1 |
NAME |
2 |
IO::AIO - Asynchronous Input/Output |
3 |
|
4 |
SYNOPSIS |
5 |
use IO::AIO; |
6 |
|
7 |
aio_open "/etc/passwd", IO::AIO::O_RDONLY, 0, sub { |
8 |
my $fh = shift |
9 |
or die "/etc/passwd: $!"; |
10 |
... |
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 |
# 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 |
DESCRIPTION |
30 |
This module implements asynchronous I/O using whatever means your |
31 |
operating system supports. It is implemented as an interface to "libeio" |
32 |
(<http://software.schmorp.de/pkg/libeio.html>). |
33 |
|
34 |
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 |
While most of this works on all types of file descriptors (for example |
45 |
sockets), using these functions on file descriptors that support |
46 |
nonblocking operation (again, sockets, pipes etc.) is very inefficient. |
47 |
Use an event loop for that (such as the EV module): IO::AIO will |
48 |
naturally fit into such an event loop itself. |
49 |
|
50 |
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 |
not well-supported or restricted (GNU/Linux doesn't allow them on normal |
56 |
files currently, for example), and they would only support aio_read and |
57 |
aio_write, so the remaining functionality would have to be implemented |
58 |
using threads anyway. |
59 |
|
60 |
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 |
|
65 |
EXAMPLE |
66 |
This is a simple example that uses the EV module and loads /etc/passwd |
67 |
asynchronously: |
68 |
|
69 |
use EV; |
70 |
use IO::AIO; |
71 |
|
72 |
# register the IO::AIO callback with EV |
73 |
my $aio_w = EV::io IO::AIO::poll_fileno, EV::READ, \&IO::AIO::poll_cb; |
74 |
|
75 |
# queue the request to open /etc/passwd |
76 |
aio_open "/etc/passwd", IO::AIO::O_RDONLY, 0, sub { |
77 |
my $fh = shift |
78 |
or die "error while opening: $!"; |
79 |
|
80 |
# stat'ing filehandles is generally non-blocking |
81 |
my $size = -s $fh; |
82 |
|
83 |
# queue a request to read the file |
84 |
my $contents; |
85 |
aio_read $fh, 0, $size, $contents, 0, sub { |
86 |
$_[0] == $size |
87 |
or die "short read: $!"; |
88 |
|
89 |
close $fh; |
90 |
|
91 |
# file contents now in $contents |
92 |
print $contents; |
93 |
|
94 |
# exit event loop and program |
95 |
EV::unloop; |
96 |
}; |
97 |
}; |
98 |
|
99 |
# possibly queue up other requests, or open GUI windows, |
100 |
# check for sockets etc. etc. |
101 |
|
102 |
# process events as long as there are some: |
103 |
EV::loop; |
104 |
|
105 |
REQUEST ANATOMY AND LIFETIME |
106 |
Every "aio_*" function creates a request. which is a C data structure |
107 |
not directly visible to Perl. |
108 |
|
109 |
If called in non-void context, every request function returns a Perl |
110 |
object representing the request. In void context, nothing is returned, |
111 |
which saves a bit of memory. |
112 |
|
113 |
The perl object is a fairly standard ref-to-hash object. The hash |
114 |
contents are not used by IO::AIO so you are free to store anything you |
115 |
like in it. |
116 |
|
117 |
During their existance, aio requests travel through the following |
118 |
states, in order: |
119 |
|
120 |
ready |
121 |
Immediately after a request is created it is put into the ready |
122 |
state, waiting for a thread to execute it. |
123 |
|
124 |
execute |
125 |
A thread has accepted the request for processing and is currently |
126 |
executing it (e.g. blocking in read). |
127 |
|
128 |
pending |
129 |
The request has been executed and is waiting for result processing. |
130 |
|
131 |
While request submission and execution is fully asynchronous, result |
132 |
processing is not and relies on the perl interpreter calling |
133 |
"poll_cb" (or another function with the same effect). |
134 |
|
135 |
result |
136 |
The request results are processed synchronously by "poll_cb". |
137 |
|
138 |
The "poll_cb" function will process all outstanding aio requests by |
139 |
calling their callbacks, freeing memory associated with them and |
140 |
managing any groups they are contained in. |
141 |
|
142 |
done |
143 |
Request has reached the end of its lifetime and holds no resources |
144 |
anymore (except possibly for the Perl object, but its connection to |
145 |
the actual aio request is severed and calling its methods will |
146 |
either do nothing or result in a runtime error). |
147 |
|
148 |
FUNCTIONS |
149 |
QUICK OVERVIEW |
150 |
This section simply lists the prototypes most of the functions for quick |
151 |
reference. See the following sections for function-by-function |
152 |
documentation. |
153 |
|
154 |
aio_wd $pathname, $callback->($wd) |
155 |
aio_open $pathname, $flags, $mode, $callback->($fh) |
156 |
aio_close $fh, $callback->($status) |
157 |
aio_seek $fh,$offset,$whence, $callback->($offs) |
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_chmod $fh_or_path, $mode, $callback->($status) |
168 |
aio_truncate $fh_or_path, $offset, $callback->($status) |
169 |
aio_allocate $fh, $mode, $offset, $len, $callback->($status) |
170 |
aio_fiemap $fh, $start, $length, $flags, $count, $cb->(\@extents) |
171 |
aio_unlink $pathname, $callback->($status) |
172 |
aio_mknod $pathname, $mode, $dev, $callback->($status) |
173 |
aio_link $srcpath, $dstpath, $callback->($status) |
174 |
aio_symlink $srcpath, $dstpath, $callback->($status) |
175 |
aio_readlink $pathname, $callback->($link) |
176 |
aio_realpath $pathname, $callback->($link) |
177 |
aio_rename $srcpath, $dstpath, $callback->($status) |
178 |
aio_mkdir $pathname, $mode, $callback->($status) |
179 |
aio_rmdir $pathname, $callback->($status) |
180 |
aio_readdir $pathname, $callback->($entries) |
181 |
aio_readdirx $pathname, $flags, $callback->($entries, $flags) |
182 |
IO::AIO::READDIR_DENTS IO::AIO::READDIR_DIRS_FIRST |
183 |
IO::AIO::READDIR_STAT_ORDER IO::AIO::READDIR_FOUND_UNKNOWN |
184 |
aio_scandir $pathname, $maxreq, $callback->($dirs, $nondirs) |
185 |
aio_load $pathname, $data, $callback->($status) |
186 |
aio_copy $srcpath, $dstpath, $callback->($status) |
187 |
aio_move $srcpath, $dstpath, $callback->($status) |
188 |
aio_rmtree $pathname, $callback->($status) |
189 |
aio_sync $callback->($status) |
190 |
aio_syncfs $fh, $callback->($status) |
191 |
aio_fsync $fh, $callback->($status) |
192 |
aio_fdatasync $fh, $callback->($status) |
193 |
aio_sync_file_range $fh, $offset, $nbytes, $flags, $callback->($status) |
194 |
aio_pathsync $pathname, $callback->($status) |
195 |
aio_msync $scalar, $offset = 0, $length = undef, flags = 0, $callback->($status) |
196 |
aio_mtouch $scalar, $offset = 0, $length = undef, flags = 0, $callback->($status) |
197 |
aio_mlock $scalar, $offset = 0, $length = undef, $callback->($status) |
198 |
aio_mlockall $flags, $callback->($status) |
199 |
aio_group $callback->(...) |
200 |
aio_nop $callback->() |
201 |
|
202 |
$prev_pri = aioreq_pri [$pri] |
203 |
aioreq_nice $pri_adjust |
204 |
|
205 |
IO::AIO::poll_wait |
206 |
IO::AIO::poll_cb |
207 |
IO::AIO::poll |
208 |
IO::AIO::flush |
209 |
IO::AIO::max_poll_reqs $nreqs |
210 |
IO::AIO::max_poll_time $seconds |
211 |
IO::AIO::min_parallel $nthreads |
212 |
IO::AIO::max_parallel $nthreads |
213 |
IO::AIO::max_idle $nthreads |
214 |
IO::AIO::idle_timeout $seconds |
215 |
IO::AIO::max_outstanding $maxreqs |
216 |
IO::AIO::nreqs |
217 |
IO::AIO::nready |
218 |
IO::AIO::npending |
219 |
|
220 |
IO::AIO::sendfile $ofh, $ifh, $offset, $count |
221 |
IO::AIO::fadvise $fh, $offset, $len, $advice |
222 |
IO::AIO::mmap $scalar, $length, $prot, $flags[, $fh[, $offset]] |
223 |
IO::AIO::munmap $scalar |
224 |
IO::AIO::madvise $scalar, $offset, $length, $advice |
225 |
IO::AIO::mprotect $scalar, $offset, $length, $protect |
226 |
IO::AIO::munlock $scalar, $offset = 0, $length = undef |
227 |
IO::AIO::munlockall |
228 |
|
229 |
API NOTES |
230 |
All the "aio_*" calls are more or less thin wrappers around the syscall |
231 |
with the same name (sans "aio_"). The arguments are similar or |
232 |
identical, and they all accept an additional (and optional) $callback |
233 |
argument which must be a code reference. This code reference will be |
234 |
called after the syscall has been executed in an asynchronous fashion. |
235 |
The results of the request will be passed as arguments to the callback |
236 |
(and, if an error occured, in $!) - for most requests the syscall return |
237 |
code (e.g. most syscalls return -1 on error, unlike perl, which usually |
238 |
delivers "false"). |
239 |
|
240 |
Some requests (such as "aio_readdir") pass the actual results and |
241 |
communicate failures by passing "undef". |
242 |
|
243 |
All functions expecting a filehandle keep a copy of the filehandle |
244 |
internally until the request has finished. |
245 |
|
246 |
All functions return request objects of type IO::AIO::REQ that allow |
247 |
further manipulation of those requests while they are in-flight. |
248 |
|
249 |
The pathnames you pass to these routines *should* be absolute. The |
250 |
reason for this is that at the time the request is being executed, the |
251 |
current working directory could have changed. Alternatively, you can |
252 |
make sure that you never change the current working directory anywhere |
253 |
in the program and then use relative paths. You can also take advantage |
254 |
of IO::AIOs working directory abstraction, that lets you specify paths |
255 |
relative to some previously-opened "working directory object" - see the |
256 |
description of the "IO::AIO::WD" class later in this document. |
257 |
|
258 |
To encode pathnames as octets, either make sure you either: a) always |
259 |
pass in filenames you got from outside (command line, readdir etc.) |
260 |
without tinkering, b) are in your native filesystem encoding, c) use the |
261 |
Encode module and encode your pathnames to the locale (or other) |
262 |
encoding in effect in the user environment, d) use |
263 |
Glib::filename_from_unicode on unicode filenames or e) use something |
264 |
else to ensure your scalar has the correct contents. |
265 |
|
266 |
This works, btw. independent of the internal UTF-8 bit, which IO::AIO |
267 |
handles correctly whether it is set or not. |
268 |
|
269 |
AIO REQUEST FUNCTIONS |
270 |
$prev_pri = aioreq_pri [$pri] |
271 |
Returns the priority value that would be used for the next request |
272 |
and, if $pri is given, sets the priority for the next aio request. |
273 |
|
274 |
The default priority is 0, the minimum and maximum priorities are -4 |
275 |
and 4, respectively. Requests with higher priority will be serviced |
276 |
first. |
277 |
|
278 |
The priority will be reset to 0 after each call to one of the |
279 |
"aio_*" functions. |
280 |
|
281 |
Example: open a file with low priority, then read something from it |
282 |
with higher priority so the read request is serviced before other |
283 |
low priority open requests (potentially spamming the cache): |
284 |
|
285 |
aioreq_pri -3; |
286 |
aio_open ..., sub { |
287 |
return unless $_[0]; |
288 |
|
289 |
aioreq_pri -2; |
290 |
aio_read $_[0], ..., sub { |
291 |
... |
292 |
}; |
293 |
}; |
294 |
|
295 |
aioreq_nice $pri_adjust |
296 |
Similar to "aioreq_pri", but subtracts the given value from the |
297 |
current priority, so the effect is cumulative. |
298 |
|
299 |
aio_open $pathname, $flags, $mode, $callback->($fh) |
300 |
Asynchronously open or create a file and call the callback with a |
301 |
newly created filehandle for the file (or "undef" in case of an |
302 |
error). |
303 |
|
304 |
The pathname passed to "aio_open" must be absolute. See API NOTES, |
305 |
above, for an explanation. |
306 |
|
307 |
The $flags argument is a bitmask. See the "Fcntl" module for a list. |
308 |
They are the same as used by "sysopen". |
309 |
|
310 |
Likewise, $mode specifies the mode of the newly created file, if it |
311 |
didn't exist and "O_CREAT" has been given, just like perl's |
312 |
"sysopen", except that it is mandatory (i.e. use 0 if you don't |
313 |
create new files, and 0666 or 0777 if you do). Note that the $mode |
314 |
will be modified by the umask in effect then the request is being |
315 |
executed, so better never change the umask. |
316 |
|
317 |
Example: |
318 |
|
319 |
aio_open "/etc/passwd", IO::AIO::O_RDONLY, 0, sub { |
320 |
if ($_[0]) { |
321 |
print "open successful, fh is $_[0]\n"; |
322 |
... |
323 |
} else { |
324 |
die "open failed: $!\n"; |
325 |
} |
326 |
}; |
327 |
|
328 |
In addition to all the common open modes/flags ("O_RDONLY", |
329 |
"O_WRONLY", "O_RDWR", "O_CREAT", "O_TRUNC", "O_EXCL" and |
330 |
"O_APPEND"), the following POSIX and non-POSIX constants are |
331 |
available (missing ones on your system are, as usual, 0): |
332 |
|
333 |
"O_ASYNC", "O_DIRECT", "O_NOATIME", "O_CLOEXEC", "O_NOCTTY", |
334 |
"O_NOFOLLOW", "O_NONBLOCK", "O_EXEC", "O_SEARCH", "O_DIRECTORY", |
335 |
"O_DSYNC", "O_RSYNC", "O_SYNC" and "O_TTY_INIT". |
336 |
|
337 |
aio_close $fh, $callback->($status) |
338 |
Asynchronously close a file and call the callback with the result |
339 |
code. |
340 |
|
341 |
Unfortunately, you can't do this to perl. Perl *insists* very |
342 |
strongly on closing the file descriptor associated with the |
343 |
filehandle itself. |
344 |
|
345 |
Therefore, "aio_close" will not close the filehandle - instead it |
346 |
will use dup2 to overwrite the file descriptor with the write-end of |
347 |
a pipe (the pipe fd will be created on demand and will be cached). |
348 |
|
349 |
Or in other words: the file descriptor will be closed, but it will |
350 |
not be free for reuse until the perl filehandle is closed. |
351 |
|
352 |
aio_seek $fh, $offset, $whence, $callback->($offs) |
353 |
Seeks the filehandle to the new $offset, similarly to perl's |
354 |
"sysseek". The $whence can use the traditional values (0 for |
355 |
"IO::AIO::SEEK_SET", 1 for "IO::AIO::SEEK_CUR" or 2 for |
356 |
"IO::AIO::SEEK_END"). |
357 |
|
358 |
The resulting absolute offset will be passed to the callback, or -1 |
359 |
in case of an error. |
360 |
|
361 |
In theory, the $whence constants could be different than the |
362 |
corresponding values from Fcntl, but perl guarantees they are the |
363 |
same, so don't panic. |
364 |
|
365 |
As a GNU/Linux (and maybe Solaris) extension, also the constants |
366 |
"IO::AIO::SEEK_DATA" and "IO::AIO::SEEK_HOLE" are available, if they |
367 |
could be found. No guarantees about suitability for use in |
368 |
"aio_seek" or Perl's "sysseek" can be made though, although I would |
369 |
naively assume they "just work". |
370 |
|
371 |
aio_read $fh,$offset,$length, $data,$dataoffset, $callback->($retval) |
372 |
aio_write $fh,$offset,$length, $data,$dataoffset, $callback->($retval) |
373 |
Reads or writes $length bytes from or to the specified $fh and |
374 |
$offset into the scalar given by $data and offset $dataoffset and |
375 |
calls the callback without the actual number of bytes read (or -1 on |
376 |
error, just like the syscall). |
377 |
|
378 |
"aio_read" will, like "sysread", shrink or grow the $data scalar to |
379 |
offset plus the actual number of bytes read. |
380 |
|
381 |
If $offset is undefined, then the current file descriptor offset |
382 |
will be used (and updated), otherwise the file descriptor offset |
383 |
will not be changed by these calls. |
384 |
|
385 |
If $length is undefined in "aio_write", use the remaining length of |
386 |
$data. |
387 |
|
388 |
If $dataoffset is less than zero, it will be counted from the end of |
389 |
$data. |
390 |
|
391 |
The $data scalar *MUST NOT* be modified in any way while the request |
392 |
is outstanding. Modifying it can result in segfaults or World War |
393 |
III (if the necessary/optional hardware is installed). |
394 |
|
395 |
Example: Read 15 bytes at offset 7 into scalar $buffer, starting at |
396 |
offset 0 within the scalar: |
397 |
|
398 |
aio_read $fh, 7, 15, $buffer, 0, sub { |
399 |
$_[0] > 0 or die "read error: $!"; |
400 |
print "read $_[0] bytes: <$buffer>\n"; |
401 |
}; |
402 |
|
403 |
aio_sendfile $out_fh, $in_fh, $in_offset, $length, $callback->($retval) |
404 |
Tries to copy $length bytes from $in_fh to $out_fh. It starts |
405 |
reading at byte offset $in_offset, and starts writing at the current |
406 |
file offset of $out_fh. Because of that, it is not safe to issue |
407 |
more than one "aio_sendfile" per $out_fh, as they will interfere |
408 |
with each other. The same $in_fh works fine though, as this function |
409 |
does not move or use the file offset of $in_fh. |
410 |
|
411 |
Please note that "aio_sendfile" can read more bytes from $in_fh than |
412 |
are written, and there is no way to find out how many more bytes |
413 |
have been read from "aio_sendfile" alone, as "aio_sendfile" only |
414 |
provides the number of bytes written to $out_fh. Only if the result |
415 |
value equals $length one can assume that $length bytes have been |
416 |
read. |
417 |
|
418 |
Unlike with other "aio_" functions, it makes a lot of sense to use |
419 |
"aio_sendfile" on non-blocking sockets, as long as one end |
420 |
(typically the $in_fh) is a file - the file I/O will then be |
421 |
asynchronous, while the socket I/O will be non-blocking. Note, |
422 |
however, that you can run into a trap where "aio_sendfile" reads |
423 |
some data with readahead, then fails to write all data, and when the |
424 |
socket is ready the next time, the data in the cache is already |
425 |
lost, forcing "aio_sendfile" to again hit the disk. Explicit |
426 |
"aio_read" + "aio_write" let's you better control resource usage. |
427 |
|
428 |
This call tries to make use of a native "sendfile"-like syscall to |
429 |
provide zero-copy operation. For this to work, $out_fh should refer |
430 |
to a socket, and $in_fh should refer to an mmap'able file. |
431 |
|
432 |
If a native sendfile cannot be found or it fails with "ENOSYS", |
433 |
"EINVAL", "ENOTSUP", "EOPNOTSUPP", "EAFNOSUPPORT", "EPROTOTYPE" or |
434 |
"ENOTSOCK", it will be emulated, so you can call "aio_sendfile" on |
435 |
any type of filehandle regardless of the limitations of the |
436 |
operating system. |
437 |
|
438 |
As native sendfile syscalls (as practically any non-POSIX interface |
439 |
hacked together in a hurry to improve benchmark numbers) tend to be |
440 |
rather buggy on many systems, this implementation tries to work |
441 |
around some known bugs in Linux and FreeBSD kernels (probably |
442 |
others, too), but that might fail, so you really really should check |
443 |
the return value of "aio_sendfile" - fewre bytes than expected might |
444 |
have been transferred. |
445 |
|
446 |
aio_readahead $fh,$offset,$length, $callback->($retval) |
447 |
"aio_readahead" populates the page cache with data from a file so |
448 |
that subsequent reads from that file will not block on disk I/O. The |
449 |
$offset argument specifies the starting point from which data is to |
450 |
be read and $length specifies the number of bytes to be read. I/O is |
451 |
performed in whole pages, so that offset is effectively rounded down |
452 |
to a page boundary and bytes are read up to the next page boundary |
453 |
greater than or equal to (off-set+length). "aio_readahead" does not |
454 |
read beyond the end of the file. The current file offset of the file |
455 |
is left unchanged. |
456 |
|
457 |
If that syscall doesn't exist (likely if your OS isn't Linux) it |
458 |
will be emulated by simply reading the data, which would have a |
459 |
similar effect. |
460 |
|
461 |
aio_stat $fh_or_path, $callback->($status) |
462 |
aio_lstat $fh, $callback->($status) |
463 |
Works like perl's "stat" or "lstat" in void context. The callback |
464 |
will be called after the stat and the results will be available |
465 |
using "stat _" or "-s _" etc... |
466 |
|
467 |
The pathname passed to "aio_stat" must be absolute. See API NOTES, |
468 |
above, for an explanation. |
469 |
|
470 |
Currently, the stats are always 64-bit-stats, i.e. instead of |
471 |
returning an error when stat'ing a large file, the results will be |
472 |
silently truncated unless perl itself is compiled with large file |
473 |
support. |
474 |
|
475 |
To help interpret the mode and dev/rdev stat values, IO::AIO offers |
476 |
the following constants and functions (if not implemented, the |
477 |
constants will be 0 and the functions will either "croak" or fall |
478 |
back on traditional behaviour). |
479 |
|
480 |
"S_IFMT", "S_IFIFO", "S_IFCHR", "S_IFBLK", "S_IFLNK", "S_IFREG", |
481 |
"S_IFDIR", "S_IFWHT", "S_IFSOCK", "IO::AIO::major $dev_t", |
482 |
"IO::AIO::minor $dev_t", "IO::AIO::makedev $major, $minor". |
483 |
|
484 |
Example: Print the length of /etc/passwd: |
485 |
|
486 |
aio_stat "/etc/passwd", sub { |
487 |
$_[0] and die "stat failed: $!"; |
488 |
print "size is ", -s _, "\n"; |
489 |
}; |
490 |
|
491 |
aio_statvfs $fh_or_path, $callback->($statvfs) |
492 |
Works like the POSIX "statvfs" or "fstatvfs" syscalls, depending on |
493 |
whether a file handle or path was passed. |
494 |
|
495 |
On success, the callback is passed a hash reference with the |
496 |
following members: "bsize", "frsize", "blocks", "bfree", "bavail", |
497 |
"files", "ffree", "favail", "fsid", "flag" and "namemax". On |
498 |
failure, "undef" is passed. |
499 |
|
500 |
The following POSIX IO::AIO::ST_* constants are defined: "ST_RDONLY" |
501 |
and "ST_NOSUID". |
502 |
|
503 |
The following non-POSIX IO::AIO::ST_* flag masks are defined to |
504 |
their correct value when available, or to 0 on systems that do not |
505 |
support them: "ST_NODEV", "ST_NOEXEC", "ST_SYNCHRONOUS", |
506 |
"ST_MANDLOCK", "ST_WRITE", "ST_APPEND", "ST_IMMUTABLE", |
507 |
"ST_NOATIME", "ST_NODIRATIME" and "ST_RELATIME". |
508 |
|
509 |
Example: stat "/wd" and dump out the data if successful. |
510 |
|
511 |
aio_statvfs "/wd", sub { |
512 |
my $f = $_[0] |
513 |
or die "statvfs: $!"; |
514 |
|
515 |
use Data::Dumper; |
516 |
say Dumper $f; |
517 |
}; |
518 |
|
519 |
# result: |
520 |
{ |
521 |
bsize => 1024, |
522 |
bfree => 4333064312, |
523 |
blocks => 10253828096, |
524 |
files => 2050765568, |
525 |
flag => 4096, |
526 |
favail => 2042092649, |
527 |
bavail => 4333064312, |
528 |
ffree => 2042092649, |
529 |
namemax => 255, |
530 |
frsize => 1024, |
531 |
fsid => 1810 |
532 |
} |
533 |
|
534 |
Here is a (likely partial - send me updates!) list of fsid values |
535 |
used by Linux - it is safe to hardcode these when $^O is "linux": |
536 |
|
537 |
0x0000adf5 adfs |
538 |
0x0000adff affs |
539 |
0x5346414f afs |
540 |
0x09041934 anon-inode filesystem |
541 |
0x00000187 autofs |
542 |
0x42465331 befs |
543 |
0x1badface bfs |
544 |
0x42494e4d binfmt_misc |
545 |
0x9123683e btrfs |
546 |
0x0027e0eb cgroupfs |
547 |
0xff534d42 cifs |
548 |
0x73757245 coda |
549 |
0x012ff7b7 coh |
550 |
0x28cd3d45 cramfs |
551 |
0x453dcd28 cramfs-wend (wrong endianness) |
552 |
0x64626720 debugfs |
553 |
0x00001373 devfs |
554 |
0x00001cd1 devpts |
555 |
0x0000f15f ecryptfs |
556 |
0x00414a53 efs |
557 |
0x0000137d ext |
558 |
0x0000ef53 ext2/ext3 |
559 |
0x0000ef51 ext2 |
560 |
0x00004006 fat |
561 |
0x65735546 fuseblk |
562 |
0x65735543 fusectl |
563 |
0x0bad1dea futexfs |
564 |
0x01161970 gfs2 |
565 |
0x47504653 gpfs |
566 |
0x00004244 hfs |
567 |
0xf995e849 hpfs |
568 |
0x958458f6 hugetlbfs |
569 |
0x2bad1dea inotifyfs |
570 |
0x00009660 isofs |
571 |
0x000072b6 jffs2 |
572 |
0x3153464a jfs |
573 |
0x6b414653 k-afs |
574 |
0x0bd00bd0 lustre |
575 |
0x0000137f minix |
576 |
0x0000138f minix 30 char names |
577 |
0x00002468 minix v2 |
578 |
0x00002478 minix v2 30 char names |
579 |
0x00004d5a minix v3 |
580 |
0x19800202 mqueue |
581 |
0x00004d44 msdos |
582 |
0x0000564c novell |
583 |
0x00006969 nfs |
584 |
0x6e667364 nfsd |
585 |
0x00003434 nilfs |
586 |
0x5346544e ntfs |
587 |
0x00009fa1 openprom |
588 |
0x7461636F ocfs2 |
589 |
0x00009fa0 proc |
590 |
0x6165676c pstorefs |
591 |
0x0000002f qnx4 |
592 |
0x858458f6 ramfs |
593 |
0x52654973 reiserfs |
594 |
0x00007275 romfs |
595 |
0x67596969 rpc_pipefs |
596 |
0x73636673 securityfs |
597 |
0xf97cff8c selinux |
598 |
0x0000517b smb |
599 |
0x534f434b sockfs |
600 |
0x73717368 squashfs |
601 |
0x62656572 sysfs |
602 |
0x012ff7b6 sysv2 |
603 |
0x012ff7b5 sysv4 |
604 |
0x01021994 tmpfs |
605 |
0x15013346 udf |
606 |
0x00011954 ufs |
607 |
0x54190100 ufs byteswapped |
608 |
0x00009fa2 usbdevfs |
609 |
0x01021997 v9fs |
610 |
0xa501fcf5 vxfs |
611 |
0xabba1974 xenfs |
612 |
0x012ff7b4 xenix |
613 |
0x58465342 xfs |
614 |
0x012fd16d xia |
615 |
|
616 |
aio_utime $fh_or_path, $atime, $mtime, $callback->($status) |
617 |
Works like perl's "utime" function (including the special case of |
618 |
$atime and $mtime being undef). Fractional times are supported if |
619 |
the underlying syscalls support them. |
620 |
|
621 |
When called with a pathname, uses utimes(2) if available, otherwise |
622 |
utime(2). If called on a file descriptor, uses futimes(2) if |
623 |
available, otherwise returns ENOSYS, so this is not portable. |
624 |
|
625 |
Examples: |
626 |
|
627 |
# set atime and mtime to current time (basically touch(1)): |
628 |
aio_utime "path", undef, undef; |
629 |
# set atime to current time and mtime to beginning of the epoch: |
630 |
aio_utime "path", time, undef; # undef==0 |
631 |
|
632 |
aio_chown $fh_or_path, $uid, $gid, $callback->($status) |
633 |
Works like perl's "chown" function, except that "undef" for either |
634 |
$uid or $gid is being interpreted as "do not change" (but -1 can |
635 |
also be used). |
636 |
|
637 |
Examples: |
638 |
|
639 |
# same as "chown root path" in the shell: |
640 |
aio_chown "path", 0, -1; |
641 |
# same as above: |
642 |
aio_chown "path", 0, undef; |
643 |
|
644 |
aio_truncate $fh_or_path, $offset, $callback->($status) |
645 |
Works like truncate(2) or ftruncate(2). |
646 |
|
647 |
aio_allocate $fh, $mode, $offset, $len, $callback->($status) |
648 |
Allocates or freed disk space according to the $mode argument. See |
649 |
the linux "fallocate" docuemntation for details. |
650 |
|
651 |
$mode can currently be 0 or "IO::AIO::FALLOC_FL_KEEP_SIZE" to |
652 |
allocate space, or "IO::AIO::FALLOC_FL_PUNCH_HOLE | |
653 |
IO::AIO::FALLOC_FL_KEEP_SIZE", to deallocate a file range. |
654 |
|
655 |
The file system block size used by "fallocate" is presumably the |
656 |
"f_bsize" returned by "statvfs". |
657 |
|
658 |
If "fallocate" isn't available or cannot be emulated (currently no |
659 |
emulation will be attempted), passes -1 and sets $! to "ENOSYS". |
660 |
|
661 |
aio_chmod $fh_or_path, $mode, $callback->($status) |
662 |
Works like perl's "chmod" function. |
663 |
|
664 |
aio_unlink $pathname, $callback->($status) |
665 |
Asynchronously unlink (delete) a file and call the callback with the |
666 |
result code. |
667 |
|
668 |
aio_mknod $pathname, $mode, $dev, $callback->($status) |
669 |
[EXPERIMENTAL] |
670 |
|
671 |
Asynchronously create a device node (or fifo). See mknod(2). |
672 |
|
673 |
The only (POSIX-) portable way of calling this function is: |
674 |
|
675 |
aio_mknod $pathname, IO::AIO::S_IFIFO | $mode, 0, sub { ... |
676 |
|
677 |
See "aio_stat" for info about some potentially helpful extra |
678 |
constants and functions. |
679 |
|
680 |
aio_link $srcpath, $dstpath, $callback->($status) |
681 |
Asynchronously create a new link to the existing object at $srcpath |
682 |
at the path $dstpath and call the callback with the result code. |
683 |
|
684 |
aio_symlink $srcpath, $dstpath, $callback->($status) |
685 |
Asynchronously create a new symbolic link to the existing object at |
686 |
$srcpath at the path $dstpath and call the callback with the result |
687 |
code. |
688 |
|
689 |
aio_readlink $pathname, $callback->($link) |
690 |
Asynchronously read the symlink specified by $path and pass it to |
691 |
the callback. If an error occurs, nothing or undef gets passed to |
692 |
the callback. |
693 |
|
694 |
aio_realpath $pathname, $callback->($path) |
695 |
Asynchronously make the path absolute and resolve any symlinks in |
696 |
$path. The resulting path only consists of directories (same as |
697 |
Cwd::realpath). |
698 |
|
699 |
This request can be used to get the absolute path of the current |
700 |
working directory by passing it a path of . (a single dot). |
701 |
|
702 |
aio_rename $srcpath, $dstpath, $callback->($status) |
703 |
Asynchronously rename the object at $srcpath to $dstpath, just as |
704 |
rename(2) and call the callback with the result code. |
705 |
|
706 |
On systems that support the AIO::WD working directory abstraction |
707 |
natively, the case "[$wd, "."]" as $srcpath is specialcased - |
708 |
instead of failing, "rename" is called on the absolute path of $wd. |
709 |
|
710 |
aio_mkdir $pathname, $mode, $callback->($status) |
711 |
Asynchronously mkdir (create) a directory and call the callback with |
712 |
the result code. $mode will be modified by the umask at the time the |
713 |
request is executed, so do not change your umask. |
714 |
|
715 |
aio_rmdir $pathname, $callback->($status) |
716 |
Asynchronously rmdir (delete) a directory and call the callback with |
717 |
the result code. |
718 |
|
719 |
On systems that support the AIO::WD working directory abstraction |
720 |
natively, the case "[$wd, "."]" is specialcased - instead of |
721 |
failing, "rmdir" is called on the absolute path of $wd. |
722 |
|
723 |
aio_readdir $pathname, $callback->($entries) |
724 |
Unlike the POSIX call of the same name, "aio_readdir" reads an |
725 |
entire directory (i.e. opendir + readdir + closedir). The entries |
726 |
will not be sorted, and will NOT include the "." and ".." entries. |
727 |
|
728 |
The callback is passed a single argument which is either "undef" or |
729 |
an array-ref with the filenames. |
730 |
|
731 |
aio_readdirx $pathname, $flags, $callback->($entries, $flags) |
732 |
Quite similar to "aio_readdir", but the $flags argument allows one |
733 |
to tune behaviour and output format. In case of an error, $entries |
734 |
will be "undef". |
735 |
|
736 |
The flags are a combination of the following constants, ORed |
737 |
together (the flags will also be passed to the callback, possibly |
738 |
modified): |
739 |
|
740 |
IO::AIO::READDIR_DENTS |
741 |
When this flag is off, then the callback gets an arrayref |
742 |
consisting of names only (as with "aio_readdir"), otherwise it |
743 |
gets an arrayref with "[$name, $type, $inode]" arrayrefs, each |
744 |
describing a single directory entry in more detail. |
745 |
|
746 |
$name is the name of the entry. |
747 |
|
748 |
$type is one of the "IO::AIO::DT_xxx" constants: |
749 |
|
750 |
"IO::AIO::DT_UNKNOWN", "IO::AIO::DT_FIFO", "IO::AIO::DT_CHR", |
751 |
"IO::AIO::DT_DIR", "IO::AIO::DT_BLK", "IO::AIO::DT_REG", |
752 |
"IO::AIO::DT_LNK", "IO::AIO::DT_SOCK", "IO::AIO::DT_WHT". |
753 |
|
754 |
"IO::AIO::DT_UNKNOWN" means just that: readdir does not know. If |
755 |
you need to know, you have to run stat yourself. Also, for speed |
756 |
reasons, the $type scalars are read-only: you can not modify |
757 |
them. |
758 |
|
759 |
$inode is the inode number (which might not be exact on systems |
760 |
with 64 bit inode numbers and 32 bit perls). This field has |
761 |
unspecified content on systems that do not deliver the inode |
762 |
information. |
763 |
|
764 |
IO::AIO::READDIR_DIRS_FIRST |
765 |
When this flag is set, then the names will be returned in an |
766 |
order where likely directories come first, in optimal stat |
767 |
order. This is useful when you need to quickly find directories, |
768 |
or you want to find all directories while avoiding to stat() |
769 |
each entry. |
770 |
|
771 |
If the system returns type information in readdir, then this is |
772 |
used to find directories directly. Otherwise, likely directories |
773 |
are names beginning with ".", or otherwise names with no dots, |
774 |
of which names with short names are tried first. |
775 |
|
776 |
IO::AIO::READDIR_STAT_ORDER |
777 |
When this flag is set, then the names will be returned in an |
778 |
order suitable for stat()'ing each one. That is, when you plan |
779 |
to stat() all files in the given directory, then the returned |
780 |
order will likely be fastest. |
781 |
|
782 |
If both this flag and "IO::AIO::READDIR_DIRS_FIRST" are |
783 |
specified, then the likely dirs come first, resulting in a less |
784 |
optimal stat order. |
785 |
|
786 |
IO::AIO::READDIR_FOUND_UNKNOWN |
787 |
This flag should not be set when calling "aio_readdirx". |
788 |
Instead, it is being set by "aio_readdirx", when any of the |
789 |
$type's found were "IO::AIO::DT_UNKNOWN". The absence of this |
790 |
flag therefore indicates that all $type's are known, which can |
791 |
be used to speed up some algorithms. |
792 |
|
793 |
aio_load $pathname, $data, $callback->($status) |
794 |
This is a composite request that tries to fully load the given file |
795 |
into memory. Status is the same as with aio_read. |
796 |
|
797 |
aio_copy $srcpath, $dstpath, $callback->($status) |
798 |
Try to copy the *file* (directories not supported as either source |
799 |
or destination) from $srcpath to $dstpath and call the callback with |
800 |
a status of 0 (ok) or -1 (error, see $!). |
801 |
|
802 |
This is a composite request that creates the destination file with |
803 |
mode 0200 and copies the contents of the source file into it using |
804 |
"aio_sendfile", followed by restoring atime, mtime, access mode and |
805 |
uid/gid, in that order. |
806 |
|
807 |
If an error occurs, the partial destination file will be unlinked, |
808 |
if possible, except when setting atime, mtime, access mode and |
809 |
uid/gid, where errors are being ignored. |
810 |
|
811 |
aio_move $srcpath, $dstpath, $callback->($status) |
812 |
Try to move the *file* (directories not supported as either source |
813 |
or destination) from $srcpath to $dstpath and call the callback with |
814 |
a status of 0 (ok) or -1 (error, see $!). |
815 |
|
816 |
This is a composite request that tries to rename(2) the file first; |
817 |
if rename fails with "EXDEV", it copies the file with "aio_copy" |
818 |
and, if that is successful, unlinks the $srcpath. |
819 |
|
820 |
aio_scandir $pathname, $maxreq, $callback->($dirs, $nondirs) |
821 |
Scans a directory (similar to "aio_readdir") but additionally tries |
822 |
to efficiently separate the entries of directory $path into two sets |
823 |
of names, directories you can recurse into (directories), and ones |
824 |
you cannot recurse into (everything else, including symlinks to |
825 |
directories). |
826 |
|
827 |
"aio_scandir" is a composite request that creates of many sub |
828 |
requests_ $maxreq specifies the maximum number of outstanding aio |
829 |
requests that this function generates. If it is "<= 0", then a |
830 |
suitable default will be chosen (currently 4). |
831 |
|
832 |
On error, the callback is called without arguments, otherwise it |
833 |
receives two array-refs with path-relative entry names. |
834 |
|
835 |
Example: |
836 |
|
837 |
aio_scandir $dir, 0, sub { |
838 |
my ($dirs, $nondirs) = @_; |
839 |
print "real directories: @$dirs\n"; |
840 |
print "everything else: @$nondirs\n"; |
841 |
}; |
842 |
|
843 |
Implementation notes. |
844 |
|
845 |
The "aio_readdir" cannot be avoided, but "stat()"'ing every entry |
846 |
can. |
847 |
|
848 |
If readdir returns file type information, then this is used directly |
849 |
to find directories. |
850 |
|
851 |
Otherwise, after reading the directory, the modification time, size |
852 |
etc. of the directory before and after the readdir is checked, and |
853 |
if they match (and isn't the current time), the link count will be |
854 |
used to decide how many entries are directories (if >= 2). |
855 |
Otherwise, no knowledge of the number of subdirectories will be |
856 |
assumed. |
857 |
|
858 |
Then entries will be sorted into likely directories a non-initial |
859 |
dot currently) and likely non-directories (see "aio_readdirx"). Then |
860 |
every entry plus an appended "/." will be "stat"'ed, likely |
861 |
directories first, in order of their inode numbers. If that |
862 |
succeeds, it assumes that the entry is a directory or a symlink to |
863 |
directory (which will be checked separately). This is often faster |
864 |
than stat'ing the entry itself because filesystems might detect the |
865 |
type of the entry without reading the inode data (e.g. ext2fs |
866 |
filetype feature), even on systems that cannot return the filetype |
867 |
information on readdir. |
868 |
|
869 |
If the known number of directories (link count - 2) has been |
870 |
reached, the rest of the entries is assumed to be non-directories. |
871 |
|
872 |
This only works with certainty on POSIX (= UNIX) filesystems, which |
873 |
fortunately are the vast majority of filesystems around. |
874 |
|
875 |
It will also likely work on non-POSIX filesystems with reduced |
876 |
efficiency as those tend to return 0 or 1 as link counts, which |
877 |
disables the directory counting heuristic. |
878 |
|
879 |
aio_rmtree $pathname, $callback->($status) |
880 |
Delete a directory tree starting (and including) $path, return the |
881 |
status of the final "rmdir" only. This is a composite request that |
882 |
uses "aio_scandir" to recurse into and rmdir directories, and unlink |
883 |
everything else. |
884 |
|
885 |
aio_sync $callback->($status) |
886 |
Asynchronously call sync and call the callback when finished. |
887 |
|
888 |
aio_fsync $fh, $callback->($status) |
889 |
Asynchronously call fsync on the given filehandle and call the |
890 |
callback with the fsync result code. |
891 |
|
892 |
aio_fdatasync $fh, $callback->($status) |
893 |
Asynchronously call fdatasync on the given filehandle and call the |
894 |
callback with the fdatasync result code. |
895 |
|
896 |
If this call isn't available because your OS lacks it or it couldn't |
897 |
be detected, it will be emulated by calling "fsync" instead. |
898 |
|
899 |
aio_syncfs $fh, $callback->($status) |
900 |
Asynchronously call the syncfs syscall to sync the filesystem |
901 |
associated to the given filehandle and call the callback with the |
902 |
syncfs result code. If syncfs is not available, calls sync(), but |
903 |
returns -1 and sets errno to "ENOSYS" nevertheless. |
904 |
|
905 |
aio_sync_file_range $fh, $offset, $nbytes, $flags, $callback->($status) |
906 |
Sync the data portion of the file specified by $offset and $length |
907 |
to disk (but NOT the metadata), by calling the Linux-specific |
908 |
sync_file_range call. If sync_file_range is not available or it |
909 |
returns ENOSYS, then fdatasync or fsync is being substituted. |
910 |
|
911 |
$flags can be a combination of |
912 |
"IO::AIO::SYNC_FILE_RANGE_WAIT_BEFORE", |
913 |
"IO::AIO::SYNC_FILE_RANGE_WRITE" and |
914 |
"IO::AIO::SYNC_FILE_RANGE_WAIT_AFTER": refer to the sync_file_range |
915 |
manpage for details. |
916 |
|
917 |
aio_pathsync $pathname, $callback->($status) |
918 |
This request tries to open, fsync and close the given path. This is |
919 |
a composite request intended to sync directories after directory |
920 |
operations (E.g. rename). This might not work on all operating |
921 |
systems or have any specific effect, but usually it makes sure that |
922 |
directory changes get written to disc. It works for anything that |
923 |
can be opened for read-only, not just directories. |
924 |
|
925 |
Future versions of this function might fall back to other methods |
926 |
when "fsync" on the directory fails (such as calling "sync"). |
927 |
|
928 |
Passes 0 when everything went ok, and -1 on error. |
929 |
|
930 |
aio_msync $scalar, $offset = 0, $length = undef, flags = 0, |
931 |
$callback->($status) |
932 |
This is a rather advanced IO::AIO call, which only works on |
933 |
mmap(2)ed scalars (see the "IO::AIO::mmap" function, although it |
934 |
also works on data scalars managed by the Sys::Mmap or Mmap modules, |
935 |
note that the scalar must only be modified in-place while an aio |
936 |
operation is pending on it). |
937 |
|
938 |
It calls the "msync" function of your OS, if available, with the |
939 |
memory area starting at $offset in the string and ending $length |
940 |
bytes later. If $length is negative, counts from the end, and if |
941 |
$length is "undef", then it goes till the end of the string. The |
942 |
flags can be a combination of "IO::AIO::MS_ASYNC", |
943 |
"IO::AIO::MS_INVALIDATE" and "IO::AIO::MS_SYNC". |
944 |
|
945 |
aio_mtouch $scalar, $offset = 0, $length = undef, flags = 0, |
946 |
$callback->($status) |
947 |
This is a rather advanced IO::AIO call, which works best on |
948 |
mmap(2)ed scalars. |
949 |
|
950 |
It touches (reads or writes) all memory pages in the specified range |
951 |
inside the scalar. All caveats and parameters are the same as for |
952 |
"aio_msync", above, except for flags, which must be either 0 (which |
953 |
reads all pages and ensures they are instantiated) or |
954 |
"IO::AIO::MT_MODIFY", which modifies the memory pages (by reading |
955 |
and writing an octet from it, which dirties the page). |
956 |
|
957 |
aio_mlock $scalar, $offset = 0, $length = undef, $callback->($status) |
958 |
This is a rather advanced IO::AIO call, which works best on |
959 |
mmap(2)ed scalars. |
960 |
|
961 |
It reads in all the pages of the underlying storage into memory (if |
962 |
any) and locks them, so they are not getting swapped/paged out or |
963 |
removed. |
964 |
|
965 |
If $length is undefined, then the scalar will be locked till the |
966 |
end. |
967 |
|
968 |
On systems that do not implement "mlock", this function returns -1 |
969 |
and sets errno to "ENOSYS". |
970 |
|
971 |
Note that the corresponding "munlock" is synchronous and is |
972 |
documented under "MISCELLANEOUS FUNCTIONS". |
973 |
|
974 |
Example: open a file, mmap and mlock it - both will be undone when |
975 |
$data gets destroyed. |
976 |
|
977 |
open my $fh, "<", $path or die "$path: $!"; |
978 |
my $data; |
979 |
IO::AIO::mmap $data, -s $fh, IO::AIO::PROT_READ, IO::AIO::MAP_SHARED, $fh; |
980 |
aio_mlock $data; # mlock in background |
981 |
|
982 |
aio_mlockall $flags, $callback->($status) |
983 |
Calls the "mlockall" function with the given $flags (a combination |
984 |
of "IO::AIO::MCL_CURRENT" and "IO::AIO::MCL_FUTURE"). |
985 |
|
986 |
On systems that do not implement "mlockall", this function returns |
987 |
-1 and sets errno to "ENOSYS". |
988 |
|
989 |
Note that the corresponding "munlockall" is synchronous and is |
990 |
documented under "MISCELLANEOUS FUNCTIONS". |
991 |
|
992 |
Example: asynchronously lock all current and future pages into |
993 |
memory. |
994 |
|
995 |
aio_mlockall IO::AIO::MCL_FUTURE; |
996 |
|
997 |
aio_fiemap $fh, $start, $length, $flags, $count, $cb->(\@extents) |
998 |
Queries the extents of the given file (by calling the Linux "FIEMAP" |
999 |
ioctl, see <http://cvs.schmorp.de/IO-AIO/doc/fiemap.txt> for |
1000 |
details). If the ioctl is not available on your OS, then this |
1001 |
request will fail with "ENOSYS". |
1002 |
|
1003 |
$start is the starting offset to query extents for, $length is the |
1004 |
size of the range to query - if it is "undef", then the whole file |
1005 |
will be queried. |
1006 |
|
1007 |
$flags is a combination of flags ("IO::AIO::FIEMAP_FLAG_SYNC" or |
1008 |
"IO::AIO::FIEMAP_FLAG_XATTR" - "IO::AIO::FIEMAP_FLAGS_COMPAT" is |
1009 |
also exported), and is normally 0 or "IO::AIO::FIEMAP_FLAG_SYNC" to |
1010 |
query the data portion. |
1011 |
|
1012 |
$count is the maximum number of extent records to return. If it is |
1013 |
"undef", then IO::AIO queries all extents of the range. As a very |
1014 |
special case, if it is 0, then the callback receives the number of |
1015 |
extents instead of the extents themselves (which is unreliable, see |
1016 |
below). |
1017 |
|
1018 |
If an error occurs, the callback receives no arguments. The special |
1019 |
"errno" value "IO::AIO::EBADR" is available to test for flag errors. |
1020 |
|
1021 |
Otherwise, the callback receives an array reference with extent |
1022 |
structures. Each extent structure is an array reference itself, with |
1023 |
the following members: |
1024 |
|
1025 |
[$logical, $physical, $length, $flags] |
1026 |
|
1027 |
Flags is any combination of the following flag values (typically |
1028 |
either 0 or "IO::AIO::FIEMAP_EXTENT_LAST" (1)): |
1029 |
|
1030 |
"IO::AIO::FIEMAP_EXTENT_LAST", "IO::AIO::FIEMAP_EXTENT_UNKNOWN", |
1031 |
"IO::AIO::FIEMAP_EXTENT_DELALLOC", "IO::AIO::FIEMAP_EXTENT_ENCODED", |
1032 |
"IO::AIO::FIEMAP_EXTENT_DATA_ENCRYPTED", |
1033 |
"IO::AIO::FIEMAP_EXTENT_NOT_ALIGNED", |
1034 |
"IO::AIO::FIEMAP_EXTENT_DATA_INLINE", |
1035 |
"IO::AIO::FIEMAP_EXTENT_DATA_TAIL", |
1036 |
"IO::AIO::FIEMAP_EXTENT_UNWRITTEN", "IO::AIO::FIEMAP_EXTENT_MERGED" |
1037 |
or "IO::AIO::FIEMAP_EXTENT_SHARED". |
1038 |
|
1039 |
At the time of this writing (Linux 3.2), this requets is unreliable |
1040 |
unless $count is "undef", as the kernel has all sorts of bugs |
1041 |
preventing it to return all extents of a range for files with large |
1042 |
number of extents. The code works around all these issues if $count |
1043 |
is undef. |
1044 |
|
1045 |
aio_group $callback->(...) |
1046 |
This is a very special aio request: Instead of doing something, it |
1047 |
is a container for other aio requests, which is useful if you want |
1048 |
to bundle many requests into a single, composite, request with a |
1049 |
definite callback and the ability to cancel the whole request with |
1050 |
its subrequests. |
1051 |
|
1052 |
Returns an object of class IO::AIO::GRP. See its documentation below |
1053 |
for more info. |
1054 |
|
1055 |
Example: |
1056 |
|
1057 |
my $grp = aio_group sub { |
1058 |
print "all stats done\n"; |
1059 |
}; |
1060 |
|
1061 |
add $grp |
1062 |
(aio_stat ...), |
1063 |
(aio_stat ...), |
1064 |
...; |
1065 |
|
1066 |
aio_nop $callback->() |
1067 |
This is a special request - it does nothing in itself and is only |
1068 |
used for side effects, such as when you want to add a dummy request |
1069 |
to a group so that finishing the requests in the group depends on |
1070 |
executing the given code. |
1071 |
|
1072 |
While this request does nothing, it still goes through the execution |
1073 |
phase and still requires a worker thread. Thus, the callback will |
1074 |
not be executed immediately but only after other requests in the |
1075 |
queue have entered their execution phase. This can be used to |
1076 |
measure request latency. |
1077 |
|
1078 |
IO::AIO::aio_busy $fractional_seconds, $callback->() *NOT EXPORTED* |
1079 |
Mainly used for debugging and benchmarking, this aio request puts |
1080 |
one of the request workers to sleep for the given time. |
1081 |
|
1082 |
While it is theoretically handy to have simple I/O scheduling |
1083 |
requests like sleep and file handle readable/writable, the overhead |
1084 |
this creates is immense (it blocks a thread for a long time) so do |
1085 |
not use this function except to put your application under |
1086 |
artificial I/O pressure. |
1087 |
|
1088 |
IO::AIO::WD - multiple working directories |
1089 |
Your process only has one current working directory, which is used by |
1090 |
all threads. This makes it hard to use relative paths (some other |
1091 |
component could call "chdir" at any time, and it is hard to control when |
1092 |
the path will be used by IO::AIO). |
1093 |
|
1094 |
One solution for this is to always use absolute paths. This usually |
1095 |
works, but can be quite slow (the kernel has to walk the whole path on |
1096 |
every access), and can also be a hassle to implement. |
1097 |
|
1098 |
Newer POSIX systems have a number of functions (openat, fdopendir, |
1099 |
futimensat and so on) that make it possible to specify working |
1100 |
directories per operation. |
1101 |
|
1102 |
For portability, and because the clowns who "designed", or shall I |
1103 |
write, perpetrated this new interface were obviously half-drunk, this |
1104 |
abstraction cannot be perfect, though. |
1105 |
|
1106 |
IO::AIO allows you to convert directory paths into a so-called |
1107 |
IO::AIO::WD object. This object stores the canonicalised, absolute |
1108 |
version of the path, and on systems that allow it, also a directory file |
1109 |
descriptor. |
1110 |
|
1111 |
Everywhere where a pathname is accepted by IO::AIO (e.g. in "aio_stat" |
1112 |
or "aio_unlink"), one can specify an array reference with an IO::AIO::WD |
1113 |
object and a pathname instead (or the IO::AIO::WD object alone, which |
1114 |
gets interpreted as "[$wd, "."]"). If the pathname is absolute, the |
1115 |
IO::AIO::WD object is ignored, otherwise the pathname is resolved |
1116 |
relative to that IO::AIO::WD object. |
1117 |
|
1118 |
For example, to get a wd object for /etc and then stat passwd inside, |
1119 |
you would write: |
1120 |
|
1121 |
aio_wd "/etc", sub { |
1122 |
my $etcdir = shift; |
1123 |
|
1124 |
# although $etcdir can be undef on error, there is generally no reason |
1125 |
# to check for errors here, as aio_stat will fail with ENOENT |
1126 |
# when $etcdir is undef. |
1127 |
|
1128 |
aio_stat [$etcdir, "passwd"], sub { |
1129 |
# yay |
1130 |
}; |
1131 |
}; |
1132 |
|
1133 |
That "aio_wd" is a request and not a normal function shows that creating |
1134 |
an IO::AIO::WD object is itself a potentially blocking operation, which |
1135 |
is why it is done asynchronously. |
1136 |
|
1137 |
To stat the directory obtained with "aio_wd" above, one could write |
1138 |
either of the following three request calls: |
1139 |
|
1140 |
aio_lstat "/etc" , sub { ... # pathname as normal string |
1141 |
aio_lstat [$wd, "."], sub { ... # "." relative to $wd (i.e. $wd itself) |
1142 |
aio_lstat $wd , sub { ... # shorthand for the previous |
1143 |
|
1144 |
As with normal pathnames, IO::AIO keeps a copy of the working directory |
1145 |
object and the pathname string, so you could write the following without |
1146 |
causing any issues due to $path getting reused: |
1147 |
|
1148 |
my $path = [$wd, undef]; |
1149 |
|
1150 |
for my $name (qw(abc def ghi)) { |
1151 |
$path->[1] = $name; |
1152 |
aio_stat $path, sub { |
1153 |
# ... |
1154 |
}; |
1155 |
} |
1156 |
|
1157 |
There are some caveats: when directories get renamed (or deleted), the |
1158 |
pathname string doesn't change, so will point to the new directory (or |
1159 |
nowhere at all), while the directory fd, if available on the system, |
1160 |
will still point to the original directory. Most functions accepting a |
1161 |
pathname will use the directory fd on newer systems, and the string on |
1162 |
older systems. Some functions (such as realpath) will always rely on the |
1163 |
string form of the pathname. |
1164 |
|
1165 |
So this functionality is mainly useful to get some protection against |
1166 |
"chdir", to easily get an absolute path out of a relative path for |
1167 |
future reference, and to speed up doing many operations in the same |
1168 |
directory (e.g. when stat'ing all files in a directory). |
1169 |
|
1170 |
The following functions implement this working directory abstraction: |
1171 |
|
1172 |
aio_wd $pathname, $callback->($wd) |
1173 |
Asynchonously canonicalise the given pathname and convert it to an |
1174 |
IO::AIO::WD object representing it. If possible and supported on the |
1175 |
system, also open a directory fd to speed up pathname resolution |
1176 |
relative to this working directory. |
1177 |
|
1178 |
If something goes wrong, then "undef" is passwd to the callback |
1179 |
instead of a working directory object and $! is set appropriately. |
1180 |
Since passing "undef" as working directory component of a pathname |
1181 |
fails the request with "ENOENT", there is often no need for error |
1182 |
checking in the "aio_wd" callback, as future requests using the |
1183 |
value will fail in the expected way. |
1184 |
|
1185 |
IO::AIO::CWD |
1186 |
This is a compiletime constant (object) that represents the process |
1187 |
current working directory. |
1188 |
|
1189 |
Specifying this object as working directory object for a pathname is |
1190 |
as if the pathname would be specified directly, without a directory |
1191 |
object. For example, these calls are functionally identical: |
1192 |
|
1193 |
aio_stat "somefile", sub { ... }; |
1194 |
aio_stat [IO::AIO::CWD, "somefile"], sub { ... }; |
1195 |
|
1196 |
To recover the path associated with an IO::AIO::WD object, you can use |
1197 |
"aio_realpath": |
1198 |
|
1199 |
aio_realpath $wd, sub { |
1200 |
warn "path is $_[0]\n"; |
1201 |
}; |
1202 |
|
1203 |
Currently, "aio_statvfs" always, and "aio_rename" and "aio_rmdir" |
1204 |
sometimes, fall back to using an absolue path. |
1205 |
|
1206 |
IO::AIO::REQ CLASS |
1207 |
All non-aggregate "aio_*" functions return an object of this class when |
1208 |
called in non-void context. |
1209 |
|
1210 |
cancel $req |
1211 |
Cancels the request, if possible. Has the effect of skipping |
1212 |
execution when entering the execute state and skipping calling the |
1213 |
callback when entering the the result state, but will leave the |
1214 |
request otherwise untouched (with the exception of readdir). That |
1215 |
means that requests that currently execute will not be stopped and |
1216 |
resources held by the request will not be freed prematurely. |
1217 |
|
1218 |
cb $req $callback->(...) |
1219 |
Replace (or simply set) the callback registered to the request. |
1220 |
|
1221 |
IO::AIO::GRP CLASS |
1222 |
This class is a subclass of IO::AIO::REQ, so all its methods apply to |
1223 |
objects of this class, too. |
1224 |
|
1225 |
A IO::AIO::GRP object is a special request that can contain multiple |
1226 |
other aio requests. |
1227 |
|
1228 |
You create one by calling the "aio_group" constructing function with a |
1229 |
callback that will be called when all contained requests have entered |
1230 |
the "done" state: |
1231 |
|
1232 |
my $grp = aio_group sub { |
1233 |
print "all requests are done\n"; |
1234 |
}; |
1235 |
|
1236 |
You add requests by calling the "add" method with one or more |
1237 |
"IO::AIO::REQ" objects: |
1238 |
|
1239 |
$grp->add (aio_unlink "..."); |
1240 |
|
1241 |
add $grp aio_stat "...", sub { |
1242 |
$_[0] or return $grp->result ("error"); |
1243 |
|
1244 |
# add another request dynamically, if first succeeded |
1245 |
add $grp aio_open "...", sub { |
1246 |
$grp->result ("ok"); |
1247 |
}; |
1248 |
}; |
1249 |
|
1250 |
This makes it very easy to create composite requests (see the source of |
1251 |
"aio_move" for an application) that work and feel like simple requests. |
1252 |
|
1253 |
* The IO::AIO::GRP objects will be cleaned up during calls to |
1254 |
"IO::AIO::poll_cb", just like any other request. |
1255 |
|
1256 |
* They can be canceled like any other request. Canceling will cancel |
1257 |
not only the request itself, but also all requests it contains. |
1258 |
|
1259 |
* They can also can also be added to other IO::AIO::GRP objects. |
1260 |
|
1261 |
* You must not add requests to a group from within the group callback |
1262 |
(or any later time). |
1263 |
|
1264 |
Their lifetime, simplified, looks like this: when they are empty, they |
1265 |
will finish very quickly. If they contain only requests that are in the |
1266 |
"done" state, they will also finish. Otherwise they will continue to |
1267 |
exist. |
1268 |
|
1269 |
That means after creating a group you have some time to add requests |
1270 |
(precisely before the callback has been invoked, which is only done |
1271 |
within the "poll_cb"). And in the callbacks of those requests, you can |
1272 |
add further requests to the group. And only when all those requests have |
1273 |
finished will the the group itself finish. |
1274 |
|
1275 |
add $grp ... |
1276 |
$grp->add (...) |
1277 |
Add one or more requests to the group. Any type of IO::AIO::REQ can |
1278 |
be added, including other groups, as long as you do not create |
1279 |
circular dependencies. |
1280 |
|
1281 |
Returns all its arguments. |
1282 |
|
1283 |
$grp->cancel_subs |
1284 |
Cancel all subrequests and clears any feeder, but not the group |
1285 |
request itself. Useful when you queued a lot of events but got a |
1286 |
result early. |
1287 |
|
1288 |
The group request will finish normally (you cannot add requests to |
1289 |
the group). |
1290 |
|
1291 |
$grp->result (...) |
1292 |
Set the result value(s) that will be passed to the group callback |
1293 |
when all subrequests have finished and set the groups errno to the |
1294 |
current value of errno (just like calling "errno" without an error |
1295 |
number). By default, no argument will be passed and errno is zero. |
1296 |
|
1297 |
$grp->errno ([$errno]) |
1298 |
Sets the group errno value to $errno, or the current value of errno |
1299 |
when the argument is missing. |
1300 |
|
1301 |
Every aio request has an associated errno value that is restored |
1302 |
when the callback is invoked. This method lets you change this value |
1303 |
from its default (0). |
1304 |
|
1305 |
Calling "result" will also set errno, so make sure you either set $! |
1306 |
before the call to "result", or call c<errno> after it. |
1307 |
|
1308 |
feed $grp $callback->($grp) |
1309 |
Sets a feeder/generator on this group: every group can have an |
1310 |
attached generator that generates requests if idle. The idea behind |
1311 |
this is that, although you could just queue as many requests as you |
1312 |
want in a group, this might starve other requests for a potentially |
1313 |
long time. For example, "aio_scandir" might generate hundreds of |
1314 |
thousands of "aio_stat" requests, delaying any later requests for a |
1315 |
long time. |
1316 |
|
1317 |
To avoid this, and allow incremental generation of requests, you can |
1318 |
instead a group and set a feeder on it that generates those |
1319 |
requests. The feed callback will be called whenever there are few |
1320 |
enough (see "limit", below) requests active in the group itself and |
1321 |
is expected to queue more requests. |
1322 |
|
1323 |
The feed callback can queue as many requests as it likes (i.e. "add" |
1324 |
does not impose any limits). |
1325 |
|
1326 |
If the feed does not queue more requests when called, it will be |
1327 |
automatically removed from the group. |
1328 |
|
1329 |
If the feed limit is 0 when this method is called, it will be set to |
1330 |
2 automatically. |
1331 |
|
1332 |
Example: |
1333 |
|
1334 |
# stat all files in @files, but only ever use four aio requests concurrently: |
1335 |
|
1336 |
my $grp = aio_group sub { print "finished\n" }; |
1337 |
limit $grp 4; |
1338 |
feed $grp sub { |
1339 |
my $file = pop @files |
1340 |
or return; |
1341 |
|
1342 |
add $grp aio_stat $file, sub { ... }; |
1343 |
}; |
1344 |
|
1345 |
limit $grp $num |
1346 |
Sets the feeder limit for the group: The feeder will be called |
1347 |
whenever the group contains less than this many requests. |
1348 |
|
1349 |
Setting the limit to 0 will pause the feeding process. |
1350 |
|
1351 |
The default value for the limit is 0, but note that setting a feeder |
1352 |
automatically bumps it up to 2. |
1353 |
|
1354 |
SUPPORT FUNCTIONS |
1355 |
EVENT PROCESSING AND EVENT LOOP INTEGRATION |
1356 |
$fileno = IO::AIO::poll_fileno |
1357 |
Return the *request result pipe file descriptor*. This filehandle |
1358 |
must be polled for reading by some mechanism outside this module |
1359 |
(e.g. EV, Glib, select and so on, see below or the SYNOPSIS). If the |
1360 |
pipe becomes readable you have to call "poll_cb" to check the |
1361 |
results. |
1362 |
|
1363 |
See "poll_cb" for an example. |
1364 |
|
1365 |
IO::AIO::poll_cb |
1366 |
Process some requests that have reached the result phase (i.e. they |
1367 |
have been executed but the results are not yet reported). You have |
1368 |
to call this "regularly" to finish outstanding requests. |
1369 |
|
1370 |
Returns 0 if all events could be processed (or there were no events |
1371 |
to process), or -1 if it returned earlier for whatever reason. |
1372 |
Returns immediately when no events are outstanding. The amount of |
1373 |
events processed depends on the settings of "IO::AIO::max_poll_req", |
1374 |
"IO::AIO::max_poll_time" and "IO::AIO::max_outstanding". |
1375 |
|
1376 |
If not all requests were processed for whatever reason, the poll |
1377 |
file descriptor will still be ready when "poll_cb" returns, so |
1378 |
normally you don't have to do anything special to have it called |
1379 |
later. |
1380 |
|
1381 |
Apart from calling "IO::AIO::poll_cb" when the event filehandle |
1382 |
becomes ready, it can be beneficial to call this function from loops |
1383 |
which submit a lot of requests, to make sure the results get |
1384 |
processed when they become available and not just when the loop is |
1385 |
finished and the event loop takes over again. This function returns |
1386 |
very fast when there are no outstanding requests. |
1387 |
|
1388 |
Example: Install an Event watcher that automatically calls |
1389 |
IO::AIO::poll_cb with high priority (more examples can be found in |
1390 |
the SYNOPSIS section, at the top of this document): |
1391 |
|
1392 |
Event->io (fd => IO::AIO::poll_fileno, |
1393 |
poll => 'r', async => 1, |
1394 |
cb => \&IO::AIO::poll_cb); |
1395 |
|
1396 |
IO::AIO::poll_wait |
1397 |
Wait until either at least one request is in the result phase or no |
1398 |
requests are outstanding anymore. |
1399 |
|
1400 |
This is useful if you want to synchronously wait for some requests |
1401 |
to become ready, without actually handling them. |
1402 |
|
1403 |
See "nreqs" for an example. |
1404 |
|
1405 |
IO::AIO::poll |
1406 |
Waits until some requests have been handled. |
1407 |
|
1408 |
Returns the number of requests processed, but is otherwise strictly |
1409 |
equivalent to: |
1410 |
|
1411 |
IO::AIO::poll_wait, IO::AIO::poll_cb |
1412 |
|
1413 |
IO::AIO::flush |
1414 |
Wait till all outstanding AIO requests have been handled. |
1415 |
|
1416 |
Strictly equivalent to: |
1417 |
|
1418 |
IO::AIO::poll_wait, IO::AIO::poll_cb |
1419 |
while IO::AIO::nreqs; |
1420 |
|
1421 |
IO::AIO::max_poll_reqs $nreqs |
1422 |
IO::AIO::max_poll_time $seconds |
1423 |
These set the maximum number of requests (default 0, meaning |
1424 |
infinity) that are being processed by "IO::AIO::poll_cb" in one |
1425 |
call, respectively the maximum amount of time (default 0, meaning |
1426 |
infinity) spent in "IO::AIO::poll_cb" to process requests (more |
1427 |
correctly the mininum amount of time "poll_cb" is allowed to use). |
1428 |
|
1429 |
Setting "max_poll_time" to a non-zero value creates an overhead of |
1430 |
one syscall per request processed, which is not normally a problem |
1431 |
unless your callbacks are really really fast or your OS is really |
1432 |
really slow (I am not mentioning Solaris here). Using |
1433 |
"max_poll_reqs" incurs no overhead. |
1434 |
|
1435 |
Setting these is useful if you want to ensure some level of |
1436 |
interactiveness when perl is not fast enough to process all requests |
1437 |
in time. |
1438 |
|
1439 |
For interactive programs, values such as 0.01 to 0.1 should be fine. |
1440 |
|
1441 |
Example: Install an Event watcher that automatically calls |
1442 |
IO::AIO::poll_cb with low priority, to ensure that other parts of |
1443 |
the program get the CPU sometimes even under high AIO load. |
1444 |
|
1445 |
# try not to spend much more than 0.1s in poll_cb |
1446 |
IO::AIO::max_poll_time 0.1; |
1447 |
|
1448 |
# use a low priority so other tasks have priority |
1449 |
Event->io (fd => IO::AIO::poll_fileno, |
1450 |
poll => 'r', nice => 1, |
1451 |
cb => &IO::AIO::poll_cb); |
1452 |
|
1453 |
CONTROLLING THE NUMBER OF THREADS |
1454 |
IO::AIO::min_parallel $nthreads |
1455 |
Set the minimum number of AIO threads to $nthreads. The current |
1456 |
default is 8, which means eight asynchronous operations can execute |
1457 |
concurrently at any one time (the number of outstanding requests, |
1458 |
however, is unlimited). |
1459 |
|
1460 |
IO::AIO starts threads only on demand, when an AIO request is queued |
1461 |
and no free thread exists. Please note that queueing up a hundred |
1462 |
requests can create demand for a hundred threads, even if it turns |
1463 |
out that everything is in the cache and could have been processed |
1464 |
faster by a single thread. |
1465 |
|
1466 |
It is recommended to keep the number of threads relatively low, as |
1467 |
some Linux kernel versions will scale negatively with the number of |
1468 |
threads (higher parallelity => MUCH higher latency). With current |
1469 |
Linux 2.6 versions, 4-32 threads should be fine. |
1470 |
|
1471 |
Under most circumstances you don't need to call this function, as |
1472 |
the module selects a default that is suitable for low to moderate |
1473 |
load. |
1474 |
|
1475 |
IO::AIO::max_parallel $nthreads |
1476 |
Sets the maximum number of AIO threads to $nthreads. If more than |
1477 |
the specified number of threads are currently running, this function |
1478 |
kills them. This function blocks until the limit is reached. |
1479 |
|
1480 |
While $nthreads are zero, aio requests get queued but not executed |
1481 |
until the number of threads has been increased again. |
1482 |
|
1483 |
This module automatically runs "max_parallel 0" at program end, to |
1484 |
ensure that all threads are killed and that there are no outstanding |
1485 |
requests. |
1486 |
|
1487 |
Under normal circumstances you don't need to call this function. |
1488 |
|
1489 |
IO::AIO::max_idle $nthreads |
1490 |
Limit the number of threads (default: 4) that are allowed to idle |
1491 |
(i.e., threads that did not get a request to process within the idle |
1492 |
timeout (default: 10 seconds). That means if a thread becomes idle |
1493 |
while $nthreads other threads are also idle, it will free its |
1494 |
resources and exit. |
1495 |
|
1496 |
This is useful when you allow a large number of threads (e.g. 100 or |
1497 |
1000) to allow for extremely high load situations, but want to free |
1498 |
resources under normal circumstances (1000 threads can easily |
1499 |
consume 30MB of RAM). |
1500 |
|
1501 |
The default is probably ok in most situations, especially if thread |
1502 |
creation is fast. If thread creation is very slow on your system you |
1503 |
might want to use larger values. |
1504 |
|
1505 |
IO::AIO::idle_timeout $seconds |
1506 |
Sets the minimum idle timeout (default 10) after which worker |
1507 |
threads are allowed to exit. SEe "IO::AIO::max_idle". |
1508 |
|
1509 |
IO::AIO::max_outstanding $maxreqs |
1510 |
Sets the maximum number of outstanding requests to $nreqs. If you do |
1511 |
queue up more than this number of requests, the next call to |
1512 |
"IO::AIO::poll_cb" (and other functions calling "poll_cb", such as |
1513 |
"IO::AIO::flush" or "IO::AIO::poll") will block until the limit is |
1514 |
no longer exceeded. |
1515 |
|
1516 |
In other words, this setting does not enforce a queue limit, but can |
1517 |
be used to make poll functions block if the limit is exceeded. |
1518 |
|
1519 |
This is a very bad function to use in interactive programs because |
1520 |
it blocks, and a bad way to reduce concurrency because it is |
1521 |
inexact: Better use an "aio_group" together with a feed callback. |
1522 |
|
1523 |
It's main use is in scripts without an event loop - when you want to |
1524 |
stat a lot of files, you can write somehting like this: |
1525 |
|
1526 |
IO::AIO::max_outstanding 32; |
1527 |
|
1528 |
for my $path (...) { |
1529 |
aio_stat $path , ...; |
1530 |
IO::AIO::poll_cb; |
1531 |
} |
1532 |
|
1533 |
IO::AIO::flush; |
1534 |
|
1535 |
The call to "poll_cb" inside the loop will normally return |
1536 |
instantly, but as soon as more thna 32 reqeusts are in-flight, it |
1537 |
will block until some requests have been handled. This keeps the |
1538 |
loop from pushing a large number of "aio_stat" requests onto the |
1539 |
queue. |
1540 |
|
1541 |
The default value for "max_outstanding" is very large, so there is |
1542 |
no practical limit on the number of outstanding requests. |
1543 |
|
1544 |
STATISTICAL INFORMATION |
1545 |
IO::AIO::nreqs |
1546 |
Returns the number of requests currently in the ready, execute or |
1547 |
pending states (i.e. for which their callback has not been invoked |
1548 |
yet). |
1549 |
|
1550 |
Example: wait till there are no outstanding requests anymore: |
1551 |
|
1552 |
IO::AIO::poll_wait, IO::AIO::poll_cb |
1553 |
while IO::AIO::nreqs; |
1554 |
|
1555 |
IO::AIO::nready |
1556 |
Returns the number of requests currently in the ready state (not yet |
1557 |
executed). |
1558 |
|
1559 |
IO::AIO::npending |
1560 |
Returns the number of requests currently in the pending state |
1561 |
(executed, but not yet processed by poll_cb). |
1562 |
|
1563 |
MISCELLANEOUS FUNCTIONS |
1564 |
IO::AIO implements some functions that might be useful, but are not |
1565 |
asynchronous. |
1566 |
|
1567 |
IO::AIO::sendfile $ofh, $ifh, $offset, $count |
1568 |
Calls the "eio_sendfile_sync" function, which is like |
1569 |
"aio_sendfile", but is blocking (this makes most sense if you know |
1570 |
the input data is likely cached already and the output filehandle is |
1571 |
set to non-blocking operations). |
1572 |
|
1573 |
Returns the number of bytes copied, or -1 on error. |
1574 |
|
1575 |
IO::AIO::fadvise $fh, $offset, $len, $advice |
1576 |
Simply calls the "posix_fadvise" function (see its manpage for |
1577 |
details). The following advice constants are available: |
1578 |
"IO::AIO::FADV_NORMAL", "IO::AIO::FADV_SEQUENTIAL", |
1579 |
"IO::AIO::FADV_RANDOM", "IO::AIO::FADV_NOREUSE", |
1580 |
"IO::AIO::FADV_WILLNEED", "IO::AIO::FADV_DONTNEED". |
1581 |
|
1582 |
On systems that do not implement "posix_fadvise", this function |
1583 |
returns ENOSYS, otherwise the return value of "posix_fadvise". |
1584 |
|
1585 |
IO::AIO::madvise $scalar, $offset, $len, $advice |
1586 |
Simply calls the "posix_madvise" function (see its manpage for |
1587 |
details). The following advice constants are available: |
1588 |
"IO::AIO::MADV_NORMAL", "IO::AIO::MADV_SEQUENTIAL", |
1589 |
"IO::AIO::MADV_RANDOM", "IO::AIO::MADV_WILLNEED", |
1590 |
"IO::AIO::MADV_DONTNEED". |
1591 |
|
1592 |
On systems that do not implement "posix_madvise", this function |
1593 |
returns ENOSYS, otherwise the return value of "posix_madvise". |
1594 |
|
1595 |
IO::AIO::mprotect $scalar, $offset, $len, $protect |
1596 |
Simply calls the "mprotect" function on the preferably AIO::mmap'ed |
1597 |
$scalar (see its manpage for details). The following protect |
1598 |
constants are available: "IO::AIO::PROT_NONE", "IO::AIO::PROT_READ", |
1599 |
"IO::AIO::PROT_WRITE", "IO::AIO::PROT_EXEC". |
1600 |
|
1601 |
On systems that do not implement "mprotect", this function returns |
1602 |
ENOSYS, otherwise the return value of "mprotect". |
1603 |
|
1604 |
IO::AIO::mmap $scalar, $length, $prot, $flags, $fh[, $offset] |
1605 |
Memory-maps a file (or anonymous memory range) and attaches it to |
1606 |
the given $scalar, which will act like a string scalar. Returns true |
1607 |
on success, and false otherwise. |
1608 |
|
1609 |
The only operations allowed on the scalar are "substr"/"vec" that |
1610 |
don't change the string length, and most read-only operations such |
1611 |
as copying it or searching it with regexes and so on. |
1612 |
|
1613 |
Anything else is unsafe and will, at best, result in memory leaks. |
1614 |
|
1615 |
The memory map associated with the $scalar is automatically removed |
1616 |
when the $scalar is destroyed, or when the "IO::AIO::mmap" or |
1617 |
"IO::AIO::munmap" functions are called. |
1618 |
|
1619 |
This calls the "mmap"(2) function internally. See your system's |
1620 |
manual page for details on the $length, $prot and $flags parameters. |
1621 |
|
1622 |
The $length must be larger than zero and smaller than the actual |
1623 |
filesize. |
1624 |
|
1625 |
$prot is a combination of "IO::AIO::PROT_NONE", |
1626 |
"IO::AIO::PROT_EXEC", "IO::AIO::PROT_READ" and/or |
1627 |
"IO::AIO::PROT_WRITE", |
1628 |
|
1629 |
$flags can be a combination of "IO::AIO::MAP_SHARED" or |
1630 |
"IO::AIO::MAP_PRIVATE", or a number of system-specific flags (when |
1631 |
not available, the are defined as 0): "IO::AIO::MAP_ANONYMOUS" |
1632 |
(which is set to "MAP_ANON" if your system only provides this |
1633 |
constant), "IO::AIO::MAP_HUGETLB", "IO::AIO::MAP_LOCKED", |
1634 |
"IO::AIO::MAP_NORESERVE", "IO::AIO::MAP_POPULATE" or |
1635 |
"IO::AIO::MAP_NONBLOCK" |
1636 |
|
1637 |
If $fh is "undef", then a file descriptor of -1 is passed. |
1638 |
|
1639 |
$offset is the offset from the start of the file - it generally must |
1640 |
be a multiple of "IO::AIO::PAGESIZE" and defaults to 0. |
1641 |
|
1642 |
Example: |
1643 |
|
1644 |
use Digest::MD5; |
1645 |
use IO::AIO; |
1646 |
|
1647 |
open my $fh, "<verybigfile" |
1648 |
or die "$!"; |
1649 |
|
1650 |
IO::AIO::mmap my $data, -s $fh, IO::AIO::PROT_READ, IO::AIO::MAP_SHARED, $fh |
1651 |
or die "verybigfile: $!"; |
1652 |
|
1653 |
my $fast_md5 = md5 $data; |
1654 |
|
1655 |
IO::AIO::munmap $scalar |
1656 |
Removes a previous mmap and undefines the $scalar. |
1657 |
|
1658 |
IO::AIO::munlock $scalar, $offset = 0, $length = undef |
1659 |
Calls the "munlock" function, undoing the effects of a previous |
1660 |
"aio_mlock" call (see its description for details). |
1661 |
|
1662 |
IO::AIO::munlockall |
1663 |
Calls the "munlockall" function. |
1664 |
|
1665 |
On systems that do not implement "munlockall", this function returns |
1666 |
ENOSYS, otherwise the return value of "munlockall". |
1667 |
|
1668 |
IO::AIO::splice $r_fh, $r_off, $w_fh, $w_off, $length, $flags |
1669 |
Calls the GNU/Linux splice(2) syscall, if available. If $r_off or |
1670 |
$w_off are "undef", then "NULL" is passed for these, otherwise they |
1671 |
should be the file offset. |
1672 |
|
1673 |
$r_fh and $w_fh should not refer to the same file, as splice might |
1674 |
silently corrupt the data in this case. |
1675 |
|
1676 |
The following symbol flag values are available: |
1677 |
"IO::AIO::SPLICE_F_MOVE", "IO::AIO::SPLICE_F_NONBLOCK", |
1678 |
"IO::AIO::SPLICE_F_MORE" and "IO::AIO::SPLICE_F_GIFT". |
1679 |
|
1680 |
See the splice(2) manpage for details. |
1681 |
|
1682 |
IO::AIO::tee $r_fh, $w_fh, $length, $flags |
1683 |
Calls the GNU/Linux tee(2) syscall, see it's manpage and the |
1684 |
description for "IO::AIO::splice" above for details. |
1685 |
|
1686 |
$actual_size = IO::AIO::pipesize $r_fh[, $new_size] |
1687 |
Attempts to query or change the pipe buffer size. Obviously works |
1688 |
only on pipes, and currently works only on GNU/Linux systems, and |
1689 |
fails with -1/"ENOSYS" everywhere else. If anybody knows how to |
1690 |
influence pipe buffer size on other systems, drop me a note. |
1691 |
|
1692 |
EVENT LOOP INTEGRATION |
1693 |
It is recommended to use AnyEvent::AIO to integrate IO::AIO |
1694 |
automatically into many event loops: |
1695 |
|
1696 |
# AnyEvent integration (EV, Event, Glib, Tk, POE, urxvt, pureperl...) |
1697 |
use AnyEvent::AIO; |
1698 |
|
1699 |
You can also integrate IO::AIO manually into many event loops, here are |
1700 |
some examples of how to do this: |
1701 |
|
1702 |
# EV integration |
1703 |
my $aio_w = EV::io IO::AIO::poll_fileno, EV::READ, \&IO::AIO::poll_cb; |
1704 |
|
1705 |
# Event integration |
1706 |
Event->io (fd => IO::AIO::poll_fileno, |
1707 |
poll => 'r', |
1708 |
cb => \&IO::AIO::poll_cb); |
1709 |
|
1710 |
# Glib/Gtk2 integration |
1711 |
add_watch Glib::IO IO::AIO::poll_fileno, |
1712 |
in => sub { IO::AIO::poll_cb; 1 }; |
1713 |
|
1714 |
# Tk integration |
1715 |
Tk::Event::IO->fileevent (IO::AIO::poll_fileno, "", |
1716 |
readable => \&IO::AIO::poll_cb); |
1717 |
|
1718 |
# Danga::Socket integration |
1719 |
Danga::Socket->AddOtherFds (IO::AIO::poll_fileno => |
1720 |
\&IO::AIO::poll_cb); |
1721 |
|
1722 |
FORK BEHAVIOUR |
1723 |
Usage of pthreads in a program changes the semantics of fork |
1724 |
considerably. Specifically, only async-safe functions can be called |
1725 |
after fork. Perl doesn't know about this, so in general, you cannot call |
1726 |
fork with defined behaviour in perl if pthreads are involved. IO::AIO |
1727 |
uses pthreads, so this applies, but many other extensions and (for |
1728 |
inexplicable reasons) perl itself often is linked against pthreads, so |
1729 |
this limitation applies to quite a lot of perls. |
1730 |
|
1731 |
This module no longer tries to fight your OS, or POSIX. That means |
1732 |
IO::AIO only works in the process that loaded it. Forking is fully |
1733 |
supported, but using IO::AIO in the child is not. |
1734 |
|
1735 |
You might get around by not *using* IO::AIO before (or after) forking. |
1736 |
You could also try to call the IO::AIO::reinit function in the child: |
1737 |
|
1738 |
IO::AIO::reinit |
1739 |
Abandons all current requests and I/O threads and simply |
1740 |
reinitialises all data structures. This is not an operation |
1741 |
supported by any standards, but happens to work on GNU/Linux and |
1742 |
some newer BSD systems. |
1743 |
|
1744 |
The only reasonable use for this function is to call it after |
1745 |
forking, if "IO::AIO" was used in the parent. Calling it while |
1746 |
IO::AIO is active in the process will result in undefined behaviour. |
1747 |
Calling it at any time will also result in any undefined (by POSIX) |
1748 |
behaviour. |
1749 |
|
1750 |
MEMORY USAGE |
1751 |
Per-request usage: |
1752 |
|
1753 |
Each aio request uses - depending on your architecture - around 100-200 |
1754 |
bytes of memory. In addition, stat requests need a stat buffer (possibly |
1755 |
a few hundred bytes), readdir requires a result buffer and so on. Perl |
1756 |
scalars and other data passed into aio requests will also be locked and |
1757 |
will consume memory till the request has entered the done state. |
1758 |
|
1759 |
This is not awfully much, so queuing lots of requests is not usually a |
1760 |
problem. |
1761 |
|
1762 |
Per-thread usage: |
1763 |
|
1764 |
In the execution phase, some aio requests require more memory for |
1765 |
temporary buffers, and each thread requires a stack and other data |
1766 |
structures (usually around 16k-128k, depending on the OS). |
1767 |
|
1768 |
KNOWN BUGS |
1769 |
Known bugs will be fixed in the next release. |
1770 |
|
1771 |
SEE ALSO |
1772 |
AnyEvent::AIO for easy integration into event loops, Coro::AIO for a |
1773 |
more natural syntax. |
1774 |
|
1775 |
AUTHOR |
1776 |
Marc Lehmann <schmorp@schmorp.de> |
1777 |
http://home.schmorp.de/ |
1778 |
|