1 /* 2 * Copyright (c) 1982, 1986, 1989, 1993 3 * The Regents of the University of California. All rights reserved. 4 * (c) UNIX System Laboratories, Inc. 5 * All or some portions of this file are derived from material licensed 6 * to the University of California by American Telephone and Telegraph 7 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 8 * the permission of UNIX System Laboratories, Inc. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 3. All advertising materials mentioning features or use of this software 19 * must display the following acknowledgement: 20 * This product includes software developed by the University of 21 * California, Berkeley and its contributors. 22 * 4. Neither the name of the University nor the names of its contributors 23 * may be used to endorse or promote products derived from this software 24 * without specific prior written permission. 25 * 26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 36 * SUCH DAMAGE. 37 * 38 * @(#)sys_generic.c 8.5 (Berkeley) 1/21/94 39 * $FreeBSD: src/sys/kern/sys_generic.c,v 1.55.2.10 2001/03/17 10:39:32 peter Exp $ 40 * $DragonFly: src/sys/kern/sys_generic.c,v 1.44 2007/02/22 15:50:49 corecode Exp $ 41 */ 42 43 #include "opt_ktrace.h" 44 45 #include <sys/param.h> 46 #include <sys/systm.h> 47 #include <sys/sysproto.h> 48 #include <sys/filedesc.h> 49 #include <sys/filio.h> 50 #include <sys/fcntl.h> 51 #include <sys/file.h> 52 #include <sys/proc.h> 53 #include <sys/signalvar.h> 54 #include <sys/socketvar.h> 55 #include <sys/uio.h> 56 #include <sys/kernel.h> 57 #include <sys/kern_syscall.h> 58 #include <sys/malloc.h> 59 #include <sys/mapped_ioctl.h> 60 #include <sys/poll.h> 61 #include <sys/queue.h> 62 #include <sys/resourcevar.h> 63 #include <sys/sysctl.h> 64 #include <sys/sysent.h> 65 #include <sys/buf.h> 66 #ifdef KTRACE 67 #include <sys/ktrace.h> 68 #endif 69 #include <vm/vm.h> 70 #include <vm/vm_page.h> 71 #include <sys/file2.h> 72 73 #include <machine/limits.h> 74 75 static MALLOC_DEFINE(M_IOCTLOPS, "ioctlops", "ioctl data buffer"); 76 static MALLOC_DEFINE(M_IOCTLMAP, "ioctlmap", "mapped ioctl handler buffer"); 77 static MALLOC_DEFINE(M_SELECT, "select", "select() buffer"); 78 MALLOC_DEFINE(M_IOV, "iov", "large iov's"); 79 80 static int pollscan (struct proc *, struct pollfd *, u_int, int *); 81 static int selscan (struct proc *, fd_mask **, fd_mask **, 82 int, int *); 83 static int dofileread(int, struct file *, struct uio *, int, int *); 84 static int dofilewrite(int, struct file *, struct uio *, int, int *); 85 86 /* 87 * Read system call. 88 * 89 * MPSAFE 90 */ 91 int 92 sys_read(struct read_args *uap) 93 { 94 struct thread *td = curthread; 95 struct uio auio; 96 struct iovec aiov; 97 int error; 98 99 aiov.iov_base = uap->buf; 100 aiov.iov_len = uap->nbyte; 101 auio.uio_iov = &aiov; 102 auio.uio_iovcnt = 1; 103 auio.uio_offset = -1; 104 auio.uio_resid = uap->nbyte; 105 auio.uio_rw = UIO_READ; 106 auio.uio_segflg = UIO_USERSPACE; 107 auio.uio_td = td; 108 109 if (auio.uio_resid < 0) 110 error = EINVAL; 111 else 112 error = kern_preadv(uap->fd, &auio, 0, &uap->sysmsg_result); 113 return(error); 114 } 115 116 /* 117 * Positioned (Pread) read system call 118 * 119 * MPSAFE 120 */ 121 int 122 sys_extpread(struct extpread_args *uap) 123 { 124 struct thread *td = curthread; 125 struct uio auio; 126 struct iovec aiov; 127 int error; 128 int flags; 129 130 aiov.iov_base = uap->buf; 131 aiov.iov_len = uap->nbyte; 132 auio.uio_iov = &aiov; 133 auio.uio_iovcnt = 1; 134 auio.uio_offset = uap->offset; 135 auio.uio_resid = uap->nbyte; 136 auio.uio_rw = UIO_READ; 137 auio.uio_segflg = UIO_USERSPACE; 138 auio.uio_td = td; 139 140 flags = uap->flags & O_FMASK; 141 if (uap->offset != (off_t)-1) 142 flags |= O_FOFFSET; 143 144 if (auio.uio_resid < 0) 145 error = EINVAL; 146 else 147 error = kern_preadv(uap->fd, &auio, flags, &uap->sysmsg_result); 148 return(error); 149 } 150 151 /* 152 * Scatter read system call. 153 * 154 * MPSAFE 155 */ 156 int 157 sys_readv(struct readv_args *uap) 158 { 159 struct thread *td = curthread; 160 struct uio auio; 161 struct iovec aiov[UIO_SMALLIOV], *iov = NULL; 162 int error; 163 164 error = iovec_copyin(uap->iovp, &iov, aiov, uap->iovcnt, 165 &auio.uio_resid); 166 if (error) 167 return (error); 168 auio.uio_iov = iov; 169 auio.uio_iovcnt = uap->iovcnt; 170 auio.uio_offset = -1; 171 auio.uio_rw = UIO_READ; 172 auio.uio_segflg = UIO_USERSPACE; 173 auio.uio_td = td; 174 175 error = kern_preadv(uap->fd, &auio, 0, &uap->sysmsg_result); 176 177 iovec_free(&iov, aiov); 178 return (error); 179 } 180 181 182 /* 183 * Scatter positioned read system call. 184 * 185 * MPSAFE 186 */ 187 int 188 sys_extpreadv(struct extpreadv_args *uap) 189 { 190 struct thread *td = curthread; 191 struct uio auio; 192 struct iovec aiov[UIO_SMALLIOV], *iov = NULL; 193 int error; 194 int flags; 195 196 error = iovec_copyin(uap->iovp, &iov, aiov, uap->iovcnt, 197 &auio.uio_resid); 198 if (error) 199 return (error); 200 auio.uio_iov = iov; 201 auio.uio_iovcnt = uap->iovcnt; 202 auio.uio_offset = uap->offset; 203 auio.uio_rw = UIO_READ; 204 auio.uio_segflg = UIO_USERSPACE; 205 auio.uio_td = td; 206 207 flags = uap->flags & O_FMASK; 208 if (uap->offset != (off_t)-1) 209 flags |= O_FOFFSET; 210 211 error = kern_preadv(uap->fd, &auio, flags, &uap->sysmsg_result); 212 213 iovec_free(&iov, aiov); 214 return(error); 215 } 216 217 /* 218 * MPSAFE 219 */ 220 int 221 kern_preadv(int fd, struct uio *auio, int flags, int *res) 222 { 223 struct thread *td = curthread; 224 struct proc *p = td->td_proc; 225 struct file *fp; 226 int error; 227 228 KKASSERT(p); 229 230 fp = holdfp(p->p_fd, fd, FREAD); 231 if (fp == NULL) 232 return (EBADF); 233 if (flags & O_FOFFSET && fp->f_type != DTYPE_VNODE) { 234 error = ESPIPE; 235 } else if (auio->uio_resid < 0) { 236 error = EINVAL; 237 } else { 238 error = dofileread(fd, fp, auio, flags, res); 239 } 240 fdrop(fp); 241 return(error); 242 } 243 244 /* 245 * Common code for readv and preadv that reads data in 246 * from a file using the passed in uio, offset, and flags. 247 * 248 * MPALMOSTSAFE - ktrace needs help 249 */ 250 static int 251 dofileread(int fd, struct file *fp, struct uio *auio, int flags, int *res) 252 { 253 struct thread *td = curthread; 254 struct proc *p = td->td_proc; 255 int error; 256 int len; 257 #ifdef KTRACE 258 struct iovec *ktriov = NULL; 259 struct uio ktruio; 260 #endif 261 262 #ifdef KTRACE 263 /* 264 * if tracing, save a copy of iovec 265 */ 266 if (KTRPOINT(td, KTR_GENIO)) { 267 int iovlen = auio->uio_iovcnt * sizeof(struct iovec); 268 269 MALLOC(ktriov, struct iovec *, iovlen, M_TEMP, M_WAITOK); 270 bcopy((caddr_t)auio->uio_iov, (caddr_t)ktriov, iovlen); 271 ktruio = *auio; 272 } 273 #endif 274 len = auio->uio_resid; 275 error = fo_read(fp, auio, fp->f_cred, flags); 276 if (error) { 277 if (auio->uio_resid != len && (error == ERESTART || 278 error == EINTR || error == EWOULDBLOCK)) 279 error = 0; 280 } 281 #ifdef KTRACE 282 if (ktriov != NULL) { 283 if (error == 0) { 284 ktruio.uio_iov = ktriov; 285 ktruio.uio_resid = len - auio->uio_resid; 286 get_mplock(); 287 ktrgenio(p, fd, UIO_READ, &ktruio, error); 288 rel_mplock(); 289 } 290 FREE(ktriov, M_TEMP); 291 } 292 #endif 293 if (error == 0) 294 *res = len - auio->uio_resid; 295 296 return(error); 297 } 298 299 /* 300 * Write system call 301 * 302 * MPSAFE 303 */ 304 int 305 sys_write(struct write_args *uap) 306 { 307 struct thread *td = curthread; 308 struct uio auio; 309 struct iovec aiov; 310 int error; 311 312 aiov.iov_base = (void *)(uintptr_t)uap->buf; 313 aiov.iov_len = uap->nbyte; 314 auio.uio_iov = &aiov; 315 auio.uio_iovcnt = 1; 316 auio.uio_offset = -1; 317 auio.uio_resid = uap->nbyte; 318 auio.uio_rw = UIO_WRITE; 319 auio.uio_segflg = UIO_USERSPACE; 320 auio.uio_td = td; 321 322 if (auio.uio_resid < 0) 323 error = EINVAL; 324 else 325 error = kern_pwritev(uap->fd, &auio, 0, &uap->sysmsg_result); 326 327 return(error); 328 } 329 330 /* 331 * Pwrite system call 332 * 333 * MPSAFE 334 */ 335 int 336 sys_extpwrite(struct extpwrite_args *uap) 337 { 338 struct thread *td = curthread; 339 struct uio auio; 340 struct iovec aiov; 341 int error; 342 int flags; 343 344 aiov.iov_base = (void *)(uintptr_t)uap->buf; 345 aiov.iov_len = uap->nbyte; 346 auio.uio_iov = &aiov; 347 auio.uio_iovcnt = 1; 348 auio.uio_offset = uap->offset; 349 auio.uio_resid = uap->nbyte; 350 auio.uio_rw = UIO_WRITE; 351 auio.uio_segflg = UIO_USERSPACE; 352 auio.uio_td = td; 353 354 flags = uap->flags & O_FMASK; 355 if (uap->offset != (off_t)-1) 356 flags |= O_FOFFSET; 357 358 if (auio.uio_resid < 0) 359 error = EINVAL; 360 else 361 error = kern_pwritev(uap->fd, &auio, flags, &uap->sysmsg_result); 362 363 return(error); 364 } 365 366 /* 367 * MPSAFE 368 */ 369 int 370 sys_writev(struct writev_args *uap) 371 { 372 struct thread *td = curthread; 373 struct uio auio; 374 struct iovec aiov[UIO_SMALLIOV], *iov = NULL; 375 int error; 376 377 error = iovec_copyin(uap->iovp, &iov, aiov, uap->iovcnt, 378 &auio.uio_resid); 379 if (error) 380 return (error); 381 auio.uio_iov = iov; 382 auio.uio_iovcnt = uap->iovcnt; 383 auio.uio_offset = -1; 384 auio.uio_rw = UIO_WRITE; 385 auio.uio_segflg = UIO_USERSPACE; 386 auio.uio_td = td; 387 388 error = kern_pwritev(uap->fd, &auio, 0, &uap->sysmsg_result); 389 390 iovec_free(&iov, aiov); 391 return (error); 392 } 393 394 395 /* 396 * Gather positioned write system call 397 * 398 * MPSAFE 399 */ 400 int 401 sys_extpwritev(struct extpwritev_args *uap) 402 { 403 struct thread *td = curthread; 404 struct uio auio; 405 struct iovec aiov[UIO_SMALLIOV], *iov = NULL; 406 int error; 407 int flags; 408 409 error = iovec_copyin(uap->iovp, &iov, aiov, uap->iovcnt, 410 &auio.uio_resid); 411 if (error) 412 return (error); 413 auio.uio_iov = iov; 414 auio.uio_iovcnt = uap->iovcnt; 415 auio.uio_offset = uap->offset; 416 auio.uio_rw = UIO_WRITE; 417 auio.uio_segflg = UIO_USERSPACE; 418 auio.uio_td = td; 419 420 flags = uap->flags & O_FMASK; 421 if (uap->offset != (off_t)-1) 422 flags |= O_FOFFSET; 423 424 error = kern_pwritev(uap->fd, &auio, flags, &uap->sysmsg_result); 425 426 iovec_free(&iov, aiov); 427 return(error); 428 } 429 430 /* 431 * MPSAFE 432 */ 433 int 434 kern_pwritev(int fd, struct uio *auio, int flags, int *res) 435 { 436 struct thread *td = curthread; 437 struct proc *p = td->td_proc; 438 struct file *fp; 439 int error; 440 441 KKASSERT(p); 442 443 fp = holdfp(p->p_fd, fd, FWRITE); 444 if (fp == NULL) 445 return (EBADF); 446 else if ((flags & O_FOFFSET) && fp->f_type != DTYPE_VNODE) { 447 error = ESPIPE; 448 } else { 449 error = dofilewrite(fd, fp, auio, flags, res); 450 } 451 452 fdrop(fp); 453 return (error); 454 } 455 456 /* 457 * Common code for writev and pwritev that writes data to 458 * a file using the passed in uio, offset, and flags. 459 * 460 * MPALMOSTSAFE - ktrace needs help 461 */ 462 static int 463 dofilewrite(int fd, struct file *fp, struct uio *auio, int flags, int *res) 464 { 465 struct thread *td = curthread; 466 struct lwp *lp = td->td_lwp; 467 struct proc *p = td->td_proc; 468 int error; 469 int len; 470 #ifdef KTRACE 471 struct iovec *ktriov = NULL; 472 struct uio ktruio; 473 #endif 474 475 #ifdef KTRACE 476 /* 477 * if tracing, save a copy of iovec and uio 478 */ 479 if (KTRPOINT(td, KTR_GENIO)) { 480 int iovlen = auio->uio_iovcnt * sizeof(struct iovec); 481 482 MALLOC(ktriov, struct iovec *, iovlen, M_TEMP, M_WAITOK); 483 bcopy((caddr_t)auio->uio_iov, (caddr_t)ktriov, iovlen); 484 ktruio = *auio; 485 } 486 #endif 487 len = auio->uio_resid; 488 if (fp->f_type == DTYPE_VNODE) 489 bwillwrite(); 490 error = fo_write(fp, auio, fp->f_cred, flags); 491 if (error) { 492 if (auio->uio_resid != len && (error == ERESTART || 493 error == EINTR || error == EWOULDBLOCK)) 494 error = 0; 495 /* Socket layer is responsible for issuing SIGPIPE. */ 496 if (error == EPIPE) { 497 get_mplock(); 498 lwpsignal(p, lp, SIGPIPE); 499 rel_mplock(); 500 } 501 } 502 #ifdef KTRACE 503 if (ktriov != NULL) { 504 if (error == 0) { 505 ktruio.uio_iov = ktriov; 506 ktruio.uio_resid = len - auio->uio_resid; 507 get_mplock(); 508 ktrgenio(p, fd, UIO_WRITE, &ktruio, error); 509 rel_mplock(); 510 } 511 FREE(ktriov, M_TEMP); 512 } 513 #endif 514 if (error == 0) 515 *res = len - auio->uio_resid; 516 517 return(error); 518 } 519 520 /* 521 * Ioctl system call 522 */ 523 /* ARGSUSED */ 524 int 525 sys_ioctl(struct ioctl_args *uap) 526 { 527 return(mapped_ioctl(uap->fd, uap->com, uap->data, NULL)); 528 } 529 530 struct ioctl_map_entry { 531 const char *subsys; 532 struct ioctl_map_range *cmd_ranges; 533 LIST_ENTRY(ioctl_map_entry) entries; 534 }; 535 536 /* 537 * The true heart of all ioctl syscall handlers (native, emulation). 538 * If map != NULL, it will be searched for a matching entry for com, 539 * and appropriate conversions/conversion functions will be utilized. 540 */ 541 int 542 mapped_ioctl(int fd, u_long com, caddr_t uspc_data, struct ioctl_map *map) 543 { 544 struct thread *td = curthread; 545 struct proc *p = td->td_proc; 546 struct ucred *cred; 547 struct file *fp; 548 struct ioctl_map_range *iomc = NULL; 549 int error; 550 u_int size; 551 u_long ocom = com; 552 caddr_t data, memp; 553 int tmp; 554 #define STK_PARAMS 128 555 union { 556 char stkbuf[STK_PARAMS]; 557 long align; 558 } ubuf; 559 560 KKASSERT(p); 561 cred = p->p_ucred; 562 563 fp = holdfp(p->p_fd, fd, FREAD|FWRITE); 564 if (fp == NULL) 565 return(EBADF); 566 567 if (map != NULL) { /* obey translation map */ 568 u_long maskcmd; 569 struct ioctl_map_entry *e; 570 571 maskcmd = com & map->mask; 572 573 LIST_FOREACH(e, &map->mapping, entries) { 574 for (iomc = e->cmd_ranges; iomc->start != 0 || 575 iomc->maptocmd != 0 || iomc->wrapfunc != NULL || 576 iomc->mapfunc != NULL; 577 iomc++) { 578 if (maskcmd >= iomc->start && 579 maskcmd <= iomc->end) 580 break; 581 } 582 583 /* Did we find a match? */ 584 if (iomc->start != 0 || iomc->maptocmd != 0 || 585 iomc->wrapfunc != NULL || iomc->mapfunc != NULL) 586 break; 587 } 588 589 if (iomc == NULL || 590 (iomc->start == 0 && iomc->maptocmd == 0 591 && iomc->wrapfunc == NULL && iomc->mapfunc == NULL)) { 592 kprintf("%s: 'ioctl' fd=%d, cmd=0x%lx ('%c',%d) not implemented\n", 593 map->sys, fd, maskcmd, 594 (int)((maskcmd >> 8) & 0xff), 595 (int)(maskcmd & 0xff)); 596 error = EINVAL; 597 goto done; 598 } 599 600 /* 601 * If it's a non-range one to one mapping, maptocmd should be 602 * correct. If it's a ranged one to one mapping, we pass the 603 * original value of com, and for a range mapped to a different 604 * range, we always need a mapping function to translate the 605 * ioctl to our native ioctl. Ex. 6500-65ff <-> 9500-95ff 606 */ 607 if (iomc->start == iomc->end && iomc->maptocmd == iomc->maptoend) { 608 com = iomc->maptocmd; 609 } else if (iomc->start == iomc->maptocmd && iomc->end == iomc->maptoend) { 610 if (iomc->mapfunc != NULL) 611 com = iomc->mapfunc(iomc->start, iomc->end, 612 iomc->start, iomc->end, 613 com, com); 614 } else { 615 if (iomc->mapfunc != NULL) { 616 com = iomc->mapfunc(iomc->start, iomc->end, 617 iomc->maptocmd, iomc->maptoend, 618 com, ocom); 619 } else { 620 kprintf("%s: Invalid mapping for fd=%d, cmd=%#lx ('%c',%d)\n", 621 map->sys, fd, maskcmd, 622 (int)((maskcmd >> 8) & 0xff), 623 (int)(maskcmd & 0xff)); 624 error = EINVAL; 625 goto done; 626 } 627 } 628 } 629 630 switch (com) { 631 case FIONCLEX: 632 error = fclrfdflags(p->p_fd, fd, UF_EXCLOSE); 633 goto done; 634 case FIOCLEX: 635 error = fsetfdflags(p->p_fd, fd, UF_EXCLOSE); 636 goto done; 637 } 638 639 /* 640 * Interpret high order word to find amount of data to be 641 * copied to/from the user's address space. 642 */ 643 size = IOCPARM_LEN(com); 644 if (size > IOCPARM_MAX) { 645 error = ENOTTY; 646 goto done; 647 } 648 649 memp = NULL; 650 if (size > sizeof (ubuf.stkbuf)) { 651 memp = kmalloc(size, M_IOCTLOPS, M_WAITOK); 652 data = memp; 653 } else { 654 data = ubuf.stkbuf; 655 } 656 if ((com & IOC_IN) != 0) { 657 if (size != 0) { 658 error = copyin(uspc_data, data, (u_int)size); 659 if (error) { 660 if (memp != NULL) 661 kfree(memp, M_IOCTLOPS); 662 goto done; 663 } 664 } else { 665 *(caddr_t *)data = uspc_data; 666 } 667 } else if ((com & IOC_OUT) != 0 && size) { 668 /* 669 * Zero the buffer so the user always 670 * gets back something deterministic. 671 */ 672 bzero(data, size); 673 } else if ((com & IOC_VOID) != 0) { 674 *(caddr_t *)data = uspc_data; 675 } 676 677 switch (com) { 678 case FIONBIO: 679 if ((tmp = *(int *)data)) 680 fp->f_flag |= FNONBLOCK; 681 else 682 fp->f_flag &= ~FNONBLOCK; 683 error = 0; 684 break; 685 686 case FIOASYNC: 687 if ((tmp = *(int *)data)) 688 fp->f_flag |= FASYNC; 689 else 690 fp->f_flag &= ~FASYNC; 691 error = fo_ioctl(fp, FIOASYNC, (caddr_t)&tmp, cred); 692 break; 693 694 default: 695 /* 696 * If there is a override function, 697 * call it instead of directly routing the call 698 */ 699 if (map != NULL && iomc->wrapfunc != NULL) 700 error = iomc->wrapfunc(fp, com, ocom, data, cred); 701 else 702 error = fo_ioctl(fp, com, data, cred); 703 /* 704 * Copy any data to user, size was 705 * already set and checked above. 706 */ 707 if (error == 0 && (com & IOC_OUT) != 0 && size != 0) 708 error = copyout(data, uspc_data, (u_int)size); 709 break; 710 } 711 if (memp != NULL) 712 kfree(memp, M_IOCTLOPS); 713 done: 714 fdrop(fp); 715 return(error); 716 } 717 718 int 719 mapped_ioctl_register_handler(struct ioctl_map_handler *he) 720 { 721 struct ioctl_map_entry *ne; 722 723 KKASSERT(he != NULL && he->map != NULL && he->cmd_ranges != NULL && 724 he->subsys != NULL && *he->subsys != '\0'); 725 726 ne = kmalloc(sizeof(struct ioctl_map_entry), M_IOCTLMAP, M_WAITOK); 727 728 ne->subsys = he->subsys; 729 ne->cmd_ranges = he->cmd_ranges; 730 731 LIST_INSERT_HEAD(&he->map->mapping, ne, entries); 732 733 return(0); 734 } 735 736 int 737 mapped_ioctl_unregister_handler(struct ioctl_map_handler *he) 738 { 739 struct ioctl_map_entry *ne; 740 741 KKASSERT(he != NULL && he->map != NULL && he->cmd_ranges != NULL); 742 743 LIST_FOREACH(ne, &he->map->mapping, entries) { 744 if (ne->cmd_ranges != he->cmd_ranges) 745 continue; 746 LIST_REMOVE(ne, entries); 747 kfree(ne, M_IOCTLMAP); 748 return(0); 749 } 750 return(EINVAL); 751 } 752 753 static int nselcoll; /* Select collisions since boot */ 754 int selwait; 755 SYSCTL_INT(_kern, OID_AUTO, nselcoll, CTLFLAG_RD, &nselcoll, 0, ""); 756 757 /* 758 * Select system call. 759 */ 760 int 761 sys_select(struct select_args *uap) 762 { 763 struct lwp *lp = curthread->td_lwp; 764 struct proc *p = curproc; 765 766 /* 767 * The magic 2048 here is chosen to be just enough for FD_SETSIZE 768 * infds with the new FD_SETSIZE of 1024, and more than enough for 769 * FD_SETSIZE infds, outfds and exceptfds with the old FD_SETSIZE 770 * of 256. 771 */ 772 fd_mask s_selbits[howmany(2048, NFDBITS)]; 773 fd_mask *ibits[3], *obits[3], *selbits, *sbp; 774 struct timeval atv, rtv, ttv; 775 int ncoll, error, timo; 776 u_int nbufbytes, ncpbytes, nfdbits; 777 778 if (uap->nd < 0) 779 return (EINVAL); 780 if (uap->nd > p->p_fd->fd_nfiles) 781 uap->nd = p->p_fd->fd_nfiles; /* forgiving; slightly wrong */ 782 783 /* 784 * Allocate just enough bits for the non-null fd_sets. Use the 785 * preallocated auto buffer if possible. 786 */ 787 nfdbits = roundup(uap->nd, NFDBITS); 788 ncpbytes = nfdbits / NBBY; 789 nbufbytes = 0; 790 if (uap->in != NULL) 791 nbufbytes += 2 * ncpbytes; 792 if (uap->ou != NULL) 793 nbufbytes += 2 * ncpbytes; 794 if (uap->ex != NULL) 795 nbufbytes += 2 * ncpbytes; 796 if (nbufbytes <= sizeof s_selbits) 797 selbits = &s_selbits[0]; 798 else 799 selbits = kmalloc(nbufbytes, M_SELECT, M_WAITOK); 800 801 /* 802 * Assign pointers into the bit buffers and fetch the input bits. 803 * Put the output buffers together so that they can be bzeroed 804 * together. 805 */ 806 sbp = selbits; 807 #define getbits(name, x) \ 808 do { \ 809 if (uap->name == NULL) \ 810 ibits[x] = NULL; \ 811 else { \ 812 ibits[x] = sbp + nbufbytes / 2 / sizeof *sbp; \ 813 obits[x] = sbp; \ 814 sbp += ncpbytes / sizeof *sbp; \ 815 error = copyin(uap->name, ibits[x], ncpbytes); \ 816 if (error != 0) \ 817 goto done; \ 818 } \ 819 } while (0) 820 getbits(in, 0); 821 getbits(ou, 1); 822 getbits(ex, 2); 823 #undef getbits 824 if (nbufbytes != 0) 825 bzero(selbits, nbufbytes / 2); 826 827 if (uap->tv) { 828 error = copyin((caddr_t)uap->tv, (caddr_t)&atv, 829 sizeof (atv)); 830 if (error) 831 goto done; 832 if (itimerfix(&atv)) { 833 error = EINVAL; 834 goto done; 835 } 836 getmicrouptime(&rtv); 837 timevaladd(&atv, &rtv); 838 } else { 839 atv.tv_sec = 0; 840 atv.tv_usec = 0; 841 } 842 timo = 0; 843 retry: 844 ncoll = nselcoll; 845 lp->lwp_flag |= LWP_SELECT; 846 error = selscan(p, ibits, obits, uap->nd, &uap->sysmsg_result); 847 if (error || uap->sysmsg_result) 848 goto done; 849 if (atv.tv_sec || atv.tv_usec) { 850 getmicrouptime(&rtv); 851 if (timevalcmp(&rtv, &atv, >=)) 852 goto done; 853 ttv = atv; 854 timevalsub(&ttv, &rtv); 855 timo = ttv.tv_sec > 24 * 60 * 60 ? 856 24 * 60 * 60 * hz : tvtohz_high(&ttv); 857 } 858 crit_enter(); 859 if ((lp->lwp_flag & LWP_SELECT) == 0 || nselcoll != ncoll) { 860 crit_exit(); 861 goto retry; 862 } 863 lp->lwp_flag &= ~LWP_SELECT; 864 865 error = tsleep((caddr_t)&selwait, PCATCH, "select", timo); 866 867 crit_exit(); 868 if (error == 0) 869 goto retry; 870 done: 871 lp->lwp_flag &= ~LWP_SELECT; 872 /* select is not restarted after signals... */ 873 if (error == ERESTART) 874 error = EINTR; 875 if (error == EWOULDBLOCK) 876 error = 0; 877 #define putbits(name, x) \ 878 if (uap->name && (error2 = copyout(obits[x], uap->name, ncpbytes))) \ 879 error = error2; 880 if (error == 0) { 881 int error2; 882 883 putbits(in, 0); 884 putbits(ou, 1); 885 putbits(ex, 2); 886 #undef putbits 887 } 888 if (selbits != &s_selbits[0]) 889 kfree(selbits, M_SELECT); 890 return (error); 891 } 892 893 static int 894 selscan(struct proc *p, fd_mask **ibits, fd_mask **obits, int nfd, int *res) 895 { 896 int msk, i, fd; 897 fd_mask bits; 898 struct file *fp; 899 int n = 0; 900 /* Note: backend also returns POLLHUP/POLLERR if appropriate. */ 901 static int flag[3] = { POLLRDNORM, POLLWRNORM, POLLRDBAND }; 902 903 for (msk = 0; msk < 3; msk++) { 904 if (ibits[msk] == NULL) 905 continue; 906 for (i = 0; i < nfd; i += NFDBITS) { 907 bits = ibits[msk][i/NFDBITS]; 908 /* ffs(int mask) not portable, fd_mask is long */ 909 for (fd = i; bits && fd < nfd; fd++, bits >>= 1) { 910 if (!(bits & 1)) 911 continue; 912 fp = holdfp(p->p_fd, fd, -1); 913 if (fp == NULL) 914 return (EBADF); 915 if (fo_poll(fp, flag[msk], fp->f_cred)) { 916 obits[msk][(fd)/NFDBITS] |= 917 ((fd_mask)1 << ((fd) % NFDBITS)); 918 n++; 919 } 920 fdrop(fp); 921 } 922 } 923 } 924 *res = n; 925 return (0); 926 } 927 928 /* 929 * Poll system call. 930 */ 931 int 932 sys_poll(struct poll_args *uap) 933 { 934 struct pollfd *bits; 935 struct pollfd smallbits[32]; 936 struct timeval atv, rtv, ttv; 937 int ncoll, error = 0, timo; 938 u_int nfds; 939 size_t ni; 940 struct lwp *lp = curthread->td_lwp; 941 struct proc *p = curproc; 942 943 nfds = uap->nfds; 944 /* 945 * This is kinda bogus. We have fd limits, but that is not 946 * really related to the size of the pollfd array. Make sure 947 * we let the process use at least FD_SETSIZE entries and at 948 * least enough for the current limits. We want to be reasonably 949 * safe, but not overly restrictive. 950 */ 951 if (nfds > p->p_rlimit[RLIMIT_NOFILE].rlim_cur && nfds > FD_SETSIZE) 952 return (EINVAL); 953 ni = nfds * sizeof(struct pollfd); 954 if (ni > sizeof(smallbits)) 955 bits = kmalloc(ni, M_TEMP, M_WAITOK); 956 else 957 bits = smallbits; 958 error = copyin(uap->fds, bits, ni); 959 if (error) 960 goto done; 961 if (uap->timeout != INFTIM) { 962 atv.tv_sec = uap->timeout / 1000; 963 atv.tv_usec = (uap->timeout % 1000) * 1000; 964 if (itimerfix(&atv)) { 965 error = EINVAL; 966 goto done; 967 } 968 getmicrouptime(&rtv); 969 timevaladd(&atv, &rtv); 970 } else { 971 atv.tv_sec = 0; 972 atv.tv_usec = 0; 973 } 974 timo = 0; 975 retry: 976 ncoll = nselcoll; 977 lp->lwp_flag |= LWP_SELECT; 978 error = pollscan(p, bits, nfds, &uap->sysmsg_result); 979 if (error || uap->sysmsg_result) 980 goto done; 981 if (atv.tv_sec || atv.tv_usec) { 982 getmicrouptime(&rtv); 983 if (timevalcmp(&rtv, &atv, >=)) 984 goto done; 985 ttv = atv; 986 timevalsub(&ttv, &rtv); 987 timo = ttv.tv_sec > 24 * 60 * 60 ? 988 24 * 60 * 60 * hz : tvtohz_high(&ttv); 989 } 990 crit_enter(); 991 if ((lp->lwp_flag & LWP_SELECT) == 0 || nselcoll != ncoll) { 992 crit_exit(); 993 goto retry; 994 } 995 lp->lwp_flag &= ~LWP_SELECT; 996 error = tsleep((caddr_t)&selwait, PCATCH, "poll", timo); 997 crit_exit(); 998 if (error == 0) 999 goto retry; 1000 done: 1001 lp->lwp_flag &= ~LWP_SELECT; 1002 /* poll is not restarted after signals... */ 1003 if (error == ERESTART) 1004 error = EINTR; 1005 if (error == EWOULDBLOCK) 1006 error = 0; 1007 if (error == 0) { 1008 error = copyout(bits, uap->fds, ni); 1009 if (error) 1010 goto out; 1011 } 1012 out: 1013 if (ni > sizeof(smallbits)) 1014 kfree(bits, M_TEMP); 1015 return (error); 1016 } 1017 1018 static int 1019 pollscan(struct proc *p, struct pollfd *fds, u_int nfd, int *res) 1020 { 1021 int i; 1022 struct file *fp; 1023 int n = 0; 1024 1025 for (i = 0; i < nfd; i++, fds++) { 1026 if (fds->fd >= p->p_fd->fd_nfiles) { 1027 fds->revents = POLLNVAL; 1028 n++; 1029 } else if (fds->fd < 0) { 1030 fds->revents = 0; 1031 } else { 1032 fp = holdfp(p->p_fd, fds->fd, -1); 1033 if (fp == NULL) { 1034 fds->revents = POLLNVAL; 1035 n++; 1036 } else { 1037 /* 1038 * Note: backend also returns POLLHUP and 1039 * POLLERR if appropriate. 1040 */ 1041 fds->revents = fo_poll(fp, fds->events, 1042 fp->f_cred); 1043 if (fds->revents != 0) 1044 n++; 1045 fdrop(fp); 1046 } 1047 } 1048 } 1049 *res = n; 1050 return (0); 1051 } 1052 1053 /* 1054 * OpenBSD poll system call. 1055 * XXX this isn't quite a true representation.. OpenBSD uses select ops. 1056 */ 1057 int 1058 sys_openbsd_poll(struct openbsd_poll_args *uap) 1059 { 1060 return (sys_poll((struct poll_args *)uap)); 1061 } 1062 1063 /*ARGSUSED*/ 1064 int 1065 seltrue(cdev_t dev, int events) 1066 { 1067 return (events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM)); 1068 } 1069 1070 /* 1071 * Record a select request. A global wait must be used since a process/thread 1072 * might go away after recording its request. 1073 */ 1074 void 1075 selrecord(struct thread *selector, struct selinfo *sip) 1076 { 1077 struct proc *p; 1078 struct lwp *lp = NULL; 1079 1080 if (selector->td_lwp == NULL) 1081 panic("selrecord: thread needs a process"); 1082 1083 if (sip->si_pid == selector->td_proc->p_pid && 1084 sip->si_tid == selector->td_lwp->lwp_tid) 1085 return; 1086 if (sip->si_pid && (p = pfind(sip->si_pid))) { 1087 FOREACH_LWP_IN_PROC(lp, p) { 1088 if (sip->si_tid == lp->lwp_tid) 1089 break; 1090 } 1091 } 1092 if (lp != NULL && lp->lwp_wchan == (caddr_t)&selwait) { 1093 sip->si_flags |= SI_COLL; 1094 } else { 1095 sip->si_pid = selector->td_proc->p_pid; 1096 } 1097 } 1098 1099 /* 1100 * Do a wakeup when a selectable event occurs. 1101 */ 1102 void 1103 selwakeup(struct selinfo *sip) 1104 { 1105 struct proc *p; 1106 struct lwp *lp = NULL; 1107 1108 if (sip->si_pid == 0) 1109 return; 1110 if (sip->si_flags & SI_COLL) { 1111 nselcoll++; 1112 sip->si_flags &= ~SI_COLL; 1113 wakeup((caddr_t)&selwait); /* YYY fixable */ 1114 } 1115 p = pfind(sip->si_pid); 1116 sip->si_pid = 0; 1117 if (p == NULL) 1118 return; 1119 FOREACH_LWP_IN_PROC(lp, p) { 1120 if (lp->lwp_tid == sip->si_tid) 1121 break; 1122 } 1123 if (lp == NULL) 1124 return; 1125 1126 crit_enter(); 1127 if (lp->lwp_wchan == (caddr_t)&selwait) { 1128 /* 1129 * Flag the process to break the tsleep when 1130 * setrunnable is called, but only call setrunnable 1131 * here if the process is not in a stopped state. 1132 */ 1133 lp->lwp_flag |= LWP_BREAKTSLEEP; 1134 if (p->p_stat != SSTOP) 1135 setrunnable(lp); 1136 } else if (lp->lwp_flag & LWP_SELECT) { 1137 lp->lwp_flag &= ~LWP_SELECT; 1138 } 1139 crit_exit(); 1140 } 1141 1142