1 /* $NetBSD: kern_event.c,v 1.104 2018/11/13 06:58:14 maxv Exp $ */ 2 3 /*- 4 * Copyright (c) 2008, 2009 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Andrew Doran. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 21 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 22 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 23 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 29 * POSSIBILITY OF SUCH DAMAGE. 30 */ 31 32 /*- 33 * Copyright (c) 1999,2000,2001 Jonathan Lemon <jlemon@FreeBSD.org> 34 * All rights reserved. 35 * 36 * Redistribution and use in source and binary forms, with or without 37 * modification, are permitted provided that the following conditions 38 * are met: 39 * 1. Redistributions of source code must retain the above copyright 40 * notice, this list of conditions and the following disclaimer. 41 * 2. Redistributions in binary form must reproduce the above copyright 42 * notice, this list of conditions and the following disclaimer in the 43 * documentation and/or other materials provided with the distribution. 44 * 45 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 46 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 47 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 48 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 49 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 50 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 51 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 52 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 53 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 54 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 55 * SUCH DAMAGE. 56 * 57 * FreeBSD: src/sys/kern/kern_event.c,v 1.27 2001/07/05 17:10:44 rwatson Exp 58 */ 59 60 #include <sys/cdefs.h> 61 __KERNEL_RCSID(0, "$NetBSD: kern_event.c,v 1.104 2018/11/13 06:58:14 maxv Exp $"); 62 63 #include <sys/param.h> 64 #include <sys/systm.h> 65 #include <sys/kernel.h> 66 #include <sys/wait.h> 67 #include <sys/proc.h> 68 #include <sys/file.h> 69 #include <sys/select.h> 70 #include <sys/queue.h> 71 #include <sys/event.h> 72 #include <sys/eventvar.h> 73 #include <sys/poll.h> 74 #include <sys/kmem.h> 75 #include <sys/stat.h> 76 #include <sys/filedesc.h> 77 #include <sys/syscallargs.h> 78 #include <sys/kauth.h> 79 #include <sys/conf.h> 80 #include <sys/atomic.h> 81 82 static int kqueue_scan(file_t *, size_t, struct kevent *, 83 const struct timespec *, register_t *, 84 const struct kevent_ops *, struct kevent *, 85 size_t); 86 static int kqueue_ioctl(file_t *, u_long, void *); 87 static int kqueue_fcntl(file_t *, u_int, void *); 88 static int kqueue_poll(file_t *, int); 89 static int kqueue_kqfilter(file_t *, struct knote *); 90 static int kqueue_stat(file_t *, struct stat *); 91 static int kqueue_close(file_t *); 92 static int kqueue_register(struct kqueue *, struct kevent *); 93 static void kqueue_doclose(struct kqueue *, struct klist *, int); 94 95 static void knote_detach(struct knote *, filedesc_t *fdp, bool); 96 static void knote_enqueue(struct knote *); 97 static void knote_activate(struct knote *); 98 99 static void filt_kqdetach(struct knote *); 100 static int filt_kqueue(struct knote *, long hint); 101 static int filt_procattach(struct knote *); 102 static void filt_procdetach(struct knote *); 103 static int filt_proc(struct knote *, long hint); 104 static int filt_fileattach(struct knote *); 105 static void filt_timerexpire(void *x); 106 static int filt_timerattach(struct knote *); 107 static void filt_timerdetach(struct knote *); 108 static int filt_timer(struct knote *, long hint); 109 static int filt_fsattach(struct knote *kn); 110 static void filt_fsdetach(struct knote *kn); 111 static int filt_fs(struct knote *kn, long hint); 112 113 static const struct fileops kqueueops = { 114 .fo_name = "kqueue", 115 .fo_read = (void *)enxio, 116 .fo_write = (void *)enxio, 117 .fo_ioctl = kqueue_ioctl, 118 .fo_fcntl = kqueue_fcntl, 119 .fo_poll = kqueue_poll, 120 .fo_stat = kqueue_stat, 121 .fo_close = kqueue_close, 122 .fo_kqfilter = kqueue_kqfilter, 123 .fo_restart = fnullop_restart, 124 }; 125 126 static const struct filterops kqread_filtops = { 127 .f_isfd = 1, 128 .f_attach = NULL, 129 .f_detach = filt_kqdetach, 130 .f_event = filt_kqueue, 131 }; 132 133 static const struct filterops proc_filtops = { 134 .f_isfd = 0, 135 .f_attach = filt_procattach, 136 .f_detach = filt_procdetach, 137 .f_event = filt_proc, 138 }; 139 140 static const struct filterops file_filtops = { 141 .f_isfd = 1, 142 .f_attach = filt_fileattach, 143 .f_detach = NULL, 144 .f_event = NULL, 145 }; 146 147 static const struct filterops timer_filtops = { 148 .f_isfd = 0, 149 .f_attach = filt_timerattach, 150 .f_detach = filt_timerdetach, 151 .f_event = filt_timer, 152 }; 153 154 static const struct filterops fs_filtops = { 155 .f_isfd = 0, 156 .f_attach = filt_fsattach, 157 .f_detach = filt_fsdetach, 158 .f_event = filt_fs, 159 }; 160 161 static u_int kq_ncallouts = 0; 162 static int kq_calloutmax = (4 * 1024); 163 164 #define KN_HASHSIZE 64 /* XXX should be tunable */ 165 #define KN_HASH(val, mask) (((val) ^ (val >> 8)) & (mask)) 166 167 extern const struct filterops sig_filtops; 168 169 /* 170 * Table for for all system-defined filters. 171 * These should be listed in the numeric order of the EVFILT_* defines. 172 * If filtops is NULL, the filter isn't implemented in NetBSD. 173 * End of list is when name is NULL. 174 * 175 * Note that 'refcnt' is meaningless for built-in filters. 176 */ 177 struct kfilter { 178 const char *name; /* name of filter */ 179 uint32_t filter; /* id of filter */ 180 unsigned refcnt; /* reference count */ 181 const struct filterops *filtops;/* operations for filter */ 182 size_t namelen; /* length of name string */ 183 }; 184 185 /* System defined filters */ 186 static struct kfilter sys_kfilters[] = { 187 { "EVFILT_READ", EVFILT_READ, 0, &file_filtops, 0 }, 188 { "EVFILT_WRITE", EVFILT_WRITE, 0, &file_filtops, 0, }, 189 { "EVFILT_AIO", EVFILT_AIO, 0, NULL, 0 }, 190 { "EVFILT_VNODE", EVFILT_VNODE, 0, &file_filtops, 0 }, 191 { "EVFILT_PROC", EVFILT_PROC, 0, &proc_filtops, 0 }, 192 { "EVFILT_SIGNAL", EVFILT_SIGNAL, 0, &sig_filtops, 0 }, 193 { "EVFILT_TIMER", EVFILT_TIMER, 0, &timer_filtops, 0 }, 194 { "EVFILT_FS", EVFILT_FS, 0, &fs_filtops, 0 }, 195 { NULL, 0, 0, NULL, 0 }, 196 }; 197 198 /* User defined kfilters */ 199 static struct kfilter *user_kfilters; /* array */ 200 static int user_kfilterc; /* current offset */ 201 static int user_kfiltermaxc; /* max size so far */ 202 static size_t user_kfiltersz; /* size of allocated memory */ 203 204 /* 205 * Global Locks. 206 * 207 * Lock order: 208 * 209 * kqueue_filter_lock 210 * -> kn_kq->kq_fdp->fd_lock 211 * -> object lock (e.g., device driver lock, kqueue_misc_lock, &c.) 212 * -> kn_kq->kq_lock 213 * 214 * Locking rules: 215 * 216 * f_attach: fdp->fd_lock, KERNEL_LOCK 217 * f_detach: fdp->fd_lock, KERNEL_LOCK 218 * f_event(!NOTE_SUBMIT) via kevent: fdp->fd_lock, _no_ object lock 219 * f_event via knote: whatever caller guarantees 220 * Typically, f_event(NOTE_SUBMIT) via knote: object lock 221 * f_event(!NOTE_SUBMIT) via knote: nothing, 222 * acquires/releases object lock inside. 223 */ 224 static krwlock_t kqueue_filter_lock; /* lock on filter lists */ 225 static kmutex_t kqueue_misc_lock; /* miscellaneous */ 226 227 static kauth_listener_t kqueue_listener; 228 229 static int 230 kqueue_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie, 231 void *arg0, void *arg1, void *arg2, void *arg3) 232 { 233 struct proc *p; 234 int result; 235 236 result = KAUTH_RESULT_DEFER; 237 p = arg0; 238 239 if (action != KAUTH_PROCESS_KEVENT_FILTER) 240 return result; 241 242 if ((kauth_cred_getuid(p->p_cred) != kauth_cred_getuid(cred) || 243 ISSET(p->p_flag, PK_SUGID))) 244 return result; 245 246 result = KAUTH_RESULT_ALLOW; 247 248 return result; 249 } 250 251 /* 252 * Initialize the kqueue subsystem. 253 */ 254 void 255 kqueue_init(void) 256 { 257 258 rw_init(&kqueue_filter_lock); 259 mutex_init(&kqueue_misc_lock, MUTEX_DEFAULT, IPL_NONE); 260 261 kqueue_listener = kauth_listen_scope(KAUTH_SCOPE_PROCESS, 262 kqueue_listener_cb, NULL); 263 } 264 265 /* 266 * Find kfilter entry by name, or NULL if not found. 267 */ 268 static struct kfilter * 269 kfilter_byname_sys(const char *name) 270 { 271 int i; 272 273 KASSERT(rw_lock_held(&kqueue_filter_lock)); 274 275 for (i = 0; sys_kfilters[i].name != NULL; i++) { 276 if (strcmp(name, sys_kfilters[i].name) == 0) 277 return &sys_kfilters[i]; 278 } 279 return NULL; 280 } 281 282 static struct kfilter * 283 kfilter_byname_user(const char *name) 284 { 285 int i; 286 287 KASSERT(rw_lock_held(&kqueue_filter_lock)); 288 289 /* user filter slots have a NULL name if previously deregistered */ 290 for (i = 0; i < user_kfilterc ; i++) { 291 if (user_kfilters[i].name != NULL && 292 strcmp(name, user_kfilters[i].name) == 0) 293 return &user_kfilters[i]; 294 } 295 return NULL; 296 } 297 298 static struct kfilter * 299 kfilter_byname(const char *name) 300 { 301 struct kfilter *kfilter; 302 303 KASSERT(rw_lock_held(&kqueue_filter_lock)); 304 305 if ((kfilter = kfilter_byname_sys(name)) != NULL) 306 return kfilter; 307 308 return kfilter_byname_user(name); 309 } 310 311 /* 312 * Find kfilter entry by filter id, or NULL if not found. 313 * Assumes entries are indexed in filter id order, for speed. 314 */ 315 static struct kfilter * 316 kfilter_byfilter(uint32_t filter) 317 { 318 struct kfilter *kfilter; 319 320 KASSERT(rw_lock_held(&kqueue_filter_lock)); 321 322 if (filter < EVFILT_SYSCOUNT) /* it's a system filter */ 323 kfilter = &sys_kfilters[filter]; 324 else if (user_kfilters != NULL && 325 filter < EVFILT_SYSCOUNT + user_kfilterc) 326 /* it's a user filter */ 327 kfilter = &user_kfilters[filter - EVFILT_SYSCOUNT]; 328 else 329 return (NULL); /* out of range */ 330 KASSERT(kfilter->filter == filter); /* sanity check! */ 331 return (kfilter); 332 } 333 334 /* 335 * Register a new kfilter. Stores the entry in user_kfilters. 336 * Returns 0 if operation succeeded, or an appropriate errno(2) otherwise. 337 * If retfilter != NULL, the new filterid is returned in it. 338 */ 339 int 340 kfilter_register(const char *name, const struct filterops *filtops, 341 int *retfilter) 342 { 343 struct kfilter *kfilter; 344 size_t len; 345 int i; 346 347 if (name == NULL || name[0] == '\0' || filtops == NULL) 348 return (EINVAL); /* invalid args */ 349 350 rw_enter(&kqueue_filter_lock, RW_WRITER); 351 if (kfilter_byname(name) != NULL) { 352 rw_exit(&kqueue_filter_lock); 353 return (EEXIST); /* already exists */ 354 } 355 if (user_kfilterc > 0xffffffff - EVFILT_SYSCOUNT) { 356 rw_exit(&kqueue_filter_lock); 357 return (EINVAL); /* too many */ 358 } 359 360 for (i = 0; i < user_kfilterc; i++) { 361 kfilter = &user_kfilters[i]; 362 if (kfilter->name == NULL) { 363 /* Previously deregistered slot. Reuse. */ 364 goto reuse; 365 } 366 } 367 368 /* check if need to grow user_kfilters */ 369 if (user_kfilterc + 1 > user_kfiltermaxc) { 370 /* Grow in KFILTER_EXTENT chunks. */ 371 user_kfiltermaxc += KFILTER_EXTENT; 372 len = user_kfiltermaxc * sizeof(*kfilter); 373 kfilter = kmem_alloc(len, KM_SLEEP); 374 memset((char *)kfilter + user_kfiltersz, 0, len - user_kfiltersz); 375 if (user_kfilters != NULL) { 376 memcpy(kfilter, user_kfilters, user_kfiltersz); 377 kmem_free(user_kfilters, user_kfiltersz); 378 } 379 user_kfiltersz = len; 380 user_kfilters = kfilter; 381 } 382 /* Adding new slot */ 383 kfilter = &user_kfilters[user_kfilterc++]; 384 reuse: 385 kfilter->name = kmem_strdupsize(name, &kfilter->namelen, KM_SLEEP); 386 387 kfilter->filter = (kfilter - user_kfilters) + EVFILT_SYSCOUNT; 388 389 kfilter->filtops = kmem_alloc(sizeof(*filtops), KM_SLEEP); 390 memcpy(__UNCONST(kfilter->filtops), filtops, sizeof(*filtops)); 391 392 if (retfilter != NULL) 393 *retfilter = kfilter->filter; 394 rw_exit(&kqueue_filter_lock); 395 396 return (0); 397 } 398 399 /* 400 * Unregister a kfilter previously registered with kfilter_register. 401 * This retains the filter id, but clears the name and frees filtops (filter 402 * operations), so that the number isn't reused during a boot. 403 * Returns 0 if operation succeeded, or an appropriate errno(2) otherwise. 404 */ 405 int 406 kfilter_unregister(const char *name) 407 { 408 struct kfilter *kfilter; 409 410 if (name == NULL || name[0] == '\0') 411 return (EINVAL); /* invalid name */ 412 413 rw_enter(&kqueue_filter_lock, RW_WRITER); 414 if (kfilter_byname_sys(name) != NULL) { 415 rw_exit(&kqueue_filter_lock); 416 return (EINVAL); /* can't detach system filters */ 417 } 418 419 kfilter = kfilter_byname_user(name); 420 if (kfilter == NULL) { 421 rw_exit(&kqueue_filter_lock); 422 return (ENOENT); 423 } 424 if (kfilter->refcnt != 0) { 425 rw_exit(&kqueue_filter_lock); 426 return (EBUSY); 427 } 428 429 /* Cast away const (but we know it's safe. */ 430 kmem_free(__UNCONST(kfilter->name), kfilter->namelen); 431 kfilter->name = NULL; /* mark as `not implemented' */ 432 433 if (kfilter->filtops != NULL) { 434 /* Cast away const (but we know it's safe. */ 435 kmem_free(__UNCONST(kfilter->filtops), 436 sizeof(*kfilter->filtops)); 437 kfilter->filtops = NULL; /* mark as `not implemented' */ 438 } 439 rw_exit(&kqueue_filter_lock); 440 441 return (0); 442 } 443 444 445 /* 446 * Filter attach method for EVFILT_READ and EVFILT_WRITE on normal file 447 * descriptors. Calls fileops kqfilter method for given file descriptor. 448 */ 449 static int 450 filt_fileattach(struct knote *kn) 451 { 452 file_t *fp; 453 454 fp = kn->kn_obj; 455 456 return (*fp->f_ops->fo_kqfilter)(fp, kn); 457 } 458 459 /* 460 * Filter detach method for EVFILT_READ on kqueue descriptor. 461 */ 462 static void 463 filt_kqdetach(struct knote *kn) 464 { 465 struct kqueue *kq; 466 467 kq = ((file_t *)kn->kn_obj)->f_kqueue; 468 469 mutex_spin_enter(&kq->kq_lock); 470 SLIST_REMOVE(&kq->kq_sel.sel_klist, kn, knote, kn_selnext); 471 mutex_spin_exit(&kq->kq_lock); 472 } 473 474 /* 475 * Filter event method for EVFILT_READ on kqueue descriptor. 476 */ 477 /*ARGSUSED*/ 478 static int 479 filt_kqueue(struct knote *kn, long hint) 480 { 481 struct kqueue *kq; 482 int rv; 483 484 kq = ((file_t *)kn->kn_obj)->f_kqueue; 485 486 if (hint != NOTE_SUBMIT) 487 mutex_spin_enter(&kq->kq_lock); 488 kn->kn_data = kq->kq_count; 489 rv = (kn->kn_data > 0); 490 if (hint != NOTE_SUBMIT) 491 mutex_spin_exit(&kq->kq_lock); 492 493 return rv; 494 } 495 496 /* 497 * Filter attach method for EVFILT_PROC. 498 */ 499 static int 500 filt_procattach(struct knote *kn) 501 { 502 struct proc *p; 503 struct lwp *curl; 504 505 curl = curlwp; 506 507 mutex_enter(proc_lock); 508 if (kn->kn_flags & EV_FLAG1) { 509 /* 510 * NOTE_TRACK attaches to the child process too early 511 * for proc_find, so do a raw look up and check the state 512 * explicitly. 513 */ 514 p = proc_find_raw(kn->kn_id); 515 if (p != NULL && p->p_stat != SIDL) 516 p = NULL; 517 } else { 518 p = proc_find(kn->kn_id); 519 } 520 521 if (p == NULL) { 522 mutex_exit(proc_lock); 523 return ESRCH; 524 } 525 526 /* 527 * Fail if it's not owned by you, or the last exec gave us 528 * setuid/setgid privs (unless you're root). 529 */ 530 mutex_enter(p->p_lock); 531 mutex_exit(proc_lock); 532 if (kauth_authorize_process(curl->l_cred, KAUTH_PROCESS_KEVENT_FILTER, 533 p, NULL, NULL, NULL) != 0) { 534 mutex_exit(p->p_lock); 535 return EACCES; 536 } 537 538 kn->kn_obj = p; 539 kn->kn_flags |= EV_CLEAR; /* automatically set */ 540 541 /* 542 * internal flag indicating registration done by kernel 543 */ 544 if (kn->kn_flags & EV_FLAG1) { 545 kn->kn_data = kn->kn_sdata; /* ppid */ 546 kn->kn_fflags = NOTE_CHILD; 547 kn->kn_flags &= ~EV_FLAG1; 548 } 549 SLIST_INSERT_HEAD(&p->p_klist, kn, kn_selnext); 550 mutex_exit(p->p_lock); 551 552 return 0; 553 } 554 555 /* 556 * Filter detach method for EVFILT_PROC. 557 * 558 * The knote may be attached to a different process, which may exit, 559 * leaving nothing for the knote to be attached to. So when the process 560 * exits, the knote is marked as DETACHED and also flagged as ONESHOT so 561 * it will be deleted when read out. However, as part of the knote deletion, 562 * this routine is called, so a check is needed to avoid actually performing 563 * a detach, because the original process might not exist any more. 564 */ 565 static void 566 filt_procdetach(struct knote *kn) 567 { 568 struct proc *p; 569 570 if (kn->kn_status & KN_DETACHED) 571 return; 572 573 p = kn->kn_obj; 574 575 mutex_enter(p->p_lock); 576 SLIST_REMOVE(&p->p_klist, kn, knote, kn_selnext); 577 mutex_exit(p->p_lock); 578 } 579 580 /* 581 * Filter event method for EVFILT_PROC. 582 */ 583 static int 584 filt_proc(struct knote *kn, long hint) 585 { 586 u_int event, fflag; 587 struct kevent kev; 588 struct kqueue *kq; 589 int error; 590 591 event = (u_int)hint & NOTE_PCTRLMASK; 592 kq = kn->kn_kq; 593 fflag = 0; 594 595 /* If the user is interested in this event, record it. */ 596 if (kn->kn_sfflags & event) 597 fflag |= event; 598 599 if (event == NOTE_EXIT) { 600 struct proc *p = kn->kn_obj; 601 602 if (p != NULL) 603 kn->kn_data = P_WAITSTATUS(p); 604 /* 605 * Process is gone, so flag the event as finished. 606 * 607 * Detach the knote from watched process and mark 608 * it as such. We can't leave this to kqueue_scan(), 609 * since the process might not exist by then. And we 610 * have to do this now, since psignal KNOTE() is called 611 * also for zombies and we might end up reading freed 612 * memory if the kevent would already be picked up 613 * and knote g/c'ed. 614 */ 615 filt_procdetach(kn); 616 617 mutex_spin_enter(&kq->kq_lock); 618 kn->kn_status |= KN_DETACHED; 619 /* Mark as ONESHOT, so that the knote it g/c'ed when read */ 620 kn->kn_flags |= (EV_EOF | EV_ONESHOT); 621 kn->kn_fflags |= fflag; 622 mutex_spin_exit(&kq->kq_lock); 623 624 return 1; 625 } 626 627 mutex_spin_enter(&kq->kq_lock); 628 if ((event == NOTE_FORK) && (kn->kn_sfflags & NOTE_TRACK)) { 629 /* 630 * Process forked, and user wants to track the new process, 631 * so attach a new knote to it, and immediately report an 632 * event with the parent's pid. Register knote with new 633 * process. 634 */ 635 memset(&kev, 0, sizeof(kev)); 636 kev.ident = hint & NOTE_PDATAMASK; /* pid */ 637 kev.filter = kn->kn_filter; 638 kev.flags = kn->kn_flags | EV_ADD | EV_ENABLE | EV_FLAG1; 639 kev.fflags = kn->kn_sfflags; 640 kev.data = kn->kn_id; /* parent */ 641 kev.udata = kn->kn_kevent.udata; /* preserve udata */ 642 mutex_spin_exit(&kq->kq_lock); 643 error = kqueue_register(kq, &kev); 644 mutex_spin_enter(&kq->kq_lock); 645 if (error != 0) 646 kn->kn_fflags |= NOTE_TRACKERR; 647 } 648 kn->kn_fflags |= fflag; 649 fflag = kn->kn_fflags; 650 mutex_spin_exit(&kq->kq_lock); 651 652 return fflag != 0; 653 } 654 655 static void 656 filt_timerexpire(void *knx) 657 { 658 struct knote *kn = knx; 659 int tticks; 660 661 mutex_enter(&kqueue_misc_lock); 662 kn->kn_data++; 663 knote_activate(kn); 664 if ((kn->kn_flags & EV_ONESHOT) == 0) { 665 tticks = mstohz(kn->kn_sdata); 666 if (tticks <= 0) 667 tticks = 1; 668 callout_schedule((callout_t *)kn->kn_hook, tticks); 669 } 670 mutex_exit(&kqueue_misc_lock); 671 } 672 673 /* 674 * data contains amount of time to sleep, in milliseconds 675 */ 676 static int 677 filt_timerattach(struct knote *kn) 678 { 679 callout_t *calloutp; 680 struct kqueue *kq; 681 int tticks; 682 683 tticks = mstohz(kn->kn_sdata); 684 685 /* if the supplied value is under our resolution, use 1 tick */ 686 if (tticks == 0) { 687 if (kn->kn_sdata == 0) 688 return EINVAL; 689 tticks = 1; 690 } 691 692 if (atomic_inc_uint_nv(&kq_ncallouts) >= kq_calloutmax || 693 (calloutp = kmem_alloc(sizeof(*calloutp), KM_NOSLEEP)) == NULL) { 694 atomic_dec_uint(&kq_ncallouts); 695 return ENOMEM; 696 } 697 callout_init(calloutp, CALLOUT_MPSAFE); 698 699 kq = kn->kn_kq; 700 mutex_spin_enter(&kq->kq_lock); 701 kn->kn_flags |= EV_CLEAR; /* automatically set */ 702 kn->kn_hook = calloutp; 703 mutex_spin_exit(&kq->kq_lock); 704 705 callout_reset(calloutp, tticks, filt_timerexpire, kn); 706 707 return (0); 708 } 709 710 static void 711 filt_timerdetach(struct knote *kn) 712 { 713 callout_t *calloutp; 714 struct kqueue *kq = kn->kn_kq; 715 716 mutex_spin_enter(&kq->kq_lock); 717 /* prevent rescheduling when we expire */ 718 kn->kn_flags |= EV_ONESHOT; 719 mutex_spin_exit(&kq->kq_lock); 720 721 calloutp = (callout_t *)kn->kn_hook; 722 callout_halt(calloutp, NULL); 723 callout_destroy(calloutp); 724 kmem_free(calloutp, sizeof(*calloutp)); 725 atomic_dec_uint(&kq_ncallouts); 726 } 727 728 static int 729 filt_timer(struct knote *kn, long hint) 730 { 731 int rv; 732 733 mutex_enter(&kqueue_misc_lock); 734 rv = (kn->kn_data != 0); 735 mutex_exit(&kqueue_misc_lock); 736 737 return rv; 738 } 739 740 /* 741 * Filter event method for EVFILT_FS. 742 */ 743 struct klist fs_klist = SLIST_HEAD_INITIALIZER(&fs_klist); 744 745 static int 746 filt_fsattach(struct knote *kn) 747 { 748 749 mutex_enter(&kqueue_misc_lock); 750 kn->kn_flags |= EV_CLEAR; 751 SLIST_INSERT_HEAD(&fs_klist, kn, kn_selnext); 752 mutex_exit(&kqueue_misc_lock); 753 754 return 0; 755 } 756 757 static void 758 filt_fsdetach(struct knote *kn) 759 { 760 761 mutex_enter(&kqueue_misc_lock); 762 SLIST_REMOVE(&fs_klist, kn, knote, kn_selnext); 763 mutex_exit(&kqueue_misc_lock); 764 } 765 766 static int 767 filt_fs(struct knote *kn, long hint) 768 { 769 int rv; 770 771 mutex_enter(&kqueue_misc_lock); 772 kn->kn_fflags |= hint; 773 rv = (kn->kn_fflags != 0); 774 mutex_exit(&kqueue_misc_lock); 775 776 return rv; 777 } 778 779 /* 780 * filt_seltrue: 781 * 782 * This filter "event" routine simulates seltrue(). 783 */ 784 int 785 filt_seltrue(struct knote *kn, long hint) 786 { 787 788 /* 789 * We don't know how much data can be read/written, 790 * but we know that it *can* be. This is about as 791 * good as select/poll does as well. 792 */ 793 kn->kn_data = 0; 794 return (1); 795 } 796 797 /* 798 * This provides full kqfilter entry for device switch tables, which 799 * has same effect as filter using filt_seltrue() as filter method. 800 */ 801 static void 802 filt_seltruedetach(struct knote *kn) 803 { 804 /* Nothing to do */ 805 } 806 807 const struct filterops seltrue_filtops = { 808 .f_isfd = 1, 809 .f_attach = NULL, 810 .f_detach = filt_seltruedetach, 811 .f_event = filt_seltrue, 812 }; 813 814 int 815 seltrue_kqfilter(dev_t dev, struct knote *kn) 816 { 817 switch (kn->kn_filter) { 818 case EVFILT_READ: 819 case EVFILT_WRITE: 820 kn->kn_fop = &seltrue_filtops; 821 break; 822 default: 823 return (EINVAL); 824 } 825 826 /* Nothing more to do */ 827 return (0); 828 } 829 830 /* 831 * kqueue(2) system call. 832 */ 833 static int 834 kqueue1(struct lwp *l, int flags, register_t *retval) 835 { 836 struct kqueue *kq; 837 file_t *fp; 838 int fd, error; 839 840 if ((error = fd_allocfile(&fp, &fd)) != 0) 841 return error; 842 fp->f_flag = FREAD | FWRITE | (flags & (FNONBLOCK|FNOSIGPIPE)); 843 fp->f_type = DTYPE_KQUEUE; 844 fp->f_ops = &kqueueops; 845 kq = kmem_zalloc(sizeof(*kq), KM_SLEEP); 846 mutex_init(&kq->kq_lock, MUTEX_DEFAULT, IPL_SCHED); 847 cv_init(&kq->kq_cv, "kqueue"); 848 selinit(&kq->kq_sel); 849 TAILQ_INIT(&kq->kq_head); 850 fp->f_kqueue = kq; 851 *retval = fd; 852 kq->kq_fdp = curlwp->l_fd; 853 fd_set_exclose(l, fd, (flags & O_CLOEXEC) != 0); 854 fd_affix(curproc, fp, fd); 855 return error; 856 } 857 858 /* 859 * kqueue(2) system call. 860 */ 861 int 862 sys_kqueue(struct lwp *l, const void *v, register_t *retval) 863 { 864 return kqueue1(l, 0, retval); 865 } 866 867 int 868 sys_kqueue1(struct lwp *l, const struct sys_kqueue1_args *uap, 869 register_t *retval) 870 { 871 /* { 872 syscallarg(int) flags; 873 } */ 874 return kqueue1(l, SCARG(uap, flags), retval); 875 } 876 877 /* 878 * kevent(2) system call. 879 */ 880 int 881 kevent_fetch_changes(void *ctx, const struct kevent *changelist, 882 struct kevent *changes, size_t index, int n) 883 { 884 885 return copyin(changelist + index, changes, n * sizeof(*changes)); 886 } 887 888 int 889 kevent_put_events(void *ctx, struct kevent *events, 890 struct kevent *eventlist, size_t index, int n) 891 { 892 893 return copyout(events, eventlist + index, n * sizeof(*events)); 894 } 895 896 static const struct kevent_ops kevent_native_ops = { 897 .keo_private = NULL, 898 .keo_fetch_timeout = copyin, 899 .keo_fetch_changes = kevent_fetch_changes, 900 .keo_put_events = kevent_put_events, 901 }; 902 903 int 904 sys___kevent50(struct lwp *l, const struct sys___kevent50_args *uap, 905 register_t *retval) 906 { 907 /* { 908 syscallarg(int) fd; 909 syscallarg(const struct kevent *) changelist; 910 syscallarg(size_t) nchanges; 911 syscallarg(struct kevent *) eventlist; 912 syscallarg(size_t) nevents; 913 syscallarg(const struct timespec *) timeout; 914 } */ 915 916 return kevent1(retval, SCARG(uap, fd), SCARG(uap, changelist), 917 SCARG(uap, nchanges), SCARG(uap, eventlist), SCARG(uap, nevents), 918 SCARG(uap, timeout), &kevent_native_ops); 919 } 920 921 int 922 kevent1(register_t *retval, int fd, 923 const struct kevent *changelist, size_t nchanges, 924 struct kevent *eventlist, size_t nevents, 925 const struct timespec *timeout, 926 const struct kevent_ops *keops) 927 { 928 struct kevent *kevp; 929 struct kqueue *kq; 930 struct timespec ts; 931 size_t i, n, ichange; 932 int nerrors, error; 933 struct kevent kevbuf[KQ_NEVENTS]; /* approx 300 bytes on 64-bit */ 934 file_t *fp; 935 936 /* check that we're dealing with a kq */ 937 fp = fd_getfile(fd); 938 if (fp == NULL) 939 return (EBADF); 940 941 if (fp->f_type != DTYPE_KQUEUE) { 942 fd_putfile(fd); 943 return (EBADF); 944 } 945 946 if (timeout != NULL) { 947 error = (*keops->keo_fetch_timeout)(timeout, &ts, sizeof(ts)); 948 if (error) 949 goto done; 950 timeout = &ts; 951 } 952 953 kq = fp->f_kqueue; 954 nerrors = 0; 955 ichange = 0; 956 957 /* traverse list of events to register */ 958 while (nchanges > 0) { 959 n = MIN(nchanges, __arraycount(kevbuf)); 960 error = (*keops->keo_fetch_changes)(keops->keo_private, 961 changelist, kevbuf, ichange, n); 962 if (error) 963 goto done; 964 for (i = 0; i < n; i++) { 965 kevp = &kevbuf[i]; 966 kevp->flags &= ~EV_SYSFLAGS; 967 /* register each knote */ 968 error = kqueue_register(kq, kevp); 969 if (!error && !(kevp->flags & EV_RECEIPT)) 970 continue; 971 if (nevents == 0) 972 goto done; 973 kevp->flags = EV_ERROR; 974 kevp->data = error; 975 error = (*keops->keo_put_events) 976 (keops->keo_private, kevp, 977 eventlist, nerrors, 1); 978 if (error) 979 goto done; 980 nevents--; 981 nerrors++; 982 } 983 nchanges -= n; /* update the results */ 984 ichange += n; 985 } 986 if (nerrors) { 987 *retval = nerrors; 988 error = 0; 989 goto done; 990 } 991 992 /* actually scan through the events */ 993 error = kqueue_scan(fp, nevents, eventlist, timeout, retval, keops, 994 kevbuf, __arraycount(kevbuf)); 995 done: 996 fd_putfile(fd); 997 return (error); 998 } 999 1000 /* 1001 * Register a given kevent kev onto the kqueue 1002 */ 1003 static int 1004 kqueue_register(struct kqueue *kq, struct kevent *kev) 1005 { 1006 struct kfilter *kfilter; 1007 filedesc_t *fdp; 1008 file_t *fp; 1009 fdfile_t *ff; 1010 struct knote *kn, *newkn; 1011 struct klist *list; 1012 int error, fd, rv; 1013 1014 fdp = kq->kq_fdp; 1015 fp = NULL; 1016 kn = NULL; 1017 error = 0; 1018 fd = 0; 1019 1020 newkn = kmem_zalloc(sizeof(*newkn), KM_SLEEP); 1021 1022 rw_enter(&kqueue_filter_lock, RW_READER); 1023 kfilter = kfilter_byfilter(kev->filter); 1024 if (kfilter == NULL || kfilter->filtops == NULL) { 1025 /* filter not found nor implemented */ 1026 rw_exit(&kqueue_filter_lock); 1027 kmem_free(newkn, sizeof(*newkn)); 1028 return (EINVAL); 1029 } 1030 1031 /* search if knote already exists */ 1032 if (kfilter->filtops->f_isfd) { 1033 /* monitoring a file descriptor */ 1034 /* validate descriptor */ 1035 if (kev->ident > INT_MAX 1036 || (fp = fd_getfile(fd = kev->ident)) == NULL) { 1037 rw_exit(&kqueue_filter_lock); 1038 kmem_free(newkn, sizeof(*newkn)); 1039 return EBADF; 1040 } 1041 mutex_enter(&fdp->fd_lock); 1042 ff = fdp->fd_dt->dt_ff[fd]; 1043 if (ff->ff_refcnt & FR_CLOSING) { 1044 error = EBADF; 1045 goto doneunlock; 1046 } 1047 if (fd <= fdp->fd_lastkqfile) { 1048 SLIST_FOREACH(kn, &ff->ff_knlist, kn_link) { 1049 if (kq == kn->kn_kq && 1050 kev->filter == kn->kn_filter) 1051 break; 1052 } 1053 } 1054 } else { 1055 /* 1056 * not monitoring a file descriptor, so 1057 * lookup knotes in internal hash table 1058 */ 1059 mutex_enter(&fdp->fd_lock); 1060 if (fdp->fd_knhashmask != 0) { 1061 list = &fdp->fd_knhash[ 1062 KN_HASH((u_long)kev->ident, fdp->fd_knhashmask)]; 1063 SLIST_FOREACH(kn, list, kn_link) { 1064 if (kev->ident == kn->kn_id && 1065 kq == kn->kn_kq && 1066 kev->filter == kn->kn_filter) 1067 break; 1068 } 1069 } 1070 } 1071 1072 /* 1073 * kn now contains the matching knote, or NULL if no match 1074 */ 1075 if (kev->flags & EV_ADD) { 1076 if (kn == NULL) { 1077 /* create new knote */ 1078 kn = newkn; 1079 newkn = NULL; 1080 kn->kn_obj = fp; 1081 kn->kn_id = kev->ident; 1082 kn->kn_kq = kq; 1083 kn->kn_fop = kfilter->filtops; 1084 kn->kn_kfilter = kfilter; 1085 kn->kn_sfflags = kev->fflags; 1086 kn->kn_sdata = kev->data; 1087 kev->fflags = 0; 1088 kev->data = 0; 1089 kn->kn_kevent = *kev; 1090 1091 KASSERT(kn->kn_fop != NULL); 1092 /* 1093 * apply reference count to knote structure, and 1094 * do not release it at the end of this routine. 1095 */ 1096 fp = NULL; 1097 1098 if (!kn->kn_fop->f_isfd) { 1099 /* 1100 * If knote is not on an fd, store on 1101 * internal hash table. 1102 */ 1103 if (fdp->fd_knhashmask == 0) { 1104 /* XXXAD can block with fd_lock held */ 1105 fdp->fd_knhash = hashinit(KN_HASHSIZE, 1106 HASH_LIST, true, 1107 &fdp->fd_knhashmask); 1108 } 1109 list = &fdp->fd_knhash[KN_HASH(kn->kn_id, 1110 fdp->fd_knhashmask)]; 1111 } else { 1112 /* Otherwise, knote is on an fd. */ 1113 list = (struct klist *) 1114 &fdp->fd_dt->dt_ff[kn->kn_id]->ff_knlist; 1115 if ((int)kn->kn_id > fdp->fd_lastkqfile) 1116 fdp->fd_lastkqfile = kn->kn_id; 1117 } 1118 SLIST_INSERT_HEAD(list, kn, kn_link); 1119 1120 KERNEL_LOCK(1, NULL); /* XXXSMP */ 1121 error = (*kfilter->filtops->f_attach)(kn); 1122 KERNEL_UNLOCK_ONE(NULL); /* XXXSMP */ 1123 if (error != 0) { 1124 #ifdef DEBUG 1125 const file_t *ft = kn->kn_obj; 1126 uprintf("%s: event type %d not supported for " 1127 "file type %d/%s (error %d)\n", __func__, 1128 kn->kn_filter, ft ? ft->f_type : -1, 1129 ft ? ft->f_ops->fo_name : "?", error); 1130 #endif 1131 1132 /* knote_detach() drops fdp->fd_lock */ 1133 knote_detach(kn, fdp, false); 1134 goto done; 1135 } 1136 atomic_inc_uint(&kfilter->refcnt); 1137 } else { 1138 /* 1139 * The user may change some filter values after the 1140 * initial EV_ADD, but doing so will not reset any 1141 * filter which have already been triggered. 1142 */ 1143 kn->kn_sfflags = kev->fflags; 1144 kn->kn_sdata = kev->data; 1145 kn->kn_kevent.udata = kev->udata; 1146 } 1147 /* 1148 * We can get here if we are trying to attach 1149 * an event to a file descriptor that does not 1150 * support events, and the attach routine is 1151 * broken and does not return an error. 1152 */ 1153 KASSERT(kn->kn_fop != NULL); 1154 KASSERT(kn->kn_fop->f_event != NULL); 1155 KERNEL_LOCK(1, NULL); /* XXXSMP */ 1156 rv = (*kn->kn_fop->f_event)(kn, 0); 1157 KERNEL_UNLOCK_ONE(NULL); /* XXXSMP */ 1158 if (rv) 1159 knote_activate(kn); 1160 } else { 1161 if (kn == NULL) { 1162 error = ENOENT; 1163 goto doneunlock; 1164 } 1165 if (kev->flags & EV_DELETE) { 1166 /* knote_detach() drops fdp->fd_lock */ 1167 knote_detach(kn, fdp, true); 1168 goto done; 1169 } 1170 } 1171 1172 /* disable knote */ 1173 if ((kev->flags & EV_DISABLE)) { 1174 mutex_spin_enter(&kq->kq_lock); 1175 if ((kn->kn_status & KN_DISABLED) == 0) 1176 kn->kn_status |= KN_DISABLED; 1177 mutex_spin_exit(&kq->kq_lock); 1178 } 1179 1180 /* enable knote */ 1181 if ((kev->flags & EV_ENABLE)) { 1182 knote_enqueue(kn); 1183 } 1184 doneunlock: 1185 mutex_exit(&fdp->fd_lock); 1186 done: 1187 rw_exit(&kqueue_filter_lock); 1188 if (newkn != NULL) 1189 kmem_free(newkn, sizeof(*newkn)); 1190 if (fp != NULL) 1191 fd_putfile(fd); 1192 return (error); 1193 } 1194 1195 #if defined(DEBUG) 1196 #define KN_FMT(buf, kn) \ 1197 (snprintb((buf), sizeof(buf), __KN_FLAG_BITS, (kn)->kn_status), buf) 1198 1199 static void 1200 kqueue_check(const char *func, size_t line, const struct kqueue *kq) 1201 { 1202 const struct knote *kn; 1203 int count; 1204 int nmarker; 1205 char buf[128]; 1206 1207 KASSERT(mutex_owned(&kq->kq_lock)); 1208 KASSERT(kq->kq_count >= 0); 1209 1210 count = 0; 1211 nmarker = 0; 1212 TAILQ_FOREACH(kn, &kq->kq_head, kn_tqe) { 1213 if ((kn->kn_status & (KN_MARKER | KN_QUEUED)) == 0) { 1214 panic("%s,%zu: kq=%p kn=%p !(MARKER|QUEUED) %s", 1215 func, line, kq, kn, KN_FMT(buf, kn)); 1216 } 1217 if ((kn->kn_status & KN_MARKER) == 0) { 1218 if (kn->kn_kq != kq) { 1219 panic("%s,%zu: kq=%p kn(%p) != kn->kq(%p): %s", 1220 func, line, kq, kn, kn->kn_kq, 1221 KN_FMT(buf, kn)); 1222 } 1223 if ((kn->kn_status & KN_ACTIVE) == 0) { 1224 panic("%s,%zu: kq=%p kn=%p: !ACTIVE %s", 1225 func, line, kq, kn, KN_FMT(buf, kn)); 1226 } 1227 count++; 1228 if (count > kq->kq_count) { 1229 goto bad; 1230 } 1231 } else { 1232 nmarker++; 1233 #if 0 1234 if (nmarker > 10000) { 1235 panic("%s,%zu: kq=%p too many markers: " 1236 "%d != %d, nmarker=%d", 1237 func, line, kq, kq->kq_count, count, 1238 nmarker); 1239 } 1240 #endif 1241 } 1242 } 1243 if (kq->kq_count != count) { 1244 bad: 1245 panic("%s,%zu: kq=%p kq->kq_count(%d) != count(%d), nmarker=%d", 1246 func, line, kq, kq->kq_count, count, nmarker); 1247 } 1248 } 1249 #define kq_check(a) kqueue_check(__func__, __LINE__, (a)) 1250 #else /* defined(DEBUG) */ 1251 #define kq_check(a) /* nothing */ 1252 #endif /* defined(DEBUG) */ 1253 1254 /* 1255 * Scan through the list of events on fp (for a maximum of maxevents), 1256 * returning the results in to ulistp. Timeout is determined by tsp; if 1257 * NULL, wait indefinitely, if 0 valued, perform a poll, otherwise wait 1258 * as appropriate. 1259 */ 1260 static int 1261 kqueue_scan(file_t *fp, size_t maxevents, struct kevent *ulistp, 1262 const struct timespec *tsp, register_t *retval, 1263 const struct kevent_ops *keops, struct kevent *kevbuf, 1264 size_t kevcnt) 1265 { 1266 struct kqueue *kq; 1267 struct kevent *kevp; 1268 struct timespec ats, sleepts; 1269 struct knote *kn, *marker, morker; 1270 size_t count, nkev, nevents; 1271 int timeout, error, rv; 1272 filedesc_t *fdp; 1273 1274 fdp = curlwp->l_fd; 1275 kq = fp->f_kqueue; 1276 count = maxevents; 1277 nkev = nevents = error = 0; 1278 if (count == 0) { 1279 *retval = 0; 1280 return 0; 1281 } 1282 1283 if (tsp) { /* timeout supplied */ 1284 ats = *tsp; 1285 if (inittimeleft(&ats, &sleepts) == -1) { 1286 *retval = maxevents; 1287 return EINVAL; 1288 } 1289 timeout = tstohz(&ats); 1290 if (timeout <= 0) 1291 timeout = -1; /* do poll */ 1292 } else { 1293 /* no timeout, wait forever */ 1294 timeout = 0; 1295 } 1296 1297 memset(&morker, 0, sizeof(morker)); 1298 marker = &morker; 1299 marker->kn_status = KN_MARKER; 1300 mutex_spin_enter(&kq->kq_lock); 1301 retry: 1302 kevp = kevbuf; 1303 if (kq->kq_count == 0) { 1304 if (timeout >= 0) { 1305 error = cv_timedwait_sig(&kq->kq_cv, 1306 &kq->kq_lock, timeout); 1307 if (error == 0) { 1308 if (tsp == NULL || (timeout = 1309 gettimeleft(&ats, &sleepts)) > 0) 1310 goto retry; 1311 } else { 1312 /* don't restart after signals... */ 1313 if (error == ERESTART) 1314 error = EINTR; 1315 if (error == EWOULDBLOCK) 1316 error = 0; 1317 } 1318 } 1319 mutex_spin_exit(&kq->kq_lock); 1320 } else { 1321 /* mark end of knote list */ 1322 TAILQ_INSERT_TAIL(&kq->kq_head, marker, kn_tqe); 1323 1324 /* 1325 * Acquire the fdp->fd_lock interlock to avoid races with 1326 * file creation/destruction from other threads. 1327 */ 1328 mutex_spin_exit(&kq->kq_lock); 1329 mutex_enter(&fdp->fd_lock); 1330 mutex_spin_enter(&kq->kq_lock); 1331 1332 while (count != 0) { 1333 kn = TAILQ_FIRST(&kq->kq_head); /* get next knote */ 1334 while ((kn->kn_status & KN_MARKER) != 0) { 1335 if (kn == marker) { 1336 /* it's our marker, stop */ 1337 TAILQ_REMOVE(&kq->kq_head, kn, kn_tqe); 1338 if (count < maxevents || (tsp != NULL && 1339 (timeout = gettimeleft(&ats, 1340 &sleepts)) <= 0)) 1341 goto done; 1342 mutex_exit(&fdp->fd_lock); 1343 goto retry; 1344 } 1345 /* someone else's marker. */ 1346 kn = TAILQ_NEXT(kn, kn_tqe); 1347 } 1348 kq_check(kq); 1349 kq->kq_count--; 1350 TAILQ_REMOVE(&kq->kq_head, kn, kn_tqe); 1351 kn->kn_status &= ~KN_QUEUED; 1352 kn->kn_status |= KN_BUSY; 1353 kq_check(kq); 1354 if (kn->kn_status & KN_DISABLED) { 1355 kn->kn_status &= ~KN_BUSY; 1356 /* don't want disabled events */ 1357 continue; 1358 } 1359 if ((kn->kn_flags & EV_ONESHOT) == 0) { 1360 mutex_spin_exit(&kq->kq_lock); 1361 KASSERT(kn->kn_fop != NULL); 1362 KASSERT(kn->kn_fop->f_event != NULL); 1363 KERNEL_LOCK(1, NULL); /* XXXSMP */ 1364 KASSERT(mutex_owned(&fdp->fd_lock)); 1365 rv = (*kn->kn_fop->f_event)(kn, 0); 1366 KERNEL_UNLOCK_ONE(NULL); /* XXXSMP */ 1367 mutex_spin_enter(&kq->kq_lock); 1368 /* Re-poll if note was re-enqueued. */ 1369 if ((kn->kn_status & KN_QUEUED) != 0) { 1370 kn->kn_status &= ~KN_BUSY; 1371 continue; 1372 } 1373 if (rv == 0) { 1374 /* 1375 * non-ONESHOT event that hasn't 1376 * triggered again, so de-queue. 1377 */ 1378 kn->kn_status &= ~(KN_ACTIVE|KN_BUSY); 1379 continue; 1380 } 1381 } 1382 /* XXXAD should be got from f_event if !oneshot. */ 1383 *kevp++ = kn->kn_kevent; 1384 nkev++; 1385 if (kn->kn_flags & EV_ONESHOT) { 1386 /* delete ONESHOT events after retrieval */ 1387 kn->kn_status &= ~KN_BUSY; 1388 mutex_spin_exit(&kq->kq_lock); 1389 knote_detach(kn, fdp, true); 1390 mutex_enter(&fdp->fd_lock); 1391 mutex_spin_enter(&kq->kq_lock); 1392 } else if (kn->kn_flags & EV_CLEAR) { 1393 /* clear state after retrieval */ 1394 kn->kn_data = 0; 1395 kn->kn_fflags = 0; 1396 kn->kn_status &= ~(KN_QUEUED|KN_ACTIVE|KN_BUSY); 1397 } else if (kn->kn_flags & EV_DISPATCH) { 1398 kn->kn_status |= KN_DISABLED; 1399 kn->kn_status &= ~(KN_QUEUED|KN_ACTIVE|KN_BUSY); 1400 } else { 1401 /* add event back on list */ 1402 kq_check(kq); 1403 kn->kn_status |= KN_QUEUED; 1404 kn->kn_status &= ~KN_BUSY; 1405 TAILQ_INSERT_TAIL(&kq->kq_head, kn, kn_tqe); 1406 kq->kq_count++; 1407 kq_check(kq); 1408 } 1409 if (nkev == kevcnt) { 1410 /* do copyouts in kevcnt chunks */ 1411 mutex_spin_exit(&kq->kq_lock); 1412 mutex_exit(&fdp->fd_lock); 1413 error = (*keops->keo_put_events) 1414 (keops->keo_private, 1415 kevbuf, ulistp, nevents, nkev); 1416 mutex_enter(&fdp->fd_lock); 1417 mutex_spin_enter(&kq->kq_lock); 1418 nevents += nkev; 1419 nkev = 0; 1420 kevp = kevbuf; 1421 } 1422 count--; 1423 if (error != 0 || count == 0) { 1424 /* remove marker */ 1425 TAILQ_REMOVE(&kq->kq_head, marker, kn_tqe); 1426 break; 1427 } 1428 } 1429 done: 1430 mutex_spin_exit(&kq->kq_lock); 1431 mutex_exit(&fdp->fd_lock); 1432 } 1433 if (nkev != 0) { 1434 /* copyout remaining events */ 1435 error = (*keops->keo_put_events)(keops->keo_private, 1436 kevbuf, ulistp, nevents, nkev); 1437 } 1438 *retval = maxevents - count; 1439 1440 return error; 1441 } 1442 1443 /* 1444 * fileops ioctl method for a kqueue descriptor. 1445 * 1446 * Two ioctls are currently supported. They both use struct kfilter_mapping: 1447 * KFILTER_BYNAME find name for filter, and return result in 1448 * name, which is of size len. 1449 * KFILTER_BYFILTER find filter for name. len is ignored. 1450 */ 1451 /*ARGSUSED*/ 1452 static int 1453 kqueue_ioctl(file_t *fp, u_long com, void *data) 1454 { 1455 struct kfilter_mapping *km; 1456 const struct kfilter *kfilter; 1457 char *name; 1458 int error; 1459 1460 km = data; 1461 error = 0; 1462 name = kmem_alloc(KFILTER_MAXNAME, KM_SLEEP); 1463 1464 switch (com) { 1465 case KFILTER_BYFILTER: /* convert filter -> name */ 1466 rw_enter(&kqueue_filter_lock, RW_READER); 1467 kfilter = kfilter_byfilter(km->filter); 1468 if (kfilter != NULL) { 1469 strlcpy(name, kfilter->name, KFILTER_MAXNAME); 1470 rw_exit(&kqueue_filter_lock); 1471 error = copyoutstr(name, km->name, km->len, NULL); 1472 } else { 1473 rw_exit(&kqueue_filter_lock); 1474 error = ENOENT; 1475 } 1476 break; 1477 1478 case KFILTER_BYNAME: /* convert name -> filter */ 1479 error = copyinstr(km->name, name, KFILTER_MAXNAME, NULL); 1480 if (error) { 1481 break; 1482 } 1483 rw_enter(&kqueue_filter_lock, RW_READER); 1484 kfilter = kfilter_byname(name); 1485 if (kfilter != NULL) 1486 km->filter = kfilter->filter; 1487 else 1488 error = ENOENT; 1489 rw_exit(&kqueue_filter_lock); 1490 break; 1491 1492 default: 1493 error = ENOTTY; 1494 break; 1495 1496 } 1497 kmem_free(name, KFILTER_MAXNAME); 1498 return (error); 1499 } 1500 1501 /* 1502 * fileops fcntl method for a kqueue descriptor. 1503 */ 1504 static int 1505 kqueue_fcntl(file_t *fp, u_int com, void *data) 1506 { 1507 1508 return (ENOTTY); 1509 } 1510 1511 /* 1512 * fileops poll method for a kqueue descriptor. 1513 * Determine if kqueue has events pending. 1514 */ 1515 static int 1516 kqueue_poll(file_t *fp, int events) 1517 { 1518 struct kqueue *kq; 1519 int revents; 1520 1521 kq = fp->f_kqueue; 1522 1523 revents = 0; 1524 if (events & (POLLIN | POLLRDNORM)) { 1525 mutex_spin_enter(&kq->kq_lock); 1526 if (kq->kq_count != 0) { 1527 revents |= events & (POLLIN | POLLRDNORM); 1528 } else { 1529 selrecord(curlwp, &kq->kq_sel); 1530 } 1531 kq_check(kq); 1532 mutex_spin_exit(&kq->kq_lock); 1533 } 1534 1535 return revents; 1536 } 1537 1538 /* 1539 * fileops stat method for a kqueue descriptor. 1540 * Returns dummy info, with st_size being number of events pending. 1541 */ 1542 static int 1543 kqueue_stat(file_t *fp, struct stat *st) 1544 { 1545 struct kqueue *kq; 1546 1547 kq = fp->f_kqueue; 1548 1549 memset(st, 0, sizeof(*st)); 1550 st->st_size = kq->kq_count; 1551 st->st_blksize = sizeof(struct kevent); 1552 st->st_mode = S_IFIFO; 1553 1554 return 0; 1555 } 1556 1557 static void 1558 kqueue_doclose(struct kqueue *kq, struct klist *list, int fd) 1559 { 1560 struct knote *kn; 1561 filedesc_t *fdp; 1562 1563 fdp = kq->kq_fdp; 1564 1565 KASSERT(mutex_owned(&fdp->fd_lock)); 1566 1567 for (kn = SLIST_FIRST(list); kn != NULL;) { 1568 if (kq != kn->kn_kq) { 1569 kn = SLIST_NEXT(kn, kn_link); 1570 continue; 1571 } 1572 knote_detach(kn, fdp, true); 1573 mutex_enter(&fdp->fd_lock); 1574 kn = SLIST_FIRST(list); 1575 } 1576 } 1577 1578 1579 /* 1580 * fileops close method for a kqueue descriptor. 1581 */ 1582 static int 1583 kqueue_close(file_t *fp) 1584 { 1585 struct kqueue *kq; 1586 filedesc_t *fdp; 1587 fdfile_t *ff; 1588 int i; 1589 1590 kq = fp->f_kqueue; 1591 fp->f_kqueue = NULL; 1592 fp->f_type = 0; 1593 fdp = curlwp->l_fd; 1594 1595 mutex_enter(&fdp->fd_lock); 1596 for (i = 0; i <= fdp->fd_lastkqfile; i++) { 1597 if ((ff = fdp->fd_dt->dt_ff[i]) == NULL) 1598 continue; 1599 kqueue_doclose(kq, (struct klist *)&ff->ff_knlist, i); 1600 } 1601 if (fdp->fd_knhashmask != 0) { 1602 for (i = 0; i < fdp->fd_knhashmask + 1; i++) { 1603 kqueue_doclose(kq, &fdp->fd_knhash[i], -1); 1604 } 1605 } 1606 mutex_exit(&fdp->fd_lock); 1607 1608 KASSERT(kq->kq_count == 0); 1609 mutex_destroy(&kq->kq_lock); 1610 cv_destroy(&kq->kq_cv); 1611 seldestroy(&kq->kq_sel); 1612 kmem_free(kq, sizeof(*kq)); 1613 1614 return (0); 1615 } 1616 1617 /* 1618 * struct fileops kqfilter method for a kqueue descriptor. 1619 * Event triggered when monitored kqueue changes. 1620 */ 1621 static int 1622 kqueue_kqfilter(file_t *fp, struct knote *kn) 1623 { 1624 struct kqueue *kq; 1625 1626 kq = ((file_t *)kn->kn_obj)->f_kqueue; 1627 1628 KASSERT(fp == kn->kn_obj); 1629 1630 if (kn->kn_filter != EVFILT_READ) 1631 return 1; 1632 1633 kn->kn_fop = &kqread_filtops; 1634 mutex_enter(&kq->kq_lock); 1635 SLIST_INSERT_HEAD(&kq->kq_sel.sel_klist, kn, kn_selnext); 1636 mutex_exit(&kq->kq_lock); 1637 1638 return 0; 1639 } 1640 1641 1642 /* 1643 * Walk down a list of knotes, activating them if their event has 1644 * triggered. The caller's object lock (e.g. device driver lock) 1645 * must be held. 1646 */ 1647 void 1648 knote(struct klist *list, long hint) 1649 { 1650 struct knote *kn, *tmpkn; 1651 1652 SLIST_FOREACH_SAFE(kn, list, kn_selnext, tmpkn) { 1653 KASSERT(kn->kn_fop != NULL); 1654 KASSERT(kn->kn_fop->f_event != NULL); 1655 if ((*kn->kn_fop->f_event)(kn, hint)) 1656 knote_activate(kn); 1657 } 1658 } 1659 1660 /* 1661 * Remove all knotes referencing a specified fd 1662 */ 1663 void 1664 knote_fdclose(int fd) 1665 { 1666 struct klist *list; 1667 struct knote *kn; 1668 filedesc_t *fdp; 1669 1670 fdp = curlwp->l_fd; 1671 list = (struct klist *)&fdp->fd_dt->dt_ff[fd]->ff_knlist; 1672 mutex_enter(&fdp->fd_lock); 1673 while ((kn = SLIST_FIRST(list)) != NULL) { 1674 knote_detach(kn, fdp, true); 1675 mutex_enter(&fdp->fd_lock); 1676 } 1677 mutex_exit(&fdp->fd_lock); 1678 } 1679 1680 /* 1681 * Drop knote. Called with fdp->fd_lock held, and will drop before 1682 * returning. 1683 */ 1684 static void 1685 knote_detach(struct knote *kn, filedesc_t *fdp, bool dofop) 1686 { 1687 struct klist *list; 1688 struct kqueue *kq; 1689 1690 kq = kn->kn_kq; 1691 1692 KASSERT((kn->kn_status & KN_MARKER) == 0); 1693 KASSERT(mutex_owned(&fdp->fd_lock)); 1694 1695 KASSERT(kn->kn_fop != NULL); 1696 /* Remove from monitored object. */ 1697 if (dofop) { 1698 KASSERT(kn->kn_fop->f_detach != NULL); 1699 KERNEL_LOCK(1, NULL); /* XXXSMP */ 1700 (*kn->kn_fop->f_detach)(kn); 1701 KERNEL_UNLOCK_ONE(NULL); /* XXXSMP */ 1702 } 1703 1704 /* Remove from descriptor table. */ 1705 if (kn->kn_fop->f_isfd) 1706 list = (struct klist *)&fdp->fd_dt->dt_ff[kn->kn_id]->ff_knlist; 1707 else 1708 list = &fdp->fd_knhash[KN_HASH(kn->kn_id, fdp->fd_knhashmask)]; 1709 1710 SLIST_REMOVE(list, kn, knote, kn_link); 1711 1712 /* Remove from kqueue. */ 1713 again: 1714 mutex_spin_enter(&kq->kq_lock); 1715 if ((kn->kn_status & KN_QUEUED) != 0) { 1716 kq_check(kq); 1717 kq->kq_count--; 1718 TAILQ_REMOVE(&kq->kq_head, kn, kn_tqe); 1719 kn->kn_status &= ~KN_QUEUED; 1720 kq_check(kq); 1721 } else if (kn->kn_status & KN_BUSY) { 1722 mutex_spin_exit(&kq->kq_lock); 1723 goto again; 1724 } 1725 mutex_spin_exit(&kq->kq_lock); 1726 1727 mutex_exit(&fdp->fd_lock); 1728 if (kn->kn_fop->f_isfd) 1729 fd_putfile(kn->kn_id); 1730 atomic_dec_uint(&kn->kn_kfilter->refcnt); 1731 kmem_free(kn, sizeof(*kn)); 1732 } 1733 1734 /* 1735 * Queue new event for knote. 1736 */ 1737 static void 1738 knote_enqueue(struct knote *kn) 1739 { 1740 struct kqueue *kq; 1741 1742 KASSERT((kn->kn_status & KN_MARKER) == 0); 1743 1744 kq = kn->kn_kq; 1745 1746 mutex_spin_enter(&kq->kq_lock); 1747 if ((kn->kn_status & KN_DISABLED) != 0) { 1748 kn->kn_status &= ~KN_DISABLED; 1749 } 1750 if ((kn->kn_status & (KN_ACTIVE | KN_QUEUED)) == KN_ACTIVE) { 1751 kq_check(kq); 1752 kn->kn_status |= KN_QUEUED; 1753 TAILQ_INSERT_TAIL(&kq->kq_head, kn, kn_tqe); 1754 kq->kq_count++; 1755 kq_check(kq); 1756 cv_broadcast(&kq->kq_cv); 1757 selnotify(&kq->kq_sel, 0, NOTE_SUBMIT); 1758 } 1759 mutex_spin_exit(&kq->kq_lock); 1760 } 1761 /* 1762 * Queue new event for knote. 1763 */ 1764 static void 1765 knote_activate(struct knote *kn) 1766 { 1767 struct kqueue *kq; 1768 1769 KASSERT((kn->kn_status & KN_MARKER) == 0); 1770 1771 kq = kn->kn_kq; 1772 1773 mutex_spin_enter(&kq->kq_lock); 1774 kn->kn_status |= KN_ACTIVE; 1775 if ((kn->kn_status & (KN_QUEUED | KN_DISABLED)) == 0) { 1776 kq_check(kq); 1777 kn->kn_status |= KN_QUEUED; 1778 TAILQ_INSERT_TAIL(&kq->kq_head, kn, kn_tqe); 1779 kq->kq_count++; 1780 kq_check(kq); 1781 cv_broadcast(&kq->kq_cv); 1782 selnotify(&kq->kq_sel, 0, NOTE_SUBMIT); 1783 } 1784 mutex_spin_exit(&kq->kq_lock); 1785 } 1786