… | |
… | |
298 | If you don't know what event loop to use, use the one returned from this |
298 | If you don't know what event loop to use, use the one returned from this |
299 | function. |
299 | function. |
300 | |
300 | |
301 | Note that this function is I<not> thread-safe, so if you want to use it |
301 | Note that this function is I<not> thread-safe, so if you want to use it |
302 | from multiple threads, you have to lock (note also that this is unlikely, |
302 | from multiple threads, you have to lock (note also that this is unlikely, |
303 | as loops cannot bes hared easily between threads anyway). |
303 | as loops cannot be shared easily between threads anyway). |
304 | |
304 | |
305 | The default loop is the only loop that can handle C<ev_signal> and |
305 | The default loop is the only loop that can handle C<ev_signal> and |
306 | C<ev_child> watchers, and to do this, it always registers a handler |
306 | C<ev_child> watchers, and to do this, it always registers a handler |
307 | for C<SIGCHLD>. If this is a problem for your application you can either |
307 | for C<SIGCHLD>. If this is a problem for your application you can either |
308 | create a dynamic loop with C<ev_loop_new> that doesn't do that, or you |
308 | create a dynamic loop with C<ev_loop_new> that doesn't do that, or you |
… | |
… | |
384 | =item C<EVBACKEND_EPOLL> (value 4, Linux) |
384 | =item C<EVBACKEND_EPOLL> (value 4, Linux) |
385 | |
385 | |
386 | For few fds, this backend is a bit little slower than poll and select, |
386 | For few fds, this backend is a bit little slower than poll and select, |
387 | but it scales phenomenally better. While poll and select usually scale |
387 | but it scales phenomenally better. While poll and select usually scale |
388 | like O(total_fds) where n is the total number of fds (or the highest fd), |
388 | like O(total_fds) where n is the total number of fds (or the highest fd), |
389 | epoll scales either O(1) or O(active_fds). The epoll design has a number |
389 | epoll scales either O(1) or O(active_fds). |
390 | of shortcomings, such as silently dropping events in some hard-to-detect |
390 | |
391 | cases and requiring a system call per fd change, no fork support and bad |
391 | The epoll mechanism deserves honorable mention as the most misdesigned |
392 | support for dup. |
392 | of the more advanced event mechanisms: mere annoyances include silently |
|
|
393 | dropping file descriptors, requiring a system call per change per file |
|
|
394 | descriptor (and unnecessary guessing of parameters), problems with dup and |
|
|
395 | so on. The biggest issue is fork races, however - if a program forks then |
|
|
396 | I<both> parent and child process have to recreate the epoll set, which can |
|
|
397 | take considerable time (one syscall per file descriptor) and is of course |
|
|
398 | hard to detect. |
|
|
399 | |
|
|
400 | Epoll is also notoriously buggy - embedding epoll fds I<should> work, but |
|
|
401 | of course I<doesn't>, and epoll just loves to report events for totally |
|
|
402 | I<different> file descriptors (even already closed ones, so one cannot |
|
|
403 | even remove them from the set) than registered in the set (especially |
|
|
404 | on SMP systems). Libev tries to counter these spurious notifications by |
|
|
405 | employing an additional generation counter and comparing that against the |
|
|
406 | events to filter out spurious ones, recreating the set when required. |
393 | |
407 | |
394 | While stopping, setting and starting an I/O watcher in the same iteration |
408 | While stopping, setting and starting an I/O watcher in the same iteration |
395 | will result in some caching, there is still a system call per such incident |
409 | will result in some caching, there is still a system call per such |
396 | (because the fd could point to a different file description now), so its |
410 | incident (because the same I<file descriptor> could point to a different |
397 | best to avoid that. Also, C<dup ()>'ed file descriptors might not work |
411 | I<file description> now), so its best to avoid that. Also, C<dup ()>'ed |
398 | very well if you register events for both fds. |
412 | file descriptors might not work very well if you register events for both |
399 | |
413 | file descriptors. |
400 | Please note that epoll sometimes generates spurious notifications, so you |
|
|
401 | need to use non-blocking I/O or other means to avoid blocking when no data |
|
|
402 | (or space) is available. |
|
|
403 | |
414 | |
404 | Best performance from this backend is achieved by not unregistering all |
415 | Best performance from this backend is achieved by not unregistering all |
405 | watchers for a file descriptor until it has been closed, if possible, |
416 | watchers for a file descriptor until it has been closed, if possible, |
406 | i.e. keep at least one watcher active per fd at all times. Stopping and |
417 | i.e. keep at least one watcher active per fd at all times. Stopping and |
407 | starting a watcher (without re-setting it) also usually doesn't cause |
418 | starting a watcher (without re-setting it) also usually doesn't cause |
408 | extra overhead. |
419 | extra overhead. A fork can both result in spurious notifications as well |
|
|
420 | as in libev having to destroy and recreate the epoll object, which can |
|
|
421 | take considerable time and thus should be avoided. |
409 | |
422 | |
410 | While nominally embeddable in other event loops, this feature is broken in |
423 | While nominally embeddable in other event loops, this feature is broken in |
411 | all kernel versions tested so far. |
424 | all kernel versions tested so far. |
412 | |
425 | |
413 | This backend maps C<EV_READ> and C<EV_WRITE> in the same way as |
426 | This backend maps C<EV_READ> and C<EV_WRITE> in the same way as |
414 | C<EVBACKEND_POLL>. |
427 | C<EVBACKEND_POLL>. |
415 | |
428 | |
416 | =item C<EVBACKEND_KQUEUE> (value 8, most BSD clones) |
429 | =item C<EVBACKEND_KQUEUE> (value 8, most BSD clones) |
417 | |
430 | |
418 | Kqueue deserves special mention, as at the time of this writing, it was |
431 | Kqueue deserves special mention, as at the time of this writing, it |
419 | broken on all BSDs except NetBSD (usually it doesn't work reliably with |
432 | was broken on all BSDs except NetBSD (usually it doesn't work reliably |
420 | anything but sockets and pipes, except on Darwin, where of course it's |
433 | with anything but sockets and pipes, except on Darwin, where of course |
421 | completely useless). For this reason it's not being "auto-detected" unless |
434 | it's completely useless). Unlike epoll, however, whose brokenness |
422 | you explicitly specify it in the flags (i.e. using C<EVBACKEND_KQUEUE>) or |
435 | is by design, these kqueue bugs can (and eventually will) be fixed |
423 | libev was compiled on a known-to-be-good (-enough) system like NetBSD. |
436 | without API changes to existing programs. For this reason it's not being |
|
|
437 | "auto-detected" unless you explicitly specify it in the flags (i.e. using |
|
|
438 | C<EVBACKEND_KQUEUE>) or libev was compiled on a known-to-be-good (-enough) |
|
|
439 | system like NetBSD. |
424 | |
440 | |
425 | You still can embed kqueue into a normal poll or select backend and use it |
441 | You still can embed kqueue into a normal poll or select backend and use it |
426 | only for sockets (after having made sure that sockets work with kqueue on |
442 | only for sockets (after having made sure that sockets work with kqueue on |
427 | the target platform). See C<ev_embed> watchers for more info. |
443 | the target platform). See C<ev_embed> watchers for more info. |
428 | |
444 | |
429 | It scales in the same way as the epoll backend, but the interface to the |
445 | It scales in the same way as the epoll backend, but the interface to the |
430 | kernel is more efficient (which says nothing about its actual speed, of |
446 | kernel is more efficient (which says nothing about its actual speed, of |
431 | course). While stopping, setting and starting an I/O watcher does never |
447 | course). While stopping, setting and starting an I/O watcher does never |
432 | cause an extra system call as with C<EVBACKEND_EPOLL>, it still adds up to |
448 | cause an extra system call as with C<EVBACKEND_EPOLL>, it still adds up to |
433 | two event changes per incident. Support for C<fork ()> is very bad and it |
449 | two event changes per incident. Support for C<fork ()> is very bad (but |
434 | drops fds silently in similarly hard-to-detect cases. |
450 | sane, unlike epoll) and it drops fds silently in similarly hard-to-detect |
|
|
451 | cases |
435 | |
452 | |
436 | This backend usually performs well under most conditions. |
453 | This backend usually performs well under most conditions. |
437 | |
454 | |
438 | While nominally embeddable in other event loops, this doesn't work |
455 | While nominally embeddable in other event loops, this doesn't work |
439 | everywhere, so you might need to test for this. And since it is broken |
456 | everywhere, so you might need to test for this. And since it is broken |
… | |
… | |
468 | might perform better. |
485 | might perform better. |
469 | |
486 | |
470 | On the positive side, with the exception of the spurious readiness |
487 | On the positive side, with the exception of the spurious readiness |
471 | notifications, this backend actually performed fully to specification |
488 | notifications, this backend actually performed fully to specification |
472 | in all tests and is fully embeddable, which is a rare feat among the |
489 | in all tests and is fully embeddable, which is a rare feat among the |
473 | OS-specific backends. |
490 | OS-specific backends (I vastly prefer correctness over speed hacks). |
474 | |
491 | |
475 | This backend maps C<EV_READ> and C<EV_WRITE> in the same way as |
492 | This backend maps C<EV_READ> and C<EV_WRITE> in the same way as |
476 | C<EVBACKEND_POLL>. |
493 | C<EVBACKEND_POLL>. |
477 | |
494 | |
478 | =item C<EVBACKEND_ALL> |
495 | =item C<EVBACKEND_ALL> |
… | |
… | |
635 | the loop. |
652 | the loop. |
636 | |
653 | |
637 | A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if |
654 | A flags value of C<EVLOOP_ONESHOT> will look for new events (waiting if |
638 | necessary) and will handle those and any already outstanding ones. It |
655 | necessary) and will handle those and any already outstanding ones. It |
639 | will block your process until at least one new event arrives (which could |
656 | will block your process until at least one new event arrives (which could |
640 | be an event internal to libev itself, so there is no guarentee that a |
657 | be an event internal to libev itself, so there is no guarantee that a |
641 | user-registered callback will be called), and will return after one |
658 | user-registered callback will be called), and will return after one |
642 | iteration of the loop. |
659 | iteration of the loop. |
643 | |
660 | |
644 | This is useful if you are waiting for some external event in conjunction |
661 | This is useful if you are waiting for some external event in conjunction |
645 | with something not expressible using other libev watchers (i.e. "roll your |
662 | with something not expressible using other libev watchers (i.e. "roll your |
… | |
… | |
1910 | |
1927 | |
1911 | |
1928 | |
1912 | =head2 C<ev_stat> - did the file attributes just change? |
1929 | =head2 C<ev_stat> - did the file attributes just change? |
1913 | |
1930 | |
1914 | This watches a file system path for attribute changes. That is, it calls |
1931 | This watches a file system path for attribute changes. That is, it calls |
1915 | C<stat> regularly (or when the OS says it changed) and sees if it changed |
1932 | C<stat> on that path in regular intervals (or when the OS says it changed) |
1916 | compared to the last time, invoking the callback if it did. |
1933 | and sees if it changed compared to the last time, invoking the callback if |
|
|
1934 | it did. |
1917 | |
1935 | |
1918 | The path does not need to exist: changing from "path exists" to "path does |
1936 | The path does not need to exist: changing from "path exists" to "path does |
1919 | not exist" is a status change like any other. The condition "path does |
1937 | not exist" is a status change like any other. The condition "path does not |
1920 | not exist" is signified by the C<st_nlink> field being zero (which is |
1938 | exist" (or more correctly "path cannot be stat'ed") is signified by the |
1921 | otherwise always forced to be at least one) and all the other fields of |
1939 | C<st_nlink> field being zero (which is otherwise always forced to be at |
1922 | the stat buffer having unspecified contents. |
1940 | least one) and all the other fields of the stat buffer having unspecified |
|
|
1941 | contents. |
1923 | |
1942 | |
1924 | The path I<should> be absolute and I<must not> end in a slash. If it is |
1943 | The path I<must not> end in a slash or contain special components such as |
|
|
1944 | C<.> or C<..>. The path I<should> be absolute: If it is relative and |
1925 | relative and your working directory changes, the behaviour is undefined. |
1945 | your working directory changes, then the behaviour is undefined. |
1926 | |
1946 | |
1927 | Since there is no standard kernel interface to do this, the portable |
1947 | Since there is no portable change notification interface available, the |
1928 | implementation simply calls C<stat (2)> regularly on the path to see if |
1948 | portable implementation simply calls C<stat(2)> regularly on the path |
1929 | it changed somehow. You can specify a recommended polling interval for |
1949 | to see if it changed somehow. You can specify a recommended polling |
1930 | this case. If you specify a polling interval of C<0> (highly recommended!) |
1950 | interval for this case. If you specify a polling interval of C<0> (highly |
1931 | then a I<suitable, unspecified default> value will be used (which |
1951 | recommended!) then a I<suitable, unspecified default> value will be used |
1932 | you can expect to be around five seconds, although this might change |
1952 | (which you can expect to be around five seconds, although this might |
1933 | dynamically). Libev will also impose a minimum interval which is currently |
1953 | change dynamically). Libev will also impose a minimum interval which is |
1934 | around C<0.1>, but thats usually overkill. |
1954 | currently around C<0.1>, but that's usually overkill. |
1935 | |
1955 | |
1936 | This watcher type is not meant for massive numbers of stat watchers, |
1956 | This watcher type is not meant for massive numbers of stat watchers, |
1937 | as even with OS-supported change notifications, this can be |
1957 | as even with OS-supported change notifications, this can be |
1938 | resource-intensive. |
1958 | resource-intensive. |
1939 | |
1959 | |
1940 | At the time of this writing, the only OS-specific interface implemented |
1960 | At the time of this writing, the only OS-specific interface implemented |
1941 | is the Linux inotify interface (implementing kqueue support is left as |
1961 | is the Linux inotify interface (implementing kqueue support is left as an |
1942 | an exercise for the reader. Note, however, that the author sees no way |
1962 | exercise for the reader. Note, however, that the author sees no way of |
1943 | of implementing C<ev_stat> semantics with kqueue). |
1963 | implementing C<ev_stat> semantics with kqueue, except as a hint). |
1944 | |
1964 | |
1945 | =head3 ABI Issues (Largefile Support) |
1965 | =head3 ABI Issues (Largefile Support) |
1946 | |
1966 | |
1947 | Libev by default (unless the user overrides this) uses the default |
1967 | Libev by default (unless the user overrides this) uses the default |
1948 | compilation environment, which means that on systems with large file |
1968 | compilation environment, which means that on systems with large file |
1949 | support disabled by default, you get the 32 bit version of the stat |
1969 | support disabled by default, you get the 32 bit version of the stat |
1950 | structure. When using the library from programs that change the ABI to |
1970 | structure. When using the library from programs that change the ABI to |
1951 | use 64 bit file offsets the programs will fail. In that case you have to |
1971 | use 64 bit file offsets the programs will fail. In that case you have to |
1952 | compile libev with the same flags to get binary compatibility. This is |
1972 | compile libev with the same flags to get binary compatibility. This is |
1953 | obviously the case with any flags that change the ABI, but the problem is |
1973 | obviously the case with any flags that change the ABI, but the problem is |
1954 | most noticeably disabled with ev_stat and large file support. |
1974 | most noticeably displayed with ev_stat and large file support. |
1955 | |
1975 | |
1956 | The solution for this is to lobby your distribution maker to make large |
1976 | The solution for this is to lobby your distribution maker to make large |
1957 | file interfaces available by default (as e.g. FreeBSD does) and not |
1977 | file interfaces available by default (as e.g. FreeBSD does) and not |
1958 | optional. Libev cannot simply switch on large file support because it has |
1978 | optional. Libev cannot simply switch on large file support because it has |
1959 | to exchange stat structures with application programs compiled using the |
1979 | to exchange stat structures with application programs compiled using the |
1960 | default compilation environment. |
1980 | default compilation environment. |
1961 | |
1981 | |
1962 | =head3 Inotify and Kqueue |
1982 | =head3 Inotify and Kqueue |
1963 | |
1983 | |
1964 | When C<inotify (7)> support has been compiled into libev (generally |
1984 | When C<inotify (7)> support has been compiled into libev and present at |
1965 | only available with Linux 2.6.25 or above due to bugs in earlier |
1985 | runtime, it will be used to speed up change detection where possible. The |
1966 | implementations) and present at runtime, it will be used to speed up |
1986 | inotify descriptor will be created lazily when the first C<ev_stat> |
1967 | change detection where possible. The inotify descriptor will be created |
1987 | watcher is being started. |
1968 | lazily when the first C<ev_stat> watcher is being started. |
|
|
1969 | |
1988 | |
1970 | Inotify presence does not change the semantics of C<ev_stat> watchers |
1989 | Inotify presence does not change the semantics of C<ev_stat> watchers |
1971 | except that changes might be detected earlier, and in some cases, to avoid |
1990 | except that changes might be detected earlier, and in some cases, to avoid |
1972 | making regular C<stat> calls. Even in the presence of inotify support |
1991 | making regular C<stat> calls. Even in the presence of inotify support |
1973 | there are many cases where libev has to resort to regular C<stat> polling, |
1992 | there are many cases where libev has to resort to regular C<stat> polling, |
1974 | but as long as the path exists, libev usually gets away without polling. |
1993 | but as long as kernel 2.6.25 or newer is used (2.6.24 and older have too |
|
|
1994 | many bugs), the path exists (i.e. stat succeeds), and the path resides on |
|
|
1995 | a local filesystem (libev currently assumes only ext2/3, jfs, reiserfs and |
|
|
1996 | xfs are fully working) libev usually gets away without polling. |
1975 | |
1997 | |
1976 | There is no support for kqueue, as apparently it cannot be used to |
1998 | There is no support for kqueue, as apparently it cannot be used to |
1977 | implement this functionality, due to the requirement of having a file |
1999 | implement this functionality, due to the requirement of having a file |
1978 | descriptor open on the object at all times, and detecting renames, unlinks |
2000 | descriptor open on the object at all times, and detecting renames, unlinks |
1979 | etc. is difficult. |
2001 | etc. is difficult. |
1980 | |
2002 | |
1981 | =head3 The special problem of stat time resolution |
2003 | =head3 The special problem of stat time resolution |
1982 | |
2004 | |
1983 | The C<stat ()> system call only supports full-second resolution portably, and |
2005 | The C<stat ()> system call only supports full-second resolution portably, |
1984 | even on systems where the resolution is higher, most file systems still |
2006 | and even on systems where the resolution is higher, most file systems |
1985 | only support whole seconds. |
2007 | still only support whole seconds. |
1986 | |
2008 | |
1987 | That means that, if the time is the only thing that changes, you can |
2009 | That means that, if the time is the only thing that changes, you can |
1988 | easily miss updates: on the first update, C<ev_stat> detects a change and |
2010 | easily miss updates: on the first update, C<ev_stat> detects a change and |
1989 | calls your callback, which does something. When there is another update |
2011 | calls your callback, which does something. When there is another update |
1990 | within the same second, C<ev_stat> will be unable to detect unless the |
2012 | within the same second, C<ev_stat> will be unable to detect unless the |
… | |
… | |
2629 | =over 4 |
2651 | =over 4 |
2630 | |
2652 | |
2631 | =item ev_async_init (ev_async *, callback) |
2653 | =item ev_async_init (ev_async *, callback) |
2632 | |
2654 | |
2633 | Initialises and configures the async watcher - it has no parameters of any |
2655 | Initialises and configures the async watcher - it has no parameters of any |
2634 | kind. There is a C<ev_asynd_set> macro, but using it is utterly pointless, |
2656 | kind. There is a C<ev_async_set> macro, but using it is utterly pointless, |
2635 | trust me. |
2657 | trust me. |
2636 | |
2658 | |
2637 | =item ev_async_send (loop, ev_async *) |
2659 | =item ev_async_send (loop, ev_async *) |
2638 | |
2660 | |
2639 | Sends/signals/activates the given C<ev_async> watcher, that is, feeds |
2661 | Sends/signals/activates the given C<ev_async> watcher, that is, feeds |
… | |
… | |
3072 | |
3094 | |
3073 | #define EV_STANDALONE 1 |
3095 | #define EV_STANDALONE 1 |
3074 | #include "ev.h" |
3096 | #include "ev.h" |
3075 | |
3097 | |
3076 | Both header files and implementation files can be compiled with a C++ |
3098 | Both header files and implementation files can be compiled with a C++ |
3077 | compiler (at least, thats a stated goal, and breakage will be treated |
3099 | compiler (at least, that's a stated goal, and breakage will be treated |
3078 | as a bug). |
3100 | as a bug). |
3079 | |
3101 | |
3080 | You need the following files in your source tree, or in a directory |
3102 | You need the following files in your source tree, or in a directory |
3081 | in your include path (e.g. in libev/ when using -Ilibev): |
3103 | in your include path (e.g. in libev/ when using -Ilibev): |
3082 | |
3104 | |
… | |
… | |
3554 | loop, as long as you don't confuse yourself). The only exception is that |
3576 | loop, as long as you don't confuse yourself). The only exception is that |
3555 | you must not do this from C<ev_periodic> reschedule callbacks. |
3577 | you must not do this from C<ev_periodic> reschedule callbacks. |
3556 | |
3578 | |
3557 | Care has been taken to ensure that libev does not keep local state inside |
3579 | Care has been taken to ensure that libev does not keep local state inside |
3558 | C<ev_loop>, and other calls do not usually allow for coroutine switches as |
3580 | C<ev_loop>, and other calls do not usually allow for coroutine switches as |
3559 | they do not clal any callbacks. |
3581 | they do not call any callbacks. |
3560 | |
3582 | |
3561 | =head2 COMPILER WARNINGS |
3583 | =head2 COMPILER WARNINGS |
3562 | |
3584 | |
3563 | Depending on your compiler and compiler settings, you might get no or a |
3585 | Depending on your compiler and compiler settings, you might get no or a |
3564 | lot of warnings when compiling libev code. Some people are apparently |
3586 | lot of warnings when compiling libev code. Some people are apparently |
… | |
… | |
3598 | ==2274== definitely lost: 0 bytes in 0 blocks. |
3620 | ==2274== definitely lost: 0 bytes in 0 blocks. |
3599 | ==2274== possibly lost: 0 bytes in 0 blocks. |
3621 | ==2274== possibly lost: 0 bytes in 0 blocks. |
3600 | ==2274== still reachable: 256 bytes in 1 blocks. |
3622 | ==2274== still reachable: 256 bytes in 1 blocks. |
3601 | |
3623 | |
3602 | Then there is no memory leak, just as memory accounted to global variables |
3624 | Then there is no memory leak, just as memory accounted to global variables |
3603 | is not a memleak - the memory is still being refernced, and didn't leak. |
3625 | is not a memleak - the memory is still being referenced, and didn't leak. |
3604 | |
3626 | |
3605 | Similarly, under some circumstances, valgrind might report kernel bugs |
3627 | Similarly, under some circumstances, valgrind might report kernel bugs |
3606 | as if it were a bug in libev (e.g. in realloc or in the poll backend, |
3628 | as if it were a bug in libev (e.g. in realloc or in the poll backend, |
3607 | although an acceptable workaround has been found here), or it might be |
3629 | although an acceptable workaround has been found here), or it might be |
3608 | confused. |
3630 | confused. |
… | |
… | |
3846 | =back |
3868 | =back |
3847 | |
3869 | |
3848 | |
3870 | |
3849 | =head1 AUTHOR |
3871 | =head1 AUTHOR |
3850 | |
3872 | |
3851 | Marc Lehmann <libev@schmorp.de>. |
3873 | Marc Lehmann <libev@schmorp.de>, with repeated corrections by Mikael Magnusson. |
3852 | |
3874 | |