1 |
root |
1.1 |
NAME |
2 |
|
|
IO::AIO - Asynchronous Input/Output |
3 |
|
|
|
4 |
|
|
SYNOPSIS |
5 |
|
|
use IO::AIO; |
6 |
|
|
|
7 |
root |
1.44 |
aio_open "/etc/passwd", IO::AIO::O_RDONLY, 0, sub { |
8 |
root |
1.21 |
my $fh = shift |
9 |
|
|
or die "/etc/passwd: $!"; |
10 |
root |
1.5 |
... |
11 |
|
|
}; |
12 |
|
|
|
13 |
|
|
aio_unlink "/tmp/file", sub { }; |
14 |
|
|
|
15 |
|
|
aio_read $fh, 30000, 1024, $buffer, 0, sub { |
16 |
|
|
$_[0] > 0 or die "read error: $!"; |
17 |
|
|
}; |
18 |
|
|
|
19 |
root |
1.18 |
# version 2+ has request and group objects |
20 |
|
|
use IO::AIO 2; |
21 |
|
|
|
22 |
|
|
aioreq_pri 4; # give next request a very high priority |
23 |
|
|
my $req = aio_unlink "/tmp/file", sub { }; |
24 |
|
|
$req->cancel; # cancel request if still in queue |
25 |
|
|
|
26 |
|
|
my $grp = aio_group sub { print "all stats done\n" }; |
27 |
|
|
add $grp aio_stat "..." for ...; |
28 |
|
|
|
29 |
root |
1.1 |
DESCRIPTION |
30 |
|
|
This module implements asynchronous I/O using whatever means your |
31 |
root |
1.38 |
operating system supports. It is implemented as an interface to "libeio" |
32 |
|
|
(<http://software.schmorp.de/pkg/libeio.html>). |
33 |
root |
1.1 |
|
34 |
root |
1.19 |
Asynchronous means that operations that can normally block your program |
35 |
|
|
(e.g. reading from disk) will be done asynchronously: the operation will |
36 |
|
|
still block, but you can do something else in the meantime. This is |
37 |
|
|
extremely useful for programs that need to stay interactive even when |
38 |
|
|
doing heavy I/O (GUI programs, high performance network servers etc.), |
39 |
|
|
but can also be used to easily do operations in parallel that are |
40 |
|
|
normally done sequentially, e.g. stat'ing many files, which is much |
41 |
|
|
faster on a RAID volume or over NFS when you do a number of stat |
42 |
|
|
operations concurrently. |
43 |
|
|
|
44 |
root |
1.20 |
While most of this works on all types of file descriptors (for example |
45 |
|
|
sockets), using these functions on file descriptors that support |
46 |
root |
1.24 |
nonblocking operation (again, sockets, pipes etc.) is very inefficient. |
47 |
root |
1.38 |
Use an event loop for that (such as the EV module): IO::AIO will |
48 |
root |
1.24 |
naturally fit into such an event loop itself. |
49 |
root |
1.19 |
|
50 |
root |
1.18 |
In this version, a number of threads are started that execute your |
51 |
|
|
requests and signal their completion. You don't need thread support in |
52 |
|
|
perl, and the threads created by this module will not be visible to |
53 |
|
|
perl. In the future, this module might make use of the native aio |
54 |
|
|
functions available on many operating systems. However, they are often |
55 |
root |
1.19 |
not well-supported or restricted (GNU/Linux doesn't allow them on normal |
56 |
root |
1.18 |
files currently, for example), and they would only support aio_read and |
57 |
root |
1.2 |
aio_write, so the remaining functionality would have to be implemented |
58 |
|
|
using threads anyway. |
59 |
root |
1.1 |
|
60 |
root |
1.24 |
Although the module will work in the presence of other (Perl-) threads, |
61 |
|
|
it is currently not reentrant in any way, so use appropriate locking |
62 |
|
|
yourself, always call "poll_cb" from within the same thread, or never |
63 |
|
|
call "poll_cb" (or other "aio_" functions) recursively. |
64 |
root |
1.18 |
|
65 |
root |
1.19 |
EXAMPLE |
66 |
root |
1.38 |
This is a simple example that uses the EV module and loads /etc/passwd |
67 |
|
|
asynchronously: |
68 |
root |
1.19 |
|
69 |
|
|
use Fcntl; |
70 |
root |
1.38 |
use EV; |
71 |
root |
1.19 |
use IO::AIO; |
72 |
|
|
|
73 |
root |
1.38 |
# register the IO::AIO callback with EV |
74 |
|
|
my $aio_w = EV::io IO::AIO::poll_fileno, EV::READ, \&IO::AIO::poll_cb; |
75 |
root |
1.19 |
|
76 |
|
|
# queue the request to open /etc/passwd |
77 |
root |
1.44 |
aio_open "/etc/passwd", IO::AIO::O_RDONLY, 0, sub { |
78 |
root |
1.21 |
my $fh = shift |
79 |
root |
1.19 |
or die "error while opening: $!"; |
80 |
|
|
|
81 |
|
|
# stat'ing filehandles is generally non-blocking |
82 |
|
|
my $size = -s $fh; |
83 |
|
|
|
84 |
|
|
# queue a request to read the file |
85 |
|
|
my $contents; |
86 |
|
|
aio_read $fh, 0, $size, $contents, 0, sub { |
87 |
|
|
$_[0] == $size |
88 |
|
|
or die "short read: $!"; |
89 |
|
|
|
90 |
|
|
close $fh; |
91 |
|
|
|
92 |
|
|
# file contents now in $contents |
93 |
|
|
print $contents; |
94 |
|
|
|
95 |
|
|
# exit event loop and program |
96 |
root |
1.38 |
EV::unloop; |
97 |
root |
1.19 |
}; |
98 |
|
|
}; |
99 |
|
|
|
100 |
|
|
# possibly queue up other requests, or open GUI windows, |
101 |
|
|
# check for sockets etc. etc. |
102 |
|
|
|
103 |
|
|
# process events as long as there are some: |
104 |
root |
1.38 |
EV::loop; |
105 |
root |
1.19 |
|
106 |
root |
1.18 |
REQUEST ANATOMY AND LIFETIME |
107 |
|
|
Every "aio_*" function creates a request. which is a C data structure |
108 |
|
|
not directly visible to Perl. |
109 |
|
|
|
110 |
|
|
If called in non-void context, every request function returns a Perl |
111 |
|
|
object representing the request. In void context, nothing is returned, |
112 |
|
|
which saves a bit of memory. |
113 |
|
|
|
114 |
|
|
The perl object is a fairly standard ref-to-hash object. The hash |
115 |
|
|
contents are not used by IO::AIO so you are free to store anything you |
116 |
|
|
like in it. |
117 |
|
|
|
118 |
|
|
During their existance, aio requests travel through the following |
119 |
|
|
states, in order: |
120 |
|
|
|
121 |
|
|
ready |
122 |
|
|
Immediately after a request is created it is put into the ready |
123 |
|
|
state, waiting for a thread to execute it. |
124 |
|
|
|
125 |
|
|
execute |
126 |
|
|
A thread has accepted the request for processing and is currently |
127 |
|
|
executing it (e.g. blocking in read). |
128 |
|
|
|
129 |
|
|
pending |
130 |
|
|
The request has been executed and is waiting for result processing. |
131 |
|
|
|
132 |
|
|
While request submission and execution is fully asynchronous, result |
133 |
|
|
processing is not and relies on the perl interpreter calling |
134 |
|
|
"poll_cb" (or another function with the same effect). |
135 |
|
|
|
136 |
|
|
result |
137 |
|
|
The request results are processed synchronously by "poll_cb". |
138 |
|
|
|
139 |
|
|
The "poll_cb" function will process all outstanding aio requests by |
140 |
|
|
calling their callbacks, freeing memory associated with them and |
141 |
|
|
managing any groups they are contained in. |
142 |
|
|
|
143 |
|
|
done |
144 |
|
|
Request has reached the end of its lifetime and holds no resources |
145 |
|
|
anymore (except possibly for the Perl object, but its connection to |
146 |
|
|
the actual aio request is severed and calling its methods will |
147 |
|
|
either do nothing or result in a runtime error). |
148 |
root |
1.1 |
|
149 |
root |
1.4 |
FUNCTIONS |
150 |
root |
1.43 |
QUICK OVERVIEW |
151 |
|
|
This section simply lists the prototypes of the most important functions |
152 |
|
|
for quick reference. See the following sections for function-by-function |
153 |
|
|
documentation. |
154 |
|
|
|
155 |
root |
1.50 |
aio_wd $pathname, $callback->($wd) |
156 |
root |
1.43 |
aio_open $pathname, $flags, $mode, $callback->($fh) |
157 |
|
|
aio_close $fh, $callback->($status) |
158 |
|
|
aio_read $fh,$offset,$length, $data,$dataoffset, $callback->($retval) |
159 |
|
|
aio_write $fh,$offset,$length, $data,$dataoffset, $callback->($retval) |
160 |
|
|
aio_sendfile $out_fh, $in_fh, $in_offset, $length, $callback->($retval) |
161 |
|
|
aio_readahead $fh,$offset,$length, $callback->($retval) |
162 |
|
|
aio_stat $fh_or_path, $callback->($status) |
163 |
|
|
aio_lstat $fh, $callback->($status) |
164 |
|
|
aio_statvfs $fh_or_path, $callback->($statvfs) |
165 |
|
|
aio_utime $fh_or_path, $atime, $mtime, $callback->($status) |
166 |
|
|
aio_chown $fh_or_path, $uid, $gid, $callback->($status) |
167 |
|
|
aio_truncate $fh_or_path, $offset, $callback->($status) |
168 |
|
|
aio_chmod $fh_or_path, $mode, $callback->($status) |
169 |
|
|
aio_unlink $pathname, $callback->($status) |
170 |
root |
1.50 |
aio_mknod $pathname, $mode, $dev, $callback->($status) |
171 |
root |
1.43 |
aio_link $srcpath, $dstpath, $callback->($status) |
172 |
|
|
aio_symlink $srcpath, $dstpath, $callback->($status) |
173 |
root |
1.50 |
aio_readlink $pathname, $callback->($link) |
174 |
|
|
aio_realpath $pathname, $callback->($link) |
175 |
root |
1.43 |
aio_rename $srcpath, $dstpath, $callback->($status) |
176 |
|
|
aio_mkdir $pathname, $mode, $callback->($status) |
177 |
|
|
aio_rmdir $pathname, $callback->($status) |
178 |
|
|
aio_readdir $pathname, $callback->($entries) |
179 |
|
|
aio_readdirx $pathname, $flags, $callback->($entries, $flags) |
180 |
|
|
IO::AIO::READDIR_DENTS IO::AIO::READDIR_DIRS_FIRST |
181 |
|
|
IO::AIO::READDIR_STAT_ORDER IO::AIO::READDIR_FOUND_UNKNOWN |
182 |
root |
1.50 |
aio_scandir $pathname, $maxreq, $callback->($dirs, $nondirs) |
183 |
|
|
aio_load $pathname, $data, $callback->($status) |
184 |
root |
1.43 |
aio_copy $srcpath, $dstpath, $callback->($status) |
185 |
|
|
aio_move $srcpath, $dstpath, $callback->($status) |
186 |
root |
1.50 |
aio_rmtree $pathname, $callback->($status) |
187 |
root |
1.43 |
aio_sync $callback->($status) |
188 |
root |
1.50 |
aio_syncfs $fh, $callback->($status) |
189 |
root |
1.43 |
aio_fsync $fh, $callback->($status) |
190 |
|
|
aio_fdatasync $fh, $callback->($status) |
191 |
|
|
aio_sync_file_range $fh, $offset, $nbytes, $flags, $callback->($status) |
192 |
root |
1.50 |
aio_pathsync $pathname, $callback->($status) |
193 |
root |
1.43 |
aio_msync $scalar, $offset = 0, $length = undef, flags = 0, $callback->($status) |
194 |
|
|
aio_mtouch $scalar, $offset = 0, $length = undef, flags = 0, $callback->($status) |
195 |
root |
1.44 |
aio_mlock $scalar, $offset = 0, $length = undef, $callback->($status) |
196 |
|
|
aio_mlockall $flags, $callback->($status) |
197 |
root |
1.43 |
aio_group $callback->(...) |
198 |
|
|
aio_nop $callback->() |
199 |
|
|
|
200 |
|
|
$prev_pri = aioreq_pri [$pri] |
201 |
|
|
aioreq_nice $pri_adjust |
202 |
|
|
|
203 |
|
|
IO::AIO::poll_wait |
204 |
|
|
IO::AIO::poll_cb |
205 |
|
|
IO::AIO::poll |
206 |
|
|
IO::AIO::flush |
207 |
|
|
IO::AIO::max_poll_reqs $nreqs |
208 |
|
|
IO::AIO::max_poll_time $seconds |
209 |
|
|
IO::AIO::min_parallel $nthreads |
210 |
|
|
IO::AIO::max_parallel $nthreads |
211 |
|
|
IO::AIO::max_idle $nthreads |
212 |
root |
1.46 |
IO::AIO::idle_timeout $seconds |
213 |
root |
1.43 |
IO::AIO::max_outstanding $maxreqs |
214 |
|
|
IO::AIO::nreqs |
215 |
|
|
IO::AIO::nready |
216 |
|
|
IO::AIO::npending |
217 |
|
|
|
218 |
|
|
IO::AIO::sendfile $ofh, $ifh, $offset, $count |
219 |
|
|
IO::AIO::fadvise $fh, $offset, $len, $advice |
220 |
root |
1.44 |
IO::AIO::madvise $scalar, $offset, $length, $advice |
221 |
|
|
IO::AIO::mprotect $scalar, $offset, $length, $protect |
222 |
|
|
IO::AIO::munlock $scalar, $offset = 0, $length = undef |
223 |
root |
1.43 |
IO::AIO::munlockall |
224 |
|
|
|
225 |
root |
1.19 |
AIO REQUEST FUNCTIONS |
226 |
root |
1.20 |
All the "aio_*" calls are more or less thin wrappers around the syscall |
227 |
|
|
with the same name (sans "aio_"). The arguments are similar or |
228 |
|
|
identical, and they all accept an additional (and optional) $callback |
229 |
root |
1.50 |
argument which must be a code reference. This code reference will be |
230 |
|
|
called after the syscall has been executed in an asynchronous fashion. |
231 |
|
|
The results of the request will be passed as arguments to the callback |
232 |
|
|
(and, if an error occured, in $!) - for most requests the syscall return |
233 |
|
|
code (e.g. most syscalls return -1 on error, unlike perl, which usually |
234 |
|
|
delivers "false"). |
235 |
|
|
|
236 |
|
|
Some requests (such as "aio_readdir") pass the actual results and |
237 |
|
|
communicate failures by passing "undef". |
238 |
root |
1.20 |
|
239 |
|
|
All functions expecting a filehandle keep a copy of the filehandle |
240 |
|
|
internally until the request has finished. |
241 |
|
|
|
242 |
|
|
All functions return request objects of type IO::AIO::REQ that allow |
243 |
|
|
further manipulation of those requests while they are in-flight. |
244 |
|
|
|
245 |
root |
1.50 |
The pathnames you pass to these routines *should* be absolute. The |
246 |
|
|
reason for this is that at the time the request is being executed, the |
247 |
|
|
current working directory could have changed. Alternatively, you can |
248 |
|
|
make sure that you never change the current working directory anywhere |
249 |
|
|
in the program and then use relative paths. You can also take advantage |
250 |
|
|
of IO::AIOs working directory abstraction, that lets you specify paths |
251 |
|
|
relative to some previously-opened "working directory object" - see the |
252 |
|
|
description of the "IO::AIO::WD" class later in this document. |
253 |
root |
1.20 |
|
254 |
|
|
To encode pathnames as octets, either make sure you either: a) always |
255 |
|
|
pass in filenames you got from outside (command line, readdir etc.) |
256 |
root |
1.50 |
without tinkering, b) are in your native filesystem encoding, c) use the |
257 |
|
|
Encode module and encode your pathnames to the locale (or other) |
258 |
|
|
encoding in effect in the user environment, d) use |
259 |
|
|
Glib::filename_from_unicode on unicode filenames or e) use something |
260 |
|
|
else to ensure your scalar has the correct contents. |
261 |
root |
1.20 |
|
262 |
|
|
This works, btw. independent of the internal UTF-8 bit, which IO::AIO |
263 |
root |
1.32 |
handles correctly whether it is set or not. |
264 |
root |
1.20 |
|
265 |
|
|
$prev_pri = aioreq_pri [$pri] |
266 |
|
|
Returns the priority value that would be used for the next request |
267 |
|
|
and, if $pri is given, sets the priority for the next aio request. |
268 |
|
|
|
269 |
|
|
The default priority is 0, the minimum and maximum priorities are -4 |
270 |
|
|
and 4, respectively. Requests with higher priority will be serviced |
271 |
|
|
first. |
272 |
|
|
|
273 |
|
|
The priority will be reset to 0 after each call to one of the |
274 |
|
|
"aio_*" functions. |
275 |
|
|
|
276 |
|
|
Example: open a file with low priority, then read something from it |
277 |
|
|
with higher priority so the read request is serviced before other |
278 |
|
|
low priority open requests (potentially spamming the cache): |
279 |
|
|
|
280 |
|
|
aioreq_pri -3; |
281 |
|
|
aio_open ..., sub { |
282 |
|
|
return unless $_[0]; |
283 |
|
|
|
284 |
|
|
aioreq_pri -2; |
285 |
|
|
aio_read $_[0], ..., sub { |
286 |
|
|
... |
287 |
|
|
}; |
288 |
|
|
}; |
289 |
|
|
|
290 |
|
|
aioreq_nice $pri_adjust |
291 |
|
|
Similar to "aioreq_pri", but subtracts the given value from the |
292 |
|
|
current priority, so the effect is cumulative. |
293 |
|
|
|
294 |
|
|
aio_open $pathname, $flags, $mode, $callback->($fh) |
295 |
|
|
Asynchronously open or create a file and call the callback with a |
296 |
|
|
newly created filehandle for the file. |
297 |
|
|
|
298 |
|
|
The pathname passed to "aio_open" must be absolute. See API NOTES, |
299 |
|
|
above, for an explanation. |
300 |
|
|
|
301 |
|
|
The $flags argument is a bitmask. See the "Fcntl" module for a list. |
302 |
|
|
They are the same as used by "sysopen". |
303 |
|
|
|
304 |
|
|
Likewise, $mode specifies the mode of the newly created file, if it |
305 |
|
|
didn't exist and "O_CREAT" has been given, just like perl's |
306 |
|
|
"sysopen", except that it is mandatory (i.e. use 0 if you don't |
307 |
root |
1.23 |
create new files, and 0666 or 0777 if you do). Note that the $mode |
308 |
|
|
will be modified by the umask in effect then the request is being |
309 |
|
|
executed, so better never change the umask. |
310 |
root |
1.20 |
|
311 |
|
|
Example: |
312 |
|
|
|
313 |
root |
1.44 |
aio_open "/etc/passwd", IO::AIO::O_RDONLY, 0, sub { |
314 |
root |
1.20 |
if ($_[0]) { |
315 |
|
|
print "open successful, fh is $_[0]\n"; |
316 |
|
|
... |
317 |
|
|
} else { |
318 |
|
|
die "open failed: $!\n"; |
319 |
|
|
} |
320 |
|
|
}; |
321 |
|
|
|
322 |
root |
1.47 |
In addition to all the common open modes/flags ("O_RDONLY", |
323 |
|
|
"O_WRONLY", "O_RDWR", "O_CREAT", "O_TRUNC", "O_EXCL" and |
324 |
|
|
"O_APPEND"), the following POSIX and non-POSIX constants are |
325 |
|
|
available (missing ones on your system are, as usual, 0): |
326 |
|
|
|
327 |
|
|
"O_ASYNC", "O_DIRECT", "O_NOATIME", "O_CLOEXEC", "O_NOCTTY", |
328 |
|
|
"O_NOFOLLOW", "O_NONBLOCK", "O_EXEC", "O_SEARCH", "O_DIRECTORY", |
329 |
|
|
"O_DSYNC", "O_RSYNC", "O_SYNC" and "O_TTY_INIT". |
330 |
|
|
|
331 |
root |
1.20 |
aio_close $fh, $callback->($status) |
332 |
|
|
Asynchronously close a file and call the callback with the result |
333 |
root |
1.26 |
code. |
334 |
root |
1.20 |
|
335 |
root |
1.27 |
Unfortunately, you can't do this to perl. Perl *insists* very |
336 |
|
|
strongly on closing the file descriptor associated with the |
337 |
root |
1.29 |
filehandle itself. |
338 |
root |
1.27 |
|
339 |
root |
1.29 |
Therefore, "aio_close" will not close the filehandle - instead it |
340 |
|
|
will use dup2 to overwrite the file descriptor with the write-end of |
341 |
|
|
a pipe (the pipe fd will be created on demand and will be cached). |
342 |
root |
1.27 |
|
343 |
root |
1.29 |
Or in other words: the file descriptor will be closed, but it will |
344 |
|
|
not be free for reuse until the perl filehandle is closed. |
345 |
root |
1.20 |
|
346 |
|
|
aio_read $fh,$offset,$length, $data,$dataoffset, $callback->($retval) |
347 |
|
|
aio_write $fh,$offset,$length, $data,$dataoffset, $callback->($retval) |
348 |
root |
1.35 |
Reads or writes $length bytes from or to the specified $fh and |
349 |
|
|
$offset into the scalar given by $data and offset $dataoffset and |
350 |
|
|
calls the callback without the actual number of bytes read (or -1 on |
351 |
|
|
error, just like the syscall). |
352 |
|
|
|
353 |
|
|
"aio_read" will, like "sysread", shrink or grow the $data scalar to |
354 |
|
|
offset plus the actual number of bytes read. |
355 |
root |
1.24 |
|
356 |
root |
1.25 |
If $offset is undefined, then the current file descriptor offset |
357 |
|
|
will be used (and updated), otherwise the file descriptor offset |
358 |
|
|
will not be changed by these calls. |
359 |
root |
1.24 |
|
360 |
|
|
If $length is undefined in "aio_write", use the remaining length of |
361 |
|
|
$data. |
362 |
|
|
|
363 |
|
|
If $dataoffset is less than zero, it will be counted from the end of |
364 |
|
|
$data. |
365 |
root |
1.20 |
|
366 |
|
|
The $data scalar *MUST NOT* be modified in any way while the request |
367 |
root |
1.24 |
is outstanding. Modifying it can result in segfaults or World War |
368 |
|
|
III (if the necessary/optional hardware is installed). |
369 |
root |
1.20 |
|
370 |
|
|
Example: Read 15 bytes at offset 7 into scalar $buffer, starting at |
371 |
|
|
offset 0 within the scalar: |
372 |
|
|
|
373 |
|
|
aio_read $fh, 7, 15, $buffer, 0, sub { |
374 |
|
|
$_[0] > 0 or die "read error: $!"; |
375 |
|
|
print "read $_[0] bytes: <$buffer>\n"; |
376 |
|
|
}; |
377 |
|
|
|
378 |
|
|
aio_sendfile $out_fh, $in_fh, $in_offset, $length, $callback->($retval) |
379 |
|
|
Tries to copy $length bytes from $in_fh to $out_fh. It starts |
380 |
|
|
reading at byte offset $in_offset, and starts writing at the current |
381 |
|
|
file offset of $out_fh. Because of that, it is not safe to issue |
382 |
|
|
more than one "aio_sendfile" per $out_fh, as they will interfere |
383 |
root |
1.48 |
with each other. The same $in_fh works fine though, as this function |
384 |
|
|
does not move or use the file offset of $in_fh. |
385 |
root |
1.20 |
|
386 |
root |
1.45 |
Please note that "aio_sendfile" can read more bytes from $in_fh than |
387 |
root |
1.48 |
are written, and there is no way to find out how many more bytes |
388 |
|
|
have been read from "aio_sendfile" alone, as "aio_sendfile" only |
389 |
|
|
provides the number of bytes written to $out_fh. Only if the result |
390 |
|
|
value equals $length one can assume that $length bytes have been |
391 |
|
|
read. |
392 |
root |
1.45 |
|
393 |
|
|
Unlike with other "aio_" functions, it makes a lot of sense to use |
394 |
|
|
"aio_sendfile" on non-blocking sockets, as long as one end |
395 |
|
|
(typically the $in_fh) is a file - the file I/O will then be |
396 |
|
|
asynchronous, while the socket I/O will be non-blocking. Note, |
397 |
|
|
however, that you can run into a trap where "aio_sendfile" reads |
398 |
|
|
some data with readahead, then fails to write all data, and when the |
399 |
|
|
socket is ready the next time, the data in the cache is already |
400 |
|
|
lost, forcing "aio_sendfile" to again hit the disk. Explicit |
401 |
root |
1.48 |
"aio_read" + "aio_write" let's you better control resource usage. |
402 |
root |
1.45 |
|
403 |
root |
1.48 |
This call tries to make use of a native "sendfile"-like syscall to |
404 |
root |
1.20 |
provide zero-copy operation. For this to work, $out_fh should refer |
405 |
root |
1.43 |
to a socket, and $in_fh should refer to an mmap'able file. |
406 |
root |
1.20 |
|
407 |
root |
1.41 |
If a native sendfile cannot be found or it fails with "ENOSYS", |
408 |
root |
1.48 |
"EINVAL", "ENOTSUP", "EOPNOTSUPP", "EAFNOSUPPORT", "EPROTOTYPE" or |
409 |
|
|
"ENOTSOCK", it will be emulated, so you can call "aio_sendfile" on |
410 |
|
|
any type of filehandle regardless of the limitations of the |
411 |
|
|
operating system. |
412 |
|
|
|
413 |
|
|
As native sendfile syscalls (as practically any non-POSIX interface |
414 |
|
|
hacked together in a hurry to improve benchmark numbers) tend to be |
415 |
|
|
rather buggy on many systems, this implementation tries to work |
416 |
|
|
around some known bugs in Linux and FreeBSD kernels (probably |
417 |
|
|
others, too), but that might fail, so you really really should check |
418 |
|
|
the return value of "aio_sendfile" - fewre bytes than expected might |
419 |
|
|
have been transferred. |
420 |
root |
1.20 |
|
421 |
|
|
aio_readahead $fh,$offset,$length, $callback->($retval) |
422 |
|
|
"aio_readahead" populates the page cache with data from a file so |
423 |
|
|
that subsequent reads from that file will not block on disk I/O. The |
424 |
|
|
$offset argument specifies the starting point from which data is to |
425 |
|
|
be read and $length specifies the number of bytes to be read. I/O is |
426 |
|
|
performed in whole pages, so that offset is effectively rounded down |
427 |
|
|
to a page boundary and bytes are read up to the next page boundary |
428 |
|
|
greater than or equal to (off-set+length). "aio_readahead" does not |
429 |
|
|
read beyond the end of the file. The current file offset of the file |
430 |
|
|
is left unchanged. |
431 |
|
|
|
432 |
|
|
If that syscall doesn't exist (likely if your OS isn't Linux) it |
433 |
|
|
will be emulated by simply reading the data, which would have a |
434 |
|
|
similar effect. |
435 |
|
|
|
436 |
|
|
aio_stat $fh_or_path, $callback->($status) |
437 |
|
|
aio_lstat $fh, $callback->($status) |
438 |
|
|
Works like perl's "stat" or "lstat" in void context. The callback |
439 |
|
|
will be called after the stat and the results will be available |
440 |
|
|
using "stat _" or "-s _" etc... |
441 |
|
|
|
442 |
|
|
The pathname passed to "aio_stat" must be absolute. See API NOTES, |
443 |
|
|
above, for an explanation. |
444 |
|
|
|
445 |
|
|
Currently, the stats are always 64-bit-stats, i.e. instead of |
446 |
|
|
returning an error when stat'ing a large file, the results will be |
447 |
|
|
silently truncated unless perl itself is compiled with large file |
448 |
|
|
support. |
449 |
|
|
|
450 |
root |
1.46 |
To help interpret the mode and dev/rdev stat values, IO::AIO offers |
451 |
|
|
the following constants and functions (if not implemented, the |
452 |
|
|
constants will be 0 and the functions will either "croak" or fall |
453 |
|
|
back on traditional behaviour). |
454 |
|
|
|
455 |
|
|
"S_IFMT", "S_IFIFO", "S_IFCHR", "S_IFBLK", "S_IFLNK", "S_IFREG", |
456 |
|
|
"S_IFDIR", "S_IFWHT", "S_IFSOCK", "IO::AIO::major $dev_t", |
457 |
|
|
"IO::AIO::minor $dev_t", "IO::AIO::makedev $major, $minor". |
458 |
|
|
|
459 |
root |
1.20 |
Example: Print the length of /etc/passwd: |
460 |
|
|
|
461 |
|
|
aio_stat "/etc/passwd", sub { |
462 |
|
|
$_[0] and die "stat failed: $!"; |
463 |
|
|
print "size is ", -s _, "\n"; |
464 |
|
|
}; |
465 |
|
|
|
466 |
root |
1.42 |
aio_statvfs $fh_or_path, $callback->($statvfs) |
467 |
|
|
Works like the POSIX "statvfs" or "fstatvfs" syscalls, depending on |
468 |
|
|
whether a file handle or path was passed. |
469 |
|
|
|
470 |
|
|
On success, the callback is passed a hash reference with the |
471 |
|
|
following members: "bsize", "frsize", "blocks", "bfree", "bavail", |
472 |
|
|
"files", "ffree", "favail", "fsid", "flag" and "namemax". On |
473 |
|
|
failure, "undef" is passed. |
474 |
|
|
|
475 |
|
|
The following POSIX IO::AIO::ST_* constants are defined: "ST_RDONLY" |
476 |
|
|
and "ST_NOSUID". |
477 |
|
|
|
478 |
|
|
The following non-POSIX IO::AIO::ST_* flag masks are defined to |
479 |
|
|
their correct value when available, or to 0 on systems that do not |
480 |
|
|
support them: "ST_NODEV", "ST_NOEXEC", "ST_SYNCHRONOUS", |
481 |
|
|
"ST_MANDLOCK", "ST_WRITE", "ST_APPEND", "ST_IMMUTABLE", |
482 |
|
|
"ST_NOATIME", "ST_NODIRATIME" and "ST_RELATIME". |
483 |
|
|
|
484 |
|
|
Example: stat "/wd" and dump out the data if successful. |
485 |
|
|
|
486 |
|
|
aio_statvfs "/wd", sub { |
487 |
|
|
my $f = $_[0] |
488 |
|
|
or die "statvfs: $!"; |
489 |
|
|
|
490 |
|
|
use Data::Dumper; |
491 |
|
|
say Dumper $f; |
492 |
|
|
}; |
493 |
|
|
|
494 |
|
|
# result: |
495 |
|
|
{ |
496 |
|
|
bsize => 1024, |
497 |
|
|
bfree => 4333064312, |
498 |
|
|
blocks => 10253828096, |
499 |
|
|
files => 2050765568, |
500 |
|
|
flag => 4096, |
501 |
|
|
favail => 2042092649, |
502 |
|
|
bavail => 4333064312, |
503 |
|
|
ffree => 2042092649, |
504 |
|
|
namemax => 255, |
505 |
|
|
frsize => 1024, |
506 |
|
|
fsid => 1810 |
507 |
|
|
} |
508 |
|
|
|
509 |
root |
1.24 |
aio_utime $fh_or_path, $atime, $mtime, $callback->($status) |
510 |
|
|
Works like perl's "utime" function (including the special case of |
511 |
|
|
$atime and $mtime being undef). Fractional times are supported if |
512 |
|
|
the underlying syscalls support them. |
513 |
|
|
|
514 |
|
|
When called with a pathname, uses utimes(2) if available, otherwise |
515 |
|
|
utime(2). If called on a file descriptor, uses futimes(2) if |
516 |
|
|
available, otherwise returns ENOSYS, so this is not portable. |
517 |
|
|
|
518 |
|
|
Examples: |
519 |
|
|
|
520 |
|
|
# set atime and mtime to current time (basically touch(1)): |
521 |
|
|
aio_utime "path", undef, undef; |
522 |
|
|
# set atime to current time and mtime to beginning of the epoch: |
523 |
|
|
aio_utime "path", time, undef; # undef==0 |
524 |
|
|
|
525 |
|
|
aio_chown $fh_or_path, $uid, $gid, $callback->($status) |
526 |
|
|
Works like perl's "chown" function, except that "undef" for either |
527 |
|
|
$uid or $gid is being interpreted as "do not change" (but -1 can |
528 |
|
|
also be used). |
529 |
|
|
|
530 |
|
|
Examples: |
531 |
|
|
|
532 |
|
|
# same as "chown root path" in the shell: |
533 |
|
|
aio_chown "path", 0, -1; |
534 |
|
|
# same as above: |
535 |
|
|
aio_chown "path", 0, undef; |
536 |
|
|
|
537 |
|
|
aio_truncate $fh_or_path, $offset, $callback->($status) |
538 |
|
|
Works like truncate(2) or ftruncate(2). |
539 |
|
|
|
540 |
|
|
aio_chmod $fh_or_path, $mode, $callback->($status) |
541 |
|
|
Works like perl's "chmod" function. |
542 |
|
|
|
543 |
root |
1.20 |
aio_unlink $pathname, $callback->($status) |
544 |
|
|
Asynchronously unlink (delete) a file and call the callback with the |
545 |
|
|
result code. |
546 |
|
|
|
547 |
root |
1.50 |
aio_mknod $pathname, $mode, $dev, $callback->($status) |
548 |
root |
1.20 |
[EXPERIMENTAL] |
549 |
|
|
|
550 |
|
|
Asynchronously create a device node (or fifo). See mknod(2). |
551 |
|
|
|
552 |
|
|
The only (POSIX-) portable way of calling this function is: |
553 |
|
|
|
554 |
root |
1.50 |
aio_mknod $pathname, IO::AIO::S_IFIFO | $mode, 0, sub { ... |
555 |
root |
1.20 |
|
556 |
root |
1.46 |
See "aio_stat" for info about some potentially helpful extra |
557 |
|
|
constants and functions. |
558 |
|
|
|
559 |
root |
1.20 |
aio_link $srcpath, $dstpath, $callback->($status) |
560 |
|
|
Asynchronously create a new link to the existing object at $srcpath |
561 |
|
|
at the path $dstpath and call the callback with the result code. |
562 |
|
|
|
563 |
|
|
aio_symlink $srcpath, $dstpath, $callback->($status) |
564 |
|
|
Asynchronously create a new symbolic link to the existing object at |
565 |
|
|
$srcpath at the path $dstpath and call the callback with the result |
566 |
|
|
code. |
567 |
|
|
|
568 |
root |
1.50 |
aio_readlink $pathname, $callback->($link) |
569 |
root |
1.20 |
Asynchronously read the symlink specified by $path and pass it to |
570 |
|
|
the callback. If an error occurs, nothing or undef gets passed to |
571 |
|
|
the callback. |
572 |
|
|
|
573 |
root |
1.50 |
aio_realpath $pathname, $callback->($path) |
574 |
root |
1.49 |
Asynchronously make the path absolute and resolve any symlinks in |
575 |
|
|
$path. The resulting path only consists of directories (Same as |
576 |
|
|
Cwd::realpath). |
577 |
|
|
|
578 |
|
|
This request can be used to get the absolute path of the current |
579 |
|
|
working directory by passing it a path of . (a single dot). |
580 |
|
|
|
581 |
root |
1.20 |
aio_rename $srcpath, $dstpath, $callback->($status) |
582 |
|
|
Asynchronously rename the object at $srcpath to $dstpath, just as |
583 |
|
|
rename(2) and call the callback with the result code. |
584 |
|
|
|
585 |
root |
1.23 |
aio_mkdir $pathname, $mode, $callback->($status) |
586 |
|
|
Asynchronously mkdir (create) a directory and call the callback with |
587 |
|
|
the result code. $mode will be modified by the umask at the time the |
588 |
|
|
request is executed, so do not change your umask. |
589 |
|
|
|
590 |
root |
1.20 |
aio_rmdir $pathname, $callback->($status) |
591 |
|
|
Asynchronously rmdir (delete) a directory and call the callback with |
592 |
|
|
the result code. |
593 |
|
|
|
594 |
|
|
aio_readdir $pathname, $callback->($entries) |
595 |
|
|
Unlike the POSIX call of the same name, "aio_readdir" reads an |
596 |
|
|
entire directory (i.e. opendir + readdir + closedir). The entries |
597 |
|
|
will not be sorted, and will NOT include the "." and ".." entries. |
598 |
|
|
|
599 |
root |
1.36 |
The callback is passed a single argument which is either "undef" or |
600 |
|
|
an array-ref with the filenames. |
601 |
|
|
|
602 |
|
|
aio_readdirx $pathname, $flags, $callback->($entries, $flags) |
603 |
root |
1.50 |
Quite similar to "aio_readdir", but the $flags argument allows one |
604 |
|
|
to tune behaviour and output format. In case of an error, $entries |
605 |
|
|
will be "undef". |
606 |
root |
1.36 |
|
607 |
|
|
The flags are a combination of the following constants, ORed |
608 |
|
|
together (the flags will also be passed to the callback, possibly |
609 |
|
|
modified): |
610 |
|
|
|
611 |
|
|
IO::AIO::READDIR_DENTS |
612 |
root |
1.47 |
When this flag is off, then the callback gets an arrayref |
613 |
|
|
consisting of names only (as with "aio_readdir"), otherwise it |
614 |
|
|
gets an arrayref with "[$name, $type, $inode]" arrayrefs, each |
615 |
root |
1.36 |
describing a single directory entry in more detail. |
616 |
|
|
|
617 |
|
|
$name is the name of the entry. |
618 |
|
|
|
619 |
|
|
$type is one of the "IO::AIO::DT_xxx" constants: |
620 |
|
|
|
621 |
|
|
"IO::AIO::DT_UNKNOWN", "IO::AIO::DT_FIFO", "IO::AIO::DT_CHR", |
622 |
|
|
"IO::AIO::DT_DIR", "IO::AIO::DT_BLK", "IO::AIO::DT_REG", |
623 |
|
|
"IO::AIO::DT_LNK", "IO::AIO::DT_SOCK", "IO::AIO::DT_WHT". |
624 |
|
|
|
625 |
|
|
"IO::AIO::DT_UNKNOWN" means just that: readdir does not know. If |
626 |
|
|
you need to know, you have to run stat yourself. Also, for speed |
627 |
|
|
reasons, the $type scalars are read-only: you can not modify |
628 |
|
|
them. |
629 |
|
|
|
630 |
|
|
$inode is the inode number (which might not be exact on systems |
631 |
root |
1.38 |
with 64 bit inode numbers and 32 bit perls). This field has |
632 |
|
|
unspecified content on systems that do not deliver the inode |
633 |
|
|
information. |
634 |
root |
1.36 |
|
635 |
|
|
IO::AIO::READDIR_DIRS_FIRST |
636 |
|
|
When this flag is set, then the names will be returned in an |
637 |
root |
1.47 |
order where likely directories come first, in optimal stat |
638 |
|
|
order. This is useful when you need to quickly find directories, |
639 |
|
|
or you want to find all directories while avoiding to stat() |
640 |
|
|
each entry. |
641 |
root |
1.36 |
|
642 |
|
|
If the system returns type information in readdir, then this is |
643 |
|
|
used to find directories directly. Otherwise, likely directories |
644 |
root |
1.47 |
are names beginning with ".", or otherwise names with no dots, |
645 |
|
|
of which names with short names are tried first. |
646 |
root |
1.36 |
|
647 |
|
|
IO::AIO::READDIR_STAT_ORDER |
648 |
|
|
When this flag is set, then the names will be returned in an |
649 |
|
|
order suitable for stat()'ing each one. That is, when you plan |
650 |
|
|
to stat() all files in the given directory, then the returned |
651 |
|
|
order will likely be fastest. |
652 |
|
|
|
653 |
|
|
If both this flag and "IO::AIO::READDIR_DIRS_FIRST" are |
654 |
|
|
specified, then the likely dirs come first, resulting in a less |
655 |
|
|
optimal stat order. |
656 |
|
|
|
657 |
|
|
IO::AIO::READDIR_FOUND_UNKNOWN |
658 |
|
|
This flag should not be set when calling "aio_readdirx". |
659 |
|
|
Instead, it is being set by "aio_readdirx", when any of the |
660 |
root |
1.50 |
$type's found were "IO::AIO::DT_UNKNOWN". The absence of this |
661 |
root |
1.36 |
flag therefore indicates that all $type's are known, which can |
662 |
|
|
be used to speed up some algorithms. |
663 |
root |
1.20 |
|
664 |
root |
1.50 |
aio_load $pathname, $data, $callback->($status) |
665 |
root |
1.22 |
This is a composite request that tries to fully load the given file |
666 |
|
|
into memory. Status is the same as with aio_read. |
667 |
|
|
|
668 |
root |
1.20 |
aio_copy $srcpath, $dstpath, $callback->($status) |
669 |
|
|
Try to copy the *file* (directories not supported as either source |
670 |
|
|
or destination) from $srcpath to $dstpath and call the callback with |
671 |
root |
1.40 |
a status of 0 (ok) or -1 (error, see $!). |
672 |
root |
1.20 |
|
673 |
root |
1.32 |
This is a composite request that creates the destination file with |
674 |
|
|
mode 0200 and copies the contents of the source file into it using |
675 |
|
|
"aio_sendfile", followed by restoring atime, mtime, access mode and |
676 |
|
|
uid/gid, in that order. |
677 |
root |
1.20 |
|
678 |
|
|
If an error occurs, the partial destination file will be unlinked, |
679 |
|
|
if possible, except when setting atime, mtime, access mode and |
680 |
|
|
uid/gid, where errors are being ignored. |
681 |
|
|
|
682 |
|
|
aio_move $srcpath, $dstpath, $callback->($status) |
683 |
|
|
Try to move the *file* (directories not supported as either source |
684 |
|
|
or destination) from $srcpath to $dstpath and call the callback with |
685 |
root |
1.40 |
a status of 0 (ok) or -1 (error, see $!). |
686 |
root |
1.20 |
|
687 |
root |
1.33 |
This is a composite request that tries to rename(2) the file first; |
688 |
|
|
if rename fails with "EXDEV", it copies the file with "aio_copy" |
689 |
|
|
and, if that is successful, unlinks the $srcpath. |
690 |
root |
1.20 |
|
691 |
root |
1.50 |
aio_scandir $pathname, $maxreq, $callback->($dirs, $nondirs) |
692 |
root |
1.20 |
Scans a directory (similar to "aio_readdir") but additionally tries |
693 |
|
|
to efficiently separate the entries of directory $path into two sets |
694 |
|
|
of names, directories you can recurse into (directories), and ones |
695 |
|
|
you cannot recurse into (everything else, including symlinks to |
696 |
|
|
directories). |
697 |
|
|
|
698 |
|
|
"aio_scandir" is a composite request that creates of many sub |
699 |
|
|
requests_ $maxreq specifies the maximum number of outstanding aio |
700 |
|
|
requests that this function generates. If it is "<= 0", then a |
701 |
|
|
suitable default will be chosen (currently 4). |
702 |
|
|
|
703 |
|
|
On error, the callback is called without arguments, otherwise it |
704 |
|
|
receives two array-refs with path-relative entry names. |
705 |
|
|
|
706 |
|
|
Example: |
707 |
|
|
|
708 |
|
|
aio_scandir $dir, 0, sub { |
709 |
|
|
my ($dirs, $nondirs) = @_; |
710 |
|
|
print "real directories: @$dirs\n"; |
711 |
|
|
print "everything else: @$nondirs\n"; |
712 |
|
|
}; |
713 |
|
|
|
714 |
|
|
Implementation notes. |
715 |
|
|
|
716 |
|
|
The "aio_readdir" cannot be avoided, but "stat()"'ing every entry |
717 |
|
|
can. |
718 |
|
|
|
719 |
root |
1.36 |
If readdir returns file type information, then this is used directly |
720 |
|
|
to find directories. |
721 |
|
|
|
722 |
|
|
Otherwise, after reading the directory, the modification time, size |
723 |
|
|
etc. of the directory before and after the readdir is checked, and |
724 |
|
|
if they match (and isn't the current time), the link count will be |
725 |
|
|
used to decide how many entries are directories (if >= 2). |
726 |
|
|
Otherwise, no knowledge of the number of subdirectories will be |
727 |
|
|
assumed. |
728 |
|
|
|
729 |
|
|
Then entries will be sorted into likely directories a non-initial |
730 |
|
|
dot currently) and likely non-directories (see "aio_readdirx"). Then |
731 |
|
|
every entry plus an appended "/." will be "stat"'ed, likely |
732 |
|
|
directories first, in order of their inode numbers. If that |
733 |
|
|
succeeds, it assumes that the entry is a directory or a symlink to |
734 |
root |
1.50 |
directory (which will be checked separately). This is often faster |
735 |
root |
1.36 |
than stat'ing the entry itself because filesystems might detect the |
736 |
|
|
type of the entry without reading the inode data (e.g. ext2fs |
737 |
|
|
filetype feature), even on systems that cannot return the filetype |
738 |
|
|
information on readdir. |
739 |
root |
1.20 |
|
740 |
|
|
If the known number of directories (link count - 2) has been |
741 |
|
|
reached, the rest of the entries is assumed to be non-directories. |
742 |
|
|
|
743 |
|
|
This only works with certainty on POSIX (= UNIX) filesystems, which |
744 |
|
|
fortunately are the vast majority of filesystems around. |
745 |
|
|
|
746 |
|
|
It will also likely work on non-POSIX filesystems with reduced |
747 |
|
|
efficiency as those tend to return 0 or 1 as link counts, which |
748 |
|
|
disables the directory counting heuristic. |
749 |
|
|
|
750 |
root |
1.50 |
aio_rmtree $pathname, $callback->($status) |
751 |
root |
1.23 |
Delete a directory tree starting (and including) $path, return the |
752 |
|
|
status of the final "rmdir" only. This is a composite request that |
753 |
|
|
uses "aio_scandir" to recurse into and rmdir directories, and unlink |
754 |
|
|
everything else. |
755 |
|
|
|
756 |
root |
1.28 |
aio_sync $callback->($status) |
757 |
|
|
Asynchronously call sync and call the callback when finished. |
758 |
|
|
|
759 |
root |
1.20 |
aio_fsync $fh, $callback->($status) |
760 |
|
|
Asynchronously call fsync on the given filehandle and call the |
761 |
|
|
callback with the fsync result code. |
762 |
|
|
|
763 |
|
|
aio_fdatasync $fh, $callback->($status) |
764 |
|
|
Asynchronously call fdatasync on the given filehandle and call the |
765 |
|
|
callback with the fdatasync result code. |
766 |
|
|
|
767 |
|
|
If this call isn't available because your OS lacks it or it couldn't |
768 |
|
|
be detected, it will be emulated by calling "fsync" instead. |
769 |
|
|
|
770 |
root |
1.50 |
aio_syncfs $fh, $callback->($status) |
771 |
|
|
Asynchronously call the syncfs syscall to sync the filesystem |
772 |
|
|
associated to the given filehandle and call the callback with the |
773 |
|
|
syncfs result code. If syncfs is not available, calls sync(), but |
774 |
|
|
returns -1 and sets errno to "ENOSYS" nevertheless. |
775 |
|
|
|
776 |
root |
1.34 |
aio_sync_file_range $fh, $offset, $nbytes, $flags, $callback->($status) |
777 |
|
|
Sync the data portion of the file specified by $offset and $length |
778 |
|
|
to disk (but NOT the metadata), by calling the Linux-specific |
779 |
|
|
sync_file_range call. If sync_file_range is not available or it |
780 |
|
|
returns ENOSYS, then fdatasync or fsync is being substituted. |
781 |
|
|
|
782 |
|
|
$flags can be a combination of |
783 |
|
|
"IO::AIO::SYNC_FILE_RANGE_WAIT_BEFORE", |
784 |
|
|
"IO::AIO::SYNC_FILE_RANGE_WRITE" and |
785 |
|
|
"IO::AIO::SYNC_FILE_RANGE_WAIT_AFTER": refer to the sync_file_range |
786 |
|
|
manpage for details. |
787 |
|
|
|
788 |
root |
1.50 |
aio_pathsync $pathname, $callback->($status) |
789 |
root |
1.28 |
This request tries to open, fsync and close the given path. This is |
790 |
root |
1.32 |
a composite request intended to sync directories after directory |
791 |
root |
1.28 |
operations (E.g. rename). This might not work on all operating |
792 |
|
|
systems or have any specific effect, but usually it makes sure that |
793 |
|
|
directory changes get written to disc. It works for anything that |
794 |
|
|
can be opened for read-only, not just directories. |
795 |
|
|
|
796 |
root |
1.39 |
Future versions of this function might fall back to other methods |
797 |
|
|
when "fsync" on the directory fails (such as calling "sync"). |
798 |
|
|
|
799 |
root |
1.28 |
Passes 0 when everything went ok, and -1 on error. |
800 |
|
|
|
801 |
root |
1.41 |
aio_msync $scalar, $offset = 0, $length = undef, flags = 0, |
802 |
|
|
$callback->($status) |
803 |
|
|
This is a rather advanced IO::AIO call, which only works on |
804 |
root |
1.43 |
mmap(2)ed scalars (see the "IO::AIO::mmap" function, although it |
805 |
|
|
also works on data scalars managed by the Sys::Mmap or Mmap modules, |
806 |
|
|
note that the scalar must only be modified in-place while an aio |
807 |
|
|
operation is pending on it). |
808 |
root |
1.41 |
|
809 |
|
|
It calls the "msync" function of your OS, if available, with the |
810 |
|
|
memory area starting at $offset in the string and ending $length |
811 |
|
|
bytes later. If $length is negative, counts from the end, and if |
812 |
|
|
$length is "undef", then it goes till the end of the string. The |
813 |
|
|
flags can be a combination of "IO::AIO::MS_ASYNC", |
814 |
|
|
"IO::AIO::MS_INVALIDATE" and "IO::AIO::MS_SYNC". |
815 |
|
|
|
816 |
|
|
aio_mtouch $scalar, $offset = 0, $length = undef, flags = 0, |
817 |
|
|
$callback->($status) |
818 |
|
|
This is a rather advanced IO::AIO call, which works best on |
819 |
|
|
mmap(2)ed scalars. |
820 |
|
|
|
821 |
|
|
It touches (reads or writes) all memory pages in the specified range |
822 |
|
|
inside the scalar. All caveats and parameters are the same as for |
823 |
|
|
"aio_msync", above, except for flags, which must be either 0 (which |
824 |
|
|
reads all pages and ensures they are instantiated) or |
825 |
|
|
"IO::AIO::MT_MODIFY", which modifies the memory page s(by reading |
826 |
|
|
and writing an octet from it, which dirties the page). |
827 |
|
|
|
828 |
root |
1.44 |
aio_mlock $scalar, $offset = 0, $length = undef, $callback->($status) |
829 |
|
|
This is a rather advanced IO::AIO call, which works best on |
830 |
|
|
mmap(2)ed scalars. |
831 |
|
|
|
832 |
|
|
It reads in all the pages of the underlying storage into memory (if |
833 |
|
|
any) and locks them, so they are not getting swapped/paged out or |
834 |
|
|
removed. |
835 |
|
|
|
836 |
|
|
If $length is undefined, then the scalar will be locked till the |
837 |
|
|
end. |
838 |
|
|
|
839 |
|
|
On systems that do not implement "mlock", this function returns -1 |
840 |
|
|
and sets errno to "ENOSYS". |
841 |
|
|
|
842 |
|
|
Note that the corresponding "munlock" is synchronous and is |
843 |
|
|
documented under "MISCELLANEOUS FUNCTIONS". |
844 |
|
|
|
845 |
|
|
Example: open a file, mmap and mlock it - both will be undone when |
846 |
|
|
$data gets destroyed. |
847 |
|
|
|
848 |
|
|
open my $fh, "<", $path or die "$path: $!"; |
849 |
|
|
my $data; |
850 |
|
|
IO::AIO::mmap $data, -s $fh, IO::AIO::PROT_READ, IO::AIO::MAP_SHARED, $fh; |
851 |
|
|
aio_mlock $data; # mlock in background |
852 |
|
|
|
853 |
|
|
aio_mlockall $flags, $callback->($status) |
854 |
|
|
Calls the "mlockall" function with the given $flags (a combination |
855 |
|
|
of "IO::AIO::MCL_CURRENT" and "IO::AIO::MCL_FUTURE"). |
856 |
|
|
|
857 |
|
|
On systems that do not implement "mlockall", this function returns |
858 |
|
|
-1 and sets errno to "ENOSYS". |
859 |
|
|
|
860 |
|
|
Note that the corresponding "munlockall" is synchronous and is |
861 |
|
|
documented under "MISCELLANEOUS FUNCTIONS". |
862 |
|
|
|
863 |
|
|
Example: asynchronously lock all current and future pages into |
864 |
|
|
memory. |
865 |
|
|
|
866 |
|
|
aio_mlockall IO::AIO::MCL_FUTURE; |
867 |
|
|
|
868 |
root |
1.20 |
aio_group $callback->(...) |
869 |
|
|
This is a very special aio request: Instead of doing something, it |
870 |
|
|
is a container for other aio requests, which is useful if you want |
871 |
|
|
to bundle many requests into a single, composite, request with a |
872 |
|
|
definite callback and the ability to cancel the whole request with |
873 |
|
|
its subrequests. |
874 |
|
|
|
875 |
|
|
Returns an object of class IO::AIO::GRP. See its documentation below |
876 |
|
|
for more info. |
877 |
|
|
|
878 |
|
|
Example: |
879 |
|
|
|
880 |
|
|
my $grp = aio_group sub { |
881 |
|
|
print "all stats done\n"; |
882 |
|
|
}; |
883 |
|
|
|
884 |
|
|
add $grp |
885 |
|
|
(aio_stat ...), |
886 |
|
|
(aio_stat ...), |
887 |
|
|
...; |
888 |
|
|
|
889 |
|
|
aio_nop $callback->() |
890 |
|
|
This is a special request - it does nothing in itself and is only |
891 |
|
|
used for side effects, such as when you want to add a dummy request |
892 |
|
|
to a group so that finishing the requests in the group depends on |
893 |
|
|
executing the given code. |
894 |
|
|
|
895 |
|
|
While this request does nothing, it still goes through the execution |
896 |
|
|
phase and still requires a worker thread. Thus, the callback will |
897 |
|
|
not be executed immediately but only after other requests in the |
898 |
|
|
queue have entered their execution phase. This can be used to |
899 |
|
|
measure request latency. |
900 |
|
|
|
901 |
|
|
IO::AIO::aio_busy $fractional_seconds, $callback->() *NOT EXPORTED* |
902 |
|
|
Mainly used for debugging and benchmarking, this aio request puts |
903 |
|
|
one of the request workers to sleep for the given time. |
904 |
|
|
|
905 |
|
|
While it is theoretically handy to have simple I/O scheduling |
906 |
|
|
requests like sleep and file handle readable/writable, the overhead |
907 |
|
|
this creates is immense (it blocks a thread for a long time) so do |
908 |
|
|
not use this function except to put your application under |
909 |
|
|
artificial I/O pressure. |
910 |
root |
1.18 |
|
911 |
root |
1.50 |
IO::AIO::WD - multiple working directories |
912 |
|
|
Your process only has one current working directory, which is used by |
913 |
|
|
all threads. This makes it hard to use relative paths (some other |
914 |
|
|
component could call "chdir" at any time, and it is hard to control when |
915 |
|
|
the path will be used by IO::AIO). |
916 |
|
|
|
917 |
|
|
One solution for this is to always use absolute paths. This usually |
918 |
|
|
works, but can be quite slow (the kernel has to walk the whole path on |
919 |
|
|
every access), and can also be a hassle to implement. |
920 |
|
|
|
921 |
|
|
Newer POSIX systems have a number of functions (openat, fdopendir, |
922 |
|
|
futimensat and so on) that make it possible to specify working |
923 |
|
|
directories per operation. |
924 |
|
|
|
925 |
|
|
For portability, and because the clowns who "designed", or shall I |
926 |
|
|
write, perpetrated this new interface were obviously half-drunk, this |
927 |
|
|
abstraction cannot be perfect, though. |
928 |
|
|
|
929 |
|
|
IO::AIO allows you to convert directory paths into a so-called |
930 |
|
|
IO::AIO::WD object. This object stores the canonicalised, absolute |
931 |
|
|
version of the path, and on systems that allow it, also a directory file |
932 |
|
|
descriptor. |
933 |
|
|
|
934 |
|
|
Everywhere where a pathname is accepted by IO::AIO (e.g. in "aio_stat" |
935 |
|
|
or "aio_unlink"), one can specify an array reference with an IO::AIO::WD |
936 |
|
|
object and a pathname instead (or the IO::AIO::WD object alone, which |
937 |
|
|
gets interpreted as "[$wd, "."]"). If the pathname is absolute, the |
938 |
|
|
IO::AIO::WD object is ignored, otherwise the pathname is resolved |
939 |
|
|
relative to that IO::AIO::WD object. |
940 |
|
|
|
941 |
|
|
For example, to get a wd object for /etc and then stat passwd inside, |
942 |
|
|
you would write: |
943 |
|
|
|
944 |
|
|
aio_wd "/etc", sub { |
945 |
|
|
my $etcdir = shift; |
946 |
|
|
|
947 |
|
|
# although $etcdir can be undef on error, there is generally no reason |
948 |
|
|
# to check for errors here, as aio_stat will fail with ENOENT |
949 |
|
|
# when $etcdir is undef. |
950 |
|
|
|
951 |
|
|
aio_stat [$etcdir, "passwd"], sub { |
952 |
|
|
# yay |
953 |
|
|
}; |
954 |
|
|
}; |
955 |
|
|
|
956 |
|
|
That "aio_wd" is a request and not a normal function shows that creating |
957 |
|
|
an IO::AIO::WD object is itself a potentially blocking operation, which |
958 |
|
|
is why it is done asynchronously. |
959 |
|
|
|
960 |
|
|
To stat the directory obtained with "aio_wd" above, one could write |
961 |
|
|
either of the following three request calls: |
962 |
|
|
|
963 |
|
|
aio_lstat "/etc" , sub { ... # pathname as normal string |
964 |
|
|
aio_lstat [$wd, "."], sub { ... # "." relative to $wd (i.e. $wd itself) |
965 |
|
|
aio_lstat $wd , sub { ... # shorthand for the previous |
966 |
|
|
|
967 |
|
|
As with normal pathnames, IO::AIO keeps a copy of the working directory |
968 |
|
|
object and the pathname string, so you could write the following without |
969 |
|
|
causing any issues due to $path getting reused: |
970 |
|
|
|
971 |
|
|
my $path = [$wd, undef]; |
972 |
|
|
|
973 |
|
|
for my $name (qw(abc def ghi)) { |
974 |
|
|
$path->[1] = $name; |
975 |
|
|
aio_stat $path, sub { |
976 |
|
|
# ... |
977 |
|
|
}; |
978 |
|
|
} |
979 |
|
|
|
980 |
|
|
There are some caveats: when directories get renamed (or deleted), the |
981 |
|
|
pathname string doesn't change, so will point to the new directory (or |
982 |
|
|
nowhere at all), while the directory fd, if available on the system, |
983 |
|
|
will still point to the original directory. Most functions accepting a |
984 |
|
|
pathname will use the directory fd on newer systems, and the string on |
985 |
|
|
older systems. Some functions (such as realpath) will always rely on the |
986 |
|
|
string form of the pathname. |
987 |
|
|
|
988 |
|
|
So this fucntionality is mainly useful to get some protection against |
989 |
|
|
"chdir", to easily get an absolute path out of a relative path for |
990 |
|
|
future reference, and to speed up doing many operations in the same |
991 |
|
|
directory (e.g. when stat'ing all files in a directory). |
992 |
|
|
|
993 |
|
|
The following functions implement this working directory abstraction: |
994 |
|
|
|
995 |
|
|
aio_wd $pathname, $callback->($wd) |
996 |
|
|
Asynchonously canonicalise the given pathname and convert it to an |
997 |
|
|
IO::AIO::WD object representing it. If possible and supported on the |
998 |
|
|
system, also open a directory fd to speed up pathname resolution |
999 |
|
|
relative to this working directory. |
1000 |
|
|
|
1001 |
|
|
If something goes wrong, then "undef" is passwd to the callback |
1002 |
|
|
instead of a working directory object and $! is set appropriately. |
1003 |
|
|
Since passing "undef" as working directory component of a pathname |
1004 |
|
|
fails the request with "ENOENT", there is often no need for error |
1005 |
|
|
checking in the "aio_wd" callback, as future requests using the |
1006 |
|
|
value will fail in the expected way. |
1007 |
|
|
|
1008 |
|
|
If this call isn't available because your OS lacks it or it couldn't |
1009 |
|
|
be detected, it will be emulated by calling "fsync" instead. |
1010 |
|
|
|
1011 |
|
|
IO::AIO::CWD |
1012 |
|
|
This is a compiletime constant (object) that represents the process |
1013 |
|
|
current working directory. |
1014 |
|
|
|
1015 |
|
|
Specifying this object as working directory object for a pathname is |
1016 |
|
|
as if the pathname would be specified directly, without a directory |
1017 |
|
|
object, e.g., these calls are functionally identical: |
1018 |
|
|
|
1019 |
|
|
aio_stat "somefile", sub { ... }; |
1020 |
|
|
aio_stat [IO::AIO::CWD, "somefile"], sub { ... }; |
1021 |
|
|
|
1022 |
root |
1.18 |
IO::AIO::REQ CLASS |
1023 |
root |
1.20 |
All non-aggregate "aio_*" functions return an object of this class when |
1024 |
|
|
called in non-void context. |
1025 |
root |
1.18 |
|
1026 |
root |
1.20 |
cancel $req |
1027 |
|
|
Cancels the request, if possible. Has the effect of skipping |
1028 |
|
|
execution when entering the execute state and skipping calling the |
1029 |
|
|
callback when entering the the result state, but will leave the |
1030 |
root |
1.37 |
request otherwise untouched (with the exception of readdir). That |
1031 |
|
|
means that requests that currently execute will not be stopped and |
1032 |
|
|
resources held by the request will not be freed prematurely. |
1033 |
root |
1.18 |
|
1034 |
root |
1.20 |
cb $req $callback->(...) |
1035 |
|
|
Replace (or simply set) the callback registered to the request. |
1036 |
root |
1.18 |
|
1037 |
|
|
IO::AIO::GRP CLASS |
1038 |
root |
1.20 |
This class is a subclass of IO::AIO::REQ, so all its methods apply to |
1039 |
|
|
objects of this class, too. |
1040 |
root |
1.18 |
|
1041 |
root |
1.20 |
A IO::AIO::GRP object is a special request that can contain multiple |
1042 |
|
|
other aio requests. |
1043 |
root |
1.18 |
|
1044 |
root |
1.20 |
You create one by calling the "aio_group" constructing function with a |
1045 |
|
|
callback that will be called when all contained requests have entered |
1046 |
|
|
the "done" state: |
1047 |
root |
1.18 |
|
1048 |
root |
1.20 |
my $grp = aio_group sub { |
1049 |
|
|
print "all requests are done\n"; |
1050 |
|
|
}; |
1051 |
|
|
|
1052 |
|
|
You add requests by calling the "add" method with one or more |
1053 |
|
|
"IO::AIO::REQ" objects: |
1054 |
|
|
|
1055 |
|
|
$grp->add (aio_unlink "..."); |
1056 |
|
|
|
1057 |
|
|
add $grp aio_stat "...", sub { |
1058 |
|
|
$_[0] or return $grp->result ("error"); |
1059 |
root |
1.1 |
|
1060 |
root |
1.20 |
# add another request dynamically, if first succeeded |
1061 |
|
|
add $grp aio_open "...", sub { |
1062 |
|
|
$grp->result ("ok"); |
1063 |
|
|
}; |
1064 |
|
|
}; |
1065 |
root |
1.18 |
|
1066 |
root |
1.20 |
This makes it very easy to create composite requests (see the source of |
1067 |
|
|
"aio_move" for an application) that work and feel like simple requests. |
1068 |
root |
1.18 |
|
1069 |
root |
1.28 |
* The IO::AIO::GRP objects will be cleaned up during calls to |
1070 |
|
|
"IO::AIO::poll_cb", just like any other request. |
1071 |
|
|
|
1072 |
|
|
* They can be canceled like any other request. Canceling will cancel |
1073 |
|
|
not only the request itself, but also all requests it contains. |
1074 |
|
|
|
1075 |
|
|
* They can also can also be added to other IO::AIO::GRP objects. |
1076 |
|
|
|
1077 |
|
|
* You must not add requests to a group from within the group callback |
1078 |
|
|
(or any later time). |
1079 |
root |
1.20 |
|
1080 |
|
|
Their lifetime, simplified, looks like this: when they are empty, they |
1081 |
|
|
will finish very quickly. If they contain only requests that are in the |
1082 |
|
|
"done" state, they will also finish. Otherwise they will continue to |
1083 |
|
|
exist. |
1084 |
|
|
|
1085 |
root |
1.32 |
That means after creating a group you have some time to add requests |
1086 |
|
|
(precisely before the callback has been invoked, which is only done |
1087 |
|
|
within the "poll_cb"). And in the callbacks of those requests, you can |
1088 |
|
|
add further requests to the group. And only when all those requests have |
1089 |
|
|
finished will the the group itself finish. |
1090 |
root |
1.20 |
|
1091 |
|
|
add $grp ... |
1092 |
|
|
$grp->add (...) |
1093 |
|
|
Add one or more requests to the group. Any type of IO::AIO::REQ can |
1094 |
|
|
be added, including other groups, as long as you do not create |
1095 |
|
|
circular dependencies. |
1096 |
|
|
|
1097 |
|
|
Returns all its arguments. |
1098 |
|
|
|
1099 |
|
|
$grp->cancel_subs |
1100 |
|
|
Cancel all subrequests and clears any feeder, but not the group |
1101 |
|
|
request itself. Useful when you queued a lot of events but got a |
1102 |
|
|
result early. |
1103 |
|
|
|
1104 |
root |
1.41 |
The group request will finish normally (you cannot add requests to |
1105 |
|
|
the group). |
1106 |
|
|
|
1107 |
root |
1.20 |
$grp->result (...) |
1108 |
|
|
Set the result value(s) that will be passed to the group callback |
1109 |
root |
1.28 |
when all subrequests have finished and set the groups errno to the |
1110 |
root |
1.20 |
current value of errno (just like calling "errno" without an error |
1111 |
|
|
number). By default, no argument will be passed and errno is zero. |
1112 |
|
|
|
1113 |
|
|
$grp->errno ([$errno]) |
1114 |
|
|
Sets the group errno value to $errno, or the current value of errno |
1115 |
|
|
when the argument is missing. |
1116 |
|
|
|
1117 |
|
|
Every aio request has an associated errno value that is restored |
1118 |
|
|
when the callback is invoked. This method lets you change this value |
1119 |
|
|
from its default (0). |
1120 |
|
|
|
1121 |
|
|
Calling "result" will also set errno, so make sure you either set $! |
1122 |
|
|
before the call to "result", or call c<errno> after it. |
1123 |
|
|
|
1124 |
|
|
feed $grp $callback->($grp) |
1125 |
|
|
Sets a feeder/generator on this group: every group can have an |
1126 |
|
|
attached generator that generates requests if idle. The idea behind |
1127 |
|
|
this is that, although you could just queue as many requests as you |
1128 |
|
|
want in a group, this might starve other requests for a potentially |
1129 |
|
|
long time. For example, "aio_scandir" might generate hundreds of |
1130 |
root |
1.50 |
thousands of "aio_stat" requests, delaying any later requests for a |
1131 |
root |
1.20 |
long time. |
1132 |
|
|
|
1133 |
|
|
To avoid this, and allow incremental generation of requests, you can |
1134 |
|
|
instead a group and set a feeder on it that generates those |
1135 |
|
|
requests. The feed callback will be called whenever there are few |
1136 |
|
|
enough (see "limit", below) requests active in the group itself and |
1137 |
|
|
is expected to queue more requests. |
1138 |
|
|
|
1139 |
|
|
The feed callback can queue as many requests as it likes (i.e. "add" |
1140 |
|
|
does not impose any limits). |
1141 |
|
|
|
1142 |
|
|
If the feed does not queue more requests when called, it will be |
1143 |
|
|
automatically removed from the group. |
1144 |
|
|
|
1145 |
root |
1.33 |
If the feed limit is 0 when this method is called, it will be set to |
1146 |
|
|
2 automatically. |
1147 |
root |
1.20 |
|
1148 |
|
|
Example: |
1149 |
|
|
|
1150 |
|
|
# stat all files in @files, but only ever use four aio requests concurrently: |
1151 |
|
|
|
1152 |
|
|
my $grp = aio_group sub { print "finished\n" }; |
1153 |
|
|
limit $grp 4; |
1154 |
|
|
feed $grp sub { |
1155 |
|
|
my $file = pop @files |
1156 |
|
|
or return; |
1157 |
root |
1.18 |
|
1158 |
root |
1.20 |
add $grp aio_stat $file, sub { ... }; |
1159 |
root |
1.1 |
}; |
1160 |
|
|
|
1161 |
root |
1.20 |
limit $grp $num |
1162 |
|
|
Sets the feeder limit for the group: The feeder will be called |
1163 |
|
|
whenever the group contains less than this many requests. |
1164 |
root |
1.18 |
|
1165 |
root |
1.20 |
Setting the limit to 0 will pause the feeding process. |
1166 |
root |
1.17 |
|
1167 |
root |
1.33 |
The default value for the limit is 0, but note that setting a feeder |
1168 |
|
|
automatically bumps it up to 2. |
1169 |
|
|
|
1170 |
root |
1.18 |
SUPPORT FUNCTIONS |
1171 |
root |
1.19 |
EVENT PROCESSING AND EVENT LOOP INTEGRATION |
1172 |
root |
1.20 |
$fileno = IO::AIO::poll_fileno |
1173 |
|
|
Return the *request result pipe file descriptor*. This filehandle |
1174 |
|
|
must be polled for reading by some mechanism outside this module |
1175 |
root |
1.38 |
(e.g. EV, Glib, select and so on, see below or the SYNOPSIS). If the |
1176 |
|
|
pipe becomes readable you have to call "poll_cb" to check the |
1177 |
|
|
results. |
1178 |
root |
1.20 |
|
1179 |
|
|
See "poll_cb" for an example. |
1180 |
|
|
|
1181 |
|
|
IO::AIO::poll_cb |
1182 |
|
|
Process some outstanding events on the result pipe. You have to call |
1183 |
root |
1.47 |
this regularly. Returns 0 if all events could be processed (or there |
1184 |
|
|
were no events to process), or -1 if it returned earlier for |
1185 |
|
|
whatever reason. Returns immediately when no events are outstanding. |
1186 |
|
|
The amount of events processed depends on the settings of |
1187 |
|
|
"IO::AIO::max_poll_req" and "IO::AIO::max_poll_time". |
1188 |
root |
1.20 |
|
1189 |
|
|
If not all requests were processed for whatever reason, the |
1190 |
root |
1.31 |
filehandle will still be ready when "poll_cb" returns, so normally |
1191 |
|
|
you don't have to do anything special to have it called later. |
1192 |
root |
1.20 |
|
1193 |
root |
1.47 |
Apart from calling "IO::AIO::poll_cb" when the event filehandle |
1194 |
|
|
becomes ready, it can be beneficial to call this function from loops |
1195 |
|
|
which submit a lot of requests, to make sure the results get |
1196 |
|
|
processed when they become available and not just when the loop is |
1197 |
|
|
finished and the event loop takes over again. This function returns |
1198 |
|
|
very fast when there are no outstanding requests. |
1199 |
|
|
|
1200 |
root |
1.20 |
Example: Install an Event watcher that automatically calls |
1201 |
root |
1.38 |
IO::AIO::poll_cb with high priority (more examples can be found in |
1202 |
|
|
the SYNOPSIS section, at the top of this document): |
1203 |
root |
1.20 |
|
1204 |
|
|
Event->io (fd => IO::AIO::poll_fileno, |
1205 |
|
|
poll => 'r', async => 1, |
1206 |
|
|
cb => \&IO::AIO::poll_cb); |
1207 |
|
|
|
1208 |
root |
1.43 |
IO::AIO::poll_wait |
1209 |
|
|
If there are any outstanding requests and none of them in the result |
1210 |
|
|
phase, wait till the result filehandle becomes ready for reading |
1211 |
|
|
(simply does a "select" on the filehandle. This is useful if you |
1212 |
|
|
want to synchronously wait for some requests to finish). |
1213 |
|
|
|
1214 |
|
|
See "nreqs" for an example. |
1215 |
|
|
|
1216 |
|
|
IO::AIO::poll |
1217 |
|
|
Waits until some requests have been handled. |
1218 |
|
|
|
1219 |
|
|
Returns the number of requests processed, but is otherwise strictly |
1220 |
|
|
equivalent to: |
1221 |
|
|
|
1222 |
|
|
IO::AIO::poll_wait, IO::AIO::poll_cb |
1223 |
|
|
|
1224 |
|
|
IO::AIO::flush |
1225 |
|
|
Wait till all outstanding AIO requests have been handled. |
1226 |
|
|
|
1227 |
|
|
Strictly equivalent to: |
1228 |
|
|
|
1229 |
|
|
IO::AIO::poll_wait, IO::AIO::poll_cb |
1230 |
|
|
while IO::AIO::nreqs; |
1231 |
|
|
|
1232 |
root |
1.20 |
IO::AIO::max_poll_reqs $nreqs |
1233 |
|
|
IO::AIO::max_poll_time $seconds |
1234 |
|
|
These set the maximum number of requests (default 0, meaning |
1235 |
|
|
infinity) that are being processed by "IO::AIO::poll_cb" in one |
1236 |
|
|
call, respectively the maximum amount of time (default 0, meaning |
1237 |
|
|
infinity) spent in "IO::AIO::poll_cb" to process requests (more |
1238 |
|
|
correctly the mininum amount of time "poll_cb" is allowed to use). |
1239 |
|
|
|
1240 |
|
|
Setting "max_poll_time" to a non-zero value creates an overhead of |
1241 |
|
|
one syscall per request processed, which is not normally a problem |
1242 |
|
|
unless your callbacks are really really fast or your OS is really |
1243 |
|
|
really slow (I am not mentioning Solaris here). Using |
1244 |
|
|
"max_poll_reqs" incurs no overhead. |
1245 |
|
|
|
1246 |
|
|
Setting these is useful if you want to ensure some level of |
1247 |
|
|
interactiveness when perl is not fast enough to process all requests |
1248 |
|
|
in time. |
1249 |
|
|
|
1250 |
|
|
For interactive programs, values such as 0.01 to 0.1 should be fine. |
1251 |
root |
1.4 |
|
1252 |
root |
1.20 |
Example: Install an Event watcher that automatically calls |
1253 |
|
|
IO::AIO::poll_cb with low priority, to ensure that other parts of |
1254 |
|
|
the program get the CPU sometimes even under high AIO load. |
1255 |
root |
1.4 |
|
1256 |
root |
1.20 |
# try not to spend much more than 0.1s in poll_cb |
1257 |
|
|
IO::AIO::max_poll_time 0.1; |
1258 |
root |
1.4 |
|
1259 |
root |
1.20 |
# use a low priority so other tasks have priority |
1260 |
|
|
Event->io (fd => IO::AIO::poll_fileno, |
1261 |
|
|
poll => 'r', nice => 1, |
1262 |
|
|
cb => &IO::AIO::poll_cb); |
1263 |
|
|
|
1264 |
root |
1.19 |
CONTROLLING THE NUMBER OF THREADS |
1265 |
root |
1.20 |
IO::AIO::min_parallel $nthreads |
1266 |
|
|
Set the minimum number of AIO threads to $nthreads. The current |
1267 |
|
|
default is 8, which means eight asynchronous operations can execute |
1268 |
|
|
concurrently at any one time (the number of outstanding requests, |
1269 |
|
|
however, is unlimited). |
1270 |
|
|
|
1271 |
|
|
IO::AIO starts threads only on demand, when an AIO request is queued |
1272 |
|
|
and no free thread exists. Please note that queueing up a hundred |
1273 |
|
|
requests can create demand for a hundred threads, even if it turns |
1274 |
|
|
out that everything is in the cache and could have been processed |
1275 |
|
|
faster by a single thread. |
1276 |
|
|
|
1277 |
|
|
It is recommended to keep the number of threads relatively low, as |
1278 |
|
|
some Linux kernel versions will scale negatively with the number of |
1279 |
|
|
threads (higher parallelity => MUCH higher latency). With current |
1280 |
|
|
Linux 2.6 versions, 4-32 threads should be fine. |
1281 |
|
|
|
1282 |
|
|
Under most circumstances you don't need to call this function, as |
1283 |
|
|
the module selects a default that is suitable for low to moderate |
1284 |
|
|
load. |
1285 |
|
|
|
1286 |
|
|
IO::AIO::max_parallel $nthreads |
1287 |
|
|
Sets the maximum number of AIO threads to $nthreads. If more than |
1288 |
|
|
the specified number of threads are currently running, this function |
1289 |
|
|
kills them. This function blocks until the limit is reached. |
1290 |
|
|
|
1291 |
|
|
While $nthreads are zero, aio requests get queued but not executed |
1292 |
|
|
until the number of threads has been increased again. |
1293 |
|
|
|
1294 |
|
|
This module automatically runs "max_parallel 0" at program end, to |
1295 |
|
|
ensure that all threads are killed and that there are no outstanding |
1296 |
|
|
requests. |
1297 |
|
|
|
1298 |
|
|
Under normal circumstances you don't need to call this function. |
1299 |
|
|
|
1300 |
|
|
IO::AIO::max_idle $nthreads |
1301 |
|
|
Limit the number of threads (default: 4) that are allowed to idle |
1302 |
root |
1.46 |
(i.e., threads that did not get a request to process within the idle |
1303 |
|
|
timeout (default: 10 seconds). That means if a thread becomes idle |
1304 |
|
|
while $nthreads other threads are also idle, it will free its |
1305 |
|
|
resources and exit. |
1306 |
root |
1.20 |
|
1307 |
|
|
This is useful when you allow a large number of threads (e.g. 100 or |
1308 |
|
|
1000) to allow for extremely high load situations, but want to free |
1309 |
|
|
resources under normal circumstances (1000 threads can easily |
1310 |
|
|
consume 30MB of RAM). |
1311 |
|
|
|
1312 |
|
|
The default is probably ok in most situations, especially if thread |
1313 |
|
|
creation is fast. If thread creation is very slow on your system you |
1314 |
|
|
might want to use larger values. |
1315 |
|
|
|
1316 |
root |
1.46 |
IO::AIO::idle_timeout $seconds |
1317 |
|
|
Sets the minimum idle timeout (default 10) after which worker |
1318 |
|
|
threads are allowed to exit. SEe "IO::AIO::max_idle". |
1319 |
|
|
|
1320 |
root |
1.30 |
IO::AIO::max_outstanding $maxreqs |
1321 |
root |
1.48 |
Sets the maximum number of outstanding requests to $nreqs. If you do |
1322 |
|
|
queue up more than this number of requests, the next call to |
1323 |
|
|
"IO::AIO::poll_cb" (and other functions calling "poll_cb", such as |
1324 |
|
|
"IO::AIO::flush" or "IO::AIO::poll") will block until the limit is |
1325 |
|
|
no longer exceeded. |
1326 |
|
|
|
1327 |
|
|
In other words, this setting does not enforce a queue limit, but can |
1328 |
|
|
be used to make poll functions block if the limit is exceeded. |
1329 |
|
|
|
1330 |
root |
1.20 |
This is a very bad function to use in interactive programs because |
1331 |
|
|
it blocks, and a bad way to reduce concurrency because it is |
1332 |
|
|
inexact: Better use an "aio_group" together with a feed callback. |
1333 |
|
|
|
1334 |
root |
1.48 |
It's main use is in scripts without an event loop - when you want to |
1335 |
|
|
stat a lot of files, you can write somehting like this: |
1336 |
|
|
|
1337 |
|
|
IO::AIO::max_outstanding 32; |
1338 |
|
|
|
1339 |
|
|
for my $path (...) { |
1340 |
|
|
aio_stat $path , ...; |
1341 |
|
|
IO::AIO::poll_cb; |
1342 |
|
|
} |
1343 |
|
|
|
1344 |
|
|
IO::AIO::flush; |
1345 |
|
|
|
1346 |
|
|
The call to "poll_cb" inside the loop will normally return |
1347 |
|
|
instantly, but as soon as more thna 32 reqeusts are in-flight, it |
1348 |
|
|
will block until some requests have been handled. This keeps the |
1349 |
|
|
loop from pushing a large number of "aio_stat" requests onto the |
1350 |
|
|
queue. |
1351 |
|
|
|
1352 |
|
|
The default value for "max_outstanding" is very large, so there is |
1353 |
|
|
no practical limit on the number of outstanding requests. |
1354 |
root |
1.1 |
|
1355 |
root |
1.19 |
STATISTICAL INFORMATION |
1356 |
root |
1.20 |
IO::AIO::nreqs |
1357 |
|
|
Returns the number of requests currently in the ready, execute or |
1358 |
|
|
pending states (i.e. for which their callback has not been invoked |
1359 |
|
|
yet). |
1360 |
|
|
|
1361 |
|
|
Example: wait till there are no outstanding requests anymore: |
1362 |
|
|
|
1363 |
|
|
IO::AIO::poll_wait, IO::AIO::poll_cb |
1364 |
|
|
while IO::AIO::nreqs; |
1365 |
|
|
|
1366 |
|
|
IO::AIO::nready |
1367 |
|
|
Returns the number of requests currently in the ready state (not yet |
1368 |
|
|
executed). |
1369 |
|
|
|
1370 |
|
|
IO::AIO::npending |
1371 |
|
|
Returns the number of requests currently in the pending state |
1372 |
|
|
(executed, but not yet processed by poll_cb). |
1373 |
root |
1.19 |
|
1374 |
root |
1.38 |
MISCELLANEOUS FUNCTIONS |
1375 |
|
|
IO::AIO implements some functions that might be useful, but are not |
1376 |
|
|
asynchronous. |
1377 |
|
|
|
1378 |
|
|
IO::AIO::sendfile $ofh, $ifh, $offset, $count |
1379 |
|
|
Calls the "eio_sendfile_sync" function, which is like |
1380 |
|
|
"aio_sendfile", but is blocking (this makes most sense if you know |
1381 |
|
|
the input data is likely cached already and the output filehandle is |
1382 |
|
|
set to non-blocking operations). |
1383 |
|
|
|
1384 |
|
|
Returns the number of bytes copied, or -1 on error. |
1385 |
|
|
|
1386 |
|
|
IO::AIO::fadvise $fh, $offset, $len, $advice |
1387 |
root |
1.44 |
Simply calls the "posix_fadvise" function (see its manpage for |
1388 |
root |
1.50 |
details). The following advice constants are available: |
1389 |
root |
1.38 |
"IO::AIO::FADV_NORMAL", "IO::AIO::FADV_SEQUENTIAL", |
1390 |
|
|
"IO::AIO::FADV_RANDOM", "IO::AIO::FADV_NOREUSE", |
1391 |
|
|
"IO::AIO::FADV_WILLNEED", "IO::AIO::FADV_DONTNEED". |
1392 |
|
|
|
1393 |
|
|
On systems that do not implement "posix_fadvise", this function |
1394 |
|
|
returns ENOSYS, otherwise the return value of "posix_fadvise". |
1395 |
|
|
|
1396 |
root |
1.44 |
IO::AIO::madvise $scalar, $offset, $len, $advice |
1397 |
|
|
Simply calls the "posix_madvise" function (see its manpage for |
1398 |
root |
1.50 |
details). The following advice constants are available: |
1399 |
root |
1.44 |
"IO::AIO::MADV_NORMAL", "IO::AIO::MADV_SEQUENTIAL", |
1400 |
|
|
"IO::AIO::MADV_RANDOM", "IO::AIO::MADV_WILLNEED", |
1401 |
|
|
"IO::AIO::MADV_DONTNEED". |
1402 |
|
|
|
1403 |
|
|
On systems that do not implement "posix_madvise", this function |
1404 |
|
|
returns ENOSYS, otherwise the return value of "posix_madvise". |
1405 |
|
|
|
1406 |
|
|
IO::AIO::mprotect $scalar, $offset, $len, $protect |
1407 |
|
|
Simply calls the "mprotect" function on the preferably AIO::mmap'ed |
1408 |
|
|
$scalar (see its manpage for details). The following protect |
1409 |
root |
1.50 |
constants are available: "IO::AIO::PROT_NONE", "IO::AIO::PROT_READ", |
1410 |
root |
1.44 |
"IO::AIO::PROT_WRITE", "IO::AIO::PROT_EXEC". |
1411 |
|
|
|
1412 |
|
|
On systems that do not implement "mprotect", this function returns |
1413 |
|
|
ENOSYS, otherwise the return value of "mprotect". |
1414 |
|
|
|
1415 |
root |
1.43 |
IO::AIO::mmap $scalar, $length, $prot, $flags, $fh[, $offset] |
1416 |
|
|
Memory-maps a file (or anonymous memory range) and attaches it to |
1417 |
|
|
the given $scalar, which will act like a string scalar. |
1418 |
|
|
|
1419 |
|
|
The only operations allowed on the scalar are "substr"/"vec" that |
1420 |
|
|
don't change the string length, and most read-only operations such |
1421 |
|
|
as copying it or searching it with regexes and so on. |
1422 |
|
|
|
1423 |
|
|
Anything else is unsafe and will, at best, result in memory leaks. |
1424 |
|
|
|
1425 |
|
|
The memory map associated with the $scalar is automatically removed |
1426 |
|
|
when the $scalar is destroyed, or when the "IO::AIO::mmap" or |
1427 |
|
|
"IO::AIO::munmap" functions are called. |
1428 |
|
|
|
1429 |
|
|
This calls the "mmap"(2) function internally. See your system's |
1430 |
|
|
manual page for details on the $length, $prot and $flags parameters. |
1431 |
|
|
|
1432 |
|
|
The $length must be larger than zero and smaller than the actual |
1433 |
|
|
filesize. |
1434 |
|
|
|
1435 |
|
|
$prot is a combination of "IO::AIO::PROT_NONE", |
1436 |
|
|
"IO::AIO::PROT_EXEC", "IO::AIO::PROT_READ" and/or |
1437 |
|
|
"IO::AIO::PROT_WRITE", |
1438 |
|
|
|
1439 |
|
|
$flags can be a combination of "IO::AIO::MAP_SHARED" or |
1440 |
|
|
"IO::AIO::MAP_PRIVATE", or a number of system-specific flags (when |
1441 |
|
|
not available, the are defined as 0): "IO::AIO::MAP_ANONYMOUS" |
1442 |
|
|
(which is set to "MAP_ANON" if your system only provides this |
1443 |
|
|
constant), "IO::AIO::MAP_HUGETLB", "IO::AIO::MAP_LOCKED", |
1444 |
|
|
"IO::AIO::MAP_NORESERVE", "IO::AIO::MAP_POPULATE" or |
1445 |
|
|
"IO::AIO::MAP_NONBLOCK" |
1446 |
|
|
|
1447 |
|
|
If $fh is "undef", then a file descriptor of -1 is passed. |
1448 |
|
|
|
1449 |
|
|
$offset is the offset from the start of the file - it generally must |
1450 |
|
|
be a multiple of "IO::AIO::PAGESIZE" and defaults to 0. |
1451 |
|
|
|
1452 |
|
|
Example: |
1453 |
|
|
|
1454 |
|
|
use Digest::MD5; |
1455 |
|
|
use IO::AIO; |
1456 |
|
|
|
1457 |
|
|
open my $fh, "<verybigfile" |
1458 |
|
|
or die "$!"; |
1459 |
|
|
|
1460 |
|
|
IO::AIO::mmap my $data, -s $fh, IO::AIO::PROT_READ, IO::AIO::MAP_SHARED, $fh |
1461 |
|
|
or die "verybigfile: $!"; |
1462 |
|
|
|
1463 |
|
|
my $fast_md5 = md5 $data; |
1464 |
|
|
|
1465 |
|
|
IO::AIO::munmap $scalar |
1466 |
|
|
Removes a previous mmap and undefines the $scalar. |
1467 |
|
|
|
1468 |
root |
1.44 |
IO::AIO::munlock $scalar, $offset = 0, $length = undef |
1469 |
|
|
Calls the "munlock" function, undoing the effects of a previous |
1470 |
|
|
"aio_mlock" call (see its description for details). |
1471 |
root |
1.43 |
|
1472 |
|
|
IO::AIO::munlockall |
1473 |
|
|
Calls the "munlockall" function. |
1474 |
|
|
|
1475 |
|
|
On systems that do not implement "munlockall", this function returns |
1476 |
|
|
ENOSYS, otherwise the return value of "munlockall". |
1477 |
|
|
|
1478 |
|
|
EVENT LOOP INTEGRATION |
1479 |
|
|
It is recommended to use AnyEvent::AIO to integrate IO::AIO |
1480 |
|
|
automatically into many event loops: |
1481 |
|
|
|
1482 |
|
|
# AnyEvent integration (EV, Event, Glib, Tk, POE, urxvt, pureperl...) |
1483 |
|
|
use AnyEvent::AIO; |
1484 |
|
|
|
1485 |
|
|
You can also integrate IO::AIO manually into many event loops, here are |
1486 |
|
|
some examples of how to do this: |
1487 |
|
|
|
1488 |
|
|
# EV integration |
1489 |
|
|
my $aio_w = EV::io IO::AIO::poll_fileno, EV::READ, \&IO::AIO::poll_cb; |
1490 |
|
|
|
1491 |
|
|
# Event integration |
1492 |
|
|
Event->io (fd => IO::AIO::poll_fileno, |
1493 |
|
|
poll => 'r', |
1494 |
|
|
cb => \&IO::AIO::poll_cb); |
1495 |
|
|
|
1496 |
|
|
# Glib/Gtk2 integration |
1497 |
|
|
add_watch Glib::IO IO::AIO::poll_fileno, |
1498 |
|
|
in => sub { IO::AIO::poll_cb; 1 }; |
1499 |
|
|
|
1500 |
|
|
# Tk integration |
1501 |
|
|
Tk::Event::IO->fileevent (IO::AIO::poll_fileno, "", |
1502 |
|
|
readable => \&IO::AIO::poll_cb); |
1503 |
|
|
|
1504 |
|
|
# Danga::Socket integration |
1505 |
|
|
Danga::Socket->AddOtherFds (IO::AIO::poll_fileno => |
1506 |
|
|
\&IO::AIO::poll_cb); |
1507 |
|
|
|
1508 |
root |
1.9 |
FORK BEHAVIOUR |
1509 |
root |
1.48 |
Usage of pthreads in a program changes the semantics of fork |
1510 |
|
|
considerably. Specifically, only async-safe functions can be called |
1511 |
|
|
after fork. Perl doesn't know about this, so in general, you cannot call |
1512 |
root |
1.49 |
fork with defined behaviour in perl if pthreads are involved. IO::AIO |
1513 |
|
|
uses pthreads, so this applies, but many other extensions and (for |
1514 |
|
|
inexplicable reasons) perl itself often is linked against pthreads, so |
1515 |
|
|
this limitation applies to quite a lot of perls. |
1516 |
|
|
|
1517 |
|
|
This module no longer tries to fight your OS, or POSIX. That means |
1518 |
|
|
IO::AIO only works in the process that loaded it. Forking is fully |
1519 |
|
|
supported, but using IO::AIO in the child is not. |
1520 |
|
|
|
1521 |
|
|
You might get around by not *using* IO::AIO before (or after) forking. |
1522 |
|
|
You could also try to call the IO::AIO::reinit function in the child: |
1523 |
|
|
|
1524 |
|
|
IO::AIO::reinit |
1525 |
root |
1.50 |
Abandons all current requests and I/O threads and simply |
1526 |
root |
1.49 |
reinitialises all data structures. This is not an operation |
1527 |
root |
1.50 |
supported by any standards, but happens to work on GNU/Linux and |
1528 |
root |
1.49 |
some newer BSD systems. |
1529 |
|
|
|
1530 |
|
|
The only reasonable use for this function is to call it after |
1531 |
|
|
forking, if "IO::AIO" was used in the parent. Calling it while |
1532 |
|
|
IO::AIO is active in the process will result in undefined behaviour. |
1533 |
|
|
Calling it at any time will also result in any undefined (by POSIX) |
1534 |
|
|
behaviour. |
1535 |
root |
1.18 |
|
1536 |
|
|
MEMORY USAGE |
1537 |
root |
1.20 |
Per-request usage: |
1538 |
root |
1.18 |
|
1539 |
root |
1.20 |
Each aio request uses - depending on your architecture - around 100-200 |
1540 |
|
|
bytes of memory. In addition, stat requests need a stat buffer (possibly |
1541 |
|
|
a few hundred bytes), readdir requires a result buffer and so on. Perl |
1542 |
|
|
scalars and other data passed into aio requests will also be locked and |
1543 |
|
|
will consume memory till the request has entered the done state. |
1544 |
|
|
|
1545 |
root |
1.25 |
This is not awfully much, so queuing lots of requests is not usually a |
1546 |
root |
1.20 |
problem. |
1547 |
|
|
|
1548 |
|
|
Per-thread usage: |
1549 |
|
|
|
1550 |
|
|
In the execution phase, some aio requests require more memory for |
1551 |
|
|
temporary buffers, and each thread requires a stack and other data |
1552 |
|
|
structures (usually around 16k-128k, depending on the OS). |
1553 |
root |
1.18 |
|
1554 |
|
|
KNOWN BUGS |
1555 |
root |
1.20 |
Known bugs will be fixed in the next release. |
1556 |
root |
1.9 |
|
1557 |
root |
1.1 |
SEE ALSO |
1558 |
root |
1.30 |
AnyEvent::AIO for easy integration into event loops, Coro::AIO for a |
1559 |
|
|
more natural syntax. |
1560 |
root |
1.1 |
|
1561 |
|
|
AUTHOR |
1562 |
root |
1.20 |
Marc Lehmann <schmorp@schmorp.de> |
1563 |
|
|
http://home.schmorp.de/ |
1564 |
root |
1.1 |
|