sys_generic.c revision 1.54.2.2
1 /* $NetBSD: sys_generic.c,v 1.54.2.2 2001/06/21 20:07:01 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 90 if ((fp = fd_getfile(fdp, fd)) == NULL) 91 return (EBADF); 92 93 if ((fp->f_flag & FREAD) == 0) 94 return (EBADF); 95 96 FILE_USE(fp); 97 98 /* dofileread() will unuse the descriptor for us */ 99 return (dofileread(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 100 &fp->f_offset, FOF_UPDATE_OFFSET, retval)); 101 } 102 103 int 104 dofileread(struct proc *p, int fd, struct file *fp, void *buf, size_t nbyte, 105 off_t *offset, int flags, register_t *retval) 106 { 107 struct uio auio; 108 struct iovec aiov; 109 long cnt, error; 110 #ifdef KTRACE 111 struct iovec ktriov; 112 #endif 113 error = 0; 114 115 aiov.iov_base = (caddr_t)buf; 116 aiov.iov_len = nbyte; 117 auio.uio_iov = &aiov; 118 auio.uio_iovcnt = 1; 119 auio.uio_resid = nbyte; 120 auio.uio_rw = UIO_READ; 121 auio.uio_segflg = UIO_USERSPACE; 122 auio.uio_procp = p; 123 124 /* 125 * Reads return ssize_t because -1 is returned on error. Therefore 126 * we must restrict the length to SSIZE_MAX to avoid garbage return 127 * values. 128 */ 129 if (auio.uio_resid > SSIZE_MAX) { 130 error = EINVAL; 131 goto out; 132 } 133 134 #ifdef KTRACE 135 /* 136 * if tracing, save a copy of iovec 137 */ 138 if (KTRPOINT(p, KTR_GENIO)) 139 ktriov = aiov; 140 #endif 141 cnt = auio.uio_resid; 142 error = (*fp->f_ops->fo_read)(fp, offset, &auio, fp->f_cred, flags); 143 if (error) 144 if (auio.uio_resid != cnt && (error == ERESTART || 145 error == EINTR || error == EWOULDBLOCK)) 146 error = 0; 147 cnt -= auio.uio_resid; 148 #ifdef KTRACE 149 if (KTRPOINT(p, KTR_GENIO) && error == 0) 150 ktrgenio(p, fd, UIO_READ, &ktriov, cnt, error); 151 #endif 152 *retval = cnt; 153 out: 154 FILE_UNUSE(fp, p); 155 return (error); 156 } 157 158 /* 159 * Scatter read system call. 160 */ 161 int 162 sys_readv(struct lwp *l, void *v, register_t *retval) 163 { 164 struct sys_readv_args /* { 165 syscallarg(int) fd; 166 syscallarg(const struct iovec *) iovp; 167 syscallarg(int) iovcnt; 168 } */ *uap = v; 169 int fd; 170 struct file *fp; 171 struct proc *p; 172 struct filedesc *fdp; 173 174 fd = SCARG(uap, fd); 175 p = l->l_proc; 176 fdp = p->p_fd; 177 178 if ((fp = fd_getfile(fdp, fd)) == NULL) 179 return (EBADF); 180 181 if ((fp->f_flag & FREAD) == 0) 182 return (EBADF); 183 184 FILE_USE(fp); 185 186 /* dofilereadv() will unuse the descriptor for us */ 187 return (dofilereadv(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 188 &fp->f_offset, FOF_UPDATE_OFFSET, retval)); 189 } 190 191 int 192 dofilereadv(struct proc *p, int fd, struct file *fp, const struct iovec *iovp, 193 int iovcnt, off_t *offset, int flags, register_t *retval) 194 { 195 struct uio auio; 196 struct iovec *iov, *needfree, aiov[UIO_SMALLIOV]; 197 long i, cnt, error; 198 u_int iovlen; 199 #ifdef KTRACE 200 struct iovec *ktriov; 201 #endif 202 203 error = 0; 204 #ifdef KTRACE 205 ktriov = NULL; 206 #endif 207 /* note: can't use iovlen until iovcnt is validated */ 208 iovlen = iovcnt * sizeof(struct iovec); 209 if ((u_int)iovcnt > UIO_SMALLIOV) { 210 if ((u_int)iovcnt > IOV_MAX) { 211 error = EINVAL; 212 goto out; 213 } 214 iov = malloc(iovlen, M_IOV, M_WAITOK); 215 needfree = iov; 216 } else if ((u_int)iovcnt > 0) { 217 iov = aiov; 218 needfree = NULL; 219 } else { 220 error = EINVAL; 221 goto out; 222 } 223 224 auio.uio_iov = iov; 225 auio.uio_iovcnt = iovcnt; 226 auio.uio_rw = UIO_READ; 227 auio.uio_segflg = UIO_USERSPACE; 228 auio.uio_procp = p; 229 error = copyin(iovp, iov, iovlen); 230 if (error) 231 goto done; 232 auio.uio_resid = 0; 233 for (i = 0; i < iovcnt; i++) { 234 auio.uio_resid += iov->iov_len; 235 /* 236 * Reads return ssize_t because -1 is returned on error. 237 * Therefore we must restrict the length to SSIZE_MAX to 238 * avoid garbage return values. 239 */ 240 if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) { 241 error = EINVAL; 242 goto done; 243 } 244 iov++; 245 } 246 #ifdef KTRACE 247 /* 248 * if tracing, save a copy of iovec 249 */ 250 if (KTRPOINT(p, KTR_GENIO)) { 251 ktriov = malloc(iovlen, M_TEMP, M_WAITOK); 252 memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen); 253 } 254 #endif 255 cnt = auio.uio_resid; 256 error = (*fp->f_ops->fo_read)(fp, offset, &auio, fp->f_cred, flags); 257 if (error) 258 if (auio.uio_resid != cnt && (error == ERESTART || 259 error == EINTR || error == EWOULDBLOCK)) 260 error = 0; 261 cnt -= auio.uio_resid; 262 #ifdef KTRACE 263 if (KTRPOINT(p, KTR_GENIO)) 264 if (error == 0) { 265 ktrgenio(p, fd, UIO_READ, ktriov, cnt, error); 266 free(ktriov, M_TEMP); 267 } 268 #endif 269 *retval = cnt; 270 done: 271 if (needfree) 272 free(needfree, M_IOV); 273 out: 274 FILE_UNUSE(fp, p); 275 return (error); 276 } 277 278 /* 279 * Write system call 280 */ 281 int 282 sys_write(struct lwp *l, void *v, register_t *retval) 283 { 284 struct sys_write_args /* { 285 syscallarg(int) fd; 286 syscallarg(const void *) buf; 287 syscallarg(size_t) nbyte; 288 } */ *uap = v; 289 int fd; 290 struct file *fp; 291 struct proc *p; 292 struct filedesc *fdp; 293 294 fd = SCARG(uap, fd); 295 p = l->l_proc; 296 fdp = p->p_fd; 297 298 if ((fp = fd_getfile(fdp, fd)) == NULL) 299 return (EBADF); 300 301 if ((fp->f_flag & FWRITE) == 0) 302 return (EBADF); 303 304 FILE_USE(fp); 305 306 /* dofilewrite() will unuse the descriptor for us */ 307 return (dofilewrite(p, fd, fp, SCARG(uap, buf), SCARG(uap, nbyte), 308 &fp->f_offset, FOF_UPDATE_OFFSET, retval)); 309 } 310 311 int 312 dofilewrite(struct proc *p, int fd, struct file *fp, const void *buf, 313 size_t nbyte, off_t *offset, int flags, register_t *retval) 314 { 315 struct uio auio; 316 struct iovec aiov; 317 long cnt, error; 318 #ifdef KTRACE 319 struct iovec ktriov; 320 #endif 321 322 error = 0; 323 aiov.iov_base = (caddr_t)buf; /* XXX kills const */ 324 aiov.iov_len = nbyte; 325 auio.uio_iov = &aiov; 326 auio.uio_iovcnt = 1; 327 auio.uio_resid = nbyte; 328 auio.uio_rw = UIO_WRITE; 329 auio.uio_segflg = UIO_USERSPACE; 330 auio.uio_procp = p; 331 332 /* 333 * Writes return ssize_t because -1 is returned on error. Therefore 334 * we must restrict the length to SSIZE_MAX to avoid garbage return 335 * values. 336 */ 337 if (auio.uio_resid > SSIZE_MAX) { 338 error = EINVAL; 339 goto out; 340 } 341 342 #ifdef KTRACE 343 /* 344 * if tracing, save a copy of iovec 345 */ 346 if (KTRPOINT(p, KTR_GENIO)) 347 ktriov = aiov; 348 #endif 349 cnt = auio.uio_resid; 350 error = (*fp->f_ops->fo_write)(fp, offset, &auio, fp->f_cred, flags); 351 if (error) { 352 if (auio.uio_resid != cnt && (error == ERESTART || 353 error == EINTR || error == EWOULDBLOCK)) 354 error = 0; 355 if (error == EPIPE) 356 psignal(p, SIGPIPE); 357 } 358 cnt -= auio.uio_resid; 359 #ifdef KTRACE 360 if (KTRPOINT(p, KTR_GENIO) && error == 0) 361 ktrgenio(p, fd, UIO_WRITE, &ktriov, cnt, error); 362 #endif 363 *retval = cnt; 364 out: 365 FILE_UNUSE(fp, p); 366 return (error); 367 } 368 369 /* 370 * Gather write system call 371 */ 372 int 373 sys_writev(struct lwp *l, void *v, register_t *retval) 374 { 375 struct sys_writev_args /* { 376 syscallarg(int) fd; 377 syscallarg(const struct iovec *) iovp; 378 syscallarg(int) iovcnt; 379 } */ *uap = v; 380 int fd; 381 struct file *fp; 382 struct proc *p; 383 struct filedesc *fdp; 384 385 fd = SCARG(uap, fd); 386 p = l->l_proc; 387 fdp = p->p_fd; 388 389 if ((fp = fd_getfile(fdp, fd)) == NULL) 390 return (EBADF); 391 392 if ((fp->f_flag & FWRITE) == 0) 393 return (EBADF); 394 395 FILE_USE(fp); 396 397 /* dofilewritev() will unuse the descriptor for us */ 398 return (dofilewritev(p, fd, fp, SCARG(uap, iovp), SCARG(uap, iovcnt), 399 &fp->f_offset, FOF_UPDATE_OFFSET, retval)); 400 } 401 402 int 403 dofilewritev(struct proc *p, int fd, struct file *fp, const struct iovec *iovp, 404 int iovcnt, off_t *offset, int flags, register_t *retval) 405 { 406 struct uio auio; 407 struct iovec *iov, *needfree, aiov[UIO_SMALLIOV]; 408 long i, cnt, error; 409 u_int iovlen; 410 #ifdef KTRACE 411 struct iovec *ktriov; 412 #endif 413 414 error = 0; 415 #ifdef KTRACE 416 ktriov = NULL; 417 #endif 418 /* note: can't use iovlen until iovcnt is validated */ 419 iovlen = iovcnt * sizeof(struct iovec); 420 if ((u_int)iovcnt > UIO_SMALLIOV) { 421 if ((u_int)iovcnt > IOV_MAX) 422 return (EINVAL); 423 iov = malloc(iovlen, M_IOV, M_WAITOK); 424 needfree = iov; 425 } else if ((u_int)iovcnt > 0) { 426 iov = aiov; 427 needfree = NULL; 428 } else { 429 error = EINVAL; 430 goto out; 431 } 432 433 auio.uio_iov = iov; 434 auio.uio_iovcnt = iovcnt; 435 auio.uio_rw = UIO_WRITE; 436 auio.uio_segflg = UIO_USERSPACE; 437 auio.uio_procp = p; 438 error = copyin(iovp, iov, iovlen); 439 if (error) 440 goto done; 441 auio.uio_resid = 0; 442 for (i = 0; i < iovcnt; i++) { 443 auio.uio_resid += iov->iov_len; 444 /* 445 * Writes return ssize_t because -1 is returned on error. 446 * Therefore we must restrict the length to SSIZE_MAX to 447 * avoid garbage return values. 448 */ 449 if (iov->iov_len > SSIZE_MAX || auio.uio_resid > SSIZE_MAX) { 450 error = EINVAL; 451 goto done; 452 } 453 iov++; 454 } 455 #ifdef KTRACE 456 /* 457 * if tracing, save a copy of iovec 458 */ 459 if (KTRPOINT(p, KTR_GENIO)) { 460 ktriov = malloc(iovlen, M_TEMP, M_WAITOK); 461 memcpy((caddr_t)ktriov, (caddr_t)auio.uio_iov, iovlen); 462 } 463 #endif 464 cnt = auio.uio_resid; 465 error = (*fp->f_ops->fo_write)(fp, offset, &auio, fp->f_cred, flags); 466 if (error) { 467 if (auio.uio_resid != cnt && (error == ERESTART || 468 error == EINTR || error == EWOULDBLOCK)) 469 error = 0; 470 if (error == EPIPE) 471 psignal(p, SIGPIPE); 472 } 473 cnt -= auio.uio_resid; 474 #ifdef KTRACE 475 if (KTRPOINT(p, KTR_GENIO)) 476 if (error == 0) { 477 ktrgenio(p, fd, UIO_WRITE, ktriov, cnt, error); 478 free(ktriov, M_TEMP); 479 } 480 #endif 481 *retval = cnt; 482 done: 483 if (needfree) 484 free(needfree, M_IOV); 485 out: 486 FILE_UNUSE(fp, p); 487 return (error); 488 } 489 490 /* 491 * Ioctl system call 492 */ 493 /* ARGSUSED */ 494 int 495 sys_ioctl(struct lwp *l, void *v, register_t *retval) 496 { 497 struct sys_ioctl_args /* { 498 syscallarg(int) fd; 499 syscallarg(u_long) com; 500 syscallarg(caddr_t) data; 501 } */ *uap = v; 502 struct file *fp; 503 struct proc *p; 504 struct filedesc *fdp; 505 u_long com; 506 int error; 507 u_int size; 508 caddr_t data, memp; 509 int tmp; 510 #define STK_PARAMS 128 511 u_long stkbuf[STK_PARAMS/sizeof(u_long)]; 512 513 error = 0; 514 p = l->l_proc; 515 fdp = p->p_fd; 516 517 if ((fp = fd_getfile(fdp, SCARG(uap, fd))) == NULL) 518 return (EBADF); 519 520 FILE_USE(fp); 521 522 if ((fp->f_flag & (FREAD | FWRITE)) == 0) { 523 error = EBADF; 524 goto out; 525 } 526 527 switch (com = SCARG(uap, com)) { 528 case FIONCLEX: 529 fdp->fd_ofileflags[SCARG(uap, fd)] &= ~UF_EXCLOSE; 530 goto out; 531 532 case FIOCLEX: 533 fdp->fd_ofileflags[SCARG(uap, fd)] |= UF_EXCLOSE; 534 goto out; 535 } 536 537 /* 538 * Interpret high order word to find amount of data to be 539 * copied to/from the user's address space. 540 */ 541 size = IOCPARM_LEN(com); 542 if (size > IOCPARM_MAX) { 543 error = ENOTTY; 544 goto out; 545 } 546 memp = NULL; 547 if (size > sizeof(stkbuf)) { 548 memp = (caddr_t)malloc((u_long)size, M_IOCTLOPS, M_WAITOK); 549 data = memp; 550 } else 551 data = (caddr_t)stkbuf; 552 if (com&IOC_IN) { 553 if (size) { 554 error = copyin(SCARG(uap, data), data, size); 555 if (error) { 556 if (memp) 557 free(memp, M_IOCTLOPS); 558 goto out; 559 } 560 } else 561 *(caddr_t *)data = SCARG(uap, data); 562 } else if ((com&IOC_OUT) && size) 563 /* 564 * Zero the buffer so the user always 565 * gets back something deterministic. 566 */ 567 memset(data, 0, size); 568 else if (com&IOC_VOID) 569 *(caddr_t *)data = SCARG(uap, data); 570 571 switch (com) { 572 573 case FIONBIO: 574 if ((tmp = *(int *)data) != 0) 575 fp->f_flag |= FNONBLOCK; 576 else 577 fp->f_flag &= ~FNONBLOCK; 578 error = (*fp->f_ops->fo_ioctl)(fp, FIONBIO, (caddr_t)&tmp, p); 579 break; 580 581 case FIOASYNC: 582 if ((tmp = *(int *)data) != 0) 583 fp->f_flag |= FASYNC; 584 else 585 fp->f_flag &= ~FASYNC; 586 error = (*fp->f_ops->fo_ioctl)(fp, FIOASYNC, (caddr_t)&tmp, p); 587 break; 588 589 case FIOSETOWN: 590 tmp = *(int *)data; 591 if (fp->f_type == DTYPE_SOCKET) { 592 ((struct socket *)fp->f_data)->so_pgid = tmp; 593 error = 0; 594 break; 595 } 596 if (tmp <= 0) { 597 tmp = -tmp; 598 } else { 599 struct proc *p1 = pfind(tmp); 600 if (p1 == 0) { 601 error = ESRCH; 602 break; 603 } 604 tmp = p1->p_pgrp->pg_id; 605 } 606 error = (*fp->f_ops->fo_ioctl) 607 (fp, TIOCSPGRP, (caddr_t)&tmp, p); 608 break; 609 610 case FIOGETOWN: 611 if (fp->f_type == DTYPE_SOCKET) { 612 error = 0; 613 *(int *)data = ((struct socket *)fp->f_data)->so_pgid; 614 break; 615 } 616 error = (*fp->f_ops->fo_ioctl)(fp, TIOCGPGRP, data, p); 617 if (error == 0) 618 *(int *)data = -*(int *)data; 619 break; 620 621 default: 622 error = (*fp->f_ops->fo_ioctl)(fp, com, data, p); 623 /* 624 * Copy any data to user, size was 625 * already set and checked above. 626 */ 627 if (error == 0 && (com&IOC_OUT) && size) 628 error = copyout(data, SCARG(uap, data), size); 629 break; 630 } 631 if (memp) 632 free(memp, M_IOCTLOPS); 633 out: 634 FILE_UNUSE(fp, p); 635 return (error); 636 } 637 638 int selwait, nselcoll; 639 640 /* 641 * Select system call. 642 */ 643 int 644 sys_select(struct lwp *l, void *v, register_t *retval) 645 { 646 struct sys_select_args /* { 647 syscallarg(int) nd; 648 syscallarg(fd_set *) in; 649 syscallarg(fd_set *) ou; 650 syscallarg(fd_set *) ex; 651 syscallarg(struct timeval *) tv; 652 } */ *uap = v; 653 struct proc *p; 654 caddr_t bits; 655 char smallbits[howmany(FD_SETSIZE, NFDBITS) * 656 sizeof(fd_mask) * 6]; 657 struct timeval atv; 658 int s, ncoll, error, timo; 659 size_t ni; 660 661 error = 0; 662 p = l->l_proc; 663 if (SCARG(uap, nd) < 0) 664 return (EINVAL); 665 if (SCARG(uap, nd) > p->p_fd->fd_nfiles) { 666 /* forgiving; slightly wrong */ 667 SCARG(uap, nd) = p->p_fd->fd_nfiles; 668 } 669 ni = howmany(SCARG(uap, nd), NFDBITS) * sizeof(fd_mask); 670 if (ni * 6 > sizeof(smallbits)) 671 bits = malloc(ni * 6, M_TEMP, M_WAITOK); 672 else 673 bits = smallbits; 674 675 #define getbits(name, x) \ 676 if (SCARG(uap, name)) { \ 677 error = copyin(SCARG(uap, name), bits + ni * x, ni); \ 678 if (error) \ 679 goto done; \ 680 } else \ 681 memset(bits + ni * x, 0, ni); 682 getbits(in, 0); 683 getbits(ou, 1); 684 getbits(ex, 2); 685 #undef getbits 686 687 if (SCARG(uap, tv)) { 688 error = copyin(SCARG(uap, tv), (caddr_t)&atv, 689 sizeof(atv)); 690 if (error) 691 goto done; 692 if (itimerfix(&atv)) { 693 error = EINVAL; 694 goto done; 695 } 696 s = splclock(); 697 timeradd(&atv, &time, &atv); 698 splx(s); 699 } else 700 timo = 0; 701 retry: 702 ncoll = nselcoll; 703 l->l_flag |= L_SELECT; 704 error = selscan(p, (fd_mask *)(bits + ni * 0), 705 (fd_mask *)(bits + ni * 3), SCARG(uap, nd), retval); 706 if (error || *retval) 707 goto done; 708 if (SCARG(uap, tv)) { 709 /* 710 * We have to recalculate the timeout on every retry. 711 */ 712 timo = hzto(&atv); 713 if (timo <= 0) 714 goto done; 715 } 716 s = splsched(); 717 if ((l->l_flag & L_SELECT) == 0 || nselcoll != ncoll) { 718 splx(s); 719 goto retry; 720 } 721 l->l_flag &= ~L_SELECT; 722 error = tsleep((caddr_t)&selwait, PSOCK | PCATCH, "select", timo); 723 splx(s); 724 if (error == 0) 725 goto retry; 726 done: 727 l->l_flag &= ~L_SELECT; 728 /* select is not restarted after signals... */ 729 if (error == ERESTART) 730 error = EINTR; 731 if (error == EWOULDBLOCK) 732 error = 0; 733 if (error == 0) { 734 735 #define putbits(name, x) \ 736 if (SCARG(uap, name)) { \ 737 error = copyout(bits + ni * x, SCARG(uap, name), ni); \ 738 if (error) \ 739 goto out; \ 740 } 741 putbits(in, 3); 742 putbits(ou, 4); 743 putbits(ex, 5); 744 #undef putbits 745 } 746 out: 747 if (ni * 6 > sizeof(smallbits)) 748 free(bits, M_TEMP); 749 return (error); 750 } 751 752 int 753 selscan(struct proc *p, fd_mask *ibitp, fd_mask *obitp, int nfd, 754 register_t *retval) 755 { 756 struct filedesc *fdp; 757 int msk, i, j, fd, n; 758 fd_mask ibits, obits; 759 struct file *fp; 760 static int flag[3] = { POLLRDNORM | POLLHUP | POLLERR, 761 POLLWRNORM | POLLHUP | POLLERR, 762 POLLRDBAND }; 763 764 fdp = p->p_fd; 765 n = 0; 766 for (msk = 0; msk < 3; msk++) { 767 for (i = 0; i < nfd; i += NFDBITS) { 768 ibits = *ibitp++; 769 obits = 0; 770 while ((j = ffs(ibits)) && (fd = i + --j) < nfd) { 771 ibits &= ~(1 << j); 772 if ((fp = fd_getfile(fdp, fd)) == NULL) 773 return (EBADF); 774 FILE_USE(fp); 775 if ((*fp->f_ops->fo_poll)(fp, flag[msk], p)) { 776 obits |= (1 << j); 777 n++; 778 } 779 FILE_UNUSE(fp, p); 780 } 781 *obitp++ = obits; 782 } 783 } 784 *retval = n; 785 return (0); 786 } 787 788 /* 789 * Poll system call. 790 */ 791 int 792 sys_poll(struct lwp *l, void *v, register_t *retval) 793 { 794 struct sys_poll_args /* { 795 syscallarg(struct pollfd *) fds; 796 syscallarg(u_int) nfds; 797 syscallarg(int) timeout; 798 } */ *uap = v; 799 struct proc *p; 800 caddr_t bits; 801 char smallbits[32 * sizeof(struct pollfd)]; 802 struct timeval atv; 803 int s, ncoll, error, timo; 804 size_t ni; 805 806 error = 0; 807 p = l->l_proc; 808 if (SCARG(uap, nfds) > p->p_fd->fd_nfiles) { 809 /* forgiving; slightly wrong */ 810 SCARG(uap, nfds) = p->p_fd->fd_nfiles; 811 } 812 ni = SCARG(uap, nfds) * sizeof(struct pollfd); 813 if (ni > sizeof(smallbits)) 814 bits = malloc(ni, M_TEMP, M_WAITOK); 815 else 816 bits = smallbits; 817 818 error = copyin(SCARG(uap, fds), bits, ni); 819 if (error) 820 goto done; 821 822 if (SCARG(uap, timeout) != INFTIM) { 823 atv.tv_sec = SCARG(uap, timeout) / 1000; 824 atv.tv_usec = (SCARG(uap, timeout) % 1000) * 1000; 825 if (itimerfix(&atv)) { 826 error = EINVAL; 827 goto done; 828 } 829 s = splclock(); 830 timeradd(&atv, &time, &atv); 831 splx(s); 832 } else 833 timo = 0; 834 retry: 835 ncoll = nselcoll; 836 l->l_flag |= L_SELECT; 837 error = pollscan(p, (struct pollfd *)bits, SCARG(uap, nfds), retval); 838 if (error || *retval) 839 goto done; 840 if (SCARG(uap, timeout) != INFTIM) { 841 /* 842 * We have to recalculate the timeout on every retry. 843 */ 844 timo = hzto(&atv); 845 if (timo <= 0) 846 goto done; 847 } 848 s = splsched(); 849 if ((l->l_flag & L_SELECT) == 0 || nselcoll != ncoll) { 850 splx(s); 851 goto retry; 852 } 853 l->l_flag &= ~L_SELECT; 854 error = tsleep((caddr_t)&selwait, PSOCK | PCATCH, "select", timo); 855 splx(s); 856 if (error == 0) 857 goto retry; 858 done: 859 l->l_flag &= ~L_SELECT; 860 /* poll is not restarted after signals... */ 861 if (error == ERESTART) 862 error = EINTR; 863 if (error == EWOULDBLOCK) 864 error = 0; 865 if (error == 0) { 866 error = copyout(bits, SCARG(uap, fds), ni); 867 if (error) 868 goto out; 869 } 870 out: 871 if (ni > sizeof(smallbits)) 872 free(bits, M_TEMP); 873 return (error); 874 } 875 876 int 877 pollscan(struct proc *p, struct pollfd *fds, int nfd, register_t *retval) 878 { 879 struct filedesc *fdp; 880 int i, n; 881 struct file *fp; 882 883 fdp = p->p_fd; 884 n = 0; 885 for (i = 0; i < nfd; i++, fds++) { 886 if ((u_int)fds->fd >= fdp->fd_nfiles) { 887 fds->revents = POLLNVAL; 888 n++; 889 } else { 890 if ((fp = fd_getfile(fdp, fds->fd)) == NULL) { 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