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