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>
51 #include <sys/types.h>
52 #ifndef WITHOUT_CAPSICUM
53 #include <sys/capsicum.h>
55 #include <sys/event.h>
59 #include <pthread_np.h>
68 #define MEV_DEL_PENDING 4
72 static pthread_t mevent_tid;
73 static int mevent_timid = 43;
74 static int mevent_pipefd[2];
75 static pthread_mutex_t mevent_lmutex = PTHREAD_MUTEX_INITIALIZER;
78 void (*me_func)(int, enum ev_type, void *);
79 #define me_msecs me_fd
87 LIST_ENTRY(mevent) me_list;
90 static LIST_HEAD(listhead, mevent) global_head, change_head;
95 pthread_mutex_lock(&mevent_lmutex);
101 pthread_mutex_unlock(&mevent_lmutex);
105 mevent_pipe_read(int fd, enum ev_type type, void *param)
107 char buf[MEVENT_MAX];
111 * Drain the pipe read side. The fd is non-blocking so this is
115 status = read(fd, buf, sizeof(buf));
116 } while (status == MEVENT_MAX);
125 * If calling from outside the i/o thread, write a byte on the
126 * pipe to force the i/o thread to exit the blocking kevent call.
128 if (mevent_pipefd[1] != 0 && pthread_self() != mevent_tid) {
129 write(mevent_pipefd[1], &c, 1);
134 mevent_kq_filter(struct mevent *mevp)
140 if (mevp->me_type == EVF_READ)
141 retval = EVFILT_READ;
143 if (mevp->me_type == EVF_WRITE)
144 retval = EVFILT_WRITE;
146 if (mevp->me_type == EVF_TIMER)
147 retval = EVFILT_TIMER;
149 if (mevp->me_type == EVF_SIGNAL)
150 retval = EVFILT_SIGNAL;
156 mevent_kq_flags(struct mevent *mevp)
160 switch (mevp->me_state) {
162 ret = EV_ADD; /* implicitly enabled */
170 case MEV_DEL_PENDING:
182 mevent_kq_fflags(struct mevent *mevp)
184 /* XXX nothing yet, perhaps EV_EOF for reads ? */
189 mevent_build(int mfd, struct kevent *kev)
191 struct mevent *mevp, *tmpp;
198 LIST_FOREACH_SAFE(mevp, &change_head, me_list, tmpp) {
199 if (mevp->me_closefd) {
201 * A close of the file descriptor will remove the
206 if (mevp->me_type == EVF_TIMER) {
207 kev[i].ident = mevp->me_timid;
208 kev[i].data = mevp->me_msecs;
210 kev[i].ident = mevp->me_fd;
213 kev[i].filter = mevent_kq_filter(mevp);
214 kev[i].flags = mevent_kq_flags(mevp);
215 kev[i].fflags = mevent_kq_fflags(mevp);
221 LIST_REMOVE(mevp, me_list);
223 if (mevp->me_state == MEV_DEL_PENDING) {
226 LIST_INSERT_HEAD(&global_head, mevp, me_list);
229 assert(i < MEVENT_MAX);
238 mevent_handle(struct kevent *kev, int numev)
243 for (i = 0; i < numev; i++) {
246 /* XXX check for EV_ERROR ? */
248 (*mevp->me_func)(mevp->me_fd, mevp->me_type, mevp->me_param);
253 mevent_add(int tfd, enum ev_type type,
254 void (*func)(int, enum ev_type, void *), void *param)
256 struct mevent *lp, *mevp;
258 if (tfd < 0 || func == NULL) {
267 * Verify that the fd/type tuple is not present in any list
269 LIST_FOREACH(lp, &global_head, me_list) {
270 if (type != EVF_TIMER && lp->me_fd == tfd &&
271 lp->me_type == type) {
276 LIST_FOREACH(lp, &change_head, me_list) {
277 if (type != EVF_TIMER && lp->me_fd == tfd &&
278 lp->me_type == type) {
284 * Allocate an entry, populate it, and add it to the change list.
286 mevp = calloc(1, sizeof(struct mevent));
291 if (type == EVF_TIMER) {
292 mevp->me_msecs = tfd;
293 mevp->me_timid = mevent_timid++;
296 mevp->me_type = type;
297 mevp->me_func = func;
298 mevp->me_param = param;
300 LIST_INSERT_HEAD(&change_head, mevp, me_list);
302 mevp->me_state = MEV_ADD;
312 mevent_update(struct mevent *evp, int newstate)
315 * It's not possible to enable/disable a deleted event
317 if (evp->me_state == MEV_DEL_PENDING)
321 * No update needed if state isn't changing
323 if (evp->me_state == newstate)
328 evp->me_state = newstate;
331 * Place the entry onto the changed list if not already there.
333 if (evp->me_cq == 0) {
335 LIST_REMOVE(evp, me_list);
336 LIST_INSERT_HEAD(&change_head, evp, me_list);
346 mevent_enable(struct mevent *evp)
349 return (mevent_update(evp, MEV_ENABLE));
353 mevent_disable(struct mevent *evp)
356 return (mevent_update(evp, MEV_DISABLE));
360 mevent_delete_event(struct mevent *evp, int closefd)
365 * Place the entry onto the changed list if not already there, and
366 * mark as to be deleted.
368 if (evp->me_cq == 0) {
370 LIST_REMOVE(evp, me_list);
371 LIST_INSERT_HEAD(&change_head, evp, me_list);
374 evp->me_state = MEV_DEL_PENDING;
385 mevent_delete(struct mevent *evp)
388 return (mevent_delete_event(evp, 0));
392 mevent_delete_close(struct mevent *evp)
395 return (mevent_delete_event(evp, 1));
399 mevent_set_name(void)
402 pthread_set_name_np(mevent_tid, "mevent");
406 mevent_dispatch(void)
408 struct kevent changelist[MEVENT_MAX];
409 struct kevent eventlist[MEVENT_MAX];
410 struct mevent *pipev;
414 #ifndef WITHOUT_CAPSICUM
418 mevent_tid = pthread_self();
424 #ifndef WITHOUT_CAPSICUM
425 cap_rights_init(&rights, CAP_KQUEUE);
426 if (caph_rights_limit(mfd, &rights) == -1)
427 errx(EX_OSERR, "Unable to apply rights for sandbox");
431 * Open the pipe that will be used for other threads to force
432 * the blocking kqueue call to exit by writing to it. Set the
433 * descriptor to non-blocking.
435 ret = pipe(mevent_pipefd);
441 #ifndef WITHOUT_CAPSICUM
442 cap_rights_init(&rights, CAP_EVENT, CAP_READ, CAP_WRITE);
443 if (caph_rights_limit(mevent_pipefd[0], &rights) == -1)
444 errx(EX_OSERR, "Unable to apply rights for sandbox");
445 if (caph_rights_limit(mevent_pipefd[1], &rights) == -1)
446 errx(EX_OSERR, "Unable to apply rights for sandbox");
450 * Add internal event handler for the pipe write fd
452 pipev = mevent_add(mevent_pipefd[0], EVF_READ, mevent_pipe_read, NULL);
453 assert(pipev != NULL);
457 * Build changelist if required.
458 * XXX the changelist can be put into the blocking call
459 * to eliminate the extra syscall. Currently better for
462 numev = mevent_build(mfd, changelist);
464 ret = kevent(mfd, changelist, numev, NULL, 0, NULL);
466 perror("Error return from kevent change");
471 * Block awaiting events
473 ret = kevent(mfd, NULL, 0, eventlist, MEVENT_MAX, NULL);
474 if (ret == -1 && errno != EINTR) {
475 perror("Error return from kevent monitor");
479 * Handle reported events
481 mevent_handle(eventlist, ret);