1 |
=head1 NAME |
2 |
|
3 |
IO::AIO - Asynchronous Input/Output |
4 |
|
5 |
=head1 SYNOPSIS |
6 |
|
7 |
use IO::AIO; |
8 |
|
9 |
aio_open "/etc/passwd", O_RDONLY, 0, sub { |
10 |
my ($fh) = @_; |
11 |
... |
12 |
}; |
13 |
|
14 |
aio_unlink "/tmp/file", sub { }; |
15 |
|
16 |
aio_read $fh, 30000, 1024, $buffer, 0, sub { |
17 |
$_[0] > 0 or die "read error: $!"; |
18 |
}; |
19 |
|
20 |
# version 2+ has request and group objects |
21 |
use IO::AIO 2; |
22 |
|
23 |
my $req = aio_unlink "/tmp/file", sub { }; |
24 |
$req->cancel; # cancel request if still in queue |
25 |
|
26 |
my $grp = aio_group sub { print "all stats done\n" }; |
27 |
add $grp aio_stat "..." for ...; |
28 |
|
29 |
# AnyEvent integration |
30 |
open my $fh, "<&=" . IO::AIO::poll_fileno or die "$!"; |
31 |
my $w = AnyEvent->io (fh => $fh, poll => 'r', cb => sub { IO::AIO::poll_cb }); |
32 |
|
33 |
# Event integration |
34 |
Event->io (fd => IO::AIO::poll_fileno, |
35 |
poll => 'r', |
36 |
cb => \&IO::AIO::poll_cb); |
37 |
|
38 |
# Glib/Gtk2 integration |
39 |
add_watch Glib::IO IO::AIO::poll_fileno, |
40 |
in => sub { IO::AIO::poll_cb; 1 }; |
41 |
|
42 |
# Tk integration |
43 |
Tk::Event::IO->fileevent (IO::AIO::poll_fileno, "", |
44 |
readable => \&IO::AIO::poll_cb); |
45 |
|
46 |
# Danga::Socket integration |
47 |
Danga::Socket->AddOtherFds (IO::AIO::poll_fileno => |
48 |
\&IO::AIO::poll_cb); |
49 |
|
50 |
=head1 DESCRIPTION |
51 |
|
52 |
This module implements asynchronous I/O using whatever means your |
53 |
operating system supports. |
54 |
|
55 |
Currently, a number of threads are started that execute your read/writes |
56 |
and signal their completion. You don't need thread support in your libc or |
57 |
perl, and the threads created by this module will not be visible to the |
58 |
pthreads library. In the future, this module might make use of the native |
59 |
aio functions available on many operating systems. However, they are often |
60 |
not well-supported (Linux doesn't allow them on normal files currently, |
61 |
for example), and they would only support aio_read and aio_write, so the |
62 |
remaining functionality would have to be implemented using threads anyway. |
63 |
|
64 |
Although the module will work with in the presence of other threads, it is |
65 |
currently not reentrant, so use appropriate locking yourself, always call |
66 |
C<poll_cb> from within the same thread, or never call C<poll_cb> (or other |
67 |
C<aio_> functions) recursively. |
68 |
|
69 |
=cut |
70 |
|
71 |
package IO::AIO; |
72 |
|
73 |
no warnings; |
74 |
use strict 'vars'; |
75 |
|
76 |
use base 'Exporter'; |
77 |
|
78 |
BEGIN { |
79 |
our $VERSION = '2.0'; |
80 |
|
81 |
our @EXPORT = qw(aio_sendfile aio_read aio_write aio_open aio_close aio_stat |
82 |
aio_lstat aio_unlink aio_rmdir aio_readdir aio_scandir aio_symlink |
83 |
aio_fsync aio_fdatasync aio_readahead aio_rename aio_link aio_move |
84 |
aio_group); |
85 |
our @EXPORT_OK = qw(poll_fileno poll_cb min_parallel max_parallel max_outstanding nreqs); |
86 |
|
87 |
@IO::AIO::GRP::ISA = 'IO::AIO::REQ'; |
88 |
|
89 |
require XSLoader; |
90 |
XSLoader::load ("IO::AIO", $VERSION); |
91 |
} |
92 |
|
93 |
=head1 FUNCTIONS |
94 |
|
95 |
=head2 AIO FUNCTIONS |
96 |
|
97 |
All the C<aio_*> calls are more or less thin wrappers around the syscall |
98 |
with the same name (sans C<aio_>). The arguments are similar or identical, |
99 |
and they all accept an additional (and optional) C<$callback> argument |
100 |
which must be a code reference. This code reference will get called with |
101 |
the syscall return code (e.g. most syscalls return C<-1> on error, unlike |
102 |
perl, which usually delivers "false") as it's sole argument when the given |
103 |
syscall has been executed asynchronously. |
104 |
|
105 |
All functions expecting a filehandle keep a copy of the filehandle |
106 |
internally until the request has finished. |
107 |
|
108 |
All requests return objects of type L<IO::AIO::REQ> that allow further |
109 |
manipulation of those requests while they are in-flight. |
110 |
|
111 |
The pathnames you pass to these routines I<must> be absolute and |
112 |
encoded in byte form. The reason for the former is that at the time the |
113 |
request is being executed, the current working directory could have |
114 |
changed. Alternatively, you can make sure that you never change the |
115 |
current working directory. |
116 |
|
117 |
To encode pathnames to byte form, either make sure you either: a) |
118 |
always pass in filenames you got from outside (command line, readdir |
119 |
etc.), b) are ASCII or ISO 8859-1, c) use the Encode module and encode |
120 |
your pathnames to the locale (or other) encoding in effect in the user |
121 |
environment, d) use Glib::filename_from_unicode on unicode filenames or e) |
122 |
use something else. |
123 |
|
124 |
=over 4 |
125 |
|
126 |
=item aio_open $pathname, $flags, $mode, $callback->($fh) |
127 |
|
128 |
Asynchronously open or create a file and call the callback with a newly |
129 |
created filehandle for the file. |
130 |
|
131 |
The pathname passed to C<aio_open> must be absolute. See API NOTES, above, |
132 |
for an explanation. |
133 |
|
134 |
The C<$flags> argument is a bitmask. See the C<Fcntl> module for a |
135 |
list. They are the same as used by C<sysopen>. |
136 |
|
137 |
Likewise, C<$mode> specifies the mode of the newly created file, if it |
138 |
didn't exist and C<O_CREAT> has been given, just like perl's C<sysopen>, |
139 |
except that it is mandatory (i.e. use C<0> if you don't create new files, |
140 |
and C<0666> or C<0777> if you do). |
141 |
|
142 |
Example: |
143 |
|
144 |
aio_open "/etc/passwd", O_RDONLY, 0, sub { |
145 |
if ($_[0]) { |
146 |
print "open successful, fh is $_[0]\n"; |
147 |
... |
148 |
} else { |
149 |
die "open failed: $!\n"; |
150 |
} |
151 |
}; |
152 |
|
153 |
=item aio_close $fh, $callback->($status) |
154 |
|
155 |
Asynchronously close a file and call the callback with the result |
156 |
code. I<WARNING:> although accepted, you should not pass in a perl |
157 |
filehandle here, as perl will likely close the file descriptor another |
158 |
time when the filehandle is destroyed. Normally, you can safely call perls |
159 |
C<close> or just let filehandles go out of scope. |
160 |
|
161 |
This is supposed to be a bug in the API, so that might change. It's |
162 |
therefore best to avoid this function. |
163 |
|
164 |
=item aio_read $fh,$offset,$length, $data,$dataoffset, $callback->($retval) |
165 |
|
166 |
=item aio_write $fh,$offset,$length, $data,$dataoffset, $callback->($retval) |
167 |
|
168 |
Reads or writes C<length> bytes from the specified C<fh> and C<offset> |
169 |
into the scalar given by C<data> and offset C<dataoffset> and calls the |
170 |
callback without the actual number of bytes read (or -1 on error, just |
171 |
like the syscall). |
172 |
|
173 |
The C<$data> scalar I<MUST NOT> be modified in any way while the request |
174 |
is outstanding. Modifying it can result in segfaults or WW3 (if the |
175 |
necessary/optional hardware is installed). |
176 |
|
177 |
Example: Read 15 bytes at offset 7 into scalar C<$buffer>, starting at |
178 |
offset C<0> within the scalar: |
179 |
|
180 |
aio_read $fh, 7, 15, $buffer, 0, sub { |
181 |
$_[0] > 0 or die "read error: $!"; |
182 |
print "read $_[0] bytes: <$buffer>\n"; |
183 |
}; |
184 |
|
185 |
=item aio_move $srcpath, $dstpath, $callback->($status) |
186 |
|
187 |
[EXPERIMENTAL due to internal aio_group use] |
188 |
|
189 |
Try to move the I<file> (directories not supported as either source or |
190 |
destination) from C<$srcpath> to C<$dstpath> and call the callback with |
191 |
the C<0> (error) or C<-1> ok. |
192 |
|
193 |
This is a composite request that tries to rename(2) the file first. If |
194 |
rename files with C<EXDEV>, it creates the destination file with mode 0200 |
195 |
and copies the contents of the source file into it using C<aio_sendfile>, |
196 |
followed by restoring atime, mtime, access mode and uid/gid, in that |
197 |
order, and unlinking the C<$srcpath>. |
198 |
|
199 |
If an error occurs, the partial destination file will be unlinked, if |
200 |
possible, except when setting atime, mtime, access mode and uid/gid, where |
201 |
errors are being ignored. |
202 |
|
203 |
=cut |
204 |
|
205 |
sub aio_move($$$) { |
206 |
my ($src, $dst, $cb) = @_; |
207 |
|
208 |
my $grp = aio_group $cb; |
209 |
|
210 |
add $grp aio_rename $src, $dst, sub { |
211 |
if ($_[0] && $! == EXDEV) { |
212 |
add $grp aio_open $src, O_RDONLY, 0, sub { |
213 |
if (my $src_fh = $_[0]) { |
214 |
my @stat = stat $src_fh; |
215 |
|
216 |
add $grp aio_open $dst, O_WRONLY, 0200, sub { |
217 |
if (my $dst_fh = $_[0]) { |
218 |
add $grp aio_sendfile $dst_fh, $src_fh, 0, $stat[7], sub { |
219 |
close $src_fh; |
220 |
|
221 |
if ($_[0] == $stat[7]) { |
222 |
utime $stat[8], $stat[9], $dst; |
223 |
chmod $stat[2] & 07777, $dst_fh; |
224 |
chown $stat[4], $stat[5], $dst_fh; |
225 |
close $dst_fh; |
226 |
|
227 |
add $grp aio_unlink $src, sub { |
228 |
$grp->result ($_[0]); |
229 |
}; |
230 |
} else { |
231 |
my $errno = $!; |
232 |
add $grp aio_unlink $dst, sub { |
233 |
$! = $errno; |
234 |
$grp->result (-1); |
235 |
}; |
236 |
} |
237 |
}; |
238 |
} else { |
239 |
$grp->result (-1); |
240 |
} |
241 |
}, |
242 |
|
243 |
} else { |
244 |
$grp->result (-1); |
245 |
} |
246 |
}; |
247 |
} else { |
248 |
$grp->result ($_[0]); |
249 |
} |
250 |
}; |
251 |
|
252 |
$grp |
253 |
} |
254 |
|
255 |
=item aio_sendfile $out_fh, $in_fh, $in_offset, $length, $callback->($retval) |
256 |
|
257 |
Tries to copy C<$length> bytes from C<$in_fh> to C<$out_fh>. It starts |
258 |
reading at byte offset C<$in_offset>, and starts writing at the current |
259 |
file offset of C<$out_fh>. Because of that, it is not safe to issue more |
260 |
than one C<aio_sendfile> per C<$out_fh>, as they will interfere with each |
261 |
other. |
262 |
|
263 |
This call tries to make use of a native C<sendfile> syscall to provide |
264 |
zero-copy operation. For this to work, C<$out_fh> should refer to a |
265 |
socket, and C<$in_fh> should refer to mmap'able file. |
266 |
|
267 |
If the native sendfile call fails or is not implemented, it will be |
268 |
emulated, so you can call C<aio_sendfile> on any type of filehandle |
269 |
regardless of the limitations of the operating system. |
270 |
|
271 |
Please note, however, that C<aio_sendfile> can read more bytes from |
272 |
C<$in_fh> than are written, and there is no way to find out how many |
273 |
bytes have been read from C<aio_sendfile> alone, as C<aio_sendfile> only |
274 |
provides the number of bytes written to C<$out_fh>. Only if the result |
275 |
value equals C<$length> one can assume that C<$length> bytes have been |
276 |
read. |
277 |
|
278 |
=item aio_readahead $fh,$offset,$length, $callback->($retval) |
279 |
|
280 |
C<aio_readahead> populates the page cache with data from a file so that |
281 |
subsequent reads from that file will not block on disk I/O. The C<$offset> |
282 |
argument specifies the starting point from which data is to be read and |
283 |
C<$length> specifies the number of bytes to be read. I/O is performed in |
284 |
whole pages, so that offset is effectively rounded down to a page boundary |
285 |
and bytes are read up to the next page boundary greater than or equal to |
286 |
(off-set+length). C<aio_readahead> does not read beyond the end of the |
287 |
file. The current file offset of the file is left unchanged. |
288 |
|
289 |
If that syscall doesn't exist (likely if your OS isn't Linux) it will be |
290 |
emulated by simply reading the data, which would have a similar effect. |
291 |
|
292 |
=item aio_stat $fh_or_path, $callback->($status) |
293 |
|
294 |
=item aio_lstat $fh, $callback->($status) |
295 |
|
296 |
Works like perl's C<stat> or C<lstat> in void context. The callback will |
297 |
be called after the stat and the results will be available using C<stat _> |
298 |
or C<-s _> etc... |
299 |
|
300 |
The pathname passed to C<aio_stat> must be absolute. See API NOTES, above, |
301 |
for an explanation. |
302 |
|
303 |
Currently, the stats are always 64-bit-stats, i.e. instead of returning an |
304 |
error when stat'ing a large file, the results will be silently truncated |
305 |
unless perl itself is compiled with large file support. |
306 |
|
307 |
Example: Print the length of F</etc/passwd>: |
308 |
|
309 |
aio_stat "/etc/passwd", sub { |
310 |
$_[0] and die "stat failed: $!"; |
311 |
print "size is ", -s _, "\n"; |
312 |
}; |
313 |
|
314 |
=item aio_unlink $pathname, $callback->($status) |
315 |
|
316 |
Asynchronously unlink (delete) a file and call the callback with the |
317 |
result code. |
318 |
|
319 |
=item aio_link $srcpath, $dstpath, $callback->($status) |
320 |
|
321 |
Asynchronously create a new link to the existing object at C<$srcpath> at |
322 |
the path C<$dstpath> and call the callback with the result code. |
323 |
|
324 |
=item aio_symlink $srcpath, $dstpath, $callback->($status) |
325 |
|
326 |
Asynchronously create a new symbolic link to the existing object at C<$srcpath> at |
327 |
the path C<$dstpath> and call the callback with the result code. |
328 |
|
329 |
=item aio_rename $srcpath, $dstpath, $callback->($status) |
330 |
|
331 |
Asynchronously rename the object at C<$srcpath> to C<$dstpath>, just as |
332 |
rename(2) and call the callback with the result code. |
333 |
|
334 |
=item aio_rmdir $pathname, $callback->($status) |
335 |
|
336 |
Asynchronously rmdir (delete) a directory and call the callback with the |
337 |
result code. |
338 |
|
339 |
=item aio_readdir $pathname, $callback->($entries) |
340 |
|
341 |
Unlike the POSIX call of the same name, C<aio_readdir> reads an entire |
342 |
directory (i.e. opendir + readdir + closedir). The entries will not be |
343 |
sorted, and will B<NOT> include the C<.> and C<..> entries. |
344 |
|
345 |
The callback a single argument which is either C<undef> or an array-ref |
346 |
with the filenames. |
347 |
|
348 |
=item aio_scandir $path, $maxreq, $callback->($dirs, $nondirs) |
349 |
|
350 |
[EXPERIMENTAL due to internal aio_group use] |
351 |
|
352 |
Scans a directory (similar to C<aio_readdir>) but additionally tries to |
353 |
separate the entries of directory C<$path> into two sets of names, ones |
354 |
you can recurse into (directories or links to them), and ones you cannot |
355 |
recurse into (everything else). |
356 |
|
357 |
C<aio_scandir> is a composite request that creates of many sub requests_ |
358 |
C<$maxreq> specifies the maximum number of outstanding aio requests that |
359 |
this function generates. If it is C<< <= 0 >>, then a suitable default |
360 |
will be chosen (currently 6). |
361 |
|
362 |
On error, the callback is called without arguments, otherwise it receives |
363 |
two array-refs with path-relative entry names. |
364 |
|
365 |
Example: |
366 |
|
367 |
aio_scandir $dir, 0, sub { |
368 |
my ($dirs, $nondirs) = @_; |
369 |
print "real directories: @$dirs\n"; |
370 |
print "everything else: @$nondirs\n"; |
371 |
}; |
372 |
|
373 |
Implementation notes. |
374 |
|
375 |
The C<aio_readdir> cannot be avoided, but C<stat()>'ing every entry can. |
376 |
|
377 |
After reading the directory, the modification time, size etc. of the |
378 |
directory before and after the readdir is checked, and if they match (and |
379 |
isn't the current time), the link count will be used to decide how many |
380 |
entries are directories (if >= 2). Otherwise, no knowledge of the number |
381 |
of subdirectories will be assumed. |
382 |
|
383 |
Then entries will be sorted into likely directories (everything without |
384 |
a non-initial dot currently) and likely non-directories (everything |
385 |
else). Then every entry plus an appended C</.> will be C<stat>'ed, |
386 |
likely directories first. If that succeeds, it assumes that the entry |
387 |
is a directory or a symlink to directory (which will be checked |
388 |
seperately). This is often faster than stat'ing the entry itself because |
389 |
filesystems might detect the type of the entry without reading the inode |
390 |
data (e.g. ext2fs filetype feature). |
391 |
|
392 |
If the known number of directories (link count - 2) has been reached, the |
393 |
rest of the entries is assumed to be non-directories. |
394 |
|
395 |
This only works with certainty on POSIX (= UNIX) filesystems, which |
396 |
fortunately are the vast majority of filesystems around. |
397 |
|
398 |
It will also likely work on non-POSIX filesystems with reduced efficiency |
399 |
as those tend to return 0 or 1 as link counts, which disables the |
400 |
directory counting heuristic. |
401 |
|
402 |
=cut |
403 |
|
404 |
sub aio_scandir($$$) { |
405 |
my ($path, $maxreq, $cb) = @_; |
406 |
|
407 |
my $grp = aio_group $cb; |
408 |
|
409 |
$maxreq = 6 if $maxreq <= 0; |
410 |
|
411 |
# stat once |
412 |
add $grp aio_stat $path, sub { |
413 |
return $grp->result () if $_[0]; |
414 |
my $now = time; |
415 |
my $hash1 = join ":", (stat _)[0,1,3,7,9]; |
416 |
|
417 |
# read the directory entries |
418 |
add $grp aio_readdir $path, sub { |
419 |
my $entries = shift |
420 |
or return $grp->result (); |
421 |
|
422 |
# stat the dir another time |
423 |
add $grp aio_stat $path, sub { |
424 |
my $hash2 = join ":", (stat _)[0,1,3,7,9]; |
425 |
|
426 |
my $ndirs; |
427 |
|
428 |
# take the slow route if anything looks fishy |
429 |
if ($hash1 ne $hash2 or (stat _)[9] == $now) { |
430 |
$ndirs = -1; |
431 |
} else { |
432 |
# if nlink == 2, we are finished |
433 |
# on non-posix-fs's, we rely on nlink < 2 |
434 |
$ndirs = (stat _)[3] - 2 |
435 |
or return $grp->result ([], $entries); |
436 |
} |
437 |
|
438 |
# sort into likely dirs and likely nondirs |
439 |
# dirs == files without ".", short entries first |
440 |
$entries = [map $_->[0], |
441 |
sort { $b->[1] cmp $a->[1] } |
442 |
map [$_, sprintf "%s%04d", (/.\./ ? "1" : "0"), length], |
443 |
@$entries]; |
444 |
|
445 |
my (@dirs, @nondirs); |
446 |
|
447 |
my ($statcb, $schedcb); |
448 |
my $nreq = 0; |
449 |
|
450 |
my $statgrp = add $grp aio_group; |
451 |
|
452 |
$schedcb = sub { |
453 |
if (@$entries) { |
454 |
if ($nreq < $maxreq) { |
455 |
my $ent = pop @$entries; |
456 |
$nreq++; |
457 |
add $statgrp aio_stat "$path/$ent/.", sub { $statcb->($_[0], $ent) }; |
458 |
} |
459 |
} elsif (!$nreq) { |
460 |
# finished |
461 |
$statgrp->cancel; |
462 |
undef $statcb; |
463 |
undef $schedcb; |
464 |
$grp->result (\@dirs, \@nondirs); |
465 |
} |
466 |
}; |
467 |
$statcb = sub { |
468 |
my ($status, $entry) = @_; |
469 |
|
470 |
if ($status < 0) { |
471 |
$nreq--; |
472 |
push @nondirs, $entry; |
473 |
&$schedcb; |
474 |
} else { |
475 |
# need to check for real directory |
476 |
add $grp aio_lstat "$path/$entry", sub { |
477 |
$nreq--; |
478 |
|
479 |
if (-d _) { |
480 |
push @dirs, $entry; |
481 |
|
482 |
if (!--$ndirs) { |
483 |
push @nondirs, @$entries; |
484 |
$entries = []; |
485 |
} |
486 |
} else { |
487 |
push @nondirs, $entry; |
488 |
} |
489 |
|
490 |
&$schedcb; |
491 |
} |
492 |
} |
493 |
}; |
494 |
|
495 |
&$schedcb while @$entries && $nreq < $maxreq; |
496 |
}; |
497 |
}; |
498 |
}; |
499 |
|
500 |
$grp |
501 |
} |
502 |
|
503 |
=item aio_fsync $fh, $callback->($status) |
504 |
|
505 |
Asynchronously call fsync on the given filehandle and call the callback |
506 |
with the fsync result code. |
507 |
|
508 |
=item aio_fdatasync $fh, $callback->($status) |
509 |
|
510 |
Asynchronously call fdatasync on the given filehandle and call the |
511 |
callback with the fdatasync result code. |
512 |
|
513 |
If this call isn't available because your OS lacks it or it couldn't be |
514 |
detected, it will be emulated by calling C<fsync> instead. |
515 |
|
516 |
=item aio_group $callback->(...) |
517 |
|
518 |
[EXPERIMENTAL] |
519 |
|
520 |
This is a very special aio request: Instead of doing something, it is a |
521 |
container for other aio requests, which is useful if you want to bundle |
522 |
many requests into a single, composite, request. |
523 |
|
524 |
Returns an object of class L<IO::AIO::GRP>. See its documentation below |
525 |
for more info. |
526 |
|
527 |
Example: |
528 |
|
529 |
my $grp = aio_group sub { |
530 |
print "all stats done\n"; |
531 |
}; |
532 |
|
533 |
add $grp |
534 |
(aio_stat ...), |
535 |
(aio_stat ...), |
536 |
...; |
537 |
|
538 |
=item IO::AIO::aio_sleep $fractional_seconds, $callback->() *NOT EXPORTED* |
539 |
|
540 |
Mainly used for debugging and benchmarking, this aio request puts one of |
541 |
the request workers to sleep for the given time. |
542 |
|
543 |
While it is theoretically handy to have simple I/O scheduling requests |
544 |
like sleep and file handle readable/writable, the overhead this creates |
545 |
is immense, so do not use this function except to put your application |
546 |
under artificial I/O pressure. |
547 |
|
548 |
=back |
549 |
|
550 |
=head2 IO::AIO::REQ CLASS |
551 |
|
552 |
All non-aggregate C<aio_*> functions return an object of this class when |
553 |
called in non-void context. |
554 |
|
555 |
A request always moves through the following five states in its lifetime, |
556 |
in order: B<ready> (request has been created, but has not been executed |
557 |
yet), B<execute> (request is currently being executed), B<pending> |
558 |
(request has been executed but callback has not been called yet), |
559 |
B<result> (results are being processed synchronously, includes calling the |
560 |
callback) and B<done> (request has reached the end of its lifetime and |
561 |
holds no resources anymore). |
562 |
|
563 |
=over 4 |
564 |
|
565 |
=item $req->cancel |
566 |
|
567 |
Cancels the request, if possible. Has the effect of skipping execution |
568 |
when entering the B<execute> state and skipping calling the callback when |
569 |
entering the the B<result> state, but will leave the request otherwise |
570 |
untouched. That means that requests that currently execute will not be |
571 |
stopped and resources held by the request will not be freed prematurely. |
572 |
|
573 |
=back |
574 |
|
575 |
=head2 IO::AIO::GRP CLASS |
576 |
|
577 |
This class is a subclass of L<IO::AIO::REQ>, so all its methods apply to |
578 |
objects of this class, too. |
579 |
|
580 |
A IO::AIO::GRP object is a special request that can contain multiple other |
581 |
aio requests. |
582 |
|
583 |
You create one by calling the C<aio_group> constructing function with a |
584 |
callback that will be called when all contained requests have entered the |
585 |
C<done> state: |
586 |
|
587 |
my $grp = aio_group sub { |
588 |
print "all requests are done\n"; |
589 |
}; |
590 |
|
591 |
You add requests by calling the C<add> method with one or more |
592 |
C<IO::AIO::REQ> objects: |
593 |
|
594 |
$grp->add (aio_unlink "..."); |
595 |
|
596 |
add $grp aio_stat "...", sub { |
597 |
$_[0] or return $grp->result ("error"); |
598 |
|
599 |
# add another request dynamically, if first succeeded |
600 |
add $grp aio_open "...", sub { |
601 |
$grp->result ("ok"); |
602 |
}; |
603 |
}; |
604 |
|
605 |
This makes it very easy to create composite requests (see the source of |
606 |
C<aio_move> for an application) that work and feel like simple requests. |
607 |
|
608 |
The IO::AIO::GRP objects will be cleaned up during calls to |
609 |
C<IO::AIO::poll_cb>, just like any other request. |
610 |
|
611 |
They can be canceled like any other request. Canceling will cancel not |
612 |
only the request itself, but also all requests it contains. |
613 |
|
614 |
They can also can also be added to other IO::AIO::GRP objects. |
615 |
|
616 |
You must not add requests to a group from within the group callback (or |
617 |
any later time). |
618 |
|
619 |
Their lifetime, simplified, looks like this: when they are empty, they |
620 |
will finish very quickly. If they contain only requests that are in the |
621 |
C<done> state, they will also finish. Otherwise they will continue to |
622 |
exist. |
623 |
|
624 |
That means after creating a group you have some time to add requests. And |
625 |
in the callbacks of those requests, you can add further requests to the |
626 |
group. And only when all those requests have finished will the the group |
627 |
itself finish. |
628 |
|
629 |
=over 4 |
630 |
|
631 |
=item $grp->add (...) |
632 |
|
633 |
=item add $grp ... |
634 |
|
635 |
Add one or more requests to the group. Any type of L<IO::AIO::REQ> can |
636 |
be added, including other groups, as long as you do not create circular |
637 |
dependencies. |
638 |
|
639 |
Returns all its arguments. |
640 |
|
641 |
=item $grp->result (...) |
642 |
|
643 |
Set the result value(s) that will be passed to the group callback when all |
644 |
subrequests have finished. By default, no argument will be passed. |
645 |
|
646 |
=item $grp->set_feeder ($callback->($grp)) |
647 |
|
648 |
[VERY EXPERIMENTAL] |
649 |
|
650 |
Sets a feeder/generator on this group: every group can have an attached |
651 |
generator that generates requests if idle. The idea behind this is that, |
652 |
although you could just queue as many requests as you want in a group, |
653 |
this might starve other requests for a potentially long time. For |
654 |
example, C<aio_scandir> might generate hundreds of thousands C<aio_stat> |
655 |
requests, delaying any later requests for a long time. |
656 |
|
657 |
To avoid this, and allow incremental generation of requests, you can |
658 |
instead a group and set a feeder on it that generates those requests. The |
659 |
feeder will be called whenever there are few enough (see C<feeder_limit>, |
660 |
below) requests active in the group itself and is expected to queue more |
661 |
requests. |
662 |
|
663 |
The feeder can queue as many requests as it likes (i.e. C<add> does not |
664 |
impose any limits). |
665 |
|
666 |
If the feeder does not queue more requests when called, it will be |
667 |
automatically removed from the group. |
668 |
|
669 |
If the feeder limit is C<0>, it will be set to C<2> automatically. |
670 |
|
671 |
Example: |
672 |
|
673 |
# stat all files in @files, but only ever use four aio requests concurrently: |
674 |
|
675 |
my $grp = aio_group sub { print "finished\n" }; |
676 |
$grp->feeder_limit (4); |
677 |
$grp->set_feeder (sub { |
678 |
my $file = pop @files |
679 |
or return; |
680 |
|
681 |
add $grp aio_stat $file, sub { ... }; |
682 |
}); |
683 |
|
684 |
=item $grp->feeder_limit ($num) |
685 |
|
686 |
Sets the feeder limit for the group: The feeder will be called whenever |
687 |
the group contains less than this many requests. |
688 |
|
689 |
Setting the limit to C<0> will pause the feeding process. |
690 |
|
691 |
=back |
692 |
|
693 |
=head2 SUPPORT FUNCTIONS |
694 |
|
695 |
=over 4 |
696 |
|
697 |
=item $fileno = IO::AIO::poll_fileno |
698 |
|
699 |
Return the I<request result pipe file descriptor>. This filehandle must be |
700 |
polled for reading by some mechanism outside this module (e.g. Event or |
701 |
select, see below or the SYNOPSIS). If the pipe becomes readable you have |
702 |
to call C<poll_cb> to check the results. |
703 |
|
704 |
See C<poll_cb> for an example. |
705 |
|
706 |
=item IO::AIO::poll_cb |
707 |
|
708 |
Process all outstanding events on the result pipe. You have to call this |
709 |
regularly. Returns the number of events processed. Returns immediately |
710 |
when no events are outstanding. |
711 |
|
712 |
Example: Install an Event watcher that automatically calls |
713 |
IO::AIO::poll_cb with high priority: |
714 |
|
715 |
Event->io (fd => IO::AIO::poll_fileno, |
716 |
poll => 'r', async => 1, |
717 |
cb => \&IO::AIO::poll_cb); |
718 |
|
719 |
=item IO::AIO::poll_wait |
720 |
|
721 |
Wait till the result filehandle becomes ready for reading (simply does a |
722 |
C<select> on the filehandle. This is useful if you want to synchronously wait |
723 |
for some requests to finish). |
724 |
|
725 |
See C<nreqs> for an example. |
726 |
|
727 |
=item IO::AIO::nreqs |
728 |
|
729 |
Returns the number of requests currently outstanding (i.e. for which their |
730 |
callback has not been invoked yet). |
731 |
|
732 |
Example: wait till there are no outstanding requests anymore: |
733 |
|
734 |
IO::AIO::poll_wait, IO::AIO::poll_cb |
735 |
while IO::AIO::nreqs; |
736 |
|
737 |
=item IO::AIO::flush |
738 |
|
739 |
Wait till all outstanding AIO requests have been handled. |
740 |
|
741 |
Strictly equivalent to: |
742 |
|
743 |
IO::AIO::poll_wait, IO::AIO::poll_cb |
744 |
while IO::AIO::nreqs; |
745 |
|
746 |
=item IO::AIO::poll |
747 |
|
748 |
Waits until some requests have been handled. |
749 |
|
750 |
Strictly equivalent to: |
751 |
|
752 |
IO::AIO::poll_wait, IO::AIO::poll_cb |
753 |
if IO::AIO::nreqs; |
754 |
|
755 |
=item IO::AIO::min_parallel $nthreads |
756 |
|
757 |
Set the minimum number of AIO threads to C<$nthreads>. The current |
758 |
default is C<8>, which means eight asynchronous operations can execute |
759 |
concurrently at any one time (the number of outstanding requests, |
760 |
however, is unlimited). |
761 |
|
762 |
IO::AIO starts threads only on demand, when an AIO request is queued and |
763 |
no free thread exists. |
764 |
|
765 |
It is recommended to keep the number of threads relatively low, as some |
766 |
Linux kernel versions will scale negatively with the number of threads |
767 |
(higher parallelity => MUCH higher latency). With current Linux 2.6 |
768 |
versions, 4-32 threads should be fine. |
769 |
|
770 |
Under most circumstances you don't need to call this function, as the |
771 |
module selects a default that is suitable for low to moderate load. |
772 |
|
773 |
=item IO::AIO::max_parallel $nthreads |
774 |
|
775 |
Sets the maximum number of AIO threads to C<$nthreads>. If more than the |
776 |
specified number of threads are currently running, this function kills |
777 |
them. This function blocks until the limit is reached. |
778 |
|
779 |
While C<$nthreads> are zero, aio requests get queued but not executed |
780 |
until the number of threads has been increased again. |
781 |
|
782 |
This module automatically runs C<max_parallel 0> at program end, to ensure |
783 |
that all threads are killed and that there are no outstanding requests. |
784 |
|
785 |
Under normal circumstances you don't need to call this function. |
786 |
|
787 |
=item $oldnreqs = IO::AIO::max_outstanding $nreqs |
788 |
|
789 |
Sets the maximum number of outstanding requests to C<$nreqs>. If you |
790 |
try to queue up more than this number of requests, the caller will block until |
791 |
some requests have been handled. |
792 |
|
793 |
The default is very large, so normally there is no practical limit. If you |
794 |
queue up many requests in a loop it often improves speed if you set |
795 |
this to a relatively low number, such as C<100>. |
796 |
|
797 |
Under normal circumstances you don't need to call this function. |
798 |
|
799 |
=back |
800 |
|
801 |
=cut |
802 |
|
803 |
# support function to convert a fd into a perl filehandle |
804 |
sub _fd2fh { |
805 |
return undef if $_[0] < 0; |
806 |
|
807 |
# try to generate nice filehandles |
808 |
my $sym = "IO::AIO::fd#$_[0]"; |
809 |
local *$sym; |
810 |
|
811 |
open *$sym, "+<&=$_[0]" # usually works under any unix |
812 |
or open *$sym, "<&=$_[0]" # cygwin needs this |
813 |
or open *$sym, ">&=$_[0]" # or this |
814 |
or return undef; |
815 |
|
816 |
*$sym |
817 |
} |
818 |
|
819 |
min_parallel 8; |
820 |
|
821 |
END { |
822 |
max_parallel 0; |
823 |
} |
824 |
|
825 |
1; |
826 |
|
827 |
=head2 FORK BEHAVIOUR |
828 |
|
829 |
This module should do "the right thing" when the process using it forks: |
830 |
|
831 |
Before the fork, IO::AIO enters a quiescent state where no requests |
832 |
can be added in other threads and no results will be processed. After |
833 |
the fork the parent simply leaves the quiescent state and continues |
834 |
request/result processing, while the child clears the request/result |
835 |
queue (so the requests started before the fork will only be handled in |
836 |
the parent). Threads will be started on demand until the limit ste in the |
837 |
parent process has been reached again. |
838 |
|
839 |
In short: the parent will, after a short pause, continue as if fork had |
840 |
not been called, while the child will act as if IO::AIO has not been used |
841 |
yet. |
842 |
|
843 |
=head2 MEMORY USAGE |
844 |
|
845 |
Each aio request uses - depending on your architecture - around 128 bytes |
846 |
of memory. In addition, stat requests need a stat buffer (possibly a few |
847 |
hundred bytes). Perl scalars and other data passed into aio requests will |
848 |
also be locked. |
849 |
|
850 |
This is now awfully much, so queuing lots of requests is not usually a |
851 |
problem. |
852 |
|
853 |
Each thread needs a stack area which is usually around 16k, sometimes much |
854 |
larger, depending on the OS. |
855 |
|
856 |
=head1 SEE ALSO |
857 |
|
858 |
L<Coro>, L<Linux::AIO> (obsolete). |
859 |
|
860 |
=head1 AUTHOR |
861 |
|
862 |
Marc Lehmann <schmorp@schmorp.de> |
863 |
http://home.schmorp.de/ |
864 |
|
865 |
=cut |
866 |
|