1 |
=head1 NAME |
2 |
|
3 |
libev - a high performance full-featured event loop written in C |
4 |
|
5 |
=head1 SYNOPSIS |
6 |
|
7 |
#include <ev.h> |
8 |
|
9 |
=head1 DESCRIPTION |
10 |
|
11 |
Libev is an event loop: you register interest in certain events (such as a |
12 |
file descriptor being readable or a timeout occuring), and it will manage |
13 |
these event sources and provide your program with events. |
14 |
|
15 |
To do this, it must take more or less complete control over your process |
16 |
(or thread) by executing the I<event loop> handler, and will then |
17 |
communicate events via a callback mechanism. |
18 |
|
19 |
You register interest in certain events by registering so-called I<event |
20 |
watchers>, which are relatively small C structures you initialise with the |
21 |
details of the event, and then hand it over to libev by I<starting> the |
22 |
watcher. |
23 |
|
24 |
=head1 FEATURES |
25 |
|
26 |
Libev supports select, poll, the linux-specific epoll and the bsd-specific |
27 |
kqueue mechanisms for file descriptor events, relative timers, absolute |
28 |
timers with customised rescheduling, signal events, process status change |
29 |
events (related to SIGCHLD), and event watchers dealing with the event |
30 |
loop mechanism itself (idle, prepare and check watchers). It also is quite |
31 |
fast (see this L<benchmark|http://libev.schmorp.de/bench.html> comparing |
32 |
it to libevent for example). |
33 |
|
34 |
=head1 CONVENTIONS |
35 |
|
36 |
Libev is very configurable. In this manual the default configuration |
37 |
will be described, which supports multiple event loops. For more info |
38 |
about various configuration options please have a look at the file |
39 |
F<README.embed> in the libev distribution. If libev was configured without |
40 |
support for multiple event loops, then all functions taking an initial |
41 |
argument of name C<loop> (which is always of type C<struct ev_loop *>) |
42 |
will not have this argument. |
43 |
|
44 |
=head1 TIME REPRESENTATION |
45 |
|
46 |
Libev represents time as a single floating point number, representing the |
47 |
(fractional) number of seconds since the (POSIX) epoch (somewhere near |
48 |
the beginning of 1970, details are complicated, don't ask). This type is |
49 |
called C<ev_tstamp>, which is what you should use too. It usually aliases |
50 |
to the double type in C. |
51 |
|
52 |
=head1 GLOBAL FUNCTIONS |
53 |
|
54 |
These functions can be called anytime, even before initialising the |
55 |
library in any way. |
56 |
|
57 |
=over 4 |
58 |
|
59 |
=item ev_tstamp ev_time () |
60 |
|
61 |
Returns the current time as libev would use it. Please note that the |
62 |
C<ev_now> function is usually faster and also often returns the timestamp |
63 |
you actually want to know. |
64 |
|
65 |
=item int ev_version_major () |
66 |
|
67 |
=item int ev_version_minor () |
68 |
|
69 |
You can find out the major and minor version numbers of the library |
70 |
you linked against by calling the functions C<ev_version_major> and |
71 |
C<ev_version_minor>. If you want, you can compare against the global |
72 |
symbols C<EV_VERSION_MAJOR> and C<EV_VERSION_MINOR>, which specify the |
73 |
version of the library your program was compiled against. |
74 |
|
75 |
Usually, it's a good idea to terminate if the major versions mismatch, |
76 |
as this indicates an incompatible change. Minor versions are usually |
77 |
compatible to older versions, so a larger minor version alone is usually |
78 |
not a problem. |
79 |
|
80 |
=item unsigned int ev_supported_backends () |
81 |
|
82 |
Return the set of all backends (i.e. their corresponding C<EV_BACKEND_*> |
83 |
value) compiled into this binary of libev (independent of their |
84 |
availability on the system you are running on). See C<ev_default_loop> for |
85 |
a description of the set values. |
86 |
|
87 |
=item unsigned int ev_recommended_backends () |
88 |
|
89 |
Return the set of all backends compiled into this binary of libev and also |
90 |
recommended for this platform. This set is often smaller than the one |
91 |
returned by C<ev_supported_backends>, as for example kqueue is broken on |
92 |
most BSDs and will not be autodetected unless you explicitly request it |
93 |
(assuming you know what you are doing). This is the set of backends that |
94 |
C<EVFLAG_AUTO> will probe for. |
95 |
|
96 |
=item ev_set_allocator (void *(*cb)(void *ptr, long size)) |
97 |
|
98 |
Sets the allocation function to use (the prototype is similar to the |
99 |
realloc C function, the semantics are identical). It is used to allocate |
100 |
and free memory (no surprises here). If it returns zero when memory |
101 |
needs to be allocated, the library might abort or take some potentially |
102 |
destructive action. The default is your system realloc function. |
103 |
|
104 |
You could override this function in high-availability programs to, say, |
105 |
free some memory if it cannot allocate memory, to use a special allocator, |
106 |
or even to sleep a while and retry until some memory is available. |
107 |
|
108 |
=item ev_set_syserr_cb (void (*cb)(const char *msg)); |
109 |
|
110 |
Set the callback function to call on a retryable syscall error (such |
111 |
as failed select, poll, epoll_wait). The message is a printable string |
112 |
indicating the system call or subsystem causing the problem. If this |
113 |
callback is set, then libev will expect it to remedy the sitution, no |
114 |
matter what, when it returns. That is, libev will generally retry the |
115 |
requested operation, or, if the condition doesn't go away, do bad stuff |
116 |
(such as abort). |
117 |
|
118 |
=back |
119 |
|
120 |
=head1 FUNCTIONS CONTROLLING THE EVENT LOOP |
121 |
|
122 |
An event loop is described by a C<struct ev_loop *>. The library knows two |
123 |
types of such loops, the I<default> loop, which supports signals and child |
124 |
events, and dynamically created loops which do not. |
125 |
|
126 |
If you use threads, a common model is to run the default event loop |
127 |
in your main thread (or in a separate thread) and for each thread you |
128 |
create, you also create another event loop. Libev itself does no locking |
129 |
whatsoever, so if you mix calls to the same event loop in different |
130 |
threads, make sure you lock (this is usually a bad idea, though, even if |
131 |
done correctly, because it's hideous and inefficient). |
132 |
|
133 |
=over 4 |
134 |
|
135 |
=item struct ev_loop *ev_default_loop (unsigned int flags) |
136 |
|
137 |
This will initialise the default event loop if it hasn't been initialised |
138 |
yet and return it. If the default loop could not be initialised, returns |
139 |
false. If it already was initialised it simply returns it (and ignores the |
140 |
flags. If that is troubling you, check C<ev_backend ()> afterwards). |
141 |
|
142 |
If you don't know what event loop to use, use the one returned from this |
143 |
function. |
144 |
|
145 |
The flags argument can be used to specify special behaviour or specific |
146 |
backends to use, and is usually specified as C<0> (or EVFLAG_AUTO). |
147 |
|
148 |
It supports the following flags: |
149 |
|
150 |
=over 4 |
151 |
|
152 |
=item C<EVFLAG_AUTO> |
153 |
|
154 |
The default flags value. Use this if you have no clue (it's the right |
155 |
thing, believe me). |
156 |
|
157 |
=item C<EVFLAG_NOENV> |
158 |
|
159 |
If this flag bit is ored into the flag value (or the program runs setuid |
160 |
or setgid) then libev will I<not> look at the environment variable |
161 |
C<LIBEV_FLAGS>. Otherwise (the default), this environment variable will |
162 |
override the flags completely if it is found in the environment. This is |
163 |
useful to try out specific backends to test their performance, or to work |
164 |
around bugs. |
165 |
|
166 |
=item C<EVBACKEND_SELECT> (value 1, portable select backend) |
167 |
|
168 |
This is your standard select(2) backend. Not I<completely> standard, as |
169 |
libev tries to roll its own fd_set with no limits on the number of fds, |
170 |
but if that fails, expect a fairly low limit on the number of fds when |
171 |
using this backend. It doesn't scale too well (O(highest_fd)), but its usually |
172 |
the fastest backend for a low number of fds. |
173 |
|
174 |
=item C<EVBACKEND_POLL> (value 2, poll backend, available everywhere except on windows) |
175 |
|
176 |
And this is your standard poll(2) backend. It's more complicated than |
177 |
select, but handles sparse fds better and has no artificial limit on the |
178 |
number of fds you can use (except it will slow down considerably with a |
179 |
lot of inactive fds). It scales similarly to select, i.e. O(total_fds). |
180 |
|
181 |
=item C<EVBACKEND_EPOLL> (value 4, Linux) |
182 |
|
183 |
For few fds, this backend is a bit little slower than poll and select, |
184 |
but it scales phenomenally better. While poll and select usually scale like |
185 |
O(total_fds) where n is the total number of fds (or the highest fd), epoll scales |
186 |
either O(1) or O(active_fds). |
187 |
|
188 |
While stopping and starting an I/O watcher in the same iteration will |
189 |
result in some caching, there is still a syscall per such incident |
190 |
(because the fd could point to a different file description now), so its |
191 |
best to avoid that. Also, dup()ed file descriptors might not work very |
192 |
well if you register events for both fds. |
193 |
|
194 |
Please note that epoll sometimes generates spurious notifications, so you |
195 |
need to use non-blocking I/O or other means to avoid blocking when no data |
196 |
(or space) is available. |
197 |
|
198 |
=item C<EVBACKEND_KQUEUE> (value 8, most BSD clones) |
199 |
|
200 |
Kqueue deserves special mention, as at the time of this writing, it |
201 |
was broken on all BSDs except NetBSD (usually it doesn't work with |
202 |
anything but sockets and pipes, except on Darwin, where of course its |
203 |
completely useless). For this reason its not being "autodetected" unless |
204 |
you explicitly specify the flags (i.e. you don't use EVFLAG_AUTO). |
205 |
|
206 |
It scales in the same way as the epoll backend, but the interface to the |
207 |
kernel is more efficient (which says nothing about its actual speed, of |
208 |
course). While starting and stopping an I/O watcher does not cause an |
209 |
extra syscall as with epoll, it still adds up to four event changes per |
210 |
incident, so its best to avoid that. |
211 |
|
212 |
=item C<EVBACKEND_DEVPOLL> (value 16, Solaris 8) |
213 |
|
214 |
This is not implemented yet (and might never be). |
215 |
|
216 |
=item C<EVBACKEND_PORT> (value 32, Solaris 10) |
217 |
|
218 |
This uses the Solaris 10 port mechanism. As with everything on Solaris, |
219 |
it's really slow, but it still scales very well (O(active_fds)). |
220 |
|
221 |
Please note that solaris ports can result in a lot of spurious |
222 |
notifications, so you need to use non-blocking I/O or other means to avoid |
223 |
blocking when no data (or space) is available. |
224 |
|
225 |
=item C<EVBACKEND_ALL> |
226 |
|
227 |
Try all backends (even potentially broken ones that wouldn't be tried |
228 |
with C<EVFLAG_AUTO>). Since this is a mask, you can do stuff such as |
229 |
C<EVBACKEND_ALL & ~EVBACKEND_KQUEUE>. |
230 |
|
231 |
=back |
232 |
|
233 |
If one or more of these are ored into the flags value, then only these |
234 |
backends will be tried (in the reverse order as given here). If none are |
235 |
specified, most compiled-in backend will be tried, usually in reverse |
236 |
order of their flag values :) |
237 |
|
238 |
=item struct ev_loop *ev_loop_new (unsigned int flags) |
239 |
|
240 |
Similar to C<ev_default_loop>, but always creates a new event loop that is |
241 |
always distinct from the default loop. Unlike the default loop, it cannot |
242 |
handle signal and child watchers, and attempts to do so will be greeted by |
243 |
undefined behaviour (or a failed assertion if assertions are enabled). |
244 |
|
245 |
=item ev_default_destroy () |
246 |
|
247 |
Destroys the default loop again (frees all memory and kernel state |
248 |
etc.). This stops all registered event watchers (by not touching them in |
249 |
any way whatsoever, although you cannot rely on this :). |
250 |
|
251 |
=item ev_loop_destroy (loop) |
252 |
|
253 |
Like C<ev_default_destroy>, but destroys an event loop created by an |
254 |
earlier call to C<ev_loop_new>. |
255 |
|
256 |
=item ev_default_fork () |
257 |
|
258 |
This function reinitialises the kernel state for backends that have |
259 |
one. Despite the name, you can call it anytime, but it makes most sense |
260 |
after forking, in either the parent or child process (or both, but that |
261 |
again makes little sense). |
262 |
|
263 |
You I<must> call this function in the child process after forking if and |
264 |
only if you want to use the event library in both processes. If you just |
265 |
fork+exec, you don't have to call it. |
266 |
|
267 |
The function itself is quite fast and it's usually not a problem to call |
268 |
it just in case after a fork. To make this easy, the function will fit in |
269 |
quite nicely into a call to C<pthread_atfork>: |
270 |
|
271 |
pthread_atfork (0, 0, ev_default_fork); |
272 |
|
273 |
At the moment, C<EVBACKEND_SELECT> and C<EVBACKEND_POLL> are safe to use |
274 |
without calling this function, so if you force one of those backends you |
275 |
do not need to care. |
276 |
|
277 |
=item ev_loop_fork (loop) |
278 |
|
279 |
Like C<ev_default_fork>, but acts on an event loop created by |
280 |
C<ev_loop_new>. Yes, you have to call this on every allocated event loop |
281 |
after fork, and how you do this is entirely your own problem. |
282 |
|
283 |
=item unsigned int ev_backend (loop) |
284 |
|
285 |
Returns one of the C<EVBACKEND_*> flags indicating the event backend in |
286 |
use. |
287 |
|
288 |
=item ev_tstamp ev_now (loop) |
289 |
|
290 |
Returns the current "event loop time", which is the time the event loop |
291 |
got events and started processing them. This timestamp does not change |
292 |
as long as callbacks are being processed, and this is also the base time |
293 |
used for relative timers. You can treat it as the timestamp of the event |
294 |
occuring (or more correctly, the mainloop finding out about it). |
295 |
|
296 |
=item ev_loop (loop, int flags) |
297 |
|
298 |
Finally, this is it, the event handler. This function usually is called |
299 |
after you initialised all your watchers and you want to start handling |
300 |
events. |
301 |
|
302 |
If the flags argument is specified as 0, it will not return until either |
303 |
no event watchers are active anymore or C<ev_unloop> was called. |
304 |
|
305 |
A flags value of C<EVLOOP_NONBLOCK> will look for new events, will handle |
306 |
those events and any outstanding ones, but will not block your process in |
307 |
case there are no events and will return after one iteration of the loop. |
308 |
|
309 |
A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if |
310 |
neccessary) and will handle those and any outstanding ones. It will block |
311 |
your process until at least one new event arrives, and will return after |
312 |
one iteration of the loop. |
313 |
|
314 |
This flags value could be used to implement alternative looping |
315 |
constructs, but the C<prepare> and C<check> watchers provide a better and |
316 |
more generic mechanism. |
317 |
|
318 |
Here are the gory details of what ev_loop does: |
319 |
|
320 |
1. If there are no active watchers (reference count is zero), return. |
321 |
2. Queue and immediately call all prepare watchers. |
322 |
3. If we have been forked, recreate the kernel state. |
323 |
4. Update the kernel state with all outstanding changes. |
324 |
5. Update the "event loop time". |
325 |
6. Calculate for how long to block. |
326 |
7. Block the process, waiting for events. |
327 |
8. Update the "event loop time" and do time jump handling. |
328 |
9. Queue all outstanding timers. |
329 |
10. Queue all outstanding periodics. |
330 |
11. If no events are pending now, queue all idle watchers. |
331 |
12. Queue all check watchers. |
332 |
13. Call all queued watchers in reverse order (i.e. check watchers first). |
333 |
14. If ev_unloop has been called or EVLOOP_ONESHOT or EVLOOP_NONBLOCK |
334 |
was used, return, otherwise continue with step #1. |
335 |
|
336 |
=item ev_unloop (loop, how) |
337 |
|
338 |
Can be used to make a call to C<ev_loop> return early (but only after it |
339 |
has processed all outstanding events). The C<how> argument must be either |
340 |
C<EVUNLOOP_ONE>, which will make the innermost C<ev_loop> call return, or |
341 |
C<EVUNLOOP_ALL>, which will make all nested C<ev_loop> calls return. |
342 |
|
343 |
=item ev_ref (loop) |
344 |
|
345 |
=item ev_unref (loop) |
346 |
|
347 |
Ref/unref can be used to add or remove a reference count on the event |
348 |
loop: Every watcher keeps one reference, and as long as the reference |
349 |
count is nonzero, C<ev_loop> will not return on its own. If you have |
350 |
a watcher you never unregister that should not keep C<ev_loop> from |
351 |
returning, ev_unref() after starting, and ev_ref() before stopping it. For |
352 |
example, libev itself uses this for its internal signal pipe: It is not |
353 |
visible to the libev user and should not keep C<ev_loop> from exiting if |
354 |
no event watchers registered by it are active. It is also an excellent |
355 |
way to do this for generic recurring timers or from within third-party |
356 |
libraries. Just remember to I<unref after start> and I<ref before stop>. |
357 |
|
358 |
=back |
359 |
|
360 |
=head1 ANATOMY OF A WATCHER |
361 |
|
362 |
A watcher is a structure that you create and register to record your |
363 |
interest in some event. For instance, if you want to wait for STDIN to |
364 |
become readable, you would create an C<ev_io> watcher for that: |
365 |
|
366 |
static void my_cb (struct ev_loop *loop, struct ev_io *w, int revents) |
367 |
{ |
368 |
ev_io_stop (w); |
369 |
ev_unloop (loop, EVUNLOOP_ALL); |
370 |
} |
371 |
|
372 |
struct ev_loop *loop = ev_default_loop (0); |
373 |
struct ev_io stdin_watcher; |
374 |
ev_init (&stdin_watcher, my_cb); |
375 |
ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ); |
376 |
ev_io_start (loop, &stdin_watcher); |
377 |
ev_loop (loop, 0); |
378 |
|
379 |
As you can see, you are responsible for allocating the memory for your |
380 |
watcher structures (and it is usually a bad idea to do this on the stack, |
381 |
although this can sometimes be quite valid). |
382 |
|
383 |
Each watcher structure must be initialised by a call to C<ev_init |
384 |
(watcher *, callback)>, which expects a callback to be provided. This |
385 |
callback gets invoked each time the event occurs (or, in the case of io |
386 |
watchers, each time the event loop detects that the file descriptor given |
387 |
is readable and/or writable). |
388 |
|
389 |
Each watcher type has its own C<< ev_<type>_set (watcher *, ...) >> macro |
390 |
with arguments specific to this watcher type. There is also a macro |
391 |
to combine initialisation and setting in one call: C<< ev_<type>_init |
392 |
(watcher *, callback, ...) >>. |
393 |
|
394 |
To make the watcher actually watch out for events, you have to start it |
395 |
with a watcher-specific start function (C<< ev_<type>_start (loop, watcher |
396 |
*) >>), and you can stop watching for events at any time by calling the |
397 |
corresponding stop function (C<< ev_<type>_stop (loop, watcher *) >>. |
398 |
|
399 |
As long as your watcher is active (has been started but not stopped) you |
400 |
must not touch the values stored in it. Most specifically you must never |
401 |
reinitialise it or call its set macro. |
402 |
|
403 |
You can check whether an event is active by calling the C<ev_is_active |
404 |
(watcher *)> macro. To see whether an event is outstanding (but the |
405 |
callback for it has not been called yet) you can use the C<ev_is_pending |
406 |
(watcher *)> macro. |
407 |
|
408 |
Each and every callback receives the event loop pointer as first, the |
409 |
registered watcher structure as second, and a bitset of received events as |
410 |
third argument. |
411 |
|
412 |
The received events usually include a single bit per event type received |
413 |
(you can receive multiple events at the same time). The possible bit masks |
414 |
are: |
415 |
|
416 |
=over 4 |
417 |
|
418 |
=item C<EV_READ> |
419 |
|
420 |
=item C<EV_WRITE> |
421 |
|
422 |
The file descriptor in the C<ev_io> watcher has become readable and/or |
423 |
writable. |
424 |
|
425 |
=item C<EV_TIMEOUT> |
426 |
|
427 |
The C<ev_timer> watcher has timed out. |
428 |
|
429 |
=item C<EV_PERIODIC> |
430 |
|
431 |
The C<ev_periodic> watcher has timed out. |
432 |
|
433 |
=item C<EV_SIGNAL> |
434 |
|
435 |
The signal specified in the C<ev_signal> watcher has been received by a thread. |
436 |
|
437 |
=item C<EV_CHILD> |
438 |
|
439 |
The pid specified in the C<ev_child> watcher has received a status change. |
440 |
|
441 |
=item C<EV_IDLE> |
442 |
|
443 |
The C<ev_idle> watcher has determined that you have nothing better to do. |
444 |
|
445 |
=item C<EV_PREPARE> |
446 |
|
447 |
=item C<EV_CHECK> |
448 |
|
449 |
All C<ev_prepare> watchers are invoked just I<before> C<ev_loop> starts |
450 |
to gather new events, and all C<ev_check> watchers are invoked just after |
451 |
C<ev_loop> has gathered them, but before it invokes any callbacks for any |
452 |
received events. Callbacks of both watcher types can start and stop as |
453 |
many watchers as they want, and all of them will be taken into account |
454 |
(for example, a C<ev_prepare> watcher might start an idle watcher to keep |
455 |
C<ev_loop> from blocking). |
456 |
|
457 |
=item C<EV_ERROR> |
458 |
|
459 |
An unspecified error has occured, the watcher has been stopped. This might |
460 |
happen because the watcher could not be properly started because libev |
461 |
ran out of memory, a file descriptor was found to be closed or any other |
462 |
problem. You best act on it by reporting the problem and somehow coping |
463 |
with the watcher being stopped. |
464 |
|
465 |
Libev will usually signal a few "dummy" events together with an error, |
466 |
for example it might indicate that a fd is readable or writable, and if |
467 |
your callbacks is well-written it can just attempt the operation and cope |
468 |
with the error from read() or write(). This will not work in multithreaded |
469 |
programs, though, so beware. |
470 |
|
471 |
=back |
472 |
|
473 |
=head2 ASSOCIATING CUSTOM DATA WITH A WATCHER |
474 |
|
475 |
Each watcher has, by default, a member C<void *data> that you can change |
476 |
and read at any time, libev will completely ignore it. This can be used |
477 |
to associate arbitrary data with your watcher. If you need more data and |
478 |
don't want to allocate memory and store a pointer to it in that data |
479 |
member, you can also "subclass" the watcher type and provide your own |
480 |
data: |
481 |
|
482 |
struct my_io |
483 |
{ |
484 |
struct ev_io io; |
485 |
int otherfd; |
486 |
void *somedata; |
487 |
struct whatever *mostinteresting; |
488 |
} |
489 |
|
490 |
And since your callback will be called with a pointer to the watcher, you |
491 |
can cast it back to your own type: |
492 |
|
493 |
static void my_cb (struct ev_loop *loop, struct ev_io *w_, int revents) |
494 |
{ |
495 |
struct my_io *w = (struct my_io *)w_; |
496 |
... |
497 |
} |
498 |
|
499 |
More interesting and less C-conformant ways of catsing your callback type |
500 |
have been omitted.... |
501 |
|
502 |
|
503 |
=head1 WATCHER TYPES |
504 |
|
505 |
This section describes each watcher in detail, but will not repeat |
506 |
information given in the last section. |
507 |
|
508 |
=head2 C<ev_io> - is this file descriptor readable or writable |
509 |
|
510 |
I/O watchers check whether a file descriptor is readable or writable |
511 |
in each iteration of the event loop (This behaviour is called |
512 |
level-triggering because you keep receiving events as long as the |
513 |
condition persists. Remember you can stop the watcher if you don't want to |
514 |
act on the event and neither want to receive future events). |
515 |
|
516 |
In general you can register as many read and/or write event watchers per |
517 |
fd as you want (as long as you don't confuse yourself). Setting all file |
518 |
descriptors to non-blocking mode is also usually a good idea (but not |
519 |
required if you know what you are doing). |
520 |
|
521 |
You have to be careful with dup'ed file descriptors, though. Some backends |
522 |
(the linux epoll backend is a notable example) cannot handle dup'ed file |
523 |
descriptors correctly if you register interest in two or more fds pointing |
524 |
to the same underlying file/socket etc. description (that is, they share |
525 |
the same underlying "file open"). |
526 |
|
527 |
If you must do this, then force the use of a known-to-be-good backend |
528 |
(at the time of this writing, this includes only C<EVBACKEND_SELECT> and |
529 |
C<EVBACKEND_POLL>). |
530 |
|
531 |
=over 4 |
532 |
|
533 |
=item ev_io_init (ev_io *, callback, int fd, int events) |
534 |
|
535 |
=item ev_io_set (ev_io *, int fd, int events) |
536 |
|
537 |
Configures an C<ev_io> watcher. The fd is the file descriptor to rceeive |
538 |
events for and events is either C<EV_READ>, C<EV_WRITE> or C<EV_READ | |
539 |
EV_WRITE> to receive the given events. |
540 |
|
541 |
Please note that most of the more scalable backend mechanisms (for example |
542 |
epoll and solaris ports) can result in spurious readyness notifications |
543 |
for file descriptors, so you practically need to use non-blocking I/O (and |
544 |
treat callback invocation as hint only), or retest separately with a safe |
545 |
interface before doing I/O (XLib can do this), or force the use of either |
546 |
C<EVBACKEND_SELECT> or C<EVBACKEND_POLL>, which don't suffer from this |
547 |
problem. Also note that it is quite easy to have your callback invoked |
548 |
when the readyness condition is no longer valid even when employing |
549 |
typical ways of handling events, so its a good idea to use non-blocking |
550 |
I/O unconditionally. |
551 |
|
552 |
=back |
553 |
|
554 |
=head2 C<ev_timer> - relative and optionally recurring timeouts |
555 |
|
556 |
Timer watchers are simple relative timers that generate an event after a |
557 |
given time, and optionally repeating in regular intervals after that. |
558 |
|
559 |
The timers are based on real time, that is, if you register an event that |
560 |
times out after an hour and you reset your system clock to last years |
561 |
time, it will still time out after (roughly) and hour. "Roughly" because |
562 |
detecting time jumps is hard, and some inaccuracies are unavoidable (the |
563 |
monotonic clock option helps a lot here). |
564 |
|
565 |
The relative timeouts are calculated relative to the C<ev_now ()> |
566 |
time. This is usually the right thing as this timestamp refers to the time |
567 |
of the event triggering whatever timeout you are modifying/starting. If |
568 |
you suspect event processing to be delayed and you I<need> to base the timeout |
569 |
on the current time, use something like this to adjust for this: |
570 |
|
571 |
ev_timer_set (&timer, after + ev_now () - ev_time (), 0.); |
572 |
|
573 |
The callback is guarenteed to be invoked only when its timeout has passed, |
574 |
but if multiple timers become ready during the same loop iteration then |
575 |
order of execution is undefined. |
576 |
|
577 |
=over 4 |
578 |
|
579 |
=item ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat) |
580 |
|
581 |
=item ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat) |
582 |
|
583 |
Configure the timer to trigger after C<after> seconds. If C<repeat> is |
584 |
C<0.>, then it will automatically be stopped. If it is positive, then the |
585 |
timer will automatically be configured to trigger again C<repeat> seconds |
586 |
later, again, and again, until stopped manually. |
587 |
|
588 |
The timer itself will do a best-effort at avoiding drift, that is, if you |
589 |
configure a timer to trigger every 10 seconds, then it will trigger at |
590 |
exactly 10 second intervals. If, however, your program cannot keep up with |
591 |
the timer (because it takes longer than those 10 seconds to do stuff) the |
592 |
timer will not fire more than once per event loop iteration. |
593 |
|
594 |
=item ev_timer_again (loop) |
595 |
|
596 |
This will act as if the timer timed out and restart it again if it is |
597 |
repeating. The exact semantics are: |
598 |
|
599 |
If the timer is started but nonrepeating, stop it. |
600 |
|
601 |
If the timer is repeating, either start it if necessary (with the repeat |
602 |
value), or reset the running timer to the repeat value. |
603 |
|
604 |
This sounds a bit complicated, but here is a useful and typical |
605 |
example: Imagine you have a tcp connection and you want a so-called idle |
606 |
timeout, that is, you want to be called when there have been, say, 60 |
607 |
seconds of inactivity on the socket. The easiest way to do this is to |
608 |
configure an C<ev_timer> with after=repeat=60 and calling ev_timer_again each |
609 |
time you successfully read or write some data. If you go into an idle |
610 |
state where you do not expect data to travel on the socket, you can stop |
611 |
the timer, and again will automatically restart it if need be. |
612 |
|
613 |
=back |
614 |
|
615 |
=head2 C<ev_periodic> - to cron or not to cron |
616 |
|
617 |
Periodic watchers are also timers of a kind, but they are very versatile |
618 |
(and unfortunately a bit complex). |
619 |
|
620 |
Unlike C<ev_timer>'s, they are not based on real time (or relative time) |
621 |
but on wallclock time (absolute time). You can tell a periodic watcher |
622 |
to trigger "at" some specific point in time. For example, if you tell a |
623 |
periodic watcher to trigger in 10 seconds (by specifiying e.g. c<ev_now () |
624 |
+ 10.>) and then reset your system clock to the last year, then it will |
625 |
take a year to trigger the event (unlike an C<ev_timer>, which would trigger |
626 |
roughly 10 seconds later and of course not if you reset your system time |
627 |
again). |
628 |
|
629 |
They can also be used to implement vastly more complex timers, such as |
630 |
triggering an event on eahc midnight, local time. |
631 |
|
632 |
As with timers, the callback is guarenteed to be invoked only when the |
633 |
time (C<at>) has been passed, but if multiple periodic timers become ready |
634 |
during the same loop iteration then order of execution is undefined. |
635 |
|
636 |
=over 4 |
637 |
|
638 |
=item ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb) |
639 |
|
640 |
=item ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb) |
641 |
|
642 |
Lots of arguments, lets sort it out... There are basically three modes of |
643 |
operation, and we will explain them from simplest to complex: |
644 |
|
645 |
=over 4 |
646 |
|
647 |
=item * absolute timer (interval = reschedule_cb = 0) |
648 |
|
649 |
In this configuration the watcher triggers an event at the wallclock time |
650 |
C<at> and doesn't repeat. It will not adjust when a time jump occurs, |
651 |
that is, if it is to be run at January 1st 2011 then it will run when the |
652 |
system time reaches or surpasses this time. |
653 |
|
654 |
=item * non-repeating interval timer (interval > 0, reschedule_cb = 0) |
655 |
|
656 |
In this mode the watcher will always be scheduled to time out at the next |
657 |
C<at + N * interval> time (for some integer N) and then repeat, regardless |
658 |
of any time jumps. |
659 |
|
660 |
This can be used to create timers that do not drift with respect to system |
661 |
time: |
662 |
|
663 |
ev_periodic_set (&periodic, 0., 3600., 0); |
664 |
|
665 |
This doesn't mean there will always be 3600 seconds in between triggers, |
666 |
but only that the the callback will be called when the system time shows a |
667 |
full hour (UTC), or more correctly, when the system time is evenly divisible |
668 |
by 3600. |
669 |
|
670 |
Another way to think about it (for the mathematically inclined) is that |
671 |
C<ev_periodic> will try to run the callback in this mode at the next possible |
672 |
time where C<time = at (mod interval)>, regardless of any time jumps. |
673 |
|
674 |
=item * manual reschedule mode (reschedule_cb = callback) |
675 |
|
676 |
In this mode the values for C<interval> and C<at> are both being |
677 |
ignored. Instead, each time the periodic watcher gets scheduled, the |
678 |
reschedule callback will be called with the watcher as first, and the |
679 |
current time as second argument. |
680 |
|
681 |
NOTE: I<This callback MUST NOT stop or destroy any periodic watcher, |
682 |
ever, or make any event loop modifications>. If you need to stop it, |
683 |
return C<now + 1e30> (or so, fudge fudge) and stop it afterwards (e.g. by |
684 |
starting a prepare watcher). |
685 |
|
686 |
Its prototype is C<ev_tstamp (*reschedule_cb)(struct ev_periodic *w, |
687 |
ev_tstamp now)>, e.g.: |
688 |
|
689 |
static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now) |
690 |
{ |
691 |
return now + 60.; |
692 |
} |
693 |
|
694 |
It must return the next time to trigger, based on the passed time value |
695 |
(that is, the lowest time value larger than to the second argument). It |
696 |
will usually be called just before the callback will be triggered, but |
697 |
might be called at other times, too. |
698 |
|
699 |
NOTE: I<< This callback must always return a time that is later than the |
700 |
passed C<now> value >>. Not even C<now> itself will do, it I<must> be larger. |
701 |
|
702 |
This can be used to create very complex timers, such as a timer that |
703 |
triggers on each midnight, local time. To do this, you would calculate the |
704 |
next midnight after C<now> and return the timestamp value for this. How |
705 |
you do this is, again, up to you (but it is not trivial, which is the main |
706 |
reason I omitted it as an example). |
707 |
|
708 |
=back |
709 |
|
710 |
=item ev_periodic_again (loop, ev_periodic *) |
711 |
|
712 |
Simply stops and restarts the periodic watcher again. This is only useful |
713 |
when you changed some parameters or the reschedule callback would return |
714 |
a different time than the last time it was called (e.g. in a crond like |
715 |
program when the crontabs have changed). |
716 |
|
717 |
=back |
718 |
|
719 |
=head2 C<ev_signal> - signal me when a signal gets signalled |
720 |
|
721 |
Signal watchers will trigger an event when the process receives a specific |
722 |
signal one or more times. Even though signals are very asynchronous, libev |
723 |
will try it's best to deliver signals synchronously, i.e. as part of the |
724 |
normal event processing, like any other event. |
725 |
|
726 |
You can configure as many watchers as you like per signal. Only when the |
727 |
first watcher gets started will libev actually register a signal watcher |
728 |
with the kernel (thus it coexists with your own signal handlers as long |
729 |
as you don't register any with libev). Similarly, when the last signal |
730 |
watcher for a signal is stopped libev will reset the signal handler to |
731 |
SIG_DFL (regardless of what it was set to before). |
732 |
|
733 |
=over 4 |
734 |
|
735 |
=item ev_signal_init (ev_signal *, callback, int signum) |
736 |
|
737 |
=item ev_signal_set (ev_signal *, int signum) |
738 |
|
739 |
Configures the watcher to trigger on the given signal number (usually one |
740 |
of the C<SIGxxx> constants). |
741 |
|
742 |
=back |
743 |
|
744 |
=head2 C<ev_child> - wait for pid status changes |
745 |
|
746 |
Child watchers trigger when your process receives a SIGCHLD in response to |
747 |
some child status changes (most typically when a child of yours dies). |
748 |
|
749 |
=over 4 |
750 |
|
751 |
=item ev_child_init (ev_child *, callback, int pid) |
752 |
|
753 |
=item ev_child_set (ev_child *, int pid) |
754 |
|
755 |
Configures the watcher to wait for status changes of process C<pid> (or |
756 |
I<any> process if C<pid> is specified as C<0>). The callback can look |
757 |
at the C<rstatus> member of the C<ev_child> watcher structure to see |
758 |
the status word (use the macros from C<sys/wait.h> and see your systems |
759 |
C<waitpid> documentation). The C<rpid> member contains the pid of the |
760 |
process causing the status change. |
761 |
|
762 |
=back |
763 |
|
764 |
=head2 C<ev_idle> - when you've got nothing better to do |
765 |
|
766 |
Idle watchers trigger events when there are no other events are pending |
767 |
(prepare, check and other idle watchers do not count). That is, as long |
768 |
as your process is busy handling sockets or timeouts (or even signals, |
769 |
imagine) it will not be triggered. But when your process is idle all idle |
770 |
watchers are being called again and again, once per event loop iteration - |
771 |
until stopped, that is, or your process receives more events and becomes |
772 |
busy. |
773 |
|
774 |
The most noteworthy effect is that as long as any idle watchers are |
775 |
active, the process will not block when waiting for new events. |
776 |
|
777 |
Apart from keeping your process non-blocking (which is a useful |
778 |
effect on its own sometimes), idle watchers are a good place to do |
779 |
"pseudo-background processing", or delay processing stuff to after the |
780 |
event loop has handled all outstanding events. |
781 |
|
782 |
=over 4 |
783 |
|
784 |
=item ev_idle_init (ev_signal *, callback) |
785 |
|
786 |
Initialises and configures the idle watcher - it has no parameters of any |
787 |
kind. There is a C<ev_idle_set> macro, but using it is utterly pointless, |
788 |
believe me. |
789 |
|
790 |
=back |
791 |
|
792 |
=head2 C<ev_prepare> and C<ev_check> - customise your event loop |
793 |
|
794 |
Prepare and check watchers are usually (but not always) used in tandem: |
795 |
prepare watchers get invoked before the process blocks and check watchers |
796 |
afterwards. |
797 |
|
798 |
Their main purpose is to integrate other event mechanisms into libev. This |
799 |
could be used, for example, to track variable changes, implement your own |
800 |
watchers, integrate net-snmp or a coroutine library and lots more. |
801 |
|
802 |
This is done by examining in each prepare call which file descriptors need |
803 |
to be watched by the other library, registering C<ev_io> watchers for |
804 |
them and starting an C<ev_timer> watcher for any timeouts (many libraries |
805 |
provide just this functionality). Then, in the check watcher you check for |
806 |
any events that occured (by checking the pending status of all watchers |
807 |
and stopping them) and call back into the library. The I/O and timer |
808 |
callbacks will never actually be called (but must be valid nevertheless, |
809 |
because you never know, you know?). |
810 |
|
811 |
As another example, the Perl Coro module uses these hooks to integrate |
812 |
coroutines into libev programs, by yielding to other active coroutines |
813 |
during each prepare and only letting the process block if no coroutines |
814 |
are ready to run (it's actually more complicated: it only runs coroutines |
815 |
with priority higher than or equal to the event loop and one coroutine |
816 |
of lower priority, but only once, using idle watchers to keep the event |
817 |
loop from blocking if lower-priority coroutines are active, thus mapping |
818 |
low-priority coroutines to idle/background tasks). |
819 |
|
820 |
=over 4 |
821 |
|
822 |
=item ev_prepare_init (ev_prepare *, callback) |
823 |
|
824 |
=item ev_check_init (ev_check *, callback) |
825 |
|
826 |
Initialises and configures the prepare or check watcher - they have no |
827 |
parameters of any kind. There are C<ev_prepare_set> and C<ev_check_set> |
828 |
macros, but using them is utterly, utterly and completely pointless. |
829 |
|
830 |
=back |
831 |
|
832 |
=head1 OTHER FUNCTIONS |
833 |
|
834 |
There are some other functions of possible interest. Described. Here. Now. |
835 |
|
836 |
=over 4 |
837 |
|
838 |
=item ev_once (loop, int fd, int events, ev_tstamp timeout, callback) |
839 |
|
840 |
This function combines a simple timer and an I/O watcher, calls your |
841 |
callback on whichever event happens first and automatically stop both |
842 |
watchers. This is useful if you want to wait for a single event on an fd |
843 |
or timeout without having to allocate/configure/start/stop/free one or |
844 |
more watchers yourself. |
845 |
|
846 |
If C<fd> is less than 0, then no I/O watcher will be started and events |
847 |
is being ignored. Otherwise, an C<ev_io> watcher for the given C<fd> and |
848 |
C<events> set will be craeted and started. |
849 |
|
850 |
If C<timeout> is less than 0, then no timeout watcher will be |
851 |
started. Otherwise an C<ev_timer> watcher with after = C<timeout> (and |
852 |
repeat = 0) will be started. While C<0> is a valid timeout, it is of |
853 |
dubious value. |
854 |
|
855 |
The callback has the type C<void (*cb)(int revents, void *arg)> and gets |
856 |
passed an C<revents> set like normal event callbacks (a combination of |
857 |
C<EV_ERROR>, C<EV_READ>, C<EV_WRITE> or C<EV_TIMEOUT>) and the C<arg> |
858 |
value passed to C<ev_once>: |
859 |
|
860 |
static void stdin_ready (int revents, void *arg) |
861 |
{ |
862 |
if (revents & EV_TIMEOUT) |
863 |
/* doh, nothing entered */; |
864 |
else if (revents & EV_READ) |
865 |
/* stdin might have data for us, joy! */; |
866 |
} |
867 |
|
868 |
ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0); |
869 |
|
870 |
=item ev_feed_event (loop, watcher, int events) |
871 |
|
872 |
Feeds the given event set into the event loop, as if the specified event |
873 |
had happened for the specified watcher (which must be a pointer to an |
874 |
initialised but not necessarily started event watcher). |
875 |
|
876 |
=item ev_feed_fd_event (loop, int fd, int revents) |
877 |
|
878 |
Feed an event on the given fd, as if a file descriptor backend detected |
879 |
the given events it. |
880 |
|
881 |
=item ev_feed_signal_event (loop, int signum) |
882 |
|
883 |
Feed an event as if the given signal occured (loop must be the default loop!). |
884 |
|
885 |
=back |
886 |
|
887 |
=head1 LIBEVENT EMULATION |
888 |
|
889 |
Libev offers a compatibility emulation layer for libevent. It cannot |
890 |
emulate the internals of libevent, so here are some usage hints: |
891 |
|
892 |
=over 4 |
893 |
|
894 |
=item * Use it by including <event.h>, as usual. |
895 |
|
896 |
=item * The following members are fully supported: ev_base, ev_callback, |
897 |
ev_arg, ev_fd, ev_res, ev_events. |
898 |
|
899 |
=item * Avoid using ev_flags and the EVLIST_*-macros, while it is |
900 |
maintained by libev, it does not work exactly the same way as in libevent (consider |
901 |
it a private API). |
902 |
|
903 |
=item * Priorities are not currently supported. Initialising priorities |
904 |
will fail and all watchers will have the same priority, even though there |
905 |
is an ev_pri field. |
906 |
|
907 |
=item * Other members are not supported. |
908 |
|
909 |
=item * The libev emulation is I<not> ABI compatible to libevent, you need |
910 |
to use the libev header file and library. |
911 |
|
912 |
=back |
913 |
|
914 |
=head1 C++ SUPPORT |
915 |
|
916 |
TBD. |
917 |
|
918 |
=head1 AUTHOR |
919 |
|
920 |
Marc Lehmann <libev@schmorp.de>. |
921 |
|