ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/IO-AIO/README
(Generate patch)

Comparing IO-AIO/README (file contents):
Revision 1.13 by root, Tue Aug 30 15:45:10 2005 UTC vs.
Revision 1.59 by root, Tue Feb 20 06:54:47 2018 UTC

1NAME 1NAME
2 IO::AIO - Asynchronous Input/Output 2 IO::AIO - Asynchronous/Advanced Input/Output
3 3
4SYNOPSIS 4SYNOPSIS
5 use IO::AIO; 5 use IO::AIO;
6 6
7 aio_open "/etc/passwd", O_RDONLY, 0, sub { 7 aio_open "/etc/passwd", IO::AIO::O_RDONLY, 0, sub {
8 my ($fh) = @_; 8 my $fh = shift
9 or die "/etc/passwd: $!";
9 ... 10 ...
10 }; 11 };
11 12
12 aio_unlink "/tmp/file", sub { }; 13 aio_unlink "/tmp/file", sub { };
13 14
14 aio_read $fh, 30000, 1024, $buffer, 0, sub { 15 aio_read $fh, 30000, 1024, $buffer, 0, sub {
15 $_[0] > 0 or die "read error: $!"; 16 $_[0] > 0 or die "read error: $!";
16 }; 17 };
17 18
18 # Event 19 # version 2+ has request and group objects
19 Event->io (fd => IO::AIO::poll_fileno, 20 use IO::AIO 2;
20 poll => 'r',
21 cb => \&IO::AIO::poll_cb);
22 21
23 # Glib/Gtk2 22 aioreq_pri 4; # give next request a very high priority
24 add_watch Glib::IO IO::AIO::poll_fileno, 23 my $req = aio_unlink "/tmp/file", sub { };
25 in => sub { IO::AIO::poll_cb; 1 }; 24 $req->cancel; # cancel request if still in queue
26 25
27 # Tk 26 my $grp = aio_group sub { print "all stats done\n" };
28 Tk::Event::IO->fileevent (IO::AIO::poll_fileno, "", 27 add $grp aio_stat "..." for ...;
29 readable => \&IO::AIO::poll_cb);
30
31 # Danga::Socket
32 Danga::Socket->AddOtherFds (IO::AIO::poll_fileno =>
33 \&IO::AIO::poll_cb);
34 28
35DESCRIPTION 29DESCRIPTION
36 This module implements asynchronous I/O using whatever means your 30 This module implements asynchronous I/O using whatever means your
37 operating system supports. 31 operating system supports. It is implemented as an interface to "libeio"
32 (<http://software.schmorp.de/pkg/libeio.html>).
38 33
34 Asynchronous means that operations that can normally block your program
35 (e.g. reading from disk) will be done asynchronously: the operation will
36 still block, but you can do something else in the meantime. This is
37 extremely useful for programs that need to stay interactive even when
38 doing heavy I/O (GUI programs, high performance network servers etc.),
39 but can also be used to easily do operations in parallel that are
40 normally done sequentially, e.g. stat'ing many files, which is much
41 faster on a RAID volume or over NFS when you do a number of stat
42 operations concurrently.
43
44 While most of this works on all types of file descriptors (for example
45 sockets), using these functions on file descriptors that support
46 nonblocking operation (again, sockets, pipes etc.) is very inefficient.
47 Use an event loop for that (such as the EV module): IO::AIO will
48 naturally fit into such an event loop itself.
49
39 Currently, a number of threads are started that execute your read/writes 50 In this version, a number of threads are started that execute your
40 and signal their completion. You don't need thread support in your libc 51 requests and signal their completion. You don't need thread support in
41 or perl, and the threads created by this module will not be visible to 52 perl, and the threads created by this module will not be visible to
42 the pthreads library. In the future, this module might make use of the 53 perl. In the future, this module might make use of the native aio
43 native aio functions available on many operating systems. However, they 54 functions available on many operating systems. However, they are often
44 are often not well-supported (Linux doesn't allow them on normal files 55 not well-supported or restricted (GNU/Linux doesn't allow them on normal
45 currently, for example), and they would only support aio_read and 56 files currently, for example), and they would only support aio_read and
46 aio_write, so the remaining functionality would have to be implemented 57 aio_write, so the remaining functionality would have to be implemented
47 using threads anyway. 58 using threads anyway.
48 59
60 In addition to asynchronous I/O, this module also exports some rather
61 arcane interfaces, such as "madvise" or linux's "splice" system call,
62 which is why the "A" in "AIO" can also mean *advanced*.
63
49 Although the module will work with in the presence of other threads, it 64 Although the module will work in the presence of other (Perl-) threads,
50 is currently not reentrant, so use appropriate locking yourself, always 65 it is currently not reentrant in any way, so use appropriate locking
51 call "poll_cb" from within the same thread, or never call "poll_cb" (or 66 yourself, always call "poll_cb" from within the same thread, or never
52 other "aio_" functions) recursively. 67 call "poll_cb" (or other "aio_" functions) recursively.
68
69 EXAMPLE
70 This is a simple example that uses the EV module and loads /etc/passwd
71 asynchronously:
72
73 use EV;
74 use IO::AIO;
75
76 # register the IO::AIO callback with EV
77 my $aio_w = EV::io IO::AIO::poll_fileno, EV::READ, \&IO::AIO::poll_cb;
78
79 # queue the request to open /etc/passwd
80 aio_open "/etc/passwd", IO::AIO::O_RDONLY, 0, sub {
81 my $fh = shift
82 or die "error while opening: $!";
83
84 # stat'ing filehandles is generally non-blocking
85 my $size = -s $fh;
86
87 # queue a request to read the file
88 my $contents;
89 aio_read $fh, 0, $size, $contents, 0, sub {
90 $_[0] == $size
91 or die "short read: $!";
92
93 close $fh;
94
95 # file contents now in $contents
96 print $contents;
97
98 # exit event loop and program
99 EV::break;
100 };
101 };
102
103 # possibly queue up other requests, or open GUI windows,
104 # check for sockets etc. etc.
105
106 # process events as long as there are some:
107 EV::run;
108
109REQUEST ANATOMY AND LIFETIME
110 Every "aio_*" function creates a request. which is a C data structure
111 not directly visible to Perl.
112
113 If called in non-void context, every request function returns a Perl
114 object representing the request. In void context, nothing is returned,
115 which saves a bit of memory.
116
117 The perl object is a fairly standard ref-to-hash object. The hash
118 contents are not used by IO::AIO so you are free to store anything you
119 like in it.
120
121 During their existance, aio requests travel through the following
122 states, in order:
123
124 ready
125 Immediately after a request is created it is put into the ready
126 state, waiting for a thread to execute it.
127
128 execute
129 A thread has accepted the request for processing and is currently
130 executing it (e.g. blocking in read).
131
132 pending
133 The request has been executed and is waiting for result processing.
134
135 While request submission and execution is fully asynchronous, result
136 processing is not and relies on the perl interpreter calling
137 "poll_cb" (or another function with the same effect).
138
139 result
140 The request results are processed synchronously by "poll_cb".
141
142 The "poll_cb" function will process all outstanding aio requests by
143 calling their callbacks, freeing memory associated with them and
144 managing any groups they are contained in.
145
146 done
147 Request has reached the end of its lifetime and holds no resources
148 anymore (except possibly for the Perl object, but its connection to
149 the actual aio request is severed and calling its methods will
150 either do nothing or result in a runtime error).
53 151
54FUNCTIONS 152FUNCTIONS
55 AIO FUNCTIONS 153 QUICK OVERVIEW
154 This section simply lists the prototypes most of the functions for quick
155 reference. See the following sections for function-by-function
156 documentation.
157
158 aio_wd $pathname, $callback->($wd)
159 aio_open $pathname, $flags, $mode, $callback->($fh)
160 aio_close $fh, $callback->($status)
161 aio_seek $fh,$offset,$whence, $callback->($offs)
162 aio_read $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
163 aio_write $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
164 aio_sendfile $out_fh, $in_fh, $in_offset, $length, $callback->($retval)
165 aio_readahead $fh,$offset,$length, $callback->($retval)
166 aio_stat $fh_or_path, $callback->($status)
167 aio_lstat $fh, $callback->($status)
168 aio_statvfs $fh_or_path, $callback->($statvfs)
169 aio_utime $fh_or_path, $atime, $mtime, $callback->($status)
170 aio_chown $fh_or_path, $uid, $gid, $callback->($status)
171 aio_chmod $fh_or_path, $mode, $callback->($status)
172 aio_truncate $fh_or_path, $offset, $callback->($status)
173 aio_allocate $fh, $mode, $offset, $len, $callback->($status)
174 aio_fiemap $fh, $start, $length, $flags, $count, $cb->(\@extents)
175 aio_unlink $pathname, $callback->($status)
176 aio_mknod $pathname, $mode, $dev, $callback->($status)
177 aio_link $srcpath, $dstpath, $callback->($status)
178 aio_symlink $srcpath, $dstpath, $callback->($status)
179 aio_readlink $pathname, $callback->($link)
180 aio_realpath $pathname, $callback->($path)
181 aio_rename $srcpath, $dstpath, $callback->($status)
182 aio_rename2 $srcpath, $dstpath, $flags, $callback->($status)
183 aio_mkdir $pathname, $mode, $callback->($status)
184 aio_rmdir $pathname, $callback->($status)
185 aio_readdir $pathname, $callback->($entries)
186 aio_readdirx $pathname, $flags, $callback->($entries, $flags)
187 IO::AIO::READDIR_DENTS IO::AIO::READDIR_DIRS_FIRST
188 IO::AIO::READDIR_STAT_ORDER IO::AIO::READDIR_FOUND_UNKNOWN
189 aio_scandir $pathname, $maxreq, $callback->($dirs, $nondirs)
190 aio_load $pathname, $data, $callback->($status)
191 aio_copy $srcpath, $dstpath, $callback->($status)
192 aio_move $srcpath, $dstpath, $callback->($status)
193 aio_rmtree $pathname, $callback->($status)
194 aio_fcntl $fh, $cmd, $arg, $callback->($status)
195 aio_ioctl $fh, $request, $buf, $callback->($status)
196 aio_sync $callback->($status)
197 aio_syncfs $fh, $callback->($status)
198 aio_fsync $fh, $callback->($status)
199 aio_fdatasync $fh, $callback->($status)
200 aio_sync_file_range $fh, $offset, $nbytes, $flags, $callback->($status)
201 aio_pathsync $pathname, $callback->($status)
202 aio_msync $scalar, $offset = 0, $length = undef, flags = MS_SYNC, $callback->($status)
203 aio_mtouch $scalar, $offset = 0, $length = undef, flags = 0, $callback->($status)
204 aio_mlock $scalar, $offset = 0, $length = undef, $callback->($status)
205 aio_mlockall $flags, $callback->($status)
206 aio_group $callback->(...)
207 aio_nop $callback->()
208
209 $prev_pri = aioreq_pri [$pri]
210 aioreq_nice $pri_adjust
211
212 IO::AIO::poll_wait
213 IO::AIO::poll_cb
214 IO::AIO::poll
215 IO::AIO::flush
216 IO::AIO::max_poll_reqs $nreqs
217 IO::AIO::max_poll_time $seconds
218 IO::AIO::min_parallel $nthreads
219 IO::AIO::max_parallel $nthreads
220 IO::AIO::max_idle $nthreads
221 IO::AIO::idle_timeout $seconds
222 IO::AIO::max_outstanding $maxreqs
223 IO::AIO::nreqs
224 IO::AIO::nready
225 IO::AIO::npending
226 $nfd = IO::AIO::get_fdlimit [EXPERIMENTAL]
227 IO::AIO::min_fdlimit $nfd [EXPERIMENTAL]
228
229 IO::AIO::sendfile $ofh, $ifh, $offset, $count
230 IO::AIO::fadvise $fh, $offset, $len, $advice
231 IO::AIO::mmap $scalar, $length, $prot, $flags[, $fh[, $offset]]
232 IO::AIO::munmap $scalar
233 IO::AIO::madvise $scalar, $offset, $length, $advice
234 IO::AIO::mprotect $scalar, $offset, $length, $protect
235 IO::AIO::munlock $scalar, $offset = 0, $length = undef
236 IO::AIO::munlockall
237
238 API NOTES
56 All the "aio_*" calls are more or less thin wrappers around the syscall 239 All the "aio_*" calls are more or less thin wrappers around the syscall
57 with the same name (sans "aio_"). The arguments are similar or 240 with the same name (sans "aio_"). The arguments are similar or
58 identical, and they all accept an additional (and optional) $callback 241 identical, and they all accept an additional (and optional) $callback
59 argument which must be a code reference. This code reference will get 242 argument which must be a code reference. This code reference will be
60 called with the syscall return code (e.g. most syscalls return -1 on 243 called after the syscall has been executed in an asynchronous fashion.
61 error, unlike perl, which usually delivers "false") as it's sole 244 The results of the request will be passed as arguments to the callback
62 argument when the given syscall has been executed asynchronously. 245 (and, if an error occured, in $!) - for most requests the syscall return
246 code (e.g. most syscalls return -1 on error, unlike perl, which usually
247 delivers "false").
248
249 Some requests (such as "aio_readdir") pass the actual results and
250 communicate failures by passing "undef".
63 251
64 All functions expecting a filehandle keep a copy of the filehandle 252 All functions expecting a filehandle keep a copy of the filehandle
65 internally until the request has finished. 253 internally until the request has finished.
66 254
255 All functions return request objects of type IO::AIO::REQ that allow
256 further manipulation of those requests while they are in-flight.
257
67 The pathnames you pass to these routines *must* be absolute and encoded 258 The pathnames you pass to these routines *should* be absolute. The
68 in byte form. The reason for the former is that at the time the request 259 reason for this is that at the time the request is being executed, the
69 is being executed, the current working directory could have changed. 260 current working directory could have changed. Alternatively, you can
70 Alternatively, you can make sure that you never change the current 261 make sure that you never change the current working directory anywhere
71 working directory. 262 in the program and then use relative paths. You can also take advantage
263 of IO::AIOs working directory abstraction, that lets you specify paths
264 relative to some previously-opened "working directory object" - see the
265 description of the "IO::AIO::WD" class later in this document.
72 266
73 To encode pathnames to byte form, either make sure you either: a) always 267 To encode pathnames as octets, either make sure you either: a) always
74 pass in filenames you got from outside (command line, readdir etc.), b) 268 pass in filenames you got from outside (command line, readdir etc.)
75 are ASCII or ISO 8859-1, c) use the Encode module and encode your 269 without tinkering, b) are in your native filesystem encoding, c) use the
76 pathnames to the locale (or other) encoding in effect in the user 270 Encode module and encode your pathnames to the locale (or other)
77 environment, d) use Glib::filename_from_unicode on unicode filenames or 271 encoding in effect in the user environment, d) use
78 e) use something else. 272 Glib::filename_from_unicode on unicode filenames or e) use something
273 else to ensure your scalar has the correct contents.
274
275 This works, btw. independent of the internal UTF-8 bit, which IO::AIO
276 handles correctly whether it is set or not.
277
278 AIO REQUEST FUNCTIONS
279 $prev_pri = aioreq_pri [$pri]
280 Returns the priority value that would be used for the next request
281 and, if $pri is given, sets the priority for the next aio request.
282
283 The default priority is 0, the minimum and maximum priorities are -4
284 and 4, respectively. Requests with higher priority will be serviced
285 first.
286
287 The priority will be reset to 0 after each call to one of the
288 "aio_*" functions.
289
290 Example: open a file with low priority, then read something from it
291 with higher priority so the read request is serviced before other
292 low priority open requests (potentially spamming the cache):
293
294 aioreq_pri -3;
295 aio_open ..., sub {
296 return unless $_[0];
297
298 aioreq_pri -2;
299 aio_read $_[0], ..., sub {
300 ...
301 };
302 };
303
304 aioreq_nice $pri_adjust
305 Similar to "aioreq_pri", but subtracts the given value from the
306 current priority, so the effect is cumulative.
79 307
80 aio_open $pathname, $flags, $mode, $callback->($fh) 308 aio_open $pathname, $flags, $mode, $callback->($fh)
81 Asynchronously open or create a file and call the callback with a 309 Asynchronously open or create a file and call the callback with a
82 newly created filehandle for the file. 310 newly created filehandle for the file (or "undef" in case of an
311 error).
83 312
84 The pathname passed to "aio_open" must be absolute. See API NOTES, 313 The pathname passed to "aio_open" must be absolute. See API NOTES,
85 above, for an explanation. 314 above, for an explanation.
86 315
87 The $flags argument is a bitmask. See the "Fcntl" module for a list. 316 The $flags argument is a bitmask. See the "Fcntl" module for a list.
88 They are the same as used by "sysopen". 317 They are the same as used by "sysopen".
89 318
90 Likewise, $mode specifies the mode of the newly created file, if it 319 Likewise, $mode specifies the mode of the newly created file, if it
91 didn't exist and "O_CREAT" has been given, just like perl's 320 didn't exist and "O_CREAT" has been given, just like perl's
92 "sysopen", except that it is mandatory (i.e. use 0 if you don't 321 "sysopen", except that it is mandatory (i.e. use 0 if you don't
93 create new files, and 0666 or 0777 if you do). 322 create new files, and 0666 or 0777 if you do). Note that the $mode
323 will be modified by the umask in effect then the request is being
324 executed, so better never change the umask.
94 325
95 Example: 326 Example:
96 327
97 aio_open "/etc/passwd", O_RDONLY, 0, sub { 328 aio_open "/etc/passwd", IO::AIO::O_RDONLY, 0, sub {
98 if ($_[0]) { 329 if ($_[0]) {
99 print "open successful, fh is $_[0]\n"; 330 print "open successful, fh is $_[0]\n";
100 ... 331 ...
101 } else { 332 } else {
102 die "open failed: $!\n"; 333 die "open failed: $!\n";
103 } 334 }
104 }; 335 };
105 336
337 In addition to all the common open modes/flags ("O_RDONLY",
338 "O_WRONLY", "O_RDWR", "O_CREAT", "O_TRUNC", "O_EXCL" and
339 "O_APPEND"), the following POSIX and non-POSIX constants are
340 available (missing ones on your system are, as usual, 0):
341
342 "O_ASYNC", "O_DIRECT", "O_NOATIME", "O_CLOEXEC", "O_NOCTTY",
343 "O_NOFOLLOW", "O_NONBLOCK", "O_EXEC", "O_SEARCH", "O_DIRECTORY",
344 "O_DSYNC", "O_RSYNC", "O_SYNC", "O_PATH", "O_TMPFILE", and
345 "O_TTY_INIT".
346
106 aio_close $fh, $callback->($status) 347 aio_close $fh, $callback->($status)
107 Asynchronously close a file and call the callback with the result 348 Asynchronously close a file and call the callback with the result
108 code. *WARNING:* although accepted, you should not pass in a perl 349 code.
109 filehandle here, as perl will likely close the file descriptor
110 another time when the filehandle is destroyed. Normally, you can
111 safely call perls "close" or just let filehandles go out of scope.
112 350
113 This is supposed to be a bug in the API, so that might change. It's 351 Unfortunately, you can't do this to perl. Perl *insists* very
114 therefore best to avoid this function. 352 strongly on closing the file descriptor associated with the
353 filehandle itself.
354
355 Therefore, "aio_close" will not close the filehandle - instead it
356 will use dup2 to overwrite the file descriptor with the write-end of
357 a pipe (the pipe fd will be created on demand and will be cached).
358
359 Or in other words: the file descriptor will be closed, but it will
360 not be free for reuse until the perl filehandle is closed.
361
362 aio_seek $fh, $offset, $whence, $callback->($offs)
363 Seeks the filehandle to the new $offset, similarly to perl's
364 "sysseek". The $whence can use the traditional values (0 for
365 "IO::AIO::SEEK_SET", 1 for "IO::AIO::SEEK_CUR" or 2 for
366 "IO::AIO::SEEK_END").
367
368 The resulting absolute offset will be passed to the callback, or -1
369 in case of an error.
370
371 In theory, the $whence constants could be different than the
372 corresponding values from Fcntl, but perl guarantees they are the
373 same, so don't panic.
374
375 As a GNU/Linux (and maybe Solaris) extension, also the constants
376 "IO::AIO::SEEK_DATA" and "IO::AIO::SEEK_HOLE" are available, if they
377 could be found. No guarantees about suitability for use in
378 "aio_seek" or Perl's "sysseek" can be made though, although I would
379 naively assume they "just work".
115 380
116 aio_read $fh,$offset,$length, $data,$dataoffset, $callback->($retval) 381 aio_read $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
117 aio_write $fh,$offset,$length, $data,$dataoffset, $callback->($retval) 382 aio_write $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
118 Reads or writes "length" bytes from the specified "fh" and "offset" 383 Reads or writes $length bytes from or to the specified $fh and
119 into the scalar given by "data" and offset "dataoffset" and calls 384 $offset into the scalar given by $data and offset $dataoffset and
120 the callback without the actual number of bytes read (or -1 on 385 calls the callback with the actual number of bytes transferred (or
121 error, just like the syscall). 386 -1 on error, just like the syscall).
387
388 "aio_read" will, like "sysread", shrink or grow the $data scalar to
389 offset plus the actual number of bytes read.
390
391 If $offset is undefined, then the current file descriptor offset
392 will be used (and updated), otherwise the file descriptor offset
393 will not be changed by these calls.
394
395 If $length is undefined in "aio_write", use the remaining length of
396 $data.
397
398 If $dataoffset is less than zero, it will be counted from the end of
399 $data.
122 400
123 The $data scalar *MUST NOT* be modified in any way while the request 401 The $data scalar *MUST NOT* be modified in any way while the request
124 is outstanding. Modifying it can result in segfaults or WW3 (if the 402 is outstanding. Modifying it can result in segfaults or World War
125 necessary/optional hardware is installed). 403 III (if the necessary/optional hardware is installed).
126 404
127 Example: Read 15 bytes at offset 7 into scalar $buffer, starting at 405 Example: Read 15 bytes at offset 7 into scalar $buffer, starting at
128 offset 0 within the scalar: 406 offset 0 within the scalar:
129 407
130 aio_read $fh, 7, 15, $buffer, 0, sub { 408 aio_read $fh, 7, 15, $buffer, 0, sub {
… …
135 aio_sendfile $out_fh, $in_fh, $in_offset, $length, $callback->($retval) 413 aio_sendfile $out_fh, $in_fh, $in_offset, $length, $callback->($retval)
136 Tries to copy $length bytes from $in_fh to $out_fh. It starts 414 Tries to copy $length bytes from $in_fh to $out_fh. It starts
137 reading at byte offset $in_offset, and starts writing at the current 415 reading at byte offset $in_offset, and starts writing at the current
138 file offset of $out_fh. Because of that, it is not safe to issue 416 file offset of $out_fh. Because of that, it is not safe to issue
139 more than one "aio_sendfile" per $out_fh, as they will interfere 417 more than one "aio_sendfile" per $out_fh, as they will interfere
140 with each other. 418 with each other. The same $in_fh works fine though, as this function
419 does not move or use the file offset of $in_fh.
141 420
421 Please note that "aio_sendfile" can read more bytes from $in_fh than
422 are written, and there is no way to find out how many more bytes
423 have been read from "aio_sendfile" alone, as "aio_sendfile" only
424 provides the number of bytes written to $out_fh. Only if the result
425 value equals $length one can assume that $length bytes have been
426 read.
427
428 Unlike with other "aio_" functions, it makes a lot of sense to use
429 "aio_sendfile" on non-blocking sockets, as long as one end
430 (typically the $in_fh) is a file - the file I/O will then be
431 asynchronous, while the socket I/O will be non-blocking. Note,
432 however, that you can run into a trap where "aio_sendfile" reads
433 some data with readahead, then fails to write all data, and when the
434 socket is ready the next time, the data in the cache is already
435 lost, forcing "aio_sendfile" to again hit the disk. Explicit
436 "aio_read" + "aio_write" let's you better control resource usage.
437
142 This call tries to make use of a native "sendfile" syscall to 438 This call tries to make use of a native "sendfile"-like syscall to
143 provide zero-copy operation. For this to work, $out_fh should refer 439 provide zero-copy operation. For this to work, $out_fh should refer
144 to a socket, and $in_fh should refer to mmap'able file. 440 to a socket, and $in_fh should refer to an mmap'able file.
145 441
146 If the native sendfile call fails or is not implemented, it will be 442 If a native sendfile cannot be found or it fails with "ENOSYS",
147 emulated, so you can call "aio_sendfile" on any type of filehandle 443 "EINVAL", "ENOTSUP", "EOPNOTSUPP", "EAFNOSUPPORT", "EPROTOTYPE" or
148 regardless of the limitations of the operating system. 444 "ENOTSOCK", it will be emulated, so you can call "aio_sendfile" on
445 any type of filehandle regardless of the limitations of the
446 operating system.
149 447
150 Please note, however, that "aio_sendfile" can read more bytes from 448 As native sendfile syscalls (as practically any non-POSIX interface
151 $in_fh than are written, and there is no way to find out how many 449 hacked together in a hurry to improve benchmark numbers) tend to be
152 bytes have been read from "aio_sendfile" alone, as "aio_sendfile" 450 rather buggy on many systems, this implementation tries to work
153 only provides the number of bytes written to $out_fh. Only if the 451 around some known bugs in Linux and FreeBSD kernels (probably
154 result value equals $length one can assume that $length bytes have 452 others, too), but that might fail, so you really really should check
155 been read. 453 the return value of "aio_sendfile" - fewer bytes than expected might
454 have been transferred.
156 455
157 aio_readahead $fh,$offset,$length, $callback->($retval) 456 aio_readahead $fh,$offset,$length, $callback->($retval)
158 "aio_readahead" populates the page cache with data from a file so 457 "aio_readahead" populates the page cache with data from a file so
159 that subsequent reads from that file will not block on disk I/O. The 458 that subsequent reads from that file will not block on disk I/O. The
160 $offset argument specifies the starting point from which data is to 459 $offset argument specifies the starting point from which data is to
… …
163 to a page boundary and bytes are read up to the next page boundary 462 to a page boundary and bytes are read up to the next page boundary
164 greater than or equal to (off-set+length). "aio_readahead" does not 463 greater than or equal to (off-set+length). "aio_readahead" does not
165 read beyond the end of the file. The current file offset of the file 464 read beyond the end of the file. The current file offset of the file
166 is left unchanged. 465 is left unchanged.
167 466
168 If that syscall doesn't exist (likely if your OS isn't Linux) it 467 If that syscall doesn't exist (likely if your kernel isn't Linux) it
169 will be emulated by simply reading the data, which would have a 468 will be emulated by simply reading the data, which would have a
170 similar effect. 469 similar effect.
171 470
172 aio_stat $fh_or_path, $callback->($status) 471 aio_stat $fh_or_path, $callback->($status)
173 aio_lstat $fh, $callback->($status) 472 aio_lstat $fh, $callback->($status)
… …
181 Currently, the stats are always 64-bit-stats, i.e. instead of 480 Currently, the stats are always 64-bit-stats, i.e. instead of
182 returning an error when stat'ing a large file, the results will be 481 returning an error when stat'ing a large file, the results will be
183 silently truncated unless perl itself is compiled with large file 482 silently truncated unless perl itself is compiled with large file
184 support. 483 support.
185 484
485 To help interpret the mode and dev/rdev stat values, IO::AIO offers
486 the following constants and functions (if not implemented, the
487 constants will be 0 and the functions will either "croak" or fall
488 back on traditional behaviour).
489
490 "S_IFMT", "S_IFIFO", "S_IFCHR", "S_IFBLK", "S_IFLNK", "S_IFREG",
491 "S_IFDIR", "S_IFWHT", "S_IFSOCK", "IO::AIO::major $dev_t",
492 "IO::AIO::minor $dev_t", "IO::AIO::makedev $major, $minor".
493
186 Example: Print the length of /etc/passwd: 494 Example: Print the length of /etc/passwd:
187 495
188 aio_stat "/etc/passwd", sub { 496 aio_stat "/etc/passwd", sub {
189 $_[0] and die "stat failed: $!"; 497 $_[0] and die "stat failed: $!";
190 print "size is ", -s _, "\n"; 498 print "size is ", -s _, "\n";
191 }; 499 };
192 500
501 aio_statvfs $fh_or_path, $callback->($statvfs)
502 Works like the POSIX "statvfs" or "fstatvfs" syscalls, depending on
503 whether a file handle or path was passed.
504
505 On success, the callback is passed a hash reference with the
506 following members: "bsize", "frsize", "blocks", "bfree", "bavail",
507 "files", "ffree", "favail", "fsid", "flag" and "namemax". On
508 failure, "undef" is passed.
509
510 The following POSIX IO::AIO::ST_* constants are defined: "ST_RDONLY"
511 and "ST_NOSUID".
512
513 The following non-POSIX IO::AIO::ST_* flag masks are defined to
514 their correct value when available, or to 0 on systems that do not
515 support them: "ST_NODEV", "ST_NOEXEC", "ST_SYNCHRONOUS",
516 "ST_MANDLOCK", "ST_WRITE", "ST_APPEND", "ST_IMMUTABLE",
517 "ST_NOATIME", "ST_NODIRATIME" and "ST_RELATIME".
518
519 Example: stat "/wd" and dump out the data if successful.
520
521 aio_statvfs "/wd", sub {
522 my $f = $_[0]
523 or die "statvfs: $!";
524
525 use Data::Dumper;
526 say Dumper $f;
527 };
528
529 # result:
530 {
531 bsize => 1024,
532 bfree => 4333064312,
533 blocks => 10253828096,
534 files => 2050765568,
535 flag => 4096,
536 favail => 2042092649,
537 bavail => 4333064312,
538 ffree => 2042092649,
539 namemax => 255,
540 frsize => 1024,
541 fsid => 1810
542 }
543
544 aio_utime $fh_or_path, $atime, $mtime, $callback->($status)
545 Works like perl's "utime" function (including the special case of
546 $atime and $mtime being undef). Fractional times are supported if
547 the underlying syscalls support them.
548
549 When called with a pathname, uses utimes(2) if available, otherwise
550 utime(2). If called on a file descriptor, uses futimes(2) if
551 available, otherwise returns ENOSYS, so this is not portable.
552
553 Examples:
554
555 # set atime and mtime to current time (basically touch(1)):
556 aio_utime "path", undef, undef;
557 # set atime to current time and mtime to beginning of the epoch:
558 aio_utime "path", time, undef; # undef==0
559
560 aio_chown $fh_or_path, $uid, $gid, $callback->($status)
561 Works like perl's "chown" function, except that "undef" for either
562 $uid or $gid is being interpreted as "do not change" (but -1 can
563 also be used).
564
565 Examples:
566
567 # same as "chown root path" in the shell:
568 aio_chown "path", 0, -1;
569 # same as above:
570 aio_chown "path", 0, undef;
571
572 aio_truncate $fh_or_path, $offset, $callback->($status)
573 Works like truncate(2) or ftruncate(2).
574
575 aio_allocate $fh, $mode, $offset, $len, $callback->($status)
576 Allocates or frees disk space according to the $mode argument. See
577 the linux "fallocate" documentation for details.
578
579 $mode is usually 0 or "IO::AIO::FALLOC_FL_KEEP_SIZE" to allocate
580 space, or "IO::AIO::FALLOC_FL_PUNCH_HOLE |
581 IO::AIO::FALLOC_FL_KEEP_SIZE", to deallocate a file range.
582
583 IO::AIO also supports "FALLOC_FL_COLLAPSE_RANGE", to remove a range
584 (without leaving a hole), "FALLOC_FL_ZERO_RANGE", to zero a range,
585 "FALLOC_FL_INSERT_RANGE" to insert a range and
586 "FALLOC_FL_UNSHARE_RANGE" to unshare shared blocks (see your
587 fallocate(2) manpage).
588
589 The file system block size used by "fallocate" is presumably the
590 "f_bsize" returned by "statvfs", but different filesystems and
591 filetypes can dictate other limitations.
592
593 If "fallocate" isn't available or cannot be emulated (currently no
594 emulation will be attempted), passes -1 and sets $! to "ENOSYS".
595
596 aio_chmod $fh_or_path, $mode, $callback->($status)
597 Works like perl's "chmod" function.
598
193 aio_unlink $pathname, $callback->($status) 599 aio_unlink $pathname, $callback->($status)
194 Asynchronously unlink (delete) a file and call the callback with the 600 Asynchronously unlink (delete) a file and call the callback with the
195 result code. 601 result code.
196 602
603 aio_mknod $pathname, $mode, $dev, $callback->($status)
604 [EXPERIMENTAL]
605
606 Asynchronously create a device node (or fifo). See mknod(2).
607
608 The only (POSIX-) portable way of calling this function is:
609
610 aio_mknod $pathname, IO::AIO::S_IFIFO | $mode, 0, sub { ...
611
612 See "aio_stat" for info about some potentially helpful extra
613 constants and functions.
614
615 aio_link $srcpath, $dstpath, $callback->($status)
616 Asynchronously create a new link to the existing object at $srcpath
617 at the path $dstpath and call the callback with the result code.
618
619 aio_symlink $srcpath, $dstpath, $callback->($status)
620 Asynchronously create a new symbolic link to the existing object at
621 $srcpath at the path $dstpath and call the callback with the result
622 code.
623
624 aio_readlink $pathname, $callback->($link)
625 Asynchronously read the symlink specified by $path and pass it to
626 the callback. If an error occurs, nothing or undef gets passed to
627 the callback.
628
629 aio_realpath $pathname, $callback->($path)
630 Asynchronously make the path absolute and resolve any symlinks in
631 $path. The resulting path only consists of directories (same as
632 Cwd::realpath).
633
634 This request can be used to get the absolute path of the current
635 working directory by passing it a path of . (a single dot).
636
637 aio_rename $srcpath, $dstpath, $callback->($status)
638 Asynchronously rename the object at $srcpath to $dstpath, just as
639 rename(2) and call the callback with the result code.
640
641 On systems that support the AIO::WD working directory abstraction
642 natively, the case "[$wd, "."]" as $srcpath is specialcased -
643 instead of failing, "rename" is called on the absolute path of $wd.
644
645 aio_rename2 $srcpath, $dstpath, $flags, $callback->($status)
646 Basically a version of "aio_rename" with an additional $flags
647 argument. Calling this with "$flags=0" is the same as calling
648 "aio_rename".
649
650 Non-zero flags are currently only supported on GNU/Linux systems
651 that support renameat2. Other systems fail with "ENOSYS" in this
652 case.
653
654 The following constants are available (missing ones are, as usual
655 0), see renameat2(2) for details:
656
657 "IO::AIO::RENAME_NOREPLACE", "IO::AIO::RENAME_EXCHANGE" and
658 "IO::AIO::RENAME_WHITEOUT".
659
660 aio_mkdir $pathname, $mode, $callback->($status)
661 Asynchronously mkdir (create) a directory and call the callback with
662 the result code. $mode will be modified by the umask at the time the
663 request is executed, so do not change your umask.
664
197 aio_rmdir $pathname, $callback->($status) 665 aio_rmdir $pathname, $callback->($status)
198 Asynchronously rmdir (delete) a directory and call the callback with 666 Asynchronously rmdir (delete) a directory and call the callback with
199 the result code. 667 the result code.
200 668
669 On systems that support the AIO::WD working directory abstraction
670 natively, the case "[$wd, "."]" is specialcased - instead of
671 failing, "rmdir" is called on the absolute path of $wd.
672
201 aio_readdir $pathname $callback->($entries) 673 aio_readdir $pathname, $callback->($entries)
202 Unlike the POSIX call of the same name, "aio_readdir" reads an 674 Unlike the POSIX call of the same name, "aio_readdir" reads an
203 entire directory (i.e. opendir + readdir + closedir). The entries 675 entire directory (i.e. opendir + readdir + closedir). The entries
204 will not be sorted, and will NOT include the "." and ".." entries. 676 will not be sorted, and will NOT include the "." and ".." entries.
205 677
206 The callback a single argument which is either "undef" or an 678 The callback is passed a single argument which is either "undef" or
207 array-ref with the filenames. 679 an array-ref with the filenames.
208 680
681 aio_readdirx $pathname, $flags, $callback->($entries, $flags)
682 Quite similar to "aio_readdir", but the $flags argument allows one
683 to tune behaviour and output format. In case of an error, $entries
684 will be "undef".
685
686 The flags are a combination of the following constants, ORed
687 together (the flags will also be passed to the callback, possibly
688 modified):
689
690 IO::AIO::READDIR_DENTS
691 When this flag is off, then the callback gets an arrayref
692 consisting of names only (as with "aio_readdir"), otherwise it
693 gets an arrayref with "[$name, $type, $inode]" arrayrefs, each
694 describing a single directory entry in more detail.
695
696 $name is the name of the entry.
697
698 $type is one of the "IO::AIO::DT_xxx" constants:
699
700 "IO::AIO::DT_UNKNOWN", "IO::AIO::DT_FIFO", "IO::AIO::DT_CHR",
701 "IO::AIO::DT_DIR", "IO::AIO::DT_BLK", "IO::AIO::DT_REG",
702 "IO::AIO::DT_LNK", "IO::AIO::DT_SOCK", "IO::AIO::DT_WHT".
703
704 "IO::AIO::DT_UNKNOWN" means just that: readdir does not know. If
705 you need to know, you have to run stat yourself. Also, for speed
706 reasons, the $type scalars are read-only: you can not modify
707 them.
708
709 $inode is the inode number (which might not be exact on systems
710 with 64 bit inode numbers and 32 bit perls). This field has
711 unspecified content on systems that do not deliver the inode
712 information.
713
714 IO::AIO::READDIR_DIRS_FIRST
715 When this flag is set, then the names will be returned in an
716 order where likely directories come first, in optimal stat
717 order. This is useful when you need to quickly find directories,
718 or you want to find all directories while avoiding to stat()
719 each entry.
720
721 If the system returns type information in readdir, then this is
722 used to find directories directly. Otherwise, likely directories
723 are names beginning with ".", or otherwise names with no dots,
724 of which names with short names are tried first.
725
726 IO::AIO::READDIR_STAT_ORDER
727 When this flag is set, then the names will be returned in an
728 order suitable for stat()'ing each one. That is, when you plan
729 to stat() all files in the given directory, then the returned
730 order will likely be fastest.
731
732 If both this flag and "IO::AIO::READDIR_DIRS_FIRST" are
733 specified, then the likely dirs come first, resulting in a less
734 optimal stat order.
735
736 IO::AIO::READDIR_FOUND_UNKNOWN
737 This flag should not be set when calling "aio_readdirx".
738 Instead, it is being set by "aio_readdirx", when any of the
739 $type's found were "IO::AIO::DT_UNKNOWN". The absence of this
740 flag therefore indicates that all $type's are known, which can
741 be used to speed up some algorithms.
742
743 aio_slurp $pathname, $offset, $length, $data, $callback->($status)
744 Opens, reads and closes the given file. The data is put into $data,
745 which is resized as required.
746
747 If $offset is negative, then it is counted from the end of the file.
748
749 If $length is zero, then the remaining length of the file is used.
750 Also, in this case, the same limitations to modifying $data apply as
751 when IO::AIO::mmap is used, i.e. it must only be modified in-place
752 with "substr". If the size of the file is known, specifying a
753 non-zero $length results in a performance advantage.
754
755 This request is similar to the older "aio_load" request, but since
756 it is a single request, it might be more efficient to use.
757
758 Example: load /etc/passwd into $passwd.
759
760 my $passwd;
761 aio_slurp "/etc/passwd", 0, 0, $passwd, sub {
762 $_[0] >= 0
763 or die "/etc/passwd: $!\n";
764
765 printf "/etc/passwd is %d bytes long, and contains:\n", length $passwd;
766 print $passwd;
767 };
768 IO::AIO::flush;
769
770 aio_load $pathname, $data, $callback->($status)
771 This is a composite request that tries to fully load the given file
772 into memory. Status is the same as with aio_read.
773
774 Using "aio_slurp" might be more efficient, as it is a single
775 request.
776
777 aio_copy $srcpath, $dstpath, $callback->($status)
778 Try to copy the *file* (directories not supported as either source
779 or destination) from $srcpath to $dstpath and call the callback with
780 a status of 0 (ok) or -1 (error, see $!).
781
782 Existing destination files will be truncated.
783
784 This is a composite request that creates the destination file with
785 mode 0200 and copies the contents of the source file into it using
786 "aio_sendfile", followed by restoring atime, mtime, access mode and
787 uid/gid, in that order.
788
789 If an error occurs, the partial destination file will be unlinked,
790 if possible, except when setting atime, mtime, access mode and
791 uid/gid, where errors are being ignored.
792
793 aio_move $srcpath, $dstpath, $callback->($status)
794 Try to move the *file* (directories not supported as either source
795 or destination) from $srcpath to $dstpath and call the callback with
796 a status of 0 (ok) or -1 (error, see $!).
797
798 This is a composite request that tries to rename(2) the file first;
799 if rename fails with "EXDEV", it copies the file with "aio_copy"
800 and, if that is successful, unlinks the $srcpath.
801
209 aio_scandir $path, $maxreq, $callback->($dirs, $nondirs) 802 aio_scandir $pathname, $maxreq, $callback->($dirs, $nondirs)
210 Scans a directory (similar to "aio_readdir") and tries to separate 803 Scans a directory (similar to "aio_readdir") but additionally tries
211 the entries of directory $path into two sets of names, ones you can 804 to efficiently separate the entries of directory $path into two sets
212 recurse into (directories), and ones you cannot recurse into 805 of names, directories you can recurse into (directories), and ones
213 (everything else). 806 you cannot recurse into (everything else, including symlinks to
807 directories).
214 808
215 "aio_scandir" is a composite request that consists of many 809 "aio_scandir" is a composite request that generates many sub
216 aio-primitives. $maxreq specifies the maximum number of outstanding 810 requests. $maxreq specifies the maximum number of outstanding aio
217 aio requests that this function generates. If it is "<= 0", then a 811 requests that this function generates. If it is "<= 0", then a
218 suitable default will be chosen (currently 8). 812 suitable default will be chosen (currently 4).
219 813
220 On error, the callback is called without arguments, otherwise it 814 On error, the callback is called without arguments, otherwise it
221 receives two array-refs with path-relative entry names. 815 receives two array-refs with path-relative entry names.
222 816
223 Example: 817 Example:
… …
231 Implementation notes. 825 Implementation notes.
232 826
233 The "aio_readdir" cannot be avoided, but "stat()"'ing every entry 827 The "aio_readdir" cannot be avoided, but "stat()"'ing every entry
234 can. 828 can.
235 829
830 If readdir returns file type information, then this is used directly
831 to find directories.
832
236 After reading the directory, the modification time, size etc. of the 833 Otherwise, after reading the directory, the modification time, size
237 directory before and after the readdir is checked, and if they 834 etc. of the directory before and after the readdir is checked, and
238 match, the link count will be used to decide how many entries are 835 if they match (and isn't the current time), the link count will be
239 directories (if >= 2). Otherwise, no knowledge of the number of 836 used to decide how many entries are directories (if >= 2).
240 subdirectories will be assumed. 837 Otherwise, no knowledge of the number of subdirectories will be
838 assumed.
241 839
242 Then entires will be sorted into likely directories (everything 840 Then entries will be sorted into likely directories a non-initial
243 without a non-initial dot) and likely non-directories (everything 841 dot currently) and likely non-directories (see "aio_readdirx"). Then
244 else). Then every entry + "/." will be "stat"'ed, likely directories 842 every entry plus an appended "/." will be "stat"'ed, likely
245 first. This is often faster because filesystems might detect the 843 directories first, in order of their inode numbers. If that
844 succeeds, it assumes that the entry is a directory or a symlink to
845 directory (which will be checked separately). This is often faster
846 than stat'ing the entry itself because filesystems might detect the
246 type of the entry without reading the inode data (e.g. ext2s 847 type of the entry without reading the inode data (e.g. ext2fs
247 filetype feature). If that succeeds, it assumes that the entry is a 848 filetype feature), even on systems that cannot return the filetype
248 directory or a symlink to directory (which will be checked 849 information on readdir.
249 seperately).
250 850
251 If the known number of directories has been reached, the rest of the 851 If the known number of directories (link count - 2) has been
252 entries is assumed to be non-directories. 852 reached, the rest of the entries is assumed to be non-directories.
853
854 This only works with certainty on POSIX (= UNIX) filesystems, which
855 fortunately are the vast majority of filesystems around.
856
857 It will also likely work on non-POSIX filesystems with reduced
858 efficiency as those tend to return 0 or 1 as link counts, which
859 disables the directory counting heuristic.
860
861 aio_rmtree $pathname, $callback->($status)
862 Delete a directory tree starting (and including) $path, return the
863 status of the final "rmdir" only. This is a composite request that
864 uses "aio_scandir" to recurse into and rmdir directories, and unlink
865 everything else.
866
867 aio_fcntl $fh, $cmd, $arg, $callback->($status)
868 aio_ioctl $fh, $request, $buf, $callback->($status)
869 These work just like the "fcntl" and "ioctl" built-in functions,
870 except they execute asynchronously and pass the return value to the
871 callback.
872
873 Both calls can be used for a lot of things, some of which make more
874 sense to run asynchronously in their own thread, while some others
875 make less sense. For example, calls that block waiting for external
876 events, such as locking, will also lock down an I/O thread while it
877 is waiting, which can deadlock the whole I/O system. At the same
878 time, there might be no alternative to using a thread to wait.
879
880 So in general, you should only use these calls for things that do
881 (filesystem) I/O, not for things that wait for other events
882 (network, other processes), although if you are careful and know
883 what you are doing, you still can.
884
885 The following constants are available (missing ones are, as usual
886 0):
887
888 "F_DUPFD_CLOEXEC",
889
890 "F_OFD_GETLK", "F_OFD_SETLK", "F_OFD_GETLKW",
891
892 "FIFREEZE", "FITHAW", "FITRIM", "FICLONE", "FICLONERANGE",
893 "FIDEDUPERANGE".
894
895 "FS_IOC_GETFLAGS", "FS_IOC_SETFLAGS", "FS_IOC_GETVERSION",
896 "FS_IOC_SETVERSION", "FS_IOC_FIEMAP".
897
898 "FS_IOC_FSGETXATTR", "FS_IOC_FSSETXATTR",
899 "FS_IOC_SET_ENCRYPTION_POLICY", "FS_IOC_GET_ENCRYPTION_PWSALT",
900 "FS_IOC_GET_ENCRYPTION_POLICY", "FS_KEY_DESCRIPTOR_SIZE".
901
902 "FS_SECRM_FL", "FS_UNRM_FL", "FS_COMPR_FL", "FS_SYNC_FL",
903 "FS_IMMUTABLE_FL", "FS_APPEND_FL", "FS_NODUMP_FL", "FS_NOATIME_FL",
904 "FS_DIRTY_FL", "FS_COMPRBLK_FL", "FS_NOCOMP_FL", "FS_ENCRYPT_FL",
905 "FS_BTREE_FL", "FS_INDEX_FL", "FS_JOURNAL_DATA_FL", "FS_NOTAIL_FL",
906 "FS_DIRSYNC_FL", "FS_TOPDIR_FL", "FS_FL_USER_MODIFIABLE".
907
908 "FS_XFLAG_REALTIME", "FS_XFLAG_PREALLOC", "FS_XFLAG_IMMUTABLE",
909 "FS_XFLAG_APPEND", "FS_XFLAG_SYNC", "FS_XFLAG_NOATIME",
910 "FS_XFLAG_NODUMP", "FS_XFLAG_RTINHERIT", "FS_XFLAG_PROJINHERIT",
911 "FS_XFLAG_NOSYMLINKS", "FS_XFLAG_EXTSIZE", "FS_XFLAG_EXTSZINHERIT",
912 "FS_XFLAG_NODEFRAG", "FS_XFLAG_FILESTREAM", "FS_XFLAG_DAX",
913 "FS_XFLAG_HASATTR",
914
915 aio_sync $callback->($status)
916 Asynchronously call sync and call the callback when finished.
253 917
254 aio_fsync $fh, $callback->($status) 918 aio_fsync $fh, $callback->($status)
255 Asynchronously call fsync on the given filehandle and call the 919 Asynchronously call fsync on the given filehandle and call the
256 callback with the fsync result code. 920 callback with the fsync result code.
257 921
… …
260 callback with the fdatasync result code. 924 callback with the fdatasync result code.
261 925
262 If this call isn't available because your OS lacks it or it couldn't 926 If this call isn't available because your OS lacks it or it couldn't
263 be detected, it will be emulated by calling "fsync" instead. 927 be detected, it will be emulated by calling "fsync" instead.
264 928
929 aio_syncfs $fh, $callback->($status)
930 Asynchronously call the syncfs syscall to sync the filesystem
931 associated to the given filehandle and call the callback with the
932 syncfs result code. If syncfs is not available, calls sync(), but
933 returns -1 and sets errno to "ENOSYS" nevertheless.
934
935 aio_sync_file_range $fh, $offset, $nbytes, $flags, $callback->($status)
936 Sync the data portion of the file specified by $offset and $length
937 to disk (but NOT the metadata), by calling the Linux-specific
938 sync_file_range call. If sync_file_range is not available or it
939 returns ENOSYS, then fdatasync or fsync is being substituted.
940
941 $flags can be a combination of
942 "IO::AIO::SYNC_FILE_RANGE_WAIT_BEFORE",
943 "IO::AIO::SYNC_FILE_RANGE_WRITE" and
944 "IO::AIO::SYNC_FILE_RANGE_WAIT_AFTER": refer to the sync_file_range
945 manpage for details.
946
947 aio_pathsync $pathname, $callback->($status)
948 This request tries to open, fsync and close the given path. This is
949 a composite request intended to sync directories after directory
950 operations (E.g. rename). This might not work on all operating
951 systems or have any specific effect, but usually it makes sure that
952 directory changes get written to disc. It works for anything that
953 can be opened for read-only, not just directories.
954
955 Future versions of this function might fall back to other methods
956 when "fsync" on the directory fails (such as calling "sync").
957
958 Passes 0 when everything went ok, and -1 on error.
959
960 aio_msync $scalar, $offset = 0, $length = undef, flags = MS_SYNC,
961 $callback->($status)
962 This is a rather advanced IO::AIO call, which only works on
963 mmap(2)ed scalars (see the "IO::AIO::mmap" function, although it
964 also works on data scalars managed by the Sys::Mmap or Mmap modules,
965 note that the scalar must only be modified in-place while an aio
966 operation is pending on it).
967
968 It calls the "msync" function of your OS, if available, with the
969 memory area starting at $offset in the string and ending $length
970 bytes later. If $length is negative, counts from the end, and if
971 $length is "undef", then it goes till the end of the string. The
972 flags can be either "IO::AIO::MS_ASYNC" or "IO::AIO::MS_SYNC", plus
973 an optional "IO::AIO::MS_INVALIDATE".
974
975 aio_mtouch $scalar, $offset = 0, $length = undef, flags = 0,
976 $callback->($status)
977 This is a rather advanced IO::AIO call, which works best on
978 mmap(2)ed scalars.
979
980 It touches (reads or writes) all memory pages in the specified range
981 inside the scalar. All caveats and parameters are the same as for
982 "aio_msync", above, except for flags, which must be either 0 (which
983 reads all pages and ensures they are instantiated) or
984 "IO::AIO::MT_MODIFY", which modifies the memory pages (by reading
985 and writing an octet from it, which dirties the page).
986
987 aio_mlock $scalar, $offset = 0, $length = undef, $callback->($status)
988 This is a rather advanced IO::AIO call, which works best on
989 mmap(2)ed scalars.
990
991 It reads in all the pages of the underlying storage into memory (if
992 any) and locks them, so they are not getting swapped/paged out or
993 removed.
994
995 If $length is undefined, then the scalar will be locked till the
996 end.
997
998 On systems that do not implement "mlock", this function returns -1
999 and sets errno to "ENOSYS".
1000
1001 Note that the corresponding "munlock" is synchronous and is
1002 documented under "MISCELLANEOUS FUNCTIONS".
1003
1004 Example: open a file, mmap and mlock it - both will be undone when
1005 $data gets destroyed.
1006
1007 open my $fh, "<", $path or die "$path: $!";
1008 my $data;
1009 IO::AIO::mmap $data, -s $fh, IO::AIO::PROT_READ, IO::AIO::MAP_SHARED, $fh;
1010 aio_mlock $data; # mlock in background
1011
1012 aio_mlockall $flags, $callback->($status)
1013 Calls the "mlockall" function with the given $flags (a combination
1014 of "IO::AIO::MCL_CURRENT" and "IO::AIO::MCL_FUTURE").
1015
1016 On systems that do not implement "mlockall", this function returns
1017 -1 and sets errno to "ENOSYS".
1018
1019 Note that the corresponding "munlockall" is synchronous and is
1020 documented under "MISCELLANEOUS FUNCTIONS".
1021
1022 Example: asynchronously lock all current and future pages into
1023 memory.
1024
1025 aio_mlockall IO::AIO::MCL_FUTURE;
1026
1027 aio_fiemap $fh, $start, $length, $flags, $count, $cb->(\@extents)
1028 Queries the extents of the given file (by calling the Linux "FIEMAP"
1029 ioctl, see <http://cvs.schmorp.de/IO-AIO/doc/fiemap.txt> for
1030 details). If the ioctl is not available on your OS, then this
1031 request will fail with "ENOSYS".
1032
1033 $start is the starting offset to query extents for, $length is the
1034 size of the range to query - if it is "undef", then the whole file
1035 will be queried.
1036
1037 $flags is a combination of flags ("IO::AIO::FIEMAP_FLAG_SYNC" or
1038 "IO::AIO::FIEMAP_FLAG_XATTR" - "IO::AIO::FIEMAP_FLAGS_COMPAT" is
1039 also exported), and is normally 0 or "IO::AIO::FIEMAP_FLAG_SYNC" to
1040 query the data portion.
1041
1042 $count is the maximum number of extent records to return. If it is
1043 "undef", then IO::AIO queries all extents of the range. As a very
1044 special case, if it is 0, then the callback receives the number of
1045 extents instead of the extents themselves (which is unreliable, see
1046 below).
1047
1048 If an error occurs, the callback receives no arguments. The special
1049 "errno" value "IO::AIO::EBADR" is available to test for flag errors.
1050
1051 Otherwise, the callback receives an array reference with extent
1052 structures. Each extent structure is an array reference itself, with
1053 the following members:
1054
1055 [$logical, $physical, $length, $flags]
1056
1057 Flags is any combination of the following flag values (typically
1058 either 0 or "IO::AIO::FIEMAP_EXTENT_LAST" (1)):
1059
1060 "IO::AIO::FIEMAP_EXTENT_LAST", "IO::AIO::FIEMAP_EXTENT_UNKNOWN",
1061 "IO::AIO::FIEMAP_EXTENT_DELALLOC", "IO::AIO::FIEMAP_EXTENT_ENCODED",
1062 "IO::AIO::FIEMAP_EXTENT_DATA_ENCRYPTED",
1063 "IO::AIO::FIEMAP_EXTENT_NOT_ALIGNED",
1064 "IO::AIO::FIEMAP_EXTENT_DATA_INLINE",
1065 "IO::AIO::FIEMAP_EXTENT_DATA_TAIL",
1066 "IO::AIO::FIEMAP_EXTENT_UNWRITTEN", "IO::AIO::FIEMAP_EXTENT_MERGED"
1067 or "IO::AIO::FIEMAP_EXTENT_SHARED".
1068
1069 At the time of this writing (Linux 3.2), this request is unreliable
1070 unless $count is "undef", as the kernel has all sorts of bugs
1071 preventing it to return all extents of a range for files with a
1072 large number of extents. The code (only) works around all these
1073 issues if $count is "undef".
1074
1075 aio_group $callback->(...)
1076 This is a very special aio request: Instead of doing something, it
1077 is a container for other aio requests, which is useful if you want
1078 to bundle many requests into a single, composite, request with a
1079 definite callback and the ability to cancel the whole request with
1080 its subrequests.
1081
1082 Returns an object of class IO::AIO::GRP. See its documentation below
1083 for more info.
1084
1085 Example:
1086
1087 my $grp = aio_group sub {
1088 print "all stats done\n";
1089 };
1090
1091 add $grp
1092 (aio_stat ...),
1093 (aio_stat ...),
1094 ...;
1095
1096 aio_nop $callback->()
1097 This is a special request - it does nothing in itself and is only
1098 used for side effects, such as when you want to add a dummy request
1099 to a group so that finishing the requests in the group depends on
1100 executing the given code.
1101
1102 While this request does nothing, it still goes through the execution
1103 phase and still requires a worker thread. Thus, the callback will
1104 not be executed immediately but only after other requests in the
1105 queue have entered their execution phase. This can be used to
1106 measure request latency.
1107
1108 IO::AIO::aio_busy $fractional_seconds, $callback->() *NOT EXPORTED*
1109 Mainly used for debugging and benchmarking, this aio request puts
1110 one of the request workers to sleep for the given time.
1111
1112 While it is theoretically handy to have simple I/O scheduling
1113 requests like sleep and file handle readable/writable, the overhead
1114 this creates is immense (it blocks a thread for a long time) so do
1115 not use this function except to put your application under
1116 artificial I/O pressure.
1117
1118 IO::AIO::WD - multiple working directories
1119 Your process only has one current working directory, which is used by
1120 all threads. This makes it hard to use relative paths (some other
1121 component could call "chdir" at any time, and it is hard to control when
1122 the path will be used by IO::AIO).
1123
1124 One solution for this is to always use absolute paths. This usually
1125 works, but can be quite slow (the kernel has to walk the whole path on
1126 every access), and can also be a hassle to implement.
1127
1128 Newer POSIX systems have a number of functions (openat, fdopendir,
1129 futimensat and so on) that make it possible to specify working
1130 directories per operation.
1131
1132 For portability, and because the clowns who "designed", or shall I
1133 write, perpetrated this new interface were obviously half-drunk, this
1134 abstraction cannot be perfect, though.
1135
1136 IO::AIO allows you to convert directory paths into a so-called
1137 IO::AIO::WD object. This object stores the canonicalised, absolute
1138 version of the path, and on systems that allow it, also a directory file
1139 descriptor.
1140
1141 Everywhere where a pathname is accepted by IO::AIO (e.g. in "aio_stat"
1142 or "aio_unlink"), one can specify an array reference with an IO::AIO::WD
1143 object and a pathname instead (or the IO::AIO::WD object alone, which
1144 gets interpreted as "[$wd, "."]"). If the pathname is absolute, the
1145 IO::AIO::WD object is ignored, otherwise the pathname is resolved
1146 relative to that IO::AIO::WD object.
1147
1148 For example, to get a wd object for /etc and then stat passwd inside,
1149 you would write:
1150
1151 aio_wd "/etc", sub {
1152 my $etcdir = shift;
1153
1154 # although $etcdir can be undef on error, there is generally no reason
1155 # to check for errors here, as aio_stat will fail with ENOENT
1156 # when $etcdir is undef.
1157
1158 aio_stat [$etcdir, "passwd"], sub {
1159 # yay
1160 };
1161 };
1162
1163 The fact that "aio_wd" is a request and not a normal function shows that
1164 creating an IO::AIO::WD object is itself a potentially blocking
1165 operation, which is why it is done asynchronously.
1166
1167 To stat the directory obtained with "aio_wd" above, one could write
1168 either of the following three request calls:
1169
1170 aio_lstat "/etc" , sub { ... # pathname as normal string
1171 aio_lstat [$wd, "."], sub { ... # "." relative to $wd (i.e. $wd itself)
1172 aio_lstat $wd , sub { ... # shorthand for the previous
1173
1174 As with normal pathnames, IO::AIO keeps a copy of the working directory
1175 object and the pathname string, so you could write the following without
1176 causing any issues due to $path getting reused:
1177
1178 my $path = [$wd, undef];
1179
1180 for my $name (qw(abc def ghi)) {
1181 $path->[1] = $name;
1182 aio_stat $path, sub {
1183 # ...
1184 };
1185 }
1186
1187 There are some caveats: when directories get renamed (or deleted), the
1188 pathname string doesn't change, so will point to the new directory (or
1189 nowhere at all), while the directory fd, if available on the system,
1190 will still point to the original directory. Most functions accepting a
1191 pathname will use the directory fd on newer systems, and the string on
1192 older systems. Some functions (such as "aio_realpath") will always rely
1193 on the string form of the pathname.
1194
1195 So this functionality is mainly useful to get some protection against
1196 "chdir", to easily get an absolute path out of a relative path for
1197 future reference, and to speed up doing many operations in the same
1198 directory (e.g. when stat'ing all files in a directory).
1199
1200 The following functions implement this working directory abstraction:
1201
1202 aio_wd $pathname, $callback->($wd)
1203 Asynchonously canonicalise the given pathname and convert it to an
1204 IO::AIO::WD object representing it. If possible and supported on the
1205 system, also open a directory fd to speed up pathname resolution
1206 relative to this working directory.
1207
1208 If something goes wrong, then "undef" is passwd to the callback
1209 instead of a working directory object and $! is set appropriately.
1210 Since passing "undef" as working directory component of a pathname
1211 fails the request with "ENOENT", there is often no need for error
1212 checking in the "aio_wd" callback, as future requests using the
1213 value will fail in the expected way.
1214
1215 IO::AIO::CWD
1216 This is a compiletime constant (object) that represents the process
1217 current working directory.
1218
1219 Specifying this object as working directory object for a pathname is
1220 as if the pathname would be specified directly, without a directory
1221 object. For example, these calls are functionally identical:
1222
1223 aio_stat "somefile", sub { ... };
1224 aio_stat [IO::AIO::CWD, "somefile"], sub { ... };
1225
1226 To recover the path associated with an IO::AIO::WD object, you can use
1227 "aio_realpath":
1228
1229 aio_realpath $wd, sub {
1230 warn "path is $_[0]\n";
1231 };
1232
1233 Currently, "aio_statvfs" always, and "aio_rename" and "aio_rmdir"
1234 sometimes, fall back to using an absolue path.
1235
1236 IO::AIO::REQ CLASS
1237 All non-aggregate "aio_*" functions return an object of this class when
1238 called in non-void context.
1239
1240 cancel $req
1241 Cancels the request, if possible. Has the effect of skipping
1242 execution when entering the execute state and skipping calling the
1243 callback when entering the the result state, but will leave the
1244 request otherwise untouched (with the exception of readdir). That
1245 means that requests that currently execute will not be stopped and
1246 resources held by the request will not be freed prematurely.
1247
1248 cb $req $callback->(...)
1249 Replace (or simply set) the callback registered to the request.
1250
1251 IO::AIO::GRP CLASS
1252 This class is a subclass of IO::AIO::REQ, so all its methods apply to
1253 objects of this class, too.
1254
1255 A IO::AIO::GRP object is a special request that can contain multiple
1256 other aio requests.
1257
1258 You create one by calling the "aio_group" constructing function with a
1259 callback that will be called when all contained requests have entered
1260 the "done" state:
1261
1262 my $grp = aio_group sub {
1263 print "all requests are done\n";
1264 };
1265
1266 You add requests by calling the "add" method with one or more
1267 "IO::AIO::REQ" objects:
1268
1269 $grp->add (aio_unlink "...");
1270
1271 add $grp aio_stat "...", sub {
1272 $_[0] or return $grp->result ("error");
1273
1274 # add another request dynamically, if first succeeded
1275 add $grp aio_open "...", sub {
1276 $grp->result ("ok");
1277 };
1278 };
1279
1280 This makes it very easy to create composite requests (see the source of
1281 "aio_move" for an application) that work and feel like simple requests.
1282
1283 * The IO::AIO::GRP objects will be cleaned up during calls to
1284 "IO::AIO::poll_cb", just like any other request.
1285
1286 * They can be canceled like any other request. Canceling will cancel
1287 not only the request itself, but also all requests it contains.
1288
1289 * They can also can also be added to other IO::AIO::GRP objects.
1290
1291 * You must not add requests to a group from within the group callback
1292 (or any later time).
1293
1294 Their lifetime, simplified, looks like this: when they are empty, they
1295 will finish very quickly. If they contain only requests that are in the
1296 "done" state, they will also finish. Otherwise they will continue to
1297 exist.
1298
1299 That means after creating a group you have some time to add requests
1300 (precisely before the callback has been invoked, which is only done
1301 within the "poll_cb"). And in the callbacks of those requests, you can
1302 add further requests to the group. And only when all those requests have
1303 finished will the the group itself finish.
1304
1305 add $grp ...
1306 $grp->add (...)
1307 Add one or more requests to the group. Any type of IO::AIO::REQ can
1308 be added, including other groups, as long as you do not create
1309 circular dependencies.
1310
1311 Returns all its arguments.
1312
1313 $grp->cancel_subs
1314 Cancel all subrequests and clears any feeder, but not the group
1315 request itself. Useful when you queued a lot of events but got a
1316 result early.
1317
1318 The group request will finish normally (you cannot add requests to
1319 the group).
1320
1321 $grp->result (...)
1322 Set the result value(s) that will be passed to the group callback
1323 when all subrequests have finished and set the groups errno to the
1324 current value of errno (just like calling "errno" without an error
1325 number). By default, no argument will be passed and errno is zero.
1326
1327 $grp->errno ([$errno])
1328 Sets the group errno value to $errno, or the current value of errno
1329 when the argument is missing.
1330
1331 Every aio request has an associated errno value that is restored
1332 when the callback is invoked. This method lets you change this value
1333 from its default (0).
1334
1335 Calling "result" will also set errno, so make sure you either set $!
1336 before the call to "result", or call c<errno> after it.
1337
1338 feed $grp $callback->($grp)
1339 Sets a feeder/generator on this group: every group can have an
1340 attached generator that generates requests if idle. The idea behind
1341 this is that, although you could just queue as many requests as you
1342 want in a group, this might starve other requests for a potentially
1343 long time. For example, "aio_scandir" might generate hundreds of
1344 thousands of "aio_stat" requests, delaying any later requests for a
1345 long time.
1346
1347 To avoid this, and allow incremental generation of requests, you can
1348 instead a group and set a feeder on it that generates those
1349 requests. The feed callback will be called whenever there are few
1350 enough (see "limit", below) requests active in the group itself and
1351 is expected to queue more requests.
1352
1353 The feed callback can queue as many requests as it likes (i.e. "add"
1354 does not impose any limits).
1355
1356 If the feed does not queue more requests when called, it will be
1357 automatically removed from the group.
1358
1359 If the feed limit is 0 when this method is called, it will be set to
1360 2 automatically.
1361
1362 Example:
1363
1364 # stat all files in @files, but only ever use four aio requests concurrently:
1365
1366 my $grp = aio_group sub { print "finished\n" };
1367 limit $grp 4;
1368 feed $grp sub {
1369 my $file = pop @files
1370 or return;
1371
1372 add $grp aio_stat $file, sub { ... };
1373 };
1374
1375 limit $grp $num
1376 Sets the feeder limit for the group: The feeder will be called
1377 whenever the group contains less than this many requests.
1378
1379 Setting the limit to 0 will pause the feeding process.
1380
1381 The default value for the limit is 0, but note that setting a feeder
1382 automatically bumps it up to 2.
1383
265 SUPPORT FUNCTIONS 1384 SUPPORT FUNCTIONS
1385 EVENT PROCESSING AND EVENT LOOP INTEGRATION
266 $fileno = IO::AIO::poll_fileno 1386 $fileno = IO::AIO::poll_fileno
267 Return the *request result pipe file descriptor*. This filehandle 1387 Return the *request result pipe file descriptor*. This filehandle
268 must be polled for reading by some mechanism outside this module 1388 must be polled for reading by some mechanism outside this module
269 (e.g. Event or select, see below or the SYNOPSIS). If the pipe 1389 (e.g. EV, Glib, select and so on, see below or the SYNOPSIS). If the
270 becomes readable you have to call "poll_cb" to check the results. 1390 pipe becomes readable you have to call "poll_cb" to check the
1391 results.
271 1392
272 See "poll_cb" for an example. 1393 See "poll_cb" for an example.
273 1394
274 IO::AIO::poll_cb 1395 IO::AIO::poll_cb
275 Process all outstanding events on the result pipe. You have to call 1396 Process some requests that have reached the result phase (i.e. they
276 this regularly. Returns the number of events processed. Returns 1397 have been executed but the results are not yet reported). You have
1398 to call this "regularly" to finish outstanding requests.
1399
1400 Returns 0 if all events could be processed (or there were no events
1401 to process), or -1 if it returned earlier for whatever reason.
277 immediately when no events are outstanding. 1402 Returns immediately when no events are outstanding. The amount of
1403 events processed depends on the settings of "IO::AIO::max_poll_req",
1404 "IO::AIO::max_poll_time" and "IO::AIO::max_outstanding".
1405
1406 If not all requests were processed for whatever reason, the poll
1407 file descriptor will still be ready when "poll_cb" returns, so
1408 normally you don't have to do anything special to have it called
1409 later.
1410
1411 Apart from calling "IO::AIO::poll_cb" when the event filehandle
1412 becomes ready, it can be beneficial to call this function from loops
1413 which submit a lot of requests, to make sure the results get
1414 processed when they become available and not just when the loop is
1415 finished and the event loop takes over again. This function returns
1416 very fast when there are no outstanding requests.
278 1417
279 Example: Install an Event watcher that automatically calls 1418 Example: Install an Event watcher that automatically calls
280 IO::AIO::poll_cb with high priority: 1419 IO::AIO::poll_cb with high priority (more examples can be found in
1420 the SYNOPSIS section, at the top of this document):
281 1421
282 Event->io (fd => IO::AIO::poll_fileno, 1422 Event->io (fd => IO::AIO::poll_fileno,
283 poll => 'r', async => 1, 1423 poll => 'r', async => 1,
284 cb => \&IO::AIO::poll_cb); 1424 cb => \&IO::AIO::poll_cb);
285 1425
286 IO::AIO::poll_wait 1426 IO::AIO::poll_wait
287 Wait till the result filehandle becomes ready for reading (simply 1427 Wait until either at least one request is in the result phase or no
288 does a "select" on the filehandle. This is useful if you want to 1428 requests are outstanding anymore.
289 synchronously wait for some requests to finish). 1429
1430 This is useful if you want to synchronously wait for some requests
1431 to become ready, without actually handling them.
290 1432
291 See "nreqs" for an example. 1433 See "nreqs" for an example.
292 1434
1435 IO::AIO::poll
1436 Waits until some requests have been handled.
1437
1438 Returns the number of requests processed, but is otherwise strictly
1439 equivalent to:
1440
1441 IO::AIO::poll_wait, IO::AIO::poll_cb
1442
293 IO::AIO::nreqs 1443 IO::AIO::flush
294 Returns the number of requests currently outstanding (i.e. for which 1444 Wait till all outstanding AIO requests have been handled.
295 their callback has not been invoked yet).
296 1445
297 Example: wait till there are no outstanding requests anymore: 1446 Strictly equivalent to:
298 1447
299 IO::AIO::poll_wait, IO::AIO::poll_cb 1448 IO::AIO::poll_wait, IO::AIO::poll_cb
300 while IO::AIO::nreqs; 1449 while IO::AIO::nreqs;
301 1450
302 IO::AIO::flush 1451 IO::AIO::max_poll_reqs $nreqs
303 Wait till all outstanding AIO requests have been handled. 1452 IO::AIO::max_poll_time $seconds
1453 These set the maximum number of requests (default 0, meaning
1454 infinity) that are being processed by "IO::AIO::poll_cb" in one
1455 call, respectively the maximum amount of time (default 0, meaning
1456 infinity) spent in "IO::AIO::poll_cb" to process requests (more
1457 correctly the mininum amount of time "poll_cb" is allowed to use).
304 1458
305 Strictly equivalent to: 1459 Setting "max_poll_time" to a non-zero value creates an overhead of
1460 one syscall per request processed, which is not normally a problem
1461 unless your callbacks are really really fast or your OS is really
1462 really slow (I am not mentioning Solaris here). Using
1463 "max_poll_reqs" incurs no overhead.
306 1464
307 IO::AIO::poll_wait, IO::AIO::poll_cb 1465 Setting these is useful if you want to ensure some level of
308 while IO::AIO::nreqs; 1466 interactiveness when perl is not fast enough to process all requests
1467 in time.
309 1468
310 IO::AIO::poll 1469 For interactive programs, values such as 0.01 to 0.1 should be fine.
311 Waits until some requests have been handled.
312 1470
313 Strictly equivalent to: 1471 Example: Install an Event watcher that automatically calls
1472 IO::AIO::poll_cb with low priority, to ensure that other parts of
1473 the program get the CPU sometimes even under high AIO load.
314 1474
315 IO::AIO::poll_wait, IO::AIO::poll_cb 1475 # try not to spend much more than 0.1s in poll_cb
316 if IO::AIO::nreqs; 1476 IO::AIO::max_poll_time 0.1;
317 1477
1478 # use a low priority so other tasks have priority
1479 Event->io (fd => IO::AIO::poll_fileno,
1480 poll => 'r', nice => 1,
1481 cb => &IO::AIO::poll_cb);
1482
1483 CONTROLLING THE NUMBER OF THREADS
318 IO::AIO::min_parallel $nthreads 1484 IO::AIO::min_parallel $nthreads
319 Set the minimum number of AIO threads to $nthreads. The current 1485 Set the minimum number of AIO threads to $nthreads. The current
320 default is 4, which means four asynchronous operations can be done 1486 default is 8, which means eight asynchronous operations can execute
321 at one time (the number of outstanding operations, however, is 1487 concurrently at any one time (the number of outstanding requests,
322 unlimited). 1488 however, is unlimited).
323 1489
324 IO::AIO starts threads only on demand, when an AIO request is queued 1490 IO::AIO starts threads only on demand, when an AIO request is queued
325 and no free thread exists. 1491 and no free thread exists. Please note that queueing up a hundred
1492 requests can create demand for a hundred threads, even if it turns
1493 out that everything is in the cache and could have been processed
1494 faster by a single thread.
326 1495
327 It is recommended to keep the number of threads low, as some Linux 1496 It is recommended to keep the number of threads relatively low, as
328 kernel versions will scale negatively with the number of threads 1497 some Linux kernel versions will scale negatively with the number of
329 (higher parallelity => MUCH higher latency). With current Linux 2.6 1498 threads (higher parallelity => MUCH higher latency). With current
330 versions, 4-32 threads should be fine. 1499 Linux 2.6 versions, 4-32 threads should be fine.
331 1500
332 Under most circumstances you don't need to call this function, as 1501 Under most circumstances you don't need to call this function, as
333 the module selects a default that is suitable for low to moderate 1502 the module selects a default that is suitable for low to moderate
334 load. 1503 load.
335 1504
… …
345 ensure that all threads are killed and that there are no outstanding 1514 ensure that all threads are killed and that there are no outstanding
346 requests. 1515 requests.
347 1516
348 Under normal circumstances you don't need to call this function. 1517 Under normal circumstances you don't need to call this function.
349 1518
1519 IO::AIO::max_idle $nthreads
1520 Limit the number of threads (default: 4) that are allowed to idle
1521 (i.e., threads that did not get a request to process within the idle
1522 timeout (default: 10 seconds). That means if a thread becomes idle
1523 while $nthreads other threads are also idle, it will free its
1524 resources and exit.
1525
1526 This is useful when you allow a large number of threads (e.g. 100 or
1527 1000) to allow for extremely high load situations, but want to free
1528 resources under normal circumstances (1000 threads can easily
1529 consume 30MB of RAM).
1530
1531 The default is probably ok in most situations, especially if thread
1532 creation is fast. If thread creation is very slow on your system you
1533 might want to use larger values.
1534
1535 IO::AIO::idle_timeout $seconds
1536 Sets the minimum idle timeout (default 10) after which worker
1537 threads are allowed to exit. SEe "IO::AIO::max_idle".
1538
350 $oldnreqs = IO::AIO::max_outstanding $nreqs 1539 IO::AIO::max_outstanding $maxreqs
351 Sets the maximum number of outstanding requests to $nreqs. If you 1540 Sets the maximum number of outstanding requests to $nreqs. If you do
352 try to queue up more than this number of requests, the caller will 1541 queue up more than this number of requests, the next call to
1542 "IO::AIO::poll_cb" (and other functions calling "poll_cb", such as
1543 "IO::AIO::flush" or "IO::AIO::poll") will block until the limit is
1544 no longer exceeded.
1545
1546 In other words, this setting does not enforce a queue limit, but can
1547 be used to make poll functions block if the limit is exceeded.
1548
1549 This is a very bad function to use in interactive programs because
1550 it blocks, and a bad way to reduce concurrency because it is
1551 inexact: Better use an "aio_group" together with a feed callback.
1552
1553 Its main use is in scripts without an event loop - when you want to
1554 stat a lot of files, you can write something like this:
1555
1556 IO::AIO::max_outstanding 32;
1557
1558 for my $path (...) {
1559 aio_stat $path , ...;
1560 IO::AIO::poll_cb;
1561 }
1562
1563 IO::AIO::flush;
1564
1565 The call to "poll_cb" inside the loop will normally return
1566 instantly, but as soon as more thna 32 reqeusts are in-flight, it
353 block until some requests have been handled. 1567 will block until some requests have been handled. This keeps the
1568 loop from pushing a large number of "aio_stat" requests onto the
1569 queue.
354 1570
355 The default is very large, so normally there is no practical limit. 1571 The default value for "max_outstanding" is very large, so there is
356 If you queue up many requests in a loop it often improves speed if 1572 no practical limit on the number of outstanding requests.
357 you set this to a relatively low number, such as 100.
358 1573
359 Under normal circumstances you don't need to call this function. 1574 STATISTICAL INFORMATION
1575 IO::AIO::nreqs
1576 Returns the number of requests currently in the ready, execute or
1577 pending states (i.e. for which their callback has not been invoked
1578 yet).
1579
1580 Example: wait till there are no outstanding requests anymore:
1581
1582 IO::AIO::poll_wait, IO::AIO::poll_cb
1583 while IO::AIO::nreqs;
1584
1585 IO::AIO::nready
1586 Returns the number of requests currently in the ready state (not yet
1587 executed).
1588
1589 IO::AIO::npending
1590 Returns the number of requests currently in the pending state
1591 (executed, but not yet processed by poll_cb).
1592
1593 MISCELLANEOUS FUNCTIONS
1594 IO::AIO implements some functions that are useful when you want to use
1595 some "Advanced I/O" function not available to in Perl, without going the
1596 "Asynchronous I/O" route. Many of these have an asynchronous "aio_*"
1597 counterpart.
1598
1599 $numfd = IO::AIO::get_fdlimit
1600 This function is *EXPERIMENTAL* and subject to change.
1601
1602 Tries to find the current file descriptor limit and returns it, or
1603 "undef" and sets $! in case of an error. The limit is one larger
1604 than the highest valid file descriptor number.
1605
1606 IO::AIO::min_fdlimit [$numfd]
1607 This function is *EXPERIMENTAL* and subject to change.
1608
1609 Try to increase the current file descriptor limit(s) to at least
1610 $numfd by changing the soft or hard file descriptor resource limit.
1611 If $numfd is missing, it will try to set a very high limit, although
1612 this is not recommended when you know the actual minimum that you
1613 require.
1614
1615 If the limit cannot be raised enough, the function makes a
1616 best-effort attempt to increase the limit as much as possible, using
1617 various tricks, while still failing. You can query the resulting
1618 limit using "IO::AIO::get_fdlimit".
1619
1620 If an error occurs, returns "undef" and sets $!, otherwise returns
1621 true.
1622
1623 IO::AIO::sendfile $ofh, $ifh, $offset, $count
1624 Calls the "eio_sendfile_sync" function, which is like
1625 "aio_sendfile", but is blocking (this makes most sense if you know
1626 the input data is likely cached already and the output filehandle is
1627 set to non-blocking operations).
1628
1629 Returns the number of bytes copied, or -1 on error.
1630
1631 IO::AIO::fadvise $fh, $offset, $len, $advice
1632 Simply calls the "posix_fadvise" function (see its manpage for
1633 details). The following advice constants are available:
1634 "IO::AIO::FADV_NORMAL", "IO::AIO::FADV_SEQUENTIAL",
1635 "IO::AIO::FADV_RANDOM", "IO::AIO::FADV_NOREUSE",
1636 "IO::AIO::FADV_WILLNEED", "IO::AIO::FADV_DONTNEED".
1637
1638 On systems that do not implement "posix_fadvise", this function
1639 returns ENOSYS, otherwise the return value of "posix_fadvise".
1640
1641 IO::AIO::madvise $scalar, $offset, $len, $advice
1642 Simply calls the "posix_madvise" function (see its manpage for
1643 details). The following advice constants are available:
1644 "IO::AIO::MADV_NORMAL", "IO::AIO::MADV_SEQUENTIAL",
1645 "IO::AIO::MADV_RANDOM", "IO::AIO::MADV_WILLNEED",
1646 "IO::AIO::MADV_DONTNEED".
1647
1648 If $offset is negative, counts from the end. If $length is negative,
1649 the remaining length of the $scalar is used. If possible, $length
1650 will be reduced to fit into the $scalar.
1651
1652 On systems that do not implement "posix_madvise", this function
1653 returns ENOSYS, otherwise the return value of "posix_madvise".
1654
1655 IO::AIO::mprotect $scalar, $offset, $len, $protect
1656 Simply calls the "mprotect" function on the preferably AIO::mmap'ed
1657 $scalar (see its manpage for details). The following protect
1658 constants are available: "IO::AIO::PROT_NONE", "IO::AIO::PROT_READ",
1659 "IO::AIO::PROT_WRITE", "IO::AIO::PROT_EXEC".
1660
1661 If $offset is negative, counts from the end. If $length is negative,
1662 the remaining length of the $scalar is used. If possible, $length
1663 will be reduced to fit into the $scalar.
1664
1665 On systems that do not implement "mprotect", this function returns
1666 ENOSYS, otherwise the return value of "mprotect".
1667
1668 IO::AIO::mmap $scalar, $length, $prot, $flags, $fh[, $offset]
1669 Memory-maps a file (or anonymous memory range) and attaches it to
1670 the given $scalar, which will act like a string scalar. Returns true
1671 on success, and false otherwise.
1672
1673 The scalar must exist, but its contents do not matter - this means
1674 you cannot use a nonexistant array or hash element. When in doubt,
1675 "undef" the scalar first.
1676
1677 The only operations allowed on the mmapped scalar are
1678 "substr"/"vec", which don't change the string length, and most
1679 read-only operations such as copying it or searching it with regexes
1680 and so on.
1681
1682 Anything else is unsafe and will, at best, result in memory leaks.
1683
1684 The memory map associated with the $scalar is automatically removed
1685 when the $scalar is undef'd or destroyed, or when the
1686 "IO::AIO::mmap" or "IO::AIO::munmap" functions are called on it.
1687
1688 This calls the "mmap"(2) function internally. See your system's
1689 manual page for details on the $length, $prot and $flags parameters.
1690
1691 The $length must be larger than zero and smaller than the actual
1692 filesize.
1693
1694 $prot is a combination of "IO::AIO::PROT_NONE",
1695 "IO::AIO::PROT_EXEC", "IO::AIO::PROT_READ" and/or
1696 "IO::AIO::PROT_WRITE",
1697
1698 $flags can be a combination of "IO::AIO::MAP_SHARED" or
1699 "IO::AIO::MAP_PRIVATE", or a number of system-specific flags (when
1700 not available, the are 0): "IO::AIO::MAP_ANONYMOUS" (which is set to
1701 "MAP_ANON" if your system only provides this constant),
1702 "IO::AIO::MAP_LOCKED", "IO::AIO::MAP_NORESERVE",
1703 "IO::AIO::MAP_POPULATE", "IO::AIO::MAP_NONBLOCK",
1704 "IO::AIO::MAP_FIXED", "IO::AIO::MAP_GROWSDOWN",
1705 "IO::AIO::MAP_32BIT", "IO::AIO::MAP_HUGETLB" or
1706 "IO::AIO::MAP_STACK".
1707
1708 If $fh is "undef", then a file descriptor of -1 is passed.
1709
1710 $offset is the offset from the start of the file - it generally must
1711 be a multiple of "IO::AIO::PAGESIZE" and defaults to 0.
1712
1713 Example:
1714
1715 use Digest::MD5;
1716 use IO::AIO;
1717
1718 open my $fh, "<verybigfile"
1719 or die "$!";
1720
1721 IO::AIO::mmap my $data, -s $fh, IO::AIO::PROT_READ, IO::AIO::MAP_SHARED, $fh
1722 or die "verybigfile: $!";
1723
1724 my $fast_md5 = md5 $data;
1725
1726 IO::AIO::munmap $scalar
1727 Removes a previous mmap and undefines the $scalar.
1728
1729 IO::AIO::munlock $scalar, $offset = 0, $length = undef
1730 Calls the "munlock" function, undoing the effects of a previous
1731 "aio_mlock" call (see its description for details).
1732
1733 IO::AIO::munlockall
1734 Calls the "munlockall" function.
1735
1736 On systems that do not implement "munlockall", this function returns
1737 ENOSYS, otherwise the return value of "munlockall".
1738
1739 IO::AIO::splice $r_fh, $r_off, $w_fh, $w_off, $length, $flags
1740 Calls the GNU/Linux splice(2) syscall, if available. If $r_off or
1741 $w_off are "undef", then "NULL" is passed for these, otherwise they
1742 should be the file offset.
1743
1744 $r_fh and $w_fh should not refer to the same file, as splice might
1745 silently corrupt the data in this case.
1746
1747 The following symbol flag values are available:
1748 "IO::AIO::SPLICE_F_MOVE", "IO::AIO::SPLICE_F_NONBLOCK",
1749 "IO::AIO::SPLICE_F_MORE" and "IO::AIO::SPLICE_F_GIFT".
1750
1751 See the splice(2) manpage for details.
1752
1753 IO::AIO::tee $r_fh, $w_fh, $length, $flags
1754 Calls the GNU/Linux tee(2) syscall, see its manpage and the
1755 description for "IO::AIO::splice" above for details.
1756
1757 $actual_size = IO::AIO::pipesize $r_fh[, $new_size]
1758 Attempts to query or change the pipe buffer size. Obviously works
1759 only on pipes, and currently works only on GNU/Linux systems, and
1760 fails with -1/"ENOSYS" everywhere else. If anybody knows how to
1761 influence pipe buffer size on other systems, drop me a note.
1762
1763 ($rfh, $wfh) = IO::AIO::pipe2 [$flags]
1764 This is a direct interface to the Linux pipe2(2) system call. If
1765 $flags is missing or 0, then this should be the same as a call to
1766 perl's built-in "pipe" function and create a new pipe, and works on
1767 systems that lack the pipe2 syscall. On win32, this case invokes
1768 "_pipe (..., 4096, O_BINARY)".
1769
1770 If $flags is non-zero, it tries to invoke the pipe2 system call with
1771 the given flags (Linux 2.6.27, glibc 2.9).
1772
1773 On success, the read and write file handles are returned.
1774
1775 On error, nothing will be returned. If the pipe2 syscall is missing
1776 and $flags is non-zero, fails with "ENOSYS".
1777
1778 Please refer to pipe2(2) for more info on the $flags, but at the
1779 time of this writing, "IO::AIO::O_CLOEXEC", "IO::AIO::O_NONBLOCK"
1780 and "IO::AIO::O_DIRECT" (Linux 3.4, for packet-based pipes) were
1781 supported.
1782
1783 Example: create a pipe race-free w.r.t. threads and fork:
1784
1785 my ($rfh, $wfh) = IO::AIO::pipe2 IO::AIO::O_CLOEXEC
1786 or die "pipe2: $!\n";
1787
1788 $fh = IO::AIO::eventfd [$initval, [$flags]]
1789 This is a direct interface to the Linux eventfd(2) system call. The
1790 (unhelpful) defaults for $initval and $flags are 0 for both.
1791
1792 On success, the new eventfd filehandle is returned, otherwise
1793 returns "undef". If the eventfd syscall is missing, fails with
1794 "ENOSYS".
1795
1796 Please refer to eventfd(2) for more info on this call.
1797
1798 The following symbol flag values are available:
1799 "IO::AIO::EFD_CLOEXEC", "IO::AIO::EFD_NONBLOCK" and
1800 "IO::AIO::EFD_SEMAPHORE" (Linux 2.6.30).
1801
1802 Example: create a new eventfd filehandle:
1803
1804 $fh = IO::AIO::eventfd 0, IO::AIO::O_CLOEXEC
1805 or die "eventfd: $!\n";
1806
1807 $fh = IO::AIO::timerfd_create $clockid[, $flags]
1808 This is a direct interface to the Linux timerfd_create(2) system
1809 call. The (unhelpful) default for $flags is 0.
1810
1811 On success, the new timerfd filehandle is returned, otherwise
1812 returns "undef". If the eventfd syscall is missing, fails with
1813 "ENOSYS".
1814
1815 Please refer to timerfd_create(2) for more info on this call.
1816
1817 The following $clockid values are available:
1818 "IO::AIO::CLOCK_REALTIME", "IO::AIO::CLOCK_MONOTONIC"
1819 "IO::AIO::CLOCK_CLOCK_BOOTTIME" (Linux 3.15)
1820 "IO::AIO::CLOCK_CLOCK_REALTIME_ALARM" (Linux 3.11) and
1821 "IO::AIO::CLOCK_CLOCK_BOOTTIME_ALARM" (Linux 3.11).
1822
1823 The following $flags values are available (Linux 2.6.27):
1824 "IO::AIO::TFD_NONBLOCK" and "IO::AIO::TFD_CLOEXEC".
1825
1826 Example: create a new timerfd and set it to one-second repeated
1827 alarms, then wait for two alarms:
1828
1829 my $fh = IO::AIO::timerfd_create IO::AIO::CLOCK_BOOTTIME, IO::AIO::TFD_CLOEXEC
1830 or die "timerfd_create: $!\n";
1831
1832 defined IO::AIO::timerfd_settime $fh, 0, 1, 1
1833 or die "timerfd_settime: $!\n";
1834
1835 for (1..2) {
1836 8 == sysread $fh, my $buf, 8
1837 or die "timerfd read failure\n";
1838
1839 printf "number of expirations (likely 1): %d\n",
1840 unpack "Q", $buf;
1841 }
1842
1843 ($cur_interval, $cur_value) = IO::AIO::timerfd_settime $fh, $flags,
1844 $new_interval, $nbw_value
1845 This is a direct interface to the Linux timerfd_settime(2) system
1846 call. Please refer to its manpage for more info on this call.
1847
1848 The new itimerspec is specified using two (possibly fractional)
1849 second values, $new_interval and $new_value).
1850
1851 On success, the current interval and value are returned (as per
1852 "timerfd_gettime"). On failure, the empty list is returned.
1853
1854 The following $flags values are available:
1855 "IO::AIO::TFD_TIMER_ABSTIME" and "IO::AIO::TFD_TIMER_CANCEL_ON_SET".
1856
1857 See "IO::AIO::timerfd_create" for a full example.
1858
1859 ($cur_interval, $cur_value) = IO::AIO::timerfd_gettime $fh
1860 This is a direct interface to the Linux timerfd_gettime(2) system
1861 call. Please refer to its manpage for more info on this call.
1862
1863 On success, returns the current values of interval and value for the
1864 given timerfd (as potentially fractional second values). On failure,
1865 the empty list is returned.
1866
1867EVENT LOOP INTEGRATION
1868 It is recommended to use AnyEvent::AIO to integrate IO::AIO
1869 automatically into many event loops:
1870
1871 # AnyEvent integration (EV, Event, Glib, Tk, POE, urxvt, pureperl...)
1872 use AnyEvent::AIO;
1873
1874 You can also integrate IO::AIO manually into many event loops, here are
1875 some examples of how to do this:
1876
1877 # EV integration
1878 my $aio_w = EV::io IO::AIO::poll_fileno, EV::READ, \&IO::AIO::poll_cb;
1879
1880 # Event integration
1881 Event->io (fd => IO::AIO::poll_fileno,
1882 poll => 'r',
1883 cb => \&IO::AIO::poll_cb);
1884
1885 # Glib/Gtk2 integration
1886 add_watch Glib::IO IO::AIO::poll_fileno,
1887 in => sub { IO::AIO::poll_cb; 1 };
1888
1889 # Tk integration
1890 Tk::Event::IO->fileevent (IO::AIO::poll_fileno, "",
1891 readable => \&IO::AIO::poll_cb);
1892
1893 # Danga::Socket integration
1894 Danga::Socket->AddOtherFds (IO::AIO::poll_fileno =>
1895 \&IO::AIO::poll_cb);
360 1896
361 FORK BEHAVIOUR 1897 FORK BEHAVIOUR
362 Before the fork, IO::AIO enters a quiescent state where no requests can 1898 Usage of pthreads in a program changes the semantics of fork
363 be added in other threads and no results will be processed. After the 1899 considerably. Specifically, only async-safe functions can be called
364 fork the parent simply leaves the quiescent state and continues 1900 after fork. Perl doesn't know about this, so in general, you cannot call
365 request/result processing, while the child clears the request/result 1901 fork with defined behaviour in perl if pthreads are involved. IO::AIO
366 queue (so the requests started before the fork will only be handled in 1902 uses pthreads, so this applies, but many other extensions and (for
367 the parent). Threats will be started on demand until the limit ste in 1903 inexplicable reasons) perl itself often is linked against pthreads, so
368 the parent process has been reached again. 1904 this limitation applies to quite a lot of perls.
1905
1906 This module no longer tries to fight your OS, or POSIX. That means
1907 IO::AIO only works in the process that loaded it. Forking is fully
1908 supported, but using IO::AIO in the child is not.
1909
1910 You might get around by not *using* IO::AIO before (or after) forking.
1911 You could also try to call the IO::AIO::reinit function in the child:
1912
1913 IO::AIO::reinit
1914 Abandons all current requests and I/O threads and simply
1915 reinitialises all data structures. This is not an operation
1916 supported by any standards, but happens to work on GNU/Linux and
1917 some newer BSD systems.
1918
1919 The only reasonable use for this function is to call it after
1920 forking, if "IO::AIO" was used in the parent. Calling it while
1921 IO::AIO is active in the process will result in undefined behaviour.
1922 Calling it at any time will also result in any undefined (by POSIX)
1923 behaviour.
1924
1925 LINUX-SPECIFIC CALLS
1926 When a call is documented as "linux-specific" then this means it
1927 originated on GNU/Linux. "IO::AIO" will usually try to autodetect the
1928 availability and compatibility of such calls regardless of the platform
1929 it is compiled on, so platforms such as FreeBSD which often implement
1930 these calls will work. When in doubt, call them and see if they fail wth
1931 "ENOSYS".
1932
1933 MEMORY USAGE
1934 Per-request usage:
1935
1936 Each aio request uses - depending on your architecture - around 100-200
1937 bytes of memory. In addition, stat requests need a stat buffer (possibly
1938 a few hundred bytes), readdir requires a result buffer and so on. Perl
1939 scalars and other data passed into aio requests will also be locked and
1940 will consume memory till the request has entered the done state.
1941
1942 This is not awfully much, so queuing lots of requests is not usually a
1943 problem.
1944
1945 Per-thread usage:
1946
1947 In the execution phase, some aio requests require more memory for
1948 temporary buffers, and each thread requires a stack and other data
1949 structures (usually around 16k-128k, depending on the OS).
1950
1951KNOWN BUGS
1952 Known bugs will be fixed in the next release :)
1953
1954KNOWN ISSUES
1955 Calls that try to "import" foreign memory areas (such as "IO::AIO::mmap"
1956 or "IO::AIO::aio_slurp") do not work with generic lvalues, such as
1957 non-created hash slots or other scalars I didn't think of. It's best to
1958 avoid such and either use scalar variables or making sure that the
1959 scalar exists (e.g. by storing "undef") and isn't "funny" (e.g. tied).
1960
1961 I am not sure anything can be done about this, so this is considered a
1962 known issue, rather than a bug.
369 1963
370SEE ALSO 1964SEE ALSO
371 Coro, Linux::AIO. 1965 AnyEvent::AIO for easy integration into event loops, Coro::AIO for a
1966 more natural syntax.
372 1967
373AUTHOR 1968AUTHOR
374 Marc Lehmann <schmorp@schmorp.de> 1969 Marc Lehmann <schmorp@schmorp.de>
375 http://home.schmorp.de/ 1970 http://home.schmorp.de/
376 1971

Diff Legend

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