linux_pipe.c revision 1.42
1/* $NetBSD: linux_pipe.c,v 1.42 1998/07/02 23:26:58 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 struct linux_sys_mremap_args /* { 523 syscallarg(void *) old_address; 524 syscallarg(size_t) old_size; 525 syscallarg(size_t) new_size; 526 syscallarg(u_long) flags; 527 } */ *uap = v; 528 struct sys_munmap_args mua; 529 size_t old_size, new_size; 530 int error; 531 532 old_size = round_page(SCARG(uap, old_size)); 533 new_size = round_page(SCARG(uap, new_size)); 534 535 /* 536 * Growing mapped region. 537 */ 538 if (new_size > old_size) { 539 /* 540 * XXX Implement me. What we probably want to do is 541 * XXX dig out the guts of the old mapping, mmap that 542 * XXX object again with the new size, then munmap 543 * XXX the old mapping. 544 */ 545 *retval = 0; 546 return (ENOMEM); 547 } 548 549 /* 550 * Shrinking mapped region. 551 */ 552 if (new_size < old_size) { 553 SCARG(&mua, addr) = (caddr_t)SCARG(uap, old_address) + 554 SCARG(uap, new_size); 555 SCARG(&mua, len) = old_size - new_size; 556 error = sys_munmap(p, &mua, retval); 557 *retval = error ? 0 : (register_t)SCARG(uap, old_address); 558 return (error); 559 } 560 561 /* 562 * No change. 563 */ 564 *retval = (register_t)SCARG(uap, old_address); 565 return (0); 566} 567 568int 569linux_sys_msync(p, v, retval) 570 struct proc *p; 571 void *v; 572 register_t *retval; 573{ 574 struct linux_sys_msync_args /* { 575 syscallarg(caddr_t) addr; 576 syscallarg(int) len; 577 syscallarg(int) fl; 578 } */ *uap = v; 579 580 struct sys___msync13_args bma; 581 582 /* flags are ignored */ 583 SCARG(&bma, addr) = SCARG(uap, addr); 584 SCARG(&bma, len) = SCARG(uap, len); 585 SCARG(&bma, flags) = SCARG(uap, fl); 586 587 return sys___msync13(p, &bma, retval); 588} 589 590/* 591 * This code is partly stolen from src/lib/libc/compat-43/times.c 592 * XXX - CLK_TCK isn't declared in /sys, just in <time.h>, done here 593 */ 594 595#define CLK_TCK 100 596#define CONVTCK(r) (r.tv_sec * CLK_TCK + r.tv_usec / (1000000 / CLK_TCK)) 597 598int 599linux_sys_times(p, v, retval) 600 struct proc *p; 601 void *v; 602 register_t *retval; 603{ 604 struct linux_sys_times_args /* { 605 syscallarg(struct times *) tms; 606 } */ *uap = v; 607 struct timeval t; 608 struct linux_tms ltms; 609 struct rusage ru; 610 int error, s; 611 612 calcru(p, &ru.ru_utime, &ru.ru_stime, NULL); 613 ltms.ltms_utime = CONVTCK(ru.ru_utime); 614 ltms.ltms_stime = CONVTCK(ru.ru_stime); 615 616 ltms.ltms_cutime = CONVTCK(p->p_stats->p_cru.ru_utime); 617 ltms.ltms_cstime = CONVTCK(p->p_stats->p_cru.ru_stime); 618 619 if ((error = copyout(<ms, SCARG(uap, tms), sizeof ltms))) 620 return error; 621 622 s = splclock(); 623 timersub(&time, &boottime, &t); 624 splx(s); 625 626 retval[0] = ((linux_clock_t)(CONVTCK(t))); 627 return 0; 628} 629 630/* 631 * NetBSD passes fd[0] in retval[0], and fd[1] in retval[1]. 632 * Linux directly passes the pointer. 633 */ 634int 635linux_sys_pipe(p, v, retval) 636 struct proc *p; 637 void *v; 638 register_t *retval; 639{ 640 struct linux_sys_pipe_args /* { 641 syscallarg(int *) pfds; 642 } */ *uap = v; 643 int error; 644 645 if ((error = sys_pipe(p, 0, retval))) 646 return error; 647 648 /* Assumes register_t is an int */ 649 650 if ((error = copyout(retval, SCARG(uap, pfds), 2 * sizeof (int)))) 651 return error; 652 653 retval[0] = 0; 654 return 0; 655} 656 657/* 658 * Alarm. This is a libc call which uses setitimer(2) in NetBSD. 659 * Fiddle with the timers to make it work. 660 */ 661int 662linux_sys_alarm(p, v, retval) 663 struct proc *p; 664 void *v; 665 register_t *retval; 666{ 667 struct linux_sys_alarm_args /* { 668 syscallarg(unsigned int) secs; 669 } */ *uap = v; 670 int s; 671 struct itimerval *itp, it; 672 673 itp = &p->p_realtimer; 674 s = splclock(); 675 /* 676 * Clear any pending timer alarms. 677 */ 678 untimeout(realitexpire, p); 679 timerclear(&itp->it_interval); 680 if (timerisset(&itp->it_value) && 681 timercmp(&itp->it_value, &time, >)) 682 timersub(&itp->it_value, &time, &itp->it_value); 683 /* 684 * Return how many seconds were left (rounded up) 685 */ 686 retval[0] = itp->it_value.tv_sec; 687 if (itp->it_value.tv_usec) 688 retval[0]++; 689 690 /* 691 * alarm(0) just resets the timer. 692 */ 693 if (SCARG(uap, secs) == 0) { 694 timerclear(&itp->it_value); 695 splx(s); 696 return 0; 697 } 698 699 /* 700 * Check the new alarm time for sanity, and set it. 701 */ 702 timerclear(&it.it_interval); 703 it.it_value.tv_sec = SCARG(uap, secs); 704 it.it_value.tv_usec = 0; 705 if (itimerfix(&it.it_value) || itimerfix(&it.it_interval)) { 706 splx(s); 707 return (EINVAL); 708 } 709 710 if (timerisset(&it.it_value)) { 711 timeradd(&it.it_value, &time, &it.it_value); 712 timeout(realitexpire, p, hzto(&it.it_value)); 713 } 714 p->p_realtimer = it; 715 splx(s); 716 717 return 0; 718} 719 720/* 721 * utime(). Do conversion to things that utimes() understands, 722 * and pass it on. 723 */ 724int 725linux_sys_utime(p, v, retval) 726 struct proc *p; 727 void *v; 728 register_t *retval; 729{ 730 struct linux_sys_utime_args /* { 731 syscallarg(char *) path; 732 syscallarg(struct linux_utimbuf *)times; 733 } */ *uap = v; 734 caddr_t sg; 735 int error; 736 struct sys_utimes_args ua; 737 struct timeval tv[2], *tvp; 738 struct linux_utimbuf lut; 739 740 sg = stackgap_init(p->p_emul); 741 LINUX_CHECK_ALT_EXIST(p, &sg, SCARG(uap, path)); 742 743 SCARG(&ua, path) = SCARG(uap, path); 744 745 if (SCARG(uap, times) != NULL) { 746 if ((error = copyin(SCARG(uap, times), &lut, sizeof lut))) 747 return error; 748 tv[0].tv_usec = tv[1].tv_usec = 0; 749 tv[0].tv_sec = lut.l_actime; 750 tv[1].tv_sec = lut.l_modtime; 751 tvp = (struct timeval *) stackgap_alloc(&sg, sizeof(tv)); 752 if ((error = copyout(tv, tvp, sizeof tv))) 753 return error; 754 SCARG(&ua, tptr) = tvp; 755 } 756 else 757 SCARG(&ua, tptr) = NULL; 758 759 return sys_utimes(p, &ua, retval); 760} 761 762/* 763 * The old Linux readdir was only able to read one entry at a time, 764 * even though it had a 'count' argument. In fact, the emulation 765 * of the old call was better than the original, because it did handle 766 * the count arg properly. Don't bother with it anymore now, and use 767 * it to distinguish between old and new. The difference is that the 768 * newer one actually does multiple entries, and the reclen field 769 * really is the reclen, not the namelength. 770 */ 771int 772linux_sys_readdir(p, v, retval) 773 struct proc *p; 774 void *v; 775 register_t *retval; 776{ 777 struct linux_sys_readdir_args /* { 778 syscallarg(int) fd; 779 syscallarg(struct linux_dirent *) dent; 780 syscallarg(unsigned int) count; 781 } */ *uap = v; 782 783 SCARG(uap, count) = 1; 784 return linux_sys_getdents(p, uap, retval); 785} 786 787/* 788 * Linux 'readdir' call. This code is mostly taken from the 789 * SunOS getdents call (see compat/sunos/sunos_misc.c), though 790 * an attempt has been made to keep it a little cleaner (failing 791 * miserably, because of the cruft needed if count 1 is passed). 792 * 793 * The d_off field should contain the offset of the next valid entry, 794 * but in Linux it has the offset of the entry itself. We emulate 795 * that bug here. 796 * 797 * Read in BSD-style entries, convert them, and copy them out. 798 * 799 * Note that this doesn't handle union-mounted filesystems. 800 */ 801int 802linux_sys_getdents(p, v, retval) 803 struct proc *p; 804 void *v; 805 register_t *retval; 806{ 807 struct linux_sys_readdir_args /* { 808 syscallarg(int) fd; 809 syscallarg(caddr_t) dent; 810 syscallarg(unsigned int) count; 811 } */ *uap = v; 812 register struct dirent *bdp; 813 struct vnode *vp; 814 caddr_t inp, buf; /* BSD-format */ 815 int len, reclen; /* BSD-format */ 816 caddr_t outp; /* Linux-format */ 817 int resid, linux_reclen = 0; /* Linux-format */ 818 struct file *fp; 819 struct uio auio; 820 struct iovec aiov; 821 struct linux_dirent idb; 822 off_t off; /* true file offset */ 823 int buflen, error, eofflag, nbytes, oldcall; 824 struct vattr va; 825 off_t *cookiebuf = NULL, *cookie; 826 int ncookies; 827 828 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 829 return (error); 830 831 if ((fp->f_flag & FREAD) == 0) 832 return (EBADF); 833 834 vp = (struct vnode *)fp->f_data; 835 836 if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p))) 837 return error; 838 839 nbytes = SCARG(uap, count); 840 if (nbytes == 1) { /* emulating old, broken behaviour */ 841 nbytes = sizeof (struct linux_dirent); 842 buflen = max(va.va_blocksize, nbytes); 843 oldcall = 1; 844 } else { 845 buflen = min(MAXBSIZE, nbytes); 846 if (buflen < va.va_blocksize) 847 buflen = va.va_blocksize; 848 oldcall = 0; 849 } 850 buf = malloc(buflen, M_TEMP, M_WAITOK); 851 852 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 853 off = fp->f_offset; 854again: 855 aiov.iov_base = buf; 856 aiov.iov_len = buflen; 857 auio.uio_iov = &aiov; 858 auio.uio_iovcnt = 1; 859 auio.uio_rw = UIO_READ; 860 auio.uio_segflg = UIO_SYSSPACE; 861 auio.uio_procp = p; 862 auio.uio_resid = buflen; 863 auio.uio_offset = off; 864 /* 865 * First we read into the malloc'ed buffer, then 866 * we massage it into user space, one record at a time. 867 */ 868 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf, 869 &ncookies); 870 if (error) 871 goto out; 872 873 inp = buf; 874 outp = SCARG(uap, dent); 875 resid = nbytes; 876 if ((len = buflen - auio.uio_resid) == 0) 877 goto eof; 878 879 for (cookie = cookiebuf; len > 0; len -= reclen) { 880 bdp = (struct dirent *)inp; 881 reclen = bdp->d_reclen; 882 if (reclen & 3) 883 panic("linux_readdir"); 884 if (bdp->d_fileno == 0) { 885 inp += reclen; /* it is a hole; squish it out */ 886 off = *cookie++; 887 continue; 888 } 889 linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen); 890 if (reclen > len || resid < linux_reclen) { 891 /* entry too big for buffer, so just stop */ 892 outp++; 893 break; 894 } 895 /* 896 * Massage in place to make a Linux-shaped dirent (otherwise 897 * we have to worry about touching user memory outside of 898 * the copyout() call). 899 */ 900 idb.d_ino = (linux_ino_t)bdp->d_fileno; 901 /* 902 * The old readdir() call misuses the offset and reclen fields. 903 */ 904 if (oldcall) { 905 idb.d_off = (linux_off_t)linux_reclen; 906 idb.d_reclen = (u_short)bdp->d_namlen; 907 } else { 908 if (sizeof (linux_off_t) < 4 && (off >> 32) != 0) { 909 compat_offseterr(vp, "linux_getdents"); 910 error = EINVAL; 911 goto out; 912 } 913 idb.d_off = (linux_off_t)off; 914 idb.d_reclen = (u_short)linux_reclen; 915 } 916 strcpy(idb.d_name, bdp->d_name); 917 if ((error = copyout((caddr_t)&idb, outp, linux_reclen))) 918 goto out; 919 /* advance past this real entry */ 920 inp += reclen; 921 off = *cookie++; /* each entry points to itself */ 922 /* advance output past Linux-shaped entry */ 923 outp += linux_reclen; 924 resid -= linux_reclen; 925 if (oldcall) 926 break; 927 } 928 929 /* if we squished out the whole block, try again */ 930 if (outp == SCARG(uap, dent)) 931 goto again; 932 fp->f_offset = off; /* update the vnode offset */ 933 934 if (oldcall) 935 nbytes = resid + linux_reclen; 936 937eof: 938 *retval = nbytes - resid; 939out: 940 VOP_UNLOCK(vp, 0); 941 if (cookiebuf) 942 free(cookiebuf, M_TEMP); 943 free(buf, M_TEMP); 944 return error; 945} 946 947/* 948 * Not sure why the arguments to this older version of select() were put 949 * into a structure, because there are 5, and that can all be handled 950 * in registers on the i386 like Linux wants to. 951 */ 952int 953linux_sys_oldselect(p, v, retval) 954 struct proc *p; 955 void *v; 956 register_t *retval; 957{ 958 struct linux_sys_oldselect_args /* { 959 syscallarg(struct linux_select *) lsp; 960 } */ *uap = v; 961 struct linux_select ls; 962 int error; 963 964 if ((error = copyin(SCARG(uap, lsp), &ls, sizeof(ls)))) 965 return error; 966 967 return linux_select1(p, retval, ls.nfds, ls.readfds, ls.writefds, 968 ls.exceptfds, ls.timeout); 969} 970 971/* 972 * Even when just using registers to pass arguments to syscalls you can 973 * have 5 of them on the i386. So this newer version of select() does 974 * this. 975 */ 976int 977linux_sys_select(p, v, retval) 978 struct proc *p; 979 void *v; 980 register_t *retval; 981{ 982 struct linux_sys_select_args /* { 983 syscallarg(int) nfds; 984 syscallarg(fd_set *) readfds; 985 syscallarg(fd_set *) writefds; 986 syscallarg(fd_set *) exceptfds; 987 syscallarg(struct timeval *) timeout; 988 } */ *uap = v; 989 990 return linux_select1(p, retval, SCARG(uap, nfds), SCARG(uap, readfds), 991 SCARG(uap, writefds), SCARG(uap, exceptfds), SCARG(uap, timeout)); 992} 993 994/* 995 * Common code for the old and new versions of select(). A couple of 996 * things are important: 997 * 1) return the amount of time left in the 'timeout' parameter 998 * 2) select never returns ERESTART on Linux, always return EINTR 999 */ 1000int 1001linux_select1(p, retval, nfds, readfds, writefds, exceptfds, timeout) 1002 struct proc *p; 1003 register_t *retval; 1004 int nfds; 1005 fd_set *readfds, *writefds, *exceptfds; 1006 struct timeval *timeout; 1007{ 1008 struct sys_select_args bsa; 1009 struct timeval tv0, tv1, utv, *tvp; 1010 caddr_t sg; 1011 int error; 1012 1013 SCARG(&bsa, nd) = nfds; 1014 SCARG(&bsa, in) = readfds; 1015 SCARG(&bsa, ou) = writefds; 1016 SCARG(&bsa, ex) = exceptfds; 1017 SCARG(&bsa, tv) = timeout; 1018 1019 /* 1020 * Store current time for computation of the amount of 1021 * time left. 1022 */ 1023 if (timeout) { 1024 if ((error = copyin(timeout, &utv, sizeof(utv)))) 1025 return error; 1026 if (itimerfix(&utv)) { 1027 /* 1028 * The timeval was invalid. Convert it to something 1029 * valid that will act as it does under Linux. 1030 */ 1031 sg = stackgap_init(p->p_emul); 1032 tvp = stackgap_alloc(&sg, sizeof(utv)); 1033 utv.tv_sec += utv.tv_usec / 1000000; 1034 utv.tv_usec %= 1000000; 1035 if (utv.tv_usec < 0) { 1036 utv.tv_sec -= 1; 1037 utv.tv_usec += 1000000; 1038 } 1039 if (utv.tv_sec < 0) 1040 timerclear(&utv); 1041 if ((error = copyout(&utv, tvp, sizeof(utv)))) 1042 return error; 1043 SCARG(&bsa, tv) = tvp; 1044 } 1045 microtime(&tv0); 1046 } 1047 1048 error = sys_select(p, &bsa, retval); 1049 if (error) { 1050 /* 1051 * See fs/select.c in the Linux kernel. Without this, 1052 * Maelstrom doesn't work. 1053 */ 1054 if (error == ERESTART) 1055 error = EINTR; 1056 return error; 1057 } 1058 1059 if (timeout) { 1060 if (*retval) { 1061 /* 1062 * Compute how much time was left of the timeout, 1063 * by subtracting the current time and the time 1064 * before we started the call, and subtracting 1065 * that result from the user-supplied value. 1066 */ 1067 microtime(&tv1); 1068 timersub(&tv1, &tv0, &tv1); 1069 timersub(&utv, &tv1, &utv); 1070 if (utv.tv_sec < 0) 1071 timerclear(&utv); 1072 } else 1073 timerclear(&utv); 1074 if ((error = copyout(&utv, timeout, sizeof(utv)))) 1075 return error; 1076 } 1077 1078 return 0; 1079} 1080 1081/* 1082 * Get the process group of a certain process. Look it up 1083 * and return the value. 1084 */ 1085int 1086linux_sys_getpgid(p, v, retval) 1087 struct proc *p; 1088 void *v; 1089 register_t *retval; 1090{ 1091 struct linux_sys_getpgid_args /* { 1092 syscallarg(int) pid; 1093 } */ *uap = v; 1094 struct proc *targp; 1095 1096 if (SCARG(uap, pid) != 0 && SCARG(uap, pid) != p->p_pid) { 1097 if ((targp = pfind(SCARG(uap, pid))) == 0) 1098 return ESRCH; 1099 } 1100 else 1101 targp = p; 1102 1103 retval[0] = targp->p_pgid; 1104 return 0; 1105} 1106 1107/* 1108 * Set the 'personality' (emulation mode) for the current process. Only 1109 * accept the Linux personality here (0). This call is needed because 1110 * the Linux ELF crt0 issues it in an ugly kludge to make sure that 1111 * ELF binaries run in Linux mode, not SVR4 mode. 1112 */ 1113int 1114linux_sys_personality(p, v, retval) 1115 struct proc *p; 1116 void *v; 1117 register_t *retval; 1118{ 1119 struct linux_sys_personality_args /* { 1120 syscallarg(int) per; 1121 } */ *uap = v; 1122 1123 if (SCARG(uap, per) != 0) 1124 return EINVAL; 1125 retval[0] = 0; 1126 return 0; 1127} 1128 1129/* 1130 * The calls are here because of type conversions. 1131 */ 1132int 1133linux_sys_setreuid(p, v, retval) 1134 struct proc *p; 1135 void *v; 1136 register_t *retval; 1137{ 1138 struct linux_sys_setreuid_args /* { 1139 syscallarg(int) ruid; 1140 syscallarg(int) euid; 1141 } */ *uap = v; 1142 struct sys_setreuid_args bsa; 1143 1144 SCARG(&bsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ? 1145 (uid_t)-1 : SCARG(uap, ruid); 1146 SCARG(&bsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ? 1147 (uid_t)-1 : SCARG(uap, euid); 1148 1149 return sys_setreuid(p, &bsa, retval); 1150} 1151 1152int 1153linux_sys_setregid(p, v, retval) 1154 struct proc *p; 1155 void *v; 1156 register_t *retval; 1157{ 1158 struct linux_sys_setregid_args /* { 1159 syscallarg(int) rgid; 1160 syscallarg(int) egid; 1161 } */ *uap = v; 1162 struct sys_setregid_args bsa; 1163 1164 SCARG(&bsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ? 1165 (uid_t)-1 : SCARG(uap, rgid); 1166 SCARG(&bsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ? 1167 (uid_t)-1 : SCARG(uap, egid); 1168 1169 return sys_setregid(p, &bsa, retval); 1170} 1171 1172int 1173linux_sys___sysctl(p, v, retval) 1174 struct proc *p; 1175 void *v; 1176 register_t *retval; 1177{ 1178 struct linux_sys___sysctl_args /* { 1179 syscallarg(struct linux___sysctl *) lsp; 1180 } */ *uap = v; 1181 struct linux___sysctl ls; 1182 struct sys___sysctl_args bsa; 1183 int error; 1184 1185 if ((error = copyin(SCARG(uap, lsp), &ls, sizeof ls))) 1186 return error; 1187 SCARG(&bsa, name) = ls.name; 1188 SCARG(&bsa, namelen) = ls.namelen; 1189 SCARG(&bsa, old) = ls.old; 1190 SCARG(&bsa, oldlenp) = ls.oldlenp; 1191 SCARG(&bsa, new) = ls.new; 1192 SCARG(&bsa, newlen) = ls.newlen; 1193 1194 return sys___sysctl(p, &bsa, retval); 1195} 1196 1197int 1198linux_sys_nice(p, v, retval) 1199 struct proc *p; 1200 void *v; 1201 register_t *retval; 1202{ 1203 struct linux_sys_nice_args /* { 1204 syscallarg(int) incr; 1205 } */ *uap = v; 1206 struct sys_setpriority_args bsa; 1207 1208 SCARG(&bsa, which) = PRIO_PROCESS; 1209 SCARG(&bsa, who) = 0; 1210 SCARG(&bsa, prio) = SCARG(uap, incr); 1211 return sys_setpriority(p, &bsa, retval); 1212} 1213