1 /* $NetBSD: sys_select.c,v 1.12 2009/01/11 02:45:52 christos Exp $ */ 2 3 /*- 4 * Copyright (c) 2007, 2008 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) 1982, 1986, 1989, 1993 34 * The Regents of the University of California. All rights reserved. 35 * (c) UNIX System Laboratories, Inc. 36 * All or some portions of this file are derived from material licensed 37 * to the University of California by American Telephone and Telegraph 38 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 39 * the permission of UNIX System Laboratories, Inc. 40 * 41 * Redistribution and use in source and binary forms, with or without 42 * modification, are permitted provided that the following conditions 43 * are met: 44 * 1. Redistributions of source code must retain the above copyright 45 * notice, this list of conditions and the following disclaimer. 46 * 2. Redistributions in binary form must reproduce the above copyright 47 * notice, this list of conditions and the following disclaimer in the 48 * documentation and/or other materials provided with the distribution. 49 * 3. Neither the name of the University nor the names of its contributors 50 * may be used to endorse or promote products derived from this software 51 * without specific prior written permission. 52 * 53 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 54 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 55 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 56 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 57 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 58 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 59 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 60 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 61 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 62 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 63 * SUCH DAMAGE. 64 * 65 * @(#)sys_generic.c 8.9 (Berkeley) 2/14/95 66 */ 67 68 /* 69 * System calls relating to files. 70 */ 71 72 #include <sys/cdefs.h> 73 __KERNEL_RCSID(0, "$NetBSD: sys_select.c,v 1.12 2009/01/11 02:45:52 christos Exp $"); 74 75 #include <sys/param.h> 76 #include <sys/systm.h> 77 #include <sys/filedesc.h> 78 #include <sys/ioctl.h> 79 #include <sys/file.h> 80 #include <sys/proc.h> 81 #include <sys/socketvar.h> 82 #include <sys/signalvar.h> 83 #include <sys/uio.h> 84 #include <sys/kernel.h> 85 #include <sys/stat.h> 86 #include <sys/poll.h> 87 #include <sys/vnode.h> 88 #include <sys/mount.h> 89 #include <sys/syscallargs.h> 90 #include <sys/cpu.h> 91 #include <sys/atomic.h> 92 #include <sys/socketvar.h> 93 #include <sys/sleepq.h> 94 95 /* Flags for lwp::l_selflag. */ 96 #define SEL_RESET 0 /* awoken, interrupted, or not yet polling */ 97 #define SEL_SCANNING 1 /* polling descriptors */ 98 #define SEL_BLOCKING 2 /* about to block on select_cv */ 99 100 /* Per-CPU state for select()/poll(). */ 101 #if MAXCPUS > 32 102 #error adjust this code 103 #endif 104 typedef struct selcpu { 105 kmutex_t sc_lock; 106 sleepq_t sc_sleepq; 107 int sc_ncoll; 108 uint32_t sc_mask; 109 } selcpu_t; 110 111 static int selscan(lwp_t *, fd_mask *, fd_mask *, int, register_t *); 112 static int pollscan(lwp_t *, struct pollfd *, int, register_t *); 113 static void selclear(void); 114 115 static syncobj_t select_sobj = { 116 SOBJ_SLEEPQ_FIFO, 117 sleepq_unsleep, 118 sleepq_changepri, 119 sleepq_lendpri, 120 syncobj_noowner, 121 }; 122 123 /* 124 * Select system call. 125 */ 126 int 127 sys___pselect50(struct lwp *l, const struct sys___pselect50_args *uap, 128 register_t *retval) 129 { 130 /* { 131 syscallarg(int) nd; 132 syscallarg(fd_set *) in; 133 syscallarg(fd_set *) ou; 134 syscallarg(fd_set *) ex; 135 syscallarg(const struct timespec *) ts; 136 syscallarg(sigset_t *) mask; 137 } */ 138 struct timespec ats; 139 struct timeval atv, *tv = NULL; 140 sigset_t amask, *mask = NULL; 141 int error; 142 143 if (SCARG(uap, ts)) { 144 error = copyin(SCARG(uap, ts), &ats, sizeof(ats)); 145 if (error) 146 return error; 147 atv.tv_sec = ats.tv_sec; 148 atv.tv_usec = ats.tv_nsec / 1000; 149 tv = &atv; 150 } 151 if (SCARG(uap, mask) != NULL) { 152 error = copyin(SCARG(uap, mask), &amask, sizeof(amask)); 153 if (error) 154 return error; 155 mask = &amask; 156 } 157 158 return selcommon(l, retval, SCARG(uap, nd), SCARG(uap, in), 159 SCARG(uap, ou), SCARG(uap, ex), tv, mask); 160 } 161 162 int 163 inittimeleft(struct timeval *tv, struct timeval *sleeptv) 164 { 165 if (itimerfix(tv)) 166 return -1; 167 getmicrouptime(sleeptv); 168 return 0; 169 } 170 171 int 172 gettimeleft(struct timeval *tv, struct timeval *sleeptv) 173 { 174 /* 175 * We have to recalculate the timeout on every retry. 176 */ 177 struct timeval slepttv; 178 /* 179 * reduce tv by elapsed time 180 * based on monotonic time scale 181 */ 182 getmicrouptime(&slepttv); 183 timeradd(tv, sleeptv, tv); 184 timersub(tv, &slepttv, tv); 185 *sleeptv = slepttv; 186 return tvtohz(tv); 187 } 188 189 int 190 sys___select50(struct lwp *l, const struct sys___select50_args *uap, 191 register_t *retval) 192 { 193 /* { 194 syscallarg(int) nd; 195 syscallarg(fd_set *) in; 196 syscallarg(fd_set *) ou; 197 syscallarg(fd_set *) ex; 198 syscallarg(struct timeval *) tv; 199 } */ 200 struct timeval atv, *tv = NULL; 201 int error; 202 203 if (SCARG(uap, tv)) { 204 error = copyin(SCARG(uap, tv), (void *)&atv, 205 sizeof(atv)); 206 if (error) 207 return error; 208 tv = &atv; 209 } 210 211 return selcommon(l, retval, SCARG(uap, nd), SCARG(uap, in), 212 SCARG(uap, ou), SCARG(uap, ex), tv, NULL); 213 } 214 215 int 216 selcommon(lwp_t *l, register_t *retval, int nd, fd_set *u_in, 217 fd_set *u_ou, fd_set *u_ex, struct timeval *tv, sigset_t *mask) 218 { 219 char smallbits[howmany(FD_SETSIZE, NFDBITS) * 220 sizeof(fd_mask) * 6]; 221 proc_t * const p = l->l_proc; 222 char *bits; 223 int ncoll, error, timo; 224 size_t ni; 225 sigset_t oldmask; 226 struct timeval sleeptv; 227 selcpu_t *sc; 228 229 error = 0; 230 if (nd < 0) 231 return (EINVAL); 232 if (nd > p->p_fd->fd_nfiles) { 233 /* forgiving; slightly wrong */ 234 nd = p->p_fd->fd_nfiles; 235 } 236 ni = howmany(nd, NFDBITS) * sizeof(fd_mask); 237 if (ni * 6 > sizeof(smallbits)) { 238 bits = kmem_alloc(ni * 6, KM_SLEEP); 239 if (bits == NULL) 240 return ENOMEM; 241 } else 242 bits = smallbits; 243 244 #define getbits(name, x) \ 245 if (u_ ## name) { \ 246 error = copyin(u_ ## name, bits + ni * x, ni); \ 247 if (error) \ 248 goto done; \ 249 } else \ 250 memset(bits + ni * x, 0, ni); 251 getbits(in, 0); 252 getbits(ou, 1); 253 getbits(ex, 2); 254 #undef getbits 255 256 timo = 0; 257 if (tv && inittimeleft(tv, &sleeptv) == -1) { 258 error = EINVAL; 259 goto done; 260 } 261 262 if (mask) { 263 sigminusset(&sigcantmask, mask); 264 mutex_enter(p->p_lock); 265 oldmask = l->l_sigmask; 266 l->l_sigmask = *mask; 267 mutex_exit(p->p_lock); 268 } else 269 oldmask = l->l_sigmask; /* XXXgcc */ 270 271 sc = curcpu()->ci_data.cpu_selcpu; 272 l->l_selcpu = sc; 273 SLIST_INIT(&l->l_selwait); 274 for (;;) { 275 /* 276 * No need to lock. If this is overwritten by another 277 * value while scanning, we will retry below. We only 278 * need to see exact state from the descriptors that 279 * we are about to poll, and lock activity resulting 280 * from fo_poll is enough to provide an up to date value 281 * for new polling activity. 282 */ 283 l->l_selflag = SEL_SCANNING; 284 ncoll = sc->sc_ncoll; 285 286 error = selscan(l, (fd_mask *)(bits + ni * 0), 287 (fd_mask *)(bits + ni * 3), nd, retval); 288 289 if (error || *retval) 290 break; 291 if (tv && (timo = gettimeleft(tv, &sleeptv)) <= 0) 292 break; 293 mutex_spin_enter(&sc->sc_lock); 294 if (l->l_selflag != SEL_SCANNING || sc->sc_ncoll != ncoll) { 295 mutex_spin_exit(&sc->sc_lock); 296 continue; 297 } 298 l->l_selflag = SEL_BLOCKING; 299 l->l_kpriority = true; 300 sleepq_enter(&sc->sc_sleepq, l, &sc->sc_lock); 301 sleepq_enqueue(&sc->sc_sleepq, sc, "select", &select_sobj); 302 error = sleepq_block(timo, true); 303 if (error != 0) 304 break; 305 } 306 selclear(); 307 308 if (mask) { 309 mutex_enter(p->p_lock); 310 l->l_sigmask = oldmask; 311 mutex_exit(p->p_lock); 312 } 313 314 done: 315 /* select is not restarted after signals... */ 316 if (error == ERESTART) 317 error = EINTR; 318 if (error == EWOULDBLOCK) 319 error = 0; 320 if (error == 0 && u_in != NULL) 321 error = copyout(bits + ni * 3, u_in, ni); 322 if (error == 0 && u_ou != NULL) 323 error = copyout(bits + ni * 4, u_ou, ni); 324 if (error == 0 && u_ex != NULL) 325 error = copyout(bits + ni * 5, u_ex, ni); 326 if (bits != smallbits) 327 kmem_free(bits, ni * 6); 328 return (error); 329 } 330 331 int 332 selscan(lwp_t *l, fd_mask *ibitp, fd_mask *obitp, int nfd, 333 register_t *retval) 334 { 335 static const int flag[3] = { POLLRDNORM | POLLHUP | POLLERR, 336 POLLWRNORM | POLLHUP | POLLERR, 337 POLLRDBAND }; 338 int msk, i, j, fd, n; 339 fd_mask ibits, obits; 340 file_t *fp; 341 342 n = 0; 343 for (msk = 0; msk < 3; msk++) { 344 for (i = 0; i < nfd; i += NFDBITS) { 345 ibits = *ibitp++; 346 obits = 0; 347 while ((j = ffs(ibits)) && (fd = i + --j) < nfd) { 348 ibits &= ~(1 << j); 349 if ((fp = fd_getfile(fd)) == NULL) 350 return (EBADF); 351 if ((*fp->f_ops->fo_poll)(fp, flag[msk])) { 352 obits |= (1 << j); 353 n++; 354 } 355 fd_putfile(fd); 356 } 357 *obitp++ = obits; 358 } 359 } 360 *retval = n; 361 return (0); 362 } 363 364 /* 365 * Poll system call. 366 */ 367 int 368 sys_poll(struct lwp *l, const struct sys_poll_args *uap, register_t *retval) 369 { 370 /* { 371 syscallarg(struct pollfd *) fds; 372 syscallarg(u_int) nfds; 373 syscallarg(int) timeout; 374 } */ 375 struct timeval atv, *tv = NULL; 376 377 if (SCARG(uap, timeout) != INFTIM) { 378 atv.tv_sec = SCARG(uap, timeout) / 1000; 379 atv.tv_usec = (SCARG(uap, timeout) % 1000) * 1000; 380 tv = &atv; 381 } 382 383 return pollcommon(l, retval, SCARG(uap, fds), SCARG(uap, nfds), 384 tv, NULL); 385 } 386 387 /* 388 * Poll system call. 389 */ 390 int 391 sys___pollts50(struct lwp *l, const struct sys___pollts50_args *uap, 392 register_t *retval) 393 { 394 /* { 395 syscallarg(struct pollfd *) fds; 396 syscallarg(u_int) nfds; 397 syscallarg(const struct timespec *) ts; 398 syscallarg(const sigset_t *) mask; 399 } */ 400 struct timespec ats; 401 struct timeval atv, *tv = NULL; 402 sigset_t amask, *mask = NULL; 403 int error; 404 405 if (SCARG(uap, ts)) { 406 error = copyin(SCARG(uap, ts), &ats, sizeof(ats)); 407 if (error) 408 return error; 409 atv.tv_sec = ats.tv_sec; 410 atv.tv_usec = ats.tv_nsec / 1000; 411 tv = &atv; 412 } 413 if (SCARG(uap, mask)) { 414 error = copyin(SCARG(uap, mask), &amask, sizeof(amask)); 415 if (error) 416 return error; 417 mask = &amask; 418 } 419 420 return pollcommon(l, retval, SCARG(uap, fds), SCARG(uap, nfds), 421 tv, mask); 422 } 423 424 int 425 pollcommon(lwp_t *l, register_t *retval, 426 struct pollfd *u_fds, u_int nfds, 427 struct timeval *tv, sigset_t *mask) 428 { 429 struct pollfd smallfds[32]; 430 struct pollfd *fds; 431 proc_t * const p = l->l_proc; 432 sigset_t oldmask; 433 int ncoll, error, timo; 434 size_t ni; 435 struct timeval sleeptv; 436 selcpu_t *sc; 437 438 if (nfds > p->p_fd->fd_nfiles) { 439 /* forgiving; slightly wrong */ 440 nfds = p->p_fd->fd_nfiles; 441 } 442 ni = nfds * sizeof(struct pollfd); 443 if (ni > sizeof(smallfds)) { 444 fds = kmem_alloc(ni, KM_SLEEP); 445 if (fds == NULL) 446 return ENOMEM; 447 } else 448 fds = smallfds; 449 450 error = copyin(u_fds, fds, ni); 451 if (error) 452 goto done; 453 454 timo = 0; 455 if (tv && inittimeleft(tv, &sleeptv) == -1) { 456 error = EINVAL; 457 goto done; 458 } 459 460 if (mask) { 461 sigminusset(&sigcantmask, mask); 462 mutex_enter(p->p_lock); 463 oldmask = l->l_sigmask; 464 l->l_sigmask = *mask; 465 mutex_exit(p->p_lock); 466 } else 467 oldmask = l->l_sigmask; /* XXXgcc */ 468 469 sc = curcpu()->ci_data.cpu_selcpu; 470 l->l_selcpu = sc; 471 SLIST_INIT(&l->l_selwait); 472 for (;;) { 473 /* 474 * No need to lock. If this is overwritten by another 475 * value while scanning, we will retry below. We only 476 * need to see exact state from the descriptors that 477 * we are about to poll, and lock activity resulting 478 * from fo_poll is enough to provide an up to date value 479 * for new polling activity. 480 */ 481 ncoll = sc->sc_ncoll; 482 l->l_selflag = SEL_SCANNING; 483 484 error = pollscan(l, fds, nfds, retval); 485 486 if (error || *retval) 487 break; 488 if (tv && (timo = gettimeleft(tv, &sleeptv)) <= 0) 489 break; 490 mutex_spin_enter(&sc->sc_lock); 491 if (l->l_selflag != SEL_SCANNING || sc->sc_ncoll != ncoll) { 492 mutex_spin_exit(&sc->sc_lock); 493 continue; 494 } 495 l->l_selflag = SEL_BLOCKING; 496 l->l_kpriority = true; 497 sleepq_enter(&sc->sc_sleepq, l, &sc->sc_lock); 498 sleepq_enqueue(&sc->sc_sleepq, sc, "select", &select_sobj); 499 error = sleepq_block(timo, true); 500 if (error != 0) 501 break; 502 } 503 selclear(); 504 505 if (mask) { 506 mutex_enter(p->p_lock); 507 l->l_sigmask = oldmask; 508 mutex_exit(p->p_lock); 509 } 510 done: 511 /* poll is not restarted after signals... */ 512 if (error == ERESTART) 513 error = EINTR; 514 if (error == EWOULDBLOCK) 515 error = 0; 516 if (error == 0) 517 error = copyout(fds, u_fds, ni); 518 if (fds != smallfds) 519 kmem_free(fds, ni); 520 return (error); 521 } 522 523 int 524 pollscan(lwp_t *l, struct pollfd *fds, int nfd, register_t *retval) 525 { 526 int i, n; 527 file_t *fp; 528 529 n = 0; 530 for (i = 0; i < nfd; i++, fds++) { 531 if (fds->fd < 0) { 532 fds->revents = 0; 533 } else if ((fp = fd_getfile(fds->fd)) == NULL) { 534 fds->revents = POLLNVAL; 535 n++; 536 } else { 537 fds->revents = (*fp->f_ops->fo_poll)(fp, 538 fds->events | POLLERR | POLLHUP); 539 if (fds->revents != 0) 540 n++; 541 fd_putfile(fds->fd); 542 } 543 } 544 *retval = n; 545 return (0); 546 } 547 548 /*ARGSUSED*/ 549 int 550 seltrue(dev_t dev, int events, lwp_t *l) 551 { 552 553 return (events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM)); 554 } 555 556 /* 557 * Record a select request. Concurrency issues: 558 * 559 * The caller holds the same lock across calls to selrecord() and 560 * selnotify(), so we don't need to consider a concurrent wakeup 561 * while in this routine. 562 * 563 * The only activity we need to guard against is selclear(), called by 564 * another thread that is exiting selcommon() or pollcommon(). 565 * `sel_lwp' can only become non-NULL while the caller's lock is held, 566 * so it cannot become non-NULL due to a change made by another thread 567 * while we are in this routine. It can only become _NULL_ due to a 568 * call to selclear(). 569 * 570 * If it is non-NULL and != selector there is the potential for 571 * selclear() to be called by another thread. If either of those 572 * conditions are true, we're not interested in touching the `named 573 * waiter' part of the selinfo record because we need to record a 574 * collision. Hence there is no need for additional locking in this 575 * routine. 576 */ 577 void 578 selrecord(lwp_t *selector, struct selinfo *sip) 579 { 580 selcpu_t *sc; 581 lwp_t *other; 582 583 KASSERT(selector == curlwp); 584 585 sc = selector->l_selcpu; 586 other = sip->sel_lwp; 587 588 if (other == selector) { 589 /* `selector' has already claimed it. */ 590 KASSERT(sip->sel_cpu = sc); 591 } else if (other == NULL) { 592 /* 593 * First named waiter, although there may be unnamed 594 * waiters (collisions). Issue a memory barrier to 595 * ensure that we access sel_lwp (above) before other 596 * fields - this guards against a call to selclear(). 597 */ 598 membar_enter(); 599 sip->sel_lwp = selector; 600 SLIST_INSERT_HEAD(&selector->l_selwait, sip, sel_chain); 601 /* Replace selinfo's lock with our chosen CPU's lock. */ 602 sip->sel_cpu = sc; 603 } else { 604 /* Multiple waiters: record a collision. */ 605 sip->sel_collision |= sc->sc_mask; 606 KASSERT(sip->sel_cpu != NULL); 607 } 608 } 609 610 /* 611 * Do a wakeup when a selectable event occurs. Concurrency issues: 612 * 613 * As per selrecord(), the caller's object lock is held. If there 614 * is a named waiter, we must acquire the associated selcpu's lock 615 * in order to synchronize with selclear() and pollers going to sleep 616 * in selcommon() and/or pollcommon(). 617 * 618 * sip->sel_cpu cannot change at this point, as it is only changed 619 * in selrecord(), and concurrent calls to selrecord() are locked 620 * out by the caller. 621 */ 622 void 623 selnotify(struct selinfo *sip, int events, long knhint) 624 { 625 selcpu_t *sc; 626 uint32_t mask; 627 int index, oflag, swapin; 628 lwp_t *l; 629 630 KNOTE(&sip->sel_klist, knhint); 631 632 if (sip->sel_lwp != NULL) { 633 /* One named LWP is waiting. */ 634 swapin = 0; 635 sc = sip->sel_cpu; 636 mutex_spin_enter(&sc->sc_lock); 637 /* Still there? */ 638 if (sip->sel_lwp != NULL) { 639 l = sip->sel_lwp; 640 /* 641 * If thread is sleeping, wake it up. If it's not 642 * yet asleep, it will notice the change in state 643 * and will re-poll the descriptors. 644 */ 645 oflag = l->l_selflag; 646 l->l_selflag = SEL_RESET; 647 if (oflag == SEL_BLOCKING && 648 l->l_mutex == &sc->sc_lock) { 649 KASSERT(l->l_wchan == sc); 650 swapin = sleepq_unsleep(l, false); 651 } 652 } 653 mutex_spin_exit(&sc->sc_lock); 654 if (swapin) 655 uvm_kick_scheduler(); 656 } 657 658 if ((mask = sip->sel_collision) != 0) { 659 /* 660 * There was a collision (multiple waiters): we must 661 * inform all potentially interested waiters. 662 */ 663 sip->sel_collision = 0; 664 do { 665 index = ffs(mask) - 1; 666 mask &= ~(1 << index); 667 sc = cpu_lookup(index)->ci_data.cpu_selcpu; 668 mutex_spin_enter(&sc->sc_lock); 669 sc->sc_ncoll++; 670 sleepq_wake(&sc->sc_sleepq, sc, (u_int)-1, 671 &sc->sc_lock); 672 } while (__predict_false(mask != 0)); 673 } 674 } 675 676 /* 677 * Remove an LWP from all objects that it is waiting for. Concurrency 678 * issues: 679 * 680 * The object owner's (e.g. device driver) lock is not held here. Calls 681 * can be made to selrecord() and we do not synchronize against those 682 * directly using locks. However, we use `sel_lwp' to lock out changes. 683 * Before clearing it we must use memory barriers to ensure that we can 684 * safely traverse the list of selinfo records. 685 */ 686 static void 687 selclear(void) 688 { 689 struct selinfo *sip, *next; 690 selcpu_t *sc; 691 lwp_t *l; 692 693 l = curlwp; 694 sc = l->l_selcpu; 695 696 mutex_spin_enter(&sc->sc_lock); 697 for (sip = SLIST_FIRST(&l->l_selwait); sip != NULL; sip = next) { 698 KASSERT(sip->sel_lwp == l); 699 KASSERT(sip->sel_cpu == l->l_selcpu); 700 /* 701 * Read link to next selinfo record, if any. 702 * It's no longer safe to touch `sip' after clearing 703 * `sel_lwp', so ensure that the read of `sel_chain' 704 * completes before the clearing of sel_lwp becomes 705 * globally visible. 706 */ 707 next = SLIST_NEXT(sip, sel_chain); 708 membar_exit(); 709 /* Release the record for another named waiter to use. */ 710 sip->sel_lwp = NULL; 711 } 712 mutex_spin_exit(&sc->sc_lock); 713 } 714 715 /* 716 * Initialize the select/poll system calls. Called once for each 717 * CPU in the system, as they are attached. 718 */ 719 void 720 selsysinit(struct cpu_info *ci) 721 { 722 selcpu_t *sc; 723 724 sc = kmem_alloc(roundup2(sizeof(selcpu_t), coherency_unit) + 725 coherency_unit, KM_SLEEP); 726 sc = (void *)roundup2((uintptr_t)sc, coherency_unit); 727 mutex_init(&sc->sc_lock, MUTEX_DEFAULT, IPL_SCHED); 728 sleepq_init(&sc->sc_sleepq); 729 sc->sc_ncoll = 0; 730 sc->sc_mask = (1 << cpu_index(ci)); 731 ci->ci_data.cpu_selcpu = sc; 732 } 733 734 /* 735 * Initialize a selinfo record. 736 */ 737 void 738 selinit(struct selinfo *sip) 739 { 740 741 memset(sip, 0, sizeof(*sip)); 742 } 743 744 /* 745 * Destroy a selinfo record. The owning object must not gain new 746 * references while this is in progress: all activity on the record 747 * must be stopped. 748 * 749 * Concurrency issues: we only need guard against a call to selclear() 750 * by a thread exiting selcommon() and/or pollcommon(). The caller has 751 * prevented further references being made to the selinfo record via 752 * selrecord(), and it won't call selwakeup() again. 753 */ 754 void 755 seldestroy(struct selinfo *sip) 756 { 757 selcpu_t *sc; 758 lwp_t *l; 759 760 if (sip->sel_lwp == NULL) 761 return; 762 763 /* 764 * Lock out selclear(). The selcpu pointer can't change while 765 * we are here since it is only ever changed in selrecord(), 766 * and that will not be entered again for this record because 767 * it is dying. 768 */ 769 KASSERT(sip->sel_cpu != NULL); 770 sc = sip->sel_cpu; 771 mutex_spin_enter(&sc->sc_lock); 772 if ((l = sip->sel_lwp) != NULL) { 773 /* 774 * This should rarely happen, so although SLIST_REMOVE() 775 * is slow, using it here is not a problem. 776 */ 777 KASSERT(l->l_selcpu == sc); 778 SLIST_REMOVE(&l->l_selwait, sip, selinfo, sel_chain); 779 sip->sel_lwp = NULL; 780 } 781 mutex_spin_exit(&sc->sc_lock); 782 } 783 784 int 785 pollsock(struct socket *so, const struct timeval *tvp, int events) 786 { 787 int ncoll, error, timo; 788 struct timeval sleeptv, tv; 789 selcpu_t *sc; 790 lwp_t *l; 791 792 timo = 0; 793 if (tvp != NULL) { 794 tv = *tvp; 795 if (inittimeleft(&tv, &sleeptv) == -1) 796 return EINVAL; 797 } 798 799 l = curlwp; 800 sc = l->l_cpu->ci_data.cpu_selcpu; 801 l->l_selcpu = sc; 802 SLIST_INIT(&l->l_selwait); 803 error = 0; 804 for (;;) { 805 /* 806 * No need to lock. If this is overwritten by another 807 * value while scanning, we will retry below. We only 808 * need to see exact state from the descriptors that 809 * we are about to poll, and lock activity resulting 810 * from fo_poll is enough to provide an up to date value 811 * for new polling activity. 812 */ 813 ncoll = sc->sc_ncoll; 814 l->l_selflag = SEL_SCANNING; 815 if (sopoll(so, events) != 0) 816 break; 817 if (tvp && (timo = gettimeleft(&tv, &sleeptv)) <= 0) 818 break; 819 mutex_spin_enter(&sc->sc_lock); 820 if (l->l_selflag != SEL_SCANNING || sc->sc_ncoll != ncoll) { 821 mutex_spin_exit(&sc->sc_lock); 822 continue; 823 } 824 l->l_selflag = SEL_BLOCKING; 825 sleepq_enter(&sc->sc_sleepq, l, &sc->sc_lock); 826 sleepq_enqueue(&sc->sc_sleepq, sc, "pollsock", &select_sobj); 827 error = sleepq_block(timo, true); 828 if (error != 0) 829 break; 830 } 831 selclear(); 832 /* poll is not restarted after signals... */ 833 if (error == ERESTART) 834 error = EINTR; 835 if (error == EWOULDBLOCK) 836 error = 0; 837 return (error); 838 } 839