libev - a high performance full-featured event loop written in C

  #include <ev.h>

Libev is an event loop: you register interest in certain events (such as a file descriptor being readable or a timeout occuring), and it will manage these event sources and provide your program with events.

To do this, it must take more or less complete control over your process (or thread) by executing the event loop handler, and will then communicate events via a callback mechanism.

You register interest in certain events by registering so-called event watchers, which are relatively small C structures you initialise with the details of the event, and then hand it over to libev by starting the watcher.

Libev supports select, poll, the linux-specific epoll and the bsd-specific kqueue mechanisms for file descriptor events, relative timers, absolute timers with customised rescheduling, signal events, process status change events (related to SIGCHLD), and event watchers dealing with the event loop mechanism itself (idle, prepare and check watchers). It also is quite fast (see this benchmark comparing it to libevent for example).

Libev is very configurable. In this manual the default configuration will be described, which supports multiple event loops. For more info about various configuration options please have a look at the file README.embed in the libev distribution. If libev was configured without support for multiple event loops, then all functions taking an initial argument of name loop (which is always of type struct ev_loop *) will not have this argument.

Libev represents time as a single floating point number, representing the (fractional) number of seconds since the (POSIX) epoch (somewhere near the beginning of 1970, details are complicated, don't ask). This type is called ev_tstamp, which is what you should use too. It usually aliases to the double type in C.

These functions can be called anytime, even before initialising the library in any way.

ev_tstamp ev_time ()

Returns the current time as libev would use it.

int ev_version_major ()

int ev_version_minor ()

You can find out the major and minor version numbers of the library you linked against by calling the functions ev_version_major and ev_version_minor. If you want, you can compare against the global symbols EV_VERSION_MAJOR and EV_VERSION_MINOR, which specify the version of the library your program was compiled against.

Usually, it's a good idea to terminate if the major versions mismatch, as this indicates an incompatible change. Minor versions are usually compatible to older versions, so a larger minor version alone is usually not a problem.

ev_set_allocator (void *(*cb)(void *ptr, long size))

Sets the allocation function to use (the prototype is similar to the realloc C function, the semantics are identical). It is used to allocate and free memory (no surprises here). If it returns zero when memory needs to be allocated, the library might abort or take some potentially destructive action. The default is your system realloc function.

You could override this function in high-availability programs to, say, free some memory if it cannot allocate memory, to use a special allocator, or even to sleep a while and retry until some memory is available.

ev_set_syserr_cb (void (*cb)(const char *msg));

Set the callback function to call on a retryable syscall error (such as failed select, poll, epoll_wait). The message is a printable string indicating the system call or subsystem causing the problem. If this callback is set, then libev will expect it to remedy the sitution, no matter what, when it returns. That is, libev will generally retry the requested operation, or, if the condition doesn't go away, do bad stuff (such as abort).

An event loop is described by a struct ev_loop *. The library knows two types of such loops, the default loop, which supports signals and child events, and dynamically created loops which do not.

If you use threads, a common model is to run the default event loop in your main thread (or in a separate thread) and for each thread you create, you also create another event loop. Libev itself does no locking whatsoever, so if you mix calls to the same event loop in different threads, make sure you lock (this is usually a bad idea, though, even if done correctly, because it's hideous and inefficient).

struct ev_loop *ev_default_loop (unsigned int flags)

This will initialise the default event loop if it hasn't been initialised yet and return it. If the default loop could not be initialised, returns false. If it already was initialised it simply returns it (and ignores the flags).

If you don't know what event loop to use, use the one returned from this function.

The flags argument can be used to specify special behaviour or specific backends to use, and is usually specified as 0 (or EVFLAG_AUTO).

It supports the following flags:

EVFLAG_AUTO

The default flags value. Use this if you have no clue (it's the right thing, believe me).

EVFLAG_NOENV

If this flag bit is ored into the flag value (or the program runs setuid or setgid) then libev will not look at the environment variable LIBEV_FLAGS. Otherwise (the default), this environment variable will override the flags completely if it is found in the environment. This is useful to try out specific backends to test their performance, or to work around bugs.

EVMETHOD_SELECT (portable select backend)

EVMETHOD_POLL (poll backend, available everywhere except on windows)

EVMETHOD_EPOLL (linux only)

EVMETHOD_KQUEUE (some bsds only)

EVMETHOD_DEVPOLL (solaris 8 only)

EVMETHOD_PORT (solaris 10 only)

If one or more of these are ored into the flags value, then only these backends will be tried (in the reverse order as given here). If one are specified, any backend will do.

struct ev_loop *ev_loop_new (unsigned int flags)

Similar to ev_default_loop, but always creates a new event loop that is always distinct from the default loop. Unlike the default loop, it cannot handle signal and child watchers, and attempts to do so will be greeted by undefined behaviour (or a failed assertion if assertions are enabled).

ev_default_destroy ()

Destroys the default loop again (frees all memory and kernel state etc.). This stops all registered event watchers (by not touching them in any way whatsoever, although you cannot rely on this :).

ev_loop_destroy (loop)

Like ev_default_destroy, but destroys an event loop created by an earlier call to ev_loop_new.

ev_default_fork ()

This function reinitialises the kernel state for backends that have one. Despite the name, you can call it anytime, but it makes most sense after forking, in either the parent or child process (or both, but that again makes little sense).

You must call this function after forking if and only if you want to use the event library in both processes. If you just fork+exec, you don't have to call it.

The function itself is quite fast and it's usually not a problem to call it just in case after a fork. To make this easy, the function will fit in quite nicely into a call to pthread_atfork:

    pthread_atfork (0, 0, ev_default_fork);

ev_loop_fork (loop)

Like ev_default_fork, but acts on an event loop created by ev_loop_new. Yes, you have to call this on every allocated event loop after fork, and how you do this is entirely your own problem.

unsigned int ev_method (loop)

Returns one of the EVMETHOD_* flags indicating the event backend in use.

ev_tstamp ev_now (loop)

Returns the current "event loop time", which is the time the event loop got events and started processing them. This timestamp does not change as long as callbacks are being processed, and this is also the base time used for relative timers. You can treat it as the timestamp of the event occuring (or more correctly, the mainloop finding out about it).

ev_loop (loop, int flags)

Finally, this is it, the event handler. This function usually is called after you initialised all your watchers and you want to start handling events.

If the flags argument is specified as 0, it will not return until either no event watchers are active anymore or ev_unloop was called.

A flags value of EVLOOP_NONBLOCK will look for new events, will handle those events and any outstanding ones, but will not block your process in case there are no events and will return after one iteration of the loop.

A flags value of EVLOOP_ONESHOT will look for new events (waiting if neccessary) and will handle those and any outstanding ones. It will block your process until at least one new event arrives, and will return after one iteration of the loop.

This flags value could be used to implement alternative looping constructs, but the prepare and check watchers provide a better and more generic mechanism.

ev_unloop (loop, how)

Can be used to make a call to ev_loop return early (but only after it has processed all outstanding events). The how argument must be either EVUNLOOP_ONCE, which will make the innermost ev_loop call return, or EVUNLOOP_ALL, which will make all nested ev_loop calls return.

ev_ref (loop)

ev_unref (loop)

Ref/unref can be used to add or remove a reference count on the event loop: Every watcher keeps one reference, and as long as the reference count is nonzero, ev_loop will not return on its own. If you have a watcher you never unregister that should not keep ev_loop from returning, ev_unref() after starting, and ev_ref() before stopping it. For example, libev itself uses this for its internal signal pipe: It is not visible to the libev user and should not keep ev_loop from exiting if no event watchers registered by it are active. It is also an excellent way to do this for generic recurring timers or from within third-party libraries. Just remember to unref after start and ref before stop.

A watcher is a structure that you create and register to record your interest in some event. For instance, if you want to wait for STDIN to become readable, you would create an ev_io watcher for that:

  static void my_cb (struct ev_loop *loop, struct ev_io *w, int revents)
  {
    ev_io_stop (w);
    ev_unloop (loop, EVUNLOOP_ALL);
  }

  struct ev_loop *loop = ev_default_loop (0);
  struct ev_io stdin_watcher;
  ev_init (&stdin_watcher, my_cb);
  ev_io_set (&stdin_watcher, STDIN_FILENO, EV_READ);
  ev_io_start (loop, &stdin_watcher);
  ev_loop (loop, 0);

As you can see, you are responsible for allocating the memory for your watcher structures (and it is usually a bad idea to do this on the stack, although this can sometimes be quite valid).

Each watcher structure must be initialised by a call to ev_init (watcher *, callback), which expects a callback to be provided. This callback gets invoked each time the event occurs (or, in the case of io watchers, each time the event loop detects that the file descriptor given is readable and/or writable).

Each watcher type has its own ev_<type>_set (watcher *, ...) macro with arguments specific to this watcher type. There is also a macro to combine initialisation and setting in one call: ev_<type>_init (watcher *, callback, ...).

To make the watcher actually watch out for events, you have to start it with a watcher-specific start function (ev_<type>_start (loop, watcher *)), and you can stop watching for events at any time by calling the corresponding stop function (ev_<type>_stop (loop, watcher *).

As long as your watcher is active (has been started but not stopped) you must not touch the values stored in it. Most specifically you must never reinitialise it or call its set method.

You can check whether an event is active by calling the ev_is_active (watcher *) macro. To see whether an event is outstanding (but the callback for it has not been called yet) you can use the ev_is_pending (watcher *) macro.

Each and every callback receives the event loop pointer as first, the registered watcher structure as second, and a bitset of received events as third argument.

The received events usually include a single bit per event type received (you can receive multiple events at the same time). The possible bit masks are:

EV_READ

EV_WRITE

The file descriptor in the ev_io watcher has become readable and/or writable.

EV_TIMEOUT

The ev_timer watcher has timed out.

EV_PERIODIC

The ev_periodic watcher has timed out.

EV_SIGNAL

The signal specified in the ev_signal watcher has been received by a thread.

EV_CHILD

The pid specified in the ev_child watcher has received a status change.

EV_IDLE

The ev_idle watcher has determined that you have nothing better to do.

EV_PREPARE

EV_CHECK

All ev_prepare watchers are invoked just before ev_loop starts to gather new events, and all ev_check watchers are invoked just after ev_loop has gathered them, but before it invokes any callbacks for any received events. Callbacks of both watcher types can start and stop as many watchers as they want, and all of them will be taken into account (for example, a ev_prepare watcher might start an idle watcher to keep ev_loop from blocking).

EV_ERROR

An unspecified error has occured, the watcher has been stopped. This might happen because the watcher could not be properly started because libev ran out of memory, a file descriptor was found to be closed or any other problem. You best act on it by reporting the problem and somehow coping with the watcher being stopped.

Libev will usually signal a few "dummy" events together with an error, for example it might indicate that a fd is readable or writable, and if your callbacks is well-written it can just attempt the operation and cope with the error from read() or write(). This will not work in multithreaded programs, though, so beware.

Each watcher has, by default, a member void *data that you can change and read at any time, libev will completely ignore it. This can be used to associate arbitrary data with your watcher. If you need more data and don't want to allocate memory and store a pointer to it in that data member, you can also "subclass" the watcher type and provide your own data:

  struct my_io
  {
    struct ev_io io;
    int otherfd;
    void *somedata;
    struct whatever *mostinteresting;
  }

And since your callback will be called with a pointer to the watcher, you can cast it back to your own type:

  static void my_cb (struct ev_loop *loop, struct ev_io *w_, int revents)
  {
    struct my_io *w = (struct my_io *)w_;
    ...
  }

More interesting and less C-conformant ways of catsing your callback type have been omitted....

This section describes each watcher in detail, but will not repeat information given in the last section.

I/O watchers check whether a file descriptor is readable or writable in each iteration of the event loop (This behaviour is called level-triggering because you keep receiving events as long as the condition persists. Remember you can stop the watcher if you don't want to act on the event and neither want to receive future events).

In general you can register as many read and/or write event watchers per fd as you want (as long as you don't confuse yourself). Setting all file descriptors to non-blocking mode is also usually a good idea (but not required if you know what you are doing).

You have to be careful with dup'ed file descriptors, though. Some backends (the linux epoll backend is a notable example) cannot handle dup'ed file descriptors correctly if you register interest in two or more fds pointing to the same file/socket etc. description (that is, they share the same underlying "file open").

If you must do this, then force the use of a known-to-be-good backend (at the time of this writing, this includes only EVMETHOD_SELECT and EVMETHOD_POLL).

ev_io_init (ev_io *, callback, int fd, int events)

ev_io_set (ev_io *, int fd, int events)

Configures an ev_io watcher. The fd is the file descriptor to rceeive events for and events is either EV_READ, EV_WRITE or EV_READ | EV_WRITE to receive the given events.

Timer watchers are simple relative timers that generate an event after a given time, and optionally repeating in regular intervals after that.

The timers are based on real time, that is, if you register an event that times out after an hour and you reset your system clock to last years time, it will still time out after (roughly) and hour. "Roughly" because detecting time jumps is hard, and soem inaccuracies are unavoidable (the monotonic clock option helps a lot here).

The relative timeouts are calculated relative to the ev_now () time. This is usually the right thing as this timestamp refers to the time of the event triggering whatever timeout you are modifying/starting. If you suspect event processing to be delayed and you *need* to base the timeout on the current time, use something like this to adjust for this:

   ev_timer_set (&timer, after + ev_now () - ev_time (), 0.);

ev_timer_init (ev_timer *, callback, ev_tstamp after, ev_tstamp repeat)

ev_timer_set (ev_timer *, ev_tstamp after, ev_tstamp repeat)

Configure the timer to trigger after after seconds. If repeat is 0., then it will automatically be stopped. If it is positive, then the timer will automatically be configured to trigger again repeat seconds later, again, and again, until stopped manually.

The timer itself will do a best-effort at avoiding drift, that is, if you configure a timer to trigger every 10 seconds, then it will trigger at exactly 10 second intervals. If, however, your program cannot keep up with the timer (because it takes longer than those 10 seconds to do stuff) the timer will not fire more than once per event loop iteration.

ev_timer_again (loop)

This will act as if the timer timed out and restart it again if it is repeating. The exact semantics are:

If the timer is started but nonrepeating, stop it.

If the timer is repeating, either start it if necessary (with the repeat value), or reset the running timer to the repeat value.

This sounds a bit complicated, but here is a useful and typical example: Imagine you have a tcp connection and you want a so-called idle timeout, that is, you want to be called when there have been, say, 60 seconds of inactivity on the socket. The easiest way to do this is to configure an ev_timer with after=repeat=60 and calling ev_timer_again each time you successfully read or write some data. If you go into an idle state where you do not expect data to travel on the socket, you can stop the timer, and again will automatically restart it if need be.

Periodic watchers are also timers of a kind, but they are very versatile (and unfortunately a bit complex).

Unlike ev_timer's, they are not based on real time (or relative time) but on wallclock time (absolute time). You can tell a periodic watcher to trigger "at" some specific point in time. For example, if you tell a periodic watcher to trigger in 10 seconds (by specifiying e.g. c<ev_now () + 10.>) and then reset your system clock to the last year, then it will take a year to trigger the event (unlike an ev_timer, which would trigger roughly 10 seconds later and of course not if you reset your system time again).

They can also be used to implement vastly more complex timers, such as triggering an event on eahc midnight, local time.

ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)

ev_periodic_set (ev_periodic *, ev_tstamp after, ev_tstamp repeat, reschedule_cb)

Lots of arguments, lets sort it out... There are basically three modes of operation, and we will explain them from simplest to complex:

* absolute timer (interval = reschedule_cb = 0)

In this configuration the watcher triggers an event at the wallclock time at and doesn't repeat. It will not adjust when a time jump occurs, that is, if it is to be run at January 1st 2011 then it will run when the system time reaches or surpasses this time.

* non-repeating interval timer (interval > 0, reschedule_cb = 0)

In this mode the watcher will always be scheduled to time out at the next at + N * interval time (for some integer N) and then repeat, regardless of any time jumps.

This can be used to create timers that do not drift with respect to system time:

   ev_periodic_set (&periodic, 0., 3600., 0);

This doesn't mean there will always be 3600 seconds in between triggers, but only that the the callback will be called when the system time shows a full hour (UTC), or more correctly, when the system time is evenly divisible by 3600.

Another way to think about it (for the mathematically inclined) is that ev_periodic will try to run the callback in this mode at the next possible time where time = at (mod interval), regardless of any time jumps.

* manual reschedule mode (reschedule_cb = callback)

In this mode the values for interval and at are both being ignored. Instead, each time the periodic watcher gets scheduled, the reschedule callback will be called with the watcher as first, and the current time as second argument.

NOTE: This callback MUST NOT stop or destroy any periodic watcher, ever, or make any event loop modifications. If you need to stop it, return now + 1e30 (or so, fudge fudge) and stop it afterwards (e.g. by starting a prepare watcher).

Its prototype is ev_tstamp (*reschedule_cb)(struct ev_periodic *w, ev_tstamp now), e.g.:

   static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now)
   {
     return now + 60.;
   }

It must return the next time to trigger, based on the passed time value (that is, the lowest time value larger than to the second argument). It will usually be called just before the callback will be triggered, but might be called at other times, too.

NOTE: This callback must always return a time that is later than the passed now value. Not even now itself will do, it must be larger.

This can be used to create very complex timers, such as a timer that triggers on each midnight, local time. To do this, you would calculate the next midnight after now and return the timestamp value for this. How you do this is, again, up to you (but it is not trivial, which is the main reason I omitted it as an example).

ev_periodic_again (loop, ev_periodic *)

Simply stops and restarts the periodic watcher again. This is only useful when you changed some parameters or the reschedule callback would return a different time than the last time it was called (e.g. in a crond like program when the crontabs have changed).

Signal watchers will trigger an event when the process receives a specific signal one or more times. Even though signals are very asynchronous, libev will try it's best to deliver signals synchronously, i.e. as part of the normal event processing, like any other event.

You can configure as many watchers as you like per signal. Only when the first watcher gets started will libev actually register a signal watcher with the kernel (thus it coexists with your own signal handlers as long as you don't register any with libev). Similarly, when the last signal watcher for a signal is stopped libev will reset the signal handler to SIG_DFL (regardless of what it was set to before).

ev_signal_init (ev_signal *, callback, int signum)

ev_signal_set (ev_signal *, int signum)

Configures the watcher to trigger on the given signal number (usually one of the SIGxxx constants).

Child watchers trigger when your process receives a SIGCHLD in response to some child status changes (most typically when a child of yours dies).

ev_child_init (ev_child *, callback, int pid)

ev_child_set (ev_child *, int pid)

Configures the watcher to wait for status changes of process pid (or any process if pid is specified as 0). The callback can look at the rstatus member of the ev_child watcher structure to see the status word (use the macros from sys/wait.h and see your systems waitpid documentation). The rpid member contains the pid of the process causing the status change.

Idle watchers trigger events when there are no other events are pending (prepare, check and other idle watchers do not count). That is, as long as your process is busy handling sockets or timeouts (or even signals, imagine) it will not be triggered. But when your process is idle all idle watchers are being called again and again, once per event loop iteration - until stopped, that is, or your process receives more events and becomes busy.

The most noteworthy effect is that as long as any idle watchers are active, the process will not block when waiting for new events.

Apart from keeping your process non-blocking (which is a useful effect on its own sometimes), idle watchers are a good place to do "pseudo-background processing", or delay processing stuff to after the event loop has handled all outstanding events.

ev_idle_init (ev_signal *, callback)

Initialises and configures the idle watcher - it has no parameters of any kind. There is a ev_idle_set macro, but using it is utterly pointless, believe me.

Prepare and check watchers are usually (but not always) used in tandem: prepare watchers get invoked before the process blocks and check watchers afterwards.

Their main purpose is to integrate other event mechanisms into libev. This could be used, for example, to track variable changes, implement your own watchers, integrate net-snmp or a coroutine library and lots more.

This is done by examining in each prepare call which file descriptors need to be watched by the other library, registering ev_io watchers for them and starting an ev_timer watcher for any timeouts (many libraries provide just this functionality). Then, in the check watcher you check for any events that occured (by checking the pending status of all watchers and stopping them) and call back into the library. The I/O and timer callbacks will never actually be called (but must be valid nevertheless, because you never know, you know?).

As another example, the Perl Coro module uses these hooks to integrate coroutines into libev programs, by yielding to other active coroutines during each prepare and only letting the process block if no coroutines are ready to run (it's actually more complicated: it only runs coroutines with priority higher than or equal to the event loop and one coroutine of lower priority, but only once, using idle watchers to keep the event loop from blocking if lower-priority coroutines are active, thus mapping low-priority coroutines to idle/background tasks).

ev_prepare_init (ev_prepare *, callback)

ev_check_init (ev_check *, callback)

Initialises and configures the prepare or check watcher - they have no parameters of any kind. There are ev_prepare_set and ev_check_set macros, but using them is utterly, utterly and completely pointless.

There are some other functions of possible interest. Described. Here. Now.

ev_once (loop, int fd, int events, ev_tstamp timeout, callback)

This function combines a simple timer and an I/O watcher, calls your callback on whichever event happens first and automatically stop both watchers. This is useful if you want to wait for a single event on an fd or timeout without having to allocate/configure/start/stop/free one or more watchers yourself.

If fd is less than 0, then no I/O watcher will be started and events is being ignored. Otherwise, an ev_io watcher for the given fd and events set will be craeted and started.

If timeout is less than 0, then no timeout watcher will be started. Otherwise an ev_timer watcher with after = timeout (and repeat = 0) will be started. While 0 is a valid timeout, it is of dubious value.

The callback has the type void (*cb)(int revents, void *arg) and gets passed an revents set like normal event callbacks (a combination of EV_ERROR, EV_READ, EV_WRITE or EV_TIMEOUT) and the arg value passed to ev_once:

  static void stdin_ready (int revents, void *arg)
  {
    if (revents & EV_TIMEOUT)
      /* doh, nothing entered */;
    else if (revents & EV_READ)
      /* stdin might have data for us, joy! */;
  }

  ev_once (STDIN_FILENO, EV_READ, 10., stdin_ready, 0);

ev_feed_event (loop, watcher, int events)

Feeds the given event set into the event loop, as if the specified event had happened for the specified watcher (which must be a pointer to an initialised but not necessarily started event watcher).

ev_feed_fd_event (loop, int fd, int revents)

Feed an event on the given fd, as if a file descriptor backend detected the given events it.

ev_feed_signal_event (loop, int signum)

Feed an event as if the given signal occured (loop must be the default loop!).

TBD.

TBD.

Marc Lehmann <libev@schmorp.de>.