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

EVENT LOOP INTEGRATION

2015       It is recommended to use AnyEvent::AIO to integrate IO::AIO
2016       automatically into many event loops:
2017
2018        # AnyEvent integration (EV, Event, Glib, Tk, POE, urxvt, pureperl...)
2019        use AnyEvent::AIO;
2020
2021       You can also integrate IO::AIO manually into many event loops, here are
2022       some examples of how to do this:
2023
2024        # EV integration
2025        my $aio_w = EV::io IO::AIO::poll_fileno, EV::READ, \&IO::AIO::poll_cb;
2026
2027        # Event integration
2028        Event->io (fd => IO::AIO::poll_fileno,
2029                   poll => 'r',
2030                   cb => \&IO::AIO::poll_cb);
2031
2032        # Glib/Gtk2 integration
2033        add_watch Glib::IO IO::AIO::poll_fileno,
2034                  in => sub { IO::AIO::poll_cb; 1 };
2035
2036        # Tk integration
2037        Tk::Event::IO->fileevent (IO::AIO::poll_fileno, "",
2038                                  readable => \&IO::AIO::poll_cb);
2039
2040        # Danga::Socket integration
2041        Danga::Socket->AddOtherFds (IO::AIO::poll_fileno =>
2042                                    \&IO::AIO::poll_cb);
2043
2044   FORK BEHAVIOUR
2045       Usage of pthreads in a program changes the semantics of fork
2046       considerably. Specifically, only async-safe functions can be called
2047       after fork. Perl doesn't know about this, so in general, you cannot
2048       call fork with defined behaviour in perl if pthreads are involved.
2049       IO::AIO uses pthreads, so this applies, but many other extensions and
2050       (for inexplicable reasons) perl itself often is linked against
2051       pthreads, so this limitation applies to quite a lot of perls.
2052
2053       This module no longer tries to fight your OS, or POSIX. That means
2054       IO::AIO only works in the process that loaded it. Forking is fully
2055       supported, but using IO::AIO in the child is not.
2056
2057       You might get around by not using IO::AIO before (or after) forking.
2058       You could also try to call the IO::AIO::reinit function in the child:
2059
2060       IO::AIO::reinit
2061           Abandons all current requests and I/O threads and simply
2062           reinitialises all data structures. This is not an operation
2063           supported by any standards, but happens to work on GNU/Linux and
2064           some newer BSD systems.
2065
2066           The only reasonable use for this function is to call it after
2067           forking, if "IO::AIO" was used in the parent. Calling it while
2068           IO::AIO is active in the process will result in undefined
2069           behaviour. Calling it at any time will also result in any undefined
2070           (by POSIX) behaviour.
2071
2072   LINUX-SPECIFIC CALLS
2073       When a call is documented as "linux-specific" then this means it
2074       originated on GNU/Linux. "IO::AIO" will usually try to autodetect the
2075       availability and compatibility of such calls regardless of the platform
2076       it is compiled on, so platforms such as FreeBSD which often implement
2077       these calls will work. When in doubt, call them and see if they fail
2078       wth "ENOSYS".
2079
2080   MEMORY USAGE
2081       Per-request usage:
2082
2083       Each aio request uses - depending on your architecture - around 100-200
2084       bytes of memory. In addition, stat requests need a stat buffer
2085       (possibly a few hundred bytes), readdir requires a result buffer and so
2086       on. Perl scalars and other data passed into aio requests will also be
2087       locked and will consume memory till the request has entered the done
2088       state.
2089
2090       This is not awfully much, so queuing lots of requests is not usually a
2091       problem.
2092
2093       Per-thread usage:
2094
2095       In the execution phase, some aio requests require more memory for
2096       temporary buffers, and each thread requires a stack and other data
2097       structures (usually around 16k-128k, depending on the OS).
2098

KNOWN BUGS

2100       Known bugs will be fixed in the next release :)
2101

KNOWN ISSUES

2103       Calls that try to "import" foreign memory areas (such as
2104       "IO::AIO::mmap" or "IO::AIO::aio_slurp") do not work with generic
2105       lvalues, such as non-created hash slots or other scalars I didn't think
2106       of. It's best to avoid such and either use scalar variables or making
2107       sure that the scalar exists (e.g. by storing "undef") and isn't "funny"
2108       (e.g. tied).
2109
2110       I am not sure anything can be done about this, so this is considered a
2111       known issue, rather than a bug.
2112

SEE ALSO

2114       AnyEvent::AIO for easy integration into event loops, Coro::AIO for a
2115       more natural syntax.
2116

AUTHOR

2118        Marc Lehmann <schmorp@schmorp.de>
2119        http://home.schmorp.de/
2120
2121
2122
2123perl v5.28.0                      2018-08-25                            AIO(3)
Impressum