Linux-AIO/AIO.pm

=head1 NAME

Linux::AIO - linux-specific aio implemented using clone

=head1 SYNOPSIS

 use Linux::AIO;

 # This module has been mostly superseded by IO::AIO.

=head1 DESCRIPTION

I<This module has been mostly superseded by IO::AIO, which is API 
compatible.>

This module implements asynchronous I/O using the means available to Linux
- clone. It does not hook into the POSIX aio_* functions because Linux
does not yet support these in the kernel (even as of 2.6.12, only O_DIRECT
files are supported) and even if, it would only allow aio_read and write,
not open, stat and so on.

Instead, in this module a number of (non-posix) threads are started that
execute your read/writes and signal their completion. You don't need
thread support in your libc or perl, and the threads created by this
module will not be visible to the pthreads library.

NOTICE: the threads created by this module will automatically be killed
when the thread calling min_parallel exits. Make sure you only ever call
min_parallel from the same thread that loaded this module.

Although the module will work with in the presence of other threads, it is
not reentrant, so use appropriate locking yourself.

=head2 API NOTES

All the C<aio_*> calls are more or less thin wrappers around the syscall
with the same name (sans C<aio_>). The arguments are similar or identical,
and they all accept an additional C<$callback> argument which must be
a code reference. This code reference will get called with the syscall
return code (e.g. most syscalls return C<-1> on error, unlike perl, which
usually delivers "false") as it's sole argument when the given syscall has
been executed asynchronously.

All functions that expect a filehandle will also accept a file descriptor.

The filenames you pass to these routines I<must> be absolute. The reason
is that at the time the request is being executed, the current working
directory could have changed. Alternatively, you can make sure that you
never change the current working directory.

=over 4

=cut

package Linux::AIO;

use base 'Exporter';

BEGIN {
   $VERSION = 1.9;

   @EXPORT = qw(aio_read aio_write aio_open aio_close aio_stat aio_lstat aio_unlink
                aio_fsync aio_fdatasync aio_readahead);
   @EXPORT_OK = qw(poll_fileno poll_cb min_parallel max_parallel nreqs);

   require XSLoader;
   XSLoader::load Linux::AIO, $VERSION;
}

=item Linux::AIO::min_parallel $nthreads

Set the minimum number of AIO threads to C<$nthreads>. The default is
C<1>, which means a single asynchronous operation can be done at one time
(the number of outstanding operations, however, is unlimited).

It is recommended to keep the number of threads low, as some linux
kernel versions will scale negatively with the number of threads (higher
parallelity => MUCH higher latency).

Under normal circumstances you don't need to call this function, as this
module automatically starts a single async thread.

=item Linux::AIO::max_parallel $nthreads

Sets the maximum number of AIO threads to C<$nthreads>. If more than
the specified number of threads are currently running, kill them. This
function blocks until the limit is reached.

This module automatically runs C<max_parallel 0> at program end, to ensure
that all threads are killed and that there are no outstanding requests.

Under normal circumstances you don't need to call this function.

=item $fileno = Linux::AIO::poll_fileno

Return the I<request result pipe filehandle>. This filehandle must be
polled for reading by some mechanism outside this module (e.g. Event
or select, see below). If the pipe becomes readable you have to call
C<poll_cb> to check the results.

See C<poll_cb> for an example.

=item Linux::AIO::poll_cb

Process all outstanding events on the result pipe. You have to call this
regularly. Returns the number of events processed. Returns immediately
when no events are outstanding.

You can use Event to multiplex, e.g.:

   Event->io (fd => Linux::AIO::poll_fileno,
              poll => 'r', async => 1,
              cb => \&Linux::AIO::poll_cb);

=item Linux::AIO::poll_wait

Wait till the result filehandle becomes ready for reading (simply does a
select on the filehandle. This is useful if you want to synchronously wait
for some requests to finish).

See C<nreqs> for an example.

=item Linux::AIO::nreqs

Returns the number of requests currently outstanding.

Example: wait till there are no outstanding requests anymore:

   Linux::AIO::poll_wait, Linux::AIO::poll_cb
      while Linux::AIO::nreqs;

=item aio_open $pathname, $flags, $mode, $callback

Asynchronously open or create a file and call the callback with the
filedescriptor (NOT a perl filehandle, sorry for that, but watch out, this
might change in the future).

The pathname passed to C<aio_open> must be absolute. See API NOTES, above,
for an explanation.

The C<$mode> argument is a bitmask. See the C<Fcntl> module for a
list. They are the same as used in C<sysopen>.

Example:

   aio_open "/etc/passwd", O_RDONLY, 0, sub {
      if ($_[0] >= 0) {
         open my $fh, "<&=$_[0]";
         print "open successful, fh is $fh\n";
         ...
      } else {
         die "open failed: $!\n";
      }
   };

=item aio_close $fh, $callback

Asynchronously close a file and call the callback with the result code.

=item aio_read  $fh,$offset,$length, $data,$dataoffset,$callback

=item aio_write $fh,$offset,$length, $data,$dataoffset,$callback

Reads or writes C<length> bytes from the specified C<fh> and C<offset>
into the scalar given by C<data> and offset C<dataoffset> and calls the
callback without the actual number of bytes read (or -1 on error, just
like the syscall).

Example: Read 15 bytes at offset 7 into scalar C<$buffer>, strating at
offset C<0> within the scalar:

   aio_read $fh, 7, 15, $buffer, 0, sub {
      $_[0] >= 0 or die "read error: $!";
      print "read <$buffer>\n";
   };

=item aio_readahead $fh,$offset,$length, $callback

Asynchronously reads the specified byte range into the page cache, using
the C<readahead> syscall.

readahead() populates the page cache with data from a file so that
subsequent reads from that file will not block on disk I/O. The C<$offset>
argument specifies the starting point from which data is to be read and
C<$length> specifies the number of bytes to be read. I/O is performed in
whole pages, so that offset is effectively rounded down to a page boundary
and bytes are read up to the next page boundary greater than or equal to
(off-set+length). aio_readahead() does not read beyond the end of the
file. The current file offset of the file is left unchanged.

=item aio_stat  $fh_or_path, $callback

=item aio_lstat $fh, $callback

Works like perl's C<stat> or C<lstat> in void context. The callback will
be called after the stat and the results will be available using C<stat _>
or C<-s _> etc...

The pathname passed to C<aio_stat> must be absolute. See API NOTES, above,
for an explanation.

Currently, the stats are always 64-bit-stats, i.e. instead of returning an
error when stat'ing a large file, the results will be silently truncated
unless perl itself is compiled with large file support.

Example: Print the length of F</etc/passwd>:

   aio_stat "/etc/passwd", sub {
      $_[0] and die "stat failed: $!";
      print "size is ", -s _, "\n";
   };

=item aio_unlink $pathname, $callback

Asynchronously unlink (delete) a file and call the callback with the
result code.

=item aio_fsync $fh, $callback

Asynchronously call fsync on the given filehandle and call the callback
with the fsync result code.

=item aio_fdatasync $fh, $callback

Asynchronously call fdatasync on the given filehandle and call the
callback with the fdatasync result code.

=cut

min_parallel 1;

END {
   max_parallel 0;
}

1;

=back

=head1 BUGS

This module has been extensively tested in a large and very busy webserver
for many years now.

   - aio_open gives a fd, but all other functions expect a perl filehandle.

=head1 SEE ALSO

L<Coro>, L<IO::AIO>.

=head1 AUTHOR

 Marc Lehmann <schmorp@schmorp.de>
 http://home.schmorp.de/

=cut

Revision:	1.37
Committed:	Wed Feb 1 23:47:26 2006 UTC (18 years, 3 months ago) by root
Branch:	MAIN
CVS Tags:	rel-1_9, HEAD
Changes since 1.36:	+1 -1 lines
Log Message:	* empty log message *
#	User	Rev	Content
1	root	1.1	=head1 NAME
2
3			Linux::AIO - linux-specific aio implemented using clone
4
5			=head1 SYNOPSIS
6
7			use Linux::AIO;
8
9	root	1.36	# This module has been mostly superseded by IO::AIO.
10
11	root	1.1	=head1 DESCRIPTION
12
13	root	1.36	I<This module has been mostly superseded by IO::AIO, which is API
14			compatible.>
15
16	root	1.27	This module implements asynchronous I/O using the means available to Linux
17			- clone. It does not hook into the POSIX aio_* functions because Linux
18			does not yet support these in the kernel (even as of 2.6.12, only O_DIRECT
19			files are supported) and even if, it would only allow aio_read and write,
20			not open, stat and so on.
21	root	1.14
22			Instead, in this module a number of (non-posix) threads are started that
23			execute your read/writes and signal their completion. You don't need
24			thread support in your libc or perl, and the threads created by this
25			module will not be visible to the pthreads library.
26	root	1.3
27	root	1.17	NOTICE: the threads created by this module will automatically be killed
28			when the thread calling min_parallel exits. Make sure you only ever call
29			min_parallel from the same thread that loaded this module.
30
31	root	1.27	Although the module will work with in the presence of other threads, it is
32			not reentrant, so use appropriate locking yourself.
33	root	1.16
34	root	1.30	=head2 API NOTES
35
36			All the C<aio_*> calls are more or less thin wrappers around the syscall
37			with the same name (sans C<aio_>). The arguments are similar or identical,
38			and they all accept an additional C<$callback> argument which must be
39			a code reference. This code reference will get called with the syscall
40			return code (e.g. most syscalls return C<-1> on error, unlike perl, which
41			usually delivers "false") as it's sole argument when the given syscall has
42			been executed asynchronously.
43
44			All functions that expect a filehandle will also accept a file descriptor.
45
46	root	1.32	The filenames you pass to these routines I<must> be absolute. The reason
47			is that at the time the request is being executed, the current working
48			directory could have changed. Alternatively, you can make sure that you
49	root	1.33	never change the current working directory.
50	root	1.32
51	root	1.1	=over 4
52
53			=cut
54
55			package Linux::AIO;
56
57	root	1.2	use base 'Exporter';
58
59	root	1.1	BEGIN {
60	root	1.37	$VERSION = 1.9;
61	root	1.1
62	root	1.29	@EXPORT = qw(aio_read aio_write aio_open aio_close aio_stat aio_lstat aio_unlink
63			aio_fsync aio_fdatasync aio_readahead);
64	root	1.2	@EXPORT_OK = qw(poll_fileno poll_cb min_parallel max_parallel nreqs);
65
66	root	1.1	require XSLoader;
67			XSLoader::load Linux::AIO, $VERSION;
68			}
69
70	root	1.14	=item Linux::AIO::min_parallel $nthreads
71	root	1.1
72	root	1.16	Set the minimum number of AIO threads to C<$nthreads>. The default is
73			C<1>, which means a single asynchronous operation can be done at one time
74			(the number of outstanding operations, however, is unlimited).
75	root	1.1
76	root	1.24	It is recommended to keep the number of threads low, as some linux
77	root	1.14	kernel versions will scale negatively with the number of threads (higher
78			parallelity => MUCH higher latency).
79
80	root	1.30	Under normal circumstances you don't need to call this function, as this
81			module automatically starts a single async thread.
82
83	root	1.24	=item Linux::AIO::max_parallel $nthreads
84
85			Sets the maximum number of AIO threads to C<$nthreads>. If more than
86			the specified number of threads are currently running, kill them. This
87			function blocks until the limit is reached.
88
89			This module automatically runs C<max_parallel 0> at program end, to ensure
90			that all threads are killed and that there are no outstanding requests.
91
92	root	1.30	Under normal circumstances you don't need to call this function.
93
94	root	1.2	=item $fileno = Linux::AIO::poll_fileno
95
96	root	1.14	Return the I<request result pipe filehandle>. This filehandle must be
97			polled for reading by some mechanism outside this module (e.g. Event
98			or select, see below). If the pipe becomes readable you have to call
99			C<poll_cb> to check the results.
100	root	1.2
101	root	1.30	See C<poll_cb> for an example.
102
103	root	1.2	=item Linux::AIO::poll_cb
104
105			Process all outstanding events on the result pipe. You have to call this
106	root	1.14	regularly. Returns the number of events processed. Returns immediately
107			when no events are outstanding.
108	root	1.2
109	root	1.3	You can use Event to multiplex, e.g.:
110
111	root	1.14	Event->io (fd => Linux::AIO::poll_fileno,
112			poll => 'r', async => 1,
113	root	1.30	cb => \&Linux::AIO::poll_cb);
114	root	1.3
115	root	1.24	=item Linux::AIO::poll_wait
116
117			Wait till the result filehandle becomes ready for reading (simply does a
118			select on the filehandle. This is useful if you want to synchronously wait
119			for some requests to finish).
120
121	root	1.30	See C<nreqs> for an example.
122
123	root	1.2	=item Linux::AIO::nreqs
124
125			Returns the number of requests currently outstanding.
126	root	1.4
127	root	1.30	Example: wait till there are no outstanding requests anymore:
128
129	root	1.32	Linux::AIO::poll_wait, Linux::AIO::poll_cb
130			while Linux::AIO::nreqs;
131	root	1.30
132			=item aio_open $pathname, $flags, $mode, $callback
133	root	1.10
134			Asynchronously open or create a file and call the callback with the
135	root	1.14	filedescriptor (NOT a perl filehandle, sorry for that, but watch out, this
136			might change in the future).
137	root	1.10
138	root	1.34	The pathname passed to C<aio_open> must be absolute. See API NOTES, above,
139			for an explanation.
140
141	root	1.30	The C<$mode> argument is a bitmask. See the C<Fcntl> module for a
142			list. They are the same as used in C<sysopen>.
143
144			Example:
145
146			aio_open "/etc/passwd", O_RDONLY, 0, sub {
147			if ($_[0] >= 0) {
148	root	1.35	open my $fh, "<&=$_[0]";
149	root	1.30	print "open successful, fh is $fh\n";
150			...
151			} else {
152			die "open failed: $!\n";
153			}
154			};
155
156	root	1.14	=item aio_close $fh, $callback
157	root	1.10
158			Asynchronously close a file and call the callback with the result code.
159
160	root	1.14	=item aio_read $fh,$offset,$length, $data,$dataoffset,$callback
161	root	1.9
162	root	1.14	=item aio_write $fh,$offset,$length, $data,$dataoffset,$callback
163	root	1.4
164			Reads or writes C<length> bytes from the specified C<fh> and C<offset>
165			into the scalar given by C<data> and offset C<dataoffset> and calls the
166	root	1.30	callback without the actual number of bytes read (or -1 on error, just
167			like the syscall).
168
169			Example: Read 15 bytes at offset 7 into scalar C<$buffer>, strating at
170			offset 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, $callback
178
179			Asynchronously reads the specified byte range into the page cache, using
180			the C<readahead> syscall.
181
182			readahead() populates the page cache with data from a file so that
183			subsequent reads from that file will not block on disk I/O. The C<$offset>
184			argument specifies the starting point from which data is to be read and
185			C<$length> specifies the number of bytes to be read. I/O is performed in
186			whole pages, so that offset is effectively rounded down to a page boundary
187			and 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
189			file. The current file offset of the file is left unchanged.
190	root	1.4
191	root	1.14	=item aio_stat $fh_or_path, $callback
192	root	1.4
193	root	1.14	=item aio_lstat $fh, $callback
194	root	1.5
195	root	1.14	Works like perl's C<stat> or C<lstat> in void context. The callback will
196			be called after the stat and the results will be available using C<stat _>
197			or C<-s _> etc...
198	root	1.5
199	root	1.32	The pathname passed to C<aio_stat> must be absolute. See API NOTES, above,
200			for an explanation.
201
202	root	1.10	Currently, the stats are always 64-bit-stats, i.e. instead of returning an
203			error when stat'ing a large file, the results will be silently truncated
204			unless perl itself is compiled with large file support.
205	root	1.17
206	root	1.30	Example: 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
215			Asynchronously unlink (delete) a file and call the callback with the
216			result code.
217
218			=item aio_fsync $fh, $callback
219
220			Asynchronously call fsync on the given filehandle and call the callback
221			with the fsync result code.
222
223			=item aio_fdatasync $fh, $callback
224	root	1.17
225	root	1.30	Asynchronously call fdatasync on the given filehandle and call the
226			callback with the fdatasync result code.
227	root	1.2
228	root	1.1	=cut
229	root	1.16
230			min_parallel 1;
231	root	1.2
232			END {
233			max_parallel 0;
234			}
235	root	1.1
236			1;
237
238			=back
239
240			=head1 BUGS
241
242	root	1.14	This module has been extensively tested in a large and very busy webserver
243			for many years now.
244	root	1.1
245	root	1.14	- aio_open gives a fd, but all other functions expect a perl filehandle.
246	root	1.6
247	root	1.1	=head1 SEE ALSO
248
249	root	1.34	L<Coro>, L<IO::AIO>.
250	root	1.1
251			=head1 AUTHOR
252
253	root	1.22	Marc Lehmann <schmorp@schmorp.de>
254	root	1.20	http://home.schmorp.de/
255	root	1.1
256			=cut
257