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 1.4 by root, Sun Jul 10 20:57:00 2005 UTC vs.
Revision 1.34 by root, Thu Aug 18 16:32:10 2005 UTC

3IO::AIO - Asynchronous Input/Output 3IO::AIO - Asynchronous Input/Output
4 4
5=head1 SYNOPSIS 5=head1 SYNOPSIS
6 6
7 use IO::AIO; 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: $!";
18 };
19
20 # Event
21 Event->io (fd => IO::AIO::poll_fileno,
22 poll => 'r',
23 cb => \&IO::AIO::poll_cb);
24
25 # Glib/Gtk2
26 add_watch Glib::IO IO::AIO::poll_fileno,
27 in => sub { IO::AIO::poll_cb; 1 };
28
29 # Tk
30 Tk::Event::IO->fileevent (IO::AIO::poll_fileno, "",
31 readable => \&IO::AIO::poll_cb);
32
33 # Danga::Socket
34 Danga::Socket->AddOtherFds (IO::AIO::poll_fileno =>
35 \&IO::AIO::poll_cb);
36
8 37
9=head1 DESCRIPTION 38=head1 DESCRIPTION
10 39
11This module implements asynchronous I/O using whatever means your 40This module implements asynchronous I/O using whatever means your
12operating system supports. 41operating system supports.
19not well-supported (Linux doesn't allow them on normal files currently, 48not well-supported (Linux doesn't allow them on normal files currently,
20for example), and they would only support aio_read and aio_write, so the 49for example), and they would only support aio_read and aio_write, so the
21remaining functionality would have to be implemented using threads anyway. 50remaining functionality would have to be implemented using threads anyway.
22 51
23Although the module will work with in the presence of other threads, it is 52Although the module will work with in the presence of other threads, it is
24currently not reentrant, so use appropriate locking yourself. 53currently not reentrant, so use appropriate locking yourself, always call
25 54C<poll_cb> from within the same thread, or never call C<poll_cb> (or other
26=head2 API NOTES 55C<aio_> functions) recursively.
27
28All the C<aio_*> calls are more or less thin wrappers around the syscall
29with the same name (sans C<aio_>). The arguments are similar or identical,
30and they all accept an additional C<$callback> argument which must be
31a code reference. This code reference will get called with the syscall
32return code (e.g. most syscalls return C<-1> on error, unlike perl, which
33usually delivers "false") as it's sole argument when the given syscall has
34been executed asynchronously.
35
36All functions that expect a filehandle will also accept a file descriptor.
37
38The filenames you pass to these routines I<must> be absolute. The reason
39is that at the time the request is being executed, the current working
40directory could have changed. Alternatively, you can make sure that you
41never change the current working directory.
42
43=over 4
44 56
45=cut 57=cut
46 58
47package IO::AIO; 59package IO::AIO;
48 60
61no warnings;
62
49use base 'Exporter'; 63use base 'Exporter';
50 64
51use Fcntl (); 65use Fcntl ();
52 66
53BEGIN { 67BEGIN {
54 $VERSION = 0.2; 68 $VERSION = 1.5;
55 69
56 @EXPORT = qw(aio_read aio_write aio_open aio_close aio_stat aio_lstat aio_unlink 70 @EXPORT = qw(aio_read aio_write aio_open aio_close aio_stat aio_lstat aio_unlink
57 aio_fsync aio_fdatasync aio_readahead); 71 aio_rmdir aio_symlink aio_fsync aio_fdatasync aio_readahead);
58 @EXPORT_OK = qw(poll_fileno poll_cb min_parallel max_parallel max_outstanding nreqs); 72 @EXPORT_OK = qw(poll_fileno poll_cb min_parallel max_parallel max_outstanding nreqs);
59 73
60 require XSLoader; 74 require XSLoader;
61 XSLoader::load IO::AIO, $VERSION; 75 XSLoader::load IO::AIO, $VERSION;
62} 76}
63 77
64=item IO::AIO::min_parallel $nthreads 78=head1 FUNCTIONS
65 79
66Set the minimum number of AIO threads to C<$nthreads>. The default is 80=head2 AIO FUNCTIONS
67C<1>, which means a single asynchronous operation can be done at one time
68(the number of outstanding operations, however, is unlimited).
69 81
70It is recommended to keep the number of threads low, as some Linux 82All the C<aio_*> calls are more or less thin wrappers around the syscall
71kernel versions will scale negatively with the number of threads (higher 83with the same name (sans C<aio_>). The arguments are similar or identical,
72parallelity => MUCH higher latency). With current Linux 2.6 versions, 4-32 84and they all accept an additional (and optional) C<$callback> argument
73threads should be fine. 85which must be a code reference. This code reference will get called with
86the syscall return code (e.g. most syscalls return C<-1> on error, unlike
87perl, which usually delivers "false") as it's sole argument when the given
88syscall has been executed asynchronously.
74 89
75Under normal circumstances you don't need to call this function, as this 90All functions expecting a filehandle keep a copy of the filehandle
76module automatically starts some threads (the exact number might change, 91internally until the request has finished.
77and is currently 4).
78 92
79=item IO::AIO::max_parallel $nthreads 93The pathnames you pass to these routines I<must> be absolute and
94encoded in byte form. The reason for the former is that at the time the
95request is being executed, the current working directory could have
96changed. Alternatively, you can make sure that you never change the
97current working directory.
80 98
81Sets the maximum number of AIO threads to C<$nthreads>. If more than 99To encode pathnames to byte form, either make sure you either: a)
82the specified number of threads are currently running, kill them. This 100always pass in filenames you got from outside (command line, readdir
83function blocks until the limit is reached. 101etc.), b) are ASCII or ISO 8859-1, c) use the Encode module and encode
102your pathnames to the locale (or other) encoding in effect in the user
103environment, d) use Glib::filename_from_unicode on unicode filenames or e)
104use something else.
84 105
85This module automatically runs C<max_parallel 0> at program end, to ensure 106=over 4
86that all threads are killed and that there are no outstanding requests.
87
88Under normal circumstances you don't need to call this function.
89
90=item $oldnreqs = IO::AIO::max_outstanding $nreqs
91
92Sets the maximum number of outstanding requests to C<$nreqs>. If you
93try to queue up more than this number of requests, the caller will block until
94some requests have been handled.
95
96The default is very large, so normally there is no practical limit. If you
97queue up many requests in a loop it it often improves speed if you set
98this to a relatively low number, such as C<100>.
99
100Under normal circumstances you don't need to call this function.
101
102=item $fileno = IO::AIO::poll_fileno
103
104Return the I<request result pipe filehandle>. This filehandle must be
105polled for reading by some mechanism outside this module (e.g. Event
106or select, see below). If the pipe becomes readable you have to call
107C<poll_cb> to check the results.
108
109See C<poll_cb> for an example.
110
111=item IO::AIO::poll_cb
112
113Process all outstanding events on the result pipe. You have to call this
114regularly. Returns the number of events processed. Returns immediately
115when no events are outstanding.
116
117You can use Event to multiplex, e.g.:
118
119 Event->io (fd => IO::AIO::poll_fileno,
120 poll => 'r', async => 1,
121 cb => \&IO::AIO::poll_cb);
122
123=item IO::AIO::poll_wait
124
125Wait till the result filehandle becomes ready for reading (simply does a
126select on the filehandle. This is useful if you want to synchronously wait
127for some requests to finish).
128
129See C<nreqs> for an example.
130
131=item IO::AIO::nreqs
132
133Returns the number of requests currently outstanding.
134
135Example: wait till there are no outstanding requests anymore:
136
137 IO::AIO::poll_wait, IO::AIO::poll_cb
138 while IO::AIO::nreqs;
139 107
140=item aio_open $pathname, $flags, $mode, $callback 108=item aio_open $pathname, $flags, $mode, $callback
141 109
142Asynchronously open or create a file and call the callback with a newly 110Asynchronously open or create a file and call the callback with a newly
143created filehandle for the file. 111created filehandle for the file.
144 112
145The pathname passed to C<aio_open> must be absolute. See API NOTES, above, 113The pathname passed to C<aio_open> must be absolute. See API NOTES, above,
146for an explanation. 114for an explanation.
147 115
148The C<$mode> argument is a bitmask. See the C<Fcntl> module for a 116The C<$flags> argument is a bitmask. See the C<Fcntl> module for a
149list. They are the same as used in C<sysopen>. 117list. They are the same as used by C<sysopen>.
118
119Likewise, C<$mode> specifies the mode of the newly created file, if it
120didn't exist and C<O_CREAT> has been given, just like perl's C<sysopen>,
121except that it is mandatory (i.e. use C<0> if you don't create new files,
122and C<0666> or C<0777> if you do).
150 123
151Example: 124Example:
152 125
153 aio_open "/etc/passwd", O_RDONLY, 0, sub { 126 aio_open "/etc/passwd", O_RDONLY, 0, sub {
154 if ($_[0]) { 127 if ($_[0]) {
161 134
162=item aio_close $fh, $callback 135=item aio_close $fh, $callback
163 136
164Asynchronously close a file and call the callback with the result 137Asynchronously close a file and call the callback with the result
165code. I<WARNING:> although accepted, you should not pass in a perl 138code. I<WARNING:> although accepted, you should not pass in a perl
166filehandle here, as perl will likely close the file descriptor itself when 139filehandle here, as perl will likely close the file descriptor another
167the filehandle is destroyed. Normally, you can safely call perls C<close> 140time when the filehandle is destroyed. Normally, you can safely call perls
168or just let filehandles go out of scope. 141C<close> or just let filehandles go out of scope.
142
143This is supposed to be a bug in the API, so that might change. It's
144therefore best to avoid this function.
169 145
170=item aio_read $fh,$offset,$length, $data,$dataoffset,$callback 146=item aio_read $fh,$offset,$length, $data,$dataoffset,$callback
171 147
172=item aio_write $fh,$offset,$length, $data,$dataoffset,$callback 148=item aio_write $fh,$offset,$length, $data,$dataoffset,$callback
173 149
174Reads or writes C<length> bytes from the specified C<fh> and C<offset> 150Reads or writes C<length> bytes from the specified C<fh> and C<offset>
175into the scalar given by C<data> and offset C<dataoffset> and calls the 151into the scalar given by C<data> and offset C<dataoffset> and calls the
176callback without the actual number of bytes read (or -1 on error, just 152callback without the actual number of bytes read (or -1 on error, just
177like the syscall). 153like the syscall).
178 154
155The C<$data> scalar I<MUST NOT> be modified in any way while the request
156is outstanding. Modifying it can result in segfaults or WW3 (if the
157necessary/optional hardware is installed).
158
179Example: Read 15 bytes at offset 7 into scalar C<$buffer>, strating at 159Example: Read 15 bytes at offset 7 into scalar C<$buffer>, starting at
180offset C<0> within the scalar: 160offset C<0> within the scalar:
181 161
182 aio_read $fh, 7, 15, $buffer, 0, sub { 162 aio_read $fh, 7, 15, $buffer, 0, sub {
183 $_[0] >= 0 or die "read error: $!"; 163 $_[0] > 0 or die "read error: $!";
184 print "read <$buffer>\n"; 164 print "read $_[0] bytes: <$buffer>\n";
185 }; 165 };
186 166
187=item aio_readahead $fh,$offset,$length, $callback 167=item aio_readahead $fh,$offset,$length, $callback
188 168
189Asynchronously reads the specified byte range into the page cache, using
190the C<readahead> syscall. If that syscall doesn't exist the status will be
191C<-1> and C<$!> is set to ENOSYS.
192
193readahead() populates the page cache with data from a file so that 169C<aio_readahead> populates the page cache with data from a file so that
194subsequent reads from that file will not block on disk I/O. The C<$offset> 170subsequent reads from that file will not block on disk I/O. The C<$offset>
195argument specifies the starting point from which data is to be read and 171argument specifies the starting point from which data is to be read and
196C<$length> specifies the number of bytes to be read. I/O is performed in 172C<$length> specifies the number of bytes to be read. I/O is performed in
197whole pages, so that offset is effectively rounded down to a page boundary 173whole pages, so that offset is effectively rounded down to a page boundary
198and bytes are read up to the next page boundary greater than or equal to 174and bytes are read up to the next page boundary greater than or equal to
199(off-set+length). aio_readahead() does not read beyond the end of the 175(off-set+length). C<aio_readahead> does not read beyond the end of the
200file. The current file offset of the file is left unchanged. 176file. The current file offset of the file is left unchanged.
177
178If that syscall doesn't exist (likely if your OS isn't Linux) it will be
179emulated by simply reading the data, which would have a similar effect.
201 180
202=item aio_stat $fh_or_path, $callback 181=item aio_stat $fh_or_path, $callback
203 182
204=item aio_lstat $fh, $callback 183=item aio_lstat $fh, $callback
205 184
224=item aio_unlink $pathname, $callback 203=item aio_unlink $pathname, $callback
225 204
226Asynchronously unlink (delete) a file and call the callback with the 205Asynchronously unlink (delete) a file and call the callback with the
227result code. 206result code.
228 207
208=item aio_rmdir $pathname, $callback
209
210Asynchronously rmdir (delete) a directory and call the callback with the
211result code.
212
229=item aio_fsync $fh, $callback 213=item aio_fsync $fh, $callback
230 214
231Asynchronously call fsync on the given filehandle and call the callback 215Asynchronously call fsync on the given filehandle and call the callback
232with the fsync result code. 216with the fsync result code.
233 217
234=item aio_fdatasync $fh, $callback 218=item aio_fdatasync $fh, $callback
235 219
236Asynchronously call fdatasync on the given filehandle and call the 220Asynchronously call fdatasync on the given filehandle and call the
237callback with the fdatasync result code. 221callback with the fdatasync result code.
222
223If this call isn't available because your OS lacks it or it couldn't be
224detected, it will be emulated by calling C<fsync> instead.
225
226=back
227
228=head2 SUPPORT FUNCTIONS
229
230=over 4
231
232=item $fileno = IO::AIO::poll_fileno
233
234Return the I<request result pipe file descriptor>. This filehandle must be
235polled for reading by some mechanism outside this module (e.g. Event or
236select, see below or the SYNOPSIS). If the pipe becomes readable you have
237to call C<poll_cb> to check the results.
238
239See C<poll_cb> for an example.
240
241=item IO::AIO::poll_cb
242
243Process all outstanding events on the result pipe. You have to call this
244regularly. Returns the number of events processed. Returns immediately
245when no events are outstanding.
246
247Example: Install an Event watcher that automatically calls
248IO::AIO::poll_cb with high priority:
249
250 Event->io (fd => IO::AIO::poll_fileno,
251 poll => 'r', async => 1,
252 cb => \&IO::AIO::poll_cb);
253
254=item IO::AIO::poll_wait
255
256Wait till the result filehandle becomes ready for reading (simply does a
257C<select> on the filehandle. This is useful if you want to synchronously wait
258for some requests to finish).
259
260See C<nreqs> for an example.
261
262=item IO::AIO::nreqs
263
264Returns the number of requests currently outstanding (i.e. for which their
265callback has not been invoked yet).
266
267Example: wait till there are no outstanding requests anymore:
268
269 IO::AIO::poll_wait, IO::AIO::poll_cb
270 while IO::AIO::nreqs;
271
272=item IO::AIO::flush
273
274Wait till all outstanding AIO requests have been handled.
275
276Strictly equivalent to:
277
278 IO::AIO::poll_wait, IO::AIO::poll_cb
279 while IO::AIO::nreqs;
280
281=item IO::AIO::poll
282
283Waits until some requests have been handled.
284
285Strictly equivalent to:
286
287 IO::AIO::poll_wait, IO::AIO::poll_cb
288 if IO::AIO::nreqs;
289
290=item IO::AIO::min_parallel $nthreads
291
292Set the minimum number of AIO threads to C<$nthreads>. The current default
293is C<4>, which means four asynchronous operations can be done at one time
294(the number of outstanding operations, however, is unlimited).
295
296IO::AIO starts threads only on demand, when an AIO request is queued and
297no free thread exists.
298
299It is recommended to keep the number of threads low, as some Linux
300kernel versions will scale negatively with the number of threads (higher
301parallelity => MUCH higher latency). With current Linux 2.6 versions, 4-32
302threads should be fine.
303
304Under most circumstances you don't need to call this function, as the
305module selects a default that is suitable for low to moderate load.
306
307=item IO::AIO::max_parallel $nthreads
308
309Sets the maximum number of AIO threads to C<$nthreads>. If more than the
310specified number of threads are currently running, this function kills
311them. This function blocks until the limit is reached.
312
313While C<$nthreads> are zero, aio requests get queued but not executed
314until the number of threads has been increased again.
315
316This module automatically runs C<max_parallel 0> at program end, to ensure
317that all threads are killed and that there are no outstanding requests.
318
319Under normal circumstances you don't need to call this function.
320
321=item $oldnreqs = IO::AIO::max_outstanding $nreqs
322
323Sets the maximum number of outstanding requests to C<$nreqs>. If you
324try to queue up more than this number of requests, the caller will block until
325some requests have been handled.
326
327The default is very large, so normally there is no practical limit. If you
328queue up many requests in a loop it often improves speed if you set
329this to a relatively low number, such as C<100>.
330
331Under normal circumstances you don't need to call this function.
332
333=back
238 334
239=cut 335=cut
240 336
241# support function to convert a fd into a perl filehandle 337# support function to convert a fd into a perl filehandle
242sub _fd2fh { 338sub _fd2fh {
243 return undef if $_[0] < 0; 339 return undef if $_[0] < 0;
244 340
245 # try to be perl5.6-compatible 341 # try to generate nice filehandles
246 local *AIO_FH; 342 my $sym = "IO::AIO::fd#$_[0]";
247 open AIO_FH, "+<&=$_[0]" 343 local *$sym;
344
345 open *$sym, "+<&=$_[0]" # usually works under any unix
346 or open *$sym, "<&=$_[0]" # cygwin needs this
347 or open *$sym, ">&=$_[0]" # or this
248 or return undef; 348 or return undef;
249 349
250 *AIO_FH 350 *$sym
251} 351}
252 352
253min_parallel 4; 353min_parallel 4;
254 354
255END { 355END {
256 max_parallel 0; 356 max_parallel 0;
257} 357}
258 358
2591; 3591;
260 360
261=back 361=head2 FORK BEHAVIOUR
262 362
263=head1 BUGS 363Before the fork, IO::AIO enters a quiescent state where no requests
264 364can be added in other threads and no results will be processed. After
265 - could be optimized to use more semaphores instead of filehandles. 365the fork the parent simply leaves the quiescent state and continues
366request/result processing, while the child clears the request/result
367queue (so the requests started before the fork will only be handled in
368the parent). Threats will be started on demand until the limit ste in the
369parent process has been reached again.
266 370
267=head1 SEE ALSO 371=head1 SEE ALSO
268 372
269L<Coro>, L<Linux::AIO>. 373L<Coro>, L<Linux::AIO>.
270 374

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines