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

Comparing IO-AIO/AIO.pm (file contents):
Revision 1.27 by root, Tue Aug 16 22:22:18 2005 UTC vs.
Revision 1.52 by root, Sat Oct 21 23:06:04 2006 UTC

14 aio_unlink "/tmp/file", sub { }; 14 aio_unlink "/tmp/file", sub { };
15 15
16 aio_read $fh, 30000, 1024, $buffer, 0, sub { 16 aio_read $fh, 30000, 1024, $buffer, 0, sub {
17 $_[0] > 0 or die "read error: $!"; 17 $_[0] > 0 or die "read error: $!";
18 }; 18 };
19
20 use IO::AIO 2; # version has aio objects
21
22 my $req = aio_unlink "/tmp/file", sub { };
23 $req->cancel; # cancel request if still in queue
24
25 # AnyEvent
26 open my $fh, "<&=" . IO::AIO::poll_fileno or die "$!";
27 my $w = AnyEvent->io (fh => $fh, poll => 'r', cb => sub { IO::AIO::poll_cb });
19 28
20 # Event 29 # Event
21 Event->io (fd => IO::AIO::poll_fileno, 30 Event->io (fd => IO::AIO::poll_fileno,
22 poll => 'r', 31 poll => 'r',
23 cb => \&IO::AIO::poll_cb); 32 cb => \&IO::AIO::poll_cb);
57=cut 66=cut
58 67
59package IO::AIO; 68package IO::AIO;
60 69
61no warnings; 70no warnings;
71use strict 'vars';
62 72
63use base 'Exporter'; 73use base 'Exporter';
64 74
65use Fcntl ();
66
67BEGIN { 75BEGIN {
68 $VERSION = 1.1; 76 our $VERSION = '1.8';
69 77
70 @EXPORT = qw(aio_read aio_write aio_open aio_close aio_stat aio_lstat aio_unlink 78 our @EXPORT = qw(aio_sendfile aio_read aio_write aio_open aio_close aio_stat
71 aio_rmdir aio_symlink aio_fsync aio_fdatasync aio_readahead); 79 aio_lstat aio_unlink aio_rmdir aio_readdir aio_scandir aio_symlink
80 aio_fsync aio_fdatasync aio_readahead aio_rename aio_link aio_move);
72 @EXPORT_OK = qw(poll_fileno poll_cb min_parallel max_parallel max_outstanding nreqs); 81 our @EXPORT_OK = qw(poll_fileno poll_cb min_parallel max_parallel max_outstanding nreqs);
73 82
74 require XSLoader; 83 require XSLoader;
75 XSLoader::load IO::AIO, $VERSION; 84 XSLoader::load ("IO::AIO", $VERSION);
76} 85}
77 86
78=head1 FUNCTIONS 87=head1 FUNCTIONS
79 88
80=head2 AIO FUNCTIONS 89=head2 AIO FUNCTIONS
88syscall has been executed asynchronously. 97syscall has been executed asynchronously.
89 98
90All functions expecting a filehandle keep a copy of the filehandle 99All functions expecting a filehandle keep a copy of the filehandle
91internally until the request has finished. 100internally until the request has finished.
92 101
102All non-composite requests (requests that are not broken down into
103multiple requests) return objects of type L<IO::AIO::REQ> that allow
104further manipulation of running requests.
105
93The filenames you pass to these routines I<must> be absolute. The reason 106The pathnames you pass to these routines I<must> be absolute and
94for this is that at the time the request is being executed, the current 107encoded in byte form. The reason for the former is that at the time the
95working directory could have changed. Alternatively, you can make sure 108request is being executed, the current working directory could have
96that you never change the current working directory. 109changed. Alternatively, you can make sure that you never change the
110current working directory.
111
112To encode pathnames to byte form, either make sure you either: a)
113always pass in filenames you got from outside (command line, readdir
114etc.), b) are ASCII or ISO 8859-1, c) use the Encode module and encode
115your pathnames to the locale (or other) encoding in effect in the user
116environment, d) use Glib::filename_from_unicode on unicode filenames or e)
117use something else.
97 118
98=over 4 119=over 4
99 120
100=item aio_open $pathname, $flags, $mode, $callback 121=item aio_open $pathname, $flags, $mode, $callback->($fh)
101 122
102Asynchronously open or create a file and call the callback with a newly 123Asynchronously open or create a file and call the callback with a newly
103created filehandle for the file. 124created filehandle for the file.
104 125
105The pathname passed to C<aio_open> must be absolute. See API NOTES, above, 126The pathname passed to C<aio_open> must be absolute. See API NOTES, above,
122 } else { 143 } else {
123 die "open failed: $!\n"; 144 die "open failed: $!\n";
124 } 145 }
125 }; 146 };
126 147
127=item aio_close $fh, $callback 148=item aio_close $fh, $callback->($status)
128 149
129Asynchronously close a file and call the callback with the result 150Asynchronously close a file and call the callback with the result
130code. I<WARNING:> although accepted, you should not pass in a perl 151code. I<WARNING:> although accepted, you should not pass in a perl
131filehandle here, as perl will likely close the file descriptor another 152filehandle here, as perl will likely close the file descriptor another
132time when the filehandle is destroyed. Normally, you can safely call perls 153time when the filehandle is destroyed. Normally, you can safely call perls
133C<close> or just let filehandles go out of scope. 154C<close> or just let filehandles go out of scope.
134 155
135This is supposed to be a bug in the API, so that might change. It's 156This is supposed to be a bug in the API, so that might change. It's
136therefore best to avoid this function. 157therefore best to avoid this function.
137 158
138=item aio_read $fh,$offset,$length, $data,$dataoffset,$callback 159=item aio_read $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
139 160
140=item aio_write $fh,$offset,$length, $data,$dataoffset,$callback 161=item aio_write $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
141 162
142Reads or writes C<length> bytes from the specified C<fh> and C<offset> 163Reads or writes C<length> bytes from the specified C<fh> and C<offset>
143into the scalar given by C<data> and offset C<dataoffset> and calls the 164into the scalar given by C<data> and offset C<dataoffset> and calls the
144callback without the actual number of bytes read (or -1 on error, just 165callback without the actual number of bytes read (or -1 on error, just
145like the syscall). 166like the syscall).
146 167
168The C<$data> scalar I<MUST NOT> be modified in any way while the request
169is outstanding. Modifying it can result in segfaults or WW3 (if the
170necessary/optional hardware is installed).
171
147Example: Read 15 bytes at offset 7 into scalar C<$buffer>, starting at 172Example: Read 15 bytes at offset 7 into scalar C<$buffer>, starting at
148offset C<0> within the scalar: 173offset C<0> within the scalar:
149 174
150 aio_read $fh, 7, 15, $buffer, 0, sub { 175 aio_read $fh, 7, 15, $buffer, 0, sub {
151 $_[0] > 0 or die "read error: $!"; 176 $_[0] > 0 or die "read error: $!";
152 print "read $_[0] bytes: <$buffer>\n"; 177 print "read $_[0] bytes: <$buffer>\n";
153 }; 178 };
154 179
180=item aio_move $srcpath, $dstpath, $callback->($status)
181
182Try to move the I<file> (directories not supported as either source or
183destination) from C<$srcpath> to C<$dstpath> and call the callback with
184the C<0> (error) or C<-1> ok.
185
186This is a composite request that tries to rename(2) the file first. If
187rename files with C<EXDEV>, it creates the destination file with mode 0200
188and copies the contents of the source file into it using C<aio_sendfile>,
189followed by restoring atime, mtime, access mode and uid/gid, in that
190order, and unlinking the C<$srcpath>.
191
192If an error occurs, the partial destination file will be unlinked, if
193possible, except when setting atime, mtime, access mode and uid/gid, where
194errors are being ignored.
195
196=cut
197
198sub aio_move($$$) {
199 my ($src, $dst, $cb) = @_;
200
201 aio_rename $src, $dst, sub {
202 if ($_[0] && $! == EXDEV) {
203 aio_open $src, O_RDONLY, 0, sub {
204 if (my $src_fh = $_[0]) {
205 my @stat = stat $src_fh;
206
207 aio_open $dst, O_WRONLY, 0200, sub {
208 if (my $dst_fh = $_[0]) {
209 aio_sendfile $dst_fh, $src_fh, 0, $stat[7], sub {
210 close $src_fh;
211
212 if ($_[0] == $stat[7]) {
213 utime $stat[8], $stat[9], $dst;
214 chmod $stat[2] & 07777, $dst_fh;
215 chown $stat[4], $stat[5], $dst_fh;
216 close $dst_fh;
217
218 aio_unlink $src, sub {
219 $cb->($_[0]);
220 };
221 } else {
222 my $errno = $!;
223 aio_unlink $dst, sub {
224 $! = $errno;
225 $cb->(-1);
226 };
227 }
228 };
229 } else {
230 $cb->(-1);
231 }
232 },
233
234 } else {
235 $cb->(-1);
236 }
237 };
238 } else {
239 $cb->($_[0]);
240 }
241 };
242}
243
244=item aio_sendfile $out_fh, $in_fh, $in_offset, $length, $callback->($retval)
245
246Tries to copy C<$length> bytes from C<$in_fh> to C<$out_fh>. It starts
247reading at byte offset C<$in_offset>, and starts writing at the current
248file offset of C<$out_fh>. Because of that, it is not safe to issue more
249than one C<aio_sendfile> per C<$out_fh>, as they will interfere with each
250other.
251
252This call tries to make use of a native C<sendfile> syscall to provide
253zero-copy operation. For this to work, C<$out_fh> should refer to a
254socket, and C<$in_fh> should refer to mmap'able file.
255
256If the native sendfile call fails or is not implemented, it will be
257emulated, so you can call C<aio_sendfile> on any type of filehandle
258regardless of the limitations of the operating system.
259
260Please note, however, that C<aio_sendfile> can read more bytes from
261C<$in_fh> than are written, and there is no way to find out how many
262bytes have been read from C<aio_sendfile> alone, as C<aio_sendfile> only
263provides the number of bytes written to C<$out_fh>. Only if the result
264value equals C<$length> one can assume that C<$length> bytes have been
265read.
266
155=item aio_readahead $fh,$offset,$length, $callback 267=item aio_readahead $fh,$offset,$length, $callback->($retval)
156 268
157C<aio_readahead> populates the page cache with data from a file so that 269C<aio_readahead> populates the page cache with data from a file so that
158subsequent reads from that file will not block on disk I/O. The C<$offset> 270subsequent reads from that file will not block on disk I/O. The C<$offset>
159argument specifies the starting point from which data is to be read and 271argument specifies the starting point from which data is to be read and
160C<$length> specifies the number of bytes to be read. I/O is performed in 272C<$length> specifies the number of bytes to be read. I/O is performed in
164file. The current file offset of the file is left unchanged. 276file. The current file offset of the file is left unchanged.
165 277
166If that syscall doesn't exist (likely if your OS isn't Linux) it will be 278If that syscall doesn't exist (likely if your OS isn't Linux) it will be
167emulated by simply reading the data, which would have a similar effect. 279emulated by simply reading the data, which would have a similar effect.
168 280
169=item aio_stat $fh_or_path, $callback 281=item aio_stat $fh_or_path, $callback->($status)
170 282
171=item aio_lstat $fh, $callback 283=item aio_lstat $fh, $callback->($status)
172 284
173Works like perl's C<stat> or C<lstat> in void context. The callback will 285Works like perl's C<stat> or C<lstat> in void context. The callback will
174be called after the stat and the results will be available using C<stat _> 286be called after the stat and the results will be available using C<stat _>
175or C<-s _> etc... 287or C<-s _> etc...
176 288
186 aio_stat "/etc/passwd", sub { 298 aio_stat "/etc/passwd", sub {
187 $_[0] and die "stat failed: $!"; 299 $_[0] and die "stat failed: $!";
188 print "size is ", -s _, "\n"; 300 print "size is ", -s _, "\n";
189 }; 301 };
190 302
191=item aio_unlink $pathname, $callback 303=item aio_unlink $pathname, $callback->($status)
192 304
193Asynchronously unlink (delete) a file and call the callback with the 305Asynchronously unlink (delete) a file and call the callback with the
194result code. 306result code.
195 307
308=item aio_link $srcpath, $dstpath, $callback->($status)
309
310Asynchronously create a new link to the existing object at C<$srcpath> at
311the path C<$dstpath> and call the callback with the result code.
312
313=item aio_symlink $srcpath, $dstpath, $callback->($status)
314
315Asynchronously create a new symbolic link to the existing object at C<$srcpath> at
316the path C<$dstpath> and call the callback with the result code.
317
318=item aio_rename $srcpath, $dstpath, $callback->($status)
319
320Asynchronously rename the object at C<$srcpath> to C<$dstpath>, just as
321rename(2) and call the callback with the result code.
322
196=item aio_rmdir $pathname, $callback 323=item aio_rmdir $pathname, $callback->($status)
197 324
198Asynchronously rmdir (delete) a directory and call the callback with the 325Asynchronously rmdir (delete) a directory and call the callback with the
199result code. 326result code.
200 327
328=item aio_readdir $pathname, $callback->($entries)
329
330Unlike the POSIX call of the same name, C<aio_readdir> reads an entire
331directory (i.e. opendir + readdir + closedir). The entries will not be
332sorted, and will B<NOT> include the C<.> and C<..> entries.
333
334The callback a single argument which is either C<undef> or an array-ref
335with the filenames.
336
337=item aio_scandir $path, $maxreq, $callback->($dirs, $nondirs)
338
339Scans a directory (similar to C<aio_readdir>) but additionally tries to
340separate the entries of directory C<$path> into two sets of names, ones
341you can recurse into (directories or links to them), and ones you cannot
342recurse into (everything else).
343
344C<aio_scandir> is a composite request that consists of many sub
345requests. C<$maxreq> specifies the maximum number of outstanding aio
346requests that this function generates. If it is C<< <= 0 >>, then a
347suitable default will be chosen (currently 8).
348
349On error, the callback is called without arguments, otherwise it receives
350two array-refs with path-relative entry names.
351
352Example:
353
354 aio_scandir $dir, 0, sub {
355 my ($dirs, $nondirs) = @_;
356 print "real directories: @$dirs\n";
357 print "everything else: @$nondirs\n";
358 };
359
360Implementation notes.
361
362The C<aio_readdir> cannot be avoided, but C<stat()>'ing every entry can.
363
364After reading the directory, the modification time, size etc. of the
365directory before and after the readdir is checked, and if they match (and
366isn't the current time), the link count will be used to decide how many
367entries are directories (if >= 2). Otherwise, no knowledge of the number
368of subdirectories will be assumed.
369
370Then entries will be sorted into likely directories (everything without
371a non-initial dot currently) and likely non-directories (everything
372else). Then every entry plus an appended C</.> will be C<stat>'ed,
373likely directories first. If that succeeds, it assumes that the entry
374is a directory or a symlink to directory (which will be checked
375seperately). This is often faster than stat'ing the entry itself because
376filesystems might detect the type of the entry without reading the inode
377data (e.g. ext2fs filetype feature).
378
379If the known number of directories (link count - 2) has been reached, the
380rest of the entries is assumed to be non-directories.
381
382This only works with certainty on POSIX (= UNIX) filesystems, which
383fortunately are the vast majority of filesystems around.
384
385It will also likely work on non-POSIX filesystems with reduced efficiency
386as those tend to return 0 or 1 as link counts, which disables the
387directory counting heuristic.
388
389=cut
390
391sub aio_scandir($$$) {
392 my ($path, $maxreq, $cb) = @_;
393
394 $maxreq = 8 if $maxreq <= 0;
395
396 # stat once
397 aio_stat $path, sub {
398 return $cb->() if $_[0];
399 my $now = time;
400 my $hash1 = join ":", (stat _)[0,1,3,7,9];
401
402 # read the directory entries
403 aio_readdir $path, sub {
404 my $entries = shift
405 or return $cb->();
406
407 # stat the dir another time
408 aio_stat $path, sub {
409 my $hash2 = join ":", (stat _)[0,1,3,7,9];
410
411 my $ndirs;
412
413 # take the slow route if anything looks fishy
414 if ($hash1 ne $hash2 or (stat _)[9] == $now) {
415 $ndirs = -1;
416 } else {
417 # if nlink == 2, we are finished
418 # on non-posix-fs's, we rely on nlink < 2
419 $ndirs = (stat _)[3] - 2
420 or return $cb->([], $entries);
421 }
422
423 # sort into likely dirs and likely nondirs
424 # dirs == files without ".", short entries first
425 $entries = [map $_->[0],
426 sort { $b->[1] cmp $a->[1] }
427 map [$_, sprintf "%s%04d", (/.\./ ? "1" : "0"), length],
428 @$entries];
429
430 my (@dirs, @nondirs);
431
432 my ($statcb, $schedcb);
433 my $nreq = 0;
434
435 $schedcb = sub {
436 if (@$entries) {
437 if ($nreq < $maxreq) {
438 my $ent = pop @$entries;
439 $nreq++;
440 aio_stat "$path/$ent/.", sub { $statcb->($_[0], $ent) };
441 }
442 } elsif (!$nreq) {
443 # finished
444 undef $statcb;
445 undef $schedcb;
446 $cb->(\@dirs, \@nondirs) if $cb;
447 undef $cb;
448 }
449 };
450 $statcb = sub {
451 my ($status, $entry) = @_;
452
453 if ($status < 0) {
454 $nreq--;
455 push @nondirs, $entry;
456 &$schedcb;
457 } else {
458 # need to check for real directory
459 aio_lstat "$path/$entry", sub {
460 $nreq--;
461
462 if (-d _) {
463 push @dirs, $entry;
464
465 if (!--$ndirs) {
466 push @nondirs, @$entries;
467 $entries = [];
468 }
469 } else {
470 push @nondirs, $entry;
471 }
472
473 &$schedcb;
474 }
475 }
476 };
477
478 &$schedcb while @$entries && $nreq < $maxreq;
479 };
480 };
481 };
482}
483
201=item aio_fsync $fh, $callback 484=item aio_fsync $fh, $callback->($status)
202 485
203Asynchronously call fsync on the given filehandle and call the callback 486Asynchronously call fsync on the given filehandle and call the callback
204with the fsync result code. 487with the fsync result code.
205 488
206=item aio_fdatasync $fh, $callback 489=item aio_fdatasync $fh, $callback->($status)
207 490
208Asynchronously call fdatasync on the given filehandle and call the 491Asynchronously call fdatasync on the given filehandle and call the
209callback with the fdatasync result code. 492callback with the fdatasync result code.
210 493
211If this call isn't available because your OS lacks it or it couldn't be 494If this call isn't available because your OS lacks it or it couldn't be
212detected, it will be emulated by calling C<fsync> instead. 495detected, it will be emulated by calling C<fsync> instead.
496
497=back
498
499=head2 IO::AIO::CB CLASS
500
501All non-aggregate C<aio_*> functions return an object of this class when
502called in non-void context.
503
504A request always moves through the following five states in its lifetime,
505in order: B<ready> (request has been created, but has not been executed
506yet), B<execute> (request is currently being executed), B<pending>
507(request has been executed but callback has not been called yet),
508B<result> (results are being processed synchronously, includes calling the
509callback) and B<done> (request has reached the end of its lifetime and
510holds no resources anymore).
511
512=over 4
513
514=item $req->cancel
515
516Cancels the request, if possible. Has the effect of skipping execution
517when entering the B<execute> state and skipping calling the callback when
518entering the the B<result> state, but will leave the request otherwise
519untouched. That means that requests that currently execute will not be
520stopped and resources held by the request will not be freed prematurely.
213 521
214=back 522=back
215 523
216=head2 SUPPORT FUNCTIONS 524=head2 SUPPORT FUNCTIONS
217 525
275 IO::AIO::poll_wait, IO::AIO::poll_cb 583 IO::AIO::poll_wait, IO::AIO::poll_cb
276 if IO::AIO::nreqs; 584 if IO::AIO::nreqs;
277 585
278=item IO::AIO::min_parallel $nthreads 586=item IO::AIO::min_parallel $nthreads
279 587
280Set the minimum number of AIO threads to C<$nthreads>. The default is 588Set the minimum number of AIO threads to C<$nthreads>. The current default
281C<1>, which means a single asynchronous operation can be done at one time 589is C<4>, which means four asynchronous operations can be done at one time
282(the number of outstanding operations, however, is unlimited). 590(the number of outstanding operations, however, is unlimited).
591
592IO::AIO starts threads only on demand, when an AIO request is queued and
593no free thread exists.
283 594
284It is recommended to keep the number of threads low, as some Linux 595It is recommended to keep the number of threads low, as some Linux
285kernel versions will scale negatively with the number of threads (higher 596kernel versions will scale negatively with the number of threads (higher
286parallelity => MUCH higher latency). With current Linux 2.6 versions, 4-32 597parallelity => MUCH higher latency). With current Linux 2.6 versions, 4-32
287threads should be fine. 598threads should be fine.
288 599
289Under normal circumstances you don't need to call this function, as this 600Under most circumstances you don't need to call this function, as the
290module automatically starts some threads (the exact number might change, 601module selects a default that is suitable for low to moderate load.
291and is currently 4).
292 602
293=item IO::AIO::max_parallel $nthreads 603=item IO::AIO::max_parallel $nthreads
294 604
295Sets the maximum number of AIO threads to C<$nthreads>. If more than 605Sets the maximum number of AIO threads to C<$nthreads>. If more than the
296the specified number of threads are currently running, kill them. This 606specified number of threads are currently running, this function kills
297function blocks until the limit is reached. 607them. This function blocks until the limit is reached.
608
609While C<$nthreads> are zero, aio requests get queued but not executed
610until the number of threads has been increased again.
298 611
299This module automatically runs C<max_parallel 0> at program end, to ensure 612This module automatically runs C<max_parallel 0> at program end, to ensure
300that all threads are killed and that there are no outstanding requests. 613that all threads are killed and that there are no outstanding requests.
301 614
302Under normal circumstances you don't need to call this function. 615Under normal circumstances you don't need to call this function.
306Sets the maximum number of outstanding requests to C<$nreqs>. If you 619Sets the maximum number of outstanding requests to C<$nreqs>. If you
307try to queue up more than this number of requests, the caller will block until 620try to queue up more than this number of requests, the caller will block until
308some requests have been handled. 621some requests have been handled.
309 622
310The default is very large, so normally there is no practical limit. If you 623The default is very large, so normally there is no practical limit. If you
311queue up many requests in a loop it it often improves speed if you set 624queue up many requests in a loop it often improves speed if you set
312this to a relatively low number, such as C<100>. 625this to a relatively low number, such as C<100>.
313 626
314Under normal circumstances you don't need to call this function. 627Under normal circumstances you don't need to call this function.
315 628
316=back 629=back
341 654
3421; 6551;
343 656
344=head2 FORK BEHAVIOUR 657=head2 FORK BEHAVIOUR
345 658
346IO::AIO handles all outstanding AIO requests before the fork, destroys all 659This module should do "the right thing" when the process using it forks:
347AIO threads, and recreates them in both the parent and the child after the
348fork.
349 660
661Before the fork, IO::AIO enters a quiescent state where no requests
662can be added in other threads and no results will be processed. After
663the fork the parent simply leaves the quiescent state and continues
664request/result processing, while the child clears the request/result
665queue (so the requests started before the fork will only be handled in
666the parent). Threads will be started on demand until the limit ste in the
667parent process has been reached again.
668
669In short: the parent will, after a short pause, continue as if fork had
670not been called, while the child will act as if IO::AIO has not been used
671yet.
350 672
351=head1 SEE ALSO 673=head1 SEE ALSO
352 674
353L<Coro>, L<Linux::AIO>. 675L<Coro>, L<Linux::AIO> (obsolete).
354 676
355=head1 AUTHOR 677=head1 AUTHOR
356 678
357 Marc Lehmann <schmorp@schmorp.de> 679 Marc Lehmann <schmorp@schmorp.de>
358 http://home.schmorp.de/ 680 http://home.schmorp.de/

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines