ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/IO-AIO/README
(Generate patch)

Comparing IO-AIO/README (file contents):
Revision 1.6 by root, Mon Jul 11 03:29:39 2005 UTC vs.
Revision 1.17 by root, Mon Jun 26 14:53:55 2006 UTC

… …
13 13
14 aio_read $fh, 30000, 1024, $buffer, 0, sub { 14 aio_read $fh, 30000, 1024, $buffer, 0, sub {
15 $_[0] > 0 or die "read error: $!"; 15 $_[0] > 0 or die "read error: $!";
16 }; 16 };
17 17
18 # AnyEvent
19 open my $fh, "<&=" . IO::AIO::poll_fileno or die "$!";
20 my $w = AnyEvent->io (fh => $fh, poll => 'r', cb => sub { IO::AIO::poll_cb });
21
18 # Event 22 # Event
19 Event->io (fd => IO::AIO::poll_fileno, 23 Event->io (fd => IO::AIO::poll_fileno,
20 poll => 'r', 24 poll => 'r',
21 cb => \&IO::AIO::poll_cb); 25 cb => \&IO::AIO::poll_cb);
22 26
23 # Glib/Gtk2 27 # Glib/Gtk2
24 add_watch Glib::IO IO::AIO::poll_fileno, 28 add_watch Glib::IO IO::AIO::poll_fileno,
25 in => sub { IO::AIO::poll_cb, 1 }; 29 in => sub { IO::AIO::poll_cb; 1 };
26 30
27 # Tk 31 # Tk
28 Tk::Event::IO->fileevent (IO::AIO::poll_fileno, "", 32 Tk::Event::IO->fileevent (IO::AIO::poll_fileno, "",
29 readable => \&IO::AIO::poll_cb); 33 readable => \&IO::AIO::poll_cb);
30 34
… …
45 currently, for example), and they would only support aio_read and 49 currently, for example), and they would only support aio_read and
46 aio_write, so the remaining functionality would have to be implemented 50 aio_write, so the remaining functionality would have to be implemented
47 using threads anyway. 51 using threads anyway.
48 52
49 Although the module will work with in the presence of other threads, it 53 Although the module will work with in the presence of other threads, it
50 is currently not reentrant, so use appropriate locking yourself. 54 is currently not reentrant, so use appropriate locking yourself, always
55 call "poll_cb" from within the same thread, or never call "poll_cb" (or
56 other "aio_" functions) recursively.
51 57
52FUNCTIONS 58FUNCTIONS
53 AIO FUNCTIONS 59 AIO FUNCTIONS
54 All the "aio_*" calls are more or less thin wrappers around the syscall 60 All the "aio_*" calls are more or less thin wrappers around the syscall
55 with the same name (sans "aio_"). The arguments are similar or 61 with the same name (sans "aio_"). The arguments are similar or
… …
57 argument which must be a code reference. This code reference will get 63 argument which must be a code reference. This code reference will get
58 called with the syscall return code (e.g. most syscalls return -1 on 64 called with the syscall return code (e.g. most syscalls return -1 on
59 error, unlike perl, which usually delivers "false") as it's sole 65 error, unlike perl, which usually delivers "false") as it's sole
60 argument when the given syscall has been executed asynchronously. 66 argument when the given syscall has been executed asynchronously.
61 67
62 All functions that expect a filehandle will also accept a file 68 All functions expecting a filehandle keep a copy of the filehandle
63 descriptor. 69 internally until the request has finished.
64 70
65 The filenames you pass to these routines *must* be absolute. The reason 71 The pathnames you pass to these routines *must* be absolute and encoded
66 is that at the time the request is being executed, the current working 72 in byte form. The reason for the former is that at the time the request
67 directory could have changed. Alternatively, you can make sure that you 73 is being executed, the current working directory could have changed.
68 never change the current working directory. 74 Alternatively, you can make sure that you never change the current
75 working directory.
69 76
77 To encode pathnames to byte form, either make sure you either: a) always
78 pass in filenames you got from outside (command line, readdir etc.), b)
79 are ASCII or ISO 8859-1, c) use the Encode module and encode your
80 pathnames to the locale (or other) encoding in effect in the user
81 environment, d) use Glib::filename_from_unicode on unicode filenames or
82 e) use something else.
83
70 aio_open $pathname, $flags, $mode, $callback 84 aio_open $pathname, $flags, $mode, $callback->($fh)
71 Asynchronously open or create a file and call the callback with a 85 Asynchronously open or create a file and call the callback with a
72 newly created filehandle for the file. 86 newly created filehandle for the file.
73 87
74 The pathname passed to "aio_open" must be absolute. See API NOTES, 88 The pathname passed to "aio_open" must be absolute. See API NOTES,
75 above, for an explanation. 89 above, for an explanation.
76 90
77 The $mode argument is a bitmask. See the "Fcntl" module for a list. 91 The $flags argument is a bitmask. See the "Fcntl" module for a list.
78 They are the same as used in "sysopen". 92 They are the same as used by "sysopen".
93
94 Likewise, $mode specifies the mode of the newly created file, if it
95 didn't exist and "O_CREAT" has been given, just like perl's
96 "sysopen", except that it is mandatory (i.e. use 0 if you don't
97 create new files, and 0666 or 0777 if you do).
79 98
80 Example: 99 Example:
81 100
82 aio_open "/etc/passwd", O_RDONLY, 0, sub { 101 aio_open "/etc/passwd", O_RDONLY, 0, sub {
83 if ($_[0]) { 102 if ($_[0]) {
… …
86 } else { 105 } else {
87 die "open failed: $!\n"; 106 die "open failed: $!\n";
88 } 107 }
89 }; 108 };
90 109
91 aio_close $fh, $callback 110 aio_close $fh, $callback->($status)
92 Asynchronously close a file and call the callback with the result 111 Asynchronously close a file and call the callback with the result
93 code. *WARNING:* although accepted, you should not pass in a perl 112 code. *WARNING:* although accepted, you should not pass in a perl
94 filehandle here, as perl will likely close the file descriptor 113 filehandle here, as perl will likely close the file descriptor
95 itself when the filehandle is destroyed. Normally, you can safely 114 another time when the filehandle is destroyed. Normally, you can
96 call perls "close" or just let filehandles go out of scope. 115 safely call perls "close" or just let filehandles go out of scope.
97 116
117 This is supposed to be a bug in the API, so that might change. It's
118 therefore best to avoid this function.
119
98 aio_read $fh,$offset,$length, $data,$dataoffset,$callback 120 aio_read $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
99 aio_write $fh,$offset,$length, $data,$dataoffset,$callback 121 aio_write $fh,$offset,$length, $data,$dataoffset, $callback->($retval)
100 Reads or writes "length" bytes from the specified "fh" and "offset" 122 Reads or writes "length" bytes from the specified "fh" and "offset"
101 into the scalar given by "data" and offset "dataoffset" and calls 123 into the scalar given by "data" and offset "dataoffset" and calls
102 the callback without the actual number of bytes read (or -1 on 124 the callback without the actual number of bytes read (or -1 on
103 error, just like the syscall). 125 error, just like the syscall).
104 126
127 The $data scalar *MUST NOT* be modified in any way while the request
128 is outstanding. Modifying it can result in segfaults or WW3 (if the
129 necessary/optional hardware is installed).
130
105 Example: Read 15 bytes at offset 7 into scalar $buffer, starting at 131 Example: Read 15 bytes at offset 7 into scalar $buffer, starting at
106 offset 0 within the scalar: 132 offset 0 within the scalar:
107 133
108 aio_read $fh, 7, 15, $buffer, 0, sub { 134 aio_read $fh, 7, 15, $buffer, 0, sub {
109 $_[0] > 0 or die "read error: $!"; 135 $_[0] > 0 or die "read error: $!";
110 print "read $_[0] bytes: <$buffer>\n"; 136 print "read $_[0] bytes: <$buffer>\n";
111 }; 137 };
112 138
139 aio_move $srcpath, $dstpath, $callback->($status)
140 [EXPERIMENTAL]
141
142 Try to move the *file* (directories not supported as either source
143 or destination) from $srcpath to $dstpath and call the callback with
144 the 0 (error) or -1 ok.
145
146 This is a composite request that tries to rename(2) the file first.
147 If rename files with "EXDEV", it creates the destination file with
148 mode 0200 and copies the contents of the source file into it using
149 "aio_sendfile", followed by restoring atime, mtime, access mode and
150 uid/gid, in that order, and unlinking the $srcpath.
151
152 If an error occurs, the partial destination file will be unlinked,
153 if possible, except when setting atime, mtime, access mode and
154 uid/gid, where errors are being ignored.
155
156 aio_sendfile $out_fh, $in_fh, $in_offset, $length, $callback->($retval)
157 Tries to copy $length bytes from $in_fh to $out_fh. It starts
158 reading at byte offset $in_offset, and starts writing at the current
159 file offset of $out_fh. Because of that, it is not safe to issue
160 more than one "aio_sendfile" per $out_fh, as they will interfere
161 with each other.
162
163 This call tries to make use of a native "sendfile" syscall to
164 provide zero-copy operation. For this to work, $out_fh should refer
165 to a socket, and $in_fh should refer to mmap'able file.
166
167 If the native sendfile call fails or is not implemented, it will be
168 emulated, so you can call "aio_sendfile" on any type of filehandle
169 regardless of the limitations of the operating system.
170
171 Please note, however, that "aio_sendfile" can read more bytes from
172 $in_fh than are written, and there is no way to find out how many
173 bytes have been read from "aio_sendfile" alone, as "aio_sendfile"
174 only provides the number of bytes written to $out_fh. Only if the
175 result value equals $length one can assume that $length bytes have
176 been read.
177
113 aio_readahead $fh,$offset,$length, $callback 178 aio_readahead $fh,$offset,$length, $callback->($retval)
114 Asynchronously reads the specified byte range into the page cache,
115 using the "readahead" syscall. If that syscall doesn't exist (likely
116 if your OS isn't Linux) the status will be -1 and $! is set to
117 ENOSYS.
118
119 readahead() populates the page cache with data from a file so that 179 "aio_readahead" populates the page cache with data from a file so
120 subsequent reads from that file will not block on disk I/O. The 180 that subsequent reads from that file will not block on disk I/O. The
121 $offset argument specifies the starting point from which data is to 181 $offset argument specifies the starting point from which data is to
122 be read and $length specifies the number of bytes to be read. I/O is 182 be read and $length specifies the number of bytes to be read. I/O is
123 performed in whole pages, so that offset is effectively rounded down 183 performed in whole pages, so that offset is effectively rounded down
124 to a page boundary and bytes are read up to the next page boundary 184 to a page boundary and bytes are read up to the next page boundary
125 greater than or equal to (off-set+length). aio_readahead() does not 185 greater than or equal to (off-set+length). "aio_readahead" does not
126 read beyond the end of the file. The current file offset of the file 186 read beyond the end of the file. The current file offset of the file
127 is left unchanged. 187 is left unchanged.
128 188
189 If that syscall doesn't exist (likely if your OS isn't Linux) it
190 will be emulated by simply reading the data, which would have a
191 similar effect.
192
129 aio_stat $fh_or_path, $callback 193 aio_stat $fh_or_path, $callback->($status)
130 aio_lstat $fh, $callback 194 aio_lstat $fh, $callback->($status)
131 Works like perl's "stat" or "lstat" in void context. The callback 195 Works like perl's "stat" or "lstat" in void context. The callback
132 will be called after the stat and the results will be available 196 will be called after the stat and the results will be available
133 using "stat _" or "-s _" etc... 197 using "stat _" or "-s _" etc...
134 198
135 The pathname passed to "aio_stat" must be absolute. See API NOTES, 199 The pathname passed to "aio_stat" must be absolute. See API NOTES,
… …
145 aio_stat "/etc/passwd", sub { 209 aio_stat "/etc/passwd", sub {
146 $_[0] and die "stat failed: $!"; 210 $_[0] and die "stat failed: $!";
147 print "size is ", -s _, "\n"; 211 print "size is ", -s _, "\n";
148 }; 212 };
149 213
150 aio_unlink $pathname, $callback 214 aio_unlink $pathname, $callback->($status)
151 Asynchronously unlink (delete) a file and call the callback with the 215 Asynchronously unlink (delete) a file and call the callback with the
152 result code. 216 result code.
153 217
218 aio_link $srcpath, $dstpath, $callback->($status)
219 Asynchronously create a new link to the existing object at $srcpath
220 at the path $dstpath and call the callback with the result code.
221
222 aio_symlink $srcpath, $dstpath, $callback->($status)
223 Asynchronously create a new symbolic link to the existing object at
224 $srcpath at the path $dstpath and call the callback with the result
225 code.
226
227 aio_rename $srcpath, $dstpath, $callback->($status)
228 Asynchronously rename the object at $srcpath to $dstpath, just as
229 rename(2) and call the callback with the result code.
230
231 aio_rmdir $pathname, $callback->($status)
232 Asynchronously rmdir (delete) a directory and call the callback with
233 the result code.
234
235 aio_readdir $pathname, $callback->($entries)
236 Unlike the POSIX call of the same name, "aio_readdir" reads an
237 entire directory (i.e. opendir + readdir + closedir). The entries
238 will not be sorted, and will NOT include the "." and ".." entries.
239
240 The callback a single argument which is either "undef" or an
241 array-ref with the filenames.
242
243 aio_scandir $path, $maxreq, $callback->($dirs, $nondirs)
244 Scans a directory (similar to "aio_readdir") and tries to separate
245 the entries of directory $path into two sets of names, ones you can
246 recurse into (directories), and ones you cannot recurse into
247 (everything else).
248
249 "aio_scandir" is a composite request that consists of many
250 aio-primitives. $maxreq specifies the maximum number of outstanding
251 aio requests that this function generates. If it is "<= 0", then a
252 suitable default will be chosen (currently 8).
253
254 On error, the callback is called without arguments, otherwise it
255 receives two array-refs with path-relative entry names.
256
257 Example:
258
259 aio_scandir $dir, 0, sub {
260 my ($dirs, $nondirs) = @_;
261 print "real directories: @$dirs\n";
262 print "everything else: @$nondirs\n";
263 };
264
265 Implementation notes.
266
267 The "aio_readdir" cannot be avoided, but "stat()"'ing every entry
268 can.
269
270 After reading the directory, the modification time, size etc. of the
271 directory before and after the readdir is checked, and if they
272 match, the link count will be used to decide how many entries are
273 directories (if >= 2). Otherwise, no knowledge of the number of
274 subdirectories will be assumed.
275
276 Then entires will be sorted into likely directories (everything
277 without a non-initial dot) and likely non-directories (everything
278 else). Then every entry + "/." will be "stat"'ed, likely directories
279 first. This is often faster because filesystems might detect the
280 type of the entry without reading the inode data (e.g. ext2fs
281 filetype feature). If that succeeds, it assumes that the entry is a
282 directory or a symlink to directory (which will be checked
283 seperately).
284
285 If the known number of directories has been reached, the rest of the
286 entries is assumed to be non-directories.
287
154 aio_fsync $fh, $callback 288 aio_fsync $fh, $callback->($status)
155 Asynchronously call fsync on the given filehandle and call the 289 Asynchronously call fsync on the given filehandle and call the
156 callback with the fsync result code. 290 callback with the fsync result code.
157 291
158 aio_fdatasync $fh, $callback 292 aio_fdatasync $fh, $callback->($status)
159 Asynchronously call fdatasync on the given filehandle and call the 293 Asynchronously call fdatasync on the given filehandle and call the
160 callback with the fdatasync result code. 294 callback with the fdatasync result code.
161 295
296 If this call isn't available because your OS lacks it or it couldn't
297 be detected, it will be emulated by calling "fsync" instead.
298
162 SUPPORT FUNCTIONS 299 SUPPORT FUNCTIONS
163 $fileno = IO::AIO::poll_fileno 300 $fileno = IO::AIO::poll_fileno
164 Return the *request result pipe filehandle*. This filehandle must be 301 Return the *request result pipe file descriptor*. This filehandle
165 polled for reading by some mechanism outside this module (e.g. Event 302 must be polled for reading by some mechanism outside this module
166 or select, see below). If the pipe becomes readable you have to call 303 (e.g. Event or select, see below or the SYNOPSIS). If the pipe
167 "poll_cb" to check the results. 304 becomes readable you have to call "poll_cb" to check the results.
168 305
169 See "poll_cb" for an example. 306 See "poll_cb" for an example.
170 307
171 IO::AIO::poll_cb 308 IO::AIO::poll_cb
172 Process all outstanding events on the result pipe. You have to call 309 Process all outstanding events on the result pipe. You have to call
173 this regularly. Returns the number of events processed. Returns 310 this regularly. Returns the number of events processed. Returns
174 immediately when no events are outstanding. 311 immediately when no events are outstanding.
175 312
176 You can use Event to multiplex, e.g.: 313 Example: Install an Event watcher that automatically calls
314 IO::AIO::poll_cb with high priority:
177 315
178 Event->io (fd => IO::AIO::poll_fileno, 316 Event->io (fd => IO::AIO::poll_fileno,
179 poll => 'r', async => 1, 317 poll => 'r', async => 1,
180 cb => \&IO::AIO::poll_cb); 318 cb => \&IO::AIO::poll_cb);
181 319
182 IO::AIO::poll_wait 320 IO::AIO::poll_wait
183 Wait till the result filehandle becomes ready for reading (simply 321 Wait till the result filehandle becomes ready for reading (simply
184 does a select on the filehandle. This is useful if you want to 322 does a "select" on the filehandle. This is useful if you want to
185 synchronously wait for some requests to finish). 323 synchronously wait for some requests to finish).
186 324
187 See "nreqs" for an example. 325 See "nreqs" for an example.
188 326
189 IO::AIO::nreqs 327 IO::AIO::nreqs
190 Returns the number of requests currently outstanding. 328 Returns the number of requests currently outstanding (i.e. for which
329 their callback has not been invoked yet).
191 330
192 Example: wait till there are no outstanding requests anymore: 331 Example: wait till there are no outstanding requests anymore:
193 332
194 IO::AIO::poll_wait, IO::AIO::poll_cb 333 IO::AIO::poll_wait, IO::AIO::poll_cb
195 while IO::AIO::nreqs; 334 while IO::AIO::nreqs;
… …
209 348
210 IO::AIO::poll_wait, IO::AIO::poll_cb 349 IO::AIO::poll_wait, IO::AIO::poll_cb
211 if IO::AIO::nreqs; 350 if IO::AIO::nreqs;
212 351
213 IO::AIO::min_parallel $nthreads 352 IO::AIO::min_parallel $nthreads
214 Set the minimum number of AIO threads to $nthreads. The default is 353 Set the minimum number of AIO threads to $nthreads. The current
215 1, which means a single asynchronous operation can be done at one 354 default is 4, which means four asynchronous operations can be done
216 time (the number of outstanding operations, however, is unlimited). 355 at one time (the number of outstanding operations, however, is
356 unlimited).
357
358 IO::AIO starts threads only on demand, when an AIO request is queued
359 and no free thread exists.
217 360
218 It is recommended to keep the number of threads low, as some Linux 361 It is recommended to keep the number of threads low, as some Linux
219 kernel versions will scale negatively with the number of threads 362 kernel versions will scale negatively with the number of threads
220 (higher parallelity => MUCH higher latency). With current Linux 2.6 363 (higher parallelity => MUCH higher latency). With current Linux 2.6
221 versions, 4-32 threads should be fine. 364 versions, 4-32 threads should be fine.
222 365
223 Under normal circumstances you don't need to call this function, as 366 Under most circumstances you don't need to call this function, as
224 this module automatically starts some threads (the exact number 367 the module selects a default that is suitable for low to moderate
225 might change, and is currently 4). 368 load.
226 369
227 IO::AIO::max_parallel $nthreads 370 IO::AIO::max_parallel $nthreads
228 Sets the maximum number of AIO threads to $nthreads. If more than 371 Sets the maximum number of AIO threads to $nthreads. If more than
229 the specified number of threads are currently running, kill them. 372 the specified number of threads are currently running, this function
230 This function blocks until the limit is reached. 373 kills them. This function blocks until the limit is reached.
374
375 While $nthreads are zero, aio requests get queued but not executed
376 until the number of threads has been increased again.
231 377
232 This module automatically runs "max_parallel 0" at program end, to 378 This module automatically runs "max_parallel 0" at program end, to
233 ensure that all threads are killed and that there are no outstanding 379 ensure that all threads are killed and that there are no outstanding
234 requests. 380 requests.
235 381
… …
239 Sets the maximum number of outstanding requests to $nreqs. If you 385 Sets the maximum number of outstanding requests to $nreqs. If you
240 try to queue up more than this number of requests, the caller will 386 try to queue up more than this number of requests, the caller will
241 block until some requests have been handled. 387 block until some requests have been handled.
242 388
243 The default is very large, so normally there is no practical limit. 389 The default is very large, so normally there is no practical limit.
244 If you queue up many requests in a loop it it often improves speed 390 If you queue up many requests in a loop it often improves speed if
245 if you set this to a relatively low number, such as 100. 391 you set this to a relatively low number, such as 100.
246 392
247 Under normal circumstances you don't need to call this function. 393 Under normal circumstances you don't need to call this function.
394
395 FORK BEHAVIOUR
396 Before the fork, IO::AIO enters a quiescent state where no requests can
397 be added in other threads and no results will be processed. After the
398 fork the parent simply leaves the quiescent state and continues
399 request/result processing, while the child clears the request/result
400 queue (so the requests started before the fork will only be handled in
401 the parent). Threats will be started on demand until the limit ste in
402 the parent process has been reached again.
248 403
249SEE ALSO 404SEE ALSO
250 Coro, Linux::AIO. 405 Coro, Linux::AIO.
251 406
252AUTHOR 407AUTHOR

Diff Legend

– Removed lines
+ Added lines
< Changed lines
> Changed lines