sys_generic.c revision 1.40
1 /* $NetBSD: sys_generic.c,v 1.40 1998/07/28 17:58:29 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 "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/proc.h> 51 #include <sys/socketvar.h> 52 #include <sys/signalvar.h> 53 #include <sys/uio.h> 54 #include <sys/kernel.h> 55 #include <sys/stat.h> 56 #include <sys/malloc.h> 57 #include <sys/poll.h> 58 #ifdef KTRACE 59 #include <sys/ktrace.h> 60 #endif 61 62 #include <sys/mount.h> 63 #include <sys/syscallargs.h> 64 65 int selscan __P((struct proc *, fd_mask *, fd_mask *, int, register_t *)); 66 int pollscan __P((struct proc *, struct pollfd *, int, register_t *)); 67 68 /* 69 * Read system call. 70 */ 71 /* ARGSUSED */ 72 int 73 sys_read(p, v, retval) 74 struct proc *p; 75 void *v; 76 register_t *retval; 77 { 78 register struct sys_read_args /* { 79 syscallarg(int) fd; 80 syscallarg(void *) buf; 81 syscallarg(size_t) nbyte; 82 } */ *uap = v; 83 int fd = SCARG(uap, fd); 84 register struct file *fp; 85 register struct filedesc *fdp = p->p_fd; 86 87 if ((u_int)fd >= fdp->fd_nfiles || 88 (fp = fdp->fd_ofiles[fd]) == NULL || 89 (fp->f_flag & FREAD) == 0) 90 return (EBADF); 91 92 return (dofileread(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 93 &fp->f_offset, FOF_UPDATE_OFFSET, retval)); 94 } 95 96 int 97 dofileread(p, fd, fp, buf, nbyte, offset, flags, retval) 98 struct proc *p; 99 int fd; 100 struct file *fp; 101 void *buf; 102 size_t nbyte; 103 off_t *offset; 104 int flags; 105 register_t *retval; 106 { 107 struct uio auio; 108 struct iovec aiov; 109 long cnt, error = 0; 110 #ifdef KTRACE 111 struct iovec ktriov; 112 #endif 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 return (EINVAL); 130 131 #ifdef KTRACE 132 /* 133 * if tracing, save a copy of iovec 134 */ 135 if (KTRPOINT(p, KTR_GENIO)) 136 ktriov = aiov; 137 #endif 138 cnt = auio.uio_resid; 139 error = (*fp->f_ops->fo_read)(fp, offset, &auio, fp->f_cred, flags); 140 if (error) 141 if (auio.uio_resid != cnt && (error == ERESTART || 142 error == EINTR || error == EWOULDBLOCK)) 143 error = 0; 144 cnt -= auio.uio_resid; 145 #ifdef KTRACE 146 if (KTRPOINT(p, KTR_GENIO) && error == 0) 147 ktrgenio(p->p_tracep, fd, UIO_READ, &ktriov, cnt, error); 148 #endif 149 *retval = cnt; 150 return (error); 151 } 152 153 /* 154 * Scatter read system call. 155 */ 156 int 157 sys_readv(p, v, retval) 158 struct proc *p; 159 void *v; 160 register_t *retval; 161 { 162 register struct sys_readv_args /* { 163 syscallarg(int) fd; 164 syscallarg(const struct iovec *) iovp; 165 syscallarg(int) iovcnt; 166 } */ *uap = v; 167 int fd = SCARG(uap, fd); 168 register struct file *fp; 169 register struct filedesc *fdp = p->p_fd; 170 171 if ((u_int)fd >= fdp->fd_nfiles || 172 (fp = fdp->fd_ofiles[fd]) == NULL || 173 (fp->f_flag & FREAD) == 0) 174 return (EBADF); 175 176 return (dofilereadv(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 177 &fp->f_offset, FOF_UPDATE_OFFSET, retval)); 178 } 179 180 int 181 dofilereadv(p, fd, fp, iovp, iovcnt, offset, flags, retval) 182 struct proc *p; 183 int fd; 184 struct file *fp; 185 const struct iovec *iovp; 186 int iovcnt; 187 off_t *offset; 188 int flags; 189 register_t *retval; 190 { 191 struct uio auio; 192 register struct iovec *iov; 193 struct iovec *needfree; 194 struct iovec aiov[UIO_SMALLIOV]; 195 long i, cnt, error = 0; 196 u_int iovlen; 197 #ifdef KTRACE 198 struct iovec *ktriov = NULL; 199 #endif 200 201 /* note: can't use iovlen until iovcnt is validated */ 202 iovlen = iovcnt * sizeof (struct iovec); 203 if ((u_int)iovcnt > UIO_SMALLIOV) { 204 if ((u_int)iovcnt > UIO_MAXIOV) 205 return (EINVAL); 206 MALLOC(iov, struct iovec *, iovlen, M_IOV, M_WAITOK); 207 needfree = iov; 208 } else { 209 iov = aiov; 210 needfree = NULL; 211 } 212 auio.uio_iov = iov; 213 auio.uio_iovcnt = iovcnt; 214 auio.uio_rw = UIO_READ; 215 auio.uio_segflg = UIO_USERSPACE; 216 auio.uio_procp = p; 217 error = copyin(iovp, iov, iovlen); 218 if (error) 219 goto done; 220 auio.uio_resid = 0; 221 for (i = 0; i < iovcnt; i++) { 222 auio.uio_resid += iov->iov_len; 223 /* 224 * Reads return ssize_t because -1 is returned on error. 225 * Therefore we must restrict the length to SSIZE_MAX to 226 * avoid garbage return values. 227 */ 228 if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) { 229 error = EINVAL; 230 goto done; 231 } 232 iov++; 233 } 234 #ifdef KTRACE 235 /* 236 * if tracing, save a copy of iovec 237 */ 238 if (KTRPOINT(p, KTR_GENIO)) { 239 MALLOC(ktriov, struct iovec *, iovlen, M_TEMP, M_WAITOK); 240 bcopy((caddr_t)auio.uio_iov, (caddr_t)ktriov, iovlen); 241 } 242 #endif 243 cnt = auio.uio_resid; 244 error = (*fp->f_ops->fo_read)(fp, offset, &auio, fp->f_cred, flags); 245 if (error) 246 if (auio.uio_resid != cnt && (error == ERESTART || 247 error == EINTR || error == EWOULDBLOCK)) 248 error = 0; 249 cnt -= auio.uio_resid; 250 #ifdef KTRACE 251 if (KTRPOINT(p, KTR_GENIO)) 252 if (error == 0) { 253 ktrgenio(p->p_tracep, fd, UIO_READ, ktriov, cnt, 254 error); 255 FREE(ktriov, M_TEMP); 256 } 257 #endif 258 *retval = cnt; 259 done: 260 if (needfree) 261 FREE(needfree, M_IOV); 262 return (error); 263 } 264 265 /* 266 * Write system call 267 */ 268 int 269 sys_write(p, v, retval) 270 struct proc *p; 271 void *v; 272 register_t *retval; 273 { 274 register struct sys_write_args /* { 275 syscallarg(int) fd; 276 syscallarg(const void *) buf; 277 syscallarg(size_t) nbyte; 278 } */ *uap = v; 279 int fd = SCARG(uap, fd); 280 register struct file *fp; 281 register struct filedesc *fdp = p->p_fd; 282 283 if ((u_int)fd >= fdp->fd_nfiles || 284 (fp = fdp->fd_ofiles[fd]) == NULL || 285 (fp->f_flag & FWRITE) == 0) 286 return (EBADF); 287 288 return (dofilewrite(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 289 &fp->f_offset, FOF_UPDATE_OFFSET, retval)); 290 } 291 292 int 293 dofilewrite(p, fd, fp, buf, nbyte, offset, flags, retval) 294 struct proc *p; 295 int fd; 296 struct file *fp; 297 const void *buf; 298 size_t nbyte; 299 off_t *offset; 300 int flags; 301 register_t *retval; 302 { 303 struct uio auio; 304 struct iovec aiov; 305 long cnt, error = 0; 306 #ifdef KTRACE 307 struct iovec ktriov; 308 #endif 309 310 aiov.iov_base = (caddr_t)buf; /* XXX kills const */ 311 aiov.iov_len = nbyte; 312 auio.uio_iov = &aiov; 313 auio.uio_iovcnt = 1; 314 auio.uio_resid = nbyte; 315 auio.uio_rw = UIO_WRITE; 316 auio.uio_segflg = UIO_USERSPACE; 317 auio.uio_procp = p; 318 319 /* 320 * Writes return ssize_t because -1 is returned on error. Therefore 321 * we must restrict the length to SSIZE_MAX to avoid garbage return 322 * values. 323 */ 324 if (auio.uio_resid > SSIZE_MAX) 325 return (EINVAL); 326 327 #ifdef KTRACE 328 /* 329 * if tracing, save a copy of iovec 330 */ 331 if (KTRPOINT(p, KTR_GENIO)) 332 ktriov = aiov; 333 #endif 334 cnt = auio.uio_resid; 335 error = (*fp->f_ops->fo_write)(fp, offset, &auio, fp->f_cred, flags); 336 if (error) { 337 if (auio.uio_resid != cnt && (error == ERESTART || 338 error == EINTR || error == EWOULDBLOCK)) 339 error = 0; 340 if (error == EPIPE) 341 psignal(p, SIGPIPE); 342 } 343 cnt -= auio.uio_resid; 344 #ifdef KTRACE 345 if (KTRPOINT(p, KTR_GENIO) && error == 0) 346 ktrgenio(p->p_tracep, fd, UIO_WRITE, &ktriov, cnt, error); 347 #endif 348 *retval = cnt; 349 return (error); 350 } 351 352 /* 353 * Gather write system call 354 */ 355 int 356 sys_writev(p, v, retval) 357 struct proc *p; 358 void *v; 359 register_t *retval; 360 { 361 register struct sys_writev_args /* { 362 syscallarg(int) fd; 363 syscallarg(const struct iovec *) iovp; 364 syscallarg(int) iovcnt; 365 } */ *uap = v; 366 int fd = SCARG(uap, fd); 367 register struct file *fp; 368 register struct filedesc *fdp = p->p_fd; 369 370 if ((u_int)fd >= fdp->fd_nfiles || 371 (fp = fdp->fd_ofiles[fd]) == NULL || 372 (fp->f_flag & FWRITE) == 0) 373 return (EBADF); 374 375 return (dofilewritev(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 376 &fp->f_offset, FOF_UPDATE_OFFSET, retval)); 377 } 378 379 int 380 dofilewritev(p, fd, fp, iovp, iovcnt, offset, flags, retval) 381 struct proc *p; 382 int fd; 383 struct file *fp; 384 const struct iovec *iovp; 385 int iovcnt; 386 off_t *offset; 387 int flags; 388 register_t *retval; 389 { 390 struct uio auio; 391 register struct iovec *iov; 392 struct iovec *needfree; 393 struct iovec aiov[UIO_SMALLIOV]; 394 long i, cnt, error = 0; 395 u_int iovlen; 396 #ifdef KTRACE 397 struct iovec *ktriov = NULL; 398 #endif 399 400 /* note: can't use iovlen until iovcnt is validated */ 401 iovlen = iovcnt * sizeof (struct iovec); 402 if ((u_int)iovcnt > UIO_SMALLIOV) { 403 if ((u_int)iovcnt > UIO_MAXIOV) 404 return (EINVAL); 405 MALLOC(iov, struct iovec *, iovlen, M_IOV, M_WAITOK); 406 needfree = iov; 407 } else { 408 iov = aiov; 409 needfree = NULL; 410 } 411 auio.uio_iov = iov; 412 auio.uio_iovcnt = iovcnt; 413 auio.uio_rw = UIO_WRITE; 414 auio.uio_segflg = UIO_USERSPACE; 415 auio.uio_procp = p; 416 error = copyin(iovp, iov, iovlen); 417 if (error) 418 goto done; 419 auio.uio_resid = 0; 420 for (i = 0; i < iovcnt; i++) { 421 auio.uio_resid += iov->iov_len; 422 /* 423 * Writes return ssize_t because -1 is returned on error. 424 * Therefore we must restrict the length to SSIZE_MAX to 425 * avoid garbage return values. 426 */ 427 if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) { 428 error = EINVAL; 429 goto done; 430 } 431 iov++; 432 } 433 #ifdef KTRACE 434 /* 435 * if tracing, save a copy of iovec 436 */ 437 if (KTRPOINT(p, KTR_GENIO)) { 438 MALLOC(ktriov, struct iovec *, iovlen, M_TEMP, M_WAITOK); 439 bcopy((caddr_t)auio.uio_iov, (caddr_t)ktriov, iovlen); 440 } 441 #endif 442 cnt = auio.uio_resid; 443 error = (*fp->f_ops->fo_write)(fp, offset, &auio, fp->f_cred, flags); 444 if (error) { 445 if (auio.uio_resid != cnt && (error == ERESTART || 446 error == EINTR || error == EWOULDBLOCK)) 447 error = 0; 448 if (error == EPIPE) 449 psignal(p, SIGPIPE); 450 } 451 cnt -= auio.uio_resid; 452 #ifdef KTRACE 453 if (KTRPOINT(p, KTR_GENIO)) 454 if (error == 0) { 455 ktrgenio(p->p_tracep, fd, UIO_WRITE, ktriov, cnt, 456 error); 457 FREE(ktriov, M_TEMP); 458 } 459 #endif 460 *retval = cnt; 461 done: 462 if (needfree) 463 FREE(needfree, M_IOV); 464 return (error); 465 } 466 467 /* 468 * Ioctl system call 469 */ 470 /* ARGSUSED */ 471 int 472 sys_ioctl(p, v, retval) 473 struct proc *p; 474 void *v; 475 register_t *retval; 476 { 477 register struct sys_ioctl_args /* { 478 syscallarg(int) fd; 479 syscallarg(u_long) com; 480 syscallarg(caddr_t) data; 481 } */ *uap = v; 482 register struct file *fp; 483 register struct filedesc *fdp; 484 register u_long com; 485 register int error; 486 register u_int size; 487 caddr_t data, memp; 488 int tmp; 489 #define STK_PARAMS 128 490 char stkbuf[STK_PARAMS]; 491 492 fdp = p->p_fd; 493 if ((u_int)SCARG(uap, fd) >= fdp->fd_nfiles || 494 (fp = fdp->fd_ofiles[SCARG(uap, fd)]) == NULL) 495 return (EBADF); 496 497 if ((fp->f_flag & (FREAD | FWRITE)) == 0) 498 return (EBADF); 499 500 switch (com = SCARG(uap, com)) { 501 case FIONCLEX: 502 fdp->fd_ofileflags[SCARG(uap, fd)] &= ~UF_EXCLOSE; 503 return (0); 504 case FIOCLEX: 505 fdp->fd_ofileflags[SCARG(uap, fd)] |= UF_EXCLOSE; 506 return (0); 507 } 508 509 /* 510 * Interpret high order word to find amount of data to be 511 * copied to/from the user's address space. 512 */ 513 size = IOCPARM_LEN(com); 514 if (size > IOCPARM_MAX) 515 return (ENOTTY); 516 memp = NULL; 517 if (size > sizeof (stkbuf)) { 518 memp = (caddr_t)malloc((u_long)size, M_IOCTLOPS, M_WAITOK); 519 data = memp; 520 } else 521 data = stkbuf; 522 if (com&IOC_IN) { 523 if (size) { 524 error = copyin(SCARG(uap, data), data, size); 525 if (error) { 526 if (memp) 527 free(memp, M_IOCTLOPS); 528 return (error); 529 } 530 } else 531 *(caddr_t *)data = SCARG(uap, data); 532 } else if ((com&IOC_OUT) && size) 533 /* 534 * Zero the buffer so the user always 535 * gets back something deterministic. 536 */ 537 bzero(data, size); 538 else if (com&IOC_VOID) 539 *(caddr_t *)data = SCARG(uap, data); 540 541 switch (com) { 542 543 case FIONBIO: 544 if ((tmp = *(int *)data) != 0) 545 fp->f_flag |= FNONBLOCK; 546 else 547 fp->f_flag &= ~FNONBLOCK; 548 error = (*fp->f_ops->fo_ioctl)(fp, FIONBIO, (caddr_t)&tmp, p); 549 break; 550 551 case FIOASYNC: 552 if ((tmp = *(int *)data) != 0) 553 fp->f_flag |= FASYNC; 554 else 555 fp->f_flag &= ~FASYNC; 556 error = (*fp->f_ops->fo_ioctl)(fp, FIOASYNC, (caddr_t)&tmp, p); 557 break; 558 559 case FIOSETOWN: 560 tmp = *(int *)data; 561 if (fp->f_type == DTYPE_SOCKET) { 562 ((struct socket *)fp->f_data)->so_pgid = tmp; 563 error = 0; 564 break; 565 } 566 if (tmp <= 0) { 567 tmp = -tmp; 568 } else { 569 struct proc *p1 = pfind(tmp); 570 if (p1 == 0) { 571 error = ESRCH; 572 break; 573 } 574 tmp = p1->p_pgrp->pg_id; 575 } 576 error = (*fp->f_ops->fo_ioctl) 577 (fp, TIOCSPGRP, (caddr_t)&tmp, p); 578 break; 579 580 case FIOGETOWN: 581 if (fp->f_type == DTYPE_SOCKET) { 582 error = 0; 583 *(int *)data = ((struct socket *)fp->f_data)->so_pgid; 584 break; 585 } 586 error = (*fp->f_ops->fo_ioctl)(fp, TIOCGPGRP, data, p); 587 *(int *)data = -*(int *)data; 588 break; 589 590 default: 591 error = (*fp->f_ops->fo_ioctl)(fp, com, data, p); 592 /* 593 * Copy any data to user, size was 594 * already set and checked above. 595 */ 596 if (error == 0 && (com&IOC_OUT) && size) 597 error = copyout(data, SCARG(uap, data), size); 598 break; 599 } 600 if (memp) 601 free(memp, M_IOCTLOPS); 602 return (error); 603 } 604 605 int selwait, nselcoll; 606 607 /* 608 * Select system call. 609 */ 610 int 611 sys_select(p, v, retval) 612 register struct proc *p; 613 void *v; 614 register_t *retval; 615 { 616 register struct sys_select_args /* { 617 syscallarg(int) nd; 618 syscallarg(fd_set *) in; 619 syscallarg(fd_set *) ou; 620 syscallarg(fd_set *) ex; 621 syscallarg(struct timeval *) tv; 622 } */ *uap = v; 623 caddr_t bits; 624 char smallbits[howmany(FD_SETSIZE, NFDBITS) * sizeof(fd_mask) * 6]; 625 struct timeval atv; 626 int s, ncoll, error = 0, timo; 627 size_t ni; 628 629 if (SCARG(uap, nd) < 0) 630 return (EINVAL); 631 if (SCARG(uap, nd) > p->p_fd->fd_nfiles) { 632 /* forgiving; slightly wrong */ 633 SCARG(uap, nd) = p->p_fd->fd_nfiles; 634 } 635 ni = howmany(SCARG(uap, nd), NFDBITS) * sizeof(fd_mask); 636 if (ni * 6 > sizeof(smallbits)) 637 bits = malloc(ni * 6, M_TEMP, M_WAITOK); 638 else 639 bits = smallbits; 640 641 #define getbits(name, x) \ 642 if (SCARG(uap, name)) { \ 643 error = copyin(SCARG(uap, name), bits + ni * x, ni); \ 644 if (error) \ 645 goto done; \ 646 } else \ 647 bzero(bits + ni * x, ni); 648 getbits(in, 0); 649 getbits(ou, 1); 650 getbits(ex, 2); 651 #undef getbits 652 653 if (SCARG(uap, tv)) { 654 error = copyin(SCARG(uap, tv), (caddr_t)&atv, 655 sizeof (atv)); 656 if (error) 657 goto done; 658 if (itimerfix(&atv)) { 659 error = EINVAL; 660 goto done; 661 } 662 s = splclock(); 663 timeradd(&atv, &time, &atv); 664 timo = hzto(&atv); 665 /* 666 * Avoid inadvertently sleeping forever. 667 */ 668 if (timo == 0) 669 timo = 1; 670 splx(s); 671 } else 672 timo = 0; 673 retry: 674 ncoll = nselcoll; 675 p->p_flag |= P_SELECT; 676 error = selscan(p, (fd_mask *)(bits + ni * 0), 677 (fd_mask *)(bits + ni * 3), SCARG(uap, nd), retval); 678 if (error || *retval) 679 goto done; 680 s = splhigh(); 681 if (timo && timercmp(&time, &atv, >=)) { 682 splx(s); 683 goto done; 684 } 685 if ((p->p_flag & P_SELECT) == 0 || nselcoll != ncoll) { 686 splx(s); 687 goto retry; 688 } 689 p->p_flag &= ~P_SELECT; 690 error = tsleep((caddr_t)&selwait, PSOCK | PCATCH, "select", timo); 691 splx(s); 692 if (error == 0) 693 goto retry; 694 done: 695 p->p_flag &= ~P_SELECT; 696 /* select is not restarted after signals... */ 697 if (error == ERESTART) 698 error = EINTR; 699 if (error == EWOULDBLOCK) 700 error = 0; 701 if (error == 0) { 702 #define putbits(name, x) \ 703 if (SCARG(uap, name)) { \ 704 error = copyout(bits + ni * x, SCARG(uap, name), ni); \ 705 if (error) \ 706 goto out; \ 707 } 708 putbits(in, 3); 709 putbits(ou, 4); 710 putbits(ex, 5); 711 #undef putbits 712 } 713 out: 714 if (ni * 6 > sizeof(smallbits)) 715 free(bits, M_TEMP); 716 return (error); 717 } 718 719 int 720 selscan(p, ibitp, obitp, nfd, retval) 721 struct proc *p; 722 fd_mask *ibitp, *obitp; 723 int nfd; 724 register_t *retval; 725 { 726 register struct filedesc *fdp = p->p_fd; 727 register int msk, i, j, fd; 728 register fd_mask ibits, obits; 729 struct file *fp; 730 int n = 0; 731 static int flag[3] = { POLLRDNORM | POLLHUP | POLLERR, 732 POLLWRNORM | POLLHUP | POLLERR, 733 POLLRDBAND }; 734 735 for (msk = 0; msk < 3; msk++) { 736 for (i = 0; i < nfd; i += NFDBITS) { 737 ibits = *ibitp++; 738 obits = 0; 739 while ((j = ffs(ibits)) && (fd = i + --j) < nfd) { 740 ibits &= ~(1 << j); 741 fp = fdp->fd_ofiles[fd]; 742 if (fp == NULL) 743 return (EBADF); 744 if ((*fp->f_ops->fo_poll)(fp, flag[msk], p)) { 745 obits |= (1 << j); 746 n++; 747 } 748 } 749 *obitp++ = obits; 750 } 751 } 752 *retval = n; 753 return (0); 754 } 755 756 /* 757 * Poll system call. 758 */ 759 int 760 sys_poll(p, v, retval) 761 register struct proc *p; 762 void *v; 763 register_t *retval; 764 { 765 register struct sys_poll_args /* { 766 syscallarg(struct pollfd *) fds; 767 syscallarg(u_int) nfds; 768 syscallarg(int) timeout; 769 } */ *uap = v; 770 caddr_t bits; 771 char smallbits[32 * sizeof(struct pollfd)]; 772 struct timeval atv; 773 int s, ncoll, error = 0, timo; 774 size_t ni; 775 776 if (SCARG(uap, nfds) > p->p_fd->fd_nfiles) { 777 /* forgiving; slightly wrong */ 778 SCARG(uap, nfds) = p->p_fd->fd_nfiles; 779 } 780 ni = SCARG(uap, nfds) * sizeof(struct pollfd); 781 if (ni > sizeof(smallbits)) 782 bits = malloc(ni, M_TEMP, M_WAITOK); 783 else 784 bits = smallbits; 785 786 error = copyin(SCARG(uap, fds), bits, ni); 787 if (error) 788 goto done; 789 790 if (SCARG(uap, timeout) != INFTIM) { 791 atv.tv_sec = SCARG(uap, timeout) / 1000; 792 atv.tv_usec = (SCARG(uap, timeout) % 1000) * 1000; 793 if (itimerfix(&atv)) { 794 error = EINVAL; 795 goto done; 796 } 797 s = splclock(); 798 timeradd(&atv, &time, &atv); 799 timo = hzto(&atv); 800 /* 801 * Avoid inadvertently sleeping forever. 802 */ 803 if (timo == 0) 804 timo = 1; 805 splx(s); 806 } else 807 timo = 0; 808 retry: 809 ncoll = nselcoll; 810 p->p_flag |= P_SELECT; 811 error = pollscan(p, (struct pollfd *)bits, SCARG(uap, nfds), retval); 812 if (error || *retval) 813 goto done; 814 s = splhigh(); 815 if (timo && timercmp(&time, &atv, >=)) { 816 splx(s); 817 goto done; 818 } 819 if ((p->p_flag & P_SELECT) == 0 || nselcoll != ncoll) { 820 splx(s); 821 goto retry; 822 } 823 p->p_flag &= ~P_SELECT; 824 error = tsleep((caddr_t)&selwait, PSOCK | PCATCH, "select", timo); 825 splx(s); 826 if (error == 0) 827 goto retry; 828 done: 829 p->p_flag &= ~P_SELECT; 830 /* poll is not restarted after signals... */ 831 if (error == ERESTART) 832 error = EINTR; 833 if (error == EWOULDBLOCK) 834 error = 0; 835 if (error == 0) { 836 error = copyout(bits, SCARG(uap, fds), ni); 837 if (error) 838 goto out; 839 } 840 out: 841 if (ni > sizeof(smallbits)) 842 free(bits, M_TEMP); 843 return (error); 844 } 845 846 int 847 pollscan(p, fds, nfd, retval) 848 struct proc *p; 849 struct pollfd *fds; 850 int nfd; 851 register_t *retval; 852 { 853 register struct filedesc *fdp = p->p_fd; 854 int i; 855 struct file *fp; 856 int n = 0; 857 858 for (i = 0; i < nfd; i++, fds++) { 859 if ((u_int)fds->fd >= fdp->fd_nfiles) { 860 fds->revents = POLLNVAL; 861 n++; 862 } else { 863 fp = fdp->fd_ofiles[fds->fd]; 864 if (fp == 0) { 865 fds->revents = POLLNVAL; 866 n++; 867 } else { 868 fds->revents = (*fp->f_ops->fo_poll)(fp, 869 fds->events | POLLERR | POLLHUP, p); 870 if (fds->revents != 0) 871 n++; 872 } 873 } 874 } 875 *retval = n; 876 return (0); 877 } 878 879 /*ARGSUSED*/ 880 int 881 seltrue(dev, events, p) 882 dev_t dev; 883 int events; 884 struct proc *p; 885 { 886 887 return (events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM)); 888 } 889 890 /* 891 * Record a select request. 892 */ 893 void 894 selrecord(selector, sip) 895 struct proc *selector; 896 struct selinfo *sip; 897 { 898 struct proc *p; 899 pid_t mypid; 900 901 mypid = selector->p_pid; 902 if (sip->si_pid == mypid) 903 return; 904 if (sip->si_pid && (p = pfind(sip->si_pid)) && 905 p->p_wchan == (caddr_t)&selwait) 906 sip->si_flags |= SI_COLL; 907 else 908 sip->si_pid = mypid; 909 } 910 911 /* 912 * Do a wakeup when a selectable event occurs. 913 */ 914 void 915 selwakeup(sip) 916 register struct selinfo *sip; 917 { 918 register struct proc *p; 919 int s; 920 921 if (sip->si_pid == 0) 922 return; 923 if (sip->si_flags & SI_COLL) { 924 nselcoll++; 925 sip->si_flags &= ~SI_COLL; 926 wakeup((caddr_t)&selwait); 927 } 928 p = pfind(sip->si_pid); 929 sip->si_pid = 0; 930 if (p != NULL) { 931 s = splhigh(); 932 if (p->p_wchan == (caddr_t)&selwait) { 933 if (p->p_stat == SSLEEP) 934 setrunnable(p); 935 else 936 unsleep(p); 937 } else if (p->p_flag & P_SELECT) 938 p->p_flag &= ~P_SELECT; 939 splx(s); 940 } 941 } 942
Indexes created Sun Sep 21 20:09:37 GMT 2025