1 /* $NetBSD: linux_misc.c,v 1.35 1997/10/10 22:16:10 fvdl Exp $ */ 2 3 /* 4 * Copyright (c) 1995 Frank van der Linden 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. All advertising materials mentioning features or use of this software 16 * must display the following acknowledgement: 17 * This product includes software developed for the NetBSD Project 18 * by Frank van der Linden 19 * 4. The name of the author may not be used to endorse or promote products 20 * derived from this software without specific prior written permission 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 23 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 24 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 25 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 26 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 27 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 28 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 29 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 30 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 31 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 32 */ 33 34 /* 35 * Linux compatibility module. Try to deal with various Linux system calls. 36 */ 37 38 #include <sys/param.h> 39 #include <sys/systm.h> 40 #include <sys/namei.h> 41 #include <sys/proc.h> 42 #include <sys/dirent.h> 43 #include <sys/file.h> 44 #include <sys/stat.h> 45 #include <sys/filedesc.h> 46 #include <sys/ioctl.h> 47 #include <sys/kernel.h> 48 #include <sys/malloc.h> 49 #include <sys/mbuf.h> 50 #include <sys/mman.h> 51 #include <sys/mount.h> 52 #include <sys/ptrace.h> 53 #include <sys/resource.h> 54 #include <sys/resourcevar.h> 55 #include <sys/signal.h> 56 #include <sys/signalvar.h> 57 #include <sys/socket.h> 58 #include <sys/time.h> 59 #include <sys/times.h> 60 #include <sys/vnode.h> 61 #include <sys/uio.h> 62 #include <sys/wait.h> 63 #include <sys/utsname.h> 64 #include <sys/unistd.h> 65 66 #include <sys/syscallargs.h> 67 68 #include <vm/vm.h> 69 #include <vm/vm_param.h> 70 71 #include <compat/linux/linux_types.h> 72 #include <compat/linux/linux_fcntl.h> 73 #include <compat/linux/linux_mmap.h> 74 #include <compat/linux/linux_signal.h> 75 #include <compat/linux/linux_syscallargs.h> 76 #include <compat/linux/linux_util.h> 77 #include <compat/linux/linux_dirent.h> 78 79 /* linux_misc.c */ 80 static void bsd_to_linux_wstat __P((int *)); 81 static void bsd_to_linux_statfs __P((struct statfs *, struct linux_statfs *)); 82 int linux_select1 __P((struct proc *, register_t *, int, fd_set *, fd_set *, 83 fd_set *, struct timeval *)); 84 85 /* 86 * The information on a terminated (or stopped) process needs 87 * to be converted in order for Linux binaries to get a valid signal 88 * number out of it. 89 */ 90 static void 91 bsd_to_linux_wstat(status) 92 int *status; 93 { 94 95 if (WIFSIGNALED(*status)) 96 *status = (*status & ~0177) | 97 bsd_to_linux_sig[WTERMSIG(*status)]; 98 else if (WIFSTOPPED(*status)) 99 *status = (*status & ~0xff00) | 100 (bsd_to_linux_sig[WSTOPSIG(*status)] << 8); 101 } 102 103 /* 104 * waitpid(2). Passed on to the NetBSD call, surrounded by code to 105 * reserve some space for a NetBSD-style wait status, and converting 106 * it to what Linux wants. 107 */ 108 int 109 linux_sys_waitpid(p, v, retval) 110 struct proc *p; 111 void *v; 112 register_t *retval; 113 { 114 struct linux_sys_waitpid_args /* { 115 syscallarg(int) pid; 116 syscallarg(int *) status; 117 syscallarg(int) options; 118 } */ *uap = v; 119 struct sys_wait4_args w4a; 120 int error, *status, tstat; 121 caddr_t sg; 122 123 if (SCARG(uap, status) != NULL) { 124 sg = stackgap_init(p->p_emul); 125 status = (int *) stackgap_alloc(&sg, sizeof status); 126 } else 127 status = NULL; 128 129 SCARG(&w4a, pid) = SCARG(uap, pid); 130 SCARG(&w4a, status) = status; 131 SCARG(&w4a, options) = SCARG(uap, options); 132 SCARG(&w4a, rusage) = NULL; 133 134 if ((error = sys_wait4(p, &w4a, retval))) 135 return error; 136 137 p->p_siglist &= ~sigmask(SIGCHLD); 138 139 if (status != NULL) { 140 if ((error = copyin(status, &tstat, sizeof tstat))) 141 return error; 142 143 bsd_to_linux_wstat(&tstat); 144 return copyout(&tstat, SCARG(uap, status), sizeof tstat); 145 } 146 147 return 0; 148 } 149 150 /* 151 * This is very much the same as waitpid() 152 */ 153 int 154 linux_sys_wait4(p, v, retval) 155 struct proc *p; 156 void *v; 157 register_t *retval; 158 { 159 struct linux_sys_wait4_args /* { 160 syscallarg(int) pid; 161 syscallarg(int *) status; 162 syscallarg(int) options; 163 syscallarg(struct rusage *) rusage; 164 } */ *uap = v; 165 struct sys_wait4_args w4a; 166 int error, *status, tstat; 167 caddr_t sg; 168 169 if (SCARG(uap, status) != NULL) { 170 sg = stackgap_init(p->p_emul); 171 status = (int *) stackgap_alloc(&sg, sizeof status); 172 } else 173 status = NULL; 174 175 SCARG(&w4a, pid) = SCARG(uap, pid); 176 SCARG(&w4a, status) = status; 177 SCARG(&w4a, options) = SCARG(uap, options); 178 SCARG(&w4a, rusage) = SCARG(uap, rusage); 179 180 if ((error = sys_wait4(p, &w4a, retval))) 181 return error; 182 183 p->p_siglist &= ~sigmask(SIGCHLD); 184 185 if (status != NULL) { 186 if ((error = copyin(status, &tstat, sizeof tstat))) 187 return error; 188 189 bsd_to_linux_wstat(&tstat); 190 return copyout(&tstat, SCARG(uap, status), sizeof tstat); 191 } 192 193 return 0; 194 } 195 196 /* 197 * This is the old brk(2) call. I don't think anything in the Linux 198 * world uses this anymore 199 */ 200 int 201 linux_sys_break(p, v, retval) 202 struct proc *p; 203 void *v; 204 register_t *retval; 205 { 206 #if 0 207 struct linux_sys_brk_args /* { 208 syscallarg(char *) nsize; 209 } */ *uap = v; 210 #endif 211 212 return ENOSYS; 213 } 214 215 /* 216 * Linux brk(2). The check if the new address is >= the old one is 217 * done in the kernel in Linux. NetBSD does it in the library. 218 */ 219 int 220 linux_sys_brk(p, v, retval) 221 struct proc *p; 222 void *v; 223 register_t *retval; 224 { 225 struct linux_sys_brk_args /* { 226 syscallarg(char *) nsize; 227 } */ *uap = v; 228 char *nbrk = SCARG(uap, nsize); 229 struct sys_obreak_args oba; 230 struct vmspace *vm = p->p_vmspace; 231 caddr_t oldbrk; 232 233 oldbrk = vm->vm_daddr + ctob(vm->vm_dsize); 234 /* 235 * XXX inconsistent.. Linux always returns at least the old 236 * brk value, but it will be page-aligned if this fails, 237 * and possibly not page aligned if it succeeds (the user 238 * supplied pointer is returned). 239 */ 240 SCARG(&oba, nsize) = nbrk; 241 242 if ((caddr_t) nbrk > vm->vm_daddr && sys_obreak(p, &oba, retval) == 0) 243 retval[0] = (register_t)nbrk; 244 else 245 retval[0] = (register_t)oldbrk; 246 247 return 0; 248 } 249 250 /* 251 * I wonder why Linux has gettimeofday() _and_ time().. Still, we 252 * need to deal with it. 253 */ 254 int 255 linux_sys_time(p, v, retval) 256 struct proc *p; 257 void *v; 258 register_t *retval; 259 { 260 struct linux_sys_time_args /* { 261 linux_time_t *t; 262 } */ *uap = v; 263 struct timeval atv; 264 linux_time_t tt; 265 int error; 266 267 microtime(&atv); 268 269 tt = atv.tv_sec; 270 if (SCARG(uap, t) && (error = copyout(&tt, SCARG(uap, t), sizeof tt))) 271 return error; 272 273 retval[0] = tt; 274 return 0; 275 } 276 277 /* 278 * Convert BSD statfs structure to Linux statfs structure. 279 * The Linux structure has less fields, and it also wants 280 * the length of a name in a dir entry in a field, which 281 * we fake (probably the wrong way). 282 */ 283 static void 284 bsd_to_linux_statfs(bsp, lsp) 285 struct statfs *bsp; 286 struct linux_statfs *lsp; 287 { 288 289 lsp->l_ftype = bsp->f_type; 290 lsp->l_fbsize = bsp->f_bsize; 291 lsp->l_fblocks = bsp->f_blocks; 292 lsp->l_fbfree = bsp->f_bfree; 293 lsp->l_fbavail = bsp->f_bavail; 294 lsp->l_ffiles = bsp->f_files; 295 lsp->l_fffree = bsp->f_ffree; 296 lsp->l_ffsid.val[0] = bsp->f_fsid.val[0]; 297 lsp->l_ffsid.val[1] = bsp->f_fsid.val[1]; 298 lsp->l_fnamelen = MAXNAMLEN; /* XXX */ 299 } 300 301 /* 302 * Implement the fs stat functions. Straightforward. 303 */ 304 int 305 linux_sys_statfs(p, v, retval) 306 struct proc *p; 307 void *v; 308 register_t *retval; 309 { 310 struct linux_sys_statfs_args /* { 311 syscallarg(char *) path; 312 syscallarg(struct linux_statfs *) sp; 313 } */ *uap = v; 314 struct statfs btmp, *bsp; 315 struct linux_statfs ltmp; 316 struct sys_statfs_args bsa; 317 caddr_t sg; 318 int error; 319 320 sg = stackgap_init(p->p_emul); 321 bsp = (struct statfs *) stackgap_alloc(&sg, sizeof (struct statfs)); 322 323 LINUX_CHECK_ALT_EXIST(p, &sg, SCARG(uap, path)); 324 325 SCARG(&bsa, path) = SCARG(uap, path); 326 SCARG(&bsa, buf) = bsp; 327 328 if ((error = sys_statfs(p, &bsa, retval))) 329 return error; 330 331 if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp))) 332 return error; 333 334 bsd_to_linux_statfs(&btmp, <mp); 335 336 return copyout((caddr_t) <mp, (caddr_t) SCARG(uap, sp), sizeof ltmp); 337 } 338 339 int 340 linux_sys_fstatfs(p, v, retval) 341 struct proc *p; 342 void *v; 343 register_t *retval; 344 { 345 struct linux_sys_fstatfs_args /* { 346 syscallarg(int) fd; 347 syscallarg(struct linux_statfs *) sp; 348 } */ *uap = v; 349 struct statfs btmp, *bsp; 350 struct linux_statfs ltmp; 351 struct sys_fstatfs_args bsa; 352 caddr_t sg; 353 int error; 354 355 sg = stackgap_init(p->p_emul); 356 bsp = (struct statfs *) stackgap_alloc(&sg, sizeof (struct statfs)); 357 358 SCARG(&bsa, fd) = SCARG(uap, fd); 359 SCARG(&bsa, buf) = bsp; 360 361 if ((error = sys_fstatfs(p, &bsa, retval))) 362 return error; 363 364 if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp))) 365 return error; 366 367 bsd_to_linux_statfs(&btmp, <mp); 368 369 return copyout((caddr_t) <mp, (caddr_t) SCARG(uap, sp), sizeof ltmp); 370 } 371 372 /* 373 * uname(). Just copy the info from the various strings stored in the 374 * kernel, and put it in the Linux utsname structure. That structure 375 * is almost the same as the NetBSD one, only it has fields 65 characters 376 * long, and an extra domainname field. 377 */ 378 int 379 linux_sys_uname(p, v, retval) 380 struct proc *p; 381 void *v; 382 register_t *retval; 383 { 384 struct linux_sys_uname_args /* { 385 syscallarg(struct linux_utsname *) up; 386 } */ *uap = v; 387 extern char ostype[], hostname[], osrelease[], version[], machine[], 388 domainname[]; 389 struct linux_utsname luts; 390 int len; 391 char *cp; 392 393 strncpy(luts.l_sysname, ostype, sizeof(luts.l_sysname)); 394 strncpy(luts.l_nodename, hostname, sizeof(luts.l_nodename)); 395 strncpy(luts.l_release, osrelease, sizeof(luts.l_release)); 396 strncpy(luts.l_version, version, sizeof(luts.l_version)); 397 strncpy(luts.l_machine, machine, sizeof(luts.l_machine)); 398 strncpy(luts.l_domainname, domainname, sizeof(luts.l_domainname)); 399 400 /* This part taken from the the uname() in libc */ 401 len = sizeof(luts.l_version); 402 for (cp = luts.l_version; len--; ++cp) 403 if (*cp == '\n' || *cp == '\t') 404 if (len > 1) 405 *cp = ' '; 406 else 407 *cp = '\0'; 408 409 return copyout(&luts, SCARG(uap, up), sizeof(luts)); 410 } 411 412 int 413 linux_sys_olduname(p, v, retval) 414 struct proc *p; 415 void *v; 416 register_t *retval; 417 { 418 struct linux_sys_uname_args /* { 419 syscallarg(struct linux_oldutsname *) up; 420 } */ *uap = v; 421 extern char ostype[], hostname[], osrelease[], version[], machine[]; 422 struct linux_oldutsname luts; 423 int len; 424 char *cp; 425 426 strncpy(luts.l_sysname, ostype, sizeof(luts.l_sysname)); 427 strncpy(luts.l_nodename, hostname, sizeof(luts.l_nodename)); 428 strncpy(luts.l_release, osrelease, sizeof(luts.l_release)); 429 strncpy(luts.l_version, version, sizeof(luts.l_version)); 430 strncpy(luts.l_machine, machine, sizeof(luts.l_machine)); 431 432 /* This part taken from the the uname() in libc */ 433 len = sizeof(luts.l_version); 434 for (cp = luts.l_version; len--; ++cp) 435 if (*cp == '\n' || *cp == '\t') 436 if (len > 1) 437 *cp = ' '; 438 else 439 *cp = '\0'; 440 441 return copyout(&luts, SCARG(uap, up), sizeof(luts)); 442 } 443 444 int 445 linux_sys_oldolduname(p, v, retval) 446 struct proc *p; 447 void *v; 448 register_t *retval; 449 { 450 struct linux_sys_uname_args /* { 451 syscallarg(struct linux_oldoldutsname *) up; 452 } */ *uap = v; 453 extern char ostype[], hostname[], osrelease[], version[], machine[]; 454 struct linux_oldoldutsname luts; 455 int len; 456 char *cp; 457 458 strncpy(luts.l_sysname, ostype, sizeof(luts.l_sysname)); 459 strncpy(luts.l_nodename, hostname, sizeof(luts.l_nodename)); 460 strncpy(luts.l_release, osrelease, sizeof(luts.l_release)); 461 strncpy(luts.l_version, version, sizeof(luts.l_version)); 462 strncpy(luts.l_machine, machine, sizeof(luts.l_machine)); 463 464 /* This part taken from the the uname() in libc */ 465 len = sizeof(luts.l_version); 466 for (cp = luts.l_version; len--; ++cp) 467 if (*cp == '\n' || *cp == '\t') 468 if (len > 1) 469 *cp = ' '; 470 else 471 *cp = '\0'; 472 473 return copyout(&luts, SCARG(uap, up), sizeof(luts)); 474 } 475 476 /* 477 * Linux wants to pass everything to a syscall in registers. However, 478 * mmap() has 6 of them. Oops: out of register error. They just pass 479 * everything in a structure. 480 */ 481 int 482 linux_sys_mmap(p, v, retval) 483 struct proc *p; 484 void *v; 485 register_t *retval; 486 { 487 struct linux_sys_mmap_args /* { 488 syscallarg(struct linux_mmap *) lmp; 489 } */ *uap = v; 490 struct linux_mmap lmap; 491 struct sys_mmap_args cma; 492 int error, flags; 493 494 if ((error = copyin(SCARG(uap, lmp), &lmap, sizeof lmap))) 495 return error; 496 497 flags = 0; 498 flags |= cvtto_bsd_mask(lmap.lm_flags, LINUX_MAP_SHARED, MAP_SHARED); 499 flags |= cvtto_bsd_mask(lmap.lm_flags, LINUX_MAP_PRIVATE, MAP_PRIVATE); 500 flags |= cvtto_bsd_mask(lmap.lm_flags, LINUX_MAP_FIXED, MAP_FIXED); 501 flags |= cvtto_bsd_mask(lmap.lm_flags, LINUX_MAP_ANON, MAP_ANON); 502 503 SCARG(&cma,addr) = lmap.lm_addr; 504 SCARG(&cma,len) = lmap.lm_len; 505 if (lmap.lm_prot & VM_PROT_WRITE) /* XXX */ 506 lmap.lm_prot |= VM_PROT_READ; 507 SCARG(&cma,prot) = lmap.lm_prot; 508 SCARG(&cma,flags) = flags; 509 SCARG(&cma,fd) = lmap.lm_fd; 510 SCARG(&cma,pad) = 0; 511 SCARG(&cma,pos) = lmap.lm_pos; 512 513 return sys_mmap(p, &cma, retval); 514 } 515 516 int 517 linux_sys_mremap(p, v, retval) 518 struct proc *p; 519 void *v; 520 register_t *retval; 521 { 522 #ifdef notyet 523 struct linux_sys_mremap_args /* { 524 syscallarg(void *) old_address; 525 syscallarg(size_t) old_size; 526 syscallarg(size_t) new_size; 527 syscallarg(u_long) flags; 528 } */ *uap = v; 529 #endif 530 531 return ENOMEM; 532 } 533 534 int 535 linux_sys_msync(p, v, retval) 536 struct proc *p; 537 void *v; 538 register_t *retval; 539 { 540 struct linux_sys_msync_args /* { 541 syscallarg(caddr_t) addr; 542 syscallarg(int) len; 543 syscallarg(int) fl; 544 } */ *uap = v; 545 546 struct sys_msync_args bma; 547 548 /* flags are ignored */ 549 SCARG(&bma, addr) = SCARG(uap, addr); 550 SCARG(&bma, len) = SCARG(uap, len); 551 552 return sys_msync(p, &bma, retval); 553 } 554 555 /* 556 * This code is partly stolen from src/lib/libc/compat-43/times.c 557 * XXX - CLK_TCK isn't declared in /sys, just in <time.h>, done here 558 */ 559 560 #define CLK_TCK 100 561 #define CONVTCK(r) (r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK)) 562 563 int 564 linux_sys_times(p, v, retval) 565 struct proc *p; 566 void *v; 567 register_t *retval; 568 { 569 struct linux_sys_times_args /* { 570 syscallarg(struct times *) tms; 571 } */ *uap = v; 572 struct timeval t; 573 struct linux_tms ltms; 574 struct rusage ru; 575 int error, s; 576 577 calcru(p, &ru.ru_utime, &ru.ru_stime, NULL); 578 ltms.ltms_utime = CONVTCK(ru.ru_utime); 579 ltms.ltms_stime = CONVTCK(ru.ru_stime); 580 581 ltms.ltms_cutime = CONVTCK(p->p_stats->p_cru.ru_utime); 582 ltms.ltms_cstime = CONVTCK(p->p_stats->p_cru.ru_stime); 583 584 if ((error = copyout(<ms, SCARG(uap, tms), sizeof ltms))) 585 return error; 586 587 s = splclock(); 588 timersub(&time, &boottime, &t); 589 splx(s); 590 591 retval[0] = ((linux_clock_t)(CONVTCK(t))); 592 return 0; 593 } 594 595 /* 596 * NetBSD passes fd[0] in retval[0], and fd[1] in retval[1]. 597 * Linux directly passes the pointer. 598 */ 599 int 600 linux_sys_pipe(p, v, retval) 601 struct proc *p; 602 void *v; 603 register_t *retval; 604 { 605 struct linux_sys_pipe_args /* { 606 syscallarg(int *) pfds; 607 } */ *uap = v; 608 int error; 609 610 if ((error = sys_pipe(p, 0, retval))) 611 return error; 612 613 /* Assumes register_t is an int */ 614 615 if ((error = copyout(retval, SCARG(uap, pfds), 2 * sizeof (int)))) 616 return error; 617 618 retval[0] = 0; 619 return 0; 620 } 621 622 /* 623 * Alarm. This is a libc call which uses setitimer(2) in NetBSD. 624 * Fiddle with the timers to make it work. 625 */ 626 int 627 linux_sys_alarm(p, v, retval) 628 struct proc *p; 629 void *v; 630 register_t *retval; 631 { 632 struct linux_sys_alarm_args /* { 633 syscallarg(unsigned int) secs; 634 } */ *uap = v; 635 int s; 636 struct itimerval *itp, it; 637 638 itp = &p->p_realtimer; 639 s = splclock(); 640 /* 641 * Clear any pending timer alarms. 642 */ 643 untimeout(realitexpire, p); 644 timerclear(&itp->it_interval); 645 if (timerisset(&itp->it_value) && 646 timercmp(&itp->it_value, &time, >)) 647 timersub(&itp->it_value, &time, &itp->it_value); 648 /* 649 * Return how many seconds were left (rounded up) 650 */ 651 retval[0] = itp->it_value.tv_sec; 652 if (itp->it_value.tv_usec) 653 retval[0]++; 654 655 /* 656 * alarm(0) just resets the timer. 657 */ 658 if (SCARG(uap, secs) == 0) { 659 timerclear(&itp->it_value); 660 splx(s); 661 return 0; 662 } 663 664 /* 665 * Check the new alarm time for sanity, and set it. 666 */ 667 timerclear(&it.it_interval); 668 it.it_value.tv_sec = SCARG(uap, secs); 669 it.it_value.tv_usec = 0; 670 if (itimerfix(&it.it_value) || itimerfix(&it.it_interval)) { 671 splx(s); 672 return (EINVAL); 673 } 674 675 if (timerisset(&it.it_value)) { 676 timeradd(&it.it_value, &time, &it.it_value); 677 timeout(realitexpire, p, hzto(&it.it_value)); 678 } 679 p->p_realtimer = it; 680 splx(s); 681 682 return 0; 683 } 684 685 /* 686 * utime(). Do conversion to things that utimes() understands, 687 * and pass it on. 688 */ 689 int 690 linux_sys_utime(p, v, retval) 691 struct proc *p; 692 void *v; 693 register_t *retval; 694 { 695 struct linux_sys_utime_args /* { 696 syscallarg(char *) path; 697 syscallarg(struct linux_utimbuf *)times; 698 } */ *uap = v; 699 caddr_t sg; 700 int error; 701 struct sys_utimes_args ua; 702 struct timeval tv[2], *tvp; 703 struct linux_utimbuf lut; 704 705 sg = stackgap_init(p->p_emul); 706 LINUX_CHECK_ALT_EXIST(p, &sg, SCARG(uap, path)); 707 708 SCARG(&ua, path) = SCARG(uap, path); 709 710 if (SCARG(uap, times) != NULL) { 711 if ((error = copyin(SCARG(uap, times), &lut, sizeof lut))) 712 return error; 713 tv[0].tv_usec = tv[1].tv_usec = 0; 714 tv[0].tv_sec = lut.l_actime; 715 tv[1].tv_sec = lut.l_modtime; 716 tvp = (struct timeval *) stackgap_alloc(&sg, sizeof(tv)); 717 if ((error = copyout(tv, tvp, sizeof tv))) 718 return error; 719 SCARG(&ua, tptr) = tvp; 720 } 721 else 722 SCARG(&ua, tptr) = NULL; 723 724 return sys_utimes(p, uap, retval); 725 } 726 727 /* 728 * The old Linux readdir was only able to read one entry at a time, 729 * even though it had a 'count' argument. In fact, the emulation 730 * of the old call was better than the original, because it did handle 731 * the count arg properly. Don't bother with it anymore now, and use 732 * it to distinguish between old and new. The difference is that the 733 * newer one actually does multiple entries, and the reclen field 734 * really is the reclen, not the namelength. 735 */ 736 int 737 linux_sys_readdir(p, v, retval) 738 struct proc *p; 739 void *v; 740 register_t *retval; 741 { 742 struct linux_sys_readdir_args /* { 743 syscallarg(int) fd; 744 syscallarg(struct linux_dirent *) dent; 745 syscallarg(unsigned int) count; 746 } */ *uap = v; 747 748 SCARG(uap, count) = 1; 749 return linux_sys_getdents(p, uap, retval); 750 } 751 752 /* 753 * Linux 'readdir' call. This code is mostly taken from the 754 * SunOS getdents call (see compat/sunos/sunos_misc.c), though 755 * an attempt has been made to keep it a little cleaner (failing 756 * miserably, because of the cruft needed if count 1 is passed). 757 * 758 * The d_off field should contain the offset of the next valid entry, 759 * but in Linux it has the offset of the entry itself. We emulate 760 * that bug here. 761 * 762 * Read in BSD-style entries, convert them, and copy them out. 763 * 764 * Note that this doesn't handle union-mounted filesystems. 765 */ 766 int 767 linux_sys_getdents(p, v, retval) 768 struct proc *p; 769 void *v; 770 register_t *retval; 771 { 772 struct linux_sys_readdir_args /* { 773 syscallarg(int) fd; 774 syscallarg(caddr_t) dent; 775 syscallarg(unsigned int) count; 776 } */ *uap = v; 777 register struct dirent *bdp; 778 struct vnode *vp; 779 caddr_t inp, buf; /* BSD-format */ 780 int len, reclen; /* BSD-format */ 781 caddr_t outp; /* Linux-format */ 782 int resid, linux_reclen = 0; /* Linux-format */ 783 struct file *fp; 784 struct uio auio; 785 struct iovec aiov; 786 struct linux_dirent idb; 787 off_t off; /* true file offset */ 788 int buflen, error, eofflag, nbytes, oldcall; 789 struct vattr va; 790 off_t *cookiebuf, *cookie; 791 int ncookies; 792 793 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 794 return (error); 795 796 if ((fp->f_flag & FREAD) == 0) 797 return (EBADF); 798 799 vp = (struct vnode *)fp->f_data; 800 801 if (vp->v_type != VDIR) /* XXX vnode readdir op should do this */ 802 return (EINVAL); 803 804 if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p))) 805 return error; 806 807 nbytes = SCARG(uap, count); 808 if (nbytes == 1) { /* emulating old, broken behaviour */ 809 nbytes = sizeof (struct linux_dirent); 810 buflen = max(va.va_blocksize, nbytes); 811 oldcall = 1; 812 } else { 813 buflen = min(MAXBSIZE, nbytes); 814 if (buflen < va.va_blocksize) 815 buflen = va.va_blocksize; 816 oldcall = 0; 817 } 818 buf = malloc(buflen, M_TEMP, M_WAITOK); 819 ncookies = buflen / 16; 820 cookiebuf = malloc(ncookies * sizeof(*cookiebuf), M_TEMP, M_WAITOK); 821 822 VOP_LOCK(vp); 823 off = fp->f_offset; 824 again: 825 aiov.iov_base = buf; 826 aiov.iov_len = buflen; 827 auio.uio_iov = &aiov; 828 auio.uio_iovcnt = 1; 829 auio.uio_rw = UIO_READ; 830 auio.uio_segflg = UIO_SYSSPACE; 831 auio.uio_procp = p; 832 auio.uio_resid = buflen; 833 auio.uio_offset = off; 834 /* 835 * First we read into the malloc'ed buffer, then 836 * we massage it into user space, one record at a time. 837 */ 838 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, cookiebuf, 839 ncookies); 840 if (error) 841 goto out; 842 843 inp = buf; 844 outp = SCARG(uap, dent); 845 resid = nbytes; 846 if ((len = buflen - auio.uio_resid) == 0) 847 goto eof; 848 849 for (cookie = cookiebuf; len > 0; len -= reclen) { 850 bdp = (struct dirent *)inp; 851 reclen = bdp->d_reclen; 852 if (reclen & 3) 853 panic("linux_readdir"); 854 if (bdp->d_fileno == 0) { 855 inp += reclen; /* it is a hole; squish it out */ 856 off = *cookie++; 857 continue; 858 } 859 linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen); 860 if (reclen > len || resid < linux_reclen) { 861 /* entry too big for buffer, so just stop */ 862 outp++; 863 break; 864 } 865 /* 866 * Massage in place to make a Linux-shaped dirent (otherwise 867 * we have to worry about touching user memory outside of 868 * the copyout() call). 869 */ 870 idb.d_ino = (linux_ino_t)bdp->d_fileno; 871 /* 872 * The old readdir() call misuses the offset and reclen fields. 873 */ 874 if (oldcall) { 875 idb.d_off = (linux_off_t)linux_reclen; 876 idb.d_reclen = (u_short)bdp->d_namlen; 877 } else { 878 if (sizeof (linux_off_t) < 4 && (off >> 32) != 0) { 879 compat_offseterr(vp, "linux_getdents"); 880 error = EINVAL; 881 goto out; 882 } 883 idb.d_off = (linux_off_t)off; 884 idb.d_reclen = (u_short)linux_reclen; 885 } 886 strcpy(idb.d_name, bdp->d_name); 887 if ((error = copyout((caddr_t)&idb, outp, linux_reclen))) 888 goto out; 889 /* advance past this real entry */ 890 inp += reclen; 891 off = *cookie++; /* each entry points to itself */ 892 /* advance output past Linux-shaped entry */ 893 outp += linux_reclen; 894 resid -= linux_reclen; 895 if (oldcall) 896 break; 897 } 898 899 /* if we squished out the whole block, try again */ 900 if (outp == SCARG(uap, dent)) 901 goto again; 902 fp->f_offset = off; /* update the vnode offset */ 903 904 if (oldcall) 905 nbytes = resid + linux_reclen; 906 907 eof: 908 *retval = nbytes - resid; 909 out: 910 VOP_UNLOCK(vp); 911 free(cookiebuf, M_TEMP); 912 free(buf, M_TEMP); 913 return error; 914 } 915 916 /* 917 * Not sure why the arguments to this older version of select() were put 918 * into a structure, because there are 5, and that can all be handled 919 * in registers on the i386 like Linux wants to. 920 */ 921 int 922 linux_sys_oldselect(p, v, retval) 923 struct proc *p; 924 void *v; 925 register_t *retval; 926 { 927 struct linux_sys_oldselect_args /* { 928 syscallarg(struct linux_select *) lsp; 929 } */ *uap = v; 930 struct linux_select ls; 931 int error; 932 933 if ((error = copyin(SCARG(uap, lsp), &ls, sizeof(ls)))) 934 return error; 935 936 return linux_select1(p, retval, ls.nfds, ls.readfds, ls.writefds, 937 ls.exceptfds, ls.timeout); 938 } 939 940 /* 941 * Even when just using registers to pass arguments to syscalls you can 942 * have 5 of them on the i386. So this newer version of select() does 943 * this. 944 */ 945 int 946 linux_sys_select(p, v, retval) 947 struct proc *p; 948 void *v; 949 register_t *retval; 950 { 951 struct linux_sys_select_args /* { 952 syscallarg(int) nfds; 953 syscallarg(fd_set *) readfds; 954 syscallarg(fd_set *) writefds; 955 syscallarg(fd_set *) exceptfds; 956 syscallarg(struct timeval *) timeout; 957 } */ *uap = v; 958 959 return linux_select1(p, retval, SCARG(uap, nfds), SCARG(uap, readfds), 960 SCARG(uap, writefds), SCARG(uap, exceptfds), SCARG(uap, timeout)); 961 } 962 963 /* 964 * Common code for the old and new versions of select(). A couple of 965 * things are important: 966 * 1) return the amount of time left in the 'timeout' parameter 967 * 2) select never returns ERESTART on Linux, always return EINTR 968 */ 969 int 970 linux_select1(p, retval, nfds, readfds, writefds, exceptfds, timeout) 971 struct proc *p; 972 register_t *retval; 973 int nfds; 974 fd_set *readfds, *writefds, *exceptfds; 975 struct timeval *timeout; 976 { 977 struct sys_select_args bsa; 978 struct timeval tv0, tv1, utv, *tvp; 979 caddr_t sg; 980 int error; 981 982 SCARG(&bsa, nd) = nfds; 983 SCARG(&bsa, in) = readfds; 984 SCARG(&bsa, ou) = writefds; 985 SCARG(&bsa, ex) = exceptfds; 986 SCARG(&bsa, tv) = timeout; 987 988 /* 989 * Store current time for computation of the amount of 990 * time left. 991 */ 992 if (timeout) { 993 if ((error = copyin(timeout, &utv, sizeof(utv)))) 994 return error; 995 if (itimerfix(&utv)) { 996 /* 997 * The timeval was invalid. Convert it to something 998 * valid that will act as it does under Linux. 999 */ 1000 sg = stackgap_init(p->p_emul); 1001 tvp = stackgap_alloc(&sg, sizeof(utv)); 1002 utv.tv_sec += utv.tv_usec / 1000000; 1003 utv.tv_usec %= 1000000; 1004 if (utv.tv_usec < 0) { 1005 utv.tv_sec -= 1; 1006 utv.tv_usec += 1000000; 1007 } 1008 if (utv.tv_sec < 0) 1009 timerclear(&utv); 1010 if ((error = copyout(&utv, tvp, sizeof(utv)))) 1011 return error; 1012 SCARG(&bsa, tv) = tvp; 1013 } 1014 microtime(&tv0); 1015 } 1016 1017 error = sys_select(p, &bsa, retval); 1018 if (error) { 1019 /* 1020 * See fs/select.c in the Linux kernel. Without this, 1021 * Maelstrom doesn't work. 1022 */ 1023 if (error == ERESTART) 1024 error = EINTR; 1025 return error; 1026 } 1027 1028 if (timeout) { 1029 if (*retval) { 1030 /* 1031 * Compute how much time was left of the timeout, 1032 * by subtracting the current time and the time 1033 * before we started the call, and subtracting 1034 * that result from the user-supplied value. 1035 */ 1036 microtime(&tv1); 1037 timersub(&tv1, &tv0, &tv1); 1038 timersub(&utv, &tv1, &utv); 1039 if (utv.tv_sec < 0) 1040 timerclear(&utv); 1041 } else 1042 timerclear(&utv); 1043 if ((error = copyout(&utv, timeout, sizeof(utv)))) 1044 return error; 1045 } 1046 1047 return 0; 1048 } 1049 1050 /* 1051 * Get the process group of a certain process. Look it up 1052 * and return the value. 1053 */ 1054 int 1055 linux_sys_getpgid(p, v, retval) 1056 struct proc *p; 1057 void *v; 1058 register_t *retval; 1059 { 1060 struct linux_sys_getpgid_args /* { 1061 syscallarg(int) pid; 1062 } */ *uap = v; 1063 struct proc *targp; 1064 1065 if (SCARG(uap, pid) != 0 && SCARG(uap, pid) != p->p_pid) { 1066 if ((targp = pfind(SCARG(uap, pid))) == 0) 1067 return ESRCH; 1068 } 1069 else 1070 targp = p; 1071 1072 retval[0] = targp->p_pgid; 1073 return 0; 1074 } 1075 1076 /* 1077 * Set the 'personality' (emulation mode) for the current process. Only 1078 * accept the Linux personality here (0). This call is needed because 1079 * the Linux ELF crt0 issues it in an ugly kludge to make sure that 1080 * ELF binaries run in Linux mode, not SVR4 mode. 1081 */ 1082 int 1083 linux_sys_personality(p, v, retval) 1084 struct proc *p; 1085 void *v; 1086 register_t *retval; 1087 { 1088 struct linux_sys_personality_args /* { 1089 syscallarg(int) per; 1090 } */ *uap = v; 1091 1092 if (SCARG(uap, per) != 0) 1093 return EINVAL; 1094 retval[0] = 0; 1095 return 0; 1096 } 1097 1098 /* 1099 * The calls are here because of type conversions. 1100 */ 1101 int 1102 linux_sys_setreuid(p, v, retval) 1103 struct proc *p; 1104 void *v; 1105 register_t *retval; 1106 { 1107 struct linux_sys_setreuid_args /* { 1108 syscallarg(int) ruid; 1109 syscallarg(int) euid; 1110 } */ *uap = v; 1111 struct sys_setreuid_args bsa; 1112 1113 SCARG(&bsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ? 1114 (uid_t)-1 : SCARG(uap, ruid); 1115 SCARG(&bsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ? 1116 (uid_t)-1 : SCARG(uap, euid); 1117 1118 return sys_setreuid(p, &bsa, retval); 1119 } 1120 1121 int 1122 linux_sys_setregid(p, v, retval) 1123 struct proc *p; 1124 void *v; 1125 register_t *retval; 1126 { 1127 struct linux_sys_setregid_args /* { 1128 syscallarg(int) rgid; 1129 syscallarg(int) egid; 1130 } */ *uap = v; 1131 struct sys_setregid_args bsa; 1132 1133 SCARG(&bsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ? 1134 (uid_t)-1 : SCARG(uap, rgid); 1135 SCARG(&bsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ? 1136 (uid_t)-1 : SCARG(uap, egid); 1137 1138 return sys_setregid(p, &bsa, retval); 1139 } 1140 1141 int 1142 linux_sys_getsid(p, v, retval) 1143 struct proc *p; 1144 void *v; 1145 register_t *retval; 1146 { 1147 struct linux_sys_getsid_args /* { 1148 syscallarg(int) pid; 1149 } */ *uap = v; 1150 struct proc *p1; 1151 pid_t pid; 1152 1153 pid = (pid_t)SCARG(uap, pid); 1154 1155 if (pid == 0) { 1156 retval[0] = (int)p->p_session; /* XXX Oh well */ 1157 return 0; 1158 } 1159 1160 p1 = pfind((int)pid); 1161 if (p1 == NULL) 1162 return ESRCH; 1163 1164 retval[0] = (int)p1->p_session; 1165 return 0; 1166 } 1167 1168 int 1169 linux_sys___sysctl(p, v, retval) 1170 struct proc *p; 1171 void *v; 1172 register_t *retval; 1173 { 1174 struct linux_sys___sysctl_args /* { 1175 syscallarg(struct linux___sysctl *) lsp; 1176 } */ *uap = v; 1177 struct linux___sysctl ls; 1178 struct sys___sysctl_args bsa; 1179 int error; 1180 1181 if ((error = copyin(SCARG(uap, lsp), &ls, sizeof ls))) 1182 return error; 1183 SCARG(&bsa, name) = ls.name; 1184 SCARG(&bsa, namelen) = ls.namelen; 1185 SCARG(&bsa, old) = ls.old; 1186 SCARG(&bsa, oldlenp) = ls.oldlenp; 1187 SCARG(&bsa, new) = ls.new; 1188 SCARG(&bsa, newlen) = ls.newlen; 1189 1190 return sys___sysctl(p, &bsa, retval); 1191 } 1192 1193 int 1194 linux_sys_nice(p, v, retval) 1195 struct proc *p; 1196 void *v; 1197 register_t *retval; 1198 { 1199 struct linux_sys_nice_args /* { 1200 syscallarg(int) incr; 1201 } */ *uap = v; 1202 struct sys_setpriority_args bsa; 1203 1204 SCARG(&bsa, which) = PRIO_PROCESS; 1205 SCARG(&bsa, who) = 0; 1206 SCARG(&bsa, prio) = SCARG(uap, incr); 1207 return sys_setpriority(p, &bsa, retval); 1208 } 1209