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 |
aioreq_pri 4; # give next request a very high priority |
24 |
my $req = aio_unlink "/tmp/file", sub { }; |
25 |
$req->cancel; # cancel request if still in queue |
26 |
|
27 |
my $grp = aio_group sub { print "all stats done\n" }; |
28 |
add $grp aio_stat "..." for ...; |
29 |
|
30 |
# AnyEvent integration |
31 |
open my $fh, "<&=" . IO::AIO::poll_fileno or die "$!"; |
32 |
my $w = AnyEvent->io (fh => $fh, poll => 'r', cb => sub { IO::AIO::poll_cb }); |
33 |
|
34 |
# Event integration |
35 |
Event->io (fd => IO::AIO::poll_fileno, |
36 |
poll => 'r', |
37 |
cb => \&IO::AIO::poll_cb); |
38 |
|
39 |
# Glib/Gtk2 integration |
40 |
add_watch Glib::IO IO::AIO::poll_fileno, |
41 |
in => sub { IO::AIO::poll_cb; 1 }; |
42 |
|
43 |
# Tk integration |
44 |
Tk::Event::IO->fileevent (IO::AIO::poll_fileno, "", |
45 |
readable => \&IO::AIO::poll_cb); |
46 |
|
47 |
# Danga::Socket integration |
48 |
Danga::Socket->AddOtherFds (IO::AIO::poll_fileno => |
49 |
\&IO::AIO::poll_cb); |
50 |
|
51 |
=head1 DESCRIPTION |
52 |
|
53 |
This module implements asynchronous I/O using whatever means your |
54 |
operating system supports. |
55 |
|
56 |
In this version, a number of threads are started that execute your |
57 |
requests and signal their completion. You don't need thread support |
58 |
in perl, and the threads created by this module will not be visible |
59 |
to perl. In the future, this module might make use of the native aio |
60 |
functions available on many operating systems. However, they are often |
61 |
not well-supported or restricted (Linux doesn't allow them on normal |
62 |
files currently, for example), and they would only support aio_read and |
63 |
aio_write, so the remaining functionality would have to be implemented |
64 |
using threads anyway. |
65 |
|
66 |
Although the module will work with in the presence of other (Perl-) |
67 |
threads, it is currently not reentrant in any way, so use appropriate |
68 |
locking yourself, always call C<poll_cb> from within the same thread, or |
69 |
never call C<poll_cb> (or other C<aio_> functions) recursively. |
70 |
|
71 |
=head1 REQUEST ANATOMY AND LIFETIME |
72 |
|
73 |
Every C<aio_*> function creates a request. which is a C data structure not |
74 |
directly visible to Perl. |
75 |
|
76 |
If called in non-void context, every request function returns a Perl |
77 |
object representing the request. In void context, nothing is returned, |
78 |
which saves a bit of memory. |
79 |
|
80 |
The perl object is a fairly standard ref-to-hash object. The hash contents |
81 |
are not used by IO::AIO so you are free to store anything you like in it. |
82 |
|
83 |
During their existance, aio requests travel through the following states, |
84 |
in order: |
85 |
|
86 |
=over 4 |
87 |
|
88 |
=item ready |
89 |
|
90 |
Immediately after a request is created it is put into the ready state, |
91 |
waiting for a thread to execute it. |
92 |
|
93 |
=item execute |
94 |
|
95 |
A thread has accepted the request for processing and is currently |
96 |
executing it (e.g. blocking in read). |
97 |
|
98 |
=item pending |
99 |
|
100 |
The request has been executed and is waiting for result processing. |
101 |
|
102 |
While request submission and execution is fully asynchronous, result |
103 |
processing is not and relies on the perl interpreter calling C<poll_cb> |
104 |
(or another function with the same effect). |
105 |
|
106 |
=item result |
107 |
|
108 |
The request results are processed synchronously by C<poll_cb>. |
109 |
|
110 |
The C<poll_cb> function will process all outstanding aio requests by |
111 |
calling their callbacks, freeing memory associated with them and managing |
112 |
any groups they are contained in. |
113 |
|
114 |
=item done |
115 |
|
116 |
Request has reached the end of its lifetime and holds no resources anymore |
117 |
(except possibly for the Perl object, but its connection to the actual |
118 |
aio request is severed and calling its methods will either do nothing or |
119 |
result in a runtime error). |
120 |
|
121 |
=cut |
122 |
|
123 |
package IO::AIO; |
124 |
|
125 |
no warnings; |
126 |
use strict 'vars'; |
127 |
|
128 |
use base 'Exporter'; |
129 |
|
130 |
BEGIN { |
131 |
our $VERSION = '2.0'; |
132 |
|
133 |
our @AIO_REQ = qw(aio_sendfile aio_read aio_write aio_open aio_close aio_stat |
134 |
aio_lstat aio_unlink aio_rmdir aio_readdir aio_scandir aio_symlink |
135 |
aio_fsync aio_fdatasync aio_readahead aio_rename aio_link aio_move |
136 |
aio_group aio_nop); |
137 |
our @EXPORT = (@AIO_REQ, qw(aioreq_pri aioreq_nice)); |
138 |
our @EXPORT_OK = qw(poll_fileno poll_cb poll_wait flush |
139 |
min_parallel max_parallel max_outstanding nreqs); |
140 |
|
141 |
@IO::AIO::GRP::ISA = 'IO::AIO::REQ'; |
142 |
|
143 |
require XSLoader; |
144 |
XSLoader::load ("IO::AIO", $VERSION); |
145 |
} |
146 |
|
147 |
=head1 FUNCTIONS |
148 |
|
149 |
=head2 AIO FUNCTIONS |
150 |
|
151 |
All the C<aio_*> calls are more or less thin wrappers around the syscall |
152 |
with the same name (sans C<aio_>). The arguments are similar or identical, |
153 |
and they all accept an additional (and optional) C<$callback> argument |
154 |
which must be a code reference. This code reference will get called with |
155 |
the syscall return code (e.g. most syscalls return C<-1> on error, unlike |
156 |
perl, which usually delivers "false") as it's sole argument when the given |
157 |
syscall has been executed asynchronously. |
158 |
|
159 |
All functions expecting a filehandle keep a copy of the filehandle |
160 |
internally until the request has finished. |
161 |
|
162 |
All requests return objects of type L<IO::AIO::REQ> that allow further |
163 |
manipulation of those requests while they are in-flight. |
164 |
|
165 |
The pathnames you pass to these routines I<must> be absolute and |
166 |
encoded in byte form. The reason for the former is that at the time the |
167 |
request is being executed, the current working directory could have |
168 |
changed. Alternatively, you can make sure that you never change the |
169 |
current working directory. |
170 |
|
171 |
To encode pathnames to byte form, either make sure you either: a) |
172 |
always pass in filenames you got from outside (command line, readdir |
173 |
etc.), b) are ASCII or ISO 8859-1, c) use the Encode module and encode |
174 |
your pathnames to the locale (or other) encoding in effect in the user |
175 |
environment, d) use Glib::filename_from_unicode on unicode filenames or e) |
176 |
use something else. |
177 |
|
178 |
=over 4 |
179 |
|
180 |
=item aioreq_pri $pri |
181 |
|
182 |
Sets the priority for the next aio request. The default priority |
183 |
is C<0>, the minimum and maximum priorities are C<-4> and C<4>, |
184 |
respectively. Requests with higher priority will be serviced first. |
185 |
|
186 |
The priority will be reset to C<0> after each call to one of the C<aio_> |
187 |
functions. |
188 |
|
189 |
Example: open a file with low priority, then read something from it with |
190 |
higher priority so the read request is serviced before other low priority |
191 |
open requests (potentially spamming the cache): |
192 |
|
193 |
aioreq_pri -3; |
194 |
aio_open ..., sub { |
195 |
return unless $_[0]; |
196 |
|
197 |
aioreq_pri -2; |
198 |
aio_read $_[0], ..., sub { |
199 |
... |
200 |
}; |
201 |
}; |
202 |
|
203 |
=item aioreq_nice $pri_adjust |
204 |
|
205 |
Similar to C<aioreq_pri>, but subtracts the given value from the current |
206 |
priority, so effects are cumulative. |
207 |
|
208 |
=item aio_open $pathname, $flags, $mode, $callback->($fh) |
209 |
|
210 |
Asynchronously open or create a file and call the callback with a newly |
211 |
created filehandle for the file. |
212 |
|
213 |
The pathname passed to C<aio_open> must be absolute. See API NOTES, above, |
214 |
for an explanation. |
215 |
|
216 |
The C<$flags> argument is a bitmask. See the C<Fcntl> module for a |
217 |
list. They are the same as used by C<sysopen>. |
218 |
|
219 |
Likewise, C<$mode> specifies the mode of the newly created file, if it |
220 |
didn't exist and C<O_CREAT> has been given, just like perl's C<sysopen>, |
221 |
except that it is mandatory (i.e. use C<0> if you don't create new files, |
222 |
and C<0666> or C<0777> if you do). |
223 |
|
224 |
Example: |
225 |
|
226 |
aio_open "/etc/passwd", O_RDONLY, 0, sub { |
227 |
if ($_[0]) { |
228 |
print "open successful, fh is $_[0]\n"; |
229 |
... |
230 |
} else { |
231 |
die "open failed: $!\n"; |
232 |
} |
233 |
}; |
234 |
|
235 |
=item aio_close $fh, $callback->($status) |
236 |
|
237 |
Asynchronously close a file and call the callback with the result |
238 |
code. I<WARNING:> although accepted, you should not pass in a perl |
239 |
filehandle here, as perl will likely close the file descriptor another |
240 |
time when the filehandle is destroyed. Normally, you can safely call perls |
241 |
C<close> or just let filehandles go out of scope. |
242 |
|
243 |
This is supposed to be a bug in the API, so that might change. It's |
244 |
therefore best to avoid this function. |
245 |
|
246 |
=item aio_read $fh,$offset,$length, $data,$dataoffset, $callback->($retval) |
247 |
|
248 |
=item aio_write $fh,$offset,$length, $data,$dataoffset, $callback->($retval) |
249 |
|
250 |
Reads or writes C<length> bytes from the specified C<fh> and C<offset> |
251 |
into the scalar given by C<data> and offset C<dataoffset> and calls the |
252 |
callback without the actual number of bytes read (or -1 on error, just |
253 |
like the syscall). |
254 |
|
255 |
The C<$data> scalar I<MUST NOT> be modified in any way while the request |
256 |
is outstanding. Modifying it can result in segfaults or WW3 (if the |
257 |
necessary/optional hardware is installed). |
258 |
|
259 |
Example: Read 15 bytes at offset 7 into scalar C<$buffer>, starting at |
260 |
offset C<0> within the scalar: |
261 |
|
262 |
aio_read $fh, 7, 15, $buffer, 0, sub { |
263 |
$_[0] > 0 or die "read error: $!"; |
264 |
print "read $_[0] bytes: <$buffer>\n"; |
265 |
}; |
266 |
|
267 |
=item aio_move $srcpath, $dstpath, $callback->($status) |
268 |
|
269 |
Try to move the I<file> (directories not supported as either source or |
270 |
destination) from C<$srcpath> to C<$dstpath> and call the callback with |
271 |
the C<0> (error) or C<-1> ok. |
272 |
|
273 |
This is a composite request that tries to rename(2) the file first. If |
274 |
rename files with C<EXDEV>, it creates the destination file with mode 0200 |
275 |
and copies the contents of the source file into it using C<aio_sendfile>, |
276 |
followed by restoring atime, mtime, access mode and uid/gid, in that |
277 |
order, and unlinking the C<$srcpath>. |
278 |
|
279 |
If an error occurs, the partial destination file will be unlinked, if |
280 |
possible, except when setting atime, mtime, access mode and uid/gid, where |
281 |
errors are being ignored. |
282 |
|
283 |
=cut |
284 |
|
285 |
sub aio_move($$$) { |
286 |
my ($src, $dst, $cb) = @_; |
287 |
|
288 |
my $grp = aio_group $cb; |
289 |
|
290 |
add $grp aio_rename $src, $dst, sub { |
291 |
if ($_[0] && $! == EXDEV) { |
292 |
add $grp aio_open $src, O_RDONLY, 0, sub { |
293 |
if (my $src_fh = $_[0]) { |
294 |
my @stat = stat $src_fh; |
295 |
|
296 |
add $grp aio_open $dst, O_WRONLY, 0200, sub { |
297 |
if (my $dst_fh = $_[0]) { |
298 |
add $grp aio_sendfile $dst_fh, $src_fh, 0, $stat[7], sub { |
299 |
close $src_fh; |
300 |
|
301 |
if ($_[0] == $stat[7]) { |
302 |
utime $stat[8], $stat[9], $dst; |
303 |
chmod $stat[2] & 07777, $dst_fh; |
304 |
chown $stat[4], $stat[5], $dst_fh; |
305 |
close $dst_fh; |
306 |
|
307 |
add $grp aio_unlink $src, sub { |
308 |
$grp->result ($_[0]); |
309 |
}; |
310 |
} else { |
311 |
my $errno = $!; |
312 |
add $grp aio_unlink $dst, sub { |
313 |
$! = $errno; |
314 |
$grp->result (-1); |
315 |
}; |
316 |
} |
317 |
}; |
318 |
} else { |
319 |
$grp->result (-1); |
320 |
} |
321 |
}, |
322 |
|
323 |
} else { |
324 |
$grp->result (-1); |
325 |
} |
326 |
}; |
327 |
} else { |
328 |
$grp->result ($_[0]); |
329 |
} |
330 |
}; |
331 |
|
332 |
$grp |
333 |
} |
334 |
|
335 |
=item aio_sendfile $out_fh, $in_fh, $in_offset, $length, $callback->($retval) |
336 |
|
337 |
Tries to copy C<$length> bytes from C<$in_fh> to C<$out_fh>. It starts |
338 |
reading at byte offset C<$in_offset>, and starts writing at the current |
339 |
file offset of C<$out_fh>. Because of that, it is not safe to issue more |
340 |
than one C<aio_sendfile> per C<$out_fh>, as they will interfere with each |
341 |
other. |
342 |
|
343 |
This call tries to make use of a native C<sendfile> syscall to provide |
344 |
zero-copy operation. For this to work, C<$out_fh> should refer to a |
345 |
socket, and C<$in_fh> should refer to mmap'able file. |
346 |
|
347 |
If the native sendfile call fails or is not implemented, it will be |
348 |
emulated, so you can call C<aio_sendfile> on any type of filehandle |
349 |
regardless of the limitations of the operating system. |
350 |
|
351 |
Please note, however, that C<aio_sendfile> can read more bytes from |
352 |
C<$in_fh> than are written, and there is no way to find out how many |
353 |
bytes have been read from C<aio_sendfile> alone, as C<aio_sendfile> only |
354 |
provides the number of bytes written to C<$out_fh>. Only if the result |
355 |
value equals C<$length> one can assume that C<$length> bytes have been |
356 |
read. |
357 |
|
358 |
=item aio_readahead $fh,$offset,$length, $callback->($retval) |
359 |
|
360 |
C<aio_readahead> populates the page cache with data from a file so that |
361 |
subsequent reads from that file will not block on disk I/O. The C<$offset> |
362 |
argument specifies the starting point from which data is to be read and |
363 |
C<$length> specifies the number of bytes to be read. I/O is performed in |
364 |
whole pages, so that offset is effectively rounded down to a page boundary |
365 |
and bytes are read up to the next page boundary greater than or equal to |
366 |
(off-set+length). C<aio_readahead> does not read beyond the end of the |
367 |
file. The current file offset of the file is left unchanged. |
368 |
|
369 |
If that syscall doesn't exist (likely if your OS isn't Linux) it will be |
370 |
emulated by simply reading the data, which would have a similar effect. |
371 |
|
372 |
=item aio_stat $fh_or_path, $callback->($status) |
373 |
|
374 |
=item aio_lstat $fh, $callback->($status) |
375 |
|
376 |
Works like perl's C<stat> or C<lstat> in void context. The callback will |
377 |
be called after the stat and the results will be available using C<stat _> |
378 |
or C<-s _> etc... |
379 |
|
380 |
The pathname passed to C<aio_stat> must be absolute. See API NOTES, above, |
381 |
for an explanation. |
382 |
|
383 |
Currently, the stats are always 64-bit-stats, i.e. instead of returning an |
384 |
error when stat'ing a large file, the results will be silently truncated |
385 |
unless perl itself is compiled with large file support. |
386 |
|
387 |
Example: Print the length of F</etc/passwd>: |
388 |
|
389 |
aio_stat "/etc/passwd", sub { |
390 |
$_[0] and die "stat failed: $!"; |
391 |
print "size is ", -s _, "\n"; |
392 |
}; |
393 |
|
394 |
=item aio_unlink $pathname, $callback->($status) |
395 |
|
396 |
Asynchronously unlink (delete) a file and call the callback with the |
397 |
result code. |
398 |
|
399 |
=item aio_link $srcpath, $dstpath, $callback->($status) |
400 |
|
401 |
Asynchronously create a new link to the existing object at C<$srcpath> at |
402 |
the path C<$dstpath> and call the callback with the result code. |
403 |
|
404 |
=item aio_symlink $srcpath, $dstpath, $callback->($status) |
405 |
|
406 |
Asynchronously create a new symbolic link to the existing object at C<$srcpath> at |
407 |
the path C<$dstpath> and call the callback with the result code. |
408 |
|
409 |
=item aio_rename $srcpath, $dstpath, $callback->($status) |
410 |
|
411 |
Asynchronously rename the object at C<$srcpath> to C<$dstpath>, just as |
412 |
rename(2) and call the callback with the result code. |
413 |
|
414 |
=item aio_rmdir $pathname, $callback->($status) |
415 |
|
416 |
Asynchronously rmdir (delete) a directory and call the callback with the |
417 |
result code. |
418 |
|
419 |
=item aio_readdir $pathname, $callback->($entries) |
420 |
|
421 |
Unlike the POSIX call of the same name, C<aio_readdir> reads an entire |
422 |
directory (i.e. opendir + readdir + closedir). The entries will not be |
423 |
sorted, and will B<NOT> include the C<.> and C<..> entries. |
424 |
|
425 |
The callback a single argument which is either C<undef> or an array-ref |
426 |
with the filenames. |
427 |
|
428 |
=item aio_scandir $path, $maxreq, $callback->($dirs, $nondirs) |
429 |
|
430 |
Scans a directory (similar to C<aio_readdir>) but additionally tries to |
431 |
separate the entries of directory C<$path> into two sets of names, ones |
432 |
you can recurse into (directories or links to them), and ones you cannot |
433 |
recurse into (everything else). |
434 |
|
435 |
C<aio_scandir> is a composite request that creates of many sub requests_ |
436 |
C<$maxreq> specifies the maximum number of outstanding aio requests that |
437 |
this function generates. If it is C<< <= 0 >>, then a suitable default |
438 |
will be chosen (currently 6). |
439 |
|
440 |
On error, the callback is called without arguments, otherwise it receives |
441 |
two array-refs with path-relative entry names. |
442 |
|
443 |
Example: |
444 |
|
445 |
aio_scandir $dir, 0, sub { |
446 |
my ($dirs, $nondirs) = @_; |
447 |
print "real directories: @$dirs\n"; |
448 |
print "everything else: @$nondirs\n"; |
449 |
}; |
450 |
|
451 |
Implementation notes. |
452 |
|
453 |
The C<aio_readdir> cannot be avoided, but C<stat()>'ing every entry can. |
454 |
|
455 |
After reading the directory, the modification time, size etc. of the |
456 |
directory before and after the readdir is checked, and if they match (and |
457 |
isn't the current time), the link count will be used to decide how many |
458 |
entries are directories (if >= 2). Otherwise, no knowledge of the number |
459 |
of subdirectories will be assumed. |
460 |
|
461 |
Then entries will be sorted into likely directories (everything without |
462 |
a non-initial dot currently) and likely non-directories (everything |
463 |
else). Then every entry plus an appended C</.> will be C<stat>'ed, |
464 |
likely directories first. If that succeeds, it assumes that the entry |
465 |
is a directory or a symlink to directory (which will be checked |
466 |
seperately). This is often faster than stat'ing the entry itself because |
467 |
filesystems might detect the type of the entry without reading the inode |
468 |
data (e.g. ext2fs filetype feature). |
469 |
|
470 |
If the known number of directories (link count - 2) has been reached, the |
471 |
rest of the entries is assumed to be non-directories. |
472 |
|
473 |
This only works with certainty on POSIX (= UNIX) filesystems, which |
474 |
fortunately are the vast majority of filesystems around. |
475 |
|
476 |
It will also likely work on non-POSIX filesystems with reduced efficiency |
477 |
as those tend to return 0 or 1 as link counts, which disables the |
478 |
directory counting heuristic. |
479 |
|
480 |
=cut |
481 |
|
482 |
sub aio_scandir($$$) { |
483 |
my ($path, $maxreq, $cb) = @_; |
484 |
|
485 |
my $grp = aio_group $cb; |
486 |
|
487 |
$maxreq = 6 if $maxreq <= 0; |
488 |
|
489 |
# stat once |
490 |
add $grp aio_stat $path, sub { |
491 |
return $grp->result () if $_[0]; |
492 |
my $now = time; |
493 |
my $hash1 = join ":", (stat _)[0,1,3,7,9]; |
494 |
|
495 |
# read the directory entries |
496 |
add $grp aio_readdir $path, sub { |
497 |
my $entries = shift |
498 |
or return $grp->result (); |
499 |
|
500 |
# stat the dir another time |
501 |
add $grp aio_stat $path, sub { |
502 |
my $hash2 = join ":", (stat _)[0,1,3,7,9]; |
503 |
|
504 |
my $ndirs; |
505 |
|
506 |
# take the slow route if anything looks fishy |
507 |
if ($hash1 ne $hash2 or (stat _)[9] == $now) { |
508 |
$ndirs = -1; |
509 |
} else { |
510 |
# if nlink == 2, we are finished |
511 |
# on non-posix-fs's, we rely on nlink < 2 |
512 |
$ndirs = (stat _)[3] - 2 |
513 |
or return $grp->result ([], $entries); |
514 |
} |
515 |
|
516 |
# sort into likely dirs and likely nondirs |
517 |
# dirs == files without ".", short entries first |
518 |
$entries = [map $_->[0], |
519 |
sort { $b->[1] cmp $a->[1] } |
520 |
map [$_, sprintf "%s%04d", (/.\./ ? "1" : "0"), length], |
521 |
@$entries]; |
522 |
|
523 |
my (@dirs, @nondirs); |
524 |
|
525 |
my ($statcb, $schedcb); |
526 |
my $nreq = 0; |
527 |
|
528 |
my $statgrp = add $grp aio_group; |
529 |
|
530 |
$schedcb = sub { |
531 |
if (@$entries) { |
532 |
if ($nreq < $maxreq) { |
533 |
my $ent = pop @$entries; |
534 |
$nreq++; |
535 |
add $statgrp aio_stat "$path/$ent/.", sub { $statcb->($_[0], $ent) }; |
536 |
} |
537 |
} elsif (!$nreq) { |
538 |
# finished |
539 |
$statgrp->cancel; |
540 |
undef $statcb; |
541 |
undef $schedcb; |
542 |
$grp->result (\@dirs, \@nondirs); |
543 |
} |
544 |
}; |
545 |
$statcb = sub { |
546 |
my ($status, $entry) = @_; |
547 |
|
548 |
if ($status < 0) { |
549 |
$nreq--; |
550 |
push @nondirs, $entry; |
551 |
&$schedcb; |
552 |
} else { |
553 |
# need to check for real directory |
554 |
add $grp aio_lstat "$path/$entry", sub { |
555 |
$nreq--; |
556 |
|
557 |
if (-d _) { |
558 |
push @dirs, $entry; |
559 |
|
560 |
if (!--$ndirs) { |
561 |
push @nondirs, @$entries; |
562 |
$entries = []; |
563 |
} |
564 |
} else { |
565 |
push @nondirs, $entry; |
566 |
} |
567 |
|
568 |
&$schedcb; |
569 |
} |
570 |
} |
571 |
}; |
572 |
|
573 |
&$schedcb while @$entries && $nreq < $maxreq; |
574 |
}; |
575 |
}; |
576 |
}; |
577 |
|
578 |
$grp |
579 |
} |
580 |
|
581 |
=item aio_fsync $fh, $callback->($status) |
582 |
|
583 |
Asynchronously call fsync on the given filehandle and call the callback |
584 |
with the fsync result code. |
585 |
|
586 |
=item aio_fdatasync $fh, $callback->($status) |
587 |
|
588 |
Asynchronously call fdatasync on the given filehandle and call the |
589 |
callback with the fdatasync result code. |
590 |
|
591 |
If this call isn't available because your OS lacks it or it couldn't be |
592 |
detected, it will be emulated by calling C<fsync> instead. |
593 |
|
594 |
=item aio_group $callback->(...) |
595 |
|
596 |
[EXPERIMENTAL] |
597 |
|
598 |
This is a very special aio request: Instead of doing something, it is a |
599 |
container for other aio requests, which is useful if you want to bundle |
600 |
many requests into a single, composite, request with a definite callback |
601 |
and the ability to cancel the whole request with its subrequests. |
602 |
|
603 |
Returns an object of class L<IO::AIO::GRP>. See its documentation below |
604 |
for more info. |
605 |
|
606 |
Example: |
607 |
|
608 |
my $grp = aio_group sub { |
609 |
print "all stats done\n"; |
610 |
}; |
611 |
|
612 |
add $grp |
613 |
(aio_stat ...), |
614 |
(aio_stat ...), |
615 |
...; |
616 |
|
617 |
=item aio_nop $callback->() |
618 |
|
619 |
This is a special request - it does nothing in itself and is only used for |
620 |
side effects, such as when you want to add a dummy request to a group so |
621 |
that finishing the requests in the group depends on executing the given |
622 |
code. |
623 |
|
624 |
While this request does nothing, it still goes through the execution |
625 |
phase and still requires a worker thread. Thus, the callback will not |
626 |
be executed immediately but only after other requests in the queue have |
627 |
entered their execution phase. This can be used to measure request |
628 |
latency. |
629 |
|
630 |
=item IO::AIO::aio_busy $fractional_seconds, $callback->() *NOT EXPORTED* |
631 |
|
632 |
Mainly used for debugging and benchmarking, this aio request puts one of |
633 |
the request workers to sleep for the given time. |
634 |
|
635 |
While it is theoretically handy to have simple I/O scheduling requests |
636 |
like sleep and file handle readable/writable, the overhead this creates is |
637 |
immense (it blocks a thread for a long time) so do not use this function |
638 |
except to put your application under artificial I/O pressure. |
639 |
|
640 |
=back |
641 |
|
642 |
=head2 IO::AIO::REQ CLASS |
643 |
|
644 |
All non-aggregate C<aio_*> functions return an object of this class when |
645 |
called in non-void context. |
646 |
|
647 |
=over 4 |
648 |
|
649 |
=item cancel $req |
650 |
|
651 |
Cancels the request, if possible. Has the effect of skipping execution |
652 |
when entering the B<execute> state and skipping calling the callback when |
653 |
entering the the B<result> state, but will leave the request otherwise |
654 |
untouched. That means that requests that currently execute will not be |
655 |
stopped and resources held by the request will not be freed prematurely. |
656 |
|
657 |
=item cb $req $callback->(...) |
658 |
|
659 |
Replace (or simply set) the callback registered to the request. |
660 |
|
661 |
=back |
662 |
|
663 |
=head2 IO::AIO::GRP CLASS |
664 |
|
665 |
This class is a subclass of L<IO::AIO::REQ>, so all its methods apply to |
666 |
objects of this class, too. |
667 |
|
668 |
A IO::AIO::GRP object is a special request that can contain multiple other |
669 |
aio requests. |
670 |
|
671 |
You create one by calling the C<aio_group> constructing function with a |
672 |
callback that will be called when all contained requests have entered the |
673 |
C<done> state: |
674 |
|
675 |
my $grp = aio_group sub { |
676 |
print "all requests are done\n"; |
677 |
}; |
678 |
|
679 |
You add requests by calling the C<add> method with one or more |
680 |
C<IO::AIO::REQ> objects: |
681 |
|
682 |
$grp->add (aio_unlink "..."); |
683 |
|
684 |
add $grp aio_stat "...", sub { |
685 |
$_[0] or return $grp->result ("error"); |
686 |
|
687 |
# add another request dynamically, if first succeeded |
688 |
add $grp aio_open "...", sub { |
689 |
$grp->result ("ok"); |
690 |
}; |
691 |
}; |
692 |
|
693 |
This makes it very easy to create composite requests (see the source of |
694 |
C<aio_move> for an application) that work and feel like simple requests. |
695 |
|
696 |
=over 4 |
697 |
|
698 |
=item * The IO::AIO::GRP objects will be cleaned up during calls to |
699 |
C<IO::AIO::poll_cb>, just like any other request. |
700 |
|
701 |
=item * They can be canceled like any other request. Canceling will cancel not |
702 |
only the request itself, but also all requests it contains. |
703 |
|
704 |
=item * They can also can also be added to other IO::AIO::GRP objects. |
705 |
|
706 |
=item * You must not add requests to a group from within the group callback (or |
707 |
any later time). |
708 |
|
709 |
=item * This does not harmonise well with C<max_outstanding>, so best do |
710 |
not combine C<aio_group> with it. Groups and feeders are recommended for |
711 |
this kind of concurrency-limiting. |
712 |
|
713 |
=back |
714 |
|
715 |
Their lifetime, simplified, looks like this: when they are empty, they |
716 |
will finish very quickly. If they contain only requests that are in the |
717 |
C<done> state, they will also finish. Otherwise they will continue to |
718 |
exist. |
719 |
|
720 |
That means after creating a group you have some time to add requests. And |
721 |
in the callbacks of those requests, you can add further requests to the |
722 |
group. And only when all those requests have finished will the the group |
723 |
itself finish. |
724 |
|
725 |
=over 4 |
726 |
|
727 |
=item add $grp ... |
728 |
|
729 |
=item $grp->add (...) |
730 |
|
731 |
Add one or more requests to the group. Any type of L<IO::AIO::REQ> can |
732 |
be added, including other groups, as long as you do not create circular |
733 |
dependencies. |
734 |
|
735 |
Returns all its arguments. |
736 |
|
737 |
=item $grp->result (...) |
738 |
|
739 |
Set the result value(s) that will be passed to the group callback when all |
740 |
subrequests have finished. By default, no argument will be passed. |
741 |
|
742 |
=item feed $grp $callback->($grp) |
743 |
|
744 |
[VERY EXPERIMENTAL] |
745 |
|
746 |
Sets a feeder/generator on this group: every group can have an attached |
747 |
generator that generates requests if idle. The idea behind this is that, |
748 |
although you could just queue as many requests as you want in a group, |
749 |
this might starve other requests for a potentially long time. For |
750 |
example, C<aio_scandir> might generate hundreds of thousands C<aio_stat> |
751 |
requests, delaying any later requests for a long time. |
752 |
|
753 |
To avoid this, and allow incremental generation of requests, you can |
754 |
instead a group and set a feeder on it that generates those requests. The |
755 |
feed callback will be called whenever there are few enough (see C<limit>, |
756 |
below) requests active in the group itself and is expected to queue more |
757 |
requests. |
758 |
|
759 |
The feed callback can queue as many requests as it likes (i.e. C<add> does |
760 |
not impose any limits). |
761 |
|
762 |
If the feed does not queue more requests when called, it will be |
763 |
automatically removed from the group. |
764 |
|
765 |
If the feed limit is C<0>, it will be set to C<2> automatically. |
766 |
|
767 |
Example: |
768 |
|
769 |
# stat all files in @files, but only ever use four aio requests concurrently: |
770 |
|
771 |
my $grp = aio_group sub { print "finished\n" }; |
772 |
limit $grp 4; |
773 |
feed $grp sub { |
774 |
my $file = pop @files |
775 |
or return; |
776 |
|
777 |
add $grp aio_stat $file, sub { ... }; |
778 |
}; |
779 |
|
780 |
=item limit $grp $num |
781 |
|
782 |
Sets the feeder limit for the group: The feeder will be called whenever |
783 |
the group contains less than this many requests. |
784 |
|
785 |
Setting the limit to C<0> will pause the feeding process. |
786 |
|
787 |
=back |
788 |
|
789 |
=head2 SUPPORT FUNCTIONS |
790 |
|
791 |
=over 4 |
792 |
|
793 |
=item $fileno = IO::AIO::poll_fileno |
794 |
|
795 |
Return the I<request result pipe file descriptor>. This filehandle must be |
796 |
polled for reading by some mechanism outside this module (e.g. Event or |
797 |
select, see below or the SYNOPSIS). If the pipe becomes readable you have |
798 |
to call C<poll_cb> to check the results. |
799 |
|
800 |
See C<poll_cb> for an example. |
801 |
|
802 |
=item IO::AIO::poll_cb |
803 |
|
804 |
Process all outstanding events on the result pipe. You have to call this |
805 |
regularly. Returns the number of events processed. Returns immediately |
806 |
when no events are outstanding. |
807 |
|
808 |
Example: Install an Event watcher that automatically calls |
809 |
IO::AIO::poll_cb with high priority: |
810 |
|
811 |
Event->io (fd => IO::AIO::poll_fileno, |
812 |
poll => 'r', async => 1, |
813 |
cb => \&IO::AIO::poll_cb); |
814 |
|
815 |
=item IO::AIO::poll_wait |
816 |
|
817 |
Wait till the result filehandle becomes ready for reading (simply does a |
818 |
C<select> on the filehandle. This is useful if you want to synchronously wait |
819 |
for some requests to finish). |
820 |
|
821 |
See C<nreqs> for an example. |
822 |
|
823 |
=item IO::AIO::nreqs |
824 |
|
825 |
Returns the number of requests currently outstanding (i.e. for which their |
826 |
callback has not been invoked yet). |
827 |
|
828 |
Example: wait till there are no outstanding requests anymore: |
829 |
|
830 |
IO::AIO::poll_wait, IO::AIO::poll_cb |
831 |
while IO::AIO::nreqs; |
832 |
|
833 |
=item IO::AIO::flush |
834 |
|
835 |
Wait till all outstanding AIO requests have been handled. |
836 |
|
837 |
Strictly equivalent to: |
838 |
|
839 |
IO::AIO::poll_wait, IO::AIO::poll_cb |
840 |
while IO::AIO::nreqs; |
841 |
|
842 |
=item IO::AIO::poll |
843 |
|
844 |
Waits until some requests have been handled. |
845 |
|
846 |
Strictly equivalent to: |
847 |
|
848 |
IO::AIO::poll_wait, IO::AIO::poll_cb |
849 |
if IO::AIO::nreqs; |
850 |
|
851 |
=item IO::AIO::min_parallel $nthreads |
852 |
|
853 |
Set the minimum number of AIO threads to C<$nthreads>. The current |
854 |
default is C<8>, which means eight asynchronous operations can execute |
855 |
concurrently at any one time (the number of outstanding requests, |
856 |
however, is unlimited). |
857 |
|
858 |
IO::AIO starts threads only on demand, when an AIO request is queued and |
859 |
no free thread exists. |
860 |
|
861 |
It is recommended to keep the number of threads relatively low, as some |
862 |
Linux kernel versions will scale negatively with the number of threads |
863 |
(higher parallelity => MUCH higher latency). With current Linux 2.6 |
864 |
versions, 4-32 threads should be fine. |
865 |
|
866 |
Under most circumstances you don't need to call this function, as the |
867 |
module selects a default that is suitable for low to moderate load. |
868 |
|
869 |
=item IO::AIO::max_parallel $nthreads |
870 |
|
871 |
Sets the maximum number of AIO threads to C<$nthreads>. If more than the |
872 |
specified number of threads are currently running, this function kills |
873 |
them. This function blocks until the limit is reached. |
874 |
|
875 |
While C<$nthreads> are zero, aio requests get queued but not executed |
876 |
until the number of threads has been increased again. |
877 |
|
878 |
This module automatically runs C<max_parallel 0> at program end, to ensure |
879 |
that all threads are killed and that there are no outstanding requests. |
880 |
|
881 |
Under normal circumstances you don't need to call this function. |
882 |
|
883 |
=item $oldnreqs = IO::AIO::max_outstanding $nreqs |
884 |
|
885 |
[DEPRECATED] |
886 |
|
887 |
Sets the maximum number of outstanding requests to C<$nreqs>. If you |
888 |
try to queue up more than this number of requests, the caller will block until |
889 |
some requests have been handled. |
890 |
|
891 |
The default is very large, so normally there is no practical limit. If you |
892 |
queue up many requests in a loop it often improves speed if you set |
893 |
this to a relatively low number, such as C<100>. |
894 |
|
895 |
This function does not work well together with C<aio_group>'s, and their |
896 |
feeder interface is better suited to limiting concurrency, so do not use |
897 |
this function. |
898 |
|
899 |
Under normal circumstances you don't need to call this function. |
900 |
|
901 |
=back |
902 |
|
903 |
=cut |
904 |
|
905 |
# support function to convert a fd into a perl filehandle |
906 |
sub _fd2fh { |
907 |
return undef if $_[0] < 0; |
908 |
|
909 |
# try to generate nice filehandles |
910 |
my $sym = "IO::AIO::fd#$_[0]"; |
911 |
local *$sym; |
912 |
|
913 |
open *$sym, "+<&=$_[0]" # usually works under any unix |
914 |
or open *$sym, "<&=$_[0]" # cygwin needs this |
915 |
or open *$sym, ">&=$_[0]" # or this |
916 |
or return undef; |
917 |
|
918 |
*$sym |
919 |
} |
920 |
|
921 |
min_parallel 8; |
922 |
|
923 |
END { |
924 |
max_parallel 0; |
925 |
} |
926 |
|
927 |
1; |
928 |
|
929 |
=head2 FORK BEHAVIOUR |
930 |
|
931 |
This module should do "the right thing" when the process using it forks: |
932 |
|
933 |
Before the fork, IO::AIO enters a quiescent state where no requests |
934 |
can be added in other threads and no results will be processed. After |
935 |
the fork the parent simply leaves the quiescent state and continues |
936 |
request/result processing, while the child frees the request/result queue |
937 |
(so that the requests started before the fork will only be handled in the |
938 |
parent). Threads will be started on demand until the limit set in the |
939 |
parent process has been reached again. |
940 |
|
941 |
In short: the parent will, after a short pause, continue as if fork had |
942 |
not been called, while the child will act as if IO::AIO has not been used |
943 |
yet. |
944 |
|
945 |
=head2 MEMORY USAGE |
946 |
|
947 |
Per-request usage: |
948 |
|
949 |
Each aio request uses - depending on your architecture - around 100-200 |
950 |
bytes of memory. In addition, stat requests need a stat buffer (possibly |
951 |
a few hundred bytes), readdir requires a result buffer and so on. Perl |
952 |
scalars and other data passed into aio requests will also be locked and |
953 |
will consume memory till the request has entered the done state. |
954 |
|
955 |
This is now awfully much, so queuing lots of requests is not usually a |
956 |
problem. |
957 |
|
958 |
Per-thread usage: |
959 |
|
960 |
In the execution phase, some aio requests require more memory for |
961 |
temporary buffers, and each thread requires a stack and other data |
962 |
structures (usually around 16k-128k, depending on the OS). |
963 |
|
964 |
=head1 KNOWN BUGS |
965 |
|
966 |
Known bugs will be fixed in the next release. |
967 |
|
968 |
=head1 SEE ALSO |
969 |
|
970 |
L<Coro::AIO>. |
971 |
|
972 |
=head1 AUTHOR |
973 |
|
974 |
Marc Lehmann <schmorp@schmorp.de> |
975 |
http://home.schmorp.de/ |
976 |
|
977 |
=cut |
978 |
|