syscall messaging 3: Expand the 'header' that goes in front of the syscall
[dragonfly.git] / sys / kern / kern_event.c
CommitLineData
984263bc
MD
1/*-
2 * Copyright (c) 1999,2000,2001 Jonathan Lemon <jlemon@FreeBSD.org>
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
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.
13 *
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
24 * SUCH DAMAGE.
25 *
26 * $FreeBSD: src/sys/kern/kern_event.c,v 1.2.2.9 2003/05/08 07:47:16 kbyanc Exp $
c7114eea 27 * $DragonFly: src/sys/kern/kern_event.c,v 1.9 2003/07/30 00:19:14 dillon Exp $
984263bc
MD
28 */
29
30#include <sys/param.h>
31#include <sys/systm.h>
32#include <sys/kernel.h>
33#include <sys/proc.h>
34#include <sys/malloc.h>
35#include <sys/unistd.h>
36#include <sys/file.h>
37#include <sys/fcntl.h>
38#include <sys/select.h>
39#include <sys/queue.h>
40#include <sys/event.h>
41#include <sys/eventvar.h>
42#include <sys/poll.h>
43#include <sys/protosw.h>
44#include <sys/socket.h>
45#include <sys/socketvar.h>
46#include <sys/stat.h>
47#include <sys/sysctl.h>
48#include <sys/sysproto.h>
49#include <sys/uio.h>
dadab5e9 50#include <sys/file2.h>
984263bc
MD
51
52#include <vm/vm_zone.h>
53
54MALLOC_DEFINE(M_KQUEUE, "kqueue", "memory for kqueue system");
55
56static int kqueue_scan(struct file *fp, int maxevents,
57 struct kevent *ulistp, const struct timespec *timeout,
90b9818c 58 struct proc *p, int *res);
984263bc 59static int kqueue_read(struct file *fp, struct uio *uio,
dadab5e9 60 struct ucred *cred, int flags, struct thread *td);
984263bc 61static int kqueue_write(struct file *fp, struct uio *uio,
dadab5e9 62 struct ucred *cred, int flags, struct thread *td);
984263bc 63static int kqueue_ioctl(struct file *fp, u_long com, caddr_t data,
dadab5e9 64 struct thread *td);
984263bc 65static int kqueue_poll(struct file *fp, int events, struct ucred *cred,
dadab5e9 66 struct thread *td);
984263bc 67static int kqueue_kqfilter(struct file *fp, struct knote *kn);
dadab5e9
MD
68static int kqueue_stat(struct file *fp, struct stat *st, struct thread *td);
69static int kqueue_close(struct file *fp, struct thread *td);
984263bc
MD
70static void kqueue_wakeup(struct kqueue *kq);
71
72static struct fileops kqueueops = {
f53ede20
MD
73 NULL, /* port */
74 0, /* autoq */
984263bc
MD
75 kqueue_read,
76 kqueue_write,
77 kqueue_ioctl,
78 kqueue_poll,
79 kqueue_kqfilter,
80 kqueue_stat,
81 kqueue_close
82};
83
84static void knote_attach(struct knote *kn, struct filedesc *fdp);
dadab5e9 85static void knote_drop(struct knote *kn, struct thread *td);
984263bc
MD
86static void knote_enqueue(struct knote *kn);
87static void knote_dequeue(struct knote *kn);
88static void knote_init(void);
89static struct knote *knote_alloc(void);
90static void knote_free(struct knote *kn);
91
92static void filt_kqdetach(struct knote *kn);
93static int filt_kqueue(struct knote *kn, long hint);
94static int filt_procattach(struct knote *kn);
95static void filt_procdetach(struct knote *kn);
96static int filt_proc(struct knote *kn, long hint);
97static int filt_fileattach(struct knote *kn);
98static void filt_timerexpire(void *knx);
99static int filt_timerattach(struct knote *kn);
100static void filt_timerdetach(struct knote *kn);
101static int filt_timer(struct knote *kn, long hint);
102
103static struct filterops file_filtops =
104 { 1, filt_fileattach, NULL, NULL };
105static struct filterops kqread_filtops =
106 { 1, NULL, filt_kqdetach, filt_kqueue };
107static struct filterops proc_filtops =
108 { 0, filt_procattach, filt_procdetach, filt_proc };
109static struct filterops timer_filtops =
110 { 0, filt_timerattach, filt_timerdetach, filt_timer };
111
112static vm_zone_t knote_zone;
113static int kq_ncallouts = 0;
114static int kq_calloutmax = (4 * 1024);
115SYSCTL_INT(_kern, OID_AUTO, kq_calloutmax, CTLFLAG_RW,
116 &kq_calloutmax, 0, "Maximum number of callouts allocated for kqueue");
117
118#define KNOTE_ACTIVATE(kn) do { \
119 kn->kn_status |= KN_ACTIVE; \
120 if ((kn->kn_status & (KN_QUEUED | KN_DISABLED)) == 0) \
121 knote_enqueue(kn); \
122} while(0)
123
124#define KN_HASHSIZE 64 /* XXX should be tunable */
125#define KN_HASH(val, mask) (((val) ^ (val >> 8)) & (mask))
126
127extern struct filterops aio_filtops;
128extern struct filterops sig_filtops;
129
130/*
131 * Table for for all system-defined filters.
132 */
133static struct filterops *sysfilt_ops[] = {
134 &file_filtops, /* EVFILT_READ */
135 &file_filtops, /* EVFILT_WRITE */
136 &aio_filtops, /* EVFILT_AIO */
137 &file_filtops, /* EVFILT_VNODE */
138 &proc_filtops, /* EVFILT_PROC */
139 &sig_filtops, /* EVFILT_SIGNAL */
140 &timer_filtops, /* EVFILT_TIMER */
141};
142
143static int
144filt_fileattach(struct knote *kn)
145{
984263bc
MD
146 return (fo_kqfilter(kn->kn_fp, kn));
147}
148
149/*ARGSUSED*/
150static int
151kqueue_kqfilter(struct file *fp, struct knote *kn)
152{
153 struct kqueue *kq = (struct kqueue *)kn->kn_fp->f_data;
154
155 if (kn->kn_filter != EVFILT_READ)
156 return (1);
157
158 kn->kn_fop = &kqread_filtops;
159 SLIST_INSERT_HEAD(&kq->kq_sel.si_note, kn, kn_selnext);
160 return (0);
161}
162
163static void
164filt_kqdetach(struct knote *kn)
165{
166 struct kqueue *kq = (struct kqueue *)kn->kn_fp->f_data;
167
168 SLIST_REMOVE(&kq->kq_sel.si_note, kn, knote, kn_selnext);
169}
170
171/*ARGSUSED*/
172static int
173filt_kqueue(struct knote *kn, long hint)
174{
175 struct kqueue *kq = (struct kqueue *)kn->kn_fp->f_data;
176
177 kn->kn_data = kq->kq_count;
178 return (kn->kn_data > 0);
179}
180
181static int
182filt_procattach(struct knote *kn)
183{
184 struct proc *p;
185 int immediate;
186
187 immediate = 0;
188 p = pfind(kn->kn_id);
189 if (p == NULL && (kn->kn_sfflags & NOTE_EXIT)) {
190 p = zpfind(kn->kn_id);
191 immediate = 1;
192 }
193 if (p == NULL)
194 return (ESRCH);
41c20dac 195 if (! PRISON_CHECK(curproc->p_ucred, p->p_ucred))
984263bc
MD
196 return (EACCES);
197
198 kn->kn_ptr.p_proc = p;
199 kn->kn_flags |= EV_CLEAR; /* automatically set */
200
201 /*
202 * internal flag indicating registration done by kernel
203 */
204 if (kn->kn_flags & EV_FLAG1) {
205 kn->kn_data = kn->kn_sdata; /* ppid */
206 kn->kn_fflags = NOTE_CHILD;
207 kn->kn_flags &= ~EV_FLAG1;
208 }
209
210 /* XXX lock the proc here while adding to the list? */
211 SLIST_INSERT_HEAD(&p->p_klist, kn, kn_selnext);
212
213 /*
214 * Immediately activate any exit notes if the target process is a
215 * zombie. This is necessary to handle the case where the target
216 * process, e.g. a child, dies before the kevent is registered.
217 */
218 if (immediate && filt_proc(kn, NOTE_EXIT))
219 KNOTE_ACTIVATE(kn);
220
221 return (0);
222}
223
224/*
225 * The knote may be attached to a different process, which may exit,
226 * leaving nothing for the knote to be attached to. So when the process
227 * exits, the knote is marked as DETACHED and also flagged as ONESHOT so
228 * it will be deleted when read out. However, as part of the knote deletion,
229 * this routine is called, so a check is needed to avoid actually performing
230 * a detach, because the original process does not exist any more.
231 */
232static void
233filt_procdetach(struct knote *kn)
234{
235 struct proc *p = kn->kn_ptr.p_proc;
236
237 if (kn->kn_status & KN_DETACHED)
238 return;
239
240 /* XXX locking? this might modify another process. */
241 SLIST_REMOVE(&p->p_klist, kn, knote, kn_selnext);
242}
243
244static int
245filt_proc(struct knote *kn, long hint)
246{
247 u_int event;
248
249 /*
250 * mask off extra data
251 */
252 event = (u_int)hint & NOTE_PCTRLMASK;
253
254 /*
255 * if the user is interested in this event, record it.
256 */
257 if (kn->kn_sfflags & event)
258 kn->kn_fflags |= event;
259
260 /*
261 * process is gone, so flag the event as finished.
262 */
263 if (event == NOTE_EXIT) {
264 kn->kn_status |= KN_DETACHED;
265 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
266 return (1);
267 }
268
269 /*
270 * process forked, and user wants to track the new process,
271 * so attach a new knote to it, and immediately report an
272 * event with the parent's pid.
273 */
274 if ((event == NOTE_FORK) && (kn->kn_sfflags & NOTE_TRACK)) {
275 struct kevent kev;
276 int error;
277
278 /*
279 * register knote with new process.
280 */
281 kev.ident = hint & NOTE_PDATAMASK; /* pid */
282 kev.filter = kn->kn_filter;
283 kev.flags = kn->kn_flags | EV_ADD | EV_ENABLE | EV_FLAG1;
284 kev.fflags = kn->kn_sfflags;
285 kev.data = kn->kn_id; /* parent */
286 kev.udata = kn->kn_kevent.udata; /* preserve udata */
287 error = kqueue_register(kn->kn_kq, &kev, NULL);
288 if (error)
289 kn->kn_fflags |= NOTE_TRACKERR;
290 }
291
292 return (kn->kn_fflags != 0);
293}
294
295static void
296filt_timerexpire(void *knx)
297{
298 struct knote *kn = knx;
299 struct callout *calloutp;
300 struct timeval tv;
301 int tticks;
302
303 kn->kn_data++;
304 KNOTE_ACTIVATE(kn);
305
306 if ((kn->kn_flags & EV_ONESHOT) == 0) {
307 tv.tv_sec = kn->kn_sdata / 1000;
308 tv.tv_usec = (kn->kn_sdata % 1000) * 1000;
309 tticks = tvtohz(&tv);
310 calloutp = (struct callout *)kn->kn_hook;
311 callout_reset(calloutp, tticks, filt_timerexpire, kn);
312 }
313}
314
315/*
316 * data contains amount of time to sleep, in milliseconds
317 */
318static int
319filt_timerattach(struct knote *kn)
320{
321 struct callout *calloutp;
322 struct timeval tv;
323 int tticks;
324
325 if (kq_ncallouts >= kq_calloutmax)
326 return (ENOMEM);
327 kq_ncallouts++;
328
329 tv.tv_sec = kn->kn_sdata / 1000;
330 tv.tv_usec = (kn->kn_sdata % 1000) * 1000;
331 tticks = tvtohz(&tv);
332
333 kn->kn_flags |= EV_CLEAR; /* automatically set */
334 MALLOC(calloutp, struct callout *, sizeof(*calloutp),
335 M_KQUEUE, M_WAITOK);
336 callout_init(calloutp);
337 callout_reset(calloutp, tticks, filt_timerexpire, kn);
338 kn->kn_hook = (caddr_t)calloutp;
339
340 return (0);
341}
342
343static void
344filt_timerdetach(struct knote *kn)
345{
346 struct callout *calloutp;
347
348 calloutp = (struct callout *)kn->kn_hook;
349 callout_stop(calloutp);
350 FREE(calloutp, M_KQUEUE);
351 kq_ncallouts--;
352}
353
354static int
355filt_timer(struct knote *kn, long hint)
356{
357
358 return (kn->kn_data != 0);
359}
360
361int
41c20dac 362kqueue(struct kqueue_args *uap)
984263bc 363{
41c20dac 364 struct proc *p = curproc;
984263bc
MD
365 struct filedesc *fdp = p->p_fd;
366 struct kqueue *kq;
367 struct file *fp;
368 int fd, error;
369
370 error = falloc(p, &fp, &fd);
371 if (error)
372 return (error);
373 fp->f_flag = FREAD | FWRITE;
374 fp->f_type = DTYPE_KQUEUE;
375 fp->f_ops = &kqueueops;
376 kq = malloc(sizeof(struct kqueue), M_KQUEUE, M_WAITOK | M_ZERO);
377 TAILQ_INIT(&kq->kq_head);
378 fp->f_data = (caddr_t)kq;
c7114eea 379 uap->sysmsg_result = fd;
984263bc
MD
380 if (fdp->fd_knlistsize < 0)
381 fdp->fd_knlistsize = 0; /* this process has a kq */
382 kq->kq_fdp = fdp;
383 return (error);
384}
385
984263bc 386int
41c20dac 387kevent(struct kevent_args *uap)
984263bc 388{
dadab5e9
MD
389 struct thread *td = curthread;
390 struct proc *p = td->td_proc;
391 struct filedesc *fdp;
984263bc
MD
392 struct kevent *kevp;
393 struct kqueue *kq;
394 struct file *fp = NULL;
395 struct timespec ts;
396 int i, n, nerrors, error;
397
dadab5e9
MD
398 KKASSERT(p);
399 fdp = p->p_fd;
400
984263bc
MD
401 if (((u_int)uap->fd) >= fdp->fd_nfiles ||
402 (fp = fdp->fd_ofiles[uap->fd]) == NULL ||
403 (fp->f_type != DTYPE_KQUEUE))
404 return (EBADF);
405
406 fhold(fp);
407
408 if (uap->timeout != NULL) {
409 error = copyin(uap->timeout, &ts, sizeof(ts));
410 if (error)
411 goto done;
412 uap->timeout = &ts;
413 }
414
415 kq = (struct kqueue *)fp->f_data;
416 nerrors = 0;
417
418 while (uap->nchanges > 0) {
419 n = uap->nchanges > KQ_NEVENTS ? KQ_NEVENTS : uap->nchanges;
420 error = copyin(uap->changelist, kq->kq_kev,
421 n * sizeof(struct kevent));
422 if (error)
423 goto done;
424 for (i = 0; i < n; i++) {
425 kevp = &kq->kq_kev[i];
426 kevp->flags &= ~EV_SYSFLAGS;
dadab5e9 427 error = kqueue_register(kq, kevp, td);
984263bc
MD
428 if (error) {
429 if (uap->nevents != 0) {
430 kevp->flags = EV_ERROR;
431 kevp->data = error;
432 (void) copyout((caddr_t)kevp,
433 (caddr_t)uap->eventlist,
434 sizeof(*kevp));
435 uap->eventlist++;
436 uap->nevents--;
437 nerrors++;
438 } else {
439 goto done;
440 }
441 }
442 }
443 uap->nchanges -= n;
444 uap->changelist += n;
445 }
446 if (nerrors) {
c7114eea 447 uap->sysmsg_result = nerrors;
984263bc
MD
448 error = 0;
449 goto done;
450 }
451
c7114eea 452 error = kqueue_scan(fp, uap->nevents, uap->eventlist, uap->timeout, p, &uap->sysmsg_result);
984263bc
MD
453done:
454 if (fp != NULL)
dadab5e9 455 fdrop(fp, p->p_thread);
984263bc
MD
456 return (error);
457}
458
459int
dadab5e9 460kqueue_register(struct kqueue *kq, struct kevent *kev, struct thread *td)
984263bc
MD
461{
462 struct filedesc *fdp = kq->kq_fdp;
463 struct filterops *fops;
464 struct file *fp = NULL;
465 struct knote *kn = NULL;
466 int s, error = 0;
467
468 if (kev->filter < 0) {
469 if (kev->filter + EVFILT_SYSCOUNT < 0)
470 return (EINVAL);
471 fops = sysfilt_ops[~kev->filter]; /* to 0-base index */
472 } else {
473 /*
474 * XXX
475 * filter attach routine is responsible for insuring that
476 * the identifier can be attached to it.
477 */
478 printf("unknown filter: %d\n", kev->filter);
479 return (EINVAL);
480 }
481
482 if (fops->f_isfd) {
483 /* validate descriptor */
484 if ((u_int)kev->ident >= fdp->fd_nfiles ||
485 (fp = fdp->fd_ofiles[kev->ident]) == NULL)
486 return (EBADF);
487 fhold(fp);
488
489 if (kev->ident < fdp->fd_knlistsize) {
490 SLIST_FOREACH(kn, &fdp->fd_knlist[kev->ident], kn_link)
491 if (kq == kn->kn_kq &&
492 kev->filter == kn->kn_filter)
493 break;
494 }
495 } else {
496 if (fdp->fd_knhashmask != 0) {
497 struct klist *list;
498
499 list = &fdp->fd_knhash[
500 KN_HASH((u_long)kev->ident, fdp->fd_knhashmask)];
501 SLIST_FOREACH(kn, list, kn_link)
502 if (kev->ident == kn->kn_id &&
503 kq == kn->kn_kq &&
504 kev->filter == kn->kn_filter)
505 break;
506 }
507 }
508
509 if (kn == NULL && ((kev->flags & EV_ADD) == 0)) {
510 error = ENOENT;
511 goto done;
512 }
513
514 /*
515 * kn now contains the matching knote, or NULL if no match
516 */
517 if (kev->flags & EV_ADD) {
518
519 if (kn == NULL) {
520 kn = knote_alloc();
521 if (kn == NULL) {
522 error = ENOMEM;
523 goto done;
524 }
525 kn->kn_fp = fp;
526 kn->kn_kq = kq;
527 kn->kn_fop = fops;
528
529 /*
530 * apply reference count to knote structure, and
531 * do not release it at the end of this routine.
532 */
533 fp = NULL;
534
535 kn->kn_sfflags = kev->fflags;
536 kn->kn_sdata = kev->data;
537 kev->fflags = 0;
538 kev->data = 0;
539 kn->kn_kevent = *kev;
540
541 knote_attach(kn, fdp);
542 if ((error = fops->f_attach(kn)) != 0) {
dadab5e9 543 knote_drop(kn, td);
984263bc
MD
544 goto done;
545 }
546 } else {
547 /*
548 * The user may change some filter values after the
549 * initial EV_ADD, but doing so will not reset any
550 * filter which have already been triggered.
551 */
552 kn->kn_sfflags = kev->fflags;
553 kn->kn_sdata = kev->data;
554 kn->kn_kevent.udata = kev->udata;
555 }
556
557 s = splhigh();
558 if (kn->kn_fop->f_event(kn, 0))
559 KNOTE_ACTIVATE(kn);
560 splx(s);
561
562 } else if (kev->flags & EV_DELETE) {
563 kn->kn_fop->f_detach(kn);
dadab5e9 564 knote_drop(kn, td);
984263bc
MD
565 goto done;
566 }
567
568 if ((kev->flags & EV_DISABLE) &&
569 ((kn->kn_status & KN_DISABLED) == 0)) {
570 s = splhigh();
571 kn->kn_status |= KN_DISABLED;
572 splx(s);
573 }
574
575 if ((kev->flags & EV_ENABLE) && (kn->kn_status & KN_DISABLED)) {
576 s = splhigh();
577 kn->kn_status &= ~KN_DISABLED;
578 if ((kn->kn_status & KN_ACTIVE) &&
579 ((kn->kn_status & KN_QUEUED) == 0))
580 knote_enqueue(kn);
581 splx(s);
582 }
583
584done:
585 if (fp != NULL)
dadab5e9 586 fdrop(fp, td);
984263bc
MD
587 return (error);
588}
589
590static int
591kqueue_scan(struct file *fp, int maxevents, struct kevent *ulistp,
90b9818c 592 const struct timespec *tsp, struct proc *p, int *res)
984263bc 593{
dadab5e9 594 struct thread *td = p->p_thread;
984263bc
MD
595 struct kqueue *kq = (struct kqueue *)fp->f_data;
596 struct kevent *kevp;
597 struct timeval atv, rtv, ttv;
598 struct knote *kn, marker;
599 int s, count, timeout, nkev = 0, error = 0;
600
601 count = maxevents;
602 if (count == 0)
603 goto done;
604
605 if (tsp != NULL) {
606 TIMESPEC_TO_TIMEVAL(&atv, tsp);
607 if (itimerfix(&atv)) {
608 error = EINVAL;
609 goto done;
610 }
611 if (tsp->tv_sec == 0 && tsp->tv_nsec == 0)
612 timeout = -1;
613 else
614 timeout = atv.tv_sec > 24 * 60 * 60 ?
615 24 * 60 * 60 * hz : tvtohz(&atv);
616 getmicrouptime(&rtv);
617 timevaladd(&atv, &rtv);
618 } else {
619 atv.tv_sec = 0;
620 atv.tv_usec = 0;
621 timeout = 0;
622 }
623 goto start;
624
625retry:
626 if (atv.tv_sec || atv.tv_usec) {
627 getmicrouptime(&rtv);
628 if (timevalcmp(&rtv, &atv, >=))
629 goto done;
630 ttv = atv;
631 timevalsub(&ttv, &rtv);
632 timeout = ttv.tv_sec > 24 * 60 * 60 ?
633 24 * 60 * 60 * hz : tvtohz(&ttv);
634 }
635
636start:
637 kevp = kq->kq_kev;
638 s = splhigh();
639 if (kq->kq_count == 0) {
640 if (timeout < 0) {
641 error = EWOULDBLOCK;
642 } else {
643 kq->kq_state |= KQ_SLEEP;
377d4740 644 error = tsleep(kq, PCATCH, "kqread", timeout);
984263bc
MD
645 }
646 splx(s);
647 if (error == 0)
648 goto retry;
649 /* don't restart after signals... */
650 if (error == ERESTART)
651 error = EINTR;
652 else if (error == EWOULDBLOCK)
653 error = 0;
654 goto done;
655 }
656
657 TAILQ_INSERT_TAIL(&kq->kq_head, &marker, kn_tqe);
658 while (count) {
659 kn = TAILQ_FIRST(&kq->kq_head);
660 TAILQ_REMOVE(&kq->kq_head, kn, kn_tqe);
661 if (kn == &marker) {
662 splx(s);
663 if (count == maxevents)
664 goto retry;
665 goto done;
666 }
667 if (kn->kn_status & KN_DISABLED) {
668 kn->kn_status &= ~KN_QUEUED;
669 kq->kq_count--;
670 continue;
671 }
672 if ((kn->kn_flags & EV_ONESHOT) == 0 &&
673 kn->kn_fop->f_event(kn, 0) == 0) {
674 kn->kn_status &= ~(KN_QUEUED | KN_ACTIVE);
675 kq->kq_count--;
676 continue;
677 }
678 *kevp = kn->kn_kevent;
679 kevp++;
680 nkev++;
681 if (kn->kn_flags & EV_ONESHOT) {
682 kn->kn_status &= ~KN_QUEUED;
683 kq->kq_count--;
684 splx(s);
685 kn->kn_fop->f_detach(kn);
dadab5e9 686 knote_drop(kn, td);
984263bc
MD
687 s = splhigh();
688 } else if (kn->kn_flags & EV_CLEAR) {
689 kn->kn_data = 0;
690 kn->kn_fflags = 0;
691 kn->kn_status &= ~(KN_QUEUED | KN_ACTIVE);
692 kq->kq_count--;
693 } else {
694 TAILQ_INSERT_TAIL(&kq->kq_head, kn, kn_tqe);
695 }
696 count--;
697 if (nkev == KQ_NEVENTS) {
698 splx(s);
699 error = copyout((caddr_t)&kq->kq_kev, (caddr_t)ulistp,
700 sizeof(struct kevent) * nkev);
701 ulistp += nkev;
702 nkev = 0;
703 kevp = kq->kq_kev;
704 s = splhigh();
705 if (error)
706 break;
707 }
708 }
709 TAILQ_REMOVE(&kq->kq_head, &marker, kn_tqe);
710 splx(s);
711done:
712 if (nkev != 0)
713 error = copyout((caddr_t)&kq->kq_kev, (caddr_t)ulistp,
714 sizeof(struct kevent) * nkev);
90b9818c 715 *res = maxevents - count;
984263bc
MD
716 return (error);
717}
718
719/*
720 * XXX
721 * This could be expanded to call kqueue_scan, if desired.
722 */
723/*ARGSUSED*/
724static int
725kqueue_read(struct file *fp, struct uio *uio, struct ucred *cred,
dadab5e9 726 int flags, struct thread *td)
984263bc
MD
727{
728 return (ENXIO);
729}
730
731/*ARGSUSED*/
732static int
733kqueue_write(struct file *fp, struct uio *uio, struct ucred *cred,
dadab5e9 734 int flags, struct thread *td)
984263bc
MD
735{
736 return (ENXIO);
737}
738
739/*ARGSUSED*/
740static int
dadab5e9 741kqueue_ioctl(struct file *fp, u_long com, caddr_t data, struct thread *td)
984263bc
MD
742{
743 return (ENOTTY);
744}
745
746/*ARGSUSED*/
747static int
dadab5e9 748kqueue_poll(struct file *fp, int events, struct ucred *cred, struct thread *td)
984263bc
MD
749{
750 struct kqueue *kq = (struct kqueue *)fp->f_data;
751 int revents = 0;
752 int s = splnet();
753
754 if (events & (POLLIN | POLLRDNORM)) {
755 if (kq->kq_count) {
756 revents |= events & (POLLIN | POLLRDNORM);
757 } else {
dadab5e9 758 selrecord(td, &kq->kq_sel);
984263bc
MD
759 kq->kq_state |= KQ_SEL;
760 }
761 }
762 splx(s);
763 return (revents);
764}
765
766/*ARGSUSED*/
767static int
dadab5e9 768kqueue_stat(struct file *fp, struct stat *st, struct thread *td)
984263bc
MD
769{
770 struct kqueue *kq = (struct kqueue *)fp->f_data;
771
772 bzero((void *)st, sizeof(*st));
773 st->st_size = kq->kq_count;
774 st->st_blksize = sizeof(struct kevent);
775 st->st_mode = S_IFIFO;
776 return (0);
777}
778
779/*ARGSUSED*/
780static int
dadab5e9 781kqueue_close(struct file *fp, struct thread *td)
984263bc 782{
dadab5e9 783 struct proc *p = td->td_proc;
984263bc 784 struct kqueue *kq = (struct kqueue *)fp->f_data;
dadab5e9 785 struct filedesc *fdp;
984263bc
MD
786 struct knote **knp, *kn, *kn0;
787 int i;
788
dadab5e9
MD
789 KKASSERT(p);
790 fdp = p->p_fd;
984263bc
MD
791 for (i = 0; i < fdp->fd_knlistsize; i++) {
792 knp = &SLIST_FIRST(&fdp->fd_knlist[i]);
793 kn = *knp;
794 while (kn != NULL) {
795 kn0 = SLIST_NEXT(kn, kn_link);
796 if (kq == kn->kn_kq) {
797 kn->kn_fop->f_detach(kn);
dadab5e9 798 fdrop(kn->kn_fp, td);
984263bc
MD
799 knote_free(kn);
800 *knp = kn0;
801 } else {
802 knp = &SLIST_NEXT(kn, kn_link);
803 }
804 kn = kn0;
805 }
806 }
807 if (fdp->fd_knhashmask != 0) {
808 for (i = 0; i < fdp->fd_knhashmask + 1; i++) {
809 knp = &SLIST_FIRST(&fdp->fd_knhash[i]);
810 kn = *knp;
811 while (kn != NULL) {
812 kn0 = SLIST_NEXT(kn, kn_link);
813 if (kq == kn->kn_kq) {
814 kn->kn_fop->f_detach(kn);
815 /* XXX non-fd release of kn->kn_ptr */
816 knote_free(kn);
817 *knp = kn0;
818 } else {
819 knp = &SLIST_NEXT(kn, kn_link);
820 }
821 kn = kn0;
822 }
823 }
824 }
825 free(kq, M_KQUEUE);
826 fp->f_data = NULL;
827
828 return (0);
829}
830
831static void
832kqueue_wakeup(struct kqueue *kq)
833{
834
835 if (kq->kq_state & KQ_SLEEP) {
836 kq->kq_state &= ~KQ_SLEEP;
837 wakeup(kq);
838 }
839 if (kq->kq_state & KQ_SEL) {
840 kq->kq_state &= ~KQ_SEL;
841 selwakeup(&kq->kq_sel);
842 }
843 KNOTE(&kq->kq_sel.si_note, 0);
844}
845
846/*
847 * walk down a list of knotes, activating them if their event has triggered.
848 */
849void
850knote(struct klist *list, long hint)
851{
852 struct knote *kn;
853
854 SLIST_FOREACH(kn, list, kn_selnext)
855 if (kn->kn_fop->f_event(kn, hint))
856 KNOTE_ACTIVATE(kn);
857}
858
859/*
860 * remove all knotes from a specified klist
861 */
862void
dadab5e9 863knote_remove(struct thread *td, struct klist *list)
984263bc
MD
864{
865 struct knote *kn;
866
867 while ((kn = SLIST_FIRST(list)) != NULL) {
868 kn->kn_fop->f_detach(kn);
dadab5e9 869 knote_drop(kn, td);
984263bc
MD
870 }
871}
872
873/*
874 * remove all knotes referencing a specified fd
875 */
876void
877knote_fdclose(struct proc *p, int fd)
878{
879 struct filedesc *fdp = p->p_fd;
880 struct klist *list = &fdp->fd_knlist[fd];
881
dadab5e9 882 knote_remove(p->p_thread, list);
984263bc
MD
883}
884
885static void
886knote_attach(struct knote *kn, struct filedesc *fdp)
887{
888 struct klist *list;
889 int size;
890
891 if (! kn->kn_fop->f_isfd) {
892 if (fdp->fd_knhashmask == 0)
893 fdp->fd_knhash = hashinit(KN_HASHSIZE, M_KQUEUE,
894 &fdp->fd_knhashmask);
895 list = &fdp->fd_knhash[KN_HASH(kn->kn_id, fdp->fd_knhashmask)];
896 goto done;
897 }
898
899 if (fdp->fd_knlistsize <= kn->kn_id) {
900 size = fdp->fd_knlistsize;
901 while (size <= kn->kn_id)
902 size += KQEXTENT;
903 MALLOC(list, struct klist *,
904 size * sizeof(struct klist *), M_KQUEUE, M_WAITOK);
905 bcopy((caddr_t)fdp->fd_knlist, (caddr_t)list,
906 fdp->fd_knlistsize * sizeof(struct klist *));
907 bzero((caddr_t)list +
908 fdp->fd_knlistsize * sizeof(struct klist *),
909 (size - fdp->fd_knlistsize) * sizeof(struct klist *));
910 if (fdp->fd_knlist != NULL)
911 FREE(fdp->fd_knlist, M_KQUEUE);
912 fdp->fd_knlistsize = size;
913 fdp->fd_knlist = list;
914 }
915 list = &fdp->fd_knlist[kn->kn_id];
916done:
917 SLIST_INSERT_HEAD(list, kn, kn_link);
918 kn->kn_status = 0;
919}
920
921/*
922 * should be called at spl == 0, since we don't want to hold spl
923 * while calling fdrop and free.
924 */
925static void
dadab5e9 926knote_drop(struct knote *kn, struct thread *td)
984263bc 927{
dadab5e9 928 struct filedesc *fdp;
984263bc
MD
929 struct klist *list;
930
dadab5e9
MD
931 KKASSERT(td->td_proc);
932 fdp = td->td_proc->p_fd;
984263bc
MD
933 if (kn->kn_fop->f_isfd)
934 list = &fdp->fd_knlist[kn->kn_id];
935 else
936 list = &fdp->fd_knhash[KN_HASH(kn->kn_id, fdp->fd_knhashmask)];
937
938 SLIST_REMOVE(list, kn, knote, kn_link);
939 if (kn->kn_status & KN_QUEUED)
940 knote_dequeue(kn);
941 if (kn->kn_fop->f_isfd)
dadab5e9 942 fdrop(kn->kn_fp, td);
984263bc
MD
943 knote_free(kn);
944}
945
946
947static void
948knote_enqueue(struct knote *kn)
949{
950 struct kqueue *kq = kn->kn_kq;
951 int s = splhigh();
952
953 KASSERT((kn->kn_status & KN_QUEUED) == 0, ("knote already queued"));
954
955 TAILQ_INSERT_TAIL(&kq->kq_head, kn, kn_tqe);
956 kn->kn_status |= KN_QUEUED;
957 kq->kq_count++;
958 splx(s);
959 kqueue_wakeup(kq);
960}
961
962static void
963knote_dequeue(struct knote *kn)
964{
965 struct kqueue *kq = kn->kn_kq;
966 int s = splhigh();
967
968 KASSERT(kn->kn_status & KN_QUEUED, ("knote not queued"));
969
970 TAILQ_REMOVE(&kq->kq_head, kn, kn_tqe);
971 kn->kn_status &= ~KN_QUEUED;
972 kq->kq_count--;
973 splx(s);
974}
975
976static void
977knote_init(void)
978{
979 knote_zone = zinit("KNOTE", sizeof(struct knote), 0, 0, 1);
980}
981SYSINIT(knote, SI_SUB_PSEUDO, SI_ORDER_ANY, knote_init, NULL)
982
983static struct knote *
984knote_alloc(void)
985{
986 return ((struct knote *)zalloc(knote_zone));
987}
988
989static void
990knote_free(struct knote *kn)
991{
992 zfree(knote_zone, kn);
993}