1 /* $OpenBSD: kern_event.c,v 1.23 2004/04/01 00:27:51 tedu Exp $ */
2
3 /*-
4 * Copyright (c) 1999,2000,2001 Jonathan Lemon <jlemon@FreeBSD.org>
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
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.
15 *
16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``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 THE AUTHOR 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
26 * SUCH DAMAGE.
27 *
28 * $FreeBSD: src/sys/kern/kern_event.c,v 1.22 2001/02/23 20:32:42 jlemon Exp $
29 */
30
31 #include <sys/param.h>
32 #include <sys/systm.h>
33 #include <sys/kernel.h>
34 #include <sys/proc.h>
35 #include <sys/malloc.h>
36 #include <sys/unistd.h>
37 #include <sys/file.h>
38 #include <sys/filedesc.h>
39 #include <sys/fcntl.h>
40 #include <sys/select.h>
41 #include <sys/queue.h>
42 #include <sys/event.h>
43 #include <sys/eventvar.h>
44 #include <sys/pool.h>
45 #include <sys/protosw.h>
46 #include <sys/socket.h>
47 #include <sys/socketvar.h>
48 #include <sys/stat.h>
49 #include <sys/uio.h>
50 #include <sys/mount.h>
51 #include <sys/poll.h>
52 #include <sys/syscallargs.h>
53
54 int kqueue_scan(struct file *fp, int maxevents,
55 struct kevent *ulistp, const struct timespec *timeout,
56 struct proc *p, int *retval);
57
58 int kqueue_read(struct file *fp, off_t *poff, struct uio *uio,
59 struct ucred *cred);
60 int kqueue_write(struct file *fp, off_t *poff, struct uio *uio,
61 struct ucred *cred);
62 int kqueue_ioctl(struct file *fp, u_long com, caddr_t data,
63 struct proc *p);
64 int kqueue_poll(struct file *fp, int events, struct proc *p);
65 int kqueue_kqfilter(struct file *fp, struct knote *kn);
66 int kqueue_stat(struct file *fp, struct stat *st, struct proc *p);
67 int kqueue_close(struct file *fp, struct proc *p);
68 void kqueue_wakeup(struct kqueue *kq);
69
70 struct fileops kqueueops = {
71 kqueue_read,
72 kqueue_write,
73 kqueue_ioctl,
74 kqueue_poll,
75 kqueue_kqfilter,
76 kqueue_stat,
77 kqueue_close
78 };
79
80 void knote_attach(struct knote *kn, struct filedesc *fdp);
81 void knote_drop(struct knote *kn, struct proc *p, struct filedesc *fdp);
82 void knote_enqueue(struct knote *kn);
83 void knote_dequeue(struct knote *kn);
84 #define knote_alloc() ((struct knote *)pool_get(&knote_pool, PR_WAITOK))
85 #define knote_free(kn) (pool_put(&knote_pool, kn))
86
87 void filt_kqdetach(struct knote *kn);
88 int filt_kqueue(struct knote *kn, long hint);
89 int filt_procattach(struct knote *kn);
90 void filt_procdetach(struct knote *kn);
91 int filt_proc(struct knote *kn, long hint);
92 int filt_fileattach(struct knote *kn);
93
94 struct filterops kqread_filtops =
95 { 1, NULL, filt_kqdetach, filt_kqueue };
96 struct filterops proc_filtops =
97 { 0, filt_procattach, filt_procdetach, filt_proc };
98 struct filterops file_filtops =
99 { 1, filt_fileattach, NULL, NULL };
100
101 struct pool knote_pool;
102 struct pool kqueue_pool;
103
104 #define KNOTE_ACTIVATE(kn) do { \
105 kn->kn_status |= KN_ACTIVE; \
106 if ((kn->kn_status & (KN_QUEUED | KN_DISABLED)) == 0) \
107 knote_enqueue(kn); \
108 } while(0)
109
110 #define KN_HASHSIZE 64 /* XXX should be tunable */
111 #define KN_HASH(val, mask) (((val) ^ (val >> 8)) & (mask))
112
113 extern struct filterops sig_filtops;
114 #ifdef notyet
115 extern struct filterops aio_filtops;
116 #endif
117
118 /*
119 * Table for for all system-defined filters.
120 */
121 struct filterops *sysfilt_ops[] = {
122 &file_filtops, /* EVFILT_READ */
123 &file_filtops, /* EVFILT_WRITE */
124 NULL, /*&aio_filtops,*/ /* EVFILT_AIO */
125 &file_filtops, /* EVFILT_VNODE */
126 &proc_filtops, /* EVFILT_PROC */
127 &sig_filtops, /* EVFILT_SIGNAL */
128 };
129
130 void kqueue_init(void);
131
132 void
kqueue_init(void)133 kqueue_init(void)
134 {
135
136 pool_init(&kqueue_pool, sizeof(struct kqueue), 0, 0, 0, "kqeuepl",
137 &pool_allocator_nointr);
138 pool_init(&knote_pool, sizeof(struct knote), 0, 0, 0, "knotepl",
139 &pool_allocator_nointr);
140 }
141
142 int
filt_fileattach(struct knote * kn)143 filt_fileattach(struct knote *kn)
144 {
145 struct file *fp = kn->kn_fp;
146
147 return ((*fp->f_ops->fo_kqfilter)(fp, kn));
148 }
149
150 int
kqueue_kqfilter(struct file * fp,struct knote * kn)151 kqueue_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
163 void
filt_kqdetach(struct knote * kn)164 filt_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*/
172 int
filt_kqueue(struct knote * kn,long hint)173 filt_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
181 int
filt_procattach(struct knote * kn)182 filt_procattach(struct knote *kn)
183 {
184 struct proc *p;
185
186 p = pfind(kn->kn_id);
187 if (p == NULL)
188 return (ESRCH);
189
190 /*
191 * Fail if it's not owned by you, or the last exec gave us
192 * setuid/setgid privs (unless you're root).
193 */
194 if ((p->p_cred->p_ruid != curproc->p_cred->p_ruid ||
195 (p->p_flag & P_SUGID)) && suser(curproc, 0) != 0)
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 return (0);
214 }
215
216 /*
217 * The knote may be attached to a different process, which may exit,
218 * leaving nothing for the knote to be attached to. So when the process
219 * exits, the knote is marked as DETACHED and also flagged as ONESHOT so
220 * it will be deleted when read out. However, as part of the knote deletion,
221 * this routine is called, so a check is needed to avoid actually performing
222 * a detach, because the original process does not exist any more.
223 */
224 void
filt_procdetach(struct knote * kn)225 filt_procdetach(struct knote *kn)
226 {
227 struct proc *p = kn->kn_ptr.p_proc;
228
229 if (kn->kn_status & KN_DETACHED)
230 return;
231
232 /* XXX locking? this might modify another process. */
233 SLIST_REMOVE(&p->p_klist, kn, knote, kn_selnext);
234 }
235
236 int
filt_proc(struct knote * kn,long hint)237 filt_proc(struct knote *kn, long hint)
238 {
239 u_int event;
240
241 /*
242 * mask off extra data
243 */
244 event = (u_int)hint & NOTE_PCTRLMASK;
245
246 /*
247 * if the user is interested in this event, record it.
248 */
249 if (kn->kn_sfflags & event)
250 kn->kn_fflags |= event;
251
252 /*
253 * process is gone, so flag the event as finished.
254 */
255 if (event == NOTE_EXIT) {
256 kn->kn_status |= KN_DETACHED;
257 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
258 return (1);
259 }
260
261 /*
262 * process forked, and user wants to track the new process,
263 * so attach a new knote to it, and immediately report an
264 * event with the parent's pid.
265 */
266 if ((event == NOTE_FORK) && (kn->kn_sfflags & NOTE_TRACK)) {
267 struct kevent kev;
268 int error;
269
270 /*
271 * register knote with new process.
272 */
273 kev.ident = hint & NOTE_PDATAMASK; /* pid */
274 kev.filter = kn->kn_filter;
275 kev.flags = kn->kn_flags | EV_ADD | EV_ENABLE | EV_FLAG1;
276 kev.fflags = kn->kn_sfflags;
277 kev.data = kn->kn_id; /* parent */
278 kev.udata = kn->kn_kevent.udata; /* preserve udata */
279 error = kqueue_register(kn->kn_kq, &kev, NULL);
280 if (error)
281 kn->kn_fflags |= NOTE_TRACKERR;
282 }
283
284 return (kn->kn_fflags != 0);
285 }
286
287 /*
288 * filt_seltrue:
289 *
290 * This filter "event" routine simulates seltrue().
291 */
292 int
filt_seltrue(struct knote * kn,long hint)293 filt_seltrue(struct knote *kn, long hint)
294 {
295
296 /*
297 * We don't know how much data can be read/written,
298 * but we know that it *can* be. This is about as
299 * good as select/poll does as well.
300 */
301 kn->kn_data = 0;
302 return (1);
303 }
304
305 int
sys_kqueue(struct proc * p,void * v,register_t * retval)306 sys_kqueue(struct proc *p, void *v, register_t *retval)
307 {
308 struct filedesc *fdp = p->p_fd;
309 struct kqueue *kq;
310 struct file *fp;
311 int fd, error;
312
313 error = falloc(p, &fp, &fd);
314 if (error)
315 return (error);
316 fp->f_flag = FREAD | FWRITE;
317 fp->f_type = DTYPE_KQUEUE;
318 fp->f_ops = &kqueueops;
319 kq = pool_get(&kqueue_pool, PR_WAITOK);
320 bzero(kq, sizeof(*kq));
321 TAILQ_INIT(&kq->kq_head);
322 fp->f_data = (caddr_t)kq;
323 *retval = fd;
324 if (fdp->fd_knlistsize < 0)
325 fdp->fd_knlistsize = 0; /* this process has a kq */
326 kq->kq_fdp = fdp;
327 FILE_SET_MATURE(fp);
328 return (0);
329 }
330
331 int
sys_kevent(struct proc * p,void * v,register_t * retval)332 sys_kevent(struct proc *p, void *v, register_t *retval)
333 {
334 struct filedesc* fdp = p->p_fd;
335 struct sys_kevent_args /* {
336 syscallarg(int) fd;
337 syscallarg(const struct kevent *) changelist;
338 syscallarg(int) nchanges;
339 syscallarg(struct kevent *) eventlist;
340 syscallarg(int) nevents;
341 syscallarg(const struct timespec *) timeout;
342 } */ *uap = v;
343 struct kevent *kevp;
344 struct kqueue *kq;
345 struct file *fp;
346 struct timespec ts;
347 int i, n, nerrors, error;
348
349 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL ||
350 (fp->f_type != DTYPE_KQUEUE))
351 return (EBADF);
352
353 FREF(fp);
354
355 if (SCARG(uap, timeout) != NULL) {
356 error = copyin(SCARG(uap, timeout), &ts, sizeof(ts));
357 if (error)
358 goto done;
359 SCARG(uap, timeout) = &ts;
360 }
361
362 kq = (struct kqueue *)fp->f_data;
363 nerrors = 0;
364
365 while (SCARG(uap, nchanges) > 0) {
366 n = SCARG(uap, nchanges) > KQ_NEVENTS
367 ? KQ_NEVENTS : SCARG(uap, nchanges);
368 error = copyin(SCARG(uap, changelist), kq->kq_kev,
369 n * sizeof(struct kevent));
370 if (error)
371 goto done;
372 for (i = 0; i < n; i++) {
373 kevp = &kq->kq_kev[i];
374 kevp->flags &= ~EV_SYSFLAGS;
375 error = kqueue_register(kq, kevp, p);
376 if (error) {
377 if (SCARG(uap, nevents) != 0) {
378 kevp->flags = EV_ERROR;
379 kevp->data = error;
380 (void) copyout((caddr_t)kevp,
381 (caddr_t)SCARG(uap, eventlist),
382 sizeof(*kevp));
383 SCARG(uap, eventlist)++;
384 SCARG(uap, nevents)--;
385 nerrors++;
386 } else {
387 goto done;
388 }
389 }
390 }
391 SCARG(uap, nchanges) -= n;
392 SCARG(uap, changelist) += n;
393 }
394 if (nerrors) {
395 *retval = nerrors;
396 error = 0;
397 goto done;
398 }
399
400 error = kqueue_scan(fp, SCARG(uap, nevents), SCARG(uap, eventlist),
401 SCARG(uap, timeout), p, &n);
402 *retval = n;
403 done:
404 FRELE(fp);
405 return (error);
406 }
407
408 int
kqueue_register(struct kqueue * kq,struct kevent * kev,struct proc * p)409 kqueue_register(struct kqueue *kq, struct kevent *kev, struct proc *p)
410 {
411 struct filedesc *fdp = kq->kq_fdp;
412 struct filterops *fops = NULL;
413 struct file *fp = NULL;
414 struct knote *kn = NULL;
415 int s, error = 0;
416
417 if (kev->filter < 0) {
418 if (kev->filter + EVFILT_SYSCOUNT < 0)
419 return (EINVAL);
420 fops = sysfilt_ops[~kev->filter]; /* to 0-base index */
421 }
422
423 if (fops == NULL) {
424 /*
425 * XXX
426 * filter attach routine is responsible for insuring that
427 * the identifier can be attached to it.
428 */
429 return (EINVAL);
430 }
431
432 if (fops->f_isfd) {
433 /* validate descriptor */
434 if ((fp = fd_getfile(fdp, kev->ident)) == NULL)
435 return (EBADF);
436 FREF(fp);
437 fp->f_count++;
438
439 if (kev->ident < fdp->fd_knlistsize) {
440 SLIST_FOREACH(kn, &fdp->fd_knlist[kev->ident], kn_link)
441 if (kq == kn->kn_kq &&
442 kev->filter == kn->kn_filter)
443 break;
444 }
445 } else {
446 if (fdp->fd_knhashmask != 0) {
447 struct klist *list;
448
449 list = &fdp->fd_knhash[
450 KN_HASH((u_long)kev->ident, fdp->fd_knhashmask)];
451 SLIST_FOREACH(kn, list, kn_link)
452 if (kev->ident == kn->kn_id &&
453 kq == kn->kn_kq &&
454 kev->filter == kn->kn_filter)
455 break;
456 }
457 }
458
459 if (kn == NULL && ((kev->flags & EV_ADD) == 0)) {
460 error = ENOENT;
461 goto done;
462 }
463
464 /*
465 * kn now contains the matching knote, or NULL if no match
466 */
467 if (kev->flags & EV_ADD) {
468
469 if (kn == NULL) {
470 kn = knote_alloc();
471 if (kn == NULL) {
472 error = ENOMEM;
473 goto done;
474 }
475 kn->kn_fp = fp;
476 kn->kn_kq = kq;
477 kn->kn_fop = fops;
478
479 /*
480 * apply reference count to knote structure, and
481 * do not release it at the end of this routine.
482 */
483 if (fp != NULL)
484 FRELE(fp);
485 fp = NULL;
486
487 kn->kn_sfflags = kev->fflags;
488 kn->kn_sdata = kev->data;
489 kev->fflags = 0;
490 kev->data = 0;
491 kn->kn_kevent = *kev;
492
493 knote_attach(kn, fdp);
494 if ((error = fops->f_attach(kn)) != 0) {
495 knote_drop(kn, p, fdp);
496 goto done;
497 }
498 } else {
499 /*
500 * The user may change some filter values after the
501 * initial EV_ADD, but doing so will not reset any
502 * filter which have already been triggered.
503 */
504 kn->kn_sfflags = kev->fflags;
505 kn->kn_sdata = kev->data;
506 kn->kn_kevent.udata = kev->udata;
507 }
508
509 s = splhigh();
510 if (kn->kn_fop->f_event(kn, 0))
511 KNOTE_ACTIVATE(kn);
512 splx(s);
513
514 } else if (kev->flags & EV_DELETE) {
515 kn->kn_fop->f_detach(kn);
516 knote_drop(kn, p, p->p_fd);
517 goto done;
518 }
519
520 if ((kev->flags & EV_DISABLE) &&
521 ((kn->kn_status & KN_DISABLED) == 0)) {
522 s = splhigh();
523 kn->kn_status |= KN_DISABLED;
524 splx(s);
525 }
526
527 if ((kev->flags & EV_ENABLE) && (kn->kn_status & KN_DISABLED)) {
528 s = splhigh();
529 kn->kn_status &= ~KN_DISABLED;
530 if ((kn->kn_status & KN_ACTIVE) &&
531 ((kn->kn_status & KN_QUEUED) == 0))
532 knote_enqueue(kn);
533 splx(s);
534 }
535
536 done:
537 if (fp != NULL)
538 closef(fp, p);
539 return (error);
540 }
541
542 int
kqueue_scan(struct file * fp,int maxevents,struct kevent * ulistp,const struct timespec * tsp,struct proc * p,int * retval)543 kqueue_scan(struct file *fp, int maxevents, struct kevent *ulistp,
544 const struct timespec *tsp, struct proc *p, int *retval)
545 {
546 struct kqueue *kq = (struct kqueue *)fp->f_data;
547 struct kevent *kevp;
548 struct timeval atv;
549 struct knote *kn, marker;
550 int s, count, timeout, nkev = 0, error = 0;
551
552 count = maxevents;
553 if (count == 0)
554 goto done;
555
556 if (tsp != NULL) {
557 TIMESPEC_TO_TIMEVAL(&atv, tsp);
558 if (tsp->tv_sec == 0 && tsp->tv_nsec == 0) {
559 /* No timeout, just poll */
560 timeout = -1;
561 goto start;
562 }
563 if (itimerfix(&atv)) {
564 error = EINVAL;
565 goto done;
566 }
567
568 s = splclock();
569 timeradd(&atv, &time, &atv);
570 timeout = hzto(&atv);
571 splx(s);
572 } else {
573 atv.tv_sec = 0;
574 atv.tv_usec = 0;
575 timeout = 0;
576 }
577 goto start;
578
579 retry:
580 if (atv.tv_sec || atv.tv_usec) {
581 timeout = hzto(&atv);
582 if (timeout <= 0)
583 goto done;
584 }
585
586 start:
587 kevp = kq->kq_kev;
588 s = splhigh();
589 if (kq->kq_count == 0) {
590 if (timeout < 0) {
591 error = EWOULDBLOCK;
592 } else {
593 kq->kq_state |= KQ_SLEEP;
594 error = tsleep(kq, PSOCK | PCATCH, "kqread", timeout);
595 }
596 splx(s);
597 if (error == 0)
598 goto retry;
599 /* don't restart after signals... */
600 if (error == ERESTART)
601 error = EINTR;
602 else if (error == EWOULDBLOCK)
603 error = 0;
604 goto done;
605 }
606
607 TAILQ_INSERT_TAIL(&kq->kq_head, &marker, kn_tqe);
608 while (count) {
609 kn = TAILQ_FIRST(&kq->kq_head);
610 TAILQ_REMOVE(&kq->kq_head, kn, kn_tqe);
611 if (kn == &marker) {
612 splx(s);
613 if (count == maxevents)
614 goto retry;
615 goto done;
616 }
617 if (kn->kn_status & KN_DISABLED) {
618 kn->kn_status &= ~KN_QUEUED;
619 kq->kq_count--;
620 continue;
621 }
622 if ((kn->kn_flags & EV_ONESHOT) == 0 &&
623 kn->kn_fop->f_event(kn, 0) == 0) {
624 kn->kn_status &= ~(KN_QUEUED | KN_ACTIVE);
625 kq->kq_count--;
626 continue;
627 }
628 *kevp = kn->kn_kevent;
629 kevp++;
630 nkev++;
631 if (kn->kn_flags & EV_ONESHOT) {
632 kn->kn_status &= ~KN_QUEUED;
633 kq->kq_count--;
634 splx(s);
635 kn->kn_fop->f_detach(kn);
636 knote_drop(kn, p, p->p_fd);
637 s = splhigh();
638 } else if (kn->kn_flags & EV_CLEAR) {
639 kn->kn_data = 0;
640 kn->kn_fflags = 0;
641 kn->kn_status &= ~(KN_QUEUED | KN_ACTIVE);
642 kq->kq_count--;
643 } else {
644 TAILQ_INSERT_TAIL(&kq->kq_head, kn, kn_tqe);
645 }
646 count--;
647 if (nkev == KQ_NEVENTS) {
648 splx(s);
649 error = copyout((caddr_t)&kq->kq_kev, (caddr_t)ulistp,
650 sizeof(struct kevent) * nkev);
651 ulistp += nkev;
652 nkev = 0;
653 kevp = kq->kq_kev;
654 s = splhigh();
655 if (error)
656 break;
657 }
658 }
659 TAILQ_REMOVE(&kq->kq_head, &marker, kn_tqe);
660 splx(s);
661 done:
662 if (nkev != 0)
663 error = copyout((caddr_t)&kq->kq_kev, (caddr_t)ulistp,
664 sizeof(struct kevent) * nkev);
665 *retval = maxevents - count;
666 return (error);
667 }
668
669 /*
670 * XXX
671 * This could be expanded to call kqueue_scan, if desired.
672 */
673 /*ARGSUSED*/
674 int
kqueue_read(struct file * fp,off_t * poff,struct uio * uio,struct ucred * cred)675 kqueue_read(struct file *fp, off_t *poff, struct uio *uio, struct ucred *cred)
676 {
677 return (ENXIO);
678 }
679
680 /*ARGSUSED*/
681 int
kqueue_write(struct file * fp,off_t * poff,struct uio * uio,struct ucred * cred)682 kqueue_write(struct file *fp, off_t *poff, struct uio *uio, struct ucred *cred)
683
684 {
685 return (ENXIO);
686 }
687
688 /*ARGSUSED*/
689 int
kqueue_ioctl(struct file * fp,u_long com,caddr_t data,struct proc * p)690 kqueue_ioctl(struct file *fp, u_long com, caddr_t data, struct proc *p)
691 {
692 return (ENOTTY);
693 }
694
695 /*ARGSUSED*/
696 int
kqueue_poll(struct file * fp,int events,struct proc * p)697 kqueue_poll(struct file *fp, int events, struct proc *p)
698 {
699 struct kqueue *kq = (struct kqueue *)fp->f_data;
700 int revents = 0;
701 int s = splhigh();
702
703 if (events & (POLLIN | POLLRDNORM)) {
704 if (kq->kq_count) {
705 revents |= events & (POLLIN | POLLRDNORM);
706 } else {
707 selrecord(p, &kq->kq_sel);
708 kq->kq_state |= KQ_SEL;
709 }
710 }
711 splx(s);
712 return (revents);
713 }
714
715 /*ARGSUSED*/
716 int
kqueue_stat(struct file * fp,struct stat * st,struct proc * p)717 kqueue_stat(struct file *fp, struct stat *st, struct proc *p)
718 {
719 struct kqueue *kq = (struct kqueue *)fp->f_data;
720
721 bzero((void *)st, sizeof(*st));
722 st->st_size = kq->kq_count;
723 st->st_blksize = sizeof(struct kevent);
724 st->st_mode = S_IFIFO;
725 return (0);
726 }
727
728 /*ARGSUSED*/
729 int
kqueue_close(struct file * fp,struct proc * p)730 kqueue_close(struct file *fp, struct proc *p)
731 {
732 struct kqueue *kq = (struct kqueue *)fp->f_data;
733 struct filedesc *fdp = p->p_fd;
734 struct knote **knp, *kn, *kn0;
735 int i;
736
737 for (i = 0; i < fdp->fd_knlistsize; i++) {
738 knp = &SLIST_FIRST(&fdp->fd_knlist[i]);
739 kn = *knp;
740 while (kn != NULL) {
741 kn0 = SLIST_NEXT(kn, kn_link);
742 if (kq == kn->kn_kq) {
743 FREF(kn->kn_fp);
744 kn->kn_fop->f_detach(kn);
745 closef(kn->kn_fp, p);
746 knote_free(kn);
747 *knp = kn0;
748 } else {
749 knp = &SLIST_NEXT(kn, kn_link);
750 }
751 kn = kn0;
752 }
753 }
754 if (fdp->fd_knhashmask != 0) {
755 for (i = 0; i < fdp->fd_knhashmask + 1; i++) {
756 knp = &SLIST_FIRST(&fdp->fd_knhash[i]);
757 kn = *knp;
758 while (kn != NULL) {
759 kn0 = SLIST_NEXT(kn, kn_link);
760 if (kq == kn->kn_kq) {
761 kn->kn_fop->f_detach(kn);
762 /* XXX non-fd release of kn->kn_ptr */
763 knote_free(kn);
764 *knp = kn0;
765 } else {
766 knp = &SLIST_NEXT(kn, kn_link);
767 }
768 kn = kn0;
769 }
770 }
771 }
772 pool_put(&kqueue_pool, kq);
773 fp->f_data = NULL;
774
775 return (0);
776 }
777
778 void
kqueue_wakeup(struct kqueue * kq)779 kqueue_wakeup(struct kqueue *kq)
780 {
781
782 if (kq->kq_state & KQ_SLEEP) {
783 kq->kq_state &= ~KQ_SLEEP;
784 wakeup(kq);
785 }
786 if (kq->kq_state & KQ_SEL) {
787 kq->kq_state &= ~KQ_SEL;
788 selwakeup(&kq->kq_sel);
789 }
790 KNOTE(&kq->kq_sel.si_note, 0);
791 }
792
793 /*
794 * walk down a list of knotes, activating them if their event has triggered.
795 */
796 void
knote(struct klist * list,long hint)797 knote(struct klist *list, long hint)
798 {
799 struct knote *kn;
800
801 SLIST_FOREACH(kn, list, kn_selnext)
802 if (kn->kn_fop->f_event(kn, hint))
803 KNOTE_ACTIVATE(kn);
804 }
805
806 /*
807 * remove all knotes from a specified klist
808 */
809 void
knote_remove(struct proc * p,struct klist * list)810 knote_remove(struct proc *p, struct klist *list)
811 {
812 struct knote *kn;
813
814 while ((kn = SLIST_FIRST(list)) != NULL) {
815 kn->kn_fop->f_detach(kn);
816 knote_drop(kn, p, p->p_fd);
817 }
818 }
819
820 /*
821 * remove all knotes referencing a specified fd
822 */
823 void
knote_fdclose(struct proc * p,int fd)824 knote_fdclose(struct proc *p, int fd)
825 {
826 struct filedesc *fdp = p->p_fd;
827 struct klist *list = &fdp->fd_knlist[fd];
828
829 knote_remove(p, list);
830 }
831
832 void
knote_attach(struct knote * kn,struct filedesc * fdp)833 knote_attach(struct knote *kn, struct filedesc *fdp)
834 {
835 struct klist *list;
836 int size;
837
838 if (! kn->kn_fop->f_isfd) {
839 if (fdp->fd_knhashmask == 0)
840 fdp->fd_knhash = hashinit(KN_HASHSIZE, M_TEMP,
841 M_WAITOK, &fdp->fd_knhashmask);
842 list = &fdp->fd_knhash[KN_HASH(kn->kn_id, fdp->fd_knhashmask)];
843 goto done;
844 }
845
846 if (fdp->fd_knlistsize <= kn->kn_id) {
847 size = fdp->fd_knlistsize;
848 while (size <= kn->kn_id)
849 size += KQEXTENT;
850 list = malloc(size * sizeof(struct klist *), M_TEMP, M_WAITOK);
851 bcopy((caddr_t)fdp->fd_knlist, (caddr_t)list,
852 fdp->fd_knlistsize * sizeof(struct klist *));
853 bzero((caddr_t)list +
854 fdp->fd_knlistsize * sizeof(struct klist *),
855 (size - fdp->fd_knlistsize) * sizeof(struct klist *));
856 if (fdp->fd_knlist != NULL)
857 free(fdp->fd_knlist, M_TEMP);
858 fdp->fd_knlistsize = size;
859 fdp->fd_knlist = list;
860 }
861 list = &fdp->fd_knlist[kn->kn_id];
862 done:
863 SLIST_INSERT_HEAD(list, kn, kn_link);
864 kn->kn_status = 0;
865 }
866
867 /*
868 * should be called at spl == 0, since we don't want to hold spl
869 * while calling closef and free.
870 */
871 void
knote_drop(struct knote * kn,struct proc * p,struct filedesc * fdp)872 knote_drop(struct knote *kn, struct proc *p, struct filedesc *fdp)
873 {
874 struct klist *list;
875
876 if (kn->kn_fop->f_isfd)
877 list = &fdp->fd_knlist[kn->kn_id];
878 else
879 list = &fdp->fd_knhash[KN_HASH(kn->kn_id, fdp->fd_knhashmask)];
880
881 SLIST_REMOVE(list, kn, knote, kn_link);
882 if (kn->kn_status & KN_QUEUED)
883 knote_dequeue(kn);
884 if (kn->kn_fop->f_isfd) {
885 FREF(kn->kn_fp);
886 closef(kn->kn_fp, p);
887 }
888 knote_free(kn);
889 }
890
891
892 void
knote_enqueue(struct knote * kn)893 knote_enqueue(struct knote *kn)
894 {
895 struct kqueue *kq = kn->kn_kq;
896 int s = splhigh();
897
898 KASSERT((kn->kn_status & KN_QUEUED) == 0);
899
900 TAILQ_INSERT_TAIL(&kq->kq_head, kn, kn_tqe);
901 kn->kn_status |= KN_QUEUED;
902 kq->kq_count++;
903 splx(s);
904 kqueue_wakeup(kq);
905 }
906
907 void
knote_dequeue(struct knote * kn)908 knote_dequeue(struct knote *kn)
909 {
910 struct kqueue *kq = kn->kn_kq;
911 int s = splhigh();
912
913 KASSERT(kn->kn_status & KN_QUEUED);
914
915 TAILQ_REMOVE(&kq->kq_head, kn, kn_tqe);
916 kn->kn_status &= ~KN_QUEUED;
917 kq->kq_count--;
918 splx(s);
919 }
920
921 void
klist_invalidate(struct klist * list)922 klist_invalidate(struct klist *list)
923 {
924 struct knote *kn;
925
926 SLIST_FOREACH(kn, list, kn_selnext) {
927 kn->kn_status |= KN_DETACHED;
928 kn->kn_flags |= EV_EOF | EV_ONESHOT;
929 }
930 }
931