X-Git-Url: https://git.llucax.com/software/libev.git/blobdiff_plain/b8c10cda08d48348f422bb3d31f019584cf252c9..678c8c54d0ce967d8f1db3e575aed61879279975:/ev_kqueue.c diff --git a/ev_kqueue.c b/ev_kqueue.c index 4bbcf39..8282fea 100644 --- a/ev_kqueue.c +++ b/ev_kqueue.c @@ -34,20 +34,14 @@ #include #include -static int kq_fd; -static struct kevent *kq_changes; -static int kq_changemax, kq_changecnt; -static struct kevent *kq_events; -static int kq_eventmax; - static void -kqueue_change (int fd, int filter, int flags, int fflags) +kqueue_change (EV_P_ int fd, int filter, int flags, int fflags) { struct kevent *ke; - array_needsize (kq_changes, kq_changemax, ++kq_changecnt, ); + array_needsize (kqueue_changes, kqueue_changemax, ++kqueue_changecnt, ); - ke = &kq_changes [kq_changecnt - 1]; + ke = &kqueue_changes [kqueue_changecnt - 1]; memset (ke, 0, sizeof (struct kevent)); ke->ident = fd; ke->filter = filter; @@ -60,7 +54,7 @@ kqueue_change (int fd, int filter, int flags, int fflags) #endif static void -kqueue_modify (int fd, int oev, int nev) +kqueue_modify (EV_P_ int fd, int oev, int nev) { if ((oev ^ nev) & EV_READ) { @@ -80,22 +74,22 @@ kqueue_modify (int fd, int oev, int nev) } static void -kqueue_poll (ev_tstamp timeout) +kqueue_poll (EV_P_ ev_tstamp timeout) { int res, i; struct timespec ts; ts.tv_sec = (time_t)timeout; ts.tv_nsec = (long)(timeout - (ev_tstamp)ts.tv_sec) * 1e9; - res = kevent (kq_fd, kq_changes, kq_changecnt, kq_events, kq_eventmax, &ts); - kq_changecnt = 0; + res = kevent (kqueue_fd, kqueue_changes, kqueue_changecnt, kqueue_events, kqueue_eventmax, &ts); + kqueue_changecnt = 0; if (res < 0) return; for (i = 0; i < res; ++i) { - if (kq_events [i].flags & EV_ERROR) + if (kqueue_events [i].flags & EV_ERROR) { /* * Error messages that can happen, when a delete fails. @@ -109,34 +103,35 @@ kqueue_poll (ev_tstamp timeout) * an event we are still processing. In that case * the data field is set to ENOENT. */ - if (kq_events [i].data == EBADF) - fd_kill (kq_events [i].ident); + if (kqueue_events [i].data == EBADF) + fd_kill (EV_A_ kqueue_events [i].ident); } else fd_event ( - kq_events [i].ident, - kq_events [i].filter == EVFILT_READ ? EV_READ - : kq_events [i].filter == EVFILT_WRITE ? EV_WRITE + EV_A_ + kqueue_events [i].ident, + kqueue_events [i].filter == EVFILT_READ ? EV_READ + : kqueue_events [i].filter == EVFILT_WRITE ? EV_WRITE : 0 ); } - if (expect_false (res == kq_eventmax)) + if (expect_false (res == kqueue_eventmax)) { - free (kq_events); - kq_eventmax = array_roundsize (kq_events, kq_eventmax << 1); - kq_events = malloc (sizeof (struct kevent) * kq_eventmax); + free (kqueue_events); + kqueue_eventmax = array_roundsize (kqueue_events, kqueue_eventmax << 1); + kqueue_events = malloc (sizeof (struct kevent) * kqueue_eventmax); } } -static void -kqueue_init (int flags) +static int +kqueue_init (EV_P_ int flags) { struct kevent ch, ev; /* Initalize the kernel queue */ - if ((kq_fd = kqueue ()) < 0) - return; + if ((kqueue_fd = kqueue ()) < 0) + return 0; /* Check for Mac OS X kqueue bug. */ ch.ident = -1; @@ -148,21 +143,22 @@ kqueue_init (int flags) * stick an error in ev. If kqueue is broken, then * kevent will fail. */ - if (kevent (kq_fd, &ch, 1, &ev, 1, 0) != 1 + if (kevent (kqueue_fd, &ch, 1, &ev, 1, 0) != 1 || ev.ident != -1 || ev.flags != EV_ERROR) { /* detected broken kqueue */ - close (kq_fd); - return; + close (kqueue_fd); + return 0; } - ev_method = EVMETHOD_KQUEUE; method_fudge = 1e-3; /* needed to compensate for kevent returning early */ method_modify = kqueue_modify; method_poll = kqueue_poll; - kq_eventmax = 64; /* intiial number of events receivable per poll */ - kq_events = malloc (sizeof (struct kevent) * kq_eventmax); + kqueue_eventmax = 64; /* intiial number of events receivable per poll */ + kqueue_events = malloc (sizeof (struct kevent) * kqueue_eventmax); + + return EVMETHOD_KQUEUE; }