linux_misc.c revision 1.121.2.7
1 /* $NetBSD: linux_misc.c,v 1.121.2.7 2004/09/24 10:53:18 skrll 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 1.121.2.7 2004/09/24 10:53:18 skrll Exp $"); 68 69 #include <sys/param.h> 70 #include <sys/systm.h> 71 #include <sys/namei.h> 72 #include <sys/proc.h> 73 #include <sys/dirent.h> 74 #include <sys/file.h> 75 #include <sys/stat.h> 76 #include <sys/filedesc.h> 77 #include <sys/ioctl.h> 78 #include <sys/kernel.h> 79 #include <sys/malloc.h> 80 #include <sys/mbuf.h> 81 #include <sys/mman.h> 82 #include <sys/mount.h> 83 #include <sys/reboot.h> 84 #include <sys/resource.h> 85 #include <sys/resourcevar.h> 86 #include <sys/signal.h> 87 #include <sys/signalvar.h> 88 #include <sys/socket.h> 89 #include <sys/time.h> 90 #include <sys/times.h> 91 #include <sys/vnode.h> 92 #include <sys/uio.h> 93 #include <sys/wait.h> 94 #include <sys/utsname.h> 95 #include <sys/unistd.h> 96 #include <sys/swap.h> /* for SWAP_ON */ 97 #include <sys/sysctl.h> /* for KERN_DOMAINNAME */ 98 99 #include <sys/ptrace.h> 100 #include <machine/ptrace.h> 101 102 #include <sys/sa.h> 103 #include <sys/syscallargs.h> 104 105 #include <compat/linux/common/linux_types.h> 106 #include <compat/linux/common/linux_signal.h> 107 108 #include <compat/linux/linux_syscallargs.h> 109 110 #include <compat/linux/common/linux_fcntl.h> 111 #include <compat/linux/common/linux_mmap.h> 112 #include <compat/linux/common/linux_dirent.h> 113 #include <compat/linux/common/linux_util.h> 114 #include <compat/linux/common/linux_misc.h> 115 #include <compat/linux/common/linux_ptrace.h> 116 #include <compat/linux/common/linux_reboot.h> 117 #include <compat/linux/common/linux_emuldata.h> 118 119 const int linux_ptrace_request_map[] = { 120 LINUX_PTRACE_TRACEME, PT_TRACE_ME, 121 LINUX_PTRACE_PEEKTEXT, PT_READ_I, 122 LINUX_PTRACE_PEEKDATA, PT_READ_D, 123 LINUX_PTRACE_POKETEXT, PT_WRITE_I, 124 LINUX_PTRACE_POKEDATA, PT_WRITE_D, 125 LINUX_PTRACE_CONT, PT_CONTINUE, 126 LINUX_PTRACE_KILL, PT_KILL, 127 LINUX_PTRACE_ATTACH, PT_ATTACH, 128 LINUX_PTRACE_DETACH, PT_DETACH, 129 #ifdef PT_STEP 130 LINUX_PTRACE_SINGLESTEP, PT_STEP, 131 #endif 132 -1 133 }; 134 135 const struct linux_mnttypes linux_fstypes[] = { 136 { MOUNT_FFS, LINUX_DEFAULT_SUPER_MAGIC }, 137 { MOUNT_NFS, LINUX_NFS_SUPER_MAGIC }, 138 { MOUNT_MFS, LINUX_DEFAULT_SUPER_MAGIC }, 139 { MOUNT_MSDOS, LINUX_MSDOS_SUPER_MAGIC }, 140 { MOUNT_LFS, LINUX_DEFAULT_SUPER_MAGIC }, 141 { MOUNT_FDESC, LINUX_DEFAULT_SUPER_MAGIC }, 142 { MOUNT_PORTAL, LINUX_DEFAULT_SUPER_MAGIC }, 143 { MOUNT_NULL, LINUX_DEFAULT_SUPER_MAGIC }, 144 { MOUNT_OVERLAY, LINUX_DEFAULT_SUPER_MAGIC }, 145 { MOUNT_UMAP, LINUX_DEFAULT_SUPER_MAGIC }, 146 { MOUNT_KERNFS, LINUX_DEFAULT_SUPER_MAGIC }, 147 { MOUNT_PROCFS, LINUX_PROC_SUPER_MAGIC }, 148 { MOUNT_AFS, LINUX_DEFAULT_SUPER_MAGIC }, 149 { MOUNT_CD9660, LINUX_ISOFS_SUPER_MAGIC }, 150 { MOUNT_UNION, LINUX_DEFAULT_SUPER_MAGIC }, 151 { MOUNT_ADOSFS, LINUX_ADFS_SUPER_MAGIC }, 152 { MOUNT_EXT2FS, LINUX_EXT2_SUPER_MAGIC }, 153 { MOUNT_CFS, LINUX_DEFAULT_SUPER_MAGIC }, 154 { MOUNT_CODA, LINUX_CODA_SUPER_MAGIC }, 155 { MOUNT_FILECORE, LINUX_DEFAULT_SUPER_MAGIC }, 156 { MOUNT_NTFS, LINUX_DEFAULT_SUPER_MAGIC }, 157 { MOUNT_SMBFS, LINUX_SMB_SUPER_MAGIC } 158 }; 159 const int linux_fstypes_cnt = sizeof(linux_fstypes) / sizeof(linux_fstypes[0]); 160 161 #ifdef DEBUG_LINUX 162 #define DPRINTF(a) uprintf a 163 #else 164 #define DPRINTF(a) 165 #endif 166 167 /* Local linux_misc.c functions: */ 168 static void bsd_to_linux_statfs __P((const struct statvfs *, 169 struct linux_statfs *)); 170 static int linux_to_bsd_limit __P((int)); 171 static void linux_to_bsd_mmap_args __P((struct sys_mmap_args *, 172 const struct linux_sys_mmap_args *)); 173 static int linux_mmap __P((struct lwp *, struct linux_sys_mmap_args *, 174 register_t *, off_t)); 175 176 177 /* 178 * The information on a terminated (or stopped) process needs 179 * to be converted in order for Linux binaries to get a valid signal 180 * number out of it. 181 */ 182 void 183 bsd_to_linux_wstat(st) 184 int *st; 185 { 186 187 int sig; 188 189 if (WIFSIGNALED(*st)) { 190 sig = WTERMSIG(*st); 191 if (sig >= 0 && sig < NSIG) 192 *st= (*st& ~0177) | native_to_linux_signo[sig]; 193 } else if (WIFSTOPPED(*st)) { 194 sig = WSTOPSIG(*st); 195 if (sig >= 0 && sig < NSIG) 196 *st = (*st & ~0xff00) | 197 (native_to_linux_signo[sig] << 8); 198 } 199 } 200 201 /* 202 * This is very much the same as waitpid() 203 */ 204 int 205 linux_sys_wait4(l, v, retval) 206 struct lwp *l; 207 void *v; 208 register_t *retval; 209 { 210 struct linux_sys_wait4_args /* { 211 syscallarg(int) pid; 212 syscallarg(int *) status; 213 syscallarg(int) options; 214 syscallarg(struct rusage *) rusage; 215 } */ *uap = v; 216 struct proc *p = l->l_proc; 217 struct sys_wait4_args w4a; 218 int error, *status, tstat, options, linux_options; 219 caddr_t sg; 220 221 if (SCARG(uap, status) != NULL) { 222 sg = stackgap_init(p, 0); 223 status = (int *) stackgap_alloc(p, &sg, sizeof *status); 224 } else 225 status = NULL; 226 227 linux_options = SCARG(uap, options); 228 options = 0; 229 if (linux_options & 230 ~(LINUX_WAIT4_WNOHANG|LINUX_WAIT4_WUNTRACED|LINUX_WAIT4_WALL| 231 LINUX_WAIT4_WCLONE)) 232 return (EINVAL); 233 234 if (linux_options & LINUX_WAIT4_WNOHANG) 235 options |= WNOHANG; 236 if (linux_options & LINUX_WAIT4_WUNTRACED) 237 options |= WUNTRACED; 238 if (linux_options & LINUX_WAIT4_WALL) 239 options |= WALLSIG; 240 if (linux_options & LINUX_WAIT4_WCLONE) 241 options |= WALTSIG; 242 243 SCARG(&w4a, pid) = SCARG(uap, pid); 244 SCARG(&w4a, status) = status; 245 SCARG(&w4a, options) = options; 246 SCARG(&w4a, rusage) = SCARG(uap, rusage); 247 248 if ((error = sys_wait4(l, &w4a, retval))) 249 return error; 250 251 sigdelset(&p->p_sigctx.ps_siglist, SIGCHLD); 252 253 if (status != NULL) { 254 if ((error = copyin(status, &tstat, sizeof tstat))) 255 return error; 256 257 bsd_to_linux_wstat(&tstat); 258 return copyout(&tstat, SCARG(uap, status), sizeof tstat); 259 } 260 261 return 0; 262 } 263 264 /* 265 * Linux brk(2). The check if the new address is >= the old one is 266 * done in the kernel in Linux. NetBSD does it in the library. 267 */ 268 int 269 linux_sys_brk(l, v, retval) 270 struct lwp *l; 271 void *v; 272 register_t *retval; 273 { 274 struct linux_sys_brk_args /* { 275 syscallarg(char *) nsize; 276 } */ *uap = v; 277 struct proc *p = l->l_proc; 278 char *nbrk = SCARG(uap, nsize); 279 struct sys_obreak_args oba; 280 struct vmspace *vm = p->p_vmspace; 281 struct linux_emuldata *ed = (struct linux_emuldata*)p->p_emuldata; 282 283 SCARG(&oba, nsize) = nbrk; 284 285 if ((caddr_t) nbrk > vm->vm_daddr && sys_obreak(l, &oba, retval) == 0) 286 ed->s->p_break = (char*)nbrk; 287 else 288 nbrk = ed->s->p_break; 289 290 retval[0] = (register_t)nbrk; 291 292 return 0; 293 } 294 295 /* 296 * Convert NetBSD statvfs structure to Linux statfs structure. 297 * Linux doesn't have f_flag, and we can't set f_frsize due 298 * to glibc statvfs() bug (see below). 299 */ 300 static void 301 bsd_to_linux_statfs(bsp, lsp) 302 const struct statvfs *bsp; 303 struct linux_statfs *lsp; 304 { 305 int i; 306 307 for (i = 0; i < linux_fstypes_cnt; i++) { 308 if (strcmp(bsp->f_fstypename, linux_fstypes[i].bsd) == 0) { 309 lsp->l_ftype = linux_fstypes[i].linux; 310 break; 311 } 312 } 313 314 if (i == linux_fstypes_cnt) { 315 DPRINTF(("unhandled fstype in linux emulation: %s\n", 316 bsp->f_fstypename)); 317 lsp->l_ftype = LINUX_DEFAULT_SUPER_MAGIC; 318 } 319 320 /* 321 * The sizes are expressed in number of blocks. The block 322 * size used for the size is f_frsize for POSIX-compliant 323 * statvfs. Linux statfs uses f_bsize as the block size 324 * (f_frsize used to not be available in Linux struct statfs). 325 * However, glibc 2.3.3 statvfs() wrapper fails to adjust the block 326 * counts for different f_frsize if f_frsize is provided by the kernel. 327 * POSIX conforming apps thus get wrong size if f_frsize 328 * is different to f_bsize. Thus, we just pretend we don't 329 * support f_frsize. 330 */ 331 332 lsp->l_fbsize = bsp->f_frsize; 333 lsp->l_ffrsize = 0; /* compat */ 334 lsp->l_fblocks = bsp->f_blocks; 335 lsp->l_fbfree = bsp->f_bfree; 336 lsp->l_fbavail = bsp->f_bavail; 337 lsp->l_ffiles = bsp->f_files; 338 lsp->l_fffree = bsp->f_ffree; 339 /* Linux sets the fsid to 0..., we don't */ 340 lsp->l_ffsid.val[0] = bsp->f_fsidx.__fsid_val[0]; 341 lsp->l_ffsid.val[1] = bsp->f_fsidx.__fsid_val[1]; 342 lsp->l_fnamelen = bsp->f_namemax; 343 (void)memset(lsp->l_fspare, 0, sizeof(lsp->l_fspare)); 344 } 345 346 /* 347 * Implement the fs stat functions. Straightforward. 348 */ 349 int 350 linux_sys_statfs(l, v, retval) 351 struct lwp *l; 352 void *v; 353 register_t *retval; 354 { 355 struct linux_sys_statfs_args /* { 356 syscallarg(const char *) path; 357 syscallarg(struct linux_statfs *) sp; 358 } */ *uap = v; 359 struct proc *p = l->l_proc; 360 struct statvfs btmp, *bsp; 361 struct linux_statfs ltmp; 362 struct sys_statvfs1_args bsa; 363 caddr_t sg; 364 int error; 365 366 sg = stackgap_init(p, 0); 367 bsp = (struct statvfs *) stackgap_alloc(p, &sg, sizeof (struct statvfs)); 368 369 CHECK_ALT_EXIST(l, &sg, SCARG(uap, path)); 370 371 SCARG(&bsa, path) = SCARG(uap, path); 372 SCARG(&bsa, buf) = bsp; 373 SCARG(&bsa, flags) = ST_WAIT; 374 375 if ((error = sys_statvfs1(l, &bsa, retval))) 376 return error; 377 378 if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp))) 379 return error; 380 381 bsd_to_linux_statfs(&btmp, <mp); 382 383 return copyout((caddr_t) <mp, (caddr_t) SCARG(uap, sp), sizeof ltmp); 384 } 385 386 int 387 linux_sys_fstatfs(l, v, retval) 388 struct lwp *l; 389 void *v; 390 register_t *retval; 391 { 392 struct linux_sys_fstatfs_args /* { 393 syscallarg(int) fd; 394 syscallarg(struct linux_statfs *) sp; 395 } */ *uap = v; 396 struct proc *p = l->l_proc; 397 struct statvfs btmp, *bsp; 398 struct linux_statfs ltmp; 399 struct sys_fstatvfs1_args bsa; 400 caddr_t sg; 401 int error; 402 403 sg = stackgap_init(p, 0); 404 bsp = (struct statvfs *) stackgap_alloc(p, &sg, sizeof (struct statvfs)); 405 406 SCARG(&bsa, fd) = SCARG(uap, fd); 407 SCARG(&bsa, buf) = bsp; 408 SCARG(&bsa, flags) = ST_WAIT; 409 410 if ((error = sys_fstatvfs1(l, &bsa, retval))) 411 return error; 412 413 if ((error = copyin((caddr_t) bsp, (caddr_t) &btmp, sizeof btmp))) 414 return error; 415 416 bsd_to_linux_statfs(&btmp, <mp); 417 418 return copyout((caddr_t) <mp, (caddr_t) SCARG(uap, sp), sizeof ltmp); 419 } 420 421 /* 422 * uname(). Just copy the info from the various strings stored in the 423 * kernel, and put it in the Linux utsname structure. That structure 424 * is almost the same as the NetBSD one, only it has fields 65 characters 425 * long, and an extra domainname field. 426 */ 427 int 428 linux_sys_uname(l, v, retval) 429 struct lwp *l; 430 void *v; 431 register_t *retval; 432 { 433 struct linux_sys_uname_args /* { 434 syscallarg(struct linux_utsname *) up; 435 } */ *uap = v; 436 struct linux_utsname luts; 437 438 strncpy(luts.l_sysname, linux_sysname, sizeof(luts.l_sysname)); 439 strncpy(luts.l_nodename, hostname, sizeof(luts.l_nodename)); 440 strncpy(luts.l_release, linux_release, sizeof(luts.l_release)); 441 strncpy(luts.l_version, linux_version, sizeof(luts.l_version)); 442 strncpy(luts.l_machine, machine, sizeof(luts.l_machine)); 443 strncpy(luts.l_domainname, domainname, sizeof(luts.l_domainname)); 444 445 return copyout(&luts, SCARG(uap, up), sizeof(luts)); 446 } 447 448 /* Used directly on: alpha, mips, ppc, sparc, sparc64 */ 449 /* Used indirectly on: arm, i386, m68k */ 450 451 /* 452 * New type Linux mmap call. 453 * Only called directly on machines with >= 6 free regs. 454 */ 455 int 456 linux_sys_mmap(l, v, retval) 457 struct lwp *l; 458 void *v; 459 register_t *retval; 460 { 461 struct linux_sys_mmap_args /* { 462 syscallarg(unsigned long) addr; 463 syscallarg(size_t) len; 464 syscallarg(int) prot; 465 syscallarg(int) flags; 466 syscallarg(int) fd; 467 syscallarg(linux_off_t) offset; 468 } */ *uap = v; 469 470 if (SCARG(uap, offset) & PAGE_MASK) 471 return EINVAL; 472 473 return linux_mmap(l, uap, retval, SCARG(uap, offset)); 474 } 475 476 /* 477 * Guts of most architectures' mmap64() implementations. This shares 478 * its list of arguments with linux_sys_mmap(). 479 * 480 * The difference in linux_sys_mmap2() is that "offset" is actually 481 * (offset / pagesize), not an absolute byte count. This translation 482 * to pagesize offsets is done inside glibc between the mmap64() call 483 * point, and the actual syscall. 484 */ 485 int 486 linux_sys_mmap2(l, v, retval) 487 struct lwp *l; 488 void *v; 489 register_t *retval; 490 { 491 struct linux_sys_mmap2_args /* { 492 syscallarg(unsigned long) addr; 493 syscallarg(size_t) len; 494 syscallarg(int) prot; 495 syscallarg(int) flags; 496 syscallarg(int) fd; 497 syscallarg(linux_off_t) offset; 498 } */ *uap = v; 499 500 return linux_mmap(l, uap, retval, 501 ((off_t)SCARG(uap, offset)) << PAGE_SHIFT); 502 } 503 504 /* 505 * Massage arguments and call system mmap(2). 506 */ 507 static int 508 linux_mmap(l, uap, retval, offset) 509 struct lwp *l; 510 struct linux_sys_mmap_args *uap; 511 register_t *retval; 512 off_t offset; 513 { 514 struct sys_mmap_args cma; 515 int error; 516 size_t mmoff=0; 517 518 if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN) { 519 /* 520 * Request for stack-like memory segment. On linux, this 521 * works by mmap()ping (small) segment, which is automatically 522 * extended when page fault happens below the currently 523 * allocated area. We emulate this by allocating (typically 524 * bigger) segment sized at current stack size limit, and 525 * offsetting the requested and returned address accordingly. 526 * Since physical pages are only allocated on-demand, this 527 * is effectively identical. 528 */ 529 rlim_t ssl = l->l_proc->p_rlimit[RLIMIT_STACK].rlim_cur; 530 531 if (SCARG(uap, len) < ssl) { 532 /* Compute the address offset */ 533 mmoff = round_page(ssl) - SCARG(uap, len); 534 535 if (SCARG(uap, addr)) 536 SCARG(uap, addr) -= mmoff; 537 538 SCARG(uap, len) = (size_t) ssl; 539 } 540 } 541 542 linux_to_bsd_mmap_args(&cma, uap); 543 SCARG(&cma, pos) = offset; 544 545 error = sys_mmap(l, &cma, retval); 546 if (error) 547 return (error); 548 549 /* Shift the returned address for stack-like segment if necessary */ 550 if (SCARG(uap, flags) & LINUX_MAP_GROWSDOWN && mmoff) 551 retval[0] += mmoff; 552 553 return (0); 554 } 555 556 static void 557 linux_to_bsd_mmap_args(cma, uap) 558 struct sys_mmap_args *cma; 559 const struct linux_sys_mmap_args *uap; 560 { 561 int flags = MAP_TRYFIXED, fl = SCARG(uap, flags); 562 563 flags |= cvtto_bsd_mask(fl, LINUX_MAP_SHARED, MAP_SHARED); 564 flags |= cvtto_bsd_mask(fl, LINUX_MAP_PRIVATE, MAP_PRIVATE); 565 flags |= cvtto_bsd_mask(fl, LINUX_MAP_FIXED, MAP_FIXED); 566 flags |= cvtto_bsd_mask(fl, LINUX_MAP_ANON, MAP_ANON); 567 /* XXX XAX ERH: Any other flags here? There are more defined... */ 568 569 SCARG(cma, addr) = (void *)SCARG(uap, addr); 570 SCARG(cma, len) = SCARG(uap, len); 571 SCARG(cma, prot) = SCARG(uap, prot); 572 if (SCARG(cma, prot) & VM_PROT_WRITE) /* XXX */ 573 SCARG(cma, prot) |= VM_PROT_READ; 574 SCARG(cma, flags) = flags; 575 SCARG(cma, fd) = flags & MAP_ANON ? -1 : SCARG(uap, fd); 576 SCARG(cma, pad) = 0; 577 } 578 579 int 580 linux_sys_mremap(l, v, retval) 581 struct lwp *l; 582 void *v; 583 register_t *retval; 584 { 585 struct linux_sys_mremap_args /* { 586 syscallarg(void *) old_address; 587 syscallarg(size_t) old_size; 588 syscallarg(size_t) new_size; 589 syscallarg(u_long) flags; 590 } */ *uap = v; 591 struct sys_munmap_args mua; 592 size_t old_size, new_size; 593 int error; 594 595 old_size = round_page(SCARG(uap, old_size)); 596 new_size = round_page(SCARG(uap, new_size)); 597 598 /* 599 * Growing mapped region. 600 */ 601 if (new_size > old_size) { 602 /* 603 * XXX Implement me. What we probably want to do is 604 * XXX dig out the guts of the old mapping, mmap that 605 * XXX object again with the new size, then munmap 606 * XXX the old mapping. 607 */ 608 *retval = 0; 609 return (ENOMEM); 610 } 611 612 /* 613 * Shrinking mapped region. 614 */ 615 if (new_size < old_size) { 616 SCARG(&mua, addr) = (caddr_t)SCARG(uap, old_address) + 617 new_size; 618 SCARG(&mua, len) = old_size - new_size; 619 error = sys_munmap(l, &mua, retval); 620 *retval = error ? 0 : (register_t)SCARG(uap, old_address); 621 return (error); 622 } 623 624 /* 625 * No change. 626 */ 627 *retval = (register_t)SCARG(uap, old_address); 628 return (0); 629 } 630 631 int 632 linux_sys_msync(l, v, retval) 633 struct lwp *l; 634 void *v; 635 register_t *retval; 636 { 637 struct linux_sys_msync_args /* { 638 syscallarg(caddr_t) addr; 639 syscallarg(int) len; 640 syscallarg(int) fl; 641 } */ *uap = v; 642 643 struct sys___msync13_args bma; 644 645 /* flags are ignored */ 646 SCARG(&bma, addr) = SCARG(uap, addr); 647 SCARG(&bma, len) = SCARG(uap, len); 648 SCARG(&bma, flags) = SCARG(uap, fl); 649 650 return sys___msync13(l, &bma, retval); 651 } 652 653 int 654 linux_sys_mprotect(l, v, retval) 655 struct lwp *l; 656 void *v; 657 register_t *retval; 658 { 659 struct linux_sys_mprotect_args /* { 660 syscallarg(const void *) start; 661 syscallarg(unsigned long) len; 662 syscallarg(int) prot; 663 } */ *uap = v; 664 unsigned long end, start = (unsigned long)SCARG(uap, start), len; 665 int prot = SCARG(uap, prot); 666 struct vm_map_entry *entry; 667 struct vm_map *map = &l->l_proc->p_vmspace->vm_map; 668 669 if (start & PAGE_MASK) 670 return EINVAL; 671 672 len = round_page(SCARG(uap, len)); 673 end = start + len; 674 675 if (end < start) 676 return EINVAL; 677 else if (end == start) 678 return 0; 679 680 if (SCARG(uap, prot) & ~(PROT_READ | PROT_WRITE | PROT_EXEC)) 681 return EINVAL; 682 683 vm_map_lock(map); 684 #ifdef notdef 685 VM_MAP_RANGE_CHECK(map, start, end); 686 #endif 687 if (!uvm_map_lookup_entry(map, start, &entry) || entry->start > start) { 688 vm_map_unlock(map); 689 return ENOMEM; 690 } 691 vm_map_unlock(map); 692 return uvm_map_protect(map, start, end, prot, FALSE); 693 } 694 695 /* 696 * This code is partly stolen from src/lib/libc/compat-43/times.c 697 */ 698 699 #define CONVTCK(r) (r.tv_sec * hz + r.tv_usec / (1000000 / hz)) 700 701 int 702 linux_sys_times(l, v, retval) 703 struct lwp *l; 704 void *v; 705 register_t *retval; 706 { 707 struct linux_sys_times_args /* { 708 syscallarg(struct times *) tms; 709 } */ *uap = v; 710 struct proc *p = l->l_proc; 711 struct timeval t; 712 int error, s; 713 714 if (SCARG(uap, tms)) { 715 struct linux_tms ltms; 716 struct rusage ru; 717 718 calcru(p, &ru.ru_utime, &ru.ru_stime, NULL); 719 ltms.ltms_utime = CONVTCK(ru.ru_utime); 720 ltms.ltms_stime = CONVTCK(ru.ru_stime); 721 722 ltms.ltms_cutime = CONVTCK(p->p_stats->p_cru.ru_utime); 723 ltms.ltms_cstime = CONVTCK(p->p_stats->p_cru.ru_stime); 724 725 if ((error = copyout(<ms, SCARG(uap, tms), sizeof ltms))) 726 return error; 727 } 728 729 s = splclock(); 730 timersub(&time, &boottime, &t); 731 splx(s); 732 733 retval[0] = ((linux_clock_t)(CONVTCK(t))); 734 return 0; 735 } 736 737 #undef CONVTCK 738 739 /* 740 * Linux 'readdir' call. This code is mostly taken from the 741 * SunOS getdents call (see compat/sunos/sunos_misc.c), though 742 * an attempt has been made to keep it a little cleaner (failing 743 * miserably, because of the cruft needed if count 1 is passed). 744 * 745 * The d_off field should contain the offset of the next valid entry, 746 * but in Linux it has the offset of the entry itself. We emulate 747 * that bug here. 748 * 749 * Read in BSD-style entries, convert them, and copy them out. 750 * 751 * Note that this doesn't handle union-mounted filesystems. 752 */ 753 int 754 linux_sys_getdents(l, v, retval) 755 struct lwp *l; 756 void *v; 757 register_t *retval; 758 { 759 struct linux_sys_getdents_args /* { 760 syscallarg(int) fd; 761 syscallarg(struct linux_dirent *) dent; 762 syscallarg(unsigned int) count; 763 } */ *uap = v; 764 struct proc *p = l->l_proc; 765 struct dirent *bdp; 766 struct vnode *vp; 767 caddr_t inp, buf; /* BSD-format */ 768 int len, reclen; /* BSD-format */ 769 caddr_t outp; /* Linux-format */ 770 int resid, linux_reclen = 0; /* Linux-format */ 771 struct file *fp; 772 struct uio auio; 773 struct iovec aiov; 774 struct linux_dirent idb; 775 off_t off; /* true file offset */ 776 int buflen, error, eofflag, nbytes, oldcall; 777 struct vattr va; 778 off_t *cookiebuf = NULL, *cookie; 779 int ncookies; 780 781 /* getvnode() will use the descriptor for us */ 782 if ((error = getvnode(p->p_fd, SCARG(uap, fd), &fp)) != 0) 783 return (error); 784 785 if ((fp->f_flag & FREAD) == 0) { 786 error = EBADF; 787 goto out1; 788 } 789 790 vp = (struct vnode *)fp->f_data; 791 if (vp->v_type != VDIR) { 792 error = EINVAL; 793 goto out1; 794 } 795 796 if ((error = VOP_GETATTR(vp, &va, p->p_ucred, l))) 797 goto out1; 798 799 nbytes = SCARG(uap, count); 800 if (nbytes == 1) { /* emulating old, broken behaviour */ 801 nbytes = sizeof (idb); 802 buflen = max(va.va_blocksize, nbytes); 803 oldcall = 1; 804 } else { 805 buflen = min(MAXBSIZE, nbytes); 806 if (buflen < va.va_blocksize) 807 buflen = va.va_blocksize; 808 oldcall = 0; 809 } 810 buf = malloc(buflen, M_TEMP, M_WAITOK); 811 812 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 813 off = fp->f_offset; 814 again: 815 aiov.iov_base = buf; 816 aiov.iov_len = buflen; 817 auio.uio_iov = &aiov; 818 auio.uio_iovcnt = 1; 819 auio.uio_rw = UIO_READ; 820 auio.uio_segflg = UIO_SYSSPACE; 821 auio.uio_lwp = NULL; 822 auio.uio_resid = buflen; 823 auio.uio_offset = off; 824 /* 825 * First we read into the malloc'ed buffer, then 826 * we massage it into user space, one record at a time. 827 */ 828 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, &cookiebuf, 829 &ncookies); 830 if (error) 831 goto out; 832 833 inp = buf; 834 outp = (caddr_t)SCARG(uap, dent); 835 resid = nbytes; 836 if ((len = buflen - auio.uio_resid) == 0) 837 goto eof; 838 839 for (cookie = cookiebuf; len > 0; len -= reclen) { 840 bdp = (struct dirent *)inp; 841 reclen = bdp->d_reclen; 842 if (reclen & 3) 843 panic("linux_readdir"); 844 if (bdp->d_fileno == 0) { 845 inp += reclen; /* it is a hole; squish it out */ 846 off = *cookie++; 847 continue; 848 } 849 linux_reclen = LINUX_RECLEN(&idb, bdp->d_namlen); 850 if (reclen > len || resid < linux_reclen) { 851 /* entry too big for buffer, so just stop */ 852 outp++; 853 break; 854 } 855 /* 856 * Massage in place to make a Linux-shaped dirent (otherwise 857 * we have to worry about touching user memory outside of 858 * the copyout() call). 859 */ 860 idb.d_ino = bdp->d_fileno; 861 /* 862 * The old readdir() call misuses the offset and reclen fields. 863 */ 864 if (oldcall) { 865 idb.d_off = (linux_off_t)linux_reclen; 866 idb.d_reclen = (u_short)bdp->d_namlen; 867 } else { 868 if (sizeof (idb.d_off) <= 4 && (off >> 32) != 0) { 869 compat_offseterr(vp, "linux_getdents"); 870 error = EINVAL; 871 goto out; 872 } 873 idb.d_off = (linux_off_t)off; 874 idb.d_reclen = (u_short)linux_reclen; 875 } 876 strcpy(idb.d_name, bdp->d_name); 877 if ((error = copyout((caddr_t)&idb, outp, linux_reclen))) 878 goto out; 879 /* advance past this real entry */ 880 inp += reclen; 881 off = *cookie++; /* each entry points to itself */ 882 /* advance output past Linux-shaped entry */ 883 outp += linux_reclen; 884 resid -= linux_reclen; 885 if (oldcall) 886 break; 887 } 888 889 /* if we squished out the whole block, try again */ 890 if (outp == (caddr_t)SCARG(uap, dent)) 891 goto again; 892 fp->f_offset = off; /* update the vnode offset */ 893 894 if (oldcall) 895 nbytes = resid + linux_reclen; 896 897 eof: 898 *retval = nbytes - resid; 899 out: 900 VOP_UNLOCK(vp, 0); 901 if (cookiebuf) 902 free(cookiebuf, M_TEMP); 903 free(buf, M_TEMP); 904 out1: 905 FILE_UNUSE(fp, l); 906 return error; 907 } 908 909 /* 910 * Even when just using registers to pass arguments to syscalls you can 911 * have 5 of them on the i386. So this newer version of select() does 912 * this. 913 */ 914 int 915 linux_sys_select(l, v, retval) 916 struct lwp *l; 917 void *v; 918 register_t *retval; 919 { 920 struct linux_sys_select_args /* { 921 syscallarg(int) nfds; 922 syscallarg(fd_set *) readfds; 923 syscallarg(fd_set *) writefds; 924 syscallarg(fd_set *) exceptfds; 925 syscallarg(struct timeval *) timeout; 926 } */ *uap = v; 927 928 return linux_select1(l, retval, SCARG(uap, nfds), SCARG(uap, readfds), 929 SCARG(uap, writefds), SCARG(uap, exceptfds), SCARG(uap, timeout)); 930 } 931 932 /* 933 * Common code for the old and new versions of select(). A couple of 934 * things are important: 935 * 1) return the amount of time left in the 'timeout' parameter 936 * 2) select never returns ERESTART on Linux, always return EINTR 937 */ 938 int 939 linux_select1(l, retval, nfds, readfds, writefds, exceptfds, timeout) 940 struct lwp *l; 941 register_t *retval; 942 int nfds; 943 fd_set *readfds, *writefds, *exceptfds; 944 struct timeval *timeout; 945 { 946 struct sys_select_args bsa; 947 struct proc *p = l->l_proc; 948 struct timeval tv0, tv1, utv, *tvp; 949 caddr_t sg; 950 int error; 951 952 SCARG(&bsa, nd) = nfds; 953 SCARG(&bsa, in) = readfds; 954 SCARG(&bsa, ou) = writefds; 955 SCARG(&bsa, ex) = exceptfds; 956 SCARG(&bsa, tv) = timeout; 957 958 /* 959 * Store current time for computation of the amount of 960 * time left. 961 */ 962 if (timeout) { 963 if ((error = copyin(timeout, &utv, sizeof(utv)))) 964 return error; 965 if (itimerfix(&utv)) { 966 /* 967 * The timeval was invalid. Convert it to something 968 * valid that will act as it does under Linux. 969 */ 970 sg = stackgap_init(p, 0); 971 tvp = stackgap_alloc(p, &sg, sizeof(utv)); 972 utv.tv_sec += utv.tv_usec / 1000000; 973 utv.tv_usec %= 1000000; 974 if (utv.tv_usec < 0) { 975 utv.tv_sec -= 1; 976 utv.tv_usec += 1000000; 977 } 978 if (utv.tv_sec < 0) 979 timerclear(&utv); 980 if ((error = copyout(&utv, tvp, sizeof(utv)))) 981 return error; 982 SCARG(&bsa, tv) = tvp; 983 } 984 microtime(&tv0); 985 } 986 987 error = sys_select(l, &bsa, retval); 988 if (error) { 989 /* 990 * See fs/select.c in the Linux kernel. Without this, 991 * Maelstrom doesn't work. 992 */ 993 if (error == ERESTART) 994 error = EINTR; 995 return error; 996 } 997 998 if (timeout) { 999 if (*retval) { 1000 /* 1001 * Compute how much time was left of the timeout, 1002 * by subtracting the current time and the time 1003 * before we started the call, and subtracting 1004 * that result from the user-supplied value. 1005 */ 1006 microtime(&tv1); 1007 timersub(&tv1, &tv0, &tv1); 1008 timersub(&utv, &tv1, &utv); 1009 if (utv.tv_sec < 0) 1010 timerclear(&utv); 1011 } else 1012 timerclear(&utv); 1013 if ((error = copyout(&utv, timeout, sizeof(utv)))) 1014 return error; 1015 } 1016 1017 return 0; 1018 } 1019 1020 /* 1021 * Get the process group of a certain process. Look it up 1022 * and return the value. 1023 */ 1024 int 1025 linux_sys_getpgid(l, v, retval) 1026 struct lwp *l; 1027 void *v; 1028 register_t *retval; 1029 { 1030 struct linux_sys_getpgid_args /* { 1031 syscallarg(int) pid; 1032 } */ *uap = v; 1033 struct proc *p = l->l_proc; 1034 struct proc *targp; 1035 1036 if (SCARG(uap, pid) != 0 && SCARG(uap, pid) != p->p_pid) { 1037 if ((targp = pfind(SCARG(uap, pid))) == 0) 1038 return ESRCH; 1039 } 1040 else 1041 targp = p; 1042 1043 retval[0] = targp->p_pgid; 1044 return 0; 1045 } 1046 1047 /* 1048 * Set the 'personality' (emulation mode) for the current process. Only 1049 * accept the Linux personality here (0). This call is needed because 1050 * the Linux ELF crt0 issues it in an ugly kludge to make sure that 1051 * ELF binaries run in Linux mode, not SVR4 mode. 1052 */ 1053 int 1054 linux_sys_personality(l, v, retval) 1055 struct lwp *l; 1056 void *v; 1057 register_t *retval; 1058 { 1059 struct linux_sys_personality_args /* { 1060 syscallarg(int) per; 1061 } */ *uap = v; 1062 1063 if (SCARG(uap, per) != 0) 1064 return EINVAL; 1065 retval[0] = 0; 1066 return 0; 1067 } 1068 1069 #if defined(__i386__) || defined(__m68k__) 1070 /* 1071 * The calls are here because of type conversions. 1072 */ 1073 int 1074 linux_sys_setreuid16(l, v, retval) 1075 struct lwp *l; 1076 void *v; 1077 register_t *retval; 1078 { 1079 struct linux_sys_setreuid16_args /* { 1080 syscallarg(int) ruid; 1081 syscallarg(int) euid; 1082 } */ *uap = v; 1083 struct sys_setreuid_args bsa; 1084 1085 SCARG(&bsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ? 1086 (uid_t)-1 : SCARG(uap, ruid); 1087 SCARG(&bsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ? 1088 (uid_t)-1 : SCARG(uap, euid); 1089 1090 return sys_setreuid(l, &bsa, retval); 1091 } 1092 1093 int 1094 linux_sys_setregid16(l, v, retval) 1095 struct lwp *l; 1096 void *v; 1097 register_t *retval; 1098 { 1099 struct linux_sys_setregid16_args /* { 1100 syscallarg(int) rgid; 1101 syscallarg(int) egid; 1102 } */ *uap = v; 1103 struct sys_setregid_args bsa; 1104 1105 SCARG(&bsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ? 1106 (uid_t)-1 : SCARG(uap, rgid); 1107 SCARG(&bsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ? 1108 (uid_t)-1 : SCARG(uap, egid); 1109 1110 return sys_setregid(l, &bsa, retval); 1111 } 1112 1113 int 1114 linux_sys_setresuid16(l, v, retval) 1115 struct lwp *l; 1116 void *v; 1117 register_t *retval; 1118 { 1119 struct linux_sys_setresuid16_args /* { 1120 syscallarg(uid_t) ruid; 1121 syscallarg(uid_t) euid; 1122 syscallarg(uid_t) suid; 1123 } */ *uap = v; 1124 struct linux_sys_setresuid16_args lsa; 1125 1126 SCARG(&lsa, ruid) = ((linux_uid_t)SCARG(uap, ruid) == (linux_uid_t)-1) ? 1127 (uid_t)-1 : SCARG(uap, ruid); 1128 SCARG(&lsa, euid) = ((linux_uid_t)SCARG(uap, euid) == (linux_uid_t)-1) ? 1129 (uid_t)-1 : SCARG(uap, euid); 1130 SCARG(&lsa, suid) = ((linux_uid_t)SCARG(uap, suid) == (linux_uid_t)-1) ? 1131 (uid_t)-1 : SCARG(uap, suid); 1132 1133 return linux_sys_setresuid(l, &lsa, retval); 1134 } 1135 1136 int 1137 linux_sys_setresgid16(l, v, retval) 1138 struct lwp *l; 1139 void *v; 1140 register_t *retval; 1141 { 1142 struct linux_sys_setresgid16_args /* { 1143 syscallarg(gid_t) rgid; 1144 syscallarg(gid_t) egid; 1145 syscallarg(gid_t) sgid; 1146 } */ *uap = v; 1147 struct linux_sys_setresgid16_args lsa; 1148 1149 SCARG(&lsa, rgid) = ((linux_gid_t)SCARG(uap, rgid) == (linux_gid_t)-1) ? 1150 (gid_t)-1 : SCARG(uap, rgid); 1151 SCARG(&lsa, egid) = ((linux_gid_t)SCARG(uap, egid) == (linux_gid_t)-1) ? 1152 (gid_t)-1 : SCARG(uap, egid); 1153 SCARG(&lsa, sgid) = ((linux_gid_t)SCARG(uap, sgid) == (linux_gid_t)-1) ? 1154 (gid_t)-1 : SCARG(uap, sgid); 1155 1156 return linux_sys_setresgid(l, &lsa, retval); 1157 } 1158 1159 int 1160 linux_sys_getgroups16(l, v, retval) 1161 struct lwp *l; 1162 void *v; 1163 register_t *retval; 1164 { 1165 struct linux_sys_getgroups16_args /* { 1166 syscallarg(int) gidsetsize; 1167 syscallarg(linux_gid_t *) gidset; 1168 } */ *uap = v; 1169 struct proc *p = l->l_proc; 1170 caddr_t sg; 1171 int n, error, i; 1172 struct sys_getgroups_args bsa; 1173 gid_t *bset, *kbset; 1174 linux_gid_t *lset; 1175 struct pcred *pc = p->p_cred; 1176 1177 n = SCARG(uap, gidsetsize); 1178 if (n < 0) 1179 return EINVAL; 1180 error = 0; 1181 bset = kbset = NULL; 1182 lset = NULL; 1183 if (n > 0) { 1184 n = min(pc->pc_ucred->cr_ngroups, n); 1185 sg = stackgap_init(p, 0); 1186 bset = stackgap_alloc(p, &sg, n * sizeof (gid_t)); 1187 kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK); 1188 lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK); 1189 if (bset == NULL || kbset == NULL || lset == NULL) 1190 return ENOMEM; 1191 SCARG(&bsa, gidsetsize) = n; 1192 SCARG(&bsa, gidset) = bset; 1193 error = sys_getgroups(l, &bsa, retval); 1194 if (error != 0) 1195 goto out; 1196 error = copyin(bset, kbset, n * sizeof (gid_t)); 1197 if (error != 0) 1198 goto out; 1199 for (i = 0; i < n; i++) 1200 lset[i] = (linux_gid_t)kbset[i]; 1201 error = copyout(lset, SCARG(uap, gidset), 1202 n * sizeof (linux_gid_t)); 1203 } else 1204 *retval = pc->pc_ucred->cr_ngroups; 1205 out: 1206 if (kbset != NULL) 1207 free(kbset, M_TEMP); 1208 if (lset != NULL) 1209 free(lset, M_TEMP); 1210 return error; 1211 } 1212 1213 int 1214 linux_sys_setgroups16(l, v, retval) 1215 struct lwp *l; 1216 void *v; 1217 register_t *retval; 1218 { 1219 struct linux_sys_setgroups16_args /* { 1220 syscallarg(int) gidsetsize; 1221 syscallarg(linux_gid_t *) gidset; 1222 } */ *uap = v; 1223 struct proc *p = l->l_proc; 1224 caddr_t sg; 1225 int n; 1226 int error, i; 1227 struct sys_setgroups_args bsa; 1228 gid_t *bset, *kbset; 1229 linux_gid_t *lset; 1230 1231 n = SCARG(uap, gidsetsize); 1232 if (n < 0 || n > NGROUPS) 1233 return EINVAL; 1234 sg = stackgap_init(p, 0); 1235 bset = stackgap_alloc(p, &sg, n * sizeof (gid_t)); 1236 lset = malloc(n * sizeof (linux_gid_t), M_TEMP, M_WAITOK); 1237 kbset = malloc(n * sizeof (gid_t), M_TEMP, M_WAITOK); 1238 if (lset == NULL || bset == NULL) 1239 return ENOMEM; 1240 error = copyin(SCARG(uap, gidset), lset, n * sizeof (linux_gid_t)); 1241 if (error != 0) 1242 goto out; 1243 for (i = 0; i < n; i++) 1244 kbset[i] = (gid_t)lset[i]; 1245 error = copyout(kbset, bset, n * sizeof (gid_t)); 1246 if (error != 0) 1247 goto out; 1248 SCARG(&bsa, gidsetsize) = n; 1249 SCARG(&bsa, gidset) = bset; 1250 error = sys_setgroups(l, &bsa, retval); 1251 1252 out: 1253 if (lset != NULL) 1254 free(lset, M_TEMP); 1255 if (kbset != NULL) 1256 free(kbset, M_TEMP); 1257 1258 return error; 1259 } 1260 1261 #endif /* __i386__ || __m68k__ */ 1262 1263 /* 1264 * We have nonexistent fsuid equal to uid. 1265 * If modification is requested, refuse. 1266 */ 1267 int 1268 linux_sys_setfsuid(l, v, retval) 1269 struct lwp *l; 1270 void *v; 1271 register_t *retval; 1272 { 1273 struct linux_sys_setfsuid_args /* { 1274 syscallarg(uid_t) uid; 1275 } */ *uap = v; 1276 struct proc *p = l->l_proc; 1277 uid_t uid; 1278 1279 uid = SCARG(uap, uid); 1280 if (p->p_cred->p_ruid != uid) 1281 return sys_nosys(l, v, retval); 1282 else 1283 return (0); 1284 } 1285 1286 /* XXX XXX XXX */ 1287 #ifndef alpha 1288 int 1289 linux_sys_getfsuid(l, v, retval) 1290 struct lwp *l; 1291 void *v; 1292 register_t *retval; 1293 { 1294 return sys_getuid(l, v, retval); 1295 } 1296 #endif 1297 1298 int 1299 linux_sys_setresuid(l, v, retval) 1300 struct lwp *l; 1301 void *v; 1302 register_t *retval; 1303 { 1304 struct linux_sys_setresuid_args /* { 1305 syscallarg(uid_t) ruid; 1306 syscallarg(uid_t) euid; 1307 syscallarg(uid_t) suid; 1308 } */ *uap = v; 1309 1310 /* 1311 * Note: These checks are a little different than the NetBSD 1312 * setreuid(2) call performs. This precisely follows the 1313 * behavior of the Linux kernel. 1314 */ 1315 1316 return do_setresuid(l, SCARG(uap, ruid), SCARG(uap, euid), 1317 SCARG(uap, suid), 1318 ID_R_EQ_R | ID_R_EQ_E | ID_R_EQ_S | 1319 ID_E_EQ_R | ID_E_EQ_E | ID_E_EQ_S | 1320 ID_S_EQ_R | ID_S_EQ_E | ID_S_EQ_S ); 1321 } 1322 1323 int 1324 linux_sys_getresuid(l, v, retval) 1325 struct lwp *l; 1326 void *v; 1327 register_t *retval; 1328 { 1329 struct linux_sys_getresuid_args /* { 1330 syscallarg(uid_t *) ruid; 1331 syscallarg(uid_t *) euid; 1332 syscallarg(uid_t *) suid; 1333 } */ *uap = v; 1334 struct proc *p = l->l_proc; 1335 struct pcred *pc = p->p_cred; 1336 int error; 1337 1338 /* 1339 * Linux copies these values out to userspace like so: 1340 * 1341 * 1. Copy out ruid. 1342 * 2. If that succeeds, copy out euid. 1343 * 3. If both of those succeed, copy out suid. 1344 */ 1345 if ((error = copyout(&pc->p_ruid, SCARG(uap, ruid), 1346 sizeof(uid_t))) != 0) 1347 return (error); 1348 1349 if ((error = copyout(&pc->pc_ucred->cr_uid, SCARG(uap, euid), 1350 sizeof(uid_t))) != 0) 1351 return (error); 1352 1353 return (copyout(&pc->p_svuid, SCARG(uap, suid), sizeof(uid_t))); 1354 } 1355 1356 int 1357 linux_sys_ptrace(l, v, retval) 1358 struct lwp *l; 1359 void *v; 1360 register_t *retval; 1361 { 1362 struct linux_sys_ptrace_args /* { 1363 i386, m68k, powerpc: T=int 1364 alpha: T=long 1365 syscallarg(T) request; 1366 syscallarg(T) pid; 1367 syscallarg(T) addr; 1368 syscallarg(T) data; 1369 } */ *uap = v; 1370 const int *ptr; 1371 int request; 1372 int error; 1373 1374 ptr = linux_ptrace_request_map; 1375 request = SCARG(uap, request); 1376 while (*ptr != -1) 1377 if (*ptr++ == request) { 1378 struct sys_ptrace_args pta; 1379 1380 SCARG(&pta, req) = *ptr; 1381 SCARG(&pta, pid) = SCARG(uap, pid); 1382 SCARG(&pta, addr) = (caddr_t)SCARG(uap, addr); 1383 SCARG(&pta, data) = SCARG(uap, data); 1384 1385 /* 1386 * Linux ptrace(PTRACE_CONT, pid, 0, 0) means actually 1387 * to continue where the process left off previously. 1388 * The same thing is achieved by addr == (caddr_t) 1 1389 * on NetBSD, so rewrite 'addr' appropriately. 1390 */ 1391 if (request == LINUX_PTRACE_CONT && SCARG(uap, addr)==0) 1392 SCARG(&pta, addr) = (caddr_t) 1; 1393 1394 error = sys_ptrace(l, &pta, retval); 1395 if (error) 1396 return error; 1397 switch (request) { 1398 case LINUX_PTRACE_PEEKTEXT: 1399 case LINUX_PTRACE_PEEKDATA: 1400 error = copyout (retval, 1401 (caddr_t)SCARG(uap, data), sizeof *retval); 1402 *retval = SCARG(uap, data); 1403 break; 1404 default: 1405 break; 1406 } 1407 return error; 1408 } 1409 else 1410 ptr++; 1411 1412 return LINUX_SYS_PTRACE_ARCH(l, uap, retval); 1413 } 1414 1415 int 1416 linux_sys_reboot(struct lwp *l, void *v, register_t *retval) 1417 { 1418 struct linux_sys_reboot_args /* { 1419 syscallarg(int) magic1; 1420 syscallarg(int) magic2; 1421 syscallarg(int) cmd; 1422 syscallarg(void *) arg; 1423 } */ *uap = v; 1424 struct sys_reboot_args /* { 1425 syscallarg(int) opt; 1426 syscallarg(char *) bootstr; 1427 } */ sra; 1428 struct proc *p = l->l_proc; 1429 int error; 1430 1431 if ((error = suser(p->p_ucred, &p->p_acflag)) != 0) 1432 return(error); 1433 1434 if (SCARG(uap, magic1) != LINUX_REBOOT_MAGIC1) 1435 return(EINVAL); 1436 if (SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2 && 1437 SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2A && 1438 SCARG(uap, magic2) != LINUX_REBOOT_MAGIC2B) 1439 return(EINVAL); 1440 1441 switch (SCARG(uap, cmd)) { 1442 case LINUX_REBOOT_CMD_RESTART: 1443 SCARG(&sra, opt) = RB_AUTOBOOT; 1444 break; 1445 case LINUX_REBOOT_CMD_HALT: 1446 SCARG(&sra, opt) = RB_HALT; 1447 break; 1448 case LINUX_REBOOT_CMD_POWER_OFF: 1449 SCARG(&sra, opt) = RB_HALT|RB_POWERDOWN; 1450 break; 1451 case LINUX_REBOOT_CMD_RESTART2: 1452 /* Reboot with an argument. */ 1453 SCARG(&sra, opt) = RB_AUTOBOOT|RB_STRING; 1454 SCARG(&sra, bootstr) = SCARG(uap, arg); 1455 break; 1456 case LINUX_REBOOT_CMD_CAD_ON: 1457 return(EINVAL); /* We don't implement ctrl-alt-delete */ 1458 case LINUX_REBOOT_CMD_CAD_OFF: 1459 return(0); 1460 default: 1461 return(EINVAL); 1462 } 1463 1464 return(sys_reboot(l, &sra, retval)); 1465 } 1466 1467 /* 1468 * Copy of compat_12_sys_swapon(). 1469 */ 1470 int 1471 linux_sys_swapon(l, v, retval) 1472 struct lwp *l; 1473 void *v; 1474 register_t *retval; 1475 { 1476 struct sys_swapctl_args ua; 1477 struct linux_sys_swapon_args /* { 1478 syscallarg(const char *) name; 1479 } */ *uap = v; 1480 1481 SCARG(&ua, cmd) = SWAP_ON; 1482 SCARG(&ua, arg) = (void *)SCARG(uap, name); 1483 SCARG(&ua, misc) = 0; /* priority */ 1484 return (sys_swapctl(l, &ua, retval)); 1485 } 1486 1487 /* 1488 * Stop swapping to the file or block device specified by path. 1489 */ 1490 int 1491 linux_sys_swapoff(l, v, retval) 1492 struct lwp *l; 1493 void *v; 1494 register_t *retval; 1495 { 1496 struct sys_swapctl_args ua; 1497 struct linux_sys_swapoff_args /* { 1498 syscallarg(const char *) path; 1499 } */ *uap = v; 1500 1501 SCARG(&ua, cmd) = SWAP_OFF; 1502 SCARG(&ua, arg) = (void *)SCARG(uap, path); 1503 return (sys_swapctl(l, &ua, retval)); 1504 } 1505 1506 /* 1507 * Copy of compat_09_sys_setdomainname() 1508 */ 1509 /* ARGSUSED */ 1510 int 1511 linux_sys_setdomainname(l, v, retval) 1512 struct lwp *l; 1513 void *v; 1514 register_t *retval; 1515 { 1516 struct linux_sys_setdomainname_args /* { 1517 syscallarg(char *) domainname; 1518 syscallarg(int) len; 1519 } */ *uap = v; 1520 int name[2]; 1521 1522 name[0] = CTL_KERN; 1523 name[1] = KERN_DOMAINNAME; 1524 return (old_sysctl(&name[0], 2, 0, 0, SCARG(uap, domainname), 1525 SCARG(uap, len), l)); 1526 } 1527 1528 /* 1529 * sysinfo() 1530 */ 1531 /* ARGSUSED */ 1532 int 1533 linux_sys_sysinfo(l, v, retval) 1534 struct lwp *l; 1535 void *v; 1536 register_t *retval; 1537 { 1538 struct linux_sys_sysinfo_args /* { 1539 syscallarg(struct linux_sysinfo *) arg; 1540 } */ *uap = v; 1541 struct linux_sysinfo si; 1542 struct loadavg *la; 1543 1544 si.uptime = time.tv_sec - boottime.tv_sec; 1545 la = &averunnable; 1546 si.loads[0] = la->ldavg[0] * LINUX_SYSINFO_LOADS_SCALE / la->fscale; 1547 si.loads[1] = la->ldavg[1] * LINUX_SYSINFO_LOADS_SCALE / la->fscale; 1548 si.loads[2] = la->ldavg[2] * LINUX_SYSINFO_LOADS_SCALE / la->fscale; 1549 si.totalram = ctob(physmem); 1550 si.freeram = uvmexp.free * uvmexp.pagesize; 1551 si.sharedram = 0; /* XXX */ 1552 si.bufferram = uvmexp.filepages * uvmexp.pagesize; 1553 si.totalswap = uvmexp.swpages * uvmexp.pagesize; 1554 si.freeswap = (uvmexp.swpages - uvmexp.swpginuse) * uvmexp.pagesize; 1555 si.procs = nprocs; 1556 1557 /* The following are only present in newer Linux kernels. */ 1558 si.totalbig = 0; 1559 si.freebig = 0; 1560 si.mem_unit = 1; 1561 1562 return (copyout(&si, SCARG(uap, arg), sizeof si)); 1563 } 1564 1565 #define bsd_to_linux_rlimit1(l, b, f) \ 1566 (l)->f = ((b)->f == RLIM_INFINITY || \ 1567 ((b)->f & 0xffffffff00000000ULL) != 0) ? \ 1568 LINUX_RLIM_INFINITY : (int32_t)(b)->f 1569 #define bsd_to_linux_rlimit(l, b) \ 1570 bsd_to_linux_rlimit1(l, b, rlim_cur); \ 1571 bsd_to_linux_rlimit1(l, b, rlim_max) 1572 1573 #define linux_to_bsd_rlimit1(b, l, f) \ 1574 (b)->f = (l)->f == LINUX_RLIM_INFINITY ? RLIM_INFINITY : (l)->f 1575 #define linux_to_bsd_rlimit(b, l) \ 1576 linux_to_bsd_rlimit1(b, l, rlim_cur); \ 1577 linux_to_bsd_rlimit1(b, l, rlim_max) 1578 1579 static int 1580 linux_to_bsd_limit(lim) 1581 int lim; 1582 { 1583 switch (lim) { 1584 case LINUX_RLIMIT_CPU: 1585 return RLIMIT_CPU; 1586 case LINUX_RLIMIT_FSIZE: 1587 return RLIMIT_FSIZE; 1588 case LINUX_RLIMIT_DATA: 1589 return RLIMIT_DATA; 1590 case LINUX_RLIMIT_STACK: 1591 return RLIMIT_STACK; 1592 case LINUX_RLIMIT_CORE: 1593 return RLIMIT_CORE; 1594 case LINUX_RLIMIT_RSS: 1595 return RLIMIT_RSS; 1596 case LINUX_RLIMIT_NPROC: 1597 return RLIMIT_NPROC; 1598 case LINUX_RLIMIT_NOFILE: 1599 return RLIMIT_NOFILE; 1600 case LINUX_RLIMIT_MEMLOCK: 1601 return RLIMIT_MEMLOCK; 1602 case LINUX_RLIMIT_AS: 1603 case LINUX_RLIMIT_LOCKS: 1604 return -EOPNOTSUPP; 1605 default: 1606 return -EINVAL; 1607 } 1608 } 1609 1610 1611 int 1612 linux_sys_getrlimit(l, v, retval) 1613 struct lwp *l; 1614 void *v; 1615 register_t *retval; 1616 { 1617 struct linux_sys_getrlimit_args /* { 1618 syscallarg(int) which; 1619 syscallarg(struct orlimit *) rlp; 1620 } */ *uap = v; 1621 struct proc *p = l->l_proc; 1622 caddr_t sg = stackgap_init(p, 0); 1623 struct sys_getrlimit_args ap; 1624 struct rlimit rl; 1625 struct orlimit orl; 1626 int error; 1627 1628 SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which)); 1629 if ((error = SCARG(&ap, which)) < 0) 1630 return -error; 1631 SCARG(&ap, rlp) = stackgap_alloc(p, &sg, sizeof rl); 1632 if ((error = sys_getrlimit(l, &ap, retval)) != 0) 1633 return error; 1634 if ((error = copyin(SCARG(&ap, rlp), &rl, sizeof(rl))) != 0) 1635 return error; 1636 bsd_to_linux_rlimit(&orl, &rl); 1637 return copyout(&orl, SCARG(uap, rlp), sizeof(orl)); 1638 } 1639 1640 int 1641 linux_sys_setrlimit(l, v, retval) 1642 struct lwp *l; 1643 void *v; 1644 register_t *retval; 1645 { 1646 struct linux_sys_setrlimit_args /* { 1647 syscallarg(int) which; 1648 syscallarg(struct orlimit *) rlp; 1649 } */ *uap = v; 1650 struct proc *p = l->l_proc; 1651 caddr_t sg = stackgap_init(p, 0); 1652 struct sys_setrlimit_args ap; 1653 struct rlimit rl; 1654 struct orlimit orl; 1655 int error; 1656 1657 SCARG(&ap, which) = linux_to_bsd_limit(SCARG(uap, which)); 1658 SCARG(&ap, rlp) = stackgap_alloc(p, &sg, sizeof rl); 1659 if ((error = SCARG(&ap, which)) < 0) 1660 return -error; 1661 if ((error = copyin(SCARG(uap, rlp), &orl, sizeof(orl))) != 0) 1662 return error; 1663 linux_to_bsd_rlimit(&rl, &orl); 1664 /* XXX: alpha complains about this */ 1665 if ((error = copyout(&rl, (void *)SCARG(&ap, rlp), sizeof(rl))) != 0) 1666 return error; 1667 return sys_setrlimit(l, &ap, retval); 1668 } 1669 1670 #ifndef __mips__ 1671 /* XXX: this doesn't look 100% common, at least mips doesn't have it */ 1672 int 1673 linux_sys_ugetrlimit(l, v, retval) 1674 struct lwp *l; 1675 void *v; 1676 register_t *retval; 1677 { 1678 return linux_sys_getrlimit(l, v, retval); 1679 } 1680 #endif 1681 1682 /* 1683 * This gets called for unsupported syscalls. The difference to sys_nosys() 1684 * is that process does not get SIGSYS, the call just returns with ENOSYS. 1685 * This is the way Linux does it and glibc depends on this behaviour. 1686 */ 1687 int 1688 linux_sys_nosys(l, v, retval) 1689 struct lwp *l; 1690 void *v; 1691 register_t *retval; 1692 { 1693 return (ENOSYS); 1694 } 1695
Indexes created Mon Sep 22 21:09:42 GMT 2025