linux_misc.c revision 1.83.2.14
1 /* $NetBSD: linux_misc.c,v 1.83.2.14 2002/05/29 21:32:40 nathanw Exp $ */ 2 3 /*- 4 * Copyright (c) 1995, 1998, 1999 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Frank van der Linden and Eric Haszlakiewicz; by Jason R. Thorpe 9 * of the Numerical Aerospace Simulation Facility, NASA Ames Research Center. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 14 * 1. Redistributions of source code must retain the above copyright 15 * notice, this list of conditions and the following disclaimer. 16 * 2. Redistributions in binary form must reproduce the above copyright 17 * notice, this list of conditions and the following disclaimer in the 18 * documentation and/or other materials provided with the distribution. 19 * 3. All advertising materials mentioning features or use of this software 20 * must display the following acknowledgement: 21 * This product includes software developed by the NetBSD 22 * Foundation, Inc. and its contributors. 23 * 4. Neither the name of The NetBSD Foundation nor the names of its 24 * contributors may be used to endorse or promote products derived 25 * from this software without specific prior written permission. 26 * 27 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 28 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 29 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 30 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 31 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 32 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 33 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 34 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 35 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 36 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 37 * POSSIBILITY OF SUCH DAMAGE. 38 */ 39 40 /* 41 * Linux compatibility module. Try to deal with various Linux system calls. 42 */ 43 44 /* 45 * These functions have been moved to multiarch to allow 46 * selection of which machines include them to be 47 * determined by the individual files.linux_<arch> files. 48 * 49 * Function in multiarch: 50 * linux_sys_break : linux_break.c 51 * linux_sys_alarm : linux_misc_notalpha.c 52 * linux_sys_getresgid : linux_misc_notalpha.c 53 * linux_sys_nice : linux_misc_notalpha.c 54 * linux_sys_readdir : linux_misc_notalpha.c 55 * linux_sys_setresgid : linux_misc_notalpha.c 56 * linux_sys_time : linux_misc_notalpha.c 57 * linux_sys_utime : linux_misc_notalpha.c 58 * linux_sys_waitpid : linux_misc_notalpha.c 59 * linux_sys_old_mmap : linux_oldmmap.c 60 * linux_sys_oldolduname : linux_oldolduname.c 61 * linux_sys_oldselect : linux_oldselect.c 62 * linux_sys_olduname : linux_olduname.c 63 * linux_sys_pipe : linux_pipe.c 64 */ 65 66 #include <sys/cdefs.h> 67 __KERNEL_RCSID(0, "$NetBSD: linux_misc.c,v 1.83.2.14 2002/05/29 21:32:40 nathanw Exp $"); 68 69 #include <sys/param.h> 70 #include <sys/systm.h> 71 #include <sys/namei.h> 72 #include <sys/lwp.h> 73 #include <sys/proc.h> 74 #include <sys/dirent.h> 75 #include <sys/file.h> 76 #include <sys/stat.h> 77 #include <sys/filedesc.h> 78 #include <sys/ioctl.h> 79 #include <sys/kernel.h> 80 #include <sys/malloc.h> 81 #include <sys/mbuf.h> 82 #include <sys/mman.h> 83 #include <sys/mount.h> 84 #include <sys/reboot.h> 85 #include <sys/resource.h> 86 #include <sys/resourcevar.h> 87 #include <sys/signal.h> 88 #include <sys/signalvar.h> 89 #include <sys/socket.h> 90 #include <sys/time.h> 91 #include <sys/times.h> 92 #include <sys/vnode.h> 93 #include <sys/uio.h> 94 #include <sys/wait.h> 95 #include <sys/utsname.h> 96 #include <sys/unistd.h> 97 #include <sys/swap.h> /* for SWAP_ON */ 98 #include <sys/sysctl.h> /* for KERN_DOMAINNAME */ 99 100 #include <sys/ptrace.h> 101 #include <machine/ptrace.h> 102 103 #include <sys/sa.h> 104 #include <sys/syscallargs.h> 105 106 #include <compat/linux/common/linux_types.h> 107 #include <compat/linux/common/linux_signal.h> 108 109 #include <compat/linux/linux_syscallargs.h> 110 111 #include <compat/linux/common/linux_fcntl.h> 112 #include <compat/linux/common/linux_mmap.h> 113 #include <compat/linux/common/linux_dirent.h> 114 #include <compat/linux/common/linux_util.h> 115 #include <compat/linux/common/linux_misc.h> 116 #include <compat/linux/common/linux_ptrace.h> 117 #include <compat/linux/common/linux_reboot.h> 118 #include <compat/linux/common/linux_emuldata.h> 119 120 const int linux_ptrace_request_map[] = { 121 LINUX_PTRACE_TRACEME, PT_TRACE_ME, 122 LINUX_PTRACE_PEEKTEXT, PT_READ_I, 123 LINUX_PTRACE_PEEKDATA, PT_READ_D, 124 LINUX_PTRACE_POKETEXT, PT_WRITE_I, 125 LINUX_PTRACE_POKEDATA, PT_WRITE_D, 126 LINUX_PTRACE_CONT, PT_CONTINUE, 127 LINUX_PTRACE_KILL, PT_KILL, 128 LINUX_PTRACE_ATTACH, PT_ATTACH, 129 LINUX_PTRACE_DETACH, PT_DETACH, 130 #ifdef PT_STEP 131 LINUX_PTRACE_SINGLESTEP, PT_STEP, 132 #endif 133 -1 134 }; 135 136 const static struct mnttypes { 137 char *bsd; 138 int linux; 139 } fstypes[] = { 140 { MOUNT_FFS, LINUX_DEFAULT_SUPER_MAGIC }, 141 { MOUNT_NFS, LINUX_NFS_SUPER_MAGIC }, 142 { MOUNT_MFS, LINUX_DEFAULT_SUPER_MAGIC }, 143 { MOUNT_MSDOS, LINUX_MSDOS_SUPER_MAGIC }, 144 { MOUNT_LFS, LINUX_DEFAULT_SUPER_MAGIC }, 145 { MOUNT_FDESC, LINUX_DEFAULT_SUPER_MAGIC }, 146 { MOUNT_PORTAL, LINUX_DEFAULT_SUPER_MAGIC }, 147 { MOUNT_NULL, LINUX_DEFAULT_SUPER_MAGIC }, 148 { MOUNT_OVERLAY, LINUX_DEFAULT_SUPER_MAGIC }, 149 { MOUNT_UMAP, LINUX_DEFAULT_SUPER_MAGIC }, 150 { MOUNT_KERNFS, LINUX_DEFAULT_SUPER_MAGIC }, 151 { MOUNT_PROCFS, LINUX_PROC_SUPER_MAGIC }, 152 { MOUNT_AFS, LINUX_DEFAULT_SUPER_MAGIC }, 153 { MOUNT_CD9660, LINUX_ISOFS_SUPER_MAGIC }, 154 { MOUNT_UNION, LINUX_DEFAULT_SUPER_MAGIC }, 155 { MOUNT_ADOSFS, LINUX_ADFS_SUPER_MAGIC }, 156 { MOUNT_EXT2FS, LINUX_EXT2_SUPER_MAGIC }, 157 { MOUNT_CFS, LINUX_DEFAULT_SUPER_MAGIC }, 158 { MOUNT_CODA, LINUX_CODA_SUPER_MAGIC }, 159 { MOUNT_FILECORE, LINUX_DEFAULT_SUPER_MAGIC }, 160 { MOUNT_NTFS, LINUX_DEFAULT_SUPER_MAGIC }, 161 { MOUNT_SMBFS, LINUX_SMB_SUPER_MAGIC } 162 }; 163 #define FSTYPESSIZE (sizeof(fstypes) / sizeof(fstypes[0])) 164 165 #ifdef DEBUG_LINUX 166 #define DPRINTF(a) uprintf a 167 #else 168 #define DPRINTF(a) 169 #endif 170 171 /* Local linux_misc.c functions: */ 172 static void bsd_to_linux_statfs __P((struct statfs *, struct linux_statfs *)); 173 static int linux_to_bsd_limit __P((int)); 174 175 /* 176 * The information on a terminated (or stopped) process needs 177 * to be converted in order for Linux binaries to get a valid signal 178 * number out of it. 179 */ 180 void 181 bsd_to_linux_wstat(st) 182 int *st; 183 { 184 185 int sig; 186 187 if (WIFSIGNALED(*st)) { 188 sig = WTERMSIG(*st); 189 if (sig >= 0 && sig < NSIG) 190 *st= (*st& ~0177) | native_to_linux_signo[sig]; 191 } else if (WIFSTOPPED(*st)) { 192 sig = WSTOPSIG(*st); 193 if (sig >= 0 && sig < NSIG) 194 *st = (*st & ~0xff00) | 195 (native_to_linux_signo[sig] << 8); 196 } 197 } 198 199 /* 200 * This is very much the same as waitpid() 201 */ 202 int 203 linux_sys_wait4(l, v, retval) 204 struct lwp *l; 205 void *v; 206 register_t *retval; 207 { 208 struct linux_sys_wait4_args /* { 209 syscallarg(int) pid; 210 syscallarg(int *) status; 211 syscallarg(int) options; 212 syscallarg(struct rusage *) rusage; 213 } */ *uap = v; 214 struct proc *p = l->l_proc; 215 struct sys_wait4_args w4a; 216 int error, *status, tstat, options, linux_options; 217 caddr_t sg; 218 219 if (SCARG(uap, status) != NULL) { 220 sg = stackgap_init(p, 0); 221 status = (int *) stackgap_alloc(p, &sg, sizeof *status); 222 } else 223 status = NULL; 224 225 linux_options = SCARG(uap, options); 226 options = 0; 227 if (linux_options & 228 ~(LINUX_WAIT4_WNOHANG|LINUX_WAIT4_WUNTRACED|LINUX_WAIT4_WALL| 229 LINUX_WAIT4_WCLONE)) 230 return (EINVAL); 231 232 if (linux_options & LINUX_WAIT4_WNOHANG) 233 options |= WNOHANG; 234 if (linux_options & LINUX_WAIT4_WUNTRACED) 235 options |= WUNTRACED; 236 if (linux_options & LINUX_WAIT4_WALL) 237 options |= WALLSIG; 238 if (linux_options & LINUX_WAIT4_WCLONE) 239 options |= WALTSIG; 240 241 SCARG(&w4a, pid) = SCARG(uap, pid); 242 SCARG(&w4a, status) = status; 243 SCARG(&w4a, options) = options; 244 SCARG(&w4a, rusage) = SCARG(uap, rusage); 245 246 if ((error = sys_wait4(l, &w4a, retval))) 247 return error; 248 249 sigdelset(&p->p_sigctx.ps_siglist, SIGCHLD); 250 251 if (status != NULL) { 252 if ((error = copyin(status, &tstat, sizeof tstat))) 253 return error; 254 255 bsd_to_linux_wstat(&tstat); 256 return copyout(&tstat, SCARG(uap, status), sizeof tstat); 257 } 258 259 return 0; 260 } 261 262 /* 263 * Linux brk(2). The check if the new address is >= the old one is 264 * done in the kernel in Linux. NetBSD does it in the library. 265 */ 266 int 267 linux_sys_brk(l, v, retval) 268 struct lwp *l; 269 void *v; 270 register_t *retval; 271 { 272 struct linux_sys_brk_args /* { 273 syscallarg(char *) nsize; 274 } */ *uap = v; 275 struct proc *p = l->l_proc; 276 char *nbrk = SCARG(uap, nsize); 277 struct sys_obreak_args oba; 278 struct vmspace *vm = p->p_vmspace; 279 struct linux_emuldata *ed = (struct linux_emuldata*)p->p_emuldata; 280 281 SCARG(&oba, nsize) = nbrk; 282 283 if ((caddr_t) nbrk > vm->vm_daddr && sys_obreak(l, &oba, retval) == 0) 284 ed->p_break = (char *)nbrk; 285 else 286 nbrk = ed->p_break; 287 288 retval[0] = (register_t)nbrk; 289 return 0; 290 } 291 292 /* 293 * Convert BSD statfs structure to Linux statfs structure. 294 * The Linux structure has less fields, and it also wants 295 * the length of a name in a dir entry in a field, which 296 * we fake (probably the wrong way). 297 */ 298 static void 299 bsd_to_linux_statfs(bsp, lsp) 300 struct statfs *bsp; 301 struct linux_statfs *lsp; 302 { 303 int i; 304 305 for (i = 0; i < FSTYPESSIZE; i++) 306 if (strcmp(bsp->f_fstypename, fstypes[i].bsd) == 0) 307 break; 308 309 if (i == FSTYPESSIZE) { 310 DPRINTF(("unhandled fstype in linux emulation: %s\n", 311 bsp->f_fstypename)); 312 lsp->l_ftype = LINUX_DEFAULT_SUPER_MAGIC; 313 } else { 314 lsp->l_ftype = fstypes[i].linux; 315 } 316 317 lsp->l_fbsize = bsp->f_bsize; 318 lsp->l_fblocks = bsp->f_blocks; 319 lsp->l_fbfree = bsp->f_bfree; 320 lsp->l_fbavail = bsp->f_bavail; 321 lsp->l_ffiles = bsp->f_files; 322 lsp->l_fffree = bsp->f_ffree; 323 /* Linux sets the fsid to 0..., we don't */ 324 lsp->l_ffsid.val[0] = bsp->f_fsid.val[0]; 325 lsp->l_ffsid.val[1] = bsp->f_fsid.val[1]; 326 lsp->l_fnamelen = MAXNAMLEN; /* XXX */ 327 (void)memset(lsp->l_fspare, 0, sizeof(lsp->l_fspare)); 328 } 329 330 /* 331 * Implement the fs stat functions. Straightforward. 332 */ 333 int 334 linux_sys_statfs(l, v, retval) 335 struct lwp *l; 336 void *v; 337 register_t *retval; 338 { 339 struct linux_sys_statfs_args /* { 340 syscallarg(const char *) path; 341 syscallarg(struct linux_statfs *) sp; 342 } */ *uap = v; 343 struct proc *p = l->l_proc; 344 struct statfs btmp, *bsp; 345 struct linux_statfs ltmp; 346 struct sys_statfs_args bsa; 347 caddr_t sg; 348 int error; 349 350 sg = stackgap_init(p, 0); 351 bsp = (struct statfs *) stackgap_alloc(p, &sg, sizeof (struct statfs)); 352 353 CHECK_ALT_EXIST(p, &sg, SCARG(uap, path)); 354 355 SCARG(&bsa, path) = SCARG(uap, path); 356 SCARG(&bsa, buf) = bsp; 357 358 if ((error = sys_statfs(l, &bsa, retval))) 359 return error; 360 361 if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp))) 362 return error; 363 364 bsd_to_linux_statfs(&btmp, <mp); 365 366 return copyout((caddr_t) <mp, (caddr_t) SCARG(uap, sp), sizeof ltmp); 367 } 368 369 int 370 linux_sys_fstatfs(l, v, retval) 371 struct lwp *l; 372 void *v; 373 register_t *retval; 374 { 375 struct linux_sys_fstatfs_args /* { 376 syscallarg(int) fd; 377 syscallarg(struct linux_statfs *) sp; 378 } */ *uap = v; 379 struct proc *p = l->l_proc; 380 struct statfs btmp, *bsp; 381 struct linux_statfs ltmp; 382 struct sys_fstatfs_args bsa; 383 caddr_t sg; 384 int error; 385 386 sg = stackgap_init(p, 0); 387 bsp = (struct statfs *) stackgap_alloc(p, &sg, sizeof (struct statfs)); 388 389 SCARG(&bsa, fd) = SCARG(uap, fd); 390 SCARG(&bsa, buf) = bsp; 391 392 if ((error = sys_fstatfs(l, &bsa, retval))) 393 return error; 394 395 if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp))) 396 return error; 397 398 bsd_to_linux_statfs(&btmp, <mp); 399 400 return copyout((caddr_t) <mp, (caddr_t) SCARG(uap, sp), sizeof ltmp); 401 } 402 403 /* 404 * uname(). Just copy the info from the various strings stored in the 405 * kernel, and put it in the Linux utsname structure. That structure 406 * is almost the same as the NetBSD one, only it has fields 65 characters 407 * long, and an extra domainname field. 408 */ 409 int 410 linux_sys_uname(l, v, retval) 411 struct lwp *l; 412 void *v; 413 register_t *retval; 414 { 415 struct linux_sys_uname_args /* { 416 syscallarg(struct linux_utsname *) up; 417 } */ *uap = v; 418 struct linux_utsname luts; 419 420 strncpy(luts.l_sysname, linux_sysname, sizeof(luts.l_sysname)); 421 strncpy(luts.l_nodename, hostname, sizeof(luts.l_nodename)); 422 strncpy(luts.l_release, linux_release, sizeof(luts.l_release)); 423 strncpy(luts.l_version, linux_version, sizeof(luts.l_version)); 424 strncpy(luts.l_machine, machine, sizeof(luts.l_machine)); 425 strncpy(luts.l_domainname, domainname, sizeof(luts.l_domainname)); 426 427 return copyout(&luts, SCARG(uap, up), sizeof(luts)); 428 } 429 430 /* Used directly on: alpha, mips, ppc, sparc, sparc64 */ 431 /* Used indirectly on: arm, i386, m68k */ 432 433 /* 434 * New type Linux mmap call. 435 * Only called directly on machines with >= 6 free regs. 436 */ 437 int 438 linux_sys_mmap(l, v, retval) 439 struct lwp *l; 440 void *v; 441 register_t *retval; 442 { 443 struct linux_sys_mmap_args /* { 444 syscallarg(unsigned long) addr; 445 syscallarg(size_t) len; 446 syscallarg(int) prot; 447 syscallarg(int) flags; 448 syscallarg(int) fd; 449 syscallarg(linux_off_t) offset; 450 } */ *uap = v; 451 struct sys_mmap_args cma; 452 int flags, fl = SCARG(uap, flags); 453 454 flags = 0; 455 flags |= cvtto_bsd_mask(fl, LINUX_MAP_SHARED, MAP_SHARED); 456 flags |= cvtto_bsd_mask(fl, LINUX_MAP_PRIVATE, MAP_PRIVATE); 457 flags |= cvtto_bsd_mask(fl, LINUX_MAP_FIXED, MAP_FIXED); 458 flags |= cvtto_bsd_mask(fl, LINUX_MAP_ANON, MAP_ANON); 459 /* XXX XAX ERH: Any other flags here? There are more defined... */ 460 461 SCARG(&cma, addr) = (void *)SCARG(uap, addr); 462 SCARG(&cma, len) = SCARG(uap, len); 463 SCARG(&cma, prot) = SCARG(uap, prot); 464 if (SCARG(&cma, prot) & VM_PROT_WRITE) /* XXX */ 465 SCARG(&cma, prot) |= VM_PROT_READ; 466 SCARG(&cma, flags) = flags; 467 SCARG(&cma, fd) = flags & MAP_ANON ? -1 : SCARG(uap, fd); 468 SCARG(&cma, pad) = 0; 469 SCARG(&cma, pos) = (off_t)SCARG(uap, offset); 470 471 return sys_mmap(l, &cma, retval); 472 } 473 474 int 475 linux_sys_mremap(l, v, retval) 476 struct lwp *l; 477 void *v; 478 register_t *retval; 479 { 480 struct linux_sys_mremap_args /* { 481 syscallarg(void *) old_address; 482 syscallarg(size_t) old_size; 483 syscallarg(size_t) new_size; 484 syscallarg(u_long) flags; 485 } */ *uap = v; 486 struct sys_munmap_args mua; 487 size_t old_size, new_size; 488 int error; 489 490 old_size = round_page(SCARG(uap, old_size)); 491 new_size = round_page(SCARG(uap, new_size)); 492 493 /* 494 * Growing mapped region. 495 */ 496 if (new_size > old_size) { 497 /* 498 * XXX Implement me. What we probably want to do is 499 * XXX dig out the guts of the old mapping, mmap that 500 * XXX object again with the new size, then munmap 501 * XXX the old mapping. 502 */ 503 *retval = 0; 504 return (ENOMEM); 505 } 506 507 /* 508 * Shrinking mapped region. 509 */ 510 if (new_size < old_size) { 511 SCARG(&mua, addr) = (caddr_t)SCARG(uap, old_address) + 512 new_size; 513 SCARG(&mua, len) = old_size - new_size; 514 error = sys_munmap(l, &mua, retval); 515 *retval = error ? 0 : (register_t)SCARG(uap, old_address); 516 return (error); 517 } 518 519 /* 520 * No change. 521 */ 522 *retval = (register_t)SCARG(uap, old_address); 523 return (0); 524 } 525 526 int 527 linux_sys_msync(l, v, retval) 528 struct lwp *l; 529 void *v; 530 register_t *retval; 531 { 532 struct linux_sys_msync_args /* { 533 syscallarg(caddr_t) addr; 534 syscallarg(int) len; 535 syscallarg(int) fl; 536 } */ *uap = v; 537 538 struct sys___msync13_args bma; 539 540 /* flags are ignored */ 541 SCARG(&bma, addr) = SCARG(uap, addr); 542 SCARG(&bma, len) = SCARG(uap, len); 543 SCARG(&bma, flags) = SCARG(uap, fl); 544 545 return sys___msync13(l, &bma, retval); 546 } 547 548 int 549 linux_sys_mprotect(l, v, retval) 550 struct lwp *l; 551 void *v; 552 register_t *retval; 553 { 554 struct linux_sys_mprotect_args /* { 555 syscallarg(const void *) start; 556 syscallarg(unsigned long) len; 557 syscallarg(int) prot; 558 } */ *uap = v; 559 unsigned long end, start = (unsigned long)SCARG(uap, start), len; 560 int prot = SCARG(uap, prot); 561 struct vm_map_entry *entry; 562 struct vm_map *map = &l->l_proc->p_vmspace->vm_map; 563 564 if (start & PAGE_MASK) 565 return EINVAL; 566 567 len = round_page(SCARG(uap, len)); 568 end = start + len; 569 570 if (end < start) 571 return EINVAL; 572 else if (end == start) 573 return 0; 574 575 if (SCARG(uap, prot) & ~(PROT_READ | PROT_WRITE | PROT_EXEC)) 576 return EINVAL; 577 578 vm_map_lock(map); 579 #ifdef notdef 580 VM_MAP_RANGE_CHECK(map, start, end); 581 #endif 582 if (!uvm_map_lookup_entry(map, start, &entry) || entry->start > start) { 583 vm_map_unlock(map); 584 return EFAULT; 585 } 586 vm_map_unlock(map); 587 return uvm_map_protect(map, start, end, prot, FALSE); 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)) 597 598 int 599 linux_sys_times(l, v, retval) 600 struct lwp *l; 601 void *v; 602 register_t *retval; 603 { 604 struct linux_sys_times_args /* { 605 syscallarg(struct times *) tms; 606 } */ *uap = v; 607 struct proc *p = l->l_proc; 608 struct timeval t; 609 struct linux_tms ltms; 610 struct rusage ru; 611 int error, s; 612 613 calcru(p, &ru.ru_utime, &ru.ru_stime, NULL); 614 ltms.ltms_utime = CONVTCK(ru.ru_utime); 615 ltms.ltms_stime = CONVTCK(ru.ru_stime); 616 617 ltms.ltms_cutime = CONVTCK(p->p_stats->p_cru.ru_utime); 618 ltms.ltms_cstime = CONVTCK(p->p_stats->p_cru.ru_stime); 619 620 if ((error = copyout(<ms, SCARG(uap, tms), sizeof ltms))) 621 return error; 622 623 s = splclock(); 624 timersub(&time, &boottime, &t); 625 splx(s); 626 627 retval[0] = ((linux_clock_t)(CONVTCK(t))); 628 return 0; 629 } 630 631 /* 632 * Linux 'readdir' call. This code is mostly taken from the 633 * SunOS getdents call (see compat/sunos/sunos_misc.c), though 634 * an attempt has been made to keep it a little cleaner (failing 635 * miserably, because of the cruft needed if count 1 is passed). 636 * 637 * The d_off field should contain the offset of the next valid entry, 638 * but in Linux it has the offset of the entry itself. We emulate 639 * that bug here. 640 * 641 * Read in BSD-style entries, convert them, and copy them out. 642 * 643 * Note that this doesn't handle union-mounted filesystems. 644 */ 645 int 646 linux_sys_getdents(l, v, retval) 647 struct lwp *l; 648 void *v; 649 register_t *retval; 650 { 651 struct linux_sys_getdents_args /* { 652 syscallarg(int) fd; 653 syscallarg(struct linux_dirent *) dent; 654 syscallarg(unsigned int) count; 655 } */ *uap = v; 656 struct proc *p = l->l_proc; 657 struct dirent *bdp; 658 struct vnode *vp; 659 caddr_t inp, buf; /* BSD-format */ 660 int len, reclen; /* BSD-format */ 661 caddr_t outp; /* Linux-format */ 662 int resid, linux_reclen = 0; /* Linux-format */ 663 struct file *fp; 664 struct uio auio; 665 struct iovec aiov; 666 struct linux_dirent idb; 667 off_t off; /* true file offset */ 668 int buflen, error, eofflag, nbytes, oldcall; 669 struct vattr va; 670 off_t *cookiebuf = NULL, *cookie; 671 int ncookies; 672 673 /* getvnode() will use the descriptor for us */ 674 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 675 return (error); 676 677 if ((fp->f_flag & FREAD) == 0) { 678 error = EBADF; 679 goto out1; 680 } 681 682 vp = (struct vnode *)fp->f_data; 683 if (vp->v_type != VDIR) { 684 error = EINVAL; 685 goto out1; 686 } 687 688 if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p))) 689 goto out1; 690 691 nbytes = SCARG(uap, count); 692 if (nbytes == 1) { /* emulating old, broken behaviour */ 693 nbytes = sizeof (idb); 694 buflen = max(va.va_blocksize, nbytes); 695 oldcall = 1; 696 } else { 697 buflen = min(MAXBSIZE, nbytes); 698 if (buflen < va.va_blocksize) 699 buflen = va.va_blocksize; 700 oldcall = 0; 701 } 702 buf = malloc(buflen, M_TEMP, M_WAITOK); 703 704 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 705 off = fp->f_offset; 706 again: 707 aiov.iov_base = buf; 708 aiov.iov_len = buflen; 709 auio.uio_iov = &aiov; 710 auio.uio_iovcnt = 1; 711 auio.uio_rw = UIO_READ; 712 auio.uio_segflg = UIO_SYSSPACE; 713 auio.uio_procp = p; 714 auio.uio_resid = buflen; 715 auio.uio_offset = off; 716 /* 717 * First we read into the malloc'ed buffer, then 718 * we massage it into user space, one record at a time. 719 */ 720 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf, 721 &ncookies); 722 if (error) 723 goto out; 724 725 inp = buf; 726 outp = (caddr_t)SCARG(uap, dent); 727 resid = nbytes; 728 if ((len = buflen - auio.uio_resid) == 0) 729 goto eof; 730 731 for (cookie = cookiebuf; len > 0; len -= reclen) { 732 bdp = (struct dirent *)inp; 733 reclen = bdp->d_reclen; 734 if (reclen & 3) 735 panic("linux_readdir"); 736 if (bdp->d_fileno == 0) { 737 inp += reclen; /* it is a hole; squish it out */ 738 off = *cookie++; 739 continue; 740 } 741 linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen); 742 if (reclen > len || resid < linux_reclen) { 743 /* entry too big for buffer, so just stop */ 744 outp++; 745 break; 746 } 747 /* 748 * Massage in place to make a Linux-shaped dirent (otherwise 749 * we have to worry about touching user memory outside of 750 * the copyout() call). 751 */ 752 idb.d_ino = bdp->d_fileno; 753 /* 754 * The old readdir() call misuses the offset and reclen fields. 755 */ 756 if (oldcall) { 757 idb.d_off = (linux_off_t)linux_reclen; 758 idb.d_reclen = (u_short)bdp->d_namlen; 759 } else { 760 if (sizeof (idb.d_off) < 4 && (off >> 32) != 0) { 761 compat_offseterr(vp, "linux_getdents"); 762 error = EINVAL; 763 goto out; 764 } 765 idb.d_off = (linux_off_t)off; 766 idb.d_reclen = (u_short)linux_reclen; 767 } 768 strcpy(idb.d_name, bdp->d_name); 769 if ((error = copyout((caddr_t)&idb, outp, linux_reclen))) 770 goto out; 771 /* advance past this real entry */ 772 inp += reclen; 773 off = *cookie++; /* each entry points to itself */ 774 /* advance output past Linux-shaped entry */ 775 outp += linux_reclen; 776 resid -= linux_reclen; 777 if (oldcall) 778 break; 779 } 780 781 /* if we squished out the whole block, try again */ 782 if (outp == (caddr_t)SCARG(uap, dent)) 783 goto again; 784 fp->f_offset = off; /* update the vnode offset */ 785 786 if (oldcall) 787 nbytes = resid + linux_reclen; 788 789 eof: 790 *retval = nbytes - resid; 791 out: 792 VOP_UNLOCK(vp, 0); 793 if (cookiebuf) 794 free(cookiebuf, M_TEMP); 795 free(buf, M_TEMP); 796 out1: 797 FILE_UNUSE(fp, p); 798 return error; 799 } 800 801 /* 802 * Linux 'readdir' call. This code is mostly taken from the 803 * SunOS getdents call (see compat/sunos/sunos_misc.c), though 804 * an attempt has been made to keep it a little cleaner (failing 805 * miserably, because of the cruft needed if count 1 is passed). 806 * 807 * The d_off field should contain the offset of the next valid entry, 808 * but in Linux it has the offset of the entry itself. We emulate 809 * that bug here. 810 * 811 * Read in BSD-style entries, convert them, and copy them out. 812 * 813 * Note that this doesn't handle union-mounted filesystems. 814 */ 815 int 816 linux_sys_getdents64(l, v, retval) 817 struct lwp *l; 818 void *v; 819 register_t *retval; 820 { 821 struct linux_sys_getdents_args /* { 822 syscallarg(int) fd; 823 syscallarg(struct linux_dirent64 *) dent; 824 syscallarg(unsigned int) count; 825 } */ *uap = v; 826 struct proc *p = l->l_proc; 827 struct dirent *bdp; 828 struct vnode *vp; 829 caddr_t inp, buf; /* BSD-format */ 830 int len, reclen; /* BSD-format */ 831 caddr_t outp; /* Linux-format */ 832 int resid, linux_reclen = 0; /* Linux-format */ 833 struct file *fp; 834 struct uio auio; 835 struct iovec aiov; 836 struct linux_dirent64 idb; 837 off_t off; /* true file offset */ 838 int buflen, error, eofflag, nbytes, oldcall; 839 struct vattr va; 840 off_t *cookiebuf = NULL, *cookie; 841 int ncookies; 842 843 /* getvnode() will use the descriptor for us */ 844 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 845 return (error); 846 847 if ((fp->f_flag & FREAD) == 0) { 848 error = EBADF; 849 goto out1; 850 } 851 852 vp = (struct vnode *)fp->f_data; 853 if (vp->v_type != VDIR) { 854 error = EINVAL; 855 goto out1; 856 } 857 858 if ((error = VOP_GETATTR(vp, &va, p->p_ucred, p))) 859 goto out1; 860 861 nbytes = SCARG(uap, count); 862 if (nbytes == 1) { /* emulating old, broken behaviour */ 863 nbytes = sizeof (idb); 864 buflen = max(va.va_blocksize, nbytes); 865 oldcall = 1; 866 } else { 867 buflen = min(MAXBSIZE, nbytes); 868 if (buflen < va.va_blocksize) 869 buflen = va.va_blocksize; 870 oldcall = 0; 871 } 872 buf = malloc(buflen, M_TEMP, M_WAITOK); 873 874 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 875 off = fp->f_offset; 876 again: 877 aiov.iov_base = buf; 878 aiov.iov_len = buflen; 879 auio.uio_iov = &aiov; 880 auio.uio_iovcnt = 1; 881 auio.uio_rw = UIO_READ; 882 auio.uio_segflg = UIO_SYSSPACE; 883 auio.uio_procp = p; 884 auio.uio_resid = buflen; 885 auio.uio_offset = off; 886 /* 887 * First we read into the malloc'ed buffer, then 888 * we massage it into user space, one record at a time. 889 */ 890 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf, 891 &ncookies); 892 if (error) 893 goto out; 894 895 inp = buf; 896 outp = (caddr_t)SCARG(uap, dent); 897 resid = nbytes; 898 if ((len = buflen - auio.uio_resid) == 0) 899 goto eof; 900 901 for (cookie = cookiebuf; len > 0; len -= reclen) { 902 bdp = (struct dirent *)inp; 903 reclen = bdp->d_reclen; 904 if (reclen & 3) 905 panic("linux_readdir"); 906 if (bdp->d_fileno == 0) { 907 inp += reclen; /* it is a hole; squish it out */ 908 off = *cookie++; 909 continue; 910 } 911 linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen); 912 if (reclen > len || resid < linux_reclen) { 913 /* entry too big for buffer, so just stop */ 914 outp++; 915 break; 916 } 917 /* 918 * Massage in place to make a Linux-shaped dirent (otherwise 919 * we have to worry about touching user memory outside of 920 * the copyout() call). 921 */ 922 idb.d_ino = bdp->d_fileno; 923 idb.d_type = bdp->d_type; 924 /* 925 * The old readdir() call misuses the offset and reclen fields. 926 */ 927 if (oldcall) { 928 idb.d_off = linux_reclen; 929 idb.d_reclen = (u_short)bdp->d_namlen; 930 } else { 931 if (sizeof (idb.d_off) < 4 && (off >> 32) != 0) { 932 compat_offseterr(vp, "linux_getdents"); 933 error = EINVAL; 934 goto out; 935 } 936 idb.d_off = off; 937 idb.d_reclen = (u_short)linux_reclen; 938 } 939 strcpy(idb.d_name, bdp->d_name); 940 if ((error = copyout((caddr_t)&idb, outp, linux_reclen))) 941 goto out; 942 /* advance past this real entry */ 943 inp += reclen; 944 off = *cookie++; /* each entry points to itself */ 945 /* advance output past Linux-shaped entry */ 946 outp += linux_reclen; 947 resid -= linux_reclen; 948 if (oldcall) 949 break; 950 } 951 952 /* if we squished out the whole block, try again */ 953 if (outp == (caddr_t)SCARG(uap, dent)) 954 goto again; 955 fp->f_offset = off; /* update the vnode offset */ 956 957 if (oldcall) 958 nbytes = resid + linux_reclen; 959 960 eof: 961 *retval = nbytes - resid; 962 out: 963 VOP_UNLOCK(vp, 0); 964 if (cookiebuf) 965 free(cookiebuf, M_TEMP); 966 free(buf, M_TEMP); 967 out1: 968 FILE_UNUSE(fp, p); 969 return error; 970 } 971 972 /* 973 * Even when just using registers to pass arguments to syscalls you can 974 * have 5 of them on the i386. So this newer version of select() does 975 * this. 976 */ 977 int 978 linux_sys_select(l, v, retval) 979 struct lwp *l; 980 void *v; 981 register_t *retval; 982 { 983 struct linux_sys_select_args /* { 984 syscallarg(int) nfds; 985 syscallarg(fd_set *) readfds; 986 syscallarg(fd_set *) writefds; 987 syscallarg(fd_set *) exceptfds; 988 syscallarg(struct timeval *) timeout; 989 } */ *uap = v; 990 991 return linux_select1(l, retval, SCARG(uap, nfds), SCARG(uap, readfds), 992 SCARG(uap, writefds), SCARG(uap, exceptfds), SCARG(uap, timeout)); 993 } 994 995 /* 996 * Common code for the old and new versions of select(). A couple of 997 * things are important: 998 * 1) return the amount of time left in the 'timeout' parameter 999 * 2) select never returns ERESTART on Linux, always return EINTR 1000 */ 1001 int 1002 linux_select1(l, retval, nfds, readfds, writefds, exceptfds, timeout) 1003 struct lwp *l; 1004 register_t *retval; 1005 int nfds; 1006 fd_set *readfds, *writefds, *exceptfds; 1007 struct timeval *timeout; 1008 { 1009 struct sys_select_args bsa; 1010 struct proc *p = l->l_proc; 1011 struct timeval tv0, tv1, utv, *tvp; 1012 caddr_t sg; 1013 int error; 1014 1015 SCARG(&bsa, nd) = nfds; 1016 SCARG(&bsa, in) = readfds; 1017 SCARG(&bsa, ou) = writefds; 1018 SCARG(&bsa, ex) = exceptfds; 1019 SCARG(&bsa, tv) = timeout; 1020 1021 /* 1022 * Store current time for computation of the amount of 1023 * time left. 1024 */ 1025 if (timeout) { 1026 if ((error = copyin(timeout, &utv, sizeof(utv)))) 1027 return error; 1028 if (itimerfix(&utv)) { 1029 /* 1030 * The timeval was invalid. Convert it to something 1031 * valid that will act as it does under Linux. 1032 */ 1033 sg = stackgap_init(p, 0); 1034 tvp = stackgap_alloc(p, &sg, sizeof(utv)); 1035 utv.tv_sec += utv.tv_usec / 1000000; 1036 utv.tv_usec %= 1000000; 1037 if (utv.tv_usec < 0) { 1038 utv.tv_sec -= 1; 1039 utv.tv_usec += 1000000; 1040 } 1041 if (utv.tv_sec < 0) 1042 timerclear(&utv); 1043 if ((error = copyout(&utv, tvp, sizeof(utv)))) 1044 return error; 1045 SCARG(&bsa, tv) = tvp; 1046 } 1047 microtime(&tv0); 1048 } 1049 1050 error = sys_select(l, &bsa, retval); 1051 if (error) { 1052 /* 1053 * See fs/select.c in the Linux kernel. Without this, 1054 * Maelstrom doesn't work. 1055 */ 1056 if (error == ERESTART) 1057 error = EINTR; 1058 return error; 1059 } 1060 1061 if (timeout) { 1062 if (*retval) { 1063 /* 1064 * Compute how much time was left of the timeout, 1065 * by subtracting the current time and the time 1066 * before we started the call, and subtracting 1067 * that result from the user-supplied value. 1068 */ 1069 microtime(&tv1); 1070 timersub(&tv1, &tv0, &tv1); 1071 timersub(&utv, &tv1, &utv); 1072 if (utv.tv_sec < 0) 1073 timerclear(&utv); 1074 } else 1075 timerclear(&utv); 1076 if ((error = copyout(&utv, timeout, sizeof(utv)))) 1077 return error; 1078 } 1079 1080 return 0; 1081 } 1082 1083 /* 1084 * Get the process group of a certain process. Look it up 1085 * and return the value. 1086 */ 1087 int 1088 linux_sys_getpgid(l, v, retval) 1089 struct lwp *l; 1090 void *v; 1091 register_t *retval; 1092 { 1093 struct linux_sys_getpgid_args /* { 1094 syscallarg(int) pid; 1095 } */ *uap = v; 1096 struct proc *p = l->l_proc; 1097 struct proc *targp; 1098 1099 if (SCARG(uap, pid) != 0 && SCARG(uap, pid) != p->p_pid) { 1100 if ((targp = pfind(SCARG(uap, pid))) == 0) 1101 return ESRCH; 1102 } 1103 else 1104 targp = p; 1105 1106 retval[0] = targp->p_pgid; 1107 return 0; 1108 } 1109 1110 /* 1111 * Set the 'personality' (emulation mode) for the current process. Only 1112 * accept the Linux personality here (0). This call is needed because 1113 * the Linux ELF crt0 issues it in an ugly kludge to make sure that 1114 * ELF binaries run in Linux mode, not SVR4 mode. 1115 */ 1116 int 1117 linux_sys_personality(l, v, retval) 1118 struct lwp *l; 1119 void *v; 1120 register_t *retval; 1121 { 1122 struct linux_sys_personality_args /* { 1123 syscallarg(int) per; 1124 } */ *uap = v; 1125 1126 if (SCARG(uap, per) != 0) 1127 return EINVAL; 1128 retval[0] = 0; 1129 return 0; 1130 } 1131 1132 #if defined(__i386__) || defined(__m68k__) 1133 /* 1134 * The calls are here because of type conversions. 1135 */ 1136 int 1137 linux_sys_setreuid16(l, v, retval) 1138 struct lwp *l; 1139 void *v; 1140 register_t *retval; 1141 { 1142 struct linux_sys_setreuid16_args /* { 1143 syscallarg(int) ruid; 1144 syscallarg(int) euid; 1145 } */ *uap = v; 1146 struct sys_setreuid_args bsa; 1147 1148 SCARG(&bsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ? 1149 (uid_t)-1 : SCARG(uap, ruid); 1150 SCARG(&bsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ? 1151 (uid_t)-1 : SCARG(uap, euid); 1152 1153 return sys_setreuid(l, &bsa, retval); 1154 } 1155 1156 int 1157 linux_sys_setregid16(l, v, retval) 1158 struct lwp *l; 1159 void *v; 1160 register_t *retval; 1161 { 1162 struct linux_sys_setregid16_args /* { 1163 syscallarg(int) rgid; 1164 syscallarg(int) egid; 1165 } */ *uap = v; 1166 struct sys_setregid_args bsa; 1167 1168 SCARG(&bsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ? 1169 (uid_t)-1 : SCARG(uap, rgid); 1170 SCARG(&bsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ? 1171 (uid_t)-1 : SCARG(uap, egid); 1172 1173 return sys_setregid(l, &bsa, retval); 1174 } 1175 1176 int 1177 linux_sys_setresuid16(l, v, retval) 1178 struct lwp *l; 1179 void *v; 1180 register_t *retval; 1181 { 1182 struct linux_sys_setresuid16_args /* { 1183 syscallarg(uid_t) ruid; 1184 syscallarg(uid_t) euid; 1185 syscallarg(uid_t) suid; 1186 } */ *uap = v; 1187 struct linux_sys_setresuid16_args lsa; 1188 1189 SCARG(&lsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ? 1190 (uid_t)-1 : SCARG(uap, ruid); 1191 SCARG(&lsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ? 1192 (uid_t)-1 : SCARG(uap, euid); 1193 SCARG(&lsa, suid) = ((linux_uid_t)SCARG(uap, suid) == (linux_uid_t)-1) ? 1194 (uid_t)-1 : SCARG(uap, suid); 1195 1196 return linux_sys_setresuid(l, &lsa, retval); 1197 } 1198 1199 int 1200 linux_sys_setresgid16(l, v, retval) 1201 struct lwp *l; 1202 void *v; 1203 register_t *retval; 1204 { 1205 struct linux_sys_setresgid16_args /* { 1206 syscallarg(gid_t) rgid; 1207 syscallarg(gid_t) egid; 1208 syscallarg(gid_t) sgid; 1209 } */ *uap = v; 1210 struct linux_sys_setresgid16_args lsa; 1211 1212 SCARG(&lsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ? 1213 (gid_t)-1 : SCARG(uap, rgid); 1214 SCARG(&lsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ? 1215 (gid_t)-1 : SCARG(uap, egid); 1216 SCARG(&lsa, sgid) = ((linux_gid_t)SCARG(uap, sgid) == (linux_gid_t)-1) ? 1217 (gid_t)-1 : SCARG(uap, sgid); 1218 1219 return linux_sys_setresgid(l, &lsa, retval); 1220 } 1221 1222 int 1223 linux_sys_getgroups16(l, v, retval) 1224 struct lwp *l; 1225 void *v; 1226 register_t *retval; 1227 { 1228 struct linux_sys_getgroups16_args /* { 1229 syscallarg(int) gidsetsize; 1230 syscallarg(linux_gid_t *) gidset; 1231 } */ *uap = v; 1232 struct proc *p = l->l_proc; 1233 caddr_t sg; 1234 int n, error, i; 1235 struct sys_getgroups_args bsa; 1236 gid_t *bset, *kbset; 1237 linux_gid_t *lset; 1238 struct pcred *pc = p->p_cred; 1239 1240 n = SCARG(uap, gidsetsize); 1241 if (n < 0) 1242 return EINVAL; 1243 error = 0; 1244 bset = kbset = NULL; 1245 lset = NULL; 1246 if (n > 0) { 1247 n = min(pc->pc_ucred->cr_ngroups, n); 1248 sg = stackgap_init(p, 0); 1249 bset = stackgap_alloc(p, &sg, n * sizeof (gid_t)); 1250 kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK); 1251 lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK); 1252 if (bset == NULL || kbset == NULL || lset == NULL) 1253 return ENOMEM; 1254 SCARG(&bsa, gidsetsize) = n; 1255 SCARG(&bsa, gidset) = bset; 1256 error = sys_getgroups(l, &bsa, retval); 1257 if (error != 0) 1258 goto out; 1259 error = copyin(bset, kbset, n * sizeof (gid_t)); 1260 if (error != 0) 1261 goto out; 1262 for (i = 0; i < n; i++) 1263 lset[i] = (linux_gid_t)kbset[i]; 1264 error = copyout(lset, SCARG(uap, gidset), 1265 n * sizeof (linux_gid_t)); 1266 } else 1267 *retval = pc->pc_ucred->cr_ngroups; 1268 out: 1269 if (kbset != NULL) 1270 free(kbset, M_TEMP); 1271 if (lset != NULL) 1272 free(lset, M_TEMP); 1273 return error; 1274 } 1275 1276 int 1277 linux_sys_setgroups16(l, v, retval) 1278 struct lwp *l; 1279 void *v; 1280 register_t *retval; 1281 { 1282 struct linux_sys_setgroups16_args /* { 1283 syscallarg(int) gidsetsize; 1284 syscallarg(linux_gid_t *) gidset; 1285 } */ *uap = v; 1286 struct proc *p = l->l_proc; 1287 caddr_t sg; 1288 int n; 1289 int error, i; 1290 struct sys_setgroups_args bsa; 1291 gid_t *bset, *kbset; 1292 linux_gid_t *lset; 1293 1294 n = SCARG(uap, gidsetsize); 1295 if (n < 0 || n > NGROUPS) 1296 return EINVAL; 1297 sg = stackgap_init(p, 0); 1298 bset = stackgap_alloc(p, &sg, n * sizeof (gid_t)); 1299 lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK); 1300 kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK); 1301 if (lset == NULL || bset == NULL) 1302 return ENOMEM; 1303 error = copyin(SCARG(uap, gidset), lset, n * sizeof (linux_gid_t)); 1304 if (error != 0) 1305 goto out; 1306 for (i = 0; i < n; i++) 1307 kbset[i] = (gid_t)lset[i]; 1308 error = copyout(kbset, bset, n * sizeof (gid_t)); 1309 if (error != 0) 1310 goto out; 1311 SCARG(&bsa, gidsetsize) = n; 1312 SCARG(&bsa, gidset) = bset; 1313 error = sys_setgroups(l, &bsa, retval); 1314 1315 out: 1316 if (lset != NULL) 1317 free(lset, M_TEMP); 1318 if (kbset != NULL) 1319 free(kbset, M_TEMP); 1320 1321 return error; 1322 } 1323 1324 #endif /* __i386__ || __m68k__ */ 1325 1326 /* 1327 * We have nonexistent fsuid equal to uid. 1328 * If modification is requested, refuse. 1329 */ 1330 int 1331 linux_sys_setfsuid(l, v, retval) 1332 struct lwp *l; 1333 void *v; 1334 register_t *retval; 1335 { 1336 struct linux_sys_setfsuid_args /* { 1337 syscallarg(uid_t) uid; 1338 } */ *uap = v; 1339 struct proc *p = l->l_proc; 1340 uid_t uid; 1341 1342 uid = SCARG(uap, uid); 1343 if (p->p_cred->p_ruid != uid) 1344 return sys_nosys(l, v, retval); 1345 else 1346 return (0); 1347 } 1348 1349 /* XXX XXX XXX */ 1350 #ifndef alpha 1351 int 1352 linux_sys_getfsuid(l, v, retval) 1353 struct lwp *l; 1354 void *v; 1355 register_t *retval; 1356 { 1357 return sys_getuid(l, v, retval); 1358 } 1359 #endif 1360 1361 int 1362 linux_sys_setresuid(l, v, retval) 1363 struct lwp *l; 1364 void *v; 1365 register_t *retval; 1366 { 1367 struct linux_sys_setresuid_args /* { 1368 syscallarg(uid_t) ruid; 1369 syscallarg(uid_t) euid; 1370 syscallarg(uid_t) suid; 1371 } */ *uap = v; 1372 struct proc *p = l->l_proc; 1373 struct pcred *pc = p->p_cred; 1374 uid_t ruid, euid, suid; 1375 int error; 1376 1377 ruid = SCARG(uap, ruid); 1378 euid = SCARG(uap, euid); 1379 suid = SCARG(uap, suid); 1380 1381 /* 1382 * Note: These checks are a little different than the NetBSD 1383 * setreuid(2) call performs. This precisely follows the 1384 * behavior of the Linux kernel. 1385 */ 1386 if (ruid != (uid_t)-1 && 1387 ruid != pc->p_ruid && 1388 ruid != pc->pc_ucred->cr_uid && 1389 ruid != pc->p_svuid && 1390 (error = suser(pc->pc_ucred, &p->p_acflag))) 1391 return (error); 1392 1393 if (euid != (uid_t)-1 && 1394 euid != pc->p_ruid && 1395 euid != pc->pc_ucred->cr_uid && 1396 euid != pc->p_svuid && 1397 (error = suser(pc->pc_ucred, &p->p_acflag))) 1398 return (error); 1399 1400 if (suid != (uid_t)-1 && 1401 suid != pc->p_ruid && 1402 suid != pc->pc_ucred->cr_uid && 1403 suid != pc->p_svuid && 1404 (error = suser(pc->pc_ucred, &p->p_acflag))) 1405 return (error); 1406 1407 /* 1408 * Now assign the new real, effective, and saved UIDs. 1409 * Note that Linux, unlike NetBSD in setreuid(2), does not 1410 * set the saved UID in this call unless the user specifies 1411 * it. 1412 */ 1413 if (ruid != (uid_t)-1) { 1414 (void)chgproccnt(pc->p_ruid, -1); 1415 (void)chgproccnt(ruid, 1); 1416 pc->p_ruid = ruid; 1417 } 1418 1419 if (euid != (uid_t)-1) { 1420 pc->pc_ucred = crcopy(pc->pc_ucred); 1421 pc->pc_ucred->cr_uid = euid; 1422 } 1423 1424 if (suid != (uid_t)-1) 1425 pc->p_svuid = suid; 1426 1427 if (ruid != (uid_t)-1 && euid != (uid_t)-1 && suid != (uid_t)-1) 1428 p->p_flag |= P_SUGID; 1429 return (0); 1430 } 1431 1432 int 1433 linux_sys_getresuid(l, v, retval) 1434 struct lwp *l; 1435 void *v; 1436 register_t *retval; 1437 { 1438 struct linux_sys_getresuid_args /* { 1439 syscallarg(uid_t *) ruid; 1440 syscallarg(uid_t *) euid; 1441 syscallarg(uid_t *) suid; 1442 } */ *uap = v; 1443 struct proc *p = l->l_proc; 1444 struct pcred *pc = p->p_cred; 1445 int error; 1446 1447 /* 1448 * Linux copies these values out to userspace like so: 1449 * 1450 * 1. Copy out ruid. 1451 * 2. If that succeeds, copy out euid. 1452 * 3. If both of those succeed, copy out suid. 1453 */ 1454 if ((error = copyout(&pc->p_ruid, SCARG(uap, ruid), 1455 sizeof(uid_t))) != 0) 1456 return (error); 1457 1458 if ((error = copyout(&pc->pc_ucred->cr_uid, SCARG(uap, euid), 1459 sizeof(uid_t))) != 0) 1460 return (error); 1461 1462 return (copyout(&pc->p_svuid, SCARG(uap, suid), sizeof(uid_t))); 1463 } 1464 1465 int 1466 linux_sys_ptrace(l, v, retval) 1467 struct lwp *l; 1468 void *v; 1469 register_t *retval; 1470 { 1471 struct linux_sys_ptrace_args /* { 1472 i386, m68k, powerpc: T=int 1473 alpha: T=long 1474 syscallarg(T) request; 1475 syscallarg(T) pid; 1476 syscallarg(T) addr; 1477 syscallarg(T) data; 1478 } */ *uap = v; 1479 const int *ptr; 1480 int request; 1481 int error; 1482 1483 ptr = linux_ptrace_request_map; 1484 request = SCARG(uap, request); 1485 while (*ptr != -1) 1486 if (*ptr++ == request) { 1487 struct sys_ptrace_args pta; 1488 1489 SCARG(&pta, req) = *ptr; 1490 SCARG(&pta, pid) = SCARG(uap, pid); 1491 SCARG(&pta, addr) = (caddr_t)SCARG(uap, addr); 1492 SCARG(&pta, data) = SCARG(uap, data); 1493 1494 /* 1495 * Linux ptrace(PTRACE_CONT, pid, 0, 0) means actually 1496 * to continue where the process left off previously. 1497 * The same thing is achieved by addr == (caddr_t) 1 1498 * on NetBSD, so rewrite 'addr' appropriately. 1499 */ 1500 if (request == LINUX_PTRACE_CONT && SCARG(uap, addr)==0) 1501 SCARG(&pta, addr) = (caddr_t) 1; 1502 1503 error = sys_ptrace(l, &pta, retval); 1504 if (error) 1505 return error; 1506 switch (request) { 1507 case LINUX_PTRACE_PEEKTEXT: 1508 case LINUX_PTRACE_PEEKDATA: 1509 error = copyout (retval, 1510 (caddr_t)SCARG(uap, data), sizeof *retval); 1511 *retval = SCARG(uap, data); 1512 break; 1513 default: 1514 break; 1515 } 1516 return error; 1517 } 1518 else 1519 ptr++; 1520 1521 return LINUX_SYS_PTRACE_ARCH(l, uap, retval); 1522 } 1523 1524 int 1525 linux_sys_reboot(struct lwp *l, void *v, register_t *retval) 1526 { 1527 struct linux_sys_reboot_args /* { 1528 syscallarg(int) magic1; 1529 syscallarg(int) magic2; 1530 syscallarg(int) cmd; 1531 syscallarg(void *) arg; 1532 } */ *uap = v; 1533 struct sys_reboot_args /* { 1534 syscallarg(int) opt; 1535 syscallarg(char *) bootstr; 1536 } */ sra; 1537 struct proc *p = l->l_proc; 1538 int error; 1539 1540 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 1541 return(error); 1542 1543 if (SCARG(uap, magic1) != LINUX_REBOOT_MAGIC1) 1544 return(EINVAL); 1545 if (SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2 && 1546 SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2A && 1547 SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2B) 1548 return(EINVAL); 1549 1550 switch (SCARG(uap, cmd)) { 1551 case LINUX_REBOOT_CMD_RESTART: 1552 SCARG(&sra, opt) = RB_AUTOBOOT; 1553 break; 1554 case LINUX_REBOOT_CMD_HALT: 1555 SCARG(&sra, opt) = RB_HALT; 1556 break; 1557 case LINUX_REBOOT_CMD_POWER_OFF: 1558 SCARG(&sra, opt) = RB_HALT|RB_POWERDOWN; 1559 break; 1560 case LINUX_REBOOT_CMD_RESTART2: 1561 /* Reboot with an argument. */ 1562 SCARG(&sra, opt) = RB_AUTOBOOT|RB_STRING; 1563 SCARG(&sra, bootstr) = SCARG(uap, arg); 1564 break; 1565 case LINUX_REBOOT_CMD_CAD_ON: 1566 return(EINVAL); /* We don't implement ctrl-alt-delete */ 1567 case LINUX_REBOOT_CMD_CAD_OFF: 1568 return(0); 1569 default: 1570 return(EINVAL); 1571 } 1572 1573 return(sys_reboot(l, &sra, retval)); 1574 } 1575 1576 /* 1577 * Copy of compat_12_sys_swapon(). 1578 */ 1579 int 1580 linux_sys_swapon(l, v, retval) 1581 struct lwp *l; 1582 void *v; 1583 register_t *retval; 1584 { 1585 struct sys_swapctl_args ua; 1586 struct linux_sys_swapon_args /* { 1587 syscallarg(const char *) name; 1588 } */ *uap = v; 1589 1590 SCARG(&ua, cmd) = SWAP_ON; 1591 SCARG(&ua, arg) = (void *)SCARG(uap, name); 1592 SCARG(&ua, misc) = 0; /* priority */ 1593 return (sys_swapctl(l, &ua, retval)); 1594 } 1595 1596 /* 1597 * Stop swapping to the file or block device specified by path. 1598 */ 1599 int 1600 linux_sys_swapoff(l, v, retval) 1601 struct lwp *l; 1602 void *v; 1603 register_t *retval; 1604 { 1605 struct sys_swapctl_args ua; 1606 struct linux_sys_swapoff_args /* { 1607 syscallarg(const char *) path; 1608 } */ *uap = v; 1609 1610 SCARG(&ua, cmd) = SWAP_OFF; 1611 SCARG(&ua, arg) = (void *)SCARG(uap, path); 1612 return (sys_swapctl(l, &ua, retval)); 1613 } 1614 1615 /* 1616 * Copy of compat_09_sys_setdomainname() 1617 */ 1618 /* ARGSUSED */ 1619 int 1620 linux_sys_setdomainname(l, v, retval) 1621 struct lwp *l; 1622 void *v; 1623 register_t *retval; 1624 { 1625 struct linux_sys_setdomainname_args /* { 1626 syscallarg(char *) domainname; 1627 syscallarg(int) len; 1628 } */ *uap = v; 1629 struct proc *p = l->l_proc; 1630 int name; 1631 int error; 1632 1633 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 1634 return (error); 1635 name = KERN_DOMAINNAME; 1636 return (kern_sysctl(&name, 1, 0, 0, SCARG(uap, domainname), 1637 SCARG(uap, len), p)); 1638 } 1639 1640 /* 1641 * sysinfo() 1642 */ 1643 /* ARGSUSED */ 1644 int 1645 linux_sys_sysinfo(l, v, retval) 1646 struct lwp *l; 1647 void *v; 1648 register_t *retval; 1649 { 1650 struct linux_sys_sysinfo_args /* { 1651 syscallarg(struct linux_sysinfo *) arg; 1652 } */ *uap = v; 1653 struct linux_sysinfo si; 1654 struct loadavg *la; 1655 1656 si.uptime = time.tv_sec - boottime.tv_sec; 1657 la = &averunnable; 1658 si.loads[0] = la->ldavg[0] * LINUX_SYSINFO_LOADS_SCALE / la->fscale; 1659 si.loads[1] = la->ldavg[1] * LINUX_SYSINFO_LOADS_SCALE / la->fscale; 1660 si.loads[2] = la->ldavg[2] * LINUX_SYSINFO_LOADS_SCALE / la->fscale; 1661 si.totalram = ctob(physmem); 1662 si.freeram = uvmexp.free * uvmexp.pagesize; 1663 si.sharedram = 0; /* XXX */ 1664 si.bufferram = uvmexp.filepages * uvmexp.pagesize; 1665 si.totalswap = uvmexp.swpages * uvmexp.pagesize; 1666 si.freeswap = (uvmexp.swpages - uvmexp.swpginuse) * uvmexp.pagesize; 1667 si.procs = nprocs; 1668 1669 /* The following are only present in newer Linux kernels. */ 1670 si.totalbig = 0; 1671 si.freebig = 0; 1672 si.mem_unit = 1; 1673 1674 return (copyout(&si, SCARG(uap, arg), sizeof si)); 1675 } 1676 1677 #define bsd_to_linux_rlimit1(l, b, f) \ 1678 (l)->f = ((b)->f == RLIM_INFINITY || ((b)->f & 0xffffffff00000000) != 0) ? \ 1679 LINUX_RLIM_INFINITY : (int32_t)(b)->f 1680 #define bsd_to_linux_rlimit(l, b) \ 1681 bsd_to_linux_rlimit1(l, b, rlim_cur); \ 1682 bsd_to_linux_rlimit1(l, b, rlim_max) 1683 1684 #define linux_to_bsd_rlimit1(b, l, f) \ 1685 (b)->f = (l)->f == LINUX_RLIM_INFINITY ? RLIM_INFINITY : (b)->f 1686 #define linux_to_bsd_rlimit(b, l) \ 1687 linux_to_bsd_rlimit1(b, l, rlim_cur); \ 1688 linux_to_bsd_rlimit1(b, l, rlim_max) 1689 1690 static int 1691 linux_to_bsd_limit(lim) 1692 int lim; 1693 { 1694 switch (lim) { 1695 case LINUX_RLIMIT_CPU: 1696 return RLIMIT_CPU; 1697 case LINUX_RLIMIT_FSIZE: 1698 return RLIMIT_FSIZE; 1699 case LINUX_RLIMIT_DATA: 1700 return RLIMIT_DATA; 1701 case LINUX_RLIMIT_STACK: 1702 return RLIMIT_STACK; 1703 case LINUX_RLIMIT_CORE: 1704 return RLIMIT_CORE; 1705 case LINUX_RLIMIT_RSS: 1706 return RLIMIT_RSS; 1707 case LINUX_RLIMIT_NPROC: 1708 return RLIMIT_NPROC; 1709 case LINUX_RLIMIT_NOFILE: 1710 return RLIMIT_NOFILE; 1711 case LINUX_RLIMIT_MEMLOCK: 1712 return RLIMIT_MEMLOCK; 1713 case LINUX_RLIMIT_AS: 1714 case LINUX_RLIMIT_LOCKS: 1715 return -EOPNOTSUPP; 1716 default: 1717 return -EINVAL; 1718 } 1719 } 1720 1721 1722 int 1723 linux_sys_getrlimit(l, v, retval) 1724 struct lwp *l; 1725 void *v; 1726 register_t *retval; 1727 { 1728 struct linux_sys_getrlimit_args /* { 1729 syscallarg(int) which; 1730 syscallarg(struct orlimit *) rlp; 1731 } */ *uap = v; 1732 struct proc *p = l->l_proc; 1733 caddr_t sg = stackgap_init(p, 0); 1734 struct sys_getrlimit_args ap; 1735 struct rlimit rl; 1736 struct orlimit orl; 1737 int error; 1738 1739 SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which)); 1740 if ((error = SCARG(&ap, which)) < 0) 1741 return -error; 1742 SCARG(&ap, rlp) = stackgap_alloc(p, &sg, sizeof rl); 1743 if ((error = sys_getrlimit(l, &ap, retval)) != 0) 1744 return error; 1745 if ((error = copyin(SCARG(&ap, rlp), &rl, sizeof(rl))) != 0) 1746 return error; 1747 bsd_to_linux_rlimit(&orl, &rl); 1748 return copyout(&orl, SCARG(uap, rlp), sizeof(orl)); 1749 } 1750 1751 int 1752 linux_sys_setrlimit(l, v, retval) 1753 struct lwp *l; 1754 void *v; 1755 register_t *retval; 1756 { 1757 struct linux_sys_setrlimit_args /* { 1758 syscallarg(int) which; 1759 syscallarg(struct orlimit *) rlp; 1760 } */ *uap = v; 1761 struct proc *p = l->l_proc; 1762 caddr_t sg = stackgap_init(p, 0); 1763 struct sys_setrlimit_args ap; 1764 struct rlimit rl; 1765 struct orlimit orl; 1766 int error; 1767 1768 SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which)); 1769 SCARG(&ap, rlp) = stackgap_alloc(p, &sg, sizeof rl); 1770 if ((error = SCARG(&ap, which)) < 0) 1771 return -error; 1772 if ((error = copyin(SCARG(uap, rlp), &orl, sizeof(orl))) != 0) 1773 return error; 1774 linux_to_bsd_rlimit(&rl, &orl); 1775 /* XXX: alpha complains about this */ 1776 if ((error = copyout(&rl, (void *)SCARG(&ap, rlp), sizeof(rl))) != 0) 1777 return error; 1778 return sys_setrlimit(l, &ap, retval); 1779 } 1780 1781 #ifndef __mips__ 1782 /* XXX: this doesn't look 100% common, at least mips doesn't have it */ 1783 int 1784 linux_sys_ugetrlimit(l, v, retval) 1785 struct lwp *l; 1786 void *v; 1787 register_t *retval; 1788 { 1789 return linux_sys_getrlimit(l, v, retval); 1790 } 1791 #endif 1792 1793 /* 1794 * This gets called for unsupported syscalls. The difference to sys_nosys() 1795 * is that process does not get SIGSYS, the call just returns with ENOSYS. 1796 * This is the way Linux does it and glibc depends on this behaviour. 1797 */ 1798 int 1799 linux_sys_nosys(l, v, retval) 1800 struct lwp *l; 1801 void *v; 1802 register_t *retval; 1803 { 1804 return (ENOSYS); 1805 } 1806
Indexes created Mon Sep 22 13:09:51 GMT 2025