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

Comparing Linux-AIO/AIO.pm (file contents):
Revision 1.12 by root, Sat May 18 21:48:36 2002 UTC vs.
Revision 1.37 by root, Wed Feb 1 23:47:26 2006 UTC

4 4
5=head1 SYNOPSIS 5=head1 SYNOPSIS
6 6
7 use Linux::AIO; 7 use Linux::AIO;
8 8
9 # This module has been mostly superseded by IO::AIO.
10
9=head1 DESCRIPTION 11=head1 DESCRIPTION
10 12
13I<This module has been mostly superseded by IO::AIO, which is API
14compatible.>
15
11This module implements asynchroneous i/o using the means available to 16This module implements asynchronous I/O using the means available to Linux
12linux - clone. It does not hook into the POSIX aio_* functions because 17- clone. It does not hook into the POSIX aio_* functions because Linux
13linux does not yet support these in the kernel. Instead, a number of 18does not yet support these in the kernel (even as of 2.6.12, only O_DIRECT
14threads are started that execute your read/writes and signal their 19files are supported) and even if, it would only allow aio_read and write,
15completion. 20not open, stat and so on.
21
22Instead, in this module a number of (non-posix) threads are started that
23execute your read/writes and signal their completion. You don't need
24thread support in your libc or perl, and the threads created by this
25module will not be visible to the pthreads library.
26
27NOTICE: the threads created by this module will automatically be killed
28when the thread calling min_parallel exits. Make sure you only ever call
29min_parallel from the same thread that loaded this module.
30
31Although the module will work with in the presence of other threads, it is
32not reentrant, so use appropriate locking yourself.
33
34=head2 API NOTES
35
36All the C<aio_*> calls are more or less thin wrappers around the syscall
37with the same name (sans C<aio_>). The arguments are similar or identical,
38and they all accept an additional C<$callback> argument which must be
39a code reference. This code reference will get called with the syscall
40return code (e.g. most syscalls return C<-1> on error, unlike perl, which
41usually delivers "false") as it's sole argument when the given syscall has
42been executed asynchronously.
43
44All functions that expect a filehandle will also accept a file descriptor.
45
46The filenames you pass to these routines I<must> be absolute. The reason
47is that at the time the request is being executed, the current working
48directory could have changed. Alternatively, you can make sure that you
49never change the current working directory.
16 50
17=over 4 51=over 4
18 52
19=cut 53=cut
20 54
21package Linux::AIO; 55package Linux::AIO;
22 56
23use base 'Exporter'; 57use base 'Exporter';
24 58
25BEGIN { 59BEGIN {
26 $VERSION = 0.111; 60 $VERSION = 1.9;
27 61
28 @EXPORT = qw(aio_read aio_write aio_open aio_close aio_stat aio_lstat); 62 @EXPORT = qw(aio_read aio_write aio_open aio_close aio_stat aio_lstat aio_unlink
63 aio_fsync aio_fdatasync aio_readahead);
29 @EXPORT_OK = qw(poll_fileno poll_cb min_parallel max_parallel nreqs); 64 @EXPORT_OK = qw(poll_fileno poll_cb min_parallel max_parallel nreqs);
30 65
31 require XSLoader; 66 require XSLoader;
32 XSLoader::load Linux::AIO, $VERSION; 67 XSLoader::load Linux::AIO, $VERSION;
33} 68}
34 69
35=item Linux::AIO::min_parallel($nthreads) 70=item Linux::AIO::min_parallel $nthreads
36 71
37Set the minimum number of AIO threads to $nthreads. You I<have> to call 72Set the minimum number of AIO threads to C<$nthreads>. The default is
38this function with a positive number at leats once, otherwise no threads 73C<1>, which means a single asynchronous operation can be done at one time
39will be started and you aio-operations will seem to hang. 74(the number of outstanding operations, however, is unlimited).
75
76It is recommended to keep the number of threads low, as some linux
77kernel versions will scale negatively with the number of threads (higher
78parallelity => MUCH higher latency).
79
80Under normal circumstances you don't need to call this function, as this
81module automatically starts a single async thread.
82
83=item Linux::AIO::max_parallel $nthreads
84
85Sets the maximum number of AIO threads to C<$nthreads>. If more than
86the specified number of threads are currently running, kill them. This
87function blocks until the limit is reached.
88
89This module automatically runs C<max_parallel 0> at program end, to ensure
90that all threads are killed and that there are no outstanding requests.
91
92Under normal circumstances you don't need to call this function.
40 93
41=item $fileno = Linux::AIO::poll_fileno 94=item $fileno = Linux::AIO::poll_fileno
42 95
43Return the request result pipe filehandle. This filehandle must be polled 96Return the I<request result pipe filehandle>. This filehandle must be
44for reading. If the pipe becomes readable you have to call C<poll_cb>. 97polled for reading by some mechanism outside this module (e.g. Event
98or select, see below). If the pipe becomes readable you have to call
99C<poll_cb> to check the results.
100
101See C<poll_cb> for an example.
45 102
46=item Linux::AIO::poll_cb 103=item Linux::AIO::poll_cb
47 104
48Process all outstanding events on the result pipe. You have to call this 105Process all outstanding events on the result pipe. You have to call this
49regularly. Returns the number of events processed. 106regularly. Returns the number of events processed. Returns immediately
107when no events are outstanding.
50 108
51You can use Event to multiplex, e.g.: 109You can use Event to multiplex, e.g.:
52 110
53 Event->io(fd => Linux::AIO::poll_fileno, 111 Event->io (fd => Linux::AIO::poll_fileno,
54 poll => 'r', async => 1, 112 poll => 'r', async => 1,
55 cb => \&Linux::AIO::poll_cb ); 113 cb => \&Linux::AIO::poll_cb);
56 114
115=item Linux::AIO::poll_wait
116
117Wait till the result filehandle becomes ready for reading (simply does a
118select on the filehandle. This is useful if you want to synchronously wait
119for some requests to finish).
120
121See C<nreqs> for an example.
57 122
58=item Linux::AIO::nreqs 123=item Linux::AIO::nreqs
59 124
60Returns the number of requests currently outstanding. 125Returns the number of requests currently outstanding.
61 126
127Example: wait till there are no outstanding requests anymore:
128
129 Linux::AIO::poll_wait, Linux::AIO::poll_cb
130 while Linux::AIO::nreqs;
131
62=item aio_open($pathname, $flags, $mode, $callback) 132=item aio_open $pathname, $flags, $mode, $callback
63 133
64Asynchronously open or create a file and call the callback with the 134Asynchronously open or create a file and call the callback with the
65filedescriptor. 135filedescriptor (NOT a perl filehandle, sorry for that, but watch out, this
136might change in the future).
66 137
138The pathname passed to C<aio_open> must be absolute. See API NOTES, above,
139for an explanation.
140
141The C<$mode> argument is a bitmask. See the C<Fcntl> module for a
142list. They are the same as used in C<sysopen>.
143
144Example:
145
146 aio_open "/etc/passwd", O_RDONLY, 0, sub {
147 if ($_[0] >= 0) {
148 open my $fh, "<&=$_[0]";
149 print "open successful, fh is $fh\n";
150 ...
151 } else {
152 die "open failed: $!\n";
153 }
154 };
155
67=item aio_close($fh, $callback) 156=item aio_close $fh, $callback
68 157
69Asynchronously close a file and call the callback with the result code. 158Asynchronously close a file and call the callback with the result code.
70 159
71=item aio_read($fh,$offset,$length, $data,$dataoffset,$callback) 160=item aio_read $fh,$offset,$length, $data,$dataoffset,$callback
72 161
73=item aio_write($fh,$offset,$length, $data,$dataoffset,$callback) 162=item aio_write $fh,$offset,$length, $data,$dataoffset,$callback
74 163
75Reads or writes C<length> bytes from the specified C<fh> and C<offset> 164Reads or writes C<length> bytes from the specified C<fh> and C<offset>
76into the scalar given by C<data> and offset C<dataoffset> and calls the 165into the scalar given by C<data> and offset C<dataoffset> and calls the
77callback without the actual number of bytes read (or undef on error). 166callback without the actual number of bytes read (or -1 on error, just
167like the syscall).
78 168
169Example: Read 15 bytes at offset 7 into scalar C<$buffer>, strating at
170offset C<0> within the scalar:
171
172 aio_read $fh, 7, 15, $buffer, 0, sub {
173 $_[0] >= 0 or die "read error: $!";
174 print "read <$buffer>\n";
175 };
176
177=item aio_readahead $fh,$offset,$length, $callback
178
179Asynchronously reads the specified byte range into the page cache, using
180the C<readahead> syscall.
181
182readahead() populates the page cache with data from a file so that
183subsequent reads from that file will not block on disk I/O. The C<$offset>
184argument specifies the starting point from which data is to be read and
185C<$length> specifies the number of bytes to be read. I/O is performed in
186whole pages, so that offset is effectively rounded down to a page boundary
187and bytes are read up to the next page boundary greater than or equal to
188(off-set+length). aio_readahead() does not read beyond the end of the
189file. The current file offset of the file is left unchanged.
190
79=item aio_stat($fh_or_path,$callback) 191=item aio_stat $fh_or_path, $callback
80 192
81=item aio_lstat($fh,$callback) 193=item aio_lstat $fh, $callback
82 194
83Works like perl's C<stat> or C<lstat> in void context, i.e. the callback 195Works like perl's C<stat> or C<lstat> in void context. The callback will
84will be called after the stat and the results will be available using 196be called after the stat and the results will be available using C<stat _>
85C<stat _> or C<-s _> etc... 197or C<-s _> etc...
198
199The pathname passed to C<aio_stat> must be absolute. See API NOTES, above,
200for an explanation.
86 201
87Currently, the stats are always 64-bit-stats, i.e. instead of returning an 202Currently, the stats are always 64-bit-stats, i.e. instead of returning an
88error when stat'ing a large file, the results will be silently truncated 203error when stat'ing a large file, the results will be silently truncated
89unless perl itself is compiled with large file support. 204unless perl itself is compiled with large file support.
90 205
206Example: Print the length of F</etc/passwd>:
207
208 aio_stat "/etc/passwd", sub {
209 $_[0] and die "stat failed: $!";
210 print "size is ", -s _, "\n";
211 };
212
213=item aio_unlink $pathname, $callback
214
215Asynchronously unlink (delete) a file and call the callback with the
216result code.
217
218=item aio_fsync $fh, $callback
219
220Asynchronously call fsync on the given filehandle and call the callback
221with the fsync result code.
222
223=item aio_fdatasync $fh, $callback
224
225Asynchronously call fdatasync on the given filehandle and call the
226callback with the fdatasync result code.
227
91=cut 228=cut
229
230min_parallel 1;
92 231
93END { 232END {
94 max_parallel 0; 233 max_parallel 0;
95} 234}
96 235
98 237
99=back 238=back
100 239
101=head1 BUGS 240=head1 BUGS
102 241
103This module has not yet been extensively tested. Watch out! 242This module has been extensively tested in a large and very busy webserver
243for many years now.
104 244
105 - perl-threads/fork interaction poorly tested.
106 - aio_open gives a fd, but all other functions expect a filehandle. 245 - aio_open gives a fd, but all other functions expect a perl filehandle.
107 246
108=head1 SEE ALSO 247=head1 SEE ALSO
109 248
110L<Coro>. 249L<Coro>, L<IO::AIO>.
111 250
112=head1 AUTHOR 251=head1 AUTHOR
113 252
114 Marc Lehmann <pcg@goof.com> 253 Marc Lehmann <schmorp@schmorp.de>
115 http://www.goof.com/pcg/marc/ 254 http://home.schmorp.de/
116 255
117=cut 256=cut
118 257

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines