1 |
root |
1.1 |
NAME |
2 |
|
|
IO::AIO - Asynchronous Input/Output |
3 |
|
|
|
4 |
|
|
SYNOPSIS |
5 |
|
|
use IO::AIO; |
6 |
|
|
|
7 |
root |
1.5 |
aio_open "/etc/passwd", O_RDONLY, 0, sub { |
8 |
|
|
my ($fh) = @_; |
9 |
|
|
... |
10 |
|
|
}; |
11 |
|
|
|
12 |
|
|
aio_unlink "/tmp/file", sub { }; |
13 |
|
|
|
14 |
|
|
aio_read $fh, 30000, 1024, $buffer, 0, sub { |
15 |
|
|
$_[0] > 0 or die "read error: $!"; |
16 |
|
|
}; |
17 |
|
|
|
18 |
root |
1.18 |
# version 2+ has request and group objects |
19 |
|
|
use IO::AIO 2; |
20 |
|
|
|
21 |
|
|
aioreq_pri 4; # give next request a very high priority |
22 |
|
|
my $req = aio_unlink "/tmp/file", sub { }; |
23 |
|
|
$req->cancel; # cancel request if still in queue |
24 |
|
|
|
25 |
|
|
my $grp = aio_group sub { print "all stats done\n" }; |
26 |
|
|
add $grp aio_stat "..." for ...; |
27 |
|
|
|
28 |
|
|
# AnyEvent integration |
29 |
root |
1.14 |
open my $fh, "<&=" . IO::AIO::poll_fileno or die "$!"; |
30 |
|
|
my $w = AnyEvent->io (fh => $fh, poll => 'r', cb => sub { IO::AIO::poll_cb }); |
31 |
|
|
|
32 |
root |
1.18 |
# Event integration |
33 |
root |
1.5 |
Event->io (fd => IO::AIO::poll_fileno, |
34 |
|
|
poll => 'r', |
35 |
|
|
cb => \&IO::AIO::poll_cb); |
36 |
|
|
|
37 |
root |
1.18 |
# Glib/Gtk2 integration |
38 |
root |
1.5 |
add_watch Glib::IO IO::AIO::poll_fileno, |
39 |
root |
1.7 |
in => sub { IO::AIO::poll_cb; 1 }; |
40 |
root |
1.5 |
|
41 |
root |
1.18 |
# Tk integration |
42 |
root |
1.5 |
Tk::Event::IO->fileevent (IO::AIO::poll_fileno, "", |
43 |
|
|
readable => \&IO::AIO::poll_cb); |
44 |
|
|
|
45 |
root |
1.18 |
# Danga::Socket integration |
46 |
root |
1.6 |
Danga::Socket->AddOtherFds (IO::AIO::poll_fileno => |
47 |
|
|
\&IO::AIO::poll_cb); |
48 |
|
|
|
49 |
root |
1.1 |
DESCRIPTION |
50 |
|
|
This module implements asynchronous I/O using whatever means your |
51 |
root |
1.2 |
operating system supports. |
52 |
root |
1.1 |
|
53 |
root |
1.19 |
Asynchronous means that operations that can normally block your program |
54 |
|
|
(e.g. reading from disk) will be done asynchronously: the operation will |
55 |
|
|
still block, but you can do something else in the meantime. This is |
56 |
|
|
extremely useful for programs that need to stay interactive even when |
57 |
|
|
doing heavy I/O (GUI programs, high performance network servers etc.), |
58 |
|
|
but can also be used to easily do operations in parallel that are |
59 |
|
|
normally done sequentially, e.g. stat'ing many files, which is much |
60 |
|
|
faster on a RAID volume or over NFS when you do a number of stat |
61 |
|
|
operations concurrently. |
62 |
|
|
|
63 |
|
|
While this works on all types of file descriptors (for example sockets), |
64 |
|
|
using these functions on file descriptors that support nonblocking |
65 |
|
|
operation (again, sockets, pipes etc.) is very inefficient. Use an event |
66 |
|
|
loop for that (such as the Event module): IO::AIO will naturally fit |
67 |
|
|
into such an event loop itself. |
68 |
|
|
|
69 |
root |
1.18 |
In this version, a number of threads are started that execute your |
70 |
|
|
requests and signal their completion. You don't need thread support in |
71 |
|
|
perl, and the threads created by this module will not be visible to |
72 |
|
|
perl. In the future, this module might make use of the native aio |
73 |
|
|
functions available on many operating systems. However, they are often |
74 |
root |
1.19 |
not well-supported or restricted (GNU/Linux doesn't allow them on normal |
75 |
root |
1.18 |
files currently, for example), and they would only support aio_read and |
76 |
root |
1.2 |
aio_write, so the remaining functionality would have to be implemented |
77 |
|
|
using threads anyway. |
78 |
root |
1.1 |
|
79 |
root |
1.18 |
Although the module will work with in the presence of other (Perl-) |
80 |
|
|
threads, it is currently not reentrant in any way, so use appropriate |
81 |
|
|
locking yourself, always call "poll_cb" from within the same thread, or |
82 |
|
|
never call "poll_cb" (or other "aio_" functions) recursively. |
83 |
|
|
|
84 |
root |
1.19 |
EXAMPLE |
85 |
|
|
This is a simple example that uses the Event module and loads |
86 |
|
|
/etc/passwd asynchronously: |
87 |
|
|
|
88 |
|
|
use Fcntl; |
89 |
|
|
use Event; |
90 |
|
|
use IO::AIO; |
91 |
|
|
|
92 |
|
|
# register the IO::AIO callback with Event |
93 |
|
|
Event->io (fd => IO::AIO::poll_fileno, |
94 |
|
|
poll => 'r', |
95 |
|
|
cb => \&IO::AIO::poll_cb); |
96 |
|
|
|
97 |
|
|
# queue the request to open /etc/passwd |
98 |
|
|
aio_open "/etc/passwd", O_RDONLY, 0, sub { |
99 |
|
|
my $fh = $_[0] |
100 |
|
|
or die "error while opening: $!"; |
101 |
|
|
|
102 |
|
|
# stat'ing filehandles is generally non-blocking |
103 |
|
|
my $size = -s $fh; |
104 |
|
|
|
105 |
|
|
# queue a request to read the file |
106 |
|
|
my $contents; |
107 |
|
|
aio_read $fh, 0, $size, $contents, 0, sub { |
108 |
|
|
$_[0] == $size |
109 |
|
|
or die "short read: $!"; |
110 |
|
|
|
111 |
|
|
close $fh; |
112 |
|
|
|
113 |
|
|
# file contents now in $contents |
114 |
|
|
print $contents; |
115 |
|
|
|
116 |
|
|
# exit event loop and program |
117 |
|
|
Event::unloop; |
118 |
|
|
}; |
119 |
|
|
}; |
120 |
|
|
|
121 |
|
|
# possibly queue up other requests, or open GUI windows, |
122 |
|
|
# check for sockets etc. etc. |
123 |
|
|
|
124 |
|
|
# process events as long as there are some: |
125 |
|
|
Event::loop; |
126 |
|
|
|
127 |
root |
1.18 |
REQUEST ANATOMY AND LIFETIME |
128 |
|
|
Every "aio_*" function creates a request. which is a C data structure |
129 |
|
|
not directly visible to Perl. |
130 |
|
|
|
131 |
|
|
If called in non-void context, every request function returns a Perl |
132 |
|
|
object representing the request. In void context, nothing is returned, |
133 |
|
|
which saves a bit of memory. |
134 |
|
|
|
135 |
|
|
The perl object is a fairly standard ref-to-hash object. The hash |
136 |
|
|
contents are not used by IO::AIO so you are free to store anything you |
137 |
|
|
like in it. |
138 |
|
|
|
139 |
|
|
During their existance, aio requests travel through the following |
140 |
|
|
states, in order: |
141 |
|
|
|
142 |
|
|
ready |
143 |
|
|
Immediately after a request is created it is put into the ready |
144 |
|
|
state, waiting for a thread to execute it. |
145 |
|
|
|
146 |
|
|
execute |
147 |
|
|
A thread has accepted the request for processing and is currently |
148 |
|
|
executing it (e.g. blocking in read). |
149 |
|
|
|
150 |
|
|
pending |
151 |
|
|
The request has been executed and is waiting for result processing. |
152 |
|
|
|
153 |
|
|
While request submission and execution is fully asynchronous, result |
154 |
|
|
processing is not and relies on the perl interpreter calling |
155 |
|
|
"poll_cb" (or another function with the same effect). |
156 |
|
|
|
157 |
|
|
result |
158 |
|
|
The request results are processed synchronously by "poll_cb". |
159 |
|
|
|
160 |
|
|
The "poll_cb" function will process all outstanding aio requests by |
161 |
|
|
calling their callbacks, freeing memory associated with them and |
162 |
|
|
managing any groups they are contained in. |
163 |
|
|
|
164 |
|
|
done |
165 |
|
|
Request has reached the end of its lifetime and holds no resources |
166 |
|
|
anymore (except possibly for the Perl object, but its connection to |
167 |
|
|
the actual aio request is severed and calling its methods will |
168 |
|
|
either do nothing or result in a runtime error). |
169 |
root |
1.1 |
|
170 |
root |
1.4 |
FUNCTIONS |
171 |
root |
1.19 |
AIO REQUEST FUNCTIONS |
172 |
root |
1.18 |
All the "aio_*" calls are more or less thin wrappers around the |
173 |
|
|
syscall with the same name (sans "aio_"). The arguments are similar |
174 |
|
|
or identical, and they all accept an additional (and optional) |
175 |
|
|
$callback argument which must be a code reference. This code |
176 |
|
|
reference will get called with the syscall return code (e.g. most |
177 |
|
|
syscalls return -1 on error, unlike perl, which usually delivers |
178 |
|
|
"false") as it's sole argument when the given syscall has been |
179 |
|
|
executed asynchronously. |
180 |
|
|
|
181 |
|
|
All functions expecting a filehandle keep a copy of the filehandle |
182 |
|
|
internally until the request has finished. |
183 |
|
|
|
184 |
root |
1.19 |
All functions return request objects of type IO::AIO::REQ that allow |
185 |
|
|
further manipulation of those requests while they are in-flight. |
186 |
root |
1.18 |
|
187 |
|
|
The pathnames you pass to these routines *must* be absolute and |
188 |
root |
1.19 |
encoded as octets. The reason for the former is that at the time the |
189 |
|
|
request is being executed, the current working directory could have |
190 |
|
|
changed. Alternatively, you can make sure that you never change the |
191 |
|
|
current working directory anywhere in the program and then use |
192 |
|
|
relative paths. |
193 |
root |
1.18 |
|
194 |
root |
1.19 |
To encode pathnames as octets, either make sure you either: a) |
195 |
root |
1.18 |
always pass in filenames you got from outside (command line, readdir |
196 |
root |
1.19 |
etc.) without tinkering, b) are ASCII or ISO 8859-1, c) use the |
197 |
|
|
Encode module and encode your pathnames to the locale (or other) |
198 |
|
|
encoding in effect in the user environment, d) use |
199 |
|
|
Glib::filename_from_unicode on unicode filenames or e) use something |
200 |
|
|
else to ensure your scalar has the correct contents. |
201 |
|
|
|
202 |
|
|
This works, btw. independent of the internal UTF-8 bit, which |
203 |
|
|
IO::AIO handles correctly wether it is set or not. |
204 |
root |
1.18 |
|
205 |
|
|
$prev_pri = aioreq_pri [$pri] |
206 |
|
|
Returns the priority value that would be used for the next |
207 |
|
|
request and, if $pri is given, sets the priority for the next |
208 |
|
|
aio request. |
209 |
|
|
|
210 |
|
|
The default priority is 0, the minimum and maximum priorities |
211 |
|
|
are -4 and 4, respectively. Requests with higher priority will |
212 |
|
|
be serviced first. |
213 |
|
|
|
214 |
|
|
The priority will be reset to 0 after each call to one of the |
215 |
|
|
"aio_*" functions. |
216 |
|
|
|
217 |
|
|
Example: open a file with low priority, then read something from |
218 |
|
|
it with higher priority so the read request is serviced before |
219 |
|
|
other low priority open requests (potentially spamming the |
220 |
|
|
cache): |
221 |
|
|
|
222 |
|
|
aioreq_pri -3; |
223 |
|
|
aio_open ..., sub { |
224 |
|
|
return unless $_[0]; |
225 |
|
|
|
226 |
|
|
aioreq_pri -2; |
227 |
|
|
aio_read $_[0], ..., sub { |
228 |
|
|
... |
229 |
|
|
}; |
230 |
|
|
}; |
231 |
|
|
|
232 |
|
|
aioreq_nice $pri_adjust |
233 |
|
|
Similar to "aioreq_pri", but subtracts the given value from the |
234 |
root |
1.19 |
current priority, so the effect is cumulative. |
235 |
root |
1.18 |
|
236 |
|
|
aio_open $pathname, $flags, $mode, $callback->($fh) |
237 |
|
|
Asynchronously open or create a file and call the callback with |
238 |
|
|
a newly created filehandle for the file. |
239 |
|
|
|
240 |
|
|
The pathname passed to "aio_open" must be absolute. See API |
241 |
|
|
NOTES, above, for an explanation. |
242 |
|
|
|
243 |
|
|
The $flags argument is a bitmask. See the "Fcntl" module for a |
244 |
|
|
list. They are the same as used by "sysopen". |
245 |
|
|
|
246 |
|
|
Likewise, $mode specifies the mode of the newly created file, if |
247 |
|
|
it didn't exist and "O_CREAT" has been given, just like perl's |
248 |
|
|
"sysopen", except that it is mandatory (i.e. use 0 if you don't |
249 |
|
|
create new files, and 0666 or 0777 if you do). |
250 |
|
|
|
251 |
|
|
Example: |
252 |
|
|
|
253 |
|
|
aio_open "/etc/passwd", O_RDONLY, 0, sub { |
254 |
|
|
if ($_[0]) { |
255 |
|
|
print "open successful, fh is $_[0]\n"; |
256 |
|
|
... |
257 |
|
|
} else { |
258 |
|
|
die "open failed: $!\n"; |
259 |
|
|
} |
260 |
|
|
}; |
261 |
|
|
|
262 |
|
|
aio_close $fh, $callback->($status) |
263 |
|
|
Asynchronously close a file and call the callback with the |
264 |
|
|
result code. *WARNING:* although accepted, you should not pass |
265 |
|
|
in a perl filehandle here, as perl will likely close the file |
266 |
|
|
descriptor another time when the filehandle is destroyed. |
267 |
|
|
Normally, you can safely call perls "close" or just let |
268 |
|
|
filehandles go out of scope. |
269 |
|
|
|
270 |
|
|
This is supposed to be a bug in the API, so that might change. |
271 |
|
|
It's therefore best to avoid this function. |
272 |
|
|
|
273 |
|
|
aio_read $fh,$offset,$length, $data,$dataoffset, |
274 |
|
|
$callback->($retval) |
275 |
|
|
aio_write $fh,$offset,$length, $data,$dataoffset, |
276 |
|
|
$callback->($retval) |
277 |
|
|
Reads or writes "length" bytes from the specified "fh" and |
278 |
|
|
"offset" into the scalar given by "data" and offset "dataoffset" |
279 |
|
|
and calls the callback without the actual number of bytes read |
280 |
|
|
(or -1 on error, just like the syscall). |
281 |
|
|
|
282 |
|
|
The $data scalar *MUST NOT* be modified in any way while the |
283 |
|
|
request is outstanding. Modifying it can result in segfaults or |
284 |
|
|
WW3 (if the necessary/optional hardware is installed). |
285 |
|
|
|
286 |
|
|
Example: Read 15 bytes at offset 7 into scalar $buffer, starting |
287 |
|
|
at offset 0 within the scalar: |
288 |
|
|
|
289 |
|
|
aio_read $fh, 7, 15, $buffer, 0, sub { |
290 |
|
|
$_[0] > 0 or die "read error: $!"; |
291 |
|
|
print "read $_[0] bytes: <$buffer>\n"; |
292 |
|
|
}; |
293 |
|
|
|
294 |
|
|
aio_sendfile $out_fh, $in_fh, $in_offset, $length, |
295 |
|
|
$callback->($retval) |
296 |
|
|
Tries to copy $length bytes from $in_fh to $out_fh. It starts |
297 |
|
|
reading at byte offset $in_offset, and starts writing at the |
298 |
|
|
current file offset of $out_fh. Because of that, it is not safe |
299 |
|
|
to issue more than one "aio_sendfile" per $out_fh, as they will |
300 |
|
|
interfere with each other. |
301 |
|
|
|
302 |
|
|
This call tries to make use of a native "sendfile" syscall to |
303 |
|
|
provide zero-copy operation. For this to work, $out_fh should |
304 |
|
|
refer to a socket, and $in_fh should refer to mmap'able file. |
305 |
|
|
|
306 |
|
|
If the native sendfile call fails or is not implemented, it will |
307 |
|
|
be emulated, so you can call "aio_sendfile" on any type of |
308 |
|
|
filehandle regardless of the limitations of the operating |
309 |
|
|
system. |
310 |
|
|
|
311 |
|
|
Please note, however, that "aio_sendfile" can read more bytes |
312 |
|
|
from $in_fh than are written, and there is no way to find out |
313 |
|
|
how many bytes have been read from "aio_sendfile" alone, as |
314 |
|
|
"aio_sendfile" only provides the number of bytes written to |
315 |
|
|
$out_fh. Only if the result value equals $length one can assume |
316 |
|
|
that $length bytes have been read. |
317 |
|
|
|
318 |
|
|
aio_readahead $fh,$offset,$length, $callback->($retval) |
319 |
|
|
"aio_readahead" populates the page cache with data from a file |
320 |
|
|
so that subsequent reads from that file will not block on disk |
321 |
|
|
I/O. The $offset argument specifies the starting point from |
322 |
|
|
which data is to be read and $length specifies the number of |
323 |
|
|
bytes to be read. I/O is performed in whole pages, so that |
324 |
|
|
offset is effectively rounded down to a page boundary and bytes |
325 |
|
|
are read up to the next page boundary greater than or equal to |
326 |
|
|
(off-set+length). "aio_readahead" does not read beyond the end |
327 |
|
|
of the file. The current file offset of the file is left |
328 |
|
|
unchanged. |
329 |
|
|
|
330 |
|
|
If that syscall doesn't exist (likely if your OS isn't Linux) it |
331 |
|
|
will be emulated by simply reading the data, which would have a |
332 |
|
|
similar effect. |
333 |
|
|
|
334 |
|
|
aio_stat $fh_or_path, $callback->($status) |
335 |
|
|
aio_lstat $fh, $callback->($status) |
336 |
|
|
Works like perl's "stat" or "lstat" in void context. The |
337 |
|
|
callback will be called after the stat and the results will be |
338 |
|
|
available using "stat _" or "-s _" etc... |
339 |
|
|
|
340 |
|
|
The pathname passed to "aio_stat" must be absolute. See API |
341 |
|
|
NOTES, above, for an explanation. |
342 |
|
|
|
343 |
|
|
Currently, the stats are always 64-bit-stats, i.e. instead of |
344 |
|
|
returning an error when stat'ing a large file, the results will |
345 |
|
|
be silently truncated unless perl itself is compiled with large |
346 |
|
|
file support. |
347 |
|
|
|
348 |
|
|
Example: Print the length of /etc/passwd: |
349 |
|
|
|
350 |
|
|
aio_stat "/etc/passwd", sub { |
351 |
|
|
$_[0] and die "stat failed: $!"; |
352 |
|
|
print "size is ", -s _, "\n"; |
353 |
|
|
}; |
354 |
|
|
|
355 |
|
|
aio_unlink $pathname, $callback->($status) |
356 |
|
|
Asynchronously unlink (delete) a file and call the callback with |
357 |
|
|
the result code. |
358 |
|
|
|
359 |
root |
1.19 |
aio_mknod $path, $mode, $dev, $callback->($status) |
360 |
|
|
[EXPERIMENTAL] |
361 |
|
|
|
362 |
|
|
Asynchronously create a device node (or fifo). See mknod(2). |
363 |
|
|
|
364 |
|
|
The only (POSIX-) portable way of calling this function is: |
365 |
|
|
|
366 |
|
|
aio_mknod $path, IO::AIO::S_IFIFO | $mode, 0, sub { ... |
367 |
|
|
|
368 |
root |
1.18 |
aio_link $srcpath, $dstpath, $callback->($status) |
369 |
|
|
Asynchronously create a new link to the existing object at |
370 |
|
|
$srcpath at the path $dstpath and call the callback with the |
371 |
|
|
result code. |
372 |
|
|
|
373 |
|
|
aio_symlink $srcpath, $dstpath, $callback->($status) |
374 |
|
|
Asynchronously create a new symbolic link to the existing object |
375 |
|
|
at $srcpath at the path $dstpath and call the callback with the |
376 |
|
|
result code. |
377 |
|
|
|
378 |
|
|
aio_rename $srcpath, $dstpath, $callback->($status) |
379 |
|
|
Asynchronously rename the object at $srcpath to $dstpath, just |
380 |
|
|
as rename(2) and call the callback with the result code. |
381 |
|
|
|
382 |
|
|
aio_rmdir $pathname, $callback->($status) |
383 |
|
|
Asynchronously rmdir (delete) a directory and call the callback |
384 |
|
|
with the result code. |
385 |
|
|
|
386 |
|
|
aio_readdir $pathname, $callback->($entries) |
387 |
|
|
Unlike the POSIX call of the same name, "aio_readdir" reads an |
388 |
|
|
entire directory (i.e. opendir + readdir + closedir). The |
389 |
|
|
entries will not be sorted, and will NOT include the "." and |
390 |
|
|
".." entries. |
391 |
|
|
|
392 |
|
|
The callback a single argument which is either "undef" or an |
393 |
|
|
array-ref with the filenames. |
394 |
|
|
|
395 |
root |
1.19 |
aio_copy $srcpath, $dstpath, $callback->($status) |
396 |
|
|
Try to copy the *file* (directories not supported as either |
397 |
|
|
source or destination) from $srcpath to $dstpath and call the |
398 |
|
|
callback with the 0 (error) or -1 ok. |
399 |
|
|
|
400 |
|
|
This is a composite request that it creates the destination file |
401 |
|
|
with mode 0200 and copies the contents of the source file into |
402 |
|
|
it using "aio_sendfile", followed by restoring atime, mtime, |
403 |
|
|
access mode and uid/gid, in that order. |
404 |
|
|
|
405 |
|
|
If an error occurs, the partial destination file will be |
406 |
|
|
unlinked, if possible, except when setting atime, mtime, access |
407 |
|
|
mode and uid/gid, where errors are being ignored. |
408 |
|
|
|
409 |
|
|
aio_move $srcpath, $dstpath, $callback->($status) |
410 |
|
|
Try to move the *file* (directories not supported as either |
411 |
|
|
source or destination) from $srcpath to $dstpath and call the |
412 |
|
|
callback with the 0 (error) or -1 ok. |
413 |
|
|
|
414 |
|
|
This is a composite request that tries to rename(2) the file |
415 |
|
|
first. If rename files with "EXDEV", it copies the file with |
416 |
|
|
"aio_copy" and, if that is successful, unlinking the $srcpath. |
417 |
|
|
|
418 |
root |
1.18 |
aio_scandir $path, $maxreq, $callback->($dirs, $nondirs) |
419 |
|
|
Scans a directory (similar to "aio_readdir") but additionally |
420 |
|
|
tries to efficiently separate the entries of directory $path |
421 |
|
|
into two sets of names, directories you can recurse into |
422 |
|
|
(directories), and ones you cannot recurse into (everything |
423 |
|
|
else, including symlinks to directories). |
424 |
|
|
|
425 |
|
|
"aio_scandir" is a composite request that creates of many sub |
426 |
|
|
requests_ $maxreq specifies the maximum number of outstanding |
427 |
|
|
aio requests that this function generates. If it is "<= 0", then |
428 |
root |
1.19 |
a suitable default will be chosen (currently 4). |
429 |
root |
1.18 |
|
430 |
|
|
On error, the callback is called without arguments, otherwise it |
431 |
|
|
receives two array-refs with path-relative entry names. |
432 |
|
|
|
433 |
|
|
Example: |
434 |
|
|
|
435 |
|
|
aio_scandir $dir, 0, sub { |
436 |
|
|
my ($dirs, $nondirs) = @_; |
437 |
|
|
print "real directories: @$dirs\n"; |
438 |
|
|
print "everything else: @$nondirs\n"; |
439 |
|
|
}; |
440 |
|
|
|
441 |
|
|
Implementation notes. |
442 |
|
|
|
443 |
|
|
The "aio_readdir" cannot be avoided, but "stat()"'ing every |
444 |
|
|
entry can. |
445 |
|
|
|
446 |
|
|
After reading the directory, the modification time, size etc. of |
447 |
|
|
the directory before and after the readdir is checked, and if |
448 |
|
|
they match (and isn't the current time), the link count will be |
449 |
|
|
used to decide how many entries are directories (if >= 2). |
450 |
|
|
Otherwise, no knowledge of the number of subdirectories will be |
451 |
|
|
assumed. |
452 |
|
|
|
453 |
|
|
Then entries will be sorted into likely directories (everything |
454 |
|
|
without a non-initial dot currently) and likely non-directories |
455 |
|
|
(everything else). Then every entry plus an appended "/." will |
456 |
|
|
be "stat"'ed, likely directories first. If that succeeds, it |
457 |
|
|
assumes that the entry is a directory or a symlink to directory |
458 |
|
|
(which will be checked seperately). This is often faster than |
459 |
|
|
stat'ing the entry itself because filesystems might detect the |
460 |
|
|
type of the entry without reading the inode data (e.g. ext2fs |
461 |
|
|
filetype feature). |
462 |
|
|
|
463 |
|
|
If the known number of directories (link count - 2) has been |
464 |
|
|
reached, the rest of the entries is assumed to be |
465 |
|
|
non-directories. |
466 |
|
|
|
467 |
|
|
This only works with certainty on POSIX (= UNIX) filesystems, |
468 |
|
|
which fortunately are the vast majority of filesystems around. |
469 |
|
|
|
470 |
|
|
It will also likely work on non-POSIX filesystems with reduced |
471 |
|
|
efficiency as those tend to return 0 or 1 as link counts, which |
472 |
|
|
disables the directory counting heuristic. |
473 |
|
|
|
474 |
|
|
aio_fsync $fh, $callback->($status) |
475 |
|
|
Asynchronously call fsync on the given filehandle and call the |
476 |
|
|
callback with the fsync result code. |
477 |
|
|
|
478 |
|
|
aio_fdatasync $fh, $callback->($status) |
479 |
|
|
Asynchronously call fdatasync on the given filehandle and call |
480 |
|
|
the callback with the fdatasync result code. |
481 |
|
|
|
482 |
|
|
If this call isn't available because your OS lacks it or it |
483 |
|
|
couldn't be detected, it will be emulated by calling "fsync" |
484 |
|
|
instead. |
485 |
|
|
|
486 |
|
|
aio_group $callback->(...) |
487 |
|
|
This is a very special aio request: Instead of doing something, |
488 |
|
|
it is a container for other aio requests, which is useful if you |
489 |
|
|
want to bundle many requests into a single, composite, request |
490 |
|
|
with a definite callback and the ability to cancel the whole |
491 |
|
|
request with its subrequests. |
492 |
|
|
|
493 |
|
|
Returns an object of class IO::AIO::GRP. See its documentation |
494 |
|
|
below for more info. |
495 |
|
|
|
496 |
|
|
Example: |
497 |
|
|
|
498 |
|
|
my $grp = aio_group sub { |
499 |
|
|
print "all stats done\n"; |
500 |
|
|
}; |
501 |
|
|
|
502 |
|
|
add $grp |
503 |
|
|
(aio_stat ...), |
504 |
|
|
(aio_stat ...), |
505 |
|
|
...; |
506 |
|
|
|
507 |
|
|
aio_nop $callback->() |
508 |
|
|
This is a special request - it does nothing in itself and is |
509 |
|
|
only used for side effects, such as when you want to add a dummy |
510 |
|
|
request to a group so that finishing the requests in the group |
511 |
|
|
depends on executing the given code. |
512 |
|
|
|
513 |
|
|
While this request does nothing, it still goes through the |
514 |
|
|
execution phase and still requires a worker thread. Thus, the |
515 |
|
|
callback will not be executed immediately but only after other |
516 |
|
|
requests in the queue have entered their execution phase. This |
517 |
|
|
can be used to measure request latency. |
518 |
|
|
|
519 |
|
|
IO::AIO::aio_busy $fractional_seconds, $callback->() *NOT EXPORTED* |
520 |
|
|
Mainly used for debugging and benchmarking, this aio request |
521 |
|
|
puts one of the request workers to sleep for the given time. |
522 |
|
|
|
523 |
|
|
While it is theoretically handy to have simple I/O scheduling |
524 |
|
|
requests like sleep and file handle readable/writable, the |
525 |
|
|
overhead this creates is immense (it blocks a thread for a long |
526 |
|
|
time) so do not use this function except to put your application |
527 |
|
|
under artificial I/O pressure. |
528 |
|
|
|
529 |
|
|
IO::AIO::REQ CLASS |
530 |
|
|
All non-aggregate "aio_*" functions return an object of this class |
531 |
|
|
when called in non-void context. |
532 |
|
|
|
533 |
|
|
cancel $req |
534 |
|
|
Cancels the request, if possible. Has the effect of skipping |
535 |
|
|
execution when entering the execute state and skipping calling |
536 |
|
|
the callback when entering the the result state, but will leave |
537 |
|
|
the request otherwise untouched. That means that requests that |
538 |
|
|
currently execute will not be stopped and resources held by the |
539 |
|
|
request will not be freed prematurely. |
540 |
|
|
|
541 |
|
|
cb $req $callback->(...) |
542 |
|
|
Replace (or simply set) the callback registered to the request. |
543 |
|
|
|
544 |
|
|
IO::AIO::GRP CLASS |
545 |
|
|
This class is a subclass of IO::AIO::REQ, so all its methods apply |
546 |
|
|
to objects of this class, too. |
547 |
|
|
|
548 |
|
|
A IO::AIO::GRP object is a special request that can contain multiple |
549 |
|
|
other aio requests. |
550 |
|
|
|
551 |
|
|
You create one by calling the "aio_group" constructing function with |
552 |
|
|
a callback that will be called when all contained requests have |
553 |
|
|
entered the "done" state: |
554 |
|
|
|
555 |
|
|
my $grp = aio_group sub { |
556 |
|
|
print "all requests are done\n"; |
557 |
root |
1.1 |
}; |
558 |
|
|
|
559 |
root |
1.18 |
You add requests by calling the "add" method with one or more |
560 |
|
|
"IO::AIO::REQ" objects: |
561 |
|
|
|
562 |
|
|
$grp->add (aio_unlink "..."); |
563 |
|
|
|
564 |
|
|
add $grp aio_stat "...", sub { |
565 |
|
|
$_[0] or return $grp->result ("error"); |
566 |
|
|
|
567 |
|
|
# add another request dynamically, if first succeeded |
568 |
|
|
add $grp aio_open "...", sub { |
569 |
|
|
$grp->result ("ok"); |
570 |
|
|
}; |
571 |
root |
1.1 |
}; |
572 |
|
|
|
573 |
root |
1.18 |
This makes it very easy to create composite requests (see the source |
574 |
|
|
of "aio_move" for an application) that work and feel like simple |
575 |
|
|
requests. |
576 |
|
|
|
577 |
|
|
* The IO::AIO::GRP objects will be cleaned up during calls to |
578 |
|
|
"IO::AIO::poll_cb", just like any other request. |
579 |
|
|
* They can be canceled like any other request. Canceling will cancel |
580 |
|
|
not only the request itself, but also all requests it contains. |
581 |
|
|
* They can also can also be added to other IO::AIO::GRP objects. |
582 |
|
|
* You must not add requests to a group from within the group |
583 |
|
|
callback (or any later time). |
584 |
|
|
|
585 |
|
|
Their lifetime, simplified, looks like this: when they are empty, |
586 |
|
|
they will finish very quickly. If they contain only requests that |
587 |
|
|
are in the "done" state, they will also finish. Otherwise they will |
588 |
|
|
continue to exist. |
589 |
|
|
|
590 |
|
|
That means after creating a group you have some time to add |
591 |
|
|
requests. And in the callbacks of those requests, you can add |
592 |
|
|
further requests to the group. And only when all those requests have |
593 |
|
|
finished will the the group itself finish. |
594 |
|
|
|
595 |
|
|
add $grp ... |
596 |
|
|
$grp->add (...) |
597 |
|
|
Add one or more requests to the group. Any type of IO::AIO::REQ |
598 |
|
|
can be added, including other groups, as long as you do not |
599 |
|
|
create circular dependencies. |
600 |
|
|
|
601 |
|
|
Returns all its arguments. |
602 |
|
|
|
603 |
|
|
$grp->cancel_subs |
604 |
|
|
Cancel all subrequests and clears any feeder, but not the group |
605 |
|
|
request itself. Useful when you queued a lot of events but got a |
606 |
|
|
result early. |
607 |
|
|
|
608 |
|
|
$grp->result (...) |
609 |
|
|
Set the result value(s) that will be passed to the group |
610 |
|
|
callback when all subrequests have finished and set thre groups |
611 |
|
|
errno to the current value of errno (just like calling "errno" |
612 |
|
|
without an error number). By default, no argument will be passed |
613 |
|
|
and errno is zero. |
614 |
|
|
|
615 |
|
|
$grp->errno ([$errno]) |
616 |
|
|
Sets the group errno value to $errno, or the current value of |
617 |
|
|
errno when the argument is missing. |
618 |
|
|
|
619 |
|
|
Every aio request has an associated errno value that is restored |
620 |
|
|
when the callback is invoked. This method lets you change this |
621 |
|
|
value from its default (0). |
622 |
|
|
|
623 |
|
|
Calling "result" will also set errno, so make sure you either |
624 |
|
|
set $! before the call to "result", or call c<errno> after it. |
625 |
|
|
|
626 |
|
|
feed $grp $callback->($grp) |
627 |
|
|
Sets a feeder/generator on this group: every group can have an |
628 |
|
|
attached generator that generates requests if idle. The idea |
629 |
|
|
behind this is that, although you could just queue as many |
630 |
|
|
requests as you want in a group, this might starve other |
631 |
|
|
requests for a potentially long time. For example, "aio_scandir" |
632 |
|
|
might generate hundreds of thousands "aio_stat" requests, |
633 |
|
|
delaying any later requests for a long time. |
634 |
|
|
|
635 |
|
|
To avoid this, and allow incremental generation of requests, you |
636 |
|
|
can instead a group and set a feeder on it that generates those |
637 |
|
|
requests. The feed callback will be called whenever there are |
638 |
|
|
few enough (see "limit", below) requests active in the group |
639 |
|
|
itself and is expected to queue more requests. |
640 |
|
|
|
641 |
|
|
The feed callback can queue as many requests as it likes (i.e. |
642 |
|
|
"add" does not impose any limits). |
643 |
|
|
|
644 |
|
|
If the feed does not queue more requests when called, it will be |
645 |
|
|
automatically removed from the group. |
646 |
|
|
|
647 |
|
|
If the feed limit is 0, it will be set to 2 automatically. |
648 |
|
|
|
649 |
|
|
Example: |
650 |
|
|
|
651 |
|
|
# stat all files in @files, but only ever use four aio requests concurrently: |
652 |
|
|
|
653 |
|
|
my $grp = aio_group sub { print "finished\n" }; |
654 |
|
|
limit $grp 4; |
655 |
|
|
feed $grp sub { |
656 |
|
|
my $file = pop @files |
657 |
|
|
or return; |
658 |
|
|
|
659 |
|
|
add $grp aio_stat $file, sub { ... }; |
660 |
|
|
}; |
661 |
|
|
|
662 |
|
|
limit $grp $num |
663 |
|
|
Sets the feeder limit for the group: The feeder will be called |
664 |
|
|
whenever the group contains less than this many requests. |
665 |
|
|
|
666 |
|
|
Setting the limit to 0 will pause the feeding process. |
667 |
root |
1.17 |
|
668 |
root |
1.18 |
SUPPORT FUNCTIONS |
669 |
root |
1.19 |
EVENT PROCESSING AND EVENT LOOP INTEGRATION |
670 |
root |
1.18 |
$fileno = IO::AIO::poll_fileno |
671 |
|
|
Return the *request result pipe file descriptor*. This |
672 |
|
|
filehandle must be polled for reading by some mechanism outside |
673 |
|
|
this module (e.g. Event or select, see below or the SYNOPSIS). |
674 |
|
|
If the pipe becomes readable you have to call "poll_cb" to check |
675 |
|
|
the results. |
676 |
|
|
|
677 |
|
|
See "poll_cb" for an example. |
678 |
|
|
|
679 |
|
|
IO::AIO::poll_cb |
680 |
root |
1.19 |
Process some outstanding events on the result pipe. You have to |
681 |
root |
1.18 |
call this regularly. Returns the number of events processed. |
682 |
root |
1.19 |
Returns immediately when no events are outstanding. The amount |
683 |
|
|
of events processed depends on the settings of |
684 |
|
|
"IO::AIO::max_poll_req" and "IO::AIO::max_poll_time". |
685 |
root |
1.18 |
|
686 |
|
|
If not all requests were processed for whatever reason, the |
687 |
|
|
filehandle will still be ready when "poll_cb" returns. |
688 |
root |
1.1 |
|
689 |
root |
1.18 |
Example: Install an Event watcher that automatically calls |
690 |
|
|
IO::AIO::poll_cb with high priority: |
691 |
root |
1.13 |
|
692 |
root |
1.18 |
Event->io (fd => IO::AIO::poll_fileno, |
693 |
|
|
poll => 'r', async => 1, |
694 |
|
|
cb => \&IO::AIO::poll_cb); |
695 |
root |
1.13 |
|
696 |
root |
1.19 |
IO::AIO::max_poll_reqs $nreqs |
697 |
|
|
IO::AIO::max_poll_time $seconds |
698 |
|
|
These set the maximum number of requests (default 0, meaning |
699 |
|
|
infinity) that are being processed by "IO::AIO::poll_cb" in one |
700 |
|
|
call, respectively the maximum amount of time (default 0, |
701 |
|
|
meaning infinity) spent in "IO::AIO::poll_cb" to process |
702 |
|
|
requests (more correctly the mininum amount of time "poll_cb" is |
703 |
|
|
allowed to use). |
704 |
|
|
|
705 |
|
|
Setting these is useful if you want to ensure some level of |
706 |
|
|
interactiveness when perl is not fast enough to process all |
707 |
|
|
requests in time. |
708 |
root |
1.13 |
|
709 |
root |
1.19 |
For interactive programs, values such as 0.01 to 0.1 should be |
710 |
|
|
fine. |
711 |
root |
1.9 |
|
712 |
root |
1.18 |
Example: Install an Event watcher that automatically calls |
713 |
|
|
IO::AIO::poll_some with low priority, to ensure that other parts |
714 |
|
|
of the program get the CPU sometimes even under high AIO load. |
715 |
root |
1.1 |
|
716 |
root |
1.19 |
# try not to spend much more than 0.1s in poll_cb |
717 |
|
|
IO::AIO::max_poll_time 0.1; |
718 |
|
|
|
719 |
|
|
# use a low priority so other tasks have priority |
720 |
root |
1.18 |
Event->io (fd => IO::AIO::poll_fileno, |
721 |
|
|
poll => 'r', nice => 1, |
722 |
root |
1.19 |
cb => &IO::AIO::poll_cb); |
723 |
root |
1.4 |
|
724 |
root |
1.18 |
IO::AIO::poll_wait |
725 |
|
|
Wait till the result filehandle becomes ready for reading |
726 |
|
|
(simply does a "select" on the filehandle. This is useful if you |
727 |
|
|
want to synchronously wait for some requests to finish). |
728 |
root |
1.4 |
|
729 |
root |
1.18 |
See "nreqs" for an example. |
730 |
root |
1.4 |
|
731 |
root |
1.19 |
IO::AIO::poll |
732 |
|
|
Waits until some requests have been handled. |
733 |
root |
1.4 |
|
734 |
root |
1.19 |
Strictly equivalent to: |
735 |
root |
1.4 |
|
736 |
root |
1.18 |
IO::AIO::poll_wait, IO::AIO::poll_cb |
737 |
root |
1.19 |
if IO::AIO::nreqs; |
738 |
root |
1.4 |
|
739 |
root |
1.18 |
IO::AIO::flush |
740 |
|
|
Wait till all outstanding AIO requests have been handled. |
741 |
root |
1.4 |
|
742 |
root |
1.18 |
Strictly equivalent to: |
743 |
root |
1.4 |
|
744 |
root |
1.18 |
IO::AIO::poll_wait, IO::AIO::poll_cb |
745 |
|
|
while IO::AIO::nreqs; |
746 |
root |
1.6 |
|
747 |
root |
1.19 |
CONTROLLING THE NUMBER OF THREADS |
748 |
root |
1.18 |
IO::AIO::min_parallel $nthreads |
749 |
|
|
Set the minimum number of AIO threads to $nthreads. The current |
750 |
|
|
default is 8, which means eight asynchronous operations can |
751 |
|
|
execute concurrently at any one time (the number of outstanding |
752 |
|
|
requests, however, is unlimited). |
753 |
root |
1.6 |
|
754 |
root |
1.18 |
IO::AIO starts threads only on demand, when an AIO request is |
755 |
root |
1.19 |
queued and no free thread exists. Please note that queueing up a |
756 |
|
|
hundred requests can create demand for a hundred threads, even |
757 |
|
|
if it turns out that everything is in the cache and could have |
758 |
|
|
been processed faster by a single thread. |
759 |
root |
1.6 |
|
760 |
root |
1.18 |
It is recommended to keep the number of threads relatively low, |
761 |
|
|
as some Linux kernel versions will scale negatively with the |
762 |
|
|
number of threads (higher parallelity => MUCH higher latency). |
763 |
|
|
With current Linux 2.6 versions, 4-32 threads should be fine. |
764 |
root |
1.11 |
|
765 |
root |
1.18 |
Under most circumstances you don't need to call this function, |
766 |
|
|
as the module selects a default that is suitable for low to |
767 |
|
|
moderate load. |
768 |
root |
1.4 |
|
769 |
root |
1.18 |
IO::AIO::max_parallel $nthreads |
770 |
|
|
Sets the maximum number of AIO threads to $nthreads. If more |
771 |
|
|
than the specified number of threads are currently running, this |
772 |
|
|
function kills them. This function blocks until the limit is |
773 |
|
|
reached. |
774 |
root |
1.4 |
|
775 |
root |
1.18 |
While $nthreads are zero, aio requests get queued but not |
776 |
|
|
executed until the number of threads has been increased again. |
777 |
root |
1.4 |
|
778 |
root |
1.18 |
This module automatically runs "max_parallel 0" at program end, |
779 |
|
|
to ensure that all threads are killed and that there are no |
780 |
|
|
outstanding requests. |
781 |
root |
1.11 |
|
782 |
root |
1.18 |
Under normal circumstances you don't need to call this function. |
783 |
root |
1.4 |
|
784 |
root |
1.19 |
IO::AIO::max_idle $nthreads |
785 |
|
|
Limit the number of threads (default: 4) that are allowed to |
786 |
|
|
idle (i.e., threads that did not get a request to process within |
787 |
|
|
10 seconds). That means if a thread becomes idle while $nthreads |
788 |
|
|
other threads are also idle, it will free its resources and |
789 |
|
|
exit. |
790 |
|
|
|
791 |
|
|
This is useful when you allow a large number of threads (e.g. |
792 |
|
|
100 or 1000) to allow for extremely high load situations, but |
793 |
|
|
want to free resources under normal circumstances (1000 threads |
794 |
|
|
can easily consume 30MB of RAM). |
795 |
|
|
|
796 |
|
|
The default is probably ok in most situations, especially if |
797 |
|
|
thread creation is fast. If thread creation is very slow on your |
798 |
|
|
system you might want to use larger values. |
799 |
|
|
|
800 |
root |
1.18 |
$oldmaxreqs = IO::AIO::max_outstanding $maxreqs |
801 |
|
|
This is a very bad function to use in interactive programs |
802 |
|
|
because it blocks, and a bad way to reduce concurrency because |
803 |
|
|
it is inexact: Better use an "aio_group" together with a feed |
804 |
|
|
callback. |
805 |
root |
1.4 |
|
806 |
root |
1.18 |
Sets the maximum number of outstanding requests to $nreqs. If |
807 |
|
|
you to queue up more than this number of requests, the next call |
808 |
|
|
to the "poll_cb" (and "poll_some" and other functions calling |
809 |
|
|
"poll_cb") function will block until the limit is no longer |
810 |
|
|
exceeded. |
811 |
root |
1.4 |
|
812 |
root |
1.18 |
The default value is very large, so there is no practical limit |
813 |
|
|
on the number of outstanding requests. |
814 |
root |
1.4 |
|
815 |
root |
1.18 |
You can still queue as many requests as you want. Therefore, |
816 |
|
|
"max_oustsanding" is mainly useful in simple scripts (with low |
817 |
|
|
values) or as a stop gap to shield against fatal memory overflow |
818 |
|
|
(with large values). |
819 |
root |
1.1 |
|
820 |
root |
1.19 |
STATISTICAL INFORMATION |
821 |
|
|
IO::AIO::nreqs |
822 |
|
|
Returns the number of requests currently in the ready, execute |
823 |
|
|
or pending states (i.e. for which their callback has not been |
824 |
|
|
invoked yet). |
825 |
|
|
|
826 |
|
|
Example: wait till there are no outstanding requests anymore: |
827 |
|
|
|
828 |
|
|
IO::AIO::poll_wait, IO::AIO::poll_cb |
829 |
|
|
while IO::AIO::nreqs; |
830 |
|
|
|
831 |
|
|
IO::AIO::nready |
832 |
|
|
Returns the number of requests currently in the ready state (not |
833 |
|
|
yet executed). |
834 |
|
|
|
835 |
|
|
IO::AIO::npending |
836 |
|
|
Returns the number of requests currently in the pending state |
837 |
|
|
(executed, but not yet processed by poll_cb). |
838 |
|
|
|
839 |
root |
1.9 |
FORK BEHAVIOUR |
840 |
root |
1.18 |
This module should do "the right thing" when the process using it |
841 |
|
|
forks: |
842 |
|
|
|
843 |
|
|
Before the fork, IO::AIO enters a quiescent state where no requests |
844 |
|
|
can be added in other threads and no results will be processed. |
845 |
|
|
After the fork the parent simply leaves the quiescent state and |
846 |
|
|
continues request/result processing, while the child frees the |
847 |
|
|
request/result queue (so that the requests started before the fork |
848 |
|
|
will only be handled in the parent). Threads will be started on |
849 |
|
|
demand until the limit set in the parent process has been reached |
850 |
|
|
again. |
851 |
|
|
|
852 |
|
|
In short: the parent will, after a short pause, continue as if fork |
853 |
|
|
had not been called, while the child will act as if IO::AIO has not |
854 |
|
|
been used yet. |
855 |
|
|
|
856 |
|
|
MEMORY USAGE |
857 |
|
|
Per-request usage: |
858 |
|
|
|
859 |
|
|
Each aio request uses - depending on your architecture - around |
860 |
|
|
100-200 bytes of memory. In addition, stat requests need a stat |
861 |
|
|
buffer (possibly a few hundred bytes), readdir requires a result |
862 |
|
|
buffer and so on. Perl scalars and other data passed into aio |
863 |
|
|
requests will also be locked and will consume memory till the |
864 |
|
|
request has entered the done state. |
865 |
|
|
|
866 |
|
|
This is now awfully much, so queuing lots of requests is not usually |
867 |
|
|
a problem. |
868 |
|
|
|
869 |
|
|
Per-thread usage: |
870 |
|
|
|
871 |
|
|
In the execution phase, some aio requests require more memory for |
872 |
|
|
temporary buffers, and each thread requires a stack and other data |
873 |
|
|
structures (usually around 16k-128k, depending on the OS). |
874 |
|
|
|
875 |
|
|
KNOWN BUGS |
876 |
|
|
Known bugs will be fixed in the next release. |
877 |
root |
1.9 |
|
878 |
root |
1.1 |
SEE ALSO |
879 |
root |
1.18 |
Coro::AIO. |
880 |
root |
1.1 |
|
881 |
|
|
AUTHOR |
882 |
root |
1.18 |
Marc Lehmann <schmorp@schmorp.de> |
883 |
|
|
http://home.schmorp.de/ |
884 |
root |
1.1 |
|