ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/IO-AIO/AIO.pm

Comparing IO-AIO/AIO.pm (file contents):
Revision 1.23 by root, Fri Jul 22 08:25:22 2005 UTC vs.
Revision 1.307 by root, Wed Feb 26 15:32:59 2020 UTC

1	=head1 NAME	1	=head1 NAME
2		2
3	IO::AIO - Asynchronous Input/Output	3	IO::AIO - Asynchronous/Advanced Input/Output
4		4
5	=head1 SYNOPSIS	5	=head1 SYNOPSIS
6		6
7	use IO::AIO;	7	use IO::AIO;
8		8
9	aio_open "/etc/passwd", O_RDONLY, 0, sub {	9	aio_open "/etc/passwd", IO::AIO::O_RDONLY, 0, sub {
10	my ($fh) = @_;	10	my $fh = shift
		11	or die "/etc/passwd: $!";
11	...	12	...
12	};	13	};
13		14
14	aio_unlink "/tmp/file", sub { };	15	aio_unlink "/tmp/file", sub { };
15		16
16	aio_read $fh, 30000, 1024, $buffer, 0, sub {	17	aio_read $fh, 30000, 1024, $buffer, 0, sub {
17	$_[0] > 0 or die "read error: $!";	18	$_[0] > 0 or die "read error: $!";
18	};	19	};
19		20
20	# Event	21	# version 2+ has request and group objects
21	Event->io (fd => IO::AIO::poll_fileno,	22	use IO::AIO 2;
22	poll => 'r',
23	cb => \&IO::AIO::poll_cb);
24		23
25	# Glib/Gtk2	24	aioreq_pri 4; # give next request a very high priority
26	add_watch Glib::IO IO::AIO::poll_fileno,	25	my $req = aio_unlink "/tmp/file", sub { };
27	in => sub { IO::AIO::poll_cb; 1 };	26	$req->cancel; # cancel request if still in queue
28		27
29	# Tk	28	my $grp = aio_group sub { print "all stats done\n" };
30	Tk::Event::IO->fileevent (IO::AIO::poll_fileno, "",	29	add $grp aio_stat "..." for ...;
31	readable => \&IO::AIO::poll_cb);
32
33	# Danga::Socket
34	Danga::Socket->AddOtherFds (IO::AIO::poll_fileno =>
35	\&IO::AIO::poll_cb);
36
37		30
38	=head1 DESCRIPTION	31	=head1 DESCRIPTION
39		32
40	This module implements asynchronous I/O using whatever means your	33	This module implements asynchronous I/O using whatever means your
41	operating system supports.	34	operating system supports. It is implemented as an interface to C<libeio>
		35	(L<http://software.schmorp.de/pkg/libeio.html>).
42		36
		37	Asynchronous means that operations that can normally block your program
		38	(e.g. reading from disk) will be done asynchronously: the operation
		39	will still block, but you can do something else in the meantime. This
		40	is extremely useful for programs that need to stay interactive even
		41	when doing heavy I/O (GUI programs, high performance network servers
		42	etc.), but can also be used to easily do operations in parallel that are
		43	normally done sequentially, e.g. stat'ing many files, which is much faster
		44	on a RAID volume or over NFS when you do a number of stat operations
		45	concurrently.
		46
		47	While most of this works on all types of file descriptors (for
		48	example sockets), using these functions on file descriptors that
		49	support nonblocking operation (again, sockets, pipes etc.) is
		50	very inefficient. Use an event loop for that (such as the L<EV>
		51	module): IO::AIO will naturally fit into such an event loop itself.
		52
43	Currently, a number of threads are started that execute your read/writes	53	In this version, a number of threads are started that execute your
44	and signal their completion. You don't need thread support in your libc or	54	requests and signal their completion. You don't need thread support
45	perl, and the threads created by this module will not be visible to the	55	in perl, and the threads created by this module will not be visible
46	pthreads library. In the future, this module might make use of the native	56	to perl. In the future, this module might make use of the native aio
47	aio functions available on many operating systems. However, they are often	57	functions available on many operating systems. However, they are often
48	not well-supported (Linux doesn't allow them on normal files currently,	58	not well-supported or restricted (GNU/Linux doesn't allow them on normal
49	for example), and they would only support aio_read and aio_write, so the	59	files currently, for example), and they would only support aio_read and
50	remaining functionality would have to be implemented using threads anyway.	60	aio_write, so the remaining functionality would have to be implemented
		61	using threads anyway.
51		62
		63	In addition to asynchronous I/O, this module also exports some rather
		64	arcane interfaces, such as C<madvise> or linux's C<splice> system call,
		65	which is why the C<A> in C<AIO> can also mean I<advanced>.
		66
52	Although the module will work with in the presence of other threads, it is	67	Although the module will work in the presence of other (Perl-) threads,
53	currently not reentrant, so use appropriate locking yourself, always call	68	it is currently not reentrant in any way, so use appropriate locking
54	C<poll_cb> from within the same thread, or never call C<poll_cb> (or other	69	yourself, always call C<poll_cb> from within the same thread, or never
55	C<aio_> functions) recursively.	70	call C<poll_cb> (or other C<aio_> functions) recursively.
		71
		72	=head2 EXAMPLE
		73
		74	This is a simple example that uses the EV module and loads
		75	F</etc/passwd> asynchronously:
		76
		77	use EV;
		78	use IO::AIO;
		79
		80	# register the IO::AIO callback with EV
		81	my $aio_w = EV::io IO::AIO::poll_fileno, EV::READ, \&IO::AIO::poll_cb;
		82
		83	# queue the request to open /etc/passwd
		84	aio_open "/etc/passwd", IO::AIO::O_RDONLY, 0, sub {
		85	my $fh = shift
		86	or die "error while opening: $!";
		87
		88	# stat'ing filehandles is generally non-blocking
		89	my $size = -s $fh;
		90
		91	# queue a request to read the file
		92	my $contents;
		93	aio_read $fh, 0, $size, $contents, 0, sub {
		94	$_[0] == $size
		95	or die "short read: $!";
		96
		97	close $fh;
		98
		99	# file contents now in $contents
		100	print $contents;
		101
		102	# exit event loop and program
		103	EV::break;
		104	};
		105	};
		106
		107	# possibly queue up other requests, or open GUI windows,
		108	# check for sockets etc. etc.
		109
		110	# process events as long as there are some:
		111	EV::run;
		112
		113	=head1 REQUEST ANATOMY AND LIFETIME
		114
		115	Every C<aio_*> function creates a request. which is a C data structure not
		116	directly visible to Perl.
		117
		118	If called in non-void context, every request function returns a Perl
		119	object representing the request. In void context, nothing is returned,
		120	which saves a bit of memory.
		121
		122	The perl object is a fairly standard ref-to-hash object. The hash contents
		123	are not used by IO::AIO so you are free to store anything you like in it.
		124
		125	During their existance, aio requests travel through the following states,
		126	in order:
		127
		128	=over 4
		129
		130	=item ready
		131
		132	Immediately after a request is created it is put into the ready state,
		133	waiting for a thread to execute it.
		134
		135	=item execute
		136
		137	A thread has accepted the request for processing and is currently
		138	executing it (e.g. blocking in read).
		139
		140	=item pending
		141
		142	The request has been executed and is waiting for result processing.
		143
		144	While request submission and execution is fully asynchronous, result
		145	processing is not and relies on the perl interpreter calling C<poll_cb>
		146	(or another function with the same effect).
		147
		148	=item result
		149
		150	The request results are processed synchronously by C<poll_cb>.
		151
		152	The C<poll_cb> function will process all outstanding aio requests by
		153	calling their callbacks, freeing memory associated with them and managing
		154	any groups they are contained in.
		155
		156	=item done
		157
		158	Request has reached the end of its lifetime and holds no resources anymore
		159	(except possibly for the Perl object, but its connection to the actual
		160	aio request is severed and calling its methods will either do nothing or
		161	result in a runtime error).
		162
		163	=back
56		164
57	=cut	165	=cut
58		166
59	package IO::AIO;	167	package IO::AIO;
60		168
61	no warnings;	169	use Carp ();
		170
		171	use common::sense;
62		172
63	use base 'Exporter';	173	use base 'Exporter';
64		174
65	use Fcntl ();
66
67	BEGIN {	175	BEGIN {
68	$VERSION = '1.0';	176	our $VERSION = 4.73;
69		177
70	@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
71	aio_fsync aio_fdatasync aio_readahead);	179	aio_stat aio_lstat aio_unlink aio_rmdir aio_readdir aio_readdirx
72	@EXPORT_OK = qw(poll_fileno poll_cb min_parallel max_parallel max_outstanding nreqs);	180	aio_scandir aio_symlink aio_readlink aio_realpath aio_fcntl aio_ioctl
		181	aio_sync aio_fsync aio_syncfs aio_fdatasync aio_sync_file_range
		182	aio_pathsync aio_readahead aio_fiemap aio_allocate
		183	aio_rename aio_rename2 aio_link aio_move aio_copy aio_group
		184	aio_nop aio_mknod aio_load aio_rmtree aio_mkdir aio_chown
		185	aio_chmod aio_utime aio_truncate
		186	aio_msync aio_mtouch aio_mlock aio_mlockall
		187	aio_statvfs
		188	aio_slurp
		189	aio_wd);
		190
		191	our @EXPORT = (@AIO_REQ, qw(aioreq_pri aioreq_nice));
		192	our @EXPORT_OK = qw(poll_fileno poll_cb poll_wait flush
		193	min_parallel max_parallel max_idle idle_timeout
		194	nreqs nready npending nthreads
		195	max_poll_time max_poll_reqs
		196	sendfile fadvise madvise
		197	mmap munmap mremap munlock munlockall);
		198
		199	push @AIO_REQ, qw(aio_busy); # not exported
		200
		201	@IO::AIO::GRP::ISA = 'IO::AIO::REQ';
73		202
74	require XSLoader;	203	require XSLoader;
75	XSLoader::load IO::AIO, $VERSION;	204	XSLoader::load ("IO::AIO", $VERSION);
76	}	205	}
77		206
78	=head1 FUNCTIONS	207	=head1 FUNCTIONS
79		208
80	=head2 AIO FUNCTIONS	209	=head2 QUICK OVERVIEW
		210
		211	This section simply lists the prototypes most of the functions for
		212	quick reference. See the following sections for function-by-function
		213	documentation.
		214
		215	aio_wd $pathname, $callback->($wd)
		216	aio_open $pathname, $flags, $mode, $callback->($fh)
		217	aio_close $fh, $callback->($status)
		218	aio_seek $fh,$offset,$whence, $callback->($offs)
		219	aio_read $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
		220	aio_write $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
		221	aio_sendfile $out_fh, $in_fh, $in_offset, $length, $callback->($retval)
		222	aio_readahead $fh,$offset,$length, $callback->($retval)
		223	aio_stat $fh_or_path, $callback->($status)
		224	aio_lstat $fh, $callback->($status)
		225	aio_statvfs $fh_or_path, $callback->($statvfs)
		226	aio_utime $fh_or_path, $atime, $mtime, $callback->($status)
		227	aio_chown $fh_or_path, $uid, $gid, $callback->($status)
		228	aio_chmod $fh_or_path, $mode, $callback->($status)
		229	aio_truncate $fh_or_path, $offset, $callback->($status)
		230	aio_allocate $fh, $mode, $offset, $len, $callback->($status)
		231	aio_fiemap $fh, $start, $length, $flags, $count, $cb->(\@extents)
		232	aio_unlink $pathname, $callback->($status)
		233	aio_mknod $pathname, $mode, $dev, $callback->($status)
		234	aio_link $srcpath, $dstpath, $callback->($status)
		235	aio_symlink $srcpath, $dstpath, $callback->($status)
		236	aio_readlink $pathname, $callback->($link)
		237	aio_realpath $pathname, $callback->($path)
		238	aio_rename $srcpath, $dstpath, $callback->($status)
		239	aio_rename2 $srcpath, $dstpath, $flags, $callback->($status)
		240	aio_mkdir $pathname, $mode, $callback->($status)
		241	aio_rmdir $pathname, $callback->($status)
		242	aio_readdir $pathname, $callback->($entries)
		243	aio_readdirx $pathname, $flags, $callback->($entries, $flags)
		244	IO::AIO::READDIR_DENTS IO::AIO::READDIR_DIRS_FIRST
		245	IO::AIO::READDIR_STAT_ORDER IO::AIO::READDIR_FOUND_UNKNOWN
		246	aio_scandir $pathname, $maxreq, $callback->($dirs, $nondirs)
		247	aio_load $pathname, $data, $callback->($status)
		248	aio_copy $srcpath, $dstpath, $callback->($status)
		249	aio_move $srcpath, $dstpath, $callback->($status)
		250	aio_rmtree $pathname, $callback->($status)
		251	aio_fcntl $fh, $cmd, $arg, $callback->($status)
		252	aio_ioctl $fh, $request, $buf, $callback->($status)
		253	aio_sync $callback->($status)
		254	aio_syncfs $fh, $callback->($status)
		255	aio_fsync $fh, $callback->($status)
		256	aio_fdatasync $fh, $callback->($status)
		257	aio_sync_file_range $fh, $offset, $nbytes, $flags, $callback->($status)
		258	aio_pathsync $pathname, $callback->($status)
		259	aio_msync $scalar, $offset = 0, $length = undef, flags = MS_SYNC, $callback->($status)
		260	aio_mtouch $scalar, $offset = 0, $length = undef, flags = 0, $callback->($status)
		261	aio_mlock $scalar, $offset = 0, $length = undef, $callback->($status)
		262	aio_mlockall $flags, $callback->($status)
		263	aio_group $callback->(...)
		264	aio_nop $callback->()
		265
		266	$prev_pri = aioreq_pri [$pri]
		267	aioreq_nice $pri_adjust
		268
		269	IO::AIO::poll_wait
		270	IO::AIO::poll_cb
		271	IO::AIO::poll
		272	IO::AIO::flush
		273	IO::AIO::max_poll_reqs $nreqs
		274	IO::AIO::max_poll_time $seconds
		275	IO::AIO::min_parallel $nthreads
		276	IO::AIO::max_parallel $nthreads
		277	IO::AIO::max_idle $nthreads
		278	IO::AIO::idle_timeout $seconds
		279	IO::AIO::max_outstanding $maxreqs
		280	IO::AIO::nreqs
		281	IO::AIO::nready
		282	IO::AIO::npending
		283	IO::AIO::reinit
		284
		285	$nfd = IO::AIO::get_fdlimit
		286	IO::AIO::min_fdlimit $nfd
		287
		288	IO::AIO::sendfile $ofh, $ifh, $offset, $count
		289	IO::AIO::fadvise $fh, $offset, $len, $advice
		290
		291	IO::AIO::mmap $scalar, $length, $prot, $flags[, $fh[, $offset]]
		292	IO::AIO::munmap $scalar
		293	IO::AIO::mremap $scalar, $new_length, $flags[, $new_address]
		294	IO::AIO::madvise $scalar, $offset, $length, $advice
		295	IO::AIO::mprotect $scalar, $offset, $length, $protect
		296	IO::AIO::munlock $scalar, $offset = 0, $length = undef
		297	IO::AIO::munlockall
		298
		299	# stat extensions
		300	$counter = IO::AIO::st_gen
		301	$seconds = IO::AIO::st_atime, IO::AIO::st_mtime, IO::AIO::st_ctime, IO::AIO::st_btime
		302	($atime, $mtime, $ctime, $btime, ...) = IO::AIO::st_xtime
		303	$nanoseconds = IO::AIO::st_atimensec, IO::AIO::st_mtimensec, IO::AIO::st_ctimensec, IO::AIO::st_btimensec
		304	$seconds = IO::AIO::st_btimesec
		305	($atime, $mtime, $ctime, $btime, ...) = IO::AIO::st_xtimensec
		306
		307	# very much unportable syscalls
		308	IO::AIO::accept4 $r_fh, $sockaddr, $sockaddr_len, $flags
		309	IO::AIO::splice $r_fh, $r_off, $w_fh, $w_off, $length, $flags
		310	IO::AIO::tee $r_fh, $w_fh, $length, $flags
		311	$actual_size = IO::AIO::pipesize $r_fh[, $new_size]
		312	($rfh, $wfh) = IO::AIO::pipe2 [$flags]
		313	$fh = IO::AIO::memfd_create $pathname[, $flags]
		314	$fh = IO::AIO::eventfd [$initval, [$flags]]
		315	$fh = IO::AIO::timerfd_create $clockid[, $flags]
		316	($cur_interval, $cur_value) = IO::AIO::timerfd_settime $fh, $flags, $new_interval, $nbw_value
		317	($cur_interval, $cur_value) = IO::AIO::timerfd_gettime $fh
		318
		319	=head2 API NOTES
81		320
82	All the C<aio_*> calls are more or less thin wrappers around the syscall	321	All the C<aio_*> calls are more or less thin wrappers around the syscall
83	with the same name (sans C<aio_>). The arguments are similar or identical,	322	with the same name (sans C<aio_>). The arguments are similar or identical,
84	and they all accept an additional (and optional) C<$callback> argument	323	and they all accept an additional (and optional) C<$callback> argument
85	which must be a code reference. This code reference will get called with	324	which must be a code reference. This code reference will be called after
86	the syscall return code (e.g. most syscalls return C<-1> on error, unlike	325	the syscall has been executed in an asynchronous fashion. The results
87	perl, which usually delivers "false") as it's sole argument when the given	326	of the request will be passed as arguments to the callback (and, if an
88	syscall has been executed asynchronously.	327	error occured, in C<$!>) - for most requests the syscall return code (e.g.
		328	most syscalls return C<-1> on error, unlike perl, which usually delivers
		329	"false").
		330
		331	Some requests (such as C<aio_readdir>) pass the actual results and
		332	communicate failures by passing C<undef>.
89		333
90	All functions expecting a filehandle keep a copy of the filehandle	334	All functions expecting a filehandle keep a copy of the filehandle
91	internally until the request has finished.	335	internally until the request has finished.
92		336
		337	All functions return request objects of type L<IO::AIO::REQ> that allow
		338	further manipulation of those requests while they are in-flight.
		339
93	The filenames you pass to these routines I<must> be absolute. The reason	340	The pathnames you pass to these routines I<should> be absolute. The
94	for this is that at the time the request is being executed, the current	341	reason for this is that at the time the request is being executed, the
95	working directory could have changed. Alternatively, you can make sure	342	current working directory could have changed. Alternatively, you can
96	that you never change the current working directory.	343	make sure that you never change the current working directory anywhere
		344	in the program and then use relative paths. You can also take advantage
		345	of IO::AIOs working directory abstraction, that lets you specify paths
		346	relative to some previously-opened "working directory object" - see the
		347	description of the C<IO::AIO::WD> class later in this document.
		348
		349	To encode pathnames as octets, either make sure you either: a) always pass
		350	in filenames you got from outside (command line, readdir etc.) without
		351	tinkering, b) are in your native filesystem encoding, c) use the Encode
		352	module and encode your pathnames to the locale (or other) encoding in
		353	effect in the user environment, d) use Glib::filename_from_unicode on
		354	unicode filenames or e) use something else to ensure your scalar has the
		355	correct contents.
		356
		357	This works, btw. independent of the internal UTF-8 bit, which IO::AIO
		358	handles correctly whether it is set or not.
		359
		360	=head2 AIO REQUEST FUNCTIONS
97		361
98	=over 4	362	=over 4
99		363
		364	=item $prev_pri = aioreq_pri [$pri]
		365
		366	Returns the priority value that would be used for the next request and, if
		367	C<$pri> is given, sets the priority for the next aio request.
		368
		369	The default priority is C<0>, the minimum and maximum priorities are C<-4>
		370	and C<4>, respectively. Requests with higher priority will be serviced
		371	first.
		372
		373	The priority will be reset to C<0> after each call to one of the C<aio_*>
		374	functions.
		375
		376	Example: open a file with low priority, then read something from it with
		377	higher priority so the read request is serviced before other low priority
		378	open requests (potentially spamming the cache):
		379
		380	aioreq_pri -3;
		381	aio_open ..., sub {
		382	return unless $_[0];
		383
		384	aioreq_pri -2;
		385	aio_read $_[0], ..., sub {
		386	...
		387	};
		388	};
		389
		390
		391	=item aioreq_nice $pri_adjust
		392
		393	Similar to C<aioreq_pri>, but subtracts the given value from the current
		394	priority, so the effect is cumulative.
		395
		396
100	=item aio_open $pathname, $flags, $mode, $callback	397	=item aio_open $pathname, $flags, $mode, $callback->($fh)
101		398
102	Asynchronously open or create a file and call the callback with a newly	399	Asynchronously open or create a file and call the callback with a newly
103	created filehandle for the file.	400	created filehandle for the file (or C<undef> in case of an error).
104		401
105	The pathname passed to C<aio_open> must be absolute. See API NOTES, above,	402	The pathname passed to C<aio_open> must be absolute. See API NOTES, above,
106	for an explanation.	403	for an explanation.
107		404
108	The C<$flags> argument is a bitmask. See the C<Fcntl> module for a	405	The C<$flags> argument is a bitmask. See the C<Fcntl> module for a
109	list. They are the same as used by C<sysopen>.	406	list. They are the same as used by C<sysopen>.
110		407
111	Likewise, C<$mode> specifies the mode of the newly created file, if it	408	Likewise, C<$mode> specifies the mode of the newly created file, if it
112	didn't exist and C<O_CREAT> has been given, just like perl's C<sysopen>,	409	didn't exist and C<O_CREAT> has been given, just like perl's C<sysopen>,
113	except that it is mandatory (i.e. use C<0> if you don't create new files,	410	except that it is mandatory (i.e. use C<0> if you don't create new files,
114	and C<0666> or C<0777> if you do).	411	and C<0666> or C<0777> if you do). Note that the C<$mode> will be modified
		412	by the umask in effect then the request is being executed, so better never
		413	change the umask.
115		414
116	Example:	415	Example:
117		416
118	aio_open "/etc/passwd", O_RDONLY, 0, sub {	417	aio_open "/etc/passwd", IO::AIO::O_RDONLY, 0, sub {
119	if ($_[0]) {	418	if ($_[0]) {
120	print "open successful, fh is $_[0]\n";	419	print "open successful, fh is $_[0]\n";
121	...	420	...
122	} else {	421	} else {
123	die "open failed: $!\n";	422	die "open failed: $!\n";
124	}	423	}
125	};	424	};
126		425
		426	In addition to all the common open modes/flags (C<O_RDONLY>, C<O_WRONLY>,
		427	C<O_RDWR>, C<O_CREAT>, C<O_TRUNC>, C<O_EXCL> and C<O_APPEND>), the
		428	following POSIX and non-POSIX constants are available (missing ones on
		429	your system are, as usual, C<0>):
		430
		431	C<O_ASYNC>, C<O_DIRECT>, C<O_NOATIME>, C<O_CLOEXEC>, C<O_NOCTTY>, C<O_NOFOLLOW>,
		432	C<O_NONBLOCK>, C<O_EXEC>, C<O_SEARCH>, C<O_DIRECTORY>, C<O_DSYNC>,
		433	C<O_RSYNC>, C<O_SYNC>, C<O_PATH>, C<O_TMPFILE>, C<O_TTY_INIT> and C<O_ACCMODE>.
		434
		435
127	=item aio_close $fh, $callback	436	=item aio_close $fh, $callback->($status)
128		437
129	Asynchronously close a file and call the callback with the result	438	Asynchronously close a file and call the callback with the result
130	code. I<WARNING:> although accepted, you should not pass in a perl	439	code.
131	filehandle here, as perl will likely close the file descriptor another
132	time when the filehandle is destroyed. Normally, you can safely call perls
133	C<close> or just let filehandles go out of scope.
134		440
135	This is supposed to be a bug in the API, so that might change. It's	441	Unfortunately, you can't do this to perl. Perl I<insists> very strongly on
136	therefore best to avoid this function.	442	closing the file descriptor associated with the filehandle itself.
137		443
		444	Therefore, C<aio_close> will not close the filehandle - instead it will
		445	use dup2 to overwrite the file descriptor with the write-end of a pipe
		446	(the pipe fd will be created on demand and will be cached).
		447
		448	Or in other words: the file descriptor will be closed, but it will not be
		449	free for reuse until the perl filehandle is closed.
		450
		451	=cut
		452
		453	=item aio_seek $fh, $offset, $whence, $callback->($offs)
		454
		455	Seeks the filehandle to the new C<$offset>, similarly to perl's
		456	C<sysseek>. The C<$whence> can use the traditional values (C<0> for
		457	C<IO::AIO::SEEK_SET>, C<1> for C<IO::AIO::SEEK_CUR> or C<2> for
		458	C<IO::AIO::SEEK_END>).
		459
		460	The resulting absolute offset will be passed to the callback, or C<-1> in
		461	case of an error.
		462
		463	In theory, the C<$whence> constants could be different than the
		464	corresponding values from L<Fcntl>, but perl guarantees they are the same,
		465	so don't panic.
		466
		467	As a GNU/Linux (and maybe Solaris) extension, also the constants
		468	C<IO::AIO::SEEK_DATA> and C<IO::AIO::SEEK_HOLE> are available, if they
		469	could be found. No guarantees about suitability for use in C<aio_seek> or
		470	Perl's C<sysseek> can be made though, although I would naively assume they
		471	"just work".
		472
138	=item aio_read $fh,$offset,$length, $data,$dataoffset,$callback	473	=item aio_read $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
139		474
140	=item aio_write $fh,$offset,$length, $data,$dataoffset,$callback	475	=item aio_write $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
141		476
142	Reads or writes C<length> bytes from the specified C<fh> and C<offset>	477	Reads or writes C<$length> bytes from or to the specified C<$fh> and
143	into the scalar given by C<data> and offset C<dataoffset> and calls the	478	C<$offset> into the scalar given by C<$data> and offset C<$dataoffset> and
144	callback without the actual number of bytes read (or -1 on error, just	479	calls the callback with the actual number of bytes transferred (or -1 on
145	like the syscall).	480	error, just like the syscall).
		481
		482	C<aio_read> will, like C<sysread>, shrink or grow the C<$data> scalar to
		483	offset plus the actual number of bytes read.
		484
		485	If C<$offset> is undefined, then the current file descriptor offset will
		486	be used (and updated), otherwise the file descriptor offset will not be
		487	changed by these calls.
		488
		489	If C<$length> is undefined in C<aio_write>, use the remaining length of
		490	C<$data>.
		491
		492	If C<$dataoffset> is less than zero, it will be counted from the end of
		493	C<$data>.
		494
		495	The C<$data> scalar I<MUST NOT> be modified in any way while the request
		496	is outstanding. Modifying it can result in segfaults or World War III (if
		497	the necessary/optional hardware is installed).
146		498
147	Example: Read 15 bytes at offset 7 into scalar C<$buffer>, starting at	499	Example: Read 15 bytes at offset 7 into scalar C<$buffer>, starting at
148	offset C<0> within the scalar:	500	offset C<0> within the scalar:
149		501
150	aio_read $fh, 7, 15, $buffer, 0, sub {	502	aio_read $fh, 7, 15, $buffer, 0, sub {
151	$_[0] > 0 or die "read error: $!";	503	$_[0] > 0 or die "read error: $!";
152	print "read $_[0] bytes: <$buffer>\n";	504	print "read $_[0] bytes: <$buffer>\n";
153	};	505	};
154		506
		507
		508	=item aio_sendfile $out_fh, $in_fh, $in_offset, $length, $callback->($retval)
		509
		510	Tries to copy C<$length> bytes from C<$in_fh> to C<$out_fh>. It starts
		511	reading at byte offset C<$in_offset>, and starts writing at the current
		512	file offset of C<$out_fh>. Because of that, it is not safe to issue more
		513	than one C<aio_sendfile> per C<$out_fh>, as they will interfere with each
		514	other. The same C<$in_fh> works fine though, as this function does not
		515	move or use the file offset of C<$in_fh>.
		516
		517	Please note that C<aio_sendfile> can read more bytes from C<$in_fh> than
		518	are written, and there is no way to find out how many more bytes have been
		519	read from C<aio_sendfile> alone, as C<aio_sendfile> only provides the
		520	number of bytes written to C<$out_fh>. Only if the result value equals
		521	C<$length> one can assume that C<$length> bytes have been read.
		522
		523	Unlike with other C<aio_> functions, it makes a lot of sense to use
		524	C<aio_sendfile> on non-blocking sockets, as long as one end (typically
		525	the C<$in_fh>) is a file - the file I/O will then be asynchronous, while
		526	the socket I/O will be non-blocking. Note, however, that you can run
		527	into a trap where C<aio_sendfile> reads some data with readahead, then
		528	fails to write all data, and when the socket is ready the next time, the
		529	data in the cache is already lost, forcing C<aio_sendfile> to again hit
		530	the disk. Explicit C<aio_read> + C<aio_write> let's you better control
		531	resource usage.
		532
		533	This call tries to make use of a native C<sendfile>-like syscall to
		534	provide zero-copy operation. For this to work, C<$out_fh> should refer to
		535	a socket, and C<$in_fh> should refer to an mmap'able file.
		536
		537	If a native sendfile cannot be found or it fails with C<ENOSYS>,
		538	C<EINVAL>, C<ENOTSUP>, C<EOPNOTSUPP>, C<EAFNOSUPPORT>, C<EPROTOTYPE> or
		539	C<ENOTSOCK>, it will be emulated, so you can call C<aio_sendfile> on any
		540	type of filehandle regardless of the limitations of the operating system.
		541
		542	As native sendfile syscalls (as practically any non-POSIX interface hacked
		543	together in a hurry to improve benchmark numbers) tend to be rather buggy
		544	on many systems, this implementation tries to work around some known bugs
		545	in Linux and FreeBSD kernels (probably others, too), but that might fail,
		546	so you really really should check the return value of C<aio_sendfile> -
		547	fewer bytes than expected might have been transferred.
		548
		549
155	=item aio_readahead $fh,$offset,$length, $callback	550	=item aio_readahead $fh,$offset,$length, $callback->($retval)
156
157	Asynchronously reads the specified byte range into the page cache, using
158	the C<readahead> syscall. If that syscall doesn't exist (likely if your OS
159	isn't Linux) the status will be C<-1> and C<$!> is set to C<ENOSYS>.
160		551
161	C<aio_readahead> populates the page cache with data from a file so that	552	C<aio_readahead> populates the page cache with data from a file so that
162	subsequent reads from that file will not block on disk I/O. The C<$offset>	553	subsequent reads from that file will not block on disk I/O. The C<$offset>
163	argument specifies the starting point from which data is to be read and	554	argument specifies the starting point from which data is to be read and
164	C<$length> specifies the number of bytes to be read. I/O is performed in	555	C<$length> specifies the number of bytes to be read. I/O is performed in
165	whole pages, so that offset is effectively rounded down to a page boundary	556	whole pages, so that offset is effectively rounded down to a page boundary
166	and bytes are read up to the next page boundary greater than or equal to	557	and bytes are read up to the next page boundary greater than or equal to
167	(off-set+length). C<aio_readahead> does not read beyond the end of the	558	(off-set+length). C<aio_readahead> does not read beyond the end of the
168	file. The current file offset of the file is left unchanged.	559	file. The current file offset of the file is left unchanged.
169		560
		561	If that syscall doesn't exist (likely if your kernel isn't Linux) it will
		562	be emulated by simply reading the data, which would have a similar effect.
		563
		564
170	=item aio_stat $fh_or_path, $callback	565	=item aio_stat $fh_or_path, $callback->($status)
171		566
172	=item aio_lstat $fh, $callback	567	=item aio_lstat $fh, $callback->($status)
173		568
174	Works like perl's C<stat> or C<lstat> in void context. The callback will	569	Works almost exactly like perl's C<stat> or C<lstat> in void context. The
175	be called after the stat and the results will be available using C<stat _>	570	callback will be called after the stat and the results will be available
176	or C<-s _> etc...	571	using C<stat _> or C<-s _> and other tests (with the exception of C<-B>
		572	and C<-T>).
177		573
178	The pathname passed to C<aio_stat> must be absolute. See API NOTES, above,	574	The pathname passed to C<aio_stat> must be absolute. See API NOTES, above,
179	for an explanation.	575	for an explanation.
180		576
181	Currently, the stats are always 64-bit-stats, i.e. instead of returning an	577	Currently, the stats are always 64-bit-stats, i.e. instead of returning an
182	error when stat'ing a large file, the results will be silently truncated	578	error when stat'ing a large file, the results will be silently truncated
183	unless perl itself is compiled with large file support.	579	unless perl itself is compiled with large file support.
		580
		581	To help interpret the mode and dev/rdev stat values, IO::AIO offers the
		582	following constants and functions (if not implemented, the constants will
		583	be C<0> and the functions will either C<croak> or fall back on traditional
		584	behaviour).
		585
		586	C<S_IFMT>, C<S_IFIFO>, C<S_IFCHR>, C<S_IFBLK>, C<S_IFLNK>, C<S_IFREG>,
		587	C<S_IFDIR>, C<S_IFWHT>, C<S_IFSOCK>, C<IO::AIO::major $dev_t>,
		588	C<IO::AIO::minor $dev_t>, C<IO::AIO::makedev $major, $minor>.
		589
		590	To access higher resolution stat timestamps, see L<SUBSECOND STAT TIME
		591	ACCESS>.
184		592
185	Example: Print the length of F</etc/passwd>:	593	Example: Print the length of F</etc/passwd>:
186		594
187	aio_stat "/etc/passwd", sub {	595	aio_stat "/etc/passwd", sub {
188	$_[0] and die "stat failed: $!";	596	$_[0] and die "stat failed: $!";
189	print "size is ", -s _, "\n";	597	print "size is ", -s _, "\n";
190	};	598	};
191		599
		600
		601	=item aio_statvfs $fh_or_path, $callback->($statvfs)
		602
		603	Works like the POSIX C<statvfs> or C<fstatvfs> syscalls, depending on
		604	whether a file handle or path was passed.
		605
		606	On success, the callback is passed a hash reference with the following
		607	members: C<bsize>, C<frsize>, C<blocks>, C<bfree>, C<bavail>, C<files>,
		608	C<ffree>, C<favail>, C<fsid>, C<flag> and C<namemax>. On failure, C<undef>
		609	is passed.
		610
		611	The following POSIX IO::AIO::ST_* constants are defined: C<ST_RDONLY> and
		612	C<ST_NOSUID>.
		613
		614	The following non-POSIX IO::AIO::ST_* flag masks are defined to
		615	their correct value when available, or to C<0> on systems that do
		616	not support them: C<ST_NODEV>, C<ST_NOEXEC>, C<ST_SYNCHRONOUS>,
		617	C<ST_MANDLOCK>, C<ST_WRITE>, C<ST_APPEND>, C<ST_IMMUTABLE>, C<ST_NOATIME>,
		618	C<ST_NODIRATIME> and C<ST_RELATIME>.
		619
		620	Example: stat C</wd> and dump out the data if successful.
		621
		622	aio_statvfs "/wd", sub {
		623	my $f = $_[0]
		624	or die "statvfs: $!";
		625
		626	use Data::Dumper;
		627	say Dumper $f;
		628	};
		629
		630	# result:
		631	{
		632	bsize => 1024,
		633	bfree => 4333064312,
		634	blocks => 10253828096,
		635	files => 2050765568,
		636	flag => 4096,
		637	favail => 2042092649,
		638	bavail => 4333064312,
		639	ffree => 2042092649,
		640	namemax => 255,
		641	frsize => 1024,
		642	fsid => 1810
		643	}
		644
		645	=item aio_utime $fh_or_path, $atime, $mtime, $callback->($status)
		646
		647	Works like perl's C<utime> function (including the special case of $atime
		648	and $mtime being undef). Fractional times are supported if the underlying
		649	syscalls support them.
		650
		651	When called with a pathname, uses utimensat(2) or utimes(2) if available,
		652	otherwise utime(2). If called on a file descriptor, uses futimens(2)
		653	or futimes(2) if available, otherwise returns ENOSYS, so this is not
		654	portable.
		655
		656	Examples:
		657
		658	# set atime and mtime to current time (basically touch(1)):
		659	aio_utime "path", undef, undef;
		660	# set atime to current time and mtime to beginning of the epoch:
		661	aio_utime "path", time, undef; # undef==0
		662
		663
		664	=item aio_chown $fh_or_path, $uid, $gid, $callback->($status)
		665
		666	Works like perl's C<chown> function, except that C<undef> for either $uid
		667	or $gid is being interpreted as "do not change" (but -1 can also be used).
		668
		669	Examples:
		670
		671	# same as "chown root path" in the shell:
		672	aio_chown "path", 0, -1;
		673	# same as above:
		674	aio_chown "path", 0, undef;
		675
		676
		677	=item aio_truncate $fh_or_path, $offset, $callback->($status)
		678
		679	Works like truncate(2) or ftruncate(2).
		680
		681
		682	=item aio_allocate $fh, $mode, $offset, $len, $callback->($status)
		683
		684	Allocates or frees disk space according to the C<$mode> argument. See the
		685	linux C<fallocate> documentation for details.
		686
		687	C<$mode> is usually C<0> or C<IO::AIO::FALLOC_FL_KEEP_SIZE> to allocate
		688	space, or C<IO::AIO::FALLOC_FL_PUNCH_HOLE | IO::AIO::FALLOC_FL_KEEP_SIZE>,
		689	to deallocate a file range.
		690
		691	IO::AIO also supports C<FALLOC_FL_COLLAPSE_RANGE>, to remove a range
		692	(without leaving a hole), C<FALLOC_FL_ZERO_RANGE>, to zero a range,
		693	C<FALLOC_FL_INSERT_RANGE> to insert a range and C<FALLOC_FL_UNSHARE_RANGE>
		694	to unshare shared blocks (see your L<fallocate(2)> manpage).
		695
		696	The file system block size used by C<fallocate> is presumably the
		697	C<f_bsize> returned by C<statvfs>, but different filesystems and filetypes
		698	can dictate other limitations.
		699
		700	If C<fallocate> isn't available or cannot be emulated (currently no
		701	emulation will be attempted), passes C<-1> and sets C<$!> to C<ENOSYS>.
		702
		703
		704	=item aio_chmod $fh_or_path, $mode, $callback->($status)
		705
		706	Works like perl's C<chmod> function.
		707
		708
192	=item aio_unlink $pathname, $callback	709	=item aio_unlink $pathname, $callback->($status)
193		710
194	Asynchronously unlink (delete) a file and call the callback with the	711	Asynchronously unlink (delete) a file and call the callback with the
195	result code.	712	result code.
196		713
		714
		715	=item aio_mknod $pathname, $mode, $dev, $callback->($status)
		716
		717	[EXPERIMENTAL]
		718
		719	Asynchronously create a device node (or fifo). See mknod(2).
		720
		721	The only (POSIX-) portable way of calling this function is:
		722
		723	aio_mknod $pathname, IO::AIO::S_IFIFO | $mode, 0, sub { ...
		724
		725	See C<aio_stat> for info about some potentially helpful extra constants
		726	and functions.
		727
		728	=item aio_link $srcpath, $dstpath, $callback->($status)
		729
		730	Asynchronously create a new link to the existing object at C<$srcpath> at
		731	the path C<$dstpath> and call the callback with the result code.
		732
		733
		734	=item aio_symlink $srcpath, $dstpath, $callback->($status)
		735
		736	Asynchronously create a new symbolic link to the existing object at C<$srcpath> at
		737	the path C<$dstpath> and call the callback with the result code.
		738
		739
		740	=item aio_readlink $pathname, $callback->($link)
		741
		742	Asynchronously read the symlink specified by C<$path> and pass it to
		743	the callback. If an error occurs, nothing or undef gets passed to the
		744	callback.
		745
		746
		747	=item aio_realpath $pathname, $callback->($path)
		748
		749	Asynchronously make the path absolute and resolve any symlinks in
		750	C<$path>. The resulting path only consists of directories (same as
		751	L<Cwd::realpath>).
		752
		753	This request can be used to get the absolute path of the current working
		754	directory by passing it a path of F<.> (a single dot).
		755
		756
		757	=item aio_rename $srcpath, $dstpath, $callback->($status)
		758
		759	Asynchronously rename the object at C<$srcpath> to C<$dstpath>, just as
		760	rename(2) and call the callback with the result code.
		761
		762	On systems that support the AIO::WD working directory abstraction
		763	natively, the case C<[$wd, "."]> as C<$srcpath> is specialcased - instead
		764	of failing, C<rename> is called on the absolute path of C<$wd>.
		765
		766
		767	=item aio_rename2 $srcpath, $dstpath, $flags, $callback->($status)
		768
		769	Basically a version of C<aio_rename> with an additional C<$flags>
		770	argument. Calling this with C<$flags=0> is the same as calling
		771	C<aio_rename>.
		772
		773	Non-zero flags are currently only supported on GNU/Linux systems that
		774	support renameat2. Other systems fail with C<ENOSYS> in this case.
		775
		776	The following constants are available (missing ones are, as usual C<0>),
		777	see renameat2(2) for details:
		778
		779	C<IO::AIO::RENAME_NOREPLACE>, C<IO::AIO::RENAME_EXCHANGE>
		780	and C<IO::AIO::RENAME_WHITEOUT>.
		781
		782
		783	=item aio_mkdir $pathname, $mode, $callback->($status)
		784
		785	Asynchronously mkdir (create) a directory and call the callback with
		786	the result code. C<$mode> will be modified by the umask at the time the
		787	request is executed, so do not change your umask.
		788
		789
		790	=item aio_rmdir $pathname, $callback->($status)
		791
		792	Asynchronously rmdir (delete) a directory and call the callback with the
		793	result code.
		794
		795	On systems that support the AIO::WD working directory abstraction
		796	natively, the case C<[$wd, "."]> is specialcased - instead of failing,
		797	C<rmdir> is called on the absolute path of C<$wd>.
		798
		799
		800	=item aio_readdir $pathname, $callback->($entries)
		801
		802	Unlike the POSIX call of the same name, C<aio_readdir> reads an entire
		803	directory (i.e. opendir + readdir + closedir). The entries will not be
		804	sorted, and will B<NOT> include the C<.> and C<..> entries.
		805
		806	The callback is passed a single argument which is either C<undef> or an
		807	array-ref with the filenames.
		808
		809
		810	=item aio_readdirx $pathname, $flags, $callback->($entries, $flags)
		811
		812	Quite similar to C<aio_readdir>, but the C<$flags> argument allows one to
		813	tune behaviour and output format. In case of an error, C<$entries> will be
		814	C<undef>.
		815
		816	The flags are a combination of the following constants, ORed together (the
		817	flags will also be passed to the callback, possibly modified):
		818
		819	=over 4
		820
		821	=item IO::AIO::READDIR_DENTS
		822
		823	Normally the callback gets an arrayref consisting of names only (as
		824	with C<aio_readdir>). If this flag is set, then the callback gets an
		825	arrayref with C<[$name, $type, $inode]> arrayrefs, each describing a
		826	single directory entry in more detail:
		827
		828	C<$name> is the name of the entry.
		829
		830	C<$type> is one of the C<IO::AIO::DT_xxx> constants:
		831
		832	C<IO::AIO::DT_UNKNOWN>, C<IO::AIO::DT_FIFO>, C<IO::AIO::DT_CHR>, C<IO::AIO::DT_DIR>,
		833	C<IO::AIO::DT_BLK>, C<IO::AIO::DT_REG>, C<IO::AIO::DT_LNK>, C<IO::AIO::DT_SOCK>,
		834	C<IO::AIO::DT_WHT>.
		835
		836	C<IO::AIO::DT_UNKNOWN> means just that: readdir does not know. If you need
		837	to know, you have to run stat yourself. Also, for speed/memory reasons,
		838	the C<$type> scalars are read-only: you must not modify them.
		839
		840	C<$inode> is the inode number (which might not be exact on systems with 64
		841	bit inode numbers and 32 bit perls). This field has unspecified content on
		842	systems that do not deliver the inode information.
		843
		844	=item IO::AIO::READDIR_DIRS_FIRST
		845
		846	When this flag is set, then the names will be returned in an order where
		847	likely directories come first, in optimal stat order. This is useful when
		848	you need to quickly find directories, or you want to find all directories
		849	while avoiding to stat() each entry.
		850
		851	If the system returns type information in readdir, then this is used
		852	to find directories directly. Otherwise, likely directories are names
		853	beginning with ".", or otherwise names with no dots, of which names with
		854	short names are tried first.
		855
		856	=item IO::AIO::READDIR_STAT_ORDER
		857
		858	When this flag is set, then the names will be returned in an order
		859	suitable for stat()'ing each one. That is, when you plan to stat() most or
		860	all files in the given directory, then the returned order will likely be
		861	faster.
		862
		863	If both this flag and C<IO::AIO::READDIR_DIRS_FIRST> are specified,
		864	then the likely dirs come first, resulting in a less optimal stat order
		865	for stat'ing all entries, but likely a more optimal order for finding
		866	subdirectories.
		867
		868	=item IO::AIO::READDIR_FOUND_UNKNOWN
		869
		870	This flag should not be set when calling C<aio_readdirx>. Instead, it
		871	is being set by C<aio_readdirx>, when any of the C<$type>'s found were
		872	C<IO::AIO::DT_UNKNOWN>. The absence of this flag therefore indicates that all
		873	C<$type>'s are known, which can be used to speed up some algorithms.
		874
		875	=back
		876
		877
		878	=item aio_slurp $pathname, $offset, $length, $data, $callback->($status)
		879
		880	Opens, reads and closes the given file. The data is put into C<$data>,
		881	which is resized as required.
		882
		883	If C<$offset> is negative, then it is counted from the end of the file.
		884
		885	If C<$length> is zero, then the remaining length of the file is
		886	used. Also, in this case, the same limitations to modifying C<$data> apply
		887	as when IO::AIO::mmap is used, i.e. it must only be modified in-place
		888	with C<substr>. If the size of the file is known, specifying a non-zero
		889	C<$length> results in a performance advantage.
		890
		891	This request is similar to the older C<aio_load> request, but since it is
		892	a single request, it might be more efficient to use.
		893
		894	Example: load F</etc/passwd> into C<$passwd>.
		895
		896	my $passwd;
		897	aio_slurp "/etc/passwd", 0, 0, $passwd, sub {
		898	$_[0] >= 0
		899	or die "/etc/passwd: $!\n";
		900
		901	printf "/etc/passwd is %d bytes long, and contains:\n", length $passwd;
		902	print $passwd;
		903	};
		904	IO::AIO::flush;
		905
		906
		907	=item aio_load $pathname, $data, $callback->($status)
		908
		909	This is a composite request that tries to fully load the given file into
		910	memory. Status is the same as with aio_read.
		911
		912	Using C<aio_slurp> might be more efficient, as it is a single request.
		913
		914	=cut
		915
		916	sub aio_load($$;$) {
		917	my ($path, undef, $cb) = @_;
		918	my $data = \$_[1];
		919
		920	my $pri = aioreq_pri;
		921	my $grp = aio_group $cb;
		922
		923	aioreq_pri $pri;
		924	add $grp aio_open $path, O_RDONLY, 0, sub {
		925	my $fh = shift
		926	or return $grp->result (-1);
		927
		928	aioreq_pri $pri;
		929	add $grp aio_read $fh, 0, (-s $fh), $$data, 0, sub {
		930	$grp->result ($_[0]);
		931	};
		932	};
		933
		934	$grp
		935	}
		936
		937	=item aio_copy $srcpath, $dstpath, $callback->($status)
		938
		939	Try to copy the I<file> (directories not supported as either source or
		940	destination) from C<$srcpath> to C<$dstpath> and call the callback with
		941	a status of C<0> (ok) or C<-1> (error, see C<$!>).
		942
		943	Existing destination files will be truncated.
		944
		945	This is a composite request that creates the destination file with
		946	mode 0200 and copies the contents of the source file into it using
		947	C<aio_sendfile>, followed by restoring atime, mtime, access mode and
		948	uid/gid, in that order.
		949
		950	If an error occurs, the partial destination file will be unlinked, if
		951	possible, except when setting atime, mtime, access mode and uid/gid, where
		952	errors are being ignored.
		953
		954	=cut
		955
		956	sub aio_copy($$;$) {
		957	my ($src, $dst, $cb) = @_;
		958
		959	my $pri = aioreq_pri;
		960	my $grp = aio_group $cb;
		961
		962	aioreq_pri $pri;
		963	add $grp aio_open $src, O_RDONLY, 0, sub {
		964	if (my $src_fh = $_[0]) {
		965	my @stat = stat $src_fh; # hmm, might block over nfs?
		966
		967	aioreq_pri $pri;
		968	add $grp aio_open $dst, O_CREAT | O_WRONLY | O_TRUNC, 0200, sub {
		969	if (my $dst_fh = $_[0]) {
		970	aioreq_pri $pri;
		971	add $grp aio_sendfile $dst_fh, $src_fh, 0, $stat[7], sub {
		972	if ($_[0] == $stat[7]) {
		973	$grp->result (0);
		974	close $src_fh;
		975
		976	my $ch = sub {
		977	aioreq_pri $pri;
		978	add $grp aio_chmod $dst_fh, $stat[2] & 07777, sub {
		979	aioreq_pri $pri;
		980	add $grp aio_chown $dst_fh, $stat[4], $stat[5], sub {
		981	aioreq_pri $pri;
		982	add $grp aio_close $dst_fh;
		983	}
		984	};
		985	};
		986
		987	aioreq_pri $pri;
		988	add $grp aio_utime $dst_fh, $stat[8], $stat[9], sub {
		989	if ($_[0] < 0 && $! == ENOSYS) {
		990	aioreq_pri $pri;
		991	add $grp aio_utime $dst, $stat[8], $stat[9], $ch;
		992	} else {
		993	$ch->();
		994	}
		995	};
		996	} else {
		997	$grp->result (-1);
		998	close $src_fh;
		999	close $dst_fh;
		1000
		1001	aioreq $pri;
		1002	add $grp aio_unlink $dst;
		1003	}
		1004	};
		1005	} else {
		1006	$grp->result (-1);
		1007	}
		1008	},
		1009
		1010	} else {
		1011	$grp->result (-1);
		1012	}
		1013	};
		1014
		1015	$grp
		1016	}
		1017
		1018	=item aio_move $srcpath, $dstpath, $callback->($status)
		1019
		1020	Try to move the I<file> (directories not supported as either source or
		1021	destination) from C<$srcpath> to C<$dstpath> and call the callback with
		1022	a status of C<0> (ok) or C<-1> (error, see C<$!>).
		1023
		1024	This is a composite request that tries to rename(2) the file first; if
		1025	rename fails with C<EXDEV>, it copies the file with C<aio_copy> and, if
		1026	that is successful, unlinks the C<$srcpath>.
		1027
		1028	=cut
		1029
		1030	sub aio_move($$;$) {
		1031	my ($src, $dst, $cb) = @_;
		1032
		1033	my $pri = aioreq_pri;
		1034	my $grp = aio_group $cb;
		1035
		1036	aioreq_pri $pri;
		1037	add $grp aio_rename $src, $dst, sub {
		1038	if ($_[0] && $! == EXDEV) {
		1039	aioreq_pri $pri;
		1040	add $grp aio_copy $src, $dst, sub {
		1041	$grp->result ($_[0]);
		1042
		1043	unless ($_[0]) {
		1044	aioreq_pri $pri;
		1045	add $grp aio_unlink $src;
		1046	}
		1047	};
		1048	} else {
		1049	$grp->result ($_[0]);
		1050	}
		1051	};
		1052
		1053	$grp
		1054	}
		1055
		1056	=item aio_scandir $pathname, $maxreq, $callback->($dirs, $nondirs)
		1057
		1058	Scans a directory (similar to C<aio_readdir>) but additionally tries to
		1059	efficiently separate the entries of directory C<$path> into two sets of
		1060	names, directories you can recurse into (directories), and ones you cannot
		1061	recurse into (everything else, including symlinks to directories).
		1062
		1063	C<aio_scandir> is a composite request that generates many sub requests.
		1064	C<$maxreq> specifies the maximum number of outstanding aio requests that
		1065	this function generates. If it is C<< <= 0 >>, then a suitable default
		1066	will be chosen (currently 4).
		1067
		1068	On error, the callback is called without arguments, otherwise it receives
		1069	two array-refs with path-relative entry names.
		1070
		1071	Example:
		1072
		1073	aio_scandir $dir, 0, sub {
		1074	my ($dirs, $nondirs) = @_;
		1075	print "real directories: @$dirs\n";
		1076	print "everything else: @$nondirs\n";
		1077	};
		1078
		1079	Implementation notes.
		1080
		1081	The C<aio_readdir> cannot be avoided, but C<stat()>'ing every entry can.
		1082
		1083	If readdir returns file type information, then this is used directly to
		1084	find directories.
		1085
		1086	Otherwise, after reading the directory, the modification time, size etc.
		1087	of the directory before and after the readdir is checked, and if they
		1088	match (and isn't the current time), the link count will be used to decide
		1089	how many entries are directories (if >= 2). Otherwise, no knowledge of the
		1090	number of subdirectories will be assumed.
		1091
		1092	Then entries will be sorted into likely directories a non-initial dot
		1093	currently) and likely non-directories (see C<aio_readdirx>). Then every
		1094	entry plus an appended C</.> will be C<stat>'ed, likely directories first,
		1095	in order of their inode numbers. If that succeeds, it assumes that the
		1096	entry is a directory or a symlink to directory (which will be checked
		1097	separately). This is often faster than stat'ing the entry itself because
		1098	filesystems might detect the type of the entry without reading the inode
		1099	data (e.g. ext2fs filetype feature), even on systems that cannot return
		1100	the filetype information on readdir.
		1101
		1102	If the known number of directories (link count - 2) has been reached, the
		1103	rest of the entries is assumed to be non-directories.
		1104
		1105	This only works with certainty on POSIX (= UNIX) filesystems, which
		1106	fortunately are the vast majority of filesystems around.
		1107
		1108	It will also likely work on non-POSIX filesystems with reduced efficiency
		1109	as those tend to return 0 or 1 as link counts, which disables the
		1110	directory counting heuristic.
		1111
		1112	=cut
		1113
		1114	sub aio_scandir($$;$) {
		1115	my ($path, $maxreq, $cb) = @_;
		1116
		1117	my $pri = aioreq_pri;
		1118
		1119	my $grp = aio_group $cb;
		1120
		1121	$maxreq = 4 if $maxreq <= 0;
		1122
		1123	# get a wd object
		1124	aioreq_pri $pri;
		1125	add $grp aio_wd $path, sub {
		1126	$_[0]
		1127	or return $grp->result ();
		1128
		1129	my $wd = [shift, "."];
		1130
		1131	# stat once
		1132	aioreq_pri $pri;
		1133	add $grp aio_stat $wd, sub {
		1134	return $grp->result () if $_[0];
		1135	my $now = time;
		1136	my $hash1 = join ":", (stat _)[0,1,3,7,9];
		1137	my $rdxflags = READDIR_DIRS_FIRST;
		1138
		1139	if ((stat _)[3] < 2) {
		1140	# at least one non-POSIX filesystem exists
		1141	# that returns useful DT_type values: btrfs,
		1142	# so optimise for this here by requesting dents
		1143	$rdxflags |= READDIR_DENTS;
		1144	}
		1145
		1146	# read the directory entries
		1147	aioreq_pri $pri;
		1148	add $grp aio_readdirx $wd, $rdxflags, sub {
		1149	my ($entries, $flags) = @_
		1150	or return $grp->result ();
		1151
		1152	if ($rdxflags & READDIR_DENTS) {
		1153	# if we requested type values, see if we can use them directly.
		1154
		1155	# if there were any DT_UNKNOWN entries then we assume we
		1156	# don't know. alternatively, we could assume that if we get
		1157	# one DT_DIR, then all directories are indeed marked with
		1158	# DT_DIR, but this seems not required for btrfs, and this
		1159	# is basically the "btrfs can't get it's act together" code
		1160	# branch.
		1161	unless ($flags & READDIR_FOUND_UNKNOWN) {
		1162	# now we have valid DT_ information for all entries,
		1163	# so use it as an optimisation without further stat's.
		1164	# they must also all be at the beginning of @$entries
		1165	# by now.
		1166
		1167	my $dirs;
		1168
		1169	if (@$entries) {
		1170	for (0 .. $#$entries) {
		1171	if ($entries->[$_][1] != DT_DIR) {
		1172	# splice out directories
		1173	$dirs = [splice @$entries, 0, $_];
		1174	last;
		1175	}
		1176	}
		1177
		1178	# if we didn't find any non-dir, then all entries are dirs
		1179	unless ($dirs) {
		1180	($dirs, $entries) = ($entries, []);
		1181	}
		1182	} else {
		1183	# directory is empty, so there are no sbdirs
		1184	$dirs = [];
		1185	}
		1186
		1187	# either splice'd the directories out or the dir was empty.
		1188	# convert dents to filenames
		1189	$_ = $_->[0] for @$dirs;
		1190	$_ = $_->[0] for @$entries;
		1191
		1192	return $grp->result ($dirs, $entries);
		1193	}
		1194
		1195	# cannot use, so return to our old ways
		1196	# by pretending we only scanned for names.
		1197	$_ = $_->[0] for @$entries;
		1198	}
		1199
		1200	# stat the dir another time
		1201	aioreq_pri $pri;
		1202	add $grp aio_stat $wd, sub {
		1203	my $hash2 = join ":", (stat _)[0,1,3,7,9];
		1204
		1205	my $ndirs;
		1206
		1207	# take the slow route if anything looks fishy
		1208	if ($hash1 ne $hash2 or (stat _)[9] == $now) {
		1209	$ndirs = -1;
		1210	} else {
		1211	# if nlink == 2, we are finished
		1212	# for non-posix-fs's, we rely on nlink < 2
		1213	$ndirs = (stat _)[3] - 2
		1214	or return $grp->result ([], $entries);
		1215	}
		1216
		1217	my (@dirs, @nondirs);
		1218
		1219	my $statgrp = add $grp aio_group sub {
		1220	$grp->result (\@dirs, \@nondirs);
		1221	};
		1222
		1223	limit $statgrp $maxreq;
		1224	feed $statgrp sub {
		1225	return unless @$entries;
		1226	my $entry = shift @$entries;
		1227
		1228	aioreq_pri $pri;
		1229	$wd->[1] = "$entry/.";
		1230	add $statgrp aio_stat $wd, sub {
		1231	if ($_[0] < 0) {
		1232	push @nondirs, $entry;
		1233	} else {
		1234	# need to check for real directory
		1235	aioreq_pri $pri;
		1236	$wd->[1] = $entry;
		1237	add $statgrp aio_lstat $wd, sub {
		1238	if (-d _) {
		1239	push @dirs, $entry;
		1240
		1241	unless (--$ndirs) {
		1242	push @nondirs, @$entries;
		1243	feed $statgrp;
		1244	}
		1245	} else {
		1246	push @nondirs, $entry;
		1247	}
		1248	}
		1249	}
		1250	};
		1251	};
		1252	};
		1253	};
		1254	};
		1255	};
		1256
		1257	$grp
		1258	}
		1259
		1260	=item aio_rmtree $pathname, $callback->($status)
		1261
		1262	Delete a directory tree starting (and including) C<$path>, return the
		1263	status of the final C<rmdir> only. This is a composite request that
		1264	uses C<aio_scandir> to recurse into and rmdir directories, and unlink
		1265	everything else.
		1266
		1267	=cut
		1268
		1269	sub aio_rmtree;
		1270	sub aio_rmtree($;$) {
		1271	my ($path, $cb) = @_;
		1272
		1273	my $pri = aioreq_pri;
		1274	my $grp = aio_group $cb;
		1275
		1276	aioreq_pri $pri;
		1277	add $grp aio_scandir $path, 0, sub {
		1278	my ($dirs, $nondirs) = @_;
		1279
		1280	my $dirgrp = aio_group sub {
		1281	add $grp aio_rmdir $path, sub {
		1282	$grp->result ($_[0]);
		1283	};
		1284	};
		1285
		1286	(aioreq_pri $pri), add $dirgrp aio_rmtree "$path/$_" for @$dirs;
		1287	(aioreq_pri $pri), add $dirgrp aio_unlink "$path/$_" for @$nondirs;
		1288
		1289	add $grp $dirgrp;
		1290	};
		1291
		1292	$grp
		1293	}
		1294
		1295	=item aio_fcntl $fh, $cmd, $arg, $callback->($status)
		1296
		1297	=item aio_ioctl $fh, $request, $buf, $callback->($status)
		1298
		1299	These work just like the C<fcntl> and C<ioctl> built-in functions, except
		1300	they execute asynchronously and pass the return value to the callback.
		1301
		1302	Both calls can be used for a lot of things, some of which make more sense
		1303	to run asynchronously in their own thread, while some others make less
		1304	sense. For example, calls that block waiting for external events, such
		1305	as locking, will also lock down an I/O thread while it is waiting, which
		1306	can deadlock the whole I/O system. At the same time, there might be no
		1307	alternative to using a thread to wait.
		1308
		1309	So in general, you should only use these calls for things that do
		1310	(filesystem) I/O, not for things that wait for other events (network,
		1311	other processes), although if you are careful and know what you are doing,
		1312	you still can.
		1313
		1314	The following constants are available and can be used for normal C<ioctl>
		1315	and C<fcntl> as well (missing ones are, as usual C<0>):
		1316
		1317	C<F_DUPFD_CLOEXEC>,
		1318
		1319	C<F_OFD_GETLK>, C<F_OFD_SETLK>, C<F_OFD_GETLKW>,
		1320
		1321	C<FIFREEZE>, C<FITHAW>, C<FITRIM>, C<FICLONE>, C<FICLONERANGE>, C<FIDEDUPERANGE>.
		1322
		1323	C<F_ADD_SEALS>, C<F_GET_SEALS>, C<F_SEAL_SEAL>, C<F_SEAL_SHRINK>, C<F_SEAL_GROW> and
		1324	C<F_SEAL_WRITE>.
		1325
		1326	C<FS_IOC_GETFLAGS>, C<FS_IOC_SETFLAGS>, C<FS_IOC_GETVERSION>, C<FS_IOC_SETVERSION>,
		1327	C<FS_IOC_FIEMAP>.
		1328
		1329	C<FS_IOC_FSGETXATTR>, C<FS_IOC_FSSETXATTR>, C<FS_IOC_SET_ENCRYPTION_POLICY>,
		1330	C<FS_IOC_GET_ENCRYPTION_PWSALT>, C<FS_IOC_GET_ENCRYPTION_POLICY>, C<FS_KEY_DESCRIPTOR_SIZE>.
		1331
		1332	C<FS_SECRM_FL>, C<FS_UNRM_FL>, C<FS_COMPR_FL>, C<FS_SYNC_FL>, C<FS_IMMUTABLE_FL>,
		1333	C<FS_APPEND_FL>, C<FS_NODUMP_FL>, C<FS_NOATIME_FL>, C<FS_DIRTY_FL>,
		1334	C<FS_COMPRBLK_FL>, C<FS_NOCOMP_FL>, C<FS_ENCRYPT_FL>, C<FS_BTREE_FL>,
		1335	C<FS_INDEX_FL>, C<FS_JOURNAL_DATA_FL>, C<FS_NOTAIL_FL>, C<FS_DIRSYNC_FL>, C<FS_TOPDIR_FL>,
		1336	C<FS_FL_USER_MODIFIABLE>.
		1337
		1338	C<FS_XFLAG_REALTIME>, C<FS_XFLAG_PREALLOC>, C<FS_XFLAG_IMMUTABLE>, C<FS_XFLAG_APPEND>,
		1339	C<FS_XFLAG_SYNC>, C<FS_XFLAG_NOATIME>, C<FS_XFLAG_NODUMP>, C<FS_XFLAG_RTINHERIT>,
		1340	C<FS_XFLAG_PROJINHERIT>, C<FS_XFLAG_NOSYMLINKS>, C<FS_XFLAG_EXTSIZE>, C<FS_XFLAG_EXTSZINHERIT>,
		1341	C<FS_XFLAG_NODEFRAG>, C<FS_XFLAG_FILESTREAM>, C<FS_XFLAG_DAX>, C<FS_XFLAG_HASATTR>,
		1342
		1343	=item aio_sync $callback->($status)
		1344
		1345	Asynchronously call sync and call the callback when finished.
		1346
197	=item aio_fsync $fh, $callback	1347	=item aio_fsync $fh, $callback->($status)
198		1348
199	Asynchronously call fsync on the given filehandle and call the callback	1349	Asynchronously call fsync on the given filehandle and call the callback
200	with the fsync result code.	1350	with the fsync result code.
201		1351
202	=item aio_fdatasync $fh, $callback	1352	=item aio_fdatasync $fh, $callback->($status)
203		1353
204	Asynchronously call fdatasync on the given filehandle and call the	1354	Asynchronously call fdatasync on the given filehandle and call the
205	callback with the fdatasync result code. Might set C<$!> to C<ENOSYS> if	1355	callback with the fdatasync result code.
206	C<fdatasync> is not available.	1356
		1357	If this call isn't available because your OS lacks it or it couldn't be
		1358	detected, it will be emulated by calling C<fsync> instead.
		1359
		1360	=item aio_syncfs $fh, $callback->($status)
		1361
		1362	Asynchronously call the syncfs syscall to sync the filesystem associated
		1363	to the given filehandle and call the callback with the syncfs result
		1364	code. If syncfs is not available, calls sync(), but returns C<-1> and sets
		1365	errno to C<ENOSYS> nevertheless.
		1366
		1367	=item aio_sync_file_range $fh, $offset, $nbytes, $flags, $callback->($status)
		1368
		1369	Sync the data portion of the file specified by C<$offset> and C<$length>
		1370	to disk (but NOT the metadata), by calling the Linux-specific
		1371	sync_file_range call. If sync_file_range is not available or it returns
		1372	ENOSYS, then fdatasync or fsync is being substituted.
		1373
		1374	C<$flags> can be a combination of C<IO::AIO::SYNC_FILE_RANGE_WAIT_BEFORE>,
		1375	C<IO::AIO::SYNC_FILE_RANGE_WRITE> and
		1376	C<IO::AIO::SYNC_FILE_RANGE_WAIT_AFTER>: refer to the sync_file_range
		1377	manpage for details.
		1378
		1379	=item aio_pathsync $pathname, $callback->($status)
		1380
		1381	This request tries to open, fsync and close the given path. This is a
		1382	composite request intended to sync directories after directory operations
		1383	(E.g. rename). This might not work on all operating systems or have any
		1384	specific effect, but usually it makes sure that directory changes get
		1385	written to disc. It works for anything that can be opened for read-only,
		1386	not just directories.
		1387
		1388	Future versions of this function might fall back to other methods when
		1389	C<fsync> on the directory fails (such as calling C<sync>).
		1390
		1391	Passes C<0> when everything went ok, and C<-1> on error.
		1392
		1393	=cut
		1394
		1395	sub aio_pathsync($;$) {
		1396	my ($path, $cb) = @_;
		1397
		1398	my $pri = aioreq_pri;
		1399	my $grp = aio_group $cb;
		1400
		1401	aioreq_pri $pri;
		1402	add $grp aio_open $path, O_RDONLY, 0, sub {
		1403	my ($fh) = @_;
		1404	if ($fh) {
		1405	aioreq_pri $pri;
		1406	add $grp aio_fsync $fh, sub {
		1407	$grp->result ($_[0]);
		1408
		1409	aioreq_pri $pri;
		1410	add $grp aio_close $fh;
		1411	};
		1412	} else {
		1413	$grp->result (-1);
		1414	}
		1415	};
		1416
		1417	$grp
		1418	}
		1419
		1420	=item aio_msync $scalar, $offset = 0, $length = undef, flags = MS_SYNC, $callback->($status)
		1421
		1422	This is a rather advanced IO::AIO call, which only works on mmap(2)ed
		1423	scalars (see the C<IO::AIO::mmap> function, although it also works on data
		1424	scalars managed by the L<Sys::Mmap> or L<Mmap> modules, note that the
		1425	scalar must only be modified in-place while an aio operation is pending on
		1426	it).
		1427
		1428	It calls the C<msync> function of your OS, if available, with the memory
		1429	area starting at C<$offset> in the string and ending C<$length> bytes
		1430	later. If C<$length> is negative, counts from the end, and if C<$length>
		1431	is C<undef>, then it goes till the end of the string. The flags can be
		1432	either C<IO::AIO::MS_ASYNC> or C<IO::AIO::MS_SYNC>, plus an optional
		1433	C<IO::AIO::MS_INVALIDATE>.
		1434
		1435	=item aio_mtouch $scalar, $offset = 0, $length = undef, flags = 0, $callback->($status)
		1436
		1437	This is a rather advanced IO::AIO call, which works best on mmap(2)ed
		1438	scalars.
		1439
		1440	It touches (reads or writes) all memory pages in the specified
		1441	range inside the scalar. All caveats and parameters are the same
		1442	as for C<aio_msync>, above, except for flags, which must be either
		1443	C<0> (which reads all pages and ensures they are instantiated) or
		1444	C<IO::AIO::MT_MODIFY>, which modifies the memory pages (by reading and
		1445	writing an octet from it, which dirties the page).
		1446
		1447	=item aio_mlock $scalar, $offset = 0, $length = undef, $callback->($status)
		1448
		1449	This is a rather advanced IO::AIO call, which works best on mmap(2)ed
		1450	scalars.
		1451
		1452	It reads in all the pages of the underlying storage into memory (if any)
		1453	and locks them, so they are not getting swapped/paged out or removed.
		1454
		1455	If C<$length> is undefined, then the scalar will be locked till the end.
		1456
		1457	On systems that do not implement C<mlock>, this function returns C<-1>
		1458	and sets errno to C<ENOSYS>.
		1459
		1460	Note that the corresponding C<munlock> is synchronous and is
		1461	documented under L<MISCELLANEOUS FUNCTIONS>.
		1462
		1463	Example: open a file, mmap and mlock it - both will be undone when
		1464	C<$data> gets destroyed.
		1465
		1466	open my $fh, "<", $path or die "$path: $!";
		1467	my $data;
		1468	IO::AIO::mmap $data, -s $fh, IO::AIO::PROT_READ, IO::AIO::MAP_SHARED, $fh;
		1469	aio_mlock $data; # mlock in background
		1470
		1471	=item aio_mlockall $flags, $callback->($status)
		1472
		1473	Calls the C<mlockall> function with the given C<$flags> (a
		1474	combination of C<IO::AIO::MCL_CURRENT>, C<IO::AIO::MCL_FUTURE> and
		1475	C<IO::AIO::MCL_ONFAULT>).
		1476
		1477	On systems that do not implement C<mlockall>, this function returns C<-1>
		1478	and sets errno to C<ENOSYS>. Similarly, flag combinations not supported
		1479	by the system result in a return value of C<-1> with errno being set to
		1480	C<EINVAL>.
		1481
		1482	Note that the corresponding C<munlockall> is synchronous and is
		1483	documented under L<MISCELLANEOUS FUNCTIONS>.
		1484
		1485	Example: asynchronously lock all current and future pages into memory.
		1486
		1487	aio_mlockall IO::AIO::MCL_FUTURE;
		1488
		1489	=item aio_fiemap $fh, $start, $length, $flags, $count, $cb->(\@extents)
		1490
		1491	Queries the extents of the given file (by calling the Linux C<FIEMAP>
		1492	ioctl, see L<http://cvs.schmorp.de/IO-AIO/doc/fiemap.txt> for details). If
		1493	the ioctl is not available on your OS, then this request will fail with
		1494	C<ENOSYS>.
		1495
		1496	C<$start> is the starting offset to query extents for, C<$length> is the
		1497	size of the range to query - if it is C<undef>, then the whole file will
		1498	be queried.
		1499
		1500	C<$flags> is a combination of flags (C<IO::AIO::FIEMAP_FLAG_SYNC> or
		1501	C<IO::AIO::FIEMAP_FLAG_XATTR> - C<IO::AIO::FIEMAP_FLAGS_COMPAT> is also
		1502	exported), and is normally C<0> or C<IO::AIO::FIEMAP_FLAG_SYNC> to query
		1503	the data portion.
		1504
		1505	C<$count> is the maximum number of extent records to return. If it is
		1506	C<undef>, then IO::AIO queries all extents of the range. As a very special
		1507	case, if it is C<0>, then the callback receives the number of extents
		1508	instead of the extents themselves (which is unreliable, see below).
		1509
		1510	If an error occurs, the callback receives no arguments. The special
		1511	C<errno> value C<IO::AIO::EBADR> is available to test for flag errors.
		1512
		1513	Otherwise, the callback receives an array reference with extent
		1514	structures. Each extent structure is an array reference itself, with the
		1515	following members:
		1516
		1517	[$logical, $physical, $length, $flags]
		1518
		1519	Flags is any combination of the following flag values (typically either C<0>
		1520	or C<IO::AIO::FIEMAP_EXTENT_LAST> (1)):
		1521
		1522	C<IO::AIO::FIEMAP_EXTENT_LAST>, C<IO::AIO::FIEMAP_EXTENT_UNKNOWN>,
		1523	C<IO::AIO::FIEMAP_EXTENT_DELALLOC>, C<IO::AIO::FIEMAP_EXTENT_ENCODED>,
		1524	C<IO::AIO::FIEMAP_EXTENT_DATA_ENCRYPTED>, C<IO::AIO::FIEMAP_EXTENT_NOT_ALIGNED>,
		1525	C<IO::AIO::FIEMAP_EXTENT_DATA_INLINE>, C<IO::AIO::FIEMAP_EXTENT_DATA_TAIL>,
		1526	C<IO::AIO::FIEMAP_EXTENT_UNWRITTEN>, C<IO::AIO::FIEMAP_EXTENT_MERGED> or
		1527	C<IO::AIO::FIEMAP_EXTENT_SHARED>.
		1528
		1529	At the time of this writing (Linux 3.2), this request is unreliable unless
		1530	C<$count> is C<undef>, as the kernel has all sorts of bugs preventing
		1531	it to return all extents of a range for files with a large number of
		1532	extents. The code (only) works around all these issues if C<$count> is
		1533	C<undef>.
		1534
		1535	=item aio_group $callback->(...)
		1536
		1537	This is a very special aio request: Instead of doing something, it is a
		1538	container for other aio requests, which is useful if you want to bundle
		1539	many requests into a single, composite, request with a definite callback
		1540	and the ability to cancel the whole request with its subrequests.
		1541
		1542	Returns an object of class L<IO::AIO::GRP>. See its documentation below
		1543	for more info.
		1544
		1545	Example:
		1546
		1547	my $grp = aio_group sub {
		1548	print "all stats done\n";
		1549	};
		1550
		1551	add $grp
		1552	(aio_stat ...),
		1553	(aio_stat ...),
		1554	...;
		1555
		1556	=item aio_nop $callback->()
		1557
		1558	This is a special request - it does nothing in itself and is only used for
		1559	side effects, such as when you want to add a dummy request to a group so
		1560	that finishing the requests in the group depends on executing the given
		1561	code.
		1562
		1563	While this request does nothing, it still goes through the execution
		1564	phase and still requires a worker thread. Thus, the callback will not
		1565	be executed immediately but only after other requests in the queue have
		1566	entered their execution phase. This can be used to measure request
		1567	latency.
		1568
		1569	=item IO::AIO::aio_busy $fractional_seconds, $callback->() *NOT EXPORTED*
		1570
		1571	Mainly used for debugging and benchmarking, this aio request puts one of
		1572	the request workers to sleep for the given time.
		1573
		1574	While it is theoretically handy to have simple I/O scheduling requests
		1575	like sleep and file handle readable/writable, the overhead this creates is
		1576	immense (it blocks a thread for a long time) so do not use this function
		1577	except to put your application under artificial I/O pressure.
207		1578
208	=back	1579	=back
209		1580
		1581
		1582	=head2 IO::AIO::WD - multiple working directories
		1583
		1584	Your process only has one current working directory, which is used by all
		1585	threads. This makes it hard to use relative paths (some other component
		1586	could call C<chdir> at any time, and it is hard to control when the path
		1587	will be used by IO::AIO).
		1588
		1589	One solution for this is to always use absolute paths. This usually works,
		1590	but can be quite slow (the kernel has to walk the whole path on every
		1591	access), and can also be a hassle to implement.
		1592
		1593	Newer POSIX systems have a number of functions (openat, fdopendir,
		1594	futimensat and so on) that make it possible to specify working directories
		1595	per operation.
		1596
		1597	For portability, and because the clowns who "designed", or shall I write,
		1598	perpetrated this new interface were obviously half-drunk, this abstraction
		1599	cannot be perfect, though.
		1600
		1601	IO::AIO allows you to convert directory paths into a so-called IO::AIO::WD
		1602	object. This object stores the canonicalised, absolute version of the
		1603	path, and on systems that allow it, also a directory file descriptor.
		1604
		1605	Everywhere where a pathname is accepted by IO::AIO (e.g. in C<aio_stat>
		1606	or C<aio_unlink>), one can specify an array reference with an IO::AIO::WD
		1607	object and a pathname instead (or the IO::AIO::WD object alone, which
		1608	gets interpreted as C<[$wd, "."]>). If the pathname is absolute, the
		1609	IO::AIO::WD object is ignored, otherwise the pathname is resolved relative
		1610	to that IO::AIO::WD object.
		1611
		1612	For example, to get a wd object for F</etc> and then stat F<passwd>
		1613	inside, you would write:
		1614
		1615	aio_wd "/etc", sub {
		1616	my $etcdir = shift;
		1617
		1618	# although $etcdir can be undef on error, there is generally no reason
		1619	# to check for errors here, as aio_stat will fail with ENOENT
		1620	# when $etcdir is undef.
		1621
		1622	aio_stat [$etcdir, "passwd"], sub {
		1623	# yay
		1624	};
		1625	};
		1626
		1627	The fact that C<aio_wd> is a request and not a normal function shows that
		1628	creating an IO::AIO::WD object is itself a potentially blocking operation,
		1629	which is why it is done asynchronously.
		1630
		1631	To stat the directory obtained with C<aio_wd> above, one could write
		1632	either of the following three request calls:
		1633
		1634	aio_lstat "/etc" , sub { ... # pathname as normal string
		1635	aio_lstat [$wd, "."], sub { ... # "." relative to $wd (i.e. $wd itself)
		1636	aio_lstat $wd , sub { ... # shorthand for the previous
		1637
		1638	As with normal pathnames, IO::AIO keeps a copy of the working directory
		1639	object and the pathname string, so you could write the following without
		1640	causing any issues due to C<$path> getting reused:
		1641
		1642	my $path = [$wd, undef];
		1643
		1644	for my $name (qw(abc def ghi)) {
		1645	$path->[1] = $name;
		1646	aio_stat $path, sub {
		1647	# ...
		1648	};
		1649	}
		1650
		1651	There are some caveats: when directories get renamed (or deleted), the
		1652	pathname string doesn't change, so will point to the new directory (or
		1653	nowhere at all), while the directory fd, if available on the system,
		1654	will still point to the original directory. Most functions accepting a
		1655	pathname will use the directory fd on newer systems, and the string on
		1656	older systems. Some functions (such as C<aio_realpath>) will always rely on
		1657	the string form of the pathname.
		1658
		1659	So this functionality is mainly useful to get some protection against
		1660	C<chdir>, to easily get an absolute path out of a relative path for future
		1661	reference, and to speed up doing many operations in the same directory
		1662	(e.g. when stat'ing all files in a directory).
		1663
		1664	The following functions implement this working directory abstraction:
		1665
		1666	=over 4
		1667
		1668	=item aio_wd $pathname, $callback->($wd)
		1669
		1670	Asynchonously canonicalise the given pathname and convert it to an
		1671	IO::AIO::WD object representing it. If possible and supported on the
		1672	system, also open a directory fd to speed up pathname resolution relative
		1673	to this working directory.
		1674
		1675	If something goes wrong, then C<undef> is passwd to the callback instead
		1676	of a working directory object and C<$!> is set appropriately. Since
		1677	passing C<undef> as working directory component of a pathname fails the
		1678	request with C<ENOENT>, there is often no need for error checking in the
		1679	C<aio_wd> callback, as future requests using the value will fail in the
		1680	expected way.
		1681
		1682	=item IO::AIO::CWD
		1683
		1684	This is a compile time constant (object) that represents the process
		1685	current working directory.
		1686
		1687	Specifying this object as working directory object for a pathname is as if
		1688	the pathname would be specified directly, without a directory object. For
		1689	example, these calls are functionally identical:
		1690
		1691	aio_stat "somefile", sub { ... };
		1692	aio_stat [IO::AIO::CWD, "somefile"], sub { ... };
		1693
		1694	=back
		1695
		1696	To recover the path associated with an IO::AIO::WD object, you can use
		1697	C<aio_realpath>:
		1698
		1699	aio_realpath $wd, sub {
		1700	warn "path is $_[0]\n";
		1701	};
		1702
		1703	Currently, C<aio_statvfs> always, and C<aio_rename> and C<aio_rmdir>
		1704	sometimes, fall back to using an absolue path.
		1705
		1706	=head2 IO::AIO::REQ CLASS
		1707
		1708	All non-aggregate C<aio_*> functions return an object of this class when
		1709	called in non-void context.
		1710
		1711	=over 4
		1712
		1713	=item cancel $req
		1714
		1715	Cancels the request, if possible. Has the effect of skipping execution
		1716	when entering the B<execute> state and skipping calling the callback when
		1717	entering the the B<result> state, but will leave the request otherwise
		1718	untouched (with the exception of readdir). That means that requests that
		1719	currently execute will not be stopped and resources held by the request
		1720	will not be freed prematurely.
		1721
		1722	=item cb $req $callback->(...)
		1723
		1724	Replace (or simply set) the callback registered to the request.
		1725
		1726	=back
		1727
		1728	=head2 IO::AIO::GRP CLASS
		1729
		1730	This class is a subclass of L<IO::AIO::REQ>, so all its methods apply to
		1731	objects of this class, too.
		1732
		1733	A IO::AIO::GRP object is a special request that can contain multiple other
		1734	aio requests.
		1735
		1736	You create one by calling the C<aio_group> constructing function with a
		1737	callback that will be called when all contained requests have entered the
		1738	C<done> state:
		1739
		1740	my $grp = aio_group sub {
		1741	print "all requests are done\n";
		1742	};
		1743
		1744	You add requests by calling the C<add> method with one or more
		1745	C<IO::AIO::REQ> objects:
		1746
		1747	$grp->add (aio_unlink "...");
		1748
		1749	add $grp aio_stat "...", sub {
		1750	$_[0] or return $grp->result ("error");
		1751
		1752	# add another request dynamically, if first succeeded
		1753	add $grp aio_open "...", sub {
		1754	$grp->result ("ok");
		1755	};
		1756	};
		1757
		1758	This makes it very easy to create composite requests (see the source of
		1759	C<aio_move> for an application) that work and feel like simple requests.
		1760
		1761	=over 4
		1762
		1763	=item * The IO::AIO::GRP objects will be cleaned up during calls to
		1764	C<IO::AIO::poll_cb>, just like any other request.
		1765
		1766	=item * They can be canceled like any other request. Canceling will cancel not
		1767	only the request itself, but also all requests it contains.
		1768
		1769	=item * They can also can also be added to other IO::AIO::GRP objects.
		1770
		1771	=item * You must not add requests to a group from within the group callback (or
		1772	any later time).
		1773
		1774	=back
		1775
		1776	Their lifetime, simplified, looks like this: when they are empty, they
		1777	will finish very quickly. If they contain only requests that are in the
		1778	C<done> state, they will also finish. Otherwise they will continue to
		1779	exist.
		1780
		1781	That means after creating a group you have some time to add requests
		1782	(precisely before the callback has been invoked, which is only done within
		1783	the C<poll_cb>). And in the callbacks of those requests, you can add
		1784	further requests to the group. And only when all those requests have
		1785	finished will the the group itself finish.
		1786
		1787	=over 4
		1788
		1789	=item add $grp ...
		1790
		1791	=item $grp->add (...)
		1792
		1793	Add one or more requests to the group. Any type of L<IO::AIO::REQ> can
		1794	be added, including other groups, as long as you do not create circular
		1795	dependencies.
		1796
		1797	Returns all its arguments.
		1798
		1799	=item $grp->cancel_subs
		1800
		1801	Cancel all subrequests and clears any feeder, but not the group request
		1802	itself. Useful when you queued a lot of events but got a result early.
		1803
		1804	The group request will finish normally (you cannot add requests to the
		1805	group).
		1806
		1807	=item $grp->result (...)
		1808
		1809	Set the result value(s) that will be passed to the group callback when all
		1810	subrequests have finished and set the groups errno to the current value
		1811	of errno (just like calling C<errno> without an error number). By default,
		1812	no argument will be passed and errno is zero.
		1813
		1814	=item $grp->errno ([$errno])
		1815
		1816	Sets the group errno value to C<$errno>, or the current value of errno
		1817	when the argument is missing.
		1818
		1819	Every aio request has an associated errno value that is restored when
		1820	the callback is invoked. This method lets you change this value from its
		1821	default (0).
		1822
		1823	Calling C<result> will also set errno, so make sure you either set C<$!>
		1824	before the call to C<result>, or call c<errno> after it.
		1825
		1826	=item feed $grp $callback->($grp)
		1827
		1828	Sets a feeder/generator on this group: every group can have an attached
		1829	generator that generates requests if idle. The idea behind this is that,
		1830	although you could just queue as many requests as you want in a group,
		1831	this might starve other requests for a potentially long time. For example,
		1832	C<aio_scandir> might generate hundreds of thousands of C<aio_stat>
		1833	requests, delaying any later requests for a long time.
		1834
		1835	To avoid this, and allow incremental generation of requests, you can
		1836	instead a group and set a feeder on it that generates those requests. The
		1837	feed callback will be called whenever there are few enough (see C<limit>,
		1838	below) requests active in the group itself and is expected to queue more
		1839	requests.
		1840
		1841	The feed callback can queue as many requests as it likes (i.e. C<add> does
		1842	not impose any limits).
		1843
		1844	If the feed does not queue more requests when called, it will be
		1845	automatically removed from the group.
		1846
		1847	If the feed limit is C<0> when this method is called, it will be set to
		1848	C<2> automatically.
		1849
		1850	Example:
		1851
		1852	# stat all files in @files, but only ever use four aio requests concurrently:
		1853
		1854	my $grp = aio_group sub { print "finished\n" };
		1855	limit $grp 4;
		1856	feed $grp sub {
		1857	my $file = pop @files
		1858	or return;
		1859
		1860	add $grp aio_stat $file, sub { ... };
		1861	};
		1862
		1863	=item limit $grp $num
		1864
		1865	Sets the feeder limit for the group: The feeder will be called whenever
		1866	the group contains less than this many requests.
		1867
		1868	Setting the limit to C<0> will pause the feeding process.
		1869
		1870	The default value for the limit is C<0>, but note that setting a feeder
		1871	automatically bumps it up to C<2>.
		1872
		1873	=back
		1874
		1875
210	=head2 SUPPORT FUNCTIONS	1876	=head2 SUPPORT FUNCTIONS
211		1877
		1878	=head3 EVENT PROCESSING AND EVENT LOOP INTEGRATION
		1879
212	=over 4	1880	=over 4
213		1881
214	=item $fileno = IO::AIO::poll_fileno	1882	=item $fileno = IO::AIO::poll_fileno
215		1883
216	Return the I<request result pipe file descriptor>. This filehandle must be	1884	Return the I<request result pipe file descriptor>. This filehandle must be
217	polled for reading by some mechanism outside this module (e.g. Event or	1885	polled for reading by some mechanism outside this module (e.g. EV, Glib,
218	select, see below or the SYNOPSIS). If the pipe becomes readable you have	1886	select and so on, see below or the SYNOPSIS). If the pipe becomes readable
219	to call C<poll_cb> to check the results.	1887	you have to call C<poll_cb> to check the results.
220		1888
221	See C<poll_cb> for an example.	1889	See C<poll_cb> for an example.
222		1890
223	=item IO::AIO::poll_cb	1891	=item IO::AIO::poll_cb
224		1892
225	Process all outstanding events on the result pipe. You have to call this	1893	Process some requests that have reached the result phase (i.e. they have
226	regularly. Returns the number of events processed. Returns immediately	1894	been executed but the results are not yet reported). You have to call
227	when no events are outstanding.	1895	this "regularly" to finish outstanding requests.
		1896
		1897	Returns C<0> if all events could be processed (or there were no
		1898	events to process), or C<-1> if it returned earlier for whatever
		1899	reason. Returns immediately when no events are outstanding. The amount
		1900	of events processed depends on the settings of C<IO::AIO::max_poll_req>,
		1901	C<IO::AIO::max_poll_time> and C<IO::AIO::max_outstanding>.
		1902
		1903	If not all requests were processed for whatever reason, the poll file
		1904	descriptor will still be ready when C<poll_cb> returns, so normally you
		1905	don't have to do anything special to have it called later.
		1906
		1907	Apart from calling C<IO::AIO::poll_cb> when the event filehandle becomes
		1908	ready, it can be beneficial to call this function from loops which submit
		1909	a lot of requests, to make sure the results get processed when they become
		1910	available and not just when the loop is finished and the event loop takes
		1911	over again. This function returns very fast when there are no outstanding
		1912	requests.
228		1913
229	Example: Install an Event watcher that automatically calls	1914	Example: Install an Event watcher that automatically calls
230	IO::AIO::poll_cb with high priority:	1915	IO::AIO::poll_cb with high priority (more examples can be found in the
		1916	SYNOPSIS section, at the top of this document):
231		1917
232	Event->io (fd => IO::AIO::poll_fileno,	1918	Event->io (fd => IO::AIO::poll_fileno,
233	poll => 'r', async => 1,	1919	poll => 'r', async => 1,
234	cb => \&IO::AIO::poll_cb);	1920	cb => \&IO::AIO::poll_cb);
235		1921
236	=item IO::AIO::poll_wait	1922	=item IO::AIO::poll_wait
237		1923
238	Wait till the result filehandle becomes ready for reading (simply does a	1924	Wait until either at least one request is in the result phase or no
239	C<select> on the filehandle. This is useful if you want to synchronously wait	1925	requests are outstanding anymore.
240	for some requests to finish).	1926
		1927	This is useful if you want to synchronously wait for some requests to
		1928	become ready, without actually handling them.
241		1929
242	See C<nreqs> for an example.	1930	See C<nreqs> for an example.
243		1931
		1932	=item IO::AIO::poll
		1933
		1934	Waits until some requests have been handled.
		1935
		1936	Returns the number of requests processed, but is otherwise strictly
		1937	equivalent to:
		1938
		1939	IO::AIO::poll_wait, IO::AIO::poll_cb
		1940
244	=item IO::AIO::nreqs	1941	=item IO::AIO::flush
245		1942
246	Returns the number of requests currently outstanding (i.e. for which their	1943	Wait till all outstanding AIO requests have been handled.
247	callback has not been invoked yet).
248		1944
249	Example: wait till there are no outstanding requests anymore:	1945	Strictly equivalent to:
250		1946
251	IO::AIO::poll_wait, IO::AIO::poll_cb	1947	IO::AIO::poll_wait, IO::AIO::poll_cb
252	while IO::AIO::nreqs;	1948	while IO::AIO::nreqs;
253		1949
		1950	This function can be useful at program aborts, to make sure outstanding
		1951	I/O has been done (C<IO::AIO> uses an C<END> block which already calls
		1952	this function on normal exits), or when you are merely using C<IO::AIO>
		1953	for its more advanced functions, rather than for async I/O, e.g.:
		1954
		1955	my ($dirs, $nondirs);
		1956	IO::AIO::aio_scandir "/tmp", 0, sub { ($dirs, $nondirs) = @_ };
		1957	IO::AIO::flush;
		1958	# $dirs, $nondirs are now set
		1959
		1960	=item IO::AIO::max_poll_reqs $nreqs
		1961
		1962	=item IO::AIO::max_poll_time $seconds
		1963
		1964	These set the maximum number of requests (default C<0>, meaning infinity)
		1965	that are being processed by C<IO::AIO::poll_cb> in one call, respectively
		1966	the maximum amount of time (default C<0>, meaning infinity) spent in
		1967	C<IO::AIO::poll_cb> to process requests (more correctly the mininum amount
		1968	of time C<poll_cb> is allowed to use).
		1969
		1970	Setting C<max_poll_time> to a non-zero value creates an overhead of one
		1971	syscall per request processed, which is not normally a problem unless your
		1972	callbacks are really really fast or your OS is really really slow (I am
		1973	not mentioning Solaris here). Using C<max_poll_reqs> incurs no overhead.
		1974
		1975	Setting these is useful if you want to ensure some level of
		1976	interactiveness when perl is not fast enough to process all requests in
		1977	time.
		1978
		1979	For interactive programs, values such as C<0.01> to C<0.1> should be fine.
		1980
		1981	Example: Install an Event watcher that automatically calls
		1982	IO::AIO::poll_cb with low priority, to ensure that other parts of the
		1983	program get the CPU sometimes even under high AIO load.
		1984
		1985	# try not to spend much more than 0.1s in poll_cb
		1986	IO::AIO::max_poll_time 0.1;
		1987
		1988	# use a low priority so other tasks have priority
		1989	Event->io (fd => IO::AIO::poll_fileno,
		1990	poll => 'r', nice => 1,
		1991	cb => &IO::AIO::poll_cb);
		1992
		1993	=back
		1994
		1995
		1996	=head3 CONTROLLING THE NUMBER OF THREADS
		1997
		1998	=over
		1999
		2000	=item IO::AIO::min_parallel $nthreads
		2001
		2002	Set the minimum number of AIO threads to C<$nthreads>. The current
		2003	default is C<8>, which means eight asynchronous operations can execute
		2004	concurrently at any one time (the number of outstanding requests,
		2005	however, is unlimited).
		2006
		2007	IO::AIO starts threads only on demand, when an AIO request is queued and
		2008	no free thread exists. Please note that queueing up a hundred requests can
		2009	create demand for a hundred threads, even if it turns out that everything
		2010	is in the cache and could have been processed faster by a single thread.
		2011
		2012	It is recommended to keep the number of threads relatively low, as some
		2013	Linux kernel versions will scale negatively with the number of threads
		2014	(higher parallelity => MUCH higher latency). With current Linux 2.6
		2015	versions, 4-32 threads should be fine.
		2016
		2017	Under most circumstances you don't need to call this function, as the
		2018	module selects a default that is suitable for low to moderate load.
		2019
		2020	=item IO::AIO::max_parallel $nthreads
		2021
		2022	Sets the maximum number of AIO threads to C<$nthreads>. If more than the
		2023	specified number of threads are currently running, this function kills
		2024	them. This function blocks until the limit is reached.
		2025
		2026	While C<$nthreads> are zero, aio requests get queued but not executed
		2027	until the number of threads has been increased again.
		2028
		2029	This module automatically runs C<max_parallel 0> at program end, to ensure
		2030	that all threads are killed and that there are no outstanding requests.
		2031
		2032	Under normal circumstances you don't need to call this function.
		2033
		2034	=item IO::AIO::max_idle $nthreads
		2035
		2036	Limit the number of threads (default: 4) that are allowed to idle
		2037	(i.e., threads that did not get a request to process within the idle
		2038	timeout (default: 10 seconds). That means if a thread becomes idle while
		2039	C<$nthreads> other threads are also idle, it will free its resources and
		2040	exit.
		2041
		2042	This is useful when you allow a large number of threads (e.g. 100 or 1000)
		2043	to allow for extremely high load situations, but want to free resources
		2044	under normal circumstances (1000 threads can easily consume 30MB of RAM).
		2045
		2046	The default is probably ok in most situations, especially if thread
		2047	creation is fast. If thread creation is very slow on your system you might
		2048	want to use larger values.
		2049
		2050	=item IO::AIO::idle_timeout $seconds
		2051
		2052	Sets the minimum idle timeout (default 10) after which worker threads are
		2053	allowed to exit. SEe C<IO::AIO::max_idle>.
		2054
		2055	=item IO::AIO::max_outstanding $maxreqs
		2056
		2057	Sets the maximum number of outstanding requests to C<$nreqs>. If
		2058	you do queue up more than this number of requests, the next call to
		2059	C<IO::AIO::poll_cb> (and other functions calling C<poll_cb>, such as
		2060	C<IO::AIO::flush> or C<IO::AIO::poll>) will block until the limit is no
		2061	longer exceeded.
		2062
		2063	In other words, this setting does not enforce a queue limit, but can be
		2064	used to make poll functions block if the limit is exceeded.
		2065
		2066	This is a very bad function to use in interactive programs because it
		2067	blocks, and a bad way to reduce concurrency because it is inexact: Better
		2068	use an C<aio_group> together with a feed callback.
		2069
		2070	Its main use is in scripts without an event loop - when you want to stat
		2071	a lot of files, you can write something like this:
		2072
		2073	IO::AIO::max_outstanding 32;
		2074
		2075	for my $path (...) {
		2076	aio_stat $path , ...;
		2077	IO::AIO::poll_cb;
		2078	}
		2079
		2080	IO::AIO::flush;
		2081
		2082	The call to C<poll_cb> inside the loop will normally return instantly, but
		2083	as soon as more thna C<32> reqeusts are in-flight, it will block until
		2084	some requests have been handled. This keeps the loop from pushing a large
		2085	number of C<aio_stat> requests onto the queue.
		2086
		2087	The default value for C<max_outstanding> is very large, so there is no
		2088	practical limit on the number of outstanding requests.
		2089
		2090	=back
		2091
		2092
		2093	=head3 STATISTICAL INFORMATION
		2094
		2095	=over
		2096
254	=item IO::AIO::flush	2097	=item IO::AIO::nreqs
255		2098
256	Wait till all outstanding AIO requests have been handled.	2099	Returns the number of requests currently in the ready, execute or pending
		2100	states (i.e. for which their callback has not been invoked yet).
257		2101
258	Strictly equivalent to:	2102	Example: wait till there are no outstanding requests anymore:
259		2103
260	IO::AIO::poll_wait, IO::AIO::poll_cb	2104	IO::AIO::poll_wait, IO::AIO::poll_cb
261	while IO::AIO::nreqs;	2105	while IO::AIO::nreqs;
262		2106
		2107	=item IO::AIO::nready
		2108
		2109	Returns the number of requests currently in the ready state (not yet
		2110	executed).
		2111
263	=item IO::AIO::poll	2112	=item IO::AIO::npending
264		2113
265	Waits until some requests have been handled.	2114	Returns the number of requests currently in the pending state (executed,
266		2115	but not yet processed by poll_cb).
267	Strictly equivalent to:
268
269	IO::AIO::poll_wait, IO::AIO::poll_cb
270	if IO::AIO::nreqs;
271
272	=item IO::AIO::min_parallel $nthreads
273
274	Set the minimum number of AIO threads to C<$nthreads>. The default is
275	C<1>, which means a single asynchronous operation can be done at one time
276	(the number of outstanding operations, however, is unlimited).
277
278	It is recommended to keep the number of threads low, as some Linux
279	kernel versions will scale negatively with the number of threads (higher
280	parallelity => MUCH higher latency). With current Linux 2.6 versions, 4-32
281	threads should be fine.
282
283	Under normal circumstances you don't need to call this function, as this
284	module automatically starts some threads (the exact number might change,
285	and is currently 4).
286
287	=item IO::AIO::max_parallel $nthreads
288
289	Sets the maximum number of AIO threads to C<$nthreads>. If more than
290	the specified number of threads are currently running, kill them. This
291	function blocks until the limit is reached.
292
293	This module automatically runs C<max_parallel 0> at program end, to ensure
294	that all threads are killed and that there are no outstanding requests.
295
296	Under normal circumstances you don't need to call this function.
297
298	=item $oldnreqs = IO::AIO::max_outstanding $nreqs
299
300	Sets the maximum number of outstanding requests to C<$nreqs>. If you
301	try to queue up more than this number of requests, the caller will block until
302	some requests have been handled.
303
304	The default is very large, so normally there is no practical limit. If you
305	queue up many requests in a loop it it often improves speed if you set
306	this to a relatively low number, such as C<100>.
307
308	Under normal circumstances you don't need to call this function.
309		2116
310	=back	2117	=back
311		2118
		2119
		2120	=head3 SUBSECOND STAT TIME ACCESS
		2121
		2122	Both C<aio_stat>/C<aio_lstat> and perl's C<stat>/C<lstat> functions can
		2123	generally find access/modification and change times with subsecond time
		2124	accuracy of the system supports it, but perl's built-in functions only
		2125	return the integer part.
		2126
		2127	The following functions return the timestamps of the most recent
		2128	stat with subsecond precision on most systems and work both after
		2129	C<aio_stat>/C<aio_lstat> and perl's C<stat>/C<lstat> calls. Their return
		2130	value is only meaningful after a successful C<stat>/C<lstat> call, or
		2131	during/after a successful C<aio_stat>/C<aio_lstat> callback.
		2132
		2133	This is similar to the L<Time::HiRes> C<stat> functions, but can return
		2134	full resolution without rounding and work with standard perl C<stat>,
		2135	alleviating the need to call the special C<Time::HiRes> functions, which
		2136	do not act like their perl counterparts.
		2137
		2138	On operating systems or file systems where subsecond time resolution is
		2139	not supported or could not be detected, a fractional part of C<0> is
		2140	returned, so it is always safe to call these functions.
		2141
		2142	=over 4
		2143
		2144	=item $seconds = IO::AIO::st_atime, IO::AIO::st_mtime, IO::AIO::st_ctime, IO::AIO::st_btime
		2145
		2146	Return the access, modication, change or birth time, respectively,
		2147	including fractional part. Due to the limited precision of floating point,
		2148	the accuracy on most platforms is only a bit better than milliseconds
		2149	for times around now - see the I<nsec> function family, below, for full
		2150	accuracy.
		2151
		2152	File birth time is only available when the OS and perl support it (on
		2153	FreeBSD and NetBSD at the time of this writing, although support is
		2154	adaptive, so if your OS/perl gains support, IO::AIO can take advantage of
		2155	it). On systems where it isn't available, C<0> is currently returned, but
		2156	this might change to C<undef> in a future version.
		2157
		2158	=item ($atime, $mtime, $ctime, $btime, ...) = IO::AIO::st_xtime
		2159
		2160	Returns access, modification, change and birth time all in one go, and
		2161	maybe more times in the future version.
		2162
		2163	=item $nanoseconds = IO::AIO::st_atimensec, IO::AIO::st_mtimensec, IO::AIO::st_ctimensec, IO::AIO::st_btimensec
		2164
		2165	Return the fractional access, modifcation, change or birth time, in nanoseconds,
		2166	as an integer in the range C<0> to C<999999999>.
		2167
		2168	Note that no accessors are provided for access, modification and
		2169	change times - you need to get those from C<stat _> if required (C<int
		2170	IO::AIO::st_atime> and so on will I<not> generally give you the correct
		2171	value).
		2172
		2173	=item $seconds = IO::AIO::st_btimesec
		2174
		2175	The (integral) seconds part of the file birth time, if available.
		2176
		2177	=item ($atime, $mtime, $ctime, $btime, ...) = IO::AIO::st_xtimensec
		2178
		2179	Like the functions above, but returns all four times in one go (and maybe
		2180	more in future versions).
		2181
		2182	=item $counter = IO::AIO::st_gen
		2183
		2184	Returns the generation counter (in practice this is just a random number)
		2185	of the file. This is only available on platforms which have this member in
		2186	their C<struct stat> (most BSDs at the time of this writing) and generally
		2187	only to the root usert. If unsupported, C<0> is returned, but this might
		2188	change to C<undef> in a future version.
		2189
		2190	=back
		2191
		2192	Example: print the high resolution modification time of F</etc>, using
		2193	C<stat>, and C<IO::AIO::aio_stat>.
		2194
		2195	if (stat "/etc") {
		2196	printf "stat(/etc) mtime: %f\n", IO::AIO::st_mtime;
		2197	}
		2198
		2199	IO::AIO::aio_stat "/etc", sub {
		2200	$_[0]
		2201	and return;
		2202
		2203	printf "aio_stat(/etc) mtime: %d.%09d\n", (stat _)[9], IO::AIO::st_mtimensec;
		2204	};
		2205
		2206	IO::AIO::flush;
		2207
		2208	Output of the awbove on my system, showing reduced and full accuracy:
		2209
		2210	stat(/etc) mtime: 1534043702.020808
		2211	aio_stat(/etc) mtime: 1534043702.020807792
		2212
		2213
		2214	=head3 MISCELLANEOUS FUNCTIONS
		2215
		2216	IO::AIO implements some functions that are useful when you want to use
		2217	some "Advanced I/O" function not available to in Perl, without going the
		2218	"Asynchronous I/O" route. Many of these have an asynchronous C<aio_*>
		2219	counterpart.
		2220
		2221	=over 4
		2222
		2223	=item $numfd = IO::AIO::get_fdlimit
		2224
		2225	Tries to find the current file descriptor limit and returns it, or
		2226	C<undef> and sets C<$!> in case of an error. The limit is one larger than
		2227	the highest valid file descriptor number.
		2228
		2229	=item IO::AIO::min_fdlimit [$numfd]
		2230
		2231	Try to increase the current file descriptor limit(s) to at least C<$numfd>
		2232	by changing the soft or hard file descriptor resource limit. If C<$numfd>
		2233	is missing, it will try to set a very high limit, although this is not
		2234	recommended when you know the actual minimum that you require.
		2235
		2236	If the limit cannot be raised enough, the function makes a best-effort
		2237	attempt to increase the limit as much as possible, using various
		2238	tricks, while still failing. You can query the resulting limit using
		2239	C<IO::AIO::get_fdlimit>.
		2240
		2241	If an error occurs, returns C<undef> and sets C<$!>, otherwise returns
		2242	true.
		2243
		2244	=item IO::AIO::sendfile $ofh, $ifh, $offset, $count
		2245
		2246	Calls the C<eio_sendfile_sync> function, which is like C<aio_sendfile>,
		2247	but is blocking (this makes most sense if you know the input data is
		2248	likely cached already and the output filehandle is set to non-blocking
		2249	operations).
		2250
		2251	Returns the number of bytes copied, or C<-1> on error.
		2252
		2253	=item IO::AIO::fadvise $fh, $offset, $len, $advice
		2254
		2255	Simply calls the C<posix_fadvise> function (see its
		2256	manpage for details). The following advice constants are
		2257	available: C<IO::AIO::FADV_NORMAL>, C<IO::AIO::FADV_SEQUENTIAL>,
		2258	C<IO::AIO::FADV_RANDOM>, C<IO::AIO::FADV_NOREUSE>,
		2259	C<IO::AIO::FADV_WILLNEED>, C<IO::AIO::FADV_DONTNEED>.
		2260
		2261	On systems that do not implement C<posix_fadvise>, this function returns
		2262	ENOSYS, otherwise the return value of C<posix_fadvise>.
		2263
		2264	=item IO::AIO::madvise $scalar, $offset, $len, $advice
		2265
		2266	Simply calls the C<posix_madvise> function (see its
		2267	manpage for details). The following advice constants are
		2268	available: C<IO::AIO::MADV_NORMAL>, C<IO::AIO::MADV_SEQUENTIAL>,
		2269	C<IO::AIO::MADV_RANDOM>, C<IO::AIO::MADV_WILLNEED>,
		2270	C<IO::AIO::MADV_DONTNEED>.
		2271
		2272	If C<$offset> is negative, counts from the end. If C<$length> is negative,
		2273	the remaining length of the C<$scalar> is used. If possible, C<$length>
		2274	will be reduced to fit into the C<$scalar>.
		2275
		2276	On systems that do not implement C<posix_madvise>, this function returns
		2277	ENOSYS, otherwise the return value of C<posix_madvise>.
		2278
		2279	=item IO::AIO::mprotect $scalar, $offset, $len, $protect
		2280
		2281	Simply calls the C<mprotect> function on the preferably AIO::mmap'ed
		2282	$scalar (see its manpage for details). The following protect
		2283	constants are available: C<IO::AIO::PROT_NONE>, C<IO::AIO::PROT_READ>,
		2284	C<IO::AIO::PROT_WRITE>, C<IO::AIO::PROT_EXEC>.
		2285
		2286	If C<$offset> is negative, counts from the end. If C<$length> is negative,
		2287	the remaining length of the C<$scalar> is used. If possible, C<$length>
		2288	will be reduced to fit into the C<$scalar>.
		2289
		2290	On systems that do not implement C<mprotect>, this function returns
		2291	ENOSYS, otherwise the return value of C<mprotect>.
		2292
		2293	=item IO::AIO::mmap $scalar, $length, $prot, $flags, $fh[, $offset]
		2294
		2295	Memory-maps a file (or anonymous memory range) and attaches it to the
		2296	given C<$scalar>, which will act like a string scalar. Returns true on
		2297	success, and false otherwise.
		2298
		2299	The scalar must exist, but its contents do not matter - this means you
		2300	cannot use a nonexistant array or hash element. When in doubt, C<undef>
		2301	the scalar first.
		2302
		2303	The only operations allowed on the mmapped scalar are C<substr>/C<vec>,
		2304	which don't change the string length, and most read-only operations such
		2305	as copying it or searching it with regexes and so on.
		2306
		2307	Anything else is unsafe and will, at best, result in memory leaks.
		2308
		2309	The memory map associated with the C<$scalar> is automatically removed
		2310	when the C<$scalar> is undef'd or destroyed, or when the C<IO::AIO::mmap>
		2311	or C<IO::AIO::munmap> functions are called on it.
		2312
		2313	This calls the C<mmap>(2) function internally. See your system's manual
		2314	page for details on the C<$length>, C<$prot> and C<$flags> parameters.
		2315
		2316	The C<$length> must be larger than zero and smaller than the actual
		2317	filesize.
		2318
		2319	C<$prot> is a combination of C<IO::AIO::PROT_NONE>, C<IO::AIO::PROT_EXEC>,
		2320	C<IO::AIO::PROT_READ> and/or C<IO::AIO::PROT_WRITE>,
		2321
		2322	C<$flags> can be a combination of
		2323	C<IO::AIO::MAP_SHARED> or
		2324	C<IO::AIO::MAP_PRIVATE>,
		2325	or a number of system-specific flags (when not available, the are C<0>):
		2326	C<IO::AIO::MAP_ANONYMOUS> (which is set to C<MAP_ANON> if your system only provides this constant),
		2327	C<IO::AIO::MAP_LOCKED>,
		2328	C<IO::AIO::MAP_NORESERVE>,
		2329	C<IO::AIO::MAP_POPULATE>,
		2330	C<IO::AIO::MAP_NONBLOCK>,
		2331	C<IO::AIO::MAP_FIXED>,
		2332	C<IO::AIO::MAP_GROWSDOWN>,
		2333	C<IO::AIO::MAP_32BIT>,
		2334	C<IO::AIO::MAP_HUGETLB> or
		2335	C<IO::AIO::MAP_STACK>.
		2336
		2337	If C<$fh> is C<undef>, then a file descriptor of C<-1> is passed.
		2338
		2339	C<$offset> is the offset from the start of the file - it generally must be
		2340	a multiple of C<IO::AIO::PAGESIZE> and defaults to C<0>.
		2341
		2342	Example:
		2343
		2344	use Digest::MD5;
		2345	use IO::AIO;
		2346
		2347	open my $fh, "<verybigfile"
		2348	or die "$!";
		2349
		2350	IO::AIO::mmap my $data, -s $fh, IO::AIO::PROT_READ, IO::AIO::MAP_SHARED, $fh
		2351	or die "verybigfile: $!";
		2352
		2353	my $fast_md5 = md5 $data;
		2354
		2355	=item IO::AIO::munmap $scalar
		2356
		2357	Removes a previous mmap and undefines the C<$scalar>.
		2358
		2359	=item IO::AIO::mremap $scalar, $new_length, $flags = MREMAP_MAYMOVE[, $new_address = 0]
		2360
		2361	Calls the Linux-specific mremap(2) system call. The C<$scalar> must have
		2362	been mapped by C<IO::AIO::mmap>, and C<$flags> must currently either be
		2363	C<0> or C<IO::AIO::MREMAP_MAYMOVE>.
		2364
		2365	Returns true if successful, and false otherwise. If the underlying mmapped
		2366	region has changed address, then the true value has the numerical value
		2367	C<1>, otherwise it has the numerical value C<0>:
		2368
		2369	my $success = IO::AIO::mremap $mmapped, 8192, IO::AIO::MREMAP_MAYMOVE
		2370	or die "mremap: $!";
		2371
		2372	if ($success*1) {
		2373	warn "scalar has chanegd address in memory\n";
		2374	}
		2375
		2376	C<IO::AIO::MREMAP_FIXED> and the C<$new_address> argument are currently
		2377	implemented, but not supported and might go away in a future version.
		2378
		2379	On systems where this call is not supported or is not emulated, this call
		2380	returns falls and sets C<$!> to C<ENOSYS>.
		2381
		2382	=item IO::AIO::mlockall $flags
		2383
		2384	Calls the C<eio_mlockall_sync> function, which is like C<aio_mlockall>,
		2385	but is blocking.
		2386
		2387	=item IO::AIO::munlock $scalar, $offset = 0, $length = undef
		2388
		2389	Calls the C<munlock> function, undoing the effects of a previous
		2390	C<aio_mlock> call (see its description for details).
		2391
		2392	=item IO::AIO::munlockall
		2393
		2394	Calls the C<munlockall> function.
		2395
		2396	On systems that do not implement C<munlockall>, this function returns
		2397	ENOSYS, otherwise the return value of C<munlockall>.
		2398
		2399	=item $fh = IO::AIO::accept4 $r_fh, $sockaddr, $sockaddr_maxlen, $flags
		2400
		2401	Uses the GNU/Linux C<accept4(2)> syscall, if available, to accept a socket
		2402	and return the new file handle on success, or sets C<$!> and returns
		2403	C<undef> on error.
		2404
		2405	The remote name of the new socket will be stored in C<$sockaddr>, which
		2406	will be extended to allow for at least C<$sockaddr_maxlen> octets. If the
		2407	socket name does not fit into C<$sockaddr_maxlen> octets, this is signaled
		2408	by returning a longer string in C<$sockaddr>, which might or might not be
		2409	truncated.
		2410
		2411	To accept name-less sockets, use C<undef> for C<$sockaddr> and C<0> for
		2412	C<$sockaddr_maxlen>.
		2413
		2414	The main reasons to use this syscall rather than portable C«accept(2)>
		2415	are that you can specify C<SOCK_NONBLOCK> and/or C<SOCK_CLOEXEC>
		2416	flags and you can accept name-less sockets by specifying C<0> for
		2417	C<$sockaddr_maxlen>, which is sadly not possible with perl's interface to
		2418	C<accept>.
		2419
		2420	=item IO::AIO::splice $r_fh, $r_off, $w_fh, $w_off, $length, $flags
		2421
		2422	Calls the GNU/Linux C<splice(2)> syscall, if available. If C<$r_off> or
		2423	C<$w_off> are C<undef>, then C<NULL> is passed for these, otherwise they
		2424	should be the file offset.
		2425
		2426	C<$r_fh> and C<$w_fh> should not refer to the same file, as splice might
		2427	silently corrupt the data in this case.
		2428
		2429	The following symbol flag values are available: C<IO::AIO::SPLICE_F_MOVE>,
		2430	C<IO::AIO::SPLICE_F_NONBLOCK>, C<IO::AIO::SPLICE_F_MORE> and
		2431	C<IO::AIO::SPLICE_F_GIFT>.
		2432
		2433	See the C<splice(2)> manpage for details.
		2434
		2435	=item IO::AIO::tee $r_fh, $w_fh, $length, $flags
		2436
		2437	Calls the GNU/Linux C<tee(2)> syscall, see its manpage and the
		2438	description for C<IO::AIO::splice> above for details.
		2439
		2440	=item $actual_size = IO::AIO::pipesize $r_fh[, $new_size]
		2441
		2442	Attempts to query or change the pipe buffer size. Obviously works only
		2443	on pipes, and currently works only on GNU/Linux systems, and fails with
		2444	C<-1>/C<ENOSYS> everywhere else. If anybody knows how to influence pipe buffer
		2445	size on other systems, drop me a note.
		2446
		2447	=item ($rfh, $wfh) = IO::AIO::pipe2 [$flags]
		2448
		2449	This is a direct interface to the Linux L<pipe2(2)> system call. If
		2450	C<$flags> is missing or C<0>, then this should be the same as a call to
		2451	perl's built-in C<pipe> function and create a new pipe, and works on
		2452	systems that lack the pipe2 syscall. On win32, this case invokes C<_pipe
		2453	(..., 4096, O_BINARY)>.
		2454
		2455	If C<$flags> is non-zero, it tries to invoke the pipe2 system call with
		2456	the given flags (Linux 2.6.27, glibc 2.9).
		2457
		2458	On success, the read and write file handles are returned.
		2459
		2460	On error, nothing will be returned. If the pipe2 syscall is missing and
		2461	C<$flags> is non-zero, fails with C<ENOSYS>.
		2462
		2463	Please refer to L<pipe2(2)> for more info on the C<$flags>, but at the
		2464	time of this writing, C<IO::AIO::O_CLOEXEC>, C<IO::AIO::O_NONBLOCK> and
		2465	C<IO::AIO::O_DIRECT> (Linux 3.4, for packet-based pipes) were supported.
		2466
		2467	Example: create a pipe race-free w.r.t. threads and fork:
		2468
		2469	my ($rfh, $wfh) = IO::AIO::pipe2 IO::AIO::O_CLOEXEC
		2470	or die "pipe2: $!\n";
		2471
		2472	=item $fh = IO::AIO::memfd_create $pathname[, $flags]
		2473
		2474	This is a direct interface to the Linux L<memfd_create(2)> system
		2475	call. The (unhelpful) default for C<$flags> is C<0>, but your default
		2476	should be C<IO::AIO::MFD_CLOEXEC>.
		2477
		2478	On success, the new memfd filehandle is returned, otherwise returns
		2479	C<undef>. If the memfd_create syscall is missing, fails with C<ENOSYS>.
		2480
		2481	Please refer to L<memfd_create(2)> for more info on this call.
		2482
		2483	The following C<$flags> values are available: C<IO::AIO::MFD_CLOEXEC>,
		2484	C<IO::AIO::MFD_ALLOW_SEALING> and C<IO::AIO::MFD_HUGETLB>.
		2485
		2486	Example: create a new memfd.
		2487
		2488	my $fh = IO::AIO::memfd_create "somenameforprocfd", IO::AIO::MFD_CLOEXEC
		2489	or die "m,emfd_create: $!\n";
		2490	=item $fh = IO::AIO::eventfd [$initval, [$flags]]
		2491
		2492	This is a direct interface to the Linux L<eventfd(2)> system call. The
		2493	(unhelpful) defaults for C<$initval> and C<$flags> are C<0> for both.
		2494
		2495	On success, the new eventfd filehandle is returned, otherwise returns
		2496	C<undef>. If the eventfd syscall is missing, fails with C<ENOSYS>.
		2497
		2498	Please refer to L<eventfd(2)> for more info on this call.
		2499
		2500	The following symbol flag values are available: C<IO::AIO::EFD_CLOEXEC>,
		2501	C<IO::AIO::EFD_NONBLOCK> and C<IO::AIO::EFD_SEMAPHORE> (Linux 2.6.30).
		2502
		2503	Example: create a new eventfd filehandle:
		2504
		2505	$fh = IO::AIO::eventfd 0, IO::AIO::EFD_CLOEXEC
		2506	or die "eventfd: $!\n";
		2507
		2508	=item $fh = IO::AIO::timerfd_create $clockid[, $flags]
		2509
		2510	This is a direct interface to the Linux L<timerfd_create(2)> system
		2511	call. The (unhelpful) default for C<$flags> is C<0>, but your default
		2512	should be C<IO::AIO::TFD_CLOEXEC>.
		2513
		2514	On success, the new timerfd filehandle is returned, otherwise returns
		2515	C<undef>. If the timerfd_create syscall is missing, fails with C<ENOSYS>.
		2516
		2517	Please refer to L<timerfd_create(2)> for more info on this call.
		2518
		2519	The following C<$clockid> values are
		2520	available: C<IO::AIO::CLOCK_REALTIME>, C<IO::AIO::CLOCK_MONOTONIC>
		2521	C<IO::AIO::CLOCK_CLOCK_BOOTTIME> (Linux 3.15)
		2522	C<IO::AIO::CLOCK_CLOCK_REALTIME_ALARM> (Linux 3.11) and
		2523	C<IO::AIO::CLOCK_CLOCK_BOOTTIME_ALARM> (Linux 3.11).
		2524
		2525	The following C<$flags> values are available (Linux
		2526	2.6.27): C<IO::AIO::TFD_NONBLOCK> and C<IO::AIO::TFD_CLOEXEC>.
		2527
		2528	Example: create a new timerfd and set it to one-second repeated alarms,
		2529	then wait for two alarms:
		2530
		2531	my $fh = IO::AIO::timerfd_create IO::AIO::CLOCK_BOOTTIME, IO::AIO::TFD_CLOEXEC
		2532	or die "timerfd_create: $!\n";
		2533
		2534	defined IO::AIO::timerfd_settime $fh, 0, 1, 1
		2535	or die "timerfd_settime: $!\n";
		2536
		2537	for (1..2) {
		2538	8 == sysread $fh, my $buf, 8
		2539	or die "timerfd read failure\n";
		2540
		2541	printf "number of expirations (likely 1): %d\n",
		2542	unpack "Q", $buf;
		2543	}
		2544
		2545	=item ($cur_interval, $cur_value) = IO::AIO::timerfd_settime $fh, $flags, $new_interval, $nbw_value
		2546
		2547	This is a direct interface to the Linux L<timerfd_settime(2)> system
		2548	call. Please refer to its manpage for more info on this call.
		2549
		2550	The new itimerspec is specified using two (possibly fractional) second
		2551	values, C<$new_interval> and C<$new_value>).
		2552
		2553	On success, the current interval and value are returned (as per
		2554	C<timerfd_gettime>). On failure, the empty list is returned.
		2555
		2556	The following C<$flags> values are
		2557	available: C<IO::AIO::TFD_TIMER_ABSTIME> and
		2558	C<IO::AIO::TFD_TIMER_CANCEL_ON_SET>.
		2559
		2560	See C<IO::AIO::timerfd_create> for a full example.
		2561
		2562	=item ($cur_interval, $cur_value) = IO::AIO::timerfd_gettime $fh
		2563
		2564	This is a direct interface to the Linux L<timerfd_gettime(2)> system
		2565	call. Please refer to its manpage for more info on this call.
		2566
		2567	On success, returns the current values of interval and value for the given
		2568	timerfd (as potentially fractional second values). On failure, the empty
		2569	list is returned.
		2570
		2571	=back
		2572
312	=cut	2573	=cut
313		2574
314	# support function to convert a fd into a perl filehandle
315	sub _fd2fh {
316	return undef if $_[0] < 0;
317
318	# try to generate nice filehandles
319	my $sym = "IO::AIO::fd#$_[0]";
320	local *$sym;
321	open *$sym, "+<&=$_[0]"
322	or return undef;
323
324	*$sym
325	}
326
327	min_parallel 4;	2575	min_parallel 8;
328		2576
329	END {	2577	END { flush }
330	max_parallel 0;
331	}
332		2578
333	1;	2579	1;
334		2580
		2581	=head1 EVENT LOOP INTEGRATION
		2582
		2583	It is recommended to use L<AnyEvent::AIO> to integrate IO::AIO
		2584	automatically into many event loops:
		2585
		2586	# AnyEvent integration (EV, Event, Glib, Tk, POE, urxvt, pureperl...)
		2587	use AnyEvent::AIO;
		2588
		2589	You can also integrate IO::AIO manually into many event loops, here are
		2590	some examples of how to do this:
		2591
		2592	# EV integration
		2593	my $aio_w = EV::io IO::AIO::poll_fileno, EV::READ, \&IO::AIO::poll_cb;
		2594
		2595	# Event integration
		2596	Event->io (fd => IO::AIO::poll_fileno,
		2597	poll => 'r',
		2598	cb => \&IO::AIO::poll_cb);
		2599
		2600	# Glib/Gtk2 integration
		2601	add_watch Glib::IO IO::AIO::poll_fileno,
		2602	in => sub { IO::AIO::poll_cb; 1 };
		2603
		2604	# Tk integration
		2605	Tk::Event::IO->fileevent (IO::AIO::poll_fileno, "",
		2606	readable => \&IO::AIO::poll_cb);
		2607
		2608	# Danga::Socket integration
		2609	Danga::Socket->AddOtherFds (IO::AIO::poll_fileno =>
		2610	\&IO::AIO::poll_cb);
		2611
		2612	=head2 FORK BEHAVIOUR
		2613
		2614	Usage of pthreads in a program changes the semantics of fork
		2615	considerably. Specifically, only async-safe functions can be called after
		2616	fork. Perl doesn't know about this, so in general, you cannot call fork
		2617	with defined behaviour in perl if pthreads are involved. IO::AIO uses
		2618	pthreads, so this applies, but many other extensions and (for inexplicable
		2619	reasons) perl itself often is linked against pthreads, so this limitation
		2620	applies to quite a lot of perls.
		2621
		2622	This module no longer tries to fight your OS, or POSIX. That means IO::AIO
		2623	only works in the process that loaded it. Forking is fully supported, but
		2624	using IO::AIO in the child is not.
		2625
		2626	You might get around by not I<using> IO::AIO before (or after)
		2627	forking. You could also try to call the L<IO::AIO::reinit> function in the
		2628	child:
		2629
		2630	=over 4
		2631
		2632	=item IO::AIO::reinit
		2633
		2634	Abandons all current requests and I/O threads and simply reinitialises all
		2635	data structures. This is not an operation supported by any standards, but
		2636	happens to work on GNU/Linux and some newer BSD systems.
		2637
		2638	The only reasonable use for this function is to call it after forking, if
		2639	C<IO::AIO> was used in the parent. Calling it while IO::AIO is active in
		2640	the process will result in undefined behaviour. Calling it at any time
		2641	will also result in any undefined (by POSIX) behaviour.
		2642
		2643	=back
		2644
		2645	=head2 LINUX-SPECIFIC CALLS
		2646
		2647	When a call is documented as "linux-specific" then this means it
		2648	originated on GNU/Linux. C<IO::AIO> will usually try to autodetect the
		2649	availability and compatibility of such calls regardless of the platform
		2650	it is compiled on, so platforms such as FreeBSD which often implement
		2651	these calls will work. When in doubt, call them and see if they fail wth
		2652	C<ENOSYS>.
		2653
		2654	=head2 MEMORY USAGE
		2655
		2656	Per-request usage:
		2657
		2658	Each aio request uses - depending on your architecture - around 100-200
		2659	bytes of memory. In addition, stat requests need a stat buffer (possibly
		2660	a few hundred bytes), readdir requires a result buffer and so on. Perl
		2661	scalars and other data passed into aio requests will also be locked and
		2662	will consume memory till the request has entered the done state.
		2663
		2664	This is not awfully much, so queuing lots of requests is not usually a
		2665	problem.
		2666
		2667	Per-thread usage:
		2668
		2669	In the execution phase, some aio requests require more memory for
		2670	temporary buffers, and each thread requires a stack and other data
		2671	structures (usually around 16k-128k, depending on the OS).
		2672
		2673	=head1 KNOWN BUGS
		2674
		2675	Known bugs will be fixed in the next release :)
		2676
		2677	=head1 KNOWN ISSUES
		2678
		2679	Calls that try to "import" foreign memory areas (such as C<IO::AIO::mmap>
		2680	or C<IO::AIO::aio_slurp>) do not work with generic lvalues, such as
		2681	non-created hash slots or other scalars I didn't think of. It's best to
		2682	avoid such and either use scalar variables or making sure that the scalar
		2683	exists (e.g. by storing C<undef>) and isn't "funny" (e.g. tied).
		2684
		2685	I am not sure anything can be done about this, so this is considered a
		2686	known issue, rather than a bug.
		2687
335	=head1 SEE ALSO	2688	=head1 SEE ALSO
336		2689
337	L<Coro>, L<Linux::AIO>.	2690	L<AnyEvent::AIO> for easy integration into event loops, L<Coro::AIO> for a
		2691	more natural syntax and L<IO::FDPass> for file descriptor passing.
338		2692
339	=head1 AUTHOR	2693	=head1 AUTHOR
340		2694
341	Marc Lehmann <schmorp@schmorp.de>	2695	Marc Lehmann <schmorp@schmorp.de>
342	http://home.schmorp.de/	2696	http://home.schmorp.de/

Diff Legend

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