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_scan(struct file *fp, int maxevents, 61 struct kevent *ulistp, const struct timespec *timeout, 62 struct proc *p); 63 static int kqueue_read(struct file *fp, struct uio *uio, 64 struct ucred *cred, int flags, struct proc *p); 65 static int kqueue_write(struct file *fp, struct uio *uio, 66 struct ucred *cred, int flags, struct proc *p); 67 static int kqueue_ioctl(struct file *fp, u_long com, caddr_t data, 68 struct proc *p); 69 static int kqueue_poll(struct file *fp, int events, struct ucred *cred, 70 struct proc *p); 71 static int kqueue_stat(struct file *fp, struct stat *st, struct proc *p); 72 static int kqueue_close(struct file *fp, struct proc *p); 73 static void kqueue_wakeup(struct kqueue *kq); 74 75 static void knote_attach(struct knote *kn, struct filedesc *fdp); 76 static void knote_drop(struct knote *kn, struct proc *p); 77 static void knote_enqueue(struct knote *kn); 78 static void knote_dequeue(struct knote *kn); 79 static void knote_init(void); 80 static struct knote *knote_alloc(void); 81 static void knote_free(struct knote *kn); 82 83 static vm_zone_t knote_zone; 84 85 #define KNOTE_ACTIVATE(kn) do { \ 86 kn->kn_status |= KN_ACTIVE; \ 87 if ((kn->kn_status & (KN_QUEUED | KN_DISABLED)) == 0) \ 88 knote_enqueue(kn); \ 89 } while(0) 90 91 #define KN_HASHSIZE 64 /* XXX should be tunable */ 92 #define KN_HASH(val, mask) (((val) ^ (val >> 8)) & (mask)) 93 94 static struct fileops kqueueops = { 95 kqueue_read, 96 kqueue_write, 97 kqueue_ioctl, 98 kqueue_poll, 99 kqueue_stat, 100 kqueue_close 101 }; 102 103 extern struct filterops so_rwfiltops[]; 104 extern struct filterops fifo_rwfiltops[]; 105 extern struct filterops pipe_rwfiltops[]; 106 extern struct filterops vn_rwfiltops[]; 107 108 static struct filterops kq_rwfiltops[] = { 109 { 1, filt_kqattach, filt_kqdetach, filt_kqueue }, 110 { 1, filt_nullattach, NULL, NULL }, 111 }; 112 113 extern struct filterops aio_filtops; 114 extern struct filterops sig_filtops; 115 extern struct filterops vn_filtops; 116 117 static struct filterops rwtype_filtops = 118 { 1, filt_rwtypattach, NULL, NULL }; 119 static struct filterops proc_filtops = 120 { 0, filt_procattach, filt_procdetach, filt_proc }; 121 122 /* 123 * XXX 124 * These must match the order of defines in <sys/file.h> 125 */ 126 static struct filterops *rwtypfilt_sw[] = { 127 NULL, /* 0 */ 128 vn_rwfiltops, /* DTYPE_VNODE */ 129 so_rwfiltops, /* DTYPE_SOCKET */ 130 pipe_rwfiltops, /* DTYPE_PIPE */ 131 fifo_rwfiltops, /* DTYPE_FIFO */ 132 kq_rwfiltops, /* DTYPE_KQUEUE */ 133 }; 134 135 /* 136 * table for for all system-defined filters. 137 */ 138 static struct filterops *sysfilt_ops[] = { 139 &rwtype_filtops, /* EVFILT_READ */ 140 &rwtype_filtops, /* EVFILT_WRITE */ 141 &aio_filtops, /* EVFILT_AIO */ 142 &vn_filtops, /* EVFILT_VNODE */ 143 &proc_filtops, /* EVFILT_PROC */ 144 &sig_filtops, /* EVFILT_SIGNAL */ 145 }; 146 147 static int 148 filt_nullattach(struct knote *kn) 149 { 150 return (ENXIO); 151 } 152 153 /* 154 * file-type specific attach routine for read/write filters 155 */ 156 static int 157 filt_rwtypattach(struct knote *kn) 158 { 159 struct filterops *fops; 160 161 fops = rwtypfilt_sw[kn->kn_fp->f_type]; 162 if (fops == NULL) 163 return (EINVAL); 164 kn->kn_fop = &fops[~kn->kn_filter]; /* convert to 0-base index */ 165 return (kn->kn_fop->f_attach(kn)); 166 } 167 168 static int 169 filt_kqattach(struct knote *kn) 170 { 171 struct kqueue *kq = (struct kqueue *)kn->kn_fp->f_data; 172 173 SLIST_INSERT_HEAD(&kq->kq_sel.si_note, kn, kn_selnext); 174 return (0); 175 } 176 177 static void 178 filt_kqdetach(struct knote *kn) 179 { 180 struct kqueue *kq = (struct kqueue *)kn->kn_fp->f_data; 181 182 SLIST_REMOVE(&kq->kq_sel.si_note, kn, knote, kn_selnext); 183 } 184 185 /*ARGSUSED*/ 186 static int 187 filt_kqueue(struct knote *kn, long hint) 188 { 189 struct kqueue *kq = (struct kqueue *)kn->kn_fp->f_data; 190 191 kn->kn_data = kq->kq_count; 192 return (kn->kn_data > 0); 193 } 194 195 static int 196 filt_procattach(struct knote *kn) 197 { 198 struct proc *p; 199 200 p = pfind(kn->kn_id); 201 if (p == NULL) 202 return (ESRCH); 203 if (p_can(curproc, p, P_CAN_SEE, NULL)) 204 return (EACCES); 205 206 kn->kn_ptr.p_proc = p; 207 kn->kn_flags |= EV_CLEAR; /* automatically set */ 208 209 /* 210 * internal flag indicating registration done by kernel 211 */ 212 if (kn->kn_flags & EV_FLAG1) { 213 kn->kn_data = kn->kn_sdata; /* ppid */ 214 kn->kn_fflags = NOTE_CHILD; 215 kn->kn_flags &= ~EV_FLAG1; 216 } 217 218 /* XXX lock the proc here while adding to the list? */ 219 SLIST_INSERT_HEAD(&p->p_klist, kn, kn_selnext); 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 */ 232 static void 233 filt_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 } 243 244 static int 245 filt_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 } 294 295 int 296 kqueue(struct proc *p, struct kqueue_args *uap) 297 { 298 struct filedesc *fdp = p->p_fd; 299 struct kqueue *kq; 300 struct file *fp; 301 int fd, error; 302 303 error = falloc(p, &fp, &fd); 304 if (error) 305 return (error); 306 fp->f_flag = FREAD | FWRITE; 307 fp->f_type = DTYPE_KQUEUE; 308 fp->f_ops = &kqueueops; 309 kq = malloc(sizeof(struct kqueue), M_TEMP, M_WAITOK); 310 bzero(kq, sizeof(*kq)); 311 TAILQ_INIT(&kq->kq_head); 312 fp->f_data = (caddr_t)kq; 313 p->p_retval[0] = fd; 314 if (fdp->fd_knlistsize < 0) 315 fdp->fd_knlistsize = 0; /* this process has a kq */ 316 kq->kq_fdp = fdp; 317 return (error); 318 } 319 320 #ifndef _SYS_SYSPROTO_H_ 321 struct kevent_args { 322 int fd; 323 const struct kevent *changelist; 324 int nchanges; 325 struct kevent *eventlist; 326 int nevents; 327 const struct timespec *timeout; 328 }; 329 #endif 330 int 331 kevent(struct proc *p, struct kevent_args *uap) 332 { 333 struct filedesc* fdp = p->p_fd; 334 struct kevent *kevp; 335 struct kqueue *kq; 336 struct file *fp = NULL; 337 struct timespec ts; 338 int i, n, nerrors, error; 339 340 if (((u_int)uap->fd) >= fdp->fd_nfiles || 341 (fp = fdp->fd_ofiles[uap->fd]) == NULL || 342 (fp->f_type != DTYPE_KQUEUE)) 343 return (EBADF); 344 345 fhold(fp); 346 347 if (uap->timeout != NULL) { 348 error = copyin(uap->timeout, &ts, sizeof(ts)); 349 if (error) 350 goto done; 351 uap->timeout = &ts; 352 } 353 354 kq = (struct kqueue *)fp->f_data; 355 nerrors = 0; 356 357 while (uap->nchanges > 0) { 358 n = uap->nchanges > KQ_NEVENTS ? KQ_NEVENTS : uap->nchanges; 359 error = copyin(uap->changelist, kq->kq_kev, 360 n * sizeof(struct kevent)); 361 if (error) 362 goto done; 363 for (i = 0; i < n; i++) { 364 kevp = &kq->kq_kev[i]; 365 kevp->flags &= ~EV_SYSFLAGS; 366 error = kqueue_register(kq, kevp, p); 367 if (error) { 368 if (uap->nevents != 0) { 369 kevp->flags = EV_ERROR; 370 kevp->data = error; 371 (void) copyout((caddr_t)kevp, 372 (caddr_t)uap->eventlist, 373 sizeof(*kevp)); 374 uap->eventlist++; 375 uap->nevents--; 376 nerrors++; 377 } else { 378 goto done; 379 } 380 } 381 } 382 uap->nchanges -= n; 383 uap->changelist += n; 384 } 385 if (nerrors) { 386 p->p_retval[0] = nerrors; 387 error = 0; 388 goto done; 389 } 390 391 error = kqueue_scan(fp, uap->nevents, uap->eventlist, uap->timeout, p); 392 done: 393 if (fp != NULL) 394 fdrop(fp, p); 395 return (error); 396 } 397 398 int 399 kqueue_register(struct kqueue *kq, struct kevent *kev, struct proc *p) 400 { 401 struct filedesc *fdp = kq->kq_fdp; 402 struct filterops *fops; 403 struct file *fp = NULL; 404 struct knote *kn = NULL; 405 int s, error = 0; 406 407 if (kev->filter < 0) { 408 if (kev->filter + EVFILT_SYSCOUNT < 0) 409 return (EINVAL); 410 fops = sysfilt_ops[~kev->filter]; /* to 0-base index */ 411 } else { 412 /* 413 * XXX 414 * filter attach routine is responsible for insuring that 415 * the identifier can be attached to it. 416 */ 417 printf("unknown filter: %d\n", kev->filter); 418 return (EINVAL); 419 } 420 421 if (fops->f_isfd) { 422 /* validate descriptor */ 423 if ((u_int)kev->ident >= fdp->fd_nfiles || 424 (fp = fdp->fd_ofiles[kev->ident]) == NULL) 425 return (EBADF); 426 fhold(fp); 427 428 if (kev->ident < fdp->fd_knlistsize) { 429 SLIST_FOREACH(kn, &fdp->fd_knlist[kev->ident], kn_link) 430 if (kq == kn->kn_kq && 431 kev->filter == kn->kn_filter) 432 break; 433 } 434 } else { 435 if (fdp->fd_knhashmask != 0) { 436 struct klist *list; 437 438 list = &fdp->fd_knhash[ 439 KN_HASH((u_long)kev->ident, fdp->fd_knhashmask)]; 440 SLIST_FOREACH(kn, list, kn_link) 441 if (kev->ident == kn->kn_id && 442 kq == kn->kn_kq && 443 kev->filter == kn->kn_filter) 444 break; 445 } 446 } 447 448 if (kn == NULL && ((kev->flags & EV_ADD) == 0)) { 449 error = ENOENT; 450 goto done; 451 } 452 453 /* 454 * kn now contains the matching knote, or NULL if no match 455 */ 456 if (kev->flags & EV_ADD) { 457 458 if (kn == NULL) { 459 kn = knote_alloc(); 460 if (kn == NULL) { 461 error = ENOMEM; 462 goto done; 463 } 464 kn->kn_fp = fp; 465 kn->kn_kq = kq; 466 kn->kn_fop = fops; 467 468 /* 469 * apply reference count to knode structure, so 470 * do not release it at the end of this routine. 471 */ 472 fp = NULL; 473 474 kn->kn_sfflags = kev->fflags; 475 kn->kn_sdata = kev->data; 476 kev->fflags = 0; 477 kev->data = 0; 478 kn->kn_kevent = *kev; 479 480 knote_attach(kn, fdp); 481 if ((error = fops->f_attach(kn)) != 0) { 482 knote_drop(kn, p); 483 goto done; 484 } 485 } else { 486 /* 487 * The user may change some filter values after the 488 * initial EV_ADD, but doing so will not reset any 489 * filter which have already been triggered. 490 */ 491 kn->kn_sfflags = kev->fflags; 492 kn->kn_sdata = kev->data; 493 kn->kn_kevent.udata = kev->udata; 494 } 495 496 s = splhigh(); 497 if (kn->kn_fop->f_event(kn, 0)) 498 KNOTE_ACTIVATE(kn); 499 splx(s); 500 501 } else if (kev->flags & EV_DELETE) { 502 kn->kn_fop->f_detach(kn); 503 knote_drop(kn, p); 504 goto done; 505 } 506 507 if ((kev->flags & EV_DISABLE) && 508 ((kn->kn_status & KN_DISABLED) == 0)) { 509 s = splhigh(); 510 kn->kn_status |= KN_DISABLED; 511 splx(s); 512 } 513 514 if ((kev->flags & EV_ENABLE) && (kn->kn_status & KN_DISABLED)) { 515 s = splhigh(); 516 kn->kn_status &= ~KN_DISABLED; 517 if ((kn->kn_status & KN_ACTIVE) && 518 ((kn->kn_status & KN_QUEUED) == 0)) 519 knote_enqueue(kn); 520 splx(s); 521 } 522 523 done: 524 if (fp != NULL) 525 fdrop(fp, p); 526 return (error); 527 } 528 529 static int 530 kqueue_scan(struct file *fp, int maxevents, struct kevent *ulistp, 531 const struct timespec *tsp, struct proc *p) 532 { 533 struct kqueue *kq = (struct kqueue *)fp->f_data; 534 struct kevent *kevp; 535 struct timeval atv, rtv, ttv; 536 struct knote *kn, marker; 537 int s, count, timeout, nkev = 0, error = 0; 538 539 count = maxevents; 540 if (count == 0) 541 goto done; 542 543 if (tsp != NULL) { 544 TIMESPEC_TO_TIMEVAL(&atv, tsp); 545 if (itimerfix(&atv)) { 546 error = EINVAL; 547 goto done; 548 } 549 if (tsp->tv_sec == 0 && tsp->tv_nsec == 0) 550 timeout = -1; 551 else 552 timeout = atv.tv_sec > 24 * 60 * 60 ? 553 24 * 60 * 60 * hz : tvtohz(&atv); 554 getmicrouptime(&rtv); 555 timevaladd(&atv, &rtv); 556 } else { 557 atv.tv_sec = 0; 558 atv.tv_usec = 0; 559 timeout = 0; 560 } 561 goto start; 562 563 retry: 564 if (atv.tv_sec || atv.tv_usec) { 565 getmicrouptime(&rtv); 566 if (timevalcmp(&rtv, &atv, >=)) 567 goto done; 568 ttv = atv; 569 timevalsub(&ttv, &rtv); 570 timeout = ttv.tv_sec > 24 * 60 * 60 ? 571 24 * 60 * 60 * hz : tvtohz(&ttv); 572 } 573 574 start: 575 kevp = kq->kq_kev; 576 s = splhigh(); 577 if (kq->kq_count == 0) { 578 if (timeout < 0) { 579 error = EWOULDBLOCK; 580 } else { 581 kq->kq_state |= KQ_SLEEP; 582 error = tsleep(kq, PSOCK | PCATCH, "kqread", timeout); 583 } 584 splx(s); 585 if (error == 0) 586 goto retry; 587 /* don't restart after signals... */ 588 if (error == ERESTART) 589 error = EINTR; 590 else if (error == EWOULDBLOCK) 591 error = 0; 592 goto done; 593 } 594 595 TAILQ_INSERT_TAIL(&kq->kq_head, &marker, kn_tqe); 596 while (count) { 597 kn = TAILQ_FIRST(&kq->kq_head); 598 TAILQ_REMOVE(&kq->kq_head, kn, kn_tqe); 599 if (kn == &marker) { 600 splx(s); 601 if (count == maxevents) 602 goto retry; 603 goto done; 604 } 605 if (kn->kn_status & KN_DISABLED) { 606 kn->kn_status &= ~KN_QUEUED; 607 kq->kq_count--; 608 continue; 609 } 610 if ((kn->kn_flags & EV_ONESHOT) == 0 && 611 kn->kn_fop->f_event(kn, 0) == 0) { 612 kn->kn_status &= ~(KN_QUEUED | KN_ACTIVE); 613 kq->kq_count--; 614 continue; 615 } 616 *kevp = kn->kn_kevent; 617 kevp++; 618 nkev++; 619 if (kn->kn_flags & EV_ONESHOT) { 620 kn->kn_status &= ~KN_QUEUED; 621 kq->kq_count--; 622 splx(s); 623 kn->kn_fop->f_detach(kn); 624 knote_drop(kn, p); 625 s = splhigh(); 626 } else if (kn->kn_flags & EV_CLEAR) { 627 kn->kn_data = 0; 628 kn->kn_fflags = 0; 629 kn->kn_status &= ~(KN_QUEUED | KN_ACTIVE); 630 kq->kq_count--; 631 } else { 632 TAILQ_INSERT_TAIL(&kq->kq_head, kn, kn_tqe); 633 } 634 count--; 635 if (nkev == KQ_NEVENTS) { 636 splx(s); 637 error = copyout((caddr_t)&kq->kq_kev, (caddr_t)ulistp, 638 sizeof(struct kevent) * nkev); 639 ulistp += nkev; 640 nkev = 0; 641 kevp = kq->kq_kev; 642 s = splhigh(); 643 if (error) 644 break; 645 } 646 } 647 TAILQ_REMOVE(&kq->kq_head, &marker, kn_tqe); 648 splx(s); 649 done: 650 if (nkev != 0) 651 error = copyout((caddr_t)&kq->kq_kev, (caddr_t)ulistp, 652 sizeof(struct kevent) * nkev); 653 p->p_retval[0] = maxevents - count; 654 return (error); 655 } 656 657 /* 658 * XXX 659 * This could be expanded to call kqueue_scan, if desired. 660 */ 661 /*ARGSUSED*/ 662 static int 663 kqueue_read(struct file *fp, struct uio *uio, struct ucred *cred, 664 int flags, struct proc *p) 665 { 666 return (ENXIO); 667 } 668 669 /*ARGSUSED*/ 670 static int 671 kqueue_write(struct file *fp, struct uio *uio, struct ucred *cred, 672 int flags, struct proc *p) 673 { 674 return (ENXIO); 675 } 676 677 /*ARGSUSED*/ 678 static int 679 kqueue_ioctl(struct file *fp, u_long com, caddr_t data, struct proc *p) 680 { 681 return (ENOTTY); 682 } 683 684 /*ARGSUSED*/ 685 static int 686 kqueue_poll(struct file *fp, int events, struct ucred *cred, struct proc *p) 687 { 688 struct kqueue *kq = (struct kqueue *)fp->f_data; 689 int revents = 0; 690 int s = splnet(); 691 692 if (events & (POLLIN | POLLRDNORM)) { 693 if (kq->kq_count) { 694 revents |= events & (POLLIN | POLLRDNORM); 695 } else { 696 selrecord(p, &kq->kq_sel); 697 kq->kq_state |= KQ_SEL; 698 } 699 } 700 splx(s); 701 return (revents); 702 } 703 704 /*ARGSUSED*/ 705 static int 706 kqueue_stat(struct file *fp, struct stat *st, struct proc *p) 707 { 708 struct kqueue *kq = (struct kqueue *)fp->f_data; 709 710 bzero((void *)st, sizeof(*st)); 711 st->st_size = kq->kq_count; 712 st->st_blksize = sizeof(struct kevent); 713 st->st_mode = S_IFIFO; 714 return (0); 715 } 716 717 /*ARGSUSED*/ 718 static int 719 kqueue_close(struct file *fp, struct proc *p) 720 { 721 struct kqueue *kq = (struct kqueue *)fp->f_data; 722 struct filedesc *fdp = p->p_fd; 723 struct knote **knp, *kn, *kn0; 724 int i; 725 726 for (i = 0; i < fdp->fd_knlistsize; i++) { 727 knp = &SLIST_FIRST(&fdp->fd_knlist[i]); 728 kn = *knp; 729 while (kn != NULL) { 730 kn0 = SLIST_NEXT(kn, kn_link); 731 if (kq == kn->kn_kq) { 732 kn->kn_fop->f_detach(kn); 733 fdrop(kn->kn_fp, p); 734 knote_free(kn); 735 *knp = kn0; 736 } else { 737 knp = &SLIST_NEXT(kn, kn_link); 738 } 739 kn = kn0; 740 } 741 } 742 if (fdp->fd_knhashmask != 0) { 743 for (i = 0; i < fdp->fd_knhashmask + 1; i++) { 744 knp = &SLIST_FIRST(&fdp->fd_knhash[i]); 745 kn = *knp; 746 while (kn != NULL) { 747 kn0 = SLIST_NEXT(kn, kn_link); 748 if (kq == kn->kn_kq) { 749 kn->kn_fop->f_detach(kn); 750 /* XXX non-fd release of kn->kn_ptr */ 751 knote_free(kn); 752 *knp = kn0; 753 } else { 754 knp = &SLIST_NEXT(kn, kn_link); 755 } 756 kn = kn0; 757 } 758 } 759 } 760 free(kq, M_TEMP); 761 fp->f_data = NULL; 762 763 return (0); 764 } 765 766 static void 767 kqueue_wakeup(struct kqueue *kq) 768 { 769 770 if (kq->kq_state & KQ_SLEEP) { 771 kq->kq_state &= ~KQ_SLEEP; 772 wakeup(kq); 773 } 774 if (kq->kq_state & KQ_SEL) { 775 kq->kq_state &= ~KQ_SEL; 776 selwakeup(&kq->kq_sel); 777 } 778 KNOTE(&kq->kq_sel.si_note, 0); 779 } 780 781 /* 782 * walk down a list of knotes, activating them if their event has triggered. 783 */ 784 void 785 knote(struct klist *list, long hint) 786 { 787 struct knote *kn; 788 789 SLIST_FOREACH(kn, list, kn_selnext) 790 if (kn->kn_fop->f_event(kn, hint)) 791 KNOTE_ACTIVATE(kn); 792 } 793 794 /* 795 * remove all knotes from a specified klist 796 */ 797 void 798 knote_remove(struct proc *p, struct klist *list) 799 { 800 struct knote *kn; 801 802 while ((kn = SLIST_FIRST(list)) != NULL) { 803 kn->kn_fop->f_detach(kn); 804 knote_drop(kn, p); 805 } 806 } 807 808 /* 809 * remove all knotes referencing a specified fd 810 */ 811 void 812 knote_fdclose(struct proc *p, int fd) 813 { 814 struct filedesc *fdp = p->p_fd; 815 struct klist *list = &fdp->fd_knlist[fd]; 816 817 knote_remove(p, list); 818 } 819 820 static void 821 knote_attach(struct knote *kn, struct filedesc *fdp) 822 { 823 struct klist *list; 824 int size; 825 826 if (! kn->kn_fop->f_isfd) { 827 if (fdp->fd_knhashmask == 0) 828 fdp->fd_knhash = hashinit(KN_HASHSIZE, M_TEMP, 829 &fdp->fd_knhashmask); 830 list = &fdp->fd_knhash[KN_HASH(kn->kn_id, fdp->fd_knhashmask)]; 831 goto done; 832 } 833 834 if (fdp->fd_knlistsize <= kn->kn_id) { 835 size = fdp->fd_knlistsize; 836 while (size <= kn->kn_id) 837 size += KQEXTENT; 838 MALLOC(list, struct klist *, 839 size * sizeof(struct klist *), M_TEMP, M_WAITOK); 840 bcopy((caddr_t)fdp->fd_knlist, (caddr_t)list, 841 fdp->fd_knlistsize * sizeof(struct klist *)); 842 bzero((caddr_t)list + 843 fdp->fd_knlistsize * sizeof(struct klist *), 844 (size - fdp->fd_knlistsize) * sizeof(struct klist *)); 845 if (fdp->fd_knlist != NULL) 846 FREE(fdp->fd_knlist, M_TEMP); 847 fdp->fd_knlistsize = size; 848 fdp->fd_knlist = list; 849 } 850 list = &fdp->fd_knlist[kn->kn_id]; 851 done: 852 SLIST_INSERT_HEAD(list, kn, kn_link); 853 kn->kn_status = 0; 854 } 855 856 /* 857 * should be called at spl == 0, since we don't want to hold spl 858 * while calling fdrop and free. 859 */ 860 static void 861 knote_drop(struct knote *kn, struct proc *p) 862 { 863 struct filedesc *fdp = p->p_fd; 864 struct klist *list; 865 866 if (kn->kn_fop->f_isfd) 867 list = &fdp->fd_knlist[kn->kn_id]; 868 else 869 list = &fdp->fd_knhash[KN_HASH(kn->kn_id, fdp->fd_knhashmask)]; 870 871 SLIST_REMOVE(list, kn, knote, kn_link); 872 if (kn->kn_status & KN_QUEUED) 873 knote_dequeue(kn); 874 if (kn->kn_fop->f_isfd) 875 fdrop(kn->kn_fp, p); 876 knote_free(kn); 877 } 878 879 880 static void 881 knote_enqueue(struct knote *kn) 882 { 883 struct kqueue *kq = kn->kn_kq; 884 int s = splhigh(); 885 886 KASSERT((kn->kn_status & KN_QUEUED) == 0, ("knote already queued")); 887 888 TAILQ_INSERT_TAIL(&kq->kq_head, kn, kn_tqe); 889 kn->kn_status |= KN_QUEUED; 890 kq->kq_count++; 891 splx(s); 892 kqueue_wakeup(kq); 893 } 894 895 static void 896 knote_dequeue(struct knote *kn) 897 { 898 struct kqueue *kq = kn->kn_kq; 899 int s = splhigh(); 900 901 KASSERT(kn->kn_status & KN_QUEUED, ("knote not queued")); 902 903 TAILQ_REMOVE(&kq->kq_head, kn, kn_tqe); 904 kn->kn_status &= ~KN_QUEUED; 905 kq->kq_count--; 906 splx(s); 907 } 908 909 static void 910 knote_init(void) 911 { 912 knote_zone = zinit("KNOTE", sizeof(struct knote), 0, 0, 1); 913 } 914 SYSINIT(knote, SI_SUB_PSEUDO, SI_ORDER_ANY, knote_init, NULL) 915 916 static struct knote * 917 knote_alloc(void) 918 { 919 return ((struct knote *)zalloc(knote_zone)); 920 } 921 922 static void 923 knote_free(struct knote *kn) 924 { 925 zfree(knote_zone, kn); 926 } 927