ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/IO-AIO/AIO.pm

Comparing IO-AIO/AIO.pm (file contents):
Revision 1.51 by root, Sat Jun 24 19:14:04 2006 UTC vs.
Revision 1.202 by root, Tue Jul 5 14:02:15 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", 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	# AnyEvent	21	# version 2+ has request and group objects
21	open my $fh, "<&=" . IO::AIO::poll_fileno or die "$!";	22	use IO::AIO 2;
22	my $w = AnyEvent->io (fh => $fh, poll => 'r', cb => sub { IO::AIO::poll_cb });
23		23
24	# Event	24	aioreq_pri 4; # give next request a very high priority
25	Event->io (fd => IO::AIO::poll_fileno,	25	my $req = aio_unlink "/tmp/file", sub { };
26	poll => 'r',	26	$req->cancel; # cancel request if still in queue
27	cb => \&IO::AIO::poll_cb);
28		27
29	# Glib/Gtk2	28	my $grp = aio_group sub { print "all stats done\n" };
30	add_watch Glib::IO IO::AIO::poll_fileno,	29	add $grp aio_stat "..." for ...;
31	in => sub { IO::AIO::poll_cb; 1 };
32
33	# Tk
34	Tk::Event::IO->fileevent (IO::AIO::poll_fileno, "",
35	readable => \&IO::AIO::poll_cb);
36
37	# Danga::Socket
38	Danga::Socket->AddOtherFds (IO::AIO::poll_fileno =>
39	\&IO::AIO::poll_cb);
40
41		30
42	=head1 DESCRIPTION	31	=head1 DESCRIPTION
43		32
44	This module implements asynchronous I/O using whatever means your	33	This module implements asynchronous I/O using whatever means your
45	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>).
46		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
47	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
48	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
49	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
50	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
51	aio functions available on many operating systems. However, they are often	57	functions available on many operating systems. However, they are often
52	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
53	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
54	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.
55		62
56	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,
57	currently not reentrant, so use appropriate locking yourself, always call	64	it is currently not reentrant in any way, so use appropriate locking
58	C<poll_cb> from within the same thread, or never call C<poll_cb> (or other	65	yourself, always call C<poll_cb> from within the same thread, or never
59	C<aio_> functions) recursively.	66	call C<poll_cb> (or other C<aio_> functions) recursively.
		67
		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
60		161
61	=cut	162	=cut
62		163
63	package IO::AIO;	164	package IO::AIO;
64		165
65	no warnings;	166	use Carp ();
66	use strict 'vars';	167
		168	use common::sense;
67		169
68	use base 'Exporter';	170	use base 'Exporter';
69		171
70	BEGIN {	172	BEGIN {
71	our $VERSION = '1.8';	173	our $VERSION = '3.93';
72		174
73	our @EXPORT = qw(aio_sendfile aio_read aio_write aio_open aio_close aio_stat	175	our @AIO_REQ = qw(aio_sendfile aio_read aio_write aio_open aio_close
74	aio_lstat aio_unlink aio_rmdir aio_readdir aio_scandir aio_symlink	176	aio_stat aio_lstat aio_unlink aio_rmdir aio_readdir aio_readdirx
75	aio_fsync aio_fdatasync aio_readahead aio_rename aio_link aio_move);	177	aio_scandir aio_symlink aio_readlink aio_realpath aio_sync aio_fsync
76	our @EXPORT_OK = qw(poll_fileno poll_cb min_parallel max_parallel max_outstanding nreqs);	178	aio_fdatasync aio_sync_file_range aio_pathsync aio_readahead
		179	aio_rename aio_link aio_move aio_copy aio_group
		180	aio_nop aio_mknod aio_load aio_rmtree aio_mkdir aio_chown
		181	aio_chmod aio_utime aio_truncate
		182	aio_msync aio_mtouch aio_mlock aio_mlockall
		183	aio_statvfs);
		184
		185	our @EXPORT = (@AIO_REQ, qw(aioreq_pri aioreq_nice));
		186	our @EXPORT_OK = qw(poll_fileno poll_cb poll_wait flush
		187	min_parallel max_parallel max_idle idle_timeout
		188	nreqs nready npending nthreads
		189	max_poll_time max_poll_reqs
		190	sendfile fadvise madvise
		191	mmap munmap munlock munlockall);
		192
		193	push @AIO_REQ, qw(aio_busy); # not exported
		194
		195	@IO::AIO::GRP::ISA = 'IO::AIO::REQ';
77		196
78	require XSLoader;	197	require XSLoader;
79	XSLoader::load ("IO::AIO", $VERSION);	198	XSLoader::load ("IO::AIO", $VERSION);
80	}	199	}
81		200
82	=head1 FUNCTIONS	201	=head1 FUNCTIONS
83		202
		203	=head2 QUICK OVERVIEW
		204
		205	This section simply lists the prototypes of the most important functions
		206	for quick reference. See the following sections for function-by-function
		207	documentation.
		208
		209	aio_open $pathname, $flags, $mode, $callback->($fh)
		210	aio_close $fh, $callback->($status)
		211	aio_read $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
		212	aio_write $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
		213	aio_sendfile $out_fh, $in_fh, $in_offset, $length, $callback->($retval)
		214	aio_readahead $fh,$offset,$length, $callback->($retval)
		215	aio_stat $fh_or_path, $callback->($status)
		216	aio_lstat $fh, $callback->($status)
		217	aio_statvfs $fh_or_path, $callback->($statvfs)
		218	aio_utime $fh_or_path, $atime, $mtime, $callback->($status)
		219	aio_chown $fh_or_path, $uid, $gid, $callback->($status)
		220	aio_truncate $fh_or_path, $offset, $callback->($status)
		221	aio_chmod $fh_or_path, $mode, $callback->($status)
		222	aio_unlink $pathname, $callback->($status)
		223	aio_mknod $path, $mode, $dev, $callback->($status)
		224	aio_link $srcpath, $dstpath, $callback->($status)
		225	aio_symlink $srcpath, $dstpath, $callback->($status)
		226	aio_readlink $path, $callback->($link)
		227	aio_realpath $path, $callback->($link)
		228	aio_rename $srcpath, $dstpath, $callback->($status)
		229	aio_mkdir $pathname, $mode, $callback->($status)
		230	aio_rmdir $pathname, $callback->($status)
		231	aio_readdir $pathname, $callback->($entries)
		232	aio_readdirx $pathname, $flags, $callback->($entries, $flags)
		233	IO::AIO::READDIR_DENTS IO::AIO::READDIR_DIRS_FIRST
		234	IO::AIO::READDIR_STAT_ORDER IO::AIO::READDIR_FOUND_UNKNOWN
		235	aio_load $path, $data, $callback->($status)
		236	aio_copy $srcpath, $dstpath, $callback->($status)
		237	aio_move $srcpath, $dstpath, $callback->($status)
		238	aio_scandir $path, $maxreq, $callback->($dirs, $nondirs)
		239	aio_rmtree $path, $callback->($status)
		240	aio_sync $callback->($status)
		241	aio_fsync $fh, $callback->($status)
		242	aio_fdatasync $fh, $callback->($status)
		243	aio_sync_file_range $fh, $offset, $nbytes, $flags, $callback->($status)
		244	aio_pathsync $path, $callback->($status)
		245	aio_msync $scalar, $offset = 0, $length = undef, flags = 0, $callback->($status)
		246	aio_mtouch $scalar, $offset = 0, $length = undef, flags = 0, $callback->($status)
		247	aio_mlock $scalar, $offset = 0, $length = undef, $callback->($status)
		248	aio_mlockall $flags, $callback->($status)
		249	aio_group $callback->(...)
		250	aio_nop $callback->()
		251
		252	$prev_pri = aioreq_pri [$pri]
		253	aioreq_nice $pri_adjust
		254
		255	IO::AIO::poll_wait
		256	IO::AIO::poll_cb
		257	IO::AIO::poll
		258	IO::AIO::flush
		259	IO::AIO::max_poll_reqs $nreqs
		260	IO::AIO::max_poll_time $seconds
		261	IO::AIO::min_parallel $nthreads
		262	IO::AIO::max_parallel $nthreads
		263	IO::AIO::max_idle $nthreads
		264	IO::AIO::idle_timeout $seconds
		265	IO::AIO::max_outstanding $maxreqs
		266	IO::AIO::nreqs
		267	IO::AIO::nready
		268	IO::AIO::npending
		269
		270	IO::AIO::sendfile $ofh, $ifh, $offset, $count
		271	IO::AIO::fadvise $fh, $offset, $len, $advice
		272	IO::AIO::madvise $scalar, $offset, $length, $advice
		273	IO::AIO::mprotect $scalar, $offset, $length, $protect
		274	IO::AIO::munlock $scalar, $offset = 0, $length = undef
		275	IO::AIO::munlockall
		276
84	=head2 AIO FUNCTIONS	277	=head2 AIO REQUEST FUNCTIONS
85		278
86	All the C<aio_*> calls are more or less thin wrappers around the syscall	279	All the C<aio_*> calls are more or less thin wrappers around the syscall
87	with the same name (sans C<aio_>). The arguments are similar or identical,	280	with the same name (sans C<aio_>). The arguments are similar or identical,
88	and they all accept an additional (and optional) C<$callback> argument	281	and they all accept an additional (and optional) C<$callback> argument
89	which must be a code reference. This code reference will get called with	282	which must be a code reference. This code reference will get called with
90	the syscall return code (e.g. most syscalls return C<-1> on error, unlike	283	the syscall return code (e.g. most syscalls return C<-1> on error, unlike
91	perl, which usually delivers "false") as it's sole argument when the given	284	perl, which usually delivers "false") as its sole argument after the given
92	syscall has been executed asynchronously.	285	syscall has been executed asynchronously.
93		286
94	All functions expecting a filehandle keep a copy of the filehandle	287	All functions expecting a filehandle keep a copy of the filehandle
95	internally until the request has finished.	288	internally until the request has finished.
96		289
		290	All functions return request objects of type L<IO::AIO::REQ> that allow
		291	further manipulation of those requests while they are in-flight.
		292
97	The pathnames you pass to these routines I<must> be absolute and	293	The pathnames you pass to these routines I<must> be absolute and
98	encoded in byte form. The reason for the former is that at the time the	294	encoded as octets. The reason for the former is that at the time the
99	request is being executed, the current working directory could have	295	request is being executed, the current working directory could have
100	changed. Alternatively, you can make sure that you never change the	296	changed. Alternatively, you can make sure that you never change the
101	current working directory.	297	current working directory anywhere in the program and then use relative
		298	paths.
102		299
103	To encode pathnames to byte form, either make sure you either: a)	300	To encode pathnames as octets, either make sure you either: a) always pass
104	always pass in filenames you got from outside (command line, readdir	301	in filenames you got from outside (command line, readdir etc.) without
105	etc.), b) are ASCII or ISO 8859-1, c) use the Encode module and encode	302	tinkering, b) are ASCII or ISO 8859-1, c) use the Encode module and encode
106	your pathnames to the locale (or other) encoding in effect in the user	303	your pathnames to the locale (or other) encoding in effect in the user
107	environment, d) use Glib::filename_from_unicode on unicode filenames or e)	304	environment, d) use Glib::filename_from_unicode on unicode filenames or e)
108	use something else.	305	use something else to ensure your scalar has the correct contents.
		306
		307	This works, btw. independent of the internal UTF-8 bit, which IO::AIO
		308	handles correctly whether it is set or not.
109		309
110	=over 4	310	=over 4
		311
		312	=item $prev_pri = aioreq_pri [$pri]
		313
		314	Returns the priority value that would be used for the next request and, if
		315	C<$pri> is given, sets the priority for the next aio request.
		316
		317	The default priority is C<0>, the minimum and maximum priorities are C<-4>
		318	and C<4>, respectively. Requests with higher priority will be serviced
		319	first.
		320
		321	The priority will be reset to C<0> after each call to one of the C<aio_*>
		322	functions.
		323
		324	Example: open a file with low priority, then read something from it with
		325	higher priority so the read request is serviced before other low priority
		326	open requests (potentially spamming the cache):
		327
		328	aioreq_pri -3;
		329	aio_open ..., sub {
		330	return unless $_[0];
		331
		332	aioreq_pri -2;
		333	aio_read $_[0], ..., sub {
		334	...
		335	};
		336	};
		337
		338
		339	=item aioreq_nice $pri_adjust
		340
		341	Similar to C<aioreq_pri>, but subtracts the given value from the current
		342	priority, so the effect is cumulative.
		343
111		344
112	=item aio_open $pathname, $flags, $mode, $callback->($fh)	345	=item aio_open $pathname, $flags, $mode, $callback->($fh)
113		346
114	Asynchronously open or create a file and call the callback with a newly	347	Asynchronously open or create a file and call the callback with a newly
115	created filehandle for the file.	348	created filehandle for the file.
…		…
121	list. They are the same as used by C<sysopen>.	354	list. They are the same as used by C<sysopen>.
122		355
123	Likewise, C<$mode> specifies the mode of the newly created file, if it	356	Likewise, C<$mode> specifies the mode of the newly created file, if it
124	didn't exist and C<O_CREAT> has been given, just like perl's C<sysopen>,	357	didn't exist and C<O_CREAT> has been given, just like perl's C<sysopen>,
125	except that it is mandatory (i.e. use C<0> if you don't create new files,	358	except that it is mandatory (i.e. use C<0> if you don't create new files,
126	and C<0666> or C<0777> if you do).	359	and C<0666> or C<0777> if you do). Note that the C<$mode> will be modified
		360	by the umask in effect then the request is being executed, so better never
		361	change the umask.
127		362
128	Example:	363	Example:
129		364
130	aio_open "/etc/passwd", O_RDONLY, 0, sub {	365	aio_open "/etc/passwd", IO::AIO::O_RDONLY, 0, sub {
131	if ($_[0]) {	366	if ($_[0]) {
132	print "open successful, fh is $_[0]\n";	367	print "open successful, fh is $_[0]\n";
133	...	368	...
134	} else {	369	} else {
135	die "open failed: $!\n";	370	die "open failed: $!\n";
136	}	371	}
137	};	372	};
138		373
		374	In addition to all the common open modes/flags (C<O_RDONLY>, C<O_WRONLY>,
		375	C<O_RDWR>, C<O_CREAT>, C<O_TRUNC>, C<O_EXCL> and C<O_APPEND>), the
		376	following POSIX and non-POSIX constants are available (missing ones on
		377	your system are, as usual, C<0>):
		378
		379	C<O_ASYNC>, C<O_DIRECT>, C<O_NOATIME>, C<O_CLOEXEC>, C<O_NOCTTY>, C<O_NOFOLLOW>,
		380	C<O_NONBLOCK>, C<O_EXEC>, C<O_SEARCH>, C<O_DIRECTORY>, C<O_DSYNC>,
		381	C<O_RSYNC>, C<O_SYNC> and C<O_TTY_INIT>.
		382
		383
139	=item aio_close $fh, $callback->($status)	384	=item aio_close $fh, $callback->($status)
140		385
141	Asynchronously close a file and call the callback with the result	386	Asynchronously close a file and call the callback with the result
142	code. I<WARNING:> although accepted, you should not pass in a perl	387	code.
143	filehandle here, as perl will likely close the file descriptor another
144	time when the filehandle is destroyed. Normally, you can safely call perls
145	C<close> or just let filehandles go out of scope.
146		388
147	This is supposed to be a bug in the API, so that might change. It's	389	Unfortunately, you can't do this to perl. Perl I<insists> very strongly on
148	therefore best to avoid this function.	390	closing the file descriptor associated with the filehandle itself.
		391
		392	Therefore, C<aio_close> will not close the filehandle - instead it will
		393	use dup2 to overwrite the file descriptor with the write-end of a pipe
		394	(the pipe fd will be created on demand and will be cached).
		395
		396	Or in other words: the file descriptor will be closed, but it will not be
		397	free for reuse until the perl filehandle is closed.
		398
		399	=cut
149		400
150	=item aio_read $fh,$offset,$length, $data,$dataoffset, $callback->($retval)	401	=item aio_read $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
151		402
152	=item aio_write $fh,$offset,$length, $data,$dataoffset, $callback->($retval)	403	=item aio_write $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
153		404
154	Reads or writes C<length> bytes from the specified C<fh> and C<offset>	405	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	406	C<$offset> into the scalar given by C<$data> and offset C<$dataoffset>
156	callback without the actual number of bytes read (or -1 on error, just	407	and calls the callback without the actual number of bytes read (or -1 on
157	like the syscall).	408	error, just like the syscall).
		409
		410	C<aio_read> will, like C<sysread>, shrink or grow the C<$data> scalar to
		411	offset plus the actual number of bytes read.
		412
		413	If C<$offset> is undefined, then the current file descriptor offset will
		414	be used (and updated), otherwise the file descriptor offset will not be
		415	changed by these calls.
		416
		417	If C<$length> is undefined in C<aio_write>, use the remaining length of
		418	C<$data>.
		419
		420	If C<$dataoffset> is less than zero, it will be counted from the end of
		421	C<$data>.
158		422
159	The C<$data> scalar I<MUST NOT> be modified in any way while the request	423	The C<$data> scalar I<MUST NOT> be modified in any way while the request
160	is outstanding. Modifying it can result in segfaults or WW3 (if the	424	is outstanding. Modifying it can result in segfaults or World War III (if
161	necessary/optional hardware is installed).	425	the necessary/optional hardware is installed).
162		426
163	Example: Read 15 bytes at offset 7 into scalar C<$buffer>, starting at	427	Example: Read 15 bytes at offset 7 into scalar C<$buffer>, starting at
164	offset C<0> within the scalar:	428	offset C<0> within the scalar:
165		429
166	aio_read $fh, 7, 15, $buffer, 0, sub {	430	aio_read $fh, 7, 15, $buffer, 0, sub {
167	$_[0] > 0 or die "read error: $!";	431	$_[0] > 0 or die "read error: $!";
168	print "read $_[0] bytes: <$buffer>\n";	432	print "read $_[0] bytes: <$buffer>\n";
169	};	433	};
170		434
171	=item aio_move $srcpath, $dstpath, $callback->($status)
172
173	[EXPERIMENTAL]
174
175	Try to move the I<file> (directories not supported as either source or destination)
176	from C<$srcpath> to C<$dstpath> and call the callback with the C<0> (error) or C<-1> ok.
177
178	This is a composite request that tries to rename(2) the file first. If
179	rename files with C<EXDEV>, it creates the destination file with mode 0200
180	and copies the contents of the source file into it using C<aio_sendfile>,
181	followed by restoring atime, mtime, access mode and uid/gid, in that
182	order, and unlinking the C<$srcpath>.
183
184	If an error occurs, the partial destination file will be unlinked, if
185	possible, except when setting atime, mtime, access mode and uid/gid, where
186	errors are being ignored.
187
188	=cut
189
190	sub aio_move($$$) {
191	my ($src, $dst, $cb) = @_;
192
193	aio_rename $src, $dst, sub {
194	if ($_[0] && $! == EXDEV) {
195	aio_open $src, O_RDONLY, 0, sub {
196	if (my $src_fh = $_[0]) {
197	my @stat = stat $src_fh;
198
199	aio_open $dst, O_WRONLY, 0200, sub {
200	if (my $dst_fh = $_[0]) {
201	aio_sendfile $dst_fh, $src_fh, 0, $stat[7], sub {
202	close $src_fh;
203
204	if ($_[0] == $stat[7]) {
205	utime $stat[8], $stat[9], $dst;
206	chmod $stat[2] & 07777, $dst_fh;
207	chown $stat[4], $stat[5], $dst_fh;
208	close $dst_fh;
209
210	aio_unlink $src, sub {
211	$cb->($_[0]);
212	};
213	} else {
214	my $errno = $!;
215	aio_unlink $dst, sub {
216	$! = $errno;
217	$cb->(-1);
218	};
219	}
220	};
221	} else {
222	$cb->(-1);
223	}
224	},
225
226	} else {
227	$cb->(-1);
228	}
229	};
230	} else {
231	$cb->($_[0]);
232	}
233	};
234	}
235		435
236	=item aio_sendfile $out_fh, $in_fh, $in_offset, $length, $callback->($retval)	436	=item aio_sendfile $out_fh, $in_fh, $in_offset, $length, $callback->($retval)
237		437
238	Tries to copy C<$length> bytes from C<$in_fh> to C<$out_fh>. It starts	438	Tries to copy C<$length> bytes from C<$in_fh> to C<$out_fh>. It starts
239	reading at byte offset C<$in_offset>, and starts writing at the current	439	reading at byte offset C<$in_offset>, and starts writing at the current
240	file offset of C<$out_fh>. Because of that, it is not safe to issue more	440	file offset of C<$out_fh>. Because of that, it is not safe to issue more
241	than one C<aio_sendfile> per C<$out_fh>, as they will interfere with each	441	than one C<aio_sendfile> per C<$out_fh>, as they will interfere with each
242	other.	442	other. The same C<$in_fh> works fine though, as this function does not
		443	move or use the file offset of C<$in_fh>.
243		444
		445	Please note that C<aio_sendfile> can read more bytes from C<$in_fh> than
		446	are written, and there is no way to find out how many more bytes have been
		447	read from C<aio_sendfile> alone, as C<aio_sendfile> only provides the
		448	number of bytes written to C<$out_fh>. Only if the result value equals
		449	C<$length> one can assume that C<$length> bytes have been read.
		450
		451	Unlike with other C<aio_> functions, it makes a lot of sense to use
		452	C<aio_sendfile> on non-blocking sockets, as long as one end (typically
		453	the C<$in_fh>) is a file - the file I/O will then be asynchronous, while
		454	the socket I/O will be non-blocking. Note, however, that you can run
		455	into a trap where C<aio_sendfile> reads some data with readahead, then
		456	fails to write all data, and when the socket is ready the next time, the
		457	data in the cache is already lost, forcing C<aio_sendfile> to again hit
		458	the disk. Explicit C<aio_read> + C<aio_write> let's you better control
		459	resource usage.
		460
244	This call tries to make use of a native C<sendfile> syscall to provide	461	This call tries to make use of a native C<sendfile>-like syscall to
245	zero-copy operation. For this to work, C<$out_fh> should refer to a	462	provide zero-copy operation. For this to work, C<$out_fh> should refer to
246	socket, and C<$in_fh> should refer to mmap'able file.	463	a socket, and C<$in_fh> should refer to an mmap'able file.
247		464
248	If the native sendfile call fails or is not implemented, it will be	465	If a native sendfile cannot be found or it fails with C<ENOSYS>,
249	emulated, so you can call C<aio_sendfile> on any type of filehandle	466	C<EINVAL>, C<ENOTSUP>, C<EOPNOTSUPP>, C<EAFNOSUPPORT>, C<EPROTOTYPE> or
		467	C<ENOTSOCK>, it will be emulated, so you can call C<aio_sendfile> on any
250	regardless of the limitations of the operating system.	468	type of filehandle regardless of the limitations of the operating system.
251		469
252	Please note, however, that C<aio_sendfile> can read more bytes from	470	As native sendfile syscalls (as practically any non-POSIX interface hacked
253	C<$in_fh> than are written, and there is no way to find out how many	471	together in a hurry to improve benchmark numbers) tend to be rather buggy
254	bytes have been read from C<aio_sendfile> alone, as C<aio_sendfile> only	472	on many systems, this implementation tries to work around some known bugs
255	provides the number of bytes written to C<$out_fh>. Only if the result	473	in Linux and FreeBSD kernels (probably others, too), but that might fail,
256	value equals C<$length> one can assume that C<$length> bytes have been	474	so you really really should check the return value of C<aio_sendfile> -
257	read.	475	fewre bytes than expected might have been transferred.
		476
258		477
259	=item aio_readahead $fh,$offset,$length, $callback->($retval)	478	=item aio_readahead $fh,$offset,$length, $callback->($retval)
260		479
261	C<aio_readahead> populates the page cache with data from a file so that	480	C<aio_readahead> populates the page cache with data from a file so that
262	subsequent reads from that file will not block on disk I/O. The C<$offset>	481	subsequent reads from that file will not block on disk I/O. The C<$offset>
…		…
268	file. The current file offset of the file is left unchanged.	487	file. The current file offset of the file is left unchanged.
269		488
270	If that syscall doesn't exist (likely if your OS isn't Linux) it will be	489	If that syscall doesn't exist (likely if your OS isn't Linux) it will be
271	emulated by simply reading the data, which would have a similar effect.	490	emulated by simply reading the data, which would have a similar effect.
272		491
		492
273	=item aio_stat $fh_or_path, $callback->($status)	493	=item aio_stat $fh_or_path, $callback->($status)
274		494
275	=item aio_lstat $fh, $callback->($status)	495	=item aio_lstat $fh, $callback->($status)
276		496
277	Works like perl's C<stat> or C<lstat> in void context. The callback will	497	Works like perl's C<stat> or C<lstat> in void context. The callback will
…		…
282	for an explanation.	502	for an explanation.
283		503
284	Currently, the stats are always 64-bit-stats, i.e. instead of returning an	504	Currently, the stats are always 64-bit-stats, i.e. instead of returning an
285	error when stat'ing a large file, the results will be silently truncated	505	error when stat'ing a large file, the results will be silently truncated
286	unless perl itself is compiled with large file support.	506	unless perl itself is compiled with large file support.
		507
		508	To help interpret the mode and dev/rdev stat values, IO::AIO offers the
		509	following constants and functions (if not implemented, the constants will
		510	be C<0> and the functions will either C<croak> or fall back on traditional
		511	behaviour).
		512
		513	C<S_IFMT>, C<S_IFIFO>, C<S_IFCHR>, C<S_IFBLK>, C<S_IFLNK>, C<S_IFREG>,
		514	C<S_IFDIR>, C<S_IFWHT>, C<S_IFSOCK>, C<IO::AIO::major $dev_t>,
		515	C<IO::AIO::minor $dev_t>, C<IO::AIO::makedev $major, $minor>.
287		516
288	Example: Print the length of F</etc/passwd>:	517	Example: Print the length of F</etc/passwd>:
289		518
290	aio_stat "/etc/passwd", sub {	519	aio_stat "/etc/passwd", sub {
291	$_[0] and die "stat failed: $!";	520	$_[0] and die "stat failed: $!";
292	print "size is ", -s _, "\n";	521	print "size is ", -s _, "\n";
293	};	522	};
294		523
		524
		525	=item aio_statvfs $fh_or_path, $callback->($statvfs)
		526
		527	Works like the POSIX C<statvfs> or C<fstatvfs> syscalls, depending on
		528	whether a file handle or path was passed.
		529
		530	On success, the callback is passed a hash reference with the following
		531	members: C<bsize>, C<frsize>, C<blocks>, C<bfree>, C<bavail>, C<files>,
		532	C<ffree>, C<favail>, C<fsid>, C<flag> and C<namemax>. On failure, C<undef>
		533	is passed.
		534
		535	The following POSIX IO::AIO::ST_* constants are defined: C<ST_RDONLY> and
		536	C<ST_NOSUID>.
		537
		538	The following non-POSIX IO::AIO::ST_* flag masks are defined to
		539	their correct value when available, or to C<0> on systems that do
		540	not support them: C<ST_NODEV>, C<ST_NOEXEC>, C<ST_SYNCHRONOUS>,
		541	C<ST_MANDLOCK>, C<ST_WRITE>, C<ST_APPEND>, C<ST_IMMUTABLE>, C<ST_NOATIME>,
		542	C<ST_NODIRATIME> and C<ST_RELATIME>.
		543
		544	Example: stat C</wd> and dump out the data if successful.
		545
		546	aio_statvfs "/wd", sub {
		547	my $f = $_[0]
		548	or die "statvfs: $!";
		549
		550	use Data::Dumper;
		551	say Dumper $f;
		552	};
		553
		554	# result:
		555	{
		556	bsize => 1024,
		557	bfree => 4333064312,
		558	blocks => 10253828096,
		559	files => 2050765568,
		560	flag => 4096,
		561	favail => 2042092649,
		562	bavail => 4333064312,
		563	ffree => 2042092649,
		564	namemax => 255,
		565	frsize => 1024,
		566	fsid => 1810
		567	}
		568
		569
		570	=item aio_utime $fh_or_path, $atime, $mtime, $callback->($status)
		571
		572	Works like perl's C<utime> function (including the special case of $atime
		573	and $mtime being undef). Fractional times are supported if the underlying
		574	syscalls support them.
		575
		576	When called with a pathname, uses utimes(2) if available, otherwise
		577	utime(2). If called on a file descriptor, uses futimes(2) if available,
		578	otherwise returns ENOSYS, so this is not portable.
		579
		580	Examples:
		581
		582	# set atime and mtime to current time (basically touch(1)):
		583	aio_utime "path", undef, undef;
		584	# set atime to current time and mtime to beginning of the epoch:
		585	aio_utime "path", time, undef; # undef==0
		586
		587
		588	=item aio_chown $fh_or_path, $uid, $gid, $callback->($status)
		589
		590	Works like perl's C<chown> function, except that C<undef> for either $uid
		591	or $gid is being interpreted as "do not change" (but -1 can also be used).
		592
		593	Examples:
		594
		595	# same as "chown root path" in the shell:
		596	aio_chown "path", 0, -1;
		597	# same as above:
		598	aio_chown "path", 0, undef;
		599
		600
		601	=item aio_truncate $fh_or_path, $offset, $callback->($status)
		602
		603	Works like truncate(2) or ftruncate(2).
		604
		605
		606	=item aio_chmod $fh_or_path, $mode, $callback->($status)
		607
		608	Works like perl's C<chmod> function.
		609
		610
295	=item aio_unlink $pathname, $callback->($status)	611	=item aio_unlink $pathname, $callback->($status)
296		612
297	Asynchronously unlink (delete) a file and call the callback with the	613	Asynchronously unlink (delete) a file and call the callback with the
298	result code.	614	result code.
299		615
		616
		617	=item aio_mknod $path, $mode, $dev, $callback->($status)
		618
		619	[EXPERIMENTAL]
		620
		621	Asynchronously create a device node (or fifo). See mknod(2).
		622
		623	The only (POSIX-) portable way of calling this function is:
		624
		625	aio_mknod $path, IO::AIO::S_IFIFO | $mode, 0, sub { ...
		626
		627	See C<aio_stat> for info about some potentially helpful extra constants
		628	and functions.
		629
300	=item aio_link $srcpath, $dstpath, $callback->($status)	630	=item aio_link $srcpath, $dstpath, $callback->($status)
301		631
302	Asynchronously create a new link to the existing object at C<$srcpath> at	632	Asynchronously create a new link to the existing object at C<$srcpath> at
303	the path C<$dstpath> and call the callback with the result code.	633	the path C<$dstpath> and call the callback with the result code.
304		634
		635
305	=item aio_symlink $srcpath, $dstpath, $callback->($status)	636	=item aio_symlink $srcpath, $dstpath, $callback->($status)
306		637
307	Asynchronously create a new symbolic link to the existing object at C<$srcpath> at	638	Asynchronously create a new symbolic link to the existing object at C<$srcpath> at
308	the path C<$dstpath> and call the callback with the result code.	639	the path C<$dstpath> and call the callback with the result code.
309		640
		641
		642	=item aio_readlink $path, $callback->($link)
		643
		644	Asynchronously read the symlink specified by C<$path> and pass it to
		645	the callback. If an error occurs, nothing or undef gets passed to the
		646	callback.
		647
		648
		649	=item aio_realpath $path, $callback->($path)
		650
		651	Asynchronously make the path absolute and resolve any symlinks in
		652	C<$path>. The resulting path only consists of directories (Same as
		653	L<Cwd::realpath>).
		654
		655	This request can be used to get the absolute path of the current working
		656	directory by passing it a path of F<.> (a single dot).
		657
		658
310	=item aio_rename $srcpath, $dstpath, $callback->($status)	659	=item aio_rename $srcpath, $dstpath, $callback->($status)
311		660
312	Asynchronously rename the object at C<$srcpath> to C<$dstpath>, just as	661	Asynchronously rename the object at C<$srcpath> to C<$dstpath>, just as
313	rename(2) and call the callback with the result code.	662	rename(2) and call the callback with the result code.
314		663
		664
		665	=item aio_mkdir $pathname, $mode, $callback->($status)
		666
		667	Asynchronously mkdir (create) a directory and call the callback with
		668	the result code. C<$mode> will be modified by the umask at the time the
		669	request is executed, so do not change your umask.
		670
		671
315	=item aio_rmdir $pathname, $callback->($status)	672	=item aio_rmdir $pathname, $callback->($status)
316		673
317	Asynchronously rmdir (delete) a directory and call the callback with the	674	Asynchronously rmdir (delete) a directory and call the callback with the
318	result code.	675	result code.
		676
319		677
320	=item aio_readdir $pathname, $callback->($entries)	678	=item aio_readdir $pathname, $callback->($entries)
321		679
322	Unlike the POSIX call of the same name, C<aio_readdir> reads an entire	680	Unlike the POSIX call of the same name, C<aio_readdir> reads an entire
323	directory (i.e. opendir + readdir + closedir). The entries will not be	681	directory (i.e. opendir + readdir + closedir). The entries will not be
324	sorted, and will B<NOT> include the C<.> and C<..> entries.	682	sorted, and will B<NOT> include the C<.> and C<..> entries.
325		683
326	The callback a single argument which is either C<undef> or an array-ref	684	The callback is passed a single argument which is either C<undef> or an
327	with the filenames.	685	array-ref with the filenames.
		686
		687
		688	=item aio_readdirx $pathname, $flags, $callback->($entries, $flags)
		689
		690	Quite similar to C<aio_readdir>, but the C<$flags> argument allows to tune
		691	behaviour and output format. In case of an error, C<$entries> will be
		692	C<undef>.
		693
		694	The flags are a combination of the following constants, ORed together (the
		695	flags will also be passed to the callback, possibly modified):
		696
		697	=over 4
		698
		699	=item IO::AIO::READDIR_DENTS
		700
		701	When this flag is off, then the callback gets an arrayref consisting of
		702	names only (as with C<aio_readdir>), otherwise it gets an arrayref with
		703	C<[$name, $type, $inode]> arrayrefs, each describing a single directory
		704	entry in more detail.
		705
		706	C<$name> is the name of the entry.
		707
		708	C<$type> is one of the C<IO::AIO::DT_xxx> constants:
		709
		710	C<IO::AIO::DT_UNKNOWN>, C<IO::AIO::DT_FIFO>, C<IO::AIO::DT_CHR>, C<IO::AIO::DT_DIR>,
		711	C<IO::AIO::DT_BLK>, C<IO::AIO::DT_REG>, C<IO::AIO::DT_LNK>, C<IO::AIO::DT_SOCK>,
		712	C<IO::AIO::DT_WHT>.
		713
		714	C<IO::AIO::DT_UNKNOWN> means just that: readdir does not know. If you need to
		715	know, you have to run stat yourself. Also, for speed reasons, the C<$type>
		716	scalars are read-only: you can not modify them.
		717
		718	C<$inode> is the inode number (which might not be exact on systems with 64
		719	bit inode numbers and 32 bit perls). This field has unspecified content on
		720	systems that do not deliver the inode information.
		721
		722	=item IO::AIO::READDIR_DIRS_FIRST
		723
		724	When this flag is set, then the names will be returned in an order where
		725	likely directories come first, in optimal stat order. This is useful when
		726	you need to quickly find directories, or you want to find all directories
		727	while avoiding to stat() each entry.
		728
		729	If the system returns type information in readdir, then this is used
		730	to find directories directly. Otherwise, likely directories are names
		731	beginning with ".", or otherwise names with no dots, of which names with
		732	short names are tried first.
		733
		734	=item IO::AIO::READDIR_STAT_ORDER
		735
		736	When this flag is set, then the names will be returned in an order
		737	suitable for stat()'ing each one. That is, when you plan to stat()
		738	all files in the given directory, then the returned order will likely
		739	be fastest.
		740
		741	If both this flag and C<IO::AIO::READDIR_DIRS_FIRST> are specified, then
		742	the likely dirs come first, resulting in a less optimal stat order.
		743
		744	=item IO::AIO::READDIR_FOUND_UNKNOWN
		745
		746	This flag should not be set when calling C<aio_readdirx>. Instead, it
		747	is being set by C<aio_readdirx>, when any of the C<$type>'s found were
		748	C<IO::AIO::DT_UNKNOWN>. The absense of this flag therefore indicates that all
		749	C<$type>'s are known, which can be used to speed up some algorithms.
		750
		751	=back
		752
		753
		754	=item aio_load $path, $data, $callback->($status)
		755
		756	This is a composite request that tries to fully load the given file into
		757	memory. Status is the same as with aio_read.
		758
		759	=cut
		760
		761	sub aio_load($$;$) {
		762	my ($path, undef, $cb) = @_;
		763	my $data = \$_[1];
		764
		765	my $pri = aioreq_pri;
		766	my $grp = aio_group $cb;
		767
		768	aioreq_pri $pri;
		769	add $grp aio_open $path, O_RDONLY, 0, sub {
		770	my $fh = shift
		771	or return $grp->result (-1);
		772
		773	aioreq_pri $pri;
		774	add $grp aio_read $fh, 0, (-s $fh), $$data, 0, sub {
		775	$grp->result ($_[0]);
		776	};
		777	};
		778
		779	$grp
		780	}
		781
		782	=item aio_copy $srcpath, $dstpath, $callback->($status)
		783
		784	Try to copy the I<file> (directories not supported as either source or
		785	destination) from C<$srcpath> to C<$dstpath> and call the callback with
		786	a status of C<0> (ok) or C<-1> (error, see C<$!>).
		787
		788	This is a composite request that creates the destination file with
		789	mode 0200 and copies the contents of the source file into it using
		790	C<aio_sendfile>, followed by restoring atime, mtime, access mode and
		791	uid/gid, in that order.
		792
		793	If an error occurs, the partial destination file will be unlinked, if
		794	possible, except when setting atime, mtime, access mode and uid/gid, where
		795	errors are being ignored.
		796
		797	=cut
		798
		799	sub aio_copy($$;$) {
		800	my ($src, $dst, $cb) = @_;
		801
		802	my $pri = aioreq_pri;
		803	my $grp = aio_group $cb;
		804
		805	aioreq_pri $pri;
		806	add $grp aio_open $src, O_RDONLY, 0, sub {
		807	if (my $src_fh = $_[0]) {
		808	my @stat = stat $src_fh; # hmm, might block over nfs?
		809
		810	aioreq_pri $pri;
		811	add $grp aio_open $dst, O_CREAT | O_WRONLY | O_TRUNC, 0200, sub {
		812	if (my $dst_fh = $_[0]) {
		813	aioreq_pri $pri;
		814	add $grp aio_sendfile $dst_fh, $src_fh, 0, $stat[7], sub {
		815	if ($_[0] == $stat[7]) {
		816	$grp->result (0);
		817	close $src_fh;
		818
		819	my $ch = sub {
		820	aioreq_pri $pri;
		821	add $grp aio_chmod $dst_fh, $stat[2] & 07777, sub {
		822	aioreq_pri $pri;
		823	add $grp aio_chown $dst_fh, $stat[4], $stat[5], sub {
		824	aioreq_pri $pri;
		825	add $grp aio_close $dst_fh;
		826	}
		827	};
		828	};
		829
		830	aioreq_pri $pri;
		831	add $grp aio_utime $dst_fh, $stat[8], $stat[9], sub {
		832	if ($_[0] < 0 && $! == ENOSYS) {
		833	aioreq_pri $pri;
		834	add $grp aio_utime $dst, $stat[8], $stat[9], $ch;
		835	} else {
		836	$ch->();
		837	}
		838	};
		839	} else {
		840	$grp->result (-1);
		841	close $src_fh;
		842	close $dst_fh;
		843
		844	aioreq $pri;
		845	add $grp aio_unlink $dst;
		846	}
		847	};
		848	} else {
		849	$grp->result (-1);
		850	}
		851	},
		852
		853	} else {
		854	$grp->result (-1);
		855	}
		856	};
		857
		858	$grp
		859	}
		860
		861	=item aio_move $srcpath, $dstpath, $callback->($status)
		862
		863	Try to move the I<file> (directories not supported as either source or
		864	destination) from C<$srcpath> to C<$dstpath> and call the callback with
		865	a status of C<0> (ok) or C<-1> (error, see C<$!>).
		866
		867	This is a composite request that tries to rename(2) the file first; if
		868	rename fails with C<EXDEV>, it copies the file with C<aio_copy> and, if
		869	that is successful, unlinks the C<$srcpath>.
		870
		871	=cut
		872
		873	sub aio_move($$;$) {
		874	my ($src, $dst, $cb) = @_;
		875
		876	my $pri = aioreq_pri;
		877	my $grp = aio_group $cb;
		878
		879	aioreq_pri $pri;
		880	add $grp aio_rename $src, $dst, sub {
		881	if ($_[0] && $! == EXDEV) {
		882	aioreq_pri $pri;
		883	add $grp aio_copy $src, $dst, sub {
		884	$grp->result ($_[0]);
		885
		886	unless ($_[0]) {
		887	aioreq_pri $pri;
		888	add $grp aio_unlink $src;
		889	}
		890	};
		891	} else {
		892	$grp->result ($_[0]);
		893	}
		894	};
		895
		896	$grp
		897	}
328		898
329	=item aio_scandir $path, $maxreq, $callback->($dirs, $nondirs)	899	=item aio_scandir $path, $maxreq, $callback->($dirs, $nondirs)
330		900
331	Scans a directory (similar to C<aio_readdir>) and tries to separate the	901	Scans a directory (similar to C<aio_readdir>) but additionally tries to
332	entries of directory C<$path> into two sets of names, ones you can recurse	902	efficiently separate the entries of directory C<$path> into two sets of
333	into (directories), and ones you cannot recurse into (everything else).	903	names, directories you can recurse into (directories), and ones you cannot
		904	recurse into (everything else, including symlinks to directories).
334		905
335	C<aio_scandir> is a composite request that consists of many	906	C<aio_scandir> is a composite request that creates of many sub requests_
336	aio-primitives. C<$maxreq> specifies the maximum number of outstanding	907	C<$maxreq> specifies the maximum number of outstanding aio requests that
337	aio requests that this function generates. If it is C<< <= 0 >>, then a	908	this function generates. If it is C<< <= 0 >>, then a suitable default
338	suitable default will be chosen (currently 8).	909	will be chosen (currently 4).
339		910
340	On error, the callback is called without arguments, otherwise it receives	911	On error, the callback is called without arguments, otherwise it receives
341	two array-refs with path-relative entry names.	912	two array-refs with path-relative entry names.
342		913
343	Example:	914	Example:
…		…
350		921
351	Implementation notes.	922	Implementation notes.
352		923
353	The C<aio_readdir> cannot be avoided, but C<stat()>'ing every entry can.	924	The C<aio_readdir> cannot be avoided, but C<stat()>'ing every entry can.
354		925
		926	If readdir returns file type information, then this is used directly to
		927	find directories.
		928
355	After reading the directory, the modification time, size etc. of the	929	Otherwise, after reading the directory, the modification time, size etc.
356	directory before and after the readdir is checked, and if they match, the	930	of the directory before and after the readdir is checked, and if they
357	link count will be used to decide how many entries are directories (if	931	match (and isn't the current time), the link count will be used to decide
358	>= 2). Otherwise, no knowledge of the number of subdirectories will be	932	how many entries are directories (if >= 2). Otherwise, no knowledge of the
359	assumed.	933	number of subdirectories will be assumed.
360		934
361	Then entires will be sorted into likely directories (everything without a	935	Then entries will be sorted into likely directories a non-initial dot
362	non-initial dot) and likely non-directories (everything else). Then every	936	currently) and likely non-directories (see C<aio_readdirx>). Then every
363	entry + C</.> will be C<stat>'ed, likely directories first. This is often	937	entry plus an appended C</.> will be C<stat>'ed, likely directories first,
		938	in order of their inode numbers. If that succeeds, it assumes that the
		939	entry is a directory or a symlink to directory (which will be checked
		940	seperately). This is often faster than stat'ing the entry itself because
364	faster because filesystems might detect the type of the entry without	941	filesystems might detect the type of the entry without reading the inode
365	reading the inode data (e.g. ext2fs filetype feature). If that succeeds,	942	data (e.g. ext2fs filetype feature), even on systems that cannot return
366	it assumes that the entry is a directory or a symlink to directory (which	943	the filetype information on readdir.
367	will be checked seperately).
368		944
369	If the known number of directories has been reached, the rest of the	945	If the known number of directories (link count - 2) has been reached, the
370	entries is assumed to be non-directories.	946	rest of the entries is assumed to be non-directories.
		947
		948	This only works with certainty on POSIX (= UNIX) filesystems, which
		949	fortunately are the vast majority of filesystems around.
		950
		951	It will also likely work on non-POSIX filesystems with reduced efficiency
		952	as those tend to return 0 or 1 as link counts, which disables the
		953	directory counting heuristic.
371		954
372	=cut	955	=cut
373		956
374	sub aio_scandir($$$) {	957	sub aio_scandir($$;$) {
375	my ($path, $maxreq, $cb) = @_;	958	my ($path, $maxreq, $cb) = @_;
376		959
		960	my $pri = aioreq_pri;
		961
		962	my $grp = aio_group $cb;
		963
377	$maxreq = 8 if $maxreq <= 0;	964	$maxreq = 4 if $maxreq <= 0;
378		965
379	# stat once	966	# stat once
		967	aioreq_pri $pri;
380	aio_stat $path, sub {	968	add $grp aio_stat $path, sub {
381	return $cb->() if $_[0];	969	return $grp->result () if $_[0];
		970	my $now = time;
382	my $hash1 = join ":", (stat _)[0,1,3,7,9];	971	my $hash1 = join ":", (stat _)[0,1,3,7,9];
383		972
384	# read the directory entries	973	# read the directory entries
385	aio_readdir $path, sub {	974	aioreq_pri $pri;
		975	add $grp aio_readdirx $path, READDIR_DIRS_FIRST, sub {
386	my $entries = shift	976	my $entries = shift
387	or return $cb->();	977	or return $grp->result ();
388		978
389	# stat the dir another time	979	# stat the dir another time
		980	aioreq_pri $pri;
390	aio_stat $path, sub {	981	add $grp aio_stat $path, sub {
391	my $hash2 = join ":", (stat _)[0,1,3,7,9];	982	my $hash2 = join ":", (stat _)[0,1,3,7,9];
392		983
393	my $ndirs;	984	my $ndirs;
394		985
395	# take the slow route if anything looks fishy	986	# take the slow route if anything looks fishy
396	if ($hash1 ne $hash2) {	987	if ($hash1 ne $hash2 or (stat _)[9] == $now) {
397	$ndirs = -1;	988	$ndirs = -1;
398	} else {	989	} else {
399	# if nlink == 2, we are finished	990	# if nlink == 2, we are finished
400	# on non-posix-fs's, we rely on nlink < 2	991	# for non-posix-fs's, we rely on nlink < 2
401	$ndirs = (stat _)[3] - 2	992	$ndirs = (stat _)[3] - 2
402	or return $cb->([], $entries);	993	or return $grp->result ([], $entries);
403	}	994	}
404		995
405	# sort into likely dirs and likely nondirs
406	# dirs == files without ".", short entries first
407	$entries = [map $_->[0],
408	sort { $b->[1] cmp $a->[1] }
409	map [$_, sprintf "%s%04d", (/.\./ ? "1" : "0"), length],
410	@$entries];
411
412	my (@dirs, @nondirs);	996	my (@dirs, @nondirs);
413		997
414	my ($statcb, $schedcb);	998	my $statgrp = add $grp aio_group sub {
415	my $nreq = 0;	999	$grp->result (\@dirs, \@nondirs);
		1000	};
416		1001
417	$schedcb = sub {	1002	limit $statgrp $maxreq;
418	if (@$entries) {	1003	feed $statgrp sub {
419	if ($nreq < $maxreq) {	1004	return unless @$entries;
420	my $ent = pop @$entries;	1005	my $entry = shift @$entries;
		1006
		1007	aioreq_pri $pri;
		1008	add $statgrp aio_stat "$path/$entry/.", sub {
		1009	if ($_[0] < 0) {
		1010	push @nondirs, $entry;
		1011	} else {
		1012	# need to check for real directory
		1013	aioreq_pri $pri;
		1014	add $statgrp aio_lstat "$path/$entry", sub {
		1015	if (-d _) {
		1016	push @dirs, $entry;
		1017
		1018	unless (--$ndirs) {
		1019	push @nondirs, @$entries;
		1020	feed $statgrp;
		1021	}
		1022	} else {
		1023	push @nondirs, $entry;
		1024	}
421	$nreq++;	1025	}
422	aio_stat "$path/$ent/.", sub { $statcb->($_[0], $ent) };
423	}	1026	}
424	} elsif (!$nreq) {
425	# finished
426	undef $statcb;
427	undef $schedcb;
428	$cb->(\@dirs, \@nondirs) if $cb;
429	undef $cb;
430	}	1027	};
431	};	1028	};
432	$statcb = sub {
433	my ($status, $entry) = @_;
434
435	if ($status < 0) {
436	$nreq--;
437	push @nondirs, $entry;
438	&$schedcb;
439	} else {
440	# need to check for real directory
441	aio_lstat "$path/$entry", sub {
442	$nreq--;
443
444	if (-d _) {
445	push @dirs, $entry;
446
447	if (!--$ndirs) {
448	push @nondirs, @$entries;
449	$entries = [];
450	}
451	} else {
452	push @nondirs, $entry;
453	}
454
455	&$schedcb;
456	}
457	}
458	};
459
460	&$schedcb while @$entries && $nreq < $maxreq;
461	};	1029	};
462	};	1030	};
463	};	1031	};
		1032
		1033	$grp
464	}	1034	}
		1035
		1036	=item aio_rmtree $path, $callback->($status)
		1037
		1038	Delete a directory tree starting (and including) C<$path>, return the
		1039	status of the final C<rmdir> only. This is a composite request that
		1040	uses C<aio_scandir> to recurse into and rmdir directories, and unlink
		1041	everything else.
		1042
		1043	=cut
		1044
		1045	sub aio_rmtree;
		1046	sub aio_rmtree($;$) {
		1047	my ($path, $cb) = @_;
		1048
		1049	my $pri = aioreq_pri;
		1050	my $grp = aio_group $cb;
		1051
		1052	aioreq_pri $pri;
		1053	add $grp aio_scandir $path, 0, sub {
		1054	my ($dirs, $nondirs) = @_;
		1055
		1056	my $dirgrp = aio_group sub {
		1057	add $grp aio_rmdir $path, sub {
		1058	$grp->result ($_[0]);
		1059	};
		1060	};
		1061
		1062	(aioreq_pri $pri), add $dirgrp aio_rmtree "$path/$_" for @$dirs;
		1063	(aioreq_pri $pri), add $dirgrp aio_unlink "$path/$_" for @$nondirs;
		1064
		1065	add $grp $dirgrp;
		1066	};
		1067
		1068	$grp
		1069	}
		1070
		1071	=item aio_sync $callback->($status)
		1072
		1073	Asynchronously call sync and call the callback when finished.
465		1074
466	=item aio_fsync $fh, $callback->($status)	1075	=item aio_fsync $fh, $callback->($status)
467		1076
468	Asynchronously call fsync on the given filehandle and call the callback	1077	Asynchronously call fsync on the given filehandle and call the callback
469	with the fsync result code.	1078	with the fsync result code.
…		…
474	callback with the fdatasync result code.	1083	callback with the fdatasync result code.
475		1084
476	If this call isn't available because your OS lacks it or it couldn't be	1085	If this call isn't available because your OS lacks it or it couldn't be
477	detected, it will be emulated by calling C<fsync> instead.	1086	detected, it will be emulated by calling C<fsync> instead.
478		1087
		1088	=item aio_sync_file_range $fh, $offset, $nbytes, $flags, $callback->($status)
		1089
		1090	Sync the data portion of the file specified by C<$offset> and C<$length>
		1091	to disk (but NOT the metadata), by calling the Linux-specific
		1092	sync_file_range call. If sync_file_range is not available or it returns
		1093	ENOSYS, then fdatasync or fsync is being substituted.
		1094
		1095	C<$flags> can be a combination of C<IO::AIO::SYNC_FILE_RANGE_WAIT_BEFORE>,
		1096	C<IO::AIO::SYNC_FILE_RANGE_WRITE> and
		1097	C<IO::AIO::SYNC_FILE_RANGE_WAIT_AFTER>: refer to the sync_file_range
		1098	manpage for details.
		1099
		1100	=item aio_pathsync $path, $callback->($status)
		1101
		1102	This request tries to open, fsync and close the given path. This is a
		1103	composite request intended to sync directories after directory operations
		1104	(E.g. rename). This might not work on all operating systems or have any
		1105	specific effect, but usually it makes sure that directory changes get
		1106	written to disc. It works for anything that can be opened for read-only,
		1107	not just directories.
		1108
		1109	Future versions of this function might fall back to other methods when
		1110	C<fsync> on the directory fails (such as calling C<sync>).
		1111
		1112	Passes C<0> when everything went ok, and C<-1> on error.
		1113
		1114	=cut
		1115
		1116	sub aio_pathsync($;$) {
		1117	my ($path, $cb) = @_;
		1118
		1119	my $pri = aioreq_pri;
		1120	my $grp = aio_group $cb;
		1121
		1122	aioreq_pri $pri;
		1123	add $grp aio_open $path, O_RDONLY, 0, sub {
		1124	my ($fh) = @_;
		1125	if ($fh) {
		1126	aioreq_pri $pri;
		1127	add $grp aio_fsync $fh, sub {
		1128	$grp->result ($_[0]);
		1129
		1130	aioreq_pri $pri;
		1131	add $grp aio_close $fh;
		1132	};
		1133	} else {
		1134	$grp->result (-1);
		1135	}
		1136	};
		1137
		1138	$grp
		1139	}
		1140
		1141	=item aio_msync $scalar, $offset = 0, $length = undef, flags = 0, $callback->($status)
		1142
		1143	This is a rather advanced IO::AIO call, which only works on mmap(2)ed
		1144	scalars (see the C<IO::AIO::mmap> function, although it also works on data
		1145	scalars managed by the L<Sys::Mmap> or L<Mmap> modules, note that the
		1146	scalar must only be modified in-place while an aio operation is pending on
		1147	it).
		1148
		1149	It calls the C<msync> function of your OS, if available, with the memory
		1150	area starting at C<$offset> in the string and ending C<$length> bytes
		1151	later. If C<$length> is negative, counts from the end, and if C<$length>
		1152	is C<undef>, then it goes till the end of the string. The flags can be
		1153	a combination of C<IO::AIO::MS_ASYNC>, C<IO::AIO::MS_INVALIDATE> and
		1154	C<IO::AIO::MS_SYNC>.
		1155
		1156	=item aio_mtouch $scalar, $offset = 0, $length = undef, flags = 0, $callback->($status)
		1157
		1158	This is a rather advanced IO::AIO call, which works best on mmap(2)ed
		1159	scalars.
		1160
		1161	It touches (reads or writes) all memory pages in the specified
		1162	range inside the scalar. All caveats and parameters are the same
		1163	as for C<aio_msync>, above, except for flags, which must be either
		1164	C<0> (which reads all pages and ensures they are instantiated) or
		1165	C<IO::AIO::MT_MODIFY>, which modifies the memory page s(by reading and
		1166	writing an octet from it, which dirties the page).
		1167
		1168	=item aio_mlock $scalar, $offset = 0, $length = undef, $callback->($status)
		1169
		1170	This is a rather advanced IO::AIO call, which works best on mmap(2)ed
		1171	scalars.
		1172
		1173	It reads in all the pages of the underlying storage into memory (if any)
		1174	and locks them, so they are not getting swapped/paged out or removed.
		1175
		1176	If C<$length> is undefined, then the scalar will be locked till the end.
		1177
		1178	On systems that do not implement C<mlock>, this function returns C<-1>
		1179	and sets errno to C<ENOSYS>.
		1180
		1181	Note that the corresponding C<munlock> is synchronous and is
		1182	documented under L<MISCELLANEOUS FUNCTIONS>.
		1183
		1184	Example: open a file, mmap and mlock it - both will be undone when
		1185	C<$data> gets destroyed.
		1186
		1187	open my $fh, "<", $path or die "$path: $!";
		1188	my $data;
		1189	IO::AIO::mmap $data, -s $fh, IO::AIO::PROT_READ, IO::AIO::MAP_SHARED, $fh;
		1190	aio_mlock $data; # mlock in background
		1191
		1192	=item aio_mlockall $flags, $callback->($status)
		1193
		1194	Calls the C<mlockall> function with the given C<$flags> (a combination of
		1195	C<IO::AIO::MCL_CURRENT> and C<IO::AIO::MCL_FUTURE>).
		1196
		1197	On systems that do not implement C<mlockall>, this function returns C<-1>
		1198	and sets errno to C<ENOSYS>.
		1199
		1200	Note that the corresponding C<munlockall> is synchronous and is
		1201	documented under L<MISCELLANEOUS FUNCTIONS>.
		1202
		1203	Example: asynchronously lock all current and future pages into memory.
		1204
		1205	aio_mlockall IO::AIO::MCL_FUTURE;
		1206
		1207	=item aio_group $callback->(...)
		1208
		1209	This is a very special aio request: Instead of doing something, it is a
		1210	container for other aio requests, which is useful if you want to bundle
		1211	many requests into a single, composite, request with a definite callback
		1212	and the ability to cancel the whole request with its subrequests.
		1213
		1214	Returns an object of class L<IO::AIO::GRP>. See its documentation below
		1215	for more info.
		1216
		1217	Example:
		1218
		1219	my $grp = aio_group sub {
		1220	print "all stats done\n";
		1221	};
		1222
		1223	add $grp
		1224	(aio_stat ...),
		1225	(aio_stat ...),
		1226	...;
		1227
		1228	=item aio_nop $callback->()
		1229
		1230	This is a special request - it does nothing in itself and is only used for
		1231	side effects, such as when you want to add a dummy request to a group so
		1232	that finishing the requests in the group depends on executing the given
		1233	code.
		1234
		1235	While this request does nothing, it still goes through the execution
		1236	phase and still requires a worker thread. Thus, the callback will not
		1237	be executed immediately but only after other requests in the queue have
		1238	entered their execution phase. This can be used to measure request
		1239	latency.
		1240
		1241	=item IO::AIO::aio_busy $fractional_seconds, $callback->() *NOT EXPORTED*
		1242
		1243	Mainly used for debugging and benchmarking, this aio request puts one of
		1244	the request workers to sleep for the given time.
		1245
		1246	While it is theoretically handy to have simple I/O scheduling requests
		1247	like sleep and file handle readable/writable, the overhead this creates is
		1248	immense (it blocks a thread for a long time) so do not use this function
		1249	except to put your application under artificial I/O pressure.
		1250
479	=back	1251	=back
480		1252
		1253	=head2 IO::AIO::REQ CLASS
		1254
		1255	All non-aggregate C<aio_*> functions return an object of this class when
		1256	called in non-void context.
		1257
		1258	=over 4
		1259
		1260	=item cancel $req
		1261
		1262	Cancels the request, if possible. Has the effect of skipping execution
		1263	when entering the B<execute> state and skipping calling the callback when
		1264	entering the the B<result> state, but will leave the request otherwise
		1265	untouched (with the exception of readdir). That means that requests that
		1266	currently execute will not be stopped and resources held by the request
		1267	will not be freed prematurely.
		1268
		1269	=item cb $req $callback->(...)
		1270
		1271	Replace (or simply set) the callback registered to the request.
		1272
		1273	=back
		1274
		1275	=head2 IO::AIO::GRP CLASS
		1276
		1277	This class is a subclass of L<IO::AIO::REQ>, so all its methods apply to
		1278	objects of this class, too.
		1279
		1280	A IO::AIO::GRP object is a special request that can contain multiple other
		1281	aio requests.
		1282
		1283	You create one by calling the C<aio_group> constructing function with a
		1284	callback that will be called when all contained requests have entered the
		1285	C<done> state:
		1286
		1287	my $grp = aio_group sub {
		1288	print "all requests are done\n";
		1289	};
		1290
		1291	You add requests by calling the C<add> method with one or more
		1292	C<IO::AIO::REQ> objects:
		1293
		1294	$grp->add (aio_unlink "...");
		1295
		1296	add $grp aio_stat "...", sub {
		1297	$_[0] or return $grp->result ("error");
		1298
		1299	# add another request dynamically, if first succeeded
		1300	add $grp aio_open "...", sub {
		1301	$grp->result ("ok");
		1302	};
		1303	};
		1304
		1305	This makes it very easy to create composite requests (see the source of
		1306	C<aio_move> for an application) that work and feel like simple requests.
		1307
		1308	=over 4
		1309
		1310	=item * The IO::AIO::GRP objects will be cleaned up during calls to
		1311	C<IO::AIO::poll_cb>, just like any other request.
		1312
		1313	=item * They can be canceled like any other request. Canceling will cancel not
		1314	only the request itself, but also all requests it contains.
		1315
		1316	=item * They can also can also be added to other IO::AIO::GRP objects.
		1317
		1318	=item * You must not add requests to a group from within the group callback (or
		1319	any later time).
		1320
		1321	=back
		1322
		1323	Their lifetime, simplified, looks like this: when they are empty, they
		1324	will finish very quickly. If they contain only requests that are in the
		1325	C<done> state, they will also finish. Otherwise they will continue to
		1326	exist.
		1327
		1328	That means after creating a group you have some time to add requests
		1329	(precisely before the callback has been invoked, which is only done within
		1330	the C<poll_cb>). And in the callbacks of those requests, you can add
		1331	further requests to the group. And only when all those requests have
		1332	finished will the the group itself finish.
		1333
		1334	=over 4
		1335
		1336	=item add $grp ...
		1337
		1338	=item $grp->add (...)
		1339
		1340	Add one or more requests to the group. Any type of L<IO::AIO::REQ> can
		1341	be added, including other groups, as long as you do not create circular
		1342	dependencies.
		1343
		1344	Returns all its arguments.
		1345
		1346	=item $grp->cancel_subs
		1347
		1348	Cancel all subrequests and clears any feeder, but not the group request
		1349	itself. Useful when you queued a lot of events but got a result early.
		1350
		1351	The group request will finish normally (you cannot add requests to the
		1352	group).
		1353
		1354	=item $grp->result (...)
		1355
		1356	Set the result value(s) that will be passed to the group callback when all
		1357	subrequests have finished and set the groups errno to the current value
		1358	of errno (just like calling C<errno> without an error number). By default,
		1359	no argument will be passed and errno is zero.
		1360
		1361	=item $grp->errno ([$errno])
		1362
		1363	Sets the group errno value to C<$errno>, or the current value of errno
		1364	when the argument is missing.
		1365
		1366	Every aio request has an associated errno value that is restored when
		1367	the callback is invoked. This method lets you change this value from its
		1368	default (0).
		1369
		1370	Calling C<result> will also set errno, so make sure you either set C<$!>
		1371	before the call to C<result>, or call c<errno> after it.
		1372
		1373	=item feed $grp $callback->($grp)
		1374
		1375	Sets a feeder/generator on this group: every group can have an attached
		1376	generator that generates requests if idle. The idea behind this is that,
		1377	although you could just queue as many requests as you want in a group,
		1378	this might starve other requests for a potentially long time. For example,
		1379	C<aio_scandir> might generate hundreds of thousands C<aio_stat> requests,
		1380	delaying any later requests for a long time.
		1381
		1382	To avoid this, and allow incremental generation of requests, you can
		1383	instead a group and set a feeder on it that generates those requests. The
		1384	feed callback will be called whenever there are few enough (see C<limit>,
		1385	below) requests active in the group itself and is expected to queue more
		1386	requests.
		1387
		1388	The feed callback can queue as many requests as it likes (i.e. C<add> does
		1389	not impose any limits).
		1390
		1391	If the feed does not queue more requests when called, it will be
		1392	automatically removed from the group.
		1393
		1394	If the feed limit is C<0> when this method is called, it will be set to
		1395	C<2> automatically.
		1396
		1397	Example:
		1398
		1399	# stat all files in @files, but only ever use four aio requests concurrently:
		1400
		1401	my $grp = aio_group sub { print "finished\n" };
		1402	limit $grp 4;
		1403	feed $grp sub {
		1404	my $file = pop @files
		1405	or return;
		1406
		1407	add $grp aio_stat $file, sub { ... };
		1408	};
		1409
		1410	=item limit $grp $num
		1411
		1412	Sets the feeder limit for the group: The feeder will be called whenever
		1413	the group contains less than this many requests.
		1414
		1415	Setting the limit to C<0> will pause the feeding process.
		1416
		1417	The default value for the limit is C<0>, but note that setting a feeder
		1418	automatically bumps it up to C<2>.
		1419
		1420	=back
		1421
481	=head2 SUPPORT FUNCTIONS	1422	=head2 SUPPORT FUNCTIONS
482		1423
		1424	=head3 EVENT PROCESSING AND EVENT LOOP INTEGRATION
		1425
483	=over 4	1426	=over 4
484		1427
485	=item $fileno = IO::AIO::poll_fileno	1428	=item $fileno = IO::AIO::poll_fileno
486		1429
487	Return the I<request result pipe file descriptor>. This filehandle must be	1430	Return the I<request result pipe file descriptor>. This filehandle must be
488	polled for reading by some mechanism outside this module (e.g. Event or	1431	polled for reading by some mechanism outside this module (e.g. EV, Glib,
489	select, see below or the SYNOPSIS). If the pipe becomes readable you have	1432	select and so on, see below or the SYNOPSIS). If the pipe becomes readable
490	to call C<poll_cb> to check the results.	1433	you have to call C<poll_cb> to check the results.
491		1434
492	See C<poll_cb> for an example.	1435	See C<poll_cb> for an example.
493		1436
494	=item IO::AIO::poll_cb	1437	=item IO::AIO::poll_cb
495		1438
496	Process all outstanding events on the result pipe. You have to call this	1439	Process some outstanding events on the result pipe. You have to call
497	regularly. Returns the number of events processed. Returns immediately	1440	this regularly. Returns C<0> if all events could be processed (or there
498	when no events are outstanding.	1441	were no events to process), or C<-1> if it returned earlier for whatever
		1442	reason. Returns immediately when no events are outstanding. The amount of
		1443	events processed depends on the settings of C<IO::AIO::max_poll_req> and
		1444	C<IO::AIO::max_poll_time>.
		1445
		1446	If not all requests were processed for whatever reason, the filehandle
		1447	will still be ready when C<poll_cb> returns, so normally you don't have to
		1448	do anything special to have it called later.
		1449
		1450	Apart from calling C<IO::AIO::poll_cb> when the event filehandle becomes
		1451	ready, it can be beneficial to call this function from loops which submit
		1452	a lot of requests, to make sure the results get processed when they become
		1453	available and not just when the loop is finished and the event loop takes
		1454	over again. This function returns very fast when there are no outstanding
		1455	requests.
499		1456
500	Example: Install an Event watcher that automatically calls	1457	Example: Install an Event watcher that automatically calls
501	IO::AIO::poll_cb with high priority:	1458	IO::AIO::poll_cb with high priority (more examples can be found in the
		1459	SYNOPSIS section, at the top of this document):
502		1460
503	Event->io (fd => IO::AIO::poll_fileno,	1461	Event->io (fd => IO::AIO::poll_fileno,
504	poll => 'r', async => 1,	1462	poll => 'r', async => 1,
505	cb => \&IO::AIO::poll_cb);	1463	cb => \&IO::AIO::poll_cb);
506		1464
507	=item IO::AIO::poll_wait	1465	=item IO::AIO::poll_wait
508		1466
		1467	If there are any outstanding requests and none of them in the result
509	Wait till the result filehandle becomes ready for reading (simply does a	1468	phase, wait till the result filehandle becomes ready for reading (simply
510	C<select> on the filehandle. This is useful if you want to synchronously wait	1469	does a C<select> on the filehandle. This is useful if you want to
511	for some requests to finish).	1470	synchronously wait for some requests to finish).
512		1471
513	See C<nreqs> for an example.	1472	See C<nreqs> for an example.
514		1473
		1474	=item IO::AIO::poll
		1475
		1476	Waits until some requests have been handled.
		1477
		1478	Returns the number of requests processed, but is otherwise strictly
		1479	equivalent to:
		1480
		1481	IO::AIO::poll_wait, IO::AIO::poll_cb
		1482
515	=item IO::AIO::nreqs	1483	=item IO::AIO::flush
516		1484
517	Returns the number of requests currently outstanding (i.e. for which their	1485	Wait till all outstanding AIO requests have been handled.
518	callback has not been invoked yet).
519		1486
520	Example: wait till there are no outstanding requests anymore:	1487	Strictly equivalent to:
521		1488
522	IO::AIO::poll_wait, IO::AIO::poll_cb	1489	IO::AIO::poll_wait, IO::AIO::poll_cb
523	while IO::AIO::nreqs;	1490	while IO::AIO::nreqs;
524		1491
525	=item IO::AIO::flush	1492	=item IO::AIO::max_poll_reqs $nreqs
526		1493
527	Wait till all outstanding AIO requests have been handled.	1494	=item IO::AIO::max_poll_time $seconds
528		1495
529	Strictly equivalent to:	1496	These set the maximum number of requests (default C<0>, meaning infinity)
		1497	that are being processed by C<IO::AIO::poll_cb> in one call, respectively
		1498	the maximum amount of time (default C<0>, meaning infinity) spent in
		1499	C<IO::AIO::poll_cb> to process requests (more correctly the mininum amount
		1500	of time C<poll_cb> is allowed to use).
530		1501
531	IO::AIO::poll_wait, IO::AIO::poll_cb	1502	Setting C<max_poll_time> to a non-zero value creates an overhead of one
532	while IO::AIO::nreqs;	1503	syscall per request processed, which is not normally a problem unless your
		1504	callbacks are really really fast or your OS is really really slow (I am
		1505	not mentioning Solaris here). Using C<max_poll_reqs> incurs no overhead.
533		1506
534	=item IO::AIO::poll	1507	Setting these is useful if you want to ensure some level of
		1508	interactiveness when perl is not fast enough to process all requests in
		1509	time.
535		1510
536	Waits until some requests have been handled.	1511	For interactive programs, values such as C<0.01> to C<0.1> should be fine.
537		1512
538	Strictly equivalent to:	1513	Example: Install an Event watcher that automatically calls
		1514	IO::AIO::poll_cb with low priority, to ensure that other parts of the
		1515	program get the CPU sometimes even under high AIO load.
539		1516
540	IO::AIO::poll_wait, IO::AIO::poll_cb	1517	# try not to spend much more than 0.1s in poll_cb
541	if IO::AIO::nreqs;	1518	IO::AIO::max_poll_time 0.1;
		1519
		1520	# use a low priority so other tasks have priority
		1521	Event->io (fd => IO::AIO::poll_fileno,
		1522	poll => 'r', nice => 1,
		1523	cb => &IO::AIO::poll_cb);
		1524
		1525	=back
		1526
		1527	=head3 CONTROLLING THE NUMBER OF THREADS
		1528
		1529	=over
542		1530
543	=item IO::AIO::min_parallel $nthreads	1531	=item IO::AIO::min_parallel $nthreads
544		1532
545	Set the minimum number of AIO threads to C<$nthreads>. The current default	1533	Set the minimum number of AIO threads to C<$nthreads>. The current
546	is C<4>, which means four asynchronous operations can be done at one time	1534	default is C<8>, which means eight asynchronous operations can execute
547	(the number of outstanding operations, however, is unlimited).	1535	concurrently at any one time (the number of outstanding requests,
		1536	however, is unlimited).
548		1537
549	IO::AIO starts threads only on demand, when an AIO request is queued and	1538	IO::AIO starts threads only on demand, when an AIO request is queued and
550	no free thread exists.	1539	no free thread exists. Please note that queueing up a hundred requests can
		1540	create demand for a hundred threads, even if it turns out that everything
		1541	is in the cache and could have been processed faster by a single thread.
551		1542
552	It is recommended to keep the number of threads low, as some Linux	1543	It is recommended to keep the number of threads relatively low, as some
553	kernel versions will scale negatively with the number of threads (higher	1544	Linux kernel versions will scale negatively with the number of threads
554	parallelity => MUCH higher latency). With current Linux 2.6 versions, 4-32	1545	(higher parallelity => MUCH higher latency). With current Linux 2.6
555	threads should be fine.	1546	versions, 4-32 threads should be fine.
556		1547
557	Under most circumstances you don't need to call this function, as the	1548	Under most circumstances you don't need to call this function, as the
558	module selects a default that is suitable for low to moderate load.	1549	module selects a default that is suitable for low to moderate load.
559		1550
560	=item IO::AIO::max_parallel $nthreads	1551	=item IO::AIO::max_parallel $nthreads
…		…
569	This module automatically runs C<max_parallel 0> at program end, to ensure	1560	This module automatically runs C<max_parallel 0> at program end, to ensure
570	that all threads are killed and that there are no outstanding requests.	1561	that all threads are killed and that there are no outstanding requests.
571		1562
572	Under normal circumstances you don't need to call this function.	1563	Under normal circumstances you don't need to call this function.
573		1564
		1565	=item IO::AIO::max_idle $nthreads
		1566
		1567	Limit the number of threads (default: 4) that are allowed to idle
		1568	(i.e., threads that did not get a request to process within the idle
		1569	timeout (default: 10 seconds). That means if a thread becomes idle while
		1570	C<$nthreads> other threads are also idle, it will free its resources and
		1571	exit.
		1572
		1573	This is useful when you allow a large number of threads (e.g. 100 or 1000)
		1574	to allow for extremely high load situations, but want to free resources
		1575	under normal circumstances (1000 threads can easily consume 30MB of RAM).
		1576
		1577	The default is probably ok in most situations, especially if thread
		1578	creation is fast. If thread creation is very slow on your system you might
		1579	want to use larger values.
		1580
		1581	=item IO::AIO::idle_timeout $seconds
		1582
		1583	Sets the minimum idle timeout (default 10) after which worker threads are
		1584	allowed to exit. SEe C<IO::AIO::max_idle>.
		1585
574	=item $oldnreqs = IO::AIO::max_outstanding $nreqs	1586	=item IO::AIO::max_outstanding $maxreqs
575		1587
576	Sets the maximum number of outstanding requests to C<$nreqs>. If you	1588	Sets the maximum number of outstanding requests to C<$nreqs>. If
577	try to queue up more than this number of requests, the caller will block until	1589	you do queue up more than this number of requests, the next call to
578	some requests have been handled.	1590	C<IO::AIO::poll_cb> (and other functions calling C<poll_cb>, such as
		1591	C<IO::AIO::flush> or C<IO::AIO::poll>) will block until the limit is no
		1592	longer exceeded.
579		1593
580	The default is very large, so normally there is no practical limit. If you	1594	In other words, this setting does not enforce a queue limit, but can be
581	queue up many requests in a loop it often improves speed if you set	1595	used to make poll functions block if the limit is exceeded.
582	this to a relatively low number, such as C<100>.
583		1596
584	Under normal circumstances you don't need to call this function.	1597	This is a very bad function to use in interactive programs because it
		1598	blocks, and a bad way to reduce concurrency because it is inexact: Better
		1599	use an C<aio_group> together with a feed callback.
		1600
		1601	It's main use is in scripts without an event loop - when you want to stat
		1602	a lot of files, you can write somehting like this:
		1603
		1604	IO::AIO::max_outstanding 32;
		1605
		1606	for my $path (...) {
		1607	aio_stat $path , ...;
		1608	IO::AIO::poll_cb;
		1609	}
		1610
		1611	IO::AIO::flush;
		1612
		1613	The call to C<poll_cb> inside the loop will normally return instantly, but
		1614	as soon as more thna C<32> reqeusts are in-flight, it will block until
		1615	some requests have been handled. This keeps the loop from pushing a large
		1616	number of C<aio_stat> requests onto the queue.
		1617
		1618	The default value for C<max_outstanding> is very large, so there is no
		1619	practical limit on the number of outstanding requests.
585		1620
586	=back	1621	=back
587		1622
		1623	=head3 STATISTICAL INFORMATION
		1624
		1625	=over
		1626
		1627	=item IO::AIO::nreqs
		1628
		1629	Returns the number of requests currently in the ready, execute or pending
		1630	states (i.e. for which their callback has not been invoked yet).
		1631
		1632	Example: wait till there are no outstanding requests anymore:
		1633
		1634	IO::AIO::poll_wait, IO::AIO::poll_cb
		1635	while IO::AIO::nreqs;
		1636
		1637	=item IO::AIO::nready
		1638
		1639	Returns the number of requests currently in the ready state (not yet
		1640	executed).
		1641
		1642	=item IO::AIO::npending
		1643
		1644	Returns the number of requests currently in the pending state (executed,
		1645	but not yet processed by poll_cb).
		1646
		1647	=back
		1648
		1649	=head3 MISCELLANEOUS FUNCTIONS
		1650
		1651	IO::AIO implements some functions that might be useful, but are not
		1652	asynchronous.
		1653
		1654	=over 4
		1655
		1656	=item IO::AIO::sendfile $ofh, $ifh, $offset, $count
		1657
		1658	Calls the C<eio_sendfile_sync> function, which is like C<aio_sendfile>,
		1659	but is blocking (this makes most sense if you know the input data is
		1660	likely cached already and the output filehandle is set to non-blocking
		1661	operations).
		1662
		1663	Returns the number of bytes copied, or C<-1> on error.
		1664
		1665	=item IO::AIO::fadvise $fh, $offset, $len, $advice
		1666
		1667	Simply calls the C<posix_fadvise> function (see its
		1668	manpage for details). The following advice constants are
		1669	avaiable: C<IO::AIO::FADV_NORMAL>, C<IO::AIO::FADV_SEQUENTIAL>,
		1670	C<IO::AIO::FADV_RANDOM>, C<IO::AIO::FADV_NOREUSE>,
		1671	C<IO::AIO::FADV_WILLNEED>, C<IO::AIO::FADV_DONTNEED>.
		1672
		1673	On systems that do not implement C<posix_fadvise>, this function returns
		1674	ENOSYS, otherwise the return value of C<posix_fadvise>.
		1675
		1676	=item IO::AIO::madvise $scalar, $offset, $len, $advice
		1677
		1678	Simply calls the C<posix_madvise> function (see its
		1679	manpage for details). The following advice constants are
		1680	avaiable: C<IO::AIO::MADV_NORMAL>, C<IO::AIO::MADV_SEQUENTIAL>,
		1681	C<IO::AIO::MADV_RANDOM>, C<IO::AIO::MADV_WILLNEED>, C<IO::AIO::MADV_DONTNEED>.
		1682
		1683	On systems that do not implement C<posix_madvise>, this function returns
		1684	ENOSYS, otherwise the return value of C<posix_madvise>.
		1685
		1686	=item IO::AIO::mprotect $scalar, $offset, $len, $protect
		1687
		1688	Simply calls the C<mprotect> function on the preferably AIO::mmap'ed
		1689	$scalar (see its manpage for details). The following protect
		1690	constants are avaiable: C<IO::AIO::PROT_NONE>, C<IO::AIO::PROT_READ>,
		1691	C<IO::AIO::PROT_WRITE>, C<IO::AIO::PROT_EXEC>.
		1692
		1693	On systems that do not implement C<mprotect>, this function returns
		1694	ENOSYS, otherwise the return value of C<mprotect>.
		1695
		1696	=item IO::AIO::mmap $scalar, $length, $prot, $flags, $fh[, $offset]
		1697
		1698	Memory-maps a file (or anonymous memory range) and attaches it to the
		1699	given C<$scalar>, which will act like a string scalar.
		1700
		1701	The only operations allowed on the scalar are C<substr>/C<vec> that don't
		1702	change the string length, and most read-only operations such as copying it
		1703	or searching it with regexes and so on.
		1704
		1705	Anything else is unsafe and will, at best, result in memory leaks.
		1706
		1707	The memory map associated with the C<$scalar> is automatically removed
		1708	when the C<$scalar> is destroyed, or when the C<IO::AIO::mmap> or
		1709	C<IO::AIO::munmap> functions are called.
		1710
		1711	This calls the C<mmap>(2) function internally. See your system's manual
		1712	page for details on the C<$length>, C<$prot> and C<$flags> parameters.
		1713
		1714	The C<$length> must be larger than zero and smaller than the actual
		1715	filesize.
		1716
		1717	C<$prot> is a combination of C<IO::AIO::PROT_NONE>, C<IO::AIO::PROT_EXEC>,
		1718	C<IO::AIO::PROT_READ> and/or C<IO::AIO::PROT_WRITE>,
		1719
		1720	C<$flags> can be a combination of C<IO::AIO::MAP_SHARED> or
		1721	C<IO::AIO::MAP_PRIVATE>, or a number of system-specific flags (when
		1722	not available, the are defined as 0): C<IO::AIO::MAP_ANONYMOUS>
		1723	(which is set to C<MAP_ANON> if your system only provides this
		1724	constant), C<IO::AIO::MAP_HUGETLB>, C<IO::AIO::MAP_LOCKED>,
		1725	C<IO::AIO::MAP_NORESERVE>, C<IO::AIO::MAP_POPULATE> or
		1726	C<IO::AIO::MAP_NONBLOCK>
		1727
		1728	If C<$fh> is C<undef>, then a file descriptor of C<-1> is passed.
		1729
		1730	C<$offset> is the offset from the start of the file - it generally must be
		1731	a multiple of C<IO::AIO::PAGESIZE> and defaults to C<0>.
		1732
		1733	Example:
		1734
		1735	use Digest::MD5;
		1736	use IO::AIO;
		1737
		1738	open my $fh, "<verybigfile"
		1739	or die "$!";
		1740
		1741	IO::AIO::mmap my $data, -s $fh, IO::AIO::PROT_READ, IO::AIO::MAP_SHARED, $fh
		1742	or die "verybigfile: $!";
		1743
		1744	my $fast_md5 = md5 $data;
		1745
		1746	=item IO::AIO::munmap $scalar
		1747
		1748	Removes a previous mmap and undefines the C<$scalar>.
		1749
		1750	=item IO::AIO::munlock $scalar, $offset = 0, $length = undef
		1751
		1752	Calls the C<munlock> function, undoing the effects of a previous
		1753	C<aio_mlock> call (see its description for details).
		1754
		1755	=item IO::AIO::munlockall
		1756
		1757	Calls the C<munlockall> function.
		1758
		1759	On systems that do not implement C<munlockall>, this function returns
		1760	ENOSYS, otherwise the return value of C<munlockall>.
		1761
		1762	=back
		1763
588	=cut	1764	=cut
589		1765
590	# support function to convert a fd into a perl filehandle
591	sub _fd2fh {
592	return undef if $_[0] < 0;
593
594	# try to generate nice filehandles
595	my $sym = "IO::AIO::fd#$_[0]";
596	local *$sym;
597
598	open *$sym, "+<&=$_[0]" # usually works under any unix
599	or open *$sym, "<&=$_[0]" # cygwin needs this
600	or open *$sym, ">&=$_[0]" # or this
601	or return undef;
602
603	*$sym
604	}
605
606	min_parallel 4;	1766	min_parallel 8;
607		1767
608	END {	1768	END { flush }
609	max_parallel 0;
610	}
611		1769
612	1;	1770	1;
613		1771
		1772	=head1 EVENT LOOP INTEGRATION
		1773
		1774	It is recommended to use L<AnyEvent::AIO> to integrate IO::AIO
		1775	automatically into many event loops:
		1776
		1777	# AnyEvent integration (EV, Event, Glib, Tk, POE, urxvt, pureperl...)
		1778	use AnyEvent::AIO;
		1779
		1780	You can also integrate IO::AIO manually into many event loops, here are
		1781	some examples of how to do this:
		1782
		1783	# EV integration
		1784	my $aio_w = EV::io IO::AIO::poll_fileno, EV::READ, \&IO::AIO::poll_cb;
		1785
		1786	# Event integration
		1787	Event->io (fd => IO::AIO::poll_fileno,
		1788	poll => 'r',
		1789	cb => \&IO::AIO::poll_cb);
		1790
		1791	# Glib/Gtk2 integration
		1792	add_watch Glib::IO IO::AIO::poll_fileno,
		1793	in => sub { IO::AIO::poll_cb; 1 };
		1794
		1795	# Tk integration
		1796	Tk::Event::IO->fileevent (IO::AIO::poll_fileno, "",
		1797	readable => \&IO::AIO::poll_cb);
		1798
		1799	# Danga::Socket integration
		1800	Danga::Socket->AddOtherFds (IO::AIO::poll_fileno =>
		1801	\&IO::AIO::poll_cb);
		1802
614	=head2 FORK BEHAVIOUR	1803	=head2 FORK BEHAVIOUR
615		1804
616	Before the fork, IO::AIO enters a quiescent state where no requests	1805	Usage of pthreads in a program changes the semantics of fork
617	can be added in other threads and no results will be processed. After	1806	considerably. Specifically, only async-safe functions can be called after
618	the fork the parent simply leaves the quiescent state and continues	1807	fork. Perl doesn't know about this, so in general, you cannot call fork
619	request/result processing, while the child clears the request/result	1808	with defined behaviour in perl. IO::AIO uses pthreads, so this applies,
620	queue (so the requests started before the fork will only be handled in	1809	but many other extensions and (for inexplicable reasons) perl itself often
621	the parent). Threats will be started on demand until the limit ste in the	1810	is linked against pthreads, so this limitation applies.
622	parent process has been reached again.	1811
		1812	Some operating systems have extensions that allow safe use of fork, and
		1813	this module should do "the right thing" on those, and tries on others. At
		1814	the time of this writing (2011) only GNU/Linux supports these extensions
		1815	to POSIX.
		1816
		1817	=head2 MEMORY USAGE
		1818
		1819	Per-request usage:
		1820
		1821	Each aio request uses - depending on your architecture - around 100-200
		1822	bytes of memory. In addition, stat requests need a stat buffer (possibly
		1823	a few hundred bytes), readdir requires a result buffer and so on. Perl
		1824	scalars and other data passed into aio requests will also be locked and
		1825	will consume memory till the request has entered the done state.
		1826
		1827	This is not awfully much, so queuing lots of requests is not usually a
		1828	problem.
		1829
		1830	Per-thread usage:
		1831
		1832	In the execution phase, some aio requests require more memory for
		1833	temporary buffers, and each thread requires a stack and other data
		1834	structures (usually around 16k-128k, depending on the OS).
		1835
		1836	=head1 KNOWN BUGS
		1837
		1838	Known bugs will be fixed in the next release.
623		1839
624	=head1 SEE ALSO	1840	=head1 SEE ALSO
625		1841
626	L<Coro>, L<Linux::AIO>.	1842	L<AnyEvent::AIO> for easy integration into event loops, L<Coro::AIO> for a
		1843	more natural syntax.
627		1844
628	=head1 AUTHOR	1845	=head1 AUTHOR
629		1846
630	Marc Lehmann <schmorp@schmorp.de>	1847	Marc Lehmann <schmorp@schmorp.de>
631	http://home.schmorp.de/	1848	http://home.schmorp.de/

Diff Legend

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