2 * Copyright (c) 1999,2000,2001 Jonathan Lemon <jlemon@FreeBSD.org>
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26 * $FreeBSD: src/sys/kern/kern_event.c,v 1.2.2.10 2004/04/04 07:03:14 cperciva Exp $
27 * $DragonFly: src/sys/kern/kern_event.c,v 1.33 2007/02/03 17:05:57 corecode Exp $
30 #include <sys/param.h>
31 #include <sys/systm.h>
32 #include <sys/kernel.h>
34 #include <sys/malloc.h>
35 #include <sys/unistd.h>
38 #include <sys/fcntl.h>
39 #include <sys/select.h>
40 #include <sys/queue.h>
41 #include <sys/event.h>
42 #include <sys/eventvar.h>
44 #include <sys/protosw.h>
45 #include <sys/socket.h>
46 #include <sys/socketvar.h>
48 #include <sys/sysctl.h>
49 #include <sys/sysproto.h>
51 #include <sys/signalvar.h>
52 #include <sys/filio.h>
54 #include <sys/thread2.h>
55 #include <sys/file2.h>
56 #include <sys/mplock2.h>
58 #include <vm/vm_zone.h>
60 MALLOC_DEFINE(M_KQUEUE, "kqueue", "memory for kqueue system");
62 static int kqueue_scan(struct kqueue *kq, struct kevent *kevp, int count,
63 struct timespec *tsp, int *errorp);
64 static int kqueue_read(struct file *fp, struct uio *uio,
65 struct ucred *cred, int flags);
66 static int kqueue_write(struct file *fp, struct uio *uio,
67 struct ucred *cred, int flags);
68 static int kqueue_ioctl(struct file *fp, u_long com, caddr_t data,
69 struct ucred *cred, struct sysmsg *msg);
70 static int kqueue_poll(struct file *fp, int events, struct ucred *cred);
71 static int kqueue_kqfilter(struct file *fp, struct knote *kn);
72 static int kqueue_stat(struct file *fp, struct stat *st,
74 static int kqueue_close(struct file *fp);
75 static void kqueue_wakeup(struct kqueue *kq);
80 static struct fileops kqueueops = {
81 .fo_read = kqueue_read,
82 .fo_write = kqueue_write,
83 .fo_ioctl = kqueue_ioctl,
84 .fo_poll = kqueue_poll,
85 .fo_kqfilter = kqueue_kqfilter,
86 .fo_stat = kqueue_stat,
87 .fo_close = kqueue_close,
88 .fo_shutdown = nofo_shutdown
91 static void knote_attach(struct knote *kn);
92 static void knote_drop(struct knote *kn);
93 static void knote_enqueue(struct knote *kn);
94 static void knote_dequeue(struct knote *kn);
95 static void knote_init(void);
96 static struct knote *knote_alloc(void);
97 static void knote_free(struct knote *kn);
99 static void filt_kqdetach(struct knote *kn);
100 static int filt_kqueue(struct knote *kn, long hint);
101 static int filt_procattach(struct knote *kn);
102 static void filt_procdetach(struct knote *kn);
103 static int filt_proc(struct knote *kn, long hint);
104 static int filt_fileattach(struct knote *kn);
105 static void filt_timerexpire(void *knx);
106 static int filt_timerattach(struct knote *kn);
107 static void filt_timerdetach(struct knote *kn);
108 static int filt_timer(struct knote *kn, long hint);
110 static struct filterops file_filtops =
111 { 1, filt_fileattach, NULL, NULL };
112 static struct filterops kqread_filtops =
113 { 1, NULL, filt_kqdetach, filt_kqueue };
114 static struct filterops proc_filtops =
115 { 0, filt_procattach, filt_procdetach, filt_proc };
116 static struct filterops timer_filtops =
117 { 0, filt_timerattach, filt_timerdetach, filt_timer };
119 static vm_zone_t knote_zone;
120 static int kq_ncallouts = 0;
121 static int kq_calloutmax = (4 * 1024);
122 SYSCTL_INT(_kern, OID_AUTO, kq_calloutmax, CTLFLAG_RW,
123 &kq_calloutmax, 0, "Maximum number of callouts allocated for kqueue");
125 #define KNOTE_ACTIVATE(kn) do { \
126 kn->kn_status |= KN_ACTIVE; \
127 if ((kn->kn_status & (KN_QUEUED | KN_DISABLED)) == 0) \
131 #define KN_HASHSIZE 64 /* XXX should be tunable */
132 #define KN_HASH(val, mask) (((val) ^ (val >> 8)) & (mask))
134 extern struct filterops aio_filtops;
135 extern struct filterops sig_filtops;
138 * Table for for all system-defined filters.
140 static struct filterops *sysfilt_ops[] = {
141 &file_filtops, /* EVFILT_READ */
142 &file_filtops, /* EVFILT_WRITE */
143 &aio_filtops, /* EVFILT_AIO */
144 &file_filtops, /* EVFILT_VNODE */
145 &proc_filtops, /* EVFILT_PROC */
146 &sig_filtops, /* EVFILT_SIGNAL */
147 &timer_filtops, /* EVFILT_TIMER */
151 filt_fileattach(struct knote *kn)
153 return (fo_kqfilter(kn->kn_fp, kn));
157 * MPALMOSTSAFE - acquires mplock
160 kqueue_kqfilter(struct file *fp, struct knote *kn)
162 struct kqueue *kq = (struct kqueue *)kn->kn_fp->f_data;
165 if (kn->kn_filter != EVFILT_READ) {
170 kn->kn_fop = &kqread_filtops;
171 SLIST_INSERT_HEAD(&kq->kq_sel.si_note, kn, kn_selnext);
177 filt_kqdetach(struct knote *kn)
179 struct kqueue *kq = (struct kqueue *)kn->kn_fp->f_data;
181 SLIST_REMOVE(&kq->kq_sel.si_note, kn, knote, kn_selnext);
186 filt_kqueue(struct knote *kn, long hint)
188 struct kqueue *kq = (struct kqueue *)kn->kn_fp->f_data;
190 kn->kn_data = kq->kq_count;
191 return (kn->kn_data > 0);
195 filt_procattach(struct knote *kn)
201 lwkt_gettoken(&proc_token);
202 p = pfind(kn->kn_id);
203 if (p == NULL && (kn->kn_sfflags & NOTE_EXIT)) {
204 p = zpfind(kn->kn_id);
208 lwkt_reltoken(&proc_token);
211 if (!PRISON_CHECK(curthread->td_ucred, p->p_ucred)) {
212 lwkt_reltoken(&proc_token);
216 kn->kn_ptr.p_proc = p;
217 kn->kn_flags |= EV_CLEAR; /* automatically set */
220 * internal flag indicating registration done by kernel
222 if (kn->kn_flags & EV_FLAG1) {
223 kn->kn_data = kn->kn_sdata; /* ppid */
224 kn->kn_fflags = NOTE_CHILD;
225 kn->kn_flags &= ~EV_FLAG1;
228 /* XXX lock the proc here while adding to the list? */
229 SLIST_INSERT_HEAD(&p->p_klist, kn, kn_selnext);
232 * Immediately activate any exit notes if the target process is a
233 * zombie. This is necessary to handle the case where the target
234 * process, e.g. a child, dies before the kevent is registered.
236 if (immediate && filt_proc(kn, NOTE_EXIT))
238 lwkt_reltoken(&proc_token);
244 * The knote may be attached to a different process, which may exit,
245 * leaving nothing for the knote to be attached to. So when the process
246 * exits, the knote is marked as DETACHED and also flagged as ONESHOT so
247 * it will be deleted when read out. However, as part of the knote deletion,
248 * this routine is called, so a check is needed to avoid actually performing
249 * a detach, because the original process does not exist any more.
252 filt_procdetach(struct knote *kn)
256 if (kn->kn_status & KN_DETACHED)
258 /* XXX locking? this might modify another process. */
259 p = kn->kn_ptr.p_proc;
260 SLIST_REMOVE(&p->p_klist, kn, knote, kn_selnext);
264 filt_proc(struct knote *kn, long hint)
269 * mask off extra data
271 event = (u_int)hint & NOTE_PCTRLMASK;
274 * if the user is interested in this event, record it.
276 if (kn->kn_sfflags & event)
277 kn->kn_fflags |= event;
280 * Process is gone, so flag the event as finished. Detach the
281 * knote from the process now because the process will be poof,
284 if (event == NOTE_EXIT) {
285 struct proc *p = kn->kn_ptr.p_proc;
286 if ((kn->kn_status & KN_DETACHED) == 0) {
287 SLIST_REMOVE(&p->p_klist, kn, knote, kn_selnext);
288 kn->kn_status |= KN_DETACHED;
289 kn->kn_data = p->p_xstat;
290 kn->kn_ptr.p_proc = NULL;
292 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
297 * process forked, and user wants to track the new process,
298 * so attach a new knote to it, and immediately report an
299 * event with the parent's pid.
301 if ((event == NOTE_FORK) && (kn->kn_sfflags & NOTE_TRACK)) {
306 * register knote with new process.
308 kev.ident = hint & NOTE_PDATAMASK; /* pid */
309 kev.filter = kn->kn_filter;
310 kev.flags = kn->kn_flags | EV_ADD | EV_ENABLE | EV_FLAG1;
311 kev.fflags = kn->kn_sfflags;
312 kev.data = kn->kn_id; /* parent */
313 kev.udata = kn->kn_kevent.udata; /* preserve udata */
314 error = kqueue_register(kn->kn_kq, &kev);
316 kn->kn_fflags |= NOTE_TRACKERR;
319 return (kn->kn_fflags != 0);
323 filt_timerexpire(void *knx)
325 struct knote *kn = knx;
326 struct callout *calloutp;
333 if ((kn->kn_flags & EV_ONESHOT) == 0) {
334 tv.tv_sec = kn->kn_sdata / 1000;
335 tv.tv_usec = (kn->kn_sdata % 1000) * 1000;
336 tticks = tvtohz_high(&tv);
337 calloutp = (struct callout *)kn->kn_hook;
338 callout_reset(calloutp, tticks, filt_timerexpire, kn);
343 * data contains amount of time to sleep, in milliseconds
346 filt_timerattach(struct knote *kn)
348 struct callout *calloutp;
352 if (kq_ncallouts >= kq_calloutmax)
356 tv.tv_sec = kn->kn_sdata / 1000;
357 tv.tv_usec = (kn->kn_sdata % 1000) * 1000;
358 tticks = tvtohz_high(&tv);
360 kn->kn_flags |= EV_CLEAR; /* automatically set */
361 MALLOC(calloutp, struct callout *, sizeof(*calloutp),
363 callout_init(calloutp);
364 kn->kn_hook = (caddr_t)calloutp;
365 callout_reset(calloutp, tticks, filt_timerexpire, kn);
371 filt_timerdetach(struct knote *kn)
373 struct callout *calloutp;
375 calloutp = (struct callout *)kn->kn_hook;
376 callout_stop(calloutp);
377 FREE(calloutp, M_KQUEUE);
382 filt_timer(struct knote *kn, long hint)
385 return (kn->kn_data != 0);
389 * Initialize a kqueue.
391 * NOTE: The lwp/proc code initializes a kqueue for select/poll ops.
396 kqueue_init(struct kqueue *kq, struct filedesc *fdp)
398 TAILQ_INIT(&kq->kq_knpend);
399 TAILQ_INIT(&kq->kq_knlist);
404 * Terminate a kqueue. Freeing the actual kq itself is left up to the
405 * caller (it might be embedded in a lwp so we don't do it here).
408 kqueue_terminate(struct kqueue *kq)
414 while ((kn = TAILQ_FIRST(&kq->kq_knlist)) != NULL) {
415 kn->kn_fop->f_detach(kn);
416 if (kn->kn_fop->f_isfd) {
417 list = &kn->kn_fp->f_klist;
418 SLIST_REMOVE(list, kn, knote, kn_link);
422 hv = KN_HASH(kn->kn_id, kq->kq_knhashmask);
423 list = &kq->kq_knhash[hv];
424 SLIST_REMOVE(list, kn, knote, kn_link);
426 TAILQ_REMOVE(&kq->kq_knlist, kn, kn_kqlink);
427 if (kn->kn_status & KN_QUEUED)
433 kfree(kq->kq_knhash, M_KQUEUE);
434 kq->kq_knhash = NULL;
435 kq->kq_knhashmask = 0;
443 sys_kqueue(struct kqueue_args *uap)
445 struct thread *td = curthread;
450 error = falloc(td->td_lwp, &fp, &fd);
453 fp->f_flag = FREAD | FWRITE;
454 fp->f_type = DTYPE_KQUEUE;
455 fp->f_ops = &kqueueops;
457 kq = kmalloc(sizeof(struct kqueue), M_KQUEUE, M_WAITOK | M_ZERO);
458 kqueue_init(kq, td->td_proc->p_fd);
461 fsetfd(kq->kq_fdp, fp, fd);
462 uap->sysmsg_result = fd;
468 * Copy 'count' items into the destination list pointed to by uap->eventlist.
471 kevent_copyout(void *arg, struct kevent *kevp, int count)
473 struct kevent_args *uap;
476 uap = (struct kevent_args *)arg;
478 error = copyout(kevp, uap->eventlist, count * sizeof *kevp);
480 uap->eventlist += count;
485 * Copy 'count' items from the list pointed to by uap->changelist.
488 kevent_copyin(void *arg, struct kevent *kevp, int count)
490 struct kevent_args *uap;
493 uap = (struct kevent_args *)arg;
495 error = copyin(uap->changelist, kevp, count * sizeof *kevp);
497 uap->changelist += count;
505 kern_kevent(int fd, int nchanges, int nevents, struct kevent_args *uap,
506 k_copyin_fn kevent_copyinfn, k_copyout_fn kevent_copyoutfn,
507 struct timespec *tsp_in)
509 struct thread *td = curthread;
510 struct proc *p = td->td_proc;
513 struct file *fp = NULL;
515 struct timespec *tsp;
516 int i, n, total, nerrors, error;
517 struct kevent kev[KQ_NEVENTS];
521 fp = holdfp(p->p_fd, fd, -1);
524 if (fp->f_type != DTYPE_KQUEUE) {
529 kq = (struct kqueue *)fp->f_data;
533 while (nchanges > 0) {
534 n = nchanges > KQ_NEVENTS ? KQ_NEVENTS : nchanges;
535 error = kevent_copyinfn(uap, kev, n);
538 for (i = 0; i < n; i++) {
540 kevp->flags &= ~EV_SYSFLAGS;
541 error = kqueue_register(kq, kevp);
544 kevp->flags = EV_ERROR;
546 kevent_copyoutfn(uap, kevp, 1);
557 uap->sysmsg_result = nerrors;
563 * Acquire/wait for events - setup timeout
568 if (tsp->tv_sec || tsp->tv_nsec) {
570 timespecadd(tsp, &ats); /* tsp = target time */
577 * Collect as many events as we can. The timeout on successive
578 * loops is disabled (kqueue_scan() becomes non-blocking).
582 while ((n = nevents - total) > 0) {
585 i = kqueue_scan(kq, kev, n, tsp, &error);
588 error = kevent_copyoutfn(uap, kev, i);
592 tsp = &ts; /* successive loops non-blocking */
596 uap->sysmsg_result = total;
608 sys_kevent(struct kevent_args *uap)
610 struct timespec ts, *tsp;
614 error = copyin(uap->timeout, &ts, sizeof(ts));
622 error = kern_kevent(uap->fd, uap->nchanges, uap->nevents,
623 uap, kevent_copyin, kevent_copyout, tsp);
629 kqueue_register(struct kqueue *kq, struct kevent *kev)
631 struct filedesc *fdp = kq->kq_fdp;
632 struct filterops *fops;
633 struct file *fp = NULL;
634 struct knote *kn = NULL;
637 if (kev->filter < 0) {
638 if (kev->filter + EVFILT_SYSCOUNT < 0)
640 fops = sysfilt_ops[~kev->filter]; /* to 0-base index */
644 * filter attach routine is responsible for insuring that
645 * the identifier can be attached to it.
647 kprintf("unknown filter: %d\n", kev->filter);
652 /* validate descriptor */
653 fp = holdfp(fdp, kev->ident, -1);
657 SLIST_FOREACH(kn, &fp->f_klist, kn_link) {
658 if (kn->kn_kq == kq &&
659 kn->kn_filter == kev->filter &&
660 kn->kn_id == kev->ident) {
665 if (kq->kq_knhashmask) {
668 list = &kq->kq_knhash[
669 KN_HASH((u_long)kev->ident, kq->kq_knhashmask)];
670 SLIST_FOREACH(kn, list, kn_link) {
671 if (kn->kn_id == kev->ident &&
672 kn->kn_filter == kev->filter)
678 if (kn == NULL && ((kev->flags & EV_ADD) == 0)) {
684 * kn now contains the matching knote, or NULL if no match
686 if (kev->flags & EV_ADD) {
698 * apply reference count to knote structure, and
699 * do not release it at the end of this routine.
703 kn->kn_sfflags = kev->fflags;
704 kn->kn_sdata = kev->data;
707 kn->kn_kevent = *kev;
710 if ((error = fops->f_attach(kn)) != 0) {
716 * The user may change some filter values after the
717 * initial EV_ADD, but doing so will not reset any
718 * filter which have already been triggered.
720 kn->kn_sfflags = kev->fflags;
721 kn->kn_sdata = kev->data;
722 kn->kn_kevent.udata = kev->udata;
726 if (kn->kn_fop->f_event(kn, 0))
729 } else if (kev->flags & EV_DELETE) {
730 kn->kn_fop->f_detach(kn);
735 if ((kev->flags & EV_DISABLE) &&
736 ((kn->kn_status & KN_DISABLED) == 0)) {
738 kn->kn_status |= KN_DISABLED;
742 if ((kev->flags & EV_ENABLE) && (kn->kn_status & KN_DISABLED)) {
744 kn->kn_status &= ~KN_DISABLED;
745 if ((kn->kn_status & KN_ACTIVE) &&
746 ((kn->kn_status & KN_QUEUED) == 0))
758 * Scan the kqueue, blocking if necessary until the target time is reached.
759 * If tsp is NULL we block indefinitely. If tsp->ts_secs/nsecs are both
760 * 0 we do not block at all.
763 kqueue_scan(struct kqueue *kq, struct kevent *kevp, int count,
764 struct timespec *tsp, int *errorp)
766 struct knote *kn, marker;
772 if (kq->kq_count == 0) {
774 kq->kq_state |= KQ_SLEEP;
775 *errorp = tsleep(kq, PCATCH, "kqread", 0);
776 } else if (tsp->tv_sec == 0 && tsp->tv_nsec == 0) {
777 *errorp = EWOULDBLOCK;
780 struct timespec atx = *tsp;
784 timespecsub(&atx, &ats);
785 if (ats.tv_sec < 0) {
786 *errorp = EWOULDBLOCK;
788 timeout = atx.tv_sec > 24 * 60 * 60 ?
789 24 * 60 * 60 * hz : tstohz_high(&atx);
790 kq->kq_state |= KQ_SLEEP;
791 *errorp = tsleep(kq, PCATCH, "kqread", timeout);
797 /* don't restart after signals... */
798 if (*errorp == ERESTART)
800 else if (*errorp == EWOULDBLOCK)
806 * Collect events. Continuous mode events may get recycled
807 * past the marker so we stop when we hit it unless no events
808 * have been collected.
810 TAILQ_INSERT_TAIL(&kq->kq_knpend, &marker, kn_tqe);
812 kn = TAILQ_FIRST(&kq->kq_knpend);
815 TAILQ_REMOVE(&kq->kq_knpend, kn, kn_tqe);
816 if (kn->kn_status & KN_DISABLED) {
817 kn->kn_status &= ~KN_QUEUED;
821 if ((kn->kn_flags & EV_ONESHOT) == 0 &&
822 kn->kn_fop->f_event(kn, 0) == 0) {
823 kn->kn_status &= ~(KN_QUEUED | KN_ACTIVE);
827 *kevp++ = kn->kn_kevent;
832 * Post-event action on the note
834 if (kn->kn_flags & EV_ONESHOT) {
835 kn->kn_status &= ~KN_QUEUED;
838 kn->kn_fop->f_detach(kn);
841 } else if (kn->kn_flags & EV_CLEAR) {
844 kn->kn_status &= ~(KN_QUEUED | KN_ACTIVE);
847 TAILQ_INSERT_TAIL(&kq->kq_knpend, kn, kn_tqe);
850 TAILQ_REMOVE(&kq->kq_knpend, &marker, kn_tqe);
860 * This could be expanded to call kqueue_scan, if desired.
865 kqueue_read(struct file *fp, struct uio *uio, struct ucred *cred, int flags)
874 kqueue_write(struct file *fp, struct uio *uio, struct ucred *cred, int flags)
883 kqueue_ioctl(struct file *fp, u_long com, caddr_t data,
884 struct ucred *cred, struct sysmsg *msg)
890 kq = (struct kqueue *)fp->f_data;
895 kq->kq_state |= KQ_ASYNC;
897 kq->kq_state &= ~KQ_ASYNC;
901 error = fsetown(*(int *)data, &kq->kq_sigio);
912 * MPALMOSTSAFE - acquires mplock
915 kqueue_poll(struct file *fp, int events, struct ucred *cred)
917 struct kqueue *kq = (struct kqueue *)fp->f_data;
922 if (events & (POLLIN | POLLRDNORM)) {
924 revents |= events & (POLLIN | POLLRDNORM);
926 selrecord(curthread, &kq->kq_sel);
927 kq->kq_state |= KQ_SEL;
939 kqueue_stat(struct file *fp, struct stat *st, struct ucred *cred)
941 struct kqueue *kq = (struct kqueue *)fp->f_data;
943 bzero((void *)st, sizeof(*st));
944 st->st_size = kq->kq_count;
945 st->st_blksize = sizeof(struct kevent);
946 st->st_mode = S_IFIFO;
951 * MPALMOSTSAFE - acquires mplock
954 kqueue_close(struct file *fp)
956 struct kqueue *kq = (struct kqueue *)fp->f_data;
960 kqueue_terminate(kq);
963 funsetown(kq->kq_sigio);
971 kqueue_wakeup(struct kqueue *kq)
973 if (kq->kq_state & KQ_SLEEP) {
974 kq->kq_state &= ~KQ_SLEEP;
977 if (kq->kq_state & KQ_SEL) {
978 kq->kq_state &= ~KQ_SEL;
979 selwakeup(&kq->kq_sel);
981 KNOTE(&kq->kq_sel.si_note, 0);
985 * walk down a list of knotes, activating them if their event has triggered.
988 knote(struct klist *list, long hint)
992 SLIST_FOREACH(kn, list, kn_selnext)
993 if (kn->kn_fop->f_event(kn, hint))
998 * remove all knotes from a specified klist
1001 knote_remove(struct klist *list)
1005 while ((kn = SLIST_FIRST(list)) != NULL) {
1006 kn->kn_fop->f_detach(kn);
1012 * remove all knotes referencing a specified fd
1015 knote_fdclose(struct file *fp, struct filedesc *fdp, int fd)
1020 SLIST_FOREACH(kn, &fp->f_klist, kn_link) {
1021 if (kn->kn_kq->kq_fdp == fdp && kn->kn_id == fd) {
1022 kn->kn_fop->f_detach(kn);
1030 knote_attach(struct knote *kn)
1033 struct kqueue *kq = kn->kn_kq;
1035 if (kn->kn_fop->f_isfd) {
1036 KKASSERT(kn->kn_fp);
1037 list = &kn->kn_fp->f_klist;
1039 if (kq->kq_knhashmask == 0)
1040 kq->kq_knhash = hashinit(KN_HASHSIZE, M_KQUEUE,
1041 &kq->kq_knhashmask);
1042 list = &kq->kq_knhash[KN_HASH(kn->kn_id, kq->kq_knhashmask)];
1044 SLIST_INSERT_HEAD(list, kn, kn_link);
1045 TAILQ_INSERT_HEAD(&kq->kq_knlist, kn, kn_kqlink);
1050 * should be called outside of a critical section, since we don't want to
1051 * hold a critical section while calling fdrop and free.
1054 knote_drop(struct knote *kn)
1061 if (kn->kn_fop->f_isfd)
1062 list = &kn->kn_fp->f_klist;
1064 list = &kq->kq_knhash[KN_HASH(kn->kn_id, kq->kq_knhashmask)];
1066 SLIST_REMOVE(list, kn, knote, kn_link);
1067 TAILQ_REMOVE(&kq->kq_knlist, kn, kn_kqlink);
1068 if (kn->kn_status & KN_QUEUED)
1070 if (kn->kn_fop->f_isfd)
1077 knote_enqueue(struct knote *kn)
1079 struct kqueue *kq = kn->kn_kq;
1082 KASSERT((kn->kn_status & KN_QUEUED) == 0, ("knote already queued"));
1084 TAILQ_INSERT_TAIL(&kq->kq_knpend, kn, kn_tqe);
1085 kn->kn_status |= KN_QUEUED;
1089 * Send SIGIO on request (typically set up as a mailbox signal)
1091 if (kq->kq_sigio && (kq->kq_state & KQ_ASYNC) && kq->kq_count == 1)
1092 pgsigio(kq->kq_sigio, SIGIO, 0);
1098 knote_dequeue(struct knote *kn)
1100 struct kqueue *kq = kn->kn_kq;
1102 KASSERT(kn->kn_status & KN_QUEUED, ("knote not queued"));
1105 TAILQ_REMOVE(&kq->kq_knpend, kn, kn_tqe);
1106 kn->kn_status &= ~KN_QUEUED;
1114 knote_zone = zinit("KNOTE", sizeof(struct knote), 0, 0, 1);
1116 SYSINIT(knote, SI_SUB_PSEUDO, SI_ORDER_ANY, knote_init, NULL)
1118 static struct knote *
1121 return ((struct knote *)zalloc(knote_zone));
1125 knote_free(struct knote *kn)
1127 zfree(knote_zone, kn);