linux_pipe.c revision 1.37
1/* $NetBSD: linux_pipe.c,v 1.37 1998/02/14 01:28:15 thorpej 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 */ 80static void bsd_to_linux_wstat __P((int *)); 81static void bsd_to_linux_statfs __P((struct statfs *, struct linux_statfs *)); 82int 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 */ 90static void 91bsd_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 */ 108int 109linux_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 */ 153int 154linux_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 */ 200int 201linux_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 */ 219int 220linux_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 */ 254int 255linux_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 */ 283static void 284bsd_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 */ 304int 305linux_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 339int 340linux_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 */ 378int 379linux_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 412int 413linux_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 444int 445linux_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 */ 481int 482linux_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 516int 517linux_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 534int 535linux_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___msync13_args bma; 547 548 /* flags are ignored */ 549 SCARG(&bma, addr) = SCARG(uap, addr); 550 SCARG(&bma, len) = SCARG(uap, len); 551 SCARG(&bma, flags) = SCARG(uap, fl); 552 553 return sys___msync13(p, &bma, retval); 554} 555 556/* 557 * This code is partly stolen from src/lib/libc/compat-43/times.c 558 * XXX - CLK_TCK isn't declared in /sys, just in <time.h>, done here 559 */ 560 561#define CLK_TCK 100 562#define CONVTCK(r) (r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK)) 563 564int 565linux_sys_times(p, v, retval) 566 struct proc *p; 567 void *v; 568 register_t *retval; 569{ 570 struct linux_sys_times_args /* { 571 syscallarg(struct times *) tms; 572 } */ *uap = v; 573 struct timeval t; 574 struct linux_tms ltms; 575 struct rusage ru; 576 int error, s; 577 578 calcru(p, &ru.ru_utime, &ru.ru_stime, NULL); 579 ltms.ltms_utime = CONVTCK(ru.ru_utime); 580 ltms.ltms_stime = CONVTCK(ru.ru_stime); 581 582 ltms.ltms_cutime = CONVTCK(p->p_stats->p_cru.ru_utime); 583 ltms.ltms_cstime = CONVTCK(p->p_stats->p_cru.ru_stime); 584 585 if ((error = copyout(<ms, SCARG(uap, tms), sizeof ltms))) 586 return error; 587 588 s = splclock(); 589 timersub(&time, &boottime, &t); 590 splx(s); 591 592 retval[0] = ((linux_clock_t)(CONVTCK(t))); 593 return 0; 594} 595 596/* 597 * NetBSD passes fd[0] in retval[0], and fd[1] in retval[1]. 598 * Linux directly passes the pointer. 599 */ 600int 601linux_sys_pipe(p, v, retval) 602 struct proc *p; 603 void *v; 604 register_t *retval; 605{ 606 struct linux_sys_pipe_args /* { 607 syscallarg(int *) pfds; 608 } */ *uap = v; 609 int error; 610 611 if ((error = sys_pipe(p, 0, retval))) 612 return error; 613 614 /* Assumes register_t is an int */ 615 616 if ((error = copyout(retval, SCARG(uap, pfds), 2 * sizeof (int)))) 617 return error; 618 619 retval[0] = 0; 620 return 0; 621} 622 623/* 624 * Alarm. This is a libc call which uses setitimer(2) in NetBSD. 625 * Fiddle with the timers to make it work. 626 */ 627int 628linux_sys_alarm(p, v, retval) 629 struct proc *p; 630 void *v; 631 register_t *retval; 632{ 633 struct linux_sys_alarm_args /* { 634 syscallarg(unsigned int) secs; 635 } */ *uap = v; 636 int s; 637 struct itimerval *itp, it; 638 639 itp = &p->p_realtimer; 640 s = splclock(); 641 /* 642 * Clear any pending timer alarms. 643 */ 644 untimeout(realitexpire, p); 645 timerclear(&itp->it_interval); 646 if (timerisset(&itp->it_value) && 647 timercmp(&itp->it_value, &time, >)) 648 timersub(&itp->it_value, &time, &itp->it_value); 649 /* 650 * Return how many seconds were left (rounded up) 651 */ 652 retval[0] = itp->it_value.tv_sec; 653 if (itp->it_value.tv_usec) 654 retval[0]++; 655 656 /* 657 * alarm(0) just resets the timer. 658 */ 659 if (SCARG(uap, secs) == 0) { 660 timerclear(&itp->it_value); 661 splx(s); 662 return 0; 663 } 664 665 /* 666 * Check the new alarm time for sanity, and set it. 667 */ 668 timerclear(&it.it_interval); 669 it.it_value.tv_sec = SCARG(uap, secs); 670 it.it_value.tv_usec = 0; 671 if (itimerfix(&it.it_value) || itimerfix(&it.it_interval)) { 672 splx(s); 673 return (EINVAL); 674 } 675 676 if (timerisset(&it.it_value)) { 677 timeradd(&it.it_value, &time, &it.it_value); 678 timeout(realitexpire, p, hzto(&it.it_value)); 679 } 680 p->p_realtimer = it; 681 splx(s); 682 683 return 0; 684} 685 686/* 687 * utime(). Do conversion to things that utimes() understands, 688 * and pass it on. 689 */ 690int 691linux_sys_utime(p, v, retval) 692 struct proc *p; 693 void *v; 694 register_t *retval; 695{ 696 struct linux_sys_utime_args /* { 697 syscallarg(char *) path; 698 syscallarg(struct linux_utimbuf *)times; 699 } */ *uap = v; 700 caddr_t sg; 701 int error; 702 struct sys_utimes_args ua; 703 struct timeval tv[2], *tvp; 704 struct linux_utimbuf lut; 705 706 sg = stackgap_init(p->p_emul); 707 LINUX_CHECK_ALT_EXIST(p, &sg, SCARG(uap, path)); 708 709 SCARG(&ua, path) = SCARG(uap, path); 710 711 if (SCARG(uap, times) != NULL) { 712 if ((error = copyin(SCARG(uap, times), &lut, sizeof lut))) 713 return error; 714 tv[0].tv_usec = tv[1].tv_usec = 0; 715 tv[0].tv_sec = lut.l_actime; 716 tv[1].tv_sec = lut.l_modtime; 717 tvp = (struct timeval *) stackgap_alloc(&sg, sizeof(tv)); 718 if ((error = copyout(tv, tvp, sizeof tv))) 719 return error; 720 SCARG(&ua, tptr) = tvp; 721 } 722 else 723 SCARG(&ua, tptr) = NULL; 724 725 return sys_utimes(p, uap, retval); 726} 727 728/* 729 * The old Linux readdir was only able to read one entry at a time, 730 * even though it had a 'count' argument. In fact, the emulation 731 * of the old call was better than the original, because it did handle 732 * the count arg properly. Don't bother with it anymore now, and use 733 * it to distinguish between old and new. The difference is that the 734 * newer one actually does multiple entries, and the reclen field 735 * really is the reclen, not the namelength. 736 */ 737int 738linux_sys_readdir(p, v, retval) 739 struct proc *p; 740 void *v; 741 register_t *retval; 742{ 743 struct linux_sys_readdir_args /* { 744 syscallarg(int) fd; 745 syscallarg(struct linux_dirent *) dent; 746 syscallarg(unsigned int) count; 747 } */ *uap = v; 748 749 SCARG(uap, count) = 1; 750 return linux_sys_getdents(p, uap, retval); 751} 752 753/* 754 * Linux 'readdir' call. This code is mostly taken from the 755 * SunOS getdents call (see compat/sunos/sunos_misc.c), though 756 * an attempt has been made to keep it a little cleaner (failing 757 * miserably, because of the cruft needed if count 1 is passed). 758 * 759 * The d_off field should contain the offset of the next valid entry, 760 * but in Linux it has the offset of the entry itself. We emulate 761 * that bug here. 762 * 763 * Read in BSD-style entries, convert them, and copy them out. 764 * 765 * Note that this doesn't handle union-mounted filesystems. 766 */ 767int 768linux_sys_getdents(p, v, retval) 769 struct proc *p; 770 void *v; 771 register_t *retval; 772{ 773 struct linux_sys_readdir_args /* { 774 syscallarg(int) fd; 775 syscallarg(caddr_t) dent; 776 syscallarg(unsigned int) count; 777 } */ *uap = v; 778 register struct dirent *bdp; 779 struct vnode *vp; 780 caddr_t inp, buf; /* BSD-format */ 781 int len, reclen; /* BSD-format */ 782 caddr_t outp; /* Linux-format */ 783 int resid, linux_reclen = 0; /* Linux-format */ 784 struct file *fp; 785 struct uio auio; 786 struct iovec aiov; 787 struct linux_dirent idb; 788 off_t off; /* true file offset */ 789 int buflen, error, eofflag, nbytes, oldcall; 790 struct vattr va; 791 off_t *cookiebuf, *cookie; 792 int ncookies; 793 794 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 795 return (error); 796 797 if ((fp->f_flag & FREAD) == 0) 798 return (EBADF); 799 800 vp = (struct vnode *)fp->f_data; 801 802 if (vp->v_type != VDIR) /* XXX vnode readdir op should do this */ 803 return (EINVAL); 804 805 if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p))) 806 return error; 807 808 nbytes = SCARG(uap, count); 809 if (nbytes == 1) { /* emulating old, broken behaviour */ 810 nbytes = sizeof (struct linux_dirent); 811 buflen = max(va.va_blocksize, nbytes); 812 oldcall = 1; 813 } else { 814 buflen = min(MAXBSIZE, nbytes); 815 if (buflen < va.va_blocksize) 816 buflen = va.va_blocksize; 817 oldcall = 0; 818 } 819 buf = malloc(buflen, M_TEMP, M_WAITOK); 820 ncookies = buflen / 16; 821 cookiebuf = malloc(ncookies * sizeof(*cookiebuf), M_TEMP, M_WAITOK); 822 823 VOP_LOCK(vp); 824 off = fp->f_offset; 825again: 826 aiov.iov_base = buf; 827 aiov.iov_len = buflen; 828 auio.uio_iov = &aiov; 829 auio.uio_iovcnt = 1; 830 auio.uio_rw = UIO_READ; 831 auio.uio_segflg = UIO_SYSSPACE; 832 auio.uio_procp = p; 833 auio.uio_resid = buflen; 834 auio.uio_offset = off; 835 /* 836 * First we read into the malloc'ed buffer, then 837 * we massage it into user space, one record at a time. 838 */ 839 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, cookiebuf, 840 ncookies); 841 if (error) 842 goto out; 843 844 inp = buf; 845 outp = SCARG(uap, dent); 846 resid = nbytes; 847 if ((len = buflen - auio.uio_resid) == 0) 848 goto eof; 849 850 for (cookie = cookiebuf; len > 0; len -= reclen) { 851 bdp = (struct dirent *)inp; 852 reclen = bdp->d_reclen; 853 if (reclen & 3) 854 panic("linux_readdir"); 855 if (bdp->d_fileno == 0) { 856 inp += reclen; /* it is a hole; squish it out */ 857 off = *cookie++; 858 continue; 859 } 860 linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen); 861 if (reclen > len || resid < linux_reclen) { 862 /* entry too big for buffer, so just stop */ 863 outp++; 864 break; 865 } 866 /* 867 * Massage in place to make a Linux-shaped dirent (otherwise 868 * we have to worry about touching user memory outside of 869 * the copyout() call). 870 */ 871 idb.d_ino = (linux_ino_t)bdp->d_fileno; 872 /* 873 * The old readdir() call misuses the offset and reclen fields. 874 */ 875 if (oldcall) { 876 idb.d_off = (linux_off_t)linux_reclen; 877 idb.d_reclen = (u_short)bdp->d_namlen; 878 } else { 879 if (sizeof (linux_off_t) < 4 && (off >> 32) != 0) { 880 compat_offseterr(vp, "linux_getdents"); 881 error = EINVAL; 882 goto out; 883 } 884 idb.d_off = (linux_off_t)off; 885 idb.d_reclen = (u_short)linux_reclen; 886 } 887 strcpy(idb.d_name, bdp->d_name); 888 if ((error = copyout((caddr_t)&idb, outp, linux_reclen))) 889 goto out; 890 /* advance past this real entry */ 891 inp += reclen; 892 off = *cookie++; /* each entry points to itself */ 893 /* advance output past Linux-shaped entry */ 894 outp += linux_reclen; 895 resid -= linux_reclen; 896 if (oldcall) 897 break; 898 } 899 900 /* if we squished out the whole block, try again */ 901 if (outp == SCARG(uap, dent)) 902 goto again; 903 fp->f_offset = off; /* update the vnode offset */ 904 905 if (oldcall) 906 nbytes = resid + linux_reclen; 907 908eof: 909 *retval = nbytes - resid; 910out: 911 VOP_UNLOCK(vp); 912 free(cookiebuf, M_TEMP); 913 free(buf, M_TEMP); 914 return error; 915} 916 917/* 918 * Not sure why the arguments to this older version of select() were put 919 * into a structure, because there are 5, and that can all be handled 920 * in registers on the i386 like Linux wants to. 921 */ 922int 923linux_sys_oldselect(p, v, retval) 924 struct proc *p; 925 void *v; 926 register_t *retval; 927{ 928 struct linux_sys_oldselect_args /* { 929 syscallarg(struct linux_select *) lsp; 930 } */ *uap = v; 931 struct linux_select ls; 932 int error; 933 934 if ((error = copyin(SCARG(uap, lsp), &ls, sizeof(ls)))) 935 return error; 936 937 return linux_select1(p, retval, ls.nfds, ls.readfds, ls.writefds, 938 ls.exceptfds, ls.timeout); 939} 940 941/* 942 * Even when just using registers to pass arguments to syscalls you can 943 * have 5 of them on the i386. So this newer version of select() does 944 * this. 945 */ 946int 947linux_sys_select(p, v, retval) 948 struct proc *p; 949 void *v; 950 register_t *retval; 951{ 952 struct linux_sys_select_args /* { 953 syscallarg(int) nfds; 954 syscallarg(fd_set *) readfds; 955 syscallarg(fd_set *) writefds; 956 syscallarg(fd_set *) exceptfds; 957 syscallarg(struct timeval *) timeout; 958 } */ *uap = v; 959 960 return linux_select1(p, retval, SCARG(uap, nfds), SCARG(uap, readfds), 961 SCARG(uap, writefds), SCARG(uap, exceptfds), SCARG(uap, timeout)); 962} 963 964/* 965 * Common code for the old and new versions of select(). A couple of 966 * things are important: 967 * 1) return the amount of time left in the 'timeout' parameter 968 * 2) select never returns ERESTART on Linux, always return EINTR 969 */ 970int 971linux_select1(p, retval, nfds, readfds, writefds, exceptfds, timeout) 972 struct proc *p; 973 register_t *retval; 974 int nfds; 975 fd_set *readfds, *writefds, *exceptfds; 976 struct timeval *timeout; 977{ 978 struct sys_select_args bsa; 979 struct timeval tv0, tv1, utv, *tvp; 980 caddr_t sg; 981 int error; 982 983 SCARG(&bsa, nd) = nfds; 984 SCARG(&bsa, in) = readfds; 985 SCARG(&bsa, ou) = writefds; 986 SCARG(&bsa, ex) = exceptfds; 987 SCARG(&bsa, tv) = timeout; 988 989 /* 990 * Store current time for computation of the amount of 991 * time left. 992 */ 993 if (timeout) { 994 if ((error = copyin(timeout, &utv, sizeof(utv)))) 995 return error; 996 if (itimerfix(&utv)) { 997 /* 998 * The timeval was invalid. Convert it to something 999 * valid that will act as it does under Linux. 1000 */ 1001 sg = stackgap_init(p->p_emul); 1002 tvp = stackgap_alloc(&sg, sizeof(utv)); 1003 utv.tv_sec += utv.tv_usec / 1000000; 1004 utv.tv_usec %= 1000000; 1005 if (utv.tv_usec < 0) { 1006 utv.tv_sec -= 1; 1007 utv.tv_usec += 1000000; 1008 } 1009 if (utv.tv_sec < 0) 1010 timerclear(&utv); 1011 if ((error = copyout(&utv, tvp, sizeof(utv)))) 1012 return error; 1013 SCARG(&bsa, tv) = tvp; 1014 } 1015 microtime(&tv0); 1016 } 1017 1018 error = sys_select(p, &bsa, retval); 1019 if (error) { 1020 /* 1021 * See fs/select.c in the Linux kernel. Without this, 1022 * Maelstrom doesn't work. 1023 */ 1024 if (error == ERESTART) 1025 error = EINTR; 1026 return error; 1027 } 1028 1029 if (timeout) { 1030 if (*retval) { 1031 /* 1032 * Compute how much time was left of the timeout, 1033 * by subtracting the current time and the time 1034 * before we started the call, and subtracting 1035 * that result from the user-supplied value. 1036 */ 1037 microtime(&tv1); 1038 timersub(&tv1, &tv0, &tv1); 1039 timersub(&utv, &tv1, &utv); 1040 if (utv.tv_sec < 0) 1041 timerclear(&utv); 1042 } else 1043 timerclear(&utv); 1044 if ((error = copyout(&utv, timeout, sizeof(utv)))) 1045 return error; 1046 } 1047 1048 return 0; 1049} 1050 1051/* 1052 * Get the process group of a certain process. Look it up 1053 * and return the value. 1054 */ 1055int 1056linux_sys_getpgid(p, v, retval) 1057 struct proc *p; 1058 void *v; 1059 register_t *retval; 1060{ 1061 struct linux_sys_getpgid_args /* { 1062 syscallarg(int) pid; 1063 } */ *uap = v; 1064 struct proc *targp; 1065 1066 if (SCARG(uap, pid) != 0 && SCARG(uap, pid) != p->p_pid) { 1067 if ((targp = pfind(SCARG(uap, pid))) == 0) 1068 return ESRCH; 1069 } 1070 else 1071 targp = p; 1072 1073 retval[0] = targp->p_pgid; 1074 return 0; 1075} 1076 1077/* 1078 * Set the 'personality' (emulation mode) for the current process. Only 1079 * accept the Linux personality here (0). This call is needed because 1080 * the Linux ELF crt0 issues it in an ugly kludge to make sure that 1081 * ELF binaries run in Linux mode, not SVR4 mode. 1082 */ 1083int 1084linux_sys_personality(p, v, retval) 1085 struct proc *p; 1086 void *v; 1087 register_t *retval; 1088{ 1089 struct linux_sys_personality_args /* { 1090 syscallarg(int) per; 1091 } */ *uap = v; 1092 1093 if (SCARG(uap, per) != 0) 1094 return EINVAL; 1095 retval[0] = 0; 1096 return 0; 1097} 1098 1099/* 1100 * The calls are here because of type conversions. 1101 */ 1102int 1103linux_sys_setreuid(p, v, retval) 1104 struct proc *p; 1105 void *v; 1106 register_t *retval; 1107{ 1108 struct linux_sys_setreuid_args /* { 1109 syscallarg(int) ruid; 1110 syscallarg(int) euid; 1111 } */ *uap = v; 1112 struct sys_setreuid_args bsa; 1113 1114 SCARG(&bsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ? 1115 (uid_t)-1 : SCARG(uap, ruid); 1116 SCARG(&bsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ? 1117 (uid_t)-1 : SCARG(uap, euid); 1118 1119 return sys_setreuid(p, &bsa, retval); 1120} 1121 1122int 1123linux_sys_setregid(p, v, retval) 1124 struct proc *p; 1125 void *v; 1126 register_t *retval; 1127{ 1128 struct linux_sys_setregid_args /* { 1129 syscallarg(int) rgid; 1130 syscallarg(int) egid; 1131 } */ *uap = v; 1132 struct sys_setregid_args bsa; 1133 1134 SCARG(&bsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ? 1135 (uid_t)-1 : SCARG(uap, rgid); 1136 SCARG(&bsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ? 1137 (uid_t)-1 : SCARG(uap, egid); 1138 1139 return sys_setregid(p, &bsa, retval); 1140} 1141 1142int 1143linux_sys_getsid(p, v, retval) 1144 struct proc *p; 1145 void *v; 1146 register_t *retval; 1147{ 1148 struct linux_sys_getsid_args /* { 1149 syscallarg(int) pid; 1150 } */ *uap = v; 1151 struct proc *p1; 1152 pid_t pid; 1153 1154 /* 1155 * NOTE: The Linux getsid(2) is different from the XPG getsid(2), 1156 * which is defined to return the process group ID of the session 1157 * leader. Insetead, Linux returns the pointer to the session. 1158 */ 1159 1160 pid = (pid_t)SCARG(uap, pid); 1161 1162 if (pid == 0) { 1163 retval[0] = (register_t)p->p_session; 1164 return 0; 1165 } 1166 1167 p1 = pfind(pid); 1168 if (p1 == NULL) 1169 return ESRCH; 1170 1171 retval[0] = (register_t)p1->p_session; 1172 return 0; 1173} 1174 1175int 1176linux_sys___sysctl(p, v, retval) 1177 struct proc *p; 1178 void *v; 1179 register_t *retval; 1180{ 1181 struct linux_sys___sysctl_args /* { 1182 syscallarg(struct linux___sysctl *) lsp; 1183 } */ *uap = v; 1184 struct linux___sysctl ls; 1185 struct sys___sysctl_args bsa; 1186 int error; 1187 1188 if ((error = copyin(SCARG(uap, lsp), &ls, sizeof ls))) 1189 return error; 1190 SCARG(&bsa, name) = ls.name; 1191 SCARG(&bsa, namelen) = ls.namelen; 1192 SCARG(&bsa, old) = ls.old; 1193 SCARG(&bsa, oldlenp) = ls.oldlenp; 1194 SCARG(&bsa, new) = ls.new; 1195 SCARG(&bsa, newlen) = ls.newlen; 1196 1197 return sys___sysctl(p, &bsa, retval); 1198} 1199 1200int 1201linux_sys_nice(p, v, retval) 1202 struct proc *p; 1203 void *v; 1204 register_t *retval; 1205{ 1206 struct linux_sys_nice_args /* { 1207 syscallarg(int) incr; 1208 } */ *uap = v; 1209 struct sys_setpriority_args bsa; 1210 1211 SCARG(&bsa, which) = PRIO_PROCESS; 1212 SCARG(&bsa, who) = 0; 1213 SCARG(&bsa, prio) = SCARG(uap, incr); 1214 return sys_setpriority(p, &bsa, retval); 1215} 1216