sys_generic.c revision 1.54.2.10
1 /* $NetBSD: sys_generic.c,v 1.54.2.10 2002/08/27 23:47:32 nathanw Exp $ */ 2 3 /* 4 * Copyright (c) 1982, 1986, 1989, 1993 5 * The Regents of the University of California. All rights reserved. 6 * (c) UNIX System Laboratories, Inc. 7 * All or some portions of this file are derived from material licensed 8 * to the University of California by American Telephone and Telegraph 9 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 10 * the permission of UNIX System Laboratories, Inc. 11 * 12 * Redistribution and use in source and binary forms, with or without 13 * modification, are permitted provided that the following conditions 14 * are met: 15 * 1. Redistributions of source code must retain the above copyright 16 * notice, this list of conditions and the following disclaimer. 17 * 2. Redistributions in binary form must reproduce the above copyright 18 * notice, this list of conditions and the following disclaimer in the 19 * documentation and/or other materials provided with the distribution. 20 * 3. All advertising materials mentioning features or use of this software 21 * must display the following acknowledgement: 22 * This product includes software developed by the University of 23 * California, Berkeley and its contributors. 24 * 4. Neither the name of the University nor the names of its contributors 25 * may be used to endorse or promote products derived from this software 26 * without specific prior written permission. 27 * 28 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 29 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 30 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 31 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 32 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 33 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 34 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 35 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 36 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 37 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 38 * SUCH DAMAGE. 39 * 40 * @(#)sys_generic.c 8.9 (Berkeley) 2/14/95 41 */ 42 43 #include <sys/cdefs.h> 44 __KERNEL_RCSID(0, "$NetBSD: sys_generic.c,v 1.54.2.10 2002/08/27 23:47:32 nathanw Exp $"); 45 46 #include "opt_ktrace.h" 47 48 #include <sys/param.h> 49 #include <sys/systm.h> 50 #include <sys/filedesc.h> 51 #include <sys/ioctl.h> 52 #include <sys/file.h> 53 #include <sys/proc.h> 54 #include <sys/socketvar.h> 55 #include <sys/signalvar.h> 56 #include <sys/uio.h> 57 #include <sys/kernel.h> 58 #include <sys/stat.h> 59 #include <sys/malloc.h> 60 #include <sys/poll.h> 61 #ifdef KTRACE 62 #include <sys/ktrace.h> 63 #endif 64 65 #include <sys/mount.h> 66 #include <sys/sa.h> 67 #include <sys/syscallargs.h> 68 69 int selscan __P((struct proc *, fd_mask *, fd_mask *, int, register_t *)); 70 int pollscan __P((struct proc *, struct pollfd *, int, register_t *)); 71 72 /* 73 * Read system call. 74 */ 75 /* ARGSUSED */ 76 int 77 sys_read(struct lwp *l, void *v, register_t *retval) 78 { 79 struct sys_read_args /* { 80 syscallarg(int) fd; 81 syscallarg(void *) buf; 82 syscallarg(size_t) nbyte; 83 } */ *uap = v; 84 int fd; 85 struct file *fp; 86 struct proc *p; 87 struct filedesc *fdp; 88 89 fd = SCARG(uap, fd); 90 p = l->l_proc; 91 fdp = p->p_fd; 92 93 if ((fp = fd_getfile(fdp, fd)) == NULL) 94 return (EBADF); 95 96 if ((fp->f_flag & FREAD) == 0) 97 return (EBADF); 98 99 FILE_USE(fp); 100 101 /* dofileread() will unuse the descriptor for us */ 102 return (dofileread(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 103 &fp->f_offset, FOF_UPDATE_OFFSET, retval)); 104 } 105 106 int 107 dofileread(struct proc *p, int fd, struct file *fp, void *buf, size_t nbyte, 108 off_t *offset, int flags, register_t *retval) 109 { 110 struct uio auio; 111 struct iovec aiov; 112 size_t cnt; 113 int error; 114 #ifdef KTRACE 115 struct iovec ktriov; 116 #endif 117 error = 0; 118 119 aiov.iov_base = (caddr_t)buf; 120 aiov.iov_len = nbyte; 121 auio.uio_iov = &aiov; 122 auio.uio_iovcnt = 1; 123 auio.uio_resid = nbyte; 124 auio.uio_rw = UIO_READ; 125 auio.uio_segflg = UIO_USERSPACE; 126 auio.uio_procp = p; 127 128 /* 129 * Reads return ssize_t because -1 is returned on error. Therefore 130 * we must restrict the length to SSIZE_MAX to avoid garbage return 131 * values. 132 */ 133 if (auio.uio_resid > SSIZE_MAX) { 134 error = EINVAL; 135 goto out; 136 } 137 138 #ifdef KTRACE 139 /* 140 * if tracing, save a copy of iovec 141 */ 142 if (KTRPOINT(p, KTR_GENIO)) 143 ktriov = aiov; 144 #endif 145 cnt = auio.uio_resid; 146 error = (*fp->f_ops->fo_read)(fp, offset, &auio, fp->f_cred, flags); 147 if (error) 148 if (auio.uio_resid != cnt && (error == ERESTART || 149 error == EINTR || error == EWOULDBLOCK)) 150 error = 0; 151 cnt -= auio.uio_resid; 152 #ifdef KTRACE 153 if (KTRPOINT(p, KTR_GENIO) && error == 0) 154 ktrgenio(p, fd, UIO_READ, &ktriov, cnt, error); 155 #endif 156 *retval = cnt; 157 out: 158 FILE_UNUSE(fp, p); 159 return (error); 160 } 161 162 /* 163 * Scatter read system call. 164 */ 165 int 166 sys_readv(struct lwp *l, void *v, register_t *retval) 167 { 168 struct sys_readv_args /* { 169 syscallarg(int) fd; 170 syscallarg(const struct iovec *) iovp; 171 syscallarg(int) iovcnt; 172 } */ *uap = v; 173 int fd; 174 struct file *fp; 175 struct proc *p; 176 struct filedesc *fdp; 177 178 fd = SCARG(uap, fd); 179 p = l->l_proc; 180 fdp = p->p_fd; 181 182 if ((fp = fd_getfile(fdp, fd)) == NULL) 183 return (EBADF); 184 185 if ((fp->f_flag & FREAD) == 0) 186 return (EBADF); 187 188 FILE_USE(fp); 189 190 /* dofilereadv() will unuse the descriptor for us */ 191 return (dofilereadv(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 192 &fp->f_offset, FOF_UPDATE_OFFSET, retval)); 193 } 194 195 int 196 dofilereadv(struct proc *p, int fd, struct file *fp, const struct iovec *iovp, 197 int iovcnt, off_t *offset, int flags, register_t *retval) 198 { 199 struct uio auio; 200 struct iovec *iov, *needfree, aiov[UIO_SMALLIOV]; 201 int i, error; 202 size_t cnt; 203 u_int iovlen; 204 #ifdef KTRACE 205 struct iovec *ktriov; 206 #endif 207 208 error = 0; 209 #ifdef KTRACE 210 ktriov = NULL; 211 #endif 212 /* note: can't use iovlen until iovcnt is validated */ 213 iovlen = iovcnt * sizeof(struct iovec); 214 if ((u_int)iovcnt > UIO_SMALLIOV) { 215 if ((u_int)iovcnt > IOV_MAX) { 216 error = EINVAL; 217 goto out; 218 } 219 iov = malloc(iovlen, M_IOV, M_WAITOK); 220 needfree = iov; 221 } else if ((u_int)iovcnt > 0) { 222 iov = aiov; 223 needfree = NULL; 224 } else { 225 error = EINVAL; 226 goto out; 227 } 228 229 auio.uio_iov = iov; 230 auio.uio_iovcnt = iovcnt; 231 auio.uio_rw = UIO_READ; 232 auio.uio_segflg = UIO_USERSPACE; 233 auio.uio_procp = p; 234 error = copyin(iovp, iov, iovlen); 235 if (error) 236 goto done; 237 auio.uio_resid = 0; 238 for (i = 0; i < iovcnt; i++) { 239 auio.uio_resid += iov->iov_len; 240 /* 241 * Reads return ssize_t because -1 is returned on error. 242 * Therefore we must restrict the length to SSIZE_MAX to 243 * avoid garbage return values. 244 */ 245 if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) { 246 error = EINVAL; 247 goto done; 248 } 249 iov++; 250 } 251 #ifdef KTRACE 252 /* 253 * if tracing, save a copy of iovec 254 */ 255 if (KTRPOINT(p, KTR_GENIO)) { 256 ktriov = malloc(iovlen, M_TEMP, M_WAITOK); 257 memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen); 258 } 259 #endif 260 cnt = auio.uio_resid; 261 error = (*fp->f_ops->fo_read)(fp, offset, &auio, fp->f_cred, flags); 262 if (error) 263 if (auio.uio_resid != cnt && (error == ERESTART || 264 error == EINTR || error == EWOULDBLOCK)) 265 error = 0; 266 cnt -= auio.uio_resid; 267 #ifdef KTRACE 268 if (ktriov != NULL) { 269 if (error == 0) 270 ktrgenio(p, fd, UIO_READ, ktriov, cnt, error); 271 free(ktriov, M_TEMP); 272 } 273 #endif 274 *retval = cnt; 275 done: 276 if (needfree) 277 free(needfree, M_IOV); 278 out: 279 FILE_UNUSE(fp, p); 280 return (error); 281 } 282 283 /* 284 * Write system call 285 */ 286 int 287 sys_write(struct lwp *l, void *v, register_t *retval) 288 { 289 struct sys_write_args /* { 290 syscallarg(int) fd; 291 syscallarg(const void *) buf; 292 syscallarg(size_t) nbyte; 293 } */ *uap = v; 294 int fd; 295 struct file *fp; 296 struct proc *p; 297 struct filedesc *fdp; 298 299 fd = SCARG(uap, fd); 300 p = l->l_proc; 301 fdp = p->p_fd; 302 303 if ((fp = fd_getfile(fdp, fd)) == NULL) 304 return (EBADF); 305 306 if ((fp->f_flag & FWRITE) == 0) 307 return (EBADF); 308 309 FILE_USE(fp); 310 311 /* dofilewrite() will unuse the descriptor for us */ 312 return (dofilewrite(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 313 &fp->f_offset, FOF_UPDATE_OFFSET, retval)); 314 } 315 316 int 317 dofilewrite(struct proc *p, int fd, struct file *fp, const void *buf, 318 size_t nbyte, off_t *offset, int flags, register_t *retval) 319 { 320 struct uio auio; 321 struct iovec aiov; 322 size_t cnt; 323 int error; 324 #ifdef KTRACE 325 struct iovec ktriov; 326 #endif 327 328 error = 0; 329 aiov.iov_base = (caddr_t)buf; /* XXX kills const */ 330 aiov.iov_len = nbyte; 331 auio.uio_iov = &aiov; 332 auio.uio_iovcnt = 1; 333 auio.uio_resid = nbyte; 334 auio.uio_rw = UIO_WRITE; 335 auio.uio_segflg = UIO_USERSPACE; 336 auio.uio_procp = p; 337 338 /* 339 * Writes return ssize_t because -1 is returned on error. Therefore 340 * we must restrict the length to SSIZE_MAX to avoid garbage return 341 * values. 342 */ 343 if (auio.uio_resid > SSIZE_MAX) { 344 error = EINVAL; 345 goto out; 346 } 347 348 #ifdef KTRACE 349 /* 350 * if tracing, save a copy of iovec 351 */ 352 if (KTRPOINT(p, KTR_GENIO)) 353 ktriov = aiov; 354 #endif 355 cnt = auio.uio_resid; 356 error = (*fp->f_ops->fo_write)(fp, offset, &auio, fp->f_cred, flags); 357 if (error) { 358 if (auio.uio_resid != cnt && (error == ERESTART || 359 error == EINTR || error == EWOULDBLOCK)) 360 error = 0; 361 if (error == EPIPE) 362 psignal(p, SIGPIPE); 363 } 364 cnt -= auio.uio_resid; 365 #ifdef KTRACE 366 if (KTRPOINT(p, KTR_GENIO) && error == 0) 367 ktrgenio(p, fd, UIO_WRITE, &ktriov, cnt, error); 368 #endif 369 *retval = cnt; 370 out: 371 FILE_UNUSE(fp, p); 372 return (error); 373 } 374 375 /* 376 * Gather write system call 377 */ 378 int 379 sys_writev(struct lwp *l, void *v, register_t *retval) 380 { 381 struct sys_writev_args /* { 382 syscallarg(int) fd; 383 syscallarg(const struct iovec *) iovp; 384 syscallarg(int) iovcnt; 385 } */ *uap = v; 386 int fd; 387 struct file *fp; 388 struct proc *p; 389 struct filedesc *fdp; 390 391 fd = SCARG(uap, fd); 392 p = l->l_proc; 393 fdp = p->p_fd; 394 395 if ((fp = fd_getfile(fdp, fd)) == NULL) 396 return (EBADF); 397 398 if ((fp->f_flag & FWRITE) == 0) 399 return (EBADF); 400 401 FILE_USE(fp); 402 403 /* dofilewritev() will unuse the descriptor for us */ 404 return (dofilewritev(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 405 &fp->f_offset, FOF_UPDATE_OFFSET, retval)); 406 } 407 408 int 409 dofilewritev(struct proc *p, int fd, struct file *fp, const struct iovec *iovp, 410 int iovcnt, off_t *offset, int flags, register_t *retval) 411 { 412 struct uio auio; 413 struct iovec *iov, *needfree, aiov[UIO_SMALLIOV]; 414 int i, error; 415 size_t cnt; 416 u_int iovlen; 417 #ifdef KTRACE 418 struct iovec *ktriov; 419 #endif 420 421 error = 0; 422 #ifdef KTRACE 423 ktriov = NULL; 424 #endif 425 /* note: can't use iovlen until iovcnt is validated */ 426 iovlen = iovcnt * sizeof(struct iovec); 427 if ((u_int)iovcnt > UIO_SMALLIOV) { 428 if ((u_int)iovcnt > IOV_MAX) { 429 error = EINVAL; 430 goto out; 431 } 432 iov = malloc(iovlen, M_IOV, M_WAITOK); 433 needfree = iov; 434 } else if ((u_int)iovcnt > 0) { 435 iov = aiov; 436 needfree = NULL; 437 } else { 438 error = EINVAL; 439 goto out; 440 } 441 442 auio.uio_iov = iov; 443 auio.uio_iovcnt = iovcnt; 444 auio.uio_rw = UIO_WRITE; 445 auio.uio_segflg = UIO_USERSPACE; 446 auio.uio_procp = p; 447 error = copyin(iovp, iov, iovlen); 448 if (error) 449 goto done; 450 auio.uio_resid = 0; 451 for (i = 0; i < iovcnt; i++) { 452 auio.uio_resid += iov->iov_len; 453 /* 454 * Writes return ssize_t because -1 is returned on error. 455 * Therefore we must restrict the length to SSIZE_MAX to 456 * avoid garbage return values. 457 */ 458 if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) { 459 error = EINVAL; 460 goto done; 461 } 462 iov++; 463 } 464 #ifdef KTRACE 465 /* 466 * if tracing, save a copy of iovec 467 */ 468 if (KTRPOINT(p, KTR_GENIO)) { 469 ktriov = malloc(iovlen, M_TEMP, M_WAITOK); 470 memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen); 471 } 472 #endif 473 cnt = auio.uio_resid; 474 error = (*fp->f_ops->fo_write)(fp, offset, &auio, fp->f_cred, flags); 475 if (error) { 476 if (auio.uio_resid != cnt && (error == ERESTART || 477 error == EINTR || error == EWOULDBLOCK)) 478 error = 0; 479 if (error == EPIPE) 480 psignal(p, SIGPIPE); 481 } 482 cnt -= auio.uio_resid; 483 #ifdef KTRACE 484 if (KTRPOINT(p, KTR_GENIO)) 485 if (error == 0) { 486 ktrgenio(p, fd, UIO_WRITE, ktriov, cnt, error); 487 free(ktriov, M_TEMP); 488 } 489 #endif 490 *retval = cnt; 491 done: 492 if (needfree) 493 free(needfree, M_IOV); 494 out: 495 FILE_UNUSE(fp, p); 496 return (error); 497 } 498 499 /* 500 * Ioctl system call 501 */ 502 /* ARGSUSED */ 503 int 504 sys_ioctl(struct lwp *l, void *v, register_t *retval) 505 { 506 struct sys_ioctl_args /* { 507 syscallarg(int) fd; 508 syscallarg(u_long) com; 509 syscallarg(caddr_t) data; 510 } */ *uap = v; 511 struct file *fp; 512 struct proc *p; 513 struct filedesc *fdp; 514 u_long com; 515 int error; 516 u_int size; 517 caddr_t data, memp; 518 int tmp; 519 #define STK_PARAMS 128 520 u_long stkbuf[STK_PARAMS/sizeof(u_long)]; 521 522 error = 0; 523 p = l->l_proc; 524 fdp = p->p_fd; 525 526 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 527 return (EBADF); 528 529 FILE_USE(fp); 530 531 if ((fp->f_flag & (FREAD | FWRITE)) == 0) { 532 error = EBADF; 533 goto out; 534 } 535 536 switch (com = SCARG(uap, com)) { 537 case FIONCLEX: 538 fdp->fd_ofileflags[SCARG(uap, fd)] &= ~UF_EXCLOSE; 539 goto out; 540 541 case FIOCLEX: 542 fdp->fd_ofileflags[SCARG(uap, fd)] |= UF_EXCLOSE; 543 goto out; 544 } 545 546 /* 547 * Interpret high order word to find amount of data to be 548 * copied to/from the user's address space. 549 */ 550 size = IOCPARM_LEN(com); 551 if (size > IOCPARM_MAX) { 552 error = ENOTTY; 553 goto out; 554 } 555 memp = NULL; 556 if (size > sizeof(stkbuf)) { 557 memp = (caddr_t)malloc((u_long)size, M_IOCTLOPS, M_WAITOK); 558 data = memp; 559 } else 560 data = (caddr_t)stkbuf; 561 if (com&IOC_IN) { 562 if (size) { 563 error = copyin(SCARG(uap, data), data, size); 564 if (error) { 565 if (memp) 566 free(memp, M_IOCTLOPS); 567 goto out; 568 } 569 } else 570 *(caddr_t *)data = SCARG(uap, data); 571 } else if ((com&IOC_OUT) && size) 572 /* 573 * Zero the buffer so the user always 574 * gets back something deterministic. 575 */ 576 memset(data, 0, size); 577 else if (com&IOC_VOID) 578 *(caddr_t *)data = SCARG(uap, data); 579 580 switch (com) { 581 582 case FIONBIO: 583 if ((tmp = *(int *)data) != 0) 584 fp->f_flag |= FNONBLOCK; 585 else 586 fp->f_flag &= ~FNONBLOCK; 587 error = (*fp->f_ops->fo_ioctl)(fp, FIONBIO, (caddr_t)&tmp, p); 588 break; 589 590 case FIOASYNC: 591 if ((tmp = *(int *)data) != 0) 592 fp->f_flag |= FASYNC; 593 else 594 fp->f_flag &= ~FASYNC; 595 error = (*fp->f_ops->fo_ioctl)(fp, FIOASYNC, (caddr_t)&tmp, p); 596 break; 597 598 case FIOSETOWN: 599 tmp = *(int *)data; 600 if (fp->f_type == DTYPE_SOCKET) { 601 ((struct socket *)fp->f_data)->so_pgid = tmp; 602 error = 0; 603 break; 604 } 605 if (tmp <= 0) { 606 tmp = -tmp; 607 } else { 608 struct proc *p1 = pfind(tmp); 609 if (p1 == 0) { 610 error = ESRCH; 611 break; 612 } 613 tmp = p1->p_pgrp->pg_id; 614 } 615 error = (*fp->f_ops->fo_ioctl) 616 (fp, TIOCSPGRP, (caddr_t)&tmp, p); 617 break; 618 619 case FIOGETOWN: 620 if (fp->f_type == DTYPE_SOCKET) { 621 error = 0; 622 *(int *)data = ((struct socket *)fp->f_data)->so_pgid; 623 break; 624 } 625 error = (*fp->f_ops->fo_ioctl)(fp, TIOCGPGRP, data, p); 626 if (error == 0) 627 *(int *)data = -*(int *)data; 628 break; 629 630 default: 631 error = (*fp->f_ops->fo_ioctl)(fp, com, data, p); 632 /* 633 * Copy any data to user, size was 634 * already set and checked above. 635 */ 636 if (error == 0 && (com&IOC_OUT) && size) 637 error = copyout(data, SCARG(uap, data), size); 638 break; 639 } 640 if (memp) 641 free(memp, M_IOCTLOPS); 642 out: 643 FILE_UNUSE(fp, p); 644 switch (error) { 645 case -1: 646 printf("sys_ioctl: _IO%s%s('%c', %lu, %lu) returned -1: " 647 "pid=%d comm=%s\n", 648 (com & IOC_IN) ? "W" : "", (com & IOC_OUT) ? "R" : "", 649 (char)IOCGROUP(com), (com & 0xff), IOCPARM_LEN(com), 650 p->p_pid, p->p_comm); 651 /* FALLTHROUGH */ 652 case EPASSTHROUGH: 653 error = ENOTTY; 654 /* FALLTHROUGH */ 655 default: 656 return (error); 657 } 658 } 659 660 int selwait, nselcoll; 661 662 /* 663 * Select system call. 664 */ 665 int 666 sys_select(struct lwp *l, void *v, register_t *retval) 667 { 668 struct sys_select_args /* { 669 syscallarg(int) nd; 670 syscallarg(fd_set *) in; 671 syscallarg(fd_set *) ou; 672 syscallarg(fd_set *) ex; 673 syscallarg(struct timeval *) tv; 674 } */ *uap = v; 675 struct proc *p; 676 caddr_t bits; 677 char smallbits[howmany(FD_SETSIZE, NFDBITS) * 678 sizeof(fd_mask) * 6]; 679 struct timeval atv; 680 int s, ncoll, error, timo; 681 size_t ni; 682 683 error = 0; 684 p = l->l_proc; 685 if (SCARG(uap, nd) < 0) 686 return (EINVAL); 687 if (SCARG(uap, nd) > p->p_fd->fd_nfiles) { 688 /* forgiving; slightly wrong */ 689 SCARG(uap, nd) = p->p_fd->fd_nfiles; 690 } 691 ni = howmany(SCARG(uap, nd), NFDBITS) * sizeof(fd_mask); 692 if (ni * 6 > sizeof(smallbits)) 693 bits = malloc(ni * 6, M_TEMP, M_WAITOK); 694 else 695 bits = smallbits; 696 697 #define getbits(name, x) \ 698 if (SCARG(uap, name)) { \ 699 error = copyin(SCARG(uap, name), bits + ni * x, ni); \ 700 if (error) \ 701 goto done; \ 702 } else \ 703 memset(bits + ni * x, 0, ni); 704 getbits(in, 0); 705 getbits(ou, 1); 706 getbits(ex, 2); 707 #undef getbits 708 709 if (SCARG(uap, tv)) { 710 error = copyin(SCARG(uap, tv), (caddr_t)&atv, 711 sizeof(atv)); 712 if (error) 713 goto done; 714 if (itimerfix(&atv)) { 715 error = EINVAL; 716 goto done; 717 } 718 s = splclock(); 719 timeradd(&atv, &time, &atv); 720 splx(s); 721 } else 722 timo = 0; 723 retry: 724 ncoll = nselcoll; 725 l->l_flag |= L_SELECT; 726 error = selscan(p, (fd_mask *)(bits + ni * 0), 727 (fd_mask *)(bits + ni * 3), SCARG(uap, nd), retval); 728 if (error || *retval) 729 goto done; 730 if (SCARG(uap, tv)) { 731 /* 732 * We have to recalculate the timeout on every retry. 733 */ 734 timo = hzto(&atv); 735 if (timo <= 0) 736 goto done; 737 } 738 s = splsched(); 739 if ((l->l_flag & L_SELECT) == 0 || nselcoll != ncoll) { 740 splx(s); 741 goto retry; 742 } 743 l->l_flag &= ~L_SELECT; 744 error = tsleep((caddr_t)&selwait, PSOCK | PCATCH, "select", timo); 745 splx(s); 746 if (error == 0) 747 goto retry; 748 done: 749 l->l_flag &= ~L_SELECT; 750 /* select is not restarted after signals... */ 751 if (error == ERESTART) 752 error = EINTR; 753 if (error == EWOULDBLOCK) 754 error = 0; 755 if (error == 0) { 756 757 #define putbits(name, x) \ 758 if (SCARG(uap, name)) { \ 759 error = copyout(bits + ni * x, SCARG(uap, name), ni); \ 760 if (error) \ 761 goto out; \ 762 } 763 putbits(in, 3); 764 putbits(ou, 4); 765 putbits(ex, 5); 766 #undef putbits 767 } 768 out: 769 if (ni * 6 > sizeof(smallbits)) 770 free(bits, M_TEMP); 771 return (error); 772 } 773 774 int 775 selscan(struct proc *p, fd_mask *ibitp, fd_mask *obitp, int nfd, 776 register_t *retval) 777 { 778 struct filedesc *fdp; 779 int msk, i, j, fd, n; 780 fd_mask ibits, obits; 781 struct file *fp; 782 static const int flag[3] = { POLLRDNORM | POLLHUP | POLLERR, 783 POLLWRNORM | POLLHUP | POLLERR, 784 POLLRDBAND }; 785 786 fdp = p->p_fd; 787 n = 0; 788 for (msk = 0; msk < 3; msk++) { 789 for (i = 0; i < nfd; i += NFDBITS) { 790 ibits = *ibitp++; 791 obits = 0; 792 while ((j = ffs(ibits)) && (fd = i + --j) < nfd) { 793 ibits &= ~(1 << j); 794 if ((fp = fd_getfile(fdp, fd)) == NULL) 795 return (EBADF); 796 FILE_USE(fp); 797 if ((*fp->f_ops->fo_poll)(fp, flag[msk], p)) { 798 obits |= (1 << j); 799 n++; 800 } 801 FILE_UNUSE(fp, p); 802 } 803 *obitp++ = obits; 804 } 805 } 806 *retval = n; 807 return (0); 808 } 809 810 /* 811 * Poll system call. 812 */ 813 int 814 sys_poll(struct lwp *l, void *v, register_t *retval) 815 { 816 struct sys_poll_args /* { 817 syscallarg(struct pollfd *) fds; 818 syscallarg(u_int) nfds; 819 syscallarg(int) timeout; 820 } */ *uap = v; 821 struct proc *p; 822 caddr_t bits; 823 char smallbits[32 * sizeof(struct pollfd)]; 824 struct timeval atv; 825 int s, ncoll, error, timo; 826 size_t ni; 827 828 error = 0; 829 p = l->l_proc; 830 if (SCARG(uap, nfds) > p->p_fd->fd_nfiles) { 831 /* forgiving; slightly wrong */ 832 SCARG(uap, nfds) = p->p_fd->fd_nfiles; 833 } 834 ni = SCARG(uap, nfds) * sizeof(struct pollfd); 835 if (ni > sizeof(smallbits)) 836 bits = malloc(ni, M_TEMP, M_WAITOK); 837 else 838 bits = smallbits; 839 840 error = copyin(SCARG(uap, fds), bits, ni); 841 if (error) 842 goto done; 843 844 if (SCARG(uap, timeout) != INFTIM) { 845 atv.tv_sec = SCARG(uap, timeout) / 1000; 846 atv.tv_usec = (SCARG(uap, timeout) % 1000) * 1000; 847 if (itimerfix(&atv)) { 848 error = EINVAL; 849 goto done; 850 } 851 s = splclock(); 852 timeradd(&atv, &time, &atv); 853 splx(s); 854 } else 855 timo = 0; 856 retry: 857 ncoll = nselcoll; 858 l->l_flag |= L_SELECT; 859 error = pollscan(p, (struct pollfd *)bits, SCARG(uap, nfds), retval); 860 if (error || *retval) 861 goto done; 862 if (SCARG(uap, timeout) != INFTIM) { 863 /* 864 * We have to recalculate the timeout on every retry. 865 */ 866 timo = hzto(&atv); 867 if (timo <= 0) 868 goto done; 869 } 870 s = splsched(); 871 if ((l->l_flag & L_SELECT) == 0 || nselcoll != ncoll) { 872 splx(s); 873 goto retry; 874 } 875 l->l_flag &= ~L_SELECT; 876 error = tsleep((caddr_t)&selwait, PSOCK | PCATCH, "select", timo); 877 splx(s); 878 if (error == 0) 879 goto retry; 880 done: 881 l->l_flag &= ~L_SELECT; 882 /* poll is not restarted after signals... */ 883 if (error == ERESTART) 884 error = EINTR; 885 if (error == EWOULDBLOCK) 886 error = 0; 887 if (error == 0) { 888 error = copyout(bits, SCARG(uap, fds), ni); 889 if (error) 890 goto out; 891 } 892 out: 893 if (ni > sizeof(smallbits)) 894 free(bits, M_TEMP); 895 return (error); 896 } 897 898 int 899 pollscan(struct proc *p, struct pollfd *fds, int nfd, register_t *retval) 900 { 901 struct filedesc *fdp; 902 int i, n; 903 struct file *fp; 904 905 fdp = p->p_fd; 906 n = 0; 907 for (i = 0; i < nfd; i++, fds++) { 908 if (fds->fd >= fdp->fd_nfiles) { 909 fds->revents = POLLNVAL; 910 n++; 911 } else if (fds->fd < 0) { 912 fds->revents = 0; 913 } else { 914 if ((fp = fd_getfile(fdp, fds->fd)) == NULL) { 915 fds->revents = POLLNVAL; 916 n++; 917 } else { 918 FILE_USE(fp); 919 fds->revents = (*fp->f_ops->fo_poll)(fp, 920 fds->events | POLLERR | POLLHUP, p); 921 if (fds->revents != 0) 922 n++; 923 FILE_UNUSE(fp, p); 924 } 925 } 926 } 927 *retval = n; 928 return (0); 929 } 930 931 /*ARGSUSED*/ 932 int 933 seltrue(dev_t dev, int events, struct proc *p) 934 { 935 936 return (events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM)); 937 } 938 939 /* 940 * Record a select request. 941 */ 942 void 943 selrecord(struct proc *selector, struct selinfo *sip) 944 { 945 struct lwp *l; 946 struct proc *p; 947 pid_t mypid; 948 int collision; 949 950 mypid = selector->p_pid; 951 if (sip->si_pid == mypid) 952 return; 953 954 collision = 0; 955 if (sip->si_pid && (p = pfind(sip->si_pid))) { 956 for (l = LIST_FIRST(&p->p_lwps); l != NULL; 957 l = LIST_NEXT(l, l_sibling)) { 958 if (l->l_wchan == (caddr_t)&selwait) { 959 collision = 1; 960 sip->si_flags |= SI_COLL; 961 } 962 } 963 } 964 965 if (collision == 0) { 966 sip->si_flags &= ~SI_COLL; 967 sip->si_pid = mypid; 968 } 969 } 970 971 /* 972 * Do a wakeup when a selectable event occurs. 973 */ 974 void 975 selwakeup(sip) 976 struct selinfo *sip; 977 { 978 struct lwp *l; 979 struct proc *p; 980 int s; 981 982 if (sip->si_pid == 0) 983 return; 984 if (sip->si_flags & SI_COLL) { 985 nselcoll++; 986 sip->si_flags &= ~SI_COLL; 987 wakeup((caddr_t)&selwait); 988 } 989 p = pfind(sip->si_pid); 990 sip->si_pid = 0; 991 if (p != NULL) { 992 for (l = LIST_FIRST(&p->p_lwps); l != NULL; 993 l = LIST_NEXT(l, l_sibling)) { 994 SCHED_LOCK(s); 995 if (l->l_wchan == (caddr_t)&selwait) { 996 if (l->l_stat == LSSLEEP) 997 setrunnable(l); 998 else 999 unsleep(l); 1000 } else if (l->l_flag & L_SELECT) 1001 l->l_flag &= ~L_SELECT; 1002 SCHED_UNLOCK(s); 1003 } 1004 } 1005 } 1006
Indexes created Mon Sep 29 21:09:56 GMT 2025