sys_generic.c revision 1.69
1 /* $NetBSD: sys_generic.c,v 1.69 2003/01/18 10:06:33 thorpej 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 1.69 2003/01/18 10:06:33 thorpej 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 com = 0; 534 goto out; 535 } 536 537 switch (com = SCARG(uap, com)) { 538 case FIONCLEX: 539 fdp->fd_ofileflags[SCARG(uap, fd)] &= ~UF_EXCLOSE; 540 goto out; 541 542 case FIOCLEX: 543 fdp->fd_ofileflags[SCARG(uap, fd)] |= UF_EXCLOSE; 544 goto out; 545 } 546 547 /* 548 * Interpret high order word to find amount of data to be 549 * copied to/from the user's address space. 550 */ 551 size = IOCPARM_LEN(com); 552 if (size > IOCPARM_MAX) { 553 error = ENOTTY; 554 goto out; 555 } 556 memp = NULL; 557 if (size > sizeof(stkbuf)) { 558 memp = (caddr_t)malloc((u_long)size, M_IOCTLOPS, M_WAITOK); 559 data = memp; 560 } else 561 data = (caddr_t)stkbuf; 562 if (com&IOC_IN) { 563 if (size) { 564 error = copyin(SCARG(uap, data), data, size); 565 if (error) { 566 if (memp) 567 free(memp, M_IOCTLOPS); 568 goto out; 569 } 570 } else 571 *(caddr_t *)data = SCARG(uap, data); 572 } else if ((com&IOC_OUT) && size) 573 /* 574 * Zero the buffer so the user always 575 * gets back something deterministic. 576 */ 577 memset(data, 0, size); 578 else if (com&IOC_VOID) 579 *(caddr_t *)data = SCARG(uap, data); 580 581 switch (com) { 582 583 case FIONBIO: 584 if ((tmp = *(int *)data) != 0) 585 fp->f_flag |= FNONBLOCK; 586 else 587 fp->f_flag &= ~FNONBLOCK; 588 error = (*fp->f_ops->fo_ioctl)(fp, FIONBIO, (caddr_t)&tmp, p); 589 break; 590 591 case FIOASYNC: 592 if ((tmp = *(int *)data) != 0) 593 fp->f_flag |= FASYNC; 594 else 595 fp->f_flag &= ~FASYNC; 596 error = (*fp->f_ops->fo_ioctl)(fp, FIOASYNC, (caddr_t)&tmp, p); 597 break; 598 599 case FIOSETOWN: 600 tmp = *(int *)data; 601 if (fp->f_type == DTYPE_SOCKET) { 602 ((struct socket *)fp->f_data)->so_pgid = tmp; 603 error = 0; 604 break; 605 } 606 if (tmp <= 0) { 607 tmp = -tmp; 608 } else { 609 struct proc *p1 = pfind(tmp); 610 if (p1 == 0) { 611 error = ESRCH; 612 break; 613 } 614 tmp = p1->p_pgrp->pg_id; 615 } 616 error = (*fp->f_ops->fo_ioctl) 617 (fp, TIOCSPGRP, (caddr_t)&tmp, p); 618 break; 619 620 case FIOGETOWN: 621 if (fp->f_type == DTYPE_SOCKET) { 622 error = 0; 623 *(int *)data = ((struct socket *)fp->f_data)->so_pgid; 624 break; 625 } 626 error = (*fp->f_ops->fo_ioctl)(fp, TIOCGPGRP, data, p); 627 if (error == 0) 628 *(int *)data = -*(int *)data; 629 break; 630 631 default: 632 error = (*fp->f_ops->fo_ioctl)(fp, com, data, p); 633 /* 634 * Copy any data to user, size was 635 * already set and checked above. 636 */ 637 if (error == 0 && (com&IOC_OUT) && size) 638 error = copyout(data, SCARG(uap, data), size); 639 break; 640 } 641 if (memp) 642 free(memp, M_IOCTLOPS); 643 out: 644 FILE_UNUSE(fp, p); 645 switch (error) { 646 case -1: 647 printf("sys_ioctl: _IO%s%s('%c', %lu, %lu) returned -1: " 648 "pid=%d comm=%s\n", 649 (com & IOC_IN) ? "W" : "", (com & IOC_OUT) ? "R" : "", 650 (char)IOCGROUP(com), (com & 0xff), IOCPARM_LEN(com), 651 p->p_pid, p->p_comm); 652 /* FALLTHROUGH */ 653 case EPASSTHROUGH: 654 error = ENOTTY; 655 /* FALLTHROUGH */ 656 default: 657 return (error); 658 } 659 } 660 661 int selwait, nselcoll; 662 663 /* 664 * Select system call. 665 */ 666 int 667 sys_select(struct lwp *l, void *v, register_t *retval) 668 { 669 struct sys_select_args /* { 670 syscallarg(int) nd; 671 syscallarg(fd_set *) in; 672 syscallarg(fd_set *) ou; 673 syscallarg(fd_set *) ex; 674 syscallarg(struct timeval *) tv; 675 } */ *uap = v; 676 struct proc *p; 677 caddr_t bits; 678 char smallbits[howmany(FD_SETSIZE, NFDBITS) * 679 sizeof(fd_mask) * 6]; 680 struct timeval atv; 681 int s, ncoll, error, timo; 682 size_t ni; 683 684 error = 0; 685 p = l->l_proc; 686 if (SCARG(uap, nd) < 0) 687 return (EINVAL); 688 if (SCARG(uap, nd) > p->p_fd->fd_nfiles) { 689 /* forgiving; slightly wrong */ 690 SCARG(uap, nd) = p->p_fd->fd_nfiles; 691 } 692 ni = howmany(SCARG(uap, nd), NFDBITS) * sizeof(fd_mask); 693 if (ni * 6 > sizeof(smallbits)) 694 bits = malloc(ni * 6, M_TEMP, M_WAITOK); 695 else 696 bits = smallbits; 697 698 #define getbits(name, x) \ 699 if (SCARG(uap, name)) { \ 700 error = copyin(SCARG(uap, name), bits + ni * x, ni); \ 701 if (error) \ 702 goto done; \ 703 } else \ 704 memset(bits + ni * x, 0, ni); 705 getbits(in, 0); 706 getbits(ou, 1); 707 getbits(ex, 2); 708 #undef getbits 709 710 timo = 0; 711 if (SCARG(uap, tv)) { 712 error = copyin(SCARG(uap, tv), (caddr_t)&atv, 713 sizeof(atv)); 714 if (error) 715 goto done; 716 if (itimerfix(&atv)) { 717 error = EINVAL; 718 goto done; 719 } 720 s = splclock(); 721 timeradd(&atv, &time, &atv); 722 splx(s); 723 } 724 725 retry: 726 ncoll = nselcoll; 727 l->l_flag |= L_SELECT; 728 error = selscan(p, (fd_mask *)(bits + ni * 0), 729 (fd_mask *)(bits + ni * 3), SCARG(uap, nd), retval); 730 if (error || *retval) 731 goto done; 732 if (SCARG(uap, tv)) { 733 /* 734 * We have to recalculate the timeout on every retry. 735 */ 736 timo = hzto(&atv); 737 if (timo <= 0) 738 goto done; 739 } 740 s = splsched(); 741 if ((l->l_flag & L_SELECT) == 0 || nselcoll != ncoll) { 742 splx(s); 743 goto retry; 744 } 745 l->l_flag &= ~L_SELECT; 746 error = tsleep((caddr_t)&selwait, PSOCK | PCATCH, "select", timo); 747 splx(s); 748 if (error == 0) 749 goto retry; 750 done: 751 l->l_flag &= ~L_SELECT; 752 /* select is not restarted after signals... */ 753 if (error == ERESTART) 754 error = EINTR; 755 if (error == EWOULDBLOCK) 756 error = 0; 757 if (error == 0) { 758 759 #define putbits(name, x) \ 760 if (SCARG(uap, name)) { \ 761 error = copyout(bits + ni * x, SCARG(uap, name), ni); \ 762 if (error) \ 763 goto out; \ 764 } 765 putbits(in, 3); 766 putbits(ou, 4); 767 putbits(ex, 5); 768 #undef putbits 769 } 770 out: 771 if (ni * 6 > sizeof(smallbits)) 772 free(bits, M_TEMP); 773 return (error); 774 } 775 776 int 777 selscan(struct proc *p, fd_mask *ibitp, fd_mask *obitp, int nfd, 778 register_t *retval) 779 { 780 struct filedesc *fdp; 781 int msk, i, j, fd, n; 782 fd_mask ibits, obits; 783 struct file *fp; 784 static const int flag[3] = { POLLRDNORM | POLLHUP | POLLERR, 785 POLLWRNORM | POLLHUP | POLLERR, 786 POLLRDBAND }; 787 788 fdp = p->p_fd; 789 n = 0; 790 for (msk = 0; msk < 3; msk++) { 791 for (i = 0; i < nfd; i += NFDBITS) { 792 ibits = *ibitp++; 793 obits = 0; 794 while ((j = ffs(ibits)) && (fd = i + --j) < nfd) { 795 ibits &= ~(1 << j); 796 if ((fp = fd_getfile(fdp, fd)) == NULL) 797 return (EBADF); 798 FILE_USE(fp); 799 if ((*fp->f_ops->fo_poll)(fp, flag[msk], p)) { 800 obits |= (1 << j); 801 n++; 802 } 803 FILE_UNUSE(fp, p); 804 } 805 *obitp++ = obits; 806 } 807 } 808 *retval = n; 809 return (0); 810 } 811 812 /* 813 * Poll system call. 814 */ 815 int 816 sys_poll(struct lwp *l, void *v, register_t *retval) 817 { 818 struct sys_poll_args /* { 819 syscallarg(struct pollfd *) fds; 820 syscallarg(u_int) nfds; 821 syscallarg(int) timeout; 822 } */ *uap = v; 823 struct proc *p; 824 caddr_t bits; 825 char smallbits[32 * sizeof(struct pollfd)]; 826 struct timeval atv; 827 int s, ncoll, error, timo; 828 size_t ni; 829 830 error = 0; 831 p = l->l_proc; 832 if (SCARG(uap, nfds) > p->p_fd->fd_nfiles) { 833 /* forgiving; slightly wrong */ 834 SCARG(uap, nfds) = p->p_fd->fd_nfiles; 835 } 836 ni = SCARG(uap, nfds) * sizeof(struct pollfd); 837 if (ni > sizeof(smallbits)) 838 bits = malloc(ni, M_TEMP, M_WAITOK); 839 else 840 bits = smallbits; 841 842 error = copyin(SCARG(uap, fds), bits, ni); 843 if (error) 844 goto done; 845 846 timo = 0; 847 if (SCARG(uap, timeout) != INFTIM) { 848 atv.tv_sec = SCARG(uap, timeout) / 1000; 849 atv.tv_usec = (SCARG(uap, timeout) % 1000) * 1000; 850 if (itimerfix(&atv)) { 851 error = EINVAL; 852 goto done; 853 } 854 s = splclock(); 855 timeradd(&atv, &time, &atv); 856 splx(s); 857 } 858 859 retry: 860 ncoll = nselcoll; 861 l->l_flag |= L_SELECT; 862 error = pollscan(p, (struct pollfd *)bits, SCARG(uap, nfds), retval); 863 if (error || *retval) 864 goto done; 865 if (SCARG(uap, timeout) != INFTIM) { 866 /* 867 * We have to recalculate the timeout on every retry. 868 */ 869 timo = hzto(&atv); 870 if (timo <= 0) 871 goto done; 872 } 873 s = splsched(); 874 if ((l->l_flag & L_SELECT) == 0 || nselcoll != ncoll) { 875 splx(s); 876 goto retry; 877 } 878 l->l_flag &= ~L_SELECT; 879 error = tsleep((caddr_t)&selwait, PSOCK | PCATCH, "select", timo); 880 splx(s); 881 if (error == 0) 882 goto retry; 883 done: 884 l->l_flag &= ~L_SELECT; 885 /* poll is not restarted after signals... */ 886 if (error == ERESTART) 887 error = EINTR; 888 if (error == EWOULDBLOCK) 889 error = 0; 890 if (error == 0) { 891 error = copyout(bits, SCARG(uap, fds), ni); 892 if (error) 893 goto out; 894 } 895 out: 896 if (ni > sizeof(smallbits)) 897 free(bits, M_TEMP); 898 return (error); 899 } 900 901 int 902 pollscan(struct proc *p, struct pollfd *fds, int nfd, register_t *retval) 903 { 904 struct filedesc *fdp; 905 int i, n; 906 struct file *fp; 907 908 fdp = p->p_fd; 909 n = 0; 910 for (i = 0; i < nfd; i++, fds++) { 911 if (fds->fd >= fdp->fd_nfiles) { 912 fds->revents = POLLNVAL; 913 n++; 914 } else if (fds->fd < 0) { 915 fds->revents = 0; 916 } else { 917 if ((fp = fd_getfile(fdp, fds->fd)) == NULL) { 918 fds->revents = POLLNVAL; 919 n++; 920 } else { 921 FILE_USE(fp); 922 fds->revents = (*fp->f_ops->fo_poll)(fp, 923 fds->events | POLLERR | POLLHUP, p); 924 if (fds->revents != 0) 925 n++; 926 FILE_UNUSE(fp, p); 927 } 928 } 929 } 930 *retval = n; 931 return (0); 932 } 933 934 /*ARGSUSED*/ 935 int 936 seltrue(dev_t dev, int events, struct proc *p) 937 { 938 939 return (events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM)); 940 } 941 942 /* 943 * Record a select request. 944 */ 945 void 946 selrecord(struct proc *selector, struct selinfo *sip) 947 { 948 struct lwp *l; 949 struct proc *p; 950 pid_t mypid; 951 int collision; 952 953 mypid = selector->p_pid; 954 if (sip->sel_pid == mypid) 955 return; 956 collision = 0; 957 if (sip->sel_pid && (p = pfind(sip->sel_pid))) { 958 for (l = LIST_FIRST(&p->p_lwps); l != NULL; 959 l = LIST_NEXT(l, l_sibling)) { 960 if (l->l_wchan == (caddr_t)&selwait) { 961 collision = 1; 962 sip->sel_flags |= SI_COLL; 963 } 964 } 965 } 966 967 if (collision == 0) 968 sip->sel_pid = mypid; 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->sel_pid == 0) 983 return; 984 if (sip->sel_flags & SI_COLL) { 985 sip->sel_pid = 0; 986 nselcoll++; 987 sip->sel_flags &= ~SI_COLL; 988 wakeup((caddr_t)&selwait); 989 return; 990 } 991 p = pfind(sip->sel_pid); 992 sip->sel_pid = 0; 993 if (p != NULL) { 994 for (l = LIST_FIRST(&p->p_lwps); l != NULL; 995 l = LIST_NEXT(l, l_sibling)) { 996 SCHED_LOCK(s); 997 if (l->l_wchan == (caddr_t)&selwait) { 998 if (l->l_stat == LSSLEEP) 999 setrunnable(l); 1000 else 1001 unsleep(l); 1002 } else if (l->l_flag & L_SELECT) 1003 l->l_flag &= ~L_SELECT; 1004 SCHED_UNLOCK(s); 1005 } 1006 } 1007 } 1008
Indexes created Mon Sep 22 10:09:38 GMT 2025