1AIO(3)                User Contributed Perl Documentation               AIO(3)
2
3
4

NAME

6       IO::AIO - Asynchronous/Advanced Input/Output
7

SYNOPSIS

9        use IO::AIO;
10
11        aio_open "/etc/passwd", IO::AIO::O_RDONLY, 0, sub {
12           my $fh = shift
13              or die "/etc/passwd: $!";
14           ...
15        };
16
17        aio_unlink "/tmp/file", sub { };
18
19        aio_read $fh, 30000, 1024, $buffer, 0, sub {
20           $_[0] > 0 or die "read error: $!";
21        };
22
23        # version 2+ has request and group objects
24        use IO::AIO 2;
25
26        aioreq_pri 4; # give next request a very high priority
27        my $req = aio_unlink "/tmp/file", sub { };
28        $req->cancel; # cancel request if still in queue
29
30        my $grp = aio_group sub { print "all stats done\n" };
31        add $grp aio_stat "..." for ...;
32

DESCRIPTION

34       This module implements asynchronous I/O using whatever means your
35       operating system supports. It is implemented as an interface to
36       "libeio" (<http://software.schmorp.de/pkg/libeio.html>).
37
38       Asynchronous means that operations that can normally block your program
39       (e.g. reading from disk) will be done asynchronously: the operation
40       will still block, but you can do something else in the meantime. This
41       is extremely useful for programs that need to stay interactive even
42       when doing heavy I/O (GUI programs, high performance network servers
43       etc.), but can also be used to easily do operations in parallel that
44       are normally done sequentially, e.g. stat'ing many files, which is much
45       faster on a RAID volume or over NFS when you do a number of stat
46       operations concurrently.
47
48       While most of this works on all types of file descriptors (for example
49       sockets), using these functions on file descriptors that support
50       nonblocking operation (again, sockets, pipes etc.) is very inefficient.
51       Use an event loop for that (such as the EV module): IO::AIO will
52       naturally fit into such an event loop itself.
53
54       In this version, a number of threads are started that execute your
55       requests and signal their completion. You don't need thread support in
56       perl, and the threads created by this module will not be visible to
57       perl. In the future, this module might make use of the native aio
58       functions available on many operating systems. However, they are often
59       not well-supported or restricted (GNU/Linux doesn't allow them on
60       normal files currently, for example), and they would only support
61       aio_read and aio_write, so the remaining functionality would have to be
62       implemented using threads anyway.
63
64       In addition to asynchronous I/O, this module also exports some rather
65       arcane interfaces, such as "madvise" or linux's "splice" system call,
66       which is why the "A" in "AIO" can also mean advanced.
67
68       Although the module will work in the presence of other (Perl-) threads,
69       it is currently not reentrant in any way, so use appropriate locking
70       yourself, always call "poll_cb" from within the same thread, or never
71       call "poll_cb" (or other "aio_" functions) recursively.
72
73   EXAMPLE
74       This is a simple example that uses the EV module and loads /etc/passwd
75       asynchronously:
76
77          use EV;
78          use IO::AIO;
79
80          # register the IO::AIO callback with EV
81          my $aio_w = EV::io IO::AIO::poll_fileno, EV::READ, \&IO::AIO::poll_cb;
82
83          # queue the request to open /etc/passwd
84          aio_open "/etc/passwd", IO::AIO::O_RDONLY, 0, sub {
85             my $fh = shift
86                or die "error while opening: $!";
87
88             # stat'ing filehandles is generally non-blocking
89             my $size = -s $fh;
90
91             # queue a request to read the file
92             my $contents;
93             aio_read $fh, 0, $size, $contents, 0, sub {
94                $_[0] == $size
95                   or die "short read: $!";
96
97                close $fh;
98
99                # file contents now in $contents
100                print $contents;
101
102                # exit event loop and program
103                EV::break;
104             };
105          };
106
107          # possibly queue up other requests, or open GUI windows,
108          # check for sockets etc. etc.
109
110          # process events as long as there are some:
111          EV::run;
112

REQUEST ANATOMY AND LIFETIME

114       Every "aio_*" function creates a request. which is a C data structure
115       not directly visible to Perl.
116
117       If called in non-void context, every request function returns a Perl
118       object representing the request. In void context, nothing is returned,
119       which saves a bit of memory.
120
121       The perl object is a fairly standard ref-to-hash object. The hash
122       contents are not used by IO::AIO so you are free to store anything you
123       like in it.
124
125       During their existance, aio requests travel through the following
126       states, in order:
127
128       ready
129           Immediately after a request is created it is put into the ready
130           state, waiting for a thread to execute it.
131
132       execute
133           A thread has accepted the request for processing and is currently
134           executing it (e.g. blocking in read).
135
136       pending
137           The request has been executed and is waiting for result processing.
138
139           While request submission and execution is fully asynchronous,
140           result processing is not and relies on the perl interpreter calling
141           "poll_cb" (or another function with the same effect).
142
143       result
144           The request results are processed synchronously by "poll_cb".
145
146           The "poll_cb" function will process all outstanding aio requests by
147           calling their callbacks, freeing memory associated with them and
148           managing any groups they are contained in.
149
150       done
151           Request has reached the end of its lifetime and holds no resources
152           anymore (except possibly for the Perl object, but its connection to
153           the actual aio request is severed and calling its methods will
154           either do nothing or result in a runtime error).
155

FUNCTIONS

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

EVENT LOOP INTEGRATION

2064       It is recommended to use AnyEvent::AIO to integrate IO::AIO
2065       automatically into many event loops:
2066
2067        # AnyEvent integration (EV, Event, Glib, Tk, POE, urxvt, pureperl...)
2068        use AnyEvent::AIO;
2069
2070       You can also integrate IO::AIO manually into many event loops, here are
2071       some examples of how to do this:
2072
2073        # EV integration
2074        my $aio_w = EV::io IO::AIO::poll_fileno, EV::READ, \&IO::AIO::poll_cb;
2075
2076        # Event integration
2077        Event->io (fd => IO::AIO::poll_fileno,
2078                   poll => 'r',
2079                   cb => \&IO::AIO::poll_cb);
2080
2081        # Glib/Gtk2 integration
2082        add_watch Glib::IO IO::AIO::poll_fileno,
2083                  in => sub { IO::AIO::poll_cb; 1 };
2084
2085        # Tk integration
2086        Tk::Event::IO->fileevent (IO::AIO::poll_fileno, "",
2087                                  readable => \&IO::AIO::poll_cb);
2088
2089        # Danga::Socket integration
2090        Danga::Socket->AddOtherFds (IO::AIO::poll_fileno =>
2091                                    \&IO::AIO::poll_cb);
2092
2093   FORK BEHAVIOUR
2094       Usage of pthreads in a program changes the semantics of fork
2095       considerably. Specifically, only async-safe functions can be called
2096       after fork. Perl doesn't know about this, so in general, you cannot
2097       call fork with defined behaviour in perl if pthreads are involved.
2098       IO::AIO uses pthreads, so this applies, but many other extensions and
2099       (for inexplicable reasons) perl itself often is linked against
2100       pthreads, so this limitation applies to quite a lot of perls.
2101
2102       This module no longer tries to fight your OS, or POSIX. That means
2103       IO::AIO only works in the process that loaded it. Forking is fully
2104       supported, but using IO::AIO in the child is not.
2105
2106       You might get around by not using IO::AIO before (or after) forking.
2107       You could also try to call the IO::AIO::reinit function in the child:
2108
2109       IO::AIO::reinit
2110           Abandons all current requests and I/O threads and simply
2111           reinitialises all data structures. This is not an operation
2112           supported by any standards, but happens to work on GNU/Linux and
2113           some newer BSD systems.
2114
2115           The only reasonable use for this function is to call it after
2116           forking, if "IO::AIO" was used in the parent. Calling it while
2117           IO::AIO is active in the process will result in undefined
2118           behaviour. Calling it at any time will also result in any undefined
2119           (by POSIX) behaviour.
2120
2121   LINUX-SPECIFIC CALLS
2122       When a call is documented as "linux-specific" then this means it
2123       originated on GNU/Linux. "IO::AIO" will usually try to autodetect the
2124       availability and compatibility of such calls regardless of the platform
2125       it is compiled on, so platforms such as FreeBSD which often implement
2126       these calls will work. When in doubt, call them and see if they fail
2127       wth "ENOSYS".
2128
2129   MEMORY USAGE
2130       Per-request usage:
2131
2132       Each aio request uses - depending on your architecture - around 100-200
2133       bytes of memory. In addition, stat requests need a stat buffer
2134       (possibly a few hundred bytes), readdir requires a result buffer and so
2135       on. Perl scalars and other data passed into aio requests will also be
2136       locked and will consume memory till the request has entered the done
2137       state.
2138
2139       This is not awfully much, so queuing lots of requests is not usually a
2140       problem.
2141
2142       Per-thread usage:
2143
2144       In the execution phase, some aio requests require more memory for
2145       temporary buffers, and each thread requires a stack and other data
2146       structures (usually around 16k-128k, depending on the OS).
2147

KNOWN BUGS

2149       Known bugs will be fixed in the next release :)
2150

KNOWN ISSUES

2152       Calls that try to "import" foreign memory areas (such as
2153       "IO::AIO::mmap" or "IO::AIO::aio_slurp") do not work with generic
2154       lvalues, such as non-created hash slots or other scalars I didn't think
2155       of. It's best to avoid such and either use scalar variables or making
2156       sure that the scalar exists (e.g. by storing "undef") and isn't "funny"
2157       (e.g. tied).
2158
2159       I am not sure anything can be done about this, so this is considered a
2160       known issue, rather than a bug.
2161

SEE ALSO

2163       AnyEvent::AIO for easy integration into event loops, Coro::AIO for a
2164       more natural syntax.
2165

AUTHOR

2167        Marc Lehmann <schmorp@schmorp.de>
2168        http://home.schmorp.de/
2169
2170
2171
2172perl v5.28.1                      2019-04-03                            AIO(3)
Impressum