1 |
=head1 NAME |
2 |
|
3 |
IO::AIO - Asynchronous Input/Output |
4 |
|
5 |
=head1 SYNOPSIS |
6 |
|
7 |
use IO::AIO; |
8 |
|
9 |
aio_open "/etc/passwd", O_RDONLY, 0, sub { |
10 |
my ($fh) = @_; |
11 |
... |
12 |
}; |
13 |
|
14 |
aio_unlink "/tmp/file", sub { }; |
15 |
|
16 |
aio_read $fh, 30000, 1024, $buffer, 0, sub { |
17 |
$_[0] > 0 or die "read error: $!"; |
18 |
}; |
19 |
|
20 |
# 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 |
|
37 |
|
38 |
=head1 DESCRIPTION |
39 |
|
40 |
This module implements asynchronous I/O using whatever means your |
41 |
operating system supports. |
42 |
|
43 |
Currently, a number of threads are started that execute your read/writes |
44 |
and signal their completion. You don't need thread support in your libc or |
45 |
perl, and the threads created by this module will not be visible to the |
46 |
pthreads library. In the future, this module might make use of the native |
47 |
aio functions available on many operating systems. However, they are often |
48 |
not well-supported (Linux doesn't allow them on normal files currently, |
49 |
for example), and they would only support aio_read and aio_write, so the |
50 |
remaining functionality would have to be implemented using threads anyway. |
51 |
|
52 |
Although the module will work with in the presence of other threads, it is |
53 |
currently not reentrant, so use appropriate locking yourself, always call |
54 |
C<poll_cb> from within the same thread, or never call C<poll_cb> (or other |
55 |
C<aio_> functions) recursively. |
56 |
|
57 |
=cut |
58 |
|
59 |
package IO::AIO; |
60 |
|
61 |
no warnings; |
62 |
|
63 |
use base 'Exporter'; |
64 |
|
65 |
use Fcntl (); |
66 |
|
67 |
BEGIN { |
68 |
$VERSION = 1.3; |
69 |
|
70 |
|
71 |
@EXPORT = qw(aio_read aio_write aio_open aio_close aio_stat aio_lstat aio_unlink |
72 |
aio_rmdir aio_symlink aio_fsync aio_fdatasync aio_readahead); |
73 |
@EXPORT_OK = qw(poll_fileno poll_cb min_parallel max_parallel max_outstanding nreqs); |
74 |
|
75 |
require XSLoader; |
76 |
XSLoader::load IO::AIO, $VERSION; |
77 |
} |
78 |
|
79 |
=head1 FUNCTIONS |
80 |
|
81 |
=head2 AIO FUNCTIONS |
82 |
|
83 |
All the C<aio_*> calls are more or less thin wrappers around the syscall |
84 |
with the same name (sans C<aio_>). The arguments are similar or identical, |
85 |
and they all accept an additional (and optional) C<$callback> argument |
86 |
which must be a code reference. This code reference will get called with |
87 |
the syscall return code (e.g. most syscalls return C<-1> on error, unlike |
88 |
perl, which usually delivers "false") as it's sole argument when the given |
89 |
syscall has been executed asynchronously. |
90 |
|
91 |
All functions expecting a filehandle keep a copy of the filehandle |
92 |
internally until the request has finished. |
93 |
|
94 |
The pathnames you pass to these routines I<must> be absolute and |
95 |
encoded in byte form. The reason for the former is that at the time the |
96 |
request is being executed, the current working directory could have |
97 |
changed. Alternatively, you can make sure that you never change the |
98 |
current working directory. |
99 |
|
100 |
To encode pathnames to byte form, either make sure you either: a) |
101 |
always pass in filenames you got from outside (command line, readdir |
102 |
etc.), b) are ASCII or ISO 8859-1, c) use the Encode module and encode |
103 |
your pathnames to the locale (or other) encoding in effect in the user |
104 |
environment, d) use Glib::filename_from_unicode on unicode filenames or e) |
105 |
use something else. |
106 |
|
107 |
=over 4 |
108 |
|
109 |
=item aio_open $pathname, $flags, $mode, $callback |
110 |
|
111 |
Asynchronously open or create a file and call the callback with a newly |
112 |
created filehandle for the file. |
113 |
|
114 |
The pathname passed to C<aio_open> must be absolute. See API NOTES, above, |
115 |
for an explanation. |
116 |
|
117 |
The C<$flags> argument is a bitmask. See the C<Fcntl> module for a |
118 |
list. They are the same as used by C<sysopen>. |
119 |
|
120 |
Likewise, C<$mode> specifies the mode of the newly created file, if it |
121 |
didn't exist and C<O_CREAT> has been given, just like perl's C<sysopen>, |
122 |
except that it is mandatory (i.e. use C<0> if you don't create new files, |
123 |
and C<0666> or C<0777> if you do). |
124 |
|
125 |
Example: |
126 |
|
127 |
aio_open "/etc/passwd", O_RDONLY, 0, sub { |
128 |
if ($_[0]) { |
129 |
print "open successful, fh is $_[0]\n"; |
130 |
... |
131 |
} else { |
132 |
die "open failed: $!\n"; |
133 |
} |
134 |
}; |
135 |
|
136 |
=item aio_close $fh, $callback |
137 |
|
138 |
Asynchronously close a file and call the callback with the result |
139 |
code. I<WARNING:> although accepted, you should not pass in a perl |
140 |
filehandle here, as perl will likely close the file descriptor another |
141 |
time when the filehandle is destroyed. Normally, you can safely call perls |
142 |
C<close> or just let filehandles go out of scope. |
143 |
|
144 |
This is supposed to be a bug in the API, so that might change. It's |
145 |
therefore best to avoid this function. |
146 |
|
147 |
=item aio_read $fh,$offset,$length, $data,$dataoffset,$callback |
148 |
|
149 |
=item aio_write $fh,$offset,$length, $data,$dataoffset,$callback |
150 |
|
151 |
Reads or writes C<length> bytes from the specified C<fh> and C<offset> |
152 |
into the scalar given by C<data> and offset C<dataoffset> and calls the |
153 |
callback without the actual number of bytes read (or -1 on error, just |
154 |
like the syscall). |
155 |
|
156 |
The C<$data> scalar I<MUST NOT> be modified in any way while the request |
157 |
is outstanding. Modifying it can result in segfaults or WW3 (if the |
158 |
necessary/optional hardware is installed). |
159 |
|
160 |
Example: Read 15 bytes at offset 7 into scalar C<$buffer>, starting at |
161 |
offset C<0> within the scalar: |
162 |
|
163 |
aio_read $fh, 7, 15, $buffer, 0, sub { |
164 |
$_[0] > 0 or die "read error: $!"; |
165 |
print "read $_[0] bytes: <$buffer>\n"; |
166 |
}; |
167 |
|
168 |
=item aio_readahead $fh,$offset,$length, $callback |
169 |
|
170 |
C<aio_readahead> populates the page cache with data from a file so that |
171 |
subsequent reads from that file will not block on disk I/O. The C<$offset> |
172 |
argument specifies the starting point from which data is to be read and |
173 |
C<$length> specifies the number of bytes to be read. I/O is performed in |
174 |
whole pages, so that offset is effectively rounded down to a page boundary |
175 |
and bytes are read up to the next page boundary greater than or equal to |
176 |
(off-set+length). C<aio_readahead> does not read beyond the end of the |
177 |
file. The current file offset of the file is left unchanged. |
178 |
|
179 |
If that syscall doesn't exist (likely if your OS isn't Linux) it will be |
180 |
emulated by simply reading the data, which would have a similar effect. |
181 |
|
182 |
=item aio_stat $fh_or_path, $callback |
183 |
|
184 |
=item aio_lstat $fh, $callback |
185 |
|
186 |
Works like perl's C<stat> or C<lstat> in void context. The callback will |
187 |
be called after the stat and the results will be available using C<stat _> |
188 |
or C<-s _> etc... |
189 |
|
190 |
The pathname passed to C<aio_stat> must be absolute. See API NOTES, above, |
191 |
for an explanation. |
192 |
|
193 |
Currently, the stats are always 64-bit-stats, i.e. instead of returning an |
194 |
error when stat'ing a large file, the results will be silently truncated |
195 |
unless perl itself is compiled with large file support. |
196 |
|
197 |
Example: Print the length of F</etc/passwd>: |
198 |
|
199 |
aio_stat "/etc/passwd", sub { |
200 |
$_[0] and die "stat failed: $!"; |
201 |
print "size is ", -s _, "\n"; |
202 |
}; |
203 |
|
204 |
=item aio_unlink $pathname, $callback |
205 |
|
206 |
Asynchronously unlink (delete) a file and call the callback with the |
207 |
result code. |
208 |
|
209 |
=item aio_rmdir $pathname, $callback |
210 |
|
211 |
Asynchronously rmdir (delete) a directory and call the callback with the |
212 |
result code. |
213 |
|
214 |
=item aio_fsync $fh, $callback |
215 |
|
216 |
Asynchronously call fsync on the given filehandle and call the callback |
217 |
with the fsync result code. |
218 |
|
219 |
=item aio_fdatasync $fh, $callback |
220 |
|
221 |
Asynchronously call fdatasync on the given filehandle and call the |
222 |
callback with the fdatasync result code. |
223 |
|
224 |
If this call isn't available because your OS lacks it or it couldn't be |
225 |
detected, it will be emulated by calling C<fsync> instead. |
226 |
|
227 |
=back |
228 |
|
229 |
=head2 SUPPORT FUNCTIONS |
230 |
|
231 |
=over 4 |
232 |
|
233 |
=item $fileno = IO::AIO::poll_fileno |
234 |
|
235 |
Return the I<request result pipe file descriptor>. This filehandle must be |
236 |
polled for reading by some mechanism outside this module (e.g. Event or |
237 |
select, see below or the SYNOPSIS). If the pipe becomes readable you have |
238 |
to call C<poll_cb> to check the results. |
239 |
|
240 |
See C<poll_cb> for an example. |
241 |
|
242 |
=item IO::AIO::poll_cb |
243 |
|
244 |
Process all outstanding events on the result pipe. You have to call this |
245 |
regularly. Returns the number of events processed. Returns immediately |
246 |
when no events are outstanding. |
247 |
|
248 |
Example: Install an Event watcher that automatically calls |
249 |
IO::AIO::poll_cb with high priority: |
250 |
|
251 |
Event->io (fd => IO::AIO::poll_fileno, |
252 |
poll => 'r', async => 1, |
253 |
cb => \&IO::AIO::poll_cb); |
254 |
|
255 |
=item IO::AIO::poll_wait |
256 |
|
257 |
Wait till the result filehandle becomes ready for reading (simply does a |
258 |
C<select> on the filehandle. This is useful if you want to synchronously wait |
259 |
for some requests to finish). |
260 |
|
261 |
See C<nreqs> for an example. |
262 |
|
263 |
=item IO::AIO::nreqs |
264 |
|
265 |
Returns the number of requests currently outstanding (i.e. for which their |
266 |
callback has not been invoked yet). |
267 |
|
268 |
Example: wait till there are no outstanding requests anymore: |
269 |
|
270 |
IO::AIO::poll_wait, IO::AIO::poll_cb |
271 |
while IO::AIO::nreqs; |
272 |
|
273 |
=item IO::AIO::flush |
274 |
|
275 |
Wait till all outstanding AIO requests have been handled. |
276 |
|
277 |
Strictly equivalent to: |
278 |
|
279 |
IO::AIO::poll_wait, IO::AIO::poll_cb |
280 |
while IO::AIO::nreqs; |
281 |
|
282 |
=item IO::AIO::poll |
283 |
|
284 |
Waits until some requests have been handled. |
285 |
|
286 |
Strictly equivalent to: |
287 |
|
288 |
IO::AIO::poll_wait, IO::AIO::poll_cb |
289 |
if IO::AIO::nreqs; |
290 |
|
291 |
=item IO::AIO::min_parallel $nthreads |
292 |
|
293 |
Set the minimum number of AIO threads to C<$nthreads>. The default is |
294 |
C<1>, which means a single asynchronous operation can be done at one time |
295 |
(the number of outstanding operations, however, is unlimited). |
296 |
|
297 |
It is recommended to keep the number of threads low, as some Linux |
298 |
kernel versions will scale negatively with the number of threads (higher |
299 |
parallelity => MUCH higher latency). With current Linux 2.6 versions, 4-32 |
300 |
threads should be fine. |
301 |
|
302 |
Under normal circumstances you don't need to call this function, as this |
303 |
module automatically starts some threads (the exact number might change, |
304 |
and is currently 4). |
305 |
|
306 |
=item IO::AIO::max_parallel $nthreads |
307 |
|
308 |
Sets the maximum number of AIO threads to C<$nthreads>. If more than |
309 |
the specified number of threads are currently running, kill them. This |
310 |
function blocks until the limit is reached. |
311 |
|
312 |
This module automatically runs C<max_parallel 0> at program end, to ensure |
313 |
that all threads are killed and that there are no outstanding requests. |
314 |
|
315 |
Under normal circumstances you don't need to call this function. |
316 |
|
317 |
=item $oldnreqs = IO::AIO::max_outstanding $nreqs |
318 |
|
319 |
Sets the maximum number of outstanding requests to C<$nreqs>. If you |
320 |
try to queue up more than this number of requests, the caller will block until |
321 |
some requests have been handled. |
322 |
|
323 |
The default is very large, so normally there is no practical limit. If you |
324 |
queue up many requests in a loop it it often improves speed if you set |
325 |
this to a relatively low number, such as C<100>. |
326 |
|
327 |
Under normal circumstances you don't need to call this function. |
328 |
|
329 |
=back |
330 |
|
331 |
=cut |
332 |
|
333 |
# support function to convert a fd into a perl filehandle |
334 |
sub _fd2fh { |
335 |
return undef if $_[0] < 0; |
336 |
|
337 |
# try to generate nice filehandles |
338 |
my $sym = "IO::AIO::fd#$_[0]"; |
339 |
local *$sym; |
340 |
|
341 |
open *$sym, "+<&=$_[0]" # usually works under any unix |
342 |
or open *$sym, "<&=$_[0]" # cygwin needs this |
343 |
or open *$sym, ">&=$_[0]" # or this |
344 |
or return undef; |
345 |
|
346 |
*$sym |
347 |
} |
348 |
|
349 |
min_parallel 4; |
350 |
|
351 |
END { |
352 |
max_parallel 0; |
353 |
} |
354 |
|
355 |
1; |
356 |
|
357 |
=head2 FORK BEHAVIOUR |
358 |
|
359 |
Before the fork IO::AIO enters a quiescent state where no requests can be |
360 |
added in other threads and no results will be processed. After the fork |
361 |
the parent simply leaves the quiescent state and continues request/result |
362 |
processing, while the child clears the request/result queue and starts the |
363 |
same number of threads as were in use by the parent. |
364 |
|
365 |
=head1 SEE ALSO |
366 |
|
367 |
L<Coro>, L<Linux::AIO>. |
368 |
|
369 |
=head1 AUTHOR |
370 |
|
371 |
Marc Lehmann <schmorp@schmorp.de> |
372 |
http://home.schmorp.de/ |
373 |
|
374 |
=cut |
375 |
|