ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/IO-AIO/AIO.pm

Comparing IO-AIO/AIO.pm (file contents):
Revision 1.21 by root, Wed Jul 13 00:13:09 2005 UTC vs.
Revision 1.315 by root, Mon Sep 5 00:03:32 2022 UTC

1	=head1 NAME	1	=head1 NAME
2		2
3	IO::AIO - Asynchronous Input/Output	3	IO::AIO - Asynchronous/Advanced Input/Output
4		4
5	=head1 SYNOPSIS	5	=head1 SYNOPSIS
6		6
7	use IO::AIO;	7	use IO::AIO;
8		8
9	aio_open "/etc/passwd", O_RDONLY, 0, sub {	9	aio_open "/etc/passwd", IO::AIO::O_RDONLY, 0, sub {
10	my ($fh) = @_;	10	my $fh = shift
		11	or die "/etc/passwd: $!";
11	...	12	...
12	};	13	};
13		14
14	aio_unlink "/tmp/file", sub { };	15	aio_unlink "/tmp/file", sub { };
15		16
16	aio_read $fh, 30000, 1024, $buffer, 0, sub {	17	aio_read $fh, 30000, 1024, $buffer, 0, sub {
17	$_[0] > 0 or die "read error: $!";	18	$_[0] > 0 or die "read error: $!";
18	};	19	};
19		20
20	# Event	21	# version 2+ has request and group objects
21	Event->io (fd => IO::AIO::poll_fileno,	22	use IO::AIO 2;
22	poll => 'r',
23	cb => \&IO::AIO::poll_cb);
24		23
25	# Glib/Gtk2	24	aioreq_pri 4; # give next request a very high priority
26	add_watch Glib::IO IO::AIO::poll_fileno,	25	my $req = aio_unlink "/tmp/file", sub { };
27	in => sub { IO::AIO::poll_cb, 1 };	26	$req->cancel; # cancel request if still in queue
28		27
29	# Tk	28	my $grp = aio_group sub { print "all stats done\n" };
30	Tk::Event::IO->fileevent (IO::AIO::poll_fileno, "",	29	add $grp aio_stat "..." for ...;
31	readable => \&IO::AIO::poll_cb);
32
33	# Danga::Socket
34	Danga::Socket->AddOtherFds (IO::AIO::poll_fileno =>
35	\&IO::AIO::poll_cb);
36
37		30
38	=head1 DESCRIPTION	31	=head1 DESCRIPTION
39		32
40	This module implements asynchronous I/O using whatever means your	33	This module implements asynchronous I/O using whatever means your
41	operating system supports.	34	operating system supports. It is implemented as an interface to C<libeio>
		35	(L<http://software.schmorp.de/pkg/libeio.html>).
42		36
		37	Asynchronous means that operations that can normally block your program
		38	(e.g. reading from disk) will be done asynchronously: the operation
		39	will still block, but you can do something else in the meantime. This
		40	is extremely useful for programs that need to stay interactive even
		41	when doing heavy I/O (GUI programs, high performance network servers
		42	etc.), but can also be used to easily do operations in parallel that are
		43	normally done sequentially, e.g. stat'ing many files, which is much faster
		44	on a RAID volume or over NFS when you do a number of stat operations
		45	concurrently.
		46
		47	While most of this works on all types of file descriptors (for
		48	example sockets), using these functions on file descriptors that
		49	support nonblocking operation (again, sockets, pipes etc.) is
		50	very inefficient. Use an event loop for that (such as the L<EV>
		51	module): IO::AIO will naturally fit into such an event loop itself.
		52
43	Currently, a number of threads are started that execute your read/writes	53	In this version, a number of threads are started that execute your
44	and signal their completion. You don't need thread support in your libc or	54	requests and signal their completion. You don't need thread support
45	perl, and the threads created by this module will not be visible to the	55	in perl, and the threads created by this module will not be visible
46	pthreads library. In the future, this module might make use of the native	56	to perl. In the future, this module might make use of the native aio
47	aio functions available on many operating systems. However, they are often	57	functions available on many operating systems. However, they are often
48	not well-supported (Linux doesn't allow them on normal files currently,	58	not well-supported or restricted (GNU/Linux doesn't allow them on normal
49	for example), and they would only support aio_read and aio_write, so the	59	files currently, for example), and they would only support aio_read and
50	remaining functionality would have to be implemented using threads anyway.	60	aio_write, so the remaining functionality would have to be implemented
		61	using threads anyway.
51		62
		63	In addition to asynchronous I/O, this module also exports some rather
		64	arcane interfaces, such as C<madvise> or linux's C<splice> system call,
		65	which is why the C<A> in C<AIO> can also mean I<advanced>.
		66
52	Although the module will work with in the presence of other threads, it is	67	Although the module will work in the presence of other (Perl-) threads,
53	currently not reentrant, so use appropriate locking yourself.	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 = 0.9;	176	our $VERSION = 4.77;
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	accept4 tee splice pipe2 pipesize
		200	fexecve memfd_create eventfd
		201	timerfd_create timerfd_settime timerfd_gettime
		202	pidfd_open pidfd_send_signal pidfd_getfd);
		203
		204	push @AIO_REQ, qw(aio_busy); # not exported
		205
		206	@IO::AIO::GRP::ISA = 'IO::AIO::REQ';
69		207
70	require XSLoader;	208	require XSLoader;
71	XSLoader::load IO::AIO, $VERSION;	209	XSLoader::load ("IO::AIO", $VERSION);
72	}	210	}
73		211
74	=head1 FUNCTIONS	212	=head1 FUNCTIONS
75		213
76	=head2 AIO FUNCTIONS	214	=head2 QUICK OVERVIEW
		215
		216	This section simply lists the prototypes most of the functions for
		217	quick reference. See the following sections for function-by-function
		218	documentation.
		219
		220	aio_wd $pathname, $callback->($wd)
		221	aio_open $pathname, $flags, $mode, $callback->($fh)
		222	aio_close $fh, $callback->($status)
		223	aio_seek $fh,$offset,$whence, $callback->($offs)
		224	aio_read $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
		225	aio_write $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
		226	aio_sendfile $out_fh, $in_fh, $in_offset, $length, $callback->($retval)
		227	aio_readahead $fh,$offset,$length, $callback->($retval)
		228	aio_stat $fh_or_path, $callback->($status)
		229	aio_lstat $fh, $callback->($status)
		230	aio_statvfs $fh_or_path, $callback->($statvfs)
		231	aio_utime $fh_or_path, $atime, $mtime, $callback->($status)
		232	aio_chown $fh_or_path, $uid, $gid, $callback->($status)
		233	aio_chmod $fh_or_path, $mode, $callback->($status)
		234	aio_truncate $fh_or_path, $offset, $callback->($status)
		235	aio_allocate $fh, $mode, $offset, $len, $callback->($status)
		236	aio_fiemap $fh, $start, $length, $flags, $count, $cb->(\@extents)
		237	aio_unlink $pathname, $callback->($status)
		238	aio_mknod $pathname, $mode, $dev, $callback->($status)
		239	aio_link $srcpath, $dstpath, $callback->($status)
		240	aio_symlink $srcpath, $dstpath, $callback->($status)
		241	aio_readlink $pathname, $callback->($link)
		242	aio_realpath $pathname, $callback->($path)
		243	aio_rename $srcpath, $dstpath, $callback->($status)
		244	aio_rename2 $srcpath, $dstpath, $flags, $callback->($status)
		245	aio_mkdir $pathname, $mode, $callback->($status)
		246	aio_rmdir $pathname, $callback->($status)
		247	aio_readdir $pathname, $callback->($entries)
		248	aio_readdirx $pathname, $flags, $callback->($entries, $flags)
		249	IO::AIO::READDIR_DENTS IO::AIO::READDIR_DIRS_FIRST
		250	IO::AIO::READDIR_STAT_ORDER IO::AIO::READDIR_FOUND_UNKNOWN
		251	aio_scandir $pathname, $maxreq, $callback->($dirs, $nondirs)
		252	aio_load $pathname, $data, $callback->($status)
		253	aio_copy $srcpath, $dstpath, $callback->($status)
		254	aio_move $srcpath, $dstpath, $callback->($status)
		255	aio_rmtree $pathname, $callback->($status)
		256	aio_fcntl $fh, $cmd, $arg, $callback->($status)
		257	aio_ioctl $fh, $request, $buf, $callback->($status)
		258	aio_sync $callback->($status)
		259	aio_syncfs $fh, $callback->($status)
		260	aio_fsync $fh, $callback->($status)
		261	aio_fdatasync $fh, $callback->($status)
		262	aio_sync_file_range $fh, $offset, $nbytes, $flags, $callback->($status)
		263	aio_pathsync $pathname, $callback->($status)
		264	aio_msync $scalar, $offset = 0, $length = undef, flags = MS_SYNC, $callback->($status)
		265	aio_mtouch $scalar, $offset = 0, $length = undef, flags = 0, $callback->($status)
		266	aio_mlock $scalar, $offset = 0, $length = undef, $callback->($status)
		267	aio_mlockall $flags, $callback->($status)
		268	aio_group $callback->(...)
		269	aio_nop $callback->()
		270
		271	$prev_pri = aioreq_pri [$pri]
		272	aioreq_nice $pri_adjust
		273
		274	IO::AIO::poll_wait
		275	IO::AIO::poll_cb
		276	IO::AIO::poll
		277	IO::AIO::flush
		278	IO::AIO::max_poll_reqs $nreqs
		279	IO::AIO::max_poll_time $seconds
		280	IO::AIO::min_parallel $nthreads
		281	IO::AIO::max_parallel $nthreads
		282	IO::AIO::max_idle $nthreads
		283	IO::AIO::idle_timeout $seconds
		284	IO::AIO::max_outstanding $maxreqs
		285	IO::AIO::nreqs
		286	IO::AIO::nready
		287	IO::AIO::npending
		288	IO::AIO::reinit
		289
		290	$nfd = IO::AIO::get_fdlimit
		291	IO::AIO::min_fdlimit $nfd
		292
		293	IO::AIO::sendfile $ofh, $ifh, $offset, $count
		294	IO::AIO::fadvise $fh, $offset, $len, $advice
		295	IO::AIO::fexecve $fh, $argv, $envp
		296
		297	IO::AIO::mmap $scalar, $length, $prot, $flags[, $fh[, $offset]]
		298	IO::AIO::munmap $scalar
		299	IO::AIO::mremap $scalar, $new_length, $flags[, $new_address]
		300	IO::AIO::madvise $scalar, $offset, $length, $advice
		301	IO::AIO::mprotect $scalar, $offset, $length, $protect
		302	IO::AIO::munlock $scalar, $offset = 0, $length = undef
		303	IO::AIO::munlockall
		304
		305	# stat extensions
		306	$counter = IO::AIO::st_gen
		307	$seconds = IO::AIO::st_atime, IO::AIO::st_mtime, IO::AIO::st_ctime, IO::AIO::st_btime
		308	($atime, $mtime, $ctime, $btime, ...) = IO::AIO::st_xtime
		309	$nanoseconds = IO::AIO::st_atimensec, IO::AIO::st_mtimensec, IO::AIO::st_ctimensec, IO::AIO::st_btimensec
		310	$seconds = IO::AIO::st_btimesec
		311	($atime, $mtime, $ctime, $btime, ...) = IO::AIO::st_xtimensec
		312
		313	# very much unportable syscalls
		314	IO::AIO::accept4 $r_fh, $sockaddr, $sockaddr_len, $flags
		315	IO::AIO::splice $r_fh, $r_off, $w_fh, $w_off, $length, $flags
		316	IO::AIO::tee $r_fh, $w_fh, $length, $flags
		317
		318	$actual_size = IO::AIO::pipesize $r_fh[, $new_size]
		319	($rfh, $wfh) = IO::AIO::pipe2 [$flags]
		320
		321	$fh = IO::AIO::eventfd [$initval, [$flags]]
		322	$fh = IO::AIO::memfd_create $pathname[, $flags]
		323
		324	$fh = IO::AIO::timerfd_create $clockid[, $flags]
		325	($cur_interval, $cur_value) = IO::AIO::timerfd_settime $fh, $flags, $new_interval, $nbw_value
		326	($cur_interval, $cur_value) = IO::AIO::timerfd_gettime $fh
		327
		328	$fh = IO::AIO::pidfd_open $pid[, $flags]
		329	$status = IO::AIO::pidfd_send_signal $pidfh, $signal[, $siginfo[, $flags]]
		330	$fh = IO::AIO::pidfd_getfd $pidfh, $targetfd[, $flags]
		331
		332	=head2 API NOTES
77		333
78	All the C<aio_*> calls are more or less thin wrappers around the syscall	334	All the C<aio_*> calls are more or less thin wrappers around the syscall
79	with the same name (sans C<aio_>). The arguments are similar or identical,	335	with the same name (sans C<aio_>). The arguments are similar or identical,
80	and they all accept an additional (and optional) C<$callback> argument	336	and they all accept an additional (and optional) C<$callback> argument
81	which must be a code reference. This code reference will get called with	337	which must be a code reference. This code reference will be called after
82	the syscall return code (e.g. most syscalls return C<-1> on error, unlike	338	the syscall has been executed in an asynchronous fashion. The results
83	perl, which usually delivers "false") as it's sole argument when the given	339	of the request will be passed as arguments to the callback (and, if an
84	syscall has been executed asynchronously.	340	error occured, in C<$!>) - for most requests the syscall return code (e.g.
		341	most syscalls return C<-1> on error, unlike perl, which usually delivers
		342	"false").
85		343
86	All functions that expect a filehandle will also accept a file descriptor.	344	Some requests (such as C<aio_readdir>) pass the actual results and
		345	communicate failures by passing C<undef>.
87		346
		347	All functions expecting a filehandle keep a copy of the filehandle
		348	internally until the request has finished.
		349
		350	All functions return request objects of type L<IO::AIO::REQ> that allow
		351	further manipulation of those requests while they are in-flight.
		352
88	The filenames you pass to these routines I<must> be absolute. The reason	353	The pathnames you pass to these routines I<should> be absolute. The
89	for this is that at the time the request is being executed, the current	354	reason for this is that at the time the request is being executed, the
90	working directory could have changed. Alternatively, you can make sure	355	current working directory could have changed. Alternatively, you can
91	that you never change the current working directory.	356	make sure that you never change the current working directory anywhere
		357	in the program and then use relative paths. You can also take advantage
		358	of IO::AIOs working directory abstraction, that lets you specify paths
		359	relative to some previously-opened "working directory object" - see the
		360	description of the C<IO::AIO::WD> class later in this document.
		361
		362	To encode pathnames as octets, either make sure you either: a) always pass
		363	in filenames you got from outside (command line, readdir etc.) without
		364	tinkering, b) are in your native filesystem encoding, c) use the Encode
		365	module and encode your pathnames to the locale (or other) encoding in
		366	effect in the user environment, d) use Glib::filename_from_unicode on
		367	unicode filenames or e) use something else to ensure your scalar has the
		368	correct contents.
		369
		370	This works, btw. independent of the internal UTF-8 bit, which IO::AIO
		371	handles correctly whether it is set or not.
		372
		373	=head2 AIO REQUEST FUNCTIONS
92		374
93	=over 4	375	=over 4
94		376
		377	=item $prev_pri = aioreq_pri [$pri]
		378
		379	Returns the priority value that would be used for the next request and, if
		380	C<$pri> is given, sets the priority for the next aio request.
		381
		382	The default priority is C<0>, the minimum and maximum priorities are C<-4>
		383	and C<4>, respectively. Requests with higher priority will be serviced
		384	first.
		385
		386	The priority will be reset to C<0> after each call to one of the C<aio_*>
		387	functions.
		388
		389	Example: open a file with low priority, then read something from it with
		390	higher priority so the read request is serviced before other low priority
		391	open requests (potentially spamming the cache):
		392
		393	aioreq_pri -3;
		394	aio_open ..., sub {
		395	return unless $_[0];
		396
		397	aioreq_pri -2;
		398	aio_read $_[0], ..., sub {
		399	...
		400	};
		401	};
		402
		403
		404	=item aioreq_nice $pri_adjust
		405
		406	Similar to C<aioreq_pri>, but subtracts the given value from the current
		407	priority, so the effect is cumulative.
		408
		409
95	=item aio_open $pathname, $flags, $mode, $callback	410	=item aio_open $pathname, $flags, $mode, $callback->($fh)
96		411
97	Asynchronously open or create a file and call the callback with a newly	412	Asynchronously open or create a file and call the callback with a newly
98	created filehandle for the file.	413	created filehandle for the file (or C<undef> in case of an error).
99		414
100	The pathname passed to C<aio_open> must be absolute. See API NOTES, above,	415	The pathname passed to C<aio_open> must be absolute. See API NOTES, above,
101	for an explanation.	416	for an explanation.
102		417
103	The C<$flags> argument is a bitmask. See the C<Fcntl> module for a	418	The C<$flags> argument is a bitmask. See the C<Fcntl> module for a
104	list. They are the same as used by C<sysopen>.	419	list. They are the same as used by C<sysopen>.
105		420
106	Likewise, C<$mode> specifies the mode of the newly created file, if it	421	Likewise, C<$mode> specifies the mode of the newly created file, if it
107	didn't exist and C<O_CREAT> has been given, just like perl's C<sysopen>,	422	didn't exist and C<O_CREAT> has been given, just like perl's C<sysopen>,
108	except that it is mandatory (i.e. use C<0> if you don't create new files,	423	except that it is mandatory (i.e. use C<0> if you don't create new files,
109	and C<0666> or C<0777> if you do).	424	and C<0666> or C<0777> if you do). Note that the C<$mode> will be modified
		425	by the umask in effect then the request is being executed, so better never
		426	change the umask.
110		427
111	Example:	428	Example:
112		429
113	aio_open "/etc/passwd", O_RDONLY, 0, sub {	430	aio_open "/etc/passwd", IO::AIO::O_RDONLY, 0, sub {
114	if ($_[0]) {	431	if ($_[0]) {
115	print "open successful, fh is $_[0]\n";	432	print "open successful, fh is $_[0]\n";
116	...	433	...
117	} else {	434	} else {
118	die "open failed: $!\n";	435	die "open failed: $!\n";
119	}	436	}
120	};	437	};
121		438
		439	In addition to all the common open modes/flags (C<O_RDONLY>, C<O_WRONLY>,
		440	C<O_RDWR>, C<O_CREAT>, C<O_TRUNC>, C<O_EXCL> and C<O_APPEND>), the
		441	following POSIX and non-POSIX constants are available (missing ones on
		442	your system are, as usual, C<0>):
		443
		444	C<O_ASYNC>, C<O_DIRECT>, C<O_NOATIME>, C<O_CLOEXEC>, C<O_NOCTTY>, C<O_NOFOLLOW>,
		445	C<O_NONBLOCK>, C<O_EXEC>, C<O_SEARCH>, C<O_DIRECTORY>, C<O_DSYNC>,
		446	C<O_RSYNC>, C<O_SYNC>, C<O_PATH>, C<O_TMPFILE>, C<O_TTY_INIT> and C<O_ACCMODE>.
		447
		448
122	=item aio_close $fh, $callback	449	=item aio_close $fh, $callback->($status)
123		450
124	Asynchronously close a file and call the callback with the result	451	Asynchronously close a file and call the callback with the result
125	code. I<WARNING:> although accepted, you should not pass in a perl	452	code.
126	filehandle here, as perl will likely close the file descriptor another
127	time when the filehandle is destroyed. Normally, you can safely call perls
128	C<close> or just let filehandles go out of scope.
129		453
130	This is supposed to be a bug in the API, so that might change. It's	454	Unfortunately, you can't do this to perl. Perl I<insists> very strongly on
131	therefore best to avoid this function.	455	closing the file descriptor associated with the filehandle itself.
132		456
		457	Therefore, C<aio_close> will not close the filehandle - instead it will
		458	use dup2 to overwrite the file descriptor with the write-end of a pipe
		459	(the pipe fd will be created on demand and will be cached).
		460
		461	Or in other words: the file descriptor will be closed, but it will not be
		462	free for reuse until the perl filehandle is closed.
		463
		464	=cut
		465
		466	=item aio_seek $fh, $offset, $whence, $callback->($offs)
		467
		468	Seeks the filehandle to the new C<$offset>, similarly to perl's
		469	C<sysseek>. The C<$whence> can use the traditional values (C<0> for
		470	C<IO::AIO::SEEK_SET>, C<1> for C<IO::AIO::SEEK_CUR> or C<2> for
		471	C<IO::AIO::SEEK_END>).
		472
		473	The resulting absolute offset will be passed to the callback, or C<-1> in
		474	case of an error.
		475
		476	In theory, the C<$whence> constants could be different than the
		477	corresponding values from L<Fcntl>, but perl guarantees they are the same,
		478	so don't panic.
		479
		480	As a GNU/Linux (and maybe Solaris) extension, also the constants
		481	C<IO::AIO::SEEK_DATA> and C<IO::AIO::SEEK_HOLE> are available, if they
		482	could be found. No guarantees about suitability for use in C<aio_seek> or
		483	Perl's C<sysseek> can be made though, although I would naively assume they
		484	"just work".
		485
133	=item aio_read $fh,$offset,$length, $data,$dataoffset,$callback	486	=item aio_read $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
134		487
135	=item aio_write $fh,$offset,$length, $data,$dataoffset,$callback	488	=item aio_write $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
136		489
137	Reads or writes C<length> bytes from the specified C<fh> and C<offset>	490	Reads or writes C<$length> bytes from or to the specified C<$fh> and
138	into the scalar given by C<data> and offset C<dataoffset> and calls the	491	C<$offset> into the scalar given by C<$data> and offset C<$dataoffset> and
139	callback without the actual number of bytes read (or -1 on error, just	492	calls the callback with the actual number of bytes transferred (or -1 on
140	like the syscall).	493	error, just like the syscall).
		494
		495	C<aio_read> will, like C<sysread>, shrink or grow the C<$data> scalar to
		496	offset plus the actual number of bytes read.
		497
		498	If C<$offset> is undefined, then the current file descriptor offset will
		499	be used (and updated), otherwise the file descriptor offset will not be
		500	changed by these calls.
		501
		502	If C<$length> is undefined in C<aio_write>, use the remaining length of
		503	C<$data>.
		504
		505	If C<$dataoffset> is less than zero, it will be counted from the end of
		506	C<$data>.
		507
		508	The C<$data> scalar I<MUST NOT> be modified in any way while the request
		509	is outstanding. Modifying it can result in segfaults or World War III (if
		510	the necessary/optional hardware is installed).
141		511
142	Example: Read 15 bytes at offset 7 into scalar C<$buffer>, starting at	512	Example: Read 15 bytes at offset 7 into scalar C<$buffer>, starting at
143	offset C<0> within the scalar:	513	offset C<0> within the scalar:
144		514
145	aio_read $fh, 7, 15, $buffer, 0, sub {	515	aio_read $fh, 7, 15, $buffer, 0, sub {
146	$_[0] > 0 or die "read error: $!";	516	$_[0] > 0 or die "read error: $!";
147	print "read $_[0] bytes: <$buffer>\n";	517	print "read $_[0] bytes: <$buffer>\n";
148	};	518	};
149		519
		520
		521	=item aio_sendfile $out_fh, $in_fh, $in_offset, $length, $callback->($retval)
		522
		523	Tries to copy C<$length> bytes from C<$in_fh> to C<$out_fh>. It starts
		524	reading at byte offset C<$in_offset>, and starts writing at the current
		525	file offset of C<$out_fh>. Because of that, it is not safe to issue more
		526	than one C<aio_sendfile> per C<$out_fh>, as they will interfere with each
		527	other. The same C<$in_fh> works fine though, as this function does not
		528	move or use the file offset of C<$in_fh>.
		529
		530	Please note that C<aio_sendfile> can read more bytes from C<$in_fh> than
		531	are written, and there is no way to find out how many more bytes have been
		532	read from C<aio_sendfile> alone, as C<aio_sendfile> only provides the
		533	number of bytes written to C<$out_fh>. Only if the result value equals
		534	C<$length> one can assume that C<$length> bytes have been read.
		535
		536	Unlike with other C<aio_> functions, it makes a lot of sense to use
		537	C<aio_sendfile> on non-blocking sockets, as long as one end (typically
		538	the C<$in_fh>) is a file - the file I/O will then be asynchronous, while
		539	the socket I/O will be non-blocking. Note, however, that you can run
		540	into a trap where C<aio_sendfile> reads some data with readahead, then
		541	fails to write all data, and when the socket is ready the next time, the
		542	data in the cache is already lost, forcing C<aio_sendfile> to again hit
		543	the disk. Explicit C<aio_read> + C<aio_write> let's you better control
		544	resource usage.
		545
		546	This call tries to make use of a native C<sendfile>-like syscall to
		547	provide zero-copy operation. For this to work, C<$out_fh> should refer to
		548	a socket, and C<$in_fh> should refer to an mmap'able file.
		549
		550	If a native sendfile cannot be found or it fails with C<ENOSYS>,
		551	C<EINVAL>, C<ENOTSUP>, C<EOPNOTSUPP>, C<EAFNOSUPPORT>, C<EPROTOTYPE> or
		552	C<ENOTSOCK>, it will be emulated, so you can call C<aio_sendfile> on any
		553	type of filehandle regardless of the limitations of the operating system.
		554
		555	As native sendfile syscalls (as practically any non-POSIX interface hacked
		556	together in a hurry to improve benchmark numbers) tend to be rather buggy
		557	on many systems, this implementation tries to work around some known bugs
		558	in Linux and FreeBSD kernels (probably others, too), but that might fail,
		559	so you really really should check the return value of C<aio_sendfile> -
		560	fewer bytes than expected might have been transferred.
		561
		562
150	=item aio_readahead $fh,$offset,$length, $callback	563	=item aio_readahead $fh,$offset,$length, $callback->($retval)
151
152	Asynchronously reads the specified byte range into the page cache, using
153	the C<readahead> syscall. If that syscall doesn't exist (likely if your OS
154	isn't Linux) the status will be C<-1> and C<$!> is set to C<ENOSYS>.
155		564
156	C<aio_readahead> populates the page cache with data from a file so that	565	C<aio_readahead> populates the page cache with data from a file so that
157	subsequent reads from that file will not block on disk I/O. The C<$offset>	566	subsequent reads from that file will not block on disk I/O. The C<$offset>
158	argument specifies the starting point from which data is to be read and	567	argument specifies the starting point from which data is to be read and
159	C<$length> specifies the number of bytes to be read. I/O is performed in	568	C<$length> specifies the number of bytes to be read. I/O is performed in
160	whole pages, so that offset is effectively rounded down to a page boundary	569	whole pages, so that offset is effectively rounded down to a page boundary
161	and bytes are read up to the next page boundary greater than or equal to	570	and bytes are read up to the next page boundary greater than or equal to
162	(off-set+length). C<aio_readahead> does not read beyond the end of the	571	(off-set+length). C<aio_readahead> does not read beyond the end of the
163	file. The current file offset of the file is left unchanged.	572	file. The current file offset of the file is left unchanged.
164		573
		574	If that syscall doesn't exist (likely if your kernel isn't Linux) it will
		575	be emulated by simply reading the data, which would have a similar effect.
		576
		577
165	=item aio_stat $fh_or_path, $callback	578	=item aio_stat $fh_or_path, $callback->($status)
166		579
167	=item aio_lstat $fh, $callback	580	=item aio_lstat $fh, $callback->($status)
168		581
169	Works like perl's C<stat> or C<lstat> in void context. The callback will	582	Works almost exactly like perl's C<stat> or C<lstat> in void context. The
170	be called after the stat and the results will be available using C<stat _>	583	callback will be called after the stat and the results will be available
171	or C<-s _> etc...	584	using C<stat _> or C<-s _> and other tests (with the exception of C<-B>
		585	and C<-T>).
172		586
173	The pathname passed to C<aio_stat> must be absolute. See API NOTES, above,	587	The pathname passed to C<aio_stat> must be absolute. See API NOTES, above,
174	for an explanation.	588	for an explanation.
175		589
176	Currently, the stats are always 64-bit-stats, i.e. instead of returning an	590	Currently, the stats are always 64-bit-stats, i.e. instead of returning an
177	error when stat'ing a large file, the results will be silently truncated	591	error when stat'ing a large file, the results will be silently truncated
178	unless perl itself is compiled with large file support.	592	unless perl itself is compiled with large file support.
		593
		594	To help interpret the mode and dev/rdev stat values, IO::AIO offers the
		595	following constants and functions (if not implemented, the constants will
		596	be C<0> and the functions will either C<croak> or fall back on traditional
		597	behaviour).
		598
		599	C<S_IFMT>, C<S_IFIFO>, C<S_IFCHR>, C<S_IFBLK>, C<S_IFLNK>, C<S_IFREG>,
		600	C<S_IFDIR>, C<S_IFWHT>, C<S_IFSOCK>, C<IO::AIO::major $dev_t>,
		601	C<IO::AIO::minor $dev_t>, C<IO::AIO::makedev $major, $minor>.
		602
		603	To access higher resolution stat timestamps, see L<SUBSECOND STAT TIME
		604	ACCESS>.
179		605
180	Example: Print the length of F</etc/passwd>:	606	Example: Print the length of F</etc/passwd>:
181		607
182	aio_stat "/etc/passwd", sub {	608	aio_stat "/etc/passwd", sub {
183	$_[0] and die "stat failed: $!";	609	$_[0] and die "stat failed: $!";
184	print "size is ", -s _, "\n";	610	print "size is ", -s _, "\n";
185	};	611	};
186		612
		613
		614	=item aio_statvfs $fh_or_path, $callback->($statvfs)
		615
		616	Works like the POSIX C<statvfs> or C<fstatvfs> syscalls, depending on
		617	whether a file handle or path was passed.
		618
		619	On success, the callback is passed a hash reference with the following
		620	members: C<bsize>, C<frsize>, C<blocks>, C<bfree>, C<bavail>, C<files>,
		621	C<ffree>, C<favail>, C<fsid>, C<flag> and C<namemax>. On failure, C<undef>
		622	is passed.
		623
		624	The following POSIX IO::AIO::ST_* constants are defined: C<ST_RDONLY> and
		625	C<ST_NOSUID>.
		626
		627	The following non-POSIX IO::AIO::ST_* flag masks are defined to
		628	their correct value when available, or to C<0> on systems that do
		629	not support them: C<ST_NODEV>, C<ST_NOEXEC>, C<ST_SYNCHRONOUS>,
		630	C<ST_MANDLOCK>, C<ST_WRITE>, C<ST_APPEND>, C<ST_IMMUTABLE>, C<ST_NOATIME>,
		631	C<ST_NODIRATIME> and C<ST_RELATIME>.
		632
		633	Example: stat C</wd> and dump out the data if successful.
		634
		635	aio_statvfs "/wd", sub {
		636	my $f = $_[0]
		637	or die "statvfs: $!";
		638
		639	use Data::Dumper;
		640	say Dumper $f;
		641	};
		642
		643	# result:
		644	{
		645	bsize => 1024,
		646	bfree => 4333064312,
		647	blocks => 10253828096,
		648	files => 2050765568,
		649	flag => 4096,
		650	favail => 2042092649,
		651	bavail => 4333064312,
		652	ffree => 2042092649,
		653	namemax => 255,
		654	frsize => 1024,
		655	fsid => 1810
		656	}
		657
		658	=item aio_utime $fh_or_path, $atime, $mtime, $callback->($status)
		659
		660	Works like perl's C<utime> function (including the special case of $atime
		661	and $mtime being undef). Fractional times are supported if the underlying
		662	syscalls support them.
		663
		664	When called with a pathname, uses utimensat(2) or utimes(2) if available,
		665	otherwise utime(2). If called on a file descriptor, uses futimens(2)
		666	or futimes(2) if available, otherwise returns ENOSYS, so this is not
		667	portable.
		668
		669	Examples:
		670
		671	# set atime and mtime to current time (basically touch(1)):
		672	aio_utime "path", undef, undef;
		673	# set atime to current time and mtime to beginning of the epoch:
		674	aio_utime "path", time, undef; # undef==0
		675
		676
		677	=item aio_chown $fh_or_path, $uid, $gid, $callback->($status)
		678
		679	Works like perl's C<chown> function, except that C<undef> for either $uid
		680	or $gid is being interpreted as "do not change" (but -1 can also be used).
		681
		682	Examples:
		683
		684	# same as "chown root path" in the shell:
		685	aio_chown "path", 0, -1;
		686	# same as above:
		687	aio_chown "path", 0, undef;
		688
		689
		690	=item aio_truncate $fh_or_path, $offset, $callback->($status)
		691
		692	Works like truncate(2) or ftruncate(2).
		693
		694
		695	=item aio_allocate $fh, $mode, $offset, $len, $callback->($status)
		696
		697	Allocates or frees disk space according to the C<$mode> argument. See the
		698	linux C<fallocate> documentation for details.
		699
		700	C<$mode> is usually C<0> or C<IO::AIO::FALLOC_FL_KEEP_SIZE> to allocate
		701	space, or C<IO::AIO::FALLOC_FL_PUNCH_HOLE | IO::AIO::FALLOC_FL_KEEP_SIZE>,
		702	to deallocate a file range.
		703
		704	IO::AIO also supports C<FALLOC_FL_COLLAPSE_RANGE>, to remove a range
		705	(without leaving a hole), C<FALLOC_FL_ZERO_RANGE>, to zero a range,
		706	C<FALLOC_FL_INSERT_RANGE> to insert a range and C<FALLOC_FL_UNSHARE_RANGE>
		707	to unshare shared blocks (see your L<fallocate(2)> manpage).
		708
		709	The file system block size used by C<fallocate> is presumably the
		710	C<f_bsize> returned by C<statvfs>, but different filesystems and filetypes
		711	can dictate other limitations.
		712
		713	If C<fallocate> isn't available or cannot be emulated (currently no
		714	emulation will be attempted), passes C<-1> and sets C<$!> to C<ENOSYS>.
		715
		716
		717	=item aio_chmod $fh_or_path, $mode, $callback->($status)
		718
		719	Works like perl's C<chmod> function.
		720
		721
187	=item aio_unlink $pathname, $callback	722	=item aio_unlink $pathname, $callback->($status)
188		723
189	Asynchronously unlink (delete) a file and call the callback with the	724	Asynchronously unlink (delete) a file and call the callback with the
190	result code.	725	result code.
191		726
		727
		728	=item aio_mknod $pathname, $mode, $dev, $callback->($status)
		729
		730	[EXPERIMENTAL]
		731
		732	Asynchronously create a device node (or fifo). See mknod(2).
		733
		734	The only (POSIX-) portable way of calling this function is:
		735
		736	aio_mknod $pathname, IO::AIO::S_IFIFO | $mode, 0, sub { ...
		737
		738	See C<aio_stat> for info about some potentially helpful extra constants
		739	and functions.
		740
		741	=item aio_link $srcpath, $dstpath, $callback->($status)
		742
		743	Asynchronously create a new link to the existing object at C<$srcpath> at
		744	the path C<$dstpath> and call the callback with the result code.
		745
		746
		747	=item aio_symlink $srcpath, $dstpath, $callback->($status)
		748
		749	Asynchronously create a new symbolic link to the existing object at C<$srcpath> at
		750	the path C<$dstpath> and call the callback with the result code.
		751
		752
		753	=item aio_readlink $pathname, $callback->($link)
		754
		755	Asynchronously read the symlink specified by C<$path> and pass it to
		756	the callback. If an error occurs, nothing or undef gets passed to the
		757	callback.
		758
		759
		760	=item aio_realpath $pathname, $callback->($path)
		761
		762	Asynchronously make the path absolute and resolve any symlinks in
		763	C<$path>. The resulting path only consists of directories (same as
		764	L<Cwd::realpath>).
		765
		766	This request can be used to get the absolute path of the current working
		767	directory by passing it a path of F<.> (a single dot).
		768
		769
		770	=item aio_rename $srcpath, $dstpath, $callback->($status)
		771
		772	Asynchronously rename the object at C<$srcpath> to C<$dstpath>, just as
		773	rename(2) and call the callback with the result code.
		774
		775	On systems that support the AIO::WD working directory abstraction
		776	natively, the case C<[$wd, "."]> as C<$srcpath> is specialcased - instead
		777	of failing, C<rename> is called on the absolute path of C<$wd>.
		778
		779
		780	=item aio_rename2 $srcpath, $dstpath, $flags, $callback->($status)
		781
		782	Basically a version of C<aio_rename> with an additional C<$flags>
		783	argument. Calling this with C<$flags=0> is the same as calling
		784	C<aio_rename>.
		785
		786	Non-zero flags are currently only supported on GNU/Linux systems that
		787	support renameat2. Other systems fail with C<ENOSYS> in this case.
		788
		789	The following constants are available (missing ones are, as usual C<0>),
		790	see renameat2(2) for details:
		791
		792	C<IO::AIO::RENAME_NOREPLACE>, C<IO::AIO::RENAME_EXCHANGE>
		793	and C<IO::AIO::RENAME_WHITEOUT>.
		794
		795
		796	=item aio_mkdir $pathname, $mode, $callback->($status)
		797
		798	Asynchronously mkdir (create) a directory and call the callback with
		799	the result code. C<$mode> will be modified by the umask at the time the
		800	request is executed, so do not change your umask.
		801
		802
		803	=item aio_rmdir $pathname, $callback->($status)
		804
		805	Asynchronously rmdir (delete) a directory and call the callback with the
		806	result code.
		807
		808	On systems that support the AIO::WD working directory abstraction
		809	natively, the case C<[$wd, "."]> is specialcased - instead of failing,
		810	C<rmdir> is called on the absolute path of C<$wd>.
		811
		812
		813	=item aio_readdir $pathname, $callback->($entries)
		814
		815	Unlike the POSIX call of the same name, C<aio_readdir> reads an entire
		816	directory (i.e. opendir + readdir + closedir). The entries will not be
		817	sorted, and will B<NOT> include the C<.> and C<..> entries.
		818
		819	The callback is passed a single argument which is either C<undef> or an
		820	array-ref with the filenames.
		821
		822
		823	=item aio_readdirx $pathname, $flags, $callback->($entries, $flags)
		824
		825	Quite similar to C<aio_readdir>, but the C<$flags> argument allows one to
		826	tune behaviour and output format. In case of an error, C<$entries> will be
		827	C<undef>.
		828
		829	The flags are a combination of the following constants, ORed together (the
		830	flags will also be passed to the callback, possibly modified):
		831
		832	=over 4
		833
		834	=item IO::AIO::READDIR_DENTS
		835
		836	Normally the callback gets an arrayref consisting of names only (as
		837	with C<aio_readdir>). If this flag is set, then the callback gets an
		838	arrayref with C<[$name, $type, $inode]> arrayrefs, each describing a
		839	single directory entry in more detail:
		840
		841	C<$name> is the name of the entry.
		842
		843	C<$type> is one of the C<IO::AIO::DT_xxx> constants:
		844
		845	C<IO::AIO::DT_UNKNOWN>, C<IO::AIO::DT_FIFO>, C<IO::AIO::DT_CHR>, C<IO::AIO::DT_DIR>,
		846	C<IO::AIO::DT_BLK>, C<IO::AIO::DT_REG>, C<IO::AIO::DT_LNK>, C<IO::AIO::DT_SOCK>,
		847	C<IO::AIO::DT_WHT>.
		848
		849	C<IO::AIO::DT_UNKNOWN> means just that: readdir does not know. If you need
		850	to know, you have to run stat yourself. Also, for speed/memory reasons,
		851	the C<$type> scalars are read-only: you must not modify them.
		852
		853	C<$inode> is the inode number (which might not be exact on systems with 64
		854	bit inode numbers and 32 bit perls). This field has unspecified content on
		855	systems that do not deliver the inode information.
		856
		857	=item IO::AIO::READDIR_DIRS_FIRST
		858
		859	When this flag is set, then the names will be returned in an order where
		860	likely directories come first, in optimal stat order. This is useful when
		861	you need to quickly find directories, or you want to find all directories
		862	while avoiding to stat() each entry.
		863
		864	If the system returns type information in readdir, then this is used
		865	to find directories directly. Otherwise, likely directories are names
		866	beginning with ".", or otherwise names with no dots, of which names with
		867	short names are tried first.
		868
		869	=item IO::AIO::READDIR_STAT_ORDER
		870
		871	When this flag is set, then the names will be returned in an order
		872	suitable for stat()'ing each one. That is, when you plan to stat() most or
		873	all files in the given directory, then the returned order will likely be
		874	faster.
		875
		876	If both this flag and C<IO::AIO::READDIR_DIRS_FIRST> are specified,
		877	then the likely dirs come first, resulting in a less optimal stat order
		878	for stat'ing all entries, but likely a more optimal order for finding
		879	subdirectories.
		880
		881	=item IO::AIO::READDIR_FOUND_UNKNOWN
		882
		883	This flag should not be set when calling C<aio_readdirx>. Instead, it
		884	is being set by C<aio_readdirx>, when any of the C<$type>'s found were
		885	C<IO::AIO::DT_UNKNOWN>. The absence of this flag therefore indicates that all
		886	C<$type>'s are known, which can be used to speed up some algorithms.
		887
		888	=back
		889
		890
		891	=item aio_slurp $pathname, $offset, $length, $data, $callback->($status)
		892
		893	Opens, reads and closes the given file. The data is put into C<$data>,
		894	which is resized as required.
		895
		896	If C<$offset> is negative, then it is counted from the end of the file.
		897
		898	If C<$length> is zero, then the remaining length of the file is
		899	used. Also, in this case, the same limitations to modifying C<$data> apply
		900	as when IO::AIO::mmap is used, i.e. it must only be modified in-place
		901	with C<substr>. If the size of the file is known, specifying a non-zero
		902	C<$length> results in a performance advantage.
		903
		904	This request is similar to the older C<aio_load> request, but since it is
		905	a single request, it might be more efficient to use.
		906
		907	Example: load F</etc/passwd> into C<$passwd>.
		908
		909	my $passwd;
		910	aio_slurp "/etc/passwd", 0, 0, $passwd, sub {
		911	$_[0] >= 0
		912	or die "/etc/passwd: $!\n";
		913
		914	printf "/etc/passwd is %d bytes long, and contains:\n", length $passwd;
		915	print $passwd;
		916	};
		917	IO::AIO::flush;
		918
		919
		920	=item aio_load $pathname, $data, $callback->($status)
		921
		922	This is a composite request that tries to fully load the given file into
		923	memory. Status is the same as with aio_read.
		924
		925	Using C<aio_slurp> might be more efficient, as it is a single request.
		926
		927	=cut
		928
		929	sub aio_load($$;$) {
		930	my ($path, undef, $cb) = @_;
		931	my $data = \$_[1];
		932
		933	my $pri = aioreq_pri;
		934	my $grp = aio_group $cb;
		935
		936	aioreq_pri $pri;
		937	add $grp aio_open $path, O_RDONLY, 0, sub {
		938	my $fh = shift
		939	or return $grp->result (-1);
		940
		941	aioreq_pri $pri;
		942	add $grp aio_read $fh, 0, (-s $fh), $$data, 0, sub {
		943	$grp->result ($_[0]);
		944	};
		945	};
		946
		947	$grp
		948	}
		949
		950	=item aio_copy $srcpath, $dstpath, $callback->($status)
		951
		952	Try to copy the I<file> (directories not supported as either source or
		953	destination) from C<$srcpath> to C<$dstpath> and call the callback with
		954	a status of C<0> (ok) or C<-1> (error, see C<$!>).
		955
		956	Existing destination files will be truncated.
		957
		958	This is a composite request that creates the destination file with
		959	mode 0200 and copies the contents of the source file into it using
		960	C<aio_sendfile>, followed by restoring atime, mtime, access mode and
		961	uid/gid, in that order.
		962
		963	If an error occurs, the partial destination file will be unlinked, if
		964	possible, except when setting atime, mtime, access mode and uid/gid, where
		965	errors are being ignored.
		966
		967	=cut
		968
		969	sub aio_copy($$;$) {
		970	my ($src, $dst, $cb) = @_;
		971
		972	my $pri = aioreq_pri;
		973	my $grp = aio_group $cb;
		974
		975	aioreq_pri $pri;
		976	add $grp aio_open $src, O_RDONLY, 0, sub {
		977	if (my $src_fh = $_[0]) {
		978	my @stat = stat $src_fh; # hmm, might block over nfs?
		979
		980	aioreq_pri $pri;
		981	add $grp aio_open $dst, O_CREAT | O_WRONLY | O_TRUNC, 0200, sub {
		982	if (my $dst_fh = $_[0]) {
		983	aioreq_pri $pri;
		984	add $grp aio_sendfile $dst_fh, $src_fh, 0, $stat[7], sub {
		985	if ($_[0] == $stat[7]) {
		986	$grp->result (0);
		987	close $src_fh;
		988
		989	my $ch = sub {
		990	aioreq_pri $pri;
		991	add $grp aio_chmod $dst_fh, $stat[2] & 07777, sub {
		992	aioreq_pri $pri;
		993	add $grp aio_chown $dst_fh, $stat[4], $stat[5], sub {
		994	aioreq_pri $pri;
		995	add $grp aio_close $dst_fh;
		996	}
		997	};
		998	};
		999
		1000	aioreq_pri $pri;
		1001	add $grp aio_utime $dst_fh, $stat[8], $stat[9], sub {
		1002	if ($_[0] < 0 && $! == ENOSYS) {
		1003	aioreq_pri $pri;
		1004	add $grp aio_utime $dst, $stat[8], $stat[9], $ch;
		1005	} else {
		1006	$ch->();
		1007	}
		1008	};
		1009	} else {
		1010	$grp->result (-1);
		1011	close $src_fh;
		1012	close $dst_fh;
		1013
		1014	aioreq $pri;
		1015	add $grp aio_unlink $dst;
		1016	}
		1017	};
		1018	} else {
		1019	$grp->result (-1);
		1020	}
		1021	},
		1022
		1023	} else {
		1024	$grp->result (-1);
		1025	}
		1026	};
		1027
		1028	$grp
		1029	}
		1030
		1031	=item aio_move $srcpath, $dstpath, $callback->($status)
		1032
		1033	Try to move the I<file> (directories not supported as either source or
		1034	destination) from C<$srcpath> to C<$dstpath> and call the callback with
		1035	a status of C<0> (ok) or C<-1> (error, see C<$!>).
		1036
		1037	This is a composite request that tries to rename(2) the file first; if
		1038	rename fails with C<EXDEV>, it copies the file with C<aio_copy> and, if
		1039	that is successful, unlinks the C<$srcpath>.
		1040
		1041	=cut
		1042
		1043	sub aio_move($$;$) {
		1044	my ($src, $dst, $cb) = @_;
		1045
		1046	my $pri = aioreq_pri;
		1047	my $grp = aio_group $cb;
		1048
		1049	aioreq_pri $pri;
		1050	add $grp aio_rename $src, $dst, sub {
		1051	if ($_[0] && $! == EXDEV) {
		1052	aioreq_pri $pri;
		1053	add $grp aio_copy $src, $dst, sub {
		1054	$grp->result ($_[0]);
		1055
		1056	unless ($_[0]) {
		1057	aioreq_pri $pri;
		1058	add $grp aio_unlink $src;
		1059	}
		1060	};
		1061	} else {
		1062	$grp->result ($_[0]);
		1063	}
		1064	};
		1065
		1066	$grp
		1067	}
		1068
		1069	=item aio_scandir $pathname, $maxreq, $callback->($dirs, $nondirs)
		1070
		1071	Scans a directory (similar to C<aio_readdir>) but additionally tries to
		1072	efficiently separate the entries of directory C<$path> into two sets of
		1073	names, directories you can recurse into (directories), and ones you cannot
		1074	recurse into (everything else, including symlinks to directories).
		1075
		1076	C<aio_scandir> is a composite request that generates many sub requests.
		1077	C<$maxreq> specifies the maximum number of outstanding aio requests that
		1078	this function generates. If it is C<< <= 0 >>, then a suitable default
		1079	will be chosen (currently 4).
		1080
		1081	On error, the callback is called without arguments, otherwise it receives
		1082	two array-refs with path-relative entry names.
		1083
		1084	Example:
		1085
		1086	aio_scandir $dir, 0, sub {
		1087	my ($dirs, $nondirs) = @_;
		1088	print "real directories: @$dirs\n";
		1089	print "everything else: @$nondirs\n";
		1090	};
		1091
		1092	Implementation notes.
		1093
		1094	The C<aio_readdir> cannot be avoided, but C<stat()>'ing every entry can.
		1095
		1096	If readdir returns file type information, then this is used directly to
		1097	find directories.
		1098
		1099	Otherwise, after reading the directory, the modification time, size etc.
		1100	of the directory before and after the readdir is checked, and if they
		1101	match (and isn't the current time), the link count will be used to decide
		1102	how many entries are directories (if >= 2). Otherwise, no knowledge of the
		1103	number of subdirectories will be assumed.
		1104
		1105	Then entries will be sorted into likely directories a non-initial dot
		1106	currently) and likely non-directories (see C<aio_readdirx>). Then every
		1107	entry plus an appended C</.> will be C<stat>'ed, likely directories first,
		1108	in order of their inode numbers. If that succeeds, it assumes that the
		1109	entry is a directory or a symlink to directory (which will be checked
		1110	separately). This is often faster than stat'ing the entry itself because
		1111	filesystems might detect the type of the entry without reading the inode
		1112	data (e.g. ext2fs filetype feature), even on systems that cannot return
		1113	the filetype information on readdir.
		1114
		1115	If the known number of directories (link count - 2) has been reached, the
		1116	rest of the entries is assumed to be non-directories.
		1117
		1118	This only works with certainty on POSIX (= UNIX) filesystems, which
		1119	fortunately are the vast majority of filesystems around.
		1120
		1121	It will also likely work on non-POSIX filesystems with reduced efficiency
		1122	as those tend to return 0 or 1 as link counts, which disables the
		1123	directory counting heuristic.
		1124
		1125	=cut
		1126
		1127	sub aio_scandir($$;$) {
		1128	my ($path, $maxreq, $cb) = @_;
		1129
		1130	my $pri = aioreq_pri;
		1131
		1132	my $grp = aio_group $cb;
		1133
		1134	$maxreq = 4 if $maxreq <= 0;
		1135
		1136	# get a wd object
		1137	aioreq_pri $pri;
		1138	add $grp aio_wd $path, sub {
		1139	$_[0]
		1140	or return $grp->result ();
		1141
		1142	my $wd = [shift, "."];
		1143
		1144	# stat once
		1145	aioreq_pri $pri;
		1146	add $grp aio_stat $wd, sub {
		1147	return $grp->result () if $_[0];
		1148	my $now = time;
		1149	my $hash1 = join ":", (stat _)[0,1,3,7,9];
		1150	my $rdxflags = READDIR_DIRS_FIRST;
		1151
		1152	if ((stat _)[3] < 2) {
		1153	# at least one non-POSIX filesystem exists
		1154	# that returns useful DT_type values: btrfs,
		1155	# so optimise for this here by requesting dents
		1156	$rdxflags |= READDIR_DENTS;
		1157	}
		1158
		1159	# read the directory entries
		1160	aioreq_pri $pri;
		1161	add $grp aio_readdirx $wd, $rdxflags, sub {
		1162	my ($entries, $flags) = @_
		1163	or return $grp->result ();
		1164
		1165	if ($rdxflags & READDIR_DENTS) {
		1166	# if we requested type values, see if we can use them directly.
		1167
		1168	# if there were any DT_UNKNOWN entries then we assume we
		1169	# don't know. alternatively, we could assume that if we get
		1170	# one DT_DIR, then all directories are indeed marked with
		1171	# DT_DIR, but this seems not required for btrfs, and this
		1172	# is basically the "btrfs can't get it's act together" code
		1173	# branch.
		1174	unless ($flags & READDIR_FOUND_UNKNOWN) {
		1175	# now we have valid DT_ information for all entries,
		1176	# so use it as an optimisation without further stat's.
		1177	# they must also all be at the beginning of @$entries
		1178	# by now.
		1179
		1180	my $dirs;
		1181
		1182	if (@$entries) {
		1183	for (0 .. $#$entries) {
		1184	if ($entries->[$_][1] != DT_DIR) {
		1185	# splice out directories
		1186	$dirs = [splice @$entries, 0, $_];
		1187	last;
		1188	}
		1189	}
		1190
		1191	# if we didn't find any non-dir, then all entries are dirs
		1192	unless ($dirs) {
		1193	($dirs, $entries) = ($entries, []);
		1194	}
		1195	} else {
		1196	# directory is empty, so there are no sbdirs
		1197	$dirs = [];
		1198	}
		1199
		1200	# either splice'd the directories out or the dir was empty.
		1201	# convert dents to filenames
		1202	$_ = $_->[0] for @$dirs;
		1203	$_ = $_->[0] for @$entries;
		1204
		1205	return $grp->result ($dirs, $entries);
		1206	}
		1207
		1208	# cannot use, so return to our old ways
		1209	# by pretending we only scanned for names.
		1210	$_ = $_->[0] for @$entries;
		1211	}
		1212
		1213	# stat the dir another time
		1214	aioreq_pri $pri;
		1215	add $grp aio_stat $wd, sub {
		1216	my $hash2 = join ":", (stat _)[0,1,3,7,9];
		1217
		1218	my $ndirs;
		1219
		1220	# take the slow route if anything looks fishy
		1221	if ($hash1 ne $hash2 or (stat _)[9] == $now) {
		1222	$ndirs = -1;
		1223	} else {
		1224	# if nlink == 2, we are finished
		1225	# for non-posix-fs's, we rely on nlink < 2
		1226	$ndirs = (stat _)[3] - 2
		1227	or return $grp->result ([], $entries);
		1228	}
		1229
		1230	my (@dirs, @nondirs);
		1231
		1232	my $statgrp = add $grp aio_group sub {
		1233	$grp->result (\@dirs, \@nondirs);
		1234	};
		1235
		1236	limit $statgrp $maxreq;
		1237	feed $statgrp sub {
		1238	return unless @$entries;
		1239	my $entry = shift @$entries;
		1240
		1241	aioreq_pri $pri;
		1242	$wd->[1] = "$entry/.";
		1243	add $statgrp aio_stat $wd, sub {
		1244	if ($_[0] < 0) {
		1245	push @nondirs, $entry;
		1246	} else {
		1247	# need to check for real directory
		1248	aioreq_pri $pri;
		1249	$wd->[1] = $entry;
		1250	add $statgrp aio_lstat $wd, sub {
		1251	if (-d _) {
		1252	push @dirs, $entry;
		1253
		1254	unless (--$ndirs) {
		1255	push @nondirs, @$entries;
		1256	feed $statgrp;
		1257	}
		1258	} else {
		1259	push @nondirs, $entry;
		1260	}
		1261	}
		1262	}
		1263	};
		1264	};
		1265	};
		1266	};
		1267	};
		1268	};
		1269
		1270	$grp
		1271	}
		1272
		1273	=item aio_rmtree $pathname, $callback->($status)
		1274
		1275	Delete a directory tree starting (and including) C<$path>, return the
		1276	status of the final C<rmdir> only. This is a composite request that
		1277	uses C<aio_scandir> to recurse into and rmdir directories, and unlink
		1278	everything else.
		1279
		1280	=cut
		1281
		1282	sub aio_rmtree;
		1283	sub aio_rmtree($;$) {
		1284	my ($path, $cb) = @_;
		1285
		1286	my $pri = aioreq_pri;
		1287	my $grp = aio_group $cb;
		1288
		1289	aioreq_pri $pri;
		1290	add $grp aio_scandir $path, 0, sub {
		1291	my ($dirs, $nondirs) = @_;
		1292
		1293	my $dirgrp = aio_group sub {
		1294	add $grp aio_rmdir $path, sub {
		1295	$grp->result ($_[0]);
		1296	};
		1297	};
		1298
		1299	(aioreq_pri $pri), add $dirgrp aio_rmtree "$path/$_" for @$dirs;
		1300	(aioreq_pri $pri), add $dirgrp aio_unlink "$path/$_" for @$nondirs;
		1301
		1302	add $grp $dirgrp;
		1303	};
		1304
		1305	$grp
		1306	}
		1307
		1308	=item aio_fcntl $fh, $cmd, $arg, $callback->($status)
		1309
		1310	=item aio_ioctl $fh, $request, $buf, $callback->($status)
		1311
		1312	These work just like the C<fcntl> and C<ioctl> built-in functions, except
		1313	they execute asynchronously and pass the return value to the callback.
		1314
		1315	Both calls can be used for a lot of things, some of which make more sense
		1316	to run asynchronously in their own thread, while some others make less
		1317	sense. For example, calls that block waiting for external events, such
		1318	as locking, will also lock down an I/O thread while it is waiting, which
		1319	can deadlock the whole I/O system. At the same time, there might be no
		1320	alternative to using a thread to wait.
		1321
		1322	So in general, you should only use these calls for things that do
		1323	(filesystem) I/O, not for things that wait for other events (network,
		1324	other processes), although if you are careful and know what you are doing,
		1325	you still can.
		1326
		1327	The following constants are available and can be used for normal C<ioctl>
		1328	and C<fcntl> as well (missing ones are, as usual C<0>):
		1329
		1330	C<F_DUPFD_CLOEXEC>,
		1331
		1332	C<F_OFD_GETLK>, C<F_OFD_SETLK>, C<F_OFD_GETLKW>,
		1333
		1334	C<FIFREEZE>, C<FITHAW>, C<FITRIM>, C<FICLONE>, C<FICLONERANGE>, C<FIDEDUPERANGE>.
		1335
		1336	C<F_ADD_SEALS>, C<F_GET_SEALS>, C<F_SEAL_SEAL>, C<F_SEAL_SHRINK>, C<F_SEAL_GROW> and
		1337	C<F_SEAL_WRITE>.
		1338
		1339	C<FS_IOC_GETFLAGS>, C<FS_IOC_SETFLAGS>, C<FS_IOC_GETVERSION>, C<FS_IOC_SETVERSION>,
		1340	C<FS_IOC_FIEMAP>.
		1341
		1342	C<FS_IOC_FSGETXATTR>, C<FS_IOC_FSSETXATTR>, C<FS_IOC_SET_ENCRYPTION_POLICY>,
		1343	C<FS_IOC_GET_ENCRYPTION_PWSALT>, C<FS_IOC_GET_ENCRYPTION_POLICY>, C<FS_KEY_DESCRIPTOR_SIZE>.
		1344
		1345	C<FS_SECRM_FL>, C<FS_UNRM_FL>, C<FS_COMPR_FL>, C<FS_SYNC_FL>, C<FS_IMMUTABLE_FL>,
		1346	C<FS_APPEND_FL>, C<FS_NODUMP_FL>, C<FS_NOATIME_FL>, C<FS_DIRTY_FL>,
		1347	C<FS_COMPRBLK_FL>, C<FS_NOCOMP_FL>, C<FS_ENCRYPT_FL>, C<FS_BTREE_FL>,
		1348	C<FS_INDEX_FL>, C<FS_JOURNAL_DATA_FL>, C<FS_NOTAIL_FL>, C<FS_DIRSYNC_FL>, C<FS_TOPDIR_FL>,
		1349	C<FS_FL_USER_MODIFIABLE>.
		1350
		1351	C<FS_XFLAG_REALTIME>, C<FS_XFLAG_PREALLOC>, C<FS_XFLAG_IMMUTABLE>, C<FS_XFLAG_APPEND>,
		1352	C<FS_XFLAG_SYNC>, C<FS_XFLAG_NOATIME>, C<FS_XFLAG_NODUMP>, C<FS_XFLAG_RTINHERIT>,
		1353	C<FS_XFLAG_PROJINHERIT>, C<FS_XFLAG_NOSYMLINKS>, C<FS_XFLAG_EXTSIZE>, C<FS_XFLAG_EXTSZINHERIT>,
		1354	C<FS_XFLAG_NODEFRAG>, C<FS_XFLAG_FILESTREAM>, C<FS_XFLAG_DAX>, C<FS_XFLAG_HASATTR>,
		1355
		1356	=item aio_sync $callback->($status)
		1357
		1358	Asynchronously call sync and call the callback when finished.
		1359
192	=item aio_fsync $fh, $callback	1360	=item aio_fsync $fh, $callback->($status)
193		1361
194	Asynchronously call fsync on the given filehandle and call the callback	1362	Asynchronously call fsync on the given filehandle and call the callback
195	with the fsync result code.	1363	with the fsync result code.
196		1364
197	=item aio_fdatasync $fh, $callback	1365	=item aio_fdatasync $fh, $callback->($status)
198		1366
199	Asynchronously call fdatasync on the given filehandle and call the	1367	Asynchronously call fdatasync on the given filehandle and call the
200	callback with the fdatasync result code. Might set C<$!> to C<ENOSYS> if	1368	callback with the fdatasync result code.
201	C<fdatasync> is not available.	1369
		1370	If this call isn't available because your OS lacks it or it couldn't be
		1371	detected, it will be emulated by calling C<fsync> instead.
		1372
		1373	=item aio_syncfs $fh, $callback->($status)
		1374
		1375	Asynchronously call the syncfs syscall to sync the filesystem associated
		1376	to the given filehandle and call the callback with the syncfs result
		1377	code. If syncfs is not available, calls sync(), but returns C<-1> and sets
		1378	errno to C<ENOSYS> nevertheless.
		1379
		1380	=item aio_sync_file_range $fh, $offset, $nbytes, $flags, $callback->($status)
		1381
		1382	Sync the data portion of the file specified by C<$offset> and C<$length>
		1383	to disk (but NOT the metadata), by calling the Linux-specific
		1384	sync_file_range call. If sync_file_range is not available or it returns
		1385	ENOSYS, then fdatasync or fsync is being substituted.
		1386
		1387	C<$flags> can be a combination of C<IO::AIO::SYNC_FILE_RANGE_WAIT_BEFORE>,
		1388	C<IO::AIO::SYNC_FILE_RANGE_WRITE> and
		1389	C<IO::AIO::SYNC_FILE_RANGE_WAIT_AFTER>: refer to the sync_file_range
		1390	manpage for details.
		1391
		1392	=item aio_pathsync $pathname, $callback->($status)
		1393
		1394	This request tries to open, fsync and close the given path. This is a
		1395	composite request intended to sync directories after directory operations
		1396	(E.g. rename). This might not work on all operating systems or have any
		1397	specific effect, but usually it makes sure that directory changes get
		1398	written to disc. It works for anything that can be opened for read-only,
		1399	not just directories.
		1400
		1401	Future versions of this function might fall back to other methods when
		1402	C<fsync> on the directory fails (such as calling C<sync>).
		1403
		1404	Passes C<0> when everything went ok, and C<-1> on error.
		1405
		1406	=cut
		1407
		1408	sub aio_pathsync($;$) {
		1409	my ($path, $cb) = @_;
		1410
		1411	my $pri = aioreq_pri;
		1412	my $grp = aio_group $cb;
		1413
		1414	aioreq_pri $pri;
		1415	add $grp aio_open $path, O_RDONLY, 0, sub {
		1416	my ($fh) = @_;
		1417	if ($fh) {
		1418	aioreq_pri $pri;
		1419	add $grp aio_fsync $fh, sub {
		1420	$grp->result ($_[0]);
		1421
		1422	aioreq_pri $pri;
		1423	add $grp aio_close $fh;
		1424	};
		1425	} else {
		1426	$grp->result (-1);
		1427	}
		1428	};
		1429
		1430	$grp
		1431	}
		1432
		1433	=item aio_msync $scalar, $offset = 0, $length = undef, flags = MS_SYNC, $callback->($status)
		1434
		1435	This is a rather advanced IO::AIO call, which only works on mmap(2)ed
		1436	scalars (see the C<IO::AIO::mmap> function, although it also works on data
		1437	scalars managed by the L<Sys::Mmap> or L<Mmap> modules, note that the
		1438	scalar must only be modified in-place while an aio operation is pending on
		1439	it).
		1440
		1441	It calls the C<msync> function of your OS, if available, with the memory
		1442	area starting at C<$offset> in the string and ending C<$length> bytes
		1443	later. If C<$length> is negative, counts from the end, and if C<$length>
		1444	is C<undef>, then it goes till the end of the string. The flags can be
		1445	either C<IO::AIO::MS_ASYNC> or C<IO::AIO::MS_SYNC>, plus an optional
		1446	C<IO::AIO::MS_INVALIDATE>.
		1447
		1448	=item aio_mtouch $scalar, $offset = 0, $length = undef, flags = 0, $callback->($status)
		1449
		1450	This is a rather advanced IO::AIO call, which works best on mmap(2)ed
		1451	scalars.
		1452
		1453	It touches (reads or writes) all memory pages in the specified
		1454	range inside the scalar. All caveats and parameters are the same
		1455	as for C<aio_msync>, above, except for flags, which must be either
		1456	C<0> (which reads all pages and ensures they are instantiated) or
		1457	C<IO::AIO::MT_MODIFY>, which modifies the memory pages (by reading and
		1458	writing an octet from it, which dirties the page).
		1459
		1460	=item aio_mlock $scalar, $offset = 0, $length = undef, $callback->($status)
		1461
		1462	This is a rather advanced IO::AIO call, which works best on mmap(2)ed
		1463	scalars.
		1464
		1465	It reads in all the pages of the underlying storage into memory (if any)
		1466	and locks them, so they are not getting swapped/paged out or removed.
		1467
		1468	If C<$length> is undefined, then the scalar will be locked till the end.
		1469
		1470	On systems that do not implement C<mlock>, this function returns C<-1>
		1471	and sets errno to C<ENOSYS>.
		1472
		1473	Note that the corresponding C<munlock> is synchronous and is
		1474	documented under L<MISCELLANEOUS FUNCTIONS>.
		1475
		1476	Example: open a file, mmap and mlock it - both will be undone when
		1477	C<$data> gets destroyed.
		1478
		1479	open my $fh, "<", $path or die "$path: $!";
		1480	my $data;
		1481	IO::AIO::mmap $data, -s $fh, IO::AIO::PROT_READ, IO::AIO::MAP_SHARED, $fh;
		1482	aio_mlock $data; # mlock in background
		1483
		1484	=item aio_mlockall $flags, $callback->($status)
		1485
		1486	Calls the C<mlockall> function with the given C<$flags> (a
		1487	combination of C<IO::AIO::MCL_CURRENT>, C<IO::AIO::MCL_FUTURE> and
		1488	C<IO::AIO::MCL_ONFAULT>).
		1489
		1490	On systems that do not implement C<mlockall>, this function returns C<-1>
		1491	and sets errno to C<ENOSYS>. Similarly, flag combinations not supported
		1492	by the system result in a return value of C<-1> with errno being set to
		1493	C<EINVAL>.
		1494
		1495	Note that the corresponding C<munlockall> is synchronous and is
		1496	documented under L<MISCELLANEOUS FUNCTIONS>.
		1497
		1498	Example: asynchronously lock all current and future pages into memory.
		1499
		1500	aio_mlockall IO::AIO::MCL_FUTURE;
		1501
		1502	=item aio_fiemap $fh, $start, $length, $flags, $count, $cb->(\@extents)
		1503
		1504	Queries the extents of the given file (by calling the Linux C<FIEMAP>
		1505	ioctl, see L<http://cvs.schmorp.de/IO-AIO/doc/fiemap.txt> for details). If
		1506	the ioctl is not available on your OS, then this request will fail with
		1507	C<ENOSYS>.
		1508
		1509	C<$start> is the starting offset to query extents for, C<$length> is the
		1510	size of the range to query - if it is C<undef>, then the whole file will
		1511	be queried.
		1512
		1513	C<$flags> is a combination of flags (C<IO::AIO::FIEMAP_FLAG_SYNC> or
		1514	C<IO::AIO::FIEMAP_FLAG_XATTR> - C<IO::AIO::FIEMAP_FLAGS_COMPAT> is also
		1515	exported), and is normally C<0> or C<IO::AIO::FIEMAP_FLAG_SYNC> to query
		1516	the data portion.
		1517
		1518	C<$count> is the maximum number of extent records to return. If it is
		1519	C<undef>, then IO::AIO queries all extents of the range. As a very special
		1520	case, if it is C<0>, then the callback receives the number of extents
		1521	instead of the extents themselves (which is unreliable, see below).
		1522
		1523	If an error occurs, the callback receives no arguments. The special
		1524	C<errno> value C<IO::AIO::EBADR> is available to test for flag errors.
		1525
		1526	Otherwise, the callback receives an array reference with extent
		1527	structures. Each extent structure is an array reference itself, with the
		1528	following members:
		1529
		1530	[$logical, $physical, $length, $flags]
		1531
		1532	Flags is any combination of the following flag values (typically either C<0>
		1533	or C<IO::AIO::FIEMAP_EXTENT_LAST> (1)):
		1534
		1535	C<IO::AIO::FIEMAP_EXTENT_LAST>, C<IO::AIO::FIEMAP_EXTENT_UNKNOWN>,
		1536	C<IO::AIO::FIEMAP_EXTENT_DELALLOC>, C<IO::AIO::FIEMAP_EXTENT_ENCODED>,
		1537	C<IO::AIO::FIEMAP_EXTENT_DATA_ENCRYPTED>, C<IO::AIO::FIEMAP_EXTENT_NOT_ALIGNED>,
		1538	C<IO::AIO::FIEMAP_EXTENT_DATA_INLINE>, C<IO::AIO::FIEMAP_EXTENT_DATA_TAIL>,
		1539	C<IO::AIO::FIEMAP_EXTENT_UNWRITTEN>, C<IO::AIO::FIEMAP_EXTENT_MERGED> or
		1540	C<IO::AIO::FIEMAP_EXTENT_SHARED>.
		1541
		1542	At the time of this writing (Linux 3.2), this request is unreliable unless
		1543	C<$count> is C<undef>, as the kernel has all sorts of bugs preventing
		1544	it to return all extents of a range for files with a large number of
		1545	extents. The code (only) works around all these issues if C<$count> is
		1546	C<undef>.
		1547
		1548	=item aio_group $callback->(...)
		1549
		1550	This is a very special aio request: Instead of doing something, it is a
		1551	container for other aio requests, which is useful if you want to bundle
		1552	many requests into a single, composite, request with a definite callback
		1553	and the ability to cancel the whole request with its subrequests.
		1554
		1555	Returns an object of class L<IO::AIO::GRP>. See its documentation below
		1556	for more info.
		1557
		1558	Example:
		1559
		1560	my $grp = aio_group sub {
		1561	print "all stats done\n";
		1562	};
		1563
		1564	add $grp
		1565	(aio_stat ...),
		1566	(aio_stat ...),
		1567	...;
		1568
		1569	=item aio_nop $callback->()
		1570
		1571	This is a special request - it does nothing in itself and is only used for
		1572	side effects, such as when you want to add a dummy request to a group so
		1573	that finishing the requests in the group depends on executing the given
		1574	code.
		1575
		1576	While this request does nothing, it still goes through the execution
		1577	phase and still requires a worker thread. Thus, the callback will not
		1578	be executed immediately but only after other requests in the queue have
		1579	entered their execution phase. This can be used to measure request
		1580	latency.
		1581
		1582	=item IO::AIO::aio_busy $fractional_seconds, $callback->() *NOT EXPORTED*
		1583
		1584	Mainly used for debugging and benchmarking, this aio request puts one of
		1585	the request workers to sleep for the given time.
		1586
		1587	While it is theoretically handy to have simple I/O scheduling requests
		1588	like sleep and file handle readable/writable, the overhead this creates is
		1589	immense (it blocks a thread for a long time) so do not use this function
		1590	except to put your application under artificial I/O pressure.
202		1591
203	=back	1592	=back
204		1593
		1594
		1595	=head2 IO::AIO::WD - multiple working directories
		1596
		1597	Your process only has one current working directory, which is used by all
		1598	threads. This makes it hard to use relative paths (some other component
		1599	could call C<chdir> at any time, and it is hard to control when the path
		1600	will be used by IO::AIO).
		1601
		1602	One solution for this is to always use absolute paths. This usually works,
		1603	but can be quite slow (the kernel has to walk the whole path on every
		1604	access), and can also be a hassle to implement.
		1605
		1606	Newer POSIX systems have a number of functions (openat, fdopendir,
		1607	futimensat and so on) that make it possible to specify working directories
		1608	per operation.
		1609
		1610	For portability, and because the clowns who "designed", or shall I write,
		1611	perpetrated this new interface were obviously half-drunk, this abstraction
		1612	cannot be perfect, though.
		1613
		1614	IO::AIO allows you to convert directory paths into a so-called IO::AIO::WD
		1615	object. This object stores the canonicalised, absolute version of the
		1616	path, and on systems that allow it, also a directory file descriptor.
		1617
		1618	Everywhere where a pathname is accepted by IO::AIO (e.g. in C<aio_stat>
		1619	or C<aio_unlink>), one can specify an array reference with an IO::AIO::WD
		1620	object and a pathname instead (or the IO::AIO::WD object alone, which
		1621	gets interpreted as C<[$wd, "."]>). If the pathname is absolute, the
		1622	IO::AIO::WD object is ignored, otherwise the pathname is resolved relative
		1623	to that IO::AIO::WD object.
		1624
		1625	For example, to get a wd object for F</etc> and then stat F<passwd>
		1626	inside, you would write:
		1627
		1628	aio_wd "/etc", sub {
		1629	my $etcdir = shift;
		1630
		1631	# although $etcdir can be undef on error, there is generally no reason
		1632	# to check for errors here, as aio_stat will fail with ENOENT
		1633	# when $etcdir is undef.
		1634
		1635	aio_stat [$etcdir, "passwd"], sub {
		1636	# yay
		1637	};
		1638	};
		1639
		1640	The fact that C<aio_wd> is a request and not a normal function shows that
		1641	creating an IO::AIO::WD object is itself a potentially blocking operation,
		1642	which is why it is done asynchronously.
		1643
		1644	To stat the directory obtained with C<aio_wd> above, one could write
		1645	either of the following three request calls:
		1646
		1647	aio_lstat "/etc" , sub { ... # pathname as normal string
		1648	aio_lstat [$wd, "."], sub { ... # "." relative to $wd (i.e. $wd itself)
		1649	aio_lstat $wd , sub { ... # shorthand for the previous
		1650
		1651	As with normal pathnames, IO::AIO keeps a copy of the working directory
		1652	object and the pathname string, so you could write the following without
		1653	causing any issues due to C<$path> getting reused:
		1654
		1655	my $path = [$wd, undef];
		1656
		1657	for my $name (qw(abc def ghi)) {
		1658	$path->[1] = $name;
		1659	aio_stat $path, sub {
		1660	# ...
		1661	};
		1662	}
		1663
		1664	There are some caveats: when directories get renamed (or deleted), the
		1665	pathname string doesn't change, so will point to the new directory (or
		1666	nowhere at all), while the directory fd, if available on the system,
		1667	will still point to the original directory. Most functions accepting a
		1668	pathname will use the directory fd on newer systems, and the string on
		1669	older systems. Some functions (such as C<aio_realpath>) will always rely on
		1670	the string form of the pathname.
		1671
		1672	So this functionality is mainly useful to get some protection against
		1673	C<chdir>, to easily get an absolute path out of a relative path for future
		1674	reference, and to speed up doing many operations in the same directory
		1675	(e.g. when stat'ing all files in a directory).
		1676
		1677	The following functions implement this working directory abstraction:
		1678
		1679	=over 4
		1680
		1681	=item aio_wd $pathname, $callback->($wd)
		1682
		1683	Asynchonously canonicalise the given pathname and convert it to an
		1684	IO::AIO::WD object representing it. If possible and supported on the
		1685	system, also open a directory fd to speed up pathname resolution relative
		1686	to this working directory.
		1687
		1688	If something goes wrong, then C<undef> is passwd to the callback instead
		1689	of a working directory object and C<$!> is set appropriately. Since
		1690	passing C<undef> as working directory component of a pathname fails the
		1691	request with C<ENOENT>, there is often no need for error checking in the
		1692	C<aio_wd> callback, as future requests using the value will fail in the
		1693	expected way.
		1694
		1695	=item IO::AIO::CWD
		1696
		1697	This is a compile time constant (object) that represents the process
		1698	current working directory.
		1699
		1700	Specifying this object as working directory object for a pathname is as if
		1701	the pathname would be specified directly, without a directory object. For
		1702	example, these calls are functionally identical:
		1703
		1704	aio_stat "somefile", sub { ... };
		1705	aio_stat [IO::AIO::CWD, "somefile"], sub { ... };
		1706
		1707	=back
		1708
		1709	To recover the path associated with an IO::AIO::WD object, you can use
		1710	C<aio_realpath>:
		1711
		1712	aio_realpath $wd, sub {
		1713	warn "path is $_[0]\n";
		1714	};
		1715
		1716	Currently, C<aio_statvfs> always, and C<aio_rename> and C<aio_rmdir>
		1717	sometimes, fall back to using an absolue path.
		1718
		1719	=head2 IO::AIO::REQ CLASS
		1720
		1721	All non-aggregate C<aio_*> functions return an object of this class when
		1722	called in non-void context.
		1723
		1724	=over 4
		1725
		1726	=item cancel $req
		1727
		1728	Cancels the request, if possible. Has the effect of skipping execution
		1729	when entering the B<execute> state and skipping calling the callback when
		1730	entering the the B<result> state, but will leave the request otherwise
		1731	untouched (with the exception of readdir). That means that requests that
		1732	currently execute will not be stopped and resources held by the request
		1733	will not be freed prematurely.
		1734
		1735	=item cb $req $callback->(...)
		1736
		1737	Replace (or simply set) the callback registered to the request.
		1738
		1739	=back
		1740
		1741	=head2 IO::AIO::GRP CLASS
		1742
		1743	This class is a subclass of L<IO::AIO::REQ>, so all its methods apply to
		1744	objects of this class, too.
		1745
		1746	A IO::AIO::GRP object is a special request that can contain multiple other
		1747	aio requests.
		1748
		1749	You create one by calling the C<aio_group> constructing function with a
		1750	callback that will be called when all contained requests have entered the
		1751	C<done> state:
		1752
		1753	my $grp = aio_group sub {
		1754	print "all requests are done\n";
		1755	};
		1756
		1757	You add requests by calling the C<add> method with one or more
		1758	C<IO::AIO::REQ> objects:
		1759
		1760	$grp->add (aio_unlink "...");
		1761
		1762	add $grp aio_stat "...", sub {
		1763	$_[0] or return $grp->result ("error");
		1764
		1765	# add another request dynamically, if first succeeded
		1766	add $grp aio_open "...", sub {
		1767	$grp->result ("ok");
		1768	};
		1769	};
		1770
		1771	This makes it very easy to create composite requests (see the source of
		1772	C<aio_move> for an application) that work and feel like simple requests.
		1773
		1774	=over 4
		1775
		1776	=item * The IO::AIO::GRP objects will be cleaned up during calls to
		1777	C<IO::AIO::poll_cb>, just like any other request.
		1778
		1779	=item * They can be canceled like any other request. Canceling will cancel not
		1780	only the request itself, but also all requests it contains.
		1781
		1782	=item * They can also can also be added to other IO::AIO::GRP objects.
		1783
		1784	=item * You must not add requests to a group from within the group callback (or
		1785	any later time).
		1786
		1787	=back
		1788
		1789	Their lifetime, simplified, looks like this: when they are empty, they
		1790	will finish very quickly. If they contain only requests that are in the
		1791	C<done> state, they will also finish. Otherwise they will continue to
		1792	exist.
		1793
		1794	That means after creating a group you have some time to add requests
		1795	(precisely before the callback has been invoked, which is only done within
		1796	the C<poll_cb>). And in the callbacks of those requests, you can add
		1797	further requests to the group. And only when all those requests have
		1798	finished will the the group itself finish.
		1799
		1800	=over 4
		1801
		1802	=item add $grp ...
		1803
		1804	=item $grp->add (...)
		1805
		1806	Add one or more requests to the group. Any type of L<IO::AIO::REQ> can
		1807	be added, including other groups, as long as you do not create circular
		1808	dependencies.
		1809
		1810	Returns all its arguments.
		1811
		1812	=item $grp->cancel_subs
		1813
		1814	Cancel all subrequests and clears any feeder, but not the group request
		1815	itself. Useful when you queued a lot of events but got a result early.
		1816
		1817	The group request will finish normally (you cannot add requests to the
		1818	group).
		1819
		1820	=item $grp->result (...)
		1821
		1822	Set the result value(s) that will be passed to the group callback when all
		1823	subrequests have finished and set the groups errno to the current value
		1824	of errno (just like calling C<errno> without an error number). By default,
		1825	no argument will be passed and errno is zero.
		1826
		1827	=item $grp->errno ([$errno])
		1828
		1829	Sets the group errno value to C<$errno>, or the current value of errno
		1830	when the argument is missing.
		1831
		1832	Every aio request has an associated errno value that is restored when
		1833	the callback is invoked. This method lets you change this value from its
		1834	default (0).
		1835
		1836	Calling C<result> will also set errno, so make sure you either set C<$!>
		1837	before the call to C<result>, or call c<errno> after it.
		1838
		1839	=item feed $grp $callback->($grp)
		1840
		1841	Sets a feeder/generator on this group: every group can have an attached
		1842	generator that generates requests if idle. The idea behind this is that,
		1843	although you could just queue as many requests as you want in a group,
		1844	this might starve other requests for a potentially long time. For example,
		1845	C<aio_scandir> might generate hundreds of thousands of C<aio_stat>
		1846	requests, delaying any later requests for a long time.
		1847
		1848	To avoid this, and allow incremental generation of requests, you can
		1849	instead a group and set a feeder on it that generates those requests. The
		1850	feed callback will be called whenever there are few enough (see C<limit>,
		1851	below) requests active in the group itself and is expected to queue more
		1852	requests.
		1853
		1854	The feed callback can queue as many requests as it likes (i.e. C<add> does
		1855	not impose any limits).
		1856
		1857	If the feed does not queue more requests when called, it will be
		1858	automatically removed from the group.
		1859
		1860	If the feed limit is C<0> when this method is called, it will be set to
		1861	C<2> automatically.
		1862
		1863	Example:
		1864
		1865	# stat all files in @files, but only ever use four aio requests concurrently:
		1866
		1867	my $grp = aio_group sub { print "finished\n" };
		1868	limit $grp 4;
		1869	feed $grp sub {
		1870	my $file = pop @files
		1871	or return;
		1872
		1873	add $grp aio_stat $file, sub { ... };
		1874	};
		1875
		1876	=item limit $grp $num
		1877
		1878	Sets the feeder limit for the group: The feeder will be called whenever
		1879	the group contains less than this many requests.
		1880
		1881	Setting the limit to C<0> will pause the feeding process.
		1882
		1883	The default value for the limit is C<0>, but note that setting a feeder
		1884	automatically bumps it up to C<2>.
		1885
		1886	=back
		1887
		1888
205	=head2 SUPPORT FUNCTIONS	1889	=head2 SUPPORT FUNCTIONS
206		1890
		1891	=head3 EVENT PROCESSING AND EVENT LOOP INTEGRATION
		1892
207	=over 4	1893	=over 4
208		1894
209	=item $fileno = IO::AIO::poll_fileno	1895	=item $fileno = IO::AIO::poll_fileno
210		1896
211	Return the I<request result pipe file descriptor>. This filehandle must be	1897	Return the I<request result pipe file descriptor>. This filehandle must be
212	polled for reading by some mechanism outside this module (e.g. Event or	1898	polled for reading by some mechanism outside this module (e.g. EV, Glib,
213	select, see below or the SYNOPSIS). If the pipe becomes readable you have	1899	select and so on, see below or the SYNOPSIS). If the pipe becomes readable
214	to call C<poll_cb> to check the results.	1900	you have to call C<poll_cb> to check the results.
215		1901
216	See C<poll_cb> for an example.	1902	See C<poll_cb> for an example.
217		1903
218	=item IO::AIO::poll_cb	1904	=item IO::AIO::poll_cb
219		1905
220	Process all outstanding events on the result pipe. You have to call this	1906	Process some requests that have reached the result phase (i.e. they have
221	regularly. Returns the number of events processed. Returns immediately	1907	been executed but the results are not yet reported). You have to call
222	when no events are outstanding.	1908	this "regularly" to finish outstanding requests.
		1909
		1910	Returns C<0> if all events could be processed (or there were no
		1911	events to process), or C<-1> if it returned earlier for whatever
		1912	reason. Returns immediately when no events are outstanding. The amount
		1913	of events processed depends on the settings of C<IO::AIO::max_poll_req>,
		1914	C<IO::AIO::max_poll_time> and C<IO::AIO::max_outstanding>.
		1915
		1916	If not all requests were processed for whatever reason, the poll file
		1917	descriptor will still be ready when C<poll_cb> returns, so normally you
		1918	don't have to do anything special to have it called later.
		1919
		1920	Apart from calling C<IO::AIO::poll_cb> when the event filehandle becomes
		1921	ready, it can be beneficial to call this function from loops which submit
		1922	a lot of requests, to make sure the results get processed when they become
		1923	available and not just when the loop is finished and the event loop takes
		1924	over again. This function returns very fast when there are no outstanding
		1925	requests.
223		1926
224	Example: Install an Event watcher that automatically calls	1927	Example: Install an Event watcher that automatically calls
225	IO::AIO::poll_cb with high priority:	1928	IO::AIO::poll_cb with high priority (more examples can be found in the
		1929	SYNOPSIS section, at the top of this document):
226		1930
227	Event->io (fd => IO::AIO::poll_fileno,	1931	Event->io (fd => IO::AIO::poll_fileno,
228	poll => 'r', async => 1,	1932	poll => 'r', async => 1,
229	cb => \&IO::AIO::poll_cb);	1933	cb => \&IO::AIO::poll_cb);
230		1934
231	=item IO::AIO::poll_wait	1935	=item IO::AIO::poll_wait
232		1936
233	Wait till the result filehandle becomes ready for reading (simply does a	1937	Wait until either at least one request is in the result phase or no
234	C<select> on the filehandle. This is useful if you want to synchronously wait	1938	requests are outstanding anymore.
235	for some requests to finish).	1939
		1940	This is useful if you want to synchronously wait for some requests to
		1941	become ready, without actually handling them.
236		1942
237	See C<nreqs> for an example.	1943	See C<nreqs> for an example.
238		1944
		1945	=item IO::AIO::poll
		1946
		1947	Waits until some requests have been handled.
		1948
		1949	Returns the number of requests processed, but is otherwise strictly
		1950	equivalent to:
		1951
		1952	IO::AIO::poll_wait, IO::AIO::poll_cb
		1953
239	=item IO::AIO::nreqs	1954	=item IO::AIO::flush
240		1955
241	Returns the number of requests currently outstanding (i.e. for which their	1956	Wait till all outstanding AIO requests have been handled.
242	callback has not been invoked yet).
243		1957
244	Example: wait till there are no outstanding requests anymore:	1958	Strictly equivalent to:
245		1959
246	IO::AIO::poll_wait, IO::AIO::poll_cb	1960	IO::AIO::poll_wait, IO::AIO::poll_cb
247	while IO::AIO::nreqs;	1961	while IO::AIO::nreqs;
248		1962
		1963	This function can be useful at program aborts, to make sure outstanding
		1964	I/O has been done (C<IO::AIO> uses an C<END> block which already calls
		1965	this function on normal exits), or when you are merely using C<IO::AIO>
		1966	for its more advanced functions, rather than for async I/O, e.g.:
		1967
		1968	my ($dirs, $nondirs);
		1969	IO::AIO::aio_scandir "/tmp", 0, sub { ($dirs, $nondirs) = @_ };
		1970	IO::AIO::flush;
		1971	# $dirs, $nondirs are now set
		1972
		1973	=item IO::AIO::max_poll_reqs $nreqs
		1974
		1975	=item IO::AIO::max_poll_time $seconds
		1976
		1977	These set the maximum number of requests (default C<0>, meaning infinity)
		1978	that are being processed by C<IO::AIO::poll_cb> in one call, respectively
		1979	the maximum amount of time (default C<0>, meaning infinity) spent in
		1980	C<IO::AIO::poll_cb> to process requests (more correctly the mininum amount
		1981	of time C<poll_cb> is allowed to use).
		1982
		1983	Setting C<max_poll_time> to a non-zero value creates an overhead of one
		1984	syscall per request processed, which is not normally a problem unless your
		1985	callbacks are really really fast or your OS is really really slow (I am
		1986	not mentioning Solaris here). Using C<max_poll_reqs> incurs no overhead.
		1987
		1988	Setting these is useful if you want to ensure some level of
		1989	interactiveness when perl is not fast enough to process all requests in
		1990	time.
		1991
		1992	For interactive programs, values such as C<0.01> to C<0.1> should be fine.
		1993
		1994	Example: Install an Event watcher that automatically calls
		1995	IO::AIO::poll_cb with low priority, to ensure that other parts of the
		1996	program get the CPU sometimes even under high AIO load.
		1997
		1998	# try not to spend much more than 0.1s in poll_cb
		1999	IO::AIO::max_poll_time 0.1;
		2000
		2001	# use a low priority so other tasks have priority
		2002	Event->io (fd => IO::AIO::poll_fileno,
		2003	poll => 'r', nice => 1,
		2004	cb => &IO::AIO::poll_cb);
		2005
		2006	=back
		2007
		2008
		2009	=head3 CONTROLLING THE NUMBER OF THREADS
		2010
		2011	=over
		2012
		2013	=item IO::AIO::min_parallel $nthreads
		2014
		2015	Set the minimum number of AIO threads to C<$nthreads>. The current
		2016	default is C<8>, which means eight asynchronous operations can execute
		2017	concurrently at any one time (the number of outstanding requests,
		2018	however, is unlimited).
		2019
		2020	IO::AIO starts threads only on demand, when an AIO request is queued and
		2021	no free thread exists. Please note that queueing up a hundred requests can
		2022	create demand for a hundred threads, even if it turns out that everything
		2023	is in the cache and could have been processed faster by a single thread.
		2024
		2025	It is recommended to keep the number of threads relatively low, as some
		2026	Linux kernel versions will scale negatively with the number of threads
		2027	(higher parallelity => MUCH higher latency). With current Linux 2.6
		2028	versions, 4-32 threads should be fine.
		2029
		2030	Under most circumstances you don't need to call this function, as the
		2031	module selects a default that is suitable for low to moderate load.
		2032
		2033	=item IO::AIO::max_parallel $nthreads
		2034
		2035	Sets the maximum number of AIO threads to C<$nthreads>. If more than the
		2036	specified number of threads are currently running, this function kills
		2037	them. This function blocks until the limit is reached.
		2038
		2039	While C<$nthreads> are zero, aio requests get queued but not executed
		2040	until the number of threads has been increased again.
		2041
		2042	This module automatically runs C<max_parallel 0> at program end, to ensure
		2043	that all threads are killed and that there are no outstanding requests.
		2044
		2045	Under normal circumstances you don't need to call this function.
		2046
		2047	=item IO::AIO::max_idle $nthreads
		2048
		2049	Limit the number of threads (default: 4) that are allowed to idle
		2050	(i.e., threads that did not get a request to process within the idle
		2051	timeout (default: 10 seconds). That means if a thread becomes idle while
		2052	C<$nthreads> other threads are also idle, it will free its resources and
		2053	exit.
		2054
		2055	This is useful when you allow a large number of threads (e.g. 100 or 1000)
		2056	to allow for extremely high load situations, but want to free resources
		2057	under normal circumstances (1000 threads can easily consume 30MB of RAM).
		2058
		2059	The default is probably ok in most situations, especially if thread
		2060	creation is fast. If thread creation is very slow on your system you might
		2061	want to use larger values.
		2062
		2063	=item IO::AIO::idle_timeout $seconds
		2064
		2065	Sets the minimum idle timeout (default 10) after which worker threads are
		2066	allowed to exit. SEe C<IO::AIO::max_idle>.
		2067
		2068	=item IO::AIO::max_outstanding $maxreqs
		2069
		2070	Sets the maximum number of outstanding requests to C<$nreqs>. If
		2071	you do queue up more than this number of requests, the next call to
		2072	C<IO::AIO::poll_cb> (and other functions calling C<poll_cb>, such as
		2073	C<IO::AIO::flush> or C<IO::AIO::poll>) will block until the limit is no
		2074	longer exceeded.
		2075
		2076	In other words, this setting does not enforce a queue limit, but can be
		2077	used to make poll functions block if the limit is exceeded.
		2078
		2079	This is a bad function to use in interactive programs because it blocks,
		2080	and a bad way to reduce concurrency because it is inexact. If you need to
		2081	issue many requests without being able to call a poll function on demand,
		2082	it is better to use an C<aio_group> together with a feed callback.
		2083
		2084	Its main use is in scripts without an event loop - when you want to stat a
		2085	lot of files, you can write something like this:
		2086
		2087	IO::AIO::max_outstanding 32;
		2088
		2089	for my $path (...) {
		2090	aio_stat $path , ...;
		2091	IO::AIO::poll_cb;
		2092	}
		2093
		2094	IO::AIO::flush;
		2095
		2096	The call to C<poll_cb> inside the loop will normally return instantly,
		2097	allowing the loop to progress, but as soon as more than C<32> requests
		2098	are in-flight, it will block until some requests have been handled. This
		2099	keeps the loop from pushing a large number of C<aio_stat> requests onto
		2100	the queue (which, with many paths to stat, can use up a lot of memory).
		2101
		2102	The default value for C<max_outstanding> is very large, so there is no
		2103	practical limit on the number of outstanding requests.
		2104
		2105	=back
		2106
		2107
		2108	=head3 STATISTICAL INFORMATION
		2109
		2110	=over
		2111
249	=item IO::AIO::flush	2112	=item IO::AIO::nreqs
250		2113
251	Wait till all outstanding AIO requests have been handled.	2114	Returns the number of requests currently in the ready, execute or pending
		2115	states (i.e. for which their callback has not been invoked yet).
252		2116
253	Strictly equivalent to:	2117	Example: wait till there are no outstanding requests anymore:
254		2118
255	IO::AIO::poll_wait, IO::AIO::poll_cb	2119	IO::AIO::poll_wait, IO::AIO::poll_cb
256	while IO::AIO::nreqs;	2120	while IO::AIO::nreqs;
257		2121
		2122	=item IO::AIO::nready
		2123
		2124	Returns the number of requests currently in the ready state (not yet
		2125	executed).
		2126
258	=item IO::AIO::poll	2127	=item IO::AIO::npending
259		2128
260	Waits until some requests have been handled.	2129	Returns the number of requests currently in the pending state (executed,
261		2130	but not yet processed by poll_cb).
262	Strictly equivalent to:
263
264	IO::AIO::poll_wait, IO::AIO::poll_cb
265	if IO::AIO::nreqs;
266
267	=item IO::AIO::min_parallel $nthreads
268
269	Set the minimum number of AIO threads to C<$nthreads>. The default is
270	C<1>, which means a single asynchronous operation can be done at one time
271	(the number of outstanding operations, however, is unlimited).
272
273	It is recommended to keep the number of threads low, as some Linux
274	kernel versions will scale negatively with the number of threads (higher
275	parallelity => MUCH higher latency). With current Linux 2.6 versions, 4-32
276	threads should be fine.
277
278	Under normal circumstances you don't need to call this function, as this
279	module automatically starts some threads (the exact number might change,
280	and is currently 4).
281
282	=item IO::AIO::max_parallel $nthreads
283
284	Sets the maximum number of AIO threads to C<$nthreads>. If more than
285	the specified number of threads are currently running, kill them. This
286	function blocks until the limit is reached.
287
288	This module automatically runs C<max_parallel 0> at program end, to ensure
289	that all threads are killed and that there are no outstanding requests.
290
291	Under normal circumstances you don't need to call this function.
292
293	=item $oldnreqs = IO::AIO::max_outstanding $nreqs
294
295	Sets the maximum number of outstanding requests to C<$nreqs>. If you
296	try to queue up more than this number of requests, the caller will block until
297	some requests have been handled.
298
299	The default is very large, so normally there is no practical limit. If you
300	queue up many requests in a loop it it often improves speed if you set
301	this to a relatively low number, such as C<100>.
302
303	Under normal circumstances you don't need to call this function.
304		2131
305	=back	2132	=back
306		2133
		2134
		2135	=head3 SUBSECOND STAT TIME ACCESS
		2136
		2137	Both C<aio_stat>/C<aio_lstat> and perl's C<stat>/C<lstat> functions can
		2138	generally find access/modification and change times with subsecond time
		2139	accuracy of the system supports it, but perl's built-in functions only
		2140	return the integer part.
		2141
		2142	The following functions return the timestamps of the most recent
		2143	stat with subsecond precision on most systems and work both after
		2144	C<aio_stat>/C<aio_lstat> and perl's C<stat>/C<lstat> calls. Their return
		2145	value is only meaningful after a successful C<stat>/C<lstat> call, or
		2146	during/after a successful C<aio_stat>/C<aio_lstat> callback.
		2147
		2148	This is similar to the L<Time::HiRes> C<stat> functions, but can return
		2149	full resolution without rounding and work with standard perl C<stat>,
		2150	alleviating the need to call the special C<Time::HiRes> functions, which
		2151	do not act like their perl counterparts.
		2152
		2153	On operating systems or file systems where subsecond time resolution is
		2154	not supported or could not be detected, a fractional part of C<0> is
		2155	returned, so it is always safe to call these functions.
		2156
		2157	=over 4
		2158
		2159	=item $seconds = IO::AIO::st_atime, IO::AIO::st_mtime, IO::AIO::st_ctime, IO::AIO::st_btime
		2160
		2161	Return the access, modication, change or birth time, respectively,
		2162	including fractional part. Due to the limited precision of floating point,
		2163	the accuracy on most platforms is only a bit better than milliseconds
		2164	for times around now - see the I<nsec> function family, below, for full
		2165	accuracy.
		2166
		2167	File birth time is only available when the OS and perl support it (on
		2168	FreeBSD and NetBSD at the time of this writing, although support is
		2169	adaptive, so if your OS/perl gains support, IO::AIO can take advantage of
		2170	it). On systems where it isn't available, C<0> is currently returned, but
		2171	this might change to C<undef> in a future version.
		2172
		2173	=item ($atime, $mtime, $ctime, $btime, ...) = IO::AIO::st_xtime
		2174
		2175	Returns access, modification, change and birth time all in one go, and
		2176	maybe more times in the future version.
		2177
		2178	=item $nanoseconds = IO::AIO::st_atimensec, IO::AIO::st_mtimensec, IO::AIO::st_ctimensec, IO::AIO::st_btimensec
		2179
		2180	Return the fractional access, modifcation, change or birth time, in nanoseconds,
		2181	as an integer in the range C<0> to C<999999999>.
		2182
		2183	Note that no accessors are provided for access, modification and
		2184	change times - you need to get those from C<stat _> if required (C<int
		2185	IO::AIO::st_atime> and so on will I<not> generally give you the correct
		2186	value).
		2187
		2188	=item $seconds = IO::AIO::st_btimesec
		2189
		2190	The (integral) seconds part of the file birth time, if available.
		2191
		2192	=item ($atime, $mtime, $ctime, $btime, ...) = IO::AIO::st_xtimensec
		2193
		2194	Like the functions above, but returns all four times in one go (and maybe
		2195	more in future versions).
		2196
		2197	=item $counter = IO::AIO::st_gen
		2198
		2199	Returns the generation counter (in practice this is just a random number)
		2200	of the file. This is only available on platforms which have this member in
		2201	their C<struct stat> (most BSDs at the time of this writing) and generally
		2202	only to the root usert. If unsupported, C<0> is returned, but this might
		2203	change to C<undef> in a future version.
		2204
		2205	=back
		2206
		2207	Example: print the high resolution modification time of F</etc>, using
		2208	C<stat>, and C<IO::AIO::aio_stat>.
		2209
		2210	if (stat "/etc") {
		2211	printf "stat(/etc) mtime: %f\n", IO::AIO::st_mtime;
		2212	}
		2213
		2214	IO::AIO::aio_stat "/etc", sub {
		2215	$_[0]
		2216	and return;
		2217
		2218	printf "aio_stat(/etc) mtime: %d.%09d\n", (stat _)[9], IO::AIO::st_mtimensec;
		2219	};
		2220
		2221	IO::AIO::flush;
		2222
		2223	Output of the awbove on my system, showing reduced and full accuracy:
		2224
		2225	stat(/etc) mtime: 1534043702.020808
		2226	aio_stat(/etc) mtime: 1534043702.020807792
		2227
		2228
		2229	=head3 MISCELLANEOUS FUNCTIONS
		2230
		2231	IO::AIO implements some functions that are useful when you want to use
		2232	some "Advanced I/O" function not available to in Perl, without going the
		2233	"Asynchronous I/O" route. Many of these have an asynchronous C<aio_*>
		2234	counterpart.
		2235
		2236	=over 4
		2237
		2238	=item $retval = IO::AIO::fexecve $fh, $argv, $envp
		2239
		2240	A more-or-less direct equivalent to the POSIX C<fexecve> functions, which
		2241	allows you to specify the program to be executed via a file descriptor (or
		2242	handle). Returns C<-1> and sets errno to C<ENOSYS> if not available.
		2243
		2244	=item $numfd = IO::AIO::get_fdlimit
		2245
		2246	Tries to find the current file descriptor limit and returns it, or
		2247	C<undef> and sets C<$!> in case of an error. The limit is one larger than
		2248	the highest valid file descriptor number.
		2249
		2250	=item IO::AIO::min_fdlimit [$numfd]
		2251
		2252	Try to increase the current file descriptor limit(s) to at least C<$numfd>
		2253	by changing the soft or hard file descriptor resource limit. If C<$numfd>
		2254	is missing, it will try to set a very high limit, although this is not
		2255	recommended when you know the actual minimum that you require.
		2256
		2257	If the limit cannot be raised enough, the function makes a best-effort
		2258	attempt to increase the limit as much as possible, using various
		2259	tricks, while still failing. You can query the resulting limit using
		2260	C<IO::AIO::get_fdlimit>.
		2261
		2262	If an error occurs, returns C<undef> and sets C<$!>, otherwise returns
		2263	true.
		2264
		2265	=item IO::AIO::sendfile $ofh, $ifh, $offset, $count
		2266
		2267	Calls the C<eio_sendfile_sync> function, which is like C<aio_sendfile>,
		2268	but is blocking (this makes most sense if you know the input data is
		2269	likely cached already and the output filehandle is set to non-blocking
		2270	operations).
		2271
		2272	Returns the number of bytes copied, or C<-1> on error.
		2273
		2274	=item IO::AIO::fadvise $fh, $offset, $len, $advice
		2275
		2276	Simply calls the C<posix_fadvise> function (see its
		2277	manpage for details). The following advice constants are
		2278	available: C<IO::AIO::FADV_NORMAL>, C<IO::AIO::FADV_SEQUENTIAL>,
		2279	C<IO::AIO::FADV_RANDOM>, C<IO::AIO::FADV_NOREUSE>,
		2280	C<IO::AIO::FADV_WILLNEED>, C<IO::AIO::FADV_DONTNEED>.
		2281
		2282	On systems that do not implement C<posix_fadvise>, this function returns
		2283	ENOSYS, otherwise the return value of C<posix_fadvise>.
		2284
		2285	=item IO::AIO::madvise $scalar, $offset, $len, $advice
		2286
		2287	Simply calls the C<posix_madvise> function (see its
		2288	manpage for details). The following advice constants are
		2289	available: C<IO::AIO::MADV_NORMAL>, C<IO::AIO::MADV_SEQUENTIAL>,
		2290	C<IO::AIO::MADV_RANDOM>, C<IO::AIO::MADV_WILLNEED>,
		2291	C<IO::AIO::MADV_DONTNEED>.
		2292
		2293	If C<$offset> is negative, counts from the end. If C<$length> is negative,
		2294	the remaining length of the C<$scalar> is used. If possible, C<$length>
		2295	will be reduced to fit into the C<$scalar>.
		2296
		2297	On systems that do not implement C<posix_madvise>, this function returns
		2298	ENOSYS, otherwise the return value of C<posix_madvise>.
		2299
		2300	=item IO::AIO::mprotect $scalar, $offset, $len, $protect
		2301
		2302	Simply calls the C<mprotect> function on the preferably AIO::mmap'ed
		2303	$scalar (see its manpage for details). The following protect
		2304	constants are available: C<IO::AIO::PROT_NONE>, C<IO::AIO::PROT_READ>,
		2305	C<IO::AIO::PROT_WRITE>, C<IO::AIO::PROT_EXEC>.
		2306
		2307	If C<$offset> is negative, counts from the end. If C<$length> is negative,
		2308	the remaining length of the C<$scalar> is used. If possible, C<$length>
		2309	will be reduced to fit into the C<$scalar>.
		2310
		2311	On systems that do not implement C<mprotect>, this function returns
		2312	ENOSYS, otherwise the return value of C<mprotect>.
		2313
		2314	=item IO::AIO::mmap $scalar, $length, $prot, $flags, $fh[, $offset]
		2315
		2316	Memory-maps a file (or anonymous memory range) and attaches it to the
		2317	given C<$scalar>, which will act like a string scalar. Returns true on
		2318	success, and false otherwise.
		2319
		2320	The scalar must exist, but its contents do not matter - this means you
		2321	cannot use a nonexistant array or hash element. When in doubt, C<undef>
		2322	the scalar first.
		2323
		2324	The only operations allowed on the mmapped scalar are C<substr>/C<vec>,
		2325	which don't change the string length, and most read-only operations such
		2326	as copying it or searching it with regexes and so on.
		2327
		2328	Anything else is unsafe and will, at best, result in memory leaks.
		2329
		2330	The memory map associated with the C<$scalar> is automatically removed
		2331	when the C<$scalar> is undef'd or destroyed, or when the C<IO::AIO::mmap>
		2332	or C<IO::AIO::munmap> functions are called on it.
		2333
		2334	This calls the C<mmap>(2) function internally. See your system's manual
		2335	page for details on the C<$length>, C<$prot> and C<$flags> parameters.
		2336
		2337	The C<$length> must be larger than zero and smaller than the actual
		2338	filesize.
		2339
		2340	C<$prot> is a combination of C<IO::AIO::PROT_NONE>, C<IO::AIO::PROT_EXEC>,
		2341	C<IO::AIO::PROT_READ> and/or C<IO::AIO::PROT_WRITE>,
		2342
		2343	C<$flags> can be a combination of
		2344	C<IO::AIO::MAP_SHARED> or
		2345	C<IO::AIO::MAP_PRIVATE>,
		2346	or a number of system-specific flags (when not available, the are C<0>):
		2347	C<IO::AIO::MAP_ANONYMOUS> (which is set to C<MAP_ANON> if your system only provides this constant),
		2348	C<IO::AIO::MAP_LOCKED>,
		2349	C<IO::AIO::MAP_NORESERVE>,
		2350	C<IO::AIO::MAP_POPULATE>,
		2351	C<IO::AIO::MAP_NONBLOCK>,
		2352	C<IO::AIO::MAP_FIXED>,
		2353	C<IO::AIO::MAP_GROWSDOWN>,
		2354	C<IO::AIO::MAP_32BIT>,
		2355	C<IO::AIO::MAP_HUGETLB>,
		2356	C<IO::AIO::MAP_STACK>,
		2357	C<IO::AIO::MAP_FIXED_NOREPLACE>,
		2358	C<IO::AIO::MAP_SHARED_VALIDATE>,
		2359	C<IO::AIO::MAP_SYNC> or
		2360	C<IO::AIO::MAP_UNINITIALIZED>.
		2361
		2362	If C<$fh> is C<undef>, then a file descriptor of C<-1> is passed.
		2363
		2364	C<$offset> is the offset from the start of the file - it generally must be
		2365	a multiple of C<IO::AIO::PAGESIZE> and defaults to C<0>.
		2366
		2367	Example:
		2368
		2369	use Digest::MD5;
		2370	use IO::AIO;
		2371
		2372	open my $fh, "<verybigfile"
		2373	or die "$!";
		2374
		2375	IO::AIO::mmap my $data, -s $fh, IO::AIO::PROT_READ, IO::AIO::MAP_SHARED, $fh
		2376	or die "verybigfile: $!";
		2377
		2378	my $fast_md5 = md5 $data;
		2379
		2380	=item IO::AIO::munmap $scalar
		2381
		2382	Removes a previous mmap and undefines the C<$scalar>.
		2383
		2384	=item IO::AIO::mremap $scalar, $new_length, $flags = MREMAP_MAYMOVE[, $new_address = 0]
		2385
		2386	Calls the Linux-specific mremap(2) system call. The C<$scalar> must have
		2387	been mapped by C<IO::AIO::mmap>, and C<$flags> must currently either be
		2388	C<0> or C<IO::AIO::MREMAP_MAYMOVE>.
		2389
		2390	Returns true if successful, and false otherwise. If the underlying mmapped
		2391	region has changed address, then the true value has the numerical value
		2392	C<1>, otherwise it has the numerical value C<0>:
		2393
		2394	my $success = IO::AIO::mremap $mmapped, 8192, IO::AIO::MREMAP_MAYMOVE
		2395	or die "mremap: $!";
		2396
		2397	if ($success*1) {
		2398	warn "scalar has chanegd address in memory\n";
		2399	}
		2400
		2401	C<IO::AIO::MREMAP_FIXED> and the C<$new_address> argument are currently
		2402	implemented, but not supported and might go away in a future version.
		2403
		2404	On systems where this call is not supported or is not emulated, this call
		2405	returns falls and sets C<$!> to C<ENOSYS>.
		2406
		2407	=item IO::AIO::mlockall $flags
		2408
		2409	Calls the C<eio_mlockall_sync> function, which is like C<aio_mlockall>,
		2410	but is blocking.
		2411
		2412	=item IO::AIO::munlock $scalar, $offset = 0, $length = undef
		2413
		2414	Calls the C<munlock> function, undoing the effects of a previous
		2415	C<aio_mlock> call (see its description for details).
		2416
		2417	=item IO::AIO::munlockall
		2418
		2419	Calls the C<munlockall> function.
		2420
		2421	On systems that do not implement C<munlockall>, this function returns
		2422	ENOSYS, otherwise the return value of C<munlockall>.
		2423
		2424	=item $fh = IO::AIO::accept4 $r_fh, $sockaddr, $sockaddr_maxlen, $flags
		2425
		2426	Uses the GNU/Linux C<accept4(2)> syscall, if available, to accept a socket
		2427	and return the new file handle on success, or sets C<$!> and returns
		2428	C<undef> on error.
		2429
		2430	The remote name of the new socket will be stored in C<$sockaddr>, which
		2431	will be extended to allow for at least C<$sockaddr_maxlen> octets. If the
		2432	socket name does not fit into C<$sockaddr_maxlen> octets, this is signaled
		2433	by returning a longer string in C<$sockaddr>, which might or might not be
		2434	truncated.
		2435
		2436	To accept name-less sockets, use C<undef> for C<$sockaddr> and C<0> for
		2437	C<$sockaddr_maxlen>.
		2438
		2439	The main reasons to use this syscall rather than portable C<accept(2)>
		2440	are that you can specify C<SOCK_NONBLOCK> and/or C<SOCK_CLOEXEC>
		2441	flags and you can accept name-less sockets by specifying C<0> for
		2442	C<$sockaddr_maxlen>, which is sadly not possible with perl's interface to
		2443	C<accept>.
		2444
		2445	=item IO::AIO::splice $r_fh, $r_off, $w_fh, $w_off, $length, $flags
		2446
		2447	Calls the GNU/Linux C<splice(2)> syscall, if available. If C<$r_off> or
		2448	C<$w_off> are C<undef>, then C<NULL> is passed for these, otherwise they
		2449	should be the file offset.
		2450
		2451	C<$r_fh> and C<$w_fh> should not refer to the same file, as splice might
		2452	silently corrupt the data in this case.
		2453
		2454	The following symbol flag values are available: C<IO::AIO::SPLICE_F_MOVE>,
		2455	C<IO::AIO::SPLICE_F_NONBLOCK>, C<IO::AIO::SPLICE_F_MORE> and
		2456	C<IO::AIO::SPLICE_F_GIFT>.
		2457
		2458	See the C<splice(2)> manpage for details.
		2459
		2460	=item IO::AIO::tee $r_fh, $w_fh, $length, $flags
		2461
		2462	Calls the GNU/Linux C<tee(2)> syscall, see its manpage and the
		2463	description for C<IO::AIO::splice> above for details.
		2464
		2465	=item $actual_size = IO::AIO::pipesize $r_fh[, $new_size]
		2466
		2467	Attempts to query or change the pipe buffer size. Obviously works only
		2468	on pipes, and currently works only on GNU/Linux systems, and fails with
		2469	C<-1>/C<ENOSYS> everywhere else. If anybody knows how to influence pipe buffer
		2470	size on other systems, drop me a note.
		2471
		2472	=item ($rfh, $wfh) = IO::AIO::pipe2 [$flags]
		2473
		2474	This is a direct interface to the Linux L<pipe2(2)> system call. If
		2475	C<$flags> is missing or C<0>, then this should be the same as a call to
		2476	perl's built-in C<pipe> function and create a new pipe, and works on
		2477	systems that lack the pipe2 syscall. On win32, this case invokes C<_pipe
		2478	(..., 4096, O_BINARY)>.
		2479
		2480	If C<$flags> is non-zero, it tries to invoke the pipe2 system call with
		2481	the given flags (Linux 2.6.27, glibc 2.9).
		2482
		2483	On success, the read and write file handles are returned.
		2484
		2485	On error, nothing will be returned. If the pipe2 syscall is missing and
		2486	C<$flags> is non-zero, fails with C<ENOSYS>.
		2487
		2488	Please refer to L<pipe2(2)> for more info on the C<$flags>, but at the
		2489	time of this writing, C<IO::AIO::O_CLOEXEC>, C<IO::AIO::O_NONBLOCK> and
		2490	C<IO::AIO::O_DIRECT> (Linux 3.4, for packet-based pipes) were supported.
		2491
		2492	Example: create a pipe race-free w.r.t. threads and fork:
		2493
		2494	my ($rfh, $wfh) = IO::AIO::pipe2 IO::AIO::O_CLOEXEC
		2495	or die "pipe2: $!\n";
		2496
		2497	=item $fh = IO::AIO::memfd_create $pathname[, $flags]
		2498
		2499	This is a direct interface to the Linux L<memfd_create(2)> system
		2500	call. The (unhelpful) default for C<$flags> is C<0>, but your default
		2501	should be C<IO::AIO::MFD_CLOEXEC>.
		2502
		2503	On success, the new memfd filehandle is returned, otherwise returns
		2504	C<undef>. If the memfd_create syscall is missing, fails with C<ENOSYS>.
		2505
		2506	Please refer to L<memfd_create(2)> for more info on this call.
		2507
		2508	The following C<$flags> values are available: C<IO::AIO::MFD_CLOEXEC>,
		2509	C<IO::AIO::MFD_ALLOW_SEALING>, C<IO::AIO::MFD_HUGETLB>,
		2510	C<IO::AIO::MFD_HUGETLB_2MB> and C<IO::AIO::MFD_HUGETLB_1GB>.
		2511
		2512	Example: create a new memfd.
		2513
		2514	my $fh = IO::AIO::memfd_create "somenameforprocfd", IO::AIO::MFD_CLOEXEC
		2515	or die "memfd_create: $!\n";
		2516
		2517	=item $fh = IO::AIO::pidfd_open $pid[, $flags]
		2518
		2519	This is an interface to the Linux L<pidfd_open(2)> system call. The
		2520	default for C<$flags> is C<0>.
		2521
		2522	On success, a new pidfd filehandle is returned (that is already set to
		2523	close-on-exec), otherwise returns C<undef>. If the syscall is missing,
		2524	fails with C<ENOSYS>.
		2525
		2526	Example: open pid 6341 as pidfd.
		2527
		2528	my $fh = IO::AIO::pidfd_open 6341
		2529	or die "pidfd_open: $!\n";
		2530
		2531	=item $status = IO::AIO::pidfd_send_signal $pidfh, $signal[, $siginfo[, $flags]]
		2532
		2533	This is an interface to the Linux L<pidfd_send_signal> system call. The
		2534	default for C<$siginfo> is C<undef> and the default for C<$flags> is C<0>.
		2535
		2536	Returns the system call status. If the syscall is missing, fails with
		2537	C<ENOSYS>.
		2538
		2539	When specified, C<$siginfo> must be a reference to a hash with one or more
		2540	of the following members:
		2541
		2542	=over
		2543
		2544	=item code - the C<si_code> member
		2545
		2546	=item pid - the C<si_pid> member
		2547
		2548	=item uid - the C<si_uid> member
		2549
		2550	=item value_int - the C<si_value.sival_int> member
		2551
		2552	=item value_ptr - the C<si_value.sival_ptr> member, specified as an integer
		2553
		2554	=back
		2555
		2556	Example: send a SIGKILL to the specified process.
		2557
		2558	my $status = IO::AIO::pidfd_send_signal $pidfh, 9, undef
		2559	and die "pidfd_send_signal: $!\n";
		2560
		2561	Example: send a SIGKILL to the specified process with extra data.
		2562
		2563	my $status = IO::AIO::pidfd_send_signal $pidfh, 9, { code => -1, value_int => 7 }
		2564	and die "pidfd_send_signal: $!\n";
		2565
		2566	=item $fh = IO::AIO::pidfd_getfd $pidfh, $targetfd[, $flags]
		2567
		2568	This is an interface to the Linux L<pidfd_getfd> system call. The default
		2569	for C<$flags> is C<0>.
		2570
		2571	On success, returns a dup'ed copy of the target file descriptor (specified
		2572	as an integer) returned (that is already set to close-on-exec), otherwise
		2573	returns C<undef>. If the syscall is missing, fails with C<ENOSYS>.
		2574
		2575	Example: get a copy of standard error of another process and print soemthing to it.
		2576
		2577	my $errfh = IO::AIO::pidfd_getfd $pidfh, 2
		2578	or die "pidfd_getfd: $!\n";
		2579	print $errfh "stderr\n";
		2580
		2581	=item $fh = IO::AIO::eventfd [$initval, [$flags]]
		2582
		2583	This is a direct interface to the Linux L<eventfd(2)> system call. The
		2584	(unhelpful) defaults for C<$initval> and C<$flags> are C<0> for both.
		2585
		2586	On success, the new eventfd filehandle is returned, otherwise returns
		2587	C<undef>. If the eventfd syscall is missing, fails with C<ENOSYS>.
		2588
		2589	Please refer to L<eventfd(2)> for more info on this call.
		2590
		2591	The following symbol flag values are available: C<IO::AIO::EFD_CLOEXEC>,
		2592	C<IO::AIO::EFD_NONBLOCK> and C<IO::AIO::EFD_SEMAPHORE> (Linux 2.6.30).
		2593
		2594	Example: create a new eventfd filehandle:
		2595
		2596	$fh = IO::AIO::eventfd 0, IO::AIO::EFD_CLOEXEC
		2597	or die "eventfd: $!\n";
		2598
		2599	=item $fh = IO::AIO::timerfd_create $clockid[, $flags]
		2600
		2601	This is a direct interface to the Linux L<timerfd_create(2)> system
		2602	call. The (unhelpful) default for C<$flags> is C<0>, but your default
		2603	should be C<IO::AIO::TFD_CLOEXEC>.
		2604
		2605	On success, the new timerfd filehandle is returned, otherwise returns
		2606	C<undef>. If the timerfd_create syscall is missing, fails with C<ENOSYS>.
		2607
		2608	Please refer to L<timerfd_create(2)> for more info on this call.
		2609
		2610	The following C<$clockid> values are
		2611	available: C<IO::AIO::CLOCK_REALTIME>, C<IO::AIO::CLOCK_MONOTONIC>
		2612	C<IO::AIO::CLOCK_CLOCK_BOOTTIME> (Linux 3.15)
		2613	C<IO::AIO::CLOCK_CLOCK_REALTIME_ALARM> (Linux 3.11) and
		2614	C<IO::AIO::CLOCK_CLOCK_BOOTTIME_ALARM> (Linux 3.11).
		2615
		2616	The following C<$flags> values are available (Linux
		2617	2.6.27): C<IO::AIO::TFD_NONBLOCK> and C<IO::AIO::TFD_CLOEXEC>.
		2618
		2619	Example: create a new timerfd and set it to one-second repeated alarms,
		2620	then wait for two alarms:
		2621
		2622	my $fh = IO::AIO::timerfd_create IO::AIO::CLOCK_BOOTTIME, IO::AIO::TFD_CLOEXEC
		2623	or die "timerfd_create: $!\n";
		2624
		2625	defined IO::AIO::timerfd_settime $fh, 0, 1, 1
		2626	or die "timerfd_settime: $!\n";
		2627
		2628	for (1..2) {
		2629	8 == sysread $fh, my $buf, 8
		2630	or die "timerfd read failure\n";
		2631
		2632	printf "number of expirations (likely 1): %d\n",
		2633	unpack "Q", $buf;
		2634	}
		2635
		2636	=item ($cur_interval, $cur_value) = IO::AIO::timerfd_settime $fh, $flags, $new_interval, $nbw_value
		2637
		2638	This is a direct interface to the Linux L<timerfd_settime(2)> system
		2639	call. Please refer to its manpage for more info on this call.
		2640
		2641	The new itimerspec is specified using two (possibly fractional) second
		2642	values, C<$new_interval> and C<$new_value>).
		2643
		2644	On success, the current interval and value are returned (as per
		2645	C<timerfd_gettime>). On failure, the empty list is returned.
		2646
		2647	The following C<$flags> values are
		2648	available: C<IO::AIO::TFD_TIMER_ABSTIME> and
		2649	C<IO::AIO::TFD_TIMER_CANCEL_ON_SET>.
		2650
		2651	See C<IO::AIO::timerfd_create> for a full example.
		2652
		2653	=item ($cur_interval, $cur_value) = IO::AIO::timerfd_gettime $fh
		2654
		2655	This is a direct interface to the Linux L<timerfd_gettime(2)> system
		2656	call. Please refer to its manpage for more info on this call.
		2657
		2658	On success, returns the current values of interval and value for the given
		2659	timerfd (as potentially fractional second values). On failure, the empty
		2660	list is returned.
		2661
		2662	=back
		2663
307	=cut	2664	=cut
308		2665
309	# support function to convert a fd into a perl filehandle
310	sub _fd2fh {
311	return undef if $_[0] < 0;
312
313	# try to be perl5.6-compatible
314	local *AIO_FH;
315	open AIO_FH, "+<&=$_[0]"
316	or return undef;
317
318	*AIO_FH
319	}
320
321	min_parallel 4;	2666	min_parallel 8;
322		2667
323	END {	2668	END { flush }
324	max_parallel 0;
325	}
326		2669
327	1;	2670	1;
328		2671
		2672	=head1 EVENT LOOP INTEGRATION
		2673
		2674	It is recommended to use L<AnyEvent::AIO> to integrate IO::AIO
		2675	automatically into many event loops:
		2676
		2677	# AnyEvent integration (EV, Event, Glib, Tk, POE, urxvt, pureperl...)
		2678	use AnyEvent::AIO;
		2679
		2680	You can also integrate IO::AIO manually into many event loops, here are
		2681	some examples of how to do this:
		2682
		2683	# EV integration
		2684	my $aio_w = EV::io IO::AIO::poll_fileno, EV::READ, \&IO::AIO::poll_cb;
		2685
		2686	# Event integration
		2687	Event->io (fd => IO::AIO::poll_fileno,
		2688	poll => 'r',
		2689	cb => \&IO::AIO::poll_cb);
		2690
		2691	# Glib/Gtk2 integration
		2692	add_watch Glib::IO IO::AIO::poll_fileno,
		2693	in => sub { IO::AIO::poll_cb; 1 };
		2694
		2695	# Tk integration
		2696	Tk::Event::IO->fileevent (IO::AIO::poll_fileno, "",
		2697	readable => \&IO::AIO::poll_cb);
		2698
		2699	# Danga::Socket integration
		2700	Danga::Socket->AddOtherFds (IO::AIO::poll_fileno =>
		2701	\&IO::AIO::poll_cb);
		2702
		2703	=head2 FORK BEHAVIOUR
		2704
		2705	Usage of pthreads in a program changes the semantics of fork
		2706	considerably. Specifically, only async-safe functions can be called after
		2707	fork. Perl doesn't know about this, so in general, you cannot call fork
		2708	with defined behaviour in perl if pthreads are involved. IO::AIO uses
		2709	pthreads, so this applies, but many other extensions and (for inexplicable
		2710	reasons) perl itself often is linked against pthreads, so this limitation
		2711	applies to quite a lot of perls.
		2712
		2713	This module no longer tries to fight your OS, or POSIX. That means IO::AIO
		2714	only works in the process that loaded it. Forking is fully supported, but
		2715	using IO::AIO in the child is not.
		2716
		2717	You might get around by not I<using> IO::AIO before (or after)
		2718	forking. You could also try to call the L<IO::AIO::reinit> function in the
		2719	child:
		2720
		2721	=over 4
		2722
		2723	=item IO::AIO::reinit
		2724
		2725	Abandons all current requests and I/O threads and simply reinitialises all
		2726	data structures. This is not an operation supported by any standards, but
		2727	happens to work on GNU/Linux and some newer BSD systems.
		2728
		2729	The only reasonable use for this function is to call it after forking, if
		2730	C<IO::AIO> was used in the parent. Calling it while IO::AIO is active in
		2731	the process will result in undefined behaviour. Calling it at any time
		2732	will also result in any undefined (by POSIX) behaviour.
		2733
		2734	=back
		2735
		2736	=head2 LINUX-SPECIFIC CALLS
		2737
		2738	When a call is documented as "linux-specific" then this means it
		2739	originated on GNU/Linux. C<IO::AIO> will usually try to autodetect the
		2740	availability and compatibility of such calls regardless of the platform
		2741	it is compiled on, so platforms such as FreeBSD which often implement
		2742	these calls will work. When in doubt, call them and see if they fail wth
		2743	C<ENOSYS>.
		2744
		2745	=head2 MEMORY USAGE
		2746
		2747	Per-request usage:
		2748
		2749	Each aio request uses - depending on your architecture - around 100-200
		2750	bytes of memory. In addition, stat requests need a stat buffer (possibly
		2751	a few hundred bytes), readdir requires a result buffer and so on. Perl
		2752	scalars and other data passed into aio requests will also be locked and
		2753	will consume memory till the request has entered the done state.
		2754
		2755	This is not awfully much, so queuing lots of requests is not usually a
		2756	problem.
		2757
		2758	Per-thread usage:
		2759
		2760	In the execution phase, some aio requests require more memory for
		2761	temporary buffers, and each thread requires a stack and other data
		2762	structures (usually around 16k-128k, depending on the OS).
		2763
		2764	=head1 KNOWN BUGS
		2765
		2766	Known bugs will be fixed in the next release :)
		2767
		2768	=head1 KNOWN ISSUES
		2769
		2770	Calls that try to "import" foreign memory areas (such as C<IO::AIO::mmap>
		2771	or C<IO::AIO::aio_slurp>) do not work with generic lvalues, such as
		2772	non-created hash slots or other scalars I didn't think of. It's best to
		2773	avoid such and either use scalar variables or making sure that the scalar
		2774	exists (e.g. by storing C<undef>) and isn't "funny" (e.g. tied).
		2775
		2776	I am not sure anything can be done about this, so this is considered a
		2777	known issue, rather than a bug.
		2778
329	=head1 SEE ALSO	2779	=head1 SEE ALSO
330		2780
331	L<Coro>, L<Linux::AIO>.	2781	L<AnyEvent::AIO> for easy integration into event loops, L<Coro::AIO> for a
		2782	more natural syntax and L<IO::FDPass> for file descriptor passing.
332		2783
333	=head1 AUTHOR	2784	=head1 AUTHOR
334		2785
335	Marc Lehmann <schmorp@schmorp.de>	2786	Marc Lehmann <schmorp@schmorp.de>
336	http://home.schmorp.de/	2787	http://home.schmorp.de/

Diff Legend

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