1 /* $NetBSD: sys_select.c,v 1.7 2008/04/30 12:45:21 ad 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 1.7 2008/04/30 12:45:21 ad 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_pselect(struct lwp *l, const struct sys_pselect_args *uap, register_t *retval) 128 { 129 /* { 130 syscallarg(int) nd; 131 syscallarg(fd_set *) in; 132 syscallarg(fd_set *) ou; 133 syscallarg(fd_set *) ex; 134 syscallarg(const struct timespec *) ts; 135 syscallarg(sigset_t *) mask; 136 } */ 137 struct timespec ats; 138 struct timeval atv, *tv = NULL; 139 sigset_t amask, *mask = NULL; 140 int error; 141 142 if (SCARG(uap, ts)) { 143 error = copyin(SCARG(uap, ts), &ats, sizeof(ats)); 144 if (error) 145 return error; 146 atv.tv_sec = ats.tv_sec; 147 atv.tv_usec = ats.tv_nsec / 1000; 148 tv = &atv; 149 } 150 if (SCARG(uap, mask) != NULL) { 151 error = copyin(SCARG(uap, mask), &amask, sizeof(amask)); 152 if (error) 153 return error; 154 mask = &amask; 155 } 156 157 return selcommon(l, retval, SCARG(uap, nd), SCARG(uap, in), 158 SCARG(uap, ou), SCARG(uap, ex), tv, mask); 159 } 160 161 int 162 inittimeleft(struct timeval *tv, struct timeval *sleeptv) 163 { 164 if (itimerfix(tv)) 165 return -1; 166 getmicrouptime(sleeptv); 167 return 0; 168 } 169 170 int 171 gettimeleft(struct timeval *tv, struct timeval *sleeptv) 172 { 173 /* 174 * We have to recalculate the timeout on every retry. 175 */ 176 struct timeval slepttv; 177 /* 178 * reduce tv by elapsed time 179 * based on monotonic time scale 180 */ 181 getmicrouptime(&slepttv); 182 timeradd(tv, sleeptv, tv); 183 timersub(tv, &slepttv, tv); 184 *sleeptv = slepttv; 185 return tvtohz(tv); 186 } 187 188 int 189 sys_select(struct lwp *l, const struct sys_select_args *uap, register_t *retval) 190 { 191 /* { 192 syscallarg(int) nd; 193 syscallarg(fd_set *) in; 194 syscallarg(fd_set *) ou; 195 syscallarg(fd_set *) ex; 196 syscallarg(struct timeval *) tv; 197 } */ 198 struct timeval atv, *tv = NULL; 199 int error; 200 201 if (SCARG(uap, tv)) { 202 error = copyin(SCARG(uap, tv), (void *)&atv, 203 sizeof(atv)); 204 if (error) 205 return error; 206 tv = &atv; 207 } 208 209 return selcommon(l, retval, SCARG(uap, nd), SCARG(uap, in), 210 SCARG(uap, ou), SCARG(uap, ex), tv, NULL); 211 } 212 213 int 214 selcommon(lwp_t *l, register_t *retval, int nd, fd_set *u_in, 215 fd_set *u_ou, fd_set *u_ex, struct timeval *tv, sigset_t *mask) 216 { 217 char smallbits[howmany(FD_SETSIZE, NFDBITS) * 218 sizeof(fd_mask) * 6]; 219 proc_t * const p = l->l_proc; 220 char *bits; 221 int ncoll, error, timo; 222 size_t ni; 223 sigset_t oldmask; 224 struct timeval sleeptv; 225 selcpu_t *sc; 226 227 error = 0; 228 if (nd < 0) 229 return (EINVAL); 230 if (nd > p->p_fd->fd_nfiles) { 231 /* forgiving; slightly wrong */ 232 nd = p->p_fd->fd_nfiles; 233 } 234 ni = howmany(nd, NFDBITS) * sizeof(fd_mask); 235 if (ni * 6 > sizeof(smallbits)) 236 bits = kmem_alloc(ni * 6, KM_SLEEP); 237 else 238 bits = smallbits; 239 240 #define getbits(name, x) \ 241 if (u_ ## name) { \ 242 error = copyin(u_ ## name, bits + ni * x, ni); \ 243 if (error) \ 244 goto done; \ 245 } else \ 246 memset(bits + ni * x, 0, ni); 247 getbits(in, 0); 248 getbits(ou, 1); 249 getbits(ex, 2); 250 #undef getbits 251 252 timo = 0; 253 if (tv && inittimeleft(tv, &sleeptv) == -1) { 254 error = EINVAL; 255 goto done; 256 } 257 258 if (mask) { 259 sigminusset(&sigcantmask, mask); 260 mutex_enter(p->p_lock); 261 oldmask = l->l_sigmask; 262 l->l_sigmask = *mask; 263 mutex_exit(p->p_lock); 264 } else 265 oldmask = l->l_sigmask; /* XXXgcc */ 266 267 sc = curcpu()->ci_data.cpu_selcpu; 268 l->l_selcpu = sc; 269 SLIST_INIT(&l->l_selwait); 270 for (;;) { 271 /* 272 * No need to lock. If this is overwritten by another 273 * value while scanning, we will retry below. We only 274 * need to see exact state from the descriptors that 275 * we are about to poll, and lock activity resulting 276 * from fo_poll is enough to provide an up to date value 277 * for new polling activity. 278 */ 279 l->l_selflag = SEL_SCANNING; 280 ncoll = sc->sc_ncoll; 281 282 error = selscan(l, (fd_mask *)(bits + ni * 0), 283 (fd_mask *)(bits + ni * 3), nd, retval); 284 285 if (error || *retval) 286 break; 287 if (tv && (timo = gettimeleft(tv, &sleeptv)) <= 0) 288 break; 289 mutex_spin_enter(&sc->sc_lock); 290 if (l->l_selflag != SEL_SCANNING || sc->sc_ncoll != ncoll) { 291 mutex_spin_exit(&sc->sc_lock); 292 continue; 293 } 294 l->l_selflag = SEL_BLOCKING; 295 l->l_kpriority = true; 296 lwp_lock(l); 297 lwp_unlock_to(l, &sc->sc_lock); 298 sleepq_enqueue(&sc->sc_sleepq, sc, "select", &select_sobj); 299 KERNEL_UNLOCK_ALL(NULL, &l->l_biglocks); /* XXX */ 300 error = sleepq_block(timo, true); 301 if (error != 0) 302 break; 303 } 304 selclear(); 305 306 if (mask) { 307 mutex_enter(p->p_lock); 308 l->l_sigmask = oldmask; 309 mutex_exit(p->p_lock); 310 } 311 312 done: 313 /* select is not restarted after signals... */ 314 if (error == ERESTART) 315 error = EINTR; 316 if (error == EWOULDBLOCK) 317 error = 0; 318 if (error == 0 && u_in != NULL) 319 error = copyout(bits + ni * 3, u_in, ni); 320 if (error == 0 && u_ou != NULL) 321 error = copyout(bits + ni * 4, u_ou, ni); 322 if (error == 0 && u_ex != NULL) 323 error = copyout(bits + ni * 5, u_ex, ni); 324 if (bits != smallbits) 325 kmem_free(bits, ni * 6); 326 return (error); 327 } 328 329 int 330 selscan(lwp_t *l, fd_mask *ibitp, fd_mask *obitp, int nfd, 331 register_t *retval) 332 { 333 static const int flag[3] = { POLLRDNORM | POLLHUP | POLLERR, 334 POLLWRNORM | POLLHUP | POLLERR, 335 POLLRDBAND }; 336 int msk, i, j, fd, n; 337 fd_mask ibits, obits; 338 file_t *fp; 339 340 n = 0; 341 for (msk = 0; msk < 3; msk++) { 342 for (i = 0; i < nfd; i += NFDBITS) { 343 ibits = *ibitp++; 344 obits = 0; 345 while ((j = ffs(ibits)) && (fd = i + --j) < nfd) { 346 ibits &= ~(1 << j); 347 if ((fp = fd_getfile(fd)) == NULL) 348 return (EBADF); 349 if ((*fp->f_ops->fo_poll)(fp, flag[msk])) { 350 obits |= (1 << j); 351 n++; 352 } 353 fd_putfile(fd); 354 } 355 *obitp++ = obits; 356 } 357 } 358 *retval = n; 359 return (0); 360 } 361 362 /* 363 * Poll system call. 364 */ 365 int 366 sys_poll(struct lwp *l, const struct sys_poll_args *uap, register_t *retval) 367 { 368 /* { 369 syscallarg(struct pollfd *) fds; 370 syscallarg(u_int) nfds; 371 syscallarg(int) timeout; 372 } */ 373 struct timeval atv, *tv = NULL; 374 375 if (SCARG(uap, timeout) != INFTIM) { 376 atv.tv_sec = SCARG(uap, timeout) / 1000; 377 atv.tv_usec = (SCARG(uap, timeout) % 1000) * 1000; 378 tv = &atv; 379 } 380 381 return pollcommon(l, retval, SCARG(uap, fds), SCARG(uap, nfds), 382 tv, NULL); 383 } 384 385 /* 386 * Poll system call. 387 */ 388 int 389 sys_pollts(struct lwp *l, const struct sys_pollts_args *uap, register_t *retval) 390 { 391 /* { 392 syscallarg(struct pollfd *) fds; 393 syscallarg(u_int) nfds; 394 syscallarg(const struct timespec *) ts; 395 syscallarg(const sigset_t *) mask; 396 } */ 397 struct timespec ats; 398 struct timeval atv, *tv = NULL; 399 sigset_t amask, *mask = NULL; 400 int error; 401 402 if (SCARG(uap, ts)) { 403 error = copyin(SCARG(uap, ts), &ats, sizeof(ats)); 404 if (error) 405 return error; 406 atv.tv_sec = ats.tv_sec; 407 atv.tv_usec = ats.tv_nsec / 1000; 408 tv = &atv; 409 } 410 if (SCARG(uap, mask)) { 411 error = copyin(SCARG(uap, mask), &amask, sizeof(amask)); 412 if (error) 413 return error; 414 mask = &amask; 415 } 416 417 return pollcommon(l, retval, SCARG(uap, fds), SCARG(uap, nfds), 418 tv, mask); 419 } 420 421 int 422 pollcommon(lwp_t *l, register_t *retval, 423 struct pollfd *u_fds, u_int nfds, 424 struct timeval *tv, sigset_t *mask) 425 { 426 char smallbits[32 * sizeof(struct pollfd)]; 427 proc_t * const p = l->l_proc; 428 void * bits; 429 sigset_t oldmask; 430 int ncoll, error, timo; 431 size_t ni; 432 struct timeval sleeptv; 433 selcpu_t *sc; 434 435 if (nfds > p->p_fd->fd_nfiles) { 436 /* forgiving; slightly wrong */ 437 nfds = p->p_fd->fd_nfiles; 438 } 439 ni = nfds * sizeof(struct pollfd); 440 if (ni > sizeof(smallbits)) 441 bits = kmem_alloc(ni, KM_SLEEP); 442 else 443 bits = smallbits; 444 445 error = copyin(u_fds, bits, ni); 446 if (error) 447 goto done; 448 449 timo = 0; 450 if (tv && inittimeleft(tv, &sleeptv) == -1) { 451 error = EINVAL; 452 goto done; 453 } 454 455 if (mask) { 456 sigminusset(&sigcantmask, mask); 457 mutex_enter(p->p_lock); 458 oldmask = l->l_sigmask; 459 l->l_sigmask = *mask; 460 mutex_exit(p->p_lock); 461 } else 462 oldmask = l->l_sigmask; /* XXXgcc */ 463 464 sc = curcpu()->ci_data.cpu_selcpu; 465 l->l_selcpu = sc; 466 SLIST_INIT(&l->l_selwait); 467 for (;;) { 468 /* 469 * No need to lock. If this is overwritten by another 470 * value while scanning, we will retry below. We only 471 * need to see exact state from the descriptors that 472 * we are about to poll, and lock activity resulting 473 * from fo_poll is enough to provide an up to date value 474 * for new polling activity. 475 */ 476 ncoll = sc->sc_ncoll; 477 l->l_selflag = SEL_SCANNING; 478 479 error = pollscan(l, (struct pollfd *)bits, nfds, retval); 480 481 if (error || *retval) 482 break; 483 if (tv && (timo = gettimeleft(tv, &sleeptv)) <= 0) 484 break; 485 mutex_spin_enter(&sc->sc_lock); 486 if (l->l_selflag != SEL_SCANNING || sc->sc_ncoll != ncoll) { 487 mutex_spin_exit(&sc->sc_lock); 488 continue; 489 } 490 l->l_selflag = SEL_BLOCKING; 491 l->l_kpriority = true; 492 lwp_lock(l); 493 lwp_unlock_to(l, &sc->sc_lock); 494 sleepq_enqueue(&sc->sc_sleepq, sc, "select", &select_sobj); 495 KERNEL_UNLOCK_ALL(NULL, &l->l_biglocks); /* XXX */ 496 error = sleepq_block(timo, true); 497 if (error != 0) 498 break; 499 } 500 selclear(); 501 502 if (mask) { 503 mutex_enter(p->p_lock); 504 l->l_sigmask = oldmask; 505 mutex_exit(p->p_lock); 506 } 507 done: 508 /* poll is not restarted after signals... */ 509 if (error == ERESTART) 510 error = EINTR; 511 if (error == EWOULDBLOCK) 512 error = 0; 513 if (error == 0) 514 error = copyout(bits, u_fds, ni); 515 if (bits != smallbits) 516 kmem_free(bits, ni); 517 return (error); 518 } 519 520 int 521 pollscan(lwp_t *l, struct pollfd *fds, int nfd, register_t *retval) 522 { 523 int i, n; 524 file_t *fp; 525 526 n = 0; 527 for (i = 0; i < nfd; i++, fds++) { 528 if (fds->fd < 0) { 529 fds->revents = 0; 530 } else if ((fp = fd_getfile(fds->fd)) == NULL) { 531 fds->revents = POLLNVAL; 532 n++; 533 } else { 534 fds->revents = (*fp->f_ops->fo_poll)(fp, 535 fds->events | POLLERR | POLLHUP); 536 if (fds->revents != 0) 537 n++; 538 fd_putfile(fds->fd); 539 } 540 } 541 *retval = n; 542 return (0); 543 } 544 545 /*ARGSUSED*/ 546 int 547 seltrue(dev_t dev, int events, lwp_t *l) 548 { 549 550 return (events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM)); 551 } 552 553 /* 554 * Record a select request. Concurrency issues: 555 * 556 * The caller holds the same lock across calls to selrecord() and 557 * selnotify(), so we don't need to consider a concurrent wakeup 558 * while in this routine. 559 * 560 * The only activity we need to guard against is selclear(), called by 561 * another thread that is exiting selcommon() or pollcommon(). 562 * `sel_lwp' can only become non-NULL while the caller's lock is held, 563 * so it cannot become non-NULL due to a change made by another thread 564 * while we are in this routine. It can only become _NULL_ due to a 565 * call to selclear(). 566 * 567 * If it is non-NULL and != selector there is the potential for 568 * selclear() to be called by another thread. If either of those 569 * conditions are true, we're not interested in touching the `named 570 * waiter' part of the selinfo record because we need to record a 571 * collision. Hence there is no need for additional locking in this 572 * routine. 573 */ 574 void 575 selrecord(lwp_t *selector, struct selinfo *sip) 576 { 577 selcpu_t *sc; 578 lwp_t *other; 579 580 KASSERT(selector == curlwp); 581 582 sc = selector->l_selcpu; 583 other = sip->sel_lwp; 584 585 if (other == selector) { 586 /* `selector' has already claimed it. */ 587 KASSERT(sip->sel_cpu = sc); 588 } else if (other == NULL) { 589 /* 590 * First named waiter, although there may be unnamed 591 * waiters (collisions). Issue a memory barrier to 592 * ensure that we access sel_lwp (above) before other 593 * fields - this guards against a call to selclear(). 594 */ 595 membar_enter(); 596 sip->sel_lwp = selector; 597 SLIST_INSERT_HEAD(&selector->l_selwait, sip, sel_chain); 598 /* Replace selinfo's lock with our chosen CPU's lock. */ 599 sip->sel_cpu = sc; 600 } else { 601 /* Multiple waiters: record a collision. */ 602 sip->sel_collision |= sc->sc_mask; 603 KASSERT(sip->sel_cpu != NULL); 604 } 605 } 606 607 /* 608 * Do a wakeup when a selectable event occurs. Concurrency issues: 609 * 610 * As per selrecord(), the caller's object lock is held. If there 611 * is a named waiter, we must acquire the associated selcpu's lock 612 * in order to synchronize with selclear() and pollers going to sleep 613 * in selcommon() and/or pollcommon(). 614 * 615 * sip->sel_cpu cannot change at this point, as it is only changed 616 * in selrecord(), and concurrent calls to selrecord() are locked 617 * out by the caller. 618 */ 619 void 620 selnotify(struct selinfo *sip, int events, long knhint) 621 { 622 selcpu_t *sc; 623 uint32_t mask; 624 int index, oflag, swapin; 625 lwp_t *l; 626 627 KNOTE(&sip->sel_klist, knhint); 628 629 if (sip->sel_lwp != NULL) { 630 /* One named LWP is waiting. */ 631 swapin = 0; 632 sc = sip->sel_cpu; 633 mutex_spin_enter(&sc->sc_lock); 634 /* Still there? */ 635 if (sip->sel_lwp != NULL) { 636 l = sip->sel_lwp; 637 /* 638 * If thread is sleeping, wake it up. If it's not 639 * yet asleep, it will notice the change in state 640 * and will re-poll the descriptors. 641 */ 642 oflag = l->l_selflag; 643 l->l_selflag = SEL_RESET; 644 if (oflag == SEL_BLOCKING && 645 l->l_mutex == &sc->sc_lock) { 646 KASSERT(l->l_wchan == sc); 647 swapin = sleepq_unsleep(l, false); 648 } 649 } 650 mutex_spin_exit(&sc->sc_lock); 651 if (swapin) 652 uvm_kick_scheduler(); 653 } 654 655 if ((mask = sip->sel_collision) != 0) { 656 /* 657 * There was a collision (multiple waiters): we must 658 * inform all potentially interested waiters. 659 */ 660 sip->sel_collision = 0; 661 do { 662 index = ffs(mask) - 1; 663 mask &= ~(1 << index); 664 sc = cpu_lookup_byindex(index)->ci_data.cpu_selcpu; 665 mutex_spin_enter(&sc->sc_lock); 666 sc->sc_ncoll++; 667 sleepq_wake(&sc->sc_sleepq, sc, (u_int)-1); 668 } while (__predict_false(mask != 0)); 669 } 670 } 671 672 /* 673 * Remove an LWP from all objects that it is waiting for. Concurrency 674 * issues: 675 * 676 * The object owner's (e.g. device driver) lock is not held here. Calls 677 * can be made to selrecord() and we do not synchronize against those 678 * directly using locks. However, we use `sel_lwp' to lock out changes. 679 * Before clearing it we must use memory barriers to ensure that we can 680 * safely traverse the list of selinfo records. 681 */ 682 static void 683 selclear(void) 684 { 685 struct selinfo *sip, *next; 686 selcpu_t *sc; 687 lwp_t *l; 688 689 l = curlwp; 690 sc = l->l_selcpu; 691 692 mutex_spin_enter(&sc->sc_lock); 693 for (sip = SLIST_FIRST(&l->l_selwait); sip != NULL; sip = next) { 694 KASSERT(sip->sel_lwp == l); 695 KASSERT(sip->sel_cpu == l->l_selcpu); 696 /* 697 * Read link to next selinfo record, if any. 698 * It's no longer safe to touch `sip' after clearing 699 * `sel_lwp', so ensure that the read of `sel_chain' 700 * completes before the clearing of sel_lwp becomes 701 * globally visible. 702 */ 703 next = SLIST_NEXT(sip, sel_chain); 704 membar_exit(); 705 /* Release the record for another named waiter to use. */ 706 sip->sel_lwp = NULL; 707 } 708 mutex_spin_exit(&sc->sc_lock); 709 } 710 711 /* 712 * Initialize the select/poll system calls. Called once for each 713 * CPU in the system, as they are attached. 714 */ 715 void 716 selsysinit(struct cpu_info *ci) 717 { 718 selcpu_t *sc; 719 720 sc = kmem_alloc(roundup2(sizeof(selcpu_t), coherency_unit) + 721 coherency_unit, KM_SLEEP); 722 sc = (void *)roundup2((uintptr_t)sc, coherency_unit); 723 mutex_init(&sc->sc_lock, MUTEX_DEFAULT, IPL_SCHED); 724 sleepq_init(&sc->sc_sleepq, &sc->sc_lock); 725 sc->sc_ncoll = 0; 726 sc->sc_mask = (1 << cpu_index(ci)); 727 ci->ci_data.cpu_selcpu = sc; 728 } 729 730 /* 731 * Initialize a selinfo record. 732 */ 733 void 734 selinit(struct selinfo *sip) 735 { 736 737 memset(sip, 0, sizeof(*sip)); 738 } 739 740 /* 741 * Destroy a selinfo record. The owning object must not gain new 742 * references while this is in progress: all activity on the record 743 * must be stopped. 744 * 745 * Concurrency issues: we only need guard against a call to selclear() 746 * by a thread exiting selcommon() and/or pollcommon(). The caller has 747 * prevented further references being made to the selinfo record via 748 * selrecord(), and it won't call selwakeup() again. 749 */ 750 void 751 seldestroy(struct selinfo *sip) 752 { 753 selcpu_t *sc; 754 lwp_t *l; 755 756 if (sip->sel_lwp == NULL) 757 return; 758 759 /* 760 * Lock out selclear(). The selcpu pointer can't change while 761 * we are here since it is only ever changed in selrecord(), 762 * and that will not be entered again for this record because 763 * it is dying. 764 */ 765 KASSERT(sip->sel_cpu != NULL); 766 sc = sip->sel_cpu; 767 mutex_spin_enter(&sc->sc_lock); 768 if ((l = sip->sel_lwp) != NULL) { 769 /* 770 * This should rarely happen, so although SLIST_REMOVE() 771 * is slow, using it here is not a problem. 772 */ 773 KASSERT(l->l_selcpu == sc); 774 SLIST_REMOVE(&l->l_selwait, sip, selinfo, sel_chain); 775 sip->sel_lwp = NULL; 776 } 777 mutex_spin_exit(&sc->sc_lock); 778 } 779 780 int 781 pollsock(struct socket *so, const struct timeval *tvp, int events) 782 { 783 int ncoll, error, timo; 784 struct timeval sleeptv, tv; 785 selcpu_t *sc; 786 lwp_t *l; 787 788 timo = 0; 789 if (tvp != NULL) { 790 tv = *tvp; 791 if (inittimeleft(&tv, &sleeptv) == -1) 792 return EINVAL; 793 } 794 795 l = curlwp; 796 sc = l->l_cpu->ci_data.cpu_selcpu; 797 l->l_selcpu = sc; 798 SLIST_INIT(&l->l_selwait); 799 error = 0; 800 for (;;) { 801 /* 802 * No need to lock. If this is overwritten by another 803 * value while scanning, we will retry below. We only 804 * need to see exact state from the descriptors that 805 * we are about to poll, and lock activity resulting 806 * from fo_poll is enough to provide an up to date value 807 * for new polling activity. 808 */ 809 ncoll = sc->sc_ncoll; 810 l->l_selflag = SEL_SCANNING; 811 if (sopoll(so, events) != 0) 812 break; 813 if (tvp && (timo = gettimeleft(&tv, &sleeptv)) <= 0) 814 break; 815 mutex_spin_enter(&sc->sc_lock); 816 if (l->l_selflag != SEL_SCANNING || sc->sc_ncoll != ncoll) { 817 mutex_spin_exit(&sc->sc_lock); 818 continue; 819 } 820 l->l_selflag = SEL_BLOCKING; 821 lwp_lock(l); 822 lwp_unlock_to(l, &sc->sc_lock); 823 sleepq_enqueue(&sc->sc_sleepq, sc, "pollsock", &select_sobj); 824 KERNEL_UNLOCK_ALL(NULL, &l->l_biglocks); /* XXX */ 825 error = sleepq_block(timo, true); 826 if (error != 0) 827 break; 828 } 829 selclear(); 830 /* poll is not restarted after signals... */ 831 if (error == ERESTART) 832 error = EINTR; 833 if (error == EWOULDBLOCK) 834 error = 0; 835 return (error); 836 } 837