ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/IO-AIO/AIO.pm

Comparing IO-AIO/AIO.pm (file contents):
Revision 1.1 by root, Sun Jul 10 17:07:44 2005 UTC vs.
Revision 1.212 by root, Thu Sep 29 22:42:15 2011 UTC

…		…
4		4
5	=head1 SYNOPSIS	5	=head1 SYNOPSIS
6		6
7	use IO::AIO;	7	use IO::AIO;
8		8
		9	aio_open "/etc/passwd", IO::AIO::O_RDONLY, 0, sub {
		10	my $fh = shift
		11	or die "/etc/passwd: $!";
		12	...
		13	};
		14
		15	aio_unlink "/tmp/file", sub { };
		16
		17	aio_read $fh, 30000, 1024, $buffer, 0, sub {
		18	$_[0] > 0 or die "read error: $!";
		19	};
		20
		21	# version 2+ has request and group objects
		22	use IO::AIO 2;
		23
		24	aioreq_pri 4; # give next request a very high priority
		25	my $req = aio_unlink "/tmp/file", sub { };
		26	$req->cancel; # cancel request if still in queue
		27
		28	my $grp = aio_group sub { print "all stats done\n" };
		29	add $grp aio_stat "..." for ...;
		30
9	=head1 DESCRIPTION	31	=head1 DESCRIPTION
10		32
11	This module implements asynchronous I/O using whatever means your	33	This module implements asynchronous I/O using whatever means your
12	operating system supports. Currently, it falls back to Linux::AIO if that	34	operating system supports. It is implemented as an interface to C<libeio>
13	module is available, or uses pthreads to emulato aio functionality.	35	(L<http://software.schmorp.de/pkg/libeio.html>).
14		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
15	Currently, in this module a number of threads are started that execute	53	In this version, a number of threads are started that execute your
16	your read/writes and signal their completion. You don't need thread	54	requests and signal their completion. You don't need thread support
17	support in your libc or perl, and the threads created by this module will	55	in perl, and the threads created by this module will not be visible
18	not be visible to the pthreads library.	56	to perl. In the future, this module might make use of the native aio
		57	functions available on many operating systems. However, they are often
		58	not well-supported or restricted (GNU/Linux doesn't allow them on normal
		59	files currently, for example), and they would only support aio_read and
		60	aio_write, so the remaining functionality would have to be implemented
		61	using threads anyway.
19		62
20	Although the module will work with in the presence of other threads, it is	63	Although the module will work in the presence of other (Perl-) threads,
21	not reentrant, so use appropriate locking yourself.	64	it is currently not reentrant in any way, so use appropriate locking
		65	yourself, always call C<poll_cb> from within the same thread, or never
		66	call C<poll_cb> (or other C<aio_> functions) recursively.
22		67
23	=head2 API NOTES	68	=head2 EXAMPLE
		69
		70	This is a simple example that uses the EV module and loads
		71	F</etc/passwd> asynchronously:
		72
		73	use Fcntl;
		74	use EV;
		75	use IO::AIO;
		76
		77	# register the IO::AIO callback with EV
		78	my $aio_w = EV::io IO::AIO::poll_fileno, EV::READ, \&IO::AIO::poll_cb;
		79
		80	# queue the request to open /etc/passwd
		81	aio_open "/etc/passwd", IO::AIO::O_RDONLY, 0, sub {
		82	my $fh = shift
		83	or die "error while opening: $!";
		84
		85	# stat'ing filehandles is generally non-blocking
		86	my $size = -s $fh;
		87
		88	# queue a request to read the file
		89	my $contents;
		90	aio_read $fh, 0, $size, $contents, 0, sub {
		91	$_[0] == $size
		92	or die "short read: $!";
		93
		94	close $fh;
		95
		96	# file contents now in $contents
		97	print $contents;
		98
		99	# exit event loop and program
		100	EV::unloop;
		101	};
		102	};
		103
		104	# possibly queue up other requests, or open GUI windows,
		105	# check for sockets etc. etc.
		106
		107	# process events as long as there are some:
		108	EV::loop;
		109
		110	=head1 REQUEST ANATOMY AND LIFETIME
		111
		112	Every C<aio_*> function creates a request. which is a C data structure not
		113	directly visible to Perl.
		114
		115	If called in non-void context, every request function returns a Perl
		116	object representing the request. In void context, nothing is returned,
		117	which saves a bit of memory.
		118
		119	The perl object is a fairly standard ref-to-hash object. The hash contents
		120	are not used by IO::AIO so you are free to store anything you like in it.
		121
		122	During their existance, aio requests travel through the following states,
		123	in order:
		124
		125	=over 4
		126
		127	=item ready
		128
		129	Immediately after a request is created it is put into the ready state,
		130	waiting for a thread to execute it.
		131
		132	=item execute
		133
		134	A thread has accepted the request for processing and is currently
		135	executing it (e.g. blocking in read).
		136
		137	=item pending
		138
		139	The request has been executed and is waiting for result processing.
		140
		141	While request submission and execution is fully asynchronous, result
		142	processing is not and relies on the perl interpreter calling C<poll_cb>
		143	(or another function with the same effect).
		144
		145	=item result
		146
		147	The request results are processed synchronously by C<poll_cb>.
		148
		149	The C<poll_cb> function will process all outstanding aio requests by
		150	calling their callbacks, freeing memory associated with them and managing
		151	any groups they are contained in.
		152
		153	=item done
		154
		155	Request has reached the end of its lifetime and holds no resources anymore
		156	(except possibly for the Perl object, but its connection to the actual
		157	aio request is severed and calling its methods will either do nothing or
		158	result in a runtime error).
		159
		160	=back
		161
		162	=cut
		163
		164	package IO::AIO;
		165
		166	use Carp ();
		167
		168	use common::sense;
		169
		170	use base 'Exporter';
		171
		172	BEGIN {
		173	our $VERSION = '4.0';
		174
		175	our @AIO_REQ = qw(aio_sendfile aio_read aio_write aio_open aio_close
		176	aio_stat aio_lstat aio_unlink aio_rmdir aio_readdir aio_readdirx
		177	aio_scandir aio_symlink aio_readlink aio_realpath aio_sync
		178	aio_fsync aio_syncfs aio_fdatasync aio_sync_file_range aio_fallocate
		179	aio_pathsync aio_readahead
		180	aio_rename aio_link aio_move aio_copy aio_group
		181	aio_nop aio_mknod aio_load aio_rmtree aio_mkdir aio_chown
		182	aio_chmod aio_utime aio_truncate
		183	aio_msync aio_mtouch aio_mlock aio_mlockall
		184	aio_statvfs
		185	aio_wd);
		186
		187	our @EXPORT = (@AIO_REQ, qw(aioreq_pri aioreq_nice));
		188	our @EXPORT_OK = qw(poll_fileno poll_cb poll_wait flush
		189	min_parallel max_parallel max_idle idle_timeout
		190	nreqs nready npending nthreads
		191	max_poll_time max_poll_reqs
		192	sendfile fadvise madvise
		193	mmap munmap munlock munlockall);
		194
		195	push @AIO_REQ, qw(aio_busy); # not exported
		196
		197	@IO::AIO::GRP::ISA = 'IO::AIO::REQ';
		198
		199	require XSLoader;
		200	XSLoader::load ("IO::AIO", $VERSION);
		201	}
		202
		203	=head1 FUNCTIONS
		204
		205	=head2 QUICK OVERVIEW
		206
		207	This section simply lists the prototypes of the most important functions
		208	for quick reference. See the following sections for function-by-function
		209	documentation.
		210
		211	aio_wd $pathname, $callback->($wd)
		212	aio_open $pathname, $flags, $mode, $callback->($fh)
		213	aio_close $fh, $callback->($status)
		214	aio_read $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
		215	aio_write $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
		216	aio_sendfile $out_fh, $in_fh, $in_offset, $length, $callback->($retval)
		217	aio_readahead $fh,$offset,$length, $callback->($retval)
		218	aio_stat $fh_or_path, $callback->($status)
		219	aio_lstat $fh, $callback->($status)
		220	aio_statvfs $fh_or_path, $callback->($statvfs)
		221	aio_utime $fh_or_path, $atime, $mtime, $callback->($status)
		222	aio_chown $fh_or_path, $uid, $gid, $callback->($status)
		223	aio_truncate $fh_or_path, $offset, $callback->($status)
		224	aio_chmod $fh_or_path, $mode, $callback->($status)
		225	aio_unlink $pathname, $callback->($status)
		226	aio_mknod $pathname, $mode, $dev, $callback->($status)
		227	aio_link $srcpath, $dstpath, $callback->($status)
		228	aio_symlink $srcpath, $dstpath, $callback->($status)
		229	aio_readlink $pathname, $callback->($link)
		230	aio_realpath $pathname, $callback->($link)
		231	aio_rename $srcpath, $dstpath, $callback->($status)
		232	aio_mkdir $pathname, $mode, $callback->($status)
		233	aio_rmdir $pathname, $callback->($status)
		234	aio_readdir $pathname, $callback->($entries)
		235	aio_readdirx $pathname, $flags, $callback->($entries, $flags)
		236	IO::AIO::READDIR_DENTS IO::AIO::READDIR_DIRS_FIRST
		237	IO::AIO::READDIR_STAT_ORDER IO::AIO::READDIR_FOUND_UNKNOWN
		238	aio_load $pathname, $data, $callback->($status)
		239	aio_copy $srcpath, $dstpath, $callback->($status)
		240	aio_move $srcpath, $dstpath, $callback->($status)
		241	aio_scandir $pathname, $maxreq, $callback->($dirs, $nondirs)
		242	aio_rmtree $pathname, $callback->($status)
		243	aio_sync $callback->($status)
		244	aio_syncfs $fh, $callback->($status)
		245	aio_fsync $fh, $callback->($status)
		246	aio_fdatasync $fh, $callback->($status)
		247	aio_sync_file_range $fh, $offset, $nbytes, $flags, $callback->($status)
		248	aio_pathsync $pathname, $callback->($status)
		249	aio_msync $scalar, $offset = 0, $length = undef, flags = 0, $callback->($status)
		250	aio_mtouch $scalar, $offset = 0, $length = undef, flags = 0, $callback->($status)
		251	aio_mlock $scalar, $offset = 0, $length = undef, $callback->($status)
		252	aio_mlockall $flags, $callback->($status)
		253	aio_group $callback->(...)
		254	aio_nop $callback->()
		255
		256	$prev_pri = aioreq_pri [$pri]
		257	aioreq_nice $pri_adjust
		258
		259	IO::AIO::poll_wait
		260	IO::AIO::poll_cb
		261	IO::AIO::poll
		262	IO::AIO::flush
		263	IO::AIO::max_poll_reqs $nreqs
		264	IO::AIO::max_poll_time $seconds
		265	IO::AIO::min_parallel $nthreads
		266	IO::AIO::max_parallel $nthreads
		267	IO::AIO::max_idle $nthreads
		268	IO::AIO::idle_timeout $seconds
		269	IO::AIO::max_outstanding $maxreqs
		270	IO::AIO::nreqs
		271	IO::AIO::nready
		272	IO::AIO::npending
		273
		274	IO::AIO::sendfile $ofh, $ifh, $offset, $count
		275	IO::AIO::fadvise $fh, $offset, $len, $advice
		276	IO::AIO::madvise $scalar, $offset, $length, $advice
		277	IO::AIO::mprotect $scalar, $offset, $length, $protect
		278	IO::AIO::munlock $scalar, $offset = 0, $length = undef
		279	IO::AIO::munlockall
		280
		281	=head2 AIO REQUEST FUNCTIONS
24		282
25	All the C<aio_*> calls are more or less thin wrappers around the syscall	283	All the C<aio_*> calls are more or less thin wrappers around the syscall
26	with the same name (sans C<aio_>). The arguments are similar or identical,	284	with the same name (sans C<aio_>). The arguments are similar or identical,
27	and they all accept an additional C<$callback> argument which must be	285	and they all accept an additional (and optional) C<$callback> argument
28	a code reference. This code reference will get called with the syscall	286	which must be a code reference. This code reference will be called after
		287	the syscall has been executed in an asynchronous fashion. The results
		288	of the request will be passed as arguments to the callback (and, if an
		289	error occured, in C<$!>) - for most requests the syscall return code (e.g.
29	return code (e.g. most syscalls return C<-1> on error, unlike perl, which	290	most syscalls return C<-1> on error, unlike perl, which usually delivers
30	usually delivers "false") as it's sole argument when the given syscall has	291	"false").
31	been executed asynchronously.
32		292
33	All functions that expect a filehandle will also accept a file descriptor.	293	Some requests (such as C<aio_readdir>) pass the actual results and
		294	communicate failures by passing C<undef>.
34		295
		296	All functions expecting a filehandle keep a copy of the filehandle
		297	internally until the request has finished.
		298
		299	All functions return request objects of type L<IO::AIO::REQ> that allow
		300	further manipulation of those requests while they are in-flight.
		301
35	The filenames you pass to these routines I<must> be absolute. The reason	302	The pathnames you pass to these routines I<should> be absolute. The
36	is that at the time the request is being executed, the current working	303	reason for this is that at the time the request is being executed, the
37	directory could have changed. Alternatively, you can make sure that you	304	current working directory could have changed. Alternatively, you can
38	never change the current working directory.	305	make sure that you never change the current working directory anywhere
		306	in the program and then use relative paths. You can also take advantage
		307	of IO::AIOs working directory abstraction, that lets you specify paths
		308	relative to some previously-opened "working directory object" - see the
		309	description of the C<IO::AIO::WD> class later in this document.
		310
		311	To encode pathnames as octets, either make sure you either: a) always pass
		312	in filenames you got from outside (command line, readdir etc.) without
		313	tinkering, b) are in your native filesystem encoding, c) use the Encode
		314	module and encode your pathnames to the locale (or other) encoding in
		315	effect in the user environment, d) use Glib::filename_from_unicode on
		316	unicode filenames or e) use something else to ensure your scalar has the
		317	correct contents.
		318
		319	This works, btw. independent of the internal UTF-8 bit, which IO::AIO
		320	handles correctly whether it is set or not.
39		321
40	=over 4	322	=over 4
41		323
42	=cut	324	=item $prev_pri = aioreq_pri [$pri]
43		325
44	package IO::AIO;	326	Returns the priority value that would be used for the next request and, if
		327	C<$pri> is given, sets the priority for the next aio request.
45		328
46	use base 'Exporter';	329	The default priority is C<0>, the minimum and maximum priorities are C<-4>
		330	and C<4>, respectively. Requests with higher priority will be serviced
		331	first.
47		332
48	BEGIN {	333	The priority will be reset to C<0> after each call to one of the C<aio_*>
49	$VERSION = 0.1;	334	functions.
50		335
51	@EXPORT = qw(aio_read aio_write aio_open aio_close aio_stat aio_lstat aio_unlink	336	Example: open a file with low priority, then read something from it with
52	aio_fsync aio_fdatasync aio_readahead);	337	higher priority so the read request is serviced before other low priority
53	@EXPORT_OK = qw(poll_fileno poll_cb min_parallel max_parallel nreqs);	338	open requests (potentially spamming the cache):
54		339
55	require XSLoader;	340	aioreq_pri -3;
56	XSLoader::load IO::AIO, $VERSION;	341	aio_open ..., sub {
57	}	342	return unless $_[0];
58		343
59	=item IO::AIO::min_parallel $nthreads	344	aioreq_pri -2;
		345	aio_read $_[0], ..., sub {
		346	...
		347	};
		348	};
60		349
61	Set the minimum number of AIO threads to C<$nthreads>. The default is
62	C<1>, which means a single asynchronous operation can be done at one time
63	(the number of outstanding operations, however, is unlimited).
64		350
65	It is recommended to keep the number of threads low, as some linux	351	=item aioreq_nice $pri_adjust
66	kernel versions will scale negatively with the number of threads (higher
67	parallelity => MUCH higher latency).
68		352
69	Under normal circumstances you don't need to call this function, as this	353	Similar to C<aioreq_pri>, but subtracts the given value from the current
70	module automatically starts a single async thread.	354	priority, so the effect is cumulative.
71		355
72	=item IO::AIO::max_parallel $nthreads
73		356
74	Sets the maximum number of AIO threads to C<$nthreads>. If more than
75	the specified number of threads are currently running, kill them. This
76	function blocks until the limit is reached.
77
78	This module automatically runs C<max_parallel 0> at program end, to ensure
79	that all threads are killed and that there are no outstanding requests.
80
81	Under normal circumstances you don't need to call this function.
82
83	=item $fileno = IO::AIO::poll_fileno
84
85	Return the I<request result pipe filehandle>. This filehandle must be
86	polled for reading by some mechanism outside this module (e.g. Event
87	or select, see below). If the pipe becomes readable you have to call
88	C<poll_cb> to check the results.
89
90	See C<poll_cb> for an example.
91
92	=item IO::AIO::poll_cb
93
94	Process all outstanding events on the result pipe. You have to call this
95	regularly. Returns the number of events processed. Returns immediately
96	when no events are outstanding.
97
98	You can use Event to multiplex, e.g.:
99
100	Event->io (fd => IO::AIO::poll_fileno,
101	poll => 'r', async => 1,
102	cb => \&IO::AIO::poll_cb);
103
104	=item IO::AIO::poll_wait
105
106	Wait till the result filehandle becomes ready for reading (simply does a
107	select on the filehandle. This is useful if you want to synchronously wait
108	for some requests to finish).
109
110	See C<nreqs> for an example.
111
112	=item IO::AIO::nreqs
113
114	Returns the number of requests currently outstanding.
115
116	Example: wait till there are no outstanding requests anymore:
117
118	IO::AIO::poll_wait, IO::AIO::poll_cb
119	while IO::AIO::nreqs;
120
121	=item aio_open $pathname, $flags, $mode, $callback	357	=item aio_open $pathname, $flags, $mode, $callback->($fh)
122		358
123	Asynchronously open or create a file and call the callback with the	359	Asynchronously open or create a file and call the callback with a newly
124	filedescriptor (NOT a perl filehandle, sorry for that, but watch out, this	360	created filehandle for the file.
125	might change in the future).
126		361
127	The pathname passed to C<aio_open> must be absolute. See API NOTES, above,	362	The pathname passed to C<aio_open> must be absolute. See API NOTES, above,
128	for an explanation.	363	for an explanation.
129		364
130	The C<$mode> argument is a bitmask. See the C<Fcntl> module for a	365	The C<$flags> argument is a bitmask. See the C<Fcntl> module for a
131	list. They are the same as used in C<sysopen>.	366	list. They are the same as used by C<sysopen>.
		367
		368	Likewise, C<$mode> specifies the mode of the newly created file, if it
		369	didn't exist and C<O_CREAT> has been given, just like perl's C<sysopen>,
		370	except that it is mandatory (i.e. use C<0> if you don't create new files,
		371	and C<0666> or C<0777> if you do). Note that the C<$mode> will be modified
		372	by the umask in effect then the request is being executed, so better never
		373	change the umask.
132		374
133	Example:	375	Example:
134		376
135	aio_open "/etc/passwd", O_RDONLY, 0, sub {	377	aio_open "/etc/passwd", IO::AIO::O_RDONLY, 0, sub {
136	if ($_[0] >= 0) {	378	if ($_[0]) {
137	open my $fh, "<&$_[0]"; # create a copy for perl
138	aio_close $_[0], sub { }; # close the aio handle
139	print "open successful, fh is $fh\n";	379	print "open successful, fh is $_[0]\n";
140	...	380	...
141	} else {	381	} else {
142	die "open failed: $!\n";	382	die "open failed: $!\n";
143	}	383	}
144	};	384	};
145		385
		386	In addition to all the common open modes/flags (C<O_RDONLY>, C<O_WRONLY>,
		387	C<O_RDWR>, C<O_CREAT>, C<O_TRUNC>, C<O_EXCL> and C<O_APPEND>), the
		388	following POSIX and non-POSIX constants are available (missing ones on
		389	your system are, as usual, C<0>):
		390
		391	C<O_ASYNC>, C<O_DIRECT>, C<O_NOATIME>, C<O_CLOEXEC>, C<O_NOCTTY>, C<O_NOFOLLOW>,
		392	C<O_NONBLOCK>, C<O_EXEC>, C<O_SEARCH>, C<O_DIRECTORY>, C<O_DSYNC>,
		393	C<O_RSYNC>, C<O_SYNC> and C<O_TTY_INIT>.
		394
		395
146	=item aio_close $fh, $callback	396	=item aio_close $fh, $callback->($status)
147		397
148	Asynchronously close a file and call the callback with the result code.	398	Asynchronously close a file and call the callback with the result
		399	code.
149		400
		401	Unfortunately, you can't do this to perl. Perl I<insists> very strongly on
		402	closing the file descriptor associated with the filehandle itself.
		403
		404	Therefore, C<aio_close> will not close the filehandle - instead it will
		405	use dup2 to overwrite the file descriptor with the write-end of a pipe
		406	(the pipe fd will be created on demand and will be cached).
		407
		408	Or in other words: the file descriptor will be closed, but it will not be
		409	free for reuse until the perl filehandle is closed.
		410
		411	=cut
		412
150	=item aio_read $fh,$offset,$length, $data,$dataoffset,$callback	413	=item aio_read $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
151		414
152	=item aio_write $fh,$offset,$length, $data,$dataoffset,$callback	415	=item aio_write $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
153		416
154	Reads or writes C<length> bytes from the specified C<fh> and C<offset>	417	Reads or writes C<$length> bytes from or to the specified C<$fh> and
155	into the scalar given by C<data> and offset C<dataoffset> and calls the	418	C<$offset> into the scalar given by C<$data> and offset C<$dataoffset>
156	callback without the actual number of bytes read (or -1 on error, just	419	and calls the callback without the actual number of bytes read (or -1 on
157	like the syscall).	420	error, just like the syscall).
158		421
		422	C<aio_read> will, like C<sysread>, shrink or grow the C<$data> scalar to
		423	offset plus the actual number of bytes read.
		424
		425	If C<$offset> is undefined, then the current file descriptor offset will
		426	be used (and updated), otherwise the file descriptor offset will not be
		427	changed by these calls.
		428
		429	If C<$length> is undefined in C<aio_write>, use the remaining length of
		430	C<$data>.
		431
		432	If C<$dataoffset> is less than zero, it will be counted from the end of
		433	C<$data>.
		434
		435	The C<$data> scalar I<MUST NOT> be modified in any way while the request
		436	is outstanding. Modifying it can result in segfaults or World War III (if
		437	the necessary/optional hardware is installed).
		438
159	Example: Read 15 bytes at offset 7 into scalar C<$buffer>, strating at	439	Example: Read 15 bytes at offset 7 into scalar C<$buffer>, starting at
160	offset C<0> within the scalar:	440	offset C<0> within the scalar:
161		441
162	aio_read $fh, 7, 15, $buffer, 0, sub {	442	aio_read $fh, 7, 15, $buffer, 0, sub {
163	$_[0] >= 0 or die "read error: $!";	443	$_[0] > 0 or die "read error: $!";
164	print "read <$buffer>\n";	444	print "read $_[0] bytes: <$buffer>\n";
165	};	445	};
166		446
		447
		448	=item aio_sendfile $out_fh, $in_fh, $in_offset, $length, $callback->($retval)
		449
		450	Tries to copy C<$length> bytes from C<$in_fh> to C<$out_fh>. It starts
		451	reading at byte offset C<$in_offset>, and starts writing at the current
		452	file offset of C<$out_fh>. Because of that, it is not safe to issue more
		453	than one C<aio_sendfile> per C<$out_fh>, as they will interfere with each
		454	other. The same C<$in_fh> works fine though, as this function does not
		455	move or use the file offset of C<$in_fh>.
		456
		457	Please note that C<aio_sendfile> can read more bytes from C<$in_fh> than
		458	are written, and there is no way to find out how many more bytes have been
		459	read from C<aio_sendfile> alone, as C<aio_sendfile> only provides the
		460	number of bytes written to C<$out_fh>. Only if the result value equals
		461	C<$length> one can assume that C<$length> bytes have been read.
		462
		463	Unlike with other C<aio_> functions, it makes a lot of sense to use
		464	C<aio_sendfile> on non-blocking sockets, as long as one end (typically
		465	the C<$in_fh>) is a file - the file I/O will then be asynchronous, while
		466	the socket I/O will be non-blocking. Note, however, that you can run
		467	into a trap where C<aio_sendfile> reads some data with readahead, then
		468	fails to write all data, and when the socket is ready the next time, the
		469	data in the cache is already lost, forcing C<aio_sendfile> to again hit
		470	the disk. Explicit C<aio_read> + C<aio_write> let's you better control
		471	resource usage.
		472
		473	This call tries to make use of a native C<sendfile>-like syscall to
		474	provide zero-copy operation. For this to work, C<$out_fh> should refer to
		475	a socket, and C<$in_fh> should refer to an mmap'able file.
		476
		477	If a native sendfile cannot be found or it fails with C<ENOSYS>,
		478	C<EINVAL>, C<ENOTSUP>, C<EOPNOTSUPP>, C<EAFNOSUPPORT>, C<EPROTOTYPE> or
		479	C<ENOTSOCK>, it will be emulated, so you can call C<aio_sendfile> on any
		480	type of filehandle regardless of the limitations of the operating system.
		481
		482	As native sendfile syscalls (as practically any non-POSIX interface hacked
		483	together in a hurry to improve benchmark numbers) tend to be rather buggy
		484	on many systems, this implementation tries to work around some known bugs
		485	in Linux and FreeBSD kernels (probably others, too), but that might fail,
		486	so you really really should check the return value of C<aio_sendfile> -
		487	fewre bytes than expected might have been transferred.
		488
		489
167	=item aio_readahead $fh,$offset,$length, $callback	490	=item aio_readahead $fh,$offset,$length, $callback->($retval)
168		491
169	Asynchronously reads the specified byte range into the page cache, using
170	the C<readahead> syscall. If that syscall doesn't exist the status will be
171	C<-1> and C<$!> is set to ENOSYS.
172
173	readahead() populates the page cache with data from a file so that	492	C<aio_readahead> populates the page cache with data from a file so that
174	subsequent reads from that file will not block on disk I/O. The C<$offset>	493	subsequent reads from that file will not block on disk I/O. The C<$offset>
175	argument specifies the starting point from which data is to be read and	494	argument specifies the starting point from which data is to be read and
176	C<$length> specifies the number of bytes to be read. I/O is performed in	495	C<$length> specifies the number of bytes to be read. I/O is performed in
177	whole pages, so that offset is effectively rounded down to a page boundary	496	whole pages, so that offset is effectively rounded down to a page boundary
178	and bytes are read up to the next page boundary greater than or equal to	497	and bytes are read up to the next page boundary greater than or equal to
179	(off-set+length). aio_readahead() does not read beyond the end of the	498	(off-set+length). C<aio_readahead> does not read beyond the end of the
180	file. The current file offset of the file is left unchanged.	499	file. The current file offset of the file is left unchanged.
181		500
		501	If that syscall doesn't exist (likely if your OS isn't Linux) it will be
		502	emulated by simply reading the data, which would have a similar effect.
		503
		504
182	=item aio_stat $fh_or_path, $callback	505	=item aio_stat $fh_or_path, $callback->($status)
183		506
184	=item aio_lstat $fh, $callback	507	=item aio_lstat $fh, $callback->($status)
185		508
186	Works like perl's C<stat> or C<lstat> in void context. The callback will	509	Works like perl's C<stat> or C<lstat> in void context. The callback will
187	be called after the stat and the results will be available using C<stat _>	510	be called after the stat and the results will be available using C<stat _>
188	or C<-s _> etc...	511	or C<-s _> etc...
189		512
…		…
191	for an explanation.	514	for an explanation.
192		515
193	Currently, the stats are always 64-bit-stats, i.e. instead of returning an	516	Currently, the stats are always 64-bit-stats, i.e. instead of returning an
194	error when stat'ing a large file, the results will be silently truncated	517	error when stat'ing a large file, the results will be silently truncated
195	unless perl itself is compiled with large file support.	518	unless perl itself is compiled with large file support.
		519
		520	To help interpret the mode and dev/rdev stat values, IO::AIO offers the
		521	following constants and functions (if not implemented, the constants will
		522	be C<0> and the functions will either C<croak> or fall back on traditional
		523	behaviour).
		524
		525	C<S_IFMT>, C<S_IFIFO>, C<S_IFCHR>, C<S_IFBLK>, C<S_IFLNK>, C<S_IFREG>,
		526	C<S_IFDIR>, C<S_IFWHT>, C<S_IFSOCK>, C<IO::AIO::major $dev_t>,
		527	C<IO::AIO::minor $dev_t>, C<IO::AIO::makedev $major, $minor>.
196		528
197	Example: Print the length of F</etc/passwd>:	529	Example: Print the length of F</etc/passwd>:
198		530
199	aio_stat "/etc/passwd", sub {	531	aio_stat "/etc/passwd", sub {
200	$_[0] and die "stat failed: $!";	532	$_[0] and die "stat failed: $!";
201	print "size is ", -s _, "\n";	533	print "size is ", -s _, "\n";
202	};	534	};
203		535
		536
		537	=item aio_statvfs $fh_or_path, $callback->($statvfs)
		538
		539	Works like the POSIX C<statvfs> or C<fstatvfs> syscalls, depending on
		540	whether a file handle or path was passed.
		541
		542	On success, the callback is passed a hash reference with the following
		543	members: C<bsize>, C<frsize>, C<blocks>, C<bfree>, C<bavail>, C<files>,
		544	C<ffree>, C<favail>, C<fsid>, C<flag> and C<namemax>. On failure, C<undef>
		545	is passed.
		546
		547	The following POSIX IO::AIO::ST_* constants are defined: C<ST_RDONLY> and
		548	C<ST_NOSUID>.
		549
		550	The following non-POSIX IO::AIO::ST_* flag masks are defined to
		551	their correct value when available, or to C<0> on systems that do
		552	not support them: C<ST_NODEV>, C<ST_NOEXEC>, C<ST_SYNCHRONOUS>,
		553	C<ST_MANDLOCK>, C<ST_WRITE>, C<ST_APPEND>, C<ST_IMMUTABLE>, C<ST_NOATIME>,
		554	C<ST_NODIRATIME> and C<ST_RELATIME>.
		555
		556	Example: stat C</wd> and dump out the data if successful.
		557
		558	aio_statvfs "/wd", sub {
		559	my $f = $_[0]
		560	or die "statvfs: $!";
		561
		562	use Data::Dumper;
		563	say Dumper $f;
		564	};
		565
		566	# result:
		567	{
		568	bsize => 1024,
		569	bfree => 4333064312,
		570	blocks => 10253828096,
		571	files => 2050765568,
		572	flag => 4096,
		573	favail => 2042092649,
		574	bavail => 4333064312,
		575	ffree => 2042092649,
		576	namemax => 255,
		577	frsize => 1024,
		578	fsid => 1810
		579	}
		580
		581
		582	=item aio_utime $fh_or_path, $atime, $mtime, $callback->($status)
		583
		584	Works like perl's C<utime> function (including the special case of $atime
		585	and $mtime being undef). Fractional times are supported if the underlying
		586	syscalls support them.
		587
		588	When called with a pathname, uses utimes(2) if available, otherwise
		589	utime(2). If called on a file descriptor, uses futimes(2) if available,
		590	otherwise returns ENOSYS, so this is not portable.
		591
		592	Examples:
		593
		594	# set atime and mtime to current time (basically touch(1)):
		595	aio_utime "path", undef, undef;
		596	# set atime to current time and mtime to beginning of the epoch:
		597	aio_utime "path", time, undef; # undef==0
		598
		599
		600	=item aio_chown $fh_or_path, $uid, $gid, $callback->($status)
		601
		602	Works like perl's C<chown> function, except that C<undef> for either $uid
		603	or $gid is being interpreted as "do not change" (but -1 can also be used).
		604
		605	Examples:
		606
		607	# same as "chown root path" in the shell:
		608	aio_chown "path", 0, -1;
		609	# same as above:
		610	aio_chown "path", 0, undef;
		611
		612
		613	=item aio_truncate $fh_or_path, $offset, $callback->($status)
		614
		615	Works like truncate(2) or ftruncate(2).
		616
		617
		618	=item aio_chmod $fh_or_path, $mode, $callback->($status)
		619
		620	Works like perl's C<chmod> function.
		621
		622
204	=item aio_unlink $pathname, $callback	623	=item aio_unlink $pathname, $callback->($status)
205		624
206	Asynchronously unlink (delete) a file and call the callback with the	625	Asynchronously unlink (delete) a file and call the callback with the
207	result code.	626	result code.
208		627
		628
		629	=item aio_mknod $pathname, $mode, $dev, $callback->($status)
		630
		631	[EXPERIMENTAL]
		632
		633	Asynchronously create a device node (or fifo). See mknod(2).
		634
		635	The only (POSIX-) portable way of calling this function is:
		636
		637	aio_mknod $pathname, IO::AIO::S_IFIFO | $mode, 0, sub { ...
		638
		639	See C<aio_stat> for info about some potentially helpful extra constants
		640	and functions.
		641
		642	=item aio_link $srcpath, $dstpath, $callback->($status)
		643
		644	Asynchronously create a new link to the existing object at C<$srcpath> at
		645	the path C<$dstpath> and call the callback with the result code.
		646
		647
		648	=item aio_symlink $srcpath, $dstpath, $callback->($status)
		649
		650	Asynchronously create a new symbolic link to the existing object at C<$srcpath> at
		651	the path C<$dstpath> and call the callback with the result code.
		652
		653
		654	=item aio_readlink $pathname, $callback->($link)
		655
		656	Asynchronously read the symlink specified by C<$path> and pass it to
		657	the callback. If an error occurs, nothing or undef gets passed to the
		658	callback.
		659
		660
		661	=item aio_realpath $pathname, $callback->($path)
		662
		663	Asynchronously make the path absolute and resolve any symlinks in
		664	C<$path>. The resulting path only consists of directories (Same as
		665	L<Cwd::realpath>).
		666
		667	This request can be used to get the absolute path of the current working
		668	directory by passing it a path of F<.> (a single dot).
		669
		670
		671	=item aio_rename $srcpath, $dstpath, $callback->($status)
		672
		673	Asynchronously rename the object at C<$srcpath> to C<$dstpath>, just as
		674	rename(2) and call the callback with the result code.
		675
		676
		677	=item aio_mkdir $pathname, $mode, $callback->($status)
		678
		679	Asynchronously mkdir (create) a directory and call the callback with
		680	the result code. C<$mode> will be modified by the umask at the time the
		681	request is executed, so do not change your umask.
		682
		683
		684	=item aio_rmdir $pathname, $callback->($status)
		685
		686	Asynchronously rmdir (delete) a directory and call the callback with the
		687	result code.
		688
		689
		690	=item aio_readdir $pathname, $callback->($entries)
		691
		692	Unlike the POSIX call of the same name, C<aio_readdir> reads an entire
		693	directory (i.e. opendir + readdir + closedir). The entries will not be
		694	sorted, and will B<NOT> include the C<.> and C<..> entries.
		695
		696	The callback is passed a single argument which is either C<undef> or an
		697	array-ref with the filenames.
		698
		699
		700	=item aio_readdirx $pathname, $flags, $callback->($entries, $flags)
		701
		702	Quite similar to C<aio_readdir>, but the C<$flags> argument allows one to
		703	tune behaviour and output format. In case of an error, C<$entries> will be
		704	C<undef>.
		705
		706	The flags are a combination of the following constants, ORed together (the
		707	flags will also be passed to the callback, possibly modified):
		708
		709	=over 4
		710
		711	=item IO::AIO::READDIR_DENTS
		712
		713	When this flag is off, then the callback gets an arrayref consisting of
		714	names only (as with C<aio_readdir>), otherwise it gets an arrayref with
		715	C<[$name, $type, $inode]> arrayrefs, each describing a single directory
		716	entry in more detail.
		717
		718	C<$name> is the name of the entry.
		719
		720	C<$type> is one of the C<IO::AIO::DT_xxx> constants:
		721
		722	C<IO::AIO::DT_UNKNOWN>, C<IO::AIO::DT_FIFO>, C<IO::AIO::DT_CHR>, C<IO::AIO::DT_DIR>,
		723	C<IO::AIO::DT_BLK>, C<IO::AIO::DT_REG>, C<IO::AIO::DT_LNK>, C<IO::AIO::DT_SOCK>,
		724	C<IO::AIO::DT_WHT>.
		725
		726	C<IO::AIO::DT_UNKNOWN> means just that: readdir does not know. If you need to
		727	know, you have to run stat yourself. Also, for speed reasons, the C<$type>
		728	scalars are read-only: you can not modify them.
		729
		730	C<$inode> is the inode number (which might not be exact on systems with 64
		731	bit inode numbers and 32 bit perls). This field has unspecified content on
		732	systems that do not deliver the inode information.
		733
		734	=item IO::AIO::READDIR_DIRS_FIRST
		735
		736	When this flag is set, then the names will be returned in an order where
		737	likely directories come first, in optimal stat order. This is useful when
		738	you need to quickly find directories, or you want to find all directories
		739	while avoiding to stat() each entry.
		740
		741	If the system returns type information in readdir, then this is used
		742	to find directories directly. Otherwise, likely directories are names
		743	beginning with ".", or otherwise names with no dots, of which names with
		744	short names are tried first.
		745
		746	=item IO::AIO::READDIR_STAT_ORDER
		747
		748	When this flag is set, then the names will be returned in an order
		749	suitable for stat()'ing each one. That is, when you plan to stat()
		750	all files in the given directory, then the returned order will likely
		751	be fastest.
		752
		753	If both this flag and C<IO::AIO::READDIR_DIRS_FIRST> are specified, then
		754	the likely dirs come first, resulting in a less optimal stat order.
		755
		756	=item IO::AIO::READDIR_FOUND_UNKNOWN
		757
		758	This flag should not be set when calling C<aio_readdirx>. Instead, it
		759	is being set by C<aio_readdirx>, when any of the C<$type>'s found were
		760	C<IO::AIO::DT_UNKNOWN>. The absence of this flag therefore indicates that all
		761	C<$type>'s are known, which can be used to speed up some algorithms.
		762
		763	=back
		764
		765
		766	=item aio_load $pathname, $data, $callback->($status)
		767
		768	This is a composite request that tries to fully load the given file into
		769	memory. Status is the same as with aio_read.
		770
		771	=cut
		772
		773	sub aio_load($$;$) {
		774	my ($path, undef, $cb) = @_;
		775	my $data = \$_[1];
		776
		777	my $pri = aioreq_pri;
		778	my $grp = aio_group $cb;
		779
		780	aioreq_pri $pri;
		781	add $grp aio_open $path, O_RDONLY, 0, sub {
		782	my $fh = shift
		783	or return $grp->result (-1);
		784
		785	aioreq_pri $pri;
		786	add $grp aio_read $fh, 0, (-s $fh), $$data, 0, sub {
		787	$grp->result ($_[0]);
		788	};
		789	};
		790
		791	$grp
		792	}
		793
		794	=item aio_copy $srcpath, $dstpath, $callback->($status)
		795
		796	Try to copy the I<file> (directories not supported as either source or
		797	destination) from C<$srcpath> to C<$dstpath> and call the callback with
		798	a status of C<0> (ok) or C<-1> (error, see C<$!>).
		799
		800	This is a composite request that creates the destination file with
		801	mode 0200 and copies the contents of the source file into it using
		802	C<aio_sendfile>, followed by restoring atime, mtime, access mode and
		803	uid/gid, in that order.
		804
		805	If an error occurs, the partial destination file will be unlinked, if
		806	possible, except when setting atime, mtime, access mode and uid/gid, where
		807	errors are being ignored.
		808
		809	=cut
		810
		811	sub aio_copy($$;$) {
		812	my ($src, $dst, $cb) = @_;
		813
		814	my $pri = aioreq_pri;
		815	my $grp = aio_group $cb;
		816
		817	aioreq_pri $pri;
		818	add $grp aio_open $src, O_RDONLY, 0, sub {
		819	if (my $src_fh = $_[0]) {
		820	my @stat = stat $src_fh; # hmm, might block over nfs?
		821
		822	aioreq_pri $pri;
		823	add $grp aio_open $dst, O_CREAT | O_WRONLY | O_TRUNC, 0200, sub {
		824	if (my $dst_fh = $_[0]) {
		825	aioreq_pri $pri;
		826	add $grp aio_sendfile $dst_fh, $src_fh, 0, $stat[7], sub {
		827	if ($_[0] == $stat[7]) {
		828	$grp->result (0);
		829	close $src_fh;
		830
		831	my $ch = sub {
		832	aioreq_pri $pri;
		833	add $grp aio_chmod $dst_fh, $stat[2] & 07777, sub {
		834	aioreq_pri $pri;
		835	add $grp aio_chown $dst_fh, $stat[4], $stat[5], sub {
		836	aioreq_pri $pri;
		837	add $grp aio_close $dst_fh;
		838	}
		839	};
		840	};
		841
		842	aioreq_pri $pri;
		843	add $grp aio_utime $dst_fh, $stat[8], $stat[9], sub {
		844	if ($_[0] < 0 && $! == ENOSYS) {
		845	aioreq_pri $pri;
		846	add $grp aio_utime $dst, $stat[8], $stat[9], $ch;
		847	} else {
		848	$ch->();
		849	}
		850	};
		851	} else {
		852	$grp->result (-1);
		853	close $src_fh;
		854	close $dst_fh;
		855
		856	aioreq $pri;
		857	add $grp aio_unlink $dst;
		858	}
		859	};
		860	} else {
		861	$grp->result (-1);
		862	}
		863	},
		864
		865	} else {
		866	$grp->result (-1);
		867	}
		868	};
		869
		870	$grp
		871	}
		872
		873	=item aio_move $srcpath, $dstpath, $callback->($status)
		874
		875	Try to move the I<file> (directories not supported as either source or
		876	destination) from C<$srcpath> to C<$dstpath> and call the callback with
		877	a status of C<0> (ok) or C<-1> (error, see C<$!>).
		878
		879	This is a composite request that tries to rename(2) the file first; if
		880	rename fails with C<EXDEV>, it copies the file with C<aio_copy> and, if
		881	that is successful, unlinks the C<$srcpath>.
		882
		883	=cut
		884
		885	sub aio_move($$;$) {
		886	my ($src, $dst, $cb) = @_;
		887
		888	my $pri = aioreq_pri;
		889	my $grp = aio_group $cb;
		890
		891	aioreq_pri $pri;
		892	add $grp aio_rename $src, $dst, sub {
		893	if ($_[0] && $! == EXDEV) {
		894	aioreq_pri $pri;
		895	add $grp aio_copy $src, $dst, sub {
		896	$grp->result ($_[0]);
		897
		898	unless ($_[0]) {
		899	aioreq_pri $pri;
		900	add $grp aio_unlink $src;
		901	}
		902	};
		903	} else {
		904	$grp->result ($_[0]);
		905	}
		906	};
		907
		908	$grp
		909	}
		910
		911	=item aio_scandir $pathname, $maxreq, $callback->($dirs, $nondirs)
		912
		913	Scans a directory (similar to C<aio_readdir>) but additionally tries to
		914	efficiently separate the entries of directory C<$path> into two sets of
		915	names, directories you can recurse into (directories), and ones you cannot
		916	recurse into (everything else, including symlinks to directories).
		917
		918	C<aio_scandir> is a composite request that creates of many sub requests_
		919	C<$maxreq> specifies the maximum number of outstanding aio requests that
		920	this function generates. If it is C<< <= 0 >>, then a suitable default
		921	will be chosen (currently 4).
		922
		923	On error, the callback is called without arguments, otherwise it receives
		924	two array-refs with path-relative entry names.
		925
		926	Example:
		927
		928	aio_scandir $dir, 0, sub {
		929	my ($dirs, $nondirs) = @_;
		930	print "real directories: @$dirs\n";
		931	print "everything else: @$nondirs\n";
		932	};
		933
		934	Implementation notes.
		935
		936	The C<aio_readdir> cannot be avoided, but C<stat()>'ing every entry can.
		937
		938	If readdir returns file type information, then this is used directly to
		939	find directories.
		940
		941	Otherwise, after reading the directory, the modification time, size etc.
		942	of the directory before and after the readdir is checked, and if they
		943	match (and isn't the current time), the link count will be used to decide
		944	how many entries are directories (if >= 2). Otherwise, no knowledge of the
		945	number of subdirectories will be assumed.
		946
		947	Then entries will be sorted into likely directories a non-initial dot
		948	currently) and likely non-directories (see C<aio_readdirx>). Then every
		949	entry plus an appended C</.> will be C<stat>'ed, likely directories first,
		950	in order of their inode numbers. If that succeeds, it assumes that the
		951	entry is a directory or a symlink to directory (which will be checked
		952	separately). This is often faster than stat'ing the entry itself because
		953	filesystems might detect the type of the entry without reading the inode
		954	data (e.g. ext2fs filetype feature), even on systems that cannot return
		955	the filetype information on readdir.
		956
		957	If the known number of directories (link count - 2) has been reached, the
		958	rest of the entries is assumed to be non-directories.
		959
		960	This only works with certainty on POSIX (= UNIX) filesystems, which
		961	fortunately are the vast majority of filesystems around.
		962
		963	It will also likely work on non-POSIX filesystems with reduced efficiency
		964	as those tend to return 0 or 1 as link counts, which disables the
		965	directory counting heuristic.
		966
		967	=cut
		968
		969	sub aio_scandir($$;$) {
		970	my ($path, $maxreq, $cb) = @_;
		971
		972	my $pri = aioreq_pri;
		973
		974	my $grp = aio_group $cb;
		975
		976	$maxreq = 4 if $maxreq <= 0;
		977
		978	# get a wd object
		979	aioreq_pri $pri;
		980	add $grp aio_wd $path, sub {
		981	$_[0]
		982	or return $grp->result ();
		983
		984	my $wd = [shift, "."];
		985
		986	# stat once
		987	aioreq_pri $pri;
		988	add $grp aio_stat $wd, sub {
		989	return $grp->result () if $_[0];
		990	my $now = time;
		991	my $hash1 = join ":", (stat _)[0,1,3,7,9];
		992
		993	# read the directory entries
		994	aioreq_pri $pri;
		995	add $grp aio_readdirx $wd, READDIR_DIRS_FIRST, sub {
		996	my $entries = shift
		997	or return $grp->result ();
		998
		999	# stat the dir another time
		1000	aioreq_pri $pri;
		1001	add $grp aio_stat $wd, sub {
		1002	my $hash2 = join ":", (stat _)[0,1,3,7,9];
		1003
		1004	my $ndirs;
		1005
		1006	# take the slow route if anything looks fishy
		1007	if ($hash1 ne $hash2 or (stat _)[9] == $now) {
		1008	$ndirs = -1;
		1009	} else {
		1010	# if nlink == 2, we are finished
		1011	# for non-posix-fs's, we rely on nlink < 2
		1012	$ndirs = (stat _)[3] - 2
		1013	or return $grp->result ([], $entries);
		1014	}
		1015
		1016	my (@dirs, @nondirs);
		1017
		1018	my $statgrp = add $grp aio_group sub {
		1019	$grp->result (\@dirs, \@nondirs);
		1020	};
		1021
		1022	limit $statgrp $maxreq;
		1023	feed $statgrp sub {
		1024	return unless @$entries;
		1025	my $entry = shift @$entries;
		1026
		1027	aioreq_pri $pri;
		1028	$wd->[1] = "$entry/.";
		1029	add $statgrp aio_stat $wd, sub {
		1030	if ($_[0] < 0) {
		1031	push @nondirs, $entry;
		1032	} else {
		1033	# need to check for real directory
		1034	aioreq_pri $pri;
		1035	$wd->[1] = $entry;
		1036	add $statgrp aio_lstat $wd, sub {
		1037	if (-d _) {
		1038	push @dirs, $entry;
		1039
		1040	unless (--$ndirs) {
		1041	push @nondirs, @$entries;
		1042	feed $statgrp;
		1043	}
		1044	} else {
		1045	push @nondirs, $entry;
		1046	}
		1047	}
		1048	}
		1049	};
		1050	};
		1051	};
		1052	};
		1053	};
		1054	};
		1055
		1056	$grp
		1057	}
		1058
		1059	=item aio_rmtree $pathname, $callback->($status)
		1060
		1061	Delete a directory tree starting (and including) C<$path>, return the
		1062	status of the final C<rmdir> only. This is a composite request that
		1063	uses C<aio_scandir> to recurse into and rmdir directories, and unlink
		1064	everything else.
		1065
		1066	=cut
		1067
		1068	sub aio_rmtree;
		1069	sub aio_rmtree($;$) {
		1070	my ($path, $cb) = @_;
		1071
		1072	my $pri = aioreq_pri;
		1073	my $grp = aio_group $cb;
		1074
		1075	aioreq_pri $pri;
		1076	add $grp aio_scandir $path, 0, sub {
		1077	my ($dirs, $nondirs) = @_;
		1078
		1079	my $dirgrp = aio_group sub {
		1080	add $grp aio_rmdir $path, sub {
		1081	$grp->result ($_[0]);
		1082	};
		1083	};
		1084
		1085	(aioreq_pri $pri), add $dirgrp aio_rmtree "$path/$_" for @$dirs;
		1086	(aioreq_pri $pri), add $dirgrp aio_unlink "$path/$_" for @$nondirs;
		1087
		1088	add $grp $dirgrp;
		1089	};
		1090
		1091	$grp
		1092	}
		1093
		1094	=item aio_sync $callback->($status)
		1095
		1096	Asynchronously call sync and call the callback when finished.
		1097
209	=item aio_fsync $fh, $callback	1098	=item aio_fsync $fh, $callback->($status)
210		1099
211	Asynchronously call fsync on the given filehandle and call the callback	1100	Asynchronously call fsync on the given filehandle and call the callback
212	with the fsync result code.	1101	with the fsync result code.
213		1102
214	=item aio_fdatasync $fh, $callback	1103	=item aio_fdatasync $fh, $callback->($status)
215		1104
216	Asynchronously call fdatasync on the given filehandle and call the	1105	Asynchronously call fdatasync on the given filehandle and call the
217	callback with the fdatasync result code.	1106	callback with the fdatasync result code.
218		1107
		1108	If this call isn't available because your OS lacks it or it couldn't be
		1109	detected, it will be emulated by calling C<fsync> instead.
		1110
		1111	=item aio_syncfs $fh, $callback->($status)
		1112
		1113	Asynchronously call the syncfs syscall to sync the filesystem associated
		1114	to the given filehandle and call the callback with the syncfs result
		1115	code. If syncfs is not available, calls sync(), but returns C<-1> and sets
		1116	errno to C<ENOSYS> nevertheless.
		1117
		1118	=item aio_sync_file_range $fh, $offset, $nbytes, $flags, $callback->($status)
		1119
		1120	Sync the data portion of the file specified by C<$offset> and C<$length>
		1121	to disk (but NOT the metadata), by calling the Linux-specific
		1122	sync_file_range call. If sync_file_range is not available or it returns
		1123	ENOSYS, then fdatasync or fsync is being substituted.
		1124
		1125	C<$flags> can be a combination of C<IO::AIO::SYNC_FILE_RANGE_WAIT_BEFORE>,
		1126	C<IO::AIO::SYNC_FILE_RANGE_WRITE> and
		1127	C<IO::AIO::SYNC_FILE_RANGE_WAIT_AFTER>: refer to the sync_file_range
		1128	manpage for details.
		1129
		1130	=item aio_pathsync $pathname, $callback->($status)
		1131
		1132	This request tries to open, fsync and close the given path. This is a
		1133	composite request intended to sync directories after directory operations
		1134	(E.g. rename). This might not work on all operating systems or have any
		1135	specific effect, but usually it makes sure that directory changes get
		1136	written to disc. It works for anything that can be opened for read-only,
		1137	not just directories.
		1138
		1139	Future versions of this function might fall back to other methods when
		1140	C<fsync> on the directory fails (such as calling C<sync>).
		1141
		1142	Passes C<0> when everything went ok, and C<-1> on error.
		1143
219	=cut	1144	=cut
220		1145
221	min_parallel 4;	1146	sub aio_pathsync($;$) {
		1147	my ($path, $cb) = @_;
222		1148
223	END {	1149	my $pri = aioreq_pri;
224	max_parallel 0;	1150	my $grp = aio_group $cb;
		1151
		1152	aioreq_pri $pri;
		1153	add $grp aio_open $path, O_RDONLY, 0, sub {
		1154	my ($fh) = @_;
		1155	if ($fh) {
		1156	aioreq_pri $pri;
		1157	add $grp aio_fsync $fh, sub {
		1158	$grp->result ($_[0]);
		1159
		1160	aioreq_pri $pri;
		1161	add $grp aio_close $fh;
		1162	};
		1163	} else {
		1164	$grp->result (-1);
		1165	}
		1166	};
		1167
		1168	$grp
225	}	1169	}
226		1170
		1171	=item aio_msync $scalar, $offset = 0, $length = undef, flags = 0, $callback->($status)
		1172
		1173	This is a rather advanced IO::AIO call, which only works on mmap(2)ed
		1174	scalars (see the C<IO::AIO::mmap> function, although it also works on data
		1175	scalars managed by the L<Sys::Mmap> or L<Mmap> modules, note that the
		1176	scalar must only be modified in-place while an aio operation is pending on
		1177	it).
		1178
		1179	It calls the C<msync> function of your OS, if available, with the memory
		1180	area starting at C<$offset> in the string and ending C<$length> bytes
		1181	later. If C<$length> is negative, counts from the end, and if C<$length>
		1182	is C<undef>, then it goes till the end of the string. The flags can be
		1183	a combination of C<IO::AIO::MS_ASYNC>, C<IO::AIO::MS_INVALIDATE> and
		1184	C<IO::AIO::MS_SYNC>.
		1185
		1186	=item aio_mtouch $scalar, $offset = 0, $length = undef, flags = 0, $callback->($status)
		1187
		1188	This is a rather advanced IO::AIO call, which works best on mmap(2)ed
		1189	scalars.
		1190
		1191	It touches (reads or writes) all memory pages in the specified
		1192	range inside the scalar. All caveats and parameters are the same
		1193	as for C<aio_msync>, above, except for flags, which must be either
		1194	C<0> (which reads all pages and ensures they are instantiated) or
		1195	C<IO::AIO::MT_MODIFY>, which modifies the memory page s(by reading and
		1196	writing an octet from it, which dirties the page).
		1197
		1198	=item aio_mlock $scalar, $offset = 0, $length = undef, $callback->($status)
		1199
		1200	This is a rather advanced IO::AIO call, which works best on mmap(2)ed
		1201	scalars.
		1202
		1203	It reads in all the pages of the underlying storage into memory (if any)
		1204	and locks them, so they are not getting swapped/paged out or removed.
		1205
		1206	If C<$length> is undefined, then the scalar will be locked till the end.
		1207
		1208	On systems that do not implement C<mlock>, this function returns C<-1>
		1209	and sets errno to C<ENOSYS>.
		1210
		1211	Note that the corresponding C<munlock> is synchronous and is
		1212	documented under L<MISCELLANEOUS FUNCTIONS>.
		1213
		1214	Example: open a file, mmap and mlock it - both will be undone when
		1215	C<$data> gets destroyed.
		1216
		1217	open my $fh, "<", $path or die "$path: $!";
		1218	my $data;
		1219	IO::AIO::mmap $data, -s $fh, IO::AIO::PROT_READ, IO::AIO::MAP_SHARED, $fh;
		1220	aio_mlock $data; # mlock in background
		1221
		1222	=item aio_mlockall $flags, $callback->($status)
		1223
		1224	Calls the C<mlockall> function with the given C<$flags> (a combination of
		1225	C<IO::AIO::MCL_CURRENT> and C<IO::AIO::MCL_FUTURE>).
		1226
		1227	On systems that do not implement C<mlockall>, this function returns C<-1>
		1228	and sets errno to C<ENOSYS>.
		1229
		1230	Note that the corresponding C<munlockall> is synchronous and is
		1231	documented under L<MISCELLANEOUS FUNCTIONS>.
		1232
		1233	Example: asynchronously lock all current and future pages into memory.
		1234
		1235	aio_mlockall IO::AIO::MCL_FUTURE;
		1236
		1237	=item aio_group $callback->(...)
		1238
		1239	This is a very special aio request: Instead of doing something, it is a
		1240	container for other aio requests, which is useful if you want to bundle
		1241	many requests into a single, composite, request with a definite callback
		1242	and the ability to cancel the whole request with its subrequests.
		1243
		1244	Returns an object of class L<IO::AIO::GRP>. See its documentation below
		1245	for more info.
		1246
		1247	Example:
		1248
		1249	my $grp = aio_group sub {
		1250	print "all stats done\n";
		1251	};
		1252
		1253	add $grp
		1254	(aio_stat ...),
		1255	(aio_stat ...),
		1256	...;
		1257
		1258	=item aio_nop $callback->()
		1259
		1260	This is a special request - it does nothing in itself and is only used for
		1261	side effects, such as when you want to add a dummy request to a group so
		1262	that finishing the requests in the group depends on executing the given
		1263	code.
		1264
		1265	While this request does nothing, it still goes through the execution
		1266	phase and still requires a worker thread. Thus, the callback will not
		1267	be executed immediately but only after other requests in the queue have
		1268	entered their execution phase. This can be used to measure request
		1269	latency.
		1270
		1271	=item IO::AIO::aio_busy $fractional_seconds, $callback->() *NOT EXPORTED*
		1272
		1273	Mainly used for debugging and benchmarking, this aio request puts one of
		1274	the request workers to sleep for the given time.
		1275
		1276	While it is theoretically handy to have simple I/O scheduling requests
		1277	like sleep and file handle readable/writable, the overhead this creates is
		1278	immense (it blocks a thread for a long time) so do not use this function
		1279	except to put your application under artificial I/O pressure.
		1280
		1281	=back
		1282
		1283
		1284	=head2 IO::AIO::WD - multiple working directories
		1285
		1286	Your process only has one current working directory, which is used by all
		1287	threads. This makes it hard to use relative paths (some other component
		1288	could call C<chdir> at any time, and it is hard to control when the path
		1289	will be used by IO::AIO).
		1290
		1291	One solution for this is to always use absolute paths. This usually works,
		1292	but can be quite slow (the kernel has to walk the whole path on every
		1293	access), and can also be a hassle to implement.
		1294
		1295	Newer POSIX systems have a number of functions (openat, fdopendir,
		1296	futimensat and so on) that make it possible to specify working directories
		1297	per operation.
		1298
		1299	For portability, and because the clowns who "designed", or shall I write,
		1300	perpetrated this new interface were obviously half-drunk, this abstraction
		1301	cannot be perfect, though.
		1302
		1303	IO::AIO allows you to convert directory paths into a so-called IO::AIO::WD
		1304	object. This object stores the canonicalised, absolute version of the
		1305	path, and on systems that allow it, also a directory file descriptor.
		1306
		1307	Everywhere where a pathname is accepted by IO::AIO (e.g. in C<aio_stat>
		1308	or C<aio_unlink>), one can specify an array reference with an IO::AIO::WD
		1309	object and a pathname instead. If the pathname is absolute, the
		1310	IO::AIO::WD objetc is ignored, otherwise the pathname is resolved relative
		1311	to that IO::AIO::WD object.
		1312
		1313	For example, to get a wd object for F</etc> and then stat F<passwd>
		1314	inside, you would write:
		1315
		1316	aio_wd "/etc", sub {
		1317	my $etcdir = shift;
		1318
		1319	# although $etcdir can be undef on error, there is generally no reason
		1320	# to check for errors here, as aio_stat will fail with ENOENT
		1321	# when $etcdir is undef.
		1322
		1323	aio_stat [$etcdir, "passwd"], sub {
		1324	# yay
		1325	};
		1326	};
		1327
		1328	This shows that creating an IO::AIO::WD object is itself a potentially
		1329	blocking operation, which is why it is done asynchronously.
		1330
		1331	As with normal pathnames, IO::AIO keeps a copy of the working directory
		1332	object and the pathname string, so you could write the following without
		1333	causing any issues due to C<$path> getting reused:
		1334
		1335	my $path = [$wd, undef];
		1336
		1337	for my $name (qw(abc def ghi)) {
		1338	$path->[1] = $name;
		1339	aio_stat $path, sub {
		1340	# ...
		1341	};
		1342	}
		1343
		1344	There are some caveats: when directories get renamed (or deleted), the
		1345	pathname string doesn't change, so will point to the new directory (or
		1346	nowhere at all), while the directory fd, if available on the system,
		1347	will still point to the original directory. Most functions accepting a
		1348	pathname will use the directory fd on newer systems, and the string on
		1349	older systems. Some functions (such as realpath) will always rely on the
		1350	string form of the pathname.
		1351
		1352	So this fucntionality is mainly useful to get some protection against
		1353	C<chdir>, to easily get an absolute path out of a relative path for future
		1354	reference, and to speed up doing many operations in the same directory
		1355	(e.g. when stat'ing all files in a directory).
		1356
		1357	The following functions implement this working directory abstraction:
		1358
		1359	=over 4
		1360
		1361	=item aio_wd $pathname, $callback->($wd)
		1362
		1363	Asynchonously canonicalise the given pathname and convert it to an
		1364	IO::AIO::WD object representing it. If possible and supported on the
		1365	system, also open a directory fd to speed up pathname resolution relative
		1366	to this working directory.
		1367
		1368	If something goes wrong, then C<undef> is passwd to the callback instead
		1369	of a working directory object and C<$!> is set appropriately. Since
		1370	passing C<undef> as working directory component of a pathname fails the
		1371	request with C<ENOENT>, there is often no need for error checking in the
		1372	C<aio_wd> callback, as future requests using the value will fail in the
		1373	expected way.
		1374
		1375	If this call isn't available because your OS lacks it or it couldn't be
		1376	detected, it will be emulated by calling C<fsync> instead.
		1377
		1378	=item IO::AIO::CWD
		1379
		1380	This is a compiletime constant (object) that represents the process
		1381	current working directory.
		1382
		1383	Specifying this object as working directory object for a pathname is as
		1384	if the pathname would be specified directly, without a directory object,
		1385	e.g., these calls are functionally identical:
		1386
		1387	aio_stat "somefile", sub { ... };
		1388	aio_stat [IO::AIO::CWD, "somefile"], sub { ... };
		1389
		1390	=back
		1391
		1392
		1393	=head2 IO::AIO::REQ CLASS
		1394
		1395	All non-aggregate C<aio_*> functions return an object of this class when
		1396	called in non-void context.
		1397
		1398	=over 4
		1399
		1400	=item cancel $req
		1401
		1402	Cancels the request, if possible. Has the effect of skipping execution
		1403	when entering the B<execute> state and skipping calling the callback when
		1404	entering the the B<result> state, but will leave the request otherwise
		1405	untouched (with the exception of readdir). That means that requests that
		1406	currently execute will not be stopped and resources held by the request
		1407	will not be freed prematurely.
		1408
		1409	=item cb $req $callback->(...)
		1410
		1411	Replace (or simply set) the callback registered to the request.
		1412
		1413	=back
		1414
		1415	=head2 IO::AIO::GRP CLASS
		1416
		1417	This class is a subclass of L<IO::AIO::REQ>, so all its methods apply to
		1418	objects of this class, too.
		1419
		1420	A IO::AIO::GRP object is a special request that can contain multiple other
		1421	aio requests.
		1422
		1423	You create one by calling the C<aio_group> constructing function with a
		1424	callback that will be called when all contained requests have entered the
		1425	C<done> state:
		1426
		1427	my $grp = aio_group sub {
		1428	print "all requests are done\n";
		1429	};
		1430
		1431	You add requests by calling the C<add> method with one or more
		1432	C<IO::AIO::REQ> objects:
		1433
		1434	$grp->add (aio_unlink "...");
		1435
		1436	add $grp aio_stat "...", sub {
		1437	$_[0] or return $grp->result ("error");
		1438
		1439	# add another request dynamically, if first succeeded
		1440	add $grp aio_open "...", sub {
		1441	$grp->result ("ok");
		1442	};
		1443	};
		1444
		1445	This makes it very easy to create composite requests (see the source of
		1446	C<aio_move> for an application) that work and feel like simple requests.
		1447
		1448	=over 4
		1449
		1450	=item * The IO::AIO::GRP objects will be cleaned up during calls to
		1451	C<IO::AIO::poll_cb>, just like any other request.
		1452
		1453	=item * They can be canceled like any other request. Canceling will cancel not
		1454	only the request itself, but also all requests it contains.
		1455
		1456	=item * They can also can also be added to other IO::AIO::GRP objects.
		1457
		1458	=item * You must not add requests to a group from within the group callback (or
		1459	any later time).
		1460
		1461	=back
		1462
		1463	Their lifetime, simplified, looks like this: when they are empty, they
		1464	will finish very quickly. If they contain only requests that are in the
		1465	C<done> state, they will also finish. Otherwise they will continue to
		1466	exist.
		1467
		1468	That means after creating a group you have some time to add requests
		1469	(precisely before the callback has been invoked, which is only done within
		1470	the C<poll_cb>). And in the callbacks of those requests, you can add
		1471	further requests to the group. And only when all those requests have
		1472	finished will the the group itself finish.
		1473
		1474	=over 4
		1475
		1476	=item add $grp ...
		1477
		1478	=item $grp->add (...)
		1479
		1480	Add one or more requests to the group. Any type of L<IO::AIO::REQ> can
		1481	be added, including other groups, as long as you do not create circular
		1482	dependencies.
		1483
		1484	Returns all its arguments.
		1485
		1486	=item $grp->cancel_subs
		1487
		1488	Cancel all subrequests and clears any feeder, but not the group request
		1489	itself. Useful when you queued a lot of events but got a result early.
		1490
		1491	The group request will finish normally (you cannot add requests to the
		1492	group).
		1493
		1494	=item $grp->result (...)
		1495
		1496	Set the result value(s) that will be passed to the group callback when all
		1497	subrequests have finished and set the groups errno to the current value
		1498	of errno (just like calling C<errno> without an error number). By default,
		1499	no argument will be passed and errno is zero.
		1500
		1501	=item $grp->errno ([$errno])
		1502
		1503	Sets the group errno value to C<$errno>, or the current value of errno
		1504	when the argument is missing.
		1505
		1506	Every aio request has an associated errno value that is restored when
		1507	the callback is invoked. This method lets you change this value from its
		1508	default (0).
		1509
		1510	Calling C<result> will also set errno, so make sure you either set C<$!>
		1511	before the call to C<result>, or call c<errno> after it.
		1512
		1513	=item feed $grp $callback->($grp)
		1514
		1515	Sets a feeder/generator on this group: every group can have an attached
		1516	generator that generates requests if idle. The idea behind this is that,
		1517	although you could just queue as many requests as you want in a group,
		1518	this might starve other requests for a potentially long time. For example,
		1519	C<aio_scandir> might generate hundreds of thousands of C<aio_stat>
		1520	requests, delaying any later requests for a long time.
		1521
		1522	To avoid this, and allow incremental generation of requests, you can
		1523	instead a group and set a feeder on it that generates those requests. The
		1524	feed callback will be called whenever there are few enough (see C<limit>,
		1525	below) requests active in the group itself and is expected to queue more
		1526	requests.
		1527
		1528	The feed callback can queue as many requests as it likes (i.e. C<add> does
		1529	not impose any limits).
		1530
		1531	If the feed does not queue more requests when called, it will be
		1532	automatically removed from the group.
		1533
		1534	If the feed limit is C<0> when this method is called, it will be set to
		1535	C<2> automatically.
		1536
		1537	Example:
		1538
		1539	# stat all files in @files, but only ever use four aio requests concurrently:
		1540
		1541	my $grp = aio_group sub { print "finished\n" };
		1542	limit $grp 4;
		1543	feed $grp sub {
		1544	my $file = pop @files
		1545	or return;
		1546
		1547	add $grp aio_stat $file, sub { ... };
		1548	};
		1549
		1550	=item limit $grp $num
		1551
		1552	Sets the feeder limit for the group: The feeder will be called whenever
		1553	the group contains less than this many requests.
		1554
		1555	Setting the limit to C<0> will pause the feeding process.
		1556
		1557	The default value for the limit is C<0>, but note that setting a feeder
		1558	automatically bumps it up to C<2>.
		1559
		1560	=back
		1561
		1562	=head2 SUPPORT FUNCTIONS
		1563
		1564	=head3 EVENT PROCESSING AND EVENT LOOP INTEGRATION
		1565
		1566	=over 4
		1567
		1568	=item $fileno = IO::AIO::poll_fileno
		1569
		1570	Return the I<request result pipe file descriptor>. This filehandle must be
		1571	polled for reading by some mechanism outside this module (e.g. EV, Glib,
		1572	select and so on, see below or the SYNOPSIS). If the pipe becomes readable
		1573	you have to call C<poll_cb> to check the results.
		1574
		1575	See C<poll_cb> for an example.
		1576
		1577	=item IO::AIO::poll_cb
		1578
		1579	Process some outstanding events on the result pipe. You have to call
		1580	this regularly. Returns C<0> if all events could be processed (or there
		1581	were no events to process), or C<-1> if it returned earlier for whatever
		1582	reason. Returns immediately when no events are outstanding. The amount of
		1583	events processed depends on the settings of C<IO::AIO::max_poll_req> and
		1584	C<IO::AIO::max_poll_time>.
		1585
		1586	If not all requests were processed for whatever reason, the filehandle
		1587	will still be ready when C<poll_cb> returns, so normally you don't have to
		1588	do anything special to have it called later.
		1589
		1590	Apart from calling C<IO::AIO::poll_cb> when the event filehandle becomes
		1591	ready, it can be beneficial to call this function from loops which submit
		1592	a lot of requests, to make sure the results get processed when they become
		1593	available and not just when the loop is finished and the event loop takes
		1594	over again. This function returns very fast when there are no outstanding
		1595	requests.
		1596
		1597	Example: Install an Event watcher that automatically calls
		1598	IO::AIO::poll_cb with high priority (more examples can be found in the
		1599	SYNOPSIS section, at the top of this document):
		1600
		1601	Event->io (fd => IO::AIO::poll_fileno,
		1602	poll => 'r', async => 1,
		1603	cb => \&IO::AIO::poll_cb);
		1604
		1605	=item IO::AIO::poll_wait
		1606
		1607	If there are any outstanding requests and none of them in the result
		1608	phase, wait till the result filehandle becomes ready for reading (simply
		1609	does a C<select> on the filehandle. This is useful if you want to
		1610	synchronously wait for some requests to finish).
		1611
		1612	See C<nreqs> for an example.
		1613
		1614	=item IO::AIO::poll
		1615
		1616	Waits until some requests have been handled.
		1617
		1618	Returns the number of requests processed, but is otherwise strictly
		1619	equivalent to:
		1620
		1621	IO::AIO::poll_wait, IO::AIO::poll_cb
		1622
		1623	=item IO::AIO::flush
		1624
		1625	Wait till all outstanding AIO requests have been handled.
		1626
		1627	Strictly equivalent to:
		1628
		1629	IO::AIO::poll_wait, IO::AIO::poll_cb
		1630	while IO::AIO::nreqs;
		1631
		1632	=item IO::AIO::max_poll_reqs $nreqs
		1633
		1634	=item IO::AIO::max_poll_time $seconds
		1635
		1636	These set the maximum number of requests (default C<0>, meaning infinity)
		1637	that are being processed by C<IO::AIO::poll_cb> in one call, respectively
		1638	the maximum amount of time (default C<0>, meaning infinity) spent in
		1639	C<IO::AIO::poll_cb> to process requests (more correctly the mininum amount
		1640	of time C<poll_cb> is allowed to use).
		1641
		1642	Setting C<max_poll_time> to a non-zero value creates an overhead of one
		1643	syscall per request processed, which is not normally a problem unless your
		1644	callbacks are really really fast or your OS is really really slow (I am
		1645	not mentioning Solaris here). Using C<max_poll_reqs> incurs no overhead.
		1646
		1647	Setting these is useful if you want to ensure some level of
		1648	interactiveness when perl is not fast enough to process all requests in
		1649	time.
		1650
		1651	For interactive programs, values such as C<0.01> to C<0.1> should be fine.
		1652
		1653	Example: Install an Event watcher that automatically calls
		1654	IO::AIO::poll_cb with low priority, to ensure that other parts of the
		1655	program get the CPU sometimes even under high AIO load.
		1656
		1657	# try not to spend much more than 0.1s in poll_cb
		1658	IO::AIO::max_poll_time 0.1;
		1659
		1660	# use a low priority so other tasks have priority
		1661	Event->io (fd => IO::AIO::poll_fileno,
		1662	poll => 'r', nice => 1,
		1663	cb => &IO::AIO::poll_cb);
		1664
		1665	=back
		1666
		1667	=head3 CONTROLLING THE NUMBER OF THREADS
		1668
		1669	=over
		1670
		1671	=item IO::AIO::min_parallel $nthreads
		1672
		1673	Set the minimum number of AIO threads to C<$nthreads>. The current
		1674	default is C<8>, which means eight asynchronous operations can execute
		1675	concurrently at any one time (the number of outstanding requests,
		1676	however, is unlimited).
		1677
		1678	IO::AIO starts threads only on demand, when an AIO request is queued and
		1679	no free thread exists. Please note that queueing up a hundred requests can
		1680	create demand for a hundred threads, even if it turns out that everything
		1681	is in the cache and could have been processed faster by a single thread.
		1682
		1683	It is recommended to keep the number of threads relatively low, as some
		1684	Linux kernel versions will scale negatively with the number of threads
		1685	(higher parallelity => MUCH higher latency). With current Linux 2.6
		1686	versions, 4-32 threads should be fine.
		1687
		1688	Under most circumstances you don't need to call this function, as the
		1689	module selects a default that is suitable for low to moderate load.
		1690
		1691	=item IO::AIO::max_parallel $nthreads
		1692
		1693	Sets the maximum number of AIO threads to C<$nthreads>. If more than the
		1694	specified number of threads are currently running, this function kills
		1695	them. This function blocks until the limit is reached.
		1696
		1697	While C<$nthreads> are zero, aio requests get queued but not executed
		1698	until the number of threads has been increased again.
		1699
		1700	This module automatically runs C<max_parallel 0> at program end, to ensure
		1701	that all threads are killed and that there are no outstanding requests.
		1702
		1703	Under normal circumstances you don't need to call this function.
		1704
		1705	=item IO::AIO::max_idle $nthreads
		1706
		1707	Limit the number of threads (default: 4) that are allowed to idle
		1708	(i.e., threads that did not get a request to process within the idle
		1709	timeout (default: 10 seconds). That means if a thread becomes idle while
		1710	C<$nthreads> other threads are also idle, it will free its resources and
		1711	exit.
		1712
		1713	This is useful when you allow a large number of threads (e.g. 100 or 1000)
		1714	to allow for extremely high load situations, but want to free resources
		1715	under normal circumstances (1000 threads can easily consume 30MB of RAM).
		1716
		1717	The default is probably ok in most situations, especially if thread
		1718	creation is fast. If thread creation is very slow on your system you might
		1719	want to use larger values.
		1720
		1721	=item IO::AIO::idle_timeout $seconds
		1722
		1723	Sets the minimum idle timeout (default 10) after which worker threads are
		1724	allowed to exit. SEe C<IO::AIO::max_idle>.
		1725
		1726	=item IO::AIO::max_outstanding $maxreqs
		1727
		1728	Sets the maximum number of outstanding requests to C<$nreqs>. If
		1729	you do queue up more than this number of requests, the next call to
		1730	C<IO::AIO::poll_cb> (and other functions calling C<poll_cb>, such as
		1731	C<IO::AIO::flush> or C<IO::AIO::poll>) will block until the limit is no
		1732	longer exceeded.
		1733
		1734	In other words, this setting does not enforce a queue limit, but can be
		1735	used to make poll functions block if the limit is exceeded.
		1736
		1737	This is a very bad function to use in interactive programs because it
		1738	blocks, and a bad way to reduce concurrency because it is inexact: Better
		1739	use an C<aio_group> together with a feed callback.
		1740
		1741	It's main use is in scripts without an event loop - when you want to stat
		1742	a lot of files, you can write somehting like this:
		1743
		1744	IO::AIO::max_outstanding 32;
		1745
		1746	for my $path (...) {
		1747	aio_stat $path , ...;
		1748	IO::AIO::poll_cb;
		1749	}
		1750
		1751	IO::AIO::flush;
		1752
		1753	The call to C<poll_cb> inside the loop will normally return instantly, but
		1754	as soon as more thna C<32> reqeusts are in-flight, it will block until
		1755	some requests have been handled. This keeps the loop from pushing a large
		1756	number of C<aio_stat> requests onto the queue.
		1757
		1758	The default value for C<max_outstanding> is very large, so there is no
		1759	practical limit on the number of outstanding requests.
		1760
		1761	=back
		1762
		1763	=head3 STATISTICAL INFORMATION
		1764
		1765	=over
		1766
		1767	=item IO::AIO::nreqs
		1768
		1769	Returns the number of requests currently in the ready, execute or pending
		1770	states (i.e. for which their callback has not been invoked yet).
		1771
		1772	Example: wait till there are no outstanding requests anymore:
		1773
		1774	IO::AIO::poll_wait, IO::AIO::poll_cb
		1775	while IO::AIO::nreqs;
		1776
		1777	=item IO::AIO::nready
		1778
		1779	Returns the number of requests currently in the ready state (not yet
		1780	executed).
		1781
		1782	=item IO::AIO::npending
		1783
		1784	Returns the number of requests currently in the pending state (executed,
		1785	but not yet processed by poll_cb).
		1786
		1787	=back
		1788
		1789	=head3 MISCELLANEOUS FUNCTIONS
		1790
		1791	IO::AIO implements some functions that might be useful, but are not
		1792	asynchronous.
		1793
		1794	=over 4
		1795
		1796	=item IO::AIO::sendfile $ofh, $ifh, $offset, $count
		1797
		1798	Calls the C<eio_sendfile_sync> function, which is like C<aio_sendfile>,
		1799	but is blocking (this makes most sense if you know the input data is
		1800	likely cached already and the output filehandle is set to non-blocking
		1801	operations).
		1802
		1803	Returns the number of bytes copied, or C<-1> on error.
		1804
		1805	=item IO::AIO::fadvise $fh, $offset, $len, $advice
		1806
		1807	Simply calls the C<posix_fadvise> function (see its
		1808	manpage for details). The following advice constants are
		1809	available: C<IO::AIO::FADV_NORMAL>, C<IO::AIO::FADV_SEQUENTIAL>,
		1810	C<IO::AIO::FADV_RANDOM>, C<IO::AIO::FADV_NOREUSE>,
		1811	C<IO::AIO::FADV_WILLNEED>, C<IO::AIO::FADV_DONTNEED>.
		1812
		1813	On systems that do not implement C<posix_fadvise>, this function returns
		1814	ENOSYS, otherwise the return value of C<posix_fadvise>.
		1815
		1816	=item IO::AIO::madvise $scalar, $offset, $len, $advice
		1817
		1818	Simply calls the C<posix_madvise> function (see its
		1819	manpage for details). The following advice constants are
		1820	available: C<IO::AIO::MADV_NORMAL>, C<IO::AIO::MADV_SEQUENTIAL>,
		1821	C<IO::AIO::MADV_RANDOM>, C<IO::AIO::MADV_WILLNEED>, C<IO::AIO::MADV_DONTNEED>.
		1822
		1823	On systems that do not implement C<posix_madvise>, this function returns
		1824	ENOSYS, otherwise the return value of C<posix_madvise>.
		1825
		1826	=item IO::AIO::mprotect $scalar, $offset, $len, $protect
		1827
		1828	Simply calls the C<mprotect> function on the preferably AIO::mmap'ed
		1829	$scalar (see its manpage for details). The following protect
		1830	constants are available: C<IO::AIO::PROT_NONE>, C<IO::AIO::PROT_READ>,
		1831	C<IO::AIO::PROT_WRITE>, C<IO::AIO::PROT_EXEC>.
		1832
		1833	On systems that do not implement C<mprotect>, this function returns
		1834	ENOSYS, otherwise the return value of C<mprotect>.
		1835
		1836	=item IO::AIO::mmap $scalar, $length, $prot, $flags, $fh[, $offset]
		1837
		1838	Memory-maps a file (or anonymous memory range) and attaches it to the
		1839	given C<$scalar>, which will act like a string scalar.
		1840
		1841	The only operations allowed on the scalar are C<substr>/C<vec> that don't
		1842	change the string length, and most read-only operations such as copying it
		1843	or searching it with regexes and so on.
		1844
		1845	Anything else is unsafe and will, at best, result in memory leaks.
		1846
		1847	The memory map associated with the C<$scalar> is automatically removed
		1848	when the C<$scalar> is destroyed, or when the C<IO::AIO::mmap> or
		1849	C<IO::AIO::munmap> functions are called.
		1850
		1851	This calls the C<mmap>(2) function internally. See your system's manual
		1852	page for details on the C<$length>, C<$prot> and C<$flags> parameters.
		1853
		1854	The C<$length> must be larger than zero and smaller than the actual
		1855	filesize.
		1856
		1857	C<$prot> is a combination of C<IO::AIO::PROT_NONE>, C<IO::AIO::PROT_EXEC>,
		1858	C<IO::AIO::PROT_READ> and/or C<IO::AIO::PROT_WRITE>,
		1859
		1860	C<$flags> can be a combination of C<IO::AIO::MAP_SHARED> or
		1861	C<IO::AIO::MAP_PRIVATE>, or a number of system-specific flags (when
		1862	not available, the are defined as 0): C<IO::AIO::MAP_ANONYMOUS>
		1863	(which is set to C<MAP_ANON> if your system only provides this
		1864	constant), C<IO::AIO::MAP_HUGETLB>, C<IO::AIO::MAP_LOCKED>,
		1865	C<IO::AIO::MAP_NORESERVE>, C<IO::AIO::MAP_POPULATE> or
		1866	C<IO::AIO::MAP_NONBLOCK>
		1867
		1868	If C<$fh> is C<undef>, then a file descriptor of C<-1> is passed.
		1869
		1870	C<$offset> is the offset from the start of the file - it generally must be
		1871	a multiple of C<IO::AIO::PAGESIZE> and defaults to C<0>.
		1872
		1873	Example:
		1874
		1875	use Digest::MD5;
		1876	use IO::AIO;
		1877
		1878	open my $fh, "<verybigfile"
		1879	or die "$!";
		1880
		1881	IO::AIO::mmap my $data, -s $fh, IO::AIO::PROT_READ, IO::AIO::MAP_SHARED, $fh
		1882	or die "verybigfile: $!";
		1883
		1884	my $fast_md5 = md5 $data;
		1885
		1886	=item IO::AIO::munmap $scalar
		1887
		1888	Removes a previous mmap and undefines the C<$scalar>.
		1889
		1890	=item IO::AIO::munlock $scalar, $offset = 0, $length = undef
		1891
		1892	Calls the C<munlock> function, undoing the effects of a previous
		1893	C<aio_mlock> call (see its description for details).
		1894
		1895	=item IO::AIO::munlockall
		1896
		1897	Calls the C<munlockall> function.
		1898
		1899	On systems that do not implement C<munlockall>, this function returns
		1900	ENOSYS, otherwise the return value of C<munlockall>.
		1901
		1902	=back
		1903
		1904	=cut
		1905
		1906	min_parallel 8;
		1907
		1908	END { flush }
		1909
227	1;	1910	1;
228		1911
		1912	=head1 EVENT LOOP INTEGRATION
		1913
		1914	It is recommended to use L<AnyEvent::AIO> to integrate IO::AIO
		1915	automatically into many event loops:
		1916
		1917	# AnyEvent integration (EV, Event, Glib, Tk, POE, urxvt, pureperl...)
		1918	use AnyEvent::AIO;
		1919
		1920	You can also integrate IO::AIO manually into many event loops, here are
		1921	some examples of how to do this:
		1922
		1923	# EV integration
		1924	my $aio_w = EV::io IO::AIO::poll_fileno, EV::READ, \&IO::AIO::poll_cb;
		1925
		1926	# Event integration
		1927	Event->io (fd => IO::AIO::poll_fileno,
		1928	poll => 'r',
		1929	cb => \&IO::AIO::poll_cb);
		1930
		1931	# Glib/Gtk2 integration
		1932	add_watch Glib::IO IO::AIO::poll_fileno,
		1933	in => sub { IO::AIO::poll_cb; 1 };
		1934
		1935	# Tk integration
		1936	Tk::Event::IO->fileevent (IO::AIO::poll_fileno, "",
		1937	readable => \&IO::AIO::poll_cb);
		1938
		1939	# Danga::Socket integration
		1940	Danga::Socket->AddOtherFds (IO::AIO::poll_fileno =>
		1941	\&IO::AIO::poll_cb);
		1942
		1943	=head2 FORK BEHAVIOUR
		1944
		1945	Usage of pthreads in a program changes the semantics of fork
		1946	considerably. Specifically, only async-safe functions can be called after
		1947	fork. Perl doesn't know about this, so in general, you cannot call fork
		1948	with defined behaviour in perl if pthreads are involved. IO::AIO uses
		1949	pthreads, so this applies, but many other extensions and (for inexplicable
		1950	reasons) perl itself often is linked against pthreads, so this limitation
		1951	applies to quite a lot of perls.
		1952
		1953	This module no longer tries to fight your OS, or POSIX. That means IO::AIO
		1954	only works in the process that loaded it. Forking is fully supported, but
		1955	using IO::AIO in the child is not.
		1956
		1957	You might get around by not I<using> IO::AIO before (or after)
		1958	forking. You could also try to call the L<IO::AIO::reinit> function in the
		1959	child:
		1960
		1961	=over 4
		1962
		1963	=item IO::AIO::reinit
		1964
		1965	Abandons all current requests and I/O threads and simply reinitialises all
		1966	data structures. This is not an operation supported by any standards, but
		1967	happens to work on GNU/Linux and some newer BSD systems.
		1968
		1969	The only reasonable use for this function is to call it after forking, if
		1970	C<IO::AIO> was used in the parent. Calling it while IO::AIO is active in
		1971	the process will result in undefined behaviour. Calling it at any time
		1972	will also result in any undefined (by POSIX) behaviour.
		1973
229	=back	1974	=back
230		1975
		1976	=head2 MEMORY USAGE
		1977
		1978	Per-request usage:
		1979
		1980	Each aio request uses - depending on your architecture - around 100-200
		1981	bytes of memory. In addition, stat requests need a stat buffer (possibly
		1982	a few hundred bytes), readdir requires a result buffer and so on. Perl
		1983	scalars and other data passed into aio requests will also be locked and
		1984	will consume memory till the request has entered the done state.
		1985
		1986	This is not awfully much, so queuing lots of requests is not usually a
		1987	problem.
		1988
		1989	Per-thread usage:
		1990
		1991	In the execution phase, some aio requests require more memory for
		1992	temporary buffers, and each thread requires a stack and other data
		1993	structures (usually around 16k-128k, depending on the OS).
		1994
231	=head1 BUGS	1995	=head1 KNOWN BUGS
232		1996
233	- aio_open gives a fd, but all other functions expect a perl filehandle.	1997	Known bugs will be fixed in the next release.
234		1998
235	=head1 SEE ALSO	1999	=head1 SEE ALSO
236		2000
237	L<Coro>, L<Linux::AIO>.	2001	L<AnyEvent::AIO> for easy integration into event loops, L<Coro::AIO> for a
		2002	more natural syntax.
238		2003
239	=head1 AUTHOR	2004	=head1 AUTHOR
240		2005
241	Marc Lehmann <schmorp@schmorp.de>	2006	Marc Lehmann <schmorp@schmorp.de>
242	http://home.schmorp.de/	2007	http://home.schmorp.de/

Diff Legend

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