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 1.21 by root, Wed Jul 13 00:13:09 2005 UTC vs.
Revision 1.107 by root, Fri Jun 1 05:52:59 2007 UTC

5=head1 SYNOPSIS 5=head1 SYNOPSIS
6 6
7 use IO::AIO; 7 use IO::AIO;
8 8
9 aio_open "/etc/passwd", O_RDONLY, 0, sub { 9 aio_open "/etc/passwd", O_RDONLY, 0, sub {
10 my ($fh) = @_; 10 my $fh = shift
11 or die "/etc/passwd: $!";
11 ... 12 ...
12 }; 13 };
13 14
14 aio_unlink "/tmp/file", sub { }; 15 aio_unlink "/tmp/file", sub { };
15 16
16 aio_read $fh, 30000, 1024, $buffer, 0, sub { 17 aio_read $fh, 30000, 1024, $buffer, 0, sub {
17 $_[0] > 0 or die "read error: $!"; 18 $_[0] > 0 or die "read error: $!";
18 }; 19 };
19 20
20 # Event 21 # version 2+ has request and group objects
22 use IO::AIO 2;
23
24 aioreq_pri 4; # give next request a very high priority
25 my $req = aio_unlink "/tmp/file", sub { };
26 $req->cancel; # cancel request if still in queue
27
28 my $grp = aio_group sub { print "all stats done\n" };
29 add $grp aio_stat "..." for ...;
30
31 # AnyEvent integration
32 open my $fh, "<&=" . IO::AIO::poll_fileno or die "$!";
33 my $w = AnyEvent->io (fh => $fh, poll => 'r', cb => sub { IO::AIO::poll_cb });
34
35 # Event integration
21 Event->io (fd => IO::AIO::poll_fileno, 36 Event->io (fd => IO::AIO::poll_fileno,
22 poll => 'r', 37 poll => 'r',
23 cb => \&IO::AIO::poll_cb); 38 cb => \&IO::AIO::poll_cb);
24 39
25 # Glib/Gtk2 40 # Glib/Gtk2 integration
26 add_watch Glib::IO IO::AIO::poll_fileno, 41 add_watch Glib::IO IO::AIO::poll_fileno,
27 in => sub { IO::AIO::poll_cb, 1 }; 42 in => sub { IO::AIO::poll_cb; 1 };
28 43
29 # Tk 44 # Tk integration
30 Tk::Event::IO->fileevent (IO::AIO::poll_fileno, "", 45 Tk::Event::IO->fileevent (IO::AIO::poll_fileno, "",
31 readable => \&IO::AIO::poll_cb); 46 readable => \&IO::AIO::poll_cb);
32 47
33 # Danga::Socket 48 # Danga::Socket integration
34 Danga::Socket->AddOtherFds (IO::AIO::poll_fileno => 49 Danga::Socket->AddOtherFds (IO::AIO::poll_fileno =>
35 \&IO::AIO::poll_cb); 50 \&IO::AIO::poll_cb);
36 51
37
38=head1 DESCRIPTION 52=head1 DESCRIPTION
39 53
40This module implements asynchronous I/O using whatever means your 54This module implements asynchronous I/O using whatever means your
41operating system supports. 55operating system supports.
42 56
57Asynchronous means that operations that can normally block your program
58(e.g. reading from disk) will be done asynchronously: the operation
59will still block, but you can do something else in the meantime. This
60is extremely useful for programs that need to stay interactive even
61when doing heavy I/O (GUI programs, high performance network servers
62etc.), but can also be used to easily do operations in parallel that are
63normally done sequentially, e.g. stat'ing many files, which is much faster
64on a RAID volume or over NFS when you do a number of stat operations
65concurrently.
66
67While most of this works on all types of file descriptors (for example
68sockets), using these functions on file descriptors that support
69nonblocking operation (again, sockets, pipes etc.) is very inefficient or
70might not work (aio_read fails on sockets/pipes/fifos). Use an event loop
71for that (such as the L<Event|Event> module): IO::AIO will naturally fit
72into such an event loop itself.
73
43Currently, a number of threads are started that execute your read/writes 74In this version, a number of threads are started that execute your
44and signal their completion. You don't need thread support in your libc or 75requests and signal their completion. You don't need thread support
45perl, and the threads created by this module will not be visible to the 76in perl, and the threads created by this module will not be visible
46pthreads library. In the future, this module might make use of the native 77to perl. In the future, this module might make use of the native aio
47aio functions available on many operating systems. However, they are often 78functions available on many operating systems. However, they are often
48not well-supported (Linux doesn't allow them on normal files currently, 79not well-supported or restricted (GNU/Linux doesn't allow them on normal
49for example), and they would only support aio_read and aio_write, so the 80files currently, for example), and they would only support aio_read and
50remaining functionality would have to be implemented using threads anyway. 81aio_write, so the remaining functionality would have to be implemented
82using threads anyway.
51 83
52Although the module will work with in the presence of other threads, it is 84Although the module will work with in the presence of other (Perl-)
53currently not reentrant, so use appropriate locking yourself. 85threads, it is currently not reentrant in any way, so use appropriate
86locking yourself, always call C<poll_cb> from within the same thread, or
87never call C<poll_cb> (or other C<aio_> functions) recursively.
88
89=head2 EXAMPLE
90
91This is a simple example that uses the Event module and loads
92F</etc/passwd> asynchronously:
93
94 use Fcntl;
95 use Event;
96 use IO::AIO;
97
98 # register the IO::AIO callback with Event
99 Event->io (fd => IO::AIO::poll_fileno,
100 poll => 'r',
101 cb => \&IO::AIO::poll_cb);
102
103 # queue the request to open /etc/passwd
104 aio_open "/etc/passwd", O_RDONLY, 0, sub {
105 my $fh = shift
106 or die "error while opening: $!";
107
108 # stat'ing filehandles is generally non-blocking
109 my $size = -s $fh;
110
111 # queue a request to read the file
112 my $contents;
113 aio_read $fh, 0, $size, $contents, 0, sub {
114 $_[0] == $size
115 or die "short read: $!";
116
117 close $fh;
118
119 # file contents now in $contents
120 print $contents;
121
122 # exit event loop and program
123 Event::unloop;
124 };
125 };
126
127 # possibly queue up other requests, or open GUI windows,
128 # check for sockets etc. etc.
129
130 # process events as long as there are some:
131 Event::loop;
132
133=head1 REQUEST ANATOMY AND LIFETIME
134
135Every C<aio_*> function creates a request. which is a C data structure not
136directly visible to Perl.
137
138If called in non-void context, every request function returns a Perl
139object representing the request. In void context, nothing is returned,
140which saves a bit of memory.
141
142The perl object is a fairly standard ref-to-hash object. The hash contents
143are not used by IO::AIO so you are free to store anything you like in it.
144
145During their existance, aio requests travel through the following states,
146in order:
147
148=over 4
149
150=item ready
151
152Immediately after a request is created it is put into the ready state,
153waiting for a thread to execute it.
154
155=item execute
156
157A thread has accepted the request for processing and is currently
158executing it (e.g. blocking in read).
159
160=item pending
161
162The request has been executed and is waiting for result processing.
163
164While request submission and execution is fully asynchronous, result
165processing is not and relies on the perl interpreter calling C<poll_cb>
166(or another function with the same effect).
167
168=item result
169
170The request results are processed synchronously by C<poll_cb>.
171
172The C<poll_cb> function will process all outstanding aio requests by
173calling their callbacks, freeing memory associated with them and managing
174any groups they are contained in.
175
176=item done
177
178Request has reached the end of its lifetime and holds no resources anymore
179(except possibly for the Perl object, but its connection to the actual
180aio request is severed and calling its methods will either do nothing or
181result in a runtime error).
182
183=back
54 184
55=cut 185=cut
56 186
57package IO::AIO; 187package IO::AIO;
58 188
189no warnings;
190use strict 'vars';
191
59use base 'Exporter'; 192use base 'Exporter';
60 193
61use Fcntl ();
62
63BEGIN { 194BEGIN {
64 $VERSION = 0.9; 195 our $VERSION = '2.4';
65 196
66 @EXPORT = qw(aio_read aio_write aio_open aio_close aio_stat aio_lstat aio_unlink 197 our @AIO_REQ = qw(aio_sendfile aio_read aio_write aio_open aio_close aio_stat
67 aio_fsync aio_fdatasync aio_readahead); 198 aio_lstat aio_unlink aio_rmdir aio_readdir aio_scandir aio_symlink
68 @EXPORT_OK = qw(poll_fileno poll_cb min_parallel max_parallel max_outstanding nreqs); 199 aio_readlink aio_fsync aio_fdatasync aio_readahead aio_rename aio_link
200 aio_move aio_copy aio_group aio_nop aio_mknod aio_load aio_rmtree aio_mkdir
201 aio_chown aio_chmod aio_utime);
202 our @EXPORT = (@AIO_REQ, qw(aioreq_pri aioreq_nice aio_block));
203 our @EXPORT_OK = qw(poll_fileno poll_cb poll_wait flush
204 min_parallel max_parallel max_idle
205 nreqs nready npending nthreads
206 max_poll_time max_poll_reqs);
207
208 @IO::AIO::GRP::ISA = 'IO::AIO::REQ';
69 209
70 require XSLoader; 210 require XSLoader;
71 XSLoader::load IO::AIO, $VERSION; 211 XSLoader::load ("IO::AIO", $VERSION);
72} 212}
73 213
74=head1 FUNCTIONS 214=head1 FUNCTIONS
75 215
76=head2 AIO FUNCTIONS 216=head2 AIO REQUEST FUNCTIONS
77 217
78All the C<aio_*> calls are more or less thin wrappers around the syscall 218All the C<aio_*> calls are more or less thin wrappers around the syscall
79with the same name (sans C<aio_>). The arguments are similar or identical, 219with the same name (sans C<aio_>). The arguments are similar or identical,
80and they all accept an additional (and optional) C<$callback> argument 220and they all accept an additional (and optional) C<$callback> argument
81which must be a code reference. This code reference will get called with 221which must be a code reference. This code reference will get called with
82the syscall return code (e.g. most syscalls return C<-1> on error, unlike 222the syscall return code (e.g. most syscalls return C<-1> on error, unlike
83perl, which usually delivers "false") as it's sole argument when the given 223perl, which usually delivers "false") as it's sole argument when the given
84syscall has been executed asynchronously. 224syscall has been executed asynchronously.
85 225
86All functions that expect a filehandle will also accept a file descriptor. 226All functions expecting a filehandle keep a copy of the filehandle
227internally until the request has finished.
87 228
229All functions return request objects of type L<IO::AIO::REQ> that allow
230further manipulation of those requests while they are in-flight.
231
88The filenames you pass to these routines I<must> be absolute. The reason 232The pathnames you pass to these routines I<must> be absolute and
89for this is that at the time the request is being executed, the current 233encoded as octets. The reason for the former is that at the time the
90working directory could have changed. Alternatively, you can make sure 234request is being executed, the current working directory could have
91that you never change the current working directory. 235changed. Alternatively, you can make sure that you never change the
236current working directory anywhere in the program and then use relative
237paths.
238
239To encode pathnames as octets, either make sure you either: a) always pass
240in filenames you got from outside (command line, readdir etc.) without
241tinkering, b) are ASCII or ISO 8859-1, c) use the Encode module and encode
242your pathnames to the locale (or other) encoding in effect in the user
243environment, d) use Glib::filename_from_unicode on unicode filenames or e)
244use something else to ensure your scalar has the correct contents.
245
246This works, btw. independent of the internal UTF-8 bit, which IO::AIO
247handles correctly wether it is set or not.
92 248
93=over 4 249=over 4
94 250
251=item $prev_pri = aioreq_pri [$pri]
252
253Returns the priority value that would be used for the next request and, if
254C<$pri> is given, sets the priority for the next aio request.
255
256The default priority is C<0>, the minimum and maximum priorities are C<-4>
257and C<4>, respectively. Requests with higher priority will be serviced
258first.
259
260The priority will be reset to C<0> after each call to one of the C<aio_*>
261functions.
262
263Example: open a file with low priority, then read something from it with
264higher priority so the read request is serviced before other low priority
265open requests (potentially spamming the cache):
266
267 aioreq_pri -3;
268 aio_open ..., sub {
269 return unless $_[0];
270
271 aioreq_pri -2;
272 aio_read $_[0], ..., sub {
273 ...
274 };
275 };
276
277
278=item aioreq_nice $pri_adjust
279
280Similar to C<aioreq_pri>, but subtracts the given value from the current
281priority, so the effect is cumulative.
282
283
95=item aio_open $pathname, $flags, $mode, $callback 284=item aio_open $pathname, $flags, $mode, $callback->($fh)
96 285
97Asynchronously open or create a file and call the callback with a newly 286Asynchronously open or create a file and call the callback with a newly
98created filehandle for the file. 287created filehandle for the file.
99 288
100The pathname passed to C<aio_open> must be absolute. See API NOTES, above, 289The pathname passed to C<aio_open> must be absolute. See API NOTES, above,
104list. They are the same as used by C<sysopen>. 293list. They are the same as used by C<sysopen>.
105 294
106Likewise, C<$mode> specifies the mode of the newly created file, if it 295Likewise, C<$mode> specifies the mode of the newly created file, if it
107didn't exist and C<O_CREAT> has been given, just like perl's C<sysopen>, 296didn't exist and C<O_CREAT> has been given, just like perl's C<sysopen>,
108except that it is mandatory (i.e. use C<0> if you don't create new files, 297except that it is mandatory (i.e. use C<0> if you don't create new files,
109and C<0666> or C<0777> if you do). 298and C<0666> or C<0777> if you do). Note that the C<$mode> will be modified
299by the umask in effect then the request is being executed, so better never
300change the umask.
110 301
111Example: 302Example:
112 303
113 aio_open "/etc/passwd", O_RDONLY, 0, sub { 304 aio_open "/etc/passwd", O_RDONLY, 0, sub {
114 if ($_[0]) { 305 if ($_[0]) {
117 } else { 308 } else {
118 die "open failed: $!\n"; 309 die "open failed: $!\n";
119 } 310 }
120 }; 311 };
121 312
313
122=item aio_close $fh, $callback 314=item aio_close $fh, $callback->($status)
123 315
124Asynchronously close a file and call the callback with the result 316Asynchronously close a file and call the callback with the result
125code. I<WARNING:> although accepted, you should not pass in a perl 317code. I<WARNING:> although accepted, you should not pass in a perl
126filehandle here, as perl will likely close the file descriptor another 318filehandle here, as perl will likely close the file descriptor another
127time when the filehandle is destroyed. Normally, you can safely call perls 319time when the filehandle is destroyed. Normally, you can safely call perls
128C<close> or just let filehandles go out of scope. 320C<close> or just let filehandles go out of scope.
129 321
130This is supposed to be a bug in the API, so that might change. It's 322This is supposed to be a bug in the API, so that might change. It's
131therefore best to avoid this function. 323therefore best to avoid this function.
132 324
325
133=item aio_read $fh,$offset,$length, $data,$dataoffset,$callback 326=item aio_read $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
134 327
135=item aio_write $fh,$offset,$length, $data,$dataoffset,$callback 328=item aio_write $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
136 329
137Reads or writes C<length> bytes from the specified C<fh> and C<offset> 330Reads or writes C<length> bytes from the specified C<fh> and C<offset>
138into the scalar given by C<data> and offset C<dataoffset> and calls the 331into the scalar given by C<data> and offset C<dataoffset> and calls the
139callback without the actual number of bytes read (or -1 on error, just 332callback without the actual number of bytes read (or -1 on error, just
140like the syscall). 333like the syscall).
141 334
335The C<$data> scalar I<MUST NOT> be modified in any way while the request
336is outstanding. Modifying it can result in segfaults or WW3 (if the
337necessary/optional hardware is installed).
338
142Example: Read 15 bytes at offset 7 into scalar C<$buffer>, starting at 339Example: Read 15 bytes at offset 7 into scalar C<$buffer>, starting at
143offset C<0> within the scalar: 340offset C<0> within the scalar:
144 341
145 aio_read $fh, 7, 15, $buffer, 0, sub { 342 aio_read $fh, 7, 15, $buffer, 0, sub {
146 $_[0] > 0 or die "read error: $!"; 343 $_[0] > 0 or die "read error: $!";
147 print "read $_[0] bytes: <$buffer>\n"; 344 print "read $_[0] bytes: <$buffer>\n";
148 }; 345 };
149 346
347
348=item aio_sendfile $out_fh, $in_fh, $in_offset, $length, $callback->($retval)
349
350Tries to copy C<$length> bytes from C<$in_fh> to C<$out_fh>. It starts
351reading at byte offset C<$in_offset>, and starts writing at the current
352file offset of C<$out_fh>. Because of that, it is not safe to issue more
353than one C<aio_sendfile> per C<$out_fh>, as they will interfere with each
354other.
355
356This call tries to make use of a native C<sendfile> syscall to provide
357zero-copy operation. For this to work, C<$out_fh> should refer to a
358socket, and C<$in_fh> should refer to mmap'able file.
359
360If the native sendfile call fails or is not implemented, it will be
361emulated, so you can call C<aio_sendfile> on any type of filehandle
362regardless of the limitations of the operating system.
363
364Please note, however, that C<aio_sendfile> can read more bytes from
365C<$in_fh> than are written, and there is no way to find out how many
366bytes have been read from C<aio_sendfile> alone, as C<aio_sendfile> only
367provides the number of bytes written to C<$out_fh>. Only if the result
368value equals C<$length> one can assume that C<$length> bytes have been
369read.
370
371
150=item aio_readahead $fh,$offset,$length, $callback 372=item aio_readahead $fh,$offset,$length, $callback->($retval)
151
152Asynchronously reads the specified byte range into the page cache, using
153the C<readahead> syscall. If that syscall doesn't exist (likely if your OS
154isn't Linux) the status will be C<-1> and C<$!> is set to C<ENOSYS>.
155 373
156C<aio_readahead> populates the page cache with data from a file so that 374C<aio_readahead> populates the page cache with data from a file so that
157subsequent reads from that file will not block on disk I/O. The C<$offset> 375subsequent reads from that file will not block on disk I/O. The C<$offset>
158argument specifies the starting point from which data is to be read and 376argument specifies the starting point from which data is to be read and
159C<$length> specifies the number of bytes to be read. I/O is performed in 377C<$length> specifies the number of bytes to be read. I/O is performed in
160whole pages, so that offset is effectively rounded down to a page boundary 378whole pages, so that offset is effectively rounded down to a page boundary
161and bytes are read up to the next page boundary greater than or equal to 379and bytes are read up to the next page boundary greater than or equal to
162(off-set+length). C<aio_readahead> does not read beyond the end of the 380(off-set+length). C<aio_readahead> does not read beyond the end of the
163file. The current file offset of the file is left unchanged. 381file. The current file offset of the file is left unchanged.
164 382
383If that syscall doesn't exist (likely if your OS isn't Linux) it will be
384emulated by simply reading the data, which would have a similar effect.
385
386
165=item aio_stat $fh_or_path, $callback 387=item aio_stat $fh_or_path, $callback->($status)
166 388
167=item aio_lstat $fh, $callback 389=item aio_lstat $fh, $callback->($status)
168 390
169Works like perl's C<stat> or C<lstat> in void context. The callback will 391Works like perl's C<stat> or C<lstat> in void context. The callback will
170be called after the stat and the results will be available using C<stat _> 392be called after the stat and the results will be available using C<stat _>
171or C<-s _> etc... 393or C<-s _> etc...
172 394
182 aio_stat "/etc/passwd", sub { 404 aio_stat "/etc/passwd", sub {
183 $_[0] and die "stat failed: $!"; 405 $_[0] and die "stat failed: $!";
184 print "size is ", -s _, "\n"; 406 print "size is ", -s _, "\n";
185 }; 407 };
186 408
409
410=item aio_utime $fh_or_path, $atime, $mtime, $callback->($status)
411
412Works like perl's C<utime> function (including the special case of $atime
413and $mtime being undef). Fractional times are supported if the underlying
414syscalls support them.
415
416When called with a pathname, uses utimes(2) if available, otherwise
417utime(2). If called on a file descriptor, uses futimes(2) if available,
418otherwise returns ENOSYS, so this is not portable.
419
420Examples:
421
422 # set atime and mtime to current time (basically touch(1)):
423 aio_utime "path", undef, undef;
424 # set atime to current time and mtime to beginning of the epoch:
425 aio_utime "path", time, undef; # undef==0
426
427
428=item aio_chown $fh_or_path, $uid, $gid, $callback->($status)
429
430Works like perl's C<chown> function, except that C<undef> for either $uid
431or $gid is being interpreted as "do not change" (but -1 can also be used).
432
433Examples:
434
435 # same as "chown root path" in the shell:
436 aio_chown "path", 0, -1;
437 # same as above:
438 aio_chown "path", 0, undef;
439
440
441=item aio_chmod $fh_or_path, $mode, $callback->($status)
442
443Works like perl's C<chmod> function.
444
445
187=item aio_unlink $pathname, $callback 446=item aio_unlink $pathname, $callback->($status)
188 447
189Asynchronously unlink (delete) a file and call the callback with the 448Asynchronously unlink (delete) a file and call the callback with the
190result code. 449result code.
191 450
451
452=item aio_mknod $path, $mode, $dev, $callback->($status)
453
454[EXPERIMENTAL]
455
456Asynchronously create a device node (or fifo). See mknod(2).
457
458The only (POSIX-) portable way of calling this function is:
459
460 aio_mknod $path, IO::AIO::S_IFIFO | $mode, 0, sub { ...
461
462
463=item aio_link $srcpath, $dstpath, $callback->($status)
464
465Asynchronously create a new link to the existing object at C<$srcpath> at
466the path C<$dstpath> and call the callback with the result code.
467
468
469=item aio_symlink $srcpath, $dstpath, $callback->($status)
470
471Asynchronously create a new symbolic link to the existing object at C<$srcpath> at
472the path C<$dstpath> and call the callback with the result code.
473
474
475=item aio_readlink $path, $callback->($link)
476
477Asynchronously read the symlink specified by C<$path> and pass it to
478the callback. If an error occurs, nothing or undef gets passed to the
479callback.
480
481
482=item aio_rename $srcpath, $dstpath, $callback->($status)
483
484Asynchronously rename the object at C<$srcpath> to C<$dstpath>, just as
485rename(2) and call the callback with the result code.
486
487
488=item aio_mkdir $pathname, $mode, $callback->($status)
489
490Asynchronously mkdir (create) a directory and call the callback with
491the result code. C<$mode> will be modified by the umask at the time the
492request is executed, so do not change your umask.
493
494
495=item aio_rmdir $pathname, $callback->($status)
496
497Asynchronously rmdir (delete) a directory and call the callback with the
498result code.
499
500
501=item aio_readdir $pathname, $callback->($entries)
502
503Unlike the POSIX call of the same name, C<aio_readdir> reads an entire
504directory (i.e. opendir + readdir + closedir). The entries will not be
505sorted, and will B<NOT> include the C<.> and C<..> entries.
506
507The callback a single argument which is either C<undef> or an array-ref
508with the filenames.
509
510
511=item aio_load $path, $data, $callback->($status)
512
513This is a composite request that tries to fully load the given file into
514memory. Status is the same as with aio_read.
515
516=cut
517
518sub aio_load($$;$) {
519 aio_block {
520 my ($path, undef, $cb) = @_;
521 my $data = \$_[1];
522
523 my $pri = aioreq_pri;
524 my $grp = aio_group $cb;
525
526 aioreq_pri $pri;
527 add $grp aio_open $path, O_RDONLY, 0, sub {
528 my $fh = shift
529 or return $grp->result (-1);
530
531 aioreq_pri $pri;
532 add $grp aio_read $fh, 0, (-s $fh), $$data, 0, sub {
533 $grp->result ($_[0]);
534 };
535 };
536
537 $grp
538 }
539}
540
541=item aio_copy $srcpath, $dstpath, $callback->($status)
542
543Try to copy the I<file> (directories not supported as either source or
544destination) from C<$srcpath> to C<$dstpath> and call the callback with
545the C<0> (error) or C<-1> ok.
546
547This is a composite request that it creates the destination file with
548mode 0200 and copies the contents of the source file into it using
549C<aio_sendfile>, followed by restoring atime, mtime, access mode and
550uid/gid, in that order.
551
552If an error occurs, the partial destination file will be unlinked, if
553possible, except when setting atime, mtime, access mode and uid/gid, where
554errors are being ignored.
555
556=cut
557
558sub aio_copy($$;$) {
559 aio_block {
560 my ($src, $dst, $cb) = @_;
561
562 my $pri = aioreq_pri;
563 my $grp = aio_group $cb;
564
565 aioreq_pri $pri;
566 add $grp aio_open $src, O_RDONLY, 0, sub {
567 if (my $src_fh = $_[0]) {
568 my @stat = stat $src_fh;
569
570 aioreq_pri $pri;
571 add $grp aio_open $dst, O_CREAT | O_WRONLY | O_TRUNC, 0200, sub {
572 if (my $dst_fh = $_[0]) {
573 aioreq_pri $pri;
574 add $grp aio_sendfile $dst_fh, $src_fh, 0, $stat[7], sub {
575 if ($_[0] == $stat[7]) {
576 $grp->result (0);
577 close $src_fh;
578
579 # those should not normally block. should. should.
580 utime $stat[8], $stat[9], $dst;
581 chmod $stat[2] & 07777, $dst_fh;
582 chown $stat[4], $stat[5], $dst_fh;
583 close $dst_fh;
584 } else {
585 $grp->result (-1);
586 close $src_fh;
587 close $dst_fh;
588
589 aioreq $pri;
590 add $grp aio_unlink $dst;
591 }
592 };
593 } else {
594 $grp->result (-1);
595 }
596 },
597
598 } else {
599 $grp->result (-1);
600 }
601 };
602
603 $grp
604 }
605}
606
607=item aio_move $srcpath, $dstpath, $callback->($status)
608
609Try to move the I<file> (directories not supported as either source or
610destination) from C<$srcpath> to C<$dstpath> and call the callback with
611the C<0> (error) or C<-1> ok.
612
613This is a composite request that tries to rename(2) the file first. If
614rename files with C<EXDEV>, it copies the file with C<aio_copy> and, if
615that is successful, unlinking the C<$srcpath>.
616
617=cut
618
619sub aio_move($$;$) {
620 aio_block {
621 my ($src, $dst, $cb) = @_;
622
623 my $pri = aioreq_pri;
624 my $grp = aio_group $cb;
625
626 aioreq_pri $pri;
627 add $grp aio_rename $src, $dst, sub {
628 if ($_[0] && $! == EXDEV) {
629 aioreq_pri $pri;
630 add $grp aio_copy $src, $dst, sub {
631 $grp->result ($_[0]);
632
633 if (!$_[0]) {
634 aioreq_pri $pri;
635 add $grp aio_unlink $src;
636 }
637 };
638 } else {
639 $grp->result ($_[0]);
640 }
641 };
642
643 $grp
644 }
645}
646
647=item aio_scandir $path, $maxreq, $callback->($dirs, $nondirs)
648
649Scans a directory (similar to C<aio_readdir>) but additionally tries to
650efficiently separate the entries of directory C<$path> into two sets of
651names, directories you can recurse into (directories), and ones you cannot
652recurse into (everything else, including symlinks to directories).
653
654C<aio_scandir> is a composite request that creates of many sub requests_
655C<$maxreq> specifies the maximum number of outstanding aio requests that
656this function generates. If it is C<< <= 0 >>, then a suitable default
657will be chosen (currently 4).
658
659On error, the callback is called without arguments, otherwise it receives
660two array-refs with path-relative entry names.
661
662Example:
663
664 aio_scandir $dir, 0, sub {
665 my ($dirs, $nondirs) = @_;
666 print "real directories: @$dirs\n";
667 print "everything else: @$nondirs\n";
668 };
669
670Implementation notes.
671
672The C<aio_readdir> cannot be avoided, but C<stat()>'ing every entry can.
673
674After reading the directory, the modification time, size etc. of the
675directory before and after the readdir is checked, and if they match (and
676isn't the current time), the link count will be used to decide how many
677entries are directories (if >= 2). Otherwise, no knowledge of the number
678of subdirectories will be assumed.
679
680Then entries will be sorted into likely directories (everything without
681a non-initial dot currently) and likely non-directories (everything
682else). Then every entry plus an appended C</.> will be C<stat>'ed,
683likely directories first. If that succeeds, it assumes that the entry
684is a directory or a symlink to directory (which will be checked
685seperately). This is often faster than stat'ing the entry itself because
686filesystems might detect the type of the entry without reading the inode
687data (e.g. ext2fs filetype feature).
688
689If the known number of directories (link count - 2) has been reached, the
690rest of the entries is assumed to be non-directories.
691
692This only works with certainty on POSIX (= UNIX) filesystems, which
693fortunately are the vast majority of filesystems around.
694
695It will also likely work on non-POSIX filesystems with reduced efficiency
696as those tend to return 0 or 1 as link counts, which disables the
697directory counting heuristic.
698
699=cut
700
701sub aio_scandir($$;$) {
702 aio_block {
703 my ($path, $maxreq, $cb) = @_;
704
705 my $pri = aioreq_pri;
706
707 my $grp = aio_group $cb;
708
709 $maxreq = 4 if $maxreq <= 0;
710
711 # stat once
712 aioreq_pri $pri;
713 add $grp aio_stat $path, sub {
714 return $grp->result () if $_[0];
715 my $now = time;
716 my $hash1 = join ":", (stat _)[0,1,3,7,9];
717
718 # read the directory entries
719 aioreq_pri $pri;
720 add $grp aio_readdir $path, sub {
721 my $entries = shift
722 or return $grp->result ();
723
724 # stat the dir another time
725 aioreq_pri $pri;
726 add $grp aio_stat $path, sub {
727 my $hash2 = join ":", (stat _)[0,1,3,7,9];
728
729 my $ndirs;
730
731 # take the slow route if anything looks fishy
732 if ($hash1 ne $hash2 or (stat _)[9] == $now) {
733 $ndirs = -1;
734 } else {
735 # if nlink == 2, we are finished
736 # on non-posix-fs's, we rely on nlink < 2
737 $ndirs = (stat _)[3] - 2
738 or return $grp->result ([], $entries);
739 }
740
741 # sort into likely dirs and likely nondirs
742 # dirs == files without ".", short entries first
743 $entries = [map $_->[0],
744 sort { $b->[1] cmp $a->[1] }
745 map [$_, sprintf "%s%04d", (/.\./ ? "1" : "0"), length],
746 @$entries];
747
748 my (@dirs, @nondirs);
749
750 my $statgrp = add $grp aio_group sub {
751 $grp->result (\@dirs, \@nondirs);
752 };
753
754 limit $statgrp $maxreq;
755 feed $statgrp sub {
756 return unless @$entries;
757 my $entry = pop @$entries;
758
759 aioreq_pri $pri;
760 add $statgrp aio_stat "$path/$entry/.", sub {
761 if ($_[0] < 0) {
762 push @nondirs, $entry;
763 } else {
764 # need to check for real directory
765 aioreq_pri $pri;
766 add $statgrp aio_lstat "$path/$entry", sub {
767 if (-d _) {
768 push @dirs, $entry;
769
770 unless (--$ndirs) {
771 push @nondirs, @$entries;
772 feed $statgrp;
773 }
774 } else {
775 push @nondirs, $entry;
776 }
777 }
778 }
779 };
780 };
781 };
782 };
783 };
784
785 $grp
786 }
787}
788
789=item aio_rmtree $path, $callback->($status)
790
791Delete a directory tree starting (and including) C<$path>, return the
792status of the final C<rmdir> only. This is a composite request that
793uses C<aio_scandir> to recurse into and rmdir directories, and unlink
794everything else.
795
796=cut
797
798sub aio_rmtree;
799sub aio_rmtree($;$) {
800 aio_block {
801 my ($path, $cb) = @_;
802
803 my $pri = aioreq_pri;
804 my $grp = aio_group $cb;
805
806 aioreq_pri $pri;
807 add $grp aio_scandir $path, 0, sub {
808 my ($dirs, $nondirs) = @_;
809
810 my $dirgrp = aio_group sub {
811 add $grp aio_rmdir $path, sub {
812 $grp->result ($_[0]);
813 };
814 };
815
816 (aioreq_pri $pri), add $dirgrp aio_rmtree "$path/$_" for @$dirs;
817 (aioreq_pri $pri), add $dirgrp aio_unlink "$path/$_" for @$nondirs;
818
819 add $grp $dirgrp;
820 };
821
822 $grp
823 }
824}
825
192=item aio_fsync $fh, $callback 826=item aio_fsync $fh, $callback->($status)
193 827
194Asynchronously call fsync on the given filehandle and call the callback 828Asynchronously call fsync on the given filehandle and call the callback
195with the fsync result code. 829with the fsync result code.
196 830
197=item aio_fdatasync $fh, $callback 831=item aio_fdatasync $fh, $callback->($status)
198 832
199Asynchronously call fdatasync on the given filehandle and call the 833Asynchronously call fdatasync on the given filehandle and call the
200callback with the fdatasync result code. Might set C<$!> to C<ENOSYS> if 834callback with the fdatasync result code.
201C<fdatasync> is not available. 835
836If this call isn't available because your OS lacks it or it couldn't be
837detected, it will be emulated by calling C<fsync> instead.
838
839=item aio_group $callback->(...)
840
841This is a very special aio request: Instead of doing something, it is a
842container for other aio requests, which is useful if you want to bundle
843many requests into a single, composite, request with a definite callback
844and the ability to cancel the whole request with its subrequests.
845
846Returns an object of class L<IO::AIO::GRP>. See its documentation below
847for more info.
848
849Example:
850
851 my $grp = aio_group sub {
852 print "all stats done\n";
853 };
854
855 add $grp
856 (aio_stat ...),
857 (aio_stat ...),
858 ...;
859
860=item aio_nop $callback->()
861
862This is a special request - it does nothing in itself and is only used for
863side effects, such as when you want to add a dummy request to a group so
864that finishing the requests in the group depends on executing the given
865code.
866
867While this request does nothing, it still goes through the execution
868phase and still requires a worker thread. Thus, the callback will not
869be executed immediately but only after other requests in the queue have
870entered their execution phase. This can be used to measure request
871latency.
872
873=item IO::AIO::aio_busy $fractional_seconds, $callback->() *NOT EXPORTED*
874
875Mainly used for debugging and benchmarking, this aio request puts one of
876the request workers to sleep for the given time.
877
878While it is theoretically handy to have simple I/O scheduling requests
879like sleep and file handle readable/writable, the overhead this creates is
880immense (it blocks a thread for a long time) so do not use this function
881except to put your application under artificial I/O pressure.
202 882
203=back 883=back
204 884
885=head2 IO::AIO::REQ CLASS
886
887All non-aggregate C<aio_*> functions return an object of this class when
888called in non-void context.
889
890=over 4
891
892=item cancel $req
893
894Cancels the request, if possible. Has the effect of skipping execution
895when entering the B<execute> state and skipping calling the callback when
896entering the the B<result> state, but will leave the request otherwise
897untouched. That means that requests that currently execute will not be
898stopped and resources held by the request will not be freed prematurely.
899
900=item cb $req $callback->(...)
901
902Replace (or simply set) the callback registered to the request.
903
904=back
905
906=head2 IO::AIO::GRP CLASS
907
908This class is a subclass of L<IO::AIO::REQ>, so all its methods apply to
909objects of this class, too.
910
911A IO::AIO::GRP object is a special request that can contain multiple other
912aio requests.
913
914You create one by calling the C<aio_group> constructing function with a
915callback that will be called when all contained requests have entered the
916C<done> state:
917
918 my $grp = aio_group sub {
919 print "all requests are done\n";
920 };
921
922You add requests by calling the C<add> method with one or more
923C<IO::AIO::REQ> objects:
924
925 $grp->add (aio_unlink "...");
926
927 add $grp aio_stat "...", sub {
928 $_[0] or return $grp->result ("error");
929
930 # add another request dynamically, if first succeeded
931 add $grp aio_open "...", sub {
932 $grp->result ("ok");
933 };
934 };
935
936This makes it very easy to create composite requests (see the source of
937C<aio_move> for an application) that work and feel like simple requests.
938
939=over 4
940
941=item * The IO::AIO::GRP objects will be cleaned up during calls to
942C<IO::AIO::poll_cb>, just like any other request.
943
944=item * They can be canceled like any other request. Canceling will cancel not
945only the request itself, but also all requests it contains.
946
947=item * They can also can also be added to other IO::AIO::GRP objects.
948
949=item * You must not add requests to a group from within the group callback (or
950any later time).
951
952=back
953
954Their lifetime, simplified, looks like this: when they are empty, they
955will finish very quickly. If they contain only requests that are in the
956C<done> state, they will also finish. Otherwise they will continue to
957exist.
958
959That means after creating a group you have some time to add requests. And
960in the callbacks of those requests, you can add further requests to the
961group. And only when all those requests have finished will the the group
962itself finish.
963
964=over 4
965
966=item add $grp ...
967
968=item $grp->add (...)
969
970Add one or more requests to the group. Any type of L<IO::AIO::REQ> can
971be added, including other groups, as long as you do not create circular
972dependencies.
973
974Returns all its arguments.
975
976=item $grp->cancel_subs
977
978Cancel all subrequests and clears any feeder, but not the group request
979itself. Useful when you queued a lot of events but got a result early.
980
981=item $grp->result (...)
982
983Set the result value(s) that will be passed to the group callback when all
984subrequests have finished and set thre groups errno to the current value
985of errno (just like calling C<errno> without an error number). By default,
986no argument will be passed and errno is zero.
987
988=item $grp->errno ([$errno])
989
990Sets the group errno value to C<$errno>, or the current value of errno
991when the argument is missing.
992
993Every aio request has an associated errno value that is restored when
994the callback is invoked. This method lets you change this value from its
995default (0).
996
997Calling C<result> will also set errno, so make sure you either set C<$!>
998before the call to C<result>, or call c<errno> after it.
999
1000=item feed $grp $callback->($grp)
1001
1002Sets a feeder/generator on this group: every group can have an attached
1003generator that generates requests if idle. The idea behind this is that,
1004although you could just queue as many requests as you want in a group,
1005this might starve other requests for a potentially long time. For
1006example, C<aio_scandir> might generate hundreds of thousands C<aio_stat>
1007requests, delaying any later requests for a long time.
1008
1009To avoid this, and allow incremental generation of requests, you can
1010instead a group and set a feeder on it that generates those requests. The
1011feed callback will be called whenever there are few enough (see C<limit>,
1012below) requests active in the group itself and is expected to queue more
1013requests.
1014
1015The feed callback can queue as many requests as it likes (i.e. C<add> does
1016not impose any limits).
1017
1018If the feed does not queue more requests when called, it will be
1019automatically removed from the group.
1020
1021If the feed limit is C<0>, it will be set to C<2> automatically.
1022
1023Example:
1024
1025 # stat all files in @files, but only ever use four aio requests concurrently:
1026
1027 my $grp = aio_group sub { print "finished\n" };
1028 limit $grp 4;
1029 feed $grp sub {
1030 my $file = pop @files
1031 or return;
1032
1033 add $grp aio_stat $file, sub { ... };
1034 };
1035
1036=item limit $grp $num
1037
1038Sets the feeder limit for the group: The feeder will be called whenever
1039the group contains less than this many requests.
1040
1041Setting the limit to C<0> will pause the feeding process.
1042
1043=back
1044
205=head2 SUPPORT FUNCTIONS 1045=head2 SUPPORT FUNCTIONS
1046
1047=head3 EVENT PROCESSING AND EVENT LOOP INTEGRATION
206 1048
207=over 4 1049=over 4
208 1050
209=item $fileno = IO::AIO::poll_fileno 1051=item $fileno = IO::AIO::poll_fileno
210 1052
215 1057
216See C<poll_cb> for an example. 1058See C<poll_cb> for an example.
217 1059
218=item IO::AIO::poll_cb 1060=item IO::AIO::poll_cb
219 1061
220Process all outstanding events on the result pipe. You have to call this 1062Process some outstanding events on the result pipe. You have to call this
221regularly. Returns the number of events processed. Returns immediately 1063regularly. Returns the number of events processed. Returns immediately
222when no events are outstanding. 1064when no events are outstanding. The amount of events processed depends on
1065the settings of C<IO::AIO::max_poll_req> and C<IO::AIO::max_poll_time>.
1066
1067If not all requests were processed for whatever reason, the filehandle
1068will still be ready when C<poll_cb> returns.
223 1069
224Example: Install an Event watcher that automatically calls 1070Example: Install an Event watcher that automatically calls
225IO::AIO::poll_cb with high priority: 1071IO::AIO::poll_cb with high priority:
226 1072
227 Event->io (fd => IO::AIO::poll_fileno, 1073 Event->io (fd => IO::AIO::poll_fileno,
228 poll => 'r', async => 1, 1074 poll => 'r', async => 1,
229 cb => \&IO::AIO::poll_cb); 1075 cb => \&IO::AIO::poll_cb);
230 1076
1077=item IO::AIO::max_poll_reqs $nreqs
1078
1079=item IO::AIO::max_poll_time $seconds
1080
1081These set the maximum number of requests (default C<0>, meaning infinity)
1082that are being processed by C<IO::AIO::poll_cb> in one call, respectively
1083the maximum amount of time (default C<0>, meaning infinity) spent in
1084C<IO::AIO::poll_cb> to process requests (more correctly the mininum amount
1085of time C<poll_cb> is allowed to use).
1086
1087Setting C<max_poll_time> to a non-zero value creates an overhead of one
1088syscall per request processed, which is not normally a problem unless your
1089callbacks are really really fast or your OS is really really slow (I am
1090not mentioning Solaris here). Using C<max_poll_reqs> incurs no overhead.
1091
1092Setting these is useful if you want to ensure some level of
1093interactiveness when perl is not fast enough to process all requests in
1094time.
1095
1096For interactive programs, values such as C<0.01> to C<0.1> should be fine.
1097
1098Example: Install an Event watcher that automatically calls
1099IO::AIO::poll_cb with low priority, to ensure that other parts of the
1100program get the CPU sometimes even under high AIO load.
1101
1102 # try not to spend much more than 0.1s in poll_cb
1103 IO::AIO::max_poll_time 0.1;
1104
1105 # use a low priority so other tasks have priority
1106 Event->io (fd => IO::AIO::poll_fileno,
1107 poll => 'r', nice => 1,
1108 cb => &IO::AIO::poll_cb);
1109
231=item IO::AIO::poll_wait 1110=item IO::AIO::poll_wait
232 1111
1112If there are any outstanding requests and none of them in the result
233Wait till the result filehandle becomes ready for reading (simply does a 1113phase, wait till the result filehandle becomes ready for reading (simply
234C<select> on the filehandle. This is useful if you want to synchronously wait 1114does a C<select> on the filehandle. This is useful if you want to
235for some requests to finish). 1115synchronously wait for some requests to finish).
236 1116
237See C<nreqs> for an example. 1117See C<nreqs> for an example.
238 1118
1119=item IO::AIO::poll
1120
1121Waits until some requests have been handled.
1122
1123Returns the number of requests processed, but is otherwise strictly
1124equivalent to:
1125
1126 IO::AIO::poll_wait, IO::AIO::poll_cb
1127
239=item IO::AIO::nreqs 1128=item IO::AIO::flush
240 1129
241Returns the number of requests currently outstanding (i.e. for which their 1130Wait till all outstanding AIO requests have been handled.
242callback has not been invoked yet).
243 1131
244Example: wait till there are no outstanding requests anymore: 1132Strictly equivalent to:
245 1133
246 IO::AIO::poll_wait, IO::AIO::poll_cb 1134 IO::AIO::poll_wait, IO::AIO::poll_cb
247 while IO::AIO::nreqs; 1135 while IO::AIO::nreqs;
248 1136
1137=back
1138
1139=head3 CONTROLLING THE NUMBER OF THREADS
1140
1141=over
1142
1143=item IO::AIO::min_parallel $nthreads
1144
1145Set the minimum number of AIO threads to C<$nthreads>. The current
1146default is C<8>, which means eight asynchronous operations can execute
1147concurrently at any one time (the number of outstanding requests,
1148however, is unlimited).
1149
1150IO::AIO starts threads only on demand, when an AIO request is queued and
1151no free thread exists. Please note that queueing up a hundred requests can
1152create demand for a hundred threads, even if it turns out that everything
1153is in the cache and could have been processed faster by a single thread.
1154
1155It is recommended to keep the number of threads relatively low, as some
1156Linux kernel versions will scale negatively with the number of threads
1157(higher parallelity => MUCH higher latency). With current Linux 2.6
1158versions, 4-32 threads should be fine.
1159
1160Under most circumstances you don't need to call this function, as the
1161module selects a default that is suitable for low to moderate load.
1162
1163=item IO::AIO::max_parallel $nthreads
1164
1165Sets the maximum number of AIO threads to C<$nthreads>. If more than the
1166specified number of threads are currently running, this function kills
1167them. This function blocks until the limit is reached.
1168
1169While C<$nthreads> are zero, aio requests get queued but not executed
1170until the number of threads has been increased again.
1171
1172This module automatically runs C<max_parallel 0> at program end, to ensure
1173that all threads are killed and that there are no outstanding requests.
1174
1175Under normal circumstances you don't need to call this function.
1176
1177=item IO::AIO::max_idle $nthreads
1178
1179Limit the number of threads (default: 4) that are allowed to idle (i.e.,
1180threads that did not get a request to process within 10 seconds). That
1181means if a thread becomes idle while C<$nthreads> other threads are also
1182idle, it will free its resources and exit.
1183
1184This is useful when you allow a large number of threads (e.g. 100 or 1000)
1185to allow for extremely high load situations, but want to free resources
1186under normal circumstances (1000 threads can easily consume 30MB of RAM).
1187
1188The default is probably ok in most situations, especially if thread
1189creation is fast. If thread creation is very slow on your system you might
1190want to use larger values.
1191
1192=item $oldmaxreqs = IO::AIO::max_outstanding $maxreqs
1193
1194This is a very bad function to use in interactive programs because it
1195blocks, and a bad way to reduce concurrency because it is inexact: Better
1196use an C<aio_group> together with a feed callback.
1197
1198Sets the maximum number of outstanding requests to C<$nreqs>. If you
1199to queue up more than this number of requests, the next call to the
1200C<poll_cb> (and C<poll_some> and other functions calling C<poll_cb>)
1201function will block until the limit is no longer exceeded.
1202
1203The default value is very large, so there is no practical limit on the
1204number of outstanding requests.
1205
1206You can still queue as many requests as you want. Therefore,
1207C<max_oustsanding> is mainly useful in simple scripts (with low values) or
1208as a stop gap to shield against fatal memory overflow (with large values).
1209
1210=back
1211
1212=head3 STATISTICAL INFORMATION
1213
1214=over
1215
249=item IO::AIO::flush 1216=item IO::AIO::nreqs
250 1217
251Wait till all outstanding AIO requests have been handled. 1218Returns the number of requests currently in the ready, execute or pending
1219states (i.e. for which their callback has not been invoked yet).
252 1220
253Strictly equivalent to: 1221Example: wait till there are no outstanding requests anymore:
254 1222
255 IO::AIO::poll_wait, IO::AIO::poll_cb 1223 IO::AIO::poll_wait, IO::AIO::poll_cb
256 while IO::AIO::nreqs; 1224 while IO::AIO::nreqs;
257 1225
1226=item IO::AIO::nready
1227
1228Returns the number of requests currently in the ready state (not yet
1229executed).
1230
258=item IO::AIO::poll 1231=item IO::AIO::npending
259 1232
260Waits until some requests have been handled. 1233Returns the number of requests currently in the pending state (executed,
261 1234but not yet processed by poll_cb).
262Strictly equivalent to:
263
264 IO::AIO::poll_wait, IO::AIO::poll_cb
265 if IO::AIO::nreqs;
266
267=item IO::AIO::min_parallel $nthreads
268
269Set the minimum number of AIO threads to C<$nthreads>. The default is
270C<1>, which means a single asynchronous operation can be done at one time
271(the number of outstanding operations, however, is unlimited).
272
273It is recommended to keep the number of threads low, as some Linux
274kernel versions will scale negatively with the number of threads (higher
275parallelity => MUCH higher latency). With current Linux 2.6 versions, 4-32
276threads should be fine.
277
278Under normal circumstances you don't need to call this function, as this
279module automatically starts some threads (the exact number might change,
280and is currently 4).
281
282=item IO::AIO::max_parallel $nthreads
283
284Sets the maximum number of AIO threads to C<$nthreads>. If more than
285the specified number of threads are currently running, kill them. This
286function blocks until the limit is reached.
287
288This module automatically runs C<max_parallel 0> at program end, to ensure
289that all threads are killed and that there are no outstanding requests.
290
291Under normal circumstances you don't need to call this function.
292
293=item $oldnreqs = IO::AIO::max_outstanding $nreqs
294
295Sets the maximum number of outstanding requests to C<$nreqs>. If you
296try to queue up more than this number of requests, the caller will block until
297some requests have been handled.
298
299The default is very large, so normally there is no practical limit. If you
300queue up many requests in a loop it it often improves speed if you set
301this to a relatively low number, such as C<100>.
302
303Under normal circumstances you don't need to call this function.
304 1235
305=back 1236=back
306 1237
307=cut 1238=cut
308 1239
309# support function to convert a fd into a perl filehandle 1240# support function to convert a fd into a perl filehandle
310sub _fd2fh { 1241sub _fd2fh {
311 return undef if $_[0] < 0; 1242 return undef if $_[0] < 0;
312 1243
313 # try to be perl5.6-compatible 1244 # try to generate nice filehandles
314 local *AIO_FH; 1245 my $sym = "IO::AIO::fd#$_[0]";
315 open AIO_FH, "+<&=$_[0]" 1246 local *$sym;
1247
1248 open *$sym, "+<&=$_[0]" # usually works under any unix
1249 or open *$sym, "<&=$_[0]" # cygwin needs this
1250 or open *$sym, ">&=$_[0]" # or this
316 or return undef; 1251 or return undef;
317 1252
318 *AIO_FH 1253 *$sym
319} 1254}
320 1255
321min_parallel 4; 1256min_parallel 8;
322 1257
323END { 1258END { flush }
324 max_parallel 0;
325}
326 1259
3271; 12601;
328 1261
1262=head2 FORK BEHAVIOUR
1263
1264This module should do "the right thing" when the process using it forks:
1265
1266Before the fork, IO::AIO enters a quiescent state where no requests
1267can be added in other threads and no results will be processed. After
1268the fork the parent simply leaves the quiescent state and continues
1269request/result processing, while the child frees the request/result queue
1270(so that the requests started before the fork will only be handled in the
1271parent). Threads will be started on demand until the limit set in the
1272parent process has been reached again.
1273
1274In short: the parent will, after a short pause, continue as if fork had
1275not been called, while the child will act as if IO::AIO has not been used
1276yet.
1277
1278=head2 MEMORY USAGE
1279
1280Per-request usage:
1281
1282Each aio request uses - depending on your architecture - around 100-200
1283bytes of memory. In addition, stat requests need a stat buffer (possibly
1284a few hundred bytes), readdir requires a result buffer and so on. Perl
1285scalars and other data passed into aio requests will also be locked and
1286will consume memory till the request has entered the done state.
1287
1288This is now awfully much, so queuing lots of requests is not usually a
1289problem.
1290
1291Per-thread usage:
1292
1293In the execution phase, some aio requests require more memory for
1294temporary buffers, and each thread requires a stack and other data
1295structures (usually around 16k-128k, depending on the OS).
1296
1297=head1 KNOWN BUGS
1298
1299Known bugs will be fixed in the next release.
1300
329=head1 SEE ALSO 1301=head1 SEE ALSO
330 1302
331L<Coro>, L<Linux::AIO>. 1303L<Coro::AIO>.
332 1304
333=head1 AUTHOR 1305=head1 AUTHOR
334 1306
335 Marc Lehmann <schmorp@schmorp.de> 1307 Marc Lehmann <schmorp@schmorp.de>
336 http://home.schmorp.de/ 1308 http://home.schmorp.de/

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines