:> although accepted, you should not pass in a perl	438	code.
121	filehandle here, as perl will likely close the file descriptor itself when
122	the filehandle is destroyed. Normally, you can safely call perls C<close>
123	or just let filehandles go out of scope.
124		439
		440	Unfortunately, you can't do this to perl. Perl I<insists> very strongly on
		441	closing the file descriptor associated with the filehandle itself.
		442
		443	Therefore, C<aio_close> will not close the filehandle - instead it will
		444	use dup2 to overwrite the file descriptor with the write-end of a pipe
		445	(the pipe fd will be created on demand and will be cached).
		446
		447	Or in other words: the file descriptor will be closed, but it will not be
		448	free for reuse until the perl filehandle is closed.
		449
		450	=cut
		451
		452	=item aio_seek $fh, $offset, $whence, $callback->($offs)
		453
		454	Seeks the filehandle to the new C<$offset>, similarly to perl's
		455	C<sysseek>. The C<$whence> can use the traditional values (C<0> for
		456	C<IO::AIO::SEEK_SET>, C<1> for C<IO::AIO::SEEK_CUR> or C<2> for
		457	C<IO::AIO::SEEK_END>).
		458
		459	The resulting absolute offset will be passed to the callback, or C<-1> in
		460	case of an error.
		461
		462	In theory, the C<$whence> constants could be different than the
		463	corresponding values from L<Fcntl>, but perl guarantees they are the same,
		464	so don't panic.
		465
		466	As a GNU/Linux (and maybe Solaris) extension, also the constants
		467	C<IO::AIO::SEEK_DATA> and C<IO::AIO::SEEK_HOLE> are available, if they
		468	could be found. No guarantees about suitability for use in C<aio_seek> or
		469	Perl's C<sysseek> can be made though, although I would naively assume they
		470	"just work".
		471
125	=item aio_read $fh,$offset,$length, $data,$dataoffset,$callback	472	=item aio_read $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
126		473
127	=item aio_write $fh,$offset,$length, $data,$dataoffset,$callback	474	=item aio_write $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
128		475
129	Reads or writes C<length> bytes from the specified C<fh> and C<offset>	476	Reads or writes C<$length> bytes from or to the specified C<$fh> and
130	into the scalar given by C<data> and offset C<dataoffset> and calls the	477	C<$offset> into the scalar given by C<$data> and offset C<$dataoffset> and
131	callback without the actual number of bytes read (or -1 on error, just	478	calls the callback with the actual number of bytes transferred (or -1 on
132	like the syscall).	479	error, just like the syscall).
133		480
		481	C<aio_read> will, like C<sysread>, shrink or grow the C<$data> scalar to
		482	offset plus the actual number of bytes read.
		483
		484	If C<$offset> is undefined, then the current file descriptor offset will
		485	be used (and updated), otherwise the file descriptor offset will not be
		486	changed by these calls.
		487
		488	If C<$length> is undefined in C<aio_write>, use the remaining length of
		489	C<$data>.
		490
		491	If C<$dataoffset> is less than zero, it will be counted from the end of
		492	C<$data>.
		493
		494	The C<$data> scalar I<MUST NOT> be modified in any way while the request
		495	is outstanding. Modifying it can result in segfaults or World War III (if
		496	the necessary/optional hardware is installed).
		497
134	Example: Read 15 bytes at offset 7 into scalar C<$buffer>, strating at	498	Example: Read 15 bytes at offset 7 into scalar C<$buffer>, starting at
135	offset C<0> within the scalar:	499	offset C<0> within the scalar:
136		500
137	aio_read $fh, 7, 15, $buffer, 0, sub {	501	aio_read $fh, 7, 15, $buffer, 0, sub {
138	$_[0] > 0 or die "read error: $!";	502	$_[0] > 0 or die "read error: $!";
139	print "read $_[0] bytes: <$buffer>\n";	503	print "read $_[0] bytes: <$buffer>\n";
140	};	504	};
141		505
		506
		507	=item aio_sendfile $out_fh, $in_fh, $in_offset, $length, $callback->($retval)
		508
		509	Tries to copy C<$length> bytes from C<$in_fh> to C<$out_fh>. It starts
		510	reading at byte offset C<$in_offset>, and starts writing at the current
		511	file offset of C<$out_fh>. Because of that, it is not safe to issue more
		512	than one C<aio_sendfile> per C<$out_fh>, as they will interfere with each
		513	other. The same C<$in_fh> works fine though, as this function does not
		514	move or use the file offset of C<$in_fh>.
		515
		516	Please note that C<aio_sendfile> can read more bytes from C<$in_fh> than
		517	are written, and there is no way to find out how many more bytes have been
		518	read from C<aio_sendfile> alone, as C<aio_sendfile> only provides the
		519	number of bytes written to C<$out_fh>. Only if the result value equals
		520	C<$length> one can assume that C<$length> bytes have been read.
		521
		522	Unlike with other C<aio_> functions, it makes a lot of sense to use
		523	C<aio_sendfile> on non-blocking sockets, as long as one end (typically
		524	the C<$in_fh>) is a file - the file I/O will then be asynchronous, while
		525	the socket I/O will be non-blocking. Note, however, that you can run
		526	into a trap where C<aio_sendfile> reads some data with readahead, then
		527	fails to write all data, and when the socket is ready the next time, the
		528	data in the cache is already lost, forcing C<aio_sendfile> to again hit
		529	the disk. Explicit C<aio_read> + C<aio_write> let's you better control
		530	resource usage.
		531
		532	This call tries to make use of a native C<sendfile>-like syscall to
		533	provide zero-copy operation. For this to work, C<$out_fh> should refer to
		534	a socket, and C<$in_fh> should refer to an mmap'able file.
		535
		536	If a native sendfile cannot be found or it fails with C<ENOSYS>,
		537	C<EINVAL>, C<ENOTSUP>, C<EOPNOTSUPP>, C<EAFNOSUPPORT>, C<EPROTOTYPE> or
		538	C<ENOTSOCK>, it will be emulated, so you can call C<aio_sendfile> on any
		539	type of filehandle regardless of the limitations of the operating system.
		540
		541	As native sendfile syscalls (as practically any non-POSIX interface hacked
		542	together in a hurry to improve benchmark numbers) tend to be rather buggy
		543	on many systems, this implementation tries to work around some known bugs
		544	in Linux and FreeBSD kernels (probably others, too), but that might fail,
		545	so you really really should check the return value of C<aio_sendfile> -
		546	fewer bytes than expected might have been transferred.
		547
		548
142	=item aio_readahead $fh,$offset,$length, $callback	549	=item aio_readahead $fh,$offset,$length, $callback->($retval)
143		550
144	Asynchronously reads the specified byte range into the page cache, using
145	the C<readahead> syscall. If that syscall doesn't exist the status will be
146	C<-1> and C<$!> is set to ENOSYS.
147
148	readahead() populates the page cache with data from a file so that	551	C<aio_readahead> populates the page cache with data from a file so that
149	subsequent reads from that file will not block on disk I/O. The C<$offset>	552	subsequent reads from that file will not block on disk I/O. The C<$offset>
150	argument specifies the starting point from which data is to be read and	553	argument specifies the starting point from which data is to be read and
151	C<$length> specifies the number of bytes to be read. I/O is performed in	554	C<$length> specifies the number of bytes to be read. I/O is performed in
152	whole pages, so that offset is effectively rounded down to a page boundary	555	whole pages, so that offset is effectively rounded down to a page boundary
153	and bytes are read up to the next page boundary greater than or equal to	556	and bytes are read up to the next page boundary greater than or equal to
154	(off-set+length). aio_readahead() does not read beyond the end of the	557	(off-set+length). C<aio_readahead> does not read beyond the end of the
155	file. The current file offset of the file is left unchanged.	558	file. The current file offset of the file is left unchanged.
156		559
		560	If that syscall doesn't exist (likely if your kernel isn't Linux) it will
		561	be emulated by simply reading the data, which would have a similar effect.
		562
		563
157	=item aio_stat $fh_or_path, $callback	564	=item aio_stat $fh_or_path, $callback->($status)
158		565
159	=item aio_lstat $fh, $callback	566	=item aio_lstat $fh, $callback->($status)
160		567
161	Works like perl's C<stat> or C<lstat> in void context. The callback will	568	Works almost exactly like perl's C<stat> or C<lstat> in void context. The
162	be called after the stat and the results will be available using C<stat _>	569	callback will be called after the stat and the results will be available
163	or C<-s _> etc...	570	using C<stat _> or C<-s _> and other tests (with the exception of C<-B>
		571	and C<-T>).
164		572
165	The pathname passed to C<aio_stat> must be absolute. See API NOTES, above,	573	The pathname passed to C<aio_stat> must be absolute. See API NOTES, above,
166	for an explanation.	574	for an explanation.
167		575
168	Currently, the stats are always 64-bit-stats, i.e. instead of returning an	576	Currently, the stats are always 64-bit-stats, i.e. instead of returning an
169	error when stat'ing a large file, the results will be silently truncated	577	error when stat'ing a large file, the results will be silently truncated
170	unless perl itself is compiled with large file support.	578	unless perl itself is compiled with large file support.
		579
		580	To help interpret the mode and dev/rdev stat values, IO::AIO offers the
		581	following constants and functions (if not implemented, the constants will
		582	be C<0> and the functions will either C<croak> or fall back on traditional
		583	behaviour).
		584
		585	C<S_IFMT>, C<S_IFIFO>, C<S_IFCHR>, C<S_IFBLK>, C<S_IFLNK>, C<S_IFREG>,
		586	C<S_IFDIR>, C<S_IFWHT>, C<S_IFSOCK>, C<IO::AIO::major $dev_t>,
		587	C<IO::AIO::minor $dev_t>, C<IO::AIO::makedev $major, $minor>.
		588
		589	To access higher resolution stat timestamps, see L<SUBSECOND STAT TIME
		590	ACCESS>.
171		591
172	Example: Print the length of F</etc/passwd>:	592	Example: Print the length of F</etc/passwd>:
173		593
174	aio_stat "/etc/passwd", sub {	594	aio_stat "/etc/passwd", sub {
175	$_[0] and die "stat failed: $!";	595	$_[0] and die "stat failed: $!";
176	print "size is ", -s _, "\n";	596	print "size is ", -s _, "\n";
177	};	597	};
178		598
		599
		600	=item aio_statvfs $fh_or_path, $callback->($statvfs)
		601
		602	Works like the POSIX C<statvfs> or C<fstatvfs> syscalls, depending on
		603	whether a file handle or path was passed.
		604
		605	On success, the callback is passed a hash reference with the following
		606	members: C<bsize>, C<frsize>, C<blocks>, C<bfree>, C<bavail>, C<files>,
		607	C<ffree>, C<favail>, C<fsid>, C<flag> and C<namemax>. On failure, C<undef>
		608	is passed.
		609
		610	The following POSIX IO::AIO::ST_* constants are defined: C<ST_RDONLY> and
		611	C<ST_NOSUID>.
		612
		613	The following non-POSIX IO::AIO::ST_* flag masks are defined to
		614	their correct value when available, or to C<0> on systems that do
		615	not support them: C<ST_NODEV>, C<ST_NOEXEC>, C<ST_SYNCHRONOUS>,
		616	C<ST_MANDLOCK>, C<ST_WRITE>, C<ST_APPEND>, C<ST_IMMUTABLE>, C<ST_NOATIME>,
		617	C<ST_NODIRATIME> and C<ST_RELATIME>.
		618
		619	Example: stat C</wd> and dump out the data if successful.
		620
		621	aio_statvfs "/wd", sub {
		622	my $f = $_[0]
		623	or die "statvfs: $!";
		624
		625	use Data::Dumper;
		626	say Dumper $f;
		627	};
		628
		629	# result:
		630	{
		631	bsize => 1024,
		632	bfree => 4333064312,
		633	blocks => 10253828096,
		634	files => 2050765568,
		635	flag => 4096,
		636	favail => 2042092649,
		637	bavail => 4333064312,
		638	ffree => 2042092649,
		639	namemax => 255,
		640	frsize => 1024,
		641	fsid => 1810
		642	}
		643
		644	=item aio_utime $fh_or_path, $atime, $mtime, $callback->($status)
		645
		646	Works like perl's C<utime> function (including the special case of $atime
		647	and $mtime being undef). Fractional times are supported if the underlying
		648	syscalls support them.
		649
		650	When called with a pathname, uses utimensat(2) or utimes(2) if available,
		651	otherwise utime(2). If called on a file descriptor, uses futimens(2)
		652	or futimes(2) if available, otherwise returns ENOSYS, so this is not
		653	portable.
		654
		655	Examples:
		656
		657	# set atime and mtime to current time (basically touch(1)):
		658	aio_utime "path", undef, undef;
		659	# set atime to current time and mtime to beginning of the epoch:
		660	aio_utime "path", time, undef; # undef==0
		661
		662
		663	=item aio_chown $fh_or_path, $uid, $gid, $callback->($status)
		664
		665	Works like perl's C<chown> function, except that C<undef> for either $uid
		666	or $gid is being interpreted as "do not change" (but -1 can also be used).
		667
		668	Examples:
		669
		670	# same as "chown root path" in the shell:
		671	aio_chown "path", 0, -1;
		672	# same as above:
		673	aio_chown "path", 0, undef;
		674
		675
		676	=item aio_truncate $fh_or_path, $offset, $callback->($status)
		677
		678	Works like truncate(2) or ftruncate(2).
		679
		680
		681	=item aio_allocate $fh, $mode, $offset, $len, $callback->($status)
		682
		683	Allocates or frees disk space according to the C<$mode> argument. See the
		684	linux C<fallocate> documentation for details.
		685
		686	C<$mode> is usually C<0> or C<IO::AIO::FALLOC_FL_KEEP_SIZE> to allocate
		687	space, or C<IO::AIO::FALLOC_FL_PUNCH_HOLE | IO::AIO::FALLOC_FL_KEEP_SIZE>,
		688	to deallocate a file range.
		689
		690	IO::AIO also supports C<FALLOC_FL_COLLAPSE_RANGE>, to remove a range
		691	(without leaving a hole), C<FALLOC_FL_ZERO_RANGE>, to zero a range,
		692	C<FALLOC_FL_INSERT_RANGE> to insert a range and C<FALLOC_FL_UNSHARE_RANGE>
		693	to unshare shared blocks (see your L<fallocate(2)> manpage).
		694
		695	The file system block size used by C<fallocate> is presumably the
		696	C<f_bsize> returned by C<statvfs>, but different filesystems and filetypes
		697	can dictate other limitations.
		698
		699	If C<fallocate> isn't available or cannot be emulated (currently no
		700	emulation will be attempted), passes C<-1> and sets C<$!> to C<ENOSYS>.
		701
		702
		703	=item aio_chmod $fh_or_path, $mode, $callback->($status)
		704
		705	Works like perl's C<chmod> function.
		706
		707
179	=item aio_unlink $pathname, $callback	708	=item aio_unlink $pathname, $callback->($status)
180		709
181	Asynchronously unlink (delete) a file and call the callback with the	710	Asynchronously unlink (delete) a file and call the callback with the
182	result code.	711	result code.
183		712
		713
		714	=item aio_mknod $pathname, $mode, $dev, $callback->($status)
		715
		716	[EXPERIMENTAL]
		717
		718	Asynchronously create a device node (or fifo). See mknod(2).
		719
		720	The only (POSIX-) portable way of calling this function is:
		721
		722	aio_mknod $pathname, IO::AIO::S_IFIFO | $mode, 0, sub { ...
		723
		724	See C<aio_stat> for info about some potentially helpful extra constants
		725	and functions.
		726
		727	=item aio_link $srcpath, $dstpath, $callback->($status)
		728
		729	Asynchronously create a new link to the existing object at C<$srcpath> at
		730	the path C<$dstpath> and call the callback with the result code.
		731
		732
		733	=item aio_symlink $srcpath, $dstpath, $callback->($status)
		734
		735	Asynchronously create a new symbolic link to the existing object at C<$srcpath> at
		736	the path C<$dstpath> and call the callback with the result code.
		737
		738
		739	=item aio_readlink $pathname, $callback->($link)
		740
		741	Asynchronously read the symlink specified by C<$path> and pass it to
		742	the callback. If an error occurs, nothing or undef gets passed to the
		743	callback.
		744
		745
		746	=item aio_realpath $pathname, $callback->($path)
		747
		748	Asynchronously make the path absolute and resolve any symlinks in
		749	C<$path>. The resulting path only consists of directories (same as
		750	L<Cwd::realpath>).
		751
		752	This request can be used to get the absolute path of the current working
		753	directory by passing it a path of F<.> (a single dot).
		754
		755
		756	=item aio_rename $srcpath, $dstpath, $callback->($status)
		757
		758	Asynchronously rename the object at C<$srcpath> to C<$dstpath>, just as
		759	rename(2) and call the callback with the result code.
		760
		761	On systems that support the AIO::WD working directory abstraction
		762	natively, the case C<[$wd, "."]> as C<$srcpath> is specialcased - instead
		763	of failing, C<rename> is called on the absolute path of C<$wd>.
		764
		765
		766	=item aio_rename2 $srcpath, $dstpath, $flags, $callback->($status)
		767
		768	Basically a version of C<aio_rename> with an additional C<$flags>
		769	argument. Calling this with C<$flags=0> is the same as calling
		770	C<aio_rename>.
		771
		772	Non-zero flags are currently only supported on GNU/Linux systems that
		773	support renameat2. Other systems fail with C<ENOSYS> in this case.
		774
		775	The following constants are available (missing ones are, as usual C<0>),
		776	see renameat2(2) for details:
		777
		778	C<IO::AIO::RENAME_NOREPLACE>, C<IO::AIO::RENAME_EXCHANGE>
		779	and C<IO::AIO::RENAME_WHITEOUT>.
		780
		781
		782	=item aio_mkdir $pathname, $mode, $callback->($status)
		783
		784	Asynchronously mkdir (create) a directory and call the callback with
		785	the result code. C<$mode> will be modified by the umask at the time the
		786	request is executed, so do not change your umask.
		787
		788
		789	=item aio_rmdir $pathname, $callback->($status)
		790
		791	Asynchronously rmdir (delete) a directory and call the callback with the
		792	result code.
		793
		794	On systems that support the AIO::WD working directory abstraction
		795	natively, the case C<[$wd, "."]> is specialcased - instead of failing,
		796	C<rmdir> is called on the absolute path of C<$wd>.
		797
		798
		799	=item aio_readdir $pathname, $callback->($entries)
		800
		801	Unlike the POSIX call of the same name, C<aio_readdir> reads an entire
		802	directory (i.e. opendir + readdir + closedir). The entries will not be
		803	sorted, and will B<NOT> include the C<.> and C<..> entries.
		804
		805	The callback is passed a single argument which is either C<undef> or an
		806	array-ref with the filenames.
		807
		808
		809	=item aio_readdirx $pathname, $flags, $callback->($entries, $flags)
		810
		811	Quite similar to C<aio_readdir>, but the C<$flags> argument allows one to
		812	tune behaviour and output format. In case of an error, C<$entries> will be
		813	C<undef>.
		814
		815	The flags are a combination of the following constants, ORed together (the
		816	flags will also be passed to the callback, possibly modified):
		817
		818	=over 4
		819
		820	=item IO::AIO::READDIR_DENTS
		821
		822	Normally the callback gets an arrayref consisting of names only (as
		823	with C<aio_readdir>). If this flag is set, then the callback gets an
		824	arrayref with C<[$name, $type, $inode]> arrayrefs, each describing a
		825	single directory entry in more detail:
		826
		827	C<$name> is the name of the entry.
		828
		829	C<$type> is one of the C<IO::AIO::DT_xxx> constants:
		830
		831	C<IO::AIO::DT_UNKNOWN>, C<IO::AIO::DT_FIFO>, C<IO::AIO::DT_CHR>, C<IO::AIO::DT_DIR>,
		832	C<IO::AIO::DT_BLK>, C<IO::AIO::DT_REG>, C<IO::AIO::DT_LNK>, C<IO::AIO::DT_SOCK>,
		833	C<IO::AIO::DT_WHT>.
		834
		835	C<IO::AIO::DT_UNKNOWN> means just that: readdir does not know. If you need
		836	to know, you have to run stat yourself. Also, for speed/memory reasons,
		837	the C<$type> scalars are read-only: you must not modify them.
		838
		839	C<$inode> is the inode number (which might not be exact on systems with 64
		840	bit inode numbers and 32 bit perls). This field has unspecified content on
		841	systems that do not deliver the inode information.
		842
		843	=item IO::AIO::READDIR_DIRS_FIRST
		844
		845	When this flag is set, then the names will be returned in an order where
		846	likely directories come first, in optimal stat order. This is useful when
		847	you need to quickly find directories, or you want to find all directories
		848	while avoiding to stat() each entry.
		849
		850	If the system returns type information in readdir, then this is used
		851	to find directories directly. Otherwise, likely directories are names
		852	beginning with ".", or otherwise names with no dots, of which names with
		853	short names are tried first.
		854
		855	=item IO::AIO::READDIR_STAT_ORDER
		856
		857	When this flag is set, then the names will be returned in an order
		858	suitable for stat()'ing each one. That is, when you plan to stat() most or
		859	all files in the given directory, then the returned order will likely be
		860	faster.
		861
		862	If both this flag and C<IO::AIO::READDIR_DIRS_FIRST> are specified,
		863	then the likely dirs come first, resulting in a less optimal stat order
		864	for stat'ing all entries, but likely a more optimal order for finding
		865	subdirectories.
		866
		867	=item IO::AIO::READDIR_FOUND_UNKNOWN
		868
		869	This flag should not be set when calling C<aio_readdirx>. Instead, it
		870	is being set by C<aio_readdirx>, when any of the C<$type>'s found were
		871	C<IO::AIO::DT_UNKNOWN>. The absence of this flag therefore indicates that all
		872	C<$type>'s are known, which can be used to speed up some algorithms.
		873
		874	=back
		875
		876
		877	=item aio_slurp $pathname, $offset, $length, $data, $callback->($status)
		878
		879	Opens, reads and closes the given file. The data is put into C<$data>,
		880	which is resized as required.
		881
		882	If C<$offset> is negative, then it is counted from the end of the file.
		883
		884	If C<$length> is zero, then the remaining length of the file is
		885	used. Also, in this case, the same limitations to modifying C<$data> apply
		886	as when IO::AIO::mmap is used, i.e. it must only be modified in-place
		887	with C<substr>. If the size of the file is known, specifying a non-zero
		888	C<$length> results in a performance advantage.
		889
		890	This request is similar to the older C<aio_load> request, but since it is
		891	a single request, it might be more efficient to use.
		892
		893	Example: load F</etc/passwd> into C<$passwd>.
		894
		895	my $passwd;
		896	aio_slurp "/etc/passwd", 0, 0, $passwd, sub {
		897	$_[0] >= 0
		898	or die "/etc/passwd: $!\n";
		899
		900	printf "/etc/passwd is %d bytes long, and contains:\n", length $passwd;
		901	print $passwd;
		902	};
		903	IO::AIO::flush;
		904
		905
		906	=item aio_load $pathname, $data, $callback->($status)
		907
		908	This is a composite request that tries to fully load the given file into
		909	memory. Status is the same as with aio_read.
		910
		911	Using C<aio_slurp> might be more efficient, as it is a single request.
		912
		913	=cut
		914
		915	sub aio_load($$;$) {
		916	my ($path, undef, $cb) = @_;
		917	my $data = \$_[1];
		918
		919	my $pri = aioreq_pri;
		920	my $grp = aio_group $cb;
		921
		922	aioreq_pri $pri;
		923	add $grp aio_open $path, O_RDONLY, 0, sub {
		924	my $fh = shift
		925	or return $grp->result (-1);
		926
		927	aioreq_pri $pri;
		928	add $grp aio_read $fh, 0, (-s $fh), $$data, 0, sub {
		929	$grp->result ($_[0]);
		930	};
		931	};
		932
		933	$grp
		934	}
		935
		936	=item aio_copy $srcpath, $dstpath, $callback->($status)
		937
		938	Try to copy the I<file> (directories not supported as either source or
		939	destination) from C<$srcpath> to C<$dstpath> and call the callback with
		940	a status of C<0> (ok) or C<-1> (error, see C<$!>).
		941
		942	Existing destination files will be truncated.
		943
		944	This is a composite request that creates the destination file with
		945	mode 0200 and copies the contents of the source file into it using
		946	C<aio_sendfile>, followed by restoring atime, mtime, access mode and
		947	uid/gid, in that order.
		948
		949	If an error occurs, the partial destination file will be unlinked, if
		950	possible, except when setting atime, mtime, access mode and uid/gid, where
		951	errors are being ignored.
		952
		953	=cut
		954
		955	sub aio_copy($$;$) {
		956	my ($src, $dst, $cb) = @_;
		957
		958	my $pri = aioreq_pri;
		959	my $grp = aio_group $cb;
		960
		961	aioreq_pri $pri;
		962	add $grp aio_open $src, O_RDONLY, 0, sub {
		963	if (my $src_fh = $_[0]) {
		964	my @stat = stat $src_fh; # hmm, might block over nfs?
		965
		966	aioreq_pri $pri;
		967	add $grp aio_open $dst, O_CREAT | O_WRONLY | O_TRUNC, 0200, sub {
		968	if (my $dst_fh = $_[0]) {
		969	aioreq_pri $pri;
		970	add $grp aio_sendfile $dst_fh, $src_fh, 0, $stat[7], sub {
		971	if ($_[0] == $stat[7]) {
		972	$grp->result (0);
		973	close $src_fh;
		974
		975	my $ch = sub {
		976	aioreq_pri $pri;
		977	add $grp aio_chmod $dst_fh, $stat[2] & 07777, sub {
		978	aioreq_pri $pri;
		979	add $grp aio_chown $dst_fh, $stat[4], $stat[5], sub {
		980	aioreq_pri $pri;
		981	add $grp aio_close $dst_fh;
		982	}
		983	};
		984	};
		985
		986	aioreq_pri $pri;
		987	add $grp aio_utime $dst_fh, $stat[8], $stat[9], sub {
		988	if ($_[0] < 0 && $! == ENOSYS) {
		989	aioreq_pri $pri;
		990	add $grp aio_utime $dst, $stat[8], $stat[9], $ch;
		991	} else {
		992	$ch->();
		993	}
		994	};
		995	} else {
		996	$grp->result (-1);
		997	close $src_fh;
		998	close $dst_fh;
		999
		1000	aioreq $pri;
		1001	add $grp aio_unlink $dst;
		1002	}
		1003	};
		1004	} else {
		1005	$grp->result (-1);
		1006	}
		1007	},
		1008
		1009	} else {
		1010	$grp->result (-1);
		1011	}
		1012	};
		1013
		1014	$grp
		1015	}
		1016
		1017	=item aio_move $srcpath, $dstpath, $callback->($status)
		1018
		1019	Try to move the I<file> (directories not supported as either source or
		1020	destination) from C<$srcpath> to C<$dstpath> and call the callback with
		1021	a status of C<0> (ok) or C<-1> (error, see C<$!>).
		1022
		1023	This is a composite request that tries to rename(2) the file first; if
		1024	rename fails with C<EXDEV>, it copies the file with C<aio_copy> and, if
		1025	that is successful, unlinks the C<$srcpath>.
		1026
		1027	=cut
		1028
		1029	sub aio_move($$;$) {
		1030	my ($src, $dst, $cb) = @_;
		1031
		1032	my $pri = aioreq_pri;
		1033	my $grp = aio_group $cb;
		1034
		1035	aioreq_pri $pri;
		1036	add $grp aio_rename $src, $dst, sub {
		1037	if ($_[0] && $! == EXDEV) {
		1038	aioreq_pri $pri;
		1039	add $grp aio_copy $src, $dst, sub {
		1040	$grp->result ($_[0]);
		1041
		1042	unless ($_[0]) {
		1043	aioreq_pri $pri;
		1044	add $grp aio_unlink $src;
		1045	}
		1046	};
		1047	} else {
		1048	$grp->result ($_[0]);
		1049	}
		1050	};
		1051
		1052	$grp
		1053	}
		1054
		1055	=item aio_scandir $pathname, $maxreq, $callback->($dirs, $nondirs)
		1056
		1057	Scans a directory (similar to C<aio_readdir>) but additionally tries to
		1058	efficiently separate the entries of directory C<$path> into two sets of
		1059	names, directories you can recurse into (directories), and ones you cannot
		1060	recurse into (everything else, including symlinks to directories).
		1061
		1062	C<aio_scandir> is a composite request that generates many sub requests.
		1063	C<$maxreq> specifies the maximum number of outstanding aio requests that
		1064	this function generates. If it is C<< <= 0 >>, then a suitable default
		1065	will be chosen (currently 4).
		1066
		1067	On error, the callback is called without arguments, otherwise it receives
		1068	two array-refs with path-relative entry names.
		1069
		1070	Example:
		1071
		1072	aio_scandir $dir, 0, sub {
		1073	my ($dirs, $nondirs) = @_;
		1074	print "real directories: @$dirs\n";
		1075	print "everything else: @$nondirs\n";
		1076	};
		1077
		1078	Implementation notes.
		1079
		1080	The C<aio_readdir> cannot be avoided, but C<stat()>'ing every entry can.
		1081
		1082	If readdir returns file type information, then this is used directly to
		1083	find directories.
		1084
		1085	Otherwise, after reading the directory, the modification time, size etc.
		1086	of the directory before and after the readdir is checked, and if they
		1087	match (and isn't the current time), the link count will be used to decide
		1088	how many entries are directories (if >= 2). Otherwise, no knowledge of the
		1089	number of subdirectories will be assumed.
		1090
		1091	Then entries will be sorted into likely directories a non-initial dot
		1092	currently) and likely non-directories (see C<aio_readdirx>). Then every
		1093	entry plus an appended C</.> will be C<stat>'ed, likely directories first,
		1094	in order of their inode numbers. If that succeeds, it assumes that the
		1095	entry is a directory or a symlink to directory (which will be checked
		1096	separately). This is often faster than stat'ing the entry itself because
		1097	filesystems might detect the type of the entry without reading the inode
		1098	data (e.g. ext2fs filetype feature), even on systems that cannot return
		1099	the filetype information on readdir.
		1100
		1101	If the known number of directories (link count - 2) has been reached, the
		1102	rest of the entries is assumed to be non-directories.
		1103
		1104	This only works with certainty on POSIX (= UNIX) filesystems, which
		1105	fortunately are the vast majority of filesystems around.
		1106
		1107	It will also likely work on non-POSIX filesystems with reduced efficiency
		1108	as those tend to return 0 or 1 as link counts, which disables the
		1109	directory counting heuristic.
		1110
		1111	=cut
		1112
		1113	sub aio_scandir($$;$) {
		1114	my ($path, $maxreq, $cb) = @_;
		1115
		1116	my $pri = aioreq_pri;
		1117
		1118	my $grp = aio_group $cb;
		1119
		1120	$maxreq = 4 if $maxreq <= 0;
		1121
		1122	# get a wd object
		1123	aioreq_pri $pri;
		1124	add $grp aio_wd $path, sub {
		1125	$_[0]
		1126	or return $grp->result ();
		1127
		1128	my $wd = [shift, "."];
		1129
		1130	# stat once
		1131	aioreq_pri $pri;
		1132	add $grp aio_stat $wd, sub {
		1133	return $grp->result () if $_[0];
		1134	my $now = time;
		1135	my $hash1 = join ":", (stat _)[0,1,3,7,9];
		1136	my $rdxflags = READDIR_DIRS_FIRST;
		1137
		1138	if ((stat _)[3] < 2) {
		1139	# at least one non-POSIX filesystem exists
		1140	# that returns useful DT_type values: btrfs,
		1141	# so optimise for this here by requesting dents
		1142	$rdxflags |= READDIR_DENTS;
		1143	}
		1144
		1145	# read the directory entries
		1146	aioreq_pri $pri;
		1147	add $grp aio_readdirx $wd, $rdxflags, sub {
		1148	my ($entries, $flags) = @_
		1149	or return $grp->result ();
		1150
		1151	if ($rdxflags & READDIR_DENTS) {
		1152	# if we requested type values, see if we can use them directly.
		1153
		1154	# if there were any DT_UNKNOWN entries then we assume we
		1155	# don't know. alternatively, we could assume that if we get
		1156	# one DT_DIR, then all directories are indeed marked with
		1157	# DT_DIR, but this seems not required for btrfs, and this
		1158	# is basically the "btrfs can't get it's act together" code
		1159	# branch.
		1160	unless ($flags & READDIR_FOUND_UNKNOWN) {
		1161	# now we have valid DT_ information for all entries,
		1162	# so use it as an optimisation without further stat's.
		1163	# they must also all be at the beginning of @$entries
		1164	# by now.
		1165
		1166	my $dirs;
		1167
		1168	if (@$entries) {
		1169	for (0 .. $#$entries) {
		1170	if ($entries->[$_][1] != DT_DIR) {
		1171	# splice out directories
		1172	$dirs = [splice @$entries, 0, $_];
		1173	last;
		1174	}
		1175	}
		1176
		1177	# if we didn't find any non-dir, then all entries are dirs
		1178	unless ($dirs) {
		1179	($dirs, $entries) = ($entries, []);
		1180	}
		1181	} else {
		1182	# directory is empty, so there are no sbdirs
		1183	$dirs = [];
		1184	}
		1185
		1186	# either splice'd the directories out or the dir was empty.
		1187	# convert dents to filenames
		1188	$_ = $_->[0] for @$dirs;
		1189	$_ = $_->[0] for @$entries;
		1190
		1191	return $grp->result ($dirs, $entries);
		1192	}
		1193
		1194	# cannot use, so return to our old ways
		1195	# by pretending we only scanned for names.
		1196	$_ = $_->[0] for @$entries;
		1197	}
		1198
		1199	# stat the dir another time
		1200	aioreq_pri $pri;
		1201	add $grp aio_stat $wd, sub {
		1202	my $hash2 = join ":", (stat _)[0,1,3,7,9];
		1203
		1204	my $ndirs;
		1205
		1206	# take the slow route if anything looks fishy
		1207	if ($hash1 ne $hash2 or (stat _)[9] == $now) {
		1208	$ndirs = -1;
		1209	} else {
		1210	# if nlink == 2, we are finished
		1211	# for non-posix-fs's, we rely on nlink < 2
		1212	$ndirs = (stat _)[3] - 2
		1213	or return $grp->result ([], $entries);
		1214	}
		1215
		1216	my (@dirs, @nondirs);
		1217
		1218	my $statgrp = add $grp aio_group sub {
		1219	$grp->result (\@dirs, \@nondirs);
		1220	};
		1221
		1222	limit $statgrp $maxreq;
		1223	feed $statgrp sub {
		1224	return unless @$entries;
		1225	my $entry = shift @$entries;
		1226
		1227	aioreq_pri $pri;
		1228	$wd->[1] = "$entry/.";
		1229	add $statgrp aio_stat $wd, sub {
		1230	if ($_[0] < 0) {
		1231	push @nondirs, $entry;
		1232	} else {
		1233	# need to check for real directory
		1234	aioreq_pri $pri;
		1235	$wd->[1] = $entry;
		1236	add $statgrp aio_lstat $wd, sub {
		1237	if (-d _) {
		1238	push @dirs, $entry;
		1239
		1240	unless (--$ndirs) {
		1241	push @nondirs, @$entries;
		1242	feed $statgrp;
		1243	}
		1244	} else {
		1245	push @nondirs, $entry;
		1246	}
		1247	}
		1248	}
		1249	};
		1250	};
		1251	};
		1252	};
		1253	};
		1254	};
		1255
		1256	$grp
		1257	}
		1258
		1259	=item aio_rmtree $pathname, $callback->($status)
		1260
		1261	Delete a directory tree starting (and including) C<$path>, return the
		1262	status of the final C<rmdir> only. This is a composite request that
		1263	uses C<aio_scandir> to recurse into and rmdir directories, and unlink
		1264	everything else.
		1265
		1266	=cut
		1267
		1268	sub aio_rmtree;
		1269	sub aio_rmtree($;$) {
		1270	my ($path, $cb) = @_;
		1271
		1272	my $pri = aioreq_pri;
		1273	my $grp = aio_group $cb;
		1274
		1275	aioreq_pri $pri;
		1276	add $grp aio_scandir $path, 0, sub {
		1277	my ($dirs, $nondirs) = @_;
		1278
		1279	my $dirgrp = aio_group sub {
		1280	add $grp aio_rmdir $path, sub {
		1281	$grp->result ($_[0]);
		1282	};
		1283	};
		1284
		1285	(aioreq_pri $pri), add $dirgrp aio_rmtree "$path/$_" for @$dirs;
		1286	(aioreq_pri $pri), add $dirgrp aio_unlink "$path/$_" for @$nondirs;
		1287
		1288	add $grp $dirgrp;
		1289	};
		1290
		1291	$grp
		1292	}
		1293
		1294	=item aio_fcntl $fh, $cmd, $arg, $callback->($status)
		1295
		1296	=item aio_ioctl $fh, $request, $buf, $callback->($status)
		1297
		1298	These work just like the C<fcntl> and C<ioctl> built-in functions, except
		1299	they execute asynchronously and pass the return value to the callback.
		1300
		1301	Both calls can be used for a lot of things, some of which make more sense
		1302	to run asynchronously in their own thread, while some others make less
		1303	sense. For example, calls that block waiting for external events, such
		1304	as locking, will also lock down an I/O thread while it is waiting, which
		1305	can deadlock the whole I/O system. At the same time, there might be no
		1306	alternative to using a thread to wait.
		1307
		1308	So in general, you should only use these calls for things that do
		1309	(filesystem) I/O, not for things that wait for other events (network,
		1310	other processes), although if you are careful and know what you are doing,
		1311	you still can.
		1312
		1313	The following constants are available and can be used for normal C<ioctl>
		1314	and C<fcntl> as well (missing ones are, as usual C<0>):
		1315
		1316	C<F_DUPFD_CLOEXEC>,
		1317
		1318	C<F_OFD_GETLK>, C<F_OFD_SETLK>, C<F_OFD_GETLKW>,
		1319
		1320	C<FIFREEZE>, C<FITHAW>, C<FITRIM>, C<FICLONE>, C<FICLONERANGE>, C<FIDEDUPERANGE>.
		1321
		1322	C<F_ADD_SEALS>, C<F_GET_SEALS>, C<F_SEAL_SEAL>, C<F_SEAL_SHRINK>, C<F_SEAL_GROW> and
		1323	C<F_SEAL_WRITE>.
		1324
		1325	C<FS_IOC_GETFLAGS>, C<FS_IOC_SETFLAGS>, C<FS_IOC_GETVERSION>, C<FS_IOC_SETVERSION>,
		1326	C<FS_IOC_FIEMAP>.
		1327
		1328	C<FS_IOC_FSGETXATTR>, C<FS_IOC_FSSETXATTR>, C<FS_IOC_SET_ENCRYPTION_POLICY>,
		1329	C<FS_IOC_GET_ENCRYPTION_PWSALT>, C<FS_IOC_GET_ENCRYPTION_POLICY>, C<FS_KEY_DESCRIPTOR_SIZE>.
		1330
		1331	C<FS_SECRM_FL>, C<FS_UNRM_FL>, C<FS_COMPR_FL>, C<FS_SYNC_FL>, C<FS_IMMUTABLE_FL>,
		1332	C<FS_APPEND_FL>, C<FS_NODUMP_FL>, C<FS_NOATIME_FL>, C<FS_DIRTY_FL>,
		1333	C<FS_COMPRBLK_FL>, C<FS_NOCOMP_FL>, C<FS_ENCRYPT_FL>, C<FS_BTREE_FL>,
		1334	C<FS_INDEX_FL>, C<FS_JOURNAL_DATA_FL>, C<FS_NOTAIL_FL>, C<FS_DIRSYNC_FL>, C<FS_TOPDIR_FL>,
		1335	C<FS_FL_USER_MODIFIABLE>.
		1336
		1337	C<FS_XFLAG_REALTIME>, C<FS_XFLAG_PREALLOC>, C<FS_XFLAG_IMMUTABLE>, C<FS_XFLAG_APPEND>,
		1338	C<FS_XFLAG_SYNC>, C<FS_XFLAG_NOATIME>, C<FS_XFLAG_NODUMP>, C<FS_XFLAG_RTINHERIT>,
		1339	C<FS_XFLAG_PROJINHERIT>, C<FS_XFLAG_NOSYMLINKS>, C<FS_XFLAG_EXTSIZE>, C<FS_XFLAG_EXTSZINHERIT>,
		1340	C<FS_XFLAG_NODEFRAG>, C<FS_XFLAG_FILESTREAM>, C<FS_XFLAG_DAX>, C<FS_XFLAG_HASATTR>,
		1341
		1342	=item aio_sync $callback->($status)
		1343
		1344	Asynchronously call sync and call the callback when finished.
		1345
184	=item aio_fsync $fh, $callback	1346	=item aio_fsync $fh, $callback->($status)
185		1347
186	Asynchronously call fsync on the given filehandle and call the callback	1348	Asynchronously call fsync on the given filehandle and call the callback
187	with the fsync result code.	1349	with the fsync result code.
188		1350
189	=item aio_fdatasync $fh, $callback	1351	=item aio_fdatasync $fh, $callback->($status)
190		1352
191	Asynchronously call fdatasync on the given filehandle and call the	1353	Asynchronously call fdatasync on the given filehandle and call the
192	callback with the fdatasync result code.	1354	callback with the fdatasync result code.
193		1355
		1356	If this call isn't available because your OS lacks it or it couldn't be
		1357	detected, it will be emulated by calling C<fsync> instead.
		1358
		1359	=item aio_syncfs $fh, $callback->($status)
		1360
		1361	Asynchronously call the syncfs syscall to sync the filesystem associated
		1362	to the given filehandle and call the callback with the syncfs result
		1363	code. If syncfs is not available, calls sync(), but returns C<-1> and sets
		1364	errno to C<ENOSYS> nevertheless.
		1365
		1366	=item aio_sync_file_range $fh, $offset, $nbytes, $flags, $callback->($status)
		1367
		1368	Sync the data portion of the file specified by C<$offset> and C<$length>
		1369	to disk (but NOT the metadata), by calling the Linux-specific
		1370	sync_file_range call. If sync_file_range is not available or it returns
		1371	ENOSYS, then fdatasync or fsync is being substituted.
		1372
		1373	C<$flags> can be a combination of C<IO::AIO::SYNC_FILE_RANGE_WAIT_BEFORE>,
		1374	C<IO::AIO::SYNC_FILE_RANGE_WRITE> and
		1375	C<IO::AIO::SYNC_FILE_RANGE_WAIT_AFTER>: refer to the sync_file_range
		1376	manpage for details.
		1377
		1378	=item aio_pathsync $pathname, $callback->($status)
		1379
		1380	This request tries to open, fsync and close the given path. This is a
		1381	composite request intended to sync directories after directory operations
		1382	(E.g. rename). This might not work on all operating systems or have any
		1383	specific effect, but usually it makes sure that directory changes get
		1384	written to disc. It works for anything that can be opened for read-only,
		1385	not just directories.
		1386
		1387	Future versions of this function might fall back to other methods when
		1388	C<fsync> on the directory fails (such as calling C<sync>).
		1389
		1390	Passes C<0> when everything went ok, and C<-1> on error.
		1391
		1392	=cut
		1393
		1394	sub aio_pathsync($;$) {
		1395	my ($path, $cb) = @_;
		1396
		1397	my $pri = aioreq_pri;
		1398	my $grp = aio_group $cb;
		1399
		1400	aioreq_pri $pri;
		1401	add $grp aio_open $path, O_RDONLY, 0, sub {
		1402	my ($fh) = @_;
		1403	if ($fh) {
		1404	aioreq_pri $pri;
		1405	add $grp aio_fsync $fh, sub {
		1406	$grp->result ($_[0]);
		1407
		1408	aioreq_pri $pri;
		1409	add $grp aio_close $fh;
		1410	};
		1411	} else {
		1412	$grp->result (-1);
		1413	}
		1414	};
		1415
		1416	$grp
		1417	}
		1418
		1419	=item aio_msync $scalar, $offset = 0, $length = undef, flags = MS_SYNC, $callback->($status)
		1420
		1421	This is a rather advanced IO::AIO call, which only works on mmap(2)ed
		1422	scalars (see the C<IO::AIO::mmap> function, although it also works on data
		1423	scalars managed by the L<Sys::Mmap> or L<Mmap> modules, note that the
		1424	scalar must only be modified in-place while an aio operation is pending on
		1425	it).
		1426
		1427	It calls the C<msync> function of your OS, if available, with the memory
		1428	area starting at C<$offset> in the string and ending C<$length> bytes
		1429	later. If C<$length> is negative, counts from the end, and if C<$length>
		1430	is C<undef>, then it goes till the end of the string. The flags can be
		1431	either C<IO::AIO::MS_ASYNC> or C<IO::AIO::MS_SYNC>, plus an optional
		1432	C<IO::AIO::MS_INVALIDATE>.
		1433
		1434	=item aio_mtouch $scalar, $offset = 0, $length = undef, flags = 0, $callback->($status)
		1435
		1436	This is a rather advanced IO::AIO call, which works best on mmap(2)ed
		1437	scalars.
		1438
		1439	It touches (reads or writes) all memory pages in the specified
		1440	range inside the scalar. All caveats and parameters are the same
		1441	as for C<aio_msync>, above, except for flags, which must be either
		1442	C<0> (which reads all pages and ensures they are instantiated) or
		1443	C<IO::AIO::MT_MODIFY>, which modifies the memory pages (by reading and
		1444	writing an octet from it, which dirties the page).
		1445
		1446	=item aio_mlock $scalar, $offset = 0, $length = undef, $callback->($status)
		1447
		1448	This is a rather advanced IO::AIO call, which works best on mmap(2)ed
		1449	scalars.
		1450
		1451	It reads in all the pages of the underlying storage into memory (if any)
		1452	and locks them, so they are not getting swapped/paged out or removed.
		1453
		1454	If C<$length> is undefined, then the scalar will be locked till the end.
		1455
		1456	On systems that do not implement C<mlock>, this function returns C<-1>
		1457	and sets errno to C<ENOSYS>.
		1458
		1459	Note that the corresponding C<munlock> is synchronous and is
		1460	documented under L<MISCELLANEOUS FUNCTIONS>.
		1461
		1462	Example: open a file, mmap and mlock it - both will be undone when
		1463	C<$data> gets destroyed.
		1464
		1465	open my $fh, "<", $path or die "$path: $!";
		1466	my $data;
		1467	IO::AIO::mmap $data, -s $fh, IO::AIO::PROT_READ, IO::AIO::MAP_SHARED, $fh;
		1468	aio_mlock $data; # mlock in background
		1469
		1470	=item aio_mlockall $flags, $callback->($status)
		1471
		1472	Calls the C<mlockall> function with the given C<$flags> (a
		1473	combination of C<IO::AIO::MCL_CURRENT>, C<IO::AIO::MCL_FUTURE> and
		1474	C<IO::AIO::MCL_ONFAULT>).
		1475
		1476	On systems that do not implement C<mlockall>, this function returns C<-1>
		1477	and sets errno to C<ENOSYS>. Similarly, flag combinations not supported
		1478	by the system result in a return value of C<-1> with errno being set to
		1479	C<EINVAL>.
		1480
		1481	Note that the corresponding C<munlockall> is synchronous and is
		1482	documented under L<MISCELLANEOUS FUNCTIONS>.
		1483
		1484	Example: asynchronously lock all current and future pages into memory.
		1485
		1486	aio_mlockall IO::AIO::MCL_FUTURE;
		1487
		1488	=item aio_fiemap $fh, $start, $length, $flags, $count, $cb->(\@extents)
		1489
		1490	Queries the extents of the given file (by calling the Linux C<FIEMAP>
		1491	ioctl, see L<http://cvs.schmorp.de/IO-AIO/doc/fiemap.txt> for details). If
		1492	the ioctl is not available on your OS, then this request will fail with
		1493	C<ENOSYS>.
		1494
		1495	C<$start> is the starting offset to query extents for, C<$length> is the
		1496	size of the range to query - if it is C<undef>, then the whole file will
		1497	be queried.
		1498
		1499	C<$flags> is a combination of flags (C<IO::AIO::FIEMAP_FLAG_SYNC> or
		1500	C<IO::AIO::FIEMAP_FLAG_XATTR> - C<IO::AIO::FIEMAP_FLAGS_COMPAT> is also
		1501	exported), and is normally C<0> or C<IO::AIO::FIEMAP_FLAG_SYNC> to query
		1502	the data portion.
		1503
		1504	C<$count> is the maximum number of extent records to return. If it is
		1505	C<undef>, then IO::AIO queries all extents of the range. As a very special
		1506	case, if it is C<0>, then the callback receives the number of extents
		1507	instead of the extents themselves (which is unreliable, see below).
		1508
		1509	If an error occurs, the callback receives no arguments. The special
		1510	C<errno> value C<IO::AIO::EBADR> is available to test for flag errors.
		1511
		1512	Otherwise, the callback receives an array reference with extent
		1513	structures. Each extent structure is an array reference itself, with the
		1514	following members:
		1515
		1516	[$logical, $physical, $length, $flags]
		1517
		1518	Flags is any combination of the following flag values (typically either C<0>
		1519	or C<IO::AIO::FIEMAP_EXTENT_LAST> (1)):
		1520
		1521	C<IO::AIO::FIEMAP_EXTENT_LAST>, C<IO::AIO::FIEMAP_EXTENT_UNKNOWN>,
		1522	C<IO::AIO::FIEMAP_EXTENT_DELALLOC>, C<IO::AIO::FIEMAP_EXTENT_ENCODED>,
		1523	C<IO::AIO::FIEMAP_EXTENT_DATA_ENCRYPTED>, C<IO::AIO::FIEMAP_EXTENT_NOT_ALIGNED>,
		1524	C<IO::AIO::FIEMAP_EXTENT_DATA_INLINE>, C<IO::AIO::FIEMAP_EXTENT_DATA_TAIL>,
		1525	C<IO::AIO::FIEMAP_EXTENT_UNWRITTEN>, C<IO::AIO::FIEMAP_EXTENT_MERGED> or
		1526	C<IO::AIO::FIEMAP_EXTENT_SHARED>.
		1527
		1528	At the time of this writing (Linux 3.2), this request is unreliable unless
		1529	C<$count> is C<undef>, as the kernel has all sorts of bugs preventing
		1530	it to return all extents of a range for files with a large number of
		1531	extents. The code (only) works around all these issues if C<$count> is
		1532	C<undef>.
		1533
		1534	=item aio_group $callback->(...)
		1535
		1536	This is a very special aio request: Instead of doing something, it is a
		1537	container for other aio requests, which is useful if you want to bundle
		1538	many requests into a single, composite, request with a definite callback
		1539	and the ability to cancel the whole request with its subrequests.
		1540
		1541	Returns an object of class L<IO::AIO::GRP>. See its documentation below
		1542	for more info.
		1543
		1544	Example:
		1545
		1546	my $grp = aio_group sub {
		1547	print "all stats done\n";
		1548	};
		1549
		1550	add $grp
		1551	(aio_stat ...),
		1552	(aio_stat ...),
		1553	...;
		1554
		1555	=item aio_nop $callback->()
		1556
		1557	This is a special request - it does nothing in itself and is only used for
		1558	side effects, such as when you want to add a dummy request to a group so
		1559	that finishing the requests in the group depends on executing the given
		1560	code.
		1561
		1562	While this request does nothing, it still goes through the execution
		1563	phase and still requires a worker thread. Thus, the callback will not
		1564	be executed immediately but only after other requests in the queue have
		1565	entered their execution phase. This can be used to measure request
		1566	latency.
		1567
		1568	=item IO::AIO::aio_busy $fractional_seconds, $callback->() *NOT EXPORTED*
		1569
		1570	Mainly used for debugging and benchmarking, this aio request puts one of
		1571	the request workers to sleep for the given time.
		1572
		1573	While it is theoretically handy to have simple I/O scheduling requests
		1574	like sleep and file handle readable/writable, the overhead this creates is
		1575	immense (it blocks a thread for a long time) so do not use this function
		1576	except to put your application under artificial I/O pressure.
		1577
194	=back	1578	=back
195		1579
		1580
		1581	=head2 IO::AIO::WD - multiple working directories
		1582
		1583	Your process only has one current working directory, which is used by all
		1584	threads. This makes it hard to use relative paths (some other component
		1585	could call C<chdir> at any time, and it is hard to control when the path
		1586	will be used by IO::AIO).
		1587
		1588	One solution for this is to always use absolute paths. This usually works,
		1589	but can be quite slow (the kernel has to walk the whole path on every
		1590	access), and can also be a hassle to implement.
		1591
		1592	Newer POSIX systems have a number of functions (openat, fdopendir,
		1593	futimensat and so on) that make it possible to specify working directories
		1594	per operation.
		1595
		1596	For portability, and because the clowns who "designed", or shall I write,
		1597	perpetrated this new interface were obviously half-drunk, this abstraction
		1598	cannot be perfect, though.
		1599
		1600	IO::AIO allows you to convert directory paths into a so-called IO::AIO::WD
		1601	object. This object stores the canonicalised, absolute version of the
		1602	path, and on systems that allow it, also a directory file descriptor.
		1603
		1604	Everywhere where a pathname is accepted by IO::AIO (e.g. in C<aio_stat>
		1605	or C<aio_unlink>), one can specify an array reference with an IO::AIO::WD
		1606	object and a pathname instead (or the IO::AIO::WD object alone, which
		1607	gets interpreted as C<[$wd, "."]>). If the pathname is absolute, the
		1608	IO::AIO::WD object is ignored, otherwise the pathname is resolved relative
		1609	to that IO::AIO::WD object.
		1610
		1611	For example, to get a wd object for F</etc> and then stat F<passwd>
		1612	inside, you would write:
		1613
		1614	aio_wd "/etc", sub {
		1615	my $etcdir = shift;
		1616
		1617	# although $etcdir can be undef on error, there is generally no reason
		1618	# to check for errors here, as aio_stat will fail with ENOENT
		1619	# when $etcdir is undef.
		1620
		1621	aio_stat [$etcdir, "passwd"], sub {
		1622	# yay
		1623	};
		1624	};
		1625
		1626	The fact that C<aio_wd> is a request and not a normal function shows that
		1627	creating an IO::AIO::WD object is itself a potentially blocking operation,
		1628	which is why it is done asynchronously.
		1629
		1630	To stat the directory obtained with C<aio_wd> above, one could write
		1631	either of the following three request calls:
		1632
		1633	aio_lstat "/etc" , sub { ... # pathname as normal string
		1634	aio_lstat [$wd, "."], sub { ... # "." relative to $wd (i.e. $wd itself)
		1635	aio_lstat $wd , sub { ... # shorthand for the previous
		1636
		1637	As with normal pathnames, IO::AIO keeps a copy of the working directory
		1638	object and the pathname string, so you could write the following without
		1639	causing any issues due to C<$path> getting reused:
		1640
		1641	my $path = [$wd, undef];
		1642
		1643	for my $name (qw(abc def ghi)) {
		1644	$path->[1] = $name;
		1645	aio_stat $path, sub {
		1646	# ...
		1647	};
		1648	}
		1649
		1650	There are some caveats: when directories get renamed (or deleted), the
		1651	pathname string doesn't change, so will point to the new directory (or
		1652	nowhere at all), while the directory fd, if available on the system,
		1653	will still point to the original directory. Most functions accepting a
		1654	pathname will use the directory fd on newer systems, and the string on
		1655	older systems. Some functions (such as C<aio_realpath>) will always rely on
		1656	the string form of the pathname.
		1657
		1658	So this functionality is mainly useful to get some protection against
		1659	C<chdir>, to easily get an absolute path out of a relative path for future
		1660	reference, and to speed up doing many operations in the same directory
		1661	(e.g. when stat'ing all files in a directory).
		1662
		1663	The following functions implement this working directory abstraction:
		1664
		1665	=over 4
		1666
		1667	=item aio_wd $pathname, $callback->($wd)
		1668
		1669	Asynchonously canonicalise the given pathname and convert it to an
		1670	IO::AIO::WD object representing it. If possible and supported on the
		1671	system, also open a directory fd to speed up pathname resolution relative
		1672	to this working directory.
		1673
		1674	If something goes wrong, then C<undef> is passwd to the callback instead
		1675	of a working directory object and C<$!> is set appropriately. Since
		1676	passing C<undef> as working directory component of a pathname fails the
		1677	request with C<ENOENT>, there is often no need for error checking in the
		1678	C<aio_wd> callback, as future requests using the value will fail in the
		1679	expected way.
		1680
		1681	=item IO::AIO::CWD
		1682
		1683	This is a compiletime constant (object) that represents the process
		1684	current working directory.
		1685
		1686	Specifying this object as working directory object for a pathname is as if
		1687	the pathname would be specified directly, without a directory object. For
		1688	example, these calls are functionally identical:
		1689
		1690	aio_stat "somefile", sub { ... };
		1691	aio_stat [IO::AIO::CWD, "somefile"], sub { ... };
		1692
		1693	=back
		1694
		1695	To recover the path associated with an IO::AIO::WD object, you can use
		1696	C<aio_realpath>:
		1697
		1698	aio_realpath $wd, sub {
		1699	warn "path is $_[0]\n";
		1700	};
		1701
		1702	Currently, C<aio_statvfs> always, and C<aio_rename> and C<aio_rmdir>
		1703	sometimes, fall back to using an absolue path.
		1704
		1705	=head2 IO::AIO::REQ CLASS
		1706
		1707	All non-aggregate C<aio_*> functions return an object of this class when
		1708	called in non-void context.
		1709
		1710	=over 4
		1711
		1712	=item cancel $req
		1713
		1714	Cancels the request, if possible. Has the effect of skipping execution
		1715	when entering the B<execute> state and skipping calling the callback when
		1716	entering the the B<result> state, but will leave the request otherwise
		1717	untouched (with the exception of readdir). That means that requests that
		1718	currently execute will not be stopped and resources held by the request
		1719	will not be freed prematurely.
		1720
		1721	=item cb $req $callback->(...)
		1722
		1723	Replace (or simply set) the callback registered to the request.
		1724
		1725	=back
		1726
		1727	=head2 IO::AIO::GRP CLASS
		1728
		1729	This class is a subclass of L<IO::AIO::REQ>, so all its methods apply to
		1730	objects of this class, too.
		1731
		1732	A IO::AIO::GRP object is a special request that can contain multiple other
		1733	aio requests.
		1734
		1735	You create one by calling the C<aio_group> constructing function with a
		1736	callback that will be called when all contained requests have entered the
		1737	C<done> state:
		1738
		1739	my $grp = aio_group sub {
		1740	print "all requests are done\n";
		1741	};
		1742
		1743	You add requests by calling the C<add> method with one or more
		1744	C<IO::AIO::REQ> objects:
		1745
		1746	$grp->add (aio_unlink "...");
		1747
		1748	add $grp aio_stat "...", sub {
		1749	$_[0] or return $grp->result ("error");
		1750
		1751	# add another request dynamically, if first succeeded
		1752	add $grp aio_open "...", sub {
		1753	$grp->result ("ok");
		1754	};
		1755	};
		1756
		1757	This makes it very easy to create composite requests (see the source of
		1758	C<aio_move> for an application) that work and feel like simple requests.
		1759
		1760	=over 4
		1761
		1762	=item * The IO::AIO::GRP objects will be cleaned up during calls to
		1763	C<IO::AIO::poll_cb>, just like any other request.
		1764
		1765	=item * They can be canceled like any other request. Canceling will cancel not
		1766	only the request itself, but also all requests it contains.
		1767
		1768	=item * They can also can also be added to other IO::AIO::GRP objects.
		1769
		1770	=item * You must not add requests to a group from within the group callback (or
		1771	any later time).
		1772
		1773	=back
		1774
		1775	Their lifetime, simplified, looks like this: when they are empty, they
		1776	will finish very quickly. If they contain only requests that are in the
		1777	C<done> state, they will also finish. Otherwise they will continue to
		1778	exist.
		1779
		1780	That means after creating a group you have some time to add requests
		1781	(precisely before the callback has been invoked, which is only done within
		1782	the C<poll_cb>). And in the callbacks of those requests, you can add
		1783	further requests to the group. And only when all those requests have
		1784	finished will the the group itself finish.
		1785
		1786	=over 4
		1787
		1788	=item add $grp ...
		1789
		1790	=item $grp->add (...)
		1791
		1792	Add one or more requests to the group. Any type of L<IO::AIO::REQ> can
		1793	be added, including other groups, as long as you do not create circular
		1794	dependencies.
		1795
		1796	Returns all its arguments.
		1797
		1798	=item $grp->cancel_subs
		1799
		1800	Cancel all subrequests and clears any feeder, but not the group request
		1801	itself. Useful when you queued a lot of events but got a result early.
		1802
		1803	The group request will finish normally (you cannot add requests to the
		1804	group).
		1805
		1806	=item $grp->result (...)
		1807
		1808	Set the result value(s) that will be passed to the group callback when all
		1809	subrequests have finished and set the groups errno to the current value
		1810	of errno (just like calling C<errno> without an error number). By default,
		1811	no argument will be passed and errno is zero.
		1812
		1813	=item $grp->errno ([$errno])
		1814
		1815	Sets the group errno value to C<$errno>, or the current value of errno
		1816	when the argument is missing.
		1817
		1818	Every aio request has an associated errno value that is restored when
		1819	the callback is invoked. This method lets you change this value from its
		1820	default (0).
		1821
		1822	Calling C<result> will also set errno, so make sure you either set C<$!>
		1823	before the call to C<result>, or call c<errno> after it.
		1824
		1825	=item feed $grp $callback->($grp)
		1826
		1827	Sets a feeder/generator on this group: every group can have an attached
		1828	generator that generates requests if idle. The idea behind this is that,
		1829	although you could just queue as many requests as you want in a group,
		1830	this might starve other requests for a potentially long time. For example,
		1831	C<aio_scandir> might generate hundreds of thousands of C<aio_stat>
		1832	requests, delaying any later requests for a long time.
		1833
		1834	To avoid this, and allow incremental generation of requests, you can
		1835	instead a group and set a feeder on it that generates those requests. The
		1836	feed callback will be called whenever there are few enough (see C<limit>,
		1837	below) requests active in the group itself and is expected to queue more
		1838	requests.
		1839
		1840	The feed callback can queue as many requests as it likes (i.e. C<add> does
		1841	not impose any limits).
		1842
		1843	If the feed does not queue more requests when called, it will be
		1844	automatically removed from the group.
		1845
		1846	If the feed limit is C<0> when this method is called, it will be set to
		1847	C<2> automatically.
		1848
		1849	Example:
		1850
		1851	# stat all files in @files, but only ever use four aio requests concurrently:
		1852
		1853	my $grp = aio_group sub { print "finished\n" };
		1854	limit $grp 4;
		1855	feed $grp sub {
		1856	my $file = pop @files
		1857	or return;
		1858
		1859	add $grp aio_stat $file, sub { ... };
		1860	};
		1861
		1862	=item limit $grp $num
		1863
		1864	Sets the feeder limit for the group: The feeder will be called whenever
		1865	the group contains less than this many requests.
		1866
		1867	Setting the limit to C<0> will pause the feeding process.
		1868
		1869	The default value for the limit is C<0>, but note that setting a feeder
		1870	automatically bumps it up to C<2>.
		1871
		1872	=back
		1873
		1874
196	=head2 SUPPORT FUNCTIONS	1875	=head2 SUPPORT FUNCTIONS
197		1876
		1877	=head3 EVENT PROCESSING AND EVENT LOOP INTEGRATION
		1878
198	=over 4	1879	=over 4
199		1880
200	=item $fileno = IO::AIO::poll_fileno	1881	=item $fileno = IO::AIO::poll_fileno
201		1882
202	Return the I<request result pipe filehandle>. This filehandle must be	1883	Return the I<request result pipe file descriptor>. This filehandle must be
203	polled for reading by some mechanism outside this module (e.g. Event	1884	polled for reading by some mechanism outside this module (e.g. EV, Glib,
204	or select, see below). If the pipe becomes readable you have to call	1885	select and so on, see below or the SYNOPSIS). If the pipe becomes readable
205	C<poll_cb> to check the results.	1886	you have to call C<poll_cb> to check the results.
206		1887
207	See C<poll_cb> for an example.	1888	See C<poll_cb> for an example.
208		1889
209	=item IO::AIO::poll_cb	1890	=item IO::AIO::poll_cb
210		1891
211	Process all outstanding events on the result pipe. You have to call this	1892	Process some requests that have reached the result phase (i.e. they have
212	regularly. Returns the number of events processed. Returns immediately	1893	been executed but the results are not yet reported). You have to call
213	when no events are outstanding.	1894	this "regularly" to finish outstanding requests.
214		1895
215	You can use Event to multiplex, e.g.:	1896	Returns C<0> if all events could be processed (or there were no
		1897	events to process), or C<-1> if it returned earlier for whatever
		1898	reason. Returns immediately when no events are outstanding. The amount
		1899	of events processed depends on the settings of C<IO::AIO::max_poll_req>,
		1900	C<IO::AIO::max_poll_time> and C<IO::AIO::max_outstanding>.
		1901
		1902	If not all requests were processed for whatever reason, the poll file
		1903	descriptor will still be ready when C<poll_cb> returns, so normally you
		1904	don't have to do anything special to have it called later.
		1905
		1906	Apart from calling C<IO::AIO::poll_cb> when the event filehandle becomes
		1907	ready, it can be beneficial to call this function from loops which submit
		1908	a lot of requests, to make sure the results get processed when they become
		1909	available and not just when the loop is finished and the event loop takes
		1910	over again. This function returns very fast when there are no outstanding
		1911	requests.
		1912
		1913	Example: Install an Event watcher that automatically calls
		1914	IO::AIO::poll_cb with high priority (more examples can be found in the
		1915	SYNOPSIS section, at the top of this document):
216		1916
217	Event->io (fd => IO::AIO::poll_fileno,	1917	Event->io (fd => IO::AIO::poll_fileno,
218	poll => 'r', async => 1,	1918	poll => 'r', async => 1,
219	cb => \&IO::AIO::poll_cb);	1919	cb => \&IO::AIO::poll_cb);
220		1920
221	=item IO::AIO::poll_wait	1921	=item IO::AIO::poll_wait
222		1922
223	Wait till the result filehandle becomes ready for reading (simply does a	1923	Wait until either at least one request is in the result phase or no
224	select on the filehandle. This is useful if you want to synchronously wait	1924	requests are outstanding anymore.
225	for some requests to finish).	1925
		1926	This is useful if you want to synchronously wait for some requests to
		1927	become ready, without actually handling them.
226		1928
227	See C<nreqs> for an example.	1929	See C<nreqs> for an example.
228		1930
		1931	=item IO::AIO::poll
		1932
		1933	Waits until some requests have been handled.
		1934
		1935	Returns the number of requests processed, but is otherwise strictly
		1936	equivalent to:
		1937
		1938	IO::AIO::poll_wait, IO::AIO::poll_cb
		1939
229	=item IO::AIO::nreqs	1940	=item IO::AIO::flush
230		1941
231	Returns the number of requests currently outstanding.	1942	Wait till all outstanding AIO requests have been handled.
232		1943
233	Example: wait till there are no outstanding requests anymore:	1944	Strictly equivalent to:
234		1945
235	IO::AIO::poll_wait, IO::AIO::poll_cb	1946	IO::AIO::poll_wait, IO::AIO::poll_cb
236	while IO::AIO::nreqs;	1947	while IO::AIO::nreqs;
237		1948
		1949	This function can be useful at program aborts, to make sure outstanding
		1950	I/O has been done (C<IO::AIO> uses an C<END> block which already calls
		1951	this function on normal exits), or when you are merely using C<IO::AIO>
		1952	for its more advanced functions, rather than for async I/O, e.g.:
		1953
		1954	my ($dirs, $nondirs);
		1955	IO::AIO::aio_scandir "/tmp", 0, sub { ($dirs, $nondirs) = @_ };
		1956	IO::AIO::flush;
		1957	# $dirs, $nondirs are now set
		1958
		1959	=item IO::AIO::max_poll_reqs $nreqs
		1960
		1961	=item IO::AIO::max_poll_time $seconds
		1962
		1963	These set the maximum number of requests (default C<0>, meaning infinity)
		1964	that are being processed by C<IO::AIO::poll_cb> in one call, respectively
		1965	the maximum amount of time (default C<0>, meaning infinity) spent in
		1966	C<IO::AIO::poll_cb> to process requests (more correctly the mininum amount
		1967	of time C<poll_cb> is allowed to use).
		1968
		1969	Setting C<max_poll_time> to a non-zero value creates an overhead of one
		1970	syscall per request processed, which is not normally a problem unless your
		1971	callbacks are really really fast or your OS is really really slow (I am
		1972	not mentioning Solaris here). Using C<max_poll_reqs> incurs no overhead.
		1973
		1974	Setting these is useful if you want to ensure some level of
		1975	interactiveness when perl is not fast enough to process all requests in
		1976	time.
		1977
		1978	For interactive programs, values such as C<0.01> to C<0.1> should be fine.
		1979
		1980	Example: Install an Event watcher that automatically calls
		1981	IO::AIO::poll_cb with low priority, to ensure that other parts of the
		1982	program get the CPU sometimes even under high AIO load.
		1983
		1984	# try not to spend much more than 0.1s in poll_cb
		1985	IO::AIO::max_poll_time 0.1;
		1986
		1987	# use a low priority so other tasks have priority
		1988	Event->io (fd => IO::AIO::poll_fileno,
		1989	poll => 'r', nice => 1,
		1990	cb => &IO::AIO::poll_cb);
		1991
		1992	=back
		1993
		1994
		1995	=head3 CONTROLLING THE NUMBER OF THREADS
		1996
		1997	=over
		1998
		1999	=item IO::AIO::min_parallel $nthreads
		2000
		2001	Set the minimum number of AIO threads to C<$nthreads>. The current
		2002	default is C<8>, which means eight asynchronous operations can execute
		2003	concurrently at any one time (the number of outstanding requests,
		2004	however, is unlimited).
		2005
		2006	IO::AIO starts threads only on demand, when an AIO request is queued and
		2007	no free thread exists. Please note that queueing up a hundred requests can
		2008	create demand for a hundred threads, even if it turns out that everything
		2009	is in the cache and could have been processed faster by a single thread.
		2010
		2011	It is recommended to keep the number of threads relatively low, as some
		2012	Linux kernel versions will scale negatively with the number of threads
		2013	(higher parallelity => MUCH higher latency). With current Linux 2.6
		2014	versions, 4-32 threads should be fine.
		2015
		2016	Under most circumstances you don't need to call this function, as the
		2017	module selects a default that is suitable for low to moderate load.
		2018
		2019	=item IO::AIO::max_parallel $nthreads
		2020
		2021	Sets the maximum number of AIO threads to C<$nthreads>. If more than the
		2022	specified number of threads are currently running, this function kills
		2023	them. This function blocks until the limit is reached.
		2024
		2025	While C<$nthreads> are zero, aio requests get queued but not executed
		2026	until the number of threads has been increased again.
		2027
		2028	This module automatically runs C<max_parallel 0> at program end, to ensure
		2029	that all threads are killed and that there are no outstanding requests.
		2030
		2031	Under normal circumstances you don't need to call this function.
		2032
		2033	=item IO::AIO::max_idle $nthreads
		2034
		2035	Limit the number of threads (default: 4) that are allowed to idle
		2036	(i.e., threads that did not get a request to process within the idle
		2037	timeout (default: 10 seconds). That means if a thread becomes idle while
		2038	C<$nthreads> other threads are also idle, it will free its resources and
		2039	exit.
		2040
		2041	This is useful when you allow a large number of threads (e.g. 100 or 1000)
		2042	to allow for extremely high load situations, but want to free resources
		2043	under normal circumstances (1000 threads can easily consume 30MB of RAM).
		2044
		2045	The default is probably ok in most situations, especially if thread
		2046	creation is fast. If thread creation is very slow on your system you might
		2047	want to use larger values.
		2048
		2049	=item IO::AIO::idle_timeout $seconds
		2050
		2051	Sets the minimum idle timeout (default 10) after which worker threads are
		2052	allowed to exit. SEe C<IO::AIO::max_idle>.
		2053
		2054	=item IO::AIO::max_outstanding $maxreqs
		2055
		2056	Sets the maximum number of outstanding requests to C<$nreqs>. If
		2057	you do queue up more than this number of requests, the next call to
		2058	C<IO::AIO::poll_cb> (and other functions calling C<poll_cb>, such as
		2059	C<IO::AIO::flush> or C<IO::AIO::poll>) will block until the limit is no
		2060	longer exceeded.
		2061
		2062	In other words, this setting does not enforce a queue limit, but can be
		2063	used to make poll functions block if the limit is exceeded.
		2064
		2065	This is a very bad function to use in interactive programs because it
		2066	blocks, and a bad way to reduce concurrency because it is inexact: Better
		2067	use an C<aio_group> together with a feed callback.
		2068
		2069	Its main use is in scripts without an event loop - when you want to stat
		2070	a lot of files, you can write something like this:
		2071
		2072	IO::AIO::max_outstanding 32;
		2073
		2074	for my $path (...) {
		2075	aio_stat $path , ...;
		2076	IO::AIO::poll_cb;
		2077	}
		2078
		2079	IO::AIO::flush;
		2080
		2081	The call to C<poll_cb> inside the loop will normally return instantly, but
		2082	as soon as more thna C<32> reqeusts are in-flight, it will block until
		2083	some requests have been handled. This keeps the loop from pushing a large
		2084	number of C<aio_stat> requests onto the queue.
		2085
		2086	The default value for C<max_outstanding> is very large, so there is no
		2087	practical limit on the number of outstanding requests.
		2088
		2089	=back
		2090
		2091
		2092	=head3 STATISTICAL INFORMATION
		2093
		2094	=over
		2095
238	=item IO::AIO::flush	2096	=item IO::AIO::nreqs
239		2097
240	Wait till all outstanding AIO requests have been handled.	2098	Returns the number of requests currently in the ready, execute or pending
		2099	states (i.e. for which their callback has not been invoked yet).
241		2100
242	Strictly equivalent to:	2101	Example: wait till there are no outstanding requests anymore:
243		2102
244	IO::AIO::poll_wait, IO::AIO::poll_cb	2103	IO::AIO::poll_wait, IO::AIO::poll_cb
245	while IO::AIO::nreqs;	2104	while IO::AIO::nreqs;
246		2105
		2106	=item IO::AIO::nready
		2107
		2108	Returns the number of requests currently in the ready state (not yet
		2109	executed).
		2110
247	=item IO::AIO::poll	2111	=item IO::AIO::npending
248		2112
249	Waits until some requests have been handled.	2113	Returns the number of requests currently in the pending state (executed,
250		2114	but not yet processed by poll_cb).
251	Strictly equivalent to:
252
253	IO::AIO::poll_wait, IO::AIO::poll_cb
254	if IO::AIO::nreqs;
255
256	=item IO::AIO::min_parallel $nthreads
257
258	Set the minimum number of AIO threads to C<$nthreads>. The default is
259	C<1>, which means a single asynchronous operation can be done at one time
260	(the number of outstanding operations, however, is unlimited).
261
262	It is recommended to keep the number of threads low, as some Linux
263	kernel versions will scale negatively with the number of threads (higher
264	parallelity => MUCH higher latency). With current Linux 2.6 versions, 4-32
265	threads should be fine.
266
267	Under normal circumstances you don't need to call this function, as this
268	module automatically starts some threads (the exact number might change,
269	and is currently 4).
270
271	=item IO::AIO::max_parallel $nthreads
272
273	Sets the maximum number of AIO threads to C<$nthreads>. If more than
274	the specified number of threads are currently running, kill them. This
275	function blocks until the limit is reached.
276
277	This module automatically runs C<max_parallel 0> at program end, to ensure
278	that all threads are killed and that there are no outstanding requests.
279
280	Under normal circumstances you don't need to call this function.
281
282	=item $oldnreqs = IO::AIO::max_outstanding $nreqs
283
284	Sets the maximum number of outstanding requests to C<$nreqs>. If you
285	try to queue up more than this number of requests, the caller will block until
286	some requests have been handled.
287
288	The default is very large, so normally there is no practical limit. If you
289	queue up many requests in a loop it it often improves speed if you set
290	this to a relatively low number, such as C<100>.
291
292	Under normal circumstances you don't need to call this function.
293		2115
294	=back	2116	=back
295		2117
		2118
		2119	=head3 SUBSECOND STAT TIME ACCESS
		2120
		2121	Both C<aio_stat>/C<aio_lstat> and perl's C<stat>/C<lstat> functions can
		2122	generally find access/modification and change times with subsecond time
		2123	accuracy of the system supports it, but perl's built-in functions only
		2124	return the integer part.
		2125
		2126	The following functions return the timestamps of the most recent
		2127	stat with subsecond precision on most systems and work both after
		2128	C<aio_stat>/C<aio_lstat> and perl's C<stat>/C<lstat> calls. Their return
		2129	value is only meaningful after a successful C<stat>/C<lstat> call, or
		2130	during/after a successful C<aio_stat>/C<aio_lstat> callback.
		2131
		2132	This is similar to the L<Time::HiRes> C<stat> functions, but can return
		2133	full resolution without rounding and work with standard perl C<stat>,
		2134	alleviating the need to call the special C<Time::HiRes> functions, which
		2135	do not act like their perl counterparts.
		2136
		2137	On operating systems or file systems where subsecond time resolution is
		2138	not supported or could not be detected, a fractional part of C<0> is
		2139	returned, so it is always safe to call these functions.
		2140
		2141	=over 4
		2142
		2143	=item $seconds = IO::AIO::st_atime, IO::AIO::st_mtime, IO::AIO::st_ctime, IO::AIO::st_btime
		2144
		2145	Return the access, modication, change or birth time, respectively,
		2146	including fractional part. Due to the limited precision of floating point,
		2147	the accuracy on most platforms is only a bit better than milliseconds
		2148	for times around now - see the I<nsec> function family, below, for full
		2149	accuracy.
		2150
		2151	File birth time is only available when the OS and perl support it (on
		2152	FreeBSD and NetBSD at the time of this writing, although support is
		2153	adaptive, so if your OS/perl gains support, IO::AIO can take advantage of
		2154	it). On systems where it isn't available, C<0> is currently returned, but
		2155	this might change to C<undef> in a future version.
		2156
		2157	=item ($atime, $mtime, $ctime, $btime, ...) = IO::AIO::st_xtime
		2158
		2159	Returns access, modification, change and birth time all in one go, and
		2160	maybe more times in the future version.
		2161
		2162	=item $nanoseconds = IO::AIO::st_atimensec, IO::AIO::st_mtimensec, IO::AIO::st_ctimensec, IO::AIO::st_btimensec
		2163
		2164	Return the fractional access, modifcation, change or birth time, in nanoseconds,
		2165	as an integer in the range C<0> to C<999999999>.
		2166
		2167	Note that no accessors are provided for access, modification and
		2168	change times - you need to get those from C<stat _> if required (C<int
		2169	IO::AIO::st_atime> and so on will I<not> generally give you the correct
		2170	value).
		2171
		2172	=item $seconds = IO::AIO::st_btimesec
		2173
		2174	The (integral) seconds part of the file birth time, if available.
		2175
		2176	=item ($atime, $mtime, $ctime, $btime, ...) = IO::AIO::st_xtimensec
		2177
		2178	Like the functions above, but returns all four times in one go (and maybe
		2179	more in future versions).
		2180
		2181	=item $counter = IO::AIO::st_gen
		2182
		2183	Returns the generation counter (in practice this is just a random number)
		2184	of the file. This is only available on platforms which have this member in
		2185	their C<struct stat> (most BSDs at the time of this writing) and generally
		2186	only to the root usert. If unsupported, C<0> is returned, but this might
		2187	change to C<undef> in a future version.
		2188
		2189	=back
		2190
		2191	Example: print the high resolution modification time of F</etc>, using
		2192	C<stat>, and C<IO::AIO::aio_stat>.
		2193
		2194	if (stat "/etc") {
		2195	printf "stat(/etc) mtime: %f\n", IO::AIO::st_mtime;
		2196	}
		2197
		2198	IO::AIO::aio_stat "/etc", sub {
		2199	$_[0]
		2200	and return;
		2201
		2202	printf "aio_stat(/etc) mtime: %d.%09d\n", (stat _)[9], IO::AIO::st_mtimensec;
		2203	};
		2204
		2205	IO::AIO::flush;
		2206
		2207	Output of the awbove on my system, showing reduced and full accuracy:
		2208
		2209	stat(/etc) mtime: 1534043702.020808
		2210	aio_stat(/etc) mtime: 1534043702.020807792
		2211
		2212
		2213	=head3 MISCELLANEOUS FUNCTIONS
		2214
		2215	IO::AIO implements some functions that are useful when you want to use
		2216	some "Advanced I/O" function not available to in Perl, without going the
		2217	"Asynchronous I/O" route. Many of these have an asynchronous C<aio_*>
		2218	counterpart.
		2219
		2220	=over 4
		2221
		2222	=item $numfd = IO::AIO::get_fdlimit
		2223
		2224	This function is I<EXPERIMENTAL> and subject to change.
		2225
		2226	Tries to find the current file descriptor limit and returns it, or
		2227	C<undef> and sets C<$!> in case of an error. The limit is one larger than
		2228	the highest valid file descriptor number.
		2229
		2230	=item IO::AIO::min_fdlimit [$numfd]
		2231
		2232	This function is I<EXPERIMENTAL> and subject to change.
		2233
		2234	Try to increase the current file descriptor limit(s) to at least C<$numfd>
		2235	by changing the soft or hard file descriptor resource limit. If C<$numfd>
		2236	is missing, it will try to set a very high limit, although this is not
		2237	recommended when you know the actual minimum that you require.
		2238
		2239	If the limit cannot be raised enough, the function makes a best-effort
		2240	attempt to increase the limit as much as possible, using various
		2241	tricks, while still failing. You can query the resulting limit using
		2242	C<IO::AIO::get_fdlimit>.
		2243
		2244	If an error occurs, returns C<undef> and sets C<$!>, otherwise returns
		2245	true.
		2246
		2247	=item IO::AIO::sendfile $ofh, $ifh, $offset, $count
		2248
		2249	Calls the C<eio_sendfile_sync> function, which is like C<aio_sendfile>,
		2250	but is blocking (this makes most sense if you know the input data is
		2251	likely cached already and the output filehandle is set to non-blocking
		2252	operations).
		2253
		2254	Returns the number of bytes copied, or C<-1> on error.
		2255
		2256	=item IO::AIO::fadvise $fh, $offset, $len, $advice
		2257
		2258	Simply calls the C<posix_fadvise> function (see its
		2259	manpage for details). The following advice constants are
		2260	available: C<IO::AIO::FADV_NORMAL>, C<IO::AIO::FADV_SEQUENTIAL>,
		2261	C<IO::AIO::FADV_RANDOM>, C<IO::AIO::FADV_NOREUSE>,
		2262	C<IO::AIO::FADV_WILLNEED>, C<IO::AIO::FADV_DONTNEED>.
		2263
		2264	On systems that do not implement C<posix_fadvise>, this function returns
		2265	ENOSYS, otherwise the return value of C<posix_fadvise>.
		2266
		2267	=item IO::AIO::madvise $scalar, $offset, $len, $advice
		2268
		2269	Simply calls the C<posix_madvise> function (see its
		2270	manpage for details). The following advice constants are
		2271	available: C<IO::AIO::MADV_NORMAL>, C<IO::AIO::MADV_SEQUENTIAL>,
		2272	C<IO::AIO::MADV_RANDOM>, C<IO::AIO::MADV_WILLNEED>,
		2273	C<IO::AIO::MADV_DONTNEED>.
		2274
		2275	If C<$offset> is negative, counts from the end. If C<$length> is negative,
		2276	the remaining length of the C<$scalar> is used. If possible, C<$length>
		2277	will be reduced to fit into the C<$scalar>.
		2278
		2279	On systems that do not implement C<posix_madvise>, this function returns
		2280	ENOSYS, otherwise the return value of C<posix_madvise>.
		2281
		2282	=item IO::AIO::mprotect $scalar, $offset, $len, $protect
		2283
		2284	Simply calls the C<mprotect> function on the preferably AIO::mmap'ed
		2285	$scalar (see its manpage for details). The following protect
		2286	constants are available: C<IO::AIO::PROT_NONE>, C<IO::AIO::PROT_READ>,
		2287	C<IO::AIO::PROT_WRITE>, C<IO::AIO::PROT_EXEC>.
		2288
		2289	If C<$offset> is negative, counts from the end. If C<$length> is negative,
		2290	the remaining length of the C<$scalar> is used. If possible, C<$length>
		2291	will be reduced to fit into the C<$scalar>.
		2292
		2293	On systems that do not implement C<mprotect>, this function returns
		2294	ENOSYS, otherwise the return value of C<mprotect>.
		2295
		2296	=item IO::AIO::mmap $scalar, $length, $prot, $flags, $fh[, $offset]
		2297
		2298	Memory-maps a file (or anonymous memory range) and attaches it to the
		2299	given C<$scalar>, which will act like a string scalar. Returns true on
		2300	success, and false otherwise.
		2301
		2302	The scalar must exist, but its contents do not matter - this means you
		2303	cannot use a nonexistant array or hash element. When in doubt, C<undef>
		2304	the scalar first.
		2305
		2306	The only operations allowed on the mmapped scalar are C<substr>/C<vec>,
		2307	which don't change the string length, and most read-only operations such
		2308	as copying it or searching it with regexes and so on.
		2309
		2310	Anything else is unsafe and will, at best, result in memory leaks.
		2311
		2312	The memory map associated with the C<$scalar> is automatically removed
		2313	when the C<$scalar> is undef'd or destroyed, or when the C<IO::AIO::mmap>
		2314	or C<IO::AIO::munmap> functions are called on it.
		2315
		2316	This calls the C<mmap>(2) function internally. See your system's manual
		2317	page for details on the C<$length>, C<$prot> and C<$flags> parameters.
		2318
		2319	The C<$length> must be larger than zero and smaller than the actual
		2320	filesize.
		2321
		2322	C<$prot> is a combination of C<IO::AIO::PROT_NONE>, C<IO::AIO::PROT_EXEC>,
		2323	C<IO::AIO::PROT_READ> and/or C<IO::AIO::PROT_WRITE>,
		2324
		2325	C<$flags> can be a combination of
		2326	C<IO::AIO::MAP_SHARED> or
		2327	C<IO::AIO::MAP_PRIVATE>,
		2328	or a number of system-specific flags (when not available, the are C<0>):
		2329	C<IO::AIO::MAP_ANONYMOUS> (which is set to C<MAP_ANON> if your system only provides this constant),
		2330	C<IO::AIO::MAP_LOCKED>,
		2331	C<IO::AIO::MAP_NORESERVE>,
		2332	C<IO::AIO::MAP_POPULATE>,
		2333	C<IO::AIO::MAP_NONBLOCK>,
		2334	C<IO::AIO::MAP_FIXED>,
		2335	C<IO::AIO::MAP_GROWSDOWN>,
		2336	C<IO::AIO::MAP_32BIT>,
		2337	C<IO::AIO::MAP_HUGETLB> or
		2338	C<IO::AIO::MAP_STACK>.
		2339
		2340	If C<$fh> is C<undef>, then a file descriptor of C<-1> is passed.
		2341
		2342	C<$offset> is the offset from the start of the file - it generally must be
		2343	a multiple of C<IO::AIO::PAGESIZE> and defaults to C<0>.
		2344
		2345	Example:
		2346
		2347	use Digest::MD5;
		2348	use IO::AIO;
		2349
		2350	open my $fh, "<verybigfile"
		2351	or die "$!";
		2352
		2353	IO::AIO::mmap my $data, -s $fh, IO::AIO::PROT_READ, IO::AIO::MAP_SHARED, $fh
		2354	or die "verybigfile: $!";
		2355
		2356	my $fast_md5 = md5 $data;
		2357
		2358	=item IO::AIO::munmap $scalar
		2359
		2360	Removes a previous mmap and undefines the C<$scalar>.
		2361
		2362	=item IO::AIO::mremap $scalar, $new_length, $flags = MREMAP_MAYMOVE[, $new_address = 0]
		2363
		2364	Calls the Linux-specific mremap(2) system call. The C<$scalar> must have
		2365	been mapped by C<IO::AIO::mmap>, and C<$flags> must currently either be
		2366	C<0> or C<IO::AIO::MREMAP_MAYMOVE>.
		2367
		2368	Returns true if successful, and false otherwise. If the underlying mmapped
		2369	region has changed address, then the true value has the numerical value
		2370	C<1>, otherwise it has the numerical value C<0>:
		2371
		2372	my $success = IO::AIO::mremap $mmapped, 8192, IO::AIO::MREMAP_MAYMOVE
		2373	or die "mremap: $!";
		2374
		2375	if ($success*1) {
		2376	warn "scalar has chanegd address in memory\n";
		2377	}
		2378
		2379	C<IO::AIO::MREMAP_FIXED> and the C<$new_address> argument are currently
		2380	implemented, but not supported and might go away in a future version.
		2381
		2382	On systems where this call is not supported or is not emulated, this call
		2383	returns falls and sets C<$!> to C<ENOSYS>.
		2384
		2385	=item IO::AIO::mlockall $flags
		2386
		2387	Calls the C<eio_mlockall_sync> function, which is like C<aio_mlockall>,
		2388	but is blocking.
		2389
		2390	=item IO::AIO::munlock $scalar, $offset = 0, $length = undef
		2391
		2392	Calls the C<munlock> function, undoing the effects of a previous
		2393	C<aio_mlock> call (see its description for details).
		2394
		2395	=item IO::AIO::munlockall
		2396
		2397	Calls the C<munlockall> function.
		2398
		2399	On systems that do not implement C<munlockall>, this function returns
		2400	ENOSYS, otherwise the return value of C<munlockall>.
		2401
		2402	=item IO::AIO::splice $r_fh, $r_off, $w_fh, $w_off, $length, $flags
		2403
		2404	Calls the GNU/Linux C<splice(2)> syscall, if available. If C<$r_off> or
		2405	C<$w_off> are C<undef>, then C<NULL> is passed for these, otherwise they
		2406	should be the file offset.
		2407
		2408	C<$r_fh> and C<$w_fh> should not refer to the same file, as splice might
		2409	silently corrupt the data in this case.
		2410
		2411	The following symbol flag values are available: C<IO::AIO::SPLICE_F_MOVE>,
		2412	C<IO::AIO::SPLICE_F_NONBLOCK>, C<IO::AIO::SPLICE_F_MORE> and
		2413	C<IO::AIO::SPLICE_F_GIFT>.
		2414
		2415	See the C<splice(2)> manpage for details.
		2416
		2417	=item IO::AIO::tee $r_fh, $w_fh, $length, $flags
		2418
		2419	Calls the GNU/Linux C<tee(2)> syscall, see its manpage and the
		2420	description for C<IO::AIO::splice> above for details.
		2421
		2422	=item $actual_size = IO::AIO::pipesize $r_fh[, $new_size]
		2423
		2424	Attempts to query or change the pipe buffer size. Obviously works only
		2425	on pipes, and currently works only on GNU/Linux systems, and fails with
		2426	C<-1>/C<ENOSYS> everywhere else. If anybody knows how to influence pipe buffer
		2427	size on other systems, drop me a note.
		2428
		2429	=item ($rfh, $wfh) = IO::AIO::pipe2 [$flags]
		2430
		2431	This is a direct interface to the Linux L<pipe2(2)> system call. If
		2432	C<$flags> is missing or C<0>, then this should be the same as a call to
		2433	perl's built-in C<pipe> function and create a new pipe, and works on
		2434	systems that lack the pipe2 syscall. On win32, this case invokes C<_pipe
		2435	(..., 4096, O_BINARY)>.
		2436
		2437	If C<$flags> is non-zero, it tries to invoke the pipe2 system call with
		2438	the given flags (Linux 2.6.27, glibc 2.9).
		2439
		2440	On success, the read and write file handles are returned.
		2441
		2442	On error, nothing will be returned. If the pipe2 syscall is missing and
		2443	C<$flags> is non-zero, fails with C<ENOSYS>.
		2444
		2445	Please refer to L<pipe2(2)> for more info on the C<$flags>, but at the
		2446	time of this writing, C<IO::AIO::O_CLOEXEC>, C<IO::AIO::O_NONBLOCK> and
		2447	C<IO::AIO::O_DIRECT> (Linux 3.4, for packet-based pipes) were supported.
		2448
		2449	Example: create a pipe race-free w.r.t. threads and fork:
		2450
		2451	my ($rfh, $wfh) = IO::AIO::pipe2 IO::AIO::O_CLOEXEC
		2452	or die "pipe2: $!\n";
		2453
		2454	=item $fh = IO::AIO::memfd_create $pathname[, $flags]
		2455
		2456	This is a direct interface to the Linux L<memfd_create(2)> system
		2457	call. The (unhelpful) default for C<$flags> is C<0>, but your default
		2458	should be C<IO::AIO::MFD_CLOEXEC>.
		2459
		2460	On success, the new memfd filehandle is returned, otherwise returns
		2461	C<undef>. If the memfd_create syscall is missing, fails with C<ENOSYS>.
		2462
		2463	Please refer to L<memfd_create(2)> for more info on this call.
		2464
		2465	The following C<$flags> values are available: C<IO::AIO::MFD_CLOEXEC>,
		2466	C<IO::AIO::MFD_ALLOW_SEALING> and C<IO::AIO::MFD_HUGETLB>.
		2467
		2468	Example: create a new memfd.
		2469
		2470	my $fh = IO::AIO::memfd_create "somenameforprocfd", IO::AIO::MFD_CLOEXEC
		2471	or die "m,emfd_create: $!\n";
		2472	=item $fh = IO::AIO::eventfd [$initval, [$flags]]
		2473
		2474	This is a direct interface to the Linux L<eventfd(2)> system call. The
		2475	(unhelpful) defaults for C<$initval> and C<$flags> are C<0> for both.
		2476
		2477	On success, the new eventfd filehandle is returned, otherwise returns
		2478	C<undef>. If the eventfd syscall is missing, fails with C<ENOSYS>.
		2479
		2480	Please refer to L<eventfd(2)> for more info on this call.
		2481
		2482	The following symbol flag values are available: C<IO::AIO::EFD_CLOEXEC>,
		2483	C<IO::AIO::EFD_NONBLOCK> and C<IO::AIO::EFD_SEMAPHORE> (Linux 2.6.30).
		2484
		2485	Example: create a new eventfd filehandle:
		2486
		2487	$fh = IO::AIO::eventfd 0, IO::AIO::EFD_CLOEXEC
		2488	or die "eventfd: $!\n";
		2489
		2490	=item $fh = IO::AIO::timerfd_create $clockid[, $flags]
		2491
		2492	This is a direct interface to the Linux L<timerfd_create(2)> system
		2493	call. The (unhelpful) default for C<$flags> is C<0>, but your default
		2494	should be C<IO::AIO::TFD_CLOEXEC>.
		2495
		2496	On success, the new timerfd filehandle is returned, otherwise returns
		2497	C<undef>. If the timerfd_create syscall is missing, fails with C<ENOSYS>.
		2498
		2499	Please refer to L<timerfd_create(2)> for more info on this call.
		2500
		2501	The following C<$clockid> values are
		2502	available: C<IO::AIO::CLOCK_REALTIME>, C<IO::AIO::CLOCK_MONOTONIC>
		2503	C<IO::AIO::CLOCK_CLOCK_BOOTTIME> (Linux 3.15)
		2504	C<IO::AIO::CLOCK_CLOCK_REALTIME_ALARM> (Linux 3.11) and
		2505	C<IO::AIO::CLOCK_CLOCK_BOOTTIME_ALARM> (Linux 3.11).
		2506
		2507	The following C<$flags> values are available (Linux
		2508	2.6.27): C<IO::AIO::TFD_NONBLOCK> and C<IO::AIO::TFD_CLOEXEC>.
		2509
		2510	Example: create a new timerfd and set it to one-second repeated alarms,
		2511	then wait for two alarms:
		2512
		2513	my $fh = IO::AIO::timerfd_create IO::AIO::CLOCK_BOOTTIME, IO::AIO::TFD_CLOEXEC
		2514	or die "timerfd_create: $!\n";
		2515
		2516	defined IO::AIO::timerfd_settime $fh, 0, 1, 1
		2517	or die "timerfd_settime: $!\n";
		2518
		2519	for (1..2) {
		2520	8 == sysread $fh, my $buf, 8
		2521	or die "timerfd read failure\n";
		2522
		2523	printf "number of expirations (likely 1): %d\n",
		2524	unpack "Q", $buf;
		2525	}
		2526
		2527	=item ($cur_interval, $cur_value) = IO::AIO::timerfd_settime $fh, $flags, $new_interval, $nbw_value
		2528
		2529	This is a direct interface to the Linux L<timerfd_settime(2)> system
		2530	call. Please refer to its manpage for more info on this call.
		2531
		2532	The new itimerspec is specified using two (possibly fractional) second
		2533	values, C<$new_interval> and C<$new_value>).
		2534
		2535	On success, the current interval and value are returned (as per
		2536	C<timerfd_gettime>). On failure, the empty list is returned.
		2537
		2538	The following C<$flags> values are
		2539	available: C<IO::AIO::TFD_TIMER_ABSTIME> and
		2540	C<IO::AIO::TFD_TIMER_CANCEL_ON_SET>.
		2541
		2542	See C<IO::AIO::timerfd_create> for a full example.
		2543
		2544	=item ($cur_interval, $cur_value) = IO::AIO::timerfd_gettime $fh
		2545
		2546	This is a direct interface to the Linux L<timerfd_gettime(2)> system
		2547	call. Please refer to its manpage for more info on this call.
		2548
		2549	On success, returns the current values of interval and value for the given
		2550	timerfd (as potentially fractional second values). On failure, the empty
		2551	list is returned.
		2552
		2553	=back
		2554
296	=cut	2555	=cut
297		2556
298	# support function to convert a fd into a perl filehandle
299	sub _fd2fh {
300	return undef if $_[0] < 0;
301
302	# try to be perl5.6-compatible
303	local *AIO_FH;
304	open AIO_FH, "+<&=$_[0]"
305	or return undef;
306
307	*AIO_FH
308	}
309
310	min_parallel 4;	2557	min_parallel 8;
311		2558
312	END {	2559	END { flush }
313	max_parallel 0;
314	}
315		2560
316	1;	2561	1;
317		2562
		2563	=head1 EVENT LOOP INTEGRATION
		2564
		2565	It is recommended to use L<AnyEvent::AIO> to integrate IO::AIO
		2566	automatically into many event loops:
		2567
		2568	# AnyEvent integration (EV, Event, Glib, Tk, POE, urxvt, pureperl...)
		2569	use AnyEvent::AIO;
		2570
		2571	You can also integrate IO::AIO manually into many event loops, here are
		2572	some examples of how to do this:
		2573
		2574	# EV integration
		2575	my $aio_w = EV::io IO::AIO::poll_fileno, EV::READ, \&IO::AIO::poll_cb;
		2576
		2577	# Event integration
		2578	Event->io (fd => IO::AIO::poll_fileno,
		2579	poll => 'r',
		2580	cb => \&IO::AIO::poll_cb);
		2581
		2582	# Glib/Gtk2 integration
		2583	add_watch Glib::IO IO::AIO::poll_fileno,
		2584	in => sub { IO::AIO::poll_cb; 1 };
		2585
		2586	# Tk integration
		2587	Tk::Event::IO->fileevent (IO::AIO::poll_fileno, "",
		2588	readable => \&IO::AIO::poll_cb);
		2589
		2590	# Danga::Socket integration
		2591	Danga::Socket->AddOtherFds (IO::AIO::poll_fileno =>
		2592	\&IO::AIO::poll_cb);
		2593
		2594	=head2 FORK BEHAVIOUR
		2595
		2596	Usage of pthreads in a program changes the semantics of fork
		2597	considerably. Specifically, only async-safe functions can be called after
		2598	fork. Perl doesn't know about this, so in general, you cannot call fork
		2599	with defined behaviour in perl if pthreads are involved. IO::AIO uses
		2600	pthreads, so this applies, but many other extensions and (for inexplicable
		2601	reasons) perl itself often is linked against pthreads, so this limitation
		2602	applies to quite a lot of perls.
		2603
		2604	This module no longer tries to fight your OS, or POSIX. That means IO::AIO
		2605	only works in the process that loaded it. Forking is fully supported, but
		2606	using IO::AIO in the child is not.
		2607
		2608	You might get around by not I<using> IO::AIO before (or after)
		2609	forking. You could also try to call the L<IO::AIO::reinit> function in the
		2610	child:
		2611
		2612	=over 4
		2613
		2614	=item IO::AIO::reinit
		2615
		2616	Abandons all current requests and I/O threads and simply reinitialises all
		2617	data structures. This is not an operation supported by any standards, but
		2618	happens to work on GNU/Linux and some newer BSD systems.
		2619
		2620	The only reasonable use for this function is to call it after forking, if
		2621	C<IO::AIO> was used in the parent. Calling it while IO::AIO is active in
		2622	the process will result in undefined behaviour. Calling it at any time
		2623	will also result in any undefined (by POSIX) behaviour.
		2624
		2625	=back
		2626
		2627	=head2 LINUX-SPECIFIC CALLS
		2628
		2629	When a call is documented as "linux-specific" then this means it
		2630	originated on GNU/Linux. C<IO::AIO> will usually try to autodetect the
		2631	availability and compatibility of such calls regardless of the platform
		2632	it is compiled on, so platforms such as FreeBSD which often implement
		2633	these calls will work. When in doubt, call them and see if they fail wth
		2634	C<ENOSYS>.
		2635
		2636	=head2 MEMORY USAGE
		2637
		2638	Per-request usage:
		2639
		2640	Each aio request uses - depending on your architecture - around 100-200
		2641	bytes of memory. In addition, stat requests need a stat buffer (possibly
		2642	a few hundred bytes), readdir requires a result buffer and so on. Perl
		2643	scalars and other data passed into aio requests will also be locked and
		2644	will consume memory till the request has entered the done state.
		2645
		2646	This is not awfully much, so queuing lots of requests is not usually a
		2647	problem.
		2648
		2649	Per-thread usage:
		2650
		2651	In the execution phase, some aio requests require more memory for
		2652	temporary buffers, and each thread requires a stack and other data
		2653	structures (usually around 16k-128k, depending on the OS).
		2654
		2655	=head1 KNOWN BUGS
		2656
		2657	Known bugs will be fixed in the next release :)
		2658
		2659	=head1 KNOWN ISSUES
		2660
		2661	Calls that try to "import" foreign memory areas (such as C<IO::AIO::mmap>
		2662	or C<IO::AIO::aio_slurp>) do not work with generic lvalues, such as
		2663	non-created hash slots or other scalars I didn't think of. It's best to
		2664	avoid such and either use scalar variables or making sure that the scalar
		2665	exists (e.g. by storing C<undef>) and isn't "funny" (e.g. tied).
		2666
		2667	I am not sure anything can be done about this, so this is considered a
		2668	known issue, rather than a bug.
		2669
318	=head1 SEE ALSO	2670	=head1 SEE ALSO
319		2671
320	L<Coro>, L<Linux::AIO>.	2672	L<AnyEvent::AIO> for easy integration into event loops, L<Coro::AIO> for a
		2673	more natural syntax.
321		2674
322	=head1 AUTHOR	2675	=head1 AUTHOR
323		2676
324	Marc Lehmann <schmorp@schmorp.de>	2677	Marc Lehmann <schmorp@schmorp.de>
325	http://home.schmorp.de/	2678	http://home.schmorp.de/

Diff Legend

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