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 extern const char *vmname;
68 static pthread_t mevent_tid;
69 static int mevent_timid = 43;
70 static int mevent_pipefd[2];
71 static pthread_mutex_t mevent_lmutex = PTHREAD_MUTEX_INITIALIZER;
74 void (*me_func)(int, enum ev_type, void *);
75 #define me_msecs me_fd
81 int me_state; /* Desired kevent flags. */
83 LIST_ENTRY(mevent) me_list;
86 static LIST_HEAD(listhead, mevent) global_head, change_head;
91 pthread_mutex_lock(&mevent_lmutex);
97 pthread_mutex_unlock(&mevent_lmutex);
101 mevent_pipe_read(int fd, enum ev_type type, void *param)
103 char buf[MEVENT_MAX];
107 * Drain the pipe read side. The fd is non-blocking so this is
111 status = read(fd, buf, sizeof(buf));
112 } while (status == MEVENT_MAX);
121 * If calling from outside the i/o thread, write a byte on the
122 * pipe to force the i/o thread to exit the blocking kevent call.
124 if (mevent_pipefd[1] != 0 && pthread_self() != mevent_tid) {
125 write(mevent_pipefd[1], &c, 1);
130 mevent_kq_filter(struct mevent *mevp)
136 if (mevp->me_type == EVF_READ)
137 retval = EVFILT_READ;
139 if (mevp->me_type == EVF_WRITE)
140 retval = EVFILT_WRITE;
142 if (mevp->me_type == EVF_TIMER)
143 retval = EVFILT_TIMER;
145 if (mevp->me_type == EVF_SIGNAL)
146 retval = EVFILT_SIGNAL;
152 mevent_kq_flags(struct mevent *mevp)
154 return (mevp->me_state);
158 mevent_kq_fflags(struct mevent *mevp)
160 /* XXX nothing yet, perhaps EV_EOF for reads ? */
165 mevent_build(int mfd, struct kevent *kev)
167 struct mevent *mevp, *tmpp;
174 LIST_FOREACH_SAFE(mevp, &change_head, me_list, tmpp) {
175 if (mevp->me_closefd) {
177 * A close of the file descriptor will remove the
182 if (mevp->me_type == EVF_TIMER) {
183 kev[i].ident = mevp->me_timid;
184 kev[i].data = mevp->me_msecs;
186 kev[i].ident = mevp->me_fd;
189 kev[i].filter = mevent_kq_filter(mevp);
190 kev[i].flags = mevent_kq_flags(mevp);
191 kev[i].fflags = mevent_kq_fflags(mevp);
197 LIST_REMOVE(mevp, me_list);
199 if (mevp->me_state & EV_DELETE) {
203 * We need to add the event only once, so we can
204 * reset the EV_ADD bit after it has been propagated
205 * to the kevent() arguments the first time.
207 mevp->me_state &= ~EV_ADD;
208 LIST_INSERT_HEAD(&global_head, mevp, me_list);
211 assert(i < MEVENT_MAX);
220 mevent_handle(struct kevent *kev, int numev)
225 for (i = 0; i < numev; i++) {
228 /* XXX check for EV_ERROR ? */
230 (*mevp->me_func)(mevp->me_fd, mevp->me_type, mevp->me_param);
234 static struct mevent *
235 mevent_add_state(int tfd, enum ev_type type,
236 void (*func)(int, enum ev_type, void *), void *param,
239 struct mevent *lp, *mevp;
241 if (tfd < 0 || func == NULL) {
250 * Verify that the fd/type tuple is not present in any list
252 LIST_FOREACH(lp, &global_head, me_list) {
253 if (type != EVF_TIMER && lp->me_fd == tfd &&
254 lp->me_type == type) {
259 LIST_FOREACH(lp, &change_head, me_list) {
260 if (type != EVF_TIMER && lp->me_fd == tfd &&
261 lp->me_type == type) {
267 * Allocate an entry, populate it, and add it to the change list.
269 mevp = calloc(1, sizeof(struct mevent));
274 if (type == EVF_TIMER) {
275 mevp->me_msecs = tfd;
276 mevp->me_timid = mevent_timid++;
279 mevp->me_type = type;
280 mevp->me_func = func;
281 mevp->me_param = param;
283 LIST_INSERT_HEAD(&change_head, mevp, me_list);
285 mevp->me_state = state;
295 mevent_add(int tfd, enum ev_type type,
296 void (*func)(int, enum ev_type, void *), void *param)
299 return (mevent_add_state(tfd, type, func, param, EV_ADD));
303 mevent_add_disabled(int tfd, enum ev_type type,
304 void (*func)(int, enum ev_type, void *), void *param)
307 return (mevent_add_state(tfd, type, func, param, EV_ADD | EV_DISABLE));
311 mevent_update(struct mevent *evp, bool enable)
318 * It's not possible to enable/disable a deleted event
320 assert((evp->me_state & EV_DELETE) == 0);
322 newstate = evp->me_state;
324 newstate |= EV_ENABLE;
325 newstate &= ~EV_DISABLE;
327 newstate |= EV_DISABLE;
328 newstate &= ~EV_ENABLE;
332 * No update needed if state isn't changing
334 if (evp->me_state != newstate) {
335 evp->me_state = newstate;
338 * Place the entry onto the changed list if not
341 if (evp->me_cq == 0) {
343 LIST_REMOVE(evp, me_list);
344 LIST_INSERT_HEAD(&change_head, evp, me_list);
355 mevent_enable(struct mevent *evp)
358 return (mevent_update(evp, true));
362 mevent_disable(struct mevent *evp)
365 return (mevent_update(evp, false));
369 mevent_delete_event(struct mevent *evp, int closefd)
374 * Place the entry onto the changed list if not already there, and
375 * mark as to be deleted.
377 if (evp->me_cq == 0) {
379 LIST_REMOVE(evp, me_list);
380 LIST_INSERT_HEAD(&change_head, evp, me_list);
383 evp->me_state = EV_DELETE;
394 mevent_delete(struct mevent *evp)
397 return (mevent_delete_event(evp, 0));
401 mevent_delete_close(struct mevent *evp)
404 return (mevent_delete_event(evp, 1));
408 mevent_set_name(void)
411 pthread_set_name_np(mevent_tid, "mevent");
415 mevent_dispatch(void)
417 struct kevent changelist[MEVENT_MAX];
418 struct kevent eventlist[MEVENT_MAX];
419 struct mevent *pipev;
423 #ifndef WITHOUT_CAPSICUM
427 mevent_tid = pthread_self();
433 #ifndef WITHOUT_CAPSICUM
434 cap_rights_init(&rights, CAP_KQUEUE);
435 if (caph_rights_limit(mfd, &rights) == -1)
436 errx(EX_OSERR, "Unable to apply rights for sandbox");
440 * Open the pipe that will be used for other threads to force
441 * the blocking kqueue call to exit by writing to it. Set the
442 * descriptor to non-blocking.
444 ret = pipe(mevent_pipefd);
450 #ifndef WITHOUT_CAPSICUM
451 cap_rights_init(&rights, CAP_EVENT, CAP_READ, CAP_WRITE);
452 if (caph_rights_limit(mevent_pipefd[0], &rights) == -1)
453 errx(EX_OSERR, "Unable to apply rights for sandbox");
454 if (caph_rights_limit(mevent_pipefd[1], &rights) == -1)
455 errx(EX_OSERR, "Unable to apply rights for sandbox");
459 * Add internal event handler for the pipe write fd
461 pipev = mevent_add(mevent_pipefd[0], EVF_READ, mevent_pipe_read, NULL);
462 assert(pipev != NULL);
466 * Build changelist if required.
467 * XXX the changelist can be put into the blocking call
468 * to eliminate the extra syscall. Currently better for
471 numev = mevent_build(mfd, changelist);
473 ret = kevent(mfd, changelist, numev, NULL, 0, NULL);
475 perror("Error return from kevent change");
480 * Block awaiting events
482 ret = kevent(mfd, NULL, 0, eventlist, MEVENT_MAX, NULL);
483 if (ret == -1 && errno != EINTR) {
484 perror("Error return from kevent monitor");
488 * Handle reported events
490 mevent_handle(eventlist, ret);