2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
4 * Copyright (c) 2011 NetApp, Inc.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
16 * THIS SOFTWARE IS PROVIDED BY NETAPP, INC ``AS IS'' AND
17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19 * ARE DISCLAIMED. IN NO EVENT SHALL NETAPP, INC OR CONTRIBUTORS BE LIABLE
20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * Micro event library for FreeBSD, designed for a single i/o thread
33 * using kqueue, and having events be persistent by default.
36 #include <sys/cdefs.h>
37 __FBSDID("$FreeBSD$");
40 #ifndef WITHOUT_CAPSICUM
41 #include <capsicum_helpers.h>
52 #include <sys/types.h>
53 #ifndef WITHOUT_CAPSICUM
54 #include <sys/capsicum.h>
56 #include <sys/event.h>
60 #include <pthread_np.h>
66 static pthread_t mevent_tid;
67 static int mevent_timid = 43;
68 static int mevent_pipefd[2];
69 static pthread_mutex_t mevent_lmutex = PTHREAD_MUTEX_INITIALIZER;
72 void (*me_func)(int, enum ev_type, void *);
73 #define me_msecs me_fd
79 int me_state; /* Desired kevent flags. */
81 LIST_ENTRY(mevent) me_list;
84 static LIST_HEAD(listhead, mevent) global_head, change_head;
89 pthread_mutex_lock(&mevent_lmutex);
95 pthread_mutex_unlock(&mevent_lmutex);
99 mevent_pipe_read(int fd, enum ev_type type, void *param)
101 char buf[MEVENT_MAX];
105 * Drain the pipe read side. The fd is non-blocking so this is
109 status = read(fd, buf, sizeof(buf));
110 } while (status == MEVENT_MAX);
119 * If calling from outside the i/o thread, write a byte on the
120 * pipe to force the i/o thread to exit the blocking kevent call.
122 if (mevent_pipefd[1] != 0 && pthread_self() != mevent_tid) {
123 write(mevent_pipefd[1], &c, 1);
128 mevent_kq_filter(struct mevent *mevp)
134 if (mevp->me_type == EVF_READ)
135 retval = EVFILT_READ;
137 if (mevp->me_type == EVF_WRITE)
138 retval = EVFILT_WRITE;
140 if (mevp->me_type == EVF_TIMER)
141 retval = EVFILT_TIMER;
143 if (mevp->me_type == EVF_SIGNAL)
144 retval = EVFILT_SIGNAL;
150 mevent_kq_flags(struct mevent *mevp)
152 return (mevp->me_state);
156 mevent_kq_fflags(struct mevent *mevp)
158 /* XXX nothing yet, perhaps EV_EOF for reads ? */
163 mevent_build(int mfd, struct kevent *kev)
165 struct mevent *mevp, *tmpp;
172 LIST_FOREACH_SAFE(mevp, &change_head, me_list, tmpp) {
173 if (mevp->me_closefd) {
175 * A close of the file descriptor will remove the
180 if (mevp->me_type == EVF_TIMER) {
181 kev[i].ident = mevp->me_timid;
182 kev[i].data = mevp->me_msecs;
184 kev[i].ident = mevp->me_fd;
187 kev[i].filter = mevent_kq_filter(mevp);
188 kev[i].flags = mevent_kq_flags(mevp);
189 kev[i].fflags = mevent_kq_fflags(mevp);
195 LIST_REMOVE(mevp, me_list);
197 if (mevp->me_state & EV_DELETE) {
201 * We need to add the event only once, so we can
202 * reset the EV_ADD bit after it has been propagated
203 * to the kevent() arguments the first time.
205 mevp->me_state &= ~EV_ADD;
206 LIST_INSERT_HEAD(&global_head, mevp, me_list);
209 assert(i < MEVENT_MAX);
218 mevent_handle(struct kevent *kev, int numev)
223 for (i = 0; i < numev; i++) {
226 /* XXX check for EV_ERROR ? */
228 (*mevp->me_func)(mevp->me_fd, mevp->me_type, mevp->me_param);
232 static struct mevent *
233 mevent_add_state(int tfd, enum ev_type type,
234 void (*func)(int, enum ev_type, void *), void *param,
237 struct mevent *lp, *mevp;
239 if (tfd < 0 || func == NULL) {
248 * Verify that the fd/type tuple is not present in any list
250 LIST_FOREACH(lp, &global_head, me_list) {
251 if (type != EVF_TIMER && lp->me_fd == tfd &&
252 lp->me_type == type) {
257 LIST_FOREACH(lp, &change_head, me_list) {
258 if (type != EVF_TIMER && lp->me_fd == tfd &&
259 lp->me_type == type) {
265 * Allocate an entry, populate it, and add it to the change list.
267 mevp = calloc(1, sizeof(struct mevent));
272 if (type == EVF_TIMER) {
273 mevp->me_msecs = tfd;
274 mevp->me_timid = mevent_timid++;
277 mevp->me_type = type;
278 mevp->me_func = func;
279 mevp->me_param = param;
281 LIST_INSERT_HEAD(&change_head, mevp, me_list);
283 mevp->me_state = state;
293 mevent_add(int tfd, enum ev_type type,
294 void (*func)(int, enum ev_type, void *), void *param)
297 return (mevent_add_state(tfd, type, func, param, EV_ADD));
301 mevent_add_disabled(int tfd, enum ev_type type,
302 void (*func)(int, enum ev_type, void *), void *param)
305 return (mevent_add_state(tfd, type, func, param, EV_ADD | EV_DISABLE));
309 mevent_update(struct mevent *evp, bool enable)
316 * It's not possible to enable/disable a deleted event
318 assert((evp->me_state & EV_DELETE) == 0);
320 newstate = evp->me_state;
322 newstate |= EV_ENABLE;
323 newstate &= ~EV_DISABLE;
325 newstate |= EV_DISABLE;
326 newstate &= ~EV_ENABLE;
330 * No update needed if state isn't changing
332 if (evp->me_state != newstate) {
333 evp->me_state = newstate;
336 * Place the entry onto the changed list if not
339 if (evp->me_cq == 0) {
341 LIST_REMOVE(evp, me_list);
342 LIST_INSERT_HEAD(&change_head, evp, me_list);
353 mevent_enable(struct mevent *evp)
356 return (mevent_update(evp, true));
360 mevent_disable(struct mevent *evp)
363 return (mevent_update(evp, false));
367 mevent_delete_event(struct mevent *evp, int closefd)
372 * Place the entry onto the changed list if not already there, and
373 * mark as to be deleted.
375 if (evp->me_cq == 0) {
377 LIST_REMOVE(evp, me_list);
378 LIST_INSERT_HEAD(&change_head, evp, me_list);
381 evp->me_state = EV_DELETE;
392 mevent_delete(struct mevent *evp)
395 return (mevent_delete_event(evp, 0));
399 mevent_delete_close(struct mevent *evp)
402 return (mevent_delete_event(evp, 1));
406 mevent_set_name(void)
409 pthread_set_name_np(mevent_tid, "mevent");
413 mevent_dispatch(void)
415 struct kevent changelist[MEVENT_MAX];
416 struct kevent eventlist[MEVENT_MAX];
417 struct mevent *pipev;
421 #ifndef WITHOUT_CAPSICUM
425 mevent_tid = pthread_self();
431 #ifndef WITHOUT_CAPSICUM
432 cap_rights_init(&rights, CAP_KQUEUE);
433 if (caph_rights_limit(mfd, &rights) == -1)
434 errx(EX_OSERR, "Unable to apply rights for sandbox");
438 * Open the pipe that will be used for other threads to force
439 * the blocking kqueue call to exit by writing to it. Set the
440 * descriptor to non-blocking.
442 ret = pipe(mevent_pipefd);
448 #ifndef WITHOUT_CAPSICUM
449 cap_rights_init(&rights, CAP_EVENT, CAP_READ, CAP_WRITE);
450 if (caph_rights_limit(mevent_pipefd[0], &rights) == -1)
451 errx(EX_OSERR, "Unable to apply rights for sandbox");
452 if (caph_rights_limit(mevent_pipefd[1], &rights) == -1)
453 errx(EX_OSERR, "Unable to apply rights for sandbox");
457 * Add internal event handler for the pipe write fd
459 pipev = mevent_add(mevent_pipefd[0], EVF_READ, mevent_pipe_read, NULL);
460 assert(pipev != NULL);
464 * Build changelist if required.
465 * XXX the changelist can be put into the blocking call
466 * to eliminate the extra syscall. Currently better for
469 numev = mevent_build(mfd, changelist);
471 ret = kevent(mfd, changelist, numev, NULL, 0, NULL);
473 perror("Error return from kevent change");
478 * Block awaiting events
480 ret = kevent(mfd, NULL, 0, eventlist, MEVENT_MAX, NULL);
481 if (ret == -1 && errno != EINTR) {
482 perror("Error return from kevent monitor");
486 * Handle reported events
488 mevent_handle(eventlist, ret);