sys_generic.c revision 1.54.2.1
1 /* $NetBSD: sys_generic.c,v 1.54.2.1 2001/03/05 22:49:44 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 "opt_ktrace.h" 44 45 #include <sys/param.h> 46 #include <sys/systm.h> 47 #include <sys/filedesc.h> 48 #include <sys/ioctl.h> 49 #include <sys/file.h> 50 #include <sys/lwp.h> 51 #include <sys/proc.h> 52 #include <sys/socketvar.h> 53 #include <sys/signalvar.h> 54 #include <sys/uio.h> 55 #include <sys/kernel.h> 56 #include <sys/stat.h> 57 #include <sys/malloc.h> 58 #include <sys/poll.h> 59 #ifdef KTRACE 60 #include <sys/ktrace.h> 61 #endif 62 63 #include <sys/mount.h> 64 #include <sys/syscallargs.h> 65 66 int selscan __P((struct proc *, fd_mask *, fd_mask *, int, register_t *)); 67 int pollscan __P((struct proc *, struct pollfd *, int, register_t *)); 68 69 /* 70 * Read system call. 71 */ 72 /* ARGSUSED */ 73 int 74 sys_read(struct lwp *l, void *v, register_t *retval) 75 { 76 struct sys_read_args /* { 77 syscallarg(int) fd; 78 syscallarg(void *) buf; 79 syscallarg(size_t) nbyte; 80 } */ *uap = v; 81 int fd; 82 struct file *fp; 83 struct proc *p; 84 struct filedesc *fdp; 85 86 fd = SCARG(uap, fd); 87 p = l->l_proc; 88 fdp = p->p_fd; 89 if ((u_int)fd >= fdp->fd_nfiles || 90 (fp = fdp->fd_ofiles[fd]) == NULL || 91 (fp->f_iflags & FIF_WANTCLOSE) != 0 || 92 (fp->f_flag & FREAD) == 0) 93 return (EBADF); 94 95 FILE_USE(fp); 96 97 /* dofileread() will unuse the descriptor for us */ 98 return (dofileread(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 99 &fp->f_offset, FOF_UPDATE_OFFSET, retval)); 100 } 101 102 int 103 dofileread(struct proc *p, int fd, struct file *fp, void *buf, size_t nbyte, 104 off_t *offset, int flags, register_t *retval) 105 { 106 struct uio auio; 107 struct iovec aiov; 108 long cnt, error; 109 #ifdef KTRACE 110 struct iovec ktriov; 111 #endif 112 error = 0; 113 114 aiov.iov_base = (caddr_t)buf; 115 aiov.iov_len = nbyte; 116 auio.uio_iov = &aiov; 117 auio.uio_iovcnt = 1; 118 auio.uio_resid = nbyte; 119 auio.uio_rw = UIO_READ; 120 auio.uio_segflg = UIO_USERSPACE; 121 auio.uio_procp = p; 122 123 /* 124 * Reads return ssize_t because -1 is returned on error. Therefore 125 * we must restrict the length to SSIZE_MAX to avoid garbage return 126 * values. 127 */ 128 if (auio.uio_resid > SSIZE_MAX) { 129 error = EINVAL; 130 goto out; 131 } 132 133 #ifdef KTRACE 134 /* 135 * if tracing, save a copy of iovec 136 */ 137 if (KTRPOINT(p, KTR_GENIO)) 138 ktriov = aiov; 139 #endif 140 cnt = auio.uio_resid; 141 error = (*fp->f_ops->fo_read)(fp, offset, &auio, fp->f_cred, flags); 142 if (error) 143 if (auio.uio_resid != cnt && (error == ERESTART || 144 error == EINTR || error == EWOULDBLOCK)) 145 error = 0; 146 cnt -= auio.uio_resid; 147 #ifdef KTRACE 148 if (KTRPOINT(p, KTR_GENIO) && error == 0) 149 ktrgenio(p, fd, UIO_READ, &ktriov, cnt, error); 150 #endif 151 *retval = cnt; 152 out: 153 FILE_UNUSE(fp, p); 154 return (error); 155 } 156 157 /* 158 * Scatter read system call. 159 */ 160 int 161 sys_readv(struct lwp *l, void *v, register_t *retval) 162 { 163 struct sys_readv_args /* { 164 syscallarg(int) fd; 165 syscallarg(const struct iovec *) iovp; 166 syscallarg(int) iovcnt; 167 } */ *uap = v; 168 int fd; 169 struct file *fp; 170 struct proc *p; 171 struct filedesc *fdp; 172 173 fd = SCARG(uap, fd); 174 p = l->l_proc; 175 fdp = p->p_fd; 176 if ((u_int)fd >= fdp->fd_nfiles || 177 (fp = fdp->fd_ofiles[fd]) == NULL || 178 (fp->f_iflags & FIF_WANTCLOSE) != 0 || 179 (fp->f_flag & FREAD) == 0) 180 return (EBADF); 181 182 FILE_USE(fp); 183 184 /* dofilereadv() will unuse the descriptor for us */ 185 return (dofilereadv(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 186 &fp->f_offset, FOF_UPDATE_OFFSET, retval)); 187 } 188 189 int 190 dofilereadv(struct proc *p, int fd, struct file *fp, const struct iovec *iovp, 191 int iovcnt, off_t *offset, int flags, register_t *retval) 192 { 193 struct uio auio; 194 struct iovec *iov, *needfree, aiov[UIO_SMALLIOV]; 195 long i, cnt, error; 196 u_int iovlen; 197 #ifdef KTRACE 198 struct iovec *ktriov; 199 #endif 200 201 error = 0; 202 #ifdef KTRACE 203 ktriov = NULL; 204 #endif 205 /* note: can't use iovlen until iovcnt is validated */ 206 iovlen = iovcnt * sizeof(struct iovec); 207 if ((u_int)iovcnt > UIO_SMALLIOV) { 208 if ((u_int)iovcnt > IOV_MAX) { 209 error = EINVAL; 210 goto out; 211 } 212 iov = malloc(iovlen, M_IOV, M_WAITOK); 213 needfree = iov; 214 } else if ((u_int)iovcnt > 0) { 215 iov = aiov; 216 needfree = NULL; 217 } else { 218 error = EINVAL; 219 goto out; 220 } 221 222 auio.uio_iov = iov; 223 auio.uio_iovcnt = iovcnt; 224 auio.uio_rw = UIO_READ; 225 auio.uio_segflg = UIO_USERSPACE; 226 auio.uio_procp = p; 227 error = copyin(iovp, iov, iovlen); 228 if (error) 229 goto done; 230 auio.uio_resid = 0; 231 for (i = 0; i < iovcnt; i++) { 232 auio.uio_resid += iov->iov_len; 233 /* 234 * Reads return ssize_t because -1 is returned on error. 235 * Therefore we must restrict the length to SSIZE_MAX to 236 * avoid garbage return values. 237 */ 238 if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) { 239 error = EINVAL; 240 goto done; 241 } 242 iov++; 243 } 244 #ifdef KTRACE 245 /* 246 * if tracing, save a copy of iovec 247 */ 248 if (KTRPOINT(p, KTR_GENIO)) { 249 ktriov = malloc(iovlen, M_TEMP, M_WAITOK); 250 memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen); 251 } 252 #endif 253 cnt = auio.uio_resid; 254 error = (*fp->f_ops->fo_read)(fp, offset, &auio, fp->f_cred, flags); 255 if (error) 256 if (auio.uio_resid != cnt && (error == ERESTART || 257 error == EINTR || error == EWOULDBLOCK)) 258 error = 0; 259 cnt -= auio.uio_resid; 260 #ifdef KTRACE 261 if (KTRPOINT(p, KTR_GENIO)) 262 if (error == 0) { 263 ktrgenio(p, fd, UIO_READ, ktriov, cnt, error); 264 free(ktriov, M_TEMP); 265 } 266 #endif 267 *retval = cnt; 268 done: 269 if (needfree) 270 free(needfree, M_IOV); 271 out: 272 FILE_UNUSE(fp, p); 273 return (error); 274 } 275 276 /* 277 * Write system call 278 */ 279 int 280 sys_write(struct lwp *l, void *v, register_t *retval) 281 { 282 struct sys_write_args /* { 283 syscallarg(int) fd; 284 syscallarg(const void *) buf; 285 syscallarg(size_t) nbyte; 286 } */ *uap = v; 287 int fd; 288 struct file *fp; 289 struct proc *p; 290 struct filedesc *fdp; 291 292 fd = SCARG(uap, fd); 293 p = l->l_proc; 294 fdp = p->p_fd; 295 if ((u_int)fd >= fdp->fd_nfiles || 296 (fp = fdp->fd_ofiles[fd]) == NULL || 297 (fp->f_iflags & FIF_WANTCLOSE) != 0 || 298 (fp->f_flag & FWRITE) == 0) 299 return (EBADF); 300 301 FILE_USE(fp); 302 303 /* dofilewrite() will unuse the descriptor for us */ 304 return (dofilewrite(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 305 &fp->f_offset, FOF_UPDATE_OFFSET, retval)); 306 } 307 308 int 309 dofilewrite(struct proc *p, int fd, struct file *fp, const void *buf, 310 size_t nbyte, off_t *offset, int flags, register_t *retval) 311 { 312 struct uio auio; 313 struct iovec aiov; 314 long cnt, error; 315 #ifdef KTRACE 316 struct iovec ktriov; 317 #endif 318 319 error = 0; 320 aiov.iov_base = (caddr_t)buf; /* XXX kills const */ 321 aiov.iov_len = nbyte; 322 auio.uio_iov = &aiov; 323 auio.uio_iovcnt = 1; 324 auio.uio_resid = nbyte; 325 auio.uio_rw = UIO_WRITE; 326 auio.uio_segflg = UIO_USERSPACE; 327 auio.uio_procp = p; 328 329 /* 330 * Writes return ssize_t because -1 is returned on error. Therefore 331 * we must restrict the length to SSIZE_MAX to avoid garbage return 332 * values. 333 */ 334 if (auio.uio_resid > SSIZE_MAX) { 335 error = EINVAL; 336 goto out; 337 } 338 339 #ifdef KTRACE 340 /* 341 * if tracing, save a copy of iovec 342 */ 343 if (KTRPOINT(p, KTR_GENIO)) 344 ktriov = aiov; 345 #endif 346 cnt = auio.uio_resid; 347 error = (*fp->f_ops->fo_write)(fp, offset, &auio, fp->f_cred, flags); 348 if (error) { 349 if (auio.uio_resid != cnt && (error == ERESTART || 350 error == EINTR || error == EWOULDBLOCK)) 351 error = 0; 352 if (error == EPIPE) 353 psignal(p, SIGPIPE); 354 } 355 cnt -= auio.uio_resid; 356 #ifdef KTRACE 357 if (KTRPOINT(p, KTR_GENIO) && error == 0) 358 ktrgenio(p, fd, UIO_WRITE, &ktriov, cnt, error); 359 #endif 360 *retval = cnt; 361 out: 362 FILE_UNUSE(fp, p); 363 return (error); 364 } 365 366 /* 367 * Gather write system call 368 */ 369 int 370 sys_writev(struct lwp *l, void *v, register_t *retval) 371 { 372 struct sys_writev_args /* { 373 syscallarg(int) fd; 374 syscallarg(const struct iovec *) iovp; 375 syscallarg(int) iovcnt; 376 } */ *uap = v; 377 int fd; 378 struct file *fp; 379 struct proc *p; 380 struct filedesc *fdp; 381 382 fd = SCARG(uap, fd); 383 p = l->l_proc; 384 fdp = p->p_fd; 385 if ((u_int)fd >= fdp->fd_nfiles || 386 (fp = fdp->fd_ofiles[fd]) == NULL || 387 (fp->f_iflags & FIF_WANTCLOSE) != 0 || 388 (fp->f_flag & FWRITE) == 0) 389 return (EBADF); 390 391 FILE_USE(fp); 392 393 /* dofilewritev() will unuse the descriptor for us */ 394 return (dofilewritev(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 395 &fp->f_offset, FOF_UPDATE_OFFSET, retval)); 396 } 397 398 int 399 dofilewritev(struct proc *p, int fd, struct file *fp, const struct iovec *iovp, 400 int iovcnt, off_t *offset, int flags, register_t *retval) 401 { 402 struct uio auio; 403 struct iovec *iov, *needfree, aiov[UIO_SMALLIOV]; 404 long i, cnt, error; 405 u_int iovlen; 406 #ifdef KTRACE 407 struct iovec *ktriov; 408 #endif 409 410 error = 0; 411 #ifdef KTRACE 412 ktriov = NULL; 413 #endif 414 /* note: can't use iovlen until iovcnt is validated */ 415 iovlen = iovcnt * sizeof(struct iovec); 416 if ((u_int)iovcnt > UIO_SMALLIOV) { 417 if ((u_int)iovcnt > IOV_MAX) 418 return (EINVAL); 419 iov = malloc(iovlen, M_IOV, M_WAITOK); 420 needfree = iov; 421 } else if ((u_int)iovcnt > 0) { 422 iov = aiov; 423 needfree = NULL; 424 } else { 425 error = EINVAL; 426 goto out; 427 } 428 429 auio.uio_iov = iov; 430 auio.uio_iovcnt = iovcnt; 431 auio.uio_rw = UIO_WRITE; 432 auio.uio_segflg = UIO_USERSPACE; 433 auio.uio_procp = p; 434 error = copyin(iovp, iov, iovlen); 435 if (error) 436 goto done; 437 auio.uio_resid = 0; 438 for (i = 0; i < iovcnt; i++) { 439 auio.uio_resid += iov->iov_len; 440 /* 441 * Writes return ssize_t because -1 is returned on error. 442 * Therefore we must restrict the length to SSIZE_MAX to 443 * avoid garbage return values. 444 */ 445 if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) { 446 error = EINVAL; 447 goto done; 448 } 449 iov++; 450 } 451 #ifdef KTRACE 452 /* 453 * if tracing, save a copy of iovec 454 */ 455 if (KTRPOINT(p, KTR_GENIO)) { 456 ktriov = malloc(iovlen, M_TEMP, M_WAITOK); 457 memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen); 458 } 459 #endif 460 cnt = auio.uio_resid; 461 error = (*fp->f_ops->fo_write)(fp, offset, &auio, fp->f_cred, flags); 462 if (error) { 463 if (auio.uio_resid != cnt && (error == ERESTART || 464 error == EINTR || error == EWOULDBLOCK)) 465 error = 0; 466 if (error == EPIPE) 467 psignal(p, SIGPIPE); 468 } 469 cnt -= auio.uio_resid; 470 #ifdef KTRACE 471 if (KTRPOINT(p, KTR_GENIO)) 472 if (error == 0) { 473 ktrgenio(p, fd, UIO_WRITE, ktriov, cnt, error); 474 free(ktriov, M_TEMP); 475 } 476 #endif 477 *retval = cnt; 478 done: 479 if (needfree) 480 free(needfree, M_IOV); 481 out: 482 FILE_UNUSE(fp, p); 483 return (error); 484 } 485 486 /* 487 * Ioctl system call 488 */ 489 /* ARGSUSED */ 490 int 491 sys_ioctl(struct lwp *l, void *v, register_t *retval) 492 { 493 struct sys_ioctl_args /* { 494 syscallarg(int) fd; 495 syscallarg(u_long) com; 496 syscallarg(caddr_t) data; 497 } */ *uap = v; 498 struct file *fp; 499 struct proc *p; 500 struct filedesc *fdp; 501 u_long com; 502 int error; 503 u_int size; 504 caddr_t data, memp; 505 int tmp; 506 #define STK_PARAMS 128 507 u_long stkbuf[STK_PARAMS/sizeof(u_long)]; 508 509 error = 0; 510 p = l->l_proc; 511 fdp = p->p_fd; 512 if ((u_int)SCARG(uap, fd) >= fdp->fd_nfiles || 513 (fp = fdp->fd_ofiles[SCARG(uap, fd)]) == NULL || 514 (fp->f_iflags & FIF_WANTCLOSE) != 0) 515 return (EBADF); 516 517 FILE_USE(fp); 518 519 if ((fp->f_flag & (FREAD | FWRITE)) == 0) { 520 error = EBADF; 521 goto out; 522 } 523 524 switch (com = SCARG(uap, com)) { 525 case FIONCLEX: 526 fdp->fd_ofileflags[SCARG(uap, fd)] &= ~UF_EXCLOSE; 527 goto out; 528 529 case FIOCLEX: 530 fdp->fd_ofileflags[SCARG(uap, fd)] |= UF_EXCLOSE; 531 goto out; 532 } 533 534 /* 535 * Interpret high order word to find amount of data to be 536 * copied to/from the user's address space. 537 */ 538 size = IOCPARM_LEN(com); 539 if (size > IOCPARM_MAX) { 540 error = ENOTTY; 541 goto out; 542 } 543 memp = NULL; 544 if (size > sizeof(stkbuf)) { 545 memp = (caddr_t)malloc((u_long)size, M_IOCTLOPS, M_WAITOK); 546 data = memp; 547 } else 548 data = (caddr_t)stkbuf; 549 if (com&IOC_IN) { 550 if (size) { 551 error = copyin(SCARG(uap, data), data, size); 552 if (error) { 553 if (memp) 554 free(memp, M_IOCTLOPS); 555 goto out; 556 } 557 } else 558 *(caddr_t *)data = SCARG(uap, data); 559 } else if ((com&IOC_OUT) && size) 560 /* 561 * Zero the buffer so the user always 562 * gets back something deterministic. 563 */ 564 memset(data, 0, size); 565 else if (com&IOC_VOID) 566 *(caddr_t *)data = SCARG(uap, data); 567 568 switch (com) { 569 570 case FIONBIO: 571 if ((tmp = *(int *)data) != 0) 572 fp->f_flag |= FNONBLOCK; 573 else 574 fp->f_flag &= ~FNONBLOCK; 575 error = (*fp->f_ops->fo_ioctl)(fp, FIONBIO, (caddr_t)&tmp, p); 576 break; 577 578 case FIOASYNC: 579 if ((tmp = *(int *)data) != 0) 580 fp->f_flag |= FASYNC; 581 else 582 fp->f_flag &= ~FASYNC; 583 error = (*fp->f_ops->fo_ioctl)(fp, FIOASYNC, (caddr_t)&tmp, p); 584 break; 585 586 case FIOSETOWN: 587 tmp = *(int *)data; 588 if (fp->f_type == DTYPE_SOCKET) { 589 ((struct socket *)fp->f_data)->so_pgid = tmp; 590 error = 0; 591 break; 592 } 593 if (tmp <= 0) { 594 tmp = -tmp; 595 } else { 596 struct proc *p1 = pfind(tmp); 597 if (p1 == 0) { 598 error = ESRCH; 599 break; 600 } 601 tmp = p1->p_pgrp->pg_id; 602 } 603 error = (*fp->f_ops->fo_ioctl) 604 (fp, TIOCSPGRP, (caddr_t)&tmp, p); 605 break; 606 607 case FIOGETOWN: 608 if (fp->f_type == DTYPE_SOCKET) { 609 error = 0; 610 *(int *)data = ((struct socket *)fp->f_data)->so_pgid; 611 break; 612 } 613 error = (*fp->f_ops->fo_ioctl)(fp, TIOCGPGRP, data, p); 614 *(int *)data = -*(int *)data; 615 break; 616 617 default: 618 error = (*fp->f_ops->fo_ioctl)(fp, com, data, p); 619 /* 620 * Copy any data to user, size was 621 * already set and checked above. 622 */ 623 if (error == 0 && (com&IOC_OUT) && size) 624 error = copyout(data, SCARG(uap, data), size); 625 break; 626 } 627 if (memp) 628 free(memp, M_IOCTLOPS); 629 out: 630 FILE_UNUSE(fp, p); 631 return (error); 632 } 633 634 int selwait, nselcoll; 635 636 /* 637 * Select system call. 638 */ 639 int 640 sys_select(struct lwp *l, void *v, register_t *retval) 641 { 642 struct sys_select_args /* { 643 syscallarg(int) nd; 644 syscallarg(fd_set *) in; 645 syscallarg(fd_set *) ou; 646 syscallarg(fd_set *) ex; 647 syscallarg(struct timeval *) tv; 648 } */ *uap = v; 649 struct proc *p; 650 caddr_t bits; 651 char smallbits[howmany(FD_SETSIZE, NFDBITS) * 652 sizeof(fd_mask) * 6]; 653 struct timeval atv; 654 int s, ncoll, error, timo; 655 size_t ni; 656 657 error = 0; 658 p = l->l_proc; 659 if (SCARG(uap, nd) < 0) 660 return (EINVAL); 661 if (SCARG(uap, nd) > p->p_fd->fd_nfiles) { 662 /* forgiving; slightly wrong */ 663 SCARG(uap, nd) = p->p_fd->fd_nfiles; 664 } 665 ni = howmany(SCARG(uap, nd), NFDBITS) * sizeof(fd_mask); 666 if (ni * 6 > sizeof(smallbits)) 667 bits = malloc(ni * 6, M_TEMP, M_WAITOK); 668 else 669 bits = smallbits; 670 671 #define getbits(name, x) \ 672 if (SCARG(uap, name)) { \ 673 error = copyin(SCARG(uap, name), bits + ni * x, ni); \ 674 if (error) \ 675 goto done; \ 676 } else \ 677 memset(bits + ni * x, 0, ni); 678 getbits(in, 0); 679 getbits(ou, 1); 680 getbits(ex, 2); 681 #undef getbits 682 683 if (SCARG(uap, tv)) { 684 error = copyin(SCARG(uap, tv), (caddr_t)&atv, 685 sizeof(atv)); 686 if (error) 687 goto done; 688 if (itimerfix(&atv)) { 689 error = EINVAL; 690 goto done; 691 } 692 s = splclock(); 693 timeradd(&atv, &time, &atv); 694 splx(s); 695 } else 696 timo = 0; 697 retry: 698 ncoll = nselcoll; 699 l->l_flag |= L_SELECT; 700 error = selscan(p, (fd_mask *)(bits + ni * 0), 701 (fd_mask *)(bits + ni * 3), SCARG(uap, nd), retval); 702 if (error || *retval) 703 goto done; 704 if (SCARG(uap, tv)) { 705 /* 706 * We have to recalculate the timeout on every retry. 707 */ 708 timo = hzto(&atv); 709 if (timo <= 0) 710 goto done; 711 } 712 s = splsched(); 713 if ((l->l_flag & L_SELECT) == 0 || nselcoll != ncoll) { 714 splx(s); 715 goto retry; 716 } 717 l->l_flag &= ~L_SELECT; 718 error = tsleep((caddr_t)&selwait, PSOCK | PCATCH, "select", timo); 719 splx(s); 720 if (error == 0) 721 goto retry; 722 done: 723 l->l_flag &= ~L_SELECT; 724 /* select is not restarted after signals... */ 725 if (error == ERESTART) 726 error = EINTR; 727 if (error == EWOULDBLOCK) 728 error = 0; 729 if (error == 0) { 730 731 #define putbits(name, x) \ 732 if (SCARG(uap, name)) { \ 733 error = copyout(bits + ni * x, SCARG(uap, name), ni); \ 734 if (error) \ 735 goto out; \ 736 } 737 putbits(in, 3); 738 putbits(ou, 4); 739 putbits(ex, 5); 740 #undef putbits 741 } 742 out: 743 if (ni * 6 > sizeof(smallbits)) 744 free(bits, M_TEMP); 745 return (error); 746 } 747 748 int 749 selscan(struct proc *p, fd_mask *ibitp, fd_mask *obitp, int nfd, 750 register_t *retval) 751 { 752 struct filedesc *fdp; 753 int msk, i, j, fd, n; 754 fd_mask ibits, obits; 755 struct file *fp; 756 static int flag[3] = { POLLRDNORM | POLLHUP | POLLERR, 757 POLLWRNORM | POLLHUP | POLLERR, 758 POLLRDBAND }; 759 760 fdp = p->p_fd; 761 n = 0; 762 for (msk = 0; msk < 3; msk++) { 763 for (i = 0; i < nfd; i += NFDBITS) { 764 ibits = *ibitp++; 765 obits = 0; 766 while ((j = ffs(ibits)) && (fd = i + --j) < nfd) { 767 ibits &= ~(1 << j); 768 fp = fdp->fd_ofiles[fd]; 769 if (fp == NULL || 770 (fp->f_iflags & FIF_WANTCLOSE) != 0) 771 return (EBADF); 772 FILE_USE(fp); 773 if ((*fp->f_ops->fo_poll)(fp, flag[msk], p)) { 774 obits |= (1 << j); 775 n++; 776 } 777 FILE_UNUSE(fp, p); 778 } 779 *obitp++ = obits; 780 } 781 } 782 *retval = n; 783 return (0); 784 } 785 786 /* 787 * Poll system call. 788 */ 789 int 790 sys_poll(struct lwp *l, void *v, register_t *retval) 791 { 792 struct sys_poll_args /* { 793 syscallarg(struct pollfd *) fds; 794 syscallarg(u_int) nfds; 795 syscallarg(int) timeout; 796 } */ *uap = v; 797 struct proc *p; 798 caddr_t bits; 799 char smallbits[32 * sizeof(struct pollfd)]; 800 struct timeval atv; 801 int s, ncoll, error, timo; 802 size_t ni; 803 804 error = 0; 805 p = l->l_proc; 806 if (SCARG(uap, nfds) > p->p_fd->fd_nfiles) { 807 /* forgiving; slightly wrong */ 808 SCARG(uap, nfds) = p->p_fd->fd_nfiles; 809 } 810 ni = SCARG(uap, nfds) * sizeof(struct pollfd); 811 if (ni > sizeof(smallbits)) 812 bits = malloc(ni, M_TEMP, M_WAITOK); 813 else 814 bits = smallbits; 815 816 error = copyin(SCARG(uap, fds), bits, ni); 817 if (error) 818 goto done; 819 820 if (SCARG(uap, timeout) != INFTIM) { 821 atv.tv_sec = SCARG(uap, timeout) / 1000; 822 atv.tv_usec = (SCARG(uap, timeout) % 1000) * 1000; 823 if (itimerfix(&atv)) { 824 error = EINVAL; 825 goto done; 826 } 827 s = splclock(); 828 timeradd(&atv, &time, &atv); 829 splx(s); 830 } else 831 timo = 0; 832 retry: 833 ncoll = nselcoll; 834 l->l_flag |= L_SELECT; 835 error = pollscan(p, (struct pollfd *)bits, SCARG(uap, nfds), retval); 836 if (error || *retval) 837 goto done; 838 if (SCARG(uap, timeout) != INFTIM) { 839 /* 840 * We have to recalculate the timeout on every retry. 841 */ 842 timo = hzto(&atv); 843 if (timo <= 0) 844 goto done; 845 } 846 s = splsched(); 847 if ((l->l_flag & L_SELECT) == 0 || nselcoll != ncoll) { 848 splx(s); 849 goto retry; 850 } 851 l->l_flag &= ~L_SELECT; 852 error = tsleep((caddr_t)&selwait, PSOCK | PCATCH, "select", timo); 853 splx(s); 854 if (error == 0) 855 goto retry; 856 done: 857 l->l_flag &= ~L_SELECT; 858 /* poll is not restarted after signals... */ 859 if (error == ERESTART) 860 error = EINTR; 861 if (error == EWOULDBLOCK) 862 error = 0; 863 if (error == 0) { 864 error = copyout(bits, SCARG(uap, fds), ni); 865 if (error) 866 goto out; 867 } 868 out: 869 if (ni > sizeof(smallbits)) 870 free(bits, M_TEMP); 871 return (error); 872 } 873 874 int 875 pollscan(struct proc *p, struct pollfd *fds, int nfd, register_t *retval) 876 { 877 struct filedesc *fdp; 878 int i, n; 879 struct file *fp; 880 881 fdp = p->p_fd; 882 n = 0; 883 for (i = 0; i < nfd; i++, fds++) { 884 if ((u_int)fds->fd >= fdp->fd_nfiles) { 885 fds->revents = POLLNVAL; 886 n++; 887 } else { 888 fp = fdp->fd_ofiles[fds->fd]; 889 if (fp == NULL || 890 (fp->f_iflags & FIF_WANTCLOSE) != 0) { 891 fds->revents = POLLNVAL; 892 n++; 893 } else { 894 FILE_USE(fp); 895 fds->revents = (*fp->f_ops->fo_poll)(fp, 896 fds->events | POLLERR | POLLHUP, p); 897 if (fds->revents != 0) 898 n++; 899 FILE_UNUSE(fp, p); 900 } 901 } 902 } 903 *retval = n; 904 return (0); 905 } 906 907 /*ARGSUSED*/ 908 int 909 seltrue(dev_t dev, int events, struct proc *p) 910 { 911 912 return (events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM)); 913 } 914 915 /* 916 * Record a select request. 917 */ 918 void 919 selrecord(struct proc *selector, struct selinfo *sip) 920 { 921 struct lwp *l; 922 struct proc *p; 923 pid_t mypid; 924 925 mypid = selector->p_pid; 926 if (sip->si_pid == mypid) 927 return; 928 if (sip->si_pid && (p = pfind(sip->si_pid))) { 929 for (l = LIST_FIRST(&p->p_lwps); l != NULL; 930 l = LIST_NEXT(l, l_sibling)) { 931 if (l->l_wchan == (caddr_t)&selwait) 932 sip->si_flags |= SI_COLL; 933 } 934 } else 935 sip->si_pid = mypid; 936 } 937 938 /* 939 * Do a wakeup when a selectable event occurs. 940 */ 941 void 942 selwakeup(sip) 943 struct selinfo *sip; 944 { 945 struct lwp *l; 946 struct proc *p; 947 int s; 948 949 if (sip->si_pid == 0) 950 return; 951 if (sip->si_flags & SI_COLL) { 952 nselcoll++; 953 sip->si_flags &= ~SI_COLL; 954 wakeup((caddr_t)&selwait); 955 } 956 p = pfind(sip->si_pid); 957 sip->si_pid = 0; 958 if (p != NULL) { 959 for (l = LIST_FIRST(&p->p_lwps); l != NULL; 960 l = LIST_NEXT(l, l_sibling)) { 961 SCHED_LOCK(s); 962 if (l->l_wchan == (caddr_t)&selwait) { 963 if (l->l_stat == LSSLEEP) 964 setrunnable(l); 965 else 966 unsleep(l); 967 } else if (l->l_flag & L_SELECT) 968 l->l_flag &= ~L_SELECT; 969 SCHED_UNLOCK(s); 970 } 971 } 972 } 973
Indexes created Tue Sep 30 07:10:03 GMT 2025