sys_generic.c revision 1.54.2.7
1 /* $NetBSD: sys_generic.c,v 1.54.2.7 2002/05/29 21:33:14 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 1.54.2.7 2002/05/29 21:33:14 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/lwp.h> 54 #include <sys/proc.h> 55 #include <sys/socketvar.h> 56 #include <sys/signalvar.h> 57 #include <sys/uio.h> 58 #include <sys/kernel.h> 59 #include <sys/stat.h> 60 #include <sys/malloc.h> 61 #include <sys/poll.h> 62 #ifdef KTRACE 63 #include <sys/ktrace.h> 64 #endif 65 66 #include <sys/mount.h> 67 #include <sys/sa.h> 68 #include <sys/syscallargs.h> 69 70 int selscan __P((struct proc *, fd_mask *, fd_mask *, int, register_t *)); 71 int pollscan __P((struct proc *, struct pollfd *, int, register_t *)); 72 73 /* 74 * Read system call. 75 */ 76 /* ARGSUSED */ 77 int 78 sys_read(struct lwp *l, void *v, register_t *retval) 79 { 80 struct sys_read_args /* { 81 syscallarg(int) fd; 82 syscallarg(void *) buf; 83 syscallarg(size_t) nbyte; 84 } */ *uap = v; 85 int fd; 86 struct file *fp; 87 struct proc *p; 88 struct filedesc *fdp; 89 90 fd = SCARG(uap, fd); 91 p = l->l_proc; 92 fdp = p->p_fd; 93 94 if ((fp = fd_getfile(fdp, fd)) == NULL) 95 return (EBADF); 96 97 if ((fp->f_flag & FREAD) == 0) 98 return (EBADF); 99 100 FILE_USE(fp); 101 102 /* dofileread() will unuse the descriptor for us */ 103 return (dofileread(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 104 &fp->f_offset, FOF_UPDATE_OFFSET, retval)); 105 } 106 107 int 108 dofileread(struct proc *p, int fd, struct file *fp, void *buf, size_t nbyte, 109 off_t *offset, int flags, register_t *retval) 110 { 111 struct uio auio; 112 struct iovec aiov; 113 long cnt, 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 long i, cnt, error; 202 u_int iovlen; 203 #ifdef KTRACE 204 struct iovec *ktriov; 205 #endif 206 207 error = 0; 208 #ifdef KTRACE 209 ktriov = NULL; 210 #endif 211 /* note: can't use iovlen until iovcnt is validated */ 212 iovlen = iovcnt * sizeof(struct iovec); 213 if ((u_int)iovcnt > UIO_SMALLIOV) { 214 if ((u_int)iovcnt > IOV_MAX) { 215 error = EINVAL; 216 goto out; 217 } 218 iov = malloc(iovlen, M_IOV, M_WAITOK); 219 needfree = iov; 220 } else if ((u_int)iovcnt > 0) { 221 iov = aiov; 222 needfree = NULL; 223 } else { 224 error = EINVAL; 225 goto out; 226 } 227 228 auio.uio_iov = iov; 229 auio.uio_iovcnt = iovcnt; 230 auio.uio_rw = UIO_READ; 231 auio.uio_segflg = UIO_USERSPACE; 232 auio.uio_procp = p; 233 error = copyin(iovp, iov, iovlen); 234 if (error) 235 goto done; 236 auio.uio_resid = 0; 237 for (i = 0; i < iovcnt; i++) { 238 auio.uio_resid += iov->iov_len; 239 /* 240 * Reads return ssize_t because -1 is returned on error. 241 * Therefore we must restrict the length to SSIZE_MAX to 242 * avoid garbage return values. 243 */ 244 if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) { 245 error = EINVAL; 246 goto done; 247 } 248 iov++; 249 } 250 #ifdef KTRACE 251 /* 252 * if tracing, save a copy of iovec 253 */ 254 if (KTRPOINT(p, KTR_GENIO)) { 255 ktriov = malloc(iovlen, M_TEMP, M_WAITOK); 256 memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen); 257 } 258 #endif 259 cnt = auio.uio_resid; 260 error = (*fp->f_ops->fo_read)(fp, offset, &auio, fp->f_cred, flags); 261 if (error) 262 if (auio.uio_resid != cnt && (error == ERESTART || 263 error == EINTR || error == EWOULDBLOCK)) 264 error = 0; 265 cnt -= auio.uio_resid; 266 #ifdef KTRACE 267 if (ktriov != NULL) { 268 if (error == 0) 269 ktrgenio(p, fd, UIO_READ, ktriov, cnt, error); 270 free(ktriov, M_TEMP); 271 } 272 #endif 273 *retval = cnt; 274 done: 275 if (needfree) 276 free(needfree, M_IOV); 277 out: 278 FILE_UNUSE(fp, p); 279 return (error); 280 } 281 282 /* 283 * Write system call 284 */ 285 int 286 sys_write(struct lwp *l, void *v, register_t *retval) 287 { 288 struct sys_write_args /* { 289 syscallarg(int) fd; 290 syscallarg(const void *) buf; 291 syscallarg(size_t) nbyte; 292 } */ *uap = v; 293 int fd; 294 struct file *fp; 295 struct proc *p; 296 struct filedesc *fdp; 297 298 fd = SCARG(uap, fd); 299 p = l->l_proc; 300 fdp = p->p_fd; 301 302 if ((fp = fd_getfile(fdp, fd)) == NULL) 303 return (EBADF); 304 305 if ((fp->f_flag & FWRITE) == 0) 306 return (EBADF); 307 308 FILE_USE(fp); 309 310 /* dofilewrite() will unuse the descriptor for us */ 311 return (dofilewrite(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 312 &fp->f_offset, FOF_UPDATE_OFFSET, retval)); 313 } 314 315 int 316 dofilewrite(struct proc *p, int fd, struct file *fp, const void *buf, 317 size_t nbyte, off_t *offset, int flags, register_t *retval) 318 { 319 struct uio auio; 320 struct iovec aiov; 321 long cnt, error; 322 #ifdef KTRACE 323 struct iovec ktriov; 324 #endif 325 326 error = 0; 327 aiov.iov_base = (caddr_t)buf; /* XXX kills const */ 328 aiov.iov_len = nbyte; 329 auio.uio_iov = &aiov; 330 auio.uio_iovcnt = 1; 331 auio.uio_resid = nbyte; 332 auio.uio_rw = UIO_WRITE; 333 auio.uio_segflg = UIO_USERSPACE; 334 auio.uio_procp = p; 335 336 /* 337 * Writes return ssize_t because -1 is returned on error. Therefore 338 * we must restrict the length to SSIZE_MAX to avoid garbage return 339 * values. 340 */ 341 if (auio.uio_resid > SSIZE_MAX) { 342 error = EINVAL; 343 goto out; 344 } 345 346 #ifdef KTRACE 347 /* 348 * if tracing, save a copy of iovec 349 */ 350 if (KTRPOINT(p, KTR_GENIO)) 351 ktriov = aiov; 352 #endif 353 cnt = auio.uio_resid; 354 error = (*fp->f_ops->fo_write)(fp, offset, &auio, fp->f_cred, flags); 355 if (error) { 356 if (auio.uio_resid != cnt && (error == ERESTART || 357 error == EINTR || error == EWOULDBLOCK)) 358 error = 0; 359 if (error == EPIPE) 360 psignal(p, SIGPIPE); 361 } 362 cnt -= auio.uio_resid; 363 #ifdef KTRACE 364 if (KTRPOINT(p, KTR_GENIO) && error == 0) 365 ktrgenio(p, fd, UIO_WRITE, &ktriov, cnt, error); 366 #endif 367 *retval = cnt; 368 out: 369 FILE_UNUSE(fp, p); 370 return (error); 371 } 372 373 /* 374 * Gather write system call 375 */ 376 int 377 sys_writev(struct lwp *l, void *v, register_t *retval) 378 { 379 struct sys_writev_args /* { 380 syscallarg(int) fd; 381 syscallarg(const struct iovec *) iovp; 382 syscallarg(int) iovcnt; 383 } */ *uap = v; 384 int fd; 385 struct file *fp; 386 struct proc *p; 387 struct filedesc *fdp; 388 389 fd = SCARG(uap, fd); 390 p = l->l_proc; 391 fdp = p->p_fd; 392 393 if ((fp = fd_getfile(fdp, fd)) == NULL) 394 return (EBADF); 395 396 if ((fp->f_flag & FWRITE) == 0) 397 return (EBADF); 398 399 FILE_USE(fp); 400 401 /* dofilewritev() will unuse the descriptor for us */ 402 return (dofilewritev(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 403 &fp->f_offset, FOF_UPDATE_OFFSET, retval)); 404 } 405 406 int 407 dofilewritev(struct proc *p, int fd, struct file *fp, const struct iovec *iovp, 408 int iovcnt, off_t *offset, int flags, register_t *retval) 409 { 410 struct uio auio; 411 struct iovec *iov, *needfree, aiov[UIO_SMALLIOV]; 412 long i, cnt, error; 413 u_int iovlen; 414 #ifdef KTRACE 415 struct iovec *ktriov; 416 #endif 417 418 error = 0; 419 #ifdef KTRACE 420 ktriov = NULL; 421 #endif 422 /* note: can't use iovlen until iovcnt is validated */ 423 iovlen = iovcnt * sizeof(struct iovec); 424 if ((u_int)iovcnt > UIO_SMALLIOV) { 425 if ((u_int)iovcnt > IOV_MAX) { 426 error = EINVAL; 427 goto out; 428 } 429 iov = malloc(iovlen, M_IOV, M_WAITOK); 430 needfree = iov; 431 } else if ((u_int)iovcnt > 0) { 432 iov = aiov; 433 needfree = NULL; 434 } else { 435 error = EINVAL; 436 goto out; 437 } 438 439 auio.uio_iov = iov; 440 auio.uio_iovcnt = iovcnt; 441 auio.uio_rw = UIO_WRITE; 442 auio.uio_segflg = UIO_USERSPACE; 443 auio.uio_procp = p; 444 error = copyin(iovp, iov, iovlen); 445 if (error) 446 goto done; 447 auio.uio_resid = 0; 448 for (i = 0; i < iovcnt; i++) { 449 auio.uio_resid += iov->iov_len; 450 /* 451 * Writes return ssize_t because -1 is returned on error. 452 * Therefore we must restrict the length to SSIZE_MAX to 453 * avoid garbage return values. 454 */ 455 if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) { 456 error = EINVAL; 457 goto done; 458 } 459 iov++; 460 } 461 #ifdef KTRACE 462 /* 463 * if tracing, save a copy of iovec 464 */ 465 if (KTRPOINT(p, KTR_GENIO)) { 466 ktriov = malloc(iovlen, M_TEMP, M_WAITOK); 467 memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen); 468 } 469 #endif 470 cnt = auio.uio_resid; 471 error = (*fp->f_ops->fo_write)(fp, offset, &auio, fp->f_cred, flags); 472 if (error) { 473 if (auio.uio_resid != cnt && (error == ERESTART || 474 error == EINTR || error == EWOULDBLOCK)) 475 error = 0; 476 if (error == EPIPE) 477 psignal(p, SIGPIPE); 478 } 479 cnt -= auio.uio_resid; 480 #ifdef KTRACE 481 if (KTRPOINT(p, KTR_GENIO)) 482 if (error == 0) { 483 ktrgenio(p, fd, UIO_WRITE, ktriov, cnt, error); 484 free(ktriov, M_TEMP); 485 } 486 #endif 487 *retval = cnt; 488 done: 489 if (needfree) 490 free(needfree, M_IOV); 491 out: 492 FILE_UNUSE(fp, p); 493 return (error); 494 } 495 496 /* 497 * Ioctl system call 498 */ 499 /* ARGSUSED */ 500 int 501 sys_ioctl(struct lwp *l, void *v, register_t *retval) 502 { 503 struct sys_ioctl_args /* { 504 syscallarg(int) fd; 505 syscallarg(u_long) com; 506 syscallarg(caddr_t) data; 507 } */ *uap = v; 508 struct file *fp; 509 struct proc *p; 510 struct filedesc *fdp; 511 u_long com; 512 int error; 513 u_int size; 514 caddr_t data, memp; 515 int tmp; 516 #define STK_PARAMS 128 517 u_long stkbuf[STK_PARAMS/sizeof(u_long)]; 518 519 error = 0; 520 p = l->l_proc; 521 fdp = p->p_fd; 522 523 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 524 return (EBADF); 525 526 FILE_USE(fp); 527 528 if ((fp->f_flag & (FREAD | FWRITE)) == 0) { 529 error = EBADF; 530 goto out; 531 } 532 533 switch (com = SCARG(uap, com)) { 534 case FIONCLEX: 535 fdp->fd_ofileflags[SCARG(uap, fd)] &= ~UF_EXCLOSE; 536 goto out; 537 538 case FIOCLEX: 539 fdp->fd_ofileflags[SCARG(uap, fd)] |= UF_EXCLOSE; 540 goto out; 541 } 542 543 /* 544 * Interpret high order word to find amount of data to be 545 * copied to/from the user's address space. 546 */ 547 size = IOCPARM_LEN(com); 548 if (size > IOCPARM_MAX) { 549 error = ENOTTY; 550 goto out; 551 } 552 memp = NULL; 553 if (size > sizeof(stkbuf)) { 554 memp = (caddr_t)malloc((u_long)size, M_IOCTLOPS, M_WAITOK); 555 data = memp; 556 } else 557 data = (caddr_t)stkbuf; 558 if (com&IOC_IN) { 559 if (size) { 560 error = copyin(SCARG(uap, data), data, size); 561 if (error) { 562 if (memp) 563 free(memp, M_IOCTLOPS); 564 goto out; 565 } 566 } else 567 *(caddr_t *)data = SCARG(uap, data); 568 } else if ((com&IOC_OUT) && size) 569 /* 570 * Zero the buffer so the user always 571 * gets back something deterministic. 572 */ 573 memset(data, 0, size); 574 else if (com&IOC_VOID) 575 *(caddr_t *)data = SCARG(uap, data); 576 577 switch (com) { 578 579 case FIONBIO: 580 if ((tmp = *(int *)data) != 0) 581 fp->f_flag |= FNONBLOCK; 582 else 583 fp->f_flag &= ~FNONBLOCK; 584 error = (*fp->f_ops->fo_ioctl)(fp, FIONBIO, (caddr_t)&tmp, p); 585 break; 586 587 case FIOASYNC: 588 if ((tmp = *(int *)data) != 0) 589 fp->f_flag |= FASYNC; 590 else 591 fp->f_flag &= ~FASYNC; 592 error = (*fp->f_ops->fo_ioctl)(fp, FIOASYNC, (caddr_t)&tmp, p); 593 break; 594 595 case FIOSETOWN: 596 tmp = *(int *)data; 597 if (fp->f_type == DTYPE_SOCKET) { 598 ((struct socket *)fp->f_data)->so_pgid = tmp; 599 error = 0; 600 break; 601 } 602 if (tmp <= 0) { 603 tmp = -tmp; 604 } else { 605 struct proc *p1 = pfind(tmp); 606 if (p1 == 0) { 607 error = ESRCH; 608 break; 609 } 610 tmp = p1->p_pgrp->pg_id; 611 } 612 error = (*fp->f_ops->fo_ioctl) 613 (fp, TIOCSPGRP, (caddr_t)&tmp, p); 614 break; 615 616 case FIOGETOWN: 617 if (fp->f_type == DTYPE_SOCKET) { 618 error = 0; 619 *(int *)data = ((struct socket *)fp->f_data)->so_pgid; 620 break; 621 } 622 error = (*fp->f_ops->fo_ioctl)(fp, TIOCGPGRP, data, p); 623 if (error == 0) 624 *(int *)data = -*(int *)data; 625 break; 626 627 default: 628 error = (*fp->f_ops->fo_ioctl)(fp, com, data, p); 629 /* 630 * Copy any data to user, size was 631 * already set and checked above. 632 */ 633 if (error == 0 && (com&IOC_OUT) && size) 634 error = copyout(data, SCARG(uap, data), size); 635 break; 636 } 637 if (memp) 638 free(memp, M_IOCTLOPS); 639 out: 640 FILE_UNUSE(fp, p); 641 switch (error) { 642 case -1: 643 printf("sys_ioctl: _IO%s%s('%c', %lu, %lu) returned -1: " 644 "pid=%d comm=%s\n", 645 (com & IOC_IN) ? "W" : "", (com & IOC_OUT) ? "R" : "", 646 (char)IOCGROUP(com), (com & 0xff), IOCPARM_LEN(com), 647 p->p_pid, p->p_comm); 648 /* FALLTHROUGH */ 649 case EPASSTHROUGH: 650 error = ENOTTY; 651 /* FALLTHROUGH */ 652 default: 653 return (error); 654 } 655 } 656 657 int selwait, nselcoll; 658 659 /* 660 * Select system call. 661 */ 662 int 663 sys_select(struct lwp *l, void *v, register_t *retval) 664 { 665 struct sys_select_args /* { 666 syscallarg(int) nd; 667 syscallarg(fd_set *) in; 668 syscallarg(fd_set *) ou; 669 syscallarg(fd_set *) ex; 670 syscallarg(struct timeval *) tv; 671 } */ *uap = v; 672 struct proc *p; 673 caddr_t bits; 674 char smallbits[howmany(FD_SETSIZE, NFDBITS) * 675 sizeof(fd_mask) * 6]; 676 struct timeval atv; 677 int s, ncoll, error, timo; 678 size_t ni; 679 680 error = 0; 681 p = l->l_proc; 682 if (SCARG(uap, nd) < 0) 683 return (EINVAL); 684 if (SCARG(uap, nd) > p->p_fd->fd_nfiles) { 685 /* forgiving; slightly wrong */ 686 SCARG(uap, nd) = p->p_fd->fd_nfiles; 687 } 688 ni = howmany(SCARG(uap, nd), NFDBITS) * sizeof(fd_mask); 689 if (ni * 6 > sizeof(smallbits)) 690 bits = malloc(ni * 6, M_TEMP, M_WAITOK); 691 else 692 bits = smallbits; 693 694 #define getbits(name, x) \ 695 if (SCARG(uap, name)) { \ 696 error = copyin(SCARG(uap, name), bits + ni * x, ni); \ 697 if (error) \ 698 goto done; \ 699 } else \ 700 memset(bits + ni * x, 0, ni); 701 getbits(in, 0); 702 getbits(ou, 1); 703 getbits(ex, 2); 704 #undef getbits 705 706 if (SCARG(uap, tv)) { 707 error = copyin(SCARG(uap, tv), (caddr_t)&atv, 708 sizeof(atv)); 709 if (error) 710 goto done; 711 if (itimerfix(&atv)) { 712 error = EINVAL; 713 goto done; 714 } 715 s = splclock(); 716 timeradd(&atv, &time, &atv); 717 splx(s); 718 } else 719 timo = 0; 720 retry: 721 ncoll = nselcoll; 722 l->l_flag |= L_SELECT; 723 error = selscan(p, (fd_mask *)(bits + ni * 0), 724 (fd_mask *)(bits + ni * 3), SCARG(uap, nd), retval); 725 if (error || *retval) 726 goto done; 727 if (SCARG(uap, tv)) { 728 /* 729 * We have to recalculate the timeout on every retry. 730 */ 731 timo = hzto(&atv); 732 if (timo <= 0) 733 goto done; 734 } 735 s = splsched(); 736 if ((l->l_flag & L_SELECT) == 0 || nselcoll != ncoll) { 737 splx(s); 738 goto retry; 739 } 740 l->l_flag &= ~L_SELECT; 741 error = tsleep((caddr_t)&selwait, PSOCK | PCATCH, "select", timo); 742 splx(s); 743 if (error == 0) 744 goto retry; 745 done: 746 l->l_flag &= ~L_SELECT; 747 /* select is not restarted after signals... */ 748 if (error == ERESTART) 749 error = EINTR; 750 if (error == EWOULDBLOCK) 751 error = 0; 752 if (error == 0) { 753 754 #define putbits(name, x) \ 755 if (SCARG(uap, name)) { \ 756 error = copyout(bits + ni * x, SCARG(uap, name), ni); \ 757 if (error) \ 758 goto out; \ 759 } 760 putbits(in, 3); 761 putbits(ou, 4); 762 putbits(ex, 5); 763 #undef putbits 764 } 765 out: 766 if (ni * 6 > sizeof(smallbits)) 767 free(bits, M_TEMP); 768 return (error); 769 } 770 771 int 772 selscan(struct proc *p, fd_mask *ibitp, fd_mask *obitp, int nfd, 773 register_t *retval) 774 { 775 struct filedesc *fdp; 776 int msk, i, j, fd, n; 777 fd_mask ibits, obits; 778 struct file *fp; 779 static int flag[3] = { POLLRDNORM | POLLHUP | POLLERR, 780 POLLWRNORM | POLLHUP | POLLERR, 781 POLLRDBAND }; 782 783 fdp = p->p_fd; 784 n = 0; 785 for (msk = 0; msk < 3; msk++) { 786 for (i = 0; i < nfd; i += NFDBITS) { 787 ibits = *ibitp++; 788 obits = 0; 789 while ((j = ffs(ibits)) && (fd = i + --j) < nfd) { 790 ibits &= ~(1 << j); 791 if ((fp = fd_getfile(fdp, fd)) == NULL) 792 return (EBADF); 793 FILE_USE(fp); 794 if ((*fp->f_ops->fo_poll)(fp, flag[msk], p)) { 795 obits |= (1 << j); 796 n++; 797 } 798 FILE_UNUSE(fp, p); 799 } 800 *obitp++ = obits; 801 } 802 } 803 *retval = n; 804 return (0); 805 } 806 807 /* 808 * Poll system call. 809 */ 810 int 811 sys_poll(struct lwp *l, void *v, register_t *retval) 812 { 813 struct sys_poll_args /* { 814 syscallarg(struct pollfd *) fds; 815 syscallarg(u_int) nfds; 816 syscallarg(int) timeout; 817 } */ *uap = v; 818 struct proc *p; 819 caddr_t bits; 820 char smallbits[32 * sizeof(struct pollfd)]; 821 struct timeval atv; 822 int s, ncoll, error, timo; 823 size_t ni; 824 825 error = 0; 826 p = l->l_proc; 827 if (SCARG(uap, nfds) > p->p_fd->fd_nfiles) { 828 /* forgiving; slightly wrong */ 829 SCARG(uap, nfds) = p->p_fd->fd_nfiles; 830 } 831 ni = SCARG(uap, nfds) * sizeof(struct pollfd); 832 if (ni > sizeof(smallbits)) 833 bits = malloc(ni, M_TEMP, M_WAITOK); 834 else 835 bits = smallbits; 836 837 error = copyin(SCARG(uap, fds), bits, ni); 838 if (error) 839 goto done; 840 841 if (SCARG(uap, timeout) != INFTIM) { 842 atv.tv_sec = SCARG(uap, timeout) / 1000; 843 atv.tv_usec = (SCARG(uap, timeout) % 1000) * 1000; 844 if (itimerfix(&atv)) { 845 error = EINVAL; 846 goto done; 847 } 848 s = splclock(); 849 timeradd(&atv, &time, &atv); 850 splx(s); 851 } else 852 timo = 0; 853 retry: 854 ncoll = nselcoll; 855 l->l_flag |= L_SELECT; 856 error = pollscan(p, (struct pollfd *)bits, SCARG(uap, nfds), retval); 857 if (error || *retval) 858 goto done; 859 if (SCARG(uap, timeout) != INFTIM) { 860 /* 861 * We have to recalculate the timeout on every retry. 862 */ 863 timo = hzto(&atv); 864 if (timo <= 0) 865 goto done; 866 } 867 s = splsched(); 868 if ((l->l_flag & L_SELECT) == 0 || nselcoll != ncoll) { 869 splx(s); 870 goto retry; 871 } 872 l->l_flag &= ~L_SELECT; 873 error = tsleep((caddr_t)&selwait, PSOCK | PCATCH, "select", timo); 874 splx(s); 875 if (error == 0) 876 goto retry; 877 done: 878 l->l_flag &= ~L_SELECT; 879 /* poll is not restarted after signals... */ 880 if (error == ERESTART) 881 error = EINTR; 882 if (error == EWOULDBLOCK) 883 error = 0; 884 if (error == 0) { 885 error = copyout(bits, SCARG(uap, fds), ni); 886 if (error) 887 goto out; 888 } 889 out: 890 if (ni > sizeof(smallbits)) 891 free(bits, M_TEMP); 892 return (error); 893 } 894 895 int 896 pollscan(struct proc *p, struct pollfd *fds, int nfd, register_t *retval) 897 { 898 struct filedesc *fdp; 899 int i, n; 900 struct file *fp; 901 902 fdp = p->p_fd; 903 n = 0; 904 for (i = 0; i < nfd; i++, fds++) { 905 if (fds->fd >= fdp->fd_nfiles) { 906 fds->revents = POLLNVAL; 907 n++; 908 } else if (fds->fd < 0) { 909 fds->revents = 0; 910 } else { 911 if ((fp = fd_getfile(fdp, fds->fd)) == NULL) { 912 fds->revents = POLLNVAL; 913 n++; 914 } else { 915 FILE_USE(fp); 916 fds->revents = (*fp->f_ops->fo_poll)(fp, 917 fds->events | POLLERR | POLLHUP, p); 918 if (fds->revents != 0) 919 n++; 920 FILE_UNUSE(fp, p); 921 } 922 } 923 } 924 *retval = n; 925 return (0); 926 } 927 928 /*ARGSUSED*/ 929 int 930 seltrue(dev_t dev, int events, struct proc *p) 931 { 932 933 return (events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM)); 934 } 935 936 /* 937 * Record a select request. 938 */ 939 void 940 selrecord(struct proc *selector, struct selinfo *sip) 941 { 942 struct lwp *l; 943 struct proc *p; 944 pid_t mypid; 945 int collision; 946 947 mypid = selector->p_pid; 948 if (sip->si_pid == mypid) 949 return; 950 951 collision = 0; 952 if (sip->si_pid && (p = pfind(sip->si_pid))) { 953 for (l = LIST_FIRST(&p->p_lwps); l != NULL; 954 l = LIST_NEXT(l, l_sibling)) { 955 if (l->l_wchan == (caddr_t)&selwait) { 956 collision = 1; 957 sip->si_flags |= SI_COLL; 958 } 959 } 960 } 961 962 if (collision == 0) { 963 sip->si_flags &= ~SI_COLL; 964 sip->si_pid = mypid; 965 } 966 } 967 968 /* 969 * Do a wakeup when a selectable event occurs. 970 */ 971 void 972 selwakeup(sip) 973 struct selinfo *sip; 974 { 975 struct lwp *l; 976 struct proc *p; 977 int s; 978 979 if (sip->si_pid == 0) 980 return; 981 if (sip->si_flags & SI_COLL) { 982 nselcoll++; 983 sip->si_flags &= ~SI_COLL; 984 wakeup((caddr_t)&selwait); 985 } 986 p = pfind(sip->si_pid); 987 sip->si_pid = 0; 988 if (p != NULL) { 989 for (l = LIST_FIRST(&p->p_lwps); l != NULL; 990 l = LIST_NEXT(l, l_sibling)) { 991 SCHED_LOCK(s); 992 if (l->l_wchan == (caddr_t)&selwait) { 993 if (l->l_stat == LSSLEEP) 994 setrunnable(l); 995 else 996 unsleep(l); 997 } else if (l->l_flag & L_SELECT) 998 l->l_flag &= ~L_SELECT; 999 SCHED_UNLOCK(s); 1000 } 1001 } 1002 } 1003
Indexes created Tue Sep 30 07:10:03 GMT 2025