1 /* $OpenBSD: kern_event.c,v 1.28 2006/12/01 07:17:25 camield 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/selinfo.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 133 kqueue_init(void) 134 { 135 136 pool_init(&kqueue_pool, sizeof(struct kqueue), 0, 0, 0, "kqueuepl", 137 &pool_allocator_nointr); 138 pool_init(&knote_pool, sizeof(struct knote), 0, 0, 0, "knotepl", 139 &pool_allocator_nointr); 140 } 141 142 int 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 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 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 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 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 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 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 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 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 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 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 ensuring 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 * filters 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 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, rtv, ttv; 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 timeout = atv.tv_sec > 24 * 60 * 60 ? 569 24 * 60 * 60 * hz : tvtohz(&atv); 570 571 getmicrouptime(&rtv); 572 timeradd(&atv, &rtv, &atv); 573 } else { 574 atv.tv_sec = 0; 575 atv.tv_usec = 0; 576 timeout = 0; 577 } 578 goto start; 579 580 retry: 581 if (atv.tv_sec || atv.tv_usec) { 582 getmicrouptime(&rtv); 583 if (timercmp(&rtv, &atv, >=)) 584 goto done; 585 ttv = atv; 586 timersub(&ttv, &rtv, &ttv); 587 timeout = ttv.tv_sec > 24 * 60 * 60 ? 588 24 * 60 * 60 * hz : tvtohz(&ttv); 589 } 590 591 start: 592 kevp = kq->kq_kev; 593 s = splhigh(); 594 if (kq->kq_count == 0) { 595 if (timeout < 0) { 596 error = EWOULDBLOCK; 597 } else { 598 kq->kq_state |= KQ_SLEEP; 599 error = tsleep(kq, PSOCK | PCATCH, "kqread", timeout); 600 } 601 splx(s); 602 if (error == 0) 603 goto retry; 604 /* don't restart after signals... */ 605 if (error == ERESTART) 606 error = EINTR; 607 else if (error == EWOULDBLOCK) 608 error = 0; 609 goto done; 610 } 611 612 TAILQ_INSERT_TAIL(&kq->kq_head, &marker, kn_tqe); 613 while (count) { 614 kn = TAILQ_FIRST(&kq->kq_head); 615 TAILQ_REMOVE(&kq->kq_head, kn, kn_tqe); 616 if (kn == &marker) { 617 splx(s); 618 if (count == maxevents) 619 goto retry; 620 goto done; 621 } 622 if (kn->kn_status & KN_DISABLED) { 623 kn->kn_status &= ~KN_QUEUED; 624 kq->kq_count--; 625 continue; 626 } 627 if ((kn->kn_flags & EV_ONESHOT) == 0 && 628 kn->kn_fop->f_event(kn, 0) == 0) { 629 kn->kn_status &= ~(KN_QUEUED | KN_ACTIVE); 630 kq->kq_count--; 631 continue; 632 } 633 *kevp = kn->kn_kevent; 634 kevp++; 635 nkev++; 636 if (kn->kn_flags & EV_ONESHOT) { 637 kn->kn_status &= ~KN_QUEUED; 638 kq->kq_count--; 639 splx(s); 640 kn->kn_fop->f_detach(kn); 641 knote_drop(kn, p, p->p_fd); 642 s = splhigh(); 643 } else if (kn->kn_flags & EV_CLEAR) { 644 kn->kn_data = 0; 645 kn->kn_fflags = 0; 646 kn->kn_status &= ~(KN_QUEUED | KN_ACTIVE); 647 kq->kq_count--; 648 } else { 649 TAILQ_INSERT_TAIL(&kq->kq_head, kn, kn_tqe); 650 } 651 count--; 652 if (nkev == KQ_NEVENTS) { 653 splx(s); 654 error = copyout((caddr_t)&kq->kq_kev, (caddr_t)ulistp, 655 sizeof(struct kevent) * nkev); 656 ulistp += nkev; 657 nkev = 0; 658 kevp = kq->kq_kev; 659 s = splhigh(); 660 if (error) 661 break; 662 } 663 } 664 TAILQ_REMOVE(&kq->kq_head, &marker, kn_tqe); 665 splx(s); 666 done: 667 if (nkev != 0) 668 error = copyout((caddr_t)&kq->kq_kev, (caddr_t)ulistp, 669 sizeof(struct kevent) * nkev); 670 *retval = maxevents - count; 671 return (error); 672 } 673 674 /* 675 * XXX 676 * This could be expanded to call kqueue_scan, if desired. 677 */ 678 /*ARGSUSED*/ 679 int 680 kqueue_read(struct file *fp, off_t *poff, struct uio *uio, struct ucred *cred) 681 { 682 return (ENXIO); 683 } 684 685 /*ARGSUSED*/ 686 int 687 kqueue_write(struct file *fp, off_t *poff, struct uio *uio, struct ucred *cred) 688 689 { 690 return (ENXIO); 691 } 692 693 /*ARGSUSED*/ 694 int 695 kqueue_ioctl(struct file *fp, u_long com, caddr_t data, struct proc *p) 696 { 697 return (ENOTTY); 698 } 699 700 /*ARGSUSED*/ 701 int 702 kqueue_poll(struct file *fp, int events, struct proc *p) 703 { 704 struct kqueue *kq = (struct kqueue *)fp->f_data; 705 int revents = 0; 706 int s = splnet(); 707 708 if (events & (POLLIN | POLLRDNORM)) { 709 if (kq->kq_count) { 710 revents |= events & (POLLIN | POLLRDNORM); 711 } else { 712 selrecord(p, &kq->kq_sel); 713 kq->kq_state |= KQ_SEL; 714 } 715 } 716 splx(s); 717 return (revents); 718 } 719 720 /*ARGSUSED*/ 721 int 722 kqueue_stat(struct file *fp, struct stat *st, struct proc *p) 723 { 724 struct kqueue *kq = (struct kqueue *)fp->f_data; 725 726 bzero((void *)st, sizeof(*st)); 727 st->st_size = kq->kq_count; 728 st->st_blksize = sizeof(struct kevent); 729 st->st_mode = S_IFIFO; 730 return (0); 731 } 732 733 /*ARGSUSED*/ 734 int 735 kqueue_close(struct file *fp, struct proc *p) 736 { 737 struct kqueue *kq = (struct kqueue *)fp->f_data; 738 struct filedesc *fdp = p->p_fd; 739 struct knote **knp, *kn, *kn0; 740 int i; 741 742 for (i = 0; i < fdp->fd_knlistsize; i++) { 743 knp = &SLIST_FIRST(&fdp->fd_knlist[i]); 744 kn = *knp; 745 while (kn != NULL) { 746 kn0 = SLIST_NEXT(kn, kn_link); 747 if (kq == kn->kn_kq) { 748 FREF(kn->kn_fp); 749 kn->kn_fop->f_detach(kn); 750 closef(kn->kn_fp, p); 751 knote_free(kn); 752 *knp = kn0; 753 } else { 754 knp = &SLIST_NEXT(kn, kn_link); 755 } 756 kn = kn0; 757 } 758 } 759 if (fdp->fd_knhashmask != 0) { 760 for (i = 0; i < fdp->fd_knhashmask + 1; i++) { 761 knp = &SLIST_FIRST(&fdp->fd_knhash[i]); 762 kn = *knp; 763 while (kn != NULL) { 764 kn0 = SLIST_NEXT(kn, kn_link); 765 if (kq == kn->kn_kq) { 766 kn->kn_fop->f_detach(kn); 767 /* XXX non-fd release of kn->kn_ptr */ 768 knote_free(kn); 769 *knp = kn0; 770 } else { 771 knp = &SLIST_NEXT(kn, kn_link); 772 } 773 kn = kn0; 774 } 775 } 776 } 777 pool_put(&kqueue_pool, kq); 778 fp->f_data = NULL; 779 780 return (0); 781 } 782 783 void 784 kqueue_wakeup(struct kqueue *kq) 785 { 786 787 if (kq->kq_state & KQ_SLEEP) { 788 kq->kq_state &= ~KQ_SLEEP; 789 wakeup(kq); 790 } 791 if (kq->kq_state & KQ_SEL) { 792 kq->kq_state &= ~KQ_SEL; 793 selwakeup(&kq->kq_sel); 794 } 795 KNOTE(&kq->kq_sel.si_note, 0); 796 } 797 798 /* 799 * walk down a list of knotes, activating them if their event has triggered. 800 */ 801 void 802 knote(struct klist *list, long hint) 803 { 804 struct knote *kn; 805 806 SLIST_FOREACH(kn, list, kn_selnext) 807 if (kn->kn_fop->f_event(kn, hint)) 808 KNOTE_ACTIVATE(kn); 809 } 810 811 /* 812 * remove all knotes from a specified klist 813 */ 814 void 815 knote_remove(struct proc *p, struct klist *list) 816 { 817 struct knote *kn; 818 819 while ((kn = SLIST_FIRST(list)) != NULL) { 820 kn->kn_fop->f_detach(kn); 821 knote_drop(kn, p, p->p_fd); 822 } 823 } 824 825 /* 826 * remove all knotes referencing a specified fd 827 */ 828 void 829 knote_fdclose(struct proc *p, int fd) 830 { 831 struct filedesc *fdp = p->p_fd; 832 struct klist *list = &fdp->fd_knlist[fd]; 833 834 knote_remove(p, list); 835 } 836 837 void 838 knote_attach(struct knote *kn, struct filedesc *fdp) 839 { 840 struct klist *list; 841 int size; 842 843 if (! kn->kn_fop->f_isfd) { 844 if (fdp->fd_knhashmask == 0) 845 fdp->fd_knhash = hashinit(KN_HASHSIZE, M_TEMP, 846 M_WAITOK, &fdp->fd_knhashmask); 847 list = &fdp->fd_knhash[KN_HASH(kn->kn_id, fdp->fd_knhashmask)]; 848 goto done; 849 } 850 851 if (fdp->fd_knlistsize <= kn->kn_id) { 852 size = fdp->fd_knlistsize; 853 while (size <= kn->kn_id) 854 size += KQEXTENT; 855 list = malloc(size * sizeof(struct klist *), M_TEMP, M_WAITOK); 856 bcopy((caddr_t)fdp->fd_knlist, (caddr_t)list, 857 fdp->fd_knlistsize * sizeof(struct klist *)); 858 bzero((caddr_t)list + 859 fdp->fd_knlistsize * sizeof(struct klist *), 860 (size - fdp->fd_knlistsize) * sizeof(struct klist *)); 861 if (fdp->fd_knlist != NULL) 862 free(fdp->fd_knlist, M_TEMP); 863 fdp->fd_knlistsize = size; 864 fdp->fd_knlist = list; 865 } 866 list = &fdp->fd_knlist[kn->kn_id]; 867 done: 868 SLIST_INSERT_HEAD(list, kn, kn_link); 869 kn->kn_status = 0; 870 } 871 872 /* 873 * should be called at spl == 0, since we don't want to hold spl 874 * while calling closef and free. 875 */ 876 void 877 knote_drop(struct knote *kn, struct proc *p, struct filedesc *fdp) 878 { 879 struct klist *list; 880 881 if (kn->kn_fop->f_isfd) 882 list = &fdp->fd_knlist[kn->kn_id]; 883 else 884 list = &fdp->fd_knhash[KN_HASH(kn->kn_id, fdp->fd_knhashmask)]; 885 886 SLIST_REMOVE(list, kn, knote, kn_link); 887 if (kn->kn_status & KN_QUEUED) 888 knote_dequeue(kn); 889 if (kn->kn_fop->f_isfd) { 890 FREF(kn->kn_fp); 891 closef(kn->kn_fp, p); 892 } 893 knote_free(kn); 894 } 895 896 897 void 898 knote_enqueue(struct knote *kn) 899 { 900 struct kqueue *kq = kn->kn_kq; 901 int s = splhigh(); 902 903 KASSERT((kn->kn_status & KN_QUEUED) == 0); 904 905 TAILQ_INSERT_TAIL(&kq->kq_head, kn, kn_tqe); 906 kn->kn_status |= KN_QUEUED; 907 kq->kq_count++; 908 splx(s); 909 kqueue_wakeup(kq); 910 } 911 912 void 913 knote_dequeue(struct knote *kn) 914 { 915 struct kqueue *kq = kn->kn_kq; 916 int s = splhigh(); 917 918 KASSERT(kn->kn_status & KN_QUEUED); 919 920 TAILQ_REMOVE(&kq->kq_head, kn, kn_tqe); 921 kn->kn_status &= ~KN_QUEUED; 922 kq->kq_count--; 923 splx(s); 924 } 925 926 void 927 klist_invalidate(struct klist *list) 928 { 929 struct knote *kn; 930 931 SLIST_FOREACH(kn, list, kn_selnext) { 932 kn->kn_status |= KN_DETACHED; 933 kn->kn_flags |= EV_EOF | EV_ONESHOT; 934 } 935 } 936